Introduction

Stroke continues to be a major global health burden, standing as the primary cause of disability and the second leading cause of death worldwide. Existing treatments for acute ischemic stroke, such as intravenous thrombolysis (IVT) and endovascular therapy (EVT), are predominantly designed to achieve reperfusion of the occluded cerebral vessels. Despite preserving at-risk tissue and minimizing additional functional impairment, these reperfusion treatments offer limited functional recovery by repairing the damaged brain tissue1,2. Moreover, reperfusion has been implicated in secondary cerebral injury, further exacerbating tissue damage3. Ischemia/reperfusion injury is attributed to multiple factors, including the generation of free radicals, inflammation spurred by white blood cell recruitment, cell-matrix breakdown, and microvascular occlusion and edema4,5. Developing therapeutic interventions to promote brain tissue regeneration post-reperfusion is imperative.

Mild hypoxia before or after an ischemic event has been shown to mitigate ischemic damage7. Hypoxic postconditioning (HPC), a novel therapeutic approach, involves subjecting the brain to brief, non-lethal hypoxic episodes following reperfusion8. HPC can be either a single hypoxic exposure or multiple intermittent hypoxic treatments spaced 24 h apart9,10,11,12,13,14. Prior research has established that HPC activates endogenous neuroprotective mechanisms against ischemia/reperfusion injury in both in vitro and in vivo models, evidenced by reduced lesion sizes, enhanced synaptic functionality, and improved long-term outcomes13,15,16. Various signaling pathways, including HIF-1α, P38/MAPK, Wnt/β-catenin, Akt/FoxO, and MEK/ERK, have been implicated in the neuroprotective effects of HPC10,11,17,18,19. Nonetheless, the underlying mechanisms of HPC-induced cerebral protection are not fully understood, especially regarding the role of glial cells, such as astrocytes and microglia20,21,22,23,24.

Astrocytes and microglia are the primary responders in cerebral injury20,22,23. Upon brain damage, astrocytes undergo reactive astrogliosis, characterized by the formation of a glial fibrillary acidic protein (GFAP)-rich glial scar, while activated microglia exhibit enhanced phagocytic activity for debris removal21,24. These activated glial cells have been conventionally categorized into binary phenotypes: pro-inflammatory (A1 astrocytes and M1 microglia) and anti-inflammatory (A2 astrocytes and M2 microglia) states. The former secretes factors detrimental to neuronal functions, while the latter promotes neuroprotection and tissue repair. Astrocytes and microglia regulate each other’s pro-inflammatory states reciprocally through inflammatory cytokines, such as IL-1β, IL-6, IL-10, IL-18, TNF-α, and TGF-β20,25. However, recent research challenges the binary classification of glial cells into pro-inflammatory (A1 astrocytes and M1 microglia) and anti-inflammatory (A2 astrocytes and M2 microglia) phenotypes, suggesting that this oversimplified view fails to capture the complexity and diversity of astrocytic and microglial roles in brain pathophysiology26,27,28.

Few studies have investigated glial reactivity in ischemic models following HPC13,29. While astrocyte and microglia were shown to facilitate HPC-mediated neuroprotection, their reactive states and intercommunications under HPC were not fully explored. A more recent study revealed that HPC reduced the proinflammatory level of immune cells with microglia transformed into the anti-inflammatory state in a mice model of ischemic stroke30. However, this simplified view of glial phenotypes is evolving, as it does not fully represent the spectrum of reactive states astrocytes and microglia can exhibit. A more representative in vitro ischemic model compatible with HPC could provide an accessible and efficient approach to validate astrocyte and microglia reactivity, further illuminating the molecular mechanisms critical for functional recovery post-ischemia.

Our group previously developed a 3D in vitrospheroidal cortical ischemic-hypoxic injury model using primary rat cortical cells, which allowed a more accurate representation of ischemic responses post oxygen-glucose deprivation-reoxygenation (OGD-R) than traditional 2D cultures31. In this study, we explored the neuroprotective effects of HPC in our 3D model, focusing on the modulation of glial cell reactive states and proinflammatory signaling. We hypothesized that altered glial reactivity under HPC is accompanied by the modulation of proinflammatory signaling. To our knowledge, this is the first study employing a 3D spheroidal cortical ischemic-hypoxic injury model to investigate HPC’s neuroprotective effects, shedding light on the roles of activated astrocytes and microglia. We exposed the model to single or multiple intermittent hypoxia treatments and measured outcomes, including cellular apoptosis, neuronal viability, spheroidal nuclear characteristics, stroke biomarker mRNA expressions, and neurite outgrowth. Our findings suggested that a single HPC episode offered the most superior neuroprotection compared to multiple intermittent treatments. Notably, single HPC promoted astrocyte reactivity, as indicated by increased GFAP and Ki67 expression, yet suppressed both A1 and A2 astrocyte markers, hinting at an intermediate astrocyte state. Conversely, despite a general decrease in the microglial activation marker Iba1, an increase in both M1 and M2 microglia markers was observed following OGD-R, with the 1HPC condition uniquely suppressing these altered expressions. This implies that 1HPC is particularly effective in mitigating aberrant microglial activation. Furthermore, HPC treatment was found to attenuate the gene expression of A1 astrocyte inducers such as TNF-α, C1q, and IL-1 within the model, potentially disrupting astrocyte-microglia crosstalk during inflammatory response.

Results

Formation of 3D cortical ischemic-hypoxic injury model for HPC

Spheroid size and circularity are essential markers of cell condition and structural integrity32,33. Under ischemic conditions, spheroids typically shrink due to cell loss and compaction. By assessing these metrics, we aimed to determine whether hypoxic postconditioning (HPC) could reverse these effects by promoting cellular survival and recovery.

To form our 3D cortical ischemic-hypoxic injury model for HPC, primary rat cortical cells were seeded onto a 96-well round-bottom ultralow attachment (ULA) plate and cultured for seven days, aggregating and compacting into structurally mature spheroids. Ischemic injury was then induced in these spheroids by oxygen-glucose deprivation-reoxygenation (OGD-R) to form ischemia-mimicking OGD-R spheroids. OGD-R involved a long-term hypoxia/reperfusion period along with the deprivation of glucose in culture media for four hours to recapitulate the interrupted blood flow and subsequent decrease of oxygen and glucose availability in ischemia. To initiate the endogenous repair mechanisms of spheroids against ischemic injury, one, three, or five cycles of HPC were applied to OGD-R spheroids to obtain 1HPC, 3HPC, and 5HPC spheroids correspondingly (as schematized in Fig. 1a). Unlike OGD-R, HPC placed spheroids to a short-term hypoxia period with the presence of glucose for two hours to stimulate the repair mechanism of cells. The morphological characteristics of the spheroids, specifically their projected areas and circularity, were quantitatively assessed on the day in vitro (DIV) 13.

