Calcium Signaling in the Photodamaged Skin: In Vivo Experiments and Mathematical Modeling

Abstract

The epidermis forms an essential barrier against a variety of insults. The overall goal of this study was to shed light not only on the effects of accidental epidermal injury, but also on the mechanisms that support laser skin resurfacing with intra-epidermal focal laser-induced photodamage, a widespread medical practice used to treat a range of skin conditions. To this end, we selectively photodamaged a single keratinocyte with intense, focused and pulsed laser radiation, triggering Ca²⁺ waves in the epidermis of live anesthetized mice with ubiquitous expression of a genetically encoded Ca²⁺ indicator. Waves expanded radially and rapidly, reaching up to eight orders of bystander cells that remained activated for tens of minutes, without displaying oscillations of the cytosolic free Ca²⁺ concentration (). By combining in vivo pharmacological dissection with mathematical modeling, we demonstrate that Ca²⁺ wave propagation depended primarily on the release of ATP, a prime damage-associated molecular patterns (DAMPs), from the hit cell. Increments of the in bystander cells were chiefly due to Ca²⁺ release from the endoplasmic reticulum (ER), downstream of ATP binding to P2Y purinoceptors. ATP-dependent ATP release though connexin hemichannels (HCs) affected wave propagation at larger distances, where the extracellular ATP concentration was reduced by the combined effect of passive diffusion and hydrolysis due to the action of ectonucleotidases, whereas pannexin channels had no role. Bifurcation analysis suggests basal keratinocytes have too few P2Y receptors (P2YRs) and/or phospholipase C (PLC) to transduce elevated extracellular ATP levels into inositol trisphosphate (IP₃) production rates sufficiently large to sustain oscillations.

Graphical Abstract

Graphical Abstract.

Introduction

The epidermis forms an essential barrier against dehydration and pathogens, as well as light, chemical and mechanical injury.^1,2 It is a stratified epithelium largely composed of keratinocytes (95%), melanocytes (that donate pigment to the keratinocytes), Langerhans’ cells, which have immunological functions, and Merkel cells, also known as Merkel-Ranvier cells or tactile epithelial cells.² It is well established that epidermis’ continuous renewal is sustained by stem cells contained in the stratum basale.^3–8 The differentiating keratinocytes are gradually displaced in the outward direction, from stratum basale through stratum spinosum to stratum corneum, where corneocytes are continually shed from the skin surface.^1,2 In different epithelia, including the epidermis, damage triggers perturbations of the that spread from cell to cell, known as (intercellular) Ca²⁺ waves (see e.g. Refs.^9–14) and considered a fundamental mechanism for coordinating multicellular responses.¹⁵ Ca²⁺ signaling is at the core of epidermal homeostasis,^16–18 however, the mechanisms underlying Ca²⁺ wave propagation and their significance in the damaged epidermis are incompletely understood. A gradient of Ca²⁺ across the epidermis is key for keratinocyte differentiation and formation of the epidermal permeability barrier,^16–18 maintenance of which relies on a delicate balance between proliferation and differentiation, two Ca²⁺-dependent cellular processes.^2,8,19 Defects of epidermal Ca²⁺ homeostasis cause skin pathologies, such as Darier's disease for which ATP2A2 has been identified as the defective gene, indicating that the sarcoplasmic/endoplasmic reticulum Ca²⁺ ATPase (SERCA)2 plays a key role.²⁰

ATP, which is present at mM concentration in the cytosol whereas its normal concentration in the extracellular environment is in low nM range, is released by most cells as an extracellular signaling molecule.^21,22 ATP release was detected in cultured human neonatal keratinocytes exposed to air,²³ as well as from injured mouse epidermal keratinocytes,²⁴ normal human epidermal keratinocytes (NHEKs) co-cultured with mouse dorsal root ganglion (DRG) neurons,²⁵ and HaCaT cells²⁶ (a spontaneously transformed aneuploid immortal keratinocyte cell line from adult human skin, with high capacity to differentiate and proliferate in vitro²⁷). Extracellular ATP can affect the by activating Ca²⁺-permeable P2XRs²⁸ and/or by promoting Ca²⁺-release from the endoplasmic reticulum (ER) via G-protein coupled P2YRs^29,30 on the surface of keratinocytes and other epidermal cells types.³¹ Therefore, ATP release and purinergic signaling have the potential to interfere with epidermal homeostasis and have been implicated in a host of processes, including pain, inflammation and wound healing.³¹

In this study, we evoked Ca²⁺ waves by selectively photodamaging a single keratinocyte of the epidermal basal layer of the earlobe skin in live anesthetized mice. Our experimental model is related not only to accidental injury, but also to the medical practice of skin resurfacing/rejuvenation based on intra-epidermal focal laser-induced photodamage that is used to treat numerous conditions such as photodamage and acne scars, hidradenitis suppurativa and posttraumatic scarring from basal cell carcinoma excision.^32–35 As detailed in the following Methods and Results sections, we used intravital multiphoton microscopy to visualize Ca²⁺ waves in mice expressing GCaMP6s, a sensitive and selective genetically encoded Ca²⁺ indicator.³⁶ To dissect the molecular components contributing to Ca²⁺ wave propagation, we performed in vivo pharmacological interference experiments by intradermal microinjection of different drugs. With the insight gained from data analysis, we formulated a mathematical model that accounts for the observed Ca²⁺ dynamics based on ATP release and diffusion, and activation of P2YRs. The significance of our findings for wound healing and skin therapy is discussed in relation to Ca²⁺ signaling pathway activation in the epidermis.

Methods

Animals

Mice were bred and genotyped at the National Research Council-Institute of Biochemistry and Cell Biology (CNR-IBBC), Infrafrontier/ESFRI-European Mouse Mutant Archive (EMMA), Specific Pathogen-Free (SPF) barrier unit (Monterotondo Scalo, Rome). All the experimental procedures were agreed upon, reviewed and approved by local animals welfare oversight bodies and were performed with the approval and direct supervision of the CNR-IBBC/Infrafrontier—Animal Welfare and Ethical Review Body (AWERB), in accordance with general guidelines regarding animal experimentation, approved by the Italian Ministry of Health, in compliance with the Legislative Decree 26/2014 (ref. Project license 603/2018-PR), transposing the 2010/63/EU Directive on protection of animals used in research). In addition, all animal experimentation was conducted in adherence to the NIH Guide for the Care and Use of Laboratory Animals and recommendations from both ARRIVE and PREPARE guidelines.^37,38 Mice were housed in individually ventilated caging systems (Tecniplast, Gazzada, Italy) at a temperature (T) of 21 ± 2°C, relative humidity (RH) of 55 ± 15% with 50–70 air changes per hour (ACH) and under controlled (12 : 12 hour) light–dark cycles (7 am–7 pm). Mice had ad libitum access to water and a standard rodent diet (Emma 23, Mucedola, Settimo Milanese, Italy).

Adult transgenic mice, both male and female, aged between 5 and 40 weeks and ubiquitously expressing the Ca²⁺ biosensor GCaMP6s³⁶ were used for in vivo Ca²⁺ imaging experiments. These mice were generated in the animal facility of the laboratory by crossing the Jackson Laboratory strain #024106 (STOCK-Gt(ROSA)26Sor^{tm96(CAG−GCaMP6s)Hze}/J) with the European Mouse Mutant Archive (EM) Cre-deleter strain EM:01135 (B6.C-Tg(CMVcre) ^1Cgn/CgnCnrm). Double mutant mice were identified amplifying tail genomic DNA by means of PCR. The presence of the CAG-GCamp6 insertion in the Rosa26 locus and the wild type allele were detected using the primers:

• 5’-ACG-AGT-CGG-ATC-TCC-CTT-TG-3’, 5'- AAG-GGA-GCT-GCA-GTG-GAG-TA-3'

and

• 5’-CCG-AAA-ATC-TGT-GGG-AAG-TC-3’.

The CRE-deleter transgene was detected using the primer pair

• 5’-CGA-GTG-ATG-AGG-TTC-GCA-AG -3’

and

• 5’-TGA-GTG-AAC-GAA-CCT-GGT-CG -3’.

PCR products were run on 2% agarose gels, visualized with ethidium bromide and photographed. Expected band sizes were 450 bp for the CAG-GCamp6 and 297 bp for the wt allele and 390bp for the Cre transgene.

