Bumepamine, a brain-permeant benzylamine derivative of bumetanide, does not inhibit NKCC1 but is more potent to enhance phenobarbital's anti-seizure efficacy

Article abstract of DOI:10.1016/j.neuropharm.2018.09.025

Based on the potential role of Na-K-Cl cotransporters (NKCCs) in epileptic seizures, the loop diuretic bumetanide, which blocks the NKCC1 isoforms NKCC1 and NKCC2, has been tested as an adjunct with phenobarbital to suppress seizures. However, because of its physicochemical properties, bumetanide only poorly penetrates through the blood-brain barrier. Thus, concentrations needed to inhibit NKCC1 in hippocampal and neocortical neurons are not reached when using doses (0.1–0.5 mg/kg) in the range of those approved for use as a diuretic in humans. This prompted us to search for a bumetanide derivative that more easily penetrates into the brain. Here we show that bumepamine, a lipophilic benzylamine derivative of bumetanide, exhibits much higher brain penetration than bumetanide and is more potent than the parent drug to potentiate phenobarbital's anticonvulsant effect in two rodent models of chronic difficult-to-treat epilepsy, amygdala kindling in rats and the pilocarpine model in mice. However, bumepamine suppressed NKCC1-dependent giant depolarizing potentials (GDPs) in neonatal rat hippocampal slices much less effectively than bumetanide and did not inhibit GABA-induced Ca²⁺ transients in the slices, indicating that bumepamine does not inhibit NKCC1. This was substantiated by an oocyte assay, in which bumepamine did not block NKCC1a and NKCC1b after either extra- or intracellular application, whereas bumetanide potently blocked both variants of NKCC1. Experiments with equilibrium dialysis showed high unspecific tissue binding of bumetanide in the brain, which, in addition to its poor brain penetration, further reduces functionally relevant brain concentrations of this drug. These data show that CNS effects of bumetanide previously thought to be mediated by NKCC1 inhibition can also be achieved by a close derivative that does not share this mechanism. Bumepamine has several advantages over bumetanide for CNS targeting, including lower diuretic potency, much higher brain permeability, and higher efficacy to potentiate the anti-seizure effect of phenobarbital.

Full text of DOI:10.1016/j.neuropharm.2018.09.025

Accepted Manuscript
Bumepamine, a brain-permeant benzylamine derivative of bumetanide, does not
inhibit NKCC1 but is more potent to enhance phenobarbital’s anti-seizure efficacy
Claudia Brandt, Patricia Seja, Kathrin Töllner, Kerstin Römermann, Philip Hampel,
Markus Kalesse, Andi Kipper, Peter W. Feit, Kasper Lykke, Trine Lisberg Toft-Bertelsen, Pauliina
Paavilainen, Inkeri Spoljaric, Martin Puskarjov, Nanna MacAulay, Kai Kaila, Wolfgang Löscher
PII:
S0028-3908(18)30677-4
10.1016/j.neuropharm.2018.09.025
NP 7350
DOI:
Reference:
To appear in:
Neuropharmacology
Received Date:
Accepted Date:
18 July 2018
16 September 2018
Please cite this article as: Claudia Brandt, Patricia Seja, Kathrin Töllner, Kerstin Römermann, Philip
Hampel, Markus Kalesse, Andi Kipper, Peter W. Feit, Kasper Lykke, Trine Lisberg Toft-Bertelsen,
Pauliina Paavilainen, Inkeri Spoljaric, Martin Puskarjov, Nanna MacAulay, Kai Kaila, Wolfgang
Löscher, Bumepamine, a brain-permeant benzylamine derivative of bumetanide, does not inhibit
NKCC1 but is more potent to enhance phenobarbital’s anti-seizure efficacy, Neuropharmacology
(2018), doi: 10.1016/j.neuropharm.2018.09.025
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
Bumepamine, a brain-permeant benzylamine derivative of bumetanide, does not inhibit
NKCC1 but is more potent to enhance phenobarbital’s anti-seizure efficacy
Claudia Brandt1,2,*, Patricia Seja , Kathrin Töllner , Kerstin Römermann , Philip Hampel ,
3,*
1,2
1,2
1,2
4
4
1
5
5
Markus Kalesse , Andi Kipper , Peter W. Feit , Kasper Lykke , Trine Lisberg Toft-Bertelsen ,
3
3
3
5
3
Pauliina Paavilainen , Inkeri Spoljaric , Martin Puskarjov , Nanna MacAulay , Kai Kaila , and
Wolfgang Löscher1,2
1Department of Pharmacology, Toxicology, and Pharmacy, University of Veterinary Medicine
Hannover, Germany,
2Center for Systems Neuroscience, Hannover, Germany
3Program in Molecular and Integrative Biosciences, and Neuroscience Center (HiLife),
University of Helsinki, Finland
4Institute for Organic Chemistry, Leibniz Universität Hannover, Germany
5Department of Neuroscience, University of Copenhagen, Copenhagen,
Denmark
*
Equal contribution
Correspondence: Dr. W. Löscher, Department of Pharmacology, Toxicology and Pharmacy,
University of Veterinary Medicine, Bünteweg 17, D-30559 Hannover, Germany; Phone: +49-
5
11-856-8720; Fax: +49-511-953-8581; E-mail: wolfgang.loescher@tiho-hannover.de
Abbreviations: ADT, afterdischarge threshold; ASD, anti-seizure drug; BMP, bumepamine;
CCC, cation–chloride cotransporter; CNS, central nervous sytem; CSF, cerebrospinal cord;
GDP, giant depolarizing potential; KCC, K–Cl-cotransporter; MEST, maximal electroshock
seizure threshold; NKCC, Na–K–Cl cotransporter; SE, status epilepticus; TLE, temporal lobe
epilepsy
1

ACCEPTED MANUSCRIPT
Abstract
Based on the potential role of Na–K–Cl cotransporters (NKCCs) in epileptic seizures, the loop
diuretic bumetanide, which blocks the isoforms NKCC1 and NKCC2, has been tested as an
adjunct with phenobarbital to suppress seizures. However, because of its physicochemical
properties, bumetanide only poorly penetrates through the blood-brain barrier. Thus,
concentrations needed to inhibit NKCC1 in hippocampal and neocortical neurons are not reached
when using doses (0.1-0.5 mg/kg) in the range of those approved for use as a diuretic in humans.
This prompted us to search for a bumetanide derivative that more easily penetrates into the brain.
Here we show that bumepamine, a lipophilic benzylamine derivative of bumetanide, exhibits
much higher brain penetration than bumetanide and is more potent than the parent drug to
potentiate phenobarbital’s anticonvulsant effect in two rodent models of chronic difficult-to-treat
epilepsy, amygdala kindling in rats and the pilocarpine model in mice. However, bumepamine
suppressed NKCC1-dependent giant depolarizing potentials (GDPs) in neonatal rat hippocampal
2
+
slices much less effectively than bumetanide and did not inhibit GABA-induced Ca transients
in the slices, indicating that bumepamine does not inhibit NKCC1. This was substantiated by an
oocyte assay, in which bumepamine did not block NKCC1a and NKCC1b after both extra- and
intracellular application, whereas bumetanide potently blocked both variants of NKCC1.
Experiments with equilibrium dialysis showed high unspecific tissue binding of bumetanide in
the brain, which, in addition to its poor brain penetration, further reduces functionally relevant
brain concentrations of this drug. These data show that CNS effects of bumetanide previously
thought to be mediated by NKCC1 inhibition can also be achieved by a close derivative that does
not share this mechanism. Bumepamine has several advantages over bumetanide for CNS
targeting, including lower diuretic potency, much higher brain permeability, and higher efficacy
to potentiate the anti-seizure effect of phenobarbital.
Key words: Epilepsy; anti-seizure drugs; neonatal seizures; GABA; giant depolarizing potentials
2

