Introduction

Large amounts of contaminant mixtures of varied composition enter the environment as—for example—industrial waste, field runoff, municipal and domestic waste and effluents (Farré and Barceló 2003). Despite of intensive treatment of wastewater, residual contamination still has a significant impact on aquatic biota as well as on human health (Heberer 2002). The most studied groups of aquatic pollutants include halogenated persistent organic pollutants (POPs), polycyclic aromatic hydrocarbons (PAHs), toxic metals, pesticides and other micropollutants such as pharmaceuticals and personal care products (Deblonde et al. 2011). Despite of existing diversity and documented adverse effects, current legislation such as the European Water Framework Directive (EU WFD) monitors and regulates only a limited number—currently 45—of so-called priority pollutants (Directive 2008/105/EC 2008).

Routine monitoring efforts of water quality in Europe currently combine two approaches, namely chemical analyses of target compounds (the priority pollutants) to evaluate chemical status and in situ assessment of biota to evaluate the biological status (Directive 2000/60/ES). Direct assessment of potential adverse (toxic) effects of water or effluents by applying bioassays is currently lacking in the EU water legislation despite of scientific recommendations and running effect-based monitoring efforts (Schulze et al. 2015; Wernersson et al. 2015; Escher et al. 2013).

While the chemical analyses inform about the identity of the chemical substances in the sample, they provide only minimum information about the overall toxicity (Wolska et al. 2007). This limitation can be overcome by applying effect-based tools and bioassays. A number of aquatic bioassays have been evaluated for testing of complex samples including diverse model organisms such as bacteria, algae, plants, invertebrates, embryonal and adult fish or amphibians as well as cell-based in vitro models relevant for human health endpoints (Rosal et al. 2010). For example, a recent international study (Escher et al. 2014) compared a series of bioassays for wastewater assessment that included acute (cyto)toxicity endpoints, biomarkers of chronic effects (such as metabolic pathways activation or receptor modulations) as well as rapid bacterial luminescence tests. Among the other assays, bacterial bioluminescence tests employing marine Vibrio fischeri (now reclassified as Aliivibrio fischeri; Urbanczyk et al. 2007) have been found highly sensitive (Escher et al. 2014). Their low costs and fast responses make them suitable for broader use in water quality assessment (Nohava et al. 1995; Wolska et al. 2007).

Bacterial bioluminescence tests have been used for decades. The most widely used model employs V. fischeri and it has been routinely employed in testing of chemicals as well as environmental samples like wastewater effluents or contaminated soil (Rodriguez-Ruiz et al. 2015, Pandard et al. 2006). The original V. fischeri test can be used for clean and transparent samples but it has limitations when testing coloured or turbid materials that interfere with the emitted bacterial bioluminescence. To overcome this limitation, a solid-phase variant of the test has been developed (Volpi Ghirardini et al. 2009) but it has also been shown to provide false positives due to the protocol, where a fraction of bacteria is adsorbed on the studied matrix and removed from the test (Lappalainen et al. 1999; Campisi et al. 2005). A suitable alternative for testing of coloured samples is based on kinetic bioluminescence measurements known as the Flash assay (Lappalainen et al. 1999; 2001). The immediate flash luminescence (measured during seconds after the injection of bacteria into the sample) is affected only by the colour or turbidity, while the actual toxic effect is recorded after prolonged exposure. This kinetic approach, in which each sample serves as its own control for colour or turbidity, has also been standardized by ISO (ISO 21338:2010). Another limitation of using marine bacterial species such as V. fischeri is the need to add relatively high concentrations of salt to the studied samples. This manipulation of the sample can change its properties and alter the toxicity (Dunlap 1985, Cook et al. 2000; Deheyn et al. 2004).

Previous research addressing the above-described problems has explored various approaches such as insertion of luminescence Lux operon into the freshwater bacteria Escherichia coli (Kurvet et al. 2011), Pseudomonas putida (Stewart and Williams 1992), or the cyanobacteria Anabaena spp. (Rosal et al. 2010). In addition to genetically modified bacteria, other naturally bioluminescent species such as Photorhabdus luminescens could be used as an alternative for toxicity testing (Tabei et al. 2013). This species was isolated from the gastrointestinal tract of nematodes and it can be cultured in general microbiological media, i.e. without a need for elevated salt concentrations (Thomas and Poinar 1979). During the environmental monitoring, it is also important to minimize sample manipulations such as additions of stabilizers, drying, sieving and prolonged storage or extractions, which could lead to volatilization, transformation or degration of chemicals. Therefore, rapid portable protocols for direct field in situ toxicity assessment have been developed, such as the assessment of algal chlorophyll fluorescence (Kumar et al. 2014; Muller et al. 2008).