Multiphoton microscopy

System description

We used a custom-made multiphoton system (Figure S1) based on a Bergamo II architecture (Thorlabs Imaging System, Sterling, VI, USA), as previously described.¹² The system was equipped with two scanning heads, one with resonant-galvo (RG) mirrors and the other with galvo-galvo (GG) mirrors, and was coupled to a mode-locked titanium-sapphire (Ti:Sa) fs pulsed laser (Chameleon Vision II Laser, Coherent, Inc., Santa Clara, CA, USA) (Figure S2). The RG scanner was used for imaging, whereas the GG scanner was used to focally photodamage a pre-defined spot in the field of view (FOV) by focusing the collimated laser beam onto the sample through a 25× water- immersion objective (XLPLN25XWMP2, NA 1.05, Olympus Corporation, Tokyo, Japan; the same objective was also used for imaging).

Multiphoton excitation of GCamp6s was performed at 920 nm, whereas its emission signal was filtered by a single band-pass filter (Cat. No. FF02-525/40-25, Semrock/IDEX, Rochester, NY, USA) placed in front of a non-descanned GaAsP detector (Cat. No. H7422–50, Hamamatsu Photonics K.K., Shizuoka, Japan). Electro-optical modulators (EOM) and mechanical ultra-fast shutters were used to control both photodamage and imaging light exposure using the ThorImage LS 3.1 software (Thorlabs). Parameters used for photodamage and image acquisition are listed in Table 1. Laser excitation intensity and frame averaging were adjusted to minimize photobleaching and phototoxicity, while achieving enough signal to noise ratio and temporal resolution.

Table 1.

Photodamage and image acquisition parameters

Parameter	Value
Excitation and photodamage wavelength	920 nm
Photodamage optical power in the focal plane	118 mW
Photodamage duration	500 ms
Average focal plane optical power used for imaging	25 mW
Frame size (pixels)	512 × 512
Pixel dwell time	81.2 ns
Raw frame rate	45.7 fps
Number of averaged frames	9
Effective frame rate	2.5 fps
Resolution	0.62 µm/pixel

Intravital multiphoton microscopy and drug delivery

Mice were anesthetized with an intraperitoneal (i.p.) injection of physiological solution containing 90 mg/kg ketamine and 0.5 mg/kg medetomidine. If the experiment duration was longer than 2 hours, half-dose of the anesthesia was reinjected to avoid mouse awaking. Mice were positioned on the stage of the multiphoton microscope equipped with a heated pad kept at 35°C for homeothermic control. Ca²⁺ signals were recorded from the basal keratinocytes of the mouse pinna (earlobe) skin, fixed under the objective with double-sided tape. For optical coupling, a drop of phosphate buffered saline (PBS) solution was placed between the skin and the 25× water-immersion objective lens. The PBS solution was composed of 10 mM Phosphate (as sodium phosphates), 2.68 mM KCl and 140 mM NaCl.

After positioning the animal under the microscope objective lens, we waited 10 minutes to allow intracellular Ca²⁺ (mobilized during earlobe manipulation) to return to baseline (as judged by low levels of GCaMP6s fluorescence). For each mouse, control photodamage experiments were repeated n  ≥ 3 times with similar results in different positions of the earlobe. Next, to block or enhance a specific pathway, the ear surface was gently cleaned and dried and a 4 µl solution containing a selected drug (Table 2) diluted in PBS was microinjected in the earlobe skin using a 10 µl NANOFIL syringe (World Precision Instruments Inc., Sarasota, FL, USA) equipped with a gauge 33 needle. The microinjected PBS solution also contained 1.78 µM of fluorescent Dextran Texas Red (molecular weight, MW = 70,000 Da; single-photon Excitation/Emission wavelengths = 595/615 nm; two-photon excitation wavelength = 920 nm) to confirm that the drug-containing fluid reached the imaged area. Thereafter, the same photodamage protocol used in the previous control recording was used to acquire image sequences in the presence of the selected drug. A dedicated series of experiments was performed to verify that photodamage-evoked Ca²⁺ waves were insensitive to the presence of the microinjected fluorescent Dextran Texas Red. Due to its large MW, Dextran diffused slowly through the tight layers of keratinocytes and accumulated mainly in the extracellular spaces below the stratum basale. The investigator was not blinded during administration of treatments or result assessment. No samples were excluded from analyses. To minimize animal stress and avoid cross-interaction between different drugs, no more than one drug was injected in the same earlobe.

Table 2.

Details of drugs used to dissect components of Ca²⁺ wave propagation mechanisms. The last column reports the serial number of the mice used for the corresponding experiments.

Reagent	Concentration	Part number	Supplier	Mouse #
Dextran Texas RedTM	1.78 µM	D1830	ThemoFisher Scientific	1–18
Apyrase	500 U/ml	A2230	Merck & Co.	3, 4
PPADS	625 µM	P178	Merck & Co.	5, 6
ARL 67156	400 µM	A265	Merck & Co.	7, 8
EGTA	5 mM	E3889	Merck & Co.	9, 10
Thapsigargin	400 nM	T9033	Merck & Co.	11, 12
CBX	400 µM	C4790	Merck & Co.	13, 14
Probenecid	4 mM	P8761	Merck & Co.	15, 16
TAT-Gap19	200 µM	A230580	GenoSphere	17, 18

Images acquired with different PMTs of the multiphoton microscope are shown in Figure S3 together with the corresponding composite images. In general, a given focal plane intercepts furrows and bulges, exposing a multitude of cell populations and structures while performing intravital microscopy in the skin. Cells corresponding to different epidermal layers can be distinguished morphologically: smaller polygonal cells are keratinocytes of the stratum basale, while increasingly larger cells correspond to the suprabasal layers. Second-harmonic generation (SHG) images were acquired at 460/50 nm emission wavelength, allowing the imaging of collagen fibers (Figure S3 A, E). GCaMP6s fluorescence (and keratinocytes autofluorescence) contributed to the signal detected in the 525/40 nm emission channel (Figure S3 B, F). Images in Figure S3 C, G were acquired in the 647/70 nm emission channel respectively before and after the microinjection of fluorescent Dextran Texas Red.

Image processing and data analysis

To establish a relationship between the and GCaMP6s fluorescence emission F, we treated GCaMP6s as a Ca²⁺ buffer (denoted as G) that is present in the cytosol at a concentration [G]. Chelation of Ca²⁺ by buffer G to form a complex CaG is described by the reaction:

and the corresponding kinetic equation:

(E1)

where square brackets are used to indicate concentration, is the rate constant for Ca²⁺ binding to G and is the rate constant for Ca²⁺ dissociation.³⁶

At chemical equilibrium , therefore

(E2)

where is the equilibrium or dissociation constant, thus

(E3)

Assuming that [G] is low enough, F can be written as the following linear combination:

(E4)

where the proportionality constant S_f and S_b lump all factors such as the illumination intensity, the efficiency of fluorescence emission, the molar absorption coefficient, and the length of the absorbing medium traversed by the illuminating beam. If c_T denotes the total (constant) fluorophore concentration, the mass balance equation is:

(E5)

Defining the emission under Ca²⁺-saturating and Ca²⁺-free conditions respectively as and , the expression for F can be re-written as

(E6)

yielding

(E7)

Combining equations E3 and E7, we conclude that

(E8)

Equation E8 can be used to estimate the change in the ,

(E9)

where is the pre-stimulus (baseline) concentration. For changes that fall within the approximately linear region of the input/output relation, we may write:

(E10)

where is pre-stimulus fluorescence and

(E11)

Thus, if imaging experiments are performed in such a way that photobleaching is negligible, and as long c_T and the optical path length do not change during the measurement, the pixel-by-pixel ratio

(E12)

is a unique function of the stimulus-induced .³⁹

On this ground, image processing and data analysis used to relate and were carried out using the open source ImageJ software, the MATLAB programming environment (R2015a, The MathWorks, Inc., Natick, MA, USA) and Vimmaging, a custom-made software developed under the MATLAB environment by Catalin D. Ciubotaru and Fabio Mammano. To automatically select cell boundaries in an image obtained by averaging all frames in a calcium wave, we used roiSelection, a Matlab graphics user interface (GUI) developed by Catalin D. Ciubotaru.

