The anxiety and ethanol intake controlling GAL5.1 enhancer is epigenetically modulated by, and controls preference for, high-fat diet

Excess maternal fat intake and obesity increase offspring susceptibility to conditions such as chronic anxiety and substance abuse. We hypothesised that environmentally modulated DNA methylation changes (5mC/5hmC) in regulatory regions of the genome that modulate mood and consumptive behaviours could contribute to susceptibility to these conditions. We explored the effects of environmental factors on 5mC/5hmC levels within the GAL5.1 enhancer that controls anxiety-related behaviours and alcohol intake. We first observed that 5mC/5hmC levels within the GAL5.1 enhancer differed significantly in different parts of the brain. Moreover, we noted that early life stress had no significant effect of 5mC/5hmC levels within GAL5.1. In contrast, we identified that allowing access of pregnant mothers to high-fat diet (> 60% calories from fat) had a significant effect on 5mC/5hmC levels within GAL5.1 in hypothalamus and amygdala of resulting male offspring. Cell transfection-based studies using GAL5.1 reporter plasmids showed that 5mC has a significant repressive effect on GAL5.1 activity and its response to known stimuli, such as EGR1 transcription factor expression and PKC agonism. Intriguingly, CRISPR-driven disruption of GAL5.1 from the mouse genome, although having negligible effects on metabolism or general appetite, significantly decreased intake of high-fat diet suggesting that GAL5.1, in addition to being epigenetically modulated by high-fat diet, also actively contributes to the consumption of high-fat diet suggesting its involvement in an environmentally influenced regulatory loop. Furthermore, considering that GAL5.1 also controls alcohol preference and anxiety these studies may provide a first glimpse into an epigenetically controlled mechanism that links maternal high-fat diet with transgenerational susceptibility to alcohol abuse and anxiety. Electronic supplementary material The online version of this article (10.1007/s00018-020-03705-6) contains supplementary material, which is available to authorized users.


Background.
In addition to their involvement in obesity (1,2) there is also evidence that diets high in fat are associated with an increase in anxiety and substance abuse in subsequent human generations (3,4). These human studies are supported by observations that high-fat diet and maternal obesity increase substance abuse in the offspring of rats (5) and anxiety related behaviour in C57BL/6 mouse models (6). Therefore, in addition to the mechanisms regulating the intake of high-fat diets, there is also a priority to uncover the genomic and epigenetic mechanisms that link high-fat diet to anxiety and substance abuse disorders.
Nutrient intake (fats and ethanol) and mood are controlled by the expression of neuropeptides in regions of the brain that include the hypothalamus and amygdala (7). The expression of one of these neuropeptides; galanin, a 30 amino acid peptide encoded by the GAL gene, contributes to the regulation of fat intake in animals (2,8-10) plays a role in modulating mood (11)(12)(13)(14) and regulates ethanol intake (15)(16)(17)(18). Considering its important role in these processes, much remains to be determined regarding the mechanisms that regulate the tissue-specific expression of the GAL gene. In order to address this knowledge gap we previously used comparative genomics to identify a highly conserved enhancer sequence (GAL5.1) that lay 42 kilobases (kb) 5' of the GAL gene transcriptional start site in humans and demonstrated its activity in galanin expressing cells of the hypothalamus including the PVN, dorsomedial hypothalamus (DMH) and arcuate nucleus (19,20). Analysis of two polymorphisms in the GAL5.1 enhancer using the UK Biobank suggested a mechanistic link between allelic variants of these polymorphisms and alcohol abuse when stratified against sex (male) and anxiety (20). Subsequent use of CRISPR genome editing, to disrupt the mouse GAL5.1 enhancer (mGAL5.1KO), demonstrated a role for mGAL5.1 in the tissue specific expression of the Gal gene in hypothalamus and amygdala as well as the modulation of alcohol intake and male anxiety-like behaviour (20).
Because of the known link between high-fat diet and increased susceptibility to anxiety and alcohol abuse in offspring of mothers fed high fat diet, we explored the hypothesis that the 4 4 anxiety and alcohol intake modulating GAL5.1 enhancer could be epigenetically influenced by environmental stimuli that include early life stress and high-fat diet. We also explored the effects of DNA-methylation on GAL5.1 activity and its interactions with, and response to, known stimuli such as PKC agonism and EGR1 expression. Finally, we asked whether deleting the GAL5.1 enhancer has any effect on metabolism or fat intake. We discuss our findings in the wider context of a role for GAL5.1 genetics and epigenetics in modulating fat intake as well as the possible role of altered GAL5.1 activity in increasing susceptibility to anxiety and alcohol abuse in future generations.

