Serial block-face scanning electron microscopy reveals neuronal-epithelial cell fusion in the mouse cornea

The cornea is the most highly innervated tissue in the body. It is generally accepted that corneal stromal nerves penetrate the epithelial basal lamina giving rise to intra-epithelial nerves. During the course of a study wherein we imaged corneal nerves in mice, we observed a novel neuronal-epithelial cell interaction whereby nerves approaching the epithelium in the cornea fused with basal epithelial cells, such that their plasma membranes were continuous and the neuronal axoplasm freely abutted the epithelial cytoplasm. In this study we sought to determine the frequency, distribution, and morphological profile of neuronal-epithelial cell fusion events within the cornea. Serial electron microscopy images were obtained from the anterior stroma in the paralimbus and central cornea of 8–10 week old C57BL/6J mice. We found evidence of a novel alternative behavior involving a neuronal-epithelial interaction whereby 42.8% of central corneal nerve bundles approaching the epithelium contain axons that fuse with basal epithelial cells. The average surface-to-volume ratio of a penetrating nerve was 3.32, while the average fusing nerve was smaller at 1.39 (p ≤ 0.0001). Despite this, both neuronal-epithelial cell interactions involve similarly sized discontinuities in the basal lamina. In order to verify the plasma membrane continuity between fused neurons and epithelial cells we used the lipophilic membrane tracer DiI. The majority of corneal nerves were labeled with DiI after application to the trigeminal ganglion and, consistent with our ultrastructural observations, fusion sites recognized as DiI-labeled basal epithelial cells were located at points of stromal nerve termination. These studies provide evidence that neuronal-epithelial cell fusion is a cell-cell interaction that occurs primarily in the central cornea, and fusing nerve bundles are morphologically distinct from penetrating nerve bundles. This is, to our knowledge, the first description of neuronal-epithelial cell fusion in the literature adding a new level of complexity to the current understanding of corneal innervation.


Introduction
The cornea is the most highly innervated tissue in the mammalian body [1]. The nerves of the cornea provide autonomic responses such as tearing and blinking and assist in maintaining corneal epithelial homeostasis through the release of trophic factors [2]. Sympathetic innervation comes from nerve fibers originating in the superior cervical ganglion while sensory information is transmitted from the corneal epithelium to cell bodies located in the trigeminal ganglion, [3][4][5][6]. It is well-established that corneal stromal nerves enter the cornea in the peripheral stroma and travel horizontally before branching to give rise to vertical axons that penetrate the epithelial basal lamina [7,8]. Penetrated nerves ramify shortly after entering the corneal epithelium (in a process known as leash formation), and these ramifications constitute the sub-basal plexus. Axons in the sub-basal plexus travel anteriorly and laterally between the wing and superficial-squamous cells of the corneal epithelium, after which they give rise to the epithelial nerve plexus in addition to axon terminals [9,10]. Corneal innervation is a dynamic process, constantly changing as a result of aging and in response to pathology or injury [11]. The mechanisms by which corneal nerve patterning is regulated are not well established.
In addition to data gathered from studies on neurotransmission, our understanding of corneal innervation is largely based on light and electron microscopic imaging. While transmission electron microscopy (TEM) makes it possible to appreciate corneal nerve ultrastructure from single ultrathin sections, it provides only a two-dimensional perspective [12]. For a three-dimensional context, serial sections are needed and while serial sectioning using TEM is possible, the technical challenge of collecting serial sections is demanding and typically limits three-dimensional (3D) reconstructions to less than 50 serial images spanning a depth of no more than 5 microns [13]. To our knowledge, no serial sectioning electron microscopy studies have been reported on the nerves of the cornea.
With the advent of a relatively new technique known as serial block-face scanning electron microscopy (SBF-SEM) it is now possible to collect 3D ultrastructural data with relative ease. Routine automated collection of a thousand or more serially-registered images spanning a depth of 50 to 100 microns allows for superior 3D reconstructions and improved ultrastructural interpretation [14]. In addition to providing the ability to produce 3D reconstruction of tissue at an ultrastructural level, the context provided by serial section imaging allows for the identification of complex cell-cell interactions at an ultrastructural level that cannot be seen using light microscopy or single section electron microscopy. As a result, SBF-SEM has been applied across a great deal of tissue in the literature, but has yet to be used to study corneal nerves.
