Lab on a Chip

View Online

C

Dynamic Article Links <

Cite this: DOI: 10.1039/c1lc20549h

PAPER

www.rsc.org/loc

Valve-based microfluidic compression platform: single axon injury and regrowth†

Downloaded by Johns Hopkins University on 07 October 2011 Published on 06 October 2011 on http://pubs.rsc.org | doi:10.1039/C1LC20549H

Suneil Hosmane,a Adam Fournier,b Rika Wright,b Labchan Rajbhandari,c Rezina Siddique,a In Hong Yang,a K. T. Ramesh,b Arun Venkatesan*c and Nitish Thakor*a Received 23rd June 2011, Accepted 9th September 2011 DOI: 10.1039/c1lc20549h We describe a novel valve-based microfluidic axon injury micro-compression (AIM) platform that enables focal and graded compression of micron-scale segments of single central nervous system (CNS) axons. The device utilizes independently controlled ‘‘push-down’’ injury pads that descend upon pressure application and contact underlying axonal processes. Regulated compressed gas is input into the AIM system and pressure levels are modulated to specify the level of injury. Finite element modeling (FEM) is used to quantitatively characterize device performance and parameterize the extent of axonal injury by estimating the forces applied between the injury pad and glass substrate. In doing so, injuries are normalized across experiments to overcome small variations in device geometry. The AIM platform permits, for the first time, observation of axon deformation prior to, during, and immediately after focal mechanical injury. Single axons acutely compressed (!5 s) under varying compressive loads (0–250 kPa) were observed through phase time-lapse microscopy for up to 12 h post injury. Under mild injury conditions (< 55 kPa) !73% of axons continued to grow, while at moderate (55–95 kPa) levels of injury, the number of growing axons dramatically reduced to 8%. At severe levels of injury (> 95 kPa), virtually all axons were instantaneously transected and nearly half (!46%) of these axons were able to regrow within the imaging period in the absence of exogenous stimulating factors.

1. Introduction Traumatic axonal injuries (TAI) are broadly defined as the focal or multi-focal damage of axons within white matter tracts of the central nervous system (CNS), and can occur in the setting of spinal cord injury (SCI) and traumatic brain injury (TBI).1 Focal axonal injury can result from mechanical forces associated with the rapid deformation (tens of milliseconds to seconds) of white matter regions during trauma.2 Through combinations of compression, stretch, and shear, axon injury arises and often results in an irreversible loss of functional neural connectivity,3 since the scope for axonal regeneration in the CNS is extremely limited.4,5 A number of experimental models have been developed to study traumatic axonal injury (TAI), including in vivo and in vitro approaches, as well as computational models.6 These studies have demonstrated that SCI and TBI not only share many commonalities in their clinical pathology but also a number of a Department of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, MD, USA. E-mail: [email protected] b Department of Mechanical Engineering, Johns Hopkins University Whiting School of Engineering, Baltimore, MD, USA c Department of Neurology, Johns Hopkins University School of Medicine, Baltimore, MD, USA. E-mail: [email protected] † Electronic supplementary information (ESI) available. See DOI: 10.1039/c1lc20549h

This journal is ª The Royal Society of Chemistry 2011

early cellular alterations, including accumulations of proteins, transport disruptions, mitochondrial perturbations, and cytoskeletal changes, that contribute to degenerative secondary injuries.7–13 Therefore, it is of great interest to determine the mechanisms linking primary to secondary injury, as prevention of secondary injuries represents a critical therapeutic target to prevent worsening axonal injury. Since elucidation of cellular mechanisms involved in axonal injury can be difficult to investigate in a complex in vivo setting, many investigators have turned to in vitro approaches. To understand the mechanisms governing axonal degeneration and regeneration after traumatic injury, experimental paradigms must allow the delivery of focal, graded injury to axons. This is critical as axons are typically located in spatially distinct microenvironments as compared to their cell bodies and are often subjected to a wide range of traumatic forces during injury.14 While primary injury results from direct mechanical strain, it remains unclear how forces directly applied to axons translate into changes in axon integrity, degeneration, and regrowth. Therefore, a reductionist approach to understanding TAI at the level of the single axon may provide new insights as to the control of axon degeneration and regrowth. Recently, it has been recognized that lab-on-chip approaches may provide unique solutions to the study of axons.15–23 One such platform introduced a microfluidic analog to the Campenot chamber, in Lab Chip

Downloaded by Johns Hopkins University on 07 October 2011 Published on 06 October 2011 on http://pubs.rsc.org | doi:10.1039/C1LC20549H

