Cancer Metastasis Reviews 2:5-23 (1983) © 1983, Martinus Nijhoff Publishers. Printed in the Netherlands Tumor heterogeneity: biological implications and therapeutic consequences Gloria H. Heppner and Bonnie E. Miller Michigan Cancer Foundation, Detroit, M1 48201, USA Keywords: tumor heterogeneity, clonal instability, tumor progression, subpopulation interactions, therapy Summary It is now appreciated that cancers can be composed of multiple clonal s~bpopulations of cancer cells which differ among themselves in many properties, including karyotype, growth rate, ability to metastasize, immunological characteristics, production and expression of markers, and sensitivity to therapeutic modalities. Such tumor heterogeneity has been demonstrated in a wide variety of animal tumors of differing etiology, tissue and cellular origin, and species. It has been shown in autochthonous, as well as transplanted, tumors. Similar results have been reported for human cancers, although much of the evidence that heterogeneity of human cancers, also reflects, at least in part, the existence of clonal subpopulations, is still indirect. Heterogeneity is not a unique property of malignancy. Preneoplastic tumors, as well as normal tissues, are also composed of cellular subpopulations. Proposed mechanisms for the origin of tumor heterogeneity include coalescence of multiple loci of cancer clones and the generation of diverse subpopulations from a single clone. This latter process could be due to genetic errors arising from classical genetic mechanisms or to the production of cellular variants as in normal tissue differentiation. Indeed, certain tumor subpopulations have been shown to produce variants at high frequency. In some cases this frequency can be modified by environmental circumstances. Nontumor cells may also contribute to production of cancer cell variants, perhaps, in the case of infiltrating phagocytic cells, by producing mutagens or by somatic hybridization with cancer cells. Production of tumor cell variants is a dynamic process which can occur at any time. Although tumors are mixed populations of cells, knowledge of the characteristics of individual components is not sufficient to predict the behavior of the whole. Individual cancer subpopulations can interact to affect each other's growth, immunogenicity, ability to metastasize, sensitivity to drugs, and clonal stability. The existence of multiple, interactive subpopulations provides a basis for the well-known phenomenon of 'tumor progression' in which tumors undergo qualitative changes in characteristics over the course of time. Selection of subpopulations better able to survive changing environmental circumstances allows for such changes as autonomy in regard to endogenous growth regulation, more "malignant' behavior, and loss of response to therapy. Tumor subpopulation interactions may play a regulatory role in this process. Tumor heterogeneity has obvious consequences to the design of effective therapy. It provides one rationale for combination therapies and suggests that initial treatment should be early and comprehensive. The continuing emergence of new clones suggests that treatment which is unsuccessful at one point might be Address for reprints: Dr. Gloria H. Heppner, Department of Immunology, Michigan Cancer Foundation, 110 E. Warren Avenue, Detroit, MI 48201, USA
`
`IMMUNOGEN 2163, pg. 1
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`effective later. Assays to predict effective therapy for individual patients need to address the multiplicity of tumor subpopulations and the ability of these subpopulations to influence each other. Subpopulation interactions may also be useful in therapy design, as may be efforts to control the extent of tumor heterogeneity by agents which effect cellular differentiation. Thus, tumor heterogeneity presents both problems and, perhaps, new solutions for control of cancer. Introduction The idea that tumors are not uniform populations of 'cancer cells' has gained new strength in the past few years. Attention is now focused on the many ways by which cancers differ and on the basis for these differences. This has led to the rediscovery of concepts of tumor biology which were known to cancer researchers in the past but which had be- come lost during the euphoria of the revolution in molecular biology. The purposes of this review are to document the increasing evidence for one such concept - tumor heterogeneity - and to speculate on its implications to tumor biology and conse- quences to cancer therapy. Definition of tumor heterogeneity Tumors are 'heterogeneous' in several ways. There is the heterogeneity among cancers in