PAPER • OPEN ACCESS
Neuroprotective effect of hypoxic preconditioning and neuronal activation
in a in vitro human model of the ischemic penumbra
To cite this article: Sara Pires Monteiro et al 2021 J. Neural Eng. 18 036016
View the article online for updates and enhancements.
Journal of Neural Engineering
OPEN ACCESS RECEIVED 5 November 2020 REVISED 22 January 2021 ACCEPTED FOR PUBLICATION 15 February 2021 PUBLISHED 16 March 2021Original content from this work may be used under the terms of the
Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
PAPER
Neuroprotective effect of hypoxic preconditioning and neuronal
activation in a in vitro human model of the ischemic penumbra
Sara Pires Monteiro1, Eva Voogd1, Lorenzo Muzzi2, Gianmarco De Vecchis1, Britt Mossink3,Marloes Levers1, Gerco Hassink1, Michel Van Putten1, Joost Le Feber1, Jeannette Hofmeijer1,4
and Monica Frega1,3,∗
1 Department of Clinical Neurophysiology, University of Twente, 7522 NB Enschede, The Netherlands
2 Department of Informatics, Bioengineering, Robotics and System Engineering, University of Genoa, 16145 Genoa, Italy
3 Department of Human Genetics, Radboudumc, Donders Institute for Brain, Cognition, and Behaviour, 6500 HB Nijmegen,
The Netherlands
4 Department of Neurology, Rijnstate Hospital, Arnhem, The Netherlands
∗
Author to whom any correspondence should be addressed. E-mail:m.frega@utwente.nl
Keywords: human induced pluripotent stem cells, ischemic penumbra, neuronal networks, optogenetic stimulation Supplementary material for this article is availableonline
Abstract
Objective. In ischemic stroke, treatments to protect neurons from irreversible damage are urgently
needed. Studies in animal models have shown that neuroprotective treatments targeting neuronal
silencing improve brain recovery, but in clinical trials none of these were effective in patients. This
failure of translation poses doubts on the real efficacy of treatments tested and on the validity of
animal models for human stroke. Here, we established a human neuronal model of the ischemic
penumbra by using human induced pluripotent stem cells and we provided an in-depth
characterization of neuronal responses to hypoxia and treatment strategies at the network level.
Approach. We generated neurons from induced pluripotent stem cells derived from healthy donor
and we cultured them on micro-electrode arrays. We measured the electrophysiological activity of
human neuronal networks under controlled hypoxic conditions. We tested the effect of different
treatment strategies on neuronal network functionality. Main results. Human neuronal networks
are vulnerable to hypoxia reflected by a decrease in activity and synchronicity under low oxygen
conditions. We observe that full, partial or absent recovery depend on the timing of re-oxygenation
and we provide a critical time threshold that, if crossed, is associated with irreversible impairments.
We found that hypoxic preconditioning improves resistance to a second hypoxic insult. Finally, in
contrast to previously tested, ineffective treatments, we show that stimulatory treatments
counteracting neuronal silencing during hypoxia, such as optogenetic stimulation, are
neuroprotective. Significance. We presented a human neuronal model of the ischemic penumbra
and we provided insights that may offer the basis for novel therapeutic approaches for patients
after stroke. The use of human neurons might improve drug discovery and translation of findings
to patients and might open new perspectives for personalized investigations.
1. Introduction
Acute ischemic stroke, defined as a sudden neurolo-gical deficit resulting from insufficient cerebral per-fusion, most often due to occlusion of a brain artery, is the leading cause of chronic adult disability and the third leading cause of death in the Western world
[1–4]. To date, the only treatment proven to benefit
functional recovery is acute recanalization by intra-venous thrombolysis or intra-arterial thrombectomy. However, due to strict patient selection, only a minor-ity of patients are eligible for these treatments and even if complete revascularization is achieved,
recov-ery rates vary widely [5,6]. There is an urgent need to
uncover new pathways for treatment, preferably with wide time-windows of opportunity.
The first consequence of cerebral ischemia is loss of neuronal functioning, which is followed either by recovery or deterioration towards irreversible dam-age, depending on the depth (i.e. level of oxygen) and
duration of ischemia [7,8]. The area that surrounds
the core of an infarct, the penumbra, is electrically silent, but structurally intact due to some remaining
blood flow from surrounding arteries [7,9]. This area
has the potential for full recovery if normal perfusion
is restored in time [10]. If oxygen and nutrients are
not timely restored, eventually, irreversible damage
occurs [10,11]. Treatments to improve neuronal
res-cue in the ischemic penumbra hold large potential to advance clinical recovery, but are unavailable.
Over the past two decades, more than 1200 animal studies have been reported providing strong proof of principle that high-grade protection of ischemic
brain tissue is an attainable goal [11]. Many of these
treatments included ion channel blockers, neuro-transmitter receptor antagonists, or suppressors of inflammation. However, in more than 500 clinical tri-als, none of these therapies were effective in patients
[11,12].
Inefficacy of previously tested neuroprotective strategies in stroke patients may be associated with the mechanism of action of treatments under study. Briefly, previously tested treatment strategies were directed at suppression of neuronal activity, in order to minimize energy consumption to preserve basic
cellular function [11]. However, suppression of
activ-ity has been associated with progression towards
irre-versible damage in vitro [13, 14] and in patients
[15]. In line with these observations, we showed that
mild stimulation improved recovery after hypoxia in
rodent neuronal cultures [16]. Secondly, failure of
translation of findings from animal studies to humans may be related to inherent differences between animal
and human neurons [17–21]. Further, there exists
a large heterogeneity in patient groups, where the pathophysiology of recovery or deterioration may
vary, which is not reflected in rodent models [22].
The recent advent of human induced pluripo-tent stem cell (hiPSC) technology with the possibil-ity to differentiate stem cells into neurons has opened the way to study human neuronal diseases in vitro
[18, 23]. The use of human neurons may allow
identification of human-specific effects and even-tually include inter-individual differences of neur-onal responses to hypoxia and neuroprotective treat-ments. To our knowledge, no studies about responses to hypoxia and treatment effects have been per-formed on hiPSCs-derived neurons at the functional level.
Here, we established a human neuronal model of the ischemic penumbra and we provided an in-depth characterization of neuronal responses to hypoxia at the network level. We show that electrophysiological responses to transient hypoxia in vitro are suppression with subsequent restoration of network activity on
timescales similar to patients with cerebral ischemia
[8]. Full, partial and absent recovery depend on
the duration of the hypoxic burden. Specifically, we observe a critical time threshold that, if crossed, is associated with irreversible impairments, even if re-oxygenation was re-established. Furthermore, we show that pre-exposure to transient, short-term hyp-oxia generated neuronal networks that are more res-istant to a second hypoxic event. Finally, we observe that counteracting neuronal silencing is neuropro-tective. Our findings encourage novel investigations of ischemic responses and treatment strategies on human neurons and open new possibilities for per-sonalized medicine.
