The Second PDZ Domain of Zonula Occludens-1 Is Dispensable for Targeting to Connexin43 Gap Junctions

Abstract

Zonula occludens (ZO)-1 is emerging as a central player in the control of gap junction (GJ) dynamics. Previously the authors reported that ZO-1 localizes preferentially to the periphery of Cx43 GJs. How ZO-1 arrives at GJ edges is unknown, but this targeting might involve we established interaction between the Cx43 C-terminus and the PDZ2 domain of ZO-1. Here the show that despite blocking the canonical PDZ2-mediated interaction by fusion of GFP to the C-terminus of Cx43, ZO-1 continued to target to domains juxtaposed with the edges of GJs comprised solely of tagged Cx43. This edge-association was not abolished by deletion of PDZ2 from ZO-1, as mutant ZO-1 also targeted to the periphery of GJs composed of either tagged or untagged Cx43. Additionally, ZO-2 was found colocalized with ZO-1 at GJ edges. These data demonstrate that ZO-1 targets to GJ edges independently of several known PDZ2-mediated interactions, including ZO-1 homodimerization, heterodimerization with ZO-2, and direct ZO-1 binding to the C-terminal residues of Cx43.

Keywords:

• connexin
• gap junction
• PDZ domain
• ZO-1
• ZO-2

INTRODUCTION

Gap junction (GJ) patterning plays a critical role in excitable tissues, most notably ventricular myocardium, where GJ plaques comprised of connexin43 (Cx43) intercellular channels occur in highly ordered spatial arrangements. The processes that determine GJ organization are conspicuously active during postnatal cardiac growth (Angst et al. 1997; Gourdie et al. 1992) and appear to go awry in certain diseases of the heart, contributing to the development of arrhythmias (Severs et al. 2006). Currently a gulf exists between our understanding of diseased states and the specific alterations in GJ organization that may contribute to them. However, the molecular mechanisms that govern GJ remodeling are starting to emerge. Elegant fluorescence imaging has indicated that accretion of connexin channels (connexons) likely occurs predominately at plaque edges (Gaietta et al. 2002; Lauf et al. 2002). Given the topology of GJs, peripheral accretion makes sense, and points to the GJ edge as a specialized assembly domain. The plaque edge also is where GJs can interface with other junctional systems, such as cadherin-based cell-cell adhesions, that may influence GJ dynamics. The challenge now is to identify the molecules that control incorporation of channels into plaques and subsequent GJ remodeling events.

One molecule that has received particular attention is the actin-binding MAGUK protein, zonula occuldens (ZO)-1 (Fanning et al. 2002; Stevenson et al. 1986). ZO-1 has been shown to interact with several connexins, but the best characterized is Cx43 (Giepmans 2004). The second PDS95/dlg/ZO-1 (PDZ2) domain of ZO-1 is known to interact with the cytoplasmic C-terminal residues of Cx43 (Giepmans and Moolenaar 1998; Sorgen et al. 2004). However, the physiological significance of direct PDZ2-mediated interaction between ZO-1 and Cx43 has remained elusive. Previously we reported that ZO-1 localizes preferentially to the periphery of Cx43 GJs (Zhu et al. 2005), and that blockade of PDZ2-mediated interaction decreased ZO-1 colocalization specifically within individual Cx43 plaques (Hunter et al. 2005). Inhibition of the Cx43-ZO-1 interaction also gave rise to a GJ remodeling defect—a loss of size control—which led to aberrantly large GJs, possibly due in part to uncontrolled accretion of channels (Hunter et al. 2005).

Localization at plaque edges places ZO-1 in a prime location to participate in the active process of GJ remodeling. Yet nothing is known about how ZO-1 arrives at the periphery of Cx43 GJs. Here we show that ZO-1 can target to the edges of Cx43 GJs independently of PDZ2-mediated interactions. Our evidence indicates that PDZ2-independent targeting to the plaque edge results in a distinct pattern of colocalization between Cx43 and ZO-1 suggestive of close proximity, but not direct interaction between the two proteins. From this we infer a targeting sequence that initially involves ZO-1 bound to junctional complexes (possibly N-cadherin–based) adjacent to GJs, followed by transfer of ZO-1 and direct engagement with Cx43 at GJ edges.

