Low Conductance Gap Junctions Mediate Specific Electrical Coupling in Body-wall Muscle Cells of Caenorhabditis elegans*

Invertebrate innexins and their mammalian homologues, the pannexins, are gap junction proteins. Although a large number of such proteins have been identified, few of the gap junctions that they form have been characterized to provide combined information of biophysical properties, coupling pattern, and molecular compositions. We adapted the dual whole cell voltage clamp technique to in situ analysis of electrical coupling in Caenorhabditis elegans body-wall muscle. We found that body-wall muscle cells were electrically coupled in a highly organized and specific pattern. The coupling was characterized by small (350 pS or less) junctional conductance (Gj), which showed a bell-shaped relationship with junctional potential (Vj) but was independent of membrane potential (Vm). Injection of currents comparable to the junctional current (Ij) into body-wall muscle cells caused significant depolarization, suggesting important functional relevance. The innexin UNC-9 appeared to be a key component of the gap junctions. Both Myc- and green fluorescent protein-tagged UNC-9 was localized to muscle intercellular junctions. Gj was greatly inhibited in unc-9(fc16), a putative null mutant. Specific inhibition of UNC-9 function in muscle cells reduced locomotion velocity. Despite UNC-9 expression in both motor neurons and body-wall muscle cells, analyses of miniature and evoked postsynaptic currents in the unc-9 mutant showed normal neuromuscular transmission. These analyses provide a relatively detailed description of innexin-based gap junctions in a native tissue and suggest that innexin-based small conductance gap junctions can play an important role in processes such as locomotion.

Gap junctions are intercellular channels that allow the passage of ions and small molecules. They play important roles in physiology and development. Mutations of gap junction proteins cause a variety of diseases, such as demyelination, deafness, female infertility, cataracts, cardiac arrhythmias or malformation, and embryonic lethality (1)(2)(3). Gap junctions may be formed by connexins, innexins, and pannexins (4,5). Pannexins are newly discovered in mammals in a search for invertebrate innexin homologues. Both the human and the mouse genomes contain three pannexin genes (Px1, Px2, and Px3) (6,7). All of the three pannexins are expressed at different levels in the brain. PX1 is also expressed in many other tissues (7,8). PX1 has been shown to form intercellular channels (7) and hemichannels (9) when expressed in Xenopus oocytes. However, it has not been confirmed that mammalian pannexins form gap junctions in native tissues.
Properties and functions of connexin-based gap junctions are relatively well understood. However, physiological functions and in vivo biophysical properties of mammalian pannexin-based gap junctions are unknown. The anatomical complexity of the brain and the possible co-existence of connexin-based gap junctions make it very difficult to study properties of pannexin-based gap junctions in vivo or in situ. To gain understanding of pannexin-based gap junctions, an alternative approach is to study gap junctions in invertebrate systems. Essentially all identified invertebrate gap junction proteins are innexins (10). Indeed, invertebrate systems have been the primary source for existing knowledge of innexin-and pannexin-based gap junctions. For example, analyses of the Drosophila shak-B and Caenorhabditis elegans eat-5 mutants established the connection between innexins and gap junctions (11,12). Drosophila Shaking-B protein was the first innexin confirmed to form gap junctions on heterologous expression in Xenopus oocytes (13). Human and mouse pannexins were discovered by data base search for invertebrate innexin homologues (6,7). Information about biophysical or functional properties of innexin-based gap junctions have been obtained through electrophysiological analyses in several invertebrate preparations, such as cultured mosquito cells (14), mechanically dissected cell pairs of Drosophila salivary glands (15), isolated leech ganglia (16), dissected body-wall muscle cells of Ascaris lumbricoides (17)(18)(19), and neurons of Aplysia californica (20) and earthworm (21). Besides Drosophila Shaking-B protein (13), at least two other invertebrate innexins have been shown to form gap junctions when expressed in paired Xenopus oocytes (22,23).
To elucidate the function of innexins/pannexins, it is critical to know the coupling pattern, biophysical and functional properties, and molecular compositions of the gap junctions that they form. However, most experimental systems do not allow combined analyses of these properties due to complex cellular organization, lack of gap junction isoformspecific inhibitors, and difficulty of molecular genetic manipulation. C. elegans body-wall muscle is potentially suitable for such analyses for several reasons: 1) C. elegans has only 95 body-wall muscle cells, which are organized in a regular pattern into four longitudinal strips called quadrants (24); 2) gap junction-like structures have been identified in the muscle cells through ultrastructural analyses (25,26); and 3) innexin expression level in the muscle may be manipulated conveniently through genetic means in C. elegans. Detailed analyses of electrical coupling in this specific system may help us to understand general properties of innexin/pannexin-based gap junctions.
