Label-free assay for the assessment of nonspecific binding of positron emission tomography tracer candidates

    Label-free assay for the assessment of nonspecific binding of positron emission tomography tracer candidates Frauke Assmus, Anna Seelig, Luca Gobbi, Edilio Borroni, Patricia Glaentzlin, Holger Fischer PII: DOI: Reference:

S0928-0987(15)30003-8 doi: 10.1016/j.ejps.2015.08.014 PHASCI 3346

To appear in: Received date: Revised date: Accepted date:

4 May 2015 2 August 2015 25 August 2015

Please cite this article as: Assmus, Frauke, Seelig, Anna, Gobbi, Luca, Borroni, Edilio, Glaentzlin, Patricia, Fischer, Holger, Label-free assay for the assessment of nonspecific binding of positron emission tomography tracer candidates, (2015), doi: 10.1016/j.ejps.2015.08.014

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Label-free assay for the assessment of nonspecific binding of positron

T

emission tomography tracer candidates

IP

Frauke Assmus,a,b1*, Anna Seeligb, Luca Gobbia, Edilio Borronia, Patricia Glaentzlina, Holger

a

SC R

Fischera

F. Hoffmann-La Roche Ltd., pRED, Pharma Research and Early Development,

Biozentrum, University of Basel, Div. of Biophysical Chemistry, Klingelbergstrasse 50/70,

MA

b

NU

Pharmaceutical Research, Innovation Center Basel, CH-4070 Basel, Switzerland.

CH-4056 Basel, Switzerland

Present address: Centre for Applied Pharmacokinetic Research, Manchester Pharmacy

D

1

TE

School, University of Manchester, Stopford Building, Oxford Road, M13 9PT Manchester,

CE P

United Kingdom

AC

*To whom correspondence should be addressed: Phone: +44 7975858459, e-mail: [email protected]

E-mail addresses of the co-authors: [email protected] [email protected] [email protected] [email protected] [email protected]

1

ACCEPTED MANUSCRIPT ABSTRACT Positron emission tomography (PET) is a valuable non-invasive technique for the

T

visualisation of drug tissue distribution and receptor occupancy at the target site in living

IP

animals and men. Many potential PET tracers, however, fail due to an unfavorably high non-

SC R

specific binding (NSB) to non-target proteins and phospholipid membranes which compromises the sensitivity of PET. Hence, there is a high demand to assess the extent of

NU

NSB as early as possible in the PET tracer development process, preferentially before ligands are radiolabeled and elaborate imaging studies are performed. The purpose of this study was

MA

to establish a novel Lipid Membrane Binding Assay (LIMBA) for assessing the tendency of potential tracers to bind non-specifically to brain tissue. The assay works with unlabeled

D

compounds and allows the medium-throughput measurement of brain tissue/water distribution

TE

coefficients, log Dbrain (pH 7.4), at minimal expense of animal tissue. To validate LIMBA,

CE P

log Dbrain (pH 7.4) values were measured and compared with NSB estimates derived from in

vivo PET studies in human brain (n = 10 tracers, literature data), and in vitro autoradiography studies in rat and mouse brain slices (n = 30 tritiated radioligands). Good agreement between

AC

log Dbrain (pH 7.4) and the volume of distribution in brain of non-specifically bound tracer in

PET was achieved, pertaining to compounds classified as non-substrates of P-glycoprotein (R2 ≥ 0.88). The ability of LIMBA for the prediction of NSB was further supported by the strong correlation between log Dbrain (pH 7.4) and NSB in brain autoradiography (R2 ≥ 0.76), whereas octanol/water distribution coefficients, log Doct (pH 7.4) were less predictive. In conclusion, LIMBA provides a fast and reliable tool for identifying compounds with unfavorably high NSB in brain tissue. The data may be used in conjunction with other parameters like target affinity, density and membrane permeability for the selection of most promising compounds to be further investigated in vivo as potential novel PET tracers. 2

ACCEPTED MANUSCRIPT

MA

NU

SC R

IP

T

GRAPHICAL ABSTRACT

TE

D

KEYWORDS

AC

CE P

Nonspecific binding, tissue binding, positron emission tomography, lipophilicity, PET tracer

3

ACCEPTED MANUSCRIPT 1 Introduction Positron Emission Tomography (PET) has emerged as a valuable tool in the drug

IP

T

development process to study the biodistribution and receptor occupancy of drug candidates

SC R

at the target site in living animals and men (Fernandes et al., 2012). The value of PET as a non-invasive imaging technique is of particular interest in the field of CNS drug discovery where large discrepancies may arise between plasma and biophase pharmacokinetics and,

NU

moreover, where animal models may not adequately reflect the complexity of human disease (Danhof et al., 2007).

MA

In order to exploit the full potential of PET for drug development, drug candidates and the corresponding PET tracers for the target site of interest should ideally be developed in

D

parallel. However, there is still a lack of suitable PET tracers for many targets, and their

TE

innovation is challenging given the large number of sometimes conflicting requirements to be

CE P

fulfilled, particularly for imaging agents that bind to specific targets in the brain (Honer et al., 2014; Laruelle et al., 2003). A high binding sites density is desirable and only the unmetabolized radiotracer should pass the blood-brain barrier to bind with high affinity and

AC

specificity to the target site of interest (specific binding, SB) (Ametamey and Schubiger, 2006). In contrast, non-specific binding (NSB) to non-target proteins and lipids should be low to ensure a high SB to NSB ratio. The higher the ratio, the higher the sensitivity of PET to monitor changes of available binding sites due to drug-related receptor occupancy (Laruelle et al., 2003; Pike, 1993). Effective radiotracers should achieve SB/NSB ratios in the range of SB/NSB = 3-10 to be successfully applied in vivo (McCarthy et al., 2009; Pike, 1993), yet, Laruelle et. al (Laruelle et al., 2003) summarized that ”most of the failures in ligand development result from an unfavorable combination of target density, ligand affinity and nonspecific binding”. 4

ACCEPTED MANUSCRIPT The full characterization of a PET tracer is time-consuming, expensive and requires expert knowledge of a radiotracer imaging team as well as special technical facilities (Wong and

T

Pomper, 2003). Less resource-demanding in vitro assays that allow the selection of

IP

compounds with optimal properties are therefore applied to facilitate the development of PET

SC R

tracers and to focus resources on the most promising compounds. In vitro autoradiography (incubation of radioligands with tissue slices), for example, can be performed to obtain a primary measure of the SB/NSB ratio, without the need to incorporate positron emitting

NU

nuclei into the ligands. Tracers are rejected when the SB/NSB is too low, i.e. when a major

MA

part of the total binding (TB = SB + NSB) in the brain region of interest (ROI) is originating from an unfavorably high NSB, as assessed in a reference region (RREF) devoid of receptors. Yet, failing at this point of the tracer development process is costly given that

sacrifice of animals.

TE

D

autoradiography studies are time-consuming, require radiolabeling of the ligand and the

CE P

To provide a faster and less resource-intensive estimation of NSB, in vitro methods working with the unlabeled compound are highly desirable. The octanol/water distribution

AC

coefficient, log Doct (pH 7.4) is still the commonly used parameter for assessing the NSB of potential PET tracers, albeit with limited prediction accuracy for structurally unrelated compounds (Dickson et al., 2011; Laruelle et al., 2003). The unbound fraction of a ligand in brain, f u ,brain , available from equilibrium dialysis with brain homogenate (Kalvass and Maurer, 2002; Liu et al., 2009; Summerfield et al., 2007), provides a physiologically more relevant NSB-estimate which has successfully been applied for the parameterization of a comprehensive radiotracer kinetic model (Guo et al., 2009). Likewise, Fridén et. al (Friden et al., 2014) proposed using the unbound volume of distribution ( Vu ,brain ) available from the brain slice uptake technique (Friden et al., 2009), to predict NSB in PET.

5

ACCEPTED MANUSCRIPT However, a drawback of the homogenate dialysis and the brain slice uptake technique is the high consumption of tissue and the relatively low throughput, respectively, which impedes

T

the screening of larger datasets. Recently, a vesicle electrokinetic chromatography (VEKC)

IP

method working with an artificial anionic detergent (Sodium bis(2-ethylhexyl) sulfosuccinate,

SC R

AOT) has been developed for the prediction of NSB showing a good correlation between chromatographic retention factors and in vitro NSB derived from equilibrium dialysis. The VEKC method has also been shown to provide a useful tool for the classification of either

NU

successful or terminated PET ligands tested in vivo, however, on a qualitative level only

MA

(Jiang et al., 2011).

