How breast cancer is diagnosed


In spite of major advances in cell and molecular biology, cancer diagnosis still depends on recognizing large-scale morphologic alterations of cells and tissues by light microscopy. It is expected that the diagnostic appearances of cancer cells and their precursors reflect basic carcinogenic mechanisms. The purpose of this mini-review is to make these diagnostic criteria accessible to the cell/molecular biologists studying breast cancer. To learn breast cancer diagnosis within roughly one hour, we will ignore lobular and stromal lesions and limit ourselves to the following:

This covers about 75% of breast cancer and the best known precursors. The reader may consult one of several comprehensive texts to read about the appearance of other breast lesions(1,2,3). 40 images are provided. Words written in italics are defined in a glossary. At the end of this session the reader should be able to recognize a typical invasive breast cancer, some related benign ductal proliferations, and know general features of glandular proliferations outside the breast. The last 7 images will be a self-assessment test. Supplementary text and answers to the self-assessment are included in the reference section.

The accumulation of diagnostic concepts in breast cancer during the past 150 years has been laborious, depending primarily on correlations of histologic patterns with clinical outcomes. In order to infer how various histologic patterns develop, pathologists have relied on finding lesions with intermediate appearances. Surprisingly there are very few models in pathology to explain why cancer cells grow in these various patterns. Without a framework, breast cancer diagnosis is reduced to pattern recognition, with little interest to the researchers who want to know about carcinogenic mechanisms.

To allow non-trained observers to appreciate diagnostic patterns within a short period of time, it has proved useful to first construct a framework for envisioning why the patterns look the way they do. In this framework, cancer is viewed as an evolutionary process in which increasingly fit clones of ductal cells emerge within the complex environment of the breast(4,5,6). The framework permits evolutionary principles to be applied to understanding the histologic patterns, similar to the way evolutionary principles are successfully applied in community-level ecology research. In essence, the patterns that distinguish the various stages of cancer progression reflect changes in "fitness". Note that in evolution, increasing fitness does not necessarily mean increasing the speed of reproduction (growth rate), nor to increasing longevity (decreased apoptosis). In evolution, increasing fitness is simply the ability to grow better or expand relative to other populations. In development of cancer, the greatest competition is often with the other cousin clones. The assumptions about the nature of the change in fitness of the ductal cells, or the key consequences of the apparent change in fitness that allow the patterns to be recognized will be marked in bold type.(7)

Normal histology and benign ductal proliferations

Figure 1 shows the H & E appearance of normal postpubertal, non-lactating female breast tissue.

Figures 1 and 2

The lobule is composed of numerous small acini that empty into a terminal ductule. The terminal duct empties into progressively larger ducts which in turn connect to the nipple. In lactation the acini produce milk and the duct system conveys the milk to the nipple. The ducts themselves are not greatly altered in appearance during lactation. Figure 2 shows a higher magnification of a duct. The environmental components of diagnostic significance include the ductal cells which form a single row on the inner-most part of the duct, myoepithelial cells external to the ductal cells, the basal lamina, extracellular matrix, periductal fibroblasts, and periductal blood vessels. In this image, the extracellular matrix contains abundant collagen, as manifest by the dense eosinophilia (asterisks). Myoepithelial cells are not always as easy to distinguish from ductal cells. A detailed description of the immunohistochemical features of each of these components, their proposed origin, and biochemical features are described elsewhere(8). Myoepithelial cells are mitotically inactive and appear to arise from terminal differentiation of ductal cells(9,10). BrdU labeling studies show that mitoses can occur anywhere along the duct(10) although there appears to be a zone of relatively higher proliferation in the terminal ductule (short arrow in Figure 1)(11).

The ducts can proliferate in response to unknown signals. Figure 3 shows such a ductal proliferation in a pattern called blunt duct adenosis.

