Sperm DNA methylation alterations from cannabis extract exposure are evident in offspring

CE exposure timing impacts magnitude of methylation change

The mean library insert size for the WGBS was 195.9 bp (SD 8.3); mean coverage was 20.2 (2.1); and the mean GC content was 22.7 (0.2). We compared methylation differences from the late exposed (LE) versus control dataset to the methylation differences in the early exposed (EE) versus control dataset to determine the role of the 56-day wash-out period on sperm DNA methylation changes. No CpG sites remained significant following conservative Bonferroni correction, so we imposed a methylation difference threshold on the top 10 K sites, retaining only those with a > 10% methylation difference in the LE relative to control datasets. This resulted in 3321 nominally significantly differentially methylated CpG (dmCpG) sites. We then analyzed methylation at those same 3321 dmCpG sites in the EE relative to control dataset. Regardless of the exposure timing, the direction of methylation change at these sites is largely the same. Linear regression of the data showed significant correlation between LE and EE mean methylation differences relative to controls (Fig. 1A, p  < 0.0001, R2 = 0.82).

Fig. 1figure 1

The impact of exposure timing on DNA methylation. A Methylation differences at the top 3321 dmCpG sites between the late exposed (LE) and control animals are plotted on the x-axis. Methylation differences between early exposed (EE) and control animals at those same CpG sites are plotted on the y-axis. Negative values represent hypomethylation relative to controls, while positive values represent hypermethylation relative to controls. Linear regression shows significant agreement between the direction of methylation change across the exposure groups (p < 0.0001, R2 = 0.82). Of the top 3321 dmCpG sites between LE and control animals, we separated those that were hypomethylated relative to controls (B) from those that were hypermethylated relative to controls (C). For B and C, the methylation difference at each CpG site is plotted on the y-axis. The x-axis represents the methylation difference at a specific CpG site between the LE and control animals and the methylation difference at that same CpG site between the EE and control animals, connected by the solid line. This demonstrates that regardless of the direction of methylation change, the magnitude of the mean methylation difference at each CpG site is significantly reduced between EE and control sperm relative to the difference present between the LE and control sperm (p < 0.0001, Bonferroni corrected)

Methylation changes are diminished following cannabis abstinence

We separated the data based on the direction of methylation change present at the 3321 CpG sites from the LE compared to control dataset. This resulted in 1844 hypomethylated CpG sites and 1477 hypermethylated CpG sites in the exposed animals relative to the controls. Comparison of the methylation change at each CpG site shows that the magnitude of the methylation difference relative to controls is greater in the LE sperm than it is at the same CpG sites for the EE sperm for both the hypomethylated (Fig. 1B) and hypermethylated CpGs (Fig. 1C).

The mean methylation difference relative to controls across the 1844 hypomethylated CpG sites was 15.5% in the LE dataset which contrasts significantly with the 8.5% difference at the same CpG sites in the EE dataset (Fig. 1B, p < 0.0001). Similarly, there was a 13.9% mean methylation difference relative to controls across the 1477 hypermethylated CpG sites in the LE dataset as compared to a 7.5% difference at the same CpG sites in the EE dataset (Fig. 1C, p < 0.0001). These differences remained significant following Bonferroni correction for the number of CpG sites analyzed.

Methylation changes occur at genes involved in development

Ingenuity Pathway Analysis (IPA) software was used to interrogate the genes associated with the top dmCpGs from both datasets. Briefly, for those CpG sites with a greater than 10% methylation difference relative to controls, an unadjusted p-value threshold of 5.0X10−5 was implemented in both the LE and EE datasets to restrict the number to those with the greatest significance. This resulted in a list of 744 CpG sites from the LE dataset and a distinct list of 317 CpG sites from the EE dataset. Notably, only five CpG sites were in common between these two lists. IPA recognized and included 492 and 203 LE and EE dataset genes, respectively, in the downstream analyses.

