Translational benchmarking of 3D in vitro models of TBI

Abstract

Traumatic brain injury (TBI) remains a major cause of neurological disability, yet effective therapies are limited. Advances in 3D in vitro neural models, including organoids, scaffold-based constructs, and brain-on-chip systems, have provided new opportunities to study neurotrauma under controlled conditions. This review evaluates recent in vitro tissue platforms through a dual framework of mechanical fidelity and biological response, benchmarking applied injury conditions against human biomechanical reconstructions and assessing the coverage of downstream responses relative to TBI pathology. Many 3D models apply clinically relevant mechanical loading regimes and reproduce select phenotypes, but remain limited by inconsistent mechanical reporting, incomplete cellular complexity, and sparse modeling of chronic and repetitive injury. Addressing these challenges will improve the translational utility of next-generation platforms.

Modeling TBI in Three Dimensions

TBI is a leading cause of death and neurological disability worldwide, affecting over 50 million individuals annually[1]. TBI is a broad term encompassing mechanistically distinct injury types, including focal contusions from direct impact, penetrating injury from objects breaching the skull, diffuse injury driven by rotational and shear loading, and blast injury from explosive pressure waves[2]. Beyond acute tissue and cellular injury, TBI contributes to chronic neurodegeneration, psychiatric disorders, and cognitive impairment, a growing public health burden driven by aging populations, military blast exposure, contact sports, and motor vehicle injuries. Despite decades of research and clinical trials, effective therapies remain limited[1]. Longitudinal clinical studies characterize TBI outcomes but are inherently limited by the heterogeneity of injury mechanisms, confounding influence of comorbidities, and an inability to directly access the tissue-level mechanical and biological processes driving the pathology[3]. Finite element (FE) (see Glossary) brain models partially bridge this gap by enabling estimation of internal tissue-level strains[4], providing biomechanical benchmarks for experimental validation. Animal models have advanced the mechanistic understanding of TBI, but species differences[5] combined with regulatory shift toward non-animal alternatives (New Approach Methodologies, NAMs) exemplified by the FDA Modernization Act 2.0[6], have motivated the development of human-relevant experimental systems capable of linking controlled mechanical insult to cellular response.

3D in vitro neural tissue models have emerged as candidate platforms for this purpose, spanning brain organoids, spheroids, scaffolds, hydrogels, microfluidic brain-on-chips , and organotypic slice cultures[7]. These platforms provide mechanical tunability, and controlled cellular inputs and injury parameters, enabling mechanistic investigations that are difficult to attain with animal models or clinical studies. Whether these platforms reproduce the mechanical and biological features of TBI with sufficient fidelity to inform clinical translation remains an open question. Existing reviews of in vitro TBI models remain largely descriptive and have not examined how applied mechanical conditions compare to clinical injury biomechanics or how biological outputs align with established human TBI pathology [8, 9]. While translational benchmarking frameworks have been applied to animal TBI models[5, 10], a comparable analysis for 3D in vitro systems is lacking. This review critically evaluates recent 3D in vitro TBI models through a dual framework of mechanical fidelity and biological coverage, benchmarking applied mechanical injury conditions against clinical reconstructions, and assessing which TBI-associated biological responses each platform reproduces.

Mechanical fidelity: applied loading against clinical biomechanics

Platform architecture determines both the mechanical loading regimes a tissue construct can sustain and the cellular complexity available for downstream injury response. Platforms differ in cellular source, spanning rodent and human cells and, among human systems, induced pluripotent or neural stem cell derivatives rather than mature tissue. Species, cell type, and developmental maturity vary independently of architecture and are reported in Table 1. Figure 1, Key figure summarizes the review framework, spanning the brain’s mechanical and cellular properties (Figure 1a), six recent 3D platform architectures applied to model TBI (Figure 1b), the applied loading mechanisms (Figure 1c), and the biological readouts used to characterize injury responses (Figure 1d). Brain tissue mechanics and the mechanical properties of each platform together constrain the achievable loading regimes (Box 1). The following subsections examine how these platforms have been mechanically loaded to simulate TBI, comparing applied conditions against FE-derived clinical tissue strains to assess mechanical fidelity. These subsections are organized by applied mechanics, the physical loading imposed on the construct, rather than injury pattern because a single loading mode can target different clinical injuries and a given injury can be produced by several distinct loads. For each mechanism, the intended clinical phenotype and the evidence linking the two are stated explicitly, while exposure schedule, whether loading is delivered once or repeatedly, is treated as a cross-cutting dimension.

Table 1.

Mechanical loading conditions, cellular composition, and biological readouts of 3D in vitro models of TBI, organized by applied mechanics