To quantify the projected area, brightfield images of spheroids were captured, and their contours were outlined for area measurement using ImageJ (Fig. 1c). The projected area of each object within the images was calculated in pixels and then converted to square micrometers using a scale reference. A comparison of the projected areas showed a significant decrease in the size of OGD-R, 3HPC, and 5HPC spheroids relative to the untreated control, with the reduction being consistent across these groups. In contrast, 1HPC spheroids displayed a reversal of this trend, with their projected area restored to a level comparable to the control and significantly larger than those subjected to OGD-R, 3HPC, and 5HPC treatments. This suggests a restorative effect of 1HPC treatment on spheroid integrity.

Using the same spheroid images, circularity was calculated in ImageJ (Fig. 1d) using the following formula: Circularity = (4π × Area) / Perimeter². Circularity quantifies how close an object is to a perfect circle, with a value of 1 representing a perfect circle and values approaching 0 indicating increasing irregularity. Our analysis revealed no significant differences in circularity across the different treatment groups, suggesting that the internal architecture of the spheroids remained relatively consistent regardless of treatment. Preserving circularity across treatments indicates that the overall spherical shape was maintained even as size varied, reflecting different responses to ischemic conditions and subsequent HPC interventions.

Overall, these findings underscore the potential of the 1HPC treatment in promoting recovery after ischemic insult within our 3D cortical ischemic-hypoxic injury model, as evidenced by its ability to restore spheroid size, an important marker of cellular health and structural integrity in this context34.

Fig. 1
figure 1

Progression and Treatment Response of 3D Cortical Spheroids in a Hypoxic Postconditioning (HPC) Model. (a) Schematic representation of the progression of cortical spheroids cultured on a 96-well round-bottom ultralow attachment (ULA) plate from different conditions: control spheroids, spheroids after oxygen-glucose deprivation-reoxygenation (OGD-R), spheroids treated with one cycle of HPC (1HPC), spheroids treated with three cycles of HPC (3HPC), and spheroids treated with five cycles of HPC (5HPC). (b) Representative brightfield images of control (b-a), OGD-R spheroids (b-b), 1HPC spheroids (b-c), 3HPC spheroids (b-d), and 5HPC spheroids (b-e) and their projected area at day in vitro (DIV) 13. Scale bars represent 200 μm. Quantitative analysis of the projected area (c) and circularity (d) of the spheroids at DIV13 is depicted in the box plots. Data are presented as means ± standard deviation (SD) with n = 8–10 independent spheroids per group (*p < 0.05 indicates statistical significance).

Full size image

Analysis of nuclear morphology and density in 3D cortical spheroids Post-HPC treatment

Changes in nuclear morphology and density often reflect stress and cellular compaction following ischemic insult35. The peripheral regions of spheroids may exhibit different responses compared to the core due to differential oxygen and nutrient availability. An in-depth examination of the 3D cortical spheroids was conducted to assess changes in nuclear morphology and density within the defined spheroid sections: the inner core, midbody, and outer shell (depicted in Fig. 2a). The analysis revealed an interesting trend of increasing nuclear area from the inner core towards the outer shell across all treatment conditions (Fig. 2b). Additionally, the nuclear aspect ratio was found to be higher in the outer shell compared to the inner core and midbody, implying that the spheroids experienced peripheral tension, leading to an elongation of nuclei (Fig. 2c). These characteristic features align with what one would expect in a mature 3D spheroidal model, reinforcing the notion that the structural integrity of the spheroids was preserved up to DIV 13, irrespective of the hypoxic treatment received35,36,37.

Furthermore, the nuclear density within the spheroids was quantitatively evaluated, revealing that the OGD-R, 3HPC, and 5HPC treatments resulted in significantly higher nuclear densities than the control and 1HPC spheroids, suggesting a more compact spheroidal architecture in these conditions (Fig. 2d,e). This compaction, while not affecting the overall spheroidal circularity, could impose a steeper gradient for solute diffusion within the spheroid, possibly affecting cell viability, particularly in the innermost regions38. Interestingly, the 1HPC treatment appeared to counteract this trend, reducing compaction and potentially promoting a more favorable environment for cell survival and recovery after ischemic stress. These observations provide valuable insights into the structural and cellular responses of 3D cortical spheroids to different HPC regimens. The data suggest that while OGD-R induces a compacted state within the spheroids, potentially detrimental to cell viability, the 1HPC treatment may alleviate such stress, contributing to a more conducive structure for recovery and repair post-ischemia.

Fig. 2
figure 2

Nuclear Morphology and Density in 3D Cortical Spheroids Across Different HPC Treatments. (a) Illustration of 3D cortical spheroid sections: inner core, midbody, and outer shell. (b,c) Box plots representing the nuclear area and aspect ratio within these regions across control, OGD-R, 1HPC, 3HPC, and 5HPC spheroids. (d) Immunofluorescent DAPI staining showcases nuclear distribution in each treatment group. Scale bars denote 200 μm. (e) Box plots of nuclear density comparisons across all groups. Data are presented as means ± standard deviation (SD) (n = 5, 8–10 spheroids from independent experiments) (*p < 0.05 indicates statistical significance).

Full size image

Effect of HPC on cell survival in a 3D cortical ischemic-hypoxic injury model

In our assessment of the impact of HPC on cell survival within a 3D cortical ischemic-hypoxic injury model, we first employed a live/dead assay to evaluate cellular viability following OGD-R and subsequent HPC treatment (Fig. 3a, c-d). Cell death via apoptosis is a major consequence of ischemic injury and contributes to neurodegeneration in stroke. We investigated whether HPC could mitigate ischemia-induced apoptosis in spheroids by measuring cell survival and apoptotic marker expression.

