The majority of research on identifying the sources of faecal contamination in the South African water resources has been done using viral biomarkers [59,60,61] The two main groups of viral biomarkers investigated are coliphages infecting E. coli and closely related bacteria; and bacteriophages of Bacteroides fragilis HSP40 [62]. Somatic coliphages include members of families Myoviridae, Siphoviridae, Podoviridae and Microviridae [62] that target bacteria through an interaction with specific receptor sites on their cell walls [59]. They can contain DNA or RNA as nucleic acids [50]. Somatic coliphages have been shown to attach to dead bacteria [62].

Male-specific coliphages infect their host through the interaction with the fertility (sex) fimbriae (F-fimbriae) of E. coli and related host bacteria [62]. If the viral receptor is located at the shaft of the F-fimbriae, then the virus generally contains RNA as the nucleic acid and such male-specific coliphages are called F-RNA coliphages [59]. If the viral receptor is located at the tip of the F-fimbriae, then the virus generally contains DNA as the nucleic acid and such male-specific coliphages are called F-DNA coliphages [50]. Both types of bacteriophages are shed in faeces of humans, thus are candidates for indicator microorganisms of faecal contamination of water resources [62]. However, only F-RNA coliphages have been used to track sources of faecal contamination in South Africa.

In the case of E. coli and related bacteria, the F-fimbriae are only produced when the temperature of the environment reaches 32 to 40 °C, which is rare in South African water resources [59]. Therefore no replication of F-RNA coliphages is expected in the water environments after a faecal contamination event and therefore this group of viruses fulfils one of the key properties of the indicator organism of faecal pollution (see Section 4 for details). Based on serological typing and genotyping, F-RNA coliphages can be separated into 4 distinct serotypes/genotypes namely I, II, III, and IV [62]. Strains of Bacteroides spp. are anaerobic and Gram-negative bacteria, which do not produce spores and its cells die-off rapidly upon exposure to O₂ [62]. Therefore viruses that infect this bacterium constitute the second group of viral biomarkers that have been studied extensively in literature for faecal source tracking. These bacteriophages are resistant to unfavourable environmental conditions [63,64] and have been examined as selective indicators of human sources of faecal contamination [65], as they have not been found in animal faeces [62].

Grabow et al. [59] investigated the applicability of F-RNA coliphages, somatic phages and the phages infecting Bacteroides fragilis HSP40 as a tool for the distinguishing between human and animal sources of faecal contamination in water resources in South Africa. The authors also measured the survival rates of these bacteriophages in the environment and compared the environmental fate of the bacteriophages to the environmental fate of indicator microorganisms and human viruses [59]. F-RNA coliphages were enumerated by the double-agar plaque assay [66], with Salmonella typhimurium WG49, and E. coli HS(pFamp)R as bacterial hosts. In 34 samples from the Klip River in the Gauteng Province, the Salmonella typhimurium WG49 host gave consistently higher counts of F-RNA coliphages than the E. coli HS(pFamp)R strain. Experimental problems of viral enumeration included the effectiveness of the recovery of viral particles and cost of laboratory consumables [50]. The presence/absence test for bacteriophages had been recommended for routine testing of water samples [19], but Grabow et al. [59] found that these were highly prone to contamination, making its applicability in routine monitoring questionable.

A comparative study was done on the application of F-RNA coliphages as biomarkers for faecal contamination from different sources in South Africa and Spain [67]. In South Africa, the samples included stool samples and wastewater samples from the Pretoria Academic Hospital (human sources of faecal contamination), samples of faeces of pigs, poultry and cows from several rural farms around the Gauteng province (animal sources of faecal contamination), settled sewage from the Zeekoegat and Baviaanspoort wastewater treatment plants in the vicinity of Pretoria (sources of faecal contamination from human dwellings but possibly from dogs, poultry, and maybe cattle), wastewater from developing communities in Atteridgeville, Soshanguwe and Botshabello (human sources of faecal contamination due to seepage from pit latrines and septic tanks), raw wastewaters from the abattoirs (animal faecal pollution sources from cattle and sheep); and secondary settled effluent from the wastewater treatment plant in Daspoort (domestic sewage of with possible mixed faecal inputs).

