Ocular Surface Pathology in Patients Suffering From Mercury Intoxication Is Consistent With Neurogenic Dry Eye Disease

Purpose: To report ocular surface pathology of patients with acute/subacute mercury intoxication. Methods: Male workers intoxicated with inorganic mercury were examined for dry eye (DE)-related symptoms. Examinations included Ocular Surface Disease Index questionnaire; tear osmolarity, tear break-up time (T-BUT) and production; mechanical and thermal corneal sensitivity; corneal nerve and dendritic cell density analysis; and analysis of 23 tear cytokines. Results: Most patients, 63.6%, had severe DE-related symptoms. Tear osmolarity was elevated in 83.4%, and T-BUT was low in 22.7% of patients. Tear production and tear lysozyme concentration were low in 13.6% and 27.3% of cases, respectively. Corneal sensitivity thresholds for mechanical, heat and cold stimuli were higher than controls. Densities of nerves, nerve branching, and dendritic cells were lower than in controls. Patient tear levels of IL-12p70, IL1-RA, RANTES, and VEGF were increased, whereas EGF, IL-6, and IP-10/CXCL10 were decreased. Based on cytokine levels, two clusters of patients were identied. Cluster 2 patients had signicantly increased tear levels of 18 cytokines, decreased tear lysozyme, lower nerve branching density, fewer dendritic cells, and higher urine mercury levels. Conclusions: Mercury poisoning produced previously undescribed ocular surface pathology, similar to neurogenic inammatory type of DE and different from the more common DE subtypes. al. Low pathway alterations and neurological toxic effects after in


Introduction
Mercury is a metallic element with a high potential for toxicity. Several public health disasters by mercury intoxication (also known as "mercury poisoning") have been reported, the most well-known of these occurred in Japan in 1956 and was known as Minamata disease. Afterwards, the Minamata Convention on Mercury made the world aware of the environmental and public health issues that mercury pollution represented and that are still present. [1] The most frequent human exposure to water-soluble forms of mercury, such as mercuric chloride or methylmercury, is caused by ingestion of any form of mercury, e.g., contaminated sh, or by inhalation of mercury vapor as an occupational exposure, including coal burning and mining, especially of mercury and gold. [1] Exposure to different forms of mercury has been associated with adverse health effects. [2] Intoxication due to inhalation of mercury vapor, a form of inorganic mercury, produces the most harmful effects, as up to 80% of the inhaled mercury is absorbed and rapidly oxidized to other forms. Oxidized mercury vapor becomes lipid soluble, so the potential exists for bioaccumulation in the renal cortex, liver, and especially in the brain, where it has been estimated that the half-life can be as long as 20 years. [3] Even though the principal target organ of mercury vapor is the brain, functional degradation of peripheral nerves, and of the renal, immune, endocrine, and muscle systems, and several types of dermatitis have been described. [4] Numerous ophthalmic ndings due to mercury toxicity and its action on the retina and optic nerve have been reported for both chronic and acute exposures. These include decreased night vision, decreased color vision and contrast sensitivity, central visual impairment, progressive visual eld constriction, and optic atrophy. [5] Less frequent symptoms and signs such as photophobia, blepharospasm, nystagmus, and mercury deposits on the anterior capsule of the lens (mercurialentis) and corneal stroma have also been reported in some cases of chronic intoxication. [5] Ocular surface pathology due to acute and chronic exposure to mercury exposure has only been vaguely described in experimental models. [6][7][8] However, there are no published studies reporting human ocular surface effects.
At the end of 2012, 49 workers in Northern Spain were accidentally exposed to dangerous levels of mercury vapor. Blood and urine levels of mercury were above the recommended biological limits for occupational exposure, and acute and subacute (acute/subacute) mercury vapor intoxication was con rmed. Most of these patients were referred to the University of Valladolid, Spain, between 1 and 2 years later, and those with ophthalmic complaints were referred to the Institute of Applied Ophthalmobiology (IOBA), University of Valladolid, Valladolid, Spain. In this study, we aimed to nd the origin of the patients' chronic ocular surface-related symptoms. Based on the known mechanisms of mercury-based pathogenesis, we hypothesized the presence of neurogenic in ammation of the ocular surface. Therefore in addition to the usual clinical tests, to assess the presence of a neurogenic component in the mercury-dependent ocular surface pathology, we added tests for corneal mechanical and temperature sensitivity, corneal nerve morphology, and in ammatory molecules in the tears. Methods

