Development of Novel Tools for Dissection of Central Versus Peripheral Dopamine D₂-Like Receptor Signaling in Dysglycemia

Dopamine (DA) is increasingly recognized as an important modulator of metabolism (1). Most studies examining dopaminergic modulation of metabolism have focused on DA D₂-like receptors (D₂, D₃, and D₄ receptors) in brain regions associated with metabolic regulation, including striatum and hypothalamus (2). Striatal and hypothalamic D₂ (D2R) and D₃ (D3R) receptors mediate appetite, feeding, and central glucose sensing (3,4). Moreover, D2R polymorphisms are associated with insulin resistance and type 2 diabetes (T2D) (5). Nevertheless, the precise mechanisms and sites of action for the metabolic effects of DA remain unclear.

Discovery of DA D₂-like receptors outside of the central nervous system (CNS) has expanded the scope of DA’s roles as a metabolic modulator (1,6). In the pancreas, D2R and D3R are expressed in glucagon-secreting α-cells and insulin-secreting β-cells (1,6,7). α-cells and β-cells produce DA, which acts locally through islet D2R and D3R to negatively modulate hormone secretion (1). Importantly, interfering with peripheral D2R/D3R signaling contributes to dysglycemia. We discovered that antipsychotic drugs (APDs) disrupt α-cell and β-cell D2R/D3R signaling, significantly elevating insulin and glucagon secretion (7). Over time, this may desensitize insulin-sensitive peripheral targets and stimulate glucose production to produce dysglycemia (1). Conversely, D2R/D3R agonists, such as bromocriptine, a novel T2D drug, act on peripheral targets, including islets, to improve insulin sensitivity (8). Although these drugs can target other receptors (e.g., serotonin and adrenergic receptors) (8–10), D2R/D3R agonism appears to be the principal mechanism by which they ameliorate dysglycemia (11). These reports suggest an important role for peripheral D2R/D3R signaling in metabolic regulation and dysglycemia, particularly within the pancreas.

Disentangling the respective metabolic contributions of CNS versus peripheral DA D₂-like receptor signaling has been difficult. Initial work to address the metabolic roles of DA D₂-like receptors used transgenic global D2R- or D3R-knockout (KO) mouse models (2,12,13). However, these KO mice exhibited complex metabolic phenotypes due to concurrent effects of receptor deletion on CNS and peripheral targets (12–14). More recently, constitutive β-cell–specific D2R-KO mice demonstrated inappropriately elevated serum insulin levels in response to meals, suggesting a role for β-cell D2R signaling in insulin release (15). Nevertheless, genetic constitutive KO strategies may still lead to compensatory effects at systemic, cellular, and/or transcriptional levels. Drugs that selectively act at CNS or peripheral targets offer an alternative approach, possessing key advantages because pharmacological manipulations are 1) reversible and 2) offer the ability to tune changes to receptor signaling by controlling drug dose and duration.

Here, we describe the first generation of new pharmacological tools intended to selectively target peripheral DA D₂-like receptors. Our aim was to use peripherally limited drugs to examine the metabolic relevance of peripheral DA D₂-like receptor signaling in dysglycemia and its treatment without the confounds of CNS actions. We used quaternary methiodide (MeI) conjugation as a strategy to diminish blood-brain barrier (BBB) permeability, selecting bromocriptine methiodide (BrMeI) as our lead compound. Systemic administration of bromocriptine, which has access to both CNS and peripheral targets, significantly improved insulin sensitivity and glucose tolerance. In contrast, selectively limiting D2R/D3R agonism to the CNS (via intracerebroventricular bromocriptine administration) or to the periphery (via BrMeI) abolished metabolic improvements. Our results suggest the importance of coordinated, tandem signaling by both peripheral and CNS DA D₂-like receptors for glycemic control, offering a new approach for studying metabolism and potentially treating dysglycemia.

All chemicals and solvents were purchased from chemical suppliers, unless otherwise stated, and used without further purification.

A solution of the free base form of each parent compound in acetone or ethanol was added with excess of methyl iodide (3 eq). After 24 h stirring at room temperature in the dark, solid products were filtered and washed with diethyl ether to yield pure quaternary ammonium salts. Descriptions for the syntheses of respective MeI analogs are provided in the online Supplementary Material, including compound purity (Supplementary Table 1).

