Metal stress-related gene expression patterns in two marine invertebrates, Hediste diversicolor (Annelida, Polychaeta) and Littorina littorea (Mollusca, Gastropoda), at a former mining site

Abstract

Abandoned mines often contaminate sediments with dissolved heavy metals and are known to impact many taxa. Physiological responses needed to avoid oxidative stress from metal toxicity include the upregulation of detoxification and metal-binding proteins such as glutathione-s-transferases and metallothioneins, which have been studied in diverse species. Fewer studies, however, have focused on gene expression changes to better understand these molecular mechanisms, especially across multiple species at a single contaminated site. To this end, the purpose of this study was to characterize metal stress-related gene expression in two species from different phyla, Hediste diversicolor (Annelida, Polychaeta) and Littorina littorea (Mollusca, Gastropoda), at a former mine site (Callahan Mine, Maine, USA). Both species and sediments were collected from a mine-affected tidal estuary (Goose Pond) and a nearby reference site. Elevated sediment metal levels were confirmed at Goose Pond. H. diversicolor individuals weighed significantly less at Goose Pond, while L. littorea weighed similarly at both sites. Transcript levels were stable in H. diversicolor but weakly upregulated in L. littorea, which likely reflect the importance of other physiological strategies for metal sequestration, or variable metal exposure at the individual level, respectively. In addition, patterns in glutathiones-transferase expression differed across isoforms in H. diversicolor, while L. littorea exhibited divergent expression patterns in foot muscle and heaptopancreas. Overall, these results reinforce that diverse species likely undergo different physiological responses to metal toxicity, and more research is needed to investigate these mechanisms.

1. Introduction

Abandoned mines are a major source of metal contamination worldwide, and in the western United States alone, represent at least 33,000 sites with confirmed environmental damage and billions of dollars in taxpayer-funded remediation efforts (Fields, 2003; Mittal, 2011). Cleanup operations often take many years to complete and include processing of both open pit sites and crushed-rock tailings (Castro and Moore, 2000; Kossoff et al., 2014). These sites represent environmental hazards for decades or centuries, even with some remediation, and contaminate groundwater and surrounding sediments with dissolved metals (Fields, 2003).

Heavy metals from soil and water sources have been well-documented to affect numerous cellular processes across both plants and animals (Malcoe et al., 2002; Tchounwou et al., 2014). Although some metals such as copper (Cu) and zinc (Zn) are vital as cofactors for many biological processes, their homeostasis and sequestration is tightly regulated to avoid metal-induced toxicity (Lippard et al., 1999; Sekler et al., 2007). This toxicity is usually due to metal cation binding to biological molecules, which induces oxidative damage and impedes normal functions (Golovanova, 2008). For instance, excess free copper ions (Cu²⁺) are known to promote formation of reactive oxygen species that induce DNA damage, mitochondrial dysfunction, and membrane lipid destruction (Gaetke et al., 2014). Free zinc ions (Zn²⁺) are also toxic and activate apoptotic pathways and free-radical generating enzymes, which contribute to oxidative stress (Frederickson et al., 2005). Overexposure of other heavy metals, including arsenic (As), cadmium (Cd), manganese (Mn), and lead (Pb), are also known to induce damage through similar pathways (Li et al., 2001; Tchounwou et al., 2014; O’Neil and Zheng 2015).

In aquatic ecosystems, heavy metal effects from mining activities have been studied in many organisms, including plants, fish, and invertebrates (Mishra et al., 2008; Kiser et al., 2010; Wadige et al., 2017). These effects have been observed to bioaccumulate in higher trophic animals, which exhibit elevated metals even when water contamination is relatively low (Kiser et al., 2010; Has-Schön et al., 2015). Benthic macroinvertebrates acquire metals through feeding and close association with sediments (Woodward et al., 1994; Farag et al., 1999), and many species have been used as pollution biomarkers (Dallinger, 1993). These invertebrates exhibit diverse negative effects, including changes in antioxidant enzyme activity and metal accumulation in hepatopancreas tissues (Golovanova, 2008). Many species respond physiologically to elevated metals by increasing synthesis of metallothioneins (MT), which are low molecular weight proteins that bind these atoms and participate in metal homeostasis, detoxification, and other environmental stressors (English and Storey, 2003; Amiard et al., 2006). Many invertebrates also exhibit elevated gene expression of other antioxidant-related enzymes such as glutathione-s-transferases (gst) and superoxide dismutases (sod) (Lee et al., 2008; Fang et al., 2010; Won et al., 2012).

