Comparative Analysis of Chemotherapy-Induced Peripheral Neuropathy in Bioengineered Sensory Nerve Tissue Distinguishes Mechanistic Differences in Early-Stage Vincristine-, Cisplatin-, and Paclitaxel-Induced Nerve Damage

Abstract

Chemotherapy-induced peripheral neuropathy (CIPN) is a well-known, potentially permanent side effect of widely used antineoplastic agents. The mechanisms of neuropathic progression are poorly understood, and the need to test efficacy of novel interventions to treat CIPN continues to grow. Bioengineered microphysiological nerve tissue (“nerve on a chip”) has been suggested as an in vitro platform for modeling the structure and physiology of in situ peripheral nerve tissue. Here, we find that length-dependent nerve conduction and histopathologic changes induced by cisplatin, paclitaxel, or vincristine in rat dorsal root ganglion-derived microphysiological sensory nerve tissue recapitulate published descriptions of clinical electrophysiological changes and neuropathologic biopsy findings in test animals and human patients with CIPN. We additionally confirm the postulated link between vincristine-induced axoplasmic transport failure and functional impairment of nerve conduction, the postulated paclitaxel-induced somal toxicity, and identify a potential central role of gliotoxicity in cisplatin-induced sensory neuropathy. Microphysiological CIPN combines the tight experimental control afforded by in vitro experimentation with clinically relevant functional and structural outputs that conventionally require in vivo models. Microphysiological nerve tissue provides a low-cost, high-throughput alternative to conventional nonclinical models for efficiently and effectively investigating lesions, mechanisms, and treatments of CIPN. Neural microphysiological systems are capable of modeling complex neurological disease at the tissue level offering unique advantages over conventional methodology for both testing and generating hypotheses in neurological disease modeling.

Impact Statement Recapitulation of distinct hallmarks of clinical CIPN in microphysiological sensory nerve validates a novel peripheral neurotoxicity model with unique advantages over conventional model systems.

Chemotherapy-induced peripheral neuropathy (CIPN) is an adverse side-effect of common antineoplastic treatments. Symptoms include hypersensitivity to touch or temperature, tingling, numbness, neuropathic pain, autonomic dysfunction, muscle weakness, and/or loss of motor coordination resulting from damage to peripheral sensory, motor, and/or autonomic nerves (Flatters et al., 2017). Dose reduction or termination of chemotherapy remains the only effective options for moderating CIPN, both of which compromise cancer treatment efficacy (Hou et al., 2018). Continuation of chemotherapy despite onset of CIPN can result in persistent, progressive, and/or permanent symptoms (Manji, 2013). Continued advances in cancer treatment have resulted in an increasingly large population of cancer survivors suffering from permanent CIPN (Brewer et al., 2016). Thus, the need is growing for basic science to understand the mechanisms responsible for the onset and progression of CIPN, determine the critical points at which CIPN can become permanent, and develop novel strategies to ameliorate or reverse its effects.

Vinca alkaloids, taxanes, and platinum-based antineoplastic agents are 3 major classes of chemotherapeutic compounds known to induce CIPN with high frequency (Staff et al., 2017). Prototypic agents for these chemical classes—vincristine (vinca alkaloid), paclitaxel (taxane), and cisplatin (platinum compound)—all induce peripheral neuropathy in a dose-dependent manner (Cavaletti et al. 1992; Forsyth et al. 1997; Sandler et al. 1969). Vincristine and paclitaxel slow cancer growth through impairment of cytoskeletal function while cisplatin-induced platinum-DNA adducts disrupt DNA tertiary structure and impair cell cycle regulation (Schiff et al., 1979; Stryckmans et al., 1973; Zwelling et al., 1979). However in peripheral nerves, vincristine- and paclitaxel-induced cytoskeletal dysfunction concurrently results in impaired axoplasmic transport, preventing organelle delivery to more distal neurite outgrowth and resulting in length-dependent distal neuropathy (Sahenk et al., 1987; Shemesh and Spira, 2010), and cisplatin-induced platinum-DNA adducts concurrently disrupt neuronal gene expression and induce apoptosis through impaired cell cycle regulation and/or mitochondrial stress (Gill and Windebank, 1998; McDonald and Windebank, 2002; Scuteri et al., 2009; Ta et al., 2006). All 3 compounds progressively impair neuroaxonal health, albeit through distinct cellular mechanisms, resulting in similar but distinct clinical presentation of CIPN.

Clinical CIPN evaluation begins with physician-led examination of patient-reported sensory, motor, and/or autonomic symptoms followed by quantitative sensory and autonomic reflex testing and electrophysiological sensorimotor nerve conduction studies (NCSs; Watson and Dyck, 2015). NCS measure sensory nerve action potentials and compound muscle action potentials generated after electrical stimulation of sensory and motor nerves, respectively (Tavee, 2019). NCS indicate that vincristine induces length-dependent mixed distal neuropathy, affecting sensory and motor nerves similarly (Guiheneuc et al., 1980). Cisplatin induces length-independent sensory neuropathy with concurrent distal and proximal sensory nerve impairment while largely sparing motor nerve function (Krarup-Hansen et al., 2006; Thompson et al. 1984). Paclitaxel induces length-dependent distal motor neuropathy, similar to vincristine, but length-independent sensory neuropathy, similar to cisplatin (Lipton et al., 1989; Pizzamiglio et al., 2020).

3D microphysiological nerve culture has unique advantages over conventional methodologies, which rely heavily on rodent and primary dorsal root ganglion (DRG) culture models, each with significant limitations (Hoke and Ray, 2014; Lehmann et al., 2020). 3D neuronal cultures better mimic in vivo nerve structure (Peretz et al., 2007; Sun et al., 2016) and functionality (Chandrasekaran et al., 2017; Lai et al., 2012) with greater efficiency than in vivo animal models (Ahadian et al., 2018; Duval et al., 2017). We recently developed a 3D, bioengineered microphysiological DRG-sensory nerve model that recapitulates the polarized, multicellular, 3D tissue structure of native peripheral nerve (Curley and Moore, 2011) and supports electrophysiological recording of long-distance compound action potentials (CAPs) analogous to clinical sensory nerve action potentials (Huval et al., 2015; Khoshakhlagh et al., 2018). We have found that electrophysiological evaluation of neurotoxicant-induced dysfunction in this model is more efficient and sensitive than conventional analysis of cell viability (Kramer et al., 2020). Here, we hypothesize that microphysiological peripheral nerve will recapitulate clinically observed, length-dependent electrophysiological characteristics of CIPN and will distinguish the functional and structural attributes that differentiate vincristine-, paclitaxel-, and cisplatin-induced CIPN.