In South Africa, 10% of human faeces samples contained F-RNA coliphages while 70% of the faecal poultry samples contained F-RNA coliphages [67]. For the human samples, the maximum value reported in South Africa is 15% [68]. Sixty six percent of the pig faeces samples were positive for the presence of F-RNA coliphages, but only 33% of cattle faeces samples contained F-RNA coliphages [67]. The F-RNA coliphage concentrations in faeces samples ranged from 6.5 × 10² plaque-forming units (PFUs)/100 g faeces in cattle samples to 1.0 × 10⁵ PFUs/100 g faeces in poultry specimens. For wastewater samples, viral concentrations ranged from 2.7 × 10³ PFUs/100 cm³ in the final effluent from wastewater treatment plants to 3.8 × 10⁶ PFUs/100 cm³ in abattoir wastewaters. F-RNA specific coliphages accounted for 62.6 to 98.7% of the plagues detected on Salmonella typhimurium WG49 plates. Out of the human faeces that were positive for F-RNA coliphages, then 90% contained serotype/genotype II and 10% contained serotype/genotype III F-RNA coliphages. Results for the presence of serotypes/genotypes in faeces from poultry confirmed previous reports that 80% of samples positive for F-RNA coliphages serotype/genotype I, while F-RNA coliphages of serotype/genotype IV were found in 20% of the positive samples. One hundred percent of F-RNA coliphages present in the cattle faeces belonged to serotype/genotype I.

The downside to routine application of this technique originates from major overlaps between the F-RNA coliphages serotypes/genotypes found in human faeces and animal faeces, as this would lead to a high rate of the faecal source misidentification. For the positive samples of pig faeces, serotype/genotype I accounted for 53% of the F-RNA coliphages present, serotype/genotype II for 28% and serotype/genotype IV for the remaining 19% of all coliphages present [67]. Two percent of the coliphage concentrations in abattoir wastewaters belonged to serotype/genotype II, while 6% of the virions from hospital wastewater were classified as serotype/genotype I [67]. The concentrations of individual serotypes/genotypes in wastewaters from developing communities and wastewater treatment plants showed substantial variations and the presence of all serotypes/genotypes [67].

In their study, Sundram et al. [60] studied the application of F-RNA coliphages to track faecal contamination sources in the Umgeni catchment in the province of KwaZulu-Natal, South Africa. The wastewaters studied included those from a cattle feedlot, a piggery and from two hospitals in Pietermaritzburg [60]. Methodologies of Grabow et al. [59] and Hsu et al. [69] were followed in this study. The authors detected serotypes III and IV in the pig slaughter house wastewaters, and serotypes I, III and IV in chicken wastewaters [60]. Serotypes II and III have been isolated from pig specimens probably due to animals’ contact with human faecal material during feeding [69].

Momba et al. [61] studied the concentrations of free residual chlorine, somatic and F-RNA coliphages in the final effluent from wastewater treatment plants in Alice, Fort Beaufort, Dimbaza and East London in the Eastern Cape province of South Africa. The average concentrations of somatic coliphages ranged from 3.88 × 10⁵ to 2.45 × 10⁶ PFUs/100 cm³, while the average concentrations of F-RNA coliphages ranged from 2.43 × 10³ to 1.82 × 10⁵ PFUs/100 cm³ [61]. Just as in previous studies on the subject, Momba et al. [61] also found an overlap of the serotypes/genotypes of F-RNA coliphages between human and animal sources of faecal contamination. Based on the above discussion, the F-RNA coliphages do not provide a clear distinction between human and animal sources of faecal contamination. At the same time, these bacteriophages are not present in the faeces of the entire population of South Africa, i.e., the lack of their detection in a water samples does not necessarily mean lack of faecal contamination. Therefore the application of these viral biomarkers should be discouraged for faecal-source tracking in the NMMP samples.