Methods.
Animal studies. All animal studies were performed in full accordance with UK Home Office guidelines. Male and female homozygous wildtype and mGAL5.1KO age matched littermates were single housed under standard laboratory conditions (12 h light/12 h dark cycle), in plastic cages with food and water available ad libitum, depending on the experiment.
Epigenetic effects of early life stress by maternal deprivation. Wildtype C57BL/6 females were housed with wild type C57BL/6 males and allowed ad-libitum access to standard CHOW diet. Once females became pregnant males were removed and females were allowed to litter down. Immediately after birth mothers were removed from the cages for two hours a day for the first 12 days as previously described (21). Once weaned litters were humanely sacrificed by euthatol injection and brain tissues (hypothalamus, hippocampus and amygdala) recovered and rapidly frozen on dry ice.
Epigenetic effects high-fat diet studies. Wildtype C57BL/6 wild type animals were housed with wild type C57BL/6 males and allowed ad libitum access to a choice of low fat diet (LFD; 22.03 kcal% protein, 68.9kcal% carbohydrate and 9.08 kcal% fat) or high fat diet (HFD; 20 kcal% protein, 20 kcal% carbohydrate and 60 kcal % fat; Research Diets Inc.) in different hoppers. Both hoppers were weighed regularly to ensure intake of HFD and LFD. Once females became pregnant males were removed and females allowed to litter down. After weaning (3-4weeks), animals were humanely sacrificed by euthatol injection and brain tissues (hypothalamus, hippocampus and amygdala) recovered and rapidly frozen on dry ice. DNA extraction bisulfite conversion and pyrosequencing. Genomic DNA was extracted and purified with the AllPrep DNA/RNA Mini Kit (Qiagen) and optimized for 10-15 mg brain tissue as per manufacturer's instructions. The concentration and purity of genomic DNA was determined using a Nanodrop One C (Thermo) and 500 ng of genomic DNA was bisulphiteconverted using the EpiMark Bisulfite Conversion Kit (New England BioLabs) as per manufacturer's instructions. The mouse GAL5.1 enhancer region analysed (chr19:3,441,054-3,441,445, GRCm38/mm10) contained 8 CpG sites (see Figure 1). Primers were used to amplify two regions of the mouse Gal5. Diet, weight gain, metabolism) with 80% power using one-way ANOVA and/or general linear modelling. Statistical significance of data sets were analysed using either two way analysis of variance (ANOVA) analysis with Bonferroni post hoc tests or using one tailed or two tailed unpaired parametric Student t-test as indicated using GraphPad PRISM version 5.02 (GraphPad Software, La Jolla, CA, USA).

Results.
Levels of CpG methylation (5mC/5hmC) within GAL5.1 varies significantly between brain regions. We have previously shown that the GAL5.1 enhancer controls anxiety and alcohol intake (20). This provided us with a unique opportunity to determine the effects of environmental factors such as dietary changes and stress on 5mC/5hmC levels within an enhancer region shown to control behaviours with a direct effect on human health. We first asked whether levels of DNA-methylation (5mC/5hmC) within GAL5.1 differed between brain regions or were similar. To address this question, DNA from 3 different brain tissues (hypothalamus, amygdala and hippocampus) was recovered from mouse pups and bisulfite converted to detect levels 5mC and 5hmC within GAL5.1 analysed using pyrosequencing. We observed that 5mC/5hmC levels within GAL5.1 varied considerably between different brain tissues such that in hippocampus 5mC levels only reached 23-32% within GAL5.1. However, in amygdala and hypothalamus 5mC levels were between 35 and 50% ( Figure 1A).

