Correspondence

Gene-Expression Profiles in Hereditary Breast Cancer

N Engl J Med 2001; 344:2028-2029June 28, 2001

Article

To the Editor:

The sample set used by Hedenfalk et al. (Feb. 22 issue)1 may not be representative of the natural distribution of the characteristics of tumors with BRCA1 mutations, tumors with BRCA2 mutations, and sporadic cases of breast cancer. The proportions of tumors with BRCA1 mutations and tumors with BRCA2 mutations that are reported to be negative or positive for estrogen receptors vary widely depending on age and ethnic background of the study cohort.2 Human mammary glands contain two distinct types of epithelial cells, basal and luminal. Perou and colleagues3 have suggested that estrogen-receptor–negative cancers encompass at least three biologically distinct subtypes of tumors: basal-like tumors, tumors with amplified expression of erb-b2, and luminal-like tumors that are negative for estrogen receptors and erb-b2 and positive for keratin 8.

High-level amplification of the erb-b2 gene occurs in about 15 to 20 percent of breast cancers, but we have not observed it in any tumors with BRCA1 mutations.4 In addition, six of seven tumors with BRCA1 mutations that we have examined stained positive for basal keratin 5, basal keratin 17, or both, a finding compatible with the basal-like pattern of gene expression.3 Consistent with our observation, in Figure 2 and Figure 3 of their article, Hedenfalk et al. also show that the level of expression of keratin 8 was low in tumors with BRCA1 mutations whereas erb-b2 was not overexpressed. In contrast, keratin 8 was highly expressed in tumors with BRCA2 mutations, a pattern of expression previously described for estrogen-receptor–positive sporadic tumors of luminal origin.3

Finally, hypermethylation of the BRCA1 promoter region is strongly correlated with negativity for estrogen receptors,5 and the pattern of expression in the case of the tumor misclassified as BRCA1-mutation–positive could be due to its negative estrogen-receptor status. Differences in the gene-expression patterns of BRCA1 and BRCA2, therefore, could also be consistent with different cellular origins of the malignant clone, a possibility that was not addressed by the authors.

Olufunmilayo I. Olopade, M.B., B.S.
Tatyana Grushko, Ph.D.
University of Chicago Pritzker School of Medicine, Chicago, IL 60637

5 References

1. 1

   Hedenfalk I, Duggan D, Chen Y, et al. Gene-expression profiles in hereditary breast cancer. N Engl J Med 2001;344:539-548
   Full Text | Web of Science | Medline

2. 2

   Phillips KA, Adrulis IL, Goodwin PJ. Breast carcinomas arising in carriers of mutations in BRCA1 or BRCA2: are they prognostically different? J Clin Oncol 1999;17:3653-3663
   Web of Science | Medline

3. 3

   Perou CM, Sorlie T, Eisen MB, et al. Molecular portraits of human breast tumours. Nature 2000;406:747-752
   CrossRef | Web of Science | Medline

4. 4

   Grushko TA, Hagos F, Olopade OI. Her2/neu gene amplification is not a feature of BRCA1-associated early onset breast and ovarian cancers. Mol Biol Cell 1999;10:Suppl:426a-426a abstract.
   Web of Science

5. 5

   Catteau A, Harris WH, Xu CF, Solomon E. Methylation of the BRCA1 promoter region in sporadic breast and ovarian cancer: correlation with disease characteristics. Oncogene 1999;18:1957-1965
   CrossRef | Web of Science | Medline

To the Editor:

Hedenfalk and colleagues claim that their results “indicate that a heritable mutation influences the gene-expression profile of a tumor.” However, the study does not allow a generalization to be made on the basis of a small subgroup of women with breast carcinomas (women with BRCA1 mutations and BRCA2 mutations make up less than 10 percent of the total number of patients presenting with breast carcinomas1). The 51 genes were found by selecting the genes that best differentiated among all three types of cancer, with no statement regarding the validity of this selection. The finding that among 5361 genes, 51 genes had levels of expression that differed among 14 patients (7 carriers of a BRCA1 mutation and 7 carriers of a BRCA2 mutation) may be due to chance.

