| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Human Cancer Biology |
Authors' Affiliations: 1 Department of Pathology, Washington University School of Medicine, St. Louis, Missouri; 2 Breast Center and 3 Department of Molecular and Cellular Biology, Baylor College of Medicine, Houston, Texas; 4 Departments of Genetics and Pathology and Laboratory Medicine, Lineberger Comprehensive Cancer Center, University of North Carolina, Chapel Hill, North Carolina; 5 Department of Pathology, Riverside Methodist Hospital, Columbus, Ohio; and 6 Department of Human Genetics, VCU, Richmond, Virginia
Requests for reprints: D. Craig Allred, Department of Pathology and Immunology, Washington University School of Medicine, 660 S. Euclid Avenue, Box 8118, St. Louis, MO 63110. Phone: 314-362-6313; Fax: 314-454-5001; Email: dcallred{at}path.wustl.edu.
 |
Abstract |
|---|
 Top Abstract Materials and MethodsResultsDiscussionReferences |
|---|
Experimental Design: The extent of diversity for conventional histologic grade and standard prognostic biomarkers assessed by immunohistochemistry was evaluated in a series of pure DCIS (n = 200) compared with a contemporaneous series of IBCs (n = 200). A subset of the DCIS (n = 25) was evaluated by DNA microarrays for the presence of luminal, basal, and erbB2 intrinsic subtypes. The extent of diversity within individual cases of DCIS (n = 120) was determined by assessing multiple regions independently for histologic (nuclear) grade and several biomarkers by immunohistochemistry, which approximate microarrays in determining intrinsic subtypes.
Results: DCIS showed a broad distribution of conventional histologic grades and standard biomarkers ranging from well to poorly differentiated, nearly identical to IBCs. Microarrays showed the same intrinsic subtypes in DCIS as in IBCs. However, higher resolution analysis showed that multiple histologic grades, biomarker phenotypes, and intrinsic subtypes often coexist within the same DCIS, and these diverse regions probably compete for dominance. Diversity within cases of DCIS was highly correlated with mutated p53 (P = 0.0007).
Conclusions: These results support the hypothesis that poorly differentiated DCIS gradually evolve from well-differentiated DCIS by randomly acquiring genetic defects resulting in increasingly abnormal cellular features. This diversity is amplified by defects resulting in genetic instability (e.g., p53 mutation), and the alterations are propagated to IBC in a manner independent of progression to invasion.
HELUs and ADH, at the near end of this evolutionary spectrum, show very well-differentiated histologic and biological features (14), although the latter features have not been studied in great detail. IBCs, at the far end of the spectrum, show enormous histologic and biological diversity, ranging from very well to very poorly differentiated, based on many comprehensive studies. DCIS reside between these extremes and are usually divided into two (e.g., noncomedo versus comedo) or three (e.g., low versus intermediate versus high grade) categories reflecting variations in the degree of differentiation (15–18). Thus, diversity of substantial degree seems to emerge at the stage of DCIS during breast cancer evolution, although detailed studies of DCIS are also relatively few and there are still many inconsistencies and unknowns. For example, ADH is widely regarded as the nonobligate precursor of noncomedo/low-grade DCIS, because they show similar well-differentiated features, but not of comedo/high-grade DCIS, because of their dissimilar poorly differentiated features, leading to speculation that the latter pursue a different course of evolution from occult precursors and the evolution of intermediate-grade lesions are essentially ignored.
These issues are important clinically because biological diversity in DCIS may influence the rate of progression to IBC, diversity propagated from DCIS to IBC may influence prognosis and sensitivity to specific therapies, and breast cancer prevention strategies should be based on an accurate understanding of how breast cancers evolve to be truly effective. The purpose of this study was to comprehensively characterize the histologic and biological diversity of DCIS compared with IBC to help further our understanding of DCIS and its role in breast cancer evolution.
|
Materials and Methods |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
Studies involving DNA microarray analysis used 25 consecutive fresh-frozen samples of pure DCIS harvested from female breast surgeries at University of Texas Health Science Center San Antonio hospitals between 1990 and 1995 (these cases were a subset of those described above). The samples were manually dissected to obtain areas in which tumor cells comprised >75% of total cells using H&E-stained frozen sections for guidance.