2. Material and methods
2.1. Human iPSC generation and neuronal differentiation
HiPSCs generated from fibroblast were kindly
provided by Mossink et al [24]. Control line 1 (C1,
30 year-old female) was reprogrammed via epi-somal reprogramming (Coriell Institute for medical research, GM25256). Control line 2 (C2, 51 year-old male) was reprogrammed via a non-integrating Sendai virus (KULSTEM iPSC core facility Leuven, Belgium, KSF-16-025). The genetically modified organism (GMO) approvals under which the lines have been used are IG 19-067 and IG 19-068. Karyo-types of hiPSCs line were verified, hiPSCs lines were tested for pluripotency and genomic integrity based
on single nucleotide polymorphism arrays [25] and
mycoplasma test was performed. HiPSCs were cul-tured on Matrigel (Corning, #356237) in E8 flex (Thermo Fisher Scientific) supplemented with
pur-omycin (0.5 µg ml−1, Sigma Aldrich) and G418
(50 µg ml−1, Sigma Aldrich) at 37 ◦C/5% CO2.
Medium was refreshed every 2–3 d and cells were pas-saged twice per week using an enzyme-free reagent (ReLeSR, Stem Cel Technologies).
GMO approvals under which the lines have been used are IG 19-067 and IG 19-068. Karyotypes of hiPSCs line were verified, hiPSCs lines were tested for pluripotency and genomic integrity based on
single nucleotide polymorphism arrays [25] and
mycoplasma test was performed. HiPSCs were cul-tured on Matrigel (Corning, #356237) in E8 flex (Thermo Fisher Scientific) supplemented with
pur-omycin (0.5 µg ml−1, Sigma Aldrich) and G418
(50 µg ml−1, Sigma Aldrich) at 37 ◦C/5% CO2.
Medium was refreshed every 2–3 d and cells were pas-saged twice per week using an enzyme-free reagent (ReLeSR, Stem Cell Technologies).
Glutamatergic cortical Layer 2/3 neurons were derived from C1 by overexpressing mouse neuronal determinant Neurogenin 2 (Ngn2) upon
doxycyc-line treatment [25, 26]. GABAergic neurons were
derived from C2 by overexpressing mouse neur-onal determinant Achaete-scute homolog 1 (Ascl1,
Addgene 97329) upon doxycycline treatment and
supplementation with Forskolin [24] (10 µM, Sigma
Aldrich). We have chosen to use this pair of hiPSCs lines since we showed that neuronal networks in which excitatory and inhibitory neurons were derived from hiPSCs from different healthy subjects (in different combinations) showed similar functional activity on micro-electrode array (MEA) during
nor-moxia [24]. Ngn2- and Ascl1-iPSCs were plated
together as single cells into a sterile 24-well MEA,
pre-coated with poly-l-ornithine (50 µg ml−1, Sigma
Aldrich) and mouse laminin (20 µg ml−1, Sigma
Aldrich), in an 80:20 ratio to obtain a final cell
density of 2000 cell mm−2, as described previously
[24] (figure1(a)). Two days after plating, astrocytes
obtained from brain cortices of newborn (P1) Wistar rats were added to the hiPSCs culture in a 1:1 ratio, to support neuronal maturation (in agreement to Dutch and European laws and the guidelines of the Dutch Animal Use Committee). The day after the addition of astrocytes, the medium was changed to Neuro-basal medium (Thermo Fisher Scientific) supplemen-ted with Forskolin (10 µM, Sigma Aldrich), B-27 (1:50 Thermo Fisher Scientific), glutaMAX (Thermo
Fisher Scientific), primocin (0.1 µg ml−1, Inivogen),
NT3 (10 ng ml−1, Bioconnect), BDNF (10 ng ml−1,
Bioconnect), and doxycycline (4 µg ml−1, Sigma
Ald-rich). Cytosine β-D-arabinofuranoside (2 µM Sigma Aldrich) was added to remove any proliferating cells. From this day onwards, half of the medium was changed three times a week. The medium was addi-tionally supplemented with 2%, 5% fetal bovine serum (FBS, Sigma Aldrich) to support astrocyte viability from day in vitro (DIV) 10 onwards and For-skolin was removed after DIV 14. Cells were kept in an incubator with a controlled atmosphere
(temper-ature of 37◦C, 100% humidity and 5% of CO2) until
the day of the experiment. 2.2. Experimental protocol
Electrophysiological recordings of neuronal network activity were performed at 7 weeks in vitro. Before the start of each experiment, the 24-well MEA (Mul-tichannel systems) (24 independent wells with 12 embedded microelectrodes (30 µm in diameter and spaced 200 µm apart)) was covered with a Breathe Easier sealing membrane (Sigma Aldrich; Z763624) to reduce evaporation during the long-lasting record-ings, while allowing gases to flow through. Three mass flow controllers were used to establish normoxia and
hypoxia through different mixtures of air and N2
(normoxia: 100% air–0% N2; hypoxia: 10% air–90%
N2) which were conveyed at a flow rate of 0.2 l min−1.
CO2(5%) was added constantly to the gas mixture.
The temperature was set at 37◦C. The level of
oxy-gen in the medium during normoxia and hypoxia was measured with a Neo-Fox-GT optical oxygen sensor (Ocean Optics, Largo, FL, USA), calibrated in
air (pO2 =21.0 per atmospheric pressure) and N2
(pO2 =0). The probe was placed into the medium
and the oxygen level was measured in three wells. The time needed to reach the desired oxygen levels was
1 h (figure1(b)), thus recordings were not performed
during this transition period.
Spontaneous electrophysiological activity of hiPSCs-derived neuronal networks was measured for 10 min during normoxia (baseline phase), after 30 min of stabilization. Afterwards, hypoxia was ini-tiated and, after 1 h of stabilization, 10 min of activity were recorded every 2 h (hypoxia phase). After the hypoxia period (duration from 6 to 48 h), neuronal networks were exposed to normoxia (re-oxygenation phase) and activity was recorded for 10 min after 6 h and 24 h of re-oxygenation. Half of the medium was changed after the first recording of the re-oxygenation phase (6 h).
After 24 h of re-oxygenation, neuronal network pre-exposed to 6, 12, 24 h of hypoxia underwent a second exposure of 6 h, following the same experi-mental protocol.
To promote recovery, three strategies were applied during the entire duration of hypoxia.
2.2.1. Optogenetic stimulation
On the seeding day, Ngn2-hiPSCs were infected with Channelrhodopsin-2 (ChR2) (AAV2-hSyn-hChR2(H134R)-mCherry, UNC Vector Core) by the addition of 0.75 µl of virus per well. After 6 h, cells were washed twice with DMEM/F12 (Thermo Fisher Scientific). Afterwards, Ascl1-hiPSCs were added into the well. Therefore, excitatory neurons were sensitive to blue light (λ = 470 nm). Blue light stimulation was delivered to the cells with the use of a Multiwell-Optogenetic prototype (Multi Channel Systems, Reutlingen, Germany). Light pulses (200 ms, 10 mA, 0.2 Hz) were applied for 3 min every 2 h. The experi-ment was conducted in the dark to avoid light inter-ference. ChR2 integration was successful, as shown by the ChR2 expression and neuronal response to stimu-lation (supplementary figures 1(b) and (c) (available
online atstacks.iop.org/JNE/18/036016/mmedia)).
