Chronic Embryo-Larval Exposure of Fathead Minnows to the Pharmaceutical Drug Metformin: Survival, Growth, and Microbiome Responses.

Metformin is a glucose-lowering drug commonly found in municipal wastewater effluents (MWWEs). The present study investigated the chronic effects of metformin in early-life stages of the fathead minnow (Pimephales promelas). Endpoints assessed were growth, survival, and deformities. The larval gut microbiome was also examined using 16 S ribosomal RNA gene amplicon sequencing to determine microbial community composition and alpha and beta diversity. Eggs and larvae were exposed to metformin measured concentrations (mean [standard deviation]) of 0.020 (0.017) μg/L (for controls) and 3.44 (0.23), 33.6 (1.6), and 269 (11) μg/L in a daily static-renewal setup, with 20 embryos per beaker. The low and middle metformin exposure concentrations represent river and MWWE concentrations of metformin. To detect small changes in growth, we used 18 replicate beakers for controls and 9 replicates for each metformin treatment. Over the 21-d exposure (5 d as embryos and 16 d posthatch [dph]), metformin did not affect survival or growth of larval fish. Hatch success, time to hatch, deformities in hatched fry, and survival were similar across all treatments. Growth (wet wt, length, and condition factor) assessed at 9 and 16 dph was also unaffected by metformin. Assessment of the microbiome showed that the larvae microbiome was dominant in Proteobacteria and Firmicutes, with small increases in Proteobacteria and decreases in Firmicutes with increasing exposure to metformin. No treatment effects were found for microbiome diversity measures. Control fish euthanized with the anesthetic tricaine methane sulfonate had decreased alpha diversity compared to those sampled by spinal severance. This experiment demonstrates that metformin at environmentally relevant concentrations (3.44 and 33.6 μg/L) and at 10 times MWWE concentrations (269 µg/L) does not adversely affect larval growth or gut microbiome in this ubiquitous freshwater fish species. Environ Toxicol Chem 2022;41:635-647.

INTRODUCTION

Metformin is a glucose‐lowering drug commonly taken by diabetics at doses ranging from 500 mg to 2 g/d. The drug is excreted unchanged in urine and is flushed down toilets with wastewater. Because of the high use of this drug and the variable breakdown in municipal wastewater‐treatment plants, metformin has been detected in municipal wastewater effluents (MWWEs; reviewed in Elizalde‐Velázquez and Gómez‐Oliván 2020). The highest concentrations in United States' MWWEs were 702 µg/L (Oliveira et al. 2015) and 82.7 µg/L (Meador et al. 2016). Concentrations of metformin in rivers are lower, with most under 5 µg/L (Elizalde‐Velázquez and Gómez‐Oliván 2020) but some samples being as high as 33.6 µg/L (Elliott et al. 2017). In a 2013 to 2014 assessment of 12 United States tributaries to the Great Lakes, metformin was found in 71% of the 292 samples taken (Elliott et al. 2017). A survey of 4 United States national parks showed metformin in 100% of samples, with a maximum concentration of 150 µg/L (Elliott and VanderMeulen 2017). Metformin was the most consistently detected pharmaceutical, found at least once at 68% of sites in a United States multiregional survey of urban streams (Bradley et al. 2020). The ubiquitous nature of this pharmaceutical has resulted in interest in its potential effects on aquatic organisms.

Some previous studies have shown that metformin decreases growth in fish. Embryo‐larval Japanese medaka (Oryzias latipes) had decreased weights and lengths after 28‐d exposure to metformin concentrations of 3.2 to 100 µg/L (Ussery et al. 2018). Fathead minnow adult male weight was decreased after 365‐d exposure to metformin from 30 d posthatch (dph) until adulthood (Niemuth and Klaper 2015). Brown trout (Salmo trutta) embryos exposed to metformin (up to 1000 µg/L) from 48 d postfertilization (dpf) to 8 wk post–yolk sac consumption showed no differences in length but some decreases in weight and increases in hepatic glycogen that were not dose‐related (Jacob et al. 2018). However, other long‐term fish exposures have shown no impacts of metformin on fish growth. For example, in a 19‐wk exposure of Japanese medaka to 50 to 373 µg/L metformin, there were no effects on growth of fish (Lee et al. 2019).

The fish microbiome is a topic of current interest in aquatic toxicology. With increasing concentrations of metformin in aquatic ecosystems, the effects on the fish microbiome are largely unknown. Although the mechanism of action for metformin (and other contaminants of concern) on the fish microbiome is unknown, metformin has been shown to alter the diversity and composition of bacteria in larval fish (Jacob et al. 2018; Rogall et al. 2020), providing the basis for our study.

The fish gut microbiome is important for host homeostasis, nutrient acquisition, and disease prevention; and it changes throughout fish development. The microbiota of early‐life stage fish is colonized at hatch, and this community continues to change until adulthood. Gibel carp (Carassius auratus gibelio) had increased alpha diversity (species evenness and/or richness) of the gut microbiota with developmental age and distinct bacterial communities at various developmental stages (Li et al. 2017). Grass carp (Ctenopharyngodon idellus), Chinese perch (Siniperca chuatsi), and southern catfish (Silurus meridionalis) had increased bacterial diversity in the gastrointestinal tract of larvae compared to adults and distinct beta diversity (degree of microbial community dissimilarity) between ages (Yan et al. 2016). The gut microbiome of carnivorous fish (Chinese perch and southern catfish) was initially dominant in Firmicutes but decreased after 18 dph, whereas Proteobacteria increased between 3 and 13 dph (Yan et al. 2016). Interestingly, the gut microbiota of larval zebrafish has been linked to gastrointestinal tract development; epithelial cell proliferation is stimulated by the presence of colonized microbiota (Cheesman et al. 2011). Furthermore, gnotobiotic (germ‐free) zebrafish had impaired gastrointestinal function, indicating that bacteria are an integral part of the host gut physiology (Lescak and Milligan‐Myhre 2017). The gut microbiota of larval fish is not well characterized, but because bacterial colonization occurs at hatch, the surrounding environment is important in determining which bacteria are present and subsequent effects on their overall health (Yan et al. 2016; Li et al. 2017).

