Ocular Surface Injury Induces Inflammation in the Brain: In Vivo and Ex Vivo Evidence of a Corneal–Trigeminal Axis

Abstract

Purpose.: To test whether a corneal injury can stimulate inflammation in the trigeminal ganglion (TG), a structure located in the brain.

Methods.: At 4 and 8 days after alkali burn induced in the right eyes of mice, in vivo magnetic resonance imaging (MRI) of the brain was done before and after ultrasmall superparamagnetic iron oxide nanoparticle (USPIO) contrast to track macrophages. Trigeminal ganglia were stained for Prussian Blue and inflammatory cell markers. Interleukin-1β, TNF-α, and VEGF-A transcripts were quantified on days 1, 4, and 8, and 4 days after corneal topical anti-inflammatory treatment with 0.2% dexamethasone. The expression of Substance P and its receptor NK-1R was also measured in the TG on day 4.

Results.: Corneal alkali burn induced leukocyte infiltration, including T cells, in the right TG at 4 and 8 days. In vivo MRI showed an increased contrast uptake in the right TG, which peaked at day 8. Prussian Blue⁺ USPIO⁺ macrophages were observed in the right TG and exhibited an M2 phenotype. The M2-macrophage infiltration was preponderant in the TG after damage. The proinflammatory cytokines Substance P and NK-1R were significantly increased in both the TGs. The expression of IL-1β and VEGF-A was significantly reduced in the right TG with dexamethasone treatment.

Conclusions.: We suggest, for the first time, inflammatory involvement of brain structures following ocular surface damage. Our findings support the hypothesis that the neuropeptide Substance P may be involved in the propagation of inflammation from the cornea to the TG through corneal nerves.

Italian Abstract

Introduction

The cornea receives the densest innervation of the whole body. Corneal sensory nerves originate in the trigeminal ganglion (TG) and reach the cornea through the first (ophthalmic) branch of the trigeminal nerve. Interestingly, a limited number of trigeminal cells (up to 450) ^1–3 give origin to up to 630,000 corneal nociceptors. ²  

Prior studies have clarified that corneal nerve terminals exert a number of functions beyond mere sensory perception; these include support for epithelial cell proliferation and/or migration ^4,5 and possibly immune regulation. ^6,7 Moreover, their loss has been described in the setting of corneal neovascularization. ⁸ Hence, it is not surprising that trigeminal ablation and/or altered function are associated with a number of corneal alterations, globally defined as neurotrophic keratitis. However, the effect of corneal nerve damage and/or ablation on TG has not been clarified yet. It has been shown that following herpetic keratitis, inflammation develops in the TG. ⁹ This is generally thought to be a response to herpes virus infiltration/replication in the TG itself. To the best of our knowledge, there is no literature showing the impact of nonherpetic corneal nerve damage on the TG. 

The fact that the TG is affected by corneal inflammation may have significant clinical implications, since corneal nerves are very close to the ocular surface and hence are the first structures damaged by ocular injuries. These represent a major cause of blindness worldwide, affecting 55 million people every year and leaving 19 million visually impaired. ¹⁰ This adds to highly prevalent ocular surface diseases in which peripheral nerves are typically altered, such as infectious keratitis and dry eye. Finally, the increasing popularity of surgical procedures involving the superficial corneal layers, where nerves are located (e.g., refractive surgery), makes corneal nerve injury even more frequent. 

In the present work, we hypothesize that corneal inflammation propagates to the TG through a cornea–trigeminal axis, which is represented by corneal nerves (Fig. 1). Further, we suggest that neuropeptide Substance P may represent an important mediator of this process. We demonstrate inflammatory infiltration in the TG through magnetic resonance imaging (MRI) with ultrasmall superparamagnetic iron oxide nanoparticles (USPIO), which allow in vivo macrophage tracking. ^11,12  

Figure 1

 

Schematic representation of the study design. The right cornea (1) of CD1 mice was burned with 1 N NaOH (red flash) to test for inflammatory changes in the homolateral (hl) TG (2), the contralateral (cl) TG (3), and the contralateral left cornea (4). A, anterior; P, posterior.

Finally, our findings may provide significant insight into the pathology of highly prevalent, potentially blinding ocular diseases and suggest that MRI may represent a powerful tool to finely examine the inflammatory process. 

Materials and Methods

Animals

Male 6- to 8-week-old CD1 mice (Charles River, Calco, Italy) were used in all experiments (42 total mice). Animals were allowed to acclimatize to their environment for 1 week prior to experimentation. Each animal was deeply anesthetized with intraperitoneal injection of tribromoethanol (250 mg/kg) before all surgical procedures. Postoperatively, all animals received a single dose of carprofen at 5 mg/kg subcutaneously. Carbon dioxide inhalation and subsequent cervical dislocation were applied to euthanize the animals. All experimental protocols were approved by the Animal Care and Use Committee of the San Raffaele Scientific Institute, in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Alkali Burn

A corneal alkali burn was created in the right eye of each mouse as previously described. ¹³ Mice were then randomized and followed for different time points after alkali burn. On days 4 and 8, six mice per time point were used for MRI analysis; then corneas and TGs were removed for immunostaining. On days 1, 4, and 8, six mice per time point were used for real-time PCR analysis. Finally, a group of six healthy animals was used as control (day 0). 

Corneal photographs were taken with the slit-lamp microscope SL 990 (C.S.O., Florence, Italy) at different time points. 

