Dissemination of antibiotic resistance genes associated with the sporobiota in sediments impacted by wastewater

Site description and sampling

Lake Geneva is the largest freshwater lake of Western Europe, with a volume of 89 km³, a surface area of 580 km² and a maximum depth of 309 m. Vidy Bay is enclosed by the shoreline near Lausanne between St.-Sulpice and Lausanne-Ouchy. The outlet pipe of the WWTP is located 700 m offshore at 35 m depth (46.51197121 N; 6.587423025 E).

The ten cores used for this study were retrieved using standard corers (50 cm length Plexiglas tubes of 6 cm diameter) between July 2011 and May 2012 in three sampling zones differently impacted by the WWTP (Bueche, 2014; Sauvain et al., 2014): 5–40 m from the outlet pipe of the WWTP (“near”, N1–N4);  134–429 m from the outlet pipe (“middle”, M1–M4); and  611–956 m from the outlet pipe (“distal”, D1–D2). Two sediment layers were analyzed for the cores with the code numbers 1 and 2 (D1, D2, M1, M2, N1, N2) [20]: 0–3 cm and 3–9 cm depth. Three sediment layers were analyzed for the cores with the code numbers 3–4 (M3, M4, N3, N4): 0–1.5 cm, 1.5–3 cm and 3–9 cm depth.

DNA extraction

Sporobiome DNA was obtained using an indirect three step extraction method: (i) extraction of cells from the sediment particles, (ii) separation of spores from the vegetative cells, (iii) DNA extraction from the spores (Wunderlin et al., 2014).

The treatment to separate spores from vegetative cells was performed on the biomass from 1.5 grams of sediment collected on one nitrocellulose filter (Merck Millipore, Darmstadt, Germany) per sample, as previously described (Wunderlin et al., 2014). The first step consisted of the lysis of vegetative cells by heat, enzymatic agents (lysozyme) and chemicals (Tris-EDTA, NaOH, SDS). The second step consisted of a DNase digestion in order to destroy the free DNA. Treated filters were stored at −20 °C until DNA was extracted from the pre-treated filters using a protocol based on the FastDNA^® SPIN kit for soil (MP Biomedicals, Santa Ana, CA, USA) (Wunderlin et al., 2013) with the following additional modifications: samples in lysing matrix were submitted to two successive bead-beating steps. Supernatant retrieved from each bead-beating was treated separately according to manufacturer’s instructions and DNA extracts were pooled together at the end of the procedure. Pooled DNA was precipitated with 0.3 M Na-acetate and ethanol (99%), stored at −20 °C overnight and centrifuged for 1 h at 21460xg and 4 °C. Supernatant was removed and the pellet was washed with 1 ml of 70% ethanol and centrifuged for 30 min at 21460xg and 4 °C. Supernatant was removed and the residual ethanol was allowed to evaporate at room temperature. Pellet was re-suspended in 50 µl of PCR-grade water. Total DNA was quantified using Qubit^® dsDNA HS Assay Kit on a Qubit^® 2.0 Fluorometer (Invitrogen, Carlsbad, CA, USA).

