Transient Cell Membrane Disruptions induce Calcium Waves in Corneal Keratocytes

The purpose of this study was to determine if transient cell membrane disruptions (TPMDs) in single keratocytes can trigger signaling events in neighboring keratocytes. Stromal cells were cultured from human corneas (HCSC) and mouse corneas (MCSC). TPMDs were produced using a multiphoton microscope in Cal-520-AM loaded cells. TPMD-induced calcium increases (Ca++i) were measured in Ca++-containing and Ca++-free solutions containing thapsigargin, ryanodine, BAPTA-AM, 18-α-glycyrrhetinic acid (18α-GA), apyrase, BCTC, AMG 9810, or AMTB. Fluorescence intensity was recorded as the number of cells responding and the area under the fluorescence versus time curve. The maximum distance of responding neighboring cells in ex vivo human corneas was measured. Connexin 43 protein in HCSC and MCSC was examined using immunofluorescence staining, and corneal rubbing was applied to confirm whether TPMDs occur following mechanical manipulation. Our results demonstrate that single cell TPMDs result in Ca++ waves in neighboring keratocytes both in culture and within ex vivo corneas. The source of Ca++ is both intra-and extra-cellular, and the signal can be mediated by ATP and/or gap junctions, and is species dependent. Stromal rubbing confirmed that TPMDs do occur following mechanical manipulation. Keratocyte TPMDs and their associated signaling events are likely common occurrences following minor or major corneal trauma.


Results
calcium signaling. Extracellular calcium. Creation of a single, small TPMD in single keratocytes/stromal cells of all control preparations resulted in large Ca ++ i increases in the targeted (source) cells and in many neighboring cells in a time-dependent fashion (i.e. Ca ++ waves; Fig. 1). A summary of all experimental results is presented in Table 1.
Initial experiments were designed to determine the Ca ++ source of neighboring cell TPMD-induced Ca ++ waves. Cultured corneal stromal cells and keratocytes within cornea rims were examined in K-SFM with or without Ca ++ to examine the influence of extracellular calcium. Figure 2a shows representative micrographs of HCSC  before and after TPMD under several experimental conditions. It is notable that Ca ++ -free K-SFM significantly reduced both the human stromal cell responding number (1.3 ± 0.22) and normalized curve area (9.83% ± 2.56) when compared to K-SFM + 1 mM calcium (6.16 ± 0.38, 100% ± 13.39; both P < 0.05) (Fig. 2b,c). Calcium wave videos corresponding to all of the still photographs in Fig. 2 can be found in Supplemental Videos S1-S5. Figure 3a demonstrates representative micrographs of MCSC before and after TPMDs. In cultured primary mouse corneal stromal cells, Ca ++ -free K-SFM resulted in a significantly lower cell number (5.03 ± 0.56) and normalized curve area (15.57% ± 2.38) compared to K-SFM + 1 mM calcium (8.86 ± 0.09 and 100% ± 6.75, respectively; P < 0.05) (Fig. 3b,c). Videos corresponding to all of the still photographs in Fig. 3 can be found in Supplemental Videos S6-S10.
The influence of extracellular calcium on TPMD-induced calcium signaling was also studied in keratocytes residing in situ within human corneal rim tissue. Our results confirm that TPMD-induced keratocyte calcium signaling is present within corneal tissue (Fig. 4a). As in the cultured cells, calcium signaling was significantly reduced in a Ca ++ -free extracellular environment (Fig. 4a,b). The mean maximum cell distance between the source cell and farthest responding cell was 143.43 ± 14.28 μm in the Ca ++ -free K-SFM group vs. 211.57 ± 13.9 um in the K-SFM + 1 mM calcium group (P < 0.05). Videos corresponding to all of the still photographs in Fig. 4 can be found in Supplemental Videos S11-S12.
Intracellular calcium. K-SFM plus the sarcoplasmic/endoplasmic reticulum Ca ++ ATPase inhibitor thapsigargin, or the intracellular Ca ++ release blocker ryanodine, were used to examine the role of intracellular Ca ++ in TPMD-induced calcium waves. K-SFM plus the calcium chelator BAPTA-AM was as a positive control to examine the combined extracellular and intracellular calcium influence on TPMD-induced calcium waves.
