Biology: Information Needs and Data Gaps
Mammary tissue undergoes a complex process of development during embryonic life. Studies in the mouse and rat have indicated that normal development and differentiation depend on interactions among three cell types (epithelial, myoepithelial, and stromal) and hormones reaching these cells through the bloodstream.
Human mammary tissue undergoes a major developmental change around puberty, when ovarian steroid hormone levels increase and menstrual cycles begin. The development process occurs in response to, and is critically dependent on, the complex interaction of a large number of systemic hormones as well as local regulatory factors. Changing levels of ovarian steroids and anterior pituitary hormones during the menstrual cycle alter all hormone target tissues in the body. Breast tissue shows differences in mitotic index, estrogen and progesterone receptor levels, and various intracellular metabolites during the follicular and luteal phases that follow the cyclic changes in estrogen and progesterone levels. Pregnancy, with its increasing steroid and lactogenic hormone secretion from the placenta, leads to increased mammary ductile development. After pregnancy, the hormonal milieu of lactation leads to synthesis and secretion of milk from epithelial cells. Thus, throughout ontogeny, mammary cells are subject to changing concentrations of pituitary, ovarian, and placental hormones that interact to modulate cellular differentiation and function. The synthetic hormones contained in oral contraceptives simulate the natural steroid hormones secreted by the ovaries and placenta.
This chapter—which raises more questions than it answers—defines an agenda for the basic research in biology that will be required to fully understand the relationship between oral contraceptives and breast cancer.
SIGNAL COMPLEXITY IN BREAST REGULATION
The complexity of regulation of breast physiology may exceed that of any other hormonal target tissue (Table 3-1). The effects of both classical hormones and a variety of growth hormones on growth and function of breast tissue and cells have been extensively studied. Classical hormones cause growth, differentiation, and milk synthesis and secretion, and include hormones from the anterior pituitary (prolactin, growth hormone), ovary (estrogen, progesterone, relaxin), placenta (placental lactogen[s], growth hormone), adrenal cortex, thyroid, and pancreas. These hormones may act directly on specific pathways of metabolism and cell division within the breast, or they may work indirectly through locally secreted growth factors (see below), which, in turn, regulate cell function. Additionally, the hormones and growth factors may interact with and modify each other's effects, changing secretion rates and actions (e.g., estrogens increasing prolactin secretion and tissue receptors for progesterone).
Growth hormones or factors are peptides secreted by most, if not all, cells. They are essential for growth and differentiation of many tissues. Unlike the classical hormones, these substances were discovered only after it was noted that many cells in culture in chemically defined media were incapable of growth, cell division, or differentiation without the addition of serum from animals. Growth factors can act as local growth regulators and can inhibit or stimulate mitogenesis of epithelial or stromal cells; they can also stimulate angiogenesis and influence cell transformation and immortalization.
In culture, normal or malignant breast tissue can secrete transforming growth factors (TGF-α) or (TGF-β), epidermal growth factor (EGF), insulin-like growth factors (IGF-I and IGF-II), platelet-derived growth factor (PDGF), and fibroblast growth factor (FGF) under various conditions. These factors can act as paracrine and autocrine growth regulators. TGF-β has been shown to act normally as an inhibitor of mammary gland growth, whereas EGF acts as a stimulator of mammary growth (Lippman and Dickson, 1989).
Recent work suggests that hormones (e.g., estrogen) may in some cases act by changing growth factor secretion locally, and perhaps by changing receptors for these factors as well. The observation that estrogen receptors in stromal tissue may be responsible for differen-
TABLE 3-1 Growth Factors and Hormones Potentially Involved in Breast Development and Differentiation
Transforming growth factor α
Transforming growth factor β
Prolactin and other lactogens
Insulin-like growth factor I
Insulin-like growth factor II
Fibroblast growth factor
Rochefort's 52-K protein
Epidermal growth factor
Platelet-derived growth factor
SOURCE: Adapted from K. S. McCarty, Jr., "Proliferative Stimuli inthe Normal Breast: Estrogen or Progestins?," Human Pathology 20(1989):1137-1138.
tiation of epithelial tissue by means of some paracrine influence speaks to the complexity of the regulation of breast tissue.
Another new paradigm is the importance of extracellular matrix (ECM) proteins in influencing cell-cell interactions and in defining “ tissues.” The extracellular matrix is the conglomeration of substances on which cells sit and which ties them together as a “tissue.” The proteins of the matrix are secreted locally and provide yet another means of cell-cell communication. The production of these proteins is regulated by the cells that produce them; in turn, the proteins regulate the function and differentiation of the cells that constitute the tissue. In the absence of extracellular matrix proteins, many cells, including cells of mammary tissue, do not show normal morphology and organization. Extracellular matrix proteins include collagens, laminin, fibronectin, glycosaminoglycans, and probably many others. They influence not only mammary tissue morphology in culture but also function, increasing greatly the synthesis of proteins responsible for milk secretion, including a primary constituent of milk, casein. In a number of tissues, including breast epithelium, it has been observed in vitro that normal epithelial function, such as secretion or differentiation, does not take place unless a normal extracellular matrix is present.
