Do environmental pharmaceuticals affect the composition of bacterial communities in a freshwater stream?: A case study of the Knivsta river in the south of Sweden

(1)

Do environmental pharmaceuticals affect the composition of bacterial communities in a freshwater stream? A case study of the Knivsta river in the south of Sweden

Aleksandra Hagberg

^a

, Shashank Gupta

^b

, Olena Rzhepishevska

^a

, Jerker Fick

^a

, Mette Burmølle

^b

, Madeleine Ramstedt

^a,

⁎

aDepartment of Chemistry, Umeå Center for Microbial Research, Umeå University, 901 87 Umeå, Sweden

bSection of Microbiology, Department of Biology, University of Copenhagen, 2100, Copenhagen, Denmark

H I G H L I G H T S

• Environmental pharmaceuticals may affect biota and fresh water ecosystems.

• Pharmaceutical content was analyzed in a fresh water stream.

• The stream received treated sewage water from a nearby town.

• Bioﬁlms were investigated by direct sequencing as well as cultivation of bacteria.

• Bioﬁlm composition differed up- and downstream from the treated efﬂuent.

G R A P H I C A L A B S T R A C T

a b s t r a c t a r t i c l e i n f o

Article history:

Received 8 September 2020

Received in revised form 7 October 2020 Accepted 7 October 2020

Available online 16 October 2020 Editor: Damia Barcelo

Keywords:

Bacterial bioﬁlm Pharmaceuticals Sequencing Bioﬁlm sampling Fresh water

Pharmaceutical substances present at low concentrations in the environment may cause effects on biological systems such as microbial consortia living on solid riverbed substrates. These consortia are an important part of the river ecosystem as they form part of the food chain. This case study aims to contribute to an increased understanding of how low levels of pharmaceuticals in freshwater streams may influence sessile bacterial consortia. An important point source for pharmaceutical release into the environment is treated household sewage water. In order to investigate what types of effects may occur, we collected water samples as well as riverbed substrates from a small stream in the south of Sweden, Knivstaån, upstream and downstream from a sewage treatment plant (STP). Data from these samples formed the base of this case study where we investigated both the presence of pharmaceuticals in the water and bacterial composition on riverbed substrates. In the water downstream from the STP, 19 different pharmaceuticals were detected at levels below 800 ng/dm³. The microbial composition was obtained from sequencing 16S rRNA genes directly from substrates as well as from cultivated isolates. The cultivated strains showed reduced species variability compared with the data obtained directly from the substrates. No systematic differences were observed following the sampling season. However, differences could be seen between samples upstream and downstream from the STP effluent. We further observed large similarities in bacterial composition on natural stones compared to sterile stones introduced into the river approximately two months prior to sampling, giving indications for future sampling methodology of biofilms.

creativecommons.org/licenses/by/4.0/).

⁎ Corresponding author.

E-mail address:madeleine.ramstedt@umu.se(M. Ramstedt).

https://doi.org/10.1016/j.scitotenv.2020.142991

Contents lists available atScienceDirect

Science of the Total Environment

j o u r n a l h o m e p a g e :w w w . e l s e v i e r . c o m / l o c a t e / s c i t o t e n v

(2)

1. Introduction

Over the last couple of decades, pharmaceutical contamination has become a concern as the impact on the environment is not yet fully understood (Küster and Adler, 2014;Berendonk et al., 2015;Segura et al., 2009;

Fick et al., 2017;Loos et al., 2013;Lindberg et al., 2014;Boxall et al., 2012).

Pharmaceutical residues have been found all over the globe: in fresh water, soils and even in the ice at the Antarctic (González-Alonso et al., 2017;Roig and D'Aco, 2016). Pharmaceuticals are compounds that were selected among other substances due to their ability to affect a speciﬁc biological target in a target organism (human or animal). However, a draw- back to this ability is that once released in the environment, they continue to be biologically active until they are degraded (Taylor, 2016). Pharma- ceuticals detected in fresh water most often originate from low levels re- maining in treated household sewage water (Roig and D'Aco, 2016;

Voulvoulis et al., 2016). A sewage treatment plant (STP) is designed to clean sewage water both from physical objects as well as chemical and biological contaminations, which include pathogenic bacteria, organic compounds, and nutrients such as phosphorous. However, many pharmaceuticals are not well captured or degraded in STPs, and remain at low levels in the treated efﬂuent that is released into the recipient (Voulvoulis et al., 2016). In the environment, the pharmaceuticals can be decomposed by degradation processes such as photo degradation or biodegradation, but may also become enriched in the bottom sediments where they may cause effects on biota (Voulvoulis et al., 2016).

In the environment, bacteria are often present in the form of bioﬁlms.

Biofilms are structures formed by microorganisms that protect them from environmental stresses (Besemer, 2015). In the biofilm, microbes are surrounded by a self-produced matrix that has several functions including protection and communication between cells inside the biofilms, as well as with the outside world (Flemming et al., 2016;Seviour et al., 2019).

Biofilms are an important part of each ecosystem as they influence the cycling of elements e.g. by enzymatic degradation of a range of substances (Stoodley et al., 2002). Biofilms are ubiquitous and can consist of bacteria, algae, and fungi (Burmølle et al., 2014). Freshwater biofilms are important in aqueous primary production as they contain microorganisms that pho- tosynthesize,fix carbon, cycle phosphate and nitrogen, and bring these elements into the fresh-water ecosystem (Lear et al., 2012;Romani et al., 2012). Furthermore, they form part of the food web in freshwater systems as a food source for grazing aquatic organisms (Lear et al., 2012). The knowledge of how pharmaceuticals affect biofilms is currently limited, but some hypotheses are made. One is that interactions between aqueous biofilms and pharmaceuticals lead to enrichment or degradation inside the biofilm. Such processes could influence the entire food web as well as the water quality close to the biofilm (Lear et al., 2012). If pharmaceuticals are enriched in biofilms, this may lead to increased exposure for grazing aquatic organisms and their predators, as previously described (Lagesson et al., 2016). However, if pharmaceuticals are instead degraded in the biofilms they may reduce this type of exposure.

This study is part of a larger initiative where we aim to investigate how pharmaceuticals interact with environmental bacterial biofilms and what effect this has on the environment. We here present an inven- tory of bacterial consortia present in a small Swedish freshwater stream, Knivstaån, during spring and summer 2018. The stream receives effluent from a local STP (Östman et al., 2019) and we investigated the water quality with respect to pharmaceutical content upstream and downstream from the effluent. Furthermore, we collected substrates from the riverbed and investigated biofilm composition, in order to observe how the effluent influenced the composition of bacterial consortia.

