Desmosomes are intercellular adhesive complexes that anchor the intermediate filament cytoskeleton to the cell membrane in epithelia and cardiac muscle cells. The desmosomal component desmoplakin plays a key role in tethering various intermediate filament networks through its C-terminal plakin repeat domain. To gain better insight into the cytoskeletal organization of cardiomyocytes, we investigated the association of desmoplakin with desmin by cell transfection, yeast two-hybrid, and/or in vitro binding assays. The results indicate that the association of desmoplakin with desmin depends on sequences within the linker region and C-terminal extremity of desmoplakin, where the B and C subdomains contribute to efficient binding; a potentially phosphorylatable serine residue in the C-terminal extremity of desmoplakin affects its association with desmin; the interaction of desmoplakin with non-filamentous desmin requires sequences contained within the desmin C-terminal rod portion and tail domain in yeast, whereas in in vitro binding studies the desmin tail is dispensable for association; and mutations in either the C-terminus of desmoplakin or the desmin tail linked to inherited cardiomyopathy seem to impair desmoplakindesmin interaction. These studies increase our understanding of desmoplakin-intermediate filament interactions, which are important for maintenance of cytoarchitecture in cardiomyocytes, and give new insights into the molecular basis of desmoplakin- and desmin-related human diseases.
Desmosomes are multi-protein complexes that play a key role in promoting cell-cell adhesion and in tethering the intermediate filament (IF) system to the cell membrane. These structures are abundant in tissue exposed to mechanical stress, such as the epidermis and heart. They contribute to maintaining the integrity of embryonic tissues during development (reviewed in Getsios et al., 2004). The adhesive core of desmosomes comprises transmembrane glycoproteins of the cadherin family, the desmogleins and the desmocollins. Two members of the armadillo family of proteins, plakoglobin and plakophilins, as well as desmoplakin (DP) link the IF cytoskeleton to the adhesive core of desmosomes by means of linear and lateral interactions between these various molecules (Getsios et al., 2004).
DP, the most abundant desmosomal plaque component, is a member of the plakin family, which also includes plectin, BP230, periplakin and epiplakin (reviewed in Jefferson et al., 2004). These proteins serve as cytolinkers and/or scaffolding proteins, connecting IF to other cytoskeletal networks and/or distinct sites at the plasma membrane. Plakins are predicted to form homodimers with a central coiled-coil domain flanked by globular end domains (Green et al., 1992a; Green et al., 1992b; Ruhrberg and Watt, 1997). The N-terminal region of DP encompasses a so-called plakin domain harboring α-helical bundles (Green et al., 1990) and mediates its localization to desmosomes by binding to plakoglobin and plakophilins (Stappenbeck and Green, 1992; Stappenbeck et al., 1993; Kouklis et al., 1994; Cowin and Burke, 1996; Kowalczyk et al., 1997; Smith and Fuchs, 1998). The C-terminus of DP contains a plakin repeat domain (PRD) consisting of three homologous subdomains, denoted A, B and C, that are interrupted by intervening sequences of varying length (Fig. 1). These subdomains consist of 4.5 copies of a 38-amino acid sequence repeat motif (Green et al., 1990; Choi et al., 2002). The DP tail is able to associate with various IF proteins, such as epidermal and simple epithelial keratins and vimentin (Stappenbeck and Green, 1992; Stappenbeck et al., 1993; Kouklis et al., 1994; Kowalczyk et al., 1997; Meng et al., 1997; Fontao et al., 2003). Recent studies have identified sequences that allow DP to bind to distinct IFs. Specifically, the C subdomain within the C-terminal extremity of DP binds to K5/K14 and K8/K18, while its linker subdomain is able to associate with K8/K18 and vimentin (Fontao et al., 2003). Furthermore, DP-IF interactions are regulated by phosphorylation of Ser2849 within the DP C-terminus (Stappenbeck at al., 1994; Fontao et al., 2003; Godsel et al., 2005). Crystallographic studies have shown that the B and C subdomains exhibit a globular structure with a conserved groove, the features of which seem to be suitable for an interaction with vimentin (Choi et al., 2002), but the importance of which has yet to be empirically tested. The importance of understanding the mechanisms governing DPIF associations is highlighted by recent studies indicating that proper attachment of IFs to the desmosomal plaque via DP is required for maintaining cell-cell adhesive strength (Huen et al., 2002).
The importance of DP in vivo is supported by the observation that DP-null mutant mice die at embryonic day 6.5 owing to defects of the extraembryonic endoderm. Furthermore, chimeric morulae expressing DP in extraembryonic tissues do not survive beyond day E9.5 as a result of defects in the developing epidermis, neuroepithelium and heart with perturbation of desmosome assembly and severing of the IF-cell membrane attachment (Gallicano et al., 1998; Gallicano et al., 2001). Finally, pathogenic mutations in the DP gene cause cardiomyopathy and palmo-plantar keratoderma (Armstrong et al., 1999; Norgett et al., 2000).
IF, which are present in all vertebrate cells and tissues, are classified into five families (Herrmann and Aebi, 2004). All IF polypeptides are built from a central rod domain, consisting of α-helical segments, flanked by globular non α-helical N-terminal head and C-terminal tail domains. The assembly of IF into long 10-12 nm caliber polymers, is initiated by the central rod of IF proteins. The process progresses by formation of dimers and tetramers, which further associate following three principal assembly modes. While the N-terminal head domains seem to control IF assembly and stabilization, their tails are thought to regulate lateral packing and stabilization of higher order filament structure (Herrmann and Aebi, 2004).
