Ocular surface inflammation induces de novo expression of substance P in the trigeminal primary afferents with large cell bodies

We evaluated the changes in substance P (SP)-expressing trigeminal neurons (TNs) innervating the cornea following ocular surface inflammation. Ocular surface inflammation was induced in Sprague–Dawley rats using 0.1% benzalkonium chloride (BAK). The corneal staining score, corneal epithelial apoptosis, conjunctival goblet cells, and density of corneal subbasal nerve plexus (SNP) were assessed, and the mRNA levels of SP, interleukin (IL)-1β, IL-6, and tumour necrosis factor-α were measured in corneas and ipsilateral trigeminal ganglia (TG). SP-immunoreactivity (IR) was measured in corneal intraepithelial nerves and TNs. The cell size of corneal TNs in the TG was calculated. All parameters were observed immediately (BAK group), at 1 week (1 w group), and 2 months (2 m group) after 2 weeks of BAK application. BAK caused an increase in the corneal staining score and the number of apoptotic cells, loss of conjunctival goblet cells, reduced density of corneal SNP, and upregulated expression of SP and inflammatory cytokines in both the cornea and TG in the BAK group but those changes were not observed in the 2 m group. On the other hand, SP-IR% and mean cell size of corneal TNs increased significantly in the BAK, 1 w, and 2 m groups, compared to the control. Our data suggest that following ocular surface inflammation, large-sized corneal TNs which normally do not express SP, expressed it and this phenotype switching lasted even after the inflammation disappeared. Long-lasting phenotypic switch, as well as changes in the expression level of certain molecules should be addressed in future studies on the mechanism of corneal neuropathic pain.


Ocular surface alteration.
After the completion of the eye wiping test, the ocular surface was assessed according to the corneal staining score, TUNEL assay, haematoxylin & eosin (H&E) and PAS stain (n = 6 in each group). Compared to the control (0 ± 0), the corneal staining score was significantly higher in the BAK group (2.8 ± 0.4), but not in the 1 w (1.3 ± 0.5) and 2 m (0.3 ± 0.5) groups (Fig. 1). TUNEL assay showed that the number of apoptotic cells in the corneal epithelium (cells per high power field [HPF]) was significantly higher in the BAK group (17.8 ± 5.8) than in the control (1.5 ± 1.5), but the number of apoptotic cells was not different in the 1 w (5.5 ± 3.6) and 2 m (3.0 ± 2.8) groups. In addition, H&E staining showed that in the BAK group, many cells with small and round nucleus, thought to be inflammatory cells, were infiltrated in the anterior stroma, unlike keratocytes, which have a spindle-like nucleus and reside between aligned lamellae. Such cells were observed in fewer numbers in the 1 w group, and rarely in the 2 m and control groups. The density of conjunctival goblet cells (cells/mm), which was identified with periodic acid-Schiff (PAS) staining, was significantly lower in the BAK (11.9 ± 10.8) and 1 w (20.12 ± 10.9) groups than in the control (57.0 ± 8.2), but there was no difference between the 2 m group (50.5 ± 9.7) and the control group. The corneal staining score and number of apoptotic cells of corneal epithelium were significantly decreased in the 2 m group than in the BAK group, and the density of conjunctival goblet cells was significantly increased in the 2 m group than in the BAK group. Taken together, the BAK group showed significant deteriorations in the corneal epithelium and conjunctival goblet cells, whereas the 2 m group showed no significant difference in all those measurements compared to the control group.
Density of corneal nerve and percentage of substance P-immunoreactivity. The density and SPimmunoreactivity (IR) of corneal nerves were assessed in a different set of the four groups (n = 6). Figure 2 shows the representative images of the corneal subbasal nerve plexus (SNP) and its density in each group. The density of corneal SNP (mm/mm 2 ) was significantly reduced in the BAK (13.9 ± 11.7) and 1 w (44.5 ± 20.8) groups, but not in the 2 m group (161.3 ± 39.09) compared to the control group (201.4 ± 37.7). The damage and recovery of corneal SNP corresponded to that of the ocular surface findings presented in Fig. 1. Most of the SNP contained SP-positive axons within their bundles; however, their fluorescence signal varied in intensity or appeared segmented. Therefore, we measured the SP-IR% of intraepithelial nerve terminals (Fig. 3), which was calculated as 46.9 ± 10.0, 48.8 ± 9.7, 52.9 ± 8.1, and 57.5 ± 5.4 in the BAK, 1 w, 2 m groups, and control, respectively. These values did not differ among all groups (P = 0.2543).