2.2.2. Electrical stimulation
Electrical stimuli (biphasic voltage pulses starting with the negative part, 250 µs for phase, 0.2 Hz,
1.5 Vpp) were delivered for 3 min every 2 h from one
electrode. The chosen electrode was the one evoking the major neuronal network response during baseline (supplementary figure 1(d)).
2.2.3. Chemical stimulation
Since ghrelin has been associated with improved syn-apse recovery after hypoxia in a rodent in vitro model
[27], we aimed at investigating whether it had a
neuroprotective effect on neuronal network func-tionality. Before the start of the experiment, ghrelin (abcam; Ab73131) was added in the medium at a final
2.3. Micro electrode array recordings and data analysis
Electrophysiological activity was measured by the Multiwell-MEA system using the Multiwell-Screen software (Multi Channel Systems, Reutlingen, Ger-many). Data were acquired at a frequency of 10 kHz and signals were filtered with a high-pass filter (2nd order Butterworth filter, 100 Hz) and a low-pass filter (4th order Butterworth filter, 3.5 kHz).
Data analysis was performed with the use of the Multiwell Analyser software (Multi Channel Systems, Reutlingen, Germany) in combination with custom made MATLAB scripts (The Mathworks, Natick, MA, USA).
2.3.1. Spike detection
Spikes were detected if exceeding 4.5 times the stand-ard deviation of the baseline noise. An electrode was
considered active if exhibiting at least 0.1 spikes s−1.
After detection, the mean firing rate was extracted by computing the number of spikes in time per electrode and averaging it among all the active electrodes of the MEA.
2.3.2. Burst detection
Single-channel bursts were detected when containing a minimum of 4 spikes with an inter spike interval of 50 ms. The minimum interval between bursts was set at 100 ms. A channel is defined as bursting channel if
exhibiting at least 0.4 burst min−1.
2.3.3. Network burst detection
The tool identifies time regions with simultan-eously occurring bursting activity on multiple single-channels. A network burst was defined as a sequence of temporally overlapping single-channel bursts. A minimum of eight distinct bursting channels in the sequence and a minimum of six channels that are bursting at the same time at some point during the sequence should be detected to qualify the sequence as network burst. After detection, parameters describ-ing the network burst have been extracted: network burst rate (number of network bursts detected in time), network burst duration (averaged duration of all network burst detected), spike frequency intra burst (number of spike detected in each network burst divided by the network burst duration and then aver-aged for all network burst detected).
2.3.4. Connectivity analysis
We obtained connectivity matrix for each culture by applying the Filtered Cross-Correlation algorithm
[29] available in the free software SPICODYN [30].
Matrices were then analysed using custom script in MATLAB. We evaluated the total number of links and the average weight of those links for each culture.
The parameters were evaluated during time win-dows in which stimulation was not delivered.
2.3.5. Post stimulus time histogram
The evoked activity was evaluated by computing the post stimulus time histogram (PSTH), through the adaptation of a custom software package (SPYCODE)
[31]. A time window of 300 ms, a bin size of 5 ms and a
blanking period of 4 ms were chosen. The area under the PSTH curve was estimated for each electrode and averaged among all active electrodes.
Neuronal networks showing a firing rate
<0.1 spike s−1, a burst rate <0.4 bursts min−1 or
no network burst during the normoxia phase were excluded from analysis. Furthermore, wells that dis-played insufficient quality, (i.e. low cell density, cell clumping) were discarded. Inclusion criteria ensure that neuronal networks belonging to different groups are comparable (supplementary figure 1(a) and sup-plementary table 1) and viable (i.e. 95% of live cells during normoxia).
2.4. Live/dead assay
For the live/dead assay, Cell Event (1:1000; Thermos-cientific) was added to the cells and incubated for
30 min at 37◦C. After that, the hypoxic period
star-ted for 24 h, 30 h, 40 h and 48 h. At the end of the hypoxic period propidium iodide (PI, 1:500 Invit-rogen) was added for 15 min at room temperature (RT) to stain the dead cells. Cells were washed with phosphate-buffered saline (PBS) and fixated with 4% paraformaldehyde (Sigma Aldrich) for 15 min at RT. Finally, DAPI (1:1000; Sigma Aldrich) was added for 20 min at RT, cells were washed with PBS and moun-ted with mowiol (Sigma Aldrich). Fluorescent pic-tures were taken at a 40× magnification with the use of a Nikon Eclipse 50i epi-fluorescence micro-scope. Cells were counted manually and considered live when only DAPI (blue) was visible, apoptotic when positive for DAPI (blue) and Cell Event (green) and dead when positive for DAPI (blue), Cell Event (green) and PI (red).
2.5. Synaptic puncta
Synaptic puncta were counted manually by staining the cells with MAP2 (1:1000 Sigma Aldrich M3696) and Synapsin (Sigma Aldrich M4403). Cells grown on coverslips were exposed to different durations of hypoxia. After the hypoxic period cells were fixed and stained. Fluorescent pictures were taken at a 60× magnification with the use of a Nikon Eclipse 50i epi-fluorescence microscope. Synapses were counted per 10 µm length of a dendrite.
2.6. Immunocytochemistry
Human neurons were fixed at 7 weeks in vitro with 4% paraformaldehyde (Sigma Aldrich) for 15 min at RT washed with PBS (homemade) and stored at
4 ◦C in PBS until stained. First, samples were
washed with PBS and to block non-specific binding sites cells were blocked with 2% BSA (Sigma Aldrich) in PBS for 30 min at RT. The cultures were stained for rabbit anti-MAP2 (1:1000; Sigma M3696), rab-bit anti-GFAP (1:500; Abcam ab7260) overnight at
4◦C in blocking buffer. Samples were washed with
PBS and stained with secondary antibodies for 1 h at RT, washed again and as a last step the nuclei were stained with DAPI (1:1000; Sigma Aldrich) 20 min at RT. Samples were washed and mounted with mowiol (Sigma Aldrich). The secondary antibodies used were goat anti-mouse Alexa Fluor 488 (1:2000, Invitrogen A-11029) and goat anti-rabbit Alexa Fluor 568 (1:2000, Invitrogen A-11036). Images were taken at a 40× magnification with the use of a Nikon Eclipse 50i Epi-Fluorescence microscope (Nikon, Japan).
2.7. Glucose assay
Culture medium was collected in normoxia and after different durations of hypoxia. It was centrifuged and supernatant was subsequently filtered over an 10 kD cut-off concentrator (Pierce; 88513) to remove proteins and debris from the medium. Hereafter, glucose assay was performed according to protocol of the manufacturer using the colorimetric glucose assay (Abcam; ab65333) and analysed at 570 nm in a Multiskan 60 (Thermo Scientific) plate reader. To make sure all read values were in the range of the calibration curve series dilutions were made from the samples. To correct for concentration effects due to evaporation osmolality of all medium samples was measured in a micro osmometer (Knauer; K-7400S). Values were corrected for osmolality differences.