Figures 3 and 4

This pattern is not generally regarded as a direct precursor to ductal cancer, and the benign nature of a proliferation such as this can be reproducibly recognized if one envisions that there is no increase in ductal cells compared to the other cellular components of their environment. All the cellular constituents of the duct microenvironment have increased in number, so there is no change in fitness of the ductal cells themselves. Figure 4 shows that the ductal cells form a single layer in close proximity to the myoepithelial cells and basal lamina compartments which in turn have a normal relationship to the stroma.

Early potential precursors to breast cancer: Proliferative fibrocystic changes, usual ductal hyperplasia, epitheliosis, and intraductal papilloma

There is evidence indicates that these can be precursors to breast cancer, and that the progression is via an evolutionary process. Patients shown in a biopsy to have marked proliferative fibrocystic changes have a slight (2 fold) increased relative risk of subsequent development of invasive ductal carcinoma(2,12). One study using FISH found single or multiple clonal cytogenetic abnormalities in short term cultures in 11 of 15 proliferative ductal lesions(6). Further, these proliferative lesions contain some of the same karyotypic abnormalities as reported in other cases of invasive breast carcinomas(6). The patterns of loss of heterozygosity have been shown to be shared between these proliferative lesions and concomitant intraductal or invasive carcinomas(5). Finally, about 25% of proliferative lesions show grossly aneuploid DNA content(14) suggesting the presence of genetic instability. There is controversy concerning the question of whether proliferative fibrocystic changes are a precursor to cancer, or merely a marker of increased risk (see reference 2 for discussion).

The morphologic features of this group is dazzlingly complex, suggesting many potential mechanisms for increasing "fitness". The complexity also suggests that numerous trophic interactions are able to influence the ductal cells(15). The patterns have been given many different names. Two classes of such early ductal proliferations are often separated. The first group is referred to variously as "proliferative fibrocystic changes", "usual ductal hyperplasia", or "epitheliosis". The hallmark of this group is that the ductal cells increase in number by proliferating as a multilayered epithelium, thus the ductal cells have achieved greater numbers relative to other cells of their environment. Figures 5 through 9 show two examples of usual ductal hyperplasia with multilayering of the ductal cells.

Figures 5 through 9

This group of ductal proliferations are still benign, and the following three features predict their benign nature: 1-Myoepithelial cells are present. Myoepithelial cells are not always easy to tell from ductal cells and it is generally sufficient to identify varied cell types within the duct to diagnose it as benign (albeit proliferative). Although other interpretations are possible, the morphologic evidence is compatible with the following: the ductal cells in these benign lesions still seem to undergo the asymmetric cell divisions that leads to the presence of the terminally differentiated myoepithelial cells. The small pyramidal shaped cells in Figure 9, distinctly different from the ductal cells, are probably myoepithelial. 2-The ductal cells that grow away from the basal lamina zone have a less active cytologic appearance compared to the cells growing in the apparently more choice area next to the basal lamina environment, as if the ductal cells maintain a partial trophic dependence on the basal lamina zone in this group of benign lesions. The term "active cytologic appearance" needs to be described. A basic theorem in cytology is that the appearance of the nucleus reflects the functional state of the cell(16). A cell that is metabolically active has abundant euchromatin (pale areas in the nucleus that presumably reflect transcriptional activity/competence), while a transcriptionally inactive terminally differentiated cell often has dark condensed heterochromatin. In the center of the ducts shown in Figure 6, the cells have inactive heterochromatic nuclei, while at the edge of the duct the cells look more active. The cytoplasm also changes appearance when the cells are away from the basal lamina environment. In most cases, the cytoplasm becomes less abundant and more eosinophilic, (presumably indicating increased numbers of intermediate filaments), while in other cases the cytoplasm can become pale and more abundant. The important benign feature is that the cytoplasm changes predictably according to the distance of the cells from the edge of the duct. 3-The ductal cells appear to align their long axes to point in the same direction(17) as if they assembled cell junctions in a coordinated manner with their neighbors. This third benign feature is shown in Figures 7-9. The typical pattern-recognition terms that are taught to pathologists in training include "swirling" of the cells in the same direction (box), or "streaming" of ductal cells with formation of "attenuated (or "weak") bridges" around the incompletely filled "slitlike spaces".

