Fax: (415) 353-9571
Clonality of lobular carcinoma in situ and synchronous invasive lobular carcinoma
Version of Record online: 6 MAY 2004
Copyright © 2004 American Cancer Society
Volume 100, Issue 12, pages 2562–2572, 15 June 2004
How to Cite
Shelley Hwang, E., Nyante, S. J., Yi Chen, Y., Moore, D., DeVries, S., Korkola, J. E., Esserman, L. J. and Waldman, F. M. (2004), Clonality of lobular carcinoma in situ and synchronous invasive lobular carcinoma. Cancer, 100: 2562–2572. doi: 10.1002/cncr.20273
- Issue online: 2 JUN 2004
- Version of Record online: 6 MAY 2004
- Manuscript Accepted: 9 MAR 2004
- Manuscript Revised: 2 MAR 2004
- Manuscript Received: 8 JAN 2004
- National Institutes of Health. Grant Numbers: CA44768, CA58207
- University of California–San Francisco Clinical Investigator Research Program Award
- lobular carcinoma in situ;
- ductal carcinoma in situ;
- comparative genomic hybridization;
- breast neoplasms
Lobular carcinoma in situ (LCIS) of the breast is considered a marker for an increased risk of carcinoma in both breasts. However, the frequent association of LCIS with invasive lobular carcinoma (ILC) suggests a precursor-product relation. The possible genomic relation between synchronous LCIS and ILC was analyzed using the technique of array-based comparative genomic hybridization (CGH).
Twenty-four samples from the University of California–San Francisco pathology archives that contained synchronous LCIS and ILC were identified. Array CGH was performed using random primer–amplified microdissected DNA. Samples were hybridized onto bacterial artificial chromosome arrays composed of approximately 2400 clones. Patterns of alterations within synchronous LCIS and ILC were compared.
A substantial proportion of the genome was altered in samples of both LCIS and ILC. The most frequent alterations were gain of 1q and loss of 16q, both of which usually occurred as whole-arm changes. Smaller regions of gain and loss were seen on other chromosome arms. Fourteen samples of LCIS were related more to their paired samples of ILC than to any other ILC, as demonstrated by a weighted similarity score.
LCIS and ILC are neoplastic lesions that demonstrate a range of genomic alterations. In the current study, the genetic relation between synchronous LCIS and ILC suggested clonality in a majority of the paired specimens. These data were consistent with a progression pathway from LCIS to ILC. The authors conclude that LCIS, which is known to be a marker for an environment that is permissive of neoplasia, may itself represent a precursor to invasive carcinoma. Cancer 2004. © 2004 American Cancer Society.
Our clinical understanding of lobular carcinoma in situ (LCIS) of the breast has undergone a gradual evolution since the term was coined by Foote and Stewart in 1941.1 Although, histologically a neoplastic lesion, LCIS has long been considered a marker for increased breast carcinoma risk rather than a precursor for invasive carcinoma, because it appears to confer an equally increased risk of breast carcinoma in both the contralateral and ipsilateral breasts.2–5 Furthermore, over half of invasive carcinomas that develop after a diagnosis of LCIS develop > 10 years subsequent to the index lesion, with most of these invasive carcinomas exhibiting ductal histology.2–4
Because of the concern regarding increased bilateral risk, the historically recommended treatment for LCIS included bilateral mastectomy. However, most clinicians now choose to follow their patients with expectant management, sometimes in conjunction with tamoxifen chemoprophylaxis. However, both the clinical significance and the biologic significance of LCIS remain uncertain. Over the past 2 decades, the greater use of screening mammography has resulted in more frequent incidental diagnoses of LCIS.6 In addition, more meticulous histologic evaluation has shown that LCIS often can be found adjacent to invasive carcinoma, particularly invasive lobular carcinoma (ILC). Epidemiologic studies conducted after the establishment of widespread mammographic screening have supported a model in which both LCIS and ductal carcinoma in situ (DCIS) confer an increased risk of contralateral breast carcinoma.6, 7 These findings again have raised the question of whether LCIS, like DCIS, may act as a precursor to invasive carcinoma.