2.8. Statistical tools
Statistical analysis was performed using GraphPad Prism 5 (GraphPad Software, Inc., CA, USA). We ensured normal distribution using a Kolmogorov– Smirnov normality test. To determine statistical
sig-nificance for the different experimental conditions
p-values <0.05 were considered to be significant. Statist-ical analysis was performed with Kruskall–Wallis and Dunn’s multiple comparisons tests. Every group was
compared to its normalized baseline (i.e. figures1(d)–
(g), (i), (j),2(c)–(f), 3(b)–(e), and S2(b)). Treated
groups were compared to their own baseline and
to not treated cultures (i.e. figures4(d)–(f), S3(b),
(e), (f), and S4(b), (c)). Details about statistics are reported in tables S2–S9. Data are presented as mean ± standard error of the mean. The num-ber of independent neuronal networks included for each experiment are reported in the figure legends. All the values of the parameters throughout the different phases of the experiment were normal-ized with respect to their baseline values to allow comparisons.
3. Results
3.1. Human neuronal networks are vulnerable to hypoxia
We differentiated hiPSCs into populations of excitat-ory and inhibitexcitat-ory neurons, by forced expression of
the transcription factors Ngn2 and Ascl1 [24,26,32]
(figure 1(a)). HiPSC derived neurons, co-cultured
with rodent astrocytes to support functional
mat-uration [26], were grown on MEAs, devices
allow-ing high-throughput long term recordallow-ing of neuronal activity by means of embedded micro-electrodes. As
previously shown [24], in conditions of normoxia
by the seventh week in vitro hiPSC-derived neur-ons grown on MEAs formed a functionally connec-ted network exhibiting electrical activity composed of spikes and bursts appearing in a synchronous manner
(figures1(a) and (c) ‘baseline’). At this time point,
the neuronal network is in a stable state of activity, where the firing and the network burst rates reach a
plateau [24,25]. To ensure consistency between
neur-onal networks, we investigated the effect of hypoxia on neuronal functionality at this stage of maturation
(figures1(a)–(c)). Although glucose deprivation was
not actively applied, we observed a decrease of gluc-ose concentration over time, resulting from glucgluc-ose consumption in the absence of supply (table S10). When exposed to hypoxia, the firing activity and syn-chronicity exhibited by hiPSCs-derived neuronal net-works immediately dropped as compared to baseline,
with different decay rates (figures 1(c)–(e)).
Signi-ficant changes were present already one hour after onset of hypoxia (p < 0.0005 and p < 0.0001, respect-ively). The spike frequency within a network burst reduced as well as a function of time (p < 0.005 after
6 h) (figure 1(f)) and moderate reduction in
net-work burst duration was observed (figure1(g)). All
the neuronal networks completely lose synchronicity within 24 h of hypoxia, and after 48 h neuronal net-works became electrically silent. Neuronal network functional connectivity was impaired during hypoxia as well, reflected by the reduced number and strength
of connections as a function of time (figures1(h)–(j)).
In particular, after 24 h, we observed a 50% decrease in the number of connections (p < 0.005), while after 48 h almost all connections were lost (p < 0.0001) and the strength of remaining connections was signific-antly decreased (p < 0.005). We evaluated cell viabil-ity during hypoxia by counting the percentage of live, apoptotic and dead cells and we found that the num-ber of apoptotic and dead cells increased over time
(figure1(k)). In particular, about 60% of the cells
were apoptotic after 24 h of hypoxia and about 40% of the cells were dead after 48 h of hypoxia.
3.2. Recovery is dependent on the timing of re-oxygenation
Here, we studied the relationship between timing of re-oxygenation and neuronal network recovery, by
Figure 1. Neuronal network activity is impaired during hypoxia in a time dependent manner. (a) Schematic presentation of the differentiation protocol on micro-electrode arrays (MEAs). At day in vitro (DIV) 0, Ngn2- and Ascl1-hiPSCs were plated together as single cells on 24-wells MEAs upon doxycycline treatment with supplementation of Forskolin (DOX and FSK). After 3 d, rodent astrocytes were added in the culture for functional support. At 49 DIV, excitatory and inhibitory neurons are
differentiated. A representative image shows hiPSCs-derived neurons stained for MAP2 (green) and rodent astrocytes stained for GFAP (red) at DIV 49 (cell nuclei are stained with DAPI (blue)). (b) Graph showing the induction of hypoxia: the level of oxygen in the medium is lowered from 21% to 2% and a stable plateau is reached after approximately 1 h (n = 3 wells). (c) Representative raster plots showing 3 min of electrophysiological activity exhibited by neuronal network during baseline and at different time points during exposure to hypoxia (6, 12, 24, 48 h) at DIV 49. (d)–(g) Graphs showing the effect of 48 h of hypoxia on the (d) firing rate (decay constant of exponential fit equal to 0.038), (e) network burst rate (decay constant of exponential fit equal to 0.148), (f) spike frequency during burst and (g) network burst duration. (h) Representative connectivity maps showing the links (red lines) and their strengths (magnitude of the grey circles) during baseline and after 12, 24 and 48 h of hypoxia. (i)–(j) Graph showing the effect of 48 h of hypoxia on neuronal network connectivity, indicated as (i) number and (j) strength of links. The values are normalized to the data of the baseline phase (n = 78 neuronal networks). Mean values of the parameters before normalization are reported in in the insets. (k) Representative images of live/dead staining during normoxia and after 24, 30, 40 and 48 h of hypoxia in hiPSCs-derived neurons stained for DAPI (blue), cell event (green) and PI (red). Scale bar: 20 µm. Bar graph showing the quantification of live cells (positive for DAPI), apoptotic cells (positive for DAPI and cell event) and dead cells (positive for DAPI, cell event and PI) during normoxia and after 24, 30, 40 and 48 h of hypoxia.∗P < 0.05,∗∗P < 0.005, ∗∗∗P < 0.0005,∗∗∗∗P < 0.0001, Kruskall–Wallis and Dunn’s multiple comparisons tests were performed between conditions. In panels (d)–(j), P < 0.0001 are not shown after the first time this value was found. Exact p-values are reported in table S2.
Figure 2. Neuronal network recovery is dependent on the timing of re-oxygenation. (a) Graph showing the experimental protocol, indicating durations of hypoxia (hpx, red tones) and re-oxygenation phase (blue). Second exposure to hypoxia is also shown (hpx2). (b) Representative raster plots showing 1 min of spontaneous activity exhibited by neuronal network after 24 h of re-oxygenation. Networks have been previously exposed to transient hypoxia for 6, 12, 24, 30 and 48 h. (c)–(f) Turkey plots showing the effect of 24 h of re-oxygenation on the (c) firing rate, (d) network burst rate, (e) spike frequency intra burst and (f) network burst duration in neuronal network previously exposed to 6, 12, 24, 30 and 48 h of transient hypoxia (dark red, red, light red, orange and light orange, respectively). The values are normalized to the data of the baseline phase (n = 4, n = 6, n = 28, n = 20, n = 20 neuronal networks for 6, 12, 24, 30 and 48 h of hypoxia, respectively).∗P < 0.05,∗∗P < 0.005,∗∗∗P < 0.0005, ∗∗∗∗P < 0.0001, Kruskall–Wallis and Dunn’s multiple comparisons tests were performed between conditions. Exact p-values are reported in table S3.
exposing neuronal networks to different periods of transient hypoxia (6–48 h) prior to restoration of
nor-moxia (figure 2(a)). After 24 h of re-oxygenation,
neuronal networks that had been exposed to short durations of hypoxia (i.e. 6–12 h) either became
hyperactive or normalized (figures 2(b)–(f)). The
synchronicity of neuronal networks exposed to 24 h of hypoxia remained lower than baseline, as indic-ated by the 50% decrease in the network burst rate
(p < 0.05) (figures2(b) and (d)). Finally, we observed
permanent functional impairments in neuronal net-works exposed to more than 30 h of hypoxia, where synchronous activity remained absent (p < 0.0001) and firing rate was statistically lower than baseline (p < 0.0005 and p < 0.0001 for 30–48 h, respectively)
(figures2(b) and (d)). Neuronal networks exposed to
normoxia for 48 h maintained a stable level of activ-ity (supplementary figure 2(a)). Functional impair-ments were stronger after 6 h than after 24 h of re-oxygenation, indicating gradual recovery over the time course of this phase (supplementary figure 2(b)). 3.3. Hypoxic preconditioning improved resistance to second insult
Next, we investigated the effect of re-oxygenation
on neuronal network connectivity (figures3(a)–(c)).
