Transmembrane voltage potential controls embryonic eye patterning in Xenopus laevis

Embryonic eye development offers the opportunity to understand the genetic and biophysical mechanisms that drive the self-assembly of complex biological structures (Erclik et al., 2009). Eye tissue formation begins with the regionalization of anterior neural plate into an ‘eye field’, as evidenced by the expression of eye field transcription factors (EFTFs) (Zaghloul et al., 2005). The eye field is further resolved into bilateral eye primordia by the dynamic expression of EFTFs (Zuber, 2010). However, relatively little is known about the regulation of EFTF bilateral refinement. EFTFs are not induced until neurulation and are not exclusively involved in eye development; Pax6 and Rx1 play important roles in brain development, neuronal migration and other processes. This suggests the existence of as yet uncharacterized additional instructive signal(s) that drive certain embryonic cells specifically toward eye formation. Recent studies identified a cocktail of factors that converts animal cap ectoderm to an eye fate (Viczian et al., 2009; Zuber et al., 2003). Could eyes be induced by an even more minimal set of signals? And, could eye formation be induced in vivo at locations far from the anterior neural field? The identification and characterization of such signals would have tremendous potential for therapeutic applications, as well as shedding light on endogenous mechanisms of lineage restriction and morphogenetic control.

Endogenous steady-state ion currents, voltage gradients and electric fields produced by ion channels and pumps in all cells (not just excitable neurons and muscle) are crucial regulators of patterning (Levin, 2009; McCaig et al., 2009). They control cell migration (McCaig et al., 2009; Prevarskaya et al., 2010; Yao et al., 2008; Yasuda and Adams, 2010), proliferation (Blackiston et al., 2011) and differentiation (Sundelacruz et al., 2009) in a wide variety of systems. Although epithelial electric fields have been implicated in adult eye wound healing (Reid et al., 2005; Wang et al., 2005; Zhao et al., 2006), the roles of ion flows in eye development have not been explored. In Xenopus laevis, we have uncovered an endogenous role for plasma membrane voltage gradients during eye development, and show that artificially induced changes in transmembrane potential (V_mem) are sufficient to induce well-formed eyes outside of the anterior neural field. Thus, V_mem is an important endogenous component of the eye induction cascade and presents a tractable control point for the therapeutic induction of visual system tissues.

Xenopus laevis embryos were fertilized in vitro according to standard protocols (Sive et al., 2000) in 0.1× Marc's Modified Ringer's (MMR; 10 mM Na⁺, 0.2 mM K⁺, 10.5 mM Cl^-, 0.2 mM Ca²⁺, pH 7.8]. Intracellular ion concentrations in Xenopus are: 21 mM Na⁺, 90 mM K⁺, 60 mM Cl^-, 0.5 mM Ca²⁺ (Gillespie, 1983). Xenopus embryos were housed at 14-18° C and staged according to Faber and Nieuwkoop (Faber and Nieuwkoop, 1967). Capped synthetic mRNAs generated using mMessage mMachine (Ambion) were resuspended in water and injected (3 nl per blastomere) into embryos in 3% Ficoll. All experiments were approved by the Tufts University Animal Research Committee in accordance with the Guide for Care and Use of Laboratory Animals.

Fresh CC2-DMPE (Molecular Probes) stocks (1 mg/ml in DMSO) were diluted 1:1000 in 0.1× MMR to 0.2 μM. Embryos were soaked in CC2-DMPE dye for 1.5 hours followed by five washes in 0.1× MMR. Embryos in 0.1× MMR were imaged using 405 nm excitation and 460 nm emission on a Nikon AZ100 microscope with an Andor Luca digital CCD camera and NIS-Elements software.

Embryos were fixed in MEMFA [100 mM MOPS (pH 7.4), 2 mM EGTA, 1 mM MgSO₄, 3.7% (v/v) formaldehyde] and in situ hybridization was performed as described (Harland, 1991). Probes used were Otx2 [gift of Dr Valerie Schneider (Pannese et al., 1995)], Pax6 [gift of Dr Ali Hemmati-Brivanlou (Hirsch and Harris, 1997)] and Rx1 [gift of Dr Michael Zuber (Andreazzoli et al., 1999)].