Gene expression approaches to evaluate metal effects represent powerful tools in understanding the molecular mechanisms of the pollutant-induced stress syndrome (Viarengo et al., 2007; Lauritano et al., 2012). Due to recent advances in RNA sequencing and transcriptome assembly, more species are being used to evaluate these expression changes, especially in laboratory settings (Sun et al., 2016). The use of transcript changes in biomonitoring of wild populations is limited, however, since: 1) many environmental factors can influence apparent expression patterns, including metal bioavailability or sequestration in tissues, 2) expression levels may not reflect relevant protein quantities or enzymatic activities, and 3) monitoring regimes are recommended to include multiple assessments across the stress syndrome (Viarengo et al., 2007). Still, a transcriptional approach in sentinel invertebrate species can provide new insights into particular isoforms that are most responsive to metal exposure, or act as foundational information to develop more comprehensive, applied biomonitoring tools such as microarrays or transcriptome sequencing (Viarengo et al., 2007). Most comparative physiological studies at metal-contaminated sites have not included expression analyses, however, and relatively few have studied multiple marine invertebrate species at a single location (Brown et al., 2004; Funes et al., 2005; Wang and Rainbow, 2008; Won et al., 2008).

The purpose of the present study was to characterize gene expression patterns during heavy metal exposure in two species, the ragworm Hediste diversicolor (Annelida, Nereididae) and common periwinkle Littorina littorea (Mollusca, Littorinidae). Both species commonly inhabit intertidal environments along the North Atlantic coast, including in USA waters and the Gulf of Maine (Chapman et al., 2007; Einfeldt et al., 2014). These species have previously been used in metal toxicity experiments, and sequences for some oxidative stress response genes, including inducible metallothionein isoforms, are available (English and Storey, 2003; McQuillan et al., 2014). Investigations into gene expression changes in these animals, however, are limited, and candidate genes in H. diversicolor and L. littorea tissues were evaluated using a quantitative PCR (qPCR) approach. To assess transcript changes, individuals were collected from contaminated estuarine habitat at the former Callahan Mine (Cape Rosier, Maine, USA), a rare, documented intertidal zone open pit site known to contaminate nearby soils, water, and fish with metals (Broadley et al., 2013). Heaviest mining activities at the Callahan Mine occurred in the late 1800s and from 1968–1972, after which it was abandoned (MACTEC, 2009; Broadley et al., 2013), and the location is now designated as a Superfund site by the U.S. Environmental Protection Agency (Ayuso et al., 2013; EPA, 2016).

2. Materials and Methods

2.1. Animal sampling

Hediste diversicolor and Littorina littorea were collected at low tide in October 2015 and September 2016, respectively, from two sites: 1) Horseshoe Cove (HC), an uncontaminated salt marsh, and 2) Goose Pond (GP), an adjacent tidal estuary impacted by the former Callahan Mine (Fig. 1). Although inner Horseshoe Cove was previously characterized by small salinity stratification and slightly lower water pH compared to Goose Pond, both sites are overall well-mixed tidal estuaries, and Horseshoe Cove was previously used a reference site (Broadley et al., 2013). Adult H. diversicolor individuals were collected from HC and GP sites (n = 34 and 25, respectively) within a single day and by hand from random sediment samples collected within approximately 0.2 km of each location (HC: 44°21’10.38”N, 68°46’17.71”W; GP: 44°20’41.20”N, 68°48’14.43”W). Individuals were briefly rinsed in seawater, weighed, and a random subsample (n= 9 adults/site) were placed in 1.5 ml microcentrifuge tubes with ice cold RNALater (Ambion, Austin, TX) for later RNA extractions. Adult L. littorea were also collected randomly by hand (n = 20/site) from rocky intertidal surfaces at both locations (HC: same as above; GP: 44°20’59.86”N, 68°48’19.35”W). The GP sampling location for L. littorea was changed to a site in the outer estuary (Dyer Point) due to the species’ preference for rocky substrate, as opposed to finer sediments that predominate in the inner estuary. Individuals were weighed with and without shells. Foot muscle and hepatopancreas were dissected, based on previous protocols (English and Storey, 2003), and all samples were immediately placed in tubes with ice cold RNALater. All samples were transported to the University of Maine at Farmington (UMF) laboratories and stored at −80°C.

Fig. 1.

(A) Map of coastal Maine, USA with highlighted region enclosing Cape Rosier. Bar represents 16 km. (B) Detail map of highlighted region with Goose Pond, Dyer Point, Horseshoe Cove, and the Callahan Mine Superfund site (cross hatch pattern). Bar represents 1.6 km. The Horseshoe Cove sampling site for both species is indicated by (1), while L. littorea and H. diversicolor sampling sites within Goose Pond are indicated by (2 – Dyer Point) and (3 – inner estuary), respectively. Maps were generated in the ArcGIS Maine Basemap viewer from the State of Maine Office of GIS.

