Received: 20 December 2016 | Revised: 4 May 2017 | Accepted: 30 May 2017

DOI: 10.1002/glia.23182

RESEARCH ARTICLE

Inhibition of glial hemichannels by boldine treatment reduces

neuronal suffering in a murine model of Alzheimer’s disease

Chenju Yi1 | Pascal Ezan1 | Paola Fern

andez2,3 | Julien Schmitt4 |
Juan C. S
aez2,3 | Christian Giaume1 | Annette Koulakoff1

1
Center for Interdisciplinary Research in
Biology (CIRB), Collège de France, CNRS,
Abstract
INSERM, PSL Research University, Paris The contribution of reactive gliosis to the pathological phenotype of Alzheimer’s disease (AD)
75005, France opened the way for therapeutic strategies targeting glial cells instead of neurons. In such context,
2
Departamento de Fisiología, Pontificia connexin hemichannels were proposed recently as potential targets since neuronal suffering is

lica de Chile, Santiago,
Universidad Cato
alleviated when connexin expression is genetically suppressed in astrocytes of a murine model of
Chile
3 AD. Here, we show that boldine, an alkaloid from the boldo tree, inhibited hemichannel activity in
Centro Interdisciplinario de Neurociencias
de Valparaíso, Valparaíso, Chile astrocytes and microglia without affecting gap junctional communication in culture and acute
4
Sorbonne Universite
s, UPMC Univ Paris 06, hippocampal slices. Long-term oral administration of boldine in AD mice prevented the increase in
INSERM, CNRS, Neurosciences Paris Seine, glial hemichannel activity, astrocytic Ca21 signal, ATP and glutamate release and alleviated
Institut de Biologie Paris Seine (NPS-IBPS),
hippocampal neuronal suffering. These findings highlight the important pathological role of
Cerebellum Navigation and Memory team
(CeZaMe), Paris 75005, France hemichannels in AD mice. The neuroprotective effect of boldine treatment might provide the basis
for future pharmacological strategies that target glial hemichannels to reduce neuronal damage in
Correspondence
neurodegenerative diseases.
Christian Giaume, CIRB, Collège de France,
11 Place Marcelin Berthelot, 75005 Paris
France. KEYWORDS
Email: [email protected]
or
Alzheimers’disease, astrocytes, connexin, hemichannel, microglia
Juan C. S
aez, Departamento de Fisiología,

lica de Chile,
Pontificia Universidad Cato
Santiago, Chile.
Email: [email protected]
Present address
Chenju Yi, Centre for Life Sciences,
National University of Singapore, 117456,
Singapore, Singapore

1 | INTRODUCTION paracrine pathway targeting neuronal activity and survival (Bennett

et al., 2012; Orellana, Retamal, Moraga-Amaro, & Stehberg, 2016).
Altered cell-to-cell communication is now emerging as an important Indeed, once activated in glial cells, HCs are involved in the release of
feature of many brain pathologies (Garden & La Spada, 2012). In partic- gliotransmitters like ATP and glutamate as well as chemokines that
ular, the crosstalk between glial cells and neurons which is crucial for have deleterious consequences on neurons as reported for murine
neuronal function and health is found to be impacted in number of models of Alzheimer’s disease (AD) (Orellana et al., 2011b; Takeuchi
acute, as well as chronic, diseases (Halassa, Fellin, & Haydon, 2007; et al., 2011; Yi et al., 2016), experimental allergic encephalomyelitis
Kettenmann, Kirchhoff, & Verkhratsky, 2013; Rossi, 2015). Indeed, (Shijie et al., 2009), amyotrophic lateral sclerosis (Almad et al., 2016;
both astrocytes and microglial cells become reactive and their ability to Takeuchi et al., 2011) and neuropathic pain (Bravo et al., 2014; Chen
release a large array of factors including pro-inflammatory cytokines et al., 2014). HCs can be formed by two families of functionally related
and gliotransmitters is strongly modified, leading to synaptic dysfunc- proteins, which share the same transmembrane topology but have
tions and neuronal damages. Among the multiple actors involved in divergent primary sequences: pannexins (Panxs) and connexins (Cxs)
this crosstalk, accumulating evidence indicates that hemichannels (HCs) (Giaume, Leybaert, Naus, & S
aez, 2013). They oligomerize into hexam-
can be important players in pathological processes by providing a eres that isolate a large nonselective pore in the membrane allowing

Glia. 2017;65:1607–1625. wileyonlinelibrary.com/journal/glia V

C 2017 Wiley Periodicals, Inc. | 1607
1608 | YI ET AL.

for the diffusion of ions, small signaling molecules and metabolic 2 | MATERIALS AND METHODS
substrates between the cell cytoplasm and the extracellular medium.
In addition, most Cx HCs dock head to head between adjacent cells 2.1 | Animals
to form intercellular channels that constitute gap junction plaques
APPswe/PS1dE9 and C57Bl/6 (WT) mice were obtained from Jackson
while endogenously expressed Panx HCs do not (Huang, Grinspan,
Laboratories. Experiments were done in mice of either sex according to
Abrams, & Scherer, 2007; MacVicar & Thompson, 2010). Both fami-
the European Community Council Directives (2010/63/UE).
lies of HCs are efficiently inhibited by different drugs, such as carbe-
noxolone (CBX) and glycyrrhetinic acid, which also inhibit gap
2.2 | Reagents
junctional communication. More selective HC blockers have been
generated, noticeably mimetic peptides targeting extracellular loops Hydrochloride boldine was prepared from boldo’s bark (Peumus boldus
of Panxs or Cxs. However, treatment with these peptides was also Molina, Monimiaceae) by Härting Company (Santiago, Chile) with a
reported to reduce gap junctional communication after 30 min of purity of 99% tested by high-performance liquid chromatography analy-
treatment likely by reducing the ability of Cx HCs to be engaged in sis, as described previously (Hern

andez-Salinas et al., 2013). ARL67156
new gap junction channels whose turnover is very rapid (Giaume & trisodium salt was from Tocris Bioscience (Bristol, UK); Fluo-4 AM,
Theis, 2010). Hence, in order to study the role of HCs in brain Sulforhodamine 101, MitoSOXTM Red and TO-PRO ®-3 Iodide were
pathology there is a need of other drugs that block HCs without from ThermoFisher Scientific (Waltham, MA) and all other drugs were
altering gap junctional communication whose maintenance may be from Sigma-Aldrich (Saint-Quentin Fallavier, France). Mimetic peptides
important to sustain the homeostatic and metabolic functions of gap Gap26 (VCYDKSFPISHVR) and 10Panx1 (WRQAAFVDSY) were synthe-
junctional cellular networks in pathological situations (Giaume, Kou- sized by Thermo Fisher Scientific (purity >95%).
lakoff, Roux, Holcman, & Rouach, 2010; Rouach, Koulakoff, Abu-
dara, Willecke, & Giaume, 2008). 2.3 | Astrocyte cell cultures
In this context, we have selected boldine, an alkaloid extracted
Astrocyte cultures were prepared from the cortex of newborn C57Bl/6
from a Chilean tree, termed boldo. This molecule is known for its cyto-
mice as previously described (Koulakoff, Ezan, & Giaume, 2008). Briefly,
protective effects mostly mediated by its potent antioxidant properties
dissected cortices were mechanically dissociated in PBS supplemented
reported in diverse in vitro preparations as well as in animal models of
with glucose (33 mM). Cells were seeded on poly-ornithine coated 100-
diabetes, hypertension and atherosclerosis (Lau, Ling, Murugan, & Mus-
mm-diameter plastic dishes at a density of 2 3 106 cells/dish in DMEM
tafa, 2015; O’Brien, Carrasco-Pozo, & Speisky, 2006; Santanam Penu-
supplemented with penicillin (10 U/ml), streptomycin (10 lg/ml, GIBCO),
metcha, Speisky, & Parthasarathy, 2004). However, it was recently
and 10% FCS. Secondary astrocyte cultures were obtained by harvesting
reported to block HCs and not gap junctions in mesangial cells in cul-
1-week-old primary subconfluent cultures with trypsin–EDTA. For
ture by a mechanism yet unknown but independent of its antioxidant
scrape-loading dye-transfer experiments, cells were seeded on
property (Hern
andez-Salinas et al., 2013). Interestingly, this molecule
polyornithine-coated 35-mm-diameter dishes (Nunc, Roskilde, Denmark)
can be administered in vivo via the drinking water without secondary
(5 3 105 cells/ml). For hemichannel experiments cells were plated on
toxic effects and reaches the brain (de Lima et al., 2017; Loghin et al.,
glass coverslips inside 24-round-well plates (1.5 3 105 cells) in the same
2003) although this issue might not be critical since in most brain
culture conditions. Cultures were treated with cytosine-arabinoside
diseases and injuries its passage to the brain parenchyma should
(1 lM) for 3 days to eliminate proliferating microglial cells. Medium was
be facilitated since the blood brain barrier (BBB) is weakened
changed twice a week and cultures were used after 1 week.
(Zlokovic, 2008).
In the present study, we analyzed the effects of boldine on HC
activity and gap junctional communication in glial cells in different
2.4 | HeLa-Panx1-YPF cells
experimental paradigms: cultured astrocytes and acute brain slices in HeLa-Panx1-YPF cells were grown in DMEM with 10% fetal bovine
which activation of HCs was triggered by pro-inflammatory treatment serum (FBS), containing G418 at a final concentration of 1 lg/ml at
in astrocytes and microglia as well as in brain slices of 9-month-old 378C in an incubator supplemented with 10% CO2. HeLa-Panx1-YPF
APP/PS1 mice, a murine model of Alzheimer’s disease (AD) in which cells were kindly provided by Dr. Felixas Bukaukas (Department of
we recently reported that HCs are chronically activated in astrocytes Neuroscience, Albert Einstein College of Medicine, Bronx, NY, USA).
(Yi et al., 2016). We show that boldine was as efficient as carbenoxo- Panx1 HCs were activated by application of mechanical streatch as
lone in inhibiting HC activity in astrocytes and in microglial cells but recently described (Johnson et al., 2016). In brief, cells cultured on glass
had no effect on astroglial gap junctional communication. Long-term coverslips were placed on an upright Olympus microscope BX 51W1I
administration of boldine in vivo in APP/PS1 mice prevented HC activa- with water immersion objectives and mechanically stimulated with
tion, reduced the release of ATP and glutamate, decreased astroglial 8 ml of recording saline solution dropped from 5 cm height drop by
Ca21
i signal to resting levels and alleviated neuronal stress, making it an drop and then were bathed with Locke’s recording saline solution
interesting new pharmacological tool to attenuate neuronal damage (in mmol/l: 154 NaCl, 5.4 KCl, 2.3 CaCl2, and 5 HEPES, pH 7.4) with
associated with amyloid pathology. 5 lmol/l ethidium bromide (EtBr). Changes in fluorescence were
YI ET AL. | 1609

