Label-free Visualization and Morphological Profiling of Neuronal Differentiation and Axonal Degeneration through Quantitative Phase Imaging

Abstract

Understanding the intricate processes of neuronal growth, degeneration, and neurotoxicity is paramount for unraveling nervous system function and holds significant promise in improving patient outcomes, especially in the context of chemotherapy-induced peripheral neuropathy (CIPN). These processes are influenced by a broad range of entwined events facilitated by chemical, electrical, and mechanical signals. The progress of each process is inherently linked to phenotypic changes in cells. Currently, the primary means of demonstrating morphological changes rely on measurements of neurite outgrowth and axon length. However, conventional techniques for monitoring these processes often require extensive preparation to enable manual or semi-automated measurements. Here, we employ a label-free and non-invasive approach for monitoring neuronal differentiation and degeneration using quantitative phase imaging (QPI). Operating on unlabeled specimens and offering little to no phototoxicity and photobleaching, QPI delivers quantitative maps of optical path length delays that provide an objective measure of cellular morphology and dynamics. Our approach enables the visualization and quantification of axon length and other physical properties of dorsal root ganglion neuronal cells, allowing greater understanding of neuronal responses to stimuli simulating CIPN conditions. Our research paves new avenues for the development of more effective strategies in the clinical management of neurotoxicity.

Graphical Abstract

Quantitative phase imaging (QPI) is employed to investigate the intricate dynamics of label-free dorsal root ganglion (DRG) neuronal cells during neuronal differentiation and degeneration. We could monitor and quantify morphological changes, including axonal density, volume, dry mass, and length, offering a comprehensive interrogation of neuronal responses to growth and neurotoxic stimuli.

Introduction

Dorsal root ganglia (DRGs), clusters of sensory neurons nestled within the spine ^[1], serve as pivotal conduits for transmitting sensory information from the body’s periphery to the central nervous system via the spinal cord ^[2]. DRG neurons are instrumental in our ability to perceive and react to the external environment, playing a multifaceted role in sensory functions, including touch, temperature, and pain perception ^[3]. Any aberrations in DRG function can culminate in a spectrum of sensory disorders, encompassing chronic pain syndromes, neuropathies, and sensory loss ^[4–5]. Notably, a prevalent culprit behind DRG dysfunction is the administration of neurotoxic drugs, particularly those employed in cancer chemotherapy, giving rise to the collective entity known as chemotherapy-induced peripheral neuropathy (CIPN) ^[6].

Various methods are used to probe the functionality of DRG neurons and decipher the intricate pathways underpinning CIPN. These approaches encompass investigations into neuronal morphology, electrical responses, and the expression of molecular markers ^[7–8]. Compromised sensory functions have been notably associated with morphological changes, including distal fiber loss, demyelination, and axonal degeneration ^[9]. In this regard, neurite length measurement stands out as one of the most commonly used assays for in vitro axon degeneration investigations, due to its simplicity ^[10]. The process involves staining cell cultures with a neuronal marker (such as β III tubulin or neurofilament) to visualize neurites, followed by imaging with a (confocal) fluorescence microscope, and subsequently measuring either the total or longest axon length using image analysis tools ^[11–13]. Despite its widespread adoption, this technique still necessitates considerable sample preparation to enable image-based analysis.

