Campylobacter Transmission in a Peruvian Shantytown: A Longitudinal Study Using Strain Typing of Campylobacter Isolates from Chickens and Humans in Ho
1Department of Tropical Medicine, Tulane School of Public Health, New Orleans, Louisiana, 2Department of International Health, Johns Hopkins School of Hygiene and Public Health, Baltimore, Maryland; 3Asociación Benéfica PRISMA, 4Universidad Peruana Cayetano Heredia, and 5Naval Medical Research Center, Detachment, Lima, Peru; 6National Laboratory for Enteric Pathogens, National Microbiology Laboratory, Winnipeg, Canada
Received 11 July 2002; revised 1 October 2002; electronically published 6 January 2003.
Campylobacter jejuni is a major cause of pediatric diarrhea in developing countries—
free-ranging chickens are presumed to be a common source. Campylobacter strains from monthly surveillance and diarrhea cases were compared by means of restriction-fragment length polymorphism (RFLP), rapid amplified polymorphic DNA, and Lior serotyping. RFLP analysis of 156 human and 682 avian strains demonstrated identical strains in chickens and humans in 29 (70.7%) of 41 families, and 35%–
39% of human isolates from diarrhea and nondiarrhea cases were identical to a household chicken isolate. Isolation of the same RFLP type from a household chicken and a human within 1 month was highly protective against diarrhea (odds ratio, 0.07; P < .005). Campylobacter strains from symptomatic humans were unlikely to be identical to strains recently carried by household chickens, limiting the potential benefits from household-based control measures.
Presented in part: 48th annual meeting of the American Society of Tropical Medicine and Hygiene, Washington, DC, November 1999 (abstract 174).
Informed consent was obtained from subjects or their parents or guardians, and the human experimentation guidelines of the US Department of Health and Human Services, Tulane Medical Center, and Asociacion Benefica PRISMA were followed in conducting this research. The opinions and assertions contained herein are those of the writers and are not to be construed as official or as reflecting the views of the US Navy.
Financial support: Thrasher Research Fund (award 02813-1); ITREID grant from the National Institutes of Health; Navy Medical Research Center (work unit 63002A.810.O.B0020).
Present affiliations: Division of Communicable Diseases and Immunology, Walter Reed Army Institute of Research, Washington, DC (D.N.T.); Department of Microbiology, University of New Hampshire, Durham (F.G.R.); Navy Environmental and Preventive Medicine Unit 6, Pearl Harbor, Hawaii (C.L.B.).
Reprints or correspondence: Dr. Richard Oberhelman, Dept. of Tropical Medicine, Tulane School of Public Health, 1440 Canal St., SL 29, New Orleans, Louisana 70112 ().
Campylobacter jejuni is a major bacterial cause of diarrhea among children aged <5 years in many developing countries [1–
4]. Campylobacter species cause both symptomatic and asymptomatic infection in chickens, cats, and dogs, and these animal species are thought to be important sources of human infection [1, 5]. In addition, consumption of undercooked meat or contact with infected animal carcasses are risk factors for campylobacteriosis [3, 6]. Seventeen surveys of retailed raw meats in the United States have indicated that broiler chickens pose the greatest risk of any meat product, with an average contamination rate of 60% [5]. Although we know much about the epidemiology of this disease from outbreak investigations, direct evidence for household transmission of Campylobacter species from animals to humans in endemic populations is lacking. This information is potentially very valuable to design effective public health strategies to reduce Campylobacter transmission and Campylobacter-related diarrhea.
The large population of children living in periurban shanty-towns or "pueblo jóvenes" surrounding greater Lima, Peru, are at high risk for C. jejuni infection. Most of these communities lack basic sanitary facilities, and fecal contamination from both human and animal sources is high. Families in pueblo jovenes often have chickens in the household, and C. jejuni has been isolated from 50% of free-ranging domestic chickens and in 88% of commercially sold chickens in studies done in pueblos jóvenes [6]. The prevalence of C. jejuni–
related diarrhea is as high as 10%–
12% in children aged <24 months, as determined by standard bacteriologic methods [3]. Previous studies in this population [7] have demonstrated that toddlers had hand contact with poultry feces an average of 2.9 times per 12-h period, and feces-to-mouth contamination was positively correlated with the amount of chicken feces in the home (r = 0.66). Although viable Campylobacter species were cultured from these stool specimens up to 48 h after deposition, free-roaming poultry often were not perceived as a health risk.
Control measures for Campylobacter species are largely nonspecific. Specific control measures such as vaccination are frequently thought to be impractical, because the causes of pediatric diarrhea are varied and any one pathogen contributes relatively little to the disease process. However, the spread of pathogens transmitted by the fecal-oral route could logically be interrupted by simple measures that reduce exposure to infectious fecal material. Since chickens are one of the major sources of C. jejuni, a frequent cause of diarrhea in children from Peruvian shantytowns, where chickens usually roam freely, a simple intervention such as corralling chickens into coops could have a major impact if household chickens, in fact, transmit their organisms to humans. To evaluate the potential efficacy of such an intervention, the goals of the present project were (1) to describe the epidemiology of endemic Campylobacter infections in a Peruvian periurban shantytown with large numbers of free-ranging chickens and (2) to analyze chicken-to-child transmission of C. jejuni strains in relation to the presence of diarrhea, using molecular strain markers and Lior serotyping.
SUBJECTS, MATERIALS, AND METHODS
Study cohort: enrollment and surveillance. Fifty families living in the Pampas de San Juan, a low-income periurban community in Lima, Peru, were identified from community surveillance data collected by the Asociacion Benefica PRISMA and were enrolled by informed consent. Each family had 2 free-roaming chickens and 2 children aged <5 years, with 1 child aged <24 months. Families were maintained in the longitudinal study as long as they had 1 free-roaming chicken in the house, and families that lost or slaughtered all chickens without replacing them within a week were excluded and replaced with another family.
On enrollment, demographic information was collected for each family. Chickens were tagged and monitored, and all household participants and chickens were cultured for C. jejuni at the start of surveillance to establish the initial (baseline) prevalence of infection. Continuous surveillance consisted of a brief study questionnaire to document recent diarrhea in any household participant by recording daily frequency of stools, stool consistency, and the presence of blood or mucus in the stool. Household members aged <5 years and 2 sentinel chickens/family were cultured for Campylobacter species each month to determine asymptomatic carriage rates (defined as rates of infection in healthy control subjects in the analysis), and additional cultures were obtained from all humans with diarrhea, according to the criteria of Black et al. [3]. In the case of children aged <24 months, an episode of diarrhea was defined as 4 liquid stools per day for 1 day. The duration of diarrhea was defined as the number of days both preceding and following the day that met the main criteria, in which 3 liquid stools were passed. By this definition, 2 days without diarrhea had to occur for each episode of diarrhea to be classified as a new episode. If a diarrheal stool specimen was culture positive for C. jejuni, stool cultures of family contacts and chickens in the household were obtained during the following week.
Recovery of C. jejuni from environmental surfaces was compared in 9 families that met the original inclusion criteria who were not part of the main surveillance cohort. We purposely selected a group of families not included in the main study, because aggressive culturing of multiple materials in the home on a repeated basis could change behavior and potentially alter the incidence of Campylobacter infections detected. Samples were collected on a quarterly basis, including culture of hand-washings, floor dirt, water, milk, stored meat, and fresh stool samples from cats and dogs living in the household. Field workers recorded the presence of chicken feces in each room of the home during each sampling period, and a sample of these specimens was cultured for C. jejuni.
Microbiological techniques. Bacteriologic assays were performed at the Naval Medical Research Center, Detachment (NMRCD; Lima, Peru). Fresh stool specimens inoculated in Cary Blair transport medium were cultured on selective agar plates (Brucella agar plus 5% defibrinated sheep blood) containing Butzler antibiotic supplement for Campylobacter (Oxoid) and Campylobacter growth supplement (Oxoid). Plates were incubated at 42°C for 48 h. Campylobacter isolates were identified by way of conventional techniques, including Gram's stain, oxidase testing, and speciation by use of hippurate hydrolysis, nitrate reduction, the rapid hydrogen sulfate test, and susceptibility to cephalothin and nalidixic acid. Environmental samples of foods or water from hand washings were placed in sterile tubes of Preston Campylobacter selective enrichment broth and were incubated for 48 h at 42°C under microaerophilic conditions [8]. After this broth incubation, broth was subcultured to selective agar plates, and confirmed strains of Campylobacter species were saved for serotyping and molecular strain typing.
Campylobacter species strain typing. To determine the prevalence of similar C. jejuni strains in participants and chickens within the same households, strains were typed in the pathology laboratory at the Universidad Peruana Cayetano Heredia (Lima, Peru) by use of 2 methods: rapid amplified polymorphic DNA (RAPD) [9–
12] and restriction–
fragment length polymorphism (RFLP) [13, 14]. For both techniques, DNA was extracted from bacteria grown on Brucella agar supplemented with 5% sheep's blood under microaerophilic conditions, using the QIAamp Tissue Kit (Qiagen). Both assays were performed by means of polymerase chain reaction (PCR), using 1× PCR buffer, MgCl2, dNTPs, primers, Taq polymerase, and sample DNA at standard concentrations. The RAPD technique uses random amplified polymorphic DNA generated by arbitrary primers to produce a unique pattern for each organism by agarose gel electrophoresis. For RAPD, samples were amplified by use of the following primers: 1281 (5′
-AACGCGCAAC), 1283 (5′
-GCGATCCCCA-3′), and 1290 (5′
-GTGGATGCGA-3′). PCR assays were performed with 45 cycles at 94°C (1 min), 36°C (1 min), and 72°C (2 min, with 10 min for final cycle). PCR products were separated electrophoretically in 2% agarose with ethidium bromide for 3 h at 60–
70 V at room temperature using 1× Tris-acetate (0.04 MTris-acetate, 0.001 M EDTA) buffer. RFLP assays were done in Peru using the flagellin (flaA) gene-based typing schema provided by Dr. I. Nachamkin from the University of Pennsylvania (Philadelphia) [13, 14]. PCR assays were performed using 4 L of template with flaA1 primer 5′
-GGATTTCGTATTAACACAAATGGTGC-3′
and flaA2 primer sequence 5′
-CTGTAGTAATCTTAAAACATTTTG-3′
in a thermal cycler under the following conditions: 94°C for 1 min and then 94°C for 15 s, 55°C for 45 s, and 72°C for 1 min and 45 s for 35 cycles, followed by 72°C for 5 min. After confirming the presence of the 1.7-kb amplicon representing the Campylobacter flagellin gene by electrophoresis, 8
L of PCR product was digested with restriction enzyme DdeI (175S; New England Biolabs), according to the manufacturer's instructions, and DdeI-digested product was analyzed by agarose gel electrophoresis and ethidium-bromide staining. For both RAPD and RFLP, electrophoretic patterns were observed under ultraviolet transillumination and photographed with a Kodak digital camera.