monitored using a digital camera (12 bits, Qimaging, Burnaby, BC, Can- sensitivity of EtBr uptake to Cx or Panx HC blockers, slices were prein-
ada) and the program of acquisition and image analysis was META- cubated in ACSF supplemented with either blocker for 15 min prior to
FLUOR software (Universal Imaging Co., Downingtown, PA, USA). dye application. The blockers used were: CBX (200 mM), mimetic pep-
10
Fluorescence was recorded every 30 s at room temperature (20–228C). tides targeting Cx43 or Panx1, Gap26 (0.3 mg/ml) and panx1
In the analysis, the regions of interest were placed over the cell nuclei (0.5 mg/ml), respectively. The control condition and diverse experimen-
selected at random and these regions were background subtracted. tal conditions were done in parallel. Following EtBr incubation, slices
Average slope values were compared using GraphPad Prism software. were rinsed three times with ACSF, fixed for 1 h in 4% paraformalde-
hyde and processed for immunostainings.
2.5 | Dye uptake in astrocyte cultures Fixed slices were preincubated in PBS containing 0.2% gelatin and
1% Triton-X100 (PBS*) for 30 min twice and then incubated overnight
EtBr uptake was measured in secondary cultures of astrocytes pre-
at 48C in primary antibodies appropriately diluted in PBS*. Mouse or
treated or not with 10 ng/ml LPS for 24 h, known to increase the open
chicken anti-GFAP (1:500, Inc. Biomedicals) rabbit anti-Ab (1:1000,
probability of Cx43 HCs. Cells were preincubated in HEPES buffer con-
Life Technologies) and rabbit or goat anti-Iba1 (1:500, Wako) antibod-
taining (in mM) NaCl (150), KCl (5.4), MgCl2 (1), CaCl2 (2), HEPES (5),
ies were used to detect astrocytes, Ab deposits and microglia, respec-
and D-glucose (10) (pH was adjusted to 7.4) in the presence or not of
tively. After 2 3 30 min washes in PBS*, sections were incubated for
50 mM CBX or 0.3 mg/ml Gap26 peptide for 5 min, then exposed to 5
2 h at RT with a mixture of Alexa-conjugated secondary antibodies
mM EtBr for 10 min at room temperature (RT). Cells were then washed
(diluted in PBS* 1:2000, Molecular Probes), washed and mounted using
with the same buffer and fixed with 4% paraformaldehyde in PBS. Fluoromount-G (Southern Biotechnology). These slices were examined
Fixed cells were examined at 403 confocal laser-scanning microscope at 403 with a confocal laser-scanning microscope (Leica SP5). All the
(Leica SP5). Stacks of 10 consecutive confocal images at 0.5 lm inter- parameters were kept the same during the acquisition for each slice
vals were acquired. Six images were captured for each experimental (same laser power, excitation intensity, same offset) that was done
condition. Images of EtBr uptake were analyzed with Image J program from the same depth below the surface of the slices (15 mm). Four
(NIH software). images covering CA1 to CA3 hippocampal region were captured for
each experimental condition. Stacks of 22 consecutive confocal images
2.6 | Determination of gap junctional communication at 0.5 mm intervals were acquired sequentially with three lasers (argon
488 nm, 561 nm for EtBr and 647 nm).
Experiments were performed by using the scrape-loading dye-transfer
^me et al., 2006). Briefly, cells EtBr fluorescent signal in astrocyte nuclei was analyzed using Mac
technique, as previously described (Me
Imaris 7.4.0 software. Cells were considered as EtBr1 astrocytes if they
were incubated at room temperature for 10 min in HEPES buffered
satisfy the following criteria: (i) cells exhibiting GFAP immunostaining;
salt solution containing (in mM): NaCl (140); KCl (5.4); CaCl2 (1.8);
(ii) cells with nuclei showing EtBr fluorescence intensity above the
MgCl2 (1); glucose (10); HEPES (5) at pH 7.4. Cells were then washed
threshold defined as the background outside the nuclei, and (iii) cells
with a Ca21-free HEPES solution for 1 min withdraw and a scrape was
21 with nuclei volume larger than 100 mm3 (diameter  6 mm) to eliminate
performed with a razor blade in the same Ca -free solution containing
incomplete objects. For each experimental condition, the percentage of
Lucifer yellow CH (dilithium salt 427 Da, 1 mg/ml). After 1 min during
EtBr1 astrocytes was calculated among the overall population of
which the dye was loaded into the cells involved in the scrape, cells
GFAP1 cells and in each EtBr1 astrocyte, the fluorescence intensity
were washed and incubated for 8 min in the HEPES solution allowing
and the volume of the nucleus were measured. Counts were done for
dye diffusion between cells connected through gap junction channels.
200 cells per condition. Then, the EtBr uptake was calculated for each
Six images were captured for each experimental condition using NIS
experimental condition by taking into account both the percentage of
Nikon software. Dye coupling was evaluated by measuring the fluores-
EtBr1 astrocytes and their mean fluorescence intensity. Finally, to nor-
cence area using Image J software and then the fluorescence area in
malize experimental data, EtBr uptake was expressed as a ratio of the
each condition was normalized to that in control condition.
control condition. EtBr uptake in microglia nuclei was analyzed in cells
positive for Iba1 with nuclei larger than 65 mm3 (diameter  5 mm) in
2.7 | Acute brain slices
the same way than in astrocytes.
Acute brain hemisphere slices (300 lm) were cut on a vibroslicer (Leica
VT 1200S) and transferred to the nylon mesh of a holding chamber 2.9 | Fluorescence recovery after photobleaching
submerged in oxygenated ACSF at RT for a stabilization period of 60 (FRAP)
min. The ACSF solution contains (in mM): NaCl (125); KCl (2.5);
To evaluate gap junction function, we have used FRAP to measure cell-
NaHCO3 (25); NaH2PO4 (1.25); glucose (25); MgCl2 (1); CaCl2 (2).
to-cell exchange of a fluorescent tracer (Cotrina, Gao, Lin, & Nedergaard,
2001; Giaume, Orellana, Abudara, & S
aez, 2012), sulforhodamine-101
2.8 | Dye uptake in acute brain slices
(SR101) that is selectively taken in by astrocytes and passes through gap
Acute slices were incubated with the HC permeable fluorescent tracer junction channels (Nimmerjahn, Kirchhoff, Kerr, & Helmchen, 2004).
EtBr for 10 min at 4 mM final concentration at RT. To determine the Acute brain hemisphere slices of 9-month-old WT and APP/PS1 mice
1610 | YI ET AL.

were prepared as described above. The living slices were incubated in 1 intensity of astrocyte soma reflecting resting intracellular Ca21 signal
mM SR101 at 378C for 30 min. After 20 min wash in ACSF, the slices was measured using Image J software. For each experimental condi-
were placed in the stage of a confocal microscope and perfused with oxy- tion, measurements were done in 150 astrocytes.
genated ACSF during the experiment. Acquisitions were performed using
a water immersion Leica 253, NA 5 0.95 objective on a Leica microscope 2.12 | Determination of L-glutamate, GABA,
(SP5 MP) controlled by the Leica Acquisition Software. After excitation and D-serine levels by CE-LIF
with the 561 nm line of laser, a baseline fluorescence was collected every
Extracellular levels of L-glutamate, D-serine, and GABA were deter-
25 s for 200 s. Then, the laser scanning was restricted to a square (15 mm
mined using acute brain hemisphere slices of 9-month-old mice treated
3 15 mm) covering the whole cell body of the target astrocyte with a
or not with boldine for 12 weeks (2 slices/tube/condition) incubated in
zoom 33. Fifteen pulses of light at full power of the laser were applied to
1 ml oxygenated ACSF as described (Fossat et al., 2012). Extracellular
bleach this target area. Then, the normal recording configuration was
medium was collected, frozen directly in liquid nitrogen and stored at
reestablished and fluorescence recovery was recorded every 25 s for 20
2808C. Intracellular levels of L-glutamate, D-serine, and GABA in slices
min. Images were analyzed offline: the mean fluorescence intensity of
were also analyzed. Briefly, pooled slices were deproteinized by adding
astrocyte cell body was measured with Image J software. F0 was defined
cold trichloroacetic acid (TCA) to a final concentration of 5%. The sus-
as the first value of the baseline recording. The fluorescence intensity of
pension was centrifuged at 16,800g for 10 min. TCA was extracted
each time point was normalized to F0 and expressed as the percentage of
from the supernatant using water-saturated diethyl ether and stored at
F0 (Ft/F0%). The percentage of recovery was calculated by subtracting the
2808C. Samples were analyzed using a commercial laser-induced
plateau value (average of the last 6 values of the recording) with the value
fluorescence capillary electrophoresis (CE-LIF) (CE: Beckman Coulter,
of the first time point after photobleaching. All studies were performed at
P/ACE MDQ; LIF: Picometrics, LIF-UV-02, 410 nm 20 mW) as follows:
45 mm below the surface to ensure that the cell of interest was not
samples were processed for micellar CE-LIF and were fluorescently
damaged during preparation. CBX that inhibits gap junctions formed by
derivatized at RT for 60 min with napthalene-2,3-dicarboxaldehyde
Cxs was used as a negative control.
before being analyzed by CE using a hydroxypropyl-b-cyclodextrin-
based chiral separation buffer. All electropherograms data were col-
2.10 | TNF-a and IL-1b measurements by ELISA
lected and analyzed using Karat 32 software v8.0 (Beckman Coulter).
Brain hemispheres were harvested from 9-month-old APP/PS1 mice Further analysis was made using GraphPad Prism software.
treated or not with boldine for 3 months and kept at 2808C until proc- The extracellular and intracellular amounts of L-glutamate, D-serine,
essing. Brain tissue was immersed in ice-cold lysis buffer containing and GABA were normalized to the protein content determined from
(in mM) Tris–HCl (50), NaCl (100), EDTA (2), 1% Triton X100 and 1% pooled slices using the BCA kit. The quantity of L-glutamate, D-serine,
protease inhibitor cocktail (Sigma-Aldrich), and sonicated for 10 s. and GABA in the samples was determined from a standardized curve
Samples were centrifuged at 13,200 rpm for 20 min at 48C and while peak identification was made by spiking the fraction with the
supernatants were kept at 2808C. A BCA kit (Thermoscientific) was amino acid. For accurate comparison between conditions, extracellular
used for protein determination. Levels of TNF-a and IL-1b were levels of each gliotransmitter (referred as [gliotransmitter]ext) were
measured using ELISA kits (Invitrogen, Thermoscientific) according to normalized to total gliotransmitter (intracellular 1 extracellular)
the manufacturer’s protocol. (referred as [gliotransmitter]tot) allowing to assess the released fraction
of the gliotransmitter
2.11 | Intracellular Ca21 signal imaging
2.13 | Determination of extracellular ATP
Acute brain hemisphere slices from 9-month-old mice were incubated
levels by bioluminescence
for 60 min at 378C in ACSF containing 0.02% Pluronic F-127 and
Fluo-4 AM (5 mM), a Ca21 indicator specifically loaded in astrocytes After equilibration, extracellular levels of ATP were determined from
(Hirase, Qian, Bartho, & Buzsaki, 2004). Fluo-4 AM loaded slices were acute brain hemisphere slices of 9-month-old mice (2 slices/tube/
transferred onto the stage of a two-photon SP5 MP microscope and condition) incubated in 1 ml oxygenated ACSF for 30 min in the
Ca21 signal was measured. Fluo-4 AM was excited at 910 nm using a presence of, 6-N,N-diethyl-b-g-dibromomethylene-dadenosine-5-
pulsed Maitai (Spectra Physics) laser. Emitted fluorescence was triphosphate FPL 67156 (ARL 67156 trisodium salt, Tocris), an ectonu-
detected using 2 nondescanned detectors, with a 525/50 nm filter. cleotidase inhibitor to inhibit ATP hydrolysis. Extracellular medium was
Acquisitions were done from the same depth below the surface of the collected and its ATP content was measured immediately using the
slice (40 mm) using a water immersion Leica 253, NA 5 0.95 objective luciferin-luciferase bioluminescent assay (ATPLite kit, Perkin Elmer).
(with a zoom 1.53) on a Leica SP5 microscope controlled by the Leica Each condition was run in triplicates. Luminescence was measured
Acquisition Software. SR101 was excited at 760 nm and the emitted using a luminometer (Berthold Technologies) according to the manufac-
fluorescence was detected using a 585/40 nm filter. All the parameters turer’s instructions. The extracellular ATP concentration was normal-
(laser power, offset) were kept the same for each slice during ized to the protein content determined from pooled brain slices by
the acquisition. Images were analyzed offline: the mean fluorescence BCA kit (as above).
YI ET AL. | 1611