To eliminate contributions independent of GCaMP6s expression, all image analyses were preceded by background subtraction. To define the background-subtraction procedure, four images of basal layer keratinocytes were recorded from a wild type mouse (not expressing GCaMP6s, Figure S4) under the same conditions used for all other experiments (see Table 1). Histograms of pixel intensity were extracted from regions of interest (ROIs) of 76×76 pixels with ImageJ and fitted with a Gaussian distribution

(E13)

The fitting parameters for wild type mouse epidermis were μ₁ = 440 ± 1 and σ₁ = 31 ± 4.

Next, we examined basal layer keratinocytes in a cohort of n = 15 GCaMP6s-expressing mice. The distribution of baseline pixel intensity values in ROIs composed of 76×76 pixels, averaged over the first 5 frames preceding the photodamage stimulus (that occurred at frame 6), followed a bimodal distribution that was well fitted by the sum of two Gaussians, G₁ + G₂, where

(E14)

The parameters μ₁ = 440 ± 2 and σ₁ = 30 ±1 (mean ± s.e.m.) of G₁ for the GCaMP6s samples were not significantly different from those of the wild type control (p = 0.98 for μ₁; p = 0.5 for σ₁; t-test). In contrast, the parameters μ₂ = 697 ± 9 and σ₂ = 206 ± 5 (mean ± s.e.m.) of G₂ differed significantly from those of the wild type control (P = 2×10^–10 for μ₁; P = 2×10^–12 for σ₁; t-test). Therefore, we attributed the pixel intensity distributions corresponding to G₁ to the sum of instrument noise/offset plus keratinocyte autofluorescence, and the distributions corresponding to G₂ to GCaMP6s pre-stimulus (baseline) fluorescence. Based on this analysis, the intersection between the two fitting curves (Figure S5, vertical dash-dotted lines), was selected as the threshold value to be subtracted from each image of the corresponding sequence. All pixel values below threshold were set to 1, thus eliminating autofluorescence and the instrumental offset. Thereafter, Ca²⁺ signals evoked by photodamage were quantified as the time-dependent background-subtracted pixel-by-pixel relative changes in fluorescence, , where F₀ is the pre-stimulus fluorescence obtained by averaging over the first 5 (pre-stimulus) frames.

Statistical analysis

Mean values are quoted ± standard error of the mean (s.e.m.). Comparisons of means for non-Gaussian sample distributions were made using the Wilcoxon Rank sum test (as implemented in the Matlab function ranksum). For samples that had Gaussian distribution, the two-sample t-test (as implemented in the Matlab function ttest2) was adopted; P = P-value < 0.05 was assumed as statistically significant.

Immunofluorescence

Adult mice were euthanized by cervical dislocation. Hair was gently removed from earlobe skin by applying hair removal cream for 3 min. Thereafter the cream was removed with cotton buds and the skin surface was rinsed with PBS. Earlobes were then excised and immersed in PBS.

For whole mount preparations, skin (dermis and epidermis) was mechanically detached from the underlining cartilage and transferred in a Petri dish containing ice-cold PBS. Dermal fat was gently peeled off with a scalpel. The skin sample was then divided in about 3 × 3 mm² portions, immersed in ice-cold PFA (4%) and shaken gently for 1 hour at 4°C.

For slice preparations, ∼1 mm-thick slices were obtained from the earlobe sample by free-hand cutting with a scalpel. Slices were then immersed in ice-cold PFA (4%) and shaken gently for 1 hour at 4°C.

After fixation as described above, whole mount and slice preparations were washed 3 times for 5 min in 1 mL of PBS and incubated for 30 min at room temperature in blocking buffer [PBS complemented with 10% heat-inactivated fetal bovine serum (FBS, 10% v/v, Cat. No. 10270-106, Thermo Fisher Scientific) and 0.2% Tryton X-100 (TX-100, #X100, Sigma)]. Samples were then incubated overnight in a humidified chamber at 4°C with primary anti-P2Y₁R or anti-P2Y₂R antibodies (#APR-009 and #APR-010, respectively, Polyclonal, knock out-validated, Alomone Labs, Jerusalem, Israel) diluted 1:100 in blocking buffer. The next day, samples were washed 3 times for 15 min at room temperature with washing buffer (2% FBS diluted in PBS with 0.2% TX-100) and incubated with cross-adsorbed Alexa Fluor 488 secondary antibody (1:500, #A-11008, Thermo Fisher Scientific) for 1 h at room temperature. Then, samples were washed 3 times for 15 minutes in washing buffer at room temperature and stained with DAPI (5 µM in washing buffer for 5 min). Samples were finally mounted in a liquid anti-fade medium (Mowiol 4–88) between a microscope slide and a coverslip. After 2 hours, coverslips were sealed with nail polish.

Fluorescence images acquisition was performed with a confocal TCS SP5 microscope (Leica) equipped with a 63× oil immersion objective (HC PL Apo, UV optimized, NA 1.4, Leica). DAPI, Alexa 488 and autofluorescence signal were excited with 405 nm, with 488 nm and with 543 nm laser light, respectively. DAPI fluorescence was collected between 415 nm and 480 nm; Alexa Fluor 488 fluorescence was collected between 500 nm and 540 nm; skin autofluorescence was collected between 625 nm and 700 nm. Images were acquired by averaging each line 32 times in a line-by-line sequential protocol (total pixel dwell time about 2.5 ms; pixel size, 120 nm).

Mathematical model

The model we developed is based on radial ATP diffusion from the damaged cell, signaling via P2YRs, PLC and IP₃Rs, and the following geometrical arrangement. The photodamaged cell was modeled as a cylinder with radius 5 µm, height 10 µm and center at r = 0 µm. The upper surface of the cell hit by laser radiation was assumed to burst open at time t = 0 seconds, so that its cytosolic ATP content diffused first in the space just above the cell, then radially in the thin (20 nm) clefts between the tightly packed sheets of epidermal keratinocytes.⁴⁰ We assumed 2-dimensional radial symmetry and discretized the inter-sheet space into 20 rings of radial length = 10 µm, corresponding to the cellular diameter. ATP in each ring stimulated IP₃ production (via P2YRs and PLC) in the adjacent cells above and below the ring, and – vice versa – these cells were allowed to release ATP via connexin HCs into the inter-sheet clefts. The equations for ATP dynamics are⁴¹  

(M1)

where A_c is the ATP concentration in the photodamaged cell, A₀ is the ATP concentration in the disk above the damaged cell, and A_j is the ATP level in ring number j of the inter-sheet space, D is the effective diffusion constant of ATP, h = 0.02 µm is the height of the space between cellular sheets,⁴⁰  k is the degradation rate of ATP, and J_HC,j is the flux of ATP via HCs, which is assumed to be a steep Hill-function of the intracellular calcium concentration (Ca) in cell j,⁴²  

(M2)

ATP binding to P2YRs activates PLC, which produces IP₃ from PIP₂. All isoforms of PLC depend on Ca²⁺,⁴³ and in particular PLC-δ, which is expressed in keratinocytes⁴⁴ and is activated by intracellular Ca²⁺.^45–47 We therefore introduced a feedback term from Ca²⁺ onto the rate of IP₃ production. IP₃ dynamics in cell j was described by the equations

(M3)

 

(M4)

where and s.

Finally, we modelled cytosolic Ca²⁺ dynamics by treating the cell cytoplasm as a well stirred closed compartment. Hence, the equation for cytosolic free Ca²⁺ in cell j reads

(M5)

where J_IP3R is IP₃-dependent Ca²⁺ flux from the ER to cytosol through IP₃Rs, which we modelled in accord with the Li-Rinzel model⁴⁸ (see below),

(M6)

represents IP₃-independent Ca²⁺ flux from the ER to cytosol, and

(M7)

represents pumping from the cytosol to the ER by sarco/endo-plasmic reticulum Ca²⁺-ATPases (SERCA pumps).

The free Ca²⁺ concentration in the ER is linked to the by the conservation equation

(M8)

where α denotes the ratio of ER to cytosol volume and Ca_TOT stands for total Ca²⁺ present within the j-th cell (part in the cytosol, part in the ER). Further, J_buff is used to model Ca²⁺ buffering, where, for simplicity, endogenous buffers are assumed to have total concentration twice the concentration of GCaMP6 and to have the same kinetics as GCaMP6. These assumptions lead to

(M9)

The Li-Rinzel model for IP₃ receptors⁴⁸ assumes that Ca²⁺ and IP₃ rapidly and independently activate the channels, whereas Ca²⁺ more slowly inactivates the channels. The Ca²⁺ flux from ER to cytosol through the IP₃Rs in cell j was modeled as

(M10)

where the inactivation variable h follows Hodgkin-Huxley-like dynamics,

(M11)

To relate model and experimental GCaMP6 signals, we used the following equation representing binding/unbinding of cytosolic Ca²⁺ to GCaMP6s,

(M12)

normalized the simulated CaG traces within the interval [0,1] and compared them to the corresponding normalized experimental traces.