2. Materials and methods 2.1. Chemicals and reagents

Nitric acid and hydrochloric acid was Suprapur (Merk, Darmstedt, Germany). Ultrapure water (MilliQ) was taken from a Merck Millipore

Advantage A10 system. All reference standards were classiﬁed as analytical grade (>98%).

2.2. Sampling

Water and bioﬁlm samples were collected from a small stream, Knivstaån, located in the southern part of Sweden (Östman et al., 2019;Pohl et al., 2018). Samples were collected from four sampling points along the small stream. Two were located before a local STP at P1: 59°43′32.9″ N 17°47′07.8″ E and P2: 59°43′25.9″ N 17°47′13.6″ E, and two after the STP efﬂuent at P3: 59°43′02.4″ N 17°47′30.3″ E, and P4: 59°42′17.0″ N 17°47′31.4″ E (Fig. 1). This STP was previously part of a pilot study in 2016, for 6 months, in which the effect of ozone treatment was investigated (Östman et al., 2019;Pohl et al., 2018). However, no ozone process was run in this STP during 2017–2018. Sampling was made three times in 2018 (4th May, 27th June, and 10th October) to en- able observation of seasonal variations. The water level in the stream varied depending on the season, with the lowest level in June.

Water samples were collected into 1dm³bottles and stored in the freezer until analysis. At all sampling occasions, the water was clear and transparent without visible coloring. Bioﬁlms were sampled by collecting small stones from the river bottom in an area of 10 × 10 cm.

In places where no stones were observed, we collected other sub- merged solids such as hard dead plant material (this happened for P3 and P4 in June, as well as P3 in October). The stones were collected

Fig. 1. The small stream, Knivstaån, with the position of the sewage treatment plant (STP) and the four sampling points.

(3)

using sterile forceps into two different sterile plastic test tubes: A and B, for each location. Tubes labeled A contained stones for cultivation on R2A media (low nutrient medium for freshwater bacteria) (Reasoner and Geldreich, 1985). The R2A medium and our cultivation design were expected to cause selective pressure reducing the number of organisms observed. Thus, we also extracted DNA directly from bioﬁlms on the stones collected in tubes labeled B. In June 2018 we introduced (planted) previously washed and autoclaved stones into the river in all four sampling locations to account for the lack of small stones in some areas of the river bed (Knivstaån runs though areas of clay-type soil). The autoclaved stones were placed on top of a small sterile plastic culture dish at the bottom of the stream. All samples were labeled according to substrate, i.e. natural stones, planted stones or plant material, as a difference in bacterial communities could be expected on the vari- ous substrates.

2.3. Water analysis

All measurements of pharmaceuticals in water were performed using a Thermo TSQ Quantum EMR triple quadrupole mass spectrome- ter according to the method byGrabic et al. (2012). In short, solid-phase extraction (SPE) columns (Oasis HLB, 200 mg, Waters Corporation, Mil- ford, MA, USA) were pre-conditioned and samples were applied to the SPE columns at aflow rate of 5 ml min⁻¹and eluted with 5 ml of meth- anol and 3 ml of ethyl acetate. Eluates were evaporated to 20μl and dis- solved in 5% acetonitrile in water to afinal volume of 1.0 ml (Grabic et al., 2012). A number of 97 substances were quantified in the water (listed in supplementary material). For quality assurance and quality control, two MS/MS transitions were used for positive identifications of analytes with the criterion that the ratio between the transitions was not allowed to deviate more than ±30% from the ratio in the corresponding calibration standard. Retention times for all analytes also had to be within ±2.5% of the retention time in the corresponding calibration standard. Carry-over effects were evaluated by injecting standards at 1000 ng l⁻¹followed by two mobile phase blanks. Several instrumen- tal andfield blanks were included in the analytical runs.

2.4. Microbial analysis

On the day of sampling, all samples in Tubes B were frozen at

−20 °C, while samples in Tubes A were treated for further cultivation.

2.4.1. Cultivation from Tubes A

To separate bioﬁlms from their substrate, the collected stones were vortexed with 1 ml PBS for 5 min. The solution was diluted 1:10 and a volume of 100μl was plated on a diluted medium designed for freshwater samples, 1/10 R2A agar (Reasoner and Geldreich, 1985). Cyclohexi- mide (50μg/ml) was present in the agar plates to avoid fungal growth (Røder et al., 2015). Thereafter, the plates were left for two days at room temperature. After incubation, 15–20 colonies were selected indi- vidually from each plate and suspended into separate sterile test tubes with 1.5 ml R2A broth. The selection of colonies from the plates was done to obtain a large variety of colony morphology and color. The tubes were left to incubate at room temperature for 24 h. Thereafter, glycerol was added (30%ﬁnal content) and the tubes were frozen to

−80 °C. Individual isolates were obtained from the frozen aliquots of bacteria by repeatedly streaking individual colonies on 1/10 R2A plates and incubating for 2–4 days and repeating the process until single isolates were obtained. These isolates were frozen again with glycerol at

−80 °C as pure isolates. This procedure resulted in a collection of 193 individual bacterial isolates.

2.4.2. DNA extraction from Tubes B and from pooled cultures

DNA was extracted directly from the frozen stones byﬁrst sonicating the stones with metal beads for 15 min before extraction using the Qiagen Powerbioﬁlm Kit. DNA extraction from cultured strains were

done by pooling the individual isolates obtained in each sampling point at a speciﬁc month as previously described (Røder et al., 2015).

Shortly, liquid cultures of strains were made by cultivating aliquots from the frozen stock in 1 ml R2A liquid broth overnight. From these overnight cultures, 66–133 μl aliquots from each isolate (15–30 per point) were pooled to form one 2 ml suspension per sampling point and time. DNA was extracted from this suspension using the Qiagen Powerbioﬁlm Kit.

