Development of a succinyl CoA:3-ketoacid CoA transferase inhibitor selective for peripheral tissues that improves glycemia in obesity

Summary

Many individuals with type 2 diabetes (T2D) cannot take current therapies due to their adverse effects. Thus, new glucose-lowering agents targeting unique mechanisms are needed. Studies have demonstrated that decreasing ketone oxidation, secondary to muscle-specific deletion of succinyl-CoA:3-ketoacid-CoA transferase (SCOT), protects mice against obesity-related hyperglycemia. In silico studies identified that the antipsychotic diphenylbutylpiperidines can inhibit SCOT and alleviate obesity-related hyperglycemia. Because ketones are a major brain fuel, whereas the diphenylbutylpiperidines have central nervous system-related adverse effects, we aimed to develop a peripheral selective SCOT inhibitor (PSSI). Using a pharmacophore derived from the diphenylbutylpiperidine-SCOT interaction, we synthesized PSSI-51, which inhibited SCOT activity in peripheral but not brain tissue, while decreasing myocardial ketone oxidation. Importantly, PSSI-51 treatment improved glycemia in obese mice and demonstrated reduced brain accumulation compared to the diphenylbutylpiperidine pimozide. We propose that PSSI-51 can lay the foundation for optimizing a new class of brain-impermeable SCOT inhibitors for treating T2D.

Graphical abstract

Highlights

• •Systemic SCOT inhibition reduces ketone oxidation and improves glycemia in obesity
• •A peripheral selective SCOT inhibitor (PSSI-51) preserves brain SCOT activity
• •PSSI-51 inhibits peripheral tissue (i.e., muscle) SCOT activity and ketone oxidation
• •PSSI-51 improves glycemia in obese mice

Pharmaceutical compounds formulation; Pharmacology; Computational molecular modelling; Medical biochemistry; Physiology; Diabetology

Introduction

Type 2 diabetes (T2D) is a major global health problem that affects >400 million people.¹ Despite the availability of several drug classes for the treatment and management of T2D, ∼40% of people with T2D are unable to achieve or maintain normoglycemia and will require combination therapy.² Furthermore, ∼15% of individuals cannot receive the first-line therapy metformin due to reduced kidney function (decreased glomerular filtration rate), and 10% of patients are unable to continue with metformin use due to nausea and abdominal pain.^2,3 Drugs that decrease hyperglycemia in T2D via stimulating insulin secretion (e.g., sulfonylureas, glucagon-like peptide-1 receptor agonists) also have limited utility in people whose T2D is concomitant with declining β cell mass/function.^2,4 Hence, the development of new treatment options is crucial to improve the quality of life for such individuals.

Previous research reported an increased protein expression and activity of succinyl-CoA:3-ketoacid-CoA transferase (SCOT), a key enzyme involved in ketone oxidation, in the skeletal muscle of obese mice.⁵ This elevation of skeletal muscle SCOT activity was deemed to be maladaptive and contribute to the exacerbation of obesity-induced hyperglycemia, since skeletal muscle-specific deficiency of SCOT improved glucose tolerance in obese mice. To demonstrate the translation of these genetic findings at a pharmacological level, in silico modeling identified that an older generation of antipsychotic agents, diphenylbutylpiperidines (DPBPs), which include pimozide, penfluridol, and others, were capable of inhibiting SCOT activity. Intriguingly, treatment of obese mice with pimozide improved glycemia in an SCOT-dependent manner, suggesting that these older generation antipsychotics that were originally developed as dopamine 2 receptor (D2R) antagonists, may have utility in being repurposed for the treatment of T2D.⁶

Unfortunately, there are several important caveats in considering a repurposing approach for pimozide or the other DPBPs. First, the doses of pimozide for glucose-lowering greatly exceed the therapeutic index for psychiatric indications (∼10-fold). This would significantly increase the likelihood of adverse pharmacological reactions, which may include neuromotor dysfunction (e.g., tardive dyskinesia) or hyperprolactinemia resulting from D2R antagonism, and a pronounced risk of QT interval prolongation due to HERG potassium channel inhibition.^7,8 The second concern relates to ketones representing a major fuel source for the brain. Indeed, during the physiology of fasting/starvation, the liver increases the formation of ketones (β-hydroxybutyrate [βOHB] and acetoacetate) as an alternative fuel source for brain neurons, which cannot oxidize fatty acids.^9,10 The adverse effects that may result from chronic inhibition of ketone oxidation within neural tissue remains inadequately characterized, with the potential detriment to overall neurological health presenting a substantial concern.¹¹

While SCOT inhibition may appear to be a promising approach for the treatment of T2D, it is unlikely that the DPBPs would represent the ideal drug class for this purpose. Nonetheless, we hypothesize that our understanding of the DPBP-SCOT interaction can pave the way for further optimization and result in the development of a distinct class of SCOT inhibitors. Our goal was to use our knowledge surrounding the DPBP-SCOT interaction to design a therapeutic candidate with decreased ability to penetrate brain tissue, while also lacking antagonism toward the D2R. In doing so, a peripheral SCOT inhibitor should lack the primary adverse effects observed with DPBP use, while preserving the brain’s ability to utilize ketones as a fuel source. Herein, we report the development and characterization of such a compound, while assessing its potential to improve glycemia in experimental obesity.

Results

Design, synthesis, and characterization of an SCOT inhibitor with peripheral selectivity and reduced brain penetration

Our drug design protocol focused on the structures of the previously discovered SCOT inhibitors pimozide, fluspirilene, and penfluridol (i.e., DPBPs⁶) as the foundational training dataset for constructing a pharmacophore model (Figure S1A). This refined pharmacophore model was used to guide the generation of candidate inhibitors through the ligand-designer module of Schrödinger. This pharmacophore model informed the ligand designer module of Schrödinger for hit generation, employing techniques, such as bioisosteric replacement and isostere scanning to refine the molecular design. For each modified structure, docking was performed to evaluate binding affinity to SCOT, and the resulting poses were analyzed for key interactions within the active site. At the same time, the compounds were docked against the D2R binding site to ensure weak binding affinity, thus reducing the potential for off-target effects. The final set of molecules underwent additional filtering based on predicted ADMET properties, including blood-brain barrier (BBB) score. This process ensured that the shortlisted candidates were unlikely to cross the BBB, aligning with the goal of developing peripherally selective inhibitors (Figure S1B). Among the 8 compounds that met these criteria, peripheral selective SCOT inhibitor-51 (PSSI-51) was selected for experimental validation owing to its strong SCOT docking score (−9.9 kcal/mol), weakest D2R docking score (−9.8 kcal/mol), and moderate BBB penetration (Figures 1A and S1B). In addition, PSSI-51 was prioritized due to its ease of synthesis, whereas the other candidates, despite meeting the selection criteria, were not synthesized due to their complex synthetic routes and lower feasibility.

Figure 1.

Development of the SCOT inhibitor PSSI-51

(A) Schematic representation of the discovery process from pharmacophore hypothesis through hit generation to the identification of compound PSSI-51.

(B) Depiction of PSSI-51 binding within the oxyanion binding pocket of SCOT, highlighting hydrophobic interactions (red dotted lines) and hydrogen bonding (green dotted lines.

(C) The binding interaction fraction of the residues in contact with PSSI-51 during the MD simulation.

(D) RMSD comparison between unbound SCOT and SCOT bound with PSSI-51.

(E) Superimposed RMSF profiles of side chain residues of unbound SCOT versus SCOT in complex with PSSI-51.

(F) PSSI-51 positional stability within the SCOT binding pocket.

(G) Detailed schematic of the synthetic route of PSSI-51.

Molecular modeling and molecular dynamics (MD) simulations revealed that PSSI-51 binds to the oxyanion pocket of SCOT in a manner analogous to the DPBPs (Figure 1B). Quantitative analyses of amino acid contributions to this interaction highlighted the specificity and robustness of the binding affinity (Figure 1C). The structural stability of SCOT in complex with PSSI-51 was evaluated by calculating the root-mean-square deviation (RMSD) of the SCOT backbone atoms. The RMSD values confirmed that there was no significant structural deviation from the unbound form of SCOT (Figure 1D). In addition, an analysis of the root-mean-square fluctuations (RMSF) across the protein residues revealed that binding of PSSI-51 induced localized fluctuations in the side chains of residues at the binding site (Figure 1E). This suggests that while the overall structure remains stable, the PSSI-51 interaction within SCOT leads to dynamic changes at the molecular level, potentially affecting the conformational flexibility of the active site residues. The continuous monitoring of PSSI-51 within the SCOT oxyanion pocket throughout the simulation trajectories suggested a sustained stability of the inhibitor within this critical binding site (Figure 1F). Comparative MD simulation data indicate that PSSI-51 interacts with the SCOT catalytic domain more effectively than pimozide, demonstrating not only a tighter binding but also a greater number of sustained protein-ligand contacts (Figure S1C, and Data S1A and S1B). Analysis of ligand torsion angles and physicochemical properties during the simulations further supported the superior binding affinity of PSSI-51 compared to pimozide (Figure S1D, and Data S1C and S1D).

