HCN2 Rescues brain defects by enforcing endogenous voltage pre-patterns

Endogenous bioelectrical signaling coordinates cell behaviors toward correct anatomical outcomes. Lack of a model explaining spatialized dynamics of bioelectric states has hindered the understanding of the etiology of some birth defects and the development of predictive interventions. Nicotine, a known neuroteratogen, induces serious defects in brain patterning and learning. Our bio-realistic computational model explains nicotine’s effects via the disruption of endogenous bioelectrical gradients and predicts that exogenous HCN2 ion channels would restore the endogenous bioelectric prepatterns necessary for brain patterning. Voltage mapping in vivo confirms these predictions, and exogenous expression of the HCN2 ion channel rescues nicotine-exposed embryos, resulting in normal brain morphology and molecular marker expression, with near-normal learning capacity. By combining molecular embryology, electrophysiology, and computational modeling, we delineate a biophysical mechanism of developmental brain damage and its functional rescue.

Improper neural patterning during development leads to highly debilitating disorders, such as open neural tube defects [spina bifida, and anencephaly (small brain)]1, brain malformations2, and susceptibility to autism and degenerative disorders like Parkinson’s and Alzheimer’s diseases3–5. There are not currently any clinically approved interventions able to rescue such brain patterning disorders. Finding regenerative and repair strategies for brain patterning and function is a critical unmet need in developmental and regenerative medicine.

In addition to well-known growth factors and chemical pathways, endogenous bioelectrical signals (spatiotemporal distributions of resting membrane voltage—Vmem—across tissues) have been shown to regulate aspects of large-scale patterning during embryonic development6–13. Endogenous bioelectric signals (produced by the function of ion channels and pumps) have long been known to have roles in cell migration, wound healing, and the direction of growth and form in vivo14,15. Bioelectric signals are implicated in vertebrate appendage regeneration and development16–18, craniofacial morphogenesis19,20, left-right patterning21,22, cancer23, heart and muscle patterning24,25, planarian head regeneration26, and eye and brain patterning27,28. Here, we explore bioelectric signaling as both a target of teratogens that induce brain defects and as a target for therapeutic strategies to rescue such malformations.

Recent studies from our laboratory and other groups (reviewed in Ref. 15) identified molecular mechanisms by which bioelectrical signaling participates in development (including transduction machinery and transcriptional targets). However, the state-of-the-art for in vivo data in this exciting field has been “arrow diagrams” that identify the cellular-level bioelectric pathways of gene products and signals required for development or regeneration. The field still largely lacks knowledge of the quantitative dynamics that are sufficient to produce the spatial patterning across tissues or organs. Widely recognized as the next key step, but still missing, are biorealistic multi-scale models that integrate cell-level electrophysiology and molecular biology with large-scale patterning information (molecular, anatomical, and bioelectric). Such a synthesis is required in order to explain, in a rigorous manner, how bioelectric events originating with ion channel proteins scale up to regulate emergent organ patterning in health and disease. Here, we present and analyze the first example of such a model in the context of teratogenesis of the Xenopus brain and use it to identify a successful repair strategy.

Nicotine is a well-known neuroteratogen: embryonic exposure to nicotine leads to severe brain morphological defects as well as significant postnatal deficits in cognitive functions29–32. The majority of these effects of nicotine occur through its action on nicotinic acetyl choline (nAChR) receptors, but it is not known how this bioelectric change at the single-cell level alters whole organ morphogenesis of the brain. Here, we establish an amphibian model for nicotine teratogenesis, and exploit the Xenopus embryo as proof-of-principle of how computational models of physiological regulatory events can help identify mechanisms of developmental defects and drive the development of therapeutic strategies targeting endogenous bioelectricity.

Hyperpolarization-activated cyclic nucleotide-gated (HCN) channels are a group of voltage-gated ion channels in which the threshold voltage is modulated by the metabolic state of the cell (levels of cyclic nucleotides, like cAMP)33, thus allowing them to regulate cellular voltage in a context-specific manner. They open at hyperpolarized (negative) Vmem, giving rise to currents that are a mix of sodium and potassium fluxes. Among the four HCN isoforms (HCN1–4), HCN2 channels show an especially high sensitivity to cAMP. These channels are primarily present in the adult nervous system and heart, but have also been detected in human and mouse embryonic cells34–37. To the best of our knowledge, these channels have not been studied in the context of embryonic brain development or as targets for developmental therapeutics.