Real-time quantitative PCR on tet(W) and sul1 genes

Real-time Taqman^®-PCR on sul1 and tet(W) genes was performed in 384-well plates using a LightCycler^®480 Instrument II (Roche, Basel, Switzerland). For sul1, the primers used were qSUL653f (5′-CCGTTGGCCTTCCTGTAAAG-3′) and qSUL719r (5′-TTGCCGATCGCGTGAAGT-3′) with tpSUL1 (FAM-CAGCGAGCCTTGCGGCGG-TAMRA) probe (Heuer & Smalla, 2007). The reaction mix for sul1 consisted of 2 µL of DNA template (between 0.08 and 1.39 ng/µL), 0.025 µM of each primer, 0.25 µM of TaqMan probe and 1 × TaqMan^® Fast Universal PCR Master Mix (Applied Biosystems, Foster City, CA, USA). Total reaction volume of 10 µL was reached with PCR-grade water. For tet (W), the primers used were tetW_f (5′-CGGCAGCGCAAAGAGAAC-3′) and tetW_r (5-CGGGTCAGTATCCGCAAGTT-3′) with tetW_s (FAM-CTGGACGCTCTTACG-TAMRA) probe (Walsh et al., 2011). The reaction mix for tet(W) consisted of 2 µL of DNA template, 0.025 µM of each primer, 0.1 µM of TaqMan probe and 1 × TaqMan^® Fast Universal PCR Master Mix (Applied Biosystems, Foster City, CA, USA). Total reaction volume of 10 µL was reached with PCR-grade water. The qPCR program was the same for both genes and started with a hold at 95 °C for 10 min, followed by 45 cycles of denaturation at 95 °C for 15 s and annealing/elongation at 60 °C for 1 min. The qPCR assays were performed in technical triplicates on samples, standards and negative controls. The negative controls consisted of PCR blanks with only the reaction mix and of PCR blanks containing the mix and 2 µL of PCR-grade water. Standard curves were prepared from serial 10-fold dilutions of plasmid DNA containing the respective target gene in a range of 5 × 10⁷ to 50 gene copies. For sul1, control plasmids and standard curves were prepared as previously described (Heuer & Smalla, 2007). For tet(W), standard curves were prepared as previously described (Walsh et al., 2011). The effect of inhibitors on amplification was tested for all the samples and for both genes. All samples were spiked with 10⁴ copies of plasmid DNA containing the tet(W) or the sul1 gene and amplified together with the same set of non-spiked samples and control DNA and the results indicated that inhibition was negligible.

Sequencing and data analysis

Purified DNA extracts were sent to Fasteris (Geneva, Switzerland) for 16S rRNA gene amplicons sequencing using Illumina MiSeq platform (Illumina, San Diego, USA), generating 250 bp paired-end reads. The hypervariable V3–V4 region was targeted using universal primers Bakt_341F (5′-CCTACGGGNGGCWGCAG-3′) and Bakt_805R (5′-GACTACHVGGGTATCTAATCC-3′) (Herlemann et al., 2011). Analysis of the dataset was performed using Mothur (Schloss et al., 2009) following the MiSeq SOP (Kozich et al., 2013). The SILVA NR v123 reference database (Quast et al., 2013) was used for the alignment of amplicons and the taxonomic assignment of representative OTUs. After quality filtering and removal of chimeras, a total of 1818238 amplicons were obtained (408758 unique sequences). Singletons were removed prior to the clustering into OTUs and corresponded to 373374 sequences. Average neighbor clustering of the 1444864 remaining sequences (35384 unique sequences) with an identity threshold of 97% led to the identification of 6390 OTUs. Sequencing data was deposited to NCBI under the Bioproject accession number PRJNA396277.