Many hormone effects on target cells are mediated by actions of ECM protein secretions, or by secretion of growth factors locally. Normal breast morphology and function can only be understood by factoring these local signals into our total understanding. Perhaps the apparent heterogeneity of breast cancer etiology (i.e., the wide range of cell types and receptors involved) may be understandable in terms of these local signals.
Until recently, the question before this committee as to whether oral contraceptives influence the risk of breast cancer could be addressed only by human epidemiological and animal studies. Human epidemiological studies (see Chapter 2) are and will remain the central source of information concerning the link, if any, between oral contraceptives and breast cancer. Rapid advances in molecular endocrinology and in the biochemistry and biology of growth factors and steroid and peptide hormones now permit scientists to gain some insight into the biological links between breast neoplasia and oral contraceptives. This knowledge should provide a rationale and hence support for epidemiological observations, and also allow the construction of biological hypotheses that can then be subjected to testing through epidemiological studies.
Numerous positive animal studies have focused attention on the estrogens in oral contraceptives as possibly being involved in breast cancer. Although much in vivo animal evidence shows that estrogens have a causative or permissive effect on mammary tumors, there is no comparable information for humans. Yet indirect evidence, mostly in the form of epidemiological data, exists for a relationship between human breast cancer and ovarian steroids, particularly estrogens. The disease is far more prevalent in women than in men, with a ratio of 100 to 1. Ovariectomy in early reproductive years decreases risk. Moreover, the relationship between the estrogen receptor in the tumor and its response to hormonal treatment implies a hormonal link to breast cancer.
The putative association between the estrogens in oral contraceptives and human breast cancer could be the result of quantitative, temporal, or qualitative factors. The quantitative relationship appears to be the least likely because the effect of the oral contraceptives is to depress the endogenous ovarian production of the natural hormone estradiol. The low-dose estrogen pill (see Appendix C) contains less estrogen than the average follicular-stage daily production rate of estradiol (about 80 micrograms per day [μg/day] during the early follicular phase) and much less than the preovulatory secretion in the human and the secondary luteal rise of the hormone, both of which are extinguished by the exogenous hormone. Pharmacokinetic studies with oral contraceptives containing 30-35 μg of estrogen suggest that the average serum concentration of ethinyl estradiol is similar to midfollicular-phase levels of estradiol and less than peak preovulatory levels. Therefore, it might be inferred that the net effect of low-dose and triphasic oral contraceptives is to decrease estrogen
“load.” However, there is marked inter- and intraindividual variability in pharmacokinetic profiles following oral contraceptive ingestion.
Oral contraceptives disrupt the cyclical nature of estrogen secretion of a normal ovulatory cycle and replace it with a relatively constant level over 20 days. Whether this temporal change might be responsible for differential cellular changes in breast cells is not known. However, pregnancy, with its noncyclical, high estrogen levels for as long as six to nine months, has not been associated with increased risk —but rather with decreased risk. Ovariectomy, which results in a constant, low background level of estrogens derived from peripheral aromatization, is also associated with decreased risk for breast cancer.
The qualitative aspects of a possible link between the synthetic estrogen in oral contraceptives and breast cancer derive from structural differences between the synthetic and the endogenous hormone. Although present knowledge indicates that the interaction of 17α-ethinyl estradiol and its prodrug, mestranol, with the estrogen receptor parallels that of estradiol in the qualitative sense, there are major differences in the metabolism of the natural hormone and ethinyl estradiol that may impinge on their biological properties. The metabolism of estradiol in the human is largely oxidative in nature. Estradiol is oxidized to estrone in a reversible reaction, but the equilibrium is greatly in favor of estrone. This metabolic transformation cannot occur in the 17α-ethinyl estrogens. Estrone is then irreversibly transformed by two largely competitive hydroxylations. Hydroxylation at C-2 leads to catechol estrogens which are ineffective as estrogen agonists, whereas hydroxylation at 16α leads to 16α-hydroxyestrone and estriol, which are potent estrogens with some unusual biological features. The 17α-ethinyl estrogens in oral contraceptives are metabolized primarily by the 2-hydroxylative route with minimal 16α hydroxylation.
The difference in metabolic pathways between the natural and the synthetic estrogens could have important biological consequences and may be relevant to the occurrence of breast cancer. Some evidence suggests that a product of 16α hydroxylation interacts in a covalent fashion with the estrogen receptor, which might lead to genomic consequences different from those elicited by the conventional reversible binding of the hormone with its receptor (Swaneck and Fishman, 1988). By this criterion, one would predict that the suppression of endogenous estradiol secretion and its replacement by ethinyl estrogens, which are principally 2-hydroxylated, would have a beneficial effect on the risk for breast cancer.
Relationships Among Cell Types
There are four critical questions about the way cell types in the breast interact with each other and the possible modulation by oral contraceptives of this interaction in relation to the development of breast cancer:
To what extent is steroid hormone function in mammary tissue mediated by synthesis and action of extracellular matrix proteins and growth factors (as opposed to a direct effect on specific cells)?
