[11C]UCB-A, a novel PET tracer for synaptic vesicle protein 2 A

Nuclear Medicine and Biology 43 (2016) 325–332

Contents lists available at ScienceDirect

Nuclear Medicine and Biology journal homepage: www.elsevier.com/locate/nucmedbio

[ 11C]UCB-A, a novel PET tracer for synaptic vesicle protein 2 A Sergio Estrada a,1, Mark Lubberink b,c,1, Alf Thibblin b,d, Margareta Sprycha d, Tim Buchanan e, Nathalie Mestdagh e, Benoit Kenda e, Joel Mercier e, Laurent Provins e, Michel Gillard e, Dominique Tytgat e,2, Gunnar Antoni a,d,⁎ a

Preclinical PET platform, Uppsala University, Uppsala, Sweden Nuclear Medicine & PET, Uppsala University, Uppsala, Sweden Medical Physics, Uppsala University Hospital, Uppsala, Sweden d PET Centre, Uppsala University Hospital, Uppsala, Sweden e UCB Pharma, Brussels, Belgium b c

a r t i c l e

i n f o

Article history: Received 13 November 2015 Received in revised form 19 February 2016 Accepted 15 March 2016 Available online xxxx Keywords: SV2A Epilepsy [11C]UCB-A Preclinical PET

a b s t r a c t Introduction: Development of a selective and specific high affinity PET tracer, [11C]UCB-A, for the in vivo study of SV2A expression in humans. Radiochemistry and preclinical studies in rats and pigs including development of a tracer kinetic model to determine VT. A method for the measurement of percent intact tracer in plasma was developed and the radiation dosimetry was determined in rats. Results: 3–5 GBq of [11C]UCB-A could be produced with radiochemical purity exceeding 98% with a specific radioactivity of around 65 GBq/μmol. In vitro binding showed high selective binding towards SV2A. [11C]UCB-A displayed a dose-dependent and reversible binding to SV2A as measured with PET in rats and pigs and the VT could be determined by Logan analysis. The dosimetry was favorable and low enough to allow multiple administrations of [11C] UCB-A to healthy volunteers, and the metabolite analysis showed no sign of labeled metabolites in brain. Conclusions: We have developed the novel PET tracer, [11C]UCB-A, that can be used to measure SV2A expression in vivo. The dosimetry allows up to 5 administrations of 400 MBq of [11C]UCB-A in humans. Apart from measuring drug occupancy, as we have shown, the tracer can potentially be used to compare SV2A expression between individuals because of the rather narrow range of baseline VT values. This will have to be further validated in human studies. © 2016 Elsevier Inc. All rights reserved.

1. Introduction Epilepsy, with an incidence estimated of 1.5 to 5% of the population, is still a clinical challenge and at the same time one of the most common neurological diseases [1]. Although a number of drugs with different mechanisms of action are available for the treatment of seizures, around 30% of patients do not have seizure control on their current medication and these patients with refractory epilepsy represent a severe unmet clinical need [2,3]. Pharmacotherapy is directed towards maintaining the patient seizure-free by controlling the pathological electric activity in the brain which is the typical hallmark of epilepsy. There are several hypotheses for the reason of pharmacoresistant epilepsy but so far, both the etiology and mechanism are unknown, although efflux systems such as P-glycoprotein (Pgp), a member of the ATP-binding cassette transporters limiting drug entry into brain, have been the focus for ⁎ Corresponding author at: Uppsala University, Preclinical PET Platform, Department of Medicinal Chemistry, Box 574, SE-751 23, Uppsala, Sweden. Tel.: +46 18 611 0640; fax: +46 18 611 0619. E-mail address: [email protected] (G. Antoni). 1 Both authors contributed equally. 2 Currently employed by Sanofi-Aventis Deutschland GmbH, Frankfurt am Main, Germany. http://dx.doi.org/10.1016/j.nucmedbio.2016.03.004 0969-8051/© 2016 Elsevier Inc. All rights reserved.

many research studies [3]. There is, however, an ongoing debate as to whether Pgp plays a role in treatment resistant epilepsy. One potential explanation for the controversy could be that local, rather than, global up-regulation of Pgp occurs, however, current diagnostic imaging tools have not been able to quantify this. Other efflux systems than Pgp could also be involved. One argument against the Pgp hypothesis is the fact that many antiepileptic drugs (AEDs) are not Pgp substrates or inhibitors [4]. Another hypothesis of AED therapy resistance involves the expression of specific drug-related molecular targets. One such drug-related molecular target is synaptic vesicle protein 2 A (SV2A) which is a 12-transmembrane glycoprotein present on synaptic vesicles of all neural cells [5]. Although its exact molecular function is still unclear, the current hypothesis is that SV2A plays a role in the exocytosis of neurotransmitters and acts as a modulator of vesicle fusion [6]. The importance of SV2A has been demonstrated in knockout mice which suffer from spontaneous seizures from birth and typically die within 3 weeks of age [7]. It has been speculated that SV2A is associated with action potential-dependent synaptic transmission and therefore reduced SV2A expression could lead to both increased excitability and higher propensity for seizures. The phenotype of homozygous SV2A knock-out mice, being characterized by uncontrolled spontaneous