Our results showed that spheroids exposed to OGD-R exhibited significantly reduced calcein-AM fluorescent intensity compared to untreated controls, indicating reduced live cells (Fig. 3c). Remarkably, spheroids receiving a single HPC cycle (1HPC) and multiple HPC cycles (3HPC and 5HPC) showed significantly higher calcein-AM fluorescent intensity relative to both control and OGD-R spheroids. This contrast suggests that HPC might confer a protective effect against OGD-R by promoting cell survival. Aligning with the reduced calcein-AM fluorescent intensity in OGD-R spheroids, OGD-R also led to significantly higher number of dead cells compared to control (Fig. 3d). In contrast, only 5HPC significantly reduced the number of dead cells, implying suppressed cell death at increasing cycle of HPC.

To gain a more comprehensive insight on the modulation of cell survival, we also investigate the apoptotic activity within spheroids via immunostaining of apoptotic marker, cleaved caspase-3 (Fig. 3b,e). Cleaved caspase-3 expression was significantly upregulated in OGD-R spheroids compared to controls (Fig. 3e), indicating an increase in programmed cell death and confirming the successful induction of ischemic-hypoxic injury. To our surprise, multiple HPC cycles maintained a significantly higher cleaved caspase-3 expression in comparison to the control similar to the OGD-R condition, whereas only single HPC cycle reduced the cleaved caspase-3 expression. This sustained apoptotic activity by multiple HPC cycles post-OGD-R could offer to a certain extent neuroprotective effect by dismantling unhealthy, redundant cells within the spheroids in an organized way that minimizes damage and disruption to neighboring cells. In addition, it may also explain the reduced number of dead cells in the spheroids as increased cleaved caspase-3 expression could be an indication of the early phase of apoptosis, signifying upcoming cell death.

Fig. 3
figure 3

Assessment of Cell Survival in 3D Cortical Spheroids After OGD-R and HPC Treatments. (a) Live/dead assay show Live (green) and dead (red) cells in control, OGD-R spheroids, and HPC spheroids with enlarged view of dead cells in the corresponding spheroids. Scale bars denote 200 μm and 50 μm for normal and enlarged view respectively. (b) Immunofluorescence images display DAPI (blue) and cleaved caspase-3 (green) staining in the respective spheroids with enlarged view of cleaved caspase-3. Scale bars measure 200 μm and 50 μm for normal and enlarged view respectively. (c) Box plot analysis of the fluorescent intensity of calcein-AM (green) in control and treated spheroids. (d) Box plot analysis of the fluorescent intensity of EthD-1 expression across all groups. (e) Box plot analysis of the fluorescent intensity of cleaved caspase-3 expression across all groups. Data represent means ± standard deviation (SD) (n = 8–10 spheroids per group) (*p < 0.05 indicates statistical significance).

Full size image

Effect of HPC on neuronal phenotypes in a 3D cortical ischemic-hypoxic injury model

Neurons are particularly vulnerable to ischemic injury, often leading to the loss of mature neuronal markers like NeuN and disruption of dendritic structures, as indicated by MAP2 staining. This experiment aims to evaluate whether HPC can preserve neuronal stability and synaptic structures, which are critical for post-stroke recovery. As shown in Fig. 3a-b, NeuN expression was significantly reduced in OGD-R and 5HPC spheroids in comparison to the untreated controls, whereas 1HPC and 3HPC maintained NeuN expression of spheroids (Fig. 3b). These findings suggest that 1HPC and 3HPC treatments may better support the stability of mature neurons following ischemic-hypoxic injury.

In addition to NeuN, the formation and maintenance of dendritic structures within the spheroids were also examined using microtubule cytoskeleton regulator, MAP2. OGD-R led to a significantly lower MAP2 fluorescent intensity relative to the control, suggesting the disruption of dendritic structure under ischemic-hypoxic injury. However, spheroids treated with multiple HPC cycles exhibited MAP2 fluorescence intensity comparable to controls at 3HPC, and even significantly higher than controls at 5HPC. Nevertheless, single HPC cycle showed no effect on the reduced MAP2 fluorescent intensity in OGD-R spheroids. Collectively, these results suggested that HPC treatments differentially regulated neuronal phenotypes following OGD-R, in which 1HPC and 3HPC better promoted the stability of mature neurons upon hypoxic-ischemic injury whereas 5HPC treatment primarily promote the development of dendritic structure that is essential to the formation of synaptic connection.

We also assessed the expression of crucial stroke-related protein biomarkers—S100b, IL-1β, and myelin basic protein (MBP)—at the mRNA level to substantiate the occurrence of ischemic-hypoxic injury by OGD-R and subsequent mitigation of damage by HPC (Fig. 3e). These markers were previously used to capture various aspects of ischemic injury associated with astrocyte activation, inflammatory response, and demyelination that could culminate in neuronal damage when overexpressed31. OGD-R spheroids demonstrated a notable upregulation of IL-1β expression when compared to controls, with modest increases in MBP and S100b expressions. In contrast, 1HPC spheroids showed a marked reduction in the expression of all stroke-related protein biomarkers. In the context of ischemic injury, the reduction in MBP levels is often indicative of demyelination, a process where the protective myelin sheath surrounding axons is damaged. MBP is a critical component of the myelin sheath, and its reduction suggests that the ischemia-reperfusion injury in our spheroid model likely leads to demyelination, consistent with clinical observations of stroke-induced white matter damage39. This demyelination could impair neural signal transmission, leading to further functional deficits and neuronal damage40,41. The observed reduction in MBP in the OGD-R group highlights the susceptibility of myelin-producing cells to ischemic injury. In contrast, the 1HPC treatment group displayed a less pronounced reduction in MBP expression, suggesting that HPC may mitigate demyelination. This protective effect could be mediated through HPC’s influence on glial cell activity, promoting remyelination or preserving existing myelin. Future studies could explore whether HPC modulates oligodendrocyte function or supports the survival of myelin-producing cells in the context of ischemic injury. Although reduced expressions of these markers were also observed in 3HPC and 5HPC spheroids relative to OGD-R spheroids, these alterations did not reach statistical significance. Taken together, these data suggest that single HPC and multiple HPC cycles differentially regulate the stability of mature neurons and their dendrite development in response to ischemic injury. This distinctive responds potentially account for the respective expression of stroke-related protein markers, in which single HPC cycle may drive an optimal balance between neural stability and dendrite formation, leading to significantly suppression of marker expression.