The aim of the present study was to assess a rapid screening bacterial luminescence test for field in situ toxicity assessment using a freshwater species P. luminescens and compare its performance and sensitivity with marine V. fischeri using a set of model toxicants and 43 wastewater effluents collected and chemically characterized in the course of pan-European monitoring campaigns.

Material and methods

Design of the study

First, the sensitivity of P. luminescens was compared with V. fischeri using a set of model chemicals and a standardized protocol (ISO 11348-3, 2009). In addition, the sensitivities of both P. luminescens and V. fischeri were tested also in the kinetic flash format (comparing the laboratory microwell plate arrangements vs. portable luminometer). Finally, the toxicity of 43 wastewater effluents collected during a pan-European monitoring study was evaluated by both microorganisms using the portable luminometer protocol, and the toxicity was compared with the results of chemical analyses of 150 organic and 20 inorganic contaminants.

Chemicals

Potassium dichromate (K2Cr2O7, purity 99.5 %); phenol (purity 99 %); and mercury chloride (HgCl2 purity 99.5 %) were purchased from Lach–Ner s.r.o (Neratovice, Czech Republic). 3,5-dichlorphenol (purity 97 %), 2-4-dichlorphenol (99 % GC purity grade), 2,3-dichloranilin (purity 99 %), and ethanol (purity 97 %) were purchased from Sigma-Aldrich (Steinheim, Germany). Sodium dodecyl sulphate (SDS, research grade purity) was from Serva GmbH (Heidelberg, Germany).

Environmental samples

Wastewater treatment plant (WWTP) effluents were collected during a pan-European study of the Joint Research Centre (EC-JRC) in Ispra, Italy (Loos et al. 2012). The 43 samples investigated in the present study covered WWTPs utilizing different treatment technologies from 13 European countries in different climatic regions. Each sample was thoroughly characterized for a range of 150 organic and 20 inorganic contaminants. Detailed information on the samples, WWTPs and methods of chemical analyses can be found in the publically available report and accompanying research paper (Loos et al. 2012, 2013). The present study used the same coding of the samples as described in the original report (Loos et al. 2012). Wastewater samples were kept frozen at −20 °C, thawed prior to toxicity testing, homogenized by vortexing and tested for toxicity as described below (direct testing of 1× concentrated sample without dilutions).

Bacterial strains

Vibrio fischeri (lyophilized NRRL-B-11177, LUMIStox, Hach Lange, Düsseldorf, Germany) were rehydrated according to the manufacturer’s instructions prior to analysis and kept on ice until the test. P. luminescens CCM7077 was obtained from the Czech collection of microorganisms (Masaryk University, Brno, Czech Republic), and was cultured on standard Tryptic soy agar in Petri dishes at 30 °C. Ready-to-use batch of bacterial suspensions were prepared as follows: after 24-h incubation on agar, five colonies were transferred into 1.5 mL of freezing medium (tryptic soy broth with 5 % of glycerol) containing several sterile glass beads and gently agitated. Glass beads coated with bacteria were aliquoted and kept at −80 °C as inoculum for the experiments. Twenty four hours before the experiment, one glass bead from the freezer was transferred into the liquid media containing 7.5 mL of phosphate buffer salt solution (PBS) and 2.5 mL of liquid tryptone soy broth (TSB, Gibco, prepared according to the manufacturer’s instructions), and then incubated for 18 h at 30 °C (Erlenmeyer flask, shaking at 120 RPM). The resulting bacterial suspension was directly used for toxicity testing. The composition of the testing media was assessed during the optimization experiments with P. luminescens with the aim to achieve good bacterial viability and stability of the luminescence signal, and at the same time, minimizing the content of TSB that might interfere with the outcome of the test. The ratio 3:1 PBS/TSB was found optimal and used in the experiments.

