Rewiring neural circuits by the insertion of ectopic electrical synapses in transgenic C. elegans

Neural circuits are functional ensembles of neurons that are selectively interconnected by chemical or electrical synapses. Here we describe a synthetic biology approach to the study of neural circuits, whereby new electrical synapses can be introduced in novel sites in the neuronal circuitry to reprogram behaviour. We added electrical synapses composed of the vertebrate gap junction protein Cx36 between Caenorhabditis elegans chemosensory neurons with opposite intrinsic responses to salt. Connecting these neurons by an ectopic electrical synapse led to a loss of lateral asymmetry and altered chemotaxis behaviour. In a second example, introducing Cx36 into an inhibitory chemical synapse between an olfactory receptor neuron and an interneuron changed the sign of the connection from negative to positive, and abolished the animal’s behavioural response to benzaldehyde. These data demonstrate a synthetic strategy to rewire behavioural circuits by engineering synaptic connectivity in C. elegans.

A central aim of neuroscience research is to reveal the function of neural circuits and how changes in synaptic connectivity modify behaviour. Recent advances in synthetic biology, whereby genetic networks of living cells have been rewired to produce alternative cellular function [1][2][3] , may revolutionize the field of cell biology. Synthetic approaches for re-engineering neural function and behaviour have also begun to be developed for neuroscience; in particular, optogenetic and other protein-based reagents have been used to manipulate the activity patterns of defined neurons in vivo 4,5 . In principle, a synthetic neurobiology approach could also be used to manipulate neural circuits by artificially modifying specific synaptic connections in the circuit and thus reprogramming behaviour.
To artificially modify the connection between neurons we sought to introduce a new synapse into an existing neural circuit. Inserting heterologous chemical synapses into a circuit would be difficult owing to their enormous complexity and hundreds of constituent proteins 6,7 . In contrast, electrical synapses, or gap junctions, consist of as little as one protein type 8 , and it has been shown that overexpression of an endogenous gap junction protein can alter neuronal connectivity in vitro 9 . The molecular constituents of vertebrate and invertebrate gap junctions belong to distinct protein families: connexins are exclusive to vertebrates and innexins to invertebrates [10][11][12] . Importantly, while proteins from the same family can interact to form heterotypic gap junctions 13,14 , no compatibility has been found between connexins and innexins 15 . Thus, we reasoned that connexins expressed heterologously in Caenorhabditis elegans neurons would form gap junctions exclusively between themselves, and that these new connections could be used to introduce novel connections between normally unconnected neurons in intact animals. Here we present data assessing the use of this approach in two neural circuits, involved in salt taste and olfactory chemotaxis.

Results
Adding synthetic electrical synapses between sensory neurons. We first investigated the possibility of rewiring neuronal connections in vivo using the neural circuit involved in salt chemotaxis. C. elegans navigates up gradients of sodium chloride and other salts in part through a biased random walk 16 , combining forward travel with stochastically occurring reorienting turnings and reversals. The primary neurons sensing attractive concentrations of salt are the bilaterally symmetric ASE neurons 17 ; ASEL is an ON cell that is stimulated by an increase in salt concentration, while ASER is an OFF cell stimulated by a salt concentration decrease 18 . Unlike most bilateral sensory neuron pairs in C. elegans, the ASEL and ASER neurons are not normally connected by gap junctions (Fig. 1a). Published connectome data likewise report no chemical synapses between ASEL and ASER 19,20 , although more recent online data based on computer-aided reconstructions 21,22 suggest some may exist. Moreover, confocal imaging of ASER and ASEL, labelled with expressed fluorescent proteins of different colours, showed that the processes of these two neurons are physically adjacent within the nerve ring (Fig. 1b).
We therefore reasoned that heterologous connexin expression could be used to synthetically connect the ASEL and ASER neurons with electrical synapses. We generated a synthetic connexinderived transgene consisting of a C. elegans codon-optimized 23 , fluorescently tagged version of the mouse connexin 36 (Cx36), a well-characterized, broadly expressed gap junction constituent 24 , under the control of promoters specific for either ASER (gcy-5) or ASEL (gcy-7). In multicopy extrachromosomal lines overexpressing Cx36 in both neurons (ASER Cx36 ASEL Cx36 ), as well as a line expressing Cx36 in ASER alone (ASER Cx36 ), this tagged Cx36 protein localized throughout the dendrite, axon and cell body in a punctate pattern (Fig. 1c). In some but not all cases, ASER Cx36 puncta were coincident to ASEL Cx36 puncta, suggesting that they might represent synthetic gap junctions ( Supplementary Fig. 1).
