Research Article

Multicolor and Electron Microscopic Imaging of Connexin Trafficking

See allHide authors and affiliations

Science  19 Apr 2002:
Vol. 296, Issue 5567, pp. 503-507
DOI: 10.1126/science.1068793


Recombinant proteins containing tetracysteine tags can be successively labeled in living cells with different colors of biarsenical fluorophores so that older and younger protein molecules can be sharply distinguished by both fluorescence and electron microscopy. Here we used this approach to show that newly synthesized connexin43 was transported predominantly in 100- to 150-nanometer vesicles to the plasma membrane and incorporated at the periphery of existing gap junctions, whereas older connexins were removed from the center of the plaques into pleiomorphic vesicles of widely varying sizes. Selective imaging by correlated optical and electron microscopy of protein molecules of known ages will clarify fundamental processes of protein trafficking in situ.

To understand how genomes give rise to cells and organisms, one would like to be able to genetically tag any protein with a nonperturbing label visible both by nondestructive optical microscopy in live cells and by electron microscopy (EM) for higher spatial resolution. The jellyfish green fluorescent protein (GFP) and its mutants and homologs have provided solutions for optical imaging, but genetically encoded tags for electron microscopy are still lacking. Furthermore, if one could pulse-label a continuously expressed protein with one label and chase the staining with a different label, older and newer copies of the same species would be distinguishable in a single image, which would help to elucidate the mechanisms of assembly and turnover.

Important examples of dynamic structures created from a single family of proteins include gap junction plaques, which are composed of an array of tightly packed individual channels formed by the docking of two hemichannels (connexons) in the extracellular space between two contacting cells [reviewed in (1–3)]. Briefly, each cell provides one hemichannel, which is formed by six connexins arranged around a central pore. Plaques ensure direct communication between contacting cells by allowing small (less than 1 kD) signaling molecules, ions, and metabolites to pass directly from one cell to the other. This connectivity enables cells in tissues and organs to efficiently communicate in order to coordinate cellular activity. Because the degree of intercellular coupling is, for the most part, determined by the number of individual channels in a gap junction plaque, cells dynamically modulate this connectivity by regulating the synthesis, transport, and turnover of the constituent connexins, which have a half-life of less than 5 hours (4–7).

Much has been learned by tagging gap junctions through the fusion of fluorescent proteins like GFP to the COOH terminus of various connexins (8–11). However, GFP and its variants are 25 to 27 kD proteins, often larger than the protein of interest. Though not necessarily a major limitation in gap junction biology, the large size of GFP can interfere with the distribution, function, and fate of other recombinant proteins. GFPs offer no advantage over epitope tags for EM because both require conventional antibody labeling methods with the associated limitations in quality of preservation, reagent penetration, and labeling efficiency, all of which hinder specificity and resolution.

Recombinant proteins in intact cells can also be labeled by genetically appending or inserting a small motif (6 to 20 residues) containing the sequence -Cys-Cys-Xaa-Xaa-Cys-Cys-, then exposing the cells to a membrane-permeant nonfluorescent biarsenical derivative of fluorescein, FlAsH-EDT2 (Fig. 1A). FlAsH binds with high affinity and specificity to the tetracysteine motif and thereby becomes strongly green fluorescent (12,13). Toxicity and binding of the trivalent arsenic atoms to endogenous thiols are minimized by simultaneous administration of micromolar concentrations of antidotes such as 1,2-ethanedithiol (EDT). More recently, we have determined in vitro rate constants for FlAsH association and dissociation from model tetracysteine-containing peptides to be in the range of 105M−1s−1 and 10−6s−1, respectively, consistent with effective dissociation constants around 10−11 M (14). Thus, once FlAsH-EDT2 is present at micromolar concentrations in the cytosol, newly synthesized tetracysteine motifs should be labeled within a few minutes. Upon removal of excess FlAsH, the fluorescent complexes should survive for days in the absence of excessive (mM) concentrations of competing EDT. We have now synthesized red- and blue-emitting analogs, the most useful of which is ReAsH (Fig. 1A), a biarsenical derivative of the red fluorophore resorufin (14). Like FlAsH, ReAsH has very little fluorescence when bound to EDT, but becomes brightly fluorescent upon binding to tetracysteine motifs. Such complexes absorb and fluoresce at considerably longer wavelengths (593 and 608 nm, respectively) than the corresponding wavelengths (508 and 528 nm) for FlAsH complexes, even though ReAsH is a smaller molecule than FlAsH.

