The impact of acute rosiglitazone on insulin pharmacokinetics at the blood‐brain barrier

Abstract Introduction CNS insulin levels are decreased and insulin receptor signalling is dampened in Alzheimer's disease (AD). Increasing CNS insulin levels through a variety of methods has been shown to improve memory. Indeed, medications routinely used to improve insulin resistance in type 2 diabetes are now being repurposed for memory enhancement. CNS insulin is primarily derived from the circulation, by an active transport system at the blood‐brain barrier (BBB). The goal of this study was to determine whether rosiglitazone (RSG), a drug used to improve insulin sensitivity in type 2 diabetes, could enhance insulin transport at the BBB, as a potential therapeutic for improving memory. Methods Using radioactively labelled insulin and the multiple‐time regression analysis technique, we measured the rate of insulin BBB transport and level of vascular binding in mice pretreated with vehicle or 10 µg RSG in the presence or absence of an insulin receptor inhibitor. Results Although we found acute RSG administration does not affect insulin transport at the BBB, it does restore BBB vascular binding of insulin in an insulin receptor–resistant state. Conclusions Acute RSG treatment does not alter insulin BBB transport in healthy mice but can restore insulin receptor binding at the BBB in an insulin‐resistant state.

insulin resistance in post-mortem human brain slices, measuring the response to ex vivo insulin stimulation and subsequent downstream insulin signalling. 7,8 As the focus of the work presented here is at the BBB, rather than the CNS or peripheral organs, insulin resistance in this study is described by the loss of insulin binding to the insulin receptor.
Endogenous CNS insulin is primarily derived from the periphery by a saturable transport system located at the BBB. 9,10 If CNS insulin resistance occurs due to impaired access from the BBB, improving transport across the BBB could improve CNS insulin resistance, 11 similar to what has been proposed for CNS leptin resistance and the development of obesity. 12,13 Insulin in the CNS can act both as a trophic factor as well as a growth factor, regulating cell growth, mitochondrial function, synaptic plasticity and cognitive function.
Therefore, alterations of insulin transport across the BBB have the potential to affect memory. Indeed, increasing the central nervous system (CNS) insulin level has been proven to be beneficial by improving memory. [14][15][16] Several pharmacologic options are used to treat peripheral insulin resistance and restore insulin receptor sensitivity, such as those used in type 2 diabetes. These drugs are now being repurposed for use in memory improvement. Antidiabetic drugs that activate peroxisome proliferator-activated receptor gamma (PPARγ) improve cognition in diabetic mice, 17 in the aged rat 18 and in AD mice. 19,20 However, the clinical data regarding the effect of PPARγ activation on memory improvement have been mixed. While some studies have shown improvements, [21][22][23] others have been negative. [24][25][26] Rosiglitazone (RSG) is a PPARγ agonist used to enhance insulin sensitivity via various mechanisms. 27 RSG can decrease phosphorylation of proteins that would otherwise dysregulate insulin receptor activated pathways. 27 Perhaps the differences in beneficial effect of RSG in clinical and preclinical studies are due to the ability of RSG to access the CNS. Indeed, RSG is able to cross the BBB in mice but it is also a ligand for P-glycoprotein. 28 In human brain endothelial cells, the brain-to-blood transport of RSG is two times greater than the blood-to-brain influx. 28 Therefore, it is possible in humans that RSG is unable to sufficiently access the CNS. Despite the link between type 2 diabetes and CNS insulin resistance, the effect of PPARγ agonists at the BBB on insulin transport have not been explored.
In the current study, our research explores the effects of RSG on insulin pharmacokinetics at the murine BBB to measure the transport rate of insulin as well as the binding capacity of insulin to the brain endothelium. Further, we used the selective antagonist to the insulin receptor, S961, 29 to determine what effects RSG might have in an insulin-resistant state. Acute treatment with S961 has been previously shown to induce hyperglycaemia and glucose intolerance. 30 We focus on regions within the whole brain that are well documented for their role in CNS insulin signalling: the olfactory bulb has the fastest transport rate for insulin across the BBB 31 and is an important memory centre in the rodent, the hippocampus is important in memory, and the hypothalamus plays a role in peripheral metabolism.

