Persistence of Bacteroides Species Populations in a River as Measured by Molecular and Culture Techniques

Abstract

Given the interest in Bacteroides species as microbial source tracking (MST) markers, and the limited knowledge of the survival of Bacteroides species in the environment, here we examine the survival of Bacteroides fragilis, B. thetaiotaomicron, and environmental species of Bacteroides by use of culture techniques and molecular tools. Two kinds of experiments were performed: (i) on-site experiments, in which bacteria were exposed to changes in the levels of several environmental parameters in a river, and (ii) microcosm assays in the laboratory, with controlled temperatures. On-site experiments showed different survival patterns for the cultivable Bacteroides strains. B. fragilis die-off rate was strongly affected by the combined effect of high temperatures and grazing predators, which were more active under warmer conditions. However, the survival rates of cultivable B. thetaiotaomicron and environmental Bacteroides spp. were more affected by dissolved oxygen (DO) concentration in water. Environmental Bacteroides strains survived longer than either type strain, due to better adaptation to environmental conditions. However, the period of their survival was shorter than that observed for fecal coliforms and enterococci, suggesting Bacteroides species as markers of recent fecal pollution. The total Bacteroides species were detected by molecular techniques throughout the experiment in winter, but they were detected on only two or three days in the summer. This indicates that temperature is the main factor affecting DNA degradation, regardless of species. The use of microcosms in the laboratory also pointed to temperature as the main factor affecting Bacteroides survival, regardless of species. However, the conditions in the laboratory may mask the effects of the environmental factors and their interactions. The observed variability in die-off rate as a function of the species analyzed, the experimental conditions, and the methodology used should be taken into consideration in future persistence studies.

Species of the order Bacteroidales are one of the most abundant bacterial groups in the human microbiota and other warm-blooded animals (18, 26, 58). Their anaerobic physiology prevents the reproduction and survival of these species in the environment (19, 28) and allows species identification between animal hosts (11, 27, 30). Bacteroidetes species or their bacteriophages have been suggested as indicators for microbial source tracking (MST) to determine the origin of fecal pollution in water (11, 14, 16, 39, 41, 43, 60). Over the last decade, authors have described various methods based on molecular-genetic tools to detect some specific sequences of uncultured species of the Bacteroidales group. Some of these sequences have been suggested as potential MST markers. These MST methods mainly rely on conventional PCR to detect specific sequences that are associated with human, ruminant, pig, or horse origins (10, 14, 16). Moreover, some authors have used quantitative PCR (qPCR) to discriminate the source and predict the concentration of the fecal markers (27, 30, 35, 39, 42, 43, 60). Although some of these methods have been tested successfully in different geographical locations (2, 20, 22, 24, 29), a certain lack of specificity has also been observed (5, 22).

Intestinal anaerobic bacterial populations may die off rapidly in environmental water or wastewater (37). Hence, they might suggest recent fecal pollution in aquatic environments. In the development of predictive MST models, the best markers and the lowest number of markers need to be selected. Besides, the resistance to water treatments and the persistence in the environment should be known (8, 12). Sequences of uncultured Bacteroidales have already been proposed as specific MST markers in predictive models (5). None of the strains associated with these molecular markers has been cultivated or identified to date, but these strains are assumed to be related to species of Bacteroidales. Therefore, there is a lack of knowledge on the ecologic, physiologic, and host-specific roles of the intestinal microbiota. Consequently, the persistence of Bacteroides spp. must be examined in order to develop proper predictive MST models. Additionally, inquiry as to the relationship between Bacteroidales or Bacteroides persistence and the persistence of other usual microbial indicators should be considered for modeling.

Biotic and abiotic environmental parameters have been shown to affect the survival and persistence of enteric bacteria and the biochemical degradation of DNA (9). For example, these occurrences can be affected by grazing protozoa, bacteriophages (6, 23, 34), temperature (36, 38, 51), solar irradiation (25, 51), and sedimentation (15), among other factors. Assays have been carried out to monitor the decay of Bacteroides spp. and Bacteroidales molecular markers and to determine the influence of biotic and abiotic parameters. These models mainly use microcosms and mesocosms under controlled parameters (9, 40, 50, 54, 55). Short survival periods, from hours to a few days, have been described for culturable Bacteroides species (19, 47, 48). In contrast, the DNA of fecal Bacteroides species remains detectable from days to weeks, depending on the conditions (9, 40, 50). Some environmental factors have been related to the survival of Bacteroidales species and the detection of the DNA in their molecular markers. Temperature is the most studied of these factors. Higher rates of die-off and nucleic acid degradation of Bacteroides populations have been found at higher temperatures (9, 28, 40, 48, 50). The presence of predators in the environment also has a major impact on the survival of Bacteroidales. Studies that compare the decreases in populations with and without predators in the water attest to the effect of predators on die-off rate (9, 28, 40). Variability among Bacteroidales species has also been reported (59). Bacteroides fragilis has a high oxygen tolerance and can even grow in low oxygen concentrations (7, 44), in which most Bacteroidales species can survive for only a few hours (49, 54). Sunlight inactivation has been reported to have a different effect on the microorganisms (6, 45, 55). Sunlight increases the reactive oxygen species in the environment, which could affect the Bacteroidales species, depending on the different oxygen protection mechanisms. The effects of sunlight on Bacteroidales are controversial (4, 55, 57). Meanwhile, other factors, such as salinity, have not shown any major effect (40).