For bacterial 16S rDNA, hypervariable regions V3-V4 was amplified through PCR using forward primer 341f (5′-CCTAYGGGRBGCASCAG- 3′) and reverse primer 806r (5-GGACTACHVGGGTWTCTAAT-3). Tem- perature set for one cycle was: 95 °C for 15 s; 56 °C for 15 s; 72 °C for 30 s. The primers were barcoded so each sample could be uniquely identified post-sequencing in the second PCR. Negative controls were included for the extraction and PCR amplification procedures. In each case, there was no indication of contaminants. Allfinal PCR products were purified using HighPrep™ PCR (MAGBIO, USA), based on para- magnetic beads technology and normalized using SequalPrep™ Nor- malization plate kit (Invitrogen, USA). Further cleaning and concentration were done by using the DNA Clean & Concentrator™-5 Kit (Zymo Research, Irvine, CA, USA). Concentrations were then determined using the Quant-iT™ High-Sensitivity DNA Assay Kit (Life Technologies).

2.4.3. Sequencing

Paired-end sequencing was performed on the Illumina MiSeq Sys- tem (Illumina Inc., CA, USA), including 5% PhiX as an internal control.

All reagents used were from the MiSeq Reagent Kits v3 (Illumina Inc., CA, USA). Automated cluster generation and paired-end sequencing with dual-index reads were performed with 2 × 300 bp. The sequencing output was generated as a demultiplexed fastq-ﬁles for downstream analysis.

2.4.4. Sequence analysis

Primers were removed from the raw paired-end FASTQﬁles generated via MiSeq using“cutadapt” (Martin, 2011). Further, reads were analyzed by QIIME2 (qiime2-2018.11) (Bolyen et al., 2019) pipeline through dada2 to infer the Amplicon Sequence Variants (ASVs) present and their relative abundances across the samples. Using read quality scores for the dataset, forward and reverse reads were truncated at 270 bp and 230 bp, followed by trimming the 5′ end till 7 bp for both forward and reverse reads, respectively; other quality parameters used dada2 default values. For 16S rRNA gene sequencing, taxonomy was assigned using a pre-trained Naïve Bayes classiﬁer (Silva database, release 132, 99% ASV) (Quast et al., 2013).

2.4.5. Statistical analysis

Data analysis was conducted in R (R Core Team, 2017). Initial pre- processing of the ASV table was conducted using the phyloseq package (v1.20.0) (McMurdie and Holmes, 2013). Furtherfiltering was done by removing ASVs classified as chloroplast, mitochondria, or without phylum-level classification, from 16S rRNA sequencing data. ASV table were normalized to generate a relative abundance of taxa present in each sample. All downstream analyses were performed on this normalized ASVs table unless mentioned. We used four alpha diversity indices i.e., observed richness, Shannon diversity index, Simpson and Chao1 index. Furthermore, beta diversity was calculated using unweighted UniFrac metric and visualized by Principal Coordinates Analysis (PCoA) for microbiome analysis. Alpha and beta diversities were calculated using phyloseq v1.20.0 and visualized with ggplot2 v2.2.1 (Wickham, 2016) in R v3.4.1. Comparison of community richness and diversity was assessed by the Kruskal–Wallis test between all the groups and comparison between the two groups were done by Wilcoxon test with Benjamini-Hochberg FDR multiple test correction.

Signiﬁcance testing between the groups for beta diversity was assessed using permutational multivariate analysis of variance (PERMANOVA)

(4)

using the“vegan” package (Oksanen et al., 2007). Principal component analysis (PCA), an unsupervised type of multivariate analysis, was done using Simca v14.0.0.1359 (Umetrics) to get an overview of which variables (sampling location, month, substrate, and DNA isolation directly from stones or from cultivated isolates) could explain the largest variability in the dataset. The key variables were presence or absence of certain bacterial genera (abundance higher than 0.1%) and/or pharmaceuticals (detected or not) without introducing concentration levels.

3. Results

The stream, Knivstaån, was sampled in two locations upstream from the Knivsta sewage treatment plant (STP) (Östman et al., 2019;Pohl et al., 2018) and two locations downstream (Fig. 1). This sampling strat- egy was chosen to investigate the spatial heterogeneity up-stream (P1, P2) as well as downstream (P3, P4) for bioﬁlms present in the river, as well as to capture effects from the STP efﬂuent. P3 is close to the outlet from the STP and P4 further downstream where the water has had time to mix throughout the water body of the small river (Fig. 1). The content of pharmaceuticals in the water was analyzed as well as the microbial composition on solid substrates in the river (Fig. 2, Supplementary Table S1).

3.1. Water chemistry

A clear difference in the content of pharmaceuticals in the water samples could be observed upstream and downstream from the STP (Fig. 2). Between P3 and P4 a dilution of all pharmaceuticals was observed as the water mixed in the stream while traveling. An overall difference in concentration of pharmaceuticals was also seen between the months and appears to mirror the volume of water in the stream. In June, the water level was very low, giving less dilution of the STP efﬂu- ent and, thus, higher concentrations of pharmaceuticals.

3.2. Microbiology

To get an overview of the variability of the bacterial composition as well as pharmaceutical concentration in Knivstaån we used multivariate

analysis. It allowed the variation with respect to all monitored variables to be observed simultaneously. However, when interpreting this PCA model, it is important to remember that in environmental systems there may be other factors influencing the dataset that have not been accounted for by variables available in this study. Thus, the PCA should be considered as an approach to identify differences between or cluster- ing of the samples based on their characteristics. The reasons for cluster- ing need to be subsequently tested and verified by other means. The PCA plot (Fig. 3) describes the factors contributing to differences observed between samples from Knivstaån. It showed the largest separation in the dataset between sampling points upstream or downstream from the STP, which is visualized in thefirst principal component (x- axis) in the PCA plot (Fig. 3a). The second-largest separation in the dataset is visualized as the second component (y-axis). Thesefirst two components of the model explain 50% of the variation (R2X (0.495 cum)) but with fairly low predictability at 30%, Q2 (0.305 cum). R2X and Q2 did not increase much by including more than two components, thus, only thefirst two components were used here. The loading plot of the model (Fig. 3c) indicates that the largest separation in the combined dataset (i.e. thefirst component) is connected to the presence of pharmaceuticals downstream (but not upstream) of the STP. Differences in microbial composition have some influence in this first component but are of lower importance for the separation of data points in the first component. Certain differences in bacterial species composition were seen, indicating that the presence of pharmaceuticals (or other components present in the STP effluent) may have produced a selective pressure beneficial for genera such as Pseudomonas, Serratia, Aeromonas, Acinetobacter, Rahnella, Acidovorax, Pedobacter, Flavobacterium, Entero- bacter. On the other hand possibly creating pressure against bacteria of the genera Crenothrix, Hyphomicrobium, family Methylomonaceae, Ferruginibacter, family Burkholder, Rhodoferax, Novosphingobium, family Rhodobacte, family Sphingomon, and Rhodobacter (Fig. 3).