MATERIALS AND METHODS

 

 

Experimental design

Initial concentration-effect experiments (analogous to in vivo dose-response experiments) were conducted to estimate the half-maximal inhibitory concentration (IC₅₀) of vincristine, paclitaxel, cisplatin, forskolin, and aspirin in microphysiological peripheral nerve based solely on high-throughput electrophysiological analysis. Chemotherapy is halted or reduced upon initial onset of peripheral neuropathy to prevent widespread neurodegeneration. We hypothesized that a near-IC₅₀ concentration would best recapitulate early-onset neurotoxicity, prior to late-stage neurodegeneration, that is most relevant to clinical CIPN. Forskolin primarily induces demyelination (Zhu and Glaser, 2008) and was included for comparison to the demyelination that occurs secondary to neurodegeneration at high chemotherapeutic doses. Aspirin is a commonly used medication that does not cause nerve lesions that was included as a negative control. Five separate concentration-effect experiments were performed, one for each agent in isolation, to estimate IC₅₀ values. A sixth and final experiment was performed for in-depth histopathological and electrophysiological comparison of neurotoxicity induced by near-IC₅₀ concentrations of vincristine, paclitaxel, cisplatin, and forskolin within a single batch of identical microphysiological nerve constructs to minimize any potential culture-to-culture variability.

Generation of 3D spheroid cultures

All animal handling and tissue harvesting procedures were performed according to guidelines set by the American Veterinary Medicine Association and approved in advance by the Institutional Animal Care and Use Committee at Tulane University. Timed-pregnant Long-Evans rats were purchased from Charles River Laboratories. All DRG from an entire embryonic day 15 (E15) rat litter (including both male and female pups) were harvested, pooled, incubated in 0.25% trypsin in phosphate-buffered saline (PBS) with EDTA, pH 7.4 (Thermo-Fisher Scientific, Waltham, Massachusetts), at 37°C for 15 min, dissociated through trituration, and passed through a 40 µM nylon mesh filter. Cells were then seeded in 96-well ultra-low attachment spheroid microplates (Corning Inc., Corning, New York) at a concentration of 45 000 cells per well in spheroid formation media composed of Neurobasal Medium supplemented with 2% v/v B27 supplement, 1% v/v GlutaMAX, 20 ng/ml nerve growth factor 2.5S native mouse protein (NGF), and 1% v/v antibiotic/antimycotic solution (all from Thermo-Fisher). Microplates were centrifuged at 500 g for 5 min, and 3D spheroid cultures aggregate after overnight incubation at 37°C and 5% CO₂.

Micropatterning microphysiological nerve cultures

Fabrication and validation of 3D peripheral nerve cultures has previously been described extensively (Bowser and Moore, 2019; Curley and Moore, 2011; Huval et al., 2015; Khoshakhlagh et al., 2018; Kramer et al., 2020; Sharma et al., 2019). Briefly, single spheroids were seeded at one end of dual-hydrogel culture scaffolds using digital projection lithography similar to a previously described method (Bowser and Moore, 2019). Outer-gels were fabricated by irradiating a solution of 10% w/v polyethylene glycol dimethacrylate (Polysciences Inc., Warrington, Pennsylvania), 1.1 mM lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP; Allevi, Philadelphia, Pennsylvania), and 0.0001% w/v TEMPO (Millipore-Sigma, St. Louis, Missouri) in PBS (pH 7.4) with ultraviolet (UV) light patterned to create a long, thin inner void surrounded by a growth-restrictive mold. A single-DRG spheroid was placed at one end of each outer-gel mold with a pipette, and the void was filled with an inner-gel solution composed of 4% w/v gelatin methacrylate (Allevi), 0.55 mM LAP, 5.76% v/v 1-vinyl-2-pyrrolidone (Millipore-Sigma), and 0.004 mg/ml natural mouse laminin (Thermo-Fisher Scientific) in PBS. Inner gels were then cured with UV light to suspend the spheroid in a growth-permissive 3D hydrogel scaffold.

Maturation of microphysiological nerve cultures

Assembled dual-hydrogel DRG-spheroid cultures were then incubated for 7 days in growth medium composed of Basal Medium Eagle (Thermo-Fisher) supplemented with 1% v/v Insulin-Transferrin-Selenium supplement, 1% v/v GlutaMAX, 2 mg/ml bovine serum albumin (Thermo-Fisher), 4 mg/ml glucose (Millipore-Sigma), 20 ng/ml NGF, and 1% antibiotic/antimycotic solution. Cultures were then matured for 24–28 days in maturation media composed of the same growth medium additionally supplemented with 15% fetal bovine serum (Thermo-Fisher) and 0.04 mg/ml ascorbic acid (Millipore-Sigma). Neurites extended unidirectionally from the site of spheroid placement along a growth-permissive channel and mature over the course of the 31- to 35-day growth and maturation period. Mature constructs are comprised of a head (or ganglion) region (“DRG”) located at the site of spheroid placement (containing the highest concentration of neuronal soma) and a tail region of neurite outgrowth (“nerve”) reaching up to 8 mm in length (Figure 1A).

Figure 1.

Electrophysiological data acquisition. A, Fully mature nerve construct stained by fluorescent immunohistochemistry with the neuron-specific marker beta-III tubulin (red) and the Schwann cell-specific marker myelin basic protein (green). The head region (“DRG”) contains the highest concentration of neuronal soma from which neurite migration extends up to 8 mm distal along the growth-permissive channel. Neurites (“nerves”) are stimulated at proximal (B), middle (C), and distal (D) sites, and the resulting compound action potentials (CAPs) are recorded as they propagate through the head region. Each CAP trace represents the average response to 10 replicate stimulations.

Concentration-effect treatment protocol

All agents were purchased from Millipore-Sigma. Mature nerve cultures were treated with increasing concentrations of vincristine (vincristine sulfate, ≥95% purity, stored at −20°C), paclitaxel (≥95% purity, stored at 4°C), cisplatin (≥99% purity, stored at room temperature [RT]), forskolin (≥98% purity, stored at RT), and aspirin (acetylsalicylic acid, ≥99%, stored at RT); agent purity was established by the manufacturer. A separate batch of constructs was manufactured for each of the 5 concentration-effect experiments yielding 30–40 identical microphysiological cultures. Between 4 and 8 identical constructs were treated with each concentration of each agent. Drug treatment was performed in Basal Medium Eagle supplemented only with 4 mg/ml glucose and 1% antibiotic/antimycotic solution to prevent conflagration of any neuroprotective effects of additional supplements with neurotoxic effects. All agents were dissolved in 0.1% dimethyl sulfoxide (DMSO) except for cisplatin; because platinum reacts with DMSO, cisplatin was dissolved directly in media. All treatments were performed for 7 days with a single media change on the fourth day of treatment.

