ARTICLE pubs.acs.org/est
Reproducible 1H NMR-Based Metabolomic Responses in Fish Exposed to Different Sewage Effluents in Two Separate Studies Linda M. Samuelsson,† Berndt Bj€orlenius,‡ Lars F€orlin,§ and D. G. Joakim Larsson*,† †
Institute of Neuroscience and Physiology, Department of Physiology, The Sahlgrenska Academy, University of Gothenburg, P.O. Box 434, SE-405 30 G€oteborg, Sweden ‡ Stockholm Water Company, V€armd€ov€agen 23, SE-131 55 Stockholm, Sweden § Department of Zoology/Zoophysiology, University of Gothenburg, P.O. Box 436, SE-405 30 G€oteborg, Sweden
bS Supporting Information ABSTRACT: Treated sewage effluents contain complex mixtures of micropollutants, raising concerns about effects on aquatic organisms. The addition of advanced treatment steps has therefore been suggested. However, some of these could potentially produce effluents affecting exposed organisms by unknown modes of action. Here, 1H NMR (proton nuclear magnetic resonance spectroscopy) metabolomics of fish blood plasma was used to explore potential responses not identified by more targeted (chemical or biological) assays. Rainbow trout was exposed in parallel to six differently treated effluents (e.g., conventional activated sludge, addition of sand filter, further addition of ozonation and/or a moving bed biofilm reactor or a separate membrane bioreactor line). Multivariate data analysis showed changes in the metabolome (HDL, LDL, VLDL and glycerol-containing lipids, cholesterol, glucose, phosphatidylcholine, glutamine, and alanine) between treatment groups. This formed the basis for postulating a hypothesis on how exposure to effluent treated by certain processes, including ozonation, would affect the metabolic profiles of exposed fish. The hypothesis withstood testing in an independent study the following year. To conclude, 1H NMR metabolomics proved suitable for identifying physiological responses not identified by more targeted assays used in parallel studies. Whether these changes are linked to adverse effects remains to be tested.
’ INTRODUCTION Many contaminants are not efficiently removed by traditional sewage treatment processes, raising concerns about their effects on aquatic organisms.1,2 The addition of more advanced treatment methods to existing plants has therefore been proposed.3-5 Ozonation and activated carbon treatment are two particularly promising technologies.4,6,7 There is, however, also a possibility that some (particularly oxidative) technologies could also create biologically active compounds. Consequently, there are studies suggesting either increased or decreased toxicity of ozonated effluents. 1,6,8-13 Other oxidative methods, mainly involving ultraviolet (UV) radiation, have also been considered.6,14-17 Evaluations of new technologies through biological effect studies are indeed useful to avoid creating one problem while solving another. 1H NMR-based metabolomics is a broad and exploratory method suitable for finding physiological responses in a situation where little is known about causative agents or potential effects, as up to 100 endogenous metabolites are analyzed simultaneously. The aims of the present study were therefore to investigate (1) whether exploratory NMR-based metabolomics can provide information about affected physiological processes in fish exposed to sewage effluents, treated either conventionally or by more advanced methods; (2) whether this information adds to existing knowledge obtained through other chemical or biological methods, including more targeted approaches; and (3) if changes in metabolite levels were reproducible by comparing responses from two independent studies. r 2011 American Chemical Society
