A total of 100 water samples were collected from different sources [pond surface water (SW)—48, HH water—29, and PSF—23] of Mathbaria subdistrict, Pirojpur. All water samples (500 ml each sample) were collected aseptically in sterile water bottle (leak-proof natural polypropylene wide mouth plastic bottles). The samples were placed in an insulated plastic box within ice after collection and transported to the laboratory.

The water samples (40% of 500 ml) were filtered with 0.22 µm polycarbonate membrane filters (Sartorius, USA) and the filter papers were then transferred into sterile alkaline peptone water (APW) (tryptone 10 g/l, sodium cholate 20 g/l; pH 8.6) labeled by sample names. Incubation was then continued at 37°C for 6–8 h for enrichment (32, 33) and approximately 10 µl of the enriched APW broth was streaked onto both thiosulfate-citrate-bile salts-sucrose (TCBS) agar (Eiken, Tokyo, Japan) and taurocholate-tellurite-gelatin agar (TTGA) media (HIMEDIA, India), which were then incubated at 37°C for 18–24 h. The suspected positive samples with specific colony morphology in specific media were then screened for the presence of cytochrome oxidase and gelatinase activity by growing on gelatin agar plate [tryptone 10.0 g/l, trypticase (BBL) 10.0 g/l, gelatin 30.0 g/l, an agar 16.0 g/l; pH: 7.4].

Immunochromatographic one step rapid test (Dipstick method) was performed for the identification of Vc O1 and O139 (34, 35). The qualitative detection of lipopolysaccharide (LPS) antigen of both serotypes was accomplished using monoclonal antibodies specific to Vc O1 and O139. In brief, four drops from APW medium after enrichment were taken into a test tube, and a dipstick test strip was placed vertically. After 20 min, the appearance of band in the strip was observed, and the identification of Vc O1 and O139 was accomplished in comparison with the positive and negative controls.

Vibrio cholerae isolates presumed as O1 serotype by Dipstick method was further tested using a serotype-specific monoclonal antibody, Inaba and Ogawa (Remel Europe Ltd., UK) by slide agglutination test (18). In brief, this was performed on a sterile glass slide by resuspending a colony from Gelatin agar plate into sterile physiological saline and subsequently mixing with antibody by tilting the glass slide back and forth. Based on the clumping appearance, the reaction is considered positive.

A 0.5 McFarland standard was prepared by mixing 0.05 ml of 1.175% barium chloride di-hydrate (BaCl₂·2H₂O), with 9.95 ml of 1% sulfuric acid (H₂SO₄) in a test tube with constant stirring (36). The tube was then sealed tightly and stored in the dark at room temperature.

Susceptibility of the identified Vc O1 strains to antimicrobial agents was determined in vitro by disc-diffusion method (36) with five different antibiotic disks (azithromycin, ciprofloxacin, erythromycin, tetracycline, and cotrimoxazole). The inocula of the test organisms were prepared by transferring loopful of colony from tryptic soy agar (Sigma, USA) medium into 9 ml of sterile Mueller Hinton broth (Difco, Sparks, MD, USA) and incubated at 37°C for 5 to 6 h. The bacterial cultures were compared with McFarland turbidity standard (10⁸ CFU/ml) and streaked evenly in three directions keeping at a 60°C angle onto the surface of the Mueller Hinton agar plate (10 mm × 40 mm) with sterile cotton swab (36). Surplus suspension was removed from the swab by rotating the swab against the side of the tube before the plate was seeded. After the moisture of the inocula dried up, the antibiotic disks were placed on the agar using sterile forceps and were gently pressed down to ensure contact. For each plate five disks were placed and were placed in a way so that they were no closer than 24 mm. Plates were kept at refrigeration temperature for 30 min for better absorption. During this time microorganisms did not grow but absorption of extracts took place. Negative controls were the disks devoid of any antibiotics. The inoculated plates containing the antibiotic disks were incubated overnight in an upright position at 37°C. The results were expressed as susceptible, intermediate or resistant based on the zone of inhibition in comparison with CLSI standard (12).