Cell size distribution and substance P-immunoreactivity of corneal trigeminal neurons. Fast blue (FB)-labelled corneal trigeminal neurons were located in the ophthalmic division of TGs, and some popu- www.nature.com/scientificreports/ lation of these had SP-IR (Fig. 4). Table 1 shows the SP-IR% of FB-labelled neurons and cell size (mean ± SD, μm 2 ) of FB-labelled neurons with or without SP-IR. Among FB-labelled cells, SP-IR% was significantly higher in all groups compared to control (P = 0.0005, 0.0012, and 0.0472 for the BAK, 1 w, and 2 m groups, respectively). Additionally, the cell size of FB-labelled cells with SP-IR was significantly greater in all groups compared to control (P = 0.049, 0.011, and 0.005 for the BAK, 1 w, and 2 m groups, respectively). In contrast, no difference was observed in the cell size of FB-labelled cells without SP-IR. Interestingly, analysis of the cell size distribution showed that the frequency of corneal trigeminal neurons with SP-IR above 800 μm 2 increased in the BAK, 1 w, and 2 m groups (12.2%, 22.2%, and 21.4%, respectively) compared to the control group (0%), although corneal trigeminal neuron has the highest frequency in range 200 to 400 μm 2 , regardless of the groups (Fig. 5).

Expression of substance P and inflammatory cytokines in corneas and trigeminal ganglia.
The mRNA expression of SP and inflammatory cytokines (IL-1β, IL-6, and TNF-α) in corneas and ipsilateral TGs was determined using the formula 2 − C T (Fig. 6  The density of corneal subbasal nerve plexus. The representative images of corneal subbasal nerve plexus (SNP) stained with βш-tubulin antibody in the control, BAK, 1 w, and 2 m groups. The density (length per area, mm/mm 2 ) of corneal SNP decreased significantly after BAK treatment compared to the control, and at 2 months it was at near-control level. *Significant difference between paired groups by Dunn's multiple comparison test as a post-hoc test of non-parametric ANOVA test (P < 0.05).

Figure 3.
Immunolocalization of substance P in the corneal intraepithelial nerve terminals. The length of intraepithelial nerve terminals stained for substance P (SP) and βш-tubulin, a pan-neuronal marker, was measured and the SP-immunoreactivity (IR) (%) of corneal intraepithelial nerve terminals was calculated using Image J software. SP-IR (%) of corneal intraepithelial nerve terminals did not show any difference among the groups. In the BAK and 1 w groups, representing the acute phase of inflammation, SP-IR or background intensity was also observed in the neighbouring cells and extracellular matrix besides the nerve fibres. Scale bar = 20 μm.