The aim of this study was to improve our understanding of the survival of Bacteroides species that occurs when these species are released into the environment. Hence, various experiments were set up to compare the persistence and inactivation of environmental Bacteroides populations with those of the Bacteroides fragilis and B. thetaiotaomicron type strains in environmental water. The influence of physical and chemical water parameters and seasonal effects on the survival of different bacterial species and environmental and laboratory strains was also analyzed. The persistence of fecal Bacteroides species was compared with the persistence of other traditional microbial indicators (fecal coliforms [FC] and enterococci [ENT]) under a variety of environmental conditions. Experiments were performed in the field (on-site) and using microcosm assays, and the results were compared. Culture and molecular methodologies were used to detect differences between cultivable and total cells.

MATERIALS AND METHODS

Growth media and type strains.

The bacterial type strains used in this study were Bacteroides fragilis DSM 2151^T and B. thetaiotaomicron DSM 2079^T. Both bacterial strains were grown using Bacteroides phage recovery medium (BPRM) (Scharlau, Barcelona, Spain) (53). This medium was prepared according to the manufacturer's instructions, although neither of the antibiotics in the formula (vancomycin and kanamycin) was added. Cultures were incubated at 37°C for 24 to 48 h anaerobically. The type strains were used as positive controls and to spike the water matrices used in the study.

Experimental design.

The inactivation of Bacteroides spp. was evaluated by two kinds of analysis: (i) to imitate a real environment subjected to different factors and fluctuations, an on-site analysis was performed in the Llobregat River, and (ii) because temperature has been identified as one of the main factors affecting the die-off rate of Bacteroides species, laboratory assays were performed in microcosm with controlled parameters, with incubation of the microcosms at different temperatures.

Three on-site assays were performed to compare the effects of different factors on the persistence of Bacteroides species: (i) a comparison of the persistence of two type strains spiked in river water, those of B. fragilis and B. thetaiotaomicron, to evaluate differences among the genus; (ii) a comparison of the levels of persistence of B. fragilis spiked in river water and that spiked in sterilized river water by autoclaving to evaluate the role of predation; and (iii) a comparison of environmental Bacteroides strains from a sewage influent, fecal coliforms, and enterococci.

For the evaluation of B. fragilis and B. thetaiotaomicron, river water was spiked with a 1:10 dilution from an overnight culture for each type strain. In parallel, river water sterilized by autoclaving was spiked with the B. fragilis. Meanwhile, to evaluate the persistence of environmental Bacteroides species compared with that of fecal coliforms and enterococci, a sample of river water was spiked with 2% raw sewage.

The aforementioned water matrices were used to fill dialysis tubes with a porous cutoff of 14 kDa (Medicell dialysis tubing; Visking, London, United Kingdom). Dialysis tubes were desalted in accordance with the manufacturer's instructions. Several dialysis tubes were prepared for each water matrix to allow sampling at various time intervals. The sampling periods were 2 weeks in winter and 1 week in summer, due to the degradation of dialysis membranes at higher temperatures. The sampling interval was every 24 h at around 9:00 to 10:00 a.m. The dialysis tubes filled with 100 ml of the water matrices were placed at a depth of 20 to 25 cm from the water surface, in the area where the river water was collected to carry out the different assays. The physical-chemical river water characteristics (temperature, pH, dissolved oxygen [DO] concentration, and conductivity) were recorded during the assays (Table 1), and seasonal differences were also measured. For this purpose, independent assays were performed four times during the summer time (two at high water temperatures of around 29°C and two at around 22°C), once in the spring (at a temperature of around 14°C), and twice during the winter (at a temperature of around 7°C).

TABLE 1.

Values for the physical and chemical parameters measured during the seasonal assays^a

Season and time point	Temp (°C)	DO concn (ppm)	pH	Conductivity (mS/cm)
Summer
1	29.40 ± 1.78	7.23 ± 1.18	8.23 ± 0.16	1.52 ± 0.25
2	29.62 ± 0.41	5.70 ± 0.66	7.84 ± 0.12	1.46 ± 0.19
3	22.33 ± 0.53	3.75 ± 0.34	8.07 ± 0.08	1.54 ± 0.03
4	22.87 ± 0.69	3.53 ± 0.27	7.99 ± 0.14	2.52 ± 0.46
Spring	14.22 ± 0.54	7.30 ± 0.32	8.16 ± 0.08	1.84 ± 0.05
Winter
1	6.29 ± 0.95	8.59 ± 0.11	8.55 ± 0.30	1.28 ± 0.10
2	8.36 ± 0.63	8.70 ± 0.12	8.53 ± 0.09	1.55 ± 0.04

^aValues shown are means ± standard deviations.

