Seasonal variability of faecal indicator bacteria numbers and die-off rates in the Red River basin, North Viet Nam

The Red River is the second largest river in Viet Nam and constitutes the main water source for a large percentage of the population of North Viet Nam. Here we present the results of an annual survey of Escherichia coli (EC) and Total Coliforms (TC) in the Red River basin, North Viet Nam. The objective of this work was to obtain information on faecal indicator bacteria (FIB) numbers over an annual cycle and, secondly, to determine the die-off rates of these bacterial indicators. Monthly observations at 10 stations from July 2013–June 2014 showed that TC and EC reached as high as 39100 cfu (colony forming units) 100 ml−1 and 15300 colonies 100 ml−1, respectively. We observed a significant seasonal difference for TC (p < 0.05) with numbers being higher during the wet season. In contrast, no significant seasonal difference was found for EC. The FIB die-off rates ranged from 0.01 d−1 to a maximum of 1.13 d−1 for EC and from 0.17 d−1 to 1.33 d−1 for TC. Die-off rates were significantly higher for free bacteria than for total (free + particle attached) bacteria, suggesting that particle attachment provided a certain level of protection to FIB in this system.

Die-off rates. At four stations, a second series of samples was collected in the same way for the determination of FIB die-off rate over time. These stations were (1) Yen Bai, located in the upstream main branch of the Red River known as the Thao River; (2) Ha Noi, after the confluence of three major upstream tributaries of the Da, Thao and Lo rivers; (3) Gian Khau, a peri-urban river system located in the Red River Delta and, (4) Truc Phuong, located in the downstream Red River on the Ninh Co River. These four stations were chosen to give a good representation of the land uses and population densities in the basin and to provide a good geographical separation over the area. For each station, duplicate 750 ml samples were incubated for five days in glass bottles in the dark and at the in situ temperature of the Hanoi station (17-29 °C). This temperature was selected for two reasons. The first being that only one incubator was available for the experiment and the second being that temperature at the Hanoi station was close to the average of the temperatures at the other stations. For the estimation of die-off rates, samples were collected from the incubations every day during 5 days (T0, T1, T2, T3, T4, T5) to determine the decrease in FIB numbers for both total and free bacteria using the method described below.
Analytical methods. Temperature, dissolved oxygen (DO), pH, and total suspended solids (TSS) were measured using a water quality probe WQC-22 A (TOA, Japan) and conductivity (Cond) was determined using a conductivity meter (Hach, USA) immediately upon sampling. Nutrients (N, P, Si) were spectrophotometrically determined on a Drell 2800 (HACH, USA) in the laboratory according to APHA 17 methods.
FIB abundance (free and attached) was measured by a direct count method using 3 M Petrifilm ™ E.coli/ Coliform Count Plate (Petrifilm EC plate), which contain Violet Red Bile (VRB) nutrients. E.coli (EC) produces beta-glucuronidase, which produces a blue precipitate and Total coliform (TC) colonies growing on the Petrifilm EC plate produce acid and the colonies are denoted by dark red points. This method has been validated by the APHA and is a technique commonly used for coliform and EC counts 18,19 .
For the total counts (free + attached bacteria; EC tot and TC tot ), 1 ml was removed from each sample (or incubation) after shaking to ensure an even distribution of bacteria, the sample was then aseptically delivered to the center of a Petrifilm EC plate. The water sample was then left to stand for 1 h and a second 1 ml aliquot was inoculated onto a second Petrifilm EC plate to estimate the number of free EC (EC free or TC free ; i.e. non-sedimented). The Petrifilm EC plates were then incubated at 37 °C for 24 hours, using a Fukusima incubator (Japan). The number of colonies (EC and TC) was determined using a Colony Counter CL-560 (Sibata, Japan). To facilitate the Scientific RepoRts | 6:21644 | DOI: 10.1038/srep21644 comparison of our data with that of previously published data and with water quality limits, we express our data as the number of colonies per 100 ml of sample.
