The functional role(s) of dual expression in tumor cell migration and invasion.

Item Type text; Dissertation-Reproduction (electronic)

Authors Chu, Yi-Wen.

Publisher The University of Arizona.

Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Download date 02/10/2021 04:10:29

Link to Item http://hdl.handle.net/10150/186143 INFORMATION TO USERS

This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer.

The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleed through, substandard margins, and improper alignment can adversely affect reproduction.

In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.

Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand corner and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book.

Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order.

U·M·I University Microfilms International A Bell & Howell Information Company 300 North Zeeb Road. Ann Arbor. M148106-1346 USA 313/761-4700 800:521-0600

Order Number 9322644

The functional role(s) of dual intermediate filament expression in tumor cell migration and invasion

Chu, Yi-Wen, Ph.D.

The University of Arizona, 1993

V·M·I 300 N. Zccb Rd. Ann Arbor. MI 48106

THE FUNCTIONAL ROLE(S) OF DUAL INTERlVIEDIA TE FILAlv1ENT

EXPRESSION IN TUMOR CELL rvnGRA TION AND INVASION

by Yi-Wen Chu

A Dissertation Submitted to the Faculty of the

COMMITTEE ON CANCER BIOLOGY (GRADUATE)

In Partial Fulfillment of the Requirements For the Degree of DOCTOR OF PHILOSOPHY

In the Graduate College THE UNIVERSITY OF ARIZONA

199 3 2

THE UNIVERSITY OF ARIZONA GRADUATE COLLEGE

As members of the Final Examination Committee, we certify that we have read the dissertation prepared by______Yi-Wen Chu __

entitled THE FUNCTIONAL ROLE (S) OF DUAL INTERMEDIATE FILA."'1ENT

EXPRESSION IN TUMOR CELL HIGRATION AND INVASION

and recommend that it be accepted as fulfilling the dissertation

requirement for the Degree of _D~o~c~t~o~r_o~f_P~hu.i~lo~s~o~p~h~y~ ______

y I r/1.3 DatJ ' tf /q/9'3 Date 1 Cj /~iS Bernstein <___ / ~ ~ ~~ Date'

Final approval and acceptance of this dissertation is contingent upon the candidate's submission of the final copy of the dissertation to the Graduate College.

I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement. 3

STATEMENT BY AUTHOR

This dissertation has been submitted in partial fulfillment of requirements for an advanced degree at The University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library. Brief quotations from this dissertation are allowable without special permission, provided that accurate acknowledgment of source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his or her judgment the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author.

SIGNED: f- ~ ~ 4 ACKNOWLEDGMENTS

At this moment, the last stage of my graduate study, I feel grateful to many of the wonderful people I have met in the past several years. First, I would like to express my deepest appreciation to my major advisor, Dr. Mary J.C. Hendrix. She has always shown great confidence in me, and provided support during the most difficult times of my life. She not only provided the guidance for my dissertation research, but also introduced me to many prestigious scientists in this field and gave me the opportunity to collaborate with people. She has spent a tremendous amount of time helping me with my writing and communication skills. With her pleasant personality, she makes me feel there are always alternatives and hopes even at the most frustrating moments. Most importantly, as a woman scientist in the modern world, she demonstrates that with intelligence, enthusiasm and endeavor, we can be what we want to be. She is definitively a world model for me in my future career. To my committee members, I would like to thank them for their valuable comments and suggestions for this work. I am very grateful to Dr. John J. Duffy for the solid training he gave me in Molecular Biology. I would also like to express my deepest gratitude to Dr. Richard B.B. Seftor and Elisabeth Seftor for their expertise and scientific discussions, and their constant encouragement and personal support through my Ph.D. career. I am very lucky to have colleagues like Catherine Frye, Adrian Winter, Hanifa Jones and Padma Sundareshan. I thank them for their help and warm friendship that made everyday of research life filled with joy and laughter. I also would like to thank the Iowa Motility Center staff and faculty, including Drs. David SoIl, Ray Runyan, Karla Daniels, and Michael Solursh, for their help in the analysis of cellular motility and their assistance with confocal microscopy. 5

DEDICATED TO MY PARENTS

SHUN CHU

AND

SHU·CHENG CHENG CHU 6 TABLE OF CONTENTS

LIST OF ILLUSTRATIONS ...... 8

LIST OF TABLE ...... 9 ABSTRACT ...... 10 I. INTRODUCTION AND OVERVIEW OF LITERATURE ...... 12 1.l Mechanisms of Tumor Cell Invasion and Metastasis ...... 12 l.2 Cytoskeletal and Neoplasia ...... 18 1.3 and Tumor Transformation ...... 24 1.4 Coexpression of IFs in Metastatic Tumor Cells ...... 28 l.5 Hypothesis: Dual IF Expression in Highly Metastatic Cells ...... 30 1.6 Dissertation Format ...... 32 ll. PRESENT STUDY ...... 34 2.1 Disruption of Filaments in Highly Metastatic Cells ...... 34 2.2 Mouse L Cell Model: Effects of Dual IF Expression ...... 43 2.3 Transfections of Keratins in Low Metastatic Melanoma Cells ...... 51 2.4 Conclusion ...... 57 Ill. APPENDIX A: Transfection of a Deleted CK18 cDNA into a Highly Metastatic Melanoma Cell Line Decreases the Invasive Potential...... 63 N. APPENDIX B: Coexpression of and Keratins by Human Melanoma Tumor Cells: Correlation with Invasive and Metastatic Potential ...... 74 7 TABLE OF CONTENTS - Continued

V. APPENDIX C: Expression of Complete Keratin Filaments in Mouse L Cells Augments Cell Migration and Invasion...... 88 VI APPENDIX D: Co expression of Vimentin and Keratins Intermediate Filaments in Human Melanoma Cells Correlates with Increased Motility and Decreased Integrin Localization in Focal Contacts ...... 127

VII. REFERENCES ...... 149 8 LIST OF ILLUSTRATIONS

FIGURE 1, Hypothesis: Dual IF Expression and Cellular Response..... 31

FIGURE 2, Plasmids Used in the Transfection Study ...... 36 FIGURE 3, Confocal Microscopy of CI070 Cells ...... 39 FIGURE 4, Attachment Ability of L Cells on Matrigel...... 46 FIGURE 5, Fluorescence Localization of F- in L Cells...... 48 FIGURE 6, Plasmids Containing Keratins 8 and 18 cDNAs ...... 52 FIGURE 7, DNA Slot Blot Analysis of Transfected A375P Cells ...... 55

FIGURE 8, Model for Dual IF Expression Augmenting Motility ...... 61 9 LIST OF TABLE

TABLE 1, Migration Rate Measured by Video Cinematography ...... 42 10 ABSTRACT

Tumor cell invasion and metastasis are very complicated biological events which involve numerous classes of proteins that participate in controling each step of the metastatic cascade. Therefore, identifying the key factor(s) that determine(s) how a tumor cell becomes more metastatic is a fascinating issue that challenges the field of cancer biology. Intermediate filaments are cytoskeletal proteins whose expression is highly regulated in a cell-type specific manner; however, recent evidence has indicated that coexpression of two different types of intermediate filaments -- specifically keratin(s) (epithelial cell marker) and vimentin (mesenchymal cell marker), in the tumor cells correlates with their invasive and metastatic potential. The focus of this dissertation is to further elucidate the functional role(s) of this dual intermediate filament expression in tumor cells. A dominant negative mutant keratin cDNA was used to transfect a highly metastatic human melanoma cell line, C8161, which contains both vimentin and keratin filaments. The resulting transfected clones showed disrupted keratin filaments by immunofluorescence microscopy. Subsequently, their migratory and invasive ability were reduced, in addition to complete abrogation of metastatic potential. Another strategy involved the transfection of and 18 DNAs into keratin-negative cells (both mouse L fibroblasts and low invasive A375P melanoma), which resulted in clones expressing the dual intermediate filament phenotype, commensurate with increased migratory and invasive ability. Furthermore, these experimental clones have a retarded spreading ability on extracellular matrix compared to the control transfectants. In addition, they do not contain any detectable uvf33, u3 or u6 integrins in the focal contact sites by 11 immunofluorescence staining. Hence, it is postulated that the mechanism responsible for differential spreading ability rests in the unique regulation of specific integrin(s) localized in focal contacts, acting either directly or indirectly, with the intermediate filaments and with extracellular matrix molecules. These results suggest that dual intermediate filament expression is important for the invasive phenotype, and their heretofore assigned role as general maintenance structural proteins has changed to that of dynamic cytoskeletal elements involved in active cellular function(s). 12 CHAPTER 1 INTRODUCTION AND OVERVIEW OF LITERATURE

The most difficult problem that challenges modern development of cancer treatment strategies is the metastatic capacity of primary tumors to disseminate to distant organs. The dispersed, tiny micrometastases are the major factors that limit the effectiveness of therapeutic modalities. Therefore, most of the patients that fail to respond to radiotherapy, chemotherapy or surgery succumb to the occurrence of multiple metastases. Unfortunately, the distant metastases are often too small to be detected at early stages, and show no clinical symptoms. In fact, it is estimated that fifty percent of the patients with solid tumors already have sub­ clinical micrometastases at the time of diagnosis of primary tumors (Liotta, 1986). One major focus in cancer research is to study the mechanisms of tumor metastasis so that we can understand how tumor cells successfully move from one organ to another in the human body. Basic research scientists have combined the techniques and knowledge from cell biology, molecular biology, biochemistry and immunology as different approaches to answer this question. The ultimate goal of this research is to find prognostic markers associated with tumor cells to reliably predict their metastatic potential, and methods to block the occurrence of metastasis, so that the appropriate therapeutic strategies can be developed and applied in a timely and effective fashion.

1.1 Mechanisms of Tumor Cell Invasion and Metastasis

Cancer metastasis is a very complicated process. It consists of sequential interrelated events, which tumor cells must accomplish and survive all the events in order to successfully form metastases. This process has been divided into the 13 following major steps: (a) initiation of the primary neoplasm via transformation; (b) extensive vascularization surrounding the tumor mass; (c) local invasion of tumor cells into the lymphatic or vascular circulatory system(s); (d) detachment and embolization of tumor aggregates; (e) survival of tumor cells exposed to the host immune system and subsequent arrest in capillary beds of distant organs; (f) extravasation; and (g) proliferation of tumor cells within the organ parachyma, thus completing one cycle of metastasis (Fidler, 1990). It is estimated that only a very small percentage « 0.01 %) of circulating tumor cells from a primary neoplasm successfully form metastatic colonies (Liotta and Stetler-Stevenson, 1991). These tumor cells must express all the required phenotypes that allow them to complete every step in the process of metastasis without being eliminated. Most recent data have suggested that tumors are biologically heterogeneous, and this is also the case for the metastatic propensity in the primary neoplasm. Fidler and Hart (1982) have viewed metastasis as a highly selective competition process, favoring the outgrowth and survival of a metastatic subpopulation that preexist within the heterogeneous primary tumor. A well­ known model demonstrating this concept is the isolation of the highly metastatic subpopulation of B 16 mouse melanoma cells by repeated isolation of the metastatic cells which were implanted into mice (Fidler, 1973). Moreover, in an in vitro model using a human melanoma cell line, A375P, more invasive and metastatic sublines can be established by several rounds of invasion selection from the original cell population (Seftor et al. 1990). Interestingly, the increase in the metastatic ability of selected tumor cells does not apparently result from the adaption of cells to preferential growth (Raz et al. 1981). Recently, Kerbel (1990) has used genetic markers to show that a metastatic subpopulation dominates the 14 primary tumor mass during early growth. Thus, it is possible to identify specific markers which will give the tumor cells a selective advantage in the metastatic process.

One critical step in this process is the ability of tumor cells to invade or penetrate biological matrices, which occurs during tumor intravasation, extravasation and dissemination. Invasion has been further subdivided into a three-step model: (1) attachment of tumor cells to a matrix substratum; (2) local degradation of the matrix by secreted proteinases; and (3) tumor cell movement into the matrix modified by proteolysis (Liotta et aI., 1986a). The occurrence of invasion is thought to result from an imbalance of positive and negative regulation (Liotta et al., 1991). Loss of this coordinated control in each step can result in the generation of invasive and metastatic phenotypes. Without a doubt, there are many key players in the process of invasion. When tumor cells escape the primary tumor, the first challenge is the barrier formed by extracellular matrix (ECM). The ECM is divided into two major compartments: and interstitial stroma. The basement membrane is composed of mainly IV, , , entactin and heparin sulfate proteoglycans which form a dense meshwork structure underlying the epithelial or endothelial boundaries. By comparison, the interstitial stroma includes other collagen types, proteoglycans and additional fibrilar components, arranged in a tissue specific manner (Tryggvason et al., 1987). Typically, cells, including tumor cells, attach to the ECM via specific receptors on the plasma membrane. One major class of cell/matrix adhesion molecules is the integrins, which are transmembrane receptors. They are heterodimers consisting of noncovalently associated a. and ~ subunits, and different combinations of a. and ~ peptides determines their 15 specificity to the ligands of ECM components (Albelda and Buck, 1990; Ruoslahti, 1991; Hynes, 1992). There is evidence to suggest the expression of integrins on tumor cells is closely related to the process of tumor progression. For instance, it has been shown that the level of

The second step of the invasion process is the proteolytic degradation of ECM. The basement membrane has a very dense matrix structure that does not allow cells to passively migrate; therefore, invasion cannot occur until the matrix has been degraded. This process requires a cascade of proteinases including the interaction of plasminogen activators and different classes of metalloproteinases (Mignatti et aI., 1986; Reich et aI., 1988; Matrisian, 1990). Plasminogen activators (P A) are highly specific proteinases that convert plasminogen into plasmin which activates latent collagenase, and leads to the degradation of collagen in ECM. Analysis of a series of cancers have revealed that the presence of PA in tumor cells is correlated with their metastatic ability (Oka et aI., 1991). Moreover, 16 transfection studies have also indicated that the expression of P A alone is sufficient to confer an invasive phenotype to mouse L cells (Cajot et aI., 1989). The matrix metalloproteinase gene family consists of three subgroups: interstitial collagenase, type IV collagenase, and stromelysins based on their substrate preference. All of these metalloproteinases are secreted in latent form and are activated by proteolysis activity. There are numerous reports regarding the correlation of increased expression of these matrix degrading enzymes and the metastatic ability of tumor cells (Matrisian et aI., 1986; Templeton and Stetler­ Stevenson, 1991; Garbisa et al., 1992; Hendrix et al., 1992; Sreenath et al., 1992; Powell et aI., 1993). These studies have suggested that up-regulation of these proteinases are one of the important factors that determine the invasive and metastatic potential of tumor cells. However, recent efforts in identifying new proteases participating in the metastatic cascade have broadened our knowledge about how the ECM is degraded. In conjunction with these proteinases are specific proteinase inhibitors that are important as well in controlling the invasive phenotype. These include plasminogen activator inhibitor type-I, type-2 (Kruithof, 1988), tissue inhibitor(s) of metalloproteinase(s) (TIMP-l; Carmichael et al., 1986), and TIMP-2 (Stetler-Stevenson et al., 1989), which inhibit the degradation activities of plasminogen activator or metalloproteinases, respectively. Consequently, a decrease in the expression of TIMP is associated with increased metastasis (Ponton et aI., 1991), and its role in tumor invasion has further been supported by recombinant TIMP (DeClerck et al., 1991) and gene targeting (Alexander and Werb, 1992) studies. Therefore, tumor invasion occurs, in part, as a result of an aberrant induction and/or "lack of suppressing" of degradation by the imbalanced activation and inhibition of proteinases. Hence, 17 this serves as a good example of the requirement for a positive and negative balance of regulation in the biological system as mentioned earlier.

Tumor cell migration is the final step in the invasion process, and it is important for cell movement across the degraded ECM (Strauli and Haemmerli, 1984). Highly metastatic cells usually have increased motility compared to low metastatic cells, which contributes to the successful completion of metastatic cascade (Volk et al., 1984; Tullberg and Burger, 1985; Verschueren et al., 1988). The mechanisms behind cell motility are less understood than the previous two processes; however, several factors have been identified associated with the locomotion of cells. These include a fibroblast-derived scatter factor that modulates epithelial cell mobility in a paracrine fashion (Gherarde et al., 1989); insulin-like growth factor II which stimulates rhabdomyosarcoma tumor motility (EI-Badry et al., 1990); and the autocrine motility factor, which acts through a receptor-activated G that enhances random tumor cell movement (Liotta et aI., 1986b; Silletti et al., 1991). In addition, locomotion is also regulated by the coordination of cytoskeletal elements and inner membrane surface, which results in pseudopodial extension at the leading edge of the migrating cell.

There are still many different classes of proteins, either identified or yet undiscovered, that are involved in the metastatic process. Oncogenes, especially ras, have been shown to play an important role in controlling the metastatic phenotype (Muschel and Liotta, 1988). Growth factors such as EGF and TGF-p can indirectly modulate the invasive ability by controlling the expression of ECM degradation enzymes (Niedbala and Sartorelli, 1989; Kerr et al., 1990). Recently, a variant of the CD44 adhesion molecule has been identified that associates with 18 metastasizing tumor cells, which provides a very useful marker of tumor progression (Arch et al., 1992). Another type of cell-cell adhesion molecule, E­ cadherin, is found to be down-regulated in high metastatic cells (Frixen et aI., 1991; Bussemakers et aI., 1992), which partially explains how metastatic cells detach and escape from primary tumor masses. The most interesting finding in the long search for the mechanisms underlying tumor metastasis probably is the discovery of metastasis suppressor genes (Ichikawa et aI., 1991; 1992). The best characterized gene is nm-23, in which reduced expression of nm-23 usually associates with highly metastatic tumors (Steeg et aI., 1988).

All of these efforts in identifying the alteration of genes and proteins in low- and high-metastatic cells have helped in providing- ;l1sight into how a cancer cell becomes metastatic. This information may allow us to develop more helpful strategies that can be used in clinical diagnosis and prognosis, and hopefully to the prevention of tumor invasion and metastasis.

1.2 Cytoskeletal Proteins and Neoplasia

The can be defined as the proteins constituting the insoluble filamentous matrix components in the cytoplasm. They provide the architectural support of the cells, including positioning of subcellular organelles, cell migration, local movements of the plasma membrane, lateral redistribution of surface components, and endocytosis. There are three major types of cytoskeletal proteins in eukaryotic cells: actin filaments, intermediate filaments (IFs) and , which are classified according to their filament sizes. The following sections will discuss their structures, associated proteins, functions and specific agents that affect each type of filament. 19 The actin filaments are composed of a single polypeptide known as G actin, which is one of the most abundant and conserved proteins in cells. Upon polymerization, they form the filamentous structure called F actin. Transmission electron microscopy analysis reveals that the actin filaments consist of two strands of globular molecules, at 4 nm in diameter each, twisted into a helix form (Pollard and Cooper, 1986). When stress tension arises in the environment, these filaments will form contractile bundles -- the stress fibers, which contain tightly packed parallel actin filaments. One end of the stress fiber is connected to the plasma membrane at focal contact sites. This feature is important for the attachment and deattachment action of cells to specific substrata. There are numerous proteins that associate with , to name a few, , pro filin, , , and . Most of these proteins are involved in the filament cross-link formation or attachment interaction of actin filaments to the cytoplasmic membrane. The most common reagents used to disrupt actin filaments are cytochalasin B which prevents actin polymerization, and phalloidin that, in contrast, inhibits depolymerization. These drugs have provided a very useful method to study the cellular activities that involve actin filament (Cooper, 1987). Actin, together with , is involved in muscle contraction. However, in the non-muscle cells, their major function is the regulation of locomotion of various cell types such as polymorphonuclear leukocytes, fibroblasts and early regenerated endothelial cells. Recent studies have shown that phosphoinositide kinase and protein kinase C are associated with actin filament system especially after EGF induction. This evidence implicates actin possibly playing a role in signal transduction by providing a matrix for the binding of the components involved in the signalling pathway (Jaken et aI., 1989; Payrastre et aI., 1991). 20 Intermediate filaments, ranging from 8 to 12 nm in diameter, are the most complicated group among cytoskeletal proteins. They consist of a heterogeneous multigene family which can be further subdivided into six groups: type I - acidic keratins; type II - neutral and basic keratins; type III - vimentin, , and glial fibrillary acidic protein (GFAP); type IV - ; type V - nuclear and the newly discovered type VI - (Steinert and Roop, 1988; Lendahl et al., 1990). Most of the IF proteins possess a central a- helical "rod" domain of conserved secondary structure and amino- and carboxyl­ terminal "end" domains of widely varying sizes and chemical characters. The individual protein molecules form coiled-coil dimers, which are then aligned in a half-stagger brick relation for eight-chain polymer complexes (Steinert et al., 1985). Most of the IF expression is highly regulated in a cell-type specific manner: keratins are expressed in epithelial cells; vimentin is found in mesenchymal cells; desmin in muscle, GFAP in astrocytes; and neurofilaments in neurons of central and peripheral nerves. The pattern of expression is usually maintained during neoplastic transformation; therefore, the typing of IFs by immunohistochemical methods is extremely important for the pathologist in determining the origin of neoplastic lesions (Osborn and Weber, 1982; Ivanyi et al., 1990). There is a growing number of proteins that have been identified associated with IFs, of which and are the best characterized. Plectin can bind to various IF proteins at the a-helical rod domain and function as a crosslinking element of IFs; filaggrins are expressed specifically in terminally differentiated epidermis, and they have been shown to cause aggregation of keratin filaments (Foisner and Wiche, 1991). The function(s) of IFs are less understood: the existence of different types of IF in different cell lineages 21 suggests that they are functionally different (Klymkowsky et al., 1989). Moreover, different types of IF expression during cell differentiation indicates the IF sUbtypes may also have distinct functions. Basically, the IF network is thought to be important for cell shape, nuclear centration, cell-cell contact, and organeller transport (Steinert et al., 1984). (The specific function of keratin filaments will be discussed further later in this chapter.) One of the reasons that it is difficult to study the function of IFs is the lack of drugs or agents which can disrupt these filaments specifically. Some of the reported agents such as acrylamide, UV radiation and teleocidin can cause IFs to collapse, but they also have effects on other components of the cytoskeleton.

Microtubules are the largest cytoskeletal components with a size of 25 nm in diameter. The major building block of microtubules is , which contains one a. and one J3 chain that form heterodimers (Dustin, 1984). The assembly and disassembly of the tubulin subunits are highly regulated 'processes during the cell cycle. In interphase cells, microtubules form a network which originate from a perinuclear region -- the organization center (Brinkley et al., 1981). Upon the onset of mitosis, this network disassembles and is subsequently rearranged into the mitotic spindle. Several microtubule associated proteins can be co-purified with tubulin. These include MAP-I, MAP-2 and Tau proteins that are important for the nucleation and elongation process during assembly (Drubin and Kirschner, 1986; Lewis et al., 1989). Microtubules participate in many cellular activities such as mitosis and cell division, cell morphogenesis, vesicle transport, and positioning of organelles. Several drugs have been shown to bind to tubulin: colchicine and co1cemid can inhibit the addition of tubulin subunits to micro tubules which leads to depolymerization; vinblastine and vincristine can 22 induce the formation of paracrystalline aggregates of tubulin; on the other hand, taxol, an alkaloid isolated from Pacific yew trees, can stablize the microtubules and block cell division (DeBrabander et al., 1982). Hence, some of these agents have been successfully tested on the treatment of melanoma, ovarian carcinoma and breast cancer in the past few years.

These three types of cytoskeletal proteins probably do not form their networks independently; there is evidence to suggest that these filaments may intereact with each other throughout the cytoplasm (Ball and Singer, 1981; Green et al., 1986). These interactions may be provided by the filament associated proteins such as MAP and plectin. However, how these filaments correspond to each other and whether they serve any functional purpose are still unknown. One of the rising fields in the study of the cytoskeleton is the structural relationship of filaments to the plasma and nuclear membranes (Luna and Hitt, 1992). The association of actin filaments with cell surface receptors involves the binding of , and a- proteins. The distribution pattern of IFs in the cytoplasm seems to provide a bridge between the cell membrane and nucleus: these filaments form a perinuclear ring around the nucleus and then extend through the entire cytoplasm and terminate at the plasma membrane. Several studies have shown that there are contact sites on both ends of the IFs connecting the plasma membrane and nuclear lamins. Recent research indicates that B represents a constitutive nuclear "receptor" site for the tail domains of peripherin IFs (Djabali et al., 1991). Georgatos and Blobel (1987a) have also shown that vimentin associates with the nuclear surface at the carboxy-terminal tail and also interacts with the plasma membrane through the amino-terminal head domain. They further suggest that lamin B is actually the attachment site for 23 vimentin at the nuclear envelope (Georgatos and Blobel, 1987b). Goldman and coworkers (1986) reported that the interconnecting network of the cytoskeletal­ nuclear matrix may be involved in signal transmission between the nucleus and cytoplasmic compartments of eukaryotic cells.

The formation of cytoskeletal filaments is a very dynamic process during the cell cycle. Recent studies have shown that these actions are controlled by phosphorylation-dephosphorylation activities. The protein p34Cdc2 plays an important part in this activity, and the vimentin and lamin have both been shown to be the substrates of this kinase (Chou et al., 1990; Ward and Kirschner, 1990).

What is the significance of cytoskeletal proteins in neoplasms and metastasis? The cytoskeletal typing is a very important method to diagnose the origin of a neoplasm. Using a whole panel of antibodies against cytoskeletal proteins can efficiently and accurately determine the type(s) of tumors and the differentiation state(s) (Osborn and Weber, 1989). Furthermore, several cytoskeleton-disruption agents have been shown to dramatically affect the metastatic behavior of tumor cells. For example, colchicine and cytochalasin B are the agents that disrupt microtubules and , respectively. Injection of mice with tumor cells treated previously with both drugs has been shown to produce fewer lung nodules than control cells (Hart et aI., 1980). Cycloheximide is one of the agents that causes the alteration of cytoskeletal filament organization, and the cells pretreated with cycloheximide have diminished capacity to form lung metastases when injected in mice (Ben-Ze'ev and Raz, 1985). Therefore, study of the cytoskeletal proteins may be an important course for better understanding tumor transformation and metastasis. 24 1.3 Keratins and Tumor Transformation

Keratins are the largest group in the IF family: there are at least 15 different type I acidic keratins (pI< 5.5) and 15 neutral-basic type 1,1 keratins (pI> 6) classified by their sizes, sequence homologies and isoelectric points (Steinert and Roop, 1988). The major difference of keratins that distinguishes them from other members of IFs is that keratins are always expressed as ht;teropolymer pairs, consisting of specific type I and type II proteins in vivo (Giudice and Fuchs, 1987). These two types of keratins form a coiled-coiled dimer at a 1: 1 ratio; two dimers then align in antiparallel fashion as a tetramer which serves as a building block for the 10 nm keratin filament formation (Coulombe and Fuchs, 1990). The keratins used to be thought of as relatively stable components of the cytoskeleton due to their "insoluble" property. However, technology using microinjection of biotinylated-keratins into cells revealed that they can incorporate into IF-like networks fairly quickly. This suggests that keratin filaments are indeed very dynamic cytoskeletal elements (Miller et aI., 1991). Indeed, the regulation of keratin filament assembly is probably by a protein phosphorylation mechanism as is the case with vimentin and nuclear lamin (Chou and Omary, 1991; Klymkowsky et aI., 1991). Recently, protein kinase C subspecies E has been identified as the potential candidate that phosphorylates keratins 8 and 18 (Omary et aI., 1992). Keratin filaments usually start at the perinuclear region and extend to the plasma membrane. In epithelial cells, keratin filaments are associated with hemidesmosomes, which are electron dense plaques localized to the cytoplasmic side of the basal membrane (Jones et aI., 1989). In addition, keratins have also been shown to connect to desmosomal proteins, I and II, at the cell-cell adherence junction (Jones and Goldman, 25 1985). An interesting study with mouse epidermal keratinocytes has found that a significant fraction of the integrin 134 subunit is associated with highly purified keratins (Gomez et aI., 1992). Unfortunately, there is still no convincing biochemical evidence to support this ultrastructural connection. Although it is not clear at this time if this association is a direct or indirect relationship, this type of study may be very important for understanding the role of keratins in the signal transduction pathway through integrin receptors.

