Savill ESR Ltd, Christchurch Science Centre, PO Box 29-181, Ilam, Christchurch, New Zealand (e-mail: firstname.lastname@example.org).
Aims: To use a published polymerase chain reaction (PCR) method for the detection and identification of thermotolerant Campylobacter species (Camp. jejuni, Camp. coli and Camp. lari) in tandem with a Most Probable Number (MPN) technique to enumerate these species in water samples.
Methods and Results: An initial study of 42 river water samples compared the use of conventional culture and PCR methods for the detection of Campylobacter in MPN enrichment tubes. It was found that all samples positive by culture were also positive by PCR. Thirty-seven percent more MPN tubes were positive by PCR compared with culture. The MPN/PCR technique was subsequently applied to 96 additional samples collected from rivers, drinking, roof and shallow ground water. Campylobacter was especially prevalent in river water (60% positive) and shallow ground water (75% positive) samples. Drinking water (29·2% positive) and roof water (37·5% positive) also contained Campylobacter, but the numbers detected were very low (maximum 0·3 and 0·56 MPN 100 ml–1, respectively).
Conclusions: River waters contained Campylobacter at higher levels than any other water type and in a high percentage of the samples. Although Campylobacter was present in treated drinking water, the levels detected were low.
Significance and Impact of the Study: These results suggest that water may act as a significant transmission route for human campylobacteriosis.
New Zealanders suffer a very high rate of campylobacteriosis, with an incidence in 1999 of 224·8 cases/100 000 population (Anon. 2000). Contaminated water has been implicated in several campylobacteriosis outbreaks in New Zealand (e.g. Stehr-Green et al. 1991; Brieseman 1994) and overseas (Koenraad et al. 1997). The use of rainwater as a source of water at home has been identified as a risk factor for sporadic campylobacteriosis in a New Zealand case-control study (Eberhardt-Phillips et al. 1997).
Little information on the distribution and numbers of Campylobacter in New Zealand waters exists, although Hudson et al. (1999) isolated the organism from selected water sources at mean rates of 75% positive in the winter and 31% positive in the summer. Overseas, a study of a river system in the UK found 43% of samples positive, with counts of Campylobacter at levels up to 230 MPN 100 ml–1 (Bolton et al. 1987), and 82·1% of German river water samples contained Campylobacter, although the majority had levels of < 10 MPN 100 ml–1 (Stelzer and Jacob 1991). It is therefore evident that Campylobacter can be isolated frequently from river water, although they are usually present in low numbers. Campylobacter have also been isolated from 13% of Spanish marine recreational waters studied (Alonso and Alonso 1993) and recently, Camp. jejuni has been isolated from ground water in the UK (Stanley et al. 1998).
Conventionally, pathogens have not been monitored in water because indicator organism testing has been considered to reflect adequately the presence of enteric pathogens. However, the correlation between Campylobacter and indicator organisms is unclear. Carter et al. (1987) found no significant correlation between Campylobacter counts and plate counts, faecal or total coliforms, or faecal streptococci. In contrast to this, Skjerve and Brennhovd (1992) used faecal coliforms and water temperature to produce a model to predict the presence of Campylobacter in water. It is possible that Campylobacter is present in waters only when faecal indicators are present, but that their density is not correlated with the actual numbers of indicators present. Alonso and Alonso (1993) found this to be the case in marine waters. Stanley et al. (1998) found Campylobacter in ground water that contained faecal indicators, but Campylobacter was at its highest count (implied from the data) when faecal indicators were at their lowest.
The work described here sought to adapt a published PCR method for the detection of Campylobacter for use with an MPN format of water enrichments. Once the PCR method had been established, it was tested for its ability to detect Campylobacter in enrichment broths and compared with the ability of conventional culturing to achieve the same purpose. After validation, the MPN/PCR method was applied to water samples from surface, drinking (treated and untreated) and shallow ground waters so that they could be assessed in terms of the number and diversity of Campylobacter species present.
MATERIALS AND METHODS
Development of the PCR assay
Primers and PCR Conditions.
