Opposite alterations of 5HT2A receptor brain density in subjects with schizophrenia: relevance of radiotracers pharmacological profile

The status of serotonin 5HT2A receptors (5HT2ARs) in schizophrenia has been controversial. In vivo positron emission tomography neuroimaging and in vitro post-mortem binding studies have reported conflicting results about 5HT2AR density. Radiotracers bind different receptor conformations depending on their agonist, antagonist or inverse agonist properties. This study investigates 5HT2AR density in the post-mortem prefrontal cortex from subjects with schizophrenia and controls using three radiotracers with a different pharmacological profile. The specific binding parameters of the inverse agonist [18F]altanserin, the agonist [3H]lysergic acid diethylamide (LSD) and the antagonist [3H]MDL100907 to brain cortex membranes from 20 subjects with schizophrenia and 20 individually matched controls were evaluated under similar methodological conditions. Ten schizophrenia subjects were antipsychotic-free at death. Saturation curve analyses were performed by non-linear regression to obtain a maximal density of binding sites (Bmax) and the affinity of the respective radiotracers (Kd). In schizophrenia subjects, 5-HT2AR density was decreased when quantified by [18F]altanserin binding, whereas increased when evaluated by [3H]LSD binding. However, [3H]MDL100907 binding was unaltered. A slight loss of affinity (higher Kd) was observed exclusively in [3H]LSD binding. The findings were more evident in antipsychotic-free subjects than in antipsychotic-treated subjects. In conclusion, a higher proportion of the 5-HT2AR-active functional conformation, which is rather identified by agonist radiotracers, was observed in schizophrenia patients. A consequent reduction of the inactive 5-HT2AR conformation, which is preferentially identified by inverse agonist radiotracers, was also obtained. Antagonist radiotracers do not distinguish between molecular conformations of the receptor, and accordingly, the absence of changes was shown. These results are compatible with the proposed increased functional activity of brain cortical 5-HT2ARs in schizophrenia.


Introduction
A role for serotonin  in the pathophysiology and therapeutics of schizophrenia is supported by converging observations 1 . First, similarities between psychotic states in psychiatric disorders and the effects of lysergic acid diethylamide (LSD) and other 5-HT receptor-mediated psychedelic drugs (e.g., mescaline and psilocybin) have been described 2 . Second, cortical serotonin 5HT 2A receptors (5-HT 2A Rs) seem to be the critical target to induce these psychosis-like responses 3 , and secondgeneration antipsychotics such as clozapine, risperidone and olanzapine, among others, display potent antagonism properties on 5-HT 2A Rs. Furthermore, since 5-HT plays a key role in emotional processing, it has been proposed that dysregulation of 5-HT neurotransmission could underlie the negative symptoms of schizophrenia 1 . The 5-HT receptor brain density is typically assessed in vivo using positron emission tomography (PET) 4 and in vitro using post-mortem tissue homogenates and sections (for a review, see Supplementary Table S1). Evidence shows a lower density of 5-HT 2A Rs in the frontal cortex of antipsychotic-naive schizophrenic patients when evaluated with the very selective (200-to 500-fold higher affinity for 5-HT 2A Rs vs. dopamine D 2 receptors (D 2 Rs)) radiotracer [ 18 F]altanserin [5][6][7] . In contrast, inconclusive results were obtained with other radiotracers such as [ 18 F]setoperone and [ 18 F]N-methyl spiperone [8][9][10][11] , which have 10-to 25fold 5-HT 2A R selectivity vs. dopamine D 2 Rs 7 . Since D 2 Rs are clearly involved in schizophrenia pathophysiology and treatment, the use of these radiotracers with substantial D 2 R affinity is considered a source of bias for 5-HT 2A R detection. On the other hand, important differences among studies have also been obtained from in vitro postmortem studies in the brain of subjects with schizophrenia 1,3,12 . Thus, while some studies described upregulation of 5HT 2A Rs, others pointed towards the absence of alterations or even a down-regulation in the number of binding sites. These apparent discrepancies among post-mortem studies have been considered in the context of different demographic and clinical parameters, the existence of diverse pharmacological treatments and the variety of methodological approaches 3,[12][13][14] . Moreover, as drug-free populations are difficult to obtain for post-mortem studies, the existence of long-term antipsychotic treatment has been considered the main explanatory factor for differences between in vivo neuroimaging studies in drug-naive patients and in vitro findings in post-mortem brain. However, less attention has been paid to the pharmacological properties, such as agonist, antagonist or inverse agonist, of the respective drugs used as radiotracer tools to identify 5HT 2A Rs. The most common 5HT 2A R drugs used to generate radioligands for in vivo PET studies ([ 18 F]altanserin and [ 11 C] M100907) are considered antagonists, and the efforts towards the development of agonist radiotracers have reported limited success 15 (Supplementary Table S1). Although unattended, the pharmacological properties, such as agonist, antagonist or inverse agonist, are determinants for the receptor conformation identified by the radiotracer and, subsequently, for the estimated binding density 16 .