Figure 1

Cx43-TC is expressed in HeLa cells and contributes to the formation of gap junction plaques. (A) Chemical structure of FlAsH-EDT2 and ReAsH-EDT2 ligands. (B) Schematic of the Cx43-TC construct. M1-4, transmembrane domains; E1-2, extracellular loop regions; CL, cytoplasmic loop; TC, tetracysteine domain; and a connexon hexamer with 6 ReAsH molecules bound to the tetracysteine domains. (C) Whole cell lysates from HeLa cells expressing Cx43-TC or wild-type Cx43 were Western blotted with a monoclonal antibody specific to Cx43. (Dthrough F) pCI-neo-Cx43-TC was transfected in HeLa cells. Expression was checked 20 to 24 hours after transfection by double labeling with FlAsH-EDT2 (D) and a Cx43-specific monoclonal antibody (E). Both labels recognized the same fusion protein, Cx43-TC as shown in (F). The cells were counterstained with propidium iodide (red) to highlight the cell bodies and nuclei. Bar, 5 μm.

We engineered a recombinant connexin43 (Cx43), named Cx43-TC, by genetically fusing a 17 amino acid–long tetracysteine receptor domain (5′-EAAAREACCRECCARA-3′) (15) to the COOH terminus of Cx43 (Fig. 1B) (16). Cx43 is one of the best known and most widely distributed member of the connexin family, found in most mammalian organs. Fusions of GFPs to the COOH termini of connexins are functional, whereas analogous fusions to the NH2termini are nonfunctional (17). Cx43-TC was expressed in intercellular communication-deficient, Cx43-negative HeLa cells. Western blots of HeLa cells expressing Cx43-TC or wild-type Cx43 revealed that Cx43-TC migrated slightly slower than wild-type Cx43 reflecting the addition of the tetracysteine-tag (Fig. 1C). Fluorescent images of live cells stained with FlAsH-EDT2 were virtually identical to the images of the same cells acquired after fixation, permeabilization, and staining with a Cx43-specific monoclonal antibody (Fig. 1, D through F) (18). Such congruence demonstrated that FlAsH staining was highly specific for the tetracysteine-tagged Cx43.

Cx43-TC assembles into functional gap junctions.

Cx43-TC expressing HeLa cells were labeled with FlAsH-EDT2and then injected with Lucifer yellow (LY), which transferred within 1 to 5 s between contacting cells revealing that Cx43-TC had assembled into functional gap junctions. A large macromolecule, rhodamine dextran (molecular weight, 70 kD) did not pass through junctions and thus acted as a control to mark injected cells. Wild-type HeLa cells injected with LY did not show dye transfer at any time. A more quantitative measure of junctional competence was to record the electrical coupling by whole-cell patch clamping of each of the neighboring cells. Junctions consisting of Cx43-TC and labeled with FlAsH-EDT2 had a mean conductance of 26 ± 16 nS (mean ± SD, n = 18), not significantly different (P > 0.1) from junctions formed from Cx43-GFP (32 ± 25 nS, n = 8) (19). Thus, the recombinant protein formed gap junctions, which were specifically recognized by FlAsH and whose ability to transfer dye and electrical current was not affected by either the insertion of the tetracysteine receptor domain or by the labeling conditions and the binding of the ligand.

Gap junction dynamics.

