Tracking the downstream impacts of inadequate sanitation in central Appalachia

Despite major advancements in wastewater treatment in the past century, inadequate sanitation remains an issue for some communities in the United States. Infrastructure for sanitation is particularly lacking in remote communities where there are fewer users to cover infrastructure costs and technical assistance is not readily available locally (Tarricone 1989; Gasteyer & Vaswani 2004). This can result in inadequate wastewater management strategies that direct untreated household waste to local surface waters with minimal pre-treatment (e.g. soakaways, straight pipes), posing potential health risks to adjacent and downstream communities.

Improper treatment of household wastewater has long been recognized as a public health hazard. In the most recently identifiable national survey of medical care providers in rural communities, surface water pollution was identified as a primary environmental health concern (Robson & Schneider 2001). Waste entering common waterways used for recreation or drinking water can be a point of human exposure for infectious disease, particularly in low population areas (Denno et al. 2009; Collier et al. 2012). Relative to less developed countries, the incidence of waterborne diseases such as hepatitis A, salmonellosis, and typhoid is very low in the United States, but real risks do exist in areas without proper water and sanitation (Gasteyer & Vaswani 2004). Eliminating these sources of pathogen contamination could result in both significant public health and economic benefits, as waterborne illnesses have multiple adverse economic effects including the cost of hospitalizations, losses in productivity at places of employment, and absences from school (Collier et al. 2012).

A particularly unique set of conditions exists in rural communities in central Appalachia. In this region, underlying karst soils are often inadequate to support traditional on-site wastewater treatment strategies such as septic systems, i.e. the soils are too thin to provide sufficient biological treatment prior to discharge to the environment. Furthermore, in narrow Appalachian valleys, residences are often located too close to streams to permit the construction of proper drain fields (Cook et al. 2015). Other on-site solutions, such as package treatment plants or sand filters have seen limited use, but are often precluded by distinctly local challenges (e.g. the necessity of shared responsibility for installation and maintenance in already resource-stressed communities; Tarricone 1989). When traditional municipal or onsite septic systems are not readily available, untreated household waste may simply be ‘straight piped’ into nearby surface waters (Banks et al. 2005). The number of straight pipes in the entire region is unknown, but the practice is not uncommon, e.g. in Letcher County, KY, one survey estimated 3,000 straight pipes serving about 40% of 30,000 total residents (Glasmeier & Farrigan 2003). Though the discharge of untreated household wastewater is intuitively expected to result in significant deterioration of receiving water quality, little monitoring data describing the magnitude of impact is available.

Standard monitoring practices to identify water quality impairments generally only target fecal indicator bacteria (FIB) such as total coliforms or Escherichia coli. While generally non-pathogenic themselves, these are considered sentinels of potential pathogen presence and associated human health risks. However, because these organisms are common to the intestinal tract of all warm-blooded animals, their presence provides limited information regarding the relative contributions of specific sources of fecal contamination (e.g. human vs. animal). Though the transfer of zoonotic pathogens is important, fecal contamination by human sources is generally considered of greatest health concern since many waterborne pathogens (viruses in particular) exhibit host specificity (Harwood et al. 2014). The inclusion of library-independent, established source-specific markers in monitoring efforts can greatly aid in watershed assessment: understanding the origin of fecal contamination can improve local understanding of potential health risks and inform decisions about appropriate remediation actions to meet water quality goals (Scott et al. 2002). At present, the genetic marker HF183 from the anaerobic Bacteroides spp. is well established as strongly indicative of the presence of human fecal contamination in waters with multiple potential contamination (Seurinck et al. 2005; Harwood et al. 2009, 2014; Boehm et al. 2013; Sidhu et al. 2013). Because the majority of prior studies have tracked Bacteroides-HF183 concentrations downstream from relatively large centralized wastewater treatment facilities or highly urbanized areas, the relationship between FIB, source tracking markers, and health risks in less populated regions is poorly understood (Harwood et al. 2009). However, application of these more advanced monitoring techniques in rural regions with known sanitation issues could be useful in tracking the downstream reach of contamination from seemingly isolated communities and justifying and/or evaluating infrastructure improvements. In addition, these relatively simple and sparsely populated systems represent a unique opportunity to evaluate the robustness of common microbial source tracking (MST) markers compared to settings in which they are more commonly used.

