Actinin-4, a Novel Actin-bundling Protein Associated with Cell Motility and Cancer Invasion

The human giant cell-lung cancer cell line, Lu-65, was established previously in our laboratory (Yamada et al., 1985). Primary cultured human foreskin keratinocytes (HFK)¹ were purchased from Seikagaku Co. (Tokyo, Japan) and were cultured as instructed by the supplier. Primary normal human uterine endometrial fibroblasts were isolated and cultured as described previously (Zhang et al., 1995). PC-10, a human squamous lung cancer cell line, was provided by Dr. Y. Hayata (Tokyo Medical College, Tokyo, Japan). A-431 is a human vulvar epidermoid cancer cell line (Giard et al., 1973), and KU7 is a urinary bladder cancer cell line. TE 4, TE 6, TE 7, and TE 10 were derived from esophageal cancers (Nishihira et al., 1993), and MCF7 and R27 were derived from breast cancers (Westley, 1979; Cavailles et al., 1988). The oral floor cancer cell line IMC2, normal embryonic lung fibroblast MRC5, and colon cancer cell line WiDr were obtained from Riken Cell Bank (Tsukuba, Japan). Colon cancer cell line SW480 was obtained from the American Type Culture Collection (Rockville, MD).

HFK were treated with 50 ng/ml wortmannin (Sigma Chemical Co., St. Louis, MO) for 15 min and then washed. After incubation for 24 h, the cells were examined by immunofluorescence microscopy as described below. For actin depolymerization, the cells were treated with 2 μg/ml cytochalasin D (Sigma Chemical Co.) for 2 h.

BALB/c mice were immunized with a lysate of Lu-65 cells, and hybridomas were produced as described previously (Hirohashi et al., 1984). The hybridoma supernatants were then screened immunohistochemically for reactivity with acetone-fixed paraffin-embedded human tissues, as described below.

Rabbit polyclonal anti–actinin-4 antibody was raised against a KLH-conjugated synthetic peptide, MGDYMAQEDDWC (containing amino acids 1–11, Fig. 1), and affinity-purified.

A lambda gt11 cDNA expression library of primary cultured HFK (CLONTECH Laboratories, Inc., Palo Alto, CA) was immunoscreened with mAb NCC-Lu-632, as described previously (Young and Davis, 1983), and a positive 950-bp cDNA clone (ck-20-2) was isolated. For isolation of full-length cDNA of actinin-4 and -1, a cDNA library of mesothelioma MS-1/CDDP cells was screened with synthetic oligonucleotides specific to each actinin. A 3.5-kb actinin-4 cDNA clone, near full-length as determined by Northern blot analysis (Fig. 2 B), was isolated, and both strands were sequenced using an autosequencer (model ABI 377; Perkin Elmer, Foster City, CA).

Multiple tissue Northern blots I and II were purchased from CLONTECH Laboratories and hybridization was performed using a ³²P-labeled oligonucleotide probe (5′-CGCTTCTGAGTTAGGGCCCCCA-3′) specific to actinin-4, as described previously (Sambrook et al., 1989). The quality and quantity of electrophoresed mRNA was determined by rehybridization of the same blot with a human β-actin cDNA probe.

Cells were extracted with lysis buffer (10 mM Hepes, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 1% NP-40, 1 mg/ml NaN₃, 0.5 mM Na₃Vo₄, 10 nM β-glycerophosphate, 1 mM PMSF, 1 μg/ml leupeptin, 1 μg/ml pepstatin A) on ice for 30 min before centrifugation (12,000 g, 15 min). Cell lysates (50 μg protein) were separated by SDS-PAGE (Laemmli, 1970) and transferred to Hybond–polyvinyl difluoride membranes (Amersham International, Buckinghamshire, UK) (Towbin et al., 1979). After incubation with antibodies at 4°C overnight, the blots were detected using a horseradish peroxidase–conjugated anti–mouse IgG+IgM (IBL, Fujioka, Japan) or anti–rabbit Ig antibody and ECL Western blotting detection reagents (Amersham International), as instructed by the suppliers. For blotting of whole cell lysates, cells were lysed directly with Laemmli buffer (62.5 mM Tris-HCl, pH 6.8, 2% SDS, 5% glycerol, 715 mM 2-mercaptoethanol, 0.0025% bromophenol blue) (Harlow and Lane, 1988).

