Immature astrocytes and astrocytoma cells contain synemin and three other intermediate filament (IF) proteins: glial fibrillary acidic protein (GFAP), vimentin and nestin. Here, we show that, after neurotrauma, reactive astrocytes produce synemin and thus propose synemin as a new marker of reactive astrocytes. Comparison of synemin mRNA and protein levels in brain tissues and astrocyte cultures from wild-type, Vim-/- and Gfap-/-Vim-/- mice showed that in the absence of vimentin, synemin protein was undetectable although synemin mRNA was present at wild-type levels. By contrast, in Gfap-/- astrocytes, synemin protein and mRNA levels, as well as synemin incorporation into vimentin IFs, were unaltered. Biochemical assays with purified proteins suggested that synemin interacts with GFAP IFs like an IF-associated protein rather than like a polymerization partner, whereas the opposite was true for synemin interaction with vimentin. In transfection experiments, synemin did not incorporate into normal, filamentous GFAP networks, but integrated into vimentin and GFAP heteropolymeric networks. Thus, alongside GFAP, vimentin and nestin, reactive astrocytes contain synemin, whose accumulation is suppressed post-transcriptionally in the absence of a polymerization partner. In astrocytes, this partner is vimentin and not GFAP, which implies a functional difference between these two type III IF proteins.
Introduction
Intermediate filament (IF) networks constitute dynamic scaffolds integrating the cytoplasm on a structural and functional level (for reviews, see Chang and Goldman, 2004; Coulombe and Wong, 2004; Lane and Pekny, 2004; Omary et al., 2004). IFs can assemble from approximately 60 proteins that share a short signature sequence and a tripartite domain, whose major structural feature is a central α-helical domain of 310 amino acids (Herrmann and Aebi, 2004). Five types of IF proteins are recognized based on sequence homologies and on the intron-exon organization of the genes.
IF proteins are expressed in a cell-type specific pattern (for reviews, see Omary et al., 2004; Coulombe and Wong, 2004). This pattern is often established after a series of developmentally regulated transitions in IF proteins, and it is often altered in diseases. For example, during development and in pathological situations, astrocytes can simultaneously express two type III [vimentin and glial fibrillary acidic protein (GFAP)] and two type IV (synemin and nestin) IF proteins. Astrocyte precursors contain vimentin, nestin and synemin (Lendahl et al., 1990; Rutka et al., 1997; Sultana et al., 2000). As astrocytes differentiate, these IF proteins are replaced progressively by GFAP. In astrocytoma cells, this process is reversed and the IF protein composition often resembles that of immature astrocytes (Dahlstrand et al., 1992; Jing et al., 2005; Rutka et al., 1997). In neurotrauma and in ischemic and neurodegenerative lesions, reactive astrocytes overexpress GFAP and re-express vimentin and nestin (for reviews, see Eng et al., 2000; Pekny and Pekna, 2004; Pekny and Wilhelmsson, 2006).
Vimentin and GFAP are ∼80% homologous at the primary protein sequence level, and each can polymerize on its own or form heteropolymers with the other (Quinlan and Franke, 1983; Wang et al., 1984). Synemin and nestin, however, exhibit only minor sequence homologies with each other and with other IF proteins, and possess a large C-terminal domain that is three to four times larger than that of vimentin or GFAP (Lendahl et al., 1990; Bellin et al., 1999; Titeux et al., 2001). In humans, the synemin gene is alternatively spliced into two isoforms called α and β, with the C-terminal domain of α-synemin containing a 312 amino-acid insertion when compared with β-synemin (Titeux et al., 2001). The C-terminal domain of α-synemin contains binding sites for α-actinin and vinculin (Bellin et al., 1999; Bellin et al., 2001), whereas β-synemin can interact with dystrobrevin (Mizuno et al., 2001). Synemin cannot assemble into IFs on its own but can copolymerize with vimentin (Bellin et al., 1999; Bilak et al., 1998; Marvin et al., 1998; Sandoval, 1983; Steinert et al., 1999; Titeux et al., 2001), resembling in this respect the type IV IF protein nestin (Eliasson et al., 1999).
Synemin is the least understood of the IF proteins expressed by astrocytes. In this study, we investigated whether reactive astrocytes in neurotrauma express synemin and how the genetic ablation of GFAP and/or vimentin influences this expression. We also analyzed in detail the interactions of synemin with GFAP and vimentin. These studies provide novel information on the IF protein composition of reactive astrocytes and identify synemin as a novel marker of reactive astrocytes. Our results also show that, in astrocytes, synemin accumulation into GFAP-containing IFs is dependent on vimentin expression, suggesting that vimentin may function as an adaptor for synemin incorporation into GFAP-containing IFs.