Following computational modeling, experimental validation was undertaken to confirm the initial predictions affirming PSSI-51 as the candidate to focus our efforts on. PSSI-51 was synthesized (Figure 1G) and recombinant human SCOT enzyme was prepared. PSSI-51’s inhibition was confirmed by a cell-free SCOT activity assay, which spectrophotometrically tracks the formation of acetoacetyl CoA via the transfer of a CoA moiety from succinyl CoA to acetoacetate. PSSI-51 prevented succinyl CoA from being converted to acetoacetyl CoA in a dose dependent manner (Figure 2A). Furthermore, treatment of isolated working mouse hearts with PSSI-51 decreased the oxidation of U-¹⁴C labeled βOHB, without impacting heart rate or ex vivo cardiac function (Figure 2B and2Data S2A–2I). Next, we utilized the parallel artificial membrane permeability assay (PAMPA) system to assess the potential cerebral uptake of PSSI-51. The permeability coefficient (P_e) of PSSI-51 (2.46 ± 0.85) was lower versus pimozide (8.17 ± 0.67), indicating that PSSI-51 has low permeability toward the artificial membranes and reduced capacity to cross the BBB (Figure 2C).

Figure 2.

PSSI-51 selectively inhibits SCOT without interacting with canonical targets of DPBPs

(A) Quantification of recombinant SCOT enzymatic activity and acetoacetyl CoA production rates in the presence of dimethyl sulfoxide (DMSO) (vehicle control) and PSSI-51 (500, 250, and 125 nmol/L) (n = 7–9 biological replicates).

(B) Evaluation of C¹⁴βOHB oxidation rates in the isolated working mouse heart in response to vehicle control or PSSI-51 treatment (n = 5–6 animals).

(C) Assessment of potential cerebral uptake of PSSI-51 and pimozide using PAMPA.

(D) Fluorescence-based assay in U2OS red _cAMPNomad D2R cells to measure inhibition of dopamine-induced cAMP mobilization by PSSI-51, compared with pimozide and the positive control blenonserin (n = 3 biological replicates).

(E) Quantitative fluorescence analysis of dopamine agonistic activity (n = 3 biological replicates).

(F) Schematic of targeted site-directed mutagenesis within SCOT binding site.

(G) SCOT enzymatic activity of the recombinant mutant SCOT (Y115A/I323A) and rate of acetoacetyl CoA production in the presence of DMSO (vehicle control) and the PSSI-51 (500 nmol/L) (n = 3 technical replicates).

(H) Circulating βOHB levels following PSSI-51 administration 16 h prior to oral ketone ester intake. Data are presented as mean ± SEM. Statistical significance was determined using Student’s t test or one-way ANOVA, followed by a Bonferroni post hoc correction where appropriate. Statistical thresholds set at ∗p < 0.05 vs. vehicle control, ^$p < 0.05 vs. pimozide, ^#p < 0.05 vs. dopamine. RMSD; root-mean-square deviation, RMSF; root-mean-square fluctuation, HP Std; high permeability standard, LP Std; low permeability standard; P_e: permeability coefficients.

PSSI-51 does not engage the canonical target of the DPBPs and appears to be a selective SCOT antagonist

Although our compounds were based on the pharmacophore model of the DPBP-SCOT interactions, PSSI-51 exhibited neither antagonistic nor agonistic effects on the D2R, in contrast to the inhibitory actions observed with pimozide against dopamine (Figures 2D and 2E). To further validate the binding interaction of PSSI-51 within the SCOT catalytic pocket, we conducted site-directed mutagenesis of two critical residues within the enzyme’s binding site (Tyr115:Ala and Ile323:Ala; mutant SCOT) (Figure 2F). We have previously shown that single SCOT binding pocket mutations (Ile323 and Lys368) can partially blunt SCOT enzymatic activity.⁵ However, our in silico modeling identified additional residues including Tyr115 as being essential for PSSI-51 binding, prompting us to investigate double mutations. Before making these modifications, MD simulations were performed to ensure that the mutations did not change overall protein structure or alter the structure of the oxyanion pocket. In the MD analysis of mutant and wildtype SCOT, the mutant SCOT first exhibits an RMSD shift in the protein backbone suggesting local conformational changes due to residue substitution. Despite these initial deviations, the overall side chain RMSF values remained consistent with those of the wildtype, indicating that the global dynamics of the protein were preserved post-mutation (Figures S2A and S2B). The secondary structure of the protein was also monitored during MD simulations and the snapshots of the wild type and mutant SCOT confirmed secondary structure integrity, showing that the mutations didn’t disrupt the protein’s structural architecture (Figures S2C and S2D). Moreover, simulation snapshots visually support the preservation of structural and functional integrity in both wild type and mutant SCOT (Figure S2E).

The interaction of PSSI-51 with mutant SCOT was characterized by marked differences from the wild type form in terms of induced conformational dynamics. This was evidenced by deviations in backbone RMSD, alterations in sidechain and ligand atom RMSF, and changes in binding interaction patterns (Figures S3A–S3D). The comparative analysis of torsional angle distributions for PSSI-51, in both its unbound state and when complexed with mutant and wild-type SCOT, offered insights into the structural and energetic determinants governing ligand specificity and stability within the active site. Notably, this analysis indicated a significant divergence from the interaction patterns observed with wild-type SCOT, suggesting that the targeted mutations critically influence the binding affinity of PSSI-51, making it improbable that it can inhibit mutant SCOT (Figures S3E and S3F). To validate our in silico findings, mutant SCOT protein was found to exhibit reduced enzymatic efficiency, but as expected PSSI-51 was unable to inhibit mutant SCOT activity (Figure 2G). This underscores the necessity of key residues within the oxyanion pocket of SCOT for binding and inhibition by PSSI-51, further validating the compound’s specific interaction with the catalytic pocket of SCOT.

As further evidence that PSSI-51 can inhibit SCOT in vivo, we performed a ketone clearance test in lean mice. Administration of PSSI-51 14 h prior to treatment with an oral ketone ester (1,719 mg/kg) caused significant elevations in circulating βOHB levels in both male and female mice when compared to their placebo treated counterparts (Figures 2H and S4A–S4D).

PSSI-51 induces glucose-lowering in experimental obesity

To determine whether PSSI-51 retained the glucose-lowering actions seen with other SCOT inhibitors such as the DPBPs, male C57BL/6J mice were fed a high-fat diet (HFD; 45% kcal from lard) for 8-week to induce obesity. A separate cohort of mice were fed a low-fat diet (10% of kcal from lard) for an equivalent duration (healthy lean controls). During the final 4-week, mice were randomized to oral gavage with vehicle control, pimozide (10 mg/kg), or PSSI-51 (10 mg/kg) once every 48 h (Figure 3A). Consistent with SCOT inhibition, PSSI-51 treatment increased fasting βOHB levels, and improved intraperitoneal glucose tolerance, mirroring the effects observed with pimozide (Figures 3B and 3C). This improvement in glycemia did not result in an elevation of circulating insulin levels at 30-min post-glucose ingestion, suggesting a potential improvement in insulin sensitivity (Figure 3D). Moreover, the PSSI-51 mediated improvement in glycemia was not associated with reductions in body weight or adiposity (Figures 3E–3G).

Figure 3.

PSSI-51 inhibits ketone oxidation and improves glycemia in obese mice

(A) Schematic diagram outlining the experimental approach.

(B) Circulating βOHB levels (n = 6–10 animals).

(C) Glucose tolerance tests (performed one day following the final treatment) (n = 6–10 animals); (D) corresponding circulating insulin levels (n = 5–6 animals); (E–G) and body weight and adiposity changes following drug treatment (n = 5–6 animals).

(H) Metabolic parameters using indirect calorimetry (24 h) to assess total food and water intake, whole-body energy expenditure, locomotor and ambulatory activity and respiratory exchange ratio (n = 4–5 animals). Data presentation includes mean ± SEM. Statistical analyses were conducted using Student’s t test or one-way ANOVA, supplemented by Bonferroni post hoc tests, with significance thresholds set at ∗p < 0.05 versus vehicle control.

