Parvin, a 42 kDa focal adhesion protein, related to the α-actinin superfamily

Proteins of the α-actinin superfamily crosslink actin filaments into tight bundles or meshworks of looser arrangement, and also connect them to the plasma membrane (Matsudaira, 1991; Matsudaira, 1994; Puius et al., 1998). The common feature of these proteins is the presence of the N-terminal actin-binding domain (ABD), a protein module of about 250 amino acids; otherwise they have diverse organizations (Puius et al., 1998). The ABDs of the α-actinin-related proteins are typically followed by rod domains, which are composed of a number of repeats that often mediate the protein dimerization, as well as by diverse C-terminal structural motifs. The rod domain repeats can have an immunoglobulin-like fold, as in the case of the filamin family (Fucini et al., 1997; McCoy et al., 1999), or double helical coiled coil motifs, as in case of cortexillins (Faix et al., 1996). The spectrin family (α-actinin, β-spectrin, dystrophin, utrophin) is characterized by rod domains composed of triple-helical coiled coil repeats (Pascual et al., 1997). Dystrophin and utrophin do not assemble as dimers, despite the structural similarity of their rod domain repeats with the those of β-spectrin. They also do not crosslink actin filaments (Winder et al., 1995b), but rather connect them to membrane-associated proteins, providing a link between the actin cytoskeleton and the extracellular matrix via the transmembrane dystroglycan complex (Matsumura et al., 1992; Tinsley et al., 1994; James et al., 1996). Likewise, dimerization and actin crosslinking activity has not been shown for plectin, dystonin and MACF (ACF7, trabeculin) and kakapo. The Drosophila protein, kakapo (Gregory and Brown, 1998; Prokop et al., 1998; Strumpf and Volk, 1998), and its mammalian homologue, MACF (Karakesisoglou et al., 2000; Leung et al., 1999), have been identified as linkers between integrins, actin filaments and microtubules. Plectin and dystonin, the most versatile cytolinkers, connect the microfilament system to intermediate filaments and microtubules (Fuchs and Yang, 1999; Leung et al., 1999; Wiche, 1998; Yang et al., 1999). As an exception, the proteins of the fimbrin (plastin) family, which lack rod domains, possess two ABDs, and therefore do not require dimerization to crosslink actin filaments into tight bundles (Matsudaira, 1994).

Actin binding is the most conserved property of the proteins of the α-actinin superfamily. The actin-binding region of α-actinin, which became a prototype of the ABDs of the other members, consists of two subdomains (calponin homology (CH) domains) in tandem, which have probably arisen by gene duplication of a single copy (Castresana and Saraste, 1995; Matsudaira, 1991). A single CH domain is found in, among others, calponin, Vav, IQGAPs and SM22-α. According to phylogenetic analysis the CH domains of different proteins were divided into three major groups (Banuelos et al., 1998; Stradal et al., 1998). The first two groups include N-terminal (CH1) and C-terminal (CH2) CH domains of the α-actinin related proteins. This indicates that the N-terminal CH domains of proteins with an ABD are more similar to each other than to the C-terminal CH domains of the same proteins. The third group is formed by the CH domains of the proteins that contain single CH domains.

The three-dimensional structures of the CH2 domain of human β-spectrin (Djinovic-Carugo et al., 1999), and of fimbrin (Goldsmith et al., 1997), utrophin (Moores et al., 2000) and dystrophin ABDs (Norwood et al., 2000) have been solved. Although the primary structure of the proteins varies significantly, the crystal structures revealed that the core CH domain structure is preserved in all molecules, with three α-helices forming a triple helical bundle and the fourth one lying perpendicular to them (Norwood et al., 2000). Despite the structural similarity, CH domains of different groups seem to have distinct functions. The N-terminal CH domain (CH1) alone is able to bind actin, although with lower affinity than the entire ABD (Way et al., 1992; Winder et al., 1995a). The C-terminal CH (CH2) domain alone has week affinity for actin, but may contribute to stabilizing the overall binding of the complete ABD (Banuelos et al., 1998; Way et al., 1992; Winder et al., 1995a). Since the single CH domain of calponin fails to bind F-actin in in vitro sedimentation assays and fails to target stress fibers or membrane ruffles of transfected fibroblasts (Gimona and Mital, 1998), the function of such domains remains unclear. Thus, structurally equivalent single and tandem CH domains seem to have different properties with respect to interaction with filamentous actin. Therefore, not only the duplication, but also the alteration of amino acid composition of CH domains, results in a distinct module capable of interacting with filamentous actin.

In addition to interacting with F-actin, ABDs can have additional functions like binding to G-actin (Andrä et al., 1998), PIP₂ (Fukami et al., 1996), the cytoplasmic domain of hemidesmosomal integrin β4 subunit (Geerts et al., 1999), vimentin (Correia et al., 1999) and extracellular regulated kinase (ERK). Thus, the ABD, originally considered as an actin-binding unit, may harbor multiple binding sites that support versatile interactions.