Recently, advances in molecular techniques have fostered an unprecedented ability to identify discrete changes in the genome. This has allowed examination at high resolution of possible genomic relations between tumors. Using array-based comparative genomic hybridization (CGH), we explored the association between LCIS and synchronous ILC to determine whether they share a clonal relation.
MATERIALS AND METHODS
The current study was evaluated and approved by the University of California–San Francisco (UCSF; San Francisco, CA) Institutional Review Board. Twenty-four samples from the pathology archives at our institution that showed concurrent ILC and LCIS diagnosed between 1994 and 2001 were selected. Each specimen was reviewed to validate the diagnosis and to select regions for microdissection. Altered immunohistochemical staining for E-cadherin was used to confirm the diagnosis of ILC in all specimens.
Microdissection and DNA Extraction
Microdissection and DNA extraction were performed as described previously (Fig. 1).8, 9 Regions of synchronous LCIS and ILC were identified in each specimen. If both were present in the same section, then the distance between the in situ carcinoma and the invasive carcinoma was determined by direct measurement. For LCIS and ILC present in separate sections, the distance between them was calculated by multiplying the number of intervening blocks with the average thickness of all blocks (4 mm).
Nine slides were cut from each block selected for the study. Six slides were stained with 0.1% methyl green for microdissection. Using an adjacent hematoxylin and eosin (H&E)-stained section as a guide, normal cells were scraped away from the cells of interest. A drop of extraction buffer (10 mM Tris, 1.5 mM MgCl2, 50 mM KCl, and 0.5% Tween-20 (Fisher Scientific, Tustin, CA) buffer with 4 mg/mL proteinase K) was placed on the cells, which were then removed and deposited in approximately 15 μL extraction buffer per 1000 cells. In situ and invasive lesions were isolated separately. Cells were incubated in a shaking water bath at 55 °C for 3 days, with the addition of 0.3 μL of 20 mg/mL proteinase K per 1000 cells on Days 2 and 3. At the end of the incubation, proteinase K was inactivated by heating at 95 °C for 15 minutes. Samples were concentrated using a Microcon YM-30 column (Amicon Millipore, Bedford, MA), after which DNA was quantitated using the TaqMan (Applied Biosystems, Foster City, CA) real-time polymerase chain reaction (PCR).10
DNA Amplification and Labeling
DNA was random primer–amplified using components from the BioPrime DNA Labeling System (Invitrogen, Carlsbad, CA). The reaction consisted of 50 ng tumor DNA, 300 ng/μL random oligodeoxyribonucleotide primers, 20 U Klenow enzyme, 200 μM deoxyadinosine triphosphate (dATP), 200 μM deoxycytidine triphosphate (dCTP), 200 μM deoxythymidine triphosphate (dTTP), and 200 μM deoxyguanine triphosphate (dGTP). Tumor DNA and the random primers were denatured at 100 °C for 10 minutes and then immediately cooled on ice. The remaining reactants were then added, and the reaction was incubated at 37 °C for 2 hours. The final reaction volume was 25 μL. Excess nucleotides were removed using the QIAquick PCR Purification Kit (Qiagen Inc., Valencia, CA).
The labeling reaction mixture consisted of amplified DNA, 370 ng/μL random primers, 64 U Klenow enzyme, 200 μM dATP, 200 μM dCTP, 200 μM dGTP, 100 μM dTTP, and 75 μM Cy dye–conjugated deoxyuridine triphosphate (dUTP). DNA and random primers were denatured at 100 °C for 10 minutes and cooled on ice immediately. The remaining reactants were then added, and the reaction was incubated for 2 hours at 37 °C. Tumor DNA was labeled with FluoroLink indocarbocyanine (Cy3)-dUTP, and reference genomic male DNA was labeled with FluoroLink indodicarbocyanine (Cy5)-dUTP (Amersham Pharmacia, Piscataway, NJ). Excess primers and nucleotides were removed using a Sephadex G-50 column (Amersham Pharmacia).
Array CGH was performed according to protocols described previously.11–13 Human Array 2.0 chromium surface arrays were obtained from the UCSF Cancer Center Array Core. Each array was composed of 2464 bacterial artificial chromosomes (BACs) printed in triplicate, representing the entire human genome. Before use, arrays were ultraviolet (UV) cross-linked at a setting of 1300 μJ using a UV Stratalinker (La Jolla, CA).
