Introduction

Stroke is the second leading cause of death and third leading cause of disability in adults worldwide, with one in four people affected over their lifetime, but with few effective therapies [1, 2]. Despite the rate of incidence and mortality is stable or even declining over decades, the number of stroke patients and survivors, disability-adjusted life-years lost due to stroke, and stroke-related deaths are increasing [3, 4]. Tissue plasminogen activator (tPA) is the only drug approved by FDA for ischemic stroke and no drug is available for hemorrhagic stroke. And, only 3%–5% stroke patients can receive tPA treatment, due to the narrow therapeutic window, contraindications and complications. Hence, it is urgent to develop effective drugs for stroke treatment.

Numerous neuroprotective agents targeting excitotoxicity, oxidative and nitrosative stress, inflammation or others have been studied from bench to bedside in the past decades, but almost all of them are failed or still in the transition of bench to bedside [5]. Many factors are related to the extremely low success rate of anti-stroke drug development. One of the major factors is the species difference between the experimental and clinical research [6, 7]. It is well-known that anti-stroke neuroprotective agents in the laboratory are primarily studied in rodent models, rather than primate or humanized models, which will be inevitable to misguide the clinical efficacy to some extent. Compared with human beings, rodents belong to anencephalic animals with difference in brain anatomy and functional organization, which determine that they cannot fully mimic the physiological, pathological and anatomic features of human brain. Although whether the lissencephalic structure of brain has an impact on the efficacy of anti-stroke drugs is still under debate, it is certain that the difference of brain anatomy and functional organization can affect the infarct localization [8]. Besides, rodent brain has lower percentage of white matter as compared to human brain. The percentage of white matter in whole brain accounts for 60% in humans, 15% in rats, and 10% in mice [7]. And the ischemic damage degree of white matter not only plays a vital role in the prognosis of stroke outcome, but also is the major cause of hemiparesis in stroke.

Compared with rodents, non-human primate animals are the better research subject for anti-stroke drug development, which have similar genetic background with human beings. However, there are disadvantages for non-human primate stroke model [9]. Firstly, the price of non-human primate animals is relatively expensive. Secondly, the in-vivo stroke model of middle cerebral artery occlusion surgery in non-human primate requires more seasoned operational experience and skills. Thirdly, the equipment for long-term monitoring of non-human primate not only takes up space, and also is expensive. More importantly, the extensive use of non-human primates is contrary to the welfare and ethical issues of experimental animals, which cannot provide enough data for the repeatability across different study subgroups. Thus, how to establish a stoke research model, that is better in line with the physiology and function of human brain, has been the bottleneck for effective anti-stroke drug development.

The recent developed method for generating three-dimensional (3D) cerebral organoids (COs) from human pluripotent stem cells provides a new sight for stroke research. The in-vitro cultured COs not only have advantages in diverse cell types (including but not limited to, neural progenitor cells/neural stem cells [NPCs/NSCs], neurons, and glia cells), easy accessibility of cell resource, enough cell number for drug screening (more than 106 cells for single CO at 55 days) [10], and 3D tissue structure (up to 4 mm in diameter) [11, 12], but also show detectable neural connectivity and brain functionality that can recapitulate features of in-vivo human brain development and maturation [11, 13,14,15]. Moreover, the cultured COs have similar cell composition, cortical structure and epigenetic genomics as compared to human brain [16]. The COs contain most neural lineage cells that are expressed in human brain, and especially have positive expression of outer radial glial cells that are not expressed in rodent brain [11]. And outer radial glial cells have a key role in the evolutionary expansion of the mammalian neocortex. Taken together, from the perspective for the drug development, the COs have a promising application in establishing humanized ischemic stroke model, and its unique features, such as high throughput, good accessibility and 3D complex structure, make them good alternatives to animal models for efficacy screening of potential anti-stroke drugs.

