Terminal type-specific cannabinoid CB1 receptor alterations in patients with schizophrenia: A pilot study

Abstract

Background:

Individuals with schizophrenia are at elevated genetic risks for comorbid cannabis use, and often experience exacerbations of cognitive and psychotic symptoms when exposed to cannabis. These findings have led a number of investigators to examine cannabinoid CB1 receptor (CB1R) alterations in schizophrenia, though with conflicting results. We recently demonstrated the presence of CB1R in both excitatory and inhibitory boutons in the human prefrontal cortex, with differential levels of the receptor between bouton types. We hypothesized that the differential enrichment of CB1R between bouton types – a factor previously unaccounted for when examining CB1R changes in schizophrenia – may resolve prior discrepant reports and increase our insight into the effects of CB1R alterations on the pathophysiology of schizophrenia.

Methods:

Using co-labeling immunohistochemistry and fluorescent microscopy, we examined total CB1R levels and CB1R levels within excitatory (vGlut1-positive) and inhibitory (vGAT-positive) boutons of prefrontal cortex samples from ten pairs of individuals (nine male pairs and one female pair) diagnosed with schizophrenia and non-psychiatric comparisons.

Results:

Significantly higher total CB1R levels were found within samples from individuals with schizophrenia. Terminal type-specific analyses identified significantly higher CB1R levels within excitatory boutons in samples from individuals with schizophrenia relative to comparisons. In contrast, CB1R levels within the subset of inhibitory boutons that normally express high CB1R levels (presumptive cholecystokinin neuron boutons) were lower in samples from individuals with schizophrenia relative to comparison samples.

Conclusion:

Given CB1R’s role in suppressing neurotransmission upon activation, these results suggest an overall shift in excitatory and inhibitory balance regulation toward a net reduction of excitatory activity in schizophrenia.

1. Introduction

Cannabis is the most widely used recreational psychoactive substance worldwide, with ongoing increase in usage (Hasin, 2018; Bhatia et al., 2022). Cannabis use is associated with various psychiatric comorbidities (Stinson et al., 2006) and represents one of the strongest environmental factors associated with schizophrenia (SZ) (Andréasson et al., 1987). In population studies, cannabis use diagnoses significantly increase the hazard ratios of developing SZ (Nielsen et al., 2017; Callaghan et al., 2012). Temporally, cannabis use is associated with younger ages of first psychotic episodes (Large et al., 2011), with severity of cannabis use dose-dependently modulating the risk of developing SZ (Zammit et al., 2002; Henquet et al., 2005), and many patients with cannabis-induced psychosis later develop SZ (Arendt et al., 2005). Importantly, individuals diagnosed with SZ use cannabis at significantly higher rates, with exposure to cannabis transiently exacerbating cognitive and positive symptoms (D’Souza et al., 2005), and a history of cannabis use being associated with worse illness prognosis (Manrique-Garcia et al., 2014; Patel et al., 2020).

Recent genetic studies investigating the relationship between cannabis use and SZ suggest that shared risk liabilities of cannabis use disorder and SZ may partially account for these observations. In particular, Mendelian randomization studies demonstrated that while SZ increases the risk of cannabis use (Gage et al., 2017), cannabis use further increases the risk of SZ beyond what can be accounted for by genetic correlations alone (Vaucher et al., 2018). These findings suggest that changes within the endocannabinoid system may affect the clinical outcomes of both cannabis use and SZ, and increasing our understanding of this system remains critical, with potential therapeutic benefits.

Δ9-tetrahydrocannabinol (THC), the major psychoactive substance in cannabis, targets the cannabinoid CB1 receptor (CB1R) (Shrivastava et al., 2014). A ubiquitous G-protein coupled receptor (GPCR) (Busquets-Garcia et al., 2018), CB1R is highly expressed within the cholecystokinin (CCK) subtype GABAergic inhibitory interneurons within the human cortex (Trettel et al., 2004; Katona et al., 1999). However, it is also found in other interneuronal subtypes and glutamatergic neurons (Bellocchio et al., 2010; Chou et al., 2022). Functionally, presynaptic CB1R activation results in reduced synaptic transmission by attenuating neurotransmitter release, a phenomenon known as depolarization-induced suppression of inhibition (DSI) at inhibitory boutons, and depolarization-induced suppression of excitation (DSE) at excitatory sites (Egertová and Elphick, 2000; Trettel and Levine, 2002; Fortin and Levine, 2007; Yoshino et al., 2011; Liu et al., 2020). Thus, CB1R appears to be critical in regulating cortical excitatory-inhibitory (E/I) balance (den Boon et al., 2015a).

