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

Our goal was not to construct a maximally-realistic simulation of the Xenopus neural tube but to identify the key drivers of this developmental physiology and understand whether and how HCN channels can underlie the propagation of resting potential changes over long distances in developing tissue. The concerted action of the HCN2 channels is simulated by an effective inward-rectifying conductance which opens to cations at hyperpolarized potentials (Biel et al., 2009; Benarroch, 2013). For strongly depolarized cells, however, this channel closes giving an almost zero residual conductance. However, given the ionic gradients in Xenopus cells (high intracellular Na+ and K+ (Gillespie, 1983) in comparison to external media), the resulting effective channel acts to polarize rather than to depolarize the Xenopus cells (Pai et al., 2018). We qualitatively described the membrane voltage states of the cells in two generic dorsal (neural) and ventral (non-neural) regions within this multicellular ensemble.

The control case (Figure 5A) represents the endogenous physiological condition. The neural region is marked by the circle and the rest represents the non-neural region of the embryo. Differences in the baseline levels of the Gpol and Gdep conductances of the cells within these regions result in a hyperpolarized neural plate region (−55 mV) in comparison to the rest of the embryo (−5 mV; Supplementary Material). This pattern is consistent with the observed membrane voltage patterns in Xenopus embryos (Figure 4 and in Pai et al., 2015b). Figure 5B corresponds to the depolarization of the neural plate due to nicotine-exposure. The depolarizing effect of nicotine is simulated by lowering the conductance Gpol in the neural region, which prevents this neural region from reaching the polarized state of Figure 5A (Pai et al., 2015b; Supplementary Material). The nicotine-exposed multicellular ensemble is then used to test the effect of externally introducing the HCN2 channel in the patch of the non-neural region.

The HCN2 channel misexpression in the non-neural tissue was simulated by increasing the context-specific conductance Gpol in the small patch of tissue (Figures 5C–F). Remarkably, the ectopic hyperpolarized patch could repolarize the neural region at a distance (Figure 5C). However, this neural repolarization did not occur if the polarized patch was too small (Figures 5C,D) or was placed too far away (Figures 5E,F). Thus, three key predictions arise from the simulations: (1) a patch of HCN2-expressing cells in the non-neural tissue should be able to restore the membrane voltage patterns in the neural plate of nicotine-exposed Xenopus embryos; (2) if the patch of HCN2-expressing cells is small, it should fail to rescue membrane voltage patterns in this neural plate of nicotine-exposed Xenopus embryos; and (3) the patch of HCN2-expressing cells needs to be sufficiently close to the neural plate to rescue the membrane voltage patterns in the nicotine-exposed Xenopus embryos. This model reveals the minimal aspects necessary for long-range bioelectric influence propagation; it explains the ability of HCN2 expression in remote regions to influence the membrane voltage of anterior neural tissues (Figure 4).

To test the model’s predictions and to determine whether HCN2’s repairing activity could operate if it was not present during the very early stages of development, we developed a tissue transplant assay (Figure 6A). We microinjected either Hcn2-WT mRNA or Hcn2-DN (dominant-negative) mRNA, with a fluorescent lineage tracer Tomato mRNA (Figure 6A, red), into donor embryos at the one-cell stage. The animal caps from donor embryos were excised and transplanted into the flank of age-matched (stage 13) nicotine-exposed host embryos (see methods for details). The embryos were allowed to heal and develop to stage 45, and brain morphology was evaluated (Figures 6B–J). This transplant assay allowed us to cleanly test the effects of an ectopic patch of HCN2-expressing cells at a well-defined point in developmental time and location.

Control (uninjected, untreated, un-transplanted) tadpoles had correctly patterned brain tissue (Pratt and Khakhalin, 2013), with well-formed nostrils, olfactory bulbs/forebrain, midbrain, and hindbrain (Figure 6B). Nicotine exposure induced a significant increase in the incidence of brain defects (58% tadpoles with brain defects) in comparison to controls (5% tadpoles with brain defects; Figures 6C,H). Strikingly, nicotine-exposed embryos with Hcn2-WT-expressing transplants (appearing in the somatic ventral region; Figure 6E) showed a significant rescue of brain defects (22% tadpoles with brain defects; Figures 6D,E,H). In contrast, the nicotine-exposed embryos with Hcn2-DN- or tomato-only expressing transplants caused no rescue (65% and 59% tadpoles with brain defects, respectively; Figures 6F–H). We conclude that HCN2-WT channel-expressing transplanted donor tissue induces repair of brain morphology defects at a distance and that this rescue is still effective when begun at stage 13. This result also validates the model’s prediction that HCN2-expressing cells in non-neural tissues should be able to rescue nicotine-induced abnormal brain morphology.