The C. elegans genome contains 25 innexin genes (27). Mutants of several of them, including inx-3, eat-5, inx-6, unc-7, and unc-9, have been analyzed. INX-3 is essential to embryonic morphogenesis (28). EAT-5 and INX-6 are expressed in the pharynx and are required for synchronized pharyngeal muscle contractions (11,29,30). Mutants of unc-7 and unc-9 share several phenotypes, including locomotion defects (31)(32)(33), suppression of the enhanced sensitivity to volatile anesthetics caused by mutations of unc-79 or unc-80 (34), and resistance to the anthelminthic ivermectin (35,36). It is unclear how mutations of UNC-7 and UNC-9 cause the pleiotropic phenotypes. Although the in vivo expression patterns of at least 17 innexins have been evaluated using specific antibodies or translational GFP 3 fusions, none of them appear to be expressed in body-wall muscle cells (37). Thus, it is yet to be confirmed whether any of the 25 innexins are expressed in C. elegans body-wall muscle cells to constitute the morphologically defined gap junctions (25,26).
We adapted the dual whole cell voltage clamp technique to in situ analyses of electrical coupling in C. elegans body-wall muscle cells. We found that low conductance gap junctions mediate cell-specific electrical coupling among body-wall muscle cells, the innexin UNC-9 is a major component of the gap junctions, and the electrical coupling is important to locomotion.

EXPERIMENTAL PROCEDURES
Electrophysiology-C. elegans was grown at room temperature (21-23°C) on agar plates with a layer of OP50 Escherichia coli (38). Adult hermaphrodites were used for physiological analysis. Body-wall muscle cells in the two ventral quadrants were exposed as previously described (39). Briefly, an animal was immobilized on a glass coverslip by applying a cyanoacrylate adhesive along the dorsal side. A longitudinal incision was made in the dorsolateral region. After clearing the viscera, the cuticle flap was folded back and glued to the coverslip, exposing the ventral nerve cord and the two adjacent muscle quadrants. A Nikon Eclipse E600FN microscope equipped with a 40ϫ water-immersion lens and 15ϫ eyepieces was used for viewing the preparation. The classic whole cell configuration was obtained with borosilicate glass electrodes (3-5 M⍀), and was used for both voltage and current clamp experiments. Seal resistance was in the range of 10 -20 G⍀. Membrane capacitance, membrane resistance, and access resistance were determined with Clampex (Axon Instruments) by applying a 20-mV depolarizing pulse from a holding potential of Ϫ60 mV. In 80 randomly picked wild-type body-wall muscle cells, membrane capacitance, membrane resistance, and series resistance (R s ) were 29.6 Ϯ 0.4 picofarads, 1.51 Ϯ 0.04 G⍀, and 9.6 Ϯ 0.4 M⍀, respectively. Clampex software was used to acquire data at a rate of 10 kHz after filtering at 2 kHz.
The dual whole cell voltage clamp technique was used to record I j . Body-wall muscle cells have high membrane resistance and low G j (see "Results"), making them suitable for analyses with this technique (5). To record I j , two neighboring body-wall muscle cells were voltage clamped. In most experiments, both cells were held at Ϫ30 mV, from which a series of voltage steps (Ϫ120 to ϩ80 mV at 10-mV intervals and 100-ms duration) was applied to one cell (cell #1), whereas the other cell (cell #2) was held constant to record I j . In one set of experiments (Fig. 2D), the holding potential was Ϫ60 mV, from which voltage steps of Ϫ60 to ϩ90 mV (100-ms duration, in 10-mV increments) were applied to cell #1, whereas cell #2 was held constant to record I j . The voltage steps were repeated four times and averaged. This protocol was alternated between the two cells in each pair.
To determine if G j is dependent on V m , pairs of cells were held at Ϫ70 to ϩ30 mV, from which 8 voltage steps (at 20-mV intervals, 4 steps on either side of the holding potential) were applied to one cell while the other cell was held constant to record I j . The sequence to apply the different holding potentials was randomized to avoid potential time-dependent effects.
Standard pipette solution contained (in mM) KCl 120, KOH 20, Tris 5, CaCl 2 0.25, MgCl 2 4, sucrose 36, EGTA 5, and Na 2 ATP 4, pH 7.2. Extracellular solution included (in mM) NaCl 140, KCl 5, CaCl 2 5, MgCl 2 5, dextrose 11, and HEPES 5, pH 7.2. The junctional potential between the two solutions, estimated to be ϳ6 mV using the Clampex software, did not affect the analyses, because the pipette offset was zeroed immediately before getting a gigohm seal and because the extracellular solution was not changed during the period of the experiment.
mPSCs and ePSCs were recorded as previously described with the muscle cell held at Ϫ60 mV (40). The standard pipette and extracellular solutions were used.
Dye Coupling Assay-5(6)-Carboxyfluorescein (Sigma) was applied through a whole cell patch clamp electrode. It was included in the pipette solution at a concentration of 100 M (pH 7.2). The recording chamber was perfused briefly with standard extracellular solution to remove any dye that had leaked into the solution from the pipette. The muscle preparation was photographed for epifluorescence ϳ15 min later.