View Online

which microfluidic channels passively guided extending axons from one compartment into another.21 Within this type of device, methods have been developed to induce focal axonal injury using advanced optical techniques such as two-photon laser ablation.24–27 However, graded axonal compression, a major component of TAI, has not been addressed.1,14 Existing models suffer from one or more of the following important limitations: 1) lack of a method to inflict focal, graded injury to axons; 2) lack of a precise understanding of the forces exerted upon the axon; 3) lack of compartmentalization of the axon from the cell body, as occurs in vivo; and 4) inability to obtain high resolution imaging of the axon in situ, during and following injury. We have developed a novel axonal injury micro-compression (AIM) platform, inspired by microfluidic valve technology28 to target this unaddressed gap in biomedical technologies. The compatibility of our platform with high-resolution optical microscopy has allowed long term (> 12 h) time-lapse microscopy of cellular events immediately preceding and following axonal injury. Utilizing this technology, we have characterized the response of single CNS axons to focal, graded compressive injury and have correlated outcome to injury forces delivered to the axon. To our knowledge, this is the first characterization of critical thresholds that result in overall maintenance of CNS axon integrity, degeneration, and regrowth.

2. Materials and methods 2.1

Fabrication and assembly of injury platform

The axon injury micro-compression (AIM) platform is an amalgamation of two distinct constructs: (A) a microfabricated chamber with three compartments for neuronal cell bodies, proximal, and distal axons; and (B) a valve-based elastomeric injury pad system for inducing graded injury to micron-scale segments of single axons. The microfabricated chamber was constructed from two layers of poly (dimethylsiloxane) (PDMS) (Sylgard 184; Dow Corning, MI) following well established replica molding protocols.29 Briefly, the master template for the first layer (for neurons and medium) was created using a threelayer microfabrication process. Silicon wafers (WRS Materials, CA) were processed with 6.25-10 mm-thick SU-8 3005 (Microchem, MA) to define two parallel linear arrays of !125 microchannels each (Fig. 1A Top). The process was immediately repeated with a 20 mm SU-8 3025 layer to define the partial height of each of the three compartments (Fig. 1A Middle) and the separation distance between the glass substrate and the injury pad. Finally, a 30 mm SU-8 3025 layer was deposited to define the height of the injury pad (Fig. 1A Bottom) and complete the composite height of all compartments. Once completed, the mold was spun-coat with !45–65 mm-thick layer of PDMS prepolymer and fully cured at 80 " C for 20 mins (ESI Fig. 1B†). The master for the second, control layer, was patterned with SU-8 3050 to define four injury controllers (Fig. 1B) connected by 50 mm-wide control lines to independently addressable access ports (D ¼ 1 mm). This master was used to create a !5 mm-thick PDMS cast (ESI Fig. 1A1†), cut into individual devices, and punched with sharpened gauge #23 needles (ESI Fig. 1A2;† McMaster-Carr, CA) to create access ports. Individual devices and the PDMS-coated wafer were introduced into an oxygen Lab Chip

plasma cleaner (Harrick Plasma, NY) and surface treated (30 Watts; 1.5 mins). Injury controllers were visually aligned to injury pad features, brought into intimate contact, and were baked overnight at 80 " C to fuse device layers (ESI Fig. 1C†). Afterwards, composite devices were removed and neuronal layer access ports were created using 3 mm biopsy punch tools (ESI Fig. 1D;† Huot Instruments, WI). Devices were sterilized by ethanol sonication, autoclaved, and sealed to 50 mm #1 glass bottom petri dishes (Wilco Wells, Netherlands) prior to use (ESI Fig. 1E†). A dye-filled AIM device and a cross-section schematic of the injury compartment can be seen in Fig. 1C,D. Details pertaining to photoresist (soft/hard/post exposure) bake, exposure, and development times can be found in the manufacturer’s technical sheet (Microchem, MA). All PDMS protocols involved standard 10 : 1 base to cross-linker ratios by mass, and detailed devices geometries can be found in Electronic Supplementary Information (ESI) Table 1.† 2.2

Finite element modeling (FEM)

A computational model of the AIM platform was constructed to assess the magnitude of the compressive load applied by the injury pad to the axon and glass substrate. The system was idealized as a two dimensional (2-D) model. Since the stiffness of the axon is several orders of magnitude smaller than that of the PDMS injury pad and glass substrate, the resistance of the axon itself to the applied loads was assumed to be negligible.30 The 2-D plane strain FEM of the device was constructed in the Abaqus 6.9-EF/Standard commercial software package (Dassault Systemes Simulia Corp., RI). The model is composed of three assembled pieces: the glass substrate, the PDMS membrane, and the PDMS control pad. Device dimensions were taken from three-dimensional (3-D) reconstructed confocal images of the device. A section of the FEM is shown in ESI Figure 2.† The types of elements for the PDMS membrane and control pad were 3-node linear triangle (CPE3H) and 4-node bilinear quadrilateral hybrid (CPE4RH) elements. The total number of elements of the PDMS membrane and PDMS control pad were 611 and 269 respectively. The glass substrate used 13578 4-node bilinear quadrilateral elements (CPE4R). Mesh sensitivity studies were conducted to ensure consistent contact force results. As the mesh density was increased, the total contact force varied less than < 1% (ESI Table 3†). The material properties for the structures in the FEM were derived from both experimental measurements and theoretical values. The glass substrate was modeled as linear elastic using properties from the literature.31 Given the nonlinear behavior of PDMS, a hyperelastic Mooney-Rivlin model was chosen to model its response. The stress-strain relationship for a hyperelastic material is ! # $ " vW vW vW vW 2 B$ (1) Iþ þ I1 B s ¼ 2J $1 I3 vI3 vI1 vI2 vI2 where s is the Cauchy stress, W is the strain energy function, B is the left Cauchy-Green deformation tensor, J is the volume ratio, and I1, I2, and I3 are first, second, and third invariants of B, respectively. For a Mooney-Rivlin material, the strain energy function, W, can be described as W ¼ C1(I1 $ 3) + C2(I2 $ 3)

