Tumor subpopulation interactions in neoplasms

Tumor subpopulation interactions in neoplasms

Biochimica et Biophysica Acta, 695 (1983) 2 1 5 - 2 2 6 215 Elsevier BBA 87110 T U M O R SUBPOPULATION INTERACTIONS IN NEOPLASMS G L O R I A H. H ...

1MB Sizes 0 Downloads 17 Views

Biochimica et Biophysica Acta, 695 (1983) 2 1 5 - 2 2 6

215

Elsevier

BBA 87110

T U M O R SUBPOPULATION INTERACTIONS IN NEOPLASMS G L O R I A H. H E P P N E R , B O N N I E E. M I L L E R a n d F R E D R. M I L L E R

Department of Immunology, Michigan Cancer Foundation, Detroit, MI 48201 (U.S.A.) (Received N o v e m b e r 3rd, 1982)

Contents I.

Introduction

.............................................................................

II.

T u m o r s u b p o p u l a t i o n i n t e r a c t i o n s in n e o p l a s m s

....................................................

215 217

III. Cellular interactions in n o r m a l tissues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

221

IV. C o n c l u s i o n s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

223

Acknowledgements ............................................................................

224

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

224

I. Introduction

Multiple subpopulations of tumor cells have been demonstrated within, and in some cases isolated from, single neoplasms of both experimental and human origin [1]. This tumor cell heterogeneity has been reported in cancers of all histological types - carcinoma [2-13], sarcoma [14,15], melanoma [16-18], glioma [19,20], lymphoma [21,22] - in premalignant, as well as malignant, neoplasms [23-25] and in tumors arising in a variety of organ systems. Both autochthonous and transplanted tumors have been shown to be heterogeneous. Tumor cell heterogeneity is manifested by differences in numerous phenotypic characteristics: cellular morphology [4] or tumor histopathology [3], expression of cell markers or production of differentiated cell products [4,5,7-11,13,16,18], growth properties in vitro, such as cloning efficiency, doubling time and saturation density [4,9,10], and tumorigenic properties in vivo, in0 3 0 4 - 4 1 9 X / 8 3 / $ 0 3 . 0 0 © 1983 Elsevier Science Publishers B.V.

cluding the number of cells required to produce tumors, latency period and growth rate [4,16,22]. The ability to induce host-immune responses is distributed heterogeneously among tumor subpopulations, as is the independent ability to be affected by such defense responses [21,26-31]. Intratumor heterogeneity in ability to invade and metastasize has been demonstrated [14,15,17, 32,33], as well as heterogeneity in other clinically important characteristics, such as sensitivity to chemotherapy [6,34-39], radiation [40-42] and immunotherapy [21,29]. Tumor cell heterogeneity provides a mechanistic basis for the concept of neoplastic progression, that is, the acquisition by tumors of new characteristics during neoplastic development. Changes in cancer behavior can be ascribed to progressive selections of different clones resulting in shifts in the predominant tumor cell subpopulations. According to this hypothesis, variant subpopulations arise from the original focus of initiation, so that even tumors of monoclonal origin become

216

heterogeneous. Nowell [43], and more recently Fidler and Hart [44], have discussed evidence for this process and have presented data indicating a relative genetic instability of neoplastic cells which would provide the basis for variant production. Other authors have discussed various molecular mechanisms which would effect genomic alterations [45]. Variant subpopulations could also be generated by mechanisms that do not necessarily alter the chemical structure of DNA, such as mechanisms involved in tissue differentiation [46] or somatic cell fusion [47]. Regardless of the mechanism of generation of neoplastic diversity, the end result is a mixture of tumor cell clones co-existing. Repeated demonstrations of clonal heterogeneity in more and more cancers have swollen the recent cancer biology literature and, indeed, tumor heterogeneity has become the explanation for numerous phenomena, including progression and incomplete responses to treatment, as well as the excuse for the problems of variability and lack of reproducibility which plague cancer research. It is not our purpose to review here the evidence relevant to cancer heterogeneity per se. Numerous recent reviews are available which detail the basis for intratumor diversity [1,44,48,49]. Clearly, not all of the variability among tumor cells of a single tumor is due to clonal diversity [50]. Recent reviews also discuss the origin of tumor subpopulations and the clinical implications of their existence [1,44,51,52]. The reader is referred to the proceedings of the Bristol-Myers Cancer Symposium, Tumor Cell Heterogeneity Origins and Implications, for a broad-based presentation of relevant observations and suggested mechanisms [53]. In this article we wish to turn attention to the mechanisms by which this variability within tumors is moderated by interactions between the varying cell subpopulations. Although many types of interaction surely exist among cells within tumors, and many of these interactions do not depend upon tumor clonal heterogeneity, we will focus on those interactions which occur between different subpopulations. Cancers are now often viewed much as bags of different colored marbles - each color indicating a different clone. Cancer behavior is seen as deriving from the number of marbles and the number and

distribution of the various colors. Cancers, however, are not composed of inert 'marbles', but rather of living cells that are capable of interacting with other living cells and cellular products. Tumors are cellular societies - ecosystems - in which the individual clones influence the overall behavior, and are themselves influenced by each other and by normal cells. Tumor cell societies resemble societies of normal cells, a point which we emphasize at the end of this review. Evidence for cell-cell interactions within neoplasms can be found even in the earliest events of tumor development. Rubin has commented recently on the dependence of in vitro transformation frequency on cell density at the time of exposure to carcinogen [54]. He pointed out that cells exposed to carcinogen undergo a variety of changes other than those associated with neoplastic transformation. Risser and Pollack [55] showed, under circumstances in which only 10% of 3T3 cells were transformed by SV40 virus, that a high frequency of clones of non-transformed cells differed from parental cells in characteristics such as saturation density, serum growth requirements, cloning efficiency, etc. This suggests that neoplastic change is accompanied by induction of extensive clonal heterogeneity within the 'field' of initiation. The number of cell types present from the beginning includes more than 'normal' and 'transformed' cells. Interactions between altered and initiated cells, as well as between normal and initiated cells, may result in a lowering of the apparent transformation frequency in high density cultures. Whether such interactions affect initiation per se, or whether they act on the growth of transformed cells, is not clear. Normal cells have been reported to inhibit and to stimulate growth of transformed or premalignant cells [56-59]. Tumor cells may either inhibit or stimulate normal cell division as well [60]. Thus, growth regulatory interactions between tumor and normal cells may be a feature of neoplastic growth and development; the effect of the interactions is not necessarily one-sided, nor always in the same direction. The above conclusion is reinforced by consideration of the well-documented interactions of endocrinological or immunological cells with neoplastic cells of many types. These interactions may be quite complex. For example, Nandi [61] has pro-

217 posed that hormones may enhance carcinogenesis by stimulating division of normal cells, thereby shortening their life span; thus, the inhibitory influence of normal cells on the growth of neoplastic cells would be decreased. Prehn [62] has proposed the theory of immunostimulation in which lymphoid cells may either enhance or retard tumor growth, depending upon the stage of tumor development. We [63] have suggested that the conversion of preneoplastic nodules to cancers may be accompanied by a shift in initiated cells from being inhibited by normal cells to being stimulated by them. Thus, the processes of initiation, establishment of preneoplastic lesions and progression to cancer all take place in an environment of multiple, and interacting, cell types.