Being the components of the subplasmalemmal cytoskeleton, many proteins of the α-actinin superfamily provide physical linkage of actin filaments to the integral and peripheral proteins of cell membranes (Puius et al., 1998). Several proteins like α-actinin (Blanchard et al., 1989), plastins (Arpin et al., 1994; Bretscher and Weber, 1980), filamin (ABP280) (Pavalko et al., 1989), plectin (Seifert et al., 1992), dystrophin and utrophin (Belkin and Burridge, 1995; Kramarcy and Sealock, 1990) have been identified at actin-rich focal contacts, specific cytoskeleton-plasma membrane molecular complexes, recruited to the clustered integrins upon interaction with extracellular matrix (ECM) (Bershadsky and Geiger, 1999; Burridge and Chrzanowska-Wodnicka, 1996). Localization of these proteins to focal adhesions was shown to be mediated via direct interaction of α-actinin and filamins with β-integrins (Loo et al., 1998; Otey et al., 1990; Sampath et al., 1998), and dystrophin and utrophin with dystroglycan (Belkin and Smalheiser, 1996).

Considering the importance of proteins of the α-actinin superfamily in organizing the subplasmalemmal cytoskeleton, we have screened the expressed sequenced tag (EST) database searching for new proteins with ABD domains. We report the cloning and characterization of α-parvin as an actin-binding protein localized to cell-matrix adhesions via its C-terminal CH domain. Together with the closely related paralogue β-parvin and the more distant γ-parvin, identified in human and mice, as well as parvin-like proteins found in Caenorhabditis elegans and Drosophila melanogaster, α-parvin forms a novel family of small α-actinin related proteins.

For α-parvin, two overlapping EST-clones (GenBank Accession Numbers AA691868 and AA791745) were identified by BLAST search (http://www.ncbi.nlm.nih.gov/BLAST/) as encoding a novel protein homologous to the actin-binding domain of α-actinin. The EST clones were obtained from the Resource Center of the German Human Genome and sequenced. The 5′-end was extended by rapid amplification of cDNA ends (RACE) PCR with adaptor-ligated mouse brain and skeletal muscle Marathon-Ready cDNA (Clontech, Heidelberg, Germany) as a template, according to the manufacturer’s protocols. The whole sequence was verified by RT-PCR of overlapping fragments (Fig. 1). The PCR products were cloned into pGEM-TEasy vector (Promega, Madison, WI) or pCR2.1-vector (Invitrogen, Carlsbad, CA) for sequencing. The cDNA sequence of α-parvin has been deposited in GenBank (Accession Number AF237774).

EST clones for the human α-, β- and γ-parvin as well as for the mouse β- and γ-parvin were identified by NCBI Xblastn program (Altschul et al., 1997) using the mouse α-parvin peptide sequence. The obtained EST sequences were used for contig assembly. The cDNAs of human and mouse parvin were cloned by reverse transcribed (RT)-PCR, with total RNA prepared by phenol/chloroform extraction (Sambrook et al., 1989) as a template. RNA from human embryonic kidney (293) cell line was used for amplifying human α- and β-parvins, RNA from the human T-cell line (Jurkat) for human γ-parvin, and from mouse heart for mouse β-parvin. Mouse embryonic cDNA (Clontech) served as template for cloning mouse γ-parvin. (The cDNA sequences have been deposited in GenBank under the Accession Numbers AF237771 for human α-parvin, AF237769 for human β-parvin, AF237772 for human γ-parvin, AF237770 for mouse β-parvin and AF312712 for mouse γ-parvin.) Reverse transcription of total RNA was carried out using the M-MLV-reverse transcriptase (Promega). The resultant cDNA was amplified with High Fidelity Expand Taq-polymerase (Boehringer, Mannheim, Germany). DNA sequencing was performed with the dideoxy nucleotide termination method using a DNA sequencer (ABIprism 377 DNA sequencer, Perkin-Elmer, Norwalk, CT) at the service facilities of the Center for Molecular Medicine Cologne.

For the analysis of the mRNA transcription pattern, multiple tissue northern blots (Clontech) from mouse embryo and adult mice were hybridized with cDNA probes 1 or 2 (Fig. 1), according to the manufacturers protocols. Hybridization with a β-actin cDNA was used as a control. Probes were ³²P-labeled by random priming using the Prime-It kit (Stratagene, La Jolla, CA) according to the instructions of the manufacturer. Trancription of α-parvin in mouse testis and myoblasts was analyzed by reverse transcriptase (RT)-PCR analysis. Mouse testis Marathon cDNA (Clontech) was used as a template to amplify the fragments marked as probes 3 and 4 (Fig. 1). Total RNA isolated from C2F3 myoblasts on differentiation days 0, 2, 4 and 6 was reverse transcribed and used as a template to amplify the probe 3.

Throughout the plasmid preparations, standard molecular biological methods (Sambrook et al., 1989) were applied. For bacterial expression of recombinant α-parvin, cDNA was amplified using primers with add-on sequences for a SphI and a PstI restriction site, which allowed cloning into the pQE-30 vector (Qiagen, Hilden, Germany). Plasmids encoding α-parvin or its deletion derivatives fused to the C or N terminus of EGFP were generated by amplifying the full-length or truncated parvin by PCR, using forward and reverse primers with extensions for restriction sites for KpnI and SmaI, respectively. The amplified products were cloned into KpnI/SmaI-cut pEGFP-N3 and pEGFP-C1 vectors (Clontech). The DNA fragments encoding parvin fused to GFP were then excised from the vector as EcoRI-NotI fragments and cloned into a retroviral vector pBMN-Z from which the lacZ gene had been removed. The pBMN-Z-vector containing the parvin-EGFP fragments was used for transfection of the packaging cell line, Phoenix-ECO or Phoenix-AMPHO, as described previously (Clemen et al., 1999). Infection efficiency was nearly 100% and cells were stably producing the entire GFP-parvin or the truncated parvin deletion mutants.