2.2. Sediment sampling and metal analyses

To confirm sediment metal levels, five sediment samples were collected from each location and sampling event. Surface sediment (~500 g each, approximately 6 cm deep) was removed nearby to animal sampling locations (< 0.1 km), and samples were randomly spaced from each other by approximately 3 m. Sediments were placed in plastic containers, transported to UMF laboratories, air-dried for 2 weeks, ground into a powder, and sieved through 0.25 mm mesh. Subsamples were then removed (50 g), placed into plastic containers, and levels of arsenic (As), cadmium (Cd), copper (Cu), manganese (Mn), lead (Pb), and zinc (Zn) were quantified using a calibrated, Alpha A4000 handheld X-Ray Fluorescence (XRF) Analyzer (Innov-X Systems, Inc., Woburn, MA) by trained staff at the Maine Department of Environmental Protection (Augusta, ME). Each sample was measured three times, and levels below the limit of detection (LOD) were considered zero for later analyses.

2.3. RNA extractions and cDNA synthesis

Hediste diversicolor individuals (n = 9/site) and L. littorea foot muscle and hepatopancreas (n = 10/tissue type from HC, 15 from GP) were removed from RNALater, blotted dry, and added to 500 μl of cold Tri Reagent (Sigma Aldrich, St. Louis, MO). Tissues were homogenized fully using 1.5 ml microcentrifuge tubes, disposable pestles, and a cordless motor (Kimble Chase, Rockwood, TN). Homogenized samples were then brought to 1.0 ml with Tri Reagent. RNA extractions were performed using standard phenol/chloroform procedures (Molecular Research Center, Cincinnati, OH) and total RNA quantity and quality were assessed using 2% agarose gel electrophoresis and an Epoch microplate reader with the Take3 Multi-Volume Plate (BioTek Instruments, Inc., Winooski, VT). All samples exhibited high RNA quality (260/280 ratio = 2.0 ± 0.1) and intact rRNA banding patterns. Total RNA (2.5 μg) was DNase-treated using the RQ1 RNase-free DNase kit (Promega Corp., Madison, WI), and cDNA was synthesized using 1.8 μg DNase-treated RNA, 2.5 μM oligo dT 20mer primer, RNaseOUT ribonuclease inhibitor (40 units), and Superscript III reverse transcriptase (200 units; Invitrogen, Carlsbad, CA).

2.4. Candidate gene identification and primer design

To characterize expression patterns in H. diversicolor, seven candidate stress response genes were chosen for later qPCR analyses, based on previous assessments of upregulation in this species (McQuillan et al. 2014). These genes included a copper transporting P-type ATPase (atp7a), copper chaperone protein for superoxide dismutase (ccs), an inducible metallothionein-like protein (mt or MTLP small in McQuillan et al., 2014), glutathione s-transferase mu class (gstm), glutathione s-transferase omega class (gsto), glutathione s-transferase theta class (gstt) and manganese superoxide dismutase (sod2). Previously identified primer sets for these genes were used (Table 1). Other selected genes that previously exhibited significant upregulation under metal exposure in this species (MTLP large and atox1; McQuillan et al. 2014) were not included due to overall unquantifiable expression levels (Mean Ct > 38.0) and possible primer sequence errors (identical to ccs; McQuillan et al., 2014, Table S1), respectively. Additionally, the high affinity copper uptake transporter gene (ctr1) was not included due to amplification of non-specific products in the present study. A commonly used reference gene, glyceraldehyde 3-phosphate dehydrogenase (gapdh), was also previously identified and included in the analyses (McQuillan et al., 2014) (Table 1).

Table 1.

Gene symbols, qPCR primer sequences, annealing temperatures (°C), product sizes (bp), and PCR efficiencies (%) for gapdh (reference gene) and 10 candidate genes. H. diversicolor primer sets were designed by McQuillan et al., 2014, while L. littorea primer sets were designed using available sequence fragments on NCBI. PCR efficiencies for L. littorea are divided into foot muscle (before slash) and hepatopancreas-validated assays (after slash).