Keratins are epithelial cell marker(s); the keratin pairs are expressed specifically in different types of epithelial cells and differentiated epidermis (van Muijen et al., 1987a; Fuchs; 1990). For example, the basal cells express keratins 5 and 14, whereas suprabasal cells have keratins 1 and 10. However, keratins 1 and 10 are only found in keratinized epithelium; keratins 3 and 12 are in corneal epithelium; keratins 4 and 14 are in stratified epithelim; keratins 6 and 16 are in hyperproliferated keratinocytes; and keratins 8 and 18 are in simple epithelial cells. The regulation of the cell type specific expression of keratin genes seems to be mainly at the transcriptional level. It has been shown that treatment of cells with 5-azacytidine resulted in the expression of keratins 8 and 18 in fibroblasts, which demonstrated that methylation of keratin genes in non-expressing cells is one of the mechanisms that controls cell-type specific gene expression (Semart et aI., 1986; Knapp and Franke, 1989). Furthermore, there seem to be a cis­ regulatory enhancer element in the 5' upstream region of bovine keratin IV which also plays a role in the cell-type-specific expression (Blessing et aI., 1989). Using a transgenic mouse model, introduction human or into the germ line resulted in a tissue-specific expression pattern. This further demonstrated that the cloned keratin gene sequences contain all cis-acting 26 regulatory elements necessary for appropriate expression (Vassar et aI., 1989; Abe and Oshima, 1990). However, transcription factors are important as well in this matter. In the epidermis, the transcription factor AP2 has been identified as one of the factors that contribute to epidermal-specific expression of the keratin 14 gene (Leask et aI., 1991). Analysis of the keratin 18 gene sequence has shown that there is an enhancer element responsible for c-fos and c-jun activation. Therefore the level of c-fos and c-jun may be the limiting factor in determining the timely expression of keratins in the differentiation of cells (Oshima et aI., 1990). Understanding the regulation of keratin gene expression can help us resolve the problem of how the induction of keratin genes are being altered during malignant transformation.

The critical question being addressed about keratin filaments concerns the determination of their cellular function. Why are there so many different sUbtypes of keratins, and are they functionally distinctive? Unfortunately, the majority of this information is still unknown. Apparently, the IFs are not involved in basic cellular functions necessary for growth and proliferation because cells that do not produce IF can still survive in culture (Venetianer et aI., 1983). Injection of anti­ keratin antibodies into epithelial cells, which disrupts keratin filaments, have no effect on cell shape (Klymkowsky et aI., 1983). Using a gene targeting approach to knock out mouse keratin 8 has shown that it is not required for simple epithelium formation of extraembryonic endoderm (Baribault and Oshima, 1991). However, in the study of the Xenopus oocytes system, keratins seem to play an important role in differentiation. First of all, keratin filaments are found to be associated with maternal mRNA in Xenopus oocytes; this association may regulate the localization and segregation of information during development 27 (Pondel and King, 1988; Klymkowsky et aI., 1991). Injection of keratin antibodies, which disrupts the deep keratin filament system of the embryo, has caused the abnormal or total faiiure of embryo gastrulation (Klymkowsky et al., 1992). Similar results have been obtained using antisense oligonucleotides to the keratins which cause a loss of the epithelial surface of the blastula in the embryo (Torpey et al., 1992). These data suggest that the 'filament system is required for the mechanical integration of the morphogeneic movements at the gastrulation stage in Xenopus. Examples of the best studied keratin function comes from two observations of human autosomal dominant skin diseases: epidermolysis bullosa simplex and epidermolytic hyperkeratosis. They both are characterized as skin blistering from either the basal or suprabasallayer of dermis after trauma. Genetic studies have shown that these patients have gene mutations at , 10 or 14, thus producing collapsed keratin filaments in the epidermis (Coulombe et aI., 1991; Fuchs and Coulombe, 1992; Rothnagel et aI., 1992). These results suggest that keratins are an important component to maintain the integrity of the epithelial sheet upon mechanical assault. Recently, Robey and coworkers (1992) have used chemotaxis chambers and immunohistochemical staining methods to show that in human retinal pigment epithelial cells, only those cells which successfully migrate through polycarbonate filters are keratin positive. They suggested that the expression of keratin(s) is associated with active migration. These results indicated that keratins are not just important for cell structural integrity, but may also playa more complicated role in cellular activities in different cell types other than epidermis.

Although keratins, in general, can be used as markers to determine cell type and differentiation state, exceptions still exist in many tumor cells which may well 28 indicate a more complex control mechanism for keratin expression. Banks­ Schlegel and Rhim (1986) have shown that chemically or virally transformed human keratinocytes express the same types of keratin(s) as normal cells when grown in vitro; however, athymic nude mouse tumors derived from these cells have additional type I keratins and no type II keratins. Other work by Trask and coworkers (1990) have shown that there are different sets of keratins expressed in normal (keratins 5, 6, 7, 14, and 17) vs. tumor-derived mammary epithelial cells (keratins 8, 18, and 19). Using differential screening of a cDNA library has also revealed that keratins 8 and 18 are expressed at high levels in tumorigenic but not in nontumorgenic clones (Modjtahedi et al., 1992). Although not yet clear, the significance in the peculiar switch of keratin expression during neoplastic transformation may be very important for understanding the "key mechanism" toward tumor formation.

1.4 Coexpression of IFs in· Metastatic Tumor Cells

There is an interesting phenomenon of anomalous dual intermediate filament expression in tumor cells, especially in highly metastatic clones. Several separate investigations have reported the expression of both keratin and vimentin in different types of metastatic tumors. For example, a spontaneous, non­ metastasizing rat pancreas adenocarcinoma expresses vimentin filaments; whereas, its metastatic cell variant expresses a complex pattern of keratins in addition to vimentin (Ben-Ze'ev et al., 1986). In a study with murine sarcoma cells, GUnther and colleagues (1984) have found the sarcoma in ascites coexpress keratins and vimentin, but the solid tumor from the same sarcoma expresses only vimentin. In rat ascites hepatoma cell lines, the spontaneously derived metastatic 29 line has both keratin and vimentin; but the low metastatic counterpart expresses only vimentin (Kinjo et al., 1984). In human tumor cells, the same correlation has also been reported. Work done by Ramaekers and others (1983) have shown that solid tumors of epithelial-origin have keratins exclusively; however, when these tumors are present in ascitic or pleural fluid, they also express vimentin. Characterization of a series of human breast cell lines with different invasive abilities have also shown that only the invasive cells have additional vimentin expression not seen in the low invasive lines (Thompson et al., 1992). Along these lines, our laboratory also found the same correlation in several human melanoma cell lines. C8161, established from the abdominal-wall metastases of a melanoma tumor, is a highly invasive and spontaneously metastatic human melanoma cell line. Indirect immunofluorescence staining indicates that these cells have a high level of keratins 8 and 18 in a filamentous form, in addition to its normal vimentin marker. A375M, a derivative from A375 melanoma, has a moderate range of invasive and metastatic potential in vitro and in vivo. These cells contain subpopulations of keratin positive and negative cells. The A375P cell line has the lowest invasive and metastatic ability among these three, and they are negative for keratin filament staining (Appendix A, B). The expression of keratin in melanoma cells is not unique; it has also been reported by others when a series of melanoma cases are scanned (Gatter et aI., 1985; Trejdosiewicz et aI., 1986; Miettinen and Franssila, 1989; Zarbo et aI., 1990), and it appears that there are always keratins 8 and 18 IFs in either highly metastatic or recurrent melanoma tumors.

Whether presence of dual IFs, keratin(s) and vimentin, In the highly metastatic melanoma and other tumor cells is an indication of tumor progression, is 30 still unknown. Further investigation is required to determine if this coexpression can be used as a phenotypic "marker" for the more aggressive and metastatic tumors. This information will be extremely useful clinically for tumor diagnosis; in addition, it will be a milestone for basic cellular research in the understanding of the function(s) of IFs.

1.5 Hypothesis: Dual IF Expression in Highly Metastatic Cells

Based on the previous observations, the expression of IF seems to play an important role in tumor cell behavior, and coexpression of two types of IFs, keratins and vimentin, in tumor cells is one of the characteristics of metastatic tumors. Since the understanding of the function of IFs is still limited, it is difficult to pinpoint at this time which step IFs affect in the invasion process. The questions I addressed for this research were: (1) does the expression of additional IFs in tumor cells give them an invasive phenotype; and (2) how does dual IFs expression affect the invasion process.

The working hypothesis for my research is summarized in Figure 1. The majority of physical responses of cells are dependent on the stimuli that come from the extracellular environment. There are transmembrane receptors, such as growth factor receptors and integrins, on the plasma membrane that can bind to their specific ligands from the environment. These receptors then transfer these messages into the cells by signal transduction pathways so that the cells can make the appropriate response for proliferation, differentiation and biosynthesis. The second messengers in the signaling pathway have been studied in depth; however, the involvement of the cytoskeleton in this action is still unclear. Due to the structural characteristic(s) of cytoskeletal proteins and the fact that they 31

Metallo- . roteinase~ DegradatIon

Cadherin Biosynthesis Integrin -i?;il>t-Attachment Invasion & AMF -fi~-Migration Metastasis AMF-R Differentiation

Cellular Cell Shape Response Proliferation (2) / Second Messenger ~-----liI~~ Cytoskeleton (cAMP, IP3) (la) (Actin, Vimentin, Kerati1l) ---___ +-~_, Plasma Cell Surface Receptor _M_e_m_b_r_a_n_e_~~

Extracellular Growth Extracellular Environment Factor Matrix

Figure 1. Hypothesis: Dual IF Expression and Cellular Response 32

are associated with protein kinases, I propose that they play a role in transferring messages from outside of the cell to the nucleus (1). They may either cooperate with second messengers by providing their cytoplasmic binding substrates (la) or directly interact with nuclear lamins and affect gene expression (lb). By the presence of additional IFs, the balance between stimulation and cellular response can be destroyed and result in the abnormal induction of proteins involved in the invasion process. On the other hand, the cytoskeletal proteins are the major components regulating cell shape and movement. Additional IFs may simply change the dynamics of random cell migration by providing extra force for the movement, thus increasing the invasive ability (2). Although the connection of IFs with integrins is not very clear, the formation of IFs has to somehow connect and end in the plasma membrane site. The additional IFS in the cell surface area may interfere with the normal interaction activities of integrins and with the ECM: an important step in the invasion process (3). Therefore, the coexpression of two different IFs may give tumor cells a selective advantage in their invasive and metastatic abilities compared with the cells that express only one type of IF. This hypothesis was tested, and the final goal of this research focused on defining the functional role(s) of dual IFs in ells with regard to the invasion process.

1.6 Dissertation Format In this chapter, I have explained the current problems and the most recent progress in the field of cancer invasion and metastasis. My research focused on the function of IFs in this aspect. I have provided a brief literature review pertinent to the cytoskeleton, as well as proposed a hypothesis for the role of IFs in invasive tumor cells. In the next Chapter, I will describe the approach and 33 methods I used to resolve the problems; I will also propose a model based on the results obtained from the study to explain the mechanism(s) of IFs in controlling cellular behavior. All of my dissertation work has been either published or submitted for publication in peer-reviewed journals, and the reprints and manuscripts will be included as Appendices of this Dissertation. The four Appendices represent research in three different cell systems. As the major investigator, I am the first author of the papers in Appendices A, C and D. Appendix A is published in Clinical Biotechnology, and this project deals with the disruption of keratin filaments in highly metastatic melanoma tumor cells. Appendix B, in which I have a co-authorship, is published in Journal of National Cancer Institute, and this paper includes the previous study of the identification of dual IFs in melanoma cell lines. My major contribution to this work was in the characterization of these cells and the transfection work. Appendix C is in press in the Proceding of the National Academy of Sciences, USA, and this project focuses on a mouse fibroblast cell model to investigate the role of keratins in this model. Appendix D is a manuscript submitted to Nature, and this work focuses on the induction of keratins in melanoma cells of low invasive potential. 34 CHAPTER 2 PRESENT STUDY

The methods, results, and conclusions of this study are presented in the papers appended to this dissertation. The following is a summary of the most important findings in these papers. In addition, results from some of the unpublished data will also be included.

2.1 Disruption of Keratin Filaments in Highly Metastatic Cells

The evidence that prompted my study was derived from the analysis of a series of human melanoma cel1lines. One interesting finding in this analysis was the correlation between the degree of metastasis and the level of expression of keratins 8 and 18. The three representative lines in this study were: C8161, a highly invasive and metastatic line; A375M, which has medium range in invasive and metastatic ability; and A375P, a low invasive line which does not form lung metastases in nude mice. Immunofluorescence microscopy showed that C8161 are positive for keratins 8 and 18 in addition to vimentin-positive (which is a classical melanoma marker); A375M contains a subpopulation of keratin-positive cells, whereas A375P does not have any keratin filaments. Northern blot, western blot and two-dimentional protein gel analysis corroborated the immunofluorescence result (Appendix B). These data show that only the highly invasive cells coexpress both keratin and vimentin IPs.

The rationale for the first project was: if the expression of keratins playa role in the invasion process of the highly invasive melanoma cells, then the disruption of the endogeneous keratin filaments should reduce their invasive 35 ability. Unfortunately, there is no easy way to inhibit keratin filament formation; most of the drugs and agents that are available do not react with keratins specifically. Microinjection of keratin antibodies is one way to knock out the filaments (Klymkowsky et aI., 1983); however, it is not possible to inhibit keratins in the whole cell population with this method, and the inhibition is only transient. The method which I chose to disrupt the keratin filaments consisted of using a dominant negative mutant keratin cDNA that codes for a truncated keratin protein. Kulesh and coworkers (1989) showed previously that transfection of this cDNA into mouse parietal endodermal cells resulted in a destabilized endogeneous mouse form of keratin proteins, and the filament formation was inhibited.

The plasmid LK442-K18-1070 was a kind gift from Dr. R.G. Oshima of the La Jolla Cancer Research Foundation. The keratin 18 cDNA in this plasmid contains only the first 1070 bps of the sequences, the remaining 275 bps of the 3' coding sequence, as well as the 68 bps of the 3' noncoding sequence; the 16 bps poly(A) tail was deleted by restriction enzyme digestion. The cDNA is under the

control of a human ~-actin promoter, and it codes for a protein containing the first 340 amino acids. The last 50 amino acids of the coil 2 structure as well as the 39- amino-acid tail region were deleted. This plasmid also has a xanthine-guanine phosphoribosyltransferase (gpt) gene for a seletion marker, and a SV 40 intron, poly(A) tail sequence for stable expression in the eukaryotic system (Fig. 2). The plasmid was used to transfect C8161 cells, and the vector pH~-Apr-l-neo (Fig. 2) was also used for transfection control. The transfection efficiency of the tumor cells using calcium phosphate method was very low; however, 14 clones of the 36

Figure 2. Plasmids Used in the Transfection Study

Plasmid pHJ3-Apr-l-neo is the vector used for control transfection. LK442-KI8- 1070 contains the fIrst 1070 bps of keratin 18 cDNA whcih codes for a protein of 340 amino acids. The a-helix coil 2 structure as well as tail region of this truncated keratin 18 were deleted, results in a protein of 38 kDa in size. 37

EcoRI

pHp-Apr-1-neo 10,000 bps Cap site

mtervening sequence Sal I Hind III amHI EcoRI

intervening sequence Hind III LK442-K18-1070 11,316 bps

BamHI

a-helix coil [::3~:::[i:::J:ll::;:·I@:···.·········1 N'i head I 1A 1B 2A ~C' K18-340 340 amino acid 38 experimental group (CI070) and 18 clones of the control group (C-NEO) that survived the selection media were able to be isolated for further study.

Not all the transfected CI070 cells expressed mutant keratins; however, two of the clones did show aberrant keratin filament staining by immunofluorescence study. Immunoprecipitation of the keratin 18 protein showed that the transfected mutant keratin DNA was expressed as a 38 kDa­ keratin protein in CI070 cells. The presence of this truncated keratin 18 interferred with normal keratin filament formation. Approximately 60% of the CI070 cells showed that their filament bundles were disrupted and appeared discontinuous in certain regions, demonstrating a striated-like pattern (Appendix A, B). This immunofluorescence staining of keratin filaments in one of the CI070 clone (CI070-10) was further analyzed with confocal microscopy to derive a more detailed three-dimensional image (Fig. 3). How the mutant keratins caused this filament pattern is not known; it may have resulted from failure in the coiled­ coil paring ability of truncated keratin 18 with the keratin 8 protein. However, the presence of truncated keratin proteins in CI070 cells did not affect the endogenous vimentin filamentous structure (data not shown). These cells portrayed a dramatically changed morphology compared to the parental and control cells; they became enlarged with irregular shape. Some cells displayed multipseudopodia and cytoplasmic processes; some cells had multiple nuclei resulting from uneven cytoplasmic separation. Staining of the cells with bromo­ deoxyurodine indicated that the cells still maintained their nucleotides metabolism (data not shown), however, the doubling time of CI070 cells increased from 24 to 39

Figure 3. Confocal Microscopy of CI070 Cells

CI070 cells were seeded onto glass covers lips to 70 % confluence, and then fixed in cold methanol for 7 min. Subsequently, the cells were treated with mouse antibodies to human keratin 18 (CK5; Sigma, MO), for 1 hr at room temperature, followed by rinsing in PBS, and incubation with rhodamine conjugated secondary antibody (Organon Teknika, PA) for 1 hr. Coverslips were then rinsed in PBS and mounted onto a glass microscope slide using gelvatol. The immunofluorescence staining was visualized by confocal microscopy. 40 41

72 hr (Appendix A). These results indicated that keratin filaments may playa role in maintaining cell shape and cell division as well as regulating normal cell cycle.

The most intriguing results in this project came from the functional analysis. The cells were tested for their ability to invade Matrigel-coated filters in the MICS (Membrane Invasion Culture System; Hendrix et al., 1987) assay; and it showed that the C1070 cells exhibited a dramatic decrease in their ability to invade. In addition, when these cells were tested for metastatic potential via subcutaneous injection in nude mice, they showed no primary tumors and no metastases (Appendix B). Further efforts were focused on identifying the mechanism(s) responsible for this decreased invasive ability of C1070 cells. The data suggested that the major change induced by keratin IF disruption was related to the migration behavior; C1070 cells showed decreased motility, as measured by their ability to move through gelatin-coated polycarbonate filters (Appendix B). In addition, microcinematography was used to measure the lateral translocation of cells across a plastic substratum also indicated a slower movement in the keratin disrupted melanoma cells compared with parental cells (Table 1). However, in an attempt to draw a correlation between the degradation of the ECM and invasive ability, there was no difference between C1070 and C­ NEO control cells in the message level as well as enzymatic activity of the metastasis-associated metalloproteinase type IV collagenase (Appendix B). This negative result further supported our contention that the difference observed in invasive and migratory ability was associated with motility properties.

In conclusion, this study provided evidence for the role of keratins in tumor cell invasion by demonstrating that the disruption of the keratin IF 42

Table 1. Migration Rate Measured By Video Cinematography*

C8161 2.5 Jlm/hr

C1070-10 1.0 Jlm/hr

* Cells were plated on plastic dishes at low density, and after 2 hr, the cells were videotaped continuously for 3 hr using an

Olympus IMT-2 inverted microscope equipped with Hoffman Optics (Hoffman Optics, Inc., NY) within an environmental

chamber. Cell movement was analyzed by tracing each cell

outlines in a particular field (7-27 cells) at 10 min interval time

points. The average speed of cell movement was measured as !Jll1/hr. 43 structure of a highly metastatic melanoma cell line resulted in a decrease in its migratory and invasive potential.

2.2 Mouse L Cell Model: Effects of Dual IF Expression

To further elucidate the importance of dual IPs in cell behavior, I used a non-tumorigenic model system -- mouse L cells, to address this question. Mouse L cells are fibroblasts that express vimentin filaments. These cells have been transfected with human keratins 8 and 18 DNAs, and their regulation of expression has been studied in detail (Kulesh et al., 1989). The reason for testing the hypothesis in a non-cancerous cell model was to avoid the complexity and heterogeneity of tumor cells so that the study could be focused specifically on the functional role of dual IFs. The rationale for this project was: if the coexpression of IFs plays a role in the metastatic ability of tumor cells, then mouse L cells (which already have vimentin) transfected with additional keratin filaments should also have increased migratory and invasive ability.

The four cell lines used in this study were obtained from Dr. R.G. Oshima. They were: L -- parental cell line; LK8 -- the L cells received the plasmid LK442- K8 containing keratin 8 cDNA; LK18 -- L cells that were co-transfected with vectors containing a neomycin resistance marker and a keratin 18 gene; and LKI8+K8 -- the LK18 transfected again with the LK442-K8 plasmid. Immunofluorescence staining of these celllines showed the parental L, LK8 and LK18 had no keratin filaments, indicating a single type of keratin alone cannot form filaments in the cells. The LK18+K8 cells demonstrated a complete filamentous keratin network that extended from the nuclear membrane to the 44 plasma membrane (Appendix C). Thus, this LK1S+KS cell line has both vimentin and keratin which mimics the IF pattern of the highly invasive CS161 melanoma cells described previously.

Interestingly, the analysis of invasive ability of these four cell lines indicated that LK1S+KS showed the highest ability to invade Matrigel-coated membrane(s). Furthennore, this cell line also had higher migratory ability through gelatin-coated filters compared to L, LKS and LKIS cells (Appendix C). To further investigate the correlation between dual IF expression and migratory behavior, I used a heterogeneous LK1S+KS population for a selection study. It appeared that after long-term culturing, LK1S+KS cells began to lose their keratin filaments, which is likely due to rearrangement and loss of one of either keratin vector(s). However, this phenomonen pennitted me to further test the hypothesis that keratin-positive transfected L cells are more invasive and migratory. The heterogeneous LK1S+KS cells were seeded into a MegaMICS chamber (Seftor et al., 1990), and the cells that migrated through the filter were collected, expanded, and selected sequentially through MegaMICS for a total of three rounds. The number of keratin-positive cells associated with each round was visually assessed with a fluorescence microscope, and their migratory ability was measured simultaneously. Interestingly, the keratin positivity was enhanced via each selection, resulting in a linear correlation between the number of keratin-positive cells and their migratory ability (Appendix C). This result provided solid evidence that keratins are associated with the increased migratory behavior of the cells.

In order to understand how the presence of normal keratin filaments in cells already expressing vimentin might contribute to increased invasive and 45 migratory ability, I tested a few factors which have been shown to playa role in these processes. First, I tested the ability of the transfected cells to attach to a Matrigel-coated membrane using 35S-methionine labelled cells. In an one-hour assay, L, LK8, LK18 and LKI8+K8 all showed approximately 60-67% of the cells attached to the matrix, with no significant difference among the four cell lines (Fig. 4). I next examined the spreading ability of the cells on the Matrigel matrix. Comparing the difference between Land LKI8+K8 with regard to the time required for the cells to spread on the substratum, LKI8+K8 were retarded in their ability to spread, and the cells maintained a more rounded morphology for a longer period. After 24 h on Matrigel, the LKI8+K8 cells are approximately 80% round in morphology, while the L cells are approximately 50% round. By 36 h, the roundness factor has diminished to 54.2% for LKI8+K8 cells and 43.1 % for L cells; and by 72 h, little difference in shape exists (18.2% for L cells vs. 26.2% for LKI8+K8 cells; Appendix C). I have also investigated whether the actin filamentous architecture may have been affected using NBD-phallacidin. The results showed that LKI8+K8 cells had diffuse actin staining primarily localized in the cortical region when the cells were seeded on Matrigel. On the other hand, L, LK8 and LK18 cells all had similar distinguished stress fiber staining. Moreover, when the cells were seeded on glass, there was no difference in the actin pattern among the four cell lines (Fig. 5). Interestingly, it has been shown that mutant cells with impaired spreading on fibronectin have decreased numbers of stress fibers (Joseph et aI., 1990). In addition, low-metastatic tumor cells usually contain tightly packed actin bundles, and high-metastatic variants have very little F-actin (Volk et aI., 1984). This also seems to be the case for LKI8+K8 which showed very poor actin organization when seeded on Matrigel matrix. 46

Figure 4. Attachment Ability of Cells on Matrigel

35 S-methionine labeled cells were seeded onto Matrigel-coated filters in the MICS chamber. After 1 hr incubation at 37oC, the attached cells were lysed with 2% SDS directly on the filters. The attachment ability was measured as the number of radioactive counts of the lysates compared to the original counts loaded. 47

Attachment Ability of Cells on Matrigel

80~------~

60

40

20

L LK8 LK18 LK18+K8 48

Figure 5. Fluorescence Localization of F-actin in L Cells

L, LK8, LK18 and LKI8+K8 cells were grown on either glass (G) or Matrigel­ coated (M) covers lips for 30 hr. For actin staining, cells were fixed with 3.7% formaldehyde for 10 min, and permeabilized for 5 min in 50% acetone/water. NBD-phallacidin (Molecular Probes, OR) was used to stain the F-actin as per the manufacturer's direction. The fluorescence was observed with a Zeiss standard 18 fluorescence microscope. (xI890) 49

L

LK8

LK18

LK18+K8 50

I hypothesized that additional keratin filaments in the cells would induce the signal transduction pathway which leads to the synthesis of proteins involved in the dynamics of invasion and migration. However, Western blot analysis of the Gai protein in the four cell lines indicated that there was no difference in the protein level among these lines, nor was there any induction of type IV collagenase message in the LK18+K8 cells. Activity analysis of other degradation enzymes, such as urokinase (uPA) and gelatinolytic enzymes, also showed similar levels in the cells. In addition, the protein level and cellular distribution of autocrine motility factor receptor were analyzed, unfortunately, these results did not correspond to the induced migratory ability seen in LK18+K8 cells. Although there are far more factors to be tested regarding the involvement of keratins in the signaling pathway(s), this does not seem likely at this point to be a key mechanism involved in the high migratory and invasive abilities ofLK18+K8 cells.

In summary, the mouse L cell model was used to determine if a direct correlation exists between dual IF expression and migration/invasion. L cells with both vimentin and keratin filaments (LK18+K8) showed a higher ability of invasion and migration compared to cells with only one type of filament. This result was strengthen by the enrichment of keratin-positive cells in a migration selection assay. This observation correlated with the spreading ability and actin organization of the cells on Matrigel matrix. These data suggest that keratins play an important role(s) in the interactions of cells with the extracellular environment, which then effects their migratory behavior. 51 2.3 Transfection of Keratins in Low Metastatic Melanoma Cells

Since the mouse L cell model has demonstrated a correlation between dual IF and cell migratory and invasive behavior, it is interesting to see if the same correlation existed in a cancer cell system. In this project, the hypothesis was tested in a low invasive and keratin-negative human melanoma cell line (A375P). The rationale of this study was: if a non-keratin expressing melanoma cell line is induced to overexpress keratins 8 and 18, the resulting cells will display of mesenchymal (vimentin) and epithelial (keratin) phenotype, a characteristic of several reported highly invasive tumor cells, therefore, these cells should have a selective advantage in their migratory and invasive ability.

The plasmids used in the transfection work are shown in Figure 6. Both keratins 8 and 18 cDNAs are controlled by the human p-actin promoter, and the respective vector contains different selection marker (neor vs. gpt). Originally, the cells were cotransfected with LK442-K8 and LK444-KI8 at a 1:1 ratio, and selected with media containing 0418 (for neor marker) and mycophenolic acid (for gpt marker). Unfortunately, no clones could be isolated which might be due to the high toxicity of mycophenolic acid to the cells. Therefore, the transfection was modified to use a 5:1 ratio of LK442-K8 to LK444-KI8, and selected with media containing only 0418. The control transfection was performed using pHP­ Apr-l-neo vector as described previously (Fig. 2) .. Several clones were isolated and expanded from each transfection. DNA slot blot analysis of the integration of the plasmid DNAs was performed using the probes that recognize either gpt or neor sequences. One of the clones -- PK-14, showed the highest copy number (approximately 10 copies) for both of the transfected plasmid DNAs compared to 52

Figure 6. Plasmids Containing Keratins 8 and 18 cDNAs

Human keratin 8 cDNA was inserted in the BamH I site of the vector which contains E. coli gpt (xanthine-guanine phosphoribosyltransferase) selection marker (LK442-K8). Keratin 18 cDNA was ligated to the Hind III site of the vector that has a neomycinr sequence (LK444-KI8). The expression of inserted DNAs is controlled by a human f3-actin promoter. 53

EcoRI

LK442-K8 BamHI

EcoRI

LK444-K18 Hind ill 54

the others (Fig. 7). Immunofluorescence microscopy showed that PK-14 cells are positive for keratin filaments in addition to vimentin. However, all other experimental clones have either minimum or negative for keratin staining (Appendix D).