CB1R alterations may also play a role in cognitive impairments in SZ – a function involving the prefrontal cortex (PFC) (Guo et al., 2019). Studies have identified reduced dendritic spine density in PFC pyramidal neurons of individuals with SZ (Garey et al., 1998; Konopaske et al., 2014; Glantz and Lewis, 2000), without a difference in pyramidal neuron density (Pierri et al., 2001), or total neuron number (Thune et al., 2001), which – given the spines’ role in forming excitatory synapses – may suggest decreased excitatory drive (Smucny et al., 2022; Glausier et al., 2020). In contrast, reduced mRNA expression of gamma-aminobutyric acid (GABA) synthesizing enzymes in SZ suggests decreased GABA synthesis (Hashimoto et al., 2003; Dienel and Lewis, 2019), again without missing neurons (Dienel et al., 2023), which may also contribute to E/I balance impairments. However, the full circuitry abnormalities leading to these PFC disturbances in SZ remain uncertain.

Considering the above, studies have investigated CB1R changes as potential mechanisms in the pathophysiology of SZ (Ferretjans et al., 2012; Ibarra-Lecue et al., 2018). Interestingly, ligand-binding autoradiography studies targeting all CB1R, including those in non-CCK interneurons and glutamatergic neurons, demonstrated increased cortical CB1R binding in postmortem samples from individuals with SZ (Dean et al., 2001a; Zavitsanou et al., 2004; Newell et al., 2006; Dalton et al., 2011; Volk et al., 2014). In contrast, immunohistochemistry (IHC) antibody-based studies showed decreased CB1R protein levels in SZ (Urigüen et al., 2009; Eggan et al., 2010; Eggan et al., 2008). Of note, existing literature indicates that the antibodies used in prior IHC studies preferentially labeled puncta with high CB1R expressions in inhibitory boutons – predominantly found to be CCK-positive cells given the high abundance and expression of CB1R within CCK subtype GABAergic interneurons (Volk et al., 2014; Eggan et al., 2008; Gonzalez-Burgos et al., 2015). We hypothesized that the discrepant results between these two methods may stem from terminal type-specific alterations of CB1R levels in SZ. We undertook to preliminarily test this idea by performing quantitative IHC using a CB1R antibody we previously showed to detect CB1R in both interneurons and glutamatergic neurons in the postmortem human PFC (Chou et al., 2022), in an existing cohort of SZ subjects previously examined with both ligand-binding autoradiography and CCK cell selective CB1R antibodies.

2. Methods and materials

2.1. Human tissue

We studied ten individuals with SZ, each matched for sex and age to an unaffected comparison subject (Ctrl) without psychiatric diagnoses (Table 1 and S1). To control for experimental variance, subjects from each pair were processed together throughout the protocol. All pairs were previously assayed for PFC CB1R levels using both ligand-binding and antibody-based approaches (Volk et al., 2014; Eggan et al., 2008). See Table 2 for each pair’s ligand-binding and antibody-based CB1R ratios.

Table 1.

Summary characteristics of individuals included in the study.

Characteristics	Ctrl, N = 10	SZ, N = 10	p-value
Age, years, mean (SD)	46.9 (15.9)	48.4 (13.7)	0.82
Sex, n
Female	1	1	1.00
Male	9	9
Race, n
Black	2	1	1.00
White	8	9
Cannabis use history
Yes	0	3	0.21
No	10	7
PMI, hours, mean (SD)	17.9 (5.9)	20.3 (11.2)	0.55
Storage time, months, mean (SD)	187.8 (16.3)	188.8 (24.4)	0.92
pH, mean (SD)	6.96 (0.24)	6.92 (0.20)	0.69
Leading cause of death (% subjects affected)	Cardiovascular (70%)	Cardiovascular (60%)

Abbreviations: Ctrl = unaffected comparison, SZ = schizophrenia, PMI = postmortem interval, SD = standard deviations.

Table 2.

Reciprocal ligand-binding and IHC protein ratio results obtained from subject pairs used in prior studies. These pairs are included in the current study (i.e., those with demographic information provided in Table 1). IHC protein ratio is calculated as the percentage of protein level (measured in optical density) in samples from subjects with schizophrenia (SZ) to samples from unaffected comparisons. Ligand-binding ratio is calculated as the percentage of OMAR ligand-binding (fm/mg) in samples from patients with SZ to samples from unaffected comparisons. Magnitude difference is calculated as the difference between the ligand binding ratio and the protein IHC ratio. Modified from Volk, et al., 2014 with permission (Volk et al., 2014).

