The ISME Journal (2014) 8, 1381–1390

& 2014 International Society for Microbial Ecology All rights reserved 1751-7362/14
www.nature.com/ismej

ORIGINAL ARTICLE
Wastewater as a point source of antibiotic-
resistance genes in the sediment of a freshwater
lake
Nadine Czekalski1, Elena Gascón Dı́ez2 and Helmut Bürgmann1
1
Department of Surface Waters—Research and Management, Eawag: Swiss Federal Institute of Aquatic
Science and Technology, Kastanienbaum, Switzerland and 2Group of Limnology and Environmental Geology,
Institut F.-A. Forel, University of Geneva, Versoix, Switzerland

Antibiotic-resistance genes (ARGs) are currently discussed as emerging environmental contami-

nants. Hospital and municipal sewage are important sources of ARGs for the receiving freshwater
bodies. We investigated the spatial distribution of different ARGs (sul1, sul2, tet(B), tet(M), tet(W)
and qnrA) in freshwater lake sediments in the vicinity of a point source of treated wastewater. ARG
contamination of Vidy Bay, Lake Geneva, Switzerland was quantified using real-time PCR and
compared with total mercury (THg), a frequently particle-bound inorganic contaminant with known
natural background levels. Two-dimensional mapping of the investigated contaminants in lake
sediments with geostatistical tools revealed total and relative abundance of ARGs in close proximity
of the sewage discharge point were up to 200-fold above levels measured at a remote reference site
(center of the lake) and decreased exponentially with distance. Similar trends were observed in the
spatial distribution of different ARGs, whereas distributions of ARGs and THg were only moderately
correlated, indicating differences in the transport and fate of these pollutants or additional sources
of ARG contamination. The spatial pattern of ARG contamination and supporting data suggest that
deposition of particle-associated wastewater bacteria rather than co-selection by, for example,
heavy metals was the main cause of sediment ARG contamination.
The ISME Journal (2014) 8, 1381–1390; doi:10.1038/ismej.2014.8; published online 6 March 2014
Subject Category: Microbial population and community ecology
Keywords: antibiotic-resistance genes; qPCR; transport; 2-D mapping; aquatic; environment

Introduction resistance traits from the environment via horizontal

gene transfer (Aminov, 2009; Martinez, 2009a).
Since their introduction into medical practice in the Moreover, various studies have demonstrated that
1930s, antibiotics have saved millions of lives, but wastewater and animal waste contain large numbers
the increasing numbers of bacterial pathogens that of resistant bacteria, which can pass through the
are becoming resistant to antibiotics is a growing wastewater treatment plant (WWTP) and reach the
cause of concern. Research initially focused on receiving water bodies (Teuber, 2001; Kümmerer,
clinical settings as the main site of the spread and 2004; Jury et al., 2011; Rizzo et al., 2013; Zurfluh
evolution of antibiotic resistance. Over the course of et al., 2013). There is growing concern that con-
the past two decades, however, researchers have tinuous discharge of these contaminants may, at
increasingly broadened their focus to include the least locally, lead to an increase in the natural
environment as a source of resistance genes and as a resistance background levels and thus enhance the
site of antibiotic-resistance evolution (Kümmerer, likelihood of antibiotic-resistance genes (ARGs)
2004). The existence of a natural environmental being transferred back to human and animal com-
resistance gene pool, which includes all known mensals or even pathogens (Kim and Aga, 2007;
clinical resistance mechanisms, has been discovered Martinez, 2009b; Cantas et al., 2013). In the long
(D’Costa et al., 2006; Aminov, 2009). Consequently, term, this ‘environmental loop’ could contribute to
bacterial pathogens are assumed to acquire the spread of resistance in pathogens and under-
mine the effectiveness of current and future
Correspondence: H Bürgmann, Department of Surface Waters— antibiotics.
Research and Management, Eawag: Swiss Federal Institute of Lake Geneva is the largest freshwater body in
Aquatic Science and Technology, Seestrasse 79, Kastanienbaum Western Europe and receives treated wastewaters
6047, Switzerland.
E-mail: [email protected] from the surrounding cities. The largest WWTP is
Received 7 August 2013; revised 18 December 2013; accepted 1 located in Lausanne and discharges treated waste-
January 2014; published online 6 March 2014 water and occasionally bypass water into Vidy Bay.
 Antibiotic-resistance genes in lake sediments
N Czekalski et al
1382
The bay’s sediments have been reported to be was carried out using a geographic information
heavily polluted by fecal-indicator bacteria, heavy system (ArcGIS). Further data on the microbial
metals, nitrogen and phosphorus (Poté et al., 2008). community structure in wastewater and sediment
Micropollutants, including antibiotics (Bonvin samples and sediment properties were considered
et al., 2011), as well as multiresistant bacteria and in order to interpret the observed spatial trends of
ARGs pass through the WWTP and are subsequently ARGs in Vidy bay sediments.