Spore formers and Bt were detected from the samples following previously described method with some modifications (37). Briefly, about 100 µl of the water samples was aseptically inoculated into 9.9 ml of L-broth (tryptone 10 g/l, yeast extract 5 g/l, and NaCl 5 g/l) supplemented with 0.25 M Na-acetate (pH 6.8) for the detection of Bt whereas no supplement was used for the Sf and incubated in an orbital shaker at 30°C and 200 rpm (New Brunswick™ Excella^® E25, USA). After 4 h, 0.5 ml of suspension was transferred into a sterile test tube and heat treated for 10 min at 80°C in a water bath. Heat treated suspension was then diluted 10-fold and inoculated onto T₃-agar medium (1.0 L: tryptone 3.0 g, tryptose 2.0 g, yeast extract 1.5 g, MnCl₂ 0.005 g, phosphate buffer 50 mM, and agar 15.0 g; pH: 6.8) (38) by spread plate method and incubated at 30°C. In case of appearance of any colony after overnight incubation, incubation period was extended up to 72 h to allow sporulation.

Statistical analyses were conducted using software R version 3.4.1. The data analysis was performed by appropriate statistical tests where feasible. The Risk analysis of the alternative water sources was carried out by the test of homogeneity of proportions. Association among Vc and Sf as well as Bt was evaluated by Boschloo’s exact test with binomial model or multinomial model, Pearson’s chi-squared test, Fisher’s exact test, etc., where applicable. Boschloo’s exact test is an unconditional exact test for which the current version of the R package “Exact” (version 1.7) was used.

Following conventional method (colony characteristics on selective media) and rapid dipstick test, it was observed that 26% of the samples (i.e., 26 out of 100 samples) contained Vc isolates (Figure 2A). The distribution of Vc among different water supplies was estimated to be 13.8, 26, and 33% in waters derived from HH reservoirs, PSF, and pond surface, respectively (Figure 2B). Among the Vc identified, 13% were pathogenic (Vc O1) and 13% were non-pathogenic (non-O1/O139) (Figure 2C). None of the sample was identified with Vc O139 and this was further confirmed by serological test performed with specific antisera.

Vibrio cholerae positive samples were identified by TCBS and TTGA selective media following enrichment in APW broth and rapid dipstick test was performed with the Vc enriched APW broth to detect the presence of serotypes O1 and/or O139. Among all the samples plated on TCBS and TTGA media after enrichment in APW broth, characteristic growth of Vc was observed in both media with 22 samples whereas in 7 cases, growth of Vc was observed either in TTGA (3) or in TCBS (4). By the dipstick test and slide agglutination test with specific antisera, four out of seven were further confirmed as Vc. In developing countries, crystal VC dipstick has been the reliable test for detection of Vc these days and other rapid tests are also available for the LPS signal to be amplified in the APW broth before testing (39). The crystal VC dipstick was reported to possess much higher specificity than previous records (91–98%) and hence, this method was suggested as a promising screening tool for cholera outbreak surveillance in resource-limited settings (34).

Out of 100 water samples collected, 77 were unique since the samples collected from the PSF (23), in fact, represent the pond SWs undergone filtration process in PSF. Among the pathogenic Vc O1, 20.8% cases (10 out of 48 samples) were from pond SW, being the maximum, followed by 13% (3 out of 23 samples) from PSFs whereas no pathogenic Vc (out of 29 samples) was detected from HH water.

On the other hand, 87% of the water samples yielded Sf and Bt was isolated from 39% of the total samples. The distributions among the sources for Sf were 76, 91, and 92% and for Bt, were 38, 43, and 37.5% in HH water, PSF water, and pond SW, respectively (Figure 2D).

Cholera outbreaks occur seasonally in the coastal areas of Bangladesh such as Mathbaria, where it is predominant during the spring (from March through May), i.e., the dry season and the inhabitants are most susceptible since the scarcity of drinking water due to the high salinity in the shallow and deep tube-well water, deteriorate the situation (9, 40). The prevalence of Vc in season before the spring should be different and the study was, therefore, conducted during the beginning of winter, in the months of November–December to assess the prevalence of Vc which in other sense, represents the safety of the alternative drinking waters.