2.14 | Immunohistochemical stainings Meylan, France). Samples were lysed by sonication (Ultrasonic cell
disrupter, Microson, Bruxelles, Belgium), and centrifuged at
Cryostat sections (15 mm) prepared as previously described were
13,000 rpm 10 min at 48C. Samples were stored at 2208C until
fixed in 4% paraformaldehyde for 30 min at RT, washed in PBS,
used. Supernatants were collected to which 53 Laemilli buffer was
treated with 70% formic acid (when mouse anti-Ab clone 4G8 wad
added and the mixture was boiled for 5 min. Proteins were meas-
used) for 10 min and preincubated in PBS containing 0.2% gelatin
ured using the PierceTM BCA kit (Thermofisher scientific). For each
and 0.25% Triton-X100 (PBS**) for 30 min. Sections were then proc-
sample, 20 lg of protein was separated on Bis–Tris 4–12% NuPAGE
essed for immunostaining by overnight incubation at 48C with pri-
gels and electro-transferred to nitrocellulose sheets as previously
mary antibodies including rabbit anti-Ab (1:1000, Life Technologies)
described (Orellana et al., 2011a,b). Nonspecific protein binding was
and mouse anti-GFAP (1:500, Inc.Biomedicals) or mouse anti-Ab
blocked by incubation of nitrocellulose sheets in Tris-buffered saline
(clone 4G8, 1:500) and rabbit anti-RTN3 (R458, 1:1000) diluted in
(TBS)-Tween-milk solution (500 ml TBS 13; 500 ll Tween 203;
PBS**, washed in PBS** and incubated in secondary goat anti-rabbit
nonfat powder milk 25 g) for 1 h. Blots were then incubated over-
and anti-mouse IgG antibodies conjugated with different Alexa
night with primary antibodies mouse anti-GFAP 1:500 (Sigma) and
FluoR (Molecular Probes) diluted 1:2000 in PBS**. Sections were
rabbit anti-Iba1 1:500 (WAKO) at 48C, followed by 4 3 15 min TBS
washed, mounted in Fluoromount-G and examined with a confocal
wash. Blots were incubated with goat anti-mouse IgG (1:2500) and
microscope.
goat anti-rabbit IgG (1:2500) antibodies conjugated to horseradish
peroxidase (Santa Cruz, Clinisciences France). Immunoreactivity was
2.15 | Quantification of the Ab plaque surface assessed by ECL detection using the SuperSignal kit (Pierce, Rock-
Mosaic images of Ab immunostaining were taken with Apotome ford, IL, USA) according to manufacture instructions. Blots were
stereomicroscope with a 203 objective (Zeiss) reconstructed with then reprobed with mouse monoclonal anti-glyceraldehyde 3-
the ZEN software. For each mouse (5 APP/PS1 age-matched mice phosphate dehydrogenase peroxidase (SIGMA 1:10,000) to check
fed with boldine or water were used), six brain sections with 120 mm protein loading. Chemiluminescence imaging was performed on a
intervals were stained and quantified. The mosaic images covered LAS4000 (Fujifilm, Stamford, CT, USA). Semiquantitative densito-
the whole area of cortex and hippocampus in each brain section. metric analysis was done by using Image J software.
Total Ab plaque surface was quantified using Image J software and
normalized with the structure area of cortex and hippocampus in 2.18 | Detection of mitochondrial superoxide
each mouse. in acute brain slices
Living brain hemisphere slices from 9-month-old mice were incu-
2.16 | Quantification of RTN3 immunoreactive
bated with 5 lM MitoSOXTM Red, a mitochondrial superoxide indi-
dystrophic neurites
cator, for 10 min. Slices were then fixed using 4% paraformaldehyde
RTN3 immunoreactive dystrophic neurites (RIDNs) were quantified on at 48C for 1 h. After washing and permeabilization in PBS*, slices
brain sections immunostained for RTN3 and Ab, as previously were further incubated with TO-PRO®-3 Iodide (diluted in PBS*
described with slight modifications (Shi, Hu, Prior, & Yan, 2009). Stacks 1:2,000) at RT for 10 min, rinsed and mounted onto slides with
of 30 consecutive optical sections (0.5 mm each) were captured using a Fluoromount-G. The slices were examined using Leica SP5 confocal
403 objective and Z maximal projections reconstructed with the Leica microscope with 40x objective. Stacks of 22 consecutive confocal
software. For each mouse (5 age-matched APP/PS1 mice fed with images taken at 0.5 lm intervals were acquired sequentially with
boldine or water were used), six brain sections at 120 mm intervals two lasers (561 nm for MitoSOX and 647 nm for TO-PRO). All
were stained and quantified. In each brain section, four maximal projec- parameters were held constant for all sections. Four images covering
tion images were captured randomly in the cortex and hippocampus. the pyramidal layer visualized with TO-PRO nuclear staining in CA1
Image J Software was used to quantify the area of RIDNs at the level hippocampal formation were captured for each mouse (5 age-
of Ab plaques. Typically, the signals of RIDNs were bright, while RTN3 matched mice per condition, WT, APP/PS1 fed with water or bol-
cell staining was dim. RIDNs contained within Ab plaque areas were dine). Image-J Software was used to quantify the mean fluorescence
selected for quantification. Percentage of RIDNs area was calculated intensity in the pyramidal cell layer.
by dividing the total RIDNs area by the total area occupied by Ab
plaques in each mouse. 2.19 | Statistical analysis
Values are presented as mean 6 SEM; n corresponding to the number
2.17 | Western blot
of independent experiments. Statistical analysis was performed on raw
Brain hemispheres from WT and APP/PS1 mice treated or not with data with Graphpad software using one-way analysis of variance
boldine were harvested and immersed in 2% SDS containing protease (ANOVA) followed by Dunnett’s test or via a two-tailed Student’s
and phosphatase inhibitors (1 mM orthovanadate; 10 mM a-glycero- t test. The level of significance was set at p < .05. Graphics were
phosphate), and complete miniprotease inhibitor (Roche Diagnostics, prepared using GraphPad prism.
1612 | YI ET AL.

FIGURE 1 Boldine inhibits hemichannels but not gap junctional communication in cultured astrocytes. (a) Representative images showing
the EtBr uptake (red) in cultured astrocytes under CTL conditions or after LPS treatment in the indicated conditions; scale bar 20 mm. (b)
Quantification of EtBr uptake in cultured astrocytes (*p < .05, vs. LPS, one-way ANOVA followed by Dunnett’s test, means 6 SEM, n 5 4),
(c) representative images of scrape loading dye transfer experiments in confluent cultured astrocytes illustrating Lucifer yellow spreading
through astrocyte gap junctions in CTL conditions or after boldine treatments; scale bar 20 mm. (d) Quantification of gap junctional
communication in astrocytes in these conditions (one-way ANOVA followed by Dunnett’s test, means 6 SEM, n 5 5)

3 | RESULTS in culture are highly coupled. However, treatment of cultures with

boldine (10 min), at concentrations that inhibited HCs had no effect
3.1 | Boldine inhibits hemichannel activity but not gap on the dye coupling (Figure 1d). Since in cultured astrocytes, Cx43 is
junctional communication in cultured astrocytes the main contributor to HC activity (Giaume et al., 2013) our data
indicate that boldine affects differentially the two cell membrane
The effect of boldine on the two main cell membrane channel func-
channel functions of Cx43, blocking HC activity but not gap junc-
tions of Cx43 was first examined in cultured astrocytes characterized
tional communication in astrocytes.
by a high level of Cx43-mediated gap junctional communication that
contrasts with the low opening probability of HCs in basal conditions.
3.2 | Boldine, an efficient blocker of astroglial
We previously showed that such weak HC activity in basal conditions
HCs in acute brain slices
could be triggered by treatment with pro-inflammatory agents despite
the presence of extracellular divalent cations (Retamal et al., 2007). Next, we studied the effect of boldine in a more integrated prepara-
Hence, astroglial cell cultures containing 10% of microglial cells tion, namely acute brain hemisphere slices from adult mice. To activate
were treated for 24 h with the bacterial endotoxin lipopolysaccharide HCs, brain hemisphere slices were first briefly incubated (15 min) in a
(LPS, 10 ng/ml) and HC activity was assessed by measuring EtBr Ca21-free solution, known to open Cx HCs (Abudara et al., 2015; Ye,
uptake, as previously described (Retamal et al., 2007). LPS treatment Wyeth, Baltan-Tekkok, & Ransom, 2003). Indeed in this condition,
increased EtBr uptake by 111% (n 5 4) compared to control condi- using EtBr uptake assay to evaluate HC activation, a strong fluorescent
tions (Figure 1a). This effect was suppressed by CBX (100 mM), a signal was detected in the nuclei of astrocytes identified postfixation
potent and commonly used nonselective blocker of HCs and gap by GFAP immunostaining (Figure 2a). As already reported (Abudara
junction channels. Boldine (100 or 500 mM) also reduced EtBr uptake et al., 2015; Rouach et al., 2008), this fluorescent signal was suppressed
to control levels (Figure 1b), indicating that this compound was as by CBX (200 mM), indicating that the uptake occurred through open
efficient as CBX to inhibit HC activity. Then, gap junctional coupling HCs. Boldine used at the same concentrations than in cell cultures, i.e.,
was assessed in scrape loading dye transfer experiments in which 100 or 500 mM, had no effect on EtBr uptake. However when higher
Lucifer yellow (1 mg/ml) was loaded in astrocytes engaged in the concentrations of boldine were tested, a small reducing effect was
scrape and allowed to diffuse during 10 min through open gap junc- observed at 1 mM while 5 mM boldine completely suppressed EtBr
tion channels (Figure 1c). As expected, we confirmed that astrocytes uptake (Figure 2b). The same blocking effect of boldine was obtained
YI ET AL. | 1613

FIGURE 2 Acute treatment with boldine blocks astroglial HCs in brain slices. (a, c) Representative images of the effects of boldine on the
EtBr fluorescent signal (red) in acute brain hemisphere slices of 3-month-old WT mice immunostained for GFAP (green) in 0 Ca21 solution
(a) or after acute LPS treatment (c); scale bar: 20 mm. (b) Quantification of EtBr uptake in 0 Ca21 solution in presence of 200 mM CBX,
or boldine at the indicated concentrations (ns, not significant; ***p < .001, vs. 0 Ca21; ANOVA Dunnett’s test; mean 6 SEM, n 5 3). (d)
Quantification of the uptake in LPS-treated slices in the indicated conditions (*p < .05; **p < .01; ***p < .001, vs. LPS; ANOVA Dunnett’s
test; mean 6 SEM, n 5 3). (e) EtBr uptake (red) in astrocytes of 9-month-old APP/PS1 brain slices stained for Ab (blue) and GFAP (green),
in presence of 200 mM CBX or 5 mM boldine; scale bar, 20lm. (f) Quantification showing that, like CBX, acute treatment with boldine
abolished the uptake in astrocytes of APP/PS1 brain slices (***p < .001, vs. CTL; ANOVA Dunnett’s test; mean 6 SEM, n 5 3). (g–i) The
FRAP technique was used to quantify gap junctional communication between astrocytes. (g) Fluorescence images of astrocytes loaded with
the gap junction permeable tracer SR101 before bleaching (left), immediately after bleaching (middle) and 20 min after laser bleach (right);
scale bar: 10 mm. (h) Analysis of FRAP in the astrocytes of 9-month-old APP/PS1 mice acutely treated or not with 5 mM boldine. Curves
showing the time-lapse percentage of fluorescence intensity (FI) normalized with F0; (i) quantification of fluorescence recovery percentage
showing no significant differences between these two conditions, while the gap junction blocker CBX inhibited gap junctional communication
significantly (***p < .001, vs. CBX; unpaired Student’s t test; mean 6 SEM, CTL, boldine n 5 6; CBX n 5 4)

when HC activation was induced by LPS treatment (1.5 h, 100 ng/ml) Then, we examined the effect of boldine in acute brain hemisphere
(Figure 2c,d). As previously described (Abudara et al., 2015) such treat- slices of a murine model of AD, the APP/PS1 mouse, in which we pre-
ment triggered mainly Cx43 HCs since the mimetic peptide Gap26 tar- viously showed that HCs are chronically activated in the hippocampal
geting the extracellular loop of Cx43 (Wang et al., 2012) strongly astrocytes (Yi et al., 2016). Such activation only occurred in mice that
reduced the uptake by 65 6 9% (n 5 3). However, we also observed have developed b-amyloid (Ab) plaques at where astrocytes exhibited
10
that panx1 mimetic peptide had a mild but significant blocking effect a reactive phenotype (Figure 2e). Interestingly in these mice, while HCs
(25 6 4%, n 5 3), indicating a contribution of Panx1 to the HC activity were only formed by Cx43 in the astrocytes located far from plaques, a
in this paradigm. Hence, it appeared that in addition to its blocking minor Panx1 component contributed to the HC activity in addition to
effect on Cx43 HCs, boldine blocked Panx1 HCs as well. In support to Cx43 HCs in the reactive astrocytes contacting plaques. Acute treat-
the latter interpretation, we found that HeLa cells expressing Panx1 ment of brain hemisphere slices from 9-month-old APP/PS1 mice with
(HeLa-Panx1-YPF cells) and not Cx showed high EtBr uptake after boldine abolished EtBr uptake as efficiently as CBX in the astrocytes
mechanical stimulation; this uptake was blocked by 5 mM CBX, a con- located both far and close to plaques (Figure 2e,f). These observations
centration that preferentially blocks Panx1 over Cx HCs (Bruzzone, indicated that boldine inhibited the activity of astroglial HCs formed by
Barbe, Jakob, & Monyer, 2005) as well as by 100 mM boldine (Support- Cx43 and Panx1. Next, we verified whether, like in cultured astrocytes,
ing Information Figure S1). boldine did not alter astroglial gap junctional communication in WT
1614 | YI ET AL.

