The impact of freshwater metal concentrations on the severity of histopathological changes in fish gills: A statistical perspective

Science of the Total Environment 599–600 (2017) 217–226

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

The impact of freshwater metal concentrations on the severity of histopathological changes in fish gills: A statistical perspective A.R. Fonseca a, L.F. Sanches Fernandes a,c, A. Fontainhas-Fernandes a,d, S.M. Monteiro a,d, F.A.L. Pacheco b,e,⁎ a

Centre for the Research and Technology of Agro-Environment and Biological Sciences, Vila Real, Portugal Chemistry Research Centre, Vila Real, Portugal Department of Engineering, University of Trás-os-Montes and Alto Douro (UTAD), Ap. 1013, 5001-801 Vial Real, Portugal d Department of Biology and Environment, UTAD, Portugal e Department of Geology, UTAD, Portugal b c

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• Gill histopathology of three fish species was evaluated in six Portuguese Rivers. • Gill histopathological analysis revealed alterations in fish. • Gill alterations act as a biomarker to toxicity of sub lethal concentrations of metals. • Filament epithelium proliferation shows the highest correlation with metal concentrations. • Boga is the most responsive species.

a r t i c l e

i n f o

Article history: Received 22 March 2017 Received in revised form 24 April 2017 Accepted 26 April 2017 Available online xxxx Editor: D. Barcelo Keywords: Heavy metal concentrations Fish gill histopathology Goodman Kruskal correlation PLS regression Freshwater quality Ecological status

a b s t r a c t The purpose of this study was to relate the severity of histopathological changes in fish gills with changes in metal concentrations of freshwater samples, and to use the relationships as premature warnings of impairment in aquatic fauna populations. The investigated species were the native barbel (Luciobarbus bocagei) and boga (Pseudochondrostoma sp.), and the introduced trout (Oncorhynchus mykiss), collected from 6 northern Portuguese rivers in a total of 249 individuals. The sampling sites have been linked to different ecological status by the official authorities. The sampling has been repeated 4 times to cover different hydrologic and environmental conditions. The analyzed metals were aluminum, arsenic, cadmium, cobalt, chromium, copper, manganese, nickel, lead and zinc. For each fish, 30 filaments of a gill arch were observed in a light microscope, and the histopathological changes evaluated according to a 6-degree gradation scale that combines the extent and severity of each lesion. The relationships between the histopathological and the chemical results were investigated by the nonparametric Goodman Kruskal gamma correlation and Partial Least Squares regression (PLS). The statistical results highlighted the importance of filament epithelium proliferation (FEP) as key biomarker to the toxicity of sub lethal concentrations of metals, because FEP was significantly correlated with all analyzed metals and explained through PLS regression by concentration changes of Cu, Zn, Mn, Cr and As. A refined regression analysis, where histopathological data on the 3 species were processed in separate, revealed that FEP severity is especially

⁎ Corresponding author at: Chemistry Research Centre, Vila Real, Portugal. E-mail address: [email protected] (F.A.L. Pacheco).

http://dx.doi.org/10.1016/j.scitotenv.2017.04.196 0048-9697/© 2017 Elsevier B.V. All rights reserved.

218

A.R. Fonseca et al. / Science of the Total Environment 599–600 (2017) 217–226

sensitive to changes in metal concentrations in boga. Thus, monitoring studies on the ecological status of northern Portuguese rivers would benefit in time and cost if FEP is used as biomarker and boga as species. Naturally, the option for this species depends on the availability of boga individuals along the stream reaches selected for the monitoring programs. © 2017 Elsevier B.V. All rights reserved.

1. Introduction Heavy metals, such as copper or zinc, are nutrients required for various biochemical and physiological functions (Rengel, 1999). However, their inadequate supply may result in a variety of deficiency diseases, while high concentrations may result in cell and tissue damage (Tchounwou et al., 2012). Pollution by heavy metals represents a major risk for public and wildlife health. The risk exists because some heavy metals may biomagnify through the aquatic food web or even bioaccumulate, with exceptions (Monroy et al., 2014). In general, the continuous exposure of lower organisms to reduced concentrations of heavy metals can expose predatory organisms, including humans, to potentially harmful concentrations (Brito et al., 2015; Stephansen et al., 2014). In most recent years, the contamination of aquatic systems by heavy metals triggered public health and ecological apprehension. The growth of human exposure to heavy metals has been recognized and related to an expanding use of these substances in industrial, agricultural, domestic and technological products (Bradl, 2005; Duffus, 2002; Fergusson, 1990). Most environmental contamination and animal exposure to heavy metals were found to result from anthropogenic activities such as mining or production and use of fertilizers in cropland (Goyer and Clarkson, 1996; He et al., 2005; Herawati et al., 2000). Heavy metal concentrations were also the focus of European Union (EU) regulatory legislation. In that context, the Water Framework Directive - WFD (Directive, 2000/60/EC) stipulates predefined concentrations for some metals in receiving water masses. Coupled with improved control and abatement techniques, the action of EU resulted in the general reduction of heavy metal emissions (Pacyna et al., 2007). Northern Portuguese rivers are subject to important anthropogenic pressures, derived from direct discharge of domestic and industrial effluents into stream water, leachates from livestock excreta and dressings of cropland fertilizers, environmental land use conflicts, human induced wildfires, among others (Oliveira et al., 2005; Pacheco and Sanches Fernandes, 2016; Pacheco et al., 2015b; Santos et al., 2015a, 2015b; Valle Junior et al., 2015; Vieira et al., 2012, 2013). It is therefore essential to monitor heavy metal emissions in these rivers, to ensure their reduction as well as the harmful ecological effects derived therefrom. The periodic analysis of heavy metal concentrations in a river allows determining the water chemical status through chemical tests, but is not adequate to determine the river's ecological status. Early methods to assess the ecological condition of a river were based on indices of biotic integrity (IBIs), which frequently link a good ecological condition to a high abundance and diversity of aquatic fauna, namely macroinvertebrates (Hauer and Resh, 1996; Santos et al., 2015c; Valle Junior et al., 2015) or fish (Fame, 2005; Hermoso and Clavero, 2013; Simpson et al., 2000). Although IBIs are easy to use, they can only detect impairment when there is a switch in the structure of aquatic fauna communities. To prevent the occurrence of local extirpations, it essential to develop and implement early warnings as regards impairment of fauna populations. The need for such rapid and sensitive tools to reveal sub-lethal effects in aquatic organisms has led to the use of biomarkers (e.g. Fonseca et al., 2016). A biomarker is a change in a biological response, either at molecular, cellular, histological, physiological or behavioral level, that can be related with the exposure to toxic environmental elements (Colin et al., 2016). Fish communities are key elements to evaluate the ecological condition of rivers based on biomarkers (Fonseca et al., 2016; Hermoso and Clavero, 2013; Scardi et al., 2008), while histopathology is the standard