A cDNA fragment of actinin-4 (encoding amino acids 410–664) and the corresponding site of actinin-1 (amino acids 418–672) were amplified by PCR and subcloned into pGEX plasmids (Pharmacia Biotech, Uppsala, Sweden). Glutathione S-transferase (GST) fusion proteins expressed in Escherichia coli were affinity-purified on glutathione–Sepharose 4B (Pharmacia Biotech) as described previously (Shibata et al., 1996).

l-[³⁵S]methionine (Amersham International) labeled polypeptides were synthesized from cDNA clones containing the entire coding regions for actinin-1 and -4, using the TNT reticulocyte lysate system (Promega Corp., Madison, WI) as instructed by the supplier. The synthesized polypeptides were analyzed by SDS-PAGE and autoradiography as described previously (Yamada et al., 1995).

Purified chicken gizzard actin (Sigma Chemical Co.) was coupled to CNBr-activated Sepharose–4B beads (Pharmacia Biotech) (Prendergast and Ziff, 1991). The beads were incubated with ³⁵S-incorporated in vitro translation products in lysis buffer overnight at 4°C. After extensive washing with lysis buffer, the beads were boiled in Laemmli buffer and analyzed by SDS-PAGE and autoradiography (Yamada et al., 1995).

A cell lysate of the lung cancer cell line PC-10 was also incubated overnight at 4°C with the actin-coupled beads. After thorough washing, the beads were boiled in Laemmli buffer and analyzed by immunoblotting as described above.

In vitro translation products of actinins were incubated with a monoclonal antibody against chicken α-actinin (BM-75.2, IgM, k; Sigma Chemical Co.), anti–actinin-4 polyclonal antibody, normal mouse IgM, or normal rabbit IgG (Sigma Chemical Co.) overnight at 4°C and precipitated with anti–mouse IgM agarose (for mouse antibodies) (Sigma Chemical Co.) or protein G plus agarose (for rabbit antibodies) (Santa Cruz Biotechnology, Santa Cruz, CA). SDS-PAGE and autoradiography were carried out as described previously (Yamada et al., 1995).

Cells were cultured on glass coverslips, fixed with 4% paraformaldehyde and 2% sucrose in PBS, and made permeable with 0.2% Triton X-100. After incubation with antibodies, actinins were detected with biotinylated anti–mouse IgM or biotinylated anti–rabbit IgG and Texas red–avidin D or FITC-conjugated avidin D (Vector Laboratories, Burlingame, CA). Actin filaments were visualized with FITC-phalloidin (Sigma Chemical Co.), and the cells were examined with a Zeiss LSM410 microscope (Thornwood, NY).

Cells were grown to confluency on glass coverslips, and a “wound” was then introduced in the monolayers with a plastic pipette tip (Glück et al., 1993; Glück and Ben-Ze'ev, 1994). After incubation for 24 h, the cells were fixed and examined as described above.

Acetone-fixed paraffin-embedded human tissues (Sato et al., 1986) were selected from the surgical pathology file of the National Cancer Center Hospital (Tokyo, Japan). 5-μm thin sections were stained with mAb NCC-Lu-632 using an immunoperoxidase avidin-biotin complex (ABC) kit (Vector Laboratories), as described previously (Sato et al., 1986). Histological subtyping of breast cancer was done by two independent certified pathologists, according to the criteria described previously (Rosen and Oberman, 1992).

The postsurgical survival curves for a total of 61 patients with stages I and II breast cancer (Rosen and Oberman, 1992) were expressed as described by Kaplan and Meier (1958). Statistical significance was determined by the generalized Wilcoxon test.

mAb NCC-Lu-632 (IgM, k) was selected on the basis of its intriguing immunohistochemical reactivity, as described below. A cDNA clone encoding the antigen molecule recognized by the mAb was isolated by immunoscreening of the HFK lambda gt11 cDNA library.

The 3.5-kb cDNA clone contained a 2,652-bp open reading frame encoding 884 amino acids (Fig. 1), with a predicted molecular mass of 102 kD. The mAb NCC-Lu-632 reacted consistently with the ∼100-kD antigen molecule in immunoblot analyses (Fig. 2,A), and in vitro translation of the full-length cDNA yielded a major protein of ∼100 kD (Fig. 3,C). This coding region and the deduced amino acid sequence showed a high degree of similarity to human α-actinin-1 (Millake et al., 1989; Youssoufian et al., 1990) (80.0% nucleotide and 86.7% amino acid similarity). It contained structures conserved among α-actinin family members, including the actin-binding domain (Kuhlman et al., 1992) (amino acids 111–125; Fig. 1), pleckstrin-homology (PH) domain containing a phosphatidylinositol 4, 5-biphosphate (PIP2)-binding site (Fukami et al., 1996) (amino acids 150–165), and two EF-hand calcium regulation domains (Witke et al., 1993; Imamura et al., 1994) (amino acids 742–770 and 783–811). Following the nomenclature proposed by Beggs et al. (1992), this gene was termed actinin-4.