Quantitative phase imaging (QPI) has emerged as a powerful technique for cellular imaging, as it offers a label-free and non-destructive approach that provides high-contrast and spatial resolution images ^[14–15]. Rooted in the interferometric principle, QPI extracts phase information, decoupled from the intensity, by overlaying an image field with a reference field, enabling the imaging of transparent specimens like biological cells with remarkable contrast and sensitivity, all while avoiding the use of exogenous labels ^[16–17]. Moreover, QPI unlocks a wealth of quantitative data on cellular properties, including cell volume, dry mass, and density ^[18–19], making it increasingly valuable for various biological applications. Driven by these salient features, QPI has been applied to monitor both physiological and pathological changes in diverse cell types, from bacteria ^[20] to red blood cells ^[21–24] and adipocytes ^[25], as well as to investigate cell dynamics ^[26–27], and inter-cellular, intra-cellular, and cell-substrate interactions ^[28–30]. In recent years, QPI has made strides in elucidating dynamic responses of neurons, uncovering subtle changes in phase shifts during neural activity ^[31], membrane fluctuations ^[32], processes of maturation and formation of neurites ^[33], and even neuron deformation during neuronal spikes ^[34]. However, a noticeable gap in the existing literature lies in the limited application of QPI to investigate morphological changes of neurons in clinically relevant scenarios, such as those induced by chemotherapeutics or to understand the degree of protection afforded by neuroprotective agents. Notably, there exists a distinct paucity of literature when it comes to the application of QPI in studying dorsal root ganglion (DRG) neurons. Studying DRG neuron morphology in a non-perturbed condition is particularly crucial, owing to the unique characteristics and role of DRG neurons in sensory transmission, coupled with their susceptibility to neurotoxicity.

In this study, we report morphological features of DRG neuronal cells and physical information revealed by optical interferometry-based QPI during differentiation and neurotoxic treatment, without the use of exogenous labels. Our experiments were conducted using the 50B11 neuronal DRG cell line, originally derived from embryonic rat DRGs ^[35]. These immortalized 50B11 cells can be induced to differentiate into a neuronal phenotype through the addition of forskolin, exhibiting sensory neuron features across biochemical, electrophysiological, and morphological dimensions. Our approach enables quantitative evaluation of axon degeneration, cell dry mass, and axon density, providing additional physical metrics for tracking neuronal responses to clinically relevant external stimuli. Our study delves into the biological implications of these findings, shedding light on how neuronal cells respond to external cues and potentially paving the way for more effective strategies in managing neurotoxicity in the context of various neurological disorders.

Results and Discussion

Experiments were conducted as shown in Figure 1. Briefly, we prepared 50B11 DRG neuronal cells for QPI measurement in response to neuronal differentiation induction and exposure to paclitaxel, a commonly used chemotherapy drug. To elucidate the dynamics of forskolin-induced neuronal differentiation, we conducted a comprehensive examination at multiple timepoints (0, 3, 6, 12, 24, 36, 48, 60, and 72 hours after differentiation induction), enabling the monitoring and quantitative assessment of morphological alterations. Additionally, we investigated morphological changes in cells exposed to paclitaxel, both in the presence and absence of a neuroprotective agent (as depicted in Figure 1A), with a primary focus on evaluating these changes using the quantifiable phase information provided by QPI images of cells.

Figure 1.

Schematic workflow of the study. Dorsal root ganglion (DRG) neuronal cells from the 50B11 cell line were used as cellular models. Neuronal differentiation was induced by the introduction of 50 µM forskolin and monitored from 0 to 72 hours of forskolin induction. (A) Three different chemotherapy cultures, treated with paclitaxel (PTX) and/or ethoxyquin (EQ) were prepared: (A1) control without chemo treatment, (A2) cells treated with both chemotherapy drug, PTX, and neuroprotective compound, EQ, and (A3) cells treated only with chemotherapy drug, PTX. Drug treatments were introduced at 6 hours after initiating neuronal differentiation through the addition of forskolin. Cells were imaged at 24 hours after the initiation of neuronal differentiation (i.e., 18 hours of drug treatment). (B) Quantitative phase imaging (QPI) was used to record quantitative morphological data. (C) Schematic representation of a Mach-Zehnder interferometer-based optical diffraction tomography system used for the measurements. (D) Measurement of the phase delay induced by specimen, DRG neuronal cells, compared with the reference, background medium, and (E) the reconstructed refractive index (RI) map based on the measured phase change.

QPI used in this study is illustrated in Figure 1B. Briefly, the system utilized a Mach-Zehnder interferometry-based off-axis optical diffraction tomography (Figure 1C). As the light source, a coherent laser beam is emitted from a diode-pumped solid-state laser with a wavelength of 532 nm, which is split into two by a 2×2 single-mode fiber coupler where one beam is directed onto the sample with its incident angle regulated by a digital micromirror device (DMD) and the other serves as a reference beam (Figure 1D). The sample beam is then projected onto a charge-coupled device (CCD) camera, where it interferes with the reference beam through a beamsplitter (BS), thus enabling quantification of phase delay induced by the sample (Figure 1E). Further details of the QPI system are described in SI.

