Here we present a method for a human in vitro model of stretch injury in a 96-well format on a timescale relevant to impact trauma. This includes methods for fabricating stretchable plates, quantifying the mechanical insult, culturing and injuring cells, imaging, and high content analysis to quantify injury.
Traumatic brain injury (TBI) is a major clinical challenge with high morbidity and mortality. Despite decades of pre-clinical research, no proven therapies for TBI have been developed. This paper presents a novel method for pre-clinical neurotrauma research intended to complement existing pre-clinical models. It introduces human pathophysiology through the use of human induced pluripotent stem cell-derived neurons (hiPSCNs). It achieves loading pulse duration similar to the loading durations of clinical closed head impact injury. It employs a 96-well format that facilitates high throughput experiments and makes efficient use of expensive cells and culture reagents. Silicone membranes are first treated to remove neurotoxic uncured polymer and then bonded to commercial 96-well plate bodies to create stretchable 96-well plates. A custom-built device is used to indent some or all of the well bottoms from beneath, inducing equibiaxial mechanical strain that mechanically injures cells in culture in the wells. The relationship between indentation depth and mechanical strain is determined empirically using high speed videography of well bottoms during indentation. Cells, including hiPSCNs, can be cultured on these silicone membranes using modified versions of conventional cell culture protocols. Fluorescent microscopic images of cell cultures are acquired and analyzed after injury in a semi-automated fashion to quantify the level of injury in each well. The model presented is optimized for hiPSCNs but could in theory be applied to other cell types.
TBI is a major cause of mortality and morbidity in the United States, causing around 52,000 deaths and 275,000 hospitalizations every year1. More than 30 clinical trials of candidate therapeutics for TBI have been conducted without a single success2. This uniform failure suggests that human-specific processes separate human TBI from the pathophysiology observed in commonly used pre-clinical rodent models.
The advent of hiPSCNs has created an opportunity to study neurotrauma in a human in vitro model. Drug screening with hiPSCN-based models may deliver results that are more predictive of clinical success than models employing rodent cells. Also, hiPSCNs can be genetically manipulated to isolate and study the effect of individual human genetic variants on pathology3.
The method described in this manuscript is designed to bring the unique advantages of hiPSCN-based disease modeling to neurotrauma. In vitro stretch injury models of neurotrauma are well established4,5,6 with primary rodent cells and human neural cancer cell lines. Most of these models generate stretch by pneumatically loading a silicone membrane. This approach is effective in a single well format but has proven difficult to scale up to a multi-well format7. As a result, there has never been a high throughput screen for agents to treat stretch injured neurons.
In this model, the membrane stretches due to indentation from underneath with a rigid indenter. This approach has been shown repeatedly to generate clinically relevant pathology in vitro in single well systems8,9,10. Our recent work has shown that it easily scales up to a 96-well format while maintaining pulse durations on the order of tens of milliseconds11, which is the time domain of closed head impact events12,13.
In summary, the key advantages of this in vitro injury model are the 96-well format, the use of hiPSCNs, and the clinically relevant time domain of the insult.
1. Silicone Detoxification
2. Plate Fabrication
3. Stretching a Plate
4. Characterizing Membrane Stretch
5. Plating the Cultured Cells
6. Injuring Cultures
7. Microscopy
The stretcher device is capable of moving the stage repeatably with pulse durations as short as 10-15 ms depending on the amplitude of the pulse (Figure 2A). The pulse amplitudes are highly repeatable, but the pulse duration varies by approximately 1 ms between repetitions. The actual pulse amplitude diverges from the prescribed pulse amplitude when a large number of wells are loaded, and the prescribed amplitude is high (see Figure 2B). As the amplitude of stage displacement is increased beyond 3 mm, the actual displacement amplitude increasingly falls short of the prescribed displacement amplitude (see Figure 2B). Careful alignment of the post block eliminates any trend in the membrane strain across rows or columns (Figure 2C). At the 3.5 mm prescribed stage displacement amplitude (3.3 mm actual displacement amplitude) with 52 wells indented, the mean Lagrangian strain across all well locations was 0.451 (standard deviation of means for all locations = 0.051, mean of standard deviations for all locations = 0.065, n = 5 measurements per well). These results are presented here for completeness although some of them have already been reported11.