Fig. 4
figure 4

Assessment of Neuronal phenotypes in 3D Cortical Spheroids After OGD-R and HPC Treatments. (a) Immunofluorescent staining of NeuN (red) and DAPI (blue) in control, OGD-R, and HPC spheroids with enlarged view of NeuN. Scale bars denote 200 μm and 50 μm for normal and enlarged view respectively. (b) Box plot analysis of the fluorescent intensity of NeuN in control and treated spheroids. (c) Immunofluorescence images display DAPI (blue) and MAP2 (green) staining in the respective spheroids with enlarged view of MAP2. Scale bars measure 200 μm and 50 μm for normal and enlarged view respectively. (d) Box plot analysis of the fluorescent intensity MAP2 expression across all groups. (e) Comparative mRNA expression level of IL-1β, MBP, and S100B in spheroids under different treatments. Data represent means ± standard deviation (SD) (n = 8–10 spheroids per group) (*p < 0.05 indicates statistical significance).

Full size image

Effect of HPC on neuronal migration and neurite outgrowth in a 3D cortical ischemic-hypoxic injury model

Neuronal migration and neurite outgrowth are essential processes for brain repair following ischemia, supporting the reconstruction of neuronal circuits. We evaluated the influence of HPC on neuronal migration and neurite outgrowth using our 3D cortical ischemic-hypoxic injury model compromised by OGD-R conditioning. On DIV 13, spheroids of distinct conditions were placed onto poly-D-lysine (PDL)-coated plates to facilitate cell adhesion and process extension, integral for neuronal circuit reconstruction post-brain injury (Fig. 4a). Over the ensuing three days, neuronal and astrocytic processes began to spread from the spheroid core onto the substrate.

In our 2D spreading assay shown in Fig. 5, control and 1HPC spheroids adhered to the PDL plates rapidly, displaying a notable increase in spreading area, indicative of active neurite and astrocyte process extension. This was in stark contrast to the OGD-R, 3HPC, and 5HPC spheroids, which exhibited limited spreading capacity (Fig. 5b,c). The spreading dynamics did not differ significantly between 3HPC and 5HPC groups. Immunofluorescent staining for β-III-tubulin and GFAP highlighted this differential spreading, with control and 1HPC spheroids demonstrating extensive neuronal and astrocytic process outgrowth while the OGD-R, 3HPC, and 5HPC spheroids showed restricted outgrowth (Fig. 5d).

These observations underscore that 1HPC treatment preserves the capacity for neuronal and astrocytic extension post-OGD-R insult, which may have important implications for the repair and regeneration of neuronal circuitry in vivo42. The 1HPC regimen appears to support cellular mechanisms underlying structural plasticity and recovery, suggesting its potential therapeutic value in the context of cerebral ischemia.

Fig. 5
figure 5

Dynamics of Neuronal and Astrocytic Spreading Post-Hypoxic Postconditioning. (a) Schematic sequence of the 2D spreading assay post-OGD-R and various HPC treatments. 3D cortical spheroids were placed on PDL-coated glass slides on DIV 13 to monitor neurite and astrocyte process extension over three days. (b) Sequential images capturing the spreading of control, OGD-R, and HPC-treated spheroids across time points. (c) Time-series plot of the normalized spreading area with statistical significance markers against OGD-R spheroids and control. Data are represented as mean ± SD (n = 8–10 spheroids per condition) (*p < 0.05). (d) Immunofluorescence staining displays β-III-tubulin (red) for neuronal processes and GFAP (green) for astrocytic processes in each group with enlarged view. Scale bars denote 200 μm.

Full size image

Impact of HPC on astrocyte reactivity in a 3D cortical ischemic-hypoxic injury model

Astrocytes play a crucial role in the brain’s injury response to ischemia, often entering reactive states marked by increased GFAP expression and proliferation (Ki67). This experiment evaluates the modulation of astrocyte reactivity by HPC, which could contribute to neuroprotection. As shown in Fig. 6a-d, astrocytes in spheroids subjected to OGD-R, 1HPC, and 5HPC exhibited a significant escalation in GFAP expression compared to the untreated control. Among these, 1HPC spheroids demonstrated the highest GFAP levels, suggesting a more pronounced state of astrocyte activation. Proliferation, as indicated by Ki67 staining, followed a similar pattern, with OGD-R, 1HPC, 3HPC, and 5HPC spheroids all showing increased levels of Ki67 compared to controls. While the spatial expressions of GFAP and Ki-67 were not identical, both markers exhibited a similar dosage-dependent trend (i.e., stronger in 1HPC compared to 3HPC or 5HPC), and their peripheral expression was more pronounced than in the inner core. This pattern of peripheral localization in both GFAP and Ki67 suggests that cells in the outer shell of the spheroids are more reactive and proliferative. The side-by-side comparison of these markers further supports the idea that cell-level co-localization—indicating astrocytes expressing GFAP are also proliferating, as shown by Ki-67—is evident. This finding reinforces the correlation between astrocyte reactivity and proliferation at a cellular level in response to HPC treatments.

In line with this, a gradient in nuclear morphology was observed across all HPC conditions, where the nuclear area increased from the inner core to the outer shell of the spheroids (Fig. 2b). This pattern suggests that cells in the outer shell experience less compression, which may promote proliferative activity, as indicated by the rise in Ki67 expression in these regions. The correlation between nuclear morphology and proliferation supports the idea that reduced compression in the outer shell could drive increased cellular division, particularly in astrocytes34.

To further clarify the astrocyte activation state post-HPC, we quantified mRNA expressions of the A1 astrocyte-specific marker C3 and the A2 astrocyte-specific marker S100a10 (Fig. 6e)43. Following OGD-R, cortical spheroids exhibited a surge in C3 mRNA expression, indicative of an A1 pro-inflammatory response. However, HPC treatments appeared to mitigate this response, with a noticeable reduction in C3 expression, particularly in the 5HPC group. Similarly, S100a10 expression, a marker for the A2 repair-associated phenotype, was lower across OGD-R, 1HPC, 3HPC, and 5HPC spheroids compared to controls, with 1HPC spheroids showing a significant decrease, suggesting a suppression of the A2 reactivity.