Labeled tumor DNA and reference DNA were combined with 100 μg CotI DNA (Invitrogen), and the mixture was precipitated using 0.1 volumes of 3 M sodium acetate and 2.5 volumes of 100% ethanol. DNA was dissolved in a solution of 33 μg/μL yeast tRNA with 9% sodium dodecyl sulfate and then mixed thoroughly with hybridization solution to yield a final composition of 50% formamide, 10% dextran sulfate, and 2X standard saline citrate (SSC). DNA was denatured at 73 °C for 15 minutes and then incubated at 37 °C for 30–60 minutes. The DNA probe was applied to the array, the slide was sealed inside a slide box humidified with 50% formamide/2X SSC, and the slide box was incubated at 37 °C for 48 hours.
After hybridization, the array was washed twice in 50% formamide/2X SSC at 45 °C for 10 minutes and then twice in phosphate buffer containing 0.1% NP-40, pH 8.0, (PN buffer) at room temperature for 10 minutes. While the slide was still wet with PN buffer, 100 μL of 3 μg/mL 4,6-diamidino-2-phenylindole (DAPI) in a 10% solution of phosphate-buffered saline in glycerol was added to the slide, which was then coverslipped and sealed.
Image Capture and Processing
Arrays were imaged using a charged coupled device camera as described previously.11 Intensity data were acquired through the DAPI, Cy3, and Cy5 channels. The SPOT 2.0 software program (available from URL: http://cc.ucsf.edu/jain/public [accessed October 12, 2003]) was used to process the image data.14 For segmentation purposes, the DAPI image field was left blank, and combined test and reference images were used to define spot size and location. Test-to-reference ratios for each spot were calculated using absolute intensities, which were determined by subtracting the local background intensity from the foreground intensity. Post-SPOT data processing included discarding spots that were less than 15 pixels, spots for which correlations were less than 0.9 or in the bottom 10th percentile of correlations, spots with a test plus reference intensity of less than 200 or in the bottom 20th percentile, and spots with a (reference intensity/[sum reference and test intensities]) value less than 0.1. For each clone, a single centered log2 ratio of test intensity over reference intensity was calculated from the three replicate spots on the array. Clones for which the ratio was derived from only 1 of 3 replicates and clones for which replicate ratios had a log2 standard deviation > 0.15 were discarded. Clones that did not yield data in at least 80% of hybridizations were excluded from further analysis, as were suspected polymorphic clones (28 clones; detailed information available from URL: http://cc.ucsf.edu/people/waldman/breast/hwang.polymorphisms.xls [accessed October 12, 2003]). An average of 1900 clones per case were included in the final data analysis. A human DNA sequence draft obtained from the University of California–Santa Cruz Genome Browser (freeze date, August 2001; available from URL: http://genome.ucsc.edu [accessed October 12, 2003]) was used to map clones.
A global threshold of ± 0.20 defined gains and losses for all log2 ratios. High-level gains and losses were scored for clones with log2 intensity ratios > 1.0 or < −0.75, respectively. Differences in continuous variables between LCIS and ILC were evaluated using a two-tailed unpaired t test. To allow comparison of array CGH results with the results of previous studies of lobular carcinoma that were performed using chromosomal CGH, genomic alterations were expressed as whole-arm changes by calculating the mean log2 ratio for all clones mapping to each chromosome arm. A threshold of ± 0.14 was then applied to the average log2 ratio to define chromosomal arm gains or losses relative to the reference sample.
The concordance between the paired in situ and invasive components was calculated for whole-arm averaged data (i.e., the log2 ratio averaged over an entire chromosome arm), as described previously,9 using the following formula:
Because 1q gain and 16q loss were observed in almost every tumor, concordance was calculated after excluding these changes and those on the X chromosome and weighting for less common changes. A log similarity score comparing each synchronous LCIS-ILC pairing with all other possible LCIS-ILC pairings also was calculated. The individual terms of the similarity score were negative when alterations were discordant (i.e., when there was a genomic alteration at a particular locus in one lesion but not at the corresponding locus in the other lesion in the pair). In contrast, similarity scores were increasingly positive when set members were increasingly alike (i.e., when both had alterations of the same type at the same locus or when neither lesion had any alteration). Probability weighting ensured that agreement at rare alteration sites was weighted more heavily than was agreement at common sites.
