Chronic trace metals effects of mine tailings on estuarine assemblages revealed by environmental DNA

Introduction

Environmental assessment studies rely on accurate detection of biodiversity of an extremely diverse and small-sized benthic fauna. For decades, morphological methods are the basis to impact assessment (IA) protocols at the cost of neglecting an enormous number of meiofaunal species that could not be accurately identified (Bhadury et al., 2006; Fonseca et al., 2010). There have been considerable advances in recent years by applying DNA-sequence based techniques, commonly referred as metabarcoding, to identify and quantify meiofaunal biodiversity (Creer et al., 2010; Bik et al., 2012; Brannock et al., 2014). Metabarcoding techniques can now be applied to access levels of richness and spatial patterns of diversity on marine Metazoans with use of homologous genes (nuclear 18S rRNA), and help uncover the extreme high biodiversity of meiofaunal benthic assemblages (Fonseca et al., 2010). These modern approaches offer fast assessments of marine Metazoan meiofaunal assemblages and are particularly useful for the identification of new species in areas with poorly reported biodiversity, which may be of special interest in IA studies.

Estuarine coastal ecosystems offer an opportunistic case to evaluate biodiversity-environmental relationships through environmental DNA (eDNA) since many estuaries are widely impacted by pollutants with deleterious effects to benthic assemblages (Lotze et al., 2006; Chariton et al., 2015; Hadlich et al., 2018; Lana & Bernardino, 2018). The Samarco mine tailing disaster that occurred in Brazil on November 2015, released near 43 million m³ of tailings in the Rio Doce river basin, which were transported for over 600 km until reaching the estuary and the Atlantic Ocean (Carmo et al., 2017; Magris et al., 2019). The tailings severely impacted the Rio Doce riverine and estuarine ecosystems causing rapid sediment accumulation, burial and death of benthic organisms, and rapidly (1 and 2 days) increased sediment heavy metal accumulation by orders of magnitude from pre-impact conditions (Gomes et al., 2017). Although the released tailings had trace metal concentrations that were within the Brazilian legislation (Segura et al., 2016); the iron tailings deposited in the estuarine soils were heavily associated with trace metals which are potentially bioavailable given the redox conditions of estuarine soils (Queiroz et al., 2018).

Trace metal accumulation in coastal ecosystems are reported to be highly associated with changes in benthic assemblages and to increase human health risks due to potential bioaccumulation in food webs (Venturini, Muniz & Rodriguez, 2002; Muniz et al., 2004; Rainbow, 2007; Hauser-Davis et al., 2012). As a result, impact assessment studies that followed the Samarco disaster were also based on traditional morphological biodiversity assessments (Gomes et al., 2017). The potential chronic pollution effects in the Rio Doce estuary will likely demand long term monitoring programs for this ecosystem. To that end, technical and taxonomic expertise will be of key importance to monitor the estuarine biodiversity, but these efforts are typically limited to the macrofaunal and megafaunal benthic taxa. Therefore, monitoring this environmental disaster by increasing its biodiversity assessment to a broader range of cryptic and meiobenthic taxa may bring valuable information on the extension of impacts.

In this study, we used an eDNA metabarcoding approach to evaluate the benthic biodiversity in the Rio Doce estuary 1.7 years after the initial impacts of the Samarco disaster. We hypothesized that spatial patterns of chronic metal contamination in the estuary would be significantly associated with patterns of meiofaunal environmental taxonomic units (eOTUs), evidencing the potential use of this technique for long term impact assessment of the estuary. We targeted benthic meiofaunal eukaryote organisms by amplifying and sequencing the V9 hypervariable region of the 18S ribosomal gene from purified eDNA. In addition, sediment variables (particle size, organic carbon content) and trace metals concentrations were used to test for spatial changes in benthic assemblages in response to contamination levels in the estuary.

Materials and methods

Study site

The Rio Doce estuary (19°38′ to 19°45′ S, 39°45′ to 39°55′ W; Fig. 1), is located on the Eastern Marine Ecoregion of Brazil that has two well-defined seasons, dry winter (April–September) and wet summer (October–March), with an average monthly rainfall of 145 mm and temperatures of 24 to 26 °C (Bernardino et al., 2018; Bissoli & Bernardino, 2018). The Rio Doce estuary has been altered by historical human occupation, but ecosystem health of the estuary was poorly known before the Samarco disaster that occurred in November 2015 (Bernardino et al., 2016; Gomes et al., 2017). The initial impacts of the Samarco disaster in the estuary were reported by Gomes et al. (2017), and a standard monitoring of benthic assemblages and contamination levels were established in 2017 with a disaster-response program funded by Brazilian government agencies (Fapes, Capes and CNPq). The first monitoring campaign occurred in August 2017 (SISBIO sampling license N 24700-1), approximately 1.7 years after the initial impacts were observed in the estuary, when we aimed to quantify the potential chronic effects of trace metal pollution that were first observed in November 2015.

