PROOF

Autocrine and ParacrineGrowth Mechanisms in Cancer

Progression and MetastasisGarth L. Nicolson

The Institute for Molecular Medicine, Huntington Beach, California

Encyclopedia of Cancer, Second Edition Copyright 2002, Elsevier Science (USA).Volume 1 165 All rights reserved.

I. IntroductionII. Cancer Progression and Paracrine Growth of

Metastatic CellsIII. Paracrine Growth Inhibitors of Metastatic CellsIV. Sources of Paracrine Growth Factors

and InhibitorsV. Autocrine Growth Mechanisms in

Cancer MetastasisVI. Gene Transfer and Metastatic Cell

Growth PropertiesVII. New Approaches to Cancer TherapyVIII. Growth Responses of Metastatic Cells and

Malignant Progression

GLOSSARY

autocrine growth factors Growth factors that are made byand that act on the same cell.

metastasis The spread of tumor cells via the lymph, blood,or body cavities to near or distant sites where new sec-ondary tumors are formed.

paracrine growth factors Growth factors that are madeand secreted by one cell and that act on adjacent cells in atissue or organ.

tumor diversification Generation of heterogeneous sub-populations of tumor cells with differing phenotypes.

tumor instability Changes in tumor cell properties causedby irreversible modifications in the coding sequences ofgenes and by quantitative changes in gene expression.

tumor progression Sequential changes in tumorigenic andmalignant properties of tumors that occur with time in vivo.The changes generally tend toward more malignant, dan-gerous states.

O ne of the most important characteristics of can-cer cells is their ability to grow in unusual locations,especially at metastatic sites. The successful prolifer-ation of cancer cells is due to their responses to local(paracrine) growth factors and inhibitors and theirproduction and responses to their own (autocrine)growth factors. As tumors grow and evolve (tumor

9905.ch018 1/22/02 8:46 AM Page 165

PROOF

progression), they undergo changes in their growthand other properties. For example, when tumor cellsinvade and spread to other sites at the early stages ofmalignant tumor progression, there is a tendency formany common cancers to metastasize and grow pref-erentially at particular sites, suggesting that uniquetissue paracrine growth mechanisms may dominatethe growth signals processed by metastatic cells. Atsomewhat later stages of tumor progression, wherewidespread dissemination to various tissues and or-gans occurs, autocrine growth mechanisms may dom-inate. The progression of malignant cells to com-pletely autonomous growth states can occur, and atthis stage of tumor progression cell proliferation maybe independent of growth factors or inhibitors.

I. INTRODUCTION

Most patients succumb to their metastatic disease,not their primary tumors; therefore, controlling thespread and growth of malignant cells at metastaticsites is an important challenge. Malignant tumors arecharacterized by differences in various properties, andthose that are functionally involved in invasion andmetastasis are among the most important propertiesof cancer cells that determine their survival andgrowth at secondary sites.

In addition to their invasive and metastatic prop-erties, highly malignant cells are characterized by pro-gressive changes in their genomes, particularly ingenes that regulate and encode products important incertain phenotypic properties. Highly malignant cellsare not particularly stable and they continually driftin their phenotypic properties. In some cases, suchphenotypic drift in cancer cell properties is virtuallyundetectable, but in other cases it can be dramaticand result in obvious tumor cell heterogeneity. Ad-vanced primary tumors that have not yet metastasizedand cancers that have metastasized are made up ofvery dynamic, unstable cellular assemblages. The in-dividual cancer cells that are unstable do not alwaysundergo phenotypic drift. In some cases they interactwith other tumor and normal cells that can stabilizethe cell population as a whole and reduce the ten-dency for individual cells to diversify and becomemore variable in a variety of properties.

If malignant cells break loose from their primarysite as individual cells or small groups of cells, someof these cells can diversity further and become evenmore heterogeneous in their properties. As stated ear-lier, this can occur because individual cells are mod-ulated in their phenotypic properties by interactionswith other cells. Once removed from these interac-tions, cancer cells undergo phenotypic diversification.Therefore, the invasion of individual cells away fromthe initial tumor site and into new tissue compart-ments can lead to cancer cells with differing proper-ties from the primary tumors from which they werederived. However, most cellular properties do notchange detectably as tumor cells in advanced primarytumors metastasize to distant sites. This may be dueto the fact that advanced primary tumors have al-ready undergone significant diversification andchange.

Tumor cells progress to the malignant state andeventually the metastatic state in a process that isthought to occur by a stepwise series of genetic andepigenetic (nongenetic) changes. These changes oc-cur apparently randomly, and accumulating enoughchanges necessary for a cancer cell to become highlymetastatic can take years. Such metastatic cells werepreviously thought to be very rare cells that occa-sionally arose within the primary tumor, and it wasthought that only these rare phenotypically stablecells were the cells capable of metastasizing. Althoughtumor progression can result in stepwise nonreversiblechanges and acquisition of a number of phenotypicproperties that are important in the process of metas-tasis, it is now thought that tumor progression to themetastatic phenotype depends less on the selection ofrare, stable metastatic cells than on the inherent in-stabilities of the tumor cells that comprise the pri-mary neoplasm. This notion was advanced by VictorLing, Ann Chambers, Richard Hill, and their col-leagues at the Ontario Cancer Institute. They envis-aged that malignant tumor cells are constantly un-dergoing rapid dynamic phenotypic changes in aprocess they called dynamic heterogeneity. Indepen-dently from the group in Canada, my colleagues andI at the University of Texas M. D. Anderson CancerCenter studied this phenomenon and termed it phe-notypic drift. In either scheme individual cancer cellsare thought to be changing constantly from the

166 AU T O C R I N E A N D PA R A C R I N E G R O W T H M E C H A N I S M S

9905.ch018 1/22/02 8:46 AM Page 166

PROOF

metastatic to nonmetastatic phenotype and back tothe metastatic phenotype. When the rate of changefavors the appearance of metastatic cells, a tumor pro-gresses to a more malignant phenotype capable ofmetastasizing.