Levels of CpG methylation (5mC/5hmC) within GAL5.1 is not affected by early life
stress. Because changes in CpG methylation (5mC/5hmC) is known to occur within enhancers as a result of early life stress and diet (21,23), we subjected newborn wildtype C57BL/6 mouse pups to maternal deprivation for 2 hours a day for the first 12 days of their lives as previously described (21). After weaning of these mice, DNA from 3 different brain tissues (hypothalamus, amygdala and hippocampus) was recovered, bisulfite converted and levels of 5mC/5hmC within GAL5.1 analysed using pyrosequencing. We did not observe any significant changes in 5mC/5hmC levels in any of the tissues recovered from animals subjected to early life stress (Fig   1B-D).
GAL5.1 displays significantly altered 5mC/5hmC levels in animals whose mothers had access to HFD. We provided pregnant mice with a choice of low-fat or high-fat diet during pregnancy and during the rearing of their pups. Following weaning, pups were humanely sacrificed and genomic DNA was recovered from tissues dissected from hypothalamus, hippocampus and amygdala. This DNA was subjected to bisulfite conversion and 1 0 1 0 pyrosequenced to determine changes in 5mC levels within the GAL5.1 enhancer in the presence of access to HFD. We were unable to detect any changes in GAL5.1 5mC levels in DNA derived from hippocampal tissues of animals whose mothers had access to high-fat diet ( Fig 1B). However, significant changes in CpG methylation were observed in DNA derived from the hypothalamic regions of animals whose mothers had been exposed to high-fat diet such that there was a significant increase in methylation in males and a clear divergence in methylation between males and females ( Figure 1C). This increase in male GAL5.1 5mC levels and sexual divergence was more pronounced in DNA derived from amygdala ( Figure 1D).
The EGR1 transcription factor physically interacts with the GAL5.1 enhancer at a highly conserved EGR1 binding site. We have previously shown that the GAL5.1 enhancer responds to expression of the early growth response factor (EGR1) transcription factor protein (20). However, these experiments stopped short of establishing a direct physical interaction. In order to determine their relationship we transfected SHSY-5Y neuroblastoma cells with equimolar equivalents of a luciferase vector containing either the complete human GAL5.1GG sequence (pLucGAL5.1GG) or GAL5.1GG lacking 10 base pairs corresponding to the highly conserved EGR1 binding site within GAL5.1(pLucGAL5.1ΔEGR1). It is important to point out that the aim of this experiment was not to determine luciferase activity but to determine levels of EGR1 interaction with GAL5.1 using the luciferase gene as a transfection control. In parallel, we also transfected an empty expression vector (pcDNA) or the same vector also expressing the cDNA of the EGR1 transcription factor (pcDNA-EGR1). We also treated these cells with the PKC agonist PMA. After incubation for 24 hours we used a ChIP-verified EGR1 antibody on cell lysates derived from transfected SHSY-5Y cells to immunoprecipitate DNA which had bound to the EGR1 protein and, after reversing cross links, assayed the amounts of GAL5.1 DNA that had formed a complex with EGR1. Using primers specific for the human GAL5.1 enhancer we used qPCR to demonstrate that the anti-EGR1 antibody immunoprecipitated significant amounts of GAL5.1 enhancer DNA and that this interaction increased in the presence of EGR1 expression or PKC activation (Fig 2A). In contrast, we were unable to immunoprecipitate little, if 1 1 1 1 any, GAL5.1 DNA from cells which had been transfected with the plasmid missing the 10bp containing the highly conserved EGR1 binding site (Fig 2B). Transfections were normalised using quantitative PCR against the firefly luciferase gene. We concluded that EGR1 physically interacted with GAL5.1 and that this interaction was stimulated by PKC signalling. In addition, our data also suggests that EGR1 binding within GAL5.1 only occurred within the highly conserved EGR1 binding site.
Methylation of the GAL5.1 enhancer suppresses its activity and its response to EGR1 binding and PKC activation. We next asked what effects CpG methylation would have on the activity of the GAL5.1 enhancer, its response to EGR1 expression and PKC activation which we have shown activates GAL5.1 (19,20). We first cloned the human GAL5.1 (GG) enhancer into the pCpGfree-promoter-Lucia luciferase vector that lacks CpG dinucleotides and contains the EF-1α promoter. We then subjected this vector to CpG methylation using the bacterial SssI enzyme and transfected methylated or unmethylated plasmid into SHSY-5Y neuroblastoma cells in the presence of an expression vector expressing the EGR1 transcription factor ( Figure 2C). Levels of plasmid methylation were monitored using digestion by the HpaII enzyme which is sensitive to methylated CpG sites ( Figure S1). These cells were then cultured for 24 hours in the absence or presence of the PKC agonist PMA. Due to evidence of transinteractions between our renilla control and the pCpG-lucia vector (probably through the EF-1α promoter), we devised an alternative transfection control based on quantitative PCR of the Lucia reporter gene. We observed that the GAL5.1 enhancer increased expression of Lucia luciferase compared to the empty vector but that this increase was negated by CpG methylation ( Figure 2C). The previously reported stimulatory effects of either EGR1 co-expression or activation of the PKC pathway (19,20 ) were similarly reduced by GAL5.1 methylation ( Figure   2C). Interestingly, we observed that the activity of the non-methylated GAL5.1 enhancer was not significantly different from that of the methylated enhancer co-stimulated by both EGR1 expression and PKC activation ( Figure 2C) suggesting that, together, EGR1 and PKC stimulation can overcome much the effects of 5mC on GAL5.1 activity.
1 2 1 2 mGAL5.1KO mice do not demonstrate significant changes in weight gain or intake of CHOW diet. We had previously shown that CRISPR deletion of GAL5.1 from the mouse genome produced mice that drank significantly less ethanol and suffered less anxiety (19,20). In order to explore whether disruption of the GAL5.1 enhancer in mice had any significant effect on low fat food intake (CHOW) or weight gain we monitored the weight gain and CHOW intake of male and female homozygous wild type (WT) and GAL5.1KO animals. We found that, after 18-24 weeks, neither male nor female animals demonstrated significant differences in weight gain compared to wild type animals ( Figure 3A). We also found that GAL5.1KO animals ate similar levels of CHOW diet to WT animals ( Figure 3B).