Gene symbol	Primer sequence (5’−3’)	Annealing temperature (°C)	Product size (bp)	Efficiency (%)
H. diversicolor
atp7a	F - CTACGAGAAGCCACGAGTCC	56	251	102.1
	R - TCTCCAGGGACCACTTTCAG
ccs	F - AGCAGTTGGAGTCAGCAGGT	56	196	107.6
	R - TGCCAGCTCTCCGTATTTCT
gapdh	F - TGTATTCACGCGGTGTCG	52	186	99.7
	R - GGAACAGGTTGTGTCCTCTGA
gstm	F - CCAGCAACTGGGTTCCTACT	56	184	104.5
	R - TGACTGTAGTCCGGACCATTC
gsto	F - CATCGCAGATTGAGGATTCA	56	176	94.1
	R - TGTCCCTATGCCCAGAGAAC
gstt	F - ACCATGAAGTTGGCTTGGTC	56	186	89.5
	R - GCTCCAAGACACCCCTTACA
mt	F - CATTGCACTGGGAATGTTTG	56	243	113.0
	R - CATCACAGCATTGGATGGAC
sod2	F - CATAAGTTGCATGATGCTTGG	56	151	95.3
	R - ACTGTGGCAAGGCTGACAAG
L. littorea
gapdh	F - AACAGCTGACGTCTCAGTGG	60	87	95.2 / 98.0
	R - GCCTCCTTCATGGCCTTCTT
gstm	F - TGGCGGAGATAATGTGACGG	60	82	95.3 / 97.9
	R - CAAGCAGCCTGGGATCATCA
mt	F - CTGTAACTGCACGGACGACT	60	135	100.7 / 104.2
	R - GAACAGCTCTGACAGCGACA
sod2	F - GGGTGGAGAGGAACCAACTG	60	102	97.1 / 100.8
	R - ACGGCAACCGTCTTACCAAT

To confirm amplification of each intended target, primer sets were used in PCR and sequenced. Briefly, 5 μl of 1/10 diluted cDNA, 0.5 μM primers, 10 mM dNTPs, 1 X REDTaq PCR buffer, and 2.5 units of REDTaq DNA polymerase (Sigma Aldrich) were combined in 50 μl total volumes. PCR was performed using standard cycling conditions (94°C for 2 min, 35 cycles of 94°C for 30 sec, 52 or 56°C for 30 sec, and 72°C for 1 min, followed by final extension step of 72°C for 5 min), and PCR products were electrophoresed in 2% agarose gels. To purify amplified products, the ExoSAP-IT PCR Product Cleanup kit (Affymetrix, Santa Clara, CA) was used, and PCR products were submitted to the Mount Desert Island (MDI) Biological Laboratory for sequencing in both forward and reverse directions using the dideoxy chain termination method with standard techniques on an ABI 3130×l DNA sequencer (Applied Biosystems, Foster City, CA). Sequence chromatographs were manually trimmed for quality, assembled, and analyzed using the blastn and blastx NCBI databases, to verify target specificity. Since sequence data from McQuillan and colleagues (2014) were not yet publicly available (pers. comm., T.S. Galloway), all sequenced PCR products > 200 bp in size (mt and atp7a) were submitted to the NCBI GenBank nucleotide database (Accession numbers KX444139 and KX444140, respectively). Sequenced products < 200 bp are available in Supplementary File 1.

To compare stress response gene expression patterns between species, three candidate genes were selected for L. littorea analyses (Table 1). Primers were designed in Primer-BLAST (NCBI) for gapdh and an inducible mt sequence available on GenBank (KM892481 and AY034179) (English and Storey, 2003; Gorbushin and Borisova, 2015). Published sequences for glutathione s-transferases in other molluscs (Biomphalaria glabrata and Ruditapes philippinarum) were used to search an L. littorea transcriptome Sequence Read Archive (SRA) (SRX092192) for a fragment with significant similarity to a glutathione s-transferase mu class gene (gstm). Similarly, a B. glabrata manganese superoxide dismutase sequence (NM_001311263) was used to identify an L. littorea sod2 fragment (see Supplementary File 1 for trace IDs). Primers were designed for all identified fragments, and amplification of intended products was confirmed using standard PCR and agarose gel electrophoresis approaches described previously.

2.5. qPCR analyses

Relative quantification SYBR green qPCR assays were conducted using a StepOne Plus Real-Time PCR system (Thermo Fisher Scientific, Waltham, MA) with 0.5–0.8 μM primer concentrations, depending on the assay. Assays were performed using the FAST SYBR Green master mix (Thermo Fisher Scientific) and run under previously published parameters for H. diversicolor (McQuillan et al., 2014; 95°C for 10 min, 40 cycles of 95°C for 15 sec, 52 or 56°C for 30 sec, and 72°C for 1 min) or standard cycling conditions for L. littorea (95°C for 10 min, 40 cycles of 95°C for 15 sec, 60°C for 1 min) with dissociation curve analysis. Samples were assayed in duplicate (1/15, 1/50, 1/100 diluted, depending on assay), and relative standard curves (1/5 – 1/1280 diluted) were made from pooled cDNAs equally represented across sites. Standard curve points were run in triplicate, and optimized curves consisted of four to six points with approximately 90–110% PCR efficiency (Table 1). All assays exhibited a single peak in dissociation curve analysis, and standard QPCR negative controls (no template and no reverse transcriptase) exhibited no contamination in all assays.