Desmin, an IF protein type III, is one of the earliest expressed muscle-specific proteins. It is detected in early phase of skeletal (Sejersen and Lendahl, 1993) and cardiac (Kachinsky et al., 1995) muscle differentiation together with vimentin and nestin, the expression of which decline during development (Carlsson et al., 2000). Desmin-null mutant mice show disorganization of myofibril architecture in mechanically stressed striated muscles, including heart (Li et al., 1996; Milner et al., 1996; Wang et al., 2001). Furthermore, desmin gene mutations cause certain forms of muscular dystrophies, with or without cardiomyopathies. Although certain mutations compromise filament assembly (Park et al., 2000; Dalakas et al., 2000; Li and Dalakas, 2001; Bär et al., 2005) at various stages, in other cases the implicated pathogenic mechanisms are unclear (Dalakas et al., 2000; Bär et al., 2004).
Ultrastructural studies have shown that desmin filaments attach in a lateral fashion to the DP-containing region of the desmosomal plaques of the heart, which suggests that special domains exist in desmin that are involved in binding (Kartenbeck et al., 1983). Recent immunoelectron microscopy studies have revealed that, at intercalated discs, desmosomes and fasciae adherentes form a highly integrated system, for which the term of area composita has been proposed (Franke et al., 2006). At the molecular level, although DP was shown to bind to desmin in yeast two-hybrid assays (Meng et al., 1997), little information is available regarding the association of DP and desmin and whether the latter is regulated as in the case of vimentin (Fontao et al., 2003). Characterization of these interactions is key to deciphering the mechanisms by which the junctional region of the intercalated discs mediates the attachment of IF bundles, generating a stress-resistant scaffolding able to integrate the mechanical forces generated during heart contraction. Recent discovery of mutations in humans (reviewed by Sen-Chowdhry et al., 2005) and engineered gene disruption of desmosomal genes (Bierkamp et al., 1996; Ruiz et al., 1996; Grossmann et al., 2004) have unequivocally confirmed the functional interdependence of desmosomal proteins and their importance for the integrity of the cytoarchitecture of cardiac muscle cells.
Therefore, the aim of our study was to investigate the interaction between DP and desmin by defining functionally important sequences implicated in the binding of these molecules with each other by comparative analysis using the yeast two-hybrid system, cell transfection studies and biochemical assays. Furthermore, we sought to gain insights into the impact of pathogenic mutations in DP and desmin on these interactions.
DP and desmin are found codistributed in the intercalated disc regions of cardiac myocytes
Previous studies have identified sequences within DP that are important for its interaction with vimentin and keratins (Stappenbeck and Green, 1992; Stappenbeck et al., 1993; Meng et al., 1997; Choi et al., 2002; Fontao et al., 2003), but the interaction of DP with desmin has not yet been characterized. Hence, we first investigated the localization of DP and desmin in longitudinal sections of monkey cardiac muscle by double immunofluorescence microscopy. As expected (Kartenbeck et al., 1983), these two proteins were both found at the level of intercalated discs. Desmin also showed the typical cross-striated labeling in a distinct sarcomeric pattern (Fig. 2).
We then examined the distribution of ectopically expressed DP in a keratinocyte cell line stably expressing desmin, in which desmin forms a keratin-independent cytoskeleton (Magin et al., 2000). To ensure the best codistribution potential of DP, we used a DP mutant carrying the S2849G mutation within the C-terminal extremity, which abrogates a potential phosphorylation site critical for the association of DP with various IF proteins (Stappenbeck et al., 1994; Meng et al., 1997; Fontao et al., 2003; Godsel et al., 2005). Immunofluorescence microscopy studies of transfected cells expressing a DP construct encompassing the B and C subdomains and the C-terminal extremity, green fluorescent protein (GFP)-tagged at its N-terminus, demonstrated that recombinant DP-BCS2849G decorated both the epidermal keratin and the desmin network (Fig. 3). However, the association with desmin appeared to be less favored than that with cytokeratins, suggesting differential binding affinities of DP for these IF networks.
To investigate the targeting of the DP tail to specific subcellular regions in cultured striated muscle cells, the GFPDP-BCS2849G construct was transiently expressed in primary neonatal rat cardiac myocytes that upon the time point of transfection variably contain assembled myofibrils with mature Z discs (van der Ven et al., 2000). Confocal laser IF microscopy studies showed a precise colocalization of DP-BCS2849G with desmin (Fig. 3): the recombinant DP-BCS2849G decorated the intercalated disc regions as well as the filamentous desmin network, that in some transfected cells has partially rearranged into a transverse orientation (Fig. 3).