Statistical and multivariate analysis

Community and statistical analyses were performed using R version 3.4.0 (R Core Team, 2014) and the phyloseq and vegan packages (McMurdie & Holmes, 2013; Oksanen et al., 2017). The distance of samples to the WWTP outflow is a combination (euclidian distance) of the distance between the sample and the pipe on the east axis, the north axis (based on CH coordinates) and the water column depth. It is expressed in meters. The difference in ARG levels between the zones (near, middle and distal) was tested using analysis of variance (ANOVA) and Tukey’s test for pairwise comparisons. Pairwise correlations between environmental parameters (organic carbon, total nitrogen, and concentration of cadmium, copper, iron, manganese, aluminum, zinc and arsenic) and ARG abundance/frequency were calculated using Spearman’s rank correlation coefficient. This correlation coefficient was selected because we did not assume a normal distribution of the variables. The organic carbon, total nitrogen, and metal concentrations were obtained from previous studies (Bueche, 2014; Sauvain et al., 2014). P-values were adjusted for family-wise error rate using the Holm method. Spearman’s rank correlation coefficient was used for calculating the correlations between ARG frequency and the relative abundance of OTUs. To provide a visual representation of the distribution of the correlation coefficient of OTUs with ARG concentration, Kernel density curves were computed from Spearman’s correlation coefficients calculated between each ARG and OTU and can be thought of as continuous histograms representing the frequency (density) of the correlation coefficients. The significance of the Spearman’s correlation coefficients was tested by correcting the p-values using the false discovery rate correction (Benjamini–Hochberg method). This method was selected because of the large number of comparisons performed. Principal component analysis (PCA) was computed on the environmental parameters and ARG abundance/frequency, after standardization (zero mean and unit variance). The environmental parameters included were C_org, N_tot, metals (listed above), distance to the outlet pipe (Dist_pipe), DNA abundance in ng per gram of sediment (DNA_ng_gSed), and ARG abundance and frequency. Sporobiota community was analyzed by principal coordinates analysis (PCoA), based on Bray-Curtis dissimilarity and Hellinger transformation of the OTU table. Environmental parameters and ARG abundance/frequency were standardized and passively fitted to the ordination. Only significant parameters were displayed (p < 0.05).

Prevalence and spatial distribution of tet(W) and sul1 in sporobiome DNA

To determine the prevalence of the two selected ARG in DNA extracted from the sporobiota and to describe their spatial distribution in sediments impacted by a WWTP, ARG quantification was carried out for three zones located at 5-40 m (near) 134-429 m (middle) and 611–956 m (distal) from the outlet pipe (Fig. 1). These zones were previously shown to be differentially impacted by the release of wastewater, according to the concentration of particulate trace and heavy metals in the sediments and to the bacterial community composition (Sauvain et al., 2014). The results of ARG quantification were expressed in two different units. The first unit, ARG abundance, refers to the number of copies of ARG per g of sediment. This reflects accumulation in sediment and represents the total ARG pool for a given ARG. However, differential microbial load in sediments could bias this measure. Therefore, we used a second unit, ARG frequency, which is defined as the number of copies of ARG per ng of extracted DNA. This allowed us to estimate the enrichment of specific ARG in DNA from the enriched sporobiota. We caution that both approaches are potentially affected by changes in the proportion of different bacterial groups that may not be causally related to either inputs or growth of ARB or local selection of ARG. For example, an increase in abundance of a population without an ARG due to nutrient inputs will result in a reduction of ARG frequency.

Both ARG, tet(W) and sul1, were detected in DNA at all sampling locations. In all samples, abundance and frequency were higher for tet(W) than for sul 1 (Table 1). The influence of the WWTP as the main source of sporobiota-associated ARG can be observed in the spatial distribution of ARG abundance. Abundance was highest in samples from the proximal zone and decreased with distance to the outlet pipe (Fig. 2, Figs. S1 and S2). This pattern was confirmed by principal component analysis (PCA; Fig. S1) and Spearman correlation tests (Fig. S3), which showed that ARG abundance was mostly correlated with environmental variables associated with deposition of treated wastewater, such as organic carbon content (C_org) and total nitrogen (N_tot) in sediments. However, unexpectedly, the R² calculated for the linear regression established between ARG abundance and distance to the outlet pipe in our case (Fig. S2) was generally low.

Measures of ARG frequency showed that tet(W) and sul1 genes had different enrichment patterns in the sporobiota. In the case of sul1, ARG frequency decreases in relation to the distance to the outlet pipe (Fig. 2, Fig. S2). A decrease in frequency could be attributed to dilution of a population with high sul1 frequency (originating from the wastewater) within the environmental sporobiota community. The frequency of tet(W) also decreased in relation to the distance to the outlet pipe, but its pattern of distribution was more complex. The highest tet(W) frequencies were found in two middle zone samples (M1_low and M2_low, Fig. 2, Fig. S1). Both correlation tests (Fig. S3) and PCA with the environmental variables (Fig. S1B) indicated a significant correlation between tet(W) frequency and certain metals (Fe, As, and Cu).