Results

The Rio Doce estuary exhibited low salinities at the time of sampling (0.1–3.7). Sediments were dominated by sand particles (>62% sand), with the exception of site two which showed less sand-sized particles (12%; Table 1). Sediment TOM varied from 1.5 to 16.8%, with the highest organic content at stations three, 13 and 22 (16.8, 13.8 and 10.2%; respectively). Several estuarine areas had TOM in a similar range of 2–6.2% (Table 1), and sediment organic content was significantly correlated to Fe content (Pearson r = 0.5043, p = 0.023; Table S4). The concentration of trace metals in the estuarine sediments also varied markedly along the studied area (Table 1). Fe concentrations ranged from 18,814 to 54,982 mg kg⁻¹ and were highly correlated with several other trace metals including Al, Cd, Cr, Co, Cu, Mn and Zn (Tables S3 and S5).

We obtained a total of 9,836,039 sequence reads, of which 6,840,886 were of high quality. The number of sequence reads per station ranged from 35,915 (St 16) to 359,718 (St 4), with an average of 207,285 total sequence variants per station. Stations that had only one replicate sequenced had a lower (e.g., stations 14 and 16) or a similar number of reads (e.g., stations 19 and 20) of sites that had two replicates sequenced. On average, 55.4% of reads corresponded to aquatic or marine Metazoan taxa (Table S1). The eOTU richness per station ranged over three-fold from 16 to 54 eOTUs (Table 1). Assemblages were dominated by Nematoda (34 eOTUs), Platyhelminthes (19), Crustacea (18), Gastrotricha and Annelida (12 eOTUs each; Table S1; Fig. 2). Most sites had over 80% of sequence variant reads represented by two to three meiofaunal taxa, including the dominant Gastrotricha, Nematoda and Crustacea. The number of unassigned Metazoan taxa was large (>50%) at stations 16 and 17, whereas it remained less than 20% in most sites.

The eDNA species accumulation curves did not reach an asymptote with addition samples suggesting an yet incomplete biodiversity characterization of the estuary (Fig. 3). Several eOTUs (N = 88) were represented by less than 0.1% of sequence variant reads. When we removed the eOTUs that had less than 0.1% of sequence reads, the species accumulation stabilized at 32 eOTUs with five to seven samples, with no additional gain of taxa. The species accumulation asymptote with dominant eOTUs was reached in about half the number of samples necessary in morphology-based studies (12–14 samples; Fig. 3). The number of meiofaunal eOTUs (eOTU richness) were largely uncorrelated to sediment grain size (p = 0.161), Fe (p = 0.647) and organic matter content (p = 0.6395; Table S4; Fig. 4).

The multivariate patterns of dominant meiofaunal (S = 32) eOTU composition were significantly related to Fe contents in sediments (F = 2.89, p = 0.018, Fig. 5; Table 2). The CAP axes 1 and 2 explained 44% and 21% of multivariate variability; respectively (Table 2). Fe contents in sediments was associated to the multivariate distribution of meiofaunal eOTUs including the Nematoda Mesodorylaimus nigritulus and Epitobrilus stefanskii, Harpacticoid copepods, the Platyhelminthes Cirrifera dumosa and Bothrioplana sinensis, and Ostracods (Chrissia dongqianhuensis). Monhysteridae and Desmodorida spp. nematode worms were negatively correlated to Fe concentrations (CAP1 score= −0.25 to −0.18). Pb and As contamination were not correlated to Fe concentrations in sediments and were not significantly associated with the meiofaunal multivariate composition (Table 2).

Discussion

Our study demonstrates that environmental DNA can be an effective method to indicate chronic contamination effects on benthic assemblages of the Rio Doce estuary, supporting our hypothesis. This first eDNA survey in the Rio Doce estuary also revealed a previously unrecognized benthic biodiversity, even with significant impacts by trace metal levels 1.7 years after the initial impacts. Although there is no baseline eDNA assessment from the Rio Doce estuary, the impacted sediments potentially supported over 32 dominant meiofaunal taxa (eOTUs), with a spatial distribution significantly related to Fe (and correlated metals) contamination.

