Comparative analysis of molecular and conventional methods for bacteriological water quality assessment in drinking water resources around Chennai

Microbial contamination in water resources is a major problem leading to several diseases. Multiple microbes lead to water contamination, whereas in most cases Escherichia coli and coliforms are applied as indicator organisms to monitor microbial contamination of water. The coliforms are Gram-negative, non-spore forming bacilli with the ability to ferment lactose to produce acid and gas. E. coli is a group of bacteria that are used as indicators of fecal contamination in water and is commonly found in many other environments such as aquatic, soil and vegetation. E. coli and coliform bacteria should not be detected in any water that is intended for drinking (Cotruvo 2017). The presence of these microorganisms implies a lack of system integrity and poor general sanitary quality of the water, leading to greater water-borne diseases (Maheux et al. 2014). Fecal substances from humans and animals are major contributors to microbial contamination in water sources, particularly the groundwater system. Disposal of untreated and improperly treated fecal and sewage contributes to the bacteriological contamination of water. In several countries, untreated sewage from towns and cities is directly disposed into open water bodies. Groundwater is a major water resource in India (Khurana & Sen 2008).

Increasing population, inadequate and unhygienic potable water supply are the foremost concerns in India. In India, only a few reports on the examination of bacterial fecal indicators have been studied. The level of fecal contaminants in the river Ganges revealed that microbial activity was abundant and varied seasonally, especially more during monsoon followed by summer and then winter (Baghel et al. 2005). The drinking water quality index of ground water in the inland region of southern Tamil Nadu was linked to numerous parameters. Fecal indicators indicated that the coliform bacteria level exceeded the guideline values (Ramesh et al. 2010). Research on the microbial contamination level in the west coast of India (Rodrigues et al. 2011) and rural and urban households of Karnataka (Mukhopadhyay et al. 2012) indicated higher microbial activity found in rural water supplies and a large proportion of isolates had multidrug resistance characteristics, which was alarming. Recently, a study conducted in Sikkim indicated possible fecal contamination of potable water in the community reservoir tanks and spring water used for household purposes. The authors suggested several monitoring and management options that should be followed by the state to maintain a hygienic drinking water supply chain (Singh et al. 2019).

Several approaches and monitoring devices have been developed by researchers for the detection and enumeration of indicator bacteria in the water supply. The conventional methods of detection are based on selective growth of indicator microbes on substrate-specific chromogenic media but suffer from long incubation periods, possibilities of contamination, inaccuracy, and lack of specificity. The modern polymerase chain reaction (PCR) based molecular techniques that detect microorganisms in the water sample are highly sensitive, specific and faster, thereby eliminating the need for long incubation time. It is widely used as a rapid detection tool in several countries to determine the quality and monitor their water systems (Fatemeh et al. 2014). The PCR-based method is not quantitative and hence cannot determine the extent of contamination. The drawbacks of the above methods warrant the need for an efficient and rapid method, namely the quantitative polymerase chain reaction (qPCR) technique.

Over a period, several genes specific to E. coli and total coliforms have been investigated as PCR primers or markers (Patel et al. 2011; Walker et al. 2017). The gene lacZ responsible for the production of enzyme β-D-galactosidase is used to detect total coliforms by PCR methods (Isfahani et al. 2017). The gene uidA encoding forβ-D-glucuronidase enzyme is largely found in E. coli (Bej et al. 1991)with frequent activities observed in Shigella and Yersinia (Molina et al. 2015).The conventional PCR methods demand further electrophoresis techniques to detect the presence of the target gene, whereas qPCR can amplify the specific gene and confirm their presence using fluorescent signals developed during the amplification of DNA during the reaction/amplification time.

Different types of drinking water resources available to the general public were initially identified in Chennai, Tamil Nadu. Most of the population in Tamil Nadu depends upon the Government water supply, which goes through chlorination for sanitization. Lakes/ponds, community wells, groundwater and overhead tanks are the major resources available to the people in every city. The current study aims to quantify the load of coliforms and E. coli in these water sources through the rapid molecular qPCR technique and compare with the plating-based method.

Water samples were collected from lakes/ponds, community wells, groundwater and overhead tanks from different areas in and outside Chennai (Figure 1). A total of 39 water samples from 12 different sites were collected and analyzed for the presence of E. coli and coliform through detection of genes uidA and lacZ using qPCR.

