HCN2 Channel-Induced Rescue of Brain Teratogenesis via Local and Long-Range Bioelectric Repair

Embryonic exposure to the teratogen nicotine results in brain defects, by disrupting endogenous spatial pre patterns necessary for normal brain size and patterning. Extending prior work in Xenopus laevis that showed that misexpression of ion channels can rescue morphogenesis, we demonstrate and characterize a novel aspect of developmental bioelectricity: channel-dependent repair signals propagate long-range across the embryo. We show that distal HCN2 channel misexpression and distal transplants of HCN2-expressing tissue, non-cell-autonomously reverse profound defects, rescuing brain anatomy, gene expression, and learning. Moreover, such rescue can be induced by small-molecule HCN2 channel activators, even with delayed treatment initiation. We present a simple, versatile computational model of bioelectrical signaling upstream of key patterning genes such as OTX2 and XBF1, which predicts long-range repair induced by ion channel activity, and experimentally validate the predictions of this model. Our results and quantitative model identify a powerful morphogenetic control mechanism that could be targeted by future regenerative medicine exploiting ion channel modulating drugs approved for human use.

Keywords: ion channel, bioelectric, teratogen, nicotine, non-local, long-range, regenerative medicine

Brain development is a paradigm case of complex organogenesis, requiring precise control of neural progenitor cell behaviors (Stanger, 2008; Joseph and Hermanson, 2010). The dorsal ectoderm differentiates into the neural plate, with the anterior region forming the brain under the influence of positive and negative effectors (De Robertis and Kuroda, 2004; Stern, 2005). Physical forces (Thompson, 1942; Stanger, 2008), gene regulatory networks (Zhao et al., 2011), and bioelectric signals (Pai et al., 2015a,b; Smith et al., 2018) regulate this complex process.

The membrane voltage potential is a fundamental property of all cells. Endogenous ion fluxes and membrane voltage patterns in embryonic and somatic cells regulate cell behavior and organ-level patterning across eukaryotes (Nuccitelli, 2003; Bates, 2015; Humphries et al., 2017; Levin and Martyniuk, 2018). They also play important roles in directing large scale growth and form in vivo, including tissue/organ regeneration (Tseng and Levin, 2012; Beane et al., 2013; Perathoner et al., 2014), left-right patterning (Levin et al., 2002; Aw et al., 2010; Pai et al., 2017), craniofacial morphogenesis (Adams et al., 2016; Belus et al., 2018), heart and muscle patterning (Lobikin et al., 2015; Pitcairn et al., 2017), and CNS patterning (Ribera, 1999; Lange et al., 2011; Aprea and Calegari, 2012; Pai et al., 2012a,b, 2015b; Avila et al., 2013; Sequerra et al., 2018; Smith et al., 2018).

Endogenous spatiotemporal patterns of cellular resting potentials regulate neural tissue induction and neural progenitor apoptosis and proliferation in the developing brain, as revealed by loss-of-function experiments in model systems (Pai et al., 2015a,b) and by naturally-occurring genetic channelopathies (Smith et al., 2018). Moreover, establishing correct spatial membrane voltage pre patterns via misexpression of specific channels to control membrane voltage can rescue neural patterning defects caused by aberrant Notch signaling, mechanical damage, or exposure to neuroteratogens, and normalize the expression of canonical brain patterning genes (Pai et al., 2015b, 2018; Herrera-Rincon et al., 2017). Changes in membrane voltage are transduced via gap-junctional signaling and calcium dynamics to control the expression of crucial transcription factors necessary for brain patterning (Pai et al., 2015b).

Hyperpolarization-activated Cyclic nucleotide-gated (HCN) channels are voltage-gated channels, with a threshold voltage that is affected by metabolic state (Biel et al., 2009; Wahl-Schott and Biel, 2009; Benarroch, 2013). The HCN channel family includes variants HCN1-4, all of which are known to be expressed in embryonic cells and early Xenopus embryos (Yasui et al., 2001; Qu et al., 2008; Vicente-Steijn et al., 2011; Spater et al., 2013; Session et al., 2016; Pai et al., 2017, 2018; Pitcairn et al., 2017). However, their roles in embryonic development and possible utility as therapeutic targets are largely unexplored (Postea and Biel, 2011; Benarroch, 2013).