Figures 10 and 11

A second group of benign ductal proliferations are called intraductal papillomas. In papillomas, the ductal cells seem to proliferate along with an increase in the stromal fibroblastic compartment, as if the stromal fibroblast proliferations were stimulated along with ductal cells. Both the ductal cells and myoepithelial cells comprising papillomas are monoclonal(18, 19). Stratification (increased numbers of cells lining the basal lamina) is not always prominent in papillomas.

Figures 10 and 11 show a small papilloma with the characteristic admixture of stromal fibroblasts. The benign nature of this is still manifest by one or more of the three features described above: 1-a mixture of myoepithelial cells (or varied cell types) within the proliferation, 2-an alteration of the cytologic features of the ductal cells depending on how close they are to the basal lamina compartment. 3-a tendency to develop a similar orientation of their long axis. (Boxes in Figure 11)

Figures 12 and 13

Figures 12-13 show a larger intraductal papilloma. See if you can find the features indicative of a benign ductal proliferation without looking at the figure legend.

Advanced probable precursor to invasive ductal carcinoma: Atypical ductal hyperplasia

Atypical ductal hyperplasia (ADH) appears more evolved or closer to intraductal carcinoma than the former groups of proliferative fibrocystic changes, and its occurrence predicts a 4 fold increased relative risk for subsequent invasive ductal adenocarcinoma(2,12). The morphologic features of ADH can overlap with the proliferative ductal lesions at one end of a spectrum, and with intraductal carcinoma at the other end(20). The increase in fitness of the ductal epithelial cells in atypical ductal hyperplasia compared to the proliferative ductal lesions previously described can be recognized given the assumptions: The ductal cells do not require the trophic support provided by the basal lamina environment, and asymmetric cell divisions that give rise to myoepithelial cells become very rare. In ADH, the ductal cells seem to have similar cytologic features whether or not they are near the basal lamina or myoepithelial environments, and the orientation of one ductal cell seems to be nearly independent of the orientation of neighboring cells. The low magnification impression is that the duct becomes expanded by a nearly homogeneous (clonal appearing) population of cells.

Figure 14

Figure 14 shows atypical ductal hyperplasia arising in the midst of a proliferative ductal lesion. The asterisk is in an area with usual hyperplasia (with more condensed nuclei in the center of the duct) and the arrow heads show two foci with an apparent clone of ADH in which the ductal population appears more uniform and the cell growth is about as robust in the center of the proliferation as it is next to the basal lamina compartment.

Figure 15

Figure 15 shows a different patient's focus of ADH. The distinction from intraductal carcinoma is somewhat arbitrary but would include the following: 1-the cells have slightly different cytologic features in one area of the duct (boxed area Figure 15) than the rest of the population, and they are focally oriented in the same direction (just above the box). Thus the cells still appear to show some very mild trophic dependence or responsiveness to their environment. 2-Arbitrarily, if the proliferation does not extend over 2 mm, ADH rather than intraductal carcinoma is diagnosed(3). It is as if the proliferation must show an ability to expand in order to warrant a diagnosis of intraductal carcinoma.

Intraductal carcinoma

This proliferation is also known as ductal carcinoma in situ or DCIS. This lesion presents no immediate threat to the patient since the cells are not "invasive" and cannot metastasize. The diagnostic features of the various forms of DCIS can be accurately appreciated given the following assumption about the ductal cells' fitness: The ductal cells have no dependence on the basal lamina environment for trophic support. Thus they grow equally well at any polarity anywhere within the duct. "Grow equally well at any polarity" means the cytologic features of the cells are the same anywhere in the duct and cells have a random orientation of their long axes relative to each other (except as occasionally noted in(17)). Figures 16 and 17 show one subtype of DCIS (called "micropapillary" DCIS).