The primers and PCR conditions used were those described by Eyers et al. (1993,1994). these primers target the 23s rRNA gene of Campylobacter and allow detection of the four thermotolerant Campylobacter species (Camp. coli, Camp. jejuni subsp. jejuni, Camp. lari, and Camp. upsaliensis) by using one primer pair, THERM 1 and THERM 2. Additional primers allow discrimination between the thermotolerant species. Campylobacter upsaliensis was not considered in this study and so the primer sequences used were (5′-3′); THERM 1, TATTCCAATACCAACATTAGT, THERM 2, CGGTACGGGCAACATTAG, THERM 3, TAAAGTAAGTA-CCGAAGCTG, LARI, ACGGCATCAGCAATTCTC, COLI, TAAATCCTAATACGAAGCG, JEJ 1, GTAAATCCTAATACAAAGCT, and JEJ 2, TAAATCCTAGTACGAAGCT.
DNA isolated from water samples was amplified initially using primers THERM 1, THERM 2 and LARI. This indicated the presence of thermotolerant Campylobacter species as well as allowing the specific identification of Camp. lari. Samples generating an amplicon, indicating the presence of thermotolerant Campylobacter, were assayed in two further reactions to demonstrate the presence of either Camp. coli (primers THERM 1 and COLI) or Camp. jejuni subsp. jejuni (primers THERM 3, JEJ 1 and JEJ 2). Variability in the 23S rRNA sequence requires two reverse primers for the detection of all Camp. jejuni (Eyres et al. 1993).
To confirm the identity of the THERM 1 : THERM 2 product derived from the primers described above, an internal primer within the THERM 1 : THERM 2 product was designed so that a nested PCR could be conducted using the amplified THERM 1:THERM 2 product as a template. This new primer was designed by aligning Campylobacter 23S rRNA DNA sequences from GenBank (http://ncbi.nlm.nih.gov/genbank), and identifying a suitable internal region. The primer was screened using BLAST (http://www.ncbi.nih.gov/blast/blast/cgi) for non-specific cross-reactivity. The sequence of the new primer was 5′-TAGCGGATGGAAGTGCTAG-3′. This was used in combination with the THERM 2 primer to produce a 167 bp product.
In samples where a product was formed with the primers specific for Camp. lari, the PCR product was re-analysed by polyacrylamide gel electrophoresis (PAGE) for enhanced resolution. This was performed as it became apparent that products slightly different in size from the Camp. lari product could be produced (i.e. false positives). Only those products of the correct size were recorded as containing Camp. lari.
PCR amplifications were performed in 0·5 ml tubes in a total reaction volume of 50 μl using PCR buffer (0·050 mol l–1 KCl, 0·010 mol l–1 Tris, pH 8·4), 5 pmol of each of the required primers, 2·5 Units of AmpliTaq, 200 μmol l–1 dNTPs, DNA from Campylobacter and 1·5 mmol l–1 MgCl2, for the general thermotolerant campylobacters, Camp. lari and Camp. jejuni reactions, or 2·0 mmol–1 MgCl2 for the Camp. coli reaction. Bovine serum albumin was added at a final concentration of 0·2 mg ml–1 to prevent inhibition of amplification by any contaminants in the cultures. Thermal cycling was carried out in a programmable DNA thermal cycler (Hybaid Omnigene, Hybaid Ltd, Ashford, UK). The thermal cycling conditions used were an initial denaturation at 94°C for 3 min, followed by denaturing at 94°C for 1 min, 54°C annealing for 1 min, 74°C extension for 1 min, over 40 cycles, with a final 8 min extension step at 74°C. For each PCR run, positive (10 μl of a mixture of DNA from Camp. jejuni, Camp. coli and Camp. lari, 100 ng of DNA each) and negative (10 μl of autoclaved double distilled [ddH2O]) controls were included.