The purpose of the current study was to use the 3D capabilities of SBF-SEM to directly examine stromal nerve penetrations into the corneal epithelium of mice. Shortly after initiating the study, we observed for the first time a novel neuronal-epithelial cell interaction in which stromal nerves approach the epithelium and fuse with basal epithelial cells. Herein we use SBF-SEM to describe and compare two types of neuronal-epithelial interactions, simple neuronal penetration into the corneal epithelium and the novel fusion event that also occurs between stromal axons and basal epithelial cells. reconstruction using the Amira 6.0.1 software was conducted by a four-person reconstruction team. Care was taken to reconstruct the electron translucent axons separately from the electron dense axons within each nerve bundle, this was accomplished by assigning a different material (i.e. color) to each structure of interest. The basal lamina was identified by its characteristic electron density (lamina densa) on the stromal face of basal epithelial cells, and neuronal mitochondria by their electron dense double membrane and size.
Morphometric analysis. Morphometric analysis using standard stereological techniques was performed as previously described [19,20]. Stereology is an aspect of morphometry that takes advantage of the inherent mathematical relationships between three-dimensional objects and their two-dimensional representations (e.g., electron micrographs) [21]. These relationships are based on the reasoning of geometric probability and statistics, and the practice of using stereological grids has been used extensively over the past 50 years to obtain unbiased and accurate estimates of geometric features such as cell/organelle number, length, surface area, and volume [22][23][24][25][26][27][28].
In order to estimate the surface-to-volume ratio of fusing and penetrating nerves, a cycloid grid was used. Briefly, serial electron images were obtained of both fusing and penetrating nerve events as they approach/interact with the corneal basal epithelium (10 animals per group, with 20 nerves assessed in the fusing group and 23 nerves assessed in the penetrating group). The serial images in which the nerve is visible were identified, and a section was selected at random for analysis. Digital micrographs were analyzed in Adobe Photoshop (Adobe Systems Inc., San Jose, CA) using a cycloid grid [29]. The vertical axis of the grid was oriented in parallel to the basal lamina within each image in order to account for the anisotropic properties of the cornea. Line intersections with the nerve bundle of interest were counted, as well as target points located within the nerve bundle (Fig 1). In order to avoid counting line intercepts and target points within nerves located on the epithelial side of the basal lamina, a restriction line was drawn from one end of the basal lamina pore to the other and counts were only made on the stromal side of each nerve. The ratio between line intersections with the nerve and target points within the nerve was used to calculate the cell surface Morphometric analysis of corneal nerve surface-to-volume ratio using a cycloid grid. A single image from an SBF-SEM series showing a nerve that has fused with a basal epithelial cell (A). A micrograph from this series was selected at random and a cycloid grid was randomly cast onto the image while maintaining the orientation of the grid (defined by the vertical white arrow) parallel to the epithelial basal lamina (B). The intersection of the grid lines with the surface of the nerve bundle are marked with blue dots (surface area) while grid points falling within the nerve bundle are marked with green dots (volume); the inset, enlarged in panel (C), offers a magnified view of the grid. Scale bars = 2 μm. density, or surface-to-volume ratio using an established stereology formula: where I is the number of intersections between the grid lines and nerve bundle, P is the number of grid points falling within the target nerve, and l/p is the length of test line per grid point (corrected for magnification) [29].
Interactions between nerve and epithelium (fusion or penetration) include a discontinuity in the basal lamina. The maximum dimension of each discontinuity (i.e., basal lamina pore diameter) was identified within each image stack and measured using Fiji (ImageJ) [30].
Transmission Electron Microscopy (TEM). Tissue blocks containing verified neuronalepithelial cell fusion and nerve penetration into the basal epithelium were removed from the Gatan 3View2 system and ultra-thin sections 100 nm thick were cut, set on single slot copper grids, and imaged on a Tecnai G2 Spirit BioTWIN electron microscope (FEI Company, Hillsboro, OR). Nerve bundles were imaged and assessed for the presence of microtubules and cellular organelles.

