Targeted metabolomics characterizes metabolite occurrence and variability in stable freshwater mussel populations

We characterized the occurrence and variability of haemolymph metabolites in two mussel species from three Indiana (USA) streams. Metabolite variability was most influenced by species, followed by site and sex. We provide reference values for metabolites from a range of metabolic pathways, which can be further developed as health biomarkers.

Abstract

Freshwater mussels (order Unionida) play a key role in freshwater systems as ecosystem engineers and indicators of aquatic ecosystem health. The fauna is globally imperilled due to a diversity of suspected factors; however, causes for many population declines and mortality events remain unconfirmed due partly to limited health assessment tools. Mussel-monitoring activities often rely on population-level measurements, such as abundance and age structure, which reflect delayed responses to environmental conditions. Measures of organismal health would enable preemptive detection of declining condition before population-level effects manifest. Metabolomic analysis can identify shifts in biochemical pathways in response to stressors and changing environmental conditions; however, interpretation of the results requires information on inherent variability of metabolite concentrations in mussel populations. We targeted metabolites in the haemolymph of two common mussels, Lampsilis cardium and Lampsilis siliquoidea, from three Indiana streams (USA) using ultra-high-performance liquid chromatography combined with quadrupole time-of-flight mass spectroscopy. The influence of species, stream and sex on metabolite variability was examined with distance-based redundancy analysis. Metabolite variability was most influenced by species, followed by site and sex. Inter- and intraspecies metabolite variability among sexes was less distinct than differences among locations. We further categorized metabolites by occurrence and variability in mussel populations. Metabolites with high occurrence (Categories 1 and 2) included those indicative of energy status (catabolism versus anabolism; arginine, proline, carnitine, nicotinic acid, pantothenic acid), oxidative stress (proline, glutamine, glutamate) and protein metabolism (thymidine, cytidine, inosine). Metabolites with lower occurrence (Category 3) are constituents of assorted metabolic pathways and can be important biomarkers with additional temporal sampling to characterize their variability. These data provide a reference for future temporal (before/after) monitoring and for studies of stressor–metabolite linkages in freshwater mussels.

Introduction

Freshwater mussels (order Unionida) provide a range of ecosystem services in aquatic habitats from water filtration to substrate stabilization to nutrient conversion (Vaughn and Hakenkamp, 2001; Haag and Williams, 2014; Vaughn, 2018). Their complex life cycle includes an infaunal, sessile adult stage, a parasitic larval stage on a vertebrate host (usually a fish), followed by transformation to a free-living benthic juvenile (McMahon and Bogan, 2001). Thus, unionids are also sentinels and umbrellas of overall aquatic system health (Geist, 2010). The fauna is in jeopardy across the globe, with 200 species on the International Union for Conservation of Nature's (IUCN) Red List, 189 of which occur in North America (Lydeard et al., 2004; Strayer et al., 2004; RéGnier et al., 2009). Suspected factors for the decline of mussel populations range from habitat degradation to contaminants and invasive species (Downing et al., 2010). However, numerous enigmatic declines and mass mussel mortality events have occurred since the 1970s without identification of causal agents due in part to the lack of health assessment tools (Neves, 1987; Starliper et al., 2011; Haag, 2019; Waller and Cope, 2019). Efforts to monitor the status of freshwater mussels generally rely on population-level assessment of abundance, species diversity and age structure. These measures can identify long-term trends in mussel populations after repeated sampling of the population; however, occurrence does not guarantee sustainability (i.e. growth and reproduction) of mussels at a location (Gray and Kreeger, 2014). During routine mussel monitoring or investigation of a mortality event, a mussel may appear outwardly healthy when in the initial stages of disease. More sensitive, real-time measures of organismal health are needed for preemptive identification of declining mussel condition before population-level effects manifest.

Metabolomic analysis is an emerging tool used to evaluate biologically relevant molecules in organismal tissues and fluids. This approach can identify metabolites linked to biochemical processes and provide insight into pathways disrupted by environmental stressors such as contaminants, starvation and pathogens (Viant, 2007; Hines et al., 2010; Lankadurai et al., 2013; Pomfret et al., 2020; Izral et al., 2021). Environmental metabolomics can be used as a real-time tool for population monitoring that may also determine mechanisms of change or identify the specific stressor (Pomfret et al., 2020). To date, a handful of studies have applied the technology to assess the health status of freshwater mussels. For example, Roznere et al. (2014) detected shifts in a subset of haemolymph metabolites in Amblema plicata after captivity and/or food limitation. Of note, changes to carbohydrate, energy, lipid, amino acid and nucleotide metabolism brought on by captivity were similar to (but less severe than) changes induced by captivity combined with fasting. In another study, Roznere et al. (2017) found that relocation of A. plicata altered amino acid and nucleotide metabolism while glucose and lipid metabolism remained relatively constant. More recently, the haemolymph metabolome of Ortmanniana pectorosa from a mass mortality site was characterized to profile moribund mussels and identify potential biomarkers of disease (Putnam et al., accepted). Thirteen metabolites were associated with field diagnosis of mussels, including gamma-linolenic acid in moribund mussels, and adenine and inosine in control mussels. Pathway analysis found that catabolic processes were more affected in mussels diagnosed as “sick.” Although limited in number and scope, these studies show the utility of metabolomic analysis for characterizing mussel health status.

Metabolite composition and abundance vary among tissue types (Hurley-Sanders et al., 2015; Nguyen et al., 2019; Izral et al., 2021), and the tissue selected depends on study objectives and metabolic pathways of interest. Haemolymph sampling has advantages and disadvantages for metabolomic studies. Generally, the compounds in haemolymph are low molecular weight and less concentrated than in tissue. Depending on the analytical platform, haemolymph metabolites may not be detected (e.g. proton nuclear magnetic resonance; Hurley-Sanders et al., 2015). On the other hand, liquid chromatography (LC) and gas chromatography-mass spectometry (GC-MS) platforms have successfully shown haemolymph metabolite shifts in response to various stressors such as captivity (Roznere et al., 2014; Roznere et al., 2017) and pathogen challenge (Nguyen et al., 2018). Additionally, haemolymph collection is a minimally invasive sampling technique and can be used for repeated sampling of an individual (Gustafson et al., 2005a; Fritts et al., 2015b), a critical consideration when monitoring threatened and endangered species.

An important first step in application of metabolomics for biomonitoring with freshwater mussels is to characterize the metabolomics of stable populations. Identifying factors that influence metabolite variability will inform an optimal sampling strategy and improve interpretation of results. For example, sex (Ji et al., 2013; Ellis et al., 2014; Nguyen et al., 2018; Dumas et al., 2020a, 2020b), life stage (Wu et al., 2017), species (Payton et al., 2016; Steinagel et al., 2018; Haider et al., 2020) and collection site (Gustafson et al., 2005b; Roznere et al., 2017; Hornbach et al., 2021) may all affect metabolomic differences in bivalves. Expanding the database on haemolymph metabolite concentrations can provide a starting point for laboratory-based studies to identify biochemical pathways of specific disease or stressors in mussels and as a reference for real-time monitoring of wild mussel populations.

Glycogen is the primary energy storage molecule in unionid mussels and is commonly used as a measure of mussel condition in response to environmental stressors (Haag et al., 1993; Patterson et al., 1999; Fritts et al., 2015a, 2015b). Several studies have measured tissue glycogen concurrently with analysis of haemolymph constituents (Gustafson et al., 2005b; Fritts et al., 2015a, 2015b). However, data are lacking to compare variability between glycogen and metabolomic data for mussel health assessment.

The objectives of the present study were 2-fold. First, a suite of haemolymph metabolites in two common freshwater mussel species, Lampsilis cardium and Lampsilis siliquoidea, were characterized to establish their “normal” operating range (Viant, 2007) in presumed “stable” populations. Metabolite and glycogen concentrations were analysed by location, species and sex to identify sources of variability within and between populations. Secondly, metabolites were categorized by occurrence and relative variability to evaluate their utility as biomarkers of mussel health. The mussel species used in this study are widely distributed and relatively common (Cummings and Cordeiro, 2012; Bogan et al., 2017); therefore, the resulting data can serve as a reference across a broad geographic range.