DNA ProSCORE (DNA ProSCAN) [14], a computerized program that analyzes restriction-fragment patterns generated by RAPD or RFLP, was used to identify similar bacterial strains. Strain profiles for each RAPD primer and for RFLP were compared with molecular-weight markers and were corrected for minor differences in banding pattern when profiles were compared. Photographs of RAPD and RFLP assays were scanned on a flatbed scanner and were imported into Adobe Photoshop (LE version 2.5; Adobe Systems), in which images were adjusted for contrast and labeled. Images were analyzed by DNA ProSCORE according to the manufacturer's instructions. Unique patterns were stored in the database so that subsequent new images could be compared with existing patterns at the 5% stringency level with 1 : 1 matching. If a match occurred, the image in the database was recalled to visually compare the 2 patterns, to ensure matching. If a complete match did not occur, the new patterns were given a unique type designation and were stored for future comparisons.
In the case of RAPD, 3 primers were used to analyze 3 clones and a pooled specimen from each Campylobacter isolate, and a number was assigned to patterns produced by each primer. The results of the 3 primers then were compared to generate a composite RAPD profile number, which was used to compare strains. RFLP studies generated a single numerical strain identifier, and results of Campylobacter isolate strain typing were compared by both methods.
Lior serotyping [15] also was performed as the third method, to compare a subset of strains previously analyzed by RAPD and RFLP. This procedure was performed in Canada under a collaborative agreement with Health Canada (Winnepeg, Manitoba) and at NMRCD. This well-recognized typing schema is based on heat-labile antigens and includes >100 serotypes.
Data management and analysis. Field workers checked the data obtained daily, after which data were entered into Access 97 databases (Microsoft). All persons and chickens were given a code for identification, which served as the link for all databases. Validation of 5% of the data was performed monthly, and data sets with coding errors >1% were reentered.
All episodes of diarrhea in humans detected in the study population were cultured, so the incidence and prevalence of Campylobacter-related diarrhea are shown for all age groups. In addition, 2 children aged <6 years in each study household were examined by monthly surveillance stool cultures to determine incidence and prevalence of infection in healthy control specimens. Because of high rates of diarrhea, some of these "control" specimens (because of the schedule of collection and by chance) also were associated with a diarrhea episode. Therefore, prevalence and incidence of infection in asymptomatic control specimens were analyzed on the basis of stool samples collected in the absence of diarrhea only. Strain typing data were analyzed by family cluster, and within families we identified "groups" of strains, defined as groups of 2 isolates with similar RFLP patterns.
Data analysis was performed at Tulane, Johns Hopkins University (Baltimore, MD), and NMRCD using SPSS version 7.5 (SPSS), Stata 7.0 (Stata), and Epi Info 6 software programs. The 2 test and, where necessary, Fisher's exact tests were used to measure strengths of association between categorical variables. A 2-tailed Student's t test was used to compare continuous variables, and logistic regression was used to assess association between multiple variables. Using a logistic regression for longitudinal data based on generalized estimating equations (GEEs), we analyzed the effect of isolating C. jejuni from a chicken within 1 month of isolating the same RFLP type from a human in the same household on the probability that the human infection was associated with diarrhea. All regressions were adjusted for age and seasonality and were controlled for longitudinal effect.
RESULTS
Demographic features and rates of Campylobacter isolation. Seventy-two families were enrolled for surveillance, of which 63 families were surveyed for 1 month beyond initial enrollment (cases that could be analyzed). The mean time of surveillance for these 63 families was 36 weeks. The age distribution of the 423 subjects surveyed was as follows: 47 (11.1%), 0–
12 months; 101 (23.9%), 1–
5 years; 48 (11.3%), 6–
10 years; 69 (16.3%), 11–
20 years; 133 (31.4%), 21–
40 years; and 25 (5.9%), >40 years. Twenty percent of families earned <1 Peruvian standard minimum wage (345 nuevos soles or $115 US monthly), 36% earned between 345 nuevos soles monthly and the community average (728 nuevos soles or $242 US monthly), and 44% earned more than the community average. Most families used water from communal indoor faucets and stored water in tanks, drums, or buckets. Indoor plumbing was present in 66% of homes, although it was frequently not functional, whereas 34% of families disposed of human waste in makeshift latrines, holes in the ground, or open fields.
A total of 3574 specimens from chickens, humans, and environmental samples were cultured for Campylobacter species. Overall, Campylobacter species was recovered from 156 (9.0%) of 1729 human specimens, 682 (39.9%) of 1711 avian specimens, and 9 (6.7%) of 134 environmental specimens. Campylobacter species was the most common bacterial pathogen isolated from 347 diarrhea specimens collected from cohort family members. Pathogens isolated from diarrhea specimens included Campylobacter species (35 cases, 10%), Shigella species (5 cases, 1.4%), Vibrio species (3 cases, 1%), Aeromonas species (2 cases, >1%), and Salmonella species (1 case, >1%).
Campylobacter infection, both with and without associated diarrhea, is very common in this population, especially in the youngest children . Both the prevalence and incidence of infection were inversely related to age, with highest rates in the 0–
12-month age group. All case subjects of Campylobacter-associated diarrhea were aged 5 years. No differences in the incidence or prevalence of Campylobacter infections were seen among children with diarrhea versus healthy control subjects, and the incidence of positive Campylobacter stool cultures in control subjects exceeded the incidence observed in children with diarrhea. Both diarrhea and control specimens positive for Campylobacter species were less common in the middle of winter (June and July in the southern hemisphere). Other than these 2 months, no seasonal variation was seen, and there was no obvious seasonal peak for either diarrhea or routine isolates.
fig.ommitted
Table 1. Prevalence and incidence of Campylobacter jejuni among children and adults living in the Pampas de San Juan, Lima, Peru.
Campylobacter species was recovered from 19 (6.1%) of 309 baseline cultures collected from well participants on enrollment. Isolation rates of Campylobacter species were highest in children aged <1 year and declined with age (0–
12 months, 6/39 culture-positive [15.4%]; 1–
2 years, 4/35 [11.3%]; 3–
5 years, 3/51 [5.9%]; 6–
10 years, 4/44 [9.1%]; 11–
20 years, 2/42 [4.8%]; and 21–
40 years, 0/87). Baseline enrollment cultures also were performed on 710 chickens from cohort households, of which 360 (50.7%) were culture-positive for Campylobacter species.
In follow-up specimens collected at 2-week intervals after the onset of Campylobacter-associated diarrhea, Campylobacter species was isolated again in about one-third of specimens tested. Although the highest rate of Campylobacter isolation after Campylobacter-associated diarrhea was again seen in the youngest children, rates were similar for older children as well (0–
12 months, 10/26 [38.5%] culture-positive specimens; 1–
2 years, 5/15 [33.3%]; and 3–
5 years, 1/4 [25%]).
Family contacts of children with Campylobacter species in stool (both with and without diarrhea) were cultured within 2 days of recovering the organism from the index case. Campylobacter species was recovered from 7.2% of these specimens. Campylobacter species was isolated more frequently from family contact specimens than from baseline specimens in both adolescents and teenagers aged 11–
20 years (3/21 [14.3%] family contact vs. 2/42 [4.8%] baseline, P = .32) and in adults aged 21–
40 years (4/60 [6.7%] vs. 0/87, P = .03).
Chicken feces found in the household were the primary environmental source of Campylobacter species in the 9 families studied. One hundred thirty-four specimens were collected and cultured for Campylobacter species, including 48 water samples from individual hand washings, 7 stored milk samples, 20 stored water samples, 20 soil samples from the interior of the house, 25 food samples, 4 samples of dog feces, and 10 samples of chicken feces collected from the floor of the house. Two of 4 samples of dog feces and 7 of 10 samples of chicken feces were culture positive, but Campylobacter species was not recovered from any of the other samples. Because chicken feces were much more abundant than dog feces in homes surveyed and because 70% of these chicken feces were culture positive for Campylobacter species, chicken feces appear to be a major environmental source of contamination.
Epidemiological correlates of Campylobacter infections. To identify risk factors for human infection with Campylobacter species, families with Campylobacter species isolated from both human and avian sources (i.e., human infection detected) were compared with families with Campylobacter species isolated from chickens only (i.e., no human infections detected). Campylobacter species was isolated from both humans and chickens in 44 (72%) of 61 families with complete demographic information and adequate follow-up, from chickens only in 16 (26%) 61 families, and from humans only in 1 (1.6%) of 61 families. Recovery of Campylobacter species from humans was more frequent in the lower socioeconomic group (SES; i.e., family income less than or at the community average). Twenty-eight (65%) of 43 families with human Campylobacter infection were in the lower SES, whereas only 4 (26.7%) of 15 families with Campylobacter species isolated from chickens only were in the lower SES (P = .009; odds ratio [OR], 5.13). Other differences that did not reach statistical significance when compared with our sample size were presence of children aged <12 months (33/283 [11.7%] for human infection group and 8/109 [7.3%] for avian infection only group; P = .21) and the proportion of families with women who had advanced post–
high school education (2/24 [8.3%] for human infection group and 3/10 [30%] for avian infection only group; P = .10). No differences were found on the basis of population density of the household, education of the head of household, type of home construction, water source and storage practices, and method of human waste disposal.
Comparison of Campylobacter isolates from family clusters by RFLP. A total of 838 Campylobacter isolates were collected from cohort surveillance, including both chicken and human isolates. RFLP analysis was conducted on all 838 strains. In addition, we studied Campylobacter isolates from 16 families with the longest period of surveillance and the largest number of isolates by RAPD and Lior serotyping. Two hundred ninety-four strains from these 16 family clusters were analyzed by RAPD using 3 primers, and 159 of these also were serotyped by the Lior method. One hundred twenty-seven strains from 16 family clusters were typed by all 3 methods.
We identified an abundance of different Campylobacter strains, including 61 distinct RFLP patterns, 162 RAPD patterns, and 34 Lior serotypes. Among the 16 family groups studied by all 3 methods, we identified an average of 5.0 RFLP patterns per 10 strains studied, 7.0 RAPD patterns per 10 strains studied, and 5.0 Lior serotypes per 10 strains studied.
Similarities among Campylobacter strains by RFLP were determined for each of the families on the basis of 4 criteria (). Denominators vary because of variations in the number of human and chicken isolates per family and in the number of persons or chickens with >1 isolate. Similarities between strains were tabulated over the entire study period, and "groups" of strains were defined as groups of 2 isolates with similar RFLP patterns.
fig.ommitted
Table 2. Strain similarities among 838 Campylobacter isolates from birds and humans in 67 households in a Peruvian periurban shantytown, as determined by restriction-fragment length polymorphism (RFLP) analysis.