II. Tumor subpopulation interactions From the above perspective, the idea that cells of different neoplastic subpopulations within the same tumor might interact is a logical extension. The existence of such interactions is suggested by the observation that one can isolate clones which differ markedly among themselves and with their parent tumor in growth properties and malignant potential. If isolation procedures per se do not change tumor behavior [1,17], why are the characteristics of a cancer not those of its most aggressive subpopulations? Despite the phenomenon of neoplastic progression, the behavior of a cancer is not necessarily determined by its 'most malignant' component. Even after prolonged serial transplantation in vivo or cell culture in vitro, one can isolate subpopulations of cells that are either more or less tumorigenic than the parental population (reviewed in Ref. 1). This is not a new observation. 30 years ago, Hauschka [64] isolated two sublines from the Krebs 2 tumor line, and noting that both were more virulent than their parent, suggested that they had been 'held in check by other elements in the parental population'. Klein and Klein [65] described the masking of cells with ascitic potential by other tumor cells which were unable to grow in this form. More recently, Woodruff and associates [66,67] also pointed out that clones from methylcholanthrene-induced tumors of mice are often less tumorigenic than their parent tumors

and suggested that some tumor subpopulations may require other subpopulations to survive. A particularly elegant analysis of tumor subpopulation interactions is that of Hauschka, Revesz and co-workers [68,69] with a set of subpopulations isolated from an Ehrlich ascites tumor line, ELD. Line ELD is composed of 96% diploid cells and 4% tetraploid cells. If this proportion was artificially upset by conditions which favored the selection of tetraploid cells, the stable 96:4 ratio was reestablished after removal of the selective influence. Addition of a few ELD cells to a tetraploid subline of ELD gradually resulted in reestablishment of a population consisting of 96% diploid, 4% tetraploid cells. These observations suggest that the diploid subpopulations have a growth advantage, as compared to the tetraploid cells. This advantage was demonstrated by experiments in which mixtures of cells, in varying proportions, were injected into mice. However, if growth of the two subpopulations had been independent, the percentage of diploid ceils should have continued to increase, approaching 100% asymptotically, unless new tetraploid cells arose from the diploid cells. Endo-reduplication resulting in the formation of new tetraploid cells did occur, but the rate of production was not sufficient to maintain a 4~ level. Analysis of mixture experiments suggested that growth of ELD was kept at the 96 : 4 balance through a mechanism responsive to the combined volume of the diploid and tetraploid tumor masses. Cell subpopulations which are not responsive to the same mechanism, and hence have different volume restrictions, would not be expected to maintain a stable balance in the face of different growth capacities. The analysis of Jansson and Revesz suggests that, whatever the mechanisms responsible for maintaining cellular heterogeneity in regard to growth within any tumor, they will depend upon a sharing of growth regulatory signals which are somehow linked to the limits in total population size attainable. This suggests further that the mechanisms of growth regulation will be 'specific' to the interacting types, a conclusion supported by the observations that some tumor cell subpopulations show independent growth in mixtures [68]. Our own interest in subpopulation interactions within neoplasms began with the isolation of five

218

distinct subpopulations from a single, spontaneously arising strain B a l b / c f C 3 H mouse mammary tumor [4,36]. Four of the subpopulations (numbers 66, 67, 168 and 68H) were derived from the autochthonous tumor; the fifth subpopulation (number 410) was isolated from a single metastatic nodule growing in the lung of a mouse bearing a subcutaneous implant of cells from the parental tumor in its tenth serial transplantation passage. In addition to many other differences, these subpopulations differ in percentage of animals in which tumors grow after injection of a given number of cells and in the time required for tumors to appear. In order to see whether subpopulation interactions could influence tumor growth, we performed a series of experiments in which syngeneic mice were injected on contralateral sides with cells of different subpopulations [70]. Control mice received either one injection or two (bilateral) injections of the same cells. The results showed that, depending upon the particular combination of subpopulations tested, the presence of one subpopulation could either retard or accelerate the appearance of another subpopulation. One particular combination of subpopulations was selected to investigate the mechanism of this growth interaction. In this combination, subpopulation 410 interfered with growth of subpopulation 168. Line 410 is an immunogenic, slowly growing tumor, whereas, by itself, 168 is rapidly growing and almost devoid of immunogenicity in hosts of the strain of origin of the parental tumor. This immunological heterogeneity between the two subpopulations provides the basis for their growth interaction [70]; line 410 not only induces immunity against itself, it does so against line 168. When the two lines are introduced into the same mouse, the immunity gives some protection against 168, seen as a partial inhibition of growth. Abrogation of the immune response by irradiating the hosts prior to transplantation results in abrogation of the growth interaction as well. In this case, the tumor subpopulation interaction is, in fact, hostmediated. Another example of tumor subpopulation growth interactions which employ host mechanisms is found in the reports of Kyner et al. [71] and Newcomb et al. [72]. These investigators used

