|
Figure 1. The challenge of anatomical homeostasis
(A) Significant remodeling processes turn tadpole heads into those of frogs. Remarkably, highly abnormal tadpole faces also make largely normal frogs as the organs rearrange and deform as needed, until a correct frog face is built.
(B) Tails transplanted onto the flank of a salamander can slowly remodel into legs—a structure more appropriate in the global context (schematized after data in Farinella-Ferruzza [1956] and Holtfreter [1955]). This includes tail-tip cells, marked in blue, which become toes, despite a normal local environment, illustrating the remodeling of tissue structure based on a global target morphology.
(C) Kidney tubules form with roughly correct diameter in newts despite polyploidy, which changes cell size. As the cells get bigger, the system adjusts to utilize a smaller number of cells to achieve the same geometrical endpoint as the cells enlarge due to polyploidy. Remarkably, when the cells get very large, development abandons normal multi-cellular coordination pathways and exploits cytoskeletal biomechanics to achieve the same lumen from just one cell bending around itself. This illustrates the use of diverse molecular mechanisms in novel circumstances to achieve the same large-scale anatomical target morphology. This figure was schematized after Fankhauser (1945).
All panels by Jeremy Guay of Peregrine Creative.
|
|
Figure 2. Developmental bioelectricity at the cellular level
(A) Neural bioelectric dynamics are created by ion channels that set the resting potential across the plasma membrane and electrical synapses that communicate electric states across networks of neurons.
(B) More ancient mechanisms present in all cells use ion channels to regulate cellular Vmem (resting potential across the plasma membrane) and selectively propagate those bioelectric states to their neighbors.
(C) The bioelectric state of a cell is a complex function of its past history and its neighborsâ bioelectrical states, due to context-sensitive (e.g., voltage-gated) ion channels determined by the genome and transcriptional mechanisms.
(D) Tissues form bioelectrical networks similar to neural networks. Spatiotemporal Vmem distributions are regulated by the temporal evolution of differential voltage states of individual cells and the changing topology of connections via electric synapses (gap junctions, which are themselves voltage sensitive). Bioelectric gradients use the movement of intracellular (e.g., neurotransmitters) and extracellular morphogens (signaling molecules) to regulate gene expression and morphogenesis. Genetic, optogenetic, and pharmacological approaches target the gap junctions and ion channels, exploiting synaptic and intrinsic plasticity non-neural cells, just as they are routinely used in neuroscience to modulate computation in neural circuits.
|
|
Figure 3. Tissue- and organ-level bioelectric patterns
(A) Spatial distributions of Vmem across tissue can be visualized by voltage-sensitive reporter dyes to reveal the depolarized head end and hyperpolarized tail end of the bioelectric circuit in a planarian flatworm fragment that determines its future regenerative anatomical polarity.
(B) Endogenous prepatterns instructively guide morphogenesis, such as this “bioelectric face†distribution on the early frog ectoderm that determines the gene-expression domains and the borders of the compartments for the eyes, mouth, arches, etc. (white signal demarcates depolarized cells, such as the embryo’s right eye, appearing slightly earlier than the left). Drugs, ion-channel misexpression, or optogenetics can be used to alter this pattern resulting in predictable changes in the domains of expression of craniofacial patterning genes and subsequent head anatomy.
(C) Individual cells transforming into cancer depolarize as a very early step in the process, leading to an electrical de-coupling from tissue-level organizational cues (here shown as an early stage of a tumor in tadpoles injected with human oncogenes; voltage-sensitive fluorescent dyes reveal location and size of the cells abandoning their participation in organogenesis in favor of tumorigenesis).
Image in (A) by Taisaku Nogi. Images in (B) and (C) are used with permission from Chernet and Levin (2013b) and Vandenberg et al. (2011).
|
|
Figure 4. Ion-channel loss- and gain-of-function developmental phenotypes reveal non-neural functions of bioelectricity across phyla
Red arrows indicate malformed features. “WT†denotes “wild type†normal morphology for each species’ feature.
(A) Drosophila wings that develop without functional ion channels such as irk2, which encodes a potassium channel or the Rpk (Ripped Pocket or ENaC) sodium channel have characteristic defects: they exhibit bristle malformations, abnormal vein patterns, hinge defects, and can be reduced in size or almost completely missing. Schematics are based on phenotypic data from Dahal et al. (2012), (2017) and George et al. (2019).
(B) Zebrafish mutants reveal ion channels essential for developmental control of proportion. Compared to wild-type fish (WT), mutations in Connexin43 (shortfin) exhibit reduced fin size, while gain-of-function mutations in kcnh2a (longfin), kcnk5b (another longfin), or kcc4a (schleier) induce increased fin size. Barbels, sensory whiskers of the fish, also show variable proportion in shortfin and other longfin/scheleir mutants, but not in longfin. This difference, as well as different characteristics of growth in the mutant classes, suggests that bioelectrics is regulating separate developmental processes during appendage formation. Schematics are based on phenotypic data from Daane et al. (2018), Iovine et al. (2005), Lanni et al. (2019), and Perathoner et al. (2014).
