Poliovirus detection in wastewater and stools following an immunization campaign in Havana, Cuba

Abstract

Background Recent outbreaks of poliomyelitis caused by vaccine-derived virus have raised concerns that vaccine-derived poliovirus may continue to circulate after eradication. In these outbreaks, the virus appears to have replicated for ≥2 years before detection. Early detection is critical for an effective response to these outbreaks. Although acute flaccid paralysis (AFP) surveillance will remain the standard for poliovirus detection, wastewater sampling could be a useful supplement. In this study, we evaluated the sensitivity of wastewater sampling by concurrently collecting stools from children aged <3 years attending two neighbourhood clinics in Havana, Cuba, and wastewater from the same neighbourhoods.

Methods Sample collection was begun during the third week after the national immunization campaign, continued weekly through the seventh week, and was repeated during weeks 15 and 19. Virus detection and titration were performed using both cell culture and polymerase chain reaction techniques.

Results Wastewater sampling was found to be at least as sensitive as stool sampling under these conditions. Poliovirus was isolated from children through week 7, suggesting that viral shedding reached undetectable levels between weeks 8 and 14. The last virus-positive wastewater sample was collected during week 15.

Conclusions Wastewater sampling under the conditions studied can be a sensitive supplement to AFP surveillance. Similar studies under different conditions are needed to determine the role of wastewater sampling in post-eradication surveillance.

The 1988 World Health Assembly of the World Health Organization (WHO) passed a resolution committing to the global eradication of the poliovirus by the end of the year 2000 (WHA41.28). Since then, the number of reported polio cases has decreased from an estimated 350 000 in 125 countries during 1988 to fewer than 1000 in 10 countries during 2001.¹ As the number of cases decreases, however, we are faced with additional challenges to our systems for detecting circulating poliovirus. Recent outbreaks of poliomyelitis on the islands of Hispaniola,^2,3 the Philippines,⁴ and Madagascar⁵ have demonstrated that vaccine-derived polioviruses can circulate in a virulent form. Furthermore, regional and global certification committees may rely on non-standard sources of information such as wastewater sampling in their task of declaring their areas polio-free. These factors accentuate the need to maintain surveillance in polio-free areas and after discontinuing oral polio vaccine (OPV) use, and to understand the sensitivities of approaches other than acute flacid paralysis (AFP) surveillance for detecting poliovirus in a population.

Maintaining the ability to detect circulating poliovirus both before and after certification will pose unique challenges. One challenge will be to sustain the motivation necessary for intense AFP surveillance. Another will be the sharp decrease in the sensitivity of AFP surveillance at low infection rates,⁶ and the lack of sufficient data to fully understand the conditions that could give rise to extended low-level transmission. Also, the future expansion of the use of inactivated polio vaccine (IPV), which provides limited protection against infection, could contribute to reduced surveillance sensitivity. The ability of genetically mutated vaccine viruses to replicate for about 2 years before their detection, as in outbreaks in Hispaniola and the Philippines, both areas that had been declared polio-free, emphasizes the difficulty and importance of sensitive post-eradication surveillance.

One approach to increasing sensitivity is to supplement AFP surveillance with methods such as sampling wastewater to test for the presence of poliovirus. To use wastewater sampling as an effective surveillance supplement, we must understand its population sensitivity, defined as the probability of obtaining at least one sample that is positive for poliovirus when the virus is circulating within the population.⁶ A number of sensitive laboratory procedures have been developed to recover virus from a sample, but few published data exist to address the sensitivity of a wastewater-based surveillance system in a population. Data from Sweden^7,8 and Finland⁹ have documented the ability to detect vaccine-derived poliovirus in wastewater during periods of routine immunization, during wild virus outbreaks, and following a national campaign. Researchers in Finland recently demonstrated the ability to detect vaccine virus in a municipal wastewater collection system for 4 days after introducing a known quantity of vaccine virus into the system.¹⁰ These data provide only limited ability to compare the results from wastewater sampling to other measures or indicators of virus circulation.