Study	Applied mechanics (Device loading)	Platform	Species	Cell source and maturity a	Vascular component b	Key mechanical parameters	Intended phenotype	Exposure schedule	Biological readouts (max timepoint)	Refs
Hinrichsen et al.	Indentation / Impact (CCI)	Scaffold	Human	hiNSC-derived, Primary, Immortalized cell line; immature	Yes	Impact 6 m/s; dwell 200 ms; depth 0.6 mm; $\epsilon$: 40% c; $\stackrel{.}{\epsilon }$: 4000 s−1 c	TBI (moderate)	Single	CD, MB, NI, BBB, NG (14 days)	[31]
Power et al.	Indentation / Impact (CCI)	Scaffold	Human	primary, immortalized cell line	Yes	Impact 6 m/s; dwell 200 ms; depth 0.6 mm; $\epsilon$: 40% c; $\stackrel{.}{\epsilon }$: 4000 s−1 c	TBI (unspecified)	Single	CD, NI, BBB (1 day)	[25]
Power et al.	Indentation / Impact (CCI)	Scaffold	Human	primary	Yes	Impact 6 m/s; dwell 200 ms; depth 0.6 mm; $\epsilon$: 40% c; $\stackrel{.}{\epsilon }$: 4000 s−1 c	TBI (unspecified)	Single	CD, NI, BBB (14 days)	[24]
Hsi et al.	Indentation / Impact (CCI)	Scaffold	Human	hiNSC-derived, primary, immortalized cell line; immature	None	Impact 6 m/s; dwell 200 ms; depth 0.6 mm; $\epsilon$: 40% c; $\stackrel{.}{\epsilon }$: 4000 s−1 c	TBI (unspecified)	Single	CD, FB (7 days)	[26]
Coe et al.	Indentation / Impact (CCI)	Scaffold	Human	hiNSC-derived, primary, immortalized cell line; immature	None	Impact 6 m/s; dwell 200 ms; depth 0.6 mm; $\epsilon$: 40% c; $\stackrel{.}{\epsilon }$: 4000 s−1 c	TBI (unspecified)	Single	CD, NW, FB (10 weeks)	[27]
Cairns et al.	Indentation / Impact (CCI)	Scaffold	Human	hiNSC-derived; immature	None	Impact 1, 3, 6 m/s; dwell 200 ms; depth 0.6 mm; $\epsilon$: 40% c; $\stackrel{.}{\epsilon }$: 666.7, 2000, 4000 s−1 c	contusion	Single	CD, NW, NI, NG (8 days)	[28]
Cairns et al.	Indentation / Impact (CHI, pneumatic)	Scaffold	Human	hiNSC-derived; immature	None	Impact 207, 414 kPa	contusion	Repeated (3 times total), 24 hours	CD, NW, NI, NG (8 days)	[28]
Rogers et al.	Indentation / Impact (CHI, pendulum)	Microfluidic Chip	Rodent	primary; embryonic-derived	None	Peak linear acceleration: 200g	contusion	Single	CD, NW, MB, NI, NG (1 day)	[36]
Pischiutta et al.	Indentation / Impact (CCI)	Organotypic Slice	Rodent	primary; ex vivo/juvenile	None	Impact 1.5 m/s; dwell 100 ms; $\epsilon$: 50% c; $\stackrel{.}{\epsilon }$: 7500 s−1 c	contusion	Single	CD, NI, FB (2 days)	[32]
Loussert-Fonta et al.	Indentation / Impact (microvalve droplet)	Organoid	Human	iPSC-derived; extended, immature	None	Not reported	TBI (mild)	Single	CD, NW, NG, FB (2 days)	[35]
Liaudanskaya et al.	Indentation / Impact (CCI)	Scaffold	Human	hiNSC-derived, primary, immortalized cell line, iPSC-derived; immature	None	Impact 6 m/s; dwell 200 ms; depth 0.6 mm; $\epsilon$: 40% c; $\stackrel{.}{\epsilon }$: 4000 s−1 c	contusion	Single	CD, MB, NI, NG (1 day)	[30]
Shi et al.	Indentation / Impact (weight drop)	Organoid	Human	iPSC-derived, immortalized cell line; immature	None	Drop 15,30 mm; $\epsilon$: >49%; $\stackrel{.}{\epsilon }$: 430,620 s−1 c	TBI (mild)	Repeated (2 times), 3 days	CD, NI (27 days)	[34]
Ramirez et al.	Indentation / Impact (CCI)	Organoid	Human	iPSC-derived; extended, immature	None	Impact 4 m/s; dwell 200 ms; depth 1 mm	TBI (unspecified)	Single	CD, NG (7 days)	[33]
Liaudanskaya et al.	Indentation / Impact (CCI)	Scaffold	Rodent	primary; embryonic-derived	None	Impact 6 m/s; dwell 200 ms; depth 0.6 mm; $\epsilon$: 40% c; $\stackrel{.}{\epsilon }$: 4000 s−1 c	contusion	Single	CD (1 day)	[29]
Shiravi et al.	Global Compression (uniaxial)	Organoid	Human	iPSC-derived; immature, genetic (MAPT V337M and IVS10+16 mutations)	None	$\epsilon$: 0–50%; $\stackrel{.}{\epsilon }$: 0–30.3 s−1 c	contusion	Single	CD, MB, NG, FB (7 days)	[37]
Gonzalez-Cruz et al.	Global Compression (centrifugation)	Spheroid	Rodent	primary; embryonic-derived	None	$\epsilon$: 10–35%; pressure 10–149 kPa; 2 min	sustained compression injury	Single	CD (1 day)	[40]
Beltran et al.	Global Compression (uniaxial)	Organoid	Human	hESC-derived; immature	None	$\epsilon$: 3%; $\stackrel{.}{\epsilon }$: 0–14 s−1	TBI (mild, moderate)	Single	CD, NW (during injury)	[38]
Shoemaker et al.	Global Compression (uniaxial)	Spheroid	Human	iPSC-derived; immature	None	$\epsilon$: 0–60%; $\stackrel{.}{\epsilon }$: ~0.1–17.8 s−1	TBI (non-penetrating)	Single	CD, NW, MB, FB (3 h)	[41]
Scimone et al.	Global Compression (uniaxial)	Hydrogel	Rodent	primary; embryonic/early primary	None	$\epsilon$: 26%; $\stackrel{.}{\epsilon }$: 89.1 s−1	TBI (mild)	Single	CD, MB (1 day)	[39]
Bobo et al.	Global Compression (uniaxial)	Microfluidic Chip	Rodent	immortalized cell line; N/A	None	$\stackrel{.}{\epsilon }$: 0.04–0.08 s−1	TBI (unspecified)	Single	NW (during injury)	[42]
Khetani et al.	Tension / Torsion / Shear (stretch, microchannel)	Microfluidic Chip	Rodent	immortalized cell line; N/A	None	$\epsilon$: 4–20%; torsion 30–300°	diffuse axonal injury	Single	CD, NG, FB (2 h)	[50]
Wu et al.	Tension / Torsion / Shear (stretch, rotation)	Hydrogel	Human	primary; primary-patient derived	None	$\epsilon$: 4, 8, 20%; 1 Hz	TBI (mild)	Repeated, immediate (50 cycles at 1Hz)	CD (2 days)	[48]
Pan et al.	Tension / Torsion / Shear (fluid shear)	Microfluidic Chip	Rodent	primary; embryonic-derived	None	Flow rate: 0.2, 1 mL/min	diffuse axonal injury	Single	CD, NW (during injury)	[49]
Tsui et al.	Tension / Torsion / Shear (rotation)	Hydrogel	Rodent	primary	None	Angular velocity 45 rad/s; time-to-peak 15 ms	TBI (mild)	Repeated, immediate (10 times)	CD, NI (2 days)	[47]
Wiseman et al.	Laceration (pipette tip)	Hydrogel	Rodent	primary; embryonic-derived	None	Lesion diameter 860 ± 41 μm	penetrating	Single	CD (3 days)	[44]
Rouleau et al.	Laceration (biopsy punch)	Scaffold	Rodent	primary; embryonic-derived	None	Lesion diameter 2,3 mm; vol 6.28,14.14 mm3	penetrating	Single	CD, NW (7 days)	[43]
Sirtori et al.	Shock-wave (pressure chamber)	Organoid	Human	iPSC-derived; immature	None	P: 900–1200 kPa; I: 600 kPa·ms	blast	Single	CD, NG (7 days)	[52]
Yokosawa et al.	Shock-wave (shockwave generator)	Hydrogel	Rodent	immortalized cell line	None	P: 131.6 kPa; I: 35.2 kPa·ms	blast	Single	CD (9 days)	[74]
Lai et al.	Shock-wave (focused ultrasound)	Organoid	Human	iPSC-derived; immature, genetic (C9ORF72)	None	P: > 600 kPa	blast	Single	CD, NW, NG, FB (7 days)	[54]
Beauclair et al.	Shock-wave (shock tube)	Microfluidic Chip	Rodent	primary; embryonic-derived	None	P: 138 kPa; I: ~172.5 kPa·ms c	blast	Single	CD, NW, MB, NI (1 day)	[55]
Rubby et al.	Shock-wave (shockwave generator)	Microfluidic Chip	Rodent	primary	None	P: 188, 439 kPa	blast	Single	CD (7 days)	[56]
Snapper et al.	Shock-wave (blast chamber)	Scaffold	Human	hiNSC-derived, primary, immortalized cell line; immature	None	P: 71.7, 128.2 kPa; I: ~640 kPa·ms c	blast	Single	CD, FB (2 days)	[77]
Silvosa et al.	Shock-wave (pressure chamber)	Organoid	Human	iPSC-derived; immature	None	P: 250, 350 kPa; 500–5000 Hz	blast	Single	CD, NW (1 day)	[53]
Hsi et al.	Cavitation (focused ultrasound)	Scaffold	Human	hiNSC-derived, primary, immortalized cell line; immature	None	3, 30, 300 pulses/mm2	blast	Single	CD, NW, NI, NG, FB (2 weeks)	[61]
Estrada et al.	Cavitation (laser cavitation)	Hydrogel	Rodent	primary; embryonic/early primary	None	$\stackrel{.}{\epsilon }$: 103–108 s−1	blast	Single	CD (1h (immediate))	[75]

^aDevelopmental maturity is inferred from cell source and pre-injury culture duration where not explicitly reported.