different individuals who nominally have the same type of disease. It is this heterogeneity which fuels the search for prognostic indicators and for methods to individualize therapy. A second type of hetero- geneity is that seen within the same patient over the course of time. The biological, as well as the clinical, characteristics of an 'early', preinvasive tumor are not the same as those exhibited by the same cancer when it has disseminated. This type of hetero- geneity is acknowledged by Fould's concept of "progression' (1). Heterogeneity is also seen within a single tumor at any one time. Histological examination of tumor samples often reveals considerable differences in the morphology of cancer cells in different areas of the same lesion. Host infiltrating and connective tissue are not evenly distributed. Areas of necrosis may be present. Depending upon tumor size, marked disturbances in vasculature can occur, leading to focal differences in oxygen tension, pH, substrate supply, and waste drainage (2). Related in part to this structural heterogeneity is heterogeneity in growth compartments. The cells within a tumor may be cycling or noncycling, quiescent or repro- ductively dead (3). If cycling, they may be at any stage in the cycle. Insofar as stage of cell cycle may influence cellular properties such as membrane biochemistry (4, 5), antigen expression (6-8), sensi- tivity to immune killing (9, 10), drug cytotoxicity (11), and ability to metastasize (12, 13), tumors will be heterogeneous in regard to those properties. The type of heterogeneity which has received the most attention, and which is the subject of this review, is that due to the simultaneous existence of multiple clonal subpopulations within the same tumor. It is well to remember that such subpopula- tions are individually subject to all the other types of heterogeneity described above: as will be describ- ed, new subpopulations can arise during neoplastic progression. Furthermore, depending upon local conditions, structural and cell-cycle heterogeneity will be present within, as well as among, subpopula- tions. In addition, subpopulation heterogeneity im- poses additional structural heterogeneity on the tumor as a whole. Cells in individual subpopula- tions may be located in distinct areas, or zones, of a tumor, rather than comingling (14-16). The zonal distribution of tumor subpopulations needs to be taken into account in devising methods of sampling tumors for various types of analysis. Investigators who serially transplant tumors in vivo with pieces of tumor, rather than cell suspensions, in reality may be transplanting only certain subpopulations. Heterogeneity of experimental tumors The coexistence of multiple subpopulations of tumor cells within single neoplasms has been re-
`
`IMMUNOGEN 2163, pg. 2
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`peatedly demonstrated in animal tumors of diverse etiology and histological type. These include me- lanoma (17-19) lymphoma-leukemia (20, 21), sarcoma (14, 22-26), and carcinoma (27-35). Heterogeneity in tumors induced by chemical agents (24, 32), physical agents (19, 25, 26), steroids (28), or viruses (20-22, 27, 30, 33-35) has been described. Long-term passaged tumors (18, 23, 24, 31), tumors of recent origin (19, 25, 26), and autochthonous tumors (20, 30) have been the source of multiple subpopulations. At this time it appears that no class of neoplasm is excluded from being heterogeneous, but quantitative differences among classes may be revealed by further ex- perience. Tumor heterogeneity is manifested by a variety of phenotypic differences. Differences in cellular morphology (30) and tumor histopathology (21, 29, 36, 37), as well as differences in growth rate, both in vivo and in vitro, have been seen (17, 19, 30, 31, 37). Tumor subpopulations can differ in expression or production of 'markers' of differentiation, includ- ing appropriate pigment (16, 17), receptors (38), cell products (21 ), and specialized biosynthetic enzymes (28). Phenotypic diversity has also been reported for immunological characteristics, including anti- gen expression, immunogenicity, and sensitivity to immune attack (14, 20, 30, 34, 3944). (Immuno- logical heterogeneity has been reviewed in depth elsewhere in this series (45).) Perhaps the most significant phenotype by which tumor subpopula- tions can differ is ability to metastasize. Following the lead of the classic experiment by Fidler and Kripke with the BI6 mouse melanoma (18)~ the existence of tumor subpopulations that vary in ability to metastasize has been demonstrated in several experimental systems, including a recently isolated u.v.