326

S. Estrada et al. / Nuclear Medicine and Biology 43 (2016) 325–332

NX C18 5 μm 150 × 10 mm column or a Reprosil PUR Basic C18 5 μm 150 × 10 mm using a mixture of 8 mM aqueous ammonium carbonate buffer and acetonitrile as the mobile phase with a flow rate of 10 mL/min. The product identification and purity check was done by analytical reversed-phase HPLC utilizing a Phenomenex Gemini NX C18 5 μm 100 × 3 mm column, also with ammonium carbonate and acetonitrile as the mobile phase, and a flow rate of 1.5 mL/min. Both HPLC systems were equipped with UV- and radioactivity detectors in series. The compounds 1–5 (Fig. 1) used to perform the in vitro and in vivo blocking/ displacement studies as well as the UCB-A precursor 6 (Scheme 1) have been synthetized by UCB from commercially available reagents according to described procedures [12,16–18].

seizures leading to premature death, suggests an agonistic or modulating action of SV2A-binding antiepileptic drugs, such as levetiracetam. Studies in brain tissue from patients that died from status epilepticus or from pharmacoresistant temporal lobe epilepsy patients showed that they had a reduced SV2A expression in hippocampus [8] compared to levels measured in the brain from nonepileptic patients. SV2A has been identified as the unique brain-specific protein binding site of the antiepileptic drug levetiracetam, therefore suggesting a major role played by SV2A in its mechanism of action [9–11]. Even though SV2A has been identified as the brain binding site of levetiracetam, the precise molecular mechanism(s) by which this binding protects against seizures is not well understood at present. In order to further investigate the role of SV2A in epilepsy, a non-invasive imaging method such as positron emission tomography (PET) and an SV2Aselective tracer would thus be of significant value. PET is a powerful nuclear imaging technique allowing the in vivo quantification of a radioligand labeled with a short-lived positron-emitting isotope, typically carbon-11 or fluorine-18. The opportunity of quantifying the SV2A expression in patients with focal epilepsy, for example temporal lobe epilepsy, and to measure the extent of SV2A occupancy resulting in a therapeutic effect would give further insight into the mechanism of action of levetiracetam and other SV2A modulating antiepileptic drugs. In a recent publication Mercier et al. presented three potential PET tracer candidates developed from the pharmacophore of levetiracetam, coined UCB-A, UCB-H and UCB-J [12]. Recently two radioligands for the study of SV2A were presented, [ 18F]UCB-H and [ 11C]levetiracetam [13–15]. Here, we present the development and preclinical evaluation, including tracer kinetic modeling, of [11C]UCB-A, as a selective SV2A PET tracer. Following development of a labeling synthesis the potential of [11C] UCB-A as an in vivo biomarker for quantification of regional SV2A expression was investigated in rats and in pigs using PET. Receptor kinetics were studied in an in vivo displacement study in rats using PET. Dosimetry was determined by ex vivo biodistribution studies in rats and extrapolated to give estimates of effective radiation doses in humans. Binding potential and specific binding was assessed in pigs by pre-treatment with brivaracetam, a selective SV2A ligand with antiepileptic properties. A tracer kinetic model, based on a metabolitecorrected input function from arterial blood, was applied to the data to create values of brain SV2A binding potential.

2.1.2. Synthesis and purification of [ 11C]UCB-A A Scanditronix MC17 cyclotron was used to produce 11C via the 14 N(p,α) 11C nuclear reaction. The target was filled with a mixture of 0.05% oxygen in nitrogen and bombarded with 17 MeV protons to produce [ 11C]CO2. The labeled carbon dioxide was reduced with 0.2 M lithium aluminium hydride in tetrahydrofuran. Treatment with 56% aqueous hydroiodic acid gave [ 11C]methyl iodide which was transferred in a stream of nitrogen through a column with phosphorus pentoxide (Sicapent ®) followed by a glass column with silver triflate at 200 °C. The produced [ 11C]methyl triflate was used for the synthesis of [ 11C] UCB-A according to Scheme 1. The [11C]methyl triflate was transferred in nitrogen gas to a solution of the UCB-A precursor 6 (3.0 mg, 10 μmol) in 250 μL of 2-pentanone. The reaction solution was then heated for 1 min at 100 °C, diluted with 350 μL of a solution of trifluoroacetic acid (TFA) in 35% ethanol (1 mL TFA/10 mL) and transferred to another vial. The protecting group was hydrolysed and the solvent was removed with a vigorous stream of nitrogen at 100 °C for 3 min. The residue was redissolved in 1.5 ml of 35% ethanol and injected onto the preparative HPLC column and the fraction containing [ 11C]UCB-A (compound 7) collected into a vial with 0.5 mL of hydroxypropyl-β-cyclodextrin (300 mg/mL in saline). The mobile phase was removed by means of a rotary evaporator and the residue was redissolved in 5 mL of sterile phosphate buffer solution (0.1 M pH 7.4). The formulated tracer was filtered through a sterile 0.2 μm filter. The product was identified by analytical HPLC using co-injection of the non-radioactive reference compound by comparing the retention times of the UV and radioactive peaks. The total production time including formulation was about 40 min. A more rigorous identification of the tracer was done by LC/MS.

2. Material and methods 2.1. Chemistry

2.1.3. Metabolite analysis A method was developed for separation of [ 11C]UCB-A from its radioactive metabolites in plasma. [ 11C]UCB-A (~50 MBq) was injected to male, Sprague–Dawley rats. They were euthanized after 5 and 40 min respectively and blood and

2.1.1. Equipment The radiosynthesis was carried out on an in-house built Synthia robot system based upon a Gilson ASPEC module. The semi-preparative HPLC purification of the tracer was carried out on a Phenomenex Gemini

F F

F F

O N

O N

NH2 O

O

F O

N NH2 O

N

NH2 O N N H

levetiracetam (1) brivaracetam (2) pKi = 5.5

pKi = 6.9

CN

seletracetam (3) compound 4 pKi =7.5

pKi = 8.2

compound 5 pKi = 8.2

Fig. 1. SV2A reference compounds used in blocking/displacement studies. pKi for human SV2A measured at 4 °C. Structure of compound 5 not shown but will be published separately by UCB.