^bVascular component indicates the presence of endothelial or vascular cells; no reviewed platform incorporated active perfusion.

^cStrain, strain rate, or impulse values were derived or calculated from reported parameters rather than measured directly; see main text for estimation methods.

Abbreviations: BBB, blood-brain barrier disruption; CCI, controlled cortical impact; CD, cell death and structural integrity; CHI, closed head injury; FB, fluid biomarker release; MB, metabolic dysfunction; N/A, not applicable; NG, neurodegeneration and proteinopathy; NI, neuroinflammation; NR, not reported; NW, ionic and network dysfunction; $\epsilon$, strain; $\stackrel{.}{\epsilon }$, strain rate; P, peak overpressure; I, impulse; hiNSC, human induced neural stem cell; iPSC, induced pluripotent stem cell; hESC, human embryonic stem cell; MAPT, microtubule-associated protein tau.

Figure 1. Key Figure. 3D in vitro platforms for modeling traumatic brain injury (TBI) across mechanical fidelity and biological coverage.

This review evaluates 3D in vitro TBI platforms along two axes, the fidelity of applied mechanical loading to clinical injury biomechanics, and the coverage of TBI-associated biological responses each platform reproduces. (A) The mechanical properties and cellular complexity of the human brain define the design requirements for physiologically relevant 3D in vitro TBI models. Translational and rotational loading transmit mechanical forces throughout the brain, affecting neurons, oligodendrocytes, astrocytes, microglia, pericytes, and endothelial cells lining the blood-brain barrier. (B) Six 3D platform architectures have recently been applied to model TBI in vitro, each with distinct mechanical characteristics, geometric constraints, and cellular composition. These include cerebral organoids, spheroids, scaffold-based constructs, hydrogel constructs, brain-on-chip microfluidic devices, and organotypic slice cultures. (C) Injury paradigms are organized by applied loading mechanics, spanning indentation and impact, global compression, tension, torsion, and shear (inertial), laceration, and shock-wave exposure and cavitation, intended to model particular TBI phenotypes. Exposure schedule, whether loading is applied once or repeatedly, is a cross-cutting dimension. (D) Biological readouts used to characterize injury span the acute-to-chronic post-injury cascade, encompassing cell death and structural damage, ionic and network dysfunction, metabolic disruption, neuroinflammation, blood-brain barrier breakdown, neurodegeneration and proteinopathy, and fluid biomarker release.

Box 1. Brain tissue mechanics and traumatic injury biomechanics.

The brain is exceptionally soft with an approximate Young's modulus (E) of 0.5 to 3 kPa, among the most compliant tissues in the body and highly susceptible to deformation [11]. Its composite architecture of neurons, glia, extracellular matrix, and interstitial fluid gives rise to viscoelastic behavior, with stiffness increasing by an order of magnitude under rapid loading relative to quasi-static conditions [11]. This rate dependence means that identical tissue strains ($\epsilon$) can produce different injury outcomes depending on the strain rate (${\epsilon }^{\cdot }$), so both are used in biomechanical reconstructions, though the relative contributions of strain, strain rate, and their product to injury remain incompletely resolved [12,13].

Brain tissue is nearly incompressible (Poisson's ratio of ~0.45 to 0.49), resisting volume change and deforming primarily in shear rather than compression during head impact [11]. Head motion combines translational (linear) acceleration, which produces localized deformation and pressure gradients underlying focal injuries such as contusions[14], and rotational acceleration, which generates distributed shear underlying diffuse axonal injury (DAI), where rapid axonal stretch disrupts cytoskeletal integrity and impairs transport[15]. This inertial loading can be represented by four components, the Euler force (angular acceleration), the linear force (linear acceleration), the centrifugal force (angular velocity), and the Coriolis force, which contribute unequally to the resulting brain strain [12,16]. In finite element models of sports impacts, brain strain is produced primarily by the Euler force, consistent with angular acceleration as the primary driver of diffuse injury [12,16].

The variable mechanical properties of 3D in vitro platforms constrain the loading regimes they support. Organoids and spheroids have Young’s moduli below 0.5 kPa [17,18] resembling immature neural tissue [17], making them well-suited for low-rate deformation but limiting them under high-velocity loading. Scaffold and hydrogel constructs offer tunable stiffness across a broader range (E ~0.1 to 20 kPa) with greater structural integrity under rapid loading. Their dimensions are set and controlled during fabrication, allowing more precise definition of applied boundary conditions than in self-organizing constructs. Microfluidic systems enable sensor integration and compartmentalization, but rigid housings may be incompatible with high-velocity impact or introduce confounding wave reflections in blast models. Reproducing these injury-relevant loading conditions, distributed shear at high strain rates delivered to soft, rate-stiffening tissue, is the central biomechanical challenge for in vitro platforms, whose stiffness, geometry, and housing determine which components of the in vivo loading environment they can recreate.

Indentation and Impact

Indentation and impact methods apply spatially localized loading intended to model focal, contusion injury, which clinically arises from direct mechanical impact producing localized tissue deformation sufficient to disrupt neuronal membranes, alter ion homeostasis, and damage cytoskeletal structures[19]. Biomechanical reconstructions estimate focal peak strains of approximately 10 - 40%, at strain rates spanning tens to several hundred s⁻¹[20, 21], with proposed injury thresholds of greater than 20% strain and 10 s⁻¹ strain rate[22, 23]. Controlled cortical impact (CCI) has been the most widely applied method, generating localized deformation in scaffold constructs at strains near 40 - 50% and strain rates of 10² - 10³ s⁻¹[24-31], with comparable application to organotypic slice cultures[32] and skull-confined organoids[33]. Weight-drop[34] and microvalve droplet-driven[35] system have produced localized injury in organoids, and pendulum or pneumatic impact has been applied neurons-on-chip and scaffolds[28] [36]. Among these platforms, several directly demonstrated a localized injury field through lesion imaging or construct-level FE modeling[26, 27, 29, 32, 34, 35], whereas others described focal or contusion injury without reporting the spatial strain distribution or pathological gradient that defines it[24, 25, 28, 31, 33]. Reported intended clinical phenotype varied, with most studies specifying only injury severity rather than a discrete phenotype. Across platforms, increasing strain magnitude and rate generally corresponded to greater injury marker upregulation, though thresholds varied with construct composition and mechanical parameter reporting remained inconsistent.