Methods

Study locations

Mussels were collected between June 18 and July 9, 2019, from three streams in West Central Indiana (Fig. 1) with relatively stable mussel populations. The drainage generally flows westerly and enters the Wabash River in Tippecanoe County, Indiana, with a total watershed size of 805 sq mi (2085 km²). Land use within the drainage is mostly cultivated crops (Supplementary Material, Table S1). Wildcat Creek samples were collected on June 18, 2019, in Carroll County, Indiana, about 48 km downstream from the city of Kokomo, Howard County, Indiana (Fig. 1). The total drainage at this location is ~966 km². The Mud Creek location is within the watershed of Wildcat Creek but is upstream from the city of Kokomo, Indiana. Mud Creek samples were collected July 9, 2019, in Tipton County, Indiana, about 32 km upstream from Kokomo (Fig. 1). The total drainage area of Mud Creek is ~193 km². Kilmore Creek is a tributary of the South Fork Wildcat Creek. Samples were collected on July 2, 2019, in Clinton County, Indiana, about 1.6 km upstream from its confluence with the South Fork Wildcat Creek. The total drainage area is ~199 km², and no cities or towns are in the watershed. Surface water analysis for a suite of constituents was included to verify that water quality criteria were within established limits and to establish baseline criteria for mussel habitat in the event of future adverse releases to the study streams (Supplementary Material, Table S2).

Figure 1.

Locations of mussel collection from three streams in Indiana (Kilmore Creek 40°19′19.10″N 86°34′21.20″W, Mud Creek 40°24′23.01″N 85°55′33.71″W, Wildcat Creek 40°31′10.02″N 86°21′23.70″W).

Sample collection

At each location, 20 mussels of each species, L. cardium and L. siliquoidea, were collected at first encounter using visual or tactile searches (either by hand grabbing or toe picking) depending on water clarity. Mussel shell length (maximum anterior–posterior axis) was measured in the field with a calliper. Immediately after collection, the valves were opened with reversible pliers, the anterior adductor muscle was swabbed with 70% isopropyl alcohol, and ~3 ml of haemolymph was withdrawn using a 6-ml syringe and a 22-gauge needle. Haemolymph was transferred to sterile tubes, immediately placed on dry ice, shipped overnight to the U. S. Geological Survey, Upper Midwest Environmental Sciences Center, La Crosse, Wisconsin, and stored at −80°C until analysis. A cross-section of the visceral mass of each mussel was collected and immersed in 10% neutral-buffered formalin for histological examination.

For glycogen analysis, a section of foot tissue, approximately 7-mm in diameter × 25-mm in length, was excised from each mussel using a scalpel and forceps. Foot samples were transferred to individual sterile tubes, placed on dry ice, shipped overnight to the Department of Animal Science at Texas A&M University, College Station, Texas, and stored at −80°C until analysis.

Water quality sampling and analysis

Duplicate water samples were collected from two locations within each of the three streams between June 19 and July 9, 2019 (Supplementary Material, Table S2). Samples were analysed for 19 metals, total hardness, chloride, sulphate, alkalinity, total solids, total dissolved solids, total suspended solids, chemical oxygen demand, ammonia, total Kjeldahl nitrogen, nitrogen-nitrate + nitrite, phosphorus, cyanide and total organic carbon. Water chemistry data were not included in the data analysis.

Histology procedures

Preserved tissues were trimmed appropriately and processed routinely for histology. Sections were stained with haematoxylin and eosin and examined under a compound microscope (×40 to ×400) to determine sex and reproductive state. Sex was determined based on presence of ovarian or spermatogenic tissue. Hermaphroditic individuals were categorized as being “male–female hermaphrodites” (H-MF), in which ovarian and spermatogenic tissues were roughly equal; “male hermaphrodites” (H-M), in which spermatogenic elements comprised the majority of the visible gonad; and “female hermaphrodites” (H-F), in which ovarian tissue comprised the majority of the visible gonad. Individuals were considered hermaphrodites even if only rare elements of the minority gonad type were present (e.g., 90% of the gonad was female, 10% of the gonad was male). Female mussels were determined to be gravid based on presence of glochidia within gill chambers.

Metabolomic analysis

A method using ultra-high-performance liquid chromatography combined with quadrupole time-of-flight mass spectroscopy (UHPLC–QTOF-MS) was developed to detect metabolomic signals in mussel haemolymph. The UHPLC-QTOF-MS analysis was performed on an Agilent 1290 Quaternary Pump (G7104A), 1290 Multi Column Temperature control (G7116B), 1290 Multisampler (G7167B), 1290 Variable wavelength detector (G1314E), and G6460A QQQ LC-/MS. The column was a Phenomenex Kinetex 1.7 µm EVO C18. Mobile phase A was 9:1 H₂O/acetonitrile + 5 mM ammonium acetate + 0.1% formic acid, and mobile phase B was 9:1 isopropanol/acetonitrile + 5 mM ammonium acetate + 0.1% formic acid. All solvents were purchased from Sigma-Aldrich (St. Louis, Missouri). The method run time was 12 min, with a two mobile phase gradient that started at 100% mobile phase A, then transitioned to 50% A and 50% mobile phase B at 3 min; the gradient changed to 100% mobile phase B at 6.5 min, and was held at 100% mobile phase B until 10 min; at 12 min the mobile phase returned to 100% mobile phase A. The samples were kept at 4 °C, and the separation took place at 45 °C. Each injection used 6 µL of sample. Two injections were completed for each sample, one in the positive and one in the negative mass spectrometry mode. Mass spectrometer settings were: 250 °C for source gas, 325 °C for sheath gas, 120 V for the fragmentor, and 1000 V for nozzle voltage. Mass discrimination range was between 60 and 1700 mass to charge (m/z) ratios. The targeted analysis compared the detected signal mass, spectrum, and retention time with a database of known standards, allowing us to identify specific metabolites. Previously frozen haemolymph samples were thawed, and a 100-μl subsample was diluted with 200-μl of cold methanol containing CUDA (CAS # 479413–68-8, 12-[[(cyclohexylamino)carbonyl]amino]-dodecanoic acid; Cayman Chemical, Ann Arbor, Michigan, US) as an internal standard. The samples were vortexed for 30 s and centrifuged at 18 787g for 7 min at 4°C. The supernatant was pipetted out from the centrifuge tubes, and the pellet was discarded. Samples were further processed in a CentriVap micro IR Vacuum Concentrator (Labconco, Kansas City, Missouri) at 1700 rpm and 35°C to evaporate the methanol and then temperature was increased to 47.5°C to evaporate the residual methanol/water. The samples were reconstituted in 125 μl of cold methanol, vortexed for 30 sec, and centrifuged at 18 787g for 7 min at 4°C. Finally, a 100-μl volume of supernatant was used for analysis.

A library of known (standard) metabolites (Mass Spectrometry Metabolite Library, Sigma-Aldrich SKU: MSMLS-1EA) was analysed to determine retention time, signal intensity, m/z, ionization mode, and chemical adduct. These data were used to calculate the reported concentration (μM) of the standard metabolites in a haemolymph sample, based on a single-point calibration.

Glycogen analysis

Foot tissue samples (200–1200 mg, depending on sample size) were extracted with 5 ml of 70% perchloric acid to precipitate the protein, followed by blending using a Polytron homogenizer (Kinematica, Switzerland) at medium setting. The homogenate was then neutralized to pH 7.0–8.0 by adding 2.5 ml of 1 M KHCO₃, and the neutralized samples were centrifuged for 15 min at 3000g. The resulting supernatant was removed for glucose, glucose-6 phosphate (G-6-P) and glycogen analysis. Glucose and G-6-P were measured using assay systems described by Bergmeyer (1974). Briefly, Tris buffer system containing 0.9 mM of nicotinamide adenine dinucleotide phosphate (NADP+) and 1 mM of adenosine triphosphate (ATP) was added to each cuvette (total volume = 1.0 ml) along with 0.2 ml of extract. Glucose-6-phosphate dehydrogenase was added to each cuvette catalysing G-6-P to 6-phosphogluconate, and the change in absorbance was measured using a Beckman DU-7400 Spectrophotometer (Palo Alto, California) set at 339 nm. In the same cuvette, free glucose was converted to G-6-P by the addition of hexokinase (HK), and the change in absorbance at 339 nm was measured. The same TRIS buffer system (Bergmeyer, 1974) was used to quantify glycogen, with modification (Rhoades et al., 2005). In digesting the homogenate, 0.3 ml of extract was combined with 2.0 ml of amyloglucosidase solution (10 mg/ml). After the samples were placed in a 40°C water bath for 2 h, 2.0 ml of 70% perchloric acid was added to stop the reaction. Subsequently, 0.2 ml of digested sample was added to a cuvette along with 1.0 ml of ATP/NADP⁺/glucose-6-phosphate dehydrogenase (G6PDH) buffer (total volume = 1.0 ml). Background absorbance was recorded, HK was added to each cuvette and the change in absorbance was recorded at 339 nm.

