Tumor metastasis formation: cell-surface proteins confer metastasis-promoting or -suppressing properties

BB

ELSEVIER

Biochimica et Biophysica Acta 1198 (1994) 1-10

Biochi PmPic~aA~ta et Biophysica

Tumor metastasis formation: cell-surface proteins confer metastasis-promoting or -suppressing properties H e l m u t P o n t a *, J o n a t h a n S l e e m a n , P e t e r H e r r l i c h Kernforschungszentrum Karlsruhe, Institut fiir Genetik, Postfach 3640, D-76021 Karlsruhe, Germany (Received 12 October 1993)

Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

2. Immune surveillance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2

3. Loss of cell-cell adhesion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2

4. Invasion into the host transport system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4

5. Contacts during migration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5

6. Lymph node metastases and variants of CD44 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5

7. Colonization of distant organs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

7

8. Putative mechanisms for CD44v protein function in metastasis induction . . . . . . . . . . . . . . . . . . . .

7

9. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

8

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

8

1. Introduction

Cancer patients die of distant metastases, a fact amply documented by the poor 5-year survival data of patients whose primary tumor has been surgically removed. Early removal enhances the chance of survival, suggesting that there are early stages when tumor cells have either not yet had the opportunity or number to spread, or have not developed the properties needed for metastasis formation. What makes tumors metasta-

* Corresponding author. Fax: + 49 7247 823354.

0304-419X/94/$26.00 © 1994 Elsevier SSD! 0 3 0 4 - 4 1 9 X ( 9 3 ) E 0 0 2 1 - E

Science B.V. All rights reserved

size? Despite an enormous wealth of information on tumor cell properties, this key question remains to be answered. Most of our progress in understanding molecular aspects of cancer has been made by studying the stages involved in cellular transformation, the process that converts a normal ceil into a cell which under conditions of sufficient density can grow as a tumor in an isogenic host. This process is driven by dominant oncogenes and the loss of recessive tumor suppressor genes, and essentially involves the alteration of cell cycle control (reviewed in Refs. 1-5). The ceils become growth-factor-independent, do not obey contact inhibition and lose features of differentiaton.

2

H. Ponta et al./Biochimica et Biophysica Acta 1198 (1994) 1-10

The development of invasive and metastatic properties is probably not completely independent from cell transformation. Transformation is, of course, a precondition for metastatic spreading. At the same time, transformed cells appear to develop properties which are also required for metastasis formation. It has been known for a long time that transformed cells look phenotypically different as compared to their normal counterparts. Fibroblasts, for example, round up, are less adherent to neighboring cells, and change their cytoskeletal organization [6-9]. Furthermore, transformed cells can reorganize the ECM. In particular, they produce elevated levels of various proteases including metalloproteases which are currently thought to endow the ceils with invasive properties (reviewed in Refs. 10,11). Thus, while cell cycle control appears to be the primary target of carcinogenesis, transformation may overlap with other stages of tumor progression. In order to form distant metastatic colonies, cancer cells have to overcome many obstacles. These include combatting immune surveillance, loss of original tissue contacts, moving through the ECM (e.g., the basement membrane), entering the lymphatic system, expanding in lymph nodes, entering the blood stream, extravasating and settling in other tissues. Tumor cells thus need to acquire a wide range of molecular properties to complete the whole metastatic process, most of which are likely to be mediated by cell-surface proteins. In order to define the molecular properties a tumor cell requires to metastasize, it is important to answer several questions. For example, no proof exists for most of the stages in the metastatic process listed above that they are indeed hinderances to metastasis formation. Which of these stages really are limiting? To what extent are the surface properties required for the formation of metastases already established during transformation, and which others are required in addition to the properties of the locally growing tumor? Can the complexity of new molecular properties which need to be accumulated be explained by the triggering of key control elements like a dominant 'metastogene', or is the hypothesis of complexity simply wrong? In this review we will discuss the few surface molecules whose causal contribution to the formation of distant metastases has been proven or strongly supported. We argue that there are a few rate-limiting cellular properties involved in the progression from local cancer to metastatic disease.

2. Immune surveillance

When tumor cells express surface antigens which are recognized as abnormal by the host's immune surveillance, then it is possible for the tumor cells to be eliminated. Naturally arising tumors are often not rec-

ognized by T cells as being foreign because they express no such aberrant surface antigens, and therefore escape from immune surveillance. Viral transformation, however, leads to the exposure of viral gene products which are recognized by cytotoxic T cells and subsequently results in destruction and elimination of the tumor cells. The adenovirus strain 12 has developed a strategy that circumvents host immune responses and leads to highly malignant transformation, in which the expression of the major histocompatibility antigen (MHC) is changed or switched off by the virus. The MHC proteins are required to present antigens to the immune system, especially to cytotoxic T cells, and only in this context are the antigens targets for immunological defense mechanisms. Elimination of expression of the MHC genes therefore allows 'hiding' of the viral gene product [12,13]. Since other less oncogenic adenoviral strains fail to switch off MHC class I expression, and the ectopic expression of MHC class I gene in ceils infected with the strongly metastasis-inducing strain 12 represses the metastatic potential of the virus [14], a crucial role of the MHC class I genes for the immune surveillance of tumor progression in this particular system has been established. Interestingly, MHC class I expression is also inversely related to metastatic capability in other tumor systems, for example, in a spontaneously arising murine lung carcinoma from which cell types differing in oncogenic potential have been isolated [15]. Ectopic expression of MHC class I genes in the metastatic lung carcinoma line interfered with tumorigenicity and metastatic competence, suggesting a crucial role for MHC class I genes in these cells [16]. MHC class I expression had a similar effect on the metastatic properties of two other tumor cell lines [17,18]. In some naturally occurring tumors, interference with expression of MHC class I genes may also play a decisive role in determining their metastatic potential. The fact that at least in some systems the recruitment of the immune system by expression of MHC class I genes effectively interferes with metastatic spread has stimulated attempts to recruit immunoprevention as a general tool to combat metastatic growth [19].

