Abstract

Most idiosyncratic drug reactions (IDRs) appear to be immune-mediated, but mechanistic events preceding severe reaction onset remain poorly defined. Damage-associated molecular patterns (DAMPs) may contribute to both innate and adaptive immune phases of IDRs, and changes in extracellular vesicle (EV) cargo have been detected post-exposure to several IDR-associated drugs. To explore the hypothesis that EVs are also a source of DAMPs in the induction of the immune response preceding drug-induced agranulocytosis, the proteome and immunogenicity of clozapine- (agranulocytosis-associated drug) and olanzapine- (non-agranulocytosis-associated drug) exposed EVs were compared in two preclinical models: THP-1 macrophages and Sprague–Dawley rats. Compared with olanzapine, clozapine induced a greater increase in the concentration of EVs enriched from both cell culture media and rat serum. Moreover, treatment of drug-naïve THP-1 cells with clozapine-exposed EVs induced an inflammasome-dependent response, supporting a potential role for EVs in immune activation. Proteomic and bioinformatic analyses demonstrated an increased number of differentially expressed proteins with clozapine that were enriched in pathways related to inflammation, myeloid cell chemotaxis, wounding, transforming growth factor-β signaling, and negative regulation of stimuli response. These data indicate that, although clozapine and olanzapine exposure both alter the protein cargo of EVs, clozapine-exposed EVs carry mediators that exhibit significantly greater immunogenicity. Ultimately, this supports the working hypothesis that drugs associated with a risk of IDRs induce cell stress, release of proinflammatory mediators, and early immune activation that precedes severe reaction onset. Further studies characterizing EVs may elucidate biomarkers that predict IDR risk during development of drug candidates.

SIGNIFICANCE STATEMENT This work demonstrates that clozapine, an idiosyncratic drug-induced agranulocytosis (IDIAG)-associated drug, but not olanzapine, a safer structural analogue, induces an acute proinflammatory response and increases extracellular vesicle (EV) release in two preclinical models. Moreover, clozapine-exposed EVs are more immunogenic, as measured by their ability to activate inflammasomes, and contain more differentially expressed proteins, highlighting a novel role for EVs during the early immune response to clozapine and enhancing our mechanistic understanding of IDIAG and other idiosyncratic reactions.

Introduction

Idiosyncratic drug reactions (IDRs) are a diverse class of immune-mediated adverse reactions that can cause organ damage, sepsis, or even death (Sernoskie et al., 2021). Due to their unpredictable nature and poorly defined etiology, IDRs may evade detection during drug development, posing concerns for patient health and successful drug discovery pipelines (Young, 1949; Stepan et al., 2011; Sun et al., 2022). Clozapine is an atypical antipsychotic that was first marketed in the early 1970s, but it was withdrawn within the decade due to the incidence of idiosyncratic drug-induced agranulocytosis (IDIAG) (Crilly, 2007). It has since regained clinical approval with prescribing limitations and is underutilized compared with its effectiveness.

Severe IDRs including IDIAG appear to involve an adaptive immune attack in susceptible individuals. Specific human leukocyte antigen (HLA) haplotypes have been positively associated with the risk of clozapine-induced IDIAG (Li et al., 2018; Legge and Walters, 2019) and other IDRs (Usui and Naisbitt, 2017), and these adaptive events can be modeled using T lymphocyte priming assays with haplotyped cells, appropriate T cell receptors, and drug-modified proteins (Faulkner et al., 2016; Ogese et al., 2020). However, this adaptive immune activation should be preceded by drug-induced cell stress, release of damage-associated molecular patterns (DAMPs), and a proinflammatory response (Cho and Uetrecht, 2017; Pichler, 2019; Ali et al., 2020; Mosedale and Watkins, 2020; Kato et al., 2021; Yokoi and Oda, 2021; Thomson et al., 2022), an early step likely to occur more predictably and more frequently in drug-treated patients than the IDR itself. Indeed, during the initial weeks of clozapine therapy, an innate immune response has frequently been reported (Blackman et al., 2021; Sernoskie et al., 2021), and our laboratory has recapitulated aspects of this response in preclinical models, with data suggesting that clozapine is unique among similar compounds (Lobach and Uetrecht, 2014a; Ng et al., 2014; Sernoskie et al., 2022, 2023). Unlike clozapine, the structural analogs olanzapine and fluperlapine did not activate inflammasomes in vitro using a THP-1 macrophage model, as indicated by a lack of caspase-1 activation and release of IL-1β (Sernoskie et al., 2022, 2023). Moreover, in vivo using a female Sprague–Dawley rat model, we observed that although fluperlapine caused a similar decrease in circulating lymphocytes compared with clozapine, no increases in circulating neutrophils or serum CXCL1 levels were detected, indicating that fluperlapine did not induce a similar innate immune response (Sernoskie et al., 2022). Although fluperlapine was associated with cases of IDIAG in early clinical trials, the true incidence of fluperlapine-induced IDIAG is unknown because it was never marketed. As a comparator drug, olanzapine has a number of advantages over fluperlapine: it is currently Food and Drug Administration-approved, commonly prescribed for the treatment of schizophrenia, and is not associated with a significant incidence of IDIAG (Sanwlani and Gangoda, 2021). Moreover, unlike fluperlapine, both clozapine and olanzapine can be directly oxidized by myeloperoxidase, an enzyme present in myeloid cells, to a similar reactive nitrenium ion (Gardner et al., 1998b).