Embryos were fixed overnight in MEMFA and stored at 4° C in PBTr (1× PBS, 0.1% Triton X-100), embedded in paraffin and sectioned at 10 μm. Sections were washed three times in PBTr, blocked with 10% goat serum, incubated with primary antibody in PBTr overnight, washed six times with PBTr, and incubated with Alexa Fluor-conjugated secondary antibody (Invitrogen; 1:1000) for 4 hours at room temperature. Sections were then washed, mounted and viewed with an Olympus BX61 spinning-disk confocal microscope with an ORCA digital CCD camera (Hamamatsu) controlled with MetaMorph software. Primary antibodies were: β-crystallin (Abcam ab90379; 1:1000) (Secchi et al., 1976); Glutamine synthetase (Sigma G2781; 1:500) (Pahuja et al., 1985); XAP2 (DSHB clone 5B9; 1:10) (Harris and Messersmith, 1992); Calbindin (Sigma C2724; 1:500) (Gargini et al., 2007); and Islet-1 (DSHB clone 39.4D5; 1:100) (Ericson et al., 1992).

Stocks of Ivermectin (IVM) (Sigma) were prepared at 10 mM in DMSO. Embryos were exposed in 0.1× MMR to: 10 μM IVM, 0.1 mM Verapamil, 10 μM Fluoxetine or 10 μM Lindane.

Membrane potentials were measured using an oocyte clamp OC-725C amplifier (Warner Instruments, Hamden, CT, USA) with a single voltage electrode. Microelectrodes were made from thin-walled borosilicate glass pulled with a flaming/brown micropipette puller (P-97, Sutter Instruments, Novato, CA, USA) and back filled with electrode solution (2 M potassium acetate, 10 mM KCl, 5 mM HEPES pH 7.5). Tip resistances were 80-100 Mω. Electrode penetration of ectodermal cells was by visual guidance on a fixed-stage microscope (Zeiss) using a three-axis micromanipulator.

EosFP mRNA (Gurskaya et al., 2006; Nienhaus et al., 2006; Wacker et al., 2007; Wiedenmann et al., 2004; Wiedenmann and Nienhaus, 2006) was injected into Xenopus embryos immediately after fertilization at the one-cell stage. At stage 17/18, the embryos were soaked in CC2-DMPE dye. The CC2-DMPE staining was documented with an Olympus BX61 microscope. The CC2-DMPE-indicated region was exposed to DAPI excitation wavelength using a 40× lens. The photoconversion was documented immediately at stage 17/18 and again at stage 28.

To document real-time changes in V_mem during eye development, we used the voltage-sensitive dye CC2-DMPE (Wolff et al., 2003) to detect hyperpolarization in the Xenopus embryo in vivo. A physiological screen for bioelectric patterns during craniofacial development (Vandenberg et al., 2011) revealed that, starting at stage 17/18, normal embryos exhibit two small areas in the anterior neural field that are bilaterally positioned and hyperpolarized relative to their neighbors (Fig. 1A, red arrowheads). Whole-cell electrophysiological recordings of V_mem from these areas and from the neighboring tissues on the flank of the same embryo (Fig. 1B) revealed that the CC2-DMPE-demarcated domains were hyperpolarized relative to neighboring tissues by ∼10 mV (also confirming the ability of the CC2-DMPE dye to detect cells with hyperpolarized V_mem). Other regions of the embryo also exhibit interesting distributions of V_mem; these are described elsewhere (Vandenberg et al., 2011) and will be investigated in subsequent work.

V_mem collapses upon cell death and thus cannot be imaged simultaneously with in situ hybridization. We imaged the hyperpolarized V_mem signal using CC2-DMPE at stage 17/18 and immediately fixed the embryos for in situ hybridization analysis of the EFTFs Otx2, Rx1 and Pax6 (Fig. 1C; data not shown). The relatively broad anterior expression of EFTFs appeared to overlap with the hyperpolarized cells (Fig. 1C; data not shown).

