A 3D Human Neuron-on-Chip Platform to Monitor Neuronal Injury Responses

In vitro models, Neuron-on-Chip, microphysiological systems, neuronal injury, calcium imaging, neurodegeneration, neuroinflammation

Traumatic brain injury (TBI) is a major cause of neurological dysfunction and long-term neurodegeneration, yet the intrinsic neuronal contributions to TBI pathophysiology remain incompletely defined. Here, we present a novel Neuron-on-Chip microfluidic platform that can be used to mechanically injure mature human prefrontal cortex neurons (hPFCs) embedded in three-dimensional (3D) hydrogels, enabling the study of injury responses in pure neuronal cultures. Real-time calcium dynamics across 13 metrics of single-cell and network activity reveals biphasic injury response: an early phase (0.5–24 h) characterized by excitotoxicity, hyper-synchronized bursting, and network collapse; and a late phase (8 d) marked by sustained depolarization and structural remodeling. Secretome profiling uncovers progressive elevations in extracellular pT181 and total Tau from days 1 to 5 post-injury. Cytokine analyses identify early (24 h) elevations in IP-10, IL-10, IFN α 2, and NCAM, and late increases (8 d) in CXCL9 and MPO, linking neuronal activity changes to stage-specific inflammatory signaling. Immunocytochemistry and immunoblotting confirmed temporally ordered upregulation of calpain-1 and caspase-3 (days 1–3), phosphorylated Tau (AT8+, days 5–8), and neurofibrillary tangle-like Tau aggregates (NFT+, day 8). These findings establish our platform as a scalable microphysiological model for probing the dynamic cellular and molecular sequelae of neuronal response to injury, offering insights into neurodegeneration and opportunities for therapeutic discovery. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 2 1. Introduction Traumatic brain injury (TBI) is a leading cause of long-term sensorimotor and cognitive impairment, largely driven by disruption of neuronal network function [1]. The pathophysiology of TBI is multifaceted, encompassing mechanical damage, oxidative stress, neuroinflammation, and cell death. In addition to these acute effects, TBI is increasingly recognized as a trigger for accelerated aging and cellular senescence [2-4], and as a significant risk factor for late-onset neurodegenerative diseases, including Alzheimer’s disease (AD)[5]. A shared hallmark between TBI and AD is tauopathy, characterized by the pathological misfolding and aggregation of Tau protein. These aggregates can propagate in a prion-like manner across neuronal circuits, amplifying neurodegenerative cascades [6, 7] . Notably, Tau accumulation is strongly correlated with cognitive decline, even in the absence of amyloid- β (Aβ ) pathology[8]. Clinical studies have consistently shown elevated levels of total Tau (tTau) and phosphorylated Tau (pT181) in cerebrospinal fluid (CSF) and extracellular fluid following TBI, with persistent accumulation of Tau aggregates in the brain linked to poor outcomes [9-13]. While post-TBI neuroinflammation has been implicated in facilitating Tau aggregation and neuronal internalization, these mechanisms are typically attributed to the actions of astrocytes, microglia, and endothelial cells [14, 15]. The intrinsic capacity of neurons to independently initiate or sustain these processes — particularly in the absence of glial or vascular components — remains unclear. Electrophysiological techniques such as whole-cell patch clamp have been widely employed to assess membrane properties and injury-induced changes at the single-neuron level [16, 17]. However, their low-throughput nature limits utility for evaluating large-scale network behavior. In contrast, calcium imaging and multi-electrode array (MEA) technologies offer simultaneous multi-neuron recordings and spatiotemporal insights into network dynamics. While MEAs lack single-unit resolution and are less suited for 3D cultures, calcium imaging—especially when combined with genetically encoded sensors and optogenetic actuators—has emerged as a powerful alternative for resolving both single-cell and network- level activity with high fidelity [18]. These recordings, when integrated with analytical frameworks such as spike estimation [19], waveform analysis [20], and graph theory-based connectivity analysis [21], provide a robust toolkit to probe the functional consequences of neuronal injury. In vitro TBI models serve as an ethical and mechanistically accessible platform for studying post-injury responses, while reducing reliance on animal experiments. Two-dimensional (2D) in vitro systems have been extensively used for drug screening and therapeutic testing [22-25], (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 3 yet they fall short in replicating the biomechanical and cellular complexity of native brain tissue. In contrast, 3D neural organoid culture models offer a more physiologically relevant microenvironment, enhancing cell-cell and cell-matrix interactions, and enabling the modeling of tissue-level shear and compression forces seen in vivo [26, 27] . Despite these advantages, neural organoid models have several key limitations. These include the need for long-term cultures to establish organoids, high variability and limited reproducibility between organoids, presence of a necrotic core, cellular heterogeneity, and large size, which prevents the use of live-cell calcium imaging methods for acquiring high-resolution single and multi- unit neuronal activity recordings. A temporally resolved 3D in vitro model that overcomes these limitations could reveal mechanistic relationships among neuronal network disruption, inflammation, and neurodegeneration in injured neurons. To address this need, we developed a novel human Neuron-on-Chip platform that facilitates controlled weight-drop injury of human pluripotent stem cells (hPSCs) derived human prefrontal cortex (hPFC) neurons embedded in 3D hydrogels. We monitored the injury progression of weight- drop injured hPFC neurons over a period of eight days in vitro. Neuronal activity was characterized via calcium imaging using 13 quantitative descriptors of single-cell and network function. These functional data were integrated with longitudinal secretome profiling, immunocytochemistry, and western blotting to assess inflammatory signaling and intracellular/extracellular Tau dynamics. Our model offers a scalable, mechanistic, and ethically aligned system to dissect neuronal contributions to TBI pathology and identify early biomarkers of neurodegeneration. 2. Results 2.1. Weight-Drop Injury Induces Force-Dependent Neuronal Cell Death in a 3D Human Neuron-on-Chip Model. We engineered a custom 3D human Neuron-on-Chip platform for long-term neuronal culture and injury induction of hydrogel-embedded hPFC neurons. A 3D-printed negative mold of the device was used to cast polydimethylsiloxane (PDMS) devices that were adhered to a 35 mm glass-bottom dish ( Figure 1 A–B). Each PDMS device consisted of included four media reservoirs (R1–R4) supplying nutrients to two cell-gel chambers (C1 and C2), into which hPSC-derived hPFC neurons were encapsulated within Geltrex™ hydrogels (Figure 1C). Media exchanges were conducted every 24 h. Following two weeks of culture, immunofluorescence confirmed neuronal network maturation, as indicated by co-expression of the neuronal maturation marker β III-tubulin and the synaptic protein Synaptophysin-1 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 4 (Syn-1), colocalized with the cytoskeletal marker F-actin (Figure 1C). Supporting this, EdU incorporation assays ( Figure S1A) showed minimal proliferation, while HuC/D+ quantification indicated robust neuronal enrichment (Figure S1B–C), and GFAP+ and Olig2+ staining revealed minimal glial contamination, confirming a highly pure neuronal population (Figure S1B–C) after 50 days in culture. To induce focal injury, we applied a mechanical weight-drop paradigm using a custom- designed impactor device (Figure 1D). A 6 g weight was dropped from 10, 15, or 20 cm heights, delivering ~6, ~9, and ~12 mJ of impact energy, respectively, to a ~0.5 mm-diameter tip positioned above the neuron-laden hydrogel. The impact transiently deformed the flexible PDMS roof, resulting in localized trauma to the underlying 3D neuronal culture. Cell viability was assessed using Hoechst/Calcein AM/Propidium Iodide (PI) staining. Results showed a graded reduction in neuronal survival and soma size corresponding to increasing impact force (Figure S2 A–B). Based on moderate cell death and preservation of network integrity, we selected the 9 mJ condition (15 cm drop) for all subsequent experiments. Longitudinal assessment of cell viability via Hoechst/PI staining revealed a modest elevation in PI+ cells (<5%) at 0.5 h post-injury compared to uninjured controls (Figure S2C). A marked increase in cell death (~10%) was observed between 24 h and 8 days post-injury (Figure S2C), suggesting a progressive, delayed degenerative response. Figure 1. (3D Human Neuron-on-Chip device design, fabrication, and characterization. (A) 3D printed negative mold of the device. (B) Schematic of PDMS human Neuron-on-Chip prototype placed on a 35 mm glass-bottom petri dish. (C) Human Neuron-on-Chip device containing hydrogel (Geltrex TM) encapsulated hPSC-derived hPFC neurons in reservoirs C1 and C2 that are fed via media reservoirs R1-4. Representative confocal image of immunocytochemically stained mature neurons ( β -III-Tubulin, F-Actin and Syn1). Scale – 1 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 5 mm. (D) Weight drop impactor design and setup. (E) Experiment timeline to assess the temporal effects of weight-drop injury on neuronal activity and network dynamics.) 