Using preclinical models, we have also demonstrated that myeloperoxidase plays a key role in the induction of this inflammatory response, specifically contributing to the release of soluble DAMPs and inflammatory mediators, as well as the chemotaxis of immune cells including neutrophils (Sernoskie et al., 2023). Yet, recent studies have also suggested that drug-modified proteins, DAMPs, and other signaling molecules can also be released encapsulated in extracellular vesicles (EVs) as opposed to directly into the extracellular space or into circulation. EVs are a heterogeneous class of nanoparticles with emerging functions in cell-to-cell communication, namely in regulating innate and adaptive immune responses, under both homeostatic and disease states (Chen et al., 2019). EVs can carry a range of cargo from the donor cell, including proteins, RNAs, and lipids, which, depending on the stimuli, can induce a plethora of effects in the recipient cell, such as cytokine release, chemotaxis, and cell polarization (Sanwlani and Gangoda, 2021). Specifically, alterations in EV cargo (proteins and microRNAs) have been demonstrated in response to exposure to idiosyncratic drug-induced liver injury (IDILI)- and severe cutaneous adverse reactions-associated drugs (Mosedale et al., 2018; Ogese et al., 2019; Salinas-Jaramillo et al., 2021), in addition to other hepatotoxicants (Momen-Heravi et al., 2015 a,b; Holman et al., 2016). Notably, in primary human hepatocytes, tolvaptan (an IDILI-associated drug) induced cellular stress and increased release of miR-122 in EVs in the absence of overt necrosis (Mosedale et al., 2018). Moreover, plasma-derived EVs from severe cutaneous adverse reactions patients exhibited an increase in proinflammatory proteins and a decrease in anti-inflammatory and protective proteins (Salinas-Jaramillo et al., 2021). Conversely, our current understanding of the role of EVs in the pathophysiology of blood-related dyscrasias like IDIAG is very limited.

In the present work, we hypothesized that EVs released in response to clozapine (an IDIAG-associated drug) and olanzapine (a non-IDIAG-associated clozapine analog) would differ significantly in terms of their specific protein cargo and relative immunogenicity, providing insight into the mechanisms of early immune activation in IDR pathogenesis. Two models that were previously shown to recapitulate clinical features of clozapine-induced inflammation were employed herein to compare changes in EVs following drug exposure: THP-1 macrophages and Sprague–Dawley rats (Fig. 1). Inflammatory mediator, toxicoproteomic, and bioinformatic analyses were performed to identify pathways involved in the relay of DAMPs and other inter-cell communication signals that may be unique to IDR-associated drug exposure, with the ultimate goal of identifying biomarkers to improve patient safety.

Fig. 1.

Overview of EV enrichment protocol. 1) In vitro model: PMA-differentiated THP-1 macrophages were exposed to clozapine (CLZ; 10 μg/mL), olanzapine (OLZ; 10 μg/mL) or vehicle control (CTR; 0.1% DMSO) for 24 hours, prior to collection of treatment conditioned media (CM). In vivo model: Female Sprague–Dawley rats were treated with clozapine (30 mg/kg), olanzapine (28.4 mg/kg), or vehicle control (saline) via IP injection, prior to serum collection at 3 hours post-dose. 2) EVs were enriched from drug-exposed CM or serum. 3) EV immunogenicity (a) was assessed by incubating EVs on drug-naïve THP-1 macrophages, collecting CM, and measuring inflammasome-dependent cytokine release. EV protein cargo (b) was examined using proteomics and subsequent ontology and network enrichment analyses.