Concentration-effect field potential recording

After completion of treatment, impulse conduction in microphysiological nerve cultures was evaluated using extracellular, field potential recording similar to previous extensive description (Kramer et al., 2020). Briefly, cultures were removed from the incubator and continuously bathed with room-temperature artificial cerebrospinal fluid continuously bubbled with 95% O₂/5% CO₂. A platinum and glass recording electrode (1–2 MΩ resistance) was placed in the “head region” containing the highest density of neuronal soma while a platinum stimulating electrode was placed 4–6 mm distal to the recording electrode in the middle region of neurite outgrowth as shown in Figure 1. Ten bipolar 15 V, 200 µs stimuli were delivered at 0.5 Hz to the middle of nerve outgrowth the resulting CAP propagation was recorded in the ganglion region. These 10 traces were then averaged into a single trace representative of each sample.

Electrophysiological data collection

Five amplitude-related metrics and 2 latency-related metrics were extracted from each 10-trace average using a custom algorithm written in Igor Pro (v7.08). This algorithm separately identifies all negative- and positive-going peaks present in each trace that are at least twice the amplitude of the prestimulation background noise and then interpolates a separate novel function through each set of peaks. Integrated trace area was calculated by integration of the area between the positive and negative peak functions to estimate total biological activity. The amplitude of each negative-going peak was calculated as the difference between the positive and negative peak functions at the time point identified for each peak. Maximum peak amplitude represents the amplitude of the largest negative-going peak. Average peak amplitude represents the average of all negative-going peak amplitudes. Summed peak amplitude represents the sum of all negative-going peak amplitudes. The peak count represents the total number of identified negative-going peaks. The latency of each peak was calculated by measuring the amount of time between the stimulus artifact and the point at which each peak reached its local maximum value. The recording distance was calculated by measuring the distance between the stimulating (“nerve”) and recording (“DRG”) electrodes using a calibrated camera, and onset velocity was calculated by dividing the recording distance by the latency to the first peak. The modal velocity was calculated by dividing the recording distance by the latency to the peak with the largest amplitude.

IC50 estimation

Concentration-effect curves were generated based solely on electrophysiological readings. Values for integrated trace area, maximum peak amplitude, average peak amplitude, summed peak amplitude, peak count, onset latency, and modal latency were calculated for each sample and averaged across all replicates at each dosage. Average values were plotted against drug concentration and resulting plots were fit with a nonlinear regression using Prism 8 (version 8.4.2; GraphPad Software, Inc, San Diego, California) to estimate an IC₅₀ value for each metric and drug condition.

Comparative microphysiological CIPN

The use of 6 replicate samples was determined to be optimal for detection of drug-induced neuropathy during the course of the 5 previous concentration-effect experiments. Therefore, another set of 30 identical, mature cultures were fabricated and randomly assigned to 1 of 5 treatment groups (n = 6 per group). Microphysiological nerve constructs were treated with either 1 nM vincristine, 100 nM paclitaxel, 6 µM cisplatin, 20 µM forskolin, or vehicle alone (0.1% DMSO) for 7 days, representing the near-IC₅₀ concentration of each chemotherapeutic and forskolin estimated by the preceding concentration-effect experiments.

In-depth comparative electrophysiology

After treatment had completed, in-depth electrophysiological analysis was performed to compare and contrast the distinct microphysiological neuropathies induced by near-IC₅₀ concentrations of vincristine, paclitaxel, cisplatin, and forskolin. To assess the length-dependence of electrophysiological impairment, CAP propagation into the neuronal cell body cluster (“DRG”) was sequentially recorded after neurite stimulation at 3 locations: 6–8 mm (distal), 4–6 mm (middle), and 2–3 mm (proximal) from the recording site. To assess changes in nerve sensitivity, a full stimulus-response curve was conducted at each stimulation site by sequentially delivering 10 bipolar, 200 µs, 0.5 Hz stimulations with increasing amplitudes of 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, and 15 V. The 10 replicate stimulations were again averaged into a single trace representative of that stimulation location and amplitude and descriptive metrics were obtained from each 10-trace average as in the previous concentration-effect experiments.

Statistical analysis of electrophysiological metrics

At each stimulation site, each of these 7 descriptive electrophysiological metrics were analyzed with a 2-way mixed model analysis of variance (ANOVA, with a within-subjects factor of “stimulus voltage” and a between-subjects factor of “drug treatment”, n = 6 per voltage and drug treatment) in Prism 8. Dunnett’s post hoc test was then utilized to probe for significant differences from the vehicle condition for each drug treatment at each stimulation voltage. To provide a complete overview of all metrics evaluated, the p-values of statistically significant post hoc tests were arranged into the heat map shown in Figure 4B. The 4 main columns of the map define the vehicle comparison to vincristine, paclitaxel, cisplatin, and forskolin treatments from left to right. The 3 main rows define these comparisons after proximal, middle, and distal stimulation locations from top to bottom. Each main column contains 12 subcolumns representing each of the 12 different stimulation amplitudes. Each main row contains 7 subrows representing each of the 7 different descriptive metrics obtained from the electrophysiological waveform. Together each color-coded cell represents the p-value obtained for the post hoc comparison to the vehicle condition of each combination of treatment, stimulation location, stimulation amplitude, and descriptive metric. Colored cells represent significant post hoc comparisons while black cells represent comparisons that were not statistically significant.

Figure 4.