Figures 16 and 17

The ductal cells have a uniform appearance throughout many duct profiles, reflecting the expansion of a clone of cells. The defining feature of malignancy is that the cells in the center of the duct well away from their natural environment of the basal lamina, have the same cytologic features as the cells near the basal lamina. A few myoepithelial cells may rarely be seen in DCIS, probably left from the previous populations that inhabited the duct before being colonized by DCIS. Figure 18 and 19 show a different example of micropapillary DCIS.

Figures 18 and 19

Again, the defining feature is the ability of the cells in the center of the duct to grow as well as the cells near the native basal lamina zone. The reader is cautioned that occasionally a benign condition called apocrine metaplasia (defined by an appearance of the cytoplasm resembling apocrine glands) can show a stable multilayering of the cells without evidence of an altered nuclear appearance (discussed in 1,2,3).

Free to orient randomly and grow equally well anywhere in the duct, highly regular geometric patterns are often formed in the ducts that are incompletely filled with the ductal cells. Figures 20 and 21 show the appearance of cribriform DCIS, with nearly spherical holes randomly placed within an incompletely filled duct.

Figures 20 and 21

Spherical holes in the incompletely filled ducts indicates that the surface area that is not in contact with a neighboring cells is minimized. In cribriform DCIS, it is as if cell junctions still cause cells to stick to each other, but the junctions were devoid of any apparent biologic tethering. The cells stick to each other to minimize the surface area, but the orientation of any cell contacts seems random. Figures 22 and 23 show "roman arches", smoothly arching trabeculae with a uniform thickness, composed of otherwise randomly oriented ductal cells.

Figures 22 and 23

Compare the cytologic and architectural uniformity of the DCIS in Figure 23 with the cells in Figure 6.

Comedo carcinoma (Figure 24 and 25) is a variant of DCIS defined by the presence of central necrosis within the duct lumen.

Figures 24 and 25

Typically the line that demarcates the necrosis is always the same distance--usually about 1 mm from the basal lamina--and it is generally assumed that the necrosis is due to a critical lack of a diffusible nutrient such as oxygen. The cells in comedo carcinoma probably need more oxygen that the cells of non-comedo DCIS because they tend to have high mitotic rates. A striking feature about comedo carcinoma is the ability of the cells to appear to maintain identical cytologic features whether the cells are located next to the basal lamina or next to the sharp line of necrosis. In Figure 25, numerous mitoses are present including some next to the zone of necrosis (arrow heads). Combinations of cribriform and comedo patterns are possible (Figure 26).

Figure 26

The basic feature of DCIS--that the cells seem to be able to grow with random polarity and with equal ability whether or not they are adjacent to a basal lamina--is also essentially diagnostic of adenocarcinoma in situ in most other glandular tissues, including the lung, the pancreas, the gall bladder and biliary tree, and the endocervix. (One difference: There are no apparent cells equivalent to myoepithelial cells in these sites.) Due to arbitrary differences in nomenclature used by pathologists specializing in different organ systems, however, the name given to the same type of lesion is sometimes different in other organs. In the gastrointestinal tract, (including the colon, small intestines, stomach, and glandular metaplasia of the esophagus--Barrett's esophagus), the proliferation that shows the features described for DCIS is currently usually called high grade dysplasia. For the ovary, the lesion that corresponds most closely to the definition of DCIS is called adenocarcinoma of borderline malignancy. The concept of in situ prostatic adenocarcinoma is controversial. The thyroid gland forms two types of adenocarcinomas, neither of which seem to progress through phases analogous to DCIS.

It is likely that other histologic patterns besides the ones recognized by pathologists are precursors to breast cancer. This is discussed in supplementary text (21).