We began with the IPA-generated list of 500 significant functions categories and their associated annotations for the LE dataset. Given the overrepresentation of background information associated with cancer in the IPA Knowledge Base, we removed categories associated with cancer from our analysis, as has been reported previously [19]. We also removed terms associated with only a single gene. This left 289 categories and their associated disease or function annotations. We then analyzed the data by first looking at categories with ten or more annotations (Fig. 2A, top). These categories included “cardiovascular system development, function, and disease”, “embryonic development”, “nervous system development, function, and disease”, and “developmental disorder”. We also identified the top ten most significant disease or function annotations (Fig. 2A, bottom).

Fig. 2figure 2

Top IPA disease and functions. (A-top) Pie chart of disease and function categories that have 10 or more associated terms from the LE dataset. (A-bottom) graph of the top ten most significant disease or function annotations from the LE dataset. (B-top) Pie chart of disease and function categories that have 10 or more associated terms from the EE dataset. (B-bottom) Graph of the top ten most significant disease or function annotations from the EE dataset. For both A and B bottom graphs, the Y-axis represents the annotation, and the x-axis is the -log of the p-value. The colors correspond to the category to which this disease or function was annotated (represented by the color legend on the bottom)

We next focused on the significant dmCpG-associated genes following the wash-out period in the EE dataset. The workflow for the IPA analysis for the EE dataset is as described above, and resulted in 319 categories remaining. Categories with ten or more diseases or functions annotated (Fig. 2B, top) included “connective tissue development, function and disorders”, “developmental disorder”, “embryonic development”, and “nervous system development, function, and disease”. The top ten most significant disease or function annotations included neurotransmission, cognition, spatial learning, and exploratory behavior (Fig. 2B, bottom).

Validation of select WGBS results

From both the LE and EE datasets, distinct genes of interest were identified for independent validation. Sites were selected based on > 10% methylation difference, having ≥ 2 significantly dmCpG sites within a 100 bp region, the associated gene having a role in early life developmental processes, and the CpG site being located within the body of the gene. The following genes were selected: Homeobox Protein Hox-B9 (Hoxb9), Methyltransferase Like 11B (Mettl11b), Slit Guidance Ligand 2 (Slit2), LDL Receptor Related Protein 1 (Lrp1), Citron Rho-Interacting Serine/Threonine Kinase (Cit), Synaptotagmin (Syt17), Synaptonemal Complex Protein 3 (Sycp3), Gamma-Aminobutyric Acid Type A Receptor Subunit Beta2 (Gabrb2), Oncostatin M (Osm), 2-Phosphoxylose Phosphatase 1 (Pxylp1), Glutamate Ionotropic Receptor NMDA Type Subunit 2A (Grin2a), Synapsin III (Syn3), Netrin G1 (Ntng1), and Metastasis Suppressor 1-Like Protein 1 (Mtss1l). We confirmed performance of each of the pyrosequencing assays using defined mixtures of unmethylated and methylated DNAs (R2 = 0.85–0.99, p < 0.0001–0.025, Additional file 1: Fig. S1).

There were no significant methylation changes found in the sperm of the LE rats compared to the sperm of the control rats at Hoxb9, Mettl11b, Slit2, Lrp1, Cit, Grin2a or Syn3 (Additional file 2: Fig. S2). A two-tailed t-test showed a significant change in DNA methylation at Syt17 by pyrosequencing at one of the two CpG sites initially identified via WGBS, however the direction of methylation change differed (Additional file 2: Fig. S2). Similarly, for Sycp3 significant methylation changes were identified in sperm by pyrosequencing at three of the four sites identified by WGBS, but in the opposite direction (Additional file 2: Fig. S2). These genes were therefore not included in subsequent analyses.

Two-tailed t-tests for Osm, Gabrb2, Pxylp1, Ntng1, and Mtss1l, confirmed significant methylation changes relative to controls by pyrosequencing, in the same direction as the WGBS for the LE (Fig. 3A-E top, p < 0.05–0.005). We then broadened our analysis to assess methylation at all the CpG sites captured by the pyrosequencing assay to see if CE exposure similarly affected neighboring sites. We included sperm from both exposure groups to determine how the timing of the exposure impacted DNA methylation changes. A two-factor ANOVA—one factor being exposure status and one being CpG site—showed a significant main effect of exposure for Osm, Gabrb2, Pxylp1, Ntng1, and Mtss1l (p < 0.0001 for all genes, Fig. 3A-E bottom). Post hoc tests revealed no additional significant methylation changes for Osm.