S. Estrada et al. / Nuclear Medicine and Biology 43 (2016) 325–332 F

Tris buffer, 3 × 1 min, slides were dried and exposed to phosphoimager plates. Within the resulting images, regions of interests (ROI) around whole brain sections were drawn manually, and the average pixel values in these ROIs were calculated. Residual [ 11C]UCB-A binding was calculated, which was defined as the percentage of remaining binding at each concentration of competing ligand. The mean residual binding was calculated and plotted against inhibitor concentration, data were fitted to a simple “one-site competition binding” equation (Hill slope set to 1) with GraphPad Prism (GraphPad Software, Inc., La Jolla, CA, USA) and IC50 values were obtained for the two SV2A ligands.

F

F O F

N 1.

N

[11C]methyl

triflate

O N

2. TFA

11CH 3

N N N

(+)-7[11C]UCB-A

(+)-6 Scheme 1. Synthesis of [11C]UCB-A.

brain were removed. The blood samples were centrifuged at 3000× g for 2 min at 4 °C. From the plasma fraction, two samples of 0.6 mL each were transferred to new tubes and an equal volume of acetonitrile was added to precipitate the proteins. The mixture was centrifuged at 16,000× g at 4 °C for 1 min. The supernatant was filtered through a 0.2 μm nylon membrane (Corning Incorporated, Corning, NY, USA) by centrifugation at 16,000× g at 4 °C for 1 min. In order to improve the peak shape, 1.5 mL filtrated plasma was diluted with 0.5 mL water. 20 μL 1 mg/mL UCB-A was added to the mixture. The sample preparation recovery was determined by measuring the radioactivity in the plasma, filters and the pellet. The brain tissue was cut in two pieces. Their weights were determined (in gram) and an equal amount of acetonitrile was added. The mixture was homogenized and then centrifuged. The supernatant was removed from the pellet and filtered as described above. 1 mL water and 20 μL 1 mg/mL UCB-A were added to 1 mL filtrated supernatant. 1.8 mL of the sample was injected onto the HPLC column (ASPEC Gilson, Middleton, USA) monitoring at 260 nm and with radiodetection. Four fractions were collected and the radioactivity in the fractions was measured by a well-type scintillation counter.

2.2. Rat studies 2.2.1. Frozen section autoradiography Frozen coronal sections were prepared (25 μm) from rat brain (male, Sprague–Dawley), placed onto glass slides (SuperFrost ® Plus, MenzelGläser, Germany) and kept at − 20 °C until use. The slides were pre-incubated in 50 mM Tris–HCl, pH 7.4 (Tris buffer) for 10 min and then pre-incubated with various concentrations (0125, 12.5, 125, and 1250 nM) of two different SV2A ligands (levetiracetam and seletracetam) in Tris buffer at room temperature for 30 min. Thereafter, [ 11C]UCB-A, at 0.03 and 0.1 MBq/mL, respectively, was added to the assays for an additional 40 min of incubation. After washing in ice-cold

A

2.2.2. Biodistribution and dosimetry Five male (309 ± 9 g) and five female (241 ± 9 g) Sprague–Dawley rats were injected in the tail vein with [ 11C]UCB-A, 7.2 ± 1.8 MBq (male) and 7.7 ± 0.2 MBq (female) in ~ 450 μL saline. One rat of each gender was sacrificed after 2, 5, 10, 20, and 40 min, respectively. Blood and 16 other organs were immediately extracted, their radioactivity measured in a well counter of NaI-type and the weight of the extracted tissues was determined. Organ radioactivity distribution from rat was extrapolated to human, assuming that similar standardized uptake values (SUV normalized to body mass) between species and residence times were calculated, which were used as an input for the OLINDA/ EXM package (Version 1.1, Vanderbilt University, USA, 2007 [19]) for estimation of organ absorbed doses and effective radiation dose. 2.2.3. PET binding study Five male Sprague–Dawley rats were anesthetized with isoflurane (2–3%) in a 50/50% mixture of oxygen and air and placed on the temperate bed of a small animal PET/CT scanner (Triumph trimodality, GE Healthcare). They were positioned with the brain in the field of view. After a CT scan for attenuation correction purposes, [ 11C]UCB-A (~10 MBq) was administered as a bolus injection in the tail vein and a dynamic PET scan was conducted for 90 min at baseline. Subsequently, 10 mg/kg of compound 4 was injected intravenously and after 15 min a second PET scan was conducted after a second injection of [11C]UCB-A (~ 10 MBq). The data were reconstructed using Maximum Likelihood Estimation Maximization (10 iterations). A whole-brain volume of interest (VOI) was drawn manually in the averaged images and transferred to all frames to generate time-activity curves. Four additional male Sprague–Dawley rats were anesthetized with isoflurane (approximately 3%) in a 50/50% mixture of oxygen and air and placed on the bed of a small animal PET scanner (microPET R4, Concorde Microsystems, Knoxville). Rats were positioned with the brain in the field of view. A transmission scan with rotating 57Co source was used to correct the emission scan for the attenuation of 511 keV photons through the tissue and scanner bed. [11C]UCB-A (~10 MBq) was administered as a bolus injection in the tail vein and a dynamic PET scan was conducted for a total of 90 min. Mid-way through (45 min) the PET scan, brivaracetam, 21 mg/kg, was administered as a short intravenous bolus,

B 4000

50000 40000

Counts (CPM)

Counts (CPM)

327

30000 20000 10000 0

3000 2000 1000 0

0

2

4

6

8

Retention time (min)

10

0

2

4

6

8

10

Retention time (min)

Fig. 2. HPLC-radio detector analysis of plasma from rat removed 5 min (A) and 40 min (B) after injection of [11C]UCB-A with a mobile phase composition of ammonium carbonate buffer 10 mM pH 8.7: methanol (45:55, v/v).