G-protein-coupled receptors (GPCRs) and, among them, 5HT 2A Rs display different molecular conformations that are interchangeable and stay in equilibrium 17 . Thus, the receptor conformation that couples to G proteins is considered to be functionally active and represents the high-affinity state of the receptor, which is preferentially identified by agonist radioligands. Conversely, inverse agonist radioligands show preference to bind the lowaffinity state, i.e., the inactive receptor conformation, which is uncoupled from G proteins. Finally, antagonist radioligands bind with similar affinity to both receptor conformations (Fig. 1). Therefore, the binding of agonist radioligands to the G-protein-coupled conformation should serve as a more accurate measure of 5HT 2A R functions and dysfunctions than antagonist binding 18,19 . In vitro studies have revealed increased functional coupling of 5HT 2A Rs to G proteins in the brain cortex of subjects with schizophrenia without alterations in total values of receptor density 20 . This finding suggests that an imbalance of 5HT 2A Rs towards the high-affinity receptor conformation might be present in schizophrenia, leading to overactive G-protein-dependent signalling. Under this 5HT 2A R overactivity, enhanced agonist radioligand binding should be expected. Conversely, a decreased binding of inverse agonist radioligands to the uncoupled conformation might indicate a decrease of the low-affinity receptor state and prove the existence of an imbalance between coupled and uncoupled 5HT 2A R conformations in schizophrenia, with receptor equilibrium displaced towards the active conformational state (Fig. 2).
Although [  properties on 5HT 2A Rs in post-mortem human brain 13,21 . In fact, [ 18 F]altanserin binding is decreased in the brain of patients with schizophrenia 5,6 and in prodromal at-risk state subjects 22 . In contrast to PET evaluations, studies in post-mortem brain of subjects with schizophrenia have mainly been performed with the agonist [ 3 H]LSD and the partial agonist [ 3 H]ketanserin radiotracers (Supplementary Table S1). Noticeably, among post-mortem studies, those developed in antipsychotic-free schizophrenia cases displayed enhanced 5HT 2A R density in the prefrontal cortex 13,23,24 . Based on all these observations, it could be hypothesized that 5HT 2A R agonists and inverse agonists identify different molecular conformations of this receptor in schizophrenia, leading to opposite findings. However, the 5HT 2A R receptor binding of agonist, antagonist and inverse agonist radiotracers has never been tested under the same methodological conditions.
The aim of the present study was to investigate the 5HT 2A R density in post-mortem prefrontal cortex of subjects with schizophrenia, using three different radiotracers with different intrinsic activity properties (agonist, antagonist and inverse agonist) on this receptor. [ 27 , whose 18 F-and 11 C-labelled forms have been used for PET studies 15,28 . Altanserin is a highly selective 5HT 2A R inverse agonist 21 , commonly considered as antagonist PET radiotracer when labelled with 18 F 15 . The three radiotracers were evaluated on saturation binding experiments in brain cortex membrane homogenates of the same subjects and under similar experimental conditions. We sought to test the hypothesis that different alterations of 5HT 2A R density are obtained in schizophrenia depending on the intrinsic activity properties of each radiotracer. The results would shed light on the unclarified status of brain 5HT 2A Rs in schizophrenia.