method for assessing both short- and long-term xenobiotic effects on fish (Hinton and Lauren, 1990). Histopathological changes in fish organs are fundamental indicators of a prior exposure to environmental stressors, probably being the result of adverse biochemical and physiological changes. Fish gills are among primary target organs of environmental pollutants because gills are in direct contact with water and are characterized by large surface area and absorption rates of chemicals (Pandey et al., 2008). Pathological alterations of gill epithelium are frequently the consequence of exposure to contaminants, especially to heavy metals (Fonseca et al., 2016), with the severity being dependent on the pollutant concentration and exposure time span (Tchounwou et al., 2012). The association of high metal concentrations to severe injuries in fish gills may be investigated from various stand points, namely the statistical perspective (Cappello et al., 2016; Nowak and Lucas, 1997). Statistical techniques are generally useful because they can handle large amounts of data. Some techniques are particularly interesting because they can distinguish a subset of predictors relevant for a specific phenomenon, among a larger group of variables. In the sequel, these algorithms can alert for predictors not accounted for in the analysis. The multiple histopathological changes usually observed in fish gills (the “phenomenon”), as well as the large number of metals responsible for their development (the “predictors”), render attraction to these approaches, especially if the dataset is substantially large. The non-parametric Goodman and Kruskal's gamma rank order correlation (Goodman and Kruskal, 1954) or the Partial Least Squares (PLS) regression (Wold, 1966) are two among the methods that may eventually set up robust associations between metal concentrations and histopathological changes, because these techniques impose few assumptions on the dataset or the underlying theory. The Goodman and Kruskal coefficient is commonly used in all sorts of studies for long. Following the precursory work by Engle et al. (1986), numerous papers based on the PLS regression model have also been published (Aneiros-Pérez et al., 2004; Chen, 1988; Ferreira et al., 2017; Pacheco and Sanches Fernandes, 2016; Pacheco et al., 2015b; Schick, 1996; Speckman, 1988). The general purpose of this study was therefore to assess the relationship between the severity of fish gill histopathological changes and exposure to heavy metals in freshwaters, using the Goodman and Kruskal correlation and the PLS regression methods. To achieve the proposed goal, a number of specific objectives had to be attained through execution of predefined tasks: 1) collection of 249 fish specimens from species representing native and introduced fauna of northern Portuguese rivers, namely the native barbel (Luciobarbus bocagei) and boga (Pseudochondrostoma sp.) and the introduced trout (Oncorhynchus mykiss). The purpose of collecting native and introduced species was to compare the sensitivity of fish species to xenobiotics according to origin. The sampling comprised four campaigns and nine locations, which accounted for seasonality and differences in ecological status among sites; 2) analysis of fish gills for identification of lesions and their severity; 3) collection of 36 freshwater samples in the same sites as those used for fish capture, for analysis of heavy metal concentrations (ten substances in total); 4) preparation of a dataset with histopathological and chemical parameters, properly combining the fish species, the campaigns and the sampling sites, to be used in STATISTICA, version 7 (Statsoft Inc., 2004), with the correlation and regression methods; 5) Identification and interpretation of statistically significant associations between specific injuries and metals, in general or considering the species involved.

A.R. Fonseca et al. / Science of the Total Environment 599–600 (2017) 217–226

2. Materials and methods 2.1. Study area The study area encompasses six hydrographic basins distributed within a sector of mainland Portugal located between the latitudes 40°30′N–42°N and longitudes 6°W–9°W (Fig. 1). This sector is limited to the East and North by the Spanish border, to the West by the Atlantic coast and to the South by the Douro River. It can be divided into the Northwest Mountains and the Northeast plateau, roughly separated by the Tâmega River. The first sector, located to the West of Tâmega, comprises the NE–SW trending mountains of Peneda, Soajo, Gerês, Larouco and Barroso. In this region, topography is craggy and climate is very humid with annual precipitations ranging from 1200 to 3000 mm·yr−1 (30 year average). In the Northeast plateau, located to the East of Tâmega, relief is undulated and topographic elevations are oriented in the NNE–SSW direction following large-scale geologic structures. Climate is moderately humid away from the main valleys, with annual precipitations varying from 800 and 2000 mm·yr−1. Land use and occupation in the studied basins are characterized by a mosaic of forests, agriculture and natural areas, representing 86% of the basin area, while human developed regions represent 3.5% (Caetano et al., 2009). The spatial distribution of human developed areas is heterogeneous, increasing from the inland (Ôlo and Pinhão) to the coastal (Cávado, Este, Ave and Vizela) watersheds. The gradient of human development is mostly expressed by escalating population densities and anthropogenic pressures (domestic sewage, effluents from agriculture and industry) towards the coast. The higher levels of urban and industrial pressure in the coastal watersheds caused water quality deterioration. In a report on the management of local hydrographic regions (PGRH, 2011), the quality status in inland watershed freshwater is generally graded as good or higher, while in the coastal watersheds is generally classified as lower than good sometimes dropping to the lowest level of bad. Similar distribution patterns were recognized in other regions around the globe (Rivera-Guzmán et al., 2014; Sindern et al., 2016). Among coastal watersheds, the most problematic is eventually the Ave River basin and its tributary Vizela River. In the mid-20th Century, the Ave River was tagged “The Great Sewer”. This label was posted because over seven hundred companies were settled down on its banks and discharged huge amounts of industrial wastewater into the river for

219

decades, without any physical, chemical or biologic treatment. Besides the industrial wastewater, the Ave River water was profoundly contaminated by domestic sewage produced by N500,000 inhabitants. The construction domestic and industrial Wastewater Treatment Plants (WWTP) in middle reaches of the basin raised the quality of Ave and Vizela Rivers, but quality problems still persist (Ferreira et al., 2017). Heavy metals are among the constituents contributing significantly to the degradation of water quality in the coastal basins. 2.2. Sampling and analysis The fish and water samples were collected from nine locations in stream reaches from the six watersheds. The sampling campaigns were repeated four times, concentrated in July 2011, January 2012, June 2012 and April 2013. The exact collecting days are listed in the Supplementary Material under the heading “Date”. The sites and associated basins are illustrated in Fig. 1b. The site coordinates and ecological status of freshwater in the sampled reaches are depicted in Table 1. A thorough inspection of this table confirms the good status of inland watershed freshwater and the poor to bad status of coastal watershed water. In each campaign, the concentrations of aluminum (Al), arsenic (As), cadmium (Cd), cobalt (Co), chromium (Cr), copper (Cu), manganese (Mn), nickel (Ni), lead (Pb) and zinc (Zn) were determined in one water sample per site where several fish specimens were also collected and analyzed for gill histopathology as described in Section 2.3. The metal concentrations in river water were determined by ion chromatography (Dionex equipment). The analytical results for the 36 water samples are presented as Supplementary Material. 2.3. Gill histopathology Depending on fish availability, several individuals of barbel, boga and trout were captured using pulsed direct current backpack electrofishing equipment with a DC-500 V generator. Fish sampling followed the Ethical Guidelines of the European Union Council (Directive, 1986) and the Portuguese Agricultural Ministry (Portaria. n°. 1005/92, 1992) for the protection of animals used for experimental and other scientific purposes. A total of 249 specimens were captured throughout the four campaigns, namely 112 barbels, 73 bogas and 64 trouts.