DNA was prepared for PCR by two methods, a phenol chloroform method and a crude heat lysis method. The phenol/chloroform method was used to extract DNA from pure cultures. A 10 ml aliquot of each culture was harvested by centrifugation at 3300 g for 20 min, the supernatant fluid removed and 300 μl extraction buffer (0·025 mol l–1 Trizma base, 0·010 mol l–1 EDTA, 0·050 mol l–1 glucose) added. Alternatively, a large loopful of fresh bacterial culture was harvested from an agar plate and suspended in 300 μl extraction buffer. Lysis was achieved by the addition of 20 μl lysozyme (50 mg ml–1 in ddH2O), with incubation for 5 min at room temperature, followed by 12 μl SDS (20%) and 4 μl proteinase K (10 mg ml–1 in ddH2O), and incubation continued at 37°C for 30 min.
DNA was prepared using a phenol chloroform extraction as previously described (Sambrook et al. 1989). The DNA was dissolved in 20 μl ddH2O and stored at – 20°C. The concentration of DNA was measured spectrophotometrically (λ=260, 280) and the sample diluted to give a final concentration of approximately 100 ng μl–1. A 10 μl aliquot of DNA prepared in this manner was added to the PCR premix.
Heat lysis was used for water sample enrichments. A 1 ml sample of each broth was centrifuged at 4000 g for 20 min to pellet the cells. The supernatant fluid was discarded and the pellet washed three times in phosphate-buffered saline (PBS). The suspension was heated (100°C for 12 min) to lyse the cells, centrifuged (11 750 g for 10 min), and 10 μl of the supernatant fluid added directly to the premix for PCR analysis.
Visualization of PCR products.
Depending on the resolution required, PCR products were resolved using either agarose electrophoresis (2%) or PAGE (9·6%). Agarose electrophoresis produces lower resolution. The samples were subjected to agarose gel electrophoresis for approximately 75 min at 100 V in TBE buffer (0·09 mol l–1 Trizma base, 0·09 mol l–1 boric acid, 0·02 mol l–1 EDTA, pH 8·0) containing 0·5 μl ml–1 ethidium bromide to enable visualization of PCR products by u.v. transillumination. Molecular weight markers were included at both ends of each gel (123 bp DNA ladder, Life Technologies, Rockville, MD, USA) to ensure the products were of the correct size and consistent interpretation of samples in the gel. Polaroid photographs were taken of the fluorescently-stained PCR product.
When enhanced resolution was required, PAGE was performed at 200 V for 3·5 h. The gel was removed, stained with Sybr®Green (FMC BioProducts, Rockland, ME USA) for 30 min in the dark, and photographed while illuminated by u.v. light.
Tests for false-positive reactions with other bacteria.
The PCR assay was tested against six Campylobacter species including the four thermotolerant species (Camp. jejuni subsp. jejuni, Camp. coli, Camp. lari, Camp. upsaliensis). Eyers et al. (1993) tested a number of organisms to ensure specificity, and this list was extended using an additional 27 organisms (Table 1). This included other bacteria that might potentially confuse interpretation of results because they were either taxonomically closely related to Campylobacter, or might be expected to be present in similar environmental niches.
Table 1. Micro-organisms tested against Campylobacter primers
Determination of sensitivity of the PCR.
A culture of Camp. jejuni was grown for 48 h in Preston Enrichment (PE) broth (Bolton et al. 1982) at 42°C. A decimal dilution series was prepared in PBS. Two 0·1 ml volumes were spread-plated onto blood agar plates and incubated at 42°C for 48 h under microaerophilic conditions generated by the Oxoid (Basingstoke, UK) CampyGenTM system. Colonies on the plates were counted to determine the number of cells present. In addition, cells from each dilution were harvested as described above and the DNA extracted by heat lysis. The DNA was used in PCR assays. The number of Campylobacter cells required for a positive PCR test was determined and the minimum number of intact dead cells which, if present in the original water, would have produced a positive result was calculated. To ensure that the number of intact dead cells needed to produce a false-positive result would be very large, two enrichment steps were used in the final format, thereby virtually guaranteeing the detection of only viable cells.