(2)

This journal is ª The Royal Society of Chemistry 2011

Downloaded by Johns Hopkins University on 07 October 2011 Published on 06 October 2011 on http://pubs.rsc.org | doi:10.1039/C1LC20549H

View Online

Fig. 1 The AIM platform was assembled from two independently fabricated templates. The neuronal template (A), which fluidically defined where neurons and medium resided, was constructed as follows: Beginning with a bare silicon wafer, (A Top) an initial resist layer was processed to yield two linear arrays of microchannels. (A Middle) Afterwards, a thicker resist layer was deposited to first define the compartment bases and the injury pad clearance height. (A Bottom) This process was repeated once again to complete the compartment height and define the injury pad geometry. (B) A separate controller template defined which areas of the final device were controlled (or deformed) by compressed air and was fabricated using a single resist layer. (C) A representative AIM device was filled with dye to visually show both the neuronal (blue/green) and control (red) layers. (D) The device cross-section depicts the relative dimensions of the injury pad relative to the compartment and microchannels. Schematics A and B are not to scale.

where C1 and C2 are material constants. The material constants were determined to be C1 ¼ 254 kPa and C2 ¼ 146 kPa by fitting the model to tension data from the literature31,32 and to experimentally obtained compression data of PDMS samples using a Dynamic Mechanical Analyzer (DMA) (ESI Fig. 3†). A comparison between the experimental and modeled stress-strain curve can be found in ESI.† The boundary conditions for the FEM were as follows. The base of the glass substrate was fixed, and the surfaces in contact between the glass substrate, the PDMS membrane, and the PDMS control pad were tied to restrict relative motion between the surfaces. A hard contact constraint was defined between the PDMS compression pad and the surface of the glass, which allowed for frictionless sliding between the PDMS membrane and the glass substrate. Pressure loads were applied uniformly across the top surface of the PDMS membrane within the control pad area in a single static ramp input. For each applied pressure load, the contact force between the PDMS injury pad and glass substrate was computed. The contact force is defined as the total force applied by the injury pad to the glass substrate. For all experiments conducted in this study, the applied pressure load was large enough to ensure that the injury pad fully contacted the glass substrate. The contact pressure was computed by dividing the total force by the contact surface area of 0.054 mm2. This contact pressure provides a quantifiable measure of the applied load to the axon. 2.3

Cell preparation

Primary hippocampal neurons were derived from embryonic day 17 (E17) pups as previously described.33 Prior to cell seeding, This journal is ª The Royal Society of Chemistry 2011

devices were coated overnight at 4 " C with 200 mg mL$1 PDL (Sigma, MO), washed 3x the next day with tissue culture grade H20, filled with neurobasal media,22 and placed in a standard humidified cell culture incubator set to 37 " C and 5% CO2 (Thermo Scientific, MA) for 15–30 mins. Primary neurons were loaded into the somal compartment at a low density (< 100 cells per mm2; 150–450 neurons/device). After 6–8 days in culture, axons could be seen extending into the middle and distal chamber of the device in sparse numbers to allow tracking of individual processes for subsequent experiments. Media was added every 3 to 4 days to maintain neuronal viability. For experiments in which cells were labeled with a fluorescent protein, dissociated neurons were nucleofected (Amaxa, MD) with a plasmid encoding the tau-TdTomato gene as per the manufacturer’s instructions. Efficiency of labeling was greater than 50%. 2.4

Experimental setup

AIM device calibration. Prior to imaging, four independent tubing lines (O.D. ¼ 1.52 mm, I.D. ¼ 0.51 mm; Cole Parmer, IL) connected to gauge #21 blunt needles (McMaster-Carr, CA) were inserted into each of the four injury control ports of the AIM platform (ESI Fig. 1E†). Before cell seeding, the input gas (CO2) pressure was calibrated with a Proportion Air electronic regulator (Equilibar, NC) to define the desired level of pressure applied between the glass substrate and injury pad, an estimate of axonal injury. CO2 was selected as the input gas because of its compatibility with cell culture and availability at most biological imaging centers. Confocal imaging was done for each AIM device to characterize injury pad deflection prior to biological experimentation. Devices were filled with fluorescein Lab Chip