Data analysis

The Profinder software (version B.06.00; Agilent Technologies, Santa Clara, California) was used to integrate and align the metabolomic signals across all the injection files. Profinder software was used to compare signals in the haemolymph samples to the standard library. A constant of 0.001 was added to each metabolite concentration to adjust for values of zero before values were log transformed (log base 2). After log scaling, data were Pareto scaled in which the square root of the standard deviation is used as the scaling factor. Targeted metabolomic data that were grouped by species were scaled and normalized. Principal component analysis (PCA), fold change (FC) analysis and a t-test were then used to look at metabolites that differed between the species. Significance was defined as |FC| > 2 and P < 0.001.

Distance-based redundancy analysis (db-RDA) was used to quantify similarities and differences between mussels based on metabolite levels. This technique is a constrained ordination method that fits multiple linear regressions on corresponding sets of response and explanatory variables (Legendre and Legendre, 2012). The response variables were metabolite levels, and the explanatory variables were the demographic variables of interest (i.e. species, sex, location). Individuals with undetermined sex were omitted from the analysis (n = 2). For each species, hermaphrodites were classified by histological assessment into three categories (H-F, H-F and H-MF; Table 1). However, due to the small sample size for each category, hermaphrodites were grouped together for statistical analysis. Once this regression was fitted, the fitted and residual response values were subject to principal coordinate analysis (PCoA), which calculates a distance or dissimilarity matrix from the n response variables. This distance/dissimilarity matrix is reduced to a matrix of dimension k, where k is specified by the user. Two dimensions (k = 2) were deemed appropriate for detecting differences among groups. Each fitted observation was assigned k scores, one score for each chosen dimension. The k scores were plotted in a scatterplot, where each point represented one observation and its canonical analysis of principal coordinates (CAP) scores. The distance between k scores (calculated from the original n variables) indicates the degree of similarity between observations. Manhattan dissimilarity was the chosen distance metric within this analysis, as this is common within db-RDA implementation in ecological and biological studies (Qi and Voit, 2017). Additional implementations of db-RDA were then utilized to examine demographic variables of interest. To compare across species and sex by each location, db-RDA was fitted at each location. Sex and species levels were the explanatory variables, and metabolite levels remained the response variables. Similarly, db-RDA was fitted across location and sex by species and across location and species by sex. Hermaphrodites were not considered in this analysis because of the small sample size. db-RDA analysis was conducted within R version 4.1.0 (R Core Team, 2021). The of db-RDA analysis was constructed using the capscale function under the vegan package version 2.5–7. Glycogen concentrations were log transformed, and analysis of variance was used to model by species, site and sex. Data have been published as a United States Geological Survey (USGS) data release (Steiner et al., 2023); the code and project work have been published as a USGS software release (Oliver et al., 2023).

Table 1. Number of individuals, shell length and sex of L. siliquoidea and L. cardium sampled from three Indiana streams.

Species and stream		Female		Male		Hermaphroditea
		Samples (gravid)	Mean length (range) (mm)	Samples	Mean length (range) (mm)	H-Fb samples (gravid)	H-Mc samples	Mean length (range) (mm)
L. siliquoidea
Kilmored		12 (6)	76 (70–87)	6	79 (64–97)	1	0	66
Mud		14 (0)	69 (59–90)	6	74 (68–86)	0	0	NA
Wildcat		4 (2)	85 (78–94)	5	83 (47–105)	11 (5)	0	80 (69–90)
L. cardium
Kilmore		11 (3)	90 (82–104)	9	115 (106–132)	0	0	NA
Mud		3 (1)	84 (78–90)	14	99 (73–118)	0	2	100 (93–106)
Wildcat		11 (2)	102 (73–132)	7	110 (84–127)	1	1	111, 124

^{a a}Table omits one hermaphrodite identified as “male–female” (H-MF) due to the presence of roughly equal ovarian and spermatogenic tissue.

^{b b}H-F = hermaphrodite with female elements in the majority

^{c c}H-M = hermaphrodite with male elements in the majority

^{d d}Sex of one mussel from this stream was undetermined.

Results

Mussel demographics

The final data set of mussels (n = 112) consisted of 58 L. cardium and 54 L. siliquoidea. Specific numbers by location and sex, along with size information, are reported in Table 1. Notably, at least one gravid female per species was collected from each location, except for L. siliquoidea at Mud Creek. Hermaphrodites of one or both species were also collected from each location including 11 of 20 L. siliquoidea and 2 of 20 L. cardium collected at Wildcat Creek.

Metabolomics

Eighty-six targeted metabolites were detected in haemolymph samples (Supplementary Material, Table S3). Mussel populations partially overlapped in metabolite levels among streams. Across species, the difference in db-RDA scores was greatest between Kilmore Creek and Mud Creek mussels, while scores were most similar between Mud and Wildcat Creek mussels (Supplementary Material, Fig. S1a). Within each species, differences among sites followed a similar pattern, with greatest intraspecies differences occurring between Kilmore Creek and Mud Creek samples in both species (Fig. 2a and b).

Figure 2.

Intraspecies metabolite differences among locations in (a) L. cardium and (b) L. siliquoidea. Each point represents one mussel. Ellipse is 95% CI.

Interspecies metabolite differences were detected in CAP 1 scores at each individual location (Fig. 3a–c) and across all locations (Supplementary Material, Fig. S1b). Variation along CAP 2 was small and did not separate species. Inter- and intraspecies metabolite variability among sexes was less distinct than differences among locations (Fig. 4a and b; Supplementary Material, Fig. S2). Across locations, interspecies differences between males and females were detected in CAP 2, but CAP 1 scores overlapped (Supplementary Material, Fig. S3). Metabolite differences were more distinct between male and female L. cardium in contrast to L. siliquoidea in which hermaphrodites showed the most separation (Fig. 4a and b). The PCoA analysis of gravid/nongravid females showed considerable overlap in metabolite levels between groups; however, the sample size of gravid females was small (Fig. 5a and b).

Figure 3.

Interspecies differences in metabolite levels by location: (a) Kilmore Creek, (b) Mud Creek, (c) Wildcat Creek. Each point represents one mussel. Ellipse represents 95% confidence limit.

Figure 4.

Intraspecies metabolite differences among sexes across locations in (a) L. cardium and (b) L. siliquoidea. Each point represents one mussel. Ellipse represents 95% confidence limit.

Figure 5.

Differences in metabolite levels by gravidity status across locations in (a) L. cardium and (b) L. siliquoidea. Each point represents one mussel. Gravid hermaphrodites (n = 5 L. siliquoidea) were excluded from this analysis. Ellipse represents 95% confidence limit.

Individual metabolites were assigned to one of three categories based on occurrence (percentage of mussels with measurable concentration of metabolite X) and variability in concentration across location, species and sex (based on CAP 1 and 2 scores from the db-RDA analysis). Metabolites categorized as “high occurrence” were found in >80% of mussels. Metabolites categorized as “high variability” had a CAP 1 or 2 score >1.13. Category 1 metabolites (n = 27) had high occurrence and low variability (Table 2). Category 2 metabolites (n = 22) had high occurrence and high variability (CAP 1 or CAP 2 score >1.13) (Table 3). Category 3 metabolites (n = 38) had low occurrence and low variability (Table 4).

Table 2. Mean concentrations (μM) with 95% confidence limits of metabolites with high occurrence and low variability (Category 1).