Are these putative local regulators responsible for heterogeneity of cell morphology, growth potential, and destiny?
Can estrogen-dependent local growth factor and ECM protein secretion lead to estrogen-independent secretion that results in local autonomy and overgrowth (a clinical observation often made)?
How relevant are ECM proteins and growth factors to the etiology of breast cancer?
A major theme of clinical observations of mammary cancer is heterogeneity —of cell types, steroid receptor distribution, oncogene expression, in vitro and in vivo responsiveness of antihormones, and prognosis. Additionally, animal studies indicate that different hormones—growth hormone and prolactin, in addition to estrogens— may be of importance in mammary cancer in different species.
Yet another dimension adding complexity in breast physiology is the time domain: ontogenetic, cyclic, pregnancy, lactation, and aging. For example, the nature of stem cells at all of these stages of breast development should be explored. This type of research may provide some clues about the nature of stem cells that give rise to breast cancer in humans, and may be important in the use of experimental models for human breast cancer. These long time periods clearly are characterized by changing serum hormone levels and changing differentiation and growth of mammary tissue per se. The local interactions outlined above may be different in different time domains, thus further increasing complexity.
The current concept of the action of any substance that serves as a “signal” to a cell, rather than as a substrate or precursor, is that it must be bound to a specific receptor in the cell membrane, cytoplasm, or nucleus. Three kinds of signals mediated by receptors are relevant to breast physiology: endocrine (hormonal), paracrine (local substance released by neighboring cells), and autocrine (substances secreted by the same cell on which they act).
Breast cells have receptors for estradiol, progesterone, prolactin, cortisol, oxytocin, transforming growth factor β, aldosterone, insulin, epidermal growth factor, placental lactogen, thyroxine, relaxin, growth hormone, calcitonin, insulin-like growth factors I and II, and fibro-blast growth factor. This extraordinary diversity of cell receptors for circulating hormones, as well as a large variety of growth factors that may be produced locally, means that investigation of the etiology of transformation in these cells is not a simple matter. In turn, ascribing a possible role of oral contraceptives in such transformation will not be readily proven.
During the process of development, there are critical periods when a given target tissue can be altered irreversibly by a signal—for example, by a hormone. Once the critical time period has passed the target tissue can no longer be affected. An example of such a critical period is the three- to seven-day period during development of the rodent brain when masculinizing hormones can permanently alter neuronal numbers in the brain. If such masculinization fails to occur during this time, it cannot ever take place. A similar critical period occurs in the human: during myelinization of the brain postnatally, thyroid hormones are critical. If they are not present before about four to six months, the brain is irreversibly stunted and mental retardation results, which cannot be reversed by thyroid hormones. Evidence also indicates that female fetuses of mothers treated with diethylstilbestrol during pregnancy have a higher incidence of vaginal cancer many years later.
Are there any similar critical time periods (e.g., peripubertal or postpubertal) in mammary tissue development when gonadal or pituitary hormones can set some process into motion that leads inexorably to altered development and irreversible differentiation along a given path? In the human breast, certain aspects of development are not initiated until puberty. Could exogenous gonadal steroids administered during the peripubertal period set some event in train that may not be manifested for a number of years, and that would not be initiated by the same steroids administered at less sensitive times to older women?
Effects of Pregnancy and Lactation
The putative protective effects of pregnancy and lactation on breast cancer add several important questions to the research agenda: What
is there about pregnancy and lactation that confers protection? Can this be seen in an animal model? Do placental hormones contribute to protection? Beyond these general questions, it has been observed that there is a cyclic difference in both mitogenesis and steroid receptors in the breast when examined using thymidine incorporation and estrogen or progesterone receptor-binding assays, respectively (Anderson et al., 1990). These data suggest additional research questions: What causes increased mitogenesis during the luteal phase? If it is progesterone, what is the mechanism? Is progesterone responsiveness permanently altered by pregnancy and lactation?
Pathological Breast Tissue
Information continues to be needed on both normal and neoplastic tissue. Although the physiology of the breast is relatively well understood, there is still no useful understanding of the transformation from benign to malignant breast epithelium. Pathologists have used biopsy material to identify a small subgroup of women at increased risk, but there is no reason to think that the majority of breast cancers arise from this type of background.
Efforts to determine the effect of oral contraceptives on the risk of developing breast cancer should probably focus on the effect of these agents on normal breast tissue. Cystic conditions—in some instances, gross cystic disease—are common during the reproductive years, and although such tissue might be considered abnormal, recent studies do not reveal an increase in breast cancer in such patients. Further studies of cyst fluid should be conducted because cyst fluid constitutes the best available reflection of breast epithelium. A number of mutagens as well as specific growth factors have been described in cyst fluid. Much of this work focuses on early detection, but it can also be used to investigate hormones that are thought to cause epithelial proliferation in the breast. Both the administration and withdrawal of hormones could be investigated because most women with gross cystic disease form additional cysts. More detailed characterization of breast cyst fluid would be beneficial; once this has been done, the effect of oral contraceptives on a wide variety of biological markers (e.g., gross cystic disease fluid protein) could be determined.