Next, we sorted the nicotine-exposed tadpoles that received Hcn2-WT+tomato mRNA transplants based on the size of the red fluorescent tomato patch and assessed the degree of rescue. Nicotine-exposed embryos that received large transplants showed a significantly better rescue (27% tadpoles with brain defects) than the nicotine-exposed embryos that received small transplants (83% tadpoles with brain defects; Figure 6I). This result validates the theoretical prediction that larger patches of HCN2-expressing cells in non-neural tissues should rescue nicotine-induced brain defects better than smaller patches of HCN2-expressing cells.

Analogously, we sorted the nicotine-exposed tadpoles that received Hcn2-WT +tomato mRNA tissue transplants based on the location of transplant (either in the fin tissue far away from CNS or in tissues close to CNS). Interestingly, nicotine-exposed embryos that received transplants near the CNS showed better rescue (0% tadpoles with brain defects) than those that received transplants in the fin region (81% tadpoles with brain defects; Figure 6J). This result agrees with the prediction that a patch of HCN2-expressing cells closer to the neural tissue should rescue nicotine-induced brain defects better than the patch of HCN2-expressing cells that are farther away from the neural tissue.

The induction of improvements in brain morphogenesis by transplants occurring well after teratogen exposure revealed that the effect is not simply prevention but also repair. To confirm this and establish a strategy that does not require tissue transplants, we next turned to human-approved ion channel activating drugs. To discover whether small molecule activation of endogenous HCN2 channels would be sufficient to rescue nicotine-induced brain defects, we used two well-known HCN channel activators: lamotrigine (LT) and gabapentin (GP; Poolos et al., 2002; Surges et al., 2003; Postea and Biel, 2011; Brennan et al., 2016). While these drugs may have additional targets besides HCN, we used these reagents not to prove the involvement of HCN2 (which we have already shown using highly specific molecular-genetic reagents; Figures 1, 6; Pai et al., 2018), but to determine whether a small molecule agent that activates native HCN2 channels at later stages is sufficient to overcome teratogen exposure. Much higher concentrations of these drugs than the ones used in this study could affect development adversely. However, we found and used a non-toxic dose of these drugs where the main effect is an improvement of developmental outcomes.

Xenopus embryos were exposed to nicotine (stages 10–35), and then either concurrently (stage 10–35) or after a delay (stage 25–35), treated with lamotrigine or gabapentin. Untreated embryos and only lamotrigine- or gabapentin-treated embryos served as respective controls. The embryos were allowed to develop to stage 45, and brain morphology was evaluated (Figure 7). Control tadpoles had correctly patterned brain tissue, with well-formed nostrils, olfactory bulbs/forebrain, midbrain, and hindbrain (Figure 7A), while nicotine exposure caused brain defects (67% and 68% tadpoles with brain defects, respectively) in comparison to controls (5% and 8% tadpoles with brain defects, respectively; Figures 7B,G,H). Concurrent lamotrigine or gabapentin treatment showed a significant rescue (15% and 23% tadpoles with brain defects, respectively) of nicotine-induced defects (Figures 7D–H). Similarly, delayed treatment with lamotrigine and gabapentin showed significant rescue (only 10% and 16% tadpoles with brain defects, respectively) of the brain defects (Figures 7G,H). Lamotrigine- and gabapentin-only treated tadpoles showed no difference in brain morphology (2% and 11% tadpoles with brain defects, respectively) compared to controls (5% and 8% tadpoles with brain defects, respectively; Figures 7A,C,E,G,H), showing that these drugs alone do not impair brain development.

To quantify brain morphology, we used geometric morphometrics (Webster and Sheets, 2010; Figure 8). Landmarks were chosen based on transition points between different brain regions and lateral and anterior outermost points of the head at stage 45 (Figure 8A), and measurements were recorded for N > 10 tadpoles for each of the following conditions: controls, nicotine, nicotine+lamotrigine, and nicotine+gabapentin. Canonical variate analyses (Figure 8B) were run on the data set, with Procrustes distances measured between each of the groups quantitatively revealing changes in the length of the brain relative to the head shape across treatment conditions (Figures 8B–D). Control, nicotine+lamotrigine, and nicotine+gabapentin embryos were not significantly different in shape (Figure 8B, orange, blue, and cyan ellipses), but were significantly different from nicotine-exposed embryos (magenta ellipse). Thus, quantitative shape analysis reveals that the treatment of nicotine-exposed embryos with lamotrigine or gabapentin well after initial teratogen exposure rescues nicotine-induced brain defects to near wild-type.