328

S. Estrada et al. / Nuclear Medicine and Biology 43 (2016) 325–332

filter, as implemented by the scanner manufacturer, applying all appropriate corrections for normalization, dead time, and random and scatter coincidences [21]. For computation of parametric images, images were reconstructed using in-house written OSEM (4 iterations, 8 subsets).

Fig. 3. Results from the quantification of frozen section autoradiography assay. [11C]UCB-A was incubated with increasing concentrations of levetiracetam and seletracetam (n = 2). Residual binding was plotted as against the concentration of the competing ligands and data were fitted to a simple competition equation. IC50 values were 0.78 μM and 0.14 μM for levetiracetam and seletracetam, respectively.

to initiate displacement. Images were reconstructed using Fourier rebinning and 2D Ordered Subsets Expectation Maximization (OSEM; 4 iterations, 16 subsets). Whole-brain VOIs were manually drawn on the PET images, and time-activity curves (TACs) were generated. 2.3. Pig studies 2.3.1. Scan protocol Six locally bred pigs (weight 20.3 ± 0.7 kg) underwent two or three 90 min dynamic PET scans (33 frames of increasing duration) after injection of 176 ± 50 MBq [ 11C]UCB-A on a Hamamatsu SHR-7700 animal PET scanner (Hamamatsu, Hamakita City, Japan) [20]. Two pigs underwent two baseline scans to assess reproducibility, two pigs underwent one scan at baseline and one after administration of compound 5, and the remaining two pigs underwent a baseline scan and two scans after different doses of compound 5. Drug concentrations were 6, 12, 30, 60, 100 and 300 μg/kg, administered as a 15 min i.v. infusion prior to injection of the PET tracer. Attenuation correction was based on a 30 min transmission scan with a rotating 68Ge rod source. Images were reconstructed using filtered back projection using a 4 mm Hann

A

2.3.2. Blood data Arterial blood radioactivity was monitored on-line at a rate of 2 mL/min during the first 10 min post injection using an automatic blood sampling system (Allogg, Sweden), and additional samples were taken at 2, 5, 10, 15, 20, 30, 40, 60, 75 and 90 min post injection for measurement of whole blood and plasma concentrations. Metabolite analysis was performed as described above on samples taken at 10, 40 and 75 min p.i. The on-line arterial blood curve was calibrated using the discrete samples taken at 2, 5 and 10 min and then combined with a multiexponential fit to the discrete sample whole blood concentrations resulting in a whole-blood time-activity curve (TAC). A plasma TAC was obtained by multiplication of this whole-blood TAC with the mean plasma/whole blood activity ratio of all samples. The metabolitecorrected arterial input function was then computed by multiplication of the plasma TAC with a sigmoid fit to the measured parent fractions. 2.3.3. Data analysis Volumes of interest were manually drawn over whole cortex and cerebellum using VOIager software (GE Healthcare, Uppsala) in 0–90 min sum images and projected onto the dynamic images. Cortex and cerebellum TACs were analyzed using Logan graphical analysis [22] as well as non-linear regression (NLR) to the operational equations of single (1 T) and two-tissue (2 T) compartment models including fitted blood volume components, giving the total volume of distribution (VT) and, in the case of the two-tissue compartment model, the binding potential BPND. In addition, parametric Logan VT images were computed. Fits were performed using in-house written software in Matlab. The optimal compartment model was assessed using the Akaike criterion [23]. Occupancy was calculated using the Lassen method [24], since no reference region devoid of SV2A is available. The assumption was made that VND was identical across scans and animals because of the limited amount of data. Occupancy values were then estimated using a two-step method: VND was varied in steps of 0.1 between 0 and 10 mL/

15

2 5 10 20 40

SUV

10

5

0

B

15

SUV

10

5

es B la dd er Te st es M us cl e B on e B ra in

LI -

ec

Fa

SI +

IS-

le e A dr n en al s K id ne y

re as

Sp

ve r

nc

ng

Pa

Li

Lu

ea rt H

B

lo

od

0

Fig. 4. Biodistribution of [11C]UCB-A in rat organs as a function of time. A; male rats and B; female rats. Scale denotes time in min between dosing and animal sacrifice. LLI wall: lower large intestine wall; ULI: upper large intestine wall; SI: small intestine; LI: large intestine: +/−: with and without contents, respectively.

S. Estrada et al. / Nuclear Medicine and Biology 43 (2016) 325–332

higher than 98% and with a specific radioactivity of 63 ± 30 GBq/μmol (n = 19, average ± SD). The chemical purity, defined as absence of precursor or other UV absorbing substances, was high, with no traces of the corresponding detritylated UCB-A precursor 6.