Here we present molecular physiology data and computational analysis explaining neuroteratogen nicotine-induced embryonic brain mispatterning. We report the first biorealistic computational model of Xenopus embryos, which predicts that nicotine disrupts the endogenous bioelectric patterns critical for brain patterning. This mispatterning mechanism is validated using voltage reporter dyes, and we model the consequences of possible methods for manipulating bioelectric gradients in vivo. Our model predicts that rescue would be effected by restoring this embryonic bioelectrical pattern via a reagent with context-specific effects on resting potential: the HCN2 channel. Remarkably, we find that misexpressing HCN2 channels in nicotine-exposed embryos rescues the endogenous spatial voltage gradient in the nascent neural tube and leads to a correction of the expression patterns of key brain transcription factors and a near complete rescue of brain morphology defects and cognitive learning abilities. Our integration of molecular developmental biology and simulation provides insight into the mechanisms of brain development and teratogenesis, and illustrates proof of principle of using computational modeling to understand and manipulate endogenous bioelectrical signaling in the context of regenerative medicine.

Nicotine is a well-known developmental neuroteratogen in humans29, causing morphological defects, decreased cognitive functions, and deficits in learning and memory. We first sought to establish a frog model of nicotine teratogenesis by exposing Xenopus embryos to nicotine (0.1 mg/mL). Nicotine exposure was targeted to stages 10–35, when neural tissue induction and patterning occurs. Untreated embryos served as controls, as neither ethanol nor dimethyl sulfoxide (DMSO) was needed to formulate the stock solution of nicotine used in our assays. Brain morphology of the embryos was assessed after they had developed to stage 45 (Fig. 1). Control tadpoles had correctly-patterned brain tissue5, with well-formed nostrils, olfactory bulbs/forebrain, midbrain, and hindbrain (Fig. 1a, d). Nicotine exposure caused a significantly high incidence of major brain morphology defects (~55%) in comparison to controls (~6%). The most striking phenotypes were the absence of nostrils, absence of forebrain, absence of both forebrain and midbrain, and occasionally misformed hindbrain (Fig. 1b, c) or truncated spinal cord. The hindbrain was the least affected of the brain regions. Eye development requires proper neural induction and development38,39. Tadpoles displayed abnormalies including missing eyes, incompletely formed eyes, fusion of eye to the brain, and pigmented optic nerves (Fig. 1b, c; see also ref. 27). We conclude that nicotine is a powerful neuroteratogen in Xenopus.Fig. 1Nicotine induces brain morphology defects in Xenopus embryos.

Representative images of stage 45 tadpoles: a control tadpole showing nostrils (blue arrowhead), forebrain (FB) indicated by the orange bracket, midbrain (MB) indicated by the yellow bracket, and hind brain (HB) indicated by the cyan bracket, b and c tadpoles from embryos exposed to nicotine (0.1 mg/mL – stage 10–35) showing severe brain morphology defects as indicated by magenta arrowheads. Cyan brackets indicate presence of hindbrain (HB). d Quantification of stage 45 tadpoles for major brain morphology phenotypes in absence or presence of nicotine exposure (0.1 mg/mL – stage 10–35). A significantly high incidence of malformed brain was observed in embryos exposed to nicotine in comparison to controls. Three independent experiments (n = 3) were conducted with N > 50 embryos per treatment group for each of those experiments collected from multiple animals across independent clutches. Data were analyzed with t-test and graphed as mean ± SD; **p < 0.01

Our previous work demonstrated that specific Vmem distributions (pre-patterns) within the developing neural tube and surrounding regions are critical for proper patterning of the developing brain28, while other studies have characterized specific ion channels in developing of neuron populations40. No existing spatial model quantitatively synthesizes the functional and electrophysiological data to explain the control of patterning by endogenous ion channel activity; such a model is needed to understand the biophysics of organ development and to uncover mechanisms of action of ion channel-targeting teratogens. As bioelectric circuits can have highly non-obvious behavior, we used a computational platform, the BioElectric Tissue Simulation Engine (BETSE41), to characterize the dynamics of voltage gradients under the relevant conditions in vivo, and to develop a predictive physiological model (see Supplementary Note). BETSE is a software simulator specially designed to study bioelectricity from a “first principles” perspective, which focuses on ion concentrations, fluxes, and transport using realistic parameters derived from molecular physiology studies of cells to gain insight into spatialized dynamics of relevance to pattern formation41.

