USNCB neuromechanics: Integrating across spatial and temporal scales

Introduction

The prevalence of neurodegenerative disorders is rapidly increasing. While Alzheimer’s disease and dementia generally correlate with longer lifespans, neurodegenerative disorders like chronic traumatic encephalopathy often affect individuals at young age. Historically, the underlying disease mechanisms of these chronic disorders—the slowly changing biochemical composition during aging and the repeated, rapidly changing biomechanical environment during head impact—have been viewed as distinct events. Recent studies suggest that Alzheimer’s disease and chronic traumatic encephalopathy share common degenerative pathways on the molecular and cellular levels.

Methods

We propose a multimodal model of neurodegeneration that can predict the onset of neurodegeneration from integrating modes of biomarker changes throughout life and provide mechanistic links between normal aging, early aging, and multiple mild traumatic brain injuries.

Results

Figure 1 illustrates the multimodal model of neurodegeneration in response to normal aging, early aging, and multiple mild traumatic brain injuries throughout an individual’s life. In normal aging, the spectrum of symptoms can be detected around the age of 85. In response to early aging in Alzheimer’s disease, the timeline of neurodegeneration shifts to the left. In response to multiple mild traumatic brain injuries in chronic traumatic encephalopathy, the timeline of neurodegeneration shifts upward in response to multiple discrete injuries within the exposure window between ages 15 and 30. However, the collective response of these events remains well below the detection threshold. It is not until years after the injuries that the spectrum of symptoms can be detected, around the age of 40.

Discussion

Understanding the pathways of neurodegeneration will likely require a multifaceted approach in which neurobiologists, neuroradiologists, and neuroscientists join forces to gain a more holistic picture of neurodegenerative diseases. Our current understanding of the spatial and temporal spectrum of neurodegeneration converges towards a view in which traumatic encephalopathy covers a landscape of phenomena between traumatic brain injury and Alzheimer’s disease, and shares common early disease mechanisms. Mechanistic modeling has the potential to provide collective insight into diagnosis, prediction, and treatment of neurodegeneration.

Acknowledgements

This work was supported by the Stanford BioX IIP 2016 seed grant ‘Molecular Mechanisms of Chronic Traumatic Encephalopathy’ and by the NSF grant CMMI 1727268 ’Understanding Neurodegeneration across the Scales’.

Reference

1. van den Bedem H, Kuhl E. (2017). Current Opinions in Biomedical Engineering, 1, pp. 23-30.


Background

The mammalian meninges are composed of three thin and heterogeneous layers of highly organized tissue with a variety of functions, from regulating the development of brain cortex to isolating the central nervous system [1]. Despite their putative role as structural dampers helping to mechanically shield the brain, their mechanical properties are still poorly investigated [2]. Here, we characterized the local mechanical heterogeneity of rat leptomeninges at the microscale via atomic force microscopy (AFM) indentation experiments to understand how microstructural variations at the tissue level can differentially affect load propagation.

Methods 

Rat pia-arachnoid complex (PAC) were isolated from rat brains immediately after death. The thin layers of tissue were peeled-off under a dissecting microscope and left to float in PBS. These were then adhered to tissue-treated glass slides. AFM indentation curves were performed on different positions of leptomeninges immersed in PBS using tipless cantilevers with spherical beads (R=30 μm). A hyperelastic material model (Fung) was used to fit the curves [3].

Results

Indentation curves were acquired for increasing indenter speeds on different positions of the PAC; increasing hysteresis (indicating viscoelastic material behavior) was observed for non-vascularized portions of the tissue, whereas indentations on capillaries yielded overlapping force-displacement cycles (Fig.1a). The approach curves were analyzed via a Fung hyperelastic model and indicated a 20-fold increase in the initial Young’s modulus E₀ of vascularized tissue (8.74 ± 0.73 kPa) as compared with non-vascularized PAC (0.38 ± 0.10 kPa); even within the latter, local heterogeneity in collagen content caused varying stiffness values (Fig.1c).