Northern hybridization was carried out with an actinin- 4–specific 22-bp oligonucleotide probe to avoid cross- hybridization with actinin-1. As shown in Fig. 2,B, intense expression of 3.5-kb actinin-4 mRNA was detected in the ovary (Fig. 2 B, lane 13) and colon (lane 15), and faint expression was detected in all tissues examined. No gross structural alterations of actinin-4 mRNA were detected in any of the human carcinoma cell lines examined (data not shown).

To determine the specific reactivity of mAb NCC-Lu-632 with actinin-4, amino acids 410–664 of actinin-4 and a corresponding site of actinin-1 (Millake et al., 1989) (amino acids 418–672) were expressed as a GST fusion protein in E. coli. Immunoblot analysis revealed that the mAb reacted only with actinin-4 GST fusion protein, and not with actinin-1 GST fusion protein or GST alone (Fig. 3 A), confirming the specificity of mAb NCC-Lu-632 for actinin-4.

Immunoprecipitation analyses revealed that an mAb against chicken actinin, BM-75.2, was reactive with the in vitro translation product of cDNA of actinin-1, but not with that of actinin-4 (Fig. 3 D).

Immunoblot analyses revealed that the polyclonal antibody raised against the NH₂-terminal amino acids of actinin-4 reacted only with a single band of the same molecular weight as NCC-Lu-632 mAb (Fig. 4,A). Immunoprecipitation analyses revealed that this polyclonal antibody was reactive with the in vitro translation product of the cDNA of actinin-4, but not with that of actinin-1 (Fig. 4 B), thus confirming its specificity.

From the deduced amino acid sequence (Fig. 1), actinin-4 was predicted to contain an actin-binding domain (Hemmings and Critchley, 1992; Kuhlman et al., 1992). To confirm this, in vitro binding experiments were performed using actin-coupled agarose beads (Prendergast and Ziff, 1991). ³⁵S-labeled actinin-4 recombinant protein, produced by in vitro translation, was incubated overnight with agarose beads coupled to actin. After extensive washing, SDS-PAGE and autoradiography revealed labeled polypeptides whose size was consistent with the predicted size of actinin-4 (Fig. 5 A).

Similarly, a cell lysate of the lung cancer cell line PC-10 was incubated overnight with actin-agarose beads and washed thoroughly. Immunoblot analysis with mAb NCC-Lu-632 revealed the retention of actinin-4 on actin-agarose beads (Fig. 5 B), confirming the actin-binding activity of actinin-4.

Confocal immunofluorescence microscopy revealed that actinin-1 was localized specifically at the ends of actin stress fibers (Fig. 6, A, C, and E) and adherens junctions (Fig. 7,A), as described previously (Lazarides, 1975; Wehland et al., 1979; Knudsen et al., 1995). In contrast to this cell membrane–associated localization of actinin-1, actinin-4 protein was colocalized with actin stress fibers and was dispersed in the cytoplasm and nucleus (Fig. 6, B, D, and F). The specific immunocytochemical reactivity of monoclonal antibody NCC-Lu-632 was further confirmed by the polyclonal antibody raised against the NH₂-terminal amino acids of actinin-4 (Fig. 4,C). In most cancer cell lines including PC10, A431, SW480, TE4, TE6, TE7, TE10, and R27, actin stress fibers were poorly developed and actinin-4 was diffusely dispersed in the cytoplasm. Characteristically, actinin-4 was highly concentrated in the cytoplasm where the cells were sharply extended (Fig. 8, B, E, and H). Actinin-4 was stained intensely at the edges of cell clusters (Fig. 8, C, F, and I). The difference in subcellular localization between these two nonmuscle isoforms suggested that they were associated with different functions.

From the above immunofluorescence observations, we speculated that the expression of actinin-4 was regulated dynamically by cell movement. An artificial linear defect was introduced in the confluent monolayers of A-431 and R27 cells, and we examined the cells forced to be motile by immunofluorescence microscopy. As shown in Fig. 9, actinin-4 was markedly induced in cells along the edges of the wound (Fig. 9,A) and in cells migrating into the wound (Fig. 9, B–D).