DiI labeling of trigeminal ganglia
Tissue processing. DiI crystals were placed on the trigeminal ganglia of 6 C57BL/6J mice. Mice were euthanized by CO 2 asphyxiation followed by cervical dislocation. The head was then removed, the skin covering the skull removed (making sure to carefully cut around the tissue surrounding the orbit), and the skull was cut down the medial line and removed along with the brain up to the cerebellum, pons, and medulla. The head was then placed in 2% paraformaldehyde overnight. The following day, the trigeminal ganglion was located [31], severed at the ophthalmic branch, and DiI crystals (1, 1-dioctadecyl-3,3,3',3'-tetramethylindocarbocyanine perchlorate, ThermoFisher, Waltham, MA) were crushed into the ganglia using surgical tweezers (Fig 2). The region around the ganglia was dried using chem-wipes prior to DiI application, as DiI is a hydrophobic substance [32, 33]. The skull was then filled with 5% lowmelting temperature agarose using a pipette, allowed to harden at 4˚C for 2 minutes. The tissue was then placed in 2% paraformaldehyde and allowed to sit at 4˚C for 26 weeks. Following this period, the eyes were enucleated, corneas isolated, stained with DAPI, and flat-mounted for imaging. Control mice, where DiI was excluded from the tissue preparation, were included in the study.
Imaging of DiI labeled corneal nerves and basal epithelial cells. Corneas were imaged using a DeltaVision wide field deconvolution fluorescence microscope (GE Life Sciences, Pittsburg, PA) with a 60x immersion oil lens. Corneas were then scanned for fusion events defined as a basal epithelial cell/cells with DiI labeled plasma membrane immediately adjacent to a DiI labeled stromal nerve. The central cornea was defined as the centermost 2 mm of the cornea. The remaining 1.5 mm region, defined at its edge by the limbal vasculature, was considered the peripheral cornea.

Statistics
GraphPad Prism (GraphPad Software. San Diego, CA, USA) was used for statistical analysis and data represented as the mean ± standard error of the mean. A two-tailed Student's t-test was performed to compare surface-to-volume ratios between penetrating and fusing nerves, while a Mann Whitney U Test (Wilcoxon Rank Sum Test) was used to compare the basal lamina pore size between the two groups. A p-value of � 0.05 was considered to be statistically significant.

SBF-SEM imaging of mouse corneal nerves revealed conventional nerve penetration as well as novel neuronal-epithelial cell fusion events
Using SBF-SEM we were able to image conventional nerve penetration through the epithelial basal lamina, where a stromal nerve bundle containing multiple axons passes through the epithelial basal lamina to form a leash point whereby the nerve bundle gives rise to multiple smaller axonal projections which extend between epithelial cells and give rise to the sub-basal and epithelial plexuses. In addition to conventional nerve penetration through the basal lamina, a novel neuronal-epithelial cell fusion event was observed (Fig 3A). Nerve bundles involved in fusion contain axons whose plasma membrane is fused and continuous with that of a basal epithelial cell such that the axoplasm comes into direct contact with the cytoplasm of the fused epithelial cell (Fig 3B). In all cases of neuronal-epithelial cell fusion (21 total events across 10 animals), the fusing axons were accompanied by conventional penetrating axons within the same nerve bundle. In other words, these nerve bundles contained a mixture of fusing and penetrating axons. Penetrating axons were easily distinguishable amongst fusing axons as their axoplasm was characteristically electron dense compared to the diffuse, electron translucent axoplasm associated with fusing axons (Fig 4). Most often, a single nerve bundle fused with multiple basal epithelial cells; however, fusion with single basal epithelial cells was also observed. Whether fusion was initiated by the nerve or the epithelium could not be determined.
After the initial discovery of neuronal-epithelial cell fusion, we sought to determine the frequency and distribution of neuronal-epithelial cell fusion events using SBF-SEM on C57BL/6J mice (n = 6). Serial transverse images were collected from the central and peripheral cornea and nerves that approached the epithelial basal lamina were identified. Of 21 stromal nerve bundles that interacted with the central corneal epithelium, 9 contained axons that fused with basal epithelial cells (42.8% of nerves observed) while the remaining 12 nerve bundles only gave rise to conventional nerve penetration and leash formation. In contrast, stromal nerve bundles that engaged the basal epithelium in the peripheral cornea (21 interactions) showed no evidence of fusion as they penetrated the basal lamina and gave rise to the sub-basal and epithelial nerve plexuses.