Neuronal networks formation is an intricate process where individual neurons respond to various internal as well as external stimuli. Neurons undergo a gradual transformation as they self-organize into groups, ultimately giving rise to functional units with specific roles. Within this complexity, we sought to leverage QPI to explore the intricacies of differentiating DRG neurons at various temporal and spatial scales. This emerging optical approach is distinguished by its high throughput and minimal invasiveness, affording us a fresh perspective into the underlying dynamics. Additionally, we harnessed QPI’s unique capability to quantify the non-aqueous content within cells, commonly referred to as dry mass, providing a distinctive opportunity to dissect the growth of cell bodies, and extensions of neurites. In this study, we initially focused on observing and quantifying the morphological changes in DRG neurons during neuronal differentiation, followed by those induced by chemotherapy.

Neuronal Differentiation Leading to Neurite Outgrowth Coincides with Changes in Cell Body Density

Our investigation commenced with the observation of both undifferentiated and differentiated 50B11 DRG neuronal cells using a QPI microscope, as shown in Figure 2. Neuronal differentiation was initiated by the addition of forskolin, and cells were tracked at multiple timepoints, including 0, 3, 6, 12, 24, 36, 48, 60, and 72 hours post-differentiation induction, allowing us to longitudinally monitor changes in cell morphology (Figure 2A). Notably, neurite outgrowth was observed as early as 3 hours post-differentiation induction, signifying the initiation of this crucial developmental process. This early neurite outgrowth aligns with known neurobiology ^[35–36], where neuronal differentiation triggers the extension of cellular processes, marking the initiation of neural network formation. Neurites continued to extend over time, with observable axonal elongation becoming prominent at the 24-hour mark. In parallel with neurite and axon growth, we noted a compaction of cell bodies, signified by an increase in refractive index (RI) values as the differentiation process advanced.

Figure 2.

Forskolin-induced neuronal differentiation of 50B11 DRG cells. (A) Three different cultures, undifferentiated control (top) and cells undergoing differentiation (middle, bottom) were monitored up to 72 hours. Differentiation induction is mediated by the addition of forskolin, with fresh replacements at every 24 hours to maintain differentiation (bottom), compared to no replacements (middle). Cells were imaged at 0, 3, 6, 12, 24, 36, 48, 60, and 72 hours of differentiation induction. Each image represents max intensity projection (MIP) of 3D RI map. Neurite outgrowth is observed as early as 3 hours post-differentiation. Scale bar: 10 µm. Changes in (B) cell volume, (C) cell dry mass, and (D) cell body density during neuronal differentiation. (***, P ≤ 0.001; **, P ≤ 0.01; *, P ≤ 0.05)

Importantly, our study extended the observation period up to 72 hours post-differentiation, in contrast to many previous studies that typically incubated cells with forskolin for 20–36 hours only followed by requisite treatments to stabilize the neuronal phenotype ^{[35, 37–38]}. We maintained the neuronal features of cells by replenishing the differentiation induction medium with fresh forskolin-containing medium every 24 hours (Figure 2A, Forskolin++). By doing so, we ensured the continued development of neurites and axon elongation, along with the sustained elevation of cell body morphology, which was notably prominent at the 24-hour mark post-differentiation. Conversely, cells induced to differentiate without medium replacement (Figure 2A, Forskolin+) reverted to their original shape, characterized by flattened cell bodies and the disappearance of neurites. Undifferentiated controls (Figure 2A, Forskolin−) exhibited consistent cell morphology throughout the monitoring period, devoid of any observable changes.