Fish exposed to environmental stressors, such as pharmaceuticals and personal care products, can have altered gut bacterial community and diversity. The gut content microbiome of adult fish (wild brown trout, Salmo trutta; common carp, Cyprinus carpio; and rainbow darter, Etheostoma caeruleum) downstream of MMWE outfalls (containing mixtures of contaminants) had both decreased and increased alpha diversity, increased bacteria associated with MWWEs, and increased abundance of Proteobacteria compared to upstream fish; such changes have been linked to dysbiosis (microbial imbalance) and altered health outcomes (Giang et al. 2018; Sakalli et al. 2018; Restivo et al. 2020). Metformin affects the intestinal microbiome of larval brown trout (Jacob et al. 2018; Rogall et al. 2020), and those exposed to metformin (up to 1000 µg/L) had shifts in Firmicutes, with the families Planococcacea disappearing and Staphlococcaceae increasing compared to controls (Rogall et al. 2020). Zebrafish larvae exposed to arsenic had distinct beta diversity and decreased alpha diversity, as well as altered abundances of amplicon sequence variants (ASVs), in their full‐body microbiome. A 7‐d exposure to triclosan (a common antimicrobial found in MWWEs) caused shifts in beta and alpha diversity in the gastrointestinal microbiota of zebrafish larvae (Narrowe et al. 2015). These studies demonstrate that fish exposed to contaminants have altered gut microbiota compared to unexposed fish, even at early developmental stages.

In humans, metformin can affect body weight and the intestinal microbiome; however, the underlying mechanisms are not well understood. Metformin reduces blood sugar by improving sensitivity to insulin and reducing gastrointestinal glucose absorption and hepatic glucose production (Davidson and Peters 1997). In a review of drugs affecting weight in humans, metformin caused on average a 1.1‐kg weight loss (Domecq et al. 2015). It reduces waist circumference (Wang et al. 2018) and body weight and body mass index (in addition to exercise; Tiwari et al. 2019) in women with polycystic ovarian syndrome. The drug has been used to counteract the weight gain associated with antipsychotic medication use (Stroup and Gray 2018). Metformin‐associated weight loss has been linked to alterations in the gut microbial community, namely changes in the ratio of Bacteroidetes and Firmicutes (bacterial phyla) after treatment (Napolitano et al. 2014). This drug changes the gut microbiota in humans (reviewed in Vich Vila et al. 2020; Weersma et al. 2020), including healthy young men (Bryrup et al. 2019). More specifically, in healthy individuals treated with metformin (850 mg tablets, twice daily for 7 d), there was a significant reduction of bacterial diversity in the gut 24 h after administration and increased severity of gastrointestinal side effects associated with the increased relative abundance (proportion) of an opportunistic pathogen (Escherichia shigella spp.; Elbere et al. 2018). Studies in people with diabetes show that metformin may increase gut bacteria that produce short chain fatty acids linked to the suppression of appetite and that this may add to the beneficial and adverse actions of this drug (Forslund et al. 2015; Wu et al. 2017). When metformin‐treated stool samples (from human donors) were transplanted into germ‐free mice, the mice had improved glucose tolerance, improved metabolism, and increased expression of bacterial genes related to short chain fatty acid metabolism (Wu et al. 2017).

The primary objective of the present study was to investigate the effect of subchronic exposures of embryo‐larval fathead minnow (Pimephales promelas) to metformin. Static‐renewal exposures of 21 d were used to assess metformin's effects on hatchability, hatch success, time‐to‐hatch, deformities at hatch, and larval fish length, weight, and condition factors at 9 and 16 dph. Metformin's effects on the community composition and alpha and beta diversity of the gut microbiome were also assessed after 16 dph using 16S ribosomal RNA (rRNA) gene amplicon sequencing. An advantage of testing these early life stages is that for many chemicals that act through a nonspecific mode of action, fish early life stages tend to be more sensitive to toxicants relative to mature life stages (McKim 1977; Hutchinson et al. 1998). The present study investigates the potential risk that metformin poses to embryo‐larval stages of this commonly used laboratory test fish species.

METHODS

Chemicals

The metformin was prepared from metformin‐HCl (C4H11N5‐HCl, MW165.62, catalog no. PHR1084‐500 mg, lot no. LRAB3694, Chemical Abstracts Service no. 1115‐70‐4; Pharmaceutical Secondary Standard Certified Reference Material) purchased from Sigma‐Aldrich. Target nominal concentrations in aquaria were 3.1, 31, and 310 µg/L of metformin. Metformin superstock of 389.9 mg metformin/L was prepared in Milli‐Q water by mixing 500 mg metformin‐HCl (one 500‐mg vial of metformin‐HCl) in 1 L of water once for the duration of the experiment. From this superstock, a daily working stock solution of 31.2 mg/L was prepared by adding 8 mL of metformin superstock to 92 mL of culture water. Exposure beakers for the 3.1, 31, and 312 µg/L metformin exposure treatments were prepared daily by mixing 80 or 800 µL or 8 mL working stock with culture water up to a volume of 800 mL. Fish exposure beakers and metformin exposure solutions were allowed to equilibrate for 2 h in a 25 °C incubator before the embryo‐larval fathead minnows were transferred to them each day. Exposures started 8 to 10 July 2019 and ended 29 to 31 July 2019.