Fig. 1. Study area (northern Portugal): (a) topography, main rivers and spatial distribution of river basins selected for sampling; (b) sampling sites (river water for analysis of metal concentrations and fish for histological assessment of gill injuries) and related watersheds. Relevant information on the sampling sites is provided in Table 1.

220

A.R. Fonseca et al. / Science of the Total Environment 599–600 (2017) 217–226

Table 1 Essential information on the sampling sites: name, associated watershed, location coordinates and ecological status of sampled river. The ecological status was compiled from the latest reports on the management plans for the relevant hydrographic regions (https://www.apambiente.pt). Name

Prado Ponte do Pingue Ponte da Junqueira Graça Caldas de Vizela Ponte Trofa Santo Adrião Pinhão Tejão

Watershed

Cávado Vizela Este Ave Vizela Ave Vizela Pinhão Olo

Coordinates

Ecological status

N

W

41°36′2.93″N 41°27′45.87″N 41°22′22.80″N 41°21′59.57″N 41°22′24.62″N 41°20′44.30″N 41°22′21.08″N 41°20′38.26″N 41°21′52.58″N

8°28′15.99″W 8°7′42.94″W 8°42′8.08″W 8°41′49.83″W 8°18′38.19″W 8°33′50.13″W 8°16′58.07″W 7°35′25.67″W 7°55′41.08″W

Immediately after the capture, fish were anaesthetized with immersion in 3-aminobenzoic acid ethyl ester methanesulfonate (MS-222) and euthanized by decapitation. For each fish, two gill arches were sampled and preserved in 10% formaldehyde. After 24 h of fixation, gills were dehydrated in graded ethanol series, embedded in paraffin and sectioned (5 μm thick). Sections were stained with hematoxylin–eosin (H&E) and observed in a light microscope. One gill arch was randomly chosen and, on average, 30 entire filaments per arch were analyzed. In a first approach, a qualitative evaluation was made and the histopathological changes recorded in each individual. Secondly, a six degree (0–5) severity gradation scale (SGS), that combines the extent and severity of each lesion, was defined according to Monteiro et al. (2008). The extent of the histopathological changes was defined as the percentage of filaments with that specific change in each fish sampled, while the different levels of severity were attributed following Monteiro et al. (2008). The severity of each lesion per average of affected filament was determined as the number of lamellar and interlamellar spaces affected by a given level of severity, divided by the number of filaments showing that type of histopathological change. The degree 0 was given to values found in fish with less histopathological changes, and to define the remaining degrees, the extent and severity levels were combined to show an increasing number of lamellae and interlamellar spaces per injured filament. All values obtained, from the extent and severity counting, were then divided into numerical intervals and combined to generate the SGS used in the histopathological changes quantification. To each histopathological change and according to the SGS defined, a degree was attributed to each fish. The histopathology results are presented as Supplementary Material. 2.4. Statistical model and data handling In this study, the calculation of Goodman and Kruskal gamma coefficients represented a precursory assessment of pairwise links between injury types and their severity in the fish gills and concentrations of specific metals in river water. The gamma coefficient was selected because rank order correlation analysis is non-parametric, which means relatively independent from dataset properties including population normality. Following the exploratory analysis, a more elaborate multivariate regression was attempted to grasp multiple associations between gill injuries and their severity (termed responses or dependent Y variables) and metal concentrations (termed predictors or independent X variables). The selected technique was Partial Least Squares (PLS) regression (Wold, 1966), because it imposes few assumptions about the dataset or the underlying theory (Falk and Tonkin, 2001). Basically, PLS regression creates an optimal linear relationship between the predictors and the response specified in a conceptual model, through the coupling of Principal Components Analysis (PCA) and Linear Regression (LR) whereby spurious correlations among the X variables are identified and minimized and the covariance between X and Y is maximized.

Poor Poor Poor Bad Bad Poor Poor Good Good

In this study, the number predictors was 10 (the ten heavy metals), which is a large number. It is expected a degree of collinearity between some predictors, because the potential sources of metals in river water (rock weathering, domestic sewage, industrial effluents, farmland fertilizers) are much smaller than the number of metal types. Because there is no simple criterion to select a specific group of metals that could be considered more relevant for the statistical analysis, the choice was to keep all metals in the PLS regression database and let the model identify which explain better the histopathological changes and their severity in the fish gills. It is worth recalling that covariates such as fish size or sex may act as confounding factors in the statistical outputs. However, they were omitted in the analyses because the aim was specifically to investigate eventual cause-effect relationships between metal concentrations and histopathological changes in the fish gills. The details of PLR regression are beyond the scope of this paper. A brief outline of the method is presented in Appendix A. The software used to run the statistical analyses was STATISTICA, version 7 (Statsoft Inc., 2004). The handling of spatial data resorted to the ArcMap GIS platform (ESRI, 2010). This software is known to aid numerous environmental applications (Pacheco and Landim, 2005; Pacheco and Van der Weijden, 2012, 2014; Pacheco et al., 2013, 2015a, 2015b; Sanches Fernandes et al., 2012; Santos et al., 2014; Valera et al., 2016). 3. Results The metal concentrations in sampled river water were always below thresholds legally imposed for environmental quality objectives of surface water (Ambiente, 1998). The exceptions were copper at Ponte Junqueira (N 0.1 mg/L) and zinc at Ponte Junqueira, Pinhão and Prado (N 0.5 mg/L), during the first campaign (summer of 2011). The highest concentrations were reported for Zn (average for n = 36: 138.2 μg·L− 1), Al (119.5 μg·L−1), Mn (27.1 μg·L−1) and Cu (18.3 μg·L−1). For the other metals, average concentrations were on average close or smaller than 5 μg·L−1. The average concentrations of all metals are illustrated in Fig. 2a. Several histopathological changes with different severity levels were observed in gills of fish collected at the various sampling locations, namely: lamellar fusion (LF), filament epithelium proliferation (FEP), lamellar epithelium proliferation (LEP), vasodilation (Vas), aneurisms (Aneu), edema (Ed), lifting (Lf) and necrosis (Ne). The higher ranks were estimated for Vas, Lif and Ed, the lowest for LF, FEP and Aneu (Fig. 2b). The microphotographs illustrating the aforementioned histopathological changes are illustrated in Fig. 3. To investigate pairwise correlations between heavy metal concentrations and severity of gill injuries, Goodman and Kruskal gamma correlation coefficients were calculated and summarized in Table 2. The coefficients depicted outside the dark grey cells were significant at p b 0.05. The lesions LF, FEP and Ne correlated positively with all metals (light shaded cells). The highest coefficients were obtained for FEP. The values were high for Cu (0.62), Zn (0.57) and Mn (0.52), moderate for Chromium (0.47) and low for Al, As, Cd, Pb and Ni (b0.35). Strikingly, Co showed a negative small correlation with the injury severity. The

A.R. Fonseca et al. / Science of the Total Environment 599–600 (2017) 217–226

221

Fig. 2. Summary of analytical results: (a) average metal concentrations and associated 95% confidence intervals in freshwater; (b) average severity ranks and associated 95% confidence intervals of gill injuries. Detailed results are provided as Supplementary Material. The abbreviations for the metal and gill injury names are described in the text.

