A Novel Radioligand Reveals Tissue Specific Pharmacological Modulation of Glucocorticoid Receptor Expression with Positron Emission Tomography

Abstract

The complexity of glucocorticoid receptor (GR) signaling cannot be measured with direct tissue analysis in living subjects, which has stifled our understanding of GR’s role in human physiology or disease and impeded the development of selective GR modulators. Herein, we report ¹⁸F-5-(4-fluorobenzyl)-10-methoxy-2,2,4-trimethyl-2,5-dihydro-1H-chromeno[3,4-f]quinoline (¹⁸F-YJH08), a radioligand that enables noninvasive measurements of tissue autonomous GR expression levels in vivo with positron emission tomography (PET). YJH08 potently binds GR (K_i ∼ 0.4 nM) with ∼100-fold selectivity compared to nuclear hormone receptors in the same subfamily. ¹⁸F-YJH08 was prepared via Cu(OTf)₂(py)₄-mediated radiofluorination of an arylboronic acid pinacol ester with ∼12% decay corrected radiochemical yield from the starting ¹⁸F-fluoride ion. We applied treatment with the tissue-wide GR agonist dexamethasone and adrenalectomy and generated an adipocyte specific GR knockout mouse to show that ¹⁸F-YJH08 specifically binds GR in normal mouse tissues, including those for which aberrant GR expression is thought to drive severe diseases (e.g., brain, adipose tissue, kidneys). Remarkably, ¹⁸F-YJH08 PET also revealed that JG231, a potent and bioavailable HSP70 inhibitor, selectively degrades GR only in the adipose tissue of mice, a finding that foreshadows how GR targeted PET might be integrated into drug discovery to screen for selective GR modulation at the tissue level, beyond the historical screening that was performed at the transcriptional level. In summary, ¹⁸F-YJH08 enables a quantitative assessment of GR expression levels in real time among multiple tissues simultaneously, and this technology is a first step toward unraveling the daunting complexity of GR signaling and rationally engineering tissue specific therapeutic modulators in vivo.

The glucocorticoid receptor (GR) regulates an exceptionally diverse spectrum of physiological processes and is expressed in nearly every normal cell type.^1,2 As a nuclear hormone receptor (NHR), GR regulates biology primarily through an inducible transcriptional mechanism, in which cytosolic corticosteroid agonists bind the isoform GRα to trigger a conformational change that encourages nuclear translocation, homodimerization, and DNA binding to regulate transcription. Agonist bound GRα can also regulate transcription through alternate mechanisms (e.g., tethering other transcription factors to DNA) or through nongenomic signaling. Moreover, other less abundant isoforms of GR, most notably the splice variant GRβ, may regulate additional dimensions of biology in a corticosteroid independent fashion. Further adding to the complexity, GR engages in both cell intrinsic and multiorgan endocrine signaling through secreted hormones (e.g., the hypothalamus–pituitary–adrenal axis).

Because of its seminal importance to the homeostasis of diverse cell types, dysregulation of GR signaling is accompanied by significant, and often highly debilitating, phenotypic changes. For instance, excess or deficient corticosteroid production and the concomitant dysregulation of cellular GRα activity lead to Cushing syndrome or Addison’s disease, respectively. Moreover, tissue changes in GR expression—either losses or gains—have been implicated in the pathobiology of a wide spectrum of maladies, including mood disorders, glomerular diseases, and more recently, cancers.³

Cell and genetically engineered animal models have been the mainstays for carefully articulating the above-mentioned features of GR (patho)biology. Fully understanding GR’s role in human physiology and disease, as well as developing next generation therapeutics to more selectively modulate GR, requires new technologies that can safely probe GR signaling in living subjects. The lack of noninvasive biomarkers to interrogate GR signaling in humans has left the field to perform very suboptimal assays, for instance, analysis of GR and target gene expression levels from tissues collected at autopsy.^4,5

With these considerations in mind, we hypothesized that direct measurement of GR expression levels in tissues with nuclear imaging could be an impactful starting point toward developing a complete repertoire of noninvasive tools to study the GR signaling axis. Radiolabeled steroid agonists targeting other NHRs in the same family have been implemented clinically with great success,^6,7 which motivated us to design a radiolabeled small molecule probe targeting the ligand binding domain (LBD) of GR.