discharged in considerable amounts into Vidy Bay
(Czekalski et al., 2012). Low levels of multiresistant
bacteria and ARGs were found even 3.2 km from the Materials and methods
WWTP discharge pipe, where lake water is pumped
for drinking water production (Czekalski et al., Study site description and sampling campaigns
2012). However, it remained unclear whether For a detailed description of the Vidy Bay study site,
this represented contamination or the (unknown) please refer to Czekalski et al. (2012). Lausanne
natural resistance background level of Lake Geneva. (population 214 000) is the biggest city in the
Molecular tools, such as PCR and quantitative Lake Geneva area. The city has no pharmaceutical
real-time PCR, have been developed for many ARGs industries but several health-care centers.
(Heuer and Smalla, 2007; Walsh et al., 2011). The largest is the Centre Hospitalier Universitaire
A limited number of studies have applied molecular Vaudois, an important site of antibiotic consump-
methods to aquatic environments. Primarily, rivers tion, which discharges B410 m3 day  1 of raw waste-
impacted by different contamination sources water into the municipal sewage system (Blanc,
have been considered (Pei et al., 2006; Pruden 2010). The conventional WWTP of Lausanne treats
et al., 2006; Graham et al., 2010). In these systems, the bulk of the wastewater it receives with a
the horizontal transport of contaminants is direc- chemical phosphate removal process and an aerobic
tional and constrained by the river channel and biological treatment, followed by clarification.
therefore relatively well defined. This is also true for Sludge is recycled or incinerated. Only part of the
most of the sampling points in the study of LaPara water receives a more elaborate biological denitrifi-
et al. (2011), which followed the input of ARGs with cation treatment (Supplementary Figure S1). Disin-
wastewater into the St Louis river and transport into fection is not performed. Treated wastewater and
Lake Superior. In freshwater lakes, complex trans- occasionally untreated sewer overflow (bypass) is
port processes prevail, dominated, for example, by discharged at an average rate of 86.631 m3 day  1 (1–
temporally variable wind regimes, currents and 3 m3 s  1 up to 5-6 m3 s  1 at low and peak flow,
stratification. Despite the high relevance of lakes as respectively) (Vioget et al., 2011). The effluent pipe
sources of drinking water, food and as a place for discharges the treated water 750 m offshore at a
recreational activities, the dynamics of antibiotic- depth of 35 m into the Vidy Bay of Lake Geneva
resistance contamination in lakes that are receiving (Figure 1, sampling point STEP (‘station d’épura-
waters for municipal wastewater has not been tion’)). The rivers Chamberonne and Flon (sampling
thoroughly investigated. It is currently unknown, points CHAM and FLON) form part of the city’s
for example, at which spatial scales a local source of combined sewer overflow system and also discharge
antibiotic-resistant bacteria affects water and sedi- into the bay (1-3 m3 s  1, Goldscheider et al., 2007).
ment in a receiving lake. The main form of agricultural land use in their
The present study aims to characterize the spatial catchment areas is vineyards. Intensive animal
impact of Lausanne’s WWTP discharge on the farming is not prevalent. About 3.2 km south-west
concentration of several plasmid-mediated ARGs from STEP, lake water is pumped for drinking water
in the lake sediment. The tested ARGs confer production (sampling point DWP).
resistance to different classes of broad-spectrum Short sediment cores were taken at 22 sites during
antibiotics (sulfonamides: sul1, sul2, tetracyclines: two sampling campaigns in August 2011 (Figure 1).
tet(M), tet(B), tet(W)) and fluoroquinolones: qnrA), Sediment cores were retrieved as described pre-
which differ in their origin (for example, natural viously (Czekalski et al., 2012). Sediment cores were
versus synthetic), time of introduction, resistance transported to the laboratory within 8 h and stored
mechanisms and clinical relevance. Sulfonamides at 4 1C for a maximum of 48 h until processing.
were the first synthetic antibacterials. Tetracyclines, Three subsamples of 0.5 g (wet weight) from the
isolated from Streptomyces, followed in 1948. sediment surface layer (0–3 cm) of each core were
Fluoroquinolones were introduced during the transferred to sterile 2-ml screw cap tubes. After
1970s and are, of the three substance classes, shock-freezing in liquid nitrogen, the tubes were
currently the most frequently used in clinics. ARGs stored at  80 1C until DNA extraction. For a subset
were compared with an already well-studied of sites, additional subsamples of surface sediments
WWTP-released contaminant, total mercury (THg), were taken, immediately preserved by freezing at
for which background levels in Lake Geneva are  20 1C and freeze-dried to analyze THg, grain size
known from previous studies (Arboille et al., 1989; and organic matter content.
Poté et al., 2008). Mapping of the two-dimensional Details on additional water samples taken in 2009
spatial pattern of the investigated ARGs and THg and 2010 from Lausanne’s sewage system and lake

