Logo of amjpatholAmerican Journal of Pathology For AuthorsAmerican Journal of Pathology SubscribeAmerican Journal of Pathology SearchAmerican Journal of Pathology Current IssueAmerican Journal of Pathology About the JournalAmerican Journal of Pathology
Am J Pathol. 2002 Dec; 161(6): 1991–1996.
PMCID: PMC1850928

Expression of Cytokeratins 17 and 5 Identifies a Group of Breast Carcinomas with Poor Clinical Outcome


While several prognostic factors have been identified in breast carcinoma, the clinical outcome remains hard to predict for individual patients. Better predictive markers are needed to help guide difficult treatment decisions. In a previous study of 78 breast carcinoma specimens, we noted an association between poor clinical outcome and the expression of cytokeratin 17 and/or cytokeratin 5 mRNAs. Here we describe the results of immunohistochemistry studies using monoclonal antibodies against these markers to analyze more than 600 paraffin-embedded breast tumors in tissue microarrays. We found that expression of cytokeratin 17 and/or cytokeratin 5/6 in tumor cells was associated with a poor clinical outcome. Moreover, multivariate analysis showed that in node-negative breast carcinoma, expression of these cytokeratins was a prognostic factor independent of tumor size and tumor grade.

A number of parameters are used to predict the clinical outcome of breast carcinoma, and to guide treatment decisions accordingly. The most important prognostic factors in current use are clinical features such as lymph node (LN) status, tumor size, and tumor grade. The expression level and staining patterns of several proteins are also useful in predicting which tumors will respond to specific therapies; tamoxifen is used to treat only estrogen receptor-positive tumors and herceptin to treat Her2/neu overexpressing tumors. Although these histological prognosticators are undeniably useful, the clinical course of any individual patient with breast carcinoma remains difficult to predict. There is little doubt that there is still considerable molecular heterogeneity within the existing tumor categories. Many studies have continued to identify and explore molecular markers that might help to better stratify patients. In multivariate analysis, however, many of these factors co-vary and are therefore not independently informative. 1,2

With the development of DNA microarray technologies for large-scale analysis of gene expression patterns, a systematic genome-wide search for molecular markers in breast carcinoma has become possible. 3-5 We recently analyzed genomic expression patterns in 78 frozen breast carcinoma specimens using DNA microarrays. 6 This study revealed at least five groups of patients that could be distinguished on the basis of their global gene expression patterns. Two traditional markers, Her2/neu and the estrogen receptor, and a group of cytokeratin genes, were notable for their differential expression among the breast cancer subgroups. Analysis of the survival data in the study showed that two subgroups had a significantly poorer prognosis; one was characterized by elevated expression of Her2/neu, the other was characterized by high levels of expression of genes characteristic of the basal epithelial cells of the normal mammary gland, including the genes that encode cytokeratins 17 and 5. The number of cases in the DNA microarray study was too small to allow multivariate analysis.

To further explore the clinical significance of these findings, we used the recently developed technique of tissue microarrays (TMA) in a retrospective immunohistochemistry evaluation of 611 breast tumor samples. 7,8 This approach allowed us to evaluate protein expression in hundreds of tumors at once, with a single staining reaction on a single glass slide. Using commercially available antibodies against cytokeratins 17 and 5/6, we carried out an immunohistochemical assay for these markers on samples from more than 600 breast carcinomas.

Materials and Methods

Tissue Microarrays

A total of 611 different paraffin-embedded breast carcinoma samples were identified in the files in the Department of Pathology at the University of Basel, Women’s Hospital Rheinfelden, and the Kreiskrankenhaus Lorrach. The specimens were obtained from patients who underwent surgery in the period between 1985 and 1994. The histological parameters for all cases were reviewed by a single pathologist (J.T.) and the histological type and grade was determined for each case according to Elston and Ellis. 9 Follow-up was obtained for 553 cases and ranged from 1 to 151 months, with a mean of 65.9 months. The use of these specimens and data for research purposes was approved by the Ethics Committee of the Basel University Hospital. After surgery, 303 patients received additional systemic therapy (193 patients received hormonal therapy, 57 patients received chemotherapy, and 53 received combined hormonal and chemotherapy). Tissue microarrays were constructed by obtaining 0.6-mm diameter tissue cores from each tumor and placing these cores in a new paraffin block in rows and columns. 7,8,10,11 Each of the 611 cases was sampled twice, once from the center of the tumor, and once from the periphery of the mass. Cores taken from the central area from each case were combined in one array and cores taken from the periphery of the tumor were combined in a second array. A more detailed description of these breast carcinoma tissue microarrays, demonstrating the reliability of these arrays for correlating clinical outcome with expression of a variety of markers (including estrogen receptor, progesterone receptor, Her2/neu [ErbB2], and p53), and including more clinical information, has been published previously. 7,8