3. Loss of cell-cell adhesion

The first step in the 'metastatic cascade' is dissemination of tumor cells from the primary tumor mass. The physiological interaction of normal cells with their neighbors is interrupted. This process resembles the release of certain cells from their context during embryogenesis. Loss of function of members of the cadherin family has been correlated with the dissemination of both embryonic cells and metastasizing tumor cells [20,21]. Vinculin and members of the N-CAM

H. Ponta et al. / Biochimica et Biophysica Acta 1198 (1994) 1-10

family are other proteins which can change the adhesive properties of tumor cells [22-24]. E-cadherin (uvomorulin) is a surface glycoprotein with an apparent molecular weight of 120 kDa, and is specifically expressed on epithelial cells where it is an integral component of adhesion plaques. It belongs to a family of closely related cadherin proteins, all of which are expressed in a highly tissue-specific manner and are thought to mediate homotypic interaction of cells (reviewed in Ref. 25). The expression of Ecadherin was found to be inversely proportional to the extent of differentiation and to the degree of malignancy of carcinoma cells [26]. First indications for a causal relationship between the loss of E-cadherin function and the dissemination of tumor cells were derived from experiments in which antibodies directed against E-cadherin increased the invasiveness of otherwise non-invasive epithelial ceils [27]. In contrast, expression of E-cadherin upon transfection of the appropriate cDNA into highly invasive, dedifferentiated carcinoma cells greatly reduced invasiveness [26,28,29]. The effect of over-expression of E-cadherin and conversely the block of its functional expression by antibodies or antisense RNA [28] has been predominantly explored in invasion assays. There has been only one study on the in vivo growth properties of tumor cells ectopically expressing the E-cadherin protein [30]. Although there is a striking correlation between loss of E-cadherin expression and increased metastatic capability, direct proof for E-cadherin as a metastasis suppressor gene is still lacking. Importantly, the adhesive function of E-cadherin can be modulated from within the cell. The cadherins are associated intracellularly with proteins called catenins [31,32]. This association is strictly required for the adhesive function of the cadherins [33], and can be regulated by phosphorylation. Thus, not only can the function of cadherins be abolished by down-regulation of the expression of the molecule or by mutational changes in the protein, but also by interference with its association with the catenins or by suppression of catenin expression. Whereas cell-cell junctions are composed of one type of cadherin, at least three different catenins are constituents, namely the a, fl and 3/catenins (reviewed in Refs. 34,35). Functional interference with adhesion properties by phosphorylation has been described for/3 catenin [36,37]. Thus, the complex organization of adhesion plaques with extracellular and intracellular components permits regulation of E-cadherin mediated adhesion on several levels. The homology of a catenin with vinculin (see below), and indirect evidence for the association of the cadherin-catenin complex with the actin network [32] has been used to postulate that there is a regulated association of the cadherins with the cytoskeleton via the catenins, which is abrogated by

3

) eg. SRC

CYTOSKELETON

Fig. 1. Regulation of cadherin-mediated cellular adhesion. The catenin complex binds to the intracellular portion of the cadherin molecules and to the cytoskeleton, allowing cadherin molecules to interact intercellularly. Upon phosphorylation, the cadherin-catenin interaction is dissociated, and intercellular adhesion is abrogated.

the dissociation of the cadherin-catenin interaction (see Fig. 1). For E-cadherin it is plausible that the loss of its function is required for tumor cell migration and metastatic spread. For P-cadherin such evidence is rather circumstantial, and is based on functional interference by phosphorylation through v-src [37]. For other tissue-specific cadherins, similar properties may well exist. Data on vinculin support the idea that loss of adhesion is an important step in the metastatic process. Vinculin is an intracellular component of adhesion plaques, and plays an important role in the assembly and stabilization of these junctions and the associated cytoskeleton [38,39]. It is also a target for oncogenedriven modification but the functional relevance of this modification has to be proven [40]. A causal role for vinculin expression in the suppression of metastasis formation has been shown by experiments with a highly metastatic tumor cell line derived from a rat pancreas carcinoma which expresses virtually no vinculin compared to a related, but only locally growing tumor line established from the same tumor. Expression of vinculin cDNA in this highly metastatic cell line drastically repressed metastasis formation in the spontaneous metastasis assay [41]. These cells also showed increased cell adhesion in culture, suggesting that their decreased metastatic potential may be due to increased adhesiveness mediated by vinculin. Members of the N-CAM family are other molecules implicated in cell-cell adhesion, and whose function appears to be abrogated in metastasizing tumor cells [42]. One of the genes encoding a member of the

H. Ponta et al. / Biochimica et Biophysica Acta 1198 (1994) 1-10

N-CAM family, DCC (deleted in colon carcinomas), is a recessive gene involved in colon tumor progression [43] and its expression is abolished in many colon carcinomas. The loss of DCC may result in phenotypic changes similar to those seen after the loss of Ecadherin expression. The function of other N-CAM proteins in cell-cell interactions appears to be modulated by sialylation [23,44]. As yet, only correlative evidence exists for a link between loss of function of N-CAM-like molecules and tumor cell migration. Taken together, the reported changes in the functions of cadherin, vinculin and N-CAM proteins suggest a crucial role for these molecules in the first steps of the metastatic process. These changes may result as an immediate consequence of oncogenic transformation (e.g., phosphorylation by src).