Tumor phenotypic instability is probably due, inpart, to essentially irreversible qualitative changes ingene structure (gene mutations, deletions, transposi-tions, amplifications, etc.) that can result in alteredgene products and to dynamic quantitative changesin gene expression. The quantitative changes in geneexpression result in transient changes in the amountsof various gene products in individual cancer cells.The net result is cellular variability within a primarytumor and eventually transient or, in some cases, evenpermanent acquisition of the metastatic phenotypeby individual malignant cells. It is probably these un-stable, highly malignant cells that ultimately give riseto tumor colonies at other sites. This could explainthe problem in determining the particular molecularcharacteristics of metastatic cells. If metastatic cellspossess unstable properties important in malignancy,these would be expected to be difficult to identify.

In contrast to the various unstable properties ofmetastatic cells, the proliferative properties requiredfor survival and growth of metastatic cells at sec-ondary sites must remain relatively stable if metas-tases grow to detectable sizes. Without stable tumorcell growth, only micrometastases would be present,and these would not be expected to kill the host.Thus an understanding of the process of malignantcell growth and its interference at metastatic sites may be important in the development of newtherapies for restraining the growth of establishedmetastases.

II. CANCER PROGRESSION ANDPARACRINE GROWTH OFMETASTATIC CELLS

As cancer progresses and individual cancer cells even-tually acquire more malignant phenotypes, under theproper circumstances, they can invade and metasta-size to near and distant sites. For many cancers thisprocess is not random and cannot be explained by theanatomic site of the cancer or mechanical properties

of cancer cells. Many cancers metastasize to sites un-expected on the basis of their circulatory or lymphaticconnections or on their ability to mechanically lodgein the first lymph node or capillary bed encounteredby cancer cells released from the primary cancer site.For example, the metastasis of cutaneous malignantmelanoma to the brain but ocular malignantmelanoma to the liver, prostate carcinoma to thebone, or colon carcinoma to the liver are examples ofnonrandom metastatic spread. In addition, during tu-mor progression the evolving malignant cells canchange in their tissue-metastatic properties. This ob-servation has been made for a number of different his-tologic classes of cancer.

At the initial stages of metastasis, many possess atendency to metastasize to particular sites. This so-called organ specificity or organ preference of metas-tasis occurs at early stages of metastatic progression,but at later stages of progression where metastasis iswidespread and many secondary sites are involved,more organs and tissues are colonized by metastaticcells. Strictly speaking, the site specificity of cancermetastasis is usually only a preference for colonizingcertain sites; it is not entirely a site-specific process,except for very few cancers. Nonetheless, it is an im-portant phenomenon that appears to be based on theunique properties of the cancer cells as well as thehost microenvironments. This notion was first ad-vanced over 100 years ago by Stephen Paget, who wastreating breast cancer patients in London. Paget ad-vanced the “seed” and “soil” hypothesis that individ-ual cancer cells or seeds can only grow in suitable soil.In terms of their growth properties, neoplastic cellsthat progress to the metastatic phenotype and colo-nize distant sites should be less dependent on theirusual growth signals and more growth responsive togrowth signals at their new metastatic sites.

The increased responsiveness of some cancer cellsto paracrine growth factors or decreased responsive-ness to paracrine growth inhibitors expressed differ-entially at particular metastatic sites could explainwhy certain cancers show a preference for metastaticgrowth at certain sites. This could also explain thefinding of increased numbers or affinities of particulargrowth factor receptors and enhanced responses tocertain growth factors. Overexpression of specificgrowth factor receptors correlates with progression

AU T O C R I N E A N D PA R A C R I N E G R O W T H M E C H A N I S M S 167

9905.ch018 1/22/02 8:46 AM Page 167

PROOF

and metastasis of certain cancers. For example, theepidermal growth factor (EGF) receptor is often asso-ciated with poor prognosis or enhanced metastasis inbreast, lung, and bladder cancers and melanomas. Insome cases, overexpression of a oncogene, such as c-erbB-2/neu-encoded putative growth factor recep-tor, is associated with poor prognosis of breast andovarian carcinomas. Other examples of alterationsexist in growth factor receptors, but in general, thosegrowth factor receptors that are probably importantin metastasis are usually the ones that change in theirexpression in highly metastatic cancers.

The overexpression of growth factor receptors hasbeen accomplished experimentally by gene transfertechniques, and the biological properties of the recip-ient cells can be tested in suitable animal hosts. Forexample, in my department at the University of TexasM. D. Anderson Cancer Center, Dihua Yu, MienchieHung, and I found that the transfer of a mutated c-erbB-2/neu oncogene into benign cells resulted inconversion to the metastatic phenotype. Althoughthe changes in metastatic properties and enhancedgrowth potential of the growth factor receptor gene-transferred cells can be explained by oncogene trans-fer, this experimental result is not always found.Changes in the metastatic properties of c-erbB-2/neugene-transferred cells occurred concomitant withchanges in several metastasis-associated properties,including increased adhesion to microvessel endothe-lial cells, particularly endothelial cells derived fromthe target organs for metastasis, increased invasive-ness of extracellular matrix and reconstituted base-ment membrane matrix, increased cell motility in re-sponse to organ-derived chemotactic factors thatstimulate directed tumor cell invasion, and increasedresponses to organ-derived paracrine growth factors.In these experiments we found that the oncogene-mediated conversion of benign tumor cells to metasta-tic cells was accompanied by changes in growth fac-tor receptor expression and growth factor responses aswell as by changes in the expression of other geneproducts involved in various steps of the metastaticprocess.