The increased activity of neuronal networks exposed
to 6 h of hypoxia (figure2(c)), was associated with a
twofold increase in the number of links and a fourfold increase in their strength (p < 0.0001 and p < 0.05,
respectively) (figures 3(b) and (c)). The increased
functional connectivity was not associated with a variation in the number of synaptic puncta (supple-mentary figure 2(c)). Neuronal networks exposed to 12 h of transient hypoxia showed a threefold increase
in connection strength (p < 0.05) (figure3(c)). When
the duration of hypoxia was longer than 12 h, neur-onal network connectivity was impaired as indicated by a statistically lower number of links (p < 0.0001)
(figure3(b)).
Then, we investigated whether the increased func-tional connectivity was associated with higher resist-ance to hypoxia. To this end, we submitted neuronal networks that underwent 6, 12 or 24 h of hypoxia to a second exposure of 6 h (see scheme in supple-mentary figure 3(a)). Cultures that underwent a first exposure of 6 or 12 h of low oxygen showed higher levels of activity and synchronicity during a second hypoxic period as compared to cultures exposed to hypoxia for the first time, unveiling increased
resist-ance (figures3(d) and (e)). Once again, we observed a
time-dependent effect, since neuronal networks pre-exposed to 6 h of hypoxia were more resistant as com-pared to neuronal networks pre-exposed for longer durations.
3.4. Neuronal activation during hypoxia is neuroprotective
We showed that neuronal network exhibiting increased levels of activity after exposure to hypoxia
Figure 3. Hypoxic preconditioning reinforced neuronal network connectivity and resistance to hypoxia. (a) Connectivity maps showing the links (red lines) and their strengths (magnitude of the grey circles) in network exposed to 6, 12, 24, 30, and 48 h of hypoxia after 24 h of re-oxygenation. (b) and (c) Quantification of the (b) number of links and (c) mean strength in networks exposed to 6, 12, 24, 30 and 48 h of hypoxia, calculated at 24 h after re-oxygenation. n = 4, n = 6, n = 28, n = 20, n = 20 neuronal networks for 6, 12, 24, 30 and 48 h of hypoxia, respectively. (d) and (e) Graphs showing the (d) firing rate and (e) network burst rate of neuronal networks exposed to hypoxia either for the first time (dark red bar), or after previous hypoxic preconditioning of 6 h, 12 h or 24 h (white, light grey, grey bars, respectively). n = 78, n = 11, n = 6 and n = 22 neuronal networks for first exposure, second exposure after pre-exposure of 6 h, second exposure after pre-exposure of 12 h and second exposure after pre-exposure of 24 h, respectively. The values are normalized to the data of the baseline phase.∗P < 0.05, ∗∗P < 0.005,∗∗∗P < 0.0005,∗∗∗∗P < 0.0001, Kruskall–Wallis and Dunn’s multiple comparisons tests were performed between conditions. Exact p-values are reported in table S4.
were more resistant to a second insult. Thus, we pos-tulated that increased activity might have protective effects during hypoxia. We tested this hypothesis by artificially activating excitatory neurons in the net-work using optogenetic stimulation. We investigated the effect of optogenetic stimulation at 40 h of hyp-oxia since functional recovery was not present for durations longer than 30 h. When neuronal networks were exposed to 40 h of hypoxia, a consistent decrease in activity was observed in the absence of stimulatory
treatment (figures4(a) and (b), red line). In contrast,
optogenetic stimulation maintained baseline levels of both firing and network bursting activities through
the entire duration of the hypoxia (figures4(a) and
(b), blue line). After 24 h of re-oxygenation, neur-onal networks treated with optogenetic stimulation showed fully recovered levels of activity, while no recovery was observed in untreated cultures (p > 0.05 and p < 0.0001 as compared to baseline level,
respect-ively) (figures 4(c) and (d)). Furthermore,
neur-onal network connectivity was stable during and after hypoxia, indicating that activation of the entire network preserved neuronal functionality by
pre-venting a decrease in activity (figures4(e) and (f)).
The protective effect of stimulation was observed after already 6 h of re-oxygenation (supplementary figure 3(b)).
Next, to investigate whether focal stimulation had a similar neuroprotective effect, we delivered electrical stimuli during hypoxia by one electrode. We observed that the network response to focal, elec-trical stimulation decreased over time, while optogen-etic stimulation led to a constant response through-out the entire duration of hypoxia (supplementary figure 3(c)). Even though neuronal networks treated with focal stimulation maintained higher levels of synchronicity as compared to untreated cultures, net-work bursting ceased after 24 h of hypoxia (supple-mentary figure 3(d)). Functional impairments were still observable after re-oxygenation as compared to baseline (supplementary figures 3(e) and (f)).
Finally, we investigated whether protection of syn-aptic transmission during hypoxia improved neur-onal network recovery. To this end, we treated neuronal network with ghrelin (a hormone and mildly excitatory neurotransmitter associated with improved synapse recovery after hypoxia in a rodent in vitro model [27]) before hypoxia exposure. Over-all, the temporal evolution of activity and synchron-icity in cultures treated with ghrelin remained very similar to untreated networks, indicating no effect of this treatment during hypoxia (supplementary figure 4(a)). However, although functional impair-ments were present after 6 h of re-oxygenation,
Figure 4. Optogenetic stimulation protects neuronal network functionality during hypoxia. (a) and (b) Graph showing the effect of 40 h of hypoxia on neuronal networks not treated (dark red line) or treated with optogenetic stimulation (blue line) during the entire hypoxia period on (a) firing rate and (b) network burst rate. (c) Representative raster plots showing 1 min of activity during baseline and after 40 h of hypoxia followed by 24 h of re-oxygenation in untreated and optogenetically treated neuronal networks. (d) Turkey plots showing the effect of 40 h of hypoxia followed by 24 h of re-oxygenation on the firing rate and network burst rate in networks not treated (dark red) and treated with optogenetic stimulation (blue). (e) Connectivity maps showing the links (red lines) and their strengths (magnitude of the grey circles) in networks not treated and treated with optogenetic stimulation during baseline, after 24–40 h of hypoxia and after 24 h of re-oxygenation. (f) Turkey plots showing the effect of 40 h of hypoxia followed by 24 h of re-oxygenation on the number of links and their strength in networks not treated (dark red) and treated with optogenetic stimulation (blue). The values are normalized to the data of the baseline phase. n = 20, n = 8 neuronal networks not treated and treated with optogenetic stimulation, respectively.∗P < 0.05,∗∗P < 0.005,∗∗∗P < 0.0005,
∗∗∗∗P < 0.0001, Kruskall–Wallis and Dunn’s multiple comparisons tests were performed between conditions. Exact p-values are reported in table S5.
improvements as compared to untreated cultures were visible after 24 h of re-oxygenation (supplement-ary figures 4(b) and (c)).