-induced melanoma (19), a variety of sarcomas (23, 25, 26), and mouse mammary tumors (31, 46, 47). Primary tumors contain subpopulations that can metastasize to specific organ sites at high, medium, or low frequency, relative to the parent tumor (23). On the other hand, subpopulations that are unable to metastasize (at least by themselves), and may even be unable to produce tumors except at high innocula and after prolonged latency periods (30, 37, 48, 49), can be isolated from highly tumorigenic parent neoplasms. As will be discussed below, the simultaneous existence within a single tumor of subpopulations that differ, when tested indepen- dently, in degree of tumorigenicity suggests that within the parent tumor there are interactive mech- anisms among the subpopulations that regulate growth and dissemination. In addition to differences among tumor cell subpopulations, nonmalignant tissue within neo- plasms may also be heterogeneous. Recent results from our laboratory suggest that normal cell heterogeneity may be related to tumor cell hetero- geneity. Infiltrating lymphocytes have been isolated from solid mammary tumors produced by a series of cell lines which were originally derived from a single strain BALB/cfC3H mouse mammary tumor. Not only did the percentage of lymphocytes iso- lated vary among the lines, but the type of lympho- cyte also differed. In particular, the relative pro- portion of T cells belonging to the helper class versus those identified as members of the killer- suppressor class was characteristic for different tumor subpopulations (50). Tumor-infiltrating cells independently isolated from two different sub- populations growing on the same mice belonged to the T cell type characteristic for the individual subpopulations. Thus, the type of T lymphocyte response was a characteristic of the tumor, not the host, and was associated with specific tumor cell subpopulations. Whether tumor cell heterogeneity similarly influences other host components of tumors remains to be determined. The wide range of phenotypic differences among tumor cell subpopulations suggests the existence of genotypic differences. Indeed, numerous investi- gators have described karyotypic differences (22, 30, 37, 51-55), as well as the presence of different marker chromosomes in different tumor subpopu- lations (37, 56). Using murine mammary tumor virus (MuMTV) DNA as a probe, cellular hetero- geneity in the location and copy number of a specific gene has been demonstrated in strain GR mouse mammary tumors (33, 35). This is in ac- cordance with the heterogeneity in expression of MuMTV-coded antigens within individual mam- mary tumors (34). Studies on the differential re-
`
`IMMUNOGEN 2163, pg. 3
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`sponse of BALB/cfCaH mammary tumor subpopu- lations to inducers of MuMTV gene expression suggest that differences in regulation of MuMTV genes also correlate with tumor subpopulation heterogeneity (57). Heterogeneity of human cancers There is considerable indirect, and increasing direct, evidence that human cancers, like their animal counterparts, are composed of heterogeneous sub- populations. Heterogeneity in histological pattern may be seen in multiple samples of breast carci- noma (58, 59) and in small cell anaplastic carci- noma of the lung (60). Histological and ultra- structural heterogeneity of tumor cells from bronchial carcinoid has been described (61). Intra- tumor heterogeneity in tumor cell DNA content has been observed in colon carcinoma (62) and small cell carcinoma of the lung (63). Expression of tumor-associated antigens has been shown to be nonuniform among cells from single neoplasms, such as osteosarcoma (64), and pancreatic (65), and breast carcinoma (66). Other markers of tumor cell differentiation have likewise been shown to be distributed heterogeneously within tumors, for example, Bz-microglobulin (67) and estrogen recep- tors (68-70) in breast cancer. Tumor cell hetero- geneity for calcitonin has been described in virulent medullary carcinoma (71). This is especially in- teresting in that it was shown that heterogeneity for calcitonin staining was seen in medullary carcino- mas with a high likelihood of metastatic spread, whereas uniform staining was seen in tumors with a small chance of recurrence. Additional evidence that human cancers contain tumor cell subpopulations comes from comparison between primary tumors and metastases. Here again one may see divergence in histological type (59). Differences in levels of histaminase and L-DOPA decarboxylase have been reported between primary small cell carcinoma of the lung and hepatic metastases (72). Different