Although a most-recent study has subjected human COs to hypoxic injury to investigate neuronal damage and regeneration after hypoxic injury [17], the features of COs-based humanized ischemic stroke model in response to oxygen-glucose deprivation (OGD) damage and its application in testing anti-stroke drugs are still unknown. Our group have accumulated a series research experience in the development of potential anti-stroke drugs and in the study of NSCs and COs [10, 15, 18,19,20,21,22]. Therefore, this study is designed to establish humanized ischemic stroke model of OGD based on COs by exploring the cell type and composition of COs that is used for model preparation and the time-effect relationship of the model, and further verifying the feasibility and applicability of the model in potential anti-stroke drug efficacy testing, ho** to provide a novel platform for anti-stroke drug translational research.

Materials and methods

Human embryonic stem cell culture

Human embryonic stem cells (hESCs) H1 (passages 45–55) were cultured as previously described with minor modifications [22, 26], and mediate the reprogramming of differentiated cells to stem cells [27]. However, in differentiated neurons, glia and other neural cell types, SOX2 is either not expressed, or important for very specific cell types. For example, a recent study reported SOX2 was strongly expressed in mouse postmitotic thalamic projection neurons [28]. Thus, the high cell proportion of SOX2+ cells made us have to wonder whether there was Tuj1+/SOX2+ neurons in cultured 85 d-CO. Further flow cytometry confirmed our hypothesis and found there was about 16.3% Tuj1+/SOX+ cells in 85 d-CO, providing more evidence for the higher similarity between 85 d-CO and in-vivo brain (Fig. 2b5). Besides, we previously demonstrated that the cell number in single 85 d-CO was more than 2 × 106 cells [22], which also provides evidence for the enough cell number in single CO for drug efficacy testing. Moreover, many studies already demonstrated COs can faithfully model the gene expression, proteomic expression, functional performance and other features of in vivo human brain, including 85 d-CO [16, 29, 30]. Thus, 85 d-CO was chosen for establishing humanized ischemic stroke model in this study.

In addition to controlling the morphology and culture time of COs that were used for model preparation, we also explored the selection criteria for the size of COs. By measuring the size of 134 COs with 85 days culture time from five culture batches, and found the majority of them were more than 2.00 mm in diameter (2.38 ± 0.03 mm, n = 98, moderate size group), and a small part had a relatively smaller diameter (1.82 ± 0.01 mm, n = 36, small size group) (Fig. 3a, b). And, with introduction of OGD model (a classical in-vitro model of ischemic stroke), we further examined the injury degree of the COs from the two groups. As shown in Fig. 3c, d, the cell cytotoxicity of LDH release and cell apoptosis of Caspase 3 activity in moderate size group were about half the levels of those in small size group, indicating the different sensitivity to OGD injury for the two groups. In order to keep the experimental study as parallel as possible, COs in moderate and small size groups are suggested not to be used together for establishing model and drug efficacy testing. In our following study, 85 d-CO with diameter more than 2.00 mm was chosen for establishing humanized ischemic stroke model.

Fig. 3: The volume size of COs at 85 days.
figure 3

a The morphology and volume size of COs at 85 days in moderate and small size groups. b The diameter of COs at 85 days in moderate and small size groups (2.38 ± 0.03 vs. 1.82 ± 0.01 mm, n = 98 and 36 respectively). c, d The LDH release and Caspase 3 activity of COs at 85 days in moderate and small size groups after 8 h oxygen-glucose deprivation (OGD) exposure. Data are shown as mean ± SEM, and are representative of three independent experiments. *P < 0.05, **P < 0.01. All scale bars are as shown.

Full size image

The time-effect relationship of COs-based humanized ischemic stroke model

Time-effect relationship is an important reference for establishing a novel research model. After separating 85 d-CO into 0 h (control group, under normal culture condition), 2 h, 4 h, 8 h, 12 h OGD exposure groups randomly, we explored the injury degree of COs (Fig. 4). H&E staining showed that the cell density in the tissue section of COs gradually decreased with the prolonged OGD exposure time (Fig. 4a). Compared with control group, the cell cytotoxicity of LDH release and cell apoptosis of Caspase 3 activity in 2 h, 4 h, 8 h, 12 h OGD exposure groups gradually increased with significant difference (Fig. 4b, c). Moreover, further examination for the expression of apoptotic cells by immunohistochemical staining Cleaved-Caspase 3, a marker of cell apoptosis, confirmed the result (Fig. 4d, e). The number of apoptotic cells gradually increased in the 0 h, 2 h, 4 h, 8 h and 12 h OGD exposure groups by quantification of Cleaved-Caspase 3+ cells in each group (Fig. 4d, e). Taken together, the ischemic injury degree of COs is gradually increased along the OGD exposure time. The establishment of the time-effect relationship of COs-based humanized ischemic stroke model provides an experimental reference for selecting proper OGD exposure time or ischemic injury degree for drug testing.