Quantitative Morphological Analysis of Differentiated Neuronal Cells

To further substantiate our morphological observations of differentiated neuronal cells (forskolin++), we conducted a quantitative analysis based on the acquired RI maps. Over time, we observed a decrease in cell volume (Figure 2B) concomitant with an increase in cell dry mass (Figure 2C), culminating in an overall rise in cell body density (Figure 2D) throughout the differentiation process. A noticeable drop in cell volume was observed after 24 hours, while there was no statistically significant difference between later timepoints. The range of observed cell volumes was notably lower at 72 hours than at the other timepoints. As for cell dry mass, there was no difference in mass between 0 and 24 hours. However, although the median mass remained consistent across these timepoints, the interquartile range of observations increased. An uptick in median dry mass was observed between 24 and 48 hours, with a slightly reduced range of distribution, which further decreased at 72 hours. Based on this trend, one can reasonably infer that while neuronal differentiation is accompanied by changes in cell volume and dry mass, thus affecting density, further changes observed in timepoints beyond 24 hours may be attributable to an increase in the number of cells that display neuronal features due to differentiation induction. At 24 hours post-induction, most cells had undergone differentiation, as evident from elongated axons, but some cells still retained undifferentiated morphology. This observation aligns with a previous study, which reported that some cells continued to proliferate without reaching terminal differentiation following forskolin induction at 50 µM or less ^[39]. However, the number of such cells was evidently reduced at later timepoints under the conditions of refreshed differentiation medium, underscoring that additional cells were induced to differentiate when provided with additional time for the process.

Dynamics of DRG Neuronal Morphology in Response to Paclitaxel-induced Neurotoxicity and Neuroprotection

Following the examination of neuronal features at various time points in rat dorsal root ganglion (DRG) cultures, our focus shifted towards the quantitative assessment of axonal morphology in response to paclitaxel (PTX)-induced neurotoxicity. PTX, a widely used chemotherapeutic agent, was employed to provoke neurotoxicity with the goal of mimicking the conditions of chemotherapy-induced peripheral neuropathy (CIPN). Briefly, PTX binds to β-tubulin within microtubule and stabilizes dynamic microtubule polymerization, thus exerting anti-cancer effects ^[40]. However, it adversely affects DRG neurons by impairing axonal transport, leading to CIPN ^[41]. For comparative analysis, we introduced ethoxyquin (EQ), a well-known neuroprotective agent recognized for its ability to safeguard axonal integrity under stress, into a parallel culture subjected to PTX treatment. EQ offers neuroprotection by modulating the chaperone activity of heat shock protein 90 (Hsp90), thereby inhibiting the binding of two client proteins, ataxin-2 and splicing factor 3B subunit 2 (SF3B2), which are associated with axonal degeneration ^[42–43]. We explored the dynamics of PTX-induced axonal alterations in a dose-dependent manner, with PTX dosages varied from 0 to 500 nM. The resulting QPI images are summarized in Figure 3.

Figure 3.

Assessment and quantification of DRG cell response to chemotherapy and neuroprotective chemotherapy. In our study, chemotherapy is induced through the addition of paclitaxel (PTX) at various concentrations (0, 50, 100, 200, and 500 nM), and neuroprotection is achieved with the inclusion of 300 nM ethoxyquin (EQ). Scale bar: 100 µm. Changes in axon morphology in response to these treatments are observed in close-up panels: (A) No treatment, (B) EQ treatment only, (C) PTX treatment only, and (D) Combined PTX and EQ treatment. Each image represents max intensity projection (MIP) of 3D RI map. Scale bar: 20 µm. QPI analysis quantifies changes in axon morphology concerning (E) longest axon length and (F) axon density (volume/dry mass of axons). n = 37–97 cells from 3 replicates.

Our observations unveiled a clear dose-dependent relationship between PTX exposure and axonal degeneration. Cells exposed to the highest PTX concentration exhibited the most severe degeneration, manifesting substantial reductions in axon length, even when EQ was introduced as a neuroprotective agent. In fact, under the conditions of 500 nM PTX exposure, nearly all cells displayed complete axon ablation. This observation agrees well with another study which demonstrated with primary DRG neurons isolated from a mouse model that underwent PTX treatment ^[44]. While EQ demonstrated some efficacy in preserving axons when combined with 500 nM PTX, many cells still exhibited minimal to no traces of axons. These axon-ablated cells were distinguishable not only by their distinct morphology but also by a unique range of refractive index (RI) values, as visually apparent in the PTX 500 nM + EQ 300 nM panel of Figure 3. Here, the two cells located in the top left corner, featuring relatively low RI values, bore the characteristics of undifferentiated cells. In contrast, cells with relatively higher RI values displayed rounded, compacted cell bodies, a hallmark of differentiated cells as previously observed (Figure 2).