Results
Synemin immunoreactivity in reactive astrocytes in adult wild-type and Gfap-/-Vim-/- mice
To assess the expression of synemin in reactive astrocytes in neurotrauma, we subjected adult wild-type and Gfap-/-Vim-/- mice to unilateral entorhinal cortex lesioning. In this injury model, the perforant pathway is interrupted, triggering astrocyte activation in the denervated outer molecular layer of the dentate gyrus of the hippocampus. In wild-type mice, GFAP immunoreactivity increased and synemin immunoreactivity was detected in the outer molecular layer on the injured side 4 days (Fig. 1A-B) and 14 days (data not shown) after lesioning. No synemin immunoreactivity was seen on the uninjured site (data not shown), which is in agreement with recent findings showing that α- and β-synemin are not present in the dentate gyrus of normal, adult mice (Izmiryan et al., 2006). Synemin and GFAP immunoreactivity colocalized in reactive astrocytes both in the denervated dentate gyrus (Fig. 1C-E) and in the lesioned entorhinal cortex (data not shown). In Gfap-/-Vim-/- mice, however, the only detectable synemin immunoreactivity was that associated with blood vessels as determined by double immunofluorescence with synemin and α-smooth muscle actin antibodies (Fig. 1H-J).
Synemin protein but not mRNA is reduced in the brains of Gfap-/-Vim-/- neonatal mice
Next, we analyzed synemin expression in the cerebral hemispheres of 2-day-old wild-type and Gfap-/-Vim-/- mice. Western blot analysis of total protein extracts from wild-type mice revealed bands of 210 and 170 kDa, corresponding to α- and β-synemin, respectively (Fig. 2A); α-synemin was more abundant than β-synemin. By contrast, α-synemin was barely detectable in cerebral extracts from Gfap-/-Vim-/- mice, and β-synemin was undetectable. α-synemin protein levels were approximately 95% lower in Gfap-/-Vim-/- mice (P<0.001, n=4 mice/group), as shown by densitometry. Actin levels were comparable in the two groups, indicating equal protein loading. As expected, GFAP and vimentin were abundant in the brain of wild-type mice, but absent in Gfap-/-Vim-/- mice (Fig. 2A).
In reverse transcription PCR (RT-PCR) reactions, fragments of 3.7 and 2.8 kb, corresponding to α- and β-synemin mRNA, respectively, were amplified at similar levels from cerebral tissues of 2-day-old wild-type and Gfap-/-Vim-/- mice (Fig. 2B). The α/β synemin mRNA ratio appears to be higher than the α/β synemin protein ratio seen on western blots. RT-PCR, however, may not provide accurate representation of the α/β synemin mRNA ratio within one sample. This is because the set of primers used is very likely to amplify more efficiently the β-synemin cDNA fragment than the 1 kb-longer α-synemin fragment.
Quantitative RT-PCR assays, performed using conditions under which GAPDH and synemin were amplified with similar efficiencies (data not shown), revealed no difference in cerebral synemin mRNA levels between wild-type and Gfap-/-Vim-/- mice (Fig. 2C).
Synemin protein does not accumulate in Vim-/- astrocytes
Next, we performed a series of experiments with primary (data not shown) and passage 1 astrocytes from 2-day-old wild-type, Gfap-/-Vim-/-, Gfap-/- and Vim-/- mice. RT-PCR showed no qualitative differences in synemin isoforms mRNA levels in cultured astrocytes of all four genotypes (Fig. 3A). Quantitative RT-PCR demonstrated no differences in synemin mRNA levels (Fig. 3B).
By contrast, western blotting analysis revealed that synemin is undetectable in the absence of vimentin, even if GFAP is present, and that synemin is present at levels comparable to wild type in astrocytes expressing vimentin, but not GFAP (Fig. 3C). Western blots also demonstrated that in both wild-type and Gfap-/- astrocyte cultures, vimentin and synemin were recovered almost entirely in the cytoskeletal fraction (Fig. 3D), implying that the lack of GFAP does not affect the polymerization state of synemin and vimentin. In wild-type cultures, a pool of approximately 10% of total GFAP was present in the cytosolic fraction (Fig. 3D).
Synemin subcellular distribution in astrocyte cultures
Wild-type astrocyte cultures displayed heterogeneity with respect to synemin expression, and synemin was present in some GFAP-positive and in some GFAP-negative cells. In both GFAP-positive and negative cells and in Gfap-/- cultures, synemin was distributed throughout the IF network, from the dense meshwork encircling the nucleus (Fig. 4A) to the dispersed fibers present at the cell periphery (Fig. 4B). The intensity of the synemin immunoreactivity varied along fibers (Fig. 4B).
Synemin has the capacity to interact with vinculin and α-actinin, and in some cell types it colocalizes with these proteins (Bellin et al., 1999; Bellin et al., 2001; Jing et al., 2005; Uyama et al., 2006). In cultured wild-type astrocytes, there was no apparent overlap between the distribution of synemin and vinculin (Fig. 5) or of synemin and α-actinin (data not shown). In Gfap-/-Vim-/- astrocytes, the subcellular distribution of vinculin (Fig. 5C,D) and α-actinin (data not shown) was similar to that observed in wild-type astrocytes (Fig. 5A,B).
Synemin binds equally well to unpolymerized GFAP and vimentin
Our finding that synemin protein is stable in Gfap-/-, but not in Gfap-/-Vim-/- or Vim-/- astrocytes raised the question of whether synemin can interact with GFAP like it does with vimentin. To address this, we performed a series of biochemical assays and transfection experiments.