Collectively, these results suggest that HPC, particularly the 1HPC treatment, modulates astrocyte activation, potentially shifting away from both the classical pro-inflammatory A1 and A2 phenotypes. This modulation may indicate a unique and less-differentiated state of astrocyte activation. This shift could play a critical role in the neuroprotective mechanism of HPC following cerebral ischemia.

Fig. 6
figure 6

HPC-Induced Alterations in Astrocyte Reactivity and Proliferation in a Cortical Ischemic-hypoxic injury model. (a) Immunofluorescence images depicting GFAP expression (green) and nuclei (DAPI, blue) in control and treated spheroids with enlarged view of GFAP. Scale bars denote 200 μm and 50 μm for normal and enlarged view respectively. (b) Immunofluorescence for Ki67 (green) indicates proliferative activity against a backdrop of nuclei (DAPI, blue). (c) Quantitative analysis of GFAP fluorescence intensity. (d) Analysis of Ki67 fluorescence intensity. (e) The bar graph shows the relative mRNA expression levels of astrocytic markers s100a10 and C3 across all groups. Scale bars represent 200 μm. Data are expressed as means ± standard deviation (SD) (n = 5, 8–10 spheroids per condition) (*p < 0.05 indicates statistical significance).

Full size image

Influence of HPC on microglial activation and inflammatory signaling

Microglia are key players in the brain’s inflammatory response to ischemia, and their activation can either exacerbate or mitigate damage. This segment of our study assessed the effects of HPC on microglial activation and their potential inflammatory signaling with astrocytes in a 3D cortical ischemic-hypoxic injury model. Utilizing Iba1 immunostaining to evaluate microglial reactivity, we observed that control spheroids unexpectedly exhibited a significantly higher expression of Iba1 compared to OGDR and all HPC-treated conditions, implying potential suppression of microglial activation following OGD-R and HPC treatments (Fig. 7a-b). Nevertheless, it is also plausible that the diminished Iba1 expression indicates a reduction in the overall microglial population within the spheroids after the treatments.

To further validate the reactive states of microglia, we quantified the mRNA expression levels of the M1 phenotype marker CD40 and the M2 phenotype marker CD206 (Fig. 7c)44. An increase in CD40 mRNA expression was evident in the spheroids following OGD-R, indicative of an M1 microglial activation. This expression was intensified in the 3HPC and 5HPC conditions compared to OGD-R alone. In contrast, the 1HPC treatment modestly subdued the CD40 expression. CD206 marker expression of spheroids exhibited a similar trend, with OGD-R spheroids displaying elevated mRNA levels relative to the control, which were slightly reduced in the 1HPC treatment, suggesting a subtle decrease in the M2 microglial activation.

Inflammatory signaling between microglia and astrocytes plays a pivotal role in the progression of ischemic injury by modulating their corresponding reactive states29,45. To acquire a glimpse of how this interaction is potentially altered in our ischemic-hypoxic injury model following HPC, we examined the whole spheroid mRNA expressions of TNF-α, C1qa, C1qb, and IL-1α (Fig. 7d), cytokines secreted by microglia that were shown to be essential in promoting the A1 astrocytic phenotype29,45. However, it is important to note that the mRNA expression of whole spheroid does necessarily reflect specific cell-cell interactions between microglia and astrocyte. The observed changes in microglial activity could also be influenced by the interactions with neurons within spheroids. Future studies should incorporate cell sorting techniques to isolate specific cell populations for a more accurate analysis. Analysis of TNF-α mRNA expression showed that OGD-R led to significantly reduction of TNF-α mRNA expression in the spheroids relative to the control. This reduced expression was further lowered by HPC treatments with no significant difference among different HPC cycles. Interestingly, unlike TNF-α mRNA expression, while OGD-R elevated the mRNA expressions of C1qa, C1qb, and IL-1α, HPC treatments demonstrated diverse regulatory effects. 3HPC increased the mRNA expression levels of these inflammatory cytokines relative to OGD-R, whereas the 1HPC and 5HPC treatments notably reduced their expressions. These findings could imply that while OGD-R predisposes microglia towards a pro-inflammatory state, HPC, especially 1HPC, seems to curb this activation. The most marked reduction in pro-inflammatory cytokine expression with 1HPC suggests its potential in mitigating the communication that fosters a pro-inflammatory environment, thereby exerting a neuroprotective influence by dampening the inflammatory response.

Fig. 7
figure 7

Analysis of Microglial Activation and Crosstalk with Astrocytes Post-HPC. (a) Immunofluorescence visualization of microglial marker Iba1 (green) against the nuclear stain DAPI (blue) in control and treated spheroids. Scale bars are set at 200 μm. (b) Box plot depicting the relative fluorescent intensity of Iba1 across different groups. (c) Bar graphs representing the mRNA expression levels of microglial activation markers CD40 and CD206 in the spheroids. (d) Quantitative mRNA expression of pro-inflammatory cytokines TNF-α, C1qa, C1qb, and IL-1α involved in microglia-astrocyte signaling. Data are presented as mean ± SD (n = 8–10 spheroids per condition) (*p < 0.05 signifies statistical significance).

Full size image

Discussions

In this study, we investigated the neuroprotective effects of HPC in a 3D cortical ischemic-hypoxic injury model, focusing on the reactivity and inflammatory signaling of glial cells, specifically astrocytes and microglia. Our findings indicated that a single HPC (1HPC) treatment confers the most substantial protection against ischemic injury induced by oxygen-glucose deprivation-reoxygenation (OGD-R). This was evidenced by the improved structural organization, enhanced cell survival and reduced apoptosis, optimal modulation of neuronal phenotypes, dampened ischemic responses, and augmented neurite outgrowth of spheroids. Additionally, we observed that HPC encourages activation of astrocytes while concurrently inhibiting both A1 and A2 phenotypes. It also led to the inactivation of microglia with a reduced M1 and M2 phenotypes. Notably, proinflammatory cytokines essential to microglia-astrocyte signaling during neuroinflammation was diminished following HPC treatment.