Laboratory microcosm assays were performed to evaluate the effect of temperature on B. fragilis and B. thetaiotaomicron inactivation. Nonsterile river water and sterile river water were spiked with each of the type strains. The four different water matrices, (i) B. fragilis-spiked river water, (ii) B. fragilis-spiked sterile river water, (iii) B. thetaiotaomicron-spiked river water, and (iv) B. thetaiotaomicron-spiked sterile river water, were used to fill 250-ml glass recipients that were covered to ensure aseptic conditions but to allow oxygen diffusion. Each set of microcosms was incubated at different temperatures: 4, 10, 22, and 30°C. In addition, a microcosm of raw, unspiked river water was evaluated for each experiment to determine the possible background concentration of Bacteroides species in the environment. Samples were collected at 24-h intervals under sterile conditions.

Enumeration of culturable populations.

Sampling analyses were performed after thorough mixing of the contents of each dialysis tube in the on-site assays or each glass bottle in the microcosm assays and serial dilution in Ringer solution (1/4 strength; Oxoid, Hampshire, England). Culturable Bacteroides species were enumerated in duplicate using two media, a selective medium (Bacteroides fragilis bilis esculin [BBE] medium) (32) and a rich medium (BPRM) with no added antibiotics (53). BPRM enables the recovery of damaged bacteria, which are otherwise difficult to recover using a selective medium under stringent conditions. BPRM allows higher numbers of anaerobic and facultative bacteria to grow. The identity of the putative Bacteroides species in BPRM was confirmed with genus-specific colony hybridization (CH), as described below. Cultures of both media were maintained at 37°C for 48 h under anaerobic conditions (GasPak; BBL, Hampshire, United Kingdom) with CO₂ atmosphere generators (Anaerocult A; Merck, Darmstadt, Germany).

Populations of fecal coliforms (FC) and enterococci (ENT) were enumerated in duplicate by membrane filtration with 0.45-μm-pore-size membranes (Ez-Pak membrane; Millipore). Membranes were transferred to m-FC agar plates (Difco, Le Pont de Claix, France) and to m-Enterococcus agar plates (Difco) for FC and ENT counts, respectively. FC were incubated at 44.5°C for 24 h and ENT at 37°C for 48 h. Enterococcus strains were confirmed by transferring the membrane to bilis esculin agar (Difco) after 1 h of incubation at 44.5°C (3).

Specific colony hybridization of Bacteroides species.

We used colony hybridization with the genus-specific probe Bac303 to differentiate Bacteroides species from the bacteria growing on BPRM. Bacterial colonies were transferred by replica plating to nylon membranes in accordance with the protocol described previously (21). The digoxigenin-labeled Bac303 probe targets the 16S rRNA gene and has the sequence 5′-GAAGGTCCCCCACATTGG-3′ (33).

DNA extraction, PCR, and real-time quantitative PCR.

Molecular detection and enumeration were performed using conventional PCR and qPCR for the on-site experiment samples. Nucleic acids were extracted and purified from 200 μl of each sample by use of a QIAamp DNA blood minikit (Qiagen, GmbH, Hilden, Germany) according to the manufacturer's indications.

Bacteroides genus-specific PCR with Bac32F and Bac708R primers (Table 2) was used to detect the presence/absence of the genus in the samples. The primers and the PCR conditions were the same as those described previously (10). Two microliters of DNA extract was used as a template in each reaction. The PCR product was visualized in 1% agarose gels stained with ethidium bromide.

TABLE 2.

Nucleotide sequences and targets of primers and probes for the real-time PCR assay

Real-time qPCR	Primer or probe	Target	Sequence (5′-3′)a	Reference or source
A	Bac32F	Most Bacteroidaceae and Prevotellaceae spp.	AACGCTAGCTACAGGCTT	11
	Bac708R	Most Bacteroidaceae and Prevotellaceae spp.	CAATCGGAGTTCTTCGTG	11
B	Bact1-F	Most Bacteroidaceae and Prevotellaceae spp.	CCGGGGCTACACACGTGT	This study
	Bact1-R	Many species of the Bacteria domain	CAAGGCCCGGGAACGTAT	This study
	Bact1-S	Most Bacteroidaceae and Prevotellaceae spp.	6-FAM-TCGCGCATCAGCCA-TAMRA	This study
C	AllBact296f	Order Bacteroidales	GAGAGGAAGGTCCCCCAC	30
	AllBact412r	Most Bacteroidaceae and Prevotellaceae spp.	CGCTACTTGGCTGGTTCAG	30
	AllBact375Bhqr	Order Bacteroidales	6-FAM-CCATTGACCAATATTCCTCACTGCTGCCT -BHQ-1	30

^a6-FAM, 6-carboxyfluorescein; TAMRA, 6-carboxytetramethylrhodamine; BHQ-1, Black Hole Quencher 1.