4. Invasion into the host transport system

The majority of tumors are carcinomas. The major difference between locally growing and metastasizing carcinoma cells is the ability of the latter to penetrate basal membranes, migrate through the interstitium and to invade lymphatic and blood vessels. A multitude of functions are required for these processes, and three different groups of molecules have been implicated as being involved. These are enzymes which disrupt the extracellular matrix and create space for invasion, adhesion molecules which mediate affinity to specific ligands (e.g., presented by the extracellular matrix), and cell-surface and cytoskeletal proteins which mediate directional cell movement. The final aim of migrating tumor cells is to invade blood or lymphatic vessels. The crucial steps in these processes need to be defined. For example, it is important to determine whether invasive tumour cells specifically interact with components of the basement membrane or endothelial cells, or whether they simply mediate enzymatic degradation. Enzymes critical for these processes also need to be identified. Additionally, it remains to be shown whether the migration towards vessels is gradient-directed or random. Although the answers to these questions are obviously of great importance in identifying steps in metastasis formation, it is astonishing that there are only few data which define molecules as rate limiting and causally involved in metastatic spread, even though such parameters can often be easily measured by using in vitro assays. The majority of the present findings are correlative. This is true for the group of matrix-degrading enzymes, partly due to the complexity of their expression pattern and to the posttranslational control of their enzymatic activity (Fig. 2). Primary human and animal tumors frequently express high amounts of proteases. Examples of such proteases are the matrix-de-

EXTRACELLULAR STIMULATORY SIGNAL

TISSUE SPECIFIC TRANSCR/PTIONAL REGULATION

1

EXTRACELLULAR STIMULATORY SIGNAL

EXTRACELLULAR ~ T O R Y SIGNAL

TISSUE SPECIFIC TISSUE SPECIFIC TRANSCRIPTIONAL TRANSCRIPTIONAL REGULATION I~GULATION

EXTRACELLULAR STIMULATORY SIGNAL

TISSUE SPECIFIC TRANI~.EPTIONAL REGULATION

I I PI~MIN

INHIBITOR

1

1

ECM DEGRADATION Fig. 2. Regulation of metalloprotease activity. Extracellular signals determine the tissue-specific transcription of the metalloprotease genes, and the genes of proteins regulating metalloproteases. These components interact through a regulatory cascade to activate or inhibit each other, and ultimately control the level of metalloprotease activity (PA, plasminogen activator; MP, metalloprotease; TIMP, tissue-specific inhibitor of metalloproteases).

grading metalloproteases, which include collagenase IV (72 and 92 kDa forms), collagenase I and the stromelysins (reviewed in Refs. 11,45). Metalloproteases are synthesized as enzymatically inactive proenzymes. As shown in Fig. 2, activation of proenzymes requires a specific protease (plasmin), which in turn is strictly controlled in its activity by another protease (plasminogen activator). Therefore, although proteases are usually quantified by measuring RNA or protein levels, conclusions about their importance during metastatic spread can only be made if their enzyme activity is measured. In an attempt to elucidate whether collagenase suffices to confer metastatic behavior onto tumor cells, we transfected a collagenase type I cDNA into rat fibroblasts which had been transformed by ras (M. Giles, P. Herrlich and H. Ponta, unpublished results). The biological activity of the collagenase expressed in the transfected cells was measured in a collagen type I degradation assay. Upon subcutaneous injection into syngeneic animals the ras-transformed fibroblasts grew exclusively as primary tumors and no metastases developed. Even clones expressing high levels of collagenase type I were no more invasive than the parental cell line. Thus, at least in this system, collagenase type I expression is not sufficient to trigger a transformed cell line to metastasize. However, collagenase type I expression had a pronounced effect on another parameter of transformation, as all collagenase type I expressing clones showed a marked increase in vascularization and growth rate of the tumors. Such an involvement of collagenase in angiogenesis has also been suggested

H. Ponta et al. / Biochirnica et Biophysica Acta 1198 (1994) 1-10

from experiments in which endothelial cells were induced to synthesize collagenase by basic fibroblast growth-factor treatment, and thereby acquired the ability to migrate into human amniotic membrane [46]. Collagenase type IV (gelatinase A, 72 kDa form) can affect the metastatic potential of tumor cells. Selected animal tumors which do not express gelatinase A and poorly colonize the lungs upon i.v. injection can be converted into effective colonizing cells after transfection with a gelatinase A expression construct (M. Cockett, A. Docherty and G. Murphy, personal communication). Transfection of other metalloproteases had no effect on the lung colonization potential in this system. However, it is unlikely that many tumors lack gelatinase as the limiting component. As will be discussed below,/32 integrins or CD44 are decisive properties in other tumors for facilitating lung colonization. After activation of proenzymes by proteolysis, the proteolytic activity of proteases is kept under tight control. Antagonists of metalloproteases, the tissuespecific inhibitors of metalloprotease (TIMP), inhibit enzyme activity (see Fig. 2). When recombinant TIMP protein was mixed with tumor cells and injected intravenously into syngeneic mice, lung colonization was strongly reduced [47-50], suggesting that TIMP had an anti metastatic effect. A similar conclusion was drawn from experiments in which the expression of TIMP in an immortalized cell line was down-regulated by antisense TIMP mRNA expression [51]. Suppression of TIMP not only led to increased invasiveness of cells as measured by an in vitro invasion assay, but also provided the cells with oncogenic and even metastatic capability [51]. Unfortunately, no further investigations in this direction have since been published.

5. Contacts during migration Our current knowledge about adhesion molecules which play a causal role in metastasis formation is poor. Since the appearance of metastases is a late event after orthotypic implantation or s.c. injection of tumor ceils, it is difficult to distinguish at which point a surface molecule is required, whether prior to vessel entry, during extravasation or for distant organ colonization. Candidates for surface molecules which are key components for metastatic spread are the integrins and the splice variants of CD44. The integrins are surface receptors composed of two different subunits, the a and/3 chains. So far, fourteen different a chains and eight different 13 chains have been detected (e.g., reviewed in Ref. 52). The combination of individual a and /3 subunits into complexes allows a variety of different receptor structures to be generated. The majority of integrins allow cells to interact with the ECM. They have affinity to matrix