Distinct sequences within the DP tail are required for its coalignment with desmin in co-transfected IF-free SW13 cells
To further define the regions of the DP tail required for its coalignment with desmin, we performed co-transfection studies with a series of DP constructs, including the wild-type DP tail, DP-BC, Myc-tagged at their N-terminus, and full-length desmin or vimentin hemagglutinin (HA)-tagged at their C-terminus using IF-free SW13 cells. In transfected SW13 cells, desmin only formed a rudimentary network with short filamentous structures and perinuclear aggregates, as described (Schweitzer et al., 2001; Bär et al., 2006). When co-expressed, both DP-BC and DP-BCS2849G colocalized with desmin (Fig. 4A-D; Table 1) suggesting that, in IF-free SW13 cells, in analogy to other cell lines (Stappenbeck et al., 1994), the presence of S2849 did not have a major impact on the ability of the DP tail to codistribute with desmin. Deletion of the last 51 amino-acid residues from DP had no detectable effect on the codistribution (Table 1). The DP mutant containing the B subdomain and the linker region, DP-BL2194-2566, also codistributed with desmin (Fig. 4E, F; Table 1). By contrast, the DP mutant protein containing the C subdomain and the C-terminal extremity, DP-CS2849G, did not colocalize with desmin but was present as cytoplasmic aggregates in transfected SW13 cells (Fig. 4G,H; Table 1). The occasional presence of clumped perinuclear material with both DP-C and desmin was most likely related to the overexpression of the recombinant proteins due to the strong CMV promotor and the inability of this cell line to maintain extended filament arrays (Bär et al., 2006), leading eventually to the formation of aggresomes. The latter have been shown to consist of misfolded proteins encircled by collapsed IF proteins (reviewed in Kopito, 2000),
Recombinant proteins encompassing either the linker region, DP-L, or the B subdomain alone, DP-B, were not detected using a polyclonal 9E10.2 antibody (Table 1), possibly owing to a toxic effect, an increased susceptibility to proteolysis of the transgene product in transfected cells, or an impaired accessibility of the antibody to the single Myc tag as suggested (Stappenbeck et al., 1993). Finally, when compared with desmin, the various DP constructs showed the same coalignment potential when co-expressed with an exogenous vimentin network (Table 1). These results suggest that a region encompassing the B subdomain with the linker region of DP contains sequences sufficient to drive the codistribution of DP with either the desmin or the vimentin network.
The DP tail associates with desmin in in vitro binding assays
Since by using a nonionic detergent buffer containing 1% Triton X-100 we were unable to solubilize DP and desmin from cultured primary rat cardiac myocytes to perform coimmunoprecipitation studies, we then tested the ability of the DP tail to associate with desmin filaments in in vitro binding assays and compared it with that of vimentin as positive control. Desmin or vimentin filaments were immobilized on nitrocellulose membranes and overlaid with in vitro transcribed and translated recombinant DP proteins that were used as fluid-phase ligands (Fig. 5). Recombinant DP-BCS2849G interacted with desmin, although with an apparent lower affinity compared with vimentin. As depicted in Fig. 5, the binding efficiency of DP-BCS2849G to both desmin and vimentin varied accordingly to the amount of immobilized filaments. Although binding of DP to desmin abruptly decreased when the amount of desmin was below 0.5′g, the association of DP with vimentin was maintained for up to 0.05′g immobilized vimentin. Furthermore, DP-BCΔ51 and DP-BL exhibited greatly reduced binding for both desmin and vimentin compared with DPBCS2849G. Finally, the recombinant proteins encompassing the linker region alone, DP-L, or the C subdomain with either the entire C-terminal extremity, DP-CS2849G, or lacking the C-terminal 51-amino acids, DP-CΔ51, did not detectably interact with either desmin or vimentin. These results, which are in agreement with those obtained in cell transfection studies, indicate that DP-BL contains sequences sufficient for the interaction of DP with both desmin and, as previously reported, vimentin. However, the presence of both B and C repeats, as well as of the C-terminal extremity of DP, are required for efficient binding. Since bacterially expressed proteins are typically not phosphorylated, the presence of Ser2849 in DP is not expected to have a negative impact on the association under the condition of the overlay assays. In keeping with this idea, the ability of wild-type DP-BC recombinant to interact with desmin and vimentin (not shown) was not significantly different from that of DP-BCS2849G (Fig. 8).
Identification of sequences within the DP tail interacting with desmin in yeast two-hybrid assays
To further map the association between DP and desmin, we performed yeast two-hybrid assays using a series of mutants of DP expressed as GAL4 DNA-BD fusion proteins (Fig. 6). Some of the truncated proteins carried the amino acid substitution S2849G within their C-terminal extremity (Stappenbeck et al., 1994; Meng et al., 1997; Fontao et al., 2003). DP-BCS2849G, but not DP-BC, bound to desmin in this assay (Fig. 6). The constructs DP-BCΔ51, DP-BL, and DP-L (encompassing residues 2442-2630 of the linker region) did weakly bind to desmin only when tested as GAL4-AD fusion proteins against desmin fused to the GAL4-BD, as inferred from the activation of the reporter gene ADE2 only (Hudson et al., 2004). By contrast, DP-CS2849G, DP-CΔ51 as well as DP-CtS2849G (consisting of the 79-residue-long C-terminal extremity of DP) never showed binding activity to desmin.
The ability of the various DP constructs to bind to vimentin seemed to be more robust when compared to that of desmin, since DP-BCΔ51, DP-BL and DP-L fused to GAL4 DNA-BD were all able to associate with vimentin (Fig. 6) (Fontao et al., 2003).
Since the C-terminal extremity in DP-BC seems to influence its binding to vimentin (Fontao et al., 2003), we next tested a chimeric protein consisting of the B and C subdomains of BP230 fused at its C-terminal extremity to the last C-terminal 51-amino acid stretch of DP with the S2849G substitution, BP230 (BC)-DP CtS2849G. The C-terminal region of BP230, another plakin family member, contains a B and C subdomain and a linker region exhibiting high homology with those of DP, but does not associate with vimentin in yeast (Fontao et al., 2003). The results indicate that the last 51-amino acid stretch within the C-terminal extremity of DP contains sequences not only affecting the association with vimentin (Fontao et al., 2003), but also to desmin, since the chimeric protein BP230 (BC)-DP CtS2849G, but not BP230-BC, was able to bind desmin (Fig. 6).