Table 1 Residence times and organ doses. Organ

Residence time, male

(min), Female

Absorbed dose, male

(μGy/MBq), Female

Adrenals Brain Breasts Gallbladder wall LI wall LLI wall Small intestine SI wall Stomach wall ULI wall Heart wall Kidneys Liver Lungs Muscle Gonads Pancreas Red marrow Bone Osteogenic cells Skin Spleen Thymus Thyroid Urinary bladder wall Uterus Effective dose (μSv/MBq)

0.05 0.76

0.06 0.89

15.0 3.3 NA 3.6

18.7 4.4 2.2 5.3

0.13

0.19

1.9 0.73

1.0 0.68

3.8 2.6

4.8 2.9

2.5 3.5 3.5 9.6 12.5 4.4 1.9 1.6 3.4 2.3

3.1 4.6 4.8 10.8 26.3 5.6 2.5 3.2 4.8 2.7

3.1 1.6 3.2 2.1 2.0 2.2 NA 2.9

3.8 1.8 4.0 2.6 2.1 2.2 2.7 4.6

0.08 0.55 4.2 0.75 5.4 0.01 0.03 0.23 0.99

0.09 0.54 6.8 0.71 4.6 0.00 0.04 0.28 0.83

0.08

0.08

0.02

0.02

3.1.2. Metabolite analysis The developed metabolite analysis method indicated a relatively slow metabolism of [ 11 C]UCB-A, with 93% and 42% intact tracer present at 5 and 40 min, respectively, after injection (Fig. 2). The metabolites were well separated from [11C]UCB-A and were easily eluted in different fractions. A mobile phase with lower amount of organic modifier (ammonium carbonate buffer 10 mM pH 8.7: methanol (55:45, v/v)), giving a retention time of 13 min for [ 11C]UCB-A, was used to assess whether any metabolite peak was not separated from [11C]UCB-A when using the ordinary mobile phase composition. No such peaks were observed. Moreover, no traces of radiolabeled metabolites were found in rat brain at these time points. The sample preparation recovery was determined by adding [11C] UCB-A to human plasma. The samples were stored at room temperature during 0 to 28 min. The recovery was 92.1% (n = 3, RSD = 1.6%). No correlation with storage time was seen indicating that the tracer was stable under these conditions. 3.2. Rat studies

cm3, and linear regression with fixed VND was performed for all six blocking scans. The value of VND that resulted in the lowest combined sum of squared residuals of all six linear regressions was then used to compute occupancy values. As a possible simplified method, occupancy was also calculated assuming zero non-specific binding (VND = 0) and compared to Lassen plot-based values using linear regression. 2.3.4. Animal Welfare All animals used in this study were handled in accordance with the guidelines set up by the Swedish Animal Welfare Agency and approved by local animal ethics committee at Uppsala University (permit numbers: C 125/11 for pigs and C 49/10 for rats). 3. Results 3.1. Chemistry 3.1.1. Synthesis of [ 11C]UCB-A [11C]UCB-A was synthesized in a two-step reaction giving 3–5 GBq of the finished product. It was obtained with a radiochemical purity

A

3.2.1. Frozen section autoradiography The uptake of [ 11C]UCB-A was dose dependently blocked when coincubated with increasing concentrations of the (reference) SV2A ligands levetiracetam and seletracetam, with IC50 values of 0.78 and 0.14 μM, respectively. Results are presented in Fig. 3. 3.2.2. Biodistribution and dosimetry [ 11C]UCB-A rapidly distributed from blood to organs and highest uptakes were observed in adrenals and brain – both expressing SV2A – and in organs involved in elimination (e.g. kidneys and liver; Fig. 4). Table 1 shows residence times and absorbed doses for a number of organs. The organ receiving the highest absorbed dose was the liver, at 12.5 and 26.3 μGy/MBq in males and females, respectively. Effective doses were 2.9 μSv/MBq and 4.6 μSv/MBq, for males and females, respectively. 3.2.3. PET binding study Brain kinetics of [11C]UCB-A were drastically changed by pre-blockade of the SV2A protein with compound 4 (10 mg/kg). As shown in Fig. 5A in the baseline study, [ 11C]UCB-A accumulated during the first 60 min, followed by a slow wash out from the brain, while in the blocking study the kinetics, after the initial five min, were dominated by a gradual release of radioactivity throughout the time of investigation. As Fig. 5B shows,

B baseline (n=5) blocked (n=4)

Brain Plasma Injection of

3.5

Standardized uptake value (SUV)

3.5

Standardized uptake value (SUV)

329

3.0 2.5 2.0 1.5 1.0 0.5

displacing drug

3.0 2.5 2.0 1.5 1.0 0.5 0.0

0.0 0

15

30

45

60

Time (min)

75

90

0

15

30

45

60

75

90

Time (min)

Fig. 5. A: Whole-brain SUV versus time p.i. in rats at baseline and after blocking with compound 4. B: Displacement of [11C]UCB-A after administration of brivaracetam, 21 mg/kg, at 45 min p.i.

330

S. Estrada et al. / Nuclear Medicine and Biology 43 (2016) 325–332

A

B 4

80

[11C]UCB-A in plasma Whole blood

3

60

SUV

Parent fraction (%)

100

40

2 1

20 0

0 0

20

40

60

80

0

20

40

60

80

100

Time p.i. (min)

Time p.i. (min)

Fig. 6. (A) Fraction of intact [11C]UCB-A in plasma versus time after injection (%). Error bars depict mean ± SD. (B) Whole blood time-activity curve and metabolite-corrected arterial input curve for a typical study. The noisy first 10 min of the curve correspond to the duration of the on-line arterial sampling, whereas the remainder of the curve is a tri-exponential fit to the discrete sample data.

brivaracetam, administered at 45 min p.i., fully and rapidly displaced [ 11C] UCB-A, demonstrating the reversibility of [11C]UCB-A binding.