(C) Xenopus embryos in which the function of the HCN4 (hyperpolarization-activated cyclic nucleotide-gated potassium channel) is abrogated exhibit cardiac malformations such as twisting, rotation, failure to loop, and double ventricles (Pai et al., 2017; Pitcairn et al., 2017). Targeting ATP6 (V-ATPase), ATP4 (H,K-ATPase), KCNJ8 (Kir6.1), or KCNA9 (KCNQ1) results in randomization of left-right asymmetry, including independent reversals of the heart, gut, and gall bladder.
(D) Kcnj2 (Kir2.1) potassium channel mutations in mice result in deformations in craniofacial and limb structures. Phenotypes include cleft palate, hypoplastic tympanic ring, nasal bones, maxilla, premaxilla, mandible, and enlarged fontanelle. Limb defects include extra digits and digits that are reduced in size. Schematics are based on phenotypic data from Belus et al. (2018) and Dahal et al. (2012).
(E) Human patients with mutations in kcnj2 (Kir2.1) potassium channels exhibit Andersen-Tawil syndrome, with craniofacial malformations that include a cleft or high arched palate, broad forehead and nose, wide-set eyes (hypertelorism), low-set ears, and a small lower jaw (micrognathia). Individuals with mutations in kcnj2 often also have digit defects such as clinodactyly (abnormal curvature), brachydactyly (shortened digits), and syndactyly (fused digits). Schematics based on Adams et al. (2016), Bates (2013), and Yoon et al. (2006).
All schematics are courtesy of Jeremy Guay of Peregrine Creative.
|
|
Figure 5. Reprogrammable biological hardware
(A) A standing bioelectrical pattern of resting potentials exhibits one depolarized region in wild-type worms (a′) and a mirror-image bipolar pattern in worms that are, or are going to be, two headed (a′′).
(B) An electrical circuit composed of a proton/potassium exchanger pump and several potassium and chloride channels is instructive for anatomy: using RNAi or drugs to target the channels to duplicate or remove the depolarization alters the downstream expression of canonical axial polarity genes and produces two-headed or no-headed worms, respectively.
(C) Two-headed worms made this way reveal a permanent revision of the target morphology: subsequent rounds of regeneration in plain water, long after the reagent is gone from the tissue, continue to make two-headed worms. The pattern to which these cells will build upon damage is stably re-written by a brief (∼2-day) modulation of their bioelectric state despite their wild-type genome.
(D) A dynamical state portrait of the combined electrochemical circuit illustrates various attractors that represent stable morphologies for this system, as stable primitive pattern memories in a non-neural connectionist network.
(E) One way to model how bioelectric circuits encode pattern memories is provided by the field of artificial neural networks, where specific memories are represented by attractors in the state space of the network. The tissue-level bioelectric patterns corresponding to each stable attractor in this network trigger different downstream changes in signaling pathways, distributions of morphogens, and transcription. These changes in turn result in emergent large-scale morphogenetic features. The connectionist paradigm from neuroscience thus facilitates models in which information processing events on multiple scales (pattern memories and cellular signaling events) are functionally integrated in the same model.
(A) and (B) are used with permission from Pezzulo et al., 2021 and Beane et al. (2011), respectively. (D) and (E) were drawn by Justin Guay of Peregrine Creative.
|
|
Figure 6. Exploring regions of morphospace belonging to different species, without genomic change
Temporary reduction in bioelectrical coupling within G. dorotocephala leads to a regeneration of heads that resemble other species of planaria, including the round heads of S. mediterranea and P. felina (A). The process is stochastic, and the frequency of each type of head is proportional to the evolutionary distance between them and the original species (Emmons-Bell et al., 2015). One conceptual way to model this is as a morphospace shaped by the bioelectric and biochemical circuits that dictate head shape. This space has several attractors, and the act of decapitation raises the system out of this stable attractor, to a high-energy configuration that it will reduce through regeneration (B). If bioelectric signaling in the network is impeded during the process, the system can stochastically drop into one of the other attractors instead of the system default shape, which it uses reliably during normal regeneration. Cracking the morphogenetic code will involve mapping the circuits to understand all of the endogenous attractors, their role in evolution, and the manipulations that can push the system toward desired ones in biomedical settings. Images of planaria courtesy of Junji Morokuma and Richard Gawne. (B) was used with permission from Sullivan et al. (2016).
|
|
Figure 7. A roadmap for electroceutical biomedicine
(A) The large set of known ion-channel drugs, many of which are human approved, can be exploited in regenerative medicine. Existing transcriptomic databases reveal which ion channels are potential targets for intervention in any tissue and serve as inputs to a computational environment that simulates bioelectric circuits and thus can predict the large-scale patterns that would result from opening or closing specific channels. These in silico models can be used to derive candidate interventions—blends of specific small-molecule compounds that target tissue-specific ion channels to induce the desired bioelectric state to trigger repair, remodeling, or normalization, as needed. The image was taken with permission from Churchill et al. (2019).
(B) The long-term strategy is to achieve a multi-scale understanding of anatomical homeostasis that integrates from transcriptional networks that produce ion-channel and gap-junction proteins, to single-cell voltage states, which scale up to tissue-level bioelectric dynamics and implement body-wide circuits that make modular decisions about large-scale anatomical features. Applying interventions at this level will make it much easier to induce and shape complex anatomical outcomes for regenerative medicine top-down, overcoming the complexity barriers that limit bottom-up rewiring approaches. Panels are by Jeremy Guay, Alexis Pietak, Daniel Lobo, and Jonathan Slack.
|
