Oligomeric Structure and Functional Characterization of Caenorhabditis elegans Innexin-6 Gap Junction Protein*

Background: Invertebrates possess gap junction channels composed of innexins. Results: The molecular dimension, oligomeric state, and permeability of C. elegans INX-6 channels were determined by structural, biochemical, and functional studies. Conclusion: INX-6 channels have a larger overall structure and higher permeability as compared with connexins. Significance: Innexin channels appear to be distinct from connexin channels in terms of structure and function. Innexin is the molecular component of invertebrate gap junctions. Here we successfully expressed and purified Caenorhabditis elegans innexin-6 (INX-6) gap junction channels and characterized the molecular dimensions and channel permeability using electron microscopy (EM) and microinjection of fluorescent dye tracers, respectively. Negative staining and thin-section EM of isolated INX-6 gap junction membranes revealed a loosely packed hexagonal lattice and a greater cross-sectional width than that of connexin26 and connexin43 (Cx43)-GFP. In gel filtration analysis, the elution profile of purified INX-6 channels in dodecyl maltoside solution exhibited a peak at ∼400 kDa that was shifted to ∼800 kDa in octyl glucose neopentyl glycol. We also obtained the class averages of purified INX-6 channels from these peak fractions by single particle analysis. The class average from the ∼800-kDa fraction showed features of the junction form with a longitudinal height of 220 Å, a channel diameter of 110 Å in the absence of detergent micelles, and an extracellular gap space of 60 Å, whereas the class averages from the ∼400-kDa fraction showed diameters of up to 140 Å in the presence of detergent micelles. These findings indicate that the purified INX-6 channels are predominantly hemichannels in dodecyl maltoside and docked junction channels in octyl glucose neopentyl glycol. Dye transfer experiments revealed that the INX-6-GFP-His channels are permeable to 3- and 10-kDa tracers, whereas no significant amounts of these tracers passed through the Cx43-GFP channels. Based on these findings, INX-6 channels have a larger overall structure and greater permeability than connexin channels.

is a temperature-sensitive allele, and inx-6 mutants exhibit feeding deficits as pharyngeal muscle contraction is unsynchronized and the anterior pharyngeal muscles are less electrically coupled than the posterior metacorpus muscles. Expression of inx-6 in the metacorpus overlaps with eat-5, but the two genes are not fully interchangeable, suggesting that inx-6 and eat-5 serve similar but not identical functions in vivo (18).
Structural studies of connexin gap junction channels were recently performed using electron crystallography and x-ray crystallography (19 -21). Knowledge of the three-dimensional structure of innexin channels, however, is very limited because neither high-resolution images nor purification of recombinant innexin channels have been reported. EM observation of invertebrate gap junctions was performed before molecular cloning of innexin genes. In freeze fracture or negatively stained images, gap junctions in crayfish appear similar to those of vertebrates, but the hexagonal arrays of the channels often exhibit swollen packing with a unit cell of up to ϳ200 Å (22). Studies using electron tomography estimated both the periodicity of the channels and the extracellular gap size between the junction membranes isolated from the lateral giant fibers of crayfish (23). Although the native gap junctions in C. elegans have been observed by thin-section EM (24,25), the molecular structure of innexin channels remains poorly resolved.
To further elucidate the structure of innexin gap junction channels, we isolated and characterized recombinant C. elegans INX-6 channels. INX-6 channels were overexpressed in Sf9 cells, and EM observation revealed the isolated gap junction plaques of INX-6. Negative staining and thin-section EM imaging methods were used to evaluate the channel distance and the width of the gap junction membranes. INX-6-GFP-His channels, in which the GFP-His tag was fused at the C terminus of INX-6, were solubilized and purified in dodecyl maltoside (DDM) detergent solution and then analyzed by gel filtration with different detergents in running buffer. Single particle analysis was performed to determine the oligomeric state and molecular dimensions of purified INX-6 channels. We also performed a dye transfer assay with various sizes of tracers, sulforhodamine 101 (SR101, 607 Da), 3-kDa Texas Red dextran (3k-TR), 10-kDa Texas Red dextran (10k-TR), and 70-kDa Texas Red dextran (70k-TR). Our results indicate that INX-6 forms similar but distinct gap junction channels from connexins in terms of the molecular dimensions and permeability. These findings are consistent with the notion that invertebrates possess a wide limiting pore diameter as compared with vertebrates (26).

