Glyphosate-based herbicides influence DNA methylation patterns in Coturnix japonica

Abstract

Glyphosate-based herbicides (GBHs) are the most commonly used herbicide worldwide, yet also the most controversial. There is increasing evidence that GBHs can affect the epigenome of non-target organisms. However, studies on the cumulative effects of GBHs remain rare. Here, we analyzed genome-wide DNA methylation data from the first long-term subtoxic GBH exposure experiment. Japanese quails (Coturnix japonica) were experimentally exposed to dietary GBHs and sampled twice at 7 and 52 weeks old for Reduce Representation Bisulfite Sequencing.
We highlighted slight hypermethylation in GBHs exposed individuals and several differentially methylated sites in promoters.
One site in CCDN1 and nineteen in SPDYA (two oncogenes) were found to be hypomethylated. Hypomethylation in CCDN1 supports a recent study assessing that CCDN1 expression was repressed in response to GBHs exposure. Hypomethylation on SPDYA, a cell-cycle regulator involved in meiosis, suggests that GBHs could seriously affect reproduction.

Full text

MSc Biodiversity, Ecology, Evolution
EcoPhysiology and EcoToxicology
Master 2 – 2020-2021
Glyphosate-based herbicides influence DNA
methylation patterns in Coturnix japonica
SERVANT Euphrasie
Under the supervision of:
Suvi RUUSKANEN – adjunct professor
Heidi M. VIITANIEMI – postdoctoral researcher
University of Turku

M2 EPET 2020-2021 SERVANT Euphrasie
2
Table of Contents
ABSTRACT 3
INTRODUCTION 4
MATERIALS AND METHODS 6
STUDY DESIGN 6
BLOOD SAMPLING AND SAMPLE COLLECTION 6
REDUCED REPRESENTATION BISULFITE SEQUENCING 7
SEQUENCE ALIGNMENT 8
QUALITY CONTROL 8
TRIMMING 11
ALIGNMENT 11
METHYLATION CALL 11
DIFFERENTIAL METHYLATION ANALYSIS 11
PROMOTER REGION AND GENE ANNOTATION 12
RESULTS 13
DISCUSSION 16
CONCLUSION 18

M2 EPET 2020-2021 SERVANT Euphrasie
3
Abstract
English
Glyphosate-based herbicides (GBHs) are the most commonly used herbicide worldwide, yet
also the most controversial. There is increasing evidence that GBHs can affect the epigenome
of non-target organisms. However, studies on the cumulative effects of GBHs remain rare.
Here, we analyzed genome-wide DNA methylation data from the first long-term subtoxic GBH
exposure experiment. Japanese quails (Coturnix japonica) were experimentally exposed to
dietary GBHs and sampled twice at 7 and 52 weeks old for Reduce Representation Bisulfite
Sequencing.
We highlighted slight hypermethylation in GBHs exposed individuals and several differentially
methylated sites in promoters.
One site in CCDN1 and nineteen in SPDYA (two oncogenes) were found to be hypomethylated.
Hypomethylation in CCDN1 supports a recent study assessing that CCDN1 expression was
repressed in response to GBHs exposure. Hypomethylation on SPDYA, a cell-cycle regulator
involved in meiosis, suggests that GBHs could seriously affect reproduction.
French
Les herbicides à base de glyphosate (HBGs) sont les herbicides les plus largement utilisés dans
le monde, mais aussi les plus controversés. Il existe de plus en plus de preuves que les HBGs
peuvent affecter l'épigénome d’organismes non-cible. Cependant, les études sur les effets
cumulatifs des HBGs restent rares. Ici, nous analysons les données de méthylation de l'ADN à
l'échelle du génome provenant de la première expérience d'exposition long-terme subtoxique
à un HBG. Des cailles japonaises (Coturnix japonica) ont été expérimentalement exposées à
un régime contaminé et échantillonnées deux fois, à l'âge de 7 et 52 semaines pour un
séquençage du méthylome.
Nous avons mis en évidence une légère hyperméthylation chez les individus exposés aux
HBGs ainsi que plusieurs promoteurs présentant des sites différentiellement méthylés.
Plusieurs sites se sont révélés être hypométhylés dans les promoteurs de deux oncogènes,
SPDYA et CCDN1, des regulateurs du cycle-cellulaire. La présence d’une site hypométhylé chez
CCDN1 soutient une étude récente évaluant que son expression est réprimée en réponse à
l'exposition aux HBGs. L'hypométhylation de SPDYA, un régulateur du cycle cellulaire impliqué
dans la méiose, suggère que les HBGs pourraient affecter la reproduction.

