Common HTR2A variants and 5‐HTTLPR are not associated with human in vivo serotonin 2A receptor levels

Abstract The serotonin 2A receptor (5‐HT2AR) is implicated in the pathophysiology and treatment of various psychiatric disorders. [18F]altanserin and [11C]Cimbi‐36 positron emission tomography (PET) allow for high‐resolution imaging of 5‐HT2AR in the living human brain. Cerebral 5‐HT2AR binding is strongly genetically determined, though the impact of specific variants is poorly understood. Candidate gene studies suggest that HTR2A single nucleotide polymorphisms including rs6311/rs6313, rs6314, and rs7997012 may influence risk for psychiatric disorders and mediate treatment response. Although known to impact in vitro expression of 5‐HT2AR or other serotonin (5‐HT) proteins, their effect on human in vivo brain 5‐HT2AR binding has as of yet been insufficiently studied. We thus assessed the extent to which these variants and the commonly studied 5‐HTTLPR predict neocortex in vivo 5‐HT2AR binding in healthy adult humans. We used linear regression analyses and likelihood ratio tests in 197 subjects scanned with [18F]altanserin or [11C]Cimbi‐36 PET. Although we observed genotype group differences in 5‐HT2AR binding of up to ~10%, no genetic variants were statistically significantly predictive of 5‐HT2AR binding in what is the largest human in vivo 5‐HT2AR imaging genetics study to date. Thus, in vitro and post mortem results suggesting effects on 5‐HT2AR expression did not carry over to the in vivo setting. To any extent these variants might affect clinical risk, our findings do not support that 5‐HT2AR binding mediates such effects. Our observations indicate that these individual variants do not significantly contribute to genetic load on human in vivo 5‐HT2AR binding.


efficacy.
A PET study from our group, performed in twins, indicates that cerebral 5-HT2AR binding is strongly genetically determined . Identifying genetic sources of variation in brain 5-HT2AR elucidates how variation emerges and highlights mechanisms mediating genetic effects on neuropsychiatric risk.
In vitro studies are inconclusive regarding these SNP's impact on 5-HT2AR protein expression. The concordant rs6311/rs6313 A/T alleles showed higher expression in some studies (Parsons, D'Souza, Arranz, Kerwin, & Makoff, 2004;Polesskaya & Sokolov, 2002), though this was contradicted by others (Bray, Buckland, Hall, Owen, & O'Donovan, 2004). rs6314 induces a histidine to tyrosine amino acid change, alters 5-HT2AR intracellular signaling (Ozaki et al., 1997), and was associated with lower expression levels in vitro and post mortem (Blasi et al., 2013). Effects of rs7997012 on 5-HT2AR protein levels or function have yet to be reported, though an association with in vivo cerebral serotonin transporter (SERT) binding was observed in humans (Laje et al., 2010). PET provides the unique opportunity to assess whether these SNPs affect 5-HT2AR binding in the living human brain. The specific impact of individual HTR2A SNPs on 5-HT2AR binding has been insufficiently elucidated. Two previous PET studies failed to demonstrate an effect of the HTR2A SNPs assessed here. However, these studies were in small samples (Hurlemann et al., 2008), particularly considering the small effect sizes attributed to singular variants in psychiatry (Bogdan et al., 2017), or only assessed one HTR2A variant (Erritzoe et al., 2009).
In addition, a study in marmosets linked a SERT gene (SLC6A4) promotor haplotype associated with low SERT expression and an anxietyprone phenotype (Santangelo et al., 2016) to reduced brain 5-HT2AR binding (Santangelo et al., 2019). In humans, 5-HTTLPR (rs4795541) and the associated rs25531 variant have been associated with depressive pathophysiology (Caspi et al., 2003) and may affect SERT binding (Praschak-Rieder et al., 2007), though null findings have been observed, including from our lab (Fisher et al., 2017;Murthy et al., 2010). Although associations between 5-HTTLPR and other serotonin receptors have been detected with PET in humans (Fisher et al., 2012;Lothe et al., 2009), the impact on 5-HT2AR binding has yet to be investigated.
Here, we aimed to assess whether the aforementioned HTR2A SNPs (rs6313, rs6314, rs7997012) and the 5-HTTLPR predict neocortex 5-HT2AR binding in the living human brain. To do so, we probed their effects in a uniquely large (n = 197) [ 18 F]altanserin and [ 11 C] Cimbi-36 PET data set available through the Center for Integrated Molecular Brain Imaging (CIMBI) database .
All subjects provided written informed consent and the study protocols were approved by the Ethics Committee of Copenhagen and Frederiksberg, Denmark (KF-11-061-03, KF-01-124-04, KF-01-2006-20, KF-02-058-99, KF-01-001-02, KF-01-156-04, H-4-2012. Healthy subjects included in the CIMBI database are deemed free from primary psychiatric disorders (according to the DSM-IV axis 1 or WHO ICD-10 diagnostic classifications), severe systemic or neurologic disease, pregnancy, head trauma, drug or alcohol abuse, and current or prior intake of psychopharmacologic substances through an interview with a trained clinician. Absence of psychopathologic symptoms is confirmed on the day of the PET-scan using the symptom check list revised SCL-90-R (Derogatis & Savitz, 1999). Additional information on CIMBI database inclusion criteria are detailed in . A largely overlapping dataset was recently utilized by Stenbaek et al., 2019) to assess the impact of openness, a specific personality trait, on 5-HT2AR binding. Erritzoe et al. (2009) previously assessed one HTR2A SNP (rs6311, in perfect LD with rs6313) in a subset (n = 95) of the data presented here.