The significant finding here is the detection of pathogenic Vc O1 in 13.0% of the total samples in the month of November–December which is half of the total Vc positive samples (26). Between March and December 2004, in a study conducted by Alam and others, 99 water samples from Mathbaria yielded 2% Vc O1, 1% O139 Bengal, and 41.8% non-O1/O139 isolates where the maximum Vc O1 was isolated in the month of December (18), similar to our case. Again, pathogenic Vc O1/O139 was not detected from the coastal areas, especially subdistricts, Dacope and Mongla of districts, Khulna and Bagerhat, respectively, as reported in another study (1). In our study, although the overall prevalence seemed to be lower than others, a sign of more protected rain-fed ponds and better hygiene practice among the inhabitants, the detection of more than 20% Vc O1 (10 out of 48) in the rain-fed pond SW alone is very alarming. Since, the isolation was accomplished following enrichment in APW broth, and the direct isolation of Vc O1 even in cholera endemic areas was reported rare (18, 41), the isolates could be considered as viable but non-culturable. The variation in prevalence especially the increase of Vc O1 by a decade indicates the augmentative risk of cholera outbreaks persisting in this region. Whether they are toxigenic or not, should be determined by M-PCR (18) and further investigation for the virulence genes such as, ctxA and rfb, is required to estimate the risk level.

The ponds, in this study, were selected in a random manner for sample collection, and those were scattered inside and in the surroundings of Mathbaria municipal arena. The ponds in the surroundings were mostly connected to the canals and were also used in utensils washing as well as in bathing purposes. Hence, the tidal waters from the estuarine that intruded into the ponds via the canals, might have contributed to the transportation of Vc reservoir, i.e., the planktons (17). Again, as the study period was at the outset of winter, gradual reduction in rainfall and shift in temperature could have affected this mangrove forest neighborhood by contributing available nutrients to promote the growth of Vc O1. Thus, the difference in locations, variation in temperatures and artifactual contaminations might be the facts behind the variable prevalence of Vc. On the other hand, HH water, even though the prevalence of Vc was not very high, could be contaminated during rainwater harvesting or from manual collection system, dirty or blocked gutter and lack of initial washing or dirty storage tanks, etc. As the PSF water samples represent the rain-fed pond SW, the variation in prevalence is correlated to the efficiency of the PSF.

Prevalence of non-pathogenic Vc in the beginning of winter season was observed to be 12.5% (6 out of 48) and 13% (3 out of 23) in pond SW and PSF water, respectively (Figure 2F). Islam and coworkers reported that non-pathogenic Vc (non-O1/non-O139) from different water sources (ponds water, PSFs water, and harvested rain water) during dry and wet seasons were 47 and 100%, respectively, for PSFs whereas for pond SW, those were 95% in both seasons (1). This indicates that the rain-fed ponds of Mathbaria are better protected than those of Dacope and Mongla, and the inhabitants used them more hygienically. Again, since Vc non-O1/non-O139 was also reported to be clinically significant since ctxA, ace, or zot genes were present in few of the tested isolates (41), the PCR detection of virulence genes should be performed with all of the Vc positive samples for a better risk analysis.

The prevalence of Vc in the water samples from HH, PSF, and rain-fed pond surface was estimated and was compared. It was found that maximum 33% of the water samples from pond surface were identified with Vc followed by 26% in PSF and 13.8% in HH water. All Vc serotypes (4 out of 4) obtained from HH water were non-pathogenic, non-O1/non-O139 whereas 50% from PSF (3 Vc O1 out of 6) and maximum 62.5% (10 Vc O1 out of 16) from rain-fed pond SW yielded pathogenic Vc O1 (Figure 2E). The risk level of alternative water supplies as revealed in this study indicates that the pond SW was riskier with higher prevalence of both pathogenic and non-pathogenic Vc than that obtained from PSF and HH reservoirs.

The estimates of the proportions of Vc were 0.33, 0.14, and 0.26 in pond SW, HH water, and PSF, respectively. The test of homogeneity of proportions was conducted, and the null hypothesis was assumed that the proportions of Vc estimated from all sources of water tested, were the same, i.e., H₀: the proportion of Vc is the same in each sources, p₁ = p₂ = p₃. And the alternative hypothesis was, H₁: at least two proportions are unequal. Based on the samples, the rejection of the null hypothesis (χ22=3.5877, p-value=0.1663) was failed at 5% level of significance, i.e., the data provide sufficient evidence to conclude that there was no significant difference in the rates of Vc from one source to another source. Hence, it could be perceived that (i)

The chances of cholera incidences are almost similar if any of the sources is used in drinking purpose.

Vibrio cholerae, both pathogenic and non-pathogenic together, were considered for the test of homogeneity of proportions since the sample size of pathogenic Vc was not significant in all three cases, i.e., for HH water, it was nil. Contrarily, the loss of virulence genes in Vc O1 (17) and detection in Vc non-O1/non-O139 in certain cases (41) emphasized both types to be considered for the statistical analysis. Nevertheless, detection of virulence genes in the metagenomic pool could be useful in providing more insights to compare the alternative water supplies more appropriately.