The ability of these clones to migrate· through gelatin-coated filters, and to invade through Matrigel coated filters were analyzed, and there was a two-to­ three fold increase in PK-14 compared to P-NEO control cells. To further analyze the migratory ability of the cells on various substrata, video-cinematography was performed on cells seeded onto plastic, gelatin or Matrigel matrix. The data demonstrated that the migration rate of PK-14 cells were higher than P-NEO cells when plated on gelatin and Matrigel; however, no significant difference was seen when they were seeded onto a tissue culture plastic (Appendix D). I next measured the differential ability ofP-NEO and PK-14 cells to spread on a Matrigel matrix. At the early time points of 2 and 4 hr after initial seeding, both cell lines were still rounded; however, a dramatic difference was seen at the 6 and 8 hr time intervals, where the PK-14 cells maintained a rounded morphology compared with the better spread P-NEO control cells (Appendix D). This observation was similar to the previously described L cell model in which the coexpressing of keratin and vimentin in the cells resulted in a retarded ability to spread on a Matrigel matrix. In order to better understand the mechanism(s) contributing to this differential spreading ability on Matrigel, the pattern of integrins localization was measured using indirect immunofluorescence microscopy. Interestingly, during the 6-8 hr spreading period, there were no detectable uv~3, u3 and (16 integrins in the PK-14 cells focal contacts, which serve as receptors for 55

Figure 7. DNA Slot Blot Analysis of Transfected A375P Cells

Total genomic DNAs were isolated and loaded (2 Jlg/slot) onto nitrocellulose membranes along with the DNAs used for copy number standard. The membranes were baked at 800 C for 2 hr and then hybridized to the probes made from the nick translated neo or gpt DNA sequences radiolabeled with 32p-dCTP (lCN,CA). PK-14 I - IA375P PK-17 P-NEO-9 Z ~ PK-3 B -- I P-NEO-23 0 copy 50 r. 100 • -1:0 copy

PK-14 A375P PK-17 P-NEO-9 C1 P'C PK-3 P-NEO-23 ~ 50 copy I ~- 11 copy 100 10

VI 0'\ 57

vitronectin, collagen and laminin, respectively, comprising Matrigel. However, these integrins are all present in the focal contacts of the P-NEO cells (Appendix D).

The results obtained in this study correlated with the results from the mouse L model, therefore, the effect of a "coexpression phenotype" in invasion and migration is likely to be a common phenomenon in non-tumorigenic and tumorigenic cell systems. These observations provided strong evidence supporting the hypothesis that coexpression of vimentin and keratins directly results in increased migratory behavior, commensurate with a diminution in spreading ability on ECMs. Most importantly, the data suggest that the mechanism responsible for differential spreading ability rests in the unique regulation of specific integrin(s) associated with focal contacts, acting either directly or indirectly, with the IFs and with ECM molecules.

2.4 Conclusion

The IFs have long been considered as structural proteins that play an important role in maintaining cellular morphology and integrity. The cell type­ specific expression of IFs is used as a marker in differentiation and pathology. However, recent studies have shown that aberrant IF expression in tumor cells correlates with tumor progression. Hence, it is very important to elucidate the role(s) of IFs in the malignant progression of various cancers for the hopeful development of new strategies for cancer diagnosis and treatment,

I have studied the involvement of dual IFs -- vimentin and keratin, in the process of tumor cell invasion and metastasis. My hypothesis was that the ability 58 to coexpress both vimentin and keratins offers a selective advantage to tumor cells in the invasion and metastasis processes compared with cells which have only one IF phenotype. Beginning with a negative approach, I used a dominant negative mutant keratin 18 DNA to disrupt the keratin filaments in a highly metastatic, dual IF expressing melanoma cell line (C8161). The resulting transfectants (CI070) demonstrated a striated- and collapsed-type of keratin IFs, concomitant with a decrease in their migratory and invasive behavior, as well as a complete abrogation of metastatic potential. On the other hand, transfection of keratin-negative, vimentin-positive mouse L cells or low invasive A375P melanoma cells with keratin 8 and 18 DNAs resulted in a dual IF expressing "phenotype" in the transfected clones (LKI8+K8 and PK-14). These cells showed a two-to-three fold increase in their migratory and invasive ability, thereby providing direct evidence that the dual IF phenotype is involved in the process of migration and invasion. However, the question remains as to whether the increase in invasive and migratory activity associated with dual IF expression is due to the combination of vimentin plus keratins 8 and 18, or is due to higher levels of IPs. This consideration is probably unlikely due to the natural in situ occurrence of vimentin with keratins 8 and 18 in several tumor cell lines and developing embryos associated with the migratory phenotype.

The manipulation of keratin filaments in the cells tested does not appear to alter the synthesis of degradative enzymes. The analyses of type IV collagenase, stromelysin and urokinase have indicated that neither message levels nor protein activities are changed by the disruption or induction of keratins in the cells. In addition, the G protein and autocrine motility factor receptor, which are involved in the signal transduction pathway and random cell motility, respectively, did not 59 show any difference in the control and experimental clones. Therefore, it is unlikely that dual IFs induce the migration and invasion of the cells by altering the signaling pathway and gene expression of these related proteins (Fig. 1-1). However, the dual IF expressing cells have a higher migratory ability through polycarbonate filters containing 10 /-lm size pores. This indicated that these cells have greater deformability allowing them to move through the pores that are smaller than their individual cell size. Indeed, it has been reported that there is a correlation between cell deformability and metastatic potential of tumor cells (Ochalek et at, 1988), which may be pertinent to the highly invasive and migratory keratin- and vimentin-positive cells in my study. How is cell deformability being changed by the additional IFs is not known, yet it can be speculated that IFs may change the dynamics of cell shape alteration, thereby affecting the deform ability of the cells (Fig. 1-2).

The major difference between the cells containing single and double IF is their spreading ability on an extracellular matrix. Unpublished data from my laboratory indicates that the highly invasive MDA-MB-231 breast cancer cells have the "coexpression phenotype", and show a decrease in their ability to spread on Matrigel compared with keratin-positive, vimentin-negative, and low invasive MCF-7 cells. Therefore, the spreading phenomenon has been observed in three different cell systems -- transfected mouse L cells, human melanoma cells, and breast cancer cells. This spreading activity requires the interaction of integrin, as well as nonintegrin, receptors with matrix components. Hence, it can be postulated that additional keratin filaments in the cells may abrogate this interaction, so that cell spreading cannot be achieved normally. Moreover, there are no detectable levels of

Coexpression of IFs has been reported in numerous malignant cells as well as in fetal tissue. It seems to be a general phenomenon for the developing embryonic cells to express two or even three different types if IPs (van Muijen, 61

Vimentin Keratin L cell o o Receptors Basement Membrane ~ + keratins 8,18

Lcell LK18+K8

receptors "fixed" by keratins Basement Membrane =='t}

Spreading Migration Invasion

Figure 8. Model for Dual IF Expression Augmenting Motility 62

1987b; Markl, 1991). Furthermore, there is a transient expression of keratins by mesenchymal cells of regenerating newt limb, specifically in the undifferentiated progenitor cells which give rise to the new tissue (Ferretti et al., 1989). Therefore, it is tempting to speculate that the coexpression of dual IFs represents a phenotype which is functionally important for the process of embryogenesis and tissue morphogenesis. The simple epithelial keratins 8 and 18 are the first IF pairs synthesized during development (Franke et al., 1982); hence, they are the major keratin sUbtypes that coexpress with other IFs in the embryo. The occurence of keratins 8 and 18 in the highly neoplastic and metastatic human melanoma cells may very well represent a "dedifferentiated" phenotype, and their behavior can be interpretated as reversion to an unique "embryonic" cell type. Lane and colleagues (1983) have reported that coexpression of keratin and vimentin in situ in the parietal endoderm of the 8.5-13.5 days old mouse embryo. They suggest that this phenotype is involved in the ability of cells to detach from cell-cell contacts and migrate away from the epithelial sheet. It is very likely that metastatic tumor cells are derived from the cells which obtain additional IFs; hence, they retain the ability to move away from the primary tumor mass.

This project has successfuly demonstrated that dual IF coexpression of vimentin and keratins leads to a phenotype with greater migratory and invasive ability than cells expressing only one IF. This research not only has a great impact on the clinical diagnosis of cancer cells, but it also provides evidence that suggests a novel function of IFs with regard to the interaction with cell surface receptors and the extracellular environment. 63 64

I hereby grant full pennission to Yi-Wen Chu to include the article published in Clinical Biotechnology 3: 27-33, 1991, in her dissertation.

7Y[avr ~. c. Irkntki= Mary J.C. Hendrix, Ph.D. 65

Volume 3, Number 1 Spring 1991

Linkjng basic bjotechnology research and th e cHnka] sci el1ces BIOTECHNOLOGY

Transplantation of Embryonic Yolk Cells 66

Volume 3 CLINICAL Number 1 BIOTECHNOLOGY Spring 1991

From the Editors iii

Currents In Clinical Biotechnology Tumor-Suppressor Genes-The Retinoblastoma Protein as a Paradigm 3 Review Molecular Diagnostics of DuchenneIBecker Muscular Dystrophy 7 Basil T. Da"as Research Articles Culture and Successful Transplantation of Embryonic Yolk-Sac Cells 15 Barbara J. Com, Michael A. Reed, Sharon L. Dishong, Yunsheng Li, and Thomas E. Wagner Transplantation of Allogeneic Hepatocytes into LDL Receptor Deficient Rabbits Leads to Tran- sient Improvement in Hypercholesterolemia 21 James M. Wilson, Namila R. Chowdllllry, Mariann Grossman, SanJeev Gupta, Jasmine Jeffers, Tian-JIIIJ lIuang, and Jayanta R. Chowdhury Transfection of a Deleted CK18 cDNA into a Highly Metastatic Melanoma Cell Line Decreases the Invasive Potential 27 y,. Wen Chu, John J. Duffy, Richard E. B. Seftor, Raymond B. Nagle, and Mary J. C. Hendrix Interleukill-4-Activated Tumor Inhibition: A T-Lymphocyte-Driven but not Fully T-Lymphocyte- Mediated Antitumor Reactivity 35 Maria Carla Bosco, Andrea Modesti, Mirella Giovarelli, Laura Masue//i, Federica Perie/e, Federica Cavallo, and Guido Forni Peptide Growth and Necrosis Immunomodulators 39 William S. Lynn, James C. Wallwork, and DalTCn Mathews Detection of Antigenuria with Monoclonal Antibodies in Typhoid Fever Patients 47 Natesan Mohan, Maharaj K. Bhan, Ravinder Kumar, and Ramesh Kumar Methodology Practlcum Centrifiltration: A Rapid Method for Changing Buffers that Maintains Enzyme Activity and Concentration 53 Catherine E. Erickson and Frank J. Castora New Products 57 Erratum 61 Instructions for Authors 65

Conr: Expansion and growlh of murine yolk,sDc cell' afler aUachmenl 10 Ihe feeder Ilyer. (See p. 17.)

o Now York 67

RESEARCH ARTICLE

Transiection of a Deleted CK18 eDNA into a Highly Metastatic Melanoma Cell line Decreases the Invasive Potential

Vi-Wen Chu*, John J. Duffy*, Richard E.B. Seftort, Raymond B. Nagle*, t and Mary J.C. Hendrlx §

·Radiation Oncology. Arizona Cancer Center. Tucson. AZ. 85724 'Department of Anatomy. University of Arizona. Tucson. AZ. 85724 'Department of Pathology. University of Arizona. Tucson. AZ. 85724

A highly invasive and metastatic human melanoma cell invasion" with associated predictions for survival line (C8161) has been shown to express high levels of (Clark, 1967). Unfortunately, little is known about the (CK) 8 and 18, which does not occur in tbe mechanisms responsible for cancer cell invasion and majority of melanomas. To study a potential relationsbip metastasis during tumor cell progression. Liotta and between the CK expression and tbe invasive ability of coworkers (1986,1987) proposed a three·step hypothesis C8161 cells, a mutant CK18 cDNA tbat contains only tbe describing the biological events of tumor cells' interac­ first 5' 1070 nucleotides was transfected into these cells. A tion with extracellular matrices during metastasis: (a) selected clone, CI070·10, sbowed a dramatic morpbolog· tumor cell attachment to a matrix substratum, (b) local ical change, and its growth rate decreased compared witb degradation of the matrix by tumor cell-secreted prote­ control cells. Indirect immunofluorescence microscopy, ases, and (c) tumor cell locomotion into the modified using an antibody against CK18, demonstrated all intra· matrix. cellular stnining pattern of sbort, discontinuous striated Many human melanoma cell lines arc available for the bundles in 60% of CI070·10 cells. The invasive ability of study of these mechanisms. In our laboratory we have tbe cells was also tested in tbe Membrane Invasion Cui· studied in depth three different melanoma cell lines that ture System using reconstituted basement membrane· differ greatly in their invasive potential in vitro and coated filters. After a 72·br incubation, CI070·10 cells metastatic ability in vivo (Hendrix et aI., 1989). One showed a 3.3·fold decrease in invasion, which suggests an interesting finding is the correlation between the dl!gree important regulatory role for proteins in of metastasis and the expression of cytokeratins (CK) 8 tumor cell invasion. and 18 (Hendrix et aI., 1990). For instance, poorly metastatic melanoma cells show barely detectable lev­ INTRODUCTION els of CKs 8 and 18; in contrast, highly metastatic Melanoma is all illsidious callcer found mostly ill the melanomas demonstrate higher amounts of CKs 8 and southwestern United States and in Australia and New 18 in the RNA and protein levels. Cytokeratins arc part Zealand. It is widely recognized to be increasing in of the intermediate filament family, the principal com­ incidence across the globe (Rigel et aI., 1987). One ponents of the cytoskeleton of mammalian cells (Nagle, major problem in the diagnosis and prognosis of this 1988; Moll et aI., 1982; Steinert and Roop, 1988). There cancer is the highly metastatic propensity of the malig. are at least 19 separate CKs ranging in size from 40 to 68 nant tumor cells. Melanoma has been categorized into kDa that can be further subdivided into two groups five levels of histological distinction based on "levels of based on the difference in their isoelectric focusing point: type I (acidic) and type II (neutral-basic) keratins. Cytokeratins are commonly expressed as particular ~Corresponding author pairs, consisting of type I and type II, which form

Volume 3. Number 1. Spring 1991 CLINICAL BIOTECHNOLOGY 27 68

heteropolymers in the cells. For example. CK18 (type I) Eb, 1973; Kingston et aI., 1984). The positive colonies is normally coexpressed with CK8 (type II). Cytokera­ were selected according to their ability to grow in tins are generally thought to have a structural role; other DMEM media containing either 25JLg!ml mycophenolic functions and intracellular involvement are still unclear. acid, 2JLg!ml aminopterin, 15JLg!ml hypoxanthine, Intermediate filaments are usually expressed in a cell­ 250JLg!ml xanthine, 1OJLg!ml thymidine, and 25JLg!ml type-specific manner, and CKs are recognized as a adenine (Sigma, MO) for the gpt-phenotype or 0.8JLg!ml marker for the origin of epithelial cells (Osborn and G418 (Gibco, NY) for the neo-phenotype. Weber, 1982). However, in addition to recent reports (Hendrix et aI., 1990; Zarbo et aI., 1990), the "unusual" DNA IsolatIon and Slot Blot expression of CKs in malignant melanoma (which nor­ Genomic DNA from transfected cells was isolated and mally expresses vimentin) was also found by others subsequently loaded (2JLg) onto a nitrocellulose mem­ (Gatter et al., 1985; Miettinen and Franssila, 1989; brane using a Minifold II slot blot apparatus (Schleicher Trejdosiewicz et al., 1986). & Schuell, NH). The membrane was baked at 80°C for Previously, Kulesh and coworkers (1989) showed that 2 hr and then hybridized to the probe made from the transfection of mouse parietal endodermal cells with a nick translated gpt DNA sequence radiolabeled with cDNA coding for a truncated CK18 protein destabilized 32p_dCTP (ICN, CA). the endogenous mouse forms of CK8 and CK18 proteins and inhibited the keratin filament formation. To inves­ Immunofluorescence StaInIng tigate whether highly expressed CKs play any function­ Cells were seeded onto glass coverslips to 70% conflu­ al role(s) in contributing to the invasiveness of a highly ence. Subsequently, the cells were rinsed three times metastatic melanoma cell line (C8161), we transfected a with PBS and fIXed in cold methanol for 5 min. Before deleted human CK18 cDNA (which contains only the being stained, cells were permeabilized with 50% ace­ first 5' 1070bps of the full length) into these cells. Our tonelwater and air dried. Indirect immunofluorescence results indicate that the expression of this mutant CK18 staining was performed using a mouse antihuman CK18 in melanoma cells disrupted the normal intermediate antibody, CK-5 (ICN, IL), at 1:50 dilution in PBS, filament structure and subsequently decreased the inva­ followed by a rhodamine conjugated rabbit antimouse sive ability of the C8161 cells in vitro. IgG (Organon Teknika, PA). The fluorescence staining was observed with a Zeiss standard 18 fluorescence MATERIALS AND METHODS microscope with an automatic rhodamine filter set. Photographs of control and experimental samples were MaIntenance of Cell Cultures exposed for equal time using the manual exposure mode C8161 is a highly metastatic human melanoma cell line on the camera. established from an amelanotic metastasis from a pa­ tient at the Arizona Cancer Center, Tucson, AZ (Breg­ Immunopreclpltatlon and Protein Gel man et al., 1983). The cells were maintained in Dulbec­ Electrophoresis co's modified Eagle's medium (DMEM) (Irvine Scien­ Cells were labeled with 35S-methionine (Arnersham, IL) tific, CA) containing a 10%, 1: 1 (v/v) ratio of fetal at 5JLCi/ml for 48 hr. The lysate preparation and immu­ bovine serum (FBS) (Irvine Scientific) to NuSerum noprecipitation procedures were performed as previ­ (Collaborative Research, MA) supplemented with 0.1% ously described (Kulesh and Oshima, 1988; Oshima, Gentamycin Sulfate (Gibco, NY). The cells were nega­ 1982). Antiserum to murine K18 was used to precipitate tive for mycoplasma (as monitored using the Fisher the K18 and truncated K18 cytokeratins. Proteins were Gen-Probe detection kit) and grown at 37°C in a humid­ separated on a 12.5% SDS/PAGE* as described by ified 5% CO2, 95% air atmosphere. Laemmli (1970). The gel was dried for 2 hr using a vacuum slab gel dryer (Bio-Rad, CA), then placed Transfectlon of CB161 Cells against Kodak XARS X-ray film at -80°C. Plasmid LK442-KI8-1070 was a kind gift from Dr. R.G. Oshima of the La Jolla Cancer Research Foundation. Invasion Assay This plasmid contains a deleted human CK18 cDNA The Membrane Invasion Culture System (MICS) (Hen­ under the control of a human J3-actin promoter (Kulesh drix et al., 1987) was used to measure cell invasion. et al., 1989). It also carries the gene that codes for E. Briefly, a reconstituted basement membrane gel (matri­ coli xanthine-guanine phosphoribosyltransferase (gpt) gel) (Collaborative Research, MA) was coated on a as a selection ma~ker. Plasmid pHf3Apr-l-neo (Gunning nucleopore filter containing 10JLm size pores (Nucle­ et al., 1987) harboring a gene for resistance to the apore, CA) and placed between the upper and lower antibiotic G418 was obtained from Dr. J. Leavitt (Linus well plates of the MICS chamber. (Matrigel is composed Pauling Institute of Science and Medicine, Palo Alto, of type IV collagen, laminin, entactin, and heparan CA) and was used for a control transfection. The plas­ sulfate proteoglycans (Kleinman et aI., 1982).) Subse­ mids were introduced into C8161 cells using the calcium quently,5 x lat cells were seeded into the upper wells, phosphate precipitation method (Graham and van der and after incubation for 72 hr at 37°C, cells that had

2B CLINICAL BIOTECHNOLOGY Volume 3, Number 1, Spring 1991 69

invaded the mat rigel-coated filters were harvested from sequence is the same as gpt in the CJ070-1O cells. the lower wells and counted_ Invasion potential was The morphology of this clone changed dramatically calculated as the percentage of cells that invaded the after three passages in culture. Fig. 2 shows the phase matrigel compared with the total number of cells seed­ contrast microscopy of the parental C8161 cells. C1070- ed. taking into account the proliferative ability of the 10 cells. and the cells transfected with the vector only as cells attached to the upper and lower surfaces of the the control (C-NEO). Since the C-NEO cells showed no filters. difference in their size or morphology compared with the parental cell line, the transfection procedure itself RESULTS was not responsible for alterations in their morpholog­ The deleted CKI8 eDNA that was used for the trans­ ical phenotype. C1070-10 cells demonstrated a normal fection study retained 1070 bps from the eDNA se­ cell morphology in the first three passages but then quence. The remainder of the 275bps of the 3' coding gradually became enlarged with irregular shapes. As sequence. as well as the 68bps of the 3' noncoding these cells elongated. many appeared to have long sequence and the 16bps poly(A) tail, were deleted by restriction-enzyme digestion. The resulting eDNA codes for a protein (KI8-340) containing the first 340 amino acids of the CK18 protein and displays a Mr of 38kDa on 12.5% SDS-PAGE. This protein does not contain the last 50 amino acids of the coil 2 structure nor the 39-amino-acid tail region. The truncated CK18 eDNA sequence was subcloned into a BamHI site of the PHf3Apr-l-gpt plasmid. This plasmid was transcribed from the human f3-actin promoter and contained the SV40 intron and polyadenylation signals_ The E. coli gpt sequence was used as a selection marker. To study the potential role(s) of CK proteins in the invasion process of a highly metastatic cell line, we transfected this mutant CK18 eDNA into C8161 cells. This human melanoma cell line is highly invasive and metastatic and also expresses high levels of CK8 and CK18 proteins. Transfection of 10" C8161 cells resulted in 14 colonies. with one of the clones. CI070-1O. select­ ed for further study. The copy number of the trans­ fected sequence in C1070-1O was investigated by DNA slot blot analysis. The results show that C1070-1O cells contain more than 10 copies of the E. coli gpt sequence (Fig. I). Since the gpt sequence and the K18-1070 sequence were contained in the same plasmid that was used for the transfection (LK442-K18-1070). we con­ clude that the copy number of the K18-1070 cDNA

Fig. 1. DNA slot blot anal· YSls of gpt genomic DNA copy numbers In CB 161 and C 1070·1 0 cells Total C8161 genomic DNA was Isolated and loaded (2"g/sI01) onto a nllrocellulose membrane Wllh the plasmid contain· C1070-10 Ing Ihe gpt sequence act· Ing as the siandard for the copy numbers (1. 10. 50 copies) The blot was hy· bndlzed to a "P·dCTP la· beled gpl DNA probe py :0 )cono.

50 Fig. 2. Phase contrast light microscopy 01 the changes In cell mor· phology with and without transfectlon (A) CB161 cells. (8) C·NEO cells. and (C) C1070·10 cells (x300).

Volume 3, Number 1, Spring 1991 CLINICAL BIOTECHNOLOGY 29 70

cytoplasmic processes or multi pseudopodia with some basement membrane in an in vitro system, MICS. Fig. 6 cells containing multiple nuclei, and other cells showed shows the percentages of invasion of C8161, C-NEO, uneven cytoplasmic separation. The growth rate of and C1070-1O cells. C8161 cells showed an invasion of C1070-1O decreased compared with C8161 and C-NEO, 14.4%, and C-NEO cell invasion was approximately as shown in Fig. 3. The doubling time of C8161 and 10%. C1070-10 cells, however, displayed a 3.0% inva­ C-NEO cells is approximately 24 to 32 hr, whereas the sion potential, which represents a 3.3-fold decrease in C1070-1O cells' doubling time increased to 72 hr. The their invasiveness compared with C-NEO cells. C1070-1O clone cannot be maintained over 11 passages in culture since the cells stop replicating and die. DISCUSSION Immunofluorescence staining of CK intermediate fil­ Many efforts have been made, using different approach­ aments was performed on C8161, C-NEO, and C1070-1O es, to elucidate the molecular mechanisms of tumor cell cells using a monoclonal antibody against human CKI8. invasion and metastasis. In this regard, our study fo­ As shown in Fig. 4, a general filamentous pattern of cused on the expression of cytoskeleton proteins in CKs was observed in both C8161 and C-NEO cells. tumor cells. In addition to their architectural role in However, approximately 60% of the C1070-1O cells maintaining cell shape, the cytoskeletal structures may were very unusually patterned in that the filament bun­ play an important role (or roles) in controlling cell dles were disrupted and appeared discontinuous in cer­ adhesion and penetration of extracellular matrices, as tain regions, demonstrating a striated-like staining. well as migration during tumor cell metastasis. Evi­ Higher magnification showed the striated regions still dence shows that agents which cause the disruption of had filaments, but to a lesser degree. This phenomenon microtubules and microfilaments lead to a decrease in was also observed for two other CK18 antibodies. The metastatic behavior of melanoma cells (Hart et ai., immunofluorescence staining procedure has been care­ 1980). Furthermore, Ben-ze'ev and Raz (1985) used fully examined to ensure that the effect is not an artifact cycloheximide to alter the synthesis and organization of of the procedure (data not shown). vimentin in B16 melanoma cells and observed a dimin­ Previous work (Kulesh et ai., 1989) shows that the ished capacity in forming lung metastases in syngeneic antiserum against murine CK18 recognizes the epitopes mice. Cytokeratins are another type of cytoskeletal both on human CK18 protein and on the truncated protein that seem to associate with the tumorigenicity CKI8. Fig. 5 shows the immunoprecipitated proteins and metastatic ability of cells. Other reports show that from 35S-methionine labeled cell Iysates of C8161 and there is differential expression of CKs in metastatic C1070-1O cells. The full-length CK18 has an apparent tumors of varying metastatic potential, including adeno­ molecular weight of 45kDa, which appears in both carcinoma (Ben-ze'ev et ai., 1986), hepatoma (Kinjo et lanes. The band with the approximate 38kDa molecular ai., 1984), melanoma (Hendrix et ai., 1990; Zarbo et ai., weight corresponds to the expected size of KI8-340, 1990), and tumor-derived mammary epithelial cells which is observed only in CI070-1O. This indicates the (Trask et a1., 1990). transfected exogenous K18-1070 sequence can be ex­ In the present study we transfected a mutant CK18 pressed in the C1070-1O cells producing a truncated CK cDNA sequence that codes for a carboxy-terminal de­ protein. leted protein into a highly metastatic C8161 melanoma To examine if the presence of a truncated CK18 and cell line. This cell line has been carefully characterized the disruption of normal keratin formation in the cells as to its invasive and metastatic potential (Hendrix et affected their invasive potential, the C1070-1O cells ai., 1989,1990), and immunohistochemical analysis were tested for their ability to invade a reconstituted showed that these cells express HLA-ABC, HMB-45, and vimentin, as well as CK8 and CK18 antigens. Furthermore, C8161 cells show a high level of type IV I. 0-0 C8161 e-e collagenase activity and are able to produce a median of 6-6 229 lung metastases within 21 days post inoculation " (i.v.) in nude mice. The transfected CK18 DNA can be 10 transcribed and translated in the C1070-1O cells and produces a protein with an approximate Mr of 38kDa that can be detected by immunoprecipitation using the antiserum against murine CKI8. The data indicate that the K18-340 is present in excess over the normal CK18 • /~ l ___r protein in C1070-1O cells. Previously it was shown that , - ~=; __6--1 ·~f--A overexpression of K18-340 in mouse cells results in the 0~1~--~~~~------~~ increased degradation of murine CK18. This also ap­ o 12 24 48 72 96 pears to be the case here in human cells. HOURS The first change observed in the transfected cells was Fig. 3. Growth curve analysis for C6161, C-NEO, and C1070-10 cells morphological: the cells became more distended with sampled over a 96-hr period. pseudopodia and the cellular volume increased several

30 CLINICAL BIOTECHNOLOGY Volume 3, Number 1, Spring 1991 71

Fig. 4. Indlrccl immunolluorescence de· lectlon 01 human cytokeratin 18 in C8161 cells A (x750) andA' (X1890); C·NEO cells: S (x750) and S' (x1890); and C1070·10 cells C (X750) and C' (x1890). uSing human antlcytokeratin 18 antibody CK5 and rhodamine conju· gated anlimouse IgG.