This study aimed to directly link microbial measures of fecal contamination with the prevalence of straight pipes in two central Appalachian streams. Specific objectives included: (1) detection of the human source tracking marker Bacteroides-HF183 in receiving waters polluted by untreated household waste from rural communities; (2) assessment of the co-occurrence of Bacteroides-HF183 with E. coli and Enterococcus, which are currently recommended by regulatory agencies as sentinels of fecal contamination and associated public health risks based on culturable and molecular targets, respectively (USEPA 2012; Haugland et al. 2013); (3) exploration of correlations between microbial indicators and physiochemical water quality parameters; and (4) development of a spatial statistical strategy to link straight pipe locations to longitudinal downstream measures of E. coli in these streams.

Water samples were collected monthly from August 2012 until August 2014 and from January 2016 to April 2016. For all months and at all sites, samples were collected for E. coli analysis (n = 28 for each site), and in-stream temperature, dissolved oxygen, pH and specific conductivity were measured on-site using a YSI Quattro Pro Plus (YSI, Yellow Springs, OH, USA). Between April 2014 and August 2014 and January 2016 to April 2016, an additional sample aliquot was collected for subsequent molecular analyses (n = 9) from five sites in each watershed. All samples for microbiological analysis were collected at the thalweg in 250 mL pre-sterilized bottles, and transported on ice back to the Biological Systems Engineering Seitz Hall laboratory at Virginia Tech for prompt analysis (i.e. within 6 hours for culturable microorganisms; filtration for molecular preservation within 48 hours).

Analysis for E. coli concentrations occurred immediately upon return to the laboratory via the Colilert defined substrate method using the Quanti-Tray/2000 (IDEXX, Westbrook, ME, USA). Following 24 hours incubation at 37 °C, a most probable number (MPN) of bacteria could be determined based on the number of positive wells (i.e. fluorescent under UV light).

For samples collected for molecular analyses, two additional 100 mL aliquots from each sample were vacuum filtered through a Millipore 0.4 μm filter (Merck KGaA, Darmstadt, Germany). Filter effluent was discarded and filters were stored in cryotubes at −80 °C prior to DNA extraction. Within six months, DNA from the filters was extracted using a PowerWater DNA Isolation Kit (MO BIO Laboratories, Carlsbad, CA, USA). Copy numbers of Enterococcus were quantified via quantitative polymerase chain reaction (qPCR) as described previously by Haugland et al. (2005). Copy numbers of the human-specific Bacteroides-HF183 MST marker were quantified by the procedures outlined in Griffith et al. (2013) and Seurinck et al. (2005) (Table 1). Samples were run in triplicate, with each well containing 20 μL of Master Mix and 5 μL of sample. A no-template control was tested along with the samples to indicate any sources of contamination. A base plasmid containing the target marker at known amounts was used to create a standard calibration curve. Results were considered valid if the run-specific standard curve had an R² value above 0.99 and efficiency between 0.90 and 1.10. The Bacteroides-HF183 master mix also included a plasmid containing an internal amplification control (IAC) sequence to quantify inhibition (Griffith et al. 2013). The limit of detection (LOD) for Bacteroides-HF183 was determined by scaling the lowest detectable amount of Bacteroides-HF183 in tested samples to the total amount of eluted sample from DNA extraction. The same process was repeated with the lowest value on the standard curve (10 copies/mL) to determine the limit of quantification (LOQ).

All statistical analyses were conducted in JMP 12 (SAS, Cary, NC, USA). Normality/non-normality of each dataset was determined via observation of the normal quantile plot prior to analysis. Correlations between molecular targets, E. coli, and general measures of water quality were assessed via Spearman's rank correlation.

Fixed variables included the sum of inverse distances to all upstream discharges of untreated household waste (D), as well as a binary variable indicating whether the season was winter (S; where December, January and February were considered winter months). Although the location of known waste discharges included both septic tanks and straight pipes, the effects from these sources were not differentiated, as local health department sanitary surveys suggest most local septics do not meet codes and/or are failing. Random variables were normally distributed and centered at 0, and accounted for dependence of measurements taken within sites (⁠⁠) as well as possible sources of random error (⁠⁠). Coefficients (⁠ and ⁠) were determined by the model to describe the behavior of the S and D variables and the general direction of the correlations.