Patients
This study was approved by the Institutional Review Board and by the Ethics Committee of the Valladolid University Clinical Hospital. All enrolled patients were informed of the aims of the study, and their written consent was obtained. All research was performed in accordance with relevant regulations and in accordance with the Declaration of Helsinki.
Acute/subacute mercury vapor intoxication occurred in 49 male workers accidentally exposed to dangerous mercury vapor levels for 14 consecutive days in a metal manufacturing plant in Northern Spain between November 19 and December 2, 2012. In the days after exposure, many of the workers reported physical complaints, mainly asthenia, headache, epigastric and abdominal pain, cough, bitter taste, dental pain, and gum in ammation and bleeding. Some of the workers also had ocular symptoms such as irritation, redness, burning, foreign body sensation, and light photophobia. Blood and urine mercury levels were measured during the second week of the exposure, and were above the recommended biological limits for occupational exposure, reaching maximum range levels between 252.62 and 507.47 µg/L in blood (normal, < 10 µg/L) and between 93.61 and 245.57 µg/g creatinine in urine (normal, < 30 µg/g creatinine). Some weeks before occupational exposure, random urine samples (n = 17) detected mercury levels below 3 µg/g creatinine in those workers. The diagnosis of acute/subacute intoxication with mercury vapor was con rmed by the post-exposure clinical analyses.
Between September 2013 and December 2014, 44 of the original 49 intoxicated patients were referred to the Clinical Toxicology Unit of the Institute of Medical Sciences at the University of Valladolid for an independent evaluation. Of those, 29 patients had ocular symptomatology and were immediately referred to IOBA and evaluated for enrollment. All referred patients had vision and retinal-related symptoms, which have been studied by our colleagues, and ocular surface-related symptoms.
All of the referred patients received medical care and were evaluated for inclusion in this study. The only inclusion criterion was that patients had to have reported their ocular surface symptoms after the mercury intoxication. The exclusion criteria were (1) pre-existing ocular surface symptoms; (2) use of arti cial tears or any other topical medication before mercury intoxication; (3) use of any topical medication other than arti cial tears 4 weeks before enrollment or 3 months before enrollment in the case of either topical cyclosporine or tacrolimus; (4) use of contact lenses in the previous 3 months; (5) any other ocular surface disease; or (6) any previous ocular surgery.

Clinical evaluation and tear sample collection
Both eyes were evaluated between 10 am and 2 pm. The temperature of the examination room was always set at 19-21ºC, and the relative humidity was between 50-60%. Clinical evaluation and subjective tests were always performed by the same examiner in a single visit and in the sequence detailed below.

Tear sample collection for molecular analysis
We followed our previous protocol in which unstimulated basal tear samples were collected non-traumatically from the external canthus. [10,11] One microliter of tear sample was collected with a glass capillary micropipette (Drummond, Broomall, PA, USA). Each sample was then diluted 1:10 in ice-cold Cytokine Assay Buffer (Milliplex, Millipore Merck Life Science SLU, Madrid, Spain). Tear samples were kept cold (4 °C) during collection and then stored at − 80 °C until assayed.