Water samples were collected in 500 mL sterile, wide mouth plastic bottles (Tarsons Product Pvt. Ltd, India). All sample containers were washed in distilled water, dried and autoclaved prior to the use. The samples are collected from faucets as per USEPA guidelines on sampling for biological contaminants to strictly ensure the inside of the bottles and the cap were not contaminated, while leaving a one inch headspace in all the containers for air and mixing before analysis (USEPA 2005). The samples were then stored in the refrigerator at 4 °C for a maximum of 24 hours until further processing. The details of the samples and their sources are given in supplementary Table S1.

The efficiency of culture-based and molecular qPCR methods to detect Escherichia coli and total coliforms was studied using strains obtained from Microbial Type Culture Collection (MTCC). The representative microorganisms for microbial contamination in water, Escherichia coli (MTCC 1687) and coliforms belonging to genera Citrobacter (Citrobacter freundii; MTCC1658), Enterobacter (Enterobacter aerogenes; MTCC 6806) and Klebsiella (Klebsiella pneumoniae; MTCC 109) were used as a consortium in this study. The microorganisms were revived according to the manual of instructions provided by MTCC in their respective media and stored as glycerol stocks in a −80 °C deep freezer for further experiments.

The assay was performed using a modified USEPA Method 1604 (USEPA 2002). HiCrome™ M-TEC agar and HiCrome™ chromogenic coliform agar (HiMedia Laboratories, India) are prepared under manufacturer's instruction and used for the detection of E. coli and total coliforms respectively. 100 ml of the water sample was filtered on a 0.45 μm pore size 47-mm diameter cellulose acetate filter (Millipore Corporation, USA) in a UV sterilized filtration unit (Tarsons Product Pvt. Ltd, India) and the filter was carefully transferred using flamed forceps on to the agar plates. The plates were incubated at 44.5 ±0.2 °C and 37 °C for 24 hours. The performance of the plates was tested with positive and negative control strains C.freundii, E. coli and P.aeruginosa. Sterility control was included for all the experiments as mentioned (APHA 2005).

The membrane filters (after filtration of samples) was transferred aseptically to a vial containing 1.5mlbuffer solution. The bacterial cells on the filter were collected by centrifugation at 13,000 rpm for 10 min. An optional upstream bead beating step using 0.1 mm diameter sterilized glass beads was included to enhance the recovery of cells from the filters. The genomic DNA (gDNA) was recovered from the bacterial pellet using the HiPurA™Bacterial Genomic DNA Purification Kit (HiMedia Laboratories, India) and the concentration of extracted gDNA yield was quantified by using NanoDrop™ 2000/2000c spectrophotometer (ThermoFisher Scientific, Massachusetts, USA). Finally, 50 μl of elution buffer was added for eluting DNA from the column and the extracted DNA is stored at −20 °C for further analysis. The gDNA standards were prepared by isolating genomic DNA from standard E. coli and coliforms (C. freundii MTCC1658, E. aerogenes MTCC 6806 and K. pneumoniae MTCC 109) purchased from culture collection. Performance characteristics of qPCR assays were determined from standard curves constructed using gDNA standard concentrations and the threshold cycle (C_T) values from qPCR.

The gene sequence for uidA targeting E. coli was acquired from NCBI (National Center for Biotechnology Information) and the corresponding gene FASTA sequence was uploaded in the Primer-BLAST to generate a specific primer for the template. A suitable primer with optimal concentration for qPCR was selected for this study. The degenerate primer of the lacZ (Martín et al. 2010) was designed using 22 different lacZ gene sequences of coliforms to generate the primer from the conserved region of all the sequences. The primers used are given in the Table 1.

The qPCR method is a robust and efficient technique eliminating the need for any post analytical processing for enumeration and detection. The efficiency of the method and the specificity of the primer were assessed with a log phase bacterial suspension of E. coli and coliforms at different dilutions. Briefly, 2 μl of genomic DNA isolated from the environmental samples was added to a 20 μl qPCR reaction mixture containing 10 μl of TB (Takara Bio) Green Premix Ex TaqII (2X) (TaKaRa Bio Inc, RR 820A), 0.4 μl of forward and reverse primer (0.4 μM) and the remaining reaction was made up with RNAse free sterile water. The ideal thermal cycling conditions for detection of E. coli were initial denaturation at 95 °C for 30sec, followed by 40 cycles of denaturation at 95 °C for 5 sec, primer annealing at 60 °C for 20 sec and extension at 72 °C for 20 sec followed by melt curve analysis. For detection of total coliforms, the conditions were followed as described elsewhere (Martín et al. 2010). A no template control (NTC) containing all the elements without the DNA template was included in all the experiments.

