Introduction

The phosphodiesterase enzyme family plays a vital role for degrading cyclic nucleotides (cyclic adenosine monophosphate (cAMP) and cGMP) in the signal transduction pathway in all cells. Phosphodiesterase 4 (PDE4) is mainly distributed in the immune cells, brain, and cardiovascular tissues [1, 2]. Roflumilast is the first PDE4 inhibitor to be approved for the treatment of severe chronic obstructive pulmonary disease [3]. Apremilast, another registered PDE4 inhibitor, has been introduced for the treatment of skin diseases such as psoriatic arthritis [4, 5].

In regard to the central nervous system (CNS), the inhibition of PDE4 has been reported to have effects in various in vivo models of behavior and inflammation [6]. By preventing the cAMP hydrolysis, PDE4 inhibitors are considered to enhance intracellular signal transduction and increase the phosphorylation of cAMP response element-binding protein (CREB), which enhances the transcription of proteins involved in synaptic plasticity and memory formation [7]. Rolipram has been reported to improve cognitive function in animal models of cognitive impairment in neuropsychiatric and neurodegenerative diseases such as schizophrenia and Alzheimer’s disease [8,9,10,11,12,13].

However, the degree of PDE4 inhibition which could be necessary to translate into improvements of cognitive domains has not been fully evaluated in vivo. Therefore, it would be helpful for further understanding and development of PDE4 inhibitors to measure brain target occupancy using clinically available PDE4 inhibitors.

In this positron emission tomography (PET) study, we aimed to investigate in nonhuman primates whether roflumilast entered the brain and demonstrated specific target occupancy of PDE4 using (R)-[11C]rolipram, a PET radioligand for PDE4 [14, 15].

Materials and Methods

Preparation of (R)-[11C] Rolipram

(R)-[11C] rolipram was synthesized as previously reported in detail [16]. Shortly, (R)-[11C]rolipram was synthesized by 11C-methylation of (R)-desmethyl-rolipram in acetone by using [11C]methyl triflate. Purification was performed on a reversed phase HPLC. Radiochemical purity of labeled ligand was more than 99 %.

PET Study

Three female rhesus monkeys (mean weight ± SD, 9.5 ± 2.4 kg, age 8.6 ± 0.1 years old) were examined. The monkeys are owned by the Centre for Psychiatry Research, Department of Clinical Neuroscience, Karolinska Institutet, and housed in the Astrid Fagraeus Laboratory of the Swedish Institute for Infectious Disease Control. The study was approved by the Animal Research Ethics Committee of the Swedish Animal Welfare Agency (Northern Stockholm Region) (N 452/11) and was performed according to the relevant guidelines of the Karolinska Institutet (“Guidelines for Planning, Conducting and Documenting Experimental Research” (Dnr 4820/06-600).

To anesthetize the monkeys during the PET experiments, ketamine hydrochloride (approximately 10 mg/kg) was initially administered intramascularily, and, after endotracheal intubation, a mixture of sevoflurane (2–8 %), O2, and medical air was administered to maintain the anesthesia. The head was immobilized to the PET bed with a fixation device [17].

The HRRT scanner (Siemens) was used to acquire PET data. A 6-min transmission scan using a single Cs-137 source was performed immediately prior to injection of the (R)-[11C]rolipram. PET emission data were acquired for 93 min as list mode just after injection of radioligand (151–178 MBq; specific radioactivity: more than 188 GBq/μmol and injected mass: less than 0. 26 μg). For the reconstruction of PET images, the ordinary Poisson three-dimensional ordered-subset expectation maximization (OSEM) algorithm was used with 10 iterations and 16 subsets, including modeling of the point spread function, after correction for attenuation, random, and scatter. The resolution of the reconstructed PET images was 1.5 mm in full width at half maximum [18]. The frame duration of the PET images was as follows: 10 s × 9, 15 s × 2, 20 s × 3, 30 s × 4, 60 s × 4, 180 s × 4, and 360 s × 12.

