Clostridium perfringens identifies source of pollution and reference streams in a tropical highland environment

Water quality evaluation using Clostridium perfringens is performed around the globe (Bisson & Cabelli 1980; Fujioka & Shizumura 1985; Sartory 1988; Sorensen et al. 1989; Byamukama et al. 2005; Goto & Yan 2011) with the aim of understanding the microbiological safety associated with surface and drinking water systems. The occurrence of C. perfringens in the gastrointestinal tract of both humans and other animals (Rood & Cole 1991; Vierheilig et al. 2013), their correlation with human enteric pathogens in environmental waters (Ferguson et al. 1996) and their ability to predict gastrointestinal diseases (Kueh et al. 1995; Viau et al. 2011) support their use as a water safety determinant. While a significant number of previous studies shows a strong link between C. perfringens and anthropogenic influences (e.g., Skanavis & Yanko 2001; Byamukama et al. 2005; Farnleitner et al. 2010), little is known regarding the indication value of C. perfringens in water systems free from anthropogenic influences. Furthermore, water systems contaminated with C. perfringens in the absence of any known anthropogenic disturbance may not correlate with human enteric pathogens and/or gastrointestinal illness risks, limiting attempts to design and initiate sustainable water quality management.

On the other hand, understanding the source of C. perfringens in anthropogenically free water systems is critical because the management of water systems contaminated with C. perfringens depends on the known source(s) of C. perfringens. For example, surface water contaminated with C. perfringens from wildlife needs completely different management strategies compared to that from sewage. Consequently, there is a critical need to comprehensively investigate the source of C. perfringens in anthropogenically free water systems in order to improve microbiological quality and the overall ecological health of surface waters.

As in other highlands around the globe, Uluguru Mountain in Tanzania, which is representative of larger areas of the East African Arch Mountains (Kimaro et al. 2008), is the main source of drinking water for Morogoro municipality and its outskirts (Jiwa et al. 1991) due to the presence of headwater streams. Although these headwater streams are free from anthropogenic influences, they are contaminated by catchment soil as a result of the unstable slopes of Uluguru Mountain (Westerberg & Christiansson 1999). Given the critical health impacts associated with stream water contaminated with soil (Baumgardner 2012), understanding the magnitude of this environmental problem is critical so that appropriate plans for precluding the bacterial contamination of stream water are put in place.

Therefore, this paper has: (1) determined the contribution of soil from the steep slope of Uluguru Mountain to the microbiological quality of stream water during the baseflow condition; (2) conveyed the way C. perfringens may support the understanding of the link between microbiological quality of the stream water and the catchment soil; and (3) determined the feasibility of using these headwater streams as reference (i.e., negative control) in order to establish long-term water quality monitoring programmes and to assist management decisions. The hypothesis was that soil from the tropical steep slope highland is the dominant source of contamination in stream water resources at the baseflow condition and that C. perfringens is able to clearly demonstrate the existence of this influence and provide the information about the feasibility of using tropical highland streams as a reference. Given that catchment soil is slowly transported through soil erosion and that its impact on headwater streams is cumulative, evaluation of this kind of influence requires an appropriate bioparametric variable reflecting this situation. Consequently, the resistant nature (Davies et al. 1995; Vierheilig et al. 2013) and accumulation potential (Byamukama et al. 2005) of C. perfringens were the basis for its use as a suitable bioparametric variable for testing the present hypothesis.

The study was conducted on the northern slope of the Uluguru Mountain (coordinates 6.8°S, 37.6°E) in Tanzania with an anticipated area of approximately 137 km² at an altitude greater than 1,450 m above sea level. This area is characterised by mean annual rainfall ranging from 900 to 2,300 mm and air temperature ranging from 19 °C to 25 °C (Kimaro et al. 2008). The soil texture ranges from sandy-clay to sandy-clay-loam (Munishi et al. 2007). The area is composed of headwater streams that drain the steep slope of the mountain environment. These headwater streams are characterised by two pronounced discharge peaks per year as a result of long (March to May) and short (November to December) rainfall events. Generally, all these headwater streams and the associated environments have not been microbiologically investigated for their suitability as source of drinking water for Morogoro's urban population in Tanzania. Extensive survey of these headwater streams revealed that anthropogenic influences were non-existent and the surveyed headwater streams satisfied 20 presumptive criteria of reference streams (see Sánchez-Montoya et al. 2012). Although the surveyed environment was covered with vegetation (mainly shrubs and grasses), soil erosion and landslides were highly pronounced as a result of the unstable nature of the mountain slope. Additionally, an in-field visual counting survey revealed the presence of various types of faecal droplets (>5 faecal droplets/km²) on the soil, signifying the potential influence from wildlife.