2.6. Statistical analyses

Quantitative PCR data were analyzed using the Pfaffl method of relative quantification (Pfaffl, 2001). Expression values were calibrated to the Horseshoe Cove (HC) group mean (set to 1.0) and normalized to gapdh levels, to compensate for differences in cDNA synthesis efficiency across samples. Relative expression values were log-transformed to meet assumptions of normality. Expression data, animal weights, and metal concentrations (ppm) were represented as the mean ± standard error and analyzed using two sample t-tests in SYSTAT13 (Systat Software, Inc. San Jose, CA). A p value < 0.05 was considered statistically significant for all analyses.

To identify possible relationships among genes, log transformed relative expression values within each species were used in least squares regression analyses in SYSTAT13. Coefficients of determination (r²) were assessed in pairwise comparisons of six genes (ccs, gstm, gsto, gstt, mt, and sod2) in H. diversicolor, using all qPCR samples (n = 18). Atp7a was not included due to missing data in three individuals (see section 3). For L. littorea analyses, a subset of individuals assessed at all genes in both foot muscle and hepatopancreas tissues (HC, n = 7; GP, n = 10) were used. To better visualize individual differences between sites, these data were also used in principal component analyses (PCA) in SYSTAT 13. Component loading plots in PCA were generated to further assess gene similarity.

3. Results

Hediste diversicolor individuals weighed significantly less at Goose Pond (mean = 10.8 ± 2.8 mg) compared to those collected at Horseshoe Cove (46.0 ± 5.2 mg) (p < 0.0001, Fig. 2A). L. littorea, in contrast, weighed similarly between sites (HC = 0.9 ± 0.1 g, GP = 1.0 ± 0.1 g) without shells (Fig. 2B). L. littorea weights were also not significantly different with shells included (4.2 ± 0.4 and 4.6 ± 0.4 g, respectively) (data not shown).

Fig. 2.

(A) H. diversicolor weight (mg) and (B) L. littorea body weight without shell (g) at Horseshoe Cove (HC) and Goose Pond (GP) sites. Each bar represents the mean ± standard error. Asterisks (**) indicate p < 0.0001.

Sediment metal content during H. diversicolor sampling in 2015 was significantly higher at the Goose Pond inner estuarine site, compared to Horseshoe Cove (Fig. 3A). As and Cu levels were below limits of detection at Horseshoe Cove and elevated at Goose Pond (49.7 ± 9.7 and 1441.7 ± 40.8 ppm, respectively). Mn, Pb, and Zn were overall significantly lower at the reference site (334.5 ± 5.3, 20.9 ± 1.9, and 62.3 ± 1.0 ppm, respectively) and higher at Goose Pond (1070.1 ± 15.1, 359.3 ± 6.5, and 4609.5 ± 124.1 ppm, respectively) (p < 0.0001). Cd levels were below limits of detection at both sites (data not shown).

Fig. 3.

Sediment metal concentrations (parts per million, ppm) of arsenic (As), copper (Cu), manganese (Mn), lead (Pb), and zinc (Zn) at Horseshoe Cove (white with diagonal lines) and Goose Pond (gray) sites during (A) 2015 H. diversicolor and (B) 2016 L. littorea sampling events. Each bar represents the mean ± standard error. Asterisks (* and **) indicate p < 0.05 and p < 0.0001, respectively. ND = not detected.

Sediment metal content during L. littorea sampling in 2016 was significantly higher at the Goose Pond Dyer Point site, compared to Horseshoe Cove (Fig. 3B). Goose Pond exhibited relatively higher levels of As, Pb, and Zn (9.1 ± 2.5, 159.1 ± 34.6, and 959.8 ± 160.7 ppm, respectively) but was overall characterized by less metal content than at the inner estuarine site sampled the previous year (Fig. 3). These elements, however, were still significantly higher than those measured at the reference site (p < 0.05) (2.2 ± 1.4, 15.9 ± 2.8, and 262.9 ± 159.4 ppm, respectively). In contrast, relatively higher Cu and Zn variability was detected at Horseshoe Cove in 2016, where one sediment sample was elevated (1600.7 and 900 ppm, respectively) while the other four were low but somewhat higher than the previous year (mean = 97.6 ± 20.7 and 103.7 ± 9.5 ppm, respectively). As a result, copper levels were not significantly different between sites (Fig. 3B). Mn levels at the Goose Pond Dyer Point location were overall highly similar to the reference site in both years (355.9 ± 24.9 ppm) and not significantly different. Similar to 2015 sampling, Cd levels were also below limits of detection at both sites (data not shown).