RNA Interference-A 0.6-kb genomic DNA fragment of unc-9 was amplified by PCR with a pair of specific primers (sense: CCG-GAAATCGTCTGAAACTG; antisense: AGTGATTGCGAATCCAC-CAT). The amplified DNA fragment was cloned into the L4440 vector and transfected into E. coli to make unc-9-specific double-stranded RNA. RNAi was achieved by feeding animals with the bacteria expressing the unc-9 double-stranded RNA (41). In control experiments determining the effect of unc-9 RNAi on locomotion, wild-type animals were fed with bacteria transfected with the L4440 vector (without the unc-9 insert). The effectiveness of RNAi was evaluated by examining GFP epifluorescence in transgenic animals expressing UNC-9::GFP fusion protein under the independent control of the neuron-specific rab-3 promoter (Prab-3) (42) and the muscle-specific myo-3 promoter (Pmyo-3) (43).
Evaluation of UNC-9 Expression Pattern and Subcellular Localization-Cells expressing UNC-9 were identified by expressing GFP or DsRed2 in vivo under the control of unc-9 promoter (Punc-9). The unc-9 gene has an unusually large first intron (ϳ10 kb). Because this intron is potentially important for cell-specific expression but difficult to be included in a typical plasmid construct, we used an alternative 3 The abbreviations used are: GFP, green fluorescent protein; G j , junctional conductance; I j , junctional currents; V j , junctional voltage; V m , membrane voltage; Px (or PX for the protein), pannexin; R s , series resistance; mPSC, miniature postsynaptic current; ePSC, evoked postsynaptic current; WT, wild-type; Punc-9, unc-9 promoter; Pmyo-3, myo-3 promoter; Prab-3, rab-3 promoter; G jss , the steady-state conductance; A, the cooperativity constant; V o , the voltage at which the decrease in G jss is half-maximal; G jmin , minimum conductance; TES, 2-{[2-hydroxy-1,1-bis(hydroxymethyl)ethyl]amino}ethanesulfonic acid (systematic); RNAi, RNA interference. approach. A 1.6-kb unc-9 genomic fragment immediately upstream of the initiation site (in the second exon) was cloned and fused in-frame to GFP or DsRed2. This plasmid (wp218 for GFP and wp268 for DsRed2) was injected into unc-9(fc16) together with cosmids R12H7 and F09D5. Cosmid R12H7 contains the 3Ј portion of the first intron and the entire coding region of unc-9, whereas cosmid F09D5 contains the entire Punc-9 and the 5Ј portion of the first intron. These two cosmids have significant overlap (Ͼ3 kb) in the first intron. Homologous recombination in vivo would result in a full-length Punc-9::gfp or Punc-9::DsRed2 transcriptional fusion (40,44,45). To confirm UNC-9 (DsRed2) expression in body-wall muscle cells and neurons, GFP was co-expressed either in neurons using Prab-3 (plasmid wp70) (42) or in muscle cells using Pmyo-3 (plasmid wp124) (43). To determine subcellular localization, two UNC-9 fusion proteins were expressed independently in body-wall muscle cells under the control of Pmyo-3. In one fusion protein, Myc was fused to the amino terminus of UNC-9 (Pmyo-3::myc::unc-9 cDNA, wp390). In the other fusion protein, GFP was fused to the carboxyl terminus of UNC-9 (Pmyo-3::unc-9 cDNA::gfp, wp206). The expression and localization patterns of DsRed2 and GFP were evaluated by examining epifluorescence of the fluorescent proteins in transgenic animals. The localization pattern of Myc::UNC-9 was determined by immunohistochemistry in whole mount animals following a published protocol (42). Rabbit anti-Myc (Sigma) and Alexa Fluor 594-conjugated goat anti-rabbit (Invitrogen) were used as the primary and secondary antibodies, respectively. The epifluorescence was visualized and photographed with a Nikon TE2000-U inverted microscope connected to a cooled monochrome digital camera (F-view II, Soft Imaging System GmbH, Germany). A 40ϫ objective, and fluorescein isothiocyanate and Texas Red filters (Chroma Technology Corp., Rockingham, VT) were used for fluorescence imaging. Measurement of Locomotion Velocity-Locomotion assays were performed at room temperature on healthy adult hermaphrodites. A piece of paper with a circular hole (7 mm in diameter) was laid on top of the bacterial lawn in the culture plate to restrict animal movements within the field of observation. A single animal was placed in the center of the field to start the assay. The plate was left undisturbed for 7 min while it was photographed at 2-s intervals with the F-view II digital camera mounted on a Nikon SMZ800 stereomicroscope. The camera was operated through analySIS Imager software (Soft Imaging System GmbH). The distance traveled by the animal from the beginning of the third minute to the end of the seventh minute was determined by measuring the length of locomotion tracks left behind by the animal using the "Polygon Length" measurement function of analySIS. Locomotion speed was calculated by dividing the distance with the time.