Consequences of head injuries are not limited to the victim alone but have impact on the society as a whole through the large costs involved, not to mention the tragedies and the suffering. It should be noted that very little is understood about the true mechanisms associated with head injury, but many theories exist. Implementation or modification of security systems, such as a new car model, is now a long and complex process. In recent years, biomechanical simulation models of head and human body acquired an increasingly larger place in the design of safety systems. One of the advantages with the finite element (FE) method is the possibility to model the anatomy with great detail, thus it is possible to study the kinematics of the head as well as the stresses and strains in the Central Nervous System (CNS) tissues. This presentation primarily focuses on summarizing current efforts, and to outline future strategies in human head injury modeling. Multiple length scales are involved in the development of traumatic brain injury, where the global mechanics of the head level are responsible for local physiological impairment of brain cells. In a multi-scale approach, finite element models of the head and at the axonal level are coupled, where it is observed that the maximum axonal strains do not always correlate directly with the brain tissue strain levels. The results indicate that cellular level heterogeneities have an important influence on the axonal strain, leading to an orientation and location-dependent sensitivity of the tissue to mechanical loads.

Mild traumatic brain injury (concussion) is a nearly unique clinical disorder: the mechanical signature of each injury is different, and these injuries occur across a varied genetic and prior exposure background for each individual. Most mechanistic studies for concussion reduce this clinical complexity by focusing on one injury mechanism in a preclinical model developed for a limited part of the mechanical exposure spectrum. Alternatively, many biomechanical studies estimate the human threshold for concussion across a broad range of mechanical exposures. In this talk, we present an approach that connects these two domains.

After reviewing the mechanical tolerance of brain components derived from experimental and computational studies, we describe efforts to predict how mechanical forces affect the activity and remodeling of neural circuits. We review past work showing the effects of these forces at the cellular and synaptic scale, and present a method to assess the impact of concussion forces on neural circuits in vitro. We utilize micro-injury models of concussion to derive thresholds for the impairment and remodeling of circuits with different wiring, composition, and glial content. We connect these data to in silico models of neural clusters at a larger scale to study the impairment of brain networks at the mesoscale, where neural circuits form connections across different brain regions. Utilizing measurements from synthetic networks in vitro, we find critical connectivity parameters and connection architectures that will allow neural circuits to synchronize and respond to patterned stimulation.

Utilizing brain architecture data from human volunteers, we examine how selected ‘injury fingerprints’ from human neuropathology data can lead to dynamic changes in neural circuit architecture. With network descriptor variables that describe control points in learning, memory, and disease progression, we estimate the emergence of functional injury phenotypes in concussion. In select cases, we compare these injury phenotypes with data from human imaging studies of concussion patients, determining if these predicted injury phenotypes can emerge from network remodeling processes at the microscopic and mesoscopic scales.

Together, these data point towards an opportunity to link concussion injury mechanisms, both individually and in combination, with a range of injury outcomes in the human population. Rather than concentrating only on the threshold for human injury, these studies allow one to develop a more granular, mechanism-based approach for connecting the unique mechanical signature of each injury with a predictive functional injury phenotype and recovery profile. When combined with molecular-scale assessment of the brain recovery process, this work will point towards a more individualized approach for concussion prevention, treatment and recovery.

Acknowledgments: Work funded, in part, by the NIH and the Allen Foundation.

Introduction

Animal injury models are widely used as human surrogates for the study of the mechanisms and the biomechanical thresholds of traumatic brain injury (TBI). However, one important question exists for translating animal responses to humans: how does the impact conditions and resulting biological response scale across species of various brain sizes and shapes. Simplified scaling methods based on structural similitude (e.g., Holbourn, 1943) have guided the interpretation of animal data in the context of humans, but these methods are unvalidated and are criticized because they do not consider the differences in brain anatomy and shape of each species. Investigations using finite element (FE) models of brains from multiple species can address some of the limitations that exist with TBI scaling methods.

Methods

Previously developed human (Mao et al., 2013), baboon, rhesus monkey (Antona-Makoshi, 2016), and porcine FE brain models (Coats et al. 2012) were modified to harmonize various modeling methods and anatomical features such as axonal tract information and brain tissue constitutive models. Where possible, the new brain models were validated against existing brain deformation test data and demonstrated good biofidelity. A parametric study was performed for each brain model simulated under a range of rotational head kinematics consistent with injury model data presented in the literature. Tissue-level injury metrics were recorded for each simulation, and model responses were compared across species under the assumption that comparable metrics result in comparable clinical outcome. Traditional scaling methods were evaluated against the predicted FE model data output (Figure 1).