Category/metabolite	L. siliquoidea			L. cardium
	Kilmore Creek	Mud Creek	Wildcat Creek	Kilmore Creek	Mud Creek	Wildcat Creek
Amino acid/peptide
S-(5′-Adenosyl)-l-homocysteine	0.22 (0.17, 0.26)	0.40 (0.25, 0.54)	0.64 (0.17, 1.11)	0.37 (0.15, 0.59)	0.34 (0.05, 0.63)	0.34 (0.15, 0.53)
l-Arginine	4.66 (3.19, 6.14)	7.96 (5.80, 10.11)	12.16 (7.22, 17.11)	4.95 (2.46, 7.45)	8.41 (5.45, 11.36)	7.38 (2.99, 11.77)
l-Glutamine	1.59 (1.11, 2.07)	2.41 (1.13, 3.69)	3.03 (2.02, 4.03)	2.65 (1.70, 3.61)	1.50 (0.86, 2.14)	6.21 (4.45, 7.97)
l-Isoleucine	199.04 (177.30, 220.78)	141.99 (88.06, 195.93)	221.80 (155.54, 288.06)	136.00 (110.00, 162.00)	177.86 (154.45, 201.26)	229.93 (180.11, 279.76)
l-Methionine	22.37 (19.00, 25.75)	21.34 (12.67, 30.01)	27.19 (18.76, 35.63)	17.98 (12.02, 23.95)	27.19 (21.47, 32.91)	33.76 (25.21, 42.30)
l-Phenylalanine	25.19 (21.25, 29.14)	19.69 (16.43, 22.95)	26.30 (21.45, 31.16)	17.61 (14.63, 20.59)	18.24 (15.64, 20.83)	24.58 (20.75, 28.42)
l-Proline	8.17 (7.46, 8.89)	7.95 (5.98, 9.92)	11.48 (9.45, 13.52)	7.21 (6.14, 8.28)	8.12 (6.89, 9.35)	10.29 (8.75, 11.82)
Tyramine	0.69 (0.54, 0.83)	0.73 (0.53, 0.94)	1.22 (0.71, 1.73)	0.85 (0.00, 2.77)	2.63 (0.12, 5.15)	2.01 (0.00, 4.60)
d-Tryptophan	6.95 (5.37, 8.53)	4.40 (3.61, 5.20)	6.64 (4.85, 8.44)	4.81 (3.95, 5.67)	4.80 (4.04, 5.57)	5.24 (4.44, 6.04)
3-Methyl-l-histidine	39.46 (29.57, 49.36)	29.10 (12.99, 45.21)	26.71 (19.37, 34.04)	20.64 (13.97, 27.31)	21.45 (15.84, 27.06)	45.91 (19.75, 72.07)
N-Methyl-l-alanine	482.32 (386.77, 577.87)	347.54 (289.83, 405.25)	518.07 (385.46, 650.67)	500.02 (423.93, 576.12)	582.20 (502.66, 661.74)	446.07 (361.67, 530.47)
2-Acetamido-2-deoxy-beta-d-glucosylamine	10.35 (7.73, 12.97)	8.25 (6.36, 10.14)	7.56 (5.93, 9.18)	8.53 (6.59, 10.47)	10.33 (7.91, 12.75)	14.56 (3.26, 25.86)
Ophthalmic acid	10.96 (9.55, 12.38)	8.82 (6.99, 10.64)	7.90 (6.28, 9.52)	9.70 (8.27, 11.13)	8.86 (7.42, 10.29)	10.55 (8.85, 12.26)
l-Pipecolic acid	2.30 (1.37, 3.23)	2.10 (1.79, 2.40)	2.26 (1.81, 2.71)	2.57 (2.02, 3.12)	3.43 (3.01, 3.86)	3.95 (2.54, 5.36)
Carbohydrate/energy production
6-Hydroxypyridine-3-carboxylic acid	2.51 (1.59, 3.42)	2.12 (1.33, 2.91)	4.31 (3.29, 5.33)	2.75 (1.93, 3.57)	1.99 (1.13, 2.85)	4.55 (3.18, 5.92)
Methyl-beta-d-galacto-pyranoside	17.12 (12.63, 21.61)	16.13 (13.73, 18.53)	29.41 (24.52, 34.29)	15.17 (11.94, 18.39)	11.94 (8.89, 15.00)	22.09 (16.62, 27.57)
d-Pantothenic acid	2.24 (1.73, 2.75)	1.14 (0.90, 1.39)	2.10 (1.72, 2.47)	1.14 (0.92, 1.37)	1.21 (0.96, 1.47)	1.89 (1.07, 2.71)
Lipid pathways
l-Carnitine	3.45 (2.85, 4.05)	3.11 (2.36, 3.85)	3.27 (2.59, 3.94)	4.19 (3.48, 4.90)	3.48 (2.87, 4.10)	4.47 (3.14, 5.79)
l-Carnosine	11.89 (10.37, 13.42)	10.03 (9.07, 10.99)	10.32 (9.10, 11.55)	7.19 (6.02, 8.35)	8.02 (6.99, 9.05)	7.78 (6.73, 8.84)
Erucic acid	2.46 (0.53, 4.40)	4.6 (2.12, 7.09)	3.15 (1.82, 4.48)	3.05 (1.84, 4.25)	2.22 (1.14, 3.30)	2.63 (1.45, 3.80)
4-Guanidinobutyric acid	1.14 (0.72, 1.57)	2.2 (1.86, 2.53)	2.23 (1.71, 2.74)	0.96 (0.60, 1.32)	1.81 (1.56, 2.07)	1.99 (1.54, 2.44)
Nucleotide
Adenine	63.92 (50.01, 77.83)	35.80 (26.90, 44.69)	31.42 (21.33, 41.52)	23.67 (17.19, 30.16)	27.48 (21.73, 33.24)	19.59 (14.87, 24.31)
Inosine	322.99 (258.52, 387.45)	296.76 (238.97, 354.55)	230.95 (158.07, 303.82)	262.30 (205.52, 319.08)	299.28 (232.50, 366.07)	225.01 (182.55, 267.47)
Neurotransmitter
dl-4-Hydroxy-3-methoxymandelic acid	96.24 (84.59, 107.88)	83.37 (64.32, 102.42)	223.60 (183.67, 263.52)	93.47 (75.96, 110.98)	55.87 (43.66, 68.08)	153.79 (108.85, 198.73)
dl-Normetanephrine	46.91 (39.83, 53.99)	35.18 (29.37, 40.99)	46.92 (38.04, 55.8)	32.41 (26.94, 37.88)	33.09 (28.21, 37.96)	44.26 (37.43, 51.10)
Vitamin/cofactor
Nicotinic acid	9.06 (7.79, 10.32)	5.81 (4.38, 7.25)	9.83 (6.72, 12.94)	5.56 (4.72, 6.39)	3.99 (3.11, 4.86)	8.44 (2.31, 14.57)
(+)-alpha-Tocopherol	28.81 (16.42, 41.21)	19.71 (11.09, 28.33)	26.52 (12.46, 40.57)	10.39 (5.37, 15.40)	13.73 (8.40, 19.05)	12.99 (6.52, 19.46)

Table 3. Mean concentrations (μM) with 95% confidence limits of metabolites with high occurrence and high variability (Category 2).