The second-largest variation (y-axis) appears also to relate to differences in bacterial composition and seems to reﬂect the large diversity of species growing on the substrates in the river both upstream and downstream from the STP (Fig. 3b). This variability is mainly seen for the microbial composition derived using 16S rRNA sequencing directly from the substrates. On the other hand, strains that were isolated by culturing

Fig. 2. Pharmaceutical content of water samples from the different sampling points at the three sampling times. P1 = blue bars, P2 = red bars, P3 = green bars, P4 = orange bars. The sampling periods are plotted chronologically (i.e. starting from May to the left, then June and October to the right) for each pharmaceutical substance. For most substances, the concentrations were higher in June than in May and October. The concentrations were also, in general, highest in P3 (green bars) and below the detection limit upstream of the STP, i.e. in P1 and P2 (blue and red bars are at baseline in the plot, except for Venlafaxine that was also detected in P2). (For interpretation of the references to color in thisﬁgure legend, the reader is referred to the web version of this article.)

(5)

showed a lower variability (green vs blue points inFig. 3b) as could be expected. The reduction in variability between isolates and stones is not surprising as not all colonies growing on the initial plates were pu- riﬁed into isolates and the procedure of cultivation selects for strains that can grow under our speciﬁc conditions: i.e. on the R2A medium, at room temperature and fast enough to appear on the culture plate within two days. However,Figs. 3b &4indicates that the cultivation procedure enabled us to cultivate both strains predominantly growing upstream and strains predominantly downstream of the STP. No systematic variation in the dataset with respect to bacterial composition was observed following sampling month (Fig. S1).

As the riverbed lacked small stones suitable for sampling biofilm in some sampling locations (especially P3), we planted washed and autoclaved stones into the river at the June sampling occasion. Thereaf- ter, these stones were sampled in parallel with natural stones in the Oc- tober sampling and the species composition compared. Unfortunately, the stones planted in P3 had disappeared at the time of the October sampling. The reason is unknown, but we hypothesize they had either sunken into the clay bed at the bottom of the stream, or they had been removed by someone from the public that observed this unusual object in the small stream and removed it.Fig. 5shows a degree of similarity between the species identified from stones that were found nat- urally and those that were placed in the river. To investigate this more closely, alpha and beta diversity indices were determined (Fig. 6). The score for PCoA1 and PCoA2 account for 23.9% of the variance in the data. This difference between the bacterial communities was significant, as determined using the ANOSIM nonparametric statistical test analysis of similarity, where R = 0.128 (p = 0.005). Alpha diversity indices (Ob- served, Shannon, Chao1 and Simpson) demonstrated that, as expected, microbial diversity in planted and natural stones was higher than what

was observed from the collection of isolates (Fig. 6a). The observed alpha diversity index suggests somewhat higher richness for natural stones compared to planted stones in October, as well as an increase in richness through the summer months on the natural stones. The very small dataset for the planted stones gives rise to a hypothesis that the richness of taxa is obtained slower downstream of the STP compared to upstream. However, a larger study would be needed to test this hypothesis. To explore the differences in the overall microbial community composition across isolates, natural stones and planted stones, the unweighted Unifrac beta diversity was calculated. The PCoA plot shows that planted and natural stones had similar microbial diversity (Fig. 6b).

The isolated strains showed a diversity that differed to some extent from the stones, which is understandable as not all initial colonies were selected and puriﬁed and only a minor fraction of all strains present may have been cultivated.

4. Discussion

The levels of pharmaceuticals in the small stream, Knivstaån, were in the range reported for pharmaceuticals in fresh water in Europe (Voulvoulis et al., 2016) and similar or slightly lower than previously reported values from the same stream (Pohl et al., 2018). The content in Knivstaån is also higher than the concentration reported for marine en- vironments (Björlenius et al., 2018).

The time-resolved pharmaceutical data for P3 and P4 in Knivstaån mirrors the changes observed in the water level of the small stream, where the water levels dropped during summer resulting in lower levels of dilution of the effluent from the STP and, thus, higher concentrations were detected (Fig. 2). The dilution and mixing of water along the stream can also be observed as the Fig. 3. a) PCA scatter plot color-coded for sampling points where the points upstream (P1 and P2) are found to the left and the downstream points (P3 and P4) to the right of the plot. b) the same score plot color-coded to show isolated strains (green) and“all” strains (blue) c) Loading plot for a and b) where pharmaceuticals are found to the right of the plot showing that their presence give a large contribution to thefirst component of the PCA. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

(6)

concentrations decrease downstream from the STP from P3 to P4.

This is likely a mixing factor possibly combined with an inﬂux of non-contaminated ground and surface water along the course of the small stream. The fact that the majority of pharmaceuticals shown in Fig. 2 could not be detected upstream in P1 and P2 indicates that the point source for the pharmaceuticals in the

river is wastewater handled in the STP. Some seasonal variations could be observed in some of the pharmaceuticals, where irbetan (antihypertension),ﬂuconazole (antifungal), bisoprolol and atenolol (beta-blockers) were highest in the spring sampling. These trends probably reﬂect changes in household consumption in the town feeding wastewater into the STP.

Fig. 4. Relative abundance of strains in different sampling and time points. Strains present with an abundance lower than 1% are not shown and represent the gap between the top of the bar and 100% (abundance 1 = 100%). First bar in each sampling and time point represents strains observed in pooled isolates and the second bar represents the strains observed from DNA sequencing directly from riverbed substrates. Missing bars represent samples that were discarded due to poor data quality during sequencing. (For interpretation of the references to color in thisﬁgure legend, the reader is referred to the web version of this article.)

Fig. 5. Relative abundance of bacterial genera on natural stones present in the river as well as planted autoclaved stones that had been in the river for more than two months. Difference between top of the bar and abundance 1.0 represents bacteria with abundance < 1%. (For interpretation of the references to color in thisﬁgure legend, the reader is referred to the web version of this article.)

(7)

The PCA analysis of the dataset indicates that STP effluent may have influenced the microbial composition down-stream the STP both with respect to the entire bacterial community as well as the composition obtained during the culturing of bacteria. Similar results have been pro- posed in a recent study of a river system in Illinois investigating the effect of a different group of micropollutants on both water microbiology and microbiology in the river bed (Gao et al., 2019) (discussed in more detail below). The STP effluent of course has the potential to change the water quality in several ways and, in this study, we have fo- cused mainly on pharmaceutical content.