FIGURE 3 Boldine inhibits HC activity in microglial cells. (a) Representative images of EtBr fluorescent signal (red) in acute brain
hemisphere slices of 3-month-old WT mice treated with LPS and stained for GFAP (green) and Iba1 (blue). The EtBr signal, very bright in
LPS condition, was quite abolished by CBX (200 lM) or boldine (5 mM); scale bar: 10 lm. (b) Representative images of EtBr uptake (red) in
9-month-old APP/PS1 brain slices stained for Iba1 (gray) and Ab (blue) in CTL conditions or after treatment with 5 mM boldine or 200 mM
CBX; scale bar: 10 lm. (c) Quantification of EtBr uptake in microglia in the indicated conditions expressed as a ratio of LPS condition
(**p < .01; ***p < .001, vs. LPS; ANOVA Dunnett’s test; mean 6 SEM, n 5 3). (d, e) Histograms showing EtBr uptake in microglia (d) and in
hippocampal neurons (e) of APP/PS1 mice in presence of a series of HC blockers expressed as a ratio of control conditions (CTL) (*p < .05;
**p < .01; ***p < .001, vs. LPS; ANOVA Dunnett’s test; mean 6 SEM, n 5 5)

mice. The latter was examined using fluorescence recovery after pho- immunostaining, confirming previous results obtained in hippocampal
tobleaching (FRAP) on SR101 loaded brain hemisphere slices (Figure slices from younger mice (Abudara et al., 2015). Indeed, although less
2g) as previously described (Yi et al., 2016). The levels and kinetics of intense than in astrocytes (85% of that in astrocytes), a fluorescent sig-
fluorescence recovery in photobleached astrocytes were unaffected by nal was detected in the nuclei of microglial cells. This signal was
boldine, reaching after 20 min 61 6 4% (n 5 6) in control conditions abolished as efficiently by either CBX or boldine, and reduced by 53 6
and 65 6 3% (n 5 6) in presence of boldine (5 mM) compared to 23 6 5% (n 5 3) by Gap26 and 37 6 5% (n 5 3) by 10
panx1 (Figure 3a,b).
4% (n 5 4) in presence of CBX (Figure 2h,i). Altogether, these results Hence, these results confirm the above interpretation and indicate that
showed that boldine acutely applied on brain slices appeared as an effi- boldine inhibits the activation of HCs made of Cx43 and Panx1 in acti-
cient blocker of HCs formed by Cx43 as well as by Panx1, but that in vated microglial cells, like in astrocytes.
contrast to CBX it did not affect gap junctional communication. EtBr uptake experiments were also performed in brain hemisphere
slices from 9-month-old APP/PS1 mice (Figure 3c,d). A fluorescent sig-
nal was also observed in microglial cells identified by Iba1 immuno-
3.3 | Boldine inhibits HC activity in microglial cells
staining nearby amyloid plaques. This signal was abolished by CBX as
In brain hemisphere slices of 3-month-old WT mice acutely treated well as by boldine and was inhibited by 61 6 2% (n 5 5) by Gap26 and
with LPS, we have observed that EtBr uptake occurred not only in 39 6 3% (n 5 5) by 10
panx1 (Figure 3d). Altogether, these results indi-
astrocytes but also in microglial cells identified by Iba1 or CD68 cate that HCs in microglial cells were activated in an acute
YI ET AL. | 1615

inflammatory situation as well as in a chronic pathological context, and untreated APP/PS1) and similar to that measured in control APP/PS1
were inhibited by boldine. As so far, no gap junctional communication mice acutely treated with CBX (Figure 4b).
has been reported between microglial cells the effect of boldine in Next, we analyzed gap junctional communication in brain
APP/PS1 microglia is limited to the inhibition of Cx43 and Panx1 hemisphere slices of WT and APP/PS1 mice treated in vivo or not with
hemichannels. boldine during 12 weeks, using FRAP on SR 101 loaded slices (Figure
Finally, since activated HCs in glial cells can induce HC activation in 4c,d). First, the kinetics and level of fluorescence recovery in bleached
neurons as previously reported in acute situations (Orellana et al., 2011a, astrocyte were similar in WT and APP/PS1 fed with water (61 6 4 vs.
b), we examined whether this was also the case in a chronic situation. 68 6 3%, n 5 8 and 9, respectively), confirming that gap junctional
EtBr uptake was measured in the neurons of brain hemisphere slices of communication was maintained in 9-month-old APP/PS1 mice. Second,
9-month-old APP/PS1 mice (Figure 3e). Indeed an EtBr fluorescent signal in both genotypes, neither the kinetics nor the level of fluorescence
blocked by CBX was detected in neurons which were identified by Neu- recovey (65 6 3% in WT vs. 68 6 3% in APP/PS1 mice, n 5 5 and 9,
N immunostaining. It was also abolished by boldine. Interestingly, it was respectively) were affected by boldine treatment. Finally, CBX reduced
strongly reduced by 10
panx1 (86 6 6, % n 5 5) and also by Gap26 (70 6 fluorescence recovery with the same efficiency in WT and APP/PS1
8% n 5 5) though to a lesser extent. The latter observation is in agree- mice fed with water (38 6 3% and 40 6 3%, n 5 5 and 3, respectively)
ment with the aforementioned work (Orellana et al., 2011b) showing that or with boldine (38 6 3% and 40 6 3%, n 5 3 and 3, respectively).

in hippocampal slices acutely treated with Ab amyloid peptide, the result- Altogether, these results showed that boldine administred in vivo

ing HC activation in glial cells activates in turn Panx1 HCs in neurons. prevents HC activation in astrocytes and microglial cells but does not
modify astroglial gap junctional communication, indicating that the
boldine dose used in the present work is innocuous on gap junctional
3.4 | Long-term in vivo administration of boldine
communication of normal and APP/PS1 brain.
prevents the increase in HC activity but does not
affect gap junctional communication in astrocytes
3.5 | In vivo boldine administration has marginal
from APP/PS1 mice
effects on the amyloid pathology
Since, in some pathological situations, e.g., diabetes or atherosclerosis
The distribution and amount of Ab plaques in APP/PS1 mice fed with
(i) boldine has been administered orally in rodents without toxic effects
water or water plus boldine was examined after Ab immunostaining
(Hern
andez-Salinas et al., 2013; Santanam et al., 2004) and (ii) boldine
(Figure 5a). Typically, Ab plaques of various sizes appeared scattered in
was reported to cross the blood–brain barrier (Loghin et al., 2003),
the cortex and the hippocampus and the areas occupied by Ab depos-
these properties prompted us to investigate the effects of a long-term
its were not significantly different between treated and untreated mice
boldine treatment of APP/PS1 mice in which HCs are chronically
(Figure 5b), indicating that the amyloid burden was not modified by
activated in astrocytes (Yi et al., 2016). Boldine was delivered in the
long-term treatment with boldine. Also, a characteristic feature of AD
drinking water of mice (30 mg/kg), starting at 6 months, an age at
is the reactive gliosis associated with Ab plaques that involves both
which these mice begin to develop the amyloid pathology (Garcia-
activated microglial cells and reactive astrocytes, exhibiting enhanced
Alloza et al., 2006; Jankowsky et al., 2004). The experiment ended after GFAP content (Figure 5c). Hence, the extent of reactive gliosis was
12 weeks treatment in 9-month-old mice, when HC activity is high (Yi assessed by measuring the expression of GFAP and Iba1 as markers of
et al., 2016) and amyloid plaques are widespread in the cortex and the astrocytes and microglial cells, respectively, in APP/PS1 mice fed either
hippocampus with deleterious consequences on neighboring neurons with water or water plus boldine. It was found that both markers
(Kuchibhotla et al., 2008; Xie et al., 2013). increased in APP/PS1 mice compared to age-matched WT mice (data
First, we examined the status of HCs in the hippocampus of these not shown) but neither GFAP nor Iba1 levels were modified in
mice (Figure 4) to determine whether the boldine property as glial HC APP/PS1 mice fed with water plus boldine, compared to untreated
blocker when acutely applied on brain slices was replicated by in vivo APP/PS1 mice (Figure 5d). Finally, since anti-inflammatory actions of
prolonged treatment. EtBr uptake experiments were carried out in boldine have been described in other tissues (Backhouse et al., 1994),
parallel in acute brain hemisphere slices of control (untreated) and we measured the level of two pro-inflammatory cytokines, TNF-a, and
boldine-treated APP/PS1 mice. As expected, in control mice, the IL-1b, potentially involved in HC activation (Retamal et al., 2007) in
fluorescent signal detected in the nuclei of astrocytes, whatever their APP/PS1 mice fed with water plus boldine or only water (Figure 5d,e).
distance from the plaques (far 50 mm or close to plaques), was TNF-a level was slightly but significantly reduced (from 96.4 6 4.6 to
abolished by acute CBX treatment (Figure 4a,b). In boldine-treated 84.4 6 2.1 pg/mg protein, n 5 5 and 5, respectively) when mice were
APP/PS1 mice, the fluorescent signal detected in the nuclei of treated with boldine compared to control mice while IL-1b level had a
astrocytes far and close to plaques was inconspicuous and represented tendency to decrease that was not significant (from 642.0 6 28.8 to
only 13 6 4% (n 5 5) and 10 6 2% (n 5 5) of the signal measured 569.6 6 18.8 pg/mg protein, n 5 5 and 5, respectively). Altogether,
in untreated APP/PS1 mice, respectively (Figure 4a,b). Also, the these observations indicate that boldine treatment of APP/PS1 mice
fluorescent EtBr signal in microglial cell nuclei of APP/PS1 mice treated did not modify the amyloid-dependent process including the reactive
with boldine was hardly detectable (8 6 4%, n 5 5, of the signal in gliosis although it slightly reduced the inflammatory context.
1616 | YI ET AL.

FIGURE 4 Long-term in vivo boldine administration in APP/PS1 mice prevents the increase in HC activity and does not affect gap
junctional communication. (a, b) Analysis of glial EtBr uptake rate in brain hemisphere slices of 9-month-old APP/PS1 mice fed with water
(CTL), or with water plus boldine. Acute CBX (200 mM) treatment of the brain slice was used as a control of HC activity inhibition (a) EtBr
uptake in astrocytes close (left) and far (right) to plaques normalized with the control condition (CTL) in APP/PS1 mice (***p < .001, vs. CTL;
ANOVA Dunnett’s test; mean 6 SEM, n 5 5). (b) EtBr uptake in microglia normalized with the control condition (CTL) in APP/PS1 mice
(***p < .001 vs. CTL; ANOVA Dunnett’s test; mean 6 SEM, n 5 5). (c–f) Gap junctional communication between astrocytes was assessed by
FRAP on SR101-loaded brain slices from 9-month-old WT (c,d) or APP/PS1 mice (e, f) treated or not with boldine. (c, e) Time-lapse percent-
age of fluorescence intensity (FI) normalized with F0 in presence or absence of CBX. (d, f) Quantification of fluorescence recovery percent-
age showing no significant differences between mice and/or treatment indicating that boldine did not affect astroglial gap junctional
coupling (ns, not significant; unpaired Student’s t-test; mean 6 SEM, WT n 5 8; WT Boldine, WT 1 CBX n 5 5; WT Boldine 1 CBX n 5 3;
APP/PS1, APP/PS1 Boldine n 5 9; APP/PS1 1 CBX, APP/PS1 Boldine 1 CBX n 5 3)

3.6 | In vivo boldine treatment reduces intracellular

APP/PS1 mice treated or not with boldine were loaded with Fluo-4AM
Ca21 signal and ATP/glutamate release
that is selectively uptaken by astrocytes (Figure 6a). As expected, the
Since we have previously shown in APP/PS1 mice that the chronic fluorescence intensity of astrocytes in APP/PS1 mice in control
activation of astroglial HCs contributed to the increase in intracellular condition (water fed) was high compared to WT mice. Treatment of
astroglial Ca21 signal (Ca21
i ) as well as to gliotransmitter release (Yi WT mice with boldine had no effect on the level of Ca21
i signal in
et al., 2016), we examined whether these parameters were impacted astrocytes of WT mice while it reduced significantly Ca21
i in APP/PS1
after in vivo boldine treatment. First, hippocampal slices from WT and mice (27 6 7%, n 5 4) (Figure 6a,b).
YI ET AL. | 1617