Subjects, materials and methods
Post-mortem human brain samples Brain samples were obtained at autopsies performed in the Basque Institute of Legal Medicine, Bilbao, Spain, in compliance with policies of research and ethical boards for post-mortem brain studies. Deaths were subjected to retrospective searching for previous medical diagnosis and treatment using the examiner's information and records from hospitals and mental health centres. A total number of 20 brains from subjects with ante-mortem diagnosis of schizophrenia according to the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV, DSM-IV-TR) were matched to 20 brains from control subjects in a paired design. Control subjects were chosen among the collected brains on the basis of the absence of neuropsychiatric disorders or drug abuse, and an appropriate sex, storage time and post-mortem interval (elapsed time between death and tissue dissection/freezing) to match each schizophrenia subject. Toxicological screening of blood samples of all subjects was performed to determine the presence at death of antipsychotics, other drugs and ethanol. According to the absence or presence of antipsychotic drugs in this toxicological screening, schizophrenia population was divided into two groups, a group of antipsychotic-free (n = 10) and a group of antipsychotic-treated (n = 10) subjects. Schizophrenia and control groups were similar for sex ratio, age, storage time and post-mortem interval values (Supplementary  Table S2). Seventeen out of the 20 subjects in the schizophrenia group committed suicide. Matched control subjects mainly died by accidental causes. Therefore, the mechanism of death indicates that almost all were violent or sudden. Specimens of the dorsolateral prefrontal cortex (Brodmann's area 9) were dissected at autopsy following standard procedures and immediately stored at −80°C until assayed. A complete description of the whole population of subjects with schizophrenia and their individually matched controls can be found in Supplementary   20 . Based on this hypothesis, agonist radioligands should identify higher binding sites and inverse agonist radioligands should label lower binding sites in schizophrenia than in controls. Antagonist radioligands would not discriminate between schizophrenia and control subjects. ketanserin binding density 13,24 and G-protein activation mediated by 5-HT 2A Rs 20,29 (see Supplementary Table S2 for details). The prefrontal cortex was selected as the region of interest based on the morphological alterations associated to schizophrenia 30 and the large expression of 5-HT 2A Rs in the area 5 .

Materials and drugs
MDL100907 (volinanserin) and altanserin were purchased from Sigma-Aldrich. [ 3 H]LSD (86.3 Ci/mmol at delivery time) was obtained from PerkinElmer Life and Analytical Sciences, Inc., and [ 3 H]MDL100907 (80 Ci/ mmol at delivery time) was obtained from ARC Radiochemicals. All other chemicals were obtained from standard sources.

Brain membranes preparation
Brain cortex samples were processed to obtain membrane-enriched homogenates as previously described 13 .

Data and statistical analyses
Data obtained from saturation binding experiments of each subject were analysed by non-linear regression using the GraphPad Prism™ software. The apparent equilibrium dissociation constant (K d ) and the maximum density of specific binding sites (B max ) were obtained. K d values were normalized to log transformation before parametric analyses. All data were subjected to a Grubbs's test in order to detect and reject possible outlier values among experimental groups. Pearson's correlation r coefficient was calculated to test for a possible association between independent covariables (age, post-mortem interval and storage time) and receptor density. When correlation was significant, analysis of covariance was performed controlling for the independent covariable. One-way analysis of variance followed by Bonferroni's post-hoc test was used to compare results between radioligands. Results are expressed as mean±standard deviation (SD) of individual values.
Statistical comparisons between groups were conducted by non-linear curve-fitting coanalysis of all individual binding experiments. The selection between a singlecurve model (absence of differences between groups) and a two-curve model (statistical differences between groups) was made by the extrasum-of-squares F test using GraphPad Prism™. When statistical differences between curves were obtained, further individual contrasts were performed to detect whether differences were attributable to changes in B max and/or K d values between groups 32,33 . The analysis that permitted one or more of the parameters to be shared without a significant increase in the residual variance was taken as the best fit. In this nonlinear analysis, results (B max and K d ) are expressed as the best estimation parameter ± SD. These values obtained from simultaneous non-linear regression analyses were not used for parametric statistical calculations. In all analyses, the level of statistical two-tailed significance was chosen as p = 0.05.