Fewer studies have attempted to relate the rate of virus shedding within the population to the likelihood of detecting virus in wastewater. One such study was conducted as part of an immunization campaign using Sabin type 3 OPV during a 1960 type 3 outbreak in Atlanta, GA.¹¹ Periodic wastewater samples were collected from neighbourhoods recently the targets of a type 3 OPV immunization campaign, while concurrent stool samples were collected from children attending daycare within the same neighbourhoods. The stool samples yielded virus for 12 weeks following the campaign, while the wastewater samples yielded virus for 14 weeks. Although these data provide useful information, they were collected as part of an immunization campaign in communities chosen because of their risk status, not because of their suitability for studying wastewater sampling. In addition, the samples were processed and tested by simpler, less sensitive methods than those currently available.

Cuba provides an ideal environment in which to investigate the sensitivity of wastewater sampling. Since the early 1960s, Cuban health officials have administered OPV solely through mass immunization campaigns;^12–15 no OPV is given through routine childhood immunization. Because of this, it is possible to estimate the proportion of recipients shedding virus starting from a known time of introduction, without background virus from routine immunization. In the study described here we compared the frequencies of poliovirus detections in wastewater with the rates of viral shedding in the community as estimated by a concurrent stool survey of vaccine recipients following a mass vaccination campaign in Havana, Cuba.

Methods

Study locations

Samples were collected following the second of the two 1997– 1998 mass immunization campaigns. Four criteria were used to select study locations in Havana. First, to minimize the presence of industrial or similar waste in the wastewater, the location had to be a residential neighbourhood. Second, the neighbourhood must have a collective wastewater system that served only that neighbourhood and have at least two accessible sampling points. Third, the wastewater at the sampling sites must have come only from the target neighbourhood. Finally, at least one polyclinic must have been present in the neighbourhood, serving only residents of that neighbourhood. The latter two criteria insure that the stool and wastewater samples represent the same population and only that population. We selected two Havana neighbourhoods that met these criteria, designated ‘A’ and ‘B’ in this report.

Stool sampling

Stool samples were collected from randomly selected children aged <3 years attending the neighbourhood polyclinics either for a well child visit or injury. This age criterion captured children targeted for vaccination in the second round of the mass campaign. Because the study’s main objective was to compare rates of viral shedding with rates of viral detection in wastewater, we only recorded each child’s date of birth and place of residence. No additional information on medical history or vaccine status was collected. However, Cuba has a long history of maintaining very high participation rates in its National Immunization Days (NID).¹⁶ We assumed, therefore, that almost all of the children in this study had received OPV prior to stool collection.

Beginning in the third week after the last national vaccination campaign (week 3), samples were collected through week 7, with additional samples collected during weeks 15 and 19 through 21. (Results for the weeks 19–21 samples were grouped and are reported here as week 19 results.) Our target was to collect stools from 50 children in each neighbourhood during each sampling week. Children were enrolled in the study after we obtained informed consent from one of the parents. The parent was given a container and instructions for stool collection. The parent kept the stool sample refrigerated until returning it to the family physician at the polyclinic. The samples were then taken from the clinic to the Instituto Pedro Kouri (IPK) for testing.

Wastewater sampling

The methods we used for collecting and testing wastewater samples were chosen over potentially more sensitive methods because they are more easily implemented in a wide variety of laboratories (WHO field guide in preparation). Wastewater samples were collected from two separate locations within each neighbourhood during the same weeks in which stools were collected. During each sampling week, samples were collected on two different days from each of the sampling sites. On each of these days, 500 ml were collected from each site in the morning and again in the afternoon. These were combined into a single 1-l composite sample for each site. Samples were collected by dipping the collection flask directly into the wastewater effluent. The outer surface of the flask was then rinsed and the flask placed in a cold box for transport to IPK.