The differential equations listed above were solved by using the stiff ode15s solver in MATLAB^TM version 2019a (The MathWorks Inc., Natick, MA, USA). Model parameters in Table 3 were chosen to give reasonable fits, both quantitatively and qualitatively, to a series of experiments.

Table 3.

List of model parameters

Parameter	Unit	Value	Description
D	µm2/s	65	Effective ATP diffusion constant
k	s–1	0.27	ATP degradation constant
a	µm	10	Cell diameter (and radial discretization step size)
h	µm	0.02	Height of space between adjacent cellular sheets
VHC	µM/s	1200	Maximal ATP release rate though connexin HCs
VPLC	µM/s	1.85	Maximal rate of PLC production
KPLC	µM	1.05	ATP EC50 value for PLC production
kIP 3	s–1	15	IP3 degradation rate
ψmax	-	0.6	Maximal feedback strength from Ca2+ onto PLC
τψ	s	30	Time-constant of feedback from Ca2+ onto PLC
L	s–1	0.5	Leak rate
VSERCA	µM/s	65	Maximal SERCA pump rate
α	-	0.1	Ratio between cytosolic and ER volumes
CaTOT	µM	2	Total amount of free cellular Ca2+
kIP 3 R	µM/s	42625	Maximal IP3R release rate
Ki	µM	1	Li-Rinzel model parameter for IP3 receptor activation by IP3
Ka	µM	0.4	Li-Rinzel model parameter for IP3 receptor activation by Ca2+
Kd	µM	0.4	Li-Rinzel model parameter for IP3 receptor inactivation by Ca2+
A	µM–1 s–1	5	Li-Rinzel model parameter for IP3 receptor inactivation by Ca2+
	µM–1 s–1	7.78	Ca2+-GCaMP6s association constant129
	s–1	1.12	Ca2+-CaGCaMP6s dissociation constant129
GT	µM	7.4	Total GCaMP6s concentration130

Results

Dynamics of Ca²⁺ waves elicited by focal intradermal photodamage in the earlobe skin of live anesthetized mice

To photodamage a single keratinocyte in the basal epidermal layer of earlobe skin in live anesthetized mice expressing the GCaMP6s Ca²⁺ indicator, we used the second scanning head of the multiphoton microscope (GG branch of the optical path, see Methods). Ca²⁺ waves propagated radially from the damaged to the surrounding/bystander epidermal cells (referred to as expansion phase, Figure 1 A). Fluorescence signals persisted for several minutes in keratinocytes invaded by the Ca²⁺ wave, and slowly returned to basal levels (referred to as waning phase, Figure 1 B). Mean ± s.e.m. of the area A(t) invaded by the Ca²⁺ wave are shown for both expansion and waning phases in Figure 1 C, D. A(t) was computed by counting all pixels in which (see Methods) exceeded an arbitrary threshold, corresponding to 30% of the maximum signal achieved as a result of photodamage in the given image sequence.

Figure 1.

Intravital multiphoton microscopy of Ca²⁺ waves elicited by focal photodamage in the epidermal basal layer of earlobe skin in live anesthetized mice. (A, B) Representative sequences of GCaMP6s fluorescence images acquired at shown time points from the instant preceding the beginning of the photodamage (t = 0 s) showing (A) Ca²⁺ wave expansion from the photodamage site (yellow asterisk in the 0.5 s image) and (B) subsequent wave waning. (C, D) Invaded area vs. time, A(t), during Ca²⁺ wave expansion (C) and waning (D). (C) Mean (solid line) ± s.e.m. (dashed lines) of n = 60 experiments in 14 mice. (D) mean (solid line) ± s.e.m. (dashed lines) of n = 3 experiments in 1 mouse. (E, F) Volume of tissue invaded by the Ca²⁺ wave after focal photodamage. (E) Views of the wave in the xy plane (focal plane) and two orthogonal (xz and yz) planes. (F) 3D rendering of the volume invaded by the wave; cyan, red and green lines indicate the xy, xz and yz planes of panel (E), respectively. (G) Invaded area vs. axial coordinate (z, along the optical axis of the objective lens); data are mean (solid line) ± s.e.m. (dashed lines) of n = 7 experiments in 1 mouse. Scale bars in A, B, E: 25 µm.

The volume of tissue invaded by the Ca²⁺ wave at the time of maximal expansion () was visualized by acquiring through-focus image sequences (also known as z-stacks) at 2 µm increments along the optical axis (z) of the objective lens (Figure 1 E, F). The graph of the invaded area vs. depth (Figure 1 G) shows that Ca²⁺ waves extended up to ∼10 µm above the photodamage plane, and occasionally stimulated cells with the typical morphology of Langerhans cells, reaching the surface of the thin earlobe mouse skin (corneocyte layer). In the opposite, dermal direction, Ca²⁺ waves reached down to ∼20 µm below the photodamage plane invading the fibroblast-populated collagen matrix. Given the approximate radial symmetry of Ca²⁺ waves during the expansion phase in the focal plane, the equivalent radius of the invaded area was computed as

(R1)

(Figure 2 A). The maximal value R_max = max[] averaged over the control experiments that gave the largest waves was 67 ± 1 µm (mean ± s.e.m.; n = 21 experiments in 6 mice).

Figure 2.

Detailed analysis of the expansion phase. (A) Equivalent radius of the area invaded during Ca²⁺ wave expansion. (B) Propagation velocity of the Ca²⁺ wave as function of time and (C) of the equivalent radius. Mean (blue solid lines) ± s.e.m. (blue dashed lines) of n = 21 experiments in 6 mice.

To further characterize the expansion phase, we computed whole-cell responses by spatially averaging pixel signals over individual bystander keratinocytes. Data from different keratinocytes in different experiments were grouped and averaged based on the distance d of the cell centroid from the photodamage site in the range from 10 ± 3 µm to 80 ± 3 µm (Figure 3 A). Note that this range exceeds the average value of R_max given above. The discrepancy is due to the 30% threshold criterion used to select pixels that contributed to the estimate of the area A(t) invaded by the Ca²⁺ wave. In general responses at 70 ± 3 µm and 80 ± 3 µm remained subthreshold, and therefore did not contribute to A(t).

Figure 3.

Analysis of GCaMP6s fluorescence changes in bystander keratinocytes at increasing distance from the photodamage site. (A) In each panel, the vertical black dash-dotted line at 0 s marks the end of the 0.5 s photodamage time interval. Data are mean (solid line) ± s.e.m. (dashed lines) vs. time for n = 60 experiments in 15 mice. (B) Amplitude (a_20s) of the signal at time t = 20 s. (C) Area (I) under the trace, computed between 0 and 20 s. (D) Slope (s) of the trace at the inflection point (mean ± s.e.m). Data in (B-D) are mean ± s.e.m. vs. bystander cell distance from the photodamage site.

From the analysis of traces, we extracted 3 parameters for each order of bystander keratinocytes. The first parameter is the amplitude reached at time t = 20 s,

(R2)

which decreased slowly with d, up to a critical value d_stop (typically comprised between 70 µm and 80 µm) beyond which wave propagation ceased abruptly and a_20s approached zero (Figure 3 B).

The second parameter is the Ca²⁺ load imparted on the cell by the Ca²⁺ wave, computed as the integral

(R3)

where t₁ = 0 and t₂ = 20 s. I increases if the amplitude of increases and/or the velocity of wave propagation increases (due to the fixed limits of the integration interval). For the same reason, due to the progressively delayed rise of the Ca²⁺ response, I decreases with d (Figure 3 C).

The third parameter is the response speed of the cell

(R4)

computed as the slope of the trace around the inflection point of its rising phase. Clearly, s depends both on the velocity of wave propagation and the kinetics of the signal transduction chain that promotes the increase in the given cell. Altogether, a cohort of 17 control mice were used to generate the data presented in Figures 1-3. The results show that s was a monotonically decreasing function of d (Figure 3 D).