The Rio Doce eDNA composition was similar to other estuarine and marine sediments assessed by metabarcoding methods (Fonseca et al., 2010; Faria et al., 2018). Nematoda, Gastrotricha and Crustaceans were highly dominant in the estuary with local changes in relative abundance across sites sampled. The marked spatial variability in assemblage composition within the estuary indicates that benthic assemblages were spatially structured; which is a similar pattern commonly observed in morphology-based assessments. Environmental filtering in benthic assemblages may result from a combination of sediment and water variables, with grain size, salinity and food availability being critical to species turnover and replacement in estuarine benthos (Menegotto, Dambros & Netto, 2019). Although our study design does not allow for an hierarchical spatial analysis of variables that determined the observed environmental filtering, the detection of spatial variance in assemblages from eDNA samples suggests that the biodiversity assessment is likely representing living benthic organisms instead of predominantly ancient or allochthone DNA. There is now strong evidence supporting that eDNA techniques can detect complex spatial variability in estuarine and coastal marine ecosystems (Chariton et al., 2015; Faria et al., 2018); and our data additionally supports its use to biodiversity assessment in a heavily impacted estuary.

Most eOTUs represented new occurrences for the estuary, but yet with several unassigned taxa, stressing the complementarity value of molecular and morphological approaches to ecological and impact assessment studies (Leasi et al., 2018). We recovered a total of 123 environmental OTUs (eOTUs) in the Rio Doce estuarine sediments, increasing by over 20-fold the previous richness of benthic taxa based on morphological identifications (Gomes et al., 2017). The species accumulation curves did not reach an asymptote with addition of eDNA samples, and most eOTUs (N = 88) were represented by less than 0.1% of sequence variant reads, suggesting an yet incomplete biodiversity assessment of the Rio Doce estuary even with high levels of trace metals. However, estuaries are highly connected to continental and marine ecosystems and it is unlikely that species accumulation curves would reach an asymptote with a single biodiversity assessment (Chariton et al., 2015; Nascimento et al., 2018). The rapid increase and stabilization of the number of dominant meiofaunal OTUs with the addition of new samples suggests a reasonable beta-diversity assessment of the Rio Doce estuary with the effort taken. Sites that had only one sequenced replicate due to low DNA stocks attained similar or lower OTUs richness if compared to other stations, but the sediment volumes used in this study (>200 g) were well over the necessary to avoid technical bias in the detection of Metazoan diversity (Brannock & Halanych, 2015; Nascimento et al., 2018).

This single eDNA survey was efficient in assembling benthic meiofaunal assemblages in the Rio Doce estuary. The species accumulation curves indicate that half of the sampling effort would be necessary to characterize the dominant meiofaunal groups in the estuary if compared to the species accumulation rate of previous morphological assessments. eDNA metabarcoding can be more efficient at characterizing marine taxa (Lobo et al., 2017), and our data supports its use on long term studies where taxonomic and technical limitations cannot be controlled (Bista et al., 2017). The lack of controlled methods may be a crucial problem to the biodiversity monitoring that followed the Samarco disaster on the Rio Doce, given the extremely large scale and diversity of impacted ecosystems. It is estimated that over 2,000 ha of terrestrial, limnetic and estuarine ecosystems along the Rio Doce basin were directly impacted by the disaster (Carmo et al., 2017); with additional potential effects on nearby coastal zones (Magris et al., 2019). Given the large area, the diversity and natural complexity of ecosystems to be monitored, it is likely that the number of biological samples needed to reach reasonable statistical power to detect biodiversity impacts would be prohibitive (Fairweather, 1991). Therefore, the massive amount of data obtained in metabarcoding techniques could have a profound contribution to environmental monitoring in this scenario, which would also increase dramatically the discovery of cryptic species on a range of aquatic and terrestrial ecosystems.

Multivariate analysis revealed that Fe contents (and other correlated metals) are partially structuring spatial patterns of dominant benthic meiofaunal assemblages in the Rio Doce estuary 1.7 years after the disaster. The sediment Fe contents were significant predictors of changes in dominant meiofaunal eOTUs including nematodes, copepods, ostracods and flatworms. These groups corresponded to over 2/3 of meiofaunal OTUs in the estuary and revealed that trace metal contents are driving spatial patterns of the Rio Doce estuarine biodiversity. Our data suggest that benthic assemblages were highly sensitive to chronic metal contamination in polluted estuaries, and partially explains a lower effect of sediment grain size and organic matter on local meiofauna (e.g., Faria et al., 2018; Menegotto, Dambros & Netto, 2019). This could indicate that the Rio Doce estuarine assemblages were strongly impacted after the disaster through the exclusion of intolerant species, although we lack baseline eDNA to fully support that hypothesis.