fold, possibly because of a failure of cell division after may be the difference in the host system. Since we DNA synthesis. Although C1070-1O cells grew more transfected a human cDNA sequence into a human cell slowly than the parental CS161 cells and the control line, the sequence homology between the transfected and transfected C-NEO cells, the relationship between the endogenous CKlS protein may have the advantage of CKs and growth regulation is not known. The expres­ maintaining normal cellular function. The immuno­ sion of K1S-340 protein may have chronic lethal effects precipitation data suggested the most abundant keratin on the cells when accumulated, which results in the species in the C1070-1O cells is the transfected K18-340 cells' inability to be maintained in culture after 11 protein. This protein would likely disrupt the normal passages. Interestingly, although only I clone out of 14 interactions between CK8 and CKI8. However, it is was studied in depth, a second clone was selected for possible that the mutant protein may be incorporated further investigation and died after two passages in into some filaments. Unfortunately, there is no specific culture. antigenic epitope on the K18-340 protein that docs not Immunofluorescence data showed that the transfected appear on the endogenous CKI8. Therefore, it is not cells still maintain the ability to form intermediate fila­ possible to differentiate the localization of K1S-340 pro­ ments, but in a fragmented form. This is different from tein from normal CK18 protein in the cells. The striated the previous report of Kulesh et a!. (1989), which showed pattern seen in C1070-1O cells by the immuno­ that the K18-340 protein expressed in mouse parietal fluorescence staining is unique, but how this pattern was endodermal cells inhibited the formation of typical kera­ formed is not clear. Since the carboxy-terminal deletion tin filament structures formed by the endogenous murine on CK18 includes the most conserved part of the coil 2 CK8 and CKI8. A possible reason for this discrepancy structure that is shared with all intermediate-filament

Volume 3, Number 1, Spring 1991 CLINICAL BIOTECHNOLOGY 31 72

o I o -0 ,..... o CD U u

45kDo - -CKIB

3BkDo - - KIB-340

Fig. 5. Autoradlograph 01 "S·methiomne tabeled cell tysates Immu­ nopreclpltated with antiserum against mUrine cytokeratln 1Band Ffg. 6. Bar graph represenllng percentage of Invasion of CB161. separated by 12.5% SOS-PAGE. The apparent dilference In width of 5 C-NED. and Cl070-10 cells through a mat rigel-coated filter In the the two fanes was the result of foadlng the same number of 3 S counts per tane In dilferent fmal volumes. Arrows indicate the poSitions of MICS. CBI61. 144% =0% (SE). C-NED. 10.0% = 1 71%. and Cl070- cytokerallns lB and K1B-340. 10.30% =1.15%.

proteins and the tail region, it is possible that this ACKNOWLEDGMENTS deletion caused the failure of certain regions on the CK8 This research was supported by a grant from the Eric T. protein to complex with their normal pairing region, Staniek family (MJCH). resulting in the discontinuous filament formation. 1m­ munonuorescence staining with an antibody specific for REFERENCES the CK8 protein, along with the ultrastructural analysis Ben-ze·ev. A. and Raz. A. 1985. Relalionship belWeen lhe organiza­ of these filaments, is necessary to prove this. lion and synlhesis of vimentin and lhe melaslalic capability of BI6 melanoma cells. Cancer Res. 45:~632-2tH I. The most noteworthy finding in our report is the Ben-ze·ev. A .. Zoller. M., and Raz. A. 19H6. Differential expression decreased invasive potential of C1070-1O cells associat­ of inlermediale filamenl proleins in melaslalic and nnnmelaslalic ed with the expression of mutant CK18 protein. The varianlS of lhe SSp73 lumor. Cancer Res_ 46:785-790. Bregman. M_D .• Funk. C .. and Fukushima. N. 1983. {" "ilm modu­ control transfected cells, C-NED, showed a small de­ lalion of human and murine melanoma growlh by proslanoid ana­ crease in invasiveness, which suggests that the trans­ logs_ Proslaglandins 27: I 7-26. Clark. W.H. 1967. A classificalion of mali~nalll melanoma in man fection of foreign DNA by the calcium phosphate pro­ correlaled wilh hislogenesis and hiological hehavior. In: Monlagna. cedure may affect their invasion in vitro to some degree. W. and lIu. F. (cds. I. The Pigmentary Syslem. pp. 621-tH7. However, C1070-1O cells showed a dramatic decrease in Pergamon Press. New York. Galler. K_C.. Ralfkiaer. E_. Skinner. J .. Brown. 0_. Hery.l. A .. their invasive ability from about 14% to 3.00/,: in the Pulford. K.A.F.. Hou-Jensen. K.. and Mason. D_Y. 19~5. An MICS; this is in the range of invasive potential of low to immunoCYlochemical sludy of malignanl melanoma and ils differ­ intermediate metastatic melanoma reported earlier ential diagnosis from Olher malignanl lumors. J. Clin. Pal hoi. 38:1353-1357. (Hendrix et aI., 1989,1990). This dramatic decrease in Graham. F.L and van der Eb. J. 1973_ A new lechnique for lhe assay invasion is likely to correlate with the disrupted keratin of infeclivily of human adennvirus 5 DNA. Virology 52:456-467. structure, which could have far-reaching effects on both Gunning. p .. Leavill. J .. Muscal. G .. Ng. S-Y .• and Kedes. L 1987. A human /3-aclin expression veclor syslem direcls high-level accu­ cell adhesion and migration. In addition, the altered mulalion of anlisense Iranscripls. I'roc. Nail. Acad. Sci. USA cytoskeleton in C1070- 10 cells could be involved in 1!4:4R31-4R35. Han. I.R .. Raz. A .. and Fidler. 1.1. 19HU. Effeci of cyloskelelon­ signal transduction for the secretion of proteases. as dlSrupling agems on lhe melaslalic behaVIOr of mcianoma cells. J. previously demonstrated by Werb et al. (1989). Further Nail. Cancer Insl. IrI:K91-R9'J. studies for the analysis of these changes at the molec­ Hendrix. M.J.C.. Schor. E.A .. Schor. R.E.D .. and Fidler. 1.1. 1987. A ~imple quanlilallve assay for sludying Ihe inv",ive pOlential of ular level will be necessary. high and low human metaslalic variants. Cancer Lell. 3R:137-147. This study provides evidence for the role of CKs in Hendrix. M.J_C.. Seflor. E.A .. Seflor. R.E.B .• MISiorowski. R.L .. tumor cell invasion by demonstrating that the disruption S"h". P_t\ .. Sundareshan. P .. and Welch. D.R. 19X9. Comparison of tumor cell invasion a~says: human amnion versus reconstituted of the CK structure of a highly metastatic melanoma cell basement mcrnhranc harriers. InvaSIOn Mcla~ta~i!-. 9:27H-:!Y7. line results in a decrease in its invasive potential. This Hendrix. M.J.C .. Seflor. E.A .• McDaniel. K.M .. SeflOr. R.E.B .• Chu, also provides a possible explanation for the differential Y-W .. Yohem. K.lI.. Leong. S.I'.L.. Trenl. J.M .. Leihovilz. A.M_. Meyskens. F.L. Jr .• Canaway. D .. Welch. D.R .. Sleller-Slevenson. expression of CKs in the high compared with the low W .. Lilllla. L.A .. and Nagle. R.G. 19'X). Coexpressinn of vimenlln metastatic tumor cells. Preliminary data (unpublished and CYlOkeralins hy human melanoma lumor cells: correia lion wilh invasive and melastalic pOlential. Am. J. Palhol.. submilled. observations) of CIl161 cells transfected with antisense KingslOn. R.E .. Kaufman. R.J .. and Sharp. P.A. 19X4. Regulalion of CKI8 show a decrease in invasive potential, which lranSCriplion of Ihe adenovirus Ell promOier hy Ela gene produCIS: complements the finding in this paper and further absence of sequence specificity. Mol. Cell. Dlol. 4:1970-1'177. Kinjn, M .. Winler. H .. and Schweizer. J. 19H4. Differenlial expres­ supports a functional role of CK8 and CK18 in tumor sion of inlermedlale filament proleins in IWO ral asciles hepaloma cells. lines of common origin. Carcinogene,is 5:124'1-1~55.

32 CLINICAL BIOTECHNOLOGY Volume 3. Number 1, Spring 1991 73

Kleinman H.K., McGarvey, M.L., Liotta, L.A., Robey, P.G., Tryg· specific markers in differentiation and pathology. Cell 31:303-306. gvason, K., and Martin, G.R. 1982. Isolation and characterization Oshima, R.G. 1982. Developmental expression of murine extraem· of type IV procollagen, laminin and heparan sulfate proteoglycan bryonic endodermal cytoskeletal proteins. J. BioI. Chern. 257:3414- from the EHS sarcoma. Biochemistry 21:6188-6193. 3421. Kulesh, D.A. and Oshima, R.G. 1988. Cloning of the human keratin Rigal, D., Kopf, A.W., and Friedman, R.I. 1987. The rate of malig· 18 gene and its expression in nonepithelial mouse cells. Mol. Cell. nant melanoma in the United States: arc we making an impact? J. BioI. 8:1540-1550. Am. Acad. Dermatol. 17:1050-1053. Kulesh, D.A., Cecena, G., Darmon, Y.M., Vasseur, M., and Oshima, Steinert, P.M. and Roop, D.R. 1988. Molecular and cellular biology of R.G. 1989. Posttranslational regulation of keratins: Degradation of intermediate filaments. Ann. Rev. Biochem. 57:593-625. mouse and human keratins 18 and 8. Mol. Cell. Bioi. 9:1553-1565. Trask, D.K., Band, V., Zajchowski, D.A., Yaswen, P., Suh, T., and Laemmli, U.K. 1970. aeavage of structural proteins during the Sager, R. 1990. Keratins as markers that distinguish normal and assembly of the head of bacteriophage T4. Nature 227:680-685. tumor-derived mammary epithelial cells. Proc. Natl. Acad. Sci. Liotta, L.A., Rao, C.N., and Wewer, U.M. 1986. Biochemical inter­ USA 87:2319-2323. actions of tumor cells with the basement membrane. Ann. Rev. Trejdosiewicz, L.K., Southgate, J., Kemshead, J.T., and Hodges, Biochem.55:1037-1057. G.M. 1986. Phenotypic analysis of cultured melanoma cells: Liotta, L.A., Guirguis, R., and Stracke, M. 1987. Biology of melano­ Expression of cytokeratin-type intermediate filaments by the MS ma invasion and metastasis. Pigment Cell Res. 1:5-15. human melanoma cell line. Exp. Cell. Res. 164:388-398. Miettinen, M. and Franssila, K. 1989. Immunohistochemical spec· Werb, Z., Tremble, P.M., Behrendtsen, 0., Crowley, E., and Dam· trum of malignant melanoma: The common presence of keratins. sky, C.H. 1989. Signal transduction through the fibronectin recep­ Lab. Invest. 61:623-628. tor induces collagenase and stromelysin gene expression. J. Cell. Moll, R., Franke, W.W., and Schiller, D.L. 1982. The catalog of Bio. 109:877-889. human cytokeratins: patterns of expression in normal epithelia, Zarbo, R.I., Gown, A.M., Nagle, R.B., Visscher, D.W., and Criss· tumors and cultured cells. Cell 31:11-24. man, J.D. 1990. Anomalous cytokeratin expression in malignant Nagle, R.B. 1988. Intermediate filaments: a review of the basic melanoma: one and two-dimensional western blot analysis and biology. Am. J. Surgical Pathol. 12 (supplement 1):4-16. immunohistochemical survey of 100 melanomas. Modem Pathol., Osborn, M. and Weber, K. 1982. Intermediate filaments: Cell.type- in press.

Volume 3, Number 1, Spring 1991 CLINICAL BIOTECHNOLOGY 33 74 75

I hereby grant full permission to Yi-Wen Chu to include the article published in J. Natl. Cancer Inst. 84: 165-174, 1992, in her dissertation.

'72l~ ~. c. 2.knJ.4$ Mary J.C. Hendrix, Ph.D. 76

.','" ': '~'."':,~ ':--''';." .. ~ :'.:-:--. :-: ... ~'.--::-~; .... :':-::'-'; ".. '" . ~! :. , ". . \. .::.\", .,:::' ',~~ -:. . ",'- ':.-:.:. ,. _. ~ .' .,...,-., f. . :.: . .. "",~/;~~:; Ai ~. "'1 ,)~-:+'~ ':2'~-;'>"1~ " ,.!: .,.1 ,( ,_~) ,il :,.,J :,1 r~-";;Jl i>~~')It~I~.\·:,;<11 >'J ',.' . "./ "'\ .... ' .. ,.;/ ·:1.,. \ ". l.",. :.,. /?J...... -:-\ .'. ,-.:. ,j, ,.' I.,-:~~., .., •. :: ~:.~~:.:)/! .~;~:.~.::::\ .< i:-7-;:';.:">\ .. IV !~:~._~<:~.~.j .'~.~~¥ j , ,TH)i:,'" NA·1;IOlYifL0-. ·;'c:.

". ~ ... ;

. , ...... !: '. '.. ':' ",~~~~.;: ,'.:, :?·;~:;,ii::;;·~ ,.:.~i;!?:~~~;':~'~: ::~~:~::~:!.~:~?:fV·~}Y~;;:I \;~};~{~:!t~,~.~>it~lQ,;:f~~11·U~t~\::~;,~?~.:.~\~S?:.:~,~" ; ... ,.,l\{~ws'.~, ':.: ,~.', ~ ·;RciJearC1iers:SenrC¥for:.G"netfi:~igJi~Ip~'B~"I5t~~~eV~op~~~~i)'."::·':;":: .;: ", ~4? .. :. .-.,' ,.' ,: .. ", -.....:~~ <;-'; ':~:' " ~ ·-:,SJat-Bite:1IIV-Rclnted'·Non-HodgKiI.l>s:Lyfupfiunm·... Benefitsof Adjmant'Ihernpy . "; ,>,,,.. .'. ····c" ' . ~,:;, :''-'' , '.:',) .~"'::~ :~{':.'JA~: ' ~ ..~.: ~~:-- .. >~ .. ;-;:_~,;-..pr·i:r.:":.-:~/;:;..I':~r:.,J '.:.;~.:~~. ':; ';.';-"':, >-'~:I_; ...·t·- .~~~~{,.:.:.' ;E.'~:.'-~; /':·':-t;:t~·:·{t·~: ." .~\. ~,.":' ... ; Coe,cpressiOn ofVimeiltinruid KeJ.-atioo by:Bumim Meiannmn'TumoiC41IIs£'.'·. >.,' .\', ,v;-:Hl5..·; . . . /.', Corre!iltron·Witliln~ll~iVe'and M'eta.;;;·,1~::~

:;';. ',:";.1'._' ,= " .. _'" :,;. ;,'~ '. ~'.' " .~., . :., .. ' ,.~:. .r ~ .', ~:' "/ • .': (:.~ • :;; + :.\" ••~;:; "\: ::.:';: • ~y.::. ''':t::: ',-'~:.~ : ~'~/Aft(~,;.~>", :.~;" :;>"~:~". ;:1, J :.)',.' r .' ,;.~ ~1,:";' :. ' " ::,' '; RNA ExpressiOilofttreWn GCiU! in Wilins~Tumors,tn,~futfOJ(tG:a:r~tology.':'-::.· .'::~7;J~··. :. ,.,.: .': ·a. Miwu..G. E. Tomlmsoii;t; :F::Tmimons;V~·ltUft.·S-~t;:-COmt; tir:al. '. \L.i:' ;~:.':"::,;' . ,", :.~" .',. • .•• :). ~ c. . • '~'. " "1'" '," .';.".",:., , " .,., ~, " •.•. ' " .l.r. _. ", .'._.. ':., .... ,.: ... " , . .. .• ,:..' .:.'. ::J3!oltancementof ~jilutinifAgeiitA!!~v~fy bySR,'if.233.m~lle~FS~Cl\~iii!!~;::.;·, !':,;: :187', .... ;.. /:,. ,; ....J:~,~.:'.': ,: :,<~~(o;~~~::r;~i~~~~.;;:i:;.~~e~~i:~·~.i~~t~~~;>';/{};!J\;k;:':,:r'~'·;:};1;;;.::. i\:,': :~.~~~~~f~:n;;~£~~~~~~~~~~~~:!~!~'t~jt.J~f]~1,r~: .. ,~ .' . .' "', ',', C.'" \.,:: .;;~;., :'7.:i;~.:'::~:::;::~',l",::·,_.·:.l,~:~.\.~,~..:.. +, ... ~.~.:.:,.. ~:.'.. :.:.{,.·~~.,~~.~.~.·.~ ..,'.::,~f: .. ~:·." ~.:~",~,,: ,. ··';:... ~';· .... ··.·.• :-.:.>:.·,' ..,> ..:~~~I.,i:, •. ,.;"~ .•', ~.'.";" i"" ~ ~_ _ ,;, .. ,~:;.~::.~;:~.<.::~.~;~ .... ;;':: .. ',. ': .. r_ ~_ - "'\". ';,' •

-".:, .if':,l;;f)"J:~···;?(··.·~~~~;&~~h;;J· 77

"

...... ' .. t :: ~ .• ", •. ,' -;:.,' -7,' ;:' 1 :>? '.~.:J

. ',. '.'.

I . ~ .' ' •....~;:.,~:~ :';.> :,,' .~. ,'~;'~;':. '.: : ":.;. ·'r·' . .', :' :. ~ .. ;·:/~~·~.~:;i. ~,' :.. :.", .. . ' :' '::. :,': .. :.' .~. 78

ARTICLES

Coexpression of Vimentin and Keratins by Human Melanoma Tumor Cells: Correlation With Invasive and Metastatic Potential

Mary J. C. Hendrix,* Elisabeth A. Seftor, Yi-Wen·Chu, Richard E. B. Seftor, Raymond B. Nagle, Kathleen M. McDaniel, Stanley P. L. Leong, Karin H. Yohem, Albert M. Leibovitz, Frank L. Meyskells, Jr., Dale H. COllaway, DallllY R. Welch, Lance A. Liotta, William Stetler-Stevenson

ability in vitro and metastatic potential in athymic nude Background: Several protein markers, including vimentin, mice. The transfectant clones CI070·10 and CI070·14 de· have been used to diagnose human melanoma. Because mel· rived from the CS161 parent line showed dramatic morpho. anoma often has metastasized by the time of diagnosis, early logical chllnges, disrupted keratin filaments, and decreased markers prognostic for metastatic potential need to be iden· invasive and metastatic potential directly correlated with a tified. Commonly, vimentin is found in mesenchymal cells, reduction in migratory activity. Conclusion: These findings and keratins are present in epithelial cells, but recent studies show a correlation between the coexpression of vimentin report coexpression of vimentin and keratin(s) in epithelial with KS and K1S keratins and the invasive and metastatic and nonepithelilll neoplasms, including some meillnomas. behavior of three representative humlln melanoma cell lines. Purpose: Our purpose was to determine whether co· [J Natl Cancer Inst84:165-174, 1992) expression of vimentin and keratin(s) is correlated with tumor cell invllsion and metastatic behavior. Methods: We evalullted nine humlln melanoma cell lines expressing Received Iune 17. 1991: revised Se",mlx:r 16. 1991: accepttd NO\'cmlx:r 19. 1991. Supported by Public Health Service granl CA·54984 from the National Cancer vimentin and other mllrkers of aggressive tumor behavior Instllule. National InSillules of Health. Departmenl of Health and Human Ser· (IIMB45, S·100, "LA·ABC cllIss I and "LA-DR class )) vices: by (he Arizona Disease Control Research Commission; and by the Eric histocomplltibllity antigens, and KS and K1S keratins). Lev· Staniek Family. M. I. C. Hendrix. Departmenl of Analomy. UniversilY of Arizona College of els of KS and IUS keratins were determined in the highly Medicine. and The Arizona Cancer Center. Tucson. Ariz. metastatic CS161 cell line, the poorly metastatic A375P line, E. A. Seflor. R. E. B. Seftor. K. H. Yoheon (Depanmenl of Analomy). K. M. and the moderlltely metastatic A375M line. To determine McDaniel (Department of Pathology). Y. Wen Chu. A. M. Leibovitz nne Ari· lana Cancer Center,. UniversilY of Arizona College of Medicine. whether the presence of keratin lIffects migratory ability, we R. B. Nagle. Departmenl of Palhology and the UniversilY of Arizona College altered the conformlltional structure of kerlltin filaments in of Medicine. The Arizona Cancer Center. CS161 cells by transfection with II mutant K1S complemen· S. P. L. Leong. Departmenl of Surgery and Ihe UniversilY of Arizona College of Medicine. The Arizona Cancer Center. tary DNA. We also determined messenger RNA levels of F. L. Meyskens. Ir .• Departmenl of Hemalology/Oncology. UniversilY of Cali· human type IV collllgenase, an enzyme mllrker for invasion fornia lJ"'r'ine Cancer Center. Irvine. and metastasis. ResulJs: In A375P cells, two·dimensionlll D. H. Conaway. Michigan Cancer Foundalion. Detroil. Mich. D. R. Welch. Departmenl of Pathology. Pennsylvania Stale UniversilY School electrophoresis with Coomassie·stained gels, immunoblott· of Medicine. Hershey. Pa. ing. and immunonuorescence staining showed no detectable L. A. Liolta. W. Sleller,Slevenson. Laboralory of Pathology. Division of Can· levels of KS or K1S. A375M cells showed low levels of KS cer Biology. Diagnosis. and Centers. National Cancer Institute, National Insti· IUles of Health. Bethesda. Md. and K1S by Western and Northern blotting. with a distinc· We thank Dr. C. George Ray for the edllonal review. Dr. Ieffrey M. Trenl and tive nuorescent subpopuilltion of cells. In comparison. KS Floyd H. Thompson for performing the karyotypic analyses. Ms. Virginia Cark for and KlS levels in CS161 cells were high in all cells. Type IV her scienlific expertise. Dr. Gloria Heppner for use of the Foundation Nude Mouse FacililY. and Dr. Roben G. Oshima for the gifl of the LK442·KI8·1070 plasmid. collagenase messenger RNA levels were lowest in A375P ·Corr"pondrncr ro: Mary I. C. Hendrix. Ph.D .• Departmenl of Analomy. cells and highest in CS161 cells, correlating with invasive College of Medicine. UniversilY of Arizona. Tucson. AZ 85724.

Vol. 84. No.3. February 5.1992 ARTICLES 165 79

A major problem in cancer management is metastasis-the by J. J. Fidler (The University of Texas M. D. Anderson Cancer migration of tumor cells in the vasculature and tissue parenchy­ Center. Houston). The CSI61 cell line was established from an ma that gives rise to tumors at multiple locations in the body. Of amelanotic metastasis from a patient with recurrent. metastasiz­ the many types of cancers. malignant melanoma has been stead­ ing malignant melanoma (28). Cell lines UACC-456. UACC- ily increasing in incidence in the United States for several de­ 383. UACC-60S. and UACC-S27 were established at the cades. It is estimated that. by the year 2000. one in 90 University of Arizona Cancer Center Core Facility from patients Americans will be afflicted with this disease (/). This incidence with metastatic melanoma. Cell lines BOWES and HS294 were has been attributed to both atmospheric ozone degradation and purchased from the American Type Culture Coliection. Rock­ the increasing aesthetic value placed on a suntan by Western ville. Md .• and have been well characterized (29JO). All cell populations. Whatever the societal and environmental contribu­ lines were stored as frozen stocks. They were maintained in cul­ tions to the etiology of melanoma may be. alanningly little is ture for no longer than 2-3 weeks for each set of experiments to known about the fundamental biology of its aggressiveness in ensure that their metastatic phenotype would not change as a re­ some individuals and quiescence in others. sult of prolonged passage in vitro. From a clinical point of view. early diagnosis of melanoma Cells were grown as adherent cultures in tissue culture flasks determines disease prognosis. Unfonunately. there is good evi­ (Falcon Plastics. Lawndale. Calif.) containing Dulbecco's modi­ dence to suggest that. at the time of diagnosis. the melanoma fied Eagle medium (DMEM) (Irvine Scientific. Santa Ana. may have already metastasized (2). depending on a variety of Calif.). DMEM was supplemented with a 10% (1:1. vol/vol) factors such as tumor level and thickness (3). Therefore. specific mixture of heat-inactivated fetal bovine serum (FBS)-calf serum markers prognostic for metastatic potential need to be identified. (GIBCO Laboratories. Grand Island. N.Y.) and 0.1% gentami­ The ability to invade or penetrate biological matrices is of pri­ cin sulfate (GIBCO Laboratories). The flasks were incubated at mary imponance in the development of malignant melanoma as 37 ·C in a humidified incubator in 5% CO2 and 95% air. Cells well as the genesis of many kinds of advanced malignancy. be­ were detached by a IO-minute incubation in 2 mM EDTA in cal­ cause these matrices are encountered by tumor cells during in­ cium- and magnesium-free phosphate-buffered saline (CMF­ travasation. extravasation. and dissemination. A three-step PBS) at pH 7.4. and a single-cell suspension was prepared by hypothesis describing the sequence of biological events in­ trituration. Cell numbers were determined by using a hemacy­ volved in tumor-cell interaction with such extracellular matrices tometer. and all cultures were found to be free of Mycoplasma has been proposed (4.5): (a) tumor cell attachment to a matrix spp. contamination with the Gen-Probe Rapid Detection System substratum. (b) local degradation of the matrix by tumor ceIl-se­ (Fisher Scientific. Tustin. Cali f.). creted proteases. and (c) tumor cell locomotion into the matrix modified by proteolysis. One enzyme marker shown to be Transfection of C8161 Cells closely associated with invasive and metastatic potential is type Transfection of CSI61 cells was performed with the plasmid IV collagenase. which specifically degrades type IV collagen. LK442-KI8-1070. which contains a deleted human KlS com­ the major structural component of basement membranes (4-/0). plementary DNA (cDNA) under the control of a human p-actin Classical diagnosis of human melanoma has relied on the promoter (3/) and produces disrupted keratin filaments. This presence of several protein markers. including HMB-45 and S- plasmid also carries the gene that codes for Escherichia coli 100 antigens and vimentin (/ I -17). Vimentin and the keratins xanthine-guanine phosphoribosyltransferase (gpt) as a drug-se­ arc intermediate filaments. which are the principal components lection marker. The plasmid pHpApr-l-neo (32). which was ob­ of the cytoskeleton of mammalian cells (see review in (18·20». tained from J. Leavitt (Linus Pauling Institute of Science and Vimentin is characteristically found in mesenchymal cells. Medicine. Palo Alto. Calif.). harbors a gene for resistance to the whereas keratins are usually found in epithelial cells. Although antibiotic G418 and was used for a control transfection (C-NEO earlier studies emphasized the use of intermediate filaments as cells). The plasmids were introduced into C8l61 cells using the cell type-specific markers in differentiation and pathology calcium phosphate precipitation method (33 J4). and positive (2122). recent repons have confounded the literature with nu­ colonies were selected according to their ability to grow in merous demonstrations of the coexpression of vimentin and ker­ DMEM containing 25 lIg/mL mycophenolic acid. 2 lIg/mL atin(s) in epithelial and nonepithelial neoplasms. including some aminopterin. 15 lIg/mL hypoxanthine. 250 lIg/mL xanthine. 10 melanomas (23-26). TIle purpose of our study was to extend lIg/mL thymidine. and 25 lIg/mL adenine (Sigma Chemical Co.. these observations by measuring specific markers in several SI. Louis. Mo.) for the gpt phenotype or O.S lIg/mL G41S human melanoma cell lines of differing metastatic potentials (GIBCO Laboratories) for the control neo phenotype. and to determine whether a correlation exists between the coexpression of vimentin and keratin(s) and tumor cell invasive Chromosomal Analysis and metastatic propenies. G- and Q-banding analyses of all cell lines were performed under routine conditions as previously described (27 J5) to ver­ Materials and Methods ify that the cell lines used were of human origin and displayed the characteristics of melanoma. Human Melanoma Cell Lines Analysis of Vimenlin and Keratins A375P. a poorly metastatic variant. and A375M. an interme­ diate metastatic variant (27). were developed from the human As described below. the intermediate filaments were identi­ melanotic A375 parental cell line. These cell lines were supplied fied by two-dimensional gel electrophoresis and Western blot