3D Reconstruction of conventional nerve penetration and neuronalepithelial cell fusion
To better characterize the ultrastructural organization of neuronal-epithelial cell fusion and document how it differed from conventional nerve penetration, SBF-SEM was used to collect serial image stacks suitable for segmentation and 3D reconstruction. When segmenting neuronal-epithelial cell fusion, care was taken to trace the electron translucent portion of the fusing axon separately from the penetrating axons with their characteristic electron dense axoplasm.
In regards to conventional penetration events, 3D reconstruction revealed a stromal nerve bundle bifurcating before extending into the epithelium through two holes, or pores, in the basal lamina (Fig 5). The basal epithelial cells protrude into the stroma through the basal lamina pore (Fig 5B and 5C) while stromal axons pass through the pore into the corneal epithelium before ramifying and establishing the sub-basal nerve plexus (Fig 5G and 5H). By comparison, 3D reconstruction of a fusing nerve bundle reveals a mixed population of fusing and penetrating axons (Fig 6). In this example, the neuronal-epithelial cell fusion event occurred at the junction between three basal epithelial cells, commonly referred to as a Y-junction or tricellular corner. The electron dense axons within this nerve bundle passed into the basal epithelium through a pore in the basal lamina at this tricellular junction, and produced four ramifications (Fig 6H). The electron translucent axons within this nerve bundle did not penetrate into the epithelium, but rather fused with three separate basal epithelial cells through this basal lamina pore (Fig 6I).

Nerve bundles containing neuronal-epithelial cell fusion were morphologically distinct from conventionally penetrating nerve bundles
Stromal nerve bundles that interacted with the epithelium were comprised of penetrating nerves only or a mixture of penetrating and fusing nerves. In addition to their more electron dense axoplasm, the diameter of penetrating nerve bundles was noticeably smaller than their fusing counterparts (Fig 7A and 7B). This resulted in a marked difference in their surface-tovolume ratio (Fig 7C). Stromal nerve bundles that only penetrated the basal lamina and extended into the epithelium exhibited a small diameter and a high surface-to-volume ratio that was more than twice that of nerve bundles containing fusion. The smaller surface-to-volume ratio of the fused nerve bundles was consistent with the "swollen" appearance of their axoplasm (Fig 7B). Despite the marked differences in surface-to-volume ratios, the basal lamina pore size through which penetrating or fusing nerve bundles passed through was not different (Fig 7D).
Volumetric and surface data was extracted from the 3D reconstruction of fusion and conventional penetration seen in Figs 5 and 6. Over the same length of reconstructed nerve, the volume of axons penetrating into the epithelium was comparable, with a volume of 28.58 μm 3 in the conventional penetration event and 24.64 μm 3 in the fusing nerve bundle. However, the volume of fusing axons within the fusing nerve bundle accounted for three-fourths of the total nerve volume, with a volume of 75.42 μm 3 .

Axons fused to basal epithelial cells lacked microtubules and mitochondria proximal to sites of fusion
In penetrating axons, mitochondria were distributed throughout the axoplasm of the stromal nerve as well as the ramified epithelial projections (Fig 8A and 8B). This was true whether the nerve bundle consisted of only penetrating axons or whether the penetrating axons were Neuronal-epithelial cell fusion in the mouse cornea grouped alongside fusing axons, (i.e., a mixed nerve bundle). Fusing nerves contained mitochondria but only in locations distal to the fusion site (Fig 9). The axoplasm in close proximity to the fusion site was always devoid of mitochondria (Fig 8C). At higher resolution, the transmission electron microscope revealed the axoplasm of penetrating nerves was not only rich in mitochondria, but also microtubules (Fig 10A and 10B). By comparison, the axoplasm of fused nerves lacked microtubules near the site of fusion; the axoplasm appeared to be composed solely of dispersed and unidentifiable material (Fig 10C and 10D). Neuronal-epithelial cell fusion in the mouse cornea Anterograde labeling confirms corneal nerve fusion with the basal epithelium SBF-SEM imaging had proved useful for documenting nerve fusion at an ultrastructural morphologic level. Because of the novelty of the observation, we sought to confirm it using a functional method. The ultrastructure suggests the plasma membrane of the nerve fuses with the plasma membrane of the basal epithelial cell (Fig 3) and predicts that a lipid membrane dye, DiI, applied to the nerve should be able to diffuse into the lipid membrane of the fused epithelial cell. DiI is a commonly used neuronal tracer because it diffuses along the plasma membrane [34, 35] and cannot pass from the neuron to another cell in fixed tissue unless their plasma membranes are contiguous and this only occurs at sites of cell-cell fusion [36].