MethodsChemicals and Reagents.

Clozapine was donated by Apotex (Toronto, ON), and olanzapine was purchased from Tokyo Chemical Industry Co., Ltd. America (Portland, OR). Fluperlapine and VX-765 were purchased from MedChemExpress (Princeton, NJ). Gibco One Shot Qualified fetal bovine serum (FBS) and exosome-depleted FBS were obtained from Thermo Fisher Scientific Inc. (Grand Island, NY) and heat-inactivated at 56°C for 30 minutes in-house. Phorbol 13-myristate 12-acetate (PMA) and bovine serum albumin were purchased from BioShop Canada Inc. (Burlington, ON). RPMI-1640 media (American Type Culture Collection modification, high glucose) was purchased from Thermo Fisher Scientific Inc. (Grand Island, NY). Amicon Ultra-15 Centrifugal Filter Units with 10 kDa molecular weight cut-off were purchased from Sigma-Aldrich Canada Co. (Oakville, ON). ExoQuick-TC Ultra and ExoQuick Ultra isolation kits were obtained from System Biosciences (Palo Alto, CA). All other reagents were commercially obtained.

THP-1 Maintenance and Drug Exposure for EV Enrichment.

THP-1 monocytes (TIB-202, American Type Culture Collection, Manassas, VA) were cultured in filter-sterilized RPMI media supplemented with 10% FBS (heat-inactivated) and maintained at 37°C, 5% CO2. Macrophage differentiation was achieved following PMA (25 ng/mL) stimulation of monocytes (4 × 105 cells/mL; passage 3–10) for 72 hours. Cells were washed with Dulbecco’s phosphate-buffered saline, rested for 24 hours, and washed again prior to drug exposure in RPMI media supplemented with 10% exosome-depleted FBS (heat-inactivated). Macrophages were exposed to clozapine or olanzapine (10 μg/mL; 30 μM) or vehicle control media [0.1% dimethyl sulfoxide (DMSO)] for 24 hours, following which conditioned media (80 mL/sample) was collected for EV enrichment (n = 6/group).

EV Enrichment of THP-1 Conditioned Media.

Media was centrifuged to remove cells (300 × g, 10 minutes) and debris (3000 × g, 10 minutes), prior to filtration (0.22 μm pore size filter) to deplete large vesicles. Media was then concentrated to < 1 mL with Amicon Ultra-15s according to the manufacturer’s recommendations. Filters were rinsed with PBS and combined retentate (5 mL) was added to ExoQuick-TC (1 mL; System Biosciences), mixed well, and incubated for 18 hours, prior to EV precipitation (centrifugation at 3000 × g, 15 minutes). EVs were resuspended in PBS for morphologic characterization and immunogenicity assays or lysed in detergent-free buffer (8 M urea, 100 mM tris hydrochloride, pH 8.2) for proteomic analysis.

Animal Treatment.

All studies were conducted in accordance with the Guide for the Care and Use of Laboratory Animals and were approved by the University of Toronto Animal Care Committee and the University Health Network Animal Care Committee. Female Sprague–Dawley rats (CRL:SD), 220–240 g, were purchased from Charles River (St. Constant, QC) and allowed to acclimatize (with handling) for > 1 week prior to study initiation. Rats were double-housed with a 12/12-hour light/dark cycle at 22°C and were provided access to rodent chow (Harlen Teklad, Madison, WI) and water ad libitum. Clozapine (30 mg/kg), an equimolar dose of olanzapine (28.4 mg/kg), and saline vehicle control were prepared as previously described (Ng et al., 2014; Sernoskie et al., 2022) and were administered between 8 and 9 am via intraperitoneal (IP) injection. Female rats were used to represent the increased incidence of IDIAG reported in female patients (Alvir et al., 1993), and our laboratory has previously demonstrated (Lobach and Uetrecht, 2014a; Sernoskie et al., 2022, 2023) that this model recapitulates immune changes observed during the first month of clozapine treatment in patients (Pollmächer et al., 1996, 1997; Löffler et al., 2010; Blackman et al., 2021; Sernoskie et al., 2021). Rats were euthanized by CO2 inhalation and exsanguination at 3 hours post-injection (n = 6/group). Blood was collected in EDTA microvettes (Sarstedt, Nümbrecht, Germany) for differential blood counts, obtained using a VETSCAN HM5 Hematology Analyzer or lithium-heparin microvettes (Sarstedt, Nümbrecht, Germany) for complete diagnostic chemistry profiles, obtained using a VETSCAN VS2 Chemistry Analyzer equipped with comprehensive diagnostic profile rotors. Blood samples were also collected for inflammatory mediator measurement or EV enrichment using Vacutainer Serum Separation Tubes (Becton, Dickinson and Company, Franklin Lakes, NJ). Blood was processed or analyzed within an hour of collection.