Quantitative in vivo NSB estimates, expressed as the volume of distribution of nonspecifically bound tracer in brain, have been derived from kinetic analysis of PET scans and

TE

D

were used for the validation of a computational NSB prediction method. A good correlation with calculated drug-lipid interaction energy was found (Dickson et al., 2011; Rosso et al.,

CE P

2008), yet, the ab initio quantum mechanical method was computationally expensive and relied on the interaction of the ligand with 1,2-dioleoyl-sn-glycero-3-phosphorylcholine

AC

(DOPC) neglecting the full complexity of a biological membrane. The purpose of the present study was to test whether a novel in vitro assay, referred to as Lipid Membrane Binding Assay (LIMBA), proves useful for estimating the NSB of potential PET tracers. LIMBA requires only small amounts of rat brain tissue for the label-free assessment of tissue binding. The medium-throughput assay was recently developed in our laboratory and yields brain tissue/water distribution coefficients, log Dbrain in excellent agreement with f u ,brain (homogenate dialysis) (Assmus, 2015; Assmus et al., Manuscript in preparation-b), even when brain tissue was replaced by a microemulsion of brain polar lipids (Belli et al., Manuscript in preparation). In the course of this study, we evaluated the ability of LIMBA for predicting NSB on a set of i) 30 publicly available and proprietary tritiated 6

ACCEPTED MANUSCRIPT radioligands for which NSB-estimates in brain were available from in vitro autoradiography studies (in-house data, in rat and mouse), and ii) 10 PET tracers for which the volume of

AC

CE P

TE

D

MA

NU

SC R

IP

was retrieved from in vivo PET studies (literature data, in human).

T

distribution of non-specifically bound drug related to the free plasma concentration ( VNS f p )

7

ACCEPTED MANUSCRIPT 2 Materials and methods Materials and

3-amino-4-[2-[(di(methyl)amino)methyl]phenyl]sulfanylbenzonitrile

IP

Fallypride

T

2.1

SC R

(DASB) were purchased from ABX (Radeberg, Germany). Diprenorphine, SCH 23390 and SR 222200 were obtained from Tocris (Bristol, UK). Flumazenil and raclopride tartrate were purchased from Sigma (Steinheim, Germany). All other drugs were available through our in-

NU

house Compound Depository Group as proprietary compounds. TRIS, TAPSO and sodium

MA

chloride were obtained from Fluka (Buchs, Switzerland). Octanol and dodecane were purchased from Sigma (Steinheim, Germany). DMSO was obtained from Acros (Geel, Belgium). Water and acetonitrile for HPLC were supplied from Merck (Darmstadt, Germany)

TE

Determination of n-octanol/water distribution coefficients

CE P

2.2

D

and were of HPLC-grade.

Unless otherwise stated, log Doct (pH 7.4) values were measured with a miniaturized shake flask technique. In brief, a DMSO stock solution of the test compound (0.5 mM) was

AC

dispensed in aqueous buffer (50 mM TAPSO /pH 7.4, final DMSO content 5 %) which was pre-saturated with octanol. The aqueous drug solution (25 μM) was filtrated and an aliquot of the filtrate was transferred to a 96-well PCR plate (Eppendorf). After adding buffer-saturated octanol on top of the filtrate, the PCR plate was sealed and incubated for two hours at room temperature (RT), while shaking. The plate was left undisturbed overnight and then centrifuged for 10 min at 3000 rpm (Eppendorf, Centrifuge 5810R). The aqueous drug solution was sampled with a robotic pipetting system (Tecan) and the equilibrium aqueous eq drug concentration, Caq , was determined with a UV plate reader (SpectraMax). In parallel, a

reference experiment was carried out under the same conditions but without octanol in order 8

ACCEPTED MANUSCRIPT 0 to obtain the initial aqueous drug concentration, Caq . LogDoct (pH 7.4) values were obtained by

mass balance according to





Eq. (1)

IP

T

0 eq  Caq  Caq Vaq  , log Doct  log eq   C  V aq oct  

2.3

SC R

where Vaq and Voct denote the volume of the aqueous and the octanol phase, respectively.

Determination of brain tissue/water distribution coefficients with LIMBA

NU

Brain tissue (rat) /water distribution coefficients, log Dbrain (pH 7.4), were determined with the in-house LIMBA assay as described below. Briefly, drug-naive female Wistar rats were

MA

sacrificed in a CO2 chamber and the whole brain was sampled and immersed in assay buffer (50 mM TRIS/114 mM NaCl/ pH 7.4) to remove adherent blood. The rat brain was weighted

D

and homogenized on ice in 2 volumes (w/v) of n-dodecane using an ultrasonic probe (Branson

TE

sonifier, G. Heinemann, Schwäbisch Gmünd, Germany). Experiments were conducted in

CE P

accordance with the current Cantonal and Federal legislation on the welfare of experimental animals. For the preparation of the aqueous phase, test compounds were introduced into the assay buffer (50 mM TRIS/114 mM NaCl/pH 7.4) as DMSO stock solutions (10 mM). The

AC

resulting aqueous drug solution (25 μM, DMSO content 0.25 %) was filtrated (NUNC Cat.No. 278752) and aliquots of the filtrate ( V aq = 50 µL) were transferred into an in-house made Teflon plate. In parallel, the rat brain homogenate ( Vbr = 1.2 μL) was coated on a PVDF filter attached to the bottom of the customized and commercially available DIFI tubes (Weidmann Plastics Technology AG, Rapperswill, Switzerland, 23358) (Fischer et al., 2006). The pipetting was conducted with a Hamilton robotic system using new pipetting tips (disposable, 10 μL) for each coating step and a sealed 96 well PCR plate (Eppendorf) for the storage of brain homogenate, which was put on a shaker. The coated DIFI tubes were added on top of the Teflon plate pre-filled with aqueous drug solution thus forming a sandwich. The 9

ACCEPTED MANUSCRIPT kit was sealed and allowed to shake at RT for 15 h. The incubation time was based on prior experiments showing that distribution equilibrium was achieved after 15 h. After removal of

T

eq the DIFI tubes, the equilibrium drug concentration in the aqueous phase, Caq , was analyzed

IP

with HPLC - UV/MS. A reference experiment was carried out in the same plate but without

SC R

brain homogenate. LogDbrain (pH 7.4) values were obtained by mass balance according to the assessment of log Doct (pH 7.4). All distribution experiments were performed in triplicate and

HPLC-UV/MS instrumentation and chromatographic conditions

MA

2.4

NU

mean log Dbrain(pH 7.4) values are reported.

Drug concentrations in aqueous solutions were quantified using an Agilent 1290 Infinity

D

HPLC-UV/MS system (Agilent Technologies, CA, USA) equipped with a DAD detector and

TE

an API single quadrupole mass spectrometer (Agilent 6140). An aliquot of each sample (5 µL) was injected onto a Kinetex 2.6 µm, 2.1 x 50 mm analytical column (Phenomenex,

CE P

Germany) operated at 60 °C. The mobile phase consisted of A (water) and B (acetonitril), both containing 0.1 % (V/V) formic acid. The gradient elution was performed as follows:

AC

initial 2 % B, 0 - 0.35 min linear gradient from 2 % B to 95 % B, 0.35 min - 0.65 min 95 % B, post time 0.5 min. After passing the DAD detector, the eluent was introduced into the electrospray interface maintaining the following source settings: capillary voltage + 3.5 kV, drying gas flow (N2) 13 L/min, nebulizer pressure 60 psi, drying gas temperature 350 °C. The integral of the [M+H]+ ion intensity was acquired in single reaction monitoring mode. Internal calibration for MS data analysis was accomplished by injection of 1, 3, and 5 µL of the sample solution. Peak areas at appropriate wavelength (UV) or mass responses (if UV sensitivity was insufficient) were used for the calculation of log Dbrain .

