A study of wound repair in Dictyostelium cells by using novel laserporation

We examined the mechanism of cell membrane repair in Dictyostelium cells by using a novel laser-based cell poration method. The dynamics of wound pores opening and closing were characterized by live imaging of fluorescent cell membrane proteins, influx of fluorescent dye, and Ca2+ imaging. The wound closed within 2–4 sec, depending on the wound size. Cells could tolerate a wound size of less than 2.0 µm. In the absence of Ca2+ in the external medium, the wound pore did not close and cells ruptured. The release of Ca2+ from intracellular stores also contributed to the elevation of cytoplasmic Ca2+ but not to wound repair. Annexin C1 immediately accumulated at the wound site depending on the external Ca2+ concentration, and annexin C1 knockout cells had a defect in wound repair, but it was not essential. Dictyostelium cells were able to respond to multiple repeated wounds with the same time courses, in contrast to previous reports showing that the first wound accelerates the second wound repair in fibroblasts.

proteins and signals also accumulate as the wound repair machinery 19,[28][29][30] . Despite several models proposed for the molecular mechanism of wound repair 17,31 , the detailed molecular mechanism remains unclear. One reason is that the involved molecular species differ among different species of cells. Therefore, additional cell types must be examined for comparison.
We previously examined wound repair in Dictyostelium cells by partially cutting them with a microneedle 32 . Even when a cell is separated into two portions by cutting, the fragment containing a nucleus can exhibit normal chemotactic movement 33,34 , suggesting that Dictyostelium cells have a wound repair system. However, it is difficult to regulate the wound size and exact timing of the wounding by manually using a microneedle. We recently invented a novel laser-based cell poration method to introduce foreign molecules into single cells 35 . Only a short-term exposure of a low power laser is sufficient for this laserporation method, although a high-power pulse laser or long-term exposure of low-power laser had been used for other cells in previous experiments. In addition, our laserporation only wounds the cell membrane, although the previous laserporation may also wound the inside of the cell.
Previous studies using other organisms have revealed that Ca 2+ in the external medium is necessary for wound repair 11,36 . However, there are few reports addressing basic questions of whether Ca 2+ certainly enters through the wound pore, whether the larger wounds lead to the greater influx of Ca 2+ , when the wound functionally closes, and when the wound becomes even with the uninjured membrane. Unlike poking or cutting with a microneedle as used in previous studies, now we can precisely regulate the wound size using our laserporation method. By using this powerful method, we observed, for the first time, the dynamics of wound pore opening and closing, Ca 2+ influx, and annexins in Dictyostelium cells. These results show that Dictyostelium cells can be an excellent new model organism for wound repair research.

Results
A new application of laserporation to wound the cell membrane. To wound the cell membrane of Dictyostelium cells, we applied a laser-based cell poration method, which we recently invented for the introduction of foreign molecules into live cells 35 . After cells were placed on a coverslip that had been coated with carbon by vapor deposition, a laser beam was focused on a small local spot beneath a single cell under a total internal reflection fluorescence (TIRF) microscope. The absorbed energy of the laser beam by the carbon made a small pore in the cell membrane that was attached to the carbon coat (Fig. 1A). It was easy to focus the laser spot on the surface of the coverslip; when the laser beam was focused on the surface of the coated coverslip, the carbon coat was peeled off, appearing as a small white spot (arrow in Fig. 1B). Figure 1C shows white spots of different sizes (0.5, 1.0, 1.5, and 2.0 µm) on the carbon coat, which were observed by light and scanning electron microscopy. The wound size could be regulated by changing the pinhole size placed in the pathway of the wound laser or changing the laser power. After the focus was set on the surface of the coverslip in this way, the laser beam was applied to the cells attached to the coverslip. Because the laser power used for the actual wound experiments was set much lower, the carbon coat was not peeled off. Thus, as described later, multiple wounding experiments were also possible at the same site.