We did not observe clear systematic differences between species composition between the three sampling months. We also investigated the composition based on substrate type, however, consider- ing that plant material mainly was collected in P3 (due to the absence of stones at the river bed) it is not possible to separate the correlation relating to sampling position and/or substrate type. In order to understand the difference in composition depending on the substrate, another sampling design would have been needed where several types of substrates would be collected in every sampling point and thereafter the composition of bacteria compared.

However, that was outside the scope of this study. The similarities between the bacterial-community composition and diversity on original stones and planted stones (Figs. 5 & 6) suggests that the method of placing substrates in advance may be a beneﬁcial methodology for sampling bioﬁlm.

The interaction between pharmaceuticals and bacteria in fresh water systems is not well understood (Gao et al., 2019). Previous studies have shown that drugs such as fexofenadine, ibuprofen, paracetamol, acetaminophen, doxycycline, tetracycline, oﬂoxacin, triclosan, and sul- famethoxazole are capable of affecting nitrogen and/or phosphorus cycling (Alvarino et al., 2014;Dokianakis et al., 2004;Katsou et al., 2016;

Jonsson et al., 2015). One study found that strains from genera such as Hydrogenophaga, and Flavobacterium were correlated positively with some micropollutants, possibly since these genera have been shown to produce enzymes capable of degrading different pollutants (Gao et al., 2019). The latter was also observed in our study, where wefind an abundance of Flavobacterium. However, in Knivstaån, it seems that strains identified as belonging to Hydrogenophaga were not correlated to increased amounts of pharmaceuticals (Fig. 3). As the two studies do not investigate the same group of pollutants, the difference is likely Fig. 6. Alpha and beta diversity analysis for different samples. A) Alpha diversity in samples collected in sampling points 1–4 in the months May, June and October. B) Principal coordinate analysis (beta diversity) of unweighted UniFrac distances, showing diversity between samples collected at different sampling points and months, M = May, J = June and O = October. (For interpretation of the references to color in thisfigure legend, the reader is referred to the web version of this article.)

(8)

related to specific degradation pathways for pollutants that may differ depending on which pollutant is investigated. Flavobacterium was also found to be positively correlated to metal pollution in a Spanish river system, together with Ferruginibacter and a few other genera (Argudo et al., 2020). In our study, Ferruginibacter was not positively correlated to pharmaceutical contamination, again suggesting that bacteria- specific detoxification and/or degradation processes or strain specific- ities may be important factors influencing the observed correlations as well as the type of selection pressure that pollutants place on freshwater microbial systems.

Not surprisingly, we observed that our cultivation and isolation procedure reduced the diversity of bacterial species. Strains that dominated among bacterial isolates in this collection were assigned to be: Enterobacter, Pseudomonas, Serratia, Aeromonas, Acinetobacter, Pantoea, Paenibacillus, and Flavobacterium genera (Fig. 3). In contrast, bacteria such as Rhodobacter, Luteolibacter, Pseudorhodobacter, Novosphingobium, family Burkholder, family Sphingomonas, Pseudoarthrobacter, Hydrogenophaga, Crenothrix, Nitrospira, Rhodoferax, Hyphomicrobium, and family Methylomonaceae, appear to have been lost in our cultivation and selection process (Fig. 3). In our selection procedure, we tried to pick colonies of many different morphologies, thus, it may be that some species were not selected as they formed colonies of very similar phenotype (e.g. small, white, semi- transparent colonies) or because they were present at low numbers.

A closer study of the composition with respect to abundance in the different sampling points and evenness between taxa (Figs. 4 & 6a) shows a selective pressure was applied during our cultivation. It resulted in low alpha diversity indices and different community composition of cultivated bacteria in comparison to those present on the stones (Fig. 6). Flavobacter species appear to have experienced a positive selection pressure, resulting in high abundance during cultivation. It is important to note that the abundance of the pooled isolated strains does not necessarily reﬂect the abundance of the same strains in the river, as we can see in the more even distribution of taxa in the Shannon alpha diversity index (Fig. 6a). Instead, the un- evenness observed in isolates is a result of the selective pressure placed on the consortium by our culturing conditions in combination with our selection criteria.

Among not-cultured bacteria with higher prevalence upstream were Rhodobacter and Rhodoferax that are both photosynthetic Gram-negative bacteria (Fig. 3). Some species of Sphingomonas were found as well as Hydrogenophaga. These genera have been reported to include strains that are facultative photoorganotroph (the former) and that use CO2as carbon source and H2as an energy source (the latter) (Whitman, 2012). Downstream bacteria that we cultured included families that contain species that may reduce ni- trate and thereby influence nitrogen cycling, e.g. Pedobacter and Flavobacterium (Whitman, 2012). Nitrospira was present predominantly downstream (but not cultured) (Fig. 3) in accordance with previous publications suggesting an increase in species involved in nitrogen removal in presence of effluent from STP (Aubertheau et al., 2017). Some genera that we observed at the riverbed downstream from this specific STP have previously been observed in effluent water from the STP during previous sampling campaigns.

These included genera such as Acidovorax, Aeromonas, Pedobacter, Flavobacterium and family Micrococca (personal communication Joakim Larsson, University of Gothenburg). Thus, some of the differences in consortium composition observed downstream could be an effect of bacterial inoculation from the STP as well as an adaptive response to water composition.

Freshwater bioﬁlms have been reported to be dominated by the phyla cyanobacteria, Proteobacteria, and Bacteroides (Besemer, 2015;

Argudo et al., 2020). This was also partly observed in Knivstaån al- though the abundance of cyanobacteria was low until October. Instead, Actinobacteria and Firmicutes showed high abundance in several sampling points (Supplementary Fig. S2). Possibly the photosynthetic

bacteria Rhodobacter (from Proteobacteria) are occupying more of the niche that cyanobacteria had in other reports. An explanation given for the abundance of Bacteroides and Alphaproteobacteria in fresh water streams is that they have been reported to degrade humic substances and also other types of complex macromolecules that form part of decaying matter in fresh water (Besemer, 2015). Similar to previous studies we observed bacteria from the classes Actinobacteria, Alphaproteobacteria, Deltaproteobacteria, Gammaproteobacteria, and Verrucomicrobiae (supplementary Fig. S3). However, the abundance of Betaproteobacteria was high in the study byGao et al. (2019)whereas it was not predominant in Knivstaån.