10

ACCEPTED MANUSCRIPT 2.5

Determination of [3H]-radioligand binding in brain by in vitro autoradiography A dataset of 30 tritiated radioligands was comprised from our in-house database and

T

included radioligands with known NSB data measured with in vitro autoradiography in brain

IP

as described elsewhere in more detail (Borroni et al., 2011). Briefly, laboratory animals (mice,

SC R

gerbil, rats) were housed at controlled temperature (20 - 22 °C) and 12 h light/dark cycle. All animals were allowed ad libitum access to food and water. The animals were killed by decapitation and the brains were rapidly removed from the skull and immediately frozen on

NU

dry ice. Sections (10 - 20 μm) from brain were cut in a cryostat and thaw-mounted onto

MA

silane-coated microscope slides (HistoBond, Paul Marienfeld GmbH, Lauda-Königshofen, Germany). Brain sections were dried at RT and stored at -20 °C until the day of the

D

autoradiography measurements. Brain sections were first pre-incubated (10 - 20 min) in assay

TE

buffer (in most cases TRIS/pH 7.4) and then in assay buffer containing the tritium-labeled ligand (concentration of the ligand: 0.08 - 23.5 nM; incubation time: 12 - 120 min; RT or

CE P

37 °C). For the assessment of NSB, additional brain sections were incubated in presence of a non-radiolabeled compound that competes with the radioligand for the same binding site

AC

(blocker concentration: 1 - 100 μM). Brain sections were subsequently rinsed in ice-cold assay buffer (washing time, twash = 2 - 10 min) and dipped briefly (n = 1 - 3) in distilled water at 4 °C to remove buffer salts. The brain sections were dried at RT and exposed, together with a tritium microscale (RPA-510, GE Healthcare, Glattbrugg, Switzerland), to tritium-sensitive imaging plates (BAS-TR 2025, Fujifilm, Dielsdorf, Switzerland) for 5 - 6 days. The imaging plate was scanned in a Fujifilm high-resolution plate scanner (BAS-5000, Bucher Biotec AG, Basel, Switzerland) and images were quantified with an MCID M2 image analysis system (Imaging Research Inc., St. Catherines, Ontario, Canada). The total amount of radioligand bound in the ROI was expressed as fmol of bound radioligand per mg of protein (total binding, TB = SB + NSB). A quantitative estimate of NSB was obtained by blocking the 11

ACCEPTED MANUSCRIPT receptors of interest with an excess of an unlabeled competitor as described above. The experimental conditions (buffer, incubation temperature, animal species, length of

T

incubation/rinsing) used for quantification of receptor binding and NSB were optimized for

Compilation of VNS/fp data from in vivo PET studies

SC R

2.6

IP

each tracer and are enclosed in detail in the Supplementary material.

A database of 10 radioligands was compiled from literature and, if not otherwise stated,

NU

included PET tracers investigated with in vivo PET in living man. Different tracer kinetic models (Slifstein and Laruelle, 2001), e.g. one tissue and two tissue compartmental models,

MA

have been used in the reported studies for the analysis of radioligand - time activity curves in PET. The volume of distribution of the non-specifically bound radioligand in a reference

D

region with respect the total plasma concentration ( VNS ) was extracted from the published

TE

kinetic PET data and was expressed with respect to the free plasma concentration according

CE P

to: VNS VND K'   1  1'  1, fp fp f p k2

Eq. (2)

AC

where f p is the unbound fraction of radioligand in plasma reported in literature and V ND is the volume of distribution of the non-displaceable radioligand (free +non-specifically bound) with respect to the total plasma concentration in a reference region (for references see Table 3). K1' and k 2' denote the rate constants for the transfer of radioligand from plasma into the brain and vice versa, respectively. The term VNS f p describes the ratio of the concentration of radioligand bound to brain tissue to the unbound concentration of radioligand in plasma. Provided that a radioligand passes the blood-brain barrier by passive diffusion, VNS f p accordingly reflects the extent of NSB in brain (Innis et al., 2007). However, if active

transport across the BBB occurs, VNS f p also reflects the contribution of active uptake/efflux 12

ACCEPTED MANUSCRIPT in addition to NSB. Further details regarding the derivation of Eq.2 including relevant background information about compartmental analysis and tracer kinetic modelling are

T

provided in the Supplementary material. Eq. 2 is similar to that derived by (Rosso et al., 2008)

IP

and (Dickson et al., 2011), however, for clarity we distinguish VNS (distribution volume

SC R

related to the total plasma concentration) from VNS f p (distribution volume related to the free plasma concentration). No VNS f p - data from in vivo PET studies in human brain was

NU

available for rolipram and therefore data for S(+)-rolipram from PET studies in porcine brain

2.7

MA

was used for regression analysis.

Statistical Analysis

TE

D

Linear regression between NSB in brain (in vitro autoradiography or in vivo PET data) and log Doct as well as log Dbrain was performed with OriginPro v7.5 (OriginLab Corporation,

CE P

Northampton, MA, USA). A straight line was fitted to the data and the coefficient of determination, R 2 , in addition to standard deviations, SD values, were calculated according

AC

to R 2  1  RSS SYY , where RSS is the residual sum of squares and SYY is the total sum of squares of the dataset. LogDbrain values for compounds with known NSB in brain could be measured for 27 of 30 compounds; the comparison between correlation statistics using either log Doct or log Dbrain as NSB surrogates were based on the same 27 - compound dataset (Table

2).

13

ACCEPTED MANUSCRIPT 3 Results and discussion A high NSB of radiotracers to lipids and non-target proteins affects the accuracy with

IP

T

which the specific binding to a target site can be measured by means of PET. Many PET

SC R

tracers fail due to an unfavorable combination of high NSB, low receptor density and insufficient receptor affinity. Notably, high NSB is one of the most common reasons for failure of novel PET tracer candidates (Laruelle et al., 2003). In this study, we have evaluated

NU

the ability of a novel medium-throughput assay (LIMBA) to predict the extent of NSB of unlabeled ligands in order to identify compounds which are likely to fail. The results are

MA

shown and discussed below in comparison with log Doct (pH 7.4), using the example of

Characterization of the in vitro autoradiography dataset

TE

3.1

D

radioligands characterized with i) in vitro autoradiography and ii) in vivo PET.

CE P

A dataset of 30 tritiated radioligands, whose NSB in presence of a blocker was estimated by autoradiography in brain sections (rat, mouse), was compiled as described in Materials and methods. The larger body of NSB autoradiographic data as compared to PET data allowed us

AC

to form a structurally diverse database covering a wide physicochemical property space. The dataset was comprised of both publicly disclosed and proprietary radiotracers in the tritiumlabelled form, yet, the measurement of log Doct (pH 7.4) and log Dbrain (pH 7.4) was performed with the unlabelled ligands. The structures for 12 ligands are disclosed in the Supplementary material

(Section

S5).

Key

physicochemical

properties

(molecular

weight,

pK a , c log P , log Doct (pH 7.4)) of all 30 compounds are provided in Table 1 along with the NSB

autoradiography data. In addition, TB values measured in ROIs under non-blocking conditions and the corresponding TB/NSB binding ratios are provided. The dataset included neutral, basic, and zwitter-ionic compounds covering a broad range of log Doct values

14

ACCEPTED MANUSCRIPT ( log Doct (pH 7.4) = -0.64 to 4.35), NSB-estimates (NSB = 7.6 to 3151.1 fmol/mg protein) as well as TB/NSB binding ratios (TB/NSB = 1 to 368). Examples of autoradiographic images

T

for radioligands with different imaging performances are provided in the Supplementary

SC R

IP

material.

Table 1

The relationship between NSB in brain (autoradiography data) and log Doct

NU

3.2

MA

LogDoct (pH 7.4) is the most commonly used parameter for estimating the NSB of potential

PET tracers, however, the datasets investigated hitherto were rather small and a trend between

D

log Doct and NSB was only observed when structurally similar ligands were compared

TE

(Laruelle et al., 2003). In order to evaluate the predictive ability of log Doct (pH 7.4) on a larger and structurally diverse set of compounds, we compared the NSB-estimates from the in vitro

CE P

autoradiography dataset with measured log Doct (pH 7.4) values. At this point it should be noted that the times brain slices were exposed to a washing

AC

solution in autoradiography were slightly different for the individual tracers. Since longer washing times are likely associated with a reduction in NSB, the autoradiography data was divided in two subsets according to a washing time of twash = 2 - 4 min and twash = 7 - 10 min. The relationship between log Doct (pH 7.4) and NSB for the two subsets is shown in Fig. 1A and the correlation statistics are summarized in Table 2.

Fig. 1, Table 2

Improved, but still rather poor fits to the experimental NSB were achieved for the two subsets as compared to the correlation based on the whole dataset. A clear trend was evident 15

ACCEPTED MANUSCRIPT towards lower NSB at longer incubation times of brain slices with the washing buffer. The discrepancy in NSB is reflected by the intercepts of the linear regression lines which are more

T

than one log unit apart. Long washing procedures of brain slices in autoradiography might be

IP

desirable to maximize TB/NSB ratios, however, the situation in vivo is lacking the benefit of

SC R

washing. Therefore, we propose limiting the exposure times of brain slices to washing solutions to maximal 4 minutes to achieve the elimination of buffer salts, without reducing NSB in an unphysiological manner. Likewise, a non-washing procedure as proposed by other

NU

authors may be applied (Patel et al., 2003), as long as a uniform protocol is used for all

MA

ligands of interest.