The recombinant His6-tagged parvin was expressed in Escherichia coli M15[pREP4]. E. coli cells transfected with the pQE-30 plasmid encoding the 6xHis-α-parvin were grown at 37°C to an A₆₀₀ of 1.6. Cells were lysed by freezing-thawing in a lysis buffer (50 mM NaHPO₄, 300 mM NaCl, 10 mM imidazole, pH 8) in the presence of benzamidine, aprotinine, leupeptine, pepstatin A, PMSF and lysozyme (Sigma, Deisenhofen, Germany) for 30 minutes. After sonication and centrifugation at 4°C for 30 minutes at 15,000 rpm the clarified supernatant was incubated with Ni-agarose (Qiagen) at 4°C overnight and then loaded onto a column. Lysis buffer in the column was gradually exchanged against 10 mM imidazole, pH 8. After washing with 20 mM imidazole, parvin was eluted with 150 mM imidazole buffer at pH 8. The protein concentration was determined by the method of Bradford (Bradford, 1976).

Rabbit skeletal muscle actin was purified from acetone-dried muscle powder, as described (Spudich and Watt, 1971), and stored in G-buffer (5 mM Tris-HCl, 0.5 mM ATP, 0.1 mM CaCl₂, 0.5 mM DTT, pH 7.6) drop-frozen in liquid nitrogen. Concentration of G-actin solutions was determined spectrophotometrically at 290 nm, using an absorption coefficient of 0.63 mg ml⁻¹ cm⁻¹ (Houk and Ue, 1974).

For F-actin co-sedimentation assay, protein samples were centrifuged at 100,000 g for 30 minutes immediately before use. Polymerization of 10 μM actin was initiated by mixing with polymerization buffer (100 mM imidazole, 20 mM MgCl₂, 2 mM CaCl₂, pH 7.4) in the absence or presence of various concentrations of α-parvin. After 1 hour of incubation at room temperature, the mixture was centrifuged at 120,000 g for 1 hour at 4°C. Corresponding amounts of the supernatant and the pellet were subjected to SDS-polyacrylamide gel electrophoresis followed by protein staining with Coomassie Brilliant Blue. Protein bands were quantified with a scanning densitometer 300A (Molecular Dynamics, Sunnyvale, CA) and used to determine the amount of α-parvin bound to F-actin.

C3H/10T1/2 fibroblasts and mouse myoblast cell line C2F3 were propagated and differentiation of C2F3 was induced as described previously (Clemen et al., 1999). Rat vascular smooth muscle cells (A7r5), human epidermal cells (A431) and primary human foreskin fibroblasts (HFF) were propagated in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal calf serum, penicillin and streptomycin. For the propagation of human embryonic kidney cells (293) pyruvate was added to the medium. Human T lymphocytes (Jurkat) were maintained in RPMI 1640 containing 10% fetal calf serum and 2 mM L-glutamine.

Monoclonal antibodies against recombinant His-tagged mouse α-parvin were produced as described (Schleicher et al., 1984), except that the ImmunEasy Mouse Adjuvant (Qiagen) was used. Culture medium from hybridoma cells producing antibody against α-parvin was taken for immunolabelling.

For immunofluorescence, C3H/10T1/2 fibroblasts were allowed to attach onto glass coverslips for 1-16 hours, rinsed with PBS, fixed in 3% paraformaldehyde and permeabilized with 0.1% Triton X-100 in PBS for 5 minutes. Alternatively, cells were fixed with methanol at –20°C for 10 minutes. After preblocking in PBG (0.5% BSA, 0.045% fish gelatine, both from Sigma, in PBS at pH 7.4) supplemented with 10% goat serum, the cells were incubated with the primary antibody, washed with PBS and incubated with the secondary Cy3-conjugated antibody. For the detection of F-actin, the cells were treated with TRITC-phalloidin. The mouse anti-vinculin, anti-talin, anti-α-actinin, TRITC-phalloidin and the secondary Cy3-conjugated antibodies were purchased from Sigma. Mouse anti-tensin antibodies were purchased from Transduction Laboratories (Lexington, KY). For mitochondria staining fibroblasts were stained with MitoTracker-Red dye (Molecular Probes, Eugene, OR) according to the manufacturers protocol. Specimens were mounted in Gelvatol/DABCO (Sigma) and examined under a confocal laser-scanning microscope TCS-SP (Leica) or a fluorescence microscope DMR (Leica, Solms, Germany) equipped with cooled charged-coupled device (CCD) camera (PCO, Kehlheim, Germany).