Statistics and Graphing-To compensate the effect of R s on clamp voltage (V clamp ), the actual membrane potential (V actual ) was calculated using the formula, V actual ϭ V clamp Ϫ I m ϫ R s , where I m is the whole cell membrane current. The mean current between the 50th and 100th ms at each V j step was determined with Clampfit (Axon Instruments) after low pass filtering at 300 Hz and is considered as steady-state I j , although it may not be truly steady state in some traces. For selected groups, instantaneous currents (the average current during the initial 2 ms of the V j step) were also determined. In the latter case, the collected data were not further filtered during the analysis. The I j -V j relationship was plotted for each experiment. The G j was determined from the slope conductance of the I j -V j relationship (Ϫ30 to ϩ30 mV when V hold was Ϫ30 mV, and Ϫ30 to 0 mV when V hold was Ϫ60 mV). V j was defined as "V m of cell #2 Ϫ V m of cell #1." The apparent steady-state G j was fitted to the Boltzmann function: where G jss is the steady-state conductance, G jmin is the minimum conductance, A is the cooperativity constant, and V o is the voltage at which the decrease in G jss is half-maximal (46). Data graphing and fitting were performed with Origin (version 7.5, OriginLab Corp., Northampton, MA). Statistical analyses were performed with Origin and SPSS (SPSS Inc., Chicago, IL). p Ͻ 0.05 was considered statistically significant. All values are shown as mean Ϯ S.E. "n" equals the number of cell pairs analyzed.

Body-wall Muscle Cells Were Electrically Coupled in a Highly Specific
Pattern-The 95 body-wall muscle cells of C. elegans are organized into two ventral quadrants and two dorsal quadrants. All electrophysiologi-FIGURE 1. Organization of body-wall muscle cells. A, diagram showing a segment of the two ventral quadrants, separated by the ventral nerve cord and the hypodermal ridge. Each quadrant has two rows of spindle-shaped muscle cells that are interlocked into a monolayer. Muscle cells located in the inner row relative to the ventral cord are defined as R1 or L1, whereas those in the outer row are R2 and L2. The shape and organization of muscle cells have been traced from a photograph of a dissected C. elegans. Note that the ventral nerve cord runs along the edge of the right quadrant. The hypodermal ridge is between the ventral cord and the left quadrant. B, diagram showing that laterally projecting muscle arms from muscle cells in the two ventral quadrants interdigitate near the ventral cord. Although each muscle cell (including R2 and L2) has three to five muscle arms (69), only one or two arms are drawn. C, photograph of a dissected C. elegans showing the ventral nerve cord and the two ventral quadrants. Two muscle arms (labeled) are in focus. The photograph was taken with an F-view II digital camera (Soft Imaging System GmbH) mounted on a Nikon E600FN microscope. MARCH 24, 2006 • VOLUME 281 • NUMBER 12 cal experiments were performed on muscle cells in the two ventral quadrants (Fig. 1).

Electrical Coupling of Body-wall Muscle Cells
At first, two adjoining muscle cells from different rows in the same quadrant (e.g. the R1R2 or L1L2 pair in Fig. 1) were analyzed. In response to a series of voltage steps applied to cell #1, I j developed in cell #2 in a V j -dependent manner ( Fig. 2A, panel a). A bell-shaped relationship was observed between G j and V j (Fig. 2B), which is characteristic of gap junctions (5,46,47), and is well fit by a Boltzmann function (46) with A ϭ 0.10, V o ϭ 40, and G jmin ϭ 0.13. In contrast, G j of the body-wall muscle was independent of V m (Fig. 2C). The amplitude of I j remained constant when the cell receiving the voltage steps was interchanged in the same pair (not shown). Thus, adjoining muscle cells from different rows in the same quadrant were electrically coupled in a non-rectifying manner. No significant I j was detected between muscle cells in sequence in the same row (the R1R1 and R2R2 pairs) (Fig. 2, A (panel c) and D). The lack of coupling  in R1R1 and R2R2 pairs suggests that second-order coupling (coupling of two R1s through a common R2 or vice versa) was undetectable in body-wall muscle cells. The coupling of adjoining muscle cells from different rows but not the same row within the same quadrant paints a wave-like coupling pattern (Fig. 2E, panel a).
The two ventral quadrants are separated by the hypodermal ridge and the ventral nerve cord (Fig. 1). To determine if inter-quadrant coupling might occur, we analyzed I j in various pairs of muscle cells from the two ventral quadrants. Significant electrical coupling was observed between muscle cells near the ventral cord at the same longitudinal or "segmental" level (the R1L1 pair) (Fig. 2, A (panel b) and D). No coupling was detected in other inter-quadrant pairs (Fig. 2D). Because muscle arms were the only structures that allow physical contacts between cells from the two quadrants (26), inter-quadrant coupling was most likely mediated by gap junctions in muscle arms.