Due to the scatter of the data shown in Fig. 1A, no clear trend in terms of the predictive

D

power of log Doct (pH 7.4) for different charge classes can be drawn. However, for the less

TE

scattered subset at washing times twash = 7 - 10 min, an underestimation of NSB by log Doct (pH 7.4) is evident for the most hydrophilic basic compound (No. 25), while the

CE P

remaining basic and neutral compounds roughly cluster. With respect to the rather poor correlation between log Doct (pH 7.4) and NSB, it is worth to

AC

note that even though octanol is widely considered as a surrogate for lipid bilayer membranes, it lacks the anisotropy of a bilayer and the charge of a phospholipid head group (Seelig, 2007). Depending on the charge state, the binding of drugs to phospholipid membranes may be driven by both hydrophobic and electrostatic interactions. At the example of propranolol and other bases, it has been shown that the neutral and cationic species of a basic drug can bind to a net neutral phospholipid membrane with a relatively large contribution of the ionized species to the overall extent of membrane binding (Avdeef et al., 1998; Balon et al., 1999). The minor contribution of ionized species to the partitioning into octanol (here observed for compound No.25) may accordingly reflect the difference between bulk phase partitioning (octanol) and drug binding in an anisotropic amphiphilic environment in which 16

ACCEPTED MANUSCRIPT charged drugs can bind to a membrane surface. The presence of negatively charged phospholipids (e.g. phosphatidylserine (PS), phosphatidylinositol (PI), phopshatidylglycerol

T

(PG), phosphatidic acid (PA)) in a biological membrane may further enhance the membrane

IP

affinity of basic drugs by electrostatic interactions with the acidic phospholipid head groups.

SC R

Pronounced distribution of basic drugs into tissues containing high concentration of negatively charged phospholipids is well documented (Obach et al., 2008; Rodgers et al., 2005a, b), and in particular the presence of PS has been identified as a main determinant of

NU

interorgan variation in tissue distribution (Yata et al., 1990). Notably, brain tissue is relatively

MA

rich in acidic phospholipids (between 9.4 and 20.8 % of total phospholipids) as estimated from the phospholipid composition in brain tissue reported for rat as: 50 mg of lipid phosphorous / g of tissue, thereof 0.23 - 5.9 mg/g PS, 2.79 mg/g PI, 1.10 mg/g PG and

The Prediction of NSB in brain (autoradiography data) using LIMBA

CE P

3.3

TE

D

0.6 mg/g PA (Rodgers et al., 2005b).

In analogy to log Doct (pH 7.4), log Dbrain (pH 7.4) values for the complete autoradiography dataset were measured with a novel in vitro assay, LIMBA, as described in Materials and

AC

methods. As opposed to the octanol shake flask technique, LIMBA works with minimal consumption of brain homogenate in dodecane (here: rat) which more likely resembles the composition and properties of brain tissue (see Section 3.2). The assay has been optimized for compounds with intermediate to high tissue binding, allowing for the measurement of log Dbrain (pH 7.4) values in the range of log Dbrain (pH 7.4) = 0.35 - 4. This range applied to 27 of the 30 compounds tested for which the results are presented in Table 1. Fig. 1B shows the correlation and the linear regression lines between log NSB and log Dbrain (pH 7.4) for the different times brain slices were exposed to the washing solution.

An inspection of the correlation statistics (Table 2) in comparison with those obtained with 17

ACCEPTED MANUSCRIPT 2

log Doct (pH 7.4) reveals a higher statistical certainty (R = 0.76 - 0.82) in the estimates of NSB

when using log Dbrain (pH 7.4). It is worth to note that tissue binding ( log Dbrain ) may also be

T

expressed in terms of the unbound fraction of drug in brain, f u ,brain , defined by

IP

fu,brain  Cu,brain Ctotal,brain , where Cu ,brain and Ctotal,brain is the unbound and the total concentration

SC R

of drug in brain, respectively. Log Dbrain and f u ,brain are non-linearly related according to: LogDbrain  logCbound,brain Cu.brain  log1  fu,brain fu,brain . However, similar to log Dbrain , NSB

NU

data obtained from in vitro autoradiography and VNS f p data obtained from PET studies (see

MA

Section 3.4) refer to the concentration ratio of bound drug, Cbound,brain , to unbound drug, Cu ,brain . Accordingly, all prediction approaches are based on linear regression between NSB

D

and distribution coefficients ( log Dbrain or log Doct ) rather than f u ,brain . For a representation of

TE

the relationship between log Dbrain and f u ,brain as well as the derivation of relevant equations, the reader is referred to the Supplementary material.

CE P

No clear outliers for compounds of different charge classes were observed in the correlation between log NSB and log Dbrain (pH 7.4) indicating that the applied brain

AC

homogenate provides a fairly good surrogate for brain tissue, despite the presence of dodecane. Alkanes (e.g. dodecane) have been shown to change the membrane fluidity and, at higher concentrations, the long range order of phospholipids in a membrane (Kirk and Gruner, 1985; Pope and Dubro, 1986; Pope et al., 1984; Sjolund et al., 1987). The effect of high concentrations of dodecane (as used herein) on the membrane structure has been investigated in great detail in our laboratory and further details will be published separately (Assmus, 2015; Assmus et al., Manuscript in preparation-a).

18

ACCEPTED MANUSCRIPT 3.4

The Prediction of VNS/fp in brain (PET data) using LIMBA Hitherto, we have based the analysis of prediction methods for radioligand binding in

T

brain on autoradiography data which were more readily available as compared to PET studies

IP

and therefore beneficial for the explanatory power of the validation. Moreover, it was

SC R

advantageous that NSB-estimates available from in vitro autoradiography were not confounded by active transport mechanisms at the BBB because brain slices were directly exposed to the radioligand incubation medium. In addition, no confounding radiometabolites

NU

were expected. However, we have shown that results obtained from the autoradiography

MA

experiments were sensitive to the assay conditions.

In order to test the applicability of LIMBA for predicting in vivo NSB observed by PET, a

D

database of 10 well characterized PET tracers studied in living human was compiled from

TE

literature and comprised tracers acting on a variety of molecular targets in the CNS (Table 3). Measured LIMBA- log Dbrain (pH 7.4) values as well as log VNS and f p data for the investigated

CE P

PET tracers are provided in Table 3, along with some key physicochemical properties

AC

including the cross-sectional areas, AD .

Table 3

It is well known that compounds with AD > 70Å2 are associated with reduced brain uptake as a result of reinforced efflux by P-gp (Fischer et al., 1998; Seelig, 2007). This would lead to VNS f p levels below those reflecting only passive drug-tissue distribution under steady state.

The relationship between VNS f p as well as log VNS f p and log Dbrain (pH 7.4) for compounds with AD values below and above the critical AD threshold is shown in Fig. 2 A and Fig. 2 B, respectively. Fig. 2 19

ACCEPTED MANUSCRIPT

Linear regression analysis of the log-transformed data demonstrated excellent correlation statistics (R2 = 0.88, SD = 0.21) for compounds with AD ≤ 70 Å2, which we classified as non-

2

(fallypride, PK 11195 and WAY 100635) were generally lower than

SC R

AD > 70 Å

IP

T

substrates for P-gp based on AD . In contrast, VNS f p values observed for compounds with

VNS f p values predicted by log Dbrain (pH 7.4). In this regard, it is important to note that

NU

LIMBA is only capable of assessing drug distribution governed by passive mechanisms since active transport proteins are not functional in the brain homogenate applied due to the

MA

presence of dodecane and the absence of ATP. Accordingly, any active efflux (influx) transport processes present in vivo will translate into an overestimation (underestimation) of

D

VNS f p values by LIMBA, as observed for fallypride, PK 11195 and WAY 100635. All three

TE

compounds have experimentally been identified as substrates (Elsinga et al., 2005; Piel et al.,

CE P

2014) or modulators (PK 11195, (Walter et al., 2005)) of the efflux transport protein P-gp which is in line with our classification based on AD . With regard to the PET tracers with 2

AD ≤ 70 Å , literature supported our classification as non-P-gp substrates in case of rolipram

AC

(Fujita et al., 2012), diprenorphine (Hassan et al., 2009) , raclopride (Ishiwata et al., 2007) and DASB (no literature with contrary information was found). For Spiperone, a borderline compound in terms of AD ( AD = 70 Å2), efflux by P - gp was reported (Xue Y. et al., 2004) and a slightly, but not significant overestimation of VNS f p by log Dbrain (pH 7.4) was observed (Fig. 2). Active efflux in vivo in rodents was reported for flumazenil (Froklage et al., 2012; Ishiwata et al., 2007), which is in contrast to our classification. However, in vitro transport studies using MDCKII cells transfected with human multidrug resistance (MDR1) genes suggested that flumazenil cannot be considered as substrate of human P-gp.