M2 EPET 2020-2021 SERVANT Euphrasie
4
Introduction
For almost 50 years, Glyphosate-Based Herbicides (GBHs) have been extensively used for
agricultural purposes. As of today, it is the most commonly used herbicide, yet also the most
controversial (Benbrook, 2016).
Despite conflicting views of European agencies and rising concerns on the carcinogenicity of
GBHs and their toxicity, it has been reapproved for use in Europe until 2022 (Székács and
Darvas, 2018). However, towards the increasing evidence of glyphosate toxicity, several
countries have already imposed limitations on the use of GBHs.
As glyphosate or its degradation products may accumulate in soil or dissolve in water, the
ubiquitous presence of glyphosate in the environment is a major concern for non-target
organisms (Bai and Ogbourne, 2016).
Numerous studies are piling up about the potential effects of GBHs on organismal health
across many taxa and on a broad range of biological levels.
There is evidence that GBHs can increase oxidative stress resulting in DNA damage (Ghisi,
Oliveira and Prioli, 2016; Kwiatkowska et al., 2017; Tarazona et al., 2017), disrupts
neurotransmitters functions (Gallegos et al., 2018; Bali et al., 2019), act as an endocrine
disruptor (Romano et al., 2010; Gomez et al., 2019; Manservisi et al., 2019), or disrupts
microbial communities (Aitbali et al., 2018; Leino et al., 2020; Ruuskanen, Rainio, Gómez-
Gallego, et al., 2020a).
However, studies investigating the effects of glyphosate on birds are lacking despite their
potential exposure to glyphosate in fields (Bai and Ogbourne, 2016).
Recently, interest emerged for the potential effects of GBHs on the epigenome to explain
GBHs' mechanism of action on non-target organisms (Rossetti et al., 2021).
Epigenetics can be defined as the study of heritable phenotype changes that do not involve
alterations in the DNA sequence. It is a set of molecular processes that can modulate the
activity of a particular gene. These variations can arise from environmental conditions,
development, aging, drug, chemicals, or even diet, allowing a new type of phenotypic
plasticity (Dupont, Armant and Brenner, 2009).
As epigenetic changes can be inherited, it challenges the usual conception that the DNA
sequence is the only source of inheritance (Bossdorf, Richards and Pigliucci, 2007).
Chromatin condensation, histones modification, and micro RNAs interference are all
epigenetics mechanisms, but DNA methylation is the most studied one (Dupont, Armant and
Brenner, 2009). It is the adding of a methyl group (-CH3) to a nucleotide; usually, a cytosine,
most often followed by a guanine (a CpG site).
When located in a promoter, methylated CpGs can repress gene transcription, mainly by
preventing transcription factors binding. Conversely, unmethylation will allow transcription
(Dupont, Armant and Brenner, 2009).
DNA methylation offers an opportunity for an adaptative response to the current
environment; it gives new perspectives to understand ecological processes and evolution
(Bossdorf, Richards and Pigliucci, 2007; Dupont, Armant and Brenner, 2009).
It is now known that many human diseases, including cancers, can be explained by DNA
methylation (Greenberg and Bourc’his, 2019), and interest in that matter is growing.
Few studies have investigated DNA methylation in relation to glyphosate exposure, especially
on a whole genome-scale.
Nonetheless, glyphosate has been shown to affect methylation in the promoter regions of
selected tumor suppressors in human cells (Kwiatkowska et al., 2017; Woźniak et al., 2020),

M2 EPET 2020-2021 SERVANT Euphrasie
5
altered genes regulating epigenetic process in fish (Smith, Vera and Bhandari, 2019), decrease
global methylation in human peripheral blood mononuclear cells (Kwiatkowska et al., 2017),
or altered methylation pattern in estrogen receptors genes and sperm cells in rats (Gomez et
al., 2019).
Most of these studies target acute doses of glyphosate and short-term exposure, which is not
comparable to what organisms can experience in the wild.
To unveil the cumulative effects and chronic exposure of GBHs on a bird species, Ruuskanen
et al. conducted a long-term exposure experiment using an environmentally relevant dose
(Ruuskanen, Rainio, Gómez-Gallego, et al., 2020a). Japanese quails were exposed to dietary
GBHs (RoundUp Flex embedded within organic food), with concentrations below the No-
Adverse-Effect-Level (NOAEL) from 10 days after hatching to 52 weeks old. As well as
collecting blood samples to conduct genome-wide analysis of methylation profile (that will be
further discussed in this thesis), Ruuskanen et al. studied intracellular oxidative status,
neurotransmitters, gut microbiome, reproductive hormones, and ultimately reproduction.
The experiment revealed that GBHs decreased the hepatic activity of catalase (an antioxidant
enzyme) but had no impact on acetylcholinesterase (a neurotransmitter enzyme). It also
revealed that GBH exposure disrupts the microbiome and decreased male testosterone but
had no impact on reproduction.
We here investigate whether long-term exposure to dietary GBHs affects DNA methylation
patterns on a genome-wide scale in Coturnix japonica at two different time points. Chicks
from this long-term experiment were sampled for blood at 7 and 52 weeks old. We used
Reduced Representation Bisulfite Sequencing (RRBS) to assess genome-wide methylation and
highlight differential methylation sites across the genome between the 4 experimental
groups. The experimental groups are females at 7 weeks old, males at 7 weeks old, females
at 52 weeks old, and males at 52 weeks old. From now on, these groups will respectively be
referred to as F7, M7, F52, and M52 in the thesis. Considering the low number of studies
assessing genome-wide methylation regarding GBHs exposure, we did not have predictions
for the following results but expected to highlight differentially methylated sites between the
experimental groups.