Regarding the efficiency, it was observed that upon filtration through the PSF, Vc was not removed in any cases (n = 16) rather persisted in all cases and in 8.33% cases (n = 12), the contamination of Vc was introduced from the filter itself. This indicates that the filter itself can be the source of pathogens for further contamination and transmission which eventually cannot be considered as safe. The reports of inefficacy, published previously (3, 10, 12), are in agreement with the results obtained. The development of biofilm in the PSF, as also detected earlier in water and stool which served as the reservoir for long-term viability of Vc O1 (17), could be similar here for contaminating fresh waters. Chitin particles entrapped in the sand bed could be colonized by Vc and utilized with the self-secreted chitinase enzyme for nutrients (21), and biofilm formation thereby might have facilitated their existence. More studies in this connection are required to resolve the factors functioning in such case.

In case of Sf, the PSF was not able to clean up spores from any samples (n = 16), i.e., in 100% samples, filtration could not make any difference rather one of the samples was found to be contaminated in this process. Although upon filtration, Bt, another spore forming bacilli, was removed from 28.5% of the samples, 33.3% of the samples free of Bt, were contaminated with that very organism. The alteration of water quality thereby is also dangerous as among the Sf, there could be pathogens such as B. cereus and Clostridium perfringens responsible for gastroenteritis and food borne intoxication.

Bacillus-like Sf are usually soil dwelling bacteria and since they are ubiquitous, it is very likely that they can be obtained from water too. The presence of Bacillus thuringiensis was reported in many habitats of Bangladesh including southern region (basically coastal area), river basin and sandy beach with Bt indices (0.87, 0.82, and 0.73, respectively) (37). Sand is an important component of the PSFs, currently in use in Bangladesh, be it coarse or fine and needs scheduled regeneration by washing, chlorination, drying, etc. (42). Change in trophic level due to accumulation of water-borne biotic and abiotic as well as soluble and insoluble components in the sand bed, might facilitate the propagation of Bacillus-like Sf and discharge during filtration, a possible cause for contamination of filtered waters with spores.

On the other hand, the practice of regeneration of the PSF sand bed with the appearance of discernible change only, might be responsible for the unpredictable efficiency of PSF. Usage time relevant data and close monitoring for cholera pathogen would be useful in this connection to estimate the shelf life of PSF too. On the other hand, research for the development of simple and rapid regeneration method of sand bed to disrupt the presumptive biofilms would be more feasible, viable, and sustainable.

In antibacterial sensitivity test, it was found that all isolated Vc O1 were sensitive (100%) to azithromycin and ciprofloxacin whereas all were resistant to erythromycin and cotrimoxazole. In case of tetracycline, about 42% were resistant, 22.5% were sensitive, and the rest were variable (Figure 3). Antibiotic sensitivity pattern of the Vc was observed to be similar to the prevailing trend revealed in last few years of study period (40, 43).

In a recent study, it was observed that susceptibility of Vc toward azithromycin and ciprofloxacin was reduced, and 95% sensitivity to CIP was evidenced (40). Extensive and unregulated administration of antibiotic in health issues, veterinary medicines even in animal feeds in Bangladesh might be the source of such resistance. The complete resistant to erythromycin was developed because of its frequent use as a growth promoter in food animal production (44).

The influence of Sf on the survival of Vc was evaluated statistically based on the reciprocal prevalence as several reports of their antagonism against Vc are also available. The percentages of presence or absence of Sf and Bt in the samples detected with as well as free of Vc provided the impression that Sf and Bt might have an ecological influence over Vc (Figure 4A). The presumption that the survival of Vc (both pathogenic and non-pathogenic) might be influenced by the Sf as well as Bt was therefore evaluated. Hence, the presence and absence of both Sf and Vc as well as B. thuringiensis (Bt) and Vc were tabulated and analyzed statistically to find any association among them. The co-association tests were, therefore, performed separately and together with all water sample types.

Overall prevalence of Sf in samples detected with Vc, Vc O1, and Vc non-O1/non-O139 were 24, 13, and 11, respectively, out of 100 samples and it was 11, 3, and 8 for Bt. Again, in samples free of Vc, Vc O1, and Vc non-O1/non-O139, prevalence of Sf were 63, 74, and 76 whereas it was 28, 36, and 31 for Bt, respectively (Figure 4A).