EXPERIMENTAL PROCEDURES
Expression and Purification of Recombinant INX-6 Gap Junction Channels-The gene coding the full-length C. elegans INX-6 was cloned into pFastBac1 (Invitrogen) for expression in Sf9 cells. A His tag (His 6 ) or GFP-His tag (GFP plus His 8 ) was inserted along with a thrombin cleavage recognition sequence at the C terminus of INX-6 and then cloned into pFastBac1. Recombinant baculoviruses of INX-6, INX-6-His, or INX-6-GFP-His were generated from recombinant bacmids and used to infect Sf9 cells at 27°C. After infection for 30 h, the cells were harvested by low speed centrifugation followed by freezing in liquid nitrogen and storage at Ϫ80°C. For membrane preparation, cells were suspended in buffer containing 10 mM Tris (pH 7.5), 150 mM NaCl, and 1 mM phenylmethylsulfonyl fluoride and disrupted by sonication for 90 s. The cell suspension was centrifuged at 22,100 ϫ g for 25 min. The collected INX-6-His or INX-6-GFP-His membranes were resuspended and solubilized in buffer containing 10 mM Tris (pH 7.5), 150 mM NaCl, and 1% DDM for 30 min. The debris was discarded by centrifugation at 21,300 ϫ g for 10 min. The supernatant was bound to the nickel-nitrilotriacetic acid-agarose (Qiagen), washed with 10 mM L-histidine, and eluted with 300 mM L-histidine. Isolated membranes and purified channels were checked by SDS-PAGE and Western blot with anti-His antibody (Qiagen). All membrane preparation and purification steps were conducted at 4°C. The human Cx26 channel preparation protocol was described previously (27).
Electron Microscopy for Negative Staining and Thin-section Imaging-All images were recorded on a JEOL 1010 with 40,000ϫ magnification operated at 100 kV. The specimens were negatively stained with 2% uranyl acetate. For cross-sectional images, we used the conventional ultrathin sectioning method. Briefly, the pelleted membranes were fixed in 2% glutaraldehyde followed by a second fixation with 2% OsO 4 at 4°C. The specimen was subsequently stained with 1% uranyl acetate, embedded in epoxy resin, and allowed to harden at 60°C overnight. The ultrathin sections were 70 -80 nm thick. Negative stain images (see Figs. 3, A-C, and 4, C and E) were obtained using a 2K slow scan charge-coupled device camera (Gatan, 24-m pixel size) with an image pixel size of 0.57 nm. Thin-section images were obtained with a 2K complementary metal oxide semiconductor TemCam-F216 (Tietz, 15.6-m pixel size) with 0.22 nm/pixel. For single particle analysis, all data were obtained with a 4K complementary metal oxide semiconductor TemCam-F416 (Tietz, 15.6-m pixel size) with an image pixel size of 0.286 nm.
Size Exclusion Chromatography and Single Particle Analysis of Purified INX-6 Channels-After digesting the GFP-His tag with thrombin, INX-6 proteins were loaded onto a Superose 6 10/300 GL column (GE Healthcare) equilibrated in 10 mM Tris (pH 7.5), 300 mM NaCl, 1 mM dithiothreitol, and 0.05% DDM. The purified INX-6 channels were also analyzed in 0.1% octyl glucose neopentyl glycol (OGNG) in gel filtration buffer to further study the oligomeric form. The column was calibrated with molecular mass standards (thyroglobulin, 670 kDa; ␥-globulin, 158 kDa; ovalbumin, 44 kDa; myoglobin, 17 kDa; and vitamin B12, 1.35 kDa, see Fig. 1A) and run at a flow rate of 0.5 ml/min. Fractions around the highest peak of UV 280 nm absorbance were analyzed by SDS-PAGE and native PAGE. The protein bands were visualized using Coomassie Brilliant Blue-R250 or SimplyBlue SafeStain (Invitrogen). Single particle analysis was performed using the negative stain EM images of purified INX-6 channels taken from gel filtration fractions at ϳ800 kDa and at ϳ400 kDa. Picked particles were aligned and classified based on multivariate statistical analysis by EMAN (29). Total numbers of picked particles were 2728 for the junction channel and 7283 for the hemichannel.