hepatic metastases from the same patient likewise vary in L-DOPA decarboxylase activity. Differences in sensitivity in vitro to antineoplastic drugs between cells from primary ovarian carcinomas and their metastases have also been seen (73). Furthermore, estrogen receptor content can vary between primary breast cancers and their metastases and among multiple metastases of the same patient (70). As with animal tumors, formal proof of the existence of tumor subpopulations requires their isolation and characterization. This has now been accomplished with a growing list of human tumors. Tumor lines that differ in drug sensitivity (74, 75), antigenicity (76, 77), or tumorigenicity in nude mice (78) have been isolated from single melanomas, both from primary lesions (74, 75, 78) and multiple metastases of the same patient (77, 78). Tumor subpopulations have also been isolated from prim- ary human colon carcinomas (79, 80). Certain of these subpopulations differ in karyotype (80), in vitro growth properties (79, 80) tumorigenicity (79) and histology of tumors in nude mice (79-80). Similar isolations of tumor subpopulations have also been reported for lung (81), ovarian (82), and pancreatic 183) cancer. Isolations of tumor subpopulations from human cancers have frequently been accomplished using cell cultures which had been maintained and pas- saged in vitro for fairly long periods of time prior to cloning. Only rarely have the subpopulations been obtained directly from the patient (77, 82). This raises the possibility that the production of hetero- geneous variants is a consequence of the in vitro environment and occurs sometime after removal of the tumor from the patient. In this regard the elegant study of Shapiro et al. (84) needs emphasis. These "investigators karyotyped tumor cells from fresh samples of human gliomas within six to 72 h after surgery. An array of unique karyotypes was found in each tumor. Simultaneously, dissociated tumor cells were cloned by dilution plating and the clones were karyotyped. By matching karyotypes of the clones with those in the fresh sample, it was possible to show that the clones were present at the time of resection. Each of eight gliomas was found in this way to have from three to 21 subpopulations - a minimal estimate since different subpopulations can have similar karyotypes. Different clones from the same tumor differed in morphology and growth kinetics. Antigenic heterogeneity has also been
`
`IMMUNOGEN 2163, pg. 4
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`reported in clones derived from a single human glioma (84). The work of Shapiro et al. (81), as well as work done with animal tumors (18, 30), suggests that heterogeneity is not induced by culture in vitro. On the other hand, it is often assumed that long-term cell lines are not heterogeneous, or minimally so, due to selection in vitro. That this is not so is shown by the ability of investigators to isolate subpopula- tions from lines such as murine L1210 (40, 41) and human tumor lines, including HT29 colon carci- noma (86), MOLT-3 malignant T-lymphoblasts (87), MCF-7 breast carcinoma (88), and other established lines (76, 80, 81, 83, 84). Origin of tumor heterogeneity A point of confusion in understanding tumor heterogeneity is reconciling its existence with the large body of evidence pointing to a single cell origin for many, if not most, neoplasms (89). Strong as this evidence is, it must be remembered that it is not universal. Some tumors, such as 'venereal' warts in man (90) and fibrosarcoma induced by relatively high doses (91) of methylcholanthrene in mice have been shown to arise from more than one clone (92). Furthermore, some human cancers are characterized by numerous loci of neoplastic change. Multiple lesions of hyperplastic, in situ, and intraductal neoplasia can often be demonstrated in breasts of women presenting with invasive breast carcinoma (59). Thus, a developing malignancy could incorporate elements from other lesions, and hence become 'heterogeneous'. Even if all cancers were truly of single cell origin, the opportunity for heterogeneity to develop occurs as soon as that single cell divides. As will be discussed, both structural and regulatory altera- tions in genetic function may contribute to cellular variation. After all, most multicellular organisms begin as single cells. Even among cells from grossly homogeneous tissues, biochemical and functional heterogeneity is apparent. The heterogeneity among hemopoietic cells, ultimately derived from clonal stem cells (93), and the multiple cell types within the lymphocyte family (94) are obvious examples. Even quite similar cells, such as