METHODS

ZO-1 Constructs

To generate DsR-ZO-1 (Fig. 1B), full-length human ZO-1 cDNA (including the alternatively spliced alpha domain) in pSK+Bluescript was digested with Not I and BstX I, then blunt-end ligated into the Sma I site of pDsRed1-C1 (Clontech Laboratories, Mountain View, CA). The PDZ2-deletion mutant DsR-ZO-1Δ PDZ2 was generated by PCR mutagenesis (Stratagene QuikChangeII Site-Directed Mutagenesis Kit; Stratagene, La Jolla, CA) of the DsR-ZO-1 plasmid using a primer pair (5′-GCTTCCAGCCAGCCTGCTAAACCTCGGGCTA CGCTATTGAATGTCCC-3′ and its complement) which resulted in deletion of ZO-1 residues K186-264E.

Figure 1 Domain structure of Cx43 and ZO-1 constructs. (A) Full-length Cx43, with or without EGFP fused to its C-terminus. The PDZ2 binding region spans the last ∼ 20 residues in Cx43, which include the C-terminal DLEI consensus sequence, as well as upstream residues that mediate non-canonical PDZ2 interactions. TM, transmembrane domain. (B) DsRed was fused to the N-terminus of full-length ZO-1 and mutant ZO-1 lacking PDZ2. SH3, Src homology domain; GUK, guanylate kinase domain.

Cell Culture

HeLa cells stably expressing full-length Cx43 or Cx43-EGFP (Fig. 1A) were cultured on no. 1.5 glass coverslips (Corning, Corning, NY) coated with poly-l-lysine (Sigma-Aldrich, St. Louis, MO). Cells were maintained at 37°C and 5% CO₂ in Dulbecco's modified Eagle's (DME) medium supplemented with 10% fetal bovine serum (FBS), penicillin, and streptomycin (Invitrogen, San Diego, CA), and transfected using Lipofectamine PLUS reagents in OptiMEM I medium (Invitrogen). After ∼ 40 h in culture, cells were fixed in 2% paraformaldehyde in phosphate-buffered saline (PBS) for 10 min.

Immunofluorescence

Fixed cells were blocked with 1% bovine serum albumin (BSA), 0.1% Triton X-100 in PBS. Immunolabeling was performed with Cx43 antibodies diluted 1:10,000 (C6219; Sigma-Aldrich) or 1:500 (IF1; Fred Hutchinson Cancer Research Center), ZO-1 antibody (61-7300; Zymed Laboratories, South San Francisco, CA) diluted 1:150, and ZO-2 antibody (Zymed Laboratories) diluted 1:150, followed by secondary antibodies conjugated with Alexa488, Alexa546 (Invitrogen), or Cy5 (Jackson ImmunoResearch Laboratories, West Grove, PA). Block buffer was used for all antibody dilutions and wash steps. Samples were mounted in SlowFade (Invitrogen).

Confocal Microscopy

Optical stacks were acquired using an UltraVIEW LCI spinning disk confocal system (PerkinElmer Life and Analytical Sciences, Waltham, MA) attached to an Axiovert 200M microscope equipped with a 63×/1.4 numerical aperture oil immersion objective (Carl Zeiss Microimaging, Thornwood, NY). Laser excitation at 488, 568, and 647 nm was AOTF-controlled, with 50 to 150 m exposures. Z-series were acquired in 100 nm increments and digital images were captured with an Orca ER CCD camera (Hamamatsu, Bridgewater, NJ), resulting in 102 × 102 × 100 nm x-y-z voxels. Alternatively, optical stacks were acquired with a TCS SP2 AOBS laser scanning confocal microscope equipped with 63×/1.4 numerical aperture oil immersion objective (Leica Microsystems, Deerfield, IL) as previously described (Hunter et al. 2005), resulting in 116 × 116 × 122 nm x-y-z voxels.