Fig. 3. Microphotographs of gill histopathological changes observed in the captured fish. A) nase gill filament showing no apparent changes; B) high severity edema (Ed) and lifting (Lf) of lamellar epithelium; C) filament epithelium proliferation (FEP) that in most instances conducted to lamellar fusion (LF); D) aneurisms with the higher level of severity; E) vasodilatation (Vas) in the base of the lamellar vascular axis and low severity edema (Ed); F) necrosis (Ne) of filament epithelial cells. FE, filament epithelium; L, lamella. Bars correspond to 50 μm.

222

A.R. Fonseca et al. / Science of the Total Environment 599–600 (2017) 217–226

Table 2 Goodman Kruskal correlation coefficients of metal concentration (e.g. Al, As) versus fish gill injuries (e.g. LF, FEP). The values outside the dark grey shaded cells are significant at p b 0.05 (n = 249). FEP is the injury better correlated with the metal concentrations, being succeeded by LF and Ne (light grey shaded cells). The data on which the coefficients are standing are provided as Supplementary Material. The abbreviations for the metal and gill injury names are described in the text.

LF

Al

As

Cd

Pb

Co

Cu

Cr

Mn

Zn

Ni

0.21

0.24

0.36

0.20

–0.15

0.54

0.47

0.45

0.52

0.33 0.27

FEP

0.29

0.33

0.29

0.16

–0.13

0.62

0.47

0.52

0.57

LEP

0.11

0.13

0.17

0.13

–0.11

0.27

0.29

0.24

0.17

0.09

Vas

0.15

0.15

–0.05

–0.25

–0.21

–0.06

0.13

0.16

–0.09

–0.03

Aneu

0.32

0.27

0.10

–0.09

–0.28

0.21

0.36

0.41

0.23

0.13

Ed

0.40

0.27

0.01

–0.18

–0.44

0.16

0.31

0.30

0.09

–0.01

Lif

0.21

0.16

–0.13

–0.19

–0.15

–0.06

0.07

0.08

–0.09

0.06

Ne

0.35

0.30

0.26

–0.12

–0.29

0.36

0.31

0.46

0.34

0.21

correlation patterns of LF and Ne were identical to the FEP fingerprint, because the highest positive coefficients were also found for Cu, Zn and Mn. The other lesions correlated with 8 (LEP, Aneu), 7 (Ed), 6 (Vas) and 5 (Lif) metals. The correlation patterns of LEP and Aneu resembled those of FEP, LF and Ne, but some changes were noted for Ed related to a higher importance of Al (0.40) and a loss of correlation with Zn. For the remaining lesions (Vas and Lif), some coefficients were significant but they were always rather low (b0.25). The method of PLS regression for determination of multiple associations between metal concentrations and gill injuries was applied twice. In the first run all collected fish were included in the analysis, regardless of species. In the second run, species (barbel, boga and trout) were treated separately. The results of PLS regression for the first run are illustrated in Fig. 4a. In this case, only the FEP has a significant portion of system variance explained by the model (close to 50%). For this injury, the PLS regression coefficients were determined and summarized in Fig. 4b. Corroborating the Goodman and Kruskal's gamma correlation analysis, Cu, Zn, Mn and Cr are heavy metals with a predominant link to the fish gill histopathological changes. In PLS regression, these metals are accompanied by As. The second run results are illustrated in Fig. 5. The upper panel (Fig. 5a–c) depicts model performances (R2) while the lower panel (Fig. 5d–f) depicts regression coefficients for injuries with R2 N 50%. The lower diagrams evidence marked differences among species as regards response to metal toxicity. In the barbel,

histopathological changes are dominated by Ed and mostly influenced by copper concentrations. In the boga, the PLS regression models show R2 N 50% for LF, FEP and LEP and high positive regression coefficients for Zn, Cr, Pb, Cd and As, although the links to Pb, Cd and As were valid solely to LF and LEP. Finally, the PLS regression models for trout repeated a R2 N 50% for LF, but added a representative result for Vas and Ne. The important metals in this case were: Cr, Al and Zn, for Vas; Zn and Ni for Ne and Zn for LF. 4. Discussion In keeping with the results of chemical and correlation analyses, metal concentrations in freshwater can cause irreversible histopathological changes in fish gills irrespective of meeting legally imposed thresholds. This is an important outcome because it may question the validity of environmental quality objectives defined by European and National authorities for river water. The impacts of Cu, Zn, Mn and Cr are prominent, because these metals correlate significantly with FEP, LF, Ne, LEP and Aneu that make up 62.5% of all studied lesions. A possible reason for the prominent impacts of Cu, Mn and Zn is the larger concentration of these metals in the collected water samples. The association of Ne to high concentrations of Mn and Zn has been reported in other studies (Al-Weher, 2008; Rajkowska and Protasowicki, 2013). The results of PLS regression portrayed in Fig. 4b confirm the link of FEP to the

Fig. 4. Results of Partial Least Squares (PLS) regression. First run: without discrimination of fish species (n = 249). (a) Model performance – only the severity of FEP are explained by increasing metal concentrations. In this case, PLS regression explains approximately 50% of FEP variance, which is noteworthy given the dataset characteristics (see details in the discussion section). Acronyms of histopathological changes (e.g. Ed for edema) are described in the text; (b) Regression coefficients – severity of FEP is most sensitive to the rise of arsenic (As) and zinc (Zn) concentrations. The role of copper (Cu) and manganese (Mn) is also noteworthy, while cadmium (Cd), lead (Pb), cobalt (Co) and nickel (Ni) seem to have little influence on the development of FEP.

A.R. Fonseca et al. / Science of the Total Environment 599–600 (2017) 217–226 Fig. 5. Results of Partial Least Squares regression. Second run: with discrimination of species (n = 112 for barbel, n = 73 for boga and n = 64 for trout). (a) Model performance – the injuries with highest performance (R2 N 50%) are dependent on the fish species: Ed performs well on barbel; LF, FEP and LEP on boga; LF, Vas and Ne on trout. (b) Regression coefficients – the response to changes in the metal concentrations are also dependent on the species, with boga responding to more metals than the other species. The abbreviations for the metal and gill injury names are described in the text.