Pair	IHC protein SZ/Ctrl ratio (%)	Ligand-binding SZ/Ctrl ratio (%)	Magnitude difference (Ligand-binding - protein)
1	−5.57	25.76	31.33
2	−11.04	41.18	52.22
3	−20.18	2.51	22.68
4	−31.61	74.85	106.46
5	−24.01	−0.74	23.27
6	−6.17	29.19	35.35
7	−30.19	43.08	73.27
8	−23.15	−5.18	17.97
9	−14.64	47.74	62.38
10	−21.75	14.52	36.27
Mean (SD)	−18.83 (9.21)	27.29 (25.27)	46.12 (27.84)

Abbreviations: Ctrl = unaffected comparison, SZ = schizophrenia, IHC = immunohistochemistry.

Brain specimens from subjects were obtained from autopsies conducted at the Allegheny County Office of the Medical Examiner, Pittsburgh, PA, following consent for donation from next of kin. Psychiatric or neurological histories were determined by an independent committee of experienced research clinicians using information obtained from clinical records and structured interviews conducted with a surviving relative, including any known history of cannabis use or use disorders. The University of Pittsburgh’s Committee for the Oversight of Research and Clinical Trials Involving Decedents and Institutional Review Board for Biomedical Research approved all procedures.

Following brain retrieval, left hemispheres were cut into 1.0–2.0 cm-thick coronal blocks, fixed for 48 h in phosphate-buffered 4% paraformaldehyde at 4 °C, immersed in graded cold sucrose solutions, and stored at −30 °C in cryoprotectant solutions until sectioning (DeGiosio et al., 2019a). PFC blocks containing the region of interest (ROI; Brodmann area 9) were sectioned coronally at 40 μm on a cryostat, and every 40th section was Nissl stained to serve as anatomical references for laminar identification. Unstained sections were stored in cryoprotectant solution at −30 °C until processed for immunohistochemistry.

2.2. Immunohistochemistry

One free-floating tissue section per subject containing the ROI was used. Sections were washed in 0.1 M phosphate-buffered saline (PBS) then incubated for 75 min in 0.01 M sodium citrate solution at 80 °C to retrieve antigens (Jiao et al., 1999). After cooling to room temperature (RT), sections were immersed in 1% sodium borohydride for 30 min at RT to reduce background autofluorescence (Clancy and Cauller, 1998), followed by membrane permeabilization with 0.3% Triton X-100 in PBS for 30 min at RT. Sections were blocked with 20% normal goat serum (NGS) in PBS for 2 h at RT to reduce nonspecific antibody binding, then incubated for 72 h at 4 °C in PBS containing 2% NGS and primary antibodies.

Primary antibodies included monoclonal mouse anti-vGAT antibody (1:500; Synaptic Systems, Göttingen, Germany; product # 131011) – which labels inhibitory boutons; polyclonal guinea pig anti-vGlut1 antibody (1:500; Millipore Sigma, Burlington, MA; product # AB5905) – which labels inhibitory boutons, and polyclonal rabbit anti-CB1R antibody (1:2000; Synaptic Systems, Göttingen, Germany; product # 258003). We previously demonstrated successful and specific vGAT and vGlut1 labeling in human and non-human primate postmortem studies using these antibodies (Rocco et al., 2017a; Moyer et al., 2013; Rocco et al., 2017b; Rocco et al., 2016). The CB1R antibody demonstrated successful co-labeling with both vGAT and vGlut1 in both neuronal cultures and postmortem human brain samples (Chou et al., 2022; Morini et al., 2015). In addition, vGAT and CB1R antibody specificities were validated through knockout samples (Martens et al., 2008; Barnes et al., 2020), and vGlut1 antibody through pre-adsorption controls (Millipore certificate of analysis, 2016).

Post primary antibody incubation, sections were rinsed for 4 × 30 min in PBS and incubated for 24 h at 4 °C in PBS containing 2% NGS and goat host secondary antibodies conjugated to Alexa-488 (1:500; vGlut1), Alexa-568 (1:500; CB1R) and Alexa-647 (1:500; vGAT; Invitrogen, Grand Island, NY, for Alexa antibodies). Sections were rinsed for 4 × 30 min in PBS, mounted on slides, cover slipped (ProLong Gold antifade reagent, Invitrogen), sealed with clear nail polish along coverslip edges, and stored at 4 °C until imaged. A sample CB1R-immunoreactive (IR) labeling within postmortem PFC is shown in Fig. 1, where CB1R-IR signals are seen co-localized with vGAT-IR and vGlut1-IR puncta. There are also CB1R-IR labeling of neuronal soma and axons not co-localized with either synaptic marker, i.e., CB1R-IR puncta that are neither vGlut-IR nor vGAT-IR.

Fig. 1.

Representative micrograph of immunohistochemical labeling of postmortem human prefrontal cortex tissue section. Left panels: Puncta with vGlut1-immunoreactive (IR) (green), vGAT-IR (blue) and CB1R-IR (red) labeling are distributed throughout the image field. Right panels: Enlarged images with arrows identifying puncta with antibody immunoreactivity toward a single or multiple proteins.