To visualize the wound pore in the cell membrane, the laser was applied to the cells expressing GFP-cAR1 (cAMP receptor), as a marker of membrane protein. Immediately after the laser application, a black spot appeared at the laser-applied position (Fig. 1D). If the coverslip was not coated with carbon, such black spots did not appear, indicating that the black spot was not due to photobleaching by the laser beam. The size of the black spot was the same as that of the white spot (0.5 µm) in the carbon coat shown in Fig. 1C. Figure 1E shows time course of the size of the black spot after the laser application. The black spot transiently expanded slightly, then shrank, and finally closed within approximately 8 sec (7.41 ± 0.95 sec, n = 10). If the size of the laser spot increased up to 2 µm, most of the cells immediately ruptured (Fig. 1F). Figure 1G shows the survival rate with different sizes of the laser spot. The cells could repair wounds with a size less than 2.0 µm. Influx of fluorescent dye from the wound pore. If the laser beam makes an actual pore in the cell membrane, the fluorescent dye in the external medium will enter the cell through the pore. Propidium iodide (PI), which emits red fluorescence when it binds to RNA and DNA, was included in the external medium BSS, a physiological saline that contains 3 mM Ca 2+ . Figure 2A shows a typical fluorescence microscopy of PI influx from the wound pore. The fluorescence began to increase at the wound site and spread over the cytoplasm. Figure 2B-E show time courses of PI influx with different wound sizes (0.5, 1.0, 1.5, and 1.75 µm). Figure 2F shows a comparison of the time course of PI influx with 4 different wound sizes. Increased fluorescence was observed with larger wound sizes. The fluorescence intensity of PI increased immediately after wounding and ceased to increase within a few seconds, which represents the time of wound closing. Figure 2G shows the half time of closing time with different wound sizes. With a larger wound size (1.75 µm), the influx of PI did not cease for a long time (Fig. 2E). Based on these experiments, we confirmed that the laserporation reliably creates a wound pore in the cell membrane. Ca 2+ influx from the wound pore. It has been reported that Ca 2+ in the external medium is essential for wound repair 8,36,37 . We examined the effect of EGTA, a chelating agent of Ca 2+ , in the external medium. In the presence of EGTA, PI influx did not cease after wounding, whereas it ceased within a short time in BSS (Fig. 3A). Figure 3B shows the survival rate with different wound sizes in BSS and EGTA buffer. In the presence of EGTA, most of the cells ruptured within 5 min after wounding. Thus, Ca 2+ in the external medium is essential for the wound repair of Dictyostelium cells.
Next, we examined whether Ca 2+ entered the cytoplasm through the wound pore from the external medium. To visualize the cytosolic Ca 2+ (Ca i 2+ ), the cells expressing Dd-GCaMP6s, a Ca 2+ sensor consisting of GFP, calmodulin, and a peptide sequence of myosin light chain kinase (M13), were wounded by laserporation. Figure 3C SCIEntIFIC REPORTS | (2018) 8:7969 | DOI:10.1038/s41598-018-26337-0 shows a typical time course of the fluorescence images of Ca 2+ influx. The fluorescence immediately began to increase at the wound site and spread over the cytoplasm, suggesting that Ca 2+ enters the cytoplasm through the wound pore. Figure 3D  . With a larger wound size, the fluorescence increased. In the presence of EGTA, the fluorescence did not increase after wounding (Fig. 3G).
Together, the influx of Ca 2+ through the wound pore contributes to the wound repair.
Regulation of cytosolic Ca 2+ after wounding. The above results in the presence of EGTA indicate that only the wound without Ca 2+ influx does not induce the elevation of Ca i 2+ . However, the influx of Ca 2+ may induce Ca 2+ release from intracellular Ca 2+ stores via a calcium-induced calcium release (CICR) mechanism 38,39 , and this release may also contribute to the wound repair. To examine this possibility, iplA null cells expressing Dd-GCaMP6s were wounded. Inositol 1,4,5-triphosphate receptor-like protein A (iplA) has a homology to inositol trisphosphate receptors, which are mainly localized in intracellular membranous structures, such as the endoplasmic reticulum (ER), although this conclusion is still controversial 40 . These mutant cells show no increase in Ca i 2+ after chemotactic stimulation 41 . Figure   . The amplitude of the fluorescence was significantly reduced compared with that of wild-type cells with 1.0 and 1.5 µm wounding, suggesting that the Ca 2+ release also partly contributes to the increase in Ca i 2+ ( Fig. 4D and F). However, following 0.5 µm wounding, there was no significant difference between wild-type and mutant cells (Fig. 4B). Presumably, a certain level of Ca i 2+ may be required for activation of the CICR mechanism.