mGAL5.1KO mice do not demonstrate significant metabolic differences to WT
animals. In order to explore the possibility that disruption of GAL5.1 significantly altered metabolism we subjected mGAL5.1KO and wild type male and female littermates to 116 hours of metabolic analysis using sealed TSE cages that monitored variables such as O2 consumption, CO2 production (ml/hour), energy expenditure ration (kcal/hour), energy balance (Kcal/hour). Over the 116 hours of the analysis we detected little or no significant difference in O 2 consumption (Fig 4A), CO 2 production (Fig 4B) or energy expenditure (Fig 4C) in male and female mice. However, we detected a significant decrease in the respiratory exchange ratio ( Fig   4D) in male animals that suggested higher activity levels. We observed an increase in distance travelled by male GAL5.1KO animals ( Figure 4E) in addition to an increase in overall speed ( Figure 4F) consistent with higher levels of exploratory behaviour associated with the reduced anxiety phenotype previously reported (20).

mGAL5.1KO mice exhibit decreased preference for high-fat diet. Previous studies
demonstrated that deletion of exons 2-6 of the Gal gene in 129Ola/Hsd mice using ES-cell targeting (24) caused a significant reduction in the intake of high fat diet in these animals compared to wild type littermates but had no significant effects on protein or carbohydrate intake (10). Because we have previously shown that disruption of the GAL5.1 enhancer resulted in a significant decrease of Gal mRNA expression in hypothalamus we tested the hypothesis 1 3 1 3 that CRISPR disruption of mGAL5.1 would affect intake of HFD in these mice. We provided singly housed, age-and sex-matched, littermate wild type and mGAL5.1KO animals with a choice of low fat diet (LFD; 6% of calories from fat) or high fat diet (HFD; 60% of calories from fat) and monitored intake of LFD and HFD over 23 days. Both male and female mGAL5.1KO mice consumed significantly less HFD overall compared to wild type littermates (Fig 5A).
Analysis of the total intake of LFD demonstrated no significant differences between the intake of wild type and mGAL5.1KO animals (Fig 5B).
1 4 1 4 Discussion. Studies have suggested that increased maternal fat intake and obesity in both humans (3,4) and animal models (5,6) increase susceptibility to substance abuse and increased anxiety in resulting offspring. In the current study we explored the hypothesis that a contributing mechanism linking maternal high-fat diet and anxiety/substance abuse in offspring could involve the epigenetic modification of tissue specific enhancer regions through DNAmethylation to form 5-methylcytosine (5mC) or 5-hydroxy-methyl cytosine (5hmC) (25-27). Our decision to explore a role for 5mC/5hmC in this process was based on previous studies were environmental challenges such as high-fat diet and early life stress have been shown to change levels and distributions of 5mC/5hmC in the genomes of experimental animals (21,28) and alter enhancer activity through interaction with DNA binding proteins (29).
We focussed our attentions on the methylation of an enhancer sequence; called GAL5.1, that we have previously shown is responsible for supporting expression of the GAL gene in the hypothalamus and amygdala and which also controls ethanol intake and anxiety in males (19,20). We first asked whether environmental factors associated with poverty and deprivation in Western countries; namely early life stress by maternal deprivation or consumption of high fat diet, altered levels of DNA-methylation within the 11 CpG dinucleotides contained within the GAL5.1 enhancer sequence. We first noted significant differences in 5mC/5hmC levels in GAL5.1 derived from different tissues such that GAL5.1 from hippocampal tissues had the lowest 5mC/5hmC levels and hypothalamus and amygdala demonstrated the highest levels.
These are interesting observations as the vast majority of large human cohort epigenome studies have based their analysis of levels and distribution of 5mC/5hmC on DNA derived from peripheral blood. However, given the observed differences in 5mC/5hmC levels observed between different brain tissues in the current study we suggest that caution should be exercised when interpreting studies based on the extrapolation of 5mC/5hmC levels in DNA derived peripheral blood to specific regions of the brain.