In a search for proteins with homology to the ABD of α-actinin in the mouse EST database, we have identified two overlapping cDNA clones encoding a novel protein, that shares low, but statistically significant, similarity with several known family members. Sequence analysis of cDNAs extended by RACE-PCR revealed two transcripts (4324 and 1591 bp) sharing a 132 bp 5′-untranslated region (UTR) and one single open reading frame (ORF) extending over 1119 bp. These two transcripts arise as a consequence of two different polyadenylation signals (AATAAA) at positions 1567 and 4299 bp, which result in 3′-UTRs of different length. In addition, a splicing variant that lacks a large stretch in the 3′-UTR, namely bases 2519-3964, was identified by RT-PCR. Further analysis of the EST database identified another variant, W18337, which spliced out a 3′-UTR stretch from 1414 to 4242 bp (Fig. 1). The α-parvin ORF encodes a 372-amino acid protein with a calculated molecular mass of 42,330.

The mouse α-parvin protein sequence was used to search human and mouse EST databases. The obtained EST sequences were assembled into three mouse and three human cDNA contigs that corresponded to α-parvin and its two paralogues, designated as β- and γ-parvins, for the closely and the more distantly related proteins, respectively. The human and mouse β-parvin cDNA encoded a 365 amino acid proteins with 74% identity and 85% similarity to mouse α-parvin (Fig. 2A). The cDNAs of human and mouse γ-parvins coded for proteins of 331 amino acids that show 42% identity and 67% similarity to paralogous α-parvins. C. elegans and D. melanogaster were also found to harbor parvin-like proteins. Parvin-related protein of C. elegans shared 48% identity and 66% similarity to mouse α-parvin (Fig. 2A). The parvin-like CG12533 predicted gene product in D. melanogaster, a peptide of 692 amino acids, showed homology to parvin in the C-terminal 357 amino acids, starting with Met 326. Given that no Drosophila EST was found to connect the cDNA upstream and downstream of the ATG for Met 326, it might well be that the CG12533 predicted gene product was a result of erroneous joining of two independent genes. Therefore, only the C-terminal 357 amino acid peptide sequence of the Drosophila gene was used in the phylogenetic analysis.

The parvins are composed of a single ABD preceded by a N-terminal stretch. The overall primary structure of parvins is very conserved with the N-terminal domain being the most variable region of the molecule (Fig. 2A) varying from 82 amino acids in γ-to 125 amino acids in β- and 133 amino acids in α-parvins. The N-terminal domains of α- and β-parvin contain two putative nuclear localization signals (NLSs) at positions corresponding to residues 20 and 39 of α-parvin (Fig. 2A), and three potential SH3-binding sites (consensus PxxP, where x is any amino acid but cysteine (Cohen et al., 1995)). The first and the second CH domains forming the ABDs share about 20% identity and are separated by a linker of about 60 residues (Fig. 2A). The linker region between the two CH domains is longer than in other related proteins and displays some similarity to the parts of β-spectrin and plectin that precede their respective CH1 domains. The sequences of C. elegans and D. melanogaster parvin-like proteins display stronger homology to mammalian α- and β-parvin than to γ-parvin.

A phylogenetic analysis of the CH domains (Fig. 2B) of parvin and various members of the α-actinin superfamily shows that CH1 and CH2 of parvins create their own branch diverging from a point between the CH1 domains of ABD-containing proteins and single CH domains. The CH1 and CH2 domains of the other α-actinin-related proteins cluster separately, reflecting the higher degree of similarity between CH1 domains of different proteins compared with CH1 and CH2 domains from the same protein, in agreement with previous findings (Stradal et al., 1998, Banuelos et al., 1998). The CH domains of proteins containing a single copy of CH form the separate branch with an exception of smoothelin, which, albeit containing only a single CH domain (van der Loop et al., 1996), shows a higher degree of similarity to the CH2 domains of proteins with two CH domains in tandem. Surprisingly, unlike the other α-actinin-related proteins, the C-terminal CH domains

of parvins shows higher degree of homology to their own CH1 than to the CH2 domains of other ABD-containing proteins (Fig. 2D). The CH1 of α-actinin shares 24% identity and 50% similarity with the second CH domain, but only 17% identity and 36% similarity with the first CH domain of α-parvin. The stronger similarity of the α-parvin CH domains to each other than to any other type of CH domain suggests that the parvins represent a separate family of proteins within the α-actinin superfamily. This is also confirmed by the sequence analysis of the whole ABDs that represent the major families of the α-actinin superfamily, together with parvins (Fig. 2C). In the evolutionary tree based on the alignment of the entire actin-binding regions, the parvins group separately from the rest of the proteins of the α-actinin superfamily. They are positioned closer to the ABDs of fimbrin (Fig. 2C), probably because fimbrins, as parvins, harbor relatively long, although different, linkers between their CH domains. The phylogenetic tree of the ABDs shows two major branches in the parvin family. One branch includes mammalian γ-parvins and the second contains the rest of the proteins with Drosophila and Caenorabditis parvins positioned closer to mammalian α-parvins. The branch of mouse α-parvin is undistinguishable from its human counterpart, owing to the highly conserved ABDs bearing a single substitution (Fig. 2C). This suggests that α-parvin represents the evolutionary most conserved family member.