Similar strains in a chicken and a human in the same household were present in 29 (70.7%) of 41 families. However, when restricted to families with the largest number of Campylobacter isolates (20 isolates/family), similar strains in a chicken and a human were present in 9 of 9 families. An example of a family group with similar RFLP patterns in chickens and humans is shown in . Similar strains in 2 different chickens in the same household were present in 51 (87.9%) of 58 families. When analysis was limited to families with 20 Campylobacter isolates, similar strains in 2 different chickens were present in 9 of 9 families. The number of RFLP-based strain groups isolated from birds in the same household was directly related to the total number of Campylobacter isolates recovered from the household (0–
10 isolates, mean 1.0 groups; 11–
20 isolates, 2.4 groups; 21–
30 isolates, 3.2 groups; and >30 isolates, 6.8 groups). Similar strains in 2 different humans in the same household (not applicable when Campylobacter species was only recovered from 1 human in the family) were present in 10 (40%) of 25 families.
fig.ommitted
Figure 1. Restriction–
fragment length polymorphism patterns for 3 human strains and 6 avian strains of Campylobacter jejuni collected from a single household (family 12). All strains were tested in duplicate. Similar patterns for human and avian strains collected from the same household are seen in lanes 4–
9 and 14–
19 (2 avian and 2 human isolate with the same pattern in each group). Lanes 1 and 20, 100-bp DNA ladder; lanes 2 and 3, CMP 956 strain from chicken; lanes 4 and 5, CMP 1263 strain from chicken; lanes 6 and 7, CMP 1265 strain from human; lanes 8 and 9, CMP strain from chicken; lanes 10 and 11, CMP 1756 from chicken; lanes 12 and 13, CMP 1264 strain from chicken; lanes 14 and 15, CMP 1320 strain from chicken; lanes 16 and 17, CMP 2235 strain from chicken; lanes 18 and 19, CMP 2421 strain from human.
Association between presence of similar strains of Campylobacter in humans and birds and risk of diarrhea. Overall, 59 (38%) of 155 Campylobacter isolates from humans had an identical RFLP profile to a chicken isolate from the same household. There were no differences between the frequency of identical chicken and human strains within households when human isolates from diarrhea cases (12/34 [>35.2%]) and nondiarrhea cases (47/121 [38.8%]) were compared. We did not find a correlation between any specific RFLP pattern and diarrhea (vs. asymptomatic infection) in the entire study group, nor was there an association between any RAPD pattern or Lior serotype and diarrhea in the subgroups with RAPD and serotyping data.
A more sophisticated analysis was performed to analyze similarities between Campylobacter strains from humans and chickens as a function of time and presence of diarrhea. A total of 840 household-months of observation were recorded in the study, and, in 100 of these, 1 human was infected with Campylobacter species. In 20 of these 100 household-months, 1 household bird shared the same RFLP strain type as the human strain within 1 month of (i.e., between 15 days before and 15 days after) the human infection. When analyzed by person-months instead of household-months, we recorded a total of 147 person-months in which a human was infected with Campylobacter species, and in 29 (19.7%) of these, 1 household bird shared the same strain as the human within a 1-month period.
We identified 17 case subjects with persistent infection which was defined as isolation of Campylobacter strains with the same RFLP type from the same person on 2 occasions within a 60-day period. Eleven of the 17 case subjects were identified by reculturing children with Campylobacter diarrhea at 2-week follow-up intervals. We also identified 18 case subjects of reinfection with a new RFLP type within a 60-day period. In 7 (39%) of these 18 case subjects, the second Campylobacter isolate was associated with diarrhea, whereas only 3 (17.6%) of 17 second isolates from persistent infections were related to diarrhea (P = .25). No differences in the frequency of similar strains in humans and chickens within a 1-month period were found when isolates from persistent infections and reinfections were compared.
Using a logistic regression for longitudinal data based on GEEs, we analyzed the effect of isolating C. jejuni from a chicken within 1 month of isolating the same RFLP type from a human in the same household on the probability that the human infection was associated with diarrhea. In this analysis, the presence of the same Campylobacter strain type in stool samples from a chicken and a human in the same household within a 1-month period was highly associated with the absence of diarrhea. Alternatively, one could say that the presence of shared Campylobacter strains between chickens and humans in a household was statistically protective against Campylobacter gastroenteritis. Of 34 Campylobacter strains from humans associated with diarrhea, 1 (3%) coincided with a strain isolated from a chicken in the same household within a 1-month period. In contrast, of 124 Campylobacter strains from humans which were not associated with diarrhea, 28 (22.6%) coincided with a strain isolated from a chicken in the same household within 1 month (P < .005; OR, 0.07). This protective effect of sharing a strain on the presence of diarrhea was more pronounced in children aged >340 days versus those aged 340 days (break point based on distribution of sample), although sample size limitations did not allow us to report significant ORs for each age-based stratification group. Neither seasonality nor age at the time of infection was significantly associated with the probability of diarrhea.
Comparison of RFLP and RAPD for the detection of Campylobacter strain variation. The correlation between RFLP and RAPD patterns were compared within individual households. Using RFLP as the reference standard, we found that, of 164 strain groups with the same RFLP pattern, 97 (59.1%) also shared the same RAPD pattern. The degree of similarity, as defined by RAPD, increased as the number of strains in the RFLP "group" increased (e.g., among "groups" of 2 strains with same RFLP: 14/42 strains [33.3%] had similar RAPD; "groups" of 3 by RFLP: 26/42 strains [61.9%] had similar RAPD; and "groups" of 4 by RFLP: 57/80 [71.2%] had similar RAPD).
Using RAPD as the reference standard, the percentage of strains that also had corresponding RFLP patterns was higher than when RFLP was used as the reference (i.e., 100 [78.1%] of 128 strains in groups with similar RAPD also had similar RFLP pattern), indicating that variability by RAPD was greater than that by RFLP. As above, similarities by RFLP increased as the number of strains in the RAPD "group" increased (e.g., among "groups" of 2 strains with same RAPD: 42/58 strains [72.4%] had similar RFLP; "groups" of 3 by RAPD: 58/70 strains [82.8%] with similar RAPD).
In most case subjects, RAPD techniques analyzed by ProSCORE showed more strain differences than by RFLP. In 3 case subjects, 6 RAPDS types corresponded to a single RFLP type among strains of Campylobacter species from a single family cluster. However, 1 RAPD type that was very common in both chickens and children in the study (strain 10) corresponded to 7 RFLP types; thus, subtle differences are not always detected more commonly with RAPD. In general, RAPD appeared to detect more strain differences than RFLP, although many of these strain variations may have been due to subtle changes in electrophoresis pattern that may be misinterpreted as a strain difference using RAPD.
We also compared results of RAPD tests done on a pooled Campylobacter isolate with results performed using 3 separate clones. Of 105 isolates tested, only 7 (6.7%) had 2 different RAPD profiles, when results from individual clones and pooled isolates were compared. Because these results show that differences between clones and pooled specimens were very uncommon, we conducted RAPD and RFLP analyses, as reported here by use of pooled isolates and 1 clone only. In >98% of cases, the pattern demonstrated by the clone was identical to the pattern demonstrated by the pooled isolate.
Comparison of Campylobacter strain variation by RFLP, RAPD, and Lior serotyping. Lior serotypes identified from the group of 159 strains tested are shown in . Strain-typing data are available by all 3 methods (Lior serotype, RFLP, and RAPD) for 127 strains from 16 households. Eight of the 159 serotyped strains agglutinated with >1 serotype, and, in 1 household, we identified 2 chickens carrying strains that agglutinated with the same 2 serotypes (C. jejuni 1 and 2).
fig.ommitted
Table 3. Frequency of Lior serotypes among Campylobacter jejuni and Campylobacter coli strains isolated from birds and humans from the study population in Peru.
The common Lior serotypes of Campylobacter strains tested in our study group were similar to those commonly reported in the United States, including outbreak-associated strains and strains from sporadic disease [16, 17]. Four of the 5 most common serotypes in our study also are commonly reported from the United States (serotypes 1, 2, 4, and 36). One exception was Lior serotype 32, which was the fourth most common serotype in our study but is uncommonly reported in data from the United States.
One hundred twenty-five strains in groups of 2 with the same RFLP type and from the same household were serotyped by the Lior method. In 59 (47%) of 125 strains, groups with the same RFLP pattern also had the same Lior serotype, whereas in 66 (53%) case subjects, groups with the same RFLP had >1 Lior serotype represented. However, many of the strains in the latter category were in groups of 3 with the same RFLP, and, in several cases, there were partial matches by Lior serotype as well—
that is, 2 of the group of 3 strains with the same RFLP also had the same serotype. If these partial matches are also included, the proportion of strains with the same RFLP and the same Lior serotype becomes 81 (65%) of 125.
RAPD data were available for 56 of 59 strains in which identical RFLP patterns corresponded exactly to identical Lior serotypes. Of these 56 strains, 31 (55%) had RAPD that also corresponded exactly to the other 2 methods. Therefore, only 31 (25%) of 125 stains, in groups of 2 with the same RFLP, also demonstrated identical matches by both Lior serotype and RAPD.
DISCUSSION
Our study population demonstrates epidemiological features typical of endemic campylobacteriosis in other parts of the developing world, including a risk for disease that is inversely proportional to age and high rates of asymptomatic infection. Our incidence rates of Campylobacter-related diarrhea in children aged 5 years (0.37 episodes/child/year) were similar to rates reported in Mexico and Thailand and similar to the rate reported elsewhere in this Peruvian community [2, 18, 19]. Campylobacter species was not associated with any of the 41 diarrhea episodes detected in persons aged 6 years, whereas it was associated among 22 (19%) of 112 diarrhea episodes detected in children aged <12 months. This very high prevalence of Campylobacter diarrhea in very young children (0.89 episodes/child/year) exceeds rates reported in previous studies and may be due to the combination of poverty and selection of households with free-ranging chickens. Although the entire study population were in the lower socioeconomic group (mean family income, $242.00 US/month) and all families had free-ranging chickens, human infection with C. jejuni was statistically associated with extreme poverty—
two-thirds of the families with human infections had a family income below the community average. In our study group, chicken feces in the home was the major environmental source of Campylobacter species, which was not detected in hand washings, stored foods, stored water, or soil samples. The high rates of asymptomatic infection are consistent with previous epidemiological data from Peru and other developing countries, where many studies have not found the striking differences in prevalence of enteropathogens between healthy children and children with diarrhea, as seen in the United States [3].
