Neuronal Responses to Ischemia: Scoping Review of Insights from Human-Derived In Vitro Models

Translation of neuroprotective treatment effects from experimental animal models to patients with cerebral ischemia has been challenging. Since pathophysiological processes may vary across species, an experimental model to clarify human-specific neuronal pathomechanisms may help. We conducted a scoping review of the literature on human neuronal in vitro models that have been used to study neuronal responses to ischemia or hypoxia, the parts of the pathophysiological cascade that have been investigated in those models, and evidence on effects of interventions. We included 147 studies on four different human neuronal models. The majority of the studies (132/147) was conducted in SH-SY5Y cells, which is a cancerous cell line derived from a single neuroblastoma patient. Of these, 119/132 used undifferentiated SH-SY5Y cells, that lack many neuronal characteristics. Two studies used healthy human induced pluripotent stem cell derived neuronal networks. Most studies used microscopic measures and established hypoxia induced cell death, oxidative stress, or inflammation. Only one study investigated the effect of hypoxia on neuronal network functionality using micro-electrode arrays. Treatment targets included oxidative stress, inflammation, cell death, and neuronal network stimulation. We discuss (dis)advantages of the various model systems and propose future perspectives for research into human neuronal responses to ischemia or hypoxia. Graphical Abstract Supplementary Information The online version contains supplementary material available at 10.1007/s10571-023-01368-y.


Introduction
It has been challenging to translate neuroprotective treatment effects from experimental animal models to patients with cerebral ischemia. While numerous rodent studies have shown efficacy of divergent "neuroprotective" treatment strategies, in clinical trials these treatments have failed to improve functional outcomes (Ginsberg 2008). Failure of treatments in clinical trials have been discussed extensively, with possible explanations ranging from methodological weaknesses of animal studies to differences in treatment protocols and from infarct sizes to patient selection (Ginsberg 2008). However, genetic differences between rodents and humans that may translate to neurophysiological differences are relatively undiscussed. While rodents and humans share part of their DNA, various genes and proteins are expressed differently (Mestas and Hughes 2004). Therefore, cellular responses to cerebral ischemia or hypoxia may differ between rodents and humans. Furthermore, mechanisms of action or pathways that are targeted with neuroprotective treatments can vary (Ginsberg 2008). Consequently, rodent models might not be the optimal starting point to investigate the effect of cerebral ischemia on neuronal functionality and treatment strategies.
Various experimental studies addressed responses to ischemia or hypoxia in human-derived cell models. These consist of non-neuronal and neuronal models. The most used human in vitro model to study neuronal responses to ischemia or hypoxia is the neuroblastoma-derived SH-SY5Y cell model ). This cell model is based on a cancerous cell line with the corresponding genetic characteristics, which may affect responses to ischemia or hypoxia and treatment strategies (Biedler et al. 1973(Biedler et al. , 1978. SH-SY5Y cells can, in principle, be differentiated into neuronlike cells upon stimulation (Shipley et al. 2016). However, the majority of research with SH-SY5Y cell models is conducted in undifferentiated SH-SY5Y cells. Only a small proportion of the research made use of protocols to differentiate SH-SY5Y cells into neuron-like cells. Furthermore, most studies focused on cell viability, neglecting neuronal functionality .
The recent advancement of human induced pluripotent stem cell (hiPSC) technology has created new opportunities to establish human neuronal in vitro models and investigate responses to ischemia or hypoxia. HiPSCs can be derived from healthy donors or patients, capturing person-specific genetic characteristics, and differentiated into neurons to generate neuronal networks (Frega et al. 2017;Mossink et al. 2021a). This allows investigation of the effect of, for example, a genetic mutation on neuronal functionality (Mossink et al. 2021b).
With this scoping review of the literature, we aim to provide an overview of (1) human neuronal in vitro models that have been used to study neuronal responses to ischemia or hypoxia, including characteristics of the various models, (2) parts of the ischemic pathophysiological cascade that have been investigated and (3) treatment targets that have been established in those human neuronal models. We will use the results as a starting point to discuss advantages and disadvantages of the various model systems, highlight current knowledge gaps, and propose possible future perspectives for research into human neuronal responses to ischemia or hypoxia.

Methods
For this scoping review, we followed the PRISMA guidelines with regard to literature search, data collection, presentation of study characteristics and results, and discussion. However, we did not include a quality appraisal or risk of bias analysis.