The number of attached EC or TC (EC att or TC att ) is determined from the difference between EC tot (or TC tot ) and EC free (or TC free ) as: The die-off rates of TC and EC were estimated by fitting an exponential equation to bacterial abundances measured over time. The equations were expressed as first order decay in the general form of: where C t = is the number of EC or TC at elapsed time t, C o is the initial number of EC or TC per ml, k is the decay constant in day −1 and t is the elapsed time in days. As in several incubations complete die-off was observed by day 4, k is calculated for the first 4 days for all incubations to ensure comparability between dates and stations. The k value was determined for both the free and total fractions (e.g. EC free , TC free , EC tot , and TC tot ). All statistical analyses were performed with XLSTAT (v. 2014). Pearson's correlation was used to test the relationships between environmental variables and FIB. Wilcoxon's non-parametric test was used to test for significant differences between variables and the Kruskal-Wallace test was used to test differences between stations  Table 1. Location and characteristics of the surrounding areas at the 10 stations. River depth (m) at the sampling site is also provided. All samples were collected from the surface layer as grab samples. and season as the data were non-normally distributed even after normalization. When statistical relationships concerning FIB number were tested, log(EC) or log(TC) was used. When a significant difference was observed, an a posteriori Dunn's all-pairwise test was used and significance is determined as p < 0.05.

Results
Physico-chemical variables. The variability of the physico-chemical parameters observed in the ten stations for the wet and dry seasons are presented in Figs 1 and 2. Temperature varied between 9.5 and 35° C and was significantly higher during the wet season (p < 0.0001). Temperature was generally highest at Ba Lat and lowest at Gian Khau. The pH ranged from ranged 7.1 to 8.8, with the lowest values observed at the downstream Ba Lat station and the highest at upstream Hoa Binh station, however, no seasonal differences were observed. Salinity was almost 0 at all stations during the entire year; the only exception was during March at the most downstream station (Ba Lat, located at the river mouth and under tidal influence) where a salinity of 8.5 was observed (data not shown). The highest conductivities were also observed at this station at this time (1410 μ Scm −1 ). At the other stations, conductivity varied between 136-423 μ Scm −1 (Fig. 1) and had no significant seasonal pattern.
Concentrations of NH 4 , PO 4 and TP are shown in Fig. 2. Ammonium varied from 0.01 to 0.6 mg N l −1 and was significantly higher during the dry season (November -April) at low dilution and in the downstream stations (Gian Khau, Nam Dinh, Truc Phuong and Ba Lat) of the highly populated delta area. Similar to NH 4 , the PO 4 tended to be higher during the dry season however the difference was not significant. TSS was significantly higher during the wet season, when particle load is higher (p < 0.0001). Overall, TSS concentrations were significantly higher at the upstream Yen Bai station (p < 0.03) than at the other stations. As for TSS, TP was significantly higher during the wet season (p < 0.012). High concentrations of TP were found at Yen Bai and Gian Khau during the dry season (January) when values of up to 1.39 and 1.59 mg PO 4 l −1 were observed.
FIB abundance: monthly observation. The EC tot , and TC tot numbers for each season at the ten different stations are shown in Fig. 3 and Table S1 (Supplementary Materials). The number of TC tot varied between 0 colonies 100 ml −1 at Hoa Binh in January to over 39100 colonies per 100 ml −1 at Yen Bai in September. During the wet season, TC tot at Yen Bai, the most upstream station, was 9783 ± 14431 (mean ± standard deviation) colonies 100 ml −1 as compared to 4383 ± 3797 colonies 100 ml −1 for this station during the dry season, with no significant difference between these two seasons. For the other upstream tributaries, relatively low mean TC tot numbers were observed during the wet season (1850 ± 715 and 800 ± 789 colonies 100 ml −1 for Hoa Binh and Vu Quang, respectively. Mean values in the downstream delta stations were 6066 ± 4506, 5408 ± 5379 and 6050 ± 5469 colonies 100 ml −1 for Nam Dinh, Truc Phuong and Ba Lat, respectively. The overall pattern was similar for the dry season, i.e. higher mean values at Yen Bai and in downstream delta stations such as the peri-urban Nam Dinh station where a seasonal average of 5016 ± 5840 colonies 100 ml −1 was found. When the data from all of the stations were combined TC tot exhibited a significant seasonal difference with lower numbers during the dry season as compared to the wet season ( Fig. 3; p = 0.042).
EC tot numbers were significantly lower than TC tot (p < 0.05) during both seasons at all of the stations. As for TC tot , low numbers of EC tot were found in the upstream tributaries during wet season (283 ± 248 and 150 ± 234 colonies 100 ml −1 at Vu Quang and Hoa Binh, respectively), and high numbers were observed at Yen Bai (3108 ± 5960 colonies 100 ml −1 ). The more anthropogenically impacted downstream stations of the delta had high values during wet season (e.g. 1733 ± 1632 colonies 100 ml −1 at Nam Dinh). During the dry season, a similar pattern was observed with high mean EC tot numbers at Nam Dinh (2266 ± 3007 colonies 100 ml −1 ) and low values in the upstream stations (167 ± 122, 83 ± 75.7 colonies 100 ml −1 for Vu Quang and Hoa Binh, respectively). Regarding the dataset as a whole, there was no significant difference between the wet and dry seasons for EC tot .