Fig. 3figure 3

Bisulfite pyrosequencing of F0 sperm. Bisulfite pyrosequencing of CpG sites initially identified via WGBS as significantly differentially methylated between LE sperm and controls for A Osm, B Gabrb2, C Pxylp1, D Ntng1, E Mtss1l. (top) CpG sites on the x-axis correspond to those that were identified as significant by WGBS, and the measured methylation level is on the y-axis. Each point represents one individual. (bottom) CpG sites are labeled on the x-axis, mean methylation levels are plotted on the y-axis ± SEM. CpG sites labeled with “*” on the x-axis correspond to those identified via WGBS that were validated in the “top” part of the panel and are labeled with the same CpG number. Black, controls; blue, EE; green,  LE. *p < 0.05, **p < 0.01, ***p < 0.005

For Gabrb2, in addition to validating WGBS methylation changes at CpG site 2, we detected significant hypomethylation at CpG site 1 in sperm from the LE animals relative to the controls (p < 0.05). At CpG sites 1 and 2, there were significant methylation differences between control and EE sperm, though the magnitude was less than that of the control and LE sperm (p < 0.05). For Pxylp1, posthoc tests were significant at all CpG sites and in both exposure groups (p < 0.005). For Ntng1, post hoc tests confirmed significant methylation changes across all three CpG sites analyzed (p < 0.05–0.005), but only for the EE group. There was no significant methylation difference between LE and control sperm for Ntng1. Lastly, for Mtss1l, post hoc tests revealed significant loss of methylation present in the sperm of the EE group at eight of the nine CpG sites analyzed by pyrosequencing (p < 0.05–0.001), but no significant changes in the LE group.

Heritability of cannabis-induced sperm methylation changes

To determine whether methylation changes in sperm of CE-exposed males were detectable in the next generation, we analyzed F1 tissues for methylation at CpG sites that were validated by pyrosequencing. We first examined Pxylp1 in F1 sperm by pyrosequencing of both exposure groups given the abundant levels of this gene’s protein product in the epididymis and in mature spermatids [25]. A two-factor ANOVA revealed a significant effect of exposure on DNA methylation in F1 sperm (p < 0.0001). Post hoc analysis revealed a significant loss of methylation in F1 sperm from animals born to LE cannabis fathers compared to controls (p < 0.05, Fig. 4A), and the difference approached significance when comparing the F1 sperm from EE and control offspring (p = 0.06).

Fig. 4figure 4

Bisulfite Pyrosequencing of F1 tissues. Bisulfite pyrosequencing detected heritable changes at A Pxylp1 in F1 sperm; B Mtss1l in F1 hippocampus; and C Mtss1l in F1 NAc. CpG sites labeled with “*” correspond to those identified via WGBS in the LE v C or EE v C datasets. Black,  control offspring; blue, EE offspring; green,  LE offspring. Two-factor ANOVA (factor 1 = exposure status, factor 2 = CpG site) were significant for all genes. Significance indicators on graphs represent post hoc two-tailed t-tests. #p = 0.05, *p < 0.05, **p < 0.01

We next analyzed changes in DNA methylation in F1 brain tissues for genes with neuronal functions. There were no significant methylation changes for Ntng1 in the hippocampus or nucleus accumbens (NAc; data not shown). For Gabrb2, methylation changes were not consistent with what was observed in F0 tissues (data not shown). Therefore, we did not continue analysis of this gene given that we could not attribute this methylation change to a change in paternal sperm.