The influence of intracellular calcium on TPMD-induced calcium signaling was also studied in keratocytes residing within human corneal rim tissue. Thapsigargin did not significantly reduce the cell distance in stromal keratocytes (151.20 ± 30.45 μm in DMEM controls vs. 85.74 ± 13.63 μm in the thapsigargin group) (Fig. 5). Videos corresponding to all of the still photographs in Fig. 5 can be found in Supplemental Videos S13-S14. Thapsigargin added to K-SFM did significantly reduce the cell distance (52.26 ± 1.80 μm in K-SFM + thapsigargin vs 211.58 ± 13.90 μm in K-SFM + 1 mM calcium; P < 0.05).

Influence of gap junctions.
To confirm the presence of gap junctions in HCSC and MCSC, Cx43 immunofluorescence staining was performed. Figure 6 demonstrates the presence of Cx43 at the cell junctions of both HCSC and MCSC.
The influence of gap junctions on TPMD-induced calcium signaling was examined using the gap junction blocker 18α-GA. In primary human corneal stromal cells, there was no significant difference between DMEM controls and 18α-GA treated cells in either the cell number or the normalized curve area (4.00 ± 0.45 vs. 3.07 ± 0.38 and 100% ± 16.99 vs. 93.33% ± 18.13, respectively) (Fig. 7b,c). Videos corresponding to all of the still photographs in Fig. 7 can be found in Supplemental Videos S15-S20.
In primary mouse corneal stromal cells, there was also no significant difference in the cell number between DMEM controls and 18α-GA treated cells (5.65 ± 0.55 versus 6.63 ± 0.42, respectively). There was, however, www.nature.com/scientificreports www.nature.com/scientificreports/ a significant decrease in the normalized curve area of control versus 18α-GA treated cells (100% ± 13.48 vs 38.96% ± 4.06; P < 0.05) (Fig. 8b,c). Videos corresponding to all of the still photographs in Fig. 8 can be found in Supplemental Videos S21-S26. In human corneal rim tissue compared to DMEM, 18α-GA significantly reduced the cell distance (117.71 ± 12.13 μm vs. 68.92 ± 7.22 μm, respectively; P < 0.05) (Fig. 9). Videos corresponding to all of the still photographs in Fig. 9 can be found in Supplemental Videos S27,S28.
Influence of extracellular ATP. The influence of extracellular ATP, presumably released by the source cell, on TPMD signaling was analyzed using apyrase, a potent hydrolyzer of ATP. In primary human corneal stromal cells, both the cell number and the normalized curve area were significantly reduced following apyrase treatment in Ca ++ -free K-SFM and K-SFM + Ca ++ before and after laser-induced TPMD. The TPMD location is shown as an arrowhead. The neighboring cell farthest from the source cell with a notable change in fluorescence was noted (white circle) and the distance from the source cell was measured. The mean maximum distance of approximately 10 target source cells from each rim was calculated and used for statistics analysis. (b) Ca ++ -free K-SFM versus K-SFM + Ca ++ cell distance. Numbers within bars indicate TPMD targeted number of cells/ number of rim. Data presented as mean ± SE. * indicates P < 0.05. www.nature.com/scientificreports www.nature.com/scientificreports/ (5.40 ± 0.58 vs. 3.66 ± 0.22 and 100% ± 13.94 vs. 64.04% ± 10.06, respectively; P < 0.05) (Fig. 7b,c). In primary mouse corneal stromal cells, apyrase had no effect on either the cell number or the curve area (6.07 ± 0.42 vs. 4.97 ± 0.39 and 100% ± 10.72 vs. 148.96% ± 25.59, respectively) (Fig. 8b,c).

Discussion
The corneal stroma harbors an abundance of quiescent keratocytes that form a cellular network interconnected via gap junctions and numerous intercellular signaling pathways. Keratocyte cell signaling has previously been studied following major disruptions to the different cell types of the cornea or the cornea itself. Transient plasma membrane disruption (TPMD), also referred to as a membrane tear, is a common form of injury to individual cells that can be created by physiological mechanical loading. In the cornea, such physiological loading could theoretically be applied through an act as simple as eye rubbing, or as extreme as corneal wounding during catastrophic injury. The purpose of this study was to determine if TPMD in a single keratocyte can trigger calcium signaling events in neighboring keratocytes.