Our molecular strain–
typing tools demonstrated an extremely large number of Campylobacter strains circulating in this community, with an average of half as many distinct strain types as Campylobacter isolates in households studied. Despite the multiplicity of circulating strains, we were able to detect similar strains in chickens and humans in the same household in 70% of all families studied, and similar stains in chickens and humans were detected in all 9 of the households with >20 Campylobacter isolates studied. However, strains shared by humans and chickens in the same household within a 1-month period were unlikely to be associated with diarrhea in humans. In fact, the presence of the same Campylobacter strain type in stool samples from a chicken and a human in the same household within a 1-month period was statistically protective against Campylobacter-associated diarrhea. Even when we compared strains in chickens and humans within households over the entire study period, human strains from case subjects with diarrhea were no more likely to match a chicken isolate from the same household than nondiarrhea strains. This would suggest that most Campylobacter-associated diarrhea is due to strains acquired outside the home and that previous exposure to strains carried by chickens in the home confers a high degree of strain-specific immunity to disease but not immunity to repeated infection. Whether this hypothetical previous exposure occurred with disease before the start of surveillance or by asymptomatic infection during surveillance remain to be clarified. It is important to emphasize, however, that the RFLP pattern is merely one method to distinguish between similar and different strains of Campylobacter, so one cannot assume that the antigens responsible for the RFLP type are the same as those responsible for the observed protective immunity. The pathogenesis of Campylobacter gastroenteritis is complex and involves adhesins, invasion factors, antigens that induce apoptosis, and enterotoxins, and the relative importance of these virulence factors in induction of protective immunity is not known.
Because Campylobacter transmission between humans and chickens is a fluid process and may occur in both directions (chicken-to-human and human-to-chicken), our study was not able to detect the direction or temporal flow of organisms within family clusters. However, our data provide strong indirect evidence of high rates of chicken-to-human transmission in this high-risk population, although this flow of organisms did not appear to contribute much to diarrheal disease. Our data suggest that Campylobacter isolates from household chickens are responsible for only a minority of Campylobacter diarrhea cases in humans living in the same household, reminiscent of recent data from the Centers for Disease Control and Prevention, which have shown decreasing rates of Campylobacter diarrhea in the United States, despite increasing rates of chicken consumption [20]. To our knowledge, this is the first time that the epidemiological link between chickens and children with Campylobacter infection has been demonstrated in an endemic disease setting by use of molecular strain markers.
Our data also suggest that molecular strain typing and serotyping are probably of little value for evaluating the effectiveness of Campylobacter control strategies in similar communities with endemic disease and high levels of transmission. Although some RFLP patterns and Lior serotypes were found more frequently than others in the study population, there were no patterns or serotypes that were statistically associated with diarrheal disease. Because of the lack of predominant strains or serotypes, intervention strategies for Campylobacter species in endemic populations are best evaluated by surveillance of all isolates and not specific strains.
Our results illustrate some of the inherent difficulties in using molecular tools to compare bacterial strains in an endemic disease setting. Although both RFLP and Lior serotyping demonstrated less variability of strain types than RAPD, about one-third of isolates that demonstrated similarity by either RFLP or serotype were found to be different by the other assay. Although some variation of results is expected because the 2 systems classify strains on the basis of different microbiological features (flagellin genotype for RFLP and heat-labile surface antigens for Lior typing), this degree of discrepancy between 2 strain-typing systems suggests that it is difficult to definitively prove that 2 Campylobacter isolates from persons or birds in the same family cluster are identical, either because of differences that are not detected or because of constant genetic mutations that may affect one classification system but not the other. Indeed, recent reports confirm that genetic exchange between both homologous and heterologous Campylobacter strains occurs during experimental infections in chickens by RFLP, amplified fragment length polymorphism, and pulsed field gel electrophoresis, resulting in the generation of new genetic rearrangements even in the absence of immunologic pressure [21]. Although consistent identical types or patterns by 2 systems that examine different antigens would provide the best proof of animal-to-human transmission, in reality this ideal may be elusive because of rapid recombinations occurring in vivo. This hypothesis is supported by the relative paucity of strains with similar patterns by 1 of the 3 systems that were identical by all 3 methods (i.e., 25% of those tested).
Because of the genomic plasticity of Campylobacter species demonstrated by de Boer et al. [21] and also suggested by our findings, are strain typing results reliable in an endemic disease situation? It is probably reasonable to assume that background spontaneous strain variation occurs at a fairly constant rate that should affect all study populations in a similar way. Although recognizing that the tools at our disposal have identifiable limitations, our data demonstrate statistically significant differences that can be supported by biologically plausible explanations, supporting the use of molecular typing in certain situations. Our findings suggest that RAPD detects more strain differences than does RFLP, but minor differences in the electrophoresis pattern (perhaps due to minor genetic recombinations) may be misinterpreted as strain differences when RAPD is used. Because the results of strain typing by RFLP were less variable and more consistent within family groups than were RAPD results and because RFLP was technically easier than serotyping for large numbers of strains, we used RFLP as the typing method of choice for our study. Further studies will be needed to determine whether RFLP or Lior serotype is less susceptible to spontaneous variability.
In summary, humans and chickens in this periurban shantytown are exposed to a veritable sea of Campylobacter strains of many different serotypes and genetic patterns. Campylobacter infections were most prevalent in the youngest children, and human infection was associated with extreme poverty, although the incidence and prevalence of symptomatic infection versus asymptomatic infection were not different. Although it is impossible to identify predominant strains associated either with diarrhea or with human isolates versus isolates from chickens, similar strains of Campylobacter in chickens and humans were found in many family clusters, especially those with the largest number of bacterial isolates. However, most of these shared strains contributed little to the burden of disease. Bacterial strain typing by RAPD, RFLP, or Lior serotyping is unlikely to be useful in assessing Campylobacter control measures in this type of community because of the high multiplicity of strains circulating in the community. Variability in results arising from the use of different methods suggests that frequent ongoing mutations are occurring in bacterial populations, making it difficult to definitively identify identical strains. New control strategies for campylobacteriosis in similar communities should target sources of infection coming from outside the home, which may be more effective in reducing rates of disease.
Acknowledgments
We thank Stephen Walz for his valuable review and suggestions; Naval Medical Research Center, Detachment (Lima, Peru) for contributing support for diagnostic microbiology; Irving Nachamkin for providing training in restriction-fragment length polymorphism typing and use of software programs for strain identification; DNA ProSCAN for contributing software for analyzing the strain-typing data; and the Health Promoters Association and the people of the Pampas de San Juan for their participation.
References
1.
Oberhelman RA, Taylor D. Campylobacter infections in developing countries. In: Blaser MJ, Nachamkin I, eds. Campylobacter. Washington, DC: American Society for Microbiology, 1999; 139–
54.
2.
Oberhelman RA, Gilman RH, Taylor DN, et al. A placebo-controlled trial of Lactobacillus GG to prevent diarrhea in Peruvian children. J Pediatr 1999; 134:15–
20.
3.
Black RE, Lopez de Romana G, Brown KH, et al. Incidence and etiology of infantile diarrhea and major routes of transmission in Huascar, Peru. Am J Epidemiol 1989; 129:785–
99.
4.
Guerrant RL, Hughes J, Lima N, Crane J. Diarrhea in developed and developing countries: magnitude, special settings, and etiologies. Rev Infect Dis 1990; 12:S41–
50.
5.
Skirrow MB, Blaser MJ. Campylobacter jejuni. In: Blaser MJ, Smith PD, Ravdin JI, Greenberg HB, Guerrant RL, eds. Infections of the gastrointestinal tract, 1st ed. New York: Raven Press, 1995; 825–
48.
6.
Grados O, Bravo N, Black RE, Butzler JP. Paediatric campylobacter diarrhea from household exposure to live chickens in Lima, Peru. Bull World Health Organ 1988; 66:369–
74.
7.
Marquis GS, Ventura G, Gilman RH, et al. Fecal contamination of shanty town toddlers in households with non-corralled poultry, Lima, Peru. Am J Public Health 1990; 80:146–
9.
8.
Hernandez J, Alonso JL, Fayos A, Amoros I, Owen RJ. Development of a PCR assay combined with a short enrichment culture for detection of Campylobacter jejuni in estaturine surface waters. FEMS Microbiol Lett 1995; 127:201–
6.
9.
Akopyans N, Bukanov N, Westblom T, Kresovich S, Berg D. DNA diversity among clinical isolates of Helicobacter pylori detected by PCR-based RAPD fingerprinting. Nucleic Acids Res 1992; 20:5137–
42.
10.
Hernandez J, Fayos A, Ferrus MA, Owen RJ. Random amplified polymorphic DNA fingerprinting of Campylobacter jejuni and C. coli isolated from human faeces, seawater, and poultry products. Res Microbiol 1995; 146:685–
96.
11.
Welsh J, McClelland M. Fingerprinting genomes using PCR with arbitrary primers. Nucleic Acids Res 1990; 18:7213–
8.
12.
Winters DK, Slavik MF. Evaluation of a PCR based assay for specific detection of Campylobacter jejuni in chicken washes. Mol Cell Probes 1995; 9:307–
10.
13.
Nachamkin I, Ung H, Patton C. Analysis of HL and O serotypes of Campylobacter strains by the flagellin gene typing system. J Clin Microbiol 1996; 34:277–
81.
14.
Nachamkin I, Bohachick K, Patton C. Flagellin typing of Campylobacter jejuni by restriction fragment length polymorphism analysis. J Clin Microbiol 1993; 31:1531–
6.
15.
Lior H, Woodward DL, Edgar JA, Laroche LJ, Gill P. Serotyping of Campylobacter jejuni by slide agglutination based on heat-labile antigenic factors. J Clin Microbiol 1982; 15:761–
8.
16.
Bopp CA, Birkness KA, Wachsmuth IK, Barrett TJ. In vitro antimicrobial susceptibility, plasmid analysis, and serotyping of epidemic-associated Campylobacter jejuni. J Clin Microbiol 1985; 21:4–
7.
17.
Patton CM, Barrett TJ, Morris GK. Comparison of the Penner and Lior methods for serotyping Campylobacter spp. J Clin Microbiol 1985; 22:558–
65.
18.
Calva JJ, Ruiz-Palacios GM, Lopez-Vidal AB, Ramos A, Bojalil R. Cohort study of intestinal infection with Campylobacter in Mexican children. Lancet 1988; 1:503–
6.
19.
Taylor DN, Echeverria P, Pitarangsi C, Seriwatana J, Bodhidatta L, Blaser M. Influence of strain characteristics and immunity on the epidemiology of Campylobacter infections in Thailand. J Clin Microbiol 1988; 26:863–
8.
20.
Preliminary FoodNet data on the incidence of foodborne illnesses: selected sites, United States 1999. MMWR Morb Mortal Wkly Rep 2000; 49:201–
5.
21.