Radioligand binding studies were conducted in HEK293 cells stably expressing human D2R, D3R, or D4R. Radioligand [³H]-(R)-(+)-7-OH-DPAT was used in all studies except for L741,626, and MPE01-06, which used [³H]-N-methylspiperone. Detailed descriptions of the binding studies are provided in the online Supplementary Material. Receptor binding data for (−)-quinpirole and sumanirole were previously reported (16).

DA D₂-like receptor–mediated mitogenesis functional assays were conducted in Chinese hamster ovary cells expressing human D2R or D3R as reported earlier (17). D2R- and D3R-expressing cell lines were seeded into 96-well plates at 5,000 cells/well. After 48–72 h, cells were incubated in serum-free α-minimum essential medium (MEM) for 24 h. On test day, serial dilutions of test compounds were made in serum-free α-MEM and incubated with cells. Following ∼24 h, 0.25 μCi [³H]thymidine in α-MEM supplemented with 10% FBS was added to each well for 2 h. Cells were trypsinized, plates filtered, and intracellular radioactivity was measured via scintillation spectrometry. (-)-Quinipirole was run each day as an internal control. Specific [³H]thymidine incorporation at each drug dose was expressed as the percentage of maximal quinpirole effect for each of the tested drugs. The half-maximal effective concentration (EC₅₀) values and maximum response (E_max) values were determined from dose-response curves via nonlinear regression analyses using GraphPad Prism (GraphPad Software, San Diego, CA).

HEK-D4.4-AC1 cells expressing human DA D4.4 receptors and adenylate cyclase type I were used. Experiments were conducted with a cAMP enzyme immunoassay (EIA) kit (Cayman, Ann Arbor, MI), as previously described (18). Basal cAMP was subtracted from all cAMP values. Maximal DA D4.4 receptor–mediated inhibition of forskolin-stimulated cAMP formation by the agonist drugs was defined with 1 µmol/L (−)-quinpirole.

INS-1E cells (gift of Dr. Pierre Maechler, Université de Geneve) were cultured as described earlier (15).

Glucose-stimulated insulin secretion (GSIS) studies in INS-1E cells were conducted as previously described (19). Briefly, INS-1E cells were seeded into a 24-well plate at a seeding density of 5.0 × 10⁵ cells/well. Culture media was exchanged 24 h later, and experiments were conducted the next day. Cells were glucose-starved (1 h at 37°C) in Krebs-Ringer buffer and subsequently stimulated with 20 mmol/L glucose in the presence of drugs (90 min at 37°C). Supernatants were collected for insulin detection.

Secreted insulin was detected using an insulin detection kit (PerkinElmer/Cisbio Bioassays, Bedford, MA) based on homogenous time-resolved fluorescence resonance energy transfer, as previously described (19). Results were read using a PheraStar FS plate reader (BMG Labtech, Ortenberg, Germany). Insulin concentrations were derived via extrapolation of ratiometric fluorescence readings (665 nm/620 nm) to a second-order quadratic polynomial curve. Dose-response curves were fit via nonlinear regression of Log[drug] versus normalized percentage maximum insulin secretion using GraphPad software. IC₅₀ and E_max values were calculated from nonlinear regression analyses.

Nanoluciferase bioluminescence resonance energy transfer (NanoBRET) assays were conducted in HEK-293T cells as described previously (8). Experiments used human D2R tagged with a HaloTag. Effector recruitment studies used human Gα_i1 or β-arrestin2 fused to nanoluciferase (NanoLuc). Detailed descriptions of the constructs and assays are provided in the Supplementary Material.

BrMeI was screened for binding to DA, serotonin, and opioid receptors as well as biogenic amine receptors. Assays for specific binding to D1R, D4R, 5-hydroxytryptamine (5-HT)_1A, 5-HT_2A, and 5-HT_2C receptors were performed as previously described (17). Detailed descriptions of the assays are provided in the Supplementary Material.

A screen to identify off-target BrMeI binding was performed commercially (Eurofins Cerep Panlabs, Celle-Lévescault, France) as described previously (20). Reference compounds were tested concurrently with BrMeI to assess assay reliability. Binding was calculated as the percentage of specific binding inhibition for each target.

BrMeI was commercially evaluated for activity at the hERG K⁺ channel for potential cardiac toxicity (Eurofins Panlabs, Redmond, WA) similar to previous studies (21) and described in detail in the Supplementary Material.

Phase I metabolic stability assays were performed in mouse liver microsomes as previously described (22). Detailed descriptions of the assays are provided in the Supplementary Material.