2.3. Tissue sampling

Sampling procedure was as previously described (Sweet et al., 2010; DeGiosio et al., 2019b). Contours outlining each section were drawn in Stereo Investigator version 8 (MicroBrightField Inc., Natick, MA). To ensure representative sampling of complete gray matter, image stacks were obtained from six to ten randomly chosen sites for each layer per subject, determined using nearby Nissl-stained sections, equally sampled within and across subject pairs. Boundaries of cortical layers were estimated as percent of distance from pial surface to white matter: Layer 1 (pia–10%), Layer 2 (10–20%), Layer 3 (20–50%), Layer 4 (50–60%), Layer 5 (60–80%), Layer 6 (80%–gray/white matter border) (Eggan et al., 2008; Pierri et al., 1999). Tissue thickness (z-axis depth) for each site was measured and divided by 40 μm (original section thickness) to correct for shrinkage during IHC.

2.4. Confocal microscopy

Microscopy equipment and capturing parameters were as previously described (McKinney et al., 2019). Data were collected using a 60 × 1.40 numerical aperture super-corrected oil immersion objective mounted on an Olympus BX51Wl upright microscope (Olympus America Inc., Center Valley, PA) equipped with an Olympus spinning disk confocal unit, Hamamatsu Orca R2 camera (Hamamatsu, Bridgewater, NJ), MBF CX9000 front mounted digital camera (MicroBrightField Inc., Natick, MA), BioPrecision2 XYZ motorized stage with linear XYZ encoders (Ludl Electronic Products Ltd., Hawthorne, NY), excitation and emission filter wheels (Ludl Electronic Products Ltd., Hawthorne, NY), Sedat Quad 89,000 filter set (Chroma Technology Corp., Bellows Falls, VT), and Lumen 220 metal halide lamp (Prior Scientific, Rockland, MA).

Equipment was controlled by SlideBook 6.0 (Intelligent Imaging Innovations, Inc., Denver, CO), which was also used for post-image processing. Three-dimensional image stacks (two-dimensional images successively captured at 0.25 μm z-dimension intervals) were acquired with a depth spanning top 20% of tissue thickness (i.e., measuring 20% of thickness beginning at the coverglass), starting from the plane furthest away from the coverglass and stepping up until reaching tissue surface. Images were 512 × 512 pixels (55 × 55 μm) in the XY dimension. Stacks were collected using optimal exposure settings (i.e., those yielding the greatest dynamic range possible for the camera without saturated pixels). Z-positions were normalized to original section thickness and exposures normalized for each capture post-image processing prior to analysis.

2.5. Image processing

Images were processed as previously described (McKinney et al., 2019; Fish et al., 2008), using SlideBook and Automation Anywhere software (Automation Anywhere, Inc., San Jose, CA). Image stacks were deconvolved using AutoQuant’s blind deconvolution algorithm (MediaCybernetics, Rockville, MD). After deconvolution, separate Gaussian channels were made for each deconvolved channel by calculating a difference of Gaussians (sigma 0.7 - sigma 2.0). These channels, which enhanced immunofluorescence edge demarcations, were used for data segmentation.

Segmentation of Gaussian channels was performed using a previously described iterative combined intensity/morphologic thresholding algorithm with MATLAB (MATLAB, The MathWorks Inc., Natick, MA) (Rocco et al., 2017a). After obtaining initial values for iterative segmentation for each channel using Otsu’s method within SlideBook, each subsequent iteration increased threshold by 50 gray levels, and object masks were size gated within 0.03–2.0μm³. After each segmentation, masked objects were merged with prior iterations, with final resulting masks copied back onto the original deconvolved channels (i.e., without Gaussian subtraction) to obtain pixel intensity information. Lipofuscin, an autofluorescent lysosomal degradation product, which may confound quantitative fluorescence measures in human postmortem tissues, was imaged using a separate channel at a constant exposure time across all sections.

After generating vGAT and vGlut1 bouton object masks, mean CB1R intensity in analog-to-digital units (ADU) underneath each masked object was obtained. Values were averaged across all boutons per sampled site for each bouton type. The resulting terminal type-specific mean CB1R intensity values for each sampled site were then averaged across each layer to obtain a single value as the dependent measure. For determining all CB1R signal irrespective of terminal type within each site, sum CB1R intensity was measured from a single 2D plane, and site values averaged for each layer as the dependent measure.