Summary of proximal, middle, and distal stimulus-response curves obtained from microphysiological nerve constructs treated with vincristine, paclitaxel, cisplatin, forskolin (positive control), or vehicle alone. A, Raster plots of the compound action potential (CAP) propagation recorded in the ganglion field during proximal, middle, and distal stimulus-response experiments. For each plot, the Y-axis indicates the increasing stimulus intensity from 0 to 15 V, whereas the X-axis represents the time since stimulus delivery. The intensity of each pixel represents the summed signal amplitude across all replicate samples for that combination of stimulus intensity and time poststimulation. B, Graphical summary of significant post hoc differences from the vehicle condition across a battery of descriptive metrics. The X-axis represents the voltage along the stimulus-response curve, repeated for vincristine, paclitaxel, cisplatin, and forskolin moving from left to right. Each row along the Y-axis represents a distinct metric to describe CAP traces, repeated for distal, middle, and proximal stimulation locations moving from bottom to top. A separate 2-way mixed model ANOVA (with a within-subjects factor of “stimulus voltage” and a between subjects factor of “drug treatment”, n = 6 cultures per treatment) was performed for data obtained after proximal, middle, and distal stimulations. The color of each cell represents the magnitude of the p-value obtained from a Dunnett’s post hoc comparison to the vehicle condition, as indicated in the color key (inset, top right), for each combination of descriptive metric, stimulus location, and stimulus intensity represented by that pixel location.

Histopathological sample preparation

Following electrophysiological testing, the same constructs were washed with 0.1 M phosphate buffer (PB) and fixed in a mixture of 2% methanol-free formaldehyde (“paraformaldehyde,” diluted from 16% paraformaldehyde in deionized water from Electron Microscopy Sciences [EMS], Hatfield, Pennsylvania) and 2.5% glutaraldehyde (diluted from 25% glutaraldehyde in deionized water from EMS) in PB at 4°C for at least 48 h. Constructs were then washed with PB containing 0.08 M glycine (Millipore-Sigma) in PB to eliminate unreacted fixative and washed again with PB to remove excess glycine. Constructs were postfixed in 1% osmium tetroxide (OsO₄, diluted from 2% OSO₄ in deionized water from EMS) for 2 h at RT with, stained with 25% UranyLess uranyl acetate replacement stain (EMS) in ethanol for 2 h at RT (Millipore-Sigma), dehydrated with ethanol, infiltrated and embedded in low-viscosity hard plastic (Spurr) resin (EMS), and cured overnight at 65°C. Sectioning as well as light microscopic (LM) and transmission electron microscopic (TEM) evaluations were performed at the Shared Instrumentation Facility at Louisiana State University (Baton Rouge, Louisiana). Briefly, ultrathin (0.5-µm thick) sections were cut using a Leica EM U27 microtome (Leica, Wetzlar, Germany) at both the ganglion (head) region containing cell soma and in the nerve (neurite outgrowth) at 4.5 mm distal from the ganglion. For LM, approximately 1-µm thick sections were stained with 0.5% toluidine blue (TB) in 2% sodium borate and then imaged using a Leica DM6 B microscope (Leica, Wetzlar, Germany). For TEM, 80- to 100-nm thick sections were placed on Formvar carbon-coated copper grids (200 mesh) and then impregnated with metal by floating on droplets of 2% uranyl acetate for 20 min at RT. They were then rinsed with deionized water droplets 3 times for 1 min. Grids were viewed using a JEOL 1400 TEM (Peabody, Massachusetts) with an accelerating voltage of 120 kV at varying magnifications.

Histopathological data collection

Findings were scored manually in cross sections of the ganglion (head) and nerve (aligned neurites) by a scorer blinded to treatment using Fiji-ImageJ (v2.0.0). Lesion descriptions and interpretations subsequently were cross-checked by an ACVP boarded-certified veterinary pathologist. LM analysis of TB-stained images of the ganglion sections were used to assess somal morphology. The entire area of a single TB-stained section containing a complete stitched cross-section of the ganglion region was analyzed for each sample. Images were converted to 32-bit grayscale and processed with the standard CLAHE local contrast enhancement to normalize the distribution of pixel intensities within each image. Block size was set to 1.5 times the somal diameter, 256 histogram bins, and a maximum slope of 3. All cell soma, nuclei, and nucleoli present in ganglion sections were outlined in Fiji-ImageJ, and all ImageJ standard descriptors of shape and staining intensity were exported for analysis including area, circularity (4π*area/perimeter²), roundness (4*area/π*major axis²), staining intensity, and solidity (area/convex area). The value of each instance of each descriptor was averaged across all values obtained from the same sample, resulting a single value for each of the 6 replicate constructs within each treatment group.

Both LM analysis of TB-stained sections and TEM analysis of ultrathin sections were used to assess morphology of nerve tissue (neurite outgrowth region). The entire area of a single TB-stained section containing a complete stitched cross-section of the nerve tissue region was analyzed for each sample. All myelin sheaths and non-neuronal cells present in nerve sections were counted in TB-stained section and were normalized to the total area of nerve tissue scored generating a single value for each of the 6 replicate constructs within each treatment group. Nerve sections were also sampled through high-magnification TEM analysis to obtain ultrastructural metrics. Approximately 25 high-magnification images were taken from all parts of each cross-section of neurite outgrowth to obtain a fair sample of the tissue cytoarchitecture for each construct, encompassing approximately 3% of the entire section area. All images were manually scored to identify instances of free vesicle accumulation, incompletely myelinated axons, and unmyelinated axons by a scorer blinded to the treatment. Aberrant organelle accumulation was scored when a myelinated axon displayed more than one vesicle/µm² axonal area. Unmyelinated axons were scored when a clearly defined axon lacked any myelination. Incomplete myelination was scored when a myelin ring did not form a complete loop around an axon. The number of instances of each of these features was normalized to the total area scored for each sample. The severity of the accumulation was further quantified as the number of vesicles counted normalized to the total axonal area scored for each sample. The distances of the outer and inner edges of the myelin sheaths from the neurite centers were also measured in ImageJ to calculate the G-ratio, or ratio of the inner radius (axon radius, center to inner margin of the sheath) to the outer radius (nerve fiber [axon + nerve sheath] radius). The G-ratio was calculated for each identifiable instance of myelination and all instances for each sample were averaged to generate a single value for each sample.

Statistical analysis of histopathological metrics

Histopathological data collection resulted in a single value for each of the 6 replicate constructs within each treatment group across all histopathological metrics collected. A 1-way ANOVA (n = 6 per drug treatment) with Dunnett’s post hoc test was conducted in Prism 8 to test for significant differences from the vehicle across treatment groups condition for each histopathological descriptor.

RESULTS

 

Electrophysiological Data Collection

Electrical stimulation anywhere along the nerve (neurite outgrowth) length of mature constructs resulted in CAP conduction through the somal head region (Figs. 1B–D). Comparison of CAP generation after distal, middle, or proximal stimulation was used to define the length-dependence of chemotherapy-induced deficits in nerve conduction, thereby providing a measurement analogous to clinical NCS.