| MR imaging
Each subject underwent one magnetic resonance imaging (MRI) session for acquisition of a high-resolution, T1-weighted structural brain scan. MRI scans were utilized for coregistration, gray matter-, white matter-, and cerebrospinal fluid segmentation, as well as region of interest (ROI) delineation. MRI scans were acquired using either a Sie-

| PET data processing
As previously described Stenbaek et al., 2019), motion correction utilized the AIR algorithm based on Woods, Cherry, and Mazziotta (1992) and PET scans were smoothed using a 12 or 10 mm within-frame Gaussian filter for the GE-Advance and HRRT scanners, respectively. The AIR algorithm was also used for coregistration of scans acquired on the GE-Advance scanner. For HRRT scans, Statistical Parametric Mapping (SPM8) was used for this purpose.
[ 18 F]altanserin BP P and [ 11 C]Cimbi-36 BP ND were calculated based on Pinborg et al. (2003) and Ettrup et al. (2016), respectively. BP P and BP ND are considered indices of 5-HT2AR density (Innis et al., 2007). The neocortex was defined as the primary ROI and delineated using corresponding subject-level MRI data and PVElab as published by Svarer et al., 2005). PVElab defines the neocortex to include the middle/inferior frontal-, middle/inferior temporal-, superior frontal-, superior temporal-, and pre/post central gyri as well as occipital-, orbitofrontal-, and parietal cortices. Several biological and methodological characteristics motivated selection of this singular ROI for analyses. Neocortex has high 5-HT2AR binding, whereas subcortical levels are low  and show lower reproducibility . [ 11 C]Cimbi-36 has serotonin 2C receptor affinity, relevant in hippocampus (Finnema et al., 2014), confounding genetic associations. In the neocortex, BP P and BP ND values from [ 18 F]altanserin and [ 11 C]Cimbi-36 scans are correlated , supporting pooling of data. Finally, 5-HT2AR binding is strongly correlated between neocortical subregions ( Figure 1). Thus, our focus on a neocortex ROI represents a statistically efficient strategy for evaluating genetic predictors of brain 5-HT2AR binding.

| Genotyping
Genotyping was done on material from either buffy coat lymphocytes or whole blood samples. Samples or extracted DNA were stored at −20 C in the CIMBI biobank until analyzed. DNA extractions were performed with Qiagen DNA extraction kits (Qiagen, Hilden, Germany, https://www.qiagen.com). Quality control of DNA samples utilized Thermo Scientific NanoDrop spectrophotometry (Thermo Fisher Scientific, Waltham, MA, https://www.thermofisher.com). In case of insufficient purity, DNA decontamination was carried out by sodium acetate precipitation, using 0.1 volume of 3 M sodium acetate and 3 volumes of ice-cold 96-99% ethanol. To verify consistent genotype results, five previously genotyped rs6313, rs6314, and rs7997012 samples were re-analyzed with perfect correspondence. rs6311 genotype information was available in 129 subjects and used to confirm perfect LD with rs6313. All other SNPs are in linkage equilibrium based on LDmatrix tool (https://ldlink.nci.nih.gov/?tab=ldmatrix).

| Statistical analyses
Hardy-Weinberg equilibrium (HWE) was evaluated with a chi-square test. HWE was not evaluated for 5-HTTLPR because this variant was an inclusion criterion in some of the original studies.