166 Journal of the National Cancer Institute 80

analysis. by immunofluorescence microscopy. and by Nonhem tometrically using a model 620 video densitometer and a one-di­ blot analysis. mensional computer analysis program (Bio-Rad Laboratories). Two·dimensional gel electrophoresis and Western blot Tumor Cell Invasion Assay analysis. The first·dimension separation of proteins was per­ formed using nonequilibrium pH gradient gel electrophoresis The Membrane Invasion Culture System (MICS) assay was (36) in tube gels (containing 2% ampholines; pH range. 3.5-9.5; used (0 evaluate the degree of tumor cell invasion through a re­ Bio-Rad Laboratories. Richmond. Calif.) at 3700 V/h. The sec­ constituted basement membrane (Matrigel; Collaborative Re­ ond dimension was run on 12.5% polyacrylamide gels contain· search. Inc .• Bedford. Mass.) in vitro as previously described ing sodium dodecyl sulfate (37). which were either stained with (44). Matrigel is composed of laminin. type IV collagen. en­ Coomassie BBR-250 (Bio-Rad Laboratories; 0.25% Coomassie tactin. and heparan sulfate proteoglycans (45). Briefly. 5 x 104 BBR-250 in 25% isopropanol-IO% acetic acid) or trans blotted tumor cells were seeded randomly onto Matrigel-coated nucle­ onto a sheet of Immobilon-P membrane (Millipore Corp .• Bed­ pore filters (Nuclepore Corp .• Pleasanton. Calif.) situated be­ ford. Mass.). Visualization of the primary antibody was facili­ tween the upper and lower well plates of each MICS chamber tated using a horseradish perollidase·conjugated secondary (sill upper wells per cell line). Subsequently. the chambers were antibody and 4-chloro·l-naphthol (Bio-Rad Laboratories). incubated with humidity for 72 hours at 37 'c. At the time of Immunofluorescence microscopy. Indirect immunofluores­ harvest. the cells that had penetrated the Matrigel-coated filters cence microscopy was performed on all three cell lines. which were collected. stained. and counted visually. Percent invasion were grown on glass coverslips to approllimately 70% conflu­ was corrected for proliferation and calculated as: (total number ence. For the detection of vimentin. the V9 mouse monoclonal of invading cells [lower well sample))\(total number of cells antibody against vimentin (DAKO Corp .• Santa Barbara. Calif.) seeded [upper well sample)) x 100. was diluted 1:50 in CMF-PBS. This antibody has been charac­ Tumor Cell Motility Assay terized previously (24). Monoclonal antibody 10.11 (25). which recognizes K8 and KI8 keratins. was diluted 1:3000 in CMF­ Unstimulated motility was determined in MICS chambers PBS. containing polycarbonate filters that had been soaked overnight All cell cultures. which were on coverslips. were rinsed in in 0.1 % bovine serum albumin by use of a modification of a CMF-PBS three times. They were then immersed in methanol procedure described by McCarthy et al. (46). Tumor cells (5 x 4 for 5 minutes. frozen. treated with acetone for 3 minutes. and 10 ) were seeded randomly in each upper well (sill wells per cell then rinsed in CMF·PBS. after which they were incubated with line). allowed to incubate at 37 'C for 6 hours. and subsequently the appropriate primary and secondary antibodies. For controls. processed as described above for the invasion assay. the primary antibody was omitted. and the cells were incubated Analysis of Metastatic Potential with the secondary antibody alone. in addition 10 incubation with normal mouse IgG followed by the appropriate rhodamine­ Cultures that were 70'70-80% confluent were detached with 2 or fluorescein-conjugated secondary antibody. mM EDTA solution. and cells were resuspended in a large vol­ For detection of S-IOO and HMB-45 antigens and histocom­ ume of medium. The mixture was centrifuged. and cells were patibility antigens HLA-ABC and HLA-DR. which are used in resuspended in a small volume of ice-cold Hanks' balanced salt detection of pathological disease. cells were prepared as de­ solution (HBSS). Cell counts were performed with a hema­ scribed above and stained with the following antibodies using cytometer. and volumes were adjusted by addition of ice-cold indirect immunofluorescence microscopy: rabbit antibody to bo­ HBSS. Four-week-old. athymic nude mice (Harlan Sprague­ vine S-IOO antigen (DAKO Corp.) (38). mouse antibody to Dawley. Inc .• Madison. Wis.) were inoculated subcutaneously human HMB-45 antigen (ENZO Diagnostics. Inc .• New York. with either 5 x 101 or I x 106 cells in 0.5 mL of HBSS in either N.Y.) (39). mouse antibody to human I-{LA-DR antigen (Becton the dorsolateral flank or the midscapular region. The animals Dickinson. San Jose. Calif.). and mouse antibody to human were maintained and housed according to the National Institutes HLA-ABC antigen. Photographs were taken with a Zeiss stan­ of Health Guide for the Care and Use of Laboratory Animals. dard 18 fluorescence microscope with automatic rhodamine and All protocols were reviewed and approved by the appropriate fluorescein filter sets. Institutional Animal Care and Use Committees. Mice were fed Northern blot analysis. Total RNA was elltracted from the Purina rodent chow and tap water (chlorine content <5 ppm) ad cells by the guanidine thiocyanate-cesium chloride method (40). libitum. TIle RNA was poly A selected and separated by gel electro­ Animals were killed by cervical dislocation at 27 or 62 days phoresis in 1'70 agarose containing 2.2 M formaldehyde (41). after inoculation. and complete gross necropsies were per­ RNA marker lanes were subsequently excised and stained with formed. Lungs were fixed in a neutral-buffered formalin­ ethidium bromide. The RNA gels were capillary blotted onto Bouin's fixative solution (5:1; vol/vol). and the number of O.I-l1m nylon membranes (GeneScreen Plus; DuPont. Wilming­ surface lung metastases was counted as previously described ton. Del.). baked. and hybridized with the following nick-trans­ (47.48). lated DNA probes: the pH3A plasmid for human type IV collagenase (42). the hp4FI probe for vimentin (43). and the pK Results 811 and pK 189 probes for K8 and K 18. respectively (31). These blots were reprobed for ll-actin to ensure equal loading. In screening the nine chosen human melanoma cell lines. we The relative amount of each message was determined densi- focused on the presence of specific markers (HMB-45. S-IOO.

Vol. 84. No.3. February 5.1992 ARTICLES 167 81

vimentin. K8 and K18. and HLA-ABC class I and HLA-DR and only subpopulations of cells stained in A375M cells (Fig. class II histocompatibility antigens) that could potentially pro­ IB'). However. a brilliant staining pattern for K8 and KI8 was vide some prognostic definition of "aggressive" melanomas. seen in C8161 cells (Fig. IC'). In addition. sections of formalin­ The information presented in Table I is a compilation of data fixed. paraffin-embedded tissues of the metastatic lesion from derived from immunohistochemical analyses and shows that which the C8161 cells were derived were positive for keratins there were three consistent markers expressed in the human mel­ (data not shown). indicating that the anomalous expression of anoma cell lines studied: HMB-45. vimentin. and HLA-ABC. keratins was not a tissue culture anifact. Specific examination of the invasive potential of the cell lines Additional documentation for the presence and type of inter­ was performed in the MICS assay during a 72-hour period mediate filaments in these cell lines included two-dimensional (Table I). The ability of the cells to penetrate Matrigel-coated gel electrophoresis (Figs. 2A. B. and C). which demonstrated filters is presented as percent invasion ± standard error. The the presence of vimentin in all three cell lines. Electrophoresis metastatic potential is also shown as the range and median num­ showed no detectable levels of KS and K 18 in A375P and ber and reflects the ability of the cells to form spontaneous me­ A375M cells. but it demonstrated K8 and KI8 levels in C8161 tastases in the lungs from a subcutaneous site 62 days after cells that were similar to levels in the normal HeLa cell control inoculation. Collectively. these data indicate that the C8161 (data not shown). cells are the most invasive and. correlatively. the most meta­ Western blot analysis (Fig. 3) showed low levels of KS and static. The A375M cells are rank-ordered second in invasive and KI8 in A375M cells and high levels in C8161 cells. In Fig. 4. metastatic potential. densitometric analysis of an accompanying Nonhem blot indi­ In a separate study (42). the abilities of A375P. A375M. and cates the relative amounts of vimentin and K8 and K IS messen­ C8161 cells to form experimental metastases from intravenous ger (mRNA) levels: low expression in A375P (1,0). relatively injection via the tail vein in athymic nude mice were measured higher levels in A375M (3.0 and 6.S). and the highest levels in (data not shown). Similarly. the C8161 cells produced the high­ C8161 (23 and 18). est median number of experimental lung colonies: >250 within An imponant enzyme associated with the metastatic behavior 21 days after inoculation versus 146 for A375M and three for of several cell types is type IV collagenase (4-10). which has A375P cells 6 weeks after inoculation (48.49). also been implicated in the directed migration of human tumor To continue our study of the potential role(s) for keratins in cells (50). We therefore measured mRNA expression levels of human melanoma cells. we focused our attention on (a) one cell this 72-kd enzyme in the three selected cell lines (Fig. 5). Densi­ line that did not express K8 and K18-A375P (low invasive and tometric analysis of the Nonhern blots showed an increase in metastatic potential). (b) one cell line that contained sub­ the relative amount of mRNA levels for A375M cells and populations of cells expressing K8 and K 18-A375M (interme­ C8161 cells compared with those for A375P (relative levels ad­ diate invasive and metastatic potential). and (e) one cell line that justed to 1.0). The corresponding levels of gelatinolytic activity strongly expressed K8 and K 18-C8161 (high invasive and of type IV collagenase. as measured by zymography. have been metastatic potential). recently published by our laboratory (42). This activity corre­ With indirect immunofluorescence microscopy. the localiza­ lates with the steady-state mRNA levels. tion patterns of virnentin and K8 and K 18 were demonstrated. as In an attempt to gather more direct evidence that keratin ex­ shown in Fig. 1. Staining for vimentin was prominent in A375P. pression in melanoma tumor cells may contribute to an invasive­ A375M. and C8161 cells (Figs. IA. B. and C. respectively). No metastatic phenotype. we asked the question: What effect on staining for K8 and KI8 was detected in A375P cells (Fig. IA'). invasive potential would occur in CSI61 cells if their normal in-

Table 1. Presence of spt"Cific prolein markers in human melanoma cclliincs·

Prolein marker Invasive Metastatic potenllal pOIential. Coli line S-IOO IIMB-45 KB.KI8 Vime"li" liLA-ABC HLA-OR in vitrot subcutaneoust

UACC-456 + + + + O.oJ ± O.llO9 0 UACC-383 ± + + + O.OHO.OI 0 UACC·608 + + + + 0.32±0.OO3 0 UACC·827 + + + + + 0.34 ±0.07 0 A375P' + + + + 1.5±0.01 0 IIS294 + + + + 1.5 ± 0.22 0 BOWES + + 3.5±0.2 0 A375M' + ± + + 4.2±0.12 3CO-22) CHI61' + + + 12.5±O.11 3)(5-187)

·PhenOlypmg of human melanoma cells was dClcnnined by dln:cl and indtrecllmmunofluorescencc: microscopy. + =pasilive marker. 6Ot;f. I 00% of populauon:­ = no pmlllve siaining; and ± = subpopulallon staining only. tlnvaslve pOIenlJal was denved by measuring percenr invasion throug:h Matrigel·co3ted filters in (he MICS assay during 72 hours. Values = % invasion ± SE for samples per cell line. iMetas13lJc potential was evaluated by the number of spontaneous lung mel35tases following subcutaneous injection of I x lOb cells into the dorsolateral flank. Values = medIan (range) for six mIce/group. Ammals were killed 62 days after inoculation.

168 Journal of Ihe National Cancer Institule 82

'Ii& BSA

v ~.

A375P A375M

A 'R ~. Fig. 2. PhOlograph< of eoomassie­ stained. two-dimensional polyacryla· mide p,elli of hlp,h·salt e~traCI.IIi from AJ75P (AI. A375M (Il). and eMI61 (e) cell~. Separallon of the protein!'! ...... in the fir~t dimension was by non­ ... equilibrium pH gradienl eleClropho­ !l.-' - resis and in the second dimemion by 12.5~ polyacrylamide gels contain­ eBIBI ing ,odium dodecyl -,ulf3le. The marker!:. indicate the location of vimenlin (VI. KH 18). and KI8 (181. IB_... alii identified Immunologically on c - identically run Western blot~. A = actin: BSA =hO\o'Ine serum albumin.

Fig. I. Indirect lmmunofluoreo;.ccnce micro\cop~ or human melanoma cells p M c p M C stamed lur vlmenlm fA.B.CI or KJ:S and K 18 jA',B',C'). AJ75P IAI. A375M (Bl. and CH IllI fel celllO 'Were all PO~llive fnr vimenlin. By ,umpan~on. keralin \1310- in~ \.loa!! neg u.ing a high.,alt eXlraClion. Ten ~g of prole in W8< loaded into each lane of the 12.5~r polyacrylamide l,!cls containing sodium Indireci immunofluorescence micro>cop), of these clones to dodecyl sulfate. and the separated proteim. werc sub\cquC'nlly eleclrophorC'tically deice I K I H demonSiraled a dramalic morphological change transferrrd to an Immobilon-P lfansblOl membrane. Detection of keratinS WiS!!' accomplished uSIOg either anll·KM (left paneh or anti-K 18 (ngtu panel) pnnlilI)' compared wilh the C·NEO Iran,feCled conlrols. as shown in Fig. antibodies followed by a ~ccondar)' anlloody conjugated with horseradi~h peru,,­ 6. The lran,feClanl mUlanl cell> were highly elongated and much idJ'Ie. TIle molecular weight markers corre~pund to Ihe putallve KR (51 kdl and KIR(4Hd). larger than Ihe C·NEO cells and demon>traled an inlracellular staining pattern of short. disconlinuous strialed bundle, of fila· men!' compared wilh the normal murphulugy seen in Ihe C· In addilion. when C1070-1O and CI070·14 cell> were tested NEO cOnlruis and in the Cll 161 wild type (Fig. I J. for mela~lalic pOlential via subculaneous administralion. they \lie Ihen dClcmlined if the presence of Ihe truncaled K 18 and were unable to form primary lumors or spontaneous melaSlases. the disruplion uf nurmal Keratin filamenl furmalion in C 1070·10 Melaslalic pOlential was evalualCd by th~ size of Ihe primary and C I 070·1-1 cell' affecled their invasive pOlenlial (Fig. 7). tumor as well as by the number of spontaneous lung melastases Cell, were lesled fur Iheir ahilily 10 invade Malrigel·coaled fil· following subculaneous injeclion of 5 x lol cells inlo Ihe mid· ters in Ihe MICS a"ay during a n·hour period. The C·NEO scapular region. Animals were killed 27 days after inoculalion. (;vlmol c:ell, demonslrated IO~,; invasion. similar IU Ihe 12q in­ Six of six mice receiving C-NEO cells showed primary tumors vaSiun profile of Ihe wild.lype C8161. In comparison. the averaging 6 mm 3 and a median of Ihree lung metastases. Six of CI070-IO cells showed a 3.0~; inva~ive pOlenliaI. while the six mice receiving C1070-1O and C1070·14 cells showed no pri­ CI07(1·I-I cells achieved unly aOA'ff inva,ive pOlenlial. mary IUmors and no melastases.

Vol. R.J. Nu. 3. Feoruar), 5.1992 ARTICLES 169 83

P M C

Vim

1.0 1.0 1.3

P M C P M C

K8 K18

1.0 3.0 23 1D 6B 18

Fis;:. 4. l'orthem bioi allaJ>,., of human \'lmenU" I Vim) and human KH and K J H 10 1\~7~P (PI. Al75\f (\11. ami CHltd Ie, melanoma t:clb. Pul) A "IC!ll!cled mRSA W3\ loaded J I-I!:! per JJn~. Jnd relative value\ denHd frolll den\IfOmClrH: '!lean') are ,hu\l." under each lane.

Fi!-:. S. !\unhem hltll 3nal~"I\ of p hum;1I1 lyre 1\' L'ollaf!ena~C' gene M c cxpre~'lun 10 AJ7~P ,1'" A.'\75~f 1\1). ilnd eX!"1 (l' melanoma cellI,. R!\A \4a, p(Jl~ ,\ ,e1ecled (I IJ!! pel IJoe I ,lOti frJ1.:IIOIlJled (In 1.1 \f fnmlaldehydc ~eh. l"ap· illaf"\ binned III mlnn memo tm.l~e,.

Intennediate filament, have been implicated in vital functions related to cellular differentiation and 'crve as imponant tran,· We lunher in\'e'tigated whether a reduction in collagena,e IV du(;cr, of signals 10 thc nucleu, (/85/). Recent ad\'ances in mo· gene expres'lOn (;ould account for the dramatic decrease in in· le(;ular biology have uncovcred at least five distinct types of vasive potential for the C1070·1O and C1070·14 cells: intennediate filament' expre",ed in various cell types. which Zymograph) ,howed no dramatic difference between the ha\'e been ascribed a variety of imponanl functiom (/1)5:21- gciatinolytic aClivlI) of the 7~·l;d type IV collagenase in these In our study. nine human melanoma cell line, were evaluated (;ells and that in the C·NEO control, (Fig. HI. Levels of mRNA for the presence of the intennediate filaments vimel1lin and ker· ,h,mn h) Northern blot analysis (data not shown) (;onllnned atin, KR and K I H. in addition 10 HMB·45. S·100. HLA·ABC. the," flOdlO!!'. and HLA·DR. We have shown a WITelatlOn between the \\"lIh no "gnifi(;ant difference shown 10 type IV collagenase coexpreSSJon of \'imenlin with KH and K I H and the invasive and e~pre"J'''1. we next mea,ured the ability 01 the celb 10 migrate meta,tatic behavior of three representati\'e human melanoma through unwated polycarbonate filters during a o·hour period cell line,. The poorl) metastatic A375P cells ,howed no KH and tFig. 'iI. The,e data show a dramatic reduction in motility of K I H by routine method, of two·dimemional electrophoresis hoth the C I mO·1 0 and C I 070·14 transfectanb. with Coomassic·stained gels. Western immunoblolling. and im·

170 Journal of (he I'\ational Cancer Inslltutc 84

,. general populalion. In comrasl. Ihe highly melaslalic C8161 cells showed a slrong presence of K8 and K 18 by Coomassie­ 12 slained gels and Weslern bloning and a prominem immuno­ fluorescence slaining panern. Funhermore. Nonhern biOI 10 Fj~. 7. Percenl in\'~lOn of analysis showed a IO-fold increase in K8 and K 18 mRNA levels C-!,\EO conlnll eel" and g CI07()'IO and C1070-14 in ell 161 cells. compared wilh A375P and A375M cells. " 0 cells mca.\~d 10 the To provide more concreie evidence for a funclional role for j MICS a~\3y durm~ J 7~­ Ihese keralins. C8161 cells were Iransfecled wilh a mUlam K I I! c hour period UlilOg l: 0 Malrigel·,oaled pol~car· cDNA. which resuhed in Ihe disruplion of normal keralin fila­ !. bonate filters. mem formalion. Two of Ihe slably Iransfecled clones. C I 070-1 0 and CI070-14. showed a dramalic decrease in inva~ivc and mel­ aSlalic pOlemial. which could be anribUied in pan ((l a decrease in mOlililY. 2 This anomalous expression of keralins in melanomas has re­ cemly been reponed in Ihe lilerature (23-26). In a recent sludy 0 CoNEO of 100 melanomas analyzed by immunohislOchemislrY. ZI'7r of frozen lissue seclions were shown 10 comain keralin. specific­ ally in melaslalic and recurrenl melanomas (53). This same sludy suggesled Ihal keralin expression in melanomas may be C-NEO CI070-IO CI070-14 relaled 10 tumor progression. which was demonslraled in our in­ vesligalion wilh in vilro and in vivo models. In anolher repon characlerizing a highly helerogeneous human melanoma cell line (JIB·MEL.]). Ihree subpopulalions were isolaled by Percoll 72kD-. gradiem cemrifugalion. and all were shown 10 e~press keralin­ posilive inlermediale filamems (54). A Ihird study found Ihal one of nine melanoma cell lines specifically e~pressed keralins 7.8.13.17. and II! (55). In Ihese repom demonslraling Ihe coexislence of keralins and vimenlins. Ihree imponanl variables are apparenl which polen­ Fig. 8. Zymographlc analYSIS of the selallnol}1JC acllVUy of the 5oecrett:d 71·kd type IV colla~e=e from supernatant> of C-f'.'ED control cells and C1070-1O and lially confound Ihe palhologisl's diagnosis of melanoma versus CI070-14 cell" numerous Iypes of carcinomas: (a) Ihe fi~alion of Ihe lissue sample. (h) Ihe mode of deleclion. and (e) Ihe inlerprelalion of Ihe dala generaled. II appears Ihal Ihe besl diagnosis could be made by using a panel of monoclonal anlibodies 10 HMB-45. S- 100. keralins. and vimenlin. in combinalion wilh Weslern im­ T munoblolling and Nonhern biOI analysi~ or in silu hybridizalion • -L for funher invesligalion. Allhe molecular level. imponanl infor­ 10 FI~, 9. Un,"mulaled mi· malion relaling gene rcgulalion could be correlaled wilh f Fralol)' .bilny of C-NED slructure and funclion. control eel" and C1070- In our sludy. all cell lines were posilive for HMB-45. which 10 and CI07()-14 cells me",ured in Ihe MICS idenlifies a premelanosomal glycoprolcin found in aClivaied and I assay dunng a 6·hour pe­ neoplaslic melanocyles and differemiales keralin-posilive mela­ 52 riod uSing geJalln,coJled polycarbonale filler ... nomas from carcinomas (/ I). In addilion. all cell lines were pos­ ! --r- ! ilive for vimenlin. which is a c1as,ical marker for melanoma (12). The e~pression of keralin, in A375M cell subpopulalions may refleci lumor cell helerogeneilY (5(>-58) and Ihe develop­ menl of clones wilh disiinci biological differences. A possiblc explanalion for Ihis phenomenon ha, been offered by Knapp 0 ~ C-NED C107(>10 C107o.14 and Franke (59). These inveSligalO" suggesl Ihal an inlrinsic in­ slabililY of Ihe inaclive slale of Ihe KR and K III genes is respon­ "iblc for Ihe occurrence of Ihc,c prOlcin producl, in cenain munofluorescence deleClion. However. a low level of mRNA nonepilhelial lissues and tumor,. Specifically. Knapp and expre~~ion of bOlh keralins was delecled by Nonhern biOI analy­ Franke found Ihal several differenl Iramfomle.: lOnepilhclial sis. The imermedialc melaslalic A375M cells showed low levels cells conslitulively express KR and K I Rand Ihallhese genes can of KR and K 18 by 2-0 gel eleCirophore~is and Weslern and be Iran>cribed wilhoul an accumulalion of deleclable amOUniS of Nonhern bloning. and dislinclive subpopulalions of celb were mRNA and wilhoul Ihe appearance of keralin-conlaining fila­ po,illn: by immunofluorescence among a basically negalive menls (59). These dala reinforce Ihe need for using more Ihan

Vol. R4. No.3. February 5.1992 ARTICLES 171 85

one technique in evaluating the presence of specific intennedi­ "simple epithelium" of many early embryonic and cenain adult ate filaments, which was demonstrated in our study. tissues (64). (e) Keratins have been detected in cenain sarcomas In allempting to draw a correlation between keratin expres­ (65), astrocytomas (66), smooth muscle tumors (67), and vari­ sion and propenies associated with invasive and metastatic be­ ous other nonepithelial tumors (see review in (68)J. havior as well as directed migration (4-10,50), we measured the Collectively. these data reinforce the suggestion by Trejdo­ expression of type IV collagenase. By Nonhem blot analysis, siewicz et a!. (55) that expression of keratins in neoplastic cells mRNA levels of the enzyme are lowest in A375P cells and high­ of neuroectodennal origin represents a reversion to a more prim­ est in C8161 cells. To show a more direct relationship between itive, embryonic phenotype. This "dedifferentiated" state is keratin expression and invasive and metastatic behavior. we compatible with a more uninhibited phenotype capable of adapt­ used C1070-1O and C1070-14 cells con!aining a stably trans­ ing to mesenchymal as well as epithelial matrices and thus able fected mutant KI8 (60) in this study. to interpret additional positional clues which may contribute to a There are at least 19 separate keratins ranging in size from 40 more aggressive behavior (69). Taking into consideration the to 68 kd. Keratins are commonly expressed as panicular pairs, fact that intermediate filaments can act as signal transducers, re­ which form heteropolymers in cells. For example, K 18 is nor­ laying information from the extracellular matrix to the nucleus mally coexpressed with K8 to form a structural filament bundle. (51), and that matrix can regulate gene expression (70.71), it is Transfection of a mUlllnt K 18 protein into the highly invasive tempting to speculate that the ability to coexpress vimentin and C8161 cells would likely disrupt the nonnal interactions be­ keratins offers a selective advantage to tumor cells. Indeed, a tween K8 and K18. Unfonunately. there is no specific antigenic general theme is beginning to emerge that suggests the phenom­ epitope on the K 18 mutant protein that does not also appear on enon of "switching" phenotypic expression and coexpression of the endogenous K 18. Therefore. it is not possible to differentiate two phenotypic markers in cells confers a selective advantage the localization of K 18 mulllnt protein from nonnal K 18 protein toward Iransfonnation. Unders!anding the genetic regulation of in the cells. Since the carboxy-terminal deletion on K 18 in­ this anomalous expression holds the key to more efficient diag­ cludes the most conserved pan of the coil 2 structure that is noses and more effective chemotherapeutic strategies. shared with all intermediate filament proteins and the !ail re­ gion, it is possible that this deletion could cause the failure of cenain regions on the K8 protein to complex with their nonnal References pairing region, resulting in discontinuous filament formation. This confonnational change in filament bundling could account (I) RIGEL D. KoPF AW. FRIEDMAN RJ: The rale of malignanl melanoma in lhe for alterations in signal transduction, thus affecting invasive and Uniled Slales: AIe we making an impacl? JAm Acad Dermalot 17:1050- metas!atic ability. 1053. t987 (2) AKSLEN LA. HOVE LM. HARTVEIT F: Melaslalic dislribulion in malignanl In a previous study, Albers and Fuchs (61) showed that the melanoma: A 30-year aulOpsy Sludy.lnvasion Melas""is 7:253-263. 1987 transient expression of a truncated mutant keratin cDNA in sim­ (3) PAUL E. PAUSCII A. BOOEKl'R RH: Speed of growlh of melanoma: SlaliSli­ ple epithelial cells and squamous carcinoma also resulted in dis­ cal analysis oflhe average ages of palieOl groups. Pigmenl Cell Res 2:475- 477.1989 ruption in the keratin filament network. Funhennore, additional (4) LIOlTA LA. GUIRGUIS R. STRACKE M: Biology of melanoma invasion and evidence implicating the involvement of K8 and KI8 in cellular mela'lasis. PigmeOl Cell Res I :5-t5. 1987 migration and invasion has recently been reponed by our labora­ (5) LIOlTA LA. RAO CN. WEWER UM: Biochemical iOleraclions of lumor cells wilh Ihe basemenl membrane. Annu Rev Biochem 55:1037-1057.1986 tory using the mouse L cell model, in which fibroblastic cells (6) Srrn.El<,STl'VENSOS WG. KRUTl.SCII HC. WAClIEl< MP: The aclivalion of transfected with K8 and K 18 were at least IO-fold more invasive human type-IV collagenase proenzyme. Sequence identification of the and migratory than the parental cells (62). major convcr.;ion product following organomercurial activation. J Bioi Chern 264: 1353-1356.1989 When type IV collagenase activity was investigated to pro­ (7) REICII R. TIIOMPSON EW. IWAMOTO Y. ET AI.: Effe'" of inhibilo,," of plas­ vide evidence of a potential mechanism that might be responsi­ minogen activator. serine proteinascs. and collagenase IVan the invasion ble for the reduction in invasive activity, we discovered that the of basemenl membranes by melaslalic cells. Cancer Res 48:3307-3312. 19M8 mRNA level of expression and the gelatinolytic activity of this (8) WANG M. STl'ARNS ME: Blocking of collagenase scerelion by eSlramuSline enzyme were no different in the C-NED control cells and the during in vilIO IUmor cell invasion. Cancer Res 48:6262~271. 1988 C 1070- 10 and C 1070-14 clones. When unstimulated migratory (9) MIGNATIl P. ROBOINS E. RIFKIN DB: Tumor invasion Ihrough Ihe human amniotic membrane: Requirement for a prOleinase cascade. Cell 47:487- ability was analyzed. there was a commensurate drop in migra­ 498.1986 tion of the C1070-1O and C1070-14 cells versus the C-NED (10) URA H. BoNtlL RD. REICII R. ET AL: Expression of lype·IV collagenase control cells, which may, in pan, explain the inability of the and procollagen genes and its correlation with the tumorigenic. invasi'Je. and metastatic abilities of oncogene-transfonned human bronchial epithe­ C1070-1O and C1070-14 cells to metas!asize. However, there lial cell>. Cancer Res 49:4615-4621. 1989 was no detection of even a primary tumor, which should be fur­ (II) ORDONEZ NG. JI XL. HICKEY RC: Comparison ofHMB-45 monoclonal an· ther investigated from the viewpoint of host immunological fac­ libody and 5-100 protein in the immunohislOchemical diagnosis of mela­ noma. Am J Clin PalhoI90:385-390. 19M8 tors. (12) CASELfTZ J. BRElTlIART E. WEBER K. ET Al: Malignant melanomas contain In trying to postulate potential role(s) for the expression of only Ihe vimentin-typc: of intennedlate filament!.. Virchows Arch 400:43- SO. 1983 keratins in melanomas, it is tempting to correlate the following (13) HUSZAR M. HALKIN H. IlERC'l£G E. ET AL: Use of an,ibod,es 10 inlermcdi· facts: (a) Keratins are the first intermediate filaments to be ex­ ate filaments In the dlJgnosi!. of metastatic amelanotic malignant mela­ pressed during embryogenesis. Therefore, epithelial as well as noma.HumPalholt4:1006-IOO8.1983 (/4) MIElTINEN M. LEHTO V·P. V,RTANEN I: Presence of fibroblasHype of in­ nonepithelial cellular derivatives originate from these precursors tennedime filaments (vimenrin) and absence of neurofilaments in (63). (h) KB and KI8 are complementary and are observed in pigmenled nevi and melanomas. J CUlan PalhollO:188-192. 1983