EV Enrichment of Rat Serum.

Coagulated blood was centrifuged to isolate serum (2000 × g, 10 minutes) and deplete debris (3000 × g, 10 minutes; 12000 × g, 10 minutes), prior to filtration (0.22 μm pore size filter) to deplete large vesicles. Purified serum (750 μL) was added to ExoQuick (100 μL), mixed well, and incubated for 30 minutes, prior to EV precipitation (centrifugation at 3000 × g, 15 minutes). EVs were further enriched using ExoQuick Ultra purification columns, according to the manufacturer’s instructions.

EV Characterization.

Resuspended EVs were analyzed for size and concentration by nanoparticle tracking analysis (NTA) using a NanoSight NS300 system (Malvern, UK). Samples were diluted in PBS to an acceptable particle/frame concentration, according to the manufacturer’s recommendations. Samples were introduced using an autoinjector with a constant flow rate (25 μL/s) at 25°C, and 3 × 30 s videos were captured with a camera level of 12 (serum-derived EVs) or 14 (THP-1 cell-derived EVs). Data were analyzed using NTA 3.4.4 software with a detection threshold of 10 and five analyses collected per sample. The NanoSight was calibrated with 100 nm polystyrene beads, and an FBS sample was run prior to samples to assess consistency of particle measurement across days.

Resuspended EVs were also visualized using transmission electron microscopy (TEM). Briefly, PFA-fixed EVs were incubated on prepared, carbon-coated grids. The grids were washed, fixed in 2% glutaraldehyde, and washed again, prior to staining with 2% filtered uranyl acetate. Images of air-dried EV grids were acquired using TEM (120vkV Hitachi HT7800, Hitachi High-Technologies Corporation, Tokyo, JP).

Immunogenicity Assessment (In Vitro Inflammasome Activation).

THP-1 macrophages (previously unexposed to treatment drugs, referred to as drug-naïve) in RPMI media (supplemented with 10% exosome-depleted FBS) were incubated with increasing concentrations of clozapine-, olanzapine-, or vehicle control-exposed EVs (0-2500 EV/cell) isolated from the in vitro or in vivo conditions described earlier. To inhibit inflammasome-dependent activation of cells by EVs, macrophages were preincubated with the selective caspase-1 inhibitor VX-765 (10 μg/mL), used in previous in vivo studies by our laboratory (Sernoskie et al., 2022). Direct drug incubations with clozapine (10 μg/mL; 30.6 μM), olanzapine (10 μg/mL; 32 μM), or drug vehicle (0.1% DMSO) were employed as controls. Cells were incubated under each condition for 24 hours, prior to the collection of conditioned media for cytokine analysis or the assessment of cell viability using a CCK-8 assay (GLPBio, Montclair, CA, USA), as described (Sernoskie et al., 2023).

Inflammatory Mediator Measurement.

Concentrations of inflammatory mediators were measured in rat serum or THP-1 macrophage conditioned media by ELISA. The following rodent kits were used for rat serum: CXCL1, growth and differentiation factor-15 (Duoset, R&D Systems, Minneapolis, MN), and corticosterone (Life Technologies Inc., Burlington, ON). The following human kits were used for THP-1 cell culture supernatants: IL-1β, CXCL1, and C-reactive protein (CRP) (Duoset, R&D Systems).

Proteomic Analysis.