Histomorphological Data Collection

Histomorphology of neuronal soma was analyzed through LM analysis of TB-stained sections taken through the head (“DRG”) region (Figure 1A). In sections of untreated constructs, healthy neuronal soma is defined by a dark cytoplasmic region averaging 18 µm in diameter. A well-defined, pale nucleus was apparent in 44.6% of soma, and a small, dark nucleolus was clearly identifiable in 48.1% of neuronal nuclei. The space between neuronal soma contained a dense mesh of both finely stippled extracellular matrix and dark-staining, small (<10 µm in diameter), oval to fusiform satellite glial cells. Histomorphology of neurite outgrowth was analyzed through both LM analysis of TB-stained thick sections (Figure 2B) and TEM analysis of ultra-thin sections (80–100 nm) (Figure 2C) of neurites (“nerve”) taken 4–5 mm distal from the head region. In TB-stained sections, myelinated axons (<5 µm in diameter) were encircled by a dark myelin sheath surrounding a lighter central axon. The space between myelinated axons was occupied by relatively dense tissue comprised of finely stippled extracellular matrix, small (<10 µm in diameter), dark staining, elliptical glial (Schwann) cells, and a small number of larger (18 µm in diameter) neurons. Counts of nonmyelinated axons and intracellular vesicles as well as accurate measurement of myelin sheath dimensions were obtained from high-magnification TEM images. Together, the clustered cells (head) and the aligned bundle of radially projecting neurites (tail) comprise a mini-PNS. The neuron-rich cell cluster in the head region approximates the neural components of a DRG, while the dense neurite outgrowth in the tail region represents its nerve. The derivation of the original tissue suggests that this mini-PNS most closely resembles an afferent somatic sensory pathway.

Figure 2.

Histomorphology of neural cell soma and neurite outgrowth. A, Densely packed neuronal soma in the head region (“DRG”) are identified through low-magnification light microscopic analysis of toluidine blue (TB)-stained sections. Healthy soma (an example outlined in red) are large in size (18 μm average diameter), show distinct separation of the cytoplasmic and nuclear compartments (arrow), and often display a dark nucleolus (dotted circle). B, Myelinated axons and glial (Schwann) cells in the neurite outgrowth (“nerve”) are identified through light microscopic analysis of TB-stained sections. Healthy myelin sheaths are narrow (<2 μm diameter) and appear as dark rings surrounding lightly stained axons (dotted circles indicate 5 examples). Healthy glial cells appear as darkly stained, oval nuclei often closely associated with myelinated axons (arrows indicate 5 examples). C, Greater subcellular detail in high-magnification transmission electron microscopic micrographs. Many small (<2 μm diameter) unmyelinated axons exhibit pale axonal cytoplasm speckled with very thin (approximately 10 nm diameter), darkly staining cytoskeletal filaments (dotted circles indicated 3 examples). Less often, larger axons are surrounded by black, multilayered myelin sheaths (arrow) and contained darkly stained intra-axonal vesicles (arrowhead, approximately 200 nm diameter, consistent with mitochondria).

The IC50 Values of Vincristine-, Paclitaxel-, and Cisplatin-Induced Neurotoxicity Can Be Rapidly Identified Through Electrophysiological Analysis of Microphysiological Nerve Tissue

Estimated IC₅₀ values did not vary appreciably across metrics, so only integrated trace area was chosen to illustrate the comparison among various chemotherapeutic agents. Vincristine was found to have the lowest toxic concentration with an IC₅₀ of 0.81 nM, followed by paclitaxel at 70.6 nM and cisplatin at 6.57 µM (Figs. 3A and 3B). IC₅₀ values for forskolin and aspirin could not be calculated; there was not a smooth reduction in integrated trace area as forskolin concentration increased, and aspirin did not result in a significant reduction in integrated trace area at any concentration (Figure 3C).

Figure 3.

Electrophysiological results of chemotherapeutic and control compound concentration-effect experiments. The integrated trace area quantifies the amplitude of compound action potentials generated by bioengineered nerves treated with increasing concentrations of vincristine, paclitaxel, cisplatin, forskolin, or aspirin. A, All 3 chemotherapeutic compounds resulted in a smooth reduction in integrated trace area indicating progressive functional impairment of nerve impulse conduction as drug concentration increased (data plotted as mean ± SEM, n = 4–8 cultures per concentration). B, The IC50 value for each chemotherapeutic compound was calculated by fitting each plot to a nonlinear function in Prism 8. Vincristine was found to be the most toxic compound in this system with an IC50 of 0.81 nM, followed by paclitaxel at 70.6 nM, and cisplatin at 6.57 μM (data plotted as IC50 ± 95% CIs). C, Treatment with increasing concentrations of aspirin (negative control) and forskolin (positive control) did not result in a smooth reduction in the resulting signal amplitude (data plotted as mean ± SEM, n = 4–8 cultures per concentration).

Direct Comparison of Electrophysiological and Histopathological Effects of Chemotherapeutic Compounds at Near-IC50 Concentrations

The raw data from all electrophysiological stimulus-response curves are plotted below (Figure 4A). Stimulus-response curves were readily obtained across all stimulation sites from vehicle-treated samples. CAP generation was detected after simulation as low as 2V, and traces grew larger in amplitude as stimulation intensity increases. Five amplitude-related and 2 latency-related descriptive metrics were calculated for each construct at each stimulation voltage. Each metric was then averaged across all constructs within each treatment group at each location of stimulation. For each stimulation site, statistical significance was calculated with a 2-way mixed model ANOVA (with a between-subjects factor of “treatment” and a within-subjects factor of “stimulus voltage”) followed by the Dunnett’s post hoc test for significant differences from the vehicle control group. Significant post hoc differences are summarized in the heat map below (Figure 4B).

Paclitaxel Treatment Prevented CAP Generation and Caused Extensive Histopathological Lesions Indicative of Severe Neurotoxicity in Both the Ganglion and Nerve Regions

CAP generation was barely detectable in constructs treated with paclitaxel across all stimulation sites (ie, length-independent; Figure 4A). Statistically significant distal impairment of CAP generation was detected with stimulations as low as 3V, while middle and proximal impairment of CAP generation was detected with stimulations as low as 2V (Figure 4B). As stimulation voltage increased at all 3 stimulation sites, significant impairment was identified across all 5 amplitude-related metrics and both NCV-related metrics.