Herein, we describe the design, synthesis, and pharmacological evaluation of ¹⁸F-5-(4-fluorobenzyl)-10-methoxy-2,2,4-trimethyl-2,5-dihydro-1H-chromeno[3,4-f]quinoline (¹⁸F-YJH08), a small molecule radiotracer that measures GR expression levels in vivo with PET. YJH08 is a potent ligand for GR with weak agonist activity. A high yielding radiosynthesis of ¹⁸F-YJH08 was achieved using Cu(II) mediated fluorination of an arylboronic acid pinacol ester precursor. ¹⁸F-YJH08 specifically bound to GR in many normal mouse tissues in vivo, including several of great interest to the GR signaling community, including the brain, adipose tissue, and adrenals. Lastly, ¹⁸F-YJH08 PET was applied to study the anti-GR effects of the bioavailable HSP70 inhibitor JG231. Remarkably, GR degradation only occurred in adipose tissue. This unexpected finding underscores how ¹⁸F-YJH08 might be applied to identify tissue specific GR modulators early in the drug discovery process, an important unmet clinical need owing to the intense side effects associated with broad spectrum GR modulators (e.g., the semisynthetic corticosteroid dexamethasone).

Results and Discussion

Design, Synthesis, and in Vitro Characterization of YJH08

The abundance of literature reporting suboptimal imaging findings with radiolabeled corticosteroids suggested to us that we ought to consider nonsteroidal structures with superior pharmacokinetics.^8,9 Agonists generally bind to NHRs with higher affinity compared to antagonists, and highly potent and specific synthetic agonists built on benzopyrano[3,4-f]quinolines (Figure 1) were previously shown to be pharmacologically active and well tolerated in animals.^10,11 Therefore, we chose to design a radioligand for GR based on this structural motif.

Figure 1.

Structure and numbering of the benzopyrano[3,4-f]quinoline and the structure of F-5-(4-fluorobenzyl)-10-methoxy-2,2,4-trimethyl-2,5-dihydro-1H-chromeno[3,4-f]quinoline (YJH08).

Prior medicinal chemistry has shown that the C5 position on the benzopyrano[3,4-f]quinoline scaffold can tolerate modification without diminishing potency and selectivity for GR, whereas the C10 position is less accepting of manipulation.^12,13 On this basis, we envisioned installing a fluorine atom on the C5 position of the core scaffold. Since ¹⁸F-alkyl fluorides can rapidly radio-defluorinate in vivo, which results in high bone uptake of the fluoride anion, we chose to couple fluorine to an aromatic ring, a more metabolically stable approach. Retrosynthetically, we anticipated that fluorine-18 could be introduced on the aryl ring through one of several reported copper mediated radiofluorination chemistries.^14,15

The synthesis of YJH08 began by preparing the methyl acetal 1 in six steps following literature precedent (Scheme 1, see also Supporting Information Methods).¹⁶ To avoid undesired oxidative side products during fluorination chemistry, the secondary amine on the benzopyrano[3,4-f]quinoline scaffold was protected with Boc₂O to confer 2. A Grignard reaction with 4-bromobenzylmagnesium bromide was applied to install an aromatic moiety on the C5 position (3). A Stille-type coupling was then invoked to convert the aryl bromide to an arylboronic acid pinacol ester (4). This functional group was desired as it has been successfully used as a substrate for regioselective copper mediated radiofluorination reactions.^17,18 Satisfyingly, we were able to convert this molecule to the aryl fluoride 5 at 68% yield within 2 h at 120 °C using Cu(OTf)₂(Py)₄, pyridine, KF, and K_2.2.2. The Boc group was then efficiently removed in 30 min with 50% TFA in DCM to confer 6. In summary, the overall yield of 4 was 25%, and the yield of 6 was 15%.

Scheme 1. Synthesis of YJH08.

Reagents and conditions: (a) n-BuLi, (Boc)₂O, THF, −45 °C, 51%; (b) 0.25 M 4-Bromobenzylmagnesium bromide in anhydrous Et₂O, BF₃·Et₂O, anhydrous DCM, −15 °C, 73%; (c) Pd(dppf)Cl₂·DCM, KOAc, B₂Pin₂, anhydrous dioxane, 100 °C, 68%; (d) Cu(OTf)₂(Py)₄, Py, KF, K₂₂₂, DMF, 120 °C, 68%; (e) 50% TFA in CH₂Cl₂, 40 °C, 30 min, 95%.