To overcome the incompatibility of in situ hybridization with in vivo physiological imaging and to more precisely characterize the timing and co-localization of the observed hyperpolarized cell cluster in relation to the formation of eye primordia, we verified the overlap by labeling the CC2-DMPE-demarcated hyperpolarized cells using a photoconvertible protein (Wacker et al., 2007). Fertilized eggs were injected with EosFP (a green fluorescent protein) mRNA and, at stage 18, as small an area as possible centered on the hyperpolarized cells was photoconverted. Observation of the resulting red signal at stage 28 confirmed that the hyperpolarized cells indeed contribute to eye formation (supplementary material Fig. S1A-C). Agarose sectioning of the embryo through the eye showed that the photoconverted cells contribute to both the lens and the retinal layers of the eye (supplementary material Fig. S1D-F). The hyperpolarized area pattern remains bilateral and largely unchanged through anterior development (stage 22; Fig. 1C; supplementary material Table S1). By contrast, the EFTFs undergo a bilateral regionalization by stage 22, as expected (Fig. 1C) (Viczian et al., 2009; Zuber et al., 2003). The distinct bioelectrical regionalization of the anterior neural field suggests that the bilateral hyperpolarized areas might be involved in the processes by which dynamic expression of the EFTFs regulates the formation of eyes.

To determine whether hyperpolarization is functionally required for eye development, we performed a loss-of-function experiment by depolarizing V_mem of the relevant cells. We used the constitutively conductive EXP1 cation channel (Beg and Jorgensen, 2003) and the glycine-gated chloride channel (GlyR) (Davies et al., 2003), which can be permanently activated by the specific GlyR opener Ivermectin (IVM) (Shan et al., 2001). The use of two different molecular reagents (channels that share no sequence or structural homology and have in common only their effect on membrane potential) specifically tested the role of V_mem per se, independent of possible other functions of a particular channel protein or ion species.

In standard 0.1× MMR medium, these channels work to depolarize cells (supplementary material Fig. S2A) (Beg and Jorgensen, 2003; Blackiston et al., 2011; Davies et al., 2003). Embryos were injected with GlyR or EXP1 mRNA in the dorsal two cells of the four-cell embryo [the blastomeres from which eyes derive (Moody, 1987)] (supplementary material Fig. S2Ai) and were stained with CC2-DMPE dye at stage 19 (supplementary material Fig. S2Aii). GlyR-injected embryos were treated with the channel opener IVM from stage 17 to 23 to enable the efflux of chloride that depolarizes GlyR-expressing cells. In contrast to uninjected embryos, which show distinct bilaterally hyperpolarized cell clusters (supplementary material Fig. S2Aiii, green arrowheads), GlyR and EXP1 misexpression strongly reduced or disrupted the hyperpolarization pattern in the injected embryos (supplementary material Fig. S2Aiv). This confirmed that our reagents produce the expected loss-of-function effect on the hyperpolarization pattern. We saw no signs of general toxicity or ill-health from ion channel mRNAs, which were titrated to the lowest levels needed to produce eye phenotypes: dorsoanterior index, midline patterning, overall growth rate and proportion, and behavior were all normal, ruling out simple toxicity as a cause of eye malformation. Consistent with existing evidence of V_mem changes being associated with specific tissue outcomes (Blackiston et al., 2011; Levin et al., 2002; Levin, 2009; Sundelacruz et al., 2009; Tseng et al., 2010), we occasionally observed additional interesting patterning phenotypes involving other organs, which will be characterized in subsequent studies.

To determine whether the observed hyperpolarization is required for normal eye development, GlyR or EXP1 mRNA was injected into the dorsal blastomeres at the four-cell stage and the embryos scored for eye development at stage 42 (Fig. 2A). Uninjected, lacZ-injected, GlyR-injected and IVM-treated embryos served as controls. Control embryos had intact, fully closed, well-formed eyes (Fig. 2B). By contrast, expression of GlyR plus IVM treatment, or expression of EXP1 mRNA, caused a high incidence of disrupted endogenous eye formation (48% and 46%, respectively; Fig. 2A). Phenotypes included incomplete eye formation (supplementary material Fig. S2Bi), small (supplementary material Fig. S2Bii) or absent (Fig. 2B) eyes, fusion to the brain (supplementary material Fig. S2Biii) and pigmented optic nerves (supplementary material Fig. S2Biv).