Calcium waves have previously been reported in corneal epithelial and endothelial cells [20][21][22][23][24][25][26][27] . Trinkaus-Randall's group have found purinergic signaling to be critical for calcium wave initiation and propagation in corneal epithelial cells, with significant input from corneal nerves 21,[25][26][27] . Importantly, corneal epithelium calcium waves could be initiated by mechanical injury, with 200-400 μm wounds in culture dishes being created using an Argon laser under a confocal microscope 24 . Gap junctions do not appear to be involved in corneal epithelial calcium wave propagation 24 . Corneal epithelial calcium waves appear to support reepithelization after corneal injury 26 .
Corneal endothelial cells show early-and late-responding calcium waves 22 . The early waves can be initiated via mechanical stimulation by deforming the membrane of a single cell with a micropipette, and these waves are propagated through purinergic signaling and gap junctions 20,23 . Late-responding calcium waves in the corneal www.nature.com/scientificreports www.nature.com/scientificreports/ endothelium occur around an hour after the fast wave, and are related to cell depolarization associated with sodium transport 22 . Endothelial calcium waves appear to play a role in preventing apoptosis 23 . It should be noted that corneal epithelial and endothelial cells are histologically classified as epithelial cell types, while corneal keratocytes are classified as fibroblasts. Corneal epithelium is embryonically derived from surface ectoderm, while keratocytes and endothelium are neural crest-derived. Each cell type has a very different group of ion channels and transporters. The current study is the first to demonstrate TPMD-induced calcium waves in fibroblast-type cells.
TRPV1 and TRPM8 are the only calcium transporters identified in keratocytes to date 18,19 , and no studies have examined cell to cell calcium signaling in keratocytes. In order to accurately analyze TPMD-induced calcium signaling, we chose to examine several parameters in human and mouse cultured stromal cells, and in keratocytes within human corneal rims. In cultured cells we counted the number of cells demonstrating an increase in Ca ++ i following source cell TPMD to represent the general spread of the response. We measured the area under the fluorescence intensity vs. time curve of the responding cells to demonstrate the magnitude and time course of the response. We chose a threshold of ≥50% change in fluorescence (from baseline) to capture definitive changes in Ca ++ i . Examining a threshold of 25% resulted in very similar, although not identical results. Ours is the first study to report and measure TPMD-induced calcium waves in cells residing within any tissue. We were limited in our ability to quantify the response within human corneal stromas, and thus chose to measure the farthest distance between the source cell and a peripheral cell with a clearly changing Ca ++ i within the visual field. A small confocal pinhole size was utilized to create a thin focal plane. The X-Y field of view typically covered the entire area of the calcium wave. The microscope does not have the temporal resolution to do calcium imaging over a large area in the z plane, which would be required to look at 3D dye spread. For this reason, we intentionally limited imaging, to the best of our ability, to the focal plane of the source cell. www.nature.com/scientificreports www.nature.com/scientificreports/ It is clear that extracellular calcium plays a significant role in the neighboring cell Ca ++ i response in both human and mouse cells and in resident human keratocytes (Table 1, Figs. 2, 3, 4). The TPMD-induced Ca ++ i response was reduced in all calcium-free medium trials. For the human cell tissue study, thapsigargin treatment resulted in a reduced Ca ++ i response, although the difference was not significant. These experiments were adequately powered to see a significant difference. Taken together with the calcium-free medium response, it appears that the TPMD-induced mouse neighboring cell calcium response is dependent on both intra-and extracellular calcium. The only treatment that affected significantly Ca ++ i in both cultured and stromal human cells was Ca ++ -free K-SFM. This would indicate that the response in human cells is more sensitive to extracellular calcium.