To understand the effect of the catchment soil on the microbiological quality of headwater streams, four independent plots (each with the size 20 m × 20 m) were randomly selected in the catchment with observable soil erosion and/or landslide. In order to obtain representative soil samples, each of the selected plots was further divided into 10 m × 10 m subplots resulting in four sub-plots per plot. Five grams of the first 10 cm top soil was sampled aseptically from each sub-plot of the selected plot to form a composite soil of 20 g per plot. Soil samples were immediately placed in sterile plastic bags before being carried to the laboratory and stored at 4 °C in the dark until analysis. Water samples from the streams corresponding to the selected plots was aseptically obtained under baseflow condition using sterile 500 mL glass bottles. Each sample matrix from the selected sites was separately sampled five times (n = 5) over the sampling period resulting in a total of 20 samples per sample matrix. Additionally, water samples from two independent river systems, one located in an agricultural environment (6.74° S, 37.77° E) and the other in an urban environment (6.8° S, 39.26° E) were sampled 25 times (n = 25) during the baseflow condition in order to evaluate the suitability of highland streams to serve as a negative control during the evaluation of tropical river systems located in anthropogenically influenced environments. All the water samples were taken to the laboratory and stored at 4 °C in the dark until analysis.

In the laboratory, soil samples were homogenised as described in Byamukama et al. (2005) before extraction of C. perfringens using a method that has proven to be efficient in recovering bacterial cells from soil matrix (Kingsley & Bohlool 1981). After extraction, the supernatant solution (10–100 mL) was analysed for C. perfringens concurrently with stream water (100–200 mL) samples using the standard technique as proposed in ISO 14189:2013 (ISO 2013). The membrane filtration method employing 47 mm diameter Whatman cellulose nitrate filters (Sartorius, Gottingen, Germany) with porosity of 0.45 μm was used for the enumeration of C. perfringens from soil extractants and stream water samples. The cellulose membrane filters that retained bacteria with greater than 0.45 μm size were placed on tryptose sulphite cycloserine (TSC) agar (Merck, Darmstadt, Germany) containing 0.4 g/L of C. perfringens supplement that was prepared according to the manufacturer's directions. It should be noted that TSC agar method was applied due to its higher specificity when compared to other available methods for detecting C. perfringens (Araujo et al. 2001). Additionally, the C. perfringens supplement test for acid phosphatase is a highly specific indicator for C. perfringens (Ryzinska-Paier et al. 2011). Triplicate TSC agar plates per sample were prepared before being incubated at 44 °C for 24 hours while in Anaerocult A anaerobic system containing anaerobic indicator (Merck). After incubation, presumptive black colonies surrounded by black precipitates that appeared light blue upon exposure to UV light were considered as C. perfringens. Enumerations were performed on Petri dishes (60 mm) containing between 20 and 200 positive colonies of C. perfringens. The counted colonies were expressed as colony forming unit per cubic centimetre (cfu/cm³) assuming that 1 mL ≈ 1 g ≈ 1 cm³.

Comparison of C. perfringens concentrations between sample matrices was performed using non-parametric Mann–Whitney U-test. Spearman rho test was used to measure the degree and statistical significance of the association between soil and water C. perfringens concentrations. The extent and significance differences among reference stream, agricultural and urban rivers, with respect to C. perfringens were determined using non-parametric Kruskal–Wallis test. Mean abundances of C. perfringens in soil and water samples were separately determined by calculating geometric means of abundances. Streams with a geometric mean below the recommended safe permissible value for tropical environment (geometric mean of 50 cfu/cm³) were classified as reference streams. All statistics were executed on the Statistical Package for the Social Sciences version 16.0 (SPSS Inc., IL, USA) and probability (P) value below 0.05 was deemed significant.