ACCEPTED MANUSCRIPT
1
. Introduction
–
In the central nervous system (CNS), the intracellular chloride concentration ([Cl ] ) determines
i
the strength and polarity of GABA-mediated neurotransmission (Kaila et al., 2014). Most mature
−
–
neurons in the CNS actively extrude Cl and thus exhibit a low [Cl ] . This unique specialization
i
is a necessary condition for the generation of hyperpolarizing inhibitory postsynaptic potentials
by GABA receptors (Kaila et al., 2014b; Medina et al., 2014; Watanabe and Fukuda, 2015). The
A
–
precise regulation of [Cl ] is achieved, in part, by cation–chloride cotransporters (CCCs) encoded
i
by the SLC12 gene family, which include the inwardly directed Na–K–2Cl cotransporter NKCC1
and the outwardly directed neuron-specific K–Cl-cotransporter 2 (KCC2). Altered chloride
homeostasis, resulting from mutation or dysfunction of NKCC1 and/or KCC2, causes
depolarizing actions of GABA leading to neuronal hyperexcitability. This has been implicated in
the pathogenesis of several brain disorders, including neonatal seizures and adult types of partial
epilepsy, autism spectrum disorders, chronic pain, spinal cord lesions, Down’s syndrome, brain
trauma, cerebral edema formation, cerebral artery occlusion, and diabetic ketoacidosis (Ben-Ari
et al., 2012; Kaila et al., 2014a,b; Ben-Ari et al., 2017; Di Cristo et al., 2018). Therefore, drugs
–
that restore low [Cl ] have been suggested to provide novel therapies for a wide range of as yet
i
difficult-to-treat disorders. In neurons, the loop diuretic bumetanide specifically blocks the
–
NKCC1 chloride importer, thereby reducing [Cl ] levels (Blaesse et al., 2009). While
i
pharmacokinetic considerations strongly suggest otherwise (Löscher et al., 2013; Puskarjov et al.,
2
014), bumetanide has been suggested to restore GABAergic inhibition in vivo in many neuronal
types in physiological and pathological conditions (Ben-Ari et al., 2017). For treatment of
neonatal seizures, it has been proposed to combine bumetanide to boost the anti-seizure drug
(ASD) phenobarbital (Dzhala et al., 2008). The hypothesis has been that blocking first the
-
depolarizing driving force of GABA currents and Cl accumulation using bumetanide, and
3

ACCEPTED MANUSCRIPT
thereafter prolonging the GABA-induced conductance using phenobarbital would lead to a potent
anticonvulsant effect. This combination of drugs has been reported to be effective in some
animal models of neonatal and adult partial seizures (Cleary et al., 2013; Töllner et al., 2014) and
neonatal hypoxia-ischemia (Liu et al., 2012), but whether their cellular mode of action in vivo is
based on the mechanism described above is not known. Furthermore, in a cerebral-ischemia
model of neonatal seizures in mice, bumetanide failed as an adjunct and significantly blunted
phenobarbital’s anticonvulsant efficacy at P10 (Kang et al., 2015). In fact, there are several
reasons to believe that bumetanide does not reach the putative targets, i.e. hippocampal and
neocortical neurons, at an effective concentration in the studies cited above. K /IC of
i
50
bumetanide for NKCC1 is often cited to be around 100-300 nM (Puskarjov et al., 2014), although
the sensitivity of mammalian NKCC1 to bumetanide may vary over almost two orders of
magnitude, depending on animal species, cell type, activation state of NKCC1 and whether
measurements are done in native/endogenous NKCC1 or in heterologous expression systems
(Russell, 2000; Alvarez-Leefmans, 2012). For selective block of NKCC1 by bumetanide,
concentrations of 1-10 µM are often used (Russell, 2000; Blaesse et al., 2009). For the present
study, we used both 100-300 nM and 1 µM of bumetanide as a cutoff for pharmacokinetic
considerations.
As previously shown in dogs and rodents (Javaheri et al., 1993; Brandt et al., 2010; Li et
al., 2011; Cleary et al. 2013), bumetanide only poorly penetrates into the cerebrospinal fluid
(
CSF) and brain, which is a consequence of its high ionization rate (>99%) at physiological pH
and its high plasma protein binding (>95%), which restrict brain entry by passive diffusion
Puskarjov et al., 2014). In addition, active efflux transport of bumetanide at the blood-brain
(
barrier (BBB) further reduces its brain levels (Römermann et al., 2017). Thus, bumetanide’s
brain:plasma concentration ratio is only 0.01-0.02 or less, so that NKCC1-inhibitory brain
4

ACCEPTED MANUSCRIPT
concentrations are not achieved after systemic administration of bumetanide in neonatal or adult
seizure models, particularly when using the low systemic doses (~0.05-0.3 mg/kg) that are
associated with maximal diuretic effects in humans (Löscher et al., 2013; Puskarjov et al., 2014).
This promoted us to design lipophilic and uncharged prodrugs of bumetanide that
penetrate the BBB more easily than the parent drug and are converted into bumetanide in the
brain (Töllner et al., 2014). The most promising bumetanide prodrug, the N,N-
dimethylaminoethylester of bumetanide (BUM5), was shown to potentiate the anticonvulsant
effect of phenobarbital markedly better than bumetanide in adult models of epilepsy (Töllner et
al., 2014; Erker et al., 2016). However, brain levels of bumetanide following administration of
BUM5 were increased by a factor of about 3-5 in rodents and only moderately exceeded the
levels needed to effectively inhibit neuronal NKCC1 (Töllner et al., 2014). Therefore, we thought
about other strategies to overcome the poor brain penetration of bumetanide. One such strategy
would be the design of a bumetanide derivative that is only minimally ionized at physiological
pH, is more lipophilic than bumetanide, and therefore penetrates better into the brain.
The clinical use of bumetanide for treating brain diseases is further complicated by severe
adverse effects, namely high diuretic activity and ototoxicity (Löscher et al., 2013). During
studies on structure-activity relationships of aminobenzoic acid diuretics and related compounds,
Nielsen and Feit (1978) evaluated various benzylamine derivatives of bumetanide for diuretic
activity in dogs. These compounds lacked the carboxylic group of bumetanide, resulting in
markedly lower diuretic activity (Nielsen and Feit, 1978). We chose one of these benzylamine
derivatives, 3-butylamino-2-phenoxy-5-phenylaminomethyl-benzenesulfonamide (termed
bumepamine in the following; Fig. 1) for pharmacokinetic and pharmacodynamic experiments.
The rationale for selection of bumepamine was its about 80% lower diuretic activity in dogs
when compared to bumetanide (Nielsen and Feit, 1978) and its much higher lipophilicity and
5

ACCEPTED MANUSCRIPT
lower ionization rate at physiological pH, which should enable improved brain penetration and
related pharmacodynamic effects. This compound is characterized in the present study in
comparison to bumetanide.
2
2
. Materials and Methods
.1 Drugs
For all in vitro and in vivo experiments, bumetanide was obtained from Sigma (Munich,
Germany). Bumepamine (previously also termed “BUM13” [Lykke et al., 2015 and 2016]; see
Fig. 1) was first described by Nielsen and Feit (1978) and resynthesized for the present
experiments as described below.
Lipophilicity (logP) and acidic dissociation constant (pKa) of bumepamine were
calculated by the Molecular Operating Environment (MOE 2012.10; Chemical Computing Group
Inc., Montreal, QC, Canada). For comparison, logP and pKa of bumetanide were calculated by
the same software and verified by published experimental approaches (Orita et al., 1976).
Depending on the experiment, bumepamine was either used as free base, as oxalate salt or
as HCl. In vitro and in vivo experiments showed that these forms of bumepamine exerted the
same effects, so that they are not further differentiated in the following. All in in vivo doses (in
mg/kg) shown in this study relate to the free base. For in vivo experiments, bumetanide and
bumepamine were dissolved in a vehicle containing 10% ethanol, 10% Solutol® HS 15 (a non-
ionic solubilizer consisting of ethoxylated 12-hydroxystearic acid; BASF, Ludwigshafen,
Germany), and 80% distilled water. Except some pharmacokinetic experiments with oral
administration, bumetanide and bumepamine were injected i.v. in all in vivo experiments. For in
vitro experiments, bumetanide and bumepamine were dissolved in DMSO and diluted with
6

