Microplastics have evolutionary and ecological impacts across species, affecting organisms’ development, reproduction, and behavior along with contributing to genotoxicity and stress. As plastic pollution is increasing and ubiquitous, gaining a better understanding of organismal responses to microplastics is necessary. Gene methylation is a heritable form of molecular regulation that is influenced by environmental conditions, including exposure to pollutants, therefore determining epigenetic responses to microplastics will reveal potential chronic consequences of this pollutant. We performed an experiment across two generations of fathead minnows (Pimephales promelas) to elucidate transgenerational effects of microplastic exposure. We exposed the first generation of fish to four different treatments of microplastics: two concentrations of each of pre-consumer polyethylene (PE) and PE collected from Lake Ontario. We then raised the second generation from these parents with no microplastic exposure. We used reduced-representation methylation sequencing on adult liver tissue and homogenized larvae to evaluate DNA methylation differences among treatments, sexes, and generations. Our findings show the origin of the plastic had a larger effect in female minnows whereas the effect of concentration was stronger in the males. We also observed transgenerational effects, highlighting a mechanism in which parents can pass on the effects of microplastic exposure to their offspring. Many of the differentially methylated genes found in our analyses are known to interact with estrogenic chemicals associated with plastic and are related to metabolism. This study highlights the persistent and potentially serious impacts of microplastic pollution on gene regulation in freshwater systems.

Experimental Design

We used fathead minnow, a toxicological model organism, throughout this experiment to determine the effects of microplastics across life stages. Fish maintenance and handling followed protocols approved by the Ontario Ministry of Environment, Conservation, and Parks (MECP) Laboratory Services Animal Care Committee (approval number: ATU-004-19). We adapted rearing methods from a previously published fathead minnow life cycle experiment (Parrot 2005). We obtained fathead minnows from the breeding stock of the MECP. From eggs through 178 DPH, we raised fish under experimental conditions with four treatments and a negative control (CTRL). The treatments were: 1) a low, environmentally relevant concentration (100 particles/L) of pre-consumer PE (PL), 2) high, future scenario concentration (2000 particles/L) of pre-consumer PE (PH), 3) low concentration of PE gathered from Lake Ontario (EL), and 4) a high concentration of PE originating in Lake Ontario (EH). We reduced microplastic size to 150-500 µm using a burr mill coffee grinder. For a full lifecycle assessment, we used 5 replicate tanks per treatment, for a total of 25 aquaria. We began with 30 fish per tank and destructively sampled to a final count of two male and three female minnows per tank. The experimental design is summarized in Figure 1. For more details see Bucci et al. (in review).

Each treatment received twice daily feedings. The initial diet consisted of newly hatched brine shrimp at a concentration of 15 nauphii/µL until 30 days post hatch (dph), followed by a gradual introduction of frozen brine shrimp into the diet until 50 dph, after which point the diet consisted solely of thawed brine shrimp. Throughout the experiment, we maintained the water bath at 25˚C and a daily light exposure of 16 hours. We completed 30% water changes three times a week and cleaned the aquaria to remove algae buildup and tested a random tank from each treatment for dissolved oxygen, pH, and conductivity to determine water quality. Additionally, tanks contained zeolite to control ammonia levels, and we checked ammonia levels in a sequential manner for each treatment during cleaning.

We encouraged mating by provided egg tiles in each tank. We obtained clutches of offspring from 23 of the 25 tanks as some tanks did not produce eggs or did not lay enough eggs. After the minnows laid eggs on the clay tiles located in each tank, we removed and counted the eggs. If more than 100 eggs were laid, we set 50 aside for the transgenerational portion of the study where we raised a second generation of minnows to 12 dph under negative control conditions (no microplastics) to isolate the transgenerational effect of parental exposure. We fed newly hatched brine shrimp to the larvae once daily starting on the second day following the hatching of the first individual. A complete water change was done every other day. The experiment concluded at 178 dph. For final processing, we euthanized fish using a lethal bath of buffered MS-222 at a concentration ≥ 250 mg/L. We left fish in the bath until two minutes after respiration ceased. We dissected liver tissue from one male and two females from each tank (n=75) and placed the tissue in 95% ethanol. Between fish, we cleaned dissection tools with 10% bleach, molecular-grade deionized water, and 90% ethanol. For the larvae, entire individuals were placed in ethanol following euthanasia.