To explore the mechanisms underlying the propagation of Ca²⁺ waves, we used different drugs (Table 2) that interfere (more or less specifically) with candidate pathways. A different cohort of 18 mice were used for these in vivo pharmacological interference experiments, whose results are summarized graphically in Figure 4 and described in detail hereafter and in the Figures S6-S14.

Figure 4.

Summary of pharmacological interference experiments results. (i) Amplitude (a_20s) of the signal at time t = 20 s. (ii) Area (I) under the trace, computed between 0 and 20 s. (iii) Slope (s) of the trace at the inflection point (mean ± s.e.m). Data are mean ± s.e.m. vs. bystander cell distance from the photodamage site in control conditions (green) and after VS microinjection (red) containing one of the following drugs: apyrase 500 U/ml (A), PPADS 625 µM (B), ARL 400 µM (C), EGTA 5 mM (D), thapsigargin 400 nM (E), carbenoxolone 200 µM (CBX, F), probenecid 4 mM (G), TAT-Gap19 200 µM (H). Significant differences are shown above each pair of bars (two-sample t-test): * = P-value < 0.05; ** = P-value < 0.005.

Initially, a dedicated set of experiments were conducted to ascertain the effects of microinjecting solely vehicle solution (VS = PBS supplemented with Dextran Texas Red, 1.78 μM; see Methods) in the earlobe skin proximal to the photodamage site. Statistical analyses of the results indicate that the parameters used to characterize wave expansion after photodamage were largely unaffected by VS microinjection (Figure S6). Because VS injection had no significant influence on the parameters used to characterize quantitatively the expansion phase of Ca²⁺ waves, all successive experiments were performed in the following way: (1) a certain number of waves (used as reference) were evoked in n ≥ 3 non-overlapping areas of the naïve mouse earlobe epidermis; thereafter, (2) different areas of the same earlobe were microinjected with the drug of interest dissolved in VS and a (3) second set of waves were evoked (in n ≥ 3 non-overlapping areas reached by Dextran Texas Red), (4) recorded and (5) compared with the previously acquired reference waves.

Based on co-injected Dextran Texas Red fluorescence, microinjected drugs diffused in an area with a diameter of about 3 mm, wide enough to contain more than 20 non-overlapping fields of view (FOVs). In each non-overlapping FOV, a single photodamage experiment was carried out generating a wave that remained completely confined within the FOV (Figure S15).

Critical role of ATP release in photodamage-induced Ca²⁺ waves

ATP in the extracellular milieu may couple to P2 purinoceptors on the cell plasma membrane to develop Ca²⁺ waves with or without involvement of IP₃ diffusion through intercellular gap junction channels (IGJCs).¹⁵ To test the hypothesis that photodamage triggered Ca²⁺ waves by promoting ATP release from the photodamaged keratinocyte, we injected 4 µl of VS containing apyrase (500 U/mL), an enzyme that catalyzes ATP hydrolysis.⁴⁹ The presence of microinjected apyrase in the extracellular milieu correlated with a highly significant reduction of the invaded area , therefore also of the equivalent Ca²⁺ wave radius (Figure S7 A, A’). From the same experiments, we derived and analyzed traces from bystander keratinocytes and confirmed that apyrase severely hampered Ca²⁺ wave propagation, such that keratinocytes at distances d > 20 μm from the photodamage site failed to respond (Figure S7 D, D’). Consequently, all three critical parameters used to characterize these responses (a_20s, I and s) were null at d > 20 μm (Figure 4A, Figure S7 E-G, E’-G’).

To determine whether P2 purinoceptors contributed to photodamage-evoked Ca²⁺ waves, we verified P2YR expression in keratinocytes of basal layer of the earlobe epidermis by immunofluorescence analysis with specific knock-out validated antibodies against P2Y₁R and P2Y₂R (Figure S16). Next, we microinjected 4 µl of VS containing pyridoxal phosphate-6-azo (benzene-2,4-disulfonic acid) tetrasodium salt hydrate (PPADS, 625 µM), a P2 purinoceptor antagonist.^31,50 PPADS reduced significantly the area invaded by Ca²⁺ waves following focal photodamage (Figure S8 A, A’) and slowed down significantly responses in all bystander keratinocytes (Figure S8 B-D, B’-D’). The parameters a_20s, I and s were significantly reduced in PPADS compared to controls at d > 30 μm, and the arrest distance d_stop decreased from 80 μm (control) to 50 μm (PPADS) (Figure 4 B, Figure S8 E-G, E’-G’).

Once released in the extracellular space, ATP is rapidly hydrolyzed by cell surface-located ectonucleotidases, in particular by members of the ecto-nucleoside triphosphate diphosphohydrolase family (NTPDase1,2,3, and 8).^51–53 Therefore, we predicted that interfering with ectonucleotidase function should influence Ca²⁺ wave propagation. To test this hypothesis, we injected 4 µl of VS containing ARL 67156 trisodium salt hydrate (shortened as ARL, 400 µM), a widely used NTPDase inhibitor.^54–56 The presence of microinject ARL in the extracellular milieu correlated with a significant increase of the area invaded by Ca²⁺ waves, hence of (Figure S9 A, A’). ARL increased also the velocity of Ca²⁺ wave propagation (Figure S9 B, C, B’, C’), confirmed by the significantly reduced lag with which traces grew after photodamage in bystander keratinocytes at d > 20 μm (Figure S9 D, D’). In addition, ARL increased significantly the parameters a_20s, I and s at d > 40 μm (Figure 4C, Figure S9 E-G, E’-G’).

Together, the results in Figures S7-S9 indicate that (ii) extracellular ATP is a key signaling molecule underlying Ca²⁺ wave propagation in the photodamaged epidermis; (ii) the area invaded by the Ca²⁺ wave and its velocity of propagation depend on degradation of extracellular ATP by NTPDases expressed by epidermal keratinocytes (which can be partially inhibited by ARL).

Role of extracellular Ca²⁺

To determine the role of extracellular Ca²⁺ in the dynamics of photodamage-evoked Ca²⁺ waves, we microinjected 4 µl of VS containing 5 mM EGTA, a widely used Ca²⁺ chelator.⁵⁷ Reduction of the extracellular free Ca²⁺ concentration ([Ca²⁺]_ex) with EGTA caused a paradoxical enhancement of the invaded area, gauged by the effective radius (Figure S10 A, A’), paralleled by a significant increase of bystander responses at all distances within 60 μm from the photodamaged cell (Figure S10 D, D’). The parameters a_20s, I and s were significantly increased in EGTA compared to controls, and the arrest distance d_stop increased from 70 μm (control) to 80 μm (EGTA) (Figure 4D, Figure S10 E-G, E’-G’).

Because EGTA increased Ca²⁺ responses, Ca²⁺ influx through plasma membrane channels (including P2XRs²⁸) did not to contribute significantly to the observed increments of the . Hence, the latter should be attributed chiefly to signal transduction through P2YRs, via the coupling to G proteins, activation of PLC, leading to the formation of IP₃ and release of Ca²⁺ from intracellular stores.^29–31,58

Role of Ca²⁺ release from the ER

To corroborate the involvement of Ca²⁺ release from the ER downstream of P2YRs activation by extracellular ATP, we microinjected 4 µl of VS containing 400 nM of thapsigargin (previously dissolved in DMSO), a potent, cell-permeable, selective and irreversible inhibitor of the sarcoplasmic and endoplasmic reticulum Ca²⁺-ATPase family (SERCA pumps).⁵⁹ As expected, thapsigargin caused a significant increase in the basal level of GCaMP6s fluorescence, , from gray levels (GLs) in control conditions to GLs in thapsigargin (P = 0.01, two-sample t-test) due to uncompensated passive leakage of Ca²⁺ from the ER. Thapsigargin caused Ca²⁺ waves to propagate over smaller distances, significantly reducing the equivalent radius of the invaded area (Figure S11 A, A’). Moreover, the increased values caused an evident reduction of bystander responses at any order of distance and consequent significantly reduced a_20s (Figure S11 D, D’, Figure 4 E(i), Figure S11 E, E’). Also the parameters I and s were significantly reduced in thapsigargin vs. controls at all distances d from the photodamage site (Figure 4 E(ii,iii), Figure S11 F, G, F’, G’).