Indirect calorimetry studies suggested that PSSI-51 treatment led to a slight increase in food intake but not water consumption, which although not significant in itself, coincided with a mild elevation of energy expenditure, while locomotor and ambulatory activity were unaffected. A mild elevation in the respiratory exchange ratio (indicative of increased carbohydrate oxidation) was also observed in response to PSSI-51 treatment, consistent with previous studies using pimozide to inhibit SCOT⁵ (Figure 3H). Similar patterns were observed when segmenting the diurnal and nocturnal phase, demonstrating that the metabolic impacts of PSSI-51 are stable across circadian cycles (Figures S4E and S4F). Importantly, PSSI-51 administration did not induce glucose-lowering in lean mice, suggesting that hypoglycemia would not be an adverse effect associated with SCOT inhibition (Figures S4G–S4I). In contrast, treatment with either PSSI-51 or pimozide failed to elevate fasting βOHB levels or improve glucose tolerance in obese female mice (Figures S4J–S4M).

PSSI-51 exhibits minimal brain penetration and lacks the adverse effects associated with DPBPs

We next performed several studies to confirm the limited cerebral permeability of PSSI-51 in obese mice treated for 2-week with PSSI-51. As anticipated, pimozide treatment led to a significant reduction in SCOT activity across several organs known to be avid metabolizers of ketones (e.g., brain, kidney, soleus muscle). Although PSSI-51 mirrored pimozide’s ability to inhibit SCOT activity in the kidney and soleus muscle, it failed to decrease SCOT activity in the brain (Figure 4A). PSSI-51-mediated SCOT inhibition was not associated with perturbations in protein expression of SCOT or other key enzymes involved in ketone oxidation (Figure S4N).

Figure 4.

Selective Inhibition of peripheral ketone body metabolism by PSSI-51

(A) SCOT activity assay in the brain, kidney and soleus muscle isolated from mouse treated PSSI-51, and pimozide (n = 3–4 animals).

(B) Representative PET images of untreated fed and untreated fasted mice 1 h post injection of S-[¹⁸F]FβOHB and time-activity curves for fasted to fed brain and muscle over 60 min (n = 4 animals).

(C) Representative PET images of vehicle treated (fasted) and PSSI-51 treated (fasted) mice 1 h post injection of S-[¹⁸F]FβOHB and time time-activity curves for untreated to treated brain and muscle over 60 min (n = 3–5 animals).

(D) Parameters of the activity wheels measured after treatment with PSSI-51 or pimozide after to 24 and 48 after treatment (n = 3 animals per group).

(E) Mass spectrometric quantification of PSSI-51:pimozide ratio following a single dose of the drug (n = 3 animals). Data are presented as mean ± SEM. Statistical analyses were conducted using Student’s t test or one-way ANOVA, supplemented by Bonferroni post hoc tests, with significance thresholds set at ∗p < 0.05 vs. vehicle control, ^#p < 0.05 vs. pimozide.

To further corroborate PSSI-51’s inability to inhibit SCOT activity, S-[¹⁸F]FβOHB, a radiolabeled analogue of βOHB, was synthesized as previously described for positron emission tomography imaging studies to monitor in vivo βOHB utilization.¹² We first performed studies in random fed versus overnight fasted C57BL/6J mice, and as expected we observed enhanced accumulation of S-[¹⁸F]FβOHB in the brains of the fasted mice, compatible with fasting-mediated elevations of ketone oxidation (Figure 4B). When overnight fasted mice were treated with PSSI-51, no decrease in brain accumulation of S-[¹⁸F]FβOHB was observed, confirming minimal impact on brain ketone oxidation (Figure 4C). Of interest, PSSI-51 treatment showed a trend toward increasing brain ketone oxidation, which is compatible with systemic SCOT inhibition increasing circulating ketone levels (Figure 3B) and subsequent brain exposure.

To further investigate the systemic effects of PSSI-51, particularly its potential impact on the CNS and anticipated neurobehavioral actions, mice were housed in cages with voluntary run wheels following PSSI-51, pimozide or VC treatment. Various activity-related parameters were monitored over a 48 h period, whereby treatment with PSSI-51 had no effect on locomotor functions, such as distance traveled, maximum speed, and acceleration, whereas mice treated with pimozide exhibited reductions in several of these parameters (Figure 4D).

To quantify the extent of cerebral uptake of PSSI-51, mice were administered a single oral dose of either PSSI-51 or pimozide, and at 16 h post-treatment their brain tissues were collected for quantitative analysis. High-performance liquid chromatography (HPLC) revealed that the concentrations of PSSI-51 in cerebral tissues were markedly reduced compared to those observed with pimozide, consistent with PSSI-51 having limited permeability across the BBB (Figure 4E and Data S3A–S3C).

Pharmacokinetics and metabolic stability of PSSI-51

In order to glean some information regarding exposure and pharmacokinetics of PSSI-51, oral doses (10 mg/kg) were administered to male Sprague-Dawley rats followed by serial blood sampling from implanted jugular vein cannulas, as rats provided the required blood volumes for reliable pharmacokinetic analysis compared to mice. Following oral dosing via gavage, the blood concentrations peaked within 2 h and were measurable for 24 h after a single dose using a liquid chromatography-tandem mass spectrometry assay (LC-MS/MS) (Figures S4O–S4Q). The profile displayed multi-exponential decline after the maximum concentration (Cmax) was reached, with a terminal phase half-life of 8 h. The blood oral clearance (CL/F) and volume of distribution (Vd/F) were both high.

Comparative information regarding the blood distribution and in vitro metabolism of PSSI-51 and pimozide were collected. The blood to plasma ratios of PSSI-51 and pimozide were 0.88 ± 0.08 and 0.72 ± 0.04, respectively. The unbound fraction in plasma of pimozide were 0.042 ± 0.004% and 0.044 ± 0.005% at spiked concentrations of 1000 and 5000 ng/mL, respectively. In contrast, unbound fractions of PSSI-51 were 0.021 ± 0.001% and 0.019 ± 0.001%, respectively for 1000 and 5000 ng PSSI-51/mL plasma.

The mean microsomal CLint and t½ of decline of pimozide in the presence of rat liver microsomes and NADPH were 8.7 μL/min/mg protein and 80 min, respectively. In human microsomes, the corresponding CLint and t½ of pimozide in the presence of human liver microsomes and NADPH were 3.4 μL/min/mg protein and 201 min, respectively. In contrast, PSSI-51 exhibited much faster declines in rat and human microsomes. Its mean microsomal CLint and t½ in rat liver microsomes with NADPH were 101 μL/min/mg protein and 3.9 min, respectively; the corresponding estimates with exposure to human liver microsomes were 39 μL/min/mg protein and 10.2 min, respectively.

Discussion

In this study we report that PSSI-51 has preferential actions to selectively inhibit peripheral versus central SCOT activity, while retaining the glucose-lowering actions in obesity that have previously been observed with the DPBPs. Because the DPBPs have CNS-related adverse effects including extrapyramidal symptoms, while ketones are a key fuel source for the brain, we reasoned that the ideal SCOT inhibitor for the treatment of T2D should have limited brain penetration. Such an approach would preserve the brain’s ability to utilize ketones for ATP production, while decreasing the potential risk of CNS-related adverse effects. Furthermore, we confirmed that PSSI-51 does not inhibit the D2R, the canonical target of the DPBPs, suggesting that the extrapyramidal symptoms associated with DPBP use would not be present with PSSI-51.

Our previous studies demonstrated that genetic deletion of SCOT in skeletal muscle improved glycemia in mice subjected to experimental obesity, suggesting that SCOT inhibitors may have utility in the clinical management of T2D. Through in silico modeling we identified that the DPBPs (e.g., pimozide) could inhibit SCOT, and observed that obese mice treated with these agents recapitulated our observations in skeletal muscle-specific SCOT deficient mice. As the glucose-lowering actions in response to SCOT inhibition were associated with lower circulating insulin levels, we propose that SCOT inhibitors may be a suitable monotherapy for T2D, or adjunct therapy for glucose-lowering agents that increase insulin secretion (e.g., glucagon-like peptide-1 receptor agonists). Intriguingly, PSSI-51 produced equivalent improvements in glucose tolerance versus pimozide in obese mice, while once again being associated with lower circulating insulin levels. While pimozide treatment decreased overall animal activity when housed in cages with voluntary run wheels, no such actions were observed following treatment with PSSI-51, suggesting the potential absence of behavioral side effects commonly associated with DPBPs, consistent with its decreased CNS uptake. Of importance, PSSI-51 did not lower glycemia in lean mice, suggesting that SCOT inhibition with PSSI-51 does not induce hypoglycemia, which mimics observations in standard chow fed skeletal muscle-specific SCOT deficient mice.¹³

It is also worth noting that both PSSI-51 and pimozide only improved glycemia in obese male mice, whereas no benefit was observed in female mice. Reasons for these discrepant findings are unknown but may relate to the lack of major increases in body weight observed in female C57BL/6J mice fed an HFD for 8-week (Figure S4M). It also remains possible that this is a true sex-specific difference whereby SCOT inhibition may only have clinical utility in the management of T2D in men but not women.