De Boer P, Wagenaar JA, Achterberg RP, van Putten JPM, Schouls LM, Duim B. Generation of Campylobacter jejuni genetic diversity in vivo. Mol Microbiol 2002; 44:351–
9., http://www.100md.com(Richard A. Oberhelman Robert H. Gilman Patricia Sheen Julianna Cordova David N. Taylor Mirko Zimic R)
Received 11 July 2002; revised 1 October 2002; electronically published 6 January 2003.
Campylobacter jejuni is a major cause of pediatric diarrhea in developing countries—
free-ranging chickens are presumed to be a common source. Campylobacter strains from monthly surveillance and diarrhea cases were compared by means of restriction-fragment length polymorphism (RFLP), rapid amplified polymorphic DNA, and Lior serotyping. RFLP analysis of 156 human and 682 avian strains demonstrated identical strains in chickens and humans in 29 (70.7%) of 41 families, and 35%–
39% of human isolates from diarrhea and nondiarrhea cases were identical to a household chicken isolate. Isolation of the same RFLP type from a household chicken and a human within 1 month was highly protective against diarrhea (odds ratio, 0.07; P < .005). Campylobacter strains from symptomatic humans were unlikely to be identical to strains recently carried by household chickens, limiting the potential benefits from household-based control measures.
Presented in part: 48th annual meeting of the American Society of Tropical Medicine and Hygiene, Washington, DC, November 1999 (abstract 174).
Informed consent was obtained from subjects or their parents or guardians, and the human experimentation guidelines of the US Department of Health and Human Services, Tulane Medical Center, and Asociacion Benefica PRISMA were followed in conducting this research. The opinions and assertions contained herein are those of the writers and are not to be construed as official or as reflecting the views of the US Navy.
Financial support: Thrasher Research Fund (award 02813-1); ITREID grant from the National Institutes of Health; Navy Medical Research Center (work unit 63002A.810.O.B0020).
Present affiliations: Division of Communicable Diseases and Immunology, Walter Reed Army Institute of Research, Washington, DC (D.N.T.); Department of Microbiology, University of New Hampshire, Durham (F.G.R.); Navy Environmental and Preventive Medicine Unit 6, Pearl Harbor, Hawaii (C.L.B.).
Reprints or correspondence: Dr. Richard Oberhelman, Dept. of Tropical Medicine, Tulane School of Public Health, 1440 Canal St., SL 29, New Orleans, Louisana 70112 ().
Campylobacter jejuni is a major bacterial cause of diarrhea among children aged <5 years in many developing countries [1–
4]. Campylobacter species cause both symptomatic and asymptomatic infection in chickens, cats, and dogs, and these animal species are thought to be important sources of human infection [1, 5]. In addition, consumption of undercooked meat or contact with infected animal carcasses are risk factors for campylobacteriosis [3, 6]. Seventeen surveys of retailed raw meats in the United States have indicated that broiler chickens pose the greatest risk of any meat product, with an average contamination rate of 60% [5]. Although we know much about the epidemiology of this disease from outbreak investigations, direct evidence for household transmission of Campylobacter species from animals to humans in endemic populations is lacking. This information is potentially very valuable to design effective public health strategies to reduce Campylobacter transmission and Campylobacter-related diarrhea.
The large population of children living in periurban shanty-towns or "pueblo jóvenes" surrounding greater Lima, Peru, are at high risk for C. jejuni infection. Most of these communities lack basic sanitary facilities, and fecal contamination from both human and animal sources is high. Families in pueblo jovenes often have chickens in the household, and C. jejuni has been isolated from 50% of free-ranging domestic chickens and in 88% of commercially sold chickens in studies done in pueblos jóvenes [6]. The prevalence of C. jejuni–
related diarrhea is as high as 10%–
12% in children aged <24 months, as determined by standard bacteriologic methods [3]. Previous studies in this population [7] have demonstrated that toddlers had hand contact with poultry feces an average of 2.9 times per 12-h period, and feces-to-mouth contamination was positively correlated with the amount of chicken feces in the home (r = 0.66). Although viable Campylobacter species were cultured from these stool specimens up to 48 h after deposition, free-roaming poultry often were not perceived as a health risk.
Control measures for Campylobacter species are largely nonspecific. Specific control measures such as vaccination are frequently thought to be impractical, because the causes of pediatric diarrhea are varied and any one pathogen contributes relatively little to the disease process. However, the spread of pathogens transmitted by the fecal-oral route could logically be interrupted by simple measures that reduce exposure to infectious fecal material. Since chickens are one of the major sources of C. jejuni, a frequent cause of diarrhea in children from Peruvian shantytowns, where chickens usually roam freely, a simple intervention such as corralling chickens into coops could have a major impact if household chickens, in fact, transmit their organisms to humans. To evaluate the potential efficacy of such an intervention, the goals of the present project were (1) to describe the epidemiology of endemic Campylobacter infections in a Peruvian periurban shantytown with large numbers of free-ranging chickens and (2) to analyze chicken-to-child transmission of C. jejuni strains in relation to the presence of diarrhea, using molecular strain markers and Lior serotyping.
SUBJECTS, MATERIALS, AND METHODS
Study cohort: enrollment and surveillance. Fifty families living in the Pampas de San Juan, a low-income periurban community in Lima, Peru, were identified from community surveillance data collected by the Asociacion Benefica PRISMA and were enrolled by informed consent. Each family had 2 free-roaming chickens and 2 children aged <5 years, with 1 child aged <24 months. Families were maintained in the longitudinal study as long as they had 1 free-roaming chicken in the house, and families that lost or slaughtered all chickens without replacing them within a week were excluded and replaced with another family.
On enrollment, demographic information was collected for each family. Chickens were tagged and monitored, and all household participants and chickens were cultured for C. jejuni at the start of surveillance to establish the initial (baseline) prevalence of infection. Continuous surveillance consisted of a brief study questionnaire to document recent diarrhea in any household participant by recording daily frequency of stools, stool consistency, and the presence of blood or mucus in the stool. Household members aged <5 years and 2 sentinel chickens/family were cultured for Campylobacter species each month to determine asymptomatic carriage rates (defined as rates of infection in healthy control subjects in the analysis), and additional cultures were obtained from all humans with diarrhea, according to the criteria of Black et al. [3]. In the case of children aged <24 months, an episode of diarrhea was defined as 4 liquid stools per day for 1 day. The duration of diarrhea was defined as the number of days both preceding and following the day that met the main criteria, in which 3 liquid stools were passed. By this definition, 2 days without diarrhea had to occur for each episode of diarrhea to be classified as a new episode. If a diarrheal stool specimen was culture positive for C. jejuni, stool cultures of family contacts and chickens in the household were obtained during the following week.
Recovery of C. jejuni from environmental surfaces was compared in 9 families that met the original inclusion criteria who were not part of the main surveillance cohort. We purposely selected a group of families not included in the main study, because aggressive culturing of multiple materials in the home on a repeated basis could change behavior and potentially alter the incidence of Campylobacter infections detected. Samples were collected on a quarterly basis, including culture of hand-washings, floor dirt, water, milk, stored meat, and fresh stool samples from cats and dogs living in the household. Field workers recorded the presence of chicken feces in each room of the home during each sampling period, and a sample of these specimens was cultured for C. jejuni.
Microbiological techniques. Bacteriologic assays were performed at the Naval Medical Research Center, Detachment (NMRCD; Lima, Peru). Fresh stool specimens inoculated in Cary Blair transport medium were cultured on selective agar plates (Brucella agar plus 5% defibrinated sheep blood) containing Butzler antibiotic supplement for Campylobacter (Oxoid) and Campylobacter growth supplement (Oxoid). Plates were incubated at 42°C for 48 h. Campylobacter isolates were identified by way of conventional techniques, including Gram's stain, oxidase testing, and speciation by use of hippurate hydrolysis, nitrate reduction, the rapid hydrogen sulfate test, and susceptibility to cephalothin and nalidixic acid. Environmental samples of foods or water from hand washings were placed in sterile tubes of Preston Campylobacter selective enrichment broth and were incubated for 48 h at 42°C under microaerophilic conditions [8]. After this broth incubation, broth was subcultured to selective agar plates, and confirmed strains of Campylobacter species were saved for serotyping and molecular strain typing.
Campylobacter species strain typing. To determine the prevalence of similar C. jejuni strains in participants and chickens within the same households, strains were typed in the pathology laboratory at the Universidad Peruana Cayetano Heredia (Lima, Peru) by use of 2 methods: rapid amplified polymorphic DNA (RAPD) [9–
12] and restriction–
fragment length polymorphism (RFLP) [13, 14]. For both techniques, DNA was extracted from bacteria grown on Brucella agar supplemented with 5% sheep's blood under microaerophilic conditions, using the QIAamp Tissue Kit (Qiagen). Both assays were performed by means of polymerase chain reaction (PCR), using 1× PCR buffer, MgCl2, dNTPs, primers, Taq polymerase, and sample DNA at standard concentrations. The RAPD technique uses random amplified polymorphic DNA generated by arbitrary primers to produce a unique pattern for each organism by agarose gel electrophoresis. For RAPD, samples were amplified by use of the following primers: 1281 (5′
-AACGCGCAAC), 1283 (5′
-GCGATCCCCA-3′), and 1290 (5′
-GTGGATGCGA-3′). PCR assays were performed with 45 cycles at 94°C (1 min), 36°C (1 min), and 72°C (2 min, with 10 min for final cycle). PCR products were separated electrophoretically in 2% agarose with ethidium bromide for 3 h at 60–
70 V at room temperature using 1× Tris-acetate (0.04 MTris-acetate, 0.001 M EDTA) buffer. RFLP assays were done in Peru using the flagellin (flaA) gene-based typing schema provided by Dr. I. Nachamkin from the University of Pennsylvania (Philadelphia) [13, 14]. PCR assays were performed using 4 L of template with flaA1 primer 5′
-GGATTTCGTATTAACACAAATGGTGC-3′
and flaA2 primer sequence 5′
-CTGTAGTAATCTTAAAACATTTTG-3′
in a thermal cycler under the following conditions: 94°C for 1 min and then 94°C for 15 s, 55°C for 45 s, and 72°C for 1 min and 45 s for 35 cycles, followed by 72°C for 5 min. After confirming the presence of the 1.7-kb amplicon representing the Campylobacter flagellin gene by electrophoresis, 8
L of PCR product was digested with restriction enzyme DdeI (175S; New England Biolabs), according to the manufacturer's instructions, and DdeI-digested product was analyzed by agarose gel electrophoresis and ethidium-bromide staining. For both RAPD and RFLP, electrophoretic patterns were observed under ultraviolet transillumination and photographed with a Kodak digital camera.