Microinjection of Fluorescent Tracers-INX-6-GFP-His recombinant viruses were infected to the attached Sf9 cells in 35-mm glass bottom dishes. After infection for 24 h, fluorescent tracers were injected into the cells forming the gap junction plaque signals. Recombinant viruses expressing intact GFP
were used as negative controls of the Sf9 cells. For Cx43-GFP, stable HeLa cells were cultured in 35-mm glass bottom dishes. The tracers were injected 24 h after induction by doxycycline and trichostatin A. Mock-transfected cells were stable HeLa cells with no induction of Cx43-GFP expression and were used as negative controls. Dye transfer was captured on a fluorescent microscope IX81 (Olympus) 1 min after injection. One set was defined as 10 -15 injections. n indicates the number of sets. The fluorescent tracers, SR101 (607 Da), 3k-TR, 10k-TR, and 70k-TR, were purchased from Molecular Probes. As for the Texas Red tracers, conjugated dextran has a molecular mass distribution range as follows: 3-kDa, 1500 -3000 Da; 10-kDa, 9000 -11,000 Da; and 70-kDa, 60,000 -90,000 Da.

RESULTS
INX-6 constructs with different C-terminal tags were generated in accordance with the experimental aim. The intact INX-6 construct was used for the negative stain EM observation of gap junction plaques and for thin-section imaging. For purification, either a His tag or a GFP-His tag was fused to INX-6 (INX-6-His, INX-6-GFP-His), and the GFP-His tag was digested by thrombin before loading proteins on a gel filtration column. INX-6-GFP-His was also used for the dye transfer assay. These constructs are summarized in Fig. 1B.
Expression and Purification of His-tagged INX-6 Gap Junction Channels-INX-6-His gap junction membranes were isolated from Sf9 cells by sonication and one-step centrifugation. This isolation protocol yielded poorer purity of innexin proteins than the alkali preparation for connexin gap junction membranes (Fig. 2, lane 1) (30). INX-6-His protein expression in the membrane fraction, however, was detected by Western blot with the anti-His antibody (Fig. 2, lane 3). The signal appeared at ϳ37 kDa, which is smaller than that estimated from the amino acid sequence, 45 kDa. This might be due to the globular form and/or charge distribution, which is specific to denatured INX-6 peptides in SDS buffer. After solubilization and purification with nickel-nitrilotriacetic acid agarose in DDM solution, the INX-6-His band clearly appeared on the Coomassie Brilliant Blue-stained gel (Fig. 2, lane 2), consistent with the signal detected by the anti-His antibody (Fig. 2, lane 4).
INX-6 Forms Loosely Packed Hexagonal Gap Junction Plaques in Sf9 Cells-Gap junction plaques of INX-6 in isolated membrane fractions observed by negative stain EM very clearly showed the channel pores (Fig. 3A). The contribution of endogenous innexin expression to the formation of recombinant INX-6 gap junctions, if any, was thought to be negligible as no gap junction plaque was observed in control Sf9 cells (data not shown). Fast Fourier transform from the selected area of INX-6 plaques resulted in blurred hexagonal diffraction spots (Fig.  3D), suggesting that the INX-6 channels form a loosely packed array but have packing similar to the gap junction plaques of Cx26 in Sf9 cells and Cx43-GFP in HeLa cells (Fig. 3, B, C, E, and F). We estimated the channel distances using the fast Fourier  transform diffraction spots from several selected images of INX-6 gap junction plaques. Assuming a hexagonal lattice, the adjacent INX-6 channels were aligned with a pitch of as much as 110.1 Ϯ 3.3 Å (n ϭ 8) ( Table 1). The channel distance of INX-6 varied depending on the selected plaques but was never less than 100 Å. Gap junction plaques of Cx26 and Cx43-GFP in our preparation revealed a channel pitch of 93.9 Ϯ 1.1 Å (n ϭ 11) and 77.2 Ϯ 1.1 Å (n ϭ 9), respectively (Table 1), consistent with the cell parameters of vertebrate gap junctions, between 75 and 95 Å (31)(32)(33)(34)(35)(36)(37)(38)(39).
