Three-Dimensional Mapping of Facial Motor Neurons with Tissue Clearing for Facial Nerve Nucleus of Guinea Pig via Long Distance Retrograde Tracing

Abstract

Facial nerve palsy profoundly affects facial expression and quality of life, with recovery frequently hindered by aberrant nerve regeneration leading to synkinesis. A comprehensive understanding of three-dimensional (3D) cytoarchitectural organization and adaptive changes within the facial motor nucleus remains constrained by conventional two-dimensional analyses. In this study, retrograde neural tracers (DiI, DiO, Dextran Alexa Fluor^TM 488/546) were integrated with tissue clearing techniques (CUBIC and LUCID) to reconstruct and analyze the 3D distribution of neurons innervating the orbicularis oris and orbicularis oculi muscles in guinea pigs. Cleared brainstem tissues were imaged using two-photon microscopy, revealing distinct region-specific neuronal clusters within the facial motor nucleus. In a facial nerve injury model, disrupted regional specificity and disorganized neuronal distribution were observed, suggesting misdirected central reinnervation. This novel 3D imaging method enables high-resolution spatial neuronal analysis and reorganization following nerve injury. This approach provides a valuable tool for elucidating facial nerve regeneration and synkinesis, and may facilitate the development of improved therapeutic strategies.

I. Introduction

Facial nerve palsy is a debilitating condition caused by diverse etiologies, including trauma or viral infection, markedly impacting a patient’s quality of life (QOL). Paralysis of the facial expression muscles impeded the ability to produce natural expressions or social smiles, and may cause food spillage from the corner of the mouth during eating, severely impairing daily functioning [1]. When complete recovery is not achieved, these deficits may persist lifelong. Aberrant nerve regeneration can lead to “synkinesis,” involuntary movements such as eye closure during oral movements or elevation of the corner of the mouth during eye closure [2, 3].

Extensive research has investigated nerve regeneration and functional recovery following disease onset. Progress in elucidating the mechanisms of nerve regeneration has identified key roles for Schwann cells, neurotrophic factors, and inflammatory mediators involved in this process [4, 5]. However, many aspects remain unresolved, particularly concerning the detailed molecular mechanisms and the safety and efficacy of regenerative therapies. Aberrant reinnervation (misdirection), occurring during regeneration and the structural reorganization within the central nervous system, remains poorly understood.

To address these gaps, visualization of neural pathways using retrograde neuronal tracers represents a powerful technique for evaluating the connectivity between peripheral and central components of nerves. Various retrograde tracers have been employed to analyze facial motor nuclei. Fernandez et al. used horseradish peroxidase (HRP), Choi et al. used Fast Blue and Fluoro Ruby, and Mendez et al. used Fluoro Gold and Fluoro Ruby to assess facial motor nuclei in rats [6–8]. Yamada et al. observed neuronal distribution within facial motor nuclei of guinea pigs using True Blue and DiI [9].

Although previous studies provided valuable insights, most assessments have relied primarily on two-dimensional (2D) sections, hindering the ability to appreciate the 3D organization of the entire facial motor nucleus. Conventional 2D analyses permit observation of individual sections, but are inherently limited by loss of information at cut surfaces and the inability to comprehensively depict the overall spatial configuration and neuronal architecture as a unified 3D structure.

Technologies that overcome these limitations, such as tissue clearing and 3D imaging, have advanced rapidly. Unlike conventional observation methods that require sectioning, tissue clearing enables the 3D visualization of internal structures by rendering the tissue transparent in its intact state. Especially in the field of neuroscience, these techniques are widely applied to the 3D analysis of neural circuits throughout the brain and can be used to evaluate the morphological changes of entire neural nuclei as well as the distances between individual cells within the nuclei [10, 11]. However, its application to the facial nerve and its central nucleus remains limited, and detailed reports on its 3D structure have not been published.

This study aimed to integrate retrograde neural tracers with tissue clearing technology to visualize and analyze, at high resolution, the 3D distribution of neurons innervating the orbicularis oris and orbicularis oculi muscles within the guinea pig facial motor nucleus. Through this approach, we sought to establish a novel method for comprehensive 3D understanding of central changes following facial nerve palsy and to advance understanding of facial nerve regeneration mechanisms.