Tear osmolarity
Ten minutes after the tear sample collection, the osmolarity was assessed by the TearLab osmometer (TearLab Corporation, San Diego, CA, USA) in a 50-nL tear sample collected from the external canthus. Values above 308 mOsm/L were considered abnormal. [12] 2.2.4. Conjunctival bulbar hyperemia Ten minutes after the tear sample collection for osmometry, the nasal and temporal conjunctivas were assessed independently with a slit-lamp biomicroscope based on the Efron scale (0-4 score). The nal score was the average of the nasal and temporal values. [12] 2.2.5. Tear break-up time (T-BUT) Five minutes after slit-lamp biomicroscopy, tear stability was assessed by T-BUT. After instillation of 5 µL of 2% sodium uorescein, the time between the last of three blinks and the appearance of the rst dry spot was measured three times, and the mean value was recorded. Values below 7 seconds are currently considered abnormal. [12] 2.2.6. Ocular surface integrity Approximately 2 minutes after uorescein instillation for T-BUT, corneal and conjunctival integrity were evaluated with uorescein and lissamine green staining, respectively. The Oxford scheme (0-5 score) for grading the staining of both areas was used. [12] 2.2.7. Tear production Five minutes after the previous test, tear production was assessed with two different tests: tear lysozyme level assay and Schirmer's test without topical anesthesia. The tear lysozyme concentration test was performed as previously detailed, and values of less than 1,000 µg/ml were considered indicative of low tear production. [13] Then, after 5 minutes, the Schirmer test was performed by placing a sterile strip (I-DEW tear strips, Entod Research Cell UK, Ltd, London, UK) in the lateral canthus of the inferior lid margin, and the length of wetting was measured after 5 min. Results below 5-mm length were considered abnormal. [12] 2.3. Corneal sensitivity Twenty minutes after, corneal sensitivity was measured with a prototype Belmonte's non-contact gas esthesiometer as previously reported by our group. [14,15] The corneal threshold for mechanical and thermal (cold and heat) sensitivities was determined in the central cornea using the method of levels. Three-second air pulses of adjustable ow rate and temperature were applied to the center of the cornea for determining corneal sensitivity thresholds. Mechanical stimulation was always rst and consisted of a series of variable ows of medicinal air (0-200 mL/min). Thermal thresholds were determined by heating or cooling the air to produce changes in basal corneal temperature of ± 0.1ºC, with a 10 mL/min ow below the mechanical threshold. Immediately after each stimulation pulse, the subject was asked to report the presence or absence of sensation. The order of heat and cold threshold measurement was randomized. Results were compared with a separate, but similarly aged (38.5 ± 8.2 years, p = 0.1622) control group of 20 healthy males from our les. [14,15,16] Each subject in the control group had an OSDI ≤ 12, corneal staining < 1 (Oxford scale), T-BUT ≥ 7 seconds, and a Schirmer test ≥ 5 mm/5 min.

In vivo confocal microscopy (IVCM)
Ten minutes after measuring sensitivity, and after applying topical anesthesia (0.1% tetracaine and 0.4% oxibuprocaine, Anestésico Doble Colirio; Alcon-Cusí, El Masnou, Barcelona, Spain), laser scanning IVCM of the cornea was performed using the Rostock cornea module of the Heidelberg Retina Tomograph 3 (Heidelberg Engineering GmbH, Heidelberg, Germany). At least three good quality, non-overlapping images from the sub-basal nerve plexus of the central cornea were obtained using sequence and/or volume scans were used for the analysis. Two masked observers analyzed the following: (1) nerve morphology parameters of density, length, branching density, and grade of tortuosity; (2) density of dendritic cells; (3) presence of neuromas; and (4) re ectivity from the confocal images, as an index of optic densitometry or transparency of cornea. [17] The mean value between the two observers for each parameter was computed for statistical analysis.
Nerve density (n/mm 2 ) and length (mm/mm 2 ) were measured using the plugin NeuronJ (http://www.imagescience.org/meijering/software/neuronj/) from the ImageJ software (https://ima gej.nih.gov/ij/), which allows semi-automated tracing of nerve bers and provides quanti cation. The number of nerve branch points and dendritic cells (identi ed in the sub-basal nerve plexus by their distinctive features, i.e., bright cell bodies with dendritic form structures), were manually determined using the multipoint tool of the ImageJ software, and the densities calculated (n/mm 2 ) as described in a previous study. [18] The grade of nerve tortuosity was evaluated according to the scale (0-4) reported by Oliveira-Soto and Efron [19] for main nerves. The histogram of each image based on the ImageJ plugin was used to obtain the mean re ectivity or optic densitometry. [17] These parameters were compared with well-established values for normal corneas and performed with the same type of confocal microscope. Speci cally, we used data from Giannacare et al. [20] for nerve length, and from our group for nerve density, density of nerve branches, density of dendritic cells, [18] and nerve tortuosity and re ectivity. [17] 2.5. Analysis of tear cytokine concentrations , tumor necrosis factor (TNF)-α, epidermal growth factor (EGF), and vascular endothelial growth factor (VEGF) were measured simultaneously with a customized 23-plex SPR assay (SPR591 HCYTO-60K, 23X-Milliplex; Millipore). We chose these molecules based on (1) altered serum values in mercury-intoxicated patients, [2,21,22] (2) altered concentrations found in DE tears by our group and others, [10,11,13,15,[23][24][25][26][27][28][29][30] and (3) in patients undergoing refractive surgery. [16,[31][32][33] The samples were analyzed following the manufacturer's low volume sample protocol that only uses 10 µl of sample/standards per assay, as previously described. 10 Data were stored and analyzed with the ''Bead View Software'' (Upstate-Millipore Corporation, Watford, UK). The minimum detectable concentration, based on manufacturer speci cations, was 1.2 pg/ml. Molecules that were detected in less than 30% of the samples were not statistically analyzed any further. Results were compared with a similarly aged (42.4 ± 8.1 years; p = 0.1523) control group of 22 healthy males from our les who met the same criteria as the control group for corneal sensitivity. [11,13,15,28,29,30] 2.6. Statistical analysis Statistical analysis was performed using R software version 3.4.1 (R Foundation for Statistical Computing, Vienna, Austria). Signi cance level was set at 5%. Quantitative data were summarized as means and standard deviations. Ordinal values were described using medians and interquartile ranges [IQR], unless otherwise speci ed in the text. The normality assumption was checked by the Shapiro-Wilk test.
Data from the study group were compared to the control groups. Student's t-tests for two independent samples were used to compare differences between mean values. Levene's test was used to check homogeneity of variance, and Welch's test was used when this assumption was not valid. When normality assumptions were not supported, the nonparametric alternative, Mann-Whitney U test, was performed.
For tear cytokine analysis values out of range, the values were imputed by the regression on order statistics method. This technique performs a regression to impute low values assuming a log-normal distribution. The detection rates in the study and control groups were compared using equality of two proportions test. Cytokine expression data were transformed using the logarithmic base 2 scale. Expression levels in the study group were compared with levels in a control group from our database. In addition, principal component and hierarchical agglomerative cluster analyses were used to explore correlation patterns among cytokine levels in the study group. To facilitate the interpretation of the clustering result, a pro le analysis was conducted, testing the differences among clusters by the same methodology as the one used to compare the study and control groups.

