Early bioelectric activities mediate redox-modulated regeneration

Understanding large-scale repair via regeneration is a fundamental issue in basic biology and regenerative medicine. Biochemical signals, such as Wnt and BMP pathways, have been extensively studied and have been shown to control regeneration (Alvarado and Tsonis, 2006; Stoick-Cooper et al., 2007). Reactive oxygen species (ROS), especially hydrogen peroxide (H₂O₂), are suggested to act as signaling cues in the wound microenvironment (Sen and Roy, 2008; Veal et al., 2007). Upon wounding, the Drosophila embryo (Moreira et al., 2010), zebrafish larva (Niethammer et al., 2009) and Xenopus tadpole (Love et al., 2013) produce a H₂O₂ gradient that precedes the oxidative burst. This gradient, in fact, attracts immune cells toward the wound edges. In addition, a redox sensor has been identified in the leading edge of neutrophils (Wittmann et al., 2012; Yoo et al., 2011, 2012) and topical application of low doses of H₂O₂ enhances wound closure in mice (Loo et al., 2012). Importantly, ROS are required for the regeneration of planarian head and tail (Pirotte et al., 2015), zebrafish caudal fin (Gauron et al., 2013) and heart (Han et al., 2014), Xenopus tadpole tail (Love et al., 2013), and gecko tail (Zhang et al., 2016).

In addition to ROS, biophysical signals such as the dynamic bioelectric activities measured during regeneration have been proposed to act as signaling cues (Levin, 2007; McCaig et al., 2005). It has been shown that transepithelial potential (TEP)-driven electric current densities (J_I) correlate with and are required for regeneration in amphibians (Borgens et al., 1977b; Jenkins et al., 1996; Nawata, 2001; Reid et al., 2009). Electrical stimulation induces some regeneration in non-permissive adult frog and rat limbs (Becker, 1972; Borgens et al., 1977a; Leppik et al., 2015; Smith, 1967). Moreover, engineered devices (Golding et al., 2016) partially induce mouse digit tip regeneration (Hechavarria et al., 2010) and are in clinical trials to improve paraplegia (Shapiro et al., 2005). Re-emerging bioelectric studies are revealing transduction mechanisms. V-ATPase-mediated repolarization of membrane potential (V_m) and voltage-gated Na⁺ channel (VGSC or Na_V) 1.2-mediated sodium influx are necessary for and sufficient to induce X. laevis tadpole tail regeneration (Adams et al., 2007; Tseng et al., 2010). V-ATPase proton efflux activity correlates with regeneration rate and ability in zebrafish caudal fin regeneration (Monteiro et al., 2014). H⁺/K⁺-ATPase-mediated V_m depolarization modulates planarian head regeneration (Beane et al., 2011). In addition, ion and V_m dynamics correlate with axolotl larvae tail regeneration (Özkucur et al., 2010).

Interestingly, redox and bioelectric activities control the expression and activity of signaling pathways, such as Wnt, FGF, BMP and Notch, in addition to cell behaviors, such as proliferation, apoptosis and innervation, which are required for regeneration (Adams et al., 2007; Gauron et al., 2013; Han et al., 2014; Love et al., 2013; Monteiro et al., 2014; Tseng et al., 2010). Redox (ROS) and bioelectric [V_m, TEP, electric fields (EF) and J_I] states therefore affect regeneration. But how do these signals integrate? Do they act in parallel or in series during regeneration? Xenopus laevis tadpole tail regeneration provides an ideal model with which to address these questions (Love et al., 2013; Reid et al., 2009). The tail comprises epidermis, muscle, blood vessels and spinal cord, making it a prime model for biomedical research (Beck et al., 2009; Deuchar, 1975). In addition, the fluctuating regenerative abilities of tails through developmental stages offer unique opportunities to study regeneration-deficient tails, in the refractory period (Beck et al., 2003), using the same model organism.

Do ROS modulate regeneration via the regulation of V_m, TEP, EF and/or J_I, all of which are known to be required for regeneration? At the cellular level, NADPH oxidases in the cell membrane catalyze electron transfer from cytoplasmic NADPH to extracellular molecular O₂, generating ROS and depolarizing V_m (Demaurex and Petheö, 2005; Lambeth and Neish, 2014). This dual property – electrogenic and catalytic – of NADPH oxidases led us to hypothesize a two-way regulation: (1) the activity of NADPH oxidases per se depolarizes V_m; and (2) the ROS produced, especially H₂O₂, regulate TEP, EF and J_I. To test this hypothesis, we performed descriptive spatiotemporal profiling and functional assays.

We first mapped the extracellular bioelectricity, TEP, EF and J_I, during regeneration to select key spatiotemporal points at which to test the two-way regulation hypothesis. Regeneration can be divided into three phases, structurally completed in 7 days (Beck et al., 2009; Tseng et al., 2010): (1) wound healing [until ∼6 h post-amputation (hpa)] – wound epithelium covers amputation plane; (2) regeneration bud formation (∼24 hpa) – bud produces progenitor cells; and (3) regenerative outgrowth (∼48 hpa and after) – outgrowth and patterning reestablishes tail structure and function (Fig. 1A, Fig. S5).

