The obesity associated polymorphism rs10767664 occurs within intron 3 of the BDNF gene and within a sequence highly conserved between humans, mice and amphibians. Because of its importance in obesity (5) and the importance of the BDNF gene in appetite control (28) we explored the hypothesis that allelic variation at the rs10767664 locus functionally changed the activity of an uncharacterised cis-regulatory element that controls aspects of the expression of the BDNF gene involved in appetite. We were intrigued to find that rs10767664 occurred within a 494 base pair region of highly conserved DNA (BE5.1; chr11:27,725,730 − 27,726,223) specifically within a 240bp region of the extreme conservation that demonstrated > 90% conservation to amphibians; a depth of conservation spanning 360 million years of divergent evolution (29) (Fig. 1). BE5.1 lay 2.4kb 5’ from promoter IV of the BDNF gene (BP4), within BDNF intron 3, (Supplementary Figure S1) and 20kb away from the recently reported + 3kb enhancer (15).
The T-allele of BE5.1 acts as an enhancer of BP4 activity in primary hypothalamic cells. To determine the effects of different alleles of BE5.1 on BDNF promoter activity we cloned 494 bp surrounding the human BE5.1 sequence using Hi-fidelity PCR and reproduced both human alleles using site directed mutagenesis. Both allelic variants were then cloned into a luciferase reporter plasmid containing the BP4 promoter region (Fig. 2A), that represented the closest promoter to BE5.1, as previously described (14). We also chose this promoter as BP4 had previously been shown to be active in hypothalamus and amygdala in transgenic animals (14). These reporter plasmids were transfected into primary rat hypothalamic cells by magnetofection as previously described (14) and incubated for 48 hours. After this time cells were lysed and lysates analysed by dual luciferase assay. In this way we were able to demonstrate that the T- allele of BE5.1 could act as an enhancer of BP4 activity but that the obesity associated A-allele could not (Fig. 2B).
BE5.1 knockout animals are healthy and viable. We used CAS9/CRISPR genome editing to delete the BE5.1 enhancer from the mouse genome by cytoplasmic injection of single guide RNA (sgRNA) and CAS9 mRNA into 1-cell C57BL/6 mouse embryos as previously described (19, 25–27). We were able to generate a single heterozygous female animal (BE5.1KO+/−) which was outbred on a C57BL/6 background for two generations to produce a colony of heterozygous male and female BE5.1 KO animals. Sequence analysis of the 5 most likely off target events for each sgRNA (Supplementary Tables S1 and S2) was carried out as previously described (22) and we were unable to detect any evidence of off-target changes within the genome of this line (Supplementary data Figure S2). These animals were then bred together to produce homozygous wild type (WT) and BE5.1KO animals, selected on the basis of PCR analysis, which proved to be healthy and viable (Supplementary data table S5).
Deletion of BE5.1 has no significant effect on expression of BDNF isoforms in the hypothalamus. To determine the possible effects of the BE5.1 deletion on the expression of BDNF mRNA splice forms at a tissue specific level we recovered total RNA from the hypothalamus of male and female WT and BE5.1KO littermates. We then used quantitative reverse transcriptase PCR (QPCR) to quantify the expression of different isoforms of BDNF derived from different promoters (Fig. 3A and Supplementary table S3) in the hypothalamus where the expression of BDNF is known to influence appetite (30). However, we were unable to detect any significant difference in the expression of BDNF isoforms expressed from promoters 1–5 (exon 1–5) or promoter 9 (exon 9) in these samples (Fig. 3B-F). In order to determine the effects of deleting BE5.1 on the expression of the receptor for BDNF we also undertook QPCR analysis of the TrkB receptor but found no change in its expression either (Fig. 3G).