Results
All of the 44 patients examined at the University Clinical Toxicology Unit had erethism mercurialis and peripheral nervous system alterations that were con rmed by electrophysiology. Twenty-nine of the patients had visual complaints and were consequently referred to IOBA. Preliminary observations reported that the retinal and neurophthalmic pathological ndings in these patients were as follows in decreasing order of frequency: (1) decreased contrast sensitivity, (2)  Of the 29 patients (65.9% of the original 44 patients referred to the Toxicology Department) with visual complaints, one was evaluated but excluded due to previous corneal refractive surgery in both eyes, and six others were clinically evaluated but declined to participate in this study. Therefore, a total of 22 male patients were nally included in this study, with a mean age of 42.0 ± 7.6 (range, 28-56) years. All of them were taking psychotropic drugs for their erethism mercurialis.

Clinical Tests
Results of the clinical tests are shown in Table 1. OSDI values were abnormal (> 12) in all patients except in one. The majority of patients, 14 (63.6%), had severe dry eye (DE)-related symptoms, 4 (18.18%) had moderate symptoms, and 3 (16.6%) had mild symptomatology. Conjunctival hyperemia and ocular surface integrity (corneal uorescein and conjunctival lissamine green staining) ndings were unremarkable. T-BUT was under normal values in 5 (22.7%) patients. Based on the Schirmer test, tear production was low in 3 (13.6%) patients, and based on tear lysozyme concentrations, it was low in 6 (27.3%) patients.
However, tear osmolarity was abnormally elevated in 83.4% of the patients.

Cornea sensitivity
Mechanical threshold and thermal thresholds for heat and cold were assessed in 21 patients (42.0 ± 7.6 years old) (one patient refused to have this test performed) of the 22 patients (Tables 1, 2) and in 20 control subjects ( Table 2). All of the sensitivity thresholds were signi cantly higher in the mercury-intoxicated patients, indicating that their corneal sensitivity was decreased  Table 3, and Table 4 presents comparisons with the control values.  Mercury-intoxicated patients had signi cantly lower nerve density and nerve branch density than did the controls ( Table 4). Density of dendritic cells in corneal stroma was also decreased in the patients compared to the control subjects. Neuromas were absent in all patients. Nerve length, nerve tortuosity, and re ectivity were not signi cantly different from controls. Representative images of a patient and a healthy control subject from our les are shown in Fig. 1.