F I G U R E 5 Long-term in vivo boldine administration has marginal effects on the amyloid pathology. (a) Distribution of Ab plaques in brain
sections of APP/PS1 mice fed with water (left) or water plus boldine (right); scale bar, 1,000 lm. (b) Quantification of the surface occupied
by amyloid plaques in the cortex and hippocampus of 9-month-old APP/PS1 mice treated or not with boldine. The plaque surface was nor-
malized to the area of cortex and hippocampus (ns, not significant; unpaired Student’s t test; mean 6 SEM, n 5 n 5 5). (c) Immunostaining
image showing activated microglial cells (green) and reactive astrocytes (blue) contacting Ab plaques (red); note the difference in the mor-
phology of glial cells contacting plaques versus those far from plaques; scale bar: 20 lm. (d) Western blots of GFAP (left) and Iba1 (right)
content in brain hemispheres of 9-month-old APP/PS1 mice fed with water or boldine (4 mice per condition). Quantitative analysis showed
that their levels were unchanged after long-term treatment with boldine (unpaired Student’s t test; mean 6 SEM, n 5 4). (e) Proinflammatory
cytokine levels in brain hemispheres from APP/PS1 mice measured by ELISA. (e) TNF-a levels (left) were significantly reduced after long
treatment with boldine (*p < .05; unpaired Student’s t test; mean 6 SEM, n 5 5). IL-1b levels (right) showed a decreasing tendency though
not significant (p 5 .06; unpaired Student’s t test; mean 6 SEM, n 5 5)

In a recent study (Yi et al., 2016) we have shown that the level of between samples from WT or APP/PS1 mice fed or not with boldine
ATP and glutamate are increased in APP/PS1 mice and that the lack of (Supporting Information Figure S2).
Cx43 in astrocytes, using a targeted Cx43 KO, reduces these levels
indicating that gliotransmission is enhanced in the AD model mouse.
3.7 | Boldine treatment of APP/PS1 mice reduces
Consequently, we next measured the level of these gliotransmitters
neuronal suffering
released from brain hemisphere slices of WT and APP/PS1 mice fed or
not with boldine. First, ATP release was directly measured in the We previously showed that knocking out Cx43 in the astrocytes of
extracellular medium of slices from WT mice fed with water or water APP/PS1 mice removed the predominant component of HC activity in
plus boldine using a luciferin–luciferase luminescence assay (Figure 6c). hippocampal astrocytes and allowed to alleviate neuronal suffering (Yi
As previously described (Yi et al., 2016), the amount of ATP released et al., 2016). Therefore, we have investigated whether the inhibition of
from brain slices of untreated APP/PS1 mice (12 6 3 nmol/mg protein, HC activity in both astrocytes and microglial cells provided by long-
n 5 5), was approximately twice that of WT mice (7 6 1 nmol/mg pro- term in vivo boldine treatment of APP/PS1 mice impacted neuronal
tein, n 5 4). The treatment of APP/PS1 mice with boldine strongly suffering. Firstly, the level of oxidative stress in hippocampal neurons
reduced the level of ATP to 8.0 6 2.8 nmol/mg protein, n 5 5) while it was assessed in living slices using MitoSOX, a fluorogenic dye targeted
had no effect in WT mice (7 6 2 nmol/mg protein, n 5 4). Second, L- to mitochondria that exhibits a red fluorescence when oxidized by
glutamate release was assessed in similar samples by measuring the superoxide ions. In APP/PS1 mice, the strong red fluorescent signal
amount of L-glutamate in the extracellular medium and in the slices. detected in hippocampal pyramidal cell layer (Figure 7a) was abolished
We then calculated the ratio of L-glutamate in the extracellular medium when mice were fed with water plus boldine. Indeed, the level of fluo-
to the total amount to determine the released fraction of L-glutamate rescence measured in APP/PS1 mice treated with boldine was compa-
in each condition (Figure 6d). L-glutamate released from slices of WT rable to that of WT mice, indicating that a long-term boldine treatment
mice was not impacted by boldine treatment. In contrast, in APP/PS1 prevented neuronal oxidative stress. Secondly, the impact of boldine
mice that exhibited a significantly higher level of L-glutamate release, treatment on the abundance of neuritic dystrophies nearby Ab plaques,
boldine treatment reduced this elevated level by 50 6 17% (n 5 5) to an indicator of neuronal injury (Sanchez-Varo et al., 2012), was also
values similar to those of WT mice. In parallel, GABA and D-serine lev- investigated by analyzing the expression of reticulon 3 (RTN3) in brain
els were evaluated as well and no significant differences were found sections of APP/PS1 mice fed with either water or water plus boldine.
1618 | YI ET AL.

FIGURE 6 In vivo long-term treatment with boldine prevents the increase in Ca21 signal and gliotransmitter release in APP/PS1 mice. (a)
Fluorescence images showing Ca21 indicator Fluo-4 AM loaded in the astrocytes of brain hemisphere slices from 9-month-old WT and
APP/PS1 treated or not with boldine; scale bar, 20 lm. (b) Resting Ca21 signal expressed as mean Fluo-4 AM fluorescence intensity in arbi-
trary units (au) in astrocytes, showing that long-term boldine treatment prevented significantly the increase in Ca21 signal in astrocytes of
APP/PS1 mice (*p < .05, v.s. APP/PS1 Water; Student’s t test; mean 6 SEM, n 5 4). (c, d) Gliotransmitter released from brain hemisphere sli-
ces of WT and APP/PS1 treated or not with boldine, respectively. ATP (c) and glutamate (d) release from APP/PS1 brain slices after long-
term treatment with boldine was as low as in brain slices of WT mice. (*p < .05, vs. APP/PS1 water; ANOVA Dunnett’s test; mean 6 SEM,
n 5 5)

Sections were immunostained for Ab and RTN3 a protein forming astroglial HCs are chronically activated in the hippocampus (Yi et al.,
aggregates that accumulate within dystrophic neurites in AD (Hu et al., 2016), have a significant neuroprotective effect as indicated by the
2007). High density spots of various sizes characteristic of RTN3 decrease in two markers of neuronal suffering in the hippocampus.
immunoreactive dystrophic neurites (RIDNs) were observed at the level Consequently, boldine represents a drug treatment that opens the way
of Ab plaques but they appeared smaller and fewer in APP/PS1 mice to design novel protective molecules allowing to reduce neuronal
treated with boldine (Figure 7c). Indeed, a significant decrease in damage associated with neurodegenerative situations, in particular the
(RIDN) was measured in boldine-treated mice (Figure 7d). These results amyloid pathology.
indicate that three months in vivo boldine treatment is neuroprotective Abnormal and sustained activation of glial HCs was reported in an
in the APP/PS1 murine model of AD. increasing number of nervous system pathologies with detrimental
effects on neuronal function and survival reviewed in (Bosch & Kielian,
2014; Davidson et al., 2013a; Orellana et al., 2016). Hence, diverse
4 | DISCUSSION strategies aimed at blocking HC activity using genetic or pharmacologi-
cal tools have been developed (Huang et al., 2012; O’Carroll et al.,
In the present study, we have shown that a natural alkaloid compound, 2013; Schulz et al., 2015; Takeuchi & Suzumura, 2014). Most of them
the boldine extracted from a Chilean tree, inhibits HC activity triggered suppressed Cx43 expression and/or function, which is considered as
in glial cells acutely or chronically by different pathological stimuli while the main HC constituent in astrocytes (Giaume et al., 2013) by
it does not modify astroglial gap junctional communication. We also knocking-out Cx43 gene, using antisense tools, antibodies or mimetic
provide evidence that boldine can be administered in vivo as a pharma- peptides. However, all these attempts likely impacted as well astroglial
cological tool with similar effects on these two channel functions. Its gap junctional communication making mechanistic interpretation of the
long-term oral administration in a murine model of familial AD, in which observations biased since both Cx channel functions were affected.
YI ET AL. | 1619

FIGURE 7 Reduced neuronal damage in the hippocampus of APP/PS1 mice treated with boldine. (a) TO-PRO staining (blue) and MitoSOX
signal (red) in hippocampal slices showing that the level of mitochondrial superoxide ions was higher in the hippocampal pyramidal cell layer
in water-fed APP/PS1 mice when compared with WT and boldine-treated APP/PS1 mice; scale bar, 20 lm. (b) Mean fluorescence intensity
of MitoSOX in the pyramidal cell layer of indicated conditions (*p < .05; Student’s t-test; mean 6 SEM, n 5 5). (c) Images of immunostainings
for Ab plaques (green) and RTN3 (red) in hippocampal sections from APP/PS1 mice treated or not with boldine. RTN3 labeled dystrophic
neurites appeared as bright red dots concentrated in and around Ab plaques and were less abundant after boldine treatment; scale bar, 20
lm. (d) Quantification of the area occupied by RTN3 stained dystrophic neurites associated with Ab plaques in APP/PS1 mice treated or
not with boldine (*p < .05; unpaired Student’s t test; mean 6 SEM, n 5 5)

Thus, an approach blocking only HC function in glial cells may repre- familial AD, the APP/PS1 mouse in which HCs formed by Cx43 with
sent a promising tool that tunes down their deleterious effects in brain an additional Panx1 component nearby Ab plaques are chronically acti-
pathologies. Also for neurodegenerative diseases, the design of a thera- vated in astrocytes (Yi et al., 2016). Moreover, we revealed that in
peutical strategy has to take into account the need of a chronic treat- APP/PS1 mice, Panx1 and Cx43 HCs were also activated in microglial
ment and the use of molecules passing the BBB. Based on our present cells and blocked by boldine. Such activation is consistent with previ-
work, boldine may fulfill these criteria. ous observations showing that an acute treatment of hippocampal sli-
First, we have shown that its ability to block HCs initially reported ces with Ab peptide induced HC activity mediated by Panx1 and Cx43
in mesanglial cells (Hern
andez-Salinas et al., 2013) also applies to HCs in microglial cells (Orellana et al., 2011b). Since in the latter experimen-
in astrocytes and microglia. Indeed, acute application of boldine in tal paradigm, glial HC activation can trigger HC openings in hippocam-
diverse in vitro models inhibited glial HCs as efficiently as CBX. This pal neurons with deleterious consequences (Orellana et al., 2011a,b),
was shown (i) in cultured astrocytes in which Cx43 HC activation was we have verified whether this also occurred in a chronic situation.
triggered by pro-inflammatory treatment; (ii) in living brain slices Indeed, we also observed that HCs were activated in hippocampal
acutely treated by pro-inflammatory treatment that triggered HCs pyramidal cells of APP/PS1 mice. It is likely that the same sequence of
formed by Cx43 or Panx1 in astrocytes and in microglial cells as well as events reported in the acute situation prevails in APP/PS1 mice since
in HeLa-Panx1-YPF cells, and (iii) in acute brain slices of a model of Gap26 mimetic peptide which inhibits Cx43 HCs strongly reduced EtBr
1620 | YI ET AL.