223

224

A.R. Fonseca et al. / Science of the Total Environment 599–600 (2017) 217–226

concentrations of Cu, Zn, Mn and Cr uncovered by the correlation analysis, but add the influence of As probably because of its high toxicity even at lower concentrations. It is therefore acceptable to think that FEP is the gill damage more sensitive to the raise of Cu, Zn, Mn, Cr and As concentrations in stream water, at least in the studied basins. For that reason, FEP is viewed as attractive biomarker to use in monitoring studies of river water quality and ecological status in northern Portuguese watersheds. The PLS regression model for FEP explains 50% of system variance (Fig. 4a). The explanation of just half the total variance may look low, but one should recall some database specificities before interpreting this value. In the Supplementary material, different scores of gill lesion severity are presented for each fish in each site and campaign. But the metal concentrations are equal for all fish individuals in the same site and campaign, because the number of samples collected in each campaign was solely one per site. The repetition of metal concentrations in the database inevitably reduces system variance and consequently the values of R2 plotted in Fig. 4a. For this reason, we are confident that R2 ≈ 0.5 is enough model performance to assume a cause-effect relationship between the increase in the Cu, Zn, Mn, Cr and As concentrations and the increase in the FEP severity. The results for aluminum are striking because the regression coefficient for Al in the FEP model is negative. It is worth recalling, however, that gill damages caused by aluminum are usually noticed for concentrations above 100–200 μg·L−1 (Dietrich and Schlatter, 1989; Slaninova et al., 2014), while the collected sample waters showed concentrations lower than that range in two thirds of the cases (24 out of 36). Returning to the results of gamma correlations (Table 2), Al seems to play a key role in development of Ed, a result also reported in other studies (Hadi and Alwan, 2012). The histopathological changes Vas and Lif showed the highest severity levels (2.70 and 2.56, respectively; Fig. 2b) but the correlation with metal concentrations were either inexistent or characterized by very low coefficients (Table 2). Besides, the PLS regression models for Vas and Lif (Fig. 4a) revealed very low performances (R2 ≈ 0.2). The suggestion to make is that Vas and Lis severity in the studied fish species and stream reaches are influenced by contaminants other than the assortment of metals included in the present statistical analyses. The results portrayed in Fig. 5 refine the analysis of Fig. 4. Boga is probably the best species to monitor FEP because the PLS regression model in this case reached a performance close to 70% (Fig. 5b). The lesions LF and Ne, which correlated significantly with all metals (Table 2), should be important biomarkers in boga and trout (Fig. 5b,c), while LEP and Ed that correlated with 8 and 7 metals, respectively, appear to be relevant biomarkers in barbel and boga (Fig. 5a,b). Finally, although the general correlation analysis revealed poor results for Vas, the use of vasodilatation as biomarker in trout seems to be statistically significant (Fig. 5c). The severity of histopathological changes in boga (LF, LEP and FEP) responded to the concentration increase of most metals (Fig. 5e) while the Ed in barbel was sensitive just to copper (Fig. 5d). The results for boga point to the use this species in water quality monitoring programs of northern Portuguese rivers based on biomarkers. Barbel is a native species in Portuguese rivers while trout has been introduced. However, both species seem to be relatively tolerant to the presence of xenobiotics in freshwater, at least when compared to boga. Segurado et al. (2011) have also classified barbel and trout as intermediately tolerant to human perturbation. So, the prevailing assumption in literature that non-native fish are more tolerant than native species to water quality degradation can be questioned. The histopathological changes of fish gills related to the presence of xenobiotics in freshwater have been widely reported as environmental biomarkers (Camargo and Martinez, 2007; Jiang et al., 2011; Mazon et al., 2002; Pawert et al., 1998). The results of our study (non-parametric correlations coupled with PLS regressions) corroborate the conclusions of former works as regards FEP, LF, Ne, LEP and Ed, but especially when

attention is given to the FEP results. Cell proliferation probably occurred as defense mechanism (Ventura and Paperna, 1985) towards heavy metal insult, leading not only to increased epithelial thickness but also, in extreme cases, to lamellar fusion. In fact, the increased thickness of epithelium prevents further chemical absorption, similarly to what has previously been observed in other biomonitoring studies (Bentivegna et al., 2015; Costa et al., 2009). On the other hand, this causes the adverse effect of compromising gas and ion exchange mechanisms. The FEP is considered a protective measure against toxicants action since it increases the diffusion distance between the organism and the environment (Mallatt, 1985). The correlation between FEP and heavy metal concentrations, also observed by other authors (Hwang and Tsai, 1993; Karlsson-Norrgren et al., 1986; Pane et al., 2004; Reid and McDonald, 1988), is however not specific from this class of contaminants. It is also reported for fish gills exposed to an ample set of pollutants, including pulp and paper mill effluents, crude oil, ammonium and detergents (Bentivegna et al., 2015; Costa et al., 2009; Mallatt, 1985). Some studies reported epithelium edema and proliferation as the more frequently observed lesions in fish gills exposed to heavy metals (Chen and Lin, 2001; Pane et al., 2004; Schwaiger et al., 2004; van Heerden et al., 2004).

5. Conclusions In this study, a number of histopathological changes and their severity in gills of barbel, boga and trout have been related to metal concentrations in freshwater collected from six river basins with ecological status ranging from good to bad. The watersheds are located in northern Portugal and were subject in the recent past to different anthropogenic pressures derived from domestic, agricultural and industrial sources. The relationship between metal concentration and gill damage was set up on the basis of pairwise Goodman Kruskal correlation analysis combined with Partial Least Squares (PLS) regression. The chemical analyses revealed high concentrations of zinc, aluminum, manganese and copper in the freshwater samples. The histological observations uncovered several alterations in the fish gills, such as lamellar fusion, filament epithelium proliferation, laminar epithelium proliferation, vasodilation, aneurisms, edema, lifting and necrosis. The damage on the gills may imply the hindering of key physiological functions such as gas exchange and osmotic balance. The analyses of pairwise correlations and of multiple regressions (PLS), exposed statistically significant links of zinc, arsenic, manganese, chromium and copper to filament epithelium proliferation (FEP). The link of metal concentration increase to FEP severity increase was especially noted in boga. Similar associations were recognized between these metals and lamellar fusion, necrosis, lamellar epithelium proliferation and aneurisms. These cause-effect relationships were considered subsidiary when compared to the link of FEP, because they were discovered by the correlation analysis but not entirely confirmed by the performed PLS regressions. The impact of zinc, aluminum, manganese and copper can be explained by the high concentrations in freshwater. The impact of arsenic is probably related to the high toxicity of this substance even at low concentrations. For the remaining lesions, no prominent associations to metal concentrations were exposed by the statistical analyses. The significant correlations existed irrespectively of metal concentrations being in keeping with legally defined quality limits. This is a matter for reflection by the environmental authorities and the general public. The overall conclusions were that gill alterations resulting from exposure to heavy metals can effectively act as biomarker to the toxicity of sub lethal heavy metal concentrations. This is especially true for FEP in boga, the reason why this injury and this species should be preferably used in monitoring programs of ecological status rivers. Although dependent on the availability of boga in the monitored streams, the option for this species would ensure reduction in the operational costs and execution time of those programs.

A.R. Fonseca et al. / Science of the Total Environment 599–600 (2017) 217–226

Acknowledgments

225

Inc., 2004). In this study we used the SIMPLS algorithm embedded in that software.

This research was funded by national funds (FCT - Portuguese Science and Technology Foundation) under the strategic project of the Vila Real Chemistry Research Center (PEst-OE/QUI/UI0616/2014) and the CITAB (Centre for the Research and Technology of Agro-Environmental and Biological Sciences) project UID/AGR/04033.

Appendix B. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.scitotenv.2017.04.196.