Predicting the permeability of small molecules across the BBB presents considerable challenges. In the case of pimozide, despite in silico predictions indicating low BBB penetration, the compound indeed penetrates the BBB. Conversely, other DPBPs such as penfluridol and fluspirilene exhibit high brain penetration scores. This discrepancy underscores the complexity of BBB permeability, which is influenced by multiple factors including protein binding in blood and brain, and possible cross-membrane transport facilitated by solute carrier and ATP binding cassette proteins.¹⁴ For PSSI-51, structural modifications aimed at reducing brain permeability such as increasing the number of hydrogen bond donors and acceptors, enhancing the topological polar surface area, decreasing lipophilicity, and improving solubility are theoretically justified.^15,16,17 However, the observed permeability of PSSI-51 may still be significantly influenced by these mechanisms, necessitating empirical validation. While in silico predictions and the PAMPA data provide valuable preliminary insights into a drug’s pharmacokinetic properties, they are not definitive. Given the complexity of BBB permeability and potential for unexpected interactions, thorough empirical testing is crucial. Thus, in vivo studies using animal models are essential to assess the pharmacokinetics and pharmacodynamics of compounds with unknown CNS permeability. Additionally, designing new drugs based solely on predicted BBB scores requires further empirical validation.

Based on the data from oral dosing, PSSI-51 appears to have a rapid CL/F and extensive distribution to tissues, although the lack of an intravenous formulation with pharmacokinetic data limits interpretation of the oral data. Nevertheless, the half-life of 8 h, despite the high CL/F, appears to be attributable to extensive distribution to tissues. In microsomes, both SCOT inhibitors (PSSI-51 and pimozide) appeared to be cleared more rapidly in the presence of rat than human microsomes. It was also apparent that the CLint of PSSI-51 exceeded that of pimozide in both species of microsomal protein. In plasma both PSSI-51 and pimozide had extensive protein binding (>99.9%), although pimozide appeared to have an unbound fraction about twice that of PSSI-51. There was no evidence of saturation of the binding over the concentration range of 1000–5000 ng/mL for either SCOT inhibitor. The uptake into blood cells for both compounds was limited, resulting in a blood to plasma concentration ratio of less than one.

Taken together, we have characterized a new chemical entity, PSSI-51, that decreases SCOT activity and ketone oxidation with preferential actions for peripheral versus brain tissue. Of clinical relevance, PSSI-51 retained the glucose-lowering actions observed with other SCOT inhibitors such as the DPBPs, thereby further validating that SCOT inhibition improves glycemia. However, comprehensive preclinical investigations are essential to confirm the long-term safety and efficacy of such an approach. With cardiovascular disease representing the number one cause of death for people living with T2D, it will be imperative to assess the actions of PSSI-51 on cardiac function. We have performed preliminary assessments in mice subjected to an experimental model of diabetic cardiomyopathy that produces diastolic dysfunction, whereby male C57BL/6J mice were fed a high-fat diet (60% from lard) for 12-week with a single low-dose injection of streptozotocin (75 mg/kg) provided at week 4. During the final 4-week, mice were randomized to treatment once every 48 h with either vehicle control or PSSI-51 (10 mg/kg) via oral gavage. Importantly, PSSI-51 treatment did not decrease systolic (e.g., ejection fraction) or diastolic (e.g., E/e′ ratio) function in mice with diabetic cardiomyopathy (Data S4), suggesting that PSSI-51 does not worsen cardiovascular risk. Even if PSSI-51 is not the ideal SCOT inhibitor for moving to translational testing in humans, we propose that the strategies we used to develop PSSI-51 can lay the foundation for further optimization of new chemical entities classified as PSSI with limited brain permeability for the potential treatment of T2D.

Limitations of the study

Although our detailed in silico studies have elucidated the binding mechanism of PSSI-51 in complex with SCOT, a limitation of our study is that our co-crystallization attempts with PSSI-51 have so far been unsuccessful. PSSI-51 may be hindering crystal nucleation, particularly evident when attempting co-crystallization, which fails to produce crystals containing the compound only yielding unbound structures. Furthermore, precipitation occurs when introducing the compound into an aqueous buffer containing the protein, indicating solubility challenges within the concentration ranges necessary for crystallization. Pre-formed crystals of SCOT soaked in PSSI-51 had a drop in diffraction quality of crystals impeding data analysis. This is likely due to the crystal lattice not tolerating conformational changes to accommodate PSSI-51 and could also be a sign of solubility issues at the concentrations required for protein crystallization.¹⁸ While the efforts to co-crystallize PSSI-51 bound to SCOT are ongoing; once achieved, the resulting crystal structure will be deposited in the Protein Data Bank to advance our research findings.

While the HPLC assay we developed indicates that PSSI-51 does not significantly penetrate the BBB in comparison to pimozide, these conclusions are based on measurements of drug concentration in brain tissue collected 16 h post-administration. This time point is considered likely to provide an accurate reflection of the drug’s concentration peak or steady state since we observed the maximum effect at this interval. It will be necessary to interrogate PSSI-51’s temporal and spatial pharmacodynamics within the systemic circulation in future studies, employing multiple time point sampling for extensive analysis of pharmacokinetic parameters, including the drug’s half-life, volume of distribution, and clearance rates. Nevertheless, PSSI-51 was still detectable in brain tissue, though it likely did not reach levels to decrease ketone oxidation, reflected by the absence of SCOT inhibition in brain lysates.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to the lead contact for this manuscript, John R. Ussher (jussher@ualberta.ca).

Materials availability

PSSI-51 is available and can be supplied for experimental use upon request from the lead contact following completion of a material transfer agreement. Recombinant proteins, plasmids, and other relevant materials used in this study are also available for experimental use upon request from the lead contact.

Data and code availability

• •All data reported in this paper will be shared by the lead contact upon reasonable request.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon reasonable request.

Acknowledgments

This research was supported by a Project Grant from the Canadian Institutes of Health Research to J.R.U. (PS 178029). J.R.U. is supported by a Tier 2 Canada Research Chair in Pharmacotherapy of Energy Metabolism in Obesity (CRC 232612). S.A.T.D. is supported by a Postdoctoral Fellowship from the CIHR (MFE-186352). M.O. is supported by grants from the Natural Sciences and Engineering Research Council of Canada (2024-06426) and the Canada Foundation for Innovation (38496). The in silico modeling studies required access to the Digital Research Alliance of Canada and SHARCNET online servers.

Author contributions

S.A.T.D. and J.R.U. conceived the project and oversaw the experimental work. S.A.T.D., K.Y., H.A.-N., A.A.G., M.W., J.W., R.A.F., C.T.S., J.S.F.C., R.K.B., I.A.M.-B., T.S., K.G., F.E., S.R.F., C.S.W., and M.E.C. performed experiments. S.A.T.D, H.A.-N., J.E.C., M.O., C.A.V.-M., J.N.M.G., F.W., D.R.B., and J.R.U. analyzed the data. S.A.T.D., and J.R.U. prepared the figures. All authors contributed to writing the manuscript and its discussion.