Before amputation, baseline trunk TEP was 28.64±2.27 mV (inside positive; n=9) and the tail tip had small inward currents of −0.04±0.005 µA cm⁻² (n=7) (Fig. 1B, Fig. S2C). EF were assumed to be 0 mV mm⁻¹, because over the measured distances (<1 mm), no consistent TEP gradients were detected. Upon amputation, a TEP short-circuit (to 0 mV) at the stump generated a strong lateral EF_Trunk→Bud of −85.50 mV mm⁻¹ (calculated using uncut TEP) and a low-resistance pathway leaking outward currents (the injury current) of 13.27±2.807 µA cm⁻² (at 0.08 hpa; n=6, P=0.005) in the bud (Fig. 1B-D). Spatial profiles showed a consistent anteroposterior (A/P) gradient either directly (for TEP) or inversely (for EF) proportional to the distance from the amputation plane and evolving circuits (for J_I and EF) during regeneration (Figs S2-S5). The spatial profile of J_I was intended to unveil circuit loops (primarily dorsoventral) more than gradients (cases of TEP and V_m).

Temporal profiles presented exponential-like curves, positive for TEP recovery and for EF (Fig. 1B,C), and negative for J_I (Fig. 1D), that correlated with regeneration. The dramatic TEP recovery (to 67.42±5.35%, n=26, P<0.0001 versus 1 hpa), decrease in EF (to −22.94±4.89 mV mm⁻¹, n=23, P=0.0002 versus 1 hpa; 73% drop versus 0.08 hpa) and J_I reversal (to −0.10±0.017 µA cm⁻², n=27, P=0.005 versus 0.08 hpa; 99% shift) by 6 hpa (Fig. 1B-D) match the wound epithelium formation phase and possible bud initiation (Fig. 1A,E, Fig. S5). Spatially, J_I reversals progressed in a centripetal way until 6 hpa, lastly in the bud (Fig. S4C), which is possibly due to centripetal healing. We descriptively termed this part of the curve as ‘slope’ (‘S’; Fig. 1B-D). Analogous TEP recovery is seen in wound healing assays in vitro (human skin) and in vivo (porcine skin) (Dubé et al., 2010). Similar drops in EF (75%) and J_I reversal (in the spinal cord) by 6 hpa occur in newt limbs and X. laevis tadpole tail regeneration, respectively (McGinnis and Vanable, 1986; Reid et al., 2009). From 6 to 48 (especially to 24) hpa, TEP, EF and J_I stabilization (P≫0.05 to all comparisons from 6 to 24 hpa; Fig. 1B-D) correlated with the regeneration bud formation phase (Fig. 1A,E, Fig. S5). We termed this part of the curve as ‘plateau’ (‘P’; Fig. 1B-D). From 48 hpa onwards, a final shift completed TEP recovery (Fig. 1B; bud versus trunk non-significant at 72 hpa, n=11, P=0.106; Fig. S2C), resolved EF (−10.74±7.12 mV mm⁻¹ at 72 hpa, n=11; Fig. 1C) and returned J_I to even lower than uncut baseline (−0.01±0.002 µA cm⁻² at 48 hpa, n=11, P<0.0001; Fig. 1D). The very small J_I from 48 hpa might be due to the increased surface area to volume ratio, which is due to the animal growth. We therefore termed this part of the curve as ‘baseline’ (‘B’; Fig. 1B-D) and it correlated with regenerative outgrowth phase (Fig. 1A,E, Fig. S5). Altogether, extracellular bioelectricity shows a dynamic correlation with regeneration and suggests that the bud at 6 hpa is a key spatiotemporal point to test the proposed hypothesis.

Next, we sought to clarify whether H₂O₂ is the main ROS involved in regeneration. We used drugs that target specific steps of the ROS ‘life cycle’ (see Fig. 6A): (1) production – by blocking NADPH oxidases with diphenyleneiodonium (DPI) (O'Donnell et al., 1993); (2) availability – by scavenging ROS with MCI-186 (MCI) (Otomo, 2003); and (3) diffusion – by preventing entry of H₂O₂ into cells through aquaporins with AgNO₃ (Niemietz and Tyerman, 2002). As previously shown (Han et al., 2014; Love et al., 2013; Niethammer et al., 2009), we confirmed that these drugs decreased levels of intracellular ROS in the bud (Fig. S6).

Similarly to another study (Love et al., 2013), DPI-treated tadpoles showed significantly inhibited regeneration, with virtually all tails presenting either a weak (35%, sixfold increase relative to vehicle control 6%) or none (64%, >50-fold increase) phenotype [regeneration index (RI) reduced from 255, n=233, to 39, n=74, P<0.0001; Fig. 2A,A′]. To establish NADPH oxidases as the main ROS source for regeneration, we used the alternative drugs apocynin (Simons et al., 1990) and VAS2870 (VAS) (ten Freyhaus et al., 2006). Besides being less penetrant than DPI, both drugs significantly impaired regeneration (Fig. S7A). Scavenging ROS with the antioxidant MCI significantly impaired regeneration with a 83% decrease in full (11%) and a fourfold increase in weak (24%) phenotypes (RI reduced from 255 to 187, n=43, P=0.005; Fig. 2A,A′). More penetrant effects of MCI have previously been shown if quantified at 3 days post-amputation (dpa), but not at 7 dpa (Love et al., 2013). Most H₂O₂ diffuses through aquaporins, such as AQP3 and AQP8 (Bienert and Chaumont, 2014; Miller et al., 2010). Highly specific small-molecule drugs for aquaporins are not yet available (Verkman et al., 2014) so we used AgNO₃. AgNO₃-treated tadpoles showed significantly impaired regeneration with a 69% reduction in the full phenotype (20%), and four- and fivefold increases in, respectively, the weak (25%) and the none (7%) phenotypes (RI reduced from 255 to 182, n=65, P<0.0001; Fig. 2A,A′). Penetrance on regeneration inhibition was therefore DPI≫AgNO₃>MCI (Fig. 2A′). Altogether, H₂O₂ is probably the main ROS responsible for the impaired regeneration and it must diffuse into cells, likely via aquaporins, to be effective.