Neuroprotection via apoptosis inhibition

A fundamental neuroprotective mechanism activated by HPC is the inhibition of apoptosis. Upton the occurrence of cerebral ischemia, the reduction in blood flow induces widespread apoptosis in the peri-infarct region46. Furthermore, reperfusion triggers the overexpression of reactive oxygen species in mitochondria, further resulting in DNA damage and apoptosis47. These enhanced apoptosis has long been shown to contribute to the neuronal cell death in ischemic injury. Therefore, inhibition of the apoptotic protein, mainly caspases, was believed to recover the neuronal loss and facilitate neuroprotection. This theory is supported by experimental studies showing that caspase inhibition with the pan-caspase inhibitor z-VAD provides neuroprotection in both transient and permanent models of cerebral ischemia in rats and mice48. Similarly, administration of the caspase inhibitor zDEVD-afc was found to reduce ischemic damage in a murine model of middle cerebral artery occlusion (MCAO)49.

In contrast to the use of anti-apoptotic pharmacological agents, HPC confers anti-apoptotic effects by inducing endogenous repair mechanisms of cells. Previous research has shown that HPC facilitates the recovery of brain function in rats with ischemic-hypoxic brain damage, potentially through the inhibition of neuronal apoptosis via the activation of HIF-1α signaling12. More recently, the reduction of apoptotic protein and subsequent induction of neuroprotective effect by HPC in a transient global cerebral ischemia model was shown to be mediated by the suppression of RNA-binding protein Piwil2 expression and CREB2 promoter methylation50. In the present study, we showed that a single HPC treatment significantly reduced cleaved caspase-3 expression, suggesting that this treatment may attenuate apoptotic cell death more effectively, a critical factor in neuronal loss post-ischemia. A potential neuroprotective mechanism activated by HPC is the inhibition of apoptosis in the ischemic injury zone. In the pathogenesis of ischemic stroke, activation of apoptosis during reperfusion is a significant problem. Therefore, the decrease in the level of activated caspase-3 noted in our study should be discussed in this context. The reduction of activated caspase-3 implies that HPC may mitigate apoptosis, a crucial contributor to neuronal death post-ischemia. This anti-apoptotic effect highlights an additional layer of neuroprotection provided by HPC alongside its modulation of glial cell activity.

Neuronal phenotypes and structural stability

In addition to reducing apoptosis, we evaluated the effects of HPC on neuronal stability through the expression of NeuN and MAP2, key markers of neuronal viability and dendritic structure. NeuN, a neuron-specific nuclear protein, is widely recognized as a post-mitotic marker51. Many studies have associated NeuN with the regulation of neural tissue development and adult brain function and showed that NeuN mutation leads to numerous neurological disorders52. On the other hand, MAP2 is a cytoskeleton protein enriched in dendrite. MAP2 expression is very weak in neuronal precursors but increases as the dendrites of neurons become more mature53. MAP2 facilitates dendrite development and stabilization, which are essential to synapse formation and neuron-neuron communication. Thus, both NeuN and MAP2, which are crucial for brain function, have been shown to be highly sensitive to ischemic injury. Upon the onset of cerebral ischemia, the immunoreactivity of NeuN and MAP2 rapidly decreases, indicating potential neuronal damage and cytoskeletal breakdown, respectively54,55.

Our study demonstrated that OGD-R treatment significantly reduced NeuN and MAP2 expression in cortical spheroids. This replication of in vivo neuro-phenotypic responses to ischemic injury validates our ischemic-hypoxic injury model as a representative in vitro ischemic model. We further observed that a single HPC cycle restored NeuN expression in spheroids to levels comparable to controls without affecting MAP2 expression, suggesting the preferential effect to maintain the healthy state of neurons. In contrast, increasing HPC cycles, especially 5HPC treatment, enhanced MAP2 expression and suppressed NeuN expression, reinforcing the synaptic connectivity and neural circuit stability. In addition to these neuronal markers, myelin integrity, as indicated by MBP levels, is also critical in the context of ischemic injury. Our findings of reduced MBP levels in the OGD-R spheroids provide evidence of ischemia-induced demyelination, a hallmark of white matter injury in stroke. Demyelination compromises axonal integrity and signal transmission, contributing to long-term neurological deficits. The significant reduction of MBP we observed suggests that ischemia-reperfusion injury disrupts myelin maintenance. Notably, the modest preservation of MBP levels in the 1HPC-treated spheroids suggests that HPC may offer protection against demyelination, potentially by modulating the inflammatory response, which is known to contribute to myelin breakdown. Pro-inflammatory cytokines produced by reactive microglia are implicated in this process. By reducing these cytokines, HPC may help preserve myelin integrity or promote remyelination. Further investigation is needed to clarify the specific mechanisms by which HPC influences myelination and whether this contributes to functional recovery in ischemic injury models.

While our assay on stroke protein biomarkers, neurite outgrowth, and astrocyte and microglia activation indicated that single HPC cycles better suppressed ischemic responses and promoted neural regeneration, maintaining neurons in a healthy state seemed to be an essential criterion for neuroprotection. The upregulated MAP2 expression following 5HPC could be a compensatory response to support the neural circuitry in an effort to alleviate the aggravating neuronal defect.

Glial reactivity and inflammatory modulation

Astrocyte and microglial reactivity play a pivotal role in the response to ischemic injury. Increased GFAP immunoreactivity and astrocyte proliferation post-HPC have been associated with improved motor and sensory functions in adult rats13, suggesting that HPC may foster astrocyte functions critical for repairing and restoring the brain post-injury. Concurrently, microglial inactivation correlates with reduced brain tissue loss, implying that inactivation may play a role in HPC-induced neuroprotection29.

However, the precise reactive states of astrocytes and microglia—whether A1 or A2 astrocytes and M1 or M2 microglia—under HPC and the molecular mechanisms driving these states are not yet fully elucidated. Our study offers insights into these reactive states and the pro-inflammatory cytokines essential to astrocyte-microglia crosstalk, potentially determining their reactivity. We found that HPC suppresses the manifestation of both pro-inflammatory (A1) and anti-inflammatory (A2) astrocyte phenotypes in our 3D cortical ischemic-hypoxic injury model, regardless of their activation state, indicating the induction of a currently undefined reactive state.

Additionally, reduced Iba1 immunostaining, along with inhibited CD40 and CD206 mRNA expression after 1HPC, suggests microglial inactivation. This finding contrasts with the results by Zhang et al., which showed a decrease of CD16+ M1 microglia and an increase of CD206+ M2 microglia following HPC in adult male mice subjected to middle cerebral artery occlusion. The absence of vascular structure and tissue architecture in our 3D cortical ischemic-hypoxic injury model may account for the observed discrepancy.