Next, the affinity of YJH08 for GR and the other type 1 subfamily 3 nuclear hormone receptors was evaluated using ³H-ligand displacement assays on cells. The K_i of YJH08 for GR was determined to be ∼0.4 nM, or about 10-fold more potent than the agonist dexamethasone (Table 1 and Figure 2A). Moreover, YJH08 bound to GR with at least 100 fold higher relative affinity than other type 1 subfamily 3 nuclear hormone receptors (Table S1 and Figure S1). We conducted cell-based studies to determine if YJH08 modulates GR activity. Real time PCR showed that YJH08 (10 μM, 6 h exposure) elevated the levels of two GR target genes (GILZ, FKBP5) in DU145, a GR positive human prostate cancer cell line (Figure 2B). The induction per gene was lower than what was observed for dexamethasone (1 μM, 6 h exposure), and one gene, Per1, was induced by dexamethasone but not YJH08. Overall, these data suggest that YJH08 may be a mild GR agonist.

Table 1. A Summary of the Binding Affinity of YJH08 Compared to Conventional Agonists for Subfamily 3 Nuclear Hormone Receptors^a.

	Ki (M)
ligand	GR	AR	PR	ER	MR
referenceb	2.03 × 10–9	4.96 × 10–10	6.27 × 10–10	6.94 × 10–9	1.41 × 10–8
YJH08	4.01 × 10–10	2.20 × 10–7	5.55 × 10–7	1.90 × 10–5	2.90 × 10–7

^aThe competitive binding data are presented in Figure 2A and Figure S2. The data were fit using a one site K_i and nonlinear regression with Prism software. The data are representative of two independent experiments.

^bReference ligands were dexamethasone (GR), dihydrotestosterone (AR), progesterone (PR), estradiol (ER), and aldosterone (MR). Abbreviations: GR = glucocorticoid receptor, AR = androgen receptor, PR = progesterone receptor, ER = estrogen receptor, MR = mineralocorticoid receptor.

Figure 2.

YJH08 is a high affinity ligand for GR with weak agonist properties. (A) A ligand displacement curve showing YJH08 and dexamethasone displace ³H-dexamethasone in DU145 cells. The data were fit using a one site nonlinear regression, and the coefficient of determination (R²) values are highlighted in the inset box. Additional competitive binding curves are presented in the Supporting Information. (B) Real time PCR data showing the relative induction of three GR target genes in DU145 cells by dexamethasone (1 μM) or YJH08 (10 μM). Cells were harvested for mRNA analysis after a 6 h exposure to drug or vehicle. The data, normalized to the ΔCt levels in vehicle treated cells, support an assessment that YJH08 may be a weaker GR agonist than dexamethasone. Statistical significance was assessed using an unpaired, two-sided Student’s t test. **P < 0.01, *P < 0.05.

Radiosynthesis of ¹⁸F-YJH08

A focused screen of reaction conditions revealed that ¹⁸F-Boc-YJH08 could be efficiently synthesized by hand in 20 min using K₂CO₃/KOTf as the base to mobilize ¹⁸F-fluoride anions from a QMA column (4.7 ± 0.8%, see Table S2, Figure S2, and Scheme 2). We also found that this reaction was amenable to automation using an ELIXYS FLEX/CHEM synthesis module, and HPLC analysis of the reaction progress showed that the reaction produces one major radioactive peak corresponding to ¹⁸F-Boc-YJH08 (Figure 3A). After purification, the Boc protecting group was removed with 50% TFA in DCM to confer ¹⁸F-YJH08 in 12.4 ± 3% decay corrected radiochemical yield (Figure 3B and Figure S3). The molar activity of ¹⁸F-YJH08 was calculated to be 0.042 ± 1 Ci/μmol over four independent radiosyntheses (Figure S4). The partition coefficient (log P) of ¹⁸F-YJH08 was assessed in instant thin layer chromatography and calculated to be 2.78.

Scheme 2. Radiosynthesis of ¹⁸F-YJH08.

Figure 3.

(A) Stacked semipreparative UV HPLC traces showing the retention of arylboronic acid pinacol ester starting material (4) and the Boc-protected aryl fluoride product (5). Below these analytical standards are shown the UV trace of the complex reaction mixture at 90 min and the corresponding trace collected using a radiation detector. A single product peak is observed in RAD-HPLC corresponding to the desired product, ¹⁸F-5, while a minor UV peak is observed at the same retention time, as desired. The peak was collected and transferred immediately to a reaction vial for Boc deprotection. (B) Stacked semipreparative UV HPLC traces showing the retention of the Boc protected aryl fluoride (5) and the deprotected final product (6). Below the analytical standards are shown the UV trace of the complex reaction mixture at 30 min and the corresponding trace collected using a radiation detector. A single product peak is observed in RAD-HPLC corresponding to the desired product, ¹⁸F-YJH08, while a minor UV peak is observed at the same retention time, as expected. After purification, the purity and identity of ¹⁸F-YJH08 were confirmed using analytical HPLC with an isocratic gradient.