We subsequently performed thin-section EM of isolated INX-6 gap junction plaques to measure the cross-sectional width (Fig. 3, G, H, and I). The width of INX-6 gap junctions was 184.1 Ϯ 4.3 Å (n ϭ 12) as compared with 139.6 Ϯ 3.3 Å (n ϭ 10) for Cx26 gap junctions and 162.1 Ϯ 3.1 Å (n ϭ 11) for Cx43-GFP junctions ( Table 1). The two extracellular loops of INX-6 are thought to be longer, i.e. ϳ50 amino acids for innexins, estimated from the alignment of innexin sequences (supplemental Fig. 1), and ϳ30 amino acids for connexins, based on the Cx26 x-ray structure (20). The greater width of INX-6 gap junction plaques may be due to greater spacing between adjacent membranes.
Oligomeric State and Molecular Dimension of Purified INX-6 Channels-INX-6-GFP-His channels were solubilized in DDM solution and purified using nickel affinity chromatography. After thrombin digestion of the GFP-His tag, the channels were analyzed by gel filtration. The highest elution peak of UV 280 nm absorbance appeared at ϳ400 kDa (386 Ϯ 8 kDa, n ϭ 13) (Fig. 4A) and was confirmed to be INX-6 peptides by SDS-PAGE (Fig. 4B). The peak fraction was collected and subjected to negative stain EM observation. The oligomeric INX-6 channels showed homogeneous particles, some of which were oriented with their pores perpendicular to the carbon support film forming a doughnut shape (Fig. 4C). When nickel affinity-purified INX-6 channels were loaded in 0.1% OGNG, the absorbance of INX-6 peaked at ϳ800 kDa (783 Ϯ 42 kDa, n ϭ 7) (Fig.  4D). Accordingly, EM observation of the fraction at ϳ800 kDa in OGNG revealed many double-barreled particles (Fig. 4E, square brackets), indicating dimerization of the channels in Fig.  4C. In native PAGE analysis of peak fractions, we observed a major band shift of INX-6 from ϳ480 kDa in DDM (Fig. 4F, lane 1) to over 720 kDa in OGNG (Fig. 4F, lane 2), consistent with the UV 280 nm absorbance profiles and negative stain EM observations. We further performed single particle analysis using the negative stain EM images. In the class averages of INX-6 channels at ϳ800 kDa, we found projection structures that clearly showed the typical features of docked junction channels as a side view with two detergent belts corresponding to the transmembrane region where the stain does not penetrate (Fig. 4G). The dimensions of the INX-6 junction structure were estimated from a representative average (Fig. 4H, left). The longitudinal height of the junction channel was 220 Å, the extracellular gap space was 60 Å, and the cytoplasmic protrusions were 45 Å wide. Because the transmembrane region, which is 140 Å in diameter, includes detergent micelles, diameters of 110 Å at the cytoplasmic and extracellular ends, where no detergent micelles are thought to contribute, may be close to the actual channel diameter. These measurements demonstrate that the molecular dimensions of the INX-6 junction channel are 1.2-1.5 times greater than those of the Cx26 channels (Fig. 4H, right) (20). This finding is consistent with the extended channel pitch and cross-sectional width of INX-6 gap junction membranes as compared with the connexins (Fig. 3 and Table 1). On the other hand, the class averages from the ϳ400-kDa fraction revealed that smaller size particles assumed a variety of orientations (Fig. 4I). A representative class average that clearly shows a channel pore in a top view has a channel diameter of ϳ140 Å (Fig. 4J), and the diameters of other class averages are similar (Fig. 4I). These particles are too small for junction forms, as shown in Fig. 4H (left). Based on these findings, we concluded that INX-6 channels at ϳ400 kDa are predominantly hemichannels, whereas the fraction at ϳ800 kDa includes more docked junction channels.