clones of tumor cells derived from the B16 melanoma. Two of the clones were tumorigenic and also produced plasminogen activator. A third clone did not produce plasminogen activator and did not grow in normal, syngeneic mice, but was tumorigenic in immunosuppressed mice. Co-cultivation of this latter clone in vitro with a plasminogen activator-producing clone suppressed plasminogen activator production. Mixtures of the two types of clone failed to grow as tumors in normal mice, although they did do so in immunosuppressed mice. The authors speculate that the ability of the non-plasminogen activator-producing clone to suppress the growth of the other clones in vivo was due to its concomitant ability to suppress plasminogen activator production. Plasminogen activator was postulated to suppress host immunity by interfering with the development of an inflammatory response at the site of the tumor. In the absence of plasminogen activator, not only was the immune response active against the 'nontumorigenic' clone, it inhibited growth of the other clones as well. A recent report of Galli et al. [73] suggests another mechanism whereby host reactivity to one tumor subpopulation could affect the growth of another. These investigators showed a growth interaction between two antigenically unrelated tumors, line 1 and line 10 hepatomas of guinea pigs. Line 1 hepatoma is an immunogenic tumor which ultimately is rejected, whereas line 10 grows rapidly and metastasizes. When these two tumors were injected on opposite sides of the same guinea pig, they each followed their characteristic growth pattern. Mixtures of line 1 and line 10 cells, however, grew and regressed in the same way as did line 1 tumors alone. Histological examination provided a basis for this interaction. Line 1 tumors are associated with an extensive stromal proliferation and a peripheral concentration of host inflammatory cells at the site of the tumorsupporting vasculature. Tumor regression occurs secondary to ischemia. Line 10 tumors induce less stromal and leucocytic infiltration, and their microvasculature is spared the injury seen in line 1 tumors. The line l-line 10 mixtures were characterized by a line 1-type host response. This situation was only seen in tumors injected into non-immune hosts. In pre-immunized hosts, the stromal pattern was altered and line 1 cells could

219

no longer suppress line 10 cells. Although this work was not done with subpopulations of the same tumor, it suggests that differential abilities of tumor clones to influence the host component of tumors could result in a masking of the different growth and behavioral characteristics of subpopulations of heterogeneous tumors. Not all growth interactions between tumor subpopulations depend upon the host. We have also detected interactions among mammary tumor subpopulations in vitro. For example, when subpopulations 168 and 68H were co-cultured in monolayer cultures, the relative proportions of the two lines remained constant, even though by themselves 168 cultures had a much shorter doubling time than did 68H cultures [74]. In individual experiments, this 'balanced growth' was achieved either by a slowing down of 168 or a speeding up of 68H. The co-cultures grew at the rate of one or the other subpopulation when cultured alone. Inspection of the co-cultures showed patches of 68H and of 168 cells - not an intermingling of the two. This suggested that the two cell types did not have to be in physical contact for the interaction to occur. However, when the subpopulations were separated by culturing them on individual coverslips within a common petri dish, a different type of growth interaction was seen. Under these circumstances, no stimulation of growth occurred; rather, line 168 cells inhibited growth of 68H cells, as well as of some other, but not all, subpopulations from the same parental tumor. Subpopulation 68H cells were not inhibitory to other cells, although cells of one other subpopulation, line 66, were. The growth inhibitory activity of 168 cells could be reproduced with cell-free media from cultures of 168 cells. Inhibitory activity was removed by centrifugation at 108000 x g for 30 min and by a 0 . 2 / t m filter and was extremely labile to heating or to storage at temperatures ranging from 37 to - 2 0 ° C [74]. Inhibition extended to unrelated mammary tumor cells, including those of the human MCF-7 breast cancer line, but not to a limited sample of other types of tumor cells nor to 3T3 cells [75]. Thus, two types of growth interaction were demonstrated in vitro. One was associated with the presence of a factor released into media of growing cells. The effect of its activity was inhibitory. The other was only demonstrated in co-cultures of

cells, where both growth stimulation and inhibition were found. The latter effect may, or may not, be related to the media factor. Interactions between tumor subpopulations have been reported to affect cellular characteristics other than growth. Mixtures of immunogenic and non-immunogenic sublines of the TA3 mouse mammary tumor line could immunize animals to both or neither subline, depending upon the ratio of the two [76]. Olsson and Ebbesen [21] showed that immunotherapy with artificial mixtures of clones isolated from an A K R leukemia was much more effective than was therapy with either clone alone or with uncloned leukemia cells. Again, the effect of the relative proportion of subtypes on the immunogenicity of the whole seems evident. W a n g et al. [32] found that immunization with individual clones from a m e t h y l c h o l a n t h r e n e - i n d u c e d sarcoma protected against the parental tumor, even though the individual clones did not necessarily immunize against each other. These authors speculated that a mixed population of antigenically-related cells may induce a qualitatively different immune response than do 'antigenically pure' cells. Using the subpopulations of a mammary tumor discussed above, we have also found that immunization with mixtures of subpopulations may result in an overall immunity different from that which would be predicted on the basis of the cross-reactive patterns determined by immunization with the individual clones [77]. The mechanisms of immunogenicity interactions have not been determined. Perhaps antigen expression of one subpopulation is altered by the presence of another subpopulation. Antigens which are dependent upon stage of cell cycle may be enhanced or decreased by alterations in growth behavior brough about by growth interactions between tumor subpopulations. Alternatively, immune interactions between tumor subpopulations may be due to factors occurring in the host. 'Antigenic competition' has been described between many antigens, and a similar phenomenon might also operate within tumors. It is also likely that different tumor subpopulatins might induce different immune effector mechanisms. Our individual mammary tumor subpopulations induce characteristic lymphocyte infiltrates which are distinct in both the absolute number and relative distribution

220 of Lyt 1 + vs. Lyt 2 + T cells [78]. Abilities among subpopulations to induce cytotoxic vs. suppressor responses can thus be distributed differentially among tumor subpopulations and could result in unexpected immune responses to mixtures, depending upon the particular subpopulations and their relative proportions. As was discussed earlier, the impact of tumor heterogeneity on host inflammatory and stromal reactions may also be at the heart of some tumor subpopulation growth interactions which do not depend on specific immune reactivity per se. Another tumor cell characteristic which can be modified by tumor subpopulation interactions is sensitivity to antineoplastic drugs. Again using our mammary tumor model, we have found that one subpopulation can alter the response of another subpopulation to such drugs as cyclophosphamide, methotrexate and thioguanine [79,80]. The mechanisms of interaction may be mediated via the host. For example, host metabolism of cyclophosphamide seems to be modified by the presence of subpopulation 168, so that subpopulation 410 is inhibited by doses of drug that are ineffective in the absence of 168 cells. The mechanisms, as in the cases of methotrexate and thioguanine, may also be dependent upon the tumor cells themselves. They may operate through 'factors' (as with methotrexate) or be dependent upon cell-cell contact (thioguanine). The mechanisms operating in any particular combination of subpopulations depend upon the specific characteristics of the cells and drugs involved. The final effect, however, is a therapeutic response that would not necessarily be predicted on the basis of the responses of single clones. The spread and growth of metastases are yet other aspects of cancer cell behavior which can be influenced by tumor subpopulation interactions. There are numerous anecdotal, as well as experimental, accounts of interactions between primary tumors and their metastases (Refs. 81-86; for a recent review, see Ref. 87). Removal of a primary mass is sometimes followed by a seeming burst in metastatic growth. Among the numerous mechanisms which may account for these observations are growth interactions (see above) between metastatic subpopulations and other subpopulations in the primary tumor. Direct evidence for this is,