Although bioelectric events can exert effects at long range (Blackiston et al., 2011; Morokuma et al., 2008), the position of the bilateral hyperpolarized regions suggested a local role. We examined the effects of inducing depolarization elsewhere in the embryo. Injection of GlyR mRNA plus IVM treatment or of EXP1 mRNA into the two ventral blastomeres (which make a minor contribution, if any, to eye fate lineages) caused no significant effect on eye development (Fig. 2A). To allow a direct comparison of the injected and uninjected sides of the same embryo and to further test the effective range of influence of hyperpolarization, only one side (one dorsal cell at the four-cell stage) of the embryo was targeted by depolarizing channel mRNA (GlyR plus IVM, Fig. 2B; EXP1, supplementary material Fig. S2Biii) while leaving the contralateral side as an internal control. The mRNA was mixed with a lineage tracer (lacZ mRNA); eye defects occurred only in regions that had been targeted by the depolarizing mRNA, whereas normal eye development occurred on the uninjected region of the same embryo (Fig. 2B; supplementary material Fig. S2Biii). Taken together, these results demonstrate that the bilateral hyperpolarized cell clusters in the anterior neural field are an important, specific component of normal eye development and that this signal works in a local fashion.

To confirm the role of specific V_mem levels in eye induction, we took advantage of the ability to control the V_mem of GlyR-expressing cells (plus IVM) by modulating the extracellular chloride content [Cl^-]_ex. The depolarizing effect of GlyR misexpression can be progressively reduced by increasing [Cl^-]_ex toward the intracellular chloride concentration (Blackiston et al., 2011; Davies et al., 2003), at which point chloride efflux (and depolarization) will cease. After GlyR mRNA injection into the dorsal two cells at the four-cell stage, we incubated the embryos (with IVM) at different levels of [Cl^-]_ex (stage 17-23). A progressive decrease in the incidence of the malformed eye phenotype (from 48% to 21% at stage 42) was seen as [Cl^-]_ex was increased toward isomolar (60 mM) levels, which abolished the normal outward gradient for these negative charges (Fig. 2C, red box). The ability to rescue the phenotype and control eye patterning by modulating [Cl^-]_ex, in a dose-responsive manner in accordance with the predictions of the Goldman equation, points to the importance of specific V_mem levels in eye development. We could not achieve 100% penetrant rescue, as the levels of [Cl^-]_ex could not be raised to higher values without perturbing other voltage-sensitive systems during patterning that might act as confounding factors (Blackiston et al., 2011; Levin et al., 2002; Levin, 2009; Sundelacruz et al., 2009).

The bilateral hyperpolarization of cell clusters precedes the bilateral regionalization of the EFTFs (Rx1 and Pax6) (Fig. 1) and this V_mem signal is important for proper eye development (Fig. 2). But where does the hyperpolarization V_mem signal function within the hierarchy of known signals, such as Otx2, Rx1 and Pax6? We analyzed the expression of these genes in V_mem-perturbed embryos. Embryos were injected in the dorsal two cells at the four-cell stage with depolarizing mRNAs (GlyR or EXP1) and fixed at stage 22. Otx2, Rx1 and Pax6 mRNA expression was detected using antisense mRNA probes (Fig. 2D). Otx2 is a marker for anterior neural plate (Acampora et al., 1995) and Pax6 and Rx1 are highly expressed in the eye field and in certain regions of the central nervous system, including the hypothalamus and the pineal gland (Bailey et al., 2004; Gehring and Ikeo, 1999). The expression pattern of Otx2 remained unchanged by depolarization (Fig. 2D, green arrowheads), suggesting that overall anterior neural/brain development [which provides crucial signals for eye induction (Lupo et al., 2002)] is not affected by the depolarization. This was confirmed by the observation that there was no change in the robust midline expression of sonic hedgehog (Shh) in treated embryos (supplementary material Fig. S2D), nor in the non-eye neural expression of Pax6 (Fig. 2D, black arrowheads). By contrast, expression of the EFTFs Rx1 and Pax6 within the eye domains was significantly reduced (in area, not expression intensity) or mispatterned in embryos in which the V_mem of hyperpolarized cell clusters had been artificially depolarized (Fig. 2D, yellow arrowheads). These results suggest that this bioelectrical signal functions to regulate the bilateral patterning of Pax6 and Rx1, confirming the functional importance of V_mem for the normal events of eye induction.

Feedback loops and cross-talk between biophysical and genetic pathways are common in patterning systems (Huang and Ingber, 2006; Ingber, 2006; Levin, 2006; Levin, 2009). We asked whether EFTFs could in turn affect the V_mem of the hyperpolarized cell clusters. Microinjection of mRNA encoding a dominant-negative mutant of Pax6 (DNPax6) resulted in embryos that lack eyes or have small eyes (Chow et al., 1999) (supplementary material Fig. S2C). We incubated the uninjected and DNPax6-injected embryos in CC2-DMPE dye to monitor the hyperpolarization signal. Interestingly, reduction in Pax6 activity caused a loss of the hyperpolarization signal in the anterior neural field (Fig. 2E), suggesting the presence of a positive-feedback loop between the hyperpolarization signal and Pax6 expression in the presumptive eye field cell cluster.