Embryo‐larval fathead minnow experiments

Animal handling and experimental procedures were approved by the Animal Care Committee, which oversees the Aquatic Life Research Facility in Burlington, Ontario, Canada (AUP 1948), operated under the approval of the Canadian Council of Animal Care. The toxicity of metformin to P. promelas was investigated using an early‐life stage assay similar to that described in Parrott et al. (2016). Eggs were purchased from Aquatox Laboratories. The eggs used in testing had been fertilized <18 h before the start of the exposures. Eggs from 21 separate breeding tiles across 3 different d were used to maximize genetic variability and diversity of the fish tested. Eggs were exposed to metformin in groups of 20 eggs per 1‐L glass beaker, with 800 mL of overlying water. The dilution water for preparing the metformin exposure beakers was from the City of Burlington. It was particle‐filtered, dechlorinated with activated carbon filtration, and sterilized using ultraviolet irradiation. Dilution water was maintained in a header tank prior to use in testing and had hardness of 125 ± 5 mg/L (mean ± standard deviation [SD]), alkalinity of 90.7 ± 4.6 mg/L, and pH of 7.20 ± 0.05. The full suite of measured physicochemical properties of the dilution water is presented in Supplemental Data, Table S1a. Nine replicate test vessels were prepared for each metformin exposure concentration, and 18 replicates were prepared for the water controls. Test vessels were aerated, covered with parafilm, and stored in an incubator at 25 ± 2 °C with 16 h light and 8 h dark during the experiment. Water quality parameters (pH, dissolved oxygen, temperature, conductivity, and ammonia) were measured weekly in test vessels during the experiment (Supplemental Data, Table S1b). The metformin exposure beakers were renewed daily by transferring eggs or hatched larvae to a newly prepared test vessel (made fresh daily and allowed to warm to 25 °C in an incubator). Eggs and larvae were kept in an egg cup (a glass cylinder with a bottom made of fine synthetic nylon mesh) within the 1‐L glass test vessel to facilitate the daily transfer to freshly prepared test vessels. Egg cups were rinsed at hatch on day 5, and the number of larvae was reduced to 10 on day 14 of the exposure (i.e., 9 dph). Exposure ended at 16 dph, which was typically at day 21 of the test.

Embryos and larvae were inspected daily for deformities or mortality. Immobilized or deformed larvae were removed, the deformities recorded, and the fish euthanized. Each day, larvae were fed a suspension containing newly hatched brine shrimp. Feeding was 10 µL/larva/d (each microliter containing ~15 nauplii) in the first week of the larval stage (0–8 dph) and 20 µL/larva/d in the second week (9–16 dph). Half of the food allotment was fed 2 h before solution changes and the remaining half following metformin solution change. At day 21 of the test (16 dph), surviving larvae were euthanized via immersion in tricaine methane sulfonate solution (MS222; 500 mg/L).

Endpoints quantified during the test included hatchability (total number of larvae hatched/total number of eggs), hatch success (total number of viable larvae hatched/total number of eggs), deformities at hatch, time to hatch, survival of eggs to 9 dph, survival of larvae from 9 to 16 dph, total survival, growth (total wet wt/surviving larvae), and biomass production (total wet wt/total number of larvae). Embryonic and larval fish were also scored using a severity scale for deformities (i.e., normal, mild, moderate, severe). Deformities assessed at hatch were behavior assessment (active swimmer, spiral swimmer, twitches, moribund), changes in edemas or circulation (pericardial edema, yolk edema, bubbles under skin, hemorrhage, other—e.g., tube heart), craniofacial abnormalities (small face, eye edema, other—e.g., small eyes or fused jaw), changes in body shape or fin morphology (helix coil, lordosis [belly out], kyphosis [belly in], scoliosis, bent tail fin, necrotic [but alive], or other).

Water samples were taken on days 3 and 4 (egg stage), days 10 and 11 (early larval stage), and days 17 and 18 (late larval stage). Pre‐exposure water samples were taken before eggs or larvae were transferred into the vessel at the beginning of a 24‐h period of exposure. After the eggs or larvae were removed at the end of the 24‐h period of exposure, postexposure water samples were taken. Water samples were taken before and after the addition of eggs and larvae to characterize any change in concentration of metformin in overlying water over the 24‐h period. Water samples were filtered (0.7 µm) and stored at 4 °C before analysis to determine the concentration of metformin. Analysis of metformin occurred within 5 d of sampling.

Analysis of metformin

Fathead minnow exposure solutions were assessed for concentrations of metformin and guanylurea. Guanylurea is the major bacterial breakdown product of metformin, occurring mainly in sewage‐treatment plants (Trautwein and Kümmerer 2011; Trautwein et al. 2014). Metformin hydrochloride, guanylurea phosphate, and ammonium formate were obtained from Sigma‐Aldrich; and the isotope‐labeled internal standards, metformin‐D6 hydrochloride and guanylurea 15N4 hydrochloride, were obtained from Toronto Research Chemicals. Methanol and water were purchased from Fisher Scientific and were Optima liquid chromatography‐mass spectrpmetry (LC‐MS)‐grade. Suprapur formic acid (98–100%) was obtained from EMD Millipore.

Samples of 20 mL exposure water were taken from each tank and filtered (to remove particles and fish waste) using a sterile syringe (Henk Sass Wolf Norm‐ject, 4200‐X00V0 20 mL) with a glass microfiber disposable syringe filter (Whatman; pore size 0.7 µm, 6890‐250725 mm GD/X). Filtered samples were stored in 20‐mL precleaned and prefired glass tubes. An aliquot of 180 µL prefiltered aquaria water was diluted with 20 µL of either a diluted labeled solution of dimethyl D6 metformin hydrochloride or 15N4 guanylurea hydrochloride. The resulting sample was vortexed, and 2 µL was directly injected onto the LC‐MS‐MS.

The LC‐MS‐MS analyses were performed using a Waters Xevo TQS mass spectrometer coupled to an Acquity ultra‐performance liquid chromatography system. The mass spectrometer was operated using an electrospray (Zspray) ionization source in positive ion and multiple reaction monitoring mode. An Agilent Eclipse Plus C18 RRD 1.8 µm, 2.1 × 100 mm analytical column was used for the analysis of metformin and guanylurea. The mobile phase consisted of water and methanol both containing 2 mM ammonium formate and 0.1% formic acid. Column temperature was maintained at 25 °C. The LC gradient conditions are presented in Supplemental Data, Table S2a. The precursor to product transitions monitored for native and labeled metformin and guanylurea are presented in Supplemental Data, Table S2b. Two ions were chosen so that a ratio between the 2 could be used as a further confirmation of compound identity (Supplemental Data, Table S2b).