Bacterially expressed human GFAP, vimentin and α- and β-synemin for biochemical assays were purified to homogeneity by repeated polymerization/depolymerization cycles and gel filtration (Fig. 6A). For overlay assays, GFAP and vimentin were loaded onto SDS-polyacrylamide gels and transferred to nitrocellulose filters, which were overlaid with α-synemin. Binding to the immobilized proteins was assessed with anti-synemin. Synemin bound equally well to GFAP and vimentin, but did not bind to actin (Fig. 6B). Similar results were obtained with β-synemin (data not shown). In the absence of synemin, no signal was detected. These findings are similar to those of Hirako et al. with GFAP and vimentin from bovine optic nerve and with synemin from rabbit stomach smooth muscle (Hirako et al., 2003).
To examine binding under nondenaturing conditions, we performed dot-blot assays. Synemin, GFAP, vimentin and actin were dialyzed against 2 mM Tris-Cl (pH 7.2) and 0.5 mM dithiothreitol (DTT), and various amounts of GFAP, vimentin and actin were adsorbed onto nitrocellulose filters. Under these low-salt, neutral-pH buffer conditions, GFAP and vimentin remain unpolymerized (Kooijman et al., 1995). α-synemin bound to GFAP and vimentin equally well and in proportion to the amount of protein immobilized on the filter, but did not bind to actin (Fig. 6C). Similar results were obtained with β-synemin (data not shown). No signal was detected in the absence of synemin (data not shown).
In vitro, synemin interacts differently with GFAP and vimentin IFs
To examine whether synemin copolymerizes with GFAP, we mixed α-synemin with GFAP (1:100 molar ratio) in 8 M urea buffer, dialyzed the mixture against a buffer of physiological ionic strength and pH to assemble IFs, subjected it to ultracentrifugation, and analyzed the pellet and supernatant by SDS-PAGE (Fig. 7A). Similar experiments were performed with vimentin, which was used as a positive control because synemin is known to copolymerize with vimentin (Bellin et al., 1999; Titeux et al., 2001). In the presence of GFAP or vimentin, α-synemin was largely recovered in the pellet. However, when α-synemin alone was subjected to this analysis, it was mostly recovered in the supernatant. Similar results were obtained with β-synemin (data not shown). Electron microscopy showed that vimentin and GFAP were capable of polymerizing into IFs (data not shown).
These results do not rule out the possibility that synemin cosediments with GFAP because it binds to GFAP IFs rather than copolymerizes with GFAP. This possibility was examined by mixing α-synemin with GFAP or vimentin IFs assembled in the absence of synemin (preformed IFs). Before addition of preformed IFs, α-synemin was dialyzed against IF assembly buffer and ultracentrifuged to remove insoluble aggregates. When this supernatant was ultracentrifuged a second time, most of the synemin was recovered in the new supernatant (Fig. 7B). Similarly, after incubation with preformed vimentin IFs, most of the α-synemin was recovered in the supernatant. However, after incubation with preformed GFAP IFs, most of the α-synemin was recovered in the pellet (Fig. 7B). Similar results were obtained with β-synemin (data not shown). This suggests that, in vitro, synemin interacts with GFAP IF like an associated protein rather than like a polymerization partner.
GFAP networks have a limited capacity to interact with synemin
To examine the capacity of synemin to associate with GFAP networks in vivo, we performed transfection experiments in SW13-cl2 adrenocortical carcinoma cells. These cells are devoid of cytoplasmic IFs and have been used extensively to study IF protein polymerization (Sarria et al., 1990). Transfection with GFAP cDNA yielded a filamentous GFAP network throughout the cytoplasm, as reported (Chen and Liem, 1994; Schweitzer et al., 2001). However, in approximately 30% of the transfected cells, part of the network consisted of large, brightly stained cables or coils (Fig. 8A). Approximately 5% of the transfected cells did not assemble a GFAP network; instead, the protein accumulated in short comma- or rod-shaped structures (Fig. 8B). These structures stained very brightly with anti-GFAP, suggesting that their formation results from GFAP overexpression.
When SW13-cl2 cells were transiently transfected with GFAP and either α- or β-synemin cDNA, the subcellular distribution of synemin depended on the nature of the GFAP network (Fig. 8C-H). When the network was filamentous, synemin formed small round structures that did not stain with GFAP antibodies (Fig. 8C-D). When high GFAP expression led to the formation of cables or to the total disruption of the GFAP network into large rhomboidal aggregates, synemin was colocalized with those GFAP-rich structures (Fig. 8G,H). When the peripheral part of the cytoplasm contained a filamentous GFAP network and the perinuclear region displayed GFAP-containing cables or coils, synemin antibodies stained only latter structures, not the filamentous portion of the network (Fig. 8E-F). The inability of synemin to co-distribute with the filamentous portion of a GFAP network contrasts with its known capacity to form an extended filamentous network with vimentin (Bellin et al., 1999; Titeux et al., 2001), a capacity that we have also observed in a series of control experiments (data not shown).
Vimentin enables synemin to incorporate into GFAP-containing IF networks
Next, we investigated whether vimentin affects the incorporation of synemin into GFAP networks. SW13-cl2 cell lines were stably transfected with GFAP cDNA (SW13-cl2/G) or vimentin cDNA (SW13-cl2/V). Immunofluorescence staining with anti-GFAP demonstrated a filamentous GFAP network in >95% of SW13-cl2/G cells and little evidence of overexpression. Similarly, anti-vimentin staining showed a filamentous vimentin network in nearly all SW13-cl2/V cells. In SW13-cl2/G cells transiently transfected with vimentin cDNA or in SW13-cl2/V cells transiently transfected with GFAP cDNA, immunofluorescence staining with anti-GFAP and anti-vimentin yielded identical patterns (not shown), consistent with the ability of GFAP and vimentin to coassemble into heteropolymeric networks (Quinlan and Franke, 1983; Wang et al., 1984).