however, lacking. Another possibility is that subpopulation interactions alter the metastatic phenotype per se. Recently published data from our laboratory [88] suggest that relatively non-metastatic subpopulations may metastasize in the presence of other, highly metastatic subpopulations. For example, when mice bearing subcutaneous tumors of subpopulation 168 were injected i.v. with saline, with 1 . 1 0 4 168 cells, or with 1 . 1 0 4 410.4 cells, only animals injected with 410.4 cells subsequently developed lung nodules, but these nodules consisted of clonogenic 168, as well as 410.4, cells. Although these results may be due to a systemic effect of the metastatic cells on host immunity, Poste and Nicolson [89] reported that fusion of membrane vesicles prepared from the metastatic F10 subpopulation of line B16 melanoma with 'non-metastatic' F1 cells resulted in F1 metastases. Since shedding of membrane vesicles can occur in vivo [90], one may speculate that they could transfer metastatic behavior under natural circumstances from one subpopulation to another. The most intriguing example of tumor subpopulation interactions which affect metastasis has recently been presented by Poste and associates [91,92]. These workers noted that individual clones of B16 melanoma were unstable with regard to the ability to metastasize to lung when propagated by themselves, either in vitro or in vivo: after relatively few generations, the clones gave rise to cells exhibiting a wide range of metastatic ability. However, when the clones were grown together, the phenotype of the individual clones 'stabilized', that is, variants with different abilities to metastasize were not detected. Such stabilization of phenotypic diversity was 'specific' in that it could not be effected by clones from histologically unrelated tumors or by normal cells and it was not dependent upon the metastatic phenotype of the participating clones. Thus, mixtures of highly metastatic clones could stabilize, as well as mixtures of low or high plus low clones. These results have been repeated with another set of B16 clones selected on the basis of metastasis to brain [93]. A possibly related observation is that of Natali et al. [94] who used a panel of monoclonal antibodies to normal and melanoma-associated antigens to define the extent of antigenic heterogeneity

221 in primary and metastatic nodular melanoma of humans. Interestingly, the metastatic lesions for a patient who still carried their primary cancer were less heterogeneous than metastases from patients whose primaries had been removed, suggesting that interactions between metastatic and nonmetastatic cells might have limited the production of further variants within the metastatic lesions. (The production of tumor cell variants within metastatic foci has been described recently by Poste et al. [92] in the B16 system.) The mechanisms whereby clonal interactions could effect stabilization of the metastatic, or any other, phenotype are as yet unknown. Since the clonal instability occurred with either in vitro or in vivo passage, host effects would not appear to be critical, at least in the B16 system. Whether the stabilization occurs at a genetic or epigenetic level is also not known. Although the high degree of clonal instability for metastatic behavior may suggest an epigenetic basis, some genetic events, such as gene amplification, can be quite unstable in the absence of continuous selective pressure [95]. Furthermore, the metastatic phenotype is multifactorial; variation in one of many different factors could alter metastatic frequency. It is also possible that the target of the interaction is not at the level of the generation of variants at all, but rather on their growth. Thus, just as with the mammary tumor and ELD subpopulations discussed earlier, growth regulatory mechanisms may exist to assure an equilibrium in the volume of the individual clones. These interactions would mask the underlying instability of the individual clones in the mixtures. III. Cellular interactions in normal tissues

The interactions discussed above suggest that cancer cells may retain the growth regulatory networks of their respective normal tissues, but in an aberrent form. Parallels of tumor subpopulation interactions have been studied by embryologists and developmental biologists with normal cell populations for years. It is of interest to compare the types of mechanism utilized by normal cells with those utilized by tumor cell subpopulations. As discussed in an excellent review by Simnett [96], the regulation of cell division in normal dif-

ferentiated tissues, including various growth phenomena such as developmental growth, cell turnover, hypertrophy, hyperplasia and cell loss during development, wound healing and regeneration, can best be explained by proposing the existence of an intracellular program or clock which determines the rate and number of cell divisions available to each cell, along with mechanisms of interaction between cells, tissues and organs, which can modify the intracellular programs. The mechanisms of interaction regulating normal cell division can be classified under several headings, including autoregulatory factors, tissue interaction factors and host factors, as well as control mechanisms acting through positional effects. Evidence for each of these has been reviewed by Simnett [96]. Briefly, autoregulatory stimulation factors or 'wound hormones' have been postulated to play a role in wound healing in liver [97], kidneys [98] and bowel [96], and to be released by damaged granulocytes and erythrocytes [97]. There are a number of growth-promoting serum proteins and growth factors from various tissues which have the ability to influence the growth of cells in culture. These factors include somatomedins and prostaglandins as well as nerve, epidermal, ovarian and fibroblast growth factors [99]. Chalones are postulated autoregulatory inhibitory factors. A deficit of cells in a tissue is believed to lead to a reduction in these inhibitors, so that the rate of cell division can increase. These substances have been described in homogenates of epidermis [100], liver [101], fibroblasts [102], kidney [103], as well as in a variety of other tissues [96]. Host control mechanisms which are believed to regulate normal cell division include those acting through hormones, which may affect tissues which are not part of recognized target organs, as well as the known endocrine target tissues. For example, parathyroid hormone appears to play a role in allowing liver cells to proliferate after partial hepatectomy [104]. The control of cell division by mechanisms acting through the nervous system is exemplified by studies on the regeneration of amphibian limbs, in which neurotropic agents appear to stimulate cell division [105]. The apparent stimulatory role of lymphocytes in liver regeneration [106] is an example of another host regulatory mechanism.