¹⁸F-YJH08 Specifically Binds GR in Vivo and Measures Receptor Saturation by a Synthetic GR Agonist

To understand the biodistribution of ¹⁸F-YJH08 in vivo, we first conducted a dynamic PET scan over 60 min in immunocompetent intact male C57Bl6/J mice. Region of interest analysis was applied to generate time activity curves, which revealed several enlightening trends (Figure 4A,B and Figure S5). First, the radiotracer rapidly cleared from serum, and fitting the data with a two phase exponential decay curve estimated the fast and slow half-lives to be ∼21 s and ∼760 s, respectively. Moreover, liver uptake was prominent at very early time points post injection (0–300 s), with continuous washout from ∼300 to 3600 s. These data suggest a hepatobiliary mechanism of clearance. Consistent with this model, significantly lower radiotracer accumulation was observed in the kidneys from 0 to 3600 s. Radiotracer uptake in the bone was low and approximately equivalent to muscle, which suggests that the compound is stable to radiodefluorination in vivo.

Figure 4.

¹⁸F-YJH08 accumulates in GR rich tissues important to normal and disease physiology, including the brain. (A) A time activity curve from a dynamic PET scan shows the magnitude of radiotracer uptake and clearance in major tissue compartments of a male C57Bl6/J mouse. The radiotracer rapidly targeted the liver within 200 s post injection, with continuous washout from 300–3000 s, suggesting a hepatobiliary clearance mechanism. An early increase and washout in kidney uptake was followed by stable accumulation. Muscle and bone uptake were significantly lower than liver and kidney. (B) Representative sagittal and coronal slices of the mouse used to derive the time activity curves provide a visual rendering of relative radiotracer biodistribution in tissues. (C) Region of interest analysis performed on the PET data from the dynamic acquisition shows high and time dependent radiotracer accumulation in the brain. At right are shown representative transaxial and sagittal images centered on the mouse brain (arrow) to highlight the uptake of the radiotracer over time. (D) Region of interest analysis on the supraspinal brown adipose tissue depot was used to calculate the depicted time activity curve. The magnitude of radiotracer uptake was statistically higher that muscle and bone from ∼300 to 3600 s post injection. At right are shown representative transaxial and sagittal images centered on the supraspinal pad of brown adipose tissue. Images from the indicated time intervals are shown to depict the accumulation of the radiotracer over time. Washout of the radiotracer from the blood pool can also be observed in the transaxial view.

Notably, we also observed clear radiotracer uptake in tissues for which aberrant GR expression is thought to drive disease pathobiology. For instance, we observed high radiotracer uptake in the brain (Figure 4C), a significant finding as the neuroimaging community has long sought to develop a noninvasive tool to measure changes in GR expression thought to be occurring in the brains of individuals suffering from mood disorders (e.g., anxiety, depression).¹⁹ Moreover, high tracer uptake occurred in supraspinal brown adipose tissue in which GR expression is expressed and hypothesized to regulate seminal processes like thermogenesis and beiging (Figure 4D).²⁰ Last, we performed a broader survey of radiotracer biodistribution by analyzing tissues post mortem. In general, numerous GR rich tissues had high radiotracer uptake, including the adrenals, lungs, and heart (Table 2).

Table 2. Biodistribution Data Collected 30, 60, 90 min after Injection of ¹⁸F-YJH08 in C57Bl6/J Mice^a.

Tissue	30 min	60 min	90 min
blood	0.85 ± 0.1	0.57 ± 0.02	0.49 ± 0.06
brain	1.68 ± 0.14	1.33 ± 0.06	0.97 ± 0.03
heart	3.04 ± 0.75	1.33 ± 0.21	0.92 ± 0.14
lung	2.65 ± 0.19	2.02 ± 0.29	1.39 ± 0.1
liver	11.03 ± 0.89	7.7 ± 0.69	8.43 ± 1.38
spleen	1.59 ± 0.13	0.83 ± 0.02	0.74 ± 0.15
pancreas	4.5 ± 0.15	2.63 ± 0.05	1.9 ± 0.41
kidney	4.5 ± 0.5	2.35 ± 0.34	1.86 ± 0.26
stomach	1.88 ± 0.96	1.63 ± 0.92	2.61 ± 1.54
small intestine	19.05 ± 3.27	8.61 ± 2.68	7.67 ± 2.34
large intestine	1.33 ± 0.35	1.23 ± 0.43	6.67 ± 4.73
supraspinal BAT	14.41 ± 3.42	16.4 ± 4.93	14.58 ± 1.91
muscle	1.49 ± 0.2	1.15 ± 0.03	0.85 ± 0.18
bone	1.11 ± 0.15	0.61 ± 0.15	0.58 ± 0.16
adrenal gland	20.02 ± 2.21	9.43 ± 0.08	5.72 ± 2.32

^aThe data are represented as mean ± standard deviation of the % ID/g.

