The cytologic criteria of malignancy

Crowding (Growth With Less Basal Lamina Contact)

Clonal expansion can take place if cells can survive with less basal lamina surface area (Fig. 1B). Cytologists refer to this as crowding (Fig. 1C), and it may bear a relation to “loss of contact inhibition.” Crowded cells often become more columnar in shape, for example, in a tubular adenoma of the colon and endocervical adenocarcinoma in situ. Crowded cells show less cytoplasm between nuclei (they show a criterion of “overlapping nuclei”) when viewed en face. Extremely crowded clones often show elongation of the nucleus within the thin columnar cytoplasm. Pseudostratification (nuclei at random positions relative to the basal aspect of the cells—see Group 3 below) is also typical.

Papillary Formation (Induction of New Basal Lamina)

A second mechanism for clonal expansion would be for an evolving neoplasm to induce more basal lamina surface area to grow on. Cytologists distinguish two types of increased basal lamina surface area. In “papillary” growth, new basal lamina surface area, supported by fibrovascular tissue, projects away from the pre-existing extracellular matrix and becomes lined on the outside by epithelium (Fig. 1D). Papillary growth patterns are typical of cutaneous human papilloma virus (HPV) warts, papillary thyroid carcinoma (PTC), a subset of breast neoplasms, early stages of some pancreatic cancers, the most common ovarian carcinomas (serous carcinomas), early stages in the evolution of most colonic neoplasms, and the most common urothelial neoplasms of the bladder. The mechanisms for an evagination of new basal lamina in papillary growth have not apparently received much attention. A papillary organization of some epithelia (e.g., urothelium or mammary epithelium) is not a recapitulation of any normal developmental process and is virtually diagnostic of neoplastic growth (including both benign and malignant papillary neoplasms).

Glandular Complexity

An outward branching of ducts, in which the epithelium is surrounded by a basal lamina and extracellular matrix is another apparent mechanism for tumor cells to increase the physical space needed for clonal expansion. Pathologists typically refer to this as glandular complexity and this is a prominent feature of the early development of several kinds of adenocarcinomas [Mutter et al., 2007]. Unlike papillary growth patterns, glandular outpouchings appear to recapitulate aspects of normal branching morphogenesis of development. The reciprocal interactions between stroma and epithelium in branching morphogenesis appear likely to be exploited by cancer cells in the process of increasing lamina surface area for clonal expansion [Wiseman and Werb, 2002; Yamada and Cukierman, 2007].

Stratification of Epithelial Cells

Loss of a requirement for attachment of epithelial cells to the basal lamina constitutes a fourth mechanism for clonal expansion within the geographic confines of an epithelium. The growth of epithelial cells without attachment to a basal lamina gives rise to three-dimensional “stratified” clusters of epithelial cells.

Stratification appears to be a manifestation of the cell biology concepts of anchorage independent growth or resistance to anoikis [Frisch and Screaton, 2001; Debnath et al., 2002], a phenotype induced by a growing number of cancer genes [Pratap et al., 2009]. Cytologists distinguish different degrees of apparent resistance to anoikis. Mammary ductal cells stratify (survive independent of a basal lamina connection) in a low-risk lesion termed “usual ductal hyperplasia” (Fig. 1E). In usual ductal hyperplasia, an admixture of ductal cells and myoepithelial cells typically maintain some degree of shared polarity (defined below) or they tend to show subtle predictable differences in their activity depending on how far they are from the basal lamina [Weaver et al., 1996]. When cells appear as if they grow with a random polarity to each other, and grow equally well whether or not they are next to a basal lamina (i.e., their nuclear and cytoplasmic features cannot be predicted based on how far away they are from the basal lamina) carcinoma in situ (non-invasive) is diagnosed (Fig. 1F). The distinction of the hyperplasias from ductal carcinoma in situ is a problem area in cytology for which insight from cell biologists would be valuable.