To test the hypothesis that organ growth propertiesare important in organ preference of metastasis, wefound that the organ preference of metastatic cellgrowth, at least at the initial stages of metastatic pro-

gression, was related to the differential responses ofmetastatic cells to paracrine growth factors and in-hibitors secreted by the target organ tissues. There-fore, differentially expressed paracrine growth factorsand inhibitors in different organs and tissues probablydetermine, to some degree, the growth potentials ofcancer cells at metastatic sites. Thus cancers, at leastat their initial stages, should be dependent onparacrine growth factors released from surroundingnormal cells. After progression to the metastatic phe-notype, growth factor responses are often changed,and they should be more compatible with the re-sponses to cytokines expressed at secondary metasta-tic sites.

Tumor models have been used to demonstrate thatthe organ preference of metastasis is related to en-hanced growth responses mediated by cytokines,growth factors, and inhibitors released at secondaryorgan sites. Using lung- and ovary-colonizing murinemelanoma sublines and liver- and lung-metastasizinglarge cell lymphoma cell lines, we demonstrated thattumor cell growth in serum-limited culture mediumwas differentially stimulated by soluble factors re-leased from different organ tissues. In these examples,lung- and liver-metastatic tumor cells were growthstimulated better by factors released from lung andliver tissue, respectively. In contrast, other tissue-conditioned media inhibited or had no effect on tu-mor cell growth.

Tissue-derived growth-promoting substances havebeen identified and partially purified from the culturemedium conditioned by certain organ tissues. For ex-ample, we purified to homogeneity a lung-derivedmetastatic cell growth factor from lung-conditionedmedium, and from its amino acid sequence we wereable to identify the growth factor as a transferrin.Transferrins are iron-transferring ferroproteins thatare required for cell growth. Some years ago PedroCuatracasas found that transferrins are more thaniron transport proteins and that they have mitogenicproperties beyond their nutrient transport. The tissue-derived transferrins that we isolated are proba-bly used as paracrine growth stimulators in several tis-sues. The transferrin isolated from lung tissue-conditioned medium was the first tumor cell growthfactor purified to homogeneity on the basis of its abil-ity to differentially stimulate the growth of highly

168 AU T O C R I N E A N D PA R A C R I N E G R O W T H M E C H A N I S M S

9905.ch018 1/22/02 8:46 AM Page 168

PROOF

metastatic cells. We examined a number of differenttumor metastatic systems and found that several other metastatic cell lines were more responsive totransferrin. To demonstrate that a transferrin-like ac-tivity in tissue-conditioned medium is responsible, inpart, for the stimulation of metastatic cell growth,transferrin-like molecules were removed from tissue-conditioned medium. After the removal of transferrinmolecules, most of the growth properties of the organtissue-conditioned medium were lost.

Other organ compartments are also important sitesof metastatic involvement and apparently have theirown set of important cytokines. For example, bone isan important metastatic site for prostate cancer, andtransferrin was found to be a major growth factor forbone-metastasizing prostatic carcinoma cells. Howare transferrins used as specific organ cytokines if theyare found at several organ sites? The answer may bethat the relative concentrations of the transferrinsand other cytokines in different tissues are different.Using a series of melanoma cell lines of differingmetastatic potentials to sites such as brain, we foundthat brain- and lung-colonizing melanoma lines re-sponded best to the lowest concentrations of trans-ferrins and expressed the highest numbers of transfer-rin receptors. Transferrin receptor numbers werehighest in brain metastatic lines and decreased in thefollowing order: high brain-metastasizing ability >high lung-metastasizing ability > intermediate lung-metastasizing ability > poor metastatic capability.Thus, a hierarchy of transferrin expression exists indifferent organs and may be important in determiningmetastatic cell growth. Cancer cells with greater num-bers of transferrin receptors (or different affinities)may be more successful at growing at sites that expresslow concentrations of transferrin molecules.

Brain is an example of an organ that is not partic-ularly susceptible to metastatic colonization. Brainmetastases are rarely produced by cancers, but in somemalignancies, such as melanoma and breast cancer,brain metastases are quite commonly found. For ma-lignant cells to metastasize to brain, it may be advan-tageous for them to express high numbers of particu-lar growth factor receptors, such as transferrinreceptors, and respond to low concentrations ofgrowth factors, such as the transferrins. Further sup-port for the evolution of enhanced transferrin re-

sponsiveness and the metastasis of various tumor cellscomes from the selection of high transferrin receptor-expressing variants from poorly metastatic cells. Thehigh transferrin receptor-expressing cells displayed in-creased spontaneous metastatic properties and grewfaster compared to low transferrin receptor-expressingcells. Brain-metastasizing cancer cells appear to alsorespond to other paracrine growth factors at secondarysites, and it is likely that paracrine sources of trans-ferrin provide only one of several growth factors im-portant in determining the organ preference ofmetastatic cell growth. Differences in the concentra-tions of various cytokines, growth factors, and in-hibitors may be important in providing the correctenvironment for metastatic cell growth.

The normal functions of paracrine growth factorsare not known, but they might be involved in con-trolling cell growth and local tissue regeneration dur-ing wounding and inflammation. Metastases often oc-cur at the sites of trauma or tissue damage. Thus thenormal role of the paracrine growth factors and growthinhibitors is probably in organ repair.

Another source of organ-derived growth-promoting molecules are extracellular matrix, tissuestroma, and basement membranes. Extracellular ma-trix and basement membranes contain tightly boundgrowth factors that can be released by tumor celldegradative enzymes and stimulate tumor cell growth.Extracellular matrix molecules themselves may mod-ulate tumor cell growth and the state of tumor celldifferentiation. For example, the maintenance of nor-mal breast cells is dependent on lactogenic hormonesand extracellular matrix, and matrix molecules canregulate gene expression and growth of particular nor-mal cells. Metastatic cells often show differentialgrowth responses to extracellular matrix molecules.Using extracellular matrix obtained from several or-gans, Lola Reid and collaborators at the Albert Ein-stein Medical Center found that metastatic mammarycarcinoma and hepatoma cells were differentiallystimulated to grow at low cell densities by organ ma-trix isolated from the target organs for metastasis for-mation. However, they did not find the same patternof growth stimulation if metastatic cells were platedat high cell densities on the various extracellular ma-trices, suggesting that extracellular matrix growthstimulatory molecules are more important at the early

AU T O C R I N E A N D PA R A C R I N E G R O W T H M E C H A N I S M S 169

9905.ch018 1/22/02 8:46 AM Page 169

PROOF

stages of cancer cell growth, such as in micrometas-tases, rather than at the later stages of cell growth athigh cell densities, such as would be expected in grossclinically detectable metastases. Thus the uniquegrowth microenvironments for cancer cells in variousorgans and tissues are probably determined collec-tively by tumor cell responses to cell-bound, matrix-bound, and soluble paracrine factors.

