Vascular Endothelial Growth Factor (VEGF) Induced Downstream Responses to Transient Receptor Potential Vanilloid 1 (TRPV1) and 3-Iodothyronamine (3-T1AM) in Human Corneal Keratocytes

This study was undertaken to determine if crosstalk among the transient receptor potential (TRP) melastatin 8 (TRPM8), TRP vanilloid 1 (TRPV1), and vascular endothelial growth factor (VEGF) receptor triad modulates VEGF-induced Ca2+ signaling in human corneal keratocytes. Using RT-PCR, qPCR and immunohistochemistry, we determined TRPV1 and TRPM8 gene and protein coexpression in a human corneal keratocyte cell line (HCK) and human corneal cross sections. Fluorescence Ca2+ imaging using both a photomultiplier and a single cell digital imaging system as well as planar patch-clamping measured relative intracellular Ca2+ levels and underlying whole-cell currents. The TRPV1 agonist capsaicin increased both intracellular Ca2+ levels and whole-cell currents, while the antagonist capsazepine (CPZ) inhibited them. VEGF-induced Ca2+ transients and rises in whole-cell currents were suppressed by CPZ, whereas a selective TRPM8 antagonist, AMTB, increased VEGF signaling. In contrast, an endogenous thyroid hormone-derived metabolite 3-Iodothyronamine (3-T1AM) suppressed increases in the VEGF-induced current. The TRPM8 agonist menthol increased the currents, while AMTB suppressed this response. The VEGF-induced increases in Ca2+ influx and their underlying ionic currents stem from crosstalk between VEGFR and TRPV1, which can be impeded by 3-T1AM-induced TRPM8 activation. Such suppression in turn blocks VEGF-induced TRPV1 activation. Therefore, crosstalk between TRPM8 and TRPV1 inhibits VEGFR-induced activation of TRPV1.

Functional TRP melastatin type 8 (TRPM8) expression has been characterized on corneal nerves and endothelial cells (11,40,41) as well as corneal epithelial cells (42) and conjunctival epithelial cells (14,43). Both TRPV1 and TRPM8 are thermosensitive channels whose activity is modulated by temperature changes encountered in daily living. TRPM8 can be activated by moderate cooling from ∼25-28 • C, as well as by menthol and icilin (44,45), whereas TRPV1 increases its activity starting at 43 • C (46).
3-Iodothyronamine (3-T 1 AM) is a thyroid hormone metabolite and was detected in rodent tissues (47) and in human serum (48). This endogenous amine stimulates TRPM8 activity, which in turn suppresses TRPV1 activity in thyroid cells as well as conjunctival epithelial and corneal epithelial and endothelial cells (49,50). Administering high doses of 3-T 1 AM in vivo induces profound hypothermia within minutes in mice and hamsters (20), whereas lower doses do not evoke hypothermia but instead other reactions (51). Recent studies clarified some mechanisms and propose anapyrexia and vasodilation instead of hypothermia (52). Interestingly, TRPM8 activation by this agonist inhibited rises in TRPV1 activity induced by capsaicin in HCEC and HCjEC (42,43).
Even though the cognate VEGF receptor (VEGFR) and TRPV1 are coexpressed on corneal fibroblasts (53), it is unclear if the Ca 2+ transients mediating VEGF-induced stromal angiogenesis stem from VEGFR-mediated activation of TRPV1 (13,54,55). A clear indication that neovascularization and the increased expression of VEGF following corneal chemical injury may depend on TRPV1 expression, is that those tissue responses were markedly attenuated in TRPV1 knockout mice (53). Besides neovascularization, VEGF upregulation contributes to increased endothelial cell proliferation and migration in a bovine wound healing model (56,57). Furthermore, it is also unknown if TRPM8 activation alters VEGF-induced TRPV1 activation.
In this study, we document that TRPM8 activation suppresses TRPV1 responsiveness toward VEGFR activation. Such disruption of VEGFR-induced TRPV1 activation suggests that TRPM8 is potentially a viable target for the suppression of TRPV1-induced inflammatory fibrosis and neovascularization, which may also be of clinical relevance.

Medium and Reagents
The Keratinocyte Basal Medium KBM R provided by Lonza (Köln, Germany) is referred to as the Keratinocyte Growth Medium KGM R . VEGF-165 human recombinant protein was purchased from ThermoFisher Scientific (Rockford, IL, USA). Icilin was obtained from the Cayman Chemical Company (Ann Arbor, Michigan, USA). BCTC and AMTB were purchased from TOCRIS Bioscience (Bristol, UK). Accutase was purchased from PAA Laboratories (Pasching, Austria). All other reagents were purchased from Sigma-Aldrich (Darmstadt, Germany).