Dye Transfer Analysis of INX-6 Channels-INX-6-GFP-His expressed in Sf9 cells formed gap junction plaques between neighboring cells (Fig. 5A, top row, yellow arrowheads) but did not form junction plaques in mammalian cells (data not shown). Recombinant viruses expressing soluble GFP were generated as negative controls (Fig. 5A, bottom row). To compare the permeability with that of connexin channels, we selected Cx43-GFP as a representative connexin because Cx43 is thought to have a relatively larger limiting pore diameter and less charge selectivity than other connexin family isoforms (40). Cx43-GFP formed gap junction signals in HeLa cells (Fig. 5B, top row, yellow arrowhead), but not in Sf9 cells (data not shown). HeLa cells without Cx43-GFP expression were used as negative controls (Fig. 5B, bottom  row). Both INX-6-GFP-His and Cx43-GFP exhibited significant transfer of SR101 within 1 min after injection (Fig. 5, A  and B, top rows, right panels). To investigate the permeability of INX-6-GFP-His and Cx43-GFP, several tracers with different sizes, SR101 (607 Da), 3k-TR (3 kDa), 10k-TR (10 kDa), and 70k-TR (70 kDa), were examined. The proportion of dye-coupled cells through INX-6-GFP-His junctions was 82.7% for SR101, 46.7% for 3k-TR, and 11.6% for 10k-TR, and the transfer of 70k-TR was as high as the background (Fig. 5C). For Cx43-GFP channels, only SR101 transferred through more than 90% of the junctions, but the other larger dyes were not significantly different from background (Fig. 5D). These findings are consistent with the general criteria that gap junctions in vertebrates allow for the passage of molecules up to 1 kDa in size (1). These results indicate that, as compared with Cx43-GFP channels, INX-6 channels allow for the passage of larger molecules.

DISCUSSION
The innexin family includes a large number of family members identified in Arthropoda, Nematoda, Annelida, and ichnovirus (12). Although recent molecular studies of innexins using electrophysiology demonstrated the helical nature of the first transmembrane domain of the innexin ShakB(L) (41), direct observation of recombinant innexin gap junction plaques and the purification of innexin channels have not yet been reported. Here we established the overexpression and purification of C. elegans INX-6 gap junction channels using Sf9 cells and successfully obtained the class averages by single particle analysis.
The alkali extraction method is often used to isolate connexin gap junction plaques in native tissue and Sf9 cells (27,30,42,43). In the present study, we used much milder conditions, such as a neutral pH and low ion concentration without reducing agents, to avoid INX-6 protein degeneration. Once solubilized in DDM, the oligomerized INX-6 channels were successfully purified with an affinity column using the similar protocol used for the purification of Cx26 (27). This may reflect the specific stability of INX-6 in the lipid bilayer and in detergent solution.
For more than three decades, a number of gap junction researchers have reported the structure of vertebrate gap junc-

TABLE 1 Estimation of channel distance and junction membrane width based on EM observation
Junction width was evaluated with edge-to-edge spacing of pixels on thin-section images. tions, comprising connexin, using isolated gap junction plaques or crystallized membrane sheets in a two-dimensional array (31)(32)(33)(34)(35)(36)(37)(38)(39). They usually form a hexagonal lattice in which the cell parameters range from 75 to 95 Å. This would be reasonable given that Cx26 is a dodecameric channel with a maximum channel diameter of 92 Å at the flexible cytoplasmic end (20). The hexagonal lattice of INX-6 gap junction plaques is similar to that of connexin (Fig. 3, A and D). The top view of the class average picked from the ϳ400-kDa fraction resembles a hexameric hemichannel (Fig. 4, I and J). However, the 1.2-1.5-fold difference in size of INX-6 and Cx26 would result in a 1.7-3.4-fold difference in volume. Even pentameric channels can form a hexagon-like lattice in two dimensions (44,45). The oligomeric number of INX-6 must be carefully determined before performing structural studies using cryo-EM. In early studies of the gap junction ultrastructure, crayfish gap junctions, probably comprising innexins, were determined by thin-section and negative stain EM to be ϳ170 Å in height with a unit cell of ϳ200 Å (22). The recent x-ray structure reveals that the height of Cx26 channels is about 155 Å with partially disordered cytoplasmic domains (20) (PDB code: 2ZW3). It is not straightforward to associate the crystal structure with thin-section EM observations due to artifacts produced during fixation or staining of the sample. The general consensus, however, with some exceptions (46), is that gap junctions from invertebrates have a larger junction membrane width, extracellular gap region, and channel diameter as compared with those from vertebrates (22,23,33,(47)(48)(49)(50)(51)(52)(53). Our measurements of INX-6 based on the class averages are consistent with these previous studies. Two-dimensional image averaging revealed that the purified vertebrate innexin homologs pannexin1 and pannexin2 channels have larger oligomer and pore diameters than Cx26 (54). The channel diameter for INX-6 is closer to that of pannexin1 than of Cx26 (140 Å for INX-6 and 120 -160 Å for pannexin1, including detergent micelles). Although other innexin isoforms must be investigated for generalization of this trend, it is possible that the junction membrane width and molecular dimensions of innexins differ from those of connexins, due in part to the longer sequences of the extracellular loops of innexins.