222

Growth control by tissue interaction, in which adjacent tissues interact to control each other's growth, is exemplified by the phenomenon of induction, which is important in the development of many embryonic organs such as liver, kidney and lung [107]. That there exists a mechanism by which cell division is regulated, not necessarily by factors released into the general circulation to which all target ceils respond, but in a way related to the location of the tissue or in response to local physiological changes, has been demonstrated by a series of experiments in which small implants of liver [108] or kidney [109] were grafted into one lung of frogs, after which the contralateral lung was removed. These experiments have shown that cell division was increased not only in the remaining lung, but in the ectopic tissue as well. Any or all of the above mechanisms may play a role in tumor subpopulation interactions. For example, the growth inhibitory factor released by mammary subpopulation 168 cells resembles the chalones in its apparent organ-specificity, species non-specificity. In addition, modification of these means of interaction may be a way by which tumor cells express their special properties. Several laboratories have described a factor or factors released by murine sarcomas and certain human tumor cells which have the ability to stimulate anchorage-independent growth, migration, and other transformation properties in normal fibroblasts [110-112]. Sarcoma growth factor has been purified and characterized by the laboratory of Todaro and has been shown to bind to epidermal growth factor (EGF) receptors on target cells [113]. DeLarco and Todaro have postulated that these factors act as mitogens to the cells producing them, as well as to normal target cells [110]. Although differential production of such diffusable growth factors by tumor subpopulations has yet to be demonstrated, one can postulate that they may be a means whereby 'fast-growing' subpopulations may act to keep their slower-growing associates in equilibrium. Another mechanism by which normal cells may interact is by means of factors passed between cells in contact. Cells of organized tissues have intercellular communication junctions through which small molecules (up to 1 kDa) may pass [115]. As a general rule, a given cell in a tissue

interconnects via such channels with many neighbors. For example, all cells in a salivary gland or thyroid acinus are interconnected and cells of liver and skin are so widely interconnected as to form a continuous network [116]. The ability to transfer molecules in this way has been identified with the phenomenon of 'metabolic cooperation', the correction of a defective phenotype of variant cells by contact with cells of the normal phenotype [117]. There is circumstantial evidence that the structure through which metabolically coupled cells communicate is the gap junction [118]. Tumor subpopulation'interactions which affect sensitivity to thioguanine (see above) appear to be due to the ability of cells which can metabolize that drug to transfer the active metabolite to otherwise insensitive variants which are insensitive due to a defect in the enzyme, hypoxanthine-guanine phosphorybosyl transferase. The data on communication competence, i.e., the ability of cells to transfer ions or small molecules through gap junctions, have been reviewed by Loewenstein [118]. All normal tissues capable of undergoing cell division, including embryological tissue, regenerating tissue and all adult tissues so far investigated except (non-dividing) skeletal muscle and nerve cells, are at least electrically coupled: The hypothesis has been formulated by Loewenstein that intercellular communication through communication channels is instrumental in control of growth [118]. If communication channels are involved in growth control, defects in the junctional channels (by inheritable defect) may result in disturbance of cellular growth control and, if so, at least some malignant tumors may be defective in junctional communication [118]. This prediction has been tested with a number of solid benign and malignant tumors in vivo using electrophysiological techniques, and in cell culture, using both electrophysiological and metabolic cooperation assays. Various results have been reported: although in some cases decreases in communication ability have been found [119-122], in other cases tumor cells have been found which are fully capable of communication with other cells [123-125]. Gap junctions have been described in a number of tumors, including metastatic tumors [126-128], al-

223 though in a few tumor systems, a reduction of gap junctions was found to be correlated with tumor progression [129,130]. We have found that the relative ability to communicate via communication channels varies with our mammary tumor subpopulations [131]. For example, on the basis of metabolic cooperation assays using a thioguanine resistant clone of subpopulation 66, we have been able to classify our subpopulations 66 and 68H as good, 410 and 67 as intermediate and 168 as poor communicators. Thus, although it appears that there are differences in the degree and specificity of contactmediated intercellular communication and in the number of gap junctions found in some tumor cells, tumor subpopulations can retain their ability to interact through this mechanism. Riley [99] has argued that a stable, co-existing homogeneous mixture of two competing populations is possible if, and only if, one exerts a greater inhibitory effect on itself than on its competitor; in other words, to permit steady-state populations when both populations occupy the same environmetnal niche, some form of self-limitation is necessary. Riley considers this 'cellular altruism' an indispensible requirement for differentiated cells to exist in multicellular organisms. Again, the mechanism for this self-limitation may be either through self-inhibiting diffusable agents (chalones, non-specific agents) or contact-mediated self-inhibition, or through stimulation of the competing population (hormones, growth factors). These same mechanisms, inherited by neoplastic cells from their normal counterparts, may act to balance competing tumor subpopulations and thus maintain tumor heterogeneity. In addition to these types of cellular interaction, Riley has made several interesting observations on the nature of competition between cells in vitro, which can be applied to both normal and neoplastic cells in vivo as well, and which suggest that it is not necessary to postulate special 'mechanisms' of interaction to explain why tumor subpopulations with a seeming growth or behavioral advantage do not dictate the characteristics of the tumor as a whole. Thus, given the clonal nature of cellular propagation and the limited mobility of cells in solid tissue or in culture on a solid support, subpopulations which may be in competition for

limited space or resources would be expected to be arranged in zones rather than completely mixed together. Indeed, several investigators have described such zonal distributions of tumor subpopulations within solid tumors [132-134]. In our coculture experiments described earlier we also noted that the interacting subpopulations, 68H and 168, were distributed within patches in the culture dish. Most conditions which limit growth act at short range, due to limits in diffusability in the case of nutrients or waste products. Therefore, two populations will compete primarily at the frontier between them, with the result that a substantial proportion of the otherwise less successful competitor may survive. Thus, what is seen as the protection of one subpopulation from overgrowth of a more aggressive neighbor may not be due to a positive interaction but rather as a consequence of their spatial relationships. IV. Conclusions Several conclusions can be drawn from this review of tumor subpopulation interactions in neoplasms. Perhaps the most general is that, although tumors may be heterogeneous populations, knowledge of the characteristics of the individual subpopulations is insufficient to predict the behavior of the whole cancer. The facts that one can isolate clonal subpopulations from tumors and show that they differ in important clinical and biological characteristics are in themselves meaningless without an appreciation that these distinctions may be lost when these same clones exist in a single tumor. Examples of tumor subpopulation interactions affecting growth, metastasis, immunogenicity and drug sensitivity have been cited. Clearly, the idea that only certain subpopulations of tumor are, by virtue of the ability to metastasize, etc., 'important' is an oversimplification. The behavior of a cancer is the result of all the subpopulations within it. The second conclusion is that numerous mechanisms for subpopulation interactions exist. In some cases the host may be a partner in the interaction. In other cases the tumor cells mediate the interaction directly, either by the production or release of diffusable 'factors', or via direct cell contact. Indeed, different clones may influence each other not by any defined

224

mechanism at all, but by presenting a border of competition for shared nutrients. A third conclusion is that, although the phenomenon of cellular interactions has been discussed in the context of neoplasia, it is not a neoplastic phenomenon per se. T h e obligation for cells to behave as members of a society is a fundamental principle of biology. Tumor cells cannot escape this obligation; even if they become 'autonomous' of many normal regulatory influences, they still must deal with spatial controls a n d with each other. As a target for regulation of neoplastic growth and metastasis, tumor subpopulation interactions may prove to be the Achilles' heel of neoplasia. At the very least, such interactions must be recognized in both experimental studies and clinical applications of tumor cell biology.