C. perfringens was frequently detected in the catchment soil (100%) and in the stream water (100%) (Figure 1) even though the sampling sites were free from anthropogenic influences and in different locations within the investigated mountain (>1 km from one site to another). The constant detection of C. perfringens in soil samples confirmed that this bacterium accumulates in the environment and thus can serve as a marker for persistent faecal contamination (Farnleitner et al. 2010; Vierheilig et al. 2013). While the concentrations of C. perfringens in the soil matrix were within the range reported in the literature for tropical environments (Byamukama et al. 2005; Goto & Yan 2011), the water matrix showed significantly lower C. perfringens concentrations than previously reported in tropical river systems (Mushi et al. 2010). Unlike the headwater streams in this study, these tropical river systems were located in an urban environment characterised by anthropogenic influences, where higher concentrations are expected due to the strong link between C. perfringens and anthropogenic influences described in previous studies (e.g., Skanavis & Yanko 2001; Byamukama et al. 2005; Farnleitner et al. 2010). C. perfringens from soil had a geometric mean of 307 cfu/cm³ (ranging from 160 to 690 cfu/cm³) as compared to the geometric mean of 33 cfu/cm³ (ranging from 10 to 140 cfu/cm³) detected in water samples (Figure 1) suggesting that soil entrance into stream water through erosion resulted in 1 log reduction of C. perfringens. The low concentrations of C. perfringens detected in the stream water suggest that the catchment soil was the source of contamination in stream water rather than anthropogenic activities. This was supported by the fact that 82% of the variations of C. perfringens concentrations in the water were explained by C. perfringens concentrations from the catchment soil (Figure 2). Supposedly, wildlife present in the studied environment deposits C. perfringens to the catchment soil through faecal droplets, which are subsequently brought to the stream water by soil erosion and/or landslide enhanced by the poor stability of the studied steep slope catchment (Westerberg & Christiansson 1999). It should be noted that autochthonous C. perfringens are unlikely to occur in the investigated environment due to the fact that these bacteria require special growth conditions (Shimizu et al. 2002), which are not offered by environmental matrices such as water and soil.

The C. perfringens concentration in stream water was below the recommended concentration for safe water in tropical environments (Fujioka & Shizumura 1985). This concentration is associated with a negligible risk of gastroenteric illnesses (Kueh et al. 1995), suggesting an acceptable level of disturbance in the studied streams. Because the stream water samples were collected in an area free from anthropogenic influences during baseflow condition (dry season), the quantified geometric mean of C. perfringens concentration is suggested to denote the baseline concentration for tropical stream waters. This outcome is consistent with previous studies (Fujioka & Shizumura 1985; Byamukama et al. 2005), which reported that C. perfringens exists in the environment at low concentrations, but tends to increase to significant levels, when there is anthropogenic influence. Consequently, the value detected in this study is suggested as reference concentration for C. perfringens in tropical environments and the use of the investigated streams – from which this reference concentration was quantified – as reference streams is proposed.

Unlike the faecal indicator bacteria Escherichia coli (Verhougstraete et al. 2015) and Enterococcus sp. (Tiefenthaler et al. 2009), which were used to define acceptable levels of disturbance and reference concentrations in freshwater systems of moderate climate, C. perfringens is a more precise indicator of fecal contamination in tropical environments (Fujioka & Shizumura 1985; Byamukama et al. 2005). In contrast to E. coli and Enterococcus sp., C. perfringens has not been shown to be able to proliferate in the environment (Davies et al. 1995; Ishii & Sadowsky 2008; Byappanahalli et al. 2012), which makes them a suitable candidate for defining microbiological reference level. This is the first time the reference level of C. perfringens is defined in the tropical developing world despite it being used frequently for evaluating faecal pollution of their water systems (Byamukama et al. 2005; Mushi et al. 2010; Macharia et al. 2015). When the concentrations of C. perfringens from the headwater streams (denoted as reference) were compared with the river waters subjected to anthropogenic influences, concentrations of C. perfringens were significantly greater (P < 0.05) in agricultural and urban river systems than in reference streams (Figure 3). This suggests that the investigated headwater streams can be used as a surrogate for negative control during evaluation of anthropogenically influenced river systems in tropical environments.

This study provides a useful basis for setting an acceptable contamination level that does not hamper critical services provided by tropical streams as there is no significant health risk associated with the low concentrations of C. perfringens detected in the investigated stream water. Despite the fact that the reference level of C. perfringens reported here can be directly applied to tropical environments similar to that of this study, the existing gaps regarding the concentration of C. perfringens contributed by each wild animal species, their accumulation and removal rates, the pathways in which these bacteria get into stream water systems, and their ecology in stream water have to be filled in order to validate their applicability in this type of environment.

This paper showed that C. perfringens was consistently detected in the catchment soil and in the stream water matrices of the investigated tropical highland with the catchment soil having significantly higher C. perfringens concentrations than the stream water matrix. The catchment soil was found to be the main source of C. perfringens in the stream water. Despite this contamination, the geometric mean of stream water was below the permissible safety level recommended for tropical environments. When the concentrations of C. perfringens in the stream water were compared with those of anthropogenically influenced rivers, significant differences were observed. Consequently, the geometric mean of C. perfringens concentration detected in the tropical highland streams was recommended as reference level and the streams in which this concentration was determined were suggested as reference streams.

I thank James Mwesongo for his assistance during sampling and processing of samples. This work was supported by funds from Ministry of Education and Vocational Training (MoEVT) of Republic of Tanzania awarded to DM.