The ISME Journal

 Antibiotic-resistance genes in lake sediments
N Czekalski et al
1383

SHL2
WTP
Lake Geneva

CHAM
© Atlas der Schweiz 2.0 7 km NC14 EG1 NC15
N N
NC5 NC2
N
NC1
4 STEP
NC6 NC4
NC7
EG2 NC3 FLON

EG3 EG14
NC9 EG4

NC11
NC13- EG6
DWP
EG15

1 km
EG9

Figure 1 Map of Vidy Bay, indicating its location in Lake Geneva, 22 sites from which sediments were sampled, and the two river
mouths of the Flon and Chamberonne (CHAM). The WWTP discharge pipe is indicated by a dashed line starting at Lausanne’s WWTP to
the point of discharge, the STEP site. Bathymetric data provided by Anh-Dao Le thi and Walter Wildi, Institut F.A. Forel. Maps used with
permission of swisstopo (Art. 30 GeoIV): PK50r2007, 2005rswisstopo.

water in Vidy Bay that were analyzed by plate (Heuer and Smalla, 2007; Heuer et al., 2008). For
counts of resistant bacteria and qPCR assays are tet(B), tet(M), tet(W) and qnrA, standard curves were
given in the Supplementary Information. prepared from cloned PCR products as described for
the 16S rRNA gene in Czekalski et al. (2012). qPCR
assays for sul, tet and 16S rRNA genes were
DNA extraction performed in 96-well plates using TaqMan Environ-
DNA from the three subsamples from each sampling mental Master Mix 2.0 (Life Technologies Corp.,
site was extracted, quantified and quality-controlled Applied Biosystems, Carlsbad, CA, USA), which is
as described in Czekalski et al. (2012). Briefly, cells optimized for samples with confirmed or expected
were lysed using bead-beating and freeze–thaw presence of PCR inhibitors. qnrA was amplified using
cycles, and DNA was purified using a phenol– QuantiTect SYBR Green PCR chemistry (Qiagen,
chloroform extraction protocol. Extraction blanks Hombrechtikon, Switzerland). In brief: each 25-ml
omitting sediment were prepared to check for reaction mixture contained 1  of either Environ-
contamination sources. DNA quality was checked mental or SYBR green Master Mix, primers, probe and
via agaose gel electrophoresis and ultraviolet-absor- MgCl2 according to the published methods
bance (Nanodrop, Thermo Fisher Scientific Inc., (Supplementary Table S1) and 5 ml of template
NanoDrop products, Wilmington, DE, USA) and DNA or standard. All qPCR assays were performed
quantified using the Quant-iT PicoGreen DNA in technical triplicates on each extract, standard and
quantification kit (Invitrogen, Basel, Switzerland). negative controls (no template control: salmon
Extraction blanks yielded negative results. DNA sperm DNA, 10 ng ml  1 in order to control for
pellets were resuspended in 50–100 ml of Tris–EDTA specificity; PCR blanks: nuclease-free water; and
buffer and stored at  20 1C until further analysis. extraction blanks: see above). qPCR reactions were
performed on a 7500 Fast Real-Time PCR System
(Applied Biosystems) and analyzed using default
qPCR assays performed with sediment, lake and settings.
wastewater DNA extracts All negative controls resulted either in no ampli-
Six different ARGs (sul1, sul2, tet(B), tet(M), tet(W) fication or a higher threshold cycle (Ct) than the
and qnrA) and bacterial 16S rRNA gene fragments most diluted quantification standard. A sample was
were quantified using the published protocols, considered to be below limit of detection (LOD) for a
primers and probes (Supplementary Table S1). target gene if X2 out of 3 technical replicates were
Standard curves were prepared from serial 10-fold negative or if sample Ct values were XCt of negative
dilutions of plasmid DNA containing the respective controls. Samples above LOD were considered to be
target gene in a range of 3  106 to 30 gene copies. The below the limit of quantification when the s.d. of Ct
preparation of control plasmids and standard values of methodological triplicates was 40.5 and
curves for sul1 and sul2 was as described previously their Ct value was higher than the Ct of the most

The ISME Journal

 Antibiotic-resistance genes in lake sediments
N Czekalski et al
1384
diluted standard whose standard deviation of Ct Sediment characterization and nutrient content
values was p0.5. The sediment grain size was measured using a
For each PCR assay, the qPCR efficiency was Coulter LS-100 laser diffractometer (Beckman
calculated from the slope of the standard curve Coulter, Fullerton, CA, USA), following 5-min
(E ¼ 10(  1/slope); Supplementary Table S2). For a ultrasonic dispersal in de-ionized water according
subset of seven samples and for one target gene to the method of Loizeau et al. (1994).
(sul2), the effect of inhibitors in the sample matrix The sediment total organic matter content was
on amplification in qPCR assays was evaluated estimated from loss on ignition at 550 1C for 1 h in a
directly. Each of the samples was spiked with 105 laboratory oven (LE14/11, Nabertherm, Lilienthal,
copies of quantification standard DNA and ampli- Germany).
fied together with the same set of non-spiked
samples and control DNA. The presence of inhibi- Data analysis, visualization and statistics
tors was evaluated according to Pei et al. (2006) by Total copy numbers of all target genes were norma-
calculating the suppression factor S. S varied lized to sediment sample wet weight (g-1 wwt). This
between 1.1 and 1.4, indicating that inhibition was is referred to as ARG concentration from here on.
negligible (Supplementary Table S3). Additionally, ARG copy numbers were normalized
to 16S rRNA gene copy numbers and reported as
percentages, as an indicator of the relative abun-
Automated ribosomal intergenic spacer analysis dance of resistance genes within the bacterial
(ARISA)-PCR and sequencing population. The number copies of the 16S rRNA
In all, 1:10 dilutions of DNA extracts were subjected genes per cell varies among bacterial species and
to ARISA using general bacterial primers may thus vary between samples, but keeping this
(Supplementary Table S1) and PCR conditions caveat in mind we refer to this measure as ‘ARG
according to Yannarell et al. (2003) with slight abundance’ from here on. Interpolation maps were
modifications (Bürgmann et al., 2011). A total of created with ArcGIS (ESRIArcMap 10, Redlands,
10 ng ml  1 of Escherichia coli DNA served as a CA, USA) with the Geostatistical Analyst extension,
positive control, 10 ng ml  1 of salmon sperm DNA using the Inverse Distance Weighted interpolation
and nuclease-free water served as no template algorithm with power functions from 1 to 10 and a
control and PCR blank, respectively. smoothing factor of 0.2.
Samples were run on an ABI 3130xl capillary In order to evaluate similarity of the distribution
sequencer as described previously (Bürgmann et al., patterns between resistance genes and between
2011). Data analysis was performed using Gene resistance genes and Hg concentrations, as well as
Mapper Software version 4.0 (Applied Biosystems) between ARGs and distance from the contamination
considering only peaks with sizes between 350 and source, correlation and regression analysis were
1250 bp and a minimum peak height of 125 performed in MS Excel 2010 (Microsoft, Redmond,
fluorescence units. ARISA data were subjected to WA, USA), using the Data Analysis Tool.
the automatic and interactive binning scripts Community similarity analysis was performed
(Ramette, 2009) for R (version 2.14.2, (2012)) using based on Hellinger-transformed ARISA peak area data
a window size of 2.5 bp. in R version 2.14.2 (R Core Team, 2012), using the
BiodiversityR and vegan packages (Oksanen et al.,
2012; Kindt and Coe, 2005). Principle coordinates
Hg analysis analysis was performed using the Bray distance
THg concentrations were determined by cold vapor metric. ARG abundance and environmental variables
atomic absorption spectrophotometry after dry were passively fitted to community principle coordi-
mineralization and pre-concentration of Hg by nates analysis data, and significance of fit was
means of amalgamation on a gold trap (Száková analyzed by means of permutation analysis (n ¼ 1000).
et al., 2004), using an automatic mercury analyzer
(Model AMA 254, Altec, Prague, Czech Republic) Results and discussion
under the following conditions: typical sample mass
50–100 mg, drying time 45 s, decomposition time Spatial trends in ARG abundance
150 s and waiting time (necessary for quantitative A central goal of this study was to characterize the
trapping of released mercury on the gold amalga- spatial impact of Lausanne’s WWTP discharge pipe
mator) 45 s. All analyses were run in duplicates. on the contamination of Vidy Bay sediments with
The relative error was usually±5% and always different ARGs. sul genes were detected at all and
under±10% (Roos-Barraclough et al., 2002). The tet genes at several of the sampled locations, whereas
limit of detection is 0.01 ng, and the working qnrA was always below detection. ARG abundance
range is between 0.05 to 600 ng (Yang and Pan, generally decreased from sul1 over sul2, tet(w),
2007). Concentrations obtained for repeated tet(M) to tet(B) (Figure 2). The prevalence of ARGs
analyses of certified reference materials never in Vidy bay sediments thus reflects their abundance
exceeded the published range of concentration in Lausanne’s wastewater: qnrA was detected in
(0.091±0.008 mg g  1 for MESS-3). low amounts and only in hospital sewage, whereas

