Milt from eight mature Atlantic salmon males kept at a broodstock facility in Central Norway was collected by gonad extraction, during spawning season. Briefly, males were culled, and the gonads were removed, cleaned, and homogenised using a mechanical grinder. The homogenate was weighed and a 1:1 amount of AquaBoost^® SpermCoat (Cryogenetics AS, Hamar, Norway) was added, before filtration and storage in 650 mL cell culture flasks (VWR International, Radnor, PA, United States). The milt samples were shipped overnight to the laboratory (Cryogenetics AS, Hamar, Norway) while kept cold on paper on top of crushed ice to ensure a temperature of about 2°C on arrival. Aliquotes of milt from each male (10 mL) were stored at 2°C and 8°C in cell culture flasks (50 mL) with filter screw caps (VWR International, Radnor, PA, United States) for 4 days. Samples were taken on day one of storage at 2°C (hereafter referred to as D1-2°C) and on day four of storage at 2°C and 8°C (hereafter referred to as D4-2°C, D4-8°C). Correspondingly aliquots of cold-stored milt were prepared for cryopreservation (hereafter referred to as D1-2°C/Cryo, D4-2°C/Cryo, D4-8°C/Cryo), resulting in a total of six experimental conditions. Samples from all six experimental conditions were analyzed for sperm viability, motility, ATP content, DNA fragmentation index (DFI), and High DNA stainability (HDS). In addition, samples from four of the males, selected based on their contrasting performance in the fertilization trials (described in 2.7), were analyzed for sperm metabolites and DNA methylation signatures by reduced representation bisulfite sequencing (RRBS). Cryopreservation was conducted using the following procedure. Sperm concentration was estimated after mixing 10 µL milt with 4 mL 0.9% NaCl using an SDM6 photometer (Minitüb GmbH, Tiefenbach, Germany) at 546 nm using a pre-determined calibration curve for Atlantic salmon sperm. Sperm concentration varied between 10.2 and 12.2 × 10⁹ sperm/mL. Based on the concentration estimate, the milt was diluted using a proprietary fish sperm extender (Cryogenetics AS, Hamar, Norway) to a standardized sperm concentration. Pre-printed 0.5 mL medium straws (IMV technologies, L’Aigle, France) were filled with milt and sealed with ultrasound using an MPP Uno automated filling and sealing machine (Minitüb GmbH, Tiefenbach, Germany). Cryopreservation was performed according to a proprietary protocol (Cryogenetics AS, Hamar, Norway) before transfer and storage in liquid N₂. The preprint ensured identification of the male, storage condition, and date of cryopreservation.

Samples from the six different experimental conditions from all individuals were treated identically according to the following procedure. Samples were analyzed in triplicate for viability, ATP content, and DFI, and in duplicate for motility assessment. For the cryopreserved milt, three doses per batch were thawed in a water bath (25°C, 30 s) and pooled. All samples were kept cold on paper on top of crushed ice before in vitro analysis. All chemicals were purchased from Sigma-Aldrich (St. Louis, MO, United States) unless otherwise noted.

The analysis of sperm plasma membrane integrity (sperm viability) was assessed using propidium iodide (PI, L-7011, LIVE/DEAD®Sperm Viability Kit, Molecular Probes, ThermoFisher Scientific, Waltham, MA United States). The spermatozoa were stained in dark (4°C, 10 min) in PBS (137 mM NaCl, 2.7 mM KCl, 1.76 mM KH₂PO₄, 8.1 mM Na₂HPO₄·2H₂O, pH 7.4) staining solution with a final concentration of 1.5 × 10⁶ spermatozoa/mL and 0.48 mM PI. The flow cytometry analysis was performed using a Cell Lab Quanta ™ SC MPL flow cytometer (Beckman Coulter, Fullerton, CA, United States), equipped with a 22 mW argon laser with an excitation wavelength of 488 nm. Data were analyzed using Cell Lab Quanta SC MPL Analysis software program (Software Version 1.0, Beckman Coulter, Fullerton, CA, United States). Side scatter (SS) signals were used for the identification of spermatozoa, and PI fluorescence in SS-gated spermatozoa was detected using a 670 nm long pass (LP) filter (FL3).