Corneas
Human corneas were obtained from cadavers (two males, three females, average age 75 ± 5 years). Donors provided written consent to the Department of Anatomy, Friedrich Alexander University Erlangen-Nürnberg (FAU), Germany. They were free of recent ocular surface trauma, eye infections or diseases. All corneas were dissected within 4 h and up to 24 h post-mortem.
Primary human cornea fibroblast cells (pHCF) were obtained from corneal peripheral remnants of transplants donated to German eye banks. Donors signed a patient consent form and the study complied with the underlying legal requirements and with the Helsinki Declaration. Epithelium was removed mechanically under a stereo microscope. Corneal stroma was cut into pieces (diameter of ≈1 mm), placed on 6-well petri dish and cultivated with Dulbecco's modified Eagle's medium (DMEM) HAMS F-12 (Biochrome AG, Berlin, Germany) containing 10% fetal bovine serum (FBS, Thermo Fisher Scientific, Waltham, USA). The medium was replaced at least every second day and primary cell culture (passage 0) was split with trypsin at a density of 80%. Cells were cultivated at 37 • C in a humidified 5% CO 2 incubator. For the present analysis, 8 different remnants of corneal transplantation from passage 1 to 5 were used.
Cell Culture of HCK SV40-immortalized human corneal keratocytes (HCK) were kindly provided by Michaela Zorn-Kruppa et al. from the Eppendorf-Hospital in Hamburg, Germany and cultivated according to their protocols (30,58,59). For the HCK cultivation, the same protocol was used as for the aforementioned pHCF cells. In brief, cells were grown in KGM R containing recombinant human epidermal growth factor (rhEGF 0.1 ng/ml), hydrocortisone (≈0.5 µg/ml), insulin (≈5 µg/ml), bovine pituitary extract (BPE) and 0.5 mM calcium chloride (CaCl 2 ) as well as penicillin/streptomycin in a humidified 5% CO 2 incubator at 37 • C (59). For the electrophysiological measurements, cell confluence ranged between 50 and 80%. At these different degrees of confluence, their electrophysiological characteristics were invariant (60).

RNA Isolation and RT-PCR
HCK, HCEC-12 and LNCaP (human prostate adenocarcinoma) cells were seeded in T75 flasks and grown to 80% confluence. Cells were harvested following three different passages. Total RNA was extracted using TRIzol R Reagent RT (Ambion, Austin, TX) according to manufacturer's instructions. The quality of extracted RNA was evaluated by NanoDrop ND-2000 spectrophotometer (PEQLAB, Germany. DNase digestion was performed and samples were stored at −80 • C. RNA (2 µg) was transcribed into cDNA by high capacity cDNA reverse transcription kit (Applied Biosystems, Darmstadt, Germany). cDNA was denatured (95 • C, 5 min), first strand synthesized (42 • C, 50 min) and the reaction was terminated by heating to 70 • C for 15 min. For RT-PCR, 2 µl cDNA mixture was used as a template in subsequent amplification reactions in a total of 30 µl volume containing specific primers for TRPM8 (sequences Fwd: CCTGTTCCTCTTTGCGGTGTGG AT; Rev: TCCTCTGAGGTGTCGTTGGCTTT) generating a 621-bp product. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as control (sequences Fwd: TCAACG ACCACTTTGTCAAGCTCA; Rev: GCTGGTGGTCCAGGGG TCTTACT) generating an anticipated 119-bp product. Each reaction also contained red PCR Master Mix (Stratec Biomedical AG, Birkenfeld, Germany). PCR reaction underwent the first cycle at 95 • C for 5 min, followed by 35 cycles of a repeat of denaturation at 95 • C for 15 s, a primer-specific annealing step at 58 • C for 30 s (TRPM8) or 60 • C for 30 s (GAPDH) and a primer-specific elongation step at 72 • C for 45 s (GAPDH) or at 72 • C for 7 min (TRPM8), followed by a final temperature holding at 4 • C. Eight microliters of the PCR products were loaded on a 1.5% agarose gel and after electrophoresis they were visualized via ethidium bromide staining under UV light.

Quantitative RT-PCR
TRPM8 specific primers (sequences Fwd: ATGGCCGGG ACGAGATGGACA; Rev: AGCCCCTGGTCTGCTCCCAAA) generated 138-bp products. Aforementioned GAPDH was also used as the reference gene in qPCR. Amplification was carried out using the Mx3000P qPCR system real-time cycler (Stratagene, Waldbronn, Germany). For detection LightCycler R 480 SYBR Green I Master (Roche, Germany) was used. Amplification was performed for 45 cycles lasting 15 s (95 • C) and 30 s (60 • C) as previously described (15,43). GAPDH expression levels normalized TRPM8 gene expression levels. Melting curve analysis was performed to confirm the specificity of the PCR reaction. Data was processed using double delta Ct analysis. Each of the three passages of every cell line was measured in triplicate together with a sample without reverse transcription to exclude genomic DNA contamination.