As expected, all trials showed a significant reduction of the Ca ++ i signal following addition of thapsigargin and ryanodine to K-SFM. Interestingly, ryanodine reduced the curve area in MCSC, but increased the corresponding cell number. This may reflect a lower dependence of internal Ca ++ i stores as the source of calcium for TPMD-induced calcium waves. This result may also reflect the more sensitive nature of the curve area measurement versus the cell number measurement. The calcium chelator BAPTA-AM, which served as a positive control for these studies, exerted a powerful role by completely inhibiting calcium signaling in HCSC and MCSC.
In examining the pattern of Ca ++ i increases in cells distributed across the visual field, it was not always clear that responding cells had an obvious physical connection to the source cell or other responding cells. This raised the question about whether or not gap junctions were involved in propagating the response between cells. There is a close relationship between gap junctions and Ca ++ . In some cells, a damage-induced increase of Ca ++ rapidly closes gap junctions, leading to isolation of healthy cells from damaged neighbors 28 . Alternatively, we have reported that elevated Ca ++ increases gap junction connectivity between corneal epithelial cells 29 . In astrocytes, mechanically-induced Ca ++ ion increases and their subsequent propagation to neighboring cells were found to have two distinct patterns: (1) persistent increases were found to propagate rapidly via gap junctions in the www.nature.com/scientificreports www.nature.com/scientificreports/ proximal region, and (2) transient increases were found to propagate slowly via extracellular ATP in the distal region 11 . It has also been reported that forced expression of gap junction connexins successfully enables gap junction-deficient cell lines to propagate intercellular Ca ++ waves, stimulating secretion of ATP 5-to 15-fold from the cells 30 . The authors of that study concluded that connexins regulate Ca ++ signaling by controlling ATP release.
We have previously demonstrated that functional gap junctions exist in human keratocytes within intact corneas 3 , and the presence of Cx43 has been previously reported in cultured human corneal fibroblasts 31 . We confirmed that the primary human keratocytes used in this study contained junctional Cx43 by immunostaining. To determine if gap junctions were propagating the TPMD-induced calcium waves in keratocytes, we used 18α-GA,  www.nature.com/scientificreports www.nature.com/scientificreports/ a gap junction communication inhibitor, to disrupt the gap junction pathway between the TPMD targeted source cell and neighboring cells. In human rim tissue, 18α-GA reduced the calcium signaling maximum distance, but no change was found in HCSC. In MCSC, the curve area was reduced by 18α-GA, while the cell number was unchanged. The difference between human cultured cells and tissue might indicate a more prominent role for gap junctions in relaying the TPMD signal between cells in tissue than in cell culture. Combined with the lack of a cultured human cell response to 18α-GA, the fact that only the curve area was affected in mouse cells indicates that gap junctions likely play only a minor role in communicating the TPMD signal in cultured stromal cells. More detailed studies will need to be conducted to confirm these results.
ATP has been shown to act as a primary signaling molecule directing cell to cell TPMD-induced calcium waves in osteocytes 7 . It was recently reported that the production and release of extracellular vesicles rich in ATP is facilitated by intracellular calcium release in osteocytes 32 . As described above, ATP has also been reported to control peripheral calcium wave propagation in astrocytes. In sea urchin embryos, membrane wounding evoked intracellular calcium ion spikes in neighboring cells, and this phenomenon was mimicked by application of ATP and inhibited by apyrase 33 . ATP efflux is observed in both paracrine and autocrine signaling. Secreted ATP binds to cell surface P2X and/or P2Y purinergic receptors, consequently activating voltage-gated calcium ion channels. Calcium channel activation leads to a calcium ion influx which can, in some cells, activate the PLC-IP3 pathway, producing calcium release from intracellular calcium ion stores 33,34 . Looking at TPMD repair, a recent study in bone cells by Mikolajewicz et al. 8 demonstrated that mechanical stresses induced reversible cell membrane injury that resulted in ATP release to the extracellular space, and that ATP-rich vesicles were critical to patching the cell membrane wound. The vesicle release was Ca2+/PLC/PKC-dependent. To examine the influence of ATP in directing keratocyte calcium waves, keratocytes were treated with apyrase. In the present study, apyrase reduced both the cell number and curve area in HCSC (rim tissue was not tested), but did not affect MCSC. These results further demonstrate a species-dependent signaling response following TPMD, as well as a major difference in the TPMD signaling responses between keratocytes and osteocytes and oocytes.