To test whether Cx36 expression could indeed generate ectopic gap junctions linking ASEL and ASER, we performed calciumimaging experiments to investigate the effects of ASER Cx36 and ASEL Cx36 on neuronal responses to chemosensory stimulation. Although genetically encoded calcium indicators lack the temporal sensitivity to detect rapid electrical potentials, they are well-suited to detect changes in the slow, graded responses to NaCl expected from the introduction of electrical coupling between ASEL and ASER. In wild-type animals, a step change from 40 to 0 mM NaCl evoked a calcium increase in ASER but no response in ASEL (Fig. 1d, blue). Expressing Cx36 in ASEL or ASER alone did not significantly alter either response (Fig. 1f,g). In contrast, in ASER Cx36 ASEL Cx36 animals both ASEL and ASER showed robust calcium increases in response to the downstep stimulus (Fig. 1d, red), as predicted if the cells were electrically coupled. We also assayed the responses of Cx36-expressing strains to increases in NaCl concentration. We observed that in wild-type, ASEL Cx36 and ASER Cx36 animals, salt upsteps led to a calcium increase in ASEL and a calcium decrease in ASER ( Fig. 1e,h,i). However, in ASER Cx36 ASEL Cx36 animals ( Fig. 1e, red), ASER showed a calcium increase in response to a salt upstep, as expected if the depolarizing current in ASEL depolarized ASER through an electrical connetion. These alterations in ASE activity correlated with a significant defect in the ability of the animals to navigate in a point-source gradient of NaCl ( Supplementary Fig. 2a). Together, these results indicate that functional electrical synapses were successfully inserted between ASEL and ASER.
Flipping the sign of an inhibitory synapse. We next tested whether existing neuronal connections could be reconfigured through ectopic Cx36 expression. To do this we focused on the AWC and AIY neurons, which mediate chemotaxis towards attractive odourants, such as benzaldehyde 25,26 . The AWC sensory neurons respond to increases in odour concentration with reduced activity and to decreases in odour concentration with elevated activity. The AWCs make inhibitory chemical synapses with the AIY interneurons (Fig. 2a); although AIY ablation does not impair chemotaxis to odourants 27 , optogenetic AIY activation has been shown to promote forward travel and inhibit reorientations 28 . To determine whether heterologous connexin expression could modify the properties of these synapses and thereby impact behaviour, we expressed Cx36 in AWC and AIY using promoters derived from odr-1 (for AWC Cx36 ) and the ttx-3 second intron (for AIY Cx36 ) and imaged calcium responses to olfactory stimuli using genetically encoded calcium indicators (Fig. 2b,c; Supplementary Fig. 3). In wild-type animals, benzaldehyde presentation led to a calcium decrease in AWC and a reciprocal calcium increase in AIY, as expected from their known inhibitory connection (Fig. 2d). In animals expressing either the AWC Cx36 or the AIY Cx36 transgenes alone, these responses were not significantly altered (Fig. 2f,g). In contrast, animals expressing Cx36 in both neurons (AWC Cx36 AIY Cx36 ) responded to benzaldehyde presentation with a calcium decrease in both AWC and AIY, suggesting that ectopic gap junctions between AWC and AIY reversed the sign of the AWC-AIY connection from negative to positive (Fig. 2d, red). A similar effect was observed when the AWC Cx36 AIY Cx36 transgenes were carried on a chromosomally integrated array ( Supplementary Fig. 4). We also observed an attenuation in the degree of AWC hyperpolarization in the AWC Cx36 AIY Cx36 worms, consistent with a shunting of the hyperpolarizing current to AIY (Fig. 2d, red). Together these results indicate that ectopic gap junctions were formed between AWC and AIY, switching an inhibitory chemical synapse to a predomininantly excitatory electrical/chemical synapse.