Having shown that a specific bioelectrical state is necessary for normal eye development, we next asked whether modulation of V_mem outside of the normal eye field could be sufficient to induce ectopic eyes. The V_mem of cells is determined by the combined activity of numerous endogenously expressed channels and pumps. Extensive expression profiling and quantitative physiological modeling would be required for a comprehensive understanding of the natively expressed conductances that give rise to distinct V_mem signals within the Xenopus face. Our interest was in probing a range of non-eye regions for their capacity to form eyes and in developing a gain-of-function technology that might be useful in regenerative applications without requiring a complete physiomic analysis. Hence, we targeted a widely expressed ion translocator.

The ATP-sensitive potassium channel (K_ATP) is a biomedically important ion channel that is involved in linking V_mem to cellular functions in many systems (Nichols, 2006). All four subunits of the K_ATP channel (Kir6.1, Kir6.2, SUR1 and SUR2) were found to be expressed in the head region (including the eye fields) and along the dorsal trunk regions of the embryo from early development (stage 15/16) to tail bud stages (data not shown). We used two well-characterized dominant-negative constructs targeting Kir6.1: either the Kir6.1 pore mutant (DNKir6.1p) or the endoplasmic reticulum (ER)-retention mutant (DNKir6.1-ER) (Ficker et al., 2000; Schwappach et al., 2000).

To verify the effect of DNKir6.1p constructs on the V_mem of injected cells, we injected the mRNAs into the dorsal two cells of four-cell embryos and then imaged them using the CC2-DMPE dye. Control (uninjected) embryos showed the characteristic hyperpolarization pattern (supplementary material Fig. S3A). DNKir6.1p-injected embryos showed highly diminished, or an absence of, hyperpolarized cell clusters (supplementary material Fig. S3A), confirming the ability of DNKir6.1p to depolarize cells.

mRNAs encoding these dominant-negative K_ATP constructs were injected into all four blastomeres of four-cell embryos. The embryos were scored for eye phenotype at stage 42. Green fluorescent protein (GFP) and lacZ mRNA negative-control injections did not produce any eye-relevant phenotypes (Fig. 3A-C). By contrast, DNKir6.1p and DNKir6.1-ER injections resulted in a high incidence (45% and 35%, respectively) of eye-related phenotypes (Fig. 3A). Approximately 25% of injected embryos had defects in their endogenous eyes, as expected when dorsal cell V_mem is perturbed (Fig. 3A,D). Strikingly, ∼20% of the embryos exhibited ectopic eye tissue (only lens or only pigmented epithelium), including patches of retinal pigmented epithelium (RPE) (Fig. 3E, red arrowheads) and lens tissue (Fig. 3I,J, blue arrowheads). We also observed embryos (7.5%) with ectopic and well-formed large complete eyes (RPE, retina and lens) (Fig. 3F-H, red arrowheads). Remarkably, ectopic eyes were found not only in the head but also on the gut (Fig. 3G, red arrowhead) and other caudal areas (Fig. 3H, red arrowheads). Ectopic lens tissue, revealed by β-crystallin antibody, occurred in the head region (Fig. 3J, inset) and even the tail (Fig. 3I, blue arrowheads), with islands of ectopic lens tissue found on both the ventral and dorsal side of the embryo (Fig. 3J, blue arrowheads). Although the majority of the ectopic tissues formed in the head (supplementary material Fig. S3B), a unique aspect of this phenotype is that eye components were frequently observed in locales well outside the anterior neural field, including the distal tail (Fig. 3H-J) and the lateral plate mesoderm (Fig. 3L). This was not accompanied by any axial or anterior duplication.

Our results suggest that perturbation of V_mem is sufficient to induce an eye fate even in regions that are outside of the normal fate map of tissues competent to become eyes (Zaghloul et al., 2005). Moreover, a single mRNA encoding a bioelectric target is sufficient to kick-start the entire eye development process in distal tissues.