Dexamethasone Treatment

A group of six mice was treated topically with 0.2% dexamethasone eye ointment (Allergan, Inc., Irvine, CA, USA) only in the alkali-burned eyes. The treatment was started immediately after injury and applied three times per day for 4 days. Twelve hours after the final instillation of dexamethasone, both corneas and TGs were removed and real-time PCR was performed. 

In Vivo MRI

Magnetic resonance imaging studies were performed on a 7-Tesla animal scanner (Bruker BioSpin, Ettlingen, Germany). Mice were kept under anesthesia during the entire MRI session with a mix of isoflurane and oxygen and were placed onto a volumic radiofrequency coil (inner diameter 3.5 cm). The anesthetic gas level was adjusted based on the constantly monitored respiration rate, and body core temperature of mice was maintained at 37°C by warm water flow. 

Anatomical reference images in the three planes were acquired using a fast spin-echo T2 weighted sequence (RARE) with a repetition time of 2500 to 3000 ms; effective echo time of 33 to 42 ms; RARE factor of 8; 0.7- to 0.85-mm slice thickness (15 coronal sections); two averages; and 93- to 130-μm in-plane resolution. The MRI signal is quite sensitive to the presence of iron particles that induce in particular a reduction of the T2 relaxation time, resulting in a loss of signal. To quantify the T2 relaxation time, a multi-spin-echo multi-echo (MSME) was used with a repetition time of 3000 ms, a train of 12 echoes from 8.3 to 99.6 ms, six averages, and a spatial resolution of 82.3 × 125 μm² with coronal sections 0.8 mm in thickness. From this acquisition, T2 relaxation time maps were generated using the software ParaVision (Bruker BioSpin). This MRI protocol was applied before and 24 hours after intravascular administration of a recently developed USPIO (9.7 μmol P904; Guerbet Laboratories, Paris, France), which was optimized for MRI of macrophages. ¹⁴  

To determine the presence of USPIO, the T2 value distribution was analyzed within both left and right TGs. For each TG area from transversal sections (n = 5–7), the percentage of pixels with a T2 lower than 43 ms was calculated. This threshold was defined from the analysis of T2 before contrast administration, for which values were found between 52 and 60 ms (mean = 55.5 ± 4.1 ms in 14 mice), and the percentage of pixels with a T2 < 43 ms was approximately 0% and 5% (mean = 3.14 ± 3.67% in 14 mice). 

Immunohistochemical Analysis in Cornea and Trigeminal Ganglion

On days 0, 4, and 8 after alkali burn, six eyes and six TGs (left and right) per group were frozen, sectioned, and processed as previously described. ¹³ Corneal and TG sections were immunostained with goat anti-CD45 (leukocyte common marker, 0.2 mg/ml; R&D Systems, Minneapolis, MN, USA), while TG sections were stained with rat anti-CD3 (T-cell marker, 0.5 mg/ml; BioLegend, San Diego, CA, USA), rabbit anti-cleaved caspase-3 (marker of apoptosis; Cell Signaling Technology, MA, Danvers, USA), rat anti-F4/80 (macrophage common marker, 0.5 mg/ml; BioLegend), goat anti-CD206 (M2-macrophage marker, 0.2 mg/ml; R&D Systems), and rabbit anti-IBA1 (M1-macrophage marker, 0.5 mg/ml; Wako Chemicals, Richmond, VA, USA) at 4°C overnight (1:200 dilution). They were subsequently stained with Alexa Fluor 488/546 donkey anti-goat IgG, Alexa Fluor 488/546 donkey anti-rat IgG, or Alexa Fluor 488/633 donkey anti-rabbit IgG (2 mg/mL; Invitrogen-Molecular Probes, Paisley, UK) in a 1:500 dilution for 2 hours at room temperature. Sections were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; Vector, Burlingame, CA, USA), mounted, and photographed using a DFC310FX digital camera attached to a CTR5500 fluorescence microscope (Leica Microsystems, Wetzlar, Germany). 

Uptake of USPIO by cells in the cornea and the TG was detected using Prussian Blue staining. Slides were incubated with a solution containing equal parts 0.25 N hydrochloric acid and 2% potassium ferrocyanide at room temperature for 30 minutes, then washed with PBS three times. Some slides were counterstained with Nuclear Fast Red (Sigma-Aldrich Corp., St. Louis, MO, USA) for 5 minutes. 

For the immunohistochemical characterization of iron-positive cells, slides were incubated first with hydrochloric acid/potassium ferrocyanide solution and secondly with goat anti-CD45 or rat anti-F4/80 and goat anti-CD206 at 4°C overnight. As secondary antibody, Alexa Fluor 488/546 donkey anti-goat IgG and Alexa Fluor 546 donkey anti-rat IgG were used for 2 hours at room temperature. Sections were counterstained with Hoechst 33342 (Invitrogen-Molecular Probes) and mounted with fluorescence mounting medium (Dako, Agilent Technologies, Inc., Glostrup, Denmark). 

Digital images taken at ×40 magnification were analyzed by ImageJ 1.44p Software (National Institute of Mental Health, Bethesda, MD, USA). Percentage areas of positive pixel staining per field were calculated for CD45 staining in both cornea and TG and for CD3 staining in the TG. Three fields per section of six longitudinal sections were taken. Cell infiltration was quantified by threshold setting, including the bright green cells and excluding the dark background. Each TG section was divided into two halves for purposes of analysis: anterior and posterior. Two images were taken along each section, one anterior and one posterior. The CD45⁺ cell distribution within both left and right TGs was analyzed by comparing the average of CD45⁺ area of six anterior fields with those of six posterior fields, taken along the whole TG thickness. 

