Region-specific and dose-specific effects of chronic haloperidol exposure on [3H]-Flumazenil and [3H]-Ro15-4513 GABAA receptor binding sites in the rat brain

Post-mortem studies suggest that schizophrenia is associated with abnormal expression of specific GABAA receptor (GABAAR) α subunits, including α5GABAAR. Positron emission tomography (PET) measures of GABAAR availability in schizophrenia, however, have not revealed consistent alterations in vivo. Animal studies using the GABAAR agonist [3H]-muscimol provide evidence that antipsychotic drugs influence GABAAR availability, in a region-specific manner, suggesting a potential confounding effect of these drugs. No such data, however, are available for more recently developed subunit-selective GABAAR radioligands. To address this, we therefore combined a rat model of clinically relevant antipsychotic drug exposure with quantitative receptor autoradiography. Haloperidol (0.5 and 2 mg/kg/day) or drug vehicle were administered continuously to adult male Sprague-Dawley rats via osmotic mini-pumps for 28 days. Quantitative receptor autoradiography was then performed post-mortem using the GABAAR subunit-selective radioligand [3H]-Ro15-4513 and the non-subunit selective radioligand [3H]-flumazenil. Chronic haloperidol exposure increased [3H]-Ro15-4513 binding in the CA1 sub-field of the rat dorsal hippocampus (p<0.01; q<0.01; d = +1.3), which was not dose-dependent. [3H]-flumazenil binding also increased in most rat brain regions (p<0.05; main effect of treatment), irrespective of the haloperidol dose. These data confirm previous findings that chronic haloperidol exposure influences the specific binding of non-subtype selective GABAAR radioligands and is the first to demonstrate a potential effect of haloperidol on the binding of a α1/5GABAAR-selective radioligand. Although caution should be exerted when extrapolating results from animals to patients, our data support a view that exposure to antipsychotics may be a confounding factor in PET studies of GABAAR in the context of schizophrenia.

Deficits in GABA neurotransmission, resulting in disruptions to normal patterns of neural oscillatory activity are implicated in the pathophysiology of schizophrenia (Benes, 2010;Benes and Berretta, 2001;Lewis et al., 2012Lewis et al., , 2005. In support of this, quantitative receptor autoradiography studies using [ 3 H]-muscimol, an orthosteric agonist at the GABA binding site on GABA A -BZR, provide consistent evidence for increased binding density in frontal and temporal cortices and the caudate nucleus in post-mortem brain tissue from patients with schizophrenia (Benes et al., 1996;Dean et al., 1999;Deng and Huang, 2006;Hanada et al., 1987;Newell et al., 2007;Verdurand et al., 2013). By contrast, post-mortem studies focusing specifically on mRNA expression of GABA A α-subunits report decreased expression of α1 (Beneyto et al., 2011;Glausier and Lewis, 2011), increased expression of α2 (Beneyto et al., 2011;Volk et al., 2002) and inconsistent results for the α5-subunit (Akbarian et al., 1995;Beneyto et al., 2011;Impagnatiello et al., 1998). A systematic review of positron emission tomography (PET) studies in schizophrenia patients using selective radiotracers for the BZ-site of the GABA A -BZR however found no consistent evidence for altered GABA A -BZR availability in schizophrenia (Egerton et al., 2017). Of note, these post-mortem data come from patients with a long duration of illness and exposure to antipsychotic medication. Similarly in most of the PET studies, the patients were also receiving antipsychotic medication (Egerton et al., 2017). Notably, different antipsychotics can directly alter the binding of ligands to GABA A -BZR, presumably by altering the expression and availability of the receptors (Frankle et al., 2015;Lee et al., 2013). Hence, distinguishing effect(s) of illness from antipsychotic exposure is challenging and medication may represent a significant source of heterogeneity in these data.