Appendix A. Partial Least Squares (PLS) regression References Concise descriptions of PLS regression are presented in Abdi (2003), Helland (1990), or Statsoft Inc. (2004) and will be synthesized here. The goal of PLS regression is to predict Y from X and to describe their common structure. Some statistical approaches attempt to reach this goal by eliminating some predictors (e.g., using stepwise methods) or by performing a Principal Component Analysis (PCA) of the X matrix and then use the principal components of X as regressors on Y. The orthogonality of the principal components eliminates the multicolinearity problem often affecting the X variables. But, the problem of choosing an optimum subset of predictors remains. A possible strategy is to keep only a few of the first components. But they are chosen to explain X rather than Y, and so nothing guarantees that the principal components, which explain X, are relevant for Y. By contrast, PLS regression finds components from X that are also relevant for Y. Specifically, PLS regression searches for a set of components (called latent vectors) that performs a simultaneous decomposition of X and Y with the constraint that these components explain as much as possible of the covariance between X and Y. This step generalizes PCA, being followed by a regression step where the decomposition of X is used to predict Y (Abdi, 2003). PLS regression decomposes both X and Y as a product of a common set of orthogonal factors (F) and a set of specific loadings (L). So, the independent and the dependent variables are decomposed as: X ¼ F x LTx þ Rx

ðA:1aÞ

Y ¼ F y LTy þ Ry

ðA:1bÞ

with FT F ¼ I

ðA:1cÞ

where the superscript “T” refers to the transpose matrix, I is the identity matrix and R is the matrix of residuals that models the noise. As with PCA, the elements of F and L are termed factor scores and loadings, respectively. The above equations represent the so-called outer relations of the PLS regression model. Just as in all regression models, it is the aim of PLS regression to minimize the residuals as much possible while also having a representation of the relationship between X and Y that generalizes well for unseen (validation) data. Because of this (inner) relationship, the principal components used for representing X and Y cannot be calculated separately. In other words, when building the PLS regression model the outer relationships should know about each other so they cooperate in building the model as a whole, i.e. build a predictive model capable of relating X to Y. This means that the outer relations cannot be separately handled but rather treated as parts of the same problem. This condition is fulfilled by re-writing the second outer equation as (Statsoft Inc., 2004): Y ¼ F x BLTy þ Ry

ðA:2Þ

where B is the matrix of coefficients that relates the matrix of x-scores (Fx) to the matrix of y-scores (Fy). There are several algorithms which can handle efficiently the PLS regression theory, namely the NIPALS algorithm of Geladi and Kowalski (1986) or the SIMPLS algorithm of Jong (1993), both available in standard software packages such as the STATISTICA version 7 (Statsoft

Abdi, H., 2003. Partial least squares (PLS) regression. In: Lewis-Beck, M., Bryman, A., Futing, T. (Eds.), Encyclopedia of Social Sciences Research Methods. Sage, Thousand Oaks (CA), pp. 792–795. Al-Weher, S., 2008. Levels of heavy metal Cd, Cu and Zn in three fish species collected from the northern Jordan valley, Jordan. Jordan Journal of Biological Sciences 1, 41–46. Ambiente, M., 1998. Decreto-Lei n° 236/98. Diário da República I Série A. pp. 3676–3722. Aneiros-Pérez, G., González-Manteiga, W., Vieu, P., 2004. Estimation and testing in a partial linear regression model under long-memory dependence. Bernoulli 10, 49–78. Bentivegna, C.S., Cooper, K.R., Olson, G., Pena, E.A., Millemann, D.R., Portier, R.J., 2015. Chemical and histological comparisons between Brevoortia sp. (menhaden) collected in fall 2010 from Barataria Bay, LA and Delaware Bay, NJ following the DeepWater Horizon (DWH) oil spill. Mar. Environ. Res. 112, 21–34. Bradl, H., 2005. Heavy Metals in the Environment: Origin, Interaction and Remediation. Academic Press. Brito, E.M.S., De la Cruz Barrón, M., Caretta, C.A., Goñi-Urriza, M., Andrade, L.H., CuevasRodríguez, G., Malm, O., Torres, J.P.M., Simon, M., Guyoneaud, R., 2015. Impact of hydrocarbons, PCBs and heavy metals on bacterial communities in Lerma River, Salamanca, Mexico: investigation of hydrocarbon degradation potential. Sci. Total Environ. 521–522, 1–10. Caetano, M., Nunes, V., Nunes, A., 2009. CORINE Land Cover 2006 for Continental Portugal. Instituto Geográfico Português. Camargo, M.M., Martinez, C.B., 2007. Histopathology of gills, kidney and liver of a neotropical fish caged in an urban stream. Neotropical Ichthyology 5, 327–336. Cappello, T., Brandão, F., Guilherme, S., Santos, M.A., Maisano, M., Mauceri, A., Canário, J., Pacheco, M., Pereira, P., 2016. Insights into the mechanisms underlying mercury-induced oxidative stress in gills of wild fish (Liza aurata) combining 1H NMR metabolomics and conventional biochemical assays. Sci. Total Environ. 548–549, 13–24. Chen, H., 1988. Convergence rates for parametric components in a partly linear model. Ann. Stat. 16, 136–146. Chen, J.-C., Lin, C.-H., 2001. Toxicity of copper sulfate for survival, growth, molting and feeding of juveniles of the tiger shrimp, Penaeus monodon. Aquaculture 192, 55–65. Colin, N., Porte, C., Fernandes, D., Barata, C., Padrós, F., Carrassón, M., Monroy, M., CanoRocabayera, O., de Sostoa, A., Piña, B., Maceda-Veiga, A., 2016. Ecological relevance of biomarkers in monitoring studies of macro-invertebrates and fish in Mediterranean rivers. Sci. Total Environ. 540, 307–323. Costa, P.M., Diniz, M.S., Caeiro, S., Lobo, J., Martins, M., Ferreira, A.M., Caetano, M., Vale, C., DelValls, T.Á., Costa, M.H., 2009. Histological biomarkers in liver and gills of juvenile Solea senegalensis exposed to contaminated estuarine sediments: a weighted indices approach. Aquat. Toxicol. 92, 202–212. Dietrich, D., Schlatter, C., 1989. Aluminium toxicity to rainbow trout at low pH. Aquat. Toxicol. 15, 197–212. Directive, C., 1986. 86/609/EEC of 24 November 1986 on the approximation of laws, regulations and administrative provisions of the Member States regarding the protection of animals used for experimental and other scientific purposes. Off. J. Eur. Communities 29, L358. Directive, W.F., 2000. Directive 2000/60. Common Implementation 23. Duffus, J.H., 2002. “Heavy metals” a meaningless term? (IUPAC technical report). Pure Appl. Chem. 74, 793–807. Engle, R.F., Granger, C.W., Rice, J., Weiss, A., 1986. Semiparametric estimates of the relation between weather and electricity sales. J. Am. Stat. Assoc. 81, 310–320. ESRI, 2010. ArcMap (Version 10). New York St., Redlands, USA. Falk, R.F., Tonkin, P., 2001. Soft Modelling the Predictors of Drug Treatment Use. Sociology at Surrey, p. 4. Fame, C., 2005. Manual for the Application of the European Fish Index–EFI. A Fish-based Method to Assess the Ecological Status of European Rivers in Support of the Water Framework Directive. Version 1. Fergusson, J.E., 1990. Heavy Elements: Chemistry, Environmental Impact and Health Effects (Pergamon). Ferreira, A.R.L., Sanches Fernandes, L.F., Cortes, R.M.V., Pacheco, F.A.L., 2017. Assessing anthropogenic impacts on riverine ecosystems using nested partial least squares regression. Sci. Total Environ. 583, 466–477. Fonseca, A.R., Sanches Fernandes, L.F., Fontainhas-Fernandes, A., Monteiro, S.M., Pacheco, F.A.L., 2016. From catchment to fish: impact of anthropogenic pressures on gill histopathology. Sci. Total Environ. 550, 972–986. Geladi, P., Kowalski, B.R., 1986. Partial least-squares regression: a tutorial. Anal. Chim. Acta 185, 1–17. Goodman, L.A., Kruskal, W.H., 1954. Measures of association for cross classifications. J. Am. Stat. Assoc. 49 (268), 732–764. Goyer, R.A., Clarkson, T.W., 1996. In: Klaassen, C.D. (Ed.), Toxic Effects of Metals. Casarett & Doull’s Toxicology. The Basic Science of Poisons, fifth ed. McGraw-Hill Health Professions Division (ISBN 71054766).