We then analyzed methylation changes at Mtss1l in offspring hippocampus (Fig. 4B) and NAc (Fig. 4C) tissues. We observed a significant effect of paternal exposure on DNA methylation in both F1 tissues (p < 0.0001). Like the methylation changes present at these CpG sites in F0 sperm, post hoc tests revealed significant losses of methylation present in EE offspring relative to control offspring. In the hippocampus, four of the nine CpG sites analyzed were significantly hypomethylated relative to controls in the EE group (p < 0.05–0.01). This included one of the CpG sites (CpG site 1) that was initially identified via WGBS as being hypomethylated in the EE paternal sperm. In the NAc, EE offspring were hypomethylated relative to controls for four of the nine CpG sites analyzed (p < 0.05). This included the other CpG site that was initially identified via WGBS (CpG site 5) as being hypomethylated in the EE paternal sperm.

Relationships between DNA methylation and gene expression

Our finding that there are heritable changes in DNA methylation for Mtss1l in brain tissues prompted us to investigate whether those methylation changes are functionally related to changes in gene expression. We first examined this relationship in hippocampal tissue of the EE and control offspring at Mtss1l CpG site 1—one of the two sites initially identified via WGBS as being affected in paternal sperm. Pearson correlations showed no significant relationship between DNA methylation and gene expression for control or exposed offspring (Additional file 3: Fig. S3A). We stratified this analysis by sex (Additional file 3: Fig. S3B and S3C) given the known sex differences in hippocampal function [26, 27] and still observed no significant relationships between DNA methylation and gene expression.

In the NAc, we examined the relationship between DNA methylation at CpG site 5 (the second site initially identified via WGBS in sperm) and gene expression for the control and EE offspring. There were no significant relationships between methylation and expression when both sexes were analyzed together (Fig. 5A). However, known sex differences in the NAc [28, 29] led us to again stratify the analysis by sex. In males (Fig. 5B), control offspring showed a significant inverse methylation-expression relationship (p < 0.05, R2 = 0.61), while exposed offspring had a significant positive methylation-expression relationship (p < 0.05, R2 = 0.53). In females (Fig. 5C), control offspring showed a significant positive methylation-expression relationship (p < 0.05, R2 = 0.52), while exposed offspring showed an inverse though non-significant relationship.

Fig. 5figure 5

Relationship between DNA methylation at CpG site 5 and gene expression for Mtss1l in the NAc. Relationship between DNA methylation and gene expression for Mtss1l in A all sexes combined; B males only; and C females only. Pearson correlation R square and p-values are reported. Control offspring,  black circles with solid regression line; EE offspring,  blue triangles with dotted regression line

Paternal CE exposure induces cardiomegaly in his offspring

There is growing evidence associating prenatal cannabis exposure and teratologies in babies and children and early life exposure to cannabis has been associated with cardiovascular defects in epidemiologic and animal studies [14, 30, 31]. We measured heart weights and body weights of each of the F1 offspring from the EE, LE and control fathers and normalized heart weight to body weight. One-factor ANOVA revealed a significant effect of paternal exposure on offspring heart weight (p = 0.0039). Post hoc tests showed significant increased heart weight relative to controls for both the EE offspring (p = 0.0013), and the LE offspring (p = 0.0099) (Fig. 6A). Based on physiological differences in cardiovascular function and disease in males and females [32, 33], we separated this analysis by sex. A one-factor ANOVA in females (Fig. 6B) showed a significant effect of paternal exposure on offspring heart weight (p < 0.05). Post hoc tests showed significant increases in heart weights for both EE (p < 0.05) and LE (p < 0.05) offspring. In males (Fig. 6C), a one-factor ANOVA just approached significance for the effect of paternal exposure status on offspring heart weight (p = 0.05). Post hoc t-tests showed a significant increase in heart weight relative to controls only in the EE offspring (p < 0.05).

Fig. 6figure 6

Paternal preconception exposure to CE is associated with cardiomegaly in offspring. Offspring heart weight as percent bodyweight is represented on the y-axis (in grams) for A all animals; B females only; and C males only. Black, control offspring; blue, EE offspring; green, LE offspring. ANOVA p-value is from a one-factor ANOVA for effect of paternal exposure status on offspring heart weight. Post hoc t-test values are reported between EE and control, and between LE and control

留言 (0)

沒有登入
gif