An optimal, uninjured culture will have few if any clumps of more than 5 cells. Neurites will be individual, long, slender, and curved with little or no sign of tension or beading (Figure 3A). Under ideal conditions, the viability of the cultures should closely approach the viability specified in the manufacturer's data sheet (typically 60-70%) and cultures on silicone should resemble those maintained on conventional rigid culture substrates (Figure 3B). Neurites may or may not be visible on a low power bright field microscope. Laminin concentration and cell density both influence cultures at baseline and after injury. Increasing the cell density increased the number and size of clumps that formed in culture. Increasing the laminin concentration often counteracted this effect (Figure 3A). However, increasing the laminin concentration too much blunted the sensitivity of the cultures to injury (Figure 4). The optimal laminin concentration for uninjured cultures was 50 µg/mL of laminin (Figure 3), but the optimal separation between the sham and stretch injured populations was obtained at 10 µg/mL of laminin (Figure 4). Higher laminin concentrations reduced the sensitivity of the cultures to injury at short time points (Figure 4), but also improved baseline cell viability at longer time points (e.g., 7 days). In summary, it is worthwhile to optimize the laminin concentration for each experimental scenario.
Membrane strain, post-injury imaging time point, laminin concentration, and cell density all exerted a highly statistically significant main effect on the neurite length per cell (ANOVA, p <0.001). The effect of membrane strain on neurite length per cell had highly statistically significant interactions with post-injury imaging time point and laminin concentration (ANOVA, p <0.001) and a statistically significant interaction with cell density (ANOVA, p <0.05). Similarly, membrane strain, post-injury imaging time point, cell density, and laminin concentration all exerted a highly statistically significant main effect on cell viability (ANOVA, p <0.001). The effect of the membrane strain on cell viability had a highly statistically significant interaction with the post-injury imaging time point (ANOVA, p <0.001) and a statistically significant interaction with the cell density (ANOVA, p <0.05). These results prove that timing, cell density, and laminin concentration exert an important influence on the relationship between the applied insult and experimental outcomes, so each should be optimized carefully.
Low cell viability and beaded neurites, along with stunted neurite growth, indicate toxic culture conditions that can arise from improperly prepared silicone. Skipping or shortening the water soak or the oven dry can leave absorbed ethanol or water in the membrane, respectively, which can diffuse into the media and stress the cells. Well-injured cultures will have reduced cell viability, shortened or missing neurites, beaded neurites, and neurites that look taut or tensioned. Injury may induce clumping in cell cultures that were well-dispersed pre-injury. Large clumps can confound the morphological analysis. For morphological analysis, the injury level should be tuned such that noticeable changes occur, but cells are still present with some neurites.
Figure 1: A labeled schematic of the injury device. (A) Top view, (B) isometric view, (C) front view, (D) right side view. The scale bar applies to the orthographic views (A, B and D). Please click here to view a larger version of this figure.
Figure 2: Kinematics of the mechanical insult. (A) The stage displacement histories over 5 pulses at a range of prescribed amplitudes (prescribed amplitudes are listed in the legend) when no wells are loaded. (B) The stage displacement histories over 10 pulses at a range of prescribed amplitudes (prescribed amplitudes are listed in the legend) when 52 wells are loaded. (C) The average strain in each well with stage displacement amplitude of 3.3 mm (n = 5 measurements per well, average standard error per well = 0.029). Note that C4-F4 and C9-F9 are unstretched control wells. This figure has been modified from Sherman et al.11 Please click here to view a larger version of this figure.