uptake in neurons that do not express Cx43 at adult age. Neurons do and release of glutamate and ATP highlight the interest of using this
express Panx1 and Cx36 (Sohl, Maxeiner, & Willecke, 2005), that can compound. In APP/PS1 mice astroglial [Ca21]i level is abnormally ele-
10
both participate to the EtBr uptake observed. Indeed panx1 was very vated (Kuchibhotla, Lattarulo, Hyman, & Bacskai, 2009). Since Cx43
efficient to reduce HC opening in neurons while CBX or boldine abol- HCs provide a pathway of Ca21 entry (De Bock et al., 2012; S

anchez,
ished such activation. Hence, these results suggest that HC activation Orellana, Verselis, & S

aez, 2009; Schalper et al., 2010), they contribute
in neurons is mostly a downstream consequence of glial HC activity
, Yang,
to rise [Ca21]i level in glial cells (Abudara et al., 2015; Garre
although we cannot exclude that a small fraction of neuronal Panx1 Bukauskas, & Bennett, 2016) as also reported in astrocytes of APP/
HCs can also be triggered directly. Second, boldine applied acutely in PS1 mice (Delekate et al., 2014; Yi et al., 2016). Indeed, we recently
cultured cells or in brain slices of WT or APP/PS1 mice did not modify reported that Ca21
i signal was reduced to WT level when Cx43 was
the level of gap junctional communication between astrocytes. Inter- knocked out in astrocytes of these mice and Ca21 transients amplitude
estingly, we confirmed that the amyloid pathology had no impact on was reduced upon CBX treatment. Accordingly, in APP/PS1 mice, bol-
this parameter since astroglial gap junctional communication is similar dine treatment prevented the increase in HC activity and intracellular
in APP/PS1 and WT mice (Cruz, Ball, & Dienel, 2010; Yi et al., 2016). In Ca21 signal in astrocytes that remained close to that of WT mice. Inter-
other cell types, boldine was shown to prevent the loss of Cx43 medi- estingly, elevated Ca21 signal in astrocytes can in turn triggers a cas-
ated gap junctional communication induced by a tumor promoting cade of [Ca21]i-dependent pathways including gliotransmitter release.
agent in liver epithelial cells (Hu, Speisky, & Cotgreave, 1995) or in a Indeed, ATP and glutamate release associated to HC activity was dem-
diabetic environment in mesangial cells (Hern
andez-Salinas et al., onstrated in glial cells reviewed in (Decrock et al., 2015): for instance in
2013), but its effect on basal gap junctional communication was not cultured cells, both Panx1 and Cx43 HCs participate to ATP release in
examined. Altogether, these results suggest that, although boldine can microglial cells (Ma et al., 2014; Orellana, Montero, & von Bernhardi,
rescue a deficient communication, it does not impact the basal level of 2013) and in astrocytes (Iglesias, Dahl, Qiu, Spray, & Scemes, 2009;
gap junctional communication. The preservation of this intercellular Kang et al., 2008); also glutamate release associated to Cx HCs involv-
communication is important given the key role played by astroglial gap ing Cx43 and Cx32 was demonstrated in astrocytes (Ye et al., 2003)
juntions in brain homeostasis via spatial buffering and trafficking of and in microglial cells (Takeuchi et al., 2006), respectively. Moreover, in
ions, glutamate energetic metabolites, and small signaling molecules acute or chronic pathological situations, sustained activation of HCs
(Giaume et al., 2010; Rouach et al., 2008). can contribute to an excessive release of ATP and/or glutamate with
The ability of boldine to affect glial HC function prompted us to neurotoxic effects that are alleviated by reducing Cx43 expression or
investigate its effect in vivo by treating APP/PS1 mice with this drug. In function (Bennett et al., 2012; Schulz et al., 2015; Takeuchi & Suzu-
view of a potential use as a therapeutic tool, the treatment began in mura, 2014). For instance, following spinal cord injury, the excessive
relatively young mice aged of 5–6 months when the first signs of the and sustained ATP release was strongly reduced in mice lacking Cx43
amyloid pathology are detectable (Garcia-Alloza et al., 2006; Mei, Ezan, in astrocytes with beneficial outcome, smaller lesion area and motor
Giaume, & Koulakoff, 2010). Mice were treated for 3 months to exam- recovery improvement (Huang et al., 2012), and blocking Cx43 expres-
ine the impact of boldine when deleterious effects of astroglial Cx43 sion with antisense oligonucleotides or Cx43 mimetic peptides applied
HCs on neurons are detected (Yi et al., 2016). First, we have shown locally or administered systemically reduced tissue damage and inflam-
that such long-term treatment allowed keeping glial HCs inactive while mation and improved functional recovery (Cronin, Anderson, Cook,
gap junctional communication was maintained at the same level as in Green, & Becker, 2008; Mao et al., 2017; O’Carroll, Alkadhi, Nicholson,
WT mice, indicating that boldine administered orally is efficiently deliv- & Green, 2008). Also the increase in glutamate levels observed after
ered into the brain. This statement is in agreement with the already perinatal brain ischemia was strongly reduced when Cx43 HCs were
reported crossing of the BBB by boldine. Indeed, administered in the inhibited by mimetic peptides (Li et al., 2015); the consequence of such
drinking water of rats at a concentration of 50 mg/kg, boldine is rapidly Cx43 targeting was a reduction of neuronal death and brain infarct size
absorbed, concentrates in the liver but reaches also the brain (Jimenez as well as an improved functional recovery (Davidson et al., 2012;
& Speisky, 2000; Loghin et al. 2003) showed an effect on the striatal Davidson, Green, Nicholson, Bennet, & Gunn, 2013b; Li et al., 2015). In
dopamine levels after one injection of boldine while recently De lima AD patients, increased level of ATP was measured in plaque bearing
et al. (2017) showed that continuous boldine treatment of mice during brain areas compared to that in control patients (Mecheri et al., 1997)
1 week after stroke was neuroprotective. However, we cannot exclude and in FAD mice, ATP and glutamate release was higher than in WT
that the entry of boldine within the brain parenchyma is also facilitated mice (Delekate et al., 2014; Yi et al., 2016). Knocking out Cx43 in astro-
by BBB weakening reported in different models of AD mice (Paul, cytes of APP/PS1 mice reduced the levels of ATP and glutamate with a
Strickland, & Melchor, 2007; Ujiie, Dickstein, Carlow, & Jefferies, 2003) concomitant reduction in neuronal suffering (Yi et al., 2016). These
and in human AD patients in which cerebral amyloid angiopathy, occur- results confirm the important role played by glial HCs in gliotransmitter
ring in 80% of cases, is correlated with a defective BBB reviewed in release. Interestingly, the use of a derivative of CBX, INI-0602, which
Bell and Zlokovic (2009) and Zlokovic (2011). inhibits glutamate release from microglial cells by blocking activated
The consequences of boldine treatment on parameters known to Cx32 HCs, improved memory impairment in another FAD mouse
be dysregulated in APP/PS1 mice, i.e., astroglial increase in Ca21
i signal model (Takeuchi et al., 2011). This drug also led to behavioral
YI ET AL. | 1621

F I G U R E 8 Schema summarizing how boldine treatment of APP/PS1 mice may have neuroprotective effects. Boldine administered in vivo
in APP/PS1 mice (1) prevents HC activation in astrocytes and (2) in microglial cells; (3) prevents the increase in [Ca21]i signal in astrocytes
by inhibiting HCs; (4) reduces TNF-a release from glia, preventing Panx HCs activation in astrocytes contacting Ab plaques; (5) reduces glu-
tamate and ATP released from astrocytes and/or from microglial cells, tuning down the excessive activation of neuronal NMDA and P2X
receptors that triggers intracellular neurotoxic cascades resulting in neurodegeneration (6). Boldine might also inhibit Panx HCs in neurons
directly (7). M, microglia; A, astrocyte; N, neuron; DN, degenerating neuron

amelioration in other models of neurodegenerative diseases, ALS and complex, does not interfere with the elevated mitochondrial superoxide
PD (Suzuki, Ono, & Sawada, 2014; Takeuchi et al., 2011), and in spinal anion levels in hippocampal neurons (Ma et al., 2011). Hence, it is likely
cord injury (Umebayashi et al., 2014). The underlying mechanisms were that the increased oxidative stress assessed by the level of superoxide
attributed to the inhibitory effect of INI-0602 on microglial Cx32 HCs anions in hippocampal neurons of APP/PS1 mice reflects a long-term
(Takeuchi et al., 2006) although direct evidence is lacking. Here, we neuronal suffering, which is mitigated by preventing Cx43 HC openings
show that boldine treatment of APP/PS1 mice reduced ATP and gluta- in the presence of boldine. It is worth noting that knocking out Cx43 in
mate levels back to WT range. Since in APP/PS1/Cx43 KO mice similar the astrocytes of these mice also resulted in a reduction in mitochon-
results were obtained, our data suggest that astroglial Cx43 HCs are drial superoxide anions (Yi et al., 2016). HC silencing, either by sup-
major actors in glutamate and ATP release while the fraction of Panx1 pressing Cx43 expression (Yi et al., 2016) or by maintaining them
HCs formed by Panx1 in astrocytes nearby Ab plaques and the HCs of inactive with boldine would prevent Ca21 inflow and activation of dif-
microglial cells have only a marginal contribution. ferent Ca21-dependent metabolic pathways that generate reactive
The chronic treatment of APP/PS1 mice with boldine presented oxygen species (Valko et al., 2007). (ii) the inhibitory effect of boldine
here had neuroprotective effects (Figure 8). Although the progression demonstrated here on HCs may be direct through a yet unknown
of the amyloid pathology was not modified, based on the abundance of mechanism but also indirect via its antioxidant property that may affect
amyloid plaques and the level of reactive gliosis, neuronal suffering the cellular redox status. Indeed, Cx43 and Panx1 hemichannel opening
was alleviated as evidenced by the decrease in the abundance of dys- and closing can be modulated by redox signaling molecules (Retamal,
trophic neurites expressing RTN3 nearby plaques and also by the 2014). An increase of ROS or/and reactive nitrogen species favours
reduction in the level of oxidative stress in hippocampal neurons which HC openings while reducing agents have an inhibitory effect. Thus bol-
returned to WT levels. Since boldine is known to have a wide array of dine can contribute to close HCs by modifying the redox status of glial
actions, we cannot exclude that some of them participate to the neuro- cells. (iii) Boldine may also exert protective effects on endothelial cells,
protective effects observed in APP/PS1 mice. First, the antioxidant as reported in models of diabetes and hypertension, by scavenging
properties of boldine (Lau et al., 2015) reported in models of hyperten- ROS via several mechanisms. (Lau et al., 2015). Indeed, AD is associ-
sion and diabetes could contribute to its neuroprotective effect in dif- ated with microvascular dysfunction, altered protein expression in par-
ferent ways: (i) by scavenging reactive oxygen species (ROS), boldine ticular Ab transporters, LRP1 and RAGE, by endothelial cells and
could reduce the oxidative stress in the hippocampal neurons of APP/ defective BBB that impact Ab clearance (Zlokovic, 2011). Although the
PS1 mice However, the antioxidant effect of boldine was shown to underlying mechanisms involved in BBB dysfunction are complex,
occur through the inhibition of NADPH superoxide axis (Lau et al., oxidative stress and diverse inflammatory mediators seem to play a
2015), whereas in APP/PS1 mice the inhibition of this axis, by using role (Zenaro, Piacentino, & Constantin, 2016). In this context, boldine
either DPI or by deleting the gp91phox subunit of NADPDH oxidase treatment could improve BBB function by its protective action on
1622 | YI ET AL.

endothelial cells and, as a consequence, could ameliorate Ab clearance R EF ER E N CE S

and reduce Ab accumulation in boldine treated animals. However, we Abudara, V., Roux Dallerac, L., Matias, G., Dulong, I. J., Mothet, J.,
observed that the progression of the amyloid deposition is not modi- Rouach, P. N., & Giaume, C. (2015). Activated microglia impairs neu-
roglial interaction by opening Cx43 hemichannels in hippocampal
fied by boldine treatment suggesting that if boldine has an effect on
astrocytes. Glia, 63, 795–811.
BBB, it is only marginal and likely not essential in its neuroprotective
Almad, A. A., Doreswamy, A., Gross, S. K., Richard, J. P., Huo, Y.,
action. Second, boldine was also reported to have anti-inflammatory Haughey, N., & Maragakis, N. J. (2016). Connexin 43 in astrocytes
effects (Lau et al., 2015), In agreement with this, we have shown that contributes to motor neuron toxicity in amyotrophic lateral sclerosis.
the level of TNFa was slightly but significantly reduced in APP/PS1 Glia, 64, 1154–1169.

mice treated with boldine while IL-1ß level was not significantly Backhouse, N., Delporte, C., Givernau, M., Cassels, B. K., Valenzuela, A.,
& Speisky, H. (1994). Anti-inflammatory and antipyretic effects of
affected. We previously showed that inflammation is involved in Panx1
boldine. Agents Actions, 42, 114–117.
HC activitation in astrocytes close to amyloid plaques in APP/PS1 mice
Bell, R. D., & Zlokovic, B. V. (2009). Neurovascular mechanisms and
(Yi et al., 2016). Indeed acute application of minocycline that reduces blood–brain barrier disorder in Alzheimer’s disease. Acta Neuropathol-
the level of TNF-a in APP/PS1 brain slices was able to inhibit Panx1 ogy, 118, 103–113.
HCs in these reactive astrocytes. Hence, the antiinflammatory effect of Bennett, M. V., Garre, J. M., Orellana, J. A., Bukauskas, F. F., Nedergaard,
boldine can be one of the mechanisms involved in the inhibition of M., Giaume, C., & Saez, J. C. (2012). Connexin and pannexin hemichan-
Panx1 HCS observed in APP/PS1 mice. nels in inflammatory responses of glia and neurons. Brain Research,
1487, 3–15.
The absence of toxicity of boldine in a large range of doses includ-
Bosch, M., & Kielian, T. (2014). Hemichannels in neurodegenerative dis-
ing the concentration used presently (Jimenez & Speisky, 2000;
eases: is there a link to pathology? Frontiers in Cellular Neuroscience,
O’Brien et al., 2006), its ability to reach the brain after in vivo adminis- 8, 242.
tration, its water solubility and thus its facility to be delivered by oral Bravo, D., Ibarra, P., Retamal, J., Pelissier, T., Laurido, C., Hernandez, A.,
administration, as well as its relatively low cost compared to mimetic & Constandil, L. (2014). Pannexin 1: A novel participant in neuro-
peptides, make it a promising therapeutic molecule to fight deleterious pathic pain signaling in the rat spinal cord. Pain, 155, 2108–2115.