Fig. 4: The time-effect relationship of COs-based humanized ischemic stroke model.
figure 4

a The H&E staining of COs in 0 h, 2 h, 4 h, 8 h and 12 h OGD groups. b, c The LDH release and Caspase 3 activity of COs in 0 h, 2 h, 4 h, 8 h and 12 h OGD groups. d, e Immunostaining and quantification of Cleaved-Caspase 3 positive cells in COs of 0 h, 2 h, 4 h, 8 h and 12 h OGD groups. Data are shown as mean ± SEM, and are representative of three independent experiments. **P < 0.01 vs. control. All scale bars are as shown.

Full size image

The feasibility and validity of COs-based humanized ischemic stroke model in drug testing

Whether the model is feasible for drug testing is still a question due to it is the first study to focus on the preparation of COs-based humanized ischemic stroke model. In order to validate the sensitivity of the model for drug efficacy testing, the proven anti-apoptotic and pro-apoptotic compounds pan-Caspase inhibitor Z-VAD-FMK (20 μM) and Bcl-2 inhibitor navitoclax (0.5 μM) were given during OGD exposure to explore their effect on OGD injury (Fig. 5). As shown in Fig. 5a, b, treatment of Z-VAD-FMK or navitoclax significantly alleviated or aggravated ischemic injury respectively, as compared to vehicle control group under OGD condition. Besides, TUNEL staining for apoptotic cells and Nissl’s staining for survival neurons further confirmed the anti-apoptotic effect of Z-VAD-FMK and pro-apoptotic effect of navitoclax in the humanized COs ischemic stroke model (Fig. 5c–f). As shown in Fig. 5c, d for TUNEL staining, compared to control group, the vehicle group had more expression of TUNEL positive cells. And Z-VAD-FMK treatment significantly decreased the expression of TUNEL+ cells as compared to vehicle group, and showed no difference with control group; navitoclax treatment significantly increased the expression of TUNEL+ cells as compared to vehicle group (Fig. 5c, d). Moreover, Nissl’s staining for survival neurons showed the vehicle group had less expression of Nissl’s positive cells than control group (Fig. 5e, f). And Z-VAD-FMK treatment significantly increased the expression of Nissl’s positive cells as compared to vehicle group and had no difference with control group, and navitoclax treatment significantly decreased the expression of Nissl’s positive cells as compared to vehicle group (Fig. 5e, f). Therefore, the COs-based humanized ischemic stroke model was sensitive to cerebral ischemic injury, and the ischemic injury can be changed with the treatment of anti-apoptotic or pro-apoptotic compounds, proving evidence for the feasibility of the model in drug efficacy testing. In addition to apoptosis, we explored the involved cell death manners in COs under OGD condition. As shown in Fig. 5g, there were positive expressions of Cleaved-Caspase 3+ apoptotic cells, Phosphor-RIP3+ necroptotic cells, LC3B+ autophagic cells and GPX4+ ferroptotic cells in the groups of control and vehicle groups, suggesting that there may be multiple cell death manners in COs induced by OGD injury.

Fig. 5: The effect of anti-apoptotic and pro-apoptotic compounds in COs-based humanized ischemic stroke model.
figure 5

a The LDH release, b Caspase 3 activity, c the representative image of TUNEL staining, d quantification of the percentage of TUNEL+ cells, e the representative image of Nissl’s staining and f quantification of the percentage of Nissl’s cells per field in humanized COs ischemic stroke model under normal or under 8 h OGD condition with or without treatment of pan-Caspase inhibitor Z-VAD-FMK (20 μM) and Bcl-2 inhibitor navitoclax (0.5 μM). g Immunostaining for the expression of Cleaved-Caspase 3, Phopho-RIP3, LC3B and GPX4 in the control and vehicle groups of COs. DAPI labels nuclei (blue). Data are shown as mean ± SEM, and are representative of three independent experiments. **P < 0.01 vs. control; #P < 0.05, ##P < 0.01 vs. vehicle. All scale bars are as shown.