Regarding routine use of the bacteriophages infecting Bacteroides fragilis HSP40, several problems have been reported in the literature. Firstly, maintenance of a viable inoculum of Bacteroides fragilis HSP40 was shown to be difficult in the accredited laboratories and the MPN counting of the phage PFUs provides results with high variability [59]. The inoculum problem could be solved using the triple agar technique with oxyrase addition [70]. Oxyrase is a mixture of mono- and dioxygenases isolated from E. coli [71] and it allows for anaerobic cultivations to be conducted under aerobic conditions. However, the enzyme was only available from a USA-based supplier at the time of writing of this article, i.e., the routine application of the Oxyrase modification is unlikely to become feasible any time soon. A PCR method developed for the detection of bacteriophages of Bacteroides fragilis HSP40 is more sensitive than the plaque assay [62]. A threshold concentration of the cells of Bacteroides fragilis HSP40 is required for the bacteriophage replication to take place and it is equal to 10⁴ CFUs/cm³ [72]. The B40-8 type-specific phage for Bacteroides fragilis HSP40 has been shown to replicate at 22 or 30 °C, but no replication has been reported in freshwater or sediment samples under aerobic or anaerobic conditions [62].

Excretion rates of the bacteriophages of Bacteroides fragilis HSP40 in human faeces are not available in South Africa, but the international literature suggests rates between 5% [73] and 10% [63] of the population. Therefore a failure to detect bacteriophages infecting Bacteroides fragilis HSP40 in a water sample does not seem to exclude the presence of faecal material/contamination, in a fashion similar to coliphages [59]. Plaque assays require the use of complex growth media with antibiotics and strict anaerobic conditions must be maintained during the entire incubation for their enumeration [62]. This requires specialised training and highly experienced laboratory staff to perform the enumerations. A severe skills shortage will preclude the widespread application of viral indicators in microbial water quality monitoring in South Africa [74].

Some papers in the international literature proposed the substitution of bacteriophages infecting Bacteroides fragilis strain RYC2056 for bacteriophages of Bacteroides fragilis HSP40 for the detection of human faecal pollution, as the latter is not found in human faeces in certain geographical areas [71,75]. However, the data on the actual application of this viral indicator in faecal-source tracking are scarce. Based on the information presented above, usability of bacteriophages of Bacteroides fragilis strains will be limited based on the limited excretion of the viruses in faeces of humans and the fact that there are currently not enough specialised laboratories in South Africa to perform this type of analysis. Therefore these viral biomarkers are not suitable for the regular faecal-source tracking in the context of the NMMP.

The NMMP is based on the enumeration of E. coli, but its concentration does not provide any information about the sources of faecal contamination. Outside the priority areas and in catchment parts that have not been considered in the prioritisation process, source identification might be virtually impossible. This disadvantage can be overcome using the microbial-source tracking methods using various properties of E. coli strains [76]. In the library dependent methods, isolates of E. coli are obtained from different sources of faecal contamination and selected properties such as the carbohydrate fermentation spectrum or the antibiotic resistance spectrum are measured [77].

Databases of the E. coli isolates’ properties should be created and maintained in every WMA monitored in the NMMP [76]. For a particular water sample, the public health risk can then be assessed based on the concentration of E. coli and the faecal contamination source can be identified using the relevant database of isolate properties and statistical analyses using e.g., the jack-knife analysis [76]. The scientific and regulatory communities in South Africa are more likely to accept the antibiotic resistance analysis of E. coli from faecally contaminated water resources as this has been studied in South Africa [78,79,80]. Obi et al. [81] reported a high degree of correlation between the antibiotic resistance spectra of E. coli isolates from the stool samples of HIV-positive and HIV-negative patients; and the drinking water available in their households. Therefore faecal-source tracking using the antibiotic spectra of E. coli could be integrated into the NMMPs and used for identification of the sources of faecal contamination after the appropriate calibration has been completed.

Based on data from international literature, the enumeration of various subgroups of Bifidobacterium spp. is a low-cost and relatively easy-to-use faecal-source tracking tool [82]. Bifidobacteria are anaerobic and Gram-positive bacteria, which are abundant among the intestinal microbiota of humans and other warm-blooded animals [83,84]. They are excreted in faeces in concentrations as high as 10¹⁰ cells/g in faeces [85] and they do not multiply in surface water resources unless the water temperature exceeds 30 °C [86]. They could be useful in South Africa where ambient water temperatures rarely exceed 30 °C [87]. Next, a literature overview of the principles of bifidobacterial faecal-source tracking is provided and suggestions for the media to be used in the NMMP are made.