Animals were housed in accordance with National Institutes of Health and Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines. Rodents were housed in cages with a 12:12 light:dark cycle and had access to food and water ad libitum at all times, unless indicated otherwise.

All animal work was approved through the Institutional Animal Care and Use Committees at the University of Pittsburgh, Albert Einstein College of Medicine, The Johns Hopkins University, and the Centre for Addiction and Mental Health.

Only male rodents were used to avoid the impact of estrus on hormone secretion and islet physiology. Additionally, our choice was based on work showing that male rodent β-cells possess transcriptomic signatures more similar to T2D versus females (23,24).

All pharmacokinetics assays were conducted in 6- to 8-week-old male CD-1 mice (Harlan, Indianapolis, IN) weighing 20–25 g as described earlier (25–27). BrMeI and unmodified bromocriptine were dosed at 10 mg/kg intravenously (i.v.). Detailed assay descriptions are provided in the Supplementary Material.

Wild-type C57BL/6J mice (3–4 months old) used in the glucose and insulin tolerance tests were fed a Western diet (5TJN: 40% fat, 44% carbohydrate, 16% protein; TestDiet) for 12 weeks.

Bromocriptine and BrMeI were dissolved in a 50% DMSO/50% distilled water (dH₂O) solution to a final concentration of 2 mg/mL and administered to mice intraperitoneally (i.p.). Vehicle was defined as 50% DMSO/50% dH₂O only. For glucose tolerance testing (GTT) and insulin tolerance testing (ITT) in mice, animals were administered the respective drugs or vehicle for 12 weeks via daily i.p. administration. In a subset of GTT studies, mice were administered bromocriptine or vehicle daily via an intracerebroventricular route for 8 weeks prior to testing. At the 8-week time point, bromocriptine GTT data were compared between intracerebroventricular versus systemic bromocriptine administration. As a control, 8 weeks of systemic bromocriptine was sufficient to improve glucose tolerance in dysglycemic mice fed a Western diet (data not shown). All treatments were conducted at the same time of day.

Intracerebroventricular drug administration in adult wild-type C57BL/6J mice was conducted as described earlier (28). Mice were anesthetized with ketamine/xylazine for stereotaxic implantation of custom indwelling 3.5-mm guide cannulas (Plastics One, Roanoke, VA) into the third cerebral ventricle (from bregma: anteroposterior, −0.3 mm; dorsoventral, −3 mm). Mice recovered for at least 1 week before intracerebroventricular drug administration.

Oral GTT (OGTT) was conducted as previously described (28). Wild-type C57BL/6J mice fed a Western diet underwent a 6-h daytime fast, followed by a glucose challenge with a glucose load of 2 mg/kg (10% glucose stock). Glucose measurements were made in duplicate from tail vein blood after glucose challenge using Contour glucometer sticks (Bayer Corp., Whippany, NJ).

Wild-type C57BL/6J mice fed a Western diet underwent a 6-h daytime fast, which was followed by i.p. administration of 1.5 units/kg human insulin. Glucose values were determined from blood collected following insulin administration via tail bleed. Blood glucose was measured in duplicate using Contour glucometer sticks. Insulin was measured via quantitative ELISA (Linco Research, St. Charles, MO).

Sprague-Dawley rats underwent intracerebroventricular operations as previously described (3,29). A cannula was stereotaxically placed into the third cerebral ventricle (from bregma: anteroposterior, −2.5 mm; mediolateral, 0.0 mm; dorsoventral, −8.0 mm) under isoflurane anesthesia. Following ∼1-week recovery, rats underwent blood vessel cannulation as described earlier (3). After recovering for ∼4 days, rats were food-restricted overnight and infused (5 µL/h) with 10 mmol/L l-3,4-dihydroxyphenylalanine (l-DOPA) or vehicle (saline) intracerebroventricularly for 3.5 h. Changes in glycemic control were determined via periodic measurement of plasma glucose with a glucose analyzer (Analox Instruments, Stourbridge, U.K.). Plasma insulin was measured via ELISA (Mercodia, Winston-Salem, NC).

We used multifactor ANOVA to compare between-group differences. This included repeated measures two-way ANOVAs, followed by Tukey multiple comparison tests for post hoc comparisons. Statistical analyses were performed with GraphPad Prism. Significance was accepted at P < 0.05.

All data and resources will be made available by the corresponding author upon request.