Particle attached FIB. The percentage of EC att and TC att varied greatly between station (Fig. 4, Table S1).
For example, the percentage of EC att varied between 9.7% and 100% with a mean ( ± se) of 46.5 ± 2.9% (Table S1). The percentage of TC att was also highly variable (2.7-100%) with a mean 34.8 ± 1.8% for the whole data set combined. The number of EC att or TC att was significantly lower than for EC free or TC free when the entire dataset was examined (p < 0.05).
The number of EC att differed significantly between station (p < 0.05) and was highest at Yen Bai (4900 cells 100 ml −1 ) and lowest at Hoa Binh (100 colonies 100 ml −1 ; Table S1), similar to what was observed for the total number (free + attached). As with EC att , the number of TC att differed significantly between station. TC att was highest at Yen Bai (13050 colonies 100 ml −1 ) and the lowest at Hoa Binh (0 colonies 100 ml −1 ). The numbers of free and attached TC and EC were positively correlated with each other and had correlation coefficients of 0.73 and 0.81, respectively. This indicates that attached and free TC and EC increased concomitantly even though the actual numbers were different.
Die-off rates. The die-off rates of EC and TC calculated over 4 days are shown in Table 2. In general, die-off rates for TC tot were significantly higher than for EC tot . For example, at Yen Bai in August, the die-off rate for TC tot was 1.33 d −1 compared to 0.50 d −1 for EC tot and 1.03 d −1 and 0.43 d −1 TC tot and EC tot respectively, at Truc Phuong during February. It is worth mentioning again the specific behavior of FIB at Yen Bai. Along with the high TC and EC numbers at this station, TC tot die-off rates were higher than at the other three stations examined e.g. 1.33 d −1 as compared to 0.70 d −1 , 1.03 and 0.42 d −1 for Hanoi, Truc Phuong and Gian Khau, respectively. Similarly, EC tot die-off rates at Yen Bai were also significantly higher (p < 0.05) than those observed for the three other stations. However, and in contrast to TC free , we observed a significant station effect for EC free with the rates found at Yen Bai and Truc Phuong being significantly higher than those found at the two other stations, Hanoi and Gian Khau.    Overall, die-off rates were significantly higher for TC free than for TC tot (p < 0.05) when the whole data set was examined. In contrast, the opposite was true for EC tot and EC free (p < 0.05).

Relationships between FIB and environmental factors.
When the data set was regarded either as a whole or by station, EC and TC (free, attached and total) numbers were only positively correlated with TSS and TP (p < 0.05; Table 3). No other significant relationship was found between EC or TC and the measured environmental variables (pH, DO, NH 4 , conductivity, or PO 4 ).
As for TC tot and EC tot numbers, the die-off rates of both TC free and EC free and TC tot and EC tot were uncorrelated with temperature, DO, NH 4 or conductivity (Table 4). However, EC tot die-off rate was significantly correlated with TP and TSS. No significant correlations were observed for TC tot and TC free and TSS.