Analysis of tear cytokine concentrations
The detection percentage and concentration for each tear cytokine is shown in Table 5. Eotaxin, IL-10, IL-4, and MIP-1α were not statistically analyzed any further due to the very low percentage of detection (< 30%). For the remaining cytokines, there were no differences in the percentage of detection between patients and controls.    For some cytokines, i.e., IL-12p70, IL-1RA, RANTES, and VEGF, the tear concentrations were signi cantly higher in the patients than in the control subjects (Table 5, Fig. 2). However, for other cytokines, i.e., EGF, IL-6, and IP-10, the tear concentrations were signi cantly lower in the patients compared to the control subjects.
To further explore correlation patterns among tear cytokine levels in the patient samples, principal component analysis (PCA) and hierarchical agglomerative cluster analysis were performed. To accommodate much of the variance in the primary dataset, PCA was used to build a few independent principal components (PC) based on interrelated levels of the 23 cytokines. In this case, two components explained 81.6% of the sample variability, suggesting that there were two principal components (PC1 and PC2) associated with the tear cytokine levels. PC1 showed high loadings on the levels of RANTES, TNF-α, IFN-γ, IL-12p70, IL-5, IL-2, IL-1β, IL-17A, IL-6, VEGF, IL-13, fractalkine, and IL-9. However, PC2 was more correlated with IP-10, IL-8, IL-1RA, EGF, MMP-9, MCP-1, eotaxin, and IL-10 levels.

The comparison of clinical parameter values between patients in each Cluster
*P value corresponding to comparison of concentration values between patient and control groups.
Signi cant P values (P < 0.05) are denoted in bold.