These findings support and extend results from the transfection and in vitro approaches, demonstrating that: (1) although the linker region and C-terminal extremity are critically involved in binding, the presence of the B and C subdomains is also required to ensure a robust interaction of DP with desmin, and (2) in analogy to what is observed with other IF proteins (Stappenbeck et al., 1994; Meng et al., 1997; Fontao et al., 2003), the amino acid substitution S2849G within the C-terminal extremity of DP favors its binding to desmin.
Identification of sequences within desmin and vimentin binding to DP by yeast two-hybrid assays: importance of both their C-terminal rod and tail regions
To characterize sequences within desmin or vimentin important for their association with DP, we generated a series of desmin and vimentin deletion constructs (Fig. 7). We first verified that the constructs containing a portion of the rod domain of desmin or vimentin could make homodimers in yeast (not shown). Whereas deletion of their head domain had no effect on binding to DP-BCS2849G, truncation of the tail domain from either desmin or vimentin completely abolished the interaction. Furthermore, DP-BCS2849G was not able to associate with the heads or tails of either desmin or vimentin with or without their entire rod domain (Fig. 7). To better map the involved interaction sites, we generated constructs encoding the C-terminal portion encompassing the linker sequence L2 and the 2B segment of these IF proteins (Fig. 7). When these constructs were tested against DP-BCS2849G, desmin287-470 and vimentin283-467 showed binding activity in yeast, while desmin287-418 and vimentin283-413 did not (Fig. 7). These findings indicate that, in yeast assays, the interaction of DPBCS2849G with either desmin or vimentin depends on their tail domain, since their deletion abrogates binding; and sequences within the C-terminal region of the rod of both desmin and vimentin contribute to the interaction.
DP interacts with tailless desmin in in vitro binding assays
To verify the participation of the tail of desmin to its binding to DP, we tested the ability of DP-BCS2849G to associate with tailless desmin in pull-down assays. Glutathione S-transferase (GST) fusion proteins encoding the B and C subdomains and the C-terminal extremity of DP and BP230 as control were tested for their ability to associate with in vitro transcribed and translated recombinant desmin proteins used as fluid-phase ligands. In apparent contrast to the findings obtained in yeast, DP-BCS2849G, but not BP230-BC, formed a complex with both wild-type and tailless desmin (Fig. 8). We also carried out comparative overlay assays using wild-type and tailless desmin filaments, that were immobilized on nitrocellulose membranes and overlaid with in vitro transcribed and translated recombinant forms of DP-BC or DP-BCS2849G. There was no detectable difference in binding activities (Fig. 8). Together, these observations indicate that under the conditions of our in vitro binding assays the tail of desmin is dispensable for DP binding.
Deletion of the desmin tail has a negative impact on the colocalization potential of DP with the desmin network in transfected cells
To further investigate in vivo the impact of the desmin tail on the ability of DP to become coaligned with the desmin network, we performed transfection studies using cDNAs for tailless desmin and DP-BCS2849G. In SW13 cells expressing tailless desmin, only short thin filamentous structures and cytoplasmic aggregates were observed (Fig. 9A,C) (Bär et al., 2006). In the majority of co-transfected cells, DP and tailless desmin were found together in spot-like or large cytoplasmic aggregates (Fig. 9C,D). However, in a percentage of transfected cells, in which transgene expression was not high (up to 10 out of 100 analyzed cells), as judged by immunofluorescence microscopy, the recombinant DPBCS2849G was diffusely distributed in the cytoplasm and did not colocalize with desmin (Fig. 9A,B), suggesting that the codistribution potential of DP with tailless desmin in these cells was impaired. In this context, it is possible that the observed perinuclear cytoplasmic aggregates containing both DP-BCS2849G and tailless desmin reflects aggresome formation due to protein overexpression rather than true codistribution. In fact, a profound redistribution of the IF network is observed during the formation of aggresomes. The latter are finally interspersed and surrounded by IF (Kopito, 2000). Together, these findings provide evidence indicating that deletion of the desmin tail exerts a negative impact on the colocalization potential of DP with the desmin network.
Effects of disease-causing mutations on DP-desmin interaction
A recessive mutation in the DP gene leading to a truncated DP protein of 2541 residues lacking the C subdomain and C-terminal extremity has been shown to cause cardiac defects (Norgett et al., 2000). When a similar truncated protein encompassing residues 2194 to 2566 of DP was tested (see DPBL) in both overlay and yeast two-hybrid assays, the truncated protein exhibited a reduced ability to interact with desmin.