Although the repressive effects of 5mC on promoter activity through mechanisms involving the recruitment of methyl-CpG-binding proteins (MBD) and histone deacetylase  (37). It is interesting, in this context, that the bisulfite sequencing protocol used in the current study is unable to differentiate 5mC from 5hmC so that it is as likely that the increased levels of 5mC observed within GAL5.1 as a consequence of maternal HFD intake actually reflects an increase in 5hmC levels. The possible role of 5hmC in upregulating enhancer activity, and our inability to differentiate 5mC and 5hmC using bisulfite conversion, may also explain the observations of increased 5mC/5hmC levels in hypothalamus and amygdala compared to the hippocampus that parallel the levels of Gal mRNA expression, which are significantly higher in hypothalamus and amygdala compared to hippocampus.
Alternative methods of differentiating between genomic 5mC and 5hmC are now available which could help clarify the situation.
Previous analysis of the transcription factors that influence activity of GAL5.1 suggested the involvement of the EGR1 transcription factor by virtue of the fact that co-expression of EGR1 induced an up-regulation of transcriptional activity driven by the GAL5.1 enhancer (20). We next designed another unique experiment to determine the effects of 5mC on the activity of the GAL5.1 enhancer and its response to EGR1 binding and PKC agonism. To achieve this we cloned the GAL5.1 enhancer into a luciferase reporter vector (pCpGfree-lucia) that contained no CpG dinucleotides. This allowed us to selectively methylate CpG dinucleotides within the GAL5.1 enhancer, using the SssI enzyme, without affecting the vector backbone. These studies showed that CpG methylation of the GAL5.1 enhancer has a significant repressive effect on its activity in SHSY-5Y cells and confirmed previous studies demonstrating the repressive effects of 5mC on regulatory activity. However, our studies could not explore the effects of 5hmC on enhancer activity, which has been associated with increased Previous studies have shown that the binding affinity of the EGR1 transcription factor to DNA is not affected by 5mC (38). However, we observed that 5mC suppressed GAL5.1 activity even in the presence of EGR1. This suggests that the binding of another transcription factor; who's binding to DNA is affected by CpG 5mC, is critical to the normal functioning of GAL5.1.
Identifying this transcription factor will be a major goal of subsequent analyses. transcription factor in modulating levels of 5mC/5hmC at the GAL5.1 locus it is also interesting that recent studies have identified EGR1 as a major recruiter of TET proteins to specific loci (39) a consideration which we will also consider in the design of future studies.
In the wider context of the relationship of maternal high-fat diet to increased susceptibility to substance abuse and anxiety, the current study suggests the possible mechanistic involvement of the GAL5.1 enhancer in this process. We have previously shown that GAL5.1 governs ethanol intake and anxiety related behaviour in mice that is mirrored by a significant association between alcohol abuse and increased anxiety in humans. In the current study we show that high-fat diet causes a significant change in 5mC/5hmC levels that are known to affect enhancer activity and that, in turn, GAL5.1 governs the decision to eat high-fat diet but also affects anxiety and alcohol intake. Taken together, we propose that maternal high-fat diet 1 8 1 8 induced changes in GAL5.1 5mC/5hmC levels may alter GAL5.1 activity in subsequent generations in such a way as to affect the ability of GAL5.1 not just to alter fat intake, but to also affect anxiety and the decision to drink alcohol.

Conclusions;
Whilst much remains to be done to differentiate the effects of environmental influence on the distribution of 5mC and 5hmC in GAL5.1 and how 5mC and 5hmC differentially affect GAL5.1 activity at a tissue specific level, the fact remains that we have identified a compelling epigenetic mechanism that may link high-fat diet to the modulation of anxiety and alcohol intake. These unique studies also provide an important stepping stone in our voyage to discover the influence of environmentally modulated regulatory mechanisms in the development of neuropsychiatric disorders Acknowledgements A. R. M. was funded by BBSRC project grant (BB/N017544/1).

Declaration of interest. None of the authors declare any conflicts of interest
References.  t  o  s  h  a  p  e  t  h  e  b  r  a  i  n  m  e  t  h  y  l  o  m  e  d  u  r  i  n  g  d  e  v  e  l  o  p  m  e  n  t  a  n  d  u  p  o  n  n  e  u  r  o  n  a  l  a  c  t  i  v  i  t  y  .  N  a  t  C  o  m  m  u  n  1  0  ,  3  8