To determine how lamotrigine or gabapentin rescue brain defects, we next tested the hypothesis that they restore the endogenous neural plate membrane voltage prepattern required for brain patterning (Pai et al., 2015b). by evaluating resting potential distributions in embryos treated from stage 10 (Figure 9). Nicotine-treated embryos showed a significantly depolarized potential (by ~28 mV) vs. controls (Figures 9A,B,E,F). In contrast, nicotine-exposed embryos treated with lamotrigine or gabapentin showed a significantly hyperpolarized neural plate (by ~33 mV) compared to controls (Figures 9C–F). Thus, both lamotrigine and gabapentin hyperpolarize the neural plate membrane voltage prepattern despite the presence of nicotine, revealing that these compounds repair brain morphology in the same way as does HCN2 overexpression.

We next assayed the prospects for the functional rescue of behavioral performance. Using an automated behavior analysis platform (Blackiston et al., 2010a; Figure 10A), a tadpole can be trained to avoid a moving red light, enabling quantification of cognitive performance (Figure 10B). A tadpole was classified as “having learned” if its preference for red light dropped below 40% of time spent in red light, averaged across the final three probe sessions of the experiment. This approach has been successfully used to quantitatively evaluate brain function and learning performance under surgical, pharmacological, and genetic manipulations (Blackiston and Levin, 2013a,b; Blackiston et al., 2017).

Forty percent of nicotine-exposed animals showed no obvious motor problems or movement abnormalities, and only these were tested in the learning assay. Morphologically normal nicotine-exposed tadpoles failed to learn to avoid red light with overall ~60% of the time spent in red light (Figure 10B). In comparison, control tadpoles learned to avoid red light with overall only ~24% of time spent in red light (Figure 10B). However, nicotine-exposed embryos receiving dorsal (local/neural) or ventral (distant/non-neural) Hcn2-WT mRNA at the four-cell stage showed significant improvement in learning ability with overall only ~32% and ~27%, respectively, of time spent in red light (Figure 10B). Similarly, nicotine-exposed embryos treated with lamotrigine (stage 10–35) or gabapentin (stage 10–35) also showed significant improvement in learning ability with overall only ~28% and ~24%, respectively, of time spent in red light (Figure 10B). Thus, both local (neural) and distant (non-neural) overexpression of HCN2 channels, as well as drug-based activation of native channels, restore cognitive learning abilities in nicotine-exposed tadpoles.

A specific spatial distribution of resting potential is an endogenous prepattern necessary for normal brain development; moreover, externally enforcing this pattern restores brain pattern despite a range of genetic, chemical, and mechanical insults (Pai et al., 2015a,b, 2018; Herrera-Rincon et al., 2017). Changes in bioelectric state are transduced by gap-junctions and calcium flux and regulate key brain patterning transcription factors (Pai et al., 2015a,b). Previously we showed that the classic neuroteratogen nicotine (Slotkin et al., 2005; Huizink and Mulder, 2006; Slotkin, 2008; Velazquez-Ulloa, 2017) induced brain morphology defects by disrupting these bioelectric prepatterns and presented a detailed model of brain cell physiology and downstream gene expression changes (Pai et al., 2018). Modulating these bioelectric patterns using HCN2 channel overexpression revealed the predicted amelioration of nicotine-induced brain defects. Overexpression of the HCN2 channel itself does not cause any phenotypic defects but instead corrects mild brain deformities seen in control embryos (Pai et al., 2018). This effect of HCN2 is mainly due to their context-specific action, where the channels open only under hyperpolarizing conditions (−40 mV to −70 mV; neural plate) and remain closed in depolarized cells (non-neural ectoderm). Thus, HCN2 channels magnify the voltage differential between two cell fields (neural and non-neural). In areas where these differences are already strong (normal embryos) they have little effect because compartment boundaries are already set and the apposition of their distinct Vmem already activates downstream mechanisms. However, in regions where boundaries have become significantly weakened (e.g., because of teratogens like nicotine), HCN2 increases the differential to normal levels, correcting morphogenesis (Pai et al., 2018).

