Long-term editing of brain circuits using an engineered electrical synapse

MainElectrical synapses enable the direct flow of ions and small molecules between two cells and play a prominent part in coupling electrical activity in multiple organs, including the brain2,3,4. Electrical synapses comprise multiple gap junction channels, each composed of two docked hemichannels embedded in the membranes of two touching cells. Each hemichannel is an oligomer that consists of six monomeric proteins called connexins, of which there are 21 isoforms in humans5,6. Most connexins can form single-isoform hemichannels that dock with themselves to create homotypic gap junctions (Fig. 1a, left).Fig. 1: Screen to identify a mutant connexin hemichannel pair with exclusively heterotypic docking.The alternative text for this image may have been generated using AI.Full size imagea, Left, schematic outlining the limitation of introducing heterologous WT connexin hemichannels (pink rectangles) to modulate specific neural circuits composed of neurons (brown and yellow). Note that connexin hemichannels produce off-target electrical synapses between presynaptic neurons (left). Right, strategy for using exclusively heterotypic docking hemichannels (green and red rectangles) to selectively modulate specific neural circuits. b, Depiction of red (iRFP670) and green (mEmerald) fluorescence-exchange profiles (left) and representative flow cytometry plots (right) for hemichannel pairs with (Cx36–Cx36; top) and without (Cx36–Cx45; bottom) docking compatibility. The pink dashed squares in the flow cytometry plots highlight the proportion of cells that express two distinct fluorescent proteins. c, Left, proportion of dual fluorescence-labelled cells for connexin pairs with known docking compatibility profiles. Right, FETCH scores for Cx43(F199L)–Cx43(F199L) and Cx26(K168V N176H)–Cx43 (ref. 26). Blue lines on the right-hand graph are the mean ± s.e.m. score for the known-negative distribution of connexin pairs with docking incompatibility. d, Schematic of M. americana Cx34.7 and Cx35 mutations in EL1 and EL2 used to screen for heterotypic-exclusive hemichannels.

Positions and mutations specific to Cx34.7 or Cx35 or common to both proteins, are shown in green, red and black, respectively. e,f, Plots showing homotypic FETCH results for Cx34.7 (e) and Cx35 (f) mutations. The locations of these mutations can be mapped back to the structure for EL2 in d by the mutation number and colour. Circular bar graphs show the Cohen’s D effect size of FETCH scores for homotypic mutant combinations compared with the heterotypic pairing of human Cx36 and Cx45, which fails to dock. The black horizontal line in the centre is the scale bar for effect sizes. Targeted residues are listed around the circle rim; substituted amino acids are listed in the interior. The intermittent black circle segregates each targeted residue, and the light purple circle corresponds to a Cohen’s D of zero. Mutations that disrupted docking are also highlighted by black arrows and letters. g, Heterotypic FETCH results for Cx34.7 and Cx35 mutant protein combinations. Bar graphs show the effect size of heterotypic mutant combinations relative to the WT Cx34.7 and Cx35 pair. The purple circle provides the reference point for an effect size of zero. The green intermittent circle corresponds to the Cx34.7 mutations identified by green labels in the outermost level around the rim of the plot. The black horizontal line in the centre is the scale bar for effect size. For n values and statistical tests, see main text. For definitions of box plots, see Methods.Neural circuit editing using gap junctions is well established in C. elegans7,8,9,10. C. elegans do not express connexins; thus, heterologous expression of the vertebrate connexin 36 (Cx36) across two connected C. elegans neurons results in the formation of an electrical synapse that does not interact with endogenous C. elegans gap junction proteins (innexins). Previous work has successfully implemented this editing approach to modify circuit physiology in multiple behavioural contexts, including C. elegans migration in response to various chemical and temperature conditions8,9,10,11.The potential for using gap junctions to repair dysfunctional circuits has also been advanced in C. elegans, as shown in experiments that used circuit editing to restore normal behaviour in animals with induced circuit disruptions11,12.

Nevertheless, this previous work highlighted a significant challenge in the use of gap junctions to edit select circuits in higher-complexity organisms. Specifically, when Cx36 was expressed in two sensory neurons of the same cell type and formed homotypic gap junctions, otherwise normal C. elegans showed disrupted behaviour in response to olfactory cues11. Because vertebrate brains are composed of many more cells of the same cell type than C. elegans, the ability of connexins to form homotypic gap junctions across more cells has the potential to substantially reduce the precision of this circuit-editing approach for mammals (for example, off-target modulation; Fig. 1a, left), which in turn produces greater behavioural disruption. Moreover, heterologous expression of connexins from other species in mammals might lead to gap junctions composed of both exogenous and endogenous connexins, thereby producing undesired connections that may impair neural circuit function.Although nearly all connexins can form homotypic channels, several connexin isoforms can dock with other connexin isoforms to generate heterotypic channels4,13 (Fig. 1a, right). We reasoned that by identifying the mechanisms that underlie the docking interactions between connexins14, we could design a hemichannel pair biased towards heterotypic gap junction formation. We also reasoned that we could engineer this pair so that it is docking-incompatible with connexins endogenous to the mammalian CNS. This strategy can therefore produce a precise approach for regulating electrical flow between distinct cell types.M. americana (white perch fish) expresses two homologues of mammalian neuronal Cx36—connexin 34.7 (Cx34.7) and connexin 35 (Cx35)—that create a heterotypic gap junction1. Notably, this electrical synapse exhibits channel-level rectification in the Cx34.7 to Cx35 direction when expressed in Xenopus oocytes1. The orthologues of Cx34.7 and Cx35 in the goldfish (Carassius auratus) CNS also create a heterotypic gap junction that shows circuit-level rectification in the same direction15. In summary, Cx34.7 and Cx35 can form heterotypic gap junctions with inherent directionality15 and can conduct currents capable of triggering action potentials1. Moreover, these connexins are potentially amenable to modification of biophysical properties through amino acid sequence mutations.