2.2. Weight-Drop Injury Elicits Persistent Neuronal Silencing and Biphasic Temporal Dynamics in Network Activity. To characterize the effects of mechanical injury on neuronal function, we performed calcium imaging using the Fluo-4 AM dye at five post-injury time points (0.5 h, 24 h, 72 h, 5 d, and 8 d) across three distinct regions of interest (ROIs). At each time point, devices were subsequently analyzed for end-point specific viability, immunocytochemistry (ICC), western blotting, and cytokine profiling as described below. Due to the need for these endpoint- specific assays, the same neuronal populations were not tracked longitudinally. Calcium imaging confirmed spontaneous spiking activity in hPFC neurons ( Figure 2 A). Neuronal calcium traces were quantified across 13 parameters encompassing single-cell dynamics and network function. These included: Calcium kinetics: mean Rise Time (mRT), Fall Time (mFT), and Bandwidth (mBW), indicative of intracellular calcium handling and burst duration; Signal magnitude: mean Amplitude (Amp), reflecting calcium influx [28, 29] ; Spike metrics: number of active neurons (AN; ≥ 5 spikes/min) and mean Firing Rate (mFR), derived from spike estimation algorithms [20]. Graph theory was employed to characterize neuronal network structures and identify the presence and connection strength of neuronal subcommunities [30]. Network properties: phase-locking value matrices [31] were used to compute weighted Node Degree (wND) and Path Length (wPL), reflecting cell-to-cell functional connectivity; number of clusters (NoC) and weighted Modularity (wM), representing network community structure; Global Synchronization Index (GSI), indicating the level of synchronized activity; weighted Global Efficiency (wGE), indicating network- wide efficiency; and weighted Clustering Coefficient (wCC), capturing local connectivity patterns [32, 33]. Hierarchical bi-clustering of covariance matrices across time and treatment revealed distinct injury-induced patterns (Figure 2B). Compared to uninjured controls, injured neurons showed consistent negative covariance with AN and mFR, suggesting diminished overall activity. Quantitatively, AN was significantly reduced at all post-injury time points except 0.5 h (Figure 2C), and mFR was significantly decreased at all time points except 72 h (Figure 2D), indicating persistent neuronal silencing over the 8-day post-injury period. Analysis of covariance patterns further identified two temporally distinct phases of network response. Early-stage injury (0.5–72 h) was characterized by positive covariance of wRT, wFT, Amp, (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 6 wND, wPL, and GSI, relative to controls, whereas these parameters shifted to negative covariance in the late stage (5–8 d). Conversely, NoC, wM, wCC, wGE, and mBW showed opposite trends - demonstrating negative covariance in early injury stages and positive covariance in late injury stages, relative to controls (Figure 2B). Dimensionality reduction via t-distributed stochastic neighbor embedding (t-SNE) and statistical significance test using permutational multivariate analysis of variance (PERMANOVA) analysis revealed significant divergence in multivariate activity profiles between injured and control groups at 0.5 h, 24 h, and 8 d ( Figure S3 ). These analyses support a biphasic trajectory in post-injury network dynamics, delineating early (0.5–24 h) and late (8 d) functional states. Figure 2. (Calcium imaging analysis indicates temporal changes in hPFC neuronal activity after injury. (A) Schematic of calcium imaging analysis workflow. (B) Heatmap of the results from bi-clustering analysis of 10 sample groups across 13 neuronal function parameters, revealing distinct neuronal function patterns of early (0.5-72 h post injury (In)) and late injury responses (5-8 d post injury (In)). Number of active neurons (C) and Mean firing rate (D) reveal continuous silencing of hPFC neurons post-injury.) 2.3. Early Injury Response Is Marked by Synchronized Burst Activity, Dysregulated Calcium Kinetics, and Excitotoxic Apoptosis. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 7 To dissect the early-stage functional response to traumatic injury, we analyzed 11 calcium imaging-derived parameters (excluding AN and mFR, which showed global reductions across all post-injury time points) within the first 72 h post-injury. A hallmark of early injury was significantly elevated global synchronization index (GSI) at both 0.5 h and 24 h post-injury relative to uninjured controls ( Figure 3A), indicating increased synchronization of neuronal firing. Temporal profiles of calcium traces showed abnormally compact and synchronized burst activity post-injury (Figure 3B), consistent with early hyperexcitability. This was accompanied by significant increases in mean Rise Time (mRT), Fall Time (mFT), and Bandwidth (mBW) (Figure 3C-E), suggesting prolonged intracellular calcium elevation and impaired calcium clearance. Notably, no significant differences were detected in calcium signal amplitude (Amp) during this period, implying that excitotoxic responses were driven by altered dynamics rather than the difference in the magnitude of calcium influx during neuronal depolarization. At the network level, early injury induced a marked increase in functional connectivity, as evidenced by elevated weighted Node Degree (wND) and Path Length (wPL) ( Figure S4A– B; Figure S5). However, this increased connectivity coincided with significant reductions in modularity (wM) and number of functional clusters (NoC) (Figure S4C–D), indicating a collapse of organized network communities. Furthermore, global network efficiency (wGE) was significantly reduced at early time points ( Figure S7 A), supporting a shift toward disorganized and energetically inefficient network activity. Waveform analysis of neuronal activity patterns suggested injury-induced excitotoxicity due to excessive calcium influx in excitatory hPFC neurons. Excessive calcium influx is known to activate calpains, a major family of calcium-dependent cysteine proteinases that can contribute to neurodegenerative Tau fragmentation and neuronal apoptosis [34, 35] (Figure 3F). The patterns of heightened synchronization and prolonged calcium transients observed in our

are consistent with excitotoxic signaling. To assess biochemical correlations of excitotoxic stress, we examined the activation of calpain-1 and caspase-3, two calcium- sensitive proteases implicated in neuronal degeneration. Western blot analysis of cell lysates collected between 0.5 and 72 h post-injury revealed a significant increase in the activated autolytic fragment of calpain-1 (~55 kDa), with a non-significant trend toward elevated 32 kDa fragments (p = 0.0573) (Figure 3G, I–J, Figure S5A-B). Full-length calpain-1 (~80 kDa) showed only a modest, non-significant increase. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 8 Consistent with calpain-mediated apoptotic signaling, both full-length (32 kDa) and cleaved (active, 19 kDa) forms of caspase-3 were significantly elevated in injured neurons compared to controls over the same time window (Figure 3G, K–L, Figure S5C-D). These findings support the conclusion that early neuronal hyperactivity leads to calcium overload, triggering calpain-1 autolysis and downstream caspase-3 activation, thereby initiating excitotoxic apoptotic pathways in the absence of glial involvement. Figure 3. (Early injury neuronal response is characterized by highly synchronized burst activity that is driven by excitotoxicity and apoptosis. (A) Global Synchronizing Index (GSI) demonstrates the highly synchronized neuronal activity observed acutely post-injury. (B) Firing event distribution from Control and Injured groups immediately (0.5 h) post-injury. (C- E) Waveform analyses showing neuronal injury-induced changes in mean rise time (C), mean fall time (D), mean band width (E), acutely post-injury. Two-way ANOVA with Fisher’s LSD test multiple comparison, *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. n = 3-5 per time (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 9 point from 4 independent experiments. (F) Schematic showing mechanisms of calcium influx induced Calpain activation and downstream effects on caspase-3 and tau truncation. (G) Western blotting results of calpain-1 and caspase-3 levels 24-72 h post injury (In). (H-L) Quantification of calpain-1 and caspase-3 fragments detected acutely (0.5-72 h) post injury. Lognormal Welch’s T test, *p < 0.05, **p < 0.01, n = 1 per time point from 2 independent experiments.) 2.4. Late Injury Response Features Sustained Neuronal Depolarization, Network Fragmentation, and Enhanced Community Structure. In contrast to the widespread alterations observed during the early injury phase, the late-stage injury response (5–8 days post-injury) was characterized by more selective but significant changes in network architecture and calcium dynamics. Notably, injured neurons exhibited sustained depolarization at both 5- and 8-days post-injury relative to uninjured controls (Figure 4A), indicating long-term disruption of membrane potential homeostasis. At the network level, injured neurons showed significant reductions in wND and wPL at late time points (Figure 4B&C; Figure S6 ), indicating weakened connectivity and loss of long- range functional links. Concurrently, an increasing trend in NoC was observed, with significant elevations at 8 days post-injury. wM also increased significantly at 5 and 8 days (Figure 4D–F), suggesting reorganization of neurons into smaller, more segregated communities. These changes were partially accompanied by a rising trend in wGE (Figure S7A) and a statistically significant increase in wCC at 5 days post-injury (Figure S7B), reflecting enhanced local neighborhood connectivity. Additionally, minor but consistent upward trends in mRT, mFT, and mBW were observed in injured neurons over this period (Figure S8 B–D), suggesting mild dysregulation in calcium kinetics remained even at later time points. Biochemical markers of cell death and excitotoxic stress diverged from the early-stage profile. Calpain-1 cleavage products remained unchanged compared to controls at late time points (Figure S5A-B), and while full-length caspase-3 levels significantly decreased in injured neurons at 5–8 days, the active cleaved form (19 kDa) was no longer detectable (Figure S5C- D). These findings suggest a resolution or exhaustion of the apoptotic response in the late injury phase. To further probe structural changes in network topology, we performed differential network analysis [36] restricted to the 8-day post-injury