To ensure that the key instructive signal was indeed carried by V_mem per se, and not by an ion-specific or ion-independent (e.g. scaffolding) role of our channel constructs, we tested several different ion channels that share no sequence or structural homology but do control V_mem. A majority of these induced ectopic eyes, as predicted (supplementary material Table S2). We also asked whether the eye phenotype (ectopic eyes and disrupted endogenous eyes) induced by the depolarizing DNKir6.1p channel could be rescued by co-injection of hyperpolarizing Kv1.5 channel. Among the embryos co-injected with Kv1.5 and DNKir6.1p, only 2% had ectopic eyes and 11% had disrupted eyes, in comparison to 20% and 25%, respectively, in embryos injected with only DNKir6.1p. The rescue of phenotype by a genetically unrelated but physiologically counteractive protein demonstrates the importance of V_mem itself as a regulatory factor in the modulation of eye development.

Further, we analyzed whether a narrow V_mem range is required for eye formation. We injected embryos with mRNA encoding a sodium channel, Nav1.5, and incubated them in medium containing a range of [Na⁺]_ex, thus generating conditions ranging from hyperpolarizing to depolarizing (as predicted by the Goldman equation). Maximum ectopic eye induction was observed within a small range of hyperpolarizing [Na⁺]_ex (Fig. 3N), confirming specific V_mem values as the information-bearing signal. Taken together, these results support the hypothesis that regulation of V_mem per se is an instructive signal in initiating eye induction.

We then performed experiments to determine whether ectopic eye induction involves: (1) the ‘migration’ of endogenous eye fate cells to ectopic locations upon perturbation of their V_mem; (2) the ‘colonization’ of ectopically injected tissues (perturbed V_mem) by attracting endogenous eye fate cells; or (3) the induction of non-eye fate cells to take on eye fate upon perturbation of their V_mem. We injected DNKir6.1p (with lacZ mRNA as a lineage label) into the dorsal and ventral blastomeres at the four-cell stage. There was no statistical difference between the incidence of ectopic eyes in dorsal (16%) versus ventral (20%) injections (data not shown). Further, whole-mount staining and sections through the ectopic eye tissue at stage 43 showed that all of the ectopic eye tissues were β-gal-positive (supplementary material Fig. S3Biii; data not shown). These results suggest that ectopic eye tissues are induced in a local manner by inducing the injected non-eye fate cells to assume eye identity and not by the migration of, or colonization by, endogenous eye-fated cells.

We next characterized in detail the ectopic eyes induced by V_mem change, to determine how closely the ectopic structures resemble normal eyes. Sectioning at stage 46 revealed that the ectopic tissues (Fig. 3K,L, red arrowheads) have a cup-shaped architecture similar to that of endogenous eyes (Fig. 3K,L, green arrows). Histological staining with Hematoxylin and Eosin showed that the ectopic cells (Fig. 3Mii) are organized in morphological layers similar to those of wild-type endogenous eyes (Fig. 3Mi).

We were able to identify and localize the key retinal cell populations in endogenous and ectopic eye tissues by immunostaining: Muller glia (Glutamine synthetase, red), amacrine cells (Islet-1, cyan), rods (XAP2, yellow) and cones (Calbindin, green) (Fig. 4A). A functional retina requires a precise arrangement of retinal cell layers. Co-immunolocalization showed that, in the ectopic eye tissue (Fig. 4B), the Muller glia, amacrine cells, rods and RPE are arranged almost identically to those in wild-type or endogenous eyes (Fig. 4B). Our histological and immunohistochemical analysis showed that bioelectric induction can give rise to most tissues present in the mature eye. Moreover, the ectopic structures formed by perturbation of V_mem contain the same tissues, in a similar arrangement, as endogenous eyes.

Pax6 has been shown to induce ectopic eyes in Xenopus embryos upon overexpression (Chow et al., 1999; Zuber et al., 2003). Does V_mem modulation induce eyes through the same EFTFs that are involved in primary (endogenous) eye development? Embryos were injected with DNKir6.1p mRNA, fixed at stage 30, and expression of the anterior neural marker Otx2 and of the EFTFs Rx1 and Pax6 detected by in situ hybridization.

DNKir6.1p injections (which produce ectopic eyes) did not induce ectopic expression of Otx2; by contrast, injected embryos showed numerous foci of ectopic Pax6 and Rx1 expression well away from the head (Fig. 4C). Thus, bioelectric induction of ectopic eye tissue involves the upregulation of key EFTFs. These data also demonstrate that V_mem changes are able to induce de novo expression of EFTFs (Fig. 4C), in addition to regulating their bilateral patterning (Fig. 1, Fig. 2D).