Invasive ductal carcinoma

In invasion, the ductal cells appear able to induce a new trophic relationship with the stromal fibroblasts and blood vessels outside the duct. This process is called desmoplasia. Figure 27 shows invasive ductal carcinoma on the left with a focus of DCIS (comedo type) on the right.

Figure 27

The stromal tissue around the DCIS is eosinophilic (pink), while the stroma around the invasive tumor has a pale purple color. The purple color is from the presence of mucopolysaccharides, a product of activated fibroblasts. Figure 28 shows the junction between the DCIS and the infiltrating carcinoma.

Figure 28

The DCIS has a smooth outer contour (like a normal duct) since the stromal cells in this vicinity have produced this architecture long ago when the gland was benign. The stromal fibroblasts around the DCIS are inactive-appearing as manifest by the relatively condensed chromatin and inapparent nucleoli (arrows). In contrast, the fibroblasts next to the invasive tumor have been recently activated as indicated by the presence of a euchromatic nucleus, nucleoli, cytoplasmic basophilia (reflecting large amounts of polyribosomes/rough endoplasmic reticulum), and increased secreted mucopolysaccharides in the extracellular space. Compare the appearance of the stromal compartment in Figure 2. Invasive tumor cells often have a shaggy outer contour and individual malignant cells can sometimes be seen in the midst of the active fibroblasts (arrow).

Infiltrating ductal carcinoma is often highly desmoplastic. It is the proliferation of the fibroblasts and the resulting production of collagen that makes some of the tumors feel hard. A consequence of the desmoplasia is that a sample of infiltrating ductal carcinoma typically contains many stromal cells compared to the number of ductal cells. Desmoplasia appears very close and may be identical to the reaction of fibroblasts in normal wound healing. Thus, the appearance of the fibroblasts in Figures 27 and 28 and their production of mucopolysaccharides closely matches the appearance of fibroblasts about one week following a wound.While collagenases and proteases appear to play a role in invasion (22), the nature of their role may not merely be to "open the barn doors" to let the cells out of the duct. There are many natural examples of in situ carcinomas in which the basal lamina is physically disrupted, but invasion does not seem to develop. For example, there is no histologic evidence that invasive carcinomas develop next to any of the multiple transected ducts filled with DCIS in the innumerable patients for whom mastectomy follows a surgical biopsy. In such cases, one may find the DCIS cells next to the interface of the transected duct, but they appear to show no ability to grow out into the stroma. This is even more surprising since the tissue in which they find themselves has activated stromal cells very similar to the activated cells that a truly invasive carcinoma is able to induce. It is as if the increasing fitness displayed by invasive tumor cells were due to an acquired resistance to a growth inhibitory activity outside of the ductal microenvironment, an idea expressed by Petersen et. al. (8).

The histologic criteria for invasion are similar for many other epithelial malignancies. In some cancers, (including some ductal carcinomas and most lobular carcinomas) the desmoplasia is not as well-developed. The features diagnostic of invasion in such cases are hard to teach in just a few pages, but would include the finding of individual detached cancer cells, occasional increases in the amount of cytoplasm in the tumor cells, and the finding of cells growing without evidence of coordination with the stroma.

How cytologic features can be used to diagnose breast cancer

Cytopathology is the subspecialty of pathology whereby the characteristics of only tiny fragments, or sometimes even individual cells (obtained by fine needle aspiration, or from a nipple discharge) can be used to diagnose malignancy. In fine needle aspirates of the breast, the ductal epithelium and myoepithelial cells (if present) tend to peel away as a group from the basal lamina or stroma. These tiny tissue fragments can be diagnosed using the same framework for diagnosing proliferative ductal lesions and DCIS in tissue sections. Thus, the relation of ductal cells to each other or to any myoepithelial cells can provide diagnostic information. For example, tiny tissue fragments from a benign duct show a mixture of myoepithelial cells with ductal cells in essentially a 2 dimensional sheet (Figure 29, Papanicolaou stained).