Hedenfalk et al. used leave-one-out cross-validation to estimate the misclassification rate of the classifier that differentiates between BRCA1-mutation–positive tumors and BRCA1-mutation–negative tumors, and the classifier that differentiates between BRCA2-mutation–positive tumors and BRCA2-mutation–negative tumors. The use of this approach for this purpose gives only an inaccurate estimate of the misclassification rate.

Jon Sudbø, M.D., Ph.D.
Albrecht Reith, M.D., Ph.D.
Norwegian Radium Hospital, 0310 Oslo, Norway

Ole Christian Lingjærde, Ph.D.
University of Oslo, N-0316 Oslo, Norway

1 References

1. 1

   Arver B, Du Q, Chen J, Luo L, Lindblom A. Hereditary breast cancer: a review. Semin Cancer Biol 2000;10:271-288
   CrossRef | Web of Science | Medline

Author/Editor Response

The authors reply:

To the Editor: As Drs. Olopade and Grushko point out, the number of samples we studied may not be large enough to represent the characteristics of all breast cancers. However, the aim of our study was to determine whether we could detect differences in gene-expression profiles among tumors with BRCA1 mutations, tumors with BRCA2 mutations, and sporadic cases of breast cancer, recognizing that a larger sample would be required to characterize these differences completely and to place them in the context of the full spectrum of breast cancers.

With regard to the subclasses of estrogen-receptor–negative breast cancers observed by Perou et al.,1 the tumors with BRCA1 mutations in our study showed low levels of expression of keratin 8, erb-b2, and estrogen receptors. However, we do not believe that these findings necessarily imply a basal-cell origin for breast tumors with BRCA1 mutations, since phenotypic changes may occur during tumor evolution. Indeed, Pechoux et al. have demonstrated that luminal cells can give rise to myoepithelial cells.2 To extend our data, we performed immunohistochemical staining for the luminal marker MUC-1 on our tissue microarrays, and more than half of the tumors with BRCA1 mutations (9 of 16) stained positively for MUC-1. Moreover, as discussed, we agree that the differential expression of hormone receptors in BRCA1-mutation–positive and BRCA2-mutation–positive breast cancers affects the gene-expression profiles of these tumors. However, we believe that this does not fully account for the observed differences, since it was possible to differentiate between tumors with BRCA1 mutations and tumors with BRCA2 mutations even after the removal of the estrogen-receptor–related genes from the analysis. This finding suggests that a substantial portion of the gene-expression profile is due to the underlying mutations in BRCA1 and BRCA2.

In addition, hypermethylation in the BRCA1 promoter region has been associated with negativity for estrogen receptors,3 but a BRCA1-like pattern of gene expression in a case of sporadic cancer is not necessarily related solely to its estrogen-receptor status. Only one of three samples of estrogen-receptor–negative sporadic cancer was classified as positive for a BRCA1 mutation, and this tumor had the lowest level of expression of BRCA1, supporting our notion that the BRCA1-like profile of gene expression was due to hypermethylation of the BRCA1 promoter region.

Dr. Sudbø and colleagues raise questions regarding the statistical techniques we used. Because of the unique challenges posed by these larger data sets, we used several methods. The analysis was performed on a set of 3226 well-measured genes. We first used three methods to derive lists of genes: weighted gene analysis, permuted F tests and t-tests, and tumor–node–metastasis scoring and mutual-information scoring. A permuted F test with an α level of 0.001 resulted in 51 differentially expressed genes among the three types of cancer, whereas only approximately 3 genes were expected to be differentially expressed on the basis of chance.