The YN-6 medium was the first medium specifically designed for the enumeration of bifidobacteria in water samples and it was designed by Resnick and Levin [88,89]. The authors studied the distribution of bifidobacterial strains in faeces of healthy human adults and agriculturally significant animals (chickens, dogs, cows, pigs, horses, cats, sheep, beavers, goats, and turkeys). Samples of the following types of wastewater were also analysed: raw, primary, secondary and chlorinated sewage from two wastewater treatment plants and in environmental water samples [88,89]. Later studies reported problems with the inhibition of the growth of certain bifidobacteria on the YN-6 medium and overgrowth by faecal streptococci [90]. Modifications of the YN-6 medium were undertaken by several authors. This led to development of the YN-17 medium for the enumeration of total bifidobacteria and design of the human bifid sorbitol agar for measurement of the sorbitol-fermenting bifidobacterial concentrations [84,90].

Bonjoch et al. [84] used these media to investigate the applicability of the ratio of the sorbitol-fermenting bifidobacteria and the total bifidobacteria (designated the tracking ratio in further text) as a criterion to distinguish between the human and animal sources of faecal pollution.

The concentrations of the sorbitol-fermenting bifidobacteria were always higher than 10⁶ CFUs/100 cm³ in the human-faecally-polluted wastewaters, while the animal wastewaters always contained concentrations lower than this number [84]. The concentrations of the total bifidobacteria ranged from 10⁶ to 10⁷ CFUs/100 cm³, but there was no systematic difference between the values in human and animal wastewaters. The tracking ratio was always higher than 0.20 for the wastewaters with human faecal pollution, and always lower 0.05 for the wastewaters with faecal pollution of animal origin [84]. Therefore the authors proposed that 0.20 be used as the cut-off value for distinguishing between human and animal sources of faecal contamination [84]. In a given water sample, the concentration of the sorbitol-utilising bifidobacteria and the total bifidobacteria is determined and then the tracking ratio is calculated. This value is subsequently used to establish the course of faecal contamination of surface water.

In the study by Blanch et al. [82], the cut-off point for the tracking ratio was 3.2. Based on this value, the authors were able to distinguish human and animal faecal pollution sources with a 95% certainty. Samples were taken across Europe and covering a larger geographical area and wider range of climatic conditions than in the study of Bonjoch et al. [84]. The cut-off value for the tracking ratio indicated that the concentration of the measured sorbitol-utilising bifidobacteria was higher than the total bifidobacterial concentration. Local differences in the survival/die-off kinetics of the respective group of bifidobacteria can lead to this observation [91,92]. At the same time, environmental stress is likely to cause greater inhibition of growth by bifidobacteria on the YN-17 medium, which contains higher concentrations of antibiotics than the human bifid sorbitol agar [90]. Therefore the cut-off value of the bifidobacterial tracking ratio will have to be calibrated for the conditions in South Africa, if this method of faecal-source tracking is to be incorporated into the NMMP battery of tests.

Jagals and Grabow [93] measured the concentrations of the sorbitol-fermenting bifidobacteria, faecal coliforms and faecal streptococci in the vicinity of an informal settlement on the banks of the Modder River in the Free State Province, South Africa. Pit and bucket latrines and limited waterborne sewagery constituted the sanitation infrastructure [93]. A non-perennial river flowed through the informal settlement and joined with a perennial river downstream of the settlement. Water samples covered the entire area and collection took place at five sampling sites during the dry and rainy seasons; and immediately after a thunderstorm. Concentrations of the sorbitol-utilising bifidobacteria would characterise human sources of faecal contamination. During the dry season, the concentration of the sorbitol-fermenting bifidobacteria ranged from 1 to 3,600 CFUs/100 cm³ in the non-perennial stream just downstream of the settlement. The concentrations decreased to 1 to 470 CFUs/100 cm³ further downstream of the settlement at the junction between the two streams. Concentrations of bifidobacteria were below 0 CFUs/100 cm³ upstream of the informal settlement, i.e., no environmental sources of bifidobacteria were observed. Samples taken during the rainy season contained the concentrations of the sorbitol-fermenting bifidobacteria concentrations inside the interval from 60 to 92,000 CFUs/100 cm³ in the non-perennial stream and just downstream of the informal settlement. At the junction of the two rivers, levels of this indicator microorganism increased to between 130 and 21,700 CFUs/100 cm³. The concentrations dropped to between 3 and 273 CFUs/100 cm³ further downstream of the informal settlement along the perennial river.