Sporobiota community composition

Several studies have demonstrated a strong impact of WWTP discharge on bacterial community composition, which is shown by changes in the community in impacted sediments compared to remote sites (LaPara et al., 2011; Czekalski, Gascón Díez & Bürgmann, 2014; Sauvain et al., 2014). In previous studies using a different sequencing technology (pyrosequencing), the total bacterial communities analyzed from the same sediments were highly divergent, forming at least four distinct groups depending on the distance to the outlet pipe and the concentration of heavy and trace metals in the sediments (Bueche, 2014; Sauvain et al., 2014). In contrast, the sporobiota community has a low degree of variability among impacted and supposedly less impacted sites (Fig. 3), suggesting that the distance to the wastewater outlet pipe had a relatively small impact on sporobiota community composition. Firmicutes are the dominant phylum in all samples (Fig. S4), accounting for at least 85.2% of the sporobiota community, except in sample taken from the middle area at a depth of 1.5–3 cm (M4_med; 65.9%). This is in agreement with a previous study demonstrating an enrichment of 83.9–90.6% in Firmicutes using the same method for the isolation of spores (Wunderlin et al., 2014). Among the non-Firmicutes representatives, we found several groups for which the production of spores is so far not reported and that could represent potential contamination with vegetative cells that withstood the extraction procedure (see below). Nevertheless, we performed the analysis with the entire community as our definition of the sporobiota considers the resilience to the extraction method, rather than phylogenetic affiliation.

In terms of relative abundance, 43.1% of the whole community was identical between all samples. The two most abundant genera, Clostridium_sensu_stricto_1 and an unclassified Peptococcaceae, were the same for all samples (except M4_med), with an average relative abundance of 57.3%. Both are widely represented in the human gut (Lozupone et al., 2012), which is also the case for other subdominant genera, including Intestinibacter (9.7%), Turicibacter (6.7%), and Ruminococcus_2 (0.9%). These findings suggest a common origin of the sporobiome and point again at the WWTP as the main source of the sporobiome community depositing in the lake sediment in the area studied here.

Despite the high homogeneity of the sporobiota community, principal coordinate analysis (PCoA, Fig. 4) reveals a slight variability among the communities. For example, in Axis 1, Bacilli is associated with samples located at a greater distance from the outlet pipe, while different Clostridia are associated to samples closer to the WWTP. While Clostridia are mostly anaerobic and are common in the microbiome of humans and other mammals, Bacilli include numerous aerobic organisms and are commonly found in soils (Madigan et al., 2015). Accordingly, Clostridia could be seen as a signature of the human influence in the sporobiota associated with wastewater discharge whereas Bacilli would represent a more environmental sporobiota community.

In addition, PCoA revealed two groups of samples of particular interest (Fig. 4). Group 1 (Gr.1 in Fig. 4; Fig. S5B), which was characterized by high load of C_org and N_tot, and high abundance/frequency of ARG, was composed of samples from the zone in close proximity to the WWTP discharge. Among the OTUs associated to this group, the most abundant belong to the genera Bifidobacteria and Collinsella, two Actinobacteria common in the human microbiome (Lamendella et al., 2008; Thorasin, Hoyles & McCartney, 2015), albeit unknown for the production of spores. Another abundant genus in this group is Trichococcus, a Firmicute found in sewerage infrastructure and WWTP (Vandewalle et al., 2012). This group reflects the direct influence of the WWTP (Fig. S5B). In contrast, Group 2 (Gr.2 in Fig. 4; Fig. S5C) clustered samples from both the near and middle zones and was related to elevated concentrations of particulate metals (Zn, Cd, Cu, As, and Fe) and a high frequency of tet(W) (Fig. 4). OTUs defining this group belonged mainly to non-Firmicutes ( Fig. S5C). Some were affiliated to organisms that have been previously isolated from WWTPs, including Synergistetes (Wang et al., 2013), Syntrophorhabdus (Saia et al., 2016), Anaerolinaceae and Leptolinae (Yamada et al., 2006). Others were affiliated to uncommon taxa previously isolated from rivers or aquifers, such as Gaiellales (Albuquerque et al., 2011) and Kaistia (Jin et al., 2011). To our knowledge, there is no evidence that these organisms are resistant to high metal contamination.