Global compression

Global compression applies spatially uniform loading across the entire construct, in contrast to the localized fields produced by indentation or impact. Uniaxial compression has been applied to organoids at strains of 3 – 50% and rates from quasi-static to approximately 90 s⁻¹ [37-39], to spheroids by centrifugation[40] and plate compression[41], and to hydrogel- and chip-based constructs[42]. Because loading is distributed across the whole construct, these platforms do not reproduce the localized strain field characteristic of focal contusion injury, a limitation confirmed in the few studies that examined the internal deformation field and found strain to vary with depth rather than concentrate at an injury center[40, 41]. Most compression studies characterized their model by the applied loading and intended severity, describing it as a mild or non-penetrating TBI, rather than mapping it to a specific clinical phenotype[38, 39, 41, 42]. When a focal phenotype was specified, compression served as a proxy for the compressive component of blunt impact rather than a reconstruction of contusion[37]. Global compression is therefore best understood as isolating a compressive loading regime relevant across TBI rather than modeling a discrete clinical injury.

Laceration and tissue disruption

Laceration models impose geometrically defined mechanical disruption to reproduce the tissue-severing component of penetrating injury, which clinically results from an object breaching the skull and lacerating the parenchyma, causing macroscale tissue disruption and axonal transection[2]. In vivo penetrating models and 2D scratch assays have traditionally been used to study bulk lacerations, as the often fatal nature of clinical penetrating injury limits experimental access. In 3D constructs, biopsy punches have been used to excise cylindrical regions of scaffold-based neural tissue, producing lesion volumes up to 25% of the construct[43], and pipette tips have been used to bore out discrete lesions in hydrogel-embedded neural cells[44]. These approaches generate defined injury boundaries that enable interrogation of the peri-lesional tissue, where secondary injury cascades are most clinically relevant. Both studies described their models as penetrating injury yet reproduced only the geometric tissue disruption of the clinical phenotype, not its cavitation, hemorrhage, foreign-material contamination, or infection[45]. These platforms are therefore more precisely described as tissue-disruption models, capturing the mechanical-laceration component of penetrating injury rather than reconstructing the full clinical entity.

Tension, torsion, and shear (inertial)

Tension, torsion, and shear reproduce the inertial loading generated by rotational head kinematics, which distribute shear strain across the brain parenchyma and clinically cause diffuse axonal injury (DAI) [15]. Biomechanical reconstructions confirm that rotational kinematics are among the strongest predictors of concussive injury, providing the clinical basis for this mechanics-to-phenotype link[46]. Imposing distributed shear rather than localized compression is more technically demanding than uniaxial impact, which may partly explain the underrepresentation of inertial paradigms among reviewed platforms. Reported approaches include rotational loading at 45 rad/s in glial hydrogel constructs[47] which is above the 50% concussion probability[46], cyclic tensile stretch of 4 – 20% at 0.25 s⁻¹ to hydrogel-encapsulated pericytes[48], and controlled fluid shear generating normal stresses on axons in a microfluidic system[49]. A neuron-on-chip system combining tensile stretch with torsion found that multiaxial loading lowered injury thresholds relative to either mode alone[50], consistent with the multiaxial character of clinical head loading. Across these platforms, studies incorporating neurons or isolated axons framed their model around DAI[49, 50], whereas the rotational and stretch constructs lacking a neurons described their injury only by severity, as mild TBI[47, 48]. Among the platforms modeling DAI, both provided evidence of axon-specific injury. Fluid-shear loading of a compartmentalized axon model produced graded axonal injury from focal swelling to axotomy, with a spatial distribution of damage along the axons, providing direct evidence linking shear loading to the axonal pathology that defines DAI[49]. Multiaxial stretch-torsion loading similarly produced axonal injury, indicated by increases in axonal markers of injury though the whole-construct geometry did not resolve the spatially distributed shear characteristics of clinical DAI[50].

Shock-wave exposure and cavitation

Shock-wave exposure and cavitation reproduce blast-related TBI (bTBI), which operates through pressure-wave propagation rather than externally applied bulk deformation. Primary blast waves induce a combination of direct pressure transmission, intracranial pressure gradients, skull flexure, cerebrospinal fluid dynamics, and cavitation, alongside secondary translational and rotational head motion, collectively representing a mechanically distinct and incompletely understood injury regime[51]. Blast loading applied across platforms spans a wide peak overpressure and impulse range. Organoids have been exposed to 900-1200 kPa at 600 kPa·ms using a tabletop blast simulator[52], 250-350 kPa at 1120 kPa·ms in pressure chambers[53], and 600 kPa with high-intensity focused ultrasound[54]. Microfluidic systems have been exposed to approximately 138-439 kPa using shock tubes[55] and shockwave generators[56], and scaffolds were subjected to 70-130 kPa at approximately 640 kPa·ms in blast chambers[57]. Collectively these conditions span sublethal to severe blast-relevant scenarios[58, 59], covering loading intensities relevant to both military and civilian blast exposure. Cavitation is a further mechanical contributor to bTBI[51]. Laser-induced inertial microcavitation within hydrogels reproduced high strain rates of 10³ -10⁸ s⁻¹, enabling quantification of cellular injury thresholds at high spatial resolution[60], and focused ultrasound-induced cavitation applied to scaffold-based neural constructs characterized bulk injury response and biomarker release over time[61].

Exposure schedule: single and repeated injury

While repeated injury is not a distinct loading mechanism, it is an exposure schedule that can be superimposed on the mechanical paradigms above. It is clinically significant, with 5.5% of individuals sustaining a recurring TBI, many within the first six months of initial injury[62]. Following the initial insult, the brain exhibits heightened sensitivity to secondary injury during a vulnerability window whose duration remains poorly defined[63], with athletes and military personnel at particular risk of cumulative effects including chronic traumatic encephalopathy (CTE)[64]. Despite this clinical importance, only four of the studies reviewed here incorporated repeat injury as impact and stretch loading, and none applied repetitive blast loading. Results were variable with repeated injury showing increased damage markers in the two impact-loading studies, where injury was repeated three times at 1-day intervals and twice across three days [28, 34], and no significant change in the two inertial loading studies where rotational injury was applied consecutively 10 and cyclic tensile stretch repeated 50 times [47, 48], likely reflecting differences in mechanical loading thresholds, inter-injury interval, strain magnitude, and platform composition. Although 2D in vitro studies have begun examining repeat blast[65] and blunt[66] injury, 3D platforms offer multicellular architecture and volumetric propagation of cumulative damage that monolayer systems cannot replicate, enabling interrogation of how repeated insults progressively disrupt recovery.

Mapping applied conditions onto clinical injury landscapes

The applied experimental mechanical conditions were mapped onto parameter landscapes benchmarked against human head injury data, with direct deformation platforms evaluated against FE reconstructions of concussive head impacts from contact sports and falls, and blast platforms evaluated against measurements from military training environments (Figure 2). FE head models provide these reconstructions by translating measured head kinematics into estimates of the tissue-level strain fields that loading produces within the brain, making them the principal bridge between external impact conditions and the internal mechanical response relevant to injury[67].

Figure 2. Physical loading of 3D in vitro traumatic brain injury (TBI) models benchmarked against clinical injury reconstructions.