Microplate bioluminescence protocol

A setup according to ISO (ISO 11348-3, 2009) was employed using both V. fischeri and P. luminescens. Bacterial suspensions were transferred into a white 96-well microplate and initial luminescence was recorded using a Biotek Synergy microplate reader (500-ms integration). Immediately after the initial luminescence recording, the tested samples (dilutions prepared in a separate microwell plate) were added using the multichannel pipette. The plate was gently agitated and the bacteria exposed for 30 min at 15 °C and room temperature for V. fischeri and P. luminescens, respectively. Then, the final luminescence was recorded. Each plate contained negative controls (2 % NaCl solution adjusted to pH 6.5–7.5 for V. fischeri and 20 % PBS for P. luminescens) as well as a positive control (potassium dichromate at concentration 1 mg/mL). The results were evaluated according to ISO 11348-3 as follows:

$$ Ht\kern-0.10em =\kern-0.20em \left(\frac{\left(ICt- It\right)}{Ict}\right)\kern-0.20em \times \kern-0.20em 100\kern2em {IC}_{\mathrm{t}}\kern-0.20em =\kern-0.20em I0\kern-0.20em \times \kern-0.20em {f}_{kt}\kern2em {f}_{kt}\kern-0.20em =\kern-0.20em {\mathrm{I}}_{kt}/{\mathrm{I}}_0 $$

where Ht is the inhibition of luminescence, It is the luminescence intensity after the exposure time t, IC t is the corrected value of bacterial luminescence of the tested sample, I0 is the luminescence intensity before addition of the samples, ƒ kt is the correction factor to eliminate natural attenuation of luminescence during the exposure, I kt is the luminescence of the negative control sample after the exposure and I 0 is the luminescence of the negative control in the beginning of exposure.

Microplate flash kinetic protocol

A miniaturized version of the previously published protocol (Lappalainen et al. 1999, 2001) included preparation of the studied samples (concentration dilutions; positive and negative controls as described above) directly in the white 96-microwell plate and brought to the appropriate temperature (15 °C or room temperature for V. fischeri and P. luminescens, respectively). The final volume of the sample in each microplate well was 80 μL. Bacterial suspension (20 μL) was then injected into the well using an automated injector, and the initial bioluminescence was immediately recorded (Biotek Synergy microplate reader with Gen 5 software employing the “Flash mode”). The microplate was then shaken inside the microplate reader at corresponding temperature for 30 s when the final luminescence was recorded. The inhibition of luminescence was calculated as previously described (Bláha et al. 2010):

$$ \begin{array}{ll}inh\%\kern-0.20em =\kern-0.20em 100-\frac{ITt}{CF\times IT0}\kern-0.20em \times \kern-0.20em 100\hfill & CF\kern-0.20em =\kern-0.20em \frac{ICt}{IC0}\hfill \end{array} $$

where CF is the correction factor controlling for the natural attenuation of luminescence, ICt is the luminescence intensity of the control sample after the contact time, IC0 is the initial luminescence intensity of the control sample, ITt is the luminescence intensity of the test sample after 30 s of the contact time and IT0 is the initial luminescence intensity of the tested sample.

Portable luminometer protocol

The 30-s Flash protocol has been adapted to commercially available cuvette luminometer (Biofix Lumi 10, Macherey-Nagel, Germany). A small hole was drilled into the original lid of the luminometer that allowed insertion of a needle with an inner diameter of 0.9 mm. This served as a holder for the second syringe needle used for direct manual injections of bacterial suspensions into the samples positioned in the measuring cell of the luminometer. The supplementary Fig. S1 shows a schematic presentation of the adapted portable luminometer. The test procedure used similar volume ratios as in the microplate arrangement. First, 0.5 mL of the studied sample (i.e. dilutions of the chemicals, wastewater samples, negative or positive controls) was prepared in the glass test tube, inserted into the luminometer, and the lid closed. Then, 0.2 mL of the bacterial suspension (either V. fischeri or P. luminescens) was injected directly into the sample using the syringe needle inserted through the lid hole. Immediately with the injection, the initial luminescence was recorded for 2-s light integration. The test tube was then removed from the luminometer and manually shaken at room temperature for 30 s. The tube was then placed back into the luminometer and the final luminescence was recorded. The luminescence inhibition was calculated as described in the previous paragraph (Microplate flash kinetic protocol).