Stress-associated gene expression patterns were overall highly stable in H. diversicolor at all candidate genes (p > 0.05) (Fig. 4), while some differences between sites were detected in L. littorea foot muscle (Fig. 5A) and hepatopancreas (Fig. 5B). In all tissues, gapdh expression was stable across sites and suitable as a reference gene (p > 0.10, data not shown). All samples exhibited quantifiable expression, except for atp7a in H. diversicolor, where one and two individuals at Horseshoe Cove and Goose Pond, respectively, exhibited no detectable expression. L. littorea gstm expression was significantly elevated in foot muscle at Goose Pond (2-fold), while no difference was detected in hepatopancreas tissue. In contrast, mt expression was significantly elevated in both foot muscle and hepatopancreas at Goose Pond, compared to the reference site (3-fold and 1.5-fold, respectively). Expression of sod2 was also somewhat elevated in Goose Pond foot muscle (~3-fold) but also highly variable and marginally non-significant (p = 0.052). No difference in sod2 expression (p > 0.10) was detected in the hepatopancreas.

Fig. 4.

Relative mRNA expression of seven candidate genes (atp7a, ccs, gstm, gsto, gstt, mt, and sod2), normalized to gapdh, for H. diversicolor sampled at Horseshoe Cove (HC) and Goose Pond (GP) sites. Each bar represents the mean ± standard error.

Fig. 5.

Relative mRNA expression of three candidate genes (gstm, mt, and sod2), normalized to gapdh, for L. littorea (A) foot muscle and (B) hepatopancreas sampled at Horseshoe Cove and Goose Pond (GP) sites. Each bar represents the mean ± standard error. Asterisks (*) indicate p < 0.05.

Some expression patterns were correlated among genes in both species. In H. diversicolor, genes exhibited some evidence of clustering in three groups: 1) ccs, gsto, and sod2, 2) gstm and gstt, and 3) mt (Table 2). For instance, significant correlations were detected between ccs and both gsto and sod2 (r² = 0.448 and 0.693, respectively), while gstm expression was not correlated to these genes but did exhibit similarity to gstt (0.568). Gsto was overall similar to the ccs gene, with a significant correlation to sod2 (0.448). Mt expression was largely correlated to genes from both groups, including ccs and gsto, as well as both gstm and gstt (0.282, 0.328, 0.390, and 0.331, respectively). In contrast, L. littorea genes were correlated only within tissues (Table 3). All three genes in foot muscle exhibited similar patterns to each other (0.323–0.630), and similar correlations were detected within the hepatopancreas (0.283–0.367).

Table 2.

Matrix of coefficients of determination (r²) from regression analyses for pairwise gene expression comparisons in H. diversicolor. Data for atp7a were not included due to non-quantifiable expression in three individuals. Bold and asterisks (*, **, or ***) indicate significance at < 0.05, < 0.01, or < 0.0001 levels, respectively.

	ccs	gstm	gsto	gstt	mt
gstm	0.110
gsto	0.448**	0.162
gstt	0.086	0.568***	0.197
mt	0.282*	0.390**	0.328*	0.331*
sod2	0.693***	0.135	0.448**	0.170	0.194

Table 3.

Matrix of coefficients of determination (r²) from regression analyses for pairwise gene expression comparisons in L. littorea foot muscle (FT) and hepatopancreas (HP) tissues. Bold and asterisks (*, **, or ***) indicate significance at < 0.05, < 0.01, or < 0.0001 levels, respectively.

	FT gstm	FT mt	FT sod2	HP gstm	HP mt
FT mt	0.323*
FT sod2	0.530**	0.630***
HP gstm	0.028	0.022	0.006
HP mt	0.067	0.106	0.100	0.283*
HP sod2	0.064	0.096	0.215	0.367*	0.351*

Genes also clustered similarly in PCA, with loading plots for H. diversicolor exhibiting three groups: 1) ccs, gsto, and sod2, 2) gstt and gstm, and 3) mt (Fig. 6A). In L. littorea, loading plots were also similar to pairwise correlations, with genes grouped into either foot muscle or hepatopancreas patterns and little similarity across tissues (Fig. 6C). When these genes were collectively assessed in PCA, very little clustering was evident between HC and GP sites in either species (Fig. 6B, D). Only two L. littorea GP individuals that exhibited consistently elevated foot muscle expression in all three genes in qPCR also clustered separately from HC individuals (dotted circle, Fig. 6D).

Fig. 6.