DNA ProSCORE (DNA ProSCAN) [14], a computerized program that analyzes restriction-fragment patterns generated by RAPD or RFLP, was used to identify similar bacterial strains. Strain profiles for each RAPD primer and for RFLP were compared with molecular-weight markers and were corrected for minor differences in banding pattern when profiles were compared. Photographs of RAPD and RFLP assays were scanned on a flatbed scanner and were imported into Adobe Photoshop (LE version 2.5; Adobe Systems), in which images were adjusted for contrast and labeled. Images were analyzed by DNA ProSCORE according to the manufacturer's instructions. Unique patterns were stored in the database so that subsequent new images could be compared with existing patterns at the 5% stringency level with 1 : 1 matching. If a match occurred, the image in the database was recalled to visually compare the 2 patterns, to ensure matching. If a complete match did not occur, the new patterns were given a unique type designation and were stored for future comparisons.
In the case of RAPD, 3 primers were used to analyze 3 clones and a pooled specimen from each Campylobacter isolate, and a number was assigned to patterns produced by each primer. The results of the 3 primers then were compared to generate a composite RAPD profile number, which was used to compare strains. RFLP studies generated a single numerical strain identifier, and results of Campylobacter isolate strain typing were compared by both methods.
Lior serotyping [15] also was performed as the third method, to compare a subset of strains previously analyzed by RAPD and RFLP. This procedure was performed in Canada under a collaborative agreement with Health Canada (Winnepeg, Manitoba) and at NMRCD. This well-recognized typing schema is based on heat-labile antigens and includes >100 serotypes.
Data management and analysis. Field workers checked the data obtained daily, after which data were entered into Access 97 databases (Microsoft). All persons and chickens were given a code for identification, which served as the link for all databases. Validation of 5% of the data was performed monthly, and data sets with coding errors >1% were reentered.
All episodes of diarrhea in humans detected in the study population were cultured, so the incidence and prevalence of Campylobacter-related diarrhea are shown for all age groups. In addition, 2 children aged <6 years in each study household were examined by monthly surveillance stool cultures to determine incidence and prevalence of infection in healthy control specimens. Because of high rates of diarrhea, some of these "control" specimens (because of the schedule of collection and by chance) also were associated with a diarrhea episode. Therefore, prevalence and incidence of infection in asymptomatic control specimens were analyzed on the basis of stool samples collected in the absence of diarrhea only. Strain typing data were analyzed by family cluster, and within families we identified "groups" of strains, defined as groups of 2 isolates with similar RFLP patterns.
Data analysis was performed at Tulane, Johns Hopkins University (Baltimore, MD), and NMRCD using SPSS version 7.5 (SPSS), Stata 7.0 (Stata), and Epi Info 6 software programs. The 2 test and, where necessary, Fisher's exact tests were used to measure strengths of association between categorical variables. A 2-tailed Student's t test was used to compare continuous variables, and logistic regression was used to assess association between multiple variables. Using a logistic regression for longitudinal data based on generalized estimating equations (GEEs), we analyzed the effect of isolating C. jejuni from a chicken within 1 month of isolating the same RFLP type from a human in the same household on the probability that the human infection was associated with diarrhea. All regressions were adjusted for age and seasonality and were controlled for longitudinal effect.
RESULTS
Demographic features and rates of Campylobacter isolation. Seventy-two families were enrolled for surveillance, of which 63 families were surveyed for 1 month beyond initial enrollment (cases that could be analyzed). The mean time of surveillance for these 63 families was 36 weeks. The age distribution of the 423 subjects surveyed was as follows: 47 (11.1%), 0–
12 months; 101 (23.9%), 1–
5 years; 48 (11.3%), 6–
10 years; 69 (16.3%), 11–
20 years; 133 (31.4%), 21–
40 years; and 25 (5.9%), >40 years. Twenty percent of families earned <1 Peruvian standard minimum wage (345 nuevos soles or $115 US monthly), 36% earned between 345 nuevos soles monthly and the community average (728 nuevos soles or $242 US monthly), and 44% earned more than the community average. Most families used water from communal indoor faucets and stored water in tanks, drums, or buckets. Indoor plumbing was present in 66% of homes, although it was frequently not functional, whereas 34% of families disposed of human waste in makeshift latrines, holes in the ground, or open fields.
A total of 3574 specimens from chickens, humans, and environmental samples were cultured for Campylobacter species. Overall, Campylobacter species was recovered from 156 (9.0%) of 1729 human specimens, 682 (39.9%) of 1711 avian specimens, and 9 (6.7%) of 134 environmental specimens. Campylobacter species was the most common bacterial pathogen isolated from 347 diarrhea specimens collected from cohort family members. Pathogens isolated from diarrhea specimens included Campylobacter species (35 cases, 10%), Shigella species (5 cases, 1.4%), Vibrio species (3 cases, 1%), Aeromonas species (2 cases, >1%), and Salmonella species (1 case, >1%).
Campylobacter infection, both with and without associated diarrhea, is very common in this population, especially in the youngest children . Both the prevalence and incidence of infection were inversely related to age, with highest rates in the 0–
12-month age group. All case subjects of Campylobacter-associated diarrhea were aged 5 years. No differences in the incidence or prevalence of Campylobacter infections were seen among children with diarrhea versus healthy control subjects, and the incidence of positive Campylobacter stool cultures in control subjects exceeded the incidence observed in children with diarrhea. Both diarrhea and control specimens positive for Campylobacter species were less common in the middle of winter (June and July in the southern hemisphere). Other than these 2 months, no seasonal variation was seen, and there was no obvious seasonal peak for either diarrhea or routine isolates.
fig.ommitted
Table 1. Prevalence and incidence of Campylobacter jejuni among children and adults living in the Pampas de San Juan, Lima, Peru.
Campylobacter species was recovered from 19 (6.1%) of 309 baseline cultures collected from well participants on enrollment. Isolation rates of Campylobacter species were highest in children aged <1 year and declined with age (0–
12 months, 6/39 culture-positive [15.4%]; 1–
2 years, 4/35 [11.3%]; 3–
5 years, 3/51 [5.9%]; 6–
10 years, 4/44 [9.1%]; 11–
20 years, 2/42 [4.8%]; and 21–
40 years, 0/87). Baseline enrollment cultures also were performed on 710 chickens from cohort households, of which 360 (50.7%) were culture-positive for Campylobacter species.
In follow-up specimens collected at 2-week intervals after the onset of Campylobacter-associated diarrhea, Campylobacter species was isolated again in about one-third of specimens tested. Although the highest rate of Campylobacter isolation after Campylobacter-associated diarrhea was again seen in the youngest children, rates were similar for older children as well (0–
12 months, 10/26 [38.5%] culture-positive specimens; 1–
2 years, 5/15 [33.3%]; and 3–
5 years, 1/4 [25%]).
Family contacts of children with Campylobacter species in stool (both with and without diarrhea) were cultured within 2 days of recovering the organism from the index case. Campylobacter species was recovered from 7.2% of these specimens. Campylobacter species was isolated more frequently from family contact specimens than from baseline specimens in both adolescents and teenagers aged 11–
20 years (3/21 [14.3%] family contact vs. 2/42 [4.8%] baseline, P = .32) and in adults aged 21–
40 years (4/60 [6.7%] vs. 0/87, P = .03).
Chicken feces found in the household were the primary environmental source of Campylobacter species in the 9 families studied. One hundred thirty-four specimens were collected and cultured for Campylobacter species, including 48 water samples from individual hand washings, 7 stored milk samples, 20 stored water samples, 20 soil samples from the interior of the house, 25 food samples, 4 samples of dog feces, and 10 samples of chicken feces collected from the floor of the house. Two of 4 samples of dog feces and 7 of 10 samples of chicken feces were culture positive, but Campylobacter species was not recovered from any of the other samples. Because chicken feces were much more abundant than dog feces in homes surveyed and because 70% of these chicken feces were culture positive for Campylobacter species, chicken feces appear to be a major environmental source of contamination.
Epidemiological correlates of Campylobacter infections. To identify risk factors for human infection with Campylobacter species, families with Campylobacter species isolated from both human and avian sources (i.e., human infection detected) were compared with families with Campylobacter species isolated from chickens only (i.e., no human infections detected). Campylobacter species was isolated from both humans and chickens in 44 (72%) of 61 families with complete demographic information and adequate follow-up, from chickens only in 16 (26%) 61 families, and from humans only in 1 (1.6%) of 61 families. Recovery of Campylobacter species from humans was more frequent in the lower socioeconomic group (SES; i.e., family income less than or at the community average). Twenty-eight (65%) of 43 families with human Campylobacter infection were in the lower SES, whereas only 4 (26.7%) of 15 families with Campylobacter species isolated from chickens only were in the lower SES (P = .009; odds ratio [OR], 5.13). Other differences that did not reach statistical significance when compared with our sample size were presence of children aged <12 months (33/283 [11.7%] for human infection group and 8/109 [7.3%] for avian infection only group; P = .21) and the proportion of families with women who had advanced post–
high school education (2/24 [8.3%] for human infection group and 3/10 [30%] for avian infection only group; P = .10). No differences were found on the basis of population density of the household, education of the head of household, type of home construction, water source and storage practices, and method of human waste disposal.
Comparison of Campylobacter isolates from family clusters by RFLP. A total of 838 Campylobacter isolates were collected from cohort surveillance, including both chicken and human isolates. RFLP analysis was conducted on all 838 strains. In addition, we studied Campylobacter isolates from 16 families with the longest period of surveillance and the largest number of isolates by RAPD and Lior serotyping. Two hundred ninety-four strains from these 16 family clusters were analyzed by RAPD using 3 primers, and 159 of these also were serotyped by the Lior method. One hundred twenty-seven strains from 16 family clusters were typed by all 3 methods.
We identified an abundance of different Campylobacter strains, including 61 distinct RFLP patterns, 162 RAPD patterns, and 34 Lior serotypes. Among the 16 family groups studied by all 3 methods, we identified an average of 5.0 RFLP patterns per 10 strains studied, 7.0 RAPD patterns per 10 strains studied, and 5.0 Lior serotypes per 10 strains studied.
Similarities among Campylobacter strains by RFLP were determined for each of the families on the basis of 4 criteria (). Denominators vary because of variations in the number of human and chicken isolates per family and in the number of persons or chickens with >1 isolate. Similarities between strains were tabulated over the entire study period, and "groups" of strains were defined as groups of 2 isolates with similar RFLP patterns.
fig.ommitted
Table 2. Strain similarities among 838 Campylobacter isolates from birds and humans in 67 households in a Peruvian periurban shantytown, as determined by restriction-fragment length polymorphism (RFLP) analysis.