To determine how V_mem changes are transduced into effector cascades during eye development, we injected Kv1.5 mRNA into two cells of four-cell embryos to induce ectopic eye tissue and performed a simple suppression screen (Table 1), blocking a range of targets known to mediate the linkage between bioelectric and transcriptional pathways in other systems (Levin, 2007). Blockade of gap junction-mediated or serotonergic (Fukumoto et al., 2005) signaling did not prevent ectopic eye tissue induction. By contrast, blockade of voltage-gated Ca²⁺ channels (VGCCs) (100 μM Verapamil), significantly reduced the ectopic eye tissue induction by ion channel misexpression. Hence, changes of V_mem that induce ectopic eye tissues are likely to be transduced into downstream cascades by VGCC-mediated Ca²⁺ influx — a common mechanism for coupling electrical signaling to changes in cell behavior (Nakanishi and Okazawa, 2006; Nishiyama et al., 2008).

Morphogenetic roles for V_mem have been demonstrated in egg-oocyte systems (Woodruff and Telfer, 1980), appendage regeneration (Adams et al., 2007) and axial patterning (Beane et al., 2011; Kurtz, and Schrank, 1955; Marsh and Beams, 1947; Shi and Borgens, 1995) in a wide variety of species (Levin, 2009; McCaig et al., 2009). Having screened for bioelectrical pre-pattern during the development of Xenopus embryos (Vandenberg et al., 2011), we observed a bilateral, distinctly hyperpolarized cluster of cells in the anterior neural field and show that this hyperpolarization is required for normal eye development.

Eye defects were induced by two independent means of depolarizing anterior neural field cells: misexpression of a depolarizing cation channel (EXP1) or of a chloride channel (GlyR). In the latter case, modulation of [Cl^-]_ex allowed control of the magnitude of the effect (in accordance with the Goldman equation) (Blackiston et al., 2011). The ability to disrupt eye development using distinct constructs that mediate the flux of different ions (sodium, potassium or chloride) and have in common only their effect on V_mem strongly supports the V_mem signal as a factor necessary for normal eye development.

Molecular analysis revealed that perturbing the endogenous V_mem signal altered the endogenous expression pattern of Rx1 and Pax6. Additionally, manipulating the V_mem of non-eye cells induces de novo Rx1 and Pax6 expression. These results suggest that the V_mem signal functions in defining the pattern of these canonical EFTFs. Analysis of the timing of the bilateral hyperpolarization signal, the bilateral regionalization of Rx1/Pax6 expression to eye domains, and the ability of dominant-negative Pax6 to disturb the normal hyperpolarization pattern, all suggest a feedback loop between the biophysical and transcriptional signals. Cross-talk between Pax6 and the V_mem signal might result in a dynamic bilateral regionalization of the eye field, similar to the role of progressive depolarization in the differentiation of human mesenchymal stem cells (Sundelacruz et al., 2009).

Consistent with the hyperpolarization-induced expression of eye genes such as Pax6 and Rx1, eye tissue and indeed complete eyes can be formed by just one ion channel mRNA that modulates V_mem. Although often only patches of individual tissue (RPE or lens) are formed, complete whole eyes can be produced that have the tissue layers and anatomical structure of native eyes. The same phenotype resulted from induced flows of different ion species and could be manipulated by direct control of ionic gradients precisely as predicted by the Goldman equation. These data reveal that eye induction is mediated by changes in V_mem itself — a physiological parameter that is not necessarily tied to any one gene product, much as substrate geometry is now known to be a physical parameter controlling cell behavior (Chen et al., 1998; Dike et al., 1999). This is a desirable property from the perspective of developing biomedical applications because interventions (e.g. small-molecule drug cocktails designed to control V_mem in key cell groups) are not dependent on any one protein but can take advantage of convenient channels or pumps expressed in the host.