A northern blot containing purified polyA⁺ RNA from various tissues was hybridized separately with probe 1, specific for both transcripts, or probe 2, specific for the 4.4-kb transcript (see Fig. 1A). Northern blot analysis with probe 1 recognized two mRNA species that were 1.6 and 4.4 kb in length (Fig. 3B), the sizes of which correlate well with the cDNA sequences. According to the sequence analysis, these two transcripts may have arisen as a consequence of the two different polyadenylation signals. Two mRNA species showed varying levels of expression in different mouse tissues. The 1.6-kb mRNA was generally expressed at lower levels than the 4.4-kb mRNA, however its tissue distribution correlated with the larger transcript (Fig. 3B). Hybridization of the same blot with the probe 2, specific for the longer mRNA, recognized only the 4.4-kb transcript (Fig. 3A). An especially high level of the 4.4-kb mRNA was detected in kidney and heart, weaker signals were observed in brain, lung and liver. As no signal was detected in mouse testis by Northern blot, we attempted to assess the expression of the α-parvin gene by means of RT-PCR. The fragments named as probe 3, corresponding to the ORF, and probe 4, corresponding to the 3′-UTR of the 4.4-kb transcript (Fig. 1), were amplified from testis cDNA, revealing the presence of both α-parvin mRNA species (Fig. 3D). This indicates that the amount of mRNA in testis was below the detection level for northern blots. The fragment corresponding to the probe 3 was also amplified using cDNA of differentiating myoblasts as a template, where we detected transcripts at all time points (Fig. 3E).

To analyze the developmental regulation of α-parvin, we have hybridized probe 1 to a northern blot containing RNA from 7-, 11-, 15- and 17-day-old mouse embryos. Both α-parvin mRNA species were expressed throughout mouse development. Again, the level of expression of the 4.4-kb transcript was significantly higher in comparison with the level of the shorter mRNA (Fig. 3G).

All proteins with ABDs composed of two CH domains known to date are able to bind filamentous actin. Considering that the ABD of α-parvin differs from the other ABDs of the other α-actinin-related proteins in its ‘reverse’ order of the CH domains and a longer linker between them, we were interested to explore whether this moderate similarity was sufficient to confer actin-binding properties to α-parvin. We tested the potential F-actin-binding properties of the full-length His-tagged recombinant α-parvin in a high-speed co-sedimentation assays with F-actin and found that α-parvin co-sedimented with F-actin (Fig. 4A, lane 4), whereas in control experiments most of the His-tagged α-parvin was found in the supernatant fraction after ultracentrifugation (Fig. 4A, lane 1). Small amounts of parvin, however, significantly less than in the presence of actin, were also seen in the pellet, probably caused by precipitation (Fig. 4A, lane 2). The actin-binding affinity was estimated by quantifying the results of four independent sedimentation assays (15 samples in total) with varying amounts of parvin (Fig. 4B). The K_d value for the interaction of the recombinant His-tagged α-parvin with actin was calculated to be 8.4±2.1 μM, which is in the same range as reported for the ABDs of other members of this superfamily.

To determine the subcellular localization of parvin, we employed GFP tagged α-parvin, because of the lack of antibodies distinguishing between parvin isoforms. Overall, expression of GFP tagged α-parvin in C3H/10T1/2 fibroblasts, even in the strongly expressing cells, did not lead to changes in actin cytoskeleton and focal contacts organization when compared with untransfected fibroblasts upon visualization with rhodamine-phalloidin and anti-vinculin immunolabelling, respectively. Cells transfected with the constructs coding for GFP fused to the N terminus of α-parvin showed identical distribution patterns to the cells with GFP at the C terminus. Therefore, data shown throughout are obtained with GFP fused to the C terminus of α-parvin. The expression of GFP-α-parvin fusion protein was verified by western blotting, using an anti-GFP monoclonal antibody (data not shown).

In well-spread fibroblasts, GFP-α-parvin preferentially appeared at the focal adhesions (Fig. 5A,B). In addition to that, GFP-parvin was localized to a distinct set of fibers in the central area of the cell running nearly parallel to one another (Fig. 5B). In fibroblasts forming contacts with several neighboring cells, these fibers may run in various directions (Fig. 5A). GFP-parvin was also observed at membrane ruffles, which were well seen in spreading fibroblasts (Fig. 5C).

To find out whether the distribution of the GFP-α-parvin observed in the C3H/10T1/2 fibroblasts reflects the general properties of the protein, or is limited to this cell type, we transfected several other cell lines, including mouse myoblasts (C2F3), rat vascular smooth muscle cells (A7r5), human epidermal carcinoma cells (A431) and primary human foreskin fibroblasts (HFF). GFP-α-parvin was associated with the cell-substrate structures (focal contacts) in all cell types studied. In epithelial cells (A431) GFP-α-parvin also accumulated at some cell-cell adhesions in addition to the focal contacts and nuclei (Fig. 5D), which implies involvement of α-parvin in general cell-adhesion events.