Two brain PET measurements were performed, in one day, under the untreated baseline condition, and after intravenous administration of roflumilast, a selective PDE4 inhibitor. Doses of roflumilast were 3.6, 12, 24, 100, and 200 μg/kg. Twelve micrograms per kilogram and 100 μg/kg of roflumilast were administered to monkey 1, 3.6 and 200 μg/kg were administered to monkey 2, and 24 μg/kg was administered to monkey 3. Roflumilast was administered 3 h before the radioligand injection. The duration of the administration was 10 min with syringe pump at the speed of 1 ml/kg. The time between two PET measurements was approximately 4.5 h. The time between two PET measurements with different doses was more than 5 weeks for monkey 1 and monkey 2.

Arterial blood was collected continuously for 3 min using an automated blood-sampling system (ABSS; Allog AB) at a speed of 3 ml/min from arterial catheter in an artery of the lower limb. One arterial blood sample (2 ml) was taken at 5 min before the radioligand injection to measure the protein binding of the radioligand.

To measure the radioactivity in blood and plasma and the fraction of the metabolite, arterial blood samples (1–2 ml) were drawn at 4, 15, 30, 45, 60, and 90 min after the injection of the radioligand.

Venous blood samples were taken at − 1, 10, 20, 40, 65, and 91 min after the radioligand injection to obtain the plasma concentration of roflumilast and a metabolite, N-oxidate roflumilast.

Metabolite Analysis of (R)-[11C]Rolipram

The amount of unchanged (R)-[11C]rolipram and its radioactive metabolites in monkey plasma was measured using a reversed-phase radio-HPLC method. After centrifugation of blood at 2000×g for 2 min, the plasma was mixed with acetonitrile (1:1.4). Then, the mixture was centrifuged at 2000×g for 4 min, and the supernatant of the mixture was injected into a HPLC system coupled to an online radioactivity detector. The used radio-HPLC system consisted of an interface module (D-7000; Hitachi: Tokyo, Japan), a L-7100 pump (Hitachi), an injector (model 7125, 5.0 ml loop; Rheodyne: Cotati, USA), and an ultraviolet absorption detector (L-7400, 254 nm; Hitachi) in series with a 150TR Packard radioactivity detector (housed in a shield of 50 mm thick lead and equipped with a 550-l flow cell). Chromatographic separation was made on a XBridge C18 column, (50 mm × 10 mm I.D., 2.5 μm + 10 mm × 10 mm I.D., 5 μm; Waters: New England, USA) using gradient elution. Acetonitrile (A) and 20 mM ammonium phosphate (pH 7) (B) were used as the mobile phase at 6.0 ml/min, according to the following program: 0–3.5 min, (A/B) 25:75 → 55:45 v/v; 3.5–4.0 min, (A/B) 55:45 v/v; 4.0–4.1 min, (A/B) 55:45 → 25:75 v/v; 4.1–5.0 min, and (A/B) 25:75 v/v. Peaks for radioactive compounds that were eluted from the column were integrated, and their areas were expressed as a percentage of the sum of the areas detected radioactive compounds (decay-corrected to the time of injection on the HPLC).

Measurement of Free Fraction of (R)-[11C]Rolipram

The free fraction, fp, of (R)-[11C]rolipram in monkey plasma was estimated using an ultrafiltration method. Initially, plasma (500 μl) or phosphate buffered saline solution (500 μl) as a control were mixed with (R)-[11C]rolipram (50 μl, ~ 1 MBq) and then incubated at room temperature for 10 min. After the incubation, 200 μl portions of the incubation mixtures were pipetted into ultrafiltration tubes (Centrifree YM-30, Millipore). Then, they were centrifuged at 1500×g for 15 min. Radioactivity in equal aliquots (20 μl) of the ultrafiltrate (Cfree) and of the plasma (Ctotal) was counted using a NaI well counter. Each determination was performed in triplicate. The free fraction was then calculated as fp = Cfree / Ctotal, and after the correction for the membrane binding measured with the control samples, the final results were obtained.tpb

MRI Measurements

An MRI system GE 1.5 T Sigma unit (Milwaukee, WI, USA) was used to obtain T1-weighted MR images. A 3-D SPGR protocol with the following settings was used as the T1 sequence: repetition time (TR) 21 ms, flip angle 35°, FOV 12.8, and matrix 256 × 256 × 128, 128 × 1.0 mm slices.