The elevated Ca i 2+ shortly returned to a resting level, suggesting that Ca 2+ is pumped out across the cell membrane or sequestered into Ca 2+ stores. Next, we examined the effects of sodium azide, an inhibitor of ATP synthesis. In the presence of sodium azide, Ca i 2+ did not return to a resting level after reaching a peak (Fig. 4G), suggesting that the energy is required for recovery to the resting level. Figure 4H shows a summary of the survival rate of mutant cells and in the presence of sodium azide with different wound sizes. Sodium azide significantly reduced the survival rate, especially with a larger wound size, suggesting that the immediate decrease in Ca 2+ is essential for cell survival. However, iplA null cells showed no significant difference in survival rate compared with wild-type cells, suggesting that the Ca 2+ release by CICR is dispensable for wound repair. Annexin C1 accumulates at the wound site in a Ca 2+ -dependent manner. Annexin, a highly conserved ubiquitous family of Ca 2+ -and phospholipid-binding proteins, has been implicated in wound repair 15,31,42 . Dictyostelium has two annexin genes, annexin C1 (annexin VII or synexin) and annexin C2 (annexin I). Annexin C2 is expressed at approximately 35-fold lower levels compared to annexin C1, whereas both are expressed throughout the developmental stages 43 . There are two isoforms by alternate splicing, 51 kDa (full length) and 46 kDa of the single annexin C1 gene. Both can bind phosphatidylserine in a Ca 2+ -dependent manner 44,45 . Here we examined the dynamics of full length annexin C1 and annexin C2 during wound repair. Figure 5A shows a typical sequence of fluorescence images of a cell expressing annexin C1 (full length)-GFP after wounding, suggesting that it accumulated at the wound site immediately after laserporation. Figure 5B show time courses of the accumulation of GFP-annexin C1, indicating that it rapidly disappeared within approximately 1 sec (1.05 ± 0.26, n = 20) after wounding. In the presence of EGTA, annexin C1-GFP did not accumulated at the wound site (Fig. 5C). Figure 5D shows a dependency of annexin C1 accumulation on the external Ca 2+ concentration, suggesting that annexin C1 significantly accumulated at free Ca 2+ higher than 10 −4 M. Annexin C2-GFP did not show any accumulation at the wound sites (Fig. 5E).
Interestingly, when the cell was ruptured with a larger wound (2 µm), the fluorescence of annexin C1-GFP rapidly spread from the wound site to the whole cell membrane within 1 sec (upper panels of Fig. 5F), whereas the fluorescence was limited at the wound site in successfully repaired cells (Fig. 5A). It has been previously reported that annexin C1 binds to the cell membrane at 10 −4 M Ca 2+ 45 . A large quantity of Ca 2+ influx seems to cause the binding of annexin C1 to the whole inner cell membrane. When the cell was ruptured in the presence of EGTA, the fluorescence of annexin C1-GFP gradually decreased, suggesting that it was released outside the cell without binding to the cell membrane (lower panels of Fig. 5F).
Next we examined wound repair in annexin C1-null cells. When annexin C1-null cells expressing Dd-GCaMP6s were wounded by laserporation (0.5 µm wound), each cell showed different irregular curves with multiple peaks, and Ca i 2+ remained at a higher level (Fig. 5G) compared with wild-type cells (red dotted line). However, they finally returned to the resting level. PI influx was also examined in annexin C1-null cells, also showing irregular curves with multiple peaks in the initial phase (Fig. 5H) compared with wild-type cells (red dotted line). Annexin C1-null cells had a significantly reduced survival rate with larger but not small wound sizes (Fig. 4H).
Taken together, annexin C1 accumulates at the wound site depending on the influx of Ca 2+ and partially contributes to the wound repair.
Cells can respond to wounds multiple times. It has been previously reported that the second wound at the same site reseals more rapidly than the initial wound in fibroblasts 11 . We examined multiple wounding at the same site in Dictyostelium cells. After wounding three times (0, 25, and 50 sec), annexin C1-GFP accumulated each time at almost the same level (Fig. 6A). Figure 6B shows the time courses of the annexin C1-GFP accumulation (n = 12). Figure 6C and D show the time courses of the influx of PI and the fluorescence intensities of Dd-GCaMP6s with three wounds, respectively (n = 8, each). Figure 6E shows a summary of time courses of the three probes, suggesting that Dictyostelium cells seem to have no such facilitated wound repair system in contrast to fibroblasts. If the cells had the second wound at different site, annexin C1-GFP showed a similar time course and amplitude as the initial wound ( Fig. 6F and G).
Together, Dictyostelium cells show reproducible responses each time with multiple wounds.