Figure 29

The myoepithelial cells often have heterochromatic nuclei, and appear in a different plane from the ductal cells (arrow). Figure 30 shows a three dimensional architecture of cells, indicating that the duct is lined by cells more than one cell layer thick.

Figure 30

The cells appear to be a uniform population with little tendency to vary in their nuclear features in any direction (reflecting the putative ability of the cells to grow well whether or not they are next to the basal lamina), and there is no common polarity between adjacent cells. The findings were "suspicious for malignancy" and found to be atypical ductal hyperplasia on biopsy. Figure 31 shows unequivocal three dimensional growth of a monoclonal-appearing cell population in a different fine needle aspiration.

Figure 31

The same features diagnostic of DCIS in tissue sections are evident in Figure 31. Invasive ductal carcinoma cannot be reliably distinguished from DCIS by cytology.

Cancer cells do not always have to be in tissue fragments to be diagnosed. In fact, the discohesion of cancer cells--presumably reflecting decreased cell-cell contacts--is itself useful for recognizing cancer in cytology smears. The "cytologic criteria of malignancy"(16) allow cancer cells to be distinguished from normal ductal cells based on nuclear changes. These changes are not absolute and can be sometimes seen to some extent in normal cells. The changes are displayed well with alcohol fixation or air drying, and are harder to see in formalin fixed tissue sections. Most cancers anywhere in the body show nuclear changes compared to normal cells. The types of nuclear changes vary from cancer to cancer and there are important examples of specific nuclear structural changes diagnostic of certain types of cancer. Figures 32 and 33 show some of the nuclear features characteristic of many cancers: sharp and variably deep indentations of the nuclear membrane contour, lobulated nuclear contours, aneurism-like swelling of parts of the nuclear membrane, variable thickness of the nuclear membrane associated heterochromatin, asymmetry in the distribution of heterochromatin/euchromatin, large variations in any one cell in the size and shape of heterochromatin aggregates, and nucleoli that have angular contours or are inappropriately large for the apparent needs of the cell.

Figure 32 and 33

The apparent needs of the cell are gauged by the degree of basophilia of the cytoplasm (reflecting the number of mRNA-containing polyribosomes), the prominence of a golgi zone, and the amount of secretory material produced. Normal cells are typically uniform and symmetric, with smooth contoured nuclear membranes and evenly textured chromatin, and nucleoli that appear rounded and predictable from cell to cell and are coordinated with the apparent needs for ribosome synthesis(16).

There are problems applying the cytologic criteria of malignancy to individual cells in breast cancer diagnosis. Under some circumstances, cells believed on the basis of their patterns in a tissue section to be benign can closely resemble cancer cells. Often, but not always, this circumstance can be deduced from the appearance and mixture of other cells present. There are not enough studies of such examples to know whether these cells are truly neoplastic(21). Another problem is that some breast cancers do not show many nuclear changes. Thus, the architectural features described above are often more diagnostically useful than the changes in individual cell nuclei. When present, the nuclear changes are seen early in the apparent evolution. Comedo carcinoma nearly always shows nuclear changes diagnostic of malignancy. The degree of nuclear abnormality has prognostic significance(3).

Some, but not all, of the malignant nuclear features appear closely related to genetic instability. Tumors in which cell to cell heterogeneity and aneuploidy are most marked often have the most asymmetric chromatin distributions. It is not clear, however whether the nuclear abnormalities are a consequence or a cause of the genetic instability. Nuclear changes do not always occur in aneuploid cells, and in cell culture systems, oncogene expression per se, rather than karyotypic changes, seem to be associated with nuclear structural changes(23). Figure 34 shows an abnormal mitotic figure (highly predictive of aneuploidy). Such asymmetric mitoses are typically found in the cells with asymmetry in interphase. Aneuploidy cannot explain all of the abnormal nuclear features in cancer cells, and the biochemical basis or functional significance any of these nuclear characteristics of cancer are unknown.