Advancements in in vitro models of traumatic brain injury

https://doi.org/10.1016/j.cobme.2022.100430Get rights and content

Abstract

Annually, forty to seventy million victims globally are estimated to sustain a traumatic brain injury (TBI). To investigate the link between different types of TBI and neurological deficits, various mechanical loading types associated with trauma to the head and brain have been replicated in animal as well as in vitro cell culture systems. In vitro models have studied how the mechanics of trauma translates to injury mechanisms; discovering potential targets for protection, prevention, and treatment. This article focuses on recent developments made in between 2017 and 2022 and the range of culture systems utilized to focus on specific injury outcomes. The review is structured to delve into the reproduction of mechanical loading used to cause injury including: uniaxial and biaxial stretch, compression, shear, weight drop, scratch, rotational acceleration, static pressure, and blast injury. We compare between 2D and 3D injury models, summarize main findings for each model, and list advantages and disadvantages of both approaches. The article also reviews advancements in the use of human induced pluripotent stem cells, new models of repetitive injury, and non-neuronal models of TBI. We identify future directions including 1) the need for simplicity in the protocol for hIPSC culture for more widespread use, 2) the need to replicate complex mechanical loading in vitro and 3) the need to develop a model using adult neuronal cell sources.

Introduction

Globally 40–70 million are estimated to sustain a traumatic brain injury (TBI) each year and an estimated 3.2–5.3 million in the US are coping with a TBI-related disability. TBI has high rates among contact sports, falls, automobile accidents and soldiers [1, 2, 3, 4∗]. Accordingly, TBI can be initiated by very different forms of dynamic mechanical loading including blunt impact, blast exposure, or a rapid inertial rotation of the head [5,6].

While a closed head injury can induce immediate tissue necrosis, edema, and hematoma, the mechanical loading of the brain tissue also initiates secondary injury mechanisms such as altered calcium homeostasis, glia activation, blood brain barrier breakdown, mitochondrial dysfunction, chronic neuroinflammation and progressive neurodegeneration [4,6]. To study how different types of TBI lead to pervasive neurological deficits, different mechanical loading modes associated with TBI have been reproduced in animal as well as in vitro models, Figure 1. In vitro models, in particular, have provided knowledge how the biomechanics of the traumatic event translates to injury mechanisms; identifying potential targets for protection, prevention and treatment. This review is organized to explore each mode of mechanical loading used to induce injury and the variety of culture systems employed to address specific injury outcomes. A literature review using the search terms of traumatic, brain, injury, TBI and in vitro, was conducted from 2016 to present and classic references were used as needed for background and completeness, Table 1.

Since the biomechanics of head injury varies widely, the relationship between mechanical loading and the evolution of structural and functional alterations of brain cellular systems needs to be investigated. Indeed, it is difficult to elucidate how the dynamic deformation of brain cells evolves into neuronal dysfunction and degeneration using in vivo models. For instance, immunohistochemical markers used in animal models offer limited insight into the intracellular mechanisms of the evolving injury. On the other hand, in vitro models have allowed us to directly study biomechanical mechanisms of injury to the individual cells of the brain in real time. Importantly, in vitro models allow for probing and assessing alterations in neuronal and glial structure and function with both novel and standard assays that can be exceedingly difficult in animal models.

In vitro models are higher throughput and less expensive than animal models where experiments can usually be completed on a shorter time scale and allow greater control of the experimental conditions (injury biomechanics and culture conditions) [7,8]. The reduced model allows study of specific molecular cascades without confounding factors such as the vascular system, intracranial pressure or immune response [5]. These models have had significant impact in the field by identifying important biological processes involved in neuronal pathology including, changes in ion homeostasis, pathologic protease activation, and electrophysiogical alterations [9, 10, 11, 12, 13].