At lower PTX dosages, axons were better preserved, yet we noted the emergence of axonal blebbing and swelling when compared to untreated controls. This phenomenon is well illustrated in the close-up images presented in Figure 3A–D. The untreated controls (Figure 3A–B) exhibited clearly defined and intact axon boundaries without any visible signs of damage. Conversely, PTX-treated cells (Figure 3C–D) displayed notable axonal blebbing and swelling, particularly pronounced in cells subjected to PTX without the protective presence of EQ. These alterations were characterized by the emergence of branches at the distal ends of axons, further emphasizing the profound impact of neurotoxic agents on axonal morphology.

Quantitative assessment of axonal changes induced by neurotoxicity and neuroprotection

We further analyzed QPI images, focusing specifically on axonal changes following PTX treatment, as detailed in Figure 3E–F. Our primary assessment involved the measurement of the longest axon length in individual neurons, a well-established method to identify axonal degeneration induced by chemotherapy drugs. The results are summarized in Figure 3E. The dose-dependent effects of PTX on the axon length presented difference in IC50 of 30.72 nM [95% confidence interval 9.218–73.23] and 22.57 nM [95% confidence interval 1.871–69.48] with and without the presence of EQ, respectively. In comparison to untreated controls subjected to normal differentiation, both with and without dimethyl sulfoxide (DMSO) serving as the vehicle medium, we observed statistically significant reductions in axon length across all treated cells. Notably, at the lowest PTX concentration of 50 nM, the difference in axon length between cells treated in the absence and presence of EQ did not yield statistical significance. However, as the PTX dosage escalated, the influence of neuroprotection became significant within the same PTX dosage groups. This observation highlighted the trend of better-preserved axon length in the presence of EQ, a phenomenon documented in previous studies ^{[42, 45–46]}.

Subsequently, we compared axonal density, calculated as the ratio of dry mass to the volume of segmented axons, under different conditions (Figure 3F). Axon mass density exhibited an opposing trend to axon length, with untreated controls demonstrating significantly smaller values than their treated counterparts. Cells with PTX treatment showed EC50 of 7.335 nM [95% confidence interval 0–69.44] and those treated with both PTX and EQ showed EC50 of 10.42 nM [95% confidence interval 0–60.22]. Within the treated groups, cellular axonal density displayed a broad distribution. Particularly, cells treated with PTX alone exhibited heightened axon mass density in comparison to cells receiving PTX and EQ. To further validate our observations, we conducted immunofluorescence analysis (Figure 4 and Figure S2). This analysis confirmed that the trends observed in immunofluorescent images closely mirrored those seen in the QPI images, thereby reinforcing the robustness of our findings regarding axon morphology.

Figure 4.

Validation of DRG cell response to chemotherapy and neuroprotective chemotherapy. Chemotherapy is mediated by the addition of paclitaxel (PTX) at various dosages (0, 50, 100, 200, and 500 nM) while neuroprotection is mediated by the addition of 300 nM ethoxyquin (EQ). Immunofluorescence assay using βIII tubulin for visualizing the microtubule cytoskeleton of axons confirms axonal outgrowth, aligned with corresponding bright field images. Top: Cells treated with 50, 100, 200, and 500 nM PTX. Bottom: Cells treated with varying dosages of PTX and 300 nM EQ. Scale bar: 20 µm.