Intraepithelial Cell Migration (Pagetoid Spread)

A fifth mechanism for clonal expansion is “pagetoid spread” (named after the 19th century surgeon/pathologist James Paget) in which a pre-invasive clone of cells appear to find room by migrating intraepithelially away from its origins. Pagetoid spread can allow an intraepithelial spread of neoplastic cells over many centimeters, for example, starting with the breast duct system and extending out into the peri-areolar skin in Paget's disease (Fig. 1G). Pagetoid spread is easier to appreciate in histologic sections. It is common feature in early melanoma development (particularly lentigo maligna) and is also relatively common in the development of breast carcinomas. In breast carcinomas it is virtually universal in lobular carcinoma in situ, a tumor in which E-cadherin expression is specifically lost [Contreras and Sattar, 2009]. However, ductal carcinomas of breast (that at least have immunohistochemically detectable E-cadherin) can also sometimes show pagetoid spread, and it is ductal rather than lobular carcinoma that extends out onto periareolar skin. Targeted overexpression of ERK 1/2 in matrigel acinar cultures of MCF-10A human breast ductal cells produces an apparently pagetoid phenotype in which cell contacts are loosened by migrating cells, associated with partial loss of membranous E-cadherin staining but without other manifestations of epithelial mesenchymal transition, and without alteration in the topography of the basal lamina [Pearson and Hunter, 2007]. Pagetoid spread necessarily requires cell migration as well as at least transient dissociation of cells from a physical connection to the basal lamina, implying that resistance to anoikis is part of the process.

Invasion

The defining feature of invasion is growth of cells without a predictable relation to pre-existing tissue landmarks established during development (Fig. 1H). To discern these large-scale architectural features generally requires a histologic section. Mitotic activity is irrelevant to the diagnosis of invasion, and it is not paradoxical that invasive tumors can show fewer mitoses than in situ carcinoma or even the normal cells from which they arose [Fischer et al., 2004b]. An important and overlooked observation of pathologists is that it is insufficient to ascribe “invasion” to a digestive process whereby cells break down basal lamina or extracellular matrix components. The process of performing an incisional biopsy necessarily breaks through the basal lamina and physically displaces non-invasive tumor cells into the extracellular matrix, and yet incisional biopsies do not transform an in situ tumor into an invasive tumor. In situ tumor cells displaced into the presumably growth factor-rich extracellular matrix of a healing biopsy wound can be sometimes be seen to involute in subsequent excisional biopsies. Though ECM modeling and breakdown is clearly experimentally important for invasion of cancer cells [Wiseman and Werb, 2002], pathology observations suggest that invasion also requires loss of responsiveness to negative growth regulatory signals from extracellular matrix.

The previous six criteria appear to bear a direct relation to solving the problem of clonal expansion within the confines of epithelial organization. The functional significance of the following criterion, for the mechanism of clonal expansion is not as easy to envision.

Loss of Intercellular Adhesion

Cytologists are able to gauge the degree of cohesiveness of adjacent epithelial cells. The brushing or fine needle biopsy procedure exposes planes of cleavage that are not evident from histologic examination of intact paraffin-embedded biopsies. Loss of intercellular adhesion compared to normal cells is a common finding in cancers of diverse origin (illustrated in examples in Figs. 2 and 3). Not all cancers (even starting from the same cell of origin) show loss of adhesion, and loss of adhesion is sometimes seen in benign tumors or even benign non-neoplastic diseases, and thus the diagnostic value of loss of cohesion is modest.

Loss of adhesion is obviously a non-specific phenotype that could reflect alteration of the diverse physiologies related to integrins, desmosomes, adherens junctions, and/or tight junctions [Burridge and Tsukita, 2001]. Such “adhesion physiologies” are inter-related to apoptotic signaling/anoikis [Gilmore et al., 2009], wnt signaling [Adler and Lee, 2001], epithelial–mesenchymal transition [Thiery et al., 2009], and tumor cell migration and invasion [Mercurio et al., 2001]. There is also extensive cross-talk between many of these adhesion physiologies and the regulators of apical–basal cell polarity [Feigin and Muthuswamy, 2009] as discussed in group 3.

Abnormal Mitotic Figures and Unpredictable Variation in Degree of Chromasia

Genetic instability was apparent to pathologists studying cancer since the early 1900s (reviewed in [Bignold, 2002]). Many familiar criteria of malignancy amount to recognizing genetically unstable cells, for example, tripolar mitoses and asymmetric metaphase plates. In interphase cells, the criteria that relate to genetic instability involve comparisons between cells, as opposed to looking for a change within a single cell. It is the unpredictable variation amongst cells that seems characteristic of genetic instability, rather than the occurrence of a specific structural feature within a population of tumor cells.