172 10umal of the National Cancer Institule 86

115) GOWN AM. Vooa AM. HOAK D. IT AL: Monoclonal antibodies specific (42) LEVY AT. OOCE V. SODa ME. IT AL: Increased e'pression of the Mr 72 for melanocytic lumors distinguish 5ubpopulalions of melanocytes. Am J 000 Iype IV collagenase in human colon adenocarcinomas. Cancer Res PathoI123:195-203.1986 51:439-444.1991 1/01 HOUGIITON AN. ALBINO AP. COROON,CAROO C. IT AL: Cell surface anti· (43) FERRARt S. BAmNi R. KAC2MARE KL. IT At.: Coding sequence and growth gens of human melanocytes and melanoma. J E,p Med 167: 197-212. 1988 regulation afthe human vimentin gene. Mol Cell Bioi 11:3614-3620. 1986 1/71 MACKtE RM. CAMPBELL I. TURBIlT ML: Use of NK I C3 monoclonal anti· (44) HENDRtX MJ. SEFTOR EA. SEFTOR REB. IT AL: A simple quantitative assay hody in the as!.Cssmenl of benign and malignant melanocyuc lesions. J for studying the invasive polemial of high and low human metasLntic vari­ Chn PathoI37:367-372. 1984 ants. Cancer Lett 38: 137-147. 1987 IIH) STElNIJIT PM. LtEM RK: Intennediate filament dynamics. Cell 60:521- (45) KLEINMAN HK. MCGARVEY ML. LtOTTA LA. IT AL: Isolation and charac· 523.1990 tenzation of type IV procollagen. laminin and heparan sulfate proteoglycan 1191 NAGLE RB: Intennediate filaments: A review of the basic biology. Am J from the EHS sarcoma. Biochemistry 21:6188-6193.1982 Surg PathoI12:4-16. 1988 (46) McCARntY JB. PALM SL. FuRCIIT LT: Migration by haptota,is of a (20) FLtCIiS E. HANUKOGlU I: Unraveling the slruclUre of the inlennedialC~ fila­ Schwann cell tumor line to the basement membr.:me glycoprotein laminin. ments. Cell 34:332-334. 1983 J Cell Bioi 97:772-777, 1983 (2/) OSBORN M. WEBER K: Intennediate filaments: Cell.type·specific markers (47) WaclI DR. NERI A. NtCOLSON GL: Comparison of "spontaneous" and In differentiation and pathology. Cell 31 :303-306. 1982 "experimental" metastases using rat 13762 mammary adenocarcinoma (22) FRANKE WW. SCHMID E. MOLL R: The inlcnnedialc filament cytoskeleton metastatic cell clones. Invasion Metastasis 3:65-80. 1983 in tissues and in cullurcd cells: Differenti;uion specificity of expression of (48) WaCit DR. BtSl JE. MtLLER BE. IT At.; Characterization of a highly inva· cell architectural elements. In Human Carcinogenesis (Hams ce. Autrup slVe and spontaneously metastatic human malignant melanoma cell line. HN. eds). New York: Academic Press. 1983. pp 3-34 Int J Cancer 47:227-237. 1991 123) MIElTlNEN M. FRANSS1LA K: Immunohisu>chemical spectrum of malignant (49) HENDRtX MJC. SEFTOR EA. SEFTOR REB. IT AL: Comparison of tumor cell melanoma: The common presence of keratins. Lab Invest 61 :623-628. invasion assays: Human amnion venus reconstituted basement membrane 1989 barriers. Invasion Metas 9:278-297.1989 (24) AlUM) N. BAm FORA H: The distribution of vimenun and keratin in epithe. (50) TERRANOVA VP. MASLOW D. MARKUS G: Directed migration of murine lial and nonepithelial neoplasms. A comprehensive immunohistochemical and human tumor cells to collagena.sc:s and olher proteases. Cancer Res Sludy on fonnalin and alcohol fi,ed tumors. Am J Clin Pathol 88:281>-296. 49:4835-4841.1989 1987 (5/) GOLDMAN RD. GOLDMAN AE. GREEN KJ: Intennediate filament networl:s: (25) CIIAN R. ROSSIlTO PV. EDWARDS BF. IT AL: Presence of proteolytically Organization and possible functions of a divern: group of cytoskelelal ele­ processed keratin in the culture medium of MCF·7 cells. Cancer Res ments. J Cell Sci 5:69-97. 1986 46:6353-6359.1986 (52) STEINERT PM. Roop DR: Molecular and cellular biology of intennediate (261 GATITJ! KC. RAlF1CIAER E. SKINNER J. IT At.; An immunocytochemical filaments. Annu Rev Biochem 57:593-625. 1988 study of malignanl melanoma and its differential diagnosis from other rna· (53) lARDO RJ. GOWN AM. NAGLE RB. IT AL: Anomalous cytokeratin e'pres· lignant tumors. J Clin PathoI38:1353-1357. 1985 sion in mali~nanl melanorTLil: One- and two-dimensional Western biOI (27) KOZLOWSKt JM. HART IR. FtDLER U. IT At.; A human melanoma line heter· analysis and immunohistochemical survey of 100 melanomas. Mod Pathol ogeneou\ with respect 10 metastatic capacity in athymic nude mice. J Natl 3:494-50 I. 1990 Cancer Inst72:913-917. 1984 (54, GUERRA L. MOROOII J. SLAVUTSKY I. IT AL: Characterization of IIB·MEL· 1281 BREGMAN MD. FUNK C. FUKUSlltMA M: In vitro modulation of human and J: A new and highly heterogeneous human melanoma cell line. Pigment murine melanoma growth by prostnl1oid analogs. Prostaglandins 27: 17-26. Cell Res 2:504-509.1989 1983 (55) TRIJOOSIEWICZ LK. SmmtGAn: J. KEMSttEAD IT. IT AL: Phenotypic analy· (29) CREASEY AA. SMmt HS. HACKETT N. IT AL: BIological propenies of sis of cultured melanoma cells: Expression of cytokeratin-IYpe imennedi­ human melanoma cells in culture. In Vitro 15:342-350. 1979 ate filaments by the M5 human melanoma cell line. E, Cell Res 1301 RIFKIN DB. LOEB IN. MOORE G. IT AL: Propenies of plasminogen activa· 164:388-398.1986 tors fonned by neoplastic human cell cultures. J E,p Med 139: 1317-1328. (56, FIDLER IJ. HART IR: Biological diversity in merastalic neoplasms: Origins 1974 and implications. Science 217:998-1003.1982 13/1 KULESII DA. CECENA G. DARMON YM. IT AL: Posuranslational regulation (57) NICOLSON Gl: Cancer melast3sis: Tumor cell and host organ propcnies of keratins: Degradallon of mouse and human keratins 18 and 8. Mol Cell imponant ill melasLlsis to specific secondary sites. Biochim Biophys ACLl Biol9: 1553-1565. 1989 948:175-224.1988 (J.?) GUNNING P. LF~vm J. MUSCAT G. E"T AL: A human p-nctin expression (58) HOUGllTON AN. REAL FX. DAVIS U. IT AL: Phenotypic heterogeneity of vector system directs high-level accumulation of antisense transcripts. Proc melanoma. Relation to the differentiation program of melanoma cells. J Natl Acad Sci USA 84:4831-4835.1987 E,p Med 165:812-829. 1987 133) GRAIIAM FL. VAN DER ED J: A new technique for the assay of infectivity of (59) KNAPP AC. FRANKE WW: Spontaneous losse, of control of cytokeratin human adenOVIrus 5 DNA. Virology 52:456-467.1973 gene eJl.pression in rransfonned. non-epilheliaJ human cells occurring at (.I~' KINGSTON RE. KAUFMAN RJ. SIIARP PA: Regulation of transcription of the different levels of regulation. Cell 59:67-79. 1989 adenovirus Ell promoter by EI gene produces: Ab~nce of sequence speci- (60) OtU YW. DuFFY JJ. SEfTOR REB. IT AL: Disruption of nonnal cytokeratin ficity. Mol Cell Biol4: I 97(}" I 977, 1984 . fonnallon by the tran,fection ofa deleted CKI8 cDNA into a highly meta· 1351 TRENT JM. ROSENFUD SB. MEYSKENS FL: Chromosome 6q involvement in stalic melanoma cell line decreases the invOl5ive potential. J Clan BiOlech human malignant melanoma. Cancer Genet Cytogenet 9: 177-180. 1983 3:27-33.1991 13M O'FARJUll PZ. GoonMAN HM. O'FARRaL PH: High resolution twCHIimc:n· (6/) ALDERS K. FUCHS E: The e'pression of mutant epidermal keratin cDNAs siona! elecuop~sis of basic as well as acidic proteins. Cdl 12:1133-1141. lIansfected in simple epilhelial and squamous cell carcinoma hnes. J Cell 1977 Bioi 105:791-806. 1987 (37) LAEMMLI UK: Cleavage of structural proteins during the assembly of the (62) CIIU YW. OSlItMA RG. HENDRIX MJ: E'preSSlon of keratins 8 and 18 in head of bacteriophage T 4. Nature 227:680-685. 1970 mouse l cells results in an increase in migratory and Invasive ability. J 13H) KINDDLOM LG. LoDDING P. ROSENGREN L. ET AL: 5·100 protein in Cell Bioi III :42A. 1990 melanocYlic tumors. An immunohistochemical investigation of benign and (63, FRANKE WW. GkUND C. KUHN C. IT AL: Fonnation of cytoskeletal ele· malignant melanocytlc tumors and melasta"iiCS of malignant melanoma and menls during mouse embryogenesis. III. Primary mesenchymal cells and a characteriuuion of the antigen in comparison to human brain. Acta the first appearance of vlmentin filaments. Differentiation 23:43-49. 1982 Pat hoi Mlcroblollmmunol Scand 92:219-230. 1984 (0./) JACKSON BW. GRUND C. SCitMID E. IT AL: Fonnallon of cYlOskeletal ele· 139) GOWN AM. VOOEL AM: Monoclonal antibodies to human inlennediale fiI· ments during mouse embryogenesis. Inlennedlate filaments of the amem prOlein!.. II. Distribution of filament proteim in nonnal human lis· c),fokeratln type and desmosomes in prelmplantalion embryos. Differentia­ sue>. Am J PathaI1l4:309-321. 1984 tion 17:161-179. 1980 I~()I CIIIRGW" JM. PRZYBYLA AE. MACDoNALD RJ. IT AL: Isolallon of biologi. (65) MIEmNEN M RAPOLA J: Immunohistochemical spectrum of cally active ribonucleic acid from sources enriched in ribonuclease. BIO' rhabdomyosarcoma and rhabdomyosarcoma· like lumors. E"pression of chemIStry 18:5294-5299.1979 cYlOkeratin and 6K·kd protein. Am J Surg Pathol 13:12(}" I~I I LEHkACII H. DIAMOND D. WOZNEY JM. l"T AL: RNA molecular weight de· 132.1989 tenninallon"ii hy gel electrophoresis under denalUnng conditions. a critical (66' COSGROVE M. FrrwlDaoNs PL. SIIERROD A. IT AL: Intennedlate filament ree»monallon. Biochemistry 16:4743-4751. 1977 e'preSSlon in astrocytic neoplasms. Am J Surg Pathol 13: 141-145. 1989

Vol. 84. No.3. February 5. 1992 ARTICLES 173 87

(671 BROWN DC. TIlEAKER JM. BANKS PM. ET Al: Cytokeralin expression in (69) HENDRlx MJC. SABA PZ. McDANIEL KM. ET Al: Culturing human mela­ smooth muscle and smooth muscle IUmol'. Histopathology 11:477-486. noma cells of varying: metastatic potential on different sUbStr.U3 can mnu­ 1987 ence invasive potenlial. J Cell BioI107:797A. 1988 (681 FRANKE WW. JAIIN L. KNAPP AC: Cytokeratins and desmosomal proteins (70) McDoNALD JA: Matrix regUlation of cell shape and gene expression. Curr in cenain epithelioid and non-epilhelial cells. In Cytoskc:lc:tal Proteins in Opin Cell Bioi 1:995-999.1989 Tumor" Diagnosis (Osborn M. Weber K. eds). Cold Spring Harbor. NY: (71) HAY ED: Theory for epithelial-mesenchymal transformation based on the ColdSpringHarborLaboratory.1989.pp 151-172 '"fi ..d cortex" cell motility model. Cell Motil CYlOskeleton 14:455-457. 1989

174 lournal of the National Cancer Institute 88 89

I hereby grant full permission to Yi-Wen Chu to include the article published in Proc. Nat!. Acad. Sci. USA 90: 4261-4265, 1993, in her dissertation.

Mary J.C. Hendrix, Ph.D. 90 Proc. Nat!. Acad. Sci. USA, In Press, 1993 Classification: Biological Sciences / Cell Biology

Expression of Complete Keratin Filaments in Mouse L Cells Augments Cell Migration and Invasion (keratin / migration / invasion)

Yi-Wen Chu*, Raymond B. Runyant, Robert G. Oshima+, and Mary J.e. Hendrix§

*Cancer Biology Program and §Department of Anatomy, College of Medicine, University of Arizona, Tucson. AZ 85724; tDepartment of Anatomy. University of Iowa, Iowa City. IA 52242; and +Cancer Research Center, La Jolla Cancer Research Foundation, La Jolla, CA 92037.

Communicated by: Dr. Herb Carter

§Corresponding Author: Dr. Mary J.C. Hendrix Department of Anatomy College of Medicine University of Arizona Tucson. AZ 85724 TEL: (602) 626-7893 FAX: (602) 626-2180 91

Abbreviations: ECM, extracellular matrix; IF, intermediate filament; MICS. Membrane Invasion Culture System. 92 ABSTRACT Intermediate filament proteins have been used to diagnose the origin of specific cells. Classically, vimentin is found in mesenchymal cells, and keratins are present in epithelial cells. However, recent evidence suggests that the coexpression of these phenotype-specific proteins augments tumor cell motility, and hence, metastasis. In the present study, we used the mouse L cell model to determine if a direct correlation exists between the expression of additional keratins in these cells, which normally express only vimentin, and their migratory ability. Mouse L cells were transfected with human keratins 8, 18 and both 8+18. The results indicate that the cells expressing complete keratin filaments have a higher migratory and invasive ability (through extracellular matrix-coated filters) compared with the parental and control-transfected clones. Furthermore, there is an enrichment of keratin-positive cells from a heterogeneous population of L clones selected over serial migrations. This migratory activity was directly correlated with the spreading ability of the cells on Matrigel matrix, in which the keratin-positive transfectants maintain a round morphology for a longer duration, compared with the other L cell populations. Collectively, these data suggest that keratins may play an important role(s) in migration, through a special interaction with the extracellular environment, thereby influencing cell shape. 93 INTRODUCTION Keratins comprise the largest group in the intermediate filament (IF) multigene family. There are at least 19 different keratin proteins which can be further subdivided into two groups: type I (acidic) and type II (neutral-basic) keratins based on their sizes and isoelectric point (1,2). One major difference associated with keratins that distinguishes them from other members of IFs, such as vimentin and desmin, is that keratins are expressed as heteropolymer pairs consisting of specific type I and type II proteins (3-5). Their expression is highly regulated during embryonic development and cellular differentiation (6-8). The functional role(s) of IFs is still unclear (for review, see 9,10). In epithelium, the keratin filaments are found attached to the cell-cell adherence junctions, known as desmosomes (11), as well as to the cell-extracellular matrix (ECM) adhesion sites, known as hemidesmosomes (12). The function of this interaction is thought to provide a mechanical scaffold for the epithelial sheet to maintain its integrity and structure. This hypothesis has been strengthened by recent investigations showing that transgenic mice expressing a mutant keratin 14 gene have a phenotype resembling a human genetic skin disease known as epidermolysis bullosa simplex, which is characterized by blistering of the skin due to the cytolysis of the basal epidermal cells (13,14). This is the first example demonstrating a relationship between abnormal keratin proteins and a human disease. This finding provides important information for understanding the normal function(s) of keratins in cells. However, there are many different keratin species, but whether they have a common function as epidermal keratins is still questionable. 94 Keratins are usually expressed in epithelia; vimentin is found in fibroblasts and endothelial cells; whereas des min and neurofilaments are markers for muscle and neuronal tissue, respectively (2). Although this cell lineage- and differentiation state- specific expression of IFs may be used in the diagnosis of tumors (15,16), noteworthy exceptions still exist in some tumor species and disease states. Work by Trask and coworkers (17) showed that there are different sets of keratins expressed in normal vs. tumor-derived mammary epithelial cells. The peculiar switch of keratin expression during neoplastic transformation may be a significant phenomenon associated with tumor progression. Another interesting observation is the unusual coexpression of vimentin and keratin in various tumor cells, which in a few documented cases correlates with metastatic potential. To list some examples: a non-metastatic rat pancreas adenocarcinoma expresses only vimentin; whereas a metastatic variant contains a high amount of keratins in addition to vimentin (18). Solid epithelial tumors contain keratin filaments exclusively; however, when these cells are present in ascitic or pleural fluid, they also express vimentin (19). In a study with murine sarcoma cells, the ascites from the sarcoma were found to coexpress keratin and vimentin, whereas the solid tumor from the same sarcoma expresses only vimentin (20). In several human breast cancer cell lines, vimentin, in addition to keratins, is expressed only in highly invasive but not in low invasive cell lines (21). Lastly, human melanoma, which expresses vimentin as its classical IF marker, coexpresses keratins in recurrent and metastatic states (22-25). Collectively, these data suggest that there is an additional type of IF expression during the process of tumor metastasis, which contributes to migratory and aggressive behavior in some manner. 95 To further elucidate if there is a direct correlation between the expression of additional keratins in cells and their migratory ability, we have used the mouse L cell model. Mouse L cells are fibroblasts that express vimentin. For this study, these cells have been transfected with human keratins 8 and 18 DNAs (26), and the formation of keratin filaments was visualized using indirect immunofluorescence staining. Invasion analysis showed that the cells with complete keratin filaments have higher invasive ability and deformability capability to penetrate polycarbonate filters (containing 10 Jlm pore size) than control cells, in addition to differential spreading ability on an ECM. This study suggests that keratins may playa role in migration, by influencing cell shape. 96 MATERIALS AND METHODS

Cells and Cell Culture Mouse L cells are a thymidine kinase-deficient derivative fibroblast line and were cultured in Dulbecco's modified Eagle's medium containing a 10%, 1:1 (v/v) fetal bovine serum to calf serum, supplemented with 0.1 % (v/v) Gentamycin Sulfate. L cells transfected with different keratin DNAs have been described previously (26,27). Briefly, LK8 cells represent L cells transfected with LK442- K8 vector, which contains the human keratin 8 cDNA, and were cultured in media containing 1.6 mM xanthine and 30 Ilg/ml mycophenolic acid. LK18 cells were obtained by co-transfection with plasmid pGC1853, which contains the human keratin 18 genomic DNA; and pSV2Neo was tranfected into L cells and selected with 400 Ilg/ml G418. LK18 cells transfected with LK442-K8 yielded the LK18+K8 clones and were grown in medium containing xanthine, mycophenolic acid, and G418.

Immunofluorescence Staining Cells were seeded onto glass coverslips to 70% confluence. Cells were washed three times with PBS and fixed in cold methanol for 7 min, and air dried. Indirect immunofluorescence staining was used for staining keratin intermediate filaments. Mouse anti-human keratin 18, CK5 (leN, IL) was incubated with the covers lips for 1 h, followed by a rhodamine-conjugated secondary antibody (Organon Teknika, PA). For staining actin filaments, cells were fixed with 3.7% formaldehyde for 10 min, and permeabilized for 5 min in 50% acetone/water. NBD-phallacidin (Molecular Probes, OR) was used to stain the F-actin filaments 97 as per the manufacturer's direction. In some cases, the cells were seeded onto Matrigel (Collaborative Research, MA)-coated glass coverslips and grown for 36 h, and then fixed accordingly. All fluorescence staining was observed with a Zeiss standard 18 fluorescence microscope equipped with automatic rhodamine and fluorescein filter sets.

Invasion and Trans-Filter Migration Assays The Membrane Invasion Culture System (MICS; 28,29) was used to measure cell invasion. A polycarbonate membrane containing 10 Jlm size pores (Poretics, CA) was coated with a reconstituted basement membrane gel (Matrigel) and placed between the upper and lower well plates of the MICS chamber. Subsequently, 5x104 cells were resuspended in 10% NuSerum (Collaborative Research, MA), and seeded into the upper wells. After incubation for 72 h at 370 C, cells that had invaded the Matrigel-coated membranes were harvested from the lower wells, stained, and counted. Invasion potential was calculated as the percentage of cells that invaded the Matrigel compared to the total number of cells seeded (after correction for proliferation). The method for measuring trans-filter migratory ability was modified from the procedure of McCarthy and coworkers (30). Instead of Matrigel coating, the polycarbonate membranes were soaked in a gelatin solution (0.1 mg/ml gelatin in 0.02 M glacial acetic acid) overnight to provide a sticky surface. The cells were seeded in the chamber for 6 h at 370 C, and those cells which had migrated through the pores to the lower wells of the MICS were harvested and counted to calculate the migratory rate. 98 Time Lapse Cinematography Cells were trypsinized in Ca++ and Mg++-free Hank's solution, washed, and resuspended in medium (FI2) containing 1% fetal bovine serum, insulin (5

~g/ml), transferrin (5 ~g/ml) and selenium (5 ng/ml) (ITS premix, Collaborative Research, MA). Aliquots of cells were applied to Matrigel-coated dishes at a 1:3 dilution of the original volume. After 2 h of incubation (5% C02, 370C), the cells were videotaped for a 3 h interval using an Olympus IMT-2 inverted microscope equipped with Hoffman Optics (Hoffman Optics, Inc., NY) and an environmental chamber designed by the microscope manufacturer. C02, humidity and temperature were maintained in the chamber throughout the 3 h observation period. Twenty-four hours after plating, additional dishes of cells were selected and videotaped for an additional 2 h period. Subsequently, videotapes were analyzed on the Dynamic Morphology System (DMS) of the Cell Motility Center of the University of Iowa (31). Each tape was analyzed by tracing the cell outlines of all cells in a particular field (7-27 cells) at 10 min intervals for 2 or 3 h.

Analysis by the DMS system included parameters of centroid movement (~m/h) and cell shape (% roundness). Data reported are the average of measurements made over the entire recording period. The scale was determined from the included image of a' stage micrometer on the videotape filmed at the time of migration.

Selection of Migratory Cells To select subpopulations from a heterogeneous population of LKlS+KS cells, as well as from the keratin-negative parental L cells, a modified MICS chamber (MegaMICS; 32) was used. 1.5xl06 cells were seeded in MegaMICS, 99 and the migration assays were performed as described above. Cells were then collected from the bottom wells and plated either on glass coverslips for immunofluorescence staining, or in culture flasks to expand for second-round selection. This procedure was repeated for third-round selection. The number of keratin filament-positive cells was detected and counted by immunofluorescence staining with CK5 antibody. The migratory abilities of all the selected populations were measured simultaneously in parallel experiments.

Cell Attachment Assay Cells were metabolically labeled with 35S-methionine (Amersham, IL) at 5 JlCi/ml for 48 h. After harvesting with EDTA-PBS, the cells were washed twice in PBS and resuspended in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. 2.5xl04 cells were loaded in the MICS chamber containing a Matrigel-coated polycarbonate membrane, and incubated at 370 C for 1 h. Non­ attached cells were gently washed away with PBS and attached cells were solubilized with 2% SDS. Radioactivity was determined by liquid scintillation counting, and the attachment ability was calculated as the percentage counts of solubilized celllysates compared to the total counts loaded.

Cell Spreading To determine the time required for the cells to spread on different matrices, cells were plated on 24-well tissue culture dishes coated with Matrigel, fibronectin, collagen I, or collagen IV (Collaborative Research). Cell shapes and morphology were monitored periodically using a wild inverted light microscope, and pictures were taken using ILFORD PANF 135 film. 100 RESULTS

Keratin Filament Formation in Mouse L Cells Although keratins are the type of IF present in epithelial cells, they have been expressed in several non-epithelial cell lines using DNA transfection techniques (26,33). Previously, we have transfected mouse L-cell fibroblasts with human simple epithelial keratins 8, 18 or both DNAs and obtained several different stably transfected lines (26). They are termed LK8 cells, which are L fibroblasts that received the plasmid (LK442-K8) containing keratin 8 cDNA driven by a human B-actin promoter; LKI8, which are L cells that were co­ transfected with vectors containing a neomycin resistance marker and a keratin 18 gene; and LKI8+K8, which represents LK18 transfected again with the LK442-K8 plasmid. The fate and expression of these transfected keratin DNAs have been analyzed in detail (26), including RNA and protein analysis (26). Figure 1 shows the results of the immunofluorescence staining of these cell lines with an anti-human keratin antibody which recognizes simple epithelial keratins: the parental L cells show negative staining as expected; LK8 and LK18 also display no staining for keratin filaments, indicating a single type of keratin alone cannot form filaments in the cells. The LKI8+K8 (clone 7) cells have a normal and complete filamentous keratin network that extends from the nuclear membrane to the plasma membrane. Furthermore, the keratin filament formation in LKI8+K8 cells does not affect the endogeneous vimentin IF network normally produced by the mouse fibroblasts (data not shown). 101 Invasive and Migratory Ability of the Transfected L Cells Does coexpression of vimentin and keratins 8 and 18 contribute to the invasive ability of the LKI8+K8, as we had observed previously in melanoma (25)? To address this question, we decided to compare the LKI8+K8 cells, which contain full length keratin IFs in addition to , with L cells, LK8 and LK18 clones which do not form additional keratin filaments. The invasive potential of these cell lines was performed in an in vitro invasion chamber over a 72 h period. The ability of the cells to invade Matrigel-coated filters is presented as a series of bar graphs showing percent invasion ± standard error (Fig. 2). Cells transfected with K8 or K18 alone have similar invasive ability as parental L cells. Interestingly, LKI8+K8 cells have the highest invasive ability, which is approximately 3.5-fold greater than the L cells. One of the major mechanisms involved in cell invasion is motility (34); therefore, cells demonstrating higher invasive activity usually show higher migratory ability. To test if this was the case with the transfected L cells, cell movement was initially measured as the ability to move in a three- dimensional manner from the upper wells of the MICS chamber through a gelatin-coated filter (containing 10 J..lm size pores) over 6 h (Fig. 2). The data reveal that the LKI8+K8 showed the highest ability to deform and move through the pores (which are much smaller than the cell size), compared to the keratin-negative L cells, as well as the LK8 and the LK18 cells. However, no significant difference between Land LKI8+K8 cell movement was measured (by time lapse cinematography) in a lateral, two-dimensional direction after 2 or 24 h on Matrigel matrix (data not shown). 102 In order to understand how the presence of nonnal keratin filaments in cells already expressing vimentin might contribute to increased invasive activity, we tested a few factors which have been shown to contribute to invasive and migratory activity. These include autocrine motility factor receptor that regulates cell movement (35), G-proteins involved in the signal transduction pathway (36), and the metalloproteinase(s) collagenase IV (72 KD; 37), in addition to stromelysin (38), and plasminogen activator activity (39), which participate in the ECM degradation cascade. However, there were no differences in the expression, response or activities of any of these proteins produced by L vs. LK18+K8 cells (data not shown).