5

structures such as collagen, laminin, fibronectin and vitronectin. Interaction with the specific amino-acid sequence Arg-Gly-Asp (RGD) that is present in many extracellular matrix components has also been observed. Additionally, certain integrins also have affinity for cell-bound receptors on leucocytes and endothelial cells. An involvement in metastasis was suggested by the observation that some integrins (intensively studied examples are LFA-1 and the vitronectin receptor) are expressed at low or undetectable levels on tumor ceils with low metastatic potential. However, more direct evidence for the involvement of integrins was derived from experiments in which their function or interaction with a ligand was blocked by the peptide RGD. This peptide not only blocked in vitro invasion [53], but also inhibited experimental metastasis formation [54,55]. Further evidence for the involvement of an integrin as a limiting factor in the metastatic process in at least one tumor cell line has been derived from transfection experiments. When a cloned integrin subunit ot2 was expressed in a human rhabdomyosarcoma cell to allow the assembly of a functional VLA-2 receptor (composed of a2fl 1 subunits), the VLA-2-positive cells formed substantially more metastases in nude mice after both intravenous and subcutaneous injection [56].

6. Lymph node metastases and variants of CD44 Most human cancers spread lymphogenically. The establishment of tumor colonies in draining lymphatic tissue appears to be obligatory, since many tumors first produce lymph node metastases, despite their ability to colonize different organs including the lungs. We have recently characterized variants of the surface glycoprotein CD44, the expression of which can be sufficient to convert otherwise non-metastasizing tumor cells into metastasizing ones. These proteins probably act within the lymph node to allow expansion and spreading of the metastasizing tumor cells. We will summarize the arguments for these conclusions, together with a brief characterization of the proteins. The family of CD44 proteins is encoded by a gene that is composed of at least nineteen exons which are 5'

-- -

vI

v2

v3

v4

v5

v6

v7

v8

vCJ

I

-

-

3'

Fig. 3. Schematic diagram of the variable exon genomic structure of the CD44 gene. Exons are indicated by boxes, lines between them indicate introns. The closed boxes refer to flanking exons commonly expressed in CD44 isoforms. T h e open boxes (vl to vl0) refer to variant exons, which are completely removed in the CD44 standard form and are expressed in various combinations in CD44 variants. In contrast to mouse and rat, the exon vl is non-functional in human. In h u m a n , the variant exons are flanked on each side by five commonly expressed exons. O n the 3' side the third c o m m o n exon encodes the t r a n s m e m b r a n e sequence [57]. In the m o u s e and rat the common exon structure has not been resolved.

6

H. Ponta et al./Biochimica et Biophysica Acta 1198 (1994) 1-10

spread over a distance of more than 50 kb of DNA [57]. The abundantly expressed small CD44 protein (CD44 standard form or CD44s) is encoded by the first five and the last five exons (Fig. 3). This protein is expressed in many mesenchymal cells, neuroectodermal cells and some epithelial ceils [58,59]. In addition to the exons encoding the CD44s isoform, there are nine (in human) and ten (in rat and mouse) 'variant' exons which can be alternatively spliced to generate a wide variety of different variants ([57,60; Fig. 3). This accounts in part for the enormous heterogeneity of the CD44 proteins [61-65]. Thus, all known CD44 proteins are composed of the same extracellular N terminal sequence and transmembrane and cytoplasmic region of the CD44s proteins. The variants of CD44 contain additional amino acids derived from the variant exons, positioned in the extracellular portion close to the transmembrane domain. For brevity, we use the term CD44v for any CD44 protein containing sequences encoded by these variant exons. Antibody interference studies suggest that CD44 functions as a lymph node homing receptor on circulating lymphocytes [58,66], is involved in lymphocyte activation [67,68] and lymphopoiesis [69], and is responsible for homotypic and heterotypic cell adhesion [70, 71]. Since the majority of these antibodies recognize epitopes on the CD44 protein common to both CD44s and CD44v molecules, it is not clear whether the antibodies interfered with CD44s or CD44v functions. Recently, antibodies have been produced which recognize sequences exclusively encoded by variant exons, and are therefore specific for subsets of CD44 molecules [61,72. 73]. One of these antibodies was used to isolate two variants of CD44 which were specifically expressed in a highly metastasizing cell line (as tested in a spontaneous metastasis protocol, that in subcutaneous injection of tumor cells) derived from a rat pancreas adenocarcinoma, but not in a related but totally non-metastatic cell line derived from the same tumor. The two CD44 variant molecules contained the variant exons v4 to v7 and v6 to v7, respectively. The epitope recognized by the antibody was mapped to exon v6. Using polyclonal sera and monoclonal antibodies specific for the variable region of human CD44v proteins, a variety of human tumor tissue sections were examined. In colon carcinomas, both at the adenocarcinoma stage and on metastases, the incidence of expression of variants containing v6 is nearly 100%. In the precancerous stage of benign polyps the incidence is only 50% and in normal colon tissue we observe no v6 expression. In mammary carcinomas, all metastases examined were positive for CD44 v6 variant expression. However, only 80% of the primary tumors were also positive. Interestingly, and for prognostic purposes highly significant, the survival probability for patients