View Online

Downloaded by Johns Hopkins University on 07 October 2011 Published on 06 October 2011 on http://pubs.rsc.org | doi:10.1039/C1LC20549H

isothiocyanate (FITC; Sigma, MO) and imaged in a Zeiss LSM 510 META Confocal Microscope (Zeiss; Germany) at 63X (NA 1.2W; C-Apochromat Korr UV-VIS-IR). Z-stack resolution of !200–250 nm allowed clear visualization of the injury pad contact with the glass substrate, which defined the point at which 0 kPa was applied to the glass substrate/axon. Subsequent increases in applied pressure were then calculated through FEM to estimate the level of axon injury (Fig. 2). Injury application. After characterization, devices were prepared and seeded with neurons as stated previously, then moved to a pre-warmed (37 " C) live-cell microscope, with the stage-cover removed to allow easy access for the control lines. Individual axons located underneath injury pad features (!625– 1250 mm from cell body) were identified and their stage positions were saved within the imaging software. For each experimental condition (e.g. injury pad X, where X ¼ 1 to 4), pressure application was done manually by first adjusting the electronic pressure regulator, then turning the stopcock ninety degrees to allow pressure translation to the injury pad (< 1 s), pressure holding for !5 s, and finally relieving the pressure by turning the stopcock back to the start position. Generally, images were taken immediately before, during, and after injury. Under all injury conditions, the pressure applied to the control network resulted in intimate, but brief contact between the injury pad and glass substrate / axons, which was seen by a loss and subsequent recovery of contrast of the injury pad feature (ESI Fig. 4; ESI

Fig. 2 (A) AIM devices were filled with FITC dye and were imaged under confocal microscopy to characterize injury pad deflection. (B, C) Representative image stacks demonstrated the near linear relationship of the normalized input to injury pad deflection (D) and close correlation to FEM modeling. For the same normalized input as the device images in (B), the corresponding FE results are shown. As membrane thickness between device layers can vary slightly between batches, FEMs were developed to quantitatively assess the relationship between input pressure, membrane thickness, and contact pressure at the glass substrate, an estimate of axonal injury.

Lab Chip

Movie 1†). After injury, tubing connectors were removed, the device was repositioned within the microscope, and the stagecover was added to provide humidified 5% CO2. After thermal equilibration (!15–20 mins), multi-positional time-lapse was initiated to observe axonal response as a function of injury level. 2.5

Axonal injury imaging and analysis

Fluorescence, bright-field and phase-contrast micrographs were captured at 40X (NA 0.6; LD PlnN DIC) or 100X (NA 1.4; PlanApochromat Oil DIC M27) magnification with a Zeiss live-cell inverted microscope (Axio Observer; Zeiss, Germany) using Zeiss Axiovision software. Changes in axon growth cone morphologies were observed by time-lapse brightfield/phase microscopy at fixed time increments (1.5 mins) under constant exposure. Due to the fine temporal resolution of image acquisition, clear assessment of cellular outcome (maintenance vs. degeneration vs. regrowth) per axon was easily accomplished. When obtaining time-lapse micrographs through fluorescence microscopy, the time interval was modified (!30 mins) and total light exposure was reduced (< 150 ms) to minimize phototoxicity. Axon growth rates were calculated in Axiovision by tracing single axon trajectories at fixed time increments. Statistical analysis was done in Prism (GraphPad Software, CA).

3. Results 3.1

Device design and operation

The AIM platform contains two independent fluidic networks that are separated by a thin (!55 mm) PDMS membrane (Fig. 1C,D). The fluid network that intimately contacts the glass substrate is termed the neuronal network and spatially defines the compartments and microchannels where the neuronal cultures reside. The overlaying top network, referred to as the injury control network, defines where pressurized gas (e.g. CO2) is allowed to flow and subsequently deform underlying PDMS features. The neuronal network is comprised of three distinct compartments (cell body, injury, and axonal) connected by arrays of microchannels. The microchannel cross-sectional area and length prevents the migration of neurons and extension of dendrites from the cell body compartment to other regions of the device, but allows unperturbed axonal outgrowth. The injury compartment (L > 8 mm; W ¼ 250 mm; H!50 mm) contains four independent ‘‘push-down’’ structures protruding from the ceiling, termed injury pads, and are aligned orthogonally to the direction of axon growth (Fig. 1D; ESI Fig. 1E†). The geometry of the injury compartment was optimized to allow visualization of individual processes, minimize visual distortion from PDMS features, and facilitate a dynamic range of graded injuries. Each device allows up to four different experimental conditions. 3.2