We aimed to study differences in diversity connected to the presence of pharmaceuticals in the stream originating from the STP.

However, the diversity indices only show a small tendency of difference in microbial diversity upstream and downstream for the STP, and only for the June samples (Fig. 6b). Possibly this reflects the higher levels of pharmaceuticals detected in the stream water for this month or is an effect caused by the substrate type as stones were not found and, hence, were replaced by hard plant material in P3 and P4 in June (Fig. 2). The richness in taxa appeared similar between the sampling points and increased through the summer months (Fig. 6a), thus, contrary to other studies, the presence of STP effluent did not seem to decrease the richness of taxonomical structures (Besemer, 2015;Corcoll et al., 2015) but may have shifted the composition of genera present. The similarity in diversity between introduced stones and natural stones (Fig. 6b) indicates that the approach to add sterile stones a few months ahead of sampling may be a suitable method for sampling biofilms in rivers where there is a shortage of suitable and comparable substrates between sampling points.

From the observed differences, we hypothesize that the presence of effluent from the STP may represent a negative selection pressure against some photosynthetic bacteria in the stream similar to what has been reported from France (Aubertheau et al., 2017). This may have caused a shift to other photosynthetic bacterial species or microorganisms. Our analysis did not include algae, thus, it may be that the disappearance of photosynthetic bacteria was compensated for by algae in river biofilms downstream from the STP (Corcoll et al., 2015). The STP effluent contains traces of substances other than pharmaceuticals (that may not be present upstream) that could po- tentially also influence the microbial consortium composition (Sabater-Liesa et al., 2019). More in-detail studies under controlled conditions (e.g. in vitro) are needed in order to decipher the relative effect that may be caused by changes in the content of nutrients and pharmaceuticals from effluent discharge into fresh water streams, as well as other influences that may have been present due to changes in sediment composition etc.

The differences observed, between species composition on the stones and in the isolates, clearly show that any in vitro model using isolated bacterial strains will be a simplification of what is present in the natural setting. The natural biofilm consortium appears to consist of a very large amount of different species of low abundance, as seen in the very large amount of species at an abundance < 1% (Fig. 4). This suggests that these biofilm consortia form very complex bacterial ecosystems in themselves. This complexity is challenging (or impossible) to reproduce in lab studies and some simplification of the system is needed when in vitro models are made. However, with this in mind, we believe that in vitro models using river isolates from the same location would be a very useful tool to better understand what types of processes and mechanisms may occur in presence of pollutants. Such model systems could give clues to decipher interactions that occurs between bacteria in biofilms, as well as how these are influenced, and influence, pharmaceuticals and other contaminants in the surrounding water. Provided with such detailed knowledge, we could better understand the com- plexities giving rise to differences in bacterial consortia, like those described here.

(9)

5. Conclusions

This case study of the small stream, Knivstaån, in central Sweden indicates that the presence of STP effluent, containing low concentrations of pharmaceuticals, affects bacterial composition in bacterial biofilms present at the bottom of the stream. From the data, it is clear that the microbial ecology in the biofilms formed in the stream is very complex and includes large numbers of species present at low levels. Based on the PCA analysis we hypothesize that pharmaceuticals and/or nutrient levels present a negative selection pressure for photosynthetic bacteria such as Rhodobacter and thereby may influence the primary production in the stream, even if the richness of bacterial taxa did not seem to change. Further studies are needed to better understand the processes and microbial ecology that underlie these observations, e.g. well- controlled lab experiments where exposure more directly can be linked to bacterial biofilm response, in comparison to biofilm responses to the dynamic natural environment of a river. Furthermore, in addition to changes in the community composition, there may be functional adap- tations within the biofilms, such as change of biofilm phenotype, EPS composition and metabolism, that are not captured through the metagenomics analysis (Besemer, 2015). This is the focus of our future work using the river isolates presented here.

Supplementary material consists of additional information for the Materials and methodssection, supplementary tables with raw data and supplementaryﬁgures showing alterative presentations of the data in order to facilitate interpretation and comparisons with previous studies. Supplementary data to this article can be found online at https://doi.org/10.1016/j.scitotenv.2020.142991.

Declaration of competing interest

The authors declare that they have no known competingﬁnancial interests or personal relationships that could have appeared to inﬂu- ence the work reported in this paper.

Acknowledgments

The Swedish Research Council Formas (grant 2017-00403) and Umeå University are acknowledged for funding. Prof Joakim Larsson at Department of Infectious Diseases, Institute of Biomedicine, The Sahlgrenska Academy at the University of Gothenburg, and Prof Tomas Brodin at the Department of Wildlife, Fish and Environmental Studies at the Swedish Agricultural University (SLU) are acknowledged for discus- sions. Dr Dmitry Shevela is acknowledged for the graphical abstract and Fig. 1.

References

Alvarino, T., Katsou, E., Malamis, S., Suarez, S., Omil, F., Fatone, F., 2014.Inhibition of bio- mass activity in the via nitrite nitrogen removal processes by veterinary pharmaceuticals. Bioresour. Technol. 152, 477–483.

Argudo, M., Gich, F., Bonet, B., Espinosa, C., Gutiérrez, M., Guasch, H., 2020.Responses of resident (DNA) and active (RNA) microbial communities inﬂuvial bioﬁlms under different polluted scenarios. Chemosphere 242, 125108.

Aubertheau, E., Stalder, T., Mondamert, L., Ploy, M.-C., Dagot, C., Labanowski, J., 2017.Im- pact of wastewater treatment plant discharge on the contamination of river bioﬁlms by pharmaceuticals and antibiotic resistance. Sci. Total Environ. 579, 1387–1398.

Berendonk, T.U., Manaia, C.M., Merlin, C., Fatta-Kassinos, D., Cytryn, E., Walsh, F., Burgmann, H., Sorum, H., Norstrom, M., Pons, M.-N., Kreuzinger, N., Huovinen, P., Stefani, S., Schwartz, T., Kisand, V., Baquero, F., Martinez, J.L., 2015.Tackling antibiotic resistance: the environmental framework. Nat. Rev. Micro 13, 310–317.

Besemer, K., 2015.Biodiversity, community structure and function of bioﬁlms in stream ecosystems. Res. Microbiol. 166, 774–781.