II. Materials and Methods

Animals

Female Hartley guinea pigs (Japan SLC, Shizuoka, Japan) aged 8–10 weeks and weighing 400–500 g were used in this study. Animals were housed under an artificial 12-hr light-dark cycle and were allowed free access to standard laboratory chow and water before and after the experiments. All experimental procedures were conducted in accordance with the guidelines of the Animal Experiment Committee of Ehime University School of Medicine and were approved (No. 05HI78-4).

Surgical procedures

Retrograde neural tracers used in this study included the following:

• DiI (1,1'-Dioctadecyl-3,3,3',3'-Tetramethylindocarbocyanine Perchlorate; DiIC18(3), D282, Thermo Fisher Scientific, Japan)

• DiO (3,3'-Dioctadecyloxacarbocyanine Perchlorate; DiOC18(3), D275, Thermo Fisher Scientific, Japan)

• Dextran, Alexa Fluor^TM 488 (10,000 MW, Anionic, Fixable; D22910, Thermo Fisher Scientific, Japan).

• Dextran, Alexa Fluor^TM 546 (10,000 MW, Anionic, Fixable; D22911, Thermo Fisher Scientific, Japan).

DiI and DiO were dissolved in dimethyl sulfoxide (FUJIFILM, Japan) at a concentration of 1 mg/ml, while Dextran Alexa Fluor^TM 488 and 546 were dissolved in PBS at the same concentration prior to use.

For tissue clearing, CUBIC (TCI, Japan) and LUCID (Photon Tech Innovations, Japan) were employed. Based on the combinations of retrograde tracers and clearing reagents, animals were categorized into four groups (Supplementary Table S1):

• Group A: DiI and DiO with LUCID clearing (n = 3).

• Group B: Dextran Alexa Fluor^TM 488 and 546 with LUCID clearing (n = 3)

• Group C, DiI and DiO with CUBIC clearing (n = 3).

• Group D: Dextran Alexa Fluor^TM 488 and 546 with CUBIC clearing (n = 3)

All surgical procedures were performed under general anesthesia using intramuscular injections of ketamine hydrochloride (35 mg/kg) and xylazine hydrochloride (7 mg/kg). Retrograde tracers were injected using a 1 ml syringe with a 29G needle (Terumo Myjector, SS-10M2913A, Terumo, Japan). For each target muscle, injections were performed at two distinct sites.

In Groups A and C, DiO (1 mg/ml) was injected into the orbicularis oculi muscle, and DiI (1 mg/ml) into the orbicularis oris muscle. For the orbicularis oculi muscle, 400 μl of tracer was administered in total: 100 μl at two sites in the upper eyelid portion and 100 μl at two sites in the lower eyelid portion. Similarly, the orbicularis oris muscle received a total of 400 μl of tracer: 100 μl at two sites in the upper lip portion and 100 μl at two sites in the lower lip portion.

In Groups B and D, the same injection procedure and volumes were used, with Dextran Alexa Fluor^TM 488 (1 mg/ml) injected into the orbicularis oculi muscle (total 400 μl) and Dextran Alexa Fluor^TM 546 (1 mg/ml) injected into the orbicularis oris muscle (total 400 μl).

Tissue preparations

Animals were sacrificed two weeks after tracer injections. Under deep anesthesia induced by intramuscular injections of ketamine hydrochloride (35 mg/kg) and xylazine hydrochloride (7 mg/kg), animals were perfused transcardially with phosphate-buffered saline (PBS) to remove blood, and then with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). The brainstem tissue, including the facial nerve nucleus, was extracted and further fixed overnight (4°C) in the same solution. The facial nerve nucleus has been reported to be approximately 2 mm thick [12]; however, in the present preliminary experiment, the observation limit for samples was 1 mm thick, so the brainstem tissue was divided into two parts. For groups A and B, each specimen was immersed in LUCID and stored in a light-shielded container for two weeks. The CUBIC protocol was followed in groups C and D. Specifically, samples were washed three times with PBS at two-hr intervals, immersed in 50% CUBIC-L for 24 hr, washed again with PBS three times at two-hr intervals, immersed in 50% CUBIC-R+ for 24 hr, and finally left in CUBIC-R+ for one week.