Acknowledgements The authors wish to thank Ms. Pat Pillon for her skillful, cheerful and repeated typing of this manuscript. Our own work discussed herein was supported by NIH grants CA-27419, CA-27437 and CA-28366, by a grant from Concern Foundation, and by the E. Walter Albachten bequest.

References 1 Heppner, G.H. and Miller, B.E. (1983) Cancer Metastasis Rev. 2, 5-23 2 Henderson, J.S. and Rous, P. (1962) J. Exp. Med. 115, 1211-1229 3 Pierce, G.B. (1974) in World Symposium on Model Studies in Chemical Carcinogenesis (T'so, P.O.P., DiPaolo, J.A., eds.), pp. 463-472, Dekker, New York 4 Dexter, D.L., Kowalski, H.M., Blazar, B.A., Fligiel, Z., Vogel, R. and Heppner, G.H. (1978) Cancer Res. 38, 3174-3181 5 Baylin, S.B., Weisburger, W.R., Eggleston, J.C., Mendelsohn, G., Beaven, M.A., Abeloff, M.D. and Eninger, D.S. (1978) N. Engl. J. Med. 299, 105-110 6 Siraeky, J. (1979) Br. J. Cancer 39, 570-577 7 Brennan, M.J., Donegan, W.L. and Appleby, D.E. (1979) Am. J. Surg. 137, 260-262 8 Macinnes, J.I., Chan, E.C.M., Percy, D.H. and Morris, V.L. (1981) Virol. 113, 119-129 9 Brattain, M.G., Fine, W.D., Khaled, F.M., Thompson, J. and Brattain, D.E. (1981) Cancer Res. 41, 1751-1756 10 Dexter, D.L., Spremulli, E.N., Fligiel, Z., Barbosa, J.A., Vogel, R., VanVorhees, A. and Calabresi, P. (1981) Am. J. Med. 71, 949-956

11 Lippman, S.M., Mendelsohn, G., Trump, D.L., Wells, S.A. and Baylin, S.B. (1982) J. Clin. Endo. Med. 54, 233-240 12 Michalides, R., Wagenaar, E. and Sluyser, M. (1982) Cancer Res. 42, 1154-1158 13 Nuti, M., Horan Hand, P., Colcher, D. and Schlom, J. (1982) Proc. Am. Assoc. Cancer Res. 23, 267 14 Varani, J., Orr, W. and Ward, P.A. (1978) Am. J. Pathol. 90, 159-172 15 Kripke, M.L., Gruys, E. and Fidler, I.J. (1978) Cancer Res. 38, 2962-2967 16 Gray, J.M. and Pierce, G.B. (1964) J. Natl. Cancer Inst. 32, 1201-1211 17 Fidler, I.J. and Kripke, M.L. (1977) Science 197, 893-895 18 Albino, A.P., Lloyd, K.O., Houghton, A.N., Oettgen, H. and Old, L.J. (1981) J. Exp. Med. 154, 1764-1778 19 Shapiro, J.R., Yung, W.A. and Shapiro, W.R. (1981) Cancer Res. 41, 2349-2359 20 Wikstrand, C.J., Bigner, S.P. and Bigner, D.D. (1982) Proc. Am. Assoc. Cancer Res. 23, 1070 21 Olsson, L. and Ebbesen, P. (1979) J. Natl. Cancer Inst. 62, 623-627 22 Mathieson, B.J., Zatz, M.M., Sharrow, S.O., Asofsky, R., Logan, W. and Kanellopoulos-Langevin, C. (1982) J. Immunol. 128, 1832-1838 23 Ogawa, K., Solt, D.B. and Farber, E. (1980) Cancer Res. 40, 725-733 24 Ashley, R.L., Cardiff, R.D., Mitchell, D.J., Faulkin, L.J. and Lund, J.K. (1980) Cancer Res. 40, 4232-4242 25 Rabes, H., Bucher, T.L., Hartmann, A., Linke, I. and Dunnwald, M. (1982) Cancer Res. 42, 3220-3227 26 Byers, V.S. and Johnston, J.O. (1977) Cancer Res. 37, 3173-3183 27 Pimm, M.V. and Baldwin, R.W. (1977) Int. J. Cancer 20, 37-43 28 Fuji, H., Mihich, E. and Pressman, D. (1977) J. Immunol. 119, 983-986 29 Killion, J.J. (1978) Cancer Immunol. lmmunother. 4, 115-119 30 Schirrmacher, V., Bosslet, K., Shantz, G., Clauer, K. and Hubsch, DI (1979) Int. J. Cancer 23, 245-252 31 Miller, F.R. and Heppner, G.H. (1979) J. Natl. Cancer Inst. 63, 1457-1464 32 Wang, N., Yu, S.H., Liener, I.E., Hebbel, R.P., Eaton, J.W. and McKhann, C.F. (1982) Cancer Res. 42, 1046-1051 33 Miller, F.R., Miller, B.E. and Heppner, G.H. (1983) Invasion Metastasis 3, 21-31 34 Barranco, S.C., Ho, D.H.W., Drewinko, B., Romsdahl, M.M. and Humphrey, R.M. (1972) Cancer Res. 32, 2733-2736 35 Hakansson, L. and Trope, C. (1974) Acta Path. Microbiol. Scand. Section A, 82, 41-47 36 Heppner, G.H., Dexter, D.L., DeNucci, T., Miller, F.R. and Calabresi, P. (1978) Cancer Res. 38, 3758-3763 37 Trope, C., Aspegren, K., Kullander, S. and Astedt, B. (1979) Acta Obstet. Gynecol. Scand. 58, 543-546 38 Tsuruo, T. and Fidler, I.J. (1981) Cancer Res. 41, 3058-3064