M2 EPET 2020-2021 SERVANT Euphrasie
6
Materials and methods
This section will aim to provide a comprehensive resource to understand the techniques I
used throughout my internship.
Study design
The data I used is based on an experiment from Ruuskanen et al. where a long-term GBH
exposure experiment on Japanese quails was conducted (Ruuskanen et al., 2020). Chicks were
randomly distributed in 2 groups where they were either fed with GBH contaminated food or
control food from the age of 10 days to 52 weeks. Prior to the experiment, individuals were
molecularly sexed, so sexes could be evenly distributed into the experimental groups. For full
details of the experimental design see Ruuskanen et al. (2020) but I’ll briefly here explain the
main points of the experiment.
The individuals are offsprings of 21 random matings between 1-year-old birds, with no history
of GBHs exposure. All individuals were fed with organic feed (organic feed for laying chickens,
"LuonnonPunaheltta," Danish Agro, Denmark). For GBH exposed quails, Roundup Flex® was
added to their feed (480 g/l glyphosate, present as 588 g/l [43.8% w.w] of potassium salt of
glyphosate, AXGD423115/7/2017 Monsanto, with surfactants alkylpolyglycoside [5% of
weight] and nitrotryl [1% of weight]). This amount of GBH is relatively low compared to birds'
exposure in contaminated areas. GBH concentration used in this experiment represents
around 160 mg/ kg feed. It is about half of the concentration found in grains on a treated field
(Eason and Scanlon, 2002).
Relative to the amount of feed used by an average Japanese quail adult, this is at least five
times less (12e20 mg glyphosate/kg body mass/day) than the NOAEL reported by the
European Food Safety Authorit (EFSA) in 2018 on poultry (100 mg/kg body mass/day) (EFSA,
2018).
The use of a GBH product rather than pure glyphosate allows quantifying the effects of
glyphosate as it is commercially distributed and can be found in the environment. However,
the effects of the adjuvants, glyphosate, or their interaction cannot be differentiated through
this experiment.
The differential methylation analysis that will be discussed in this thesis is part of the effort
of the Ruuskanen team to assess the potential effects of GBH in food diet on Japanese quails,
presented in the introduction.
Blood sampling and sample collection
Individuals were sampled at 7 and 52 weeks old, allowing to follow GBH impact at short and
long-time exposure, at two different life points of the birds. While Japanese quails are still
young individuals at 7 weeks, they are entirely adults by 1 year old.
DNA was sampled from blood. Blood is a convenient tissue to conduct repeated sampling at
different time point on one individual, as it can be easily collected without harming.

M2 EPET 2020-2021 SERVANT Euphrasie
7
Most epigenetics studies conducted on vertebrates use blood tissue type as a proxy.
However, methylation patterns are tissue and cell-specific, hindering inferences on the
phenotypic expression of change in methylation pattern (Husby, 2020).
Blood samples (ca 140µl) were collected from the brachial vein in the wing, centrifuged and
red blood cells (RBCs) were then frozen at -80 C. Unlike mammals, avian red blood cells are
nucleated (Husby, 2020).
DNA was extracted from 10-20µl RBCs using the salt extraction method modified from
Aljanabi, 1997. Extracted DNA was treated with RNase-I according to the manufacturer's
protocol. DNA concentration was measured fluorometrically with a Qubit High Sensitivity kit
(ThermoFisher Scientific) and DNA integrity was assessed by running each DNA sample on an
agarose gel.
Reduced Representation Bisulfite Sequencing
To highlight and compare methylation patterns, DNA was sequenced using RRBS. This high-
throughput technique allows determining the DNA methylation status of single cytosines on
a genome-wide scale. This technique enriches CpG rich segments by digesting DNA with the
restriction enzyme Msp1, which cleaves DNA at the recognition sites 5'C^CGG'3. Before
sequencing, the DNA is treated with bisulfite salt, which converts unmethylated cytosines into
uracil, allowing to differentiate methylated from unmethylated Cs (Meissner, 2005; Gu et al.,
2011).
Msp1
Msp1
1. DNA is digested with Msp1, at the
recognition sites 5'C^CGG'3,
targeting CG rich regions.
2. Unmethylated Cs are converted
into Us, eventually replaced by Ts
after PCR amplification
3. After sequencing, reads are
aligned against the reference
genome and methylation state of
Cs are deduced from nucleotides
mismatch
Figure 1 RRBS process

M2 EPET 2020-2021 SERVANT Euphrasie
8
The library preparation was performed by The Finnish Functional Genomics Centre (FFGC)
according to the reduced representation bisulfite sequencing protocol from Boyle et al., 2012.
DNA was sequenced on an Illumina NovaSeq 6000 with a single-ended 100 bp run.
Hundred samples were pooled into 5 different groups. Each group contains five libraries of
each of the four conditions tested in this experiment (GBH~age) (20 samples in total). Each
pool was run on five different lanes, except one batch, which was run on two separate lanes,
producing tow libraries for one sample. Up to this date, the FFGC has not provided any
answers about why this specific batch was run on two lanes when asked. However, the report
stated that 2 different flow-cell types were used to process the samples, one being “SP” and
the other being “S1”. As the “SP” type produce twice less data than the “S1” according to the
Illumina website, the FFGC probably ran the batch twice so it could meet the reads
requirements. After merging the resulting libraries, quality was similar to other samples.
Sequence alignment
Quality control
A quality control check was performed on the raw data with FastQC (v0.11.9), a quality control
tool for high throughput sequence data (Andrews S., 2010). It identifies biases in base
composition, read duplication, base quality, and read length. From the FastQC reports, the
library quality was good, and its content as expected. Representative plots are depicted in
Figure 2.