To review investigated human neuronal in vitro models of cerebral ischemia or hypoxia, we applied a search in PubMed and SCOPUS databases until November 2021. We conducted the literature research with several combinations of key words and MeSH terms. We searched the literature with general terms for "human neuronal in vitro models" and the following specific search terms: "NT2-N", "SK-N-SH", "SH-SY5Y" or "hiPSC-derived neurons" . For selection of disease we used the MeSH terms "ischemic stroke" or "cerebral ischemia" and search terms "hypoxia" or "oxygen-glucose deprivation (OGD)" in combination with the different human neuronal in vitro models. One reviewer (EV) screened articles for eligibility based on the abstracts and methods. Review articles were used to screen reference lists. We only included studies with modeling of cerebral ischemia by oxygen-glucose deprivation or hypoxia. Studies were excluded when chemical simulation of ischemia was investigated. Flow chart is provided in the supplementary materials. For analyses of effects of potential neuroprotective treatments, we included studies on treatments that were applied during or after ischemia/ hypoxia, and excluded studies that investigated the effect of pre-treatment (i.e. a treatment started before initiation of hypoxia or ischemia). We extracted the cell type, the level of differentiation towards neurons (if applicable), the way in which ischemia or hypoxia was modelled (including duration and recovery period), treatment strategy (if applicable), outcome measures, and results from the included studies. All results of the scoping review are presented in a descriptive way.

Results
We included 147 papers on four different human neuronal in vitro models of cerebral ischemia/hypoxia.

Model Characteristics
Model characteristics are summarized Fig. 1 and Table 1. Of the 147 included papers, 145 were on cell models derived from unhealthy donors (neuroblastoma SK-N-SH (n = 11) and SH-SY5Y (n = 132)) or carcinoma (differentiated NT2-N (n = 2)). SK-N-SH is a cell model directly derived from a neuroblastoma cell line. SH-SY5Y cell model is based on a subclone derived from the SK-N-SH cell line. These can both be differentiated into neuron-like cells by stimulation with retinoic acid (RA). However, the majority of studies on neuronal response to ischemia/hypoxia is conducted in undifferentiated cell models (SK-N-SH undifferentiated n = 9 and differentiated n = 2; SH-SY5Y undifferentiated n = 119, and differentiated n = 13). Two studies were conducted on neurons differentiated from hiPSCs of healthy donors Pires Monteiro et al. 2021). Examples of cultured undifferentiated SH-SY5Y and hiPSC-derived neuronal networks are presented in Fig. 2, showing that undifferentiated SH-SY5Y cells have short truncated processes and do not express the neuronal markers microtubule-associated protein 2 (MAP2) and Synapsin 1/2 positive synaptic puncta. While hiPSC-derived neuronal Fig. 1 Pie chart representing the distribution of human neuronal in vitro models used in experimental cerebral ischemia studies. The majority of the research is conducted in SH-SY5Y cell models. hiPSCs human induced pluripotent stem cells. Differentiated NT2-N = teratocarcinoma-derived Ntera2/D1 neuron-like cells networks show long neurites and express MAP2 and Synapsin 1/2 positive synaptic puncta. Cerebral ischemia was modeled by oxygen-glucose deprivation (OGD; n = 142) or hypoxia (n = 5). All studies used immunocytochemical techniques to investigate cell survival, protein expression linked with cell death or inflammation, or factors related to oxidative stress. Neuronal functionality was assessed in only one study in which electrophysiological measurements were performed with micro-electrode arrays (MEAs) (Pires Monteiro et al. 2021).

Effects of Ischemia or Hypoxia on Human Cell Models
The pathophysiological cascade that follows upon cerebral ischemia/hypoxia is extensive and complex. In SH-SY5Y, SK-N-SH, NT2-N and hiPSC-derived cell models, several steps of this pathophysiological cascade have been investigated. These steps are summarized in Table 2 and explained below.

SH-SY5Y and SK-N-SH Cell Models
Calcium Homeostasis Dysregulation Ischemia induced membrane depolarization leads to opening of calcium channels, ultimately resulting in an increase in intracellular Ca 2+ . With insufficient ATP production, Ca 2+ extruders (e.g. Na + /Ca 2+ exchanger) stop working, causing a pathological increase of the intracellular and mitochondrial Ca 2+ concentration. The Na + /Ca 2+ exchanger 1 (NCX1) has been investigated in one study using undifferentiated SH-SY5Y cells. The results showed that NCX1 was repressed during OGD by the RE1-silencing transcription factor (REST) ). In undifferentiated SK-N-SH cells increased levels of intracellular calcium was found during OGD (n = 1) (Lehane et al. 2013).