Figure 3: Optimization of culture conditions on silicone. (A) The effect of varying cell density and laminin concentration on hiPSCN cultures on silicone. Clumping increases with increasing cell density and decreasing laminin concentration. Cell density and laminin concentration must be optimized to achieve mono-dispersed cultures. Mono-dispersed cultures are less vulnerable to artifacts during quantification. Note that the dynamic range has been adjusted to optimize visualization of the neurites. As a consequence, the much brighter soma are saturated. This presentation is preferred to the alternative of optimizing the dynamic range with respect to the soma, which renders the much dimmer neurites almost invisible. The condition highlighted by the red square was deemed to be optimal for in vitro stretch injury experiments. (B) Under optimal conditions, cultures on silicone membranes appear similar to cultures on conventional rigid substrates. The left panel shows hiPSCNs cultured at 33,750 cells/cm2 with 3.3 µg/mL of laminin on a conventional, rigid 96-well plate (the cell culture substrate is a tissue culture treated cyclic olefin co-polymer). The right panel reproduces the panel outlined in red from (A). Scale bars = 100 µm. Please click here to view a larger version of this figure.
Figure 4: The injury phenotype and its dependence on laminin concentration. (A) Healthy culture, using 10 µg/mL laminin and 67,500 cells/cm2. Neurites are long with no beads. There are few dead nuclei, and few clumps. (B) Culture using the same culture conditions, injured with 57% peak strain and imaged after 4 h. Neurites are shortened or missing, and some have beads (indicated by arrows). There are fewer Calcein AM-positive cells and more Calcein AM-negative (i.e., dead) nuclei. Injury has increased clumping among the surviving cells. (C) 4 h after injury, neurite length per cell declines with increasing strain in a manner that depends on the laminin concentration. (D) 4 h after injury, cell viability declines with increasing strain in a manner that depends on the laminin concentration. (E) 24 h after injury, neurite length per cell declines with increasing strain in a manner that depends on the laminin concentration. (F) 24 h after injury, cell viability declines with increasing strain in a manner that depends on the laminin concentration. (n = 4 per bar, error bars are ± 1 standard deviation, Scale bars = 100 µm). Strain values are deduced from stage displacement using data from a prior publication by Sherman et al.11 Please click here to view a larger version of this figure.
Supplementary Figure 1: Technical drawing of the indenter. Please click here to download this figure.
Supplementary Table 1: Custom Built Devices. Please click here to download this table.
Supplementary Table 2: 96 Well Plate-loader Pinout Wiring Diagram. Please click here to download this table.
Supplementary Code File 1: Computer-aided design drawings of the injury device. Please click here to download this file.
Supplementary Code File 2: Computer-aided design drawings of the plate fabrication clamp. Please click here to download this file.
Supplementary Code File 3: 3D representation of the stamp geometry, suitable for use with a 3D printer. Please click here to download this file.
Supplementary Code File 4: SubVI for MuStLiMo_si_initialize.vi, which is a SubVI for motion_control.vi. Converts entries in dialog boxes into parameters for motion. Please click here to download this file.
Supplementary Code File 5: SubVI for Multiple Straight Line Moves_simplified.vi, which is a SubVI for motion_control.vi. Converts entries in dialog boxes into parameters for motion. Please click here to download this file.
Supplementary Code File 6: SubVI for position_tracker.vi. Counter tracks displacement input from linear encoder. Please click here to download this file.
Supplementary Code File 7: Base LabVIEW Project. Please click here to download this file.
Supplementary Code File 8: Top level VI that moves the device. Please click here to download this file.
Supplementary Code File 9: SubVI for motion_control.vi. Executes the rapid displacement that stretches the plate. Please click here to download this file.
Supplementary Code File 10: SubVI for motion_control.vi. Executes the slow displacement that moves the stage. Please click here to download this file.
Supplementary Code File 11: SubVI for motion_control.vi. Plots a (usually undersampled) displacement history in the motion_control.vi control panel. Please click here to download this file.
Supplementary Code File 12: Top level VI that records the displacement history. Please click here to download this file.
Supplementary Code File 13: Contains Variable2, which communicates between motion_control.vi and position_tracker.vi. Please click here to download this file.