After transient transfection of SW13-cl2/G cells with α- or β-synemin cDNA, synemin failed to incorporate into the GFAP network but instead formed numerous punctate aggregates throughout the cytoplasm (Fig. 9). However, after cotransfection of these cells with vimentin and α- or β-synemin cDNA, synemin was incorporated into the vimentin/GFAP IF network in >90% of the cotransfected cells (Fig. 10A,B). In SW13-cl2/V cells cotransfected with GFAP and α- or β-synemin, synemin and GFAP co-distributed within the same IF network (Fig. 10C-F). Each of the above transfection experiments were performed with three separate SW13-Cl2/G or SW13-cl2/V clones, and each clone yielded similar results.
Discussion
In this study, we demonstrate that reactive astrocytes express synemin in wild-type but not Gfap-/-Vim-/- mice subjected to neurotrauma. Despite normal synemin mRNA levels, synemin failed to accumulate in the brain tissues and astrocyte cultures from newborn Gfap-/-Vim-/- and Vim-/- mice, whereas in Gfap-/- astrocytes, synemin accumulated normally. Biochemical assays with purified proteins and stable and transient transfection experiments demonstrated that GFAP IF networks have limited capacity to interact with synemin, which contrasts with known synemin ability to interact with vimentin networks (Bellin et al., 1999; Titeux et al., 2001). These results suggest that, in astrocytes, vimentin functions as an adaptor that enables synemin to incorporate into GFAP-containing networks.
The induction of synemin synthesis in reactive astrocytes in wild-type mice after neurotrauma is consistent with our recent finding that human reactive astrocytes found around ischemic lesions or epileptic foci stain with synemin antibodies (Jing et al., 2005). We propose that synemin expression is part of the response of astrocytes to neurotrauma and other brain and spinal cord pathologies. Thus, along with nestin (Dahlstrand et al., 1992) and endothelin B receptors (Rogers et al., 1997), synemin might be a useful marker of reactive astrocytes.
Our immunofluorescence studies show that, in astrocyte cultures, synemin localizes over the entire IF network, irrespective of whether this network is assembled of vimentin or of vimentin and GFAP. In cultured astrocytes, synemin did not associate with α-actinin-rich cellular regions, as it does in muscle and astrocytoma cells (Sandoval et al., 1983; Bellin et al., 1999; Jing et al., 2005), or with focal contacts, as it does in hepatic stellate cells (Uyama et al., 2006). The factors determining synemin subcellular distribution in different cell types remain to be identified.
Synemin was not found in reactive astrocytes in Gfap-/-Vim-/- mice after entorhinal cortex lesion or in brain tissue and cultured astrocytes from newborn Gfap-/-Vim-/- and Vim-/- mice. This synemin expression profile is comparable to that of nestin, another type IV IF protein expressed by immature, reactive and malignant astrocytes (Dahlstrand et al., 1992; Lendahl et al., 1990; Frisen et al., 1995; Eliasson et al., 1999). In reactive astrocytes, nestin protein levels are high in wild-type mice, but are markedly reduced in Gfap-/-Vim-/- and Vim-/- mice (Eliasson et al., 1999). Similarly, nestin protein levels are 80% lower in Gfap-/-Vim-/- and Vim-/- cultured astrocytes than in wild-type and Gfap-/- cultured astrocytes (Eliasson et al., 1999).
Gene targeting studies in mice suggest that IF protein levels are decreased in the absence of a polymerization partner in cell types other than astrocytes. In the central nervous system, targeted gene ablation of neurofilament (NF)-M reduces NF-L protein levels by 30 to 90%, depending on the brain region (Jacomy et al., 1999; Elder et al., 1998). Genetic ablation of keratin 18 (K18) or K8 in mice also reduced protein levels for some of their keratin polymerization partners (Magin et al., 1998; Tao et al., 2003), and K2e protein was absent in transgenic mice expressing truncated, polymerization-incompetent K10 (Reichelt et al., 1997). In transfection experiments with cultured fibroblasts, the stability of a type II keratin required a polymerization partner (Kulesh et al., 1989).
Two lines of evidence strongly suggest that the stability of synemin in astrocytes also depends on the presence of a polymerization partner. First, because synemin incorporates into vimentin IF networks (Bellin et al., 1999; Titeux et al., 2001), it is significant that synemin was not detected in Gfap-/-Vim-/- and Vim-/- astrocytes, but was present in Gfap-/- astrocytes at levels identical to those in wild-type astrocytes. Moreover, our immunofluorescence and biochemical studies revealed that synemin incorporation into the vimentin IF network in Gfap-/- astrocytes was indistinguishable from that in wild-type astrocytes. Second, synemin did not accumulate in Vim-/- cells, even though they possess a filamentous GFAP network (Eliasson et al., 1999). That synemin does not incorporate into a GFAP network was demonstrated directly by the finding that in SW13-cl2 cells cotransfected with synemin and GFAP cDNAs, synemin accumulated as punctate inclusions that were independent of the GFAP network.