Cysts generally appear during the 30s and 40s; thus, there is also a need for tissue that will provide epithelium from younger women. The best source of such tissue would be fibroadenomas, which are most common between the ages of 15 and 25. These tumors are commonplace but have received little study because they are not associated with cancer. Despite the lack of association, however, efforts
should be initiated to identify biological markers that might be affected by hormone administration. Because each patient would only be studied once—second fibroadenomas are far less common than second cysts—methods must be developed for repeated epithelial sampling. Induced nipple secretions have been used for such sampling, but the cytology is extremely variable and may reflect shed rather than viable cells. A better method might be fine-needle aspiration, which can be performed repeatedly with minimal discomfort. Currently, it is used routinely in the diagnosis of neoplasms; however, it can also be used to aspirate normal breast epithelial cells. Epithelium sampled by this method has been examined histologically, but biological studies (with and without cultures) are also possible, provided the appropriate markers (either functional or cytological) can be developed, and would cast additional light on the effect of hormones on normal breast epithelium.
The important question concerning oral contraceptives is their effect on normal breast epithelium, as breast cancer has already been studied intensively. Consequently, the committee emphasizes this necessary concentration on the largest gap in our present knowledge: the transformation of benign to malignant epithelium.
Another area that requires new information relates to family history and efforts to identify oral contraceptive users who might be at increased risk. The study of breast cancer should consider tissue-typing techniques that have made it possible, for example, to identify children who will be affected in families with juvenile diabetes. The search for linkage in breast cancer families is an active field, and improved techniques (i.e., analysis of linkage using polymorphic DNA probes) are now available. These methods should make it possible to better define the genetic makeup of the small group of women with a relatively high risk of breast cancer as a result of specific family history of the disease.
Carcinoma of the breast presents a variable histological appearance, but current classification systems place almost 90 percent of all breast cancers in the invasive (or infiltrating) ductal category. Foote and Stewart described invasive lobular carcinoma—the last new histological subgroup—in 1941; a miscellaneous collection of unusual breast cancers has been recognized for at least 50 years. The largest group would be medullary cancers, but colloid (or mucinous), adenoid cystic, papillary, tubular, and even cancers with carcinoid features or squamous and osseous metaplasia have also been noted. Lobular carcinoma in situ and hyperplasia with atypia also fall into this category. Unlike other cancers induced by carcinogens, no new histological type of breast cancer can be tied to the introduction of oral
contraceptives. However, there has been no systematic examination of the rare cancers and use of the pill, and even more important, no study of certain benign lesions associated with increased risk.
A final point relates to the deposition of calcium in the breast. The presence of calcification signals in situ breast cancer on screening mammograms, but only 20 to 30 percent of all breast cancers show microcalcifications. There is no information on the possible role of oral contraceptives in the development of these calcifications.
Recent studies of the molecular biology of oncogenic transformation (i.e., the process by which a normal cell becomes malignant) have identified various mechanisms. Using animal cell model systems, these studies define various genes, or oncogenes, whose dysregulated expression results in transformation. In some cases, abnormally high levels of specific oncogene products cause transformation. These levels can be achieved in the laboratory for study by amplifying the number of gene copies, by mutating the gene so that it does not respond properly to regulatory elements, or by genetically altering the regulatory elements themselves. This third type of oncogene is dominant because only one allele need be affected to produce abnormally high levels of gene product. In other cases, transformation results when a gene product necessary for normal cellular behavior is missing. Generally, both alleles of the gene must be affected (either deleted or mutated) for the functional gene product to be absent; hence, such genes are often termed recessive oncogenes.
If the effects of estrogens and antiestrogens on mitogenesis and various aspects of breast cell intracellular metabolic pathways are mediated by secondary factors like locally secreted growth factors and extracellular matrix protein production, then any autonomous constitutive secretion of these factors could mediate transformation of an initially estrogen-dependent tumor into an estrogen-independent tumor. This conversion might result from the action of an oncogene.
Because different genes can induce oncogene transformation of animal cells, however, there must be more than one mechanism by which a cell can become malignant. Therefore, it is important to determine which, if any, of the known dominant or recessive oncogenes are actually involved in particular human cancers. To answer this question, human tumor biopsies have been characterized for various genetic aberrations.
Because most biopsies are quite small, technological breakthroughs allowing evaluation of small amounts of nucleic acid have only re-
cently made it possible to study human tumors. Two techniques that have been particularly important are the polymerase chain reaction (PCR) and analysis of restriction fragment length polymorphisms (RFLPs). PCR is a method by which any known gene sequence can be specifically amplified so that as few as 10 to 100 copies of a specific sequence can be detected. RFLP analysis allows researchers to distinguish between two alleles of a given gene because of small variations in DNA fragment size that are detected as differences in migration through acrylamide gels after cleavage with site-specific DNAases. One can deduce that a given allele is missing when one fragment is absent from the gel.