Device characterization

Confocal imaging. As the membrane thickness can fluctuate from one device batch to another, an experimental and computational approach was taken to normalize device performance. Optical imaging was utilized to measure the membrane thickness and provide data for computational modeling as seen by Fig. 2. This journal is ª The Royal Society of Chemistry 2011

Downloaded by Johns Hopkins University on 07 October 2011 Published on 06 October 2011 on http://pubs.rsc.org | doi:10.1039/C1LC20549H

View Online

Fig. 3 Representative images of axons that continued to grow, degenerate, or regrow as a function of injury level. (A) Under mild injuries (!25 kPa), axons generally continued to grow from left to right as evidenced by progressing growth cones (red triangles). (B) At medium levels of injury (!68 kPa), more axons begun to undergo degeneration as shown by axoplasm disruption and nodal swellings. (C) Under severe compression (!192 kPa), which led to rapid transection, a fraction of axons were able to regrow. These axons were often seen retracting, stuttering, and pausing prior to reformation of the growth cone. Axons (green) were false colored to enhance contrast and allow clear visualization. Scale bar 20 mm.

Here we show both the experimental and FEM-generated crosssections of a representative device under varying pressure loads. The AIM device geometry was optimized such that injury pad deflection under normal operating pressures did not result in non-specific pinching at the microchannel interface (ESI Fig. 5†). Longitudinal (side) image reconstructions demonstrate flat injury pad profiles during pad deflection. Input pressure was normalized to the pressure required for the injury pad to fully contact the glass substrate, and was referred to as the percent deflected.

was created with the same dimensions as the experimental device. The deflection of the injury pad was measured from confocal images of the experimental device for a given input pressure. The same input pressure was applied to the FE model, and the resulting deflection of the injury pad was computed. The computationally determined injury pad deflections showed good agreement with the experimentally measured deflection, demonstrating the accuracy of the finite element solution (Fig. 2). A graph showing the applied input pressure vs. percent deflection for a membrane thickness of 55 mm is shown in ESI Figure 6.†

FEM. For each fabricated injury device, our objective was to quantify the contact pressure between the injury pad and the system of the axon and glass substrate. The applied contact pressure is dependent on the input fluidic pressure from the control network and the thickness of the PDMS membrane. A parametric finite element analysis was conducted, and a linear relationship was determined between the contact pressure, input pressure, and the PDMS membrane thickness using a Trustregion algorithm from MATLAB v7.9.0.529 (R2009b) (The MathWorks Inc., Natick, MA) having 95% confidence bounds:

3.3

f(x,y) ¼ p00 + p10x +p10y

(3)

where f is the contact pressure (kPa), x is the input pressure (kPa), y is the membrane thickness (mm), and pij are the fitting coefficients (ESI Table 4†). This relationship was used to tune the input pressures to achieve the desired magnitudes of applied loading for each device assembly. To validate the FEM solution, the computational results were compared to experimental measurements. A finite element model This journal is ª The Royal Society of Chemistry 2011

Graded axonal response to focal injury

Time-lapse of injury response. Under mild injuries (< 55 kPa), the vast majority of axons remained healthy and continued to grow even after focal compression (Fig. 3A; ESI Movie 2†). As injury level was increased (55–95 kPa), more axons began to degenerate, as seen by a combination of nodal axonal swellings (similar to that seen in Wallerian degeneration34) and thinning of the axon membrane. Both the proximal and distal segments underwent axonal swelling and rapid degeneration (Fig. 3B; ESI Movie 3†). However, above 95 kPa, complete transection, or rapid severing of the axon was seen in all cases. Beyond this injury threshold, a number of axons were able to spontaneously regrow as evidenced by the reformation of the growth cone after an initial presence of dystrophic axonal end-bulbs. Axons that eventually regrew initially retracted, stuttered, paused, and then within 1 to 6 h post injury, were able to reform a growth cone and continue to extend past the point of injury (Fig. 3C; ESI Movie 4†). To further verify the presence of regrowing axons, lowdensity primary hippocampal neurons were transfected with Lab Chip

View Online

Downloaded by Johns Hopkins University on 07 October 2011 Published on 06 October 2011 on http://pubs.rsc.org | doi:10.1039/C1LC20549H

a tau fusion protein (red), which is a microtubule-associated protein that specifically localizes to axons as opposed to dendrites. Time-lapse imaging in the setting of severe compression injury (235 kPa) confirmed axonal regrowth following transection (Fig. 4). Quantification of axonal response. Axonal response was binned into one of three potential outcomes per experimental condition. Individual axons and growth cones were examined frame-byframe and labeled healthy, degenerating, or regrowing as depicted in Fig. 3. Due to the fine temporal resolution (!1.5 mins) between acquired images, axon morphology could be precisely tracked to correctly bin each individual axonal response (Fig. 5). Overall, the percentage of degenerating axons rose as the injury level increased until !95 kPa. However, beyond this threshold, a significant fraction of injured axons began to regrow after injury (!46%; Fig. 5A). Axon growth rates were quantified for both regrowing and control (uninjured) axons by measuring axon displacement at three separate time intervals and averaging the calculated rates for said intervals. Only individual axons that were not growing atop other axons were used for quantification, and therefore only represent a subset of the overall regrowing population. Regrowing axons tended to extend !40% faster (22.0 & 5.8 mm/h), on average, than uninjured axons (15.8 & 4.0 mm/h) within the 12–16 h imaging window (ESI Fig. 7†).