’ MATERIALS AND METHODS Two exposure studies were carried out: April 17-May 3, 2007 (A) and March 3-17, 2008 (B) under animal ethics permits to D.G.J.L. For study A, general water quality parameters, concentrations of estrogenic chemicals, estrogenic gene responses of exposed fish, their liver and heart somatic indexes, and expression of hepatic carbonyl reductase/20β hydroxysteroid dehydrogenase mRNA have been reported elsewhere.18,19 Also, two other studies have been published using the same experimental setup as in study A, preceding it by a few weeks. These include concentrations of 19 pharmaceuticals and some other chemicals20,21 as well as effluent toxicities for the bacterium Vibrio fischeri (bioluminescence), the micro algae Pseudokirchneriella subcapitata (growth inhibition), the macro algae Ceramium tenuicorne (growth inhibition), the harpactidoid copepod Nitocra spinipes (larval development), and the fish Danio rerio (embryotoxicity).21 The effects of the effluents in study B on the population dynamics of exposed N. spinipes have also been described.22 Only effects of undiluted effluents were investigated here. We considered a complete risk characterization and evaluation of implementation and running costs to be outside the scope of this paper. Sewage Treatment Technologies. Both studies were carried out at Henriksdal sewage treatment plant (STP; Stockholm, Received: December 8, 2010 Accepted: January 4, 2011 Published: January 24, 2011 1703
dx.doi.org/10.1021/es104111x | Environ. Sci. Technol. 2011, 45, 1703–1710
Environmental Science & Technology
ARTICLE
Scheme 1. Schematic Overview of the Sewage Treatment Setup Used in (a) Study A (2007) and (b) Study B (2008)
Sweden) using a semilarge scale pilot plant with parallel treatment lines (Scheme 1). A full description of the six parallel treatment lines in study A is available elsewhere (19; Table S1). Briefly, the raw influent (sewage from the Henriksdal inlet) was divided into two separate lines: one using membrane bioreactor (MBR) technology and one using conventional treatment (C) including preliminary treatment (removal of large objects, grit, and dense inorganic solids), primary sedimentation followed by an activated sludge process (sludge age of 6 d), and finally a secondary sedimentation procedure (Scheme 1a). This type of conventional treatment is common in Sweden. In the membrane bioreactor (MBR) line, the primary sedimentation was replaced by a drum filter, whereas the activated sludge and secondary sedimentation steps were substituted by a membrane bioreactor.
A sand filter, a commonly used add-on treatment, was added to the conventional line (C) to generate our reference effluent (Cs). Three different tertiary treatments were added onto the reference effluent (for details see Table S1): a moving bed biofilm reactor (CsþMBBR), ozonation at 15 mg/L, (CsþO3,high) and ozonation (15 mg/L) plus a moving bed biofilm reactor (CsþO3,highþMBBR). In addition, tap water containing 2% Cs effluent was used in two aquaria (TA1 and TA2) in this study, as our experience is that 100% tap water is not as well accepted by fish in general. In study B the incoming sewage came from the inlets Henriksdal and Sickla (Scheme 1b). The conventionally treated reference effluent (C2) was similar to Cs in study A, however the C2 process was operated in full scale with a sludge age of 20 d. 1704
dx.doi.org/10.1021/es104111x |Environ. Sci. Technol. 2011, 45, 1703–1710