It is important to note that Iba1 immunostaining was used to identify and quantify microglia within the spheroids. In contrast, cytokine expression analysis was performed using total mRNA extracted from the entire spheroid. Although this approach provides a general overview of the cytokine environment, it may not accurately reflect the specific cytokine profile of microglia due to the presence of other cell types in the spheroid. Furthermore, this approach does not allow for a detailed examination of cell-cell interactions, specifically between microglia and astrocytes. The mRNA expression data reflect the combined contributions of all cell types within the spheroid, limiting our ability to specifically analyze interactions between microglia and astrocytes.

Moreover, the observed changes in microglial activity could be influenced by interactions with other cell types, including neurons, within the spheroid. In the context of ischemic injury and neuroprotection, we recognize the importance of understanding the crosstalk between different cell types, particularly between microglia, astrocytes, and neurons. Future studies implementing cell-type-specific analyses, such as sorting Iba1-positive cells prior to cytokine quantification, will enhance our ability to dissect the complex cellular interactions and signaling pathways involved.

Lastly, the downregulation of pro-inflammatory factors, which are crucial to M1 microglia-A1 astrocyte signaling, suggests that HPC may disrupt pro-inflammatory signaling pathways between microglia and astrocytes, potentially inhibiting their pro-inflammatory phenotypes and contributing to neuroprotection against ischemic/reperfusion injury. Previous studies have established various HPC protocols in different ischemic-hypoxic injury models, demonstrating the versatility of HPC’s neuroprotective benefits. For instance, Leconte et al. reported the neuroprotective effects of single hypoxia in 2D neocortical cultures using 0.1% and 1% O2 post-OGD-R14. Zhan et al. uncovered multiple signaling pathways that mediate brain protection by single hypoxia using 8% O2 in a rat craniectomy model9. Contrastingly, repetitive intermittent hypoxia has been shown to provide neuroprotection by stimulating the pentose phosphate pathway in rats exposed to severe hypobaric hypoxia11and by enhancing astrocyte function and reducing microglia reactivity in models of neonatal hypoxia-ischemic brain injury13,29. These studies emphasize the dependency of HPC’s neuroprotective benefits on the ischemic-hypoxic injury model and the ischemia-inducing mechanisms.

The superiority of single HPC treatment

Overall, our findings support the superiority of single HPC (1HPC) over multiple cycles of HPC in neuroprotection. The significant reduction of ischemic markers enhanced neuronal viability, and suppression of pro-inflammatory cytokines after 1HPC highlight its potential to promote brain regeneration. This outcome aligns with studies showing that single hypoxia regimens offer significant neuroprotection in both in vitro and in vivo models. Our extended recovery period (up to DIV13) may have contributed to the observed enhanced recovery, differentiating our findings from prior research where samples were collected within 24 h post-HPC.

Conclusion

Our study demonstrates that HPC can enhance neuroprotection and modulate the reactivity of astrocytes and microglia in a 3D cortical ischemic hypoxic injury model. HPC improved structural organization, enhanced cell survival, reduced apoptosis, optimal modulation of neuronal phenotypes, mitigated ischemic responses, and promoted neurite outgrowth, which may collectively aid functional recovery. The differential modulation of glial cell reactivity by HPC—promoting astrocyte activation while inhibiting A1 and A2 astrocyte phenotypes and inactivating microglia with reduced M1 phenotype—underscores the complexity of glial cell dynamics post-injury. The potential suppression of pro-inflammatory signaling from microglia to astrocytes further suggests that the inhibition of pro-inflammatory glial phenotypes is instrumental in mediating HPC-induced neuroprotection. Moreover, the observed reduction in activated caspase-3 levels points to an anti-apoptotic mechanism of action by HPC, which is critical in counteracting reperfusion-induced neuronal death. While additional research is required to decipher the molecular mechanisms governing HPC’s regulation of glial cell reactivity and communication, our findings affirm the utility of the 3D cortical ischemic-hypoxic injury model as a platform for further studies on HPC mechanisms, with potential applications extending to brain regeneration, drug screening, and the development of disease models.

Materials and methods

All methods were performed according to the relevant guidelines and regulations.

Isolation of prenatal rat cortical cells

Primary cortices were obtained from postnatal day 1–2 Sprague-Dawley rat pups (DBL, South Korea). The ARRIVE guidelines were implemented in all experimental protocols, and the current study is reported accordingly. Cortical cells were dissociated following the protocols developed previously. In brief, the isolated cortical tissues were dissected into small fragments and placed in papain solution (Worthington, US) for 10 min at 37℃. Following the removal of papain solution, fragmented cortical tissues were triturated with 1000, 100, and 10 µl pipette tips by pipetting them up and down ten times each sequentially in Dulbecco’s Modified Eagle’s Medium (DMEM; WELGENE, South Korea), supplemented with 10% fetal bovine serum (FBS; Gibco, UK) and 1% penicillin-streptomycin (Gibco, UK). The remaining debris was removed by filtering the cell solution through a 22 μm cell strainer. The filtered cell solution was centrifuged at 1500 rpm for 3 min, and the supernatant was removed. The cortical cell pellet was then resuspended and maintained in neural basal medium-A (Gibco, UK), supplemented with 2% B-27 supplement (Gibco, UK), 1% GlutaMAXTM (Gibco, UK), and 1% penicillin-streptomycin (Gibco, UK).

3D cortical spheroid formation

Cortical spheroids were formed using a 96-well round-bottom ULA plate (Corning, US) and an agarose plate. To fabricate the round-bottom agarose plate, 50 µl of molten 3% agarose (Young Sciences, South Korea) solution was poured onto each well of the 96-well cell culture plate (SPL, Korea). The agarose plate was equilibrated with the culture medium for at least three hours in a humidified incubator. Prior to seeding the cells, the medium was removed from the agarose plate. The dissociated cortical cell suspension was seeded onto the ULA and agarose plates with 3 × 105 cells/well. The cultures were then incubated at 37℃ and 5% CO2 for 7 days. Half of the culture medium was replaced with a fresh medium every 3–4 days. The dissociated cortical cell suspension was seeded onto the 6-well transparent flat bottom ULA plate (Corning, US) and agarose plate with 3 × 105 cells/ml to closely observe the cell-substrate interaction during cortical spheroid formation. In this case, 1 ml of molten 3% agarose solution was poured onto each well of the 6-well cell culture plate (SPL, Korea) to fabricate the agarose plate.