BETSE software was used to construct models of Vmem patterns in the early Xenopus neurula and to examine the influence of various interventions (Supplementary Note). Initial (i.e., simulation time zero) ion and substance concentrations in the cytosol, intercellular space, and global environmental media are summarized in Table 1. BETSE modeled an exterior sheet of cells representing an anterior view of a stage 15 Xenopus neurula, where cells were assumed to be a 10-μm-thick layer facing 0.1 × MMR (Fig. 2a). Voltage-sensitive gap junctions (GJs) were employed in all models, using the kinetic model of voltage sensitivity fit to amphibian gap junctions42.Table 1Initial simulation concentrations in the intracellular space (cytosol), extracellular space (intercellular regions), and global environmental space surrounding cells (0.1 × MMR)IonIntracellular (mM)Extracellular (mM)0.1 × MMR (mM)Na+21.3101.810.0K+91.03.70.5Cl−40.237.69.0Ca2+50e-61.50.1Anionic protein47.010.00.0ATP2.50.00.0ADP0.10.00.0Pi0.10.00.0H+4.0e-54.0e-54.0e-5HCO3−10.010.01.0Charge balancing anions15.147.91.5Fig. 2BETSE model of endogenous resting potential gradients that instructively pattern embryonic neural tissue. a The BETSE model of neurula stage Xenopus embryo featured three different cell groups (referred to as tissue profiles) as indicated by color and labels. The neural tube profile was made to have 5× higher leak membrane permeability to K+ ions and 15× lower membrane permeability to Cl− ions, compared to the other two profiles, leading to hyperpolarized resting Vmem in neural tube cells as has been experimentally observed28, 48. The model is assumed to be the outer layer of cells which face 0.1 × MMR. See supplemental document for detailed description of model. b The regulatory network describing the main components of the model.

All cells of all simulations expressed equal levels of Na/K-ATPase ion pumps, inward-rectifying K+ channels (Kir), and nAChR cation channels. Some simulations included HCN2 channels equally expressed on all cells, and/or nicotine introduced via environmental exposure

A depiction of the complete bioelectric regulatory network underlying all BETSE models reported herein (including the optional elements: nicotine, IT, and HCN2 channels) is shown in Fig. 2b (Supplementary Note). The control model represents the fundamental (endogenous baseline) physiological situation, without nicotine treatment or IT misexpression. The control model considered three different spatial regions of cells, roughly demarcating the ectoderm, neural folds, and neural tube areas of the embryo from an anterior view of the outer surface (Fig. 2a). Variations in the base level of membrane permeability to K+ and Cl−(Fig. 2a) simulate the presence of leak channels, creating the observed endogenous relative hyperpolarization in the neural tube compared to the neural folds and ectoderm. This results in open Kir channels in the hyperpolarized neural tube region leading to a Vmem pattern of approximately −50 mV in the neural tube region, while the lateral neural folds and ectoderm remain at ~−5 mV (Fig. 3a, e). This hypothesis of generation of the endogenous Vmem pattern in the Xenopus neural tube is consistent with observed Vmem patterns in Xenopus embryos and the effect of dominant-negative Kir channels on them27,28.Fig. 3BETSE model predicts that nicotine suppresses neural tube hyperpolarization and HCN2 recovers neural tube hyperpolarization in presence of nicotine. a–e BETSE model of a stage 15 Xenopus embryo where cells are assumed to be facing 0.1 × MMR. a Control model exhibits a characteristic Vmem pattern featuring relative hyperpolarization in the neural tube area and depolarization everywhere else. This Vmem pattern is actually seen in the Xenopus embryos with voltage reporter dyes28. b Nicotine treatment of the control model is predicted to preferentially depolarize the neural tube area to suppress the Vmem gradient.