Table 1 shows the overall and source based results of appropriate association tests between Vc (pathogenic and non-pathogenic) and Sf as well as Bt (45, 46). Overall test results indicated that no association between Sf and Vc (both pathogenic and non-pathogenic) could be found considering all the samples, and the similar outcome was observed in case of Bt over Vc and its types.

Whether the association is varied over the sample types or not was also assessed, and the statistical analyses were performed separately for all sample types.

In PSF, prevalence of Sf in samples detected with Vc, Vc O1, and Vc non-O1/non-O139 were 6, 3, and 3, respectively, out of 23 samples and it was 2, 0, and 2 for Bt. Again, in samples free of Vc, Vc O1, and Vc non-O1/non-O139, prevalence of Sf was 15, 18, and 18 whereas it was 9, 10, and 12 for Bt (Figure 4B). Prevalence was expressed in percent in the diagram in all similar cases if not otherwise stated. Table 1 shows the results of association test between various attributes obtained from the PSF water samples. Based on the data, it can be concluded that neither Sf nor Bt influence the distribution of Vc (both pathogenic and non-pathogenic) in the PSF water.

In HH reservoir, prevalence of Sf in samples detected with Vc, Vc O1, and Vc non-O1/non-O139 was 2, 0, and 2, respectively, out of 29 samples, and it was 2, 0, and 2 for Bt. Again, in samples free of Vc, Vc O1, and Vc non-O1/non-O139, prevalence of Sf were 20, 22, and 20 whereas it was 9, 11, and 9 for Bt (Figure 4C). Table 1 shows the results of association test between various attributes obtained from the HH water sample. The data provide sufficient evidence to conclude that neither Sf nor Bt influence the distribution of Vc (both pathogenic and non-pathogenic) in the HH water.

In pond SW, prevalence of Sf in samples detected with Vc, Vc O1, and Vc non-O1/non-O139 were 16, 10, and 6, respectively, out of 48 samples and it was 7, 3, and 4 for Bt. Again, in samples free of Vc, Vc O1, and Vc non-O1/non-O139, prevalence of Sf were 28, 34, and 38 whereas it was 11, 15, and 14 for Bt (Figure 4D). Table 1 shows the results of association test between various attributes obtained from the SW sample. Based on the data, it can be concluded that Sf or Bt do not influence the distribution of Vc (both pathogenic and non-pathogenic) in the pond SW.

In a 2 × 2 contingency table, if no expected frequency is less than 5, a Pearson’s chi-squared test was conducted and on the contrary, an exact test was done (47). In case of both margins being fixed of a 2 × 2 contingency table, Fisher’s exact test was conducted. If single margin was fixed, Boschloo’s exact test was used with a binomial model, and if no margin was fixed, Boschloo’s exact test with a multinomial model was performed (45). Fisher’s exact test was also performed in cases where all cells of a row or column are 0 (48).

The test results of appropriate association, although the impression of correlation could be perceived from graphical representation (Figure 4), revealed that neither the Bacillus-like Sf nor the Bt has any influence on the prevalence of Vc, be in the pond surface, HH reservoir or PSF derived water. The p-values in all cases were greater than 0.05 which nullified the chances of any co-association (Table 1).

Usually, Vc O1 and O139 remain preponderantly as viable but non-culturable cells in water and culturable cells in biofilm consortia either free-swimming or attached to plankton or other aggregates (17, 18). In this study, only water samples were considered, and no plankton samples were used in isolation of either Vc or Sf or Bt. Hence, the assessment of association was directed solely to the drinking water of alternative supplies. Although the binary data of presence or absence, analyzed statistically provided no clue of association between the bacteria of interests, estimation of CFU counts could provide a deep insight into any correlation prevailing among them or not. Contrarily, biofilm consortia could be the potential arena where Vc could be affected by Bacillus spp. due to complex microbial interaction as reported by Augustine and coworkers from their in vitro analysis (24). Although it was observed that the introduction or retention of Vc from the PSF was occurred in cases where Bt was neither present nor introduced and introduction or retention of Bt was occurred where Vc was absent, a larger sample size associated with the detection of aiiA gene in the metagenomes is essential to conclude any in situ inhibitory effect against biofilm formation. At the same time, prevalence of Vc and other Bacillus spp. from plankton and other biofilm harboring aggregates should be determined in this connection.