Discussion
This study demonstrated that the proportion and mean cell size of corneal trigeminal neurons with SP-IR were significantly increased in the BAK, 1 w, and 2 m groups, compared to the control. Unlike these changes in the SP-expressing corneal trigeminal neurons observed in all groups, the ocular surface deteriorations including corneal epithelial apoptosis, loss of conjunctival goblet cells, decreased density of intraepithelial corneal nerves, and increased expression of inflammatory cytokines were observed only in the BAK group, and not in the 2 m group. These findings suggest that BAK induced ocular surface inflammation/damage resulted in de novo SP expression in large-sized corneal trigeminal neurons, which normally do not express SP, and this change persisted even when ocular surface inflammation/damage was apparently resolved.   [38][39][40][41][42] . It has been reported that peripheral nerve injury or inflammation can induce de novo expression of SP in neuronal cells, which are normally negative for SP [43][44][45] . Bae et al. 's study of rat TG reported that 98% of neuronal cell bodies with SP-IR are small to medium-sized with the cross-sectional area < 800 μm 238 . Our study showed that SP-IR neurons with the cross-sectional area > 800 μm 2 were absent in control, but increased to 12.2%, 22.2%, and 21.4% in the BAK, 1 w, and 2 m groups, respectively. Therefore, it supports our assumption that the increase in proportion and mean cell size of SP-IR corneal trigeminal neurons indicates the phenotype switching of large neurons to the ones expressing it, although distinct phenotype of SP-expressing neurons was not characterised in our study. This is an important finding because the phenotype switching of a sensory neuron is known to contribute to nociceptive sensitisation, in which noxious stimuli generate greater response and even innocent stimuli elicit pain, and pain sensitivity can be extended to surrounding undamaged tissues 46 .  www.nature.com/scientificreports/ Unlike our speculation, the pattern of SP-IR% in intraepithelial nerve terminals was not consistent with that in TG. SP-IR% of intraepithelial nerve terminals was relatively constant among all groups. SP is synthesised in the perikaryon of primary afferent neurons in TG or DRG, then packed into storage vesicles, and axonally transported to both central and peripheral processes 47 . Owing to the process of its synthesis, transport, and release, the distribution of SP can vary along a neuron. Additionally, SP may be transported predominantly to the central process in large-sized cells with SP-IR. Neumann's study 43 showed that Aβ fibre, newly expressing SP after inflammation, could contribute to the excitability of spinal cord neurons, rather than exerting an efferent function at the periphery. Although these reports explain the inconsistency in the pattern of SP-IR% between the peripheral endings and cell bodies of corneal trigeminal neurons, studies on the expression pattern at the centre is required to elucidate it. Additionally, investigations of branching and tortuosity of corneal nerves, which have been widely accepted as important markers of peripheral neuropathy 48 can be helpful in discovering peripheral changes corresponding to TG neuronal cell bodies.
Our results showed that inflammatory cytokines were slightly upregulated in the ipsilateral TG immediately and at 1 week after BAK treatment on the ocular surface, along with significant upregulation of inflammatory cytokines in the corneas. This result corresponds to previous studies, which have demonstrated that ocular inflammation/injury leads to the trigeminal and central inflammation along the cornea-trigeminal axis 36,49 . It has been widely accepted that SP can promote the production of inflammatory cytokines by activating immune cells, dependent or independent of their specific receptor, NK-1. In turn, inflammatory cytokines lead to SP production and release from sensory neurons and non-neuronal cells 50,51 . In this study, a slight increase in SP was observed in the BAK group along with an increase in the inflammatory cytokines in the BAK and 1 w groups, suggesting a possible link between inflammatory cytokines and SP expression. However, further studies into their sources and interaction will be needed to determine the connections between corneal inflammation and its expansion to the TG and neurochemical modification of neurons.
This study has some limitations. First, we were not able to determine whether changes in the SP expression profile of the corneal trigeminal neurons are related to an abnormal pain response. The eye wiping behaviour was used to indirectly assess pain sensitivity, but it did not show a significant difference between groups. The eye wipes appeared to be interrupted by other pain behaviours such as spasmodic blinking, prolonged eye closure, and burying heads in the sawdust. Therefore, standardised method assessing ocular pain or electrophysiological test should be included in future studies, as there is a recent advancement in developing experimental models for corneal pain studies 52 . Second, we did not evaluate the distinct subtype of large cells expressing SP due to limited resources. Recently, the latest analysis methods such as single-cell sequencing have further classified sensory neurons with the same neurochemical properties into distinct subtypes 53 . Although previous studies support our results making them interpretable, the distinct subtypes should be evaluated to expand our understanding of SP-related pain mechanism and approach the possible target for pain control.