Since the

log Dbrain (pH 7.4) value for flumazenil was below the applicability range of LIMBA a direct

20

ACCEPTED MANUSCRIPT comparison between log Dbrain (pH 7.4) and the VNS f p data could not be made. Further research is therefore required and it is worth noting that the classification of whether or not a

T

compound is a substrate of P-gp is heavily dependent on the compound concentration (Li-

IP

Blatter et al., 2009). In summary, LIMBA- log Dbrain (pH 7.4) values were highly predictive for

SC R

in vivo VNS f p data in human brain with respect to compounds classified as non-P-gp substrates ( AD ≤ 70 Å2). This is in line with the results obtained from in vitro autoradiography

NU

experiments showing excellent performance of LIMBA for predicting the NSB in rodent brain

3.5

MA

for tritium-labelled compounds. Limitations and future perspectives

D

The likely development success of potential PET tracers depends on a multitude of factors

TE

apart from NSB. The list of requirements is long and includes e.g. sufficient target density and affinity, target selectivity, low protein binding, brain permeability, resistance to rapid

CE P

metabolism and absence of brain penetrant metabolites (Ametamey and Schubiger, 2006). The aim for achieving low NSB, on the one hand, and a high affinity for the target site as well

AC

as high brain permeability, on the other hand, are potentially conflicting and thus the need for balancing lipophilicity cannot be overemphasized (Wong and Pomper, 2003) . At Roche, a guideline for optimizing potential PET tracers is currently applied to identify compounds with undesirably high NSB: As a rule of thumb, the rejection of compounds for later later stage in vivo studies is indicated for compounds showing log Dbrain (LIMBA) values > 2.1. Further details regarding the rationale for the classification system are provided in the Supplementary material. For future applications, log Dbrain (pH 7.4) values available through LIMBA should not be considered in isolation but rather within the framework of an integrative strategy. Recently, a bio-mathematical modelling approach has been successfully applied to predict the image 21

ACCEPTED MANUSCRIPT quality in PET by integrating the properties of both, the tracer candidate and the biological system, into one prediction model (Guo et al., 2009). An important input parameter was the

T

unbound fraction of PET tracers in brain, as determined by equilibrium dialysis. LIMBA

IP

allows for the estimation of non-specific brain tissue binding at higher throughput and with

SC R

lower consumption of brain homogenate as compared to homogenate dialysis and we therefore propose LIMBA as a novel assay for the prediction of in vitro and in vivo NSB of

AC

CE P

TE

D

MA

NU

PET tracer candidates.

22

ACCEPTED MANUSCRIPT 4 Conclusions In summary, the novel LIMBA assay yields log Dbrain (pH 7.4) values in good agreement

IP

T

with NSB-estimates in brain, as demonstrated on a 30-compound dataset studied with in vitro

SC R

autoradiography. Moreover, excellent prediction accuracy was achieved for the estimation of VNS f p in in vivo PET, provided that tracers are not classified as P-gp substrates. Since

LIMBA works with minimum consumption of brain homogenate (1.2 μL) and, moreover,

NU

does not require radioactive labeling, reliable NSB-estimates can be obtained at higher

MA

throughput as compared to autoradiography and equilibrium dialysis studies. Tracer candidates with an unfavourably high NSB can be identified with LIMBA, however, other factors such as brain permeability as well as sufficient receptor density and target affinity are

TE

D

equally important for the success of a PET tracer. Log Dbrain (pH 7.4) values may thus be used in conjunction with other physiological and physicochemical parameters to predict the image

CE P

quality in PET by means of tracer kinetic modelling and simulation. Therefore, LIMBA holds much promise to accelerate the optimization of novel PET tracers in order to exploit the full

AC

potential of PET neuroimaging for drug discovery and development.

23

ACCEPTED MANUSCRIPT FIGURE LEGENDS Fig.1. Nonspecific binding under blocking conditions (autoradiography data) as a function

T

of A: octanol/water distribution coefficient, log Doct (pH 7.4), and B: brain tissue/water

IP

distribution coefficient, log Dbrain (pH 7.4), measured with the novel assay (LIMBA). The color

SC R

and the shape of the symbols correspond to the charge class of the compounds (neutral:grey; basic: blue; zwitter-ionic:green) and the exposure time of brain slices to a washing solution

NU

after incubation with the tritiated tracer (twash = 2 - 4 min: asterisk; twash = 7 - 10 min: triangle). Black lines are linear fits to the data (twash = 2 - 4 min: black solid line; twash = 7 - 10 min:

MA

black dashed line). The red dash-dotted line indicates the line of identity.

D

Fig.2. In vivo non-specific binding in PET as a function of the brain tissue / water

TE

distribution coefficient, log Dbrain (pH 7.4), for the PET tracers investigated (Table 3). Non-

CE P

specific binding in PET was expressed as A: volume of distribution of non-specifically bound radiotracer with respect to the unbound plasma concentration, VNS f p and B: log VNS f p . Size

AC

of the symbols according to the cross-sectional area, AD , of the ligands; compounds with

AD > 70 Å2 (grey triangles) and AD ≤ 70 Å2 (black triangles) have a high likelihood to be apparent P-gp substrates and to be non-substrates for P-gp, respectively. The lines indicate linear fits to the data (only non-substrates for P-gp) yielding the following correlation statistics: A: VNS f p (PET) = 98.1 * logDbrain (LIMBA) - 65.0 (R2 = 0.89, SD = 23.7, n = 6); B: Log VNS f p (PET) = 0.83 * logDbrain (LIMBA) + 0.46 (R2 = 0.88, SD = 0.21, n = 6).

24

ACCEPTED MANUSCRIPT FIGURES

4

14

13 12

15 30

27 25 4 24

2

7

21 16

1

2

9 26 3

23 18 19 17 20

11

29 28 8 22

0

T 12

2

5

1

4 7

13 15

30 11

9

27

14

8

25

28

26 21 23 3 18 24 22 19 20 2 17 16

29

0

0

1

2

3

4

5

200

D

Fig.1.

A

180 160

TE

1

2

3

4

LogDbrain LIMBA (pH 7.4)

DASB

80

CE P

100

MDL-100907

60

Raclopride

40

Diprenorphine Rolipram

0.5

AC

0 0.0

PK 11195

Fallypride WAY-100635

1.0

1.5

2.0

LogDbrain (pH 7.4), LIMBA

B

2.5

Spiperone

DASB

2.0

Log VNS/fp,PET

120

20

2.5

Spiperone

140

Fig.2.

0

MA

LogDoct (pH 7.4)

VNS/fp,PET

3

SC R

3 5

Log NSB autoradiography (fmol/mg protein)

4

B

IP

A

NU

Log NSB autoradiography (fmol/mg protein)

5

MDL-100907

1.5

PK 11195

Diprenorphine

1.0

Fallypride

Rolipram Raclopride

WAY-100635

0.5 0.0 0.0

0.5

1.0

1.5

2.0

2.5

LogDbrain (pH 7.4), LIMBA

25

ACCEPTED MANUSCRIPT

AC

CE P

TE

D

MA

NU

SC R

IP

T

TABLES

26

ACCEPTED MANUSCRIPT

Table 1. Target sites, key physicochemical properties (including log Doct and log Dbrain at pH 7.4) and in vitro autoradiography binding data for the

IP

T

ligands investigated. The compounds are listed below according to washing time, twash, and within each class according to increasing order of NSB.

Target

MW

Charge classa

pKa b

clogPc

logDoct (pH 7.4)d

Striatum

0.43 ± 0.02

46.4 ± 11.1

4

11.8

Cortex

MA N

Hippocampus

2.56

2.73

1.23 ± 0.12

214.1 ± 17.1

4

1.5

Cortex

405.2 ± 14.0

4

1.9

Cortex

388.5

B

9.14

6

D3 receptor

367.5

B

8.79

7

GABA(A) receptor

326.3

N

8

GABA(A) receptor

351.4

N

9

NK 3 receptor

447.9

N

10

NK3 recpetor

11

NK3 recpetor

380.5

N

12

NK3 recpetor

458.9

N

13

NK3 recpetor

602.5

B

14

NK3 recpetor

475.3

N

15

mGluR5 receptor

16

D2 receptor

347.2

Z

5.95/9.46

4.06

1.10

17

OX2 receptor

454.5

N

4.54

3.43

18

GlyT1 transporter

468.0

N

3.71

2.05

19

confidential

401.0

N

5.12

3.41

d1

-0.35

-0.64

2.12

1.66

d2

2.82

2.99

1.61

64.3 ± 5.0

4

4.1

Hippocampus

1.71

0.35 ± 0.08

65.8 ± 17.6

4

2.2

Striatum

0.95

<0.35

70.2 ± 11.4

4

1.8

Striatum

3.14

2.00

1.39

3.77

2.55

6.26

4.35

2.08 ± 0.01

590.7 ± 150.1

4

1.3

Cortex

3.25

3.09

1.93 ± 0.12

957.8 ± 108.3

4

2.4

Cortex

2.45 ± 0.11

2249.7 ± 158

4

2.8

Cortex

AC

CE P

2.47

4.15

4.1

Hippocampus

4.97

N

4

3.8

N

D3 receptor

343.8

13.9 ± 0.7

28.6

362.4

5

b3

Hippocampus

4

4.67

8.49

367.6

4

N

GABA(A) receptor

b2

2

75.2 ± 9.6

559.7

4

3.89

7.6 ± 2.4

119.6 ± 11.1

9.47/7.31

b3

<0.35

1.70 ± 0.02

B

NK3 recpetor

7.98

[min]

0.80 ± 0.07

221.3

3

B

[fmol/mg]

3.35

10.7 /7.73

582.5

ROI

3.12

B

D3 receptor

b2

TB (ROI) / NSB (blocking)

0.70 ± 0.02

351.4

2

b2

twash

1.94

M1 recpetor

b1

NSB (blocking)

0.57 ± 0.16

1

b1

Autoradiography data

TE D

[g/mol]

LIMBA logDbrain (pH 7.4)

US

No.