All standards and working solutions for both analytes were prepared in initial mobile‐phase (water:methanol [90:10] gradient with 2 mM ammonium formate and 0.1% formic acid). Calibration plots containing 13 points and 7 points were generated for metformin and guanylurea, respectively (Supplemental Data, Figure S1).

Quantifications were performed using external calibration. Concentration range R 2 values were 0.98 or better. Labeled standards were used as performance standards to monitor instrument stability. Waters Targetlynx software was used to integrate chromatograms, generate calibration curves, and calculate sample concentrations.

The limit of detection (LOD) for metformin was 0.0186 µg/L, which was calculated as 3 times the SD of 11 blank water samples. Samples of fish exposure water that were below the LOD for metformin were assigned a value of one‐half the LOD for calculation of mean metformin concentrations. This occurred on 4 occasions for the control exposure water only. For the 3 metformin treatments, metformin concentrations were always above the LOD. For guanylurea the LOD was 1.3 g/L.

Analysis of the larval microbiome

Fathead minnow larvae were sampled at 16 dph using MS222, with the exception of 20 larvae which were euthanized by spinal severance to understand the potential effects of MS222 on the larval microbiome. The gut was isolated by removing the larva's head and tail and dipping the body in 70% ethanol (EtOH), then in nanopore water, to remove external bacteria. Although the subsample of the larvae included both the intestine (with its transient and indigenous bacteria) and somatic tissues (potentially with some external bacteria remaining), we assumed most of the bacteria DNA would be from the gut and hereafter refer to the results as the “gut microbiome.” Larvae were sampled from 6 replicates of control beakers and 3 replicates of each metformin concentration. Each beaker contained 10 larvae that were pooled in groups of 3 to 4 into tubes containing 900 μL of buffer, prepared as in Restivo et al. (2020) and inverted. Samples were held at room temperature for approximately 3 h and then stored at –20 °C until genomic DNA (gDNA) was extracted. All tools were sterilized between each larva with 30% bleach, 70% EtOH, and nanopore water. Next, gDNA was extracted, nested polymerase chain reaction amplification of the 16 S rRNA V3 to V4 region was performed, and read processing was conducted as in Restivo et al. (2020). The sequencing data (FASTq) are publicly available. Microbiome sequences have been uploaded to the National Center for Biotechnology Information (project code PRJNA661307).

Statistical analysis

All data for larval survival, deformities, and growth were analyzed using Systat, Ver 11.0 (Systat Software). Data were checked for normality using the Shapiro‐Wilk normality test. In 100% of the cases, the growth data at 9 and 16 dph (weight, length, condition factor, and tail length) were distributed normally, so we proceeded with analysis of variance (ANOVA) using untransformed data. Significant differences from control treatments were assessed using 2‐sample t tests (separate variances, Bonferroni's adjusted probabilities) to determine levels of significance. All p values for these tests are shown in the text, figures, and tables. Where ANOVA p values or Tukey's p values are used, they are indicated in the text.

For the assessment of survival data from the metformin exposures of fathead minnow (time to hatch, hatchability, hatch success, deformities at hatch, and survival from the egg stage to 16 dph), the data were not normal (only 8.3% of these data were normal as determined by the Shapiro‐Wilk normality test), so nonparametric statistical tests were used. To assess for differences among metformin treatments and controls, Kruskal‐Wallis one‐way ANOVA assessed whether there were differences among the ranked hatching, time‐to‐hatch, deformity, or survival data. The p values in the text are from chi‐squared distributions of the data with 3 degrees of freedom.

Data from the analysis of fathead minnow microbiome were assessed statistically using R (R Development Core Team 2019). The ASVs were analyzed using Phyloseq (Ver 1.30.0) in R (Ver 3.6.1). The ASVs are high‐resolution, single DNA sequences that are used to classify species. To analyze gut bacterial composition and diversity, larval microbiome samples were grouped by treatment: control (MS222, n = 30 samples), control (spinal severance, n = 6 samples), 3.4 µg/L metformin (n = 15 samples), 33.6 µg/L metformin (n = 21 samples), and 269 µg/L metformin (n = 18 samples). The relative abundance of bacteria was calculated at the phyla, family, and genus levels to understand changes in gut bacterial composition. Alpha diversity was calculated using various measures: Shannon diversity, evenness and richness of ASVs; observed species, ASV richness; Chao1, ASV richness independent of read depth. Statistical significance was determined using a one‐way ANOVA between treatments and pairwise differences were determined using Tukey's post hoc tests. Finally, beta diversity using a Bray‐Curtis principal coordinate analysis (PCoA) was conducted to discern differences in bacterial communities between treatments. Using the Adonis function in the package vegan (Ver 2.5.6), statistical significance was determined using a permutational multivariate analysis of variance (PERMANOVA), and a pairwise PERMANOVA was conducted to understand differences between pairs of treatments.

RESULTS

Measured metformin in fish‐exposure solutions

Measured concentrations of metformin in fish exposure beakers were very close to nominal concentrations. Nominal concentrations were 3.12, 31, and 312 µg/L; and mean measured concentrations (averaging the pre‐and post‐exposure solutions) were 3.44, 33.6, and 269 µg/L (Table 1). Measured concentrations ranged from 86 to 111% of nominal metformin concentrations (mean 102 ± 14% SD). There was very little metformin detected in the control beakers, with a mean concentration of 0.02 µg/L metformin over the 3‐wk exposure period. The full data set of measured metformin concentrations in exposure beakers is shown in Supplemental Data, Table S3.

Table 1

Mean measured concentrations of metformin (and standard deviation [SD]) in fish exposure beakersa

Nominal concentration of metformin (µg/L)	n	Preexposure concentration of metformin (mean)	SD	Postexposure concentration of metformin (mean)	SD	Mean metformin exposure concentration (µg/L)	SD
0	6	0.0125	0.0078	0.0283	0.0215	0.020	0.017
3	3	3.42	0.25	3.46	0.25	3.44	0.23
31	3	32.6	0.9	34.6	1.7	33.6	1.6
312	3	269	15	269	8	269	11

The concentration of metformin was measured 6 to 12 times in each treatment over the 21‐d embryo‐larval exposure to metformin; concentrations were measured 3 times for each preexposure solution and 3 times for each postexposure solution for the 3 metformin exposure concentrations and 6 times for each preexposure solution and 6 times for each postexposure solution for water controls.