Images

Image manipulations, including false coloration, adjustments to contrast and brightness (gamma was unaltered), maximum projections and 3D rendering, were performed using Volocity (Improvision, Waltham, MA), with final images exported as TIF files. Channel visibility and image size were adjusted (without altering pixel dimensions) in Photoshop (Adobe Systems, San Jose, CA).

RESULTS AND DISCUSSION

In this study we used HeLa cells, which express essentially no connexins, to allow us to control via transfection the expression of GFP-tagged and untagged Cx43 (Figure 1A) and DsRed-tagged ZO-1 (Figure 1B) constructs. Confocal immunofluorescence of HeLa cells stably expressing full-length untagged Cx43 (HeLa Cx43) revealed a typical arrangement of Cx43 plaques with peripherally associated endogenous ZO-1 (Figure 2A). When tagged with DsRed and transiently expressed in HeLa Cx43 cells, ZO-1 localization (Figure 2B) remained qualitatively indistinguishable from the localization pattern of endogenous ZO-1 (Figure 2A), consistent with other studies using tagged ZO-1 constructs (McNeil et al. 2006; Riesen et al. 2002; Utepbergenov et al. 2006). Both endogenous ZO-1 and transiently expressed DsR-ZO-1 exhibited edge localization that included regions of partial overlap with gap junctional Cx43 immunofluorescence (Figure 2A and Figure 2B), indicating colocalization of the two molecules and the possibility of direct physical interaction through ZO-1 PDZ2 binding to the C-terminus of Cx43 (Giepmans and Moolenaar 1998). However, the majority of ZO-1 at GJ peripheries does not directly overlap with junctional Cx43, but rather appears to accumulate in membrane domains that are adjacent to, but not part of, the GJ plaque. Perijunctional localization suggests that ZO-1 targets to GJ edges independently of its PDZ-mediated interaction with Cx43, although it is unclear if these adjacent domains include unincorporated connexons that are below the level of detection. This led us to ask whether physical interaction with Cx43 is required for targeting of ZO-1 to GJ edge domains.

Figure 2 Full-length ZO-1 targets to edge domains of gap junctions composed of either Cx43 or Cx43-GFP. (A) Immunofluorescence of Cx43 (red) and endogenous ZO-1 (green) in HeLa Cx43 cells. HeLa Cx43 (B) and HeLa Cx43-GFP (C) cells transiently expressing DsR-ZO-1, with Cx43 (red) detected by immunofluorescence (B) or GFP fluorescence (C), and ZO-1 (green) by DsRed fluorescence (B, C). Note that both endogenous ZO-1 (A, inset) and DsR-ZO-1 (B) display stretches of continuous overlap with the edges of Cx43 plaques, whereas DsR-ZO-1 associated with Cx43-GFP GJs manifests as punctate rather than diffuse accumulations along plaque edges that exhibit minimal overlap with Cx43. All images are maximum projections of z-series acquired by laser scanning (A) or spinning disk (B, C) confocal microscopy. Scale bars, 5 μ m.

We first tested the requirement of PDZ2-mediated interaction for ZO-1 targeting to GJs by stably expressing GFP-tagged Cx43 (Jordan et al. 1999) in HeLa cells (HeLa Cx43-GFP), absent wild-type Cx43. The fusion of green florescent protein (GFP) to the C-terminus of Cx43 has been shown to interfere with ZO-1 binding (Giepmans et al. 2001), presumably by blocking the critical isoleucine residue (Giepmans and Moolenaar 1998). We reported previously that C-terminal tagging of Cx43 resulted in the formation of aberrantly large GJs (Hunter et al. 2003) and loss of partial colocalization of ZO-1 at plaque edges (Hunter et al. 2005). Here we confirm that C-terminal tagging of Cx43 largely abolishes ZO-1 colocalization with plaque edges (Figure 2C); however, ZO-1 continues to target to domains adjacent to Cx43-GFP plaques (Figure 2C and Figure 4C), despite its presumed inability to interact with the tagged Cx43 C-terminus. The accumulation of DsR-ZO-1 in domains juxtaposed with Cx43-GFP plaque edges (Figure 2C) appeared reduced when compared to ZO-1 accumulation at the periphery of native Cx43 GJs, as assessed by either DsR-ZO-1 fluorescence (Figure 2B) or immunofluorescence detection of endogenous ZO-1 (Figure 2A; Hunter et al. 2005). Furthermore, DsR-ZO-1 manifested as discrete fluorescent punctae that did not appear to overlap directly with plaque edges defined by Cx43-GFP fluorescence (Figure 2C and Figure 4C), as opposed to the more diffuse but overlapping distribution of endogenous and DsRed-tagged ZO-1 around native Cx43 GJs (Figure 2A and Figure 2B; Hunter et al. 2005; Zhu et al. 2005). However, we cannot make definitive statements about the extent of true colocalization (or apparent lack thereof) at plaque edges due to the resolution limits of the light microscope.