Since our findings suggested an implication of the desmin tail in its binding to DP, we then investigated the effect of a dominant desmin mutation associated with cardiomyopathy. The mutation consisted of a substitution of an Ile to a Met at position 451 within the desmin tail, desminI451M (Li et al., 1999; Dalakas et al., 2003). In yeast two-hybrid assays, the mutated molecule, which was able to self-dimerize (not shown) showed no binding activity with DP-BCS2849G (Fig. 7). We next carried out co-transfection studies using IF-free SW13 cells (Fig. 9). When single transfected cells expressing desminI451M were examined, the majority of cells exhibited, in addition to small spot-like aggregates, thin and short filamentous structures, which suggests that the mutated desminI451M molecule retained some ability to assemble into higher ordered structures (Fig. 9E) (Dalakas et al., 2003). In the majority of co-transfected cells, DP-BCS2849G was found together with desminI451M in cytoplasmic aggregates. However, in cells in which transgene expression was not high (5-10% of the cells), as judged by immunofluorescence microscopy, DP-BCS2849G exhibited a diffuse cytoplasmic distribution with no obvious colocalization with desminI451M similarly to that observed with tailless desmin (see above) (Fig. 9E,F). This observation provides further support to the idea that the desmin tail contributes to its interaction with DP. Together, mutations in either DP or the desmin tail may impair their association.
Binding of the desmin IF system to the intercalated disc regions in cardiac myocytes is thought to critically participate in the establishment of a stress-resistant scaffold that integrates the mechanical forces generated during heart contraction. Here we have dissected the interaction of DP, a junctional plaque protein of intercalated discs (Franke et al., 2006), with desmin. Our findings show that DP is able to associate directly with desmin as assessed by biochemical and yeast two-hybrid assays. In extension to previous studies (Meng et al., 1997; Fontao et al., 2003), we show that sequences contained in the linker region and the C-terminus of DP are not only critical for binding to vimentin and cytokeratins (Fontao et al., 2003), but also to desmin. However, the presence of both B and C repeats and flanking sequences is essential for efficient DP binding to desmin. Furthermore, the interaction of DP with nonfilamentous desmin and vimentin in yeast assays is dependent on sequences contained within the C-terminal portion of the rod and the tail domain, whereas in in vitro binding studies DP is also able to associate with polymerized tailless desmin. This observation suggests that the assembly state of desmin affects its binding to DP. Finally, we obtained evidence that disease-causing mutations in either the DP or desmin gene may have an as yet unrecognized impact on the ability of DP and desmin to associate with each other.
Identification of sequences within the DP tail required for its association with desmin
Here we demonstrate that regions of DP important for vimentin binding (Stappenbeck et al., 1993; Fontao et al., 2003) (this study) also contain recognition sites contributing to the interaction with desmin. First, in transfection studies, we found that the DP tail is not only able to coalign with the desmin network in non-myogenic cell types, but is also targeted to the subcellular region of cardiac myocytes, where the endogenous desmin network is found. Furthermore, combined yeast and ligand-binding studies indicate that the linker region and the C-terminus of DP are critically implicated in the association with desmin, but that the B and C plakin repeat domains are required to ensure robust binding. This situation is thus consistent with the results of in vitro binding assays showing that a recombinant DP protein encompassing both the B and C subdomains associates better with vimentin than the isolated B or C subdomain (Choi et al., 2002). In this context, it should be noted that although the regions within the DP tail implicated in binding to vimentin and desmin were similar, the association of DP with vimentin appeared to be invariably stronger than that with desmin. Overall the presented results provide further support to the theory that the linker region between the B and C subdomain and C-terminal extremity of DP encompass sequences critical for binding to both desmin and vimentin, while the B and C subdomains participate to ensure a robust and efficient association (Fontao et al., 2003) (Fig. 10).
The presence of Ser at position 2849 within the C-terminal extremity of DP affects the association with desmin
Phosphorylation of the tail of DP and plectin seems to critically modulate their interaction with various IFs (Fontao et al., 2003; Godsel et al., 2005). The 68-residue long COOH extremity of DP contains smaller repeating units of G-S-R-X, the last of which is modified to G-S-R-R-G-S and may serve as a target sequence for protein kinases. Evidence has been provided indicating that the C-terminal extremity of DP is indeed subject to phosphorylation in vivo in distinct cell lines (Stappenbeck et al., 1994; Godsel et al., 2005) and in the PJ-49A yeast strain utilized here (Fontao et al., 2003). Our results in yeast reveal that the DP-BC mutant carrying the substitution S2849G, but none of the recombinant proteins containing the wild-type COOH extremity of DP, showed binding activity to desmin and vimentin. Recent studies have demonstrated that a DP molecule containing the S2849G substitution shows an enhanced association with IFs in living cells. This phosphorylation-deficient mutant was found at a five-times greater ratio in the Triton X-100 insoluble fraction when compared with the wild-type construct (Godsel et al., 2005). Together, these findings further support the idea that phosphorylation of the DP tail represents a general means by which the association of DP with various IF types is differentially modulated. In this context, evidence has been recently provided indicating that the interaction between cadherin-catenin complexes with the actin cytoskeleton is not static and stable, but rather dynamic, a means which would enable the cells to undergo constant morphogenetic changes (Yamada et al., 2005). Therefore, further studies are needed to test whether the linkage of desmosomal molecule-containing cell junctions with the IF system is also more dynamic than currently appreciated and to precise the role of phosphorylation events in local regulation of this connection.