time point—chosen due to the significant (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 10 divergence between injured and control groups across all 13 functional parameters as confirmed by PERMANOVA. This analysis revealed that NoC functioned as a central network hub negatively correlated with Amp, wPL, and GSI, and positively correlated with wGE (Figure 4G). These results indicate that injury-induced fragmentation of large-scale connectivity is accompanied by a compensatory increase in modular community structures, and that community reorganization is closely linked to signal synchronization, calcium amplitude, and network efficiency. Figure 4. (Weight-drop injury induces neuronal depolarization, loss of long-range connectivity, and rearrangement of neuronal community structure. (A) Amp of calcium signal demonstrates neuronal depolarization 5-8 days post injury. (B, C) nD and PL reveal acute increase and long-term loss of functional connections. (D) Representative heatmaps of pair- wise phase locking matrix from control and injured groups 0.5 h and 8 d post injury, reveal the remodeling of functional network structure in response to the injury. (E, F) NoC (E, neuronal subcommunities) identified by an eigen-value-based algorithm, and Modularity (F, presence of neuronal sub-communities) demonstrate acute disruption and eventual rearrangement of neuronal community structure in response to injury. (G) Differential network analysis of injured Vs control groups 8-days post-injury reveals injury specific correlation between NoC, AmP, wPL, and GSI. Two-way ANOVA with Fisher’s LSD test multiple comparison, *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. n = 3-5 per time point across 4 independent experiments.) (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 11 2.5. Secretome Profiling Reveals Extracellular Tau Accumulation and Temporal Regulation of Inflammaging Associated Proteins Post-Injury. To identify injury-induced changes in secreted neurodegenerative and inflammatory markers, we analyzed the conditioned media from injured and uninjured hPFC neuron cultures over four time points: 24 h, 72 h, 5 d, and 8 d post-injury. Log /i4 fold changes were computed to compare the abundance of secreted proteins between injured and control groups. Among neurodegeneration-related proteins, we observed a sustained increase in extracellular phosphorylated Tau (pT181) across the first 5 days post-injury in the injured group, indicating persistent release of pathologically modified Tau isoforms ( Figure 5A). This was accompanied by elevated levels of total Tau, suggesting broad Tau dysregulation in response to mechanical insult. In contrast, extracellular levels of A β 40 and A β 42 progressively declined post-injury, diverging from Tau trends and suggesting distinct regulatory mechanisms governing amyloid and Tau dynamics in neuron-only cultures (Figure 5A). Temporal clustering of 24 inflammatory and neurodegeneration-associated proteins revealed three secretion profiles: 12 proteins exhibited early release (peak at 24 h), 8 proteins showed mid-phase release (peak at 5 d), and 4 proteins displayed late release (peak at 8 d), indicating a structured temporal cascade in the neuronal secretome post-injury. Spectral clustering of protein-protein interaction networks identified two major clusters with distinct functional and temporal signatures (Figure 5B). The yellow cluster (early-phase hub) contained IP-10, IL-10, NCAM, and M-CSF—factors that formed central nodes connected to both mid-phase (IL-4, VEGF-A) and late-phase factors (CXCL9, PDGF-AA). The purple cluster included predominantly mid-release proteins, such as FGF2 and IL-1 family cytokines, which also showed the highest fold changes across all time points. Cathepsin D (CatD) was identified as a key regulator bridging interactions between the two clusters. To explore relationships between secretome dynamics and functional neuronal outcomes, we performed sparse canonical correlation analysis (CCA) [37] using the 13 calcium imaging- derived activity parameters (Figure 5C). CCA separated these parameters into two functional groups: Group 1 (Red): Early injury-associated metrics, including AMP, GSI, wND, and wPL; and Group 2 (Blue): Mid-to-late stage parameters, including wGE, NoC, and wM. Secreted proteins strongly correlated with Group 1 included early-phase cytokines IP-10, IL- 10, NCAM, and IFN α 2 (CCA weight > 0.25), indicating tight association with initial network hyperactivity and connectivity changes. In contrast, Group 2 activity parameters correlated most strongly with late-phase proteins such as CXCL9 and MPO, along with tPAI1, which peaked earlier but exhibited delayed associations with network remodeling. Additional (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 12 secreted factors-VEGF-A, IL-4, BDNF, MDC, pT181 Tau, tTau, PDGF-AA, Eotaxin, and IL- 1demonstrated significant correlations with neuronal activity but were assigned lower CCA weights (<0.25), suggesting more modest or indirect contributions to functional dynamics. Together, these findings indicate that Tau release and injury-triggered inflammatory signaling are temporally structured, and that secreted factors—particularly those in the early-phase yellow cluster—are tightly coupled to specific features of network activity and remodeling. This reinforces the functional interdependence between neuronal secretome composition and electrophysiological state following mechanical injury. Figure 5. (Secretome analysis of inflammatory and neurodegenerative proteins validates extracellular Tau localization, and shows temporal regulation of inflammatory factors, post- injury. (A) Log2 Fold change of secreted proteins in injured vs control groups across time indicates temporal regulation and classification into three clusters - Early, Mid, and Late release. Data are representative of 2-3 independent experiments. (B) Network analysis and spectral clustering analysis reveal 2 functional clusters within 28 secreted factors and their hub proteins. (C) Sparse canonical correlation analysis identified the high association of IP10, IL10 and NCAM with early injury state neuronal function; and CXCL9, MPO, tPAI1, with late injury state neuronal function.) (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 13 2.6. Weight-Drop Injury Induces Intracellular Tau Aggregation that is Correlated with Neuronal Activity Disruption. Given the persistent elevation of extracellular total Tau and phosphorylated Tau (pT181) identified in secretome analyses, we next investigated intracellular Tau dynamics in injured hPFC neurons using western blotting and immunocytochemistry (ICC). Western blot analysis revealed a significant reduction in intracellular tTau (total Tau, Tau46) at 24 h, 5 d, and 8 d post-injury compared to uninjured controls ( Figure 6A&B), suggesting enhanced Tau release or degradation following neuronal injury. In contrast, levels of intracellular phosphorylated Tau (pT181) were significantly elevated across all time points post-injury (Figure 6A&C), consistent with abnormal Tau phosphorylation and aggregation. Other phospho-Tau species, including pTau231 and pS396, were also examined. While pTau231 showed no significant change, pS396 displayed a trend toward upregulation in injured neurons (p = 0.0594; Figure S9), indicating potential isoform-specific involvement in injury-induced Tau pathology. To further characterize Tau aggregation and maturation, we performed ICC on neurons previously analyzed by calcium imaging. Matched-pair analysis using antibodies against hyperphosphorylated Tau (AT8; paired helical filament-positive) and neurofibrillary tangle (NFT)-positive Tau revealed significant increases in AT8+ Tau in injured neurons at 5 and 8 days post-injury (Figure 6D&E). Additionally, NFT+ Tau was markedly elevated at 8 days in the injured group compared to controls (Figure 6F&G), confirming the late-stage accumulation of aggregated Tau structures. These findings establish a clear link between injury-induced changes in neuronal activity and the intracellular accumulation and aggregation of pathogenic Tau species, aligning with extracellular Tau release dynamics and supporting the progressive development of tauopathy- like features in pure neuronal cultures following weight-drop neuronal injury. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 14 Figure 6. (Weight-drop injury induced the intracellular accumulation of pathological Tau isoforms in hPFC neurons seeded in 3D Neuron-on-Chip devices. (A) Western blots of pT181 Tau, total Tau (Tau46) and β -actin loading controls from injured and control groups across time points. (B-C) Quantification of results from (A). Two-way ANOVA with Fisher’s LSD, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. Data represent 3 independent experiments. (D) Representative confocal image of immunocytochemically stained AT8 (pathological Tau maker). (E) Quantification of AT8 area demonstrating significant AT8 upregulation 5- and 8- day post injury in injury group, compared to the control. (F) Representative confocal image of immunocytochemically stained NFT (neurofibrillary tangles). (G) Quantification of NFT area demonstrating significant NFT upregulation 8-day post injury in injury group, compared to the control. Scale – 1 mm. Two-way ANOVA with Bonferroni’s multiple comparison, *p<0.05. n = 8-10 images per device, 3 devices per group, from 2 independent experiments.) 3. Discussion Neuronal injury is characterized by acute neurotoxicity and widespread disruption of activity patterns, often resulting in the rewiring of neuronal circuits. While these phenomena have been well documented in in vivo and mixed cell culture models, the direct effects of injury and injury-induced neurodegeneration in genetically unaltered, neuron-only cultures remain less understood. Using a 3D hydrogel-based Neuron-on-Chip platform incorporating hPFC neurons, we demonstrate that mechanical injury via weight-drop leads to temporally dynamic changes in both single-unit and network-level activity. These activity changes are accompanied by a tightly regulated and temporally distinct inflammatory secretome, which together appear to promote neurodegenerative tau pathology. Immediately post-injury, neurons exhibited hyperexcitation driven by glutamate-induced excitotoxicity and the acute loss of inhibitory input-hallmarks of TBI pathology [38, 39] . By (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 15 modeling injury responses using a purified excitatory hPFC neuronal population, we aimed to isolate cell-intrinsic mechanisms of damage response. These hPSC-derived neurons, predominantly cortical projection neurons [40], are well-suited for modeling neurodegenerative responses[41-43]. Through calcium imaging, we captured real-time intracellular Ca²/i4 dynamics and network-level firing patterns, providing insights beyond the temporal and spatial