Discussion
Distribution of FIB. Land use affects the abundance and distribution of FIB in both tropical 20,21 and temperate regions 22,23 . It is also probably the case in the Red River system. In general, mean FIB numbers (EC, TC) in the upstream stations (e.g. Vu Quang, Hoa Binh, Son Tay) were lower than the downstream stations (Nam Dinh, Truc Phuong and Ba Lat). Of the ten stations investigated, the upstream Hoa Binh station on the Da River had the lowest FIB levels. The Da River sub-basin is primarily forest (74% of total sub-basin area) and population density is sparse (< 100 people per km −2 ) and this probably explains the low FIB numbers at this station. In contrast, Yen Bai on the Thao River has the highest and most variable levels of contamination, despite being upstream. The Thao River basin has a relatively high population density (190 people km −2 ) and a large proportion of the total    sub-basin area (33%) is used for agriculture and livestock grazing. Livestock are known to be an important source of FIB contamination in streams and rivers 20 and this along with non-treated human wastewater from the surrounding population may well explain the high and variable FIB numbers observed at this site during both seasons. Furthermore, the Van Yen industrial zone which processes agricultural products is located in the upstream Thao River and this is probably also a source of FIB. Indeed, the high and variable FIB numbers observed during the wet season also suggest that FIB originated mostly from diffuse runoff sources. The higher FIB abundances observed in the downstream stations (Nam Dinh, Ba Lat and Truc Phuong) are also probably related to the surrounding land use. These three sites are in peri-urban areas of the delta with high population densities (1200 people km −2 compared to 100-150 people km −2 in the upstream basin 14 ), agriculture and industry all of which release untreated wastewater into the system. Although the Hanoi station is located in the city and the FIB numbers are relatively low (Fig. 3), they are still above the acceptable limits for individual, non-commercial water supplies in Viet Nam (20 EC colonies 100 ml −1 and 150 TC colonies 100 ml −1 ) 24 . Domestic wastewater from Hanoi city is mostly discharged into small urban rivers as shown by the high ammonium concentrations in the To Lich and Nhue Rivers 10,25,26 and other work from the urban rivers in Hanoi city has reported very high TC numbers. For example, TC numbers at Lien Mac, at Phu Van Bridge (Nhue River) and at Thanh Liet Dam (To Lich river) were 7820, 9820 and 148480 MPN 100 ml −1 , respectively 27 meaning that the water at these locations is considered as being unfit even for irrigation due to its high FIB load.
Free and attached FIB. Understanding the relative numbers of attached and free FIB and their differential fate allows us to better estimate the time these indicator organisms remain viable in the environment. In soils and sediments, bacteria tend to be associated with particles as opposed to in the free-state e.g. 28 whereas in aquatic systems the percentages of particle associated bacteria are highly variable and can range from 10% in clear waters with very low organic particle loads to over 70% in estuaries with high particle loads 29,30 . Suter et al. 31 have shown that high proportions of FIB are associated with particles (52.9% ± 20.9% and over 90% in some areas) and that these values were related to turbidity levels.
In our work, free-living bacteria generally predominated (only 36% ± 20% of TC were particle attached and 50% ± 26.9% of EC particle attached). However, the %EC att and %TC att were highly variable (7.8% to 100% and 2.7% to 80% for EC att and TC att , respectively). Previous studies from temperate systems have indicated that the attachment to particles by FIB in the aquatic environment is influenced by various factors, including temperature, bacterial genotype, soil particle size, organic matter, pH, ionic strength, dissolved nutrients and turbidity 32,33 . We also found significant correlations among TC att , EC att , TSS and TP. In one of the few articles investigating the factors controlling FIB concentrations in the tropics, Byamukama et al. 34 working in Uganda, also found that FIB were correlated with TSS concentrations as we show here. The Red River system has high turbidity levels 8,14 and this may explain the relatively high %EC att and %TC att observed in this system. Moreover, attachment to particles probably plays a strong role in controlling the transport of FIB in the system as well 35 . Particles in aquatic systems are often associated with nutrients and particulate carbon and phosphorus in particular, both of which can be limiting substrates for bacterial growth and in turn, can affect FIB survival in non-host environments 36 and in systems with high particle loads, the TP pool is generally dominated by the particulate fraction, as is the case here. This, along with higher carbon concentrations in the particles probably confers a competitive advantage to the bacteria thereby enhancing their survival. However, for the moment, little other information exists on the proportions of attached and free FIB in other tropical water bodies and so it is difficult to compare our results with other tropical ecosystems. Nevertheless it also appears that TSS and TP plays an important role in determining the proportions of attached and free FIB in this highly turbid, tropical riverine system. Die-off rates. Schumacher 37 in an in-stream incubation in the Upper Shoal Creek Basin, Southwestern Missouri, found that fecal coliform and EC densities decreased more than 90% from initial densities over a 42 h period. Troussellier et al. 38 also reported rapid die-offs with over 90% of FIB lost over a 128 h period. We also observed a rapid decrease in EC and TC over the 5 day incubation, with in many cases up 100% loss after 120 h. We therefore chose to calculate the decay constants over the first 4 days. Our TC and EC die-off rate decay constants (k) ranged from a minimum of 0.01 d −1 to a maximum of 1.13 d −1 for EC and from 0.17 d −1 to 1.33 d −1 for TC, with a mean of 0.36 ± 0.21 d −1 for EC free and 0.44 ± 0.23 d −1 for EC tot and 0.83 ± 0.19 d −1 for TC free and 0.61 ± 0.28 d −1 for TC tot . These rates are similar to other studies from temperate aquatic environments. For example, Menon et al. 39 , working in the Seine river, France observed die-off rates of 0.19 to 0.82 × 10 −3 h −1 for EC and Blaustein et al. 40 , in a review on EC survival in a range of temperate aquatic environments reported an average rate of 0.725 ± 0.078 d −1 . In one of the few articles dealing with die-off rates in sub-tropical systems, Chan et al. 41 also found values of between 0.85 to 1.50 d −1 for the coastal water around Hong Kong, similar to the values we report in this work.