An increased importance of the tail domain of desmin for the interaction with DP in yeast, but not in ligand-binding assays
Evidence exists indicating that the rod domain of different IF proteins can mediate their association with distinct cytolinkers, such as DP, BP230, plectin and plakophilins (Foisner et al., 1988; Hofmann et al., 2000; Fontao et al., 2003). Specifically, two studies suggested that the rod domain of vimentin mediates the interaction with DP (Meng et al., 1997; Choi et al., 2002). Crystallographic studies of Choi et al. (Choi et al., 2002) showed that the B and C subdomains of DP exhibit a conserved basic groove, a feature that would potentially allow an interaction with the rod of vimentin. Although our overall findings do not exclude a critical role of the rod domain of desmin and vimentin for binding to DP, they provide evidence that there are additional important sequences in their tail domain that contribute to the interaction. In fact, deletion of their tail domain abrogated their binding to DP in yeast. Furthermore, the colocalization potential of DP with tailless desmin was occasionally impaired in transfected cells. The presence of a missense mutation in the desmin tail linked to cardiomyopathy (see below) had similar consequences. Hence, our findings suggest that the tails of desmin and vimentin participate in the establishment of cytoskeletal architecture by favoring their connection with DP and thus membrane sites.
Intriguingly, yeast two-hybrid assays appeared to be more stringent than in vitro binding assays and transfection studies, in which DP was still able to associate and become coaligned with tailless desmin, respectively. It is possible that the binding sites on the C-terminal rod region are sufficient to ensure the association of DP under the experimental conditions of the overlay or transfection studies, in which DP is tested against a polymerized and reconstituted IF network. By contrast, in yeast, desmin and vimentin are only expected to form either homodimers or tetramers (Meng et al., 1997). Therefore, the assembly state of desmin and vimentin may have a critical impact on the conformation and/or the number of available recognition sites important for their interaction with DP. In this regard, the situation is reminiscent of previous studies indicating that the nonfilamentous tetrameric form of desmin has substantially different binding abilities for calponin from that of polymeric filamentous desmin (Mabuchi et al., 1997).
Recent yeast two-hybrid assays from our laboratory (our unpublished data) indicate that the N-terminal half of the desmin rod mediates its association with plectin, another desmin-binding protein of the plakin family that is also expressed in cardiomyocytes (Reipert et al., 1999). These preliminary observations and current findings support the idea that desmin binds to DP and plectin by using a different set of sequences located in the C-terminal and N-terminal portion of the rod domain, respectively. This may provide a means by which desmin participates in multiple and non-competitive molecular interactions with various cytolinkers to strengthen IF connections. This seems essential in vivo, since the presence of plectin at intercalated disc regions is not sufficient to fully compensate defects of DP function in heart of either humans or mice (Gallicano et al., 2001).
DP and desmin mutations can impair the DP-desmin interaction: evidence for the molecular basis of DP- and desmin-related cardiomyopathies
DP gene mutations have been linked to cardiomyopathies (Armstrong et al., 1999; Norgett at al., 2000; Alcalai et al., 2003). Notably, in myocardial sections from one patient with a mutation leading to a truncated DP protein lacking the C subdomain and C-terminal extremity, desmin did not localize to intercalated discs, which suggests an impairment in the DP-desmin interaction is the underlying causative mechanism (Kaplan et al., 2004). Our findings provide strong support for this idea, since a recombinant DP protein (DP-BL) similar to the above mentioned inherited DP tail truncation (Norgett et al., 2000) showed reduced binding activities to desmin. Strikingly, we found that a recessively inherited missense mutation in the B subdomain (Alcalai et al., 2003) also impairs DP-desmin association, since a DP-BC construct carrying the G2375R substitution was unable to interact with desmin in yeast (not shown). The substitution of this Gly, which adopts a backbone conformation with formation of a sharp turn at the end of the third plakin repeat (Choi et al., 2002), by an Arg is expected to profoundly affect the structure of the B subdomain and thus of the DP tail. This observation further underlines the participation of the plakin repeats to IF-binding.
Desmin mutations may also lead to cardiomyopathy (Goldfarb et al., 1998; Li et al., 1999; Dalakas et al., 2003). These mutations, most of which are located within the 2B segment of the rod, interfere with the IF assembly process at distinct stages, although some mutants form normal-looking IFs (Bär et al., 2005; Bär et al., 2006). In this context, a desmin mutant with a missense mutation in the tail, desminI451M, exhibited only a partial impairment of IF network formation, which raises the possibility that certain mutations, besides impairing IF assembly, contribute to disease by other mechanisms (Li et al., 1999; Dalakas et al., 2000). Our results supports this idea. First, when tested in yeast, desminI451M was no longer able to interact with DP, whereas another desmin mutant with a L385P substitution within the 2B segment did (not shown). Furthermore, in transfection studies, the potential of DP to coalign with desminI451M appeared reduced. Hence, mutations in the desmin tail may have an as yet unrecognized impact on DP-desmin interaction, impairing IF-membrane attachments. The latter have been found to regulate intercellular adhesive strength (Huen et al., 2002). In line with this idea, it should be mentioned that, in desmin-null mutant mice, changes at intercalated discs have been observed, where intercellular gaps form between opposing cardiomyocytes (Thornell et al., 1997).
In conclusion, our study demonstrates for the first time that, in analogy to what was found with vimentin, binding of DP to desmin depends on sequences within the linker region and C-terminal extremity of DP with its B and C subdomains contributing to efficient binding. Furthermore, the C-terminal rod region and tail of desmin contain recognition sites implicated in the interaction with DP. Finally, inherited mutations in these proteins may critically impair their association. These studies further increase our understanding of DP-IF interactions important for maintenance of cytoarchitecture in cardiac cells and give new insights into the molecular basis of desmosomal proteins and desmin-related human cardiomyopathies.