EV protein samples (10 μg) were reconstituted in 50 mM ammonium bicarbonate with 10 mM Tris(2-carboxyethyl)phosphine hydrochloride (Thermo Fisher Scientific Inc., Grand Island, NY), and vortexed for 1 hour at 37°C. Chloroacetamide (Sigma-Aldrich Canada Co.) was added for alkylation to a final concentration of 55 mM. Samples were vortexed for another hour at 37°C. Trypsin (1 μg) was added, and digestion was performed for 8 hours at 37°C. Samples were evaporated to dryness using a Savant SpeedVac Vacuum Concentrator (Thermo Fisher Scientific Inc., Waltham, MA) and solubilized in 5% acetonitrile-4% formic acid. The samples were loaded on a 1.5 µL pre-column (Optimize Technologies, Oregon City, OR). Peptides were separated on a homemade reversed-phase column (150-μm i.d. by 200 mm) with a 56-minute gradient from 10% to 30% acetonitrile-0.2% formic acid and a 600-nL/min flow rate on an Easy nLC-1000 connected to a Exploris 480 (Thermo Fisher Scientific Inc., San Jose, CA). Each full mass spectrometry (MS) spectrum acquired at a resolution of 120,000 was followed by tandem-MS spectra acquisition on the most abundant multiply charged precursor ions for 3 seconds. Tandem-MS experiments were performed using higher energy collision dissociation at a collision energy of 27%. The data were processed using PEAKS X Pro (Bioinformatics Solutions, Waterloo, ON) and a Uniprot homo sapiens database (20,349 entries) or rattus norvegicus database (47,945 entries). Mass tolerances on precursor and fragment ions were 10 ppm and 0.01 Da, respectively. Fixed modification was carbamidomethyl (C). Variable selected posttranslational modifications were acetylation (N-ter), oxidation (M), deamidation (NQ), and phosphorylation (STY).

Bioinformatic Analysis.

All proteomic data were analyzed using Microsoft Excel and visualized using GraphPad Prism software (version 9.5.1) unless otherwise specified. Raw intensities extracted by PEAKS were log2 transformed, and missing values were replaced by imputation with 0 values replaced by random values in the 5% lower part of the Gaussian distribution of intensities. A fold change (FC) > 2 (|log2FC|>1) and a Pvalue<0.05 (log10P value>1.30103) were considered significant, and only these proteins were considered in subsequent analyses, compared against all identified proteins as background. Hierarchical clustering was performed with an average linker clustering method and a Pearson correlation distance measurement method on Heatmapper (http://www.heatmapper.ca/expression/).(Babicki et al., 2016) Principal component analysis was performed on SRplot ((http://www.bioinformatics.com.cn/srplot), an online platform for data analysis and visualization. Vesiclepedia (Pathan et al., 2019) and ExoCarta (Keerthikumar et al., 2016) protein databases (http://microvesicles.org/ or http://www.exocarta.org/, respectively) were used to compare previously reported EV proteins. Metascape (Zhou et al., 2019) (https://metascape.org, v3.5.20230501) was used to perform functional enrichment, interactome analysis, and gene annotation, leveraging multiple independent knowledgebases within one integrated portal.

Statistical Analysis.

Results are expressed as means ± SD, with an adjusted Pvalue < 0.05 considered statistically significant. Differences between groups were assessed using Student’s t test or one- or two-way ANOVA with the Holm–Sidak’s multiple comparison test. All tests were performed using GraphPad Prism software version 9.5.1 (GraphPad, San Diego, CA).

ResultsClozapine, But Not Olanzapine, Induces a Proinflammatory Response In Vivo.