We next tested if ¹⁸F-YJH08 accumulation in tissues is due to specific GR binding. First, biodistribution studies were conducted to determine if dexamethasone treatment lowered ¹⁸F-YJH08 uptake in normal mouse tissues due to competitive GR binding. Previous studies, including our own experience with a radiofluorinated corticosteroid termed ¹⁸F-GR02, have shown that it can be challenging to engage GR with a cointravenous injection of excess cold ligand to an extent that measurably alters tracer biodistribution in tissues.⁹ We proposed that the solubility limits of hydrophobic steroids and steroid-like molecules in intravenous formulations account for this observation, as we also found that administering a competitive ligand for GR via gavage hours to days in advance of the imaging study very effectively suppressed radiotracer uptake. With this experience in mind, we administered dexamethasone via oral gavage in two ways. One treatment arm of C57Bl6/J mice received dexamethasone once (50 mg/kg) at 4 h prior to ¹⁸F-YJH08 injection, and another cohort received the drug once daily (100 mg/kg) via oral gavage for 3 days prior to the radiotracer injection. Biodistribution studies showed that a single dose of dexamethasone reduced radiotracer uptake in the small intestine, while a higher dose administered over several days suppressed radiotracer uptake in many tissues compared to mice receiving vehicle (Figure 5A and Figures S6, S7). To confirm that high dose dexamethasone did not unexpectedly reduce GR expression levels, we examined GR expression in select tissues in which a significant degree of radiotracer suppression was observed. We found that GR expression levels were equivalent in several normal tissues from treated and untreated mice, which provides confidence that the suppression of radiotracer uptake was due to competitive GR binding (Figure S8). We next evaluated radiotracer uptake in adrenalectomized (adx) mice, as disruption of the HPA axis is well understood to ablate circulating corticosteroids while triggering normal cells to autonomously upregulate GR. ¹⁸F-YJH08 uptake was higher in the tissues of adx mice compared to intact mice (Figure 5A and Figure S9). Collectively, these data show that tissue uptake of the radiotracer is a reflection of GR abundance and occupancy by endogenous ligand.

Figure 5.

¹⁸F-YJH08 specific binding to GR in vivo and application to measure receptor occupancy by endogenous or exogenous ligands. (A) A summary of the percent change in radiotracer uptake for mouse tissues in three separate cohorts comparing (1) the impact of dexamethasone treatment at 50 mg/kg via gavage 4 h before the radiotracer injection, (2) the impact of dexamethasone treatment at 100 mg/kg/day via gavage 3 days before the radiotracer injection, and (3) the impact of adrenalectomy (adx.) on ¹⁸F-YJH08 tissue biodistribution. (B) Representative PET/CT slices showing the suppression of radiotracer uptake due to prior dexamethasone treatment in the brain (top white arrow), supraspinal brown fat (middle white arrow), and the kidneys (bottom white arrow). A tan arrow was placed over a submandibular region that showed evidence of specific binding that could represent anterior cervical or supraclavicular brown fat. These data were taken from the cohort that received 100 mg/kg/day. (C) Digital autoradiography showing the relative radiotracer binding in coronal sections of the brains of mice from adx mice or intact mice treated with dexamethasone. (D) Maximum intensity projections showing the distribution of ¹⁸F-YJH08 in wild type female C57Bl6/J mice and a female littermate with adipocyte specific GR knockout. White arrows are positioned on the supraspinal brown fat depot and the bilateral scapular brown fat depots on the transverse view of the wild type mouse. A white bracket is positioned over the region of the abdomen expected to be encompassed by the omentum on the sagittal view of the wild type mouse. Radiotracer uptake is absent in all highlighted regions in the GR^fl/fl, Adipoq-Cre mouse.