Histopathological analysis of paclitaxel-treated cultures identified multiple morphological abnormalities in both the ganglion and nerve regions. Paclitaxel treatment altered somal morphology relative to vehicle-treated constructs in several ways including increased somal area (F_[4,25] = 14.56, p < .0001, Dunnett’s p < .0001); decreased nuclear circularity (F_[4,25] = 5.034, p = .0041, Dunnett’s p = .0031); decreased nuclear roundness (F_[4,25] = 9.786, p < .0001, Dunnett’s p = .0007); decreased cytoplasmic TB staining intensity (F_[4,25] = 12.72, p < .0001, Dunnett’s p < .0001]; and decreased nuclear TB staining intensity (F_[4,25] = 4.300, p = .0088, Dunnett’s p = .0047; Figs. 5A, 5B, and 5D–H). There was also a significant effect of drug treatment on the proportion of soma with identifiable nuclei (F_[4,25] = 2.907, p = .0420), although none of the Dunnett’s post hoc tests were significant, the largest reduction occurred after paclitaxel treatment (p = .0888) (Figure 5I). Paclitaxel treatment significantly reduced the density of detectable myelin sheaths in the region of neurite outgrowth (F_[4,25] = 6.023, p = .0015, Dunnett’s p = .0017) while having no effect on the density of glial cells in the region of neurite outgrowth (Figs. 6A, 6B, 6D, and 6E). Finally, high-magnification TEM imaging indicated that paclitaxel treatment significantly increased the G-ratio (ie, axon diameter relative to nerve fiber [axon + myelin sheath] diameter), indicating a decrease in myelin thickness, relative to vehicle-treated constructs (F_[4,25] = 6.386, p = .0012, Dunnett’s p = .0016) and significantly decreased the presence of identifiable unmyelinated axons (F_[4,25] = 4.174, p = .0100, Dunnett’s p = .0254; Figs. 7A, 7C, 7F, and 7H).

Figure 5.

Toluidine blue-stained light micrographs reveal histopathology of paclitaxel and cisplatin in neuronal soma in the head region. Representative light micrographs of constructs treated with (A) vehicle alone, (B) paclitaxel, or (C) cisplatin. Treatment with paclitaxel induced multiple neuron somal abnormalities including (D) enlarged soma, (E) decreased nuclear circularity, (F) decreased nuclear roundness, (G) decreased cytoplasmic staining intensity, (H) decreased nuclear staining intensity, and (I) a trend toward reduction in the proportion of soma with identifiable nuclei. Treatment with cisplatin also induced neuron somal abnormalities including (J) a trend toward reduction in the number of identifiable nucleoli and a (K) a significant reduction in solidity. Data were analyzed with a 1-way ANOVA followed by Dunnett’s post hoc test, n = 6 cultures per treatment, and plotted as mean ± SEM (*p < .05 vs vehicle condition).

Figure 6.

Toluidine blue-stained light micrographs reveal histopathology of paclitaxel and cisplatin in the region of neurite outgrowth (“nerve”). Representative light micrographs of constructs treated with (A) vehicle alone, (B) paclitaxel, or (C) cisplatin. Treatment with paclitaxel resulted in (D) a significant decrease in the number of identifiable myelin sheaths around axons but (E) had no effect on the presence of glial cells. Treatment with cisplatin had (D) no effect on the presence of myelinated axons but (E) resulted in a significant reduction in the number of glial cells. Data were analyzed with a 1-way ANOVA followed by Dunnett’s post hoc test, n = 6 cultures per treatment, and plotted as mean ± SEM (*p < .05 vs vehicle condition).

Figure 7.

Transmission electron micrographs reveal pathology of vincristine, paclitaxel, and forskolin in neurite outgrowth (“nerve”). Representative electron micrographs of constructs treated with (A) vehicle alone, (B) vincristine, (C) paclitaxel, and (D) forskolin. E, The severity of aberrant organelle accumulation was significantly increased in constructs treated with vincristine relative to constructs treated with vehicle alone. F, The density of unmyelinated axons identifiable at high magnification was significantly reduced in paclitaxel-treated constructs. G, The incidence of incomplete myelin sheaths was significantly increased in constructs treated with forskolin relative to constructs treated with vehicle alone. H, The G-ratio (ie, axon diameter relative to nerve fiber [axon + myelin sheath] diameter) was significantly increased in constructs treated with paclitaxel or forskolin relative to constructs treated with vehicle alone, indicating that these 2 treatments reduced the average myelin sheath thickness. Data were analyzed with a 1-way ANOVA followed by Dunnett’s post hoc test, n = 6 cultures per treatment and plotted as mean ± SEM (*p < .05 vs vehicle condition).

Cisplatin Treatment Results in Length-Independent Impairment of Action Potential Generation and Loss of Nonneuronal Cell Density

Cisplatin-treated nerve constructs demonstrated length-independent impairment of CAP generation, with similar severity of impairment at proximal, middle, and distal stimulation sites (Figure 4A). Statistically significant distal impairment of CAP generation was evident at stimulation voltages as low as 5V, and impairment was detected in 4 of the 5 amplitude-related metrics but neither NCV-related metric as stimulation voltage increased. Middle and proximal impairment of CAP generation was evident at stimulation voltages as low as 3V, and impairment of CAP generation was detected across all 5 amplitude-related metrics and both NCV-related metrics as stimulation voltage increased (Figure 4B).

Histopathological analysis of cisplatin-treated constructs identified both neuron somal and nonneuronal cell abnormalities. Cisplatin treatment resulted in a significant decrease in solidity in the shape of neuron somal borders, indicating an increase in concave invaginations relative to vehicle treatment alone (F_[4,25] = 3.842, p = .0144, Dunnett’s p = .0426; Figs. 5A, 5C, and 5K). There was also a significant effect of drug treatment on the proportion of nuclei with identifiable nucleoli (F_[4,25] = 4.073, p = .0112), although none of the Dunnett’s post hoc tests were significant, the largest reduction occurred after cisplatin treatment (p = .0898; Figs. 5A, 5C, and 5J). Cisplatin treatment also resulted in a significant decrease in the density of glial cells in the region of neurite outgrowth relative to vehicle treatment alone (F_[4,25] = 4.771, p = .0053, Dunnett’s p = .0052) but had no effect on the density of myelinated axons (Figs. 6A and 6C–E). Cisplatin treatment had no effect on organelle accumulation, myelin sheath morphology, or unmyelinated axon density (Figure 7).