Love et al. have shown that ROS production for the first 3 days is enough to impair regeneration (Love et al., 2013). We tested the role for ROS earlier, in wound epithelium and regeneration bud form, because both are limiting factors of regeneration (Fig. S7B-C′). We started by testing DPI until 6 and 24 hpa. These experiments still robustly inhibited regeneration (until 6 hpa, RI reduced from 246, n=94, to 160, n=52, P<0.0001; until 24 hpa, RI reduced to 89, n=50, P<0.0001; Fig. S7B). VAS treatment confirmed this early requirement (Fig. S7A). Next, we extended the temporal screening and found that ROS are also necessary for the last phase of regeneration (48 hpa onwards), although not for morphogenesis (RI) but for growth (area) (Fig. S7B-B″). This is congruent with impaired proliferation produced by inhibition of NADPH oxidases (Gauron et al., 2013; Han et al., 2014; Love et al., 2013). Finally, by increasing DPI dose, we found that ROS are immediately necessary for regeneration (Fig. S7C,C′). Altogether, ROS presented multi-phase pleiotropic effects during regeneration and NADPH oxidase-mediated ROS production is, to our knowledge, potentially the earliest biochemical limiting factor of regeneration.

Cellular bioelectricity has previously been characterized in the same X. laevis model (Adams et al., 2007) and in axolotl larvae (Özkucur et al., 2010). These studies showed a robust V_m depolarization in the bud at 6 hpa that should repolarize by 24 hpa for complete regeneration (Adams et al., 2007). However, the origin of depolarization is unknown. V_m depolarization (Adams et al., 2007) and H₂O₂ concentration (Love et al., 2013) appear to overlap spatiotemporally. Activity of NADPH oxidases generates an electron flow that depolarizes V_m in immune cells that can be used as a direct quantification of holoenzyme activity (Demaurex and Petheö, 2005).

To test whether the activity of NADPH oxidases induced the V_m depolarization, we blocked them with DPI and imaged V_m with the membrane potential-sensitive dye DiBAC₄(3) (Fig. S8A). DPI-treated tails at 6 hpa were, as in vehicle controls, still depolarized in the bud in relation to the rest of the tail (bud≫shoulder>trunk; Fig. S8A′), but had a significantly decreased depolarization (46% reduction, n=20, P=0.033; Fig. 2B,B′). Consequently, the spatial A/P gradient was also reduced (Fig. S8A′).

To exclude a possible effect of ROS/H₂O₂ in V_m, we next scavenged and prevented them from diffusing into cells as before. MCI- and AgNO₃-treated tails showed no V_m variation in the bud at 6 hpa (n=16, P>0.05 to both; Fig. 2B,B′). Therefore, V_m depolarization in the bud at 6 hpa is originated by the activity of NADPH oxidases per se.

Next, we tested whether H₂O₂ regulates TEP, EF and J_I. To achieve this, we blocked ROS/H₂O₂ production, availability and diffusion as before and measured TEP and J_I in the bud at 6 hpa. DPI-treated tails showed a significantly decreased TEP (58% reduction, from 17.42±1.11, n=38, to 7.33±1.08 mV, n=33, P<0.0001; Fig. 2C). DPI reversed the direction of some TEPs to negative inside (e.g. 12% of total at 6 hpa). MCI- and AgNO₃-treated tadpoles also had significantly decreased TEPs (MCI, 16% reduction, n=24, P=0.046; AgNO₃, 30% reduction, n=23, P=0.002; Fig. 2C).

Drug-treated shoulder and trunk (amputated or uncut) TEP shifted in the same way as bud TEP (Figs S8B, S12C,D), suggesting a systemic baseline shift and effect on housekeeping translocators. As a consequence, spatial A/P gradients remained relatively unaffected (Fig. S8B), leading to constant TEP recoveries and long-range EF_Trunk→Bud (Fig. S12E). However, short-range EF_{Shoulder→Bud} presented increased magnitudes (Fig. S16A), potentially powered by J_I direction (below). For these reasons, we mainly compared absolute TEP in bud.

The TEP decrease, caused by depleted H₂O₂, anticipated an effect of H₂O₂ on J_I. Almost all (91.60±3.38%) vehicle-control bud currents reversed at 6 hpa, giving a net inward J_I (−0.10±0.025 µA cm⁻², n=64). Only 18.05±8.32% reversed in DPI-treated tadpoles, maintaining a significant outward J_I (0.34±0.096 µA cm⁻², n=65, P<0.0001; Fig. 2D). Indeed, steady outward J_I were measured for days (Fig. S9A). Further analysis showed the reversibility of DPI-prevented reversal (Fig. S9B). MCI and AgNO₃ treatments also prevented reversal at 6 hpa with significant outward J_I (MCI, n=42, P<0.01, 58.33±10.41% reversal; AgNO₃, n=25, P=0.026, 40.00±11.55% reversal; Fig. 2D). Unlike DPI, J_I in MCI- and AgNO₃-treated tails reversed by 24 hpa (Fig. S8C), mimicking the refractory period (Fig. S11C).

Spatially, drug-sustained outward current at 6 hpa precluded an evolving electrical circuit (Figs S8C, S9C) and probably contributed to maintaining EF direction (cathode in the bud) and to increasing EF magnitude (Fig. S16A). Importantly, the penetrance of the drugs on TEP decrease and J_I reversal was equal to impaired regeneration (DPI≫AgNO₃>MCI; Fig. 2A′,C,D). Altogether, the TEP increase and J_I reversal are regulated by H₂O₂ that diffuses into cells, likely via aquaporins.