ACCEPTED MANUSCRIPT
medium (see below). Phenobarbital (used as sodium salt; Serva, Heidelberg) was dissolved in
distilled water and injected i.p.
AqB013, a 4-aminopyridine derivative of bumetanide (Fig. 1) that has previously been
described as an inhibitor of aquaporin-4 (AQP4; Migliati et al., 2009), was kindly provided by
Dr. Andrea J. Yool (University of Adelaide, Australia) and dissolved in DMSO.
2
.2 Synthesis of bumepamine
Bumetanide (500 mg, 1.37 mmol) was dissolved in 9 mL dry DMF and DiPEA (765 μL,
.39 mmol), followed by aniline (250 μL, 2.74 mmol) were added and the mixture cooled on an
4
ice-bath. COMU (707 mg, 1.65 mmol) was added in one portion and the reaction was gradually
warmed to room temperature and stirred for 16 h. The reaction was quenched with saturated
aqueous NaHCO solution and extracted twice with EtOAc. The combined organic layers were
3
washed with water, brine and dried over Na SO . The crude product was purified by column
2
4
chromatography (toluene:EtOAc:Et N, 3:2:0.01). The desired product bumepamine (540 mg,
3
1
1
.22 mmol) was obtained in 89% yield as a slightly yellow solid. H NMR (400 MHz, d -MeOH)
4
δ 7.79 (d, J = 2.1 Hz, 1H), 7.68 – 7.71 (m, 2H), 7.50 (d, J = 2.1 Hz, 1H), 7.36 – 7.40 (m, 2H),
7
6
.28 – 7.32 (m, 2H), 7.14 – 7.19 (m, 1H), 7.05 – 7.09 (m, 1H), 6.93 – 6.96 (m, 2H), 3.17 (t, J =
1
3
.7 Hz, 2H), 1.40 – 1.48 (m, 2H), 1.13 – 1.21 (m, 2H), 0.83 (t, J = 7.3 Hz, 3H). C NMR (100
MHz) d -MeOH δ 167.83, 157.87, 144.02, 140.63, 139.74, 138.38, 133.97, 130.61, 129.82,
4
1
25.77, 124.01, 122.45, 116.66, 115.22, 114.88, 43.71, 32.03, 20.87, 14.03. HRMS (ESI) calcd
+
for C H N O S [M+H] 440.1644, found 440.1646.
2
3
26
3
4
All reactions were carried out under positive pressure of argon, with oven-dried glassware
using standard Schlenk techniques. Dry solvents were obtained from Acros Organics in Acroseal
bottles and used without further purification. Aniline was distilled over CaH and used directly
2
7

ACCEPTED MANUSCRIPT
after distillation. Bumetanid was purchased from Alfa Aesar and used without further
purification. Reactions were monitored by TLC (0.2 mm, silica gel 60, F , aluminium-backed,
2
54
Machery-Nagel), with detection by UV light (254 nm). Flash chromatography was performed on
Merck silica (60M). NMR spectra were recorded on an AMX-400 instrument (Bruker) at 400
1
13
MHz or at 100 MHz for H and C, respectively. Deuterated methanol was used as solvent and
1
spectra were calibrated against the residual solvent peak (3.31 ppm and 49.00 ppm for H and
13C, respectively). Data are presented as follows: chemical shift (δ, ppm), multiplicity (s =
singlet, d = doublet, t = triplet, q = quadruplet, m = multiplet), coupling constant (reported in
Hertz (Hz)), integration. Electrospray (ESI-HRMS) mass spectra were obtained using an Applied
Biosystems API 150EX LC/MS system.
2
.3 Animals
NMRI mice were obtained from Charles River (Sulzfeld, Germany) at a weight of 21-25 g and
adapted at least one week to the laboratory before being used in experiments. Wistar Unilever
rats were purchased at a body weight of 200–220 g from Harlan Laboratories (Horst,
Netherlands). Experiments were performed according to the EU council directive 210/63/EU and
the German Law on Animal Protection (“Tierschutzgesetz”). Ethical approval for the study was
granted by an ethical committee (according to §15 of the Tierschutzgesetz) and the government
agency (Lower Saxony State Office for Consumer Protection and Food Safety; LAVES)
responsible for approval of animal experiments in Lower Saxony (reference numbers for this
project: 14/1659 and 15/1825). Frog experiments were performed according to the same
principles (reference number for this project 15/1825). All efforts were made to minimize both
the suffering and the number of animals.
Rodents were housed in groups of maximal 10 under controlled conditions (temperature:
8

ACCEPTED MANUSCRIPT
2
3 ± 1°C; humidity: 50%-60%), under a 12-h light-dark cycle (lights on at 6:00 h). Standard
laboratory chow (Altromin 1324 standard diet, Altromin Spezialfutter GmbH, Lage, Germany)
and tap water were provided ad libitum. Female mice and rats were used to allow comparisons
with our previous studies on bumetanide and its derivatives (Brandt et al., 2010; Töllner et al.,
2
014; Töpfer et al., 2014; Erker et al., 2016; Römermann et al., 2017). Furthermore, female mice
and rats are more easily to maintain in groups during prolonged experiments on chronic epilepsy,
which is an advantage in terms of animal welfare. Mice and rats were housed in rooms without
males in order to keep them acyclic or asynchronous with respect to their estrous cycle (Kücker
et al., 2010).
2
.4 Dose selection for in vivo experiments
Previous experiments with bumetanide had shown that mice and rats metabolize bumetanide
much more rapidly than humans, so that high doses (at least 5-10 mg/kg) of bumetanide have to
be administered to induce diuresis and to obtain any meaningful brain concentrations in rats
(Olsen, 1977; Brandt et al., 2010; Töllner et al., 2014). Therefore, for the present study, the
intended dosage for all in vivo experiments was 10 mg/kg injected via the i.v. route except for the
diuresis experiments, in which also a lower dose (1 mg/kg) was administered i.v. For
bumepamine, the same doses were used. Injection volumes were 10 ml/kg in mice and 3 ml/kg in
rats; injection volumes were reduced by 50% in case of combined drug administration. Because
of the higher MW of bumepamine (see below), its molar concentration at identical
weight/volume is 86% of that of bumetanide.
2
.5 Pharmacokinetic experiments with bumetanide, bumepamine, and phenobarbital in
mice
9