| Animals
Eight-week-old male CD-1 mice were purchased from Charles River Laboratories. CD-1 mice are an established model for BBB transport studies. 9,32,33 Mice had free access to food and water and were maintained on a 12-hour dark (18:00-06:00 hour)/12-hour light (06:00-18:00 hour) cycle in a room with a controlled temperature (24 ± 1°C) and humidity (55 ± 5%). All studies were approved by the certified local Animal Care and Use Committee and were performed in a facility approved by the Association for Assessment and Accreditation of Laboratory Animal Care. For all studies, mice were anaesthetized with a 0.15 mL intraperitoneal injection of 40% urethane at the beginning of each study to minimize pain and distress.

| Radioactive labeling
Ten micrograms of human insulin (Sigma-Aldrich) were radioactively labelled by the chloramine-T (Sigma-Aldrich) method with 1 mCi 125 I (Perkin Elmer) as previously described. 34 Radiolabelled insulin was separated from free iodine on a Sephadex G-10 column (Sigma-Aldrich). 125 I-insulin was prepared on the day prior or day of experiment. The specific activity of 125 I-insulin is 55 Ci/g as previously calculated. 9 One milligram of bovine serum albumin (Sigma-Aldrich) was radioactively labelled via stannous tartrate (Fisher, MP Biomedicals) with 1 mCi 99m Tc (GE Healthcare). 34 Radiolabelled albumin was separated from free 99m Tc on a Sephadex G-10 column. This agent has a half-life of 6 hours and therefore was prepared freshly on the day of each experiment. Radioactivity was consistently over 90% in the 15% trichloroacetic acid precipitated fractions for both insulin and albumin to confirm successful radioactive labelling.

| Pretreatment with RSG
All mice were pretreated with a 0.2 mL iv injection of 10% dime- Arbor, MI) in 10% DMSO/LR thirty minutes prior to measurement of 125 I-insulin BBB transport by either the blood-to-brain transport method or transcardiac brain perfusion method. This experimental set-up was based on previous studies investigating the acute effect of rosiglitazone. 35,36 2.4 | Blood-to-brain 125 I-insulin transport Following pretreatment with DMSO or RSG, a second 0.1 mL jugular vein iv injection containing 1 × 10 6 counts per minute (cpm) 125 I-insulin and 5 × 10 5 cpm 99m Tc-albumin in 1% bovine serum albumin (BSA)/ LR ± 1 µg S961 (Novo Nordisk) was administered. 99m Tc-albumin was co-injected as a marker for vascular space. 37 At 0.5-10 minutes after the injection, blood was collected from the carotid artery. Mice were decapitated, and the brain was collected immediately, dissected into the hypothalamus, olfactory bulb and remaining whole brain and weighed. Blood was centrifuged at 5400g for 10 min and serum collected. Radioactivity in 50 µL of serum and entire brain samples were counted in a gamma counter (Wizard2; Perkin Elmer). The brain/serum (B/S) ratio (μL/g) of 125 I-insulin and 99m Tc-albumin in each gram of brain sample was calculated separately.

| Transcardiac brain perfusion
Following pretreatment with DMSO or RSG, the thorax was opened, heart exposed, both jugulars severed and the descending thoracic aorta clamped. A 26-gauge butterfly needle was inserted into the left ventricle of the heart, and Zlokovic's buffer (7.19 g/L NaCl, 0.3 g/L KCl, 0.28 g/L CaCl 2 , 2.1 g/L NaHCO 3 , 0.16 g/L KH 2 PO 4 , 0.17 g/L anhydrous MgCl 2 , 0.99 g/L D-glucose and 1% BSA) containing 2 × 10 5 cpm 125 I-insulin was infused at a rate of 2 mL/min for 1-5 minutes. The perfusate was freshly prepared for each study day. Perfusate was collected throughout the study to determine the average cpm/µL of perfusate. After perfusion, the olfactory bulb was collected and the brain dissected into regions (frontal/parietal/ occipital cortex, striatum, hypothalamus, hippocampus, thalamus, cerebellum, midbrain and pons/medulla) according to Glowinski and Iversen; 38 each region was weighed separately. The amount of radioactivity was determined by a gamma counter (Wizard2, Perkin Elmer). Brain/perfusate ratios are calculated by dividing the cpm in a gram of brain by the cpm in a µL of perfusate to yield units of µL/g. In order to determine whether the 10% DMSO solution was affecting 125 I-insulin transport, we also pre-injected another set of mice with 1% BSA/LR prior to transcardiac brain perfusion.