Analysis of Cytokine Transcripts by Real-Time PCR

On days 0, 1, 4, and 8, six corneas and six TGs (right and left) per group were homogenized with Ultra-Turrax T8 (IKA, Wilmington, NC, USA). Total RNA extraction, DNAse treatment, retrotranscription, and real-time PCR were performed as previously described. ¹³ We used Taqman Gene Expression Assays (Applied Biosystems, Foster City, CA, USA) for tumor necrosis factor-α (TNF-α, Mm00443258_m1), interleukin-1β (IL-1β, Mm01336189_m1), vascular endothelial growth factor-A (VEGF-A, Mm01281449_m1), tachykinin precursor 1 (TAC1, Mm01166996_m1), and tachykinin receptor 1 (TAC1R, Mm00436892_m1). Glyceraldehyde 3-phosphate dehydrogenase (GAPDH, Mm99999915_g1) transcript was used as endogenous control. Results are presented as relative expression for six normal corneas or TGs (ΔΔC_T method). 

Corneal Trephination Model

The right cornea of six mice was trephined as previously described. ¹⁵ Briefly, using a 2-mm trephine to delineate the area, a circular incision was made through corneal epithelium and penetrated approximately into the anterior stroma to cut the sub-basal corneal nerves. Corneal photographs were taken with slit-lamp microscope SL 990 (C.S.O.) at different time points. 

After 4 days, freshly enucleated eyes were prepared into corneal flat mounts, and corneal nerves were immunostained with rabbit anti-mouse β III tubulin primary antibody (Chemicon, Temecula, CA, USA) as previously described. ⁸ Corneal whole mounts were imaged using confocal immunofluorescence microscopy (model TCS SP5; Leica Microsystems). 

Finally, the TGs were removed and used for the analysis of IL-1β, VEGF-A, TAC1, TAC1R, and GAPDH transcript by real-time PCR with the ΔΔC_T method as previously described. 

Statistical Analysis

Unpaired t-test was used to evaluate the difference in leukocyte, T-lymphocyte, and macrophage infiltration and cytokine expression in corneas and TGs, and also for the difference in T2 values and in apoptosis. Significance was defined as a P value < 0.05. All results are presented as mean ± standard error of the mean (SEM). All data were processed using GraphPad Prism software 5.0 (GraphPad Software, Inc., San Diego, CA, USA). 

Results

Alkali Burn in the Cornea Induces Leukocyte, Including T-Lymphocyte, Infiltration in the Trigeminal Ganglion

First, we evaluated whether an alkali burn in the cornea is able to induce inflammatory activity in the TG. Specifically, we quantified the inflammatory cell infiltration in both the cornea and the TG. 

Slit-lamp examination revealed clear damage in all alkali-burned corneas at different time points (Fig. 2A, upper). In particular, corneal transparence reduction and neovessels were observed on day 4 and increased on day 8. The contralateral eyes did not show any pathological changes. Moreover, as expected, the alkali burn induced leukocyte infiltration in the cornea, which increased over time. Leukocyte and T-lymphocyte infiltration was also observed in the homolateral TG (Fig. 2A, lower). CD45⁺CD3⁺/CD45⁺ cell ratio was 4.89 ± 0.29% (mean ± SEM) before and after injury. The cryosections of contralateral TGs did not show a significant increase of CD45⁺ or CD3⁺ cells as compared to values in healthy animals. 

Figure 2

 

Inflammatory cell infiltration detected in the alkali-burned cornea and in the homolateral trigeminal ganglion. (A) Slit-lamp examination of corneas 4 and 8 days after alkali burn. Reduced corneal transparency and growth of neovessels are detectable in the alkali-burned eye. The injury induced infiltration of CD45⁺ leukocytes (green positive cells) in both cornea and TG sections and of CD3⁺ T lymphocytes (red positive cells) in the TG. (B) The CD45⁺ cell increase was statistically significant in the alkali-burned (ab) eye in comparison to the contralateral (cl) eye and to the control (day 0) on both days 4 and 8 (n = 6). (C, D) The homolateral (hl) TG showed a significant CD45⁺ (C) and CD3⁺ (D) cell increase in comparison to the contralateral (cl) TG and to the control (day 0) over time (n = 6). (E, F) CD45⁺ leukocyte infiltration (green positive cells) was significantly increased in the anterior versus the posterior part of the TG on day 4. At day 8, no difference in CD45⁺ cell infiltration was detectable between the anterior and posterior TG (n = 6). Histograms represent mean values ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001.

CD45⁺ leukocytes were significantly increased (10-fold) on day 4 in the alkali-burned cornea in comparison with contralateral or healthy eyes. Moreover, CD45⁺ cell number significantly (P < 0.05) increased (2-fold) in the burned corneas from day 4 to day 8 (Fig. 2B). 

Similarly, TG homolateral to the damaged cornea showed an increase (2-fold) of CD45⁺ cells (Fig. 2C) and of CD3⁺ T cells (Fig. 2D) as compared to the contralateral TG and to healthy control. These differences were similar and statistically significant at both days 4 and 8 after the alkali burn. Of note, we found few positive CD3 cells in the control and contralateral TG as has been previously reported. ¹⁶  

Interestingly, CD45⁺ leukocyte infiltration was dominant in the anterior part of the TG on day 4 (Fig. 2E). On day 8, however, CD45⁺ cells were homogeneously distributed along the TG, suggesting antero-posterior inflammatory cell migration. This difference in leukocyte infiltration between the anterior and posterior part of the TG was statistically significant (P < 0.05) on day 4, but not on day 8 (Fig. 2F). 