4. Discussion and conclusion Fig. 4 A tau-labeled (microtubule marker; red) axon was subjected to severe (235 kPa) compression injury and images were collected every 30 mins for 8 h post injury. Immediately after injury, the distal segment of the transected axon underwent classic Wallerian degeneration (nodal swellings), while the axon tip (white arrows) first retracted (!30 mins), then began to reform a growth cone (!1 h 30 mins). After reformation, the axon was able to extend past the site of injury. Dotted white lines demarcate the region of the injury pad. Scale bar 25 mm.

Since its discovery, PDMS-based elastomeric valves have become increasingly used in many advanced microfluidic devices. Within the context of neurobiology, this technology has been used in novel device architectures to enable high-throughput nanoaxotomy,24,27 screening of axon-protective small molecules,26 and to facilitate dynamic neuron-glia co-cultures.35 However, to date no microfluidic platforms have utilized the valve structure as a means to deliver mechanical injury to cells or sub-cellular

Fig. 5 Individual axons were binned into one of three following categories: continued growth, degenerating, or regrowing and separated into ranges of injuries: Mild (< 55 kPa), Medium (55–95 kPa), and Severe (> 95 kPa). Quantification of control (uninjured) axons was also done to determine baseline levels of continued growth, degeneration, and regrowth. All experimental conditions were completed in triplicate. Statistical analysis was performed by Tukey pairwise 1-way ANOVA: * ¼ p-value < 0.05, ** ¼ p-value < 0.01, *** ¼ p-value < 0.001. Error bars on graphs correspond to standard errors.

Lab Chip

This journal is ª The Royal Society of Chemistry 2011

Downloaded by Johns Hopkins University on 07 October 2011 Published on 06 October 2011 on http://pubs.rsc.org | doi:10.1039/C1LC20549H

View Online

elements. The AIM platform presented here demonstrates an innovative approach to studying graded axonal injury in vitro. We chose primary rat embryonic CNS cells for these experiments due to their reproducibility of isolation, their extensive characterization in the field of neuroscience, and their ability to survive in culture for extended periods of time. Using these cells, we have demonstrated the ability to examine the relationship between injury thresholds and axon response within our AIM platform. In the case of embryonic CNS cells, our studies suggest that there are unique thresholds that govern the balance between axon survival, degeneration, or regrowth. This study provides a strong starting point for comparative studies in the adult CNS system, which is less permissive to regrowth. A future comparative study between the two systems, as well as with peripheral nervous system neurons in which successful regeneration typically occurs, could enable a deeper understanding of the specific processes responsible for inhibitory or successful regrowth in the adult CNS. This work builds upon findings from other previous in vitro studies, such as one in which graded axonal compression was applied to the squid giant axon.36 Due to the far larger sizes (> 0.5 mm in diameter) of the squid axons as compared to CNS axons, physical manipulation could be readily performed. Their work elegantly showed that as focal axon compression increased, the integrity of the axoplasm diminished up to a critical threshold. Past this injury level, full transection was seen in conjunction with nodal swellings on the proximal axon, a phenomenon we also observe. In another study, axon bundles of superior cervical ganglion (SCG) cells were compressed, which led to many insights to both degenerative and regenerative responses.37 Interestingly, the growth rate of regrowing SCG axons were roughly 595 mm/day, a value that very closely matches what we have seen in our experiments (528 mm/day). While these studies and many others have shed light on cellular mechanisms of axon injury, there has been little quantitative correlation between applied stress and the injury response of CNS axons. Apart from a recent study investigating the mechanical role of microtubules in stretch injury to the axon,38 previous studies have either focused on stress application to the whole neuron,39 thereby masking axon-specific effects, or have utilized stress levels that induce complete axon transection.21,40 While axon regeneration has previously been shown to occur after axotomy within CNS dissociated cultures,41 it is intriguing that our data suggests that only severe, as opposed to mild or moderate, compression was required to induce axonal regrowth in a large fraction of axons. In the peripheral nervous system, the transport of nerve injury signals, both intrinsic and extrinsic, is thought to be crucial to the initiation of regeneration, as the cell bodies of injured neurons may require information regarding the extent and location of injury in order to create a successful regeneration response.42 However, data from our time-lapse studies demonstrate axonal regeneration as quickly as 1–6 h post injury, thus potentially implicating molecular and cellular events at the site of injury.43 Further investigations using the AIM platform may shed light on the secondary injury mechanisms that dictate much of the irreversible axonal damage following trauma. Intracellular calcium fluctuations,44,45 cytoskeletal changes,49,50 and excitotoxicity, all processes implicated in secondary injury, can be This journal is ª The Royal Society of Chemistry 2011