225 39 Yung, W.A., Shapiro, J.R. and Shapiro, W.R. (1982) Cancer Res. 42, 992-998 40 Hill, H.Z., Hill, G.J., Miller, C.F., Kwong, F. and Purdy, J. (1979) Radiat. Res. 80, 259-276 41 Leith, J.T., Brenner, H.J., DeWyngaert, J.K., Dexter, D.L., Calabresi, P. and Glicksman, A.S. (1981) Int. J. Radiat. Oncol. Biol. Phys. 7, 943-947 42 Leith, J.T., Dexter, D.L., DeWyngaert, J.K., Zeman, E.M., Chu, M.Y., Calabresi, P. and Glicksman, A.S. (1982) Cancer Res. 42, 2556-2561 43 Nowell, P.C. (1976) Science 194, 23-28 44 Fidler, l.J. and Hart, I.F. (1982) Science 217, 998-1002 45 Sager, R. (1982) in Bristol-Myers Cancer Symposia, Vol. 4 (Owens, A.H., Coffey, D.S. and Baylin, S.B., eds.), pp. 411-423, Academic Press, New York 46 Pierce, G.B. and Cox, W.F., Jr. (1978) in Cell Differentiation and Neoplasia (Saunders, G.F., ed.), pp. 57-66, Raven Press, New York 47 Kerbel, R.S., Dennis, J.W., Largarde, A.E. and Frost, P. (1982) Cancer Metastasis Rev. 1, 99-140 48 Poste, G. and Greig, R. (1982) Invasion Metastasis 2, 137-176 49 Dexter, D.L. and Calabresi, P. (1982) Biochim. Biophys. Acta 695, 97-112 50 Weiss, L. (1980) in Cell Biology of Breast Cancer (McGrath, C.M., Brennan, M.J. and Rich, M.A., eds.), pp. 189-205, New York, Academic Press 51 Calabresi, P., Dexter, D.L. and Heppner, G.H. (1979) Biochem. Pharm. 28, 1933-1941 52 Miller, F.R. (1982) Cancer Metastasis Rev. 1, 319-334 53 Owens, A.H., Coffey, D.S. and Baylin, S.B. (1982) BristolMyers Cancer Symposia, Vol. 4 (Owens, A.H., Coffey, D.S. and Baylin, S.B., eds.), Academic Press, New York 54 Rubin, H. (1980) J. Natl. Cancer Inst. 64, 995-1000 55 Risser, R. and Pollack, R. (1974) Virology 59, 477-489 56 Stoker, M. (1964) Virology 24, 165-174 57 DeOme, K.B., Miyamoto, M.J., Osborn, R.C., Guzman, R.C. and Lum, K. (1978) Cancer Res. 38, 2103-2111 58 Medina, D., Shepherd, F. and Gropp, T. (1978) J. Natl. Cancer Inst. 60, 1121-1126 59 Slemmer, G. (1974) J. Invest. Dermatol. 63, 27-47 60 Ranadive, K.J. and Bhide, S.V. (1962) in Biological Interactions in Normal and Neoplastic Growth (Brennan, M.J. and Simpson, W.L., eds.), pp. 337-354, Little, Brown & Co., Boston, MA 61 Nandi, S. (1978) J. Environ. Pathol. Tox. 2, 13-20 62 Prehn, R.T. (1972) Science 176, 170-171 63 Miller, F.R., Medina, D. and Heppner, G. (1981) Cancer Res. 41, 3863-3867 64 Hauschka, T.S. (1953) J. Natl. Cancer Inst. 14, 723-736 65 Klein, G. and Klein, E. (1956) Ann. N.Y. Acad. Sci. 63, 640-661 66 Woodruff, M. (1982) Br. J. Cancer 46, 313-322 67 Woodruff, M.F.A., Ansell, J.D., Forbes, G.M., Gordon, J.C., Burton, D.I. and Micklem, H.S. (1982) Nature 299, 822-824 68 Jansson, B. and Revesz, L. (1976) in Methods in Cancer Research (Busch, H., ed.), Vol. XIII, pp. 227-290, Academic Press, New York

69 Hauschka, T.S., Grinnell, S.T., Revesz, L. and Klein, G. (1957) J. Natl. Cancer Inst. 19, 13-32 70 Miller, B.E., Miller, F.R., Leith, J. and Heppner, G.H. (1980) Cancer Res. 40, 3977-3981 71 Kyner, D., Christman, J., Acs, G., Silagi, S., Newcomb, E.W. and Silverstein, S.C. (1978) J. Cell Physiol. 95, 159-167 72 Newcomb, E.W., Silverstein, S.C. and Silagi, S. (1978) J. Cell Physiol. 95, 169-177 73 Galli, S.J., Bast, R.C., Bast, B.S., Isomura, T., Zbar, B., Rapp, H.J. and Dvorak, H.F. (1982) J. Immunol. 129, 890-899 74 Heppner, G., Miller, B., Cooper, D.N. and Miller, F.R. (1980) in Cell Biology of Breast Cancer (McGrath, C.M., Brennan, M.J. and Rich, M.A., eds.), pp. 161-172, Academic Press, New York 75 Heppner, G.H. (1982) in Bristol-Myers Cancer Symposia, Vol. 4 (Owens, A.H., Coffey, D.S. and Baylin, S.B., eds.), pp. 225-236, Academic Press, New York 76 Nowotny, A. and Grohasman, J. (1973) Int. Arch. Allergy 44, 434-440 77 Miller, F.R. and Heppner, G.H. (1979) Proc. Am. Assoc. Cancer Res. 21, 201 78 Rios, A., Miller, F.R. and Heppner, G. (1983) Cancer Immunol. Immunother. 15, 87-91 79 Miller, B.E., Miller, F.R. and Heppner, G.H. (1981) Cancer Res. 41, 4378-4381 80 Miller, B.E. and Heppner, G.H. (1981) Breast Cancer Res. Treat. 1, 163 81 DeWys, W.D. (1972) Cancer Res. 32, 374-379 82 Greene, H.S.N. and Harvey, E.K. (1960) Cancer Res. 20, 1094-1100 83 Milas, L., Hunter, N., Mason, K. and Withers, H.R. (1974) Cancer Res. 34, 61-71 84 Yuhas, J.M. and Pazmino, N.H. (1974) Cancer Res. 34, • 2005-2010 85 Goldie, H., Walker, M., Kelley, L. and Gaines, J. (1956) Cancer Res. 16, 553-558 86 Sugarbaker, E.V., Thorntwaite, J. and Ketcham, A.S. (1977) in Cancer Invasion and Metastasis: Biologic Mechanisms and Therapy (Day, S.D., eel.), pp. 227-238, Raven Press, New York 87 Gorelik, E., Segal, S., Shapiro, J., Katzav, S., Ron, Y. and Feldman, M. (1982) Cancer Metastasis Rev. 1, 83-94 88 Miller, F.R. (1983) Invasion Metastasis, in the press 89 Poste, G. and Nicolson, G. (1980) Proc. Natl. Acad. Sci. U.S.A. 77, 399-403 90 Dvorak, H.F., Quay, S.C., Orenstein, N.S., Dvorak, A.M., Hahn, P. and Bitzer, A.M. (1981) Science 212, 923-924 91 Poste, G., Doll, J. and Fidler, I.J. (1981) Proc. Natl. Acad. Sci. U.S.A. 78, 6626-6230 92 Poste, G., Tzeng, J., Doll, J., Greig, R., Rieman, D., Zeidman, I. (1982) Proc. Natl. Acad. Sci. U.S.A. 79, 6574-6578 93 Nicoison, G.L. and Poste, G. (1982) Curr. Probl. Cancer 6, in the press 94 Natali, P.G., Cavaliere, R., Bigotti, A., Nicotra, M.R., Russo, C., Ng, A.K., Giacomini, P. and Ferrone, S. (1983) J. Immunol. 130, 1462-1466