After demonstrating that H₂O₂ is required for regeneration and regulates TEP and J_I, we performed stimulation assays. We first attempted to rescue DPI-impaired regeneration, TEP and J_I using exogenous H₂O₂. To impose a suitable spatiotemporal redox state was challenging due to diverse cell responses (physiological and biochemical) to disparate ROS levels. Briefly, high levels might produce oxidative stress, whereas low levels produce more signaling effects (Lambeth and Neish, 2014). Long-term exposure to H₂O₂ could turn counterproductive by setting too much of a steady state in the very dynamic process of regeneration. In addition, regeneration-specific events are more likely to be present within 24 hpa. Therefore, we focused on application of H₂O₂ in low doses (µM range) and for a short exposure (first 24 h).

H₂O₂ until 24 hpa significantly rescued DPI-impaired regeneration, TEP increase and J_I reversal (Fig. 2A,A′,C,D). H₂O₂ increased twofold the frequency of full phenotypes compared with DPI-treated tails (from 7 to 15%) and decreased 37% the frequency of none phenotypes (from 41 to 26%). Overall, regeneration was significantly rescued (RI from 90, n=194, to 124, n=134, P=0.016; Fig. 2A,A′). Long-term (7 days) exposure to H₂O₂ did not rescue regeneration (Fig. S13), pointing to the importance of redox state dynamics (Loo et al., 2012).

TEP magnitude in the bud at 6 hpa was entirely rescued to values slightly higher than vehicle control [dimethyl sulfoxide (DMSO): 17.42±1.11 mV, n=38; DPI+H₂O₂: 19.15±1.72 mV, n=12; P=0.435; Fig. 2C]. H₂O₂ significantly rescued 161% of TEP (from 7.33±1.08, n=33, to 19.15±1.72 mV, n=12, P<0.0001; Fig. 2C) and enhanced 45% of TEP (from 17.42±1.11 to 25.20±2.84 mV, n=7, P=0.009; Fig. 2C). As for DPI, MCI and AgNO₃, shoulder and trunk (amputated or uncut) TEP shifted in the same way as bud TEP (Figs S8B, S12C-E).

J_I magnitude and J_I reversal in the bud at 6 hpa were dramatically rescued to values closer to vehicle control (DMSO: −0.10±0.025 µA cm⁻², 91.60±3.38% reversal, n=64; DPI+H₂O₂: −0.02±0.086 µA cm⁻², 44.76±10.05% reversal, n=36; P=0.395; Fig. 2D). H₂O₂ significantly rescued 105% of J_I magnitude to inward current (from 0.34±0.096, n=65, to −0.02±0.086 µA cm⁻², n=36, P=0.005; Fig. 2D). Importantly, current reversals more than doubled (from 18.05±8.32 to 44.76±10.05%, P=0.029; Fig. 2D). Unlike the changes induced in TEP, H₂O₂ neither enhanced J_I magnitude nor J_I reversal frequency in vehicle control. Nonetheless, most currents still reversed by 6 hpa (69.05±13.52%), resulting in a large variance (0.36±0.204 µA cm⁻², n=39, P=0.114; Fig. 2D). Altogether, short-term exposure to low-dose H₂O₂ rescued DPI-impaired regeneration, TEP increase and J_I reversal. In addition, these results suggest that the electrogenicity of NADPH oxidases may not affect TEP and J_I.

The X. laevis non-regenerative refractory period provides a unique advantage among regeneration models. We explored it to test whether we could induce regeneration, TEP and J_I using exogenous H₂O₂. First, we verified whether the refractory period has impaired bioelectricity. Remarkably, TEP decrease and postponed J_I reversal in the bud mimicked the effects of depleted ROS (Fig. S11). Blocked J_I reversal has previously been shown in the spinal cord (Reid et al., 2009). Application of H₂O₂ until 24 hpa increased the frequency of full phenotypes by 2.4-fold (from 13 to 31%) and decreased the frequency of none phenotypes by 62% (from 39 to 15%). Overall, regeneration was significantly induced (RI from 110, n=55, to 174, n=51, P=0.004; Fig. 3A,A′). As for rescue, long-term (7 days) exposure to H₂O₂ did not induce regeneration (Fig. S13).

The magnitude of the TEP in the bud at 6 hpa increased to values slightly higher than those found during the regenerative period (DMSO stage 40-41: 23.10±1.74 mV, n=14; DMSO stage 45-46+H₂O₂: 25.92±1.54 mV, n=18; P=0.235; Fig. 3B). H₂O₂ raised the TEP by 115% (from 12.04±1.45, n=14, to 25.92±1.54 mV, n=18, P<0.0001; Fig. 3B). In addition, H₂O₂ shoulder and trunk (amputated or uncut) TEP shifted in the same way as bud TEP (Figs S10A and S12C-E).

J_I magnitude and J_I reversal in the bud at 6 hpa were dramatically induced to values closer to those found during the regenerative period (DMSO stage 40-41: −0.10±0.065 µA cm⁻², 82.34±6.27% reversal, n=23; DMSO stage 45-46+H₂O₂: 0.68±0.296 µA cm⁻², 43.81±11.18% reversal, n=37; P=0.013; Fig. 3C, Fig. S10B). H₂O₂ significantly decrease J_I magnitude in 81% to a still outward current (from 3.64±0.897, n=31, to 0.68±0.296 µA cm⁻², n=37, P=0.003; Fig. 3C). Importantly, current reversals increased threefold (from 14.29±7.19 to 43.81±11.18%, P=0.044; Fig. 3C).

We then attempted to induce ectopic tail formation during the regenerative period. Upon dorsal incisions, H₂O₂ treatment for 24 h slightly induced significant ectopic (abnormal or complete) tail formation (up to a frequency of 19%) in a dose-independent way (Fig. S15). Altogether, short-term exposure to low-dose H₂O₂ is sufficient to induce regeneration, TEP increase and J_I reversal during the refractory period, in addition to inducing ectopic tail formation in the regenerative period.