Synemin mRNA levels were not reduced in astrocytes deficient in GFAP and/or vimentin. This finding establishes that synemin protein levels in vimentin-deficient astrocytes are regulated post-transcriptionally. This may be a general mechanism by which IF protein levels are regulated in the absence of polymerization partners, as post-transcriptional regulation is also responsible for decreased nestin, NF-L and keratin protein levels after the genetic ablation of vimentin, NF-M and of specific keratin pair members, respectively (Reichelt et al., 1997; Elder et al., 1998; Magin et al., 1998; Eliasson et al., 1999; Jacomy et al., 1999; Tao et al., 2003).
Although synemin did not accumulate in astrocytes lacking vimentin, it did form cytoplasmic inclusions in SW13-cl2 adrenocarcinoma cells lacking vimentin. In this cell line, the lack of IF protein is not caused by targeted IF gene disruption, but by the inactivation of two members of the trithorax pathway, resulting in the inactivation of many genes involved in maintaining a differentiated state (Yamamichi-Nishina et al., 2003). Conceivably, the products of some of these genes may also be involved in post-transcriptional control of IF protein accumulation. In their absence, synemin may accumulate in SW13-cl2 cells without a polymerization partner. In any case, our experiments with GFAP-transfected SW13-cl2 cells showed that synemin mRNA can be translated in the absence of vimentin. Thus, in Gfap-/-Vim-/- and Vim-/- astrocytes, synemin may fail to accumulate because of post-translational regulatory mechanisms. Such mechanisms are responsible for the failure of K8 protein to accumulate in fibroblasts transfected with a K8 cDNA (Kulesh et al., 1989).
Our biochemical experiments and those of others (Bellin et al., 1999) suggest that synemin associates with vimentin IF networks by copolymerizing with vimentin rather than by binding to vimentin IFs after vimentin polymerization. Although synemin can integrate into vimentin IF networks (Bellin et al., 1999; Titeux et al., 2001), our transfection experiments demonstrate that synemin does not associate with a normal fibrillar GFAP network in vivo. This is surprising because synemin had some capacity to interact with GFAP in our in vitro biochemical experiments. However, those experiments also suggested that synemin may not copolymerize with GFAP. Instead, it may only associate with GFAP IFs after GFAP polymerization, as similar amounts of synemin were recovered in the GFAP IF pellet in ultracentrifugation assays, regardless of whether synemin was mixed with GFAP before GFAP polymerization or added to GFAP IFs after GFAP polymerization. Thus, in the copolymerization experiments, the synemin in the pellets may reflect its association with GFAP IFs after GFAP polymerization rather than its incorporation into GFAP IFs during GFAP polymerization. Our transfection experiments showed that this binding of synemin to GFAP IFs did not take place along a GFAP filamentous network, but occurred in structures arising from GFAP overexpression, such as GFAP aggregates and cables. Synemin does not play a role in the formation of these structures because they are present in cells transfected with GFAP only, as we showed here, and in the case of GFAP cables, in Vim-/- astrocytes (Eliasson et al., 1999), which we showed are devoid of synemin.
Other differences between the in vitro and in vivo behavior of IF proteins have been reported. NF-L, for instance, forms homopolymers in vitro, but not in vivo (Ching and Liem, 1993; Lee et al., 1993). Such differences may occur because IF assembly in cells commences during the translation of IF proteins (Moon and Lazarides, 1983; Isaacs et al., 1989; Chang et al., 2006). Pulse-chase studies in avian erythroid cells showed that the elongation of vimentin IF generates sites for the incorporation of synemin (Moon and Lazarides, 1983). Such translational interactions may be especially important for understanding how synemin incorporates into IF networks in cells.
Transient cotransfection of synemin and vimentin into SW13-cl2 cells stably transfected with GFAP showed that synemin can incorporate into a pre-existing GFAP network in the presence of vimentin. Similarly, transient cotransfection of synemin and GFAP into SW13-cl2 cells stably transfected with vimentin showed that a pre-existing vimentin IF network can serve as a scaffold for the incorporation of GFAP and synemin. These results suggest that one function of vimentin in astrocytes is to permit the incorporation of synemin into GFAP-containing IFs. The pattern of synemin expression in astrocytes also supports this possibility. Synemin is expressed in astrocytes when vimentin is also present, such as during development, in astrocytomas and, as we have shown here, in reactive astrocytes in neurotrauma. In adults, synemin was not detected in mature cerebral astrocytes (Sultana et al., 2000; Jing et al., 2005) (this study), but is present in the optic nerve astrocytes (Hirako et al., 2003), which coexpress GFAP and vimentin (Dahl et al., 1981; Calvo et al., 1990).
Our findings establish that GFAP and vimentin interact differently with synemin in vivo. First, in astrocytes, expression of vimentin, but not of GFAP, was crucial for synemin accumulation, as shown by differences in synemin protein levels in the brains of wild-type, Vim-/-, Gfap-/-Vim-/- and Gfap-/- mice. Second, synemin was not incorporated into a GFAP filamentous network as it was into a vimentin network. This finding may reflect differences in the dynamic properties of GFAP and vimentin filamentous networks and/or the fact that the head domains of vimentin and GFAP, which are important for polymerization, diverge at the primary sequence level (Traub and Vorgias, 1984; Ralton et al., 1994; Herrmann et al., 1996).