Isolation and identification of poliovirus in stool

Virus isolation was performed according to the following methods. First, for each stool specimen a 20% (w/v) suspension was made using phosphate buffered saline (PBS) with antibiotics. Chloroform was then added to the mixture, agitated vigorously, and the mixture clarified by centrifugation at 10 000 r.p.m. in an Eppendorf centrifuge. The supernatant was kept frozen at –20°C until innoculation onto cell monolayers. For innoculation, 0.2 ml of the supernatant was innoculated on rhabdomyosarcoma (RD) and L20b cell lines. The culture tubes were then incubated at 37°C and examined every 24 hours for evidence of cytopathic effect (CPE). Samples showing CPE were frozen and thawed for passage onto new cells. A sample was observed for 12 days before being considered negative. For samples showing CPE in L20b cells, poliovirus serotypes were identified by neutralization with hyperimmune serum pools. Four preparations were formed by mixing sera with antibodies against types 1 and 2, types 1 and 3, types 2 and 3, or types 1, 2, and 3 in a solution containing 50 neutralization units of each type of antiserum and approximately 100 TCD₅₀ of the isolate (32–320 TCD₅₀). Because the L20b cells are less sensitive than the RD cells, all samples showing CPE only in RD cells were passed onto L20b cells also. If CPE resulted in the L20b cells, the identification proceeded as previously described.

Extraction of virus particles from sewage samples

Virus was extracted and concentrated by the method recommended by WHO. The sewage was centrifuged at 4°C for 30 minutes at 5000 × g to sediment (pellet) the sewage solids. The supernatant was recovered for further processing. The solids were then re-suspended in 5 volumes of elution medium (3% beef extract), and 1/2 volume of chloroform added. This mixture was centrifuged at 4°C for 30 minutes at 5000 × g to sediment the chloroform and sewage solids, and the aqueous supernatant was recovered. The chloroform and sewage solids were supplemented with another volume of elution medium, the mixture re-extracted and centrifuged, and the resulting aqueous supernatant recovered. The supernatant extracts from the sewage solids were combined with the original sewage supernatant and viruses were precipitated from the mixture with polyethylene glycol 8% (PEG) and sodium chloride 0.3 M (NaCl) at 4°C overnight. The precipitated viruses were sedimented by centrifugation at 5000 × g for 1 hour at 4°C. The resulting sediment was re-suspended in 2 ml of PBS and the mixture was extracted with chloroform. After centrifuging to remove the chloroform, the aqueous supernatant was recovered and the remaining chloroform was re-extracted with a volume of elution medium. After centrifugation, this elution medium was recovered and combined with the previously collected aqueous supernatant to obtain the concentrated virus in a final volume of 6 ml. The sample was then treated with penicillin (1000 units/ml) and streptomycin (200 μg/ml). A sample was considered negative if no virus was detected in the undiluted concentrate using PCR or cell innoculation methods. The amount of virus in the positive samples was titrated using serial dilution with virus detection performed using both PCR and virus isolation.

For the PCR procedures, RNA extraction was performed with TRIZOL (Life technologies TM, Gibco BRL) according to the manufacturer’s instructions. Briefly, 0.25 ml of concentrated sewage samples were homogenized in 0.75 ml of TRIZOL and 0.2 ml of chloroform. After centrifugation at 4°C for 15 minutes at 12 000 × g, the aqueous phase containing RNA was precipitated with isopropanol (v/v) at –20°C overnight. After centrifugation for 10 minutes at 16 000 × g the supernatant was discarded and the pellet washed in l ml of 75% ethanol, dried and suspended in 50 μl of diethylpyrocarbonate (DEPC)-treated water.

Sequences of a primer pair matching highly conserved sites within the 5’-non-coding regions of enterovirus genomes were EV PCR-1 (A, 539 to 565) 5’ACACGGACACCCAAAGTAGTCGGTTCC-3’, and EV PCR-2 (S, 452 to 476) 5’-TCCGGCCCCTGAATGCGGCTAATCC-3’.