On PET, visually obvious decreases in radiotracer uptake due to dexamethasone treatment (100 mg/kg) were observed in several tissues (e.g., brain, supraspinal brown adipose tissue, and kidneys, see Figure 5B). Moreover, digital autoradiography was applied to compare the intensity of radiotracer binding in the brain sections from adx mice to those from intact mice treated with dexamethasone (Figure 5C). Comparison of the relative intensity of the pseudocolor showed that radiotracer binding was uniformly higher in adx mice, as expected. Radiotracer binding was high in the isocortex, a relatively large anatomical area known to have high GR expression in the mouse brain based on in situ hybridization data (Figure S10). GR is highly expressed in smaller structures within the brain, for example the hippocampal formation, although this was challenging to detect with the limited resolution of autoradiography. Studies in nonhuman primates or humans will be required the interrogaion of the regions of specific GR binding within complex organs like the brain.

Finally, we conducted an exploratory study to determine if genetic knockout of GR reduced ¹⁸F-YJH08 uptake. Mice bearing loxP sites flanking exon 3 of the Nr3cl gene were crossed with mice bearing Cre recombinase under the control of the adiponectin promoter to induce adipocyte specific GR knockout. PET/CT data clearly showed that ¹⁸F-YJH08 uptake was significantly reduced in several fat depots within the GR^fl/fl, Adipoq-Cre knockout mice, most obviously the supraspinal and scapular brown fat depots (Figure 5D). Moreover, overall radiotracer binding in the abdomen was significantly reduced in the knockout mice compared to wild type litter mate control, which is likely attributable to a loss of GR in fat depots as well as the greater omentum. To our knowledge, this is the first reported evidence of specific radioligand binding to GR in the adipose tissue.

¹⁸F-YJH08 Reveals Tissue Specific Modulation of GR Expression by JG231, a Potent HSP70 Inhibitor

The limitations of broad-spectrum GR modulators like corticosteroids are well-known, and the severe side effects associated with their long-term use (e.g., diabetes mellitus, osteoporosis) motivated drug development campaigns starting in the 1990s to identify safer GR modulators that selectively modulate GR activity.²¹ We hypothesized that ¹⁸F-YJH08 PET could complement these ongoing efforts by revealing tissue specific drug–GR interactions. To test this concept, we evaluated the biodistribution of ¹⁸F-YJH08 in mice after treatment with JG231, a potent HSP70 inhibitor that we have shown can induce GR degradation (Figure 6A and B).²² JG231 has a well-defined pharmacokinetic profile, and our prior experience with this molecule showed it to have antitumor activity in prostate cancer mouse models.²³ Using the same treatment strategy, mice were subjected to 4 mg/kg/day for 3 days intraperitoneally and subsequently treated with ¹⁸F-YJH08. On PET, the only visually obvious alterations in radiotracer distribution were noted in the supraspinal brown fat depot and in a portion of the abdomen that we interpreted to be the greater omentum (Figure 6C). In these regions, it appeared that JG231 suppressed radiotracer binding, perhaps indicative of GR degradation. Post mortem imaging of the supraspinal brown fat depot more clearly showed that JG231 dramatically reduced ratiotracer binding compared to the same tissue harvested from mice treated with vehicle (Figure 6C). Biodistribution studies also suggested that JG231 most significantly induced GR degradation in brown fat (Figure 6D). Immunoblot studies corroborated the biodistribution data, showing approximately ∼50% reduction in GR expression levels due to JG231 treatment in brown fat, and no change in GR levels from other tissues, for example, kidneys, liver, and muscle (Figure 6E and Figure S11). To our knowledge, this is first example of tissue specific GR degradation. More generally, these early data suggest that ¹⁸F-YJH08 PET could be applied to identify therapies with unexpected tissue tropism or prospectively engineer desirable tissue specificity into experimental GR modulators.

Figure 6.

¹⁸F-YJH08 revealing unexpected tissue specific GR degradation for an HSP70 inhibitor. (A) The structure of JG231, a recently developed potent HSP70 inhibitor with excellent pharmacokinetics and antitumor activity. (B) Immunoblot data showing that JG231 degrades GR in MCF7 and MDA MB 231, two human breast cancer cell lines. Cells were treated with the indicated dose for 24 h. (C) Maximum intensity projections of viable mice showing the suppression of radiotracer uptake by JG231 treatment in the supraspinal brown fat depot and the omentum, but not in other tissues with high radiotracer uptake. At right is shown a maximum intensity projection of supraspinal BAT and muscle from a representative mouse in either treatment arm. Mice (n = 5/treatment arm) were treated with JG231 or vehicle once daily for 3 days via intraperitoneal injection. (D) Biodistribution data showing the tissue uptake of ¹⁸F-YJH08 in mice from the treatment study. As predicted by the PET/CT data, JG231 suppressed radiotracer uptake exclusively in the supraspinal brown fat pad. *P < 0.01. The biodistribution data are representative of two independent experiments. (E) Immunoblot data showing GR degradation in brown fat after treatment with JG231. GR levels were not impacted in liver, kidneys, or muscle, as predicted based on the radiotracer biodistribution data.