Library preparation and quality filtration

We homogenized the liver tissue of individual adults (N=75) and pooled larvae from familial clutches (N=21) using a bead mill with steel beads prior to extraction. We a bead-based method to extract and purify genomic DNA (Ali et al., 2016). We used the NEBNext Enzymatic Methyl-seq kit (#E7120L) for enzymatic conversion of methylated nucleotides, individual barcoding, and library preparation (New England Biolabs, Inc, Ipswich, MA). The samples (N=96) were pooled into a single library and sequenced twice on the Novaseq6000 on one lane of a SP flowcell (PE 2x150) at the RTSF Genomics Core at Michigan State University.

Following sequencing, reads from both sequencing runs were trimmed using TrimGalore! (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) to remove adapter-contaminated sequences and the first ten base pairs from the 5’ and 3’ ends of the fragments. We converted the fathead minnow reference genome (GenBank accession GCA_016745375.1) (Burns et al., 2016 & Saari et al., 2017) using Bismark (Krueger & Andrews 2011) and aligned reads to the converted genome using Bowtie2 (Langmead & Salzberg 2012) within Bismark with a modified stringency setting of --score_min L,0,-0.6. Alignments were deduplicated from each library using deduplicate_bismark and samples from each sequencing run merged into a single file using SAMtools version 1.16.1 (Danecek et al., 2021). Prior to further analysis, we removed samples from the dataset with less than one million reads. We subsequently extracted methylation calls using bismark_methylation_extractor.

Relatedness determination

To determine relatedness among individuals, we used the merged and sorted bam files to extract SNPs using BCFtools mpileup (Danecek et al., 2021) piped to BCFtools call (Li, 2011). We quality filtered the resulting file using VCFtools v0.1.15 (Danecek et al., 2011), initially removing genotypes not represented in at least 50% of individuals and any SNPs with a count less than five. The resulting data was then further filtered to remove genotypes with fewer than three reads, genotypes with a call rate of less than 90% and SNPs with a minor allele frequency less than 0.05. We removed individuals with a missingness greater than 15% and then calculated pairwise φ among all samples using the relatedness2 estimator. Samples with pairwise φ ranges [0.177, 0.354], were denoted as first-degree relatives, which includes full siblings and parent-offspring pairs. Kinship coefficient ranges [0.0884, 0.177] and [0.0442, 0.0884] were considered 2nd-degree and 3rd-degree relationships, respectively (see https://www.kingrelatedness.com/manual.shtml).

Differential methylation analysis

We used the R packages bsseq (Hansen et al., 2012), dmrseq (Korthauer et al., 2017), DSS (Wu et al., 2013; 2015; Feng et al., 2014; Park and Wu, 2016), and MethylSig (Park et al., 2014) to quality filter our methylation data and determine regions of differential methylation. Briefly, we read in the coverage files from Bismark to a BS-seq object, first filtering out any loci with coverage of less than 10 and then filtering out any individuals with an average read count of less than 0.5. Our final filtration step involved filtering out any loci that did not occur in at least 90% of individuals from the parental generation (n= 58) and the F1 generation (n=21). We tested for differentially methylated loci (DMLs) and regions (DMRs) across the variables of and interactions within sex (female, male), exposure (treated, control), concentration (high, low, control), plastic type (environmental, pre-consumer, control), individual treatments (environmental high, environmental low, pre-consumer high, pre-consumer low, control), and generation (parental, F1). The stringency values for a locus/region to be called differentially methylated were a delta of 0.1 and false discovery rate (FDR) of 0.05. We used the NCBI Genome Data Viewer for the Pimephales promelas annotation release 100 to search for genes within the DMRs (or DMLs if there were not statistically significant DMRs).