The in vitro model indeed reproduces similar pathological features and mechanisms of human TBI allowing it to pose many questions about the neuronal response from mechanical stretch injury [14,15]. For instance, in vitro models have helped establish the evolution of DAI (diffuse axonal injury) including calpain activation, sodium channel proteolysis, and interruption of axonal transport [9,16, 17, 18, 19, 20]. Models have also elucidated mechanisms related to synaptic receptor changes, cleaved spectrin and neurofilaments, and Aβ accumulation; some now used as biomarkers [18,21, 22, 23]. A 2011 review evaluated the fidelity of in vitro modeling of TBI by comparing 26 compounds tested in vitro and in vivo and concluded that in-vitro models were predictive of 88% of in vivo results [24].

Most established models have adapted two-dimensional (2D) cell culture techniques to study injury mechanisms induced by mechanical trauma. 2D in vitro models have used immortalized cell lines and dissociated primary cells leading to ease of implementation and rapid turnaround of experiments needed for high throughput analysis. These systems, however, have not recapitulated important features of the neural micro-environment to study more complex questions [25,26]. Recent advancements in 3D culture techniques have offered significant improvements to in vitro injury models such as replicating tissue stiffness, cell–cell interaction, native extracellular matrix, and hetero-cellular composition [27]. Table 1 identifies recent studies using 2D and 3D cultures. 3D culture models are physically essential to the advancement of this field. Biomechanically, 3D cultures enable injury loading that can only be replicated with a 3D structure including shear and compression. They will allow the incorporation of human neuronal and multi-cell systems requiring the 3D culture environment to differentiate into appropriate form and function.

Ex vivo models offer a hybrid between the controlled conditions of in vitro culture and the complexity of in vivo tissue architecture. Two ex vivo models have been used in the study of TBI: organotypic cultures and acute live brain slices. Fresh brain slices from injured animals (within hours post-injury) have been primarily used to study the electrophysiological function of brain circuits such as the hippocampus. We did not include these in this review, but readers are directed to a recent review [21]. Organotypic cultures are derived from tissue slices and have the advantage of preserving the original tissue architecture and cellular constituents in culture. While the culture techniques for slices are more demanding and lead to lower experimental yield, these cultures preserve some of the original tissue architecture and neuronal circuits. Organotypic hippocampal slice cultures on deformable silicone membranes have been injured with equi-biaxial deformation [5,28, 29, 30, 31, 32, 33, 34], impact using a weight drop device [34] and for the study of blast TBI [5,30, 31, 32, 33,35]. While these models provide the advantage of probing the system in real time, there is loss of control of the mechanical loading compared to culture models. They also retain confounding factors of the in vivo system that cannot be easily separated from the system.

The type and degree of damage to the brain and individual neurons are directly related to the kinematics, boundary conditions, and type of loading to the head. For instance, skull fracture is related to direct impacts and linear motions whereas subdural hematoma and diffuse axonal injury are correlated to non-impact rotational motions and the associated inertial motion of the brain [36, 37, 38, 39]. In this review, we classify in-vitro models by the mechanical mode of injury.

Scratches have been used to replicate injuries such as cerebral lacerations from penetrating injuries. One model design produces concentric mechanical cuts in a cultured cell layer. The model has been used to show glial cells are more resistant to injury than neurons where glutamatergic and gabaergic antagonists were neuroprotective [25,40]. Scratch has also been applied manually with pipette tips to recreate pathological features of penetrating TBI, cell infiltration into lesion cite biomaterials, and astrocytes hypertrophy with GFAP upregulation [41]. The scratch model can be useful to study laceration of nervous tissue associated with penetrating brain injuries. We note that the majority of TBI models replicate non-penetrating injuries as closed head type injuries rarely result in disconnection of axons from the injury event. Accordingly, one has to take care in generalizing findings to all types of TBI. Importantly, disruption of the cell membranes, like necrosis, is a well-known activator of inflammatory and regulated cell death processes.