In this study, we demonstrated that BAK-induced ocular surface inflammation caused SP expression in some population of large-sized TG neurons innervating the cornea, which normally do not express SP, and this phenotype switch lasted even after the ocular surface inflammation disappeared. This is a novel finding because to the best of our knowledge, no study has reported the long-lasting neuroplasticity related to ocular surface inflammation/damage. Clinically, many patients suffer from chronic dry eye symptoms or persistent postsurgical pain with minimal/no ocular surface signs. Although this phenotype switch of primary sensory neurons has already been addressed as a mechanism of pain pathology in other tissues, verification of it in the eye is an important step to expand our understanding of corneal nociceptive pain. Future studies on functional and neurochemical characterization of SP-expressing corneal neurons and inflammatory mechanisms that trigger their phenotypic change will provide insights into new therapeutic targets and approaches to corneal neuropathic pain.  54 . Twentyseven rats in each group were evaluated immediately (BAK group), at 1 week (1 w group), and 2 months (2 m group) after 2 weeks of BAK application. As the control group, 27 rats were assessed immediately after phosphate buffer saline (PBS) was applied to their unilateral eyes instead of BAK twice a day for 2 weeks. In each group, 6 eyes were used for the eye wiping test, corneal vital staining, TUNEL assay, HE and PAS staining; 6 for immunofluorescence staining of corneal nerves, 9 for RNA isolation, and 6 for immunofluorescence staining of TG with retrograde labelling.

Methods
Eye wiping test. The eye wiping behaviour was observed to assess pain sensitivity according to Farazifard's protocol 55 . Briefly, rat placed in a transparent cage was allowed to adapt to the experimental environment for 30 min, and then the number of eye wipes performed with forepaws was counted for 30 s after dropping 5 M NaCl solution into their eyes.
TUNEL assay and H&E and PAS stainings. Enucleated eyeballs (n = 6 per group) were immersed in fresh 2.5% glutaraldehyde solution overnight at 4 °C and were cut into 10 μm-thick paraffin sections from the vertical meridian of the eye. TUNEL assay was performed to discern the apoptotic cells in the cornea using the Click-iT TUNEL assay kit (C10246; Thermo-Fisher Scientific, Waltham, MA, USA), according to the manufacturer's instructions. Images were captured using confocal laser scanning microscope (LSM800, Carl Zeiss), and TUNEL-positive cells (cells in an area of 600 × 600 μm 2 ) in the corneal epithelium were counted. H&E and PAS staining (Sigma-Aldrich) were conducted to identify the presence of inflammatory cells in the cornea and the goblet cell density in the conjunctiva, respectively. Representative images were taken at the central cornea and the inferior conjunctival fornix in each group using a camera-equipped inverted light microscope (DMI 5000B, Leica, Wetzlar, Germany). The goblet cell density (cells/mm in an area of 400 × 300 μm 2 ) was calculated as the number relative to the length of the conjunctival epithelial line using ZEN software (blue edition, Carl Zeiss) and Image J, ver. 1.47 (NIH, Bethesda, MA, USA).
Immunofluorescence staining of corneal nerves. Corneas were processed for a flat mount, as described previously 58 . Briefly, corneas (n = 6 in each group) were trephined from the enucleated eyeballs and fixed in 4% paraformaldehyde (PFA) on ice for 1 h. After washing with PBS, the corneas were incubated in 0.1% EDTA (Sigma-Aldrich) and 0.02% hyaluronidase (type IV-S; Sigma-Aldrich) at 37 °C overnight, and further permeabilised in PBS with 0.3% Triton-X (PBST) containing 10% donkey serum and 1% bovine serum albumin (BSA; Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA) for 2 h at RT. Further, corneas were incubated with primary antibodies in 0.1% PBST for 48 h at RT, followed by secondary antibodies in 0.1% PBST for 24 h at 4 °C. After incubation with antibodies, the corneas were washed extensively for 1 h in PBS, and then flattened by six radial cuts. Flat corneas were mounted on glass slides (Muto Pure Chemical Co., Ltd., Tokyo, Japan) using mounting media containing DAPI (Vector Laboratories, Burlingame, CA, USA), and images were captured using confocal laser scanning microscope. Mouse anti-β tubulin III, a pan-neuronal marker (1:200; ab78078, Abcam, Cambridge, UK) and rabbit anti-SP (1:400; ab67006, Abcam) were used as the primary antibodies. Fluorophore-conjugated secondary antibodies (1:400; Alexa Fluor 488 and 594, Abcam) were used according to the host species of the primary antibodies. Negative control staining was performed by omitting primary antibodies. Representative images (320 μm × 320 μm) were taken at the mid-periphery of each cornea using a confocal laser scanning microscope, and then processed with the z-axis projection to form the image corresponding to the corneal SNP and intraepithelial nerve terminals, following which, the density (total length/area, mm/mm 2 ) of corneal SNP and SP-IR% (length/length) of intraepithelial nerve terminals were calculated using Image J.