Discussion
In this study, we examined wound repair in Dictyostelium cells by using our novel laserporation method. Previous laserporation mainly uses an expensive high-power UV laser. Otherwise, a longer time exposure (1-5 sec) of the lower-power laser is used, but the longer exposure cannot be used for fast moving cells such as Dictyostelium cells and the exact wound timing may be obscure. In the present method, a short-term exposure (4-40 msec) of a low-power and low-cost laser is sufficient. As another merit, only the spot area of laser beam in the cell membrane is wounded, whereas the inside of the cell could also be wounded in previous methods. Since the laser used in the present study could not wound the cell membrane without the carbon coat, presumably the energy absorbed in the carbon creates a pore due to thermal energy (heat). The size of a wound can be precisely regulated with our laserporation method. By changing wound sizes, we found that the cells could tolerate less than 2 µm wounds. Although the limit of the tolerable wound size has not been defined in other organisms, the wound sizes vary in other experiments: 25-50 µm for fruit fly embryos, 10-20 µm for frog eggs, 10-25 µm for sea urchin eggs, 0.5-2.5 µm for animal cultured cells, and 0.5 µm for yeast. Larger cells such as eggs and embryos seem to tolerate larger wound sizes. As one simple explanation, since larger cells have a larger volume, they could have a capacity to tolerate greater wounding without irreversible rupture than smaller cells. Another explanation is that larger cells have an additional mechanism, such as an actomyosin-based closing mechanism to repair larger wound, because the surface tension of larger cells is larger than that of smaller cells 32 . In fact, no accumulation of myosin II in wound sites has been reported in smaller cells such as fibroblasts, yeast, and Dictyostelium cells 9,32,46 , whereas larger cells, such as the fruit fly embryo and frog egg, form an actomyosin ring 19,37 .
After laserporation in cell expressing GFP-cAR1, the black area transiently expanded and then shrank ( Fig. 1D and E), which might represent the dynamics of the wound hole. Similar transient expansion of the wound hole was observed in Drosophila embryo 6 . The transient expansion may be caused by the cell membrane tension immediately after wounding. However, the black area persisted approximately 8 sec until closing. In contrast, the PI influx experiments showed that PI influx ceased within 2-3 sec following a 0.5 µm wound. It is difficult to estimate a precise closing time from the Ca 2+ influx experiments because the pumping out of Ca 2+ or sequestering into Ca 2+ stores should be taken into consideration. From the peak time in the presence of sodium azide (Fig. 4G  membrane, excluding the GFP-cAR1 although the wound pore functionally closes within 2-3 sec. Some intracellular small vesicles may fuse to the wound cell membrane, which may exclude GFP-cAR1. In the presence of EGTA, PI influx continued at slower rate after an immediate increase, whereas it ceased in BSS (Fig. 3A). This observation suggests that wound repair without Ca 2+ influx partially functions but it is incomplete. There seems to be at least two mechanisms in wound repair. One is Ca 2+ influx-dependent, which evokes the recruitment of annexin C1, potentially leading to the vesicles accumulation to reseal the wound pore. The other is a Ca 2+ influx-independent pathway, such as rab-5 and rab-11 47 . The present and previous 32 studies showed that at least 0.1 mM extracellular Ca 2+ is required for wound repair for Dictyostelium cells. However, the soil where these cells are normally found may not have such high levels of free Ca 2+ , which also supports the existence of Ca 2+ influx-independent pathway. In addition, it is very interesting from the aspect to evolution that a Ca 2+ -dependent primordial repair mechanism exists in cells that may not normally be exposed to such levels of Ca 2+ .
The present results clarify why annexin C1-GFP locally accumulates only at the wound site. When the cells ruptured with a larger wound, annexin C1-GFP bound to the whole cell membrane (Fig. 5F). In vitro, annexin C1 binds to the cell membrane at 10 −4 M Ca 2+ , but not in the presence of EGTA 45 . Because annexin C1-GFP accumulates in the wound site at concentrations higher than 10 −4 M Ca 2+ in the external medium (Fig. 5D), the concentration of Ca 2+ must locally increase to levels higher than 10 −4 M around the wound site. The fluorescence of annexin C1-GFP increased only at the wound site, but not at the other cell membrane, whereas Ca i 2+ spread and increased throughout the cytoplasm ( Fig. 2A). Presumably, this level of Ca i 2+ will not cause the binding of annexin C1-GFP to the other cell membrane. If the level of Ca i 2+ exceeds 10 −4 M, annexin C1 will bind to the entire cell membrane.