c Model with ion translocator (IT) expression in all cells, acting in a context-specific manner showing strongly enhanced neural tube Vmem pattern. d Adding the ion translocator (IT) to the nicotine treated model shows the Vmem pattern significantly retained in comparison to the only nicotine treated model of b. e Line graph showing the information retrieved along the white dotted lines depicted in (a–d), which allows for easy comparison between models. f The effect of the HCN2 channel on the Vmem of a series of cells with different membrane leak permeabilities (Pmem). The ability of homogeneously expressed HCN2 channels to act in a context specific manner is predicted to stem from the channels’s opening at hyperpolarized Vmem with a hyperpolarizing effect upon opening. This selectively hyperpolarizes cells while leaving relatively more depolarized cell’s Vmem unchanged. This further emphasizes that HCN2 channels amplify the endogenous Vmem pattern of the neural tube and maintain the Vmem pattern against nicotine depolarization exactly like IT in (c, d and e)

Simulation of nicotine exposure was built upon the control model by incorporation of nicotine-sensitive ionotropic nAChR receptors, expressed homogenously throughout the embryo43 (Supplementary Note). While nicotine is a well-known agonist of the nicotinic acetylcholine receptor (nAChR)44, it can also directly block a variety of K+ ion channels45, including the HCN channels46. Nicotine block of HCN channels is only partial (<39%). All this information was included in the models using a standard Hill function to describe the influence of nicotine on the channel state as a function of its concentration (Supplementary Note). Nicotine exposure simulated 0.62 mM diffusion of nicotine from the environment, which the model predicted would depolarize the Vmem by increasing Na+ membrane permeability (Fig. 3b, e). Treating the model with nicotine predicted to diminish the endogenous neural tube hyperpolarization pattern in comparison to controls (Fig. 3a, b, and e), thus reducing the spatial contrast of Vmem between the hyperpolarized neural tube and the depolarized neural folds and ectoderm (Fig. 3e). We tested the BETSE model predictions about nicotine effect on Vmem patterns in stage 15 embryos, using a combination of in vivo voltage reporter dye (CC2-DMPE:DiBAC4(3)) imaging47 and whole cell Vmem recordings27,28. Whole-cell electrophysiological recordings of Vmem from neural plate cells and flanking ectodermal cells provided calibration points for the voltage reporter dye images; the fluorescence intensities were analyzed against these calibration points to approximate membrane voltages at different points within the developing embryos (Fig. 4 and Supplementary Figure 1). Nicotine-treated embryos showed significant reduction in endogenous neural Vmem signal (depolarization by ~15 mV) in comparison to controls (Figs. 4a, b, e, f, and Supplementary Figure 1).

This observation matches the BETSE model predictions of nicotine effect on Vmem both in direction (depolarization) and magnitude (~15 mV). Our model explains why nicotine induces brain defects: it correctly predicts the disruptive effects of this teratogen on the endogenous bioelectric prepattern required for brain development.Fig. 4Validation of the model prediction of nicotine effects and HCN2-induced recovery of membrane voltage prepatterns. Representative CC2-DMPE:DiBAC4(3) images of stage ~15 Xenopus embryos: a untreated controls, b nicotine-exposed (0.1 mg/mL – stage 10–35), c Hcn2-WT mRNA microinjected (0.75 ng/injection) in both blastomeres at 2-cell stage, and d nicotine-exposed (0.1 mg/mL – stage 10–35) and Hcn2-WT mRNA microinjected (0.75 ng/injection) in both blastomeres at 2-cell stage. Control embryos show the characteristic hyperpolarization (solid yellow arrow) as previously reported28. Nicotine-treated embryos show reduced signal (depolarized) within the neural tube (hollow yellow arrow). Hcn2-WT mRNA microinjection show enhanced signal (hyperpolarized) within the neural tube (magenta arrows), in presence or absence of nicotine exposure. e Quantification of CC2-DMPE:DiBAC4(3) images of stage ~15 Xenopus embryos along the red dotted line as indicated in the inset illustration, along with electrophysiology based membrane voltage approximations (as previously reported in refs.27,28). The fluorescence intensity/membrane voltage pattern within the neural tube (indicated by the black dotted line in the inset illustration and corresponding lack dotted line in the graph) is significantly reduced (depolarization) in nicotine-exposed embryos in comparison to controls. Hcn2-WT mRNA microinjection significantly enhances the fluorescence intensity/membrane voltage patterns within the neural tube in comparison to controls.