Björlenius, B., Ripszám, M., Haglund, P., Lindberg, R.H., Tysklind, M., Fick, J., 2018.Pharma- ceutical residues are widespread in Baltic Sea coastal and offshore waters– screening for pharmaceuticals and modelling of environmental concentrations of carbamaze- pine. Sci. Total Environ. 633, 1496–1509.

Bolyen, E., Rideout, J.R., Dillon, M.R., Bokulich, N.A., Abnet, C.C., Al-Ghalith, G.A., Alexander, H., Alm, E.J., Arumugam, M., Asnicar, F., Bai, Y., Bisanz, J.E., Bittinger, K., Brejnrod, A., Brislawn, C.J., Brown, C.T., Callahan, B.J., Caraballo-Rodríguez, A.M., Chase, J., Cope, E.K., Da Silva, R., Diener, C., Dorrestein, P.C., Douglas, G.M., Durall, D.M., Duvallet, C., Edwardson, C.F., Ernst, M., Estaki, M., Fouquier, J., Gauglitz, J.M., Gibbons, S.M.,

Gibson, D.L., Gonzalez, A., Gorlick, K., Guo, J., Hillmann, B., Holmes, S., Holste, H., Huttenhower, C., Huttley, G.A., Janssen, S., Jarmusch, A.K., Jiang, L., Kaehler, B.D., Kang, K.B., Keefe, C.R., Keim, P., Kelley, S.T., Knights, D., Koester, I., Kosciolek, T., Kreps, J., Langille, M.G.I., Lee, J., Ley, R., Liu, Y.X., Loftﬁeld, E., Lozupone, C., Maher, M., Marotz, C., Martin, B.D., McDonald, D., McIver, L.J., Melnik, A.V., Metcalf, J.L., Morgan, S.C., Morton, J.T., Naimey, A.T., Navas-Molina, J.A., Nothias, L.F., Orchanian, S.B., Pearson, T., Peoples, S.L., Petras, D., Preuss, M.L., Pruesse, E., Rasmussen, L.B., Rivers, A., Robeson, M.S., Rosenthal, P., Segata, N., Shaffer, M., Shiffer, A., Sinha, R., Song, S.J., Spear, J.R., Swafford, A.D., Thompson, L.R., Torres, P.J., Trinh, P., Tripathi, A., Turnbaugh, P.J., Ul-Hasan, S., van der Hooft, J.J.J., Vargas, F., Vázquez-Baeza, Y., Vogtmann, E., von Hippel, M., Walters, W., Wan, Y., Wang, M., Warren, J., Weber, K.C., Williamson, C.H.D., Willis, A.D., Xu, Z.Z., Zaneveld, J.R., Zhang, Y., Zhu, Q., Knight, R., Caporaso, J.G., 2019.Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2. Nat. Biotechnol. 37, 852–857.

Boxall, A., Rudd, M., Brooks, B., Caldwell, D., Choi, K., Hickmann, S., Innes, E., Ostapyk, K., Staveley, J., Verslycke, T., Ankley, G., Beazley, K., Belanger, S., Berninger, J., Carriquiriborde, P., Coors, A., DeLeo, P., Dyer, S., Ericson, J., Gagne, F., Giesy, J., Gouin, T., Hallstrom, L., Karlsson, M., Larsson, D., Lazorchak, J., Mastrocco, F., McLaughlin, A., McMaster, M., Meyerhoff, R., Moore, R., Parrott, J., Snape, J., Murray-Smith, R., Servos, M., Sibley, P., Straub, J., Szabo, N., Topp, E., Tetreault, G., Trudeau, V., Van Der Kraak, G., 2012.Pharmaceuticals and personal care products in the environment:

what are the big questions? Environ. Health Perspect. 120, 1221–1229.

Burmølle, M., Ren, D., Bjarnsholt, T., Sørensen, S.J., 2014.Interactions in multispecies bioﬁlms: do they actually matter? Trends Microbiol. 22, 84–91.

Corcoll, N., Casellas, M., Huerta, B., Guasch, H., Acuña, V., Rodríguez-Mozaz, S., Serra- Compte, A., Barceló, D., Sabater, S., 2015.Effects ofﬂow intermittency and pharmaceutical exposure on the structure and metabolism of stream bioﬁlms. Sci. Total En- viron. 503–504, 159–170.

Dokianakis, S.N., Kornaros, M.E., Lyberatos, G., 2004.On the effect of pharmaceuticals on bacterial nitrite oxidation. Water Sci. Technol. 50, 341–346.

Fick, J., Brodin, T., Heynen, M., Klaminder, J., Jonsson, M., Grabicova, K., Randak, T., Grabic, R., Kodes, V., Slobodnik, J., Sweetman, A., Earnshaw, M., Barra Caracciolo, A., Lettieri, T., Loos, R., 2017. Screening of benzodiazepines in thirty European rivers.

Chemosphere 176, 324–332.

Flemming, H.-C., Wingender, J., Szewzyk, U., Steinberg, P., Rice, S.A., Kjelleberg, S., 2016.

Bioﬁlms: an emergent form of bacterial life. Nat. Rev. Microbiol. 14, 563–575.

Gao, H., LaVergne, Jeanne M., Carpenter, C.M.G., Desai, R., Zhang, X., Gray, K., Helbling, D.E., Wells, G.F., 2019.Exploring co-occurrence patterns between organic micropollutants and bacterial community structure in a mixed-use watershed. Environ Sci Process Impacts 21, 867–880.

González-Alonso, S., Merino, L.M., Esteban, S., López de Alda, M., Barceló, D., Durán, J.J., López-Martínez, J., Aceña, J., Pérez, S., Mastroianni, N., Silva, A., Catalá, M., Valcárcel, Y., 2017.Occurrence of pharmaceutical, recreational and psychotropic drug residues in surface water on the northern Antarctic Peninsula region. Environ. Pollut. 229, 241–254.

Grabic, R., Fick, J., Lindberg, R.H., Fedorova, G., Tysklind, M., 2012.Multi-residue method for trace level determination of pharmaceuticals in environmental samples using liquid chromatography coupled to triple quadrupole mass spectrometry. Talanta 100, 183–195.

Jonsson, M., Ershammar, E., Fick, J., Brodin, T., Klaminder, J., 2015.Effects of an antihista- mine on carbon and nutrient recycling in streams. Sci. Total Environ. 538, 240–245.

