The first two authors contributed equally to this paper.
Cancer Cell Biology
Stat5a is tyrosine phosphorylated and nuclear localized in a high proportion of human breast cancers
Article first published online: 7 NOV 2003
Copyright © 2003 Wiley-Liss, Inc.
International Journal of Cancer
Volume 108, Issue 5, pages 665–671, 20 February 2004
How to Cite
Cotarla, I., Ren, S., Zhang, Y., Gehan, E., Singh, B. and Furth, P. A. (2004), Stat5a is tyrosine phosphorylated and nuclear localized in a high proportion of human breast cancers. Int. J. Cancer, 108: 665–671. doi: 10.1002/ijc.11619
- Issue published online: 19 DEC 2003
- Article first published online: 7 NOV 2003
- Manuscript Accepted: 9 SEP 2003
- Manuscript Revised: 22 JUL 2003
- Manuscript Received: 3 JUN 2003
- US Army Medical Research. Grant Numbers: DAMD17-01-1-0310, DAMD 17-03-1-0185
- NCI. Grant Number: P30-CA51008
- breast cancer;
- nuclear localization
Signal transducers and activators of transcription (STATs) are latent cytoplasmic transcription factors that are activated and translocated into the nucleus after phosphorylation at a conserved tyrosine residue. Mouse model studies have demonstrated that activated Stat5a acts as a critical survival factor for normal, preneoplastic and malignant mammary epithelial cells. Very limited information is available, however, on the expression, tyrosine phosphorylation status and nuclear localization of Stat5a in human breast cancers. In our study, the pattern of Stat5a cellular localization was analyzed by immunohistochemistry in a series of 83 randomly selected primary human breast adenocarcinomas. Immunoprecipitation/Western blotting and immunohistochemistry assays employing different phospho-specific antibodies verified Stat5a tyrosine phosphorylation status. Stat5a was nuclear localized and tyrosine phosphorylated in 59 of 78 (76%) breast cancers examined; 38 of 78 (49%) demonstrated Stat5a nuclear localization in more than 25% of the breast cancer cells within the adenocarcinomas. Nuclear localized Stat5a was associated positively with increased levels of histologic differentiation (p = 0.03). A statistically significant positive association with p27 nuclear localization also was identified (p = 0.05). No relationship was found between nuclear localized Stat5a and menopausal status, tumor size, ploidy, percentage of cells in S-phase, lymph node metastases, ER, ErbB2, nuclear localized p21 or nuclear localized Stat5b/Stat3. As its role in human breast cancer progression and response to therapy is defined, Stat5a could become a new molecular target for breast cancer therapy. © 2003 Wiley-Liss, Inc.
Discovered originally as a mammary gland factor (MGF),1 signal transducer and activator of transcription (Stat) 5a plays the most prominent role of all STAT family members during mouse mammary gland development.2 Activated Stat5a is required for lactation and contributes to the survival and terminal differentiation of mammary epithelial cells during pregnancy and lactation.3 Basal activation of Stat5a in virgin mice and in normal non-pregnant human breast tissues may contribute to early differentiation and survival of mammary epithelial cells outside of pregnancy.4
Like other family members, Stat5a is activated through phosphorylation at an invariant tyrosine residue in the cytoplasm and then translocated into the nucleus where it can act as a transcription factor.5, 6, 7 The cytokine-inducible SH2 protein (CIS)/suppressor of cytokine signaling (SOCS) family in concert with protein inhibitor of activated Stat (PIAS) family members and various protein tyrosine phosphatases negatively regulate the activity of STATs.8, 9 Serine phosphorylation of STATs contributes to the fine modulation of their transcriptional activity.7, 9, 10 In normal mammary epithelial cells, the activation of both Stat5a and its closely related family member arising from gene duplication, Stat5b,11 is governed by the prolactin-signaling pathway.9, 12 Activated Stat5a/b also act to promote the survival and differentiation of hematopoietic progenitor cells of both myeloid and lymphoid lineages.13, 14, 15, 16 Aberrant activation of Stat5 through BCR-ABL has been identified as a key oncogenic event in chronic myelogenous leukemia17, 18 and impairs the apoptotic response to chemotherapeutic agents and DNA damage in various hematopoietic cancer cell lines.19, 20, 21