Sperm motility analysis was performed using a SCA evolution CASA system (Sperm Class Analyzer^® version 6.1, Microptic SL, Barcelona, Spain). The CASA system was equipped with a phase contrast Eclipse Ci-S/Ci-L microscope (Nikon, Japan) with a temperature-controlled stage using a water circulation pump, and a Basler digital camera (Basler Vision Technologies, Ahrensburg, Germany). Chamber slides were kept on the cooled microscopic stage before analysis. Samples were diluted to a working sperm concentration of 87 × 10⁶ cells/mL in AquaBoost^® Dilutor (Cryogenetics AS, Hamar, Norway) prior to analysis. Sperm activation was initiated by the addition of AquaBoost^® Activator (Cryogenetics AS, Hamar, Norway) simultaneously with loading the sample to the microscope slide. Using a Picus (0.2–10 µL) electronic 1-channel pipette (Sartorius Stedim Biotech GmbH, Göttingen, Germany) in Multi-Aspirating Mode, aspiration was done sequentially in the following order: 1.2 µL diluted milt—air—1.2 µL AquaBoost^® Activator. For each CASA measurement, the aspiration volume was dispensed into a single chamber on a Leja 20 µm 4 chamber microscope slide (Leja, Nieuw-Vennep, the Netherlands). Image analysis was performed on four consecutive fields within approximately 8 s after activation, with a total of at least 400 cells analyzed per sample. The parameters recorded were total motility (TM, %), progressive motility (PM, %), straight-line velocity (VSL, mm/sec), average path velocity (VAP, mm/sec), curvilinear velocity (VCL, mm/sec), straightness (STR, ratio VSL/VAP) and linearity (LIN, ratio VSL/VCL). Sperm cells were identified by sperm head area of 10–50 μm², a 50 Hz frame rate was used, and 45 frames were captured per object. Motile sperm cells were defined based on VAP (µm/sec) 15< Slow < 30 < Medium < 50 < Rapid. Cells were classified as progressive motile if STR >45%.

The sperm ATP content was determined using the CellTiter-Glo^® Luminescent Cell Viability Assay (G7571, Promega, Madison, WI, United States), a FLUOstar OPTIMA multiwell plate reader (BMG Labtech GmbH, Offenburg, Germany) and MARS data analysis software (Version 1.10, BMG Labtech GmbH, Offenburg, Germany). In brief, milt samples were diluted to 5 × 10⁶ spermatozoa/mL in AquaBoost^® SpermCoat (Cryogenetics AS, Hamar, Norway), and each sample (50 µL) was transferred to wells in an opaque 96-well microtiter plate (NUNC™, Roskilde, Denmark). CellTiter-Glo^® Reagent (50 µL) was added to each well and the mixture was gently shaken for 2 min in a rotary shaker (300 rpm) (IKA^® MS 3 digital, IKA®-Werke GmbH & Co., Staufen, Germany), followed by 10 min incubation in room temperature (RT). The data recorded was measured in relative luminescence units (RLU), and converted to corresponding ATP values (nM) according to a prepared standard curve.

Sperm chromatin integrity was analyzed using the Sperm Chromatin Structure Assay (SCSA), as previously described for bovine semen (Narud et al., 2020), with minor modifications. In brief, milt samples were diluted to about 1.5 × 10⁷ spermatozoa/mL in TNE buffer (10 mM Tris-HCL, 0.1 M NaCl, 1 mM EDTA, pH 7.4) in a final volume of 200 μL. Next, 400 μL acid detergent solution [0.38 M NaCl, 120 mM HCL, 0.1% (w/v) Triton X-100, pH 1.2] was added, followed by incubation for 30 s in RT. Next, 1.2 mL acridine orange (AO) staining solution [6 mg/mL AO (A3568, Life Technologies, OR, United States)] in a buffer containing 37 mM citric acid, 0.126 M Na₂HPO₄, 1.1 mM EDTA, and 0.15 M NaCl (pH 6) was added. Data acquisition started after a 3 min setup mode, in which 5000 events were captured for each sample at a rate of ∼200 events/sec. Analysis in FCS Express 6 Flow cytometry Software (Denovo Software, Los Angeles, CA, United States) revealed the percentage of red (ssDNA) and green (dsDNA) fluorescence. Based on a histogram of the fluorescence ratio red/(red + green), the sperm cells with fragmented DNA (DFI, %) were calculated. The percentage of spermatozoa with high DNA stainability (HDS, %) was determined by the bivariate cytogram.