In a limited number of cell lines, including the breast cancer cell line MCF7, oral floor cancer IMC2, and bladder cancer KU7, it was noticed that actinin-4 was localized exclusively in the nucleus (Fig. 7,B), in contrast to the association of actinin-1 with the cell membrane in these cells (Fig. 7,A). The nuclear staining was not due to cross-reactivity of the mAb with other nuclear proteins because NCC-Lu-632 mAb reacted only with a protein of ∼100 kD in whole-cell lysates of these cell lines (Fig. 7 C).

Actinin-4 possesses a conserved PIP2-binding site. PIP2 binds α-actinin through the pleckstrin homology domain and regulates its actin-binding activity (Fukami, 1992; Fukami et al., 1996). PIP2 is one of the substrates of phosphatidylinositol 3 kinase (PI3 kinase). By treating HFK with a PI3 kinase inhibitor, wortmannin (Vemuri, 1994), actinin-4 was steadily translocated into the nucleus (Fig. 10). A bladder cancer cell line, KU7, which is deficient in PI3 kinase expression, also showed nuclear localization of actinin-4 without treatment (data not shown). These findings suggest that inhibition of the PI3 kinase–mediated signaling pathway is involved in the nuclear translocation of actinin-4.

Cytochalasin D inhibits actin polymerization (Casella et al., 1981). Treatment with cytochalasin D also induced nuclear translocation (data not shown), suggesting that the nuclear translocation of actinin-4 is also caused by loss of its association with the cytoplasmic actin cytoskeleton.

The distribution of actinin-4 was determined immunohistochemically in human tissue using the mAb, NCC-Lu-632. Actinin-4 was expressed in a limited population of normal cells, including erythrocytes, endothelial cells, and epithelial cells in various tissues at their border with stromal connective tissue.

In the ducts of normal mammary glands, the epithelio– stromal and epithelio–myoepithelial borders were stained, but the nucleus of normal mammary gland cells did not stain (Fig. 11,A). Scirrhous carcinoma and invasive lobular carcinoma of the breast manifest highly infiltrative growth into the fibrous stroma, in contrast to low-infiltrative histological types including papillotubular and solid-tubular carcinomas (Rosen and Oberman, 1992). To determine the relationship between the growth pattern of breast cancer and the subcellular localization of actinin-4, 83 cases of invasive breast cancer were examined immunohistochemically. The reactivity of mAb NCC-Lu-632 was classified into three types (Table I): nuclear (type A, Fig. 11,B), combined (type B), and cytoplasmic (type C, Fig. 10, C and D). Histologically, all type A cases were papillotubular or solid-tubular carcinomas that showed low-infiltrative extension (20/20). In type B, the majority of cases were papillotubular and solid tubular carcinomas (25/33) and fewer were scirrhous carcinomas (8/33). In contrast to these two types, most type C cases were scirrhous (15/30) (Fig. 11,C) and invasive lobular carcinomas (6/30) (Fig. 11 D), both of which extended locally in a highly infiltrative manner.

We then determined whether these staining patterns were correlated with prognosis in patients with breast cancer. The relationship between the actinin-4 staining pattern and disease-free and overall survival of 61 patients with relatively early breast cancer (clinical stages I and II [Rosen and Oberman, 1992]) was examined. A significant difference in disease-free survival was observed between types A+B and type C (P < 0.01) (Fig. 12,A). In addition, a significant difference in overall survival was observed between types A + B and type C (P < 0.01) (Fig. 12 B). These findings suggest that the subcellular distribution of actinin-4 may indicate the biological behavior of breast cancer and may be considered a new risk factor for relapse.

α-Actinin is a member of the superfamily of actin-binding proteins that includes spectrin, filamin, dystrophin, and fimbrin, by virtue of the similarity of their actin-binding domains (Matsudaira, 1991). α-Actinin binds to the integrin β1 cytoplasmic domain (Otey et al., 1990, 1993) and is believed to link the actin cytoskeleton to focal adhesions in cooperation with other cytoplasmic proteins, including tensin, talin, and vinculin (Burridge et al., 1990; Miyamoto et al., 1995). Actinin is also localized at adherens junctions in connection with α-catenin (Knudsen et al., 1995). Assembly of these structural proteins is believed to play an important role in stabilizing cell adhesion (Burridge et al., 1990; Miyamoto et al., 1995) and suppressing cell motility (Glück et al., 1993; Glück and Ben-Ze'ev, 1994).