Supplementary Code File 14: Schematic for printed circuit board. Please click here to download this file.
Supplementary Code File 15: Layout for printed circuit board. Please click here to download this file.
The key to obtaining a consistent, biofidelic phenotype in this model is applying a consistent biofidelic mechanical insult. This model can generate pulse durations as short as 10-15 ms, which are similar to the pulse durations for human head impacts according to cadaveric experiments12,13. The consistency of this insult depends on the alignment of the plate with the indenter block and consistent lubrication of the indenters. When the indenter block is well aligned, there is no trend in the applied strain across rows or columns (Figure 2C). A thin layer of lubricant typically creates less friction than a thick layer, and viscous greases are not recommended because they foul the silicone and obstruct the passage of light during microscopy. The actual stage displacement amplitude can fall substantially short of the prescribed displacement amplitude when many indenters are used, and the prescribed stage displacement amplitude is large (> 3 mm). However, while the actual displacement is less than the prescribed displacement at large amplitudes, it remains repeatable (Figure 2B). Therefore, large, actual displacements amplitudes can be reliably obtained by entering a prescribed value in excess of the desired value. Displacement amplitude matters only because it is an easily recorded proxy for the peak membrane strain, which directly measures the mechanical insult that induces pathology. Therefore, the procedure described for determining membrane strain from stage displacement is critical. This process should be repeated if any major changes are made to the system that affect the interaction between the plate and the indenters, for example if different diameter indenters, different indenter materials or coatings, or different types of silicone bottomed plate are used. The process of realigning the indenter block and determining the zero position should be repeated at the start of each experiment. A schematic of the stretching device is shown in Figure 1. CAD models required to reproduce the device are provided as supplemental materials: 'Injury Device – FULL ASSEMBLY – Generic 3D.STEP'; the associated bill of materials provided as 'Supplementary Table 1: Custom Built Devices – BOM.xlsx. Also see Supplementary Table 2 96 Well Plate _loader – Pinout Wiring Diagram.xlsx, which describes the cabling connections that connect various components of the systems. 'Interconnector_circuit_board.dip' describes a circuit board that interconnects the cables.
If the device is deactivated with the stage near the middle of its travel, the stage will move after the power is cut off because it is spring-loaded. When the power is restored, the feedback loop will detect a large difference between the last known prescribed position and the actual position. This will cause the stage to move suddenly to the position it was in when the device was deactivated. This sudden motion can cause errors in the output of the encoder, so care should be taken to deactivate the device only when it is in its unpowered resting position at the top of its travel.
The fabrication clamp is designed to bring the plate body and silicone bottom together in a manner that allows optimal bonding. To this end, there are three key features in the design presented in the supplemental file 'Press Die – Generic 3D.STEP'. First, the clamp plate body holder is parallel to the silicone bottom. If this is properly built, it will require no adjustment after the initial setup. Second, the layer of foam rubber in the clamp provides a small amount of compliance under the plate, as a completely rigid system would theoretically experience a sudden increase from zero clamping force to infinite clamping force when the clamp was closed. The position of the crossbar and set screw of the clamp are adjustable so that the distance between the two sides of the clamp can be fine-tuned.
Every effort should be made to provide a bright, white background behind the dot on the well bottom during strain characterization experiments. The better the contrast in these images, the easier it will be to automate the process of measuring the height and width of the dot, which can become tedious for a human operator analyzing a large experiment. High-speed videography of the bottom of a well in a 96-well plate presents challenges because the walls of the well tend to cast shadows. The use of a dome light or diffuse axial light that can illuminate along the line of sight of the camera without obscuring the image eliminates shadows or specular reflections that would arise with a conventional light source. The brightest available light source should be used because bright illumination allows images to be acquired with a short exposure time. Short exposure times minimize motion blur. Upgrading the light emitting diodes (LEDs) in the diffuse axial light allows shorter exposure times during high-speed video acquisition. The LEDs can be upgraded by opening the diffuse axial light, removing the stock LEDs, mounting 4 high power LED arrays to the back pane using LEDs holders, connecting them to a constant current power supply, and reassembling the diffuse axial light (see Table of Materials for catalog numbers). The disadvantage of upgrading the LEDs is that the passively cooled LEDs cannot be kept on for more than a few seconds due to the risk of overheating. Therefore, a different light is needed for alignment of the post-block and camera adjustment.