Immunohistochemistry and Scoring

Double-staining of normal breast epithelium and tumors in conventional paraffin sections was performed by first staining luminal cells with CAM5.2 using alkaline phosphatase/fast blue staining and subsequent staining of basal cells with CK17 using horseradish peroxidase/DAB staining. Sections of tissue microarrays were stained with monoclonal antibodies specific for cytokeratin 17 (DAKO, Carpenteria, CA, clone E3, dilution 1:10) and cytokeratin 5/6 (Boehringer Mannheim, Indianapolis, IN, dilution 1:10) after antigen retrieval by microwave treatment in citrate buffer. 12 Staining results were scored as follows: 1, invasive tumor cells present in tissue core and no staining seen; 2, invasive tumor cells present and weak staining; 3, invasive tumor cells present with strong staining. Only those cores showing invasive carcinoma were included in the outcome analysis. Cases that either had no tissue present on the array sections or cases in which the material sampled consisted only of fat, fibrosis, normal mammary glands, or in situ carcinoma, were omitted from further analysis. Immunohistochemical staining for cytokeratins either in tissue microarray samples or conventional paraffin sections, often produced only focal staining of tumor cells. To account for the focal expression of CK17 and CK5/6, each of the 611 breast tumors was analyzed four times: with anti-CK17 and anti-CK5/6 antibody on the “central sample” array, and with anti-CK17 and anti-CK5/6 antibody on the “peripheral sample” array. A breast tumor sample was scored as staining positive for the keratins if infiltrating carcinoma in one or more of the cores from that sample reacted with either of the antibodies. To aid in recognizing infiltrating carcinoma in the core samples, sections of each array were also stained with a cytokeratin mix reacting with cytokeratins 8 and 18 (CAM5.2, Becton Dickinson, Franklin Lakes, NJ, dilution 1:20) after antigen unmasking by trypsin digestion to highlight invasive carcinoma cells. Sections of arrays were also stained with antibodies against estrogen receptor (Ventana, Tuscon, AZ), Her2/neu (DAKO), and GATA-3 (Santa Cruz Biotechnology, Santa Cruz, CA). Scoring for Her2/neu was performed following FDA approved manufacturer’s instructions. Nuclear staining was scored for estrogen receptor (ER) and GATA-3 staining, with tumors with less than 5% nuclear staining scored as negative, those with 5 to 20% staining as weak, and those with more than 20% staining scored as strongly positive.

Rabbit Antiserum

A rabbit polyclonal antiserum was raised by injecting three peptides derived from cytokeratin 17 protein sequence. The peptides were synthesized by standard 9-fluorenylmethloxcarbonyl (FMOC) chemistry: peptide 1 KKEPVTTRQVRTIVEE, peptide 2 QDGKVISSREQVHQTTR, peptide 3 SSSIKGSSGLGGGSS. The peptides were conjugated to keyhole limpet hemocyanin (KLH). The peptide-KLH conjugate was injected into two out-bred rabbits. The serum was harvested after the rabbits demonstrated significant anti-peptide titer. Affinity-purified antiserum was obtained by binding the antiserum to an affinity column conjugated with the three peptides; the bound antibodies were eluted with a pH gradient.