The ability to label Cx43-TC in live cells quickly and stoichiometrically with green FlAsH-EDT2 and red ReAsH-EDT2 enabled us to examine how Cx43 gap junction channels are added and removed from a gap junction plaque (20). We first labeled all preexisting Cx43-TC with FlAsH-EDT2, then washed away unbound dye and incubated the cells for either a 4- (Fig. 2, A and B) or 8-hour (Fig. 2, C and D) interval. This labeling protocol pulse-labeled all preexisting Cx43-TC within the cell, but it would not label Cx43 molecules synthesized during the incubation interval of 4 or 8 hours. Subsequent labeling of the newly synthesized pool of Cx43-TC with red ReAsH-EDT2 would not stain the older Cx43-TC molecules because their tetracysteine sites were still occupied by FlAsH. This pulse-chase labeling thus revealed the differential distributions of connexins synthesized at different times. The oldest proteins, stained green, were concentrated in the central region of the gap junction plaques flanked, as seen in thin optical sections, by two regions of red fluorescence from the most recently synthesized connexins. Three-dimensional reconstructions (Fig. 2, E through G) showed that the red regions were actually rings surrounding roughly circular green core zones. The partitioning between green and red fluorescence showed that younger protein molecules were added adjacent to older ones with remarkably little mixing. As the time interval between FlAsH-EDT2 and ReAsH-EDT2 applications was increased to 8 hours, the red periphery gradually displaced the central green core, leaving small intracellular punctae (presumably endocytic products) as the only remaining sites of green fluorescence (Fig. 2D). Other intracellular red punctae carrying young Cx43-TC might represent exocytic vesicles. As a control, the temporal order of FlAsH-EDT2 and ReAsH-EDT2applications was reversed, which produced a reversed color pattern (Fig. 2H). Almost no intracellular punctae contained both colors, which is consistent with the separation of endocytic and exocytic traffic (Fig. 3A). Furthermore, young Cx43-TC were taken to the cell surface not necessarily in proximity to a preexisting plaque (Fig. 3A). Small junctions showed similar partition of green- and red-labeled Cx43-TC, indicating that the size of the plaque did not affect the segregation pattern of young and old protein (Fig. 3B). However, when two junctions were closely spaced, the segregation of green- and red-labeled Cx43-TC occurred asymmetrically so that the apposed portions of the plaques elongated less than the distal portions (Fig. 3C). This suggested that less Cx43-TC is available in the smaller space between two adjacent plaques, reinforcing the hypothesis that new channels are transported to the plasma membrane and then flow to the sides of the plaque. Thus, under steady-state conditions of connexin expression, newly synthesized connexons are continually being taken to the cell surface, assembled into channels specifically at the edges of the gap junction; from there, they migrate in an orderly progression toward the center of the plaque, where they are removed by endocytosis. This inward radial flow is in the opposite direction from what is usually pictured at synapses, where central exocytosis is thought to be balanced by peripheral endocytosis (21).

Figure 2

FlAsH and ReAsH label two temporally separated pools of Cx43-TC and allow recording of junctional plaque renewal over time. HeLa cells stably expressing Cx43-TC were stained with FlAsH-EDT2, incubated for 4 (A andB) or 8 (C and D) hours in complete medium at 37°C, stained with ReAsH-EDT2, and imaged using a BioRad MRC-1024 confocal microscope. Panels (A) through (D) show Cx43-TC junctional plaques at different stages of refurbishing, as indicated by the different ratios of FlAsH (green) and ReAsH (red) stains. Panels (E) through (G) show a 3D volume projection of two Cx43-TC containing plaques rotated through various angles, from 0° to 90° showing that the newer protein (red) is added as a ring around the periphery of existing junctions (green). Reversing the order of staining with FlAsH-EDT2 and ReAsH-EDT2 results in the reversal of the staining pattern (H), showing that the color pattern reflects the age of the connexin rather than any intrinsic bias of ReAsH-EDT2 for connexins at the periphery or FlAsH-EDT2 for connexins at the center of a plaque. Bars, 1 μm.