Stretching of neurons has been long implicated as a major mechanical mechanism of TBI induction. Biaxial stretch is reproduced with 2D circular elastic polydimethylsiloxane (PDMS) membranes where the substrate is stretched in both the radial and circumferential directions. One injury model design uses an air pulse to stretch the PDMS membrane like the surface of a balloon. It has been the most widely employed in vitro model of TBI and has been used to study neuronal and astrocyte responses to strain magnitude and rate [15,42], 43], 44, 45]. A disadvantage of the air pulse model is that the biaxial stretch is not uniform across the culture substrate. A second injury model has been designed to produce equi-biaxial stretch (radial and circumferential strains are equal) using a linear actuator system to stretch the PDMS over a cylindrical post keeping the membrane deformation within a single plane [8,46].

Biaxial stretch has been employed to multi-well injury devices to increase sample throughput. Models have been designed to stretch several culture wells at one time or a multiwell plate adapted with a PDMS substrate [7,8]. Both air pulse and linear actuator designs have been used to stretch multiple wells simultaneously. While throughput is increased, these systems have some variability in the well-to-well induction of stretch. Due to the developmental cost and complexity of these systems, their use has been limited.

The model has been particularly useful in separating levels of stretch that result in loss of plasma membrane integrity representing an injury that results in immediate necrosis vs. inducing secondary cascades; an effect that is sensitive to both strain magnitude and rate [47,48]. The biaxial stretch model has been used to study cell viability, intracellular calcium dynamics, electrophysiological changes, as well as many intracellular secondary mechanisms of injury [13,18,49].

When considering the biaxial stretch, it is important to distinguish the difference from a uniaxial deformation (section 1.2.3) or simple shear (section 1.2.4). In a one-dimension stretch, the neuron shrinks in the other directions due to the effect of Poisson's ratio. This is a higher level concept of complex loading in mechanics and the reader is directed here for further details [50]. Briefly, when a material, tissue or nerve cell is stretched in more than one direction, the shrinking is constrained and the effective stiffness of the material or tissue behavior increases requiring more stress to achieve the same amount of strain. In the case of biaxial stretch, the applied strain occurs within the 2D plane and the neuronal cells can only shrink in the out of plane direction. This will result in higher stresses in the neuron that can cause more damage than an equal strain magnitude under a uniaxial stretch. For instance, mechanoporation of the cell membrane due to stretch often occurs in biaxial stretch at strains of 50% whereas mechanoporation does not occur in the uniaxial stretch of axons up to 65% strain [48,51].

Recently, the biaxial stretch injury device was used to demonstrate that Exendin-4 (a glucagon-like peptide 1 receptors (GLP-1R) antagonist) has neurotrophic and neuroprotective actions, increasing cell viability over the vehicle injury group [52]. Another study used biaxial stretch for screening five drugs (indapamide, captopril, rifampicin, camphor and etoposide); all of them improved viability through measurement of lactate dehydrogenase (LDH) and cell metabolism with the MTT Assay responses in injured cells. Further, Rifampicin had the best outcome by reducing cell necrosis and apoptosis as well as improved cellular bioenergetics [53]. Another study used biaxial stretch to investigate if postsynaptic density 95 (PSD-95) protein facilitates intracellular signaling through NMDA receptors. Their results show that PSD-95 expression has no effect on neuronal viability and that morphological changes to injured neurons can be replicated with a mutant PSD-95 [54].

It has been widely hypothesized that brain deformation reduces to a uniaxial stretch of axons that induces axonal degeneration and disconnection associated with the pathology of diffuse axonal injury [13,55]. In this model, a region of uniaxially aligned axons are grown within a cell free region of a PDMS membrane that is deformed uniaxially in the direction of the axons using and air pulse [51] or a linear actuator [56]. This model provides the advantage of isolating the axonal and somatic compartments of the neurons and isolates injury to the axons only. In this model, degeneration processes have been followed in real time with axons developing swellings along their length from interrupted axon transport [51]. Recently, uniaxial stretch has been used to study the effects of low levels of stretch that replicate mild TBI on the viability and function of primary isolated rat brain microvascular endothelial cells. This study showed that stretch injury led to a reduction in tight junction protein expression and increased permeability [3].