thymocytes (87) or mammary epithelial cells (95) are heterogeneous in regard to enzymatic activity or antigen expression. Griffin et al. (96) have demonstrated that normal cells can be cloned into heterogeneous subpopula- tions; different clones of genital skin fibroblasts display a wide range of activity of 5~-reductase, the enzyme that catalyzes the conversion of testoste- rone to dihydrotestosterone. If normal tissues exhibit cellular heterogeneity, it is not surprising that minimally transformed or preneoplastic tissues would do so also. Hetero- geneity in expression of a battery of marker en- zymes within loci of hyperplastic, preneoplastic hepatocytes has been demonstrated at the earliest time of recognition of such ~esions (97). Intranodule heterogeneity in expression of MuMTV antigens was seen in mammary hyperplastic alveolar nodules of MuMTV-infected mice (98). Similarly, chromo- somal analysis of tumors produced by subcutane- ous implantation of C3H/10T~ cells attached to plastic suggests that they arose from minor sub- populations within the original culture (99), indi- cating a heterogeneity within that line in regard to induction of tumorigenicity. That such hetero- geneity can have a genetic basis was shown for susceptibility to ultraviolet light-induced transfor- mation by cloning differentially susceptible variants from BAEB/3T3 cells (100). Thus, cellular hetero- geneity is present before, as well as after tumor production, and is itself a factor in tumorigenesis. Clearly such heterogeneity is not unique to cancers, and tumor heterogeneity does not necessarily re- quire any special explanation. Numerous mechanisms have been proposed for the production of diverse subpopulations within a developing tumor. The most pervasive ideas are those of Nowell (55) who theorized that con- comitant with the initiation of neoplasia within a single cell is the acquisition of genetic instability beyond that seen in normal cells and not due only to loss of growth restraints. Nowell cited studies showing a higher frequency of genetic errors in neoplastic than in normal cells and futher suggested that genetic instability becomes greater as a neo- plasm evolves. Direct evidence for this latter hypo- thesis has recently been presented by Cifone and
`
`IMMUNOGEN 2163, pg. 5
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`10 Fidler (101) who showed that the rate of spon- taneous mutation is higher in fibrosarcoma cells of a metastatic clone than in similar cells of non- metastatic subpopulations. The type of genetic errors in neoplastic cells could include mutations in structural genes, mutations in regulatory genes, major chromosomal rearrangements and losses, gene amplification, and subtle rearrangements in specific gene positions resulting in alterations in gene regulation (55, 102). Evidence for the role of genomic rearrangements in contributing to variant production during tumor development (and being related to the multiple steps observed in tumori- genesis) has recently been described by Smith and Sager (103). These rearrangements can be non- random, structured alterations which reproduce precisely from experiment to experiment (102). Nowell also indicated a special role for viral onco- genes in this process, suggesting that variability in number and place of insertion sites could result in position effects on gene regulation. Subpopulations of cells with such differences in viral gene integra- tion have been described in mouse leukemias and mammary tumors (35, 104) and, in the latter case, related to specific clones within a single tumor (35). Other mechanisms for the generation of tumor heterogeneity exist which do not necessarily require structural alterations in the genome. Pierce (29) was one of the first investigators to suggest that neo- plastic stem cells could give rise to variants through a process resembling normal tissue differentiation. Single cells isolated from a murine teratocarcinoma differentiated in vivo into a wide variety of tissues, representing all three germ layers. The progeny of the malignant stem cells were nonmalignant. Pierce stated that teratocarcinoma was a 'caricature of embryogenesis'. That teratocarcinoma is not a special case is shown by similar results with tumors of diverse origin, including a chemically induced rat squam- ous cell carcinoma (29), two types of chemically induced rat mammary adenocarcinomas (105, 106), a virus-induced mouse mammary adenocarcinoma (37), a chemically induced rat neurotumor (107), and the MOPC-315 murine myeloma line (108). Some interesting differences among these various systems are apparent. The teratocarcinoma and squamous cell