Clinical characteristics of the study population are summarized in Table 1. None of the patients had bilateral disease, and all patients were female. Eleven of 24 patients (46%) had pathologic confirmation of lymph node metastasis. No patient had distant metastasis at initial diagnosis. The LCIS and ILC were in the same breast in all patients, although the distance between the two components varied (Table 1). Immunohistochemical staining of the invasive tumor component showed loss of E-cadherin in all but two specimens; weak, variable membrane staining was observed in both of these tumors, one of which was a mixed ductolobular carcinoma. All ILCs were positive for estrogen receptor staining.
|Patient no.||Age (yrs)||LICS extenta||Invasive size (cm)||Invasive gradeb||Lymph node status||E-cadherin expression (ILC)c||Approximate distance between LCIS and ILC (cm)|
Genomic Alterations in LCIS and ILC
Twenty-four pairs of concurrent LCIS and ILC lesions were analyzed by array CGH (log2 data are available at http://cc.ucsf.edu/people/waldman/breast/hwang.data.xls). A substantial proportion of the genome was altered in both LCIS and ILC lesions; 310 (± 105 SD) and 403 (± 122 SD) of 1945 BACs were altered in LCIS and ILC, respectively (P = 0.01), representing 16% and 21% of the genome (P = 0.04). For both LCIS and ILC, the most frequently altered individual BACs were parts of larger regions of gain or loss (Fig. 2A,B). The most common alterations usually involved the entire chromosome arm. Loss of the entire 16q arm, which includes the locus for E-cadherin (CDH1), was seen in 100% of samples. The second most common genomic change was gain of 1q, which was noted in almost 90% of both LCIS and ILC lesions. Nearly the entire 8p arm was lost in > 20% of LCIS and ILC lesions; however, it is noteworthy that the most proximal portion (8p11.1–11.2; 48.1–52.9 megabases [Mb]), which included 2 clones that mapped to the locus for fibroblast growth factor receptor 1 (FGFR1), was gained in > 10% of samples. Similarly, although 40% of LCIS samples and 60% ILC samples exhibited loss of the distal portion of 11q (11q14–11qter; 94.1–160 Mb), the proximal portion of 11q (11q11–11q13; 62.3–76.2 Mb) was lost less frequently, with the exception of the region that contained the CCND1 gene (79.2–79.9 Mb), which was gained in approximately 30% of all samples.
Loss of the distal portion of chromosome 1p (1p36–1p33; 1.0–40.6 Mb) was observed in 25% of LCIS lesions and in 40% of ILC lesions. This region included a clone for E2F2 (1p36.12; 26.2 Mb), which was lost in almost 30% of lesions. A second distinct region of loss was noted at 1p22–1p13 (92.5–128.7 Mb), with almost 40% of lesions exhibiting loss at this locus.
Overall, 64 clones had high-level gains (log2 intensity ratio > 1.0) in at least 1 instance. Ten LCIS lesions and 7 ILC lesions had at least 1 high-level gain, with a mean of 2.8 gains (± 4.8 gains) per sample for LCIS and 2.2 gains (± 5.5 gains) per sample for ILC. The highest gene copy numbers (9.2-fold increase) were observed at 11q13, which includes the loci for the genes CCND1, FGF19, FGF4, and FGF3.
One hundred thirty clones showed high-level losses (log2 intensity ratio < −0.75) in at least 1 ILC lesion or 1 LCIS lesion. Ten LCIS lesions (mean, 6.4 ± 14.7 high-level losses) and 10 ILC lesions (mean, 4.0 ± 9.9 high-level losses) had at least 1 such loss. The regions that most commonly contained BACs with high-level losses generally belonged to larger loci of genomic loss, primarily on 11q, 8p, and 16q, as described above.