Fertilization trials were performed using milt from all six different experimental conditions (2.1). For each trial, eggs from one female were stripped and shipped overnight on ice from a broodstock facility to Cryogenetics AS in Hamar. Eggs were kept in ovarian fluid and shipped in 650 mL cell culture flasks (filled with approximately 300 mL per flask). In vitro fertilization was carried out using two different sperm:egg ratios, 75 × 10³ sperm cells/egg (duplicates) and 500 × 10³ sperm cells/egg. Each fertilization unit consisted of approximately 150 eggs estimated by weight (18 g). Cryopreserved straws were thawed for 30 s in a water bath at 25°C. Fertilization was performed in a climate-controlled room at 4°C. The temperature difference between eggs and other media throughout the fertilization trial was at all times below 2.5°C. For each fertilization, sperm and eggs were gently mixed in a plastic cup while adding AquaBoost^® Activator sperm activating solution to slightly cover the eggs. After 2 min, eggs were rinsed in physiological saline solution (0.9% NaCl) and transferred to nylon mesh-bottomed incubation rings. Incubation rings were then incubated in fresh water. After incubation for approximately 150–200 h-degrees post-fertilization, eggs were transferred to a plastic cup and AquaBoost^® Quattro (Cryogenetics AS, Hamar, Norway) was added. After fixation for 15 min in this solution, each egg was visually inspected and categorized as fertilized or unfertilized, based on the presence or absence of cell division.

Samples from all six experimental conditions were analyzed for targeted metabolites. For the cold-stored samples, 2 mL milt prediluted 1:1 with AquaBoost^® SpermCoat was centrifuged (800 × g, 10 min) to separate the supernatant (seminal plasma and extender, hereafter referred to as seminal plasma) from the pellet (sperm cells). For the cryopreserved samples, one straw of frozen-thawed milt was centrifuged (500 × g, 5 min). All pellet and supernatant samples were snap-frozen in liquid N₂ and shipped on dry ice to SINTEF (SINTEF Industry, Trondheim, Norway) for analyses of targeted amines and amino acids. Samples of the proprietary extenders used were also analyzed for background levels of the targeted metabolites.

At SINTEF the samples were thawed, and an aliquot was collected for sample preparation and analysis. Sample preparation consisted of protein precipitation by the addition of four volumes of ice-cold methanol. Following centrifugation, a 100 μL aliquot of the supernatant was collected for amine analysis, dried using a speed-vac, and dissolved in an aqueous internal standard mix prior to analysis. A 10 μL aliquot of the supernatant was collected for amino acid analysis. The sample was mixed with the amino acid internal standard mixture and derivatized using an AccQ-Tag Derivatization Kit (Waters, Milford, MA, United States) prior to analysis.

Metabolite analysis of amino acids and amines was performed on an Agilent 1290 Infinity II LC system (Agilent, Santa Clara, United States) coupled to an Agilent 6495 QqQ mass spectrometer, using one method for amino acids and one for amines. The QqQ-MS was equipped with a jet-stream ESI source operated in positive mode. The QqQ-MS was operated in dynamic multiple-reaction monitoring (MRM) mode (delta Rt = 1 min) with a unit mass resolution for both mass filters. The MRM transitions for standards and internal standards and the employed collision energies, gas temperatures, and flows are given in Supplementary Tables S1, S2. For the amines (choline, creatine, and L-carnitine), the chromatographic separation was performed in HILIC mode employing a BEH Amide (2.1 × 150 mm, 2.7 μM) column (Waters, Milford, United States) and gradient elution using an 80:20 mixture of 25 mM Formic acid and 50 mM ammonium acetate as eluent A, and acetonitrile as eluent B, at a flow rate of 0.3 mL/min. The gradient used started at 90% B and was decreased stepwise to 5% B at 6.5 min. The column thermostat was maintained at 35°C and the autosampler at 8°C. The injection volume was 1 μL. Mixed standards at 0, 0.5, 1, 10, 50, 100, 500, 1,000, and 5,000 nM were used for calibration and quantitation. Amino acids (alanine, arginine, asparagine, GABA, glutamate, glutamine, glycine, Histidine, isoleucine, leucine, lysine, methionine, ornithine, phenylalanine, proline, serine, threonine, tryptophan, tyrosine, and valine) were analyzed in reversed-phase mode employing an Ascentis Express C18 (2.1 × 150 mm, 2.7 μM) column (Sigma-Aldrich), and gradient elution using 25 mM Formic acid as eluent A, and acetonitrile as eluent B, at a flow rate of 0.3 mL/min. The gradient used started at 5% B and was increased stepwise to 50% B at 15 min. The column thermostat was maintained at 20°C and the autosampler at 6°C. The injection volume was 2 μL. Mixed standards at 0, 0.1, 1, 5, 10, 25, 50, 100, 250, 500, and 1,000 μM were used for calibration and quantitation.