Fluorescence Calcium Imaging
HCK cells were cultivated on 15 mm diameter glass cover slips placed in a culture plate with wells until they reached a semi confluent stage (≈60-80%). The culture conditions used to prepare cells for immunofluorescence were the same as those used for Ca 2+ imaging. The cells were loaded with fura-2/AM (1 µM) at 37 • C for 20-40 min. Loading was stopped with a Ringer-like (control) solution containing (mM): 150 NaCl, 6 CsCl, 1 MgCl 2 , 10 glucose, 10 HEPES, and 1.5 CaCl 2 at pH 7.4 and 317 mOsM (62). Potassium was replaced by cesium and the coverslips were placed on an inverted microscope stage (Olympus BW50WI, Olympus Europa Holding GmbH, Hamburg, Germany) inside an experimental chamber containing the same solution. This setup was connected to a digital imaging system (TILL Photonics, Munich, Germany), outfitted for UV excitation. Fura-2/AM fluorescence was alternately excited at 340 nm and 380 nm and monitored for different times at 500 ms intervals (63). The corresponding emission wavelength was 510 nm. The fluorescence ratio (f 340nm /f 380nm ) was calculated by the software. This ratio is a relative index of intracellular Ca 2+ ([Ca 2+ ] i ) levels (63). The f 340nm /f 380nm dynamic ratio range was small [0.2 (42)] because single fluorescence response signals at 340 and 380 nm were set to a fixed value to avoid distortions of the Ca 2+ responses patterns. A TRPM8 control was provided by overexpressing TRPM8 in transfected cells and exposing them to described agonists (42). The measuring field was adapted to the number of cells (TILL Photonics view finding system). Before each experiment, cells were routinely tested to determine whether the calcium homeostasis (control baseline) remained constant for 10 min. The control measurements are shown with open circles in the figures. All experiments were performed at a constant room temperature (≈23 • C). If stabilization had not occurred within the first 5 min, adaptation to room temperature was prolonged. Results are shown as mean traces of the f 340nm /f 380nm ratio ± SEM (error bars in both directions) with n-values indicating the number of experiments per data point. For 10 min, measurements were obtained from groups of 5-10 cells at least three times. The fluorescence ratios were normalized (control set to 1.2) and averaged (with error bars). In addition, the fura-2-induced fluorescence signals were alternatively acquired and evaluated using Life Science fluorescence cell imaging software "cellSens" (Olympus, Hamburg, Germany) in conjunction with a digital camera (Olympus XM-10) (Figures 4G,I, 5D,F, 8F,H, 10E,I; control set to 0.1 and 0.2, resp.). For the fluorescence excitation wavelengths (340 and 380 nm), specific filters and a LED light source were used (LED-Hub by Omikron, Rodgau-Dudenhoven, Germany). Fura-2 fluorescence was alternately excited at 340 and 380 nm and emission was detected every 5 s at 510 nm (250-3,800 ms exposure time). The relatively slow rise times of Ca 2+ transients in all experiments were a consequence of drugs being pipetted into a stationary bath rather than a flow through system. In such a configuration, the time delay was 1-2 min between drug addition and peak response. When using a blocker, cells were preincubated with the blocker for ∼30 min before the measurement and all test solutions contained the blocker. Drugs were dissolved in dimethyl sulfoxide (DMSO) to obtain a stock solution and diluted to provide a working concentration that did not exceed 0.1%. This DMSO concentration was nontoxic based on stable f 340nm /f 380nm levels (data not shown).