We placed the lipophilic dye DiI at the trigeminal ganglia of 6 C57BL/6J mice and allowed it to diffuse along and label neuronal projections that reached into the cornea. (Fig 11). DiI labeling revealed axons penetrating the corneal basal lamina, ramifying, and giving rise to the sub-basal plexus (Fig 11A). Importantly, DiI labeling was also seen in the plasma membrane of a sub-population of basal epithelial cells associated with stromal nerves at the level of the basal lamina (Fig 11B-11D). DiI labeled basal epithelial cells were found primarily in the central 2 mm of each cornea. Single labeled cells as well as clusters of labeled cells were observed (Fig 12). Cross-sectional projections of DiI labeled epithelial cells revealed the continuity of DiI labeling from stromal nerve to basal epithelial cell (Fig 12D).

Discussion
The purpose of this study was to describe and compare two types of neuronal-epithelial interactions, conventional neuronal penetration into the corneal epithelium and the novel neuronal-epithelial cell fusion that also occurs between corneal neurons and basal epithelial cells. To Fusing nerves had smaller surface-to-volume ratios than penetrating nerves. A penetrating nerve bundle which has passed through the epithelial basal lamina giving rise to the sub-basal plexus (A). The axoplasm was electron dense and contained numerous mitochondria. A nerve bundle containing fusion that has merged with a basal epithelial cell (B). The axoplasm was electron translucent and devoid of mitochondria. The surface-to-volume ratio of nerve bundles containing fusion were significantly smaller than that of penetrating nerve bundles consistent with their "swollen" appearance (C). The diameter of the basal lamina pores through which these nerves interact with the corneal epithelium was similar (D). Scale bars = 2 μm. https://doi.org/10.1371/journal.pone.0224434.g007 Neuronal-epithelial cell fusion in the mouse cornea our knowledge, this is the first study to document fusion between neurons and basal epithelial cells in the cornea. Segmentation and reconstruction of serial images collected using SBF-SEM allowed us to unequivocally identify neuronal-epithelial cell fusion events as the merging of neuronal and epithelial plasma membranes and respective cytoplasms. Plasma membrane fusion was independently confirmed by fluorescence microscopy imaging of lipid membrane dye transfer from the neuronal plasma membrane to the epithelial cell plasma membrane. Documenting neuronal-epithelial cell fusion in the mouse cornea adds a new layer of complexity to our understanding of corneal innervation and offers new insight into the regulation of corneal nerve patterning.
Using SBF-SEM we were able to visualize the penetration of stromal nerves through the epithelial basal lamina to contribute to the epithelial plexus. These nerves were electron dense, had a high surface-to-volume ratio, and contained abundant microtubules as well as mitochondria. The high surface-to-volume ratio of these penetrating nerves is characteristic of nerves throughout the body, and conducive to the cellular processes required for neuronal signaling [37-39]. Often, nerve bundles approaching the epithelium consist of a mixed bundle of penetrating and fusing neurons. Despite the intimate contact between penetrating and fusing axons within these bundles, no obvious morphological changes were seen in the penetrating axons. Whatever the mechanism responsible for neuronal-epithelial cell fusion, it is selective even within the same nerve bundle. Penetrating axons within a bundle containing fusion appear morphologically indistinguishable from axons present within penetrating bundles. Neuronal-epithelial cell fusion in the mouse cornea Despite the termination of fusing axons, these fusing bundles are still able to contribute to the epithelial plexus through their subpopulation of penetrating axons.
Within the cornea, neuronal-epithelial cell fusion is fairly common and occurs primarily in the central cornea between stromal nerves that are morphologically distinct from nerves that simply penetrate into the epithelium. Fusing nerves were shown to have a significantly lower surface-to-volume ratio, an electron translucent appearance, and a distinct lack of microtubules and mitochondria in close proximity to sites of fusion. Fusion was defined as the continuity between neuronal and epithelial plasma membrane such that the epithelial cytoplasm and neuronal axoplasm are in direct contact.