Studies assessing histologic nuclear grade used available routine H&E-stained slides of consecutive clinical cases of pure DCIS (n = 120) at Baylor College of Medicine hospitals between 1996 and 2002 (average, 3.4 slides per case). A subset (n = 112) of these cases with available formalin-fixed paraffin-embedded tissue blocks were used in experiments assessing intrinsic subtypes by immunohistochemistry (average, 1.3 blocks per case).
The tissue harvesting and scientific studies had separate Baylor College of Medicine Institutional Review Board approval (protocols H-10493 and H-12585, respectively).
Conventional histologic scoring and grading. The Elston-Ellis modified Scarff-Bloom-Richardson method was used for conventional histologic scoring and grading of IBCs (19, 20). Briefly, scores from 1 to 3 were assigned to each of three histologic features, and the total histologic score (H-score) was used to determine overall grade. The features and general scoring criteria included (a) the degree (area) of tubule/gland formation in the tumor (1, >75%; 2, 10-75%; 3, <10%), (b) the degree of nuclear atypia or pleomorphism in the worst 10% of the tumor cells (1, mild; 2, moderate; 3, severe), and (c) the average mitotic count per high power field (HPF) based on counting a minimum of 10 HPFs (1, <1/HPF; 2, 1-2/HPF; 3, >2/HPF). Thus, for IBCs, H-scores ranged from 3 (the best differentiated) to 9 (the worst), and scores of 3 to 5 defined grade 1, 6 to 7 for grade 2, and 8 to 9 for grade 3.
The Elston-Ellis modified Scarff-Bloom-Richardson method was further modified for evaluating DCIS in this study by also scoring the extent (area) of central necrosis (1, <10%; 2, 10-50%; 3, >50%) because this histologic feature (i.e., "comedo" necrosis) has played such a prominent role in the historical classification of DCIS. Thus, H-scores for DCIS ranged from 4 (the best differentiated) to 12 (the worst), and H-scores of 4 to 7 defined grade 1, 8 to 9 for grade 2, and 10 to 12 for grade 3.
Histologic nuclear grading. Black's nuclear grading method as modified by Fisher (21) was used to assess diversity of histologic differentiation within individual cases of pure DCIS. All available H&E-stained slides for each case were evaluated and scored for the presence and estimated percentage of tumor cells showing grade 1 (mild), grade 2 (moderate), or grade 3 (severe) nuclear atypia involving at least 5% of total tumor cells. This strategy was chosen over the modified Elston-Ellis modified Scarff-Bloom-Richardson method used for conventional histologic grading described above because of the superior accuracy and ease of scoring nuclear grade compared with other features in regions comprising a relatively small percentage of a tumor (e.g., it is difficult to score mitotic count if the area was smaller than 10 HPFs, etc.).
Assessment of biomarkers by immunohistochemistry. Each of the 200 consecutive cases of pure DCIS, IBCs, and the DCIS component of the IBCs evaluated for conventional histologic score and grade were also evaluated by immunohistochemistry for standard prognostic biomarkers, including estrogen receptor (ER), progesterone receptor (PGR), erbB2, p53, and proliferation rate (Ki67). The immunohistochemistry used routine clinically validated methodology as previously described (4, 22–26). All biomarkers except Ki67 were quantified by the so-called Allred score (range, 0-8; refs. 24, 27) and "positive" was defined as scores 
3. Ki67 was quantified by absolute counting (>500 cells per case) and reported as average percentage of positive.
Each of the 112 consecutive cases of pure DCIS evaluated for diversity of histologic nuclear grade within individual tumors were also evaluated by immunohistochemistry for a set of biomarkers previously shown to be a reasonably accurate (80-90%) surrogate strategy for identifying intrinsic subtypes as originally defined by microarray studies of IBCs at the RNA level (28–30), including ER, GATA-binding protein 3 (GATA3), erbB2, cytokeratin 5/6 (CK5/6), CK18, and p53. The immunohistochemistry used routine methodology as previously described (23, 24, 28–30) and was quantified by the Allred score. For the purposes of this study, positive was defined as Allred scores 2 (any positive cells) and negative as Allred score of 0 (entirely negative). Diversity was defined as a change from positive to negative, or the opposite, for one or more biomarkers within an individual tumor, comparing regions with different histologic nuclear grades.