Misexpression of HCN2 channels in Xenopus embryos restored the bioelectric prepattern and brain patterning disrupted by nicotine (Pai et al., 2018). We characterized in detail how nicotine’s teratogenic effects occur via the disruption of endogenous bioelectric patterns and presented a quantitative model of the physiology of the affected cells and their rescue under HCN2 dynamics (Pai et al., 2018). Here, we use this established neuroteratogen (Slotkin et al., 2005; Huizink and Mulder, 2006; Slotkin, 2008; Velazquez-Ulloa, 2017) to ask important new questions about the spatial properties of non-cell-autonomous bioelectric controls of brain patterning and test the hypothesis that repair can be induced without the need for gene therapy. Remarkably, we found that HCN2-activated repair can be triggered at considerable distance from the brain. We constructed and tested a computational model which explains how repair-inducing bioelectric states can propagate across tissues. Lastly, we found that FDA-approved small-molecule drugs targeting ion channels can induce brain repair, counteracting teratogenic exposure without the need for transgenes. Together, the predictive computational model of bioelectric signal propagation across tissues and the functional data identify a novel long-range regulator of brain organogenesis and suggest molecular bioelectric strategies as interventions to induce repair in a roadmap for regenerative applications targeting birth defects (Mathews and Levin, 2018; McLaughlin and Levin, 2018).

Xenopus laevis embryos were fertilized in vitro according to standard protocols 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; Sive et al., 2000). Xenopus embryos of both sexes were housed at 14–18°C and staged according to Nieuwkoop and Faber (1958). For animals used in behavior trials, individuals of both sexes were raised under 12 h:12 h light:dark cycle at a temperature of 16°C at no more than 30 individuals per 100 × 25 mm Petri dish. After stage 46, tadpoles were fed twice per day on standard sera micron powdered food until behavioral testing. All experiments were approved by the Tufts University Animal Research Committee (M2017-53) following the guide for the care and use of laboratory animals.

Capped synthetic mRNAs generated using the mMessage mMachine kit (Ambion) were dissolved in nuclease-free water and injected into embryos immersed in 3% Ficoll using standard methods (Sive et al., 2000). Each injection delivered between 0.5–1 ng of mRNA (per blastomere) into the embryos at the indicated stages into the middle of a cell in the animal hemisphere. Hcn2-WT and Hcn2-DN were mammalian (mouse) hyperpolarization-activated cyclic nucleotide-gated channel 2, modified as detailed in Pai et al. (2018).

Embryos were incubated in chemicals dissolved in 0.1× MMR during the stages of interest as indicated in the respective experiments followed by several washes with 0.1× MMR. Embryos were exposed to 0.1 mg/ml nicotine (sigma—N3876) from stage 10–35 unless otherwise specified (targeting neurodevelopment while allowing normal cleavage and gastrulation). Embryos were exposed to 200 μM lamotrigine (tocris—2289) and 175 μM gabapentin (tocris—0806) at the specified stages. The dose of lamotrigine and gabapentin drugs were titrated to a level at which no general toxicity was observed, and the survival rate of embryos was similar to untreated controls (Supplementary Figure S1).

Tadpoles used for morphometric analysis were imaged with a Nikon SMZ1500 microscope with a Retiga 2000R camera and Q-capture imaging software. Landmarks were chosen and annotated using ImageJ software (Schneider et al., 2012) based on biological relevance and reproducibility across tadpoles with the varying brain and head morphologies: (1) anterior tip of the head; (2) space between the olfactory bulbs at the beginning of the forebrain; (3) center of the transition line between forebrain and midbrain; (4) center of the transition line between midbrain and hindbrain; (5) transition point between the hindbrain and spinal cord, and (6); and (7) lateral-most points of the head at the eyes. MorphoJ (Klingenberg, 2011) was used for Canonical Variate Analysis to quantify and graphically represent changes in brain regions relative to head shape. MorphoJ was also used to calculate Procrustes distances and perform statistical analysis. Our analysis was conservative, as nicotine-treated embryos that had such severe defects that we could not identify the landmarks were excluded.