III. PARACRINE GROWTHINHIBITORS OF METASTATIC CELLS

Cancer cells also receive and process paracrine nega-tive growth signals. Only a few organ-derivedparacrine growth inhibitory molecules have beenidentified. In most cases these have turned out to bewell-known cytokines, such as the transforminggrowth factor-� (TGF-�) family. Certain organ cellscan release potent growth inhibitors that prevent thegrowth of malignant cells, and these factors could beimportant in determining metastatic cell growth atparticular sites. For example, kidney cell-conditionedmedium is particularly inhibitory for many cancercells, and most cancers and metastatic model systemsfail to metastasize to the kidney. The most potentgrowth inhibitor released by kidney tissue has beenidentified as TGF-�1. TGF-�1 can inhibit the growthof several highly metastatic cell lines. Not all metasta-tic cells are growth inhibited by TGF-�, but this fam-ily of cytokines is very important in cancer metasta-sis. An interesting finding is the growth stimulationof metastatic cells by TGF-� and the inhibition ofgrowth of tumor cells from primary sites, but this doesnot appear to be a general phenomenon.

The growth responses of malignant cells can changeduring progression to more malignant phenotypes.For example, the growth responses of early lateralgrowth phase human melanoma cells and more ad-vanced vertical growth phase melanoma cells havebeen studied by Robert Kerbel and Meenhard Herlyn.Only the vertical growth phase melanomas are dan-gerous, and patients with these lesions are at risk todevelop metastases. Studies on the responses ofmelanoma cells to positive and negative growth cy-tokines indicated that the more progressed melanomacells lose responsiveness to negative growth inhibitors.

The molecule responsible for differentially inhibitingthe growth of early lateral growth phase melanomacells was purified and shown to be interleukin-6 (IL-6), a well-known hemopoietic cytokine. This cytokine is produced by a variety of tissues, amongthem keratinocytes, endothelial cells, fibroblasts,macrophages, and monocytes. The inhibitory re-sponses to recombinant IL-6 were not duplicated withmore advanced vertical growth phase melanoma cells.It was subsequently established that the more highlyprogressed metastatic cells uniformly lost responsive-ness to a variety of growth inhibitors. Loss of paracrineinhibitor responses could be as important to the for-mation of metastasis as changes in paracrine growthfactor responsiveness.

IV. SOURCES OF PARACRINEGROWTH FACTORS AND INHIBITORS

The sources of paracrine growth factors and in-hibitors in various organ tissues are largely undeter-mined. Parenchymal cells, fibroblasts, endothelialcells, mast cells, and macrophages, among other celltypes, and factors from acellular sources, includinginterstitial extracellular matrix and basement mem-branes, could collectively provide various growthfactors and inhibitors. Some of the growth factorsexpressed by parenchymal cells are released as solu-ble molecules, whereas some are not released and re-quire cell–cell or cell–matrix contact. When lung-and liver-colonizing malignant melanoma cells werecocultured with normal hepatocytes, Max Burgerand colleagues found that only liver-colonizingmelanoma cells were growth stimulated. The stimu-lation required cell-to-cell contact with the hepato-cytes and was not duplicated with liver tissue-con-ditioned medium. Thus in some metastatic systems,organ parenchymal cells are important sources of tu-mor cell growth stimulation.

Microvascular endothelial cells are important in metastatic cell growth. Using lung- and liver-colonizing large cell lymphoma cells, we showed thatconditioned medium from organ-derived microvesselendothelial cells could substitute for organ tissue-conditioned medium in tumor cell proliferation as-

170 AU T O C R I N E A N D PA R A C R I N E G R O W T H M E C H A N I S M S

9905.ch018 1/22/02 8:46 AM Page 170

PROOF

says. Interestingly, liver-colonizing large cell lym-phoma cells responded best to conditioned mediumfrom liver sinusoidal endothelial cells, whereas lung-colonizing lymphoma cells responded best to condi-tioned medium from lung microvessel endothelialcells. Removal of transferrin from the lung endothe-lial cell-conditioned medium resulted in a reductionof mitogenic activity, but some activity remained thatwas not associated with transferrin. Endothelial cellscan respond, in turn, to growth and motility factors,called angiogenesis factors, released by malignant cells.A reciprocal relationship may exist between tumorcells and specific organ-derived normal cells (Fig. 1).This relationship extends to other cell types as well asto extracellular matrix. Thus malignant cells can stim-ulate as well as be stimulated by normal host cells.

Fibroblasts isolated from different tissues have beenused to differentially stimulate the growth of cancercells. Using lateral and vertical growth phasemelanoma cells, the growth responses of these cells tofibroblasts from various tissue sources have been testedin cocultures. Fibroblasts isolated from dermal tissuegenerally inhibited the growth of early lateral growthphase human melanoma cells but had stimulatory orlittle effect on more advanced vertical growth phasemelanoma cells. In this case, the inhibitory molecule