Vincristine Treatment Resulted in Length-Dependent Impairment of CAP Generation and Increased the Accumulation of Organelles in Axons

Vincristine-treated nerve constructs demonstrated length-dependent impairment of CAP generation, with more severe impairment distally rather than proximally (Figure 4A). Statistically significant distal impairment of CAP generation was severe with stimulations as low as 4V, and significant impairment was identified across all 5 amplitude-related metrics and both NCV-related metrics as stimulation voltage increased (Figure 4B). Impairment of CAP generation at the middle site was detected with stimulations as low as 6V, and evidence of impairment in both amplitude and NCV-related metrics was sporadic and modest as stimulation voltage increased. No significant impairment of proximal CAP generation was detected at any stimulation voltage.

Histopathological analysis of vincristine-treated constructs identified a significant increase in aberrant organelle accumulation by 1-way AVOVA (F_[4,25] = 5.637, p = .0057) and the Dunnett’s post hoc test determined it was significantly higher in constructs treated with vincristine that vehicle alone (p = .0045; Figs. 7A, 7B, and 7E). Vincristine treatment had no effect on neuron somal morphology in the ganglion region (Figure 5) and hand no effect on neuronal or glial cell density in the region of neurite outgrowth (Figure 6) or myelin morphology (Figs. 7A, 7B, 7G, and 7H).

Forskolin Treatment Impaired Distal CAP Generation after Low-Voltage Stimulation and Increased the Prevalence of Thin and Incompletely Myelinated Axons

Forskolin-induced length-dependent impairment of CAP generation was limited to low-voltage stimulation at the most distal stimulation site (Figure 4A). Statistically significant distal impairment of CAP generation was evident at stimulation voltages as low as 3V (Figure 4B). Significant impairment in 3 of the 5 amplitude-related metrics and both latency-related metrics were detected as stimulation voltage increased to 6V but was not detected at stimulation voltages above 6V. No significant impairment of middle or proximal CAP generation was detected. The effects of forskolin treatment on CAP generation were subtle and distinct from the effects of the 3 antineoplastic chemotherapeutics in that forskolin-induced impairment was only evident after low-intensity stimulation of distal neurite outgrowth while CIPN-induced impairment was evident after both low and high-intensity stimulation.

Histopathological analysis of forskolin-treated constructs specifically identified myelin-related abnormalities rather than neuronal or axonal damage. Forskolin treatment resulted in a significant increase in the number of incomplete myelin sheaths (F_[4,25] = 4.270, p = .0095, Dunnett’s p = .0099) and a significant increase in G-ratio (F_[4,25] = 6.386, p = .0012, Dunnett’s p = .0169) relative to treatment with vehicle alone but had no effect on unmyelinated axon density or aberrant organelle accumulation (Figs. 7A and 7D–H). Forskolin treatment had no effect on neuron somal morphology (Figure 5) and had no effect on the density of myelinated nerve fibers or glial cells (Figure 6).

DISCUSSION

Clinical chemotherapy is reduced or halted upon initial onset of CIPN to prevent permanent nerve dysfunction from widespread neurodegeneration (of ganglionic neurons and/or neurites and/or myelinating cells). It is therefore critical to clinical relevance that nonclinical CIPN modeling represents nerve function in this therapeutic window, which has proven difficult when modeling CIPN in rodents (Hoke and Ray, 2014). Electrophysiological analysis of microphysiological nerve tissue rapidly generates quantitative neurotoxic concentration-effect curves from which the degree of early and late-stage CIPN can be predicted objectively. Our model estimates an IC₅₀ of 0.81 nM for vincristine, 70.6 nM for paclitaxel, and 6.57 µM for cisplatin, consistent with clinically defined cumulative neurotoxic doses of 5–15 mg/m² for vincristine, 200 mg/m² for paclitaxel, and 300–400 mg/m² for cisplatin (Grisold et al., 2012). Subsequent in-depth electrophysiological and histopathological analysis confirmed that neuropathy induced by near-IC₅₀ concentrations of vincristine and cisplatin in microphysiological nerve tissue resembled early-stage CIPN, whereas neuropathy induced by a near-IC₅₀ concentration of paclitaxel resembled late-stage neurodegeneration. Similar to clinical NCS (Guiheneuc et al., 1980; Krarup-Hansen et al., 2006; Thompson et al., 1984), vincristine-induced microphysiological nerve conduction impairment was length-dependent and more severe at more distal stimulation sites while cisplatin-induced microphysiological nerve conduction impairment was length-independent, occurring to a similar degree after distal, middle, and proximal stimulations. The near-IC₅₀ concentration of paclitaxel resulted in severe impairment of nerve conduction across all stimulation sites. Although the length-independence of paclitaxel-induced neuropathy described here correlates with reported clinical outcomes (Lipton et al., 1989; Pizzamiglio et al., 2020), the severity of functional impairment and the associated histopathological observations suggest that the concentration of paclitaxel used in this study induced neurodegeneration more consistent with late-stage CIPN. These data support our hypothesis that treatment of microphysiological peripheral nerve with chemotherapeutics recapitulates the length-dependence of nerve conduction impairment seen in clinical CIPN.

Microphysiological nerve tissue also resolved similar but distinct histopathological abnormalities induced by these 3 chemotherapeutic compounds and the control comparison, forskolin. Vincristine-induced microphysiological CIPN was characterized by a specific and unique increase in the severity of axoplasmic transport failure with no apparent abnormalities of neuronal soma or glial cells, cisplatin-induced microphysiological CIPN was characterized by mild abnormalities of neuronal soma, undetectable axonopathy, and extensive gliotoxicity, paclitaxel-induced microphysiological CIPN was characterized by extensive neuroaxonal toxicity but undetectable gliotoxicity, and forskolin-induced primary demyelination while having no detectable effect on neuronal morphology. Thus, the data presented here support the hypothesis that microphysiological nerve tissue distinguishes the functional and structural attributes that differentiate vincristine-, paclitaxel-, and cisplatin-induced CIPN and primary demyelination.