Two generalizations are emerging from these molecular characterizations of cancer biopsies that may be relevant to the problem of identifying a possible role for oral contraceptives in the etiology of breast cancer. First, it is clear that the molecular lesions associated with cancers differ depending on the organ of origin. For example, some of the molecular lesions found in colon cancers and certain lung cancers are not detected in breast cancers. This observation provides evidence at the molecular level for the conclusion that different etiologic factors may be responsible for these three common human cancers. Second, among breast cancers there is a great deal of heterogeneity in the types of genetic lesions detected, which suggests that breast cancer may be more than one disease.
Molecular Changes Associated with Human Breast Cancer
Activated ras Mutations
The products of ras oncogenes are functionally and structurally similar to the G-proteins. G-proteins normally act by transmitting proliferative signals initiated by extracellular hormones and growth factors. They bind guanine nucleotides and mediate signal transduction through effectors such as adenylate cyclase. Certain mutations in G-proteins induce autonomous activity by inactivating control regions of the protein; these mutations cause loss of cellular growth capability. ras is the only one of the G-protein family that has been extensively studied in human cancers. There are three ras genes, Kl-ras, Ha-ras and N-ras. Mutations at codons 12, 13, or 61 are known to activate each of these ras genes, resulting in transformation of various cell types (for reviews, see Milburn et al., 1990; Bos, 1989).
The overall incidence of ras activation in human cancer has been estimated at 10 to 15 percent. This figure is much higher, however, for specific solid tumors such as adenocarcinomas of the lung (50
percent) and gastrointestinal tract (40 percent), or acute myeloid leukemias. In contrast, ras activation is rarely seen in either primary or metastatic breast cancer. Thus, it is unlikely that ras activation by gene mutation has any significant role in the initiation of metastatic progression of human breast cancer in vivo. These negative findings are significant in at least two regards. First, they provide evidence of etiologic differences between spontaneous human breast tumors and carcinogen-induced rat models of breast cancer, which have a high incidence of ras mutations. Second, they illustrate the molecular biological differences between breast adenocarcinomas and morphologically similar adenocarcinomas of the lung and gastrointestinal tract, which have a high incidence of ras mutations.
Overexpression of Tyrosine Kinases
The G-proteins are one example of signal transducers; another common mechanism of signal transduction involves kinases and phosphatases. Protein kinases and phosphatases add or remove phosphate either from serine and threonine residues or from tyrosine residues. Phosphorylation affects the enzymatic activity of a variety of proteins. Many growth factor receptors function as tyrosine kinases. They normally become transiently activated by binding a growth factor; mutations in their transmembrane domains, however, can result in constitutive activity.
In breast cancer, there has been no evidence to date of activating mutations in the transmembrane domains of tyrosine kinase growth factor receptors. Yet abnormal cell growth may also result from overexpression of normal growth factor receptors. Three related tyrosine kinase transmembrane proteins are expressed at increased levels in breast cancer: the epidermal growth factor, or EGF, receptor, the receptor-like product of the erbB-2 oncogene, and the product of a recently described gene, erbB-3.
The EGF receptor is a 170-kilodalton (kD) transmembrane glyco-protein, which is found on many epithelial cell types. It also binds transforming growth factor alpha (TGF-α), which results in a growth stimulatory effect similar to that observed with EGF. In patients whose breast cancers are negative for estrogen receptor, a high level of EGF receptors is associated with poor prognosis (Huebner et al., 1988), but this issue is still controversial (Foekens et al., 1989). In breast cancers, the cause of EGF receptor overexpression is unknown; only rarely is overexpression caused by EGF receptor gene amplification.
The erbB-2 gene (also known as neu or her-2) is amplified in ap-
proximately 10 to 30 percent of breast carcinomas and also in adenocarcinomas of salivary gland, stomach, and ovary. erbB-2 is structurally related to the EGF receptor; hence, its gene product is thought to be a growth factor, but its ligand is unknown. Some investigators have shown that erbB-2 amplification and protein overexpression correlate with poor prognosis; others have found no such correlation. The evidence for association of erbB overexpression with poor prognosis is stronger in patients with lymph node metastases than in node-negative patients.
Much less is known about the role in breast cancers of erbB-3, a third member of the erb family. It was identified because it is homologous to the EGF receptor gene and to erbB-2. Sequence analysis of the cDNA predicts that it encodes a 148-kD transmembrane polypeptide. Like the other members of this family, overexpression of erbB-3 occurs in a percentage of breast cancers.
Amplification of int-2 and myc
There are four known int genes, which are defined as the sites of common integration of the mouse mammary tumor virus genome. Their sequences fall into two groups: int-1 and int-2. Although the two groups have no sequence similarity, both are implicated in mouse mammary tumorigenesis and are essential in early embryogenesis.
int-1 related genes are not amplified, translocated, or even expressed in human breast cancer. In contrast, the int-2 gene is amplified in 15 percent of breast cancers. It is unlikely, however, that the int-2 gene itself is important in breast cancer. Whenever gene amplification occurs, the gene that confers a selective advantage is coamplified, together with 100 to 1,000 kilobases of surrounding DNA. The int-2 gene is usually coamplified with certain other genes that have been implicated in human cancer, including hst, bcl-1, and sea. In a minority of breast cancers, bcl-1 is the only gene of the cluster that is amplified; in some cases, none of the three genes is expressed. There is hence some question of which gene is the driving force in the amplicon, and it has been hypothesized that an as yet unknown gene in the region, when amplified, confers a selective growth advantage on breast cancers.