Role connexin HCs and pannexin 1 channels

Open connexin HCs mediate the diffusive release of paracrine messengers, most importantly ATP.^60–64 Also pannexin 1 (Panx1) channels⁶⁵ are permeable to ATP,⁶⁶ expressed in the epidermis and important for differentiation of keratinocytes and wound healing.⁶⁷ In addition, Panx1 channels can be activated by extracellular ATP acting through purinergic receptors of the P2Y group, as well as by cytoplasmic Ca²⁺.⁶⁸ Therefore, a positive-feedback loop involving the control by cytoplasmic Ca²⁺ of Panx1⁶⁸ and/or connexin HCs^42,69,70 may lead to an overall augmented ATP-induced ATP-release^68,71 that would increase both the area invaded by Ca²⁺ waves and the increments in individual keratinocytes. Testing these hypotheses required further pharmacological interventions, as summarized hereafter.

Carbenoxolone (CBX) is a broad-spectrum inhibitor of connexin-made channel (i.e. HCs and IGJCs)⁷² that inhibits also Panx1 channels^65,66 and IP₃-dependent Ca²⁺ release from the ER.⁷³ We injected 4 µl of VS containing CBX (400 µM), which reduced significantly both the area invaded by Ca²⁺ waves following focal photodamage (Figure S12 A, A’) and the responses in all bystander keratinocytes, except those adjected to the photodamage site (Figure S12 D, D’). The parameters a_20s, I and s were significantly reduced by CBX compared to controls at d > 10 μm, and the arrest distance d_stop decreased by 20 μm with respect to control conditions (Figure 4F, Figure S12 E-G, E’-G’).

To discriminate between Panx1 and connexin-made channels, we microinjected 4 µl of VS containing 4 mM of probenecid, which affects primarily Panx1 channels.⁷⁴ Probenecid modified slightly the time course of the equivalent radius without affecting significantly the area invaded by the Ca²⁺ waves at their maximal expansion (Figure S13 A, A’). In addition, signals were largely unaffected, with the exception of bystander keratinocytes located at 40 μm ≤ d ≤ 50 μm where responses appeared increased rather than decreased by the drug (Figure S13 D, D’). Only in this range of distances, the parameters a_20s, I and s were significantly larger in probenecid compared to controls, and the arrest distance d_stop was not affected (Figure 4G, Figure S13 E-G, E’-G’). These experiments lead us to conclude that (i) Panx1 channels were not involved in the Ca²⁺ signaling triggered by focal photodamage and (ii) the effect of CBX may reflect inhibition of connexin-made channel⁷² and/or IP₃-dependent Ca²⁺ release from the ER.⁷³

Various expression profiles of connexins through epidermis layers are reported in the literature, the majority of them agreeing in the predominance of Cx43 in the basal layer of the epidermis.^75–86 To discriminate between connexin HCs and IGJCs, we microinjected 4 µl of VS containing 200 µM TAT-Gap19, a peptide that selectively inhibits Cx43 HCs.⁸⁷ TAT-Gap19 did not change significantly the area invaded by Ca²⁺ waves (Figure S14 A, A’), whereas signals in bystander keratinocytes were attenuated at larger distances (60, 70 µm) (Figure S14 D, D’) compared to controls. The parameters a_20s, I and s were significantly reduced by Tat-Gap19 compared to controls at d > 50 μm, but the arrest distance d_stop was not affected (Figure 4H, Figure S14 E-G, E’-G’).

Comparison of experimental results with mathematical model

To overcome the limitations of pharmacological dissection, gain further mechanistic insight into these photodamage-evoked Ca²⁺ waves and determine whether the conclusions drawn from the experimental results can be reproduced coherently, we formulated a mathematical model (see Methods). To solve the system of model differential equations numerically, we assumed radial symmetry and reduced the problem to one dimension (see Methods). Considering the inter-animal variation (Figures S6-S14) the model with the chosen parameters (Table 3) reproduced well the experimental traces averaged over all control experiments (Figure 5A).

Figure 5.

Model simulations for GCaMP6s fluorescence responses and equivalent radius of the area invaded by Ca²⁺ waves under control condition. (A) Schematic representation of model geometry. ATP dynamics is simulated as being released from the photodamaged cell (grey cylinder; A_c) into a thin disc above the cell (A₀). ATP diffusion in the intercellular space is simulated by discretizing in the radial direction assuming radial symmetry. The ATP concentration in each ring (A_j) communicates with the cells above and below (indicated by the red dashed rectangles). (B) Cartoon representing main actors of Ca²⁺ signaling in this model. HC, connexin hemichannels; SERCA, sarco-endoplasmic reticulum Ca²⁺ ATPases; IP₃R, inositol (1,4,5) triphosphate receptors; PLC, phospholipases C; IP_3,j, flux of IP₃; Ca_j, flux of ionized calcium out of endoplasmic reticulum (ER) through IP₃R; A_j, extracellular ATP concentration; J_HC,j, flux of ATP via connexin HCs; Ca_ER,j, flux of ionized calcium into the ER through SERCA; solid arrows represent net fluxes; dash-dotted arrows represent extracellular diffusion down the concentration gradient; dashed lines represent interactions: arrowheads for stimulatory interactions, diamond for inhibitory interaction. A_j stimulates IP₃ production via PLCβ. IP₃ in turn stimulates Ca²⁺ release from the ER via IP₃Rs. Intracellular Ca²⁺ is assumed to stimulate ATP release via HC and to stimulate PLC. In addition, Ca²⁺ feeds back onto IP₃R, and is pumped into the ER via SERCA pumps. (C) Simulated GCaMP6s responses at increasing distances, in 10 µm steps, from the center of the damaged cell. Circles show the experimental results at the same distance as indicated by identical colors and specified in the legend. The black dashed curves show s.e.m of the experimental data points. (D) The simulation from panel C is shown here on an extended time scale to include the waning phase of the Ca²⁺ wave (note change of units from seconds to minutes in the abscissae). (E) Radius of the equivalent Ca²⁺ wave vs. time computed as the distance at which the response is 30% of the simulated response in panel C (black). The red curve shows corresponding experimental data with s.e.m. indicated by the dashed curves. (F) The simulation from panel E is shown here on an extended time scale to include the waning phase of the Ca²⁺ wave.

In the model, the simulated signal raised rapidly close to the damaged cell, followed by gradually more shallow and delayed responses at greater distances. The maximum model responses were achieved within 1–2 minutes after photodamage, thereafter signals slowly returned towards baseline on a timescale of more than 20 minutes for the cells closest to the damage site (Figure 5B). In other words, the simulations show a propagating wave that halts at approximately 80 µm from the damaged cell (Figure 5A) in agreement with the experiments, and with an equivalent radius (i.e., location of 30% response) as a function of time that mimics the experimental behavior (Figure 5C), in particular making allowance for the variability between different experimental animals (Figure S17). The model wave then slowly dissipated so that the equivalent radius reduced to 10 μm after about 18 min (Figure 5D).

After photodamage, the relatively large amounts of ATP present in the cytosol of the damaged cell diffused into the tight intercellular space. Near the damaged cells, the model ATP concentration reached tens to hundreds of µM (Figure 6A, B), hence the IP₃ producing machinery was completely saturated by ATP (the EC₅₀ parameter in the model is K_PLC = 1.05 µM). As a result of the saturation of IP₃ production near the damaged cell, steady-state IP₃ levels are very similar at d <50 µm (Figure 6C). The long-lived responses with a duration of tens of minutes (Figure 5B) are due to the high ATP concentrations, which decrease only slowly over time due to the modest degradation rate and relatively low effective diffusion constant, D. We used D = 65 µm²/s, which is substantially lower than the diffusion constant of ATP in solution (350–400 µm²/s).⁸⁸ However, the extracellular space between adjacent cells is about 20 nm,⁴⁰ i.e., only an order of magnitude larger than an ATP molecule, and is likely a crowded environment, which may slow down diffusion of ATP.⁸⁹ Indeed, the intracellular diffusion coefficient of ATP is ∼3.5 times lower than the extracellular value.⁹⁰ Further, ATP is buffered as it binds to e.g. ectonucleotidases, P2YRs and P2XRs on the cell membranes, and it is well-known that fixed buffers can reduce the effective diffusion constant substantially.⁹¹

Figure 6.