Concentrations of metformin were stable over the 24‐h time between solution changes. Concentrations of metformin in preexposure beakers and postexposure beakers were similar (p = 0.338, one‐sample t test of percent change pre‐ vs postexposure measured metformin concentrations = 0, Bonferroni's adjusted p value), and there was no detectable loss of metformin during the time between exposure solution changes. There was no guanylurea (the major bacterial breakdown product of metformin [Trautwein and Kümmerer 2011]) detected in any samples of exposure beaker water.

Water quality parameters measured weekly in the 21‐d exposures were similar among controls and metformin treatments. Averages ± SDs were as follows: temperature 24.7 ± 0.3 °C, pH 8.40 ± 0.06, dissolved oxygen 8.01 ± 0.05 mg/L, conductivity 364 ± 9 µS/cm, and free ammonia 0.00 ± 0.00 mg/L. Full water quality parameters for each replicate beaker over time are in Supplemental Data, Table S1b. Water quality parameters for dilution water (ions, sulfates, nitrates, dissolved organic carbon, dissolved inorganic carbon) are shown in Supplemental Data, Table S1a.

Hatch success and deformities

There were no effects of metformin on egg hatching success or time to hatch (Table 2). Eggs exposed to metformin hatched at similar times as control water‐exposed eggs, in approximately 4.2 d (Kruskal‐Wallis one‐way ANOVA p ≥ 0.923). Hatchability (percentage of eggs that hatched) and hatch success (percentage of eggs that hatched into viable fry) were similar between metformin‐exposed embryos and control embryos (Kruskal‐Wallis one‐way ANOVA p ≥ 0.255 for hatchability and p ≥ 0.639 for hatch success). Metformin did not affect the percentage of deformities in newly hatched fry. Mean deformity rates were low (3.9–4.5%) and similar among control and metformin‐exposed embryos and fry (Kruskal‐Wallis one‐way ANOVA p ≥ 0.958).

Table 2

Time to hatch, deformities at hatch, hatching success, and survival of embryo‐larval fathead minnows (until 9 and 16 d posthatch) exposed to metformin measured concentrations for up to 21 da

Metformin (µg/L)	n	Hatchability (%)	SD	Hatch success (%)	SD	Survival from egg to 9 dph (%)	SD
9 dph
Controls	18	99.2	1.9	95.0	4.2	94.7	4.0
3.4	9	97.8	3.6	94.4	6.3	93.3	6.6
33.6	9	99.4	1.7	97.2	2.6	96.1	2.2
269	9	100.0	0.0	96.1	3.3	96.1	3.3
16 dph
Controls	18	94.7	4.0	4.2	4.3	4.18	0.18
3.4	9	93.3	6.6	4.0	5.5	4.22	0.29
33.6	9	96.1	2.2	4.5	3.0	4.20	0.13
269	9	96.1	3.3	3.9	3.3	4.23	0.23

Data are treatment means (and standard deviation) for n exposure beakers, and each beaker started with 20 fathead minnow eggs. Hatchability is the percentage of eggs that hatched. Hatch success is the percentage of eggs that hatched into viable fry.

dph = days posthatch; SD = standard deviation.

Larval survival and growth

Larval fish survival until 9 or 16 dph was unaffected by metformin exposures (Kruskal‐Wallis one‐way ANOVA p ≥ 0.577; Figure 1). Overall, from the fertilized egg stage until 16 dph, survival was very good (94.4% for controls and 93.3–96.1% for the metformin treatments; Table 2). Survival of all larval fish between 9 and 16 dph was 100% in controls and in all metformin treatments. The full data sets for time to hatch, deformities at hatch, egg hatch, and larval survival parameters are in Supplemental Data, Table S4.

Figure 1

Survival until 16 d posthatch of fathead minnow larvae exposed to water control (blue circles) or to 3 metformin concentrations (3.4, 33.6, or 269 µg/L) from the fertilized egg stage. Mean survival is shown by black rectangles connected by a dotted line, with error bars showing standard deviation. Note that the y‐axis begins at 80% survival. Individual data points are jittered on the x‐axis to show overlapping values. dph = days posthatch.

There were no significant effects of metformin exposure on larval growth at 9 or 16 dph (Figure 2). Larval fish mean wet weights were approximately 4.6 to 4.7 mg at 9 dph and 15.3 to 16.3 mg at 16 dph (Table 3). There were no differences in mean wet weights among controls and metformin‐exposed fish (ANOVA p = 0.882 for 9 dph and 0.364 for 16 dph). Mean lengths of fish were also unaffected by metformin exposure; total lengths of fish were 9.8 to 10.0 mm at 9 dph and 14.0 to 14.1 mm at 16 dph (ANOVA p = 0.513 for 9 dph and p = 0.687 for 16 dph). There were observations of slightly higher weights and lengths of fish from the 33.6 µg/L metformin treatments (Table 3), but these were not significantly different from controls (2‐sample t test Bonferroni's adjusted p value with separate variances, p = 0.054 for wet wt and p = 0.236 for total length).

Figure 2

Wet weight versus total length at 9 and 16 d posthatch (dph) of fathead minnow larvae exposed to water control (blue circles) or to 3 metformin concentrations (3.4, 33.6, or 269 µg/L) from the fertilized egg stage. The top panel is 9 dph and the bottom panel is 16 dph. Each point on the figure is a replicate beaker mean calculated from 7 to 10 larval fish at 9 dph and 9 to 10 larval fish at 16 dph.