We aimed to establish a family of pharmacological tools to selectively manipulate DA D₂-like receptor signaling in the periphery while limiting their actions in the CNS. Based on earlier MeI quaternization to produce peripherally-limited drugs (30,31), we hypothesized that addition of a quaternary MeI group would similarly render D₂-like receptor agonists less BBB permeable. We synthesized quaternary MeI salts of full DA D₂-like receptor agonists that broadly target both D2R and D3R including (−)-quinpirole (Fig. 1A) and bromocriptine (Fig. 1B), producing MeI analogs AB01-59 and AB01-60, respectively. Focusing on D2R, we quaternized sumanirole, a D2R-selective agonist, to generate MPE01-05 (Fig. 1C). We also synthesized MeI analogs of D3R-preferential agonists (±)-7-OH-DPAT (AB01-62) (Fig. 1D) and (+)-PD128,907 (AB01-117) (Fig. 1E). To target peripheral D4R, we generated a quaternary MeI analog of CAB03-015, a D4R-selective agonist (32) (AB01-102) (Fig. 1F). Additionally, we synthesized an MeI analog of DA D₂-like receptor partial agonist and APD, aripiprazole (AB01-61) (Fig. 1G). Finally, we generated an MeI analog of a DA D₂-like receptor antagonist by quaternizing L741,226 (MPE01-06) (Fig. 1H). MeI-modified compounds were >95% pure based on nuclear magnetic resonance spectroscopy, elemental analysis, high-resolution mass spectroscopy (HRMS, which agrees within ±5 ppm), and tandem MS (MS/MS) (ESI in positive mode) (Supplementary Table 1).

We ascertained whether MeI quaternization altered the D₂-like receptor binding affinities of the MeI analogs by measuring inhibitory constants (K_i). K_i values were higher for all MeI analogs at human D2R and D3R compared with unmodified compounds (Table 1), albeit to differing degrees. The MeI analog of bromocriptine, AB01-60, continued to bind to D2R and D3R with relatively high affinity. In contrast, the MeI analog of (±)-7-OH-DPAT, AB01-62, showed 179-fold and 365-fold decreased binding affinity at D2R and D3R versus unmodified (±)-7-OH-DPAT. MeI modification also altered the relative balance of D2R versus D3R binding affinities for some analogs. AB01-59 showed preferential binding to D2R over D3R compared with unmodified (−)-quinpirole. In contrast, AB01-102, AB01-61, and MPE01-06 showed increased D3R affinity versus D2R (Table 1), suggesting that local changes in polarity following MeI addition may impact D2R/D3R receptor-binding properties. Finally, we measured the binding affinities of D4R-selective agonist CAB03-015 and its MeI analog AB01-102 at human D4R, finding a 61.7-fold loss of binding affinity for AB01-102 versus its unmodified parent (Supplementary Table 2).

We evaluated the activity of our MeI analogs via the mitogenesis functional assay by measuring D2R- and D3R-mediated increases in [³H]thymidine incorporation (17) (Fig. 2). D2R/D3R agonists (−)-quinpirole and bromocriptine and their MeI analogs produced dose-dependent increases in [³H]thymidine incorporation (Fig. 2A–D). However, while the efficacies of AB01-59 and AB01-60 were largely unchanged compared with their unmodified parents, both MeI analogs displayed diminished potency (Supplementary Table 3). MPE01-05 did not markedly differ from sumanirole in efficacy or potency in D2R- and D3R-mediated mitogenesis (Fig. 2E and F). In contrast, AB01-62 and AB01-117, MeI analogs of preferential D3R agonists, demonstrated substantial potency loss in D3R-mediated mitogenesis despite retaining their efficacies (Figs. 2G and H and Supplementary Table 3). Consistent with aripiprazole’s partial agonism of D2R and D3R, MeI analog AB01-61 also displayed partial agonist properties in D2R- and D3R-mediated mitogenesis with nearly identical efficacies. However, AB01-61 showed reduced potencies compared with aripiprazole, suggesting that this analog has little functional activity at D2R or D3R (Fig. 2I and J and Supplementary Table 3). As expected, MPE01-06, the MeI analog of D2R/D3R antagonist L741,626, did not impact mitogenesis [<1% of the maximum (−)-quinpirole response] (Fig. 2K and L and Supplementary Table 3). Lastly, since CAB03-015 is a D4R-selective agonist, we tested the drug and its MeI analog AB01-102 in a D4R-mediated adenylate cyclase activity/cAMP formation assay. Both drugs behaved as partial agonists, exhibiting similar efficacies in D4R-driven decreases in forskolin-stimulated adenylate activity and cAMP formation. However, AB01-102 showed 732-fold diminished potency compared with unmodified CAB-03-015 (Supplementary Table 4).