Full size image

The sensitivity of COs-based humanized ischemic stroke model for neuroprotective compounds

The efficacies of four neuroprotective compounds (edaravone, butylphthalide, P7C3-A20, and ZL006) were tested in the humanized COs ischemic stroke model (Fig. 6a). According to the result of cell cytotoxicity of LDH release and cell apoptosis of Caspase 3 activity as shown in Fig. 6b, c, 10 μM edaravone, 10 μM butylphthalide, 10 μM P7C3-A20 and 10 μM ZL006 protected COs from ischemic injury under 8 h OGD condition. Therefore, the COs-based humanized ischemic stroke model can be also used for potential neuroprotective anti-stroke drug efficacy testing.

Fig. 6: The effect of neuroprotective compounds in COs-based humanized ischemic stroke model.
figure 6

a The chemical structure of edaravone, butylphthalide, P7C3-A20 and ZL006. b The LDH release and c Caspase 3 activity of COs under 8 h OGD condition with or without edaravone (10 μM), butylphthalide (10 μM), P7C3-A20 (10 μM) and ZL006 (10 μM) treatment. *P < 0.05, **P < 0.01 vs. vehicle group. Data are shown as mean ± SEM, and are representative of at least three independent experiments.

Full size image

Discussion

It is well-known that a good experimental model should be reliable and effective, which can not only produce consistent and reproducible results, but also predict clinical drug efficacy. At present, numerous in-vivo stroke research is mostly carried out in rodent animals, wherein stroke models in mouse and rat account for 27% and 66% of total in-vivo stroke model respectively [31]. And the OGD model is usually used as in-vitro stroke model to explore the effect of potential drug on the specific brain cell types, such as neurons, glial cells, etc., for efficacy evaluation and target validation at the molecular level [31]. Although almost all potential neuroprotective drugs have been studied for pharmacodynamics and mechanisms in the in-vitro and in-vivo models of stroke, all of them are still in the clinical translational research stage and few effective neuroprotective drug is available for stroke treatment so far [5]. Besides, although mouse or rat brain slices have been widely used as 3D model to study the role of different neural cell types in stroke, there are many disadvantages for the model, including additional mechanical injury during cutting, severed neural connections or loss of connections from distant brain areas to the area of interest, limited in-vitro culture time and so on [32,33,34]. The cultured COs not only have advantage in human species, but also show advantage in its 3D tissue structure, diverse cell types, detectable neural connectivity and projection that recapitulate features of in-vivo human brain. Therefore, we proposed whether it is possible to establish a humanized stroke model based on COs to provide a novel platform for anti-stroke drug efficacy testing and pathophysiological mechanism research. A most-recent study exposed COs to low-oxygen concentrations to model hypoxic injury [17], but the features of COs-based humanized ischemic stroke model in response to OGD damage and its application in testing anti-stroke drugs are still not explored. Thus, this study planned to establish an in-vitro humanized ischemic stroke model of OGD and explored its application in testing anti-stroke drugs (Fig. 7).

Fig. 7
figure 7

Diagram showing the proposed application of COs-based humanized ischemic stroke model in studying anti-stroke drugs.

Full size image

How to standardize the process of model preparation is the key problem to be solved for a new model. Due to the heterogeneity of COs in morphology and volume, we explored the selection criteria of COs for ischemic stroke model preparation. Firstly, we examined the cell type and composition of COs at different time. With the prolongation of culture time, expression of neural stem cells in COs gradually decreased and the terminally differentiated neurons and astrocytes gradually increased. And, there was the formation of forebrain and choroid plexus in 75 d-CO, indicating the successful generation of COs. Thus, we further prolonged the culture time of COs and employed 85 d-CO for model preparation, which had higher cell maturation degree and more expression of terminal differentiated neural cells than 75 d-CO. Although we found the formation of apoptotic cavity in tissue center of 140 d-CO, whether COs with longer culture time are suitable for OGD model preparation remains to be further verified. After measuring the diameter of 85 d-CO from different culture batch, we found that most COs were more than 2 mm in diameter and a small part of COs was <2 mm in diameter. Therefore, we grouped COs from same culture batch into moderate size group and small size group respectively. Due to the different injury degree of COs between moderate and small size groups under OGD condition, we set the selection criteria of COs for establishing humanized stroke model, that is, in our study only employing COs with diameter more than 2 mm for the model preparation. As for a new research model, it is important to establish the time-effect relationship of the model. Thus, we explored the time-effect relationship of COs-based humanized stroke model, and found the ischemic injury degree of COs gradually aggravated with the extension of OGD time, providing an experimental reference for selecting the proper OGD time and ischemic injury degree during drug testing.