Statistical Analysis

Univariate (Kaplan-Meier) analysis of patient survival for subgroups defined on the basis of cytokeratin expression was performed using WinSTAT software (www.winstat.com). Subsequent multivariate analyses were performed using Cox’s proportional hazards model for survival data. 13


Basal Keratin Staining in Normal Breast and Breast Carcinoma

In normal breast, CK17 and CK5/6 stain the basal layer of breast ductal epithelium while keratins 8 and 18 stain luminal cells (Figure 1A) . An examination of whole paraffin sections of breast carcinoma showed that cytokeratin 17 and 5/6 expression in paraffin-embedded tissue, when present, was focal (Figure 1B) with often less than 10% of tumor cells reacting. To further investigate the focal reactivity of the monoclonal antibodies against the basal-type cytokeratins, and as an attempt to improve the reliability of this test, we raised a rabbit antiserum against CK17. This antiserum was tested using a separate tissue microarray with over 300 hundred breast samples (M. van de Rijn, unpublished results). The new antiserum and the monoclonal antibody against CK17 showed very similar reactivity with epithelial cells in the breast sections. Both reagents stained the same fraction of tumor cells suggesting that neither is a significantly better reagent (data not shown). These results suggest that the focal reactivity seen with monoclonal anti-CK17 was not due to weak reactivity of the monoclonal antibody but to the expression of this basal keratin in only a subset of tumor cells at a level detectable by immunohistochemistry.

Figure 1.
A: Normal mammary gland simultaneously stained with CAM5.2 monoclonal antibody [specific for keratins 8 and 18] (blue), and monoclonal anti-cytokeratin 17 (brown). Note that the CAM5.2 antibody specifically stains the lumenal epithelial ...

Analysis of Basal Keratin Expression in Breast Carcinoma Using Tissue Microarrays

Because the individual tumor samples examined in tissue microarray cores were significantly smaller than those in conventional tissue sections, we were concerned that the focal reactivity of basal type cytokeratins might cause positive tumors to be missed. We decided to maximize the chance of detecting basal keratin expression in the breast tumors on the arrays by staining them with monoclonal antibodies directed at both CK5/6 and CK17 and by examining arrays made with cores taken from both central and peripheral areas of the tumors. By combining the results from the “central” array and the “peripheral” array, 532 tumors were available for CK17 analysis, 535 were available for CK5/6 analysis, and 564 were available for either CK17 or CK5/6. The remainder of the tumors represented on the arrays were either lost in transfer during sectioning of the tissue microarray block, or showed no convincing invasive carcinoma in the sections we examined. After combining results from the peripheral and central microarrays, 75 and 63 tumors scored positive (either weak or strongly) for CK17 and CK5/6, respectively. By combining the results from the stains for CK17 and CK5/6, 90 cases (16%) of the 564 tumors examined reacted with either CK17 and/or CK5/6. Of these 90 cases, 51 stained for both antibodies, while the remainder reacted with either of the two antibodies. Follow-up data were available for 505 of the 564 cases on which CK staining data were obtained. The follow-up period ranged from 1 to 151 months with a mean of 65.9 months and a median of 63 months.

Kaplan-Meier survival analysis for all patients with follow-up showed that the absence of detectable cytokeratin 17 or cytokeratin 5 was associated with a significantly better prognosis than the presence of either of these cytokeratins (Figure 2A , P = 0.012). The lymph node status was known in 474 patients. In the group of 229 patients with known lymph node metastases, the expression of CK17 and CK5/6 had no predictive value. In contrast, in the group of 245 patients without lymph node metastases at presentation, CK17 and/or CK5/6 expression was associated with significantly shorter survival (Figure 2B , P = 0.006). The percentage of basal keratin-positive tumors was similar in patients with and without lymph node metastases. Multivariate analysis on all patients taken together showed that the prognostic association of basal cytokeratin expression with poor outcome was not independent from tumor size, LN status, and histological grade. When only patients who presented without lymph node metastases were considered, however, the expression of basal cytokeratins was not only a significant prognostic factor, but its prognostic significance was independent of tumor size, tumor grade, Her2/neu status, ER status, and GATA-3 status.

Figure 2.
A: Kaplan-Meier survival curve showing poor outcome in cytokeratin 17- and/or 5/6-positive tumors (P = 0.012). Clinical follow-up was available for 505 patients (mean, 65.9 months). A: No expression of CK17 or cytokeratin 5/6 in tumor cells. ...