The degenerate primers used for amplifying poliovirus sequences were pan PV PCR-1 (A, 2935 to 2954) 5’-TTIAIIGC(A/G)TGICC(A/G)TT(A/G)TT-3’, and pan PV PCR-2 (S, 2875 to 2894) 5’-CITAITCI(A/C)GITT(C/T)GA(C/T)ATG-3’. The number in parentheses indicates the genomic intervals matching the primers (A, antisense or antigenome polarity; S, sense or genome polarity), according to the numbering system of Toyoda et al.¹⁷ Deoxyinosine residues are indicated by the letter I. Primer positions having equimolar amounts of two different nucleotides are enclosed in parentheses.

In vitro amplification by PCR was performed by a modification of methods described previously.¹⁸ Amplification reactions were carried out in 50 μl reaction mixture containing RNA template in PCR buffer (67 mM Tris-HCL; pH 8.8, 17 mM NH₄SO₄, 6 μM EDTA, 2 mM MgCl₂, 1 mM 2-mercaptoethanol), 100 ng of each primer, 100 μM (each) dATP, dGTP, dGTP, dTTP (Pharmacia Biotech, Piscataway, NJ), 5 U of placental RNase inhibitor (Boehringer Mannheim Biochemicals, Indianapolis, IN.), 1.5 U of avian myeloblastosis virus reverse transcriptase (Boehringer Mannheim), and 1.25 U of Taq DNA polymerase (Boehringer Mannheim). The reaction mixture were overlaid with mineral oil, and incubated at 42°C for 30 minutes prior to 30 cycles of programmed amplification (denaturation at 94°C for 45 sec, annealing at 42°C for 45 seconds, and extension at 70°C for 45 seconds) in a DNA thermal cycler (Perkin-Elmer-Cetus). Conditions for polyacrylamide gel electrophoresis and detection of amplified products by ethidium bromide staining were as described previously.¹⁹

Equal numbers of test samples and negative controls were processed in each test batch to detect possible extraneous nucleic acid sequences contaminating the RNA extraction and polymerase chain reaction (PCR) reagents were amplified by PCR. Additional procedures to avoid PCR contamination were adopted as elsewhere.²⁰

For virus found in wastewater, serotype identification was performed using hyperimmune sera and serotype-specific PCR primers.

Results

Stool samples

Stools were collected from 677 children ranging in age from 0 through 35 months. The percentage of children positive for at least one poliovirus serotype remained relatively constant from week 3 through week 6, and then declined beginning week 7. None was found to be positive during weeks 15 or 19 (Table 1). Roughly the same pattern was observed for each of the three individual serotypes. To provide a more conservative point of comparison for determining the sensitivity of wastewater sampling, we also calculated the 90% upper confidence limit for each week’s percentage of children shedding. The lowest upper limit was 2% during weeks 15 and 19. The upper bound after combining these two weeks was 1%.

The pattern for non-polio enteroviruses (NPEV), however, was opposite that for polioviruses. The percentage of samples positive for NPEV increased sharply during week 7, and remained at this level during weeks 15 and 19. During each week, the percentage of children positive for poliovirus decreased with age. Combining all time periods, 41% (20/49) of the children under 6 months of age were positive for poliovirus, 22% (28/130) of those 6–11 months of age, 16% (36/231) of those 12–23 months of age, and 12% (31/258) of those 24–35 months of age.

Wastewater samples

By either virus isolation or PCR, the percentage of wastewater samples positive for poliovirus declined with time (Table 2). During weeks 3 and 4 after the campaign, all samples were found to contain virus regardless of testing method. The percentage containing virus declined after that; the last positive sample was collected during week 15. Only one positive sample was found for this week, and it was found to be positive by both virus isolation and PCR.