Three real-time qPCRs were evaluated to enumerate Bacteroides type strains and environmental Bacteroidales and to follow the DNA degradation in the environment. The first qPCR was performed using the specific Bacteroidales primers Bac32F and Bac708R (10) with SYBR green (qPCR A). Each 25-μl PCR mixture was composed of 1× SYBR green 10× buffer (Molecular Probes, Eugene, OR), 0.750 U of Taq polymerase, 0.2 mM deoxynucleoside triphosphates (dNTPs), the specific primers at a concentration of 0.2 μM, 1 μl of the DNA extraction, and different MgCl₂ concentrations from 1 mM to 3 mM. The conditions used were an initial denaturation at 94°C for 10 min, followed by 40 cycles at 94°C for 30 s, different annealing temperatures (from 54°C to 61°C) for 30 s, and 72°C for 40 s. After the amplification step, a melting curve analysis was performed to distinguish the targeted PCR product from the nontargeted one. Two qPCRs with hydrolysis probe (TaqMan) chemistry were also tested. The primers and probe for qPCR B were designed using Primer Express software version 2.0 (Applied Biosystems, Foster City, CA) (Table 2). The PCR mixture (20 μl) was composed of TaqMan universal PCR master mix (2×; Applied Biosystems), 0.2 μM each specific primer, 0.1 μM TaqMan probe, and 2 μl of the DNA extraction. For qPCR C, we used the AllBac primers and probe described by Layton et al. (30). In this case, 0.6 μM each specific primer and 0.2 μM TaqMan probe were used (30). The PCR program for both reactions was an initial step of 95°C for 10 min, followed by 50 cycles at 95°C for 30 s and 60°C for 45 s. The analyses were performed using an ABI PRISM 7700 sequence detection system (Applied Biosystems). All the samples were analyzed in triplicate, and 1:10 dilutions of each sample were made to determine any possible inhibition of the reaction by the sample's components.

Standard curves of the cycle threshold (C_T) versus the number of CFU ml⁻¹ were used to calculate the viable-cell numbers of unknown samples. The standard curve was generated from genomic DNA extracted from several dilutions of a B. fragilis DSM 2151^T culture (with cell numbers ranging from 3·10¹ to 3·10⁸ CFU ml⁻¹). The standard curve was added in each qPCR performed to analyze the environmental samples, and the amplification efficiency was calculated from the slope. The qPCR quantification limit was established using the standard curve.

Quantitative PCR detection limit.

The qPCR detection limit was determined using 10-fold dilutions of a pure culture of B. fragilis DSM 2151^T (from 1:1 [corresponding to 200 μl of the culture] to 1:10,000,000 dilutions). The initial number of viable cells was determined using BPRM. The LIVE/DEAD BacLight test (Molecular Probes, Eugene, OR) was used to estimate the number of dead cells in the culture according to the manufacturer's instructions. DNA extractions from the dilutions of each culture were performed using DNA extraction with a QIAamp DNA blood kit (Qiagen). Two microliters of the DNA extracted from each dilution was used to determine the qPCR detection limit.

Inactivation value estimations and statistical comparisons.

The enumeration results from the selective culture media and from qPCR were used to calculate the inactivation kinetics of the culturable populations and of DNA detection for each experimental treatment. Individual measurements of the concentrations at day 0 (t₀) and daily until the end of the assay were log₁₀ transformed. The daily reduction was used to perform regression studies (13). The regression lines of the inactivation value were calculated for the different bacterial populations and for the different assays (summer, winter, and spring). The following equations were used to calculate the decay rates (K_s values) and the amounts of time required for 90% of the initial population to decay (T₉₀ values [h]):

where the N_t value is the cell concentration/ml or the number of 16S rRNA gene copies/ml at time t and the N₀ value is the initial cell concentration/ml or the number of 16S rRNA gene copies/ml at t₀.

Analysis of variance (ANOVA) and the F test were performed using the software program Statgraphics Plus (version 5.1; Rockville, MD) to make comparisons between (i) strains of the two Bacteroides species (B. fragilis and B. thetaiotaomicron), (ii) environmental Bacteroides strains, FC, and ENT, (iii) the presence and absence of predation in the water, and (iv) seasonal differences. Regression analyses were carried out to determine the possible effects of physical and chemical river water factors, such as temperature, pH, dissolved oxygen (DO) concentration, conductivity, and inactivation kinetics.

RESULTS

Sensitivity and amplification efficiency of real-time quantitative PCR methods.