Her2/neu, Estrogen Receptor and GATA-3 Staining on Breast Carcinoma Arrays

Sections of the tissue microarrays made with peripheral cores were stained for estrogen receptor and Her2/neu. As expected, expression of estrogen receptors was associated with a better clinical outcome. This finding was independent of tumor grade, LN status, and size (data not shown). In contrast, Her2/neu expression was associated with a poor prognosis. These results are compatible with previous reports and are similar to those of two prior studies 7,8 conducted using the same breast tumor tissue microarrays used here. We also stained sections of the arrays with antibodies specific for GATA-binding protein 3, a protein that has previously been observed to be co-expressed with the estrogen receptor. 14 Expression of GATA-3 was associated with a good clinical outcome, and highly correlated (χ2 = 720.3 on 9 degrees of freedom) with estrogen receptor expression. Our immunohistochemical staining results for estrogen receptor, GATA-3, and Her2/neu thus confirm results of prior studies, and provide further evidence of the utility and reliability of tissue microarray-based studies. A comparison was made between the basal cytokeratin staining and Her2/neu staining on 304 tumors for which both cytokeratin and Her2/neu staining data were available. As shown in Table 1 , no statistically significant correlation was found between basal cytokeratin staining and Her2/neu expression, a finding further confirmed by our multivariate analysis (above).

Table 1.
Comparison of Her2-Neu and Cytokeration Staining Results


The sequencing of the human genome and the development of massively parallel technologies for analyzing gene expression have opened a new era of molecular diagnostic medicine. DNA microarray analysis now allows the rapid determination of mRNA levels for many thousands of genes in tumor samples while tissue microarrays can be used to analyze large numbers of tumors by immunohistochemical or other staining methods. In a previous study we examined the molecular profiles of breast carcinomas from 42 patients using DNA microarrays representing more than 8000 genes. 15 That study identified a subset of carcinomas that was distinguished from the other tumors by their relatively high level of expression of a specific set of genes characteristic of the basal epithelial cells of normal mammary ducts. Basal keratin expression distinguished this set of tumors from the majority of the tumor samples, which expressed keratin types typically expressed in normal luminal breast epithelial cells. A subsequent study on a larger group of 78 breast tumors showed that the carcinomas in which these “basal” keratins were expressed had a significantly poorer prognosis. 1

Using tissue microarrays made with over 600 breast carcinoma cases, we show here that the presence of basal epithelial cytokeratins in breast carcinoma cells is associated with a poor prognosis. Because of the relatively small size of the tissue samples available for analysis in tissue microarrays, the interpretation of stains that are only focally or heterogeneously reactive can be ambiguous. This was a special concern in our study, because the expression of basal keratins, as detected by immunohistochemical staining, is often very localized, with only scattered tumor cells in a cross-section of the tumor mass showing detectable reactivity. To minimize this ambiguity, we combined results from samples taken from two different areas (central and peripheral) of each tumor, and stained samples from each site with two different antibodies (anti-CK17 and anti-CK5/6). Sixteen percent of the tumors we examined expressed detectable CK17 and/or CK5/6. This frequency was similar to that found in an independent patient population, in which high levels of mRNA for cytokeratins 5 and 17 were detected in 18% of the breast carcinomas studied. In both studies the expression of these markers was associated with poor clinical outcome. 6,15

Because of the large number of patient samples that we were able to analyze using tissue microarrays, we were able to test separately the prognostic significance of these markers in patients with lymph node metastases and those without evident metastases. In patients with metastatic disease to the lymph nodes, the expression of the basal cytokeratins was not associated with a significant difference in clinical outcome. However, in patients without detectable lymph node metastases, expression of “basal” cytokeratins was associated with a poor prognosis independent of tumor size, tumor grade, or immunostain reactivity for ER, Her2/neu, or GATA-3. Taken together with the DNA microarray results, these findings support the idea that anti-cytokeratin antibodies may identify a distinct form of breast cancer, derived from basal cells rather than the luminal cells from which the majority of mammary cancers appear to arise. Further studies are needed to define the cellular origin of each of these groups of tumors.

