Multi-physics 3D human brain models elucidate mechanical mechanisms in Alzheimer’s disease

Introduction: : Alzheimer’s disease (AD), the leading cause of dementia, involves cognitive decline, neuronal loss, amyloid β (Aβ) plaque accumulation, alongside dysfunction of the blood-brain barrier (BBB) and meningeal lymphatics (ML) [1-3]. Together, the BBB and ML form the brain’s entry-exit vascular system, playing critical roles in brain health (Fig. 1a). Disruption of this system exacerbates toxic protein buildup and accelerates AD progression [2-4]. However, the underlying mechanisms of how this system interacts and fails in AD remain unknown, partially due to limited access to human brain tissue and the poor clinical translation of animal models [1].
Here, we developed physiologically relevant 3D in-vitro human brain models that integrate perfusable BBB microvasculature, sprouting ML vessels, and AD-mimicking neurons within one microfluidic platform (Fig. 1c). Our models recapitulate key structural and functional features of the human brain, enabling quantitative assessment of BBB permeability, ML drainage, and Aβ clearance under AD conditions. Furthermore, we incorporated mechanobiological tools onto our models to explore therapeutic targets, including Piezo1, a mechanosensitive channel [5], and employed optical tweezers (OT) and force spectrum microscopy [6, 7] to assess cellular mechanical changes during AD progression. We observed distinct alterations in BBB permeability, ML morphology and drainage capacity, interstitial flows, and cellular mechanics in AD.
Our versatile multi-physics platform enables systematic investigation of neurovascular-lymphatic interactions and mechanical mechanisms in AD, offering a powerful tool for both fundamental research and therapeutic discovery. It also lays the groundwork for broader applications in modeling other brain-related disorders.

Materials and

Methods: : Microfluidic Platform: The device mold was designed using Autodesk Fusion 360, milled via a micro-CNC milling machine (Bantam Tools), and made from polydimethylsiloxane (PDMS) by established protocols [8]. The device includes one media channel, one BBB channel, one ML channel, one neuron channel, and four gel loading ports (Fig. 1b).
Cell Culture: Human astrocytes, pericytes (ScienCell), brain microvascular endothelial cells (Angio-Proteomie), and lymphatic cells were cultured in cell-specific media. Neurons were cultured and differentiated using established protocols [9]. BBB cells were then seeded in fibrin gels within the BBB channel to form perfusable microvasculature. Neurons were embedded in Matrigel within the neuron channel. The ML channel was filled with fibrin gel and a monolayer of lymphatic cells was deposited onto the side of the ML channel gel. Interstitial flow was induced via hydrostatic pressure gradients.
Microscopy and Function Quantification: Confocal images were acquired using Olympus FV1200 and Leica DMI8 and analyzed with ImageJ and MATLAB. BBB permeability was measured by tracking intravascular fluorescent tracer diffusion into matrix over time (Fig. 2a). ML drainage capacity was quantified by tracking fluorescent tracers moving through lymphatic sprouts over time (Fig. 2b).
Mechanics Measurement: Optical tweezers (OT) were used to measure intracellular mechanical property changes in AD by dragging endocytosed beads and analyzing force-displacement responses (Fig. 2e) [7]. Force spectrum microscopy was used to measure overall intracellular mechanical activities by tracking endocytosed beads and analyzing their mean squared displacement (MSD) and cytoplasmic spring constant from OT [6].

Results, Conclusions, and Discussions:: We have successfully developed a unified 3D in-vitro human brain model integrating perfusable BBB networks, sprouting ML vessels, and AD-mimicking neurons (Fig. 1c). Under controlled interstitial flow, BBB cells form functional microvasculature within five days, and ML sprouts extend over 500 μm toward the BBB. Aβ accumulation and plaque formation are successfully replicated in both BBB and ML regions.
Quantitative analysis revealed increased BBB permeability and reduced ML drainage capacity in the AD model. Aβ42 across the BBB shows a lower permeability than passive permeability measured using dextran of similar molecular weight, suggesting involvement of active transport mechanisms (Fig. 2c). ML growth, morphology, and drainage capacity are all impaired in the AD model with a significant reduction in tracer clearance rates.
Mechanical cues are essential for proper vascular formation. Aβ exposure disrupts ML sprouting and alters flow dynamics. Optical tweezers-based microrheology (Fig. 2e) showed an increase in astrocyte stiffness from 0.5 Pa to 1.0 Pa under Aβ treatment (Fig. 2f), indicating a mechanobiological response. Ongoing experiments aim to correlate these changes with Piezo1 channel activity (Fig. 2d) and further quantify 3D tissue mechanical alterations.
Our physiologically relevant platform enables simultaneous, quantitative evaluation of Aβ dynamics and cellular mechanical changes under AD conditions. It offers a powerful tool for dissecting the neurovascular-lymphatic interface and identifying mechanosensitive pathways involved in disease progression. Future work will focus on fully coupling BBB permeability and ML drainage to better characterize clearance deficits in AD and facilitate the development of targeted therapeutic strategies.

Acknowledgements and/or References (Optional):: We thank Georgios Pavlou and Shun Zhang for their assistance with the microfluidic platform design, Marie Floryan for training in microfluidic fabrication, and Lani Lee and Haiqian Yang for support with optical tweezers and force-spectrum setup and training. This work was supported by the Cure Alzheimer’s Fund and NIH R01 grant NS121078 (to R.D.K. and S.H.C.).
[1] 2024 Alzheimer’s Disease Facts and Figures, Alzheimer’s Association. [2] Shin, Y. et al., Advanced Science, 6(20): 1900962, 2019. [3] Louveau, A. et al. Nature 523, 337–341, 2015. [4] Da Mesquita, S. et al. Neuron 100, 375–388, 2018. [5] Choi, D. et al., Nat Neurosci 27, 913–926 2024. [6] Guo, M. et al. Cell 158 (4), 822-832, 2014. [7] Gupta, S.K. et al., Multi-scale Extracellular Matrix Mechanics and Mechanobiology 283-310, 2020. [8] Hajal, C. et al., Nature Protocols, 17(1): 95–128, 2022. [9] Kim, Y.H. et al., Nature Protocols, 10(7): 985–1006, 2015. [10] Sweeney, M.D. et al. Nature 560, 172–174, 2018.