Sample treatment prior to detection of Campylobacter
Water samples were tested using different MPN formats depending on the likely presence of Campylobacter in the sample. For recreational water (rivers) and shallow ground water, 1 × 500 ml, 3 × 100 ml and 3 × 10 ml samples were tested. For collected roof water, 3 × 500 ml and 3 × 100 ml volumes, and for reticulated drinking water, 3 × 500 ml samples, were tested. The water samples were filtered through sterile 0·45 μm membrane filters and the filters were transferred to 25 ml PE broth and incubated at 42°C for 48 h. Following this, a 100 μl volume was used to inoculate a second PE broth (10 ml) which was incubated at 42°C for a further 24 h. This second broth was processed for detection by either culture or PCR methods as required.
The use of PE broth significantly reduces the ability of Camp. upsaliensis to be detected because of sensitivity to the antibiotics used (Aspinall et al. 1996). Of note, however, is that this organism has not been found in New Zealand clinical cases to date to the knowledge of the authors.
The method was initially trialled in a study where Campylobacter was enumerated in 42 river water samples collected in the morning and afternoon from three different rivers on seven occasions over three months in autumn and winter. MPNs were performed by both culture and PCR methods. To isolate Campylobacter by culture, a loopful of each MPN enrichment broth was streaked onto Campylobacter selective plates (modified CCDA, Oxoid) and incubated microaerophilically at 42°C for 48 h. Colonies which were characteristic of Campylobacter were purified and identified by their Gram stain, motility, oxidase reaction, hippurate hydrolysis and sensitivity to nalidixic acid and cephalothin. In addition, a 1 ml volume from the same enrichment series was tested by PCR as described.
Water samples from five rivers, reticulated drinking water in four towns and ground water from three bores were collected once each month for six months. Also, 24 different roof water samples were collected over the same period at a rate of four samples per month. Samples were collected in sterile 2l containers. The bottles used to collect chlorinated water contained sodium thiosulphate to neutralize chlorine. Surface water sites were selected to include the three used in the pilot study. The four reticulated drinking water sites were selected to represent a range of different water treatment methods. Two of the shallow ground water sites were infiltration galleries supplying water for treatment and reticulation, samples of which were tested in the reticulated drinking water set. The third ground water sample was extracted from an unprotected aquifer. Roof water samples were collected from 24 different roof water tanks in one area of New Zealand.
All samples were enriched in the MPN formats described above; Campylobacter species were detected and identified by PCR methods. Total coliform and Escherichia coli counts were obtained for 100 ml of each water sample using the Colilert Quanti-Tray MPN method (Clesceri et al. 1998).
The correlation between total coliforms, E. coli and Campylobacter spp. was determined using the Spearman rank correlation test. As there were significant differences between the microbiological contamination of different water types, a separate correlation matrix was calculated for each.
Characteristics of the PCR developed
All four thermotolerant Campylobacter species tested yielded a 222 bp product when tested with the THERM 1 : THERM 2 primer set. In addition, Camp. hyoilei gave a positive result. This organism has been shown to be synonymous with Camp. coli (Vandamme et al. 1997). This primer set failed to generate PCR products from DNA isolated from Camp.fetus subsp. fetus and other non-Campylobacter species tested (Table 1). This result was independent of template preparation. The three thermotolerant Campylobacter species, Camp. jejuni, Camp. coli and Camp. lari, which were the focus of this study, all produced the correct PCR product with the primers specified for their individual species. Campylobacter lari yielded a 177 bp product with the THERM 1, THERM 2 and LARI primers, Camp. coli a 345 bp product with the THERM 1 and COLI primers, and Camp. jejuni a 600 or 700 bp product with the THERM 3, JEJ 1 and JEJ 2 primers.
These PCR products were not of the sizes indicated by Eyers et al. (1993), but inspection of 23S rRNA gene sequences for Camp. jejuni, Camp. coli and Camp. lari (GenBank: http://ncbi.nlm.nih.gov/genbank) revealed that the sizes produced experimentally were those that would be predicted by the DNA sequences available.
For samples that showed a visible product corresponding to thermotolerant Campylobacter spp. but did not yield a corresponding product for the species Camp. jejuni, Camp. coli or Camp. lari, it was necessary to confirm the THERM 1 : THERM 2 PCR product. These samples were re-tested using the nested primer described, which in all cases confirmed the THERM1:THERM 2 result.