carcinoma stem cells differentiate to benign cells, whereas the others give rise to neo- plastic variants. The rat mammary carcinoma tumors show a differentiation of epithelial-like cells to spindle-shaped, fibroblast-appearing cells, whereas the mouse mammary carcinoma, neuro- tumor, and myeloma have a broader potential, giving rise to a spectrum of variants. Some of the variants produced by the mouse mammary tumor are also variant producers. Interestingly, although this mammary tumor produces variants at high frequency only in vivo, the variant-producing cells produced by it can produce further variants in vitro. The rat mammary tumor lines and the neuro- tumor produce variants in vitro, although the frequency of variant production has been shown to vary from clone to clone in at least one of the rat mammary tumors (106). Taken in the aggregate these data suggest that the degree of differentiation potential of a neoplastic stem cell is a reflection of the potential of its normal counterpart. The fre- quency of differentiation, however, may reflect additional genetic and environmental factors. The role of environmental versus genetic factors in the generation of tumor heterogeneity is a com- plex problem. Several investigators have assumed that the high frequency of variant production argues against somatic mutation as a primary mechanism (106, 109). However, quantitation of variant production, which can involve many pheno- typic changes and, undoubtedly, numerous pleio- tropic effects, is not as straight forward as quanti- tation of mutation frequency. Expectation of mu- tation rates in mammalian cells are, in the main, based on experiments with normal cells. As already mentioned, mutation frequency in tumorigenic and malignant cells can be considerably higher (55, 101). Furthermore, it is usually assumed that the ~environment" in which variant production is oc- curing is free of mutagens. That this may not be the case is suggested by recent experiments performed by Scott Loveless in our laboratory (110). Starting with an observation that human monocytes can increase the mutation rate of Salmonella o~phi- murium in the Ames assay (111), Loveless isolated macrophages from a series of mouse mammary tumors and similarly assayed their mutagenic activ-
`
`IMMUNOGEN 2163, pg. 6
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`ity. Macrophages isolated from tumors capable of spontaneous metastasis increased the mutation rate over ten times above background. Nonmetastatic tumors were less mutagenic. Results of Weitzman and Stossel (112) indicate that the mutagenic activity of phagocytes is related to their produc- tion of oxygen radicals. Whatever the mechanism, it would seem that the conjunction of host in- filtrating cells capable of inducing mutation with tumor cells known to be genetically unstable might result in mutation rates considerably greater than what has been thought likely. Reports on the necessity, in some systems, of in vivo passage to induce variant production may indicate a role for tumor-associated host cells in the generation of heterogeneity (37, 109, 113). Another role for normal host cells in the genera- tion of tumor variants may be through the process of somatic cell hybridization (59). DeBaetselier et al. (114) have shown that hybridization of non- metastasizing murine plasmacytoma cells with normal spleen B lymphocytes results in the pro- duction of variants capable of spontaneous meta- stasis. Furthermore, these variants exhibit distinct organ specific patterns of spread, growing pre- ferentially in liver or spleen. Other investigators have shown that fusion between tumor and host cells can occur in vivo (115-117). One can imagine that tumor cell variants could be generated by such an event followed by genomic rearrangements, unequal distribution of chromosomes, or chromo- some loss at division. This sort of mechanism may also be responsible for the observations of 'carcino- genic tumors', in which transplantation of tumor cells results in tumor formation by host cells (118- 120). A most interesting example of this phenom- enon is that of Kerbel and associates ( 119, 120) who showed that injection of any of five, independent strain A mouse tumors into DBA/2 mice results in the production of DBA tumors, at a 100% fre- quency, after a very short time. Interestingly, the new DBA tumors are very similar to each other and are always highly metastatic. Clearly, however, mechanisms other than somatic cell fusion may be responsible for these observations. This brief review of the origin and mechanisms of tumor heterogeneity reveals two major points: 11 heterogeneity is not a property exclusive