The ISME Journal

 Antibiotic-resistance genes in lake sediments
N Czekalski et al
1385
1.E+01
sul1 rB  0.17 for ARG concentration and r B  0.36 for
sul2 ARG abundance, Supplementary Tables S4 and S5),
1.E+00
tet(B) but ARG abundance followed an exponential decay
1.E-01 R² = 0.27
tet(M) function (Figure 2). The decreasing trend appeared
tet(W)
to be intact across the entire range of distances
THg
sampled. ARG concentration trends were similar as
ARG abundance

1.E-02 sul1
sul2 16S rRNA gene copy numbers did not show a clear
1.E-03 tet(B) distance-related trend (Supplementary Figure S4).
R² = 0.54 tet(M) However, for each ARG, the R2 was higher for
1.E-04 R² = 0.58 tet(W) abundance data compared with concentration data.
THg
A possible explanation would be sample-to-sample
1.E-05
variations in PCR inhibition in spite of our best
R² = 0.48 efforts to minimize such effects. We concluded that
1.E-06
R² = 0.57
R² = 0.68
ARG abundance is slightly more robust for the
1.E-07
purpose of our analysis. The R2 for ARG abundance
0.01 0.1 1 10 varied between 0.48 and 0.68, indicating that
distance from STEP (km) considerable amounts of variation are not explained
Figure 2 Log–log plot of ARG abundance (ARG copy numbers by distance (Figure 2).
normalized to bacterial 16S rRNA gene copy numbers) versus We therefore used spatial mapping to investigate
distance from the STEP site. STEP was arbitrarily given a distance whether deposition patterns deviate from a simple
of ½ the distance to the next closest site to allow it to be included
in the analysis. Lines with R2 values are for fitted functions of
distance relationship, indicating directional transport
type y ¼ a  xb. (Figure 3 and Supplementary Figure S5). The maps
clearly show the accumulation of ARGs around the
WWTP discharge pipe. Designating a sul1 abundance
sul and tet were more abundant in both WWTP 41% as ‘contaminated’, Figure 3a indicates an area of
influent and effluent samples, with sul1 abun- at least B0.3 km2 has been impacted. All sampling
dance always being the highest (Supplementary points within this area are hereafter referred to as
Figure S5). This abundance pattern may be related ARG-contaminated sediments.
to the longer history of clinical application of ARG maps indicated a strong decrease in ARG
sulfonamides and tetracyclines. sul and tet genes abundance and concentration towards the deep lake
have previously been found to be quite abundant (EG2 to EG9) and towards the south east (sampling
in sewage and receiving waters (Pruden et al., locations EG14 and EG15). Ct values obtained for EG6,
2006) and in livestock waste sludge (Zhang et al., EG9, EG15 and the center of the lake (reference point
2013). Plasmid-mediated quinolone resistance SHL2) were either below the LOQ (sul genes) or even
encoded by qnrA was first reported in 1998 below LOD (tet genes). In contrast, our data indicated
(Martinez-Martinez et al., 1998), which may a less dramatic decrease in ARG levels along the
explain its low abundance in Lausanne’s wastewater transect following the shore line in south-western
system. Other quinolone-resistance mechanisms were direction (NC5–NC13, Figures 2a–e and Figures 3a–e).
likewise infrequent, as shown by low colony counts Finally, apart from site NC14 (close to the
on Ofloxacin- and Norfloxacin-amended agar plates Chamberonne river—CHAM), samples taken close
inoculated with samples from Lausanne’s sewage to shore (NC15, EG1) showed low levels of ARGs.
and wastewater treatment system (Supplementary These observations indicate a directionality of the
Figures S6 and S7). pollutant transport in the water column of Vidy Bay.
For all the detected ARGs, highest gene concen- Hydrodynamic transport in lakes is often complex.
trations and abundances occurred at sites in close Currents are influenced by changing wind regimes,
proximity to the WWTP discharge (Figure 2), espe- stratification and bathymetry and vary in direction
cially at sites STEP, EG2 and NC4 (Supplementary and speed with depth (Wüest and Lorke, 2003;
Figures S2 and S3). At site EG2, a maximum of Righetti et al., 2011). For Vidy Bay, it has recently
2.2  109 (sul1) and of 1.5  106 (tet(B)) ARG been shown that near-shore gyres can occur (which
copies g  1 wwt were determined (Supplementary enhance the retention time of water and pollutants
Figure S2). Abundance of sul1 at this site reached in the bay) and that neither of two dominant wind
12% (Supplementary Figure S3). ARG concentration regimes produce a consistent current pattern (Razmi
and abundance of both sul and tet genes were also et al., 2013). However, westward shore-parallel
high at sites NC1, NC2, NC3, NC5 and EG3 currents can disrupt gyre formation and move water
(Supplementary Figures S2 and S3), which are all from near the STEP site to the water column above
located within 350 m of the STEP coordinate the DWP, with a dilution factor of about 100 (Razmi
(Figure 1). Only sul1 was also detected at compara- et al., 2013). In previous studies, tracer released
tively high concentration and abundance at more from the WWTP during holomixis reached surface
remote sites (EG4, NC7, NC9 and NC14). waters at the DWP site (NC13) within 48 h, whereas
The relationship of ARGs with distance from the no tracers were detected when thermal stratification
WWTP discharge pipe was not linear (Pearson’s precluded upwelling (Wildi and Rossi, 1997;

The ISME Journal

 Antibiotic-resistance genes in lake sediments
N Czekalski et al
1386

Figure 3 Spatial interpolation maps showing the distributions of ARG abundance (ARG copy numbers normalized to bacterial 16S
rRNA gene copy numbers) and total mercury (THg, in mg kg  1) in Vidy Bay sediments. (a) sul1, (b) sul2, (c) tet(W), (d) tet(M), (e) tet(B)
and (f) THg.