Ischemia would initiate the ischemic cascade which involves a series of biochemical events, including energy failure, ion imbalance and excitotoxicity, oxidative stress, cell death, complement system activation, initiation of the inflammation and immune response [35]. And increasing evidence proves that multiple cell death pathways are persistently present in the ischemic core and penumbral area, which play either beneficial, deleterious or dual roles in the progression of post-stroke brain damage [35]. The final death or survival state of ischemic injured cells is directly correlated with the prognosis after stroke. In order to validate the feasibility and applicability of the model for potential drug testing, this study introduced the proven anti-apoptotic and pro-apoptotic compounds pan-Caspase inhibitor Z-VAD-FMK and Bcl-2 inhibitor navitoclax as positive control to demonstrate that the model was sensitive to ischemic apoptotic injury and related treatment. In addition to apoptosis, this study provided early evidence that the cell death manners in COs induced by OGD injury may involve apoptosis, necroptosis, autophagy and ferroptosis. Future study needs to provide more evidence for the involved cell death manners of the model, and explore the vulnerability of different cell types and their sites inside the sphere of COs to different cell death manners under OGD condition.

Besides, four neuroprotective compounds (edaravone, butylphthalide, P7C3-A20 and ZL006) were introduced for further efficacy testing in the humanized COs ischemic stroke model. Thereinto, edaravone, a free radical scavenger, has been first approved in Japan by the Ministry of Health, Labour and Welfare for the treatment of ischemic stroke within 24 h of onset, and now is widely used for acute ischemic stroke within 24 h of onset following the guidelines for stroke management [36, 37]; butylphthalide is an anti-cerebral-ischemia drug which has been approved for ischemic stroke treatment by China Food and Drug Administration in 2002 [38]; P7C3-A20 has been demonstrated with neuroprotection in rodent model of ischemic stroke by our group [24] and of other disease models by several research groups [20]; ZL006, a small-molecular inhibitor of nNOS-PSD-95 interaction, has been demonstrated with neuroprotection in rodent model of ischemic stroke [39]. Here, we demonstrated that treatment of edaravone, butylphthalide, P7C3-A20 and ZL006 can protect COs from OGD injury. Therefore, COs-based humanized ischemic stroke model can be used for potential neuroprotectant efficacy testing.

Although this study explored the feasibility and applicability of the COs-based humanized ischemic stroke model in testing anti-stroke drugs, there are still many questions need to be answered. One of those major questions is lack of compounds as negative control to compare the difference of its efficacy in this model from that in clinical trial. It is known that numerous neuroprotective compounds have failed in clinical trials. However, until now, most of these compounds have failed due to safety or failure to show significant protection, and there are few reports on whether they failed due to differences in action targets caused by species difference between preclinical experiments and clinical trials. Future study needs to provide answer for the question. Besides, blood vessel unit, blood brain barrier and immune cells play crucial role in human ischemic stroke, but are absent in COs, which hamper the application of the model [40, 41]. Although many works have been done to resolve the problems in recent years [42, 43], there is still a long way to obtain an ideal COs containing vascular network, blood brain barrier and immune cells to achieve high-throughput screening in drug development. Thus, the culture method of COs needs to be further optimized to achieve better bioequivalence in the 3D spatial structure and cell tissue composition as compared to human brain.

In summary, the study establishes a humanized ischemic stroke model based on COs, and validates the feasibility and applicability of the model in studying anti-stroke drugs, providing a new research platform for anti-stroke drug development.