Nicotine-exposed embryos that are microinjected with Hcn2-WT mRNA maintain a significantly enhanced fluorescence intensity/membrane voltage pattern within the neural tube in comparison to only nicotine-treated embryos. N = 10 embryos for each treatment group at each of the indicated spatial distance in pixels were collected from multiple animals across independent clutches. Data is plotted as mean ± S.E.M. Data at black line (400 Pixels) was analyzed using one way ANOVA, *p < 0.05, **p < 0.01. (f) Quantification of peak fluorescence intensity and electrophysiology based membrane voltage approximations (as previously reported in ref. 27,28) from voltage reporter dye images (CC2-DMPE:DiBAC4(3)) of ~stage 15 Xenopus embryos within the neural tube at the intersection of the red and black dotted lines in the inset illustration in (a). Nicotine exposure significantly reduced (depolarizes) the neural tube peak intensity/membrane voltage in comparison to controls. Hcn2-WT mRNA microinjection, both in presence or absence of nicotine, significantly enhances (hyperpolarizes) the neural tube peak intensity/membrane voltage in comparison to controls. N = 10 embryos for each treatment group at each of the indicated spatial distance in pixels were collected from multiple animals across independent clutches. Data is plotted as mean ± SD and was analyzed using one way ANOVA, *p < 0.05, **p < 0.01, ***p < 0.001

Simulations of Vmem manipulation using an ion translocator (IT) involved the use of the control model with IT equally expressed in all cells. Simulations indicated that a homogenously expressed IT that opened preferentially in the hyperpolarized neural tube region, while remaining closed in the two lateral regions of the model, would lead to accentuation of neural tube Vmem pattern to ~−80 mV, while keeping remaining two regions of the model at ~−5 mV (Fig. 3c, e)—the spatial prepattern observed in wild-type embryos. The spatial patterning is crucial, as both depolarization of the middle region, or hyperpolarization of the lateral regions, both cause brain defects—it is the precise differential bioelectric state that induces the required balance of proliferation, apoptosis, and gene expression to build a normal brain28,48. Our model predicted that misexpression of an IT that amplified endogenous Vmem patterns (hyperpolarization) in the neural tube without affecting the Vmem patterns in the surrounding ectoderm would lead to maintenance of the endogenous neural tube Vmem pattern even in nicotine-exposed model embryos (Fig. 3c vs. d, e).

BETSE predicted HCN2 channels as an ideal reagent to meet this requirement. Its Vmem pattern amplification arises from the fundamental dynamics of the HCN2 channel in 0.1 × MMR (which is a low ionic environment). HCN2 channels (mostly studied in adult mammalian cardiac and neural tissues), in conjugation with voltage-gated Na+ and K+ channels, create self-oscillating Vmem signals33. The frequency of these Vmem oscillations is regulated by cAMP, which shifts the opening threshold of HCN2 channels to more positive Vmem values and couples HCN2-mediated Vmem oscillation frequency to cAMP modulators and cell metabolism33,49. Under Xenopus ionic parameters and external media, HCN2 channels hyperpolarize (Table 1 and Fig. 3f), which is also consistent with predictions using the GHK equation. HCN channels are primarily permeable to Na+ and K+ in a Na+:K+ ratio of ~0.2–0.33 (ref. 33). In this first model of the role of HCN2 channels in Xenopus development, the voltage and cAMP sensitivity of HCN2 channels were accounted for using kinetic models of HCN2 voltage sensitivity50 in combination with the assumption that the presence of cAMP shifts the V1/2 opening threshold of the HCN2 channel by + 20 mV (ref. 49) from its baseline value of V1/2 = 99.0 in the absence of cAMP50. HCN2 channel models used an Na+:K+ permeability ratio of 0.2 (ref. 33).

Our modeling predicted that HCN2 would be able to amplify the endogenous Vmem gradient patterns in Xenopus neurula (Fig. 3f). HCN2 channels are predicted to open at hyperpolarized Vmem, generating more hyperpolarization with opening, thereby serving as a Vmem-controlled switch of Vmem state, capable of amplifying small Vmem gradients, similar to IT effect (Fig. 3c, e, f). Thus, HCN2 channel transitions the Vmem of sufficiently electronegative cells to very hyperpolarized state, while leaving the Vmem of other relatively depolarized regions unchanged (Fig. 3c, e, f). Simulations indicated that homogenously expressed HCN2 channels open preferentially in the neural tube region, while remaining closed in the other two regions of the model. This leads to accentuation of neural tube Vmem pattern ~−80 mV, while keeping remaining two regions of the model at ~−5 mV, similar to IT (Fig. 3c, e, f). Importantly, our model predicts that HCN2 expression leads to maintenance of neural tube Vmem pattern even in nicotine-exposed model embryos, similar to the action of IT (Fig. 3c vs. d, e, f). This result suggests that HCN2 channels should amplify the intrinsic Vmem pattern of the neural tube, leading to a maintenance of Vmem pattern in spite of the depolarizing influence of nicotine. Thus, our computational model begins with a quantitative physiological explanation of the brain defects induced by nicotine and identifies an intervention strategy, predicting repair by the HCN2 channel.