Rodent models offer the means to study direct effects of antipsychotic drugs on GABA A R radioligand binding, in the absence of illness effects and other confounds, and with strict control of genetic and environmental factors. Combining such models with quantitative receptor autoradiography (QAR) to assess drug influences on radioligand binding is also highly advantageous. Specifically, this method provides greater spatial resolution as compared to in vivo microPET, whilst keeping within a translational framework for comparison to clinical research using the same radioligands, which other techniques such as histology do not allow (Onwordi et al., 2020). Previous QAR studies in naïve adult rats provide evidence that chronic exposure to different antipsychotics influences the binding of both [ 3 H]-muscimol (indexing GABA A R binding) and [ 3 H]-flunitrazepam (indexing BZ-site binding) in a region-specific manner that is also dependent on the duration of drug exposure, the mode of administration and the sex of the animal (Dean et al., 2001;McLeod et al., 2008;See et al., 1990See et al., , 1989Shirakawa and Tamminga, 1994;Skilbeck et al., 2008Skilbeck et al., , 2007Zink et al., 2004). For example, a 21 day oral exposure to aripiprazole (1mg/kg), olanzapine (1mg/kg) or risperidone (03.mg/kg) increased [ 3 H]-muscimol binding in the striatum and nucleus accumbens of adolescent male rats, whilst in adolescent female rats binding was only increased in the prefrontal cortex (Lian and Deng, 2019). By contrast, a 12-day exposure to haloperidol (1 mg/kg/d) via intraperitoneal injections was reported to decrease [ 3 H]-flumazenil binding in several regions of the rat brain (McLeod et al., 2008). The binding sites for the GABA A -BZR allosteric ligand, [ 3 H]-flumazenil, in the rodent and human brain comprise both "zolpidem-sensitive" and "zolpidem-insensitive" sites, with the latter suggested to correspond to GABA A Rs that contain the α 5 subunit (Mcleod et al., 2002). Consistent with the aforementioned data on [ 3 H]-flumazenil binding, chronic (12d) systemic exposure to haloperidol resulted in a significant reduction in zolpidem-sensitive binding sites (α1,2,3GABA A R (Sancar et al., 2007), but had no effect on the insensitive-binding sites, suggesting a lack of effect of haloperidol on α 5GABA A Rs (McLeod et al., 2008).
Notably, Kapur et al. reported that daily systemic (intraperitoneal) injections of antipsychotics, including haloperidol, result in no or inappropriately low occupancy at central dopamine D2 receptors at trough plasma levels (24 hours post-injection), since the half-life of antipsychotics in rodents is 4 to 6 times shorter as compared to humans (Kapur et al., 2003). This raises questions about the inferences drawn from these previous studies that have used doses unrepresentative of the clinical situation (Kapur et al., 2003). Furthermore, to date, no studies have examined the potential impact of antipsychotic drug exposure using radioligands with greater selectivity for GABA A -BZR containing α1/α5 subunits, such as Ro15-4513 (Lingford-hughes et al., 2002;Maeda et al., 2003). This is of clinical relevance, as a recent PET study in schizophrenia found reduced volume of distribution (V T ) of [ 11 C]-Ro15-4513 relative to healthy controls only in antipsychotic-naïve schizophrenia patients, whilst no group differences were found in antipsychotic-medicated schizophrenia patients relative to healthy controls . Furthermore, convergent lines of evidence from animal models strongly suggest that allosteric modulators of α5GABA A R have potential as novel, non-dopaminergic antipsychotic compounds, by balancing hippocampal excitation via tonic inhibition of pyramidal neurons (Bonin et al., 2007;Caraiscos et al., 2004;Donegan et al., 2019;Gerdjikov et al., 2008;Gill et al., 2011;Hauser et al., 2005;Semyanov et al., 2004;Towers et al., 2004). Importantly, the rescue of amphetamineinduced hyperlocomotion in the methylazoxymethanol acetate (MAM) neurodevelopmental disruption model of schizophrenia by the α5GABA A R positive allosteric modulator (PAM) SH-053-2'F-R-CH3, was abolished following prior exposure to the D2R antagonist haloperidol (Gill et al., 2014(Gill et al., , 2011. In the present study we therefore determined the impact of chronic exposure to haloperidol on GABA A R binding using post-mortem quantitative receptor autoradiography with [ 3 H]-Ro15-4513 to assess α1/α5GABA A R and [ 3 H]-flumazenil to assess BZ-sensitive α1-3;5GABA A R using a validated rat model of clinically comparable drug exposure (Kapur et al., 2003;Vernon et al., 2011). Based on the results of McLeod and colleagues (2008) who observed decreases in zolpidem-sensitive binding sites and no change in zolpidem-insensitive sites after haloperidol exposure, we hypothesized that chronic haloperidol exposure would decrease [ 3 H]-flumazenil binding, with no effect on [ 3 H]-Ro15-4513 binding.