Table 3

Wet weight, total length, condition factor, and tail length of embryo‐larval fathead minnows (sampled at 9 and 16 d posthatch) exposed to metformin measured concentrations for up to 21 da

Metformin (µg/L)	n	Wet wt (mg)	SD	Total length (mm)	SD	CF	SD	Tail length (mm)	SD
9 dph
Controls	18	4.61	0.50	9.89	0.30	0.463	0.027	1.20	0.08
3.4	9	4.69	0.36	9.91	0.18	0.472	0.025	1.22	0.07
33.6	9	4.55	0.69	9.78	0.38	0.474	0.032	1.18	0.10
269	9	4.74	0.60	10.01	0.38	0.466	0.027	1.22	0.10
16 dph
Controls	16	15.34	1.08	13.96	0.29	0.551	0.017	2.48	0.07
3.4	9	15.74	1.60	14.04	0.43	0.555	0.022	2.49	0.12
33.6	9	16.27	1.07	14.12	0.33	0.568	0.021	2.52	0.07
269	9	15.69	1.25	13.98	0.32	0.562	0.014	2.47	0.09

Data are treatment means (and standard deviation) for n exposure beakers, and each beaker started with 20 fathead minnow eggs.

CF = condition factor; dph = days posthatch.

Metformin did not affect condition factors of fish at 9 or 16 dph (ANOVA p = 0.758 and 0.156 for 9 and 16 dph, respectively; Table 3). The condition factor of fish from the 33.6 µg/L metformin treatment at 16 dph (condition factor = 0.568) was higher than control condition factor (condition factor = 0.551), but this difference was not significant (p = 0.055, 2‐sample t test Bonferroni's adjusted p value with separate variances).

There were no differences in mean tail length among controls and metformin‐exposed fish (ANOVA p = 0.833 for 9 dph and 0.568 for 16 dph). Tail lengths were approximately 1.2 mm at 9 dph and 2.5 mm at 16 dph for all fish in the experiment, regardless of treatment. Fish exposed to 33.6 µg/L metformin and sampled at 16 dph had slightly longer tails (2.52 mm) than control fish (2.48 mm), but this difference was not significant (p = 0.143, 2‐sample t test Bonferroni's adjusted p value with separate variances). The full data set for all of the growth parameters at 9 and 16 dph is in Supplemental Data, Table S5.

Microbiome

There was a sum of 6 645 967 reads across all larval gut samples, with an average of 73 844 reads per sample (minimum of 38 452 and maximum of 106 206). In total there were 899 ASVs detected across larval samples; there was a minimum of 22, a maximum of 105, and a mean of 51 ASVs per sample. There were 16 phyla, 28 classes, 63 orders, 100 families, and 225 genera of bacteria detected across samples. Generally, the larvae were dominated by the bacterial phyla Proteobacteria (335 ASVs) and Firmicutes (312 ASVs), with a lower abundance of Bacteroidetes (120 ASVs). At the family level, larvae were dominated by Lachnospiraceae (phylum Firmicutes; 80 ASVs) and Ruminococcaceae (phylum Firmicutes; 62 ASVs). The top 5 genera were Bacteroides (phylum Bacteroidetes; 40 ASVs), Streptococcus (phylum Firmicutes; 20 ASVs), Legionella (phylum Proteobacteria; 18 ASVs), Staphylococcus (phylum Firmicutes; 17 ASVs), and Bacillus (phylum Firmicutes; 14 ASVs).

There was an average of 13 to 33 ASVs found in larval guts among the treatments (Table 4). Proteobacteria dominated in all treatments except for the MS222 controls, which were dominated by Firmicutes. Burkholderiaceae was dominant in the spinal controls and in 3.4, 33.6, and 269 µg/L metformin‐exposed fish, whereas Ruminococcaceae and Lachnospiraceae dominated the MS222 controls and 33.6 µg/L groups, respectively. Bacteroides was the top genus present in all treatment groups except the 3.4 µg/L exposed larvae, which were most abundant in Legionella.

Table 4

Total number and mean of amplicon sequence variants across samples and top phylum, family, and genus by treatment group including number of measures

Treatment (no. samples)	Total ASVs across all samples	Mean ASVs in larvae among treatments	Top phylum (ASVs)	Top family (ASVs)	Top genus (ASVs)
Controls–spinal (n = 6)	196	33	Proteobacteria (108)	Burkholderiaceae (28)	Bacteroides (7)
Controls–MS222 (n = 30)	396	13	Firmicutes (154)	Ruminococcaceae (40)	Bacteroides (15)
3.4 µg/L (n = 15)	320	21	Proteobacteria (131)	Burkholderiaceae (26)	Legionella (14)
33.6 µg/L (n = 21)	421	20	Proteobacteria (177)	Lachnospiraceae (31), Burkholderiaceae (31)	Bacteroides (14)
269 µg/L (n = 18)	276	15	Proteobacteria (131)	Burkolderiaceae (27)	Bacteroides (13)

ASV = amplicon sequence variant; MS222 = tricaine methane sulfonate.

The relative abundance of bacteria was calculated and takes into account number of reads per given ASV; for this reason, there are differences between number of ASVs present and relative abundance. The larval gut was dominant in Proteobacteria and Firmicutes; Proteobacteria increased and Firmicutes decreased with increased metformin exposure (Figure 3). All larvae were dominated by Family_XII and Burkholderiaceae (Figure 4). At the genus level, Exiguobacterium dominated in all treatments, and there was a small increase in the genera Acinetobacter and Enterobacter in the gut microbiome of larvae exposed to 33.6 µg/L of metformin (Figure 5).

Figure 3

Relative abundance of the 2 most abundant phyla (Firmicutes, red; Proteobacteria, blue) in fathead minnow larvae gut by treatment (from left to right: controls–spinal, controls–tricaine methane sulfonate, 3.4 µg/L of metformin, 33.6 µg/L of metformin, and 269 µg/L of metformin). Upper, middle, and lower lines represent first, second, and third quartiles, respectively; whiskers represent a 1.5 interquartile range. MS222 = tricaine methane sulfonate.