The presence of ZO-1 at the periphery of plaques comprised solely of Cx43-GFP argues strongly against a role for PDZ2-mediated binding of ZO-1 to Cx43 in the process of ZO-1 targeting to GJ edges. Yet recent studies have demonstrated that the interaction of Cx43 with ZO-1 appears to involve conserved residues in Cx43 that are upstream of the canonical C-terminal PDZ binding motif (Fanning et al. 2007; Sorgen et al. 2004), and further indicate that Cx43 C-termini might interface with ZO-1 homodimerized by PDZ2 interactions (Fanning et al. 2007). Thus it remains possible that abrogation of PDZ-mediated Cx43-ZO-1 interaction by C-terminal tagging of Cx43 is due not only to masking of the terminal isoleucine residue (Giepmans and Moolenaar 1998), but also results from disruption of critical residues further upstream (and/or loss of ZO-1 dimerization). This could explain why we see continued, although reduced, ZO-1 accumulation at the periphery of Cx43-GFP plaques (Figure 2C): conditions in cells, unlike those of in vitro binding assays, could remain favorable for Cx43-ZO-1 interactions that do not require a known PDZ binding sequence. Alternatively, the presence of low levels of ZO-1 at the edge of Cx43-GFP plaques could result indirectly via macromolecular associations between GJs and other cell-cell junctions that incorporate ZO-1. Adherens junctions are logical candidates in this respect as they are known to play a pivotal role in GJ stability (Hertig et al. 1996; Jongen et al. 1991; Li et al. 2005; Meyer et al. 1992; Musil et al. 1990; Wei et al. 2005) and cadherin-associated catenin complexes have been shown to bind both Cx43 (Ai et al. 2000; Wei et al. 2005) and ZO-1 (Itoh et al. 1997).

To determine definitively whether PDZ2-mediated interactions are required for ZO-1 targeting to the edges of Cx43 plaques, we expressed DsR-ZO-1 that lacks the PDZ2 domain (DsR-ZO-1Δ PDZ2) in HeLa Cx43 cells. Surprisingly, deletion of the PDZ2 domain did not abolish ZO-1 edge localization—confocal imaging revealed extensive stretches of DsR-ZO-1Δ PDZ2 fluorescence arrayed along the edges of native Cx43 plaques (Figure 3A). DsR-ZO-1Δ PDZ2 also was found associated with the periphery of Cx43-GFP plaques (Figure 3B). Relative to endogenous ZO-1 (Figure 2A) or DsR-ZO-1 (Figure 2B) colocalized with native Cx43 GJs, there appeared to be less DsR-ZO-1Δ PDZ2 associated with regions interior to the edges of Cx43-GFP plaques. Interestingly, the qualitative distribution of DsR-ZO-1Δ PDZ2 at plaque edges appeared to differ depending on whether or not the Cx43 was C-terminally tagged, with mutant ZO-1 manifesting predominately as distinct punctae at the edges of Cx43-GFP plaques (Figure 3B) rather than the more diffuse accumulation seen at the edges of untagged Cx43 plaques (Figure 3A). This difference in patterning indicates that C-terminal tagging of Cx43 and removal of the PDZ2 domain from ZO-1 may elicit distinct effects, suggesting that ZO-1 distribution at plaque edges involves elements of the Cx43 C-terminus and ZO-1 PDZ2 domain that act independently of one another. Regardless of the mechanistic details, these results confirm that ZO-1 targets to Cx43 GJs independently of PDZ2-mediated interactions involving either homodimerization of ZO-1 (Fanning et al. 2007; Utepbergenov et al. 2006) or physical interaction between ZO-1 and Cx43, and hint at the existence of discrete ZO-1 complexes associated with GJ edges. The molecular composition of ZO-1 complexes juxtaposed with GJ edges is unclear, but preliminary colocalization data (not shown) suggest a prominent role for N-cadherin, which may participate in the delivery of ZO-1 to GJs.