The microphysiological data described here also offer novel insights into early development of vincristine- and cisplatin-induced CIPN. Lesions similar to the vincristine-induced organelle accumulation shown here have been described in other neuropathies (Cappello et al., 2016; Smith et al., 2016) and it has long been hypothesized that failed axoplasmic transport results in both axonal organelle accumulation and length-dependent distal nerve conduction failure. The data presented here demonstrate for the first time that impaired axoplasmic transport and length-dependent nerve conduction impairment occur concurrently in vincristine-treated nerve tissue and that such effects develop prior to any significant morphological evidence of axonal or somal degeneration, confirming this long-held hypothesis. Paclitaxel is also known to disrupt cytoskeletal function, but axoplasmic transport failure could not be assessed in paclitaxel-treated constructs because neurodegeneration was so severe that axoplasmic structures could not be identified. It is possible that lesions consistent with axoplasmic transport failure would develop at lower concentrations of paclitaxel. Furthermore, we found that cisplatin-treated microphysiological nerves demonstrated significant loss in glial cell density prior to extensive neuronal cell abnormalities. This finding is consistent with previous in vivo histopathological (Roytta and Raine, 1986; ter Laak et al., 2000; Xiao et al., 2011) and in vitro studies (Guo et al., 2017; Imai et al., 2017; Konings et al., 1994; Melli et al., 2008) indicating that treatment with cisplatin, but not paclitaxel or vincristine, alters the morphology and function of satellite glial cells (SGCs) surrounding DRG neuronal soma. These studies speculate that gliotoxicity may result in impaired nerve conduction, but the data presented here demonstrate for the first time that glial cell loss and length-independent reductions in nerve conduction occur concurrently in cisplatin-treated nerve tissue prior to significant neurodegeneration. The contribution of nonneuronal cell toxicity to CIPN is not yet well understood and represents an interesting avenue for further investigation using microphysiological nerve tissue.

Paclitaxel has long been known to disrupt cytoskeletal function in axonal extensions (Lipton et al., 1989; Pizzamiglio et al., 2020) but may also be directly toxic to sensory ganglia outside the protection of the blood brain barrier (Barriere et al., 2012; Cavaletti et al., 2000). Here we identified histopathological abnormalities of both the ganglion region and nerve outgrowth in microphysiological nerve tissue following paclitaxel treatment. Paclitaxel-induced somal abnormalities included swollen neuron soma, asymmetric neuronal nuclei, and decreased cytoplasmic and nuclear staining intensity and nerve tissue abnormalities included decreased myelinated and unmyelinated nerve fibers densities, decreased myelin thickness, and an increased presence of incomplete myelin structures. Although these observations are consistent with previous descriptions in nonclinical literature, we believe the extensive neurodegeneration and lack of electrophysiological active are more consistent with late-stage neurodegeneration than early-onset CIPN.

Finally, the data here demonstrate a functional difference between primary demyelination and myelin abnormalities that develop secondary to primary axonal degeneration. With sufficient exposure, vincristine (Balayssac et al., 2005; Chine et al., 2019; Ja’afer et al., 2006), paclitaxel (Andoh et al., 2017; Chen et al., 2015; Tasnim et al., 2016; Wang et al., 2003), and cisplatin (Yoon et al., 2009) all induce axon degeneration and reduce myelinated fiber density in vivo. However, it is not always clear if this reflects primary or secondary myelinopathy. Forskolin is known to induce demyelination in vitro (Zhu and Glaser, 2008) and was included here to directly compare primary demyelination to myelin-related abnormalities that occur secondary to neurotoxicity. We confirmed that forskolin treatment disrupted myelin sheath morphology as a primary effect, resulting in an increased G-ratio and increased presence of incompletely myelinated axons in the absence of identifiable effects on neuroaxonal morphology. We also observed that forskolin treatment produced a distinct length- and stimulus intensity-dependent electrophysiological signature. Unlike the chemotherapeutics, forskolin did not yield a progressive inhibition of CAP generation with increasing concentration. We only observed moderate signs nerve conduction impairment in forskolin-treated constructs after low volt simulation at the distal stimulation site. We hypothesize that disruption of myelin integrity in the absence of axonal damage resulted in decreased sensitivity of the most distal neurite growth and a reduction in the ability of small, distally generated CAPs to be propagated the full length of the nerve construct for detection in the head region. Further investigation is necessary to confirm this hypothesis.

The present description of microphysiological CIPN has its limitations. These cultures lack a motor nerve population and are therefore unable to assess chemotherapy-induced motor neuropathy, which may be addressed by microphysiological evaluation of human stem cell-derived motor neurons (Sharma et al., 2019). Additionally, clinical NCS separately evaluate reductions in amplitude and velocity nerve impulses which was not distinguished in this work. This limitation can be addressed by incorporation of microelectrode arrays to track nerve impulses across multiple electrodes and generate more accurate estimations of nerve conduction velocity. Histopathological analysis of additional concentrations and more distal nerve sections would provide a wealth of information to help define patterns that could discriminate early and late-stage CIPN. Such assessments are not presently feasible given the time and cost associated with generating and analyzing these images. Our current results were also limited by an inability to sufficiently evaluate mitochondrial structure. In vivo treatment with vincristine (Chine et al., 2019), paclitaxel (Chen et al., 2015; Flatters and Bennett, 2006; Xiao et al., 2011), and cisplatin (Yoon et al., 2009) have all been shown to induce distort mitochondrial structure, but we did not identify any significant mitochondrial abnormalities in our in vitro constructs. The optical clarity of the cells and nerve fibers in this microphysiological system conceivably enables high-throughput functional evaluation of both axoplasmic transport and mitochondrial health repeatedly and in real time using in vivo microscopy. Therefore, this model may more directly assess the onset and progression of chemical effects than is feasible through conventional histopathological analysis.

Despite these limitations, in vitro microphysiological platforms have unique advantages over conventional model systems for high-throughput screening of novel chemical entities and neurological disease modeling (Pollard et al., 2019). The results presented here demonstrate that our fully in vitro, bioengineered, microphysiological nerve system recapitulates clinical descriptions of length-dependence associated with CIPN induced by 3 key chemotherapy agents with divergent mechanisms of action: vincristine, paclitaxel, and cisplatin. Histopathological analysis of our organized 3D nerve constructs confirmed the presence of previously described anatomic and functional hallmarks of each distinct CIPN scenario that normally requires confirmation through in vivo experimentation. This represents an innovative in vitro means for both exploring basic ganglion and nerve biology as well as reliably, quickly and cost-effectively obtaining information to distinguish neurotoxic findings in CIPN including axonopathy, myelinopathy, neuronopathy, and neuro/glial cytotoxicity.

FUNDING

United States National Institutes of Health/National Center for Advancing Translational Sciences (NIH/NCATS) (Grant No. R42-TR001270).

DECLARATION OF CONFLICTING INTERESTS

K.J.P. and B.B. report no conflicts of interest. M.J.M. is an inventor on a patent for “nerve-on-a-chip” technology and is a co-founder, officer, and equity stakeholder in a startup company that has commercialized this technology.