| Multiple-time regression analysis
Multiple-time regression analysis was used as detailed previously 37,39,40 to calculate the rate of unidirectional influx for 125 I-insulin. For blood-to-brain transport studies, the B/S ratios for 125 I-insulin are corrected for vascular space by subtracting the corresponding ratio for 99m Tc-albumin, yielding a delta B/S ratio. Exposure time is calculated by the formula: where Cp(t) is the level of radioactivity (cpm) in serum at time (t).
Exposure time corrects for the clearance of peptide from the blood.
For blood-to-brain studies, the B/S ratios are plotted against the exposure time to calculate the influx of 125 I-insulin using the same formula where Am is level of radioactivity (cpm) per g of brain tissue at time t, Cpt is the level of radioactivity (cpm) per μL serum at time t, K i (μL/gmin) is the unidirectional solute influx from blood to brain and V i (μL/g) is the level of rapid and reversible binding for the brain vasculature.
The slope of the linearity measures K i and is reported with its standard error term. The y-intercept of the linearity measures V i , the initial volume of distribution in brain at t = 0. 37 Blasberg, Fenstermacher and Patlak originally described V i as the volume of test substance that rapidly and reversibly exchanges with the plasma . For cardiac perfusion studies, formula 2 is employed, using the brain/perfusate ratios and the clock time is used in place of exposure time.

| Capillary depletion in mice
To determine whether RSG altered brain capillary sequestration of 125 I-insulin, we performed capillary depletion as adapted to mice. 41,42 Following pretreatment with DMSO or RSG, a second 0.1 mL jugular vein iv injection containing 1 × 10 6 counts per minute (cpm) 125 I-insulin and 5x10 5 cpm 99m Tc-albumin in 1% BSA/LR. Two and a half minutes later, blood was collected from the carotid, the mice decapitated and the whole brain removed. The brain was homogenized with ten strokes of a glass homogenizer in 0.

| Data analysis
, and y-intercepts (V i ) were compared statistically with the Prism 8.0 software package. Differences in V i were compared by two-way analysis of variance (ANOVA) followed by Sidak's post hoc test to determine differences due to RSG and S961 treatment.

| Serum clearance of 125 I-insulin following RSG treatment
Following pretreatment with DMSO or 10 µg RSG, we measured the rate of clearance from blood of 125 I-insulin clearance ± 1 µg S961 using a logarithmic scale. All groups had similar exponential decay curves of insulin ( Figure 1). There was no effect of RSG or S961 on 125 I-insulin clearance. The average half-time clearance for 125 I-insulin was 1.54 minutes.

| 125 I-insulin BBB pharmacokinetics due to RSG: serum contribution
Whole brain, olfactory bulb and the hypothalamus were analysed for differences in the rate of 125 I-insulin transport (slope, K i ) and binding of 125 I-insulin to brain capillaries (y-intercept, V i ). Pretreatment with RSG did not significantly alter the rate of 125 I-insulin BBB transport in whole brain, olfactory bulb or hypothalamus ( Figure 2). As shown previously, 34 inhibition of the insulin receptor with S961 did not affect 125 I-insulin BBB transport in whole brain (Figure 2A). In addition, S961 did not affect 125 I-insulin transport in the olfactory bulb ( Figure 2B) or hypothalamus ( Figure 2C). Transport rates (K i ) were calculated based on the linear transport data of Figure 2 and are presented in Table 1. Average rates of transport (K i ) were 1.10 µL/gmin (whole brain), 2.25 µL/g-min (olfactory bulb) and 1.75 µL/g-min (hypothalamus).
In order to verify 125 I-insulin was fully crossing the brain endothelial cell, we separated the brain capillary fraction from the brain parenchyma to measure the percentage of radioactivity present in each fraction. The majority of 125 I-insulin was present in the brain parenchyma and there were no overt differences due to RSG treatment (data not shown).