To confirm the presence of GFP-parvin in focal adhesions and to explore the relationship between focal contacts and parvin in more detail, we labeled the transfected fibroblasts with antibodies against the focal contact proteins vinculin, talin and α-actinin. Notably, GFP-α-parvin was not always associated with focal contacts positive for vinculin. This can be observed particularly well during fibroblast spreading, which involves formation of focal adhesions at the cell periphery (Fig. 6). In some spreading cells, fixed within 30-90 minutes after plating, GFP-α-parvin was not recruited to nascent or to the well-developed focal adhesions, as observed by the intensity of anti-vinculin labeling (Fig. 6A). Instead, GFP-α-parvin was distributed diffusely throughout the cytoplasm with enrichment at lamellipodial ruffling. At a later stage of cell spreading, GFP-α-parvin can be seen at the cell membrane and also weekly incorporated at vinculin-positive focal adhesions (Fig. 6B). When the cell membrane was depleted of GFP-α-parvin, the GFP-α-parvin fluorescence became stronger at the focal contacts (Fig. 6C). In addition, it localized to the central fibers, which were weakly labeled with vinculin (Fig. 6C). In contrast, in fibroblasts, which were allowed to spread overnight, the GFP-α-parvin fluorescence overlapped significantly with the anti-vinculin pattern (Fig. 7D,F). Occasionally, GFP-α-parvin labeling at focal contacts was found to be more elongated towards the cell center when compared with the anti-vinculin staining. The anti-vinculin labeling of the α-parvin-positive central fibers appeared to be weaker than the GFP-α-parvin fluorescence. These observations suggest that α-parvin was not necessary for the formation of focal adhesions, as it was not associated with nascent newly formed adhesion complexes.

To investigate the nature of the GFP-α-parvin-rich fibers located in the central cell area we stained the transfected fibroblasts with TRITC-phalloidin to label F-actin. In some cells GFP-α-parvin fluorescence was extended from focal contacts towards the center of the cell, showing partial co-localization with stress fibers (Fig. 7A-C). Additionally, GFP-α-parvin fluorescence revealed that not all filaments appear to be associated with F-actin. This was even more obvious in specimen immunolabeled for α-actinin (Fig. 7H,I), which localizes to focal contacts and actin stress fibers. We compared the GFP-α-parvin distribution with the anti-α-actinin pattern and found that GFP-α-parvin co-localized with α-actinin in focal contacts. As GFP-α-parvin extended from the focal adhesions, it followed the stress fibers marked by the typical for α-actinin striated pattern. However, not all GFP-α-parvin-fibers in the center of the cell appear to overlap with anti-α-actinin staining. Thus, the data imply that association of α-parvin with central fibers was not mediated by actin. Next, we examined whether GFP-α-parvin was also localized to focal contacts stained by anti-talin antibodies and found that parvin co-localized completely with talin in focal adhesions as well as with fibers in the central area of the fibroblasts (Fig. 7K,L). The resemblance of the α-parvin-rich central fibers to tensin-associated fibrillar adhesions, a distinct type of adhesions involved in matrix remodeling (Katz et al., 2000; Zamir et al., 2000), led us to examine the co-distribution of GFP-α-parvin and tensin. Anti-tensin antibodies were hardly detected in the focal contacts, but strongly labeled the central α-parvin fibers (Fig. 7N,O). Thus, α-parvin appears to be associated with various types of cell-matrix adhesion structures, including mature focal contacts and fibrillar adhesion-like structures.

In addition, GFP-α-parvin was strongly enriched in nuclei of some cells (Figs 7G,J), but not in others (Fig. 7A,D). This difference appears to correlate with the expression level of α-parvin in the transfected cells. Cells expressing lower level of parvin did not accumulate it in their nuclei, although it was clearly associated with the focal-contact-like structures (Fig. 7A,D). The appearance of GFP-α-parvin in the nucleus could be due to the presence of the two nuclear localization signals identified in the N-terminal part of α-parvin. Nuclear staining was also observed in control cells expressing GFP alone, probably as a result of diffusion of the 28-kD GFP through the nuclear pores. Although the molecular mass of GFP-α-parvin fusion protein (71 kDa) is higher than the generally accepted threshold of 30 kDa for proteins that can passively diffuse through the nuclear pores, we employed methanol fixation to exclude unspecific nuclear targeting. Indeed, after this fixation, no GFP fluorescence was observed in control cells expressing GFP alone. The GFP fusion protein remained, in contrast, associated with intranuclear structures (data not shown).

To compare the distribution of the endogenous parvin with the GFP-fused protein we employed the monoclonal antibodies raised against mouse α-parvin. The immunolabeling overlapped with the GFP-α-parvin pattern in the focal contacts (Fig. 8A-C) and the nuclei (not shown). The staining of endogenous parvin in untransfected fibroblasts with the anti-parvin antibodies resulted in a pattern similar to the one of the GFP-α-parvin-expressing cells (Fig. 8D). The fiber-like structures, also recognized by the antibodies, could result from the cross-reaction with β- and γ-parvin.

To further characterize the function of α-parvin and to map the domains responsible for targeting different cell compartments, a set of overlapping fragments of parvin fused to GFP was generated and prepared for retroviral transfection into C3H/10T1/2 fibroblasts. A schematic representation of the domain structure with the expressed fragments is shown in Fig. 9. None of the deletion mutants affected the apparent actin filament organization and focal contact distribution as monitored by labeling with TRITC-phalloidin and anti-vinculin, respectively. Deletion of the second CH domain (GFP-α-parvin^1-273) abolished its localization to focal adhesions and centrally located fibers. GFP fluorescence in cells transfected with this construct was diffuse throughout the cytoplasm in fully spread cells (Fig. 10A) and could not be found at focal contacts visualized with anti-vinculin antibody (Fig. 10B,D). However, the truncated protein was enriched at cell ruffles in moving or spreading cells (Fig. 10C).