ACCEPTED MANUSCRIPT
Bumetanide and bumepamine were dissolved as described above and injected at a dose of 10
mg/kg i.v. in groups of NMRI mice. Based on previous pharmacokinetic experiments with
bumetanide in mice (Töllner et al., 2014), blood and brain were sampled at 30 and 60 min
following i.v. injection. Bumetanide and bumepamine were analyzed in plasma and brain by high
performance liquid chromatography (HPLC) as recently described in detail (Brandt et al., 2010;
Römermann et al., 2017). The mobile phase consisted of 0.05 M phosphate buffer/acetonitrile
(70:30 for bumetanide, 35:65 for bumepamine by volume, pH adjusted at 5.6 for both). As
standard for HPLC analysis, we used a commercial solution (0.5 mg/ml) of the sodium salt of
bumetanide (Burinex), which was kindly provided by Leo Pharma (Ballerup, Denmark).
In additional experiments, the oral bioavailability of bumetanide and bumepamine was
determined in mice. For this purpose, groups of mice received 10 mg/kg of either drug p.o. (by
gavage using a stomach tube) and blood was sampled at 30 and 60 min after administration.
Bioavailability was calculated by comparing drug levels after i.v. and p.o. administration, using
PK Solutions 2.0TM (Summit Research Services, Montrose, CO, USA). This program was also
used to calculate the elimination half-life and apparent volume of distribution of bumetanide and
bumepamine after i.v. administration.
Furthermore, because we evaluated combinations of phenobarbital with either bumetanide
or bumepamine on seizure threshold (see below), we performed experiments in which we
determined phenobarbital in plasma and brain after administration of phenobarbital alone (30 min
after 10 mg/kg i.v.) or combinations of bumetanide or bumepamine (10 mg/kg i.v., administered
3
0 min before phenobarbital) and phenobarbital (10 mg/kg i.v.). Phenobarbital levels in plasma
and brain were always determined 30 min after administration by HPLC as described previously
Brandt et al., 2010).
For determining drug levels in brain, mice were decapitated, and therefore the residual
(
1
0

ACCEPTED MANUSCRIPT
blood in the brain may lead to erroneously high brain tissue levels, particularly in the case of
bumetanide. We therefore made experiments in which brain levels of bumetanide were compared
in brains from decapitated mice vs. mice that were perfused over 3 min with 20 ml 0.01 M PBS
(under anesthesia) before decapitation to eliminate blood from the brain. Average brain levels (30
min after i.v. administration of 10 mg/kg bumetanide) were 0.15 ± 0.03 µg/g following
decapitation vs. 0.14 ± 0.02 µg/g following perfusion, thus excluding residual blood as a
significant bias.
2
.6 Determination of fraction unbound in mouse plasma and brain by equilibrium dialysis
Equilibrium dialysis was performed as described previously (Kalvass and Maurer, 2002; Waters
et al., 2008) using Single-Use RED (Rapid Equilibrium Dialysis) Plates with 48 inserts
(ThermoScientific; Waltham, MA, USA). Each insert is comprised of two side-by-side sample
and buffer chambers separated by an O-ring-sealed vertical cylinder of dialysis membrane with a
molecular weight cutoff of 8.0 kDa, validated for minimal nonspecific binding. Plasma and brain
tissue were gathered from adult female CD1-mice. Plasma was not diluted, while brain tissue was
homogenized and diluted with two volumes of phosphate-buffered saline (PBS; pH = 7.2) and
stored at -20°C until dialysis. Additional experiments were performed with 1:10 and 1:20 brain
tissue dilutions to determine whether dilution affected the extent of tissue binding. Plasma and
brain homogenate samples were spiked with bumetanide (40 µg/ml) and the equimolar
concentration of bumepamine, respectively (50.8 µg/ml). After buffer compartments were loaded
with 400 µl of PBS, aliquots of 600 µl spiked plasma or brain homogenate were loaded into
sample compartments (n=8). Dialysis was performed at 37°C at 150 rpm, and 100 µl samples
were taken at 4 h (n=3), 6 h (n=3) and 8 h (n=2) from each compartment and stored at -20°C until
HPLC analysis. All experiments were repeated at least once.
1
1

ACCEPTED MANUSCRIPT
Equilibrium between buffer compartments and plasma/brain compartments was reached
within 4 h, so that data from 4 h, 6 h and 8 h did not differ and were averaged. The mean of the
concentrations, at which equilibrium was reached, were used to calculate the unbound fractions
(fu) of brain homogenate using the following equation
1
퐷
푓푢,푏푟푎푖푛
=
1
1
‒
1 +
(
(
푓
)
)
퐷
푢.푑ℎ
in which D is the dilution factor of brain homogenate and fu,dh is the concentration ratio between
buffer and diluted homogenate (Kalvass and Maurer, 2002; Reichel, 2009). Unbound fractions in
undiluted plasma equal the concentration ratio between buffer and plasma. Unbound
concentrations (cu,plasma and cu,brain) were calculated from the total plasma and brain concentrations
(ctotal,plasma and ctotal,brain) determined in the in vivo experiments (cf., section 2.5) using fu,plasma and
fu,brain as follows: 푐 = 푐 ∗ 푓푢.
푢
푡표푡푎푙
2
.7 Diuretic effect of bumetanide and bumepamine in mice
The diuretic effect of bumetanide and bumepamine, injected i.v. at either 1 or 10 mg/kg,
was determined in groups of NMRI mice. After drug injection, animals were placed on filter
paper (grade 1F, Munktell, Falum, Sweden) in a macrolon cage type 1 (one animal per cage). At
1
5, 30, 60, 90 and 120 min after injection, the weight of the filter paper was determined and the
animals were placed on a fresh filter paper. The difference between the initial weight and the
weight of the urine-drenched filter paper was calculated for each time point. The amount of urine
[g] for each time point was calculated by summing up the respective value with the preceding
values.
2
.8 Effect of bumetanide and bumepamine alone or in combination with phenobarbital on
1
2

ACCEPTED MANUSCRIPT
electrically induced seizures in epileptic mice
We tested the effects of bumetanide (10 mg/kg i.v.) and bumepamine (10 mg/kg i.v.) alone or in
combination with a low dose of phenobarbital (10 mg/kg i.v.) on electrically induced seizures in
epileptic and nonepileptic mice. The dose of phenobarbital was based on previous dose-effect
experiments in this model (Erker et al., 2016). Bumepamine was also tested at a lower dose (1
mg/kg i.v.). Mice were made epileptic by pilocarpine.
Systemic administration of convulsant doses of pilocarpine induces a status epilepticus
(SE) that is followed by development of epilepsy with spontaneous recurrent seizures in
subsequent weeks (Curia et al., 2008). As previously described for induction of SE by pilocarpine
in NMRI mice (Gröticke et al., 2007), we used a ramping-up dosing protocol that allows a more
individual dosing of pilocarpine compared to injection of the same high dose in each animal per
group, thereby reducing inter-individual variability in SE induction and mortality. In order to
avoid peripheral cholinergic effects, methyl scopolamine (1 mg/kg i.p.) was administered 30 min
before the application of pilocarpine. For induction of seizures, 100 mg/kg pilocarpine were
injected i.p. every 20 min until onset of SE. SE was defined as continuous limbic seizure activity,
which was occasionally interrupted by clonic forelimb seizures, generalized clonic-tonic seizures
or running and jumping seizures. The experiments were performed subsequently in three groups
of mice with 40 animals injected with pilocarpine and 20 sham-treated controls per group, which
received saline instead of pilocarpine. In a total of 120 mice in which pilocarpine was injected, all
animals developed SE. All mice received diazepam (10 mg/kg i.p.) after 90 min of SE (or 90 min
after the last saline injection in controls) to decrease mortality. Despite diazepam, about 40% of
the mice died during or after SE. On the days following SE, mice received subcutaneous saline
injections (1 ml/day) and moist chow to facilitate recovery.
Similar to the study of Blanco et al. (2009), we did not implant EEG electrodes for
1
3