Carboxyfluorescein is often used to show the existence of intercellular channels, as has been done in pharyngeal muscle cells of C. elegans (11). However, when it was loaded into a single body-wall muscle cell through the whole cell voltage clamp electrode, the dye did not diffuse into neighboring cells (Fig. 2F). This observation suggests that measurement of I j is more sensitive than examination of dye coupling in analyzing coupling of body-wall muscle cells.
Junctional Currents Could Change Membrane Potential Significantly-The I j of body-wall muscle cells was relatively small. The maximal steady-state I j was ϳ15 pA for intra-quadrant coupling and 5 pA for inter-quadrant coupling. To evaluate the effect of the small I j on muscle cells, we measured the resting membrane potential, and tested the effect of injecting depolarizing current comparable to the I j on the membrane potential in the same cell. The resting membrane potential was Ϫ27.1 Ϯ 0.8 mV (n ϭ 25), which is similar to that of the nematode Ascaris lumbricoides body-wall muscle cells (17). Membrane potential changes in response to a series of depolarizing current steps (2.5-25 pA in 2.5-pA increments, 1-s pulse duration) were determined. Injection of 5-and 15-pA currents depolarized the membrane by ϳ5 and 12 mV, were measured in wild-type (WT), unc-9(fc16), unc-7(e5), unc-7(e5); unc-9(fc16) double mutant, WT treated with unc-9 RNAi, the RNAi-supersensitive rrf-3(pk1426) strain treated with unc-9 RNAi, and unc-9(fc16) rescued by expressing WT UNC-9 in muscle cells. A, properties of intra-quadrant coupling (the R1R2 or L1L2 pair). Panel a, representative traces of I j . Panels b and c, the relationship between I j and junctional potential (V j ). unc-9, unc-7;unc-9, WT plus RNAi, and rrf-3 plus RNAi were significantly different from WT and unc-7; there was no difference between WT and unc-7, and between unc-9 and unc-7;unc9 (Kolmogorov-Smirnov test). The same WT data were plotted in both graphs for comparison. Panel d, comparison of junctional conductance (G j ). The asterisk indicates a statistically significant difference compared with WT (one-way analysis of variance with Bonferroni post hoc tests). The numbers of samples analyzed were: WT (6), unc-9 (5), unc-7 (5), unc-7;unc-9 (8), WT plus RNAi (5), and rrf-3 plus RNAi (5). B, properties of inter-quadrant coupling (the R1L1 pair). Panel a, representative traces of I j ; panel b, I j -V j relationships of WT, unc-9, unc-7, and unc-9 rescue. The gray y-axis on the right is only for unc-9 rescue, whereas the black y-axis on the left is for the remaining three groups. Panel c, I j -V j relationships of WT, WT treated with unc-9 RNAi, and rrf-3 treated with unc-9 RNAi. For comparison, the WT data shown in b is re-plotted here. In b and c, unc-9, WT plus RNAi, and rrf-3 plus RNAi but not unc-7 were significantly different from WT; unc-9 rescue was also significantly different from WT (Kolmogorov-Smirnov test). Panel d, comparison of G j . G j was essentially absent in the unc-9 mutant, which was rescued completely by expressing WT UNC-9 in muscle cells; G j was normal in the unc-7 mutant; unc-9 RNAi inhibited the coupling partially in WT but as effectively as the unc-9 mutation in the rrf-3 strain. The asterisk indicates a statistically significant difference compared with WT (one-way analysis of variance with Bonferroni post hoc tests). The numbers of samples analyzed were: WT (7), unc-9 (8), Rescue (9), unc-7 (5), WT plus RNAi (7), and rrf-3 plus RNAi (7). respectively (Fig. 3). Thus, small I j was sufficient to cause significant changes of membrane potential in a coupled cell.

UNC-9 Was a Key Component of Gap Junctions in Body-wall Muscle Cells-
The electrophysiological analyses suggest that body-wall muscle cells are electrically coupled through gap junctions. However, molecular compositions of the gap junctions were unknown. We reasoned that electrical coupling of body-wall muscle cells is likely important to locomotion; the innexin(s) forming the gap junctions may be encoded by an unc gene(s) for which mutants had already been isolated based on uncoordinated phenotypes. Only two innexin genes have been associated with the Unc phenotype: unc-7 and unc-9 (31)(32)(33). We first examined electrical coupling in unc-9(fc16), a putative null mutant resulting from a premature stop in the intracellular loop between the second and third membrane spanning domains (33). Indeed, analyses of steady-state currents show that intra-quadrant coupling was significantly reduced (ϳ50% of the wild type) (Fig. 4A), and inter-quadrant coupling was essentially absent (Fig. 4B) in the mutant. The coupling defect was completely rescued when the wild-type UNC-9 was expressed specifically in muscle cells using Pmyo-3 (Fig. 4B). In fact, G j was significantly larger in rescued animals than in wild-type, possibly due to overexpression. Instantaneous I j was similarly inhibited by the UNC-9 mutation (Fig. S1).