CR

Physicochemical properties

d2

<0.35

e1

5.33

3.92

3.89

1.88

2.16 ± 0.04

3151.1 ± 385.3

4

1.8

Cortex

4.86

3.78

2.54 ± 0.03

1560.2 ± 220.2

7

1.8

Striatum

1.06 ± 0.06

15.8 ± 1.9

10

82.5

Striatum

2.27

1.19 ± 0.03

16.8 ± 0.6

10

6.4

Hippocampus

2.26

0.92 ± 0.11

22.1 ± 1.3

10

162.8

Colliculus

2.66

0.87 ± 0.03

22.7 ± 1.6

10

1.4

Lateral septal

d2

27

ACCEPTED MANUSCRIPT

Target

MW

Charge classa

pKa b

clogPc

logDoct (pH 7.4)d

LIMBA logDbrain (pH 7.4)

20

confidential

409.9

N

5.2

3.59

2.51

1.03 ± 0.12

500.4

B N

10.27/9 3.75

TB (ROI) / NSB (blocking)

ROI

24.9 ± 3.1

10

3.0

Lateral septal

1.13 ± 0.10

37.3 ± 12.0

10

53.2

Hippocampus

3.16

d2

1.13 ± 0.63

46.9 ± 5.3

10

167.7

Colliculus

4.70

2.61

d2

1.53 ± 0.08

50.4 ± 4.4

10

64.8

Colliculus

2.92

2.52

1.22 ± 0.19

55.5 ± 5.8

10

70.7

Pons

1.39 ± 0.02

61.8 ± 15.3

10

95.3

Striatum

1.46 ± 0.05

79.6 ± 17.7

10

173.3

Striatum

3.15

23

GlyT1 transporter

350.5

B

24

GlyT1 transporter

497.5

N

25

D1 receptor

287.8

B

9.26/7.73

4.02

1.82

26

PDE-10 enzyme

388.4

N

5.74

1.37

2.61

27

PDE-10 enzyme

402.4

N

2.92

2.48

MA N

8.3

twash

2.23

4.30

b3

IP

GlyT1 transporter

339.5

CR

22

NMDA 2B receptor

NSB (blocking)

d3

US

21

T

No.

1.18 ± 0.00

93.1 ± 5.0

10

82.9

S.nigra

d2

GlyT1 transporter

472.5

N

3.30

3.18

2.22 ± 0.02

10

22.8

Colliculus

29

GlyT1 transporter

428.5

N

3.71

3.45d2

2.41 ± 0.00

180.4 ± 34.8

10

16.8

Colliculus

30

PDE-10 enzyme

394.4

N

4.00

3.17

2.05 ± 0.01

787.7 ± 214.2

10

41.0

S.nigra

a

TE D

28

155.8 ± 22.3

N: neutral (less than 3 % ionization at pH 7.4); A: acid with ApK a < 8.9; B: base BpKa >6; Z: zwitterions with ApKa <8.9 and BpKa >6. If not otherwise stated, the pKa was measured potentiometrically as described elsewhere (Bendels et al., 2006). The first pKa is the acidic pKa, the second is the basic pKa. If only one value is given, this is the basic pKa. b1 Potentiometric pKa in 13% methanol; b2 Calculated pKa (Moka, continuously trained with in-house data); b3 In-house SGA pKa in 10% methanol. c ClogP v4.71 Daylight. d Measured in-house with a miniaturized shake flask assay (50mM TAPSO, 5% (v/v) DMSO, pH 7.4).d1 large-scale shake flask, d2 CAMDIS-logDoct (in-house assay, 50mM TAPSO, 1% (v/v) DMSO, pH 7.4), d3 pH-metric e1 Compound probably precipitated

AC

CE P

b

28

ACCEPTED MANUSCRIPT

Table 2. Correlation statistics for the data shown in Fig. 1A and Fig.1B. logDoct(pH 7.4)

logDbrain(pH 7.4)

T

twash

R2

SD

Regression equation

R2

SD

2-4

logNSB = 0.36*logDoct+1.33 (n=11)

0.23

0.69

logNSB = 0.88*logDbrain+1.15 (n=11)

0.82

0.34

7-10

logNSB = 0.65*logDoct+0.13 (n=16)

0.56

0.39

0.76

0.29

SC R

IP

Regression equation

CE P

TE

D

MA

NU

logNSB = 0.92*logDbrain+0.47 (n=16)

AC

[min]

29

ACCEPTED MANUSCRIPT

Table 3. Target sites, key physicochemical properties (including log Doct and LIMBA - log Dbrain at pH 7.4) and in vivo PET data for the ligands

MW

ADa

[g/mol]

[Å2]

Charge class

pKa b

clogP

IP CR

Physicochemical properties

Target

logDoct (pH 7.4)c

US

No.

MA N

Compound

T

investigated.

LIMBA logDbrain (pH 7.4)

PET data fp

Log VNS

Reference regione

Model

Reference

[ml plasma/ g tissue]

1

SERT transporter

283.4

65

B

8.73c1

3.21

2.52±0.03

1.72 ± 0.04

0.094

2.00

Cerebellume1

1TC model

(Frankle et al., 2006)

Diprenorphinee

2

Opiate receptor

425.6

70

B

9.57/8.48

2.66

2.43±0.01

1.00 ±0 .04

0.300

1.14

Blockinge2

2TC model

(Jones et al., 1994)

Fallypride

3

D2/D3 receptor

364.5

90

B

8.22

3.18

2.02±0.03

1.20 ± 0.03

0.070

0.95

Cerebellume3

1TC model

(Slifstein et al., 2010)

Flumazenil

4

GABA(A) recpetor

303.3

55

N

1.29

1.08±0.09

<0.35

0.049

-0.41

Blockinge4

2TC model

(Price et al., 1993)

MDL-100907

5

5-HT2A receptor

373.5

56

B

3.29

1.87±0.04

1.22 ± 0.05

0.305d1

1.74

Blockinge5

2TC model

(Hinz et al., 2007)

PK 11195

6

TSPO receptor

352.9

76

4.62

3.79±0.06

2.12 ± 0.05

0.012d1

1.43

Cortexe6

2TC model

(Kropholler Marc et al., 2006)

Raclopride

7

D2 recpetor

347.2

65

4.06

1.13±0.04

1.05 ± 0.04

0.0344

1.08

Cerebellume7

Ratio method

(Mawlawi et al., 2001)

S(+)-rolipram

8

PDE-4 enzyme

275.3

43

N

1.72

1.91±0.05

0.44 ± 0.14

0.265

0.92

Blockinge8

Logan method

(Parker et al., 2005)

Spiperone

9

D2 recpetor

395.5

70

B

8.29

2.82

2.55±0.04

2.18 ± 0.02

0.045

2.22

Cerebellume9

1TC model

(Perlmutter et al., 1997)

WAY-100635

10

5-HT1A receptor

422.6

98

B

6.71

4.09

3.05±0.03

1.39 ± 0.06

0.066

0.87

Cerebellume10

Extrapolation

(Giovacchini et al., 2009)

CE P

TE D

DASB

AC

9.02

N

Z

5.95/9.46

30

ACCEPTED MANUSCRIPT

a

Cross sectional area at pH 7.4 calculated as described by (Gerebtzoff and Seelig, 2006) if not otherwise stated, the pKa was measured potentiometrically. The first pKa is the acidic pKa, the second is the basic pKa. If only one value is given, this is the basic pKa.c1calulated pKa (Moka). c CAMDIS-logDoct (in-house assay, 50mM TAPSO, 1% (v/v) DMSO, pH 7.4) d fraction unbound in plasma taken from (Summerfield et al., 2008) e1 from mean reported VT estimate e2from regional values for K1,k2,k5 and k6, mean value from 18 brain regions, e3from reported VND estimate (baseline scan), e4from mean reported fND estimate (unbound fraction of tracer in brain) across volunteers (grey matter), e5from reported VND (mirtazapine scan) , e6from reported K1 and k2 estimates e7from reported V2 estimates, e8from reported VND(F+NS), mean from 6 brain regions, porcine brain, e9from mean reported fND estimate; e10from reported VWM (volume in cerebral white matter).