Ischemic-hypoxic injury induction of 3D cortical spheroid by oxygen-glucose deprivation reoxygenation (OGD-R)

On DIV 07, cortical spheroids cultured on ULA and agarose plates were subjected to oxygen-glucose deprivation (OGD) to mimic stroke. For this purpose, cortical spheroids were maintained in a culture medium lacking glucose and B-27 supplement and exposed to hypoxia (1% O2) in the hypoxic chamber for four hours. Upon the completion of OGD, cortical spheroids underwent reoxygenation to attain OGD-R. In reoxygenation, the culture was maintained in a standard culture medium and incubated under optimal incubation conditions (37ºC and 5% CO2) for 24 h.

HPC of the 3D cortical spheroid

After reperfusion (DIV08), cortical spheroids were subjected to single hypoxia (1HPC) or repeated intermittent hypoxia for three days (3HPC) or five days (5HPC). In 1HPC, cortical spheroids were exposed to hypoxia (1% O2) in the hypoxic chamber for two hours while maintained in a standard culture medium. The samples were returned to the optimal incubation conditions following hypoxia. Similarly, cortical spheroids were exposed to hypoxia (1% O2) in the hypoxic chamber for two hours for 3 or 5 days, spaced at 24-hour intervals for 3HPC and 5HPC, respectively. All samples were collected on DIV 13.

Antibodies

For immunostaining of NeuN, class III β-tubulin, GFAP, integrin α5, fibronectin, Ki-67, Iba1, and Olig2, we used purified anti-NeuN antibody [EPR12763] (Abcam, ab177487), anti-β-tubulin (Sigma, T8578), GFAP (D1F4Q) XP® (Cell Signaling TECHNOLOGY®, 12389), anti-Integrin alpha five antibodies [EPR7854] (Abcam, ab150361), anti-fibronectin antibody (Abcam, ab23750, anti-Ki67 antibody (ab15580), anti-Iba1 antibody (FUJIFILM, 019-19741), and anti-Olig2 antibody [EPR2673] (ab109186) respectively.

Whole spheroid fixation and immunostaining

The cortical spheroids were fixed in 4% v/v paraformaldehyde and 8% w/v sucrose in phosphate-buffered saline (PBS) overnight at 4℃, followed by one h PBS wash three times. The fixed spheroids were permeabilized and blocked with 1% v/v Triton X-100 and 4% w/v bovine serum albumin (BSA) in PBS for two hours at room temperature and washed for one hour with PBS 3 times. Primary antibodies (1:200 neuron-specific class III β-tubulin and 1:200 GFAP in PBS) were incubated with cortical spheroids overnight at 4℃. After incubation, cortical spheroids were washed twice for two hours in PBS. Subsequently, the spheroids were incubated with secondary antibodies (Alexa Fluor 488 or 594; Invitrogen, US) overnight at 4℃, followed by another two hours of washing in PBS twice. Counterstaining staining of cortical spheroids was performed with DAPI for one hour at room temperature.

Cryosectioning of cortical spheroids and immunostaining

Cortical spheroids were fixed with 4% v/v paraformaldehyde and 8% w/v sucrose in PBS overnight and washed in PBS for one hour three times. The spheroids were then immersed sequentially in 15% and 30% sucrose in PBS for three hours per step at room temperature. Samples were embedded in the Tissue-Tek® O.C.T.TM compound (SAKURA, US) and stored at -80℃. Frozen samples were sectioned into approximately 20 μm-thick slices using Lecia CM3050 S Cryostat (Lecia, Germany) and placed on Histobond adhesive glass slides (Marienfeld, Germany). Spheroid sections on the adhesive glass slides were dried at room temperature for 30 min and washed twice with PBS. Dried sections were permeabilized with 0.1% Triton X-100 in PBS for 15 min, followed by washing in PBS twice. Samples were then blocked in 3% BSA solution with 0.1% Triton X-100 for one hour and washed twice with PBS. Incubation of samples with primary antibodies was performed overnight at 4℃. Following two washing steps in PBS for 30 min, samples were stained with secondary antibodies (Alexa Fluor 488 or 594) for one hour at room temperature. The sections were washed twice with PBS for 5 min and incubated in a DAPI solution for 5 min.

RNA isolation and quantitative polymerase chain reaction (qPCR) analysis

Cortical spheroids were lysed with RNAiso Plus reagent (TaKaRa, Japan), and total RNA was extracted according to the manufacturer’s instructions. In brief, the lysed samples were incubated with chloroform, followed by centrifugation (12,000 g, 15 min, and 4℃). The resulting supernatant was further incubated with isopropanol and centrifuged (12 000 g, 15 min, and 4℃). After removing the supernatant, the pellet was cleaned with 75% cold ethanol and centrifuged (7 500 g, 5 min, and 4℃). Extracted RNA pellets were resuspended in ultrapure water (WELGENE, South Korea), and the concentrations were measured using a NanoDrop 1000 Spectrophotometer (ThermoScientific, USA). Reverse transcription of RNA into complementary DNA (cDNA) was performed using the iScript cDNA synthesis kit (Bio-Rad, US). qPCR analysis was then performed with primers and SYBR Green Supermix (Bio-Rad, US) using CFX96TM Real-Time PCR Detection System (Biorad, US). The fold expression was calculated by the 2 − ΔΔCτ formula. The gene expression level was normalized to the housekeeping gene, glyceraldehyde-3-phosphate-dehydrogenase (GAPDH). The sequences of the primers used in this study are shown in Table 1.

Table 1 Primers for qPCR.
Full size table

2D spreading assay

On DIV13, cortical spheroids completing HPC were collected and plated on the PDL-coated 35 mm glass-bottom dishes. The Projected area expansion was quantified using ImageJ software.

Statistical analysis

All quantitative measurements were presented as means ± standard deviations. Statistical significance was determined by one-way ANOVA using the statistical software IBM SPSS Statistics 25 (SPSS, US). P-values below 0.05 were considered statistically significant.