In Vivo MRI Evidence of Macrophage Infiltration in the Trigeminal Ganglion After Corneal Alkali Burn

To evaluate the infiltration of inflammatory cells following corneal damage, we performed in vivo MRI follow-up (Fig. 3A) with administration of USPIO contrast, which is uptaken by activated macrophages and detectable as signal loss on T2 weighted MR images. Indeed, a reduction of the T2 relaxation time was clearly differentiated within the TG homolateral to the corneal damage, specifically at day 8 post injury (Fig. 3B). This USPIO uptake within the TG was further confirmed by quantitative analysis of T2 reduction at both days 4 and 8 (Fig. 3C). Uptake of USPIO was predominant (P < 0.05) in the anterior part of the TG at day 4 and with a significant (P < 0.05) extension toward the posterior part at day 8 (Fig. 3C). This is in line with the TG CD45 immunostaining results (Fig. 2). 

Figure 3

 

In vivo MRI after corneal alkali burn: evidence of macrophage infiltration in the trigeminal ganglion by USPIO contrast uptake. (A) In vivo MRI view along the TG from anterior to posterior (A to P) used as reference image to define the coronal views (hl, homolateral TG). (B) Compared to the precontrast, a clear T2 reduction was observed inside the homolateral TG (arrowheads), which was more pronounced at day 8 post USPIO administration (n = 6). (C) The analysis of T2 values highlighted the increased area of the homolateral TG with USPIO uptake between days 4 and 8, with a particular increase toward the posterior part of TG. A significant difference was found between the anterior and posterior TG at day 4 (§P < 0.05) and between day 4 and day 8 for the posterior TG (*P < 0.05; n = 6). (D) Prussian Blue staining confirmed USPIO uptake in the homolateral TG (blue stain) as shown at 3× magnification (I₁ and I₂). The contralateral TG was negative. Histograms represent mean values ± SEM.

To confirm the presence of iron uptake by macrophages in the homolateral TG as observed by MRI, Prussian Blue staining was performed. As shown in Figure 3D, a well-defined blue stain around nuclei was visible in TG cryosections, confirming USPIO uptake. The contralateral TG was negative. 

Characterization of Macrophage Infiltration in the Trigeminal Ganglion

In order to better characterize the USPIO-laden cells infiltrating the cornea and the TG, we performed additional immunostaining. Iron-laden cells (positive for Prussian Blue) infiltrating the cornea and the homolateral TG after alkali burn were positive for CD45 (Figs. 4A, 4C). Staining for F4/80 (generic macrophage marker) and for CD206 (M2-subtype marker) identified the Prussian Blue-positive cells as M2 macrophages, in the cornea (Fig. 4B) and in the TG (Fig. 4D). 

Figure 4

 

Immunohistochemical characterization of USPIO-laden macrophages in the cornea and trigeminal ganglion. Four days after alkali burn, some of the infiltrating CD45⁺ cells were positive for Prussian Blue (DAPI⁺ PB⁺ CD45⁺ cells), in both the homolateral cornea (A) and TG (C). Prussian Blue colocalized also with the macrophage markers F4/80 and CD206, in both the homolateral cornea (B) and TG (D), suggesting that USPIO-laden cells were M2 macrophages. (E) Representative image of triple labeling with different macrophage markers in the homolateral TG after 4 days: the common marker F4/80 (green), the M2-marker CD206 (red), and the M1-marker IBA1 (blue). (F) The predominant macrophage phenotype in the TG was the M2 subtype (F4/80⁺CD206⁺IBA1⁻) in all conditions. The number of M2 macrophages was significantly increased in the TG homolateral (hl) to the alkali-burned cornea on day 4 (**P < 0.01) in comparison to both control (day 0) and contralateral (cl) TGs. The M1 and M1/M2 subtypes (F4/80⁺CD206⁻IBA1⁺ and F4/80⁺CD206⁺IBA1⁺, respectively) were significantly increased on day 4 (#P < 0.05 and §§§P < 0.001, respectively) and on day 8 (##P < 0.01 and §§P < 0.01) in comparison to the control TG. (G) Representative image of caspase-3 staining in the homolateral TG after 4 days. Following corneal damage, cells in apoptosis (caspase-3⁺, green) were detected in the TG; these cells tested negative for the macrophage/microglia marker F4/80 (red). (H) The corneal alkali burn induced a significant increase of caspase-3⁺ cells in the homolateral (hl) TG after 4 and 8 days (**P < 0.01; ***P < 0.001). Histograms represent mean values ± SEM.