Here, we sought additional insight into endogenous bioelectric controls of brain development and established a proof-of-principle for regenerative medicine approaches, by tackling two key questions that remained open: (1) can patterning be controlled by bioelectric states of distant tissues; and (2) can a repair be induced by targeting native channels instead of over-expressing native HCN2. First, by targeting HCN2 mRNA to dorsal or ventral tissues, followed by marker analysis, physiological profiling, and morphometrics, we found that either local or long-range HCN2 overexpression can restore the normal voltage pattern, the expression of classic neural transcription factors otx2 and xbf1, and brain morphology—all of which are otherwise disrupted by nicotine exposure (Figures 3, 4). Our marker analysis revealed the bioelectric influence to operate upstream of the known regulators of brain size and shape (Otx2, Xbf1), likely using the same transduction mechanisms we previously identified (Pai et al., 2015a,b, 2018). In the future, high-resolution analysis of HCN2 expression in individual blastomeres of 16 and 32 cell embryos could identify the smallest regions that are still sufficient for inducing repair.

How does distant (outside neural plate/non-CNS) HCN2 overexpression in nicotine-exposed embryos restore the membrane voltage pattern? As with many gene-regulatory circuits, membrane voltage control involves many complex (positive and negative) feedback loops, making it difficult to directly predict the time evolution of multicellular tissue (system-level) distributions of resting potentials (Adams and Levin, 2013; Levin and Martyniuk, 2018); thus, computational modeling is helpful to understand the dynamics and possible behaviors of bioelectric tissues (Cervera et al., 2018; Pietak and Levin, 2018). We constructed a simulation to map spatial membrane voltage changes to analyze and understand the observed experimental results (Supplementary Material; Cervera et al., 2016b; Cervera et al., 2018).

The distant HCN2 overexpression was simulated as a patch of polarized cells in the non-neural region of nicotine-exposed embryos. This polarized patch of cells repolarizes the neural region in nicotine-exposed embryos via the known presence of intercellular gap-junction connections between the polarized patch and the neural region (Figure 5). The patch polarization spreads mainly in the direction of the neural region because the bulk of the region ventral to the patch is depolarized and dynamically buffers against the spread of polarization in this region. A patch of sufficient size spreads polarization into the neural region, while small HCN2-expressing polarized patches fail to restore the neural membrane voltage (Figure 5). The plasticity of Xenopus embryonic development allowed us to test the model’s predictions using embryonic transplant experiments (Figure 6). An ectopic patch of HCN2 tissue transplanted onto nicotine-exposed embryos was sufficient to rescue the brain morphology, definitively confirming the non-cell-autonomous nature of the effect as well as other predictions of the model concerning size and distance of the patch. Our model explains the minimal dynamics necessary for the highly surprising finding that introducing a hyperpolarizing region on an embryo enables control of membrane voltage states (and downstream morphogenesis) in a different region.

Certainly, we do not rule out the possibility of additional biochemical, mechanical, or other signals such as exosomal communication (Danilchik and Tumarkin, 2017), electrically coupled nanotubes (Wang et al., 2010), and/or cytonemes (Danilchik et al., 2013; Kornberg and Roy, 2014) that might propagate the effects of distant membrane voltage changes into downstream effects in vivo. Because this simple equivalent circuit model is compatible with a variety of biological contexts (underlying channel profiles), such long-range propagating influences may be sought in other cases, for example in the propagation of bioelectric states during limb injury (Busse et al., 2018) or cancer (Chernet and Levin, 2014; Chernet et al., 2015). Moreover, it provides the minimal necessary condition to recapitulate such tissue propagation signals in synthetic bioengineering contexts (McNamara et al., 2016, 2018). Using the insights of such models, future efforts will focus on developing stimulation protocols to maximize the distance at which reparative effects can be induced in a target region in vivo.

Our data show that distant membrane voltage states can rescue brain patterning as long as they are close to neural tissue and are significantly large. These signals can regulate the transcription of key genes, likely due to small molecules or electric current propagating through gap junctions during this long-range communication. One of the remaining open questions concerns why some cells along with the path act as mere relays of information while other cells interpret this information and act because of their different transcriptional states. This is currently a major area of investigation, using synthetic minimal bioelectric tissues to understand how voltage boundaries form and how voltage differences are transduced to regulate downstream transcriptional responses (Cervera et al., 2018, 2016b; McNamara et al., 2018, 2020).

