Stretch Injury of Human Induced Pluripotent Stem Cell Derived Neurons in a 96 Well Format

Abstract

Traumatic brain injury (TBI) is a major cause of mortality and morbidity with limited therapeutic options. Traumatic axonal injury (TAI) is an important component of TBI pathology. It is difficult to reproduce TAI in animal models of closed head injury, but in vitro stretch injury models reproduce clinical TAI pathology. Existing in vitro models employ primary rodent neurons or human cancer cell line cells in low throughput formats. This in vitro neuronal stretch injury model employs human induced pluripotent stem cell-derived neurons (hiPSCNs) in a 96 well format. Silicone membranes were attached to 96 well plate tops to create stretchable, culture substrates. A custom-built device was designed and validated to apply repeatable, biofidelic strains and strain rates to these plates. A high content approach was used to measure injury in a hypothesis-free manner. These measurements are shown to provide a sensitive, dose-dependent, multi-modal description of the response to mechanical insult. hiPSCNs transition from healthy to injured phenotype at approximately 35% Lagrangian strain. Continued development of this model may create novel opportunities for drug discovery and exploration of the role of human genotype in TAI pathology.

Figure 1. Kinematics of the injury device.

(A) Stage displacement histories over 10 pulses at a range of target amplitudes. (B) Average strain in each well in misaligned configuration with displacement of 2.9 mm (n = 5 measurements per well, average standard error per well = 0.056). Note that C4, D4, E4, F4, C9, D9, E9 and F9 are uninjured control wells. (C) The Lagrangian strain in the membrane increased with increasing stage displacement (n = 200–260 wells over 10 plates, bars = standard deviation). (D) Average strain in each well in optimally aligned configuration with displacement of 3.3 mm (n = 5 measurements per well, average standard error per well = 0.029). (E) Distribution of strains in optimally aligned configuration (circle = average strain in a single well location, square = average strain across all well locations, error bars = 1 standard deviation).

Figure 2. Purity of iCell neurons.

(A) Representative image of uninjured iCell neurons stained with MAP2, GFAP, and Hoechst 33342 demonstrating the absence of astrocytes. (B) Rat astrocyte culture stained and imaged with an identical protocol as a positive control.

Figure 3. Evolution of injury phenotype with increasing strain.

(A) Representative images of neurons stained with calcein AM (green) and Hoechst 333342 (blue) 4 hours after injury at various levels of strain. As strain increases, the neurite network becomes less extensive and the number of calcein-AM negative nuclei increases, indicating cell death. (B) In the control condition, neurites have a large, constant thickness. (C) In injured neurons, neurites are shorter and thinner with beads (see white arrows) distributed along their length.

Figure 4. Injury phenotypes increase with increasing strain.

(A) Mean neurite length per cell declines with increasing strain. (B) Cell viability declines with increasing strain. (C) Processes/cell declines with increasing strain. (D) Branches/cell declines with increasing strain. (E) Viable cells/image declines with increasing strain. (F) Dead cells/image increases with increasing strain. Injury metrics are plotting against well-specific strain. For phenotype measurements, n = 160 wells over 5 plates. For strain measurements, n = 800 wells over 50 plates (i.e. each point represents the average of 5 measurements (average standard error = 0.056). In both cases, the fit line represents a generalized logistic regression of the data (see Methods). Estimates for the coefficients of each fit along with confidence intervals and R² values are presented in Table 2.

Figure 5. Injury with 2 mm stage displacement alters synaptic density.

(A) Representative image of uninjured iCell neurons stained with MAP2, Synaptophysin, and Hoechst 33342. Synaptophysin staining was punctate and distributed across soma and along neurites. (B) Injury did not change the ratio of synaptophysin positive cell area to total cell area (n = 12). (C) Injury did not significantly change the ratio of synaptophysin positive neurite area to total neurite area, although there was a modest downward trend (n = 12). (D) Injury significantly increased the ratio of synaptophysin positive soma area to total soma area (n = 12, *=t-test with significance criterion Bonferroni corrected to p < 0.05/3, bars = standard error).

Figure 6. The in vitro neuronal stretch injury model.

(A) The injury device consists of a stage positioned above an array of Teflon-coated, aluminum posts and driven vertically by an electromagnetic voice coil. (B) The silicone-bottomed plate consists of a commercially-distributed plate top covalently bonded to a sheet of silicone. Since no sandwich construction or air tight gaskets are employed, the standard geometry of the 96 well plate is preserved. (C) Schematic depiction of the injury process. The plate and post array are shown in cross section. Initially, the plate is positioned so that the silicone membrane touches the posts. To induce injury, the plate is lowered, stretching the membrane over the rims of the posts. Posts can be omitted from the post array to create unstretched, control wells.

Figure 7. Cell segmentation.

(A) Neurons injured with 38% strain and imaged 4 hours post injury with calcein AM (green) and Hoechst 33342 staining (blue) (B) Automated segmentation of the image into cell bodies and neurites. Note that beads on neurites are rejected as cell bodies based on their size and the absence of Hoechst-positive nuclei. Extracellular nuclei are rejected as cell bodies based on the absence of calcein AM staining.