DNA content can be roughly estimated by the total integrated amount of hematoxylin in the nucleus [Schulte and Fink, 1995], at least when comparing adjacent cells of the same cell type. Some commercialized hematoxylin staining protocols are sufficiently stoichiometric that hematoxylin is used for computerized screening based at least partly on DNA content abnormalities in cervical Pap tests (Hologic, Incorporated, Marlborough, MA). Such an estimate of DNA content takes into account both nuclear volume and the darkness of staining. Thus, resting lymphocytes that have a small heterochromatic (dark staining) nucleus have the same DNA content as neurons that have large, pale euchromatic nucleus. DNA content is harder to estimate in tissue sections than it is in direct smears of cytology samples, because tissue sections can cut off part of the nucleus.

Many forms of cancer show obvious variation from cell to cell in DNA content. A rapidly dividing normal population will typically have fewer than about 25% of cells in G2 (with twice normal DNA content) and S-phase. Thus, the majority of cells will have a discernable uniform diploid DNA content. Except that cells in G2 have a twice normal overall nuclear size and a commensurate approximate doubling of cytoplasmic volume, there are typically no specific differences that distinguish a G1 cell from a G2 cell in terms of chromatin morphology or cytoplasmic features in cytology material. In many tumors, a high degree of genetic instability is present, and for these tumors, a modal DNA content is not evident with hematoxylin staining (Fig. 2A). The majority of aneuploid tumors have an overall increased average DNA content, and accordingly, overall “hyperchromasia” is typical. Variation in the degree of chromasia (or objectively measured DNA content) is relatively subtle, apparently reflecting gains or loss of individual chromosomes—chromosomal instability. The amount of variation suggests that for many cancers, the majority of chromosome segregations are unequal?

Cytologists are incorporating objective markers of chromosomal instability into clinical practice, including especially fluorescent in situ hybridization (FISH) for chromosome numeration. Supernumerary centrosomes, or abnormally shaped centrosomes are other potential markers [Pihan et al., 2003]. HPV infection itself appears to induce genetic instability [Munger et al., 2004]. The correlation of increased p53 immunostaining (reflecting loss of p53 function) [Tomasini et al., 2008] and FISH abnormalities is substantantial, and in some highly unstable tumors, p53 status and aneuploidy can often be predicted based only on the H&E or Papanicolaou staining appearance [Cho, 2009]. Aneuploidy is not always apparent based on hematoxylin staining, and the presence of aneusomy detected by FISH can be useful for diagnosing cancer [Halling et al., 2000; Brankley et al., 2006]. Aneusomy has been shown to be a marker for risk of cancer development in histologically normal cells at a considerable distance from the site of subsequent evolving neoplasms [Bronner et al., 2008].

Distinction from polyploidization

In contrast to the relatively subtle variation in chromasia reflected in chromosomal instability, marked variation in chromasia can possibly reflect polyploidization, or completion of multiple rounds of DNA replication without a mitosis, to produce 4 n, 8 n, 16 n, etc. DNA contents. Polyploidization appears to be a relatively common benign change, seen in some normal cells and in some benign tumors. It is a prominent and characteristic of particular types of benign tumors, for example, schwannoma [Takeshita et al., 1985] (Fig. 2B). Polyploidization is characterized by commensurate increase in both nuclear chromasia and cytoplasic volume without unpredictable segregation of these two features. It is as if polyploid cells were merely bigger versions of diploid cells, with a similar chromatin pattern from cell to cell in spite of polyploidization. Polyploidization is physiological in megakaryocytes, and it has been well-documented in aging liver cells and cardiomyocytes subjected to mechanical stress (e.g., in hypertension or following a heart attack which decreases total cardiomyocyte content) (reviewed in [Ravid et al., 2002]). Tetraploidy is commonly observed in otherwise benign urothelial cells examined in FISH based assays [Bubendorf et al., 2001]. The occurrence of polyploidization has not been actively explored in other body sites, but it appears likely that tetraploid cells are common in many organs. For example, based on Papanicoalou-stained morphology, some terminally differentiated squamous cells from Pap tests of perimenopausal women appear commonly to be tetraploid (Fig. 2C). Such changes have no relation to HPV infection, and are without apparent risk of progression to cervical cancer [Cibas et al., 2005]. Some malignant tumors, however, seem to show a combination of polyploidization and chromosomal instability.