Tadpoles were fed directly before trials and provided food ad-lib during the trials. All behavioral trials involved a color-based associative learning assay, performed as previously described, using an automated behavior analysis platform (Blackiston et al., 2010a; Blackiston and Levin, 2012). Briefly, individual tadpoles were introduced to the chamber which was half-illuminated with red light and half blue light to probe their innate color preferences for 30 min in the absence of any punishment. In the learning acquisition phase, each tadpole received a 1.2 mA shock whenever it occupied the red area. This duration was 20 min, with colors in the chamber being inverted every 5 min to ensure that immobility is not scored as a success. Then, tadpoles were given a 90 min rest in which the entire chamber is illuminated with blue light and no shock is delivered. Learning was scored by giving the tadpoles a choice between red and blue light for 5 min with no punishment. The entire block of acquisition-rest-probe was repeated six times. A tadpole was determined to have learned if their preference for a red light was below 40% as averaged across the final three probe sessions of the experiment.

Xenopus embryos were collected and fixed in MEMFA (1 h at room temperature; Sive et al., 2000), and in situ hybridization was performed as previously described (Sive et al., 2000). in situ antisense probes were generated in vitro from linearized templates using a DIG (Digoxigenin)-labeling mix (Roche). Chromogenic reaction times were optimized for signal-to-background ratio. Probes used were otx2 (Pannese et al., 1995) and xbf1 (Eagleson and Theisen, 2008). A stock solution of each probe was used for all experimental groups, which were incubated together for the same amount of reaction time, to avoid any variability due to probe concentration or length of reaction time.

CC2-DMPE and DiBAC4(3) ratiometric voltage reporter dyes were obtained from Invitrogen and used as per the standard protocol (Adams and Levin, 2013). Briefly, CC2-DMPE stock (5 mM) was dissolved 1:1,000 in 0.1× MMR and the embryos were incubated in the dark in this solution for at least 1 h followed by five washes with 0.1× MMR. DiBAC4(3) stock (1.9 mM) was dissolved 1:1,000 in 0.1× MMR and the CC2-DMPE-stained embryos were then incubated in the dark in this solution for at least 30 min washed thoroughly in 0.1× MMR followed by visualization under the microscope. An Olympus BX-61 microscope equipped with a Hamamatsu ORCA AG CCD camera and controlled by MetaMorph software (Molecular Devices), was used to collect images. ImageJ was used to quantify the fluorescence intensities of the CC2-DMPE: DiBAC signal along the red line across the image as indicated in the illustrations in Figures 4G, 9E). Fluorescence values at each point along this line were used to plot graphs.

Transplants were performed as previously described (Viczian and Zuber, 2010). Animal caps were excised from tdTomato-expressing stage 9 donors in 0.75× MMR and were allowed to heal overnight in 0.75× MMR at 14°C. Sibling uninjected host embryos were treated with nicotine (0.1 ml/ml) beginning stage 10 and incubated overnight at 14°C. At stage 13, the vitelline membranes of nicotine-exposed recipient embryos were removed in 0.1× MMR+nicotine (0.1 mg/ml). A small square piece of tissue was removed from the flank of the recipient, taking care that the excision was only superficial, not exposing the body cavity of the embryo to the external medium. The donor animal cap was cut in half and placed into the recipient at the excision site. The recipient embryos were then carefully placed in agarose molds submerged in 0.1× MMR with nicotine (0.1 mg/ml). The molds hold the graft in place and allow the embryo to heal. Embryos were allowed to heal for 4–8 h, then washed and placed in fresh 0.1× MMR with nicotine at 14°C overnight. Embryos were then cultured at 18°C until analyzed.

Animals injected with β-galactosidase (β-gal) mRNA were fixed (30 min in MEMFA at RT) at the relevant stages, washed twice in PBS with 2 mM MgCl2, and stained with X-gal (Roche Applied Sciences, Indianapolis, IN, USA) staining solution at 37°C for 3 h. Tadpoles were then rinsed three times in PBS followed by dehydration through sequential in incubation in 25%, 50%, 75%, and 100% methanol. Tadpoles were then incubated in 30% H2O2 in methanol overnight for bleaching. Tadpoles were then washed in 100% methanol sequentially rehydrated to PBS and imaged.