Because β-cell DA D₂-like receptors are negative modulators of GSIS (1), we tested the ability of several MeI analogs to signal through these receptors by examining their impacts on GSIS in INS-1E cells, a rat pancreatic β-cell line (33). As expected, (−)-quinpirole and bromocriptine dose-dependently reduced GSIS [(−)-quinpirole IC₅₀ = 1.18 μmol/L; bromocriptine IC₅₀ = 0.18 μmol/L] (Fig. 3A and B and Table 2). AB01-59 similarly decreased GSIS, but with reduced efficacy (48.4% decrease) and potency (IC₅₀ = 6.94 μmol/L) versus (−)-quinpirole (Fig. 3A and Table 2). By comparison, AB01-60 functioned as a full agonist, retaining its efficacy, albeit with decreased potency (IC₅₀ = 10.17 μmol/L, 57.7-fold decrease) compared with bromocriptine (Fig. 3B and Table 2). Interestingly, the D2R-preferential agonist sumanirole or D3R-preferential agonist PD128,907 and their MeI analogs exhibited markedly less efficacy and potency versus nonselective D2R/D3R agonists (−)-quinpirole or bromocriptine (Fig. 3C and D and Table 2). This is consistent with our findings that D2R and D3R work in combination to effectively modulate GSIS (15). Overall, given AB01-60’s preservation of D2R/D3R binding affinity and efficacy in both GSIS and mitogenesis assays, we chose this drug (henceforth termed bromocriptine MeI or BrMeI) as our lead compound.

We determined BrMeI’s ability to initiate D2R-mediated signaling by examining recruitment of intracellular effectors such as G protein Gα_i1 and β-arrestin2, given their roles in the regulation of hormone secretion (7,34). We used nanoBRET technology to determine whether BrMeI-stimulated recruitment of Gα_i1 and β-arrestin2 to D2R differed from bromocriptine or DA. Bromocriptine (EC₅₀ = 15.07 nmol/L) was more potent than DA (EC₅₀ = 288.67 nmol/L) at eliciting dose-dependent Gα_i1 recruitment to D2R, albeit with 47% less efficacy. This suggested that bromocriptine functioned as a partial agonist in stimulating Gα_i1 recruitment to D2R. BrMeI’s efficacy resembled unmodified bromocriptine, but with diminished potency (EC₅₀ = 1514.71 nmol/L) (Fig. 4A and Supplementary Table 5). Importantly, we discovered that BrMeI more potently elicited β-arrestin2 recruitment to D2R (EC₅₀ = 8.37 nmol/L) compared with bromocriptine (EC₅₀ = 20.13 nmol/L) and DA (EC₅₀ = 27.63 μmol/L), while retaining bromocriptine’s reduced efficacy (Fig. 4B and Supplementary Table 5). These data suggest that BrMeI is an agonist of D2R that preferentially recruits β-arrestin2 over G proteins to D2R.

We characterized BrMeI’s binding specificity to several families of G protein–coupled receptors (e.g., DA, serotonin, opioid receptors) and biogenic amine transporters in an initial screen using radioligand binding competition assays. Radioligand binding was successfully validated using established standards (Supplementary Tables 6–9). At 100 nmol/L, BrMeI did not exhibit significant hits (Supplementary Fig. 1). However, at the higher 10 µmol/L concentration, BrMeI demonstrated strong D2R and D3R binding. In contrast, BrMeI did not significantly bind to D1R or D4R (Supplementary Fig. 1). We additionally observed BrMeI binding to serotonin 5-HT_1A, 5-HT_2A, and 5-HT_2C receptors as well as to δ-, κ-, and µ-opioid receptors at the 10 µmol/L concentration, in line with bromocriptine’s ability to similarly bind non-dopaminergic receptors, including serotonergic and opioid receptors (8–10). We found no biogenic amine transporter binding at either tested BrMeI concentrations (Supplementary Fig. 1).