Model simulations for the concentrations of extracellular ATP and intracellular IP₃ under control conditions. (A) Simulated extracellular ATP responses at increasing distances, in 10 µm steps, from the center of the damaged cell as indicated by the color codes in the legend. (B) First 30 sec of the traces plotted in panel A. (C) Simulated intracellular IP₃ concentrations at increasing distances in the first 30 sec post damage and in 10 µm steps from the center of the damaged cell (traces color coded as in panel A).

The slowly increasing Ca²⁺ concentrations seen in the experiments even at very low distances are reproduced by the model because of slow feedback from cytosolic Ca²⁺ on PLC activity.^43–47 Note that this behavior cannot be explained by increased ATP levels, for example due to release from HCs, since any further increase in ATP beyond saturation would not lead to a raise in IP₃ and downstream levels. Simulating increased ATP degradation, as in the apyrase experiments, reduced the model responses substantially at d >10 µm, and response were basically suppressed at d >30 µm (Figure 7A). The lack of model Ca²⁺ responses follows from the greatly reduced ATP levels at all distances (Figure 7B). Consequently, the waves stopped at 20–30 µm from the damaged cell (Figure 7C). In contrast, reducing the rate of ATP consumption to mimic the ARL experiments yielded larger model responses, in particular at larger distances (Figure 7D). The increased model ATP levels seen at all distances (Figure 7E) affected the cells at greater distances more, since their IP₃ production was far from saturation. Quantifying, the simulation corresponding to the ARL experiments shows a greater equivalent radius than the control simulation, corresponding to waves that terminate at ∼100 µm (Figure 7F). Together, these model results show that the effects of apyrase and ARL can be explained entirely by interference with extracellular ATP degradation, which controls the extent of extracellular ATP levels.

Figure 7.

Model simulation of interference with extracellular ATP handling. (A-C) Simulated apyrase experiment performed by increasing 10-fold the ATP degradation rate k (full curves). The simulations under control conditions (from Figures 5 and 6) are shown for comparison as dashed curves with the same color code. (A) Simulated normalized GCaMP6s fluorescence responses, as indicated by the color codes in the legend. (B) Model extracellular ATP levels. (C) Equivalent radii from apyrase (blue solid line) and control (black dashed line, same as black solid line in Figure 5E) simulations. (D-F) Simulated ARL experiment performed by decreasing the ATP degradation rate k by 50% (solid lines). Legends as in panels A-C. Traces in panels A, B, D, E were computed at increasing distances, in 10 µm steps, from the center of the damaged cell.

Inhibition of connexin HCs, mimicking the TAT-Gap19 experiments, reduced the model responses at larger distances compared to the control scenario (Figure 8A), in agreement with experiments. Model cells with closed HCs did not release ATP, hence the extracellular ATP levels decreased (Figure 8B) causing a drop in IP₃ production and Ca²⁺ release from the ER. In control conditions, ATP release from HCs was larger close to the damaged cell since the was higher there than further away. However, the extracellular ATP was already saturating even without the contribution from the HCs, so blocking them had virtually no effect on variations at short distances from the photodamaged cell. On the contrary, at larger distances the contribution to extracellular ATP from HCs was relatively important compared to the ATP diffusion from the site of damage. Consequently, inhibiting HCs has a substantial effect on the simulated response. This explains why HCs blockade lowered the extent of wave propagation in the model (Figure 8C). Increasing the maximal rate of IP₃ production, V_PLC, yielded simulated responses that resembled the experimental data with zero extracellular Ca²⁺ and EGTA. At all distances, importantly also at the shortest distances from the damaged cell, the signals increased (Figure 8D) partly because of the positive feedback from Ca²⁺ on ATP release from HCs, which increased ATP levels at the larger distances (Figure 8E). Consequently, the wave propagated ∼10 µm further than in control simulations (Figure 8F).

Figure 8.

Model simulation of connexin HCs inhibition and extracellular Ca ²⁺ chelation. (A-C) Simulated TAT-Gap19 experiment performed by decreasing the rate of ATP release through connexin HCs (full curves). The simulations under control conditions (from   Figures 5 and 6) are shown for comparison as dashed curves with the same color code. (A) Simulated normalized GCaMP6s fluorescence responses, as indicated by the color codes in the legend. (B) Model extracellular ATP levels. (C) Equivalent radii from TAT-Gap19 (blue solid line) and control (black dashed line, same as black solid line in Figure 5E) simulations. (D-F) Simulated EGTA experiment performed by increasing the maximal rate of IP₃ production V_PLC by 7% (see main text). Legends as in panels A-C. Traces in panels A, B, D, E were computed at increasing distances, in 10 µm steps, from the center of the damaged cell.

Discussion

We have shown here that ATP is released from epidermal keratinocytes of live anesthetized mice following intradermal photodamage caused by intense pulsed IR laser radiation. Kumamoto et al. obtained results compatible with those reported here using point laser stimulation of keratinocytes in ex vivo human epidermis. In particular, they observed different responses in the various layers of the epidermis after stimulating a keratinocyte of the stratum granulosum, with higher responses in the basal layer.¹³

Insight provided by in vivo pharmacological dissection experiments

The largest effects were caused by drugs that interfere with degradation of extracellular ATP or P2 purinoceptors, suggesting that Ca²⁺ waves in the photodamaged epidermis are primarily due to release of ATP from the target cell whose plasma membrane integrity was compromised by laser irradiation. As signals in bystander keratinocytes were augmented by exposure to the Ca²⁺ chelator EGTA in the extracellular medium, the corresponding transient increments of the should be ascribed primarily to Ca²⁺ release from the ER, downstream of ATP binding to P2Y purinoceptors with Ca²⁺ entry through plasma membrane channels playing a comparatively negligible role. The effect of thapsigargin supports this conclusion.

PPADS antagonizes P2Y₁,⁹² P2Y₄,⁹³ P2Y₆⁹⁴ and P2Y₁₃⁹⁵ receptors (see also ref.²⁹). In contrast, P2Y₂Rs are PPADS-insensitive.^29,93 Published work on P2Y mRNA expression in murine keratinocytes identified P2Y₁R, P2Y₂R, P2Y₄R and P2Y₆R subtypes,⁹⁶ therefore P2Y₂Rs are likely responsible for the incomplete effect of PPADS administration.

The ineffectiveness of probenecid suggests pannexin channels had no role, whereas the limited effect of TAT-Gap19 suggests ATP-dependent ATP release though connexin HCs affected Ca²⁺ wave propagation only at larger distances, where the concentration of ATP released from the photodamaged cell was reduced by the combined effect of passive diffusion and hydrolysis due to the action of ecto-ATPases. The fact that CBX exerted a stronger inhibitory action compared to TAT-Gap19 may suggest that IGJCs played a more prevalent role than Cx43 HCs in Ca²⁺ wave propagation. However, CBX has been recently reported to interference also with IP₃Rs,⁹⁷ which would instead lend further support to a fundamental role played by the P2Y-dependent pathway and IP₃-dependent Ca²⁺ release from intracellular stores.

Insight provided by the mathematical model

Our mathematical model reproduced the experimental observations by assuming that extracellular ATP diffusion carried the signal from the damaged to bystander cells, where it activated PLC and increased IP₃ levels, which in turn triggered Ca²⁺ release via IP₃Rs. Ca²⁺ feedback onto PLC caused the slowly rising response during the expansion phase of the Ca²⁺ waves. Close to the damaged cell, ATP concentrations were saturating for the P2YRs/PLC machinery, whereas the wave stopped further away where degradation and dilution of ATP lowered the ATP concentration to levels insufficient to cause substantial Ca²⁺ release. Changes in the model degradation rate reproduced experiments with apyrase and ARL. ATP release through connexin HCs contributed comparatively more at longer distances, where the flux of ATP from cytosol raised the extracellular ATP levels sufficiently to yield an intracellular Ca²⁺ response. However, HCs contribution was eclipsed in the proximity of the damage site by the far preponderant contribution from photodamage.

We also propose a plausible explanation for how chelation of extracellular Ca²⁺ with EGTA might increase PLC activity. In keratinocytes, extracellular Ca²⁺ partly activates PKC,⁹⁸ thus removal of extracellular Ca²⁺ may ensue in a reduced PKC activity. Since PKC inhibits PLC,^44,99,100 PKC deactivation would remove the brake on the rate of IP₃ production. Note that alternative explanations invoking an increase in extracellular ATP levels following chelation of extracellular Ca²⁺, for example because of a direct effect of low Ca²⁺ on connexin HCs, are incompatible with the saturating ATP levels near the damaged cell.