M2 EPET 2020-2021 SERVANT Euphrasie
9
A
B
C
D
E
F

M2 EPET 2020-2021 SERVANT Euphrasie
10
Figure 2 Representative example of signatures and quality of RRBS sequenced read as provided by FastQC
(sample: 0/B5).
(A) Quality scores across all bases. The second and third bases usually have a lower score than the average,
but their scores remain good. (B) Quality per tile. Flow-cell tiles from which each read came are
represented. The plot shows the deviation from the average quality for each tile. The decreased in quality
per tile for the first few bases explains the previous plot A. (C) Quality score distribution over all sequences.
(D) Sequence content across all bases. The first base is always either a T or a C, which is expected from
Msp1 digested segments. The number of Cs remains low, which corresponds to the amount of Cs
methylated that would be expected in a Coturnix japonica’s genome (~1-2%). (E) GC distribution over all
sequences. (F) N content across all bases. If the sequencer cannot make a base call with enough confident,
it will substitute it with an N. (G) Distribution of sequence length over all sequences. The vast majority of
reads are 100bp, which is expected from a sequencing of a 100bp run. (H) Sequences duplicates. This plot
shows the proportion of duplicate sequences in the library. Considering the enrichment biased of an RRBS
library, a high level of what appears as PCR duplicates is normal, and they should not be removed. (I)
Adapters sequences contamination. This plot shows the possible contamination from Illumina adapters.
No, or very weak contamination with adapters sequences here.
G
H
I

M2 EPET 2020-2021 SERVANT Euphrasie
11
Trimming
Trim Galore! (v0.6.5) was used to remove low-quality bases and reads and adapters
contamination prior to the reads alignment against the reference genome (Felix Krueger,
2019). On average, sequencing resulted in 44,23 million reads per library, of which 44,18
remained after quality filtering. As reads were of good quality before trimming, read lengths
after trimming were hardly affected.
Alignment
The reads were mapped to the Coturnix japonica 2.1 genome (GCF_001577835.2) using
Bismark (v0.21.0), a program allowing the mapping of bisulfite treated sequencing reads to
a reference genome as well as identifying methylation sites {Citation}.Out of the 44,18
million trimmed reads, 46.5% ± 4,0% were mapped to the reference genome. This rate can
be considered low but is comparable to the alignment rate found in other studies. Mäkinen
et al., who conducted RRBS on red blood cells as it was done in this study, found a mapping
efficiency of 52.0 ± 2.0 (Mäkinen et al., 2019). Indeed, RRBS alignment rates are usually
lower than for regular sequencing. First, the alignment efficiency is decreased because it
relies on three nucleotides, rather than 4 (Baheti et al., 2016). Also, the published Japanese
quail's genome contains many scaffolds (~2500), which is known to make the mapping
challenging (Tran et al., 2014; Baheti et al., 2016).
Methylation call
In total, 7 881 349 CpG sites were identified of which 2 362 801 had at least 10x coverage
(Ziller et al., 2015; Lindner et al., 2021). In order to minimize the amount of missing data we
only kept sites that were present in all of the hundred samples (Mäkinen et al., 2019). After
filtering, 440 977 sites remained.
Differential methylation analysis
The average methylation level for each CpG site was calculated over the samples for each
sex~treatment group within the two age groups. Sites having at least 10% of difference in
methylation level between any of the different age~sex~conditions groups (Leenen, Muller
and Turner, 2016; Lindner et al., 2021) were included in the differential methylation analysis
reducing the data set to 52 594 sites.
To identify differentially methylated sites, a generalized linear mixed model (GLMM) was
applied to the data set (Baheti et al., 2016; Zhang, Baheti and Sun, 2016) enabling to take into
account all the grouping factors and parameters that could influence methylation level, as
specified in the Equation 1.
Equation 1
sitei∼Binomial(n= 1, probsite=1 =

P)
log [
P
1−
P]=αj[i]
αj∼N(γα
0+γα
1(sex) + γα
2(treatmentglyph) + γα
3(treatmentglyph ×sex), σ2
αj),!for!relatedness!j!=!1,…,J

M2 EPET 2020-2021 SERVANT Euphrasie
12
Equation 1 was applied to the dataset through the R package lme4qtl (Ziyatdinov et al., 2018),
allowing to fit GLMM while accounting for relatedness between individuals. The convergence
was optimized, and the computation was sped up by using quadratic approximation (bobyqa
optimiser) (Lindner et al., 2021).
During the model fitting, four types of warning came out: singular fit warnings, 'unable to
evaluate scaled gradient' warnings, warnings about models failing to converge, and warnings
about large eigen values. Compare to the number of sites analyzed, there were very few
warnings (820 and 760 respectively for the analysis on the 7 weeks old and on the 52 weeks
old out of 52 594 fits). To prevent false discovery, all sites that were issued a warning were
discarded.
For each CpG site, the fitted model was used to infer the estimated marginal means (EMMs)
for all factor combinations and then calculate contrasts between the two glyphosate
treatments within a sex category. P-values were normally distributed, as expected.
P-values were then corrected using the python package combined-pvalues (v0.50.4)
(Pedersen et al., 2012) using a distance of 750bp with a p-value threshold of 0.05. The
combined-pvalues approach combines the p-value associated with the target site with the p-
values of its neighbors: i.e., neighboring p-values influence the correction. Out of the
combined-pvalues correction, 938 and 585 significant sites remains respectively in M7 and
M52, and 1962 and 1160 in F7 and F12.
Promoter regions and gene annotation
This study focuses on CpG sites in promoter regions as CpG site methylation in promoters is
negatively correlated with gene expression (The Great Tit HapMap Consortium et al., 2016).
Promoters were defined as a region span from 3000bp upstream of the transcription site of a
gene to 300bp downstream (Viitaniemi et al., 2019). The R package rtracklayer (v1.52.0)
(Michael Lawrence, 2017) was used to export the Coturnix japonica GFF3 file
(GCF_001577835.2) (a file format used for describing genes and other features of DNA, RNA
and protein sequences) into R. The genomicfeatures package was used to define promoters'
regions as previously defined and export the results as GFF3 files (M. Carlson, 2017).
The significantly differentially methylated sites were then located into the promoter regions
using bedtools (v2.30.0) (Quinlan and Hall, 2010). The numbers of significantly differentially
methylated sites found in promoters are in Table 1. Average CpG site methylation levels of
these sites are provided in Table 1.