226 95 Kaufman, R.J., Brown, P.C. and Schimke, R.T. (1979) Proc. Natl. Acad. Sci. U.S.A. 76, 5669-5673 96 Simnett, J.D. (1981) in Regulation of Growth in Neoplasia (Sherbet, G.V., ed.), pp. 1-51, Karger, Basel 97 Teir, H., Lahtiharju, A., Alho, A. and Forsell, K.J. (1967) in Control of Cellular Growth in Adult Organisms (Teir, H. and Rytomaa, T., eds.), pp. 67-82, Academic Press, New York 98 Argyris, T.S. and Trimble, M.E. (1964) Anat. Rec. 150, 1-9 99 Riley, P.A. (1981) in Regulation of Growth in Neoplasia (Sherbet, G.V., ed.), pp. 131-198, Karger, Basel 100 Bullough, W.S. (1965) Cancer Res. 25, 1683-1727 101 Loginov, A.S., Speranskii, M.D., Arvin, L.I., Matyushina, E.D. and Magnitskii, G.S. (1976) Bull. Exp. Biol. Med. 82, 1852-1854 102 Houck, J.C. (1976) Prog. Clin. Biol. Res. 5, 193-210 103 Chopra, D.P. and Simner, J.D. (1969) Exp. Cell Res. 58, 319-322 104 Rixon, R.H. and Whitfield, J.F. (1974) Proc. Soc. Exp. Biol. Med. 146, 926-930 105 Singer, M. (1974) Ann. N.Y. Acad. Sci. 228, 308-322 106 Davies, A.J.S., Leuchars, E., Doak, S.M.A. and Cross, A.M. (1964) Nature 201, 1097-1101 107 Balmsky, B.I. (1975) An Introduction to Embryology, Saunders, Eastbourne 108 Simnett, J., Oates, C. and Walton, J. (1977) Experientia 33, 1457-1458 109 Simnett, J., Walton, J. and Oates, C. (1977) Anat. Rec. 187, 273-279 110 DeLarco, J.E. and Todaro, G.J. (1978) Proc. Natl. Acad. Sci. U.S.A. 75, 4001-4005 111 Todaro, G.J., Fryling, C. and DeLarco, J.E. (1980) Proc. Natl. Acad. Sci. U.S.A. 77, 5258-5262 112 Burk, R.B. (1980) in Control Mechanisms in Animal Cells (DeAsua, L.J., Levi-Montalcini, R., Shields, R. and Iacobelli, S., eds.), pp. 245-257, Raven Press, New York

113 Todaro, G.J. and DeLarco, J.E. (1978) Cancer Res. 38, 4147-4154

114 Twardzik, D.R., Ranchalis, J.E. and Todaro, G.J. (1982) Cancer Res. 42, 590-593 115 Staehelin, L.A. (1974) Int. Rev. Cytol. 39, 191-283 116 Loewenstein, W.R. (1980) Ann. N.Y. Acad. Sci. 339, 39-45 117 Subak-Sharpe, H., Burk, R.R. and Pitts, J.D. (1969) J. Cell Sci. 4, 353-367 118 Loewenstein, W.R. (1979) Biochim. Biophys. Acta 560, 1-65 119 Azarnia, R., Michalke, W. and Loewenstein, W.R. (1972) J. Membrane Biol. 10, 247-258 120 Nelson, P.G., Peacock, J.H. and Amano, T. (1971) J. Cell Physiol. 77, 353-362 121 Borek, C., Higashino, S. and Loewenstein, W.R. (1969) J. Membrane Biol. 1,274-293 122 Azarnia, R. and Loewenstein, W.R. (1976) J. Membrane Biol. 30, 175-186 123 Sheridan, J.D. (1970) J. Cell Biol. 45, 91-99 124 Johnson, I. and Sheridan, J.D. (1971) Science 174, 717-719 125 Harris, A.J. and Dennis, M.J. (1970) Science 167, 1255 126 Weinstein, R.S., Merk, F.B. and Alroy, J. (1976) Adv. Cancer Res. 23, 23-89 127 Letourneau, R.J., Li, J.J., Rosen, S. and Villee, C.A. (1975) Cancer Res. 35, 6-10 128 Gondos, B. (1969) Cancer 24, 954-958 129 McNutt, N.S., Hershberg, R.A. and Weinstein, R.S. (1971) J. Cell Biol. 51, 805-825 130 Pauli, B.U. and Weinstein, R. (1981) Experientia 37, 248-250 131 Miller, B.E., Roi, L.D., Howard, L.M. and Miller, F.R. (1983) Cancer Res. 43, 4102-4107 132 Prehn, R.T. (1970) J. Natl. Cancer Inst. 45, 1039-1045 133 Hakansson, L. and Trope, C. (1974) Acta Pathol. Microbiol. Scand. Suppl. (Sect. A) 82, 35-40 134 Fidler, l.J. and Hart, 1.R. (1981) Cancer Res. 41, 3266-3267