Our multi-tissue study demonstrates the utility of our rhesus macaque model to study the impact of prenatal THC use on fetal development and future offspring health, and to investigate the underlying molecular mechanisms. We focus our discussion on several key findings. First, differential methylation following THC exposure was present in all 5 tissue types, with the greatest effect measured in placenta. Second, prenatal THC exposure altered placental and fetal tissue epigenetic signatures at genes involved in embryonic development that may influence longer-term offspring outcomes. Third, changes in methylation in the placenta were associated with changes in RNA expression in genes enriched for tissue development and morphogenesis-related processes. Fourth, loci differentially methylated with THC were enriched for candidate ASD genes in all tissues. Lastly, we found significant overlap at the gene level when comparing differentially methylated loci in THC-exposed placentas from this study with DMRs in human placentas from pregnancies where the newborn was later diagnosed with ASD.
A total of 573 FDR (< 0.05) DMCs and 403 DMRs were associated with THC exposure in placenta. The top DMC and DMR in placenta were annotated to MEGF10 (multiple EGF like domains 10), a gene highly expressed in brain tissue that encodes a protein critical to synaptic number and function in postnatal brain [55]. Interestingly, SNPs in the transcription regulatory region of MEGF10 have been associated with increased risk of autism in a Chinese Han cohort, and MEGF10 expression was lower in peripheral blood of autistic individuals compared to healthy controls [55]. In this study, we observed increased methylation over the transcription start site with THC exposure in placental tissue, which would suggest potential downregulation of expression. However, MEGF10 expression in our placental RNA-seq dataset was relatively low in all placental samples, in agreement with the Human Protein Atlas [56]. Therefore, although we observed lower expression in the THC group relative to CON, the difference was not statistically significant (log2FC = − 1.08; p value = 0.21). Further work is necessary to determine the functional significance of MEGF10 differential methylation in THC-exposed placentas.
We detected fewer differentially methylated CpGs with THC exposure in fetal tissues after multiple testing correction relative to placenta. This may be because placental DNA contains a higher frequency of partially methylated domains relative to blood and other tissues [57] and CpG sites with intermediate levels of methylation can be measured with greater reliability and precision than at extreme values (i.e., near β = 0 or fully unmethylated and near β = 1 or fully methylated) [58]. This may also be related to a lower relative THC exposure of fetal tissues compared with the placenta. However, the FDR DMCs identified in the fetal tissues warrant further investigation based on their biological relevance in those respective tissues. For example, in cerebellum we identified one hypomethylated CpG in the 3’UTR of neuron-specific gene family member 1 (NSG1; aka NEEP21), which is highly enriched in developing cerebellum [59]. Loss of this gene in mice is associated with increased anxiety-related behavior in certain tasks such as spending 50% less time in the open arms of the elevated plus maze [60]. In the right ventricle of the heart, we identified FDR significant hypomethylation of one CpG upstream of the heart and neural crest derivatives expressed 1 (HAND1) gene, which is essential to the formation of the right ventricle [61]. Additionally in the right ventricle, we observed hypermethylation of one CpG annotated to the neutral sphingomyelinase activation associated factor gene (NSMAF), which may play a role in activation of neutral sphingomyelinase in response to cardiac ischemia or reperfusion injury [62].
We observed several placental DMRs annotated to a large block of protocadherin (PCDH) genes and individual CpGs with decreased methylation that were negatively correlated with increased expression of PCDHB8 following THC exposure. Consistent with our observed effect of THC on placental methylation, a study of placental DNA methylation in autism cases versus control identified several differentially methylated CpGs in this region, the majority of which were hypomethylated [63]. The PCDH gene block is organized into 3 clusters containing PCDHA, PCDHB, and PCDHG genes, which are highly expressed in the developing nervous system where they play an important role in neuronal cellular diversity [64]. In mice, there is evidence that differential expression of Pcdh isoforms is regulated by differential methylation of their promoter regions [65], and in humans dysregulated PCDH methylation has been observed in several neurological and psychiatric disorders [66].
Our findings are also consistent with a recent study that used whole genome bisulfite sequencing to identify novel differentially methylated regions in placenta samples from a prospective ASD study. The study by Zhu et al. [67] included functional characterization of a novel transcript within a hypomethylated block at 22q13.33, renamed NHIP (neuronal hypoxia inducible, placenta associated). Within this region, we observed increased methylation and expression of SHANK3, a gene involved in glutamatergic synapse formation with known mutations, epigenetic dysregulation, and structural variations observed in patients with ASD and other neurological syndromes [68,69,70,71].