Goldscheider et al., 2007). In the case of particle- previous clone library-based observations from Vidy
bound contaminants, the contamination record in Bay (Haller et al., 2011). The communities at the
sediments is expected to integrate over such tempo- most strongly ARG-contaminated sites STEP and
rally variable transport processes. EG2 were highly similar to some untreated waste-
water communities. Fitting of ARG abundances and
Contamination effects on the microbial community environmental variables to the sediment community
Analysis of bacterial community similarity composition showed that these community changes
(Figure 4a) showed a clearly distinct community in correlated with indicators of contamination, such
sites close to the discharge pipe. This confirms as ARG abundance, THg and total organic matter

The ISME Journal

 Antibiotic-resistance genes in lake sediments
N Czekalski et al
1387
a 0.4 b
ARG contam. sed. 0.4
Western transect sed.
Near-shore sed. NC1
Other sed.
Sewage
0.2 WTP out NC2
Lake water NC3
0.2 NC4

Clay (n.s.) sul2

Dim2

Dim2
0.0
tet(B)
Sewage EG2 0.0 STEP
NC2
sul1
NC1
-0.2 tet(M)
STEP tet(W) THg
-0.2 TOM

NC3
Silt EG2
-0.4 NC4

-0.4 -0.2 0.0 0.2 -0.4 -0.2 0.0 0.2 0.4

Dim1 Dim1
Figure 4 (a) Principle coordinates analysis of ARISA community profiles from sediment and water. Both biological and technical
replicates are included. Dispersion ellipses are shown for sediment sampling sites near the WWTP discharge pipe and for untreated
hospital and municipal wastewater (Sewage) (b) Principle coordinates analysis of ARISA community profiles from selected sediments
with passively fitted environmental variables and ARG abundance. For this analysis, technical replicates were averaged, omitting peaks
only present in one replicate. Samples were selected to match ARGs and sediment data availability. Symbols indicate samples grouped
according to sample type and location. n.s., not significant (P40.05).

(Figure 4b). In contrast, particle composition para- bacteria could proliferate or spread ARGs to the
meters appeared either unrelated to the distinction indigenous community via horizontal gene transfer.
between contaminated and uncontaminated sites If the latter mechanisms dominate, a diffuse and
(silt) or were not significantly correlated (clay, more widespread pattern of ARG concentrations
P40.05). Vectors fitted for sul2 and tet(B) indicated compared with THg would be expected. Differences
a slightly different distribution within the microbial as well as similarities in the spatial patterns of both
community compared with the other ARGs. Overall, contaminants may thus be informative.
community analysis indicated that changes in ARG Compared with previous studies (8 mg kg  1 sedi-
abundance were related to changes in community ment (Poté et al., 2008)), a relatively low maximum
composition, which can be explained by deposition THg concentration of 1.3 mg kg-1 sediment was
of bacteria originating in wastewater. However, an measured during our survey (EG2, compare
impact of confounding factors, such as heavy metal Figure 2f), indicating reduced recent deposition. We
pollution, cannot be ruled out by this analysis. found elevated THg levels to be the highest around
the STEP site but above background levels up to
600 m away in direction of the deep lake (EG4), 535 m
Comparison of ARGs and Hg contamination away in the south-western direction (NC6) and 877 m
The spatial trends in ARG distribution were in the south-eastern direction (EG14, Figure 2f).
compared with THg levels (Figures 2 and 3). Hg is The similarity of the vectors of THg and ARG
a pollutant that is known to be released from abundances fitted to community data in Figure 4b
Lausanne’s WWTP (Loizeau et al., 2004). Back- indicated a close link, which could either be
ground concentrations in Lake Geneva sediments deposition from a common source or because the
are B0.2 mg kg  1. Although the chemistry and heavy metal is a constraining factor of community
transport behavior of Hg is complex, particulate composition. Nevertheless, there are notable
transport is dominant in most systems (Glass et al., differences in the spatial distribution, as shown by
1990; Wang et al., 2004). Assuming that particle- relatively low R2 of linear regressions of THg and
associated bacteria or bacterial aggregates released ARG concentration (R2 ¼ 0.32–0.41, Po0.05,
from the WWTP are in turn primarily responsible for Supplementary Table S6) and abundance
ARG deposition in the sediment, we expected a (R2 ¼ 0.21–0.63, Po0.01, except for sul2 with
similar pollution pattern for THg and ARGs. P ¼ 0.05, Supplementary Table S7). The steeper
However, ARG dynamics may be affected by bacterial slope of ARG abundance with distance compared
mortality, for example, through predation, ultra- with THg (Figure 2) indicates restrictions on ARG
violet radiation or starvation. This could lead to a dissemination compared with THg. However, the
more restricted distribution compared with the south-western transect deviated from this general
heavy metal. On the other hand, released resistant observation. Here, ARG levels decreased more