Much less work has been done on the role of c-myc in breast cancer. C-myc is a nuclear protein that may be involved in transcriptional regulation of other genes important for cellular growth control. Amplification of the c-myc gene has been observed in approximately 5 to 30 percent of breast cancers.
Deletions of Genes
Researchers are increasingly aware that mutations resulting in loss of function play an important role in the pathogenesis of human malignancies. Deletion of one allele, measured by loss of heterozygosity for restriction fragment length polymorphism, is thought to unmask mutations in the corresponding normal allele. Thus, it is thought that recessive oncogenes are located in chromosomal regions showing a high incidence of allele loss.
In breast cancer, investigators report nonrandom loss of heterozygosity for a number of chromosomal loci. The frequency of loss of heterozygosity ranged from approximately 50 percent for chromosome 17p, to 20 to 30 percent for regions on chromosomes 1q, 3p, 11p, 13q, 17q, and 18q. Moreover, it has been suggested that deletions at several loci tend to occur within the same tumors.
At some of the deleted chromosomal loci, researchers have tentatively identified the recessive oncogene involved. The target for loss on chromosome 13q is thought to be the Rb gene, which was originally isolated as the recessive oncogene that causes retinoblastoma. The Rb gene encodes a protein that is thought to be involved in cell cycle regulation. The region deleted on chromosome 17p includes the p53 gene, which encodes a protein that binds to DNA as a homodimer. In vivo, it may function by binding to and thereby inactivating the suppressor gene products.
To address the basic biological questions in the relationship of breast cancer and oral contraceptives, appropriate model systems need to be developed. A recurrent issue in biological research is whether findings in studies of subprimates are predictive of similar findings in humans. Animal models (see Appendix D) have been valuable in studies of breast cancer; in both rodents and dogs, investigators have shown that estrogens can increase the rate of mammary cancer. However, there may be fundamental differences in overall endocrine physiology among various species that preclude direct extrapolation of data from animal models to humans. For example, one pituitary factor, prolactin, is necessary for production of estrogen-induced tumors in mice, whereas a related but different pituitary factor, growth hormone, serves the same function in beagle dogs (see Appendix D). Thus, it is possible that species-specific effects of exogenous steroids such as oral contraceptives will be of primary importance with regard to induction of mammary cancers.
Because many types of physiological studies are impossible in humans, there is a continuing need for animal studies. The conclusions from these studies, however, must be reinforced using human tissue and cells. Consequently, relevant in vitro human systems are essential. There are four types of in vitro human systems currently being used: (1) established tumor cell lines, (2) short-term mammary epithelial cell cultures, (3) organ cultures, and (4) athymic mice. Although tremendous progress has been made using these systems, they are still far from adequate for detecting slight increases in transformation or subtle effects on differentiation. Because it is likely that the effects of many exogenous agents such as oral contraceptives will be small and subtle with respect to cancer induction, further development and refinement of appropriate culture systems are essential. Therefore, the committee recommends that a sustained, long-term basic research effort to develop and refine appropriate culture systems is warranted. Furthermore, the specific steroids contained in oral contraceptives should be studied using these systems. Finally, it is crucially important to refer to whole-animal in vivo studies to verify that in vitro findings have relevance to the whole animal.
Breast Cancer Cell Lines
Breast cancer established cell lines are valuable because they provide a readily available source of proliferating cells with infinite growth potential. In many cases, these lines retain in vivo properties, thereby providing useful substrates for many important physiological studies. For example, much information on the cellular biology of estrogen receptors has been gathered because some breast cancer cell lines retain estrogen receptors in culture. This readily available source of estrogen receptor-positive cells permitted extensive characterization of the mechanisms by which estrogen receptors control gene expression (for a review, see Dickson and Lippman, 1987). Studies using these cell lines have also been directly useful in the clinic because the cell lines support more rapid, detailed characterization of potential estrogen receptor inhibitors, such as tamoxifen.
Although established cell lines derived from breast carcinomas have provided tools for many informative and important studies, the interpretation of this information must take into account the limitations of such cell lines. Normal human mammary cells do not develop into established cell lines. Furthermore, the existing breast cancer cell lines represent only a small subset of breast cancer cells because (1) only occasional cells within a given tumor survive as the
cell line, and (2) only rare breast cancer specimens develop into cell lines.
To develop a cell line, the tumor cells are usually held in a maintenance state for prolonged periods, sometimes months, before a cell population emerges that can grow continuously in culture. During the initial period, most of the carcinoma cells obtained from the malignant tissue proliferate a few times and then undergo a phenomenon, or “crisis, ” in which most of them stop proliferating, deteriorate, and disappear from the culture vessel. The cell line subsequently emerges from a subpopulation of the remaining cells.