Parallel samples from four of the eight individuals belonging to D1-2°C, D1-2°C-Cryo, D4-2°C, and D4-8°C groups were considered for the construction of RRBS libraries using a gel-free multiplexed technique (Boyle et al., 2012), previously optimized to study sperm DNA methylation in boar (Khezri et al., 2019) and bull (Khezri et al., 2020). In brief, 100 ng genomic DNA isolated from milt samples was digested overnight at 37°C using MspI and Taqα1 enzymes (New England Biolabs). Fragmented DNA was subjected to Gap filling, A-tailing, and size selection (300–500 bp). Size selected DNA samples were adapter-ligated using NEXTflex™ Bisulfite-Seq barcodes (Bio Scientific Corporation) and were bisulfite converted using EpiTect kit (QIAGEN, Germany) following the manufacturer’s protocol. The product was cleaned up according to recommendations in the QIAGEN EpiTect kit (Gu et al., 2011) and PCR amplified. The PCR product was further enriched by adding 1× SPRI AMPure XP beads. The DNA concentration of eluted and cleaned RRBS libraries was measured using the PicoGreen dsDNA absorbance method. The prepared libraries were sent to the Norwegian Sequencing Centre where sequencing was performed by Illumina HiSeq 4000 in paired-end (2 × 150 bp) mode.

The statistical analyses of the different in vitro sperm quality traits, metabolites, and fertility performance were performed using SAS Version 9.4 for Microsoft Windows (SAS Institute, Cary, NC, United States). The data were tested for normal distribution by the Shapiro-Wilk test. Parameters that did not show a normal distribution were log-transformed to approach a normal distribution prior to further statistical analysis. The mixed procedure in SAS was used to perform the least square means analyses. For the sperm quality parameters, the effect of the explanatory variables on the outcome variables; motility parameters, viability, DFI, HDS, and ATP, were estimated by the following mixed linear model: Y_ijk = µ + C_i + S_j + M_k e_ijk, where: Y_ijk = observation of in vitro sperm parameter per semen sample; µ = overall mean of the in vitro sperm parameter; C_i = fixed effect of storage condition, i = 1 (day one milt), 2 (stored at two degrees) or 3 (stored at eight degrees); S_j = fixed effect of semen state, j = 0 (fresh) or 1 (frozen-thawed); M_k = random effect of male, k = 1 to 8; e_ijk = random error. The same model was used to perform a least square means analysis for the metabolites in sperm and seminal plasma, using the different amines and amino acids as outcome variables. For the fertilization outcome, the same model was used, however, the outcome variables were the fertilization outcome either from a sperm:egg ratio of 75 or 500 × 10³, respectively. The Tukey test was applied for pairwise comparisons between means. The correlation between in vitro sperm quality parameters or metabolites and fertilization outcome was assessed, where correlation coefficients (Spearman) were calculated. The significance level in all statistical tests was set to 0.05.

The sperm viability of cold-stored milt in D1-2°C was on average 95.1% ± 0.7 and decreased significantly in D4-2°C (p < 0.0001) and D4-8°C (p < 0.0002) (Table 2).

Frozen-thawed milt samples showed significantly lower viability than the cold-stored samples (p < 0.0001), for all experimental conditions. There was a significant higher post-thaw sperm viability in D1-2°C/Cryo compared to D4-2°C/Cryo and D4-8°C/Cryo samples (p < 0.002).

Sperm motility of cold-stored milt samples measured as both total (TM) and progressive (PM) motility, were higher in D1-2°C compared to D4-2°C and D4-8°C (p < 0.0001) (Table 2). The TM and PM measured in D4-2°C were not significantly different from D4-8°C. The sperm velocity parameters VAP, VCL, and VSL were higher in D1-2°C compared to D4-2°C and D4-8°C (p < 0.0005). The parameters STR and LIN were not significantly different between D1-2°C and D4-2°C. However, both parameters decreased in D4-8°C (p < 0.0001).