Head rotations during injury results in large shear deformation of the brain tissue [7]. To model the effects of simple shear, a linear actuator was used to apply shear to the top plane of a 3D neuronal-astrocytic co-culture and established that neurons are sensitive to shear strain magnitude and rate [57]. In mechanics, shearing deformation is a deformation mode consisting of both tensile and compressive strains along the principal axes within the plane [50]. Accordingly, axons in the 3D construct can be aligned in the direction of stretch, compression or a mix of the two. Due to the complexity of the model and 3D culture conditions, its use has been limited.

Alternatively, fluid shear stress has been applied to a 2D neuronal culture using a microfluidic chamber and a fast pressure servo. It is important to delineate that fluid shear as a loading mechanism is very different that shearing a 3D culture. Applying shear to a 3D culture creates a shape change affecting the entire culture creating a mixed deformation mode on the individual cells. Fluid shear on the other hand places a shear force on the cell membrane rather than a deformation of the whole neuron [58]. A recent study found that fluid shear stress increased intracellular calcium depending on the magnitude, duration, and rise time of the stimulus [59]. Interestingly, fluid shear has also been used to model blast exposure. This model demonstrated that fluid shear on cells having sub-millisecond rise times and amplitudes of 8–21Pa can trigger a purinergic cellular calcium response in primary human neuronal-astrocyte cultures and blocking with the antagonist PPADS reduces intracellular calcium [60].

Recent in vitro models have been introduced to deliver a compressive deformation to 3D culture systems using an impactor via a linear actuator or weight drop. Biological tissue is considered incompressible and therefore does not deform under purely compressive loads. Therefore, it is not surprising that 3D cultures have been found more sensitive to bulk shear deformation than compression with respect to viability and cell membrane permeability [57]. It is noted, however, that while the term compression has been used, impactors that push on neural tissue result in the tissue being squeezed out in the lateral directions. This deformation leads to mixed loading of compression, tension and shear [61]. One study considered repeated impacts to 3D neuronal cultures to show that neurite swelling is highly strain rate dependent in compression. Unlike for other modes of injury, mechanoporation seems to be absent in compression [62]. Under compressive impact, hypothermia is neuroprotective, reducing apoptosis and increasing calpastatin expression, an endogenous neuroprotectant [63]. Using a weight drop device, differentiated PC12 cells in 3D collagen scaffolds were used to measure calcium dynamics [64]. In a novel design, 3D cultures of human brain spheroids were impacted to show compression results in reduced neuronal viability and decreased spontaneous electrical activity [65]. Since these models attempt to replicate the pre-clinical animal model (section 2.2), they can be useful in terms of questions in relationship to the in vivo injury model, but should not be used to draw conclusions on the relationship between compression and the induction of injury.

Trauma from an explosive blast is classified as primary, secondary, tertiary or quaternary injury [66]. In vitro models have been created to model primary blast injury that results from the direct action of the shock wave on cells of the brain. Reproduction of the shock wave is modeled by the Friedlander waveform, where the peak overpressure (maximum pressure) and the duration are the main contributors to injury [5,32]. There have been two approaches to recreating primary blast to in vitro cultures. 1) A shock tube apparatus where a membrane is ruptured with compressed nitrogen or helium to expose a liquid chamber that houses the in vitro system [5,67,68]. 2) Detonating a charge within a liquid chamber containing the in vitro system. In this system, a capacitor is discharged that vaporizes a small thin wire, producing a repeatable high-rate compression wave through the culture [60]. One consideration of these in vitro models is the use of liquid media to deliver the shock wave. While it is a practical issue, particularly with the liquid components in an in vitro system, a shock wave propagates differently in liquid than air. Indeed, the propagation of the primary blast wave through the various media of the head (skin, skull, meninges, and fluids) remains to be investigated.