All the F4/80⁺ infiltrating cells were also CD45 positive. Moreover, all the iron-laden cells were positive for CD206, but not all the F4/80⁺CD206⁺ cells were positive for Prussian Blue (no USPIO uptake), suggesting that not every macrophage internalized USPIO, as reported previously. ¹¹  

We performed additional triple labeling in the TG (Figs. 4E, 4F) with F4/80, CD206, and IBA1 (M1-subtype marker). In the control, M2 (F4/80⁺CD206⁺IBA1⁻) macrophages were preponderant (approximately 80% of the F4/80⁺ macrophages) and were moderately decreased after injury (66%–72% after 4–8 days) in the homolateral TG. Instead, the percentage of M1 (F4/80⁺IBA1⁺CD206⁻) macrophages slightly increased after injury (from 11%–14%–17% after 4–8 days). A population of triple-positive macrophages (F4/80⁺CD206⁺IBA1⁺) was also found, which increased after damage on day 4 (from 8%–20%) and then decreased on day 8 (from 20%–11%). Although there was a decrease in the M2-macrophage percentage of the total number of cells over time, the absolute number of M2 macrophages in the homolateral TG after 4 (24 ± 6 cells/field) and 8 (27 ± 5) days was significantly higher compared to that in healthy (16 ± 2) and contralateral TG (P < 0.01) at 4 days (13 ± 5) and 8 days (15 ± 3). M1- and M1/2-macrophage analysis showed a significant increase of these cells only in comparison to the control (P < 0.05 and P < 0.01, respectively). No differences were found in the total M2-macrophage number in the homolateral TG from day 4 to day 8 after injury, but a significant decrease in M1/2 macrophages could be measured (from 7 ± 3 to 4 ± 2, P < 0.05) in favor of a modest increase in M1 subtype, suggesting a switch from M2 to M1 phenotype. 

As neuronal injury can cause retrograde neural degeneration, we checked immunohistochemical expression of activated caspase-3 in the TG sections as a marker of apoptosis (Fig. 4G). Indeed, the number of apoptotic cells was almost tripled in the TG homolateral to the alkali burn as opposed to the contralateral TG (Fig. 4H). Since these caspase-3⁺ cells were also CD45 negative (data not shown) and F4/80 (Fig. 4G) negative, we may exclude the possibility that these cells were bone marrow-derived and, more specifically, macrophage/microglia infiltrating cells. 

Increased Inflammatory Cytokine Expression in the Trigeminal Ganglion After Corneal Alkali Burn

Following our detection of infiltrating macrophages in the TG after corneal alkali burn, we investigated whether this might be associated with altered expression of proinflammatory cytokines. Indeed, alkali burn induced a strong upregulation of a panel of proinflammatory cytokines in the injured cornea on days 1, 4, and 8 in comparison to the control (day 0) and to the contralateral cornea (Supplementary Fig. S1A). In particular, IL-1β and VEGF-A were significantly (P < 0.001) upregulated on days 1, 4, and 8, while TNF-α was significantly increased on day 4. Interleukin-1β exhibited an early 24-hour peak expression (200-fold), which then slowly decreased; VEGF-A level was constantly upregulated over time (2-fold). Finally, TNF-α expression decreased on day 1 and then reached a peak on day 4 (3-fold). Similarly, the contralateral cornea showed a significant increase in IL-1β (on days 4 and 8, P < 0.01) and TNF-α (on day 4, P < 0.001) in comparison to the control. 

Intriguingly, proinflammatory cytokine expression was increased also in the TG homolateral to the damaged cornea (Fig. 5A). In particular, gene expression of IL-1β and VEGF-A was significantly upregulated in the homolateral TG on days 4 and 8 in comparison to the control (P < 0.01) and to the contralateral TG (P < 0.05); IL-1β expression in the TG was significantly higher than in the control as soon as 24 hours after corneal damage (5-fold). Tumor necrosis factor-alpha expression significantly increased on day 8 (8-fold) compared to control. Surprisingly, the contralateral TG showed a significant increase of IL-1β and VEGF-A on day 4, and of IL-1β and TNF-α on day 8, in comparison to controls. 

Figure 5

 

Upregulation of proinflammatory cytokines IL-1β, TNF-α, and VEGF-A in the trigeminal ganglion after corneal alkali burn and their downregulation following dexamethasone treatment. (A) The gene expression of IL-1β and VEGF-A was significantly upregulated in the homolateral (hl) TG at 1, 4, and 8 days post alkali burn in comparison to the control (day 0) and the contralateral (cl) TG; TNF-α mRNA level was significantly increased after 8 days in comparison to the control (day 0) and the contralateral (cl) TG. The contralateral TG showed an increase of the same cytokines in comparison to the control. (B) Anti-inflammatory, topical dexamethasone (DEXA) treatment of the alkali-burned corneas for 4 days significantly reduced the expression of IL-1β and VEGF-A in the homolateral TG in comparison to the untreated TG on day 4. Interleukin-1β expression was downregulated also in the contralateral TG. No difference was observed in TNF-α expression after treatment. Histograms represent mean values ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001 (n = 6).

Inflammatory Cytokine Expression Reduction in the Trigeminal Ganglion Following Anti-Inflammatory Dexamethasone Treatment of the Cornea

To further prove the existence of a cornea–TG inflammatory axis, we tested whether inhibition of inflammation in the cornea could be replicated in the TG. To this end, we treated the alkali-burned cornea with an anti-inflammatory corticosteroid-based ointment (0.2% dexamethasone). This treatment in the cornea significantly (P < 0.001) reduced the expression of IL-1β and TNF-α in comparison to the untreated cornea on day 4. Tumor necrosis factor-alpha expression was downregulated also in the contralateral cornea (P < 0.001) (Supplementary Fig. S1B). No difference was observed in VEGF-A expression after treatment. 

