S100A4, a Mediator of Metastasis*

The transition from benign tumor growth to malignancy is manifested by the ability of tumor cells to traverse tissue barriers and invade surrounding tissues. The traversal of basement membrane barriers is thought to be the critical event in the initiation of themetastatic cascade and, as described by the three-step hypothesis of metastasis, involves attachment to the extracellular matrix, local proteolysis, and subsequentmigration (1). In recent years, a number of specific genes have been identified, the expressions of which correlate with tumorigenesis and/or metastatic potential. S100A4 is a member of the S100 family of calcium-binding proteins that is directly involved in tumor metastasis. This report reviews the animal and cellular studies implicating S100A4 in the establishment of the metastatic phenotype and examines the biochemical properties of S100A4 that are relevant to its role as a metastasis factor.

The transition from benign tumor growth to malignancy is manifested by the ability of tumor cells to traverse tissue barriers and invade surrounding tissues. The traversal of basement membrane barriers is thought to be the critical event in the initiation of the metastatic cascade and, as described by the three-step hypothesis of metastasis, involves attachment to the extracellular matrix, local proteolysis, and subsequent migration (1). In recent years, a number of specific genes have been identified, the expressions of which correlate with tumorigenesis and/or metastatic potential. S100A4 is a member of the S100 family of calcium-binding proteins that is directly involved in tumor metastasis. This report reviews the animal and cellular studies implicating S100A4 in the establishment of the metastatic phenotype and examines the biochemical properties of S100A4 that are relevant to its role as a metastasis factor.

S100 Protein Family
In humans there are now over 20 S100 family members that are distributed tissue specifically (2)(3)(4). The first family members were discovered in the brain by Moore (5) and given the name S100 because they are soluble in 100% saturated ammonium sulfate (5). As for most S100 family members, S100A4 is a symmetric homodimer stabilized by noncovalent interactions between two helices from each subunit (helices 1, 4, 1Ј, 4Ј) that form an X-type four-helix bundle (Fig. 1). Each S100A4 subunit has two EF-hand Ca 2ϩ -binding domains: a pseudo-EF-hand or S100-hand and a typical EF-hand ( Ca K D ϭ 2.6 M) (6), which are brought into close proximity by a small two-stranded antiparallel ␤-sheet (7). The pseudo-EF-hand differs from prototypical EF-hands because it is composed of 14 residues (rather than 12) and coordinates calcium weakly via backbone carbonyl oxygen atoms rather than with side chain oxygen atoms of Asp, Glu, Asn, or Gln (8 -10). For S100A4, the pseudo-EF-hand is formed by helices 1 and 2 and is similar in conformation to other members of the S100 family. The orientation of helices 3 and 4 in the S100A4 typical EF-hand is similar to only one other member of the S100 protein family, S100A6 (11), and less like other S100 family members for which three-dimensional structures are available in the calcium-free state (12,13). The difference in the position of helix 3 in the apo-state of these S100 proteins is due to variations in the amino acid sequence in the C terminus of helix 4 and in loop 2 (the hinge region) and is used to subclassify the S100 protein family (7). For S100A4, there is a C-terminal loop following helix 4 that is quite long and very basic, which makes it particularly unique when compared with other S100 proteins. In general, the "hinge region" and the C-terminal loop of S100 proteins are involved in target protein binding; therefore, the lack of sequence homology in these regions provides a means for S100 proteins to interact with specific binding partners (14). S100A4 and most other dimeric S100 proteins undergo a relatively large conformational change in the typical EF-hand upon binding calcium, which exposes a hydrophobic binding pocket comprising residues in helices 3 and 4, the hinge region, and the C-terminal loop. This calcium-dependent conformational change is necessary for S100A4 to interact with its protein targets and generate a biological effect (11,(15)(16)(17)(18). Similar to other S100 proteins (19), target binding significantly enhances the calcium binding affinity (K D ϳ 0.2 M) of S100A4 (6). This provides a mechanism for the formation of a high affinity complex in vivo that occurs only when calcium, S100A4, and one of its target proteins are present (Fig. 1).