As illustrated in Figure 2(a), Bacteroides-HF183 concentrations were negligible at the most upstream point in Callahan Creek, which was expected given that this sampling site is above the known discharge points of untreated household waste. The next sampling site is directly below a residential community, and shows a spike in the concentrations of Bacteroides-HF183. Samples taken directly from the household waste effluent discharge to Callahan Creek ranged from detectable but not quantifiable (DNQ) to 59,400 copies per mL. Though this value may appear high, it is notably lower than typical values of Bacteroides-HF183 in raw sewage which have been observed between copies/mL (e.g. Seurinck et al. 2005). In Callahan Creek itself, Bacteroides-HF183 levels remained between 10 and 10³ copies/mL (43% positive samples and 13% quantifiable samples) until the furthest downstream sampling site, where it was again below the LOD. Observed levels are similar to those reported by Tambalo et al. (2012), in a rural environment with known discharges of human waste in Saskatchewan, Canada. In this study, Bacteroides was detected via qPCR of the genetic marker BacH, at concentrations at or below 10⁴ copies/mL in different sites at positive detection rates between 25 and 47% of collected samples. In contrast, a recent urban study by Sidhu et al. (2013), reported Bacteroides-HF183 in over 90% of the samples adversely impacted by untreated or partially treated human waste. Denser population centers with sewage intrusion appear to result in more successful detection of Bacteroides-HF183 and higher concentrations in receiving waters.

Both Enterococcus and E. coli are universally present in the human digestive system, and may also originate from wildlife or domestic animals, however, additional non-human significant discharges of E. coli are considered to be negligible in this area. Agricultural land use is low (0.5%), and samples collected in wholly forested areas upstream from the straight-piped communities generally showed barely detectable E. coli (geometric mean = 4 MPN/100 mL), suggesting low levels of background/wildlife. Longer longitudinal persistence in Enterococcus and E. coli levels therefore may be the result, at least partially, of longitudinal decay. Unlike Bacteroides, which is an obligatory anaerobe, E. coli and Enterococcus are facultative anaerobic bacteria, and therefore can survive in the aerobic surface water environment. Highly oxygenated mountain streams may cause particular stress to Bacteroides.

Seasonal trends are not immediately apparent in the recovery of Bacteroides-HF183 and Enterococcus from Callahan Creek, though values varied considerably by month (Table 2). E. coli was generally lower during winter and early spring months (i.e. January–April) than late spring and summer months (i.e. May–August). There were two notably high values in June 2014 in consecutive sampling sites (i.e. CC09 and CC06). This is consistent with expectations since flow rate should generally increase during the colder/wetter month versus the warmer/drier months while bacterial loads remain constant, thereby resulting in negative correlations with the measured concentration of microbial indicators. However, positive samples were collected even during winter months, when the highest streamflow values would be expected to dilute discharges, and only weak correlations were found between E. coli and flow rate in Callahan Creek (Table 3). Therefore, while seasonal water quality parameters may provide some anecdotal explanation of variability in Bacteroides-HF183, Enterococcus and E. coli, additional factors such as sample collection timing are also likely important (i.e. collection directly following straight pipe discharge).

To investigate further, a correlational analysis was conducted between microbial indicators and several general water quality parameters and flow rate (Spearman's rho values, Table 3). Bacteroides-HF183 and conductivity were significantly and positively correlated in Callahan Creek (i.e. Spearman's rho = 0.37). This is consistent with expectations since both parameters can originate from the same source: untreated household waste. However, in Roaring Fork, conductivity and Enterococcus were inversely correlated (i.e. Spearman's rho = −0.44). This seems contradictory to expectations and perhaps suggested additional unattributed sources of conductivity (i.e. other than the untreated household waste discharges) in this stream (e.g. additional residences, adjacent railroad tracks).

A weak but significant correlation was observed between E. coli and pH in this stream (i.e. Spearman's rho = 0.21). In Callahan Creek, a slight spike in pH values was observed directly below the untreated household waste discharges, and then not far downstream a return to similar values as in the headwaters. The elevated pH near the waste discharges may be related to household chemical use or high pH in the water source for these residences (relative to the native stream pH). No significant correlation was found between pH and microbial indicators in Roaring Fork.

In both streams E. coli and temperature were significantly positively correlated (i.e. Spearman's rho of 0.24 in Callahan Creek and 0.26 in Roaring Fork) which may reflect higher detection during low flow warmer months. In keeping with this, flow rate was identified as an important seasonal variable in Callahan Creek. A negative correlation between the flow rate and E. coli (i.e. Spearman's rho = −0.19), as well as depth and E. coli (i.e. Spearman's rho = −0.23), was observed for that stream. In both streams, a significant and negative correlation was also observed between E. coli and dissolved oxygen (i.e. Spearman's rho of −0.25 in Callahan Creek and −0.22 in Roaring Fork), which could be related to biological and/or chemical oxygen demand contributed by untreated household waste.