Declaration of interests

The University of Alberta has filed a patent application regarding the subject matter of this article.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-mouse IgG	Cell Signaling Technology	Cat# 7076S; RRID: AB_330924
Anti-rabbit IgG	Cell Signaling Technology	Cat# 7074S; RRID: AB_2099933
Mouse monoclonal anti-Hsp90	BD Biosciences	Cat# 610418; RRID: AB_397799
Rabbit polyclonal anti-ACAT1	Cell Signaling Technology	Cat# 44276S; RRID: AB_2799262
Rabbit polyclonal anti-BDH1	Novusbio	Cat# NBP1-88673; RRID: AB_11009202
Rabbit polyclonal anti-SCOT	ProteinTech	Cat# 12175-1-AP; RRID: AB_2157444
Recombinant DNA
pNIC-CTHF	pNIC-CTHF was a gift from Opher Gileadi	plasmid# 26105; http://n2t.net/addgene:26105; RRID:Addgene_26105
OXCT1	OXCT1 was a gift from Nicola Burgess-Brown	plasmid# 38922; http://n2t.net/addgene:38922; RRID:Addgene_38922
pRSET_A_hOXCT1-Y115A I-323A	ThermoFisher-GeneArt	Cat# VLGS02KDE
Experimental models: Organisms/strains
C57BL/6J mice	The Jackson Laboratory	Strain# 000664
Bacterial and virus strains
Escherichia coli BL21 Star (DE3)	ThermoFisher	Cat# C600003
Sprague-Dawley rats	Charles River Laboratories	Strain Code 001
U2OS red cAMPNomad_D2R cells	Innoprot, Spain	P20316
Chemicals, peptides, and recombinant proteins
Pimozide	Sigma-Aldrich	Cat# P1793
Streptozotocin	Sigma-Aldrich	Cat# S0130
Nicotinamide adenine dinucleotide phosphate tetrasodium (NADPH)	Sigma-Aldrich	Cat# 481973
Insulin (Humulin)	Eli Lilly and Co.Bio	Cat# VL7510
[U-14C] βOHB	American Radiolabeled Chemicals	Cat# ARC 1545
Kryptofix® 222 (K222)	TCI chemicals	Cat# H0932
CH3CN (anhydrous)	Fisher Scientific	Cat#AC610961000
(2S)-(+)-glycidyl tosylate	Sigma-Aldrich	Cat #540129
Recombinant nitrilase enzyme	Sigma-Aldrich	Cat# 04529
Pooled male human liver microsomes	Xenotech	Cat# MX00801
Pooled male Sprague-Dawley rats’ liver microsomes	Xenotech	Cat# MX000100
Ketone Monoester	HVMN	https://ketone.com
2-(3-bromopropyl)-1H-isoindole-1,3(2H)-dione	Combi-blocks	Cat# HC-4156
1-(bis(4-fluorophenyl)methyl)piperazine	Sigma-Aldrich	Cat# 552402
Succinyl CoA	Sigma-Aldrich	Cat# 108347973
DMEM F12	Sigma-Aldrich	D6421
Opti-MEM	ThermoFisher	31985070
Dopamine	Sigma-Aldrich	H8502
Blonanserin	Sigma-Aldrich	B7188
Software and algorithms
ImageJ	N/A	https://imagej.nih.gov/ij/
Prism v.10.2.3	GraphPad	https://www.graphpad.com/scientificsoftware/prism/
ChimeraX v.1.7	Meng et al.19	https://www.cgl.ucsf.edu/chimerax/
Molstar	RCSB PDB20	https://doi.org/10.1093/nar/gkab314
AutoDock Vina v.1.2.3	Forli et al.21	https://vina.scripps.edu/
Schrödinger Maestro Release 2020-3	Maestro, Schrödinger, LLC, New York, NY, 2020.	https://www.schrodinger.com/
Desmond	Schrödinger, Inc.	https://www.schrodinger.com/platform/products/desmond/
COOT	N/A	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
PHENIX	Liebschner et al.22	https://www.phenix-online.org/
EZStart	Shimadzu	https://www.shimadzu.com/
CalRapp	Mina et al.23	https://calrapp.org/
BIO-ACTIV2-SOFT	BIOSEB	https://bioseb.com/en/content/6-software
VMD	Humphrey et al.24	http://www.ks.uiuc.edu/Research/vmd/
ROVER v.2.0.51	ABX GmbH, Radeberg, Germany	https://abx.de/
Critical Commercial Assays-Kits
MCSG-1 crystal screen	Molecular Dimensions	Cat# MCSG-1
MCSG-2 crystal screen	Molecular Dimensions	Cat# MCSG-2
MCSG-3 crystal screen	Molecular Dimensions	Cat# MCSG-3
Top96 crystal screen	Molecular Dimensions	Cat# TOP96-10ML
Equilibrium dialysis plate	Life Technologies, Burlington ON Canada	Cat# 90006
Ultrasensitive Mouse Insulin ELISA Kit	ALPCO	Cat# 80-INSMSU-E01
Rapid Equilibrium Dialysis (RED) Device	Thermofisher
SuperQuick® BBB-PAMPA Kit	Creative Biogene Inc	DPK-YS003
Other
Low-fat diet (10% kcal from lard)	Research Diets	Cat# D12450J
High-fat diet (45% kcal from lard)	Research Diets	Cat# D12451
High-fat diet (60% kcal from lard)	Research Diets	Cat# D12492
LUNA ® C18(2) column (100 Å, 250 × 10 mm, 10 mm)	Phenomenex	Part# 00G-4253-N0

Experimental model and study participant details

Animal models, care and experimentation

The Health Sciences Animal Welfare Committee at the University of Alberta approved all experimental protocols involving animals, ensuring care agreed with the Canadian Council on Animal Care’s guidelines. Mice were accommodated in an environment maintained at 22°C with a consistent light-dark cycle of 12 h each, starting with lights on at 6 a.m. Their living conditions included chip bedding, a plastic hideaway, crinkle paper weighing 8 ounces, and Nestlet square bedding made from short fiber cotton. They also received unrestricted access to food and water. For the obesity study, male C57BL/6J (The Jackson Laboratory) mice at 8 weeks of age were put on a diet with either low-fat (10% calories from lard, D12450J) or high-fat content (45% calories from lard, D12451) for a period of 12 weeks. Upon the study’s conclusion, following an overnight fast and a 2 h refeed, the mice were euthanized, their tissues were immediately frozen in liquid nitrogen with liquid nitrogen pre-cooled Wollenberger clamps, and stored at −80°C. Regarding the cardiac function studies, 12-week-old male C57BL/6J mice were fed a high-fat diet (60% calories from lard, D12492) for 12-week, with streptozotocin (75 mg/kg) administered at week 4, while lean mice received standard chow. For the pharmacokinetics studies, male Sprague-Dawley rats cannulated at the jugular vein were purchased from Charles River Laboratories (Raleigh, NC). Although we performed studies to validate the glucose-lowering actions of PSSI-51 in obesity in both male and female mice, because glycemia was not improved in females, the majority of our follow-up studies characterizing PSSI-51 in mice were only performed in males. All animals were randomized into our specific treatment groups and further randomized across cages to avoid cage-specific effects.

Human microsomes

Pooled male human liver microsomes at a concentration of 20 mg/mL in 250 mM sucrose buffer were purchased from XenoTech-BIOIVT (MX00801).

Cell lines

U2OS red cAMPNomad_D2R cells (Innoprot, Spain) were maintained in DMEM/F-12 (Sigma-Aldrich D6421) supplemented with 10% FBS. The cell line was tested and confirmed to be free of mycoplasma, bacterial, and fungal contamination through functional assays. Short tandem repeat (STR) profiling or additional authentication methods were not performed.

Bacterial strains

Escherichia coli was used for protein expression, specifically the Bl21 DE3 strain.

Method details

Chemistry

All reagents and solvents were purchased from Sigma-Aldrich/Combi-blocks and used as received. Monitoring of all chemical reactions was achieved through the use of RediSep thin-layer chromatography (TLC) plates, with UV light for visualization. The Electrothermal melting point apparatus (Thermofisher, USA) provided melting point data, which were not corrected. The spectroscopic analysis of 1H, 13C, was performed using a Bruker FT-600 MHz spectrometer (operating at 600 MHz, 150 MHz, and 565 MHz, respectively) with Water-d2 as the solvent and tetramethylsilane for calibration. Chemical shifts (δ) and coupling constants (J) were reported in parts per million and Hertz, respectively. Descriptions of signal multiplicity included s (singlet), d (doublet), t (triplet), and m (multiplet). Microwave-assisted synthetic processes were conducted in a Biotage Initiator Reactor.

2-(3-(4-(bis(4-fluorophenyl)methyl)piperazin-1-yl)propyl)-1H-isoindole-1,3(2H)-dione (PSSI-51): The reaction mixture was prepared by mixing 2-(3-bromopropyl)-1H-isoindole-1,3(2H)-dione (1.1 equivalents, 0.038 mol), 1-(bis(4-fluorophenyl)methyl)piperazine (1 equivalent, 0.035 mol), and potassium carbonate (3 equivalents, 0.104 mol) in Dimethylformamide (DMF) (20 mL). The mixture was transferred to a microwave vial and heated at 90°C in a microwave reactor for 12 h under continuous magnetic stirring. Thin Layer Chromatography (TLC) was used to monitor the reaction progress. Upon completion, the reaction mixture was extracted into the ethyl acetate and the organic phase was evaporated under reduced pressure. The crude product was then purified using flash column chromatography with ethyl acetate and hexane as the mobile phase, resulting in the isolation of PSSI-51 as a brown semisolid. % Yield: 86. The dihydrochloride salt of the PSSI-51 was then formed by dissolving the purified fraction in 1 M HCl (2 equivalents) in diethyl ether and the precipitate was washed multiple times with a cold solution of diethyl ether. % Yield: 95. 1H NMR (600 MHz, water-D2O) δ 7.76 (ddt, J = 21.1, 6.0, 3.1 Hz, 4H), 7.59 (ddd, J = 8.3, 5.2, 2.3 Hz, 4H), 7.18 (td, J = 10.6, 2.6 Hz, 4H), 5.46 (d, J = 1.8 Hz, 1H), 3.70 (td, J = 6.3, 3.5 Hz, 2H), 3.61 (s, 4H), 3.52–3.49 (m, 4H), 3.32 (td, J = 8.1, 2.2 Hz, 2H), 2.13–2.04 (m, 2H). δ. 13C NMR (151 MHz, D2O) δ 170.47, 163.99, 162.35, 134.86, 131.22, 130.38, 129.38, 123.47, 116.93, 74.03, 54.44, 48.81, 48.44, 34.64, 22.90. HRMS (m/z): [M]− calcd for C28H27F2N3O2, 475.2071; found, 476.2143.