Various diagnostic parameters including amplification efficiency, limit of detection (LOD), specificity, sensitivity, positive and negative predictive values, positive and negative likelihood ratios were determined to test the efficiency of the qPCR analysis (Alberg et al. 2004; Caraguel et al. 2011; Nutz et al. 2011; Sedighi 2013; Gensberger et al. 2014a). The positive and negative samples were adjudicated in comparison with the culture method based on the diagnostic parameters (true positives and true negatives, false positive and false negatives)

The acceptable values of slope are (−3.67 ≤ slope ≤ −3.1) and the amplification efficiency (87–110%) in both the standard curves of Escherichia coli and total coliforms to establish qPCR assay performance are provided in the USEPA method 1611 (USEPA 2013)

The association between the culture-based and qPCR method was calculated by Fisher's exact tests (odds ratios and 95% confidence intervals) using positive and negative sources using GraphPad Prism version 6.0.

The presence of E. coli and coliforms in water samples is ascertained by conventional microbiological method using differential media. Purple to magenta colonies indicate E. coli and pink colonies indicate coliforms (Figure 2). Out of 39 samples, a higher number (CFU/ml) of E. coli or coliforms was detected in 31 different samples by qPCR compared to the culture method (Table S1)

The standard curve was linear for E. coli over 5 log units, namely from 10⁵ to 10¹ CFU and for total coliforms; it was linear over 3 log units, from 10⁴ to 1 CFU. In order to establish the analytical detection limit, genomic DNA of serially diluted E. coli and total coliforms were isolated and subjected to qPCR analysis (Table S2-S3). C_T values for NTC (No Template Control) were set at maximum cycles and hence the limit of detection was set as 37. The robustness of the qPCR assay along with its amplification efficiency was assessed using DNA of the target bacterial indicators (E. coli and coliforms), isolated from their mid-log phase cultures (Table 2). The amplification efficiency was calculated using the slope obtained from the linear regression equation. The r² values were ≥0.9 and a slope of −3.33 and −3.24, which were between the acceptable limits mentioned for a standard curve to determine the efficiency of the experiment (USEPA 2013). The percentage of amplification efficiency (AE %) for standard curve of E. coli is 99.6% and for total coliforms is 103.39%, which is well within the acceptable amplification efficiency (Bustin et al. 2009). High amplification efficiency denotes the reproducibility, the analytical sensitivity, and the robustness of the assay adopted.

Out of 39 samples analyzed from different water resources in the city through the molecular qPCR method, 50% in the OHT (overhead tank) samples tested positive for E. coli and 70% for the presence of coliform (Table 3). Lake samples were found to be positive for the target microorganisms, though the number of samples selected for the study was less. Well samples (36%) tested positive for E. coli and less than 10% for coliforms. Lake and ground water samples also tested positive for both the organisms, but the number of samples collected in the locations were less. The culture-based method indicated significant presence of the organisms in OHT and wells. The results indicate that the source of water and regular cleaning is important to prevent the growth of these organisms in OHT.

Multiple diagnostic parameters were determined for the molecular based assay, to understand its efficiency. The development of false positive and false negative signals was observed for both the indicator organisms in the qPCR experiment when compared to the culture-based technique. A non-significant association (E. coli Odds ratio: 0.73, 95% confidence interval: 0.29–1.8 and coliform Odds ratio: 0.58, 95% confidence interval: 0.23–1.46) was observed between the culture-based and qPCR method. False positive and false negative results from qPCR were registered for E. coli (17.9% and ∼10.2%) and total coliforms (15.3% and 2.5%). There are a number of factors that can lead to varied results. The sensitivity and specificity of the qPCR ranged from 73 to 92% for detecting E. coli and total coliforms in water samples. Detection of total coliforms and E. coli retrieved positive predictive values of 66.67 and 59% respectively. The likelihood ratio for the detection of coliforms was better in comparison to E. coli detection.

The association between the microbiological culture-based assay using specific chromogenic agar and molecular qPCR analysis for the detection of E. coli and total coliforms were compared. A total of 39 samples from various water sources (overhead tank, ground water, wells and lakes) available to the public were collected for the study and analyzed. The specific medium used in our study was able to detect 13 positive samples (8 OHT, 3 well and 2 lake samples) both E. coli and 14 positive samples (10 OHT, 1 well, 1 lake and 2 groundwater samples) in coliforms. This accounts for 35.8% of the samples containing pathogenic bacteria. The use of membrane filtration for screening the presence of E. coli and coliforms in water samples has been shown to possess an increased recovery rate of 90% with the use of a selective differential medium. The high number of positive OHT samples warrants the application of proper disinfection methods by the government, filtration, boiling of water before consumption, or regular cleaning of the tanks.