Oxidative/Nitrosative Stress Oxidative stress results from imbalance between pro-and anti-oxidants, that in turn results in excessive formation of reactive oxygen species (ROS). Various negative effects of ROS on DNA, proteins or lipids have been established in SH-SY5Y and SK-N-SH cell models of cerebral ischemia.
The first ROS that is produced during ischemia/hypoxia is superoxide anion (O 2 − ), which is the precursor of most other ROS. One study in undifferentiated SH-SY5Y cells found an increase in O 2 − (Marmol et al. 2021). Superoxide dismutases (SODs) convert superoxide to hydrogen peroxide (H 2 O 2 ), which can be removed by glutathione peroxidase (GPx). GPx was decreased after ischemia/hypoxia in undifferentiated SH-SY5Y and SK-N-SH cells Wang and Xu 2021). SODs prevent the formation of highly aggressive ROS. Ten studies in undifferentiated SH-SY5Y cells and one in undifferentiated SK-N-SH cells found decreased levels of SODs during ischemia/hypoxia (Di et al. 2021;Dong et al. 2021;Feng et al. 2021;Gao et al. 2017Gao et al. , 2020Liu et al. 2019b;Wang et al. 2019bWang et al. , 2020aWang and Xu 2021;Wen et al. 2020;Xu et al. 2020b).
Glutathione (GSH) is an antioxidant that can prevent damage to cellular components by ROS. The ratio of reduced GSH to oxidized glutathione (GSSG) is a measure of cellular oxidative stress. This ratio has been investigated in one study using undifferentiated SH-SY5Y cells (Guo et al. 2009). The results showed a decreased ratio by increased oxidized glutathione. This study also investigated the level of cellular nicotinamide adenine Reference list for this table is provided in the supplementary materials OGD oxygen glucose deprivation; LDH lactate dehydrogenase; MTT 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; CCK-8 cell counting kit 8; IF immunofluorescence; RT-PCR reverse transcription polymerase chain reaction; ELISA enzyme-linked immunosorbent assay; CELISA cellular ELISA; ROS reactive oxygen species; PTEN phosphatase and tensin homolog; ATP adenosine triphosphate dinucleotide phosphate (NADPH) content. NADPH provides oxidation-reduction which protects against ROS allowing the regeneration of GSH. There was no difference in NADPH content between normoxia or hypoxia treated undifferentiated SH-SY5Y cells (Guo et al. 2009). NADPH oxidase was involved in post-ischemic cell death in differentiated SH-SY5Y cells (n = 1) (Beske and Jackson 2012). One other study investigated the level of GSH in undifferentiated SH-SY5Y cells and found that there was a decrease in GSH during hypoxia, leading to increased cellular oxidative stress (Wang et al. 2018a). Another anti-oxidant that has been investigated in undifferentiated SH-SY5Y cells is catalase. This enzyme was reduced after ischemia/hypoxia .
Besides anti-oxidants, free radical scavengers were investigated in an undifferentiated SH-SY5Y cell model of cerebral ischemia, revealing that spermine, spermidine and other polyamines were elevated (Shin et al. 2016). ROS can also be decreased by hypoxia inducible factor 1 (HIF-1). In undifferentiated SH-SY5Y cells it was found that HIF-1 regulated redox status after hypoxic exposure (Guo et al. 2009).
ROS regulates cell ferroptosis by multiple signaling pathways like the Janus kinase and signal transducer and activator of transcription (JAK/STAT). JAK2 has been investigated in undifferentiated SH-SY5Y cells in OGD. The results showed increased expression of JAK2 and STAT3 (n = 2) Xu et al. 2020b).
The transforming growth factor (TGF)/SMAD pathway has been implicated in oxidative stress. TGF expression was increased in an undifferentiated SH-SY5Y cell model of cerebral ischemia, together with increased phosphorylation of SMAD2 and SMAD3. Ischemia/hypoxia promoted translocation of SMAD3 from the cytoplasm to the nucleus (Yang et al. 2021b).
Oxidative stress can cause DNA damage. A multifunctional protein that is involved in the base-excision repair of oxidative DNA damage is Apurinic/apyrimidinic endonuclease 1 (APE1). The expression level of APE1 decreased during ischemia/hypoxia in undifferentiated SH-SY5Y cells (n = 1) ).