Statistical analyses

Model chemicals (dilution series) were tested in three replicates. At least three independent experiments were performed for each compound and each of the three compared protocols (Microplate bioluminescence protocol, Microplate flash kinetic protocol, Portable luminometer protocol). Inhibitions and dose-response data (ECx calculations) were evaluated using Microsoft Excel and GraphPad Prism (GraphPad software, San Diego, USA). Statistical analyses were done in Statistica (StatSoft, Tulsa, OK, USA). The relationships between toxic responses and chemical contamination in the effluent samples were assessed by non-parametric Spearman correlation. Concentrations of chemicals below the limit of detection (LOD) were replaced by 1/2 of LOD value (Jarošová et al. 2014). Principal component analysis (PCA) was run with log-transformed contaminant data, which improved symmetry and normality. Toxicity data for both V. fischeri and P. luminescence were originally symmetric and were not log-transformed for PCA. Toxicity results (EC50 values) were inverted (1/EC50) for PCA to better represent actual effects (higher toxicity indicated by higher 1/EC50 value). For PCA, all data were standardized to mean = 0 and SD = 1 (i.e. mean subtracted from each individual value and then divided by the standard deviation, SD).

Results

To evaluate the suitability of P. luminescens for toxicity testing and to compare its sensitivity with V. fischeri, the experiments were performed with a range of model chemicals (Table 1) using all three compared protocols.

Table 1 Comparison of V. fischeri and P. luminescens sensitivity (EC50 values ± SD) to model chemicals in various experimental setups. (Note: EC50 values are in mg/mL with the exception of HgCl2 where μg/mL concentrations are used)
Full size table

When comparing the results from laboratory microplate arrangements (upper four rows in Table 1), the Flash kinetic protocol (30 s) showed lower sensitivity at most of the toxicants in both V. fischeri and P. luminescens. The exception was 3,5-chlorophenol, which was comparably toxic in V. fischeri at both variants (30 min vs. 30-s Flash). Ethanol was the least toxic compound and its EC50 values were comparable in both species and evaluated protocols. In the microplate-based protocols P. luminescens seemed to be less sensitive (higher EC50 values). Using the portable luminometer protocol (results in the two bottom lines in Table 1), the trend in the toxicity of individual chemicals remained the same as for the laboratory microplate protocols but both V. fischeri and P. luminescens showed comparable responses and sensitivities. The quality and reproducibility of the P. luminescens assay was confirmed by repeated testing of the standard chemical potassium dichromate in the portable luminometer protocol during eight repeated experiments covering a period of 9 months. The resulted EC50 values ranged from 0.26 to 0.68 mg/mL, with the mean EC50 = 0.5 ± SD = 0.14.

After the experiments with model chemicals, the toxicity of 43 wastewater effluents were tested with both bacterial species using the portable luminometer protocol (Fig. 1). For V. fischeri, most of the samples caused weak inhibitions of luminescence (grey columns in Fig. 1). One undiluted sample (code no. 159) caused an inhibition of around 50 %, while strong 50 % stimulations were recorded for sample no. 233. The responses of V. fischeri to effluents were not significantly correlated with the responses of P. luminescens (Fig. 1), where most of the samples caused luminescence stimulations, and one sample (no. 184) caused inhibition greater than 20 %.

Fig. 1
figure 1

Effects of wastewater effluents on V. fischeri and P. luminescens as determined with a novel portable luminometer flash kinetic protocol. Codes of samples on X-axis correspond to a previously published report (Loos et al. 2012). Columns represent mean ± SD of three replicated analyses

Full size image

Statistical analyses were performed with the effluent samples to disclose relationships between the toxic responses of V. fischeri and P. luminescens and chemical contamination. Summary data on contamination, i.e. groups of analysed compounds, are presented in Table 2 (for full details, see Loos et al. 2012).