	L. siliquoidea			L. cardium
Category/metabolite	Kilmore Creek	Mud Creek	Wildcat Creek	Kilmore Creek	Mud Creek	Wildcat Creek
Amino acid/peptide
l-Citrulline	12.14 (8.04, 16.25)	25.05 (16.46, 33.64)	33.38 (20.91, 45.85)	10.14 (4.73, 15.55)	17.24 (11.00, 23.48)	14.17 (5.31, 23.04)
l-Glutamic acid (glutamate)	72.46 (62.98, 81.94)	46.90 (33.67, 60.12)	69.12 (54.69, 83.55)	85.67 (68.01, 103.33)	48.89 (40.98, 56.80)	103.41 (88.34, 118.48)
d-Ornithine	1.32 (0.65, 1.99)	3.98 (2.73, 5.23)	6.67 (2.37, 10.97)	1.14 (0.12, 2.15)	3.26 (2.14, 4.39)	4.26 (0.40, 8.12)
2,6-Diaminopimelic acid	2.79 (1.30, 4.29)	3.65 (1.10, 6.20)	3.68 (2.15, 5.22)	3.45 (2.47, 4.44)	1.84 (1.29, 2.40)	2.73 (1.54, 3.93)
2-Pyrrolidone-5-carboxylic acid	256.52 (219.51, 293.54)	162.36 (116.52, 208.20)	233.81 (179.12, 288.51)	302.60 (238.73, 366.47)	172.57 (143.74, 201.41)	356.58 (297.97, 415.19)
3-Nitro-l-tyrosine	0.65 (0.40, 0.89)	0.43 (0.24, 0.62)	0.66 (0.34, 0.97)	0.27 (0.12, 0.42)	0.17 (0.09, 0.25)	0.30 (0.16, 0.44)
Carbohydrate/energy production
d-Malic acid	157.57 (98.60, 216.54)	225.09 (154.33, 295.85)	162.15 (82.36, 241.93)	137.54 (79.09, 195.99)	257.70 (203.50, 311.90)	159.58 (112.11, 207.05)
Salsolinol	0.64 (0.35, 0.94)	1.47 (1.05, 1.88)	2.97 (2.31, 3.62)	0.63 (0.47, 0.78)	1.14 (1.00, 1.28)	3.39 (2.85, 3.94)
Lipid pathways
O-Acetyl-l-carnitine	0.26 (0.16, 0.35)	0.67 (0.45, 0.89)	0.14 (0.10, 0.19)	0.15 (0.00, 0.37)	0.85 (0.70, 1.00)	0.45 (0.18, 0.73)
Glutaryl-l-carnitine	0.45 (0.36, 0.54)	0.46 (0.33, 0.59)	0.55 (0.43, 0.68)	0.68 (0.53, 0.84)	0.75 (0.55, 0.95)	1.24 (1.07, 1.42)
dl-3-Aminoisobutyric acid	3.19 (2.28, 4.09)	2.22 (1.36, 3.08)	4.53 (3.56, 5.50)	3.94 (3.01, 4.86)	4.73 (3.79, 5.66)	4.25 (3.29, 5.22)
gamma-Linolenic acid	0.66 (0.41, 0.92)	1.35 (0.35, 2.35)	4.66 (0.47, 8.84)	1.89 (0.00, 7.39)	5.21 (0.00, 12.51)	1.40 (0.00, 2.91)
(+/−)-Mevalonolactone	129.84 (106.28, 153.40)	74.79 (57.99, 91.59)	105.29 (79.39, 131.19)	91.31 (71.26, 111.36)	133.92 (114.27, 153.58)	102.96 (90.28, 115.65)
Palmitic acid (hexadecenoic acid)	237.23 (0.00, 480.14)	344.86 (203.81, 485.92)	385.42 (236.36, 534.47)	655.18 (338.80, 971.56)	535.30 (121.11, 949.50)	491.96 (381.81, 602.11)
Petroselinic acid (octadecenoic acid)	24.67 (9.74, 39.59)	106.13 (26.48, 185.78)	320.82 (42.55, 599.09)	64.52 (0.00, 365.28)	251.42 (0.00, 653.41)	97.84 (0.00, 196.01)
Nucleotide
Cytosine	1.39 (0.89, 1.89)	1.58 (1.28, 1.88)	1.87 (1.61, 2.13)	2.02 (1.53, 2.52)	2.44 (1.89, 2.99)	1.96 (1.50, 2.42)
Guanosine	7.62 (6.50, 8.73)	3.82 (2.70, 4.93)	7.27 (6.23, 8.30)	6.28 (4.72, 7.83)	3.27 (1.77, 4.77)	8.03 (6.60, 9.45)
Hypoxanthine	109.65 (82.53, 136.77)	51.04 (34.27, 67.8)	50.98 (30.38, 71.58)	33.16 (19.36, 46.96)	30.99 (18.32, 43.65)	18.35 (10.07, 26.62)
Thymidine	6.1 (3.68, 8.52)	5.24 (3.64, 6.84)	4.88 (3.32, 6.45)	3.61 (1.80, 5.41)	5.02 (3.45, 6.59)	6.3 (4.12, 8.49)
2′-Deoxyguanosine	290.82 (234.11, 347.53)	262.59 (214.21, 310.96)	212.97 (147.71, 278.22)	213.13 (158.44, 267.82)	271.54 (206.92, 336.16)	204.64 (166.45, 242.83)
5′-Deoxyadenosine	139.58 (60.50, 218.65)	120.97 (56.95, 185.00)	58.91 (35.52, 82.30)	308.93 (214.69, 403.16)	243.83 (155.10, 332.55)	288.58 (156.41, 420.76)
Vitamin/cofactor
Nicotinamide	62.03 (50.61, 73.46)	33.78 (21.49, 46.06)	74.02 (54.1, 93.94)	30.98 (23.53, 38.42)	6.39 (2.65, 10.13)	34.5 (22.17, 46.83)

A single-point calibration was used to obtain a micromolar concentration (μM) with 95% confidence limits.

Table 4. Table 3 Mean concentrations (μM) with 95% confidence limits of metabolites with low occurrence and low variability (Category 3).