Planar Patch-Clamp Recordings
Whole-cell currents were measured using a planar patchclamp setup (Port-a-Patch R , Nanion, Munich, Germany) in conjunction with an EPC 10 patch-clamp amplifier (HEKA, Lamprecht, Germany) and controlled by PatchMaster software (Version 2.6; HEKA, Lamprecht, Germany). A standard intracellular solution containing (mM): 50 CsCl, 10 NaCl, 60 CsF, 20 EGTA, and 10 HEPES at pH ≈7.2 and ≈288 mOsM was applied to the microchip (both provided by Port-a-Patch R , Nanion, Munich, Germany). The external solution contained (mM): 140 NaCl, 4 KCl, 1 MgCl 2 , 2 CaCl 2 , 5 D-glucose monohydrate, and 10 HEPES, pH ≈7.4 and osmolarity ≈298 mOsM. A single cell suspension of 5-10 µl was placed onto a microchip having a 2.5-3 M resistance (aperture ≈1-3 µm). A negative pressure applied by a software-controlled pump (Nanion) fixed one single cell atop the aperture. Mean membrane capacitance (19 ± 2 pF; n = 38) and mean access resistance (18 ± 2 M ; n = 38) were software calculated. Series resistances, fast and slow capacitance transients were compensated by the software of the patch-clamp amplifier. Series resistance did not change markedly during an experiment. The liquid junction potential was calculated (≈3.8 mV) (64), which is in close agreement with the measured range (≈4-7 mV). Its mean value was used to correct the measurements analyzed with the Patch-Master software. Current recordings were all leak-subtracted and cells with leak currents above 100 pA were excluded from analysis. All experiments were performed at 22 • C room temperature in an air-conditioned room to avoid confounding responses by other thermo-sensitive TRPs. The current response patterns were generated through application of specific voltage step-protocols, which induced typical TRP channel whole-cell current patterns (62). The holding potential (HP) was set to 0 mV in order to eliminate any possible contribution by voltage-dependent Ca 2+ channel activity. After confirming the control settings, experiments were started ∼10 min after breaking into a whole-cell configuration (65). Whole-cell currents were recorded using 10 mV voltage steps over a range from −60 to +130 mV (10 mV increments) for 400 ms each. Currents were also recorded through a voltage ramp protocol of −60 to +130 mV range and 500 ms duration every 5 s. Resulting currents were normalized with respect to cell membrane capacitance to obtain current density (pA/pF) values displayed in current vs. voltage plots.

Statistical Analysis
Significance was determined using Student's t-test for paired data (p-values: two-tailed) provided they passed a software available normality test. If the normality test failed, non-parametric Wilcoxon matched pairs were used. For non-paired data, Student's t-test for unpaired data was used, if it passed a normality test. Alternatively, the non-parametric Mann-Whitney-U test was performed. Welch's correction was applied if data variance of the two groups were not at the same level. Probabilities of p < 0.05 [indicated by asterisks ( * ) and hashtags (#)] were considered to be significant. In addition, normally distributed data of more than two groups were statistically analyzed using one-way ANOVA (one-way analysis of variance). Selected pairs of columns were tested using Bonferroni post-test. If the data were not normally distributed, Kruskal-Wallis test was used and selected pairs of columns were tested using Dunn's post-test. Significance level alpha = 0.05 (95% confidence intervals). The number of repeats is shown in each case in brackets, near the traces or bars. All values are means ± SEM (error bars in both directions). All plots were generated with SigmaPlot software version 12.5 for Windows (Systat Software, San Jose, California, U.S.A.). Bar charts were plotted and statistical analyses were performed using GraphPad Prism (version 5.00 for Windows) (La Jolla, California, USA.).  ] i with CAP (n = 9; p < 0.05; paired tested). The hashtag (#) denotes a statistically significant difference in fluorescence ratios between CAP with and without CPZ (n = 9-12; p < 0.05; non-paired tested).

Validation of HCK Identity
The expression of lumican (LUM), a biomarker of keratocytes, was discernible along with LUM derived green fluorescence, which was localized in the cytoplasm, and also in the endoplasmic reticulum (ER) (66, 67) ( Figure 1A). This expression pattern is similar to results obtained with an anti-keratocan (KTN) antibody ( Figure 1C), which is a more specific keratocyte marker than LUM. Alpha-smooth muscle actin (α-SMA) expression patterns were different from the patterns with the LUM and KTN biomarker. The α-SMA expression patterns are indicative of myofibroblast coexpression (67). In this case, up to 30% of HCK cells were immunofluorescent (IF). Another 20-30% of these cells displayed a diffuse IF signal whereas ∼40-50% of the HCK cells did not express α-SMA ( Figure 1B). In controls, omission of the primary antibody eliminated IF ( Figure 1D). The specificity of the LUM, KTN and α-SMA antibodies was verified by showing absence of any IF staining in telomeraseimmortalized human corneal epithelial (hTCEpi) cells (31). LUM specificity agrees with its failure to stain pHCF cells (data not shown). CD90 is a biomarker of mesenchymal stem cells (MSC) derived from keratocytes in vitro. It was also detectable in the cultured stromal cells (HCK) (Figure 2) (68). Immunostaining patterns were eliminated by omitting the primary antibody ( Figure 2E). The MSC marker expression panel shows that keratocytes constitute a majority of the cell types in the HCK cell line. They are accompanied by a subpopulation of fibroblasts and myofibroblasts and some mesenchymal stem cells. Such a mixture is unavoidable especially in a serum containing medium since under this condition keratocytes spontaneously undergo transformation acquiring a (myo)fibroblast phenotype (66,69).