Prior to analyses, data were filtered to ensure accurate representation of receptor labeling. Based upon examination of antibody signal penetrance across tissue thickness, only objects falling within 10–14 μm from tissue surface after correcting for tissue shrinkage were included for analysis. To prevent potential spherical aberration confounding measurements, a virtual counting frame inclusive of signals falling between the upper and lower 2% of XY dimensions was used (i.e., between 10 and 502 units for each dimension). To ensure accurate capture of terminal type-specific measurements, objects overlapping the lipofuscin and both vGlut1 and vGAT masks were excluded from analysis.

2.6. Statistical analysis

Demographic data were analyzed using Fisher’s exact test (categorical variables) and Student’s two-tailed t-test (continuous variables). To analyze sum CB1R intensity, analysis of covariance (ANCOVA) models were performed. Sum CB1R intensity values at all sampled sites per cortical layer per subject were averaged to obtain a single measure as the dependent variable. Subject group, cortical layer, and subject group × cortical layer two-way interaction were entered as fixed effects, and subject pair entered as a blocking factor. To assess possible confounding effects of cohort variables (sex, race, age, postmortem interval, and tissue storage time), a second unpaired ANCOVA model was performed to validate the first model, using subject group, cortical layer, and subject group × cortical layer two-way interaction as fixed effects, and cohort variables as covariates. Results for both paired and unpaired models were reported.

As existing literature indicates that within GABAergic interneurons, CB1R is most abundant and expressed at highest levels within CCK-positive cells, and given that prior IHC studies used an anti-CB1R antibody that specifically labeled inhibitory boutons with high CB1R expressions determined to be predominantly CCK-positive subtype interneurons (Volk et al., 2014; Eggan et al., 2008; Gonzalez-Burgos et al., 2015; Karson et al., 2008; Iball and Ali, 2011), we separated inhibitory bouton populations for terminal type-specific analysis of mean CB1R intensity. Inhibitory boutons were categorized as high- or low-CB1R-expressors, using the median value of non-psychiatric subjects’ mean CB1R intensities in vGAT-IR boutons (712 ADU) to define groups after reviewing total intensity distribution (see Fig. S2). We then compared mean CB1R intensities within excitatory (i.e., vGlut1-IR), high-CB1R-expressing, and low-CB1R-expressing inhibitory (i.e., vGAT-IR) boutons between groups.

To analyze mean CB1R intensity, ANCOVA models were performed. Mean CB1R intensity values for each terminal type at all sampled sites per cortical layer per subject were averaged to obtain a single measure as the dependent variable. Subject group, cortical layer, terminal type, subject group × terminal type two-way interaction, and subject group × cortical layer × terminal type three-way interaction were entered as fixed effects, and subject pair entered as a blocking factor. A second unpaired ANCOVA model was performed to validate the first model, using the same fixed effects as the first model, and cohort variables as covariates. Results for both models were reported.

To analyze effects of cannabis and medications on CB1R, independent-samples two-tailed t-test was used to compare within-pair ratios of mean or sum CB1R intensities (Ctrl/SZ CB1R intensity ratio) between pairs with and without cannabis or medication exposure histories in SZ subjects. When appropriate, significant differences were followed by post hoc Bonferroni tests to correct for increased risk of a type I error when making multiple comparisons. For all analyses, p < 0.05 was considered statistically significant.

3. Results

3.1. Global CB1R alterations in the PFC of patients with SZ

Sum CB1R intensity (encompassing vGlut1-IR, vGAT-IR, and non-vGlut1-IR/vGAT-IR populations) was significantly +26.8% higher in SZ compared to Ctrl, F(1,99) = 18.702, p < 0.001 for paired analysis; F (1,103) = 9.130, p = 0.003 for unpaired analysis (Fig. 2a and b). There was a significant main effect of layers using paired analysis, F(5,99) = 3.700, p = 0.004, with post hoc comparison indicating significantly lower sum CB1R intensity in layer VI compared to layers I and II. However, this effect was not present using unpaired analysis, F(5,103) = 1.996, p = 0.085. There was no significant condition × layer interaction in both paired and unpaired analyses, p = 0.863 for paired; p = 0.960 for unpaired (Fig. 2c). Significant results persisted in analyses without outlier pair (Pair 2; see Tables S1 and S2).

Fig 2.

a Sum CB1R intensity from postmortem PFC samples of subjects with schizophrenia (SZ) and unaffected comparisons (Ctrl). Each individual data point represents the sum intensity averaged across all sampled sites across a single subject. Central line indicates the median, box boundaries extend from the 25th to 75th percentiles, and whiskers extend from the minimum to maximum value. There was a main effect of subject group, p < 0.001. **p < 0.001.

b Comparison of sum CB1R intensity in matched pairs of comparison subjects (Ctrl) and subjects with schizophrenia (SZ). Mean values of sum CB1R intensities for each subject group are indicated by the X. Markers above the dashed unity line indicate pairs for which the subject with schizophrenia disorder had higher sum CB1R intensity than the matched comparison subject.

c Sum CB1R intensity for individual subjects across cortical layers. Each individual data point represents the sum intensity averaged across all sampled sites for a single subject. Central line indicates the median, box boundaries extend from the 25th to 75th percentiles, and whiskers extend from the minimum to maximum value. There was a main effect of subject group, p < 0.001. **p < 0.001.