	L. siliquoidea			L. cardium
Pathway/metabolite	Kilmore Creek	Mud Creek	Wildcat Creek	Kilmore Creek	Mud Creek	Wildcat Creek
Amino acid/peptide
l-Anserine	0.75 (0.00, 2.14)	0.46 (0.12, 0.81)	0.48 (0.11, 0.85)	0.15 (0.00, 0.33)	0.08 (0.00, 0.16)	0.28 (0.00, 0.59)
l-Histidine	24.94 (15.00, 34.88)	26.12 (8.36, 43.88)	19.13 (12.15, 26.11)	46.03 (25.83, 66.24)	21.00 (7.61, 34.39)	12.65 (3.65, 21.64)
l-Histidinol	0.72 (0.29, 1.14)	0.06 (0.00, 0.13)	0.53 (0.20, 0.86)	0.21 (0.08, 0.34)	0.10 (0.02, 0.18)	0.36 (0.15, 0.56)
N-Acetyl-l-leucine	0.87 (0.69, 1.06)	0.08 (0.00, 0.17)	0.62 (0.17, 1.07)	0.26 (0.01, 0.52)	0.64 (0.37, 0.91)	0.22 (0.07, 0.37)
l-Lysine	4.51 (1.64, 7.37)	8.36 (5.82, 10.89)	9.83 (6.56, 13.10)	10.95 (8.62, 13.29)	11.13 (8.93, 13.33)	21.80 (9.30, 34.31)
Thymine	11.89 (7.19, 16.60)	5.63 (3.53, 7.74)	7.41 (4.32, 10.49)	8.22 (4.72, 11.72)	7.76 (4.41, 11.11)	9.23 (5.83, 12.64)
l-Valine	9.91 (0.00, 30.16)	48.54 (10.88, 86.19)	29.12 (0.00, 82.66)	2.57 (0.00, 40.29)	40.08 (0.00, 90.16)	47.28 (8.32, 86.23)
l-Cysteinesulfinic acid	10.85 (6.17, 15.53)	0.00	0.61 (0.00, 1.86)	8.58 (5.37, 11.79)	0.00	1.72 (0.00, 3.56)
Cytidine diphosphate (CDP)-ethanolamine	1.75 (0.75, 2.76)	0.14 (0.06, 0.21)	0.72 (0.25, 1.18)	0.99 (0.44, 1.54)	0.26 (0.07, 0.45)	0.63 (0.00, 1.25)
trans-4-Hydroxy-l-proline	0.71 (0.00, 4.10)	0.93 (0.00, 2.84)	0.78 (0.00, 2.38)	1.26 (0.07, 2.45)	0.00	0.67 (0.00, 1.64)
l-Kynurenine	0.87 (0.33, 1.41)	0.76 (0.23, 1.30)	0.88 (0.17, 1.59)	0.04 (0.00, 0.26)	0.28 (0.00, 0.58)	0.28 (0.00, 0.60)
Nɛ,Nɛ,Nɛ-Trimethyllysine hydrochloride	0.58 (0.00, 1.23)	1.49 (0.47, 2.50)	1.00 (0.27, 1.73)	0.00 (0.00, 0.16)	0.35 (0.12, 0.57)	0.53 (0.00, 1.19)
Spermidine	2.04 (0.73, 3.35)	0.08 (0.00, 0.25)	2.77 (0.00, 5.79)	0.13 (0.00, 0.32)	0.00	10.09 (0.00, 28.24)
1-Aminocyclopropanecarboxylic acid	10.76 (3.44, 18.07)	12.94 (3.03, 22.85)	8.38 (0.61, 16.15)	7.38 (0.00, 16.51)	17.89 (7.16, 28.63)	35.97 (0.00, 91.13)
(3-Carboxypropyl)trimethylammonium chloride	0.23 (0.09, 0.36)	0.10 (0.02, 0.18)	0.16 (0.08, 0.25)	0.34 (0.21, 0.47)	0.20 (0.12, 0.28)	0.21 (0.02, 0.40)
Carbohydrate/energy production
N-Acetylneuraminic acid	7.08 (2.82, 11.34)	10.24 (4.03, 16.46)	13.56 (4.67, 22.44)	1.89 (0.00, 4.72)	3.73 (0.09, 7.37)	0.27 (0.00, 0.64)
Azelaic acid	0.10 (0.01, 0.18)	0.08 (0.02, 0.14)	0.13 (0.04, 0.22)	0.08 (0.01, 0.16)	0.08 (0.01, 0.15)	0.12 (0.03, 0.21)
Dulcitol	6.84 (2.26, 11.41)	6.13 (1.86, 10.40)	2.89 (0.00, 5.82)	4.62 (0.90, 8.34)	5.47 (1.23, 9.71)	6.64 (2.88, 10.40)
3-Hydroxy-3-methylglutaric acid	20.4 (13.5, 27.29)	3.40 (1.68, 5.13)	24.25 (0.00, 51.29)	19.42 (0.00, 55.61)	32.70 (0.00, 81.23)	13.85 (4.24, 23.47)
d-(+)-Raffinose	0.12 (0.00, 0.46)	1.82 (0.00, 4.31)	5.56 (0.60, 10.53)	1.21 (0.00, 5.75)	4.27 (0.00, 10.34)	2.51 (0.00, 5.75)
Stachyose	0.03 (0.00, 0.07)	0.82 (0.00, 1.9)	2.28 (0.79, 3.76)	0.50 (0.00, 2.52)	1.62 (0.00, 4.31)	1.12 (0.00, 2.52)
d-(+)-Trehalose	2.01 (0.42, 3.61)	4.50 (1.66, 7.35)	9.59 (0.00, 20.82)	6.02 (0.00, 24.37)	21.99 (0.00, 45.61)	2.85 (0.00, 7.03)
Lipid/pathways
7-Dehydro-cholesterol	5.97 (0.00, 14.04)	3.41 (1.41, 5.42)	7.99 (3.97, 12.02)	0.60 (0.00, 1.24)	0.81 (0.01, 1.62)	1.11 (0.27, 1.95)
1,2-Didecanoyl-sn-glycero-3-phosphocholine	0.09 (0.00, 0.19)	0.3 (0.16, 0.44)	0.15 (0.05, 0.25)	0.34 (0.14, 0.54)	0.63 (0.42, 0.84)	0.42 (0.29, 0.55)
Methyl jasmonate	50.29 (0.00, 135.57)	8.43 (5.58, 11.29)	4.08 (0.40, 7.77)	3.65 (0.00, 12.17)	14.87 (3.98, 25.77)	3.99 (1.38, 6.60)
Sphingomyelin	0.00	0.01 (0.00, 0.03)	0.02 (0.00, 0.06)	0.00 (0.00, 0.80)	1.77 (0.81, 2.72)	0.22 (0.00, 0.64)
Stearic acid	1.10 (1.10, 1.10)	36.06 (28.71, 43.40)	41.85 (25.66, 58.04)	42.52 (22.60, 62.44)	65.91 (44.44, 87.37)	52.02 (39.54, 64.50)
Nucleotide
2′-Deoxycytidine	0.14 (0.04, 0.25)	0.13 (0.04, 0.22)	0.16 (0.07, 0.24)	0.34 (0.22, 0.46)	0.28 (0.15, 0.41)	0.23 (0.10, 0.35)
2-Deoxy-d-ribose	0.00	2.67 (0.00, 8.15)	68.15 (39.81, 96.50)	129.12 (101.49, 156.74)	114.52 (81.58, 147.46)	191.62 (152.95, 230.29)
5′-Deoxy-5′-(methylthio)adenosine	0.02 (0.02, 0.02)	0.79 (0.59, 0.99)	0.37 (0.05, 0.68)	0.24 (0.04, 0.44)	0.53 (0.36, 0.71)	0.59 (0.26, 0.92)
2′-Deoxyuridine	21.86 (7.30, 36.41)	8.57 (0.49, 16.64)	18.87 (7.20, 30.55)	51.56 (31.29, 71.83)	38.58 (16.68, 60.47)	25.73 (12.40, 39.07)
2′-Deoxyguanosine 5′-monophosphate	36.42 (14.02, 58.82)	47.36 (18.30, 76.42)	77.52 (35.87, 119.17)	28.03 (0.00, 62.35)	48.35 (3.79, 92.90)	9.36 (3.57, 15.15)
Caffeine	0.00	0.15 (0.00, 0.46)	4.2 (0.00, 10.21)	0.5 (0.00, 1)	0.00	0.00
Neurotransmitter
Dopamine	3.05 (1.99, 4.11)	2.57 (1.33, 3.82)	3.84 (2.13, 5.55)	0.64 (0.00, 1.51)	3.03 (2.24, 3.83)	3.9 (2.76, 5.05)
Vitamin/cofactor
Coenzyme Q10	0.02 (0.02, 0.02)	0.33 (0.19, 0.48)	0.07 (0.00, 0.13)	0.11 (0.00, 0.26)	0.35 (0.18, 0.53)	0.39 (0.00, 0.9)
Ergocalciferol	0.37 (0.02, 0.72)	1.87 (0.00, 3.87)	2.74 (1.54, 3.93)	3.57 (2.05, 5.09)	2.89 (1.65, 4.13)	14.94 (8.34, 21.53)

A single-point calibration was used to obtain a micromolar concentration (μM) with 95% confidence limits.

Metabolite differences between species were detected, based on fold change and P value, in five metabolites, all in Category 3, including 2-deoxy-d-ribose, ergocalciferol, 5′-deoxyadenosine, sphingomyelin and 1,2-didecanoyl-sn-glycero-3-phosphocholine. Most notably, sphingomyelin was elevated in 8 L. cardium from Mud Creek, none of which were females (not shown).

Glycogen concentration

Glycogen varied widely among individuals, and a few outlier values overly influenced differences among groups (Table 5 and Supplementary Material, Fig. S4). No predictors or their interactions (species F = 0.784, df = 1, P = 0.378; location F = 1.924, df = 2, P = 0.151; and sex F = 0.566, df = 6, P = 0.756) explained glycogen concentrations. Notably, L. siliquoidea had greater glycogen concentration than L. cardium, but uncertainty was high around this estimate, and the 95% confidence interval (CI) included zero (1.05, 95% CI, 0.00 to 2.29). All other estimated coefficients were near zero.

Table 5. Mean (SD) and range of glycogen concentrations in L. cardium and L. siliquoidea (wet weight, wwt) sampled from three Indiana streams .

	Sex	Number	Mean (SD) glycogen (mg/g wwt)
L. cardium			1.36 (1.40)
Kilmore Creek
	Female	11	1.27 (0.71)
	Male	8	1.31 (1.47)
Mud Creek
	Female	3	0.50 (0.24)
	Herm	2	1.36 (0.86)
	Male	14	1.71 (1.82)
Wildcat Creek
	Female	11	1.75 (1.91)
	Herm	2	0.36 (0.07)
	Male	7	0.91 (0.74)
L. siliquoidea			1.87 (1.32)
Kilmore Creek
	Female	11	1.18 (0.84)
	Herm	1	0.66 (NA)
	Male	6	1.34 (0.74)
Mud Creek
	Female	8	1.44 (0.64)
	Male	4	3.14 (2.14)
Wildcat Creek
	Female	3	3.02 (1.69)
	Herm	10	2.64 (1.45)
	Male	3	1.78 (1.13)

Sample numbers differ from Table 1 due to loss of some samples during analysis.

Discussion

The present study provides baseline metabolomic data from robust populations of two Lampsilis species, which are relatively common and widely distributed, and thus could be used throughout their ranges as site-specific sentinels of mussel health, as well as environmental monitors of water quality. The variability of haemolymph metabolites was most influenced by species, followed by site and sex. Thus, single species sampling for metabolomic analysis and monitoring would be more informative than mixed species sampling, and spatial differences within a watershed may be less important than species differences. Even so, few specific metabolites showed detectable interspecies differences based on P value and fold change. Based on the similarity in occurrence between species, Categories 1 and 2 metabolites may be most useful in targeted metabolomic biomonitoring.

The sex ratio of mussels varied among stream locations, and the occurrence of hermaphroditism was notably greater at Wildcat Creek (Table 1). The incidence of hermaphroditism in unionids is reportedly low and considered accidental or occasional (van der Schalie, 1970; Downing et al., 1989; Garner et al., 1999; Haag and Staton, 2003), so the greater frequency in our study was unexpected. Hermaphroditism has been associated with low-density populations (Kat, 1983; Bauer, 1987) and translocation to a different habitat, presumably with improved reproductive conditions (Heinricher and Layzer, 1999), neither of which occurred in the present study. The reason for the greater number of hermaphrodites in our study was not explained by metabolomic differences. Additional sampling may determine whether these mussels were representative of the larger population or other factors (e.g. temperature, sex ratio) affected gonadal differentiation in these streams.