To better understand the mechanisms preceding severe IDR onset, the early immune response to olanzapine was investigated in vivo and compared with that induced by clozapine. In patients, the average daily dose of clozapine can be more than 50 times higher than that of olanzapine (Taylor, 2004), a difference that has been proposed to contribute to the risk of IDIAG for clozapine (Ng et al., 2014). Thus, to remove this possibility, olanzapine and clozapine were administered to rats at equimolar doses (28.7 mg/kg and 30 mg/kg, respectively), such that the clozapine dose approximated the therapeutic drug levels reported in patients (Lobach and Uetrecht, 2014a; Sernoskie et al., 2022). Changes in blood leukocyte populations, chemistry profiles, and serum inflammatory mediators were assessed at 3 hours post-dose, with the endpoint corresponding to the peak of the acute clozapine-induced inflammatory response. Much like our fluperlapine observations, a similar reduction in absolute lymphocyte counts was detectable with both clozapine and olanzapine, while only clozapine induced significant increases in monocytes and neutrophils (Fig. 2). In line with these cell changes, serum levels of corticosterone, the major circulating glucocorticoid in rodents (Schwab et al., 2005), were increased with both drugs, while the neutrophil chemokine CXCL1 was only elevated with clozapine. Growth and differentiation factor-15, a stress-responsive cytokine and hormone in the transforming growth factor (TGF)-β superfamily (Wang et al., 2021) that we have found to be upregulated by IDR-associated drugs nevirapine and carbamazepine (unpublished observations), was elevated with both drugs, albeit to a greater level with clozapine. Several chemistry profile markers were also selectively or more significantly altered by clozapine, including decreases in albumin, globulin, and sodium, and increases in alanine transaminase (ALT), blood urea nitrogen, and glucose. Other blood parameters related to erythrocytes and platelets were unchanged (Supplemental Fig. 1).

Fig. 2.

Clozapine and olanzapine exhibit different but overlapping effects on circulating leukocytes and blood chemistry markers in rats. Differential blood counts, presented as absolute cell numbers (A–D), blood chemistry diagnostic profiles (E–H), and serum inflammatory mediator measurements (I–K) from female Sprague–Dawley rats following 3 h treatment with vehicle control (saline, IP injection), olanzapine (28.4 mg/kg, IP injection), or clozapine (30 mg/kg, IP injection). Differential blood counts and chemistry profiles were obtained using a VETSCAN HM5 Hematology Analyzer and VS2 Chemistry Analyzer equipped with comprehensive diagnostic profile rotors, respectively. Serum mediators were measured using rodent ELISA kits. Results are presented as the mean ± SD and statistical difference between groups was determined by a one-way ANOVA with the Holm–Sidak’s test for multiple comparisons, where ***, p < 0.001; ****, p < 0.0001 (clozapine vs. control); ^, p < 0.05; ^^, p < 0.01; ^^^, p < 0.001; ^^^^, p < 0.0001 (clozapine vs. olanzapine); +, p < 0.05; ++, p < 0.01; +++, p < 0.001; ++++, p < 0.0001 (olanzapine vs. control). BUN, blood urea nitrogen; CLZ, clozapine; OLZ, olanzapine; CTR, control. N = 6 replicates/treatment.

Having established that olanzapine did not induce a proinflammatory response in our in vitro or in vivo models, EVs were enriched from relevant clozapine-, olanzapine-, or drug-vehicle-exposed biofluids: conditioned cell culture media or serum, respectively. Enriched THP-1 cell conditioned media-derived extracellular vesicles (TEVs) and rat serum-derived EVs (SEVs) were characterized using TEM for common EV markers. Using NTA to determine EV size and concentration, an increase in particle concentration was observed for both clozapine-exposed TEVs and SEVs, but the mean diameter of clozapine-exposed TEVs was increased, while that of SEVs was decreased (Fig. 3). TEM micrographs are available in Supplemental Fig. 2, and all NTA parameters are available in Supplemental Tables 1 and 2.

Fig. 3.

Clozapine exposure significantly increases the particle concentration and alters the particle size of Eversus. Nanoparticle tracking analysis of EVs enriched from in vitro (A–C) and in vivo (D–F) models. (A–C) EVs were enriched from conditioned media of THP-1 macrophage cultures following 24 h incubation with vehicle control (0.1% DMSO), olanzapine (10 μg/mL), or clozapine (10 μg/mL). (D–F) EVs were enriched from serum of female Sprague–Dawley rats following 3 h treatment with vehicle control (saline, IP injection), olanzapine (28.4 mg/kg, IP injection), or clozapine (30 mg/kg, IP injection). Mean (A, D) and mode (B, E) EV diameters represent particle size, and EV concentration (C, F) indicates the upgraded number of particles in the initial drug-conditioned biofluid. Results are presented as the mean ± SD and statistical difference between groups was determined by a one-way ANOVA with the Holm–Sidak’s test for multiple comparisons, where *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001, with comparison groups indicated by a horizontal line on graphs. CLZ, clozapine; OLZ, olanzapine; CTR, control. N = 6 replicates/treatment.

Clozapine-exposed EVs Induce a Proinflammatory Response In Vitro.