Polyploidization has been proposed as a precursor to chromosomal instability, and is proposed to require altered p53 for continued cell cycling [Storchova and Pellman, 2004; Ganem et al., 2007], but the apparent widespread nature of polyploidization puts this proposal into question. If polyploidization is a precursor to chromosomal instability, it would appear that the rate of progression would have to be very low.

Micronuclei

Micronuclei are sometimes seen in genetically unstable tumor cells as a diagnostic trait (Fig. 2D). Amplified DNA in tumor cells is known to be able to become extruded from the nucleus during S-phase, enclosed by a nuclear envelope as a “micronucleus” [Shimizu et al., 1998; Shimizu et al., 2007]. Micronuclei also form from acentric chromosomes, or lagging chromosomes in cells with mitotic spindle abnormalities [Rao et al., 2008].

Overall Cellular Pleomorphism

Regardless of the type of genetic instability, unpredictable variation from cell to cell in the amount and character of cytoplasm (Fig. 2A,D) is an important criterion that seems to reflect a genetically unstable population prone to tumor progression.

Unpredictable Chromatin Pattern Variation

Unpredictable variation in the chromatin texture in adjacent cells is a criterion of malignancy (Fig. 2E). Large-scale chromatin packaging patterns evident to pathologists appear likely to be determined by histone modification (see also below) [Zinner et al., 2006], and it is tempting to hypothesize that epigenetic instability is a characteristic of some tumors [Stein et al., 2009]. Evidence for epigenetic instability in X chromosome silencing in breast cancers has been reported [Pageau et al., 2007].

Disorganization of Intermediate Filaments

Pseudostratification of Nuclei

Normal glandular cells tend to show a basally located nucleus with a regular amount of apical cytoplasm. Pseudostratification refers to positioning of the nucleus at various planes above the basal aspect of non-stratified epithelium (Fig. 1B,C). Some types of normal epithelium (e.g., respiratory and fallopian tube) can look pseudostratified, due to the admixture of different cell types in the epithelium. For glandular epithelia composed normally of just one cell type, pseudostratification can be seen in regenerating benign epithelium, but commonly it is also an early neoplastic change preceding carcinoma in situ.

Pseudostratification may be related to the emerging cell biology concept of nuclear positioning (reviewed in [Burke and Roux, 2009]), in which there is aberrant nuclear positioning relative to the apical–basal polarized cytoplasm. The mechanism for abnormal nuclear positioning in human cancers has apparently not been explored. Pseudostratification typically precedes true stratification in the evolution of many adenocarcinomas, suggesting that aberrant nuclear positioning would typically precede loss of resistance to anoikis. Nuclear positioning involves active physiological processes during development. For example, the nuclei of developing neuroepithelial cells migrate from an apical to basal and back to apical during the cell cycle, and nuclear positioning is crucial to the arrest of neuronal nuclei in stereotyped positions in the various layers of cortex. Wound healing involves directed movement of the nucleus in relation to the microtubule organizing center [Gomes et al., 2005].

The molecular basis of nuclear positioning relative to the apical–basal polarity involves a growing number of LINC (LInker of Nucleoskeleton and Cytoskeleton) proteins that span the inner and outer nuclear membrane to form a mechanical connection extending from the cytoplasm through the two membranes of nuclear envelope to chromatin [Starr and Han, 2003; Burke and Roux, 2009]. SUN domain-containing proteins are inner nuclear membrane transmembrane proteins that associate (within the perinuclear space) with outer nuclear membrane integral membrane proteins bearing KASH domains. SUN–KASH pairs interact with microtubules (directly or via dynein or kinesin), actin, and apparently intermediate filaments (via plectin). Remarkably, a direct mechanical link mediated by a SUN–KASH pair connects the cytoskeleton (possibly microtubules) to individual meitotic chromosomes, contributing to homologous chromosome pairing prior to meiosis 1 prophase (prior to nuclear envelope breakdown; reviewed in [Hiraoka and Dernburg, 2009]). Some SUN–KASH pairs show cell-type specificity, including the KASH-containing protein Nesprin 4 which was recently suggested to maintain the normal basal localization of exocrine glandular nuclei [Roux et al., 2009].