responsible was identified as IL-6 released by dermalfibroblasts. We also found that metastatic and non-metastatic mammary carcinoma cells responded dif-ferentially to tissue-derived fibroblast-conditionedmedium, and this was related to the organ preferenceof metastasis. The highest growth stimulation wasfrom fibroblasts isolated from lung and mammarygland, targets for metastatic cell growth in this sys-tem. The fibroblast-derived growth factor was not re-lated to transferrin because we demonstrated that theorgan-derived fibroblasts did not synthesize detectableamounts of transferrin. Peter Jones has examined thegrowth properties of bladder carcinomas of differentgrade and invasive properties. Using bladder carcino-mas of differing grade, differentiation, and invasiveproperties, only the poorly differentiated high-grade,invasive bladder carcinoma cells were growth stimu-lated by bladder-derived fibroblasts. When prostatecarcinoma cells were cocultured with fibroblasts fromdifferent tissue sources, an interesting relationshipwas found. Similar to the reciprocal relationship be-tween cancer cells and endothelial cells, bidirectionalstimulation of growth was seen by Lelung Chung whenconditioned medium from prostate carcinoma cellswas tested with conditioned medium from bone fi-broblasts. Because bone is a common site of prostatemetastases, this suggested that prostate carcinomacells stimulate and are stimulated by target bone fi-broblasts. This reciprocal or bilateral relationship be-tween metastatic cells and host cells in the target sitefor metastasis appears to be an important feature ofmetastatic cell colonization (Fig. 2).

Mast cells are another source of mitogens and motil-ity factors for metastatic cells. Similar to the processof inflammation, mast cells are attracted to tumorsites by the release of mast cell mediators. Mast cellsoften associate at the periphery of tumors and can re-lease tumor cell mitogens and motility factors thatcan differentially affect the growth and motility ofmalignant cells. With Mustafa Dabbous we found thatonly highly metastatic mammary cells attracted largenumbers of mast cells into the tumor periphery. Mastcells isolated from the tumor periphery were found torelease factors that differentially stimulated thegrowth of the highly metastatic but not poorlymetastatic cells. A function for the mast cell mito-gens was shown in vivo by administering drugs that are

AU T O C R I N E A N D PA R A C R I N E G R O W T H M E C H A N I S M S 171

FIGURE 1 Reciprocal interactions between malignant cells andtheir cellular microenvironments. Cancer cells release paracrine fac-tors that can affect host cells, such as parenchymal cells, endothelialcells, fibroblasts, mast cells, granulocytes, macrophages, and tissue ex-tracellular matrix. In turn, the host cells and matrix can releaseparacrine factors that stimulate or inhibit tumor cell proliferation. Tumor cells can also synthesize autocrine factors that can act in-side the cell (private or intracellular) or outside the cell (public or extracellular).

9905.ch018 1/22/02 8:46 AM Page 171

PROOF

mast cell stabilizers to animals receiving metastaticcell implants. Mast cell stabilizers prevented the re-lease of mast cell contents and reduced the growthand metastases of the highly metastatic cells growingat their normal organ sites.

V. AUTOCRINE GROWTHMECHANISMS IN CANCER METASTASIS

Cancer cells have the capacity to synthesize and re-lease multiple growth factors that can act on the tu-mor cells themselves in tumor cell autocrine loopmechanisms. This can occur either by extracellularrelease of the growth factor and its binding to an ap-propriate extracellular receptor on the same cell (pub-lic or extracellular autocrine mechanism) or by ligand–receptor interaction inside the cell (private orintracellular autocrine mechanism, Fig. 1). Whenhighly malignant and metastatic cells are examinedfor the synthesis of autocrine growth factors, they are

commonly found to make and use a variety of growthfactors. Interestingly, these are often the same growthfactors that normally stimulate the growth of normalcells from which the tumor cells were derived.

Autocrine growth factors play an important role inneoplastic transformation, but their role in cancermetastasis is somewhat less clear. Studies on tumorcells derived from primary and metastatic sites or se-quential selection of malignant or metastatic variantshave yielded important but inconclusive data on therole of autocrine growth factors in metastasis. As tu-mors progress to more malignant or metastatic phe-notypes, they generally become less dependent onserum-derived growth factors for their growth andthey begin to produce polypeptide growth factors,suggesting that autocrine growth mechanisms may beinvolved in metastasis formation. There are, however,examples where the autocrine production of a growthfactor by malignant cells did not correlate with tumorprogression or stage. For example, the production andsecretion of melanoma growth-stimulating activity bymelanoma cells and bombesin-like (gastrin-releasing

172 AU T O C R I N E A N D PA R A C R I N E G R O W T H M E C H A N I S M S

FIGURE 2 Tumor microenvironment and the role of paracrine cytokines in regulating tumor cell growth, differentiation, and malignant prop-erties. Reciprocal or bidirectional interactions between various cells in the tumor microenvironment are important in the establishment of thetumor microenvironment.

9905.ch018 1/22/02 8:46 AM Page 172

PROOF

peptide) activity by small cell lung carcinoma cellswere not related to tumor stage or tumor progression.The usual result is a loss of growth factor responsive-ness with tumor progression to the metastatic pheno-type. For example, Meenhard Herlyn and collabora-tors found that cell lines established from dysplasticnevi had similar growth factor requirements to nor-mal melanocytes, whereas malignant melanoma celllines had reduced requirements for a variety of growthfactors. Only cell lines established from melanomametastases could be quickly adapted to in vitro growthin serum-free medium. When they sequentially se-lected human melanoma cell lines that were estab-lished from a primary melanoma lesion for their abil-ity to invade a reconstituted basement membrane,they showed that only the most invasive tumor cellvariants were spontaneously metastatic in nude miceand that these same cell lines were less serum depen-dent for growth in tissue culture. Selection for growthin serum-free medium resulted in increases in inva-sive and metastatic properties, but the selectedmelanoma cell lines were unstable and in the absenceof continued selective pressure reverted back to thephenotype of the parental cell line. The relative in-stability of metastasis-associated properties was men-tioned earlier, and transient changes in gene expres-sion are expected when a strong selective pressure isremoved from a population of inherently unstablemalignant cells.