Following the parallel responses of regeneration, TEP and J_I to the different conditions used, we tested whether bioelectricity predicts regeneration efficiency. Plotting all conditions together showed a significant linear correlation of regeneration efficiency with TEP magnitude (r=0.74, P=0.036; Fig. S14) and with the frequency of J_I reversals (r=0.86, P=0.006; Fig. 4).

To unveil the mechanisms by which H₂O₂ modulates regeneration, we searched for ion translocators that ideally affect regeneration and bioelectricity. Na_V1.2 is required for X. laevis tadpole tail regeneration via Na⁺ influx control (Tseng et al., 2010). Therefore, we tested whether Na_V channels act downstream of H₂O₂ using complementary epistasis assays. First, we verified that the inhibition of Na_V channels impairs regeneration. As expected (Tseng et al., 2010), the Na_V inhibitor tricaine (Frazier and Narahashi, 1975) significantly impaired regeneration (250 µM: RI reduced from 277, n=59, to 113, n=45, P<0.0001; 250 µM until 24 hpa: RI reduced to 202, n=64, P=0.008). A lower dose (150 µM) had less penetrant effects (Fig. 5A,A′). Next, we designed the epistasis assays: (1) rescue DPI-impaired regeneration using the Na_V activator aconitine; and (2) loss H₂O₂-induced regeneration in the refractory period using the Na_V inhibitor tricaine.

Treatment with aconitine (Ameri, 1998) until 24 hpa significantly rescued DPI-impaired regeneration (Fig. 5B,B′). Aconitine increased 3.6-fold the frequency of tails with full phenotypes in relation to DPI-treated tails (from 9 to 34%) and decreased by ∼80% the frequency of none phenotypes (from 34 to 7%). Overall, regeneration was significantly rescued (RI from 105, n=57, to 179, n=56, P<0.0001; Fig. 5B,B′). Interestingly, aconitine (in the µM range) was lethal in the absence of DPI, pointing to antagonistic competition.

Treatment with tricaine until 24 hpa significantly impaired H₂O₂-induced regeneration (Fig. 5C,C′). Tricaine virtually eliminated full phenotypes in relation to H₂O₂-treated tails (from 30 to 2%) and increased 3.4-fold the frequency of none phenotypes (from 17 to 57%). Overall, regeneration was significantly lost (RI from 168, n=135, to 65, n=55, P<0.0001; Fig. 5C,C′). Therefore, Na_V channels act downstream of H₂O₂ and mediate H₂O₂-modulated regeneration – a new mechanism.

Bioelectric and, more recently, redox activities have been shown, independently, to modulate regeneration in widespread models: planaria (Beane et al., 2011; Pirotte et al., 2015), zebrafish (Gauron et al., 2013; Monteiro et al., 2014) and amphibians (Love et al., 2013; Reid et al., 2009). Seeking a biochemical and biophysical integration, we aimed to determine whether and how these pervasive activities interact during regeneration.

Do ROS modulate regeneration via regulation of bioelectric activities that are known to be required for regeneration? Our results support the two-way regulation hypothesis proposed to address this question. First we mapped the spatiotemporal dynamics of extracellular bioelectricity (Fig. 1, Figs S2-S5) and the temporal requirement for NADPH oxidase-mediated ROS production (Fig. 2A,A′, Fig. S7). These profiles identified the bud at 6 hpa as key to testing the proposed hypothesis, owing to its bioelectric state. Next, we blocked NADPH oxidases and imaged V_m. We found that V_m depolarization resulted from NADPH oxidase activity per se (Fig. 2B,B′). This supported the first part of the two-way regulation hypothesis, arguing that the electrogenic property of NADPH oxidases is the mechanism by which V_m depolarization occurs. As depolarization is not required for regeneration (Adams et al., 2007), we then focused on the extracellular bioelectricity. Drug-depleted ROS levels decreased TEP and prevented or postponed J_I reversal, maintaining a steady outward current at 6 hpa (Fig. 2C,D). This mimicked the altered TEP and J_I measured during the refractory period (Fig. S11). Exogenous H₂O₂ rescued and induced regeneration, TEP and J_I (Figs 2, 3). It also induced ectopic tail formation (Fig. S15). Regeneration efficiency was directly proportional to TEP increase and J_I reversals (Fig. 4, Fig. S14). These results supported the second part of the two-way regulation hypothesis, arguing that H₂O₂ regulates TEP, EF and J_I. Ion translocators are known to mediate bioelectricity (McCaig et al., 2005), thus H₂O₂ is likely to be epistatic to translocators. Indeed, epistasis assays revealed that Na_V channels act downstream of H₂O₂ (Fig. 5). Altogether, the convergent results from disparate methods provide considerable consilience to conclude that redox-modulated regeneration is mediated by early bioelectric activities (Figs 6, 7).

Endogenous electric activities arise due to energetically expensive resting electrochemical potentials, primarily V_m and TEP. Can these biological batteries be, in part, dormant sensors and effectors ready to detect and to react to ubiquitous injuries? In fact, the idea that endogenous batteries play a role in wound healing (primarily guiding closure) is not entirely new (Barker et al., 1982; Borgens, 1984; Lykken, 1971). For an integrative view, we propose three phases of extracellular bioelectric contributions in regeneration: passive, active and homeostatic (Fig. 1E, Fig. S5).