In addition to the region that included the locus for E2F2, which was lost in almost 30% of lesions, several other loci that contained genes known to be associated with disease progression were frequently altered. Epidermal growth factor receptor (EGFR; 7p11.2; 59.7 Mb) was gained in > 70% of LCIS lesions and almost 60% of ILC lesions. Cyclin D1 (CCND1; 11q13; 79.2 Mb) was gained in almost 40% of both LCIS and ILC lesions. p53 (17p13.1; 75.1 Mb) was lost in 30% of LCIS lesions and in > 50% of ILC lesions, and MDM2 (12q15; 79.8 Mb) was gained in 20% of both LCIS and ILC lesions. Only 10% of LCIS lesions and 15% of ILC lesions gained ERBB2 (17q21.1; 41.9 Mb). E-cadherin (CDH1; 16q22.1; 81.6 Mb) was lost in nearly every lesion.
Chromosomal arm changes were calculated for each lesion based on thresholded averages of copy number ratios along the chromosome so that array data could be compared with data previously obtained using chromosomal CGH (see Materials and Methods). The mean number of chromosome arms altered in ILC was 8.9, compared with 5.0 in LCIS (P = 0.004). The most common changes in LCIS and ILC were 1q gain and 16q loss, which were present according to whole-arm averaged analysis in 87.5% and 100% of lesions, respectively (Table 2).
|Event||Frequency in in LCIS (%)a||Frequency in in ILC (%)|
Relation between Synchronous in Situ and Invasive Tumors
Array-based CGH allowed discrimination of unique alteration patterns shared by synchronous in situ and invasive carcinomas at high resolution. Figure 3 shows examples of the pattern of alteration seen in two matched pairs of LCIS and ILC lesions. The striking similarity in genomic changes between the in situ and invasive components of these lesions clearly demonstrated the common clonality of the two lesions.
Two approaches were used to quantify the degree of similarity between LCIS and its associated ILC. A weighted similarity score was calculated from high-resolution data (Fig. 4); strong evidence for the common clonality of synchronous specimens was indicated when the similarity score for the pair of specimens was higher than the score for any other different-case pairing of invasive and in situ carcinoma. In 14 of 24 cases, the synchronous lesions were more similar to each another than to any of the other lesion pairings. In addition, concordance of whole-arm alterations was calculated for each LCIS-ILC pair; these calculations revealed clear similarities between synchronous pairs in a large subset of cases (Table 3).
|Patient no.||Changes in common||LCIS only||ILC only||Concordancea|
|LC28||12q−, 16q−, 17−||1.00|
|LC69||1q+, 5q−, 7q−, 8p−, 9p−, 10p+, 11q−, 12p−, 15−, 16p+, 16q−, 17p−, 18q−, 20p+, 22−||1p−, 6p+, 13+||0.90|
|LC56||1p−, 1q+, 4−, 7q−, 11q−, 15−, 16q−, 18q−, 22−||6p+||11p−, 18p−||0.84|
|LC57||3−, 13−, 16p+, 16q−, 22−||7p+, 10q−||1q+||0.83|
|LC03||1q+, 5p+, 8p−, 8q+, 11q−, 16q−||17p−, 22−||0.80|
|LC55||1q+, 8p−, 11q−, 13−, 16p+, 16q−||8q+, 11p−||0.80|
|LC53||1q+, 7+, 8p−, 8q+, 11p+, 13−, 16p+, 16q−||1p−, 17p−, 19−, 22−||0.74|
|LC45||1q+, 16p+, 16q−||19q−||0.67|
|LC23||1q+, 6q−, 11q−, 16q−,||16p+, 18−, 22−||0.50|
|LC39||1q+, 16q−, 17p−||22−||11q+||0.50|
|LC48||12p−, 16q−, 20q+||5+, 12q+, 13−||0.50|
|LC04||1q+, 11q−, 13−, 16q−||4q−; 17p+, 19p+, 22+||8q+, 16p+||0.40|
|LC33||1q+, 6q−, 16q−||3p−, 17−, 22−||0.40|
|LC07||1q+, 8p−, 16q−, 17p−||8q+, 13−, 17q+||6−, 9+||0.36|
|LC12||1q+, 8p−, 16q−||12p−||6−, 17p−||0.33|
|LC19||1q+, 16q−, 18q−||18p−||1p−, 8p−, 11−, 12p−, 16p+, 22−||0.20|
|LC02||1q+, 16q−, 17p−||4p−, 6q−, 7+, 9+, 16p+, 17q−, 18−, 22−||0.15|
|LC38||16q−, 17p−||1q+||1p−, 8+, 11−, 13−, 16p+, 17q+, 19q−, 20p+, 22−||0.15|
|LC06||1q+, 16q−||16p+, Xp+||0.00|
|LC31||1q+, 16q−||7q−, 13−, 17p−||0.00|
|LC42||1q+, 16q−||1p−, 2−, 4−, 5p+, 6−, 7p+, 13−, 14+, 16p+, 17−, 18−, 19+, 20+, 22−||0.00|
|LC43||1q+, 16q−||1p−, 3p−, 5+, 6p−, 7p+, 11q−, 12+, 13−, 16p+, 17p−, 18+, 19q+, 20+, 21−, 22−||0.00|