TRPV1 Protein Expression
TRPV1 was identified in corneal stromal cells of human corneal cross sections (HCCS) as well as in primary human corneal fibroblast (pHCF) cells at different cell passages. The anti-TRPV1 antibody exhibited a consistent cytoplasmic staining pattern in all analyzed cell passages 1-5 (Figures 3A,F). Furthermore, α-SMAderived green fluorescence was localized in the cytoplasm in about 50% of pHCF (passage 1) (Figures 3G,H). Some HCF cells were not stainable with any of the aforementioned antibodies (arrows), which is consistent with the presence of fibroblasts and myofibroblasts derived from keratocytes. Figures 4A,B demonstrate the localization of TRPV1 in SV40-immortalized HCK cells, which were used in the following functional assays.

Functional TRPV1 Channel Expression
Raising the bath solution temperature to >43 • C (Figure 4C, upper trace) increased the fluorescence ratio from 1.2000 ± 0.0001 to 1.2050 ± 0.0013 after 590 s in immortalized HCK (n = 4; Figure 4C). Similarly, activation of TRPV1 by 20 µM µM capsaicin (CAP) irreversibly increased the fluorescence ratio from 1.2000 ± 0.0002 to 1.213 ± 0.0028 after 590 s (n = 8; p < 0.01; Figure 4D), whereas this rise was blocked by 20 µM capsazepine (CPZ) (1.201 ± 0.0007; n = 6; p < 0.001; Figures 4E,F). To avoid desensitization and cell death with over stimulation, experiments were repeated using a lower concentration of CAP and CPZ (both 10 µM) in combination with an alternative fluorescence signal measuring setup (Figures 4G,H,I). The results are comparable. However, a recovery of the CAP-induced Ca 2+ increase could be clearly suppressed in the presence of 10 µM CPZ ( Figure 4H). More specifically, activation of TRPV1 by 10 µM CAP reversibly increased the fluorescence ratio from 0.0969 ± 0.0031 to 0.2050 ± 0.0394 after 132 s (n = 9; p < 0.05; Figure 4G), whereas this rise could be blocked by 10 µM CPZ (0.1041 ± 0,0017; n = 12; p < 0.05; Figures 4H,I). The changes and the underlying whole-cell currents were determined by measuring the time dependent changes in currents along with evaluating the plots of the corresponding current voltage relationships at the indicated time points: A, B and C (Figures 5A,B). At positive pipette potentials, CAP (10 µM) activated outwardly rectifying currents that were larger than those at negative voltages ( Figure 5B). As already observed in HCjEC, L-carnitine (1 mM) is a TRPV1 antagonist (15) and it reduced maximal negative and maximal positive current amplitudes induced by a voltage step from −60 to +130 mV (% of control) (15) shown in Figure 5C. In each measured cell, 10 µM CAP increased both in-and outward currents, whereas 1 mM L-carnitine suppressed them (n = 6; p < 0.05 at the minimum). In addition, functional TRPV1 expression could be also documented with the classical TRPV1 antagonist CPZ (10 µM) (n = 5; p < 0.05; Figure 5C). In some previous studies as well as this study, 20 µM CAP was used to elicit physiological responses in human corneal endothelial cells (11), corneal epithelial cells (60), retinoblastoma cells (70) and in neuroendocrine tumor cells (71). Regarding the effect of L-carnitine, fluorescence Ca 2+ imaging results mirrored those obtained in the patch-clamp experiments (Figures 5D,F). The CAP-induced Ca 2+ increase (n = 55; control n = 9) could be clearly suppressed in the presence of 1 mM L-carnitine (Figures 5E,F  Ten micromolar CAP induced an increase of normalized maximal in-and outward whole-cell currents (for −60 mV control was set to −100%; for 130 mV control was set to +100%; n = 13). This increase could be suppressed by 1 mM L-carnitine and 10 µM CPZ, respectively (n = 5-6). Data passed the normality test and were statistically analyzed using one-way ANOVA (one-way analysis of variance). Selected pairs of columns were tested using Bonferroni post-test. The asterisks (*) designate significance differences between the columns (n = 5-13; p < 0.05 at the minimum). Significance level alpha = 0.05 (95% confidence intervals). (D) CAP (10 µM) induced an irreversible increase of intracellular Ca 2+ concentration (n = 55) whereas non-treated control cells maintained a constant Ca 2+ baseline (n = 9).
(E) Same experiment as shown in (D), but in the presence of L-carnitine. L-carnitine (1 mM) suppressed the CAP-induced Ca 2+ increase (n = 55). (F) Summary of the experiments with CAP and L-carnitine. Data were statistically analyzed using one-way ANOVA (and Nonparametric) (one-way analysis of variance) (Kruskal-Wallis test).