Barrier breaking disrupts the in/out discontinuum so ions leak down their electrochemical gradients, resulting in injury current and lateral EF (primary) with the cathode at the amputation plane. The trunk TEP remains unchanged with amputation, thus the A/P gradient is resolved at this position. Thereby, the ‘bioelectric microenvironment’ – spatial range of physiologically meaningful bioelectricity – capable of influencing regeneration spans ∼335 µm anterior to the amputation plane. Importantly, this microenvironment matches maximum mobilization distances, ∼400 µm, of cells for wound closure in urodeles (Lash, 1955).

Presumptive barrier restoration re-establishes the discontinuum so ions diffuse with or without energy costs through the wound epithelium. Leaky epithelia are not highly efficient ion-flux barriers due to paracellular movement. Thus, the timing from passive to active activity is difficult to determine precisely. Rather, according to the TEP increase and J_I reversal frequency, a transition should occur from 1 to 6 hpa. Epithelial resistance plus J_I reversal leads to an exponential build-up of TEP, provoking an equivalent drop in EF. The current reversal possibly reversed short-range EF (secondary) locally in the bud (to anode). The plateau part in both TEP and J_I suggests that inward current magnitude should go beyond TEP recovery.

Ion fluxes should now be mainly for housekeeping functions, as TEP and J_I return to uncut baseline, resolving EF. Uncut and complete regeneration tails have an inside-positive TEP and a small inward J_I, in agreement with the salt uptake (osmoregulatory) function of amphibian skin (Koefoed-Johnsen and Ussing, 1958). Remarkably, TEP recovers to the new stage-specific level, not to the level before amputation (Fig. S2C), implying systemic control synchronizing development and regeneration. Bioelectric and regeneration phases dynamically correlate (Figs 1A,E, 4, Fig. S5), suggesting a control of regeneration by bioelectric events.

We showed that the characteristic V_m depolarization in the bud at 6 hpa mechanistically mimics the well-known NADPH oxidase-driven depolarization in immune cells during an oxidative burst (Bankers-Fulbright et al., 2003). This makes it attractive to speculate that V_m depolarization is a by-product of ROS production, which is reinforced by the non-requirement of V_m depolarization for regeneration. V_m should be repolarized by H⁺ efflux at 24 hpa to allow regeneration (Adams et al., 2007). We noticed that, irrespective of the drugs used, bud repolarization occurred by 24 hpa (Fig. S8A′), showing that redox state modulates regeneration in a V_m-independent way (Fig. 7). Similarly, robust V_m depolarization around the wound was measured in the caudal fin inter-ray wound model in zebrafish (F.F., unpublished). Importantly, Gauron et al. showed increased ROS levels produced by NADPH oxidases around wounds of the same model (Gauron et al., 2013). Thus, NADPH oxidase-driven V_m depolarization might be a universal injury-specific by-product of ROS production. However, depolarization is not excluded from having functions (Tseng et al., 2010) not addressed in the present study. In the past two decades, non-injury related V_m polarizations are being unveiled as being key in diverse developmental and patterning processes (Levin, 2014). Future work may unveil whether the activities of NADPH oxidases also have a role in these contexts, when ROS production is involved.

Reid et al. showed a correlation between electric current reversals (from outward to inward) and X. laevis tadpole tail regeneration. Ion substitution assays revealed that Na⁺ contributed to approximately two-thirds of the reversed current. However, neither ion nor translocator modulations were able to actually maintain the outward current (Reid et al., 2009). Remarkably, using non-ionic modulation, we either prevented or postponed current reversal in the bud at 6 hpa, mimicking the refractory period. This led to a dramatic inhibition of regeneration. We showed that H₂O₂ diffuses into the cell and stimulates the reversal. Thus, in an electrical equivalent circuit, we propose that H₂O₂ is a switcher that permits Na⁺-carried current to diffuse into the stump (Fig. 6B). In the same model, Tseng et al. unveiled part of the molecular basis for Na⁺ influx, the channel Na_V1.2 (Tseng et al., 2010). We found that H₂O₂ acts upstream of the Na_V channel family. Future work is necessary to reveal the mechanism of this activation, which can be indirect (e.g. redox-sensitive pathways or gene transcription of channels proteins/regulators) or direct (e.g. oxidation of channels cysteine residues, if available) (Fig. 6A). Na_V1.2 is present by 18 hpa in the bud, accumulating cations there (Tseng et al., 2010). The mismatch with our measured reversal at 6 hpa might imply other Na_V channel(s) in action, possibly in the wound epithelium. This would complete the path of Na⁺ from outside to bud, as barrier restoration prevents most ion leakage and reduces paracellular flux. In principle, however, some Na⁺ transporter(s) should be present in the wound epithelium to move the ions against the electrochemical gradient. The sharp TEP increase until 6 hpa supports this and might suggest the ubiquitous Na⁺/K⁺-ATPase as a candidate (Dubé et al., 2010). Thus, it is possible that H₂O₂ (or other ROS) modulates additional ion translocators (Ma, 2011; Matalon et al., 2003). Exogenous H₂O₂ rescued and induced regeneration and induced ectopic tails, suggesting that H₂O₂ provides indispensable morphogenetic information. H₂O₂ robustly rescued and induced TEP increase and J_I reversal at 6 hpa. Therefore, we propose that H₂O₂-switched current reversal is a hallmark of regeneration. As a limiting and stimulating factor of regeneration, this hallmark can be used as a diagnostic characteristic of regeneration and as a marker and prognosis of its efficiency (Fig. 4). Furthermore, it can be used to promote regeneration in regenerative-deficient animals.