We detected potential off-target binding by BrMeI across a broad range of human receptors, channels, and enzymes. As above, we found no significant hits at 100 nmol/L BrMeI (Supplementary Figs 2 and 3). At 10 µmol/L, we identified BrMeI binding to α₂-adrenergic receptors, consistent with work showing bromocriptine’s ability to also signal via α₂-adrenergic receptors (8,10,35). We also found positive hits at A_2A, α₁, α₂, D_4.4R, κ-opioid, µ-opioid, M, M₂, NK₁, and NK₂ receptors, l-type Ca²⁺ (diltiazem) and Na⁺ (site 2) channels, as well as acetylcholinesterase (Supplementary Figs 2 and 3). Finally, we tested BrMeI for hERG channel activity, finding concentration-dependent hERG tail current inhibition (IC₅₀ = 8.65 µmol/L) (Supplementary Fig. 4), similar to bromocriptine (36).

We compared BrMeI’s phase I metabolic stability to bromocriptine using mouse liver microsomes fortified with NADPH. Incubation of BrMeI and bromocriptine in the presence of NADPH led to the complete disappearance of bromocriptine within 30 min (Fig. 5A). Similarly, we found almost complete BrMeI loss within 30 min and total loss by 60 min (Fig. 5B). These results indicate that BrMeI undergoes significant hepatic metabolism in a manner identical to bromocriptine.

We evaluated the pharmacokinetics of BrMeI versus unmodified bromocriptine in vivo in wild-type CD-1 mice via liquid chromatography–MS/MS (LC/MS/MS) measurements of plasma and brain drug levels (Fig. 5C–E, Supplementary Fig. 5, and Supplementary Table 10). At 15 min after administration, BrMeI’s brain-to-plasma ratio was 2.5-fold lower compared with bromocriptine. This suggested that, acutely, BrMeI had lower brain penetrance and was preferentially localized to the periphery (Fig. 5C and D). However, by 60 min, brain-to-plasma ratios of BrMeI and bromocriptine were similar, suggesting that BrMeI was not completely restricted from the CNS (Fig. 5E). Rather, quaternary MeI modification significantly slowed BBB penetration of the drug. Within 60 min, most unmodified bromocriptine was cleared from brain and plasma (Supplementary Table 10). In contrast, significantly higher plasma and brain levels of BrMeI still remained, with plasma BrMeI levels approximately eightfold greater compared with bromocriptine (Fig. 5). These data indicated that BrMeI possessed an in vivo pharmacokinetic profile distinct from bromocriptine which enabled increased duration of action.

We examined the impacts of BrMeI versus bromocriptine on glucose homeostasis in vivo in a diet-induced obesity model where wild-type C57BL/6J mice were chronically maintained on a Western diet to induce dysglycemia and obesity. As expected, vehicle-treated mice exhibited fasting hyperglycemia and impaired glucose tolerance during OGTT. Systemic bromocriptine administration significantly improved glucose tolerance (F_2,29 = 13.10, P < 0.0001) and showed an interaction of drug × time (F_8,116 = 4.178, P = 0.0002). In contrast, systemic administration of BrMeI did not significantly modify glucose tolerance versus vehicle (P > 0.05) (Fig. 6A). Systemic bromocriptine also lowered fasting blood glucose (F_2,29 = 6.994, P = 0.0033) compared with vehicle (P = 0.0163) or BrMeI (P = 0.0035); BrMeI showed no significant effect (P > 0.05) (Fig. 6B). ITT similarly revealed a significant effect of drug on blood glucose (F_2,21 = 6.761, P = 0.0054) driven by bromocriptine-induced improvements in insulin sensitivity (Fig. 6C). Notably, BrMeI significantly reduced blood glucose versus vehicle at the 60 min time point (P = 0.022) (Fig. 6C), suggesting the drug had partial efficacy in improving insulin sensitivity.

We also measured the impacts of systemic bromocriptine and BrMeI treatment on blood insulin during the initial 15 min of the ITT time course. There was a significant drug × time interaction (F_4,42 = 10.57, P < 0.0001) on blood insulin driven by bromocriptine’s ability to enhance circulating plasma insulin levels within 5 min of insulin administration versus vehicle (P = 0.0002) or BrMeI (P = 0.0003); BrMeI treatment did not significantly alter plasma insulin (P > 0.05) (Fig. 6D). Interestingly, neither systemic bromocriptine nor BrMeI treatments had significant effects on weight at 12 weeks (P > 0.05) (Fig. 6E), suggesting that drug-induced effects on dysglycemia could be dissociated from effects on weight.