Table 2 Chemical analyses of WWTP effluents—all concentrations are in nanograms per litre. Presented are the sum concentrations of groups of chemicals (see at the bottom of the table; values below the limit of quantification LOQ were replaced by 1/2 of LOQ, ng/L), empty cells - samples not analysed
Full size table

For V. fischeri, there were no statistically significant correlations between the toxic responses and the concentrations of pollutant groups (Spearman’s R, p > 0.05). On the other hand, comparison of contamination and responses (% of bioluminescence change) of P. luminescens showed significant correlations. When considering all available data, i.e. both inhibitions and stimulations of luminescence, statistically significant (Spearman rank, p < 0.05) were correlations of the toxicity with the sum of pesticides (Spearman’s Rs = 0.33), the sum of pharmaceuticals (Rs = 0.32) and the sum of musk fragrances (Rs = 0.44).

The multivariate presentation by PCA is shown in Fig. 2. V. fischeri toxicity seemed to be associated with inorganic contaminants (Inorg), nitrophenols and benzotriazoles for both “all data” (Fig. 2a) and inhibitions only (Fig. 2b). Alignments of a vector of toxic response of i (Fig. 2c, d) with vectors of chemical contaminants were more variable but—similar to V. fischeri—associations of toxicity with inorganics, nitrophenols and benzotriazoles was observable as the major general trend.

Fig. 2
figure 2

Principal component analyses (PCA) relating the toxic effects of wastewater effluents to concentrations of 13 groups of chemical contaminants. a, b V. fischeri toxicity - inhibitions and all samples, respectively; c, d P. luminescens toxicity - inhibitions and all samples, respectively. Sweet sweeteners; PFs perfluoralkyl substances; Pharm pharmaceuticals (without vet. antibiotics); Pest pesticides; PCPs personal care products (sum of triclosan, DEET, caffeine); Contrast contrast media; Gad gadolinium; NPs nitrophenols; BTZ benzotriazoles; Musk musk fragrances; OPFR organophosphate flame retardants (for details on chemical groups see Table 2)

Full size image

Discussion

The present work aimed to develop a simple portable battery-supplied luminometer tool for field in situ toxicity assessment with freshwater bacterium Photorhabdus luminescens. The initial investigation of sensitivity of P. luminescens with model compounds and employing standardized assays in microplates indicated rather lower sensitivity of this strain in comparison with the standard marine bacteria V. fischeri. Interestingly, in the portable luminometer, protocol sensitivities of both bacterial strains were similar indicating thus good applicability of P. luminescens. Significant correlations were observed between the responses of V. fischeri and P. luminescens (e.g. Spearman’s Rs = 0.97, p < 0.001 for portable luminometer protocol). The interspecies differences in sensitivities had also been observed before for studied bacterial species (Jennings et al. 2001; Schmitz et al., 1999). Nevertheless, statistical analyses of larger datasets often showed systematic correlations between different model organisms (Kaiser 1998).

Our further comparisons focused on protocols with different exposure durations, namely comparing standard 30-min microplate luminometer protocol with V. fischeri (ISO standard) with fast Flash kinetic 30-s exposure, which allows for the assessment of turbid or coloured samples (Lappalainen et al. 1999, 2001). As it could be expected, the shorter exposure in the Flash kinetic protocol resulted in lower sensitivity (higher EC50 values) and this was apparent for both V. fischeri and P. luminescens. Interestingly, both dichlorophenols tested (3,5- and 2,4-dichlorophenol) appeared to be comparably toxic to both V. fischeri and P. luminescens irrespectively of the protocol or exposure duration. This may be related to their known rapid mechanism of toxicity affecting membranes via polar narcosis (Zhao et al. 1998). On the other hand, large differences between the 30-min and 30-s exposures were observed at metals, particularly at potassium dichromate. This could be linked to known interactions of metals with proteins and other macromolecules, which may manifest only after longer exposures (van Assche and Clijsters 1990).

The comparisons showed some non-systematic trends where explanation is not straightforward. For example, 3,4-dichloroaniline was 10 times less toxic to V. fischeri in the portable luminometer (EC50 = 0.2 mg/mL) compared to microplate mode (EC50 = 0.017 mg/mL) while for all the other model compounds, the sensitivity of V. fischeri in both 30-s arrangements was similar. This difference was confirmed by repeated independent experiments, and we can only hypothesise about possible causes such as different volumes-to-surface ratios in different arrangements affecting the sorption/bioavailability or differing oxygen content that might aid to fast transformation of the studied compound.