Animals and treatment protocol
Male Sprague-Dawley rats (N=36, Charles River, UK; ~ 10 weeks of age) were administered haloperidol (0.5 or 2 mg/kg/day; haloperidol; n=12/group: Sigma-Aldrich, Gillingham, Dorset, UK) or vehicle (β-hydroxypropylcyclodextrin, 20% w/v, acidified to pH 6 using ascorbic acid; n=12/group) using osmotic minipumps for 28 days (Vernon et al., 2011). Dyskinetic behavior, i.e., vacuous chewing movements, was assessed once at 26 days after the start of haloperidol exposure. This involved a simple measurement of purposeless chewing jaw movements in a 2-minute period, outside the home cage as described previously (Vernon et al., 2011). All experimental procedures were performed in accordance with the relevant guidelines and regulations, specifically, the Home

Quantitative receptor autoradiography with [ 3 H]Ro15-4513 and [ 3 H]flumazenil
On completion of drug or vehicle exposure, rats were culled by rising CO 2 concentration and perfused transcardially with 100-200 ml of cold (+4°C) heparinized saline (50 IU/ml). The brains were then quickly dissected from the skull on a chilled platform, hemisected along the midline and flash frozen in isopentane. A plasma sample was collected from trunk blood for estimation of drug levels (see supplementary material). From one frozen hemisphere, coronal sections (20 µm-thick) from the leftbrain hemisphere (Bregma +2.7 to -7.0mm) were cut in series using a cryostat (Leica CM1950), mounted onto glass slides (Superfrost TM ) and stored at -80ºC until used for autoradiography. The remaining frozen hemisphere was utilized in separate experiments not related to this study.
Sections were pre-incubated at room temperature in Tris buffer (50 mM flumazenil, for specific and non-specific binding, respectively, and exposed for 4 weeks before development.

Statistical Analyses
All statistical analyses were performed in Prism software (v8.0.0 for Macintosh, GraphPad Software, La Jolla California USA, www.graphpad.com). Tables 1 and 2 contain the final n-values per treatment group for each radioligand. In total, four rats were excluded from the complete dataset (n=2 for flumazenil and n=2 for Ro15-4513, respectively) for technical reasons. Specifically, data for one rat was missing due to a broken slide and artifacts were present in the films from the other three rats, which prevented the acquisition of images from these films for analysis. To check for the presence of possible outliers, we applied Grubb's test once (not iteratively) to each dataset. Although some rats were identified as potential outliers from this analysis, we had no biological or technical reason to exclude them and they were thus included in the statistical analysis as described. The data were also checked for Gaussian distribution using the Shapiro-Wilk normality test. Group-level differences in ligand binding were assessed using a mixed-effects model, with ROI as within-subject factor and treatment (vehicle, haloperidol 0.5 or 2 mg/kg/day) as between-subject factor, using the specific binding (nCi/mg) of either [ 3 H]-Ro15-4513 or [ 3 H]-flumazenil as the dependent variable. Vacuous chewing movements scores were analyzed using non-parametric Kruskal-Wallis test (p<0.001). Post-hoc tests were performed where appropriate and corrected for multiple comparisons using the 2-stage set-up method of Benjamini, Krieger and Yekutieli, with the false discovery rate set at 5% (q<0.05) (Verhoeven et al., 2005). Relationships between vacuous chewing movements and ligand binding were modeled using non-parametric Spearman's Rho correlation (2-tailed).