Figure 4

Relative abundance of the 10 most abundant bacterial families found in the gut of fathead minnow larvae by treatment (from left to right: controls–spinal, controls–tricaine methane sulfonate, 3.4 µg/L of metformin, 33.6 µg/L of metformin, and 269 µg/L of metformin). MS222 = tricaine methane sulfonate.

Figure 5

Relative abundance of the 10 most abundant bacterial genera found in the gut of fathead minnow larvae by treatment (from left to right: controls–spinal, controls–tricaine methane sulfonate, 3.4 µg/L of metformin, 33.6 µg/L of metformin, and 269 µg/L of metformin). MS222 = tricaine methane sulfonate.

In general, there were no changes in alpha diversity with increased metformin exposure, but there was increased diversity of the gut bacteria in unexposed larvae euthanized using spinal severance compared to those euthanized with MS222 (Figure 6). There was no significant difference in Shannon alpha diversity between treatment groups (one‐way ANOVA, df = 4, F statistic = 1.727, p = 0.152). There was a difference in observed species between treatment groups (one‐way ANOVA, df = 4, F statistic = 3.175, p = 0.0176); Tukey's post hoc test indicated a significant difference between spinal and MS222 controls (p = 0.03). There was also a difference in Chao1 between treatments (one‐way ANOVA, df = 4, F statistic = 3.193, p = 0.0171), particularly between spinal and MS222 controls (Tukey's honestly significant difference, p = 0.03). It is likely that significant differences observed in Chao1 and observed species indices were driven by the spinal and MS222 control groups.

Figure 6

Alpha diversity metrics (observed, Chao1, and Shannon) of bacterial species in the gut of larval fathead minnow by treatment: controls–spinal, controls–tricaine methane sulfonate, 3.4 µg/L of metformin, 33.6 µg/L of metformin, and 269 µg/L of metformin). MS222 = tricaine methane sulfonate.

Beta diversity was determined using a Bray‐Curtis PCoA; axes 1 and 2 accounted for 71.8% of the variation explained (Figure 7). There were no significant dissimilarities between treatments (PERMANOVA, df = 4, F statistic = 1.3185, p = 0.196). There was one pairwise difference between the spinal control and 3.4 µg/L exposed larvae, indicating distinct microbial communities (pairwise PERMANOVA, df = 1, F statistic = 2.4409, p = 0.047).

Figure 7

Beta diversity of fathead minnow larvae exposed to various concentrations of metformin calculated using the Bray‐Curtis distance metric with principal coordinate analysis by treatment: controls–spinal (circle), controls–tricaine methane sulfonate (open circle), 3.4 µg/L of metformin (black triangle), 33.6 µg/L of metformin (white triangle), and 269 µg/L of metformin (down triangle). Square brackets on axes 1 and 2 denote percentage of variation explained by treatment. MS222 = tricaine methane sulfonate.

DISCUSSION

Fathead minnow embryos and larvae were exposed under daily static‐renewal conditions to the antidiabetic drug metformin for 21 d to assess effects on hatching, deformities, survival, growth of fish, and the gut microbiome. Exposures were well replicated (with 9–18 replicate beakers containing 20 embryos/larvae each) to detect any small changes in growth of larval fish.

Measured concentrations of metformin in exposure solutions were 102 ± 14% of nominal concentrations (mean ± SD), and there was no difference in concentrations of metformin measured in preexposure beakers and postexposure beakers (assessed after 24‐h exposure time between solution changes). There was no detected guanylurea (the major bacterial breakdown product of metformin [Trautwein and Kümmerer 2011]) in any metformin pre‐ or postexposure beaker.

Replication in our metformin embryo‐larval exposure was high, to detect small changes in fish weight or length due to metformin exposure. With such high replication, we had power of 0.8 to detect an 11.1% change in fish weight and a 3.3% change in fish length (at 16 dph) with p ≤ 0.05. Previous studies assessing metformin's effects on medaka growth also used large numbers of larval fish to detect a negative effect (8–40 larval fish sampled per replicate and 4 replicate tanks per metformin exposure concentration; Ussery et al. 2018).

There were no significant effects of metformin on any parameter assessed in embryo‐larval fathead minnows. Hatching success and deformities at hatch were no different from controls. There were also no effects of metformin on survival and growth of larval fathead minnows up to 16 dph (day 21 of the exposure). Survival (egg–16 dph) and mean growth parameters (at 16 dph) were similar to those of all control fish (from this 21‐d test) in our lab that have been performed since 2014. Average (± SD) survival for controls (n = 180 beakers over 6 yr of tests performed 2014–2019) was 94.1 ± 6.3%, and average growth at 16 dph was weight 15.6 ± 2.0 mg, length 13.6 ± 0.6 mm, condition factor 0.61 ± 0.4, and tail length 2.4 ± 0.2 mm. The mean values for controls in the metformin test are within 4% of the overall 6‐yr control means for survival, weight, length, and tail length and within 10% for condition factor. The mean values for all control and treatment beakers in the metformin test are within 4% of the overall 6‐yr control means for survival, weight, length, and tail length and within 8.5% for condition factor.

The lack of growth changes in the current metformin exposures was different from one other study with embryo‐larval fish exposed to metformin at similar concentrations; that study showed significant dose‐related decreases in growth. Ussery et al. (2018) found that embryo‐larval medaka exposed to 1 to 100 µg/L metformin had decreased lengths and weights when they were 14 and 28 d old. The reason for the discrepancy may be differences in species sensitivity; however, other studies assessing growth in the fathead minnow and Japanese medaka have shown they are similarly sensitive when comparing median lethal concentrations (LC50s). Fathead minnow and Japanese medaka exposed to several concentrations of the synthetic androgen spironolactone both showed growth increases after exposure to 50 µg/L (Lalone et al. 2013). Medaka weight decreases were observed after 28‐d larval exposures to 8 organic chemicals (allyl isothiocyanate, aniline, benzyl acetate, 1,2‐dibromoethane, 4‐chloroaniline, 2,4‐diaminotoluene, 2,4‐dichlorophenoxyacetic acid, and phenol; Holcombe et al. 1995). But overall, for the organic chemicals tested, the authors concluded that the acute and chronic sensitivities of medaka were similar to those of fathead minnows when the LC50s were compared between the 2 species (Holcombe et al. 1995).