to tumors, but is seen in normal tissues as well, and the potential mechanisms for generation of tumor heterogeneity are many and interrelated. Are any of the mechanisms unique to tumors, or are the differences in the origin of tumor versus normal tissue heterogeneity matters of differences in fre- quency or control? Somatic cell fusion can occur between normal cells in vivo (117). Furthermore, recent discoveries in molecular genetics, such as 'jumping genes' and generalized RNA to DNA pathways, point to a flexibility in the normal genome that was previously thought impossible. The possibility, however, that the generation of tumor heterogeneity is a reflection of 'normal' processes should be remembered when considering methods to limit it. Stability of tumor cell subpopulations In view of the above discussion on the origins of tumor cell subpopulations, it may seem that a consideration of subpopulation stability is unneces- sary. Many investigators, however, seem to feel that the processes generating diversity somehow cease at the moment they have obtained a cell subpopula- tion. This expectation is the source of experimental frustration, scientific conflict, and intellectual error. For example, the continuing capacity of tumor cells to generate diversity is not appreciated by investi- gators who think that cells from metastatic foci should be more metastatic or less heterogeneous than their parent tumors (121). Although this can be so, particularly for parent tumors which are not already metastasizing at a very high rate (122), cells in individual metastases may also be heterogeneous for all the reasons discussed above and for some to be described below. Tumor heterogeneity is a dynamic process! Suffice it to say that our experience (47), as well as that of others (123-125), shows that individual subpopulations and clones thereof are heteroge- neous in their stability. Furthermore, normal cell clones can be similarly unstable (96). Tumor sub- population change~ can be sudden (47, 124) or gradual (47) and to a more (47, 123-125) or less (47,
`
`IMMUNOGEN 2163, pg. 7
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`12 124) malignant phenotype. Changes can occur after in vitro (47, 123, 124) or in vivo (113, 125) passage. The degree of stability seems to vary with the phenotype; in our experience, at least, ability to metastasize is one of the most labile characteristics. Whether this is due to the multifactorial require- ments of the metastatic process, where any of many changes could affect the outcome, or whether it reflects differences in the mechanisms responsible for generation of variation in different phenotypes, requires further study. A most interesting finding is that ofPoste et al. (126) who showed that instability in the metastatic phenotype of B16 melanoma cells became evident after cloning, whereas individual clones growing together as mixed cell populations retained their characteristic degree of metastatic ability. This observation is one of a series showing that tumor cell interactions can influence sub- population behavior. Tumor cell subpopulation interactions Demonstration of tumor cell heterogeneity focuses on differences among multiple cell subpopulations. These differences are shown most convincingly when the subpopulations are grown and compared in isolation from each other. Tumor cells and cell subpopulations do not, however, exist indepen- dently of each other, but rather as parts of mixed cell populations. Cancer development and growth are 'group phenomena'. Cellular interactions can affect the frequency of initiation, as well as growth into overt cancers (127). Not only can normal and tumor cells influence each other's growth (128- 134), there are well-described interactions between embryonic cells (127) and between normal adult cells (135). Thus, the existence of interactions be- tween tumor cell subpopulations is not unexpected. Early investigators of tumor heterogeneity, most prominently Hauschka (136), commented that the growth of isolated sublines of a tumor was some- times faster than the growth of the parent tumor from which they came and suggested the existence of mechanisms within the parent tumor to control growth of the more vigorous subpopulations. Simi- lar observations have been made by more recent workers (21, 137). Makino showed that two dif- ferent sarcomas could control each other's growth (138), and Cheshire noted that the growth rate of single, spontaneously arising C3H mouse mammary tumors was generally faster than were the rates of the first tumors appearing on mice which had developed multiple tumors (139). Our laboratory has described a number of s