Microarray analysis. Total RNA for the microarray experiments was purified from the samples using the microRNA isolation kit (Stratagene), which is based on an aqueous/organic phase separation in guanidine isothiocyanate-phenol/chloroform extraction buffer. The isolated RNAs were then purified using RNeasy cleanup kit (Qiagen). The purified total RNA was then linear amplified two rounds and converted to cDNA using Eberwine's RNA amplification method, which is based on an oligo-dT primer and T7 RNA polymerase (31, 32). The cDNA was then in vitro transcribed to biotin-labeled cRNA with the BioArray high-yield RNA transcript labeling kit (Enzo).
Gene expression was measured using U95Av2 human GeneChip oligo-based microarrays (Affymetrix). These earlier-generation microarrays contain transcripts for an estimated 63,000 probe sets interrogating 
54,000 UniGene clusters (many clusters are represented by multiple probe sets), corresponding to 10,000 full-length annotated genes. Briefly, amplified biotinylated cRNA was further purified with the RNeasy mini kit (Qiagen) and fragmented (20 µg cRNA) in 5x fragmentation buffer (200 mmol/L Tris-acetate + 500 mmol/L KOAc + 150 mmol/L MgOAc at pH 8.1) at a 4:1 volume ratio (cRNA/buffer) for 10 min to a range of 35 to 200 bases to optimize hybridization. Fragmented cRNA (20 µg) was then mixed with hybridization controls (bioB, bioC, bioD, and cre) and hybridized to the microarrays for 16 h. After hybridization, the microarrays were washed, stained, and scanned to generate quantitative digital files.
Data analyses. dChip software7 was used to evaluate results from DNA microarray experiments. Normalization of raw data was by the invariant set normalization method, and expression was estimated by the perfect-match-only model. Analyses relevant to this study were restricted to an unsupervised hierarchical clustering using a list of 392 intrinsic genes (Supplementary Table S1) expressed in the samples. The intrinsic genes were distilled by one author (C.M.P.) from the results of previous microarray studies of IBC to optimally identify luminal, erbB2, and basal subtypes based on shared patterns of gene expression (33–35).
The results of experiments assessing protein expression by immunohistochemistry for the biomarkers ER, GATA3, erbB2, CK5/6, and CK18 were also subjected to unsupervised hierarchical clustering using dChip software for the purpose of identifying luminal, basal, and intrinsic subtypes (28–30).
Studies assessing diversity within individual cases of DCIS identified three categories: (a) cases showing no diversity, (b) cases showing diversity of histologic nuclear grade, and (c) cases showing diversity of both histologic nuclear grade and one or more biomarkers. The Student's t test was used to determine the significance of differences in the extent of diversity between these categories and the status of individual biomarkers.
|
Results |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
20% in cases with the highest H-scores, as well as the average proportion and intensity of positive cells in the cases (not shown). In contrast, alterations of erbB2 and p53 were uncommon (<10%) in DCIS with low H-scores, but gradually increased to 60% in cases with the highest H-scores. Average proliferation rates (Ki67) also gradually increased from <5% to 35% in direct correlation with H-scores. The distribution of H-scores and biomarkers were nearly identical in the IBCs and their DCIS component.
|
|
H&E-stained slides from 120 recent consecutive cases of pure DCIS were evaluated for the presence of cells showing different nuclear grades comprising at least 5% of total tumor cells (Fig. 3 ; Supplementary Fig. S1). No diversity of uclear grade was identified in 54.2% of the cases, including 29.2%, 22.5%, and 2.5% composed entirely of nuclear grades 1, 2, and 3, respectively. The remaining 45.8% of cases showed areas of diversity, including 30.0% with grades 1 and 2, 6.6% with grades 2 and 3, and 9.2% with grades 1, 2, and 3.