| Genotype grouping
Subjects were grouped for each HTR2A variant based on genotype, resulting in three groups for each SNP with the exception of the rs6314, for which we identified only one YY homozygote (i.e., rs6313 TT vs. CT vs. CC; rs6314 HH vs. Y-carriers; rs7997012 AA vs. AG vs. GG). For the 5-HTTLPR, L G or S alleles were coded together as an S 0 allele and genotypes were grouped accordingly (i.e., S 0 S 0 , L A S 0 , L A L A ).
This grouping is based on evidence that the L G and S alleles are associated with similarly low SERT expression levels (Hu et al., 2006). See

| Exploratory analyses of genotype effects on insula 5-HT2AR binding
In addition, exploratory analysis probing for effects of 5-HTTLPR genotype on insula 5-HT2AR binding was performed in the n = 197 pooled sample based on animal studies reporting effects of SLC6A4 genotype on 5-HT2AR binding (Santangelo et al., 2019). Again, linear regression analyses and likelihood ratio testing were implemented using the same procedures as for the neocortex ROI.

| DISCUSSION
We did not find evidence for a statistically significant effect of three common HTR2A SNPs (rs6313, rs6314, rs7997012 (Parsons et al., 2004), post mortem (Blasi et al., 2013;Polesskaya, Aston, & Sokolov, 2006;Polesskaya & Sokolov, 2002;Turecki et al., 1999) and in animal studies (Santangelo et al., 2019) demonstrate only very limited translation to 5-HT2AR binding in healthy adult humans. Although nominal effects detected for rs6313 and rs6314 were consistent with related in vitro reports (Blasi et al., 2013;Polesskaya & Sokolov, 2002), all observations were not statistically significant. Thus, despite evidence from twin studies for a substantial impact of genetic load on 5-HT2AR binding , our findings suggest only marginal penetrance of the individual variants considered here. The potential explanations for this disparity differ between the individual SNPs due to their diverging effects on the HTR2A gene and how it is regulated.
The rs6311/rs6313 A/T-alleles have been linked to increased promotor activity (Parsons et al., 2004), mRNA levels (Polesskaya & Sokolov, 2002) and post mortem 5-HT2AR binding (Turecki et al., 1999). In our sample, TC and TT individuals showed nominally increased 5-HT2AR binding relative to CC individuals, but observed group differences were ≤2.5% and not statistically significant. Thus, although we did not reject our null hypothesis, we observed effects nominally similar to previous in vitro and post mortem studies (Parsons et al., 2004;Polesskaya & Sokolov, 2002;Turecki et al., 1999). Importantly, rs6311/ rs6313 may regulate transcription in a manner moderated by epigenetic mechanisms, including gene methylation (Polesskaya et al., 2006) and interaction with transcription factors (Falkenberg, Gurbaxani, Unger, & Rajeevan, 2011;Smith et al., 2013). Thus, we cannot rule out that epigenetic processes obscured main effects of this variant on 5-HT2AR binding. Although epigenetic markers can be assayed from peripheral blood samples, this information is not available for the current data and the correspondence between peripheral blood and relevant neuronal populations is unclear. The above mentioned positive post mortem studies suggest that genotype effects override epigenetic-induced variance. However, negative post mortem studies have also been published (Blasi et al., 2013;Bray et al., 2004). The susceptibility of these variants to epigenetic influence may explain these discrepancies, particularly as the corresponding studies were performed in small samples. In our uniquely large healthy adult sample, we estimate a small, nonsignificant main effect of rs6313 on 5-HT2AR binding.
Previous studies linked rs6314 HY genotype to lower post mortem prefrontal mRNA levels (Blasi et al., 2013). Although not statistically significant (p = .15), we estimated a nominally consistent 10.2% lower 5-HT2AR binding in Y-carriers compared to HH homozygotes.
rs6314 results in an amino acid change (histidine to tyrosine) (Ozaki et al., 1997). rs6313 is also exonic and could thus potentially also alter  (Blasi et al., 2013), yet not statistically significant.
We did not observe a statistically significant impact of rs7997012 on 5-HT2AR binding. rs7997012 A allele was previously associated with improved antidepressant treatment response (McMahon et al., 2006). Neither in vitro nor in vivo effects of rs7997012 on HTR2A gene expression have been reported previously, but an earlier PET study reported a marginal effect on SERT binding (Laje et al., 2010). We estimated 8.9 and 7.1% increases in 5-HT2AR binding in AG and GG groups compared to AA individuals, respectively; AG and GG individuals showed similar 5-HT2AR binding (2% difference). In two cohorts, our lab has reported a nonlinear relation between 5-HT2AR and SERT binding Haahr et al., 2015). Thus, it is plausible that small rs7997012 effects on 5-HT2AR binding mediate effects on SERT, a model that requires further evaluation in independent samples.