| Effect of DMSO on 125 I-insulin BBB transport
DMSO is an organic reagent commonly used to dissolve substances.

| 125 I-insulin BBB pharmacokinetics due to RSG: absence of serum
To determine if RSG could affect the BBB directly in the context of 125 I-insulin transport, we wanted to eliminate all serum factors and investigate the impact of RSG on 125 I-insulin BBB pharmacokinetics in a cardiac perfusion model, using Zlokovic's buffer as the physiological perfusate. To make sure there were not regional differences in the effect of RSG, we investigated 125 I-insulin transport in individual brain regions and summed up these data to derive whole-brain transport data. There was no difference in the rate of 125 I-insulin transport across the BBB or changes in vascular binding due to RSG pretreatment in whole brain, olfactory bulb or hypothalamus ( Figure 5). Transport rates (K i ) were calculated based on the linear transport data of Figure 5 and are presented in Table 2. Of the other regions investigated (striatum, frontal/parietal/occipital cortex, hippocampus, thalamus, cerebellum, midbrain and pons/medulla), there was no effect of RSG on 125 I-insulin BBB pharmacokinetics (data not shown).

| D ISCUSS I ON
Thiazolidinediones are routinely used in the treatment of type 2 diabetes to improve insulin resistance. Increasing insulin sensitivity at the receptor of insulin-sensitive tissues increases their uptake of glucose, causing a reduction in glucose, as exemplified by the use of PPARγ agonists. 43 There has recently been a repurposing of diabetic drugs for improvements in memory, 44 likely due to the link between diabetes and AD. 45 Patients with type 2 diabetes have almost double the risk for developing AD. 1 This association between CNS and peripheral insulin resistance demonstrates the rationale for exploring insulin-sensitizing drugs, such as those used for type 2 diabetes. 46 Here, we investigated the impact of RSG on insulin transport across the murine BBB, the primary means for CNS access to insulin. 47 While RSG did not affect the transport rate of insulin across the BBB, it did improve vascular binding under conditions in which the insulin receptor was acutely inhibited. In our studies, we include a marker for vascular volume, 99m Tc-Alb, and can cor-  Note: 125 I-insulin BBB transport rate (K i ), correlation coefficient (r) and level of vascular binding (V i ) are derived from Figure 2 and expressed ± SE, n = 6-10/group. **P < .001, ***P < .0001. receptors. Restoration of vascular binding could in and of itself act in an endocrine-like manner by altering release of abluminal factors that impact the CNS. 48 We explored the impact of RSG on insulin BBB transport both on the direct effect at the BBB as well as the impact due to bloodborne factors. In the presence of blood factors, RSG did not affect 125 I-insulin transport or binding by itself. However, we also explored the effect of RSG when the insulin receptor was inhibited using the selective antagonist, S961. 29 S961 treatment in mice and rats has been shown to induce features of type 2 diabetes including hyperglycaemia, glucose intolerance and impaired insulin sensitivity. 30,49 Similar to previous studies, there was no change in the rate of 125 I-insulin transport across the BBB when the receptor was