In contrast, GFP-α-parvin^257-372, composed solely of the second CH domain, was targeted to focal adhesions and central fibrillar structures in fully spread fibroblasts (Fig. 10E,F). In migrating cells, GFP-α-parvin^257-372 showed a distribution similar to the full-length construct as it was attracted to the leading edges of migrating cells (Fig. 10G,H). The distribution of this GFP-α-parvin^257-372 was similar to the staining pattern of vinculin, although the fluorescence signal at the focal contacts seen with this construct appeared to be generally weaker than the one with the full-length GFP-α-parvin. This implies that the second CH domain is responsible for the attraction of parvin to the fibrillar adhesions and thus could represent a module interacting with molecular complexes mediating cell-matrix adhesion. The first CH domain of α-parvin appears to reinforce or stabilize the binding of the second CH domain, as the focal contact fluorescence of the latter was weaker then the one of full-length GFP-α-parvin.

Transfection with the construct coding for GFP-α-parvin^1-93, composed of the most N-terminal part of parvin, resulted in weak diffuse staining in the cytoplasm after paraformaldehyde-Triton X-100 fixation, and fluorescence was completely abolished in methanol fixed cells. Yet, the GFP-α-parvin^1-93 fluorescence was clearly seen in the nuclei of the intact unpermeabilized cells. Thus, these observations indicate that the predicted NLSs found in the N-terminal part of α-parvin may be responsible for the import to the nucleus, but not sufficient to confer targeting of parvin to its nuclear binding partner. All constructs of GFP-α-parvin and its fragments, with an exception of GFP-α-parvin^188-372 and GFP-α-parvin^188-273 (see below), were found to localize to nuclei, with the full-length GFP-α-parvin resulting in the strongest nuclear fluorescence. Taken together, these data may indicate that the sites involved in association with nuclear targets are located in the C-terminal two thirds of α-parvin, whereas the NLSs are positioned in the N-terminal region.

An unexpected fluorescence pattern was observed with GFP-α-parvin^188-372, which was attracted to reticular structures identified as mitochondria by co-staining with MitoTracker-Red marker, whereas the localization to focal or fibrillar adhesions, to lamellae and to the nuclei was completely abolished (Fig. 10I,J). Further truncation of this construct yielded GFP-α-parvin^188-273 and identified the amino acids 188 through 273 of α-parvin as being sufficient for targeting to mitochondria. Since no mitochondria targeting was observed with the full-length GFP-α-parvin, it is possible that the association of GFP-α-parvin^188-372 and GFP-α-parvin^188-273with mitochondria was an artefact caused by exposure of residues mediating mitochondrial targeting.

We have identified and characterized α-parvin, a novel 42-kD actin-binding protein associated with focal adhesions. Sequence analysis of α-parvin revealed a pair of CH domains forming an ABD, which is characteristic for a large α-actinin/β-spectrin superfamily of proteins. In contrast to other known proteins of this superfamily, the C-terminally located ABD of α-parvin is not followed by any of the protein motifs typical for the other members. Since α-parvin appears to be the smallest member of the α-actinin-superfamily known so far, it was given a name derived from the Latin term ‘parvus’ for ‘small’.

In view of the differences between amino acid sequences of the CH domains of parvin and other α-actinin-related proteins, it was interesting to examine actin-binding properties of α-parvin. For this we used His-tagged bacterially expressed α-parvin. His-tagged proteins or their actin-binding domains were used previously for estimating actin-binding properties, although these data should be evaluated with caution, since at least in one case the direct comparison of nonfusion versus His-tagged protein (cortexillin fragment) showed that the binding to actin was enhanced, probably due to electrostatic interactions of the positively charged residues of the tag with the negatively charged actin (Stock et al., 1999). Yet, the affinities of His-tagged ABD of dystrophin to α- and γ-actin, estimated to 13.7 and 10.6 μM, respectively (Renley et al., 1998) were highly consistent with the data obtained with the same dystrophin fragment expressed without the tag and β-actin (13 μM) (Winder et al., 1995a). The affinity of His-α-parvin to F-actin was determined to be 8.4 μM, which is comparable with the K_d value of 4.7 μM determined for α-actinin (Way et al., 1992). The regions of the primary sequences of conventional ABDs involved in direct interaction with actin filaments in dystrophin (Corrado et al., 1994; Fabbrizio et al., 1993; Levine et al., 1992), α-actinin (Hemmings et al., 1992), ABP-120 (Bresnick et al., 1990) and β-spectrin (Karinch et al., 1990) were mapped to three short stretches designated as actin-binding sites (ABSs). ABS1 and ABS2 located in the CH1 domain of α-actinin-related proteins are conserved in the C-terminal CH domain of parvin, which shows higher overall homology to CH1 (Fig. 2D). It was also shown that the CH1 domains of α-actinin (Way et al., 1992) and dystrophin (Winder et al., 1995a) harboring ABS1 and ABS2, are sufficient for binding to F-actin, although the affinity of interaction was lower than the one determined for the whole ABDs. It could therefore well be, in contrast to the other members of the α-actinin superfamily, that the C-terminal CH domain represents the major actin-binding region in parvins. The structural analysis suggest that linkers between CH domains are not directly involved in the interaction with actin filaments, but confer conformational flexibility to ABDs and play a role in arranging the CH domains to form an actin-binding surface (Norwood et al., 2000). Therefore an exclusively long linker region of parvin might not interfere with actin interaction and may be involved in positioning of the CH domains upon association with microfilaments. Thus, we demonstrated that the ABD of α-parvin is a functional module capable of binding to filamentous actin in spite of the relatively low similarity to the conventional ABDs of α-actinin-related proteins.