of MEA or patch clamp approaches. Our results reproduced classical excitotoxic responses to injury, including calcium dyshomeostasis and synchronized burst firing. These findings are consistent with previous work reporting increased membrane permeability [44, 45], dysregulation of voltage-gated Na /i4 channels[46, 47], and glutamate-induced excitotoxicity [48]. The increased cleavage of calpain-1 observed via western blot further confirmed calcium dysregulation. Calpain, a calcium- activated protease, undergoes autolysis in the presence of elevated Ca² /i4 levels [49], thereby facilitating excitotoxic damage by proteolyzing voltage-gated Na /i4 channels [50, 51] and promoting tauopathy. Our functional connectivity analysis revealed that early post-injury synchrony was followed by significant reductions in connectivity strength, depolarization, and reorganization of neuronal clusters. These late-phase changes suggest long-term depression (LTD)-like plasticity, potentially mediated by decreased calcium influx, a known trigger for shifting synaptic strength from potentiation (LTP) to LTD [52, 53] . The reduced network amplitude observed aligns with these LTD-associated mechanisms, characterized by lower wPL and increased community isolation. These data present a comprehensive view of the temporal evolution of neuronal injury responses across both cellular and network dimensions. To investigate molecular correlations of these functional transitions, we profiled secreted proteins across injury time points. While neuronal production of inflammatory cytokines is known [54-71] , our study provides one of the first high-resolution temporal maps of cytokine secretion from pure neuronal populations. We identified 24 secreted proteins, many of which play roles in regulating neuronal function and inflammatory signaling [65, 66, 72-78] . Secretome dynamics revealed temporally defined waves of release: early (e.g., PAI-1, BDNF, IL-8, IP- 10), mid (e.g., IL-1 family, PDGF-AA), and late-phase (e.g., IL-13, MPO) factors. Notably, many of these proteins are also upregulated within 3 days post-TBI in both human and animal studies [79-83]. We propose that early-stage excitotoxicity and hyperactivity likely drive the release of activity-regulated factors, as suggested by high canonical correlations between IP-10, IL-10, IFN α 2, NCAM, and injury-induced activity changes. IP-10 and NCAM, in particular, are (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 16 upregulated via calcium-dependent gene expression mechanisms[84-86], and the involvement of CDK5 and calpain in tau hyperphosphorylation has been well established. While direct activity-dependence of IL-10 and IFN α 2 remains uncertain, their tight association with early injury responses suggests indirect regulation through other neuronal signaling cascades. Network analysis further revealed IP-10, NCAM, IL-10, M-CSF as central hubs in autocrine regulatory circuits, indicating a mechanistic link between neuronal activity and inflammatory secretome progression. Interestingly, we found that many injury-induced secreted proteins are components of the senescence-associated secretory phenotype (SASP), which exerts autocrine and paracrine effects in aged tissues [87-91]. These include early-phase proteins (e.g., IL-8, GM-CSF), mid- phase (e.g., FGF2, IL-4), and late-phase (e.g., IL-13, VEGF-A). In parallel, we identified non- SASP factors such as MPO and Cathepsin D, which have been directly linked to tau pathology[15, 54, 92-96] . Together, these findings suggest that neuronal injury alone can recapitulate aging-like secretory patterns and drive neurodegenerative changes, independent of glial involvement. These molecular profiles coincide temporally with the observed shift from early hyperexcitability to late-phase LTD, supporting existing links between aging, synaptic weakening, and tau aggregation [97-100]. The extracellular release of A β was decreased following injury, consistent with human CSF profiles post-TBI[101, 102] . This reduction may reflect enhanced intracellular accumulation of Aβ due to upregulated APP processing via injury-induced neuronal activity[103-106]. In contrast, extracellular tau (total and pT181) increased significantly, confirming previous TBI findings[9, 10], and implicating neuronal activity in tau release regulation [107, 108] . The increase in extracellular Tau along with simultaneous decline in intracellular total Tau and rise in intracellular pT181 suggests accumulation of hyperphosphorylated tau species that potentially contribute to the formation of tau aggregates and neurofibrillary tangles. Importantly, we observed significant upregulation of calpain-1, cathepsin D, and active caspase-3,enzymes known to cleave tau at specific sites and generate neurotoxic fragments [109-111]. Tau truncation products such as Tau151-391 have been shown to exhibit high aggregation propensity and pathological phosphorylation [112]. Calpain-1 and caspase-3 are primarily responsible for the formation of tau fragments [109, 113] . Calpain-1 produces a cleaved form of ~ 78 kDa [113] and activated forms at ~ 55 and 32 kDa [109, 114-116] by autolysis. In our study, we observed that these activated forms of calpain-1 were elevated post-injury followed by increased active caspase-3 early in injury. Indeed, studies have shown that calpain-1 knockout mice lack the ability to increase caspase-3 and its active form after injury (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 17 [35], suggesting that calpain-1 directly produces active caspase-3 in our model. Additionally, the total Tau antibody Tau46, detects AA404-441 on the C-terminal of Tau protein, which can be cleaved by the three proteinases. The significant decrease of intracellular Tau46+ staining post-injury is consistent with enhanced tau truncation. This study demonstrates that a 3D Neuron-on-Chip model using hPSC-derived excitatory hPFC neurons can recapitulate key features of neuronal injury over an extended post-injury timeline. We report acute excitotoxic responses, progressive functional connectivity loss, and the emergence of a neurodegenerative and SASP-like secretome, all in the absence of glial contribution. These findings highlight an intrinsic neuronal mechanism for tauopathy development following neuronal injury and offer a unique platform to dissect cell- autonomous injury responses relevant to TBI and Alzheimer’s disease progression. 4. Methods Device Fabrication: Negative device molds were 3D printed using Clear-V4 resin (Formlabs™, Cat# RS-F2-GPCL-04) on a Formlabs 3 printer. PDMS (Sylgard™ 184, Dow Corning) was mixed in a 10:1 base-to-curing agent ratio, and 7 mL of the mixture was poured into 60 mm Petri dishes containing seven molds that were held in place with mineral oil. Device molds and PDMS mixture were degassed under vacuum (30 min) and cured at 80°C for 3 hours to result in polymerized PDMS devices that were ~22 mm in height. Polymerized PDMS using the above ratio of base polymer and curing agent has been previously characterized to have an elastic modulus of ~1.3-3 MPa [117]. The devices were removed from 60 mm petri dish by a biopsy punch with a radius of 7 mm and then immobilized to 35 mm glass bottom dishes (Cellvis) using PDMS base mixture. The devices were returned to the oven to bake for another 3 hours at 80°C, then sterilized in 70% ethanol and air-dried before seeding cells. Cell Culture of hPSC-Derived Prefrontal Cortex Neurons: Human pluripotent stem cells (hPSCs) derived prefrontal cortex neurons (hPFCs) were generated following methods described previously [40]. hPSCs were dissociated into single cells and plated on Matrigel- coated plates in Essential 8 medium with 10 μ M ROCK inhibitor. From day 0–2, cultures were treated with Essential 6 medium containing 2 μ M XAV939, 100 nM LDN193189, and 10 μ M SB431542. XAV939 was withdrawn from day 2 onward. At day 6–8, cells were replated as high-density droplets on poly-ornithine/laminin/fibronectin-coated dishes and cultured in N2 medium with B27 (1:1000, no vitamin A), FGF8 (50 ng/mL), and SHH (25 ng/mL) for 4 days, until neuroepithelial rosettes were visible. Droplets were then passaged 1:2 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 18 with trypsin onto poly-ornithine/laminin/fibronectin coated plates and cultured in the same media. On day 16, cells were passaged using Accutase and plated at 1×10 /i4 cells/cm² in PFC medium (Neurobasal™, B27 1:50, N2 1:100, GlutaMAX 1:100). FGF8 (50 ng/mL) was supplemented during the first two weeks. Cells were passaged again on day 22, replated at 100,000 cells/cm2 and used for experiments on day 35. Hydrogel encapsulation of hPFC neurons: iPSC-derived hPFC neurons were suspended in GeltrexTM at a concentration of 20,000 cells/ μ L. 10 μ L of mixture was quickly dispensed into each cell chamber (B1, B2) of the device. The device was then incubated at 37°C/5% CO2 for 30 minutes to facilitate polymerization before 200 μ L of PFC media was added into the device. Media change was performed daily thereafter. Weight-Drop Injury Induction: hPFC neurons in the device were allowed to mature for 2 weeks before being subjected to weight-drop injury. A weight drop apparatus was designed and built based on Feeney’s weight drop injury model [118, 119] . Both chambers within each Neuron-on-Chip device received a focal injury with a 6-gram weight dropped from a height of 15 cm (impact velocity ≈ 1.72 m