3.3. Pig study Fig. 6 shows the parent fraction in plasma versus time after injection for all pigs, as well as typical whole blood time-activity and metabolitecorrected arterial input curves. At 75 min p.i., 77 ± 3% of plasma radioactivity concentration was still due to intact tracer. Plasma to whole blood radioactivity concentration ratio was constant after injection at 1.01 ± 0.02 (mean ± SD) across all scans. Fig. 7 shows typical cortex and cerebellum time-activity curves at baseline and after administration of compound 5, clearly indicating the blocking effect. NLR fits and Logan plots of the same pig are shown in Fig. 8. According to the Akaike criterion, tracer kinetics of [ 11C]UCB-A were best described by a 1 T model at baseline and a 2 T model after blocking. Cortex VT values at baseline based on the 1 T model ranged from 16 to 30, with a mean standard error of 0.1; K1 values ranged from 0.06 to 0.20 mL/cm 3/min with a mean standard error of 0.004 and blood volume values ranged from 0.03 to 0.16 with a mean standard error of 0.01. The 1 T model could not fit the data after blocking. However, both at baseline and after blocking, uncertainty in VT was large for the two-tissue model and BPND could not be determined robustly, with standard errors of the same magnitude as the parameter value itself in all cases. Hence, no reliable occupancy values could be estimated using NLR. Logan analysis (30–90 min p.i.) was preferred in terms of reproducibility, with a mean absolute difference between test and re-test V T values of 16%. Therefore,

6

Cortex Cerebellum

SUV

4

2

0 0

20

40

60

80

100

Time p.i. (min) Fig. 7. Time-activity curves in cortex (closed circles) and cerebellum (open circles) for a typical pig at baseline (solid lines) and after administration of 300 μg/kg of compound 5 (dashed lines).

results from Logan analysis were used in the remainder of the present work. VT values at baseline ranged between 21 and 26 in cortex and between 19 and 22 in cerebellum, with cerebellum VT values always lower than those in the cortex. Fig. 9 shows parametric Logan VT images of a single pig at baseline and after two different doses of compound 5. Figs. 10 shows the results of the blocking study. A Lassen plot with common x-intercept (Fig. 10A) resulted in a VND of 2.6, comparable to a mean VND of 3.0 ± 1.3 when Lassen plots were made for each pig individually. Occupancy values based on a common VND correlated well with those based on individually fitted VND values (R 2 = 0.96), with no significant difference (p 0.81, Wilcoxon test). Assuming negligible non-specific binding, i.e. VND = 0, resulted in occupancy values that correlated well with the values based on the Lassen plot (R2 = 0.96), albeit with a significant negative bias (y intercept −10.9, SD 7.0; p = 0.03).

4. Discussion The purpose of the present work was to evaluate [ 11C]UCB-A as a PET tracer for SV2A in rats and pigs. [ 11C]UCB-A has a high brain uptake as measured in rats and pigs using PET. In the latter an SUV exceeding 6 was measured at 90 min p.i. In both species, the apparent brain kinetics suggested irreversible binding of [ 11C]UCB-A during the time course of the PET investigation. It could, however, be demonstrated by a displacement study in rat using the selective SV2A drug brivaracetam, that the binding of [ 11C]UCB-A to SV2A was indeed reversible. This was shown by the rapid wash-out of the tracer from the target protein and brain after drug challenge (Fig. 5). A potential explanation for the apparent irreversible kinetics, as shown by the time-activity-curve (Fig. 7), could be the high plasma stability and slow excretion of the tracer which resulted in a sustained blood level of intact tracer. As a result this caused an apparent steady influx of [ 11C]UCB-A from blood to brain tissue, even though the kinetics of the tracer binding to SV2A was reversible. In the metabolite analysis in rats, only more hydrophilic radiolabelled metabolites, as compared to [ 11C]UCB-A, were found while no traces of radiolabelled metabolites in the brain could be detected. This is consistent with the rapid wash-out of radioactivity from the brain as shown in the displacement study. The results show that [11C]UCB-A potentially can be used to measure the SV2A expression in brain and thus be a non-invasive tool for the study of the dose–occupancy relationship for selective SV2A drugs. The results also support a means to investigate the potential differences in SV2A expression in patients with epilepsy as compared with healthy controls. The latter could be of value in light of the purported association between SV2A and refractory epilepsy [25,26]. Because of the small

S. Estrada et al. / Nuclear Medicine and Biology 43 (2016) 325–332

A

B 100

80

80 60

Baseline Post-drug 2k 4k

40 20

Int(Ctis)/Ctis

Activity concentration (kBq/mL)

331

60 40

Baseline 20

Post-drug

0

0 0

20

40

60

80

100

0

10

20

30

40

50

Int(Cpl)/Ctis

Time p.i. (min)

Fig. 8. (A) Cortex time-activity curve at baseline (open circles) and post-drug (closed circles), and corresponding fits with single tissue (1 T-2 k) and two-tissue (2 T-4 k) compartment models. Fits overlap at baseline. (B) Logan plots of the same data as in (A), with fits to the 30–90 min portion of the graphs.