The presented method of quantifying membrane strain by measuring the dilation of a dot stamped on to the membrane is relatively crude, but it can scale up to multiple wells in a robust manner. The strain field across the well bottom can be characterized in more detail using digital image correlation. This technique involves spraying a speckled pattern onto the base of the well and then imaging it at high-speed during deformation. Commercial software can then be used to quantify strain at every point in the image by tracking the evolution of the speckled pattern.
This protocol produces a multi-faceted, clinically relevant, stretch injury phenotype in hiPSCNs. Cell death, neurite degeneration, and neurite beading are all well-documented sequelae of TBI in humans and animal models15. The key to success in this model is establishing and maintaining healthy cultures. Generally speaking, a cell culture protocol developed with conventional rigid plates is a worthwhile starting point for stretchable plate culture. However, the possibility that the cells in question may respond differently on silicone must always be considered. This is particularly true of hiPSCNs, which are very sensitive to culture conditions. Some examples of optimizing cell density and laminin concentration are supplied in the Representative Results section (Figure 3, Figure 4). Activation of the silicone with plasma treatment is vital. Silicone is hydrophobic and unreactive; in its natural state, it will not bind to laminin or other molecules used to promote cell attachment. Plasma treatment renders the surface hydrophilic and exposes reactive groups. These changes allow adhesion molecules to bind to the silicone and promote cell attachment. It is important to note that the plasma treatment effect dissipates within minutes unless the surface is submerged in liquid, and so procedures that involve drying the activated surface should be performed as quickly as possible. A simple way to check if the effect of plasma treatment has worn off is to place a droplet of water on the surface. On untreated silicone, the droplet will bead up whereas on plasma treated silicone, it will spread out. With the hiPSCNs that we used (see Table of Materials), the manufacturer recommends adding the laminin with the cell suspension rather than pre-coating. This protocol has incorporated this approach successfully. While segmentation can, in theory, be accomplished with open source software or general-purpose programming languages, a high degree of proficiency with these tools is required to obtain good results. Neurites are frequently difficult to distinguish from background signal because they are so slender. Therefore, we recommend the use of commercial software tools distributed by high content microscopy companies with dedicated modules for segmentation and quantification of neurons, if they are available. Even with commercial software, it is wise to export images of the segmentation to visually verify accuracy.
There are some limitations associated with working in stretchable plates compared to working with conventional, rigid plates. Stretchable plates can be imaged as normal with air objectives. However, imaging with immersion objectives is very difficult. Lens oil may damage the silicone. Additionally, the objective exerts pressure on the silicone membrane as it moves upwards. This pressure displaces the membrane vertically, making it difficult to bring the sample into focus. The silicone membranes currently used in fabricating the plates are approximately 250 µm thick. This thickness exceeds the focal distance of many high power, immersion objectives. Special care must be taken to lay the membranes perfectly flat before clamping to achieve the flatness required for microscopy. Autofocus systems can compensate for deviations in the flatness of the finished plate to some extent. Future versions of the protocol may pre-tension the membrane before it is bonded to the plate top to ensure flatness. The adhesive-free procedure for bonding the silicone membrane to the plate top14 is considered an important strength of the current protocol. It eliminates the risk of neurotoxicity from the adhesive as well as any deviations in flatness due to non-uniform thickness of the adhesive layer.