Discussion
In this study, we use a wide variety of techniques to describe for the rst time the chronic ocular surface pathology caused by acute/subacute mercury poisoning in workers accidentally exposed to toxic doses of mercury. Brie y, we showed that most patients were highly symptomatic and had increased tear osmolarity, corneal hypoesthesia, altered corneal sub-basal nerve and dendritic cell parameters, and altered tear levels of some in ammation-related cytokines. We concluded that the pathology encountered 1-2 years after the acute/subacute event is consistent with a neurogenic-based DE disease that was more severe in patients with higher urine levels of mercury.
The chief target organ of mercury vapor is the brain, where it causes apoptosis and ischemia of nerve bers. [4] The eye and visual pathways are especially susceptible to neurologically-driven diseases, and the ocular effects of poisoning due to mercury exposure are not unexpected because of the extraordinarily abundant innervation of the eye. This is especially true in the cornea, the most highly innervated tissue in the whole human body. This innervation is sensitive and is delivered by the ophthalmic branch of the trigeminal or V cranial nerve. [34] Because the damage caused by mercury poisoning could target this rich innervation, we evaluated corneal sensitivity and the morphology of the sub-basal corneal nerves by non-contact esthesiometry and IVCM, respectively. Both techniques are minimally invasive, and although not regularly performed in the clinical setting, they can provide invaluable information about some ocular surface diseases. We have accumulated experience with these techniques in contact lens-related discomfort [18] and stem cell therapy for corneal pathology. [35] Mercury poisoning, which could be responsible for the neurotoxicity and subsequent damage to the corneal nerves, could also be why the vast majority of the patients had DE-related symptoms, most of which were strongly experienced. Aside from changes in some tear cytokines (as discussed below), there were no signs of alteration in tear production and/or tear quality that can cause epithelial damage to the ocular surface as would be typical in tear-de cient and/or evaporative DE. In fact, a disparity between signs visualized with the slit lamp and symptoms is one of the most striking aspects of DE disease and has been reported in many types of DE patients. [12,34,36] This is especially true after corneal refractive surgery (the so-called "pain without stain") in which there is an unavoidable lesion to the corneal nerves as part of the required laser treatment. [37] In post-refractive surgery patients, and most likely in our mercury-intoxicated patients, neurogenic in ammation due to corneal nerve damage results in the release of the in ammatory mediators. [37] This in ammation could cause the patients to have DE symptoms without manifestation of compromised tear production, and therefore not causing an obvious ocular surface integrity problem. It is not clear why corneal nerve damage and hypoesthesia caused by refractive surgery or poisons like mercury develop into neurogenic DE without epithelial involvement. This problem is especially confounding because other deleterious causes, e.g., neurotrophic keratitis, also have corneal nerve damage and reduction of corneal sensation but result in epithelial damage. We hypothesize that neurogenic DE represents either another stage or another cause of neurotrophic keratitis. Clearly, more research is needed to better understand these conditions at the molecular and genetic levels. [38] Tear osmolarity was elevated in 19 of the 22 patients. Although tear hyperosmolarity has been implicated in the pathogenesis of DE, [12] there is no obvious explanation as to why the osmolarity was high in these patients, especially since there was no accompanying damage to the integrity of the ocular surface. Consistent with our ndings, Yi et al. [39] reported a signi cant positive correlation between tear osmolarity and ocular symptoms, including cold sensitivity, foreign body sensation, and light sensitivity; however in our patients, T-BUT, corneal staining, eyelid hyperemia, and tear secretion volume were not signi cantly correlated with tear osmolarity. Similarly, Gjerdrum et al. [40] also found tear hyperosmolarity with normal tear production in patients after nerve alteration caused by corneal laser surgery.
Using Belmonte's gas esthesiometer to measure corneal sensitivity thresholds, we found an increase in the mechanical threshold and in the heat and cold thermal thresholds in the mercury-intoxicated patients. This means that the overall corneal sensitivity was diminished. Benitez del Castillo et al. [36] found similar results for mechanical and thermal sensitivities in Sjögren syndrome-associated DE disease. Likewise, Bourcier et al. [41] reported corneal hypoesthesia with mechanical and thermal stimuli in a more mixed sample of DE patients.
The changes in corneal sensitivity are probably due to the nerve damage that we detected by IVCM. Corneal nerves not only protect the ocular surface through the mechanism of sensation, but they also release trophic factors that regulate wound healing, epithelial integrity, and cell proliferation. [19,34,42] Thus, nerve damage caused by mercury intoxication could be responsible for an alteration in neuronal stimulation and a delay in the transmission of nerve impulses of the affected bers. This would explain the decrease in mechanical and thermal sensitivity that we found.
Regarding nerve morphology, corneal nerve density and nerve branch density were signi cantly lower in the mercury-intoxicated patients, and these changes were associated with higher levels of mercury in the urine of Cluster 2 patients. These results are in agreement with most studies in which there was a signi cant reduction in the sub-basal nerve density in DE patients compared with controls. [23,43] There are, however, two studies [44,45] that show no difference in sub-basal nerve density, but instead, the DE patients had abnormal nerve morphology. Finally, one study of patients with aqueous-de cient DE disease found increased sub-basal nerve density, suggesting the possibility of corneal nerve regeneration in this form of DE. [46] In general, regenerative activity is manifested by nerve branches from endbulbs, and in our patients, the density of branches was diminished. All of these ndings support the notion that mercury intoxication adversely affects nerve function and also the capacity for nerve regeneration.