In the TG homolateral to the topical treatment, expression of IL-1β and VEGF-A was significantly (P < 0.001 and P < 0.01, respectively) reduced after dexamethasone treatment in comparison to the untreated mice on day 4 (Fig. 5B). Interestingly, the treatment significantly reduced IL-1β RNA levels also in the contralateral TG (P < 0.01). No difference was observed in TNF-α expression after treatment. 

Increased Expression of Substance P and NK-1R in the Trigeminal Ganglion

Substance P, the major peptide encoded by the TAC1 gene, is one of the sensory neurotransmitters released from the trigeminal nerve in the cornea. It is known that after alkali burn, Substance P is markedly induced in the cornea between days 1 and 5 (Son Y, et al. IOVS 2004;45:ARVO E-Abstract 1423 and Ferrari G, et al. IOVS 2014;55:ARVO E-Abstract 3243). Substance P is deeply implicated in neuroinflammation and hence represents an excellent candidate as a mediator of inflammation from the cornea to the TG. With regard to the hypothesis of a corneal–TG axis, we evaluated whether the expression of TAC1 and TAC1R (coding for the Substance P receptor, NK-1R) could be altered in the TG following chemical burn of the cornea. 

Indeed, alkali burn induced an upregulation of TAC1 and TAC1R in the homolateral TG (Fig. 6). In particular, TAC1 and TAC1R expression was significantly upregulated in the homolateral TG on day 4 in comparison to controls (P < 0.01) and to the contralateral TG (P < 0.05). Similarly, the contralateral TG showed a lower but significant increase of TAC1 and TAC1R on day 4 in comparison to the control. 

Figure 6

 

Upregulation of TAC1 and TAC1R in the trigeminal ganglion after corneal alkali burn. The gene expression of TAC1 and TAC1R was significantly upregulated in the homolateral (hl) TG 4 days after alkali burn in comparison to the control (day 0) and the contralateral (cl) TG. The contralateral TG showed an increase of TAC1 and TAC1R in comparison to the control. Histograms represent mean values ± SEM; *P < 0.05, **P < 0.01 (n = 6).

Increased Inflammatory Cytokine Expression in the Trigeminal Ganglion After Corneal Trephination

Since the alkali burn model is not specific for nerve ablation, we used trephination to selectively sever corneal nerves (Supplementary Fig. S2A). Beta-III tubulin immunostaining on corneal whole mounts confirmed the efficacy of this model to induce nerve disruption (Supplementary Fig. S2B). After 4 days, the sub-basal nerve fibers were not detected inside the trephined area. 

Real-time PCR analysis confirmed the increased expression of proinflammatory cytokines in the TG homolateral to the corneal trephination. Similarly to what we observed in the alkali burn model, there was a significant increase of IL-1β (P < 0.05), VEGF-A (P < 0.01), neuropeptide Substance P (TAC1, P < 0.05), and its receptor NK-1R (TAC1R, P < 0.01) transcripts in the homolateral TG in comparison to the controls (Supplementary Figs. S2C, S2D). TAC1R RNA levels increased in the contralateral TG in comparison to the healthy TG (P < 0.01). 

Discussion

In this paper, we show that an injury to the ocular surface (i.e., the cornea) is associated with inflammation in the brain (i.e., the TG). In particular, alkali burn of the cornea is followed by a rapid increase in the expression of proinflammatory cytokines, the neuropeptide Substance P, and its receptor NK-1R, and by the infiltration of T cells and macrophages and neural apoptosis in the TG. In addition, we confirmed the enhanced expression of these markers of neuroinflammation in a specific mouse model of corneal nerve injury (Supplementary Fig. S2), suggesting involvement of sensory nerves in the corneal–TG inflammatory axis. 

In our study, we found that the TG is populated with macrophages and T lymphocytes, similarly to what has been reported for other ganglia. ^16,17 Furthermore, we observed that the density of macrophages was considerably higher than that of T cells (Fig. 2A). The existing literature suggests that after peripheral nerve injury, macrophages and T lymphocytes infiltrate the injury site and also the corresponding ganglia where cytokines such as IL-1β and TNF-α are released and are thought to play a role in the genesis of neuropathic pain. ¹⁸ These findings are in agreement with our results (Fig. 5A, Supplementary Fig. S1A). Specifically, the increase of IL-1β and TNF-α observed in the TG after alkali burn may be explained by T-cell (Fig. 2D) and M1-“proinflammatory” macrophage infiltration (Fig. 5A). In this vein, TNF-α induces ectopic activity in nociceptive primary afferent fibers, ¹⁹ while IL-1β induces the expression of cyclooxygenase-2 leading to prostanoid release in the central nervous system (CNS). ²⁰  

Increasing evidence, however, suggests that the role of macrophages in the injured CNS may be more complex. More specifically, it is suggested that they may be beneficial or destructive depending on their activation into one of at least two phenotypes, M1 or M2. For example, macrophage-derived molecules such as TNF-α, ²¹ IL-1β, ²² and MHC II ²³ have been shown to promote oligodendrocyte progenitor cells directly or indirectly. On one hand, M1 macrophages release proinflammatory mediators that contribute to neuronal dysfunction and cell death. ^24–26 On the other hand, M2 macrophages show an enhanced phagocytic activity and have anti-inflammatory properties. ²⁷ Following peripheral nerve damage, the infiltrating M2 macrophages facilitate recovery by removing debris of degenerating axons and dead cells, hence stimulating tissue regeneration. ²⁸ These findings are in accordance with our results (Fig. 4F), showing that the total macrophage number increased in the TG. M2 macrophages were indeed prevalent over M1 macrophages in normal conditions and following corneal damage. 