We observed higher die-off rates for TC free than for TC tot , however, this was not the case for EC, where die-off was significantly higher for EC tot than for EC free . Some authors 42,43 have shown lower decay rates for fecal bacteria attached to sediments as compared to free fecal bacteria, as we show here. Attachment to particles may protect FIB against grazing by small heterotrophic nanoflagellates (the main grazers of bacteria in aquatic systems) as well as providing a micro-environment rich in nutrients. Although, according to Sinton et al. 44 and Chan et al. 41 , exposure to solar radiation and predation are among the most important factors controlling die-off, although it seems from our work that attachment to particles can also be an important factor at least for TC.
We observed systematically higher die-off rates for TC tot than for EC tot . Why this might be the case in this system merits some reflection. The appropriateness of using TC and EC as indicators of faecal contamination in tropical systems has been questioned by several authors e.g. 45,46 . Indeed, it has been shown that E.coli may be able to persist and even proliferate for some time in tropical freshwaters [47][48][49] particularly in those with high temperatures and elevated nutrient and organic matter concentrations. This may well explain the differences in die-off rates between EC and TC. It may also be why die-off in EC free was lower than that for EC att . Although the technique used in this work allowed us to determine the presence or absence of TC or EC, it provides no information on the sources of the TC or EC. Therefore, if the FIB are from different origins with different levels of adaptation to the environment, then this may explain some of the differences between the die-off rates of free and attached TC and EC. Nevertheless, this clearly is a hypothesis that merits further investigation in situ using some of the newer microbial source tracking techniques 36 .
Many studies have indicated that seasonal variations can influence FIB die-off rates through modifications of temperature, pH, nutrient concentration, and dissolved oxygen 50,51 . In this study, die-off rates were determined over the whole year with temperatures varying between 9.5 and 35 °C and we did not observe any significant relationship with in situ temperature. Moreover, we only observed a significant correlation between temperature and TC free in contrast to what has been found by other investigators 23,52 . We did, however, find significant differences in FIB numbers and die-offs between the wet and dry seasons, emphasizing the dominant role of the hydrology. Why we did not find a clear temperature relationship for FIB number or die-off rate is not clear from the data, however, it is probably related to other environmental and source related factors not investigated in this work.

Conclusions
There have been many studies on the survival of EC and other indicators of faecal pollution in freshwater, marine and estuarine habitats in temperate ecosystems however less information exists for tropical aquatic environments. Crane and Moore 53 observed that the identification of relationships between environmental and physical parameters and FIB survival was a fundamental topic for future research in the field. Although their work is dated, and much work has been published since on temperate ecosystems, it is still the case that very little information is available on die-off rates in tropical and sub-tropical ecosystems where the management of FIB concentration is most needed and where the risks to human populations are highest 36 . In our work from the sub-tropical Red River system in Viet Nam, we found that FIB numbers exceed Vietnamese water quality guidelines of 20 and 150 colonies 100 ml −1 for EC and TC, respectively, throughout the whole year at almost all of the 10 stations investigated. Moreover, many values exceeded 500 colonies 100 ml −1 above which the World Health Organization considers that there is a 10% risk of gastro-intestinal illness after one single exposure. Therefore, the use of water from sites with high FIB numbers such as those in the downstream sites pose a real risk to public health. We also found a significant correlation between particles (as TSS or TP) and FIB as has been observed in temperate riverine ecosystems. However, in contrast to other studies we did not find a significant correlation between FIB and temperature for this subtropical environment, but instead a correlation with discharge (wet vs. dry season). Indeed, the highest TSS concentrations and FIB numbers were found during the wet season at high discharge. It is therefore probable that in developing countries where sanitation facilities are deficient and where people and their livestock live in close proximity to the river, FIB contamination mostly originates from diffuse sources 21 . Nevertheless, the data presented here, notably that of FIB numbers and their respective die-off rates, provides a base for the application of a model that will allow the parameterization of FIB dynamics in Red River and delta and potentially in other large river basins of the sub-tropical and tropical belt.