TRPM8 Gene and Protein Expression
RT-PCR and semi-quantitative real-time PCR identified TRPM8 gene expression in HCK (Figures 6A,B) based on generating the predicted TRPM8 621-bp amplicon ( Figure 6A). qPCR confirmed its identity because the size of this product was identical with its positive control in the LNCaP cell line (72,73) and corneal endothelial cells (41). In addition, TRPM8 immunostaining expression was identified in both the cell membrane and peri-nuclear regions (Figures 6C,D,E). Absence of immunostaining caused by omission of the primary antibody excluded nonspecific secondary antibody staining. IF analyses also detected TRPM8 expression, on keratocytes in HCCS (Figures 6F,G).
The increases in the inward whole-cell currents induced by VEGF (10 ng/ml) shown in Figure 11 are similar to those shown in Figure 12. Figure 12A shows maximum in-and outward whole-cell currents under control conditions (marked as A) and in the presence of 10 ng/ml VEGF (shown in green and marked as B). After extracellular application of 1 µM 3-T 1 AM, the currents decreased after a short delay (shown in red and marked as C). The corresponding current voltage relationships at the aforementioned points of time (A, B, and C) are shown in Figure 12B. This experiment was repeated using a step protocol instead of a voltage ramp protocol (from −60 to +130 mV in 10 mV steps at a holding potential of 0 mV). Figure 12C shows the corresponding whole-cell currents under control conditions, which were increased in the presence of VEGF ( Figure 12D). Notably, the inward currents were completely suppressed in the presence of 3-T 1 AM since the inward currents are mainly attributable to increases in Ca 2+ influx (Ca 2+ free internal solution) ( Figure 12E). The corresponding current voltage relationships from the recordings shown in Figures 12C,D,E are summarized in Figure 12F, which clearly demonstrate that the VEGF-induced current increase could be blocked by 3-T 1 AM. Figure 12G summarizes the statistical analysis of all patch-clamp experiments. Specifically, the inward currents rose from −15 ± 3 pA/pF to −24 ± 4 pA/pF (n = 16; p < 0.0005). Similar to CPZ, 1 µM 3-T 1 AM suppressed the inward currents (−19 ± 5 pA/pF; n = 16; p < 0.05; Figure 12G). The same pattern was seen in the outward currents. VEGF increased the mean outward currents from 98 ± 14 pA/pF to 129 ± 14 pA/pF (n = 16; p < 0.01), which were suppressed to 100 ± 16 pA/pF by 3-T 1 AM (n = 16; p < 0.05; Figure 12G). Taken together, 3-T 1 AM suppressed VEGF-induced increases in the whole-cell currents.

Limitations of Therapeutic Intervention of Corneal Neovascularization
Corneal neovascularization (CNV) is a severe pathological condition that may profoundly impair vision. Although its clinical impact is very well known, novel strategies are still needed to prevent infiltration of incompletely formed blood vessels. This maladaptive response to either infection, trauma, toxic agents or inflammation accounts for a significant number of patients seen in clinical practice (81). Angiogenesis is also a significant risk factor for keratoplasty because neovascularization develops in 41% of the cases undergoing this procedure (82). Furthermore, medical interventions are only partially successful because they do not fully reduce neovascularization through topical, subconjunctival, and intraocular application of bevacizumab [reviewed (83)]. Alternatively, glucocorticoids are used but they do not regress preexisting CNV [reviewed (84)]. Furthermore, chronic usage of steroids to reduce inflammation often has ocular side effects such as cataracts, glaucoma and superinfection (85). Another limitation involves their short duration of action. Rapamycin also has limited effectiveness in inhibiting CNV induced by a chemical burn through suppressing the mammalian target of rapamycin (mTOR) signaling pathway in a mouse corneal wound healing model (86,87). Even though CNV declined, it was accompanied by increases in proinflammatory cytokines and decreases in anti-inflammatory cytokines. These latter effects may instead aggravate infection driven immuneinduced inflammation (88). Such complications account for continuing efforts to develop alternative strategies for more effective suppression of neovascularization (89).

Biomarker Validation of HCK Identity
Biomarker expression patterns confirmed the practicality of using the HCK cell line as an in vitro test model (Figure 1). Even though, LUM is also expressed by other cell types in the heart and lung (32,90), it is coexpressed with KTN, which is a more selective keratocyte biomarker than LUM (32) (Figure 1C). Furthermore, CD90, a mesenchymal stem cell marker, was evident (Figure 2). This expression pattern documents the heterogeneity of cell types found in stromal isolates of human corneal tissue (25,26,91). Myofibroblasts were also detected based on a more intense α-SMA staining pattern than that of KTN ( Figure 1B). Furthermore, fibroblast and myofibroblast coexistence is consistent with numerous reports showing that keratocyte phenotype maintenance is very condition dependent. For example, environmental stress or tissue culture medium content changes can cause keratocytes to undergo a differentiation program. As a result, this transition can lead to fibroblasts and ultimately myofibroblast differentiation (66,68).