Selection of Keratin-Positive Cells from a Heterogeneous LK18+K8 Population Not all LK18+K8 cells or other double transfected fibroblasts contain immunofluorescence detectable keratin filaments even after subc10ning (26). The heterogeneity of the population of a particular clone is likely due to rearrangement and loss of one of either keratin vectors. However, the heterogeneity of the LK18+K8 cells permitted a test of our hypothesis that keratin-positive transfected L cells are more invasive and migratory. We perfonned selection assays using heterogeneous LK18+K8 populations (clone 7 and an additional clone 5) that contained only a certain percentage of keratin filament-positive cells, which were seeded into a MegaMICS chamber for migration analyses. Interestingly, the majority of the clone 5 transfectants displayed "incomplete" keratin-positive filaments (i.e. the filaments were more concentrated in the perinuclear area with fragmented extensions throughout the 103 cytoplasm). This is probably due to lower levels of expression of keratins in these cells, which provided an additional parameter to test in the selection experiments. The cells that migrated (over 6 h) through the filters were collected, expanded, and selected sequentially through MegaMICS for a total of three rounds. The number of keratin-positive cells associated with each round was visually assessed with a fluorescence microscope, and their migratory ability was measured simultaneously. Interestingly, a starting population containing 44.2% keratin­ positive clone 7 cells was .enriched to 58.3% after the first-round of selection (clone 7-1), and enhanced again to 68.0% and 74.8%, respectively for clones 7-2 and 7-3. The migration analysis of these cells showed there is an increase in migration rate which is directly proportional to the percentage of keratin positivity in each cell population (Fig. 3A). Similarly, keratin positive cells in LK18+K8-clone 5 were enriched from 33.9% to 40.0% after the first-round selection (clone 5-1), and then increased to 46.5% (clone 5-2) and 57.4% (clone 5-3) after further selections. Their migration rate also gradually increased with the enhancement of the number of keratin-positive cells (Fig. 3B). We have also performed control selection experiments using keratin-negative L cells. The results showed no increase in the migratory ability of migration-selected L cells from the MegaMICS, compared with the parental L cells (data not shown).

Cell Attachment, Spreading, and F -actin Formation in Response to Basement Membrane Substrate The invasion of cells through ECM substrates requires several sequential steps: cell attachment to ECM, cell spreading via cell-matrix specific receptors, and cell migration through degraded matrix (34). We tested the ability of the 104 transfected cells to attach to a Matrigel-coated membrane using 35S-methionine labeled cells. In an one-hour assay, L, LK8, LKI8, and LKI8+K8 all showed approximately 60% of the cells attached to the matrix, with no significant difference among the four cell lines (data not shown). However, observations of cell spreading indicated there is a great difference between L and LKI8+K8 with regard to the time required for the cells to spread on the Matrigel matrix. Figure 4 shows the phase contrast microscopy pictures taken at several time points after the cells were seeded on Matrigel-coated dishes. At 12 h, both cell lines still maintained a rounded morphological shape; after 24 h, L cells started to extend their pseudopodia; and by 36 h, they had formed a webnet type of organization. On the other hand, LKI8+K8 cells did not spread as well after 36 h. However, the morphology and spreading area of both cell lines are similar after 72 h, indicating they both have the ability to interact with Matrigel components for spreading, but it takes much longer for the LKI8+K8 cells to approximate a flattened morphology. We also investigated whether the actin filamentous architecture may have been affected in companion cultures seeded on either glass or Matrigel-coated coverslips for 36 h, using NBD-phallacidin (data not shown). In response to the Matrigel matrix, LKI8+K8 cells remain rounded, and the fluorescence staining was diffuse and primarily localized in the cortical region, with few cells displaying stress fibers; whereas the keratin-negative L, LK8 and LK18 cells similarly displayed stress fiber staining on either glass or Matrigel. The spreading data were further quantified by computerized digital analysis of the photomicrographs shown in Figure 4. After 24 h on Matrigel, the LKI8+K8 cells are approximately 80% round in morphology, while the L cells are approximately 50% round. By 36 h, the roundness factor has diminished to 54.2% for 105 LKI8+K8 cells and 43.1 % for L cells; and by 72 h, little difference in shape exists (18.2% for L cells vs. 26.2% for LKI8+K8 cells). LK8 and LK18 showed the same pattern of rapid spreading as L cells (data not shown). We have also used different substrates to test this difference in spreading phenomenon. Both cell lines (L and LKI8+K8) respond very quickly and similary to type I and IV , as well as fibronectin (data not shown), suggesting that the spreading response on the more complex Matrigel matrix is unique. 106 Discussion Investigation of the functional role(s) of IFs poses a challenge as well as a dilemma. Recent breakthroughs, however, for keratin studies, have advanced our observational opportunities; specifically using a transgenic mouse model containing a mutation in keratin gene 14, which results in the development of the human genetic disease epidermolysis bullosa simplex (14), and the keratin knockout approach, which allows us to study the requirement of keratins in cellular differentiation (40). For the present study, we imployed gene transfection technology to address a fundamental question regarding the role(s) of keratin filaments in migration and invasion. Transfection experiments have provided significant information concerning keratin filament formation, maintenance and degradation, which provided a firm foundation for our investigation. Interestingly, the present study is an extension of an observation made in our laboratory with human melanoma tumor cells in which we have reported the unusual coexpression of two different types of IFs -- vimentin (which is normally expressed) and keratins 8 and 18 in highly invasive and metastatic melanoma (25). We hypothesized that additional IF filament expression may offer a selective advantage for cells to be more migratory. To test this hypothesis, we transfected a deleted keratin cDNA in a previous study, into the highly metastatic melanoma cells and showed: 1) the disruption of endogenous keratin filament organization; and 2) a dramatic decrease in invasive and metastatic ability (41). These data suggested that the additional keratin filament expression in advanced stages of melanoma may be necessary for their aggressive behavior. Recently, we have successfully transfected a human melanoma cell line, A375P, of low invasive/metastatic ability 107 (which normally expresses only vimentin) with keratin 8+18 DNAs, and tested its ability to migrate. These experiments have shown a two-fold increase in the ability of the transfected cells to migrate compared to the neo-transfected controls, which provides additional confirmatory evidence for the role of keratins augmenting motility (unpublished data). Hence, we have set forth to further test our hypothesis in the L cell model, and found many similarities in the data generated from this study and the previous melanoma results. Transfection of L cells (which normally express only vimentin) with a single type of keratin did not result in the formation of keratin filaments in either the LK8 or LK18 stable transfectants (26,27), as expected. Whereas, L cells transfected with both keratins 8 and 18 (LKI8+K8) demonstrated a complete keratin filamentous network with immunofluorescence staining. We next discerned the ability of the L cell lines to invade Matrigel­ coated filters over a 72 h period, as well as migrate through gelatin-coated filters over 6 h. The results showed the invasive ability of the cells more or less coincided with the trans-filter migration rates. To further delineate potential differences in migration across the surface of the Matrigel matrix, we measured the rate of lateral migration (j.lm/hr) of L vs. LKI8+K8 cells at 2 and 24 h, using microcinematography. We found no significant difference between the cell lines, which is consistant with previous studies of cultured epithelial cells in which keratin filaments had been disrupted by antibody injection, but had no effect on cell movement in tissue culture (42). However, this measurement of lateral migration is different from trans-filter migration and invasion in that the latter two events encompass deformability in a three-dimensional manner through 10 j.lm size pores. Recently, Robey and colleagues (43) have also shown that the 108 presence of keratins 18 and 19 in human retinal pigment epithelial cells plays an important role in the active migration ability of the cells using a similar type of trans-filter migration assay protocol. Interestingly, a correlation between cell deformability and metastatic potential has been proposed (44,45), which indicates that more metastatic cells have a greater ability to deform and migrate. Although it seems difficult to imagine how additional IFs in cells would contribute to deformability, it is tempting to speculate that perhaps the keratin filaments do not form a rigid structure as previously thought. A recent report has shown that the incorporation of microinjected keratins into endogenous keratin filaments is a very rapid process, indicating the keratin IFs are very dynamic cytoskeletal elements (46). Cells with additional keratins may actually provide flexibility to cell structure due to the energetic assembly-disassembly process of keratin filaments. This view of a dynamic IF infrastructure supports the recent insights into IPs provided by Skalli and Goldman (47). The invasion and trans-filter migration rates of LKI8+K8 were the highest among all the transfected cells, which supports our hypothesis proposed from the melanoma study. In addition, the most striking result in the present study is the enrichment of keratin-positive populations by selection via trans-filter migration. The heterogeneity of keratin filament expression in LKI8+K8 cells is likely due to the recombination and segregation of the many copies of the vectors that are integrated. However, the number of keratin-positive cells can be enhanced by several rounds of selection from the migratory cells. In addition, the higher percentage of keratin-positive cells in the population have a higher migratory rate, in fact a linear correlation. Clone 5 of LKI8+K8 cells displayed a lower migratory ability compared to clone 7, which is probably due to a lower 109 percentage of keratin positive cells in this unselected clone, in addition to a high percentage of "incomplete" keratin filaments in these same cells. However, keratin positivity was still enhanced via migration selection. A control experiment using migration selected L cells showed that there is no increase in the migratory ability of the selected L cells. This indicates that the observed increase in migration rate is not a general phenomenon associated with the selection procedure. On the other hand, it is associated uniquely with the number of keratin-positive cells in the population. These results indicate that indeed keratins are associated with the increased migratory behavior of the cells. Our data strongly suggest that cells with both vimentin and "complete" keratin IFs can move and invade faster than the ones with only one type of filament. The mechanisms involved in the invasion process, proposed by Liotta (34), can be simplified to consist of: a) attachment to an ECM; b) degradation of the ECM; and c) motility of cells through the .degraded matrix. We, therefore, focused our attention on measuring potential differences in the degradative ability of the cells that could account for their invasive behavior through Matrigel. However, measurement of type IV collagenase (72 KD), mouse stromelysin and urokinase indicated no differential expression of these proteases between the parental L cells and the transfected clones. We then focused on measuring potential differences in attachment ability among the cell lines. In these experiments, we found no differential ability for attachment to Matrigel. These findings then prompted us to focus on the process of cell spreading, which involves the interaction of cell surface receptors with the ECM. It is also an important step in the process of tumor cell metastasis in which low-metastatic cells spread and attach firmly to the matrix, while highly metastatic cells do not spread 110 as well, thus facilitating their ability to move freely and detach more easily from the primary tumor mass. Interestingly, the L cells with complete keratin filaments (LKI8+K8) have very different spreading ability on Matrigel compared to parental L cells. Our data indicate that a longer time is required for the LKI8+K8 cells to spread on a basement membrane matrix composed of a complex meshwork of laminin, type IV collagen, entactin, heparan sulfate proteoglycans and vitronectin (48,49): at 12 h post-seeding, both cell lines still have rounded shapes; however, L cells start to extend their cellular processes at 24 h, whereas LKI8+K8 do not show this until after 36 h. After 72 h, there is no difference between the two. It is still unknown how keratins may affect the spreading process; nevertheless, the retarded spreading of LKI8+K8 correlates very well with their invasive ability. We speculate that the less spread morphology exhibited by LKI8+K8 cells is more conducive for their migration through 10 Jl.m pores. Preliminary data indicate a preferential migratory ability of L cells toward laminin, the major component of Matrigel, which might facilitate rapid spreading and attachment, compared with LK18+K8 cells. The actin organization in cells is also involved in the spreading process as well as maintaining cell shape. It has been shown that mutant cells with impaired spreading on fibronectin have decreased numbers of stress fibers in addition to reduced levels of fibronectin receptors compared to wild-type cells (50). In tumor cells, low-metastatic cell variants usually contain tightly packed actin bundles, and high-metastatic variants have very little F-actin (51,52). This is also the case for LKI8+K8 which showed very poor actin organization on Matrigel matrix; while L, LK8 and LK18 have well organized stress fibers. These findings provide 111 additional support for cell shape differences and invasive ability through Matrigel. The occurrence of IF coexpression has been reported in several other species (53) as well as in specific cases of neoplastic progression. Moreover, it is a prevailing phenomenon in early embryogenesis. It has been suggested that the coexpression phenotype in parietal endoderm cells of the early mouse embryo may be involved in the ability of cells to datach and migrate away from an epithelial sheet (54).The embryonic cells are thought to be very motile and communicate differently with the ECM compared with differentiated cells. Therefore, the cells that have both filaments, such as metastatic tumor cells, may obtain "embryonic" characteristics in their behavior. Further investigation of the differential migratory and spreading ability of cells containing single and dual IF networks will focus on cell surface receptors and linkages to the cytoskeletal proteins. Although the connection between IFs and cell surface receptors is not fully understood, we suspect involvement of one of the following scenarios: 1) in a "positive" model, keratin filaments, as shown in LK1S+KS cells, may anchor directly or indirectly into a specific type of receptor(s) in the cell membrane, which in turn may "fix" the receptors in a manner which may inhibit clustering and, thus, facilitate'movement; or 2) alternatively, in a "negative" model, additional IFs may interfere with the binding of receptors to the cytoskeleton, thus preventing them from retarding the mobility of the cell. It is too premature to speculate the underlying mechanisms responsible for increased motility in cells coexpressing vimentin and keratins. However, our report has provided a novel approach in studying the function of IFs, and has made a simple observation that enrichment for cells with enhanced ability to 112 migrate through small pores also selects for an increased frequency of keratin filament expression. It appears these cytoskeletal proteins have more complicated roles than just simply providing structural support in cells. They may, in fact, be a critical determinant in motility, affecting both normal and abnormal cell migration. 113 Acknowledgements: The authors would like to gratefully acknowledge the scientific contributions of Dr. David SolI and the Dynamic Morphology System (DMS) of the Cell Motility Center at the University of Iowa. We also thank Dr. Richard E.B. Seftor and Mr. Hanifa A. Jones for the preparation of the illustrations. This work was supported by Public Health Service grants CA-54984 (from the NCI to MJCH) and CA-42302 (from the NCI to ROO). 114 REFERENCE

1. Eichner, R., Bonitz, P., & Sun, T.-T. (1984) 1. Cell Bioi. 98, 1388-1396. 2. Steinert, P.M., & Roop, D.R. (1988) Ann. Rev. Biochem. 57,593-625. 3. Fuchs, E., & Marchuk, D. (1983) Proc. Nat!. Acad. Sci. USA 80,5857-5861. 4. Eichner, R, Sun, T.-T., & Aebi, U. (1986) 1. Cell Bioi. 102, 1767-1777. 5. Steinert, P.M. (1990) 1. Bioi. Chem. 265,8766-8774. 6. Franke, W.W., Schmid, E., Schiller, D.L., Winter, S., Jarasch, E.D., Moll, R, Denk, H., Jackson, B.W., & Illmensee, K. (1981) Cold Spring Harbor Symp. Quant. Bioi. 46, 431-453. 7. van Muijen, G.N.P., Warnaar, S.O., & Ponec. M. (1987) Exp. Cell Res. 171, 331-345. 8. Fuchs, E. (1990) 1. Cell Bioi. 111,2807-2814. 9. Klymkowsky, M.W., Bachant, J.B., & Domingo, A. (1989) Cell Moti!. Cytoskeleton 14, 309-331. 10. Oshima, R.G. (1992) Curro Opin. Cell Bioi. 4, 110-116. 11. Jones, J.C.R., & Goldman, R.D. (1985) 1. Cell Bioi. 101,506-517. 12. Quaranta, V., & Jones, J.C.R (1991) Trends Cell Bioi. 1,2-4. 13. Coulombe, P.A., Hutton, M.E., Letai, A., Hebert, A., Paller, A.S., & Fuchs, E. (1991) Cell 66, 1301-1311. 14. Vassar, R., Coulombe, P.A., Degenstein, L., Albers, K. & Fuchs, E. (1991) Cell 64, 365-380. 15. Osborn, M., & Weber, K. (1982) Cell 31, 303-306. 16. Osborn, M., & Weber, K. (1986) Trends Biochem. Sci. 11,469-472. 115 17. Trask, D.K., Band, V., Zajchowski, D.A., Yaswen, P., Suh, T., & Sager, R. (1990) Proc. Natl. Acad. Sci. USA 87, 2319-2323. 18. Ben-Ze'ev, A., Zoller, M., & Raz, A. (1986) Cancer Res. 46, 785-790. 19. Ramaekers, F.C.S., Haag, D., Kant, A., Moesker, 0., Jap, P.H.K., & Vooijs, G.P. (1983) Proc. Nat!. Acad. Sci. USA 80, 2618-2622. 20. GUnther, A., Kinjo, M., Winter, H., Sonka, J. & Volm, M. (1984) Cancer Res. 44, 2590-2594. 21. Thompson, E.W., Paik, S., Brunner, N., Sommers, c.L., Zugmaier, G., Clarke, R., Shima, T.B., Torti, J., Donahue, S., Lippman, M.E., Martin, G.R., & Dickson, R.B. (1992) J. Cell Physio!. 150, 534-544. 22. Trejdosiewicz, L.K., Southgate, J., Kemshead, J.T., & Hodges, a.M. (1986) Exp. Cell Res. 164, 388-398. 23. Miettinen, M., & Franssila, K. (1989) Lab. Invest. 61,623-628. 24. Zarbo, R.I., Gown, A.M., Nagle, R.B., Visscher, D.W., & Crissman, J.D. (1990) Modern Patho!. 3, 494-501. 25. Hendrix, M.I.C., Seftor, E.A., Chu, Y.-W., Seftor, R.E.B., Nagle, R.B., McDaniel, K.M., Leong, S.P.L., Yohem, K.H., Leibovitz, A.M., Meyskens, F.L., Conaway, D.H., Welch, D.R., Liotta, L.A. & Stetler-Stevenson, W. (1992) 1. Nat!. Cancer Inst. 84, 165-174. 26. Kulesh, D.A., Cecelia, G., Darmon, Y.M., Vasseur, M. & Oshima, R.G. (1989) Mol. Cell. Bioi. 9, 1553-1565. 27. Kulesh, D.A., & Oshima, R.G. (1988) Mol. Cell. Bioi. 8, 1540-1550. 28. Hendrix, M.J.C., Seftor, E.A., Seftor, R.E.B., & Fidler, 1J. (1987) Cancer Lett. 38, 137-147. 116 29. Hendrix, M.J.C., Seftor, E.A., Seftor, R.E.B., Misiorowski, R.L., Saba, P.Z., Sundareshan, P. & Welch, D.R. (1989) Invas. Metas. 9,278-297. 30. McCarthy, J.B., Palm, S.L., & Furcht, L.T. (1983) J. Cell Bioi. 97, 772-777. 31. SoIl, D.R. (1988) Cell MotU. Cytoskeleton 10, 91-106. 32. Seftor, E.A., Seftor, R.E.B., & Hendrix, M.1.C. (1990) BioTechniques 9,324- 331. 33. Domenjoud, L., Jorcano, J.L., Breuer, B., & Alonso, A. (1988) Exp. Cell Res. 179,352-361. 34. Liotta, L.A. (1986) Cancer Res. 46, 1-7. 35. Nabi, I.R., Watanabe, H., & Raz, A. (1990) Cancer Res. 50,409-414. 36. Lester, B.R., McCarthy, J.B., Sun, Z., Smith, R.S., Furcht, L.T., & Spiegel, A.M. (1989) Cancer Res. 49, 5940-5948. 37. Collier, I.E., Wilhelm, S.1\1., Eisen, A.Z., Marmer, B.L., Grant, G.A., Seltzer, IL., Kronberger, A., He. C., Bauer, E.A. & Goldberg, G.I. (1988)J. Bioi. Chern. 263, 6579-6587. 38. Matrisian, L.M., Bowden, G.T., Krieg, P., Fiirstenberger, G., Briand, J.-P., Leory, P., & Breathnach, R. (1986) Proc. Nat!. Acad. Sci. USA 83, 9413- 9417. 39. Quax, P.H.A., van Muijen, G.N.P., Weening-Verhoeff, E.D.1., Lund, L.R., Danp, Ruiter, D.1., & Verheijen, J.H. (1991) J. Cell Bioi. 115, 191-199. 40. Baribault, H., & Oshima, R.G. (1991)J. Cell Bioi. 115, 1675-1684. 41. Chu, Y.-W., Duffy, J.1., Seftor, R.E.B., Nagle, R.B., & Hendrix, M.J.C. (1991) CUn. Biotechnology 3, 27-33. 42. Klymkowsky, M.W., Miller, R.H., & Lane, E.B. (1983) J. Cell Bioi. 96,494- 509. 117 43. Robey, H.L., Hiscott, P.S., & Grierson, I. (1992) 1. Cell Sci. 102,329-340. 44. Tullberg, K.F., & Burger, M.M. (1985) Invas. Metas. 5, 1-15. 45. Ochalek, T., Nordt, F.J., Tullberg, K. & Burger, M.M. (1988) Cancer Res. 48, 5124-5128. 46. Miller, R.K., Vikstrom, K., & Goldman, R.D. (1991) 1. Cell Bioi. 113,843-855. 47. Skalli, 0., & Goldman, R.D. (1991) Cell Moti!. Cytoskeleton 19,67-79. 48. Kleinman, H.K., McGarvey, M.L., Liotta, L.A., Robey, P.G., Tryggvason, K., & Martin, G.R. (1982) Biochemistry 21,6188-6193. 49. Seftor, R.E.B., Seftor, E.A., Gehlsen, K.R., Stetler-Stevenson, W.G., Brown, P.D., Ruoslahti, E., & Hendrix, M.J.C. (1992) Proc. Nat!. Acad. Sci. USA 89, 1557-1561. 50. Joseph, L.B., Kreutzer, D.L., & Tanzer, M.L. (1990) 1. Cell Sci. 96,519-526. 51. Raz, A., & Geiger, B. (1982) Cancer Res. 42,5183-5190. 52. Volk, T., Geiger, B., & Raz, A. (1984) Cancer Res. 44,811-824. 53. Markl, J. (1991) 1. Cell Sci. 98,261-264. 54. Lane, E.B., Hogan, B.L.M., Kurkinen, M., & Garrels, 1.I. (1983) Nature 303, 701-704. 118

Fig. 1. Immunofluorescence localization of keratins 8 and 18. Cells were grown on glass coverslips for 24 h, fixed with methanol, and stained with mouse anti­ human simple epithelial keratin antibody followed by rhodamine conjugated goat anti-mouse IgG. The cell lines are designated: parental L cells (L); L cells transfected with keratin 8 cDNA (LK8); L cells transfected with keratin 18 DNA (LKI8); L cells transfected with keratin 18 DNA and keratin 8 cDNA (LKI8+K8). (xI890). 119

Fig. 2. Invasive and migratory ability of L cells. Parallel-lined bar graphs depict the percent invasion of L (9.28 ± 1.77 %), LK8 (13.94 ± 3.60 %), LK18 (10.61 ± 1.86 %) and LKI8+K8 (32.54 ± 2.80 %) cells measured over 72 h using Matrigel­ coated filters. Cross-hatched bar graphs show the percent migration of L (5.47 ± 0.15 %), LK8 (5.98 ± 1.37 %), LK18 (6.26 ± 0.41 %) and LKI8+K8 (18.39 ± 1.98 %) cells measured over 6 h using gelatin-coated filters. Standard error bars represent data generated from three experiments, each containing n=4. 120

Fig. 3. Selection of keratin-positive cells and their migratory abilities. The migratory cells from two heterogeneous LK18+K8 clones: clone 7 (A) and clone 5 (B) were selected by their ability to move through the polycarbonate filter. The parental clone 7 containing 44.2% keratin-positive cells has 15.0 ± 1.3% migration rate; 58.3% (clone 7-1) keratin-positivity following one migration passage resulting in 18.7 ± 1.5% migration; 68.0% (clone 7-2) keratin-positivity after second selection give 20.2 ± 1.8% migration; and 74.8% (clone 7-3) keratin­ positivity after third selection achieve 24.4 ± 2.4% migration ability. The parental clone 5, starting with 33.9% keratin positive cells, has 3.44 ± 0.85% migration rate; one round of selection results in 40.0% (clone 5-1) keratin-positivity and 4.17 ± 0.57% migration; after the second selection, keratin positive cells increase to 46.5% (clone 5-2) and the migration rate is 6.45 ± 0.68%; the third selection demonstrates a population of 57.4% (clone 5-3) keratin-positivity with migration increasing to 8.74 ± 0.69%. 121

Fig. 4. Morphology of L cells on Matrigel. Phase contrast micrographs of parental L cells compared with LK1S+KS cells after 12, 24, 36 and 72 h attachment and spreading on Matrigel. Percent roundness (%) from an average of 7-27 cells measured at each time point and parameter with the Dynamic Morphology System. Standard error for each value is equal to or less than 10% of the calculated % roundness. 122 123

L

LKS

LK1S

LK1S+KS

40 30 20 10 0 5 10 15 20 25 Percent Invasion Percent Migration -~

A 30T 0: Clone 7 8: Clone 7-1 6: Clone 7-2 A: Clone 7-3

25 I 1 z 0 I-« n::: -C) 20 :2

~ 15

10+1----~----~----~~----r-----+-----~-----r----~ 40 45 50 55 60 65 70 75 80 % KERATIN POSITIVE CELLS In N

B 10 T 0: Clone 5 ~: Clone 5-1 6.: Clone 5-2 A.: Clone 5-3 8'

z 0 6 «f- a:: (9 :2 4

~ 2

O+I------r------r----~------+_----_+------+_----~ 25 30 35 40 45 50 55 60

% KERATIN POSITIVE CELLS 126

L ._.L~JB±X8

o

Q .• 0 e 12h. ~ 0 , " ...~ 13 ..:l ·c

80% 87% .... L-______... i. ____. . _ . ,.~ - ---_.-. ----_._----- r..····. ----...... ~~. .'.J. I:~~ . . . o'~· I, t I .', 24 I .0 .' I I I I I c ! 52% I

, ; .. to

36 \ I' .: ... .. /.,

. 43% L-.______, -1. ____ 54%

72

::~·.i~~: ",.' . .. t " I I >-.:..-: ,;. -'~: .. 18% ,.-.. , 127 128 Submitted To: NATURE

Coexpression of Vimentin and Keratins Intermediate Filaments in Human Melanoma Cells Correlates with Increased Motility and Decreased Integrin Localization in Focal Contacts

Yi-Wen Chu*, Robert G. Oshimat, Erkki Ruoslahtit, and Mary J.e. Hendrix:j:§

*Cancer Biology Program and :j:Department of Anatomy, College of Medicine, University of Arizona, Tucson, Arizona 85724, USA; and tLa Jolla Cancer Research Foundation, La Jolla, California 92037, USA