S100A4 Discovery, Expression, and the Metastatic Phenotype
The S100A4 gene was cloned independently by several groups under various names, including metastasin (Mts1), fibroblast-specific protein (FSP1), 18A2, pEL98, p9Ka, 42A, CAPL, and calvasculin (20 -27). Initial cloning efforts identified S100A4 as a highly expressed transcript in growth-stimulated cultured cells (23,28) and metastatic tumor cell lines (20) and during morphogenic conversion from an epithelial to mesenchymal phenotype (21). In addition, S100A4 expression levels are up-regulated during oncogenic transformation (29,30). Northern blot analyses indicate that S100A4 is expressed in several adult mouse and rat tissues, with high expression found in the spleen, thymus, bone marrow, T-lymphocytes, neutrophils, and macrophages (31,32). Immunohistochemical analysis of rat tissues shows that S100A4 is expressed in some absorptive and keratinized epithelia, in the parietal cells of the stomach, and in a subset of cells of the immune system in the blood, bone marrow, spleen, and lymph nodes (33). Similar studies examining S100A4 expression during mouse development demonstrate that S100A4 is highly expressed in embryonic macrophages, as well as in differentiating mesenchymal tissues (34). S100A4 expression is also observed in metastatic cancers. Immunohistochemical analyses of human cancers shows significant S100A4 expression in breast, pancreatic, prostate, gallbladder, esophageal, gastric, lung, and thyroid carcinomas (35)(36)(37)(38)(39)(40)(41)(42). In these cancers S100A4 expression is found at elevated levels compared with normal tissue, suggesting that enhanced S100A4 expression contributes to manifestation of a metastatic phenotype.
Studies in rodents have provided evidence supporting the direct involvement of S100A4 in tumor progression and metastasis. The role of S100A4 in cancer has been examined most widely in breast cancer models, which have demonstrated that overexpression of S100A4 in nonmetastatic mammary tumor cells confers a metastatic phenotype (43,44). Transgenic mice that overexpress S100A4 in the mammary epithelium are phenotypically indistinguishable from wild-type mice (45), demonstrating that S100A4 itself is not tumorigenic; however, transgenic mouse models of breast cancer have shown that S100A4 expression correlates with metastasis. MMTV-neu and GRS/A animals are characterized by a high incidence of mammary tumors that rarely metastasize; overexpression of S100A4 in the mammary epithelium of these animals causes more invasive primary tumors and the appearance of metastases in the lungs (45,46). The link between S100A4 and metastasis is further supported by knockdown experiments, as inhibition of S100A4 expression by antisense or anti-ribozyme techniques suppresses the metastatic capacity of S100A4-expressing tumor cells in animal models of lung carcinoma and osteosarcoma (47,48). In addition, recent studies demonstrate that when transgenic mice expressing the polyoma virus middle T antigen are crossed with mice carrying null alleles for S100A4, they display a significant decrease in lung metastases (49). Altogether these observations suggest that S100A4 is not simply a marker for metastatic disease but rather has a causal role in mediating this process.
The association between S100A4 expression and metastasis observed in animal studies has led to a number of studies examining the utility of S100A4 expression as a prognostic marker in human cancers. In a retrospective study of 349 invasive human breast cancer specimens, S100A4 expression and other variables were evaluated for their prognostic significance over a period of 14 -20 years (50,51). Analysis of patients with carcinomas that stain positively for S100A4 expression demonstrated that S100A4 expression is highly correlated with patient death. In addition to breast cancer, S100A4 has been shown to be a prognostic marker in a number of human cancers, including esophageal-squamous cancers (39), non-small lung cancers (41), primary gastric cancers (40), malignant melanomas (52), prostate cancers (37,53), bladder cancers (54), and pancreatic carcinomas (36,55). The universality of S100A4 expression in a variety of cancers illustrates the potential use of S100A4 as a marker for tumor metastasis and disease progression.