J_I reversal prior to blastema formation was also seen in newt limbs (unquantified frequency, apparently occasional due to net outward current) (Borgens et al., 1977b) and tails (Nawata, 2001), and in X. laevis tadpole limb regeneration (G.L., unpublished). Reversal in newt tails was assumed to be a return to baseline (Nawata, 2001), despite the higher magnitude than before amputation. Surface potential measurements revealed stump potential reversal (from positive to negative) prior to blastema formation in urodels, but not in non-permissive anuran limb regeneration (Becker, 1961; Rose and Rose, 1974), showing a correlation with regeneration. However, another study found potential reversal in both taxa (Lassalle, 1979). We measured instead transepithelial potentials that, apart from sporadic reversals, had no overall mean reversal. Short-range EF_{Shoulder→Bud} presented more reversal frequency (21%) than did long-range EF_Trunk→Bud (9%) at 6 hpa (increasing thereafter; Fig. S3C), indicating the predominance of local reversal (in the bud). Indeed, in newt limbs, the surface potential reversal occurred only a short distance from the amputation plane (Rose and Rose, 1974). The small and soft bud hindered the spatial resolution required to fully characterize local reversals. Unambiguous current reversal supports the short-range potential and EF reversals in the bud. Altogether, our results showed a close correlation between reversals, bud formation and regeneration (Figs 4, 6, 7).

How can the H₂O₂-switched current reversal mechanistically impact regeneration? Cell migration precedes proliferation (Adams et al., 2007) and is essential for wound closure in the first phase of regeneration (Beck et al., 2009). Thus, the early and vectorial J_I reversal suggests migration as a prime candidate. However, other cell behaviors are not excluded. Cumulative evidence shows that endogenous or exogenous EF act as guiding cues for migration during wound healing (McCaig et al., 2005; Zhao et al., 1996, 2006). Importantly, most cells, including keratinocytes and neurons, migrate or grow towards the cathode (Song, 2004; Sun et al., 2013). ROS-depleted and refractory period tails at 6 hpa had a sustained outward J_I that led to lower bud TEP and higher short-range EF, establishing a long-term cathode at amputation plane (Figs 2C, 3B, Figs S11, S16A). Therefore, we postulate that a non-reversed J_I by 6 hpa might lead to cell overmigration. In short, cells are deluded into behaving as if the wound was still open (Fig. S16B). Wound epithelium from the refractory period is thicker than regenerative X. laevis tadpole tails (Beck et al., 2003; Reid et al., 2009; Tseng et al., 2010). Mature skin flapped or grafted to amputated newt limbs or children's fingertips blocks regeneration (Illingworth, 1974; Mescher, 1976). Interestingly, these skin-like structures have altered currents (Altizer et al., 2002; Reid et al., 2009) (Fig. S11) and might disrupt epithelial-mesenchymal crosstalk (Lee et al., 2009), both of which are important for regeneration. Thereby, cell overmigration might produce a non-permissive refractory-like wound epithelium. This might explain why J_I reversal by 24 hpa in MCI- and AgNO₃-treated tails (Fig. S8C) was not sufficient for regeneration (Fig. 2A,A′). Although we cannot exclude direct ROS-induced chemotaxis (Hurd et al., 2012; Pan et al., 2011), ROS-modulated electrotaxis might be the major cue (Zhao, 2009). This rationale could apply to axonal sprouting and to axonal growth (electrotropism). Neuronal presence is a long-recognized limiting and stimulating factor in regeneration (Kumar and Brockes, 2012; Sidman and Singer, 1951). Importantly, applied currents enhance nerve growth and regeneration in adult frogs (Borgens et al., 1979). Observed J_I reversal per se or the establishment of a steady inward current might be independent cues. Reversal can be a transitory electric pole discretely sensed by cells. Therefore, J_I reversal may fine-tune (mitigate) migration of cells and axonal growth, setting the conditions for or triggering bud initiation around 6 hpa and subsequent proliferation and de novo tail formation. Indeed, a permissive epithelium neuronally supplied (Thornton, 1954) is the basis of the accessory limb formation in axolotls (Endo et al., 2004). H₂O₂-induced ectopic tails might thus result from bioelectric effects on neuronal tissue.

We aimed to conceptualize a comprehensive model that integrates redox and bioelectric states during regeneration (Fig. 7). We showed that NADPH oxidase-mediated ROS production is pleiotropic (early on it affects morphogenesis; later on it affects growth) and is required immediately upon amputation. Thus, an elusive ultra-fast signal needs to trigger the activity of NADPH oxidases. Probably the fastest injury-induced signals are bioelectric in nature. Exogenous EF activate NADPH oxidases in cells in vitro (Chatterjee et al., 2012; Li et al., 2013). Thus, we propose that endogenous EF activate NADPH oxidases (Fig. 7). Another candidate is Ca²⁺ signaling. Wounding induces fast Ca²⁺ influx and Ca²⁺ flashes in X. laevis oocytes and Drosophila embryo, respectively (Luxardi et al., 2014; Razzell et al., 2013). Importantly, those flashes activate a NADPH oxidase (Duox). Excitingly, this points to a feedback module, where bioelectric signals (primary EF and/or Ca²⁺) might activate redox signals (NADPH oxidases and generated H₂O₂) that in turn regulate bioelectricity (V_m, TEP, secondary EF and J_I) to modulate regeneration (Fig. 7). This might be a starting point from which to test whether bioelectricity-modulated redox activities also occur during regeneration. By 6 hpa, the electrogenic property of NADPH oxidases depolarizes V_m and the catalytic property stimulates TEP recovery and switches on J_I reversal (Figs 6B, 7). H₂O₂-modulated regeneration is then V_m independent. However, both paths might interact. For example, preventing V_m repolarization downregulates Na_V1.2 expression (Tseng et al., 2010). H₂O₂ diffusion into cells revealed for the first time, to our knowledge, that aquaporins are potential new targets to modulate epimorphic regeneration (Fig. 6A). H₂O₂-switched current reversal, potentially via Na_V channels, may fine-tune cell migration and axonal growth that is essential for regeneration (Figs 6, 7).