Our data show that the bilateral hyperpolarizing V_mem signal in the anterior neural field plays a crucial role in regionalization of EFTF expression and thus eye development. Upon perturbation of the bilateral V_mem signal, a major proportion of the malformed eye phenotypes (74%) represent failure of eye field separation (e.g. eyes fused to brain, eyes fused along midline, and eyes with pigmented optic nerve/stalk). We propose a hypothetical model in which feedback between V_mem and Pax6 establishes a stable, bilaterally regionalized pattern of Pax6 and other EFTF expression (Fig. 5A). Conversely, induction of an appropriate V_mem signal in non-eye cells results in ectopic induction and maintenance of EFTFs, driving the formation of ectopic eye tissues, perhaps by establishing the same stable feedback loop.

Neural tissue and Otx2 expression are thought to be crucial for eye induction (Zuber et al., 2003). The ability to form ectopic eyes without induction of additional primary axes or central nervous system tissue (lack of ectopic Otx2) is very interesting and suggests a more direct (downstream) linkage of V_mem to eye-specific cascades (Fig. 5A). Indeed, bioelectrically induced eyes can be formed on the gut or within lateral mesoderm — non-neural tissues outside the known competence lineage for eyes. Similarly, patches of lens/retina readily form in the tail. By contrast, Pax6 alone is incapable of initiating an eye field in regions outside the competent neurectoderm, suggesting that additional factors might also be involved in positive-feedback interactions with V_mem. However, V_mem is able to initiate not only Pax6 expression but also the conditions that make tissues competent to form eyes upon Pax6 transcription.

Temporally and spatially restricted hyperpolarization has previously been shown to drive cellular fate (Hinard et al., 2008; Konig et al., 2004; Konig et al., 2006; Sundelacruz et al., 2009). Moreover, specific cellular responses to windows of V_mem have also been documented (Gruler and Nuccitelli, 1991; Jurkat-Rott et al., 2009; McDonald et al., 1972). Our data suggest the presence of a narrow hyperpolarization range that specifies the V_mem signal for eye induction (Fig. 5C). Cells that are normally more strongly hyperpolarized than the eye-specific V_mem values can be driven into this range by depolarizing channels, whereas cells that are normally depolarized will be pushed into this range by hyperpolarizing channels. Accordingly, both depolarizing (dominant-negative K_ATP) and hyperpolarizing (Kv1.5) channels are capable of inducing ectopic eyes. The consequences of changes in V_mem (by altering [Na⁺]_ex) in Nav1.5-injected embryos also support the induction of ectopic eye tissues when cells are moved into a narrow range of V_mem. Together, these data suggest a specific range of V_mem values for eye induction — changes in V_mem to levels above or below this target range do not result in eye induction. Also in line with this model, depolarization of endogenous eye field cells represses or disrupts eye formation by driving the V_mem of these cells out of the ‘normal’ permissive range. The narrow range of eye-compatible V_mem values is in line with our observation that only a small subset of injected cells within any tadpole form ectopic eye tissues.

In conclusion, we have characterized a V_mem-mediated signal that is necessary for patterning of Pax6 and Rx1 expression during normal eye development. This V_mem signal is also sufficient for induction of ectopic EFTF expression and eye formation. V_mem-induced ectopic eyes are formed in regions far outside the anterior neural tissues, which were conventionally thought to be the only region competent to form eyes. The discovery of bioelectrical signals that initiate the development of an organ as complex as the eye paves the way to important future work on the developmental and evolutionary roles of ion channel- and pump-derived gradients as mediators of patterning during embryogenesis. Moreover, the ability of changes in V_mem to initiate the formation of a complete vertebrate organ suggests a potential class of novel strategies for regenerative medicine and repair of birth defects (Levin, 2009).

We thank Punita Koustubhan, Amber Currier and Claire Stevenson for Xenopus husbandry and general laboratory assistance; Daryl Davies and Miriam Fine for the GlyCl expression construct; Franz Oswald for EosFP, Sally Moody for Xenopus Shh and Manuel Kukuljan for Nav1.5 constructs; Ali Hemmati-Brivanlou for DNPax6 and the Pax6 antisense probe; Florian Lang for Kv1.5 plasmid; Erik Jorgensen for EXP1 plasmid; Michael Zuber for Rx1 antisense probe; Valerie Schneider for Otx2 antisense probe; and Dany S. Adams and Brook Chernet for microscopy assistance. The XAP2 and anti-Islet-1 antibodies developed by D. S. Sakaguchi and W. A. Harris and by T. M. Jessell and S. Brenner-Morton, respectively, were obtained from the Developmental Studies Hybridoma Bank (DSHB) developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biology, Iowa City, IA 52242, USA.