M2 EPET 2020-2021 SERVANT Euphrasie
13
Results
Table 1 Results from the differential methylation analysis.
DM = Differentially methylated.
The analysis identified 346 differentially methylated sites between control and treated
samples in females and 541 in males. Among these sites, respectively 139 and 175 are located
in promoter (Table 1). At 52 weeks, the number of sites localized in promoters increases in
both sexes to 1160 in females and 585 in males. While the number of sites found in promoters
rises proportionally in females at 7 and 52 weeks, it decreases in males.
At 7 weeks, females seemed to be less impacted than males: we identified 43% more
differentially methylated CpGs in males than in females. While on average, 77% of these sites
were hypermethylated in the males’ GBHs treated group, it drops to 53% in females (Table
1). At 52 weeks, females showed almost 50% more methylated sites than males, with a higher
amount of hypermethylated CpGs (80% against 65% for males) (Table 1).
If most of the differentially methylated CpGs identified are hypermethylated in the GBHs
treated group (i.e., in comparison, the average methylation level is higher), this amount
decreases in promoters, except for F52. For the F7, F52, and M52 groups, there are between
4.6 to 5.9% more hypermethylated sites in promoters than in the rest of the genome. In M7,
it goes up 13,6%.
Figure 3 Mean methylation percentage per groups at 7 and 52 weeks old
As shown in Figure 3, there was a steeper increase in methylation in the female GBHs treated
groups.

M2 EPET 2020-2021 SERVANT Euphrasie
14
Sites in promoters are distributed across 26 different genes in F7 and 72 at F52 weeks. In
males, they are spread in 24 genes at 7 weeks and 29 at 52 (Table 1). A few genes are shared
between groups of the same sex: 8 genes are found in F7 and F52, and 8 others in M7 and
M52. No genes were shared between all the groups. However, 2 genes were found to be in
F52 and M52, an unknown gene and the gene PIN4 (peptidylprolyl cis/trans isomerase, NIMA-
interacting 4 [Source:NCBI gene;Acc:107312540]). Genes shared between groups, and their
descriptions are presented in Table 2 and Table 3.
Table 2 Shared genes in F7 and F52 including differentially methylated sites
Table 3 Shared genes in M7 and M52 including differentially methylated sites
The ontology analysis performed on the genes covered by differentially methylated sites in
promoter did not highlight any enriched terms in any experimental groups or between the
shared genes, except in F7, where the GO term “positive regulation of cyclin-dependent
protein serine/threonine kinase activity” (GO:0045737) and “positive regulation of cyclin-
dependent kinase activity” (GO:1904031) where find to be significant (p-value < 0.05). These
two GO terms are associated with the genes CCDN1, encoding a cyclin-D1 protein and forming
a complex with proteins CDK4 and CDK6, regulating the G0/G1 checkpoint, and SPDYA
(Speedy/RINGO Cell Cycle Regulator Family Member A), a protein-coding gene regulating the
G1/S phase transition of the cell cycle by binding and activating CDK1 and CDK2. The
differentially methylated site found in the promoters of these two genes are both
hypomethylated compare to the control treatment.
Besides these two cell-cycle regulators, GO terms in our analysis cover a broad and non-
specific range of semantic.

M2 EPET 2020-2021 SERVANT Euphrasie
15
Figure 4 Mean methylation percentage in CCND1’ differentially methylated site
Figure 5 Mean methylation percentage across SPDYA' differentially methylated sites
Female, control
Male, control
Female, glyphosate
Male, glyphosate
Female, control
Male, control
Female, glyphosate
Male, glyphosate