While this study is the first to our knowledge to describe heterotypic neuronal-epithelial fusion in normal adult tissue, the history of cell-cell fusion can be traced back to Schwann in 1839. Ironically, given that Schwann contributed so much to the study of neurons and their associated cells, he noted this cell-cell fusion between myoblasts while studying superficial dorsal muscle in pig embryos [40]. Cell-cell fusion has since been described in many other cellular systems [41][42][43][44][45][46][47]. A search of the literature reveals that neuron fusion has been reported to . The importance of cell-cell fusion in development and disease cannot be overstated, being involved in a wide array of biological processes, ranging from fertilization to the development of bone, muscle, and placenta, it has been implicated in the immune response, tumorigenesis, as well as aspects of stem cell-mediated tissue regeneration [53-62]. Regarding heterotypical cell-cell fusion, the fusion between neurons and stem cells during development is particularly noteworthy in relation to the fusion events outlined in this paper. Within the cornea there is a population of cells known as transient amplifying cells (TACs) which retain stem-like properties. TACs retain the ability to divide as they migrate towards the cornea center [63]. While it is not known whether these fused epithelial cells are in fact TACs, this is a possibility that warrants further study.
Neuronal-epithelial cell fusion occurs within nearly half of all nerve bundles penetrating the epithelial basal lamina in the central cornea. To our knowledge, no prior electron microscopic study has identified neuronal-epithelial cell fusion in either the cornea or other tissues within the body. Two factors likely account for this, and the first is the sparse and random nature of sampling inherent in transmission electron microscopy. To this point, in 2005 it was estimated that if all material that had ever come into focus in all of the transmission electron microscopes worldwide were gathered together the total tissue volume would account for less than one cubic centimeter of volume [29]. The likelihood of a section passing through a corneal nerve bundle just as it penetrates or fuses with the epithelium is rare given the small size of the nerve, the small size of the tissue block and the random nature of sampling. The second factor is the 2D nature of routine transmission electron microscopy and the uncertainty of identifying a cell profile as a neuron rather than an epithelial cell, a leukocyte or a keratocyte. The interpretation of a single electron micrograph is often subjective and always open to criticism. Such is not the case with SBF-SEM where the three dimensional context allows, for the first time, accurate and unambiguous ultrastructural detection of the neuron and its interaction with the basal epithelium.
Regarding light microscopy, the lack of a body of literature on neuronal-epithelial cell fusion can be linked to two primary factors. First, without an ultrastructural understanding of the morphology of fusing nerves, any detection of neuronal-epithelial cell fusion at the light Neuronal-epithelial cell fusion in the mouse cornea microscopic level would be difficult to interpret as such. For example, Al-Aqaba et al. may have observed neuronal fusion when noting "the termination of sub-basal nerves into characteristic bright bulb-like thickenings" roughly the size of basal epithelial cells using confocal microscopy in human corneas [64]. These characteristic bulb-like thickenings are visible, but not discussed, in several other published confocal images [65][66][67][68]. Second, the common fluorescent markers used to locate and study corneal nerves typically do not target membranes (e.g. Thy1-YFP and anti-beta-tubulin III antibody). Towards this point, detection of neuronalepithelial cell fusion using fluorescence microscopy necessitates using a continuous plasmamembrane bound dye or antibody specific to the neuronal lipid bilayer within the corneal tissue. And while DiI administered at the trigeminal ganglion fulfills this requirement, the technical and temporal requirements for this methodology are a limiting factor in its use. For most studies of corneal nerves an endogenous fluorescent marker such as Thy1-YFP, or an easily applied fluorescent antibody such as beta-tubulin III suffice for nerve localization, are well established methodologies within the tissue, and require marginal time and effort to use [69,70]. For this reason, anterograde DiI labeling of corneal nerves remains an esoteric technique. However, given the extensive use of DiI in the literature for studying cell-cell fusion, this methodology was uniquely suited for our purposes [71][72][73][74][75][76]. Neuronal-epithelial cell fusion in the mouse cornea When viewed using electron microscopy, fusing nerve bundles are morphologically distinct from nerve bundles simply penetrating into the basal epithelium. Fusing neurons exhibit an electron translucent "salt and