Based on the experiments using iplA null mutants (Fig. 4), the source of Ca i 2+ elevation is not only Ca 2+ influx but also Ca 2+ release from Ca 2+ stores. The present data clearly shows the existence of the CICR system in Dictyostelium cells, as suggested previously 39 . The observation that the survival rate after wounding was not reduced in iplA null cells suggests that the additional increase in Ca i 2+ by CICR followed by Ca 2+ influx is dispensable for wound repair. Moreover, the elevation of Ca 2+ throughout the cytosol also may not be necessary for local wound repair.
The experiments using sodium azide suggest that energy is required to return the Ca i 2+ to the resting level, which is essential for cell survival. The responsible Ca 2+ stores and membrane pump remain to be determined in the future. The present wound method can serve also as a powerful tool to investigate cytosolic Ca 2+ homeostasis.
Dictyostelium cells can survive after microneedle cutting as shown in previous experiments [32][33][34] . The wound size by microneedle cutting could be much larger than the limit of wound size examined in the present study (less than 2 µm). However, for microneedle cutting, the cell is gently pressed with a microneedle and then cut by swiftly pulling it. This pressing may fuse the upper and lower cell membranes; thereby the actual wound pore may be smaller than the apparent.
In our previous experiments by microneedle cutting, actin begins to accumulate at approximately 2-3 sec and peaks at about 6-10 sec after wounding 32 , which is significantly slower than the timing of annexin C1 accumulation. In future, the connection between Ca 2+ influx, annexin C1 accumulation, and cytoskeletal accumulation must be investigated.
The reaction of wound repair is attractive for cell biology because it is inducible, dynamic, and repeatable, involving membrane trafficking, membrane fusion, and dynamics of the cytoskeleton 48 . Dictyostelium has been recognized as a model organism for research of cell biology, such as cytokinesis, phagocytosis, cell migration, and chemotaxis. There are many available resources of knockout mutants of cytoskeletal proteins, intracellular signals, and membrane traffic proteins, most of which are almost the same as those found in higher animal cells. Since the genome of Dictyostelium is haploid, it is easy to generate knockout mutants and expression of tag genes. It is also valuable to know the evolution of wound repair mechanism.
We believe that Dictyostelium cells will become a new member of excellent model organisms for research in wound repair.
Plasmid construction. The vectors of annexin C1-GFP and annexin C2-GFP were newly constructed. Each gene was amplified from cDNA library (a gift from Dr. D. Robinson) and inserted into the pA15GFP expression vector including a G418 resistance gene. Dd-GCaMP6s was used as a Ca 2+ probe, in which the codon usage in the original GCaMP6s 51 was optimized for Dictyostelium. To generate the knockout construct of annexin C1, annexin C1 fragment was subcloned outside of two loxP sites in pLRBLP. The left arm (nucleotides 50-600) and right arm (nucleotides 619-1196) of annexin C1 gene were amplified from the cDNA by PCR using the following primer sets with restriction enzyme sites (underscore): 5′-ATGTCGACCACAACAAGGTTATCCACCACAACAAGGCT-3′ (SalI) and 5′-ATATCGATGTGAATTTGTTCAACATCGAAATGAGCTGG-3′ (ClaI) (for the left arm of the KO construct); and 5′-ATGGATCCGGTACCAACGAGAACACTATAATTGAAATTTTAG-3′ (BamHI) and 5′-ATACTAGTGCTAATGAATTCTTGAAGAGAGTTGAATAAGC-3′ (SpeI) (for the right arm of the KO construct). The amplified left arm fragments were subcloned between SalI and ClaI sites lying upstream of N-terminal loxP site in pLRBLP. The right arm fragments were subsequently subcloned between BamH1 and SpeI sites lying downstream of C-terminal loxP site. After transformation with the knockout construct, cells were selected in HL5 medium containing 10 µg/ml of blasticidin S hydrochloride (Wako).