Loss of Cell Polarity

Loss of polarity is an important large-scale criterion for cancer diagnosis. Normal cells show a predictable orientation in adjacent cells. Loss of polarity is estimated by Cytologists based on the long axis of the nucleus of a cell, or by positioning of exocrine products. Random polarity of epithelial cells is potent evidence for carcinoma (Fig. 3D), as if cells had no predictable relation to each other or to environmental cues. Stratification of carcinoma cells (see above) generally occurs coincidental to loss of predictable orientation of the long axis of adjacent nuclei, suggesting a mechanistic relation between these criteria.

It seems likely that Cytologists' concept of polarity is related to cell biologists' concept of apical–basal polarity. If so, objective markers of apical and basal cytoplasm, or assessment of the determinants of apical–basal polarity should have diagnostic value. Loss of basal–apical polarity is induced in vitro by a number of cancer genes commonly mutated in a wide variety of human tumors. Loss of basal–apical polarity contributes to resistance to anoikis [Aranda et al., 2006] (reviewed in [Feigin and Muthuswamy, 2009]), further suggesting a mechanistic relation between loss of basal–apical polarity and stratification. One of the apical–basal determinants, Scribble, functions as a tumor suppressor in mammary tumorigenesis [Zhan et al., 2008]. The interplay between loss of basal–apical polarity and carcinogenesis is further highlighted by the ability of the transforming gene HPV E6 to induce degradation of apical–basal polarity determinant Scribble, and mutations in E6 that lack the ability to degrade Scribble are defective in “transforming” cells [Nakagawa and Huibregtse, 2000].

Cytoplasmic Auer Rods

Auer rods are elongated, needle-like cytoplasmic crystalline inclusions in immature myeloid cells (Fig. 3E). They are composed of fused lysosomes containing myeloperoxidase and acid and alkaline phosphatase—normal components of differentiating myeloblasts—and are virtually diagnostic of either acute myeloid leukemia or clonal pre-leukemic myelodysplastic syndromes (reviewed in [Yoshida et al., 2009]). They are hard to detect by H&E staining or Papanicolaou staining, but are readily apparent by a Giemsa stain used on air-dried fixed smears (see www.cytopathology.org for protocol). They are found in at least some cells of about 50% of acute myelogenous leukemia patients, and when present, their number can vary from many per cell to rare [Zittoun et al., 1984]. The genetic basis for their formation and functional significance are obscure [Yoshida et al., 2009].

Ringed Sideroblasts

Abnormal iron accumulation in the mitochondria of erythroid precursors, is the defining characteristic of sideroblastic anemias [Sheftel et al., 2009]. Iron is detectable with cytochemical stains and the mitochondria are (normally) arranged in a perinuclear ring (Fig. 3F). Several hereditary forms of sideroblastic anemias have been defined genetically, characterized by mutations in the pathways of heme synthesis (which takes place in mitochondria). Heme synthesis has to be completed before iron is incorporated, and feedback loops cause accumulation of iron in the mitochondria. Acquired forms of sideroblastic anemias have been described, for example, due to drugs or vitamin B6 deficiency that interfere with heme synthesis. These acquired forms as well as the hereditary forms of sideroblastic anemia are not clonal or pre-malignant. However, an idiopathic acquired form of sideroblastic anemia is characteristic of a clonal, pre-leukemic myelodysplastic syndrome called refractory anemia with ringed sideroblasts (RARS). Of the heterogeneous groups of myelodysplastic syndromes, RARS has a relatively low rate of progression to acute leukemia. The genetic basis of sideroblast formation in RARS remains unclear, and there are no clear biochemical clues as to why sideroblasts are associated with clonal evolution only in RARS. Heme synthesis in RARS is rarely impaired, suggesting that a more downstream mitochondrial event is affected, such as reduction of ferric to ferrous iron, incorporation of ferrous iron into the heme, or possibly export of iron-bound heme from the mitochondria for incorporation into cytoplasmic hemoglobin [Gattermann, 2000]. Sideroblasts are evident at earlier stages of erythyroid differentiation in RARS compared to the other non-clonal forms of sideroblastic anemia [Sheftel et al., 2009], and mitochondrial autophagy is more pronounced and occurs at earlier stages of erythroid differentiation in RARS as well as other myelodysplastic patients compared to normal subjects [Houwerzijl et al., 2009].