We recently reported that these THC-exposed pregnancies demonstrated findings suggestive of placental insufficiency including decreased amniotic fluid volume, placental perfusion, and fetal oxygen availability [72]. The placenta plays a vital role in overall fetal health, development, and growth through the supply of oxygen and nutrients, gas and waste exchange, and endocrine signaling [73, 74]. Impaired placental function can result in adverse offspring outcomes including growth restriction, stillbirth, and miscarriage [75]. The placenta also plays an important role in producing neurotransmitters that may directly affect fetal brain development, and there is growing evidence linking placental dysfunction with adverse neurobehavioral outcomes in offspring [76,77,78]. This intimate relationship between placental function and fetal brain development has led to the coinage of terms such as the “placenta-brain-axis” [79] and “neuroplacentology” [80].
To better understand the potential underlying mediators of our previous findings [72], in the current study we examined correlation of DNA methylation with THC-associated gene expression [72] and examined the potential functional relevance of these eQTM loci using network analysis. Out of 631 genes differentially expressed with THC exposure, 520 were associated with methylation at one or more CpG. Notably, placental eQTM genes were enriched for GO terms related to morphogenesis, response to wounding, and response to oxygen-containing compounds. Taken together, these results suggest that prenatal THC exposure dysregulates placental gene expression and function through effects on the epigenome, and these effects are also likely to impact brain development.
Several studies have also focused on using placental ‘omics to gain mechanistic insight into the relationship between the in utero environment and birth outcomes such as fetal growth, preterm delivery, and birthweight, and infant and childhood health outcomes such as neurocognition and behavior [81, 82]. As observed in our study from prenatal cannabis exposure, dysregulation of placental DNA methylation (DNAm) in response to the in utero environment is a common finding [81]. A prior literature review that included studies of placental DNAm and/or transcriptomics in response to the prenatal environment identified 28 articles meeting their search criteria. In 16 of these studies, there was evidence of mediation or biological plausibility linking these exposures to fetal and infant health outcomes through the placental epigenome [82]. More recently, the term “placenta epigenome–brain axis” was used to describe the relationship between disrupted placental function (linked to placental gene regulation) and neurocognitive function later in life [83]. Altogether, studies from this field of research suggest that neurobehavioral disorders such as ASD, likely originate with dysregulated placental function.
Although the current evidence is not sufficient to conclude that prenatal cannabis exposure is a cause or a risk factor for development of ASD in offspring, there are a few large cohort studies that examine long-term outcomes related to prenatal cannabis exposure longitudinally. A retrospective Canadian study linked pregnancy and birth data to provincial health administrative databases to ascertain child neurodevelopmental outcomes and found an association between maternal cannabis use in pregnancy and increased incidence of ASD in the offspring [30]. However, due to limitations in data availability, this study was not able to account for key risk factors that can increase the likelihood of developing ASD. Our study provides biological plausibility for the findings of increased likelihood of developing ASD reported by this large Canadian study.
Another recent study assessed the effects of maternal cannabis use on psychosocial and physiological measures in young children along with the potential relevance of the in utero environment reflected in the placental transcriptome [35]. This study identified a relationship between maternal cannabis use and transcriptome changes in the placenta as a potential mediator of risk for anxiety-related problems in early childhood. We have now identified DNA methylation as a potential mediator of gene expression changes following prenatal THC exposure related to placental function and offspring neurobehavioral development, which may be related to this study’s observations of increased offspring anxiety-related problems in early childhood with in utero cannabis exposure.
Strengths of our study are that it utilized a translational rhesus macaque model to overcome the limitations of existing studies such as the inability to measure cannabis exposure including timing, duration, frequency, dose, type of product, and administration route. Our study used weight-based THC dosing for rigor and reproducibility, and chose THC edibles as the mode of administration to recapitulate typical human THC use in pregnancy and to examine the direct effects of THC-only on the placental and fetal epigenome without other confounders. Additionally, our study also overcomes the paucity of human epigenetic studies on prenatal cannabis exposure due to lack of feasibility in obtaining fetal tissue and ethical challenges related to research on maternal use of federally illegal drugs, such as cannabis. We were able to extend our investigation beyond the placenta to include 4 additional fetal tissues, not accessible in human pregnancies, for comparison of epigenetic signatures in response to prenatal THC exposure. All placental and fetal tissue were collected at time of cesarean section delivery from pregnancies with similar environmental exposures, including diet. This controlled tissue collection minimizes potential confounders, such as inflammation. We utilized the Illumina MethlyationEPIC platform in order to allow for direct comparison of our results with those from human studies. Future studies should consider whole genome or reduced representation bisulfite sequencing in order to identify potentially novel loci not covered by MM valid EPIC probes. Additionally, this study was limited by a smaller cohort size, which did not provide the power to examine fetal sex as a biological variable.