contribution of non-displaceable tracer uptake as seen in pigs, with VND approximately 10% of the total distribution volume VT at baseline, the total distribution volume VT could be used as a marker of SV2A expression. Furthermore, the narrow range of baseline VT values as shown in Fig. 10A may allow for comparison of SV2A expression between groups with relatively small numbers of subjects, or even between individuals. The developed method for quantification of [ 11C] UCB-A binding to SV2A requires a metabolite-corrected arterial input function. This complicates the use of [ 11C]UCB-A for routine studies in patients with epilepsy. However, since white matter is devoid of SV2A it could be possible to develop a reference tissue model based on white matter as reference [27], in a similar manner as has been suggested for other PET tracers without a suitable gray matter reference region, e.g. [ 11C]flumazenil [28] and recently suggested for another SV2A ligand, [ 11C]UCB-J [29]. This could, however, not be evaluated in this study due to the small brains of the animals used, but will be investigated in a future study in humans. Further simplifications, such as the use of SUV ratios, are dependent on whether (transient) equilibrium between target and reference tissue is reached within the duration of the scan, which will be subject of future studies in humans. The effects of the slow kinetics of [ 11C]UCB-A on future clinical application are much depending on the type of image evaluation to be used. If clinical application lies in the assessment of defect areas rather than in absolute quantification of SV2A, a late static scan could be used and the apparent irreversibility may actually be advantageous because it creates a high signal and consequently high image quality. For example, in epileptic patients, heterogeneity in SV2A expression, and hence tracer signal, may indicate epileptic lesions. The dosimetry study suggests higher effective doses in females than in males due to a considerably higher liver uptake in females.

However, even the higher effective dose in females, 4.6 μSv/MBq, would lead to an effective dose of no more than 1.8 mSv per typical administration of 400 MBq. This means that an individual could undergo 5 scans before reaching the general accepted highest dose of 10 mSv for healthy volunteers in medical research. The option of performing up to 5 tracer administrations in the same subject gives the opportunity to study the dose–occupancy relationship with a range of different doses of an SV2A drug as well as follow the duration of action on the target protein following a single dose of drug in the same subject. Furthermore, the short half-life of carbon-11 allows, in contrast to fluorine-18, repeated administration in the same subject on the same day. The synthesis and quality control methods for [ 11C]UCB-A have been validated according to good manufacturing practice and process validation batches analyzed with respect to; radiochemical purity, chemical purity, pH, residual solvents, endotoxins and sterility. [ 11C]UCB-A, formulated for intravenous injection, is thus formally qualified for use in man.

5. Conclusions We have developed the novel PET tracer, [ 11C]UCB-A, that can be used to measure intra-individual heterogeneity in SV2A expression and potentially to measure absolute SV2A expression in vivo. Apart from measuring drug occupancy, as we have shown, the tracer can potentially be used to compare SV2A expression between individuals because of the rather narrow range of baseline VT values. This will have to be further validated in human studies. The dosimetry allows up to 5 administrations of 400 MBq of [ 11C]UCB-A in humans.

Fig. 9. Parametric [11C]UCB-A Logan VT images of a single pig brain at baseline (left) and after two different doses of compound 5 resulting in 52% and 81% occupancy, respectively.

332

S. Estrada et al. / Nuclear Medicine and Biology 43 (2016) 325–332

A

B 100

300 3 0 100 0 6 60 3 30 12 6

20 0

10 0

Occupancy (%)

VT, baseline - VT, drug

30 0

0

80 60 40 20 0

0 VND

10

20

30

VT, baseline

0

100

200

300

Drug dose (µg/kg)

Fig. 10. A: Lassen plots (VT, baseline – VT, drug versus VT, baseline) for six different doses of compound 5 in μg/kg. The solid lines are multi-linear regressions with a common x-axis intercept equal to VND. B: Occupancy versus drug dose based on Lassen plots with common VND.

Acknowledgements This work was performed as a collaboration between Uppsala University and UCB Biopharma, which financially supported the study. The authors want to thank the staff at the Pre-clinical PET platform, Uppsala University, for their assistance in this study. References [1] Sander JW, Shorvon SD. Epidemiology of the epilepsies. J Neurol Neurosurg Psychiatry 1996;61:433–43. [2] Kwan P, Brodie MJ. Early identification of refractory epilepsy. N Engl J Med 2000; 342:314–9. [3] Robey RW, Lazarowski A, Bates SE. P-glycoprotein–a clinical target in drugrefractory epilepsy? Mol Pharmacol 2008;73:1343–6. [4] Luna-Tortos C, Fedrowitz M, Loscher W. Several major antiepileptic drugs are substrates for human P-glycoprotein. Neuropharmacology 2008;55:1364–75. [5] Buckley K, Kelly RB. Identification of a transmembrane glycoprotein specific for secretory vesicles of neural and endocrine cells. J Cell Biol 1985;100:1284–94. [6] Madeo M, Kovacs AD, Pearce DA. The human synaptic vesicle protein, SV2A, functions as a galactose transporter in Saccharomyces cerevisiae. J Biol Chem 2014;289: 33066–71. [7] Crowder KM, Gunther JM, Jones TA, Hale BD, Zhang HZ, Peterson MR, et al. Abnormal neurotransmission in mice lacking synaptic vesicle protein 2 A (SV2A). Proc Natl Acad Sci U S A 1999;96:15268–73. [8] Crevecoeur J, Kaminski RM, Rogister B, Foerch P, Vandenplas C, Neveux M, et al. Expression pattern of synaptic vesicle protein 2 (SV2) isoforms in patients with temporal lobe epilepsy and hippocampal sclerosis. Neuropathol Appl Neurobiol 2014;40:191–204. [9] Gillard M, Chatelain P, Fuks B. Binding characteristics of levetiracetam to synaptic vesicle protein 2 A (SV2A) in human brain and in CHO cells expressing the human recombinant protein. Eur J Pharmacol 2006;536:102–8. [10] Lynch BA, Lambeng N, Nocka K, Kensel-Hammes P, Bajjalieh SM, Matagne A, et al. The synaptic vesicle protein SV2A is the binding site for the antiepileptic drug levetiracetam. Proc Natl Acad Sci U S A 2004;101:9861–6. [11] Klitgaard H, Matagne A, Gobert J, Wulfert E. Evidence for a unique profile of levetiracetam in rodent models of seizures and epilepsy. Eur J Pharmacol 1998;353:191–206. [12] Mercier J, Archen L, Bollu V, Carre S, Evrard Y, Jnoff E, et al. Discovery of heterocyclic nonacetamide synaptic vesicle protein 2 A (SV2A) ligands with single-digit nanomolar potency: opening avenues towards the first SV2A positron emission tomography (PET) ligands. ChemMedChem 2014;9:693–8. [13] Bretin F, Bahri MA, Bernard C, Warnock G, Aerts J, Mestdagh N, et al. Biodistribution and radiation dosimetry for the novel SV2A radiotracer [F]UCB-H: first-in-human study. Mol Imaging Biol 2015.