AC

CE P

TE D

MA N

US

CR

IP

T

b

31

ACCEPTED MANUSCRIPT Acknowledgements The authors thank Celine Sutter for providing a part of the autoradiography data and for

AC

CE P

TE

D

MA

NU

SC R

IP

T

valuable discussions.

32

ACCEPTED MANUSCRIPT References Ametamey, S.M., Schubiger, P.A., 2006. PET radiopharmaceuticals for neuroreceptor

T

imaging. Nuclear Science and Techniques 17, 143-147.

IP

Assmus, F., 2015. Artificial Tissue Binding Models: Development and Comparative

Optimization. PhD thesis, University of Basel.

SC R

Evaluation of High -Throughput Lipophilicity Assays and their Use for PET Tracer

Assmus, F., Fischer, H., Seelig, A., Manuscript in preparation-a. 31P and 1H-NMR studies

NU

on the molecular organization of lipids in the PAMPA permeation barrier. Assmus, F., Seelig, A., Fischer, H., Manuscript in preparation-b. Evaluation of a Novel Lipid

MA

Membrane Binding Assay (LIMBA) for the Assessment of Brain Tissue Distribution. Avdeef, A., Box, K.J., Comer, J.E.A., Hibbert, C., Tam, K.Y., 1998. pH-metric logP 10.

Res. 15, 209-215.

TE

D

Determination of liposomal membrane-water partition coefficients of ionizable drugs. Pharm.

CE P

Balon, K., Riebesehl, B.U., Muller, B.W., 1999. Drug liposome partitioning as a tool for the prediction of human passive intestinal absorption. Pharm. Res. 16, 882-888.

AC

Belli, S., Assmus, F., Wagner, B., Honer, M., Alvarez-Sanchez, R., Manuscript in preparation. Methodology and in vivo application of a partitioning assay with brain polar lipid microemulsion for the estimation of drug binding to brain tissue. Bendels, S., Tsinman, O., Wagner, B., Lipp, D., Parrilla, I., Kansy, M., Avdeef, A., 2006. PAMPA Excipient Classification Gradient Map. Pharmaceutical Research 23, 2525-2535. Borroni, E., Zhou, Y., Ostrowitzki, S., Alberati, D., Kumar, A., Hainzl, D., Hartung, T., Hilton, J., Dannals Robert, F., Wong Dean, F., 2011. Pre-clinical characterization of [(11)C]R05013853 as a novel radiotracer for imaging of the glycine transporter type 1 by positron emission tomography. NeuroImage.

33

ACCEPTED MANUSCRIPT Danhof, M., de Jongh, J., De Lange, E.C.M., Della Pasqua, O., Ploeger, B.A., Voskuyl, R.A., 2007.

Mechanism-based

pharmacokinetic-pharmacodynamic

modeling:

biophase

T

distribution, receptor theory, and dynamical systems analysis. Annual Review of

IP

Pharmacology and Toxicology 47, 357-400.

SC R

Dickson, C.J., Gee, A.D., Bennacef, I., Gould, I.R., Rosso, L., 2011. Further evaluation of quantum chemical methods for the prediction of non-specific binding of positron emission tomography tracers. Phys Chem Chem Phys 13, 21552-21557.

NU

Elsinga, P.H., Hendrikse, N.H., Bart, J., van Waarde, A., Vaalburg, W., 2005. Positron

MA

emission tomography studies on binding of central nervous system drugs and P-glycoprotein function in the rodent brain. Mol Imaging Biol 7, 37-44.

D

Fernandes, E., Barbosa, Z., Clemente, G., Alves, F., Abrunhosa, A.J., 2012. Positron emitting

TE

tracers in pre-clinical drug development. Current Radiopharmaceuticals 5, 90-98. Fischer, H., Gottschlich, R., Seelig, A., 1998. Blood-brain barrier permeation: molecular

CE P

parameters governing passive diffusion. Journal of Membrane Biology 165, 201-211. Fischer, H., Kansy, M., Wagner, B., 2006. Determination of high lipophilicity values.

AC

Frankle, W.G., Slifstein, M., Gunn, R.N., Huang, Y., Hwang, D.-R., Darr, E.A., Narendran, R., Abi-Dargham, A., Laruelle, M., 2006. Estimation of serotonin transporter parameters with 11C-DASB in healthy humans: reproducibility and comparison of methods. Journal of Nuclear Medicine 47, 815-826. Friden, M., Wennerberg, M., Antonsson, M., Sandberg-Ställ, M., Farde, L., Schou, M., 2014. Identification of positron emission tomography (PET) tracer candidates by prediction of the target-bound fraction in the brain. EJNMM Research 4, 50. Froklage, F.E., Syvanen, S., Hendrikse, N.H., Huisman, M.C., Molthoff, C.F., Tagawa, Y., Reijneveld, J.C., Heimans, J.J., Lammertsma, A.A., Eriksson, J., de Lange, E.C., Voskuyl,

34

ACCEPTED MANUSCRIPT R.A., 2012. [11C]Flumazenil brain uptake is influenced by the blood-brain barrier efflux transporter P-glycoprotein. EJNMMI Res 2, 12.

T

Fujita, M., Hines, C.S., Zoghbi, S.S., Mallinger, A.G., Dickstein, L.P., Liow, J.S., Zhang, Y.,

IP

Pike, V.W., Drevets, W.C., Innis, R.B., Zarate, C.A., Jr., 2012. Downregulation of brain

SC R

phosphodiesterase type IV measured with 11C-(R)-rolipram positron emission tomography in major depressive disorder. Biol Psychiatry 72, 548-554.

Gerebtzoff, G., Seelig, A., 2006. In silico prediction of blood-brain barrier permeation using

MA

Information and Modeling 46, 2638-2650.

NU

the calculated molecular cross-sectional area as main parameter. Journal of Chemical

Giovacchini, G., Conant, S., Herscovitch, P., Theodore William, H., 2009. Using cerebral

D

white matter for estimation of nondisplaceable binding of 5-HT1A receptors in temporal lobe

TE

epilepsy. Journal of nuclear medicine 50, 1794-1800. Guo, Q., Brady, M., Gunn, R.N., 2009. A biomathematical modeling approach to central

1715-1723.

CE P

nervous system radioligand discovery and development. Journal of Nuclear Medicine 50,

AC

Hassan, H.E., Myers, A.L., Coop, A., Eddington, N.D., 2009. Differential involvement of Pglycoprotein (ABCB1) in permeability, tissue distribution, and antinociceptive activity of methadone, buprenorphine, and diprenorphine: in vitro and in vivo evaluation. J Pharm Sci 98, 4928-4940. Hinz, R., Bhagwagar, Z., Cowen, P.J., Cunningham, V.J., Grasby, P.M., 2007. Validation of a tracer kinetic model for the quantification of 5-HT2A receptors in human brain with [11C]MDL 100,907. Journal of Cerebral Blood Flow & Metabolism 27, 161-172. Honer, M., Gobbi, L., Martarello, L., Comley, R.A., 2014. Radioligand development for molecular imaging of the central nervous system with positron emission tomography. Drug Discovery Today, Ahead of Print. 35

ACCEPTED MANUSCRIPT Innis, R.B., Cunningham, V.J., Delforge, J., Fujita, M., Gjedde, A., Gunn, R.N., Holden, J., Houle, S., Huang, S.-C., Ichise, M., Iida, H., Ito, H., Kimura, Y., Koeppe, R.A., Knudsen,

T

G.M., Knuuti, J., Lammertsma, A.A., Laruelle, M., Logan, J., Maguire, R.P., Mintun, M.A.,

IP

Morris, E.D., Parsey, R., Price, J.C., Slifstein, M., Sossi, V., Suhara, T., Votaw, J.R., Wong,

SC R

D.F., Carson, R.E., 2007. Consensus nomenclature for in vivo imaging of reversibly binding radioligands. Journal of Cerebral Blood Flow & Metabolism 27, 1533-1539. Ishiwata, K., Kawamura, K., Yanai, K., Hendrikse, N.H., 2007. In vivo evaluation of P-

NU

glycoprotein modulation of 8 PET radioligands used clinically. J Nucl Med 48, 81-87.

MA

Jiang, Z., Reilly, J., Everatt, B., Briard, E., 2011. A rapid vesicle electrokinetic chromatography method for the in vitro prediction of non-specific binding for potential PET

D

ligands. J Pharm Biomed Anal 54, 722-729.

TE

Jones, A.K.P., Cunningham, V.J., Ha-Kawa, S.-K., Fujiwara, T., Qi, L., Luthra, S.K., Ashburner, J., Osman, S., Jones, T., 1994. Quantitation of [11C]diprenorphine cerebral

CE P

kinetics in man acquired by PET using presaturation, pulse-chase and tracer-only protocols. Journal of Neuroscience Methods 51, 123-134.