M2 EPET 2020-2021 SERVANT Euphrasie
16
Discussion
This study investigated whether long-term exposure to GBH affects DNA methylation of CpGs
on a whole-genome level in Japanese quails.
First, the results highlighted in Figure 3 andTable 1 demonstrate a natural age-related
evolution in DNA methylation, as in both sexes, the control groups showed an increase in
mean methylation. It was shown in mammals that age was positively correlated with gain in
DNA methylation(BI Ageing Clock Team et al., 2017). During development, DNA methylation
plays an essential role as an epigenetic barrier repressing (when methylated) or enabling
(when unmethylated) the expression of genes (Lou et al., 2014). Development genes acting
in the early stage of life, it is also a general tendency to gain in DNA methylation over time, as
genes are being repressed(BI Ageing Clock Team et al., 2017).
DNA methylation was globally increased by the GBHs treatment, even if the global change
remains relatively weak. If other studies corroborate that GBHs induces hypermethylation
(Gomez et al., 2019; Woźniak et al., 2020), others don’t, (Kwiatkowska et al., 2017; Duforestel
et al., 2019) and the effect of GBHs on DNA methylation remains unclear. However, these
studies target mammals (as in Gomez et al., where the expression of an estrogen receptors
was altered by hypermethylation in rats); often specific proto-oncogenes (as in Woźniac et al.
targeting the genes CCND1, P16, P21 and TP53 inolved in cell cyle) and different tissues than
blood which makes the results not always reliable in our context. Up to this date, this is the
only study testing the effects of long-term GBHs exposure on DNA methylation on a genome-
scale, in a bird species.
The response to GBHs treatment was drastically different between males and females,
especially at 7 weeks when looking at hypermethylated sites (Table 1). If there is no evidence
to explain why the responses between the sexes are different, it is interesting to see that
when the Japanese quail’s preference towards GBH contaminated food was tested, females
were preferably drawn to contaminated food (Ruuskanen, Rainio, Kuosmanen, et al., 2020).
Perhaps, during the experiment, females were eating more of the contaminated food than
males, resulting in higher exposure to GBHs. Also, female quails being more prominent than
male, thus eating more, makes females again more exposed to GBHs, potentially explaining
the differences observed from the Table 1.
The broad range of ontologies covered by the genes from the analysis, and the differences
between the groups, suggest GBHs may have a general effect on DNA methylation, rather
than targeted.
Nonetheless, the enrichment in the F7 group of cell-cycles-related ontologies (Figure 4 and
Figure 5) supports numerous studies stating the tumorigenic potential of GBHs, especially
Woźniak et al. who highlighted the effects of GBHs on CCDN1. Woźniak et al., 2020 tested
methylation levels within selected gene promoters involved in proliferation, tumorigenesis,
and apoptosis processes (among which CCDN1) to low concentrations of GBHs on human
cells. If the gene was overly expressed when cells were exposed to GBHs, they did not observe
significant changes in DNA methylation in CCDN1 promoter. However, in our study, the one
site differentially methylated present in CCDN1 promoter was found to be hypomethylated
in the F7 group when treated with GBHs (~11% decrease in methylation level), suggesting that
there could be an increase in the expression of CCDN1, as promoter methylation has been
found to be inversely correlated with gene expression. Even though the site was not found to

M2 EPET 2020-2021 SERVANT Euphrasie
17
be significantly differentially methylated in other groups, the methylation level was lower in
all of them.
The same observations were made on SPDYA, the other oncogene highlighted from the
enrichment analysis in the F7 group. SPDYA is a protein-coding gene for a cell cycle regulator
involved in meiosis. It regulates the cell cycle by interacting with cyclin-dependant kinase.
SPDYA binding with CDK-2 allows the initial assembly of the meiotic telomere complex,
beginning prophase 1 (Mikolcevic et al., 2016). Knockout of SPDYA leads to loss of male and
female germ cells (Tu et al., 2017). A decrease in SPDYA expression due to hypomethylation
in its promoters could have serious consequences on the species reproduction.

M2 EPET 2020-2021 SERVANT Euphrasie
18
Conclusion
Finally, the analysis of methylation on a genome-scale did not yet allow to back up the
observations from Ruuskanen, Rainio, Gómez-Gallego, et al., 2020, including the decrease in
males’ testosterone, decrease in antioxidant enzymes, or corroborate hypotheses that
glyphosate might induce oxidative stress or disturbance in reproductive hormones, which
would have been difficult at whole-genome scale. To better identify effects on targeted
biological pathways, we could in the future focus on differentially methylated analysis on a
specific set of genes, as well as assessing the expression profile of these genes to link
methylation level and gene expression.
on However, we did highlight differentially methylated sites, confirming that GBHs impact
methylation pattern in birds’ DNA, suggesting potentially dangerous effect on non-target
species, especially reproduction.