[14] Bretin F, Warnock G, Bahri MA, Aerts J, Mestdagh N, Buchanan T, et al. Preclinical radiation dosimetry for the novel SV2A radiotracer [18F]UCB-H. EJNMMI Res 2013;3:35. [15] Cai H, Mangner TJ, Muzik O, Wang MW, Chugani DC, Chugani HT. Radiosynthesis of (11)C-levetiracetam: A potential marker for PET Imaging of SV2A expression. ACS Med Chem Lett 2014;5:1152–5. [16] Kenda B, Jnoff E. (UCB Pharma S.A., Brussels, Belgium) PCT Int. Pub. WO 2014/ 012563; 2014. [17] Michel P, Differding E, Pasau PM, Kenda B, Lallemand B, Matagne A, et al. (UCB Pharma S.A., Brussels, Belgium) PCT Int. Pub. WO 01/62726, WO 01/64637; 2001. [18] Kenda BM, Matagne AC, Talaga PE, Pasau PM, Differding E, Lallemand BI, et al. Discovery of 4-substituted pyrrolidone butanamides as new agents with significant antiepileptic activity. J Med Chem 2004;47:530–49. [19] Stabin MG, Sparks RB, Crowe E. OLINDA/EXM: the second-generation personal computer software for internal dose assessment in nuclear medicine. J Nucl Med 2005;46:1023–7. [20] Watanabe M, Okada H, Shimizu K, Omura T, Yoshikawa E, Kosugi T, et al. A high resolution animal PET scanner using compact PS-PMT detectors. IEEE Trans Nucl Sci 1997;44:1277–82. [21] Lubberink M, Kosugi T, Schneider H, Ohba H, Bergstrom M. Non-stationary convolution subtraction scatter correction with a dual-exponential scatter kernel for the Hamamatsu SHR-7700 animal PET scanner. Phys Med Biol 2004;49:833–42. [22] Logan J, Fowler JS, Volkow ND, Wolf AP, Dewey SL, Schlyer DJ, et al. Graphical analysis of reversible radioligand binding from time-activity measurements applied to [N- 11 C-methyl]-(−)-cocaine PET studies in human subjects. J Cereb Blood Flow Metab 1990;10:740–7. [23] Akaike H. A new look at the statistical model identification. IEEE Trans Autom Control 1974;19:716–23. [24] Lassen NA, Bartenstein PA, Lammertsma AA, Prevett MC, Turton DR, Luthra SK, et al. Benzodiazepine receptor quantification in vivo in humans using [11C]flumazenil and PET: application of the steady-state principle. J Cereb Blood Flow Metab 1995; 15:152–65. [25] van Vliet EA, Aronica E, Redeker S, Boer K, Gorter JA. Decreased expression of synaptic vesicle protein 2 A, the binding site for levetiracetam, during epileptogenesis and chronic epilepsy. Epilepsia 2009;50:422–33. [26] Mohanraj R, Parker PG, Stephen LJ, Brodie MJ. Levetiracetam in refractory epilepsy: a prospective observational study. Seizure 2005;14:23–7. [27] Crevecoeur J, Foerch P, Doupagne M, Thielen C, Vandenplas C, Moonen G, et al. Expression of SV2 isoforms during rodent brain development. BMC Neurosci 2013;14:87. [28] Klumpers UM, Veltman DJ, Boellaard R, Comans EF, Zuketto C, Yaqub M, et al. Comparison of plasma input and reference tissue models for analysing [(11)C]flumazenil studies. J Cereb Blood Flow Metab 2008;28:579–87. [29] Finnema SJ, Nabulsi N, Mercier J, Lin S, Najafzadeh S, Ropchan J, et al. Evaluation of [11C]UCB-J as a novel PET radioligand for imaging synaptic vesicle glycoprotein 2 A (SV2A) in the human brain. 17th International Symposium on Cerebral Blood Flow, Metabolism and Function. Vancouver; 2015 [Brain-0522].