Figure 3

Distribution, segregation and asymmetry of newer Cx43-TC in refurbishing plaques. Cells were labeled with FlAsH-EDT2 (green) for 1 hour, incubated for 4 hours in complete medium at 37°C, stained with ReAsH-EDT2, and imaged using a BioRad MRC-1024 confocal microscope. (A) Stereopair confocal image shows that the older protein is predominantly found in the center of a gap junction plaque and near the cell center (green), whereas the newer protein is found at the periphery of the plaque and dispersed in vesicles throughout the cytoplasm (red). These red vesicles can also be seen near the plasma membrane of both cells. (B) Plaques of different sizes (arrows) segregate similarly. (C) When two gap junctions occur in close apposition, an asymmetry in the segregation occurs with smaller areas of newly incorporated channels occurring at the proximal portions of the two plaques (arrowheads) and larger areas at the distal portions (arrows). Bars, 2 μm.

Correlated electron microscopy.

ReAsH is not only fluorescent but can also be used for photoconversion of diaminobenzidine (DAB) to allow direct correlation of live cell images with high-resolution EM detection (22). Photoconversion consists of intense illumination of a dye in fixed specimens in the presence of oxygen and DAB. Dye-catalyzed formation of singlet oxygen causes highly localized polymerization of DAB into an insoluble osmiophilic precipitate visible by EM (23–25). Previously, detection of specific macromolecules by photoconversion has required diffusion of antibodies, protein toxins, or antisense oligonucleotides into the fixed tissue (25–28). Neither GFPs nor FlAsH are capable of photooxidizing DAB with any efficiency. Live cells labeled with ReAsH could be imaged for long times at multiple focal planes with the use of confocal microscopy (Fig. 4A) or multiphoton excitation. Subsequent fixation with high concentrations of strong cross-linking fixatives such as glutaraldehyde (2%) did not affect the stability of the ReAsH-Cx43-TC complex (Fig. 4A). The fixed cells were then placed in a solution of oxygenated DAB, illuminated until a reaction product became visible by bright field microscopy, treated with osmium tetroxide, and prepared for EM. This preparation showed an electron-dense reaction product at the same gap junctions previously identified by fluorescence (Fig. 4B). At high magnification, structures corresponding in size and shape to individual channels at apposed plasma membranes were visible in cross sections of gap junction plaques (Fig. 4C). For comparison, a high-quality conventional immunogold image (Fig. 4D) showed that only a small, random fraction of the densely packed connexins were labeled. Maintenance of antigenicity and diffusional access of antibodies are antithetical to good ultrastructural preservation. By contrast, ReAsH-EDT2staining was performed while the cells were alive, enabling easy correlation of optical and EM images and the use of vigorous fixatives. The only molecules required for the amplification process after fixation are oxygen, DAB, and osmium tetroxide, all of which are small and highly permeant.

Figure 4

ReAsH can photoconvert DAB to produce an electron dense reaction product allowing for tetracysteine-tagged connexins to be imaged in the same cells by both fluorescent and electron microscopy. (A) Confocal image of two ReAsH-labeled Cx43-TC containing gap junction plaques. (B) Electron micrograph of the same area after photoconversion. (C) The high magnification view of a portion of a gap junction reveals dense packing of Cx43 hemichannels within gap junction plaques and highlights the high resolution obtainable with this labeling method. For comparison, a Lowicryl-embedded gap junction labeled for Cx43 with the traditional immunogold is shown in (D). Bar in (A) 10 μm, in (B) 1 μm, and in (C) and (D) 100 nm.

Junctional trafficking.