Carbon coating and chamber preparation. The surface of the coverslip of a glass-bottom chamber was coated with carbon by vapor deposition using a vacuum evaporator (JEOL, JEE-400). The thickness of the carbon layer was approximately 20 nm. To make the surface hydrophilic, the surface of the carbon-coated coverslip was SCIEntIFIC REPORTS | (2018) 8:7969 | DOI:10.1038/s41598-018-26337-0 activated with a plasma treatment. The chamber was sterilized with 70% ethanol and dried if necessary. The cells were settled on the surface of the carbon-coated coverslip, and they were slightly compressed with agarose block (1.5%, dissolved in BSS, 1 mm thick) to observe the ventral cell surface 52 . Laserporation. The fluorescence images of cells expressing GFP-proteins were observed under a total internal reflection fluorescence microscope (TIRF, based on IX71 microscope, Olympus) as previously described 53 . For laserporation, a nanosecond-pulsed laser (FDSS532-Q, CryLas) beam was directed toward the sample using a dichroic mirror through the TIRF microscope as previously described 35 . The laser beam (wavelength, 532 nm) was operated at a 15 mW output power and pulse width of 1 nsec and attenuated to 1/300 via passage through several neutral density filters. A 60 × (PLAPON60XOTIRFM, Olympus, NA = 1.45) objective was used to focus the laser beam on the surface of the carbon-coat. When the laser beam was applied by properly adjusting the focus, the carbon coat was peeled off, appearing as a small white spot (Fig. 1C). The x-y position of the laser spot was fixed in the center of the microscopic field. The position of the cells was moved using a piezo-motorized x-y stage on the microscope. The duration of the laser beam application was set at 4-40 msec and controlled by a shutter. The wound size was regulated by changing the laser power or the size of pinhole, which was inserted on the pathway between the laser and dichroic mirror. Time-lapse fluorescence images were acquired with 50-100 msec exposure times at 200-500 msec intervals using a cooled CCD camera (Orca ER, Hamamatsu Photonics).
Scanning electron microscopy (SEM). The laser beam was applied on the surface of carbon-coated coverslip to generate various-sized white spots. After acquiring the images of white spots under a light microscope, the same coverslip was directly attached to the SEM mount folder with double-sided carbon sticky tapes to remove negative charges. The coverslip was observed under a scanning electron microscope (JEOL, JSM-6360LA).

Measurement of the influx of propidium iodide and Ca 2+ .
To characterize the opening and closing of wound pores, 0.15 mg/ml of propidium iodide (PI, Sigma) was included in the external medium (BSS). When PI enters the cytoplasm, it emits red fluorescence by an argon laser (488 nm) under TIRF microscopy.
To monitor the cytosolic Ca 2+ , cells expressing Dd-GCaMP6s were wound by laserporation. This Ca 2+ probe emits green fluorescence (>515 nm) under illumination by an argon laser (488 nm), depending on the Ca 2+ concentration.
Sodium azide (Wako) was dissolved in distilled water to generate a 100 mM stock solution. A small aliquot of the stock solution was applied to the surface of the agarose blocks to make the indicated final concentration (5 mM). Cells were incubated with sodium azide from 30 min before the wound experiments. Image analysis. The acquired images were analyzed with Image J software (http://rsbweb.nih.gov/ij/). The time course of relative fluorescence intensity at the wound region was calculated by dividing the fluorescence intensity at the wound site by that of the other cytoplasmic region after each value was subtracted from the background. The graphs of the time course of the relative fluorescence intensity were created using GraphPad Prism 7 (GraphPad Inc., USA) after calculation using Microsoft Excel.
To examine the half time of PI influx to reach a maximum after wounding, the time course of relative fluorescence intensity (Fr) was fitted to following formula (1) for a single exponential rise to maximum: where a is amplitude of exponential, b is rate constant, t is time (sec), and c is constant. The half time (t 1/2 ) is calculated using the following formula (2) Ca-EGTA buffer. To formulate medium containing the indicated free Ca 2+ , a Ca-EGTA buffer (10 mM KCl, 10 mM NaCl, 3 mM MES, 10 mM EGTA and an appropriate concentration of CaCl 2 , pH 6.3) was used. The concentration of CaCl 2 in the Ca-EGTA buffer was calculated by Ca-EGTA Calculator v1.3 (http://maxchelator. stanford.edu/CaEGTA-TS.htm). The agarose block for the agar-overlay was made by dissolving 1.5% agarose in Ca-EGTA buffer or BSS as previously described 52 .
Statistical analysis. Statistical analysis was performed using GraphPad Prism 7. Data are presented as the mean ± SD and analyzed by one-way ANOVA with Tukey's multiple comparison test.
Data availability. All relevant data are available from the authors on reasonable request.