Agarose sectioning was performed as previously described (Blackiston et al., 2010b). Briefly, stage 45 tadpoles were fixed for 2 days in MEMFA (Sive et al., 2000), washed in PBT, dehydrated in methanol, rehydrated back in PBT, embedded in 4% agarose and sectioned at 100 μm using a Leica vibratome, followed by hematoxylin and eosin staining.

Statistical analyses were performed using GraphPad Prism7. At least three independent experiments were conducted with N > 50 embryos for each treatment group, with embryos collected from multiple animals across independent clutches. Data were analyzed by t-test (for two groups) or ANOVA (for more than two groups, with Tukey’s multiple comparison test) as indicated with each experiment.

All data generated or analyzed during this study are included in this article and Supplementary Material is available from the corresponding author upon request.

To test whether HCN2 expression in local or non-local tissues can rescue nicotine-induced brain defects, we injected Hcn2-WT mRNA into precursors of the brain [dorsal blastomeres (Moody, 2018)] or ventral tissues (ventral blastomeres) at four-cell stage followed by exposure of Xenopus embryos to nicotine (Pai et al., 2018). Untreated, uninjected embryos and untreated, Hcn2-DN (Dominant Negative) mRNA-injected embryos were used as controls. We injected mRNA at the four-cell stage, using the fate map distinguishing dorsal from ventral precursors to establish the first proof-of-principle of long-range activity. To optimize the likelihood of efficient rescue, we sought to target either the whole neural plate region or equivalently a large portion of the non-neural region to achieve the maximum demonstration of long-range repair. Co-injection of lineage tracer β-galactosidase mRNA validated our targeting of neural and non-neural regions (Figures 1A–D). Brain morphology was evaluated at stage 45 (Figures 1E–Q). Tadpoles from control embryos exhibited correctly patterned (Pratt and Khakhalin, 2013) brain tissue (Figures 1E,K,N,Q). Nicotine exposure caused an increased incidence of abnormal brain morphology (59% tadpoles with brain defects) in comparison to controls (7% tadpoles with brain defects; Figures 1F,L,O,Q). Phenotypes included the absence of nostrils, absence of forebrain, absence of forebrain and midbrain, truncated and mispatterned nostrils, forebrain, and midbrain. Eye development was also affected, resulting in absent or incompletely formed eyes (Pai et al., 2012a). The average distribution of these phenotypes in nicotine-exposed tadpoles is listed in Supplementary Table S1.

In contrast, embryos with dorsal blastomere (neural) Hcn2-WT mRNA injection showed significant rescue or protection of nicotine-induced abnormal brain morphology (27% tadpoles with brain defects), with a large scale brain patterning similar to that of controls (Figures 1H,Q). Embryos with only dorsal Hcn2-WT mRNA injection (no nicotine exposure) were not significantly different from the controls (5% tadpoles with brain defects; Figures 1G,Q). Thus, overexpression of HCN2-WT channels in neural tissue is able to significantly counteract the teratogenic effects of nicotine exposure on brain patterning.

Remarkably, embryos with Hcn2-WT mRNA injected into ventral blastomeres (non-neural precursors, distant from the brain) also showed a near-complete rescue/protection of brain morphology (16% tadpoles with brain defects) with large-scale brain patterning similar to controls (Figures 1J,M,P,Q). Embryos with only ventral blastomere Hcn2-WT mRNA injection (no nicotine exposure) were not significantly different from the controls (6% tadpoles with brain defects; Figures 1I,Q), confirming that HCN2-WT expression in ventral tissues does not harm normal development. Thus, overexpression of HCN2-WT channels in non-CNS tissues can significantly counteract the teratogenic effects of nicotine exposure on brain patterning.