We expected to find a distinct separation of metabolite scores by sex but found considerable overlap among sexes in both species. Additionally, sex was not a factor in the individual mussels that contributed more than 5% of the variability at a location (Supplementary Material, Table S4). In unionid mussels, gametogenesis occurs in gonadal visceral tissue. Female mussels brood larvae in gill marsupium where nutrients are exchanged (Schwartz and Dimock, 2001). Thus, haemolymph of Lampsilis mussels, as was used in the present study, may be a poor source of metabolites indicative of sex or reproductive state compared to gonadal or gill tissue. Many unionid mussel species are not sexually dimorphic, and determination of sex requires sacrifice or biopsy of gonadal tissue, an undesirable outcome for imperilled species. When possible, controlling for potential differences among sexes is beneficial; however, when sex determination is not possible, the results presented herein indicate that sex-related differences in haemolymph metabolites are less important than those related to site and species in stable populations. In the present study, mussels were sampled in the summer when field sampling/monitoring of populations is most common; future studies could evaluate temporal influences on metabolites, not only from reproductive condition but also from shifts in diet composition, water temperature and flow regime.

Metabolite responses in aquatic invertebrates have been documented for a range of stressors, including contaminants, hypoxia, heavy metals, disease and starvation (Lankadurai et al., 2013; Nguyen and Alfaro, 2020; Dumas et al., 2020a, 2020b). The 86 metabolites in the present analysis are components of a range of metabolic pathways, such as energy production (catabolism of carbohydrates, lipids, protein), cell growth (nucleotide synthesis), neuroendocrine synthesis, as well as antioxidants, osmolytes and cell signalling molecules, and could detect mussel response to various perturbations or environmental exposures. A comprehensive review of stressors to metabolite linkages is beyond the scope of this paper, but the following discussion provides an overview of functions and potential indications of many of the targeted metabolites detected in Lampsilis spp. in our study. It should be noted that metabolite function and pathways in freshwater mussels are largely assumed from literature on marine organisms (molluscs and crustaceans), the Human Metabolome Database (https://hmdb.ca), and KEGG pathways of marine bivalves (https://www.genome.jp).

The state of energy metabolism (catabolic versus anabolic) in a mussel is indicated by cofactors/vitamins, the amino acid pool, tricarboxylic acid (TCA) cycle intermediates, energy substrates, and ATP components. Target metabolites tryptophan, nicotinic acid and nicotinamide are precursors for nicotinamide adenine dinucleotide biosynthesis, an important cofactor for the multiple redox reactions in energy production. Pantothenic acid is a constituent of coenzyme A, the starting point for the TCA cycle and utilization of carbohydrates, fat and protein for ATP production. Adenine is a component of ATP, and malic acid (malate) is an intermediate in the TCA cycle. Decreased nicotinamide and pantothenic acid, along with energy substrates, were suggested as potential biomarkers of heavy metal toxicity in the crab Procambarus clarkii (Gago-Tinoco et al., 2014). Many amino acids are converted to intermediates in the TCA cycle, including target metabolites asparagine, glutamic acid (glutamate, ionic form) and glutamine, or for glucogenesis, such as isoleucine and valine. Decreases in these amino acids may indicate their use for energy production and/or protein synthesis in response to food limitation (Roznere et al., 2014), in response to xenobiotics (Hines et al., 2010) or pathogen exposure (Ericson et al., 2022).

An adequate pool of amino acids is clearly needed for protein synthesis, but additionally, free amino acids also function as osmolytes in bivalves (Hosoi et al., 2008; Zhang et al., 2022; Song et al., 2023). A shift in amino acid concentrations involving both those needed for osmoregulation and to meet increased energy demand is a generalized response in mussels to a range of stressors (e.g. thermal, pollution, pH, pathogen challenge) (Day et al., 1990; Lankadurai et al., 2013; Liu et al., 2013; Ellis et al., 2014; Campillo et al., 2015; Haider et al., 2020; Nguyen and Alfaro, 2020). The baseline levels of target amino acids established herein could be used to monitor relative changes as a general stress indicator in mussel populations.

Shifts in some amino acids are related to specific pathways or functions. For example, proline and its derivative hydroxyproline comprise two-thirds of the amino acids in collagen. Increased levels of proline and hydroxyproline are considered indicators of connective tissue degradation or decreased collagen synthesis from disease or toxicant exposure (Leonard et al., 2014b; Nguyen et al., 2018; Haider et al., 2020). Phosphoarginine (arginine + ATP) is the primary phosphagen in invertebrates. An increase in the ratio of arginine/phosphoarginine may be indicative of cellular energy demands and a switch to anaerobic metabolism (Hines et al., 2010). Target metabolites arginine, ornithine, proline, citrulline and polyamines (e.g. spermidine) are involved in urea production and mark shifts in nitrogen catabolism and excretion. Hypoxia-induced accumulation of urea cycle intermediates arginine and ornithine in Mytilus edulis indicated suppression of urea production (Haider et al., 2020). Exposure of Lampsilis fasciola to fadrozole (aromatase inhibitor) (Leonard et al., 2014b) and 17α-ethinylestradiol (synthetic oestrogen) (Leonard et al., 2014a) significantly altered these urea cycle metabolites. Spermidine is derived from ornithine and as a polyamine binds to RNA and DNA, playing a role in cell division, transcription, cell signalling and immune response (Bae et al., 2018). Levels of putrescine, another polyamine and ornithine decreased in A. plicata during captivity and food limitation, in association with reduced growth (Roznere et al., 2014, 2017). Spermidine and spermine, interconvertible polyamines, were upregulated and downregulated, respectively, in response to lipopolysaccharide exposure in the pearl oyster (Pinctada fucata martensii) (Cao et al., 2022), indicating their potential use as immunomarkers.

Target metabolite glutamine, interconvertible by deamidation to glutamic acid (ionized form is glutamate), has multiple cellular functions including protein and nucleotide synthesis and neuronal activity. Glutamine and methionine are precursors of glutathione, an antioxidant that can be an indicator of oxidative stress from temperature, xenobiotics, salinity and hypoxia (Hines et al., 2010; Hellou et al., 2012).

Several targeted metabolites are precursors or components of the neuroendocrine system in bivalves. Glutamic acid is a potent excitatory neurotransmitter and is decarboxylated to gamma-aminobutyrate (GABA), an inhibitory neurotransmitter (Cochran et al., 2012). GABA can be synthesized and released from hemocytes as an immunomodulator (Li et al., 2016). Tryptophan is a precursor of serotonin, a neurohormone associated with reproductive and digestive regulation. The catecholamine dopamine is a precursor for norepinephrine and epinephrine, all of which are found in haemolymph (Liu et al., 2018). These components of the neuroendocrine system have various triggers for synthesis and release. Shifts in pathway metabolites or end products could indicate an immune response, neurotoxicant exposure and hypoxia, among other stressors (see review by Liu et al., 2018; Dumas et al., 2020b). For example, dopamine and serotonin pathways were downregulated in Mytilus galloprovincialis exposed to wastewater treatment plant effluent (Dumas et al., 2020a). Higher temperature triggered increased glutamate and decreased GABA in the haemolymph of Perna canaliculus (Ericson et al., 2022). Glutamate increased in L. fasciola gill tissue in response to fadrozole hydrochloride exposure, indicating its neurotoxic effects on mussels (Leonard et al., 2014b).

Shifts in nucleotide metabolism can indicate reduced cell proliferation and protein synthesis for growth, tissue repair, reproduction and immune response. Targeted metabolites in pathways for nucleotide metabolism included adenine, inosine, cytosine, guanosine, hypoxanthine, thymidine., 5′-deoxyadenosine, 2′-deoxyguanosine, 2′-deoxycytidine, 2′-deoxyuridine and 2′-deoxyguanosine 5′-monophosphate. Levels of several nucleosides, including thymidine, guanosine and inosine, decreased in A. plicata during food limitation (Roznere et al., 2014), suggesting their use as bioindicators of growth. Purine metabolite levels were altered in the haemolymph of Pacific white shrimp (Penaeus vannamei) after exposure to elevated ammonia (Liu et al., 2019) and in the gills of M. edulis and Magallana (Crassostrea gigas) in response to hypoxia (Haider et al., 2020). Target metabolite 5′-methylthioadenosine initiates the methionine salvage pathway and is produced during synthesis of polyamines, such as spermidine. It is considered a key regulator of gene expression, proliferation, differentiation and apoptosis (Avila et al., 2004; Albers, 2009).