s-1, acceleration ≈ 1 g, force ≈ 9 mJ, force/area on PDMS shell ≈ 11.4 mJ mm-2). Calcium imaging: Calcium imaging was performed immediately after injury (0.5h), 24 h and 72 h post-injury. Fluo-4 AM (Thermo Fisher) was dissolved in 20% Pluronic F-127 (in DMSO) and diluted (10 μ L/mL) in PFC media to reach a 10 μ M final concentration. A loading solution was made by adding 8 μ L of stock solution to 1 mL of PFC media. Media in the device was replaced with loading solution and the device was incubated for 60 minutes. The device was transferred back to PFC media for recording. Calcium recordings were acquired using Leica DM-IRBE inverted microscope with a FITC filter (Ex/Em: 488/525 mm). Images from 3 regions of interest (ROIs; B1, B2, and region between the two) were obtained separately for 5 minutes at 20 fps with 50 ms exposure. The device and cells were maintained under physiological conditions (37°C with 5% CO2 medical grade air mixture (Airgas, PA)) in a microscope stage top incubator (Tokai hit live cell culture stage adapter) for the entire recording session. Neuronal Activity and Network Analysis: Calcium traces were extracted using custom MATLAB® scripts. Deconvolution and spike estimation was performed using MATLAB script developed by Pnevmatikakis [19]. Waveform features (rise/fall time, pulse width, and amplitude) were calculated using methods published previously [20]. Network analysis was assessed via pair-wise phase-locking synchronization matrix using previously published methods [31]. Briefly, each neuron j has a discrete sequence of spike (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 19 times, /g1675/g3037 /g4666/g1666/g4667 , and a time-varying instantaneous phase was assigned to each neuron j within the nth inter-spike interval (1). /g5030 /g3037 /g4666 /g1675, /g1666 /g4667 /g34042 /g5054 /g1666/g33∑72 /g5054 /g1675/g33∑6/g1675 /g3037 /g4666/g1666/g4667 /g1675/g3037 /g4666 /g1666/g33∑71 /g4667 /g33∑6/g1675 /g3037 /g4666/g1666/g4667 /g46661/g4667 Pair-wise phase-locking synchronization matrix C of neurons with total number of M was generated by calculating circular variance of the phase difference between pair-wise neurons j and k (2). /g1645 /g3037/g3038 /g3404 /g3657 /g1657 /g3036/g4666/g3101 /g3285/g4666 /g3047 /g4667 /g2879/g3101 /g3285/g3286/g4666 /g3047 /g4667 /g4667 /g3657 /g46662/g4667 To obtain a normalized value of global synchronization index (GSI), which ranges from 0 (non-coordinated activity) to 1 (perfectly synchronized activity), and independent of the number of neurons, we followed previously published eigen-value based decomposition approaches [120, 121]. The eigenvalue decomposition of C can be given by (3) /g165∑/g1674 /g3036 /g3404/g501∑ /g3036 /g1674 /g3036 /g46663/g4667 Where /g501∑ /g3036 is the eigenvalue and /g501∑ /g2869 /g340∑/g501∑ /g2870 /g340∑/g1710/g340∑/g501∑ /g3014 , and /g1674 /g3036 is the eigenvector corresponding to /g501∑ /g3036 . Surrogate matrices of matrix C were generated using the amplitude-adjusted Fourier transform (AAFT) and repeated 10 times. The mean of eigen-value calculated across 10 surrogates is denoted as /g501∑ /g11∑1/g3036 . The normalized GSI can be computed by (4) /g1633/g1645/g1635 /g3404 /g4666 /g501∑ /g3014 /g33∑6/g501∑ /g11∑1/g3014 /g163∑/g33∑6/g501∑ /g11∑1/g3014 ,/g1661 /g1656 /g501∑ /g3014 /g3406/g501∑ /g11∑1/g3036 0, /g1667/g1675/g1660/g1657/g1670/g1675/g1661/g1671/g1657 /g1 /g46664/g4667 The equation to compute the number of clusters is expressed as (5) /g1640/g1673/g1665/g165∑/g1664/g1673/g1671/g1675/g1657/g1670 /g3404 /g3533 /g1671/g165∑/g1666/g4666 /g3038 /g501∑ /g3038 /g3406 /g3435 /g501∑ /g11∑1/g3038 /g33∑7/g1637/g3400/g1645 /g1630 /g3038 /g343∑ /g46665/g4667 Where sng is 1 if /g501∑ /g3038 /g3406/g4666 /g501∑ /g11∑1/g3036 /g33∑7/g1637/g3400/g1645 /g1630 /g3038 /g4667 is true, K is a constant (here K = 2, giving 95% confidence levels), /g501∑ /g11∑1/g3038 is the average and /g1645/g1630 /g3038 is the standard deviation of surrogate eigenvalues. Weighted Modularity (/g1643 /g3050 ) describes the degree to which the network may be subdivided into such clearly delineated and nonoverlapping groups, which was measured following previously published method [32, 33] (6). /g1643 /g3050 /g3404 1 /g1664 /g3050 /g3533/g4660 /g1675 /g3036/g3037 /g33∑6 /g1663 /g3036 /g3050 /g1663 /g3037 /g3050 /g1664 /g3050 /g4661/g5015 /g3040 /g3284,/g3040 /g3285 /g3036,/g3037/g3106/g3015 /g46666/g4667 Where i, j are the nodes in the network, and N is the set of all nodes in the network. /g1675 /g3036/g3037 is the weight of the link (i,j); Sum of all weights in the network, /g1664 /g3050 /g3404 ∑ /g1675 /g3036/g3037/g3036,/g3037/g3106/g3015 ; Weight degree of i, (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 20 /g1663 /g3036 /g3050 /g3404 ∑ /g1675 /g3036/g3037/g3037/g3106/g3015 ; /g1665 /g3036 is the module containing node i, which is determined by Louvain clustering, and /g5015 /g3040 /g3284,/g3040 /g3285/g34041 if /g1665 /g3036 /g3404/g1665 /g3037 , otherwise /g5015 /g3040 /g3284,/g3040 /g3285/g34040 . Weighted Global Efficiency ( /g1631 /g3050 ) is the average inverse shortest path length. It describes the integration of the network and how efficiently the network exchange information. /g1631 /g3050 was calculated following previously published method [33, 122] (7). /g1631 /g3050 /g3404 1 /g1666 /g3533 ∑ /g4666/g1656 /g3036/g3037 /g3050 /g4667 /g2879/g2869 /g3037/g3106/g3015,/g3037/g2999/g3036 /g1666/g33∑61/g3036/g3106/g3015 /g46667/g4667 Where i, j are the nodes in the network, and n is the number of nodes in the network; Shortest weighted path length between i and j, /g1656 /g3036/g3037 /g3050 /g3404 ∑ /g1656/g4666/g1675 /g3048/g3049 /g4667/g3028 /g3296/g3297/g3354/g3282/g3284/g1374/g3285 /g3298 , where f is a map from weight to length and /g165∑ /g3036/g1374/g3037 /g3050 is the shortest weighted path between i and j. Immunocytochemical analysis: The devices were fixed with 4% (w/v) paraformaldehyde and 0.4 M sucrose in PBS for 20 min at RT. Next, the devices were washed with PBST (0.5% (v/v) TritonX-100 in PBS) and permeabilized for 1 h with blocking solution (5% (w/v) BSA in PBST). Primary antibody incubation was performed overnight at 4°C in blocking solution. The devices were then washed with PBST twice and incubated with blocking solution for 1 h at RT. Secondary antibody incubations were performed for 3 h at RT in blocking solution, followed by 30 min incubation with Hoechst. Unbound antibodies and Hoechst were washed away by washing 2 times with PBS. 10X images were acquired on a Zeiss LSM 900 confocal microscope (Zeiss, Germany) and analyzed using custom MATLAB scripts. Primary antibodies include: pS396 Tau (Abcam #ab109390), pT231 Tau (Abcam #ab151559), Tau 46 (Cell Signaling #4019S), AT8 (ThermoFisher #MN1020), Beta III-Tubulin (EMD Millipore #ab9354), Synaptophysin-1 (Abcam #ab14692). Secondary antibodies include: 488 Goat anti-Chicken IgY (ThermoFisher #A11039), 594 Goat anti-Rabbit IgG(H+L) (ThermoFisher #A21428), 647 Goat anti-Mouse IgG1 (ThermoFisher #20240), 647 Goat anti- Rabbit IgG(H+L) (ThermoFisher #21244). Western Blotting: Cells were extracted from hydrogels, washed once with ice-cold DPBS without Ca/Mg, and then lysed in RIPA buffer (ThermoFisher, Waltham) supplemented with 1X Halt Protease & Phosphatase Inhibitor cocktail (Thermo Scientific, Waltham), on ice for 30 min, with intermittent vortexing. Cell lysates were spun down at 14,000 RPM for 20 minutes, at 4 /i4 C. Supernatants were collected and protein amounts were estimated using Pierce BCA Protein Assay kit (ThermoFisher, Waltham). Sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) was run to resolve proteins (10 μ g per lane) and transferred onto nitrocellulose membranes. The membranes were blocked for 1 h using (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 21 5% non-fat dry milk in TBST and incubated with primary antibodies – Calpain-1 (#MA3-940, Invitrogen), Caspase-3 (#9662, Cell Signaling), pT181 (#MN1050, Thermo), Tau46 (#4019S, CST), pS396 (#ab109390, Abcam), AT8 (#MN1020, Thermo), pT231 (#ab151559, Abcam) and β -actin (#4970S, CST) in 5% BSA in TBST, at 4°C overnight. This was followed by secondary antibody incubation for 1 h at room temperature. Immunoblots were developed using Clarity Western ECL substrate (170-5060, Bio-Rad) and digital images were acquired using the ChemiDoc MP imaging system (Bio-Rad). Quantification of protein bands was performed using ImageJ software (NIH) and band intensities were normalized to respective β - actin and Tau46 loading controls. Secretome Analysis: Conditioned media was collected before each calcium imaging session at 24 h, 72 h, 5d and 8d time points. Pooled replicates were analyzed using 48-plex, human supplemental biomarker 10-plex, and human amyloid beta and tau 4-plex discovery assays (Eve Technologies). Low-abundance proteins were excluded. Proteins with log /i4 fold change >0.3 (injury vs. control) at any time point were retained, identifying 24 cytokines and 4 tau- related markers. These log2 fold change values of each marker at each time point were subjected to hierarchical clustering using the Ward algorithm [123] for their release profile. Next, a cross-correlation matrix of absolute fluorescence values from the 28 markers was used to compute an adjacency matrix after removing non-significant correlations (p-value < 0.5). Functional clusters were visualized using a spectral clustering algorithm, and the functional association of the markers was plotted using a force directed graph. To correlate secretome with neuronal function, sparse canonical correlation analysis (sCCA) was applied to averaged calcium imaging metrics from corresponding biological replicates.