The ISME Journal

 Antibiotic-resistance genes in lake sediments
N Czekalski et al
1388
slowly, whereas sediments were classified as unim- carried out along a section of the Almendares River,
pacted in terms of THg (Figure 3f). which is affected by different urban and industrial
Finally, low levels of THg were determined at near- contamination sources, good correlations were
shore sites (0.04 and 0.05 mg kg  1 for EG1 and NC14, demonstrated for sediment ARGs (tet and b-lacta-
respectively), most likely related to the comparatively mases) and copper levels (Graham et al., 2010).
large grain sizes and low organic matter content Other heavy metals (Pb, Co, Zn) also partly
(Supplementary Table S8) of sediments deposited in correlated with ARGs, but neither Hg nor sul genes
this area by the Chamberonne and Flon rivers. were considered in this study.
In contrast, site NC14 near the Chamberonne river However, for the present study we conclude that the
mouth showed elevated ARG levels (Figures 3a, b available evidence favors the hypothesis that the
and f). Previous studies monitoring pollutant trans- observed ARG levels are mainly a result of direct
port in the water column (Wildi and Rossi, 1997; deposition of resistant bacteria rather than being due
Goldscheider et al., 2007; Haldimann, 2009; Bonvin to co-selection by heavy metals: (a) The correlations of
et al., 2011) document that at least under certain ARG levels with THg were relatively poor. The
circumstances site NC13 might be affected by con- observed correlation can be explained by the parallel
taminants released from the WWTP, including ARGs release of both types of contaminants from the WWTP.
(Czekalski et al., 2012). However, the Chamberonne (b) The impact of the WWTP discharge on the
river is known as a discontinuous source of fecal deposition of fecal indicator bacteria and contrasting
bacteria (Haldimann, 2009; Haller et al., 2009), bacterial community composition has been shown
whereas its THg load is too low to explain elevated previously (Haller et al., 2011). (c) Our analysis
concentrations measured in sediments of Vidy Bay of community similarity using ARISA (Figure 4)
(Howa and Vernet, 1988). Our data show it to be a confirmed a distinct community composition of
likely source of resistant bacteria and ARG contami- ARG-contaminated sediments in the vicinity of the
nation (compare Supplementary Figures S5, S6f and discharge pipe that, furthermore, exhibits a high
S7f). Low ARG levels detected at locations close to similarity to sewage communities (Figure 4a), support-
the shore further east (EG1 and NC15) would support ing the notion that bacteria from these environments
the hypothesis that the Chamberonne plume is partly are deposited in the sediments. (d) A more rapid
responsible for the extension of the ARG contamina- decrease of ARG abundance with distance compared
tion in the south-western direction. This is also with THg (Figure 2) is best explained by different
supported by current measurements in Vidy Bay in dynamics during transport or deposition. Finally, (e)
2010, which revealed 45% of the currents moving we found highly similar patterns of ARG abundance
west-southwest (Razmi et al., 2013). in wastewater/sewage and sediment (Supplementary
Figure S4). Nevertheless, the role of the heavy metal
contamination in Vidy Bay for selection or persistence
Direct and indirect causes of ARG contamination of ARGs warrants further study.
Independently of transport, sediment properties
might also influence the natural abundance of ARGs
and the further fate of deposited allochthonous Implications
ARGs. Sediments containing a high fraction of clay We present detailed results on the spatial distribu-
minerals can bind DNA after cell death and protect it tion of different ARGs in sediments of a freshwater
from degradation by DNAse (Aardema et al., 1983). In lake that can be convincingly linked to the inflow of
the sampled sediments, the proportion of clay was municipal wastewater. Our results demonstrate that
quite low (0.01–3.92%, Supplementary Table S8). the combination of qPCR and geostatistical tools is a
ARGs have previously been detected in extracellular suitable approach to quantitatively detect and assess
DNA in livestock waste sludge but were one to two the dissemination of ARGs from local sources. The
orders of magnitude less frequent than the intra- release of wastewater into Vidy Bay has resulted in a
cellular fraction, and both fractions showed similar high ARG accumulation in sediments in the vicinity
dynamics (Zhang et al., 2013). Therefore we assume of the contamination source, but a diffuse zone of
that the detected ARGs reflect the ARGs in the viable elevated ARG concentrations extends further west
bacterial biomass in our samples. towards the location of the drinking water pumping
It has been demonstrated that heavy metals can station, indicating that transport takes place over
cross- or co-select for ARGs (Berg et al., 2010) as the considerable distances. Uptake of ARGs via drinking
same mechanisms often underlie both resistances, water is considered a risk for human health as
or resistance genes are located on the same mobile horizontal transfer to pathogens may occur. Treat-
genetic element (Baker-Austin et al., 2006; Knapp ment and quality control of waste and drinking
et al., 2011). As sediments in Vidy Bay are water should be reconsidered with this in mind. In
contaminated not only with Hg but also with other Switzerland and many other European countries,
heavy metals in excess of background levels (Poté the majority of WWTPs operate without a final
et al., 2008), it could be argued that the observed disinfection treatment, and many rely on basic
elevated ARG levels are caused by heavy metal biological treatment. As modernization of many of
pollution rather than bacterial deposition. In a study these facilities (including Lausanne’s WWTP) is