Nuclear Lamina Alterations

There are a wide variety of nuclear lamina changes diagnostic of various types of tumors. Many of the distinctive lamina alterations are apparently independent of genetic instability, due to their conservation through clonal evolution, or their appearance in relatively genetically stable tumor cells. Some nuclear lamina alterations appear to be a consequence of activation of particular cancer genes. However, a precise structural explanation for any of the nuclear lamina alterations in cancer is lacking. What follows is a morphologic description of different types of nuclear lamina alteration and the cancer genes that apparently signal for a reorganized lamina. At the end is a brief summary of the possible structural basis of lamina alterations.

Cell Biology basis of nuclear irregularity

The large-scale organization of the nuclear lamina defines nuclear shape (reviewed in [Gruenbaum et al., 2005]). The lamina is the proteinaceous layer associated with the inner surface of the inner nuclear membrane, and it comprises non-transmembrane lamins A/C, B1, and B2 as well as about 80 transmembrane proteins specifically localized at the lamina [Schirmer et al., 2003]. Many of them are still poorly characterized and the majority have >10-fold variation in expression amongst different cell types [Wilkie and Schirmer, 2006]. These proteins intersect a wide variety of nuclear physiologies [Gruenbaum et al., 2005], and it is tempting to relate changes in nuclear shape in tumor cells to alterations in physiologies associated with these proteins. Lamina proteins are mutated in a growing number of “laminopathies”—degenerative diseases involving principally muscle, nerves, and fat. For example, mutations in Lamin A/C or the transmembrane protein emerin is found in various neuromuscular degenerative diseases, and nuclear irregularity has been hypothesized to be important in the cellular degeneration characteristic of these diseases [Broers et al., 2004; Lammerding et al., 2004] (reviewed in [Dauer and Worman, 2009]).

However, the relation of lamina proteins to abnormal nuclear shape in cancer cells, or more broadly, a relation of nuclear lamina irregularity to an altered lamina physiology in tumor cells is far from clear. Most cells in patients with laminopathies have no discernable abnormality of shape [Lammerding et al., 2005], and conspicuously absent from the disease associations in the laminopathies is predisposition to cancer [Dauer and Worman, 2009]. While experimental mutations in lamins lead to abnormal nuclear shapes [Schirmer et al., 2001; Favreau et al., 2003], the mutations described so far induce aggregation of lamins. Aggregated lamina proteins have not, to the best of our knowledge, been observed in any tumor cells with nuclear irregularity. Lamina immunostaining shows a regular thickness of the lamina for lamin A/C, Lamin B1, emerin, and LAP 2 throughout the irregular folds and intranuclear inclusions of the nuclei of PTC, with a regular distribution of nuclear pores [Fischer et al., 2001]. Similar smooth distribution of lamina proteins has been shown in breast cancers with nuclear shape abnormality [Bussolati, 2008; Bussolati et al., 2008]. Overexpression of lamin B1 can cause expansion of the nuclear lamina with resulting redundant folds [Favreau et al., 2003]. However, normal amounts of lamin B1, have been detected in the irregular (diploid) nuclei of PTC with nuclear irregularity [Fischer et al., 2001]. Depletion of the newly discovered transmembrane nuclear lamina protein LEM2 in tissue culture cells leads to nuclear lamina irregularity, but with reduced cell survival [Ulbert et al., 2006]. A proteomic approach [Schirmer et al., 2003] to compare nuclear lamina composition between normal cells and genetically stable tumor cells with lamina irregularity (e.g., normal thyroid epithelium compared to PTC) would be a direct means to begin to test whether lamina proteins contribute to nuclear shape abnormalities.

The irregularity of the lamina induced by RET/PTC develops during interphase following microinjection of RET/PTC, without an intervening post-mitotic nuclear lamina reassembly [Fischer et al., 2003]. These results point to the possibility of an interphase force exerted on the lamina from chromatin or the cytoskeleton (reviewed in [Fischer et al., 2004a]). One single study identified beta catenin accumulation in many of the intranuclear cytoplasmic inclusions of PTC [Rezk et al., 2004]. Electron microscopy studies have not shown consistently distinctive traits of the cytoplasm or nuclear lamina associated with the intranuclear cytoplasmic inclusions in PTC [Johannessen et al., 1983; Oyama, 1989]. Distortion of the interphase lamina by chromatin-based forces could arise from large-scale interphase chromatin repositioning [Chuang et al., 2006], or theoretically from interphase condensations and decondensations of chromatin.