Similar strains in a chicken and a human in the same household were present in 29 (70.7%) of 41 families. However, when restricted to families with the largest number of Campylobacter isolates (20 isolates/family), similar strains in a chicken and a human were present in 9 of 9 families. An example of a family group with similar RFLP patterns in chickens and humans is shown in . Similar strains in 2 different chickens in the same household were present in 51 (87.9%) of 58 families. When analysis was limited to families with 20 Campylobacter isolates, similar strains in 2 different chickens were present in 9 of 9 families. The number of RFLP-based strain groups isolated from birds in the same household was directly related to the total number of Campylobacter isolates recovered from the household (0–
10 isolates, mean 1.0 groups; 11–
20 isolates, 2.4 groups; 21–
30 isolates, 3.2 groups; and >30 isolates, 6.8 groups). Similar strains in 2 different humans in the same household (not applicable when Campylobacter species was only recovered from 1 human in the family) were present in 10 (40%) of 25 families.
fig.ommitted
Figure 1. Restriction–
fragment length polymorphism patterns for 3 human strains and 6 avian strains of Campylobacter jejuni collected from a single household (family 12). All strains were tested in duplicate. Similar patterns for human and avian strains collected from the same household are seen in lanes 4–
9 and 14–
19 (2 avian and 2 human isolate with the same pattern in each group). Lanes 1 and 20, 100-bp DNA ladder; lanes 2 and 3, CMP 956 strain from chicken; lanes 4 and 5, CMP 1263 strain from chicken; lanes 6 and 7, CMP 1265 strain from human; lanes 8 and 9, CMP strain from chicken; lanes 10 and 11, CMP 1756 from chicken; lanes 12 and 13, CMP 1264 strain from chicken; lanes 14 and 15, CMP 1320 strain from chicken; lanes 16 and 17, CMP 2235 strain from chicken; lanes 18 and 19, CMP 2421 strain from human.
Association between presence of similar strains of Campylobacter in humans and birds and risk of diarrhea. Overall, 59 (38%) of 155 Campylobacter isolates from humans had an identical RFLP profile to a chicken isolate from the same household. There were no differences between the frequency of identical chicken and human strains within households when human isolates from diarrhea cases (12/34 [>35.2%]) and nondiarrhea cases (47/121 [38.8%]) were compared. We did not find a correlation between any specific RFLP pattern and diarrhea (vs. asymptomatic infection) in the entire study group, nor was there an association between any RAPD pattern or Lior serotype and diarrhea in the subgroups with RAPD and serotyping data.
A more sophisticated analysis was performed to analyze similarities between Campylobacter strains from humans and chickens as a function of time and presence of diarrhea. A total of 840 household-months of observation were recorded in the study, and, in 100 of these, 1 human was infected with Campylobacter species. In 20 of these 100 household-months, 1 household bird shared the same RFLP strain type as the human strain within 1 month of (i.e., between 15 days before and 15 days after) the human infection. When analyzed by person-months instead of household-months, we recorded a total of 147 person-months in which a human was infected with Campylobacter species, and in 29 (19.7%) of these, 1 household bird shared the same strain as the human within a 1-month period.
We identified 17 case subjects with persistent infection which was defined as isolation of Campylobacter strains with the same RFLP type from the same person on 2 occasions within a 60-day period. Eleven of the 17 case subjects were identified by reculturing children with Campylobacter diarrhea at 2-week follow-up intervals. We also identified 18 case subjects of reinfection with a new RFLP type within a 60-day period. In 7 (39%) of these 18 case subjects, the second Campylobacter isolate was associated with diarrhea, whereas only 3 (17.6%) of 17 second isolates from persistent infections were related to diarrhea (P = .25). No differences in the frequency of similar strains in humans and chickens within a 1-month period were found when isolates from persistent infections and reinfections were compared.
Using a logistic regression for longitudinal data based on GEEs, we analyzed the effect of isolating C. jejuni from a chicken within 1 month of isolating the same RFLP type from a human in the same household on the probability that the human infection was associated with diarrhea. In this analysis, the presence of the same Campylobacter strain type in stool samples from a chicken and a human in the same household within a 1-month period was highly associated with the absence of diarrhea. Alternatively, one could say that the presence of shared Campylobacter strains between chickens and humans in a household was statistically protective against Campylobacter gastroenteritis. Of 34 Campylobacter strains from humans associated with diarrhea, 1 (3%) coincided with a strain isolated from a chicken in the same household within a 1-month period. In contrast, of 124 Campylobacter strains from humans which were not associated with diarrhea, 28 (22.6%) coincided with a strain isolated from a chicken in the same household within 1 month (P < .005; OR, 0.07). This protective effect of sharing a strain on the presence of diarrhea was more pronounced in children aged >340 days versus those aged 340 days (break point based on distribution of sample), although sample size limitations did not allow us to report significant ORs for each age-based stratification group. Neither seasonality nor age at the time of infection was significantly associated with the probability of diarrhea.
Comparison of RFLP and RAPD for the detection of Campylobacter strain variation. The correlation between RFLP and RAPD patterns were compared within individual households. Using RFLP as the reference standard, we found that, of 164 strain groups with the same RFLP pattern, 97 (59.1%) also shared the same RAPD pattern. The degree of similarity, as defined by RAPD, increased as the number of strains in the RFLP "group" increased (e.g., among "groups" of 2 strains with same RFLP: 14/42 strains [33.3%] had similar RAPD; "groups" of 3 by RFLP: 26/42 strains [61.9%] had similar RAPD; and "groups" of 4 by RFLP: 57/80 [71.2%] had similar RAPD).
Using RAPD as the reference standard, the percentage of strains that also had corresponding RFLP patterns was higher than when RFLP was used as the reference (i.e., 100 [78.1%] of 128 strains in groups with similar RAPD also had similar RFLP pattern), indicating that variability by RAPD was greater than that by RFLP. As above, similarities by RFLP increased as the number of strains in the RAPD "group" increased (e.g., among "groups" of 2 strains with same RAPD: 42/58 strains [72.4%] had similar RFLP; "groups" of 3 by RAPD: 58/70 strains [82.8%] with similar RAPD).
In most case subjects, RAPD techniques analyzed by ProSCORE showed more strain differences than by RFLP. In 3 case subjects, 6 RAPDS types corresponded to a single RFLP type among strains of Campylobacter species from a single family cluster. However, 1 RAPD type that was very common in both chickens and children in the study (strain 10) corresponded to 7 RFLP types; thus, subtle differences are not always detected more commonly with RAPD. In general, RAPD appeared to detect more strain differences than RFLP, although many of these strain variations may have been due to subtle changes in electrophoresis pattern that may be misinterpreted as a strain difference using RAPD.
We also compared results of RAPD tests done on a pooled Campylobacter isolate with results performed using 3 separate clones. Of 105 isolates tested, only 7 (6.7%) had 2 different RAPD profiles, when results from individual clones and pooled isolates were compared. Because these results show that differences between clones and pooled specimens were very uncommon, we conducted RAPD and RFLP analyses, as reported here by use of pooled isolates and 1 clone only. In >98% of cases, the pattern demonstrated by the clone was identical to the pattern demonstrated by the pooled isolate.
Comparison of Campylobacter strain variation by RFLP, RAPD, and Lior serotyping. Lior serotypes identified from the group of 159 strains tested are shown in . Strain-typing data are available by all 3 methods (Lior serotype, RFLP, and RAPD) for 127 strains from 16 households. Eight of the 159 serotyped strains agglutinated with >1 serotype, and, in 1 household, we identified 2 chickens carrying strains that agglutinated with the same 2 serotypes (C. jejuni 1 and 2).
fig.ommitted
Table 3. Frequency of Lior serotypes among Campylobacter jejuni and Campylobacter coli strains isolated from birds and humans from the study population in Peru.
The common Lior serotypes of Campylobacter strains tested in our study group were similar to those commonly reported in the United States, including outbreak-associated strains and strains from sporadic disease [16, 17]. Four of the 5 most common serotypes in our study also are commonly reported from the United States (serotypes 1, 2, 4, and 36). One exception was Lior serotype 32, which was the fourth most common serotype in our study but is uncommonly reported in data from the United States.
One hundred twenty-five strains in groups of 2 with the same RFLP type and from the same household were serotyped by the Lior method. In 59 (47%) of 125 strains, groups with the same RFLP pattern also had the same Lior serotype, whereas in 66 (53%) case subjects, groups with the same RFLP had >1 Lior serotype represented. However, many of the strains in the latter category were in groups of 3 with the same RFLP, and, in several cases, there were partial matches by Lior serotype as well—
that is, 2 of the group of 3 strains with the same RFLP also had the same serotype. If these partial matches are also included, the proportion of strains with the same RFLP and the same Lior serotype becomes 81 (65%) of 125.
RAPD data were available for 56 of 59 strains in which identical RFLP patterns corresponded exactly to identical Lior serotypes. Of these 56 strains, 31 (55%) had RAPD that also corresponded exactly to the other 2 methods. Therefore, only 31 (25%) of 125 stains, in groups of 2 with the same RFLP, also demonstrated identical matches by both Lior serotype and RAPD.
DISCUSSION
Our study population demonstrates epidemiological features typical of endemic campylobacteriosis in other parts of the developing world, including a risk for disease that is inversely proportional to age and high rates of asymptomatic infection. Our incidence rates of Campylobacter-related diarrhea in children aged 5 years (0.37 episodes/child/year) were similar to rates reported in Mexico and Thailand and similar to the rate reported elsewhere in this Peruvian community [2, 18, 19]. Campylobacter species was not associated with any of the 41 diarrhea episodes detected in persons aged 6 years, whereas it was associated among 22 (19%) of 112 diarrhea episodes detected in children aged <12 months. This very high prevalence of Campylobacter diarrhea in very young children (0.89 episodes/child/year) exceeds rates reported in previous studies and may be due to the combination of poverty and selection of households with free-ranging chickens. Although the entire study population were in the lower socioeconomic group (mean family income, $242.00 US/month) and all families had free-ranging chickens, human infection with C. jejuni was statistically associated with extreme poverty—
two-thirds of the families with human infections had a family income below the community average. In our study group, chicken feces in the home was the major environmental source of Campylobacter species, which was not detected in hand washings, stored foods, stored water, or soil samples. The high rates of asymptomatic infection are consistent with previous epidemiological data from Peru and other developing countries, where many studies have not found the striking differences in prevalence of enteropathogens between healthy children and children with diarrhea, as seen in the United States [3].