To investigate whether clozapine-exposed EVs exhibit increased immunogenicity, PMA-differentiated THP-1 macrophages were incubated with drug- or vehicle control-exposed EVs for 24 hours, following which the release of IL-1β and other proinflammatory cytokines was assessed in conditioned media. Direct incubation of macrophages with clozapine or olanzapine (10 μg/mL) or vehicle control (0.1% DMSO) were included as positive or negative controls for inflammasome activation, based on prior data (Sernoskie et al., 2022), and cell viability was unaltered by treatment (Supplemental Fig. 3). Concentration-dependent increases in IL-1β, CXCL1, and CRP release were observed in response to incubation with either clozapine-exposed TEVs or SEVs, although not to the same levels as direct clozapine incubation (Fig. 4). Notably, mediator release was greatly attenuated by pretreatment with the caspase-1 inhibitor VX-765, indicating that the response was generated in an inflammasome-dependent manner. While control- or olanzapine-exposed TEVs did not induce an inflammatory response, higher concentrations of both SEV groups increased IL-1β and CXCL1 release. This increase in mediator release from incubations with rat-derived EVs likely represents inherent immunogenicity of the xenogeneic model (Huang et al., 2021a), as compared with the autologous incubations of THP-1 cell-derived EVersus. Still, at the highest EV concentrations tested, clozapine-exposed TEVs and SEVs exhibited the highest mediator release. Together, these findings demonstrate that clozapine induces significant alterations to EVs, both with respect to the number of EVs produced and the ability of these EVs to promote an inflammatory response.

Fig. 4.

Clozapine-exposed EVs induce a proinflammatory response and inflammasome activation in drug-naïve THP-1 macrophages. Conditioned media inflammatory mediator concentrations of IL-1β (A–C, J–L), CXCL1 (D–F, M-O), and CRP (G–I, P–R) collected from THP-1- macrophages after 24 h of incubation with TEVs (A–I) or SEVs (J–R). Groups include clozapine-, olanzapine-, or control-EV (0-2500 EV/cell), coincubation of EV and the caspase-1 inhibitor VX-765 (10 μg/mL), or direct drug incubation with vehicle control (0.1% DMSO), olanzapine (10 μg/mL), clozapine (10 μg/mL), or VX-765 (10 μg/mL). At the highest EV concentrations (250 or 2500 EV/cell), clozapine-exposed EVs exhibit significantly greater immunogenicity compared with olanzapine- or control-exposed EVs (S-T). Results are presented as the mean ± SD and statistical difference between groups was determined by a one-way ANOVA with the Holm–Sidak’s test for multiple comparisons, where *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001 (treatment group vs. 0 EV/cell); ++, p < 0.01; +++, p < 0.001; ++++, p < 0.0001 (VX-765 + [highest EV]/cell vs. [highest EV]/cell). V, VX-765.CLZ, clozapine; OLZ, olanzapine; CTR, control. N= 6 replicates/treatment.

Clozapine and Olanzapine Exposure Differentially Alter the EV proteome In Vitro and In Vivo.

After highlighting differences in the inflammatory responses with EV treatment following exposure to IDR-associated versus non-IDR-associated drugs, the next focus was to identify specific proteins that may be responsible for these observed changes. Label-free proteomic profiling of EVs was performed using tandem-MS. Through PEAKS peptide mapping, a total of 1694 proteins (1693 isoforms with unique intensities) were identified in TEVs, and 786 proteins (524 isoforms with unique intensities) were identified in SEVs (Supplemental Fig. 1). Study-identified TEV proteins exhibited substantial overlap with published Vesiclepedia and ExoCarta database-identified mammalian vesicle proteins, including 80% of the top 100 most frequently identified EV markers (Fig. 5). Conversely, about 40% of study-identified SEV proteins were previously unreported in the databases, including more than 150 rat Ig-like domain-containing proteins. Only proteins with significantly different expression among comparison groups (FC > 2; |log2FC|>1, plus a Pvalue < 0.05; log10P value > 1.30103) were considered for subsequent analyses.

Fig. 5.