Specificity of TRPV1 Activation
TRPV1 protein expression in stromal keratocytes, HCCS, pHCF (Figure 3) and HCK (Figures 4A,B), agrees with its presence in human corneal fibroblasts (13). The CAP-induced Ca 2+ transients in corneal epithelial cells closely resemble those obtained with the HCK cell line (16). The irreversibility of these transients in some tissues may be related to differences  in its tissue specific binding to TRPV1. Functional TRPV1 expression was also confirmed by showing that raising the bath temperature above 43 • C induced rises in Ca 2+ levels that were partially reversible (92). Differences in reversibility were also found between pterygial cells and HCjEC. The poor reversibility in the immortalized HCK cell line corresponds with what was described in non-malignant pterygial cells and healthy HCjEC (16). As shown in Figure 3, not all HCK cells were stainable. This staining inhomogeneity is consistent with the presence of fibroblasts and myofibroblasts derived from keratocytes. Notably, TRPV1 and α-SMA are coexpressed, which is characteristic of myofibroblasts (38). Furthermore, their coexpression was described in a report showing that TRPV1 upregulation promotes α-SMA expression during myofibroblast transdifferentiation (93). Such an association may be needed for TRPV1 to mediate control of actin assembly and disassembly that occurs during upregulation of this contractile protein. This type of dependence was described in F11 cells transduced with retroviral particles obtained from a triple transfection of human embryonic kidney (HEK)293T cells with TRPV1 plasmids (93). CAP concentrations (10-20 µM) led to Ca 2+ transients and whole-cell current increases (Figures 4D,I, 5A,D). These effects are in line with those reported in studies using human corneal endothelial cells (11), corneal epithelial cells (60), retinoblastoma cells (70), and in neuroendocrine tumor cells (71). The CAPinduced changes of ≈20 pA/pF in either direction are moderate (e.g., Figures 5A,B,C) and may be mixed by heterogeneous responses of different cell types in the HCK cell line. As previously described, L-carnitine is able to suppress TRPV1 activation (Figures 5A,B,C) and Ca 2+ influx (Figures 5D,E,F), which makes this osmoprotectant relevant for possible use in a clinical setting (e.g., dry eye) (15,(94)(95)(96). Taken together, CAP has adequate TRPV1 selectivity to be used as a marker of its expression (24,46). Even though CPZ only blocks the vanilloid site on TRPV1, it is the first identified competitive antagonist of capsaicin, which is a sensory neuronal excitant. Accordingly, it is universally accepted as a relevant TRPV1 probe (24,46,97).

Specificity of TRPM8 Activation
TRPM8 gene and protein expression levels (Figures 6A,B,C,D,E) in HCK and HCCS cells (Figures 6F,G) are similar to those described in human corneal (HCEC) and conjunctival epithelial cells (HCjEC) (42,43). Even though there is a poor correspondence between the extensive and rather intense TRPM8 immunostaining and low level of TRPM8 gene expression, this disparity may be attributable to reported limited selectivity of the commercially available TRPM8 antibodies (16,42,43). Nevertheless, menthol (100-500 µM) induced Ca 2+ transients and increases in whole-cell currents corresponding to those reported in HCEC and HCjEC (42,43). These high menthol concentrations were appropriate because concentrations as high as 1 mM were used for describing functional TRPM8 expression in some other studies (74). Similarly, inhibition by AMTB of these menthol induced responses confirmed TRPM8 involvement ( Figure 7A) (98). Non-specific menthol effects cannot be excluded because the current rises were not fully reversible at higher menthol concentrations (> 100 µM). At these concentrations, an irreversible rise in intracellular Ca 2+ levels occurred in human corneal endothelial cells (HCEC-12) and in TRPM8-overexpressing HEK293 cells (99) whereas in HCEC-12 a reversible effect was detectable (41). BCTC and AMTB are well-established TRPM8 antagonists, but AMTB is only TRPM8 selective whereas BCTC is also a TRPV1 antagonist (21,24,100). Nevertheless, both BCTC and AMTB had similar inhibitory effects on TRPM8 activation induced by either menthol or icilin (Figures 7-9) (98,101). Specifically, icilin, which is a mixed TRPM8/ TRPA1 agonist (79) also irreversibly increased intracellular Ca 2+ in HCEC and HCjEC (42)(43)(44) whereas these increases were also inhibited by AMTB confirming TRPM8 involvement in HCK (Figures 8D,G) (42)(43)(44). As icilin is both a super-cooling TRPM8 agonist, and a weak TRPA1 agonist (44,102), its effects may include TRPA1 activation at either a high icilin/menthol concentration or cooling to ≈14 • C ( Figure 8A) (79). Nevertheless, the menthol/AMTB effects are very likely solely attributable to TRPM8 activation because icilin had effects that can only be accounted for by TRPM8 activation in mutant mice (98). In the current study, icilin induced increases in the intracellular Ca 2+ levels and the whole-cell currents which were inhibited by both BCTC and AMTB (Figure 9). These inhibitory effects were similar to those occurring in HCEC (42). 3-T 1 AM was used as probe of thermo-sensitive TRPM8 involvement in mediating temperature lowering in rodents (20). Even though the responses induced by this thyroxine metabolite also included an interaction of TRPM8 with a GPCR in some other cell types, we did not deal with this possibility (4).