Mitochondrial Dysfunction Neuronal ischemia/hypoxia causes mitochondrial dysfunction along several ways: oxidative stress with mitochondrial DNA mutations, dysfunction of the mitochondrial respiratory chain, altered membrane permeability with Ca 2+ accumulation, and disturbance of mitochondrial defense systems (Ham and Raju 2017).
One study in undifferentiated SH-SY5Y cells showed that the mitochondrial permeability transition pore is opened with OGD by translocation of P53 to mitochondria and activation of Cyclophilin D ). The consequent mitochondrial membrane depolarization has been found in undifferentiated SH-SY5Y cells (n = 3) (Agudo-Lopez et al. 2010;Park et al. 2017;Wang et al. 2018b). Mitochondrial membrane permeabilization by opening of the mitochondrial permeability transition pore and loss of the mitochondrial membrane potential (MMP) has also been established in three studies in undifferentiated and in two studies differentiated SH-SY5Y cells (Lin et al. 2020;Liu et al. 2019a;Park et al. 2017;Sriwastva et al. 2020;Wang et al. 2019b). Normal mitochondrial function was reduced in undifferentiated SH-SY5Y cells after OGD (Herrmann et al. 2013).
Nuclear respiratory factor-1 and mitochondrial transcription factor A were decreased in undifferentiated SH-SY5Y cells after ischemia/hypoxia (Lin et al. 2020). Cytochrome oxidase activity was decreased (Lin et al. 2020). When mitochondrial dysfunction occurs, cells usually remove the dysfunctional mitochondria with mitophagy. Ischemia/hypoxia induced mitophagy has been established in one study in undifferentiated SH-SY5Y cells (Ping et al. 2021).
Anti-inflammatory cytokines can be induced by V-maf musculoaponeurotic fibrosarcoma oncogene homolog B. This transcriptional activator was found to be decreased after ischemia/hypoxia in one study using undifferentiated SH-SY5Y cells ). The anti-inflammatory factor interleukin-10 was reduced after ischemia/hypoxia in one study with undifferentiated SH-SY5Y cells .
Inhibitor of κB (IκB) kinase is an enzyme involved in propagating the cellular response to inflammation. It activates NF-κB. This kinase is investigated in four studies using undifferentiated SH-SY5Y cell, two of which investigated IκB kinase in combination with the activity of NF-κB Huang et al. 2021;Jiang et al. 2012;. The results showed an increase in IκB kinase.
Autophagy Intracellular autophagy is activated in different cell types of the brain upon nutrient starvation or metabolic stress. It is the phagocytosis of cytoplasmic materials, which are captured by autophagosomes and fuse with lysosomes to form autolysosomes and are then degraded. Several factors related to autophagy after ischemia/hypoxia have been investigated in human neuronal in vitro models. Ischemia/hypoxia increased the level of microtubule-associated protein 1 light chain 3 (LC3), as was found in undifferentiated (n = 9) Liu et al. , 2021Niu et al. 2018;Shi et al. 2012;Wang et al. 2017Wang et al. , 2019cZhi et al. 2020;Zhou et al. 2019) and differentiated SH-SY5Y cells (n = 2) Zhang et al. 2019b). Con-nexin32 (CX32) was found to inhibit the autophagic effect of nuclear receptor 4A1 (NUR77) during ischemia/hypoxia in one study using undifferentiated SH-SY5Y cells (Ping et al. 2021). Increased Beclin-1 expression was found to increase the level of autophagy in four studies using undifferentiated SH-SY5Y cells (Shi et al. 2012;Wang et al. 2017Wang et al. , 2019cZhi et al. 2020) and one study using differentiated SH-SY5Y cells during ischemia/hypoxia .

Cell Death
In addition to anti-apoptotic factors, the BCL-2 family has members that act as pro-apoptotic factors, such as BCL2 associated X (BAX). Expression of this pro-apoptotic protein increased in undifferentiated SH-SY5Y cell models of cerebral ischemia (n = 23) Dong et al. 2019;Feng et al. 2021;Gao et al. 2020;Guo 2019;He et al. 2014;Hu et al. 2020;Huang et al. 2018;Li et al. , 2020aLiu et al. 2018Liu et al. , 2020Wang et al. 2013Wang et al. , 2018aWang et al. , 2019cWang et al. , 2021bWu et al. 2020a;Xu et al. 2020b;Yan et al. 2020;Zhang et al. 2016b;Zhao and Wang 2020;Zhou et al. 2019). Expression of BAX is only investigated in one study using differentiated SH-SY5Y cells. These results showed increased expression of BAX during and after hypoxia (Sriwastva et al. 2020). BAX induces cell death via mitochondrial membrane permeabilization. This results in the release of small molecules such as cyt-c and apoptosis inducing factor (AIF), among others.