Platform architecture is denoted by marker color across both panels, and marker shape denotes the loading method. Each data point represents a tested loading condition. (A) Applied strain and strain rate conditions across direct-deformation models. Shapes indicate uniaxial compression (global compression)[37-41], CCI[24-30] and weight drop[34] (indentation and impact), and stretch (inertial loading)[48]. Gray bands indicate proposed transition zones of cellular and tissue-level brain injury, spanning 10 - 20% strain and 1 - 10 s⁻¹ strain rate[19, 22, 23, 97]. The orange envelope represents FE-computed tissue-level strain and strain rate ranges from reconstructed head impacts spanning concussion through severe TBI[20, 21, 68]. (B) Peak overpressure and positive phase impulse across shock-wave models, with shapes indicating blast platforms that include shock tube[55], shockwave generator[56, 74], blast chamber[77], and pressure chamber[52, 53] systems. The gray shaded region represents the range of peak overpressure and impulse measured during sublethal blast exposures in operational training environments[58, 69]. The dashed red line (600 kPa) indicates an estimated reference overpressure above which lethal blast TBI outcomes are predicted to occur[59, 70]. Plotted overpressures are incident, applied values, not the transmitted or tissue-level pressures experienced by cells. Where not reported, derived strain, strain-rate, and impulse values are estimated as described in the main text. Shaded zones are illustrative benchmarks rather than absolute injury boundaries, as TBI occurrence and severity depend on injury mechanism, brain region, measurement uncertainty, and patient-specific variables.

For direct deformation platforms (Figure 2a), scaffold-based systems consistently achieved high strain and strain rates, reflecting their structural integrity under rapid loading, while organoids, spheroids, and hydrogel constructs spanned a broader range. The reviewed platforms also revealed a recurring gap between applied mechanics and stated clinical intent. Focal contusion injury was frequently named as the target yet was most often modeled through either localized impact without spatial characterization or spatially uniform compression. Many studies specified only injury severity (mild, moderate, severe) rather than a mechanism-defined phenotype. Most experimental conditions met or exceeded at least one proposed injury threshold spanning 10 - 20% strain and 1 - 10 s⁻¹ strain rate[19, 22, 23, 67] and a subset of hydrogel and organoid conditions overlapped with clinical concussion reconstructions[20, 21, 68]. The lower boundary of the clinical FE envelope aligns closely with the upper limit of the proposed strain-rate transition zone, indicating that concussive head impacts operate at or just above the strain rates at which tissue- and cellular-level injury thresholds have been proposed. Blast conditions generally operated in an intermediate zone between sublethal human training exposures measured by body-worn sensors[58, 69] and estimated lethal overpressure thresholds[59, 70], reflecting a key advantage of in vitro platforms for investigating injury regimes that cannot be ethically replicated in human subjects (Figure 2b).

Comparisons across these platforms are shaped by how mechanical loading was quantified. Strain and strain rate were reported directly as the deformation and rate applied to the construct, derived by the study through FE modeling, or calculated from reported parameters such as construct thickness, impact velocity, and penetration depth when strain and strain rate were not stated directly. Because these are spatiotemporal tensor fields reduced to scalar values through differing definitions and calculation schemes, the reported magnitudes are method-dependent and not strictly comparable across studies[12, 13]. Blast impulse, where not reported, was estimated from peak overpressure and positive-phase duration using a triangular approximation of the pressure-time waveform. This also separates the reported values from clinical references, as most in vitro studies report the nominal strain applied at the construct surface, but clinical FE reconstructions estimate tissue-level strain within the brain. A spheroid FE study showed this directly, finding that maximum principal strain within the construct exceeded surface-layer values as compression increased[41]. The clinical FE envelope itself also carries uncertainty, as the estimated tissue strains depend on the accuracy of the measured input kinematics[71], the type of head impact[72], and individual differences in brain morphology[73], so the envelope represents an approximate reference rather than a fixed boundary. The strain and strain-rate values in Table 1 and Figure 2a are therefore best read as approximate estimates. An analogous distinction applies to the blast landscape. Reported and plotted overpressures are incident or applied values, whereas the pressure reaching neural tissue is modified by transmission through the skull or device housing, intracranial pressure gradients, and cavitation, so the values in Figure 2b represent exposure conditions rather than tissue-level pressure. The sublethal human blast reference data in Figure 2b share this incident basis, having been recorded as body-worn sensor overpressure and impulse data, so the in vitro and human values are compared at the level of applied exposure. Additionally, several studies could not be placed in either landscape because mechanical parameters were unreported or given in device-specific units from which strain, strain rate, overpressure, or impulse could not be derived (Table 1).

Biological responses against human TBI pathology

The biological hallmarks of TBI span acute cell death and ionic disruption through neuroinflammation, neurodegeneration, and biomarker release, a progression examined in the following subsections and summarized across platforms in Figure 3. In evaluating these responses, it is useful to distinguish phenotypic coverage, whether a platform assesses a given injury response and detects a change in the expected direction, from biological fidelity, the quantitative, spatial, and temporal alignment with human TBI pathology. This section characterizes biological responses in terms of phenotypic coverage because most studies report the directional presence of injury effect, and because roughly half of the reviewed platforms use rodent-derived cells for which human alignment is not a uniform benchmark.

Figure 3. Phenotypic coverage of applied mechanics and biological readouts across 3D in vitro traumatic brain injury (TBI) platform architectures (n = 35 studies).

Each bar represents the number of studies assessing a given applied loading mechanics or biological readout category, with stacked sections indicating contributing platform architectures (color). Bar height within each platform zone is proportional to study count, aligned to a shared baseline reflecting total platform representation (leftmost bar, n = 35). Gaps within each platform bar indicate mechanics or readouts not assessed by that architecture. A readout category was counted as assessed regardless of whether a statistically significant difference from control was reported, so the figure reflects phenotypic coverage. Indentation and impact loading is the most represented across platform types, while laceration and tension, torsion, and shear (inertial) loading are the least represented, despite inertial loading being the predominant mechanism of clinical TBI. Among biological readouts, cell death and structural integrity is assessed across nearly every platform. Ionic and network dysfunction appears in organoid, scaffold, and microfluidic platforms, which provide the electrophysiological and imaging access these readouts require. Neuroinflammation is most frequent in scaffold systems, consistent with their multicellular composition and ability to incorporate microglia. Neurodegeneration and proteinopathy concentrate in organoid and scaffold platforms, reflecting their capacity for extended culture and genetic manipulation. Fluid biomarker release is assessed across most architectures, whereas metabolic dysfunction appears in relatively few studies and blood-brain barrier disruption is the least represented readout, reflecting the limited number of mechanically loaded 3D neurovascular models.

Cell death and disruption of cytoskeletal structural integrity

Neuronal loss and axonal injury are key pathological features of TBI[15]. Acute cell death and structural disruption are the most universally reported readouts, with viability loss and cytoskeletal fragmentation observed across platform architectures and applied mechanics. In studies that assessed temporal dynamics, cell death persisted or progressively increased, consistent with ongoing secondary injury cascades[26, 30, 34, 35, 37]. An exception was in oligodendrocyte progenitor cells, which showed no viability loss following blast at mild TBI-relevant conditions[74], reflecting differential cell-type vulnerability. High strain-rate loading revealed that dendritic structures fail at lower mechanical thresholds compared to microtubule and actin elements, reflecting distinct subcellular mechanical vulnerability[75]. Localized focal injury models reproduced spatially graded damage, with structural disruption diminishing with distance from the focal lesion[29, 32, 43, 44, 75], mirroring the topographic distribution of axonal pathology in postmortem human TBI tissue[76]. Axonal varicosities, a morphological hallmark of DAI[15], also appeared across blast[77], compression[40], and fluid shear loading regimes[49], with fluid shear producing a graded spectrum from focal swelling to complete axotomy. Multiaxial loading paradigms also triggered upregulation of injury markers tau and growth-associated protein 43 (GAP-43) at lower thresholds than uniaxial loading alone[50], and molecular indicators of oxidative axonal damage were elevated before gross cytoskeletal disruption was detectable[36, 55]. Genetic background further modulated vulnerability, with a microtubule-associated protein tau (MAPT) mutation exacerbating viability loss following compressive injury without amplifying tau pathology itself[37].