PSSI-51 was synthesized to a minimum >95% purity (Data S5A and S5B).

Assessment of glucose/ketone tolerance

Mice were randomly selected to be administered, through oral gavage, either vehicle control (corn oil), pimozide (10 mg/kg) or PSSI-51 (10 mg/kg) acutely (single dose) or every other day for 4 weeks. All animals underwent intraperitoneal glucose (2 g/kg) tolerance tests at 2 weeks post-treatment, with samples collected from mouse tail whole blood at 0, 5, 10, 15, 30-, 60-, 90-, and 120-min post-glucose administration using the Contour Next blood glucose monitoring system (Bayer). In a separate cohort of animals, lean mice received a single dose of PSSI-51 (10 mg/kg) after 3 h of fasting, and after 2 h, all mice were administered either a placebo drink (provided by Health Via Modern Nutrition, HVMN) or a ketone monoester (1,719 mg/kg, HVMN). Mouse tail whole blood samples were collected at 0, 5, 10, 15, 30-, 60-, 90-, and 120-min post-gavage for the measurement of circulating βOHB levels using the FreeStyle Precision Neo blood ketone monitoring system (Abbott) and insulin levels using an enzyme-linked immunosorbent assay kit (Alpco Diagnostics).

SCOT activity assay

For in-tissue activity assay kidney, soleus, and brain tissues from the vehicle, pimozide, and PSSI-51 treated mice were homogenized in phosphate-buffered saline (pH 7.2) supplemented with protease inhibitors (Halt Protease Inhibitor Cocktail; ThermoFisher). The homogenates or recombinant protein samples were centrifuged at 20,000 x g for 20 min at 4°C. The supernatant was collected for the SCOT activity assay. The enzymatic reaction was initiated by adding tissue lysates (100 mg) or recombinant SCOT enzyme (300 μg) to a reaction mixture containing 50 mM Tris HCl (pH 8.0), 10 mM MgCl2, 0.2 mM succinyl CoA, 0.1 mM lithium acetoacetate, and 4 mM iodoacetamide. Absorbance changes at 313 nm were recorded every 30 s from 0 to 3 min at room temperature. To quantify the amount produced, the absorbance values were plotted and normalized against a standard curve generated using known concentrations of chemically synthesized acetoacetyl CoA. The resulting data were then used to calculate the rate of acetoacetyl CoA synthesis.⁶

Isolated working heart perfusion

Mice were anesthetized with an intraperitoneal injection of sodium pentobarbital (60 mg/kg). Following this, their hearts were excised and immediately immersed in an ice-cold Krebs-Henseleit bicarbonate solution. The aorta was then cannulated, and the hearts were equilibrated under Langendorff conditions before being transitioned to a working heart model, as previously described.²⁵ During perfusion, the hearts were infused through the left atrium with an oxygenated Krebs-Henseleit solution containing 0.8 mM [9,10-3H] palmitate complexed with 3% fatty acid-free bovine serum albumin, 5.0 mM glucose, 10 nM PSSI-51, and 0.8 mM [U-¹⁴C] βOHB at a preload pressure of 11.5 mmHg. The outflow from the hearts was subjected to a 50-mmHg hydrostatic afterload via the aortic outflow line. This aerobic perfusion was maintained for 60 min, throughout which the oxidation of βOHB was closely monitored. Immediately following perfusion, the hearts were rapidly frozen in liquid nitrogen pre-cooled Wollenberger clamps and subsequently stored at −80°C.

Recombinant protein production and purification

Wildtype and mutant SCOT (GeneArt, Invitrogen) were purified and cloned with an N-terminal six histidine tag.⁶ Protein expression was induced using 0.2 M IPTG at 18°C, and cells (Bl21 DE3) were harvested after 16 h. Pellets were resuspended in 50 mM Tris pH 7.5, 200 mM NaCl, 5 mM β-mercaptoethanol, 20 mM imidazole and sonicated. Lysed cells were centrifugated, the supernatant was incubated with 5 mL of Ni2+Sepharose (GE Healthcare Life Sciences) and bound protein was eluted using lysis buffer supplemented with 500 mM imidazole. Additional purification was conducted utilizing a Superdex 200 column, which was pre-equilibrated with gel filtration buffer consisting of 20 mM 2-[4-(2-hydroxyethyl)piperazin-1-yl]ethanesulfonic acid at pH 7.0, 50 mM NaCl, and 1 mM dithiothreitol. Subsequently, SCOT was concentrated using a 30 kDa MWCO filter (Amicon-Millipore) to achieve a final concentration of 30 mg/mL.

SCOT structural studies and X-Ray crystallography

Human SCOT has been previously crystallized (PDB: 3DLX). Co-crystallization trials were carried out with 30 mg/mL of SCOT incubated with an excess of 2x, 4x, and 10x of PSSI-51 and incubated for 24 h, and crystallization screens were set up. We made use of Top96, MCSG-1, MCSG-2, and MCSG-3 crystal screen kits by Anatrace. Crystal growth was observed in two conditions: 25% PEG 3350, and 15% PEG 3350 with 0.2 M NaCl and 0.1 M Tris pH 8.7. Crystals were looped and cryoprotected with the addition of glycerol to a final concentration of 25% v/v and flash-frozen in liquid nitrogen. Data was collected on our home source (Rigaku MicroMax-007 HF with a Dectris PILATUS3 R 200K-A detector), and structures were solved using Python-based Hierarchical Environment for Integrated crystallography (PHENIX),²² specifically molecular replacement utilizing the previously solved structure 3DLX at 1.92 A. The structure was further analyzed in the Crystallographic Object-Oriented Toolkit (COOT).²⁶ Unbound crystals of SCOT were reproducible and concurrent to co-crystallization trials, soaking trials using pre-formed crystals of SCOT were carried out. Unbound crystals were looped and soaked in a reservoir buffer solution containing PSSI-51 (500 μM–1 mM) for 1 h, 6 h, and 24 h.

Magnetic resonance imaging

Body composition, including lean and fat mass, was assessed through quantitative nuclear magnetic resonance relaxometry using an EchoMRI-4in1/700 body composition analyze.^27,28

In silico modeling

A pharmacophore model for known SCOT inhibitors was meticulously constructed using the Pharmit, enabling the identification of essential molecular features critical for effective inhibition.²⁹ This model was rigorously evaluated and refined based on its alignment with the structural features of the SCOT oxyanion pocket, a process facilitated by the tools available in Schrödinger 's Maestro software suite.³⁰ Subsequently, the refined pharmacophore model guided the generation of candidate inhibitors through the ligand-designer module of Schrödinger. This module employed techniques such as bioisosteric replacement and isostere scanning to enhance the molecular design, thereby optimizing the candidates for improved efficacy and target specificity. For the MD simulations, the human SCOT structure (PDB: 3DLX) was prepared for simulations using the Maestro Schrödinger suite’s Protein Preparation Wizard. This involved adding missing side chains with Epik and optimizing ionizable residues based on pKa predictions using PROPKA. Water molecules located more than 5 Å away from protein residues were removed, and the proteins were subsequently minimized using the OPLS4 force field under restraint. Molecular dynamics (MD) simulations were conducted using Desmond within an orthorhombic box, surrounded by a 10 Å buffer of TIP3P water, to maintain physiological charge neutrality. These simulations ran for 80 ns under the NPT ensemble, maintaining a constant temperature of 300 K and pressure of 1.01325 bar. For docking studies, ligands were prepared using UCSF Chimera version 1.10.2 using the AMBER99SB force field framework.³¹ The structure of DRD2 (PDB: 6CM4) was prepared similarly, after removing the bound ligand risperidone.³² Docking was performed with AutoDock Vina, which enclosed the oxyanion pocket of SCOT and the risperidone binding site of DRD2 in a 30 × 30 × 30 Å grid with 0.375Å spacing.²¹ Rotatable bonds were allowed free rotation, and 20 docking runs were executed for each ligand. The poses with the lowest binding energy were selected, imported into Schrödinger Maestro, and subjected to MD simulation as described for SCOT. The resulting data from these simulations were analyzed using the Schrödinger Maestro suite and Visual Molecular Dynamics (VMD) software,²⁴ with graphical representations generated through Mol∗ (Molstar)²⁰ and ChimeraX.¹⁹

Parallel artificial membrane permeability assay (PAMPA)