Our molecular strain–
typing tools demonstrated an extremely large number of Campylobacter strains circulating in this community, with an average of half as many distinct strain types as Campylobacter isolates in households studied. Despite the multiplicity of circulating strains, we were able to detect similar strains in chickens and humans in the same household in 70% of all families studied, and similar stains in chickens and humans were detected in all 9 of the households with >20 Campylobacter isolates studied. However, strains shared by humans and chickens in the same household within a 1-month period were unlikely to be associated with diarrhea in humans. In fact, the presence of the same Campylobacter strain type in stool samples from a chicken and a human in the same household within a 1-month period was statistically protective against Campylobacter-associated diarrhea. Even when we compared strains in chickens and humans within households over the entire study period, human strains from case subjects with diarrhea were no more likely to match a chicken isolate from the same household than nondiarrhea strains. This would suggest that most Campylobacter-associated diarrhea is due to strains acquired outside the home and that previous exposure to strains carried by chickens in the home confers a high degree of strain-specific immunity to disease but not immunity to repeated infection. Whether this hypothetical previous exposure occurred with disease before the start of surveillance or by asymptomatic infection during surveillance remain to be clarified. It is important to emphasize, however, that the RFLP pattern is merely one method to distinguish between similar and different strains of Campylobacter, so one cannot assume that the antigens responsible for the RFLP type are the same as those responsible for the observed protective immunity. The pathogenesis of Campylobacter gastroenteritis is complex and involves adhesins, invasion factors, antigens that induce apoptosis, and enterotoxins, and the relative importance of these virulence factors in induction of protective immunity is not known.
Because Campylobacter transmission between humans and chickens is a fluid process and may occur in both directions (chicken-to-human and human-to-chicken), our study was not able to detect the direction or temporal flow of organisms within family clusters. However, our data provide strong indirect evidence of high rates of chicken-to-human transmission in this high-risk population, although this flow of organisms did not appear to contribute much to diarrheal disease. Our data suggest that Campylobacter isolates from household chickens are responsible for only a minority of Campylobacter diarrhea cases in humans living in the same household, reminiscent of recent data from the Centers for Disease Control and Prevention, which have shown decreasing rates of Campylobacter diarrhea in the United States, despite increasing rates of chicken consumption [20]. To our knowledge, this is the first time that the epidemiological link between chickens and children with Campylobacter infection has been demonstrated in an endemic disease setting by use of molecular strain markers.
Our data also suggest that molecular strain typing and serotyping are probably of little value for evaluating the effectiveness of Campylobacter control strategies in similar communities with endemic disease and high levels of transmission. Although some RFLP patterns and Lior serotypes were found more frequently than others in the study population, there were no patterns or serotypes that were statistically associated with diarrheal disease. Because of the lack of predominant strains or serotypes, intervention strategies for Campylobacter species in endemic populations are best evaluated by surveillance of all isolates and not specific strains.
Our results illustrate some of the inherent difficulties in using molecular tools to compare bacterial strains in an endemic disease setting. Although both RFLP and Lior serotyping demonstrated less variability of strain types than RAPD, about one-third of isolates that demonstrated similarity by either RFLP or serotype were found to be different by the other assay. Although some variation of results is expected because the 2 systems classify strains on the basis of different microbiological features (flagellin genotype for RFLP and heat-labile surface antigens for Lior typing), this degree of discrepancy between 2 strain-typing systems suggests that it is difficult to definitively prove that 2 Campylobacter isolates from persons or birds in the same family cluster are identical, either because of differences that are not detected or because of constant genetic mutations that may affect one classification system but not the other. Indeed, recent reports confirm that genetic exchange between both homologous and heterologous Campylobacter strains occurs during experimental infections in chickens by RFLP, amplified fragment length polymorphism, and pulsed field gel electrophoresis, resulting in the generation of new genetic rearrangements even in the absence of immunologic pressure [21]. Although consistent identical types or patterns by 2 systems that examine different antigens would provide the best proof of animal-to-human transmission, in reality this ideal may be elusive because of rapid recombinations occurring in vivo. This hypothesis is supported by the relative paucity of strains with similar patterns by 1 of the 3 systems that were identical by all 3 methods (i.e., 25% of those tested).
Because of the genomic plasticity of Campylobacter species demonstrated by de Boer et al. [21] and also suggested by our findings, are strain typing results reliable in an endemic disease situation? It is probably reasonable to assume that background spontaneous strain variation occurs at a fairly constant rate that should affect all study populations in a similar way. Although recognizing that the tools at our disposal have identifiable limitations, our data demonstrate statistically significant differences that can be supported by biologically plausible explanations, supporting the use of molecular typing in certain situations. Our findings suggest that RAPD detects more strain differences than does RFLP, but minor differences in the electrophoresis pattern (perhaps due to minor genetic recombinations) may be misinterpreted as strain differences when RAPD is used. Because the results of strain typing by RFLP were less variable and more consistent within family groups than were RAPD results and because RFLP was technically easier than serotyping for large numbers of strains, we used RFLP as the typing method of choice for our study. Further studies will be needed to determine whether RFLP or Lior serotype is less susceptible to spontaneous variability.
In summary, humans and chickens in this periurban shantytown are exposed to a veritable sea of Campylobacter strains of many different serotypes and genetic patterns. Campylobacter infections were most prevalent in the youngest children, and human infection was associated with extreme poverty, although the incidence and prevalence of symptomatic infection versus asymptomatic infection were not different. Although it is impossible to identify predominant strains associated either with diarrhea or with human isolates versus isolates from chickens, similar strains of Campylobacter in chickens and humans were found in many family clusters, especially those with the largest number of bacterial isolates. However, most of these shared strains contributed little to the burden of disease. Bacterial strain typing by RAPD, RFLP, or Lior serotyping is unlikely to be useful in assessing Campylobacter control measures in this type of community because of the high multiplicity of strains circulating in the community. Variability in results arising from the use of different methods suggests that frequent ongoing mutations are occurring in bacterial populations, making it difficult to definitively identify identical strains. New control strategies for campylobacteriosis in similar communities should target sources of infection coming from outside the home, which may be more effective in reducing rates of disease.
Acknowledgments
We thank Stephen Walz for his valuable review and suggestions; Naval Medical Research Center, Detachment (Lima, Peru) for contributing support for diagnostic microbiology; Irving Nachamkin for providing training in restriction-fragment length polymorphism typing and use of software programs for strain identification; DNA ProSCAN for contributing software for analyzing the strain-typing data; and the Health Promoters Association and the people of the Pampas de San Juan for their participation.
References
1.
Oberhelman RA, Taylor D. Campylobacter infections in developing countries. In: Blaser MJ, Nachamkin I, eds. Campylobacter. Washington, DC: American Society for Microbiology, 1999; 139–
54.
2.
Oberhelman RA, Gilman RH, Taylor DN, et al. A placebo-controlled trial of Lactobacillus GG to prevent diarrhea in Peruvian children. J Pediatr 1999; 134:15–
20.
3.
Black RE, Lopez de Romana G, Brown KH, et al. Incidence and etiology of infantile diarrhea and major routes of transmission in Huascar, Peru. Am J Epidemiol 1989; 129:785–
99.
4.
Guerrant RL, Hughes J, Lima N, Crane J. Diarrhea in developed and developing countries: magnitude, special settings, and etiologies. Rev Infect Dis 1990; 12:S41–
50.
5.
Skirrow MB, Blaser MJ. Campylobacter jejuni. In: Blaser MJ, Smith PD, Ravdin JI, Greenberg HB, Guerrant RL, eds. Infections of the gastrointestinal tract, 1st ed. New York: Raven Press, 1995; 825–
48.
6.
Grados O, Bravo N, Black RE, Butzler JP. Paediatric campylobacter diarrhea from household exposure to live chickens in Lima, Peru. Bull World Health Organ 1988; 66:369–
74.
7.
Marquis GS, Ventura G, Gilman RH, et al. Fecal contamination of shanty town toddlers in households with non-corralled poultry, Lima, Peru. Am J Public Health 1990; 80:146–
9.
8.
Hernandez J, Alonso JL, Fayos A, Amoros I, Owen RJ. Development of a PCR assay combined with a short enrichment culture for detection of Campylobacter jejuni in estaturine surface waters. FEMS Microbiol Lett 1995; 127:201–
6.
9.
Akopyans N, Bukanov N, Westblom T, Kresovich S, Berg D. DNA diversity among clinical isolates of Helicobacter pylori detected by PCR-based RAPD fingerprinting. Nucleic Acids Res 1992; 20:5137–
42.
10.
Hernandez J, Fayos A, Ferrus MA, Owen RJ. Random amplified polymorphic DNA fingerprinting of Campylobacter jejuni and C. coli isolated from human faeces, seawater, and poultry products. Res Microbiol 1995; 146:685–
96.
11.
Welsh J, McClelland M. Fingerprinting genomes using PCR with arbitrary primers. Nucleic Acids Res 1990; 18:7213–
8.
12.
Winters DK, Slavik MF. Evaluation of a PCR based assay for specific detection of Campylobacter jejuni in chicken washes. Mol Cell Probes 1995; 9:307–
10.
13.
Nachamkin I, Ung H, Patton C. Analysis of HL and O serotypes of Campylobacter strains by the flagellin gene typing system. J Clin Microbiol 1996; 34:277–
81.
14.
Nachamkin I, Bohachick K, Patton C. Flagellin typing of Campylobacter jejuni by restriction fragment length polymorphism analysis. J Clin Microbiol 1993; 31:1531–
6.
15.
Lior H, Woodward DL, Edgar JA, Laroche LJ, Gill P. Serotyping of Campylobacter jejuni by slide agglutination based on heat-labile antigenic factors. J Clin Microbiol 1982; 15:761–
8.
16.
Bopp CA, Birkness KA, Wachsmuth IK, Barrett TJ. In vitro antimicrobial susceptibility, plasmid analysis, and serotyping of epidemic-associated Campylobacter jejuni. J Clin Microbiol 1985; 21:4–
7.
17.
Patton CM, Barrett TJ, Morris GK. Comparison of the Penner and Lior methods for serotyping Campylobacter spp. J Clin Microbiol 1985; 22:558–
65.
18.
Calva JJ, Ruiz-Palacios GM, Lopez-Vidal AB, Ramos A, Bojalil R. Cohort study of intestinal infection with Campylobacter in Mexican children. Lancet 1988; 1:503–
6.
19.
Taylor DN, Echeverria P, Pitarangsi C, Seriwatana J, Bodhidatta L, Blaser M. Influence of strain characteristics and immunity on the epidemiology of Campylobacter infections in Thailand. J Clin Microbiol 1988; 26:863–
8.
20.
Preliminary FoodNet data on the incidence of foodborne illnesses: selected sites, United States 1999. MMWR Morb Mortal Wkly Rep 2000; 49:201–
5.
21.
De Boer P, Wagenaar JA, Achterberg RP, van Putten JPM, Schouls LM, Duim B. Generation of Campylobacter jejuni genetic diversity in vivo. Mol Microbiol 2002; 44:351–
9., http://www.100md.com(Richard A. Oberhelman Robert H. Gilman Patricia Sheen Julianna Cordova David N. Taylor Mirko Zimic R)