The ISME Journal

 Antibiotic-resistance genes in lake sediments
N Czekalski et al
1389
imminent or already under way, it will be important Blanc L. (2010). Characterisation of hospital wastewater: the
to consider modern procedures for disinfection and case of the CHUV. MSc thesis, Cranfield University:
contaminant removal that may be able to remove Cranfield, UK and EPFL: Lausanne, Switzerland.
ARGs from waste and drinking water (Dodd, 2012). Bonvin F, Rutler R, Chèvre N, Halder J, Kohn T. (2011).
Effective ARGs and pollutant removal may help to Spatial and temporal presence of a wastewater-derived
micropollutant plume in Lake Geneva. Environ Sci
break the cycle of ARG exchange between the Technol 45: 4702–4709.
human and the aquatic environment. The available Bürgmann H, Jenni S, Vazquez F, Udert KM. (2011).
technologies should be studied with regards to their Regime shift and microbial dynamics in a sequencing
effectiveness in reducing both the load of resistant batch reactor for nitrification and anammox treatment
bacteria and potentially transferable DNA. It should of urine. Appl Environ Microbiol 77: 5897–5907.
be kept in mind, however, that chemical disinfec- Cantas L, Shah SQA, Cavaco LM, Manaia C, Walsh F,
tion methods may themselves cause an undesirable Popowska M et al. (2013). A brief multi-disciplinary
selection for antibiotic resistance. Even wastewater review on antimicrobial resistance in medicine and its
receiving elaborate treatment and disinfection has linkage to the global environmental microbiota. Front
been shown to be a source of ARG contamination Microbiol 4: 96.
Czekalski N, Berthold T, Caucci S, Egli A, Buergmann H.
(LaPara et al., 2011). (2012). Increased levels of multiresistant bacteria and
resistance genes after wastewater treatment and their
dissemination into Lake Geneva, Switzerland. Front
Conflict of Interest Microbiol 3: 106.
The authors declare no conflict of interest. D’Costa VM, McGrann KM, Hughes DW, Wright GD. (2006).
Sampling the antibiotic resistome. Science 311: 374–377.
Dodd MC. (2012). Potential impacts of disinfection
Acknowledgements processes on elimination and deactivation of anti-
biotic resistance genes during water and wastewater
We thank: Nina Bircher, José Santos Cáceres and Francisco treatment. J Environ Monit 14: 1754–1771.
Vazquez for substantial support of our laboratory work; Glass GE, Sorensen JA, Schmidt KW, Rapp GR. (1990). New
Philippe Arpagaus and Janine Halder for assistance during source identification of mercury contamination in the
field campaigns; Jean-Luc Loizeau and Andrew Clarke for Great Lakes. Environ Sci Technol 24: 1059–1069.
kindly revising the manuscript; Holger Heuer, Fiona Goldscheider N, Haller L, Pote J, Wildi W, Zopfi J. (2007).
Walsh and Laurent Poirel for providing positive control Characterizing water circulation and contaminant
strains; Fiona Walsh for support in establishing tet gene transport in Lake Geneva using bacteriophage tracer
qPCR; Anh-Dao Le thi and Walter Wildi for providing the experiments and limnological methods. Environ Sci
bathymetry map of Vidy Bay; Rosi Siber for substantial Technol 41: 5252–5258.
help with geostatistical analysis; Luca Rossi and Nathalie Graham DW, Olivares-Rieumont S, Knapp CW, Lima L,
Chèvre for organizing access to sampling sites and back- Werner D, Bowen E. (2010). Antibiotic resistance gene
ground data; Gregor Maurer and Anoys Magnet (STEP abundances associated with waste discharges to the
Vidy) for access to their facilities and support; and COST Almendares River near Havana, Cuba. Environ Sci
action TD0803 ‘Detecting evolutionary hot spots of Technol 45: 418–424.
antibiotic resistances in Europe (DARE)’ for discussions Haldimann S. (2009). Modelling the Spreading of Urban
and valuable feedback. Funding was provided by the Water Discharges in the Coastal Zone of Lausanne.
Swiss National Science Foundation Grant PFMP2-123048. EPFL: Lausanne, Switzerland.
Haller L, Poté J, Loizeau JL, Wildi W. (2009). Distribution
and survival of faecal indicator bacteria in the sediments
of the Bay of Vidy, Lake Geneva, Switzerland. Ecol
References Indicators 9: 540–547.
Haller L, Tonolla M, Zopfi J, Peduzzi R, Wildi W, Poté J.
Aardema BW, Lorenz MG, Krumbein WE. (1983). Protection (2011). Composition of bacterial and archaeal
of sediment-adsorbed transforming DNA against communities in freshwater sediments with different
enzymatic inactivation. Appl Environ Microbiol 46: contamination levels (Lake Geneva, Switzerland).
417–420. Water Res 45: 1213–1228.
Aminov RI. (2009). The role of antibiotics and antibiotic Heuer H, Smalla K. (2007). Manure and sulfadiazine
resistance in nature. Environ Microbiol 11: 2970–2988. synergistically increased bacterial antibiotic resistance
Arboille D, Howa H, Span D, Vernet J-P. (1989). Etude in soil over at least two months. Environ Microbiol 9:
générale de la pollution par les métaux et répartition 657–666.
des nutriments dans les sédiments du Léman. Rapp. Heuer H, Focks A, Lamshöft M, Smalla K, Matthies M,
Comm. Int. prot. eaux Léman contre pollut. Campagne Spiteller M. (2008). Fate of sulfadiazine administered
1988: 139–172. to pigs and its quantitative effect on the dynamics of
Baker-Austin C, Wright MS, Stepanauskas R, McArthur JV. bacterial resistance genes in manure and manured
(2006). Co-selection of antibiotic and metal resistance. soil. Soil Biol Biochem 40: 1892–1900.
Trends Microbiol 14: 176–182. Howa H, Vernet H-P. (1988). Metaux lourds dans les
Berg J, Thorsen MK, Holm PE, Jensen J, Nybroe O, sediments des affluents vaudois de Léman. Institut F.
Brandt KK. (2010). Cu exposure under field conditions A. Forel, Université de Genève: Versoix, Switzerland.
coselects for antibiotic resistance as determined by a Jury KL, Khan SJ, Vancov T, Stuetz RM, Ashbolt NJ. (2011).
novel cultivation-independent bacterial community Are sewage treatment plants promoting antibiotic
tolerance assay. Environ Sci Technol 44: 8724–8728. resistance? Crit Rev Environ Sci Technol 41: 243–270.