Two previous immunohistochemistry studies have suggested a correlation between basal cell type markers and poor prognosis. Dairkee and colleagues 16 reported four cases of breast carcinoma with expression of the marker 312C8–1 for myoepithelial cells in the tumor cells and noted a poor clinical outcome. In a study of 51 patients with clinical follow-up, Malzahn et al 17 found a statistically significant association of basal/myoepithelial cell keratin expression with poor prognosis. In contrast to our findings, this association was found to be statistically significant in LN-positive patients but not in LN-negative patients.

The interpretation of immunohistochemical staining results for the basal keratins is complicated by the focal and often weak reactivity of monoclonal antibodies against these proteins, limiting their use in clinical settings. We have therefore begun searching for better immunohistochemical markers for this group of breast cancers. We considered the possibility that alternative antibodies against these cytokeratins might provide better performance. Analysis of more than 300 breast carcinoma samples in a separate array showed that the number of staining cells, the focal staining pattern, and the intensity of staining were similar for a new polyclonal antiserum against CK17 and the commercial monoclonal antibodies. This result suggests that the basal keratins are indeed only focally expressed at a level detectable by immunohistochemistry and that the low numbers of cells stained with antibodies are not due to a weak reactivity of the monoclonal antibodies with the protein. We are currently developing antibodies specific to the products of other genes that were found in the DNA microarray studies to be expressed specifically in the same tumors that expressed basal keratins.

Several studies have now reported that breast cancers expressing basal cytokeratins are not uncommon (>10%), and that they are associated with a poor prognosis. 1 Patients with metastatic breast carcinoma to the axillary lymph nodes are at high risk for recurrence and most receive adjuvant therapy. The situation for “node-negative” patients is less clear; depending on the size and grade of the tumor, the reported recurrence rate varies between five and thirty percent. In patients who present without detectable lymph-node metastases, the clinical decision to give or withhold systemic therapy is therefore a difficult one and hence it is for this group of patients that the need for new prognostic markers is the most acute. The relative size of this group of patients is expected to increase, due to continuing advances in screening and diagnostic techniques that allow detection of smaller breast tumors. Most of these smaller tumors have not metastasized to the “sentinel” lymph node. This group of patients therefore faces a difficult choice among a variety of treatment options, including lumpectomy, mastectomy, chemotherapy, radiation therapy, and hormonal therapy. Expression of cytokeratins 17 and 5/6 appears to define a group of breast tumors with a relatively high mortality rate: clearly a significant consideration in the treatment decisions for node-negative breast carcinoma patients. Whether more aggressive treatment procedures will improve the outcome for these patients, and which of the available options provide the greatest benefit, is an important question for future studies. The challenge in the future will be to develop therapies directed specifically at this molecularly and clinically distinct form of breast cancer.


Address reprint requests to Matt van de Rijn, M.D., Ph.D., L235 Department of Pathology, Stanford University Medical Center, 300 Pasteur Drive, Stanford, CA 94305. E-mail: .ude.drofnats@njirm

Supported by National Institutes of Health grant CA 85129 and the Howard Hughes Medical Institute. P.O.B. is an associate investigator of the Howard Hughes Medical Institute.