The release of autocrine cytokines can also produceeffects on neighboring cancer cells. The release of acytokine from one tumor cell and its effect on neigh-boring cells is a form of paracrine signaling among tu-mor cell populations that may be important in modi-fying the properties of the tumor. In a few cases, theclonal effects of tumor cells on surrounding tumorcells have been examined and found to affect tumorcell properties, especially malignant cell properties.When we examined the interclonal interactions of aseries of melanoma cell lines, we found that the ex-pression and the display of a cell surface glycoproteinthat later was identified as a growth factor receptorwere affected by clonal cell interactions along withmetastatic properties. Here the interclonal tumor cellinteractions stabilized the phenotypic properties ofthe malignant cells, but when the malignant cellswere grown separately, they quickly lost their metasta-tic and organ growth properties. Thus tumor cell–cell

and host cell–tumor cell interactions are important inthe metastasis and growth of malignant cells.

In some tumor systems, both paracrine and au-tocrine signals are important in metastasis. For exam-ple, hepatocyte growth factor, which is also known asscatter factor (HGF/SF), plays autocrine and paracineroles in both cell motility and growth. The HGF/SFreceptor at the cell surface is encoded by the onco-gene met or its normal cell counterpart, c-met.HGF/SF has been shown to be synthesized by a num-ber of mesenchymal cell types, and it can act as aparacrine stimulator of metastasis at specific organsites, such as liver and lung. Certain melanoma cellsselected for liver-specific metastasis overexpress theHGF/SF receptor, and exposure of these cells to HGFaugments cell motility and invasive behavior. Upreg-ulation of c-met in these cells and its HGF/SF recep-tor by various methods increases liver colonizationability but does not change the organ metastatic speci-ficities of these cells. Treatment of certain mammaryadenocarcinoma cells with HGF stimulates their abil-ity to metastasize to lung.

VI. GENE TRANSFER ANDMETASTATIC CELL GROWTH PROPERTIES

Transfer of a growth factor gene or a growth factor re-ceptor gene into suitable untransformed recipient cellscan result in these cells acquiring malignant growthcharacteristics. For example, we found that transfec-tion of the transferrin receptor gene resulted in an in-creased ability to grow in serum-free medium withtransferrin as the sole supplement along with acquisi-tion of the metastatic phenotype. In other experi-ments, transfer of the oncogene c-sis that encodes theplatelet-derived growth factor (PDGF) receptor, c-erbB-1 that encodes the EGF receptor, or c-erbB-2/neuthat encodes a growth factor-related receptor by useof viruses increases the growth potential of the gene-transferred cells, resulting in oncogenic transforma-tion. In some cases, however, the transfer of a growthfactor gene does not result in neoplastic transforma-tion unless the encoded growth factor is synthesizedin a form that can be cell secreted and bind to a cellsurface receptor. For example, Michael Klagsbrunfound that the addition of basic fibroblast growth

AU T O C R I N E A N D PA R A C R I N E G R O W T H M E C H A N I S M S 173

9905.ch018 1/22/02 8:46 AM Page 173

PROOF

factor (bFGF) by itself or transfer of the bFGF genealone was not transforming to recipient cells. How-ever, when the bFGF gene was fused to a secretionsignal sequence to facilitate bFGF secretion, only thechimeric signal peptide–bFGF gene was transforming.In this example, the appropriate location of thegrowth factor was important to its ability to signalgrowth. In addition to tumor cell growth, FGFs arealso important in endothelial cell proliferation andangiogenesis.

Using gene transfer techniques to stimulate cells tosecrete growth factors and become transformed in theprocess assumes that the technique itself does notcause other changes in the recipient cell. This turnsout to be important because highly malignant cellsare also unstable, and the gene transfer techniquescan destabilize cells in a process that leads to cellularinstability, diversification, and cellular heterogeneity.Controversy exists as to whether additional genomicchanges are required for the transformation of cellsthat have received growth factor genes. An exampleof this is the use of transforming sequence containinga gene for one of the PDGF polypeptide chains. ThePDGF molecule is made up of two polypeptide (Aand B) chains. Only the B chain gene was found tohave a transforming sequence in at least two retro-viruses, a class of transforming virus that has a propen-sity to pick up critical host gene sequences that cantransform normal cells to neoplastic cells. The PDGFreceptor homologue that constitutes the c-sis retrovi-ral gene is oncogenic, suggesting the importance ofgrowth factor genes and their receptors in neoplastictransformation. Although overexpression of the c-sisoncogene can result in neoplastic transformation, theaddition of excess PDGF to untransformed cells doesnot reproduce this phenomenon. This apparent para-dox can be explained by considering that the in-creased rate of cell proliferation caused by PDGF prob-ably results in expansion of a subpopulation ofpreneoplastic cells that are potential targets for onco-genic transformation. Alternatively, the insertion ofthe growth factor gene itself causes mutation of cel-lular genes or other changes that are necessary forneoplastic transformation.

Gene transfer techniques have been used to test forthe involvement of autocrine growth mechanisms inmetastasis formation. Transfection of human breast

cancer cells with the HGF/SF gene establishes an au-tocrine loop, and these cells demonstrate a greaterpropensity to metastasize to the lung. Cotransfectionof murine 3T3 cells with genes encoding HGF/SF andits receptor results in cells that acquire the metastaticphenotype, especially lung-metastatic ability. ArnoldGreenberg and collaborators used a transformingchimeric bFGF gene construct to transfect untrans-formed but highly unstable preneoplastic ells. Thesignal peptide–bFGF chimeric gene-transfected cellsformed experimental metastases, whereas the bFGFgene-transfected cells without the signal peptide didnot. In addition to experiments where untransformedcells were converted to malignancy by transfer of agene encoding a growth factor that the recipient cellscan respond to, cells have been transfected withgrowth factor genes encoding factors that they nor-mally cannot respond to. For example, untransformedNIH-3T3 cells are not normally responsive to colony-stimulating factor-1 (CSF-1) because they lack theCSF-1 receptor. Using v-fms-transfected cells (v-fmsis an oncogene encoding a CSF-1-like receptor),Greenberg and collaborators found that an additionof exogenous CSF-1 stimulated cell growth. In vivothe v-fms-transformed cells formed experimentalmetastases, whereas the untransformed cells did not.The addition of exogenous CSF-1 to v-fms-transformed cells before injection into animals re-sulted in greater numbers of experimental metastasesin mice. However, if v-fms-transformed cells were al-lowed to grow under culture conditions where au-tocrine growth factor-conditioning could occur, andthen the cells were treated with CSF-1, the oppositeeffect was obtained and the v-fms-transformed cellsformed fewer metastases. This inhibitory effect wasprobably due to downregulation of the CSF-1 recep-tors on the v-fms-transfected cells by excess exoge-nous CSF-1. The differing effects of growth factors onthe transfected NIH-3T3 cells were explained as be-ing due to receptor occupancy and growth factor sat-uration effects. Although the levels of secreted au-tocrine growth factors, their receptors, and receptoroccupancy were not determined, the saturation ofgrowth factor receptors by exogenous growth factorscould cause effects other than growth stimulation,such as growth inhibition or differentiation. Differentconcentrations of certain growth factors can have op-