Release of cyt-c is investigated in four studies in undifferentiated SH-SY5Y cells Li et al. 2020a;Lin et al. 2020;Wang et al. 2019b). The results showed increased cyt-c release, which led to increased apoptotic rate. Bcl2interacting killer (BIK) enhances apoptosis, it was found to be upregulated in undifferentiated SH-SY5Y cells . Another pro-apoptotic protein of the BCL-2 family has been investigated in undifferentiated SH-SY5Y cells, namely Bcl-2-like protein 11 (BIM). The expression of BIM was found to be increased during ischemia/hypoxia (n = 2) Song et al. 2019). In addition to BIM, apoptosis facilitator Bcl-2-like protein 14 (BCL2L14) has been investigated in undifferentiated SH-SY5Y cells and the results showed increased expression, which induced apoptosis during ischemia/hypoxia (Yao et al. 2019).
AIF translocate to the nucleus upon release by mitochondria, where it triggers chromatin condensation and DNA fragmentation to induce parthanatos. AIF translocation is investigated in undifferentiated SH-SY5Y cells (n = 2). The results showed nuclear translocation of AIF induced by ischemia/hypoxia Wang et al. 2018a). One study investigated the expression of AIF in differentiated SH-SY5Y cells and the results showed increased expression induced by hypoxia ). In addition to AIF release from mitochondria, poly(ADP-ribose) polymerase-1 (PARP-1) overactivation mediates AIF release. Normal function of PARP-1 is DNA repair. Cleaved PARP induces parthanatos, which has been investigated in undifferentiated SH-SY5Y cells (n = 5), differentiated SH-SY5Y cells (n = 3), and undifferentiated SK-N-SH (n = 1). Cleaved PARP was increased, which was in turn associated with ischemia/hypoxia induced parthanatos Lehane et al. 2013;Li et al. 2020a;Ruan et al. 2020;Shi et al. 2020;Sriwastva et al. 2020;Wang et al. 2018a;Xing et al. 2021;Xu et al. 2020a). Endothelial protein C receptor (EPCR) showed anti-cell death features by signaling via protease-activated receptor 1 (PAR1) and PAR3. Levels of these receptors were decreased in differentiated SH-SY5Y cells during and after ischemia/hypoxia (n = 1) .
Another key player in apoptosis is the tumor suppressor gene P53. This gene is activated by hypoxia and can induce apoptosis. P53 is investigated in one study using undifferentiated SH-SY5Y cells and showed that P53 activity was induced by ischemia/hypoxia ). PI3K/AKT/mTOR pathway is an intracellular signaling pathway that plays a role in regulating the cell cycle and apoptosis. This has been investigated in sixteen studies using undifferentiated SH-SY5Y cell models of cerebral ischemia. Overall, a decrease in phosphorylated protein kinase B (AKT), phosphorylated phosphoinositide 3-kinases (PI3K), mTOR and forkhead box proteins (FOXO) were found Gao et al. 2020;Guo et al. 2018;Liu et al. 2019bLiu et al. , 2021Marutani et al. 2012;Song et al. 2019;Wang et al. 2019b;Wen et al. 2020;Wu et al. 2020a;Yi et al. 2020;Zhi et al. 2020;Zhou et al. 2019). Decreased levels of AKT were also found in two studies using differentiated SH-SY5Y cell models Han et al. 2017). Epidermal growth factor receptor showed antiapoptotic effects by activation of the PI3K/AKT pathway in one study with undifferentiated SH-SY5Y and one study with differentiated SH-SY5Y cell models of cerebral ischemia (Lin et al. 2011;Wu et al. 2020b). Glycogen synthase kinase 3β (GSK-3β) was found to be dependent on several signaling pathways, like PI3K/AKT. One study in differentiated SH-SY5Y cells found that increased levels of GSK-3β led to increased apoptotic rate (Lin et al. 2011). Upregulation of phosphatase and tensin homolog (PTEN) was found in four undifferentiated SH-SY5Y and one undifferentiated SK-N-SH cell models of cerebral ischemia Guo 2019;Guo et al. 2018;Jin et al. 2020;Yi et al. 2020).