The persistence of some degree of intra-quadrant coupling in the unc-9 null mutant suggests that at least one additional innexin functions in body-wall muscle cells. Among the 24 other innexins in C. elegans, UNC-7 was the most obvious candidate, because unc-7 and unc-9 mutants reportedly have similar phenotypes (32)(33)(34)(35)(36)48). However, both intra- (Fig. 4A) and inter-quadrant (Fig. 4B) couplings were indistinguishable between wild type and the null mutant unc-7(e5) (32). The apparently normal electrical coupling in the unc-7 mutant was not due to functional redundancy between UNC-7 and UNC-9, because intraquadrant coupling was similar between the unc-7(e5);unc-9(fc16) double mutant and the unc-9(fc16) single mutant (Fig. 4A). Thus, UNC-7 does not play a significant role in electrical coupling of body-wall muscle cells.
To further confirm the function of UNC-9, we also performed RNAi experiments. unc-9 RNAi inhibited both intra-and inter-quadrant couplings, although its effect on G j appeared somewhat weaker than the unc-9 null mutation (Fig. 4, A and B). In the RNAi-supersensitive rrf-3(pk1426) strain (49), unc-9 RNAi inhibited the coupling as effectively as the unc-9 null mutation (Fig. 4, A and B). Thus, both mutant analysis and RNAi experiments suggest that UNC-9 is a key component of gap junctions in body-wall muscle cells. Because significant intra-quadrant  coupling remained in the unc-9 mutant and in unc-7;unc-9 double mutant, there must be at least one additional innexin mediating electrical coupling in body-wall muscle cells.

UNC-9 Was Expressed in Both Muscle Cells and Neurons-An in vivo homologous recombination approach (details under "Experimental
Procedures") was used to express a Punc-9::gfp or Punc-9::DsRed2 transcriptional fusion. Transgenic animals showed that UNC-9 was expressed in many neurons and various muscle cells (Fig. 5), which was confirmed by co-labeling neurons or muscle cells with GFP (Fig. S2). The expression of UNC-9 in body-wall muscle cells is consistent with its function in electrical coupling.
UNC-9 Was Localized to Muscle Arms and Intercellular Junctions between Muscle Cell Bodies-To determine the subcellular localization of UNC-9, we expressed two different UNC-9 fusion proteins specifically in muscle cells using Pmyo-3 (43). One of the fusion proteins was Myc::UNC-9, in which Myc was fused to the amino terminus of UNC-9. The other fusion protein was UNC-9::GFP, in which GFP was fused to the carboxyl terminus of UNC-9. In transgenic animals, both Myc::UNC-9 and UNC-9::GFP appeared as puncta or plaques at intercellular junctions between muscle cell bodies, and near the ventral and dorsal nerve cords where muscle arms interdigitate, but were conspicuously absent at membrane regions where there were neither muscle cell body contacts nor muscle arms (Fig. 6). This observation is consistent with results of a previous ultrastructural study showing the presence of morphologically defined gap junctions between muscle arms and between muscle cell bodies (25).
It should be noted the puncta were also observed at intercellular junctions between muscle cells in sequence (Fig. 6), which did not show detectable coupling (Fig. 2, A (panel c) and D). There are two possible interpretations for this discrepancy: 1) gap junctions are normally localized to these junctions; however, they do not cause detectable coupling, because the length of intercellular junctions (thus the number of gap junctions) between muscle cells in sequence are generally shorter than that between muscle cells from the two different rows (Fig. 1); 2) the fusion proteins were localized somewhat more abundantly and indiscriminately compared with wild-type UNC-9. We do not know which of the two interpretations is true at this time.