F I G U R E 3 125 I-insulin
inhibited, yet the amount of vascular binding for 125 I-insulin was decreased. 34,50 When mice were pretreated with RSG, this decrease in binding was reversed. In a separate study, pioglitazone, another PPARγ agonist, was able to reverse the effects of S961, restoring insulin sensitivity. 30  Following these studies, we wanted to verify our vehicle (10% DMSO/LR) was not affecting 125 I-insulin transport. Following an iv 30-minutes pretreatment of 10% DMSO/LR or a more physiological injectate solution, 1% BSA/LR, we measured 125 I-insulin pharmacokinetics. We did not observe any difference in 125 I-insulin transport or vascular binding, confirming that the vehicle the RSG is dissolved in does not impact our primary outcome.
By studying 125 I-insulin transport in the presence of blood, we could not deduce whether the impact of RSG was due to a change in serum factors in mediating insulin transport. Therefore, we  Note: 125 I-insulin BBB transport rate (K i ), correlation coefficient (r) and level of vascular binding (V i ) are derived from Figure 5 and expressed ± SE, n = 7-11/group. examined the direct effects of RSG on 125 I-insulin brain uptake from cerebral circulation in the absence of blood-borne factors by brain perfusion. Again, there was no difference in the transport rate or amount of vascular binding due to RSG. By eliminating any contribution of serum factors, we can conclude RSG does not have a direct effect on the BBB to alter 125 I-insulin pharmacokinetics in healthy mice. The transport rate and level of vascular binding for 125 I-insulin were similar between the two methods employed to measure pharmacokinetics (in the presence and absence of serum) in the whole brain and olfactory bulb, and similar values to previous studies reporting on the whole brain. 54 However, in the absence of serum, the transport rate of insulin across the BBB in the hypothalamus nearly tripled and the amount of vascular binding was over 3-fold lower (similar to the level of vascular binding when S961 is present in serum). This suggests that there is likely a regulatory factor in the serum involved in hypothalamic insulin BBB transport and receptor binding. Indeed, it is known serum factors such as triglycerides, free fatty acids and glucose can alter insulin BBB transport. 6,11 Previous studies have observed beneficial effects of acute RSG in cardiac ischemia/reperfusion models when administered 5 minutes prior to injury. 55,56 In brain endothelial cells, PPARβ and PPARδ are the predominant genes expressed compared to other members of the PPAR family. 57 PPARγ is most abundantly expressed in adipose tissue compared to other metabolic organs. However, PPARγ is still expressed in endothelial cells and can regulate the release of nitric oxide 58 which has been shown to affect insulin BBB transport. 59 Another study investigated the acute response of RSG treatment on endothelial function in healthy men and found that a single dose did not affect serum glucose or insulin levels 6 or 24 hours following treatment. 60 Importantly though, endothelial function as measured by flow-mediated endothelium-dependent vasodilation was significantly increased following RSG. These data suggest a direct effect of RSG on the endothelium, independent of metabolic action, that can occur with a single administration in a short time period.
While RSG does not greatly penetrate the murine BBB (0.045% of an iv injected dose), uptake does occur rapidly, within 1 minutes. 28 In addition, RSG is a substrate of p-glycoprotein and, therefore, is readily transported out of the brain. However, we were interested in the direct effect that would be mediated at the luminal surface of the brain endothelial cell, rather than an effect within the CNS. Indeed, PPARγ activation is known to have a direct effect on cerebral vascular expression of proteins, including various adhesion molecules and metalloproteinases. 61 Therefore, we know brain endothelial cells can respond directly to RSG and alter expression of proteins at the brain endothelial cell surface.
Finding methods to increase transport of insulin into the brain could elucidate potential pharmacologic solutions to conditions in which CNS insulin resistance occurs. An underlying question remains: What roles do current type 2 diabetes medications play in central insulin transport? Therefore, it is worth investigating whether RSG influences insulin BBB transport in a chronic insulin-resistant state, both peripherally and centrally, in addition to how the ability of RSG to enhance insulin binding to brain endothelial cells might improve CNS insulin resistance. Our studies focus on a PPARγ agonist in determining the effect on insulin at the BBB, both regarding transport and binding. However, whether other type 2 diabetes medications have an acute or chronic impact on insulin BBB transport or binding remain to be determined.

ACK N OWLED G EM ENTS
This work was supported by the University of Washington Diabetes Research Center (to EMR), the Medical Student Training in Aging Research (MSTAR) programme (NIA T35 AG26736-14 to DCG) and the Veterans Affairs Puget Sound Research and Development (to EMR and WAB).

CO N FLI C T S O F I NTE R E S T
The authors have no conflicts of interest to disclose.

AUTH O R ' S CO NTR I B UTI O N S
EMR conceived and designed the study. DCG and EMR planned and executed the experiments, analysed the data, interpreted the results and drafted the manuscript. WAB aided in data interpretation and critically reviewed the manuscript.

E TH I C S A PPROVA L
The present study was designed and conducted in compliance with

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.