Similarly, qPCR retrieved 9 samples positive for both E. coli and coliforms. In this study, we have successfully used the uidA and lacZ genes as a target for the detection of E. coli-coliform bacteria respectively. These molecular methods are accurate and allow rapid detection of pathogenic bacteria in water sources with minimal limitations. The primer used to amplify the specific gene of pathogenic bacteria in qPCR generated a fluorescent signal on amplification of the target gene through the use of SYBR green q-PCR mix in the reaction. A gene fragment of 196 bp was obtained using the designed primer for E. coli. Different gene targets for detecting bacterial indicators in different water resources have also been reported. Potable water and riverine system samples of river Gomti were studied for the presence of Enterobacter using their conserved sequences of 16S rRNA and 23S rRNA genes in qPCR analysis, which showed very high specificity for the detection of Enterobacter (Patel et al. 2016). A SYBR green qPCR assay with LT1 and ST1 primers for Enterotoxigenic E. coli (Patel et al. 2011) and with stx2 primers for Shiga toxin-producing E. coli (Ram et al. 2011) confirmed their presence in potable waters of a major city of northern India. Besides, several other bacterial pathogens (Entamoeba histolytica (16SrRNA primer), Giardia lamblia (16SrRNA primer), and Salmonella spp. (invasive A primer) have been detected in municipality water supplies (Shankar et al. 2019). High incidence of multidrug resistance and class 1 integrons of E. coli in the urban waters of river Yamuna water of India were also detected raising the concern for multidrug resistant (MDR) coliforms (Kaushik et al. 2019).

Alternatively, the real-time PCR technique has drawbacks including underestimation of pathogens due to losses in bacteria during sample filtration and nucleic acid extraction steps. A few of the samples were also found to be overestimated due to the inability to discriminate between live and dead forms of E. coli (Na et al. 2006). DNA is known to remain intact after cell death and may persist for a few days and up to 3 weeks after cell death. Therefore overestimation or false positive results may occur in using this technique (Nocker et al. 2007). Reports suggest that higher bacterial counts were observed in the qPCR method when compared to the culture method for the quantification of Listeria monocytogenes and Enterobacteriaceae in food samples (Takahashi et al. 2017). Persistence of dead fecal bacteria is still an indicator of water-borne disease, therefore qPCR could be viewed as the best assay for the detection of E. coli and coliforms. Negative and positive samples generated the odds of association between the two techniques, which was found to be insignificant (E.coli p-value: 0.64 and Coliform p-value: 0.35). The reference culture method also has drawbacks of its own as it underestimates the results of the qPCR method. Therefore, detection of E. coli and coliforms in complex environments requires the use of multiple gene primer probes. Multiple water quality indicator bacteria, larger sample size, and different assessment techniques should be used to obtain multiple lines of evidence on the occurrence of fecal contamination.

In order to study the efficiency and accuracy of the molecular method, several diagnostic parameters are to be determined. The drawbacks in the PCR could affect the sensitivity of the molecular detection method. The presence of inhibitors relative to the type of sample hinders both the culture and qPCR methods. The complexity of the sample influences the amplification efficiency and primer binding ability, often leading to false positive and false negative results (Schrader et al. 2012). Higher false positive (12%) and false negatives (13%) were assigned to Enterobactericiae detection in drinking water, where specific primers against uidA and 23SrRNA genes were employed for the detection of E. coli and Enterobactericiae. This was attributed to the drawbacks in the reference test, Colilert®-18 (Gensberger et al. 2014b). The sensitivity and specificity of qPCR was found to be on the lower end for the detection of E. coli (71.42 and 72%) and Enterbactericiae (92.31 and 76.92%). Similarly a lower positive predictive value for E. coli (59%) and higher value for Enterobactericiae (66.67%) was observed. Both the drinking and processed water samples showed lower sensitivity, specificity, negative and positive predictive values in qPCR analysis. Inefficient sample preparation was predicted as the reason for lower diagnostic parameters (Gensberger et al. 2014b). Both β-D-glucuronidase activity and uidA amplification were not found to effectively identify E. coli, which led to the design of alternative primers for detection, including yaiO (Molina et al. 2015), and ybbW (Walker et al. 2017). A qPCR assay based on uidA had lower sensitivity (71.42%) in comparison to 100% sensitivity using ybbW primers. The authors claim that none of the existing uidA qPCR assays possesses complete inclusivity and exclusivity (specificity and sensitivity). This is due to the detection of Shigella along with E. coli and false positive results for non-E.coli Escherichia species (Walker et al. 2017). Most of the studies do not report the diagnostic parameters in comparison to a reference method, therefore the reliability of such results is to be questioned. Ours is one of the few studies which make a comparison between two methods.