with primary tumors which do not express CD44 v6 variants is dramatically increased compared to patients with CD44 v6-positive primary tumors (M. Kaufmann, H.-P. Sinn, K.-H. Heider, H. Ponta and P. Herrlich, manuscript in preparation). It thus appears that the expression of CD44 variants is correlated in human cancers with transformation and tumor progression. We established a causal relationship between CD44v expression and metastasis formation in rats by two different approaches. Firstly, when antibodies specific for CD44 variant exon v6 were used to block the function of these variants by injection into tumor bearing animals, the spread of the tumor cells was retarded and in some instances completely blocked [74,75]. Secondly, CD44v cDNA clones were stably transfected into non-metastasizing tumor cells, and the metastatic capacity of the resulting CD44v-positive cells was tested in a spontaneous metastasis assay. As recipient cells we used the non-metastasizing rat pancreas adenocarcinoma cell line mentioned above and a fibrosarcoma cell line derived from a spontaneous sarcoma isolated from the same rat strain. Transfection and expression of several cDNA clones of CD44v which all contained the exon v6 sequence confered the complete metastatic phenotype onto both of the recipient cell lines when tested in the spontaneous metastasis assay. The smallest variant domain in a CD44v protein so far shown to trigger metastasis is encoded by exon v6 and v7, while the largest is encoded by all ten variable exons [61,75-77]. Overexpression of the CD44 standard isoform did not induce metastasis formation in this system, demonstrating the specificity and importance of variant domains in CD44v for spontaneous metastasis formation. The metastatic spread of the highly metastatic pancreatic adenocarcinoma cell and of the CD44v transfected cells occurs via the lymphatic system, since we first observe the outgrowth of lymph node metastases after s.c. injection of the tumor cells, and only at later times the appearance of lung metastases. Treatment of injected animals with the CD44 v6-specific antibody prevents the outgrowth of lymph nodes metastases but does not interfere with the migration of the tumor cells to the lymph node [75]. This is the first indication that the function of the CD44v protein is required for a process in the lymph node. A second line of evidence for such a lymph-node-specific function of CD44v is derived from the observation that CD44v proteins also play an important role in the normal immune response. The surveillance for 'foreign' molecules is mediated by lymphocytes which do not express CD44v, but highly express CD44s. Recognition of foreign epitopes induces the expression of CD44v. In peripheral lymphocytes, the respective variant is one that contains only v6 in the variant portion. This expression appears to be crucial, since CD44v-specific antibodies are able to

H. Ponta et al. / Biochimica et Biophysica A cta 1198 (1994) 1-10

repress an immune response [78]. It is very likely that the antibodies interfere with a reaction of CD44v in the lymph node because expansion of the activated T lymphocytes occurs in this compartment. In this respect activated T lymphocytes differ from resting ones, which just pass through the lymph node. Variant CD44 expression is not sufficient per se to induce metastasis, and other events such as transformation are apparently required. CD44 variants are expressed during embryogenesis, in some epithelial tissues such as keratinocytes and pancreatic ducts, and in antigen-stimulated lymphocytes [72,78,79], but obviously do not initiate metastatic behaviour in these cells. Furthermore, non-transformed rat fibroblasts do not metastasize when transfected with CD44v cDNA, but can do so if they are also transformed with ras (M. Hofmann, M. Z611er, P. Herrlich and H. Ponta, unpublished results). Additionally, expression of CD44 variants does not seem to be an obligatory requirement for metastasis formation in all tumors, as invasive melanomas do not express detectable levels of exon v6, although there is some weak expression of exons v8, v9, and vl0 (K.-H. Heider, J.C. Simon, P. Herrlich and H. Ponta, unpublished results), suggesting that metastasis may occur in these cases in the absence of significant levels of variant CD44 proteins. Similarly, expression of CD44 variants in gastric carcinomas does not always correlate with metastasis [80]. Nevertheless, the experimental evidence overwhelmingly indicates that in the majority of carcinomas and other types of tumour tested so far, variant CD44 proteins are an intrinsic and often limiting component of the cellular attributes required for metastasis formation, and that the acquisition of expression of these proteins is in many cases sufficient to initiate metastasis. In the rat, CD44v is clearly critical for the metastasis of tumors with origins as diverse as a pancreatic carcinoma and a fibrosarcoma.

7. Colonization of distant organs

Although CD44s is apparently not a key molecule in the metastatic spread out of a subcutaneous location, it might play a decisive role in other systems a n d / o r other steps of the metastatic cascade. The Namalwa Burkitt lymphoma cell line lacks expression of either CD44s and CD44v, and shows weak tumorigenic activity. Transfection and subsequent expression of CD44s in these cells results in the faster appearance of subcutaneous tumors and increases the metastatic potential as judged by the experimental metastasis assay (e.g., intravenous injection of tumor cells; [81]). Similar experiments in which a CD44v protein containing sequences encoded by variant exons 8-10 was expressed in the Namalwa cells had no such effect. Compared to

7

non-transfected Namalwa cells, the growth rate in vitro and the tissue distribution of metastases was not altered in the CD44s transfectants, suggesting the function of CD44s might depend on the interaction with a so far unidentified ligand. Such a ligand interaction is supported by the result that a soluble CD44s-Ig fusion protein suppresses the effect of CD44s on tumor growth when simultaneously injected with CD44s-transfected Namalwa cells [82]. A CD44v-Ig fusion protein that contained sequences derived from variant exons v8 to vl0 (but not v6) had no influence on tumor growth and metastasis formation in a similar assay. It should be noted that in the rat pancreas tumor system all cells already express CD44s, but the ones which metastasize after s.c. injection additionally express CD44v proteins, whereas in the Namalwa cells all CD44 isoforms are absent except for the one introduced by transfection. CD44s alone is not sufficient for lymphogenic spreading of tumor cells, but it might be required in addition to CD44v to provide tumor metastases a growth advantage at distant sites.

8. Putative mechanisms for CD44v protein function in metastasis induction

CD44v could function during metastasis in several ways (see Fig. 4). It seems most likely that the additional sequences allow the variant CD44 protein to interact with other molecules. Such interactions could be purely adhesive, or may also directly or indirectly involve signal transduction. Changes in adhesive properties a n d / o r signal transduction of CD44v as compared to CD44s would then trigger metastatic behaviour. Another possibility is that CD44v expression provides a growth advantage to the metastasizing cells. There is no evidence that CD44v itself acts as a ligand-dependent or -independent 'growth factor receptor' or as a signal transducing molecule. Direct signal transduction is probably ruled out, as a CD44v protein lacking most of the cytoplasmic portion initiates metastasis as efficiently as the intact protein (J. Sleeman, P. Herrlich and H. Ponta, unpublished results). Growth promotion could, however, be indirect, for example, by facilitating aggregation and attachment. This could lead to the stimulation of signal transduction through molecular interactions not involving CD44v, or to the stimulation of stromal cells to synthesize and release growth factors. Alternatively, heteromolecular interactions between CD44 variants and molecules on the same membrane could result in indirect signal transduction. The regulation of the interaction of standard CD44 molecules with the cytoskeleton is apparently functionally important in certain cells [83]. However, CD44v is