Specific binding sites for [ 18 F]altanserin, [ 3 H]LSD and [ 3 H] MDL100907
The individual non-linear analysis of each radioligand binding fitted to a saturation curve displaying a single specific binding site of high affinity, compatible with selective detection of 5-HT 2A Rs. The receptor density (B max ) in the overall population was similar when identified by [

Comparison between schizophrenia and control groups
The co-analysis of saturation curves obtained with [ 18 F] altanserin showed a statistically significant reduction of the density of the binding sites in the schizophrenia group compared to matched controls. No differences in affinity were detected between schizophrenia and control groups (Table 1). When the presence of antipsychotic drugs was considered, the co-analysis demonstrated a significant reduction of [ 18 F]altanserin binding sites in antipsychoticfree subjects with schizophrenia (B max = 329 ± 24 fmol/mg) vs. matched controls (B max = 410 ± 25 fmol/mg) (p < 0.05) (Fig. 3A). In contrast, subjects with schizophrenia and presence of antipsychotic treatment displayed density values (B max = 376 ± 18 fmol/mg) closer to the respective matched control group (B max = 411 ± 23 fmol/mg) (Fig. 3A). No changes were observed in the affinity parameters (K d = 0.30 ± 0.07 nM in antipsychotic-free schizophrenia group vs. K d = 0.34 ± 0.07 nM in matched controls; K d = 0.33 ± 0.05 nM in antipsychotic-treated schizophrenia group vs. K d = 0.29 ± 0.06 nM in matched controls). As expected, these findings were confirmed by analysis of covariance controlling B max value for age.
The co-analysis of saturation curves obtained with [ 3 H] LSD demonstrated a significant increase of binding sites in subjects with schizophrenia compared to matched controls The co-analysis also demonstrated higher K d values of this radioligand in the schizophrenia group than in controls (Table 1). When the presence of antipsychotic drugs in blood was considered, the enhanced receptor density was maintained in antipsychotic-free schizophrenia subjects (B max = 791 ± 69 fmol/mg) when compared with matched controls (B max = 646 ± 34 fmol/mg) (p < 0.05) (Fig. 3B). Conversely, antipsychotic-treated schizophrenia subjects displayed receptor density values (B max = 735 ± 63 fmol/mg) that did not differ from those in respective control group (B max = 635 ± 30 fmol/mg) (Fig. 3B) For each radioligand, the two sets of data (schizophrenia and control) were first analysed separately. The overall value for the sum of squares and the degrees of freedom (d.f.) was the sum of the individual values of each fit. Next, the two sets of data were pooled and analysed simultaneously constraining them to share one or two common parameters (B max , K d ). The pooled fit yielded values for the sum of squares and degrees of freedom. The analysis that permitted one or two parameters to be shared without a significant increase in the residual variance was taken as the best fit. correlation (r = 0.68, p = 0.04) was obtained in antipsychotic-treated subjects. The co-analysis of curves obtained with [ 3 H] MDL100907 showed no differences of binding sites between subjects with schizophrenia and matched controls. The affinities were also similar ( Table 1). No differences were detected neither in antipsychotic-free schizophrenia subjects (B max = 324 ± 37 fmol/mg; K d = 0.34 ± 0.13 nM) with respective controls (B max = 328 ± 23 fmol/mg; K d = 0.32 ± 0.07 nM) nor in antipsychotictreated subjects (B max = 330 ± 28 fmol/mg; K d = 0.26 ± 0.07 nM) vs. matched controls (B max = 344 ± 22 fmol/mg; K d = 0.25 ± 0.05 nM) (Fig. 3C). Age at death did not influence the results in the analysis of covariance.

Discussion
The present study demonstrates in post-mortem human frontal cortex that alterations of 5-HT 2A Rs observed in schizophrenia are dependent on the pharmacological properties of the radiotracer used to identify this receptor.