PET Data Analysis

Regions of interest (ROIs) were delineated manually on the putamen, caudate, hippocampus, amygdale, thalamus, pons, frontal cortex, temporal cortex, anterior cingulate cortex, occipital cortex, parietal cortex, and cerebellum on the co-registered MRI/PET images. Regional uptake was expressed as percentage of standard uptake value (%SUV), which equals uptake (MBq/ml)/injected radioactivity (MBq) × body weight (g) × 100.

The total distribution volume (VT) and VT/fp of (R)-[11C]rolipram was calculated with two-tissue compartment models using metabolite-corrected plasma input [19, 20]. Because there were no appropriate regions for the reference region, Lassen plot approach was applied for the current data to estimate the target occupancy and the distribution volume (VND) of free and nonspecifically bound radioligand [21, 22].

The relationship between the target occupancy and the plasma concentration of the drug (roflumilast, N-oxide roflumilast, and the sum of roflumilast and N-oxide roflumilast) was examined using the following hyperbolic function.

$$ \mathrm{Occupancy}\ \left(\%\right)={\mathrm{Occ}}_{\mathrm{max}}\times \mathrm{Cp}/\left(\mathrm{Kd}+\mathrm{Cp}\right) $$

where Cp is the concentration of the drug, and Kd is the plasma concentration required to produce 50 % of the maximal target occupancy. The area under the curve (AUC) during the 90 min PET measurements divided by the time length of 90 min was used as the plasma level of the drug at the PET measurements.

Occmax was evaluated in two ways. One was a priori fixed to 100 % while the other was estimated by fitting data.

Results

Summed PET images at baseline and after administration of 200 μg/kg of roflumilast are shown in Fig. 1. There was high accumulation of the radioligand in the thalamus and striatum (caudate and putamen) and moderate accumulation in the cortex at baseline. After administration of roflumilast, accumulation was decreased in all the brain regions.

Fig. 1.
figure 1

PET images of (R)-[11C] rolipram summed from 15 to 90 min. a Baseline. b After administration of 200 μg/kg of roflumilast.

Full size image

Time activity curves of the brain regions at baseline and after administration of 200 μg/kg of roflumilast are shown in Fig. 2. Compared with the baseline, initial brain uptake of (R)-[11C]rolipram was a little higher after administration of roflumilast, but the washout from the brain was faster. For the cerebellum, the peak value was 534 %SUV and the time to decrease to half was approximately 16.5 min at the baseline while the peak value was 585 % SUV and the time to decrease to half was approximately 8 min after roflumilast administration.

Fig. 2.
figure 2

Brain regional time activity curves of (R)-[11C] rolipram at a baseline and b after 200 μg of roflumilast. CB cerebellum, CAU caudate, PUT putamen, HIP hippocampus, AMY amygdala, THA; thalamus, FC frontal cortex, TC temporal cortex, ACC anterior cingulated cortex, OC occipital cortex, PC parietal cortex. Percentage of standard uptake value (%SUV) equals uptake (MBq/ml)/injected radioactivity (MBq) × body weight (g) × 100.

Full size image

The parent fraction of (R)-[11C]rolipram in the plasma decreased to 32.3 ± 6.4 % at 15 min and 14.9 ± 0.1 % at 90 min, and the metabolite rate did not change significantly after roflumilast treatment. There were no radioactive metabolites that were more lipophilic than (R)-[11C]rolipram. Free fraction of the radioligand was 39.5 ± 6.0 % at baseline and 44.7 ± 3.6 % after roflumilast administration. There was no correlation between the change of the free fraction and roflumilast concentration.

The calculated VT and VT/fp and the percent change of one monkey for whom one baseline PET and post-treatment PET with 200 μg/kg of roflumilast are shown in Table 1. All the brain regions showed the decrease of the VT and VT/fp. There was no brain region which can be used as reference region for simplified quantification.