4. Discussion
In this study, we established a human neuronal model of the ischemic penumbra to (a) investigate neuronal networks response to hypoxia, (b) study the role of re-oxygenation timing on neuronal functioning and (c) test different treatment strategies to support neuronal functionality. We showed that neuronal networks derived from hiPSCs respond to hypoxia with revers-ible and irreversrevers-ible network failure, sequentially. Human neuronal network activity and connectivity decreased as a function of time and full, partial, or absent neuronal recovery depended on the timing of re-oxygenation. We observed a critical time threshold of approximately 24 h: longer durations of hypoxia invariably led to irreversible functional impairments in the absence of neuroprotective treatment. Hypoxic preconditioning made neuronal network resistant to a second hypoxic insult. Optogenetic stimulation during hypoxia had a protective effect, by maintain-ing the neuronal network levels of activity and con-nectivity stable during the entire duration of hypoxia.
4.1. Hypoxia induces reversible and irreversible neuronal network failure
We found that hiPSCs-derived neuronal networks are vulnerable to hypoxia since their level of activ-ity decreased along with the synchronicactiv-ity under low oxygen conditions. After 6 and 12 h of hypoxia, neur-ons remained active and synchronized. However, the frequency of the synchronous events decreased indic-ating possible changes in synaptic functioning, as
observed in vivo [33]. After re-oxygenation, these
neuronal networks showed levels of activity that were higher than baseline, suggesting that homeo-static plasticity occurred, leading to increase cell excitability. The number of links and their strength increased, indicating that neuronal network con-nectivity changed in order to preserve neuronal com-munication. Increased excitability has already been observed in rat models of ischemic stroke and in
patients after stroke [34]. Moreover, it has been
shown that the synaptic density of the inhibitory neurons decreased within the first 12 h of hypoxia leading to an increased excitation/inhibition ratio
[35].
If low oxygen persisted for 24 h, network burst activity ceased, together with severe firing activity and connectivity decreases. This was associated with
a decreased cell viability and an increased number of apoptotic cells. Irreversible neuronal network failure occurred with low oxygen condition lasting longer than 24 h. After this time window, the majority of hypoxic neurons were either apoptotic or dead. This is in line with results of recanalization studies in patients showing efficacy only if applied within 24 h
from the insult [36,37].
We have observed that firing and network burst rates follow first-order exponential decay kinetics
during hypoxia (see figure1). This suggests that their
decays are linked to limited amount of substrates that critically depends on oxygen to be replenished (i.e. glucose, adenosine triphosphate (ATP)). It is also pos-sible that the firing and synchronous activities are related to distinguishable energy supports, hence the difference between their decay kinetics. The steeper decay in the network burst rate shows that hypoxia primarily affects synchronous activity and indicates that synaptic ‘network’ transmission is more expens-ive in terms of energy consumption than spontan-eous spiking activity. Furthermore, given that syn-aptic transmission requires a set of coordinated pre-and post-synaptic ATP-dependent processes, our res-ults suggest that network bursting requires a min-imum or quantal amount of energy to occur. Thus, it might be a good endpoint measure of energy usage in neuronal networks.
4.2. Hypoxic preconditioning increases resistance to second insult
Neuronal networks were more resistant to a second period of low oxygen after hypoxic preconditioning. Studies on rat brains claim that a prior exposure to hypoxia reduces ischemic vulnerability by a decrease in the distance between capillaries as a consequence
of angiogenesis [38–40]. Although this process is
absent in our in vitro model, we found an increased resistance to a second hypoxic insult in neuronal networks after a pre exposure. Hypoxic precondition-ing improved functional connectivity among neur-ons whose mechanisms need to be better investig-ated. Thus, our results suggest that reduced neuronal vulnerability involves mechanisms other than or in addition to angiogenesis and may open new perspect-ives for treatment. Oxygen-dependent signalling cas-cades that regulate cell-specific response to hypoxia and activate neuroprotective systems (i.e. hypoxia-inducible transcription factors and their specific tar-get genes), warrant further exploration for future
dia-gnostic and therapeutic options [41].
4.3. Optogenetic stimulation preserves neuronal functionality
We employed treatment strategies counteracting pro-longed neuronal silencing. Optogenetic stimulation of excitatory neurons helped networks maintaining physiological activity levels throughout the entire hypoxic period, while strong neuronal impairments
were observed in untreated cultures. This suggests that a certain amount of neuronal activity is prob-ably crucial for neuronal viability and that early stim-ulatory treatment may be beneficial, even during tem-porary low oxygen conditions such as present in the
ischemic penumbra [42, 43]. However, prolonged
increased activity levels in neuronal cultures might be a marker of detrimental effects and thus should be carefully interpreted. Under ordinary conditions, the lactate shuttle that occurs mostly between astrocytes and neurons is a primary form of energy
produc-tion in the brain [44–46]. Furthermore, it has been
found that lactate administered at early time points after ischemia can protect against cell death, decrease lesion size, and improve neurologic outcome in vitro
and in vivo [47]. Since the lactate shuttle is
neuronal-activity dependent [44–46], the lack of neuronal
activity occurring during hypoxia in untreated cul-tures would prevent it. Thus, our results suggest that optogenetic stimulation might preferentially main-tain or trigger the shuttle during hypoxia, by
activ-ating signalling molecules (i.e. glutamate, NH4[48])
and consequently having a neuroprotective effect. Electrical stimulation was less effective than opto-genetic stimulation, suggesting that a global neur-onal activation is more effective than focal activation. The low effectiveness of electrical stimulation might be caused by impaired propagation of the electrical impulse through the entire network as a result of
hyp-oxic synaptic failure [49–51].
Treatment with ghrelin improved network recov-ery at 24 h of re-oxygenation, but it was not effect-ive in maintaining stable levels of activity during hyp-oxia. These findings are in line with previous studies showing beneficial effects of ghrelin on synaptic
dens-ity after hypoxia [27] and suggest other effective
path-way than neuronal stimulation, as well.
Together, our results suggest that treatment strategies directed at early neuronal stimulation, for instance before and during application of recanaliz-ation therapies, may enhance recovery after ischemic stroke. A non-invasive and safe way to stimulate and modulate neuronal activity in patients is by trans-cranial magnetic stimulation or transtrans-cranial
direct-current stimulation [52]. Further investigation using
electromagnetic field stimulation on hiPSC-derived neuronal networks can be performed to enhance the translation of our findings. This might implicate that strategies based on non-invasive brain stimulation comprise a novel, high potential treatment approach for patients after stroke.
4.4. Human iPSC may help explaining inter-individual variation
In our model of human cultured neurons, loss of neuronal activity during hypoxia occurred slower
than in previous studies with rodent cells [51].