In vivo α-parvin fused to GFP interacted with the site where actin filaments are associated with cell-matrix adhesion complexes. Focal contacts or focal adhesions represent molecular complexes at the points of the closest apposition between the cell membrane and extracellular substrate that mediates integrin-dependent cell-matrix adhesion. The assembly and maturation of focal adhesions is a complex process (Bershadsky and Geiger, 1999; Burridge and Chrzanowska-Wodnicka, 1996), where the order of appearance and topology is not clear. In contrast to fimbrin, a protein with two ABDs of the α-actinin type that is involved in early stages of attachment (Correia et al., 1999), α-parvin does not seem to be involved in the maturation of the focal contacts (Fig. 6). GFP α-parvin appears at already formed adhesions and is also associated with the net of fibers in the central area of the cell. There is a growing body of evidence that focal adhesions are diverse and highly motile structures, which move in an actomyosin-dependent fashion, and that this movement is responsible for generation of different types of adhesion complexes (Katz et al., 2000; Smilenov et al., 1999; Zamir et al., 2000). The central fibers, positive for anti-tensin, may correspond to recently described tensin-rich fibrillar adhesions, whose molecular composition is distinct from classical focal contacts and which were shown to be involved in matrix remodeling (Katz et al., 2000; Zamir et al., 1999). Fibrillar adhesions, which are elongated structures associated with fibronectin fibrils, contain little or no paxillin, vinculin, talin and focal-adhesion kinase, and are assembled at proximal margins of focal contacts and translocated towards the cell center in an actomyosin-dependent manner (Zamir et al., 2000). The location of parvin not only at classical focal contacts but also at tensin-rich central fibers poses the implication of α-parvin in matrix remodeling and in turnover of adhesion complexes, although the exact molecular mechanism remains to be examined. Our results also indicate that the CH2 domain of α-parvin is sufficient for forming complexes with the proteins of focal contacts, since it was recruited to cell-adhesion sites when the domain alone was expressed in a fibroblast cell line. The CH2 domain of α-parvin also harbors the conserved regions ABS1 and ABS2, which are involved in F-actin binding (see above). This observation suggests that parvin may be complexed with proteins of focal contacts by virtue of binding F-actin, and that it associates with actin filaments in an adhesion-dependent manner. Although the C-terminal CH domain of parvin was responsible for both focal contacts and fibrillar adhesions, the mechanism that targeted α-parvin to the central tensin-rich fibers may differ from the association with focal adhesions, given that the central fibers were not often co-localized with actin filaments.

Actin seems to be an accepted nuclear component, and an increasing number of actin-binding proteins have been reported to reside in the nucleus or to translocate to the nucleus under regulatory control (reviewed by Rando et al., 2000). The observed nuclear targeting of GFP-α-parvin is in agreement with the two putative nuclear localization signals, which were identified in the N-terminal part of α-parvin. Strong nuclear localization of GFP-α-parvin, which is too large to diffuse passively into the nucleus, suggests that the identified NLSs are functional. Analysis of the distribution of the truncated α-parvin shows that the fragments of α-parvin lacking both NLSs are also observed in nuclei, although not to the same extent as the full-length GFP-α-parvin. The N-terminal α-parvin fragment encompassing the first 93 residues harboring both predicted NLSs is detected in nuclei of the intact cells, but not after fixation and permeabilization. A likely explanation for this could be that the GFP-α-parvin^1-93 contains the NLSs responsible for the transport to the nucleus, but lacks the CH domains that may encompass the sites responsible for the interaction with intranuclear targets.

Our data identify α-parvin as a novel component of focal adhesions recruited to the sites of cell-matrix adhesion via its second CH domain. Given that this CH domain of mouse α-parvin shares 84% identity and 94% similarity with paralogous

β-parvin, and 50% identity and 71% similarity with γ-parvin, we speculate that parvins constitute a novel emerging family of actin-binding proteins involved in cell-matrix adhesion. Although the exact biological function of parvins remains to be determined, parvins probably represent structural components of a link between actin filaments and transmembrane complexes associated with extracellular matrix.

This work was supported by a grant from the Center for Molecular Medicine Cologne. T.O. is a member of the Graduiertenkolleg ‘Molekularbiologische Grundlagen pathophysiologischer Vorgänge’. We thank Dr M. Aumailley and Dr M. Schleicher for helpful discussions and sharing of reagents and Dr F. Rivero for critical reading of the manuscript.

The protein identical to mouse α-parvin was identified as paxillin-binding protein and named actopaxin by Turner et al. (Turner (2000) J. Cell Sci. 113, 4139-4140).