Increased nuclear fragmentation has been found in undifferentiated SH-SY5Y cell model (n = 1) ). FMS-like tyrosine kinase-3 (FLT3) has been implicated in cell survival. One study into this kinase showed that FLT3 expression increased in undifferentiated SH-SY5Y cell model of cerebral ischemia .
The pro-apoptotic factor 'programmed cell death 4' (PDCD4) was found to be increased in an undifferentiated SH-SY5Y cell model of cerebral ischemia (Shan et al. 2022). Increased expression of inhibitor of growth protein 5 (ING5) was correlated with an increased apoptotic rate in an undifferentiated SH-SY5Y cell model of cerebral ischemia ). Hypoxia-inducible factor-1 showed to be upregulated, followed by downregulation during ischemia/hypoxia and played a role in ischemia/hypoxia induced cell death in undifferentiated SH-SY5Y cells (n = 4) and SK-N-SH cells (n = 1) Niu et al. 2018;Olechnowicz et al. 2012;Ruan et al. 2021;Zhang et al. 2016c). Modulation of apoptosis-1 is an apoptotic pathway that was found to be increased in undifferentiated SH-SY5Y cells (n = 1) .

NT2-N Cell Model
Two studies investigated ischemia/hypoxia in differentiated NTERA-2 clone D1 (NT2-N) cell model. These studies investigated complement system activation upon ischemia/ hypoxia induced stress. The complement system is part of the immune system that promotes inflammation. The first study into expression of several complement regulators, such as CD35, CD45, CD55, CD59, C3a and C5a, showed decreased expression of CD55 and an increased vulnerability to hypoxia (Pedersen et al. 2007b).
The second study aimed to investigate whether complement regulatory protein CD59 was expressed in the NT2-N cell model and what its function was. The results showed that CD59 was expressed and that this protein played a role in protecting the cells from complement attack and eventually cell death (Pedersen et al. 2007a).

hiPSC-Derived Neuronal Cell Model
Two studies investigated ischemia/hypoxia in healthy hiPSC-derived neuronal cell models. One study investigated hiPSC-derived neuronal networks consisting of excitatory and inhibitory neurons. Electrophysiological measurements showed decreased neuronal activity, as well as decreased synchronous activity, throughout the network. Furthermore, the results showed increased levels of apoptosis on time scales of 24-48 h of hypoxia, followed by an increased level of dead cells (Pires Monteiro et al. 2021).
The other study compared the effect of ischemia/hypoxia on hiPSC-derived neurons with that on differentiated SH-SY5Y cells. The results showed less cell survival of SH-SY5Y cells than of hiPSC-derived neurons .
One study with hiPSC-derived neurons investigated the effect of neuronal network activation by the mildly excitatory hormone/neurotransmitter ghrelin (Pires Monteiro et al. 2021). The results showed partial preservation of neuronal network functioning after hypoxia (Pires Monteiro et al. 2021).
Oxidative/nitrosative stress was investigated as a treatment target in twenty studies using undifferentiated SH-SY5Y cells and one study using differentiated SH-SY5Y cells. These studies showed that various chemical compounds could reduce oxidative stress by reducing the formation of ROS or increasing anti-oxidant defense mechanisms, which led to increased cell survival (Agudo-Lopez et al.
Autophagy was investigated as a treatment target in five studies using undifferentiated SH-SY5Y cells and two studies using differentiated SH-SY5Y cells Lai et al. 2020;Liu 2016;Wang et al. 2019c;Zhang et al. 2019b;Zhi et al. 2020). These studies showed that various pharmacological compounds could maintain healthy autophagic flux.

Discussion
With this scoping review, we provide an overview of human neuronal in vitro models that have been used to study human neuronal responses to ischemia or hypoxia, including the established steps in the pathophysiological cascade and potential treatment targets. The scoping review shows that models based on SH-SY5Y cells are the most widely used in this field. Almost all studies used immunocytochemical techniques to investigate cell survival and protein expression. Neuronal functioning was investigated in only one study using hiPSC-derived neuronal networks. Cell death was the most investigated pathomechanism, followed by oxidative stress and inflammation. Studies on apoptosis revealed the importance of diverse pro-and anti-apoptotic factors such as the BCL-2 protein family that lead to apoptosis and eventually cell death. Studies on oxidative stress showed higher levels of ROS and lower levels of ROS scavenging molecules after ischemia or hypoxia. The most important results on inflammatory responses were increased levels of proinflammatory and decreased levels of anti-inflammatory cytokines.