To determine whether specific candidate genes may play a role in the progression of in situ carcinoma to invasive carcinoma, we searched for BACs that were either gained or lost in invasive tumors, but not in the matched in situ tumors, in at least 4 of 14 lesions that were related clonally according to the weighted similarity score (Table 4). For each BAC, the mean log2 ratio for the ILC that exhibited this change was calculated. Most BACs that were altered only in ILC lesions and not in LCIS lesions exhibited low-level alterations. These changes were found in scattered regions throughout the genome. The candidate genes associated with the altered BACs were identified using the human DNA sequence draft (freeze date, August 2001).
|Clone||Chromosome||Position (Mb)||Associated gene(s)a||Alteration||Average log2 ratio in ILCb|
|RP11-93O2||2||34.3||LOC84661, MGC11061, SRD5A2, XDH, SPG4, FLJ20837, BIRC6, LOC51072||Gain||0.317|
|RP11-133J6||6||46.0||KCNK5, KCNK16, KCNK17||Loss||−0.275|
Among breast carcinomas of ductal histology, there is consensus that DCIS is likely to be a direct precursor of invasive disease; however, the role of LCIS remains poorly characterized. Early studies suggested that LCIS was unlikely to lead to invasive carcinoma, given that it appeared to increase breast carcinoma risk bilaterally.2–5 Furthermore, it was observed that patients with LCIS commonly developed invasive ductal carcinoma rather than ILC.2, 4 However, the data obtained in the current study support the idea of a progression pathway from LCIS to ILC and call into question the prevailing model, in which LCIS simply represents a marker for a field defect, rather than a lesion that possesses its own invasive potential.
Other studies also have challenged this accepted paradigm. Nemoto et al.15 reported results from a series of cases of LCIS with microinvasion, suggesting that LCIS sometimes may give rise to ILC. Additional histologic evidence of this relation was reported by DeLeeuw et al.16 as well as our own group,17 which noted a loss of E-cadherin and catenins in both LCIS and ILC, but not in ductal lesions. Moran and Haffty18 reported on a series of 1096 patients who underwent breast-conserving surgery before 1992. In specimens that contained LCIS, those investigators found a 53% prevalence of ILC, compared with a 5% prevalence of ILC in patients without LCIS (P < 0.001).
We have observed that LCIS is an advanced neoplastic state with alterations that, on average, involve > 20% of the genome. Whole-arm averaging of the log2 ratio across each chromosome arm was used to compare our array CGH data with results reported previously using chromosomal CGH; this analysis showed that the average number of chromosomal alterations in LCIS (5.0) was less than the average number of alterations in ILC (8.9), DCIS (8.8), or invasive ductal carcinoma (8.6), but greater than the corresponding figure in atypical ductal hyperplasia (3.6).8, 19 These findings suggest that LCIS has greater genomic stability compared with other breast neoplasms and may indicate that LCIS has a lower propensity for disease progression than does DCIS.
Loss of chromosome arm 16q was seen in 100% of both ILC and LCIS lesions. Genomic alteration of 16q is an early change that also has been described in a variety of other types of breast neoplasms, such as atypical lobular hyperplasia, atypical ductal hyperplasia, tubular carcinoma, invasive papillary carcinoma, and low-grade DCIS.8, 9, 20–22 In these typically well differentiated, low-grade tumors, this loss frequently is due to a 1;16 fusion (as detected by fluorescence in situ hybridization), which results in loss of the distal arm of 16q combined with a translocation and gain of 1q.23 For low-grade breast carcinomas, including LCIS and ILC, this translocation may be an important early event in carcinogenesis.