This indicates that the concentration of the sorbitol-utilising bifidobacteria correctly indicates how human sources of faecal pollution affect microbial water quality in surface water resources. Similar results were obtained later by Jagals [94]. To the best of the authors’ knowledge, the studies by Jagals and Grabow and Jagals are the only two to date in South African literature on the application of bifidobacteria as an indicator of faecal contamination. This limited data does, however, point to the correct identification of human sources of faecal contamination based on bifidobacteria. Implementation of the anaerobic bifidobacteria into the NMMP battery of tests will be feasible in South Africa as all the consumables are commercially available from companies such as Merck and Oxoid. At the same time, the costs are comparable to the currently used standard methods of microbial water quality monitoring. The cut-off value of the tracking ratio will have to be determined for local conditions in South Africa, along with the necessary survival rates. Since 48 h is required to obtain the value of the ratio of the sorbitol-fermenting bifidobacteria and the total bifidobacteria, the time frame is comparable to the current indicator microorganism tests.

Many of the surface water resources are located in remote and rural parts of South Africa [13], thus adherence to the 6 h holding time for E. coli and other indicator microbial indicators [19,95] is likely not to be feasible there. This might be one of the reasons that only 70% of the total territory of South Africa was covered by the NMMP in 2009 [56]. Therefore a surrogate test(s) is needed to perform regular microbial water quality monitoring in remote areas. Extensive literature search was conducted to find substitute method. The qualitative and modified H₂S strip test is the ideal test for this purpose [13]. The original H₂S strip test has been shown to be in agreement with the Colilert™ system results and the concentrations of E. coli in Indonesia [96] and showed high correlation with sources of drinking water contamination in South Africa [13]. In this test, a water sample is mixed with a dehydrated cultivation medium which are released upon sealing of the cultivation bottle and hand-shaking [18,97]. The medium contains Na₂S₂O₃ and bacteria of faecal origin reduce it to H₂S, which subsequently reacts with ferric iron forming a black precipitate [98] Sodium thiosulphate also neutralises the residual chlorine and the H₂S strip test is therefore suitable for testing the drinking water sources normally found in South Africa. A positive reaction, i.e., detection of faecal contamination of the water body, is based on a visual detection of a colour change from the original brown colour into a black colour of the ferric sulphide precipitate. Incubations are performed away from direct sunlight and at room temperature. There is no need for an incubator and a sample is considered faecally contaminated if black colour develops within 72 h [18,99]. The incubation period is longer than found with other indicator microorganisms. However, the test can be performed with low costs and after limited training (see below). Thus testing of the microbial surface water quality using the H₂S strip test will provide public health information about water consumption in areas where this information would otherwise be lacking.

Test kits (Pathoscreen) can be obtained from the Hach Company [98] or prepared by the municipal officials, e.g., environmental health officers. This is feasible after a one-day training workshop which can be performed by academic personnel or any other qualified individual. First, the following items are purchased: the H₂S media components, sterile urine jars and automatic pipettes/tips. The kits are then assembled as demonstrated during the training workshop [13]. The sample collection and transport can subsequently be performed on site with minimum processing and low labour intensity [37,99]. The cost per sample with the H₂S strip test is around 10.00 ZAR or less (approximately 1.32 USD or less; this amount includes transport, labour and all other costs). As a result, the test is easily affordable for rural municipalities, even under conditions of financial distress.