8

H. Ponta et al./Biochimica et Biophysica Acta 1198 (1994) 1-10

Hyaluronate

Ligand

Extra-cellular space

Cytoplasm

Indirect Signal Transduction

Cytoskeleton Fig. 4. Diagram summarizing possible molecular mechanisms by which variant CD44 proteins could induce metastasis formation. The incorporation of variant domains into the CD44 protein may induce signal transduction directlyor indirectly,perhaps through interaction with other molecules. Interactions with ligands may also bring about changes in adhesive properties, which could additionally be influenced by the abilityof the protein to bind to hyaluronate. For further details, see text. able to initiate metastasis both when it is dissociated from the cytoskeleton and when it is permanently attached as a chimeric protein fused at the C-terminal end to the spectrin binding domain of ankyrin (J. Sleeman, P. Herrlich and H. Ponta, unpublished results). The most intriguing possibility is that the alternative and combinatorial usage of more than 10 variant exons allows each CD44 variant to present a 'code'. Each code could be read by a specific ligand. Such ligand binding may be in conjunction with binding to hyaluronate, a property already present on the standard portion of the molecule. Similarly, the ligand affinity of CD44v proteins may be potentiated by self association or by interaction with other surface molecules to form molecular aggregates.

9. Conclusions

Despite properties associated volvement

enormous amounts of literature about the of metastasizing tumor cells and molecules with these properties, proof of causal inof such molecules in the development of

invasiveness and in the formation of distant metastases is surprisingly limited. Metastatic spread is dependent on the surface properties of tumor cells. Many such properties, e.g., loss of E-cadherin function, are already established with transformation, long before metastasis occurs. The observations obtained by studying CD44v expression suggests that in some circumstances only one additional surface molecule is required to convert a non-metastasizing tumor cell into a metastasizing one. We postulate that metastatic development is limited by a few key restricting molecules. The expression of these molecules may be required to allow the tumor cell to proceed to the next step in the metastatic pathway, or their suppression may be required to alleviate repression of metastasis. For example, the expression or suppression of the restricting molecule may permit the occupation of a new environment. Additional functions accessory to metastasis development may then further be acquired or induced after exposure to this new environment. It appears that CD44 v6 splice variants allow tumor cells to be exposed to such a new environment, namely lymphatic tissue. In this case, it is possible that the CD44v protein allows the tumor cells to mimic lymphocytes (reviewed in Ref. 84), perhaps permitting them to be retained and to proliferate in lymph nodes. Integrins may fulfill similar functions in other parts of the metastatic pathway. Only a subset of the many available proto-oncogenes needs to be activated or repressed in order to transform cells, and different subsets suffice to transform the same cell type. By analogy, the expression or repression of only a subset of potential metastasis-restricting molecules may need to be activated to allow tumor cells to metastasize. Such redundancy would mean that a particular molecule is absolutely required for metastasis formation in some tumors, while the same molecule is either not required or is already present in other tumors, where different molecules are restricting for metastasis formation. This would explain why, for example, CD44 v6 expression is not always sufficient or necessary for metastasis formation. In summary, only a few molecules have been causally shown to be involved in the metastatic process. As yet, the mechanisms by which several of these proteins are able to stimulate metastasis is unclear, although rapid progress in this respect has been made in recent months. Insights into the molecular basis of metastasis promotion promise to vastly increase our understanding of the metastatic process, and more importantly offer exciting possibilities for the diagnosis and therapy of cancer. References

[1] Weinberg, R.A. (1989) Cancer Res. 49, 3713-3721. [2] Bishop, J.M. (1991) Cell 64, 235-248.