Macrophages are able to respond dynamically to nerve injury, switching from a transient M2 phenotype to a sustained M1 phenotype. ²⁷ During this transition, macrophages express both M1 and M2 markers. Indeed, we detected a subtype of macrophages expressing both M1 and M2 markers, which progressively decreased in number after the injury. The M1 subtype slightly increased over time, suggesting a transition from M2 to M1 phenotype. A similar M2-to-M1 shift has already been reported, ^27,29,30 suggesting that this change of phenotype may be a common pathologic mechanism in various CNS injuries. 

Finally, we observed progressive leukocyte infiltration, which proceeded from the anterior (i.e., proximal to the cornea) to the posterior part of the TG. This could be explained given that cell bodies of the ophthalmic branch of the trigeminal nerve are found anteriorly within the TG. 

Our data and substantial literature suggest that Substance P, a neuropeptide well known for its roles in neuroinflammation and pain processing, may represent an ideal candidate involved in mediating inflammatory progression from the cornea to the TG. 

Substance P is considered a major player in neurogenic inflammation. ³¹ It is produced in the soma of a subset of sensory neurons mainly localized in the dorsal root and TGs. Substance P is then transported to both central and peripheral processes of these cells. ^32,33 It is well documented that the stimulation and/or injury of sensory peripheral terminals of these neurons results in the peripheral release of Substance P. ^33,34 Lymphocytes, granulocytes, and macrophages have receptors for Substance P (NK-1R), and they can be stimulated by Substance P to produce proinflammatory mediators and cytokines such as IL-1β, IL-6, and TNF-α. ³²  

Substance P also acts as a potent chemotactic agent itself, further recruiting inflammatory cells. All these molecular events ultimately sustain the synthesis and release of new Substance P, progressively increasing its concentration in inflamed tissues. Interestingly, Substance P has been shown to extend its effect also contralateral to the primary injury site. This is particularly true in the eye, as previous data indicate that Substance P allows propagation of inflammation from the damaged eye to the contralateral eye following a retinal laser burn. ³⁵ Those authors found an increase of IL-1β, IL-6, TNF-α, and NK-1R in both eyes following a unilateral injury. Similarly, we found an increase of proinflammatory cytokines (IL-1β, VEGF-A, and TNF-α), Substance P, and its receptor NK-1R in the TG, which was predominant in the homolateral side, after corneal alkali burn/trephination. Interestingly, we found a very high increase of IL-1β expression in both alkali-burned cornea and homolateral ganglion. It is known that IL-1β may represent an intermediate proinflammatory mediator. Interleukin-1β production is enhanced in crushed peripheral nerve, ³⁶ and may cause an upregulation of neuropeptides such as Substance P and growth factors in neurons. ³⁷ Moreover, IL-1β is able to enhance the axonal transport of Substance P in sensory neurons ³⁸ and its release. ³⁹  

The role of nerve-secreted peptides such as Substance P and nerve growth factor has been described in the setting of corneal infection ⁴⁰ and inflammation. ⁴¹ However, mechanisms regulating nerve secretion of these peptides into the cornea have not been clarified yet. It is tempting to speculate that corneal nerves may secrete these mediators in response to specific modifications of the corneal milieu, such as inflammation. In this vein, corneal nerves may provide a channel for the transmission of inflammation to the TG. This would in turn allow secretion in the cornea of proinflammatory neuropeptides. In summary, we propose that a corneal–TG axis exists and is represented anatomically by trigeminal/corneal nerves and functionally by specific neurotransmitters, such as Substance P. 

Although the final implications of this axis deserve further study, it should be noted that it appears to work bidirectionally. In fact, we previously reported that an injury involving the TG induces inflammatory cell infiltration in the cornea. ⁷ Our findings may provide better understanding of the pathophysiology of corneal diseases from “ocular only” to neurological disorders (i.e., involvement the TG). Indeed, the vast majority of corneal diseases are considered merely ocular, although neurological involvement such as chronic pain, nausea, and vomiting can occur. 

Surprisingly, we found that proinflammatory cytokines were increased not only in the TG homolateral to the injury, but also in the contralateral one, although no significant inflammatory cell infiltration was detected in the latter. In this regard, there is increasing evidence that unilateral nerve injury evokes contralateral responses, but the underlying mechanisms are largely unknown. The bilateral effects after unilateral nerve damage have recently been reported in a mouse model of neurotrophic keratopathy. ⁴² Yamaguchi et al. ⁴² observed some corneal epithelial defects followed by corneal neovascularization and a sub-basal corneal nerve reduction in the untreated contralateral eyes. Bilateral nerve alteration has been also found in patients with unilateral herpes simplex keratitis ⁴³ and herpes zoster ophthalmicus. ⁴⁴ These midline-crossing effects were also previously documented in contralateral undamaged neurons after peripheral nerve lesions. ^45,46 Interestingly, the release of inflammatory mediators from neurons, such as TNF-α, IL-1β, Substance P, and others, has been suggested to mediate these bilateral effects. ^45–47 Intriguingly, we found these inflammatory mediators increased also in the TG. Corneal innervation is generally considered to travel along a unilateral nerve pathway ⁴⁸ ; however, impairment of contralateral nerve function has been noticed in monocular herpetic keratitis. ⁴⁹ This was further confirmed by others ⁴⁴ and is likely mediated by corneal nerves, although the precise mechanism is still unclear. Projections of trigeminal neurons innervating bilateral brainstem areas and crossing between trigeminal nuclei of both sides have been observed, ^50,51 suggesting anatomical structures involved in these midline-crossing effects. Additionally, it has been proposed that Substance P may mediate loss of immune privilege to the contralateral eye following a retinal laser burn. ³⁵ For this reason, we used healthy animals as controls. Finally, contralateral involvement induced by Substance P and other neuropeptides has also been described in monoarthritis. ⁵²  