Bac32F/Bac708R PCR was used to evaluate the presence/absence of Bacteroidales in the samples and would also be the best choice for quantifying the DNA decrease. Although the length of the amplicon with these primers may be problematic (∼675 bp) for a qPCR assay, the qPCR was tested to be validated or rejected. A low level of efficiency was achieved with this qPCR assay. Hence, qPCR A was discarded from further analyses. High levels of efficiency of amplification were achieved with qPCR B and qPCR C: 83 to 97% and 93 to 105%, respectively. The correlation coefficients of the different standard curves (r²) were always higher than 0.98. The limit of quantification was lower in qPCR B than in qPCR C (2.8·10³ CFU/ml versus 2.8·10¹ CFU/ml). Consequently, AllBact qPCR (qPCR C) (30) was selected from the evaluated qPCRs to monitor the die-off rate of Bacteroidales DNA in the environment. The detection limit of qPCR C was found to be 2.79·10⁴ CFU/ml.

Persistence of B. fragilis and B. thetaiotaomicron.

The initial populations measured with BBE medium were 1.01·10⁹ ± 0.94·10⁹ CFU 100 ml⁻¹ in the B. fragilis experiments and 1.29·10⁹ ± 0.81·10⁹ CFU 100 ml⁻¹ for B. thetaiotaomicron. The inactivation of B. fragilis following treatment with the selective medium (BBE medium) was faster during the summer than during the winter, with an intermediate value during the spring (Table 3). Significant seasonal differences, with a P value of 0.046, were reported following the use of a simple ANOVA. A regression analysis performed between the environmental factors and the T₉₀ values showed a strong correlation with temperature (P = 0.02; r = −0.88; n = 7). However, no correlation was observed with DO concentration (P = 0.165; n = 6).

TABLE 3.

T₉₀ values obtained from regression lines of inactivation models^a

Medium or method	Season	B. thetaiotaomicron in river water			B. fragilis in river water			B. fragilis in sterile river water
		Mean T90 (h) ± SD	P	Length of DNA detection (h)	Mean T90 (h) ± SD	P	Length of DNA detection (h)	Mean T90 (h) ± SD	P	Length of DNA detection (h)
BBE	Summer	17.9 ± 1.7			13.9 ± 2.3			27.5 ± 5.6
	Spring	15.5			19.7			27.2
	Winter	7.6	0.063		24.9	0.046		29.5	>0.100
BPRM with CH	Summer	47.2 ± 5.0			36.7 ± 15.2			50.1 ± 24.7
	Spring	69.9			129.9			32.4
	Winter	13.6 ± 6.6	0.028		104.7 ± 91.6	>0.100		41.8 ± 18.6	<0.001
qPCR	Summer	12.0 ± 5.7			11.2 ± 1.5			15.2 ± 1.1
	Spring	15.4			18.3			15.5
	Winter	47.3 ± 8.4	<0.001		42.0	<0.001		66.6 ± 10.5	<0.001
Conventional PCR	Summer			24-48			48-72			48-72
	Spring			48			72
	Winter			>216			>216			>216

^aP values represent comparisons among seasons and were calculated using ANOVA.

The B. thetaiotaomicron strain showed the opposite profile of persistence. The decay measured with BBE medium was more gradual during the summer, showing higher T₉₀ values (Table 3). During the winter, a decay rate of 5 log₁₀ was reported in the first 24 h of the experiment and lower T₉₀ values were reported (Table 3). In this case, the seasonal statistical differences were less apparent than those observed in the B. fragilis assay (P = 0.063). However, a correlation between T₉₀ value and temperature was reported by regression analysis (P = 0.089; r = 0.82; n = 5).

The die-off analysis results obtained using the rich medium (BPRM) plus specific colony hybridization confirmed the putative Bacteroides species and showed higher T₉₀ values than the results obtained using BBE medium, although both methods showed the same tendency (Table 3). However, a high number of unspecific colonies grew on BPRM. Hence, this method was not used in further analyses.

The DNA of both strains was detected by PCR for 24 to 72 h in the spring and summer (Table 3). In contrast, the DNA was detected throughout the experiment in the winter. The qPCR results showed similar concentrations of the initial population for both type strains, as did the results obtained using BBE medium when targeted DNA was used to estimate the number of CFU. The average initial concentration of the B. fragilis population was 3.72·10⁹ ± 3.22·10⁹ CFU 100 ml⁻¹, and that of B. thetaiotaomicron was 3.27·10⁹ ± 2.44·10⁹ CFU 100 ml⁻¹. The T₉₀ values were similar for B. fragilis and B. thetaiotaomicron (Table 3). The differences between the results for summer and spring and those for winter for each type strain were statistically significant (P < 0.0001).

Persistence of B. fragilis in sterile river water.