In this study, using a newly established mAb, we isolated a cDNA clone encoding a novel isoform of α-actinin, designated actinin-4. Immunofluorescence microscopy demonstrated that the classical isoform, actinin-1, was concentrated in actin stress fiber ends and adherens junctions. In contrast, this novel nonmuscle α-actinin, actinin-4, binds actin filaments at a different subcellular location, thereby exerting functions distinct from those of actinin-1. Actinin-4 is highly concentrated in sharply stretched cells. In the wound assay, the cells forced to be motile along wound edges and cells migrating into the wound overexpressed actinin-4. These findings suggest the involvement of actinin-4 in cell movement. Glück et al. demonstrated that the expression of actinin-1 was reduced in SV-40–transformed 3T3 cells (Glück et al., 1993; Glück and Ben-Ze'ev, 1994). Transfection with actinin-1 cDNA suppressed tumorigenicity. Transfection with actinin-1 cDNA in an antisense orientation increased cell motility, supporting the contradictory roles of the two nonmuscle isoforms with respect to cell movement.

Various growth factors provide stimuli for cell motility in addition to mitogenic activity (Klagsbrun, 1996). Recently, Hsu et al. (1996) identified a new murine nonmuscle α-actinin isoform, FR-17, as a fibroblast growth factor-1–inducible gene in NIH-3T3 cells. Although it has not been fully sequenced, the deduced COOH-terminal amino acids (1–109; sequence data available from GenBank/ EMBL/DDBJ under accession number U41415/6) available for FR-17 showed 97.2% identity with amino acids 776–884 of human actinin-4, suggesting that this isoform may be a murine homologue of human actinin-4. In addition to FR-17, a computer search revealed that several partial sequences identical to actinin-4 were registered in dbEST databases. These ESTs were isolated from human cDNAs of placenta (accession number AA368907), fetal brain (M85377), Jurkat T-cells (AA312012), and pancreatic islets (C06273, D83843).

The level of another actin-binding protein, thymosin β15, was reported to be elevated in human metastatic prostate cancer and correlated positively with the Gleason score (Bao et al., 1996). In contrast to actinins, thymosins inhibit actin polymerization. Cell motility is regulated by both disassembly and reformation of the actin cytoskeleton. Thymosin β15 has been considered to degrade the static anchorage of the actin cytoskeleton and liberate actin monomers. Actinin-4 may then cross-link actin filaments and reorganize the cytoskeleton, which is essential for cell movement. Further studies are necessary to elucidate how these molecules regulate the actin cytoskeleton and participate in cancer invasion and metastasis.

In this study, we found that actinin-4 existed in the nucleus of a certain population of breast cancers and in several cell lines. From the deduced amino acid sequence, it seems that actinin-4 does not possess any apparent nuclear localization signals (Boulikas, 1993). Although the precise mechanisms of the nuclear transition are still under investigation, we were able to reproduce the nuclear transition of actinin-4 in tissue culture. Actinin-4 was translocated from the cytoplasm to the nucleus by treatment with the PI3 kinase inhibitor, wortmannin (Vemuri, 1994), or by actin depolymerization using cytochalasin D (Casella et al., 1981). PI3 kinase is thought to be one of the key molecules involved in the signaling pathways of activated receptor tyrosine kinases and regulates cell growth, motility, and morphogenesis (Raffioni and Bradshaw, 1992; Royal and Park, 1995). The nuclear localization of actinin-4 was observed mainly in low-infiltrative histological subtypes of breast carcinoma. The abnormal activation of this enzyme may determine the biological behavior of cancer cells and confer metastatic potential to them.

Although histological subtyping is a common procedure in the pathological diagnosis of breast cancer, the correlation between histological subtypes and the outcome of patients in this study did not reach statistical significance (data not shown). However, we demonstrated a significantly poorer disease-free survival and overall survival in breast cancer patients with a nonnuclear localization of actinin-4 (Fig. 12). This difference was probably due to cases of papillotubular and solid-tubular carcinomas, in which the location of actinin-4 was exceptionally cytoplasmic. These cases may have had a high degree of metastatic potential that could not be detected by conventional histological examination.

From the above experimental and clinical observations, we conclude that a cytoplasmic localization may indicate active actin-bundling by actinin-4 and enhanced cell motility. Because no definite marker is currently available for assessment of the metastatic potential of cancer, the subcellular localization of the actinin-4 molecule may represent a potential new biological marker for cancer metastasis and may predict early relapse of breast cancer. Large-scale prospective analyses are necessary to examine this issue, not only for breast cancer but also for other cancers.

GST

glutathione S-transferase

HFK

human foreskin keratinocyte(s)

PH

pleckstrin homology

PI3

phosphatidylinositol 3

PIP2

phosphatidylinositol 4,5-biphosphate