The low positive predictive values for E. coli and coliforms (59 and 66.67% respectively) from the current study is due to the cross specificity with other microbes (Shigella, Yersenia, Salmonella) possessing target genes or due to the detection of genes from dead bacteria. Besides, PCR retrieves positive signal for strains that possess the gene sequence but negative for the corresponding gene. Negative predictive values were estimated to be 81.82 and 95.23% for E. coli and coliform respectively, which ascertains the probability of no indicator bacteria present in the samples. Both positive and negative predictive nature of tests depends on the existence of genes in the strains tested. On the contrary, likelihood ratio does not depend on the prevalence of genes in the strains, but is a true indicator of the test. A positive coliform qPCR test significantly enhances the probability of its presence (LR⁺: 3.83) in comparison to low prediction of E. coli (LR⁺: 2.53). All the diagnostic parameters are predicted in comparison to the culture based method, therefore the limitations of the technique will affect the outcome. These low values arise due to signals from DNA of viable and dead cells, possible presence of chemical inhibitors in environmental samples (Wolffs et al. 2005), inefficient sensitivity/specificity with the current uidA/lacZ primers and sometimes due to contamination or improper handling of experiments. An insignificant association between the two techniques justifies the low predictability of the indicator organisms. A higher false positive ratio is acceptable only if the results of the study are screened further through confirmatory assays. The chances of reducing the rate of false positives in qPCR detection can also be achieved through proper sample preparation and culture enrichment methods such as immuno-magnetic separation (Nogva et al. 2000) of cell and free DNA and floatation (Wolffs et al. 2004). Propidiummonoazide coupled with DNA of E. coli inhibited PCR amplification of DNA derived from killed cells (Deshmukh et al. 2020). In few cases, the samples collected from environment were treated with sodium thiosulfate for removal of chlorine residues to increase the detection efficiency through qPCR (Isfahani et al. 2017).

The limitations involved in the traditional culture based microbiological method are their long incubation time for results and labor intensiveness. The method lacks sensitivity, specificity over molecular tools for enumeration of pathogens, which are often the reasons to promote the latter for detection. The culture-based methods in chromogenic agar plates have constraints that include interaction and growth of other microorganisms, inefficiency to identify and detect VBNC (viable but non-culturable) bacteria in the medium due to their low metabolic activity (Park et al. 2011). Moreover, the recovery rate of microbes in the chromogenic agar can change based on the method of concentrating the microorganisms in the sample and a change in colony specification might arise which needs further confirmation steps.

The majority of people in the city of Chennai, Tamil Nadu, depend on the water sources mentioned in the study. Overhead Tanks are one of the major sources of water supply to the society. The tanks are built around the city for storage and supply of water from time to time. The study reports 50 and 70% of OHT samples were detected with E. coli and coliform respectively, which indicates sanitation measures for controlling the spread of disease through these water supplies are not adequately taken care. The disinfection strategies of basic chlorination within permissible limits and membrane filtration should be adopted, especially during the monsoon season (July to September). It is recommended that monitoring through sampling of these resources be done twice in a year. Sewage run off into wells and ponds needs to be addressed to prevent the contamination of water bodies. Consumer or public education is the foremost criterion that is required to reduce the development of waterborne diseases.

The presence of E. coli-coliform in water resources from different locations in and around the city of Chennai (South India) was identified using molecular qPCR method and compared with the plating method, which is slow and time consuming. It is significant to monitor bacteriological indicators in OHT, which is one of the major sites of contamination, as identified in the present study. Measures to device the contamination through point of care testing equipment are essential to promote appropriate preventive measures to the community. The rapid molecular testing is compared with the standard microbiology technique to determine the specificity and efficiency of the study. Large-scale validation studies with substantial and extensive data points from different sources will be required to dictate the efficacy of the molecular method. Furthermore, a focus on advanced qPCR detection such as multiplex PCR and ddPCR (Digital Droplet PCR) for multiple target detection using meticulous design of primers in the environmental water samples is needed. Finally, the results of the current research are pertinent to untreated water resources available to the community; and it could be different for other water types.

The authors thank Hermes i Tickets Private Limited for the financial support.

The authors of the manuscript confirm that there are no known conflicts associated with this publication.

All relevant data are included in the paper or its Supplementary Information.