The ubiquitous loss of 16q in both LCIS and ILC includes the locus for E-cadherin (16q22.1) and is associated with the loss of E-cadherin immunohistochemical staining. Promoter methylation, mutation, and allelic loss all have been proposed as mechanisms to explain loss of E-cadherin staining in breast carcinomas. Our data support a model in which it is likely that regional genomic loss plays an important role. The finding that 16q is lost in other tumors that nonetheless have normal E-cadherin staining21, 22 suggests that genomic loss is an early event. This genomic alteration may be followed by a “second hit” through mutation or methylation, resulting in the reduced E-cadherin staining phenotype seen in both LCIS and ILC. E-cadherin is central to cell adhesion and tissue morphogenesis and recently has been reported to play a role in the suppression of invasion.24 Therefore, its loss may play an important role in tumorigenesis for lobular carcinomas.
The region that contains the p53 gene (17p13.1) was lost in one-third of LCIS lesions and one-half of ILC lesions, whereas cyclin D1 (11q13) was gained in almost 40% of lesions. Also noteworthy are the differences between our own findings and the genomic alterations described in invasive ductal carcinomas. The region that codes for EGFR (7p11.2) is gained more frequently in lobular carcinomas than in ductal carcinomas (> 60% in the current study vs. 27% reported in a series of 1029 invasive ductal breast carcinomas25). Gains in the locus coding for ERBB2 (17q21.1) were conspicuously rare in the current data set (occurring in 10% of LCIS lesions and 15% of ILC lesions), whereas it has been shown that > 30% of invasive ductal carcinomas amplify and overexpress this gene. This finding is consistent with previous reports, which also have shown that ERBB2 amplification is uncommon in lobular carcinomas,26, 27 and indicates that genomic alteration of ERBB2 is not critical for lobular carcinoma progression.
The relation between LCIS and synchronous ILC was evaluated with two different metrics—a weighted similarity score and a concordance score (see Materials and Methods). Both analyses confirmed that many paired LCIS and ILC lesions were genomically similar. We have referred to synchronous LCIS/ILC lesions with the highest similarity scores as clonal, because the extent of shared genomic alterations was highly indicative that these lesions shared a common cell of origin. However, a possible alternate interpretation of the data may be that the presence of unrecognized polymorphisms may cause lesions from the same patient to appear more closely related to one another than to other lesions on the basis of these polymorphisms alone. We minimized this potential confounding influence by excluding all known polymorphic BACs from the analysis. After excluding a total of 28 BACs, the majority of LCIS/ILC pairs still were related more closely to one another than to all other lesions. Another potential confounder in our study was possible cross-contamination of in situ and invasive carcinomas. However, clonality was observed even in cases in which the LCIS and ILC lesions were separated by > 2 cm or were present in separate tissue blocks; this is the most compelling evidence to date of a precursor-product relation between LCIS and ILC.
Among the clonal pairs, there were several regions that were altered in ILC but not in LCIS (Table 4). Although some of these genomic changes may reflect only the overall increased genomic instability associated with disease progression, it is possible that genes located within these loci may be important to lobular carcinoma progression. Protein tyrosine phosphatase receptor type N, which is located at 7q36.3 (170.3 Mb), is a member of the protein tyrosine phosphatase family, which is known to play an important role in the regulation of cell growth and oncogenic differentiation. It has been shown that the protein kinase inhibitor α gene, which is located at 8q21.11 (89.8 Kb), is a potent competitive inhibitor of cyclic AMP–dependent protein kinase activity.28 It also has been reported that a group of ion channel genes, KCNF1 (2p25; 11.8 Mb), KCNK5, and KCNK17 (6p21; 46.0 Mb), were gained in invasive carcinomas but not in paired in situ carcinomas. It is known that some of these genes affect electrolyte metabolism, whereas others play a role in cell cycle regulation.