To quantify brain morphology, we used geometric morphometrics at stage 45 (Webster and Sheets, 2010; Figure 2). Landmarks were chosen based on transition points between different brain regions and lateral and anterior outermost points of the head (Figure 2A) and recorded for tadpoles for each of the controls, nicotine, nicotine + dorsal (local/within neural tissue) Hcn2-WT mRNA, and nicotine + ventral (non-local/within non-neural tissue) Hcn2-WT mRNA. Canonical variate analyses with Procrustes distances between each of the groups quantitatively revealed changes in the length of the brain relative to head shape between treatment conditions (Figure 2B) and the direction of changes in shape (Figures 2C,D). Control and nicotine + dorsal-HCN2-WT embryos were not significantly different in shape, but both were significantly different from nicotine-exposed embryos (Figure 2B). Also, control and nicotine + ventral-HCN2-WT embryos were not significantly different in shape (Figure 2B), but both were significantly different from nicotine-exposed embryos. Thus, both dorsal (local, within neural tissue) or ventral (non-local, in non-neural tissue) Hcn2-WT mRNA microinjections protect or rescue brain morphology to the wild-type state despite nicotine exposure.

To gain insight into the molecular mechanisms and determine whether local (neural) and distant (ventral) HCN2 channel interventions affect the known brain patterning transcriptional regulators, we analyzed the expression of canonical factors otx2 (forebrain and midbrain; Acampora et al., 1995) and xbf1 (forebrain; Bourguignon et al., 1998). Embryos were analyzed at stage 25 by in situ hybridization (Figure 3). Nicotine-exposed embryos showed significantly reduced expression of otx2 (both in area and intensity; Figure 3C) and reduced or mispatterned xbf1 expression (Figure 3D) compared to controls (uninjected and untreated embryos; Figures 3I,J). Crucially, both dorsal blastomere (local/neural) and ventral blastomere (distant) Hcn2-WT mRNA-injected embryos showed normal otx2 and xbf1 expression (Figures 3E–J). Thus, both neural and distal HCN2 channel misexpression induces correction of otx2 and xbf1 expression despite nicotine exposure, suggesting that the action of the bioelectric repair functions upstream of these early brain patterning genes.

To elucidate the biophysical mechanism by which HCN2 overexpression induced corrective changes in brain gene expression, we characterized the effect of nicotine and Hcn2 microinjections on the endogenous bioelectric prepattern (distribution of resting potentials) known to be a critical regulator of the neural transcription factors and brain patterning (Pai et al., 2015b). We evaluated embryos between stages 15–17 using a combination of in vivo imaging, voltage reporter dyes (Adams and Levin, 2012), and whole-cell membrane voltage recordings (Pai et al., 2015b; Figure 4). Whole-cell electrophysiological recordings of membrane voltage from neural plate cells and flanking ectodermal cells were used as calibration points for the voltage reporter dye images, and the fluorescence intensities were analyzed against these calibration points to approximate membrane voltages at different points within the developing embryos (Figure 4).

Nicotine-treated embryos showed a significantly depolarized neural resting membrane potential (depolarized by ~30 mV) in comparison to controls (uninjected and untreated embryos; Figures 4A,B,G,H). Hcn2-WT mRNA injected embryos (either dorsal or ventral microinjections) were not significantly different from the controls (Figures 4A,C,E,G,H). Interestingly, nicotine-exposed embryos that were injected with Hcn2-WT mRNA, either dorsally or ventrally, both showed restored neural plate hyperpolarization in comparison to nicotine-only treated embryos (Figures 4A,B,D,F,G,H). Thus, both local (neural) and distant (ventral) HCN2 channel expression restores the correct neural membrane voltage prepattern despite the presence of nicotine. We conclude that bioelectric changes induced in a distant region of the embryo can affect the bioelectric prepattern of the nascent brain and that ventral HCN2 injections are inducing repair by correcting the nicotine-mediated disruption of the neural plate bioelectric prepattern.