This work was supported by National Institute of Health (NIH)- National Institute of Neurological Disorders and Stroke (R01NS099596, and R21NS130468), and Alliance for Regenerative Rehabilitation Research and Training (AR3T) technology development awards to Lohitash Karumbaiah. This research was supported in part by NIH R01ES033892 and the Dianne Isakson Distinguished Professorship (JRR). Data Availability Statement Computer-aided design (CAD) files for the Neuron-on-Chip device, data, and code for data analysis will be made available by the lead contact upon request. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 22 Author contributions R. T., C-F L., and L. K. designed the study. R. T., C. K., and L. K. wrote the manuscript. R. T. and H. W. performed neuronal differentiation and cell culture. R. T., C-F L., and A. C. performed device design, manufacturing, maintenance, injury induction, calcium imaging, and immunofluorescence staining experiments. M. S., N. G., and C. K. performed western blotting. R. T. and A. M performed data analysis. I. M-W., N. Z, J. R, C-F L, and L.K. assisted with overall experimental design and manuscript revisions. Declaration of interests The authors declare no conflicts of interest. Received: ((will be filled in by the editorial staff)) Revised: ((will be filled in by the editorial staff)) Published online: ((will be filled in by the editorial staff))

1. Hayes, J. P., Bigler, E. D., and Verfaellie, M., "Traumatic Brain Injury as a Disorder of Brain Connectivity". J Int Neuropsychol Soc (2016). 22(2): p. 120-37. 2. Dennis, E. L., Vervoordt, S., Adamson, M. M., et al., "Accelerated Aging after Traumatic Brain Injury: An ENIGMA Multi-Cohort Mega-Analysis". Ann Neurol (2024). 96(2): p. 365-377. 3. Amgalan, A., Maher, A. S., Ghosh, S., et al., "Brain age estimation reveals older adults’ accelerated senescence after traumatic brain injury". GeroScience (2022). 44(5): p. 2509-2525. 4. Cole, J. H., Leech, R., Sharp, D. J., and Initiative, f. t. A. s. D. N., "Prediction of brain age suggests accelerated atrophy after traumatic brain injury". Annals of Neurology (2015). 77(4): p. 571-581. 5. Assiociation, A. s., Traumatic Brain Injury (TBI). 6. Mudher, A., Colin, M., Dujardin, S., et al., "What is the evidence that tau pathology spreads through prion-like propagation?". Acta Neuropathologica Communications (2017). 5(1): p. 99. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 23 7. Edwards, G., 3rd, Zhao, J., Dash, P. K., Soto, C., and Moreno-Gonzalez, I., "Traumatic Brain Injury Induces Tau Aggregation and Spreading". J Neurotrauma (2020). 37(1): p. 80-92. 8. Saha, P. and Sen, N., "Tauopathy: A common mechanism for neurodegeneration and brain aging". Mech Ageing Dev (2019). 178: p. 72-79. 9. Johnson, V. E., Stewart, W., and Smith, D. H., "Widespread τ and amyloid-β pathology many years after a single traumatic brain injury in humans". Brain Pathol (2012). 22(2): p. 142-9. 10. Edwards, G., Zhao, J., Dash, P. K., Soto, C., and Moreno-Gonzalez, I., "Traumatic Brain Injury Induces Tau Aggregation and Spreading". Journal of Neurotrauma (2020). 37(1): p. 80-92. 11. Ost, M., Nylen, K., Csajbok, L., et al., "Initial CSF total tau correlates with 1-year outcome in patients with traumatic brain injury". Neurology (2006). 67(9): p. 1600- 1604. 12. Magnoni, S., Esparza, T. J., Conte, V., et al., "Tau elevations in the brain extracellular space correlate with reduced amyloid-β levels and predict adverse clinical outcomes after severe traumatic brain injury". Brain (2012). 135(4): p. 1268-1280. 13. Shahim, P., Tegner, Y., Gustafsson, B., et al., "Neurochemical aftermath of repetitive mild traumatic brain injury". JAMA neurology (2016). 73(11): p. 1308-1315. 14. Collins-Praino, L. E. and Corrigan, F., "Does neuroinflammation drive the relationship between tau hyperphosphorylation and dementia development following traumatic brain injury?". Brain, Behavior, and Immunity (2017). 60: p. 369-382. 15. Zhou, F., Sun, Y., Xie, X., and Zhao, Y., "Blood and CSF chemokines in Alzheimer’s disease and mild cognitive impairment: a systematic review and meta-analysis". Alzheimer's Research & Therapy (2023). 15(1): p. 107. 16. Finley, M., Fairman, D., Liu, D., et al., "Functional validation of adult hippocampal organotypic cultures as an in vitro model of brain injury". Brain Research (2004). 1001(1): p. 125-132. 17. Chen, W., Sheng, J., Guo, J., et al., "Cytokine cascades induced by mechanical trauma injury alter voltage-gated sodium channel activity in intact cortical neurons". Journal of Neuroinflammation (2017). 14(1): p. 73. 18. Robbins, M., Christensen, C. N., Kaminski, C. F., and Zlatic, M., "Calcium imaging analysis - how far have we come?". F1000Res (2021). 10: p. 258. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 24 19. Pnevmatikakis, E., Merel, J., Pakman, A., and Paninski, L., "Bayesian spike inference from calcium imaging data". Conference Record - Asilomar Conference on Signals, Systems and Computers (2013). 20. Clarkson, B. D. S., Kahoud, R. J., McCarthy, C. B., and Howe, C. L., "Inflammatory cytokine-induced changes in neural network activity measured by waveform analysis of high-content calcium imaging in murine cortical neurons". Scientific Reports (2017). 7(1): p. 9037. 21. Sporns, O., "Graph theory methods: applications in brain networks". Dialogues Clin Neurosci (2018). 20(2): p. 111-121. 22. Luo, C.-L., Chen, X.-P., Li, L.-L., et al., "Poloxamer 188 Attenuates in vitro Traumatic Brain Injury-Induced Mitochondrial and Lysosomal Membrane Permeabilization Damage in Cultured Primary Neurons". Journal of Neurotrauma (2012). 30(7): p. 597-607. 23. Inyang, E., Abhyankar, V., Chen, B., and Cho, M., "Modulation of in vitro Brain Endothelium by Mechanical Trauma: Structural and Functional Restoration by Poloxamer 188". Scientific Reports (2020). 10(1): p. 3054. 24. Bao, H., Yang, X., Zhuang, Y., et al., "The effects of poloxamer 188 on the autophagy induced by traumatic brain injury". Neuroscience Letters (2016). 634: p. 7-12. 25. López-García, I., Ger ő , D., Szczesny, B., et al., "Development of a stretch-induced neurotrauma model for medium-throughput screening in vitro: identification of rifampicin as a neuroprotectant". British Journal of Pharmacology (2018). 175(2): p. 284-300. 26. Wu, Y.-H., Rosset, S., Lee, T.-r., et al., "In Vitro Models of Traumatic Brain Injury: A Systematic Review". Journal of Neurotrauma (2021). 38(17): p. 2336-2372. 27. Hanna, M. E. and Pfister, B. J., "Advancements in in vitro models of traumatic brain injury". Current Opinion in Biomedical Engineering (2023). 25: p. 100430. 28. Berridge, M. J., "Calcium microdomains: Organization and function". Cell Calcium (2006). 40(5): p. 405-412. 29. Zhu, X., Beal, M. F., Wang, X., et al., "Neuronal Calcium Signaling, Mitochondrial Dysfunction, and Alzheimer's Disease". Journal of Alzheimer’s Disease (2010). 20(s2): p. S487-S498. 30. Schwarz, A. J., Gozzi, A., and Bifone, A., "Community structure and modularity in networks of correlated brain activity". Magnetic Resonance Imaging (2008). 26(7): p. 914-920. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 25 31. Patel, T., Ventre, S., and Meaney, D., "Dynamic Changes in Neural Circuit Topology Following Mild Mechanical Injury In Vitro". Annals of biomedical engineering (2012). 40: p. 23-36. 32. Newman, M. E. J., "Analysis of weighted networks". Physical Review E (2004). 70(5): p. 056131. 33. Rubinov, M. and Sporns, O., "Complex network measures of brain connectivity: Uses and interpretations". NeuroImage (2010). 52(3): p. 1059-1069. 34. Vosler, P. S., Brennan, C. S., and Chen, J., "Calpain-mediated signaling mechanisms in neuronal injury and neurodegeneration". Mol Neurobiol (2008). 38(1): p. 78-100. 35. Yamada, K. H., Kozlowski, D. A., Seidl, S. E., et al., "Targeted gene inactivation of calpain-1 suppresses cortical degeneration due to traumatic brain injury and neuronal apoptosis induced by oxidative stress". J Biol Chem (2012). 287(16): p. 13182-93. 36. Xue, S., Rogers, L. R. K., Zheng, M., et al., "Applying differential network analysis to longitudinal gene expression in response to perturbations". Frontiers in Genetics (2022). 13. 37. González, I., Déjean, S., Martin, P. G. P., and Baccini, A., "CCA: An R Package to Extend Canonical Correlation Analysis". Journal of Statistical Software (2008). 23(12): p. 1 - 14. 38. Carron, S. F., Alwis, D. S., and Rajan, R., "Traumatic Brain Injury and Neuronal Functionality Changes in Sensory Cortex". Frontiers in Systems Neuroscience (2016). 10. 39. Carron, S. F., Alwis, D. S., and Rajan, R., "Traumatic Brain Injury and Neuronal Functionality Changes in Sensory Cortex". Front Syst Neurosci (2016). 10: p. 47. 40. Cederquist, G. Y., Tchieu, J., Callahan, S. J., et al., "A Multiplex Human Pluripotent Stem Cell Platform Defines Molecular and Functional Subclasses of Autism-Related Genes". Cell Stem Cell (2020). 27(1): p. 35-49.e6. 41. Valadez-Barba, V., Cota-Coronado, A., Hernández-Pérez, O. R., et al., "iPSC for modeling neurodegenerative disorders". Regen Ther (2020). 15: p. 332-339. 42. Zhang, X., Hu, D., Shang, Y., and Qi, X., "Using induced pluripotent stem cell neuronal models to study neurodegenerative diseases". Biochimica et Biophysica Acta (BBA) - Molecular Basis of Disease (2020). 1866(4): p. 165431. 43. Penney, J., Ralvenius, W. T., and Tsai, L.-H., " Modeling Alzheimer’s disease with iPSC-derived brain cells". Molecular Psychiatry (2020). 25(1): p. 148-167. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 26 44. Slemmer, J. E., Matser, E. J. T., De Zeeuw, C. I., and Weber, J. T., "Repeated mild injury causes cumulative damage to hippocampal cells". Brain (2002). 125(12): p. 2699-2709. 45. Ravin, R., Morgan, N. Y., Blank, P. S., et al., "Response to Blast-like Shear Stresses Associated with Mild Blast-Induced Brain Injury". Biophysical Journal (2019). 117(7): p. 1167-1178. 46. Wolf, J. A., Stys, P. K., Lusardi, T., Meaney, D., and Smith, D. H., "Traumatic Axonal Injury Induces Calcium Influx Modulated by Tetrodotoxin-Sensitive Sodium Channels". The Journal of Neuroscience (2001). 21(6): p. 1923-1930. 47. Brittain, M. K., Brustovetsky, T., Sheets, P. L., et al., "Delayed calcium dysregulation in neurons requires both the NMDA receptor and the reverse Na+/Ca2+ exchanger". Neurobiology of Disease (2012). 46(1): p. 109-117. 48. Baracaldo-Santamaría, D., Ariza-Salamanca, D. F., Corrales-Hernández, M. G., et al., "Revisiting Excitotoxicity in Traumatic Brain Injury: From Bench to Bedside". Pharmaceutics (2022). 14(1). 49. Li, H., Thompson, V. F., and Goll, D. E., "Effects of autolysis on properties of μ - and m-calpain". Biochimica et Biophysica Acta (BBA) - Molecular Cell Research (2004). 