Furthermore, only a small number of breast cancer specimens, most of which are derived from metastatic lesions, contain a cell subpopulation capable of proliferating subsequent to crisis. Even among effusion metastases, the most widely studied type of metastatic lesion, less than 10 percent actually develop cell lines. Among primary breast cancers, the frequency of cell line development is much lower, reaching incidences of approximately 1 in 200 cases.
The ability to survive crisis and become an immortalized cell line is not random, either in relation to culture technique or to tumor progression. In one study (Smith et al., 1987), the properties in culture of breast cancer effusion metastases, obtained over approximately two years from the same patient, were examined. Despite repeated attempts with cryopreserved cells, only the last specimen reproducibly exhibited immortality in culture; the first two specimens grew initially but failed to survive crisis. Each specimen was unique in morphology, growth properties, and oncogene aberrations, although karyotypic markers indicated a common origin. The observation that the last effusion metastasis could develop reproducibly into a cell line when prior malignant effusions from the same patient could not suggests that the capacity for infinite life in culture depends on inherent changes in the biological phenotype of the tumor rather than on irreproducible vagaries of cell culture. This study, together with numerous observations that metastatic specimens develop into cell lines much more commonly than primary breast cancers, indicates that the capacity for infinite life in vitro results from a phenotype that is usually acquired by breast cancer cells at a late stage of malignant progression.
Short-term Culture of Normal Mammary Epithelium
Studies on breast cancer cell lines cannot address many of the relevant questions related to the potential link of oral contraceptives and breast cancer. Therefore, it will be critical to develop in vitro model systems that use normal human mammary epithelium.
There are two main sources for culturing normal mammary epithelial cells: breast milk and reduction mammoplasties. Large numbers of epithelial cells can be obtained from milk, particularly during early stages of weaning, when the mammary ducts and alveoli are involuting and being sloughed into the luminal contents (Russo et al., 1975; Kirkland et al., 1979). The epithelial cells derived from milk have the advantage of being free of fibroblast contamination, although they generally grow less well than those isolated from reduction mammoplasties.
To isolate cells from reduction mammoplasties, researchers must separate the cells from massive amounts of connective tissue and fat by enzymatic digestion. The epithelial cells, which are connected by junctional complexes that are insensitive to these enzymes, remain as clumps; the stromal fibroblasts and connective tissues are dissociated to single cells. The epithelial cells can then easily be isolated free of fibroblasts by sedimentation at unit gravity or by filtration through nylon mesh filters. This technique also isolates capillary endothelium, but the endothelial cells do not grow in the media formulations developed for the epithelium.
Mammary epithelial cells proliferate in a variety of culture conditions, including collagen coating of the culture surface, the presence of various growth factors such as epidermal growth factor and cholera toxin, reduced calcium concentrations, or conditioned media from specific cell lines. The cells also grow to some extent in a variety of completely defined media containing high-density lipoproteins, extracellular matrix, various hormones, and growth factors.
A number of criteria have been used to verify the epithelial origin of the cultured cells. Cultured breast cells have a typical cuboidal morphology and form secretory domes and ductlike, three-dimensional ridges at confluence. Ultrastructurally, the cells show junctional complexes and evidence of secretory activity. They also have a distinctive punctate pattern of cell-associated fibronectin and express epithelial membrane antigens as defined by antibodies raised to milkfat globules. The epithelial origin of the mammary epithelial cells has been further verified using antibodies to cytokeratins, the intermediate filament proteins, characteristic of epithelial cells (for reviews, see Osborn and Weber, 1983; Taylor-Papadimitriou and Lane, 1987).
Although the epithelial nature of cultured mammary cells has been clearly established, the type of mammary epithelium being cultured is controversial. In vivo, normal mammary epithelium is organized into ducts and alveoli, and within the ducts, there are both basal (sometimes referred to as myoepithelial) and luminal epithelial cells. The different types of epithelium have not been successfully sepa-
rated prior to culture. After culture, it has been difficult to identify which cells were derived from the different epithelial components. In vivo, basal cells of the mammary ducts are positive for the cell surface marker CALLA (Gusterson et al., 1986), and for cytokeratin 14 (Dairkee et al., 1985), a member of the cytokeratin family. Luminal cells express epithelial antigens derived from milk fat, and other cytokeratin markers (for a review, see Taylor-Papadimitriou et al., 1977). When reduction mammoplasty cells are cultured, all of the cells are positive for cytokeratin 14 and CALLA, suggesting that basal epithelial cells are preferentially grown in culture. Many of the same cells, however, are also positive for a luminal antigen (Dairkee et al., 1986). Therefore, one cannot rule out the possibility that luminal dedifferentiation occurs after culture. An excellent analysis of numerous markers of mammary epithelial differentiation in culture (Petersen and van Deurs, 1988) concluded that the mammary differentiation phenotype is plastic and can be modulated by growth factors and other media components.