The observed difference in response between species might affect drug testing and translation of findings
to patients. Intrinsic differences between rodent and human neurons have already been observed. In par-ticular, functional divergences have been found, such as different rates of network maturation, levels of
activity and response to compounds [18, 20, 21,
53]. In addition, although many neuronal genes are
highly conserved across species, it has been shown that human neurons rely on human-specific gene
expression patterns and cellular mechanisms [17,
19]. Therefore, moving beyond the reliance on
non-human sources of neuronal cells towards the use of human-cell based systems might improve drug discovery and translation. However, hiPSCs derived neuronal models are subjected to variations mainly driven by differences in genetic background between
hiPSCs donors [54] and this might possibly
inter-fere with the extent of neuronal response to hypoxia, re-oxygenation and treatment strategies. Therefore, further investigation on hiPSCs derived from differ-ent healthy donors should be performed to under-stand whether genetic background influences neur-onal vulnerability to hypoxia. The use of hiPSCs might open new perspectives for investigations of differences between patients with regard to vulner-ability to ischemia. Indeed, variation in recovery amongst patients with ischemic stroke cannot be fully explained by variation in infarct size, infarct location,
or levels of reperfusion [55,56]. If genetic
constitu-tions contribute to variaconstitu-tions in for example synaptic failure, cortical spreading depression, excitotoxicity,
or formation of heat shock proteins [8,10,57,58],
our hiPSCs based model opens new perspectives for understanding the variability of neuronal vulnerabil-ity to ischemia and recovery of patients with ischemic stroke.
5. Conclusion
Combining hiPSCs derived human neuronal models with neuronal network dynamics constitutes a prom-ising tool to investigate human neuronal responses to hypoxia. Hypoxic preconditioning and early stim-ulatory treatment hold potential to modify success-ive reversible and irreversible neuronal damage and improve recovery. The use of hiPSCs opens new perspectives for understanding variation of recovery amongst patients with ischemic stroke.
Acknowledgments
We thank Dr N N Kasri for providing the hiPSCs used in this study and for his valuable comments on this work.
Author contribution statement
M F conceived and supervised the study. S P M and M F designed all the experiments. S P M, E V, G D V, M L, G H and M F performed all experiments. S P M,
E V, L M, G D V, B M and M F performed data ana-lysis. J H, M V P and M F provided resources, concep-tualization, and intellectual content. S P M and M F wrote the paper. B M, J H, J L F and M V P edited the paper.
ORCID iDs
Sara Pires Monteiro
https://orcid.org/0000-0002-2530-5535
Lorenzo Muzzi
https://orcid.org/0000-0003-2893-6057
Britt Mossink
https://orcid.org/0000-0002-4393-5506
Joost Le Feber
https://orcid.org/0000-0002-0605-1437
Monica Frega
https://orcid.org/0000-0002-9697-3282
References
[1] Bush C K, Kurimella D, Cross L J S, Conner K R, Martin-Schild S, He J, Li C, Chen J and Kelly T 2016 Endovascular treatment with stent-retriever devices for acute ischemic stroke: a meta-analysis of randomized controlled trials PLoS One11 e0147287
[2] Holloway R G et al 2005 Prognosis and decision making in severe stroke JAMA294 725–33
[3] Kelly A G, Hoskins K D and Holloway R G 2012 Early stroke mortality, patient preferences, and the withdrawal of care bias Neurology79 941–4
[4] Lozano R et al 2012 Global and regional mortality from 235 causes of death for 20 age groups in 1990 and 2010: a systematic analysis for the Global Burden of Disease Study 2010 Lancet380 2095–128
[5] Berkhemer O A et al 2015 A randomized trial of intraarterial treatment for acute ischemic stroke New Engl. J. Med. 372 11–20
[6] Hacke W et al 2008 Thrombolysis with alteplase 3 to 4.5 hours after acute ischemic stroke New Engl. J. Med. 359 1317–29
[7] Astrup J, Siesjo B K and Symon L 1981 Thresholds in cerebral ischemia—the ischemic penumbra Stroke 12 723–5
[8] Hofmeijer J and Van Putten M J 2012 Ischemic cerebral damage: an appraisal of synaptic failureStroke 43 607–15
[9] Symon L, Branston N M, Strong A J and Hope T D 1977 The concepts of thresholds of ischaemia in relation to brain structure and function J. Clin. Pathol. Suppl. (R. Coll. Pathol.)11 149–54
[10] Del Zoppo G J, Sharp F R, Heiss W-D and Albers G W 2011 Heterogeneity in the penumbra J. Cereb. Blood Flow Metab. 31 1836–51
[11] Ginsberg M D 2008 Neuroprotection for ischemic stroke: past, present and future Neuropharmacology55 363–89 [12] Nielsen N et al 2013 Targeted temperature management at
33 degrees C versus 36 degrees C after cardiac arrest New Engl. J. Med.369 2197–206
[13] Ghosh A, Carnahan J and Greenberg M E 1994 Requirement for BDNF in activity-dependent survival of cortical neurons Science263 1618–23
[14] Mao Z et al 1999 Neuronal activity-dependent cell survival mediated by transcription factor MEF2 Science 286 785–90
[15] Ruijter B J et al 2019 Early electroencephalography for outcome prediction of postanoxic coma: a prospective cohort study Ann. Neurol.86 203–14
[16] Muzzi L, Hassink G, Levers M, Jansman M, Frega M, Hofmeijer J, Van Putten M and Le Feber J 2019 Mild stimulation improves neuronal survival in an in vitro model of the ischemic penumbra J. Neural. Eng.17 016001 [17] Hawrylycz M et al 2015 Canonical genetic signatures of the
adult human brain Nat. Neurosci.18 1832–44
[18] Keller J M and Frega M 2019 Past, present, and future of neuronal models in vitro Adv. Neurobiol.22 3–17 [19] Khaitovich P et al 2004 Regional patterns of gene
expression in human and chimpanzee brains Genome Res. 14 1462–73
[20] Tukker A M, Wijnolts F M J, De Groot A and Westerink R H S 2020 Applicability of hiPSC-derived neuronal co-cultures and rodent primary cortical cultures for in vitro seizure liability assessment Toxicol. Sci.178 71–87 [21] Hyvarinen T, Hyysalo A, Kapucu F E, Aarnos L,
Vinogradov A, Eglen S J, Ylä-Outinen L and Narkilahti S 2019 Functional characterization of human pluripotent stem cell-derived cortical networks differentiated on laminin-521 substrate: comparison to rat cortical cultures Sci. Rep. 9 17125
[22] Collaborators G B D N 2019 Global, regional, and national burden of neurological disorders, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016 Lancet Neurol.18 459–80