Deletion of the BE5.1 enhancer had no significant effect on weight gain and only marginal effects on food intake. Because the of the association of the rs10767664 polymorphism within BE5.1 with obesity, we explored the hypothesis that deletion of BE5.1 from the mouse genome would produce a significant change in food intake or weight gain. We allowed WT and BE5.1KO littermates access to a choice of either standard CHOW or high fat diet (HFD) for 28 days. Animals were weighed at regular intervals and analysed at the beginning and end of 28 days using echo MRI analysis. Although we detected a marginally significant increase in the intake of CHOW diet in BE5.1KO animals (Fig. 4B and E), consistent with previously described effects of decreased in BDNF expression (28), we could not identify any significant differences in the intake of high fat diet, weight gain or fat distribution in these animals after 28 days suggesting that BE5.1 had only marginally significant effects on food intake and no significant effect on weight gain over the period of time studied (Fig. 4A, C, and D).
Deletion of BE5.1 significantly up-regulates expression of BDNF exons in the female amygdala. In addition to the hypothalamus we also recovered tissues from the amygdala of male and female WT and BE5.1KO littermates. We used QPCR to quantify the expression of different exons of BDNF in the amygdala where the expression of BDNF is known to influence anxiety-like behaviour (31). Although we observed a trend in the data towards increased expression of BDNF exons 4 and 9 in these samples in males the increase was only marginally significant for exon 4 (by t-test) and fell just short of significant for exon 9 (p = 0.051) (Fig. 5A-E). However, compared to WT littermates we detected a significant increase in the expression of exons 1, 2 and 4, (excluding exon 3), and exon 9 in amygdala tissues derived from BE5.1KO female mice (Fig. 5A-E).
Female BE5.1KO mice demonstrate increased anxiety-like behaviour in the elevated zero maze. Because we detected increased levels of 4 different BDNF exons in the amygdala of female BE5.1KO mice we asked whether deletion of BE5.1 would influence anxiety-like behaviour in these animals. We subjected male and female BE5.1KO and WT littermates to the elevated zero maze (EZM) which is a robust method of detecting anxiety-like behaviour in mice (32). However, we were unable to detect any significant difference between the time spent in the open quadrants (Fig. 6A), total distance travelled (Fig. 6B), Line crossings (Fig. 6C) speed (Fig. 6D) or freezing time (Fig. 6E) between male WT or BE5.1KO littermates in the EZM. By contrast, we observed highly significant changes in all of these behaviours in female BE5,1KO mice who spent significantly less time in the open quadrants of the EZM, showed reduced line crossings and overall speed with a highly significant increase in freezing time that accounted for much of the perceived lack of mobility in these animals (Fig. 6A-E).
Deletion of BE5.1 decreases marble burying in female mice. In order to further test the effects of BE5.1 deletion on anxiety-like behaviour in mice we exposed male and female WT and BE5.1KO animals to a second anxiety test called the marble burying test (MBT); a well characterised test of anxiety-like behaviour in rodents (33). Mice habitually bury marbles placed in their environment; a behaviour which decreases with increased anxiety. The advantage of the MBT is that, unlike the EZM; where animals become habituated to the test, the MBT can be repeated many times with the same animals with minimal change in the numbers of marbles each animal buries(33). Briefly, the animal were placed in a cage with 3cm depth of fresh bedding (wood shavings) onto which 20 x 1.5 cm glass marbles were placed. After 30 minutes the animal was removed from the cage and the number of marbles buried were recorded. In the case of male animals, no significant difference was observed in the number of marbles buried by either the wild type or the BE5.1KO animals where both groups buried and average of 6–7 marbles each (Fig. 7A). However, consistent with our observations of the EZM, we detected a significant difference between the female WT and BE5.1KO whereby WT female mice buried an average of 5 marbles each whereas female BE5.1KO animals only buried between 1–2 marbles within the same time frame (Fig. 7A).
The effects of BE5.1 deletion on female anxiety like behaviour could be reversed using diazepam. Although marble burying has been widely used as a test of anxiety there are concerns that extrapolating anxiety-like behaviour from marble burying may be compromised by the parallel manifestation of obsessive-compulsive disorder related behaviours(33). To better understand the contribution of anxiety-like behaviours in our results we undertook these tests again in females after treating a subset of them with either a vehicle (saline) or saline containing 10mg/kg− 1 of the anxiolytic drug diazepam. As before, female WT and BE5.1KO mice treated with the vehicle displayed a significant difference in the numbers of marbles buried within the 10 minutes of the study (Fig. 7B). However, following treatment with diazepam, both groups of mice buried significantly more marbles and displayed no significant difference in the numbers of marbles buried (8–9; Fig. 7B).