Ionic and network dysfunction

Mechanical disruption and compromise of the cell membrane drives depolarization, excessive glutamate release, and excitotoxic calcium influx [78], leading to network-level dysfunction [79]. Across platforms, glutamate was elevated for at least seven days following injury[27-29, 43, 77], consistent with cerebral microdialysis in patients[80], and showed a biphasic pattern with a second rise after 4 weeks coinciding with an excitatory shift[27], paralleling the excitatory/inhibitory imbalance in clinical post-traumatic epileptogenesis[81]. Electrophysiological recordings revealed broadly dysregulated activity, with high-frequency blast producing enhanced firing and desynchronized oscillations[53], conditioned media from injured constructs suppressing naive network activity[36, 55], and network recovery occurring within 48 hours[35]. Calcium dysregulation was strain dependent[38, 41, 42] and localized to focal axonal swellings sites[49].

Metabolic dysfunction

Metabolic dysfunction is characterized by an initial surge in glucose metabolism followed by prolonged depression persisting for weeks[78, 82]. In 3D platforms, mitochondrial depolarization occurred in a strain-dependent manner following compressive injury[37, 41], with partial recovery at one week[37], consistent with the transient metabolic crisis documented clinically. In multicellular scaffolds, mitochondrial fragmentation in microglia propagated dysfunction to astrocytes and neurons[30]. Selective inhibition of mitochondrial fission with P110 preserved network integrity and attenuated neuroinflammation in co-cultures but not monocultures[30], identifying glial mitochondrial crosstalk as a pharmacologically targetable secondary injury mechanism. Despite its clinical significance, oxidative stress, a downstream consequence of mitochondrial dysfunction, remains largely unassessed across current 3D platforms, limiting interrogation of pathways central to chronic injury progression.

Neuroinflammation

Mechanical insult elicited neuroinflammatory responses across 3D platforms[24, 25, 28, 30, 32, 34, 36, 44, 55], including pro-inflammatory cytokine release and glial activation consistent with patterns in human TBI postmortem tissue[83]. Most studies assessed neuroinflammation within the first 72 hours after injury, with few extending beyond one week[28, 31, 34], short of the months-to-years persistence documented clinically[83]. Cellular composition strongly shaped inflammatory response. A glia-only model subjected to rotational loading showed no significant cytokine or morphological changes[47], while neuron-only systems produced tumor necrosis factor alpha (TNF-α) upregulation following injury[36, 55], supporting a model in which neurons act as injury initiators and glia as amplifiers. Consistent with this, a neurovascular construct showed that inclusion of neurons and microglia was necessary for cytokine resolution and partial vascular repair after CCI, with tricultures lacking these populations exhibiting sustained inflammation without angiogenic re-engagement[31]. Sustained microglial activation that propagates neuroinflammation in TBI remains underrepresented across current platforms, and where microglia are included, rodent-derived microglia[32, 44, 47] and immortalized cell lines[24, 25, 31] predominate over human induced pluripotent stem cell (iPSC)-derived microglia[30]. Given well-documented divergence in transcriptional profiles and activation kinetics between human and rodent microglia[84], integration of human iPSC-derived microglia represents a necessary step toward capturing the neuroinflammatory dynamics most relevant to chronic post-TBI pathology.

Blood-brain barrier disruption

Blood-brain barrier (BBB) disruption following TBI occurs in an acute mechanically induced phase and a delayed phase driven by secondary injury cascades, persisting from hours to years and scaling with injury severity[85]. While 3D BBB models have advanced considerably, mechanical loading of these models remains limited, reflecting the technical demands of replicating cellular complexity alongside controlled injury delivery. CCI applied to scaffold-based constructs demonstrated that barrier disruption is cell-composition dependent. Pericyte inclusion preserved tight junction markers, while microglial paracrine signaling contributed to tight junction loss[24, 25]. A neurovascular construct extended this work, demonstrating acute junctional disruption and incomplete vascular recovery over five weeks, and showed that tracer diffusion was more restricted in the intact five-cell construct than in simpler configurations, consistent with the emergence of barrier-like properties[31]. Barrier-relevant functional readouts compatible with 3D neural constructs represent a critical next step toward connecting in vitro findings to the vascular dysfunction observed in patients.

Neurodegeneration and proteinopathy

TBI, particularly when repetitive, is an established risk factor for chronic neurodegenerative conditions including CTE, Alzheimer’s disease (AD) and related dementias, each characterized by distinct but overlapping proteinopathies involving tau, TAR DNA-binding protein 43 (TDP-43), and β-amyloid[86]. Compressive injury induced tau hyperphosphorylation and oligomerization in wild-type organoids in a maturity and severity dependent manner, generating tau oligomers without genetic manipulation[37]. Blast exposure produced tau pathology alongside TDP-43 nuclear egress and downstream splicing dysfunction[52, 54], resembling the co-pathology observed in human CTE postmortem tissue[86]. Repetitive mechanical injury activated latent herpes simplex virus type 1 (HSV-1) infection, driving β-amyloid plaque accumulation and AD-like pathology[28], demonstrating that mechanical trauma can trigger amyloid pathology through viral reactivation in this model. Organoids carrying disease-associated mutations showed neurodegenerative responses to mechanical injury[37, 54], enabling interrogation of genotype-specific injury susceptibility. This interaction is consistent with clinical evidence that pre-existing vulnerabilities such as genetic variants and comorbid conditions modulate TBI severity and recovery[1]. A limitation across these findings is temporal scope, as the relevant proteinopathies unfold over months to years clinically. Incorporating aged or maturation-accelerated iPSC lines may better approximate the vulnerability and trajectory of chronic post-TBI neurodegeneration in older populations.

Fluid biomarker release

Fluid biomarkers, including protein markers and extracellular vesicles[26], are an expanding tool in clinical TBI diagnosis and management, with glial fibrillary acidic protein (GFAP) and ubiquitin carboxyl-terminal hydrolase L1 (UCH-L1) receiving FDA clearance, and S100 calcium-binding protein B (S100B) integrated into European diagnostic frameworks[87]. Clinically these markers follow distinct temporal profiles reflecting their cellular sources and injury kinetics, with acute peaks for GFAP and UCH-L1 contrasting with the delayed rise of neurofilament light chain (NfL) over days to weeks[87]. A key strength of 3D platforms is their capacity to release canonical clinical TBI biomarkers in conditioned media across architectures[27, 32, 35, 37, 41, 50, 54, 77], with release showing mechanical dose dependence. NfL elevation increased with applied strain[41], phosphorylated-tau/tau ratios rose above a blast pressure threshold[54], and multiaxial loading increased NfL and tau at lower strain magnitudes than uniaxial loading alone[50]. NfL and UCH-L1 showed early peaks that declined over time[50, 77], consistent with their acute clinical profiles, while S100B showed delayed and sustained elevation[77], likely reflecting ongoing glial injury and the absence of clearance mechanisms[88]. Platform cellular composition constrains interpretability, as GFAP and S100B are predominantly glial in origin and neuron-only platforms cannot produce these markers regardless of injury severity. While acute temporal dynamics partially recapitulate clinical profiles, chronic biomarker trajectories remain poorly characterized, limiting direct comparison to the chronic profiles observed in patients.