Results

Relationships between tissue-level metrics, including maximum principal strain (MPS), the cumulative strain damage measure (CSDM) and maximum axonal strain (MAS), and applied head kinematics were formulated in angular acceleration-angular velocity space in a manner similar to Gabler et al. (2017).  The metrics predicted by the FE models were different compared with those measured using traditional kinematic scaling methods. 

Discussion

Compared with traditional scaling methods, using FE models to study cross-species TBI mechanics enabled the consideration for the differences between species in brain physical morphology, anatomy, tissue properties. Further efforts will focus on investigating the correlation between strain responses and pathology results. Overall, this study provided an improved method for scaling animal experiments and tissue injury metrics to humans.

References

Antona-Makoshi, J. (2016). PhD Thesis, Chalmers.
Coats, B. et al. (2012).Int. J. Dev. Neurosci, 30(3),191. 
Gabler, L. F. et al. (2017). J Biomech Eng.
Holbourn, A. H. S. (1943). Lancet, 242(6267), 438-441.
Mao H, et al.( 2013). J Biomech Eng.135(11):111002.

Introduction

The finite element (FE) derived injury metrics are used increasingly as predictive tools in traumatic brain injury (TBI) field. However, these metrics have mainly focused on predicting absence or presence of TBI rather than estimating the location of injury. In this study, a novel method was developed to quantify the utility of FE tools and different FE-derived tissue injury metrics in estimating the location of axonal injury under rapid head rotation. 

Methods

Piglets (28 four-week-old) experiencing rapid non-impact head rotation (113-203 rad/s) about the axial (n=10) or sagittal (n=18) were sacrificed at 6-hours post-injury, brains were perfusion-fixed, sectioned in coronal slices, stained for beta-amyloid-precursor-protein, and axonal injury (AI) maps were generated. An idealized pig head FE model [1] was enhanced by adding lateral ventricles and embedding axonal fiber tractography obtained from diffusion tensor images into the model (Fig.1A). The FE model-predicted deformations were validated against in-situ hemisection experiments [2] (Fig.1B). The model was scaled to match each piglet’s brain mass, and measured rotational velocity time-histories were used as input loading conditions. Brain deformation parameters maximum principal strain (MPS), strain rate (SR), and strain times strain rate (SxSR) were examined as potential regional AI predictors. The histopathological reports of AI were mapped onto equivalent locations in the scaled FE models (Fig.1C). For each of the brain deformation parameter, the average value experienced by 10% of the brain (POP10) over all simulations was defined as the threshold. The minimum distance between locations of brain elements exceeded the parameter threshold in each simulation and the regions of AI was averaged for each animal, and then across animals. We propose that a smaller distance represents a better biomechanical parameter for AI. The effect of brain anisotropy modeling on the injury location predictive capability was also studied.


Results

The selected threshold values for MPS, SR, and SxSR were 0.43, 254.5 s^-1, 53.98 s^-1, respectively. Brain MPS and SR estimated AI location within 5.18±1.85mm and 5.82±0.79mm from actual AI region, and were statistically better AI predictors (one-way ANOVA, p-value<0.05) than SxSR (6.49±1.36mm).

Discussion

Axonal injury can be attributed to local tissue deformation, blood flow, bioenergetics, and signaling pathways. We find that values of brain MPS and SR within 5-6mm may be valuable for predicting regions of AI acutely after TBI. Further efforts will focus on understanding and optimizing FE-neuropathology co-localization and examining additional AI-related metrics.

Acknowledgements

Funding provided by Biocore Co., Ltd. and NIH-R01NS097549.

References

1. Sullivan, S., et al., (2015). Biomech Model Mechanobiol, 14(4) p877
2. Ibrahim, N.G., et al, (2010). J Biomech Eng, 132(4) p044501

Introduction
Mild traumatic brain injury is a health threat that is under-reported and poorly understood [1]. Skull acceleration measurements have been crucial in understanding brain tolerances to trauma. Efforts to measure skull accelerations have leveraged sensor devices mounted to headbands, ear plugs, helmets, bite bars, and mouth-guards. However, the accuracy of these devices has been debated [2-3]. Our objective is to compare the accuracy among 1. a six accelerometer helmet (Head Impact Telemetry system, HITs), 2. a three accelerometer, three gyroscope mouthguard with mandible isolation (Stanford MiG 2.0).