The extremely high Fe contents allied to covariance of several potentially toxic trace metals that are adhered to iron oxides present in the tailings strongly suggest that the tailings have led to major changes in the estuarine benthic biodiversity since the initial impact (Queiroz et al., 2018). The Rio Doce basin was previously polluted by historical mining and urban activities, so the trace metals that rapidly accumulated in the estuary were likely transported downstream attached to Fe oxides from the released tailings. The initial impacts in the Rio Doce estuary were observed immediately with the arrival of tailings, which led to significant (2–20 times) increases in sediment Fe, Mn, Cr, Ni, Cu, Zn and As (Gomes et al., 2017). Sediment concentrations of Fe, Pb and selected trace metals in the Rio Doce in August 2017 continued to be 2–20 times higher compared to preserved (Piraquê-Açu-Mirim estuary) or polluted estuaries such as the Vitoria Bay, located in a major metropolitan and industrial area ~100 km to the south (Hadlich et al., 2018).

The statistical lack of As and Pb effects on the multivariate distribution and composition of meiofaunal assemblages have important implications for future environmental monitoring in the estuary. One plausible cause is that not all elements that are accumulated in the sediments are bioavailable and have toxicity to the estuarine biota. Given the amplitude of trace metals accumulated in the Rio Doce sediments since the mining tailing impacts that occurred in 2015, it is very likely that a combination of these contaminants lead to further changes in the estuarine benthos. The sediment concentrations of Pb in August 2017 were over 20 times higher than baseline values (Gomes et al., 2017); and several other trace metals also increased with time since the impact. Queiroz et al. (2018) reported a significant correlation between Fe, Pb and other trace metals in tailing deposits after the initial impacts in the Rio Doce estuary in 2015. The iron oxides from tailings deposited in the estuary have a strong capacity of metal retention (Cornell & Schwertmann, 2003; Yin et al., 2016); and they are likely to be released due dissimilatory iron reduction under estuarine conditions (Bonneville, Behrends & Van Cappellen, 2009; Queiroz et al., 2018; Xia et al., 2019). Although anoxic estuarine soils favor the formation of sulfides (e.g., Pyrite; AVS) which have strong affinity and role in the chelation of metals (Machado et al., 2010; Nóbrega et al., 2013), the Rio Doce estuary contrast to several other estuaries due to low salinity, low tidal influence and an apparent limited sulfate availability and sulfate reduction that reduce the formation of sulfides (Queiroz et al., 2018). As a result, the observed relationship of meiofaunal assemblages with Fe contents (and other trace metals) suggest that the tailings have some toxicity to benthic organisms even though a number of contaminants may not achieve alarming concentrations.

The effects of trace metal contents on the Rio Doce benthic assemblages resemble impacts in other areas that are highly polluted with trace metals, but these effects could be confounded with the constant environmental changes that typically occur in these ecosystems (Krull et al., 2014; Martins et al., 2015). Our approach of selecting dominant meiofaunal OTUs to multivariate analysis led to positive detection of Fe contents effects and associated trace metals. This approach was justifiable given that we detected 88 eOTUs with less than 0.1% of sequence variant reads, which could be potentially associated with allochthone DNA from connected river or ocean ecosystems and would not be under influence of local contaminants. The use of indicator taxa or functional groups to eDNA biodiversity assessment studies is becoming practice in ecological studies (e.g., Bista et al., 2017) and our approach offers an important methodological approach for detection of trace metals effects in aquatic biota that need to be further investigated in other case studies.

Conclusions

In conclusion, our eDNA survey of benthic meiofaunal assemblages in the Rio Doce estuary detected environmental filtering with strong influence of contamination by Fe and other correlated trace metals, supporting chronic mine tailing impacts in the estuary. Our study is also in agreement with previous assertions that ecological inferences from eDNA analysis may increase the performance of biodiversity assessments in marine ecosystems by capturing a range of cryptic taxa, thus greatly improving current short and long-term impact assessment studies. The use of eDNA to the Samarco mine tailing disaster would benefit monitoring assessments with standard techniques and dramatically increase our knowledge of the biodiversity of cryptic aquatic species. The continued sampling and monitoring would also increase the precision of the eDNA assessments, especially if allied to detailed morphological work.

Supplemental Information