In this study, we harnessed the power of QPI to characterize the morphological properties of dorsal root ganglion (DRG) neuronal cells as they underwent neuronal differentiation and degeneration induced by forskolin and chemotherapy, respectively. Our approach allowed for label-free quantification of key neuronal features, including changes in axonal density, volume, cell body dry mass, and axonal length, offering a comprehensive view of neuronal responses to growth and neurotoxic stimuli. Specifically, we observed an increase in cell dry mass during neuronal differentiation, coupled with neurite outgrowth and cell body rounding. Subsequently, our investigation into PTX treatment revealed dose-dependent alterations in axon length and density. The inclusion of the neuroprotective agent EQ yielded statistically significant differences in these metrics when used alongside PTX treatment, a finding consistent with our prior study that identified EQ’s effectiveness in preventing chemotherapy-induced peripheral neuropathy (CIPN)-associated axon degeneration, as validated by multiple targeted assays ^[47]. This suggests that our optical-based, label-free approach can shed light on the relationship between CIPN and effective dosages, potentially enhancing anti-cancer and other therapeutic strategies. Determining the appropriate dosage is crucial for preventing CIPN, particularly considering its impact on up to 60% of cancer patients, a number expected to rise as cancer survival rates increase ^[6]. Common CIPN symptoms, such as pain in the hands and feet, numbness, and dysesthesia, often necessitate dosage adjustments, highlighting the importance of finding better methods for assessing and optimizing therapeutic strategies ^[48].

Traditional neuroscientific investigations often rely on electrophysiological methods or fluorescent imaging to dissect neural development and activity. While electrophysiology delivers a high-fidelity readout of neural function, its spatial throughput limitations hinder in-depth exploration of intra- and intercellular interactions among individual neurons in a network. On the other hand, fluorescent imaging provides functional insights, including electrical activity, within both single neurons and neural networks. However, this approach necessitates meticulous sample preparation and poses challenges for extended imaging due to the issues of phototoxicity and photobleaching. QPI offers an alternate tool that enables us to scrutinize the intricate dynamics of developing neural networks across multiple temporal and spatial scales with high throughput and minimal invasiveness. Indeed, emergent QPI techniques promise further improvements in image resolution, wider field-of-view, and higher throughput ^[49–51].

Several indicators of neural growth and degeneration are currently used, with morphological demonstrations playing a central role in assessing neuron health ^{[12, 52–53]}. While these methods have successfully provided (semi-)quantification of neuron morphology, they require additional substrates and/or labeling, potentially disrupting neurons and their surrounding environment. As an alternative, some nonlinear optical approaches, including multi-photon, coherent anti-Stokes Raman scattering (CARS), and stimulated Raman scattering (SRS) microscopes, have demonstrated the capacity for high-resolution and specific neuronal imaging ^[54]. Yet, such non-linear optical microscopy methods are prone to phototoxicity due to the high-intensity laser light they employ and are limited by their often-restricted field of view, making them less suitable for long-term imaging of larger neural networks or dynamic processes. Additionally, these methods often require expensive and elaborate instrumentation, further restricting their broader application in routine research settings. In contrast, QPI effectively captures cell dynamics and processes with high spatial and temporal resolution while remaining a cost-effective and accessible tool for a wider range of research applications. When coupled with deep learning analysis, it allows for the selective visualization and assessment of intracellular organelles based on their physical properties from each QPI image, all without the constraints of limited available channels ^[55].

Conclusion

In conclusion, this study demonstrates the potential of quantitative phase imaging (QPI) as a valuable and accessible tool for investigating the intricate dynamics of developing neural networks, particularly in the context of dorsal root ganglion (DRG) neuronal cells. We employed QPI to monitor and quantify morphological changes during neuronal differentiation and degeneration induced by forskolin and paclitaxel (PTX), respectively. The label-free nature of QPI allowed for the quantitative assessment of various cellular properties, including axonal density, volume, dry mass, and length, providing a comprehensive view of neuronal responses to growth and neurotoxic stimuli. Additionally, we confirmed the efficacy of the neuroprotective agent ethoxyquin (EQ) in preventing PTX-induced axonal degeneration. By shedding light on the relationship between chemotherapy-induced peripheral neuropathy (CIPN) and effective dosages, this study not only offers insights into neural growth and network formation but also holds promise for optimizing anti-cancer and therapeutic strategies. Furthermore, QPI’s label-free, cost-effectiveness, and minimal invasiveness make it a compelling alternative to traditional fluorescence-based methods, enabling broader applications in neuroscience research and potentially addressing critical questions at various spatial and temporal scales.