Blast models have employed organotypic slices in addition to dissociated cultures. Using rat hippocampal slice cultures, cell death was found to scale with peak pressure of the shock wave [5]. Blast injury induced axonal beading (degenerative morphology), but interestingly there was no significant difference between a single or repeated blast. The model was used to show that Fluoro-Jade staining is absent after blast. Fluoro-Jade is a common marker of acute degenerating neurons in vivo models of TBI [33]. Organotypic hippocampal slice cultures from 57BL/6N mice have been used to demonstrate that Xenon gas (50% atm) applied 1 h after blast exposure reduced cell death over 72 h post injury, compared with untreated injured cultures [32].

Using the advantage of an in vitro model, the time course of deficits in neuronal activity was explored. Injured slices required 1 h to develop deficits in long term potentiation which recovered over 10 days following exposure depending on blast intensity. Blast injury reduced AMPA and GluR1 subunit expression, reduced the expression of postsynaptic density protein-95 (PSD-95) and phosphorylation of stargazin protein. Finally, they found that activation of the cyclic adenosine monophosphate (cAMP) pathway by rolipram/forskolin ameliorated electrophysiological and protein-expression changes caused by blast [31].

A couple of additional in vitro models were found that attempt to replicate inertial deformation from rapid acceleration and static pressure associated with elevated intracranial pressure (ICP). Rotational acceleration/deceleration of the head has been well described as leading to large brain tissue deformations. A custom-made apparatus consisting of a rotational motion stage and a spring-loaded impactor was used to rapidly accelerate a 3D culture of mixed glia to rotational speeds thought to cause mild TBI. No significant glial activation was observed in the model [6]. Another study considered static pressure-induced brain injury associated with elevated intracranial pressure (ICP), present in almost every category of intracranial pathology like concussive injuries, mass lesions, hydrocephalus, sudden vascular injury, or hypertensive hemorrhages. The model developed a pressure-controlled incubation system using 3D culture constructs with a small fluid layer representing the intraventricular space and the bottom against a hard surface of the well-plate representing the skull. In vitro cultures subjected to sustained pathologic pressures induced an acute stress response through ATP release in neurons [69].

Section snippets

Use human neuronal cultures

All drug treatments derived from preclinical animal models have failed in the treatment of TBI [15]. The introduction of human derived neurons to in vitro TBI models are a major advancement to the translation of mechanisms studied in rodent TBI models. Human induced pluripotent stem cells (hiPSC) derived from brain are differentiated into neural progenitor cells (NPC) then over a period of 8 weeks mature into neurons and glial cells that replicate synaptogenesis, spontaneous neuronal electric

Limitations and future directions

There are distinct advantages to the in vitro model of TBI in addressing questions that are not possible in the in vivo model. However, with a reduced model, there are several limitations that must be considered in regards to the translational relevance to the preclinical animal TBI model and to the human condition. From our experience, we found the following limitations should be considered:

  • 1)

    Derivation/isolation of neuronal tissue: Ex-vivo brain slices and in vitro neuronal and glial cultures

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (80)

  • S.A. Pasquesi

    Predictions of neonatal porcine bridging vein rupture and extra-axial hemorrhage during rapid head rotations

    J Mech Behav Biomed Mater

    (2020)
  • R.H. Basit

    In vitro model of traumatic brain injury to screen neuro-regenerative biomaterials

    Mater Sci Eng C

    (2021)
  • P.M. Abdul-Muneer

    High Ca2+ influx during traumatic brain injury leads to caspase-1-dependent neuroinflammation and cell death

    Mol Neurobiol

    (2017)
  • E. Chierto

    Mechanical stretch of high magnitude provokes axonal injury, elongation of paranodal junctions, and signaling alterations in oligodendrocytes

    Mol Neurobiol

    (2019)
  • D.H. Smith

    High tolerance and delayed elastic response of cultured axons to dynamic stretch injury