SH-SY5Y cells are a subclone from SK-N-SH, originating from a single 4-year-old female neuroblastoma patient (Biedler et al. 1973). The wide spread use of SH-SY5Y cells indicates that pathomechanisms of human neuronal responses to ischemia or hypoxia have mainly been investigated in a cancerous cell line . It is known that the genetic background of a cell line affects functionality, as well as responses to interventions (Mossink et al. 2021b). Thus, a cell line with cancerous characteristics will probably respond to ischemia or hypoxia differently than healthy (neuronal) cells.
Moreover, although SH-SY5Y cells can be differentiated into neuron-like cells upon stimulation with RA (Khwanraj et al. 2015;Korecka et al. 2013;Shipley et al. 2016), the majority of studies has been conducted in undifferentiated SH-SY5Y cells. Undifferentiated and differentiated SH-SY5Y cells differ in morphology and function. Undifferentiated SH-SY5Y cells have large cell bodies with short truncated processes (Kovalevich and Langford 2013). These undifferentiated cells express immature neuronal markers, such as SOX2, and lack the expression of mature neuronal markers, such as synapsin1/2 positive puncta. Differentiated SH-SY5Y cells form and extent neuritic processes (Kovalevich and Langford 2013) and show increased expression of synaptic markers, such as SNAP25 and SYN1 (Forster et al. 2016). Differentiated SH-SY5Y cells acquired the neuronal ability to produce trains of spikes upon prolonged stimulation in currentclamp experiments (Tosetti et al. 1998) and showed spontaneous firing, bursting, and network behaviour on microelectrode arrays (MEAs) (Yoon et al. 2020), that were not observed in undifferentiated SH-SY5Y cells. These differences may lead to divergent responses to ischemia/ hypoxia between undifferentiated and differentiated SH-SY5Y cells. Thus, it is questionable whether results from studies in undifferentiated SH-SY5Y cells can be extrapolated to neurons or neuronal networks.
OGD directly models ischemic stress and allows for the exploration of divergent components of the pathophysiological cascade caused by ischemic stress. Furthermore, modelling of OGD is straightforward and can be replicated in various laboratories (Cimarosti and Henley 2008). Other proposed cerebral ischemia models are based on chemical or enzymatic modulation of specific pathophysiological pathways. These can be used to study specific components of the ischemic cascade, such as excitotoxicity or glutathione depletion. These techniques are commonly used to investigate the effect of increased glutamate release or induced oxidative stress caused by ischemic stroke, respectively (Nicholls and Ward 2000). In turn, these may lead to other downstream effects, for example calcium dysregulation (by excitotoxicity) or glutathione depletion (by oxidative stress) (Mattson et al. 1995;Mytilineou et al. 1998). However, it is important to note that these techniques probably model only parts of the ischemic stress cascade (Cimarosti and Henley 2008). Moreover, it is unclear whether chemical or enzymatic induction of downstream effects has the desired effect on undifferentiated SH-SY5Y cells, which are often used in ischemic stroke research. For example, none of the studies included in this scoping review touched upon excitotoxicity. This highlights the need for a careful assessment of the suitability of the various cell models and modes of ischemic stress induction in light of the research question at hand.
The majority of human neuronal in vitro studies on effects of ischemia/hypoxia focused on cell survival and protein expression. Expression of proteins that might have deleterious effects (such as pro-apoptotic factors, pro-inflammatory cytokines, oxidative stress factors) apparently increase and expression of proteins that might have protective effects (such as anti-apoptotic factors and anti-inflammatory factors) decrease. Various studies showed this in relationship to cell death, apoptosis, oxidative stress, or inflammation. This focus on microscopic measures of relatively downstream processes neglects early steps in the pathophysiological cascade of ischemia, such as energy failure caused by OGD (Galkin 2019), loss of cell ion homeostasis (Caplan 2009;Taoufik and Probert 2008), and depression of excitatory synaptic transmission. It also neglects effects on neuronal functionality and activity (Bolay et al. 2002;Hofmeijer and van Putten 2012;Pires Monteiro et al. 2021).