Immunofluorescence staining of trigeminal ganglia with retrograde labelling. Retrograde tracing method was used to label the TG neurons innervating the corneas (corneal trigeminal neurons) at 1 week before the immunofluorescence staining of TGs in each group (n = 6) according to the modified Gonzalez-Coto's protocol 59 . Briefly, a 5 mm diameter-disc of Surgicel (Johnson & Johnson, New Brunswick, NJ, USA) soaked with 5% FB (FB; Polysciences, Inc., Warrington, PA, USA) was placed on the entire cornea for 2 h under general anaesthesia. After 1 week, which allowed FB to label the corneal trigeminal neurons retrogradely, TG was dissected following perfusion, fixation with 4% PFA, and processing for frozen sections. Ten-micrometer sections on the glass slides were washed thrice for 15 min and blocked in 0.1% PBST containing 10% donkey serum and 1% BSA for 30 min at RT. Then, the samples were incubated with anti-SP (1:400; Abcam) antibody in 0.1% PBST overnight at 4 °C, followed by secondary antibody in 0.1% PBST for 1 h at RT. After washing thrice for 5 min with PBS, samples on the glass slides were mounted in a medium without DAPI and examined using a confocal laser scanning microscope. The frequency of FB-labelled corneal trigeminal neurons with/without SP-IR and their cell size (the cross-sectional area, μm 2 ) were measured using Image J.
RNA isolation and quantitative real-time PCR. Nine cornea and ipsilateral TG each group were used to isolate RNA and analyse the mRNA level of SP and inflammatory cytokines using quantitative real-time PCR, according to the manufacturer's instructions. Three corneas and TGs were used for a single test, and the result was averaged from three tests in each group. Tissue was homogenised in lysis buffer in a 1.5 mL tube. RNA extraction was done using RNeasy Mini Kit (#74,106, Qiagen, Hilden, Germany). The cDNA was synthesised from 1 μg total RNA using the RT2 First-strand cDNA synthesis kit (#330,401, Qiagen) and was mixed with primers and RT2 SYBR Green Mastermix (330,500, Qiagen). Each sample was assayed in duplicates using a realtime cycler (Bio-Rad, Hercules, CA, USA) under the conditions of 10 min at 95 °C, followed by 40 cycles of 95 °C for 15 s, and 60 °C for 1 min. The primers used were Tac1 for SP (PPR42621A, Qiagen), IL-1β (PPR06480B, Qiagen), IL-6 (PPR06483B, Qiagen), and TNF-α (PPR06411, Qiagen); and glyceraldehyde-3-phosphate dehydrogenase (GAPDH, PPR06557B, Qiagen) was used as the reference gene. The fold change with respect to control was Scientific RepoRtS | (2020) 10:15210 | https://doi.org/10.1038/s41598-020-72295-x www.nature.com/scientificreports/ calculated using the 2 − C T method. More than twofold upregulation or less than 0.5-fold downregulation was considered as a significant change.
Statistical analyses. All experimental data were analysed by non-parametric one-way analysis of variance (the Kruskal-Wallis test) after the normality of data was determined by D' Agostino & Pearson omnibus normality test. Pearson's Chi-squared test was used to compare the frequency among the groups. Two-sided P values < 0.05 were considered significant. All data are presented as mean ± SD. The data were analysed using the SPSS software (version 15.0, SPSS Inc., Chicago, IL, USA) and GraphPad Prism ver. 5.0 (GraphPad Software Inc., San Diego, CA, USA).

Data availability
The datasets used in the current study are available from the corresponding author on reasonable request.