We asked whether the mechanisms responsible for systemic bromocriptine’s improvements in dysglycemia were mainly via actions on targets in the CNS versus those in the periphery or through concurrent actions at both sites. To test whether bromocriptine’s effects are primarily due to CNS actions, we limited its actions to the brain via intracerebroventricular administration in mice fed a Western diet. Unlike systemic administration, intracerebroventricular bromocriptine had no significant effect on glucose tolerance (P > 0.05) (Fig. 6F). We validated these results by intracerebroventricular infusion of DA precursor l-DOPA into brains of Sprague-Dawley rats for an extended 3.5-h period to permit adequate time for conversion into DA. Similar to our intracerebroventricular bromocriptine data, l-DOPA infusion did not significantly impact plasma glucose or insulin levels (P > 0.05) (Supplementary Fig. 6). Overall, these data suggest that coordinated actions on targets in the CNS and periphery are required to effectively modify glycemic control.

There is emerging evidence that DA D₂-like receptors in the CNS play important roles in glycemic control and treatment of dysglycemia. However, CNS dopaminergic signaling is not sufficient to fully explain the mechanisms by which DA and D₂-like receptors modulate metabolism (37). Altered glucose homeostasis in response to D2R/D3R blockade occurs even in the absence of increased food intake or psychiatric disease (38). Recent discoveries showing local DA signaling in pancreatic islets suggest that dopaminergic regulation of metabolism also relies on peripheral targets (1,6). However, it has been challenging to unravel central versus peripheral contributions of D2R/D3R signaling in metabolic regulation due to the paucity of pharmacological tools targeting these respective compartments. Therefore, our goal was to generate a peripherally-limited D2R/D3R agonist to selectively examine metabolic roles of D2R/D3R signaling within the periphery.

We used MeI quaternization to render DA D₂-like receptor–targeted drugs less BBB permeable, identifying BrMeI as our lead compound. BrMeI concentrations sufficient for interacting with DA D₂-like receptors were within the range of the plasma and tissue concentrations identified in our tissue and in vivo analyses. Since bromocriptine’s therapeutic plasma concentrations are in the nanomolar range (39), these concentrations were similarly sufficient for bromocriptine to bind DA D₂-like receptors (40) as well as additional receptors, including α_2A-adrenergic and serotonin receptors (8,10,35,37). These receptors also modulate metabolism (1), likely further contributing to bromocriptine’s efficacy. Consistent with this, higher concentrations of BrMeI shared bromocriptine’s binding to the same dopaminergic and non-dopaminergic receptors.

BrMeI was preferentially distributed to the periphery within 15 min of administration. Resulting differences between brain and plasma ratios for BrMeI versus bromocriptine were likely driven by higher BrMeI plasma concentrations. Because bromocriptine and BrMeI both exhibited comparable phase I metabolic stability, differences in plasma concentration between the two drugs were most likely not due to differences in hepatic phase I degradation. Rather, differences in drug distribution may be accounted for by BrMeI binding to plasma proteins (e.g., albumin) as well as tissue binding in the periphery. Contrary to our original hypothesis, we discovered that BrMeI was equally distributed between the brain and periphery 1 h after administration. This suggests that MeI modification did not fully block BrMeI’s entrance into the CNS over time; further work is needed to determine what functionality, if any, BrMeI exhibits in brain. Nevertheless, compared with bromocriptine, BrMeI was present in the periphery and brain at much higher concentrations long after unmodified bromocriptine was cleared. Thus, despite its lower potency, BrMeI has a longer opportunity to signal, including in the periphery.

Our in vivo metabolic data validate bromocriptine’s ability to improve dysglycemia, consistent with preclinical and clinical data (41). We recently demonstrated that bromocriptine acts directly on peripheral dopaminergic targets, including α-cell and β-cell D2R/D3R to reduce glucagon and insulin secretion. Such actions may resensitize insulin-sensitive tissues and reduce β-cell stress (1,8). This provides a rationale for using BrMeI to dissect central versus peripheral mechanisms of D2R/D3R-mediated glycemic control. Our studies comparing systemic versus intracerebroventricular drug administration strongly suggest that stimulation of CNS dopaminergic targets by either bromocriptine or l-DOPA is insufficient to modify glycemic control. Instead, these data emphasize the importance of tandem targeting of both central and peripheral targets for effective glycemic control as well as improved dysglycemia. Indeed, there is coordination between the hypothalamus and the autonomic nervous system to control energy metabolism (42). Hypothalamic-autonomic nervous system coordination that mediates fat browning in relation to DA receptor agonism (43) may also explain the lack of net effects on weight with improved glucose tolerance.