    J Neurosci

    (1999)
  • A. Menichetti

    A machine learning approach to investigate the uncertainty of tissue-level injury metrics for cerebral contusion

    Front Bioeng Biotechnol

    (2021)
  • A. Aravind

    Behavioral deficits in animal models of blast traumatic brain injury

    Front Neurol

    (2020)
  • Kim, J., et al., Mechanical stretch induces myelin protein loss in oligodendrocytes by activating Erk1/2 in a...
  • M.C. Dewan

    Estimating the global incidence of traumatic brain injury

    Journal of Neurosurgery JNS

    (2019)
  • W. Shi

    Design and evaluation of an in vitro mild traumatic brain injury modeling system using 3D printed mini impact device on the 3D cultured human iPSC derived neural progenitor cells

    Adv Healthc Mater

    (2021)
  • R. Campos-Pires

    A novel in vitro model of blast traumatic brain injury

    JoVE

    (2018)
  • C.T. Tsui

    Applying a novel 3D hydrogel cell culture to investigate activation of microglia due to rotational kinematics associated with mild traumatic brain injury

    J Mech Behav Biomed Mater

    (2021)
  • S.A. Sherman

    Stretch injury of human induced pluripotent stem cell derived neurons in a 96 well format

    Sci Rep

    (2016)
  • A. Iwata

    Traumatic axonal injury induces proteolytic cleavage of the voltage-gated sodium channels modulated by tetrodotoxin and protease inhibitors

    J Neurosci

    (2004)
  • C.Q. Kao

    Potentiation of GABA(A) currents after mechanical injury of cortical neurons

    J Neurotrauma

    (2004)
  • D.M. Geddes et al.

    Susceptibility of hippocampal neurons to mechanically induced injury

    Exp Neurol

    (2003)
  • X. Di

    Mechanical injury alters volume activated ion channels in cortical astrocytes

    Acta Neurochir Suppl

    (2000)
  • V.E. Johnson et al.

    Axonal pathology in traumatic brain injury

    Exp Neurol

    (2013)
  • E. Su et al.

    Diffuse axonal injury

  • C.R. von Reyn

    Mechanisms of calpain mediated proteolysis of voltage gated sodium channel alpha-subunits following in vitro dynamic stretch injury

    J Neurochem

    (2012)
  • J.A. Wolf

    Traumatic axonal injury induces calcium influx modulated by tetrodotoxin-sensitive sodium channels

    J Neurosci

    (2001)
  • C.R. von Reyn

    Calpain mediates proteolysis of the voltage-gated sodium channel alpha-subunit

    J Neurosci

    (2009)
  • K.A. Hamilton et al.

    Current ex vivo and in vitro approaches to uncovering mechanisms of neurological dysfunction after traumatic brain injury

    Curr Opin Biomed Eng

    (2020)
  • M.D. Tang-Schomer

    Partial interruption of axonal transport due to microtubule breakage accounts for the formation of periodic varicosities after traumatic axonal injury

    Exp Neurol

    (2012)
  • N. Rouleau

    A 3D tissue model of traumatic brain injury with excitotoxicity that is inhibited by chronic exposure to Gabapentinoids

    Biomolecules

    (2020)
  • C. Chen

    Develop a 3D neurological disease model of human cortical glutamatergic neurons using micropillar-based scaffolds

    Acta Pharm Sin B

    (2019)
  • V. Liaudanskaya

    Modeling controlled cortical impact injury in 3D brain-like tissue cultures

    Adv Healthc Mater

    (2020)
  • B. Morrison

    An in vitro model of traumatic brain injury utilising two-dimensional stretch of organotypic hippocampal slice cultures

    J Neurosci Methods

    (2006)
  • A.P. Miller

    Acute death of astrocytes in blast-exposed rat organotypic hippocampal slice cultures

    PLoS One

    (2017)
  • E.W. Vogel

    Primary blast injury depressed hippocampal long-term potentiation through disruption of synaptic proteins

    J Neurotrauma

    (2017)
  • View full text