Assessment of Cellular End of Life Management in Acute Brain Injury: Clearing Dying Cells After Injury

Abstract

A Dynamic Balance Between Neuronal Death and Clearance in an In Vitro Model of Acute Brain Injury

Balena T, Lillis K, Rahmati N, Bahari F, Dzhala V, Berdichevsky E, Staley K. J. Neurosci. 2023;43(34):6084–6107. doi:10.1523/JNEUROSCI.0436-23.2023.

In in vitro models of acute brain injury, neuronal death may overwhelm the capacity for microglial phagocytosis, creating a queue of dying neurons awaiting clearance. Neurons undergoing programmed cell death are in this queue and are the most visible and frequently quantified measure of neuronal death after injury. However, the size of this queue should be equally sensitive to changes in neuronal death and the rate of phagocytosis. Using rodent organotypic hippocampal slice cultures as a model of acute perinatal brain injury, serial imaging demonstrated that the capacity for microglial phagocytosis of dying neurons was overwhelmed for 2 weeks. Altering phagocytosis rates (eg, by changing the number of microglia) dramatically changed the number of visibly dying neurons. Similar effects were generated when the visibility of dying neurons was altered by changing the membrane permeability for stains that label dying neurons. Canonically neuroprotective interventions, such as seizure blockade, and neurotoxic maneuvers, such as perinatal ethanol exposure, were mediated by effects on microglial activity and the membrane permeability of neurons undergoing programmed cell death. These canonically neuroprotective and neurotoxic interventions had either no or opposing effects on healthy surviving neurons identified by the ongoing expression of transgenic fluorescent proteins.

Commentary

In acute brain injury, individual cells die and must be cleared via programmed cell death or apoptosis. The speed at which clearance occurs depends on the number of damaged cells and the rate of phagocytosis. Many previous assays to evaluate the effectiveness of neuroprotective therapies have been static, examining the process at a single time point. However, programmed cell death is a multifactorial and dynamic process. ¹ Novel, more dynamic, approaches are needed to evaluate the entire process to more accurately determine neuroprotective mechanisms to target. Such strategies might be employed immediately after certain types of injuries or perhaps applied prophylactically in individuals at high risk for brain injury (eg, military combat, American football, radiation therapy, etc). An emerging model for assessing both immediate and delayed phases of injury is rodent organotypic hippocampal slice preparation.

Balena et al ² performed a rather elegant study, with numerous experimental paradigms that cannot be described in detail in this commentary. Briefly, to understand the reasons that in their previous study, ³ the number of dying cells could have been overestimated, they first postulated that the dye employed, propidium iodide (PI), could be staining healthy neurons. They determined that this was not the case. They next evaluated the rates of entry and exit into the pool of dying cells, which can be labeled with PI and other markers that enter cells with increased membrane permeability. They determined that in organotypic slices, neurons wait an average of 1.5 days to undergo phagocytosis by microglia, or efferocytosis and that the rate of efferocytosis is proportional to the number of PI + neurons in the programmed death pathway. The organotypic slice is an interesting model well-employed by the lab in which the act of creating the brain slice is the injury, then the organotypic slice is used to model sequelae of acute brain injury. They noted that the ratio of dying neurons to active microglia varies between slices, whereby slices with a greater abundance of dying neurons have longer time constants. Via longitudinal assessment of neurons in their fluorescent imaging paradigm, they determined that there is a very low death rate for neurons in the healthy pool, despite ongoing seizures.

To determine the number of neurons fated to programmed death, but not yet PI+, they examined organic acetoxymethyl ester (AM)-conjugated dye uptake in relation to other programmed death biomarkers. AM dyes enter and become trapped in cells with increased membrane permeability, thus selectively staining dying neurons. They verified this empirically and determined that AM dye-positive cells that were undergoing programmed death demonstrated concurrent quenching of transgenic or virally expressed fluorescent labels. Programmed death is also accompanied by the shrinking of soma and condensation of chromatin, providing additional biomarkers to identify cells undergoing apoptosis.

Since efferocytosis is a key step in apoptosis, they used a dye that selectively stains activated microglia along with the AM dyes. In this manner, they were able to elegantly visualize cells in the programmed death pathway being completely engulfed by microglia and were able to follow this process dynamically over time. Thus, AM-dye staining is a useful biomarker to track neurons throughout the programmed death pathway from the initial fluorescence quenching to ultimate efferocytosis.

They were able to further differentiate dying cells from healthy cells by examining membrane permeability and determined that there are two distinct cell populations, with 1 having very high membrane permeability indicative of cells undergoing programmed death. By tracking membrane permeability, they were able to determine that after acute injury programmed death progresses over days to weeks. Thus, membrane permeability could be another useful biomarker against which to benchmark neuroprotective prowess.

They determined that seizure suppression by kynurenic acid reduced the number of PI + neurons, suggesting a neuroprotective effect. Conversely, depletion of microglia with clodrosome increased the number of PI + neurons, indicating that the density of dying cells at any time point reflects a balance between the rates of entry to and exit from the cell death pathway. Through a complex series of experiments using additional cell death markers, the investigators conclude that seizure suppression both enhances entry into the programmed death pathway and accelerates exit from the programmed death pathway, thus explaining prior paradoxical and unexpected findings. This is a nice example of how changing rates of cell death and clearance can significantly alter cell counts at a specific time, leading to incorrect conclusions.

Finally, they examined whether altering the membrane permeability of neurons already in the programmed death pathway would affect biomarker positivity. For this, they examined the effect of alcohol, which is well known to increase membrane permeability. They determined that ethanol only increased the permeability of already dying cells. Healthy cells were not affected. This implies that any treatment that alters membrane permeability will lead to false readouts with these dyes.

This is an elegant study that demonstrates that since neuronal injury is a dynamic process, static means to assess injury and improvement from injury by neuroprotective agents are insufficient. They also demonstrate that the hippocampal organotypic slice studied via longitudinal two-photon imaging could be a viable assay to assess the effects of injury and the appropriateness of neuroprotective therapies across many facets of neurological disease. They provide a variety of indicators to follow dying cells through the pathway, including both their association with various dyes, as well as morphological, biochemical, and electrophysiological/membrane permeability markers. Here they employed dyes that are taken up throughout the process as opposed to dyes such as PI that are taken up 1 time. They meticulously and systematically examined factors during the time from the cell entering the programmed death pathway and being cleared.

These are important observations. The development of neuroprotective therapies has been of great interest. In theory, a neuroprotective therapy could be something simple and safe to use regularly/prophylactically and would protect against neural injury. However, numerous neuroprotective trials have been unsuccessful. Perhaps some of the failures could be attributed to the inappropriateness of using a static assay to study a dynamic process. Inclusion of stereological cell counts could strengthen such experiments by providing a means to accurately assess the number of neurons present at a given time point. So, this study provides a better, more accurate, way to assay sequelae of injury accounting for temporal dynamics. These findings also provide a better understanding of how the factors involved regulate the clearance of injured cells. Future studies will need to differentiate specific contributions of hypoxic–ischemic injury and traumatic injury, perhaps with the aid of artificial intelligence tools. ⁴ Given the somewhat surprising effect of seizure cessation, this study also brings into question whether prolonged seizures are neurotoxic and how much the treatments themselves contribute to toxicity. ⁵ The current study provides a pathway forward for better differentiating which components of seizure cessation may be helpful/harmful for use in this model and others^6–8 and lends hope that reexamination of some therapeutics using a more appropriate longitudinal assessment may yet reveal promising neuroprotective candidates.