Transient receptor potential (TRP) channels are non-selective cation channels with variable calcium ion selectivity. The plasma membrane TRPV1 and TRPM8 receptors are highly permeable to calcium, and are the only calcium transporters described to date in keratocytes 18 . To determine if TRPV1 or TRPM8 are involved in the keratocyte TPMD-induced calcium wave response, AMG 9810, a selective antagonist for TRPV1, AMTB, a selective antagonist for TRPM8, and BCTC, an antagonist of both the TRPV1 and TRPM8 channels, were applied to cultured human and mouse stromal cells. In MCSC, AMG 9810 resulted in a small but significant reduction in the cell number measurement, but had no effect on the curve area. No other measurements were affected by the TRP inhibitors. Taken together, this indicates that TRPV1 and TRPM8 do not play a role in the regulation of TPMD-induced keratocyte calcium signaling in HCSC, while TRPV1 may play a small role in MCSC signaling.
Cornea rubbing experiments confirmed that TPMDs do indeed occur in keratocytes following minor mechanical manipulation. It was notable that it took a relatively small amount of rubbing to create TPMD in resident keratocytes versus very serious rubbing to initiate keratocyte cell death in both the human and mouse corneas.
In conclusion, this study is the first to describe TPMD and TPMD-induced calcium signaling in corneal keratocytes. It is also the first to describe ATP-mediated calcium signaling and the involvement of intracellular www.nature.com/scientificreports www.nature.com/scientificreports/ calcium signaling in keratocytes. It is clear that both intracellular and extracellular calcium are involved in the response, and that gap junctions and ATP play important roles in transmitting the signal from the source cell to neighboring cells and between neighboring cells. TRPV1 and TRPM8 do not appear to be involved in the response. The TPMD-induced signaling pathways are species dependent, and possibly culture versus tissue dependent. Given that keratocyte TPMD is likely a common occurrence following eye rubbing, in dry eye, and following corneal trauma, the TPMD-induced signaling response is likely a normal aspect of corneal physiology and pathophysiology.

Materials and Methods
Reagents. Active reagents and the concentrations used were selected from previous TPMD literature 7,[35][36][37] and from the TRP channel literature 38  Informed consent was obtained from all donors or their families in accordance with tissue donation protocols. Isolation and culture of primary mouse corneal stromal cells from 4 week old C57BL/6 mouse were performed as above. The culture medium was replaced with fresh DMEM medium containing 10% FBS every 2 days. Cells of passages 3-6 were used in this study. Cells were routinely screened using PCR for the presence of the corneal stromal cell-specific marker thy-1 39,40 . All protocols for animal use and euthanasia were reviewed and approved by the Augusta University Institutional Animal Care and Use Committee and followed the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research.
Immunofluorescence staining. Primary human and mouse corneal stromal cells (HCSC and MCSC, respectively) were fixed in cold 4% paraformaldehyde for 10 min followed by washing in phosphate buffer. Cells were incubated in goat serum (10% in phosphate buffer or PBS) for 60 min, washed in 0.1% Triton X-100/PBS, and exposed overnight to anti-connexin 43 (Cx43; 1:200). After rinsing with 0.1% Triton X-100/PBS, cultures were exposed to secondary antibody conjugated to Alexa Fluor 488 (1:1000) for 1 h and washed in 0.1% Triton X-100/PBS. DAPI was added to each slide with mounting medium. Slides were examined using a Zeiss LSM 780 upright laser-scanning confocal microscope equipped with ZEN lite software.