ACCEPTED MANUSCRIPT
monitoring of spontaneous seizures developing after SE, because it is known that pilocarpine-
induced SE leads to development of epilepsy in almost all animals under the conditions used in
our experiments, which is an advantage of this model for the purpose of our study (Gröticke et
al., 2007). Furthermore, we wanted to avoid that the BBB disruption associated with EEG
electrode implantation affects the otherwise poor brain penetration of bumetanide (Löscher et al.,
2
013).
About 6 weeks following SE, i.e., at a time where all mice have developed epilepsy with
spontaneous recurrent seizures in this model (Gröticke et al., 2007), we started to test drugs for
their effects on the maximal electroshock seizure threshold (MEST; see below). Age-matched,
sham-control mice were used in parallel for comparison. The epileptic mice (and age-matched
controls) were repeatedly used in the MEST test at minimum intervals of one week. As
previously shown, repeated determination of MEST at such intervals does not affect seizure
threshold in this model (Löscher et al., 1991). Animal testing regulations for this study restricted
the number of drug or vehicle administrations to 6 in each individual mouse.
For testing drug and drug vehicle effects on MEST, drugs were dissolved as described
above and injected either 30 min (phenobarbital) or 60 min (bumetanide, bumepamine) before
onset of MEST determination in groups of nonepileptic and epileptic mice, respectively. When
bumetanide or bumepamine were administered together with phenobarbital, they were injected 30
min prior to phenobarbital and MEST was determined 30 min after phenobarbital. This
administration protocol was based on previous experiments with bumetanide and phenobarbital
(Töllner et al., 2014) with the aim to blocking first the depolarizing driving force of GABA
-
currents and Cl accumulation using bumetanide, and thereafter prolonging the GABA-induced
conductance using phenobarbital. Thus phenobarbital was administered at time of maximal brain
concentrations of bumetanide and bumepamine (see Results). In all experiments, the vehicles
1
4

ACCEPTED MANUSCRIPT
used for drug solutions were administered to determine vehicle control MESTs prior to and after
drug experiments.
MEST was determined by a staircase procedure as previously described (Löscher et al.,
1
991; Erker et al., 2016) by means of a stimulator (BMT Medizintechnik, Berlin, Germany) that
delivered a constant current (adjustable from 1-200 mA regardless of the impedance of the
animal) with sinusoidal pulses (50/sec) for 0.2 sec. Current administration was performed via
bilateral trans-corneal stimulation (using copper electrodes). Before trans-corneal stimulation, a
drop of tetracaine solution (2%) was administered to the eyes of the mouse for local anaesthesia.
Two minutes later, the mouse was restrained by hand to press the copper electrodes on the
corneae while the stimulus was applied by stepping on a foot pedal switch connected with the
stimulator. Electrodes were covered with soft leather and soaked with saline before each current
application. Directly after stimulation, the mouse was released from restraint in order to permit
observation of seizures. The stimulus intensity was varied by an up-and-down method in which
the current was lowered or raised by 0.06 log intervals according to whether the preceding mouse
did or did not exert a tonic hind limb extension. The data thus generated from groups of 17-20
nonepileptic control and 20-26 epileptic mice were used to calculate the CC (convulsant current
5
0
that induces a tonic hind limb seizure in 50% of the mice per group with confidence limits for
5% probability) as described previously (Erker et al., 2016). The group size decreased during
9
the course of the experiments because some mice died during MEST. In all experiments, mice
were closely observed for behavioral adverse effects.
2
.9 Effect of phenobarbital and bumepamine alone or in combination on fully amygdala-
kindled seizures in rats
We determined whether bumepamine, injected alone or in combination with phenobarbital, alters
1
5

ACCEPTED MANUSCRIPT
the seizure susceptibility of fully amygdala-kindled rats. Wistar rats were electrically kindled via
a bipolar depth electrode implanted in the basolateral nucleus of the amygdala as described in
detail previously (Töllner et al., 2011). Following kindling acquisition, the focal seizure threshold
(afterdischarge threshold, ADT) was determined by an ascending staircase method (Töllner et al.,
2
011). The ADT was defined as the current inducing an afterdischarge of at least 3 sec in the
electroencephalogram (EEG) recorded from the depth electrode in the amygdala. The ADT was
repeatedly determined twice per week after kindling acquisition until it was stable in each rat.
Then, bumepamine (10 mg/kg i.v.; with or without phenobarbital) was administered and ADT
was determined 30 min later. In experiments with combined treatment, bumepamine was injected
first, followed 30 min later by injection of phenobarbital, and then 30 min later ADT
determination. For phenobarbital, we chose a dose (10 mg/kg i.p.) that did not increase ADT
when injected alone. Each rat served as its own control with vehicle control ADT determination
3
-4 days before and after each drug experiment (see Töllner et al., 2011). In all experiments, rats
were closely observed for behavioral adverse effects.
2
.10 Comparison of the effects of bumetanide and bumepamine on giant depolarizing
potentials (GDPs) in the neonatal hippocampal slice
Extracellular recordings of field giant depolarizing potentials (fGDPs) in slices from Wistar rat
pups (P2 – 5) were performed as described in detail before (Töllner et al., 2014). In short,
postnatal day 2-5 (P2 - P5) Wistar rat pups were decapitated and the brains were quickly removed
and immersed in cold (0-4 °C) sucrose-based cutting solution containing (in mM): 87 NaCl, 2.5
KCl, 0.5 CaCl , 25 NaHCO -, 1.25 NaH PO , 7 MgCl , 50 sucrose, and 25 D-glucose,
2
3
2
4
2
equilibrated with 95 % O and 5 % CO . Horizontal hippocampal slices (thickness 400 µm) were
2
2
cut with a vibrating microtome (7000 smz-2, Campden instruments). The slices were allowed to
1
6

ACCEPTED MANUSCRIPT
recover in standard solution containing (in mM) 124 NaCl, 3.0 KCl, 2.0 CaCl , 25 NaHCO -, 1.1
2
3
NaH PO , 1.3 MgSO , and 10 ᴅ-glucose (pH 7.4 at 32 °C) for one hour at 34 °C and then stored
2
4
4
for up to 5 hours at room temperature.
Drugs were dissolved in DMSO (10 mM stock for bumepamine) and stored in -20˚C.
To avoid precipitation of bumepamine, the stock was applied into slightly alkaline standard
solution (before equilibration with 95% O + 5% CO ) at room temperature, after which the
2
2
solution was sonicated for 10 minutes.
Slices were transferred to a submerged-type recording chamber (volume <1 ml) perfused
with standard solution (32–33°C) at a rate of 3.5 ml/min (double-sided perfusion) and were let to
recover for a minimum of 10 minutes before starting the recording. In all recordings the
+
extracellular K concentration was 3.5 - 4 mM (by adding KCl from a 1 M stock solution) since a
slightly higher fGDP frequency (see Sipilä et al., 2005) was considered advantageous in
experiments where a reduction of fGDP frequency was expected to take place. Extracellular field
potential recordings were performed with conventional NaCl-filled (150 mM) glass capillary
microelectrodes (tip diameter 3-8 µm) placed in the stratum pyramidale of area CA3. Due to the
much slower effect of bumepamine in comparison to bumetanide (see Results), the duration of
these recordings was 40 minutes.
Data were aquired with 5-20 kHz sampling frequency and 0.1-1000 Hz filtering with the
Strathclyde Electrophysiology WinEDR (John Dempster, Glasgow, UK) program. Baseline was
recorded for a minimum of 10 minutes before bumepamine (2 µM) or bumetanide (2 µM, Tocris)
were bath applied.
fGDPs were manually detected from the local field potential recordings after 1-10 Hz
band pass filtering with a threshold individually selected for each recording (Clampfit, Molecular
Devices, Foster City, CA). fGDP area, which takes into account changes in event amplitude and
1
7