Cryopreservation of milt samples after cold storage did not affect the average sperm motility measured as TM and PM (p > 0.05). The post-thaw TM and PM were higher in D1-2°C/Cryo compared to D4-2°C/Cryo and D4-8°C/Cryo (p < 0.0001). Post-thaw TM and PM did not differ between D4-2°C/Cryo and D4-8°C/Cryo (p > 0.05). Average VAP, VSL, and VCL in frozen-thawed samples increased (p < 0.02) compared to cold-stored samples, except for VCL in D1-2°C/Cryo (p > 0.05). The velocity parameters VAP, VCL, and VSL were higher in D1-2°C/Cryo compared to D4-2°C/Cryo and D4-8°C/Cryo (p < 0.001). Samples cryopreserved after cold storage had higher post-thaw STR and LIN compared to cold-stored samples (p ≤ 0.05). The parameter LIN was lower in D4-2°C/Cryo and D4-8°C/Cryo compared to D1-2°C/Cryo (p < 0.03). Post-thaw STR decreased in D4-8°C/Cryo compared to D1-2°C/Cryo and D4-2°C/Cryo (p < 0.0005).

Sperm ATP content was significantly lower in D4-2°C compared to D1-2°C and was further reduced in D4-8°C (p < 0.0001) (Table 2). The ATP content of frozen-thawed milt samples was not significantly different from the cold-stored samples. Post-thaw ATP content was lower in D4-2°C/Cryo compared to D1-2°C/Cryo and was further reduced in D4-8°C/Cryo (p < 0.0001).

Higher percentages of the sperm chromatin integrity parameters DFI (p < 0.003) and HDS (p < 0.0001) were observed in milt from D4-2°C and D4-8°C compared with D1-2°C. The percentages of DFI and HDS were higher in D4-8°C compared to D4-2°C (p < 0.0001).

Samples cryopreserved after cold storage showed a higher percentage of DFI compared to cold-stored samples (p < 0.05). The post-thaw percentages of DFI and HDS were higher in D4-2°C/Cryo compared to D1-2°C/Cryo (p < 0.0001) and was further increased in D4-8°C (p < 0.05).

There was a significant influence of cold storage duration and temperature using both sperm:egg ratios (Figure 1). For the samples cryopreserved after cold storage, there was a reduction in fertilized eggs from D1-2°C/Cryo to D4-2°C/Cryo and a further reduction in D4-8°C/Cryo (p < 0.05). The cold stored samples in D4-2°C were not significantly different from D1-2°C. However, D4-8°C samples significantly reduced the number of fertilized eggs. Overall, cryopreservation of cold-stored samples resulted in a significant reduction of fertilized eggs compared to cold-stored samples, except for D1-2°C using sperm:egg ratio of 500 × 10³.

Spearman correlations between all in vitro sperm parameters and fertilization rate were significantly different from zero (p < 0.05) using a sperm:egg ratio of 75 × 10³ for both cold-stored milt and milt cryopreserved after cold storage (Supplementary Table S3). The highest correlations were found for the motility parameters TM (r_s = 0.96) and PM (r_s = 0.96), ATP (r_s = 0.93) content, and the kinematic parameters VAP (r_s = 0.91), VCL (r_s = 0.91) and VSL (r_s = 0.91) of frozen-thawed samples. The post-thaw chromatin integrity parameters DFI (r_s = −0.82) and HDS (r_s = −0.66) were negatively correlated with the fertilization outcome. No significant correlations were found for cold-stored samples using sperm:egg ratio of 500 × 10³. For the frozen-thawed samples, all correlations were weaker for fertilizations with a sperm:egg ratio of 500 × 10³ compared with 75 × 10³ (Supplementary Table S3).

The effects of storage duration and temperature on amine and amino acid concentrations were assessed in sperm cells (Table 3) and seminal plasma (Supplementary Table S4) from all experimental conditions. For the cold-stored sperm samples, the concentration of the amines choline and creatine increased in D4-2°C and D4-8°C compared with samples in D1-2°C (p < 0.05). The concentration of L-carnitine increased from D1-2°C to D4-2°C (p < 0.0001) but was not further increased in D4-8°C (p > 0.05). There was a significant overall effect of cryopreservation on all amine concentrations (p < 0.002). The concentration of choline was not significantly affected by storage duration and temperature in samples cryopreserved after cold storage. The concentration of carnitine/L-carnitine was significantly reduced from D1-2°C/Cryo to D4-2°C/Cryo but was not further reduced in D4-8 °C/Cryo.