readily studied in the AIM platform. In addition, the glial response to axon injury, which mediates scarring and inflammation, can also be investigated.8,10 We have previously studied the response of microglia to axon degeneration following chemical injury within microfluidic systems.22 Our AIM platform allows the application of both chemical and physical insults and will enable insights into the glial response after direct physical injury, an incompletely understood aspect of traumatic injury. Overall, we have demonstrated that the AIM platform allows unprecedented control of graded injury application and in situ imaging of injury dynamics within a compartmentalized microfluidic environment. We believe that this system and its future derivatives will be invaluable tools in the study of axonal regrowth, axon-glia interactions, and discovery of novel axon protective therapies.

5. Acknowledgements The authors would like to thank Nishant Ganesh Kumar and Clark Zhang for their assistance and Dr Joseph Steiner for his generous gift of the TdT-tau plasmid. This work was funded by the Johns Hopkins Institute for Nanobiotechnology (N.T. and A.V.), Maryland Technology Development Corporation Grant (N.T.), US National Institutes of Health grant 1F31NS06675301 (S.H.), NIDA K08DA22946 (A.V.) and Howard Hughes Medical Institute Early Career Physician-Scientist Award (A.V.).

References 1 K. H. Taber and R. A. Hurley, J. Neuropsychiatry Clin. Neurosci., 2007, 19, iv-104. 2 M. C. LaPlaca, C. M. Simon, G. R. Prado and D. K. Cullen, Prog. Brain Res., 2007, 161, 13–26. 3 T. Koike, Y. Yang, K. Suzuki and X. Zheng, Neurochem. Int., 2008. 4 M. Mattson, Nature, 2004. 5 N. De Stefano, P. Matthews, L. Fu and S. Narayanan, Brain, 1998. 6 H. C. Wang and Y. B. Ma, J. Clin. Neurosci., 2010, 17, 157–162. 7 H. Ouyang, W. Sun, Y. Fu, J. Li, J. X. Cheng, E. Nauman and R. Shi, J. Neurotrauma, 2010, 27, 1109–1120. 8 K. P. Horn, S. A. Busch, A. L. Hawthorne, N. van Rooijen and J. Silver, J. Neurosci., 2008, 28, 9330–9341. 9 J. W. Fawcett, A. Curt, J. D. Steeves, W. P. Coleman, M. H. Tuszynski, D. Lammertse, P. F. Bartlett, A. R. Blight, V. Dietz, J. Ditunno, B. H. Dobkin, L. A. Havton, P. H. Ellaway, M. G. Fehlings, A. Privat, R. Grossman, J. D. Guest, N. Kleitman, M. Nakamura, M. Gaviria and D. Short, Spinal Cord, 2006, 45, 190–205. 10 S. A. Busch and J. Silver, Curr. Opin. Neurobiol., 2007, 17, 120–127. 11 C. M. Simon, S. Sharif, R. P. Tan and M. C. LaPlaca, J. Neurotrauma, 2009, 26, 563–574. 12 Y. C. Chen, D. H. Smith and D. F. Meaney, J. Neurotrauma, 2009, 26, 861–876. 13 V. E. Johnson, W. Stewart and D. H. Smith, Nat. Rev. Neurosci., 2010. 14 I. M. Medana and M. M. Esiri, Brain, 2003, 126, 515–530. 15 A. J. Blake, T. M. Pearce, N. S. Rao, S. M. Johnson and J. C. Williams, Lab Chip, 2007, 7, 842–849. 16 B. G. Chung, L. A. Flanagan, S. W. Rhee, P. H. Schwartz, A. P. Lee, E. S. Monuki and N. L. Jeon, Lab Chip, 2005, 5, 401–406. 17 J. El-Ali, P. K. Sorger and K. F. Jensen, Nature, 2006, 442, 403–411. 18 T. M. Pearce and J. C. Williams, Lab Chip, 2007, 7, 30–40. 19 L. J. Millet, M. E. Stewart, J. V. Sweedler, R. G. Nuzzo and M. U. Gillette, Lab Chip, 2007, 7, 987–994. 20 S. W. Rhee, A. M. Taylor, C. H. Tu, D. H. Cribbs, C. W. Cotman and N. L. Jeon, Lab Chip, 2005, 5, 102–107. 21 A. M. Taylor, M. Blurton-Jones, S. W. Rhee, D. H. Cribbs, C. W. Cotman and N. L. Jeon, Nat. Methods, 2005, 2, 599–605.