[23] Juntunen M, Hagman S, Moisan A, Narkilahti S and Miettinen S 2020 In vitro oxygen-glucose
deprivation-induced stroke models with human neuroblastoma cell- and induced Pluripotent stem cell-derived neurons Stem Cells Int.2020 8841026 [24] Mossink B et al 2020 Cadherin-13 is a critical regulator of
GABAergic modulation in human stem cell derived neuronal networks Biorxiv (https://doi.org/10.1101/
2020.05.07.082453)
[25] Frega M et al 2019 Neuronal network dysfunction in a model for Kleefstra syndrome mediated by enhanced NMDAR signaling Nat. Commun.10 4928
[26] Frega M, Van Gestel S H C, Linda K, Van Der Raadt J, Keller J, Van Rhijn J-R, Schubert D, Albers C A and Nadif Kasri N 2017 Rapid neuronal differentiation of induced pluripotent stem cells for measuring network activity on micro-electrode arrays J. Vis. Exp.119 e54900 [27] Stoyanova I I, Hofmeijer J, Van Putten M J A M and
Le Feber J 2016 Acyl ghrelin improves synapse recovery in an in vitro model of postanoxic encephalopathy Mol. Neurobiol. 53 6136–43
[28] Stoyanova I I and Le Feber J 2014 Ghrelin accelerates synapse formation and activity development in cultured cortical networks BMC Neurosci.15 49
[29] Pastore V P, Massobrio P, Godjoski A and Martinoia S 2018 Identification of excitatory-inhibitory links and network topology in large-scale neuronal assemblies from multi-electrode recordings PLoS Comput. Biol. 14 e1006381
[30] Pastore V P, Godjoski A, Martinoia S and Massobrio P 2018 SPICODYN: a toolbox for the analysis of neuronal network dynamics and connectivity from multi-site spike signal recordings Neuroinformatics16 15–30
[31] Bologna L L, Pasquale V, Garofalo M, Gandolfo M, Baljon P L, Maccione A, Martinoia S and Chiappalone M 2010 Investigating neuronal activity by SPYCODE multi-channel data analyzer Neural Netw.23 685–97 [32] Zhang Y et al 2013 Rapid single-step induction of functional
neurons from human pluripotent stem cells Neuron 78 785–98
[33] Bolay H, Gürsoy-özdemir Y, Sara Y, Onur R, Can A and Dalkara T 2002 Persistent defect in transmitter release and synapsin phosphorylation in cerebral cortex after transient moderate ischemic injury Stroke 33 1369–75
[34] Schiene K, Bruehl C, Zilles K, Qu M, Hagemann G, Kraemer M and Witte O W 1996 Neuronal hyperexcitability and reduction of GABAA-receptor expression in the surround of
cerebral photothrombosis J. Cereb. Blood Flow Metab. 16 906–14
[35] Le Feber J, Dummer A, Hassink G C, Van Putten M J A M and Hofmeijer J 2018 Evolution of excitation-inhibition ratio in cortical cultures exposed to hypoxia Front. Cell. Neurosci.12 183
[36] Sairanen T, Karjalainen-Lindsberg M-L, Paetau A, Ijäs P and Lindsberg P J 2006 Apoptosis dominant in the periinfarct area of human ischaemic stroke—a possible target of antiapoptotic treatments Brain129 189–99
[37] Sairanen T, Szepesi R, Karjalainen-Lindsberg M-L, Saksi J, Paetau A and Lindsberg P J 2009 Neuronal caspase-3 and PARP-1 correlate differentially with apoptosis and necrosis in ischemic human stroke Acta Neuropathol. 118 541–52
[38] LaManna J C, Chavez J C and Pichiule P 2004 Structural and functional adaptation to hypoxia in the rat brain J. Exp. Biol. 207 3163–9
[39] LaManna J C, Vendel L M and Farrell R M 1992 Brain adaptation to chronic hypobaric hypoxia in rats J. Appl. Physiol.72 2238–43
[40] Schurr A, Reid K H, Tseng M T, West C and Rigor B M 1986 Adaptation of adult brain tissue to anoxia and hypoxia in vitro Brain Res.374 244–8
[41] Trollmann R and Gassmann M 2009 The role of hypoxia-inducible transcription factors in the hypoxic neonatal brain Brain Dev.31 503–9
[42] Furlan M, Marchal G, Derlon J-M, Baron J-C and Viader F 1996 Spontaneous neurological recovery after stroke and the fate of the ischemic penumbra Ann. Neurol.40 216–26 [43] Ramos-Cabrer P, Campos F, Sobrino T and Castillo J 2011
Targeting the ischemic penumbra Stroke42 S7–S11 [44] Barros L F and Weber B 2018 CrossTalk proposal: an
important astrocyte-to-neuron lactate shuttle couples neuronal activity to glucose utilisation in the brain J. Physiol. 596 347–50
[45] Pellerin L and Magistretti P J 2003 Food for thought: challenging the dogmas J. Cereb. Blood Flow Metab. 23 1282–6
[46] Wyss M T, Jolivet R, Buck A, Magistretti P J and Weber B 2011 In vivo evidence for lactate as a neuronal energy source J. Neurosci.31 7477–85
[47] Berthet C, Lei H, Thevenet J, Gruetter R, Magistretti P J and Hirt L 2009 Neuroprotective role of lactate after cerebral ischemia J. Cereb. Blood Flow Metab.29 1780–9 [48] Lerchundi R et al 2015 NH4(+) triggers the release of
astrocytic lactate via mitochondrial pyruvate shunting Proc. Natl Acad. Sci. USA112 11090–5
[49] Hofmeijer J, Mulder A T B, Farinha A C, Van Putten M J A M and Le Feber J 2014 Mild hypoxia affects synaptic connectivity in cultured neuronal networks Brain Res. 1557 180–9
[50] Le Feber J, Erkamp N, Van Putten M J A M and Hofmeijer J 2017 Loss and recovery of functional connectivity in cultured cortical networks exposed to hypoxia J. Neurophysiol.118 394–403
[51] Le Feber J, Tzafi Pavlidou S, Erkamp N, Van Putten M J A M and Hofmeijer J 2016 Progression of neuronal damage in an in vitro model of the ischemic penumbra PLoS One 11 e0147231
[52] Rossini P M et al 2015 Non-invasive electrical and magnetic stimulation of the brain, spinal cord, roots and peripheral nerves: basic principles and procedures for routine clinical and research application. An updated report from an I.F.C.N. Committee Clin. Neurophysiol.126 1071–107 [53] Napoli A and Obeid I 2016 Comparative analysis of human
and rodent brain primary neuronal culture spontaneous activity using micro-electrode array technology J. Cell. Biochem.117 559–65
[54] Germain P L and Testa G 2017 Taming human genetic variability: transcriptomic meta-analysis guides the experimental design and interpretation of iPSC-based disease modeling Stem Cell Rep.8 1784–96
[55] Hofmeijer J, Algra A, Kappelle L J and Van Der Worp H B 2008 Predictors of life-threatening brain edema in middle cerebral artery infarction Cerebrovasc. Dis.25 176–84 [56] Jansen I G H, Mulder M J H L and Goldhoorn R J B 2018
Endovascular treatment for acute ischaemic stroke in routine clinical practice: prospective, observational cohort study (MR CLEAN Registry) Br. Med. J.360 k949
[57] Lauritzen M 1994 Pathophysiology of the migraine aura. The spreading depression theory Brain117 199–210
[58] Wang Y, Denisova J V, Kang K S, Fontes J D, Zhu B T and Belousov A B 2010 Neuronal gap junctions are required for NMDA receptor-mediated excitotoxicity: implications in ischemic stroke J. Neurophysiol. 104 3551–6