VEGF Interacts With TRPV1
The Ca 2+ transients and underlying ionic currents induced by VEGF are similar to those in microvascular endothelial cells and podocytes in which these responses were instead mediated by crosstalk with TRP canonical 6 (TRPC6) rather than TRPV1 (103,104). In HCK, TRPV1 is solely involved because CPZ completely blocked these effects whereas AMTB had no effect (Figure 10). Unlike CPZ, AMTB instead slightly enhanced VEGF-induced whole-cell current, which supports the notion of TRPM8 suppressing crosstalk between TRPV1 and VEGFR ( Figures 11D,E,F). The time delays seen in some cases in VEGFinduced rises in whole-cell currents and Ca 2+ transients are probably due to drug diffusion delays ( Figure 11A).

3-T 1 AM Involvement in VEGF Signaling
In this study, 3-T 1 AM blunting of the crosstalk between VEGFR and TRPV1 is similar to its inhibitory effect on TRPV1 activation by CAP in HCEC and HCjEC (42,43). Other evidence supportive of crosstalk among these receptor triad members stems from the fact that AMTB increased inward currents induced by VEGF (Figures 11D,E,F) whereas 3-T 1 AM blunted VEGF-induced increases in whole-cell currents using a voltage ramp protocol (Figures 12A,B) or a voltage step protocol (Figures 12C,F). The current voltage relationships are consistent with TRP behavior described in previous studies of corneal cells (e.g., reversal potential, outwardly rectifying currents) (e.g., (13,60)). Interestingly, 3-T 1 AM had a delayed response which may be possibly attributable to a diffusion delay to an intracellular binding site. Another possibility requiring further study is that 3-T 1 AM instead indirectly targets TRPV1 via directly promoting the inhibitory effect of the (soluble) VEGF receptor 1 on VEGFR, which is also expressed in the cornea (55,105). Such control by TRPM8 can be even more complex and may also involve interactions with adrenergic receptors (106).

Potential Therapeutic Targets
This study provides additional supportive evidence that TRPV1 is a potential drug target for improving treatment of VEGFinduced neovascularization. Such an effect may be obtainable by suppressing VEGFR crosstalk with TRPV1 through increasing TRPM8 activity with an agonist such as thyronamine (3-T 1 AM). This strategy may provide a selective approach to inhibit TRPV1 upregulation and activation, which in turn suppresses angiogenesis, fibrosis as well as inflammatory processes in different pathophysiological conditions such as dry eye disease (DED), pterygium, and conjunctivitis sicca or red eyes (16,94,107). 3-T 1 AM may also be beneficial in the treatment of DED (42,43,49) since TRPM8 activation triggers increases in tear fluid production by the lacrimal gland via activation of the Central Nervous System (CNS) (40). In this context, borneol is another TRPM8 agonist, which provided symptomatic relief to DED patients when it was applied as eye drops (108)(109)(110).

CONCLUSIONS
VEGF-induces Ca 2+ transients by transactivating TRPV1, whereas TRPM8 activation suppresses this response by blocking TRPV1 activation. Accordingly, VEGF-induced corneal neovascularization may be inhibited by novel TRPM8 agonists such as 3-T 1 AM.

AUTHOR CONTRIBUTIONS
SM, ErT, and PR designed the study, analyzed the data, wrote, and edited the manuscript. FG and TB performed immunofluorescence analysis and keratocyte validation. FG managed primary human corneal fibroblast cell culture. JK contributed with his expertise on thyroid hormone metabolites, discussed data and their interpretation, and helped edit the manuscript. UP also contributed with his expertise in medical issues and helped edit the manuscript. NK performed PCR analysis and immunohistochemistry. ErT, SM, AL, PJ, EliT, DC, FS, RR, and NL performed calcium measurements and planar patch-clamp recordings as well as plot analyses. ErT, PJ, NK, AL, EliT, and SM created diagrams.

FUNDING
SM is supported by DFG (ME 1706/18-1) for a TRP channel related research project. The planar patch-clamp equipment was partially funded by Sonnenfeld-Stiftung (Berlin, Germany). FG is supported by DFG (PA 738/9-2). TB is participant in the BIH-Charité Clinical Scientist Program funded by the Charité-Universitätsmedizin Berlin and the Berlin Institute of Health. This work was supported by DFG, priority program SPP1629 Thyroid Trans Act, KO 922/16-2 and KO 922/17-2 to JK and ME 1706/13-1 to SM.