effects of HC activity in diverse brain pathologies. Indeed, a neuropro- Bruzzone, R., Barbe, M. T., Jakob, N. J., & Monyer, H. (2005). Pharmaco-
logical properties of homomeric and heteromeric pannexin hemichan-
tective effect of boldine in a model of stroke induced by a permanent
nels expressed in Xenopus oocytes. Journal of Neurochemistry, 92,
MCAO in mice was recently described (de Lima et al., 2017). In mice 1033–1043.
treated with 25 mg/kg boldine during 5 days after stroke, the size of Chen, G., Park, C. K., Xie, R. G., Berta, T., Nedergaard, M., & Ji, R. R.
the infarct area and the neuroinflammatory response were reduced, (2014). Connexin-43 induces chemokine release from spinal cord
while cell viability and neurological scores of the mice were improved. astrocytes to maintain late-phase neuropathic pain in mice. Brain: A
Journal of Neurology, 137, 2193–2209.
The mechanisms underlying such neuroprotection were not examined,
Contreras, J. E., S
anchez, H. A., Eugenin, E. A., Speidel, D., Theis, M., Wil-
but an inhibition of glial HCs that are activated in ischemic conditions
lecke, K., . . . S
aez, J. C. (2002). Metabolic inhibition induces opening
(Contreras et al., 2002; Giaume et al., 2013) could likely contribute to of unapposed connexin 43 gap junction hemichannels and reduces
its beneficial effects. Moreover, boldine is present in many nutraceuti- gap junctional communication in cortical astrocytes in culture. Pro-
cal preparations indicated in the treatment of diverse human disorders ceedings of National Academic Sciences USA, 99, 495–500.

in particular digestive ones without secondary deleterious effect Cotrina, M. L., Gao, Q., Lin, J. H., & Nedergaard, M. (2001). Expression and
function of astrocytic gap junctions in aging. Brain Research, 901, 55–61.
(Speisky & Cassels, 1994). Hence, boldine appears as a new
Cronin, M., Anderson, P. N., Cook, J. E., Green, C. R., & Becker, D. L.
neuroprotective drug that could be administered in therapeutical
(2008). Blocking connexin43 expression reduces inflammation and
treatment of both acute and chronic brain diseases to reduce neuronal improves functional recovery after spinal cord injury. Molecular Cell &
damages. Neuroscience, 39, 152–160.
Cruz, N. F., Ball, K. K., & Dienel, G. A. (2010). Astrocytic gap junctional
ACKNOWLE DGMENT communication is reduced in amyloid-beta-treated cultured astro-
cytes, but not in Alzheimer’s disease transgenic mice. ASN Neuro, 2,
We thank Edwige Amigou and Isabelle Matias for technical assis- e00041.
tance, J. Teillon for confocal image analysis training, the animal facil- Davidson, J. O., Green, C. R., Bennet, L., Nicholson, L. F., Danesh-Meyer,
ity staff and Prof. Riqiang Yan for the gift of anti-RTN3 antibodies. H., O’Carroll, S. J., & Gunn, A. J. (2013a). A key role for connexin
~es, Härting Co., Santiago, Chile, for prepar-
We thank Dr Cecilia Bran hemichannels in spreading ischemic brain injury. Currents Drug
Targets, 14, 36–46.
ing boldine hydrochloride. C. Yi was supported by FRM and LECMA.
Davidson, J. O., Green, C. R., Nicholson, L. F., Bennet, L., & Gunn, A. J.
This work was funded by CRPCEN and LECMA and ICM-Economía
(2013b). Connexin hemichannel blockade is neuroprotective after,
P09–22F Centro Interdisciplinarion de Neurociencias de Valparaíso but not during, global cerebral ischemia in near-term fetal sheep.
(to JCS). Experimental Neurology, 248, 301–308.
Davidson, J. O., Green, C. R., Nicholson, L. F., O’Carroll, S. J., Fraser, M.,
Bennet, L., & Gunn, A. J. (2012). Connexin hemichannel blockade
CONFLICT OF INTERE ST
improves outcomes in a model of fetal ischemia. Annals of Neurology,
The authors declare no conflict of interest. 71, 121–132.
YI ET AL. | 1623

De Bock, M., Wang, N., Bol, M., Decrock, E., Ponsaerts, R., Bultynck, G., . . . Huang, Y., Grinspan, J. B., Abrams, C. K., & Scherer, S. S. (2007). Pan-
Leybaert, L. (2012). Connexin 43 hemichannels contribute to cytoplas- nexin1 is expressed by neurons and glia but does not form functional
mic Ca21 oscillations by providing a bimodal Ca21-dependent Ca21 gap junctions. Glia, 55, 46–56.
entry pathway. Journal of Biological Chemistry, 287, 12250–12266. Iglesias, R., Dahl, G., Qiu, F., Spray, D. C., & Scemes, E. (2009). Pannexin
de Lima, N. M., Ferreira, E. O., Fernandes, M. Y., de Lima, F. A., Neves, 1: The molecular substrate of astrocyte “hemichannels”. Journal of
K. R., do Carmo, M. R., & de Andrade, G. M. (2017). Neuroinflamma- Neuroscience, 29, 7092–7097.
tory response to experimental stroke is inhibited by boldine. Behav- Jankowsky, J. L., Fadale, D. J., Anderson, J., Xu, G. M., Gonzales, V., Jen-
ioural Pharmacology, 28, 223–237. kins, N. A., . . . Copeland, N. G. (2004). Mutant presenilins specifically
Decrock, E., De Bock, M., Wang, N., Bultynck, G., Giaume, C., Naus, C. elevate the levels of the 42 residue beta-amyloid peptide in vivo: Evi-
C., . . . Leybaert, L. (2015). Connexin and pannexin signaling path- dence for augmentation of a 42-specific gamma secretase. Human
ways, an architectural blueprint for CNS physiology and pathology? Molecular Genetics, 13, 159–170.
Cellular and Molecular Life Sciences: CMLS, 72, 2823–2851. Jimenez, I., & Speisky, H. (2000). Biological disposition of boldine: In vitro
Delekate, A., Fuchtemeier, M., Schumacher, T., Ulbrich, C., Foddis, M., & and in vivo studies. Phytotherapy Research, 14, 254–260.
Petzold, G. C. (2014). Metabotropic P2Y1 receptor signalling medi- Johnson, R. G., Le, H. C., Evenson, K., Loberg, S. W., Myslajek, T. M.,
ates astrocytic hyperactivity in vivo in an Alzheimer’s disease mouse Prabhu, A., . . . Sheridan, J. D. (2016). Connexin hemichannels:
model. Nature Communications, 5, 5422. Methods for dye uptake and leakage. Journal of Membrane Biology,
Fossat, P., Turpin, F. R., Sacchi, S., Dulong, J., Shi, T., Rivet, J. M., . . . 249, 713–741.
Mothet, J. P. (2012). Glial d-serine gates NMDA receptors at excita- Kang, J., Kang, N., Lovatt, D., Torres, A., Zhao, Z., Lin, J., & Nedergaard,
tory synapses in prefrontal cortex. Cerebral Cortex, 22, 595–606. M. (2008). Connexin 43 hemichannels are permeable to ATP. Journal
Garcia-Alloza, M., Robbins, E. M., Zhang-Nunes, S. X., Purcell, S. M., of Neurosciences, 28, 4702–4711.
Betensky, R. A., Raju, S., Frosch, M. P. (2006). Characterization of Kettenmann, H., Kirchhoff, F., & Verkhratsky, A. (2013). Microglia: New
amyloid deposition in the APPswe/PS1dE9 mouse model of Alzhei- roles for the synaptic stripper. Neuron, 77, 10–18.
mer disease. Neurobiology Disease, 24, 516–524.
Koulakoff, A., Ezan, P., & Giaume, C. (2008). Neurons control the expres-
Garden, G. A., & La Spada, A. R. (2012). Intercellular (mis)communication sion of connexin 30 and connexin 43 in mouse cortical astrocytes.
in neurodegenerative disease. Neuron, 73, 886–901. Glia, 56, 1299–1311.

, J. M., Yang, G., Bukauskas, F. F., & Bennett, M. V. (2016). FGF-1
Garre Kuchibhotla, K. V., Goldman, S. T., Lattarulo, C. R., Wu, H. Y., Hyman, B.
triggers pannexin-1 hemichannel opening in spinal astrocytes of T., & Bacskai, B. J. (2008). Abeta plaques lead to aberrant regulation
rodents and promotes inflammatory responses in acute spinal cord of calcium homeostasis in vivo resulting in structural and functional
slices. Journal of Neuroscience, 36, 4785–4801. disruption of neuronal networks. Neuron, 59, 214–225.
Giaume, C., Koulakoff, A., Roux, L., Holcman, D., & Rouach, N. (2010). Kuchibhotla, K. V., Lattarulo, C. R., Hyman, B. T., & Bacskai, B. J. (2009).
Astroglial networks: A step further in neuroglial and gliovascular Synchronous hyperactivity and intercellular calcium waves in astro-
interactions. Nature Reviews Neuroscience, 11, 87–99. cytes in Alzheimer mice. Science, 323, 1211–1215.
Giaume, C., Leybaert, L., Naus, C., & S
aez, J. C. (2013). Connexin and Lau, Y. S., Ling, W. C., Murugan, D., & Mustafa, M. R. (2015). Boldine
pannexin hemichannels in brain glial cells: Properties, pharmacology, ameliorates vascular oxidative stress and endothelial dysfunction:
and roles. Frontiers in Pharmacology, 4, 88. Therapeutic implication for hypertension and diabetes. Journal of Car-
Giaume, C., Orellana, J. A., Abudara, V., & Sa
ez, J. C. (2012). Connexin- diovascular Pharmacology, 65, 522–531.
based channels in astrocytes: How to study their properties. Methods Li, X., Zhao, H., Tan, X., Kostrzewa, R. M., Du, G., Chen, Y., . . . Xu, X.
in Molecular Biology, 814, 283–303. (2015). Inhibition of connexin43 improves functional recovery after
Giaume, C., & Theis, M. (2010). Pharmacological and genetic approaches ischemic brain injury in neonatal rats. Glia, 63, 1553–1567.
to study connexin-mediated channels in glial cells of the central nerv- Loghin, F., Chagraoui, A., Asencio, M., Comoy, E., Speisky, H., Cassels, B.
ous system. Brain Research & Review, 63, 160–176. K., & Protais, P. (2003). Effects of some antioxidative aporphine
Halassa, M. M., Fellin, T., & Haydon, P. G. (2007). The tripartite synapse: derivatives on striatal dopaminergic transmission and on MPTP-
Roles for gliotransmission in health and disease. Trends in Molecular induced striatal dopamine depletion in B6CBA mice. European Journal
Medicine, 13, 54–63. of Pharmaceutical Science, 18, 133–140.
Hern
andez-Salinas, R., Vielma, A. Z., Arismendi, M. N., Boric, M. P., S
aez, Ma, T., Hoeffer, C. A., Wong, H., Massaad, C. A., Zhou, P., Iadecola, C.,
J. C., & Velarde, V. (2013). Boldine prevents renal alterations in dia- . . . Klann, E. (2011). Amyloid beta-induced impairments in hippocam-
betic rats. Journal of Diabetes Research, 2013, 593672. pal synaptic plasticity are rescued by decreasing mitochondrial super-
Hirase, H., Qian, L., Bartho, P., & Buzsaki, G. (2004). Calcium dynamics oxide. Journal of Neuroscience, 31, 5589–5595.
of cortical astrocytic networks in vivo. PLoS Biology, 2, E96. Ma, Y., Cao, W., Wang, L., Jiang, J., Nie, H., Wang, B., . . . Ying, W.
Hu, J., Speisky, H., & Cotgreave, I. A. (1995). The inhibitory effects of (2014). Basal CD38/cyclic ADP-ribose-dependent signaling mediates
boldine, glaucine, and probucol on TPA-induced down regulation of ATP release and survival of microglia by modulating connexin 43
gap junction function. Relationships to intracellular peroxides, protein hemichannels. Glia, 62, 943–955.
kinase C translocation, and connexin 43 phosphorylation. Biochemical MacVicar, B. A., & Thompson, R. J. (2010). Non-junction functions of
Pharmacology, 50, 1635–1643. pannexin-1 channels. Trends in Neuroscience, 33, 93–102.
Hu, X., Shi, Q., Zhou, X., He, W., Yi, H., Yin, X., . . . Yan, R. (2007). Trans- Mao, Y., Tonkin, R., Nguyen, S. T., O’Carroll, S. J., Nicholson, L. F., Green,
genic mice overexpressing reticulon 3 develop neuritic abnormalities. C., . . .Gorrie, C. A. (2017). Systemic administration of connexin43
EMBO J, 26, 2755–2767. mimetic peptide improves functional recovery after traumatic spinal
Huang, C., Han, X., Li, X., Lam, E., Peng, W., Lou, N., . . . Takano, T. cord injury in adult rats. Journal of Neurotrauma, 34, 707–719.
(2012). Critical role of connexin 43 in secondary expansion of trau- Mecheri, G., Marie-Cardine, M., Sappey-Marinier, D., Bonmartin, H.,
matic spinal cord injury. Journal of Neuroscience, 32, 3333–3338. Albrand, G., Ferry, G., . . . Courpron, P. (1997). In vivo hippocampal
1624 | YI ET AL.