Correlation tests between OTUs and ARG

To further investigate the relationship between the sporobiota community and ARG distribution, pairwise spearman correlation tests were performed between tet(W) frequency and the relative abundance of each OTU. For both tet(W) and sul1, the results indicated that only a small number of OTUs correlate with the ARG studied here (Figs. 5A–5B). Interestingly, the 10 most positively correlated OTUs were specific to each ARG. Although environmental correlations alone are not sufficient to identify carriers of antibiotic resistance, the results can help to highlight potential groups that merit further investigation. For example, the three most tet(W)- positively correlated OTUs (correlation coefficients ranging from 0.84 to 0.71) all belonged to Firmicutes (Table 2). In particular, one of these OTUs (OTU00002; Clostridium_sensu_stricto_1) was highly abundant (over 20% of the relative abundance) in all samples (Fig. 5C). The known bacterial species most closely related to this OTU were Clostridium celatum, Clostridium saudiense, and Clostridium disporicum, all isolated from the feces of mammals, including humans, swine and rats (Horn, 1987; Angelakis et al., 2014; Agergaard et al., 2016). Although in most Clostridium spp. studied so far, tetracycline resistance is conferred by the tet(M) gene (Adams et al., 2002), tet(W) has been detected in a medical isolate from Clostridium difficile (Spigaglia, Barbanti & Mastrantonio, 2008). A similarity search using this gene as a query showed the detection of homologues in various uncultured environmental Clostridium spp. (Supplemental Information 2). The other two most correlated OTUs (Otu00045 and Otu00062) both belonged to the genus Ruminococcus_1, closely related to Ruminococcus callidus and unclassified Ruminococcus spp., respectively. All of these have been reported in human and other mammal feces (Wang, Cao & Cerniglia, 1997; Leser et al., 2002; Schmidt et al., 2011). Another OTU of interest was OTU00008, which represented up to 1.75% of the total community (0.87% in average). OTU00008 was closely related to Ruminococcus bromii, one of the predominant species in human, swine, and cattle gut (Moore, Cato & Holdeman, 1972; Klieve et al., 2007). Although the genus Ruminococcus was originally considered as to comprise non-spore forming species, a recent report has demonstrated the production of spores by Ruminococcus isolated from human colon and rumen (Mukhopadhya et al., 2018). This gives additional support to our approach and shows the importance of defining the sporobiota based on its properties, rather than on the reports of spore production in cultured isolates.

In spite of the widespread phylogenetic distribution of sul1 (Aminov, 2010), the 10 most sul1-positively correlated OTUs belonged mainly to Actinobacteria (Table 2). The other most correlated OTUs belonged to Firmicutes, Proteobacteria, and Saccharibacteria (all with a correlation coefficient above 0.8). Unsurprisingly, all these highly correlated OTUs were closely related to organisms commonly reported as part of the human/mammal microbiome and/or are commonly found in WWTPs. Compared to tet(W), none of the best correlated OTUs represented a major fraction of the sporobiome community (<0.25% in average, Fig. 5D). However, one has to be careful in the interpretation of these correlations. The system under study is mainly driven by the influence of the WWTP. Most sul1-correlated OTUs were related to Group 1 and their abundance correlates with high concentrations of C_org and N_tot and other environmental indicators of the direct influence of treated wastewater release (Fig. S5).