|
|
|
|
|
Discussion |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
Although this model has been very useful, it oversimplifies a complex process, and there are still important inconsistencies and unknowns. For example, ADH as currently defined are uniformly composed of cells with very well-differentiated histologic and biological features. DCIS are more diverse than ADH and are usually divided into two (e.g., comedo versus noncomedo) or three (e.g., low versus intermediate versus high grade) categories reflecting variations in the degree of differentiation (15–18). ADH is widely accepted as the nonobligate precursor of low-grade/noncomedo DCIS because they show similar well-differentiated features, but not of high-grade/comedo DCIS because of their poorly differentiated features, leading to speculation that the latter pursue a different course of evolution from occult precursors.
The results of this study show that DCIS vary on a broad continuum of differentiation, similar to IBCs and, along with other often overlooked evidence, support the hypothesis that most higher-grade DCIS gradually evolve from lower-grade DCIS and, thus, indirectly from ADH, by the random accumulation of genetic defects, which are propagated to IBC in a manner that is largely independent of progression to invasion (Fig. 6 ). This hypothesis is entirely compatible with current theories about breast cancer stem or progenitor cells, as discussed in more detail below.
|
Other supporting evidence comes from this (Fig. 1) and earlier studies (15, 38–40) demonstrating that DCIS show a broad range of differentiation in terms of conventional histologic scores/grades and standard prognostic biomarkers nearly identical to IBCs, suggesting that diversity for these features evolves first in DCIS and is later propagated to IBC, which was proposed by Gupta and colleagues nearly a decade ago (37). Although these features do not seem to influence the ability of DCIS to progress to invasive disease, they are associated with the rate of progression, as shown by studies showing a much higher rate of short-term local recurrence in higher-grade DCIS compared with lower-grade lesions treated by lumpectomy, although the rates converge with longer follow-up (41). DCIS and IBCs have also been shown to be very similar at the high resolution of global gene expression assessed by microarrays and other types of high throughput technologies (42–44).
The enormous range of histologic and biological diversity in DCIS (and IBCs for that matter) makes it nearly impossible to categorize them into two or three subtypes in any fundamentally meaningful way because cases exist with nearly every combination of features imaginable. This diversity also argues against the notion that ADH is a precursor of low-grade, but not high-grade, DCIS in the sense that it fails to explain the cellular origin of the majority of DCIS and because low-grade and high-grade DCIS do not really exist as distinct categories. Gradual change from well-differentiated to poorly differentiated DCIS seems more likely for the majority of cases, although some cases may bypass sequential change and evolve in a more direct or entirely different manner.
If it is true that higher-grade DCIS evolve from lower-grade DCIS, then there must be diversity within individual lesions at some point in time. A previous study by Lennington and colleagues noted the presence of multiple histologic growth patterns (e.g., cribriform, micropapillary, solid, etc.) in 46% of 100 DCIS evaluated (45). This study (Fig. 3) showed the presence of multiple histologic nuclear grades in 46% of 120 DCIS evaluated. Furthermore, nearly a third of the cases in our study showing diversity of nuclear grade also showed diversity for several important biomarkers when comparing areas of different nuclear grade within individual tumors (Table 1 and Fig. 4). The presence and magnitude of any type of diversity was highly correlated (P = 0.007) with positive immunohistochemistry staining for p53, which is a surrogate for mutation of the gene (22, 46) but not with the other biomarkers evaluated. Mutation of p53 leads to genetic instability (47), suggesting that randomly acquired defects with this outcome (and there are many possibilities) may enable and/or accelerate the development of diversity in DCIS. This argument may be somewhat semantic in the sense that acquiring genetic instability in the first place may be fundamentally responsible for the characteristics pathologists use to define DCIS. Earlier precursors, such as HELUs and ADH, show far less genetic diversity than DCIS (at least in terms of magnitude if not specific defects; refs. 3, 9, 48), and it is feasible that alterations of development and/or differentiation, perhaps adaptive in nature, are more important than heritable defects in the evolution of these lesions.