How does altering the electrophysiological properties of distant cells (HCN2 misexpression in ventral regions) correct the membrane voltage prepattern in the anterior neural plate? Cell-cell connections across tissues can drive highly complex, non-intuitive, spatiotemporal changes in membrane voltage patterns that are best understood via simulations (Cervera et al., 2015, 2016a,b, 2018; Pietak and Levin, 2016, 2017; Brodsky and Levin, 2018). Hence, we formulated a minimal computational physiological model of spatial membrane voltage patterns to understand long-range rescue/protection by distant ion channel misexpression (Supplementary Material and Figure 5). This model incorporated both, past results (Pai et al., 2015a,b, 2018) and the novel data reported here (Figures 1–4). This model focused specifically on the membrane voltage dynamics, not restricting absolute tissue size or timescale: the objective was specifically to understand non-cell-autonomous membrane voltage propagation. We then extracted from this model specific predictions for experimental membrane voltage perturbations that should or should not be successful in reversing or preventing nicotine-induced brain defects.

The key experimental results on which this model is formulated are as follows. Beginning neurulation the neural plate cells are significantly hyperpolarized in comparison to the surrounding depolarized ectodermal cells (Pai et al., 2015a,b, 2018). We found that it is not the individual ion fluxes (such as Na+ or K+ flux) but the membrane voltage of these cells that regulate embryonic patterning (Pai et al., 2012a, 2015b). In the case of brain patterning, we found that it is not the absolute values of membrane voltage but the contrast/difference in membrane voltage pattern between neural plate and ectoderm that is crucial for proper brain patterning (Pai et al., 2015b). Eliminating this contrast/difference results in serious brain defects irrespective of the absolute membrane voltage of any given tissue. These membrane voltage patterns act locally and over long-range to control key cell behaviors such as proliferation and apoptosis during brain patterning (Pai et al., 2015a,b). Also, these membrane voltage signals are transduced through GJs and regulate canonical gene regulatory networks and biochemical signals orchestrating brain patterning (Pai et al., 2015b). Lastly, these membrane voltage patterns are altered during many developmental brain deformities and these deformities can be corrected by precise modulation of these membrane voltage patterns (Pai et al., 2015b, 2018).

In our model (Supplementary Material), complex developmental membrane voltage patterning is depicted at the multicellular level in cell ensembles. We used an equivalent circuit strategy, sufficient to qualitatively reproduce the observed behavior of the importance of contrast (difference) of membrane voltage patterns during brain patterning as described above and shown in the results of this manuscript (Pai et al., 2015b, 2018; and Figures 1–4). Since individual ion fluxes are important due to their contribution to overall Vmem, a single cell’s membrane voltage is represented as controlled by two counteracting voltage-gated ion channel aggregates of maximum conductances Gpol and Gdep. Gpol is regarded as an effective sum of channels that promote the polarized (pol) cell state while Gdep is regarded as an effective sum of channels that promote the depolarized (dep) state. A key characteristic of our approach is that the single-cell state can be modulated at the ensemble level because of the coupling of a given cell with the neighboring cells (Cervera et al., 2016b, 2018, 2019). This coupling is allowed by the intercellular gap junction conductance Gij that permits the transfer of ionic currents and signaling molecules between two adjacent cells (Cervera et al., 2016b, 2018, 2019). Such gap-junction connections between embryonic cells are crucial for proper embryonic tissue patterning (Spray et al., 1981; Warner, 1985; Pai et al., 2015b; Mathews and Levin, 2017).

In the multicellular ensemble, the membrane voltage states are modulated by: (1) the relative values of the single-cell channel conductances Gpol and Gdep promoting the polarized and depolarized cell states, respectively; (2) the relative values of the maximum intercellular GJ conductance G0 and the single-cell conductances of the connected cells; (3) the relative size and proximity of the polarized dorsal region (representing the polarized neural tube) and the polarized tissue patch (representing the HCN2-expressing non-neural tissue) that is created within the depolarized non-neural region. A limitation of our approach is that external actions should also produce additional diffusion-reaction processes characterized by experimental transient times much higher than those obtained with purely electrical mechanisms, as previously explained (Cervera et al., 2016b, 2018). The results of Figure 5 concern only the steady-state system states. Besides, the particular current-voltage curves of all ion channels and gap-junctions used are also shown in the respective Supplementary Figures S1–S3 of Supplementary Material.