Target metabolites involved in lipid pathways include l-carnitine, acetyl-l-carnitine and glutaryl-carnitine. These function to transport acyl-groups to the mitochondria for beta-oxidation of fatty acids. Metal accumulation in P. clarkii was associated with shifts in lipid metabolites, including decreased carnosine and increased acetyl carnitine (Gago-Tinoco et al., 2014). Exposure of Penaeus vannamei to elevated ammonia also resulted in increased acetyl carnitine (Liu et al., 2019), and carnitine metabolism was upregulated in Ruditapes philippinarum in response to hypoxia (Sun et al., 2021). In contrast, food limitation led to a decrease in acetyl carnitine in A. plicata indicating its metabolism for energy production (Roznere et al., 2014). l-Carnitine and l-carnosine are concentrated in muscles and other tissues that metabolize fatty acids. Elevated levels in haemolymph could also result from tissue degradation. Targeted metabolites involved in lipid synthesis pathways include palmitic acid, gamma-linoleic acid and mevalonolactone. Palmitic acid, one of the most common saturated fats in animals, is synthesized from excess carbohydrates and the precursor to longer fatty acids. Gamma-linoleic acid, produced from linoleic acid, is used to synthesize prostaglandins; mevalonolactone readily converts to mevalonic acid and is used in synthesis of steroids and terpenes. Changes in lipid metabolic pathways involving these metabolites were detected in L. fasciola after exposure to synthetic oestrogen (Leonard et al., 2014a) and fadrozole hydrochloride (Leonard et al., 2014b), in Perna canaliculus infected with Vibrio spp. (Nguyen et al., 2018) and in Sinonovacula constricta in response to osmotic stress (Li et al., 2021).

Sphingomyelin levels were significantly greater in a group of non-female mussels, all of which were L. siliquoidea collected at Mud Creek, except for one of each species collected from Wildcat Creek. Sphingomyelin is a sphingolipid synthesized from ceramide and phosphatidylcholine. In vertebrates, sphingomyelin is a key component of cell membranes, especially the myelin sheath of neurons, and also plays a role in signal transduction, fatty acid uptake and initiation of an inflammatory response (Sakamoto et al., 2017). Sphingomyelin may not be present in invertebrates (Kobayashi et al., 2001), but more than likely, the metabolite detected in Lampsilis spp. in this study was an analogue, which occurs widely in invertebrates, ceramide phosphoethanolamine or ceramide aminoethylphosphonate (CAEP) (Hori et al., 1967; Le Grand et al., 2011; Panevska et al., 2019). The high proportions of CAEP in hemocytes of several marine bivalves (C. gigas and R. philippinarum) are thought to contribute to their role in an immune response (Le Grand et al., 2011, 2013). Potential reasons for the high levels of a sphingomyelin-type metabolite in a small subset of mussels are unclear. Further work on the composition and proportions of sphingolipids in freshwater mussels and their potential functions would be beneficial.

Our library of targeted metabolites also included exogenous compounds, such as caffeine, azelaic acid, dulcitol and stachyose, detected in some mussels (Category 3, Table 4). Their occurrence in mussel haemolymph is evidence of anthropogenic inputs to these streams and highlights the value of freshwater mussels and metabolomics for environmental monitoring and risk assessment.

Metabolomics is a snapshot of an organism's state—the where and how of sample collection influences the composition and variability of metabolites. Most metabolomic-based studies are conducted in controlled laboratory settings to isolate responses of individuals to known stressors (Ellis et al., 2014; Leonard et al., 2014a, 2014b; Haider et al., 2020). Organisms are often collected in the field and then transported to a laboratory for a holding period before tissue samples are collected. Yet, Hines et al. (2007) found that mussels sampled in the field had lower metabolic variability than those held in the laboratory for a stabilizing period. After collection of a tissue or fluid sample, metabolism quenching is critical to halt reactions or prevent chemical degradation or transformation. Freezing samples soon after collection by immersion in liquid nitrogen or placement on dry ice was reported as optimal for maintaining metabolite integrity (Venter et al., 2021). In the present study, haemolymph was collected from mussels in the field, to capture the real-time metabolome, and placed on dry ice to stabilize metabolite composition. Standard reporting requirements for metabolomic studies were followed to provide context for interpretation of the results and to ensure repeatability of sampling (Morrison et al., 2007).

Glycogen is a major energy storage molecule in invertebrates, including unionid mussels, and it has frequently been used as a measure of mussel condition. Significant shifts in glycogen concentration were demonstrated after prolonged captivity (Naimo et al., 1998), quarantine conditions (Patterson et al., 1999) and zebra mussel infestation (Haag et al., 1993; Patterson et al., 1997; Hallac and Marsden, 2000; Sousa et al., 2011; Beason and Schwalb, 2022). Glycogen content varies among tissues in unionids and is most evenly distributed in the foot compared to mantle (Naimo et al., 1998). Despite the use of foot tissue in our study, glycogen concentrations varied widely and did not show a discrete difference among sites, species or sex. Overall mean glycogen concentration was greater in L. siliquoidea than L. cardium (Supplementary Material, Fig. S4), but not at every site. Mean glycogen concentration was greatest at Wildcat Creek, and interspecies differences were greatest at this site. However, there was considerable overlap in the range of glycogen values (Table 5). Glycogen levels may vary between sexes and with reproductive status as females utilize glycogen during brooding of glochidia (Schwartz and Dimock, 2001). We found no detectable difference in glycogen concentration between males and females. Hornbach et al. (2021) also found no sex-related differences in glycogen concentration in L. cardium, but results were confounded by gravidity differences among locations. Similarly, carbohydrate concentration did not differ among brooding and non-brooding Actinonaias ligamentina (Baker and Hornbach, 2001). Non-gravid females in our study had greater glycogen stores, on average, than gravid females (Table 5); however, the influence of gravidity on glycogen stores was inconclusive given the high variability and limited number of gravid females in our sample.

Conservation activities for unionid mussels include augmentation and restoration of populations at select sites (FMCS, 2016). To select and evaluate suitable sites for these efforts, resource managers need assessment tools that can distinguish thriving from declining mussels, not merely mussel survival. Assessing the health status of a freshwater mussel is subjective at best, as biologists most often rely on behavioural signs, such as position in the substrate, valve gaping and response to physical touch (Neves, 1987; Richard et al., 2020). Metabolomics offers a quantitative and objective assessment of an organism's status. This technology is increasingly being used to identify biomarkers and disrupted pathways indicative of stressors such as contaminants, disease and environmental extremes (e.g. elevated temperature, hypoxia) (e.g. Lankadurai et al., 2013; Leonard et al., 2014a, 2014b; Nguyen and Alfaro, 2020; Dumas et al., 2020a; Haider et al., 2020). With the probable increase in metabolomics for mussel health assessment, more data on the metabolome of “normal, healthy” mussels will improve sampling strategies and interpretation of results. Baseline data on metabolite occurrence and variability provided herein can serve as a reference for a before/after monitoring approach (Pomfret et al., 2020). Categories 1 and 2 metabolites may be most suitable for monitoring temporal trends because they were detected in most mussels and had relatively low variability across location, sex and species. The low occurrence of Category 3 metabolites does not disqualify them as potentially important indicators as many are critical in metabolic processes (e.g. dopamine, amino acids, nucleosides) and may show significant temporal variability. We sampled mussels in the summer when field sampling/monitoring of populations is most common; continued monitoring over a full year would greatly increase the value of these data by characterizing temporal variation in metabolite concentrations. Finally, field monitoring of these metabolites with associated controlled laboratory trials can help link stressors to metabolite pathways and a resulting organism response.

Disclaimer

Use of trade, firm or product names is for descriptive purposes only and does not imply endorsement by US Government.

Ethics and Animal Welfare Statement

All biological samples were obtained in accordance with all federal, state and local laws and policies.

Author Contributions

Conceptualization, N.B.; data curation, J.P., J.N.S., D.W.; formal analysis, J.P., J.N.S., J.O., R.E.; investigation, B.F., G.N.B., S.B.S., A.R., N.B., J.P., J.N.S.; methodology, N.B., J.P.; project administration, N.B.; writing—original draft preparation, D.W., J.P., J.N.S.; writing—review and editing, all authors; visualization, J.O., R.E; supervision, N.B., D.W.; project administration, N.B., D.W.; funding acquisition, N.B., D.W. All authors have read and agreed to the published version of the manuscript.

Supplementary Material

Supplementary material is available at Conservation Physiology online.

Funding

The work was supported by the United States Fish and Wildlife Service Fund (190F161MD) and United States Geological Survey, Ecosystems Mission Area.

Data Availability Statement

The data in support of this publication are available at https://doi.org/10.5066/P9Y2T31F.