Gene Structure and Regulation
The majority of the human, mouse, and rat S100 genes, including S100A4, are located as a gene cluster on chromosomes 1q21, 3f3, and 2q34, respectively (56). The mouse S100A4 gene consists of three exons and two introns (57). The first exon is noncoding and contains the 5Ј-UTR 3 of the gene, the second exon contains the start codon and encodes the N-terminal EF-hand, and the third exon encodes the C-terminal EF-hand. The human S100A4 gene contains 4 exons, and the additional exon is located within the 5Ј-UTR and is noncoding. In addition, there are two variants of the human S100A4 as a result of alternative splicing within the 5Ј-UTR (58). The two splice variants display different expression profiles in human tissues and tumor cell lines (57); however, the significance of this observation with respect to gene or protein activity is not known. The S100A4 promoter contains an ErbB2 response element (59), and transcription of the S100A4 gene is controlled by both positive and negative regulatory elements located within the first intron, which bind several transcription factors (60 -63). Furthermore, hypermethylation of the S100A4 gene is associated with transcriptional silencing (64 -66).
Genetic deletion of S100A4 (S100A4Ϫ/Ϫ mice) has no overt phenotype in the postnatal period but results in a low incidence of spontaneous tumors late in life. The gender ratio of the knock-out mice is affected also, with fewer Ϫ/Ϫ females being born than Ϫ/Ϫ males (67). In addition, recent studies show that following orthotopic injection of a highly metastatic, S100A4-positive, carcinoma cell line into S100A4Ϫ/Ϫ mice, tumor development and the formation metastases are suppressed (68). These observations suggest that S100A4 expression in host-derived stroma may also contribute to tumor progression and metastasis.

Intracellular Roles and Targets
Several proteins have been identified as S100A4 targets, including liprin ␤1 (69), methionine aminopeptidase (70), the p53 tumor suppressor protein (71), and proteins involved in cytoskeletal rearrangement and cell motility such as F-actin, tropomyosin, and the heavy chain of nonmuscle myosin II (72)(73)(74). In all of these cases, the interaction of these protein targets with S100A4 is calciumdependent and thus links the cellular functions of these proteins to changes in the intracellular calcium concentration (Fig. 1). It has been a general observation that two-hybrid screens utilizing S100 proteins as bait have detected primarily other S100 family members as targets (75)(76)(77)(78). Heterodimerization is not calcium-dependent, as was confirmed in vitro for several S100 pairs (13,76), and likely results from the low dissociation constant for dimerization (Յpicomolar) for the S100 proteins (79). The S100A1-S100A4 complex, which was identified using two-hybrid screening methods (75,76), is not readily detected with in vitro mixtures of S100A4 and S100A1 (75) casting some doubt about the relevance of this interaction in vivo.
Several S100 proteins, including S100A4, regulate the p53 tumor suppressor protein. Similar to S100B, S100A4 interacts with the C terminus of p53 and inhibits protein kinase C (PKC) phosphorylation of the tumor suppressor in vitro (71,80,81). Likewise, the interaction between p53 and either S100B or S100A4 inhibits p53 from binding to its consensus DNA-binding sequence (71,82); thus it was expected that S100A4 would be a general inhibitor of p53 function. However, an examination of p53-regulated genes in S100A4-expressing cells indicates that the expression of several genes are up-regulated (e.g. bax); others genes are down-regulated initially and then later up-regulated (e.g. mdm2), and some genes are inhibited (e.g. p21, thrombospondin-1) (71). Several of these p53-responsive genes are pro-survival, whereas others are pro-apoptotic. These seemingly opposite effects of S100A4 expression on p53regulated genes could result from differences in the cell lines used in the various studies. Notwithstanding the potential interaction between S100A4 and p53, S100A4 expression consistently promotes the acquisition of an invasive phenotype. This is likely not due to the interactions of S100A4 with p53 but is likely the consequence of direct interactions with other targets such as cytoskeletal proteins. S100A4 selectively binds to the A isoform of nonmuscle myosin II (myosin IIA) and promotes the monomeric, unassembled state (83,84). The S100A4 binding site maps to residues 1909 -1924 in the C-terminal end of the coiledcoil of the myosin IIA heavy chain (84,85). Recent studies demonstrate that although the S100A4 binding site overlaps a PKC phosphorylation site at Ser-1917, S100A4 binding is not affected by PKC phosphorylation. However, phosphorylation on the CK2 site at Ser-1944, which is located downstream of the S100A4 binding site in the tail piece, inhibits S100A4 binding (86). Moreover, CK2 phosphorylation protects against S100A4-induced inhibition of filament assembly and S100A4-induced disassembly of myosin IIA filaments. These findings demonstrate that heavy chain phosphorylation at the CK2 site provides, in addition to calcium binding, another regulatory mechanism for the S100A4-myosin IIA interaction. With respect to its effects on cell locomotion, S100A4 modulates cellular motility by effecting the orientation and localization of cellular protrusions. S100A4-expressing cells display few side protrusions and extensive forward protrusions during chemotaxis as compared with cells that do not express S100A4. 4 Moreover, these effects are due to the specific interaction of S100A4 with myosin IIA. Altogether, these data suggest that S100A4 may modulate the metastatic phenotype through the regulation of cellular motility and myosin IIA function.