However, the tail and rod domains of GFAP and vimentin share ∼80% homology. They are also highly homologous with the tail and rod domains of desmin, the muscle-specific type III IF protein that coexists with synemin in muscle cells during development and in adulthood (Price and Lazarides, 1993; Bilak et al., 1998). In its in vitro interaction with synemin, desmin resembles vimentin more than GFAP because, in biochemical assays, synemin copolymerizes with desmin but binds only minimally to preformed desmin IFs (Bellin et al., 1999). In addition, desmin may, like vimentin, stabilize synemin because, in smooth muscle cells co-expressing desmin and vimentin, synemin is found in Vim-/- or Des-/-, but not in Vim-/-Des-/- cells (Xue et al., 2004). Skeletal and cardiac muscle cells also co-express synemin and desmin, but not vimentin, further suggesting that vimentin is not required for synemin to incorporate into a desmin IF network. In these cells, the absence of desmin leads to greatly reduced synemin protein levels (Carlsson et al., 2000; Xue et al., 2004).
In conclusion, this study demonstrates that neurotrauma induces synemin expression in reactive astrocytes. Importantly, our results in transgenic mice lacking vimentin and/or GFAP also shed light on the functional significance of vimentin upregulation in reactive astrocytes. Specifically, our findings suggests that vimentin is required for synemin to accumulate in astrocytes in response to injury. However, the closely related IF protein GFAP does not perform this function, because vimentin and GFAP differ in their capacity to interact with synemin and to incorporate it into a normal filamentous IF network. The functional significance of synemin expression in reactive astrocytes remains to be determined. However, our findings in cultured astrocytes, whose IF protein composition resembles that of reactive astrocytes, suggest that synemin does not link the IF network to structures rich in α-actinin or vinculin in astrocytes, as it does in other cell types (Sandoval et al., 1983; Bellin et al., 1999; Jing et al., 2005; Uyama et al., 2006). Synemin, however, also binds α-dystrobrevin (Mizuno et al., 2001) protein kinase A regulatory subunit II (Russell et al., 2006), as well as dystrophin and urotrophin (Bhosle et al., 2006). Future investigations will examine whether synemin interacts with some of these proteins in reactive astrocytes.
Materials and Methods
Mice
Mice carrying a null mutation in the GFAP and/or vimentin gene and wild-type controls have been described (Colucci-Guyon et al., 1994; Pekny et al., 1995; Pekny et al., 1999). All mice were on a mixed genetic background (C57Bl/6-129Sv-129Ola).
Surgery
Unilateral entorhinal cortex lesioning was performed in 15-month-old female mice. This lesion partly interrupts the perforant pathway, which innervates the molecular layer in the ipsilateral dentate gyrus of the hippocampus. Axonal degeneration triggers astrocyte activation in the dentate gyrus, an area not directly affected by the trauma. The surgical procedure was performed as described (Stone et al., 1998; Wilhelmsson et al., 2004), with the coordinates +3.6 mm lateral and -0.2 mm posterior to lambda for insertion of the wireknife. The brains were fixed and dissected 4 or 14 days after lesioning, as described (Wilhelmsson et al., 2004).
Cell culture
Primary astrocytes were obtained from the brain hemispheres of 2-day-old mice as described (Pekny et al., 1998). SW13-cl2 adrenocortical carcinoma cells (Sarria et al., 1990) were a kind gift from Robert Evans (Department of Pathology, University of Colorado Health Sciences Center, Denver, CO). Cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented as described (Sultana et al., 2000).
Antibodies and immunofluorescence
Rabbit antibodies against the C-terminus of human α-synemin were raised and affinity-purified as reported (Jing et al., 2005). Other primary antibodies included: mouse monoclonal anti-human vimentin clone V9 (Sigma), mouse monoclonal anti-vinculin (ICN), chicken anti-human and rodent vimentin (Chemicon), mouse monoclonal anti-GFAP clone GA5 (Sigma), mouse monoclonal anti-α-smooth muscle actin (Sigma), and rabbit anti-actin (Sigma). The secondary antibodies were affinity-purified goat anti-mouse or anti-rabbit IgG or anti-chicken IgY conjugated to peroxidase (Kirkegaard & Perry), or Alexa Fluor-488 or Alexa Fluor-588 (Invitrogen). Nuclei were stained with TOPRO-3 (Invitrogen).
Brain sections and cells plated on glass coverslips were fixed and stained for immunofluorescence as detailed earlier (Jing et al., 2005; Wilhelmsson et al., 2004). Confocal microscope observations were performed as described (Jing et al., 2005; Wilhelmsson et al., 2004).
Preparation of total, cytoskeletal and cytosolic protein samples
Total protein samples for SDS-PAGE were prepared as described by Sultana et al. (Sultana et al., 1998). For cytoskeletal and cytosolic protein samples, cells were lysed in lysis buffer (150 μl per 10 cm2) [10 mM Tris-Cl (pH 7.4), 140 mM NaCl, 1% (v/v) Triton X-100, 2 mM EGTA] and protease inhibitors as described (Jing et al., 2005). Centrifugation at 14,000 g for 15 minutes yielded a pellet enriched in cytoskeletal proteins and a supernatant containing cytosolic proteins. Gel loading buffer (150 μl) was added to each supernatant, and pelleted proteins were dissolved in 300 μl of a 1:1 mixture of lysis buffer and gel loading buffer.