We also analysed the effect of ectopic connexin expression on responses to benzaldehyde removal. As expected, in wild-type worms this led to a calcium increase in AWC and a small calcium decrease in AIY, consistent with an inhibitory chemical synapse (Fig. 2e). In contrast, AWC Cx36 AIY Cx36 worms showed a calcium increase in AIY, an attenuated calcium increase in AWC (Fig. 2e, red) and a strong defect in benzaldehyde chemotaxis ( Supplementary Fig. 2b), consistent with the presence of an ectopic electrical connection. We also assayed AWC Cx36 and ARTICLE AIY Cx36 strains individually. We observed that AIY Cx36 worms showed normal responses to benzaldehyde removal and normal benzaldehyde chemotaxis, whereas AWC Cx36 worms showed reduced calcium influx and a partial deficiency in chemotaxis (Fig. 2h,i; Supplementary Fig. 2b). Since the left and right AWCs, unlike most amphid neurons, are not normally linked by gap junctions, we reasoned that AWC Cx36 expression might alter the neurons' electrical properties by electrically coupling the AWC pair. Consistent with this possibility, when we expressed Cx36 in only one of the two AWCs, (AWC off ) using the srsx-3 promoter 29 , we observed wild-type calcium responses and normal chemotaxis ( Supplementary Fig. 5  AWCL. Alternatively, although our odr-1 promoter fragment appears AWC-specific in our lines, if the odr-1::Cx36 transgene is driving low-level Cx36 expression in other neurons, these might form ectopic gap junctions with the AWCs and thereby attenuate AWC responses by shunting.

Discussion
In summary, we have shown using two different neural circuits in C. elegans that vertebrate connexins can be used to create synthetic electrical synapses in an intact nervous system. By transgenically targeting connexin expression to physically adjacent neurons, this manipulation can be used as a general technique for artificially modifying neural connectivity. This approach complements several existing methods for manipulating the excitability 30 or efficacy of synaptic transmission 31,32 from genetically targeted neurons. Ectopic connexin expression makes it possible not only to generate new connections between unconnected neurons, but also to modify the strength of existing electrical synapses 33 . Thus, synthetic modification of neural connections should prove to be a useful tool to obtain insight into the workings of neural circuits in genetically tractable animals, in which the precise connectivity and the relative position of each circuit component is known.
The approach described here takes advantage of the fact that vertebrate and invertebrate gap junctions are formed from distinct, most likely incompatible protein families. Thus, an ectoptic connexin hemichannel is not expected to pair with endogenous innexin hemichannels in C. elegans neurons; indeed, the observation that AWC Cx36 does not on its own lead to electrical coupling to AIY (which express the endogenous innexin inx-18 (ref. 34)) supports this conjecture. Although this would appear to limit the exact method described here to invertebrate nervous systems, it has been found that invertebrate innexins can readily be expressed to form hemichannels in vertebrate Xenopus oocytes 14 . It is therefore feasible that a mirror image of the method we describe here (that is, ectopic expression of innexins in vertebrates) might be used to modify the connectivity of mammalian brain circuits. Synthetic synaptic modification approaches of this sort have a broad range of potential applications in neuroscience research, including the implementation of theoretical neural circuit models in real live circuits, the ability to test the sufficiency (as opposed to necessity) of particular connections for neuronal circuit function and even the engineering of novel circuits and behaviours.
Visualizing Cx36 expression in transgenic animals. The complementary DNA sequence of Mus musculus gap junction protein, delta 2 (Gjd2) was codon optimized 23 to produce a synthetic Cx36 gene indicated in the strain list as Cx36* (GeneArt). We fused to Cx36 an upstream promoter (gcy-5 for ASER 35 , gcy-7 for ASEL 35 , odr-1 for AWC 25 , srsx-3 for AWC off29 and the second intron of ttx-3 for AIY 36 ) and a downstream gene encoding a fluorescent protein (YFP or mCherry). Plasmids were injected at 70 ng ml À 1 each, together with a coinjection marker (unc-122::gfp, unc-122::mCherry or elt-2::mCherry at 40, 40 or 20 ng ml À 1 , respectively) to generate transgenic worms carrying an extrachromosomal array. To express an electrical synapse linking two neurons, we co-injected plasmids for both neurons in the same worm. Transgenic Cx36 expression was visualized using an LSM 510 or LSM 780 confocal laser scanning microscope (Zeiss), captured and processed by Zen (Zeiss) software and further processed using Volocity, Fiji and Adobe Illustrator CS3. After scanning a stack of B1-mm-thick slices, either a maximum projection image or a three-dimensional rendering was generated and uniformly adjusted for brightness and contrast. Strains for imaging the overlap between the ASE neurons were prepared by co-injecting gcy-7::gfp and gcy-5::mCherry at 85 ng ml À 1 each, with 40 ng ml À 1 of unc-122::mCherry. These images were acquired and processed as mentioned above. For the deconvolution images in Supplementary Fig. 1, the samples were prepared by immobilizing the worms on a 3% pad of agarose using 200 mM of sodium azide. The worms were imaged using API Deltavision with a 100 Â objective and processed using PerkinElmer Volocity.