pepper" axoplasm that is devoid of mitochondria and microtubules in the cytoplasm immediately surrounding the site of fusion. Stromal nerves involved in fusion have a significantly smaller surface-to-volume ratio, indicative of a large or swollen axon. Distal to the site of fusion however, these nerve bundles are morphologically indistinguishable from other stromal nerves, containing both mitochondria and microtubules. These observations may be linked to a calcium effect. It is well known that membrane fusion is often accompanied by an increase in intracellular calcium near the site of fusion [77][78][79][80]. Increased levels of intracellular calcium have been shown to lead to the breakdown of microtubules and the inability of mitochondria to associate with kinesin and dynein (motor proteins responsible for intracellular transport), which may explain why neither mitochondria nor microtubules are present proximal to sites of neuronal-epithelial cell fusion, but can be seen in abundance distal to sites of fusion [81][82][83]. Without mitochondrial support or functional microtubules to traffic mitochondria and intracellular proteins near the site of fusion, axonal swelling occurs. However, the fate of these fusing axons is not known [84]. Given that fusing nerves appear morphologically typical distal to sites of fusion, the fate of these neurons cannot be assumed. In fact, similar axonal responses have been seen to be both transient and reversible in a variety of models [85][86][87][88][89]. It is possible that fusion with basal epithelial cells denies a subpopulation of stromal nerves the ability to innervate the epithelium, causing them to undergo a form of Wallerian degeneration followed by continued growth, and subsequent attempts to penetrate into the basal epithelium [90].
While the lack of mitochondria and microtubules near sites of fusion suggest the fused axons may be neurologically inactive, it is important to consider this alternative. If fused axons are neurologically active, gap junction communication between a fused basal epithelial cell and its neighbors would surely "short-circuit" transmission unless the gap junctions switched to a "closed" state. The switch from an "open" to a "closed" state can occur in response to a variety of stimuli, including changes in the levels of intracellular calcium, pH, transjunctional applied voltage, phosphorylation, and in response to activation of membrane receptors [91][92][93]. Gap junction closure would also serve to mitigate the risk of infectious agent and/or toxin transfer from basal epithelial cells into fused stromal axons. If fused axons are capable of creating action potentials, then the fused epithelial cell may function as its terminal.
While the function of neuronal-epithelial cell fusion in the cornea is open to speculation, we favor the idea that this interaction plays a role in limiting and shaping the neuronal network. The rationale behind this idea comes from noting that although the stromal nerve plexus does not change with age, the basal and epithelial nerve plexuses are constantly in flux, changing tortuosity and losing density as we age [94][95][96][97][98][99]. This suggests that axonal rearrangement occurs even in the normal, uninjured cornea. Given the relatively high frequency of fusion in the normal mouse cornea, it seems reasonable to suppose that neuronal-epithelial cell fusion is a determinant of axonal patterning which in turn would affect corneal sensitivity and epithelial proliferation (e.g., through neuropeptide release). Additionally, as the corneal epithelial cells migrate towards the central cornea, the subbasal and epithelial plexuses are dragged along with them [65]. This creates the possibility of overabundant or improper innervation of the central cornea and the necessity of neuronal rearrangement. It is possible that neuronal-epithelial cell fusion plays a role in this, and this may account for the localization of fusion events within the central cornea. Rather than the complete degeneration and loss of a neuron spanning the distance between trigeminal ganglion and corneal surface, neuronal-epithelial cell fusion would allow a neuron to maintain the integrity of its soma during the process of axonal rearrangement.

Conclusion
Here we provide evidence for the novel neuronal-epithelial cell fusion event within the cornea. This event is primarily defined by the fusion between the plasma membrane of a stromal nerve with that of one or more basal epithelial cells such that axoplasm and cytoplasm are no longer separate. This event is morphologically distinct in that fusing nerves exhibit electron translucency, a lack of mitochondria and microtubules proximal to the site of fusion, and a significantly smaller surface-to-volume ratio. This cell-cell interaction may play a role in regulating neuronal patterning changes that accompany aging and tissue damage. Conceivably, within the cornea, neuronal fusion may influence corneal sensitivity and epithelial homeostasis throughout the life of an individual.