226

A.R. Fonseca et al. / Science of the Total Environment 599–600 (2017) 217–226

Hadi, A.A., Alwan, S.F., 2012. Histopathological changes in gills, liver and kidney of fresh water fish, Tilapia zillii, exposed to aluminum. Int. J. of Pharm. & Life Sci. 3 (11), 2071–2081. Hauer, F.R., Resh, V.H., 1996. Benthic macroinvertebrates. In: Hauer, F.R., Lamberti, G.A. (Eds.), Methods in Stream Ecology. Academy Press, New York, USA, pp. 339–369. He, Z.L., Yang, X.E., Stoffella, P.J., 2005. Trace elements in agroecosystems and impacts on the environment. J. Trace Elem. Med. Biol. 19, 125–140. Helland, I.S., 1990. Partial least regression and statistical models. Scand. J. Stat. 17, 97–114. Herawati, N., Suzuki, S., Hayashi, K., Rivai, I., Koyama, H., 2000. Cadmium, copper, and zinc levels in rice and soil of Japan, Indonesia, and China by soil type. Bull. Environ. Contam. Toxicol. 64, 33–39. Hermoso, V., Clavero, M., 2013. Revisiting ecological integrity 30 years later: non-native species and the misdiagnosis of freshwater ecosystem health. Fish Fish. 14, 416–423. Hinton, D.E., Lauren, D.J., 1990. Integrative histopathological approaches to detecting effects of environmental stressors on fishes. Am. Fish. Soc. Symp. 51–66. Hwang, P., Tsai, Y., 1993. Effects of arsenic on osmoregulation in the tilapia Oreochromis mossambicus reared in seawater. Mar. Biol. 117, 551–558. Jiang, J., Gu, X., Song, R., Zhang, Q., Geng, J., Wang, X., Yang, L., 2011. Time-dependent oxidative stress and histopathological changes in Cyprinus carpio L. exposed to microcystin-LR. Ecotoxicology 20, 1000–1009. Jong, S., 1993. SIMPLS: an alternative approach to partial least squares regression. Chemom. Intell. Lab. Syst. 18 (3), 251–263. Karlsson-Norrgren, L., Dickson, W., Ljungberg, O., Runn, P., 1986. Acid water and aluminium exposure: gill lesions and aluminium accumulation in farmed brown trout, Salmo trutta L. J. Fish Dis. 9, 1–9. Mallatt, J., 1985. Fish gill structural changes induced by toxicants and other irritants: a statistical review. Can. J. Fish. Aquat. Sci. 42, 630–648. Mazon, A., Cerqueira, C., Fernandes, M., 2002. Gill cellular changes induced by copper exposure in the South American tropical freshwater fish Prochilodus scrofa. Environ. Res. 88, 52–63. Monroy, M., Maceda-Veiga, A., Sostoa, A., 2014. Metal concentration in water, sediment and four fish species from Lake Titicaca reveals a large-scale environmental concern. Sci. Total Environ. 487, 24–233. Monteiro, S., Rocha, E., Fontaínhas-Fernandes, A., Sousa, M., 2008. Quantitative histopathology of Oreochromis niloticus gills after copper exposure. J. Fish Biol. 73, 1376–1392. Nowak, B.F., Lucas, C.T., 1997. Diagnosis of structural changes in fish gills—can biopsy replace necropsy? Aquaculture 159, 1–10. Oliveira, R., Lima, M., Vieira, J., 2005. An Indicator System for Surface Water Quality in River Basins. Pacheco, F.A.L., Landim, P.M.B., 2005. Two-way regionalized classification of multivariate data sets and its application to the assessment of hydrodynamic dispersion. Math. Geol. 37 (4), 393–417. Pacheco, F.A.L., Van der Weijden, C.H., 2012. Integrating topography, hydrology and rock structure in weathering rate models of spring watersheds. J. Hydrol. 428–429, 32–50. Pacheco, F.A.L., Van der Weijden, C.H., 2014. Modeling rock weathering in small watersheds. J. Hydrol. 513C, 13–27. Pacheco, F.A.L., Landim, P.M.B., Szocs, T., 2013. Anthropogenic impacts on mineral weathering: a statistical perspective. Appl. Geochem. 36, 34–48. Pacheco, F.A.L., Pires, L.M.G.R., Santos, R.M.B., Sanches Fernandes, L.F., 2015a. Factor weighting in DRASTIC modeling. Sci. Total Environ. 505, 474–486. Pacheco, F.A.L., Sanches Fernandes, L.F., 2016. Environmental land use conflicts in catchments: a major cause of amplified nitrate in river water. Sci. Total Environ. 548, 173–188. Pacheco, F.A.L., Santos, R.M.B., Sanches Fernandes, L.F., Pereira, M.G., Cortes, R.M.V., 2015b. Controls and forecasts of nitrate fluxes in forested watersheds: a view over mainland Portugal. Sci. Total Environ. 537, 421–440. Pacyna, E.G., Pacyna, J.M., Fudala, J., Strzelecka-Jastrzab, E., Hlawiczka, S., Panasiuk, D., Nitter, S., Pregger, T., Pfeiffer, H., Friedrich, R., 2007. Current and future emissions of selected heavy metals to the atmosphere from anthropogenic sources in Europe. Atmos. Environ. 41, 8557–8566. Pandey, S., Parvez, S., Ansari, R.A., Ali, M., Kaur, M., Hayat, F., Ahmad, F., Raisuddin, S., 2008. Effects of exposure to multiple trace metals on biochemical, histological and ultrastructural features of gills of a freshwater fish, Channa punctata Bloch. Chem. Biol. Interact. 174, 183–192. Pane, E.F., Haque, A., Wood, C.M., 2004. Mechanistic analysis of acute, Ni-induced respiratory toxicity in the rainbow trout (Oncorhynchus mykiss): an exclusively branchial phenomenon. Aquat. Toxicol. 69, 11–24. Pawert, M., Müller, E., Triebskorn, R., 1998. Ultrastructural changes in fish gills as biomarker to assess small stream pollution. Tissue Cell 30, 617–626. PGRH, 2011. Plano de Gestão da Região Hidrográfica do Cávado, Ave e Leça. Technical Report [in Portuguese], Administração da Região Hidrográfica do Norte, I.P. 214p. Portaria. n°. 1005/92, 1992. Ministérios da Agricultura, da Educação, da Saúde e do Comércio e Turismo. Diário da Répública I. Rajkowska, M., Protasowicki, M., 2013. Distribution of metals (Fe, Mn, Zn, Cu) in fish tissues in two lakes of different trophy in Northwestern Poland. Environ. Monit. Assess. 185, 3493–3502.