The ISME Journal

 Antibiotic-resistance genes in lake sediments
N Czekalski et al
1390
Kim S, Aga DS. (2007). Potential ecological and human lacustrine embayment: Vidy Bay (Lake Geneva,
health impacts of antibiotics and antibiotic-resistant Switzerland). Aquat Sci: 1–13.
bacteria from wastewater treatment plants. J Toxicol Righetti M, Toffolon M, Lucarelli C, Serafini M. (2011).
Environ Health Part B Crit Rev 10: 559–573. Sediments as tracers for transport and deposition
Kindt R, Coe R. (2005). Tree diversity analysis. A manual processes in peri-alpine lakes: a case study. J Hydrol
and software for common statistical methods for 411: 1–11.
ecological and biodiversity studies: World Agrofores- Rizzo L, Manaia C, Merlin C, Schwartz T, Dagot C,
try Centre (ICRAF), Nairobi. Ploy MC et al. (2013). Urban wastewater treatment
Knapp CW, McCluskey SM, Singh BK, Campbell CD, plants as hotspots for antibiotic resistant bacteria and
Hudson G, Graham DW. (2011). Antibiotic resistance gene genes spread into the environment: a review. Sci Total
abundances correlate with metal and geochemical condi- Environ 447: 345–360.
tions in archived Scottish soils. PLoS One 6: e27300. Roos-Barraclough F, Givelet N, Martinez-Cortizas A,
Kümmerer K. (2004). Resistance in the environment. Goodsite ME, Biester H, Shotyk W. (2002). An
J Antimicrob Chemother 54: 311–320. analytical protocol for the determination of total
LaPara TM, Burch TR, McNamara PJ, Tan DT, Yan M, mercury concentrations in solid peat samples. Sci
Eichmiller JJ. (2011). Tertiary-treated municipal waste- Total Environ 292: 129–139.
water is a significant point source of antibiotic Száková J, Kolihová D, Miholová D, Mader P. (2004).
resistance genes into Duluth-Superior Harbor. Environ Single-purpose atomic absorption spectrometer AMA-
Sci Technol 45: 9543–9549. 254 for mercury determination and its performance in
Loizeau JL, Pardos M, Monna F, Peytremann C, Haller L, analysis of agricultural and environmental materials.
Dominik J. (2004). The impact of a sewage treatment Chem Papers 58: 311–315.
plant’s effluent on sediment quality in a small bay in Lake Teuber M. (2001). Veterinary use and antibiotic resistance.
Geneva (Switzerland-France). Part 2: Temporal evolution Curr Opin Microbiol 4: 493–499.
of heavy metals. Lakes Reserv Res Manage 9: 53–63. Vioget P, Jaquerod C-A, Ruegg E, Poget E. (2011). Bilans
Loizeau JL, Arbouille D, Santiago S, Vernet JP. (1994). 2010 de l’épuration vaudoise. Service des eaux, sols et
Evaluation of a wide range laser diffraction grain size assainissement, Canton de Vaud: Orbe, Savigny,
analyser for use with sediments. Sedimentology 41: Switzerland.
353–361. Walsh F, Ingenfeld A, Zampicolli M, Hilber-Bodmer M,
Martinez-Martinez L, Pascual A, Jacoby G. (1998). Quino- Frey JE, Duffy B. (2011). Real-time PCR methods for
lone resistance from a transferable plasmid. Lancet quantitative monitoring of streptomycin and tetra-
351: 797–799. cycline resistance genes in agricultural ecosystems.
Martinez JL. (2009a). The role of natural environments in J Microbiol Methods 86: 150–155.
the evolution of resistance traits in pathogenic bacteria. Wang Q, Kim D, Dionysiou DD, Sorial GA, Timberlake D.
Proc R Soc Lond Ser B Biol Sci 276: 2521–2530. (2004). Sources and remediation for mercury contami-
Martinez JL. (2009b). Environmental pollution by anti- nation in aquatic systems—a literature review.
biotics and by antibiotic resistance determinants. Environ Pollut 131: 323–336.
Environ Pollut 157: 2893–2902. Wildi W, Rossi P. (1997). Assainissement de la baie de
Oksanen J, Blanchet FG, Kindt R, Legendre P, O’Hara RB et al. Vidy: tarcage par bactériophages 19-25 février 1997
(2012). vegan: Community Ecology Package. R package 6-11 mai 1997. Ville de Lausanne, Direction de la
version 2.0-4. sécurité cociale et de lénvironnement, Service d’assai-
Pei R, Kim S-C, Carlson KH, Pruden A. (2006). Effect of nissement, Station d’épuration de Vidy, Université de
river landscape on the sediment concentrations of Genève, Institut F.A. Forel, Section des sciences de la
antibiotics and corresponding antibiotic resistance terre, Université Neuchâtel, Laboratoire de Micro-
genes (ARG). Water Res 40: 2427–2435. biologie: Lausanne, Switzerland.
Poté J, Haller L, Loizeau JL, Garcia Bravo A, Sastre V, Wildi W. Wüest A, Lorke A. (2003). Small-scale hydrodynamics in
(2008). Effects of a sewage treatment plant outlet pipe lakes. Annu Rev Fluid Mech 35: 373–412.
extension on the distribution of contaminants in the Yang H-m, Pan W-P. (2007). Transformation of mercury
sediments of the Bay of Vidy, Lake Geneva, Switzerland. speciation through the SCR system in power plants.
Bioresour Technol 99: 7122–7131. J Environ Sci 19: 181–184.
Pruden A, Pei R, Storteboom H, Carlson KH. (2006). Yannarell AC, Kent AD, Lauster DH, Kratz TK, Triplett W.
Antibiotic resistance genes as emerging contaminants: (2003). Temporal patterns in bacterial communities in
studies in Northern Colorado. Environ Sci Technol 40: three temperate lakes of different trophic status.
7445–7450. Microb Ecol 46: 391–405.
R Core Team. (2012). R: A Language and Environment for Zhang Y, Snow DD, Parker D, Zhou Z, Li X. (2013).
Statistical Computing. R Foundation for Statistical Intracellular and extracellular antimicrobial
Computing: Vienna, Austria. resistance genes in the sludge of livestock waste
Ramette A. (2009). Quantitative community fingerprinting management structures. Environ Sci Technol 47:
methods for estimating the abundance of operational 10206–10213.
taxonomic units in natural microbial communities. Zurfluh K, Hächler H, Nüesch-Inderbinen M, Stephan R.
Appl Environ Microbiol 75: 2495–2505. (2013). Characteristics of extended-spectrum b-Lacta-
Razmi AM, Barry DA, Lemmin U, Bonvin F, Kohn T, mase- and Carbapenemase-producing Enterobacteria-
Bakhtyar R. (2013). Direct effects of dominant winds ceae isolates from rivers and lakes in Switzerland.
on residence and travel times in the wide and open Appl Environ Microbiol 79: 3021–3026.

Supplementary Information accompanies this paper on The ISME Journal website (http://www.nature.com/ismej)

The ISME Journal