The PCR assay detected Camp. jejuni in broth culture down to a dilution which equated to 75 cells per PCR (data not shown). It was calculated that at least 1·5 × 106 non-viable cells 100 ml–1 would have to be present in a 500 ml volume of sample to produce a false-positive result. The same calculations showed that 7·5 × 106 and 7·5 × 107 cells 100 ml–1 would be required to produce false-positive results in the 100 and 10 ml samples, respectively.
A pilot study comparing conventional culture and enrichment PCR methods demonstrated a good correlation between the two methods. Of the 42 samples tested, 27 (64·3%) gave the same result for the two detection methods. Values varied from < 0·12 MPN 100 ml–1 to > 11 MPN 100 ml–1. In five samples (11·9%), detection by culture yielded values of < 0·12 MPN 100 ml–1 (i.e. not detected), while detection by PCR gave measurable values. Examination of the data for the 294 enrichment tubes showed that all samples positive by culture were positive by PCR, but PCR detected more positive results than conventional culture, with 100 of the 294 enrichments positive by culture and 137 positive by PCR.
The distributions of the PCR/MPN values are shown in Fig. 1 (since conventional culture/MPN data were very similar to MPN/PCR data they are not shown for clarity). These results show that one river, River A, contained Campylobacter at a generally higher level than the other two.
Given that there was good agreement between the two methods, it was concluded that the PCR method was of use for the enumeration of Campylobacter in water samples, and was applied to another set of samples as described below.
Thermotolerant Campylobacter spp. were detected by PCR in each of the four water types tested in the study. Summarized results of Campylobacter, faecal coliforms and E. coli analyses are shown in Table 2. Campylobacter were detected more frequently in shallow ground water and surface water than in roof waters and reticulated drinking water supplies. Three-quarters of the shallow ground water samples and 60% of the river water samples contained Campylobacter spp.
Table 2. Summary statistics of Campylobacter and indicators organisms by water type
River waters were the most contaminated of the four types tested. Total coliforms and E. coli were always detected at each of the surface water sites. Campylobacter were detected in 60% of the samples and at levels higher than in the other water types. For example, the highest number estimated in any other source was 0·72 MPN 100 ml–1, whereas 37% of the river water samples had MPN values in excess of this.
The microbiological quality of the roof waters was generally poor, with E. coli being detected in nine of the 24 samples (38%) and total coliforms detected in 18/24 (75%). Campylobacter were detected in 9/24 (37·5%) of roof waters sampled; they were detected in the absence of E. coli on five occasions and in the absence of total coliforms on two occasions.
All of the reticulated drinking water samples collected in the study complied with New Zealand’s drinking water standards (Ministry of Health 1995) in that E. coli was not detected in 100 ml of any sample. A single sample contained a total coliform count of 1 MPN 100 ml–1.
Distribution of different thermotolerant Campylobacter spp.
The occurrence of different thermotolerant Campylobacter spp. in the different water types is shown in Fig. 2. Campylobacter coli and Camp. lari were detected in each of the water types tested. Campylobacter jejuni was detected in some of the surface water, shallow ground water and reticulated drinking water samples but in none of the roof water samples.
Relationship between Campylobacter and indicator organisms
A significant correlation was found only between the concentration of total coliforms and Camp. jejuni in shallow ground water (P=0·026). Significant correlations were not observed between thermotolerant Campylobacter, or the individual Campylobacter species and E. coli or total coliforms in the other three water types. However, a low Spearman’s rs value does not allow ‘no correlation’ to be inferred, but merely denotes that there may be insufficient evidence to conclude that a relationship exists.
Campylobacter were detected in 19/52 (37%) samples where E. coli was absent. Furthermore, the concentrations of Campylobacter did not exceed 1 MPN 100 ml–1 until total coliform and E. coli concentrations exceeded 500 MPN 100 ml–1 and 100 MPN 100 ml–1, respectively.