To evaluate the effect of grazing predators on inactivation, sterile river water samples were spiked with B. fragilis and differences from B. fragilis in nonsterile river water were studied. The average initial population among the different seasonal experiments was 4.78·10⁹ ± 7.84·10⁹ CFU 100 ml⁻¹. When predators were removed from the river water by autoclave sterilization, nonseasonal differences in the decay rates of B. fragilis monitored were reported by a simple ANOVA (P = 0.946) (Table 3). However, the lowest T₉₀ value was obtained during the first summer (21.32 h), when high temperatures coincided with high DO concentrations (Table 3). The comparison between sterile and nonsterile river water samples by the F test revealed significant differences between matrices (P = 0.003). No correlation between T₉₀ values and temperature or DO concentration was reported. However, a multiple regression showed that the T₉₀ values interacted with both factors (for temperature, P = 0.086; for DO, P = 0.089; n = 5).

Although the T₉₀ values were higher when inactivation was detected using BPRM and confirmation by CH, the same behavior was reported (Table 3).

Conventional PCR detected the presence of Bacteroides DNA for the first 48 to 72 h during the summer and spring experiments and for the whole period during the winter assay for both water matrices. The qPCR technique detected an initial population of 4.13·10⁹ ± 3.22·10⁹ CFU 100 ml⁻¹. Similar T₉₀ values were obtained during the summer and spring and higher values during winter (Table 3). Seasonal differences between the results for summer and spring and those for winter were reported (P < 0.0001).

Persistence of environmental Bacteroides strains.

The initial environmental Bacteroides population obtained from sewage and enumerated with BBE medium was higher in the winter (2.83·10⁵ ± 0.90·10⁵ CFU 100 ml⁻¹) than in the summer (5.38·10⁴ ± 4.19·10⁴ CFU 100 ml⁻¹) (Table 4). However, lower T₉₀ values were registered in the winter than in the summer, and an intermediate value was observed during the spring (Table 4). During the first summer, a T₉₀ value of 20.0 h was reported and the observed inactivation rate was closer to the low values obtained in the winter. When this value was excluded from the analysis due to the unusually high temperature that year, seasonal differences were reported, with a P value of 0.088. A correlation between the T₉₀ values obtained with BBE medium and the DO in the river water was reported by simple regression (P = 0.0598; r = −0.73; n = 7). Inactivation was also evaluated using enumeration on BPRM and colony hybridization. However, other bacterial species were shown to interfere with this technique (data not shown).

TABLE 4.

T₉₀ values obtained from regression lines of inactivation models^a

Season	Culturable FC		Fecal ENT		BBEb
	Mean T90 (h) ± SD	P	Mean T90 (h) ± SD	P	Mean T90 (h) ± SD	P
Summer	118.9 ± 7.1		27.9 ± 3.6		41.4 ± 11.0
Spring	92.6		54.9		27.6
Winter	88.1 ± 0.6	0.0193	116.1 ± 10.4	0.0002	21.3 ± 0.9	0.0881

^aP values represent comparisons among seasons and were calculated using ANOVA.

^bEnvironmental Bacteroides spp. measured with BBE medium.

The fecal indicators FC and ENT persisted longer than Bacteroides species, regardless of the season (Table 4). For FC, the highest T₉₀ values were reported for summer; slightly lower values were obtained in spring and winter (Table 4). The opposite survival pattern was detected for ENT, which persisted longer in winter than in summer (Table 4). Significant seasonal differences were reported, with P values of 0.0193 for FC and 0.0002 for ENT. A strong correlation between the two bacterial fecal indicators and the temperature was reported by linear regression (for FC [P = 0.0104; r = 0.92; n = 6] and ENT [P = 0.007; r = −0.89; n = 7]).

PCR detection of Bacteroidales with specific primers was negative for all the samples. Quantification with qPCR showed an initial population close to the detection limit of the technique (2.8·10⁴ CFU/ml). Therefore, no reliable values were obtained for the following days.

Microcosm assays.

When the persistence of Bacteroides species was analyzed under controlled conditions, both species showed the highest inactivation rate at the highest incubation temperature. When BBE medium was used to analyze B. fragilis in river water at 30°C, a lag of 48 h was observed before the reduction of the initial population began (Fig. 1). A decrease of 6 log₁₀ was reached after 192 h of the experiment. The microcosm at 22°C showed a longer lag phase and a lower level of reduction. When the microcosms were maintained at 10°C or 4°C, no significant decrease occurred. When the survival of B. fragilis was analyzed in sterile river water a significant decrease was reported only in the 30°C incubation experiment. No substantial reduction was reported when the microcosms were kept at 22, 10 and 4°C (Fig. 1).

FIG. 1.

Survival of B. fragilis DSM 2151^T studied in microcosm experiments using two water matrices: river water monitored with BBE medium (a) and BPRM (b) and sterile river water monitored with BBE (c) and BPRM (d). Results are indicated as logarithmic values for the decreases of the measured populations at different incubation temperatures: 4°C (•), 10°C (○), 22°C (▾), and 30°C (▵).