Reid, S., McDonald, D., 1988. Effects of cadmium, copper, and low pH on ion fluxes in the rainbow trout, Salmo gairdneri. Can. J. Fish. Aquat. Sci. 45, 244–253. Rengel, Z., 1999. Heavy metal stress in plants. Heavy Metals as Essential Nutrients. Springer, pp. 231–251. Rivera-Guzmán, N.E., Moreno-Casasola, P., Ibarra-Obando, S.E., Sosa, V.J., Herrera-Silveira, J., 2014. Long term state of coastal lagoons in Veracruz, Mexico: effects of land use changes in watersheds on seagrasses habitats. Ocean Coast. Manag. 87, 30–39. Sanches Fernandes, L.F., Seixas, F., Oliveira, P., Leitao, S., Moura, J., 2012. Climate-change impacts on nitrogen in a hydrographical basin in the northeast of Portugal. Fresenius' Environ. Bull. 21 (12), 3643–3650. Santos, R.M.B., Sanches Fernandes, L.F., Moura, J.P., Pereira, M.G., Pacheco, F.A.L., 2014. The impact of climate change, human interference, scale and modeling uncertainties on the estimation of aquifer properties and river flow components. J. Hydrol. 519, 1297–1314. Santos, R.M.B., Sanches Fernandes, L.F., Pereira, M.G., Cortes, R.M.V., Pacheco, F.A.L., 2015a. Water resources planning for a river basin with recurrent wildfires. Sci. Total Environ. 526, 1–13. Santos, R.M.B., Sanches Fernandes, L.F., Pereira, M.G., Cortes, R.M.V., Pacheco, F.A.L., 2015b. A framework model for investigating the export of phosphorus to surface waters in forested watersheds: implications to management. Sci. Total Environ. 536, 295–305. Santos, R.M.B., Sanches Fernandes, L.F., Varandas, S.G.P., Pereira, M.G., Sousa, R., Teixeira, A., Lopes-Lima, M., Cortes, R.M.V., Pacheco, F.A.L., 2015c. Impacts of climate change and land-use scenarios on Margaritifera margaritifera, an environmental indicator and endangered species. Sci. Total Environ. 511, 477–488. Scardi, M., Cataudella, S., Di Dato, P., Fresi, E., Tancioni, L., 2008. An expert system based on fish assemblages for evaluating the ecological quality of streams and rivers. Ecological informatics 3, 55–63. Schick, A., 1996. Root-n consistent estimation in partly linear regression models. Statistics & Probability Letters 28, 353–358. Schwaiger, J., Ferling, H., Mallow, U., Wintermayr, H., Negele, R., 2004. Toxic effects of the non-steroidal anti-inflammatory drug diclofenac: part I: histopathological alterations and bioaccumulation in rainbow trout. Aquat. Toxicol. 68, 141–150. Segurado, P., Santos, J.M., Pont, D., Melcher, A.H., Jalon, D.G., Hughes, R.M., Ferreira, M.T., 2011. Estimating species tolerance to human perturbation: expert judgment versus empirical approaches. Ecol. Indic. 11, 1623–1635. Simpson, J., Norris, R., Wright, J., Sutcliffe, D., Furse, M., 2000. Biological assessment of river quality: development of AUSRIVAS models and outputs, assessing the biological quality of fresh waters: RIVPACS and other techniques. Proceedings of an International Workshop Held in Oxford, UK, on 16–18 September 1997. Freshwater Biological Association (FBA), pp. 125–142. Sindern, S., Tremöhlen, M., Dsikowitzky, L., Gronen, L., Schwarzbauer, J., Siregar, T.H., Ariyani, F., Irianto, H.E., 2016. Heavy metals in river and coast sediments of the Jakarta Bay region (Indonesia) — Geogenic versus anthropogenic sources. Mar. Pollut. Bull. 110, 624–633. Slaninova, A., Machova, J., Svobodova, Z., 2014. Fish kill caused by aluminium and iron contamination in a natural pond used for fish rearing: a case report. Vet. Med. 59, 573–581. Speckman, P., 1988. Kernel smoothing in partial linear models. J. R. Stat. Soc. Ser. B Methodol. 413–436. Statsoft Inc., 2004. STATISTICA, Version 7. Tusla, USA. Stephansen, D.A., Nielsen, A.H., Hvitved-Jacobsen, T., Arias, C.A., Brix, H., Vollertsen, J., 2014. Distribution of metals in fauna, flora and sediments of wet detention ponds and natural shallow lakes. Ecol. Eng. 66, 43–51. Tchounwou, P.B., Yedjou, C.G., Patlolla, A.K., Sutton, D.J., 2012. Molecular, clinical and environmental toxicology. Heavy Metal Toxicity and the Environment. Springer, pp. 133–164. Valera, C., Valle Junior, R.F., Varandas, S.G.P., Sanches Fernandes, L.F., Pacheco, F.A.L., 2016. The role of environmental land use conflicts in soil fertility: a study on the Uberaba River basin, Brazil. Sci. Total Environ. 562, 463–473. Valle Junior, R.F., Varandas, S.G.P., Pacheco, F.A.L., Pereira, V.R., Santos, C.F., Cortes, R.M.V., Sanches Fernandes, L.F., 2015. Impacts of land use conflicts on riverine ecosystems. Land Use Policy 43, 48–62. van Heerden, D., Vosloo, A., Nikinmaa, M., 2004. Effects of short-term copper exposure on gill structure, metallothionein and hypoxia-inducible factor-1α (HIF-1α) levels in rainbow trout (Oncorhynchus mykiss). Aquat. Toxicol. 69, 271–280. Ventura, M., Paperna, I., 1985. Histopathology of Ichthyophthirim multifiliis infections in fishes. J. Fish Biol. 27, 185–203. Vieira, J., Fonseca, A., Vilar, V.J., Boaventura, R.A., Botelho, C.M., 2012. Water quality in Lis river, Portugal. Environ. Monit. Assess. 184, 7125–7140. Vieira, J., Fonseca, A., Vilar, V.J., Boaventura, R.A., Botelho, C.M., 2013. Water quality modelling of Lis River, Portugal. Environ. Sci. Pollut. Res. 20, 508–524. Wold, H., 1966. Multivariate analysis. Estimation of Principal Components and Related Models by Iterative Least squares. Academic Press, pp. 391–420.