On the basis of these findings, here we use Cx34.7 and Cx35 to engineer a new electrical synapse for editing mammalian circuits.In vitro assay of connexin hemichannel dockingTo establish a method for evaluating connexin docking specificity and to ultimately engineer our electrical synapse, we leveraged the natural cellular turnover of docked connexin hemichannels. In mammalian cells, connexin hemichannels can be removed from the membrane through a coordinated endocytic and exocytic process that results in the internalization of fully docked gap junctions in double-bilayer vesicles called connexosomes16,17,18,19,20,21 (Fig. 1b, top left). By fusing connexin monomers with a fluorescent protein tag, internalization of labelled gap junctions from a fluorescent cell to a non-fluorescent cell can be visualized21,22 (Fig. 1b, top).Our approach used separate populations of HEK293FT cells that were transiently transfected with individual connexins as either mEmerald or iRFP670 fluorescent fusion proteins (Fig. 1b). We then co-plated and incubated the HEK293FT cells that expressed connexin counterparts. Finally, we evaluated their fluorescence exchange by flow cytometry (Extended Data Fig. 1a–d). As hemichannel docking is a prerequisite for internalizing fluorescently tagged connexins expressed by other cells, docking can be quantified as the proportion of transfected cells that are labelled by dual fluorescence in the co-plated sample (Fig. 1b,c). We first established the utility of our assay (termed flow-enabled tracking of connexosomes in HEK293FT cells (FETCH)) by testing well-characterized connexin isoforms: Cx26, Cx36, Cx43 and Cx45. Given that each of these is capable of homotypic docking21,22,23,24, we tested them under homotypic pairings (FETCH mean ± s.e.m.

= 24.8 ± 1.8%, 15.2 ± 1.1%, 19.5 ± 0.4% and 14.4 ± 0.5% dual-labelled cells for Cx26–Cx26, Cx36–Cx36, Cx43–Cx43 and Cx45–Cx45, respectively; Fig. 1c, left). We also tested them in paired combinations for which there was previous evidence13 of heterotypic docking-incompatibility (for example, Cx26–Cx43, Cx36–Cx43 and Cx36–Cx45, which had FETCH values of 2.5 ± 0.1%, 0.8 ± 0.1% and 0.9 ± 0.1%, respectively; Fig. 1c, left). Notably, the proportion of dual-labelled cells in the population of docking-compatible versus docking-incompatible pairs was significantly different (t40 = 14.5, P = 1.6 × 10−17, two-tailed unpaired t-test). These results establish FETCH as a method that can be used to broadly assess connexin hemichannel docking compatibility.Second, we analysed the utility of FETCH by testing two connexin mutations that affect gap junction formation. Specifically, we tested a Cx43(F199L) mutant that has previously been shown to disrupt trafficking to the cell membrane25. We also evaluated a Cx26(K168V N176H) mutant that confers heterotypic docking compatibility with wild-type (WT) Cx43 (ref. 26). In both cases, we tested whether the FETCH score for the mutants exceeded the scores for cell pairs under conditions in which docking was not anticipated (FETCH = 1.5 ± 0.2% for this ‘known negative’ distribution; n = 92 new cell pairs; Fig. 1c right, blue lines, and Methods).

The homotypically paired Cx43(F199L) mutant did not exhibit a level of fluorescence exchange that was higher than the known-negative distribution (FETCH = 2.1 ± 0.3%, t96 = 0.85, P = 0.20, one-tailed unpaired t-test). By contrast, the Cx26(K168V N176H) mutant that heterotypically paired with Cx43 showed fluorescence exchange that was significantly increased (FETCH = 26.6 ± 1.3%, t96 = 30.9, P = 5.2 × 10−52, unpaired one-tailed t-test; Fig. 1c, right). Thus, we established that our FETCH assay can be used to identify connexin mutations that disrupt or enable docking compatibility.Cx34.7 and Cx35 mutant hemichannel dockingWe used FETCH to assay a library of Cx34.7 and Cx35 mutants for their impact on hemichannel docking. Although the precise interactions that guide hemichannel docking are incompletely characterized for most connexins, structure–function and sequence analyses indicate that both the extracellular loops (EL1 and EL2) play a part in hemichannel docking14,27,28,29. To identify Cx34.7 and Cx35 variants that are unable to form homotypic gap junctions, we introduced around 70 individual mutations at 16 positions on both EL1 and EL2 of each connexin (Fig. 1d, Methods (for the design of the library of mutants) and Supplementary Fig. 1). We then compared the homotypic pairing FETCH scores of the mutants to a docking-incompatible heterotypic pair30 (for example, Cx36 paired with Cx45) (Fig.