Lab Chip

Downloaded by Johns Hopkins University on 07 October 2011 Published on 06 October 2011 on http://pubs.rsc.org | doi:10.1039/C1LC20549H

View Online

22 S. Hosmane, I. H. Yang, A. Ruffin, N. Thakor and A. Venkatesan, Lab Chip, 2010, 10, 741–747. 23 I. H. Yang, R. Siddique, S. Hosmane, N. Thakor and A. Hoke, Exp. Neurol., 2009, 218, 124–128. 24 F. Zeng, C. B. Rohde and M. F. Yanik, Lab Chip, 2008, 8, 653–656. 25 Y. T. Kim, K. Karthikeyan, S. Chirvi and D. P. Dave, Lab Chip, 2009, 9, 2576–2581. 26 C. B. Rohde, F. Zeng, R. Gonzalez-Rubio, M. Angel and M. F. Yanik, Proc. Natl. Acad. Sci. U. S. A., 2007, 104, 13891–13895. 27 S. X. Guo, F. Bourgeois, T. Chokshi, N. J. Durr, M. A. Hilliard, N. Chronis and A. Ben-Yakar, Nat. Methods, 2008, 5, 531–533. 28 M. A. Unger, H. P. Chou, T. Thorsen, A. Scherer and S. R. Quake, Science, 2000, 288, 113–116. 29 S. K. Sia and G. M. Whitesides, Electrophoresis, 2003, 24, 3563–3576. 30 R. Bernal, P. A. Pullarkat and F. Melo, Phys. Rev. Lett., 2007, 99, 018301. 31 G. R. Anstis, P. Chantikul, B. R. Lawn and D. B. Marshall, J. Am. Ceram. Soc., 1981, 64, 533–538. 32 S. J. Lee, et al., J. Micromech. Microeng., 2007, 17, 843. 33 G. B. K. Goslin, Culturing Nerve Cells, 2nd edn, The MIT Press, Cambridge, MA, 1998. 34 M. Coleman, Nat. Rev. Neurosci., 2005, 6, 889–898. 35 D. Majumdar, Y. Gao, D. Li and D. J. Webb, J. Neurosci. Methods, 2011, 196, 38–44. 36 P. E. Gallant, J. Neuropathol. Exp. Neurol., 1992, 51, 220–230. 37 I. Fayaz and C. H. Tator, J. Neurosci. Methods, 2000, 102, 69–79.

Lab Chip

38 M. D. Tang-Schomer, A. R. Patel, P. W. Baas and D. H. Smith, FASEB J., 2009, 24, 1401–1410. 39 M. C. LaPlaca, G. R. Prado, D. K. Cullen and H. R. Irons, Conf. Proc. IEEE Eng. Med. Biol. Soc., 2006, 1, 2384–2387. 40 D. Kilinc, J. M. Peyrin, V. Soubeyre, S. Magnifico, L. Saias, J. L. Viovy and B. Brugg, Neurotoxic. Res., 2010, 19, 149–161. 41 F. Sun and Z. He, Curr. Opin. Neurobiol., 2010, 20, 510–518. 42 N. Abe and V. Cavalli, Curr. Opin. Neurobiol., 2008, 18, 276–283. 43 D. Willis, K. W. Li, J. Q. Zheng, J. H. Chang, A. Smit, T. Kelly, T. T. Merianda, J. Sylvester, J. van Minnen and J. L. Twiss, J. Neurosci., 2005, 25, 778–791. 44 J. A. Wolf, P. K. Stys, T. Lusardi, D. Meaney and D. H. Smith, J. Neurosci., 2001, 21, 1923–1930. 45 J. Knoferle, J. C. Koch, T. Ostendorf, U. Michel, V. Planchamp, P. Vutova, L. Tonges, C. Stadelmann, W. Bruck, M. Bahr and P. Lingor, Proc. Natl. Acad. Sci. U. S. A., 2010, 107, 6064–6069. 46 T. Koike, Y. Yang, K. Suzuki and X. Zheng, Neurochem. Int., 2008, 52, 751–760. 47 M. A. Hilliard, J. Neurochem., 2009, 108, 23–32. 48 J. Q. Zheng, T. K. Kelly, B. Chang, S. Ryazantsev, A. K. Rajasekaran, K. C. Martin and J. L. Twiss, J. Neurosci., 2001, 21, 9291–9303. 49 J. A. Chuckowree and J. C. Vickers, J. Neurosci., 2003, 23, 3715–3725. 50 J. Chetta, C. Kye and S. B. Shah, Cytoskeleton, 2010, 67, 650–665. 51 J. Silver, Curr. Biol., 2009, 19, R444–446. 52 M. P. Coleman and V. H. Perry, Trends Neurosci., 2002, 25, 532–537.

This journal is ª The Royal Society of Chemistry 2011