H. Ponta et al. / Biochimica et Biophysica Acta 1198 (1994) 1-10

[3] Cantley, L.C., Auger, K.R., Carpenter, C., Duckworth, B., Graziani, A., Kapeller, R. and Soltoff, S. (1991) Cell 64, 281-302. [4] Lewin, B. (1991) Cell 64, 303-312. [5] Marshall, C.J. (1991) Cell 64, 313-326. [6] Bar-Sagi, D. and Feramisco, J.R. (1985) Cell 42, 841-848. [7] Greig, R.G., Koestler, T.P., Trainer, D.L, Corwin, S.P., Miles, L., Kline, T., Sweet, R., Yokoyama, S. and Poste, G. (1985) Proc. Natl. Acad. Sci. USA 82, 3698-3701. [8] Hurlin, P.J. and Fry, D.G. (1987) Cancer Res. 47, 5752-5757. [9] Kamps, M.P., Buss, J.E. and Sefton, B.M. (1985) Proc. Natl. Acad. Sci. USA 82, 4625-4628. [10] Liotta, L.A., Steeg, P.S. and Stetler-Stevenson, W.G. (1991) Cell 64, 327-336. [11] Birkedal-Hansen, H., Moore, W.G.I., Bodden, M.K., Windsor, L.J., Birkedal-Hansen, B., DeCarlo, A. and Engler, J.A. (1993) Crit. Rev. Oral Biol. Med. 42, 197-250. [12] Schrier, P.I., Bernards, R., Vaessen, R.T.M.J., Houweling, A. and van der Eb, A.J. (1983) Nature 305, 771-775. [13] Bernards, R., Schrier, P.I., Houweling, A., Bos, J.L., van der Eb, A.J., Zijlstra, M. and Melief, C.J.M. (1983) Nature 305, 776-779. [14] Tanaka, K., Isselbacher, K.J., Khoury, G. and Jay, G. (1985) Science 228, 26-30. [15] Eisenbach, L., Segal, S. and Feldman, M. (1983) Int. J. Cancer 32, 113-120. [16] Plaksin, D., Gelber, C., Feldman, M. and Eisenbach, L. (1988) Proc. Natl, Acad. Sci. USA 85, 4463-4467. [17] Wallich, R., Bulbuc, N,, H~immerling, G.J., Katzav, S., Segal, S. and Feldman, M. (1985) Nature 315, 301-305. [18] Porgador, A., Feldman, M. and Eisenbach, L. (1989) J. lmmunogenet. 16, 291-303. [19] Mandelboim, O., Feldman, M. and Eisenbach, L. (1992) J. Immunol. 148, 3666-3673. [20] Behrens, J., Birchmeier, W., Goodman, S.L. and Imhof, B.A. (1985) J. Cell Biol. 1001, 1307-1315. [21] Shimoyama, y. and Hirohashi, S. (1991) Cancer Res. 51, 21852192. [22] Burridge, K., Fath, K., Kelly, R., Nuckolls, G. and Turner, C. (1988) Annu. Rev. Cell Biol. 4, 487-525. [23] Roth, J., Zuber, C., Wagner, P., Taatjes, D.G., Weisgerber, C., Heitz, P.U., Goridis, C. and Bitter-Suermann, D. (1988) Proc. Natl. Acad. Sci. USA 85, 2999-3003. [24] Lipinsky, M., Braham, K., Philip, J., Wills, J., Philip, T., Goridis, C., Lenoir, G. and Tursz, T. (1987) Cancer Res. 47, 183-187. [25] Takeichi, M. (1991) Science 251, 1451-1455. [26] Frixen, U.H., Behrens, J., Sachs, M., Eberle, G., Voss, B., Warda, A., L6chner, D. and Birchmeier, W. (1991) J. Cell Biol. 113, 173-185. [27] Behrens, J., Mareel, M.M., van Roy, F.M. and Birchmeier, W. (1989) J. Cell Biol. 198, 2435-2447. [28] Vleminckx, K., Vakaet Jr., L., Mareel, M., Fiefs, W. and Van Roy, F. (1991) Cell 66, 107-119. [29] Chen, W. and Obrink, B. (1991) J. Cell Biol. 114, 319-327. [30] Navarro, P., G6mez, M., Pizarro, A., Gamallo, C., Quintanilla, M. and Cano, A. (1991) J. Cell Biol. 115, 517-533. [31] Nagafuchi, A., Takeichi, M. and Tsukita, S.H. (1991) Cell 65, 849-857. [32] Ozawa, M., Ringwald, M. and Kemler, R. (1990) Proc. Natl. Acad. Sci. USA 87, 4246-4250. [33] Hirano, S., Kimoto, N., Shimoyama, Y., Hirohashi, S. and Takeichi, M. (1992) Cell 70, 293-301. [34] Tsukita, Sh., Tsukita, S., Nagafuchi, A. and Yonemura, S. (1992) Curr, Biol. 4, 834-839. [35] Magee, A. and Buxton, R.S. (1991) Curr. Opin. Cell Biol. 3, 854-861. [36] Behrens, J., Vakaet, L., Friis, R., Winterhager, E., van Roy, F., Mareel, M.M. and Birchmeier, W. (1993) J. Cell Biol. 120, 757-766.

9

[37] Matsuyoshi, N., Hamaguchi, M., Taniguchi, S., Nagafuchi, A., Tsukita, S. and Takeichi, M. (1992) J. Cell Biol. 118, 703-714. [38] Bendori, R.D., Saiomon, D. and Geiger, B. (1989) J. Cell Biol. 108, 2383-2393. [39] Geiger, B. and Ginsberg, D. (1991) Cell Motil. Cytoskeleton 20, 1-6. [40] Sefton, B.M., Hunter, T., Ball, E.H. and Singer, S.J. (1981) Cell 24, 165-174. [41] Fernfindez, J.L.R., Geiger, B., Salomon, D., Sabanay, I., Zfller, M. and Ben-Ze'ev, A. (1992) J. Cell Biol. 119, 427-438. [42] Michalides, R.J.A.M., Moolenaar, K.C.E.C., Kibbelaar, R.E. and Mooi, W.J. (1990) in Molecular Biology of Cancer Genes (Sluyer, M., ed.), Ellis Horwood, NY, pp. 223-236. [43] Fearon, E.R., Cho, K.R., Nigro, J.M., Kern, S.E., Simons, J.W., Ruppert, J.M., Hamilton, S.R., Preisinger, A.C., Thomas, G., Kinzler, K.W. and Vogelstein, B. (1990) Science 247, 49-56. [44] Moolenaar, C.E.C.K., Muller, E.J., Schol, D.J., Figdor, C.G., Bock, E., Bitter-Suermann, D. and Michalides, R.J.A.M. (1990) Cancer Res. 50, 1102-1106. [45] Murphy, G., Docherty, A.J.P., Hembry, R.M. and Reynolds, J.J. (1991) Br. J. Rheum. 30, 25-31. [46] Mignatti, P., Tsuboi, R., Robbins, E. and Rifldn, D.B. (1989) J. Cell Biol. 108, 671-682. [47] Schultz, R.M., Silberman, S., Persky, B., Bajkowski, A.S. and Carmichael, D.F. (1988) Cancer Res. 48, 5539-5545. [48] Alvarez, O.A., Carmichael, D.F. and CeClerk, Y.A. (1990) J. Natl. Cancer Inst. 82, 589-595. [49] Albini, A., Melchiori, A., Santi, L, Liotta, L.A., Brown, P.D. and Stetler-Stevenson, W.G. (1991) J. Natl. Cancer Inst. 83, 775-779. [50] DeClerck, Y.A., Yean, T.-D., Chan, D., Shimada, H. and Langley, K.E. (1991) Cancer Res. 51, 2151-2157. [51] Khokha, R., Waterhouse, P., Yagel, S., Lala, P.K., Overall, C.M., Norton, G. and Denhardt, D.T. (1989) Science 243, 947950. [52] Hynes, R.O. (1992) Cell 69, 11-25. [53] Gehlsen, K.R., Argraves, W.S., Pierschbacher, M.D. and Ruoslahti, E. (1988) J. Cell. Biol. 106, 925-930. [54] Humphries, M.J., Olden, K. and Yamada, K.M. (1986) Science 233, 467-470. [55] Saiki, I., Iida, J., Murata, J., Ogawa, R., Nishi, N., Sugimura, K., Tokura, S. and Azuma, I. (1989) Cancer Res. 49, 3815-3822, [56] Chart, B.M.C., Matsuura, N., Takada, Y., Zetter, B.R. and Hemler, M.E. (1991) Science 251, 1600-1602. [57] Screaton, G.R., Bell, M.V., Jackson, D.G., Cornelis, F.B., Gerth, U. and Bell, J.I. (1992) Proc. Natl. Acad. Sci. USA 89, 1216012164. [58] Jalkanen, S., Bargatze, R., Herron, L. and Butcher, E.C. (1986) Eur. J. Immunol. 16, 1195-1202. [59] Picker, L.J., Nakache, M. and Butcher, E.C. (1989) J. Cell Biol. 109, 927-937. [60] T61g, C., Hofmann, M., Herrlich, P. and Ponta, H. (1993) Nucleic Acids Res. 21, 1225-1229. [61] Giinthert, U., Hofmann, M., Rudy, W., Reber, S., Zfller, M., HauBmann, I., Matzku, S., Wenzel, A., Ponta, H. and Herrlich, P. (1991) Cell 65, 13-24. [62] Stamenkovic, I., Aruffo, A., Amiot, M. and Seed, B. (1991) EMBO J. 10, 343-348. [63] Omary, M.B., Trowbridge, I.S., Letarte, M., Kagnoff, M.F. and Isacke, C.M. (1988) Immunogenetics 27, 460-464. [64] Dougherty, G.J., Lansdorp, P.M., Cooper, D.L. and Humphries, R.K. (1991) J. Exp. Med. 174, 1-5. [65] Jackson, D.G., Buckley, J. and Bell, J.I. (1992) J. Biol. Chem. 267, 4732-4739. [66] Jalkanen, S., Steere, A.C., Fox, R.1. and Butcher, E.C. (1986) Science 233, 556-558.