AC

Kalvass, J.C., Maurer, T.S., 2002. Influence of nonspecific brain and plasma binding on CNS exposure: implications for rational drug discovery. Biopharmaceutics & Drug Disposition 23, 327-338.

Kirk, G.L., Gruner, S.M., 1985. Lyotropic effects of alkanes and headgroup composition on the Lα-HII lipid liquid crystal phase transition: hydrocarbon packing versus intrinsic curvature. J. Phys. (Les Ulis, Fr.) 46, 761-769. Kropholler Marc, A., Boellaard, R., Schuitemaker, A., Folkersma, H., van Berckel Bart, N.M., Lammertsma Adriaan, A., 2006. Evaluation of reference tissue models for the analysis of [11C](R)-PK11195 studies. Journal of cerebral blood flow and metabolism 26, 1431-1441.

36

ACCEPTED MANUSCRIPT Laruelle, M., Slifstein, M., Huang, Y., 2003. Relationships between radiotracer properties and image quality in molecular imaging of the brain with positron emission tomography.

T

Molecular imaging and biology 5, 363-375.

IP

Li-Blatter, X., Nervi, P., Seelig, A., 2009. Detergents as intrinsic P-glycoprotein substrates

SC R

and inhibitors. Biochim. Biophys. Acta, Biomembr. 1788, 2335-2344.

Liu, X., Van Natta, K., Yeo, H., Vilenski, O., Weller, P.E., Worboys, P.D., Monshouwer, M., 2009. Unbound drug concentration in brain homogenate and cerebral spinal fluid at steady

NU

state as a surrogate for unbound concentration in brain interstitial fluid. Drug Metab Dispos

MA

37, 787-793.

Mawlawi, O., Martinez, D., Slifstein, M., Broft, A., Chatterjee, R., Hwang, D.-R., Huang, Y.,

D

Simpson, N., Ngo, K., Van Heertum, R., Laruelle, M., 2001. Imaging human mesolimbic

TE

dopamine transmission with positron emission tomography: I. Accuracy and precision of D2 receptor parameter measurements in ventral striatum. Journal of Cerebral Blood Flow and

CE P

Metabolism 21, 1034-1057.

McCarthy, D.J., Halldin, C., J.D., A., M.E., P., 2009. Chapter 24 Discovery of Novel

513.

AC

Positron Emission Tomography Tracers. Annual Reports in Medicinal Chemistry 44, 501-

Obach, R.S., Lombardo, F., Waters, N.J., 2008. Trend analysis of a database of intravenous pharmacokinetic parameters in humans for 670 drug compounds. Drug Metab. Dispos. 36, 1385-1405. Parker, C.A., Matthews, J.C., Gunn, R.N., Martarello, L., Cunningham, V.J., Dommett, D., Knibb, S.T., Bender, D., Jakobsen, S., Brown, J., Gee, A.D., 2005. Behaviour of [11C]R(-)and [11C]S(+)-rolipram in vitro and in vivo, and their use as PET radiotracers for the quantificative assay of PDE4. Synapse (Hoboken, NJ, United States) 55, 270-279.

37

ACCEPTED MANUSCRIPT Patel, S., Hamill, T., Hostetler, E., Burns, H.D., Gibson, R.E., 2003. An in vitro assay for predicting successful imaging radiotracers. Mol Imaging Biol 5, 65-71.

T

Perlmutter, J.S., Stambuk, M.K., Markham, J., Black, K.J., McGee-Minnich, L., Jankovic, J.,

IP

Moerlein, S.M., 1997. Decreased [18F]spiperone binding in putamen in idiopathic focal

SC R

dystonia. Journal of Neuroscience 17, 843-850.

Piel, M., Schmitt, U., Bausbacher, N., Buchholz, H.G., Grunder, G., Hiemke, C., Rosch, F., 2014. Evaluation of P-glycoprotein (abcb1a/b) modulation of [(18)F]fallypride in MicroPET

NU

imaging studies. Neuropharmacology 84, 152-158.

MA

Pike, V.W., 1993. Positron-emitting radioligands for studies in vivo-probes for human psychopharmacology. J Psychopharmacol 7, 139-158.

D

Pope, J.M., Dubro, D.W., 1986. The interaction of n-alkanes and n-alcohols with lipid bilayer

TE

membranes: a deuterium NMR study. Biochim. Biophys. Acta, Biomembr. 858, 243-253. Pope, J.M., Walker, L.W., Dubro, D., 1984. On the ordering of n-alkane and n-alcohol

CE P

solutes in phospholipid bilayer model membrane systems. Chem. Phys. Lipids 35, 259-277. Price, J.C., Mayberg, H.S., Dannals, R.F., Wilson, A.A., Ravert, H.T., Sadzot, B., Rattner, Z.,

AC

Kimball, A., Feldman, M.A., Frost, J.J., 1993. Measurement of benzodiazepine receptor number and affinity in humans using tracer kinetic modeling, positron emission tomography, and [11C]flumazenil. Journal of Cerebral Blood Flow and Metabolism 13, 656-667. Rodgers, T., Leahy, D., Rowland, M., 2005a. Physiologically based pharmacokinetic modeling: Predicting the tissue distribution of moderate-to-strong bases. J. Pharm. Sci. 94, 1259-1276. Rodgers, T., Leahy, D., Rowland, M., 2005b. Tissue distribution of basic drugs: Accounting for enantiomeric, compound and regional differences amongst β-blocking drugs in rat. J. Pharm. Sci. 94, 1237-1248.

38

ACCEPTED MANUSCRIPT Rosso, L., Gee, A.D., Gould, I.R., 2008. Ab initio computational study of positron emission tomography ligands interacting with lipid molecule for the prediction of nonspecific binding.

T

J Comput Chem 29, 2397-2405.

SC R

fatty acids. Journal of Molecular Neuroscience 33, 32-41.

IP

Seelig, A., 2007. The role of size and charge for blood-brain barrier permeation of drugs and

Sjolund, M., Lindblom, G., Rilfors, L., Arvidson, G., 1987. Hydrophobic molecules in lecithin-water systems. I. Formation of reversed hexagonal phases at high and low water

NU

contents. Biophys J 52, 145-153.

MA

Slifstein, M., Kegeles, L.S., Xu, X., Thompson, J.L., Urban, N., Castrillon, J., Hackett, E., Bae, S.A., Laruelle, M., Abi-Dargham, A., 2010. Striatal and extrastriatal dopamine release

D

measured with PET and [18F] fallypride. Synapse (Hoboken, NJ, United States) 64, 350-362.

TE

Slifstein, M., Laruelle, M., 2001. Models and methods for derivation of in vivo neuroreceptor parameters with PET and SPECT reversible radiotracers. Nuclear Medicine and Biology 28,

CE P

595-608.

Summerfield, S.G., Lucas, A.J., Porter, R.A., Jeffrey, P., Gunn, R.N., Read, K.R., Stevens,

AC

A.J., Metcalf, A.C., Osuna, M.C., Kilford, P.J., Passchier, J., Ruffo, A.D., 2008. Toward an improved prediction of human in vivo brain penetration. Xenobiotica 38, 1518-1535. Summerfield, S.G., Read, K., Begley, D.J., Obradovic, T., Hidalgo, I.J., Coggon, S., Lewis, A.V., Porter, R.A., Jeffrey, P., 2007. Central nervous system drug disposition: the relationship between in situ brain permeability and brain free fraction. J Pharmacol Exp Ther 322, 205-213. Walter, R.B., Pirga, J.L., Cronk, M.R., Mayer, S., Appelbaum, F.R., Banker, D.E., 2005. PK11195, a peripheral benzodiazepine receptor (pBR) ligand, broadly blocks drug efflux to chemosensitize leukemia and myeloma cells by a pBR-independent, direct transportermodulating mechanism. Blood 106, 3584-3593. 39

ACCEPTED MANUSCRIPT Wong, D.F., Pomper, M.G., 2003. Predicting the success of a radiopharmaceutical for in vivo imaging of central nervous system neuroreceptor systems. Mol Imaging Biol 5, 350-362.

T

Xue Y., Yap C. W., Sun L. Z., Cao Z. W., Wang J. F., Z., C.Y., 2004. Prediction of P-

IP

Glycoprotein Substrates by a Support Vector Machine Approach. Journal of Chemical

SC R

Information and Modeling 44, 1497-1505.

Yata, N., Toyoda, T., Murakami, T., Nishiura, A., Higashi, Y., 1990. Phosphatidylserine as a determinant for the tissue distribution of weakly basic drugs in rats. Pharm. Res. 7, 1019-

AC

CE P

TE

D

MA

NU

1025.

40