174 AU T O C R I N E A N D PA R A C R I N E G R O W T H M E C H A N I S M S

9905.ch018 1/22/02 8:46 AM Page 174

PROOF

posite effects on the same cell system. Often low con-centrations of cytokines and growth factors can bestimulatory, whereas high concentrations of the samefactor can be inhibitory. In addition, as discussed ear-lier, the transfer of genes into unstable cells can resultin cellular diversification and heterogeneity.

Another example of loss of responsiveness with tu-mor progression is the loss of hormone responses withprogression of breast cancer. Loss in estrogen depen-dency of MCF-7 breast cancer cells after transfectionwith the v-H-ras oncogene has been seen by MarcLippman and collaborators. In contrast to parentalMCF-7 cells, ras-transfected MCF-7 cells were tu-morigenic in the absence of infused estrogens, andthe transfected cells secreted TGF-�, TGF-�, and insulin-like growth factor-I, without a change in theircell surface growth factor receptor densities. Becauseadvanced breast cancers often become refractory toestrogens and other hormones, this suggests that asmalignant cells become more advanced, they becomeless dependent on systemic hormones and they possi-bly secrete higher amounts of autocrine growth fac-tors. In support of this notion is the loss of hormoneresponsiveness of breast cancers as they progress invivo. Breast cancers that are initially responsive to17�-estradiol lose their hormone responsiveness as they progress to more malignant and aggressive,hormone-independent phenotypes. They can also in-crease their synthesis and secretion of autocrinegrowth factors. These finding are consistent with ageneralized loss of growth factor regulation and an in-crease in autocrine growth mechanisms with progres-sion to more malignant cellular phenotypes.

VII. NEW APPROACHES TOCANCER THERAPY

The information on growth requirements, expressionof growth factor receptors, and growth inhibitor re-ceptors by malignant cells should be useful in devel-oping new therapeutic approaches to limiting thegrowth of metastatic cancers. Specific approaches usedto limit the proliferation of tumor cells include theadministration of growth inhibitors or analogs ofgrowth factors or the use of antibodies against growthfactor receptors to deliver toxins to highly metastatic

cells. For example, John Mulshine and collaboratorsfound that a bombesin/gastrin-releasing peptide is acommonly found autocrine growth factor made bysmall cell lung cancer cells, and bombesin analogshave been used to inhibit the growth of the lung car-cinoma cells in vitro in clonogenic growth assays andin vivo in xenographs in nude mice. Alternatively, toinhibit autocrine/paracrine growth pathways, anti-bodies have been administered that bind to the growthfactors and remove or prevent them from interactingwith tumor cells. For example, bombesin/gastrin-releasing peptides can be eliminated by administeringmonoclonal antibodies against these factors. A phaseI clinical trial using a monoclonal antibody against abombesin/gastrin-releasing peptide indicated that theeffects of the bombesin/gastrin-releasing peptide canbe partially blocked without apparent toxicity.

The employment of antibodies against growth fac-tor receptors has been effective in preventing tumorgrowth. John Mendelson and collaborators and RalphReisfeld independently were among the first to usemonoclonal antibodies against the EGF receptor toinhibit the growth of human tumors in nude mice.One of the monoclonal antibodies that Mendelsonand colleagues used was also found to activatemacrophage- and complement-dependent lysis of can-cer cells in vitro, suggesting that the effects of growthfactor receptor antibodies on the growth of tumorcells in vivo could be due, in part, to blocking host re-sponses against the tumor. The formation of metas-tases from implants of human melanoma cells in se-verely immunodeficient mice has been blocked witha monoclonal antibody against the EGF receptor.When the antibody Fc or tail portion was removed,the tailless monoclonal antibody did not inhibitmetastasis, suggesting that the Fc or tail portion ofthe antibody contributed to the anti-metastatic ef-fect, possibly by an antibody-dependent host cell ef-fector mechanism.

Various investigators have coupled the antibodiesto toxins or toxin subunits to increase the effective-ness of antigrowth factor receptor monoclonal anti-bodies in suppressing the growth and metastasis of hu-man tumors. Using extremely toxic moleculescovalently bound to the targeting monoclonal anti-body, specific killing of malignant cells has beenachieved. The chimeric toxin–antibody molecules

AU T O C R I N E A N D PA R A C R I N E G R O W T H M E C H A N I S M S 175

9905.ch018 1/22/02 8:46 AM Page 175

PROOF

can produce antitumor effects without apparent sideeffects and suppression of white blood cells, eventhough these normal cells also express the growth fac-tor receptors. Although it remains to be demonstratedthat every last malignant cell can be killed with suchtoxin–antibody conjugates, the use of potent toxinconjugates to specifically kill metastatic cells at sec-ondary sites may be an achievable goal that is wellwithin our technical ability.