HiPSC-derived neuronal networks on MEAs are currently the only functional human neuronal cell model used to investigate neuronal network activity in cerebral ischemia (Pires Monteiro et al. 2021). HiPSCs can be derived from healthy individuals or patients and differentiated in excitatory neurons and inhibitory neurons. This allows establishment of healthy neuronal networks with physiological excitatory -inhibitory (E/I) ratio's (Frega et al. 2017;Mossink et al. 2021a). Genetic characteristics of the donor are preserved in the model, making it possible to study patient-specific network behaviour, including responses to ischemia/hypoxia (Mossink et al. 2021b). By the use of MEAs, functionality of many neurons and synapses in the network can be studied simultaneously. Despite the lack of normal brain architecture, neuronal network functioning can be readily derived, in addition to microscopic measures. The dynamics of neuronal network failure, from reversible to irreversible damage or recovery, can be monitored continuously (Pires Monteiro et al. 2021). This allows identification of early determinants of irreversibility, including possible effects of interventions.
Studies on interventions to ameliorate pathophysiological processes during or after ischemia/hypoxia focused on a wide range of treatment targets. Evidence from studies in SH-SY5Y cells mostly suggests effects of treatments targeting apoptosis or other modes of cell death, because this was the most targeted pathomechanism (Agudo-Lopez et al. 2010;Chen et al. 2021b;Cheng et al. 2014;Dong et al. 2021;Gao et al. 2017;Hao et al. 2015;Huang et al. 2021;Jiang et al. 2012;Li et al. , 2020aLin et al. 2011Lin et al. , 2020Liu et al. 2019a, b;Luan et al. 2013;Marutani et al. 2012;Park et al. 2017;Song et al. 2019;Sriwastva et al. 2020;Wang et al. 2013Wang et al. , 2019bWang et al. , c, 2021aWu et al. 2020a, b;Xu et al. 2020b;Yang et al. 2014;Zappala et al. 2021;Zhang et al. 2019b;Zhi et al. 2020). One study in hiPSC-derived neuronal networks on MEAs showed beneficial effects of early neuronal network stimulation on functional network recovery and cell survival (Pires Monteiro et al. 2021). All intervention studies were small sampled proof of principle experiment and none of the intervention effects was conformed in follow-up studies.
In vitro models of OGD provide a versatile platform for replicating diverse aspects of ischemic stroke pathology in a reproducible manner. Compared to in vivo models, these models are often quicker, cheaper and easier to establish. Also, the use of human-derived models may contribute to a reduction of animal model use. However, for clinical translation, combinations of in vitro and in vivo models will continue to be necessary (Trotman-Lucas and Gibson 2021). With the increasing amount of modelling possibilities, it will remain important to consider the optimal model based on the research goal at hand.
The strengths of this scoping review are the prospective literature research and systematic data extraction. Limitations include that the review was not pre-registered, data extraction was conducted by only one reviewer (EV), and lack of quality assessment of the included articles. Data on effort, costs, and reproducibility of the models are not systematically presented in the included research papers.
In conclusion, the majority of human neuronal in vitro studies on experimental cerebral ischemia/hypoxia exploited the cancerous SH-SY5Y cell line, mostly in a relatively immature stage and not differentiated into neuronal-like cells. Most studies focused on microscopic measures of single downstream pathophysiological processes, such as cell death. Intervention studies targeted a myriad of treatment targets and none of the intervention effects presented in this review have been confirmed in follow-up experiments. We suggest the use of networks of differentiated neurons on 1 3 MEAs. These provide new opportunities to analyze neuronal network functioning, in addition to microscopic measures. This allows coupling of neuronal structure and function (which is not possible in unfunctional neurons) in order to identify drivers of neuronal network (dys)functioning. The use of hiPSC-derived neuronal networks provides additional opportunities to study networks with physiological E/I ratios and identify person-specific characteristics. For intervention studies, we favor standardization to collect strong evidence of (in)efficacy to improve translation from bench to bedside.
Acknowledgements Graphical abstract was created with Biorender. com.
Author Contributions EV screened all articles for eligibility. EV and JH wrote the main manuscript text and created supplementary Fig. 1. EV prepared graphical abstract, Fig. 1, 2 and Tables 1, 2. All authors reviewed the manuscript.

Funding This research has been supported by an institutional research grant.
Data Availability Enquiries about data availability should be directed to the authors.

Declarations
Conflict of interest All authors certify that they have no affiliations with or involvement in any organization or entity with any financial interest or non-financial interest in the subject matter or materials discussed in this manuscript.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.