Study-identified EV proteins overlap with Vesiclepedia and ExoCarta database-identified vesicle proteins. Venn diagrams illustrating comparison of all TEV proteins (A, B) and SEV proteins (C, D) identified in this study against all mammalian proteins recorded in the ExoCarta and Vesiclepedia databases (A, C) or comparison of all SEV proteins identified in this study against top 100 most commonly identified EV markers recorded in the ExoCarta and Vesiclepedia databases (B, D). Of the TEV proteins identified in the current study, 80% were previously identified EV markers, including 80% of the most commonly identified EV markers. Of the SEV proteins identified in the current study, 60% were previously identified EV markers, including 20% of the most commonly identified EV markers.

Unsupervised correlational analyses successfully and independently delineated clozapine-, olanzapine-, and control-exposed TEV and SEV groups by sample, as indicated by principal component analysis plots and hierarchical clustering heatmaps (Fig. 6). In TEVs, a total of 86 differentially expressed proteins (DEPs) were identified when comparing clozapine versus control (CLZ/CTR: 61↑, 25↓); more than double the number of DEPs identified in the comparison of olanzapine versus control (OLZ/CTR: 29↑, 13↓). Similar trends were observed in SEVs, with 80 CLZ/CTR DEPs (47↑, 33↓) and 59 OLZ/CTR DEPs (19↑, 40↓) identified. Interestingly, in the comparisons of clozapine versus olanzapine (CLZ/OLZ), 49 DEPs were identified TEVs (28↑, 21↓), while 90 DEPs were identified in SEVs (81↑, 9↓), the vast majority of which were upregulated.

Fig. 6.

Clozapine exposure markedly increases the number of DEPs identified in EVersus. Principal component analyses (PCA) (A, F), hierarchical clustering heatmaps (B,G), and volcano plots for TEV (A–E) and SEV (F–J) analyses based on all DEPs identified from significant fold change values for clozapine versus control (C, H), clozapine versus olanzapine (D, I), or olanzapine versus control (E, J), of EV samples based on significantly differentially expressed proteins for the respective study. A FC > 2 (|log2FC|>1) and a P value < 0.05 (log10p-value > 1.30103) were considered statistically significant when analyzing log2-transformed protein intensities. PCAs illustrate the relationship between the technical replicates within a given treatment group (clozapine, olanzapine, or control) and variation between different treatments, with each treatment group surrounded by a 70% confidence ellipse. The hierarchical clustering heatmap was performed with an average linker clustering method and a Pearson correlation distance measurement method, with representative dendrograms along the X and Y axes, respectively. The cutoffs for fold change size (FC > 2) and fold change significance (p < 0.05) are indicated on the volcano plots by vertical blue (downregulation) and red (upregulation) lines on the X-axis and a horizontal gray line on the Y-axis. Total upregulated and downregulated DEPs are noted in red and blue, respectively. Please refer to Supplemental File 1 for a complete list of identified proteins, fold changes, and P values. CLZ, clozapine-exposed EVs; CTR, control-exposed EVersus; OLZ, olanzapine-exposed EVs.

To identify candidate biomarkers unique to clozapine treatment, DEPs conserved among clozapine-based comparisons (CLZ/CTR and CLZ/OLZ) were evaluated. Although almost 20 clozapine DEPs from both TEV and SEVs were identified, no proteins were identified in both in vitro and in vivo models (Fig. 7, Tables 1 and 2). When comparing DEPs conserved with antipsychotic treatment (i.e., differentially expressed with both clozapine and olanzapine versus control), again, no proteins were found to overlap across models (Tables 3 and 4). A greater number of DEPs were found to overlap between clozapine and olanzapine exposure in SEVs, compared with TEVs (21 vs. 9, respectively). Finally, clozapine-upregulated DEPs were also cross-referenced with published literature of DAMPs and related signaling molecules (Santoni et al., 2015; Akdis et al., 2016; Grazioli and Pugin, 2018; Roh and Sohn, 2018; Denning et al., 2019; Gong et al., 2020; Picca et al., 2020; Zindel and Kubes, 2020) to identify mediators that may be important in the inflammatory responses observed (Tables 5 and 6). In TEVs, a number of recognized inflammatory proteins were identified, including heat shock proteins (HSPs; HSP70, adaptor ST13, HSP40/DNAJ), chemokine CCL20, and interferon I-related proteins (IFI16, PINLYP). In SEVs, TGF-β proteins (TGFB1, LTBP1), histone 2B, and kininogens (KNG1, KNG2) were among the identified inflammation-related markers. Overall, clozapine and olanzapine exposure exhibit largely nonoverlapping effects on the EV proteome, independent of the model employed.