In the proposed model, we neglected intercellular Ca²⁺ and IP₃ diffusion via inter-cellular gap junction channels since our experiments suggested that extracellular ATP diffusion is the major route of signal propagation, in contrast to other physiological systems (e.g. Refs^101–103). Thus, in our model, the waves are propagating with no communication between the bystander cells, except from a near-negligible contribution due to ATP diffusion through hemichannels.

The mechanism for wave block of the simulated abortive waves that we studied here differs from – but also has similarities to – other systems. For example, in gap-junction coupled pancreatic beta-cells, an externally imposed glucose gradient causes abortive, repetitive Ca²⁺ waves,^104,105 which have been proposed to stop because of the gradual reduction in excitability of the cells,^106,107 resembling the ATP gradient observed here, although the waves between beta-cells are driven by electrical communication via gap junctions. Heterogeneity in cellular and gap-junction properties in combination with discretization effects may also lead to propagation failure between coupled cells.^108–111 Such heterogeneity was not considered here, and the model waves stop simply when the concentration of diffusing ATP becomes too low to stimulate sufficient IP₃ production to increase the intracellular Ca²⁺ concentration.

The observed near-steady and long-lived elevations were quantitatively different from the oscillations often observed, for example, in the sensory epithelium of the cochlea following photodamage^11,12 or IP₃ uncaging.^71,112–114 Why are such oscillations not seen in the photodamaged skin? Also in this case, the mathematical model provides a possible explanation. The core of the Ca²⁺ signaling part of the model is the Li-Rinzel model,⁴⁸ which is well-known to sustain oscillations, but only for a range of IP₃ concentrations. We therefore investigated how the single-cell model depends on the IP₃ production rate V_PLC (assuming that ATP levels were saturating). The bifurcation diagram in Figure 9 shows that the model has a single stable equilibrium at low V_PLC values, which becomes unstable due to a subcritical Hopf bifurcation at V_PLC = 2.02. The equilibrium becomes stable again for values above V_PLC = 2.44, where another subcritical Hopf bifurcation occurs. The unstable limit cycles born in the Hopf bifurcations become stable at saddle-node of limit cycles bifurcations, thus the model generates stable oscillations for values of V_PLC between 2.0 and 2.45 (approximately). Hence, the default value of V_PLC = 1.85, which we used to reproduce the experiments, is too low to sustain oscillations. Biologically, this explanation corresponds to the hypothesis that keratinocytes have too few P2YRs and/or amounts of PLC to transduce the high ATP levels into sufficiently high rates of IP₃ production to cause oscillations.

Figure 9.

Bifurcation diagram of the single-cell model. The steady-state response, normalized to the response at the default value of V_PLC= 1.85 µM/s, is shown as a function of the maximal rate of IP₃ production, V_PLC, under the assumption of saturating ATP levels. At low V_PLC values, the steady-state is stable (blue solid line), but loses stability in a Hopf bifurcation (HB) at V_PLC≈2.02. The unstable fixpoint (red dashed line) regains stability in another HB at V_PLC≈2.44. Unstable limit cycles (black asterisks) are born at both HBs, but due to saddle-node-of-limit-cycles bifurcations at V_PLC≈2.0 and V_PLC≈ 2.45, stable limit cycles are present between these values. The black circle indicates the default value of V_PLC used in the model, which is located below the interval where the system may oscillate.

Significance of the present findings for skin therapy and wound healing

In laser skin resurfacing with intra-epidermal focal laser-induced photodamage, the targeted micro-regions are thought to promote formation of new dermal collagen and elastin during the repair process,^32–35 but the underlying mechanisms were poorly understood before this study. Our results clearly indicate that P2YRs are activated by ATP downstream of photodamage, and P2YRs regulate cell proliferation, differentiation and migration,^115–117 besides being involved in wound healing.³¹ A prime target of the associated signaling activity is likely to be the extracellular matrix (ECM). In cultured keratinocytes, ATP caused a rapid and strong but transient activation of hyaluronan synthase 2 (HAS2) expression (via PKC-, CaMKII-, MAPK- and CREB-dependent pathways) by activating P2Y₂Rs.¹¹⁸ Hyaluronan, also named hyaluronic acid, is a major component of the ECM, and its synthesis is rapidly upregulated after tissue wounding.¹¹⁹ Thus, epidermal insults are associated with extracellular ATP release, as well as hyaluronan synthesis, and the two phenomena are linked.¹¹⁸

In addition, ATP is well known as one of the DAMPs that, once released from injured cells, activate the immune system.¹²⁰ Besides alerting the body about danger and initiate the immune response, DAMPs also promote regeneration processes.¹²⁰ P2YRs are important also in this context, because they regulate many aspects of immune cell function, including phagocytosis and killing of pathogens, antigen presentation, chemotaxis, degranulation, cytokine production, and lymphocyte activation.¹²¹ In the epidermis, Langerhans cells, characterized by the expression of Langerin (CD207) and CD1a, are considered to be macrophages that retain the function of dendritic cells, playing a clear role of antigen-presenting cells in skin inflammation, triggering a series of immune responses by migrating from the epidermis to the lymph nodes and presenting antigens to T-regulatory cells.¹²² A Langerhans cell line (XS106) was shown to express mRNA for P2X₁, P2X₇, P2Y₁, P2Y₂, P2Y₄, and P2Y₁₁ receptors,¹²³ which is consistent with the increase in GCaMP6s fluorescence in Langerhans cells we observed in some of our in vivo experiments. Therefore, it would be interesting, in future studies, to trace the movements of the Langerhans cells invested by damage-evoked Ca²⁺ waves.

In our experimental model, Ca²⁺ waves reached into the fibroblast-populated collagen matrix, and this event may activate and coordinate the keratinocyte-fibroblast interaction that is key to wound healing and skin remodeling.^124,125 Indeed, fibroblasts are well known not only for being the major producers of extracellular matrix, but also as a cell type actively involved in synthesis of inflammatory mediators and tissue repair. ATP or UTP increased the levels of ECM-related proteins through the activation of P2Y₂Rs in mouse fibroblasts.¹²⁶ Therefore, it is tempting to speculate that response coordination after accidental injury or laser treatment occurs via propagation of the ATP-dependent Ca²⁺ waves we visualized and analyzed in this article.

Concluding remarks

Numerous other molecular players contribute to epidermal Ca²⁺ homeostasis, therefore Ca²⁺ dynamics may involve a variety of other processes we did not take into account, including Ca²⁺-sensing receptor, transient receptor potential channels, store-operated calcium entry channel Orai1, endoplasmic Ca²⁺ depletion sensor (stromal interaction molecule 1 [STIM1]), and voltage-gated calcium channels such as L-type calcium channels.^18,127 In addition, further experiments are required to better characterize the dynamics of ATP after it is released into the extracellular space and is degraded rapidly to ADP, AMP, and further to adenosine,⁶⁴ which (with the exception of AMP) activate specific receptors even in a self-sustaining fashion.¹²⁸

Data and code availability

Data were archived in a repository handled by the University of Padova and can be accessed at: http://researchdata.cab.unipd.it/id/eprint/547.

The model code is available in the repository managed by the University of Padova at http://researchdata.cab.unipd.it/id/eprint/559 and at http://www.dei.unipd.it/∼pedersen.

The roiSelection GUI code is available on GitHub at https://github.com/DonatiViola/ROISELECTION.git.

Author contributions

Conceptualization: F.M. Data Curation: V.D., C.P. Formal Analysis: V.D., C.P. Funding Acquisition: F.M. Investigation: V.D., C.P., C.N. Methodology: M.B., M.G.P. Project Administration: F.M. Resources: F.M., F.S., M.R. Software: F.M., C.D.C., V.D., C.P., M.B., M.G.P. Supervision: F.M., M.B.

Validation: V.D., C.P. C.N. Visualization: V.D., C.P., M.B., M.G.P. Writing - Original Draft Preparation: F.M. Writing - Review and Editing: F.M., V.D., C.P., C.N., M.B., M.G.P.

Funding

This work was supported by the University of Padova (Grant No. BIRD187130 to FM) and Fondazione Telethon (Grant No. GGP19148 to FM). The funders had no role in study design, data collection, data analysis, interpretation or writing of the report.

Conflict of interest statement. The authors declare no competing interests.