Comparing virus isolation with PCR, we found that the percentage of samples positive by PCR was equal to or greater than the percentage positive by virus isolation during each week. However, we did not find virus in any of the week 19 samples using either method.

We also determined the virus titre for each positive sample. From week 6 forward, virus was detected only in the undiluted sample by either detection method, so the virus titre added no additional information. Examining the mean log virus titres from weeks 3 to 5 reveals a different pattern, however. For weeks 3 and 4, all of the samples contained poliovirus, but the geometric mean viral titre increased sharply between these weeks. During week 3, the geometric mean viral titres were 1:7.5 by both virus isolation and PCR. By week 4, however, the titres were 1:177.8 and 1:74.0 by isolation and PCR respectively. By week 5, only 5 of the 8 samples collected were positive for poliovirus. The geometric mean viral titers for these positive samples were 1:4.0 by using virus isolation and 1:10.0 by using PCR.

Relationship of wastewater isolation to rate of viral shedding

To compare the percentage of children shedding virus with the probability of a positive wastewater sample, we fitted a probit cubit curve to the weekly wastewater results using the proportion of children shedding poliovirus as the independent variable using SAS Proc Probit {SAS, 1999 #1561}. Using the best fit-curve (intercept = 1.72, slope = –7.76), we estimated that a shedding rate of 31% is required to have a 0.75 probability of obtaining a positive wastewater sample.

Serotype-specific results

Overall, type 1 was found most often (8% of stools; 27% of wastewater samples), followed by type 3 (8% and 16%), with type 2 found least often (4% and 9%). Comparing serotype-specific isolations from stool by week, we found that the percentage of children shedding type 1 virus peaks during weeks 5 (15%) and 6 (19%), dropping to only 6% during week 7. The percentage shedding type 2 virus is 6% or lower for all but week 5, during which 11% where shedding type 2. The pattern for type 3 shedding was slightly different. During weeks 3 and 4, type 3 was the most frequently isolated serotype (15% and 16% of children tested, respectively). Also, the shedding pattern for type 3 did not have a distinct peak as was seen for the other serotypes.

The serotype-specific temporal patterns for virus isolation from wastewater differed from the stool isolation patterns. In wastewater, isolation of each of the serotypes showed a distinct peak. The highest isolation rate for types 1 and 3 where during week 4 (88% and 50%, respectively). The highest rate for type 2, however, was during week 3 (38%). During all weeks but week 6, the percentage of samples positive for type 1 was equal to or higher than the percentages for types 2 or 3. During week 6, one sample was found to contain a type 2 virus, while neither of the other serotypes was found. The last sample found to contain virus was from week 15; this sample contained type 1 virus.

Discussion

In this study we compared the frequencies of viral detection in wastewater with concurrent estimates of the proportions of children shedding poliovirus. The results indicate that sampling wastewater from a collective system as is found in Havana can be a sensitive procedure for detecting vaccine-derived poliovirus in a population. The results also show that PCR can be at least as sensitive for detecting poliovirus as the use of cell culture and hyperimmune sera. Finally, this study supports earlier findings that vaccine virus is present in the Cuban population for only a few weeks following a mass campaign.^21,22

We were able to detect type 1 virus in the wastewater when only about 6% of the sampled children were shedding this serotype. Furthermore, we found virus in wastewater samples in the absence of positive stool 15 weeks following the mass campaign. At the other extreme, all of the wastewater samples tested positive for at least one virus serotype when about 25% of the sampled children were shedding. These findings indicate that wastewater sampling under these conditions can be a sensitive tool for moderately low levels of poliovirus circulation.

The wastewater samples from this study may not reflect virus excretion from children wearing diapers. We assume, however, that Havana is similar to a large number of cities throughout the world in which many infants will not be represented in a wastewater sample. The absence of these children from our samples then, does not compromise our objective of evaluating this method for virus detection.