Extracellular Roles and Targets
There is accumulating evidence that S100 proteins have extracellular functions; however, their mechanism of secretion remains unknown. S100A4 is secreted constitutively by fibroblasts in culture, and co-culture of tumor cells and fibroblasts stimulates the release of S100A4 by fibroblasts (27,87). In FIGURE 1. The calcium-dependent interaction of S100A4 with protein targets. Ribbon diagrams of the NMR solution structure of apo-S100A4 (left) and a homology model of calcium-bound S100A4 (right) illustrating the movement of helix 3 (shown in orange) upon the addition of calcium (red spheres) as described by Vallely et al. (7) are shown. This conformational change is required for S100A4 binding to protein targets such as myosin IIA, p53, and annexin II. addition, in S100A4 transgenic mice, S100A4 accumulates in the blood of aging animals (88). Extracellular S100A4 stimulates MMP-13 activity and acts as a moderate prometastatic factor of tumor cells (87). In osteosarcoma cells, inhibition of S100A4 mRNA expression significantly reduces MMP-2, MT1-MMP, and TIMP-1 mRNA expression (89), and oncogenic transformation of Madin-Darby canine kidney cells increases MT1-MMP and S100A4 expression (90), suggesting that the coordinated action of S100A4 and matrix metalloproteinases contributes to the invasive phenotype. Recently, S100A4 has been shown to bind annexin II, an endothelial plasminogen co-receptor. This interaction accelerates plasminogen activation by tissue plasminogen activator on the endothelial cell surface to generate plasmin, which is known to activate pro-MMPs (91). Thus S100A4 may also promote metastasis by inducing remodeling of the extracellular matrix and as a result facilitate angiogenesis and tumor invasion.

Perspectives
Although S100A4 is established as a regulator of metastasis, the mechanisms by which it promotes an invasive phenotype are not well understood. Biochemical and cell-based studies point to the involvement of S100A4 in at least two steps in the metastatic cascade: remodeling of the extracellular matrix and cellular motility. However, the details of how S100A4 regulates its protein targets and how these interactions influence cellular functions remains to be determined. Identification and characterization of the molecular mechanisms by which S100A4 regulates its effectors will provide the biochemical foundation for understanding the contribution of S100A4 to normal and metastatic processes and will provide important new insights into the molecular basis for the invasive behavior of tumor cells. This mechanistic information can be applied to the use of S100A4 as a diagnostic marker and as a target for novel therapies.