Because steroid hormones are intimately involved in mammary gland differentiation, it is critical that future studies concentrate on developing better culture systems for maintaining the differentiated state to allow distinguishing among ductal, alveolar, basal, and luminal epithelium. Only then will it be possible to explore in vitro the critical research questions that relate to the actions of oral contraceptives on normal breast tissue function.
In vitro cell culture has been invaluable for studying the effects of various stimuli directly on synthesis and secretion of many cell products. Usually such cultures contain predominantly one cell type, which enhances their usefulness for examining specificity of stimulus and response. But this simple system has been proved inadequate to examine regulation when two or more cell types communicate with each other and act as a coordinated system. For example, research has shown that, in developing gonadal steroid target tissues, the steroid receptor develops first in the stromal tissue and the steroid acts on the epithelial cell by means of a local signal generated by the stroma. Such cell-cell interactions occur in all organs. Increasingly, in vitro systems using pieces of organs that contain several cell types are being used as more “normal” models to examine putative regulators of synthesis, secretion, and morphology. These breast-tissue organculture systems must be explored more fully than in the past, however, particularly in light of increased evidence of the importance of
locally secreted growth factors and extracellular matrix proteins in determining epithelial cell function. It is possible that local misregulation is actually the mechanism by which transformation of specific cells takes place. One negative aspect of organ culture is that the heterogeneity of cells means that certain effects need to be measured in situ on the cells, in addition to measuring factors released into the medium.
Nude Mouse Model
The athymic nude mouse model system is an immunologically incompetent mouse that does not reject tissue from other species. Normal mammary epithelium can grow in cleared fat pads of nude mice and forms normal ductal structures. Some breast cancer cell lines form tumors readily in these animals. Hence, the nude mouse provides a useful “in vivo/in vitro” system to study factors that contribute to tumor growth. Athymic mice do not secrete much estrogen because their ovaries undergo premature failure and thus exogenous estrogen must be given for tumor growth of human breast cell lines. This exogenous estrogen can act locally —that is, directly on the breast tumor cells. The model has also shown that mutagenesis by estrogen of human breast cancer cells can be potentiated by cotransplants of a pituitary cell line, GH3 cells, suggesting that a pituitary factor (not GH or prolactin) may also be necessary for this process (Dembinski et al., 1985).
The usefulness of the nude mouse is that the animal itself provides the “culture milieu” for its transplanted tissue, and this natural medium is renewed appropriately by the circulatory system. These culture conditions may simulate the intact mammary tissue better than in vitro conditions. Furthermore, given the mouse's low endogenous estrogen levels, this model may be valuable for studying the effects of cyclic administration of various oral contraceptives on normal mammary epithelium.
SUMMARY AND CONCLUSIONS
Although recent studies of mammary gland development, physiology, cell biology, and molecular biology have increased our knowledge considerably, there are no definitive answers as yet that enable us to understand, in all of its aspects, the biological etiology of breast cancer. A number of leads should be followed. Does the 16α-hydroxysteroid metabolic pathway lead to production of a hormone metabolite that binds the estrogen receptor more tightly than normal, leading to abnormal or enhanced gene products? Do exogenous
estrogens regulate this pathway or act differently than endogenous estrogens? Are local growth factors and the newly detected extracellular matrix proteins, such as tenascin, important in the etiology of breast cancer? If not, can they serve at least as markers of the stage or etiology of the disease? Are oncogenes always involved ultimately in breast cancer (regardless of the initial insult)? Will the different oncogenes that are possibly involved act through the same final common pathway to cause transformation? Questions such as these, and many others, need to be vigorously pursued in a multidisciplinary setting to expand our understanding of breast biology in a way that will be relevant to the etiology of breast cancer.
A number of different genetic aberrations have been seen in breast cancers, but in each instance, the lesions are found only in a proportion of all such cancers. Preliminary observations suggest that some lesions may be coordinately expressed; thus, it may be possible to define subsets of breast cancer by their constellations of molecular aberrations. Breast cancers are unusual in that tumors of similar histology and staging may have widely varying clinical courses. This variability and the existence of molecular subsets suggest that breast cancer may be more than one disease, each with differing etiologies. If so, insights into putative etiologic agents—for example, oral contraceptives —may be acquired by determining whether their use is correlated with specific molecular breast cancer subsets.
With respect to oral contraceptive formulations (see Appendix C), four questions emerge as immediate research priorities: Do individual variations in blood levels in ethinyl estradiol and the progestin component of oral contraceptives affect the risk of breast cancer? What are the effects of the progestin component of oral contraceptives in modulating estrogen action? Do the inherent androgenic or antiestrogenic properties of different oral contraceptive formulations affect normal breast tissue response? How will the overall estrogen dominance of the new oral contraceptives affect breast tissue response?
Cancerous breast tissue has been well studied. For the 1990s, intense focus on normal rather than neoplastic epithelium is warranted. Because the significant issue concerning oral contraceptives is their effect on normal breast epithelium, the committee emphasizes the importance of concentrating on the largest gap in present knowledge: the transformation of benign to malignant epithelium.