Environmental Science & Technology The added tertiary treatments processes were ozonation at 15 or 5 mg/L (C2þO3,high and C2þO3,low, respectively); ozone (5 mg/L) plus moving bed biofilm reactor (C2þO3,lowþMBBR); activated carbon (C2þAC); and finally UV/hydrogen peroxide (C2þUV/H2O2) (Table S1). Carbon-filtered tap water plus 2% C2 reference effluent was used in two aquaria (TB1 and TB2). Total organic carbon (TOC), suspended solids (SS), 7-day biochemical oxygen demand (BOD7), total nitrogen (Tot-N), and total phosphorus (Tot-P) were measured in all effluents in both studies to evaluate differences in removal efficiency between treatments (Table S2). Fish Exposure and Sampling. Juvenile rainbow trout (Oncorhynchus mykiss), of both sexes (approximate weights 80 g (A) and 110 g (B)) were supplied by Antens Laxodling Sweden). Upon arrival at Henriksdal STP, the fish AB (Alingsas, were transferred to tanks supplied with tap water þ 2% reference effluent (carbon filtered for study B). After an acclimatization period of 1 (A) or 4 d (B), the fish were distributed randomly among eight flow-through 100-L aquaria supplied with the different effluents or tap water þ 2% reference effluent according to Scheme 1 until there were 22 (A) or 25 (B) fish in each aquarium. The aquaria were aerated throughout the 14-d exposure period and the fish were not fed. In study A, a total of 33 fish died (15 from TA1, 14 from TA2, two from Cs, one from MBR, and one from the CsþO3 treatment groups). The high mortality in TA1 and TA2 was most probably caused by an unannounced raise in the chlorine concentration by the water supplier (see also ref 19). After installing a carbon filter on day four no more fish died. All data from TA1 and TA2 in experiment A were nevertheless excluded from all NMR analyses as the high initial mortality could have affected some of the analyzed metabolites. In Study B, only one fish died (C2þUV/H2O2). Temperature, pH, conductivity, oxygen saturation, and flow rate were measured 5 d per week in all aquaria (Supporting Information). In both studies, sampling took place over two consecutive days. Fish were killed by a blow to the head in random order. Blood was collected in heparinized syringes from the caudal vessels, and plasma was separated by centrifugation at 10 000 rpm for 2 min, frozen on dry ice, and then stored at -80 °C. Fish were weighed, fork length was measured, and sex was determined. NMR Sample Preparation and Analyses and NMR Data Preprocessing. All steps from sample preparation to data analysis were performed separately for studies A and B. In general, sample preparation and acquisition of presaturation (“presat”) and Carr-Purcell-Meiboom-Gill (“CPMG”) spectra of fish blood plasma were performed as described elsewhere (ref 23 and Supporting Information). Multivariate Data Analysis of NMR Data. Data were meancentered and Pareto-scaled prior to analysis using SIMCA-P Sweden). Principal Component (version 11.0, Umetrics AB, Umea, 24 Analysis (PCA ), an unsupervised data reduction method, was performed as a first step giving an overview of the data sets, showing trends, groupings, and outliers. Next, supervised data analysis (Partial Least Squares-Discriminant Analysis (PLS-DA25) was used to find potential differences pairwise between treatment groups. In cases where more than one PLS component was obtained in the model, orthogonal PLS (O-PLS) was used instead to remove variation in the data matrix not related to the treatment method.26 The probability to obtain models with a similar or better explained variance (R2) or predictive capability (Q2) by chance was tested by Monte Carlo randomizations (n = 9990 per comparison, see
ARTICLE
Supporting Information). Model validity was assessed by calculating p-values for R2 and Q2 and rejecting models with p > 0.05. All buckets differing significantly between two groups (95% confidence level) in a valid model were listed and the corresponding metabolites were identified, but only if all peaks arising from one particular metabolite were significant. A hypothesis regarding the effects of advanced treatment technologies on fish plasma metabolite profiles was built based on the found metabolites from study A. The ten most regulated metabolites in B were then compared to the corresponding list from study A to investigate its validity.