The presence of Campylobacter in two-thirds of shallow ground water samples is probably a reflection of the nature of the ground water sites used in the study. Two of the three sites were infiltration galleries used as sources of drinking water, although the water is treated prior to reticulation. This result highlights the fact that infiltration galleries should not be used as the sole means of treating drinking water. Coliforms were detected often at two of the shallow ground water sites (the infiltration galleries), and E. coli was only isolated from the infiltration gallery waters. No unprotected aquifer sample yielded faecal indicators.
The roof water samples were all collected from one district in the North Island where few communities have reticulated water supplies. The presence of Campylobacter in roof water samples is therefore not necessarily indicated by E. coli or total coliforms. The presence of Campylobacter in roof water is of concern as there are 416 registered roof water supplies in New Zealand and many more unregistered, most of which are untreated (Ball 2000). However, the concentration of Campylobacter was low and did not exceed 0·56 100 ml–1. Water derived from storage tanks has been implicated in outbreaks of campylobacteriosis. For example, 234 pupils and 23 staff were affected over an 8 week outbreak at a boarding school where the storage tank was thought to be contaminated by roosting birds (Palmer et al. 1983).
The presence of Campylobacter in 29% of reticulated drinking water samples is of concern, but this concern is largely mitigated because the concentration of Campylobacter was low, with a maximum MPN of 0·3 100 ml–1 in any of the 24 samples tested (Table 2). Most of the isolates from reticulated water which were identified were Camp. lari and it is possible that this organism has different survival characteristics compared with the two major pathogenic Campylobacter species. This observation will be tested in further studies.
The number of dead but intact Campylobacter cells required to produce a false-positive result was very high (at least 1·5 × 106) in the MPN format used. While it is difficult to estimate the number of intact but dead cells that may be found in water, a worst case for live cells may be represented by the influx of a sewage works. Keonraad et al. (1994) found a maximum of approximately 104Campylobacter cells 100 ml–1 in such influx water. Assuming that all of these cells passed through treatment and were killed but remained intact and were undiluted, it is still unlikely that the MPN method used would have yielded a false-positive result.
Teunis et al. (1997) produced a sigmoidal dose/dependence curve for Campylobacter, relating numbers ingested to the probability of infection. The data they presented indicated that, for one of the isolates tested, a dose of 800 cells gave a 50% probability of infection. Unfortunately, the data on which the curve is based do not include values below 800 cells (Black et al. 1988) and so extrapolation below this number is subject to wide variations in the 95% confidence intervals.
Table 3 shows estimations of the volume of water that would have to be consumed for 800 Camp. jejuni cells to be ingested and hence, to give a 50% probability of infection. In the case of roof water, drinking water and shallow ground water, the estimate based on the available data would be that large, undrinkable quantities would have to be consumed for there to be a 50% probability of infection. In the case of surface water, the risks are much more real, especially if the concentrations in those samples recorded as >11 100 ml–1 were significantly in excess of 11 100 ml–1. Further studies on river waters should accommodate the enumeration of Campylobacter at higher levels than this to obtain a better understanding of the risks that may exist.
Table 3. Estimates of the quantity of water required to be consumed for a 50% probability of infection
These values are based on the species ingested being Camp. jejuni, but this organism is not predominant in the species that were identified. Ingestion of Camp. coli may result in a similar risk, but the risk from Camp. lari is probably much less.
The results of this work indicate that Campylobacter is frequently present in water. In water sources regarded as potable, the number of Campylobacter present, and the fact that a significant proportion of the samples contained Camp. lari, indicate that the risk posed to human health from these sources is small. Further studies into the efficacy of water treatments on different species of Campylobacter are warranted, as the organism was detected at very low concentrations in treated drinking water.
The authors wish to acknowledge the assistance of the Health Protection Officers and personnel from Canterbury, South Canterbury, West Coast, Tairawhiti and Waikato, who helped in the collection of water samples, the Public Health Laboratory for their assistance in processing the water samples, Els Maas for carrying out the sequence alignments, Nick Garrett for assistance with statistical analyses and Jan Gregor for advice on reticulated water supplies. They thank the Ministry of Health for financial support.