Component loading plots and principal component analyses (PCA) for H. diversicolor (A and B, respectively) and L. littorea (C and D, respectively) genes. Data for atp7a were not included due to non-quantifiable expression in three individuals. Horseshoe Cove (HC) and Goose Pond (GP) individuals are represented by black circles and gray diamonds, respectively. Black triangles represent each gene, and FT and HP indicate foot muscle and hepatopancreas tissues, respectively. The dotted circle encloses two L. littorea GP individuals with consistent, elevated expression in all three foot muscle gene assays.

4. Discussion

Gene expression changes in Hediste divericolor and Littorina littorea at the Callahan Mine site were overall absent or weak, despite sediment metal levels being significantly elevated and similar to previously documented patterns (MACTEC, 2009; Broadley et al., 2013; EPA, 2016). Indeed, Cu, Pb, and Zn levels at both the Goose Pond inner site and Dyer Point were somewhat higher than these respective locations previously evaluated by Broadley and colleagues (2013). This was likely due to each sampling location being relatively closer to the tailings pile. For example, the inner estuary site exhibited higher Zn levels (~4600 ppm) than previously found (~2000) but was approximately 100–150 m closer to the highest contamination levels (~6000, Broadley et al., 2013). These data reinforce that metal contamination at this site is enriched on a fine spatial scale (Ayuso et al., 2013, Broadley et al., 2013). In addition, while sediments from outer locations such as Dyer Point were previously found to be similar to Horseshoe Cove (Broadley et al. 2013), that was not the case in the present study. Instead, L. littorea at a location < 100 m southeast were exposed to elevated As, Pd, and Zn, likely due to being closer to an historical mining effluent spot (MACTEC, 2009). In addition, Horseshoe Cove in 2016 exhibited high copper variability, unlike prior assessments. The source of this variation is unknown, but estuarine sediments often act as sinks for heavy metals from runoff and human activities (Lytle and Lytle, 2001), and we cannot exclude a point contamination event in 2016 that was not present in 2015. Both years overall though exhibited significant enrichment of most heavy metals in Goose Pond. However, this was largely not associated with dramatic transcript changes related to metal homeostasis, detoxification, and oxidative stress.

Both H. diversicolor and L. littorea are largely metal-tolerant (De Wolf et al., 2000; 2001; Daka and Hawkins, 2004), as both were found in Goose Pond despite the sediments being known to induce mortality in other polychaete worms, amphipods, and sea urchins (MACTEC, 2009; Broadley et al., 2013). Expression changes in stress-related genes, however, have previously been documented in both species. In H. diversicolor, elevated metallothionein, copper transporter, and chaperone gene expression was associated with a contaminated English estuary (Restronguet Creek, Cornwall, UK) (McQuillan et al., 2014). This site, however, was characterized by higher Cu levels than those in the present study (Bryan and Gibbs, 1983), and other individuals at a lesser contaminated site (Mylor Bridge) did not exhibit similar upregulation (McQuillan et al., 2014). Instead, those individuals were characterized by high expression of gst genes (McQuillan et al., 2014), which was also documented in other polychaetes at relatively lower metal exposures (Won et al., 2012). However, gst upregulation was not evident in this study and responsiveness may vary, as the three isoforms assessed (gstm, gsto, and gstt) exhibit somewhat inconsistent patterns across invertebrates (Lee et al., 2008; Won et al., 2012; Zhang et al., 2012; McQuillan et al., 2014).

The lack of expression changes in H. diversicolor at Goose Pond may reflect the importance of other strategies in the pollutant-induced stress syndrome, such as metal cation sequestration in body tissues. For instance, Restronguet Creek individuals are genetically adapted to high toxicity and exhibit significant increases in insoluble metal inclusions within body tissues, as well as a mucus response that may bind dissolved metals and reduce uptake (Mouneyrac et al., 2003; McQuillan et al., 2014). There is likely still a metabolic cost to metal tolerance, though, as individuals also exhibited reduced fecundity, energy reserves, and weight (Pook et al., 2009). Similar mechanisms may be present in other polychaetes such as Alitta virens, which exhibit low tissue metal concentrations but are characterized by smaller oocytes and reduced developmental success (Watson et al., 2013, 2018). Since Goose Pond individuals in this study exhibited reduced weight, similar compensatory mechanisms may be ongoing, but this will require further investigations using histological and physiological assessments.

In contrast to ragworm, L. littorea in this study were likely exposed to overall less metal contamination and weighed similarly between sites, but exhibited some evidence of upregulation in gstm, mt, and sod2. Although a wide range of environmental stressors can increase mt expression in both foot muscle and hepatopancreas (English and Storey, 2003), elevated levels of all three genes in this study are likely due to metal toxicity, as gstm and sod2 upregulation is also associated with metal exposure in other molluscs (Kim et al., 2007; Zhang et al., 2012). In addition, heavy metals are known to induce synthesis of MT-like proteins in littorine species (Bebianno and Langston, 1995; Park et al., 2002), but protein levels were not measured in the present study and require further investigation.