There were no group differences in Ctrl/SZ sum CB1R intensity ratios between pairs including SZ subjects with or without cannabis, antipsychotic, antidepressant, benzodiazepine or valproic acid exposures (Table S3).

3.2. Terminal type-specific CB1R alterations in the PFC of patients with SZ

There was a significant main effect of terminal type in mean CB1R intensity, F(2,315) = 827.566, p < 0.001 for paired analysis; F(2,319) = 786.746, p < 0.001 for unpaired analysis; and a significant terminal type × subject group interaction, F(2,315) = 22.875, p < 0.001 for paired analysis; F(2,319) = 21.747, p < 0.001 for unpaired analysis (Fig. 3a and b). Post hoc pairwise comparisons showed 35.3% higher mean CB1R intensity in SZ compared to Ctrl within vGlut1-IR bouton populations, p < 0.001 for both paired and unpaired analyses, and 14.9% lower mean CB1R intensity in SZ compared to Ctrl within high-CB1R-expressing vGAT-IR bouton populations, p < 0.001 for both paired and unpaired analyses (Table 3). Significant results persisted in analyses without outlier pair (Pair 2; see Tables S1 and S2).

Fig 3.

a Mean CB1R intensity within excitatory (vGlut1-IR), high-CB1R-expressing inhibitory (vGAT-IR), and low-CB1R-expressing inhibitory boutons from postmortem PFC samples of subjects with schizophrenia (SZ) and unaffected comparisons (Ctrl). Each individual data point represents mean intensity averaged across all sampled sites across a single subject. Central line indicates the median, box boundaries extend from the 25th to 75th percentiles, and whiskers extend from the minimum to maximum value. There was a significant terminal type × subject group interaction, p < 0.001. Mean CB1R intensity in SZ was significantly higher compared to Ctrl in excitatory boutons, and significantly lower compared to Ctrl in high-CB1R-expressing inhibitory boutons. ** p < 0.001.

b Comparison of mean CB1R intensity in excitatory (vGlut1-IR) boutons (left) and high-CB1R-expressing inhibitory (vGAT-IR) boutons (right) in matched pairs of comparison subjects (Ctrl) and subjects with schizophrenia (SZ). Mean values on mean CB1R intensities for each subject group are indicated by the X. Markers below the dashed unity line indicate pairs for which the subject with schizophrenia disorder had lower mean CB1R intensity than the matched comparison subject.

c Within-pair ratios of mean CB1R intensity in vGlut-IR boutons to high-CB1R-expressing vGAT-IR boutons for individual subjects. Each individual data point represents the within-pair ratio averaged across all sampled sites for a single subject. Central line indicates the median, box boundaries extend from the 25th to 75th percentiles, and whiskers extend from the minimum to maximum value. There was a main effect of subject group, p < 0.001. **p < 0.001.

Table 3.

Mean CB1R intensity in analogue-to-digital units (ADU) for all boutons across subject groups and terminal types. Values are represented as mean ± SEM.

Terminal type	Subject group	Mean CB1R intensity (ADU ± SEM)	p-value
Excitatory (vGlut1-IR)	Ctrl	770.537 ± 48.073	p < 0.001
	SZ	1042.193 ± 48.073
High-CB1R-expressing Inhibitory (vGAT-IR)	Ctrl	2510.098 ± 48.073	p < 0.001
	SZ	2137.309 ± 48.073
Low-CB1R-expressing Inhibitory (vGAT-IR)	Ctrl	435.461 ± 48.073	p = 0.714
	SZ	460.420 ± 48.073

Abbreviations: Ctrl = unaffected comparison, SZ = schizophrenia, ADU = analog-to-digital units, SEM = standard error of mean.

To further validate terminal type-specific CB1R alterations between subject groups, we performed within subject comparisons of mean CB1R intensity ratio in vGlut1-IR to high-CB1R-expressing vGAT-IR bouton populations (Fig. 3c). There was a significant main effect of subject group, F(1,99) = 53.702, p < 0.001. The ratio of mean CB1R intensity in vGlut1-IR boutons to high-CB1R-expressing vGAT-IR boutons was 15.6% higher in SZ compared to Ctrl, indicating smaller CB1R enrichment differentials between terminal types in SZ.