1. Henson DE, Fielding LP, Grignon DJ, Page DL, Hammond ME, Nash G, Pettigrew NM, Gorstein F, Hutter RV: College of American Pathologists Conference XXVI on clinical relevance of prognostic markers in solid tumors: Summary Members of the Cancer Committee. Arch Pathol Lab Med 1995, 119:1109-1112 [PubMed]
2. Allred DC, Harvey JM, Berardo M, Clark GM: Prognostic and predictive factors in breast cancer by immunohistochemical analysis. Mod Pathol 1998, 11:155-168 [PubMed]
3. Lockhart DJ, Dong H, Byrne MC, Follettie MT, Gallo MV, Chee MS, Mittmann M, Wang C, Kobayashi M, Horton H, Brown EL: Expression monitoring by hybridization to high-density oligonucleotide arrays. Nat Biotechnol 1996, 14:1675-1680 [PubMed]
4. Chee M, Yang R, Hubbell E, Berno A, Huang XC, Stern D, Winkler J, Lockhart DJ, Morris MS, Fodor SP: Accessing genetic information with high-density DNA arrays. Science 1996, 274:610-614 [PubMed]
5. Cheung VG, Morley M, Aguilar F, Massimi A, Kucherlapati R, Childs G: Making and reading microarrays. Nat Genet 1999, 21:15-19 [PubMed]
6. Sørlie T, Perou CM, Tibshirani R, Aas T, Geisler S, Johnsen H, Eisen MB, van de Rijn M, Jeffrey SS, Quist H, Rees CA, Brown PO, Botstein D, Lønning PE, Børresen-Dale A-L: Gene expression patterns of breast carcinomas distinguish tumor subclasses with potential clinical implications. Proc Natl Acad Sci USA 2001, 98:10869-10874 [PMC free article] [PubMed]
7. Bucher C, Torhorst J, Kononen J, Haas P, Schraml L, Bubendorf L, Zuber M, Kochli OR, Mross F, Dieterich H, Askaa J, Godtfredsen SE, Seelig S, Moch H, Mihatsch M, Kallioniemi O, Sauter G: Prognostic significance of HER-2 amplification and overexpression in breast cancer: methodological comparison of fluorescence in situ hybridization and immunohistochemistry using tissue microarrays of 611 primary breast cancers. J Pathol , in press
8. Torhorst J, Bucher C, Kononen J, Haas P, Zuber M, Kochli OR, Mross F, Dieterich H, Moch H, Mihatsch M, Kallioniemi O, Sauter G: Tissue microarrays for rapid linking of molecular changes to clinical endpoints. Am J Pathol 2001, 159:2249-2256 [PMC free article] [PubMed]
9. Elston CW, Ellis IO: Pathological prognostic factors in breast cancer. I. The value of histological grade in breast cancer: experience from a large study with long-term follow-up. Histopathology 1991, 19:403-410 [PubMed]
10. Kononen J, Bubendorf L, Kallioniemi A, Barlund M, Schraml P, Leighton S, Torhorst J, Mihatsch MJ, Sauter G, Kallioniemi OP: Tissue microarrays for high-throughput molecular profiling of tumor specimens. Nat Med 1998, 4:844-847 [PubMed]
11. Schraml P, Kononen J, Bubendorf L, Moch H, Bissig H, Nocito A, Mihatsch MJ, Kallioniemi OP, Sauter G: Tissue microarrays for gene amplification surveys in many different tumor types. Clin Cancer Res 1999, 5:1966-1975 [PubMed]
12. Natkunam Y, Warnke RA, Montgomery K, Falini M, van de Rijn M: Analysis of MUM1/RF4-protein expression utilizing tissue microarrays and immunohistochemistry. Mod Pathol 2001, 7:686-694 [PubMed]
13. Cox DR: Regression models and life tables. J R Stat Soc 1972, 74:187-220
14. Hoch RV, Thompson DA, Baker RJ, Weigel RJ: GATA-3 is expressed in association with estrogen receptor in breast cancer. Int J Cancer 1999, 84:122-128 [PubMed]
15. Perou CM, Sørlie T, Eisen MB, van de Rijn M, Jeffrey SS, Rees CA, Pollack JR, Ross DT, Johnsen H, Akslen LA, Fluge, Pergamenschikov A, Williams C, Zhu SX, Lønning PE, Børresen-Dale A-L, Brown PO, Botstein D: Molecular portraits of human breast tumors. Nature 2000, 406:747-752 [PubMed]
16. Dairkee SH, Mayall BH, Smith HS, Hackett AJ: Monoclonal marker that predicts early recurrence of breast cancer. Lancet 1987, 1:514 [PubMed]
17. Malzahn K, Mitze M, Thoenes M, Moll R: Biological and prognostic significance of stratified epithelial cytokeratins in infiltrating ductal breast carcinomas. Virchows Arch 1998, 433:119-129 [PubMed]

Articles from The American Journal of Pathology are provided here courtesy of American Society for Investigative Pathology
PubReader format: click here to try


Save items

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...