PAMPA was conducted using a Bioarray Kit, following the manufacturer’s protocol. Artificial lipid membranes were prepared by applying the provided lipid solution to the filter support of the donor plate. Test compounds (10 μM) were prepared and added to the donor wells. The acceptor plate, containing 5% DMSO in phosphate buffered saline (PBS), was then assembled with the donor plate to create a two-compartment system. The assembled plates were incubated at 25°C for 4 h to allow passive diffusion. Following incubation, samples from both the donor and acceptor compartments were examined with UV spectrophotometry to measure the concentration of the test compound, utilizing standard curves for quantification. The permeability coefficients (P_e) were calculated based on the difference in compound concentration between compartments using the following formula:

$$\log {\mathrm{P}}_{\mathrm{e}}=\log \{\mathrm{C}\times \ln \left(1-\frac{\left[\text{drug}\right]\text{acceptor}}{\left[\text{drug}\right]\text{equilibrium}}\right)]$$

$$\text{Where}:\mathrm{C}=\left(\frac{\text{VD}\times \text{VA}}{\left(\text{VD}+\text{VA}\right)\text{Area}\times \text{time}}\right)\text{.}$$

$$\text{Where}:\left[\text{drug}\right]\text{equilibrium}=\left(\left[\text{drug}\right]\text{donor}\times \text{VD}+\left[\text{drug}\right]\text{acceptor}\times \text{VA}\right)/\left(\text{VD}+\text{VA}\right)\text{.}$$

$$\text{where}:\text{VD}=0.15\text{mL},\text{VA}=0.30\text{mL};\text{Area}=0.25\text{cm}2;\text{time}=4\mathrm{h}\left(14400\mathrm{s}\right)\text{.}$$

Dopamine 2 receptor cAMP mobilization assay

U2OS red cAMPNomad_D2R cells (Innoprot, Spain) were cultured in DMEM/F-12 (Sigma-Aldrich D6421) supplemented with 10% FBS (Sigma-Aldrich D7524) in T75 flasks. Cultures were maintained at 37°C in a humidified atmosphere with 5% CO₂. Cells were subcultured at a ratio of 1:6 when they reached >95% confluence. For the agonism assay, cells were incubated with PSSI-51 (ranging from 100 μM to 1 nM) for 24 h in Opti-MEM (Thermo-Fisher Scientific 31985070). A 30 μM concentration of Dopamine was used as the positive control, while vehicle (DMSO) was used as the negative control. For the antagonism assay, cells were incubated with PSSI-51 or pimozide (ranging from 100 μM to 1 nM) for 24 h in Opti-MEM containing 30 μM Dopamine. A 10 μM concentration of Blenonserin (Sigma-Aldrich B7188) diluted in Opti-MEM containing 30 μM Dopamine was used as the positive control. Opti-MEM containing 30 μM Dopamine (Sigma-Aldrich H8502) was used as the negative control, and vehicle (DMSO) was used as the zero control. Fluorescence intensity was acquired using the Synergie II microplate reader. Changes in the fluorescence intensity of the red cAMPNomad biosensor were detected using 485/20 nm and 528/20 nm filters.

Western blotting

Frozen tissues were pulverized using homogenization buffer (50mM Tris-Base, 1mM EDTA, 10% glycerol, 0.02% Brij) supplemented with protease inhibitors (Sigma Aldrich, 87785) and phosphatase inhibitor cocktail (ThermoFisher, P0044). The protein concentration was quantified with the Bradford method before 30 μg of protein were applied to a 12% SDS-PAGE gel. After electrophoresis, proteins were transferred to a nitrocellulose membrane (Thermofisher) and blocked with 5% skim milk in tris-buffered saline with 0.1% Tween 20 (TBS-T) for 1 h. Overnight incubation at 4°C with primary antibodies was followed by 2 × 5 min washes in TBS-T and a 1 h incubation with the corresponding secondary antibodies the next day. The membrane was then washed three times with TBS-T for 5 min each, before protein bands were detected using the ChemiDoc MP Imaging System (Bio-Rad) with SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermofisher). Primary antibodies targeting SCOT (ProteinTech, 12175-1-AP), BDH1 (Novusbio, NBP1-88673), ACAT1 (Cell Signaling, 44276S), and HSP90 (BD Biosciences, 610418) were used at a dilution of 1/1000 in 5% BSA. Secondary antibodies, anti-rabbit IgG (Cell Signaling, 7074S) and anti-mouse IgG (Cell Signaling, 7076S), were diluted at 1/10,000 in TBS-T.

Distribution of PSSI-51 and pimozide within blood of rats

Blood to plasma ratio: The blood to plasma ratio of PSSI-51 and pimozide were evaluated in vitro using fresh blood (collected in heparinized tubes) from male Sprague-Dawley rats. Blood was spiked with either pimozide or PSSI-51 (n = 3 per drug) to provide a concentration of 100 ng/mL in 1 mL of blood (n = 3). Each drug was allowed to equilibrate in the blood for 1 h at 37°C in a shaking water bath. A volume of 0.2 mL of whole blood was transferred to new tubes. The rest of the blood sample was centrifuged at 3000 xg for 10 min to obtain plasma, then 0.2 mL of plasma was transferred to another test tube. Each of the 0.2 mL specimens of plasma and blood were then assayed for either pimozide or PSSI-51 using LC-MS/MS.

Plasma protein binding: Fresh blood collected from Sprague-Dawley rats was spiked with 1000 and 5000 ng/mL of PSSI-51 or pimozide. The blood was incubated in a water bath (37°C) for 1 h, then centrifuged at 3000 xg for 10 min. The obtained plasma was subjected to rapid equilibrium dialysis using a rapid equilibrium dialysis plate with 8K molecular weight cut-off (Life Technologies, 90006). A volume of 200 μL plasma and 400 μL PBS were added to corresponding chambers in equilibrium dialysis. The unit was covered with sealing tape and incubated at 37°C on the shaker at 300 rpm for 100 min. A 100 μL volume of incubated plasma and PBS were assayed using LC-MS/MS to calculate the unbound fractions. Plasma protein binding was calculated using the following equation:

$$UnboundFraction(\%)=\frac{\text{Concentration}\text{in}\text{filtrate}\left({\mathrm{C}}_{\text{unbound}}\right)}{\text{Concentration}\text{in}\text{plasma}\left({\mathrm{C}}_{\text{total}}\right)}\times 100$$

NADPH-dependent oxidative metabolism

The relative rates of depletion of PSSI-51 and pimozide were examined using rat and human liver microsomes (XenoTech-BIOIVT, M000100 and MX00801, respectively). A concentration of 100 ng/mL was added to the tubes containing 1 mg/mL of microsomal protein and incubation buffer.^33,34 Once temperature equilibration was completed in a shaking water bath at 37°C, reactions were initiated by adding 1 mM NADPH (Sigma Aldrich, 481973) to each tube. The tubes were shaken and reactions stopped at 10 and 30 min afterward. Each sample was assayed using LC-MS/MS and the following equation was used to estimate the intrinsic clearance (CLint) by utilizing the half-life (t_1⁄2) and the concentration of microsomal protein (M)^33,34:

$$CLint=\frac{0.693}{T_{1/2}}\times M$$

Each determination was performed in triplicate.

Pharmacokinetic study in rats

Sprague-Dawley rats (300–350 g) were administered 10 mg/kg of PSSI-51 by oral gavage. PSSI-51 was suspended in 1% methylcellulose and sonicated for 75 min just before dosing. Serial blood samples were collected for up to 24 h from the surgically implanted jugular-veins. Samples were stored at −80°C until the day of analysis by reverse phase LC-MS/MS. The assay involved extraction from blood components using ethyl acetate and yielded highly linear (r² > 0.99) calibration curved in blood and a coefficient of variation and mean error of less than 11% at concentrations ranging from 0.1 to 40 ng/mL based on 0.1 mL of rat whole blood.

Ultrasound echocardiography

Mice were initially anesthetized with 2–3% isoflurane and maintained on 1–1.5% isoflurane for the remainder of the evaluation. Heart rate, respiration, and body temperature were continuously recorded throughout the procedure. Transthoracic echocardiography using a VisualSonics Vevo 3100 rodent ultrasound machine with an MX 550S probe was performed in mice, with ultrasound sonograms recorded and interpreted using VisualSonics Vevo Lab version 5.7.1.^27,28 Parameters of systolic function (left ventricular ejection fraction, left ventricular fractional shortening) were evaluated using the following formulas;

$$LVVolume\left(Vol\right)\left(\mu L\right)=\left(\frac{7.0}{2.4+LVID}\right)\times LVID$$

$$LVEF(\%)=100\times \left(\frac{LVVol;d-LVVol;s}{LVVol;d}\right)$$

$$LVFS(\%)=100\times \left(\frac{LVID;d-LVID;s}{LVID;d}\right)$$

where: EF = Ejection fraction, FS = Fractional Shortening, LV = Left ventricular, LVID = LV Internal Diameter.