M2 EPET 2020-2021 SERVANT Euphrasie
19
Aitbali, Y. et al. (2018) ‘Glyphosate based- herbicide exposure affects gut microbiota, anxiety and depression-
like behaviors in mice’, Neurotoxicology and Teratology, 67, pp. 44–49. doi: 10.1016/j.ntt.2018.04.002.
Aljanabi, S. (1997) ‘Universal and rapid salt-extraction of high quality genomic DNA for PCR- based techniques’,
Nucleic Acids Research, 25(22), pp. 4692–4693. doi: 10.1093/nar/25.22.4692.
Andrews S. (2010) FastQC: a quality control tool for high throughput sequence data. Available at:
http://www.bioinformatics.babraham.ac.uk/projects/fastqc.
Baheti, S. et al. (2016) ‘Targeted alignment and end repair elimination increase alignment and methylation
measure accuracy for reduced representation bisulfite sequencing data’, BMC Genomics, 17(1), p. 149. doi:
10.1186/s12864-016-2494-8.
Bai, S. H. and Ogbourne, S. M. (2016) ‘Glyphosate: environmental contamination, toxicity and potential risks to
human health via food contamination’, Environmental Science and Pollution Research, 23(19), pp. 18988–19001.
doi: 10.1007/s11356-016-7425-3.
Bali, Y. A. et al. (2019) ‘Learning and memory impairments associated to acetylcholinesterase inhibition and
oxidative stress following glyphosate based-herbicide exposure in mice’, Toxicology, 415, pp. 18–25. doi:
10.1016/j.tox.2019.01.010.
Benbrook, C. M. (2016) ‘Trends in glyphosate herbicide use in the United States and globally’, Environmental
Sciences Europe, 28(1), p. 3. doi: 10.1186/s12302-016-0070-0.
BI Ageing Clock Team et al. (2017) ‘Multi-tissue DNA methylation age predictor in mouse’, Genome Biology,
18(1), p. 68. doi: 10.1186/s13059-017-1203-5.
Bossdorf, O., Richards, C. L. and Pigliucci, M. (2007) ‘Epigenetics for ecologists’, Ecology Letters, 0(0), pp.
071117033013002-??? doi: 10.1111/j.1461-0248.2007.01130.x.
Duforestel, M. et al. (2019) ‘Glyphosate Primes Mammary Cells for Tumorigenesis by Reprogramming the
Epigenome in a TET3-Dependent Manner’, Frontiers in Genetics, 10, p. 885. doi: 10.3389/fgene.2019.00885.
Dupont, C., Armant, D. and Brenner, C. (2009) ‘Epigenetics: Definition, Mechanisms and Clinical Perspective’,
Seminars in Reproductive Medicine, 27(05), pp. 351–357. doi: 10.1055/s-0029-1237423.
Eason, T. H. and Scanlon, P. F. (2002) ‘Effects of Atrazine and Glyphosate ingestion on body weight and
nutritional well-being of Coturnix quail’, Zeitschrift für Jagdwissenschaft, 48(S1), pp. 281–285. doi:
10.1007/BF02192419.
European Food Safety Authority (EFSA) (2018) ‘Evaluation of the impact of glyphosate and its residues in feed
on animal health’, EFSA Journal, 16(5). doi: 10.2903/j.efsa.2018.5283.
Felix Krueger (2019) Trim Galore! Available at:
https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/.
Gallegos, C. E. et al. (2018) ‘Perinatal Glyphosate-Based Herbicide Exposure in Rats Alters Brain Antioxidant
Status, Glutamate and Acetylcholine Metabolism and Affects Recognition Memory’, Neurotoxicity Research,
34(3), pp. 363–374. doi: 10.1007/s12640-018-9894-2.
Ghisi, N. de C., Oliveira, E. C. de and Prioli, A. J. (2016) ‘Does exposure to glyphosate lead to an increase in the
micronuclei frequency? A systematic and meta-analytic review’, Chemosphere, 145, pp. 42–54. doi:
10.1016/j.chemosphere.2015.11.044.
Gomez, A. L. et al. (2019) ‘Male mammary gland development and methylation status of estrogen receptor alpha
in Wistar rats are modified by the developmental exposure to a glyphosate-based herbicide’, Molecular and
Cellular Endocrinology, 481, pp. 14–25. doi: 10.1016/j.mce.2018.11.005.

M2 EPET 2020-2021 SERVANT Euphrasie
20
Greenberg, M. V. C. and Bourc’his, D. (2019) ‘The diverse roles of DNA methylation in mammalian development
and disease’, Nature Reviews Molecular Cell Biology, 20(10), pp. 590–607. doi: 10.1038/s41580-019-0159-6.
Husby, A. (2020) ‘On the Use of Blood Samples for Measuring DNA Methylation in Ecological Epigenetic Studies’,
Integrative and Comparative Biology, 60(6), pp. 1558–1566. doi: 10.1093/icb/icaa123.
Kwiatkowska, M. et al. (2017) ‘DNA damage and methylation induced by glyphosate in human peripheral blood
mononuclear cells ( in vitro study)’, Food and Chemical Toxicology, 105, pp. 93–98. doi:
10.1016/j.fct.2017.03.051.
Leenen, F. A. D., Muller, C. P. and Turner, J. D. (2016) ‘DNA methylation: conducting the orchestra from exposure
to phenotype?’, Clinical Epigenetics, 8(1), p. 92. doi: 10.1186/s13148-016-0256-8.
Leino, L. et al. (2020) Classification of the glyphosate target enzyme (5-enolpyruvylshikimate-3-phosphate
synthase). preprint. Bioinformatics. doi: 10.1101/2020.05.27.118265.
Lindner, M. et al. (2021) ‘Rapid changes in DNA methylation associated with the initiation of reproduction in a
small songbird’, Molecular Ecology, p. mec.15803. doi: 10.1111/mec.15803.
Lou, S. et al. (2014) ‘Whole-genome bisulfite sequencing of multiple individuals reveals complementary roles of
promoter and gene body methylation in transcriptional regulation’, Genome Biology, 15(7), p. 408. doi:
10.1186/s13059-014-0408-0.
M. Carlson, H. P. (2017) GenomicFeatures. Bioconductor. doi: 10.18129/B9.BIOC.GENOMICFEATURES.
Mäkinen, H. et al. (2019) ‘Temporally replicated DNA methylation patterns in great tit using reduced
representation bisulfite sequencing’, Scientific Data, 6(1), p. 136. doi: 10.1038/s41597-019-0136-0.
Manservisi, F. et al. (2019) ‘The Ramazzini Institute 13-week pilot study glyphosate-based herbicides
administered at human-equivalent dose to Sprague Dawley rats: effects on development and endocrine system’,
Environmental Health, 18(1), p. 15. doi: 10.1186/s12940-019-0453-y.
Michael Lawrence, V. C. (2017) rtracklayer. Bioconductor. doi: 10.18129/B9.BIOC.RTRACKLAYER.
Mikolcevic, P. et al. (2016) ‘Essential role of the Cdk2 activator RingoA in meiotic telomere tethering to the
nuclear envelope’, Nature Communications, 7(1), p. 11084. doi: 10.1038/ncomms11084.
Pedersen, B. S. et al. (2012) ‘Comb-p: software for combining, analyzing, grouping and correcting spatially
correlated P-values’, Bioinformatics, 28(22), pp. 2986–2988. doi: 10.1093/bioinformatics/bts545.
Quinlan, A. R. and Hall, I. M. (2010) ‘BEDTools: a flexible suite of utilities for comparing genomic features’,
Bioinformatics, 26(6), pp. 841–842. doi: 10.1093/bioinformatics/btq033.
Romano, R. M. et al. (2010) ‘Prepubertal exposure to commercial formulation of the herbicide glyphosate alters
testosterone levels and testicular morphology’, Archives of Toxicology, 84(4), pp. 309–317. doi: 10.1007/s00204-
009-0494-z.
Rossetti, M. F. et al. (2021) ‘Epigenetic Changes Associated With Exposure to Glyphosate-Based Herbicides in
Mammals’, Frontiers in Endocrinology, 12, p. 671991. doi: 10.3389/fendo.2021.671991.
Ruuskanen, S., Rainio, M. J., Kuosmanen, V., et al. (2020) ‘Female Preference and Adverse Developmental Effects
of Glyphosate-Based Herbicides on Ecologically Relevant Traits in Japanese Quails’, Environmental Science &
Technology, 54(2), pp. 1128–1135. doi: 10.1021/acs.est.9b07331.
Ruuskanen, S., Rainio, M. J., Gómez-Gallego, C., et al. (2020a) ‘Glyphosate-based herbicides influence
antioxidants, reproductive hormones and gut microbiome but not reproduction: A long-term experiment in an
avian model’, Environmental Pollution, 266, p. 115108. doi: 10.1016/j.envpol.2020.115108.