We found it interesting that the levels of type 2 poliovirus were low both in stool and in wastewater. Generally, type 2 Sabin virus produces a greater immunological response than the other serotypes, especially after the first dose. Almost all of the children who participated in the second round of NID had already received at least one dose of trivalent OPV, so it is likely that most had responded well to the type 2 component of earlier doses, and were less likely to shed virus. Unfortunately, we have no data from this study to either support or refute this hypothesis.

We also found it interesting that in the stool specimens, the rates of polio and non-polio enterovirus isolations were inversely related. The pattern was found across all age groups (data not presented), suggesting that the OPV virus may have interfered with NPEV in the gut or in the cell culture. Unfortunately, we do not have data that would allow us to investigate this observation further.

Our findings about the sensitivity of wastewater sampling are encouraging, but we must be cautious about generalizing too broadly. The first caveat is that wastewater collection systems differ widely among countries and communities, and sensitivities for detecting poliovirus may differ considerably among these systems. In the neighbourhoods we sampled in Havana, wastewater is carried away from the houses by an inter-connected, closed system that empties into an open canal. The areas were specifically chosen to be free of industrial waste or agricultural runoff. Other methods that are used to dispose of wastewater include collective systems that carry the waste to a central treatment facility, open channels that lead from houses or communities to larger open canals, septic tanks, and pit latrines. The sensitivities of sampling from these different systems have not been systematically compared. The second caveat is that different laboratory processing and detection methods may have different detection sensitivities. For example, the data from our study indicate that detection by using PCR may have been more sensitive than detection using cell culture.

From the standpoint of the polio eradication programme, an important question is whether wastewater sampling can become a standard tool to supplement AFP surveillance. If this is to happen, we must have available a robust procedure that can be implemented in a wide variety of laboratories. Our finding that the technically simpler cell culture method produced results similar to those using PCR is encouraging. Additional systematic comparisons of methods with different types of samples are still needed.

The third caveat is that the sensitivity of any detection system can be increased by collecting and testing more samples. Because of this, evaluation of candidate supplemental surveillance approaches should also compare the amount of effort and resources each would require to achieve equal sensitivities. During the first week of sampling in our study, for example, 26% of the children were shedding at least one of the poliovirus serotypes, while 100% of the wastewater samples were positive. With this level of shedding, testing a single stool sample would have a sensitivity of 26%, while testing a single wastewater sample would have a sensitivity near 100%. Under these same conditions, however, a random sample of 16 children would raise the sensitivity of the stool survey to more that 99% (i.e. a probability > 0.99 of detecting at least one poliovirus).

As a final caveat, it must be understood that we can learn nothing about the transmission or circulation of vaccine-derived polioviruses from these data. The patterns of poliovirus isolations could reflect decreasing levels of circulation, latency within the wastewater system, the normal range of viral shedding durations, or all of these.

In conclusion, this study has provided important data that have expanded our understanding of the sensitivity of wastewater sampling and testing. Before we attempt to implement this approach more widely, we need additional studies and data to estimate the sensitivity of these methods for early detection of an outbreak. We especially need more data that inform us of the sensitivity of wastewater sampling relative to AFP surveillance at different levels of circulation. As part of this work, we must identify which types of wastewater collection systems are reasonable to sample, which processing and testing methods are feasible and cost-effective for more widespread use, and what sampling schemes can yield the best results. Only after we obtain a better understanding of the limits of wastewater-based surveillance will we be able to determine its appropriate role in the final stages of polio eradication.

The research reported here was funded by the Pan American Health Organization. The authors wish to acknowledge the invaluable assistance of Dr Ciro de Quadros, Dr Claudio Silviera, and Ms Nina Toro of PAHO; Dr Guadalupe Guzman and Dr Alina Llop of the Instituto Pedro Kouri; and, Ms Yvonne Stifel of the Centers for Disease Control and Prevention.

References