in vitro expression studies would support inference of the nominal differences that we observed, resolving rs7997012 functional effects.
We did not observe a statistically significant effect of 5-HTTLPR on 5-HT2AR binding. Our data nominally indicate that L A S 0 and S 0 S 0 status are associated with 4.2 and 7.4% higher 5-HT2AR binding compared to L A L A individuals, a somewhat linear effect. Notably, this remains not statistically significant even when considering L A L A versus S 0 -carriers (p = .27, data not shown), emphasizing a large degree of statistical uncertainty. We performed an exploratory analysis probing for an effect of 5-HTTLPR on insula 5-HT2AR binding, which was also nonsignificant (p = .24). This was motivated by a recent marmoset study showing an SLC6A4 haplotype associated with heightened anxiety and reduced insula 5-HT2AR binding (Santangelo et al., 2016;Santangelo et al., 2019). The convergence with an anxiety phenotype is intriguing considering similar 5-HTTLPR effects (Lesch et al., 1996).  Laje et al., 2010;Polesskaya et al., 2006), allele frequencies, as well as reported associations with disease risk (Caspi et al., 2003;Choi et al., 2004;Petit et al., 2014;Williams et al., 1996) and treatment outcomes of affective and psychotic disorders from candidate gene studies (Anttila et al., 2007;Arranz et al., 1995;Arranz et al., 1998;Blasi et al., 2013;Chen, Shen, & Chen, 2009;Choi, Kang, Ham, Jeong, & Lee, 2005;Kato et al., 2006;McMahon et al., 2006;Minov et al., 2001;Peters et al., 2009). This was done to allow for cautious interpretation of genotype effects vis-à-vis clinical findings. 5-HT2AR binding appears altered in depression (Mintun et al., 2004) and schizophrenia (Rasmussen et al., 2010), and receptor function is modulated by antidepressant and antipsychotic drugs (Artigas, 2013;Meltzer & Massey, 2011 tracer as a covariate. Notably, we observed nonsignificant effects considering each radioligand separately. However, such a split-sample strategy also negatively affects statistical power. Our healthy population varied in, for example, age and BMI. Although this can introduce measurement error and decrease statistical power, we note that collecting 200 datasets at only two PET scanners is a particularly homogenous sample than is otherwise available. Variation in ancestry was addressed by excluding from analyses persons self-reported to be of non-European ancestry. A limitation of the current study is that genetic relatedness and ethnic heterogeneity is not available and not directly modeled. The limitation of evaluating candidate genetic variants as predictors of brain imaging measures has been considered at length (Bogdan et al., 2017). In particular, recent GWAS do not support significant independent roles for the variants assessed here in depression (Wray et al., 2018) or schizophrenia (Pardinas et al., 2018).
However, the sample sizes necessary for GWAS are currently not feasible in conjunction with human in vivo PET data. In the absence of available genome-wide data, focusing on variants selected based on in vitro evidence and clinical candidate gene studies is a potentially useful alternative strategy. In addition, we point out that our outcome measure (5-HT2AR binding) is more proximal to a genetic effect on protein levels vis-à-vis functional and structural brain MRI markers.
In conclusion, this study assessed the effect of three common HTR2A SNPs and the 5-HTTLPR on 5-HT2AR binding in the largest sample size to date (n = 197). Although we do not find evidence for statistically significant differences, we note effects nominally consistent with previous studies linking these variants to 5-HT2AR levels and the 5-HT system more broadly. Our study suggests not more than a limited direct effect of these variants on 5-HT2AR binding in the adult human brain.

ACKNOWLEDGMENTS
The authors thank L. Freyr for assistance in scheduling and data collection at the MR and PET centers. The authors also thank S. Haugbøl and H. Eiberg for genotyping analyses. Then, they would like to thank the John and Birthe Meyer Foundation for the donation of the Cyclotron and PET scanner as well as the Simon Spies Foundation for the donation of the Siemens Trio MRI scanner. This study was supported by a grant from the Augustinusfonden to P. M. F. (19-0489) and by a grant from the Lundbeck Foundation (R90-A7722).

CONFLICT OF INTEREST
Within the last 3 years, M. S. has received travel grants and speaker honoraria from Janssen and Austroplant. All other authors declare no potential conflict of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author and CIMBI database (http:// www.cimbi.dk/db). The data are not publicly available due to privacy or ethical restrictions.