To test the predictions made by the BETSE model about nicotine’s and HCN2’s effects on membrane voltage prepatterns, we evaluated embryos between stage 15–17 using a combination of in vivo imaging with voltage reporter dye pair CC2-DMPE:DiBAC4(3) (ref. 47) and whole cell Vmem recordings (Fig. 4; refs. 27,28,]). Whole-cell electrophysiological recordings of Vmem 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 (Fig. 4 and Supplementary Figure 1). Uninjected and untreated embryos (controls), embryos exposed to nicotine (stage 10 onward), embryos microinjected with Hcn2-WT (wild type) mRNA (both blastomeres at two-cell stage for an overall uniform expression) and embryos exposed to nicotine that are also injected with Hcn2-WT mRNA were analyzed.

Nicotine-treated embryos showed significant reduction in endogenous neural Vmem signal (depolarization by ~15 mV) in comparison to controls (Fig. 4a, b, e, f, and Supplementary Figure 1). This observation matches the BETSE model predictions of nicotine effect on Vmem both in direction (depolarization) and extent (~15 mV). Consistent with the BETSE model, voltage imaging in Hcn2-WT-injected embryos revealed a significantly enhanced endogenous neural signal (hyperpolarization by ~15 mV) in comparison to controls (Fig. 4a, c, e, f, and Supplementary Figure 1). However, the extent of enhanced hyperpolarization predicted by BETSE model for IT and HCN2 was higher (~30 mV) than actually observed (~15 mV) (Figs. 3e and 4). Crucially, nicotine-exposed embryos that were injected with Hcn2-WT channel mRNA showed remarkably enhanced (hyperpolarized better than controls by ~20 mV) endogenous neural Vmem signals in comparison to only nicotine treated embryos (hyperpolarized by ~30 mV) (Fig. 4a, b, d, e, f, and Supplementary Figure 1). This observation largely fits BETSE model predictions in both, direction of Vmem change, and extent of Vmem change (Fig. 3 and 4). As predicted, nicotine exposure suppresses the neural brain-specific Vmem pattern, while HCN2 channel expression enhances the neural Vmem pattern and restores it even in presence of the otherwise teratogenic nicotine. Having established that HCN2 can repair the bioelectric prepattern, we next assayed its effects on brain structure and function.

As a prelude to testing its repair function, we examined the effects of HCN2 channel expression on brain morphology and patterning in normal embryos. We injected Hcn2-WT mRNA into both blastomeres of wild-type embryos at the two-cell stage and assayed brain morphology at stage 45 (swimming tadpoles). Hcn2-WT mRNA injection increased the HCN2 channel levels by ~3.1 times (310%) in the injected embryos in comparison to controls as observed by immunostaining followed by quantitative analysis of control and injected embryos (Supplementary Figure 3). Even very small instances of morphological deviation from perfectly normal brain morphology were documented, including slightly shortened nostril distances, and unequal forebrain/olfactory bulbs. In untreated control embryos, a normal basal level of minor incidences of deviation from ideal brain morphology was observed in this very strict assay (~9%). Remarkably, we noticed that Hcn2-WT channel-injected embryos showed a significant reduction of these minor brain morphology phenotypes (~1.5%, **p < 0.01, t-test n = 3 with N > 50 embryos for each group for each of the experiments collected from multiple animals across independent clutches) (Fig. 5a, b): the channel mRNA-injected embryos were better than the controls. Thus, misexpression of HCN2 channels resulted in an improvement of developmental patterning even beyond that observed in control populations of unperturbed embryos raised under optimal conditions. This result is consistent with the use of our BETSE model to identify Vmem-modifying interventions that reinforce endogenous patterns and might be useful candidates for improving anatomical structure (Fig. 3c, e).Fig. 5HCN2 channels rescue nicotine-induced brain morphology defects in Xenopus embryos.