Multi-electrode arrays are commonly used in experiments with hiPSCNs to assess their maturity and functionality. Unfortunately, these systems are incompatible with this model because the cell culture substrate is rigid. It is possible to create a stretchable multi-electrode array, although this has so far only been demonstrated in a single well format16,17. Note that indenters can be removed individually from the indenter block so that some wells are not indented and can serve as shams. Removing the indenter prevents indentation but does not completely eliminate mechanical loading since there is still inertial motion of the fluid in the wells while the stage is moving. It is worth comparing these wells to wells in plates that were never subject to stage motion to measure any pathological influence of fluid motion. Also, the array of indenters in the block should be bisymmetric (symmetrical from front to back and side to side). This precaution ensures that the plate is evenly loaded during indentation, so that the stage does not tilt sideways and cause the rods to bind in their bearings.
One of the primary challenges to therapeutic innovation in neurotrauma is the complexity and heterogeneity of the condition. Trauma applies multi-modal stress to every cell type in the central nervous system simultaneously. Neurons have been reliably generated from human induced pluripotent stem cells (hiPSCs) and are now widely available from commercial vendors. Innovation is proceeding quickly in this field, and other neural cell types such as astrocytes18 and microglia19 are also being derived from hiPSCs. It may soon be possible to isolate the cell-autonomous responses of each of these cell types to trauma in vitro and then to co-culture different cell types to understand how they communicate after trauma. In this way, it may ultimately be possible to recreate the clinical challenge from the bottom up to thoroughly understand it in a human system. This approach is distinct from the conventional approach relying on rodent models and has the potential to generate novel insights that lead to the first therapies for this common, devastating, and intractable condition.
The authors have nothing to disclose.
This work is supported in part by a grant from the National Institutes of Health (R21NS098129). We would like to acknowledge excellent technical assistance from SueSan Chen, Jonathan Tan, Courtney Cavanaugh, Shi Kai Ng and Feng Yuan Bu, who designed and built a structure to support lights used during high speed imaging experiments described in this manuscript.
.010" Silicone Sheet | Specialty Manufacturing, Inc | #70P001200010 | Polydimethylsiloxane (PDMS) sheet |
Sparkleen | Fisher Scientific | #043204 | |
Nunc 256665 | Fisher Scientific | #12-565-600 | Bottomless 96 Well Plate |
Kim Wipes | ULINE | S-8115 | |
Plasma Cleaner | Harrick Plasma | #PDC-001-HC | |
(3-Aminopropyl) triethoxysilane | Sigma-Aldrich | #440140 | APTES |
Parchment Paper | Reynolds | N/A | |
Dome Light | CCS inc | LFX2-100SW | |
Dome Light Power Supply | CCS inc | PSB-1024VB | |
Axial Diffuse Lighting Unit | Siemens | Nerlite DOAL-75-LED | Diffuse axial light |
High Power LED Array | CREE | XLamp CXA2540 | High Power LED Array |
LED holder | Molex | 1807200001 | LED Holder |
LED power supply | Mean Well | HLG-320H-36B | Constant Current Power Supply |
FastCam Viewer software | Photron | camera softeware | |
Fastcam Mini UX50 | Photron | N/A | High Speed Camera |
Micro-NIKKOR 105mm f/2.8 | Nikon | #1455 | High Speed Camera Lens |
0.1 mg/mL Poly-L-Ornithine | Sigma-Aldrich | #P4597 | |
iCells | Cellular Dynamics International | #NRC-100-010-001 | |
iCell media | Cellular Dynamics International | #NRM-100-121-001 | |
iCell supplement | Cellular Dynamics International | #NRM-100-031-001 | |
Laminin | Sigma-Aldrich | #L2020 | |
Hoechst 33342 | Fisher Scientific | #H3570 | |
Calcein AM | Fisher Scientific | #C3099 | |
voice coil actuator | BEI Kimco | LA43-67-000A | |
optical linear encoder | Renishaw | T1031-30A | |
servo drive | Copley Controls | Xenus XTL | |
Controller | National Instruments | cRIO 9024 Real Time PowerPC Controller | |
cRIO chassis | National Instruments | cRIO 9113 | |
digital input module | National Instruments | NI 9411 | |
data acquistion chassis | National Instruments | NI 9113 | |
LabVIEW | National Instruments | instrument control software | |
hiPSCNs | Cellular Dynamics International |