M2 EPET 2020-2021 SERVANT Euphrasie
21
Ruuskanen, S., Rainio, M. J., Gómez-Gallego, C., et al. (2020b) ‘Glyphosate-based herbicides influence
antioxidants, reproductive hormones and gut microbiome but not reproduction: A long-term experiment in an
avian model’, Environmental Pollution, 266, p. 115108. doi: 10.1016/j.envpol.2020.115108.
Smith, C. M., Vera, M. K. M. and Bhandari, R. K. (2019) ‘Developmental and epigenetic effects of Roundup and
glyphosate exposure on Japanese medaka (Oryzias latipes)’, Aquatic Toxicology, 210, pp. 215–226. doi:
10.1016/j.aquatox.2019.03.005.
Székács, A. and Darvas, B. (2018) ‘Re-registration Challenges of Glyphosate in the European Union’, Frontiers in
Environmental Science, 6, p. 78. doi: 10.3389/fenvs.2018.00078.
Tarazona, J. V. et al. (2017) ‘Glyphosate toxicity and carcinogenicity: a review of the scientific basis of the
European Union assessment and its differences with IARC’, Archives of Toxicology, 91(8), pp. 2723–2743. doi:
10.1007/s00204-017-1962-5.
The Great Tit HapMap Consortium et al. (2016) ‘Evolutionary signals of selection on cognition from the great tit
genome and methylome’, Nature Communications, 7(1), p. 10474. doi: 10.1038/ncomms10474.
Tran, H. et al. (2014) ‘Objective and Comprehensive Evaluation of Bisulfite Short Read Mapping Tools’, Advances
in Bioinformatics, 2014, pp. 1–11. doi: 10.1155/2014/472045.
Tu, Z. et al. (2017) ‘Speedy A–Cdk2 binding mediates initial telomere–nuclear envelope attachment during
meiotic prophase I independent of Cdk2 activation’, Proceedings of the National Academy of Sciences, 114(3),
pp. 592–597. doi: 10.1073/pnas.1618465114.
Viitaniemi, H. M. et al. (2019) ‘Seasonal Variation in Genome-Wide DNA Methylation Patterns and the Onset of
Seasonal Timing of Reproduction in Great Tits’, Genome Biology and Evolution. Edited by M. Meyer, 11(3), pp.
970–983. doi: 10.1093/gbe/evz044.
Woźniak, E. et al. (2020) ‘Glyphosate affects methylation in the promoter regions of selected tumor suppressors
as well as expression of major cell cycle and apoptosis drivers in PBMCs (in vitro study)’, Toxicology in Vitro, 63,
p. 104736. doi: 10.1016/j.tiv.2019.104736.
Zhang, Y., Baheti, S. and Sun, Z. (2016) ‘Statistical method evaluation for differentially methylated CpGs in base
resolution next-generation DNA sequencing data’, Briefings in Bioinformatics, p. bbw133. doi:
10.1093/bib/bbw133.
Ziller, M. J. et al. (2015) ‘Coverage recommendations for methylation analysis by whole-genome bisulfite
sequencing’, Nature Methods, 12(3), pp. 230–232. doi: 10.1038/nmeth.3152.
Ziyatdinov, A. et al. (2018) ‘lme4qtl: linear mixed models with flexible covariance structure for genetic studies
of related individuals’, BMC Bioinformatics, 19(1), p. 68. doi: 10.1186/s12859-018-2057-x.