Katsou, E., Alvarino, T., Malamis, S., Suarez, S., Frison, N., Omil, F., Fatone, F., 2016.Effects of selected pharmaceuticals on nitrogen and phosphorus removal bioprocesses. Chem.

Eng. J. 295, 509–517.

Küster, A., Adler, N., 2014.Pharmaceuticals in the environment: scientiﬁc evidence of risks and its regulation. Philos. Trans. R. Soc. B Biol. Sci. 369, 20130587.

Lagesson, A., Fahlman, J., Brodin, T., Fick, J., Jonsson, M., Byström, P., Klaminder, J., 2016.

Bioaccumulation ofﬁve pharmaceuticals at multiple trophic levels in an aquatic food web - insights from aﬁeld experiment. Sci. Total Environ. 568, 208–215.

Lear, G., Dopheide, A., Ancion, P.-Y., Roberts, K., Washington, V., Smith, J., Lewis, G.D., 2012.Bioﬁlms in freshwater: their importance for the maintenance and monitoring of freshwater health. In: Lear, G., Lewis, G. (Eds.), Microbial Bioﬁlms Current Research and Applications. Caister Academic Press, Norfolk, UK, pp. 129–151.

Lindberg, R., Ostman, M., Olofsson, U., Grabic, R., Fick, J., 2014.Occurrence and behaviour of 105 active pharmaceutical ingredients in sewage waters of a municipal sewer collection system. Water Res. 58, 221–229.

Loos, R., Carvalho, R., Antonio, D., Cornero, S., Locoro, G., Tavazzi, S., Paracchini, B., Ghiani, M., Lettieri, T., Blaha, L., Jarosova, B., Voorspoels, S., Servaes, K., Haglund, P., Fick, J., Lindberg, R., Schwesig, D., Gawlik, B., 2013.EU-wide monitoring survey on emerging polar organic contaminants in wastewater treatment plant efﬂuents. Water Res. 47, 6475–6487.

Martin, M., 2011.Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet.journal 17, 10–12.

McMurdie, P.J., Holmes, S., 2013.phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data. PLoS One 8, e61217.

Oksanen, J., Kindt, R., Legendre, P., O’Hara, B., Simpson, G.L., Solymos, P., Stevens, M.H.H., Wagner, H., 2007.The vegan package. Commun. Ecol. Pack. 10, 719.

Östman, M., Björlenius, B., Fick, J., Tysklind, M., 2019.Effect of full-scale ozonation and pilot-scale granular activated carbon on the removal of biocides, antimycotics and an- tibiotics in a sewage treatment plant. Sci. Total Environ. 649, 1117–1123.

Pohl, J., Björlenius, B., Brodin, T., Carlsson, G., Fick, J., Larsson, D.G.J., Norrgren, L., Örn, S., 2018.Effects of ozonated sewage efﬂuent on reproduction and behavioral endpoints in zebraﬁsh (Danio rerio). Aquat. Toxicol. 200, 93–101.

Quast, C., Pruesse, E., Yilmaz, P., Gerken, J., Schweer, T., Yarza, P., Peplies, J., Glöckner, F.O., 2013.The SILVA ribosomal RNA gene database project: improved data processing and web-based tools. Nucleic Acids Res. 41, D590–D596.

(10)

R Core Team, 2017. R: A Language and Environment for Statistical Computing.https://

www.R-project.org/. (Accessed 20 October 2020).

Reasoner, D.J., Geldreich, E.E., 1985.A new medium for the enumeration and subculture of bacteria from potable water. Appl. Environ. Microbiol. 49, 1–7.

Røder, H.L., Raghupathi, P.K., Herschend, J., Brejnrod, A., Knøchel, S., Sørensen, S.J., Burmølle, M., 2015.Interspecies interactions result in enhanced bioﬁlm formation by co-cultures of bacteria isolated from a food processing environment. Food Microbiol. 51, 18–24.

Roig, B., D’Aco, V., 2016.Distribution of pharmaceutical residues in the environment.

Pharmaceuticals in the Environment. The Royal Society of Chemistry, pp. 34–69.

Romani, A.M., Artigas, J., Ylla, I., 2012.Extracellular enzymes in aquatic bioﬁlms: microbial interactions versus water quality effects in the use of organic matter. In: Lear, G., Lewis, G. (Eds.), Microbial Bioﬁlms Current Research and Applications. Caister Aca- demic Press, Norfolk, UK, pp. 153–174.

Sabater-Liesa, L., Montemurro, N., Font, C., Ginebreda, A., González-Trujillo, J.D., Mingorance, N., Pérez, S., Barceló, D., 2019.The response patterns of stream bioﬁlms to urban sewage change with exposure time and dilution. Sci. Total Environ. 674, 401–411.

Segura, P., Francois, M., Gagnon, C., Sauve, S., 2009.Review of the occurrence of anti- infectives in contaminated wastewaters and natural and drinking waters. Environ.

Health Perspect. 117, 675–684.

Seviour, T., Derlon, N., Dueholm, M.S., Flemming, H.-C., Girbal-Neuhauser, E., Horn, H., Kjelleberg, S., van Loosdrecht, M.C.M., Lotti, T., Malpei, M.F., Nerenberg, R., Neu, T.R., Paul, E., Yu, H., Lin, Y., 2019.Extracellular polymeric substances of bioﬁlms: suffering from an identity crisis. Water Res. 151, 1–7.

Stoodley, P., Sauer, K., Davies, D.G., Costerton, J.W., 2002.Bioﬁlms as complex differenti- ated communities. Annu. Rev. Microbiol. 56, 187–209.

Taylor, D., 2016.The pharmaceutical industry and the future of drug development. Phar- maceuticals in the Environment. The Royal Society of Chemistry, pp. 1–33.

Voulvoulis, N., Barcelo, D., Verlicchi, P., 2016.Pharmaceutical residues in sewage treatment works and their fate in the receiving environment. In: Hester, R., Harrison, R.

(Eds.), Pharmaceuticals in the Environment. The Royal Society of Chemistry, Cam- bridge, UK, pp. 120–179.

Whitman, W.B., 2012.Bergey’s Manual of Systematics of Archaea and Bacteria (BMSAB).

John Wiley & Sons, Inc.

Wickham, H., 2016.ggplot2: Elegant Graphics for Data Analysis. 2nd ed. Springer, Hous- ton, Texas, USA.