Table 1. VT and VT/fp calculated by two tissue compartment model and the percent change in a monkey at the baseline and after 200 μg/kg of roflumilast i.v. administration
Full size table

Lassen plots of PET data after 200 μg/kg of roflumilast administration are shown in Fig. 3. The occupancy at 200 μg/kg of roflumilast was 43.5 % calculated using VT and 45.5 % calculated using VT/fp. Estimated PDE4 occupancy and the dose and mean plasma concentration of roflumilast at PET measurements are shown in Table 2. Regardless of the outcome measures of VT and VT/fp, PDE4 occupancy was lower than 50 % when up to 200 μg/kg of rofulmilast was administered. Plasma concentrations during drug treatment PET measurements for each monkey are shown in Suppl. Fig. 1 (see ESM). The plasma concentrations of roflumilast and N-oxide roflumilast were fairly constant during PET measurements.

Fig. 3.
figure 3

The occupancy estimation using the Lassen plot. The slope corresponds to the occupancy and the intersection with x axis corresponds to VND. aVT is used as the outcome measures. The equation of the linear fitting: Y = 0.4353X − 0.7409. bVT/fp is used as the outcome measure. The equation of the linear fitting: Y = 0.4548X − 1.8163.

Full size image
Table 2. PDE4 occupancy by roflumilast calculated with two different outcome measures (VT and VT/fp)
Full size table

The relationship between PDE4 occupancy and the plasma concentration of the drug at the PET measurement is shown in Fig. 4 and Suppl. Fig. 2. The curve fitting was better when the target occupancy was calculated using VT. Curve fitting was better when maximal occupancy was estimated by the model than when the occupancy was fixed to 100 %. The estimated Kd values and estimated maximal occupancy are summarized in Table 3. The estimated maximal occupancy was 53.1 to 57.4 %.

Fig. 4.
figure 4

The relationship between PDE4 occupancy and the plasma concentration of roflumilast. a The relationship based on the target occupancy using VT. b The relationship based on the target occupancy using VT/fp. Dotted lines show the hyperbola where the maximal occupancy was fixed to 100 %. Solid lines show the hyperbola where the maximal occupancy was estimated by the fitting. Kd values and estimated Occmax values are shown in Table 3.

Full size image
Table 3. Kd values and estimated Occumax of the plotted data
Full size table

Discussion

This PET study showed that roflumilast, a PDE4 inhibitor used for the treatment of COPD, enters the brain and binds to PDE4 in a dose-dependent manner in nonhuman primate brain. However, PDE4 occupancy in the brain was lower than 50 % even when up to 200 μg/kg of rofulmilast was administered.

Roflumilast is the first PDE4 inhibitor to be licensed for the treatment of COPD and was approved in Europe in 2010 and in the USA in 2011 as an oral add-on treatment of COPD. Roflumilast is metabolized principally via the liver into N-oxide roflumilast, which is pharmacologically active with a lower potency but longer half-life than the parent compound [23, 24]. Roflumiast was reported to have no PDE4 subtype selectivity apart from PDE4C while N-oxide rofulmilast has no selectivity for PDE4 subtypes [25]. Therefore, both roflumilast and N-oxide roflumilat have potentials to contribute to PDE4 occupancy.

In this study, roflumilast was administered 3 h before the PET radioligand administration in order to imitate the relationship between roflumilast and N-oxide roflumilast in clinical settings based on internal data (data were not shown).

N-oxide roflumilast was detected under all the drug treatment conditions in this study, and the ratio of the plasma concentration between roflumilast and N-oxide roflumilast was rather constant during PET measurements under all the drug treatment conditions (Suppl. Table 1). Time-dependent different contributions of roflumilast and N-oxide roflumilast to the target binding could not be evaluated due to the constant ratio between them.

In this study, the doses of roflumilast were selected in order to imitate the plasma concentration in the clinical setting. The plasma concentration of roflumilast and N-oxide roflumilast after oral administration of 500 μg of roflumilast, the recommended clinical dose, in human subjects has been reported to be approximately 5 and 9 ng/ml at peak concentration, respectively [23]. As shown in Table 2, the doses up to 200 μg/kg of rofulmilast covered the range of the plasma concentration in the clinical settings. Although we have to take into consideration the species difference between the human and NHPs, the estimated PDE4 occupancy of 500 μg of roflumilast would be up to approximately 30–40 % if we apply the human plasma data to the present NHP results of the relationship between the occupancy and the plasma concentration.