Materials and Methods
Plasmid inserts were generated by restriction digestion or PCR using the proofreading Pfu DNA polymerase (Promega, Madison, WI) and gene-specific sense and antisense primers containing restriction site tags. Primer design was based on human DP, vimentin and desmin sequences (GenBank acc. no. m77830, bc030573 and nm-001927, respectively). Mutagenesis was carried out using the Quick Change site-directed mutagenesis kit (Stratagene, La Jolla, CA). The various DP deletion mutants and IF proteins were cloned into either the yeast GAL4 DNABD vector pAS2-1 or GAL4-AD vector pACT2 (Clontech, Palo Alto, CA), into the eukaryotic expression vector pEGFP-C3 (Clontech), pcDNA3-myc vector (Fontao et al., 2003), the prokaryotic expression vector pET15b (Novagen, Madison, WI) and pGEX-2T (Amersham Pharmacia Biotech, Piscataway, NJ). The chimeric construct BP(BC)-DP(Ct) consisting of residues 2077 to 2649 of BP230 fused to residues 2821 to 2871 of DP with or without the S2849G mutation has been previously described (Fontao et al., 2003). Sequences were confirmed by nucleotide sequencing.
Cell culture, transfection and immunofluorescence microscopy studies
The SW13 clone 2l human adrenal carcinoma cell line has been previously described (Sarria et al., 1990). Cells were cultured in DMEM and Ham's F12 media, supplemented with 5% FBS, 100 U/ml glutamine, 100 U/ml penicillin and 100 U/ml streptomycin. Cells were grown at 37°C in a humidified 5% CO2 atmosphere. The human KEB-3D keratinocyte cell line stably expressing desmin was cultured as described (Magin et al., 2000). Cells were grown to 40-60% confluence on glass coverslips in 6-well tissue-culture plates. Transient transfections were performed with 0.8 μg cDNA using 2 μl of Lipofectamine 2000 (Invitrogen) according to the manufacturer's procedure. Neonatal cardiac cells were isolated from one to twoday-old Wistar rats ventricles by digestion with trypsin-EDTA and type 2 collagenase as described (Springhorn and Clayhorn, 1989). Once the sequential digestions were terminated, the cells were pooled in Dulbecco's modified Eagle's medium (DMEM, Invitrogen, Groningen, Netherlands) supplemented with 10% FBS (GibcoBRL, Switzerland), penicillin (100 units/ml) and streptomycin (10 μg/ml) and seeded in 150 cm2 flasks to allow selective adhesion of cardiac fibroblasts (Sadoshima and Izumo, 1993). Thereafter, cardiomyocytes were decanted from the flasks. Electroporation was performed with cells using Nucleofector™ (Amaxa, Cologne, Germany) according to the manufacturer's protocol. Briefly, 1 μg of pEGFP-C3 (Clontech) or GFP-DP-BC per 106 cardiomyocytes were resuspended in a mixture of 100 μl Nucleofector solution and electroporated in the Nucleofector electroporator with the cardiomyocyte specific program (G09). Thereafter, cardiomyocytes were plated on fibronectin-gelatin-coated 12-mm glass slides. After 4 days of culture, transfected rat cardiomyocytes were rinsed twice in PBS, fixed with 2% paraformaldehyde for 15 minutes, rinsed twice in PBS and permeabilized with 0.1% Triton X-100 in PBS for 5 minutes at room temperature. For the other immunofluorescence microscopy studies, cell grown on glass coverslips were fixed with 1% paraformaldehyde in MTSB (0.1 M Pipes, 1 mM EGTA, 4% PEG 4000, pH 6.9) for 10 minutes at 37°C and permeabilized with 0.1% Triton X-100 in PBS for 5 minutes at room temperature. After rinsing in PBS and blocking with 1% BSA in PBS for 30 minutes at 37°C, the cells were incubated with primary antibodies for 30 minutes at 37°C and then washed twice with PBS. Cells were subsequently incubated with secondary antibodies for 30 minutes at 37°C, washed twice, mounted in DAKO medium and viewed under a Zeiss inverted microscope Axiovert 200 (Zeiss, Oberkochen, Germany), under a Zeiss LSM410 confocal inverted laser scanning microscope (Zeiss) for the desmin-transfected keratinocyte cell line and under a LSM 510 Meta confocal scanner mounted on an upright Axioskop 2FS microscope for the transfected cardiomyocytes. Tissue sections of monkey heart were purchased (Inova Diagnostics Inc., San Diego, CA).
The following immunoreagents were used: mouse monoclonal antibody (mAb) directed against the HA epitope tag (12CA5), mAb 9E10 against the Myc epitope tag (Boehringer Mannheim Corporation, CA), mAb GFP-2 against GFP (Santa Cruz Biotechnology, Santa Cruz, CA), mAb D33 against desmin (Dako, Hamburg, Germany), mAb anti desmin (LabVision, CA), the rabbit NW161 anti-desmoplakin antiserum (Bornslaeger et al., 1996), mAb RCK107 directed against K14 (Monosan, Uden, The Netherlands), and anti-vimentin mAb clone V9 (Immunotech, Marseille, France); rabbit SC-805 anti-serum against hemagglutinin (HA) epitope tag, rabbit anti-GFP antiserum and rabbit H76 anti-desmin antiserum (Santa Cruz Biotechnology, Santa Cruz, CA), GP53 guinea pig anti-vimentin antiserum (Progen Biotechnik GMBH, Heidelberg, Germany). Secondary antibodies were purchased from Molecular Probes (Eugene, OR), Alexa Fluor 488-conjugated goat anti-rabbit IgG, Alexa Fluor 488 goat anti-guinea pig, Alexa Fluor 568 goat anti-rabbit, Alexa Fluor 488 FITC goat anti-mouse, TRITC-conjugated donkey anti-mouse IgG and Cy5-conjugated AffiniPure Donkey anti-Mouse IgG (Jackson ImmunoResearch Laboratories, PA).