Image acquisition

Cleared tissue sections were imaged using light microscopy and two-photon microscopy (TPM) (AX R MP and A1 R MP, Nikon, Japan, with Alcor 920-2, SPARK LASERS, France) with a ×20 APO LWD 20XC NA1.0 lens or a ×25 APO LWD 25XC NA1.10 water-immersion lens. The excitation laser for TPM imaging is set to 920 nm (ALCOR). Second-harmonic generation (SHG) signals from collagen fibers were captured by splitting the signals with dichroic mirrors at 488 nm and a short-pass filter at 460 nm for 920 nm excitation. Fluorescence signals from Alexa488 or DiI and Alexa546 or DiO were detected at 500–550 nm and 563–593 nm, using a GaAsP-type photomultiplier module. Image stacks were acquired from the surface through the entire depth of tissue slices.

Image analysis

Images were processed using NIS-Elements version 5.21 software (Nikon, Japan). To enhance image clarity, all images were applied median 3 × 3 filtering and rolling ball background subtraction (radius 37.98 μm). All the samples were evaluated along the vertical axis, covering the entire depth of the sections or tissue blocks. Maximum intensity projection images were saved as a single file in tagged image format (TIF, 8-bit).

Facial paralysis model

Animal models of facial nerve palsy were created according to the previously described protocol from our laboratory [13, 14]. General anesthesia was induced with intramuscular injections of ketamine hydrochloride (35 mg/kg) and xylazine hydrochloride (7 mg/kg). A postauricular skin incision was made on the left side to expose the temporal bone cavity, allowing visualization of the facial nerve canal. Using a freezing spray (Oken, Fukuoka, Japan), the distance between the tip of the nozzle and the facial nerve canal was set to 5 mm, and the left vertical portion of the facial nerve canal was cooled for 5 sec. Tracer injections were performed 15 weeks after facial nerve injury, and the animals were sacrificed 2 weeks later for tissue extraction (n=3).

III. Results

Injection of neural tracer

In preliminary experiments, neurons within the facial motor nucleus were not stained when tracers were injected from the external skin surface. This was particularly evident in cells innervating the orbicularis oculi muscle. Upon detaching the orbicularis oculi muscle from the skin, the palpebral conjunctival approach provided a closer access route to the muscle. When the injection sites were switched from the palpebral conjunctival side and the gingival side, tracer uptake improved markedly, resulting in consistent and reproducible neuronal labeling (Fig. 1).

Fig. 1.

Injection technique of retrograde neural tracers. Schematic representation of tracer injection sites from the palpebral conjunctival side (a, b) and from the gingival side (c, d). Dashed lines indicate the regions of the orbicularis oculi muscle (b) and orbicularis oris muscle (d).

Visualization of the facial motor nucleus

Initially, we attempted to visualize neurons in the facial motor nucleus using DiI/DiO as a retrograde neural tracer with the LUCID [15, 16] clearing method for divided fixed spinal cord slices. However, no neuronal labeling was observed in any samples (Fig. 2). This is attributable to the lipid-soluble nature of DiI/DiO, which were washed out because of the strong amphipathic effects of LUCID. To address this limitation, we employed water-soluble Dextran, Alexa Fluor^TM 488/546 for fixable. Using TPM, clear fluorescent signals from neurons labeled with long-distance tracers were successfully obtained from all three animals. The labeled neurons exhibited a region-specific distribution in thick slices of the cleared spinal cord. Moreover, labeled neurons could be visualized in arbitrary sections, and 3D reconstruction enabled comprehensive visualization of the spatial distribution within the facial motor nucleus (Fig. 3). This method proved effective for simultaneous multi-muscle labeling within a single animal. Furthermore, 3D analysis revealed that in both the rostral (Fig. 3a) and caudal (Fig. 3f) sections, neurons innervating the orbicularis oculi muscle were observed not only in their original lateral region but also in the medial region of the facial motor nucleus.

Fig. 2.

Labeling of the facial nerve nucleus (Clearing reagent: LUCID; Neural tracer: DiI/DiO). Representative images of brainstem sections after clearing with LUCID and labeling with DiI/DiO. No fluorescent signals were detected. (a) and (b) show the facial nerve nucleus region; however, no fluorescent signals were detected in either DiI (red) or DiO (green) channels.

Fig. 3.

Labeling of the facial nerve nucleus (Clearing reagent: LUCID; Neural tracer: Dextran Alexa Fluor^TM 488/546). Representative brainstem sections were divided into two tissue blocks for observation. Panels (a–f) show labeled sections obtained with Dextran Alexa Fluor^TM 488 (green) and 546 (red).