Pharmacological Intervention

A subset of platforms has been used to test whether injury responses can be pharmacologically modified (Table 2). Reported interventions span ion channel modulation[54], mitochondrial fission inhibition[30], acrolein scavenging[36, 55], gabapentinoids[43], a stem-cell secretome[32], hypothermia[39], and a biomaterial implant [44], each reducing specific injury markers in the platform tested. These studies establish that 3D platforms can register intervention effects, but their translational interpretation is limited. Most applied a single agent at a single timescale in one platform, toxicity was assessed in five of the eight studies, and clinically achievable exposure was not reported. Of the interventions tested, only therapeutic hypothermia has been evaluated in human TBI trials, where it did not improve outcomes[89, 90] despite reducing injury in the model reviewed[39]. The other agents are either preclinical or approved for other indications, so these platforms have not been tested against interventions of known clinical TBI outcome, a prerequisite for predictive drug screening.

Table 2.

Pharmacological and physical interventions tested in 3D in vitro TBI platforms

Study	Intervention (class)	Target / mechanism	Platform; injury	Concentration	Treatment timing	Primary endpoint	Significant finding	Toxicity assessed a	Development status b	Refs
Rouleau et al.	Gabapentin, pregabalin (Ca2+ channel modulators)	α2δ subunit (voltage-gated Ca2+ channel)	Scaffold; laceration	100 μM	Immediately post-injury for 24 h, or redosed every 3 d for 10 d	Viability; electrical activity	Rescued electrical activity and prevented excitotoxicity	Yes	FDA-approved (epilepsy, neuropathic pain); not established as disease-modifying in TBI	[43]
Scimone et al.	Therapeutic hypothermia (physical)	Multiple (temperature-dependent)	Hydrogel; global compression	35, 33, 31.5 °C	Immediately post-injury, or delayed 1–12 h	Viability (24 h); caspase 3/7	Cooling to 33 °C improved viability and reduced caspase 3/7 within 4 h	Yes	Trialed extensively in human TBI; RCTs showed no consistent outcome benefit	[39]
Beauclair et al.	Hydralazine (acrolein scavenger)	Acrolein	Microfluidic Chip; shockwave	50, 100 μM	15 min post-injury	Neuronal activity (MEA spiking recovery)	Improved spiking recovery (87% vs 70% at 6 h); reduced acrolein and TNF-α	NR	FDA-approved (hypertension); not clinically tested in TBI	[55]
Rogers et al.	Hydralazine (acrolein scavenger)	Acrolein	Microfluidic Chip; impact	50 μM	4 h exposure immediately post-injury	Neuronal activity (MEA spiking); injury markers	Reduced acrolein, TNF-α, and Aβ42; improved spiking activity	Yes	FDA-approved (hypertension); not clinically tested in TBI	[36]
Liaudanskaya et al.	P110 (mitochondrial fission inhibitor)	Drp1	Scaffold; impact	1 μM	Immediately post-injury	Network integrity; neuroinflammatory cytokine release	Preserved network integrity and attenuated neuroinflammation in co-cultures but not monocultures	Yes	Preclinical (experimental peptide inhibitor)	[30]
Lai et al.	ML133 (KCNJ2/Kir2.1 inhibitor) c	KCNJ2 (Kir2.1)	Organoid; shock-wave	ML133 30 μM; ASO 10 μM	Immediately post-injury	Neuronal survival; TDP-43 nuclear localization	Reduced TDP-43 mislocalization and improved neuronal survival	Yes	Preclinical (experimental / genetic tool)	[54]
Pischiutta et al.	MSC secretome (biologic)	Multiple (paracrine)	Organotypic slice; impact	Conditioned medium	1 h post-injury	Cell death (contusion core); NfL release	Reduced cell death in the contusion core, lowered NfL release, improved microglial survival	NR	Preclinical (MSC-based; early-stage)	[32]
Wiseman et al.	Duragel implant (biomaterial)	Structural	Hydrogel; laceration	–	Immediately post-injury	Microglial infiltration; implant integration	Microglia infiltrated the implant region, supporting cellular integration at the lesion	NR	Biomaterial (repair scaffold); not a pharmacological agent	[44]

^aToxicity was scored as assessed only where the agent was applied to uninjured constructs or an explicit agent-only cytotoxicity control was included.

^bDevelopment status reflects each agent’s broader clinical or preclinical record, not validation within the cited platform.

^cKCNJ2 was identified by a CRISPR interference screen and validated with an antisense oligonucleotide (ASO); ML133 is the small-molecule inhibitor.

Abbreviations: ASO, antisense oligonucleotide; MEA, multielectrode array; MSC, mesenchymal stromal cell; NfL, neurofilament light; NR, not reported; RCT, randomized controlled trial; TNF-α, tumor necrosis factor alpha; TDP-43, TAR DNA-binding protein 43; Aβ42, amyloid-beta 42.

Concluding remarks and future perspectives

Across the recent 3D in vitro TBI literature reviewed, many platforms span clinically relevant strain, strain-rate, or pressure regimes, and reproduce selected TBI-associated phenotypes, with a subset integrating well-characterized mechanical insults alongside broad biological readouts. However, critical limitations persist, including inconsistent mechanical parameter reporting, insufficient cellular complexity and culture duration to capture chronic cascades and longitudinal pharmacological responses, and limited coverage of repeat injury paradigms. No single platform architecture is uniquely suited to all TBI research questions, and selection should be driven by the mechanical regime, cellular complexity, temporal scale, and species-specific biology of the question at hand.

Platform choice also involves cellular source. Nearly half of all platforms reviewed used rodent-derived cells, which retain value for modeling conserved injury mechanisms and for linking in vitro findings to the extensive in vivo rodent literature. Human-specific questions, including microglial injury responses, viral reactivation, and genotype-dependent vulnerability, require human iPSC-derived systems. However, human origin alone does not equate to adult-human fidelity and maturity. Clinical injury responses can differ between pediatric and adult TBI[91], yet responses that depend on mature or aging tissue are not captured by current iPSC-derived systems, which retain fetal-like immature characteristics[92]. Modeling aged or chronic phenotypes will require maturation and aging strategies[93, 94]. For patient-specific genotype effects, which so far have been examined in only a small number of platforms spanning tauopathy-associated MAPT and C9ORF72 amyotrophic lateral sclerosis (ALS) – frontotemporal dementia (FTD) backgrounds[37, 54], human iPSC-derived systems are essential as these features define the molecular targets and patient populations of candidate therapies. Advancing maturation, and pairing human genotypes with the appropriate injury paradigm, is therefore as important as the choice between human and rodent cells.