Chemotaxis assays. Salt chemotaxis assays were performed essentially as described 37 . Briefly, we poured 10 ml of buffered agar (2% agar, 1 mM CaCl2, 1 mM MgSO4, 5 mM KPO4) into 10-cm diameter Petri dishes. To form the gradient, 16 h before the assay we applied 10 ml of 250 mM NaCl salt solution adjusted to pH 6.0 to the attractant spot and 10 ml of double-distilled H 2 O to the control spot. Four hours before the assay, we added another 4 ml of salt solution or double-distilled H 2 O to the same spots. Ten minutes before the assay, we applied 1 ml of 1 M sodium azide to both attractant and control spots to immobilize the worms once they reached these spots. We used synchronized animals that were grown on 6-cm nematode growth media plates with food, washed three times with CTX solution (1 mM CaCl2, 1 mM MgSO4, 5 mM KPO4) and pipetted onto a nylon membrane to dry. Subsequently B200 worms were placed in the center of the assay plate inside a circle of diameter of 1.5 cm. We allowed the animals to move about the agar surface for 1 h. The chemotaxis index was calculated as the number of animals contained in a circle of radius 1.5 cm around the salt spot minus the number of worms in a circle of radius 1.5 cm around the control spot, divided by the total number of animals in the two circles. Benzaldehyde chemotaxis assays were performed as described 38 , except that they were performed in smaller 6-cm plates, and with N ¼ 20 worms, which were picked and transferred to the assay plate after being released from food on an empty plate. A drop of 1% benzaldehyde in ethanol and ethanol alone were placed at two opposite ends of the assay plate together with 1 mM sodium azide for trapping the worms. Chemotaxis was scored the following day by counting the number of paralyzed worms near the odour spot N(odour) and the number of paralyzed worms near the control spot N(control), and calculating the chemotaxis index 38 equal to: (N(odour) À N(control))/(N(odour) þ N(control)).
Calcium imaging. Imaging was performed on a Zeiss Axioscope upright microscope using an Andor iXon EM camera or a Hamamatsu ORCA-ER camera in combination with neutral density filters to reduce the intensity of ultraviolet illumination. Movies were captured by IQ1.9 software (Andor) or a custom-written Matlab program and analysed using another custom-written Matlab (Mathworks) script. A rectangular region of interest (ROI) was drawn surrounding the cell body (AWC and ASE) or the neurite (AIY), and for every frame the ROI was shifted according to the new position of the center of mass (in the case of the cell body) or point of maximum intensity (for the AIY neural processes). The fluorescence intensity, F, was computed as the difference between the sum of pixel intensities and the faintest 10% pixels (background) within the ROI. For ratiometric imaging, ROI Y tracked the neuron in the yellow channel, and in the cyan channel, ROI C moved at a fixed offset from ROI Y . F was computed as F Y /F C after correcting for bleed through 37 . No correction for bleaching was necessary. DF for calcium traces was equal to (F À F 0 )/F 0 Â 100, where F 0 equals the average F within the first 3 s of recording. For statistical quantification DF was computed as (F 1 À F 0 )/F 0 Â 100, where F 0 is the average F over 10 s before odour switching and F 1 is the average over 10 s after odour switching. All imaging strains performed normally in chemotaxis assays. A two-tailed unpaired t-test was used to analyse the data. Where 41 comparison was made, a Bonferroni t-test was used instead.
To image salt responses, single animals were placed in a perfusion chamber (RC-26GLP,Warner Instruments) under a constant flow rate (0.3 ml min À 1 ) of CTX buffer using a perfusion pencil (AutoMate). Outflow was regulated using a peristaltic pump (Econo Pump, Bio-Rad). Upsteps or downsteps in NaCl concentration were delivered using the perfusion pencil, and switch between control and stimulus solutions was done using manually controlled valves. Solutions contained the indicated amount of NaCl and in addition 1 mM MgSO 4 , 1 mM CaCl 2 and 5 mM KPO 4 . The stimulus was delivered for 10 s starting on the 10th second from the onset of the movie.
To image olfactory responses, worms were inserted into a microfluidic polydimethylsiloxane chip, a generous gift from N. Chronis, designed to deliver odour stimuli under a fluorescent microscope 39 . Ten seconds after the beginning of each recording, we presented the odour (10 À 4 benzaldehyde diluted in S-basal lacking cholesterol) and continued imaging for an additional 30 s. We kept the odour channel open for 5 min and then resumed imaging. After an additional 10 s, we switched the odour off and completed the recording following an additional 30 s. We used buffer instead of fluorescein in the chip to enhance the signal to noise ratio and eliminate switching artifacts, and unlike previous reports 25, 26 we did not pre-immerse the worms in buffer before each recording.