(31)P NMR metabolites in Alzheimer’s disease and ageing. European Sanchez-Varo, R., Trujillo-Estrada, L., Sanchez-Mejias, E., Torres, M.,
Psychiatry, 12, 140–148. Baglietto-Vargas, D., Moreno-Gonzalez, I., . . . Gutierrez, A. (2012).
Mei, X., Ezan, P., Giaume, C., & Koulakoff, A. (2010). Astroglial connexin Abnormal accumulation of autophagic vesicles correlates with axonal
immunoreactivity is specifically altered at beta-amyloid plaques in and synaptic pathology in young Alzheimer’s mice hippocampus. Acta
beta-amyloid precursor protein/presenilin1 mice. Neuroscience, 171, Neuropathology, 123, 53–70.
92–105. Santanam, N., Penumetcha, M., Speisky, H., & Parthasarathy, S. (2004). A
Meme, W., Calvo, C. F., Froger, N., Ezan, P., Amigou, E., Koulakoff, A., & novel alkaloid antioxidant, Boldine and synthetic antioxidant, reduced
Giaume, C. (2006). Proinflammatory cytokines released from micro- form of RU486, inhibit the oxidation of LDL in vitro and atheroscle-
glia inhibit gap junctions in astrocytes: Potentiation by beta-amyloid. rosis in vivo in LDLR (–/–) mice. Atherosclerosis, 173, 203–210.
FASEB Journal, 20, 494–496. Schalper, K. A., S
anchez, H. A., Lee, S. C., Altenberg, G. A., Nathanson,
Nimmerjahn, A., Kirchhoff, F., Kerr, J. N., & Helmchen, F. (2004). Sulfo- M. H., & S
aez, J. C. (2010). Connexin 43 hemichannels mediate the
rhodamine 101 as a specific marker of astroglia in the neocortex in Ca21 influx induced by extracellular alkalinization. American Journal of
vivo. Nature Methods, 1, 31–37. Physiology & Cell Physiology, 299, C1504–C1515.

O’Brien, P., Carrasco-Pozo, C., & Speisky, H. (2006). Boldine and its anti- Schulz, R., Gorge, P. M., Gorbe, A., Ferdinandy, P., Lampe, P. D., & Ley-
oxidant or health-promoting properties. Chemical Biology & Interac- baert, L. (2015). Connexin 43 is an emerging therapeutic target in
tion, 159, 1–17. ischemia/reperfusion injury, cardioprotection and neuroprotection.
Pharmacological Therapy, 153, 90–106.
O’Carroll, S. J., Alkadhi, M., Nicholson, L. F., & Green, C. R. (2008). Connexin
43 mimetic peptides reduce swelling, astrogliosis, and neuronal cell Shi, Q., Hu, X., Prior, M., & Yan, R. (2009). The occurrence of aging-
death after spinal cord injury. Cell Community & Adhesion, 15, 27–42. dependent reticulon 3 immunoreactive dystrophic neurites decreases
cognitive function. Journal of Neuroscience, 29, 5108–5115.
O’Carroll, S. J., Becker, D. L., Davidson, J. O., Gunn, A. J., Nicholson, L.
F., & Green, C. R. (2013). The use of connexin-based therapeutic Shijie, J., Takeuchi, H., Yawata, I., Harada, Y., Sonobe, Y., Doi, Y., . . .
approaches to target inflammatory diseases. Methods in Molecular Suzumura, A. (2009). Blockade of glutamate release from microglia
Biology, 1037, 519–546. attenuates experimental autoimmune encephalomyelitis in mice.
Tohoku Journal of Experimental Medicine, 217, 87–92.
Orellana, J. A., Froger, N., Ezan, P., Jiang, J. X., Bennett, M. V., Naus, C.
C., Giaume, C., & S
aez, J. C. (2011a). ATP and glutamate released via Sohl, G., Maxeiner, S., & Willecke, K. (2005). Expression and functions of
astroglial connexin 43 hemichannels mediate neuronal death through neuronal gap junctions. Nature Reviews Neuroscience, 6, 191–200.
activation of pannexin 1 hemichannels. Journal of Neurochemistry, Speisky, H., & Cassels, B. K. (1994). Boldo and boldine: An emerging
118, 826–840. case of natural drug development. Pharmacological Research: The Offi-
Orellana, J. A., Montero, T. D., & von Bernhardi, R. (2013). Astrocytes cial Journal of the Italian Pharmacological Society, 29, 1–12.
inhibit nitric oxide-dependent Ca(21) dynamics in activated microglia: Suzuki, H., Ono, K., & Sawada, M. (2014). Protective effect of INI-0602,
Involvement of ATP released via pannexin 1 channels. Glia, 61, a gap junction inhibitor, on dopaminergic neurodegeneration of mice
2023–2037. with unilateral 6-hydroxydopamine injection. Journal Neural Transmis-
Orellana, J. A., Retamal, M. A., Moraga-Amaro, R., & Stehberg, J. (2016). Role sion (Vienna), 121, 1349–1355.
of astroglial hemichannels and pannexons in memory and Takeuchi, H., Jin, S., Wang, J., Zhang, G., Kawanokuchi, J., Kuno, R., . . .
neurodegenerative diseases. Frontiers in Integrative Neuroscience, 10, 26. Suzumura, A. (2006). Tumor necrosis factor-alpha induces neurotoxic-
Orellana, J. A., Shoji, K. F., Abudara, V., Ezan, P., Amigou, E., S
aez, P. J., ity via glutamate release from hemichannels of activated microglia in an
. . . Giaume, C. (2011b). Amyloid beta-induced death in neurons autocrine manner. Journal of Biological Chemistry, 281, 21362–21368.
involves glial and neuronal hemichannels. Journal of Neuroscience, 31, Takeuchi, H., Mizoguchi, H., Doi, Y., Jin, S., Noda, M., Liang, J., . . . Suzu-
4962–4977. mura, A. (2011). Blockade of gap junction hemichannel suppresses
Paul, J., Strickland, S., & Melchor, J. P. (2007). Fibrin deposition disease progression in mouse models of amyotrophic lateral sclerosis
accelerates neurovascular damage and neuroinflammation in mouse and Alzheimer’s disease. PloS One, 6, e21108.
models of Alzheimer’s disease. Journal of Experimental Medicine, 204, Takeuchi, H., & Suzumura, A. (2014). Gap junctions and hemichannels
1999–2008. composed of connexins: Potential therapeutic targets for neurodege-
Retamal, M. A. (2014). Connexin and Pannexin hemichannels are regu- nerative diseases. Frontiers in Cellular Neuroscience, 8, 189.
lated by redox potential. Frontiers in Physiology, 5, 80. Ujiie, M., Dickstein, D. L., Carlow, D. A., & Jefferies, W. A. (2003).
Retamal, M. A., Froger, N., Palacios-Prado, N., Ezan, P., S

aez, P. J., S
aez, Blood–brain barrier permeability precedes senile plaque formation in
J. C., & Giaume, C. (2007). Cx43 hemichannels and gap junction an Alzheimer disease model. Microcirculation, 10, 463–470.
channels in astrocytes are regulated oppositely by proinflammatory Umebayashi, D., Natsume, A., Takeuchi, H., Hara, M., Nishimura, Y.,
cytokines released from activated microglia. Journal of Neuroscience, Fukuyama, R., . . . Wakabayashi, T. (2014). Blockade of gap junction
27, 13781–13792. hemichannel protects secondary spinal cord injury from activated
Rossi, D. (2015). Astrocyte physiopathology: At the crossroads of inter- microglia-mediated glutamate exitoneurotoxicity. Journal of Neuro-
cellular networking., inflammation and cell death. Progression in Neu- trauma, 31, 1967–1974.
robiology, 130, 86–120. Valko, M., Leibfritz, D., Moncol, J., Cronin, M. T., Mazur, M., & Telser, J.
Rouach, N., Koulakoff, A., Abudara, V., Willecke, K., & Giaume, C. (2008). (2007). Free radicals and antioxidants in normal physiological func-
Astroglial metabolic networks sustain hippocampal synaptic transmis- tions and human disease. International Journal of Biochemistry & Cell
sion. Science, 322, 1551–1555. Biology, 39, 44–84.
Sanchez, H. A., Orellana, J. A., Verselis, V. K., & Saez, J. C. (2009). Wang, N., De Bock, M., Antoons, G., Gadicherla, A. K., Bol, M., Decrock,
Metabolic inhibition increases activity of connexin-32 hemichannels E., . . . Leybaert, L. (2012). Connexin mimetic peptides inhibit Cx43
permeable to Ca21 in transfected HeLa cells. American Journal of hemichannel opening triggered by voltage and intracellular Ca21
Physiology & Cell Physiology, 297, C665–C678. elevation. Basic Research in Cardiology, 107, 304.
YI ET AL. | 1625

Xie, H., Hou, S., Jiang, J., Sekutowicz, M., Kelly, J., & Bacskai, B. J. Zlokovic, B. V. (2011). Neurovascular pathways to neurodegeneration in
(2013). Rapid cell death is preceded by amyloid plaque-mediated Alzheimer’s disease and other disorders. Nature Reviews Neuroscience,
oxidative stress. Proceedings of National Academic Sciences USA, 110, 12, 723–738.
7904–7909.
Ye, Z. C., Wyeth, M. S., Baltan-Tekkok, S., & Ransom, B. R. (2003). Func- S UPPORTING INF ORMATION
tional hemichannels in astrocytes: A novel mechanism of glutamate
release. Journal of Neuroscience, 23, 3588–3596. Additional Supporting Information may be found online in the sup-
Yi, C., Mei Ezan, X., Mato, P., Matias, S., Giaume, I. C., & Koulakoff, A. porting information tab for this article.
(2016). Astroglial connexin43 contributes to neuronal suffering in a
mouse model of Alzheimer’s disease. Cell Death and Differentiation
23, 1691–1701.
How to cite this article: Yi C, Ezan P, Fern

andez P, et al. Inhibi-
Zenaro, E., Piacentino, G., & Constantin, G. (2016). The blood–brain bar-
tion of glial hemichannels by boldine treatment reduces neuro-
rier in Alzheimer’s disease. Neurobiology & Disease. doi: 10.1016/j.
nbd.2016.07.007.
nal suffering in a murine model of Alzheimer’s disease. Glia.
2017;65:1607–1625. https://doi.org/10.1002/glia.23182
Zlokovic, B. V. (2008). The blood–brain barrier in health and chronic neu-
rodegenerative disorders. Neuron, 57, 178–201.