Our immunohistochemistry studies also showed the presence of multiple intrinsic subtypes within individual cases of DCIS (Fig. 5). Intratumor diversity of intrinsic subtypes has not been observed in microarray studies to date, possibly because they were based on RNA isolated from whole-tumor homogenates, which would mix and obscure underlying diversity. The presence of this diversity suggests that there is considerable plasticity during tumor progression and presumably regions with different characteristics would compete for dominance and eventually the most aggressive or poorly differentiated area would prevail. Well-differentiated DCIS may progress to more poorly differentiated DCIS by this general mechanism. The magnitude of intratumor diversity observed in this study at all levels evaluated (histologic grade, biomarkers, and intrinsic subtypes) must be the lowest possible estimate of the true diversity in DCIS, because such a small (<1%) volume of tumor was evaluated in these experiments and there is probably more than enough to account for the broad continuum of diversity observed between cases overall. It seems likely that similar intratumor diversity exists and evolves in IBCs as well.
Other evidence supporting the idea that higher-grade DCIS evolve from lower-grade DCIS comes from studies showing that diversity increases with time. For example, a diagnosis of DCIS was rare (<5%) before but common (20-30%) after the introduction of screening mammography (49, 50). On average, DCIS were larger and more poorly differentiated in the premammography era than they are today (50–53), which is also true for IBCs (50, 52, 54). For example, in this study, the average conventional histologic score/grade of the series of DCIS in Fig. 1 (diagnosed between 1980 and 1995) and Fig. 3 (diagnosed between 1996 and 2002) were 8.4/2.0 and 7.2/1.6, respectively (P = 0.0003). Screening mammography became routine in our hospitals in the late 1980s, and thus, the first series contained a substantial number of cases that were not detected by mammography whereas cases in the second series nearly all were. The significant decrease in histologic score/grade observed in the second series is consistent with the idea that early detection due to screening mammography identified lower-grade tumors before they progressed to higher-grade lesions.
A similar conclusion can be drawn from studies showing that DCIS contain many of the same specific genetic defects regardless of histologic differentiation, although the absolute number of defects is higher in more poorly differentiated lesions. For example, previous studies from our laboratory (9) assessing loss of heterozygosity (LOH) in DCIS categorized into high grade and non–high grade showed similar frequencies of LOH for the majority (80%) of the 15 cancer-associated loci evaluated, but there was a higher average number of LOH in the high-grade lesions (2.4/case versus 1.8/case; P = 0.13; Supplementary Fig. S2). Studies assessing chromosomal alterations by comparative genomic hybridization have also shown that high-grade and non–high-grade DCIS have many more similarities than differences in terms of specific defects (48). One of the loci in our LOH study spanned p53 (D17/s960 at 17p13), and DCIS with LOH at this locus showed significantly higher average LOH than those without (3.0 LOH/case versus 1.6 LOH/case; P = 0.0001). This is consistent with our immunohistochemistry results showing higher diversity in DCIS with mutated p53, further supporting the idea that defects leading to genetic instability promote diversity in DCIS, which, again, is a plausible mechanism for the progression of higher from lower-grade lesions.
There are two prominent hypotheses in current thinking about breast cancer evolution: the Wellings Jensen model emphasized in this paper proposes a prolonged evolution of worsening precursor lesions through the random accumulation of genetic mutations (1, 2), whereas the other proposes that subtypes of tumors arise through the expansion of categorical stem or progenitor cells, which may then also undergo mutations (55, 56). Both theories are based on compelling and reproducible observations, and, in our opinion, they are not intrinsically incompatible. For example, perhaps cancer stem cells are created from normal stem cells which slowly acquire and accumulate random mutations, explaining the prominent intratumor diversity and decades-long premalignant evolution proposed by Wellings and Jensen. The same mutation might have different consequences depending on the stage of commitment of the stem or progenitor cell it occurs in. Mutations in an early multipotential stem cell might result in more diverse and rapidly evolving tumors than defects in a more committed progenitor cell. Alternatively, perhaps cancer stem cells can be newly created from somatic cells by mutations which reactivate genes regulating stem cell behavior (self-renewal and multipotential differentiation). Furthermore, creating cancer stem cells by mutating normal stem cells or by reactivating stem cell properties in somatic cells are not mutually exclusive possibilities, and both could help explain the enormous biological diversity observed within and between breast cancers. Some cancers may even contain multiple stem cells, contributing to intratumor diversity. Even genetically transformed somatic cells without stem cell properties may have the potential to create a mass large enough to be potentially lethal.