The concentrations of the amino acids arginine, ornithine, proline, serine, and valine measured in sperm from cold-stored milt samples, significantly increased with both storage time and temperature (p < 0.001). The concentrations of alanine, lysine, phenylalanine, and tyrosine were significantly increased in D4-2°C compared to D1-2°C (p < 0.05). However, there was no significant difference between D4-2°C and D4-8°C samples. The glutamate concentration was significantly higher in D4-8°C, compared to D1-2°C (p < 0.01). For the amino acids glutamine, histidine and tryptophan, the concentration was significantly higher in D4-2°C compared to D1-2°C (p < 0.05). However, there was no significant effect of increasing the storage temperature to eight degrees. Overall, there was a significant effect of cryopreservation for all amino acid concentrations except for asparagine in D4-8°C and ornithine in D1-2°C. For the cryopreserved samples, only the concentrations of alanine (p < 0.05) and GABA (p < 0.02) were significantly reduced during storage. No significant changes were observed after increasing the storage temperature to eight degrees. The results for metabolite concentrations in seminal plasma showed similar patterns to the sperm cell samples, where several metabolites were affected by storage conditions during cold storage (Supplementary Table S4). None of the metabolites in the seminal plasma of frozen-thawed samples were significantly affected by the storage conditions prior to cryopreservation.

Correlations (Spearman correlation) between sperm metabolites and fertilization outcome using a sperm:egg ratio of 75 × 10³ is presented in Table 4. For the cold-stored samples, the highest positive correlations were found for asparagine (r_s = 0.88) and GABA (r_s = 0.86), while choline (r_s = −0.87) showed the highest negative correlation. For the cryopreserved samples, all the amines and several amino acids showed significant correlations to the fertilization outcome. The highest positive correlation was observed for GABA (r_s = 0.90), creatine (r_s = 0.88), and L-carnitine (r_s = 0.85). Arginine (r_s = −0.81) and choline (r_s = −0.78) showed the highest negative correlation.

Overall, minimal variation was observed between libraries regarding retrieved clean data, mapping efficiency, and global CpG methylation levels (Table 5). Using Trim Galore default options, we were able to recover, on average, 99.7% of the reads, corresponding to 83.8% of the total bases in the original reads. Libraries mapped against the Salmo salar reference genome, showed on average 59.3% unique mapping efficiency. Furthermore, the libraries had a high degree of global methylation percentage (largely close to 87.3%). In this study, only CpGs with equal or greater than 10x coverage were considered for methylation analysis. The number of CpG_10X varied between samples from 2 million to 4.5 million Cs, corresponding to an average of 23.6% of all covered Cs in the libraries. To compare the methylation profiles, all samples were considered for correlation and cluster analyses based on CpG_10X methylation values. Pearson’s correlation revealed a high similarity between samples (Pearson’s correlation coefficient > 0.96) and cluster analysis showed that test and control groups from the same individuals always were clustered together (Supplementary Figure S1; Supplementary Table S5).

Differential methylation analysis was performed using CpG_10X after applying a linear regression model and 10% methylation cut-off as well as a q-value < 0.05. This resulted in identifying diverse numbers of DMCs (from 226 to 3483), with different levels of methylation difference ranging from −64% to 59% among different experiments (Table 6). In experiments B and E, test groups were less methylated compared to control groups, while in experiments C and D an opposite trend was observed.

Both hypo- and hyper-methylated Cs in all experiments were separately annotated with different genomic and CpG features. Considering gene features, results showed that the majority of DMCs were annotated with intergenic regions followed by introns (Table 7). Furthermore, a higher percentage of DMCs (both in hypo- and hyper-methylated groups) was annotated to promoter regions compared to exons. Regarding CpG features, on average, at least 95% of filtered DMCs in both hypomethylation and hypermethylation groups were annotated within regions outside of CpG islands (CGI)/CpG shores.

Genes whose TSSs were annotated with DMCs were first identified and subsequently subjected to functional annotation analysis. In the current study, significantly enriched terms were observed for all experiments, except experiment A. For each experiment, the top five significant terms and pathways per group (hypo and hyper) are shown in Table 8. A full list of significant terms and pathways can be found in (Supplementary Table S6). We observed the highest number of significant terms and pathways for experiment C, followed by D, B, and E (Figure 2). The analysis clearly showed that a wide variety of significant terms related to signaling, membrane structure, ion exchange, energy metabolism, cell adhesion, hormone, and embryo development can be linked to genes proximal to DMCs (Supplementary Table S6). The term entitled leucine-rich repeats, was the only mutually identified term in the hypo group of experiments B, C, and D (Figure 2A). No terms or pathways were found mutual among hyper groups between experiments (Figure 2B).