Haloperidol plasma levels and vacuous chewing movement behavior
Administration of haloperidol by osmotic pump achieved plasma levels (mean ± s.d.) of 2.96 ± 0.52 ng/mL and 12.2 ± 1.96 ng/mL, for the 0.5 and 2 mg/kg/day doses, respectively. Stereotypical vacuous chewing movement (VCM) behaviors were significantly different across treatment group (Kruskal-Wallis statistic = 9.98; p<0.001; Fig. S2). Post-hoc testing revealed a statistically significant increase in VCMs in rats exposed to 2 mg/kg/day haloperidol after 26 days exposure, relative to vehicle (p<0.01; q<0.05). There were no statistically significant differences between the haloperidolexposed groups (p>0.05; q>0.05). The development of VCMs was not related to the binding of either [ 3 H]-Ro15-4513 (Table S2) (Table   S2). Haloperidol plasma levels were also significantly correlated with binding of either of the ligands used (Table S3).

Discussion
To our knowledge, this is the first study to investigate the effects of chronic (28d) exposure to haloperidol on GABA A R availability in a receptor subtype-specific manner using quantitative autoradiography. We observed that the specific binding of [ 3 H]-Ro15-4513, a radioligand selective for α 1/5-containing GABA A R is increased, but only in the dCA1 sub-region of the rat hippocampus an effect that was not dose-dependent.
By contrast, chronic exposure to haloperidol robustly increased the specific binding of the non-subtype selective GABA A R radioligand [ 3 H]-flumazenil across the rat brain ROIs examined, irrespective of the dose administered. There were no statistically significant relationships between the specific binding of either radioligand in any brain ROI and either vacuous chewing movements, a proxy measure for haloperidol-induced tardive dyskinesia, or haloperidol plasma levels. Collectively, these data provide evidence that chronic exposure to haloperidol has a limited effect on the availability of α 1/5-containing GABA A R, whilst it robustly affects the availability of α 1-3;5 containing GABA A R as a whole in the naïve male rat brain.
Our findings were not consistent with our original hypotheses, based on prior work in this area. Specifically for [ 3 H]-flumazenil, McLeod and Colleagues (2008) reported a decrease in the specific binding of this ligand in the rat frontal cortex, striatum, parietal cortex, dentate gyrus, CA1, 2 and 3 hippocampus subfields and thalamic nuclei following chronic haloperidol exposure (McLeod et al., 2008). By contrast, our data suggest a generalized increase in [ 3 H]-flumazenil binding across these brain ROIs following chronic haloperidol exposure. Our data are however consistent with increased GABA A -BZR density in cortical areas measured using [ 3 H]-flunitrazepam following chronic haloperidol exposure (Skilbeck et al., 2008(Skilbeck et al., , 2007. With regard to effects of haloperidol on α 1/5-containing GABA A R, McLeod and colleagues (2008) reported no effect of this drug on zolpidem-insensitive [ 3 H]-flumazenil binding, which is suggested to reflect α5GABA A R binding sites (Mcleod et al., 2002;McLeod et al., 2008). By contrast, we observed a localized effect of haloperidol on specific binding of the The mechanism driving drug-induced changes in GABA A R binding availability also remains unclear. Since haloperidol does not have affinity for the GABA A receptor, it is likely that these are indirect as a result of the effects of haloperidol on neurotransmitter release through antagonism of dopamine D2-receptor (D2R) (McLeod et al., 2008). In this context, the role of dopamine in controlling GABA release, in which both D1 and D2R are involved, is well established (Starr, 1987). Increases in GABA also enhance the affinity of GABA A R for BZ-ligands such as flumazenil via a conformational change (Frankle et al., 2009;Miller et al., 1988;Tallman et al., 1978). In contrast, for BZR inverse agonists such as Ro15-4513, increased GABA levels appear to decrease the affinity of GABA A R for this ligand (Stokes et al., 2014). Changes in GABA release during haloperidol exposure thus likely account for the effect of haloperidol on GABA A R availability observed herein and by others (McLeod et al., 2008). Studies of bulk tissue GABA levels in the frontal cortex of schizophrenia patients using proton magnetic resonance spectroscopy ( 1 H-MRS), however, report either no effect (Bojesen et al., 2019;Tayoshi et al., 2010) or a normalisation of elevated GABA levels (de la