Parallel analyses were carried out using BPRM. A reduction of the initial population was only observed in the microcosms incubated at 30°C. The decrease was 1 log₁₀ after 192 h of the experiment.

B. thetaiotaomicron showed similar behavior to B. fragilis (Fig. 2). The microcosm of B. thetaiotaomicron in river water incubated at 30°C showed a lag phase of 24 h before the start of the reduction in the initial population, as measured with BBE medium. A reduction of 5.5 log₁₀ was detected after 168 h of experiment. At 22°C the decrease reached 4.2 log₁₀ after 192 h. No significant decrease was observed at 10 and 4°C. When sterilized river water was used, very slight decreases in the microcosms at 30°C and 22°C were reported. No reduction was observed in the microcosms kept at 10 and 4°C. A slight reduction was observed through the utilization of the rich medium (BPRM) in nonsterile river water at 30°C and 22°C and sterile river water at 30°C. No significant reduction was observed at the other incubation temperatures.

FIG. 2.

Survival of B. thetaiotaomicron DSM 2079^T studied in microcosm experiments using two water matrices: river water monitored with BBE (a) and BPRM (b) and sterile river water monitored with BBE medium (c) and BPRM (d). Results are indicated as logarithmic values for the decreases of the measured populations at different incubation temperatures: 4°C (•), 10°C (○), 22°C (▾), and 30°C (▵).

A negative-control experiment using river water was performed in parallel with the microcosm experiment. Non-Bacteroidales colonies were detected using BBE medium in negative-control experiments using river water at the different temperatures.

DISCUSSION

The use of Bacteroidales as markers to detect the possible source of fecal pollution has been widely applied, showing a broad geographical distribution (1, 20, 22, 46). In order to develop proper models to detect the source of fecal pollution, the survival and detection of the markers must first be described. Although there is increasing research done to understand the survival of Bacteroidales in the environment (4, 55, 57), there are still many gaps in our knowledge.

As none of the current Bacteroidales markers have been cultivated yet, two of the main Bacteroides species in the human gut (B. fragilis and B. thetaiotaomicron) were used in this study to simulate the survival pattern observed when they are released in the environment. The persistence of both Bacteroides species was compared with that of environmental Bacteroides strains from sewage and fecal bacterial indicators. The linear decay measurement used to calculate and compare the inactivation rates was as described elsewhere (13, 57). Nevertheless, mathematical functions other than linear decay measurement may better fit the observed kinetics.

Although B. fragilis and B. thetaiotaomicron belong to the same genus, their culturable cells showed different survival patterns over the seasons when experiments were performed on-site. B. fragilis persisted longer during winter, whereas B. thetaiotaomicron did so in summer. B. fragilis might be more susceptible to higher temperatures, and B. thetaiotaomicron survival might be more related to dissolved oxygen (DO) concentration in water. DO concentration in water depends on physical parameters, such as temperature and atmospheric pressure, and on biological parameters, such as photosynthetic and respiratory activities of organisms in water. Higher DO concentration in water is associated with lower temperature (31). Therefore, differential resistance to oxygen concentration could explain the variation in seasonal decay. In this case, B. thetaiotaomicron was less tolerant to aerobic conditions than B. fragilis. B. fragilis is the most frequently isolated organism from intra-abdominal infections. It can survive for 48 to 72 h under aerobic conditions, although it grows in and can benefit from oxygen at nanomolar concentrations (7, 44, 52). In addition, two enzymes for detoxifying reactive oxygen species have been described to occur in B. fragilis: catalase and superoxide bismutase. Only catalase has been detected in B. thetaiotaomicron (59). Consequently, the higher level of resistance to aerobic conditions shown by one of the two species could partially explain the differences in the kinetics of die-off.

Environmental Bacteroides strains from sewage survived better than the laboratory strains studied. This indicates that environmental strains were better adapted to selective pressure induced by environmental factors. These environmental strains persisted better in the summer, when higher temperatures and lower DO concentrations were detected in water. This suggests that environmental strains may be more sensitive to DO concentration than B. fragilis laboratory strains. Variance in survival rates among different environmental Bacteroides spp. and B. fragilis in aerobic environments and also different responses to sunlight irradiation have already been reported (48, 55). Sunlight inactivation increases the reactive oxygen species in the environment. Therefore, species with different levels of aerotolerance can be affected to various extents.

A comparison of the levels of B. fragilis persistence in river water and sterile river water on-site confirmed the role of indigenous microorganisms in Bacteroides persistence (9, 40, 47). A reduced seasonal effect on B. fragilis survival was detected when autoclaved water was used, which implies that the highest level of predation activity occurred under warmer conditions. The same effect was observed in microcosm assays: there was a higher level of inactivation in nonsterile water matrices and at a higher incubation temperature.