Figure 3 PDZ2-mediated interactions are not required for ZO-1 targeting to Cx43 gap junctions. HeLa Cx43 (A) and HeLa Cx43-GFP (B) cells transiently expressing mutant ZO-1 lacking PDZ2 (DsR-ZO-1Δ PDZ2), with Cx43 (red) detected by immunofluorescence (A) or GFP fluorescence (B), and mutant ZO-1 (green) by DsRed fluorescence (A, B). Despite the absence of a PDZ2 domain, DsR-ZO-1Δ PDZ2 localizes at the periphery of Cx43 plaques (A), with some areas of edge overlap similar to that seen with full-length ZO-1 (A, arrowheads). DsR-ZO-1Δ PDZ2 also targets to the edges of Cx43-GFP plaques (B); however, 3D rotations show that the mutant ZO-1 accumulates in punctae juxtaposed but not overlapping with plaque edges (B, arrowheads). All images are maximum projections of z-series acquired by spinning disk confocal microscopy. Z-series were rendered in 3D and rotated to provide en face plaque perspectives and reveal the extent of true colocalization (3D panels). Scale bars, 5 μ m.

Earlier immunofluorescence studies revealed that ZO-2 associates with Cx43 GJs and that the second PDZ2 of ZO-2 binds to the same stretch of C-terminal residues in Cx43 as ZO-1 (Singh et al. 2005). We extend these findings by showing that ZO-2 localizes predominately to GJ edges in a manner similar to ZO-1 (Figure 4A). Moreover, ZO-2 localizes to the periphery of plaques composed solely of GFP-tagged Cx43 (Figure 4B and Figure 4C), indicating that, as is the case for ZO-1, binding of the ZO-2 PDZ2 domain to the C-terminus of Cx43 is likely dispensable for targeting of ZO-2 to GJ edges. ZO-1 and ZO-2 have been shown to form heterodimers via PDZ2 interactions (Fanning et al. 1998; Itoh et al. 1999). The presence of DsR-ZO-1Δ PDZ2 at GJ peripheries (Figure 3) provides solid evidence that PDZ2-mediated heterodimerization of ZO-1 and ZO-2 is not necessary for targeting of ZO-1 to plaque edges. Nor does ZO-1/ZO-2 heterodimerization appear to be required for maintaining edge localization of ZO-1, as there are regions of non-overlapping signal along GJ edges (Figure 4B and Figure 4C). However, much (if not all) of the edge-localized ZO-2 appears to colocalize with ZO-1 (Figure 4B and Figure 4C); thus it remains to be determined if heterodimerization with ZO-1 is required for targeting and/or maintenance of ZO-2 at plaque edges.

Figure 4 ZO-2 colocalizes with ZO-1 at the periphery of Cx43 and Cx43-GFP gap junctions. (A) Immunofluorescence of Cx43 (cyan) and endogenous ZO-2 (red) in HeLa Cx43 cells, showing ZO-2 at the periphery of Cx43 plaques. (B) HeLa Cx43-GFP cells transiently expressing DsR-ZO-1, with Cx43 (cyan) detected by GFP fluorescence, ZO-1 (green) by DsRed fluorescence, and ZO-2 (red) by immunofluorescence. Note that the majority of ZO-2 at the edges of Cx43-GFP plaques is colocalized with ZO-1 (B, arrowheads), whereas some ZO-1 at edges does not overlap significantly with ZO-2 (B, arrow). (C) Z-series (same as shown in B) was rendered in 3D and rotated to show the relative distributions of DsR-ZO-1 and ZO-2 at plaque edges from an en face perspective. (Note: rotation of cytoplasmic structures into the foreground creates the false impression of ZO-1 and ZO-2 localization in the plaque interior.) All images are maximum projections of z-series acquired by laser scanning (A) or spinning disk (B, C) confocal microscopy. Scale bars, 5 μ m.