Fuente-Sandoval et al., 2017) following antipsychotic exposure, although this method only measures bulk tissue GABA and not synaptic levels. In contrast, rodent slice electrophysiology data suggest that D2R mediate synaptic GABA release onto pyramidal neurons in the PFC, whereby GABA release is decreased following dopamine administration (Xu and Yao, 2010). D2R antagonists, such as haloperidol, may therefore be predicted to increase GABA levels in the rodent frontal cortex, which could lead to elevated [ 3 H]-flumazenil binding. In support of this view, GABA-immunoreactivity is increased in the axosomatic terminals of neurons in layers II, III, V, and VI in the frontal cortex of rats exposed chronically to 0.5 mg/kg/day haloperidol over 4 months (Vincent et al., 1994). On the other hand, an in vivo microdialysis study reported a decrease in the extracellular levels of GABA in the rat nucleus accumbens following chronic haloperidol exposure (See et al., 1992), which may perhaps suggest a compensatory upregulation of GABA A R in this region. Supporting these data, increased mRNA expression of GABA A R is reported in the rat NAc after chronic haloperidol exposure, (Pan et al., 2016b(Pan et al., , 2016a. Further studies are however required using additional in vivo methods and post-mortem immunohistochemistry to understand how the effects of haloperidol on GABA A R availability observed herein relate to brain GABA and dopamine levels, and confirm these effects at the protein level, including localisation of these changes to specific cell types. Some limitations of our study should be noted. First, while [ 3 H]-Ro15-4513 binds predominantly to diazepam-sensitive GABA A R sites, it also binds to a diazepaminsensitive site in the cortex and hippocampus with lower affinity (Turner et al., 1991). as olanzapine and clozapine, on GABA A R availability in the rat brain (Farnbach-Pralong et al., 1998). In our previous studies comparing the effects of haloperidol and olanzapine on rat brain volume, radioligand binding (e.g. [ 3 H]-UCB-J) and post-mortem cellular markers, we have consistently found no clear differences between these compounds (Cotel et al., 2015;Onwordi et al., 2020;Vernon et al., 2014). Whilst we therefore have no reason to believe that olanzapine for example, would not induce similar effects to haloperidol, this should be explicitly tested in future studies. Fourth, the effect of haloperidol on [ 3 H]-Ro15-4513 binding in the dCA1 did not show true dose-dependence, with the effect only seen at the lower dose (0.5 mg/kg/d). The plasma levels of HAL in this group were slightly below the generally accepted therapeutic range (3 ng/ml c.f. 5-20 ng/ml), whilst the higher dose (2 mg/kg/d) fell within this, but had no effect on [ 3 H]-Ro15-4513 binding in the dCA1. This study also had relatively small numbers and replication of this finding in a larger number of subjects will therefore be an important advance. Addressing the effect of sex as a biological variable is also necessary, since only male animals were used.
In conclusion, our findings suggest that exposure to haloperidol (and perhaps other antipsychotics) should be considered when measuring and interpreting global GABA A R binding availability using non-selective radioligands in the context of schizophrenia.
Our data also provide initial evidence that haloperidol exposure has only a modest influence on α1/5-GABA A R selective radioligand binding. Of note, these data were collected in naïve animals, lacking any features relevant to the pathophysiology of schizophrenia. In this context, a recent PET study reported reduced [ 11 C]-Ro15-4513 V T in the hippocampus of antipsychotic-naïve schizophrenia patients, whilst no group differences were found in a second cohort of patients who were taking antipsychotics . One interpretation of these data is that antipsychotics may have a "normalizing" effect and it is tempting to speculate that this is consistent with the increases seen in GABA A R availability herein. Since the mechanism underlying these effects remains however, unknown, until this is better understood or convincingly refuted, one should be very cautious in drawing any clinical inferences from our data.
Consequently, further studies to determine the effects of different antipsychotics on GABA A R availability in animal models reflective of genetic, environmental or pharmacological risk factors for schizophrenia are required to confirm or refute this hypothesis.