’ RESULTS There were no significant effects on body length or weight in either study. Study A. Representative 1H NMR spectra of fish blood plasma are presented in Figure S1. The presaturation spectrum (Figure S1a) is dominated by broad resonances originating mainly from lipoprotein lipids. In the CPMG spectrum, lowintensity, sharp peaks originating from smaller metabolites become more visible (Figure S1b). Outliers can greatly influence PLS analyses. Three fish (out of 128) were identified as outliers (one in Cs, two in CsþMBBR) based on deviating PCA scores or standard deviation of the distance to model and were removed from further analysis. The Hotelling’s T2 statistics, included in the SIMCA-P software, also confirmed these outliers (99% confidence level). Results of (O)PLS-DA are presented in Table 1 along with metabolites responsible for group separation. The metabolite profiles of fish exposed to the two conventional effluents (C and Cs) were similar (Table 1, entry 1). However, Cs differed from all three advanced treatments (entries 2-4), consistently through similar metabolite changes, i.e., HDL (high density lipoprotein) and LDL (low density lipoprotein) lipids, cholesterol and lipids were increased due to advanced treatment, and VLDL (very low density lipoprotein) lipids and glycerol-containing lipids were decreased. This was confirmed through pairwise comparisons of the advanced treatments (entries 5 and 6). Thus, these three groups were pooled and compared with a pool of C and Cs. Again, HDL and LDL lipids and cholesterol increased along with R- and β-glucose, phosphatidylcholine, glutamine, methionine, alanine, and taurine, whereas VLDL lipids, VLDL cholesterol, glycerol-containing lipids (“glycerolipids”), and creatinine decreased in fish exposed to advanced treated effluent (entry 7). A scores plot from the OPLS analysis of entry 7 is presented in Figure S2. Although the separation is incomplete and the R2 and Q2 values are relatively low, the model is still highly significant as shown by the very low p-value (p < 0.0001). Finally, a possible difference in metabolite profiles was found between fish exposed to Cs and MBR effluent (entry 8). However the Monte Carlo randomization of R2 and Q2 values indicated a less confident model (p > 0.05 in some cases). No differences between sexes were found (note - juvenile fish). From these results, we hypothesized that exposure of fish to effluent treated with ozone and/or MBBR would mainly affect lipid metabolism (HDL, LDL, and VLDL lipids, glycerolipids, and HDL/ VLDL cholesterol), glucose, phosphatidylcholine, and some amino acids. This hypothesis was tested using data from study B. Study B. Effluents used in study B differed somewhat from study A (Scheme 1a and b). Seven outliers (out of 199 fish) were detected by PCA and removed from further analysis (three from TB1 and TB2, one from C2, one from C2þO3,high, and one from C2þUV/H2O2), again in agreement with the Hotelling’s T2 outlier identification. 1705
dx.doi.org/10.1021/es104111x |Environ. Sci. Technol. 2011, 45, 1703–1710
Environmental Science & Technology
ARTICLE
Table 1. Results from PLS-DA or O-PLS Analysis of NMR Data (Presat and CPMG) from Study B Showing Main Metabolite Changes (v Increased or V Decreased) in Fish Blood Plasma after Exposure to Treated Sewage b Compared with Exposure to aa entry
exposure groups
main metabolite changes
PLS-DA or O-PLS model Presat/CPMG NMR data
a
b
vs
no. comp
R2X
R2Y (p-values)
Q2 (p-values)
1
C
vs
Cs
0/0
-/-
-/-
-/-
-
2
Cs
vs
CsþMBBR
1/1b
0.30/0.35
0.30/0.20 (0.0059/
0.18/0.08 (0.0064/
v: HDL þ LDL lipids, cholesterol, lipids. V: VLDL lipids, glycerolipids,
0.0019)
0.0013)
VLDL cholesterol, Gly, Trp, creatine/creatinine, lipids.
3
Cs
vs
CsþO3,high
1/1
0.30/0.33
0.36/0.31 (0.0016/ 0.014)
0.23/0.20
v: HDL þ LDL lipids, cholesterol, R-glucose,
(0.0022/
HDL cholesterol, phosphatidylcholine,
0.0035)
Ile, taurine, lipids, unknown (1.12 ppm). V: VLDL lipids, glycerolipids, lipids.
4
Cs
vs
CsþO3,highþ MBBR
1/1
0.26/0.27
0.35/0.30 (0.0019/ 0.0072)
0.28/0.12 (0.00050/
v: HDL þ LDL lipids, cholesterol, HDL cholesterol, Ile, lipids.
0.023) V: VLDL lipids, glycerolipids, lipids.
5
CsþO3,high
6
CsþO3,high
vs
CsþMBBR
0/0
-/-
-/-
-/-
-
vs
CsþO3,high
0/0
-/-
-/-
-/-
-
vs
CsþO3,highþ MBBR, CsþO3,high
1b/1b
0.29/0.37
0.17/0.17 (