The lobulated nuclei of cell lines from pancreatic cancer, melanoma, and HeLa cells were shown to be demarcated by bands of perinuclear intermediate filaments, suggesting that lamina indentations can sometimes be caused by disorganized bundles of intermediate filaments pushing on the lamina [Kamei, 1994].

It is important to note that dramatic nuclear shape changes are physiological in the differentiation of lymphoid cells.

The molecular basis of the lack of resiliency of the lamina of small cell carcinoma is not known. Small cell carcinomas have very scant cytoplasm with a dot-like aggregation of their scant keratin as a diagnostic trait (see above). The absence of a robust cytoplasm could contribute to a lack of rigidity of the nuclei in small cell carcinomas, but bare nuclei stripped of cytoplasm from other cell types do not appear to crush as easily. Small cell carcinomas lack lamins A and C [Broers et al., 1997]. However, other normal cells and tumors (including the histogenetically related low-grade neuroendocrine carcinomas) also lack lamin A and C and they do not show fragile nuclei. Small cell carcinomas also show a distinctive chromatin organization with respect to the lamina (see below), and this may also bear a relation to the fragility of the nucleus.

Chromatin Alterations

A wide variety of chromatin texture alterations have diagnostic value in specific contexts. Chromatin changes are difficult to quantify and we only describe a few relatively clear-cut examples. Some chromatin texture changes develop as a consequence of cancer gene signaling, but the actual structural basis of altered chromatin packaging in these examples is not understood. What follows is a morphologic description of a few criteria involving chromatin, with a description of the associated cancer genes and observations relevant to finding their precise structural basis.

Abnormalities of Nucleoli or Ribosomes

A common important criterion is nucleolar prominence in the absence of an apparent increase in the number of ribosomes in the cytoplasm [Frost, 1986] (Fig. 2D). A probably related criterion is “high nuclear to cytoplasmic ratio.” In physiologically activated cells (e.g., in epithelial cells proliferating to cover an ulceration), nucleolar prominence is matched by an increase in the amount of cytoplasm and cytoplasmic protein synthesis. The quantity of ribosomes and mRNA can apparently be roughly estimated by the amount of hematoxylin staining in the cytoplasm. Histochemical studies from the 1940s documented a disproportionate amount of nucleolar RNA to cytoplasmic RNA and protein as a specific feature of cancers [Caspersson and Santesson, 1943]. Explaining this finding, Caspersson et al. wrote: “A cell can be diagnosed as a malignant tumour cell…when external factors are so unfavorable for growth that no cytoplasmic increase takes place although the system of protein formation gives evidence of intense stimulation” [Caspersson and Santesson, 1943]. The molecular mechanism and genetic underpinning of the disproportionate increase in nucleolar size compared to cytoplasmic protein production or accumulation remains obscure [Fischer et al., 2004a], but the cross-talk between the protein synthetic machinery and other cell physiologies, including cell cycle progression, is extensive [Young et al., 2007] (reviewed in [Ruggero and Pandolfi, 2003; Lempiainen and Shore, 2009]). The effects of increased nucleolar size are not likely to be trivial to the cell, given the high energy requirement for producing the protein synthetic machinery [Warner, 1999].

It is important to note that inconspicuous nucleoli are actually characteristic of some cancers and pre-malignant neoplastic processes. Examples include small cell carcinoma (Fig. 3L) and intraepithelial squamous dysplasia (Fig. 2A), respectively. Both are genetically unstable and have a high growth fraction, yet throughout the clonal proliferation there is a uniform absence of nucleolar prominence.

Generally, cellular level cytologic features do not predict whether a cancer cell is invasive or whether it is still growing in situ. A major exception is invasive squamous cell carcinoma (e.g., the most common form of cervical cancer). While the pre-invasive intraepithelial neoplasias characteristically lack nucleoli, invasive squamous cell carcinomas often acquire an enlarged nucleolus along with more euchromatin, and increased amounts of cytoplasm (Fig. 1H) [Cibas and Ducatman, 2009].