Protein concentration determination, SDS-PAGE, western blotting and densitometry
Protein concentrations were measured as described (Jing et al., 2005). For comparison of protein expression levels, 5 μg of total cellular proteins were loaded per lane. To determine whether IF proteins partition between cytoskeletal and cytosolic compartments, equal volumes of protein samples were loaded. Proteins were separated on 4-15% gradient SDS-polyacrylamide gels. Gels were stained with GelCode Blue (Pierce) or processed for western blotting (Jing et al., 2005). Densitometric and statistical analysis were performed as described (Sultana et al., 1998).
Qualitative and quantitative PCR
Total RNA was purified from tissues with the MELT Total RNA Isolation System (Ambion) and from cultured cells with the High Pure RNA Isolation Kit (Roche). Genomic DNA contamination was avoided by treating samples with RNase-free DNase I. cDNA synthesis was performed with Thermoscript reverse transcriptase (Invitrogen) using an oligo(dT)20 primer. RT-PCR of murine synemin isoforms was performed using the conditions and the P2 and P5 primers described by Xue et al. (Xue et al., 2004). These primers amplify all synemin isoforms.
Quantitative PCR for synemin was performed with a TaqMan assay utilizing the fluorigenic probe 5′-AGCCTTCTTCCCTGACCCGTGTGT-3′ conjugated to TAMRA and 6FAM, and the primers 5′-GAGGATGCCATCTTCAGGTAGG-3′ and 5′-AGCCGTTCTGTGTTTCTCTTATAC-3′. These primers and the probe allow quantitation of all synemin isoforms in one reaction. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) served as endogenous control, using the primers and probe provided by ABI. Quantitative PCR was performed with an ABI Prism 7700 and data were analyzed with the ABI 7700 Sequence Detection System using the comparative Ct method. Values are expressed as means ± s.e.m. of three to four separate experiments; differences were analyzed by t-test or one-way analysis of variance (ANOVA).
Cloning human GFAP, vimentin and α- and β-synemin cDNAs into expression vectors
mRNA purification from U-373 MG human glioblastoma cells and cDNA synthesis were performed following standard procedures, using the following gene-specific reverse primers: 5′-AGCCCAGGAGTTCAAGGTCA-3′ (GFAP), 5′-AAACTGCAGAAAGGCACTTGA-3′ (vimentin) and 5′-AAATTGACCATGCCGGGAGTA-3′ (synemin). PCR reactions were conducted with Advantage-GC cDNA polymerase mix (BD Biosciences) and, for each cDNA, with the reverse primer used for cDNA synthesis and the corresponding gene-specific forward primers: 5′-ACGATGGAGAGGAGACGCAT-3′ (GFAP), 5′-ATGTCCACCAGGTCCGTGTC-3′ (vimentin) and 5′-GAGGACGAGACCGGGACAAGA-3′ (synemin). For synemin, this yielded products of 4.9 kb (α-synemin) and 3.9 kb (β-synemin). PCR products were inserted by topocloning into the eukaryotic expression vector pcDNA3.1/V5-His TOPO (Invitrogen). The PCR products included a stop codon at the end of the reading frame for protein expression without the V5-His tag.
Vimentin and GFAP PCR products were also inserted by topocloning into the prokaryotic expression vector pTrcHis2-TOPO (Invitrogen). For α- and β-synemin, we inserted into this vector cDNA obtained by PCR using the primers 5′-ATGCTGTCCTTGCGGCTG-3′ and 5′-AAATTGACCATGCCGGGAGTA-3′. The PCR products included the stop codon at the end of the reading frame, for protein expression without the tag. Because of the sequence of the cloning site and, for GFAP, of the forward primer, the translated proteins contained at their N-terminus the additional amino acids MAL, for vimentin and α- and β-synemin, and MALT, for GFAP. For bacterial expression, pTrcHis2-TOPO constructs were transduced into BL21-CodonPlus-RIL Escherichia coli (Stratagene).
DNA sequencing and BLAST search revealed that our human GFAP and vimentin cDNAs encode amino-acid sequences identical to J04569 and X56134, respectively. For α- and β-synemin, our cDNAs encode proteins identical to O15061 and NP 056101, respectively.
Purification of bacterially expressed human GFAP, vimentin and α- and β-synemin
All buffers included the protease inhibitors described in Jing et al. (Jing et al., 2005). GFAP, vimentin and α- and β-synemin synthesis was induced for 3 hours with 1 mM IPTG. The bacteria were pelleted at 5000 g for 10 minutes, resuspended in 10 mM MES, pH 6.0, 5 mM EDTA and 0.5 mM DTT, lysed with a French press, sonicated and centrifuged at 15,000 g for 15 minutes. The pellets were resuspended in 10 mM HEPES, pH 7.0, 1.5 M KCl, 5 mM EDTA, 1% Triton X-100 and 0.5 mM DTT and centrifuged at 15,000 g for 15 minutes. This procedure was repeated twice, and the final pellet was resuspended in the same buffer, but without Triton X-100, and centrifuged at 15,000 g for 15 minutes. The pellet was extracted with buffer G (6M guanidine HCl, 20 mM Tris-HCl, pH 7.8, 500 mM NaCl, 5 mM EDTA and 0.5 mM DTT).