Thus, binding assays with an agonist ([ 3 H]LSD), an antagonist ([ 3 H]MDL100907) and an inverse agonist ([ 18 F]altanserin) radiotracer conducted in similar incubation conditions from identical samples lead to different
results. The present study provides evidence that the agonist radioligand binding to 5-HT 2A Rs is increased in schizophrenia, whereas inverse agonist radioligand binding is reduced, and the antagonist radioligand binding remains unaltered. This differential pattern is remarkable in antipsychotic-free subjects, whereas the presence of these drugs in blood tends to reverse the 5-HT 2A R density alterations to control values. Until recently, an equivalent PET study was not feasible, mainly due to the lack of suitable agonist radiotracers for selective 5-HT 2A R identification 36 . However, the recent development of [ 11 C] Cimbi-36 15,18 , a non-selective 5-HT 2A/B/C R agonist 37 , could help to confirm through head-to-head in vivo comparisons between antagonist and inverse agonist radiotracers the findings here reported in the postmortem tissue. Nevertheless, the technical and ethical feasibility of in vivo identification of 5-HT 2A Rs by three different PET radiotracers in the same patient and short time frame is limited. Therefore, in vitro post-mortem studies could help to overcome these drawbacks in the study of neurotransmitter receptor molecular alterations. However, other potentially confounding factors, especially the existence of previous treatment with antipsychotic drugs, add limitations to the conclusions of post mortem studies. In fact, the number of antipsychotic-free subjects included in this type of studies is very limited 38 (Supplementary Table S1).
The increased density of 5-HT 2A Rs identified by the agonist radiotracer [ 3 H]LSD confirms the proposed higher functional sensitivity of this receptor in schizophrenia. Previous studies with the partial agonist radioligand [ 3 H] ketanserin have reported similar findings in antipsychoticfree subjects 13,24 . Endogenous and exogenous agonists bind preferentially to the high-affinity state of the receptor, which represents the active functional conformation coupled to cell signalling pathways (Fig. 1). Recently, the assessment of 5-HT 2A R coupling to G proteins in postmortem frontal cortex has demonstrated an enhanced sensitivity of inhibitory G i1 proteins in response to the agonist (±)DOI (2,5-dimethoxy-4-iodoamphetamine) in schizophrenia 20 . In concordance, the prolactin response to the 5-HT-releasing drug D-fenfluramine 39 , which is considered a functional in vivo test dependent on 5-HT 2A/C Rs activation, is enhanced in drug-free schizophrenia subjects 40 LSD. The compound altanserin has been classically regarded as the selective 5-HT 2A R antagonist. However, in the brain cortex, this drug shows inverse agonist properties 13,21 , which means preferential labelling to the receptor conformation uncoupled from G proteins. In this way, the pharmacological profile of [ 18 F]altanserin explains the reduced 5-HT 2A R density reported in schizophrenia as a reduction of the uncoupled conformation of this receptor. G-protein-coupled (higher affinity for agonists than for inverse agonists) and G-protein-uncoupled (higher affinity for inverse agonists than for agonists) receptor conformations are interchangeable molecular states of GPCRs.
In brains of subjects with schizophrenia, the imbalance between 5-HT 2A R conformations in favour of the Gprotein-coupled state would be expressed as increased density of the agonist [ 3 H]LSD binding and reduced density of the inverse agonist [ 18 F]altanserin binding, as observed in the present study (Fig. 2) 24,42 . Receptor oligomers coexist with monomeric forms (Fig. 4). Dimeric receptor complex crosstalk to each other promoting differential modulation of the ligand access to the respective binding pockets 36 . In fact, the binding of partial agonist and antagonists, as ketanserin and MDL100907, to one of the two binding sites in the 5-HT 2A R homodimer introduces negative cooperativity effects on the propensity of a second molecule of the same drug to bind the dimer 43 . In contrast, the hallucinogenic 5-HT 2A R agonist (±)DOB shows a similar affinity for the two binding sites of the dimer 43 (Fig. 4). This molecular mechanism would be reflected in a two-fold higher density when estimated by radiotracers bound to the full homodimer with respect to the density obtained by radiotracers bound only to one of the monomers that conform to the dimer. The results shown in the present study together with those in previous studies with the radiotracer [ 3 H]ketanserin 13,24 agree with this hypothesis of a homodimeric 5-HT 2A R structure and function. Certainly, the assumption of the receptor oligomerization paradigm should affect future comparisons between radiotracer binding properties.
The observed decline of 5-HT 2A R density with ageing is a repeated finding in previous post-mortem 13,24,44 and PET 45 studies. This profound effect of ageing provides the rationale for experimental designs based on one-to-one individual matching of each schizophrenia case with respective control, as performed here, rather than the usual and less rigorous group-based matching.