VIII. GROWTH RESPONSES OFMETASTATIC CELLS ANDMALIGNANT PROGRESSION

During the progression of malignant tumors there isoften a tendency for the most malignant cells in a tu-mor cell population to lose expression of growth fac-tor or inhibitor receptors and lose responsiveness toparticular growth factors or inhibitors. The loss ofthese responses in highly metastatic cancer cells andthe ability of such cancers to colonize and grow at dis-tinct secondary sites may be explainable by consider-ing each stage of cancer progression. At the earlystages of metastasis, many cancers show restricted or-

gan preference of metastatis, whereas at the finalstages near host death, these same cancers often col-onize multiple organ and tissues sites. The explana-tion for this is that cancers progress from mainlyparacrine growth stimulatory and inhibitory mecha-nisms at the initial stages of metastatic progression tomainly autocrine stimulatory mechanisms at the finalterminal stages (Fig. 3). As discussed earlier, highlyadvanced cancers can secrete a variety of growth fac-tors that could serve as autocrine sources of growthstimulation independent of their microenvironments.This could explain the loss of organ preference ofmetastasis seen at the later stages of cancer progres-sion, whereas at earlier stages of cancer progressionfewer organs are usually involved. Eventually metasta-tic progression continues and alterations in tumorcells may produce what has been termed an acrine(lack of regulation) state where malignant cells havelost their usual growth factor and inhibitor responsesand are refractory to growth regulation by cytokinesand growth factors and inhibitors. Because acrine ma-lignant cells are not expected to respond to endoge-nous or exogenous growth factors or inhibitors, theyshould be the ultimate autonomous cells.

The concept that growth stimulatory and inhibitory

176 AU T O C R I N E A N D PA R A C R I N E G R O W T H M E C H A N I S M S

FIGURE 3 Progression of malignant cancer cells results in alterations in their responsiveness to host-derived growth stimulatory and inhibitory molecules and synthesis of autocrine growth factors. Tumor celldifferentiation is shown as a possible process to the left and tumor cell progression to the right. Tumor pro-gression can result initially in an increase in paracrine growth factor responses, as in the case of organ pref-erence of metastatic colonization. As tumor progression continues, however, loss of paracrine growth factorand growth inhibitor responsiveness occurs and increases in autocrine growth factor production occur withmore widespread tissue metastatic colonization. Eventually, tumor progression can result in a complete loss ofgrowth factor/inhibitor responses (acrine state) and metastatic colonization of virtually any tissue can occur.

9905.ch018 1/22/02 8:46 AM Page 176

PROOF

responses are altered during cancer progression hasimportant significance for the development of newtherapies for metastatic cancers. If highly progressedmetastatic cells are more refractory to growth regula-tion, then it is unlikely that therapeutic interventionusing analogs of growth inhibitors, monoclonal anti-bodies against growth factors or their receptors, orother means of suppressing growth-stimulating mole-cules or enhancing growth-inhibitory molecules willbe useful in highly advanced cancers. Because thesehighly progressed cancers are often unstable and pro-duce cell progeny that are unstable and undergo rapidchanges in their gene expression programs, it is un-likely that therapies based on growth properties ofmalignant cells will succeed at the terminal stages ofcancer progression. Their use must be at the earlieststages of cancer progression and metastasis, wheremalignant cells are still responsive to growth signals.

See Also the Following Articles

HEMATOPOIETIC GROWTH FACTORS • MOLECULAR MECHA-NISMS OF CANCER INVASION • TUMOR CELL MOTILITY AND

INVASION

Bibliography

Aaronson, S. A. (1991). Growth factors and cancer. Science254, 1146–1153.

Cavanaugh, P. G., Jia, L., Zou, Y. Y., and Nicolson, G. L.

(1999). Transferrin receptor overexpression enhances trans-ferrin responsiveness and the metastatic growth of a ratmammary adenocarcinoma line. Breast Cancer Res. Treat.56, 203–217.

Nicolson, G. L. (1987). Tumor cell instability, diversification,and progression to the metastastic phenotype: From onco-gene to oncofetal expression. Cancer Res. 47, 1473–1487.

Nicolson, G. L. (1991). Gene expression and tumor progres-sion to the metastatic phenotype. Bioessays 13, 337–342.

Nicolson, G. L. (1988). Cancer metastasis: Tumor cell andhost organ properties important in colonization of specificsecondary sites. Biochim. Biophys. Acta 948, 175–224.

Nicolson, G. L. (1991). Molecular mechanisms of cancermetastasis: Tumor and host properties and the role of onco-genes and suppressor genes. Curr. Opin. Oncol. 3, 75–92.

Nicolson, G. L. (1993). Cancer progression and growth: Rela-tionship of paracrine and autocrine growth mechanisms toorgan preference of metastasis. Exp. Cell Res. 204, 171–180.

Nicolson, G. L. (1993). Paracrine and autocrine growth mech-anisms in tumor metastasis to specific sites with particularemphasis on brain and lung metastasis. Cancer Metast. Rev.12, 325–343.

Nicolson, G. L., Menter, D., Herrmann, J., Cavanaugh, P., Jia,L.-B., Hamada, J., Yun, Z., and Marchetti, D. (1994). Tu-mor metastasis to brain: Role of endothelial cells, neu-trophins and paracrine growth factors. Crit. Rev. Oncogen,5, 451–471.

Rodeck, U., and Herlyn, M. (1991). Growth factors inmelanoma. Cancer Metast. Rev. 10, 89–101.

Vande Woude, G. F., Jeffers, M., Cortner, J., Alvord, G., Tsar-faty, I., and Resau, J. (1997). Met-HGF/SF: Tumorigenesis,invasion and metastasis. Ciba Found. Symp. 212, 119–130.

Volpe, J. P. (1988). Genetic instability of cancer: Why ametastatic tumor is unstable and a benign tumor is stable.Cancer Genet. Cytogenet. 34, 125–134.

AU T O C R I N E A N D PA R A C R I N E G R O W T H M E C H A N I S M S 177

9905.ch018 1/22/02 8:46 AM Page 177