10

H. Ponta et aL /Biochimica et Biophysica Acta 1198 (1994) 1-10

[67] Denning, S.M., Le, P.T., Singer, K.H. and Haynes, B.F. (1990) J. Immunol. 144, 7-15. [68] Huet, S., Groux, H., Caillou, B., Valentin, H., Prieur, M. and Bernard, A. (1989) J. Immunol. 143, 798-801. [69] Miyake, K., Underhill, C.B., Lesley, J. and Kincade, P.W. (1990) J. Exp. Med. 172, 69-75. [70] Koopman, G., van Kooyk, Y., de Graaff, M., Meyer, C.J.L.M., Figdor, C.G. and Pals, S.T. (1990) J. Immunol. 145, 3589-3593. [71] Shimizu, Y., van Seventer, G.A., Siraganian, R., Wahl, L. and Shaw, S. (1989) J. Immunol. 143, 2457-2463. [72] Heider, K.-H., Hofmann, M., Horst, E., van den Berg, F., Ponta, H., Herrlich, P. and Pals, S.T. (1993) J. Cell Biol. 120, 227-233. [73] Koopman, G., Heider, K.-H., Horst, E., Adolf, G.R., van den Berg, F., Ponta, H., Herrlich, P. and Pals, S.T. (1993) J. Exp. Med. 177, 897-904. [74] Reber, S., Matzku, S., Giinthert, U., Ponta, H., Herrlich, P. and Z611er, M. (1990) Int. J. Cancer 46, 919-927. [75] Seiter, S., Arch, R., Komitowski, D., Hofmann, M., Ponta, H., Herrlich, P., Matzku, S. and Z611er, M. (1993) J. Exp. Med. 177, 443-455. [76] Rudy, W., Hofmann, M., Schwartz-Albiez, R., Z611er, M., Heider, K.-H., Ponta, H. and Herrlich, P. (1993) Cancer Res. 53, 1262-1268.

[77] Herrlich, P., Rudy, P., Hofmann, M., Arch, R., Z611er, M., Zawadzki, V., T61g, C., Hekele, A., Koopman, G., Pals, S., Heider, K.-H., Sleeman, J. and Ponta, H. (1993) in Cell adhesion molecules (Hemler, M.E. and Mihich, E., eds.), Plenum Press, New York, pp. 265-288. [78] Arch, R., Wirth, K., Hofmann, M., Ponta, H., Matzku, S., Herrlich, P. and Z611er, M. (1992) Science 257, 682-685. [79] Wirth, K., Arch, R., Somasundaram, C., Hofmann, M., Weber, B., Herrlich, P., Matzku, S. and Z611er, M. (1993) Eur. J. Cancer 29A, 1177-1183. [80] Heider, K.-H., D~immrich, J., Skroch-Angel, P., Miiller-Hermelink, H.-K., Vollmers, H.P., Herrlich, P. and Ponta, H. (1993) Cancer Res., in press. [81] Sy, M.S., Guo, Y.-J. and Stamenkovic, I. (1991) J. Exp. Med. 174, 859-866. [82] Sy, M.-S., Guo, Y.-J. and Stamenkovic, I. (1992) J. Exp. Med. 176, 623-627. [83] Camp, R.L., Kraus, T.A. and PurL, E. (1991) J. Cell Biol. 115, 1283- 1292. [84] Herrlich, P., Pals, S.T., Z611er, M. and Ponta, H. (1993) Immunol. Today 14, 395-399.