The fact that multiple significant correlations were found between E. coli and other parameters, whereas the genetic markers were only correlated to conductivity, is likely due to the relatively higher number of quantifiable data points available for E. coli. The higher number of stronger correlations observed between microbial indicators and other parameters in Callahan Creek as compared to Roaring Fork may simply be related to the relatively larger number of untreated household waste discharges, and thus presumably higher waste volume.

All of the microbial indicators targeted can originate from human fecal contamination, and therefore a positive correlation between them would be expected; however, previous examinations of potential correlations between culturable indicator bacteria and genetic markers used in source tracking have been inconsistent, which may be a result of the inherent differences in analytical targets (Harwood et al. 2014). Culture-based methods only indicate the presence of viable organisms while non-culture based methods may detect non-viable strands of DNA; the two types of targets may have different rates of decay and/or transport mechanisms. In a laboratory study in Gulf of Mexico, Harwood et al. (2009) did not find any correlation between Enterococcus and human associated Bacteroidales, nor did Sidhu et al. (2013) find a significant correlation between E. coli and Enterococcus in a field study in commercial and residential catchments. Kapoor et al. (2013) found a moderate correlation (r² = 0.17) between human associated Bacteroidales and E. coli genetic markers (i.e. uidA gene) downstream from combined sewage overflows in an urban environment. The only attempt at a similar analysis in a rural environment (Tambalo et al. 2012) found no correlation between the Bacteroides marker BacH and E. coli.

The sum of the inverse distances to all upstream sources of untreated household waste and the concentration of E. coli was significantly and positively correlated (p < 0.05) for both Callahan Creek and Roaring Fork (Table 4). There was also a significant, positive relationship between the concentrations of E. coli and seasonality for Callahan Creek, though this relationship was not significant for Roaring Fork. Water temperature in the streams and flow in the Powell River (the receiving river for both Callahan Creek and Roaring Fork) were tested in both models but did not show a significant relationship with the concentrations of E. coli and therefore were not included. Similar analyses with Bacteroides-HF183 or Enterococcus as the dependent variables were not significant due to the high level of non-detectable results that skewed the dataset.

The significant waste discharge number variable indicated that with more discharges upstream from a sampling site, the concentration of E. coli is likely to be higher, as would be expected. It also indicated that E. coli concentrations decrease as distance from discharges increases, which may be attributable to die-off and/or dilution, given the previously mentioned lack of other appreciable inputs of fecal contamination (e.g. wildlife). The model also confirmed that concentrations of E. coli are significantly lower during the winter in Callahan Creek, which is consistent with the expectation that these point sources should be diluted by generally higher streamflow values in the winter months and/or lower temperatures may be affecting levels of persistence. However, it is important to recognize that the model does not explain all variability in longitudinal measures of E. coli concentrations on Callahan Creek or Roaring Fork (⁠ = 0.29 and 0.24, respectively).

Although the genetic marker Bacteroides-HF183 has routinely been detected downstream from highly populated areas contaminated by human sewage (Seurinck et al. 2005; Sidhu et al. 2013; Harwood et al. 2014), there has been little previous demonstration of its application to rural areas. This study demonstrated that human-specific Bacteroides-HF183 can be found in rural watersheds with known upstream direct discharges of untreated household waste (e.g. via straight-pipes). However, given its sporadic recovery, it may be most appropriate to use this marker in conjunction with more traditional culture-based FIB following initial sanitary surveys to identify areas most at risk. Used concurrently with E. coli (as has been previously suggested by Harwood et al. (2009) and Sidhu et al. (2013)), human-specific Bacteroides may provide strong evidence of human fecal contamination, and contribute to justification or evaluation of infrastructure improvements. Ideally, epidemiological studies and water quality information at the most common point of exposure (i.e. private drinking water sources) alongside water surveys would allow for more informed decisions on the scale of investment in wastewater infrastructure.

Funding for this project was provided by the Appalachian Research Initiative for Environmental Science. The views here are those of the authors and do not necessarily represent the views of the sponsoring organization, nor has the sponsoring organization taken any role in the research. Statistical assistance was provided by the Laboratory for Interdisciplinary Statistical Analysis, as well as Karen Kline from the Department of Biological Systems Engineering at Virginia Tech. The Virginia Department of Health provided information on the location of straight pipes and septic systems in the study watersheds.