There was no significant difference in mean CB1R intensity between SZ and Ctrl within the low-CB1R-expressing vGAT-IR bouton population, p = 0.714 for paired analysis; p = 0.802 for unpaired analysis. There was also no significant main effect of cortical layer (Fig. S3; F (5,315) = 1.287, p = 0.269 for paired analysis; F(5,319) = 1.223, p = 0.298 for unpaired analysis), or significant subject group × cortical layer × terminal type three-way interaction (F(25,315) = 0.758, p = 0.794 for pair analysis; F(25,319) = 0.721, p = 0.836 for unpaired analysis).

There were no group differences in Ctrl/SZ sum CB1R intensity ratios between pairs including SZ subjects with or without cannabis, antipsychotic, antidepressant, benzodiazepine or valproic acid exposures for any terminal type (Table S4).

4. Discussion

4.1. Summary of current findings

This preliminary study compared terminal type-specific distributions of CB1R within postmortem human PFC in individuals with SZ and non-psychiatric comparisons. We focused on this region given its involvement in the cognitive symptoms of SZ, to expand upon prior knowledge regarding CB1R alterations in this illness. When examining total CB1R, which includes not only CB1R on excitatory and inhibitory boutons, but also other locations in which CB1R are present (e.g., cholinergic, serotonergic terminals, axon segments, astrocytes) – accounting for a smaller but substantial and functionally important portion of total CB1R (Lu and Mackie, 2021) – we identified significantly higher overall CB1R levels in individuals with SZ than non-psychiatric comparisons.

Interestingly, when examining terminal type-specific distributions of CB1R levels, we identified a significant terminal type by subject group interaction. Specifically, mean CB1R intensity in excitatory boutons was significantly higher in SZ samples, while mean CB1R intensity in high-CB1R-expressing inhibitory boutons was significantly lower in SZ samples compared to controls.

4.2. Comparison to prior findings

The subject pairs included in the present study were chosen based on their prior findings suggesting reciprocal alterations in CB1R protein and ligand binding. Here, our result of higher overall CB1R levels in individuals with SZ is consistent with prior results using CB1R ligand binding assays. Using postmortem brain samples, groups have assessed CB1R levels within the PFC of individuals with SZ using various radioligands, including agonist (i.e., [³H]CP-55940) (Dalton et al., 2011; Dean et al., 2001b), and inverse agonist (i.e., [³H]SR141716A, [³H] MePPEP and [³H]-OMAR) (Zavitsanou et al., 2004; Volk et al., 2014; Jenko et al., 2012). Irrespective of differences in binding affinity or specificity, all studies reported higher ligand binding in samples from individuals with SZ compared to unaffected counterparts, with multiple studies controlling for covariates including age, sex, postmortem interval, THC history, and antipsychotic history. Although the same ligands as used in postmortem studies (i.e., [¹¹C]MePPEP, [¹¹C]OMAR) demonstrated lower global binding in individuals with SZ in vivo (Borgan et al., 2019; Ranganathan et al., 2016), higher global CB1R radioligand binding in SZ had similarly been demonstrated when assessed in vivo using the inverse agonist [¹⁸F]MK-9470 (Ceccarini et al., 2013). Notably, the in vivo studies did not specifically examine the PFC, which may account for the differing results from postmortem findings.

In addition, we expanded upon prior IHC studies examining CB1R in postmortem PFC samples from individuals with SZ, which utilized anti-CB1R antibodies that preferentially targeted high-CB1R-expressing inhibitory neurons confirmed to be CCK-positive (Eggan et al., 2010). We separated mean CB1R levels in inhibitory boutons between low- and high-CB1R-expressing populations based on the median value of CB1R intensities within Ctrl samples. Our results again complemented prior findings. Specifically, we identified lower CB1R levels in postmortem PFC samples from individuals with SZ relative to comparisons when assessing the subset of high-CB1R-expressing inhibitory boutons.

4.3. CB1R within excitatory and inhibitory neuronal populations

The current findings expand our understanding of terminal type-specific CB1R alterations in the PFC of SZ. Here, we note that CB1R changes in SZ appear to be terminal type-specific, with increased CB1R in excitatory terminals compared to unaffected individuals. As PFC pathology is implicated in the cognitive dysfunctions of SZ, CB1R alterations may directly contribute to symptom development by disturbing the excitatory and inhibitory balance (den Boon et al., 2015b) – a mechanism known to contribute to impaired salience learning (Insel et al., 2018). Considering CB1R’s role in suppressing neurotransmitter release, it is possible that these findings of higher CB1R levels in excitatory boutons of SZ represent a stronger suppression of excitatory neurotransmission (i.e., DSE), a phenomenon previously demonstrated in the somatosensory cortex (Fortin and Levine, 2007). Alternatively, CB1R is also implicated in the long-term depression of excitatory synapses within the PFC (Lafourcade et al., 2007), and these two mechanistic possibilities both complement the glutamatergic hypofunction theory as a contributor to the pathology of SZ (Dixon et al., 2022; Dauvermann et al., 2017).