We subsequently applied this technique to a facial nerve injury model to further evaluate its utility and robustness [13, 14]. In a preliminary experiment, we injected retrograde tracers two weeks after nerve injury, and two weeks later we extracted the facial nerve nucleus to examine the presence of fluorescent neuronal labeling. In this condition, no fluorescent neuronal labeling was detected in the facial nerve nucleus (data not shown). Conversely, in the model 15 weeks after nerve injury, labeled neurons were clearly detected, although their distribution was irregular and the normal topographic organization was disrupted (Fig. 4). These results indicate that even after nerve injury and subsequent regeneration, retrograde tracers can label neurons. Additionally, combinations of DiI/DiO with CUBIC (Fig. 5) and Dextran Alexa Fluor with CUBIC (Fig. 6) were examined, given that CUBIC is a hydrophilic clearing reagent. Under both conditions, neuronal labeling was observed in two of three animals, demonstrating that tissue clearing with CUBIC also allowed for successful tracer visualization, albeit with slightly lower consistency. Representative images of all tracer–clearing reagent combinations are summarized in Figure 7.

Fig. 4.

Labeling of the facial nerve nucleus in a facial paralysis model (Clearing reagent: LUCID; Neural tracer: Dextran Alexa Fluor^TM 488/546). Representative brainstem sections of the facial paralysis model. Panels (a–f) show labeled sections obtained with Dextran Alexa Fluor^TM 488 (green) and 546 (red).

Fig. 5.

Labeling of the facial nerve nucleus (Clearing reagent: CUBIC; Neural tracer: DiI/DiO). Representative brainstem sections were divided into two tissue blocks for observation. Panels (a–f) show labeled sections obtained with DiI (green) and DiO (red).

Fig. 6.

Labeling of the facial nerve nucleus (Clearing reagent: CUBIC; Neural tracer: Dextran Alexa Fluor^TM 488/546). Representative brainstem sections. Panels (a–d) show labeled sections obtained with Dextran Alexa Fluor^TM 488 (green) and 546 (red).

Fig. 7.

Combinations of clearing reagents and retrograde neural tracers. Comparison of tracer labeling patterns obtained with different clearing reagents (LUCID, CUBIC) and tracers (DiI/DiO, Dextran Alexa Fluor^TM 488/546). Representative fluorescence images are shown. Bars = 100 μm (a–d).

IV. Discussion

In this study, we established a method for 3D observation and analysis of the facial nerve nucleus by integrating retrograde neuronal tracers with tissue-clearing techniques. By injecting different neuronal tracers into the orbicularis oculi and orbicularis oris muscles and clearing the brainstem tissue, we successfully visualized the 3D organization of neuronal populations innervating each muscle within the facial motor nucleus. Furthermore, in a facial nerve injury model, the normal topographic arrangement of these neuronal populations was disrupted, and regional specificity was lost. This likely reflects processes such as nerve regeneration and misdirection that occur after neural injury, and represents an important finding for understanding central nervous system changes in facial nerve palsy.

In this study, both lipid-soluble dyes DiI and DiO, and water-soluble dye Dextran, Alexa Fluor^TM 488/546, were employed. DiI and DiO are incorporated into the lipid bilayer of the neuronal cell membranes and selectively stain the membrane without entering the cytoplasm or nucleus [17]. Contrarily, Dextran, Alexa Fluor ^TM488/546 are taken up into the cell via membrane permeability and accumulate in the endosomes and cytoplasm [18]. Additionally, by using fixable versions of these tracers, fluorescent signals are more likely to be retained even after the fixation and clearing processes.

Initially, the lipophilic tracers DiI/DiO were combined with LUCID (Group A); however, no fluorescent signals were detected, suggesting that the lipid-targeted fluorophores may have been lost because of the lipid extraction and/or amphipathic properties of LUCID. Subsequently, the hydrophilic tracers Dextran, Alexa Fluor^TM 488/546, were combined with LUCID (Group B), resulting in clear labeling of facial nerve neurons in all cases. The labeled neurons exhibited region-specific distribution, and multiple injection sites could be targeted within the same animal. Moreover, 3D reconstruction enabled detailed stereoscopic visualization of neuronal distribution.