ACCEPTED MANUSCRIPT
duration, was quantified using numerical integration in Clampfit for each event (Spoljaric et al.,
2
017). The baseline fGDP area was normalized for the last 600 s before the drug application
(time -600 – 0; where t = 0 is the time of drug application). A sliding average of the fGDP
frequency was calculated for each recording using a window of 120 s moving by 10 s. The data
are presented as mean ± SEM and n indicates the number of slices (slices from 5-6 animals per
group). Because of a difference in their time course, the effect of the applied drugs was quantified
as a time-average for 5 minutes, starting 10-15 minutes (bumetanide) or 35 minutes (bumepamine
and DMSO) after the beginning of application. DMSO did not have any effect on fGDPs, even at
concentrations 5 times higher (100 μl/100 ml) than the maximum concentration added with the
drugs (20 µl/100ml).
2
.11 Comparison of the effects of bumetanide and bumepamine on GABA-induced Ca2+
transients in the neonatal hippocampal slice
To load hippocampal CA3 neurons with Fluo-4, horizontal brain slices (400 µm, from P0 - P5
rats) were incubated in standard solution with Fluo-4 AM (10 µM, Invitrogen/MolecularProbes)
for 15 minutes at RT. Slices were transferred to the recording chamber (submerged, one-sided
+
constant perfusion with standard solution (3.5 mM K ), perfusion at 3.5 ml/min, recording
temperature 32 ± 0.5°C) and washed for ≥ 20 minutes before recordings. The fluorescence
imaging system consisted of a LED light source (Cairn, UK; excitation wavelength 500 ±15 nm,
a 535 ± 12.5 nm band pass filter), an upright Zeiss AxioScope II (60x water immersion objective
n.a. 0.9) and an ORCA Flash 4.0 camera (Hamamatsu). Images were captured at 0.5 Hz and
analyzed with ImageJ (National Institutes of Health, USA).
GABA R activation was done by focal pressure application of GABA (100 µM) in the
A
presence of TTX (0.5 µM), CGP55845 (1 µM), D-AP5 (20 µM) and CNQX (10 µM).
1
8

ACCEPTED MANUSCRIPT
Throughout the experiment 1 puff (puff duration 600-1000 ms) of GABA was delivered once per
minute.
2
+
For analysis of GABA R-induced changes in [Ca ] traces from individual cells were
A
i,
2
+
normalized to the peak responses of the evoked Ca transient at the first two GABA puffs
before drug application (at minutes -2 and 0). Bumetanide and bumepamine were applied via the
bath perfusion at minute 0. All other details were as described above for the GDP experiments.
2
.12 Effects of bumetanide and bumepamine on activated NKCC1 in an oocyte assay
Frogs (Xenopus laevis) were obtained from Laboratory Animal Science (Hannover Medical
School, Germany). The surgical removal and preparation of defolliculated oocytes was performed
essentially as previously described (Fenton et al., 2010) and complies with the European
Community guidelines for the use of experimental animals. Human NKCC1a (hNKCC1a;
obtained from Prof. Biff Forbush, Yale School of Medicine, CT) was subcloned into the oocytes
expression vector pXOOM, linearized downstream from the poly-A segment, and in vitro
transcribed using T7-mMessage Machine according to manufacturer’s instructions (Ambion,
Austin, TX). cRNA was then extracted with MEGAclear (Ambion, Austin, TX) and micro-
injected into defolliculated Xenopus laevis oocytes (25 ng RNA/oocyte). The oocytes were kept
in Kulori medium (in mM: 90 NaCl, 1 KCl, 1 CaCl , 1 MgCl , 5 Hepes, pH 7.4) for 5 days at 19
2
2
°
C prior to experiments.
In cells and tissues from different species, including humans, two alternatively spliced
RNA variants of NKCC1 have been identified: a full-length NKCC1 transcript (NKCC1a) and a
shorter splice variant (NKCC1B); both form functional cotransporters and vary in their tissue
distribution in humans (Markadieu and Delpire, 2014). Both hNKCC1a and hNKCC1b are
expressed in the human brain, but expression of hNKCC1b is several fold larger than that of
1
9

ACCEPTED MANUSCRIPT
hNKCC1a (Vibat et al., 2001). We therefore also performed transport experiments with
hNKCC1b. The shorter hNKCC1b variant was prepared by site-directed mutagenesis from
hNKCC1a (verified by DNA sequencing).
To activate NKCC1 prior to the uptake experiment, hNKCC1a- or hNKCC1b-expressing
oocytes or uninjected control oocytes (8-10 oocytes per well) were pre-incubated for 30 min at
+
room temperature in a hyperosmolar K -free solution (containing in mM: 5 choline chloride, 95
NaCl, 1 MgCl , 1 CaCl , 10 Hepes, pH 7.4, 207 mOsm), which causes shrinkage of the oocyte
2
2
+
and thus activation of NKCC1 (Zeuthen and MacAulay, 2012). To measure K influx, oocytes
were exposed to an isosmotic test solution in which KCl (5 mM) was substituted for choline
8
6
+
chloride and 2 μCi/mL Rb (NEZ072, PerkinElmer, Rodgau, Germany) included as a tracer for
+
K . Osmolarities of the test media were verified by using an automatic osmometer Type 15
(Löser; Berlin, Germany). Bumetanide, bumepamine or control vehicle (0.1%; ensuring equal
exposure to relevant drug solvent of all tested oocytes in the given experiment) was added to the
preincubation and to the test solution to allow cell penetration for a prolonged time (35 min). The
uptake assay was performed at room temperature for 5 min. The influx experiments were
8
6
+
terminated by 3 times rapid wash in ice-cold Rb -free assay solution after which the oocytes
were individually dissolved in 50 µl 10% sodium dodecyl sulfate in scintillation vials. The
radioactivity present was determined by liquid scintillation β-counting with Aquasafe 300 Plus
scintillation cocktail (Zinsser Analytic GmbH, Frankfurt, Germany) using a Microbeta Trilux
+
(
Perkin Elmer). NKCC1-mediated K uptake was assessed as ([flux
in
NKCC1-expressing oocytes
presence of x µM drug]-[fluxuninjected oocytes in presence of x µM drug]), in order to correct for
endogenous NKCC activity. All experiments were repeated at least three times. Sigmoidal curves
were fitted to the data for determination of the IC value for loop diuretics using GraphPad
5
0
Prism 7.0, according to a dose-response inhibition curve with log (inhibitor vs. response, variable
2
0

ACCEPTED MANUSCRIPT
slope; Y=Min + (Max-Min)/(1+10^((LogIC -X)*HillSlope)) and for experiments with fewer
5
0
concentrations assuming the curves would reach complete inhibition (going from 100% to 0
according to a dose-response inhibition curve with log(inhibitor) vs. normalized response,
variable slope; Y=100/(1+10^(X-LogIC )).
5
0
In addition to adding bumetanide and bumepamine extracellularly to the medium,
additional experiments were performed in which the two compounds were injected into the
oocytes. The experiments were conducted as described previously, except for the addition of
bumetanide and bumepamine into the media. Instead, 50 nl of bumetanide (2 µM or 20 µM
solution), bumepamine (2 µM or 20 µM solution) or control vehicle were injected using the
Nanoject II (Drummond Scientific Company, Broomall, USA) into the oocytes prior to exposure
to the preincubation medium.
2
.13 Effect of loop diuretics and bumepamine on AQP4-mediated water permeability in an
oocyte assay
Oocytes from Xenopus laevis were obtained and prepared as described above but with Xenopus
laevis frogs housed at University of Copenhagen, Denmark (all procedures complies with the
European Community guidelines for the use of experimental animals and have been approved by
the animal inspectorate at the Danish Ministry of Justice). Rat AQP4.M23 was obtained from
Søren Nielsen (Aalborg University, Denmark) and was subcloned into the oocyte expression
vector pXOOM. The cDNA was linearized downstream from the poly-A segment and in vitro
transcribed using T7 mMessage Machine (Ambion, Austin, TX). The cRNA was then extracted
with MEGAclear (Ambion, Austin, TX) and micro-injected into defolliculated Xenopus oocytes
(8 ng rAQP4.M23 RNA/oocyte). The oocytes were kept in Kulori medium (in mM: 90 NaCl, 1
KCl, 1 CaCl , 1 MgCl , 5 Hepes, pH 7.4) for 4-6 days at 19°C prior to experiments.
2
2
2
1