1691(2): p. 91-103. 50. von Reyn, C. R., Mott, R. E., Siman, R., Smith, D. H., and Meaney, D. F., "Mechanisms of calpain mediated proteolysis of voltage gated sodium channel α - subunits following in vitro dynamic stretch injury". J Neurochem (2012). 121(5): p. 793-805. 51. Brocard, C., Plantier, V., Boulenguez, P., et al., "Cleavage of Na(+) channels by calpain increases persistent Na(+) current and promotes spasticity after spinal cord injury". Nat Med (2016). 22(4): p. 404-11. 52. Evans, R. C. and Blackwell, K. T., "Calcium: amplitude, duration, or location?". Biol Bull (2015). 228(1): p. 75-83. 53. Inglebert, Y. and Debanne, D., "Calcium and Spike Timing-Dependent Plasticity". Frontiers in Cellular Neuroscience (2021). 15. 54. Green, P. S., Mendez, A. J., Jacob, J. S., et al., "Neuronal expression of myeloperoxidase is increased in Alzheimer's disease". Journal of Neurochemistry (2004). 90(3): p. 724-733. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 27 55. Masliah, E., Mallory, M., Alford, M., Deteresa, R., and Saitoh, T., "PDGF is associated with neuronal and glial alterations of Alzheimer's disease". Neurobiology of Aging (1995). 16(4): p. 549-556. 56. Watson, A. E. S., Goodkey, K., Footz, T., and Voronova, A., "Regulation of CNS precursor function by neuronal chemokines". Neuroscience Letters (2020). 715: p. 134533. 57. Li, S., olde Heuvel, F., Rehman, R., et al., "Interleukin-13 and its receptor are synaptic proteins involved in plasticity and neuroprotection". Nature Communications (2023). 14(1): p. 200. 58. Huang, Y., Smith, D. E., Ibáñez-Sandoval, O., Sims, J. E., and Friedman, W. J., "Neuron-specific effects of interleukin-1β are mediated by a novel isoform of the IL-1 receptor accessory protein". J Neurosci (2011). 31(49): p. 18048-59. 59. Zhao, X., Wang, H., Sun, G., et al., "Neuronal Interleukin-4 as a Modulator of Microglial Pathways and Ischemic Brain Damage". The Journal of Neuroscience (2015). 35(32): p. 11281-11291. 60. Ouyang, X., Wani, W. Y., Benavides, G. A., et al., "Cathepsin D overexpression in the nervous system rescues lethality and Aβ 42 accumulation of cathepsin D systemic knockout in vivo". Acta Pharmaceutica Sinica B (2023). 13(10): p. 4172-4184. 61. Rosenstein, J. M., Krum, J. M., and Ruhrberg, C., "VEGF in the nervous system". Organogenesis (2010). 6(2): p. 107-14. 62. Noda, M., Takii, K., Parajuli, B., et al., "FGF-2 released from degenerating neurons exerts microglial-induced neuroprotection via FGFR3-ERK signaling pathway". Journal of Neuroinflammation (2014). 11(1): p. 76. 63. Takeuchi, A., Miyaishi, O., Kiuchi, K., and Isobe, K.-i., "Macrophage colony- stimulating factor is expressed in neuron and microglia after focal brain injury". Journal of Neuroscience Research (2001). 65(1): p. 38-44. 64. Vogel, D. Y., Kooij, G., Heijnen, P. D., et al., "GM-CSF promotes migration of human monocytes across the blood brain barrier". Eur J Immunol (2015). 45(6): p. 1808-19. 65. Herrmann, K. A. and Broihier, H. T., "What neurons tell themselves: autocrine signals play essential roles in neuronal development and function". Curr Opin Neurobiol (2018). 51: p. 70-79. 66. Yang, J., Ding, H., Shuai, B., Zhang, Y., and Zhang, Y., "Mechanism and effects of STING–IFN-I pathway on nociception: A narrative review". Frontiers in Molecular Neuroscience (2023). 15. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 28 67. Fryer, A. D., Stein, L. H., Nie, Z., et al., "Neuronal eotaxin and the effects of CCR3 antagonist on airway hyperreactivity and M2 receptor dysfunction". J Clin Invest (2006). 116(1): p. 228-36. 68. Michael, B. D., Bricio-Moreno, L., Sorensen, E. W., et al., "Astrocyte- and Neuron- Derived CXCL1 Drives Neutrophil Transmigration and Blood-Brain Barrier Permeability in Viral Encephalitis". Cell Rep (2020). 32(11): p. 108150. 69. Yamamoto, K., Takeshita, K., Kojima, T., Takamatsu, J., and Saito, H., "Aging and plasminogen activator inhibitor-1 (PAI-1) regulation: implication in the pathogenesis of thrombotic disorders in the elderly". Cardiovascular Research (2005). 66(2): p. 276-285. 70. Westphal, D., Sytnyk, V., Schachner, M., and Leshchyns'ka, I., "Clustering of the neural cell adhesion molecule (NCAM) at the neuronal cell surface induces caspase- 8- and -3-dependent changes of the spectrin meshwork required for NCAM-mediated neurite outgrowth". J Biol Chem (2010). 285(53): p. 42046-57. 71. Robinson, K. F., Narasipura, S. D., Wallace, J., Ritz, E. M., and Al-Harthi, L., "Negative regulation of IL-8 in human astrocytes depends on β -catenin while positive regulation is mediated by TCFs/LEF/ATF2 interaction". Cytokine (2020). 136: p. 155252. 72. Sil, S., Periyasamy, P., Thangaraj, A., Chivero, E. T., and Buch, S., "PDGF/PDGFR axis in the neural systems". Mol Aspects Med (2018). 62: p. 63-74. 73. Del Rey, A., Verdenhalven, M., Lörwald, A., et al., "Brain-borne IL-1 adjusts glucoregulation and provides fuel support to astrocytes and neurons in an autocrine/paracrine manner". Molecular psychiatry (2016). 21(9): p. 1309-1320. 74. Ogunshola, O. O., Antic, A., Donoghue, M. J., et al., "Paracrine and Autocrine Functions of Neuronal Vascular Endothelial Growth Factor (VEGF) in the Central Nervous System*". Journal of Biological Chemistry (2002). 277(13): p. 11410-11415. 75. Zamburlin, P., Ruffinatti, F. A., Gilardino, A., et al., "Calcium signals and FGF-2 induced neurite growth in cultured parasympathetic neurons: spatial localization and mechanisms of activation". Pflügers Archiv - European Journal of Physiology (2013). 465(9): p. 1355-1370. 76. Runge, E. M., Setter, D. O., Iyer, A. K., et al., "Cellular Sources and Neuroprotective Roles of Interleukin-10 in the Facial Motor Nucleus after Axotomy". Cells (2022). 11(19): p. 3167. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 29 77. Schäbitz, W.-R., Krüger, C., Pitzer, C., et al., "A Neuroprotective Function for the Hematopoietic Protein Granulocyte-Macrophage Colony Stimulating Factor (GM- CSF)". Journal of Cerebral Blood Flow & Metabolism (2008). 28(1): p. 29-43. 78. Petrisko, T. J., Bloemer, J., Pinky, P. D., et al., "Neuronal CXCL10/CXCR3 Axis Mediates the Induction of Cerebral Hyperexcitability by Peripheral Viral Challenge". Front Neurosci (2020). 14: p. 220. 79. Gustafsson, D., Klang, A., Thams, S., and Rostami, E., "The Role of BDNF in Experimental and Clinical Traumatic Brain Injury". Int J Mol Sci (2021). 22(7). 80. Hayakawa, M., Tsuchida, T., Honma, Y., et al., "Fibrinolytic system activation immediately following trauma was quickly and intensely suppressed in a rat model of severe blunt trauma". Sci Rep (2021). 11(1): p. 20283. 81. Helmy, A., Carpenter, K. L., Menon, D. K., Pickard, J. D., and Hutchinson, P. J., "The Cytokine Response to Human Traumatic Brain Injury: Temporal Profiles and Evidence for Cerebral Parenchymal Production". Journal of Cerebral Blood Flow & Metabolism (2011). 31(2): p. 658-670. 82. Di Battista, A. P., Rhind, S. G., Hutchison, M. G., et al., "Inflammatory cytokine and chemokine profiles are associated with patient outcome and the hyperadrenergic state following acute brain injury". Journal of Neuroinflammation (2016). 13(1): p. 40. 83. Tweedie, D., Karnati, H. K., Mullins, R., et al., "Time-dependent cytokine and chemokine changes in mouse cerebral cortex following a mild traumatic brain injury". eLife (2020). 9: p. e55827. 84. Lee, P. R., Cohen, J. E., Iacobas, D. A., Iacobas, S., and Fields, R. D., "Gene networks activated by specific patterns of action potentials in dorsal root ganglia neurons". Scientific Reports (2017). 7(1): p. 43765. 85. Kiss, J. Z., Wang, C., Olive, S., et al., "Activity-dependent mobilization of the adhesion molecule polysialic NCAM to the cell surface of neurons and endocrine cells". Embo j (1994). 13(22): p. 5284-92. 86. Liu, T., Zhang, L., Joo, D., and Sun, S.-C., "NF- κ B signaling in inflammation". Signal Transduction and Targeted Therapy (2017). 2(1): p. 17023. 87. Ohtani, N., "The roles and mechanisms of senescence-associated secretory phenotype (SASP): can it be controlled by senolysis?". Inflammation and Regeneration (2022). 42(1): p. 11. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 30 88. Coppé, J. P., Desprez, P. Y., Krtolica, A., and Campisi, J., "The senescence-associated secretory phenotype: the dark side of tumor suppression". Annu Rev Pathol (2010). 5: p. 99-118. 89. Li, X., Chentao, L., Zhang, W., et al., "Inflammation and aging: signaling pathways and intervention therapies". Signal Transduction and Targeted Therapy (2023). 8. 90. Li, R., Teng, Y., Guo, Y., et al., "Aging-related decrease of histone methyltransferase SUV39H1 in adipose-derived stem cells enhanced SASP". Mechanisms of Ageing and Development (2023). 215: p. 111868. 91. Demaria, M., Ohtani, N., Youssef, S. A., et al., "An essential role for senescent cells in optimal wound healing through secretion of PDGF-AA". Dev Cell (2014). 31(6): p. 722-33. 92. Gellhaar, S., Sunnemark, D., Eriksson, H., Olson, L., and Galter, D., "Myeloperoxidase-immunoreactive cells are significantly increased in brain areas affected by neurodegeneration in Parkinson's and Alzheimer's disease". Cell Tissue Res (2017). 369(3): p. 445-454. 93. Ginsberg, S. D., Hemby, S. E., Lee, V. M. Y., Eberwine, J. H., and Trojanowski, J. Q., "Expression profile of transcripts in Alzheimer's disease tangle‐ bearing CA1 neurons". Annals of neurology (2000). 48(1): p. 77-87. 94. Fernández-Montoya, J. and Pérez, M., "Cathepsin D in a murine model of frontotemporal dementia with Parkinsonism-linked to chromosome 17". Journal of Alzheimer’s Disease (2015). 45(1): p. 1-14. 95. Khurana, V., Elson-Schwab, I., Fulga, T. A., et al., "Lysosomal dysfunction promotes cleavage and neurotoxicity of tau in vivo". PLoS genetics (2010). 6(7): p. e1001026. 96. Vidoni, C., Follo, C., Savino, M., Melone, M. A. B., and Isidoro, C., "The Role of Cathepsin D in the Pathogenesis of Human Neurodegenerative Disorders". Medicinal Research Reviews (2016). 36(5): p. 845-870. 97. Wang, Z., Kennedy, B. K., Wong, L.-W., and Sajikumar, S., "Aging inverts the effects of p75-modulated mTOR manipulation on hippocampal neuron synaptic plasticity in male mice". The FASEB Journal (2023). 