Our results also identified significantly different CB1R levels in inhibitory boutons between samples from individuals with SZ and non-psychiatric comparisons, and suggested GABAergic subtype specific alterations of CB1R in SZ. Current literature supports the predominance of CB1R within CCK-containing interneurons using non-psychiatric postmortem human brain samples, with lower levels of CB1R detected in parvalbumin (PV)-positive cells using rodent studies (Katona et al., 1999). CB1R associated DSI appears to be present only within CCK-positive interneurons and not identified within other GABAergic subtypes despite low levels of CB1R being present in these interneuron populations (e.g., PV neurons) (Gonzalez-Burgos et al., 2015; Karson et al., 2008; Iball and Ali, 2011). Thus, our finding GABAergic CB1R alterations only within high-CB1R-expressing boutons in SZ suggest a predominant disruption of CB1R in presumptive CCK-containing interneurons, potentially contributing to the pathophysiology of the illness through attenuated DSI.

4.4. Relationship with cannabis use in individuals with SZ

Our findings of increased CB1R in excitatory boutons and decreased CB1R in putative DSI associated inhibitory boutons in individuals with SZ may offer a potential explanation for the clinical observations of THC exposure exacerbating symptoms in SZ. An increase in CB1R within excitatory boutons may strengthen DSE following THC activation of the receptors, while a decrease in CB1R within inhibitory boutons may reduce DSI following THC exposure. It is possible that these alterations may then lead to further intensification of the glutamatergic hypoactivity present in individuals with SZ, and subsequent symptom worsening.

This is partially supported by a recent study on terminal type-specific CB1R dependent behavioral effects using knock-out mice that underwent CB1R rescues in either dorsal telencephalic glutamatergic or forebrain GABAergic neurons (de Giacomo et al., 2020). In CB1R knock-out mice that underwent glutamatergic CB1R rescue – a condition relevant to what we observed at present in individuals with SZ (i.e., increased glutamatergic CB1R and decreased GABAergic CB1R), THC exposure was sufficient to produce hypolocomotion. It is possible that alterations in terminal type-specific CB1R distribution led to a disruption in E/I homeostasis, which is then exacerbated by exogenous CB1R activation through THC exposure. Additional studies using rodent manipulations would be necessary to understand how terminal type-specific CB1R alterations may affect SZ related behaviors, and whether cannabis use leads to further terminal type-specific behavioral disturbances under these conditions.

4.5. Limitations

While this study expanded our understanding of CB1R alterations in SZ, its clinical generalizability is limited given its small scale and restricted subject selection. With only one pair of female subjects, three Black individuals, and predominantly middle-aged adult samples, our selection was inadequate for detecting sex, race, or age-related outcomes (also see Table S5 for additional analyses conducted without the inclusion of the female subject pair). However, by including these variables as covariates in our analyses, we were able to identify unique terminal type-specific CB1R changes after controlling for these factors. Similarly, by conducting t-tests to compare results from subject pairs with or without cannabis and medication exposures, we were able to clarify that terminal type-specific CB1R alterations observed were independent of medication or cannabis histories. However, these latter results should be interpreted in the context of limited samples. Future studies with larger sample size are needed to allow for more robust comparisons of the influence of these and other potential confounding variables.

In addition, by assessing only vGlut1 and vGAT colocalization with CB1R, the results provide only a broad overview of CB1R distributions in excitatory and inhibitory boutons, with the understanding that these groups are comprised of additional subpopulations. Future larger scale work incorporating GABAergic subtype specific markers would be necessary to fully elucidate more nuanced cell type specificity. The identity of CB1R-positive puncta not colocalized with these two markers were also unknown, and these may represent other contributors to the development of psychiatric symptoms (Han et al., 2017).

5. Conclusion and future directions

Our study replicated prior findings of higher overall CB1R levels within postmortem PFC of individuals with SZ. We also identified the presence of terminal type-specific CB1R alterations, namely increased CB1R levels in excitatory boutons, and decreased CB1R levels in high-CB1R-expressing (presumptive CCK) inhibitory boutons in SZ. These changes suggest possible net attenuation of excitatory neurotransmission in SZ, supporting the prefrontal glutamatergic dysfunction hypothesis, lending strength to the idea that CB1R alterations disrupt PFC E/I balance in SZ. Though limitations exist, these results support the importance of conducting more in-depth CB1R examinations in SZ to elucidate the relationship between the endocannabinoid system, cannabis exposure and psychotic illnesses.

Supplementary data to this article can be found online at https://doi.org/10.1016/j.nbd.2023.106262.

Data availability

Data will be made available on request.