UNC-9 Function in Body-wall Muscle Was Required for Normal Locomotion-To evaluate the function of body-wall muscle UNC-9 in locomotion, we took advantage of the differential sensitivities of neurons and muscle cells to RNAi-induced gene silencing. Body-wall muscle cells are sensitive, whereas neurons are generally resistant to RNAi (50,51). We produced transgenic animals expressing UNC-9::GFP in both neurons and body-wall muscle cells. As expected, UNC-9::GFP puncta in body-wall muscle cells were largely eliminated by RNAi, whereas those in neurons were indistinguishable from the control (Fig.  7A). Compared with the control, animals subjected to unc-9 RNAi showed a significant reduction of locomotion velocity (Fig. 7B), suggesting that the function of UNC-9 in body-wall muscle cells is important to locomotion. The less severe locomotion defect compared with the null mutant might be due to UNC-9 function in neurons and/or incomplete inhibition of UNC-9 function in body-wall muscle cells. Although the rrf-3 mutant strain appeared to be more sensitive to RNAi than the wild-type (Fig. 4), it could not be used to determine the specific function RNAi inhibited UNC-9::GFP expression in muscle cells but showed no detectable effect in neurons. Panel a, expression at body-wall muscle intercellular junctions (arrowheads) disappeared, whereas that in the nerve ring (arrow) was indistinguishable from the control. Panel b, RNAi eliminated fluorescent puncta between muscle cell bodies and greatly reduced puncta density along the ventral nerve cord (arrows) where gap junctions between muscle arms are located. The bright fluorescent area in the upper part of the right panel was due to autofluorescence of the gut. B, wild-type animals subjected to unc-9 RNAi showed a significant reduction in locomotion velocity, although they were still faster than unc-9(fc16) animals. Panel a, photographs of locomotion tracks after the worms had been singly placed in the center of the field for 7 min. Panel b, comparison of locomotion velocities. The asterisk indicates a statistically significant difference compared with the WT, and the triangle indicates a statistically significant difference compared with both WT and RNAi (one-way analysis of variance with Bonferroni post hoc tests). The results suggest that normal locomotion requires the function of UNC-9 in muscle cells. FIGURE 8. unc-9 mutation did not alter neuromuscular transmission. A, representative traces of miniature postsynaptic currents (mPSCs) from wild-type (WT) and unc-9(fc16). B, representative traces of evoked postsynaptic currents (ePSCs) from wild-type (WT) and unc-9(fc16). C, comparisons of ePSC amplitude, mPSC amplitude, and mPSC frequency between WT and the unc-9 mutant. There was no statistically significant difference. The numbers of samples analyzed were: WT ePSCs (9), unc-9 ePSCs (7), WT mPSCs (11), and unc-9 mPSCs (17).
of muscle UNC-9 in locomotion, because neurons are also sensitive to RNAi in this strain (49).
Neuromuscular Transmission Was Normal in unc-9 Mutant-Because UNC-9 was expressed in both presynaptic motoneurons and postsynaptic body-wall muscle cells, neuromuscular transmission could be altered in the unc-9 mutant. To examine this possibility, we recorded mPSCs and ePSCs at the neuromuscular junction of the unc-9(fc16) mutant. Both mPSCs and ePSCs were indistinguishable from those of the wild type (Fig. 8). Thus, UNC-9 function was not required for neuromuscular transmission under our experimental conditions.

DISCUSSION
Our analyses show that electrical coupling among C. elegans bodywall muscle cells existed not only within the same quadrant but also between two distinct quadrants. The coupling was characterized by a highly organized and specific coupling pattern. Although intra-quadrant coupling likely occurred through gap junctions between muscle cell bodies, inter-quadrant coupling occurred through gap junctions between muscle arms. Although the analyses were performed in the two ventral quadrants, the findings are likely applicable to the two dorsal quadrants as well because ventral and dorsal muscle cells are identically organized (24,26).
Junctional conductance was ϳ300 pS for intra-quadrant coupling, and 75 pS for inter-quadrant coupling. Because multiple gap junctions likely contributed to the G j , single channel conductance of the gap junction is probably smaller. Because similar in situ analyses of innexin-or pannexin-based gap junctions have not been performed in other systems, no direct comparisons can be made. Nevertheless, biophysical properties of innexin-or pannexin-based gap junctions have been examined in cultured insect cells and on heterologous expression in Xenopus oocytes. In the insect cell line Sf9 (Spodoptera frugiperda), gap junctions have two single channel conductance states of 224 and 42 pS, respectively (52). In gap junctions formed de novo by manually opposing two cultured mosquito cells, the average conductance of a fully open gap junction channel is 375 pS (14). Hemichannels formed by human PX1 in Xenopus oocytes have a single channel conductance of 475 pS (9). Compared with these gap junctions and hemichannels, the conductance of gap junctions in C. elegans body-wall muscle cells appeared small. However, biophysical properties of gap junctions in cultured cells and in the heterologous expression system could be very different from those in vivo or in situ. For connexin-based gap junctions, single channel conductance ranges from 10 to 300 pS in native tissues (5,53). G j is also highly variable in mammalian preparations. For example, the mean G j between green cones in squirrel retina is 220 pS (54). It is 100 pS or less in rat electrical synapses between Bergmann glial cells and Purkinje neurons (55), between dopaminergic neurons (56), and between striatal ␥-aminobutyric acidergic output neurons (57). In contrast, relatively large G j (ϳ1 nS) is observed in rat electrical synapses between dendrites of retinal ganglion cells (58) and between sympathetic preganglionic neurons (59). Thus, although the G j of C. elegans body-wall muscle is relatively small compared with some electrical synapses, it is not a unique property.
Although the I j was small, injection of comparable depolarizing currents changed the resting membrane potential considerably. Furthermore, partial inhibition of muscle gap junction function by unc-9 RNAi significantly reduced locomotion velocity. Therefore, gap junctions in body-wall muscle cells are of physiological importance. Body-wall muscle cells of the nematode A. lumbricoides are also electrically coupled (60). It has been speculated that not all muscle cells form neuromuscular junctions with motoneurons in Ascaris; the function of gap junctions in body-wall muscle might be "a means of transmitting the electrical signal to neighboring cells so that their activity tends to be synchronized" (60). The specific coupling patterns revealed by this study suggest that the wave-like intra-quadrant coupling might contribute to the sinusoidal locomotion wave form, whereas the cross-coupling between selected muscle cells from separate quadrants might serve to synchronize ventral or dorsal muscle activities at the same longitudinal level.