The presence or absence at death of antipsychotic drugs in the blood of subjects with schizophrenia represents another relevant confounding factor in radioligand binding studies (Supplementary Table S1). In the present study, the absence of antipsychotic drugs in the toxicological analysis does not mean that these subjects termed as antipsychotic-free were antipsychotic-naive, but rather that they were untreated in the nearest antemortem period. The more [ 18 F]altanserin binding and the less [ 3 H]LSD binding densities in antipsychotic-treated respect to antipsychotic-free schizophrenia subjects suggest that antipsychotic treatment would counterbalance the 5-HT 2A Rs alterations observed in schizophrenia.
Long-term treatment with second-generation antipsychotics modulates 5-HT 2A R expression in animals 14,24 and could modify binding parameters due to residual presence of antipsychotics acting as 5-HT 2A R antagonists 20 . However, the possibility that observed alterations of 5-HT 2A R density (B max ) in schizophrenia represent a consequence of the current or past long-term antipsychotic treatment is improbable. First, the changes of density are more evident in recent antipsychotic-free than in antipsychotic-treated subjects. Second, the differential up-or down-regulation of 5-HT 2A Rs associated with recent antipsychotic treatment in function of the different radiotracers makes unlikely a residual competitive effect between the antipsychotic and the radioligand to bind the receptor pocket. In contrast to density (B max ), the apparent affinity (K d value) was sensitive to the residual presence of antipsychotic drugs, although this effect was only observed for [ 3 H]LSD binding assays. The finding reasserts the use of agonist radiotracers to better detect the 5-HT 2A R occupation by psychedelic drugs 46 . Receptor exploration in drug-free conditions is more feasible by PET imaging than by retrospective post-mortem binding studies. Indeed, most of the post-mortem studies evaluating 5-HT 2A Rs in schizophrenia have been performed in the brain of subjects under antipsychotic treatment, which probably led to inconclusive results (Supplementary Table  S1). Besides this, when schizophrenia subjects were differentiated between those under antipsychotic treatment and those antipsychotic-free, the post-mortem radioligand studies demonstrated up-regulation of brain 5-HT 2A Rs identified by agonist/partial agonist radiotracers 13,20,23,24 . Therefore, in order to discard eventual bias in post-mortem studies, independent and wellmatched groups of antipsychotic-free and antipsychotictreated subjects should be selected and independently analysed.
Another potential confounding factor to consider in the present study is the fact that schizophrenia subjects died mostly from violent suicide mechanisms. Suicide has been proposed as a condition that could influence the evaluation of 5-HT 2A Rs 47 . However, there are several studies in the frontal cortex of suicide victims with a variety of psychiatric disorders supporting that suicide unlikely represents a major confounder in 5-HT 2A R binding studies 1,13,48,49 .
In conclusion, the study and interpretation of 5-HT 2A R dysfunctions in schizophrenia requires a deep knowledge of the pharmacological properties of the candidate radiotracers. The distinction of 5-HT 2A R radiotracers between agonist, antagonist and inverse agonist may shed light on the, up to now, contradictory results. According to the different pharmacological profile, the present results and most of the studies would demonstrate an upregulation of the active functional 5-HT 2A R conformation in the brain of subjects with schizophrenia.
The present results support the hypothesis that 5-HT 2A R molecular conformation and/or the receptor interaction with other synaptic proteins might be altered in schizophrenia. Moreover, as previously described, the antipsychotic treatment seems also to modify the The 5-HT 2A R is expressed as single monomers but also as receptor complexes constituted by two or more receptor units. The binding of a drug to one of the monomers of the complex could induce positive, negative or no cooperativity for the binding to other monomers with the complex. For example, the binding of the antagonist MDL100907 or the inverse agonist altanserin to one of the binding pockets of the homodimer introduces negative allosteric effects on the binding of the same drug to the second binding pocket 43 . This negative cooperativity would prevent the identification of the total number of binding sites. In contrast, the binding of hallucinogenic 5-HT 2A R agonist such as (±)DOB and LSD to receptor homodimers does not modify the estimation of the total number of binding sites 43 . Higher order 5-HT 2A R oligomeric and heteromeric complexes of 5-HT 2A Rs with other GPCRs are also feasible 24,42 . functional state of 5-HT 2A Rs, trying to revert the alterations found in antipsychotic-free schizophrenia subjects. Therefore, the development and in vivo use of agonist radiotracers in antipsychotic-naive patients should be encouraged to validate the 5-HT 2A R overactivity here proposed.