To explore the possibility that systemic, human-safe bioelectric drugs can be exploited for new strategies in regenerative medicine, we utilized two FDA-approved molecular activators of HCN channels: lamotrigine (Goldsmith et al., 2003; Postea and Biel, 2011) and gabapentin (Surges et al., 2003; Wiffen et al., 2017), both of which are in wide use in patients (Dolk et al., 2016). While these drugs may have additional targets besides HCN, our molecularly-specific experiments with HCN2 mRNA demonstrate that the HCN2 activation activity is sufficient to induce repair. While high doses of these compounds might affect development on their own, we show that it is possible to identify a dose where the main effect is to strengthen membrane voltage boundaries and rescue complex organogenesis. Both lamotrigine and gabapentin successfully rescued brain defects in nicotine-exposed embryos (Figures 7–10). Importantly, we did not have to precisely target the drug exposure to a specific region or tissue of the embryo. Soaking the entire embryo in drug-containing media resulted in localized effects without requiring localized exposure, due to the sensitivity of HCN2 to local voltage conditions—it is a context-sensitive reagent having different effects on the membrane voltage of cells in the middle vs. surrounding of the neural tube.

It is possible that lamotrigine and gabapentin might also affect other HCN channels or indeed other ion translocators (e.g., inhibitors of Na+ and Ca2+ channels; Goldsmith et al., 2003; Eroglu et al., 2009); however, our data (Figures 1–6, Pai et al., 2018) show that HCN2 is sufficient to induce brain repair, while the drugs do not induce any defects on their own. Also, given the ionic gradients in Xenopus cells (high intracellular Na+ and Ca2+ (Gillespie, 1983) in comparison to external media), Na+ and Ca2+ channel opening would cause hyperpolarization, and hence the non-HCN action of drug (inhibition of Na+ and Ca2+ channels) would actually prevent the hyperpolarization we observed, further pointing towards HCN2 mediated action of these drugs. As numerous ion channel blockers and openers are already approved for human use for arrhythmias, epilepsy, etc., future efforts at identifying novel and more specific HCN2 drugs may be warranted, as part of a strategy to identify useful electroceuticals (Churchill et al., 2019).

Two other aspects should be noted. First, although the nicotine exposure started at stage 10, reparative drug treatment could be initiated after a significant delay (stage 25). We do not know the limit of the delay, but the ability to repair defects significantly later than teratogen exposure began is an attractive feature of this strategy. Second, the repair was not merely morphological but also functional. Remarkably, complex learning behavior can be restored by simple drug treatment, channel misexpression, or implant of HCN2-expressing cells, not requiring high-resolution spatiotemporally patterned manipulation (Figure 10). Thus, a relatively simple bioelectric prepattern can initiate a complex downstream program of gene expression, tissue morphogenesis, and learning behavior (Figure 11). The ability to rescue the function of such a complex, integrated organ, despite the presence of a potent teratogen, underscores the power of capitalizing on biophysical controls of developmental modules. Strategies taking advantage of bioelectric models of health and disease states, as well as the plethora of well-characterized ion channel drugs, could offer a new way to trigger complex patterning modules for applications in birth defects, regenerative medicine, cancer, and bioengineering.

All datasets generated for this study are included in the article/Supplementary Material.

All experiments were approved by the Tufts University Animal Research Committee (M2017-53) in accordance with the guide for care and use of laboratory animals.

VP performed experiments. VP and ML designed the experiments and interpreted the data. VW assisted with the geometric morphometric analysis and in situ hybridization assays. EL assisted in membrane voltage dye analysis. JC and SM designed and performed the simulations. VP and ML wrote the manuscript together.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

We thank Erin Switzer, and Rakela Colon for Xenopus husbandry and general lab assistance, Dany Adams for help with microscopy, Douglas Blackiston for help with learning behavior assay and analysis, Valerie Schneider for otx2 anti-sense probe, Gerald Eagleson for xbf1 anti-sense probe, Nian-Qing Shi and Bin Ye for Hcn2-WT and Hcn2-DN constructs. We thank Jean-Francois Pare and Joan Lemire for assistance with the cloning of cDNAs into injection vectors PCS2. We thank Kelly McLaughlin and Joshua Finkelstein for their helpful suggestions and comments on the manuscript.

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fncel.2020.00136/full#supplementary-material.

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

All datasets generated for this study are included in the article/Supplementary Material.