Limitations include incomplete restriction of BrMeI to the periphery with BBB penetrance over time. Thus, we cannot completely rule out that BrMeI in the brain contributes metabolically. While possible, this is unlikely given absence of significant effects in our OGTT studies. It is also possible that BrMeI’s diminished potency may account for the lack of an effective metabolic response. Evidence of a partial in vivo response in our ITT assay suggests that BrMeI retains some potency. Future efforts are required to create the next generation of drugs with tighter brain exclusion. Another limitation is lack of functional characterization of BrMeI’s actions on non-dopaminergic targets, including serotonergic and adrenergic receptors. Additionally, our in vivo metabolic assays did not examine the impact of circadian rhythms by sampling at multiple zeitgebers or evaluate metabolic impacts of bromocriptine versus BrMeI on lean mice. Lastly, we did not directly examine potential sex differences. Interestingly, a small study demonstrated that men and women responded similarly to bromocriptine therapy for APD-induced hyperprolactinemia (44). These data offer a rationale for comparing effects of BrMeI versus bromocriptine in males and females, including examining end points according to estrous phase.

Ultimately, we propose that future generations of peripherally limited D2R/D3R agonists can treat APD-induced dysglycemia. Because APDs act on peripheral dopaminergic targets (1,45), co-administration of a peripherally-limited D2R/D3R agonist could outcompete peripheral APD actions to overcome APD-induced dysglycemia without interfering with APDs’ intended therapeutic actions in the CNS.

Our results suggest that coordinated signaling via CNS and peripheral DA D₂-like receptors is required for bromocriptine’s metabolic effects, underscoring the importance of both peripheral and CNS dopaminergic metabolic regulation. Moreover, the design of peripherally limited dopaminergic agonists opens the door to new classes of drugs for more effective treatment of dysglycemia.

Acknowledgments. The authors gratefully thank Drs. Jeffrey Deschamps (Naval Research Laboratory), Caitlin Burzynski (The Johns Hopkins University), Jonathan Javitch (Columbia University), Vijay Yechoor (University of Pittsburgh), and George Gittes (University of Pittsburgh) for assistance and helpful discussions throughout this study. The authors also thank Raghunath Singh (Centre for Addiction and Mental Health), Emily Au (University of Toronto), and Sally Wu (University of Toronto) for their technical and experimental assistance. The graphical abstract and illustrations featured in Fig. 6 and Supplementary Fig. 6 were created with Biorender.com.

Funding. Funding support was provided by National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases grant R01DK124219 (Z.F.) and National Institute on Drug Abuse (NIDA) grants K08DA031241 (Z.F.) and DP1DA058385 (C.A.B.), Department of Defense grants PR141292 (Z.F., G.J.S.) and PR210207 (Z.F.), U.S. Department of Justice/Drug Enforcement Agency (Interagency agreement D-15-OD-0002) (A.J.), Department of Veterans Affairs Merit Review (I01BX002758) and Department of Veterans Affairs Award Senior Research Career Scientist programs (1IK6BX005754) (A.J.), and National Institutes of Health/NIDA (Interagency agreement ADA12013) (A.J.), the John F. and Nancy A. Emmerling Fund of The Pittsburgh Foundation (Z.F.), the NIDA Intramural Research Program Z1ADA000424 (A.H.N.), the NIDA Medications Development Program Z1ADA000611 (A.H.N.), and the NIDA Addiction Treatment Discovery Program (ATDP).

The contents do not represent the views of the U.S. Department of Veterans Affairs, U.S. Department of Justice, Drug Enforcement Administration, or the U.S. Government.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. A.B., J.O.M.-R., G.J.S., A.H.N., and Z.F. wrote the manuscript. A.B., M.E., Z.J.F., D.A., R.R., S.P., J.O.M.-R., C.A.B., A.J.E., A.J., M.K.H., G.J.S., B.S.S., A.H.N., and Z.F. reviewed and edited the manuscript. A.B., M.E., C.A.B., and A.H.N. performed synthesis and characterization of test chemicals. Z.J.F. performed insulin secretion assays. D.A. performed nanoBRET assays. R.R. and B.S.S. performed metabolic stability and pharmacokinetic studies. S.P., M.K.H., and G.J.S. performed rodent metabolic studies. A.J.E. and A.J. performed mitogenesis assays. Z.F. provided project administration and supervision. Z.F. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis..

Prior Presentation. Parts of this study were presented in abstract form at the 82nd Scientific Sessions of the American Diabetes Association, New Orleans, LA, 3–7 June 2022.