For GFAP and vimentin, buffer G extracts were dialyzed against 2 mM Tris-HCl, pH 8.0, 1 mM EDTA and 0.5 mM DTT, and ultracentrifuged at 100,000 g for 30 minutes. GFAP or vimentin IF assembly was induced by adding 140 mM NaCl to the supernatants. After 1 hour of incubation, GFAP or vimentin IFs were collected by ultracentrifugation at 100,000 g for 30 minutes. The IF pellet was solubilized in buffer U (8 M urea, 50 mM Tris-HCl, pH 8.0, 1 mM DTT and 5 mM EDTA), and two to three additional polymerization cycles were performed.
For α- and β-synemins, buffer G extracts were dialyzed against buffer P (10 mM Tris-HCl, pH 7.8, 5 mM EDTA, 140 mM NaCl), and centrifuged at 15,000 g for 30 minutes. Urea was added to the supernatant to reach 6 M final concentration, and purified vimentin in buffer U was added to this solution using the empirical formula, A [vimentin (mg)]=VD/60, where V (ml) is the volume of the supernatant and D is its optical density at 280 nm. The rationale for this step is to use synemin capacity to copolymerize with vimentin (Bellin et al., 1999; Titeux et al., 2001) to separate it from bacterial proteins. Copolymerization was induced by dialysis against buffer P, and IFs were pelleted by ultracentrifugation at 100,000 g for 30 minutes.
The final IF protein pellets dissolved in buffer U were loaded onto either a 2.5×100 cm Sephacryl S-200 (GFAP and vimentin) or a 1.5×100 cm Sephacryl S-300 (α- and β-synemin) gel filtration column equilibrated with buffer U. Chromatography was run at a flow rate of 7.5 ml/hour (GFAP and vimentin) or 9 ml/hour (α- and β-synemin). Fractions containing a single band of appropriate size and immunoreactivity were identified by SDS-PAGE and western blotting.
Binding and cosedimentation assays
For blot overlays, 0.2 μg each of GFAP, vimentin and platelet actin (Cytoskeleton) were loaded onto a 4-15% SDS-polyacrylamide gel and transferred to nitrocellulose filters. For dot-blot assays, proteins were dialyzed against 2 mM Tris-HCl (pH 7.2) and 0.5 mM DTT, and 5, 1, 0.5 or 0.1 μg of each protein was spotted onto nitrocellulose filters. The filters were then overlayed with synemin and incubated with antibodies as described (Hirako et al., 2003).
For copolymerization assays, mixtures of synemin and either GFAP or vimentin were assembled in buffer U (see above) at a 1:100 molar ratio (synemin:GFAP or vimentin), with GFAP or vimentin at a final concentration of 0.5 mg/ml. Control samples contained only synemin at the same final concentration as in the protein mixtures. Vimentin or GFAP IF assembly was obtained by dialysis against buffer P (2 mM Tris-HCl, pH 7.2, 140 mM NaCl and 0.5 mM DTT), and the samples were ultracentrifuged at 100,000 g for 30 minutes.
To determine whether synemin binds to GFAP or vimentin IFs after they have polymerized, synemin (0.1 mg/ml), GFAP (0.5 mg/ml) and vimentin (0.5 mg/ml) in buffer U (see above) were individually dialyzed against buffer P. This induces the assembly of GFAP or vimentin IFs, but not of synemin (Bilak et al., 1998). The synemin sample was ultracentrifuged at 100,000 g for 30 minutes. The supernatant (25 μl) was added to 75 μl of either GFAP or vimentin IF, or buffer P for control. Under these conditions, the molar ratio of synemin:GFAP or vimentin was in the 1:50-1:100 range. After a 1-hour incubation at 25°C, samples were ultracentrifuged at 100,000 g for 30 minutes.
SDS gel-loading buffer was added to supernatant and pellet proteins to the same final volume, and equal volumes were loaded on gels for comparison.
Transfection of SW13-cl2 cells
Plasmid DNA was obtained with the endotoxin-free plasmid purification kit from Qiagen (Valencia, CA). For stable transfection, the plasmids were linearized with MluI (for α- and β-synemin) or MfeI (for GFAP and vimentin). Transfections were performed on SW13-cl2 cells by adding for 4 hours in 10 cm2 well 2 μg of DNA forming a 1:3 (w/v) complex with Fugene 6 reagent (Roche). After 1-2 days, cells were stained by immunofluorescence or, for stable transfection, trypsinized and plated into medium containing 1.5 mg/ml geneticin. After ∼2 weeks, geneticin-resistant clones were lifted off the dishes by trypsinization using cloning cylinders.
Acknowledgements
We thank Paula Polk and Lijia Yin from the LSU-Shreveport core facility for assistance with quantitative RT-PCR experiments. This work was supported by National Institutes of Health grant NS-35317 to O.S., the Swedish Medical Society (16850) to U.W. and the Swedish Research Council (11548), Hjärnfoden, LUA/ALF Göteborg, Stroke Foundation and Torsten and Ragnar Söderbergs Foundations to M.P.