The rs10767664 polymorphism is associated with feelings of anxiety and risk taking in human cohort studies. Next, we explored whether there was evidence for a functional role of rs10767664 in anxiety-like behaviour in humans. To this end, we extracted association results for rs10767664 from all genome-wide association studies (GWAS) included in the MRC IEU OpenGWAS database [https://europepmc.org/article/ppr/ppr199104]. We restricted the GWAS results to traits related to anxiety, worry, risk taking and sexual history. In total, we extracted 51 associations (Supplementary Table S4) of which 4 relating to anxiety/worry were statistically significant after adjustment for multiple testing and found that the A allele of rs10767664 was statistically significantly associated with anxious feelings and feelings of worry (FDR < 0.05, Table 1) consistent with the role for BE5.1 in anxiety observed with our mouse model. In addition, the analysis also emphasised a high association with higher levels of overall risk taking, earlier age at first sexual intercourse and with a higher lifetime number of sexual partners (Supplementary Table S4).
eQTL analysis of the rs10767664 polymorphism suggest a role for BE5.1 in regulating the expression of the BDNF-AS transcript. The current study has demonstrated that the BE5.1 sequence acts as an enhancer in primary cells but that deleting BE5.1 from the mouse genome resulted in an increase in the expression of BDNF exonic mRNA in amygdala tissues in female mice. On the face of it, these observations appear to contradict each other as deleting an enhancer should not result in an increase in gene expression. For this reason, we asked whether different alleles of the rs10767664 polymorphism had an effect on the expression of genes surrounding BE5.1 (including the BDNF gene) in humans. We interrogated the GTEx database with the rs10767664 polymorphism and found a significant change in the expression of a number of genes flanking the BDNF locus including METTL15 (p = 0.03 in amygdala) and LIN7C (strongly affected in vascular tissue (Aorta; p = 4.5x10− 7, tibial artery; p = 1.2x10− 4) and less strongly in esophagus (p = 1.7x10− 4), muscle (p = 4.9x10− 3) and visceral adipose tissue (p = 0.006). In contrast, we were unable to find data supporting a change in the expression of the BDNF gene in GTEx as a result of the rs10767664 polymorphism. Intriguingly, rs10767664 had a significant effect on the expression and splicing of a gene called BDNF-AS; that expresses an alternatively spliced antisense RNA whose exon 5 sequence is complimentary to exon 9 of the BDNF mRNA transcript (34) (Table 2). This observation is especially interesting as siRNA knockdown of BDNF-AS results in an increase in both BDNF mRNA and protein (34). The most significant effects of rs10767664 on BDNF-AS expression occurred in the nucleus accumbens; the addiction centre of the brain, and a number of cortical regions including the cingulate cortex that is a critical component of the limbic system and controls anxiety and fear (Table 2). We also observed that allelic variation at the rs10767664 polymorphism had significant effects on the splicing of the BDNF-AS primary transcript (Table 2).
QPCR analysis of BE5.2 deletion mice supports its role in modulating expression of BDNF-AS. In order to explore the possibility that BE5.1 supports the expression of BDNF-AS in mice and was consistent with the observed human eQTL analysis, we examined the expression of the BDNF-AS gene in the amygdala of wild type and BE5.1KO mice using quantitative PCR as previously described (34). We demonstrated that QPCR could detect the BDNF-AS transcript in wild type mice but that there was a significant reduction in the detection of the BDNF-AS transcript in BE5.1KO mice (Fig. 8). This observation is consistent with the hypothesis that at least part of the role of the BE5.1 enhancer is to control BDNF expression through modulation of the BDNF-AS transcript.