The clinical implications of these findings merit discussion. Although complete surgical excision of LCIS to clear margins is not the current standard of care, do our data support such a shift in practice? In the National Surgical Adjuvant Breast Project (NSABP) data set, only 4 of 182 women with LCIS developed invasive carcinoma at 5 years.29 Thus, at best, four incidents of ILC may have been prevented by more aggressive treatment for LCIS. However, the impact on long-term outcome remains uncertain, because it is unknown how many of those four patients would have had a different outcome as a result of earlier treatment. Furthermore, aggressive local treatment would not be expected to affect the contralateral risk of carcinoma conferred by LCIS. Fisher et al. showed that at 4 years of follow-up, women with LCIS were afforded a 56% overall reduction in invasive carcinoma risk with tamoxifen compared with placebo.30 Thus, for the majority of patients who are diagnosed with LCIS, current epidemiologic and clinical evidence supports the established approach of close expectant management and consideration of prophylactic tamoxifen therapy for bilateral risk reduction.
Nevertheless, although we do not advocate an abrupt change in the surgical management of LCIS, our data suggest a reevaluation of the role that LCIS plays in the breast carcinoma progression model. For patients with preinvasive lobular carcinoma, disease progression can lead to ILC. Previous reports indicated that any ipsilateral invasive carcinoma appearing after a diagnosis of LCIS was more likely to have ductal histology rather than lobular histology.2, 3 The apparently increased incidence of ipsilateral invasive ductal carcinoma after LCIS in earlier studies may be attributable to the absence of standardized criteria for LCIS, with the possibility that many cases of DCIS were incorrectly diagnosed as LCIS. Studies that were conducted after more comprehensive histologic criteria for lobular neoplasias were established indicate that ipsilateral ILC is more common after LCIS than after DCIS. In the NSABP-B17 data set, all four invasive ipsilateral carcinomas that developed after LCIS were of the lobular type.29 It is noteworthy that the only predictor of invasive carcinoma was the degree of lobular distortion (a possible marker of tumor grade) in the LCIS. Others have found that when LCIS is present in the setting of invasive carcinoma, the histology of the invasive component is more than twice as likely to be lobular than it is to be ductal.2, 4, 18 These clinical observations are entirely consistent with our genomic findings.
We propose that in addition to histologic type, the grade of preinvasive disease also may be an important determinant of both the rate and the likelihood of disease progression (Fig. 5). It is probable that some subtypes of in situ disease have a limited likelihood of ever progressing to invasion. Other subtypes, such as high-grade DCIS and pleomorphic LCIS, may be much more likely to progress. The samples of LCIS evaluated in the current study were selected specifically because of the accompanying ILC; thus, our data may characterize a subset of LCIS with an elevated invasive potential. However, even in the current data set, we have shown that LCIS exhibits significantly fewer genomic aberrations than have been reported in DCIS. Over the long term, this relative genomic stability may be manifested as both a decreased likelihood and a slower rate of disease progression (Fig. 5). Because most published studies cover relatively short follow-up periods, there may be an ascertainment bias in favor of in situ lesions that progress at a faster rate. It is certain that aggressive treatment is warranted for any in situ lesion that has a high propensity for progression. However, for an in situ lesion with limited invasive potential (such as most LCIS lesions and low-grade DCIS lesions), prevention strategies aimed at global risk reduction may represent the optimal long-term approach.
We conclude that genomic alterations, particularly 16q loss and 1q gain, are highly prevalent in both LCIS and ILC, with invasive lesions exhibiting more alterations overall compared with in situ lesions. High-resolution data were particularly powerful in identifying the genomic regions that may be important in the progression from in situ disease to invasive disease. Despite the overall greater frequency of genomic changes in ILC versus LCIS, our data suggest that progression to the invasive phenotype does not require a specific genomic change and is not dependent on the cumulative burden of genomic changes, thus implicating multiple potential pathways for progression. The clonal relation exhibited by most of these matched pairs supports the idea of a precursor-product relation between LCIS and ILC and indicates that LCIS is more than simply a marker for an increased risk of breast carcinoma.
The authors thank the University of California–San Francisco Cancer Center Array Core, directed by Dr. Donna Albertson and Dr. Dan Pinkel, for providing Human Array 2.0 genomic arrays.