Single TPMDs were produced in Cal-520-AM pre-stained corneal stromal cells using a Zeiss 780 upright multi-photon microscope, using a procedure slightly modified from that described previously by one of the authors (MML) 9 . Briefly, an 820 nm laser was used to produce a TPMD in a single targeted cell using the microscope's single-time bleaching setting. A circular 9 μm diameter region was targeted on the cell and a single 100% power pulse created the damage. Cal-520-AM fluorescence was visualized and recorded for 50 seconds under a 20x objective (Zeiss W Plan-Apochromat N.A. = 1, working distance = 2.4 mm) using a 488 nm laser. A series of 50 scans was performed at a scan speed of 1 scan/second. 5-10 source cells per culture plate in isolated fields of view were separately targeted for TPMD. The fluorescence intensity of other cells within that field of view that had a notable increase in fluorescent intensity was measured using ZEN 2.1 software by manually tracing those cells and measuring the fluorescent intensity of each (2020) 10:2840 | https://doi.org/10.1038/s41598-020-59570-7 www.nature.com/scientificreports www.nature.com/scientificreports/ cell over the time course of the trial. For the final analysis, only cells that increased their baseline fluorescent intensity by a minimum of 50% were considered positive responders. Both the number of positively responding neighboring cells (cell number) and the area under the fluorescence intensity versus time curve (curve area) of responding cells, which takes into account both the change in Ca ++ concentration along with the duration of that change, were measured. Curve area values were normalized to control for comparisons. The baseline-corrected curve area was calculated using the formula Area = ∑ (I n+1 + I n )/2*(T n+1 − T n ), where, I: fluorescent intensity, T: time. Each experiment included at least 3 plates.
Laser wounding and calcium signaling in fresh human corneal tissue. For all trials except the K-SFM + thapsigargin trial, human keratoplasty corneal rims were divided in half; one half was used as the DMEM control and the other half used for treatment. The rims treated with K-SFM + thapsigargin were compared to unmatched rims in K-SFM + 1 mM calcium. Cal-520-AM staining and drug treatments were performed as described above. After washing with PBS, the corneal tissue was pinned onto a silicone pad within a 35 mm plate with the endothelium surface facing up. The 820 nm laser was used to produce a TPMD in a single targeted cell using the microscope's multiple-time bleaching setting. The bleaching region was set as above with 100% power, with repeated bleaching after every 3 scans. The confocal pinhole setting was set to 1.26 Airy units, corresponding to a 1.80 μm optical section, to minimize out of focal plane cell imaging. All scan settings were set as above.
Image resolution limitations prevented us from using the same measurements used in the cultured cell studies. Instead, for these experiments, the neighboring cell farthest from the source cell with a notable change in fluorescence within the visual field was noted, and the distance from the source cell to that cell to was measured (cell distance). The mean maximum distance from approximately 10 targeted source cells per rim section was calculated and used for statistical analysis. The mean maximum cell distance was compared between DMEM control groups and treatment groups. Assays were repeated in at least 3 rims per experimental group. tpMD after cornea rubbing. TPMDs have never been shown in cells residing within their native tissue. To determine if keratocyte TPMDs can occur as a result of the equivalent of eye rubbing, mice were sacrificed and a mixture of fluorescein-labelled dextran (10 kDa, 0.5 mg/ml) and propidium iodide (PI, Sigma-Aldrich, St. Louis, MO) (0.5 mg/ml) was injected into the anterior chamber. Corneas were then rubbed 30 times with a cotton-tipped swab using moderate pressure to mimic moderate and heavy eye rubbing 42,43 . For positive controls of cell damage/ death, corneas were rubbed 60 times with the rounded bottom end of a scalpel. We determined that this relatively extreme method was required to fatally damage cells using this method. Corneas were then isolated, fixed in 4% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA), stained with DAPI (Santa Cruz Biotechnology, Dallas, TX), and mounted on a microscope slide for multiphoton fluorescence imaging. FITC-dextran can only label cells that have a membrane disruption, while PI stains the nuclei of dead cells, providing a view of which cells had FITC-dextran staining as a result of TPMDs, and which as a result of plasma membrane disruption following cell death. Human corneal rim tissue was subjected to rubbing with a cotton-tipped swab. The same FITC-dextran/PI mixture was added to DMEM in a cell culture dish, the rim was rubbed 0, 30 and 45 times, and the cornea was fixed, stained with DAPI and imaged as above. This protocol was modeled after the mechanical injury (glass beads) protocol described in 7 .
Statistical analysis. All data are provided as the mean ± SE of at least three experiments. The difference between the DMEM control groups and treatment groups was compared using the Student's t-test (GraphPad Prism 7.0), where P < 0.05 was considered statistically significant.

Data availability
The datasets generated during in this study are available from the corresponding author on reasonable request.