If it is true that most breast cancers evolve from precursors which gradually change over time, then identifying biological alterations associated with early precursors before the development of substantial diversity may reveal effective strategies for the prevention of the majority of breast cancers, independent of differentiation. Hopefully, future studies will shed light on these important issues.
|
Footnotes |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).
Received 5/ 8/07; revised 8/16/07; accepted 9/17/07.
|
References |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  R. E. Roses, E. C. Paulson, A. Sharma, J. E. Schueller, H. Nisenbaum, S. Weinstein, K. R. Fox, P. J. Zhang, and B. J. Czerniecki HER-2/neu Overexpression as a Predictor for the Transition from In situ to Invasive Breast Cancer Cancer Epidemiol. Biomarkers Prev., May 1, 2009; 18(5): 1386 - 1389. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Vincent-Salomon and O. Delattre Role of Genetic, Genomic, and Transcriptomic Factors Am. Assoc. Cancer Res. Educ. Book, April 18, 2009; 2009(1): 97 - 103. [Full Text] [PDF]
|
|
|
|
|
|
P. Simpson Linear and Nonlinear Models of Progression from In situ to Invasive Breast Cancer Am. Assoc. Cancer Res. Educ. Book, April 18, 2009; 2009(1): 105 - 109. [Full Text] [PDF]
|
|
|
|
 |
|
 R. Natrajan, M. B. Lambros, S. M. Rodriguez-Pinilla, G. Moreno-Bueno, D. S.P. Tan, C. Marchio, R. Vatcheva, S. Rayter, B. Mahler-Araujo, L. G. Fulford, et al. Tiling Path Genomic Profiling of Grade 3 Invasive Ductal Breast Cancers Clin. Cancer Res., April 15, 2009; 15(8): 2711 - 2722. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. L. Kleinberg, T. L. Wood, P. A. Furth, and A. V. Lee Growth Hormone and Insulin-Like Growth Factor-I in the Transition from Normal Mammary Development to Preneoplastic Mammary Lesions Endocr. Rev., February 1, 2009; 30(1): 51 - 74. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. M. Kuerer, C. T. Albarracin, W. T. Yang, R. D. Cardiff, A. M. Brewster, W. F. Symmans, N. M. Hylton, L. P. Middleton, S. Krishnamurthy, G. H. Perkins, et al. Ductal Carcinoma in Situ: State of the Science and Roadmap to Advance the Field J. Clin. Oncol., January 10, 2009; 27(2): 279 - 288. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. L. Rosenberg Polysomy 17 and HER-2 Amplification: True, True, and Unrelated J. Clin. Oncol., October 20, 2008; 26(30): 4856 - 4858. [Full Text] [PDF]
|
|
|
|
 |
|
 L. Melchor and J. Benitez An integrative hypothesis about the origin and development of sporadic and familial breast cancer subtypes Carcinogenesis, August 1, 2008; 29(8): 1475 - 1482. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Polyak Is Breast Tumor Progression Really Linear? Clin. Cancer Res., January 15, 2008; 14(2): 339 - 341. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Cancer Research | Clinical Cancer Research |
| Cancer Epidemiology Biomarkers & Prevention | Molecular Cancer Therapeutics |
| Molecular Cancer Research | Cancer Prevention Research |
| Cancer Prevention Journals Portal | Cancer Reviews Online |
| Annual Meeting Education Book | Meeting Abstracts Online |