Methods
Laboratory testing was performed on a twin-wire drop test consisting of an aluminum drop carriage that sled along twin wires and hit an aluminum plate. A helmeted 50^th percentile dummy head (ADT) on a HIII neck was attached to the drop carriage. The ADT with unconstrained jaw was instrumented with a 6a3ω sensor package and impacted at facemask, forehead, side, rear high and jaw pad at 3.3, 4.5 and 6.1 m/s.

The device performance was evaluated for impact location, peak values and error over time, compared to gold standard laboratory instrumentation (6a3ω). The accuracy of impact direction (ID) was measured by the solid angle between the reference (  and the estimated impact versor (  ):

The error over time was the root mean square error normalized by the maximum of the golden standard ( ): 



Peak resultant linear acceleration, angular acceleration, and angular velocity were correlated to gold standard. The accuracy of the devices was measured in terms of .

Results
In terms of peak values accuracy, Stanford MiG 2.0 had the highest  ( ), followed by HITs ( ) (Fig 1). MiG 2.0 also showed low average NRMS error ( ) and impact direction error ( ).

Discussion
Stanford MiG 2.0 showed higher accuracy than today’s most commonly used sensing system HITs. Since the instrumented mouth-guard rigidly couples to the skull via the upper dentition, it has the advantage of being immune to any helmet/head slipping or separation experienced by HITs. Thanks to its shock absorbers, Stanford MiG 2.0 could account for jaw disturbances in the unconstrained mandible condition, and provide researchers with the ability to collect high quality head impact acceleration measurements.

References
[1] Faul M, Xu L, Wald MM, Coronado VG (2010). Traumatic Brain Injury in the United States: Emergency Department Visits, Hospitalizations and Deaths 2002–2006. Atlanta (GA): CDC.

[2] Kuo C, Wu LC, Hammoor BT, Luck JF, Cutcliffe HC, Lynall RC, Kait JR, Campbell KR, Mihalik JP, Bass CR, Camarillo DB (2016). Effect of the mandible on mouthguard measurements of head kinematics. J Biomech 49: 1845-1853.

[3] Siegmund GP, Guskiewicz KM, Marshall SW, DeMarco AL, Bonin SJ. Laboratory Validation of Two Wearable Sensor Systems for Measuring Head Impact Severity in Football Players. Ann Biomed Eng(2015), 1257-1274.

Introduction: Magnetic Resonance Elastography (MRE) is a technique for estimating in-vivo mechanical properties of the brain [1]. Harmonic excitation of the skull induces brain motion, which can be decomposed into bulk motion and dynamic deformation. Commonly, the bulk motion is removed and the wave field is analyzed to estimate the stiffness of different brain regions [1,2,3]. Previously we measured bulk motion and dynamic deformation in the human cerebrum during MRE [4]. Here we compare the motion of the cerebellum and the cerebrum, as the fiber bundles that couple them to each other and the brainstem are potential injury sites [5].

Methods:  Fifteen adult human subjects were scanned using a 3D spiral MRE sequence [2] with 50 Hz harmonic excitation applied at the back of the skull [4]. The MRE volume was 240x240x120 mm³ with 2 mm isotropic voxels and 8 phase offsets. T1-weighted images were also obtained and segmented to create  cerebrum and cerebellum masks, which were then applied to the MRE volume (Fig. 1A). Bulk motion of the cerebrum and cerebellum were individually estimated and subtracted to obtain dynamic deformation in each region. Amplitude-weighted shear wave propagation directions in each region were estimated by directionally filtering the curl field [6].