Overall, expression changes in L. littorea were weak and little distinction was evident between Horseshoe Cove and Goose Pond individuals when all genes were assessed using a multivariate approach. Since L. littorea prefer rocky substrate and are not buried in marine sediments, their exposure may depend on microhabitat characteristics and fine-scale spatial differences in metal contamination. As a result, apparent metal-induced stress in this species likely vary at the individual level. Indeed, only two Goose Pond individuals clustered separately from other samples, and these were characterized by consistent, high relative expression of all three foot muscle genes but low transcript levels in hepatopancreas. This tissue-specific expression pattern was possibly due to a combination of two factors: 1) the foot muscle is more exposed to contaminated water and sediments, and 2) the hepatopancreas may differentially accumulate metals. For instance, Cd readily accumulates in the hepatopancreas of another gastropod (Helix pomatia), while Cu was more broadly deposited across tissues (Chabicovsky et al., 2003). Since Goose Pond was characterized by low Cd and elevated Cu, a hepatopancreas-specific signal may have been minimized. In addition, sod2 patterns in other molluscs exhibit relatively more shifts in foot muscle than hepatopancreas (Kim et al., 2007), while gstm is not strongly responsive in the hepatopancreas over longer term exposures (Zhang et al., 2012). Although more research is needed to further evaluate tissue-specific differences, these patterns suggest that some genes related to oxidative stress and metal toxicity may be highly responsive in foot muscle.

Stress-associated genes in both species also exhibited some novel expression patterns, which may be useful in future biomarker development and advance our understanding of these molecular mechanisms. For instance, in H. diversicolor, not all gst genes exhibited similar patterns across individuals, with gsto divergent from the both gstt and gstm isoforms. Invertebrates express many glutathione-s-transferases, which can vary greatly in tissue-specific expression, detoxification functions, and responsiveness to metals (Lee et al., 2008; Zhang et al., 2012). In another polychaete (Alitta succinea), Cu exposure induces gsto upregulation, while gstt remains largely stable (Rhee et al., 2007). Although gsto in this study did not exhibit similar upregulation, the overall expression pattern across individuals was correlated to other known Cu responsive genes (ccs, mt, and sod2; Rhee et al., 2011; McQuillan et al., 2014). As such, this gene may warrant further research as a Cu responsive, glutathione s-transferase biomarker in polychaetes, while gstt or gstm isoforms may regulate other stress-associated pathways. This pattern differs in L. littorea, where a gstm isoform was both correlated to other Cu responsive genes and exhibited significant upregulation in metal-exposed individuals. In other studies, gstm transcript changes also occur during Cu exposure in molluscs (Zhang et al., 2012), but largely not in copepods (Lee et al., 2008), which demonstrates that different taxa likely utilize divergent molecular mechanisms. This is also evident in other physiological comparisons, where MT protein synthesis positively correlates with metal exposure in mussels and clams but not in polychaetes under longer term exposures (Ng and Wang, 2004; Ng et al., 2007; 2008; Won et al., 2012). Physiological assessments in these invertebrates should therefore not only rely on a small suite of gene expression biomarkers, but also use protein, enzymatic activity, and histological examinations to more fully assess the metal-induced stress response. Overall though, broad differences in gene expression patterns were evident between metal-exposed H. diversicolor and L. littorina, and taxa-specific transcript changes may be useful as a complementary approach to better understand these mechanisms in diverse invertebrate species.

5. Conclusion

Two marine invertebrate species, Hediste diversicolor and Littorina littorea, were characterized by different gene expression patterns in response to metal contamination at the Callahan Mine site. H. diversicolor were exposed to high metal levels in inner estuary sediments but did not exhibit transcript changes, possibly due to other physiological strategies associated with metal cation sequestration that require further investigation. In contrast, L. littorea exhibited significant upregulation in several transcripts, but changes were overall weak, possibly due to fine scale differences in sediment exposure among individuals. Several relationships among genes in both species were also identified, including divergent patterns in glutathione-s-transferase isoforms in H. diversicolor, and tissue-specific patterns in L. littorea. However, further physiological assessments will be needed in both species, including protein assays and histological examinations, to more fully characterize these stress responses associated with heavy metal toxicity.

Highlights.

• Stress-related gene expression was assessed in invertebrates at a metal-contaminated estuary.

• Hediste diversicolor exhibited no expression changes but exhibited reduced weight.

• Littorina littorea exhibited elevated expression of genes related to metal toxicity.

• Glutathione-s-transferase isoforms had divergent expression patterns in H. diversicolor.

• L littorea foot muscle and hepatopancreas expression differed.