Conclusions

Here, we report the design, synthesis, and characterization of ¹⁸F-YJH08, a first in class radioligand to measure GR expression noninvasively with PET. In vivo studies showed that ¹⁸F-YJH08 specifically binds to virtually all normal tissues in mice, including tissue compartments like the brain for which the imaging community has long desired biomarkers to study the GR signaling axis. Moreover, the level of GR specific binding in tissues observed with ¹⁸F-YJH08 stands out compared to previously reported biodistribution data for radiotracers targeting GR, including nonsteroidal agonists.²⁴ We further show the radiotracer can be used to study the pharmacology of GR modulators, which led to the unexpected finding of tissue specific anti-GR activity for an inhibitor of HSP70, a protein expressed in virtually every tissue in the body. Overall, ¹⁸F-YJH08 provides an unprecedented multiorgan view of GR expression in real time, and its clinical translation will likely open new frontiers in our understanding of the (patho)biology of GR in humans.

Experimental Section

Radiosynthesis of ¹⁸F-YJH08

To identify optimal conditions for radiofluorination, the following workflow was applied by hand. An aqueous solution of H¹⁸F (20 mCi) was passed through a Sep-Pak light QMA cartridge, which was previously activated with distilled water (3 mL). A ¹⁸F fluoride ion was eluted with 0.5 mL of a K₂CO₃/KOTf solution (0.1 mg mL^–1 and 6 mg mL^–1 in 1:1 CH₃CN/H₂O). The solution was dried at 110 °C under a gentle stream of argon gas and then azeotropically dried twice with anhydrous acetonitrile (1 mL) at 110 °C under a gentle stream of argon. To the reaction vial containing ¹⁸F was added a solution of precursor 4 (2 mg), Cu(Py)₄(OTf)₂ (10.9 mg, 5 equiv), and pyridine (8 μL, 30 equiv) in anhydrous DMF (0.5 mL). The mixture was heated to 110 or 120 °C and stirred for 30 min. The reaction was carefully monitored by radio-TLC for conversion.

On an ELIXYS FLEX/CHEM synthesis module (Sofie Biosciences), ¹⁸F-Boc-YJH08 was prepared in 90 min and then transferred from the ELIXYS on a C18 Sep-Pak cartridge. The ¹⁸F-Boc-YJH08 was eluted from the cartridge using anhydrous CH₃CN and purified by semipreparative HPLC (50% to 95% CH₃CN in water over 12 min and 95% CH₃CN in water for another 8 min) to obtain the desired compound in a decay-corrected radiochemical yield of 12.4 ± 3% (n = 10). The purity of the compound was further verified by reinjection on the semipreparative HPLC.

¹⁸F-Boc-YJH08 was dissolved in a 1.0 mL solution of 50% trifluoroacetic acid in CH₂Cl₂, and the reaction was stirred at 40 °C for 30 min to remove the Boc protecting group. After the reaction was complete on TLC, the reaction vial was then cooled to 0 °C and quenched by adding DMF (0.5 mL), followed by the dropwise addition of 5 M NaOH solution (aq, 0.5 mL). The DCM was removed under reduced pressure, and the mixture was subjected to the semipreparative HPLC (30% to 95% CH₃CN in water over 12 min, then 95% CH₃CN for 8 min) for the purification, which provided the radioactive production in a decay-corrected radiochemical yield of 6.7 ± 1% (n = 10). The purity of the compound was further verified by reinjection of pure compound on semipreparative HPLC.

To prepare the radiotracer for animal studies, ¹⁸F-YJH08 was trapped on the cartridge and eluted with CH₃CN (1 mL). After removing the organic solvent, the pure product was dissolved in an injection formulation of 1/1/8 (v/v/v) DMSO/Tween 80/Saline. The stability, UV purity, and radiochemical purity of ¹⁸F-YJH08 in formulation was further confirmed by analytical HPLC equipped with a gamma ray radio detector and a UV detector at 254 nm. The analytical HPLC condition was as follows: Waters C18 (4.6 mm × 150 mm, 5 μm), mobile phase acetonitrile/H₂O, 8:2 (v/v); flow rate = 1 mL/min. Both the UV and Rad spectrum comigrated with the analytical standard, and the purity of the tracer was ∼97%.