Action potentials may occur either spontaneously or response to current injections in body-wall muscle cells of Ascaris, and they correlate with visible muscle contractions (19,61). It is still questionable whether action potentials occur in C. elegans body-wall muscle cells, and, if they do, what the threshold is. Occasional spike-like activities (with peak amplitude of up to ϩ15 mV) may be observed in C. elegans body-wall muscle cells (62). However, current injections into body-wall muscle cells do not induce typical action potentials (62). Under our experimental conditions, muscle contractions were only observed occasionally, even during stimulation of motoneurons. Thus, more experiments are needed to elucidate the relationship between electrical coupling and body-wall muscle contractility.
Differentiated vertebrate skeletal muscles do not have gap junctions (63). However, motoneurons innervating the muscles are electrically coupled in a spatially restricted manner. The mean coupling potential, which is action potential-independent membrane depolarization induced by antidromic stimulation of an electrically coupled neighboring motoneuron, is 1.3-3.4 mV in neonatal rat lumbar spinal cord preparations (64). Gap junctions in the spinal cord can produce a stable rhythmic motor pattern independent of action potentials and chemical neurotransmission (65). Thus, electrical coupling may have importance to locomotion in other species besides nematodes.
The gap junction is a head-to-head assembly of two hemichannels, with each hemichannel consisting of six subunits. Gap junctions could be homotypic (consisting of a single isoform of connexin, innexin, or pannexin), heterotypic (consisting of two different homomeric hemichannels), or heteromeric (consisting of two heteromeric hemichannels). There are 25 innexin genes in C. elegans. Although ultrastructural analyses revealed the presence of gap junctions in C. elegans body-wall muscle 30 years ago (25), molecular compositions of the gap junctions have remained elusive. Our analyses provide compelling evidence that UNC-9 is a key component of the gap junctions. Despite unc-7 and unc-9 mutants reportedly have similar behavioral and pharmacological phenotypes (32)(33)(34)(35)(36), body-wall muscle electrical coupling was normal in the unc-7 mutant. Because significant intraquadrant coupling remained in the unc-9 null mutant, at least one additional innexin functions in body-wall muscle cells. The identity of the one or more innexins is yet to be determined.
Although UNC-9 was expressed in both presynaptic motoneurons and postsynaptic muscle cells, both mPSCs and ePSCs were indistinguishable between the unc-9 mutant and wild-type, suggesting that UNC-9 is not directly involved in chemical transmission at the neuromuscular junction. Compared with the unc-9 mutation, inhibition of body-wall muscle UNC-9 by RNAi only had a weak effect on locomotion. A major cause for this difference could be the function of UNC-9 in the nervous system, because neurons are resistant to RNAi (49,50). Thus, UNC-9 in the nervous system might play an important role in locomotion by functioning upstream of the neuromuscular junction.
Ultrastructural analyses of C. elegans body-wall muscle cells in the head region show that gap junctions between muscle arms only occur among muscle cells from distinct quadrants but not from the same quadrant, even though both sets of arms are equally accessible. Muscle cells in the same quadrants are connected by gap junctions at their cell bodies (66). This pattern of gap junction localization is likely preserved among other body-wall muscle cells. Consistent with the ultrastructural analysis, both Myc::UNC-9 and UNC-9::GFP, when expressed specifically in muscle cells, appeared as puncta at intercellular junctions between muscle cell bodies as well as near the ventral and dorsal nerve cords where muscle arms interdigitate. These analyses collectively suggest that intra-and inter-quadrant couplings might be mediated by two populations of gap junctions that were localized between muscle cell bodies and muscle arms, respectively. The differential inhibition of intra-and inter-quadrant couplings by the unc-9 mutation suggests that the two populations of gap junctions might have different molecular compositions or functional properties.
Gap junctions play critical roles in generating rhythmic oscillations in the nervous system. For example, electrical coupling among respiratory neurons in the brainstem is important to respiratory rhythm generation (67). Electrical coupling among motoneurons in the spinal cord appears to be the basis for a motor rhythm (65). Gap junction-mediated rhythmic activities in hippocampus appear to underlie several cognitive functions, including memory and sensory processing (68). The generation of rhythmic oscillation patterns in the nervous system might require populations of cells to function in a spatially and temporally organized manner. How could cells achieve this spatially and temporally organized coupling? Obviously, it would be impossible if all cells were excited or inhibited at the same time due to strong electrical coupling. C. elegans body-wall muscle cells contract coordinately and sequentially to produce the sinusoidal locomotion body bends, even though they are all connected through gap junctions. Low conductance electrical coupling could potentially be important to such spatially and temporally organized couplings.