a Representative image of stage 45 tadpole with an example of a minor brain morphology defect showing smaller forebrain and nostril on the left (magenta arrowhead), normal midbrain (MB - yellow bracket) and normal hindbrain (HB - cyan bracket). b Quantification of stage 45 tadpoles for subtle changes in overall brain morphology with or without microinjecting Hcn2-WT (wild-type) mRNA (0.75 ng/injection) in both blastomeres at 2-cell stage as indicated in the illustrations. Hcn2-WT mRNA injections significantly suppress the minor brain defects seen in uninjected control embryos. Three independent experiments (n = 3) were conducted with N > 50 embryos per treatment group for each of those experiments, collected from multiple animals across independent clutches. Data were analyzed with t-test and graphed as mean ± SD, **p < 0.01. Representative images of stage 45 tadpoles: c control tadpole showing nostrils (blue arrowhead), forebrain (FB) indicated by the orange bracket, midbrain (MB) indicated by the yellow bracket, and hind brain (HB) indicated by the cyan bracket, d tadpole from embryos exposed to nicotine (0.1 mg/mL – stage 10–35) showing severe brain morphology defects as indicated by magenta arrowheads. e tadpole from embryos exposed to nicotine (0.1 mg/mL – stages 10–35) and microinjected with Hcn2-WT mRNA (0.75 ng/injection) in both blastomeres at 2-cell stage showing intact nostrils (blue arrowheads), forebrain (FB - orange brackets), midbrain (MB - yellow brackets), and hindbrain (HB - cyan brackets). f Quantification of stage 45 tadpoles for major brain morphology phenotypes in absence or presence of nicotine exposure (0.1 mg/mL – stage 10–35) with or without microinjection of Hcn2-WT or Hcn2-DN (dominant-negative) mRNA (0.75 ng/injection) in both blastomeres at 2-cell stage as indicated in the illustrations.

A significantly high incidence of malformed brain was observed in embryos exposed to nicotine in comparison to controls. Hcn2-WT mRNA injection significantly reduced the incidence of malformed brain, while Hcn2-DN mRNA injection had no significant effect on nicotine exposure induced malformed brain. Three independent experiments (n = 3) were conducted with N > 50 embryos per treatment group for each of those experiments collected from multiple animals across independent clutches. Data were analyzed with one way ANOVA and Tukey’s post-test and graphed as mean ± SD,***p < 0.001, n.s. non-significant

To test the predicted effect of HCN2 on nicotine’s teratogenic influences on embryonic brain patterning, we exposed Xenopus embryos to nicotine (0.1 mg/mL – stage 10–35). We tested the effect of both Hcn2-WT and Hcn2-DN (dominant-negative) construct. Hcn2-DN was generated by mutating the highly conserved cation selective pore-domain GYG (amino acid 402–404) to AAA, a strategy known to drastically reduce conductance and generate dominant-negative mutants of HCN4 channels24,51 (Supplementary Figure 2). Hcn2-WT or Hcn2-DN mRNA was microinjected into both blastomeres at two-cell stage along with nicotine exposure. Untreated (uninjected) embryos served as controls. After the embryos were developed to stage 45, we quantified brain morphology (Fig. 5c–e). Control tadpoles had correctly patterned brain tissue5, including normally developed nostrils, olfactory bulbs, forebrain, midbrain, and hindbrain (Fig. 5c, f). As before (Fig. 1), nicotine exposure caused a significantly high incidence of major brain morphology defects (~60%) in comparison to controls (~7%) (Fig. 5c–f). Remarkably, Hcn2-WT mRNA-injected embryos showed a complete rescue of brain morphology defects from nicotine exposure, to levels (~6%) indistinguishable from that of untreated control embryos (~7%) (Fig. 5f). The nicotine-exposed, Hcn2-WT mRNA injected tadpoles showed nostrils, olfactory bulbs and forebrain, midbrain, and hindbrain that had formed similarly to controls (Fig. 5e). In contrast, nicotine-exposed embryos injected with a mutant, dominant-negative form of HCN2 failed to show rescue of brain morphology defects (~70%) (Fig. 5f).