The permeability of gap junction channels was previously estimated using fluorescent probes with various sizes and properties such as charge and shape, which established a molecular size cut-off of ϳ1 kDa for vertebrate gap junctions and ϳ2 kDa for invertebrate gap junctions (55,56). Another study using fluorescent molecules conjugated with oligosaccharides, neutral branched glycopeptides, and charged linear peptides showed the transfer of molecules up to ϳ3 kDa (26). It is therefore not surprising that the 3k-TR tracer passed through the INX-6 channels in this study. The low but significant transfer of 10k-TR, with a dextran distribution range between 9000 and 11,000 Da, was however, contrary to our expectations (Fig. 5C). The permeability limits of pores would depend on substrate shape rather than molecular mass. Conjugated dextran is thought to have an elongated polysaccharide structure. Although linear peptides as large as 1.8 kDa effectively permeate Cx43 channels, circular 8-mer peptides do not (57). In fact, fluorescent dye-conjugated aprotinin, which has a molecular mass of ϳ6.5 kDa, did not pass through INX-6 channels (data not shown), probably because of the highly folded molecular structure with an ␣-helix and ␤-sheet that are larger than the pore diameter. Our results indicate that the INX-6 channel pore is more permeable than Cx43 and would probably be more permeable than most connexins as Cx43 is considered to have one of the largest limiting pore diameters in the connexin fam- ily (40) with some exceptions for positively charged molecules (58). It remains unclear why gap junction channels with greater permeability predominate in invertebrates. The electrical synapse is common in invertebrates, and high permeability might be advantageous for simple and fast transmission. Because of the demand for complex communication, vertebrates may have developed unique coupling systems with distinct permeability, such as connexin.
Endogenous gap junction channels in Sf9 cells were inferred by electrophysiology (59). In all of our microinjection experiments, however, no high background activity was observed (Fig.  5). We cannot rule out some expression of endogenous innexin in Sf9 cells because Sf-inx2, which is close to Drosophila inx-2, has been identified (12). The expression of endogenous gap junction channels in Sf9 cells might not be high enough to affect the dye transfer background in our study. Alternatively, the electrical conductance of insect gap junction channels might not correlate with pore size. For connexin channels, there is likely no direct relation between channel conductance and ionic selectivity or dye permeability (60), which is supported by other studies (61,62).
We successfully established the overexpression and purification of recombinant C. elegans INX-6 gap junction channels. Studies of the purified INX-6 gap junction channels revealed that they form a loosely packed hexagonal gap junction plaque that is similar to but distinct from connexins. The purified INX-6 channels, predominantly hemichannels in DDM and junction channels in OGNG, can be used for structural studies using high-resolution analysis. The class averages of purified INX-6 channels revealed that the molecular dimensions of the junction form of INX-6 are greater than those of Cx26. Future cryo-EM studies could elucidate the oligomeric number of INX-6 channels, which remains unclear in the present study. Dye transfer assays revealed that INX-6 channels have higher permeability than Cx43-GFP channels. Our findings provide insight into the morphology and characteristic permeability of recombinant innexin channels. The biologic significance of INX-6 pores with higher permeability remains to be determined. Further structural and functional studies of innexin channels will enhance our understanding of the mysteriously fast gating channels observed in crayfish, such as those that can close within 1.0 ms (63,64).