A group of the H₂S strip test bacteria include all bacterial strains of faecal origin which produce a positive signal in the qualitative H₂S strip test [18]. Such strains belong to the species of Citrobacter spp. [100], E. coli [99], Proteus mirabilis [101], Proteus vulgaris [102], Aeromonas spp., Clostridium spp. and Salmonella spp. [99]. Thus E. coli will be detected in the H₂S strip test, i.e., the public health prediction with respect to surface water will be based on similar scientific basis as the Colilert^®18 and the NMMP. Proteus vulgaris and Proteus mirabilis have been reported to cause diarrhoeal diseases in infants [103]. Clostridium perfringens has been detected in the patients’ faeces who suffered from food poisoning from undercooked meat or gravy [104]. Strains of Salmonella spp. have been reported to cause diarrhoeal infections in humans [105]. Most strains of the species Citrobacter spp. are commonly found in the intestinal tract of humans [106], and certain strains have been isolated from the intestinal tract of dairy cows [107]. Literature data therefore indicates that a positive H₂S strip test for a given water sample will clearly indicate faecal contamination of that water resource.

The coliform group of bacteria has been reported to contain the following species: Citrobacter spp., Enterobacter spp., Escherichia spp., Hafnia spp., Klebsiella spp., Serratia spp. and Yersinia spp. [95]. Therefore the acceptance of the H₂S strip test by the regulatory community should not pose a problem to its implementation in the NMMP context as there is a partial species overlap between the standard indicator microorganisms, i.e., coliform bacteria, and the H₂S strip test bacteria [97]. Environmental isolates of the H₂S strip test bacteria have been reported from soil, water, tree bark and logging operations [106], but these accounted for only 12% of the total concentrations in surface water samples [108]. Thus the influence of environmental sources on the detection of faecal contamination using the H₂S strip test will be negligible and likely similar to the effects observed with respect to the E. coli concentration in the current NMMP framework.

In the South African literature, the most used medium for the H₂S strip test is the one of Venkobachar et al. [37]. Genthe and Franck [37] examined and reported the rate of correspondence between the positive H₂S strip test based on that medium and the detection of non-zero concentrations of faecal coliforms in environmental water samples. The water resources examined included surface streams, several drinking water sources, fire hydrant water, and borehole water [37]. The overall rate of correspondence of positive results between the H₂S strip test and faecal coliforms ranged from 86 to 94% [37]. However, the H₂S strip test lacked the ability to provide a positive signal when the concentrations of FC were below 10 cells/100 mL [37]. This drawback can be eliminated by the addition of 0.5% taurocholate or deoxycholate into the H₂S strip test medium [99], which increases the probability of successful detection of faecal contamination to 64% if the E. coli concentration ranges from 0–10 cells/100 cm³ [109]. The specificity of the H₂S strip test could be further improved by optimisation of the bile acid concentrations in the dehydrated form of the cultivation medium. If the activity is successful the H₂S strip tests will have comparable sensitivity and reliability to the standard methods for assessing faecal contamination of surface water in South Africa. A mechanism will have to be put in place to guarantee validity of the microbial water testing based on the H₂S strip test in the context of the NMMP. Therefore a certification mechanism must be implemented for the H₂S strip test based monitoring, similar in scope and format to the South African Scoring System certification [110]. Recent data from the Philippines and the USA compared reliability of the Colilert^TM system and the H₂S strip test based that the medium of Venkobachar et al. [37,111]. For the WHO-defined improved and unimproved sources of drinking water, the Colilert^TM system correctly predicted faecal contamination in 75% of 521 samples taken from drinking water sources similar to the ones used in South Africa [111].

At the same time, the Colilert^TM system showed a six-percent rate of false positive results, while 19% of samples were false-negative for faecal contamination. On the other hand, the H₂S strip test performed best if the manually-prepared sampling kits were used and the sample size was equal to 20 mL [18,111]. In that case, 84% of 521 samples were correctly identified as faecally-contaminated in the H₂S strip test, while the rate of false positive results was equal to 10% [111]. At the same time, only 6% of samples were incorrectly classified as faecally-contaminated and the H₂S strip test increased the error of the faecal contamination detection to a lower extent than the Colilert^TM system [111]. These data rates will be further improved based on the addition of 0.5% of deoxycholate [18,109] and will have to be verified under South African conditions. They do, however, point to the comparable reliability of the H₂S strip test and the current NMMP testing framework. This in turn indicates that H₂S strip test will form a valuable addition to that framework and will allow for widening of its geographical scope.