When predators were removed from the water matrix, it was easier to evaluate the role of physical and chemical parameters in the persistence of B. fragilis. Our results suggest that temperature has a synergetic effect and DO modulates the survival of Bacteroides spp. In the presence of higher oxygen concentrations (6.2 to 9.4 ppm) and high water temperatures (27.3°C to 32.6°C), the persistence of cultivable B. fragilis cells decreases significantly. Therefore, the lower survival rate of B. fragilis at higher temperatures is the combined effect of the temperature and the higher level of predation activity accentuated by soaring DO levels in water. In the environment, many physical, chemical, and biological parameters can interact to modulate population survival, in this case, temperature, DO concentration, and predation. These parameters interact, making it difficult to distinguish the role of each one. A holistic view should be taken, rather than a reduction and simplification of the effects, when ecological studies are performed.

For instance, the periods of persistence of B. fragilis and B. thetaiotaomicron in microcosms under controlled conditions were longer than the periods observed in the environment. The higher rates of survival of both species were linked to cooler conditions, whereas on-site experiments showed that B. thetaiotaomicron survived better in warmer seasons. With reliance on microcosm experiments, temperature, followed by predation, might be considered the main factor affecting the survival of both species. Generally, studies analyzing the persistence of Bacteroidales in the environment are performed using microcosms and mesocosms with controlled environmental parameters (9, 40). Therefore, the use of these experimental conditions can lead researchers to misinterpret the environmental pattern of bacterial survival. However, all the various environmental factors acting in on-site experiments are difficult to monitor, and replication of the experiments with the same conditions is difficult. Thus, the combination of the two kinds of experiments is important for understanding Bacteroides survival in the environment.

Higher bacterial numbers were recorded when the rich medium (BPRM) was used instead of the selective medium (BBE medium). Damaged bacterial cells can become viable but not cultivable (VBNC) under stressful conditions. A selective medium with astringent conditions hampers their recovery. The damaged cells may be easier to recover in a rich medium. However, nonspecific detection was reported with the methodology used, which can interfere with the interpretation of the results.

By using PCR and qPCR, we were able to detect and quantify the total Bacteroides cells. Although the initial cell counts were similar following the use of BBE medium and quantification by qPCR, the decay rates were different. DNA was detected for longer periods during winter than during summer, suggesting a strong correlation between temperature and DNA degradation regardless of the Bacteroides species. In contrast, cultivable B. thetaiotaomicron and environmental strains decayed faster in winter during on-site experiments. The molecular techniques used in this study could not distinguish between cultivable, viable, and nonviable cells. Therefore, we cannot discern whether temperature affects the viability or culturability of the cells. However, the recent use of propidium monoazide when DNA is extracted may enable us to differentiate intact (live) cells from dead cells or extracellular DNA (4). The variability in DNA detection between summer and winter has to be considered for the detection of Bacteroides species as MST indicators by PCR and qPCR. The fast DNA degradation under high-temperature conditions may lead to misdetection of the marker when fecal pathogens are still viable, or, in contrast, the persistence of the DNA at lower temperatures may lead to an overestimation of the risk. Studies examining the correlation between Bacteroides species and fecal pathogens should be performed under different environmental conditions.

In this study, differences were observed between the detection and quantification thresholds, which could be due to the limitations of the DNA extraction protocol that occur when high or low DNA concentrations of bacteria are extracted, as reported by others (17).

Bacteroides spp. from sewage showed higher die-off rates than fecal coliforms and enterococci, regardless of the season. Hence, Bacteroides spp. are indicative of recent fecal pollution. The persistence of molecular targets for Bacteroides spp., and perhaps for other intestinal bacterial species within the Bacteroidales group, could limit the application of these species in predictive MST models. However, higher correlations between detection and persistence of enteric pathogens and Bacteroidales than for some fecal indicator bacterial groups (49, 56, 57).

Bacteroides genus persistence in the environment is a complex issue, where many factors and interactions have to be considered. Experimental designs with artificial conditions might mask the effect of the environmental factors and their interactions, but on the other hand, on-site experiments are demanding and difficult to reproduce. The Bacteroidales group includes a large number of species with different responses to environmental stresses. Different host-associated probes and primers in this taxonomical group, where many included species have not yet been identified, have been proposed as MST markers. Therefore, the different survival rates among species included within Bacteroidales have to be considered in order to develop predictive models to detect the source of the fecal pollution that occurs when these molecular methods are used. The period of persistence of these bacterial populations in environmental water was very short; an analysis using improved qPCR methods could provide longer detection periods, especially in winter, when water temperatures are lower. However, the inability to discern between viable, VBNC, and dead cells could be solved by the combined use of propidium monoazide in the DNA extraction. The persistence of MST indicators is an essential factor for their reliability when they are applied in environmental case studies. Consequently, the environmental persistence of any suggested MST marker, together with specificity and sensitivity, should be included when proposed.