§To whom correspondence should be addressed: Dr. Mary J.C. Hendrix Department of Anatomy College of Medicine University of Arizona Tucson, AZ 85724 U.S.A. TEL: (602) 626-7893 FAX: (602) 626-2180 129 ABSTRACT Intermediate filaments (IFs) have been used as cell type-specific markers in differentiation and pathology; however, recent reports have demonstrated the coexpression of vimentin (a mesenchymal marker) and keratins (an epithelial markers) in numerous neoplasms, including melanoma, which has been linked to metastatic disease1-9. To test the hypothesis that dual IF expression of vimentin and keratins by melanoma cells contributes to a more migratory and invasive phenotype, we transfected a vimentin-positive human melanoma cell line, A375P (of low invasive ability), with DNAs for both keratins 8 + 18. The resultant stable transfectants express vimentin- and keratin-positive IFs, and show a two-to-three fold increase in their ability to invade basement membrane matrix in vitro, commensurate with a two and half-fold increase in cell motility. Concomitantly, the transfectants containing dual IFs maintain a more rounded morphology and do not contain detectable levels of CJ.v P3, CJ.3 or CJ.6 integrins (receptors for vitronectin, collagen and laminin, respectively) in their focal contacts during the process of attachment and spreading on basement membrane matrix; however, the vimentin-positive control cell focal contacts are positive for these integrins. From these data, we postulate that a diminution in spreading ability combined with less focal contacts during the initial stages of tumor cell attachment contribute to a more migratory phenotype. 130 TEXT The functional role(s) of IPs is still unc1ear lO-12. It has been suggested that IPs can act as signal transducers, relaying information from the extracellular matrix (ECM) to the nUc1eus I3 . Furthermore, there is prevailing evidence demonstrating that the ECM can regulate gene expression, and, hence, modulate cellular function(s)14,15. Therefore, it is tempting to speculate that the ability to coexpress vimentin and keratin IPs offers a selective advantage to tumor cells in the interpretation of cues from mesenchymal and epithelial matrices. Classically, diagnosis of melanoma has relied on the presence of several protein markers, including vimentin5. Although earlier studies emphasized the use of IFs as cell type-specific markers, recent reports have confounded the literature with numerous demonstrations of the coexpression of vimentin and keratins in epithelial and nonepithelial neoplasmsl-9. Specifically, our laboratory and others have recently reported the coexpression of vimentin and keratins 8 and 18 in melanoma to be associated with recurrent and metastatic disease4,6. In a previous study, we attempted to inhibit keratin filament formation in a highly metastatic human melanoma cell line coexpressing vimentin and keratins by transfection with a dominant negative mutant keratin 18 cDNA4. The clones generated with this genetic mutation manifested aberrant keratin IPs and showed a dramatic reduction in invasive and migratory ability with complete abrogation of metastatic potential. Moreover, in an experimental L cell model, in which only vimentin is expressed, we have tested the invasive and migratory ability of clones transfected with keratins 8 and 18, and found that this dual expression of IPs resulted in a two-to three-fold increase in migratory activity, which correlated with reduced spreading ability on basement membrane matrix 16. Hence, in the 131 present study, we set forth to further test our hypothesis that vimentin and keratin dual IF expression contributes to a more invasive and migratory phenotype, via an unique modulation of cell shape and spreading ability on extracellular matrices, by overexpressing keratins 8+18 in a vimentin-positive, human melanoma A375P cell line (of low invasive potential). A375P cells were cotransfected with both cDNAs for keratins 8 and 18. A DNA slot blot analysis of the integration of the plasmid DNAs indicated that one clone (PK-14) contained approximately 10 copies of each plasmids, while other experimental clones demonstrated a lower copy number of both DNAs. (Data not shown.) Indirect immunofluorescence microscopy demonst.rates that the PK-14 clone is positive for keratins 8+18 (Fig. lA), as well as vimentin (Fig. IB); and P­ NEO control cells are negative for keratins 8+ 18 (Fig. 1C), but positive for vimentin (Fig. ID). Next, we analyzed the ability of these cells to migrate (over 6 h) through gelatin-coated filters, and to invade (over 48 h) through Matrigel­ coated filters (Matrigel is a reconstituted basement membrane matrix containing laminin, collagen IV, vitronectin, heparan sulfate proteoglycan, and entactin as solid matrix molecules17). As shown in Table 1, a two and half-fold increase in migratory ability exists between the PK-14 cells vs. the two P-NEO transfectants and the parental A375P cells. Moreover, a two-to-three fold increase in invasive ability is observed in the PK-14 cells compared with the P-NEO control transfectants. To further analyze the migratory ability of the cells on various substrata, video-cinematography was performed on cells seeded onto plastic, gelatin or Matrigel matrix (Table 2). These data demonstrate the rate of movement in j..lm/h and further corroborate the increased migratory ability of PK- 14 cells over the control transfectants (P-NEO-9), both on gelatin and Matrigel 132 substrata. However, no significant difference is seen between the control and experimental cells when they are seeded onto a tissue culture plastic substratum, which implicates the involvement of cell surface receptors and their extracellular matrix ligands. Since previous observations from our laboratory documented that vimentin and keratin dual IF in the L cell model directly correlated with decreased spreading ability on Matrigel16, we next measured the differential ability of P­ NEO and PK-14 cells to spread on Matrigel matrix over 2, 4, 6, 8 and 22 h after initial seeding (Fig. 2). These data were recorded by inverted phase photomicrographs, which show no significant difference at the early time points of 2 and 4 h, in which the cells are rounded; however, a dramatic difference is seen at the 6 and 8 h time intervals, where the PK-14 cells maintain a rounded morphology on Matrigel compared with the better spread P-NEO control cells. By 22 h, both cell populations have similar spread morphologies. The percent of rounded cells was calculated for each group during the spreading process, which correlated with the differential morphological configurations at 6 and 8 h. In order to better understand the mechanism(s) contributing to this differential spreading ability on Matrigel, as well as other matrices, we focused on measuring the pattern of specific integrins localized in the focal contacts, using indirect immunofluorescence microscopy (Fig. 3). We found, during the 6-8 h spreading period, a lack of detectable

and their linking proteins, particularly between the keratins and the

diminution in spreading ability on ECMs and a lack of detectable a.v~3,

The authors gratefully acknowledge the scientific expertise of Dr. Raymond Runyan in the analysis of cellular motility; and Dr. Richard Seftor and Mr. Hanifa Jones for their skillful printing of the photomicrographs. This work was supported by the National Cancer Institute, NIH. 136 REFERENCES 1. Raymond, W.A., & Leong, A.S.-Y. J. Pathol. 158, 107-114 (1989). 2. Osborn, M., & Weber, K. Cell 31, 303-306 (1982). 3. Franke, W.W., Schmid, E. & Moll, R In Human Carcinogenesis (eds Harris, C.C., & Autrup, H.N.) 3-34 (Academic Press, NY, 1983). 4. Hendrix, M.J.C. et al. J. Natl. Cancer Inst. 84, 165-174 (1992). 5. Miettinen, M., & Franssila, K. Lab. Invest. 61,623-628 (1989). 6. Zarbo, R.I., Gown, A.M., Nagle, RB., Visscher, D.W., & Crissman, J.D. Mod. Pathol. 3, 494-501 (1990). 7. Thompson, E.W. et al. J. Cell. Physiol. 150,534-544 (1992). 8. Ben-Ze'ev, A., Zoller, M., & Raz, A. Cancer Res. 46,785-790 (1986). 9. Ramaekers, F.C.S. et al. Proc. natn. Acad. Sci. USA 80,2618-2622 (1983). 10. Steinert, P.M., & Liem, RK. Cell 60, 521-523 (1990). 11. Oshima, R.G. Curro Opin. Cell Bioi. 4, 110-116 (1992). 12. Klymkowsky, M.W., Bachant, J.B., & Domingo, A. Cell Moti!. Cytoskeleton 14,309-331 (1989). 13. Skalli, 0., & Goldman, R.D. Cell Moti!. Cytoskeleton 19, 67-79 (1991). 14. Hay, E.D. Cell Moti!. Cytoskeleton 14, 455-457 (1989). 15. McDonald, J.A. Curro Opin. Cell Bioi. 1,995-999 (1989). 16. Chu, Y.-W., Runyan, RB., Oshima, RG.,& Hendrix, M.I.C. Proc. natn. Acad. Sci. USA (in press). 17. Kleinman, H.K. et al. Biochem. 24, 6188-6193 (1982). 18. Ruoslahti, E. J. CUn. Invest. 87, 1-5 (1991). 19. Tawil, N., Wilson, P., & Carbonetto, S. J. Cell Bioi. 120,261-271 (1993). 20. Quaranta, V., & Jones, J.C.R. Trend Cell Bioi. 1,2-4 (1991). 137 21. Wayner, E.A., Orlando, RA., & Cheresh, D.A. 1. Cell Bioi. 113,919-929 (1991). 22. Raz, A., & Ben-Ze'ev, A. Science 221, 1307-1310 (1983). 23. Repesh, L.A. et aI. 1. din. Exp. Metas. 11,91-102 (1993). 24. Lane, E.B., Hogan, B.L.M., Kurkinen, M., & Garrels, J,l. Nature 303, 701-703 (1983). 138 FIG. 1 Immunofluorescence staining of keratins 8+18 and vimentin IFs in experimental (PK-14) and control transfectants (P-NEO). A, PK-14 clone is positive for keratins 8+18, as well as vimentin, B. C, P-NEO control cells are negative for keratins 8+18, but positive for vimentin, D. For experimental transfection, plasmid LK442-K8 contains a human keratin 8 cDNA under the control of a human f3-actin promoter, and a gene that codes for E. coli xanthine­ guanine phosphoribosyltransferase (gpt) as a selection marker. Plasmid LK444- K18 has a similar construct, except that it has human keratin 18 cDNA and a neomycin resistance marker. For control transfection, LK444 vector, which has no cDNA insert, was used. A375P cells were cotransfected with LK442-K8 and LK444-K18 (at a ratio of 5:1) using Lipofectin Reagent (BRL, MD). The transfected cells were selected with G418 at I mg/ml concentration. Subsequently, 24 colonies from the transfected A375P cells (PK clones) and 25 colonies from the control transfection (P-NEO clones) were selected for further analysis. From these clones, PK-14 showed the strongest level of expression for keratins 8 and 18, and was used in the remainder of the experiments in addition to several P-NEO control transfectants. Indirect immunofluorescence microscopy was performed on PK-14 and P-NEO clones seeded onto glass coverslips to 70 % confluence, and then fixed in 100 % cold methanol for 7 min. Subsequently, the cells were treated with rabbit antibody to human vimentin (V9; DAKO Corp., CA), or mouse antibody to human keratin 18 (CK5; Sigma, MO), for I h at room temperature, followed by rinsing in PBS, and incubation with the appropriate secondary antibodies (conjugated to fluorescein or rhodamine; Organon Teknika, PA) for 1 h. Coverslips were then rinsed in PBS and mounted onto a glass microscope slide using gelvatol. Controls for nonspecific staining consisted of 139 incubation with either nonspecific rabbit or mouse IgG control antisera, followed by the appropriate secondary antibody. All samples were observed with a Zeiss epi fluorescence microscope, equipped with automatic rhodamine and fluorescein filter sets. Photographs of control and experimental samples were exposed for equal time using the manual exposure mode. (x 1890) 140 FIG. 2. Inverted light microscopy of the spreading ability of control transfectants (P-NEO) and experimental transfectants (PK-14, containing vimentin plus keratins 8+18 IFs) over 22 h post-seeding on Matrigel matrix (2 mg/ml; composed of laminin, collagen IV, heparin sulfate proteoglycan, vitronectin and entactin17). A, C, E, G and I are representative photomicrographs of P-NEO cells at 2, 4, 6, 8 and 22 h, respectively, spread on Matrigel. B, D, F, H, and J are typical PK-14 cells at 2, 4, 6, 8 and 22 h, respectively, on Matrigel matrix. Percentage of rounded cells (%) was calculated for each time point during the spreading process on an average of 25-30 cells per group. All photomicrographs were recorded with a Pentax ME, mounted onto a WILD inverted light microscope. (x 400) 141 FIG. 3. Immunofluorescence localization of integrins in focal contacts during spreading on Matrigel after 8 h. A, C, E are representative photographs of P-NEO cells stained for avf33, a3 and

P-NEO PK-14

'i'

2h.

I: t 100 ----_.. _._-_ ..

. ".

4

58 100

6

100

8

22 146

P-NEO PK-14 147

TABLE 1 Migration and Invasion Abilities

% Migration * % Invasion t

A375P 0.56 ± 0.17 % 1.49 ± 0.21 %

P-NEO-9 0.66 ± 0.12 % 1.18 ± 0.07 %

P-NEO-23 0.67 ± 0.10 % 1.69 ± 0.38 % PK-14 1.71 ± 0.01 % 3.61 ± 0.24 % 148

TABLE 2 Migration Rate on Different Substrata Measured By Video Cinematography (Jlm/h) * Plastic Gelatin Matrigel

P-NEO-9 6.51 ± 0.80 5.04 ± 0.69 4.30 ± 0.56 (n=9) (n=12) (n=11)

PK-14 5.01 ± 0.52 8.50 ± 0.75 11.54 ± 1.22 (n=ll) (n=10) (n=13) 149 REFERENCES

Abe, M., and Oshima, R.G. 1990. A single human keratin 18 gene is expressed in diverse epithelial cells of transgenic mice. J. Cell BioI. 111: 1197-1206. Albelda, S.M., and Buck, C.A. 1990. Integrins and other cell adhesion molecules. FASEB J. 4: 2868-2880. Albelda, S.M., l\1ette, S.A., Elder, D.E., Stewart, R., Damjanovich, L., Herlyn, M. and Buck, C.A. 1990. Integrin distribution in malignant melanoma: association of the J33 subunit with tumor progression. Cancer Res. 50: 6757-6764. Alexander, C.M., and Werb, Z. 1992. Targeted disruption of the tissue inhibitor of metalloproteinases gene increases the invasive behavior of primitive mesenchymal cells derived from embryonic stem cells in vitro. J. Cell BioI. 118: 727-739. Arch, R., Wirth, K., Hofmann, M., Ponta, H., Matzh..'U, S., Herrlich, P., and ZOller, M. 1992. Participation in normal immune responses of a metastasis-inducing splice variant of CD44. Science 257: 682-685. Ball, E.H., and Singer, SJ. 1981. Association of microtubules and intermediate filaments in normal fibroblasts and its disruption upon transformation by a temperature-sensitive mutant of Rous sarcoma virus. Proc. NatI. Acad. ScLUSA 78: 6986-6990. Banks-Schlegel, S.P., and Rhim, J.S. 1986. Keratin expression of both chemically and virally transformed human epidermal keratinocytes during the process of neoplastic conversion. Carcinogenesis 7: 153-157. Baribault, H., and Oshima, R.G. 1991. Polarized and functional epithelia can form after the targeted inactivation of both mouse keratin 8 alleles. J. Cell BioI. 115: 1675-1684. 150 Ben-Ze'ev, A., and Raz, A. 1985. Relationship between the organization and synthesis of vimentin and the metastatic capability of B 16 melanoma cells. Cancer Res. 45: 2632-2641. Ben-Ze'ev, A., Zoller, M., and Raz, A. 1986. Differential expression of intermediate filament proteins in metastatic and non metastatic variants of the BSp73 tumor. Cancer Res. 46: 785-790. Blessing, M., Jorcano, J.L., and Franke, W.W. 1989. Enhancer elements directing cell-type-specific expression of cytokeratin genes and changes of the epithelial cytoskeleton by transfections of hybrid cytokeratin genes. EMBO J. 8: 117-126. Bloch, R.J. 1992. Clusters of neural cell adhesion molecule at sites of cell-cell contact. J. Cell BioI. 116: 449-463. Brinkley, B.R., Cox, S.M., Pepper, D.A., Wible, L., Brenner, S.L., and Pardue, R.L. 1981. Tubulin assembly sites and the organization of cytoplasmic microtubules in cultured mammalian cells. J. Cell BioI. 90:554-562. Bussemakers, M.J.G., van Moorselaar, RJ.A., Giroldi, L.A., Ichikawa, T., Isaacs, J.T., Takeichi, M., Debruyne, F.MJ., and Schalken, J.A. 1992. Decreased expression of E-cadherin in the progression of rat prostatic cancer. Cancer Res. 52: 2916-2922. Cajot, J.-F., Schleuning, W.-D., Medcalf, R.L., Barnat, J., Testuz, J., Liebermann, L., and Sordat, B. 1989. Mouse L cells expressing human prourokinase-type plasminogen activator: effects on extracellular matrix degradation and invasion. J. Cell BioI. 109: 915-925. Carmichael, D.F., Sommer, A., Thompson, R.C, Anderson, D.C., Smith, CG., Welgus, H.G., and Stricklin, G.P. 1986. Primary structure and cDNA cloning of 151 human fibroblast collagenase inhibitor. Proc. Natl. Acad. Sci. USA 83: 2407- 2411. Chou, C.F., and Omary, M.B. 1991. Phorbol acetate enhances the phospgorylation of cytokeratins 8 and 18 in human colonic epithelial cells. FEBS Lett. 282: 200-204. Chou, Y.-H., Bischoff, J.R., Beach, D., and Goldman, R.D. 1990. Intermediate filament reorganization during mitosis is mediated by p34Cdc2 phosphorylation of vimentin. Cell 62: 1063-1071. Cooper, J.A. 1987. Effects of cytochalasin and phalloidin on actin. J. Cell BioI. 105: 1473-1478. Coulombe, P.A., and Fuchs, E. 1990. Elucidating the early stages of keratin filament assembly. J. Cell BioI. 111: 153-169. Coulombe, P.A., Hutton, M.E., Letai, A., Hebert, A., Paller, A.S., and Fuchs, E. 1991. Point mutations in human keratin 14 genes of epidermolysis bullosa simplex patients: genetic and functional analyses. Cell 66: 1301-1311. DeBrabander, M., Geuens, G., Nuydens, R., Willebrords, R., and DeMary, J. 1982. Microtubule stability and assembly in living cells: the influence of metabolic inhibitors, taxol and pH. Cold Spring Harbor Symp. Quant. BioI. 46: 227-240. DeClerck, Y.A., Yean, T.-D., Chan, D., Shimada, H., and Langley, K.E. 1991. Inhibition of tumor invasion of smooth muscle cell layers by recombinant human metalloproteinase inhibitor. Cancer Res. 51: 2151-2157. Djabali, K., Portier, M.-M., Gros, E, Blobel, G., and Georgatos, S.D. 1991. Network antibodies identify nuclear lamin B as a physiological attachment site for peripherin intermediate filaments. Cell 64:109-121. 152 Drubin, D.G., and Kirschner, M. 1986. function in living cells. J. Cell BioI. 103: 2739-2746. Dustin, P. 1984. Microtubules. Springer-Verlag, New York. EI-Badry, O.M., Minniti, c., Kohn, E.C., Houghton, PJ., Daughaday, W.H., and Helman, L.J. 1990. Insulin-like growth factor II acts as an autocrine growth and motility factor in human rhabdomyosarcoma tumors. Cell Growth Differ. 1: 325-331. Ferretti, P., Fekete, D.M., Patterson, M., and Lane, E.B. 1989. Transient expression of simple epithelial keratins by mesenchymal cells of regenerating newt limb. Dev. BioI. 133: 415-424. Fidler,I.J. 1973. Selection of successive tumor lines for metastasis. Nat. New BioI. 242: 148-149. Fidler,I.J. 1990. Critical factors in the biology of human cancer metastasis: twenty-eight G.H.A. Clowes memorial award lecture. Cancer Res. 50: 6130- 6138. Fidler, I.J., and Hart, I.R. 1982. Biologic diversity in metastatic neoplasms- origins and implications. Science 217: 998-1001. Foisner R, and Wiche, G. 1991. Intermediate filament-associated proteins. Curro Opin. Cell BioI. 3: 75-81. Franke, W.W., Schmid, E., Schiller, D.L., Winter, S., Jarasch, E.D., Moll, R, Denk, H., Jackson, B.W., and Illmensee, K. 1982. Differentiation-related patterns of expression of proteins of intermediate-sized filaments in tissue and cultured cells. Cold Spring Harbor Symp. Quant. BioI. 46: 431-454. 153 Frixen, V.H., Behrens, 1., Sachs, M., Eberle, G., Voss, B., Warda, A., Lochner, D., and Birchmeier, W. 1991. E-cadherin-mediated cell-cell adhesion prevents invasiveness of human carcinoma cells. J. Cell BioI. 113: 173-185. Fuchs, E. 1990. Epidermal differentiation: the bare essentials. J. Cell BioI. 111: 2807-2814. Fuchs, E., and Coulombe, P.A. 1992. Of mice and men: genetic skin diseases of keratin. Cell 69: 899-902. Garbisa, S., Scagliotti, G., Masiero, L., Francesco, C.D., Caenazzo, C., Onisto, M., Micela, M., Stetler-Stevenson, W.G., and Liotta, L.A. 1992. Correlation of serum metalloproteinase levels with lung cancer metastasis and response to therapy. Cancer Res. 52: 4548-4549. Gatter, K.C., Ralfldaer, E., Skinner, J., Brown, D., Heryet, A., Pulford, K.A.F., Hou­ Jensen, K., and Mason, D.Y. 1985. An immunocytochemical study of malignant melanoma and its differential diagnosis from other malignant tumours. J. Clin. PathoI. 38: 1353-1357. Gehlsen, K.R., Argraves, W.S., Pierschbacher, M.D., and Ruoslahti, E. 1988. Inhibition of in vitro tumor cell invasion by arg-gly-asp-containing synthetic peptides. J. Cell BioI. 106: 925-930. Gehlsen, K.R., Davis, G.E., and Sriramarao, P. 1992. Integrin expression in human melanoma cells with differing invasive and metastatic properties. Clin. Exp. Metastasis 10: 111-120. Georgatos, S.D., and Blobel, G. 1987a. Two distinct attachment sites for vimentin along the plasma membrane and the nuclear envelope in avian erythrocytes: a basis for a vectorial assembly of intermediate filaments. J. Cell BioI. 105: 105- 115. 154 Georgatos, S.D., and Blobel, G. 1987b. Lamin B constitutes an intermediate filament attachment site at the nuclear envelope. J. Cell BioI. 105: 117-125. Gherardi, E., Gray, J., Stoker, M., Perryman, M., and Furlong, R. 1989. Purification of scatter factor, a fibroblast-derived basic protein that modulates epithelial interactions and movement. Proc. Natl. Acad. Sci. USA 86: 5844-5848. Giudice, G.J., and Fuchs, E. 1987. The transfection of epidermal keratin genes into fibroblasts and simple epithelial cells: evidence for inducing a type I keratin by a type II gene. Cell 48: 453-463. Goldman, A.E., Maul, G., Steinert, P.M., Yang, H.-Y., and Goldman, R.D. 1986. Keratin-like proteins that coisolate with intermediate filaments of BHK-21 cells are nuclear lamins. Proc. Natl. Acad. Sci. USA 83: 3839-3843. Gomez, M., Navarro, P., Quintanilla, M., and Cano, A. 1992. Expression of

Rothnagel, J.A., Dominey, A.M.~ Dempsey, L.D., Longley, M.A., Greenhalgh, D.A., Gagne, T.A., Huber, M., Frenk, E., Hohl, D., Roop, D.R. 1992. Mutations in the rod domains of keratins 1 and 10 in epidermolytic hyperkeratosis. Science 257: 1128-1130. Ruoslahti, E. 1991. Integrins. J. Clin. Invest. 87: 1-5. Seftor, E.A., Seftor, R.E.B., and Hendrix, M.1.C. 1990. Selection of invasive and metastatic subpopulations from a heterogeneous human melanoma cell line. Biotechniques 9: 324-331. Seftor, R.E.B., Seftor, E.A., Gehlsen, K.R., Stetler-Stevenson, W.G., Brown, P.D., Ruoslahti, E., and Hendrix, M.J.C. 1992. Role of the uv/33 integrin in human melanoma cell invasion. Proc. Natl. Acad. Sci. USA 89: 1557-1561. Semart, A., Duprey, P., Vasseur, M., and Darmon, M. 1986. Mesenchymal-epithelial conversions induced by 5-azacytidine: appearance of cytokeratin Endo-A messenger RNA. Differentiation 31:61-66. Silletti, S., Watanabe, H., Hogan, V., Nabi, I.R., and Raz, A. 1991. Purification of BI6-Fl melanoma autocrine motility factor and its receptor. Cancer Res. 51: 3507-3511. Sreenath, T., Matrisian, L.M., Stetler-Stevenson, W., Gattone-Celli, S., and Pozzatti, R.O. 1992. Expression of matrix metalloproteinase genes in transformed rat cell lines of high and low metastatic potential. Cancer Res. 52: 4942-4947. Steeg, P.S., Bevilacqua, G., Kopper, L., Thorgeirsson, U.P., Talmadge, J.E., Liotta, L.A., and Sobel, M.E. 1988. Evidence for a novel gene associated with low tumor metastatic potential. J. Natl. Cancer Inst. 80: 200-204. 162 Steinert, P.M., and Roop, D.R. 1988. Molecular and cellular biology of intermediate filaments. Ann. Rev. Biochem. 57: 593-625. Steinert, P.M., Jones, J.C.R., and Goldman, R.D. 1984. Intermediate filaments. J. Cell BioI. 99: 22s-27s. Steinert, P.M., Steven, A.C., and Roop, D.R. 1985. The molecular biology of intermediate filaments. Cell 42: 411-419. Stetler-Stevenson, W.G., Krutzsch, H.L., and Liotta, L.A. 1989. Tissue inhibitor of metalloproteinases (TIMP-2). J. BioI. Chern. 264: 17374-17378. Strauli, P., and Haemmerli, G. 1984. The role of cancer cell motility in invasion. Cancer Metas. Rev. 3: 127-141. Tawil, N., Wilson, P., and Carbonetto, S. 1993. Integrins in point contacts mediate cell spreading: factors that regulate integrin accumulation in point contacts vs. focal contacts. J. Cell BioI. 120: 261-271. Templeton, N.S., and Stetler-Stevenson, W.G. 1991. Identification of a basal promoter for the human Mr 72,000 type IV collagenase gene and enhanced expression in a highly metastatic cell line. Cancer Res. 51: 6190-6193. Thompson, E.W., Paik, S., Brunner, N., Sommers, C.L., Zugmaier, G., Clarke, R., Shima, T.B., Tom, J., Donahue, S., Lippman, M.E., Martin, G.R., and Dickson, R.B. 1992. Association of increased basement membrane invasiveness with absence of estrogen receptor and expression of vimentin in human breast cancer cell lines. J. Cell. PhysioI. 150: 534-544. Torpey, N., Wylie, C.C., and Heasman, J. 1992. Function of maternal cytokeratin in Xenopus development. Nature 357: 413-415. 163 Trask, D.K., Band, V., Zajchowski, D.A., Yaswen, P., Suh, T., and Sager, R. 1990. Keratins as markers that distinguish normal and tumor-derived mammary epithelial cells. Proc. Natl. Acad. Sci. USA 87: 2319-2323. Trejdosiewicz, L.K., Southgate, J., Kemshead, J.T., and Hodges, G.:M., 1986. Phenotypic analysis of cultured melanoma cells: expression of cytokeratin­ type intermediate filaments by the M5 human melanoma cell line. Exp. Cell Res. 164: 388-398. Tryggvason, K., Hoyhtya , M., and Salo, T. 1987. Proteolytic degradation of extracellular matrix in tumor invasion. Biochim. Biophys. Acta 907: 191-217. Tullberg, K.F., and Burger, M.M. 1985. Selection of B16 melanoma cells with increased metastatic potential and low intercellular cohesion using nucIepore filters. Invasion Metastasis 5: 1-15. van Muijen, G.N.P., Warnaar, S.O., and Ponee, M. 1987a. Differentiation-related changes of cytokeratin expression in cultured keratinocytes and in fetal, newborn, and adult epidermis. Exp. Cell Res. 171: 331-345. van Muijen, G.N.P., Ruiter, D.J., and Warnaar, S.O. 1987b. Coexpression of intermediate filament polypeptides in human fetal and adult tissues. Lab. Invest. 57: 359-369. Vassar, R., Rosenberg, M., Ross, S., Tyner, A., and Fuchs, E. 1989. Tissue specific and differentiation-specific expression of a human K14 keratin gene in transgenic mice. Proc. Natl. Acad. Sci. USA 86: 1563-1567. Venetianer, A., Schiller, D.L., Magin, T., and Franke, W.W. 1983. Cessation of cytokeratin expression in a rat hepatoma cell line lacking differentiated functions. Nature 305: 730-733. 164 Verschueren, H., Dekegel, D., and DeBaetselier, P. 1988. Difference in motile behavior between lymphoma variants with different invasive and metastatic capabilities. Invasion Metastasis 8: 31-44. Volk, T., Geiger, B., and Raz, A. 1984. Motility and adhesive properties of high­ and low-metastatic murine neoplastic cells. Cancer Res. 44: 811-824. Ward, G.E., and Kirschner, M.W. 1990. Identification of cell cycle-regulated phosphorylation sites on nuclear lamin C. Cell 61: 561-577. Yamada, K.M., Kennedy, D.W., Yamada, S.S., Gralnick, H., Chen, W.-T., and Akiyama, S.K. 1990. Monoclonal antibody and synthetic peptide inhibitors of human tumor cell migration. Cancer Res. 50: 4485-4496. Zarbo, R.J., Gown, A.M., Nagle, R.B., Visscher, D.W., and Crissman, J.D. 1990. Anomalous cytokeratin expression in malignant melanoma: one and two­ dimensional Western blot analysis and immunohistochemical survey of 100 melanomas. Modem PathoI. 3: 494-501.