Although the evidence supporting H₂O₂ is compelling, the role of other ROS in controlling bioelectric and regenerative outcomes should not be excluded (Li et al., 2013). Moreover, the possible importance of a H₂O₂ extracellular gradient is not excluded by this study. Indeed, it might be translated into a gradient of bioelectric activity that can be a blueprint for positional information and axes patterning (Jaffe, 1981; Tseng and Levin, 2013). Engineered solutions for controlled release of H₂O₂ (Garland et al., 2014) may prove useful for future work, especially in non-aquatic models.

In conclusion, this study unveils the interplay between biochemical and biophysical signals during regeneration, suggesting a novel mechanism that mediates regeneration: early H₂O₂-switched J_I reversal, potentially via Na_V channels. It also highlights the opportunities for interdisciplinary integration of apparently disparate states (Serena et al., 2009; Tandon et al., 2014) that may reserve promising advances for translational medicine.

Animal procedures and pharmacological treatments were approved by local Institutional Animal Care and Use Committee (protocols 16822 and 18601). Xenopus laevis (Daudin, 1802) tadpoles acquired from Xenopus Express (www.xenopus.com) were staged (Nieuwkoop and Faber, 1967), sorted and transferred to fresh Marc's modified Ringer (MMR) 0.1× culture medium, composed of (in mM): NaCl 10, CaCl₂·2H₂O 0.2, KCl 0.2, MgCl₂·6H₂O 0.1 and HEPES 0.5 (pH 7.1-7.2) (Sigma-Aldrich). Surgery was executed as previously described (Adams et al., 2007; Reid et al., 2009). For further details, see Supplementary Materials and Methods.

Tail regeneration assays were performed as previously described (Adams et al., 2007; Tseng et al., 2010). For further details, see Supplementary Materials and Methods.

Ectopic tails were induced by a dorsal incision severing down to the spinal cord (inclusive; Fig. S15A) (Sugiura et al., 2009). For further details, see Supplementary Materials and Methods.

Drugs and respective vehicle reconstituted in MMR 0.1× were applied via immersion (bath) at the doses and exposures specified. Most drugs were stocked in DMSO (Sigma-Aldrich, cat. no. D2650) at −20°C: DPI (Sigma-Aldrich, cat. no. D2926) 1 and 10 mM; MCI (Cayman chemical, cat. no. 13320) 200 mM; AgNO₃ (Sigma-Aldrich, cat. no. 209139) 24 mM; apocynin (Sigma-Aldrich, cat. no. W508454) 200 mM; VAS (EMD Millipore, cat. no. 492000) 1.4 mM; and aconitine (Alamone Labs, cat. no. A-150) 1 mM. Vehicle (DMSO) had non-significant effects on readouts (Fig. S1). H₂O₂ (EMD Millipore, cat. no. HX0635) was kept as a 1 M stock in deionized water at 4°C. Occasionally, stocks of 25 and 50 mM were kept for short periods at −20°C. Tricaine was kept as a 1 mM stock in MMR 0.1× (pH readjusted to 7.2) at −20°C. Working solutions were freshly prepared. For further details, see Supplementary Materials and Methods.

Intracellular relative ROS and V_m were measured non-invasively by fluorescence imaging of vital dyes as previously described (Adams et al., 2007; Owusu-Ansah et al., 2008). For further details, see Supplementary Materials and Methods.

Transepithelial potential was measured invasively by glass microelectrode impalement through the epithelial layers as previously described (Luxardi et al., 2014; McCaig and Robinson, 1982). For further details, see Supplementary Materials and Methods.

Extracellular net electric current density was measured non-invasively with a locked-in one dimensional vibrating probe, prepared and used as previously described (Reid et al., 2007, 2009). For further details, see Supplementary Materials and Methods.

Regeneration efficiencies were compared using the Fisher's exact test. Frequencies (in J_I reversals and ectopic tails) were compared using the Mann-Whitney U-test. Correlations were tested with the Pearson's correlation coefficient r. Linear regressions (coefficient of determination r²) were applied to fit data to a statistical model. All remaining comparisons were analyzed using the unpaired Student's t-test.

Data are presented as mean±s.e.m. Sample size (n, biological replicates) used in each experiment are stated in figures and text. At least two independent batches of tadpoles were used per readout. Differences were significant if P<0.05 and the levels of significance are: NS, non-significant; *P<0.05; **P<0.01 and ***P<0.001. Means and standard errors were calculated and plotted using Excel, and statistics were performed using GraphPad Prism 5 (GraphPad Software). For further details, see Supplementary Materials and Methods.

We apologize for not being able to cite all the relevant studies due to space or context constraints. We are grateful to Dr Andreia Gomes (Departamento de Biologia, CBMA, Universidade do Minho, Portugal) and to Dr VijayKrishna Raghunathan [Department of Surgical and Radiological Sciences, University of California (UC) Davis; The Ocular Surface Institute, College of Optometry, University of Houston, TX, USA] for helpful discussion. We thank Dr Paul Fitzgerald (Department of Cell Biology and Human Anatomy, UC Davis) and Dr Andrew Ishida [Department of Neurobiology, Physiology, and Behavior (NPB), UC Davis] for providing access to confocal microscopes. We are grateful to Dr Tyler Stradleighand (Department of NPB, UC Davis) and to Dr Qizhi Gong (Department of Cell Biology and Human Anatomy, UC Davis) for training and suggestions in confocal microscopy. We thank Dr Yao-Hui Sun for providing zebrafish and all other lab members (past and present) for helpful discussion.