Similar to many freshwater fish, the gastrointestinal microbiome of fathead minnow larvae was dominated by the phyla Proteobacteria and Firmicutes. Generally, the gut of freshwater fish is abundant in the bacterial phyla Proteobacteria, Firmicutes, and Fusobacteria (Eichmiller et al. 2016; Tarnecki et al. 2017; Turner and Bucking 2019; Restivo et al. 2020), including larvae (Jacob et al. 2018). Assessment of the 20 most common families, genera, and species of bacteria in control fathead minnows showed that 70% were Proteobacteria, 10% Bacteriodetes, and 10% Fusobacteria (Narrowe et al. 2015). In another study, fathead minnow controls were dominated by the phyla Proteobacteria and Fusobacteria (Bridges et al. 2018; DeBofsky et al. 2020). The most abundant families vary between studies of fathead minnows: Aeromonadaceae (phylum Proteobacteria) and Shewanallaceae (phylum Proteobacteria) were dominant in Narrowe et al. (2015), whereas Burkholderiaceae (phylum Proteobacteria) and Fusobacteriaceae (phylum Fusobacteria) were dominant in DeBofsky et al. (2020).

Interestingly, in the present study there were small increases in Proteobacteria in the larvae exposed to increasing concentrations of metformin, similar to observations in brown trout larvae exposed to the same contaminant (Jacob et al. 2018). Increases in Proteobacteria have been linked to altered health outcomes in mammals (Shin et al. 2015; Méndez‐Salazar et al. 2018), and it is possible that fish may have similar outcomes (Restivo et al. 2020). Furthermore, Proteobacteria may play a role in the transformation of xenobiotics (Reis et al. 2018), although this has not yet been studied in fish.

There was no effect of metformin exposure on the community (beta diversity) or alpha diversity of the fathead minnow larvae microbiome examined in the present study, and these results contrast with other studies on metformin and other waterborne contaminants. Brown trout larvae exposed to 1 and 1000 µg/L of metformin had an increase in bacterial alpha diversity of their gut, whereas those exposed to 10 and 100 µg/L of metformin had decreased alpha diversity (observed species, Chao1, and Shannon) compared to controls; however, similar to our study these changes were not significant (Rogall et al. 2020). Beta diversity of larval zebrafish was altered after 20‐d exposure to arsenic (larval microbiome; Dahan et al. 2018) and after 7 d of exposure to triclosan (gut microbiome [Narrowe et al. 2015]); communities clustered by low and high exposure in the latter. Zebrafish larvae exposed to arsenic or triclosan had increased alpha diversity compared to controls (observed species [Dahan et al. 2018]; Shannon diversity [Narrowe et al. 2015]). It is possible that differences in the methods used in the aforementioned studies contributed to differences in alpha diversity results. For example, brown trout larvae were 156 d old and exposed to metformin for 108 d (Rogall et al. 2020), zebrafish larvae exposed to arsenic were sampled at 20 dpf, and zebrafish larvae exposed to triclosan were sampled 9 wk posthatch (Narrowe et al. 2015), whereas larvae in the present study were sampled at 16 dph. Alpha diversity of the microbiome differs with age in grass carp (Li et al. 2017); thus, it is possible that age differences may be a factor in the variable results among studies.

Although there were no changes in the microbiome of fathead minnow exposed to metformin in the laboratory, field studies are required to understand the effects of contaminant mixtures (downstream from MWWE outfalls) under natural conditions. In addition, there is uncertainty in the effects of natural diets and environmental stressors, which may influence the effect of metformin (and other chemicals) on the fish microbiome.

Interestingly, there was a difference in the alpha diversity (richness measures: observed species and Chao1) of larvae euthanized using spinal severance or MS222. At present, studies examining the fish gut microbiome do not have a standard for euthanasia and, therefore, various methods are used. As examples, others examining the larval microbiome have used spinal severance (Narrowe et al. 2015), MS222 (Dahan et al. 2018), or both (Jacob et al. 2018; Rogall et al. 2020). Further research into the effects of MS222 on the microbiome of larvae may be necessary.

CONCLUSIONS

There were no significant changes in survival, deformities, or growth in embryo‐larval fathead minnows exposed to metformin for 21 d, suggesting that this life stage of fathead minnow was not affected by this pharmaceutical at environmentally relevant concentrations. Compared to control fish, there were no large changes in the gastrointestinal bacterial community of fathead minnow larvae exposed to metformin. There was a small increase in Proteobacteria and a decrease in Firmicutes with increased exposure to metformin, which may be an indicator of exposure to environmental contaminants. The lack of responses of the larval microbiome to metformin suggests that such compounds are unlikely to be linked to the changes in microbial community composition and diversity observed in the gut of wild fish living downstream of MWWE outfalls. Furthermore, there were changes in the alpha diversity of control larvae euthanized using MS222 compared to spinal severance, indicating that the use of MS222 may impact the gut microbial community of fish sampled in the laboratory and warrants further consideration.

Supplemental Data

The Supplemental Data are available on the Wiley Online Library at https://doi.org/10.1002/etc.5054.

Disclaimer

The authors declare that they have no conflicts of interest related to the reporting of these results.

Author Contributions Statement

J.L. Parrott and K.A. Kidd conceived and designed the experiments; V.E. Restivo, J. Zhu, K. Shires, S. Clarence, H. Khan, and C. Sullivan performed the experiments; J.L. Parrott and V.E. Restivo performed statistical analysis; G. Pacepavicius and M. Alaee performed chemical analysis; J.L. Parrott, V.E. Restivo, and K.A. Kidd wrote the manuscript; J. Zhu, K. Shires, S. Clarence, H. Khan, C. Sullivan, G. Pacepavicius, and M. Alaee provided technical and editorial assistance.

This article includes online‐only Supplemental Data.

Supporting information.

Click here for additional data file.