Animal Studies

All animal experiments were conducted under the approval of the Institutional Animal Care and Use Committee (IACUC) at UCSF. Male nu/nu or C57BL6/J mice (4–6 weeks) were purchased from Charles River. All the mice were housed in a dedicated vivarium with free access to food and water. For dexamethasone treatment studies, mice were treated via oral gavage with vehicle (0.5% hydroxy-propyl-methylcellulose and 0.2% Tween 80 in water) or dexamethasone (100 mg/kg or 50 mg/kg). For JG231 treatment studies, mice were treated via intraperitoneal injection of JG231 (4 mg/kg) in injectable formulation (10% DMSO, 18% Tween80, 3.6% dextrose, 68.4% 1 M HEPES) or vehicle. The mice were treated once daily for 3 days. At day 3, the mice received the radiotracer after the final gavage or injection. Adrenalectomized mice were purchased from Charles River and provided drinking water supplemented with NaCl (aq) per instructions until the time of the tracer biodistribution study.

Small Animal PET/CT

Prior to injection, the formulated dose of ¹⁸F-YJH08 was passed through a sterile 0.22 μm membrane. Approximately 300 μCi was injected via tail vein in a volume of 100–150 μL per mouse. After the indicated period of uptake time, the mice were anesthetized with 2–3% isoflurane and imaged with a Siemens Inveon micro PET/CT. For dynamic acquisitions, the mice were anesthetized prior to injection and injected while positioned on the scanner bed. All imaging data were decay corrected, reconstructed, and analyzed with AMIDE software. Maximum intensity projections (MIPs) were generated by AMIDE software. Regions of interest (ROI) were manually placed to calculate SUV_mean data from the dynamic acquisitions.

Biodistribution Studies

Mice were euthanized after radiotracer injection with CO₂ (g) asphyxiation and dissected at dedicated time points post injection. The blood and tissues were removed, washed, dried, and weighed. The activity of each tissue was measured with a gamma counter. All data was decay corrected. PRISM software was used to express a percentage (% ID/g) of the injected dose per gram of tissue.

Autoradiography

Twenty minutes postinjection of ¹⁸F-YJH08, mice were anesthetized and were perfused with cold PBS via cardiac puncture. Tissues were immediately collected and flash frozen in OCT on dry ice. Tissues were sectioned on a microtome at a thickness of 20 μm and immediately mounted on glass slides. The slides were then exposed on a GE phosphor storage screen, and the screen was developed on an Amersham Typhoon 9400 phosphorimager. H&E staining was performed by the Pathology Core Facility at UCSF. The autoradiography images were processed using ImageJ software.

Generation of the Transgenic Mice with Adipocyte Specific GR Knockout

To generate mice with Nr3c1 selectively deleted in mature adipocytes, female GR floxed mice (Jackson Laboratory, cat. no. 021021) were crossed to a male mouse carrying a transgene that expresses the Cre recombinase enzyme (Cre) under the regulation of the adiponectin promoter and enhancer elements (Jackson Laboratory, cat. no. 028020). This adiponectin-Cre (adipo-Cre) mouse line has been validated by a number of laboratories to selectively express Cre in mature adipocytes.²⁵ Progeny from this cross with both adipo-Cre and floxed Nr3c1 had the Nr3c1 gene deleted in their adipocytes.²⁶ Knockout mice from this cross are born in normal Mendelian ratios and have no obvious health issues.

Statistics

All statistical analysis was performed using PRISM v6.0 software. An unpaired, two-tailed Student’s t test was used to determine statistically significant differences in the data. Changes at the 95% confidence level (P < 0.05) were reported as statistically significant. For the determination of K_i, the data were fit with a one site nonlinear regression model.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acschembio.9b01043.

• Additional experimental details, figures, and schemes (PDF)

Author Contributions

Synthesis of YJH08: Y.H., N.Z., Y.W. In vitro affinity assays: Y.H., Y.W., C.T. Radiosynthesis of YJH08: Y.H., N.Z., J.B., H.V.B., D.W. Execution of PET and biodistributions studies: Y.H., N.Z., Y.W., J.W., Y.S. Image reconstruction and analysis: Y.H., Y.S., R.A., S.B., J.V.M. Autoradiography: Y.H., Y.W., J.V.M. Construction of the adipose specific GR KO mouse: M.S., B.F. Synthesis, in vitro, and animal studies with JG231: Y.H., H.S., J.G. Supervised the study: M.E.

The authors declare the following competing financial interest(s): M.J.E. declares an equity stake in ORIC Pharmaceuticals, Inc.