Results: Bulk motion amplitudes in the cerebrum and cerebellum were 10–28 µm in the anterior-posterior (AP) direction. Amplitudes of dynamic deformation were smaller than bulk motion (Fig. 1B, 1D). Median values of AP dynamic deformation were generally smaller in the cerebellum than in the cerebrum of each subject (slope of 0.89 in Fig. 1E), but inferior-superior and right-left components are larger. The phase and amplitude of AP dynamic deformation differed between the inferior cerebrum and the superior cerebellum (Fig. 1B). In the cerebrum, shear waves propagate inwards from the skull; in the cerebellum, waves originate from the tentorium and back of the skull (Fig. 1C).

Discussion:  Skull vibration induces bulk motion and dynamic deformation in both cerebrum and cerebellum. Differences in phase and amplitude of motion on opposite sides of the tentorium indicates sliding motion of these brain structures relative to the tentorium and each other. Relative motion between cerebellum and cerebrum may be important in TBI [5]; current observations could be used to guide and evaluate computer models of brain biomechanics [7].

Acknowledgements: NIH R01-NS055951, ONR N0001417P7002

References:

[1] Arani et al. Neuroimage 111:59-64 (2015)

[2] Johnson et al. Magn Reson Med 70: 404-412 (2014)

[3] Zhang et al. J Biomech 44:1909-1913 (2011)

[4] Badachhape et al. J Biomech Eng 139:051002 (2017)

[5] MacDonald et al. NEJM, 364:2091-2100 (2011)

[6] Clayton et al. J R Soc Interface 9:2899-2910 (2012)

[7] McGrath et al. Magn Reson Med 78:341-356 (2017)

Introduction

Peripheral nerves are continuously subjected to mechanical forces without suffering functional loss. Inappropriate and traumatic loading are associated with disabling functional losses, such as iatrogenic brachial plexus injury [1]. Factors contributing to loss of nerve function include altered blood supply [2], and cell-level effects, where conduction blocks occur above 5-20% strain [3]. The multi-scale link between tissue deformation and loss of function is poorly understood, and we aim to develop in vitro techniques to measure multi-scale strains in peripheral nerves.

X-Ray diffraction is an ideal modality for in-situ strain measurement of quasi-crystalline microscopic structures within tissues, such as collagen and myelin [4]. We hypothesise that compared with whole tissue, myelin will show much reduced radial compression on axial elongation of peripheral nerve.

Methods

Sciatic nerves harvested from adult Sprague-Dawley rats were imaged in-situ by x-ray diffraction during axial tensile loading, at physiological hydration levels. Diffraction patterns were used to provide fibril level axial collagen strain, and from changes in intra-lamellar spacing radial myelin strain was inferred. Strains were calculated from changes in diffraction peak position.

Results

Following recruitment, axially aligned collagen strained linearly with applied load, and the whole tissue strained non-linearly, similar to other mechanically significant connective tissues. Whole tissue radial compression was non-linear with applied loading (fig. (a)). Microscopically, the myelin sheath compressed with axial elongation (fig. (b)), indicating a partitioning of radial strain to microscopic structures. Myelin radial strain was significantly lower than whole tissue compression.

 

Discussion

Since only epineurial collagen in peripheral nerves is axially aligned these results suggest a major load-bearing role at these strains. Partitioning between macro and microscopic strains suggests that tissue axial strain is a combination of collagen deformation and molecular rearrangements. At low loads, tissue radial strain increased faster than at higher load, suggesting a multi-scale structural rearrangement. The myelin sheath was compressed but at a fraction of induced whole tissue compression, suggesting a high radial stiffness, and a possible mechanically protective function, allowing for nerves to compress as a result of axial strain without the onset of axon damage. Understanding how macroscopic tissue strain is transferred to the microscopic level is key to understand the mechanisms behind conduction blocks and loss of nerve function.

Acknowledgements

The authors acknowledge Diamond Light Source I22 Beamline (proposal SM125-18) and EPSRC (F.B. funding through DTP) 

References

[1] Lalkhen, A. G., & Bhatia, K. (2011). Critical Care & Pain, 12(1), 38-42.

[2] Ogata, K., & Naito, M. (1986 The Journal of Hand Surgery: British & European Volume, 11(1), 10-14.

[3] Takai, Shinro, et al. (2002). Journal of orthopaedic research 20(6), 1311-1314.

[4] Schmitt, Francis O., Richard S. Bear, and Kenneth J. Palmer. Journal of Cellular Physiology 18.1 (1941): 31-42.