To quantify brain shape, we used geometric morphometrics on stage 45 tadpoles52,53 (Fig. 6). Landmarks were chosen based on transition points between different brain regions, and lateral and anterior outermost points of the head (Fig. 6b) and recorded for n > 26 tadpoles for controls, nicotine, and nicotine + Hcn2-WT mRNA. Canonical variate analyses (Fig. 6a) were run on the data set, with Procrustes distances between each of the groups to characterize shape changes in each condition.Fig. 6HCN2 channels restore the brain size relation to anterior head shape in nicotine-exposed embryos. Morphometrics canonical variate analysis of brain size relation to anterior head shape of stage 45 tadpoles. a Graphical output, showing confidence ellipses for means, at a 0.95 probability, of shape data from controls and nicotine-exposed (0.1 mg/mL – stage 10–35) tadpoles with or without Hcn2-WT mRNA (0.75 ng/injection) microinjected in both blastomeres at 2-cell stage. Ellipses are colored to correspond with treatment as indicated. N > 26 for each group. Procrustes distances show objective alteration of brain size in relation to head morphology by nicotine exposure in relation to controls as indicated by the red-blue arrow. Hcn2-WT mRNA microinjection along with nicotine exposure moves the brain size—head shape relation closer to controls as indicated by the blue-green arrow. b Stage 45 control tadpole image illustrating the 7 landmarks considered for this analysis comparing brain size and head shape. Landmarks 2, 3, 4, and 5 show the start of forebrain, transition to midbrain, transition to hindbrain, and end of hindbrain, respectively. Landmarks 1, 6, and 7 indicate the anterior most, and lateral most points of the head shape. c Canonical variate 1 axis legends (ball and stick diagrams) showing movement of each of the 7 landmarks in mainly anterior-posterior direction.

Each ball represents the landmark as indicated by the number and the accompanying stick represents the direction and extent of movement of that particular landmark. d Canonical variate 2 axis legends (ball and stick diagrams) showing movement of each of the 7 landmarks in mainly lateral direction. Each ball represents the landmark as indicated by the number and the accompanying stick represents the direction and extent of movement of that particular landmark

The quantification revealed whether the length of the brain was changed in relation to the head shape in nicotine-exposed embryos and whether Hcn2-WT mRNA injections restored wild-type proportions. Canonical variate analysis and Procrustes distances confirmed that control and nicotine + HCN2-WT group were not different in shape (Fig. 6a, green and red ellipses), but significantly different from nicotine-exposed embryos (blue ellipse) (Procrustes distances: Control vs. nicotine + HCN2-WT = 0.0446, Control vs. nicotine = 0.1222, and nicotine vs. nicotine + HCN2-WT = 0.1200). ANOVA of centroid shape between the controls and nicotine + HCN2-WT in relation to nicotine treatment confirmed significant differences between the groupings (F = 12.45, p < 0.0001). Thus, nicotine significantly changes brain shape, but Hcn2-WT mRNA microinjections restore brain morphology to the wild-type state.

Using an automated behavior analysis platform54,55 (Fig. 7a), a robust red-avoidance behavior can be produced and quantified in normal tadpoles (Fig. 7b, c). A tadpole was classified as ‘having learned’ if their preference for red light drops 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 manipulations56–58. We tested nicotine-exposed tadpoles (0.1 mg/mL – stage 10–35) with or without the Hcn2-WT mRNA microinjection (0.75 ng/injection) in both blastomeres at two-cell stage.Fig. 7HCN2 channels restore associative learning capacity in nicotine-exposed embryos. Associative learning capacity analysis for stage 45–50 tadpoles either left untreated (controls) or exposed to nicotine (0.1 mg/mL – stage 10–35) with or without microinjections of Hcn2-WT mRNA (0.75 ng/injection) in both blastomeres at 2-cell stage as indicated in the illustrations. a The training regime used consisted of an innate preference test, a training phase, a rest period, and a learning probe. Tadpoles were placed individually in the behavior analysis robot. Motion tracking cameras under each tadpole recorded its position/behavior, and automated software executed a training cycle where animals received a shock when occupying the red half of the arena. Training, rest and testing sessions were repeated a total of six times across the trial. b Quantification of time spent in the red color during the final testing probe for tadpoles in each treatment group. N > 20 for each experimental group. Error bars indicate ± S.E.M. Data were analyzed using one way ANOVA ***p < 0.001. c Representation of individual tadpole’s associative learning test. N > 20 for each experimental group.