With regard to the bilateral inflammatory extension that we observed, it is well known that responses to contralateral injuries are usually qualitatively similar but smaller in magnitude and have a shorter time course compared to homolateral changes. ⁴⁵ This is in accordance with our results for cytokine expression. Specifically, IL-1β reaches highest expression level in the alkali burn cornea 24 hours after injury, in both ganglia, on day 4 (but with a lower expression in the contralateral one) and, finally, in the contralateral cornea on day 8. Tumor necrosis factor-alpha reached highest expression in the alkali burn cornea on day 4 but in the contralateral cornea and ganglion on day 8. Finally, VEGF-A RNA levels were persistently high in both corneas from day 1 on while peaking later on day 4 in both ganglia; in both cases, the contralateral changes were always less than the homolateral ones. 

Finally, we observed a bilateral effect even after topical treatment with dexamethasone. Indeed, topical treatment in the alkali-burned eye reduced inflammatory markers also in the contralateral cornea. It must be pointed out that the effects of corticosteroids on corneal nerves are mixed. Their use has been associated with both delayed nerve regeneration ^53,54 and enhanced peripheral nerve repair. ⁵⁵ Specifically, topical application of corticosteroid in the cornea delays re-epithelization and hence reinnervation. Although we cannot exclude some systemic effect of unilateral dexamethasone absorption, involvement of corneal nerves should be considered. In fact, the only anatomical structures directly connecting the corneal surface with the TG are represented by sensory nerves. We observed a more pronounced reduction of IL-1β RNA levels in the ganglion homolateral to the treatment in comparison to the contralateral TG (approximately 14 times versus 2 times). A similar pattern of reduction was observed for VEGF-A (2.5 times reduction in the homolateral TG versus no reduction in the contralateral TG). This suggests that although a systemic effect may indeed occur, homolateral topical application plays a relevant role. With regard to the dexamethasone treatment, we did not observe reduction of VEGF-A expression levels in the cornea after topical administration. Interestingly, a 4-day treatment course did not reduce the growth of neovessels observed at slit-lamp examination. Previous literature suggests that topical dexamethasone in the alkali burn model may need a longer time to be effective against corneal neovascularization (possibly due to delayed VEGF-A reduction). ^56,57 This hypothesis is supported by the finding that bevacizumab, a selective anti-VEGF inhibitor, reduced alkali burn–induced corneal neovascularization only from day 5 onward, while VEGF-A level remained high for the first 4 days. ⁵⁷ Moreover, it should also be noted that the alkali burn model induces a stronger increase of VEGF-A levels in comparison to other mouse models of corneal injury. ⁵⁸  

In this study, we demonstrated that MRI could be used to track inflammatory cell infiltration in vivo along the TG. More specifically, we found that the significant uptake of USPIO was related to M2 macrophages, which have a high phagocytic capacity. This noninvasive imaging modality coupled with USPIO contrast, which is currently being evaluated in phase 1 clinical trials, could be translated into the clinic by using advanced MRI. This technology, although not available in many primary care centers, is becoming increasingly accessible. It is well recognized that MRI could image many of the pathological processes involving the trigeminal nerve with standard imaging protocols. ^59,60 Advanced MRI including high magnetic field and three-dimensional fast imaging allowed improved image resolution and thus accurate observation of trigeminal structures. ^61,62 Furthermore, brain activation during trigeminal neuralgia could be assessed with functional MRI methods in a complementary manner to explore the trigeminal involvement after corneal injury. ⁶³  

In summary, our study shows firstly that injuries to the cornea induce pathological changes far beyond the eye, specifically in the brain, and secondly that a monolateral injury also affects the contralateral eye and TG. We anticipate that our findings may have significant clinical implications for the pathophysiology and finally the treatment of ocular surface diseases, since corneal nerve damage or loss occurs in more than 100 million people worldwide every year. ¹⁰ Additionally, trigeminal neuropathic pain is a common and feared complication of corneal nerve injury as it negatively affects quality of life. In this vein, it is known that IL-1β ⁶⁴ and TNF-α ⁶⁵ play a pivotal role in the generation and/or maintenance of chronic neuropathic pain. Intriguingly, we found a significant increase of both these cytokines in the TG following corneal nerve injury. Finally, based on our in vivo imaging results, we propose that noninvasive brain MRI study of patients affected with corneal diseases could be helpful in better understanding the pathophysiology of ocular surface disorders. 

Supplementary Materials

pdf S1 

pdf S2 

Acknowledgments

We thank Claire Corot, PhD (Laboratoire Guerbet, Paris), for providing the USPIO (P904). 

Disclosure: G. Ferrari, None; F. Bignami, None; C. Giacomini, None; E. Capitolo, None; G. Comi, None; L. Chaabane, None; P. Rama, None 

Footnotes

 GF and FB contributed equally to the work presented here and should therefore be regarded as equivalent authors.

Footnotes

 LC and PR are joint senior authors.