37(8): p. e23067. 98. Shetty, M. S., Sharma, M., and Sajikumar, S., "Chelation of hippocampal zinc enhances long ‐ term potentiation and synaptic tagging/capture in CA 1 pyramidal neurons of aged rats: implications to aging and memory". Aging Cell (2017). 16(1): p. 136-148. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 31 99. Norris, C. M., Korol, D. L., and Foster, T. C., "Increased susceptibility to induction of long-term depression and long-term potentiation reversal during aging". Journal of Neuroscience (1996). 16(17): p. 5382-5392. 100. Marttinen, M., Kurkinen, K. M. A., Soininen, H., Haapasalo, A., and Hiltunen, M., "Synaptic dysfunction and septin protein family members in neurodegenerative diseases". Molecular Neurodegeneration (2015). 10(1): p. 16. 101. Kay, A. D., Petzold, A., Kerr, M., et al., "Alterations in Cerebrospinal Fluid Apolipoprotein E and Amyloid β -Protein after Traumatic Brain Injury". Journal of Neurotrauma (2003). 20(10): p. 943-952. 102. Franz, G., Beer, R., Kampfl, A., et al., "Amyloid beta 1-42 and tau in cerebrospinal fluid after severe traumatic brain injury". Neurology (2003). 60(9): p. 1457-61. 103. Bero, A. W., Yan, P., Roh, J. H., et al., "Neuronal activity regulates the regional vulnerability to amyloid-β deposition". Nature Neuroscience (2011). 14(6): p. 750-756. 104. Hefter, D., Ludewig, S., Draguhn, A., and Korte, M., "Amyloid, APP, and Electrical Activity of the Brain". Neuroscientist (2020). 26(3): p. 231-251. 105. Itoh, T., Takao, S., Shozo, N., et al., "Expression of amyloid precursor protein after rat traumatic brain injury". Neurological Research (2009). 31(1): p. 103-109. 106. Chen, X. H., Johnson, V. E., Uryu, K., Trojanowski, J. Q., and Smith, D. H., "A lack of amyloid beta plaques despite persistent accumulation of amyloid beta in axons of long-term survivors of traumatic brain injury". Brain Pathol (2009). 19(2): p. 214-23. 107. Pooler, A. M., Phillips, E. C., Lau, D. H. W., Noble, W., and Hanger, D. P., "Physiological release of endogenous tau is stimulated by neuronal activity". EMBO reports (2013). 14(4): p. 389-394-394. 108. Berlind, J. E., Lai, J. D., Lie, C., et al., "KCTD20 suppression mitigates excitotoxicity in tauopathy patient organoids". Neuron (2025). 109. Liu, M. C., Kobeissy, F., Zheng, W., et al., "Dual vulnerability of tau to calpains and caspase-3 proteolysis under neurotoxic and neurodegenerative conditions". ASN Neuro (2011). 3(1): p. e00051. 110. Xia, Y., Lloyd, Grace M., and Giasson, Benoit I., "Targeted proteolytic products of τ and α -synuclein in neurodegeneration". Essays in Biochemistry (2021). 65(7): p. 905- 912. 111. Chu, D., Yang, X., Wang, J., et al., "Tau truncation in the pathogenesis of Alzheimer’s disease: a narrative review". Neural Regeneration Research (2024). 19(6). (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 32 112. Gu, J., Xu, W., Jin, N., et al., "Truncation of Tau selectively facilitates its pathological activities". Journal of Biological Chemistry (2020). 295(40): p. 13812-13828. 113. Chen, H.-H., Liu, P., Auger, P., et al., "Calpain-mediated tau fragmentation is altered in Alzheimer’s disease progression". Scientific Reports (2018). 8(1): p. 16725. 114. Hosfield, C. M., Elce, J. S., Davies, P. L., and Jia, Z., "Crystal structure of calpain reveals the structural basis for Ca2+‐dependent protease activity and a novel mode of enzyme activation". The EMBO Journal (1999). 18(24): p. 6880-6889. 115. Kitagaki, H., Tomioka, S., Yoshizawa, T., et al., "Autolysis of calpain large subunit inducing irreversible dissociation of stoichiometric heterodimer of calpain". Biosci Biotechnol Biochem (2000). 64(4): p. 689-95. 116. Elce, J. S., Hegadorn, C., and Arthur, J. S., "Autolysis, Ca2+ requirement, and heterodimer stability in m-calpain". J Biol Chem (1997). 272(17): p. 11268-75. 117. Johnston, I. D., McCluskey, D. K., Tan, C. K. L., and Tracey, M. C., "Mechanical characterization of bulk Sylgard 184 for microfluidics and microengineering". Journal of Micromechanics and Microengineering (2014). 24: p. 035017. 118. Feeney, D. M., Boyeson, M. G., Linn, R. T., Murray, H. M., and Dail, W. G., "Responses to cortical injury: I. Methodology and local effects of contusions in the rat". Brain Research (1981). 211(1): p. 67-77. 119. Ma, X., Aravind, A., Pfister, B. J., Chandra, N., and Haorah, J., "Animal Models of Traumatic Brain Injury and Assessment of Injury Severity". Molecular Neurobiology (2019). 56(8): p. 5332-5345. 120. Li, X., Cui, D., Jiruska, P., et al., "Synchronization measurement of multiple neuronal populations". J Neurophysiol (2007). 98(6): p. 3341-8. 121. Lee, S. H., Park, Y. M., Kim, D. W., and Im, C. H., "Global synchronization index as a biological correlate of cognitive decline in Alzheimer's disease". Neurosci Res (2010). 66(4): p. 333-9. 122. Latora, V. and Marchiori, M., "Efficient Behavior of Small-World Networks". Physical Review Letters (2001). 87(19): p. 198701. 123. Ward Jr, J. H., "Hierarchical Grouping to Optimize an Objective Function". Journal of the American Statistical Association (1963). 58(301): p. 236-244. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 33 Supporting Information A 3D Human Neuron-on-Chip Platform to Monitor Neuronal Injury Responses Ruiping Tang, Charles-Francois Latchoumane, Avi Chopra, Marzan Sarkar, Chunki Kim, Nathan Gonsalves, Hsueh-Fu Wu, Isha Mhatre-Winters, Aditya Mishra, Nadja Zeltner, Jason R. Richardson, Lohitash Karumbaiah* Figure S1. (hPFC cells demonstrate decreased proliferation and increased neuronal differentiation 9 days post device-seeding. (A) Edu+ staining of hPFC cells reveals significantly reduced cell proliferation 9 days post device seeding. Two-way ANOVA. Uncorrected Fisher’s LSD, ***P<0.001, ****P<0.0001. n = 5 images per sample, data represents 3 independent experiments. (B) Representative confocal image of immunohistochemically stained HuC/D (pan-neuronal marker), GFAP (astrocyte marker), and Olig2 (oligodendrocytes marker), scale – 200 µm. (C) Quantification of (B) demonstrating a major presence of HuC/D+ neurons in PFC culture 2 weeks post device seeding. n = 5 images per sample from 3 independent experiments.) Figure S2. (Weight-drop injury induces changes in cell body retraction and morbidity in response to increasing impact force. Plots demonstrating significant changes in cell body retraction (A), and cell death (B) with increasing impact force. Two-way ANOVA. Uncorrected Fisher’s LSD, ***P<0.001, ****P<0.0001. (C) Scatter dot plot demonstrating significantly enhanced cell death in the weight drop injured group, compared to uninjured (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 34 control. Two-way ANOVA, Ptime < 0.0001, Ptreatment < 0.0001, Pinteraction = 0.1541 with Fisher’s LSD, ***p<0.001, ****p<0.0001. n = 5 images per ROI, 2 ROIs per device, 3 devices per group from 3 independent experiments.) Figure S3. (Injury induces significant changes in neuronal activity at 0.5h, 24h, and 8d post injury. (A-D) tSNE plot of sample distribution and their 95% CI eclipses on neuronal functions described by 13 neuronal activity parameters reveals significant differences in neuronal activity patterns between injury and control groups at 0.5h, 24h and 8d post injury. PERMANOVA, P-Group = 0.001, iter = 1000. Bonferroni pairwise comparison, P value indicated in each plot. Transparent makers in each plot indicates samples from other time points.) (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 35 Figure S4. (Injury induces an acute increase in neuronal connectivity and disruption of network structure. (A, B) nD and PL indicate acute increase and subsequent loss of functional connections. (C, D) NoC (C, neuronal subcommunities) identified by an eigen-value-based algorithm, and Modularity (D, sub-community presence) demonstrate acute disruption of community structure in response to injury. (E) Representative heatmaps of pair-wise phase locking matrix from control and injured groups 0.5 h post injury indicate remodeling of functional network structure in response to the injury. Two-way ANOVA with Fisher’s LSD test multiple comparison, *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. n = 3-5 per time point from 4 independent experiments.) (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 36 Figure S5. (Western blot analysis of calpain-1 and caspase-3. (A, B) Western blotting results of calpain-1 levels 0.5 hr – 8 days post injury. (C, D) Western blotting results of caspase-3 levels 0.5 hr – 8 days post injury.) (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 37 Figure S6. (Loss of overall connectivity and long-range connections post injury. Histograms demonstrate the loss of neurons with high connectivity (A-E) and long-range connection (F-J) over time post-injury in the injured groups when compared to controls.) Figure S7. (Acute decrease in global efficiency and cluster coefficient post-injury. (A-B) Network analysis demonstrated an injury induced acute alteration of Global Efficiency (A), and Cluster Coefficient (B). Two-way ANOVA with Fisher’s LSD test multiple comparison, *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. n = 3-5 per time point across 4 independent experiments.) Figure S8. (Dysregulation of calcium trafficking 5-8 days post injury. (A-D) GSI (A), mRT (B), mFT (C), mBW (D) revealing a statistically insignificant but increasing trend in injured neurons 5-8 days post injury compared to control. Two-way ANOVA with Fisher’s LSD test multiple comparison, *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. n = 3-5 per time points across 4 independent experiments.) (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint 38 Figure S9. (Intracellular accumulation of pathologically relevant tau. (A) Western blotting

of pT231, pT396 tau and total tau levels. (B-C) Densitometric quantification of results from A. Two-way ANOVA with fisher’s LSD multiple comparison, *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. Data represents 3 independent experiments.) (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted August 9, 2025. ; https://doi.org/10.1101/2025.08.06.667201doi: bioRxiv preprint