In this study, we have addressed the molecular elements that direct ZO-1 to Cx43 GJs. The data presented here effectively eliminate binding interactions that require the ZO-1 PDZ2 domain, thereby putting constraints of the types of mechanisms that might mediate ZO-1 targeting to membrane domains that encompass or juxtapose GJ edges. For instance, it is improbable that ZO-1 is delivered to plaque edges bound directly to Cx43. Still unresolved however is the specific activity of ZO-1 after it arrives at GJ edges.

Current evidence points to an important role for ZO-1 in modulating the patterning of Cx43 GJs. Previously we showed that Cx43-ZO-1 interaction increased following enzymatic dissociation of ventricular muscle cells (Barker et al. 2001). Based on these data, we concluded that ZO-1 interaction with Cx43 is dynamic and likely related to actin-based remodeling of GJ complexes. Subsequently, we showed that disruption of Cx43-ZO-1 interaction in myocytes and HeLa cells led to decreased ZO-1 colocalization with GJs, concomitant with increased plaque size; we proposed that, as a consequence of disengagement of ZO-1 from Cx43 at the GJ edge, plaques expanded due to unregulated channel accretion (Hunter et al. 2005). It has been variously proposed that ZO-1 enhances GJ assembly (Laing et al. 2005), plays a role in GJ disassembly (Akoyev and Takemoto 2007), or mediates GJ internalization (Barker et al. 2002; Segretain et al. 2004). ZO-1 also has been implicated in the regulation of intercellular communication through Cx43 GJs (Akoyev and Takemoto 2007; Girao and Pereira 2007; Laing et al. 2005; Toyofuku et al. 2001; van Zeijl et al. 2007). Given the predominant edge-localization of ZO-1, it is possible that many of these processes are integrated at plaque edges, with modality determined by which molecules ZO-1 actively links to (or disengages from) Cx43.

Work in vitro on the physiological significance of ZO-1 association with Cx43 GJs is increasingly being translated in vivo, emphasized by several studies in the heart. The present findings regarding ZO-1 engagement of Cx43 at plaque edges accord with our previous finding that only a fraction of the total ZO-1 present in intercalated disks of ventricular cardiomyocytes colocalizes with Cx43 plaques (Barker et al. 2002). More recently, others have focused on the spatial arrangement of ZO-1 relative to Cx43 GJs within cardiac tissues. Maass et al. (2007) reported that haplodeficient expression of a Cx43 truncation mutant missing a large stretch of C-terminal residues, including those required for ZO-1 binding, resulted in the apparent segregation of ZO-1 from Cx43 and generation of abnormally large plaques at the margins of intercalated disks. Likewise, it was reported that increased levels of Cx43-ZO-1 interaction in patients with congestive heart failure correlated with a decrease in the extent of Cx43 GJs at the intercalated disk (Bruce et al. 2008). Related studies in diseased heart also suggest that changes in ZO-1 expression levels affect GJ remodeling (Kostin 2007; Laing et al. 2007). For the most part, these in vivo data agree with our earlier in vitro data that showed Cx43-ZO-1 interaction influences GJ size control (Hunter et al. 2005).

Although much remains to be resolved, identification of the key proteins and processes that define the mechanistic framework for remodeling of the Cx43 GJ plaque is progressing—and continued advances in our understanding of the mechanisms that control GJ patterning is certain to have important implications for several human diseases, including cardiac arrhythmias and epilepsy.

We thank Jane Jourdan and Dr. Yuhua Zhang for technical assistance; Dr. Melvin Anderson for human ZO-1 cDNA; Dr. Paul Lampe for IF1 antibody; and Dr. Dale Laird for Cx43-GFP. This work was supported by National Institutes of Health grants HL07260 and K12GM081265 (to A.W.H), and HL56728 and HL082802 (to R.G.G).