Yeast two-hybrid assays
Yeast two-hybrid assays were performed as previously described (Fontao et al., 2003; Gontier et al., 2005). The vectors used were the yeast GAL4-AD and GAL4-BD expression vectors pACT2 and pAS2.1, respectively (Clontech).
Metabolic labeling of c-myc and HA-tagged proteins
35S-methionine-labeled recombinant forms of DP and desmin were generated by coupled in vitro transcription/translation of pcDNA3-myc constructs using the Quick TNT coupled reticulocyte lysate system (Promega). Non-incorporated amino acids were removed from the in vitro translation mixture (50 μl) using Ultrafree 0.5-Biomax 5K (Millipore) filters. The translation mixture was diluted into 2 ml binding buffer (20 mM HEPES, 10 mM PIPES, 0.2 mM CaCl2, 2 mM MgCl2, 50 mM KCl, pH 7.2) supplemented with 0.1% (wt/vol) of BSA.
Expression and purification of recombinant proteins
cDNA fragments encoding the B and C subdomains of DP and BP230 were subcloned in frame with the C-terminus sequence of GST into pGEX-2T (Amersham Pharmacia Biotech). Escherichia coli BL21(DE3) (Novagen) was transformed with these constructs or the vector pGEX-2T without insert. Expression of GST or GST-fusion proteins were induced by adding 0.5 mM IPTG in the broth medium for 3 hours. Bacteria were collected by centrifugation, resuspended in phosphate-buffered saline (PBS) supplemented with 1% Triton X-100 and 5 mM EDTA and lyzed by sonication. Purification of GST-fusion proteins was performed as previously described (Geerts et al., 1999). Protein concentration was determined with the protein assay reagent from Bio-Rad (Bio-Rad, Hercules) using bovine serum albumin (BSA) as a standard.
IF proteins (human vimentin, a gift of H. Herrmann, Heidelberg, Germany, and human desmin, Progen) were first equilibrated by dialysis against a buffer consisting of 6 M urea, 10 mM Tris-HCl, 10 mM β-mercaptoethanol, pH 8.0 for 1 hour at 4°C. Additional dialyses were performed with 3 M urea, 10 mM Tris-HCl, 10 mM β-mercaptoethanol, pH 8.0 for 4 hours at 4°C, then with 5 mM Tris-HCl, 10 mM β-mercaptoethanol, pH 8.0 for overnight at 4°C and finally with 10 mM Tris-HCl, 2 mM β-mercaptoethanol, 5 mM EDTA, pH 8.0 for 3 hours at 4°C. The proteins were polymerized by addition of 1/10 (vol/vol) of 0.2 M Tris/HCl, 1.6 M NaCl, pH 7.0 for 1 hour at room temperature. The quality of the filaments was verified by electron microscopy after negative staining with uranyl acetate. One to 0.01 μg of the polymerisation mixture was spotted onto a nitrocellulose membrane using a Dot-Blot apparatus (Schleider and Schuell, Keene, NH). Membranes were subsequently washed in binding buffer and incubated overnight at 4°C in blocking buffer [binding buffer supplemented with 2% (wt/vol) of heat-treated BSA]. Nitrocellulose strips were then incubated overnight at 4°C with 35S-methionine-labeled proteins prepared as described above. After subsequent washes with binding buffer supplemented with 0.1% BSA and with binding buffer, the nitrocellulose strips were air-dried and bound proteins were visualised by autoradiography.
GST pull-down assays
To reduce non-specific binding to the GST moiety in subsequent steps, 35S-methionine-labeled recombinant forms of desmin, prepared as described above, were diluted in binding buffer supplemented with 1% heat inactivated BSA and preabsorbed for 1 hour at room temperature to 100 μg of GST immobilized to glutathione beads. The preabsorbed mixture was then incubated for 1 hour at room temperature with 10 μg of GST-DP-BC or GST-BP230-BC immobilized on glutathione. After four washes in binding buffer, beads were resuspended in SDS-sample buffer, heated five minutes at 100°C and loaded on a 10% polyacrylamide gel. Bound proteins were visualized by autoradiography.
The authors are indebted to Prof. R. Evans, University of Colorado (USA) for kindly providing the IF-free SW13 cell line, to Prof. B. Lane, Dundee (UK) and Prof. T. Magin, Bonn (Germany) for the human EBS keratinocyte cell line stably expressing desmin, to Prof. H. Herrmann, Heidelberg for the recombinant vimentin and desmin proteins and the various cDNAs, and Prof. F. C. Ramaekers, Maastrich (The Netherlands) for his generous gifts of antibodies. This work was supported by grants from the Swiss National Foundation for Research (3100-067860 and 3100-109811 to L.B.), Téléthon Action Suisse (to L.B.), Swiss Foundation for Research on Muscle Diseases (to L.B.), Association Française contre les Myopathies (Paris, to K.L.), and NIH AR43380 (to K.J.G.).
- Accepted September 11, 2006.
- © The Company of Biologists Limited 2006