Realizing this translational potential will require advances in mechanical realism, biological completeness, and functional measurement. Rotational, inertial loading is the predominant mechanism of clinical TBI, yet most reviewed platforms apply compression or direct impact. Few implement the tension, torsion, or shear that reproduce inertial injury. Expanding this capability, along with sub-concussive and repeated exposures, would align platform mechanics with the loading that drives most human injury. Repeated-exposure paradigms would also clarify the post-injury vulnerability window, cumulative damage thresholds, and the mechanisms linking repeated insult to CTE. Standardized mechanical reporting, accompanied where feasible by FE characterization of construct-level strain fields, would make these comparisons rigorous across platforms. Biological alignment is a parallel need. Integrating vascular components, human immune cells, and extended culture timescales into a single platform with physiological perfusion, would enable interrogation of the neurovascular, inflammatory, and neurodegenerative cascades that define chronic TBI morbidity. Recent neurovascular constructs demonstrate that this integration is achievable[31], though functional barrier quantification remains a methodological challenge. Longer-lived platforms would also enable testing of emerging therapeutic strategies that promote regeneration and recovery after TBI[95]. Functional network-level readouts remain less common than structural and viability measures, reflecting the difficulty of electrophysiological recording in 3D, though emerging multielectrode arrays may better capture network dysfunction throughout the tissue[96].

Properly integrated and standardized 3D in vitro platforms, particularly those that utilize human cell sources will complement animal models and clinical studies by contributing a controlled setting that links defined mechanical inputs to human-specific molecular and cellular responses. Pharmacological studies show that these platforms can measure intervention effects, but their value for therapeutic prediction remains unproven until they are benchmarked against agents of known clinical outcome (Table 2). Even so, 3D in vitro methods are well-suited to pharmacological screening, patient-specific stratification of injury susceptibility using iPSC-derived platforms carrying disease-associated variants, and dose-response mapping across clinically relevant loading regimes. Decades of TBI clinical trials have shown how difficult it is to translate preclinical findings into effective therapies. Closing the gaps in cellular complexity, mechanical reporting, and injury paradigm coverage identified here will determine whether 3D in vitro platforms achieve the combined mechanical and biological validity that translation demands (see Outstanding Questions).

Outstanding Questions.

• How can vascular, immune, and extended culture timescales be integrated into a single 3D TBI platform without compromising mechanical fidelity or culture stability?

• What culture durations, cellular compositions, and longitudinal readouts are required to capture chronic neuroinflammatory, neurodegenerative, and pharmacological responses defining long-term TBI morbidity?

• How do sources of interindividual variability, including age, sex, comorbid conditions, prior injury history, and disease-associated genotypes, modulate mechanical injury responses, and can patient-derived iPSC systems capture this biological variability systematically?

• How do outcomes from single injury events extrapolate to repetitive exposures, and what inter-injury intervals define the window of cellular vulnerability?

• Can 3D platforms be used to identify therapeutic targets that translate to clinical efficacy, and what validation framework should connect in vitro pharmacological screens to subsequent in vivo and clinical testing?

• What cellular and molecular conditions promote regeneration and recovery following mechanical injury in 3D platforms, and do these recapitulate repair mechanisms observed clinically?

• How does the presence of latent viral infections, microbial products, or microbiome-derived signals interact with mechanical injury to shape neuroinflammatory and neurodegenerative outcomes, and can 3D neural platforms be adapted to model these host-pathogen interactions under defined injury conditions?

• What functional network-level measurements, from cellular membrane potential dynamics to population-level activity captured by emerging 3D multielectrode array technologies, most closely predict the cognitive and behavioral symptoms observed in TBI patients?

• What standardized reporting framework for mechanical parameters, cellular composition, and biological endpoints would enable rigorous cross-platform comparison and meta-analysis across 3D TBI studies?

Highlights.

• TBI causes immediate mechanical damage that triggers acute and progressive biological cascades driving long-term neurological disability.

• 3D in vitro neural platforms reproduce clinically relevant mechanical loading conditions and selected TBI-associated phenotypes under controlled conditions, offering a complement to animal models and clinical studies.

• Multicellular composition and injury paradigm complexity are critical determinants of biological phenotype coverage in TBI modeling.

• Mechanically injured 3D platforms reproduce selected post-traumatic proteinopathies, linking controlled mechanical insult to molecular markers associated with neurodegeneration.

• Pharmacological intervention in mechanically injured 3D platforms has reduced specific injury markers, illustrating their potential for mechanism-focused target screening.

Glossary

Blood-brain barrier (BBB)

A specialized vascular interface of endothelial cells, pericytes, and astrocytes that regulates molecular and cellular exchange between blood and brain.

Brain organoid

A self-organizing 3D neural tissue derived from stem cells that recapitulates selected developmental, cellular, and architectural features of the brain.

Cavitation

The rapid nucleation and collapse of microscopic bubbles in fluid or soft tissue under transient negative pressure, generating extreme local strain rates.

Controlled cortical impact (CCI)

An injury paradigm originally developed for in vivo rodent TBI in which a rigid piston impacts the brain surface at a defined velocity and depth.

Diffuse axonal injury (DAI)

A form of TBI pathology characterized by widespread axonal damage distributed across the brain, arising from rotational kinematics and shear strains.

Fidelity

The degree to which a model quantitatively reproduces clinically relevant aspects of human TBI, including both the applied biomechanical loading conditions and the resulting biological responses.

Finite element (FE) modeling

A computational method that divides a complex structure into discrete elements to simulate mechanical behavior under applied loading.

Hydrogel

A water-swollen polymer network used as a 3D cell culture substrate; used here for gel-phase systems such as collagen, Matrigel, or hyaluronic acid.

Impulse

The integral of pressure over time during the positive phase of a blast wave, measured in kPa·ms, reflecting the combined magnitude and duration of the pressure loading.

Microfluidic brain-on-chip

A microscale culture platform that uses channels, chambers, or controlled fluid flow to model selected brain tissue functions or injury responses.

Organotypic slice culture

An ex vivo brain tissue slice maintained in culture that preserves native tissue architecture and multicellular organization.

Peak overpressure

The maximum pressure above ambient atmospheric pressure generated by a blast wave, typically measured in kilopascals (kPa) or megapascals (MPa).

Phenotypic coverage

Which TBI-associated biological responses a platform assesses and reproduces. Distinct from biological fidelity.

Poisson's ratio

The ratio of transverse contraction to longitudinal extension under loading. Values near 0.5, as in brain tissue, indicate near-incompressibility.

Scaffold

A 3D biomaterial architecture providing structural support for embedded or seeded cells; used here for sponge-based constructs.

Shear

Deformation produced by forces acting parallel to a surface, causing adjacent layers of tissue to slide relative to one another.

Spheroid

A 3D cellular aggregate with simpler organization than an organoid, typically lacking defined regional patterning or complex cytoarchitecture.

Strain ($\epsilon$)

A dimensionless measure of deformation defined as the change in length divided by the original length of a material.

Strain rate ($\stackrel{.}{\epsilon }$)

The rate at which strain changes over time, measured in inverse seconds (s⁻¹).

Young’s modulus (E)

A measure of a material's stiffness, the ratio of tensile stress to strain, expressed in pascals (Pa). Lower values indicate softer tissue.