The density of dendritic cells in the corneal stroma of our mercury-intoxicated patients was decreased. Dendritic cells are in contact with the sensory nerve bers, and play an important role in corneal homeostasis. [23,47,48] Elevated density of dendritic cells is a common nding in in ammatory disorders such as DE disease, [23] after refractive surgery, [49] in diabetic neuropathy, [49] and in infectious keratitis. [23] Consequently, we initially expected to nd a higher density of these cells in the cornea of our patients. However, where more centralized nerve damage occurs, such as in patients with bromyalgia syndrome where the corneal sub-basal nerve plexus is also damaged, corneal dendritic cell density is similaly decreased. [47] In animals, after trigeminal denervation, there is a depletion of dendritic cells, and corneal sensitivity is signi cantly reduced, delaying corneal recovery during wound healing. [48] So the decrease in the density of dendritic cells in our mercury-intoxicated patients could be due to the damage we demonstrated in corneal sub-basal nerve plexus.
Lastly, we found alterations in some tear cytokine levels in the mercury-intoxicated patients. Damage caused by mercury intoxication could be responsible for the alterations in nerve stimulation and impulse transmission. It could also cause nerve in ammation, resulting in liberation of several in ammatory cytokines. Indeed, neuro-in ammation is one of the main pathways of methyl mercury-induced central nervous system impairment. [2] Furthermore, in addition to affecting the nervous system, there is accumulating evidence that exposure to mercury alters immunomodulation, although with differences in the mechanism of action depending on the speci c form of mercury (inorganic or organic), the species, and even the cell type or tissue. [2,50] Also, other cell types apart from nerves, such as ocular epithelial and/or immune cells, could participate in the ocular surface in ammatory response to mercury exposure. These responses are related to interactions of metals, such as mercury, with electrophilic groups that are not solely restricted to the central nervous system, but are also ubiquitously present in several systems and organs. [2,6,7,50] Other studies have already shown tear molecule alterations in several ocular pathologies. [10,11,13,15,[23][24][25][26][27][28][29][30] Additionally, some studies reported alteration of tear cytokine levels after corneal refractive surgery. [16,[31][32][33] While there are several published studies regarding serum and/or tissue cytokine/chemokine levels or gene expression in mercury-intoxicated patients, [2,21,22] to our knowledge, this study is the rst to address tear cytokine levels in these patients.
We found that mercury-intoxicated patients had signi cantly increased tear levels of lL-12p70, IL-1RA, RANTES, and VEGF. Similar ndings have been described in DE patients [13,26,51] and in tears from advanced surface ablation refractive surgery patients. [16] The increase in these molecules is in agreement with the increase in serum cytokines in mercury-exposed patients, [2,21,22] and it re ects an in ammatory response at the ocular surface of these patients.
On the other hand. EGF, IP-10, and IL-6 tear levels were signi cantly decreased in mercury-intoxicated patients. EGF tear levels usually decrease in DE patients, particularly in the more severe forms. [10,27,29,30] A decrease in tear IP-10 levels has also been described by our group in patients with severe DE associated with ocular graft vs host disease [30] and by others in primary Sjögren syndrome and in Stevens-Johnson syndrome patients. [52] The nding of decreased tear IL-6 levels is in contrast to the increased concentration in DE patients. [24,26,27,29] However, a decrease in serum cytokines, including IL-6, has been reported among anti-nuclear antibody-positive subjects after mercury exposure. Nyland et al. suggested that the decrease in cytokine concentration was a speci c phenotype of mercury susceptibility. [21] A recent study in mice intoxicated with mercury showed increased levels of IL-6 in serum but not in the cerebellum. [53] IL-6 is a pleiotropic cytokine that acts as a neurotrophic factor and is crucial in the differentiation of oligodendrocytes and regeneration of peripheral nerves. [54] The decrease in IL-6 levels could be a factor contributing to the lower density of corneal nerves, branching density, and dendritic cell density observed in our patients.
In addition to the measurement of tear cytokine levels in the mercury-intoxicated patients, we performed PC and hierarchical agglomerative cluster analyses to explore correlation patterns among cytokine tear levels and the associations with the clinical ndings. Based on the tear cytokine levels, we identi ed two patient clusters. Patients in Cluster 2 had signi cantly increased tear levels for 18 out of the 23 assayed cytokines, indicating a higher degree of ocular surface in ammation in this group. In agreement with this, the Cluster 2 patients also had signi cantly decreased tear lysozyme levels, indicating reduced tear production, compared to the patients in Cluster 1. Interestingly, in the same group, the nerve branching density and dendritic cell density were also lower than in Cluster 1. Because the maximum urine mercury levels were signi cantly higher in patients belonging to Cluster 2, this probably indicates a more intense mercury intoxication in dose and/or exposure time, and/or a higher susceptibility to mercury toxicity. [21] Regarding our control groups, like the mercury-intoxicated patients, they were composed of males whose ages were not signi cantly different from that of the patients. IVCM was the only parameter compared with control values taken from three publications (two of them from our group), that included females and the ages were signi cantly different (see Table 4). While acknowledging that this is a limitation, we however feel that the comparisons are reliable because the corneal IVCM parameters that we analyzed do not change with age and are not associated with sex. [55][56][57] Another potential limitation is that our patients were taking psychotropic drugs for their erethism mercurialis. Because all of the patients had a similar clinical picture and were taking similar drugs, it is unlikely that this unavoidable potential bias had any effect of the results of our evaluations.
In summary, we described a range of unreported ocular surface pathologies produced by mercury poisoning. We hypothesize that the DE-related symptoms experienced by the patients are due to mercury-related damage to the corneal innervation, corneal sensitivity, and tear cytokine disturbances. Thus, this DE could be described as neurogenic in origin, in contrast to the more classic tear-de cient and/or evaporative-DE subtypes.