To begin characterizing how the traffic routes of nascent and senile connexins are spatially segregated at the ultrastructural level, it was desirable to image Cx43-TC within transport intermediates and organelles at EM resolution. The pulse-labeling protocol described earlier and the ability of ReAsH but not FlAsH to photooxidize DAB (as shown in Fig. 5, A through C) provided the opportunity to selectively examine older or newer connexons and their subcellular environment by EM. For example, if green FlAsH-EDT2 incubation was followed 4 hours later by red ReAsH-EDT2, then most recently synthesized red connexins could be followed preferentially throughout the exocytosis pathway by light microscopy and then be photoconverted and identified by EM. These presumably outbound vesicles typically ranged in size from 100 to 150 nm (120 ± 22.8, mean ± SD; n= 53; area examined, 100 μm2) and could be seen near the Golgi apparatus (Fig. 5D), dispersed throughout the cytoplasm (Fig. 5E) and fused to the plasma membrane (Fig. 5F). The distribution of these vesicles at various locations, not necessarily adjacent to an existing plaque, suggests that connexons go first to nonjunctional sites of the plasma membrane and then flow to the plaque. To highlight the Cx43 degradation pathway, Cx43-TC was labeled with ReAsH-EDT2and after 4 hours, newly synthesized Cx43-TC was labeled with FlAsH-EDT2. The latter labeling step was not necessary for EM but was helpful for correlative fluorescence images. Internalization of plaque material could be visualized by EM (Fig. 5, G through I), which revealed stained lysosomal-like bodies (Fig. 5I). Some internalized gap junctions were clearly large double-membrane annular gap junctions, as previously described (Fig. 5H) (29), and internalized vesicles were quite variable in size.

Figure 5

Reaction products formed by the photoconversion of DAB by ReAsH allow detailed electron microscopy analysis of plaque refurbishing during double labeling experiments. When FlAsH-EDT2 was used first to saturate preexisting Cx43-TC in transport vesicles as well as previously assembled gap junctions, ReAsH-EDT2 only labeled newly synthesized recombinant Cx43. (A) Confocal image of a labeled junctional plaque. (B) Correlated EM image after photoconversion. (C) Higher magnification of the boxed region in (B), showing staining of the peripheral portions of the junction (arrows). (D) through (F) show staining of vesicles (arrows) associated with the Golgi apparatus (D), putative transport vesicles (E), and an exocytic event (F). By reversing the labeling order so that ReAsH was used first and was followed by an adequate incubation period in absence of free label, preexisting recombinant protein mainly associated to the degradative pathways and assembly intermediates were labeled, as shown in panel (Gthrough I). Staining was observed at plaque material being internalized (G) and (H), and at lysosomal-like structures (I). Bars in (A) through (C) 2 μm and in (D) through (I) 0.5 μm.


The approach described here for studying the life cycle of proteins, including assembly and internalization, by correlated light and electron microscopy has considerable potential for generalization because the targeting motif is tiny, genetically encoded, and versatile. Previous techniques for distinguishing the age of a genetically encoded tag include the spontaneous green-to-red ripening of a mutated Discosoma fluorescent protein (“fluorescent timer”) (30) and photobleaching or photoenhancement ofAequorea and Discosoma protein fluorescence (31–33). The spontaneous color change is elegantly simple but occurs gradually with a fixed time constant, so the time window of sensitivity is not easily manipulated, and abrupt boundaries such as in Fig. 2, A through C and E through G would be blurred. Moreover, aggregation of the DsRed prevents functional expression of many fusion proteins including connexin chimeras. The latter could only be rescued by cotransfection with an excess of unlabeled or GFP-labeled connexins (34). Optical marking of fluorescent proteins has faster time resolution but less contrast than our pulse-labeling protocol, whose ability to correlate multicolor with EM images is a unique advantage.

We have used this novel pulse-chase labeling approach at both the light and electron microscopy levels to show for the first time that gap junction plaques are assembled from the outer edges and removed from the central core. High-resolution images have revealed that 100- to 150-nm vesicles participate in the delivery of connexins to the cell surface. Gap junction internalization involves larger structures of more varied sizes and morphologies. These studies open avenues to uncover the complexity and sophistication of membrane protein assembly and turnover.

  • * Howard Hughes Medical Institute, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093–0648 USA.

  • To whom correspondence should be addressed. E-mail: mhellisman{at}


View Abstract

Navigate This Article