ACCEPTED MANUSCRIPT
The experimental setup for measuring cell swelling of oocytes has been described in
detail previously (Zeuthen et al., 2006). Oocytes were preincubated 1-2 hours in control solution
(
containing: 75 mM NaCl, 2 mM KCl, 1 mM CaCl , 1 mM MgCl , 10 mM Hepes, 50 mM
2 2
mannitol, pH 7.4) at 19 °C. An oocyte was then placed in a small chamber with a glass bottom
and perfused with the control solution at room temperature. Images of the oocytes were captured
continuously from below at a rate of 25 images/s. The oocytes were challenged with a
hyposmotic assay solution (control solution without mannitol; Δ -50 mOsm) with or without drug
in order to determine the rate of cell swelling (mean from two consecutive measurements 2-6
minutes apart). All osmolarities were validated using a cryoscopic osmometer Type 15 (Löser,
Berlin, Germany).
Effects of bumetanide (100 µM), AqB013 (20 µM) or bumepamine (2 and 20 µM) were
determined by preincubation of the oocytes in these media (or vehicle control) 1-2h prior to water
measurements (or in the case of acute bumetanide exposure; only introduced around 10 minutes
before the cell swelling measurements). The concentrations of bumetanide and AqB013 were
based on the study of Migliati et al. (2009).
2
.14 Statistics
Statistical analyses were performed using GraphPad Prism 7.0 software (San Diego, CA, US).
Either parametric or nonparametric tests were used for statistical evaluation, depending on data
distribution. For comparison of two groups, either Student’s t-test or the Mann-Whitney U-test
was used. In case of more than two groups we used analysis of variance (ANOVA) with post hoc
testing and correction for multiple comparisons. Depending on data distribution, either the
ANOVA F-test, followed posthoc by Dunnett’s multiple comparison test, or the Kruskal-Wallis
test followed posthoc by Dunn’s multiple comparisons test were used. A p≤0.05 was considered
2
2

ACCEPTED MANUSCRIPT
significant. G*Power Data Analysis (Faul et al., 2007) was used to calculate the necessary sample
size for a specified power of 80%. Pharmacokinetic analyses were performed by the software PK-
Solutions (Summit Research Services, Montrose, CO, U.S.A.).
3
3
. Results
.1 Physicochemical properties of bumepamine
Selection of bumepamine for the present experiments was based on experiments in a large series
of bumetanide derivatives, in which solubility in solvents tolerable for in vivo testing, diuretic
potency compared to bumetanide, inhibition of NKCC1, and brain penetration were assessed. As
shown in Fig. 1, the carboxylic group of bumetanide has been replaced by a phenylamino-methyl
group in bumepamine. Consequently, whereas bumetanide is a relatively strong acid (pKa 3.6)
with relatively low lipophilicity (logP 2.7), bumepamine is a weak acid (pKa 7.0) with high
lipophilicity (logP 4.5). Molecular weight of bumepamine is 425.54 compared to 364.42 of
bumetanide.
3
.2 Dose selection and route of administration for in vivo experiments
Previous experiments with bumetanide had shown that mice and rats metabolize bumetanide
much more rapidly than humans, so that high doses (at least 5-10 mg/kg) of bumetanide have to
be administered to induce diuresis and to obtain any meaningful brain concentrations in rats
(Olsen, 1977; Brandt et al., 2010; Töllner et al., 2014). Therefore, for the present study, the
intended dosage for all in vivo experiments was 10 mg/kg injected via the i.v. route except for the
diuresis experiments, in which also a lower dose (1 mg/kg) was administered i.v. For
bumepamine, the same doses were used. The reason for using i.v. administration for bumetanide
and bumepamine was that previous experiments showed that – because of rapid metabolism of
2
3

ACCEPTED MANUSCRIPT
bumetanide in rodents – its bioavailability is low after i.p. injection (Brandt et al., 2010; Töllner
et al., 2014). Some pharmacokinetic experiments were also performed with oral drug
administration (see below).
3
.3 Pharmacokinetic in vivo experiments with bumetanide and bumepamine in mice
As shown in Fig. 2A, bumetanide was rapidly eliminated in mice with an average half-life of
about 30 min, which is in line with our previous studies in mice (Töllner et al, 2014; Töpfer et al.,
2
014). Apparent volume of distribution was small (0.16 l/kg), indicating that the drug was
primarily distributed extracellularly. Only low brain concentrations of bumetanide were
determined after systemic (i.v.) administration, resulting in an average brain:plasma ratio of
0
.017 (30 min) and 0.022 (60 min), respectively. Despite the high dose (10 mg/kg) of bumetanide
that was administered i.v., average brain concentrations of bumetanide at 30 and 60 min after
injection were in the range of only 0.26-0.77 µg/g (0.71-2.1 µM; Table 1). However, these were
total brain concentrations, not taking unspecific tissue binding into account (see below).
Following i.v. injection of bumepamine, plasma concentrations of this compound were
relatively low (Fig. 2A), which might either indicate a rapid metabolism or a high tissue
distribution or both. Apparent volume of distribution was large (1.92 l/kg), indicating extensive
extra- and intracellular distribution and high non-specific tissue binding of bumepamine (cf.,
Reichel, 2009). Elimination half-life in plasma was ~26 min on average and thus similar to the
half-life of bumetanide in mice. Since bumepamine could be metabolized to the corresponding
benzoic acid bumetanide, we also determined bumetanide concentrations after administration of
bumepamine. Compared to maximum bumetanide plasma levels after administration of
bumetanide (29-35 µg/ml), maximum bumetanide concentrations after administration of
bumepamine were markedly lower (1.1-1.7 µg/ml), indicating that metabolism to bumetanide
2
4

ACCEPTED MANUSCRIPT
played no significant role in the pharmacodynamic effects of bumepamine (see below). In
contrast to the low plasma levels of bumepamine, brain levels were markedly higher (Fig. 2B),
indicating that the low plasma levels were due to extensive tissue distribution of the lipophilic
compound. Brain:plasma ratio ranged between 0.8 and 2.1 and was significantly higher than the
brain:plasma ratio of bumetanide (Fig. 2C). Brain concentrations of bumepamine at 30 and 60
min after injection were in the range of 1.4-4.4 µg/g, which corresponds to 3.3-10.4 µM. In
comparison to these relatively high brain concentrations of bumepamine, very low concentrations
(~0.1 µg/g) of bumetanide were determined in the brain (Fig. 2B), substantiating that metabolism
to bumetanide is not important for the pharmacodynamic effects of bumepamine.
In addition to i.v. drug administration, bumetanide and bumepamine were also
administered p.o. to allow calculation of oral bioavailability. As shown in Fig. 2D, plasma
concentrations of bumetanide were considerably lower after p.o. vs. i.v. administration, resulting
in an average oral bioavailability of 42%. Similar data were determined for bumepamine (Fig.
3
E), resulting in an average oral bioavailability of 41%.
3
.4 Determination of fraction unbound in plasma and brain of mice
When comparing drug brain levels obtained in vivo with effective drug concentrations
determined in in vitro assays, one has to consider unspecific brain tissue binding of drugs to
proteins and lipids, which may markedly reduce the brain drug concentration that is available for
effects on pharmacological targets (Reichel, 2009). Thus, unbound brain drug concentrations
should be distinguished from total brain concentrations to allow comparison with drug potencies
in in vitro assays (Reichel, 2009). We therefore determined the brain tissue binding of
bumetanide and bumepamine by equilibrium dialysis of the compounds between buffer and brain
homogenate. Furthermore, the plasma protein binding of both compounds was determined, since
2
5