As mentioned, one of the limitations of the assays with marine V. fischeri is the need to add high concentrations of sodium chloride into the test media. Previous studies showed that elevated osmomolarity interfered with luminescence emission in other bacteria such as Photobacterium leiognathi (Dunlap 1985) or lux-transfected Pseudomonas fluorescens (Cook et al. 2000), and it also has substantial effects on metal toxicity (Deheyn et al. 2004). To overcome this limitation, some authors evaluated freshwater bioluminescent bacteria such as Vibrio qinghaiensis (Ma et al. 1999) or lux genes-transfected non-luminous species such as P. putida or E. coli (Stewart & Williams 1992; Kurvet et al. 2011). Another viable approach, as shown also in the present study, is the use of naturally bioluminescent freshwater species P. luminescens.

In the standardized microplate protocols, lower sensitivity of P. luminescens was observed in comparison with V. fischeri. This could be caused by possible species-specific differences (Jennings et al. 2001; Schmitz et al. 1999) but also by other factors such as composition of the exposure media. With this respect, both compared organisms had specific limitations. V. fischeri requires additions of high salt concentrations that might interfere with toxicity of metals (Deheyn et al. 2004; Rüdel et al. 2015). On the other hand, stable luminescence signal at P. luminescens was achieved only in the presence of 25 % organic media TSB, which could decrease bioavailability and lower thus toxicity of both inorganic (Thavamani et al. 2015) and organic toxicants (Beesley et al. 2010). Interestingly, these differences the species diminished in the Flash cuvette protocol in portable luminometer, where the influence of media could be minimized during very short 30-s exposures.

Further assessment of the assay included testing of complex environmental samples, i.e. waste water effluents. Various whole effluent toxicity (WET) tests have been explored in the past including, e.g. Daphnia magna, fathead minnow, zebrafish embryos (Chapman 2000) as well as bacterial bioluminescence tests (Mendonça et al. 2009). The importance of WET testing has been highlighted especially for arid areas, where insufficient dilution of the effluent may directly affect biotic communities within the recipient stream (Diamond and Daley 2000). Recently, toxicity limits for WET have been discussed (Huybrechts et al. 2014).

With respect to bacterial assays, Kováts et al. (2012) employed V. fischeri Flash kinetic testing of effluents, and reported EC50 values around the 50 % dilution of the original effluent. In contrast, the present study found weaker luminescence inhibitions in both V. fischeri and P. luminescens with maxima around 20 % effect at undiluted effluents. These observations are in general agreement with a previously published European interlaboratory study (Farré et al. 2006), where generally minor responses were detected in a number of V. fischeri-based protocols (Microtox, Tox alert, Biofix lumi and Tox tracer tests). Despite of weaker effects of effluents in the present study, multivariate PCA analysis revealed clear associations between toxicity and elevated levels of inorganic compounds, chlorophenols and benzotriazole anticorrosives. Effect-based monitoring thus provides additional value to routine chemical analyses and may serve for sample prioritization or identification of major drivers of toxic effects (Wernersson et al. 2015; Escher et al. 2013). However, limitations also exist in the interpretation of bacterial bioluminescence assays in WET testing. For example, the inhibition of luminescence is the expected and evaluated endpoint but practical testing often shows both inhibitions and stimulations (Dizer et al. 2002; Pessala et al. 2004). This was also confirmed in the present study. Although some of the factors affecting luminescence in complex matrices have been suggested such as pH, potassium and calcium ions or salinity (Berglind et al. 2010; Cook et al. 2000), further research and debate on the interpretation of stimulatory responses in bacterial bioluminescence is needed.

Conclusions

The present study demonstrated the development of a rapid bioluminescence-based testing tool employing a small battery-supplied luminometer. The sensitivity of freshwater P. luminescens was comparable with standard marine bacterium V. fischeri using the portable luminometer protocol. The applicability and reproducibility of the assay was further confirmed in the study of 43 European effluents, where elevated levels of inorganic compounds (metals), chlorophenols and benzotriazoles were the main drivers of toxicity in both compared bacterial species. The use of P. luminescens in combination with a portable luminometer brings several advantages such as applicability at room temperatures, fast 30-s response and a freshwater setup.