When this method was applied to the facial nerve injury model, the distribution of labeled neurons was irregular and did not preserve the regional organization observed under normal conditions. In a preliminary experiment, we examined a model in which retrograde tracers were injected two weeks after nerve injury. The facial motor nucleus was then extracted two weeks later, and no fluorescent labeling was detected. This suggests that surviving fibers were insufficient for retrograde tracer transport at this early stage and that substantial re-innervation had not yet progressed. Therefore, the labeled neurons observed in the regenerated model are unlikely to result from residual intact innervation and instead likely reflect re-innervation processes accompanying nerve regeneration, including the possibility of aberrant synaptic connections. These results indicate that this approach is useful for visualizing and assessing nerve regeneration after injury. Additionally, combinations using CUBIC (Groups C and D) were tested, yielding successful labeling in two of the three cases for each group. Collectively, these findings underscore that the choice of an appropriate tissue-clearing reagent in accordance with the physicochemical properties of the tracer is critical for successful 3D evaluation.

Previous 2D analyses have classified the facial motor nucleus into six subnuclear structures: medial, medial-intermediate, dorsolateral, lateral, intermediate, and ventromedial [19, 20]. These structures are intricately intertwined, and their anatomical boundaries are often indistinct. With respect to the spatial relationship between these subnuclei and the neurons innervating facial muscles, the orbicularis oculi is thought to be located mainly in the dorsolateral region, and the orbicularis oris in the intermediate region. Our study demonstrated a comparable distribution, although we observed orbicularis oculi neurons in the medial region of the rostral and caudal parts of the tissue (Fig. 3a, f). This finding suggests that their distribution was closer to a spherical shape than tubular, with the lateral region forming a 3D structure that enveloped the medial region. This insight became clear only through 3D observation. Contrarily, the 3D observations in our study not only allowed us to avoid double-counting of cells, which occurs in conventional thin-section analysis, but also preserved information about the tissue plane, permitting higher-precision results.

On the other hand, in the facial nerve palsy model, the regional specificity observed under normal conditions was lost, and the labeled neurons were irregularly distributed within the facial nerve nucleus. This distribution pattern seems to be related to the misdirected reinnervation underlying the pathological synkinesis observed after facial nerve palsy. During nerve regeneration, the axons of motor neurons that originally innervate different muscles mistakenly reinnervate other muscles, resulting in pathological synkinesis [21, 22]. Our findings suggest that such misdirection may also originate at the central level. Previous reports have described such abnormal cells as “misplaced cells” and their numbers have been measured two-dimensionally [3–6]. Moving forward, the accumulation of additional data from nerve injury models will expand this analysis beyond simple neuronal counts to include morphometric and spatial parameters, such as distribution bias, density, and intercellular distances.

Our study has several limitations. First, the reproducibility of intramuscular injections is a challenge in the injection of retrograde neuronal tracers. This is because there are no available methods to confirm whether the dye was injected into all the projecting nerves, potentially leading to variability in the number of stained neurons observed in the brainstem. In our experiments, injections into the eyelids and lips were stabilized by injections from the palpebral conjunctival and gingival sides, respectively, which shortened the distance from the injection site to the muscle and contributed to the stability of the procedure. However, for more reliable staining, direct exposure and dye impregnation of the nerve under direct vision may be an alternative approach for future studies. Second, the two-photon laser microscope used for 3D observation is an expensive and technically demanding instrument that limits the number of facilities and skilled technicians capable of performing such experiments.

Nevertheless, the methodology established in this study offers a novel approach for the 3D observation of reconstructive processes in the central nervous system after facial nerve injury. With further analyses using this method, it may be possible to better understand the actual phenomenon of misdirection during the regeneration process and to evaluate neural reconstruction after therapeutic intervention.

V. Conclusion

This study established a method for clearly visualizing and analyzing the 3D distribution of neurons innervating the orbicularis oris and orbicularis oculi muscles within the guinea pig facial motor nucleus by applying retrograde neural tracers and tissue-clearing techniques. The usefulness of this approach is demonstrated. Future research based on this method is expected to contribute to the development of novel therapeutic strategies by elucidating the mechanisms of nerve regeneration after facial nerve injury.

VI. Conflicts of Interest

The authors declare no conflicts of interest.

VII. Acknowledgments

This work was supported by JSPS KAKENHI Grant Numbers 25H00848, 24K02567, 25K10284, 22H04926, 20K09714, JP16H06280, and a Grant-in-Aid for Scientific Research on Innovative Areal Platforms for Advanced Technologies and Research Resources from the “Advanced Bioimaging Support” program. This work was supported by JST A-STEP, Grant Number JPMJTR234G, Japan. TI was also supported by the Naito Foundation, the Uehara Memorial Foundation, and the Research Grant 399 of the Princess Takamatsu Cancer Research Fund (23-255005).

Supplementary Material

Supplementary Table S1