The relationship between the target occupancy and plasma concentration suggested that the plotted data fitted better when the maximal occupancy were not fixed as shown in Fig. 4. On the other hand, [3H]rolipram binding was reported to be fully blocked by roflumilast [26]. (R)-[11C]rolipram rat PET study that the brain uptake of (R)-[11C]rolipram was almost completely blocked by cold rolipram [27]. Considering non-PDE4 subtype selectivity of both rolipram and roflumilast [24, 28], full blocking of (R)-[11C]rolipram would be expected by roflumilast administration. Higher doses of roflumilast than what we tested in this study would be needed to investigate in order to confirm whether in vivo maximal target occupancy with (R)-[11C]rolipram can reach 100 % or not.

As outcome measures, VT and VT/fp of (R)-[11C]rolipram were calculated in this study because VT and VT/fp of (R)-[11C]rolipram were reported to have comparable reproducibility [19], and there was possible change of fp after roflumilast administration. The relationship between the roflumilast and the PDE4 occupancy was weaker when VT/fp was used to calculate the target occupancy than VT was used as shown in Fig. 4. In theory, as only free (R)-[11C]rolipram is considered to enter the brain, correction for fp (VT/fp) should more accurately reflect binding than that without correction (VT). However, the inclusion of fp in the calculation may introduce additional variability associated with measurement of protein binding. In this study, fp values at the baseline in the same monkeys (monkey 1 and monkey 2) showed rather high variability at different study days (Suppl. Fig. 3). High variability might arrive from technical issues with measurement of protein binding as well as potential physiological conditions in monkeys under anesthesia of induction by ketamine and maintenance by sevoflurane. There was no clear dose-dependent change of fp before and after roflumilast administration although the limited number of the data sets and high variability of fp at the baseline condition might obscure the relation (Suppl. Fig. 3). The target occupancy estimated by VT/fp was influenced by fp values directly. Considering high variability of fp values at the baseline condition and no clear trend of fp changes before and after roflumilast administration, the target occupancy estimated by VT/fp would be overestimated or underestimated easily. Therefore, the outcome measures without fp were considered to be more robust in this study.

Compared with other PDE4 inhibitors whose development was stopped due to side effects of nausea and emesis, side effects are typically mild to moderate with roflumilast [29]. Emesis induced by PDE4 inhibitors was reported to be related to the degree of PDE4 inhibition based on the ex vivo evaluation of [3H]rolipram binding in the brain although roflumilast was not included in the study [30]. Although species difference is needed to consider, the present study may suggest that 30–40 % of in vivo brain PDE4 occupancy, which corresponds to clinical dose of roflumilast, might be an acceptable level of PDE4 inhibition not to induce severe nausea and emesis if the in vivo brain target occupancy can be used as an objective parameter. In order to confirm it, direct evaluation between the brain occupancy of PDE4 inhibitors and occurrence of emesis would be necessary in human subjects.

In a recent meta-analysis of roflumilast clinical data [31], 2–5 % of the patients who took roflumiast showed potentially CNS-related side effects such as insomnia, dizziness, and headache. This may support our results of brain penetration of roflumilast.

There has been no clinical data to show the improvement of cognition by PDE4 inhibitors in clinical settings although several lines of research using animal models suggest that PDE4 inhibition could improve some cognitive domains such as executive functions and memory [32, 33]. Regarding PDE4 subtypes, PDE4D is considered to be related to cognitive improvement [34, 35] while PDE4B is also considered to be involved in the cognitive function [36]. It would be worth investigating in future the relationship among treatment with roflumilast, PDE4 occupancy, and effects on cognitive domains in preclinical and clinical settings.

Rolipram binds nonselectively to all PDE4 subtypes (Bruno et al., 2009). As the subtypes differ with respect to their regulatory behavior and tissue expression patterns, PET radioligands with more selective to the subtypes of PDE4 would help to understand the function of PDE4 subtypes in future.

Each dose of roflumilast was investigated only in one monkey. Investigation of the same dose in multiple monkeys would help to evaluate more detailed Kd values.

Conclusions

This PET study showed a certain level of in vivo brain PDE4 occupancy after administration of clinically relevant doses of roflumilast, in nonhuman primates. The results would facilitate investigating the relationship between brain PDE4 occupancy and putative procognitive properties of PDE4 inhibitors in human subjects.