Diastolic function (E/A ratio, e′/a′ ratio, E/e′ ratio) was assessed using pulsed-wave and tissue Doppler imaging of the mitral inflow and mitral annular velocities, respectively, during early diastolic filling (pulsed-wave Doppler denoted as E; tissue Doppler denoted e′) and late diastolic filling due to atrial contraction (pulsed-wave Doppler denoted as A; tissue Doppler denoted a′).

Dynamic PET imaging

Male C57BL/6J mice fed ad libitum or fasted for 24 h were used for the PET imaging experiments, which utilized a radiolabeled ketone body S-[¹⁸F]FβOHB and were performed on an INVEON PET/CT scanner (Siemens Preclinical Solutions, Knoxville, TN, U.S.A). Radiosynthesis of S-[¹⁸F]FβOHB was performed as previously described.¹² In brief, the dried residue of [18F]KF/K222, (2S)-(+)-glycidyl tosylate (0.01 g (0.044 mmol)) in CH3CN (0.4 mL) was reacted at 95°C for 25 min. Epoxide opening was then completed with an aqueous solution of KCN (50 μL, 3.0 M) at 95°C for 15 min. The nitrile intermediate was then purified using HPLC on a Phenomenex LUNA C18 column (100 Å, 250 × 10 mm, 10 mm). HPLC purified nitrile intermediate in water was then concentrated using rotary evaporation and then combined with E. coli-derived nitrilase enzyme (1 mg) in TRIS buffer (10–20 μL of 10 mM, pH 9). The enzymatic reaction was completed at 30°C, 750 rpm for 60 min. S-[¹⁸F]FβOHB was additionally purified by HPLC and reformulated into 0.1 M NaOAc, pH 5.4 buffer for animal injection. Prior to radiotracer injection, mice were anesthetized through inhalation of isoflurane in 40% oxygen/60% nitrogen (gas flow 1 L/min), and body temperature was kept constant at 37°C. Mice were placed in prone position into the center field of view of the scanner. A transmission scan for attenuation correction was not acquired. After the emission scan was started, mice were injected with 3–8 MBq of S-[¹⁸F]FβOHB in ∼100–150 μL saline solution injected with a delay of ∼15 s through a previously placed tail vein catheter. Radioactivity present in the injection solution (0.5 mL syringe) was determined using a dose calibrator (AtomlabTM 500; Biodex Medical Systems, New York, NY, U.S.A.). Data acquisition was performed over 60 min in 3D list mode. The dynamic list mode data were sorted into sinograms with 54-time frames (10 × 2, 8 × 5, 6 × 10, 6 × 20, 8 × 60, 10 × 120, 6 × 300 s). The frames were reconstructed using maximum a posteriori (MAP) as reconstruction mode. No correction for partial volume effects was applied. The image files were further processed using the ROVER v2.0.51 software (ABX GmbH, Radeberg, Germany). Masks defining 3D regions of interest (ROIs) were set, and the ROIs were defined by 50% thresholding (for analysis of the heart at 80%). Mean standardized uptake values [SUVmean = (activity/mL tissue)/(injected activity/body weight), μL/g] were calculated for each ROI.

Behavioral analysis

The voluntary activity of the animals was recorded using the Spontaneous Activity Wheel system (BIOSEB). Mice were acclimatized for 24 h before receiving a single dose of PSSI-51 or pimozide. Subsequently, their spontaneous activity was tracked using the BIOSEB software. Four parameters were monitored and analyzed over a 48 h period: distance traveled, mean speed, maximum acceleration, access count, and active time. The cage system resembled the animals' original housing and environmental enrichments including chip bedding, a plastic hideaway, 8 ounces of crinkle paper, and Nestlet square bedding made from short fiber cotton were maintained.

Whole body in vivo metabolic assessment

in vivo metabolic assessment was conducted using indirect calorimetry with the Oxymax CLAMS system (Columbus Instruments) and analyzed using CalR.²³ Animals were acclimatized in the system for 24 h, followed by a 24 h data collection period.²⁵

Chromatographic separation and quantification

Liquid-liquid extraction was used to extract PSSI-51 and pimozide from mice brain tissue. A volume of 300 μL double-distilled water was added to a small portion of brain tissue (100 mg). Then, the IS was added to the mixture. That was followed with homogenization of the tissue with IS (500 ng/mL). The homogenized mixture was transferred to clean test tubes using Pasteur pipettes. A volume of 3 mL hexane was added to the mixture followed with vortexing 30 s and centrifuged at 3000 g for 3 min. The supernatant layer was transferred to small test tubes and evaporated in vacuo. The residues were reconstituted using 150 μL of mobile phase and 75 μL being injected into the HPLC.

Stock solutions of PSSI-51 were prepared at 10 mM, and working standards were created by serial dilution in acetonitrile to concentrations ranging from 0.475 to 0.000475 ng/mL. Pimozide stock solutions were made by dissolving 2.7 mg in 9 mL of acetonitrile for a final concentration of 0.3 mg/mL. These were stored at −20°C. Standard curves were constructed from brain samples spiked with various concentrations of the analytes. After brain tissue homogenization, liquid-liquid extraction was employed using hexane after the addition of an internal standard. The organic layer was then separated, dried in vacuo, and reconstituted in the mobile phase for chromatographic analysis. Pimozide served as the internal standard for PSSI-51 and vice versa. After extraction with hexane, the supernatant was dried and reconstituted for HPLC injection. The chromatographic system comprised a Waters 600E multi-solvent delivery system pump, Waters 717 autosampler with a variable injection valve, and a Waters 486 UV–visible tunable absorbance detector. Data collection and processing were facilitated by EZStart software on a Windows-based computer. Chromatographic separation of PSSI-51 and pimozide was achieved using an Alltima C18-column (150 × 4.6mm i.d., 5 μm particle size) from Alltech, complemented by a Grace Alltech All-Guard Guard Cartridge pre-column. The isocratic mobile phase consisted of acetonitrile and potassium dihydrogen phosphate buffer in a 40:60 v/v ratio, delivered at 1 mL/min at room temperature. The mobile phase was freshly prepared, degassed through a 0.45 μm nylon filter daily, and detection was performed at a UV wavelength of 210 nm to maximize absorption readings of PSSI-51 and pimozide, with the analytical run time being under 19 min.

Quantification and statistical analysis

Values are reported as mean ± the standard error of the mean (SEM). Group comparisons were conducted using a one-way or two-way analysis of variance (ANOVA) followed by Bonferroni post hoc analysis, or Student’s t-test when suitable for the experimental design and data distribution. A p-value of less than 0.05 was deemed indicative of statistical significance. Statistical details of the experiments, including the type of statistical test applied, the precise sample size (n) and its meaning, are provided in the respective figure legends. No data points were removed as outliers either through testing or arbitrary exclusion. All statistical analyses were performed using GraphPad Prism version 10.

Published: April 3, 2025

Supplemental information

Data S1. Detailed interaction and conformational dynamics of Pimozide in complex with SCOT, related to Figure 1

(A) Graphical representation of the SCOT-pimozide interaction fraction throughout the simulation, with normalized bar graphs detailing the trajectory. (B) Overview of the specific protein-ligand contacts between pimozide and SCOT during the molecular dynamics (MD) simulation, illustrating the interaction points and their persistence. (C) Comparison of the torsional dynamics of unbound pimozide (displayed in gray) and pimozide when bound to SCOT (color-coded), highlighting the conformational changes and restrictions imposed by enzyme binding. (D) Evaluation of key biophysical properties of PSSI-51 in complex with SCOT, examining its stability, flexibility, and dynamic behavior during the simulation. RMSD; root-mean-square deviation, rGyr; radius of gyration, MolSa; molecular surface area, SASA; solvent-accessible surface area

Data S2. Cardiac function assessment following isolated heart perfusion with [U-14C] βOHB and PSSI-51, related to Figure 2

(A) heart rate, (B), peak systolic pressure, (C), developed pressure, (D), heart rate x peak systolic pressure, (E), heart rate x developed pressure, (F), cardiac output, (G), aortic outflow, (H), coronary flow, and (I), cardiac work in isolated working mouse hearts perfused with 10 nM PSSI-51 (n = 7 animals). Data are presented as mean ± SEM. Statistical analyses were conducted using Student’s t-test

Data S3. Determination of pimozide and PSSI-51 concentration in the brain, related to Figure 4

HPLC-UV chromatograms of (A) a blank mouse brain, and mouse brain samples following (B) PSSI-51, and (C) pimozide administration

Data S4. Cardiac function assessment following PSSI-51 treatment, related to Discussion

Cardiac function in lean and T2D mice before/after vehicle or PSSI-51 treatment (n = 10–11 animals)

Data S5. NMR spectra, related to STAR Methods

¹H (A) and ¹³C (B) spectra of PSSI-51