It is correct that misclassification rates based on cross-validation are not precise; larger studies are needed to obtain more precise estimates. However, the fact that the number of cross-validated misclassifications was significantly smaller (P=0.003 for the classification of BRCA1-mutation–positive tumors and P=0.04 for the classification of BRCA2-mutation–positive tumors) than would be expected by chance establishes that the gene-expression profiles of mutation-positive and mutation-negative tumors differ and that these differences are not spurious.

Ingrid Hedenfalk, M.S.
Richard Simon, D.Sc.
Jeffrey Trent, Ph.D.
National Institutes of Health, Bethesda, MD 20892

3 References

1. 1

   Perou CM, Sorlie T, Eisen MB, et al. Molecular portraits of human breast tumours. Nature 2000;406:747-752
   CrossRef | Web of Science | Medline

2. 2

   Pechoux C, Gudjonsson T, Ronnov-Jessen L, Bissell MJ, Petersen OW. Human mammary luminal epithelial cells contain progenitors to myoepithelial cells. Dev Biol 1999;206:88-99
   CrossRef | Web of Science | Medline

3. 3

   Catteau A, Harris WH, Xu CF, Solomon E. Methylation of the BRCA1 promoter region in sporadic breast and ovarian cancer: correlation with disease characteristics. Oncogene 1999;18:1957-1965
   CrossRef | Web of Science | Medline

Citing Articles (9)

Citing Articles

1. 1

   Melinda L. Telli, Ellen T. Chang, Allison W. Kurian, Theresa H. M. Keegan, Laura A. McClure, Daphne Lichtensztajn, James M. Ford, Scarlett L. Gomez. (2011) Asian ethnicity and breast cancer subtypes: a study from the California Cancer Registry. Breast Cancer Research and Treatment 127:2, 471-478
   CrossRef

2. 2

   F. Fang, S. Turcan, A. Rimner, A. Kaufman, D. Giri, L. G. T. Morris, R. Shen, V. Seshan, Q. Mo, A. Heguy, S. B. Baylin, N. Ahuja, A. Viale, J. Massague, L. Norton, L. T. Vahdat, M. E. Moynahan, T. A. Chan. (2011) Breast Cancer Methylomes Establish an Epigenomic Foundation for Metastasis. Science Translational Medicine 3:75, 75ra25-75ra25
   CrossRef

3. 3

   Qin Wang, Hong-Dong Li, Qing-Song Xu, Yi-Zeng Liang. (2011) Noise incorporated subwindow permutation analysis for informative gene selection using support vector machines. The Analyst 136:7, 1456
   CrossRef

4. 4

   Lisa A. Carey. (2010) Through a Glass Darkly: Advances in Understanding Breast Cancer Biology, 2000-2010. Clinical Breast Cancer 10:3, 188-195
   CrossRef

5. 5

   Ping Tang, Kristin A. Skinner, David G. Hicks. (2009) Molecular Classification of Breast Carcinomas by Immunohistochemical Analysis. Diagnostic Molecular Pathology 18:3, 125-132
   CrossRef

6. 6

   Shannon R. Morris, Lisa A. Carey. (2007) Molecular profiling in breast cancer. Reviews in Endocrine and Metabolic Disorders 8:3, 185-198
   CrossRef

7. 7

   Jeffrey. Peppercorn, Charles M. Perou, Lisa A. Carey. 2007. Molecular Subtypes in Breast Cancer Evaluation and Management: Divide and Conquer. , 103-120.
   CrossRef

8. 8

   Evan Matros, Zhigang C. Wang, Gabriela Lodeiro, Alexander Miron, J. Dirk Iglehart, Andrea L. Richardson. (2005) BRCA1 promoter methylation in sporadic breast tumors: relationship to gene expression profiles. Breast Cancer Research and Treatment 91:2, 179-186
   CrossRef

9. 9

   Aurélie Catteau, Joanna R. Morris. (2002) BRCA1 methylation: a significant role in tumour development?. Seminars in Cancer Biology 12:5, 359-371
   CrossRef