ACCEPTED MANUSCRIPT
only the unbound fraction is available for passive diffusion through the BBB (Reichel, 2009).
As shown in Table 1 for bumetanide, fraction unbound (f ) in plasma was only 0.012,
u
indicating a plasma protein binding of 98%, which is in line with previous reports (Puskarjov et
al., 2014). Extensive unspecific tissue binding, most likely to lipids, was determined for brain
tissue (corrected for dilution), resulting in f of only 0.17 (Table 1). Thus, using tissue dilution of
u
1
:3 (see Methods), only 17% of the total brain concentration of bumetanide was unbound and
thus available for pharmacological effects on NKCC1 in the brain. Slightly lower f values were
u
obtained with tissue dilutions of 1:10 (f 0.094 ± 0.014) and 1:20 (0.12 ± 0.023). When f was
u
u
used to calculate unbound brain concentrations of bumetanide for the brain values determined in
the in vivo mice experiments described above, bumetanide concentrations of 0.12-0.36 µM were
obtained, which was in the range of NKCC1 IC s often cited in the literature (100-300 nM) but
5
0
below the cutoff of 1 µM needed for selective block of NKCC1 (Table 1). The ratio between
unbound brain levels and total plasma levels of bumetanide ranged between 0.003 and 0.0038.
Bumepamine was even higher bound to plasma proteins and tissue than bumetanide
(
Table 1). Plasma protein and tissue binding was >99%, resulting in extremely low free
unbound) concentrations in plasma and brain.
(
3
.5 Diuretic effect of bumetanide and bumepamine in mice
As shown in Fig. 3, i.v. doses of 1 and 10 mg/kg of bumetanide induced a significant diuretic
effect. At 1 mg/kg, urine volume increased 1.9-fold compared to vehicle controls within 120 min
after injection (Fig. 3A), while a 3.2–fold increase was determined after the higher dose (Fig.
3
B). Bumepamine, on the other hand, was significantly less diuretic. At 1 mg/kg, no significant
effect on diuresis was observed (Fig. 3A), while 10 mg/kg led to a 2.1–fold increase in urine
production. Thus, 10 mg/kg bumepamine induced a similar diuretic effect than 1 mg/kg
2
6

ACCEPTED MANUSCRIPT
bumetanide, indicating that bumetanide is about 10-times more potent as a systemic diuretic than
bumepamine, which corresponds to the 7-fold difference between the diuretic potency of
bumetanide and bumepamine previously reported for dogs (Nielsen and Feit, 1978; Lykke et al.,
2
015).
3
.6 Effect of bumetanide and bumepamine alone or in combination with phenobarbital on
electrically induced seizures in nonepileptic vs. epileptic mice
Increase in hippocampal (CA1) brain NKCC1 and decrease in KCC2 expression have been
described in the pilocarpine model of TLE in adult mice (Li et al., 2008). One may therefore
hypothesize that NKCC1 inhibitors are more effective in epileptic mice of this model than in
nonepileptic mice. This has been shown recently with the bumetanide prodrug BUM5, which
potentiated the anticonvulsant effect of phenobarbital on electrically induced seizures in epileptic
but not in nonepileptic mice (Erker et al., 2016). The use of acutely induced seizures rather than
spontaneous seizures for drug testing in epileptic mice was based on a previous study by Blanco
et al. (2009), who proposed that the induction of acute seizures in animals pretreated with
pilocarpine might constitute an effective method to test putative ASDs for human use in epileptic
rodents.
In line with the recent study of Erker et al. (2016), bumetanide (10 mg/kg) did not
potentiate the anticonvulsant effect of phenobarbital (10 mg/kg) on the electroconvulsive seizure
threshold (MEST) in nonepileptic or epileptic mice (Fig. 4A). In contrast, bumepamine (10
mg/kg), which did not increase MEST when administered alone, induced a moderate but
statistically significant (64%) increase in phenobarbital’s MEST-increasing effect in nonepileptic
mice, whereas a marked (5.6-fold) increase was determined in epileptic mice (Fig. 4A). The
increase of phenobarbital’s anticonvulsant efficacy by bumepamine in epileptic mice was
2
7

ACCEPTED MANUSCRIPT
significantly higher than the increase observed in nonepileptic mice. When administered at a
lower dose (1 mg/kg), bumepamine had no potentiating effect on phenobarbital’s anticonvulsant
efficacy (Fig. 4A). When the dose of phenobarbital was decreased to 5 mg/kg, bumepamine (10
mg/kg) significantly potentiated its anticonvulsant efficacy only in epileptic mice (Fig. 4A). No
behavioral adverse effects were observed following combined treatment with any dose
combination of bumepamine and phenobarbital. As reported previously (Klein et al., 2014), the
ethanol-containing vehicle (used for dissolving bumetanide and bumepamine) did not
significantly increase MEST.
To exclude the possibility that the effect of pretreatment with bumepamine and
phenobarbital’s anticonvulsant activity was only due to a pharmacokinetic interaction, we
repeated the experiments and determined phenobarbital in plasma and brain at time of MEST
determination (30 min after drug injection) in groups of mice that were either treated with vehicle
or with bumetanide or bumepamine 30 min before phenobarbital. As shown in Fig. 4B-D,
combined treatment did not significantly alter phenobarbital levels in plasma and brain or the
brain:plasma ratio of phenobarbital. In all groups, phenobarbital levels in plasma were at the
lower border of the therapeutic plasma concentration range of this drug (10-40 µg/ml) in patients
with epilepsy (Patsalos et al., 2008). These data thus substantiated that the synergistic interaction
between bumepamine and phenobarbital was pharmacodynamic and not pharmacokinetic in
nature.
3
.7 Effect of phenobarbital and bumepamine alone or in combination on fully amygdala-
kindled seizures in rats
Increases in hippocampal and piriform cortex NKCC1 expression have been described in the
kindling model of TLE in adult rats (Okabe et al., 2002, 2003). One may thus hypothesize that
2
8

ACCEPTED MANUSCRIPT
seizures in this model are particularly responsive to NKCC1 inhibitors. In a previous study, we
showed that neither bumetanide alone (10 mg/kg i.v.) nor phenobarbital alone (10 or 20 mg/kg
i.v.) affect focal (afterdischarge) seizure threshold (ADT) in amygdala kindled rats, but a
moderate (~30%) increase of ADT was observed when combining both drugs (Töllner et al.,
2
014). In the present experiments, bumepamine (10 mg/kg i.v.) did not significantly increase
ADT when given alone or in combination with a low dose of phenobarbital (10 mg/kg i.p.) in a
group of 9 fully kindled rats (Fig. 5A). We have previously shown that individual kindled rats
differ in their response to ASDs, resulting in ASD responders and nonresponders (Löscher, 1997;
Baraban and Löscher, 2014). We therefore checked whether a similar phenomenon exists for the
bumepamine/phenobarbital combination. Responders were defined by an at least 40% increase in
ADT after drug treatment (Löscher et al., 1998). Three of the 9 kindled rats responded to the
combined treatment with bumepamine and phenobarbital with an average 3.1-fold ADT increase
(Fig. 5B). This differed significantly from nonresponders, in which ADT was not affected after
combined treatment (Fig. 5B). No behavioral adverse effects were observed with any treatment.
3
.8 Comparison of in vitro effects of bumetanide and bumepamine on GDPs in the neonatal
hippocampal slice
A characteristic feature of the immature hippocampus is the occurrence of spontaneous network
events that have been termed giant depolarizing potentials (GDPs) (Ben-Ari et al., 1989). GDP
activity is extremely sensitive to NKCC1 block by bumetanide (Sipilä et al., 2005; Spoljaric et
al., 2017) and therefore we used it for comparing the pharmacodynamic effects of bumetanide
and bumepamine. As shown previously (Sipilä et al., 2005; Töllner et al., 2014; Spoljaric et al.,
2
017), bumetanide rapidly and potently blocked GDPs in the neonate (P2 - P5) rat hippocampus
(Fig. 6A,C). We first compared the effect of bumetanide (2 µM, a saturating concentration to
2
9