Stat3 is another family member located on the same chromosome22 that is expressed and activated in human breast cancers and breast cancer cell lines.7, 23, 24, 25 Interestingly, activated Stat3 induces physiological apoptosis of mammary epithelial cells in the mouse mammary gland during involution,26, 27 but constitutive activation of Stat3 promotes cell survival and growth in breast cancer cell lines.7, 28 Nevertheless, activated Stat3 has been correlated recently with better prognosis in node-negative breast cancers.25
Experiments in mouse models of mammary cancer progression have demonstrated that activated Stat5a is an important survival factor for both preneoplastic and malignant mammary epithelial cells, which promotes cancer progression.29, 30 Only limited information is available, however, on the expression, tyrosine phosphorylation status and nuclear localization of Stat5a in primary human breast cancers. No immunohistochemical studies of Stat5a in breast cancer have been reported to date. One group of investigators was able to detect by electrophoretic mobility shift assays (EMSAs) high Stat5a DNA binding activity only in 1 of 63 human breast cancer samples, but the authors comment that the DNA binding activity was either low or below the limit of the sensitivity of their detection method in the other tumor samples. They were not able to detect tyrosine phosphorylation of Stat5 by Western blotting.24 In contrast, activation of Stat5a/b by both prolactin and epidermal growth factor (EGF)-related pathways clearly can be found in human breast cancer cell lines.31, 32, 33, 34, 35 Moreover, adenoviral-mediated gene delivery of dominant-negative Stat5 induced apoptosis in T47D breast cancer cells.36
As its receptor lacks intrinsic tyrosine kinase activity, prolactin activates Stat5a/b indirectly through recruitment of a tyrosine kinase (i.e., Janus kinase 2 [Jak2]) upon receptor dimerization.6, 9 Both endocrine and autocrine/paracrine sources of prolactin can contribute to Stat5a activation. Although the inhibition of pituitary secretion of prolactin by bromocriptine is effective as a prophylactic, as well as a curable means against rodent mammary carcinomas,37 this strategy failed in several clinical trials.9 Because the majority of human breast cancers and various breast cancer cell lines are able to synthesize prolactin,38, 39 it is possible that local production of prolactin and paracrine/autocrine activity contribute to the failure of this therapeutic approach.
In addition, members of the ErbB (epidermal growth factor receptor) family possess intrinsic tyrosine kinase activity and therefore could activate Stat5a/b directly. Epidermal growth factor (EGF), TGFα and Neu differentiation factor (NDF) can all activate Stat5a/b through various combinations between EGFR, ErbB2 and ErbB4 receptors.29, 35, 40, 41 This activation, however, is dependent on the recruitment of Src to the EGFR/Stat5 complex and results in phosphorylation of additional tyrosine residues, beside the invariant critical residue required for activation.40 Notably, Src activation in COS-1 or HeLa cells results in tyrosine phosphorylation of both Stat5a and Stat5b, but nuclear translocation of only Stat5b, as opposed to prolactin that induces the phosphorylation and nuclear translocation of both Stat5a and Stat5b.42
In our study, Stat5a expression, nuclear localization, and tyrosine phosphorylation status were evaluated in a large series of primary breast adenocarcinomas through a combination of immunohistochemical, immunoprecipitation and Western blotting techniques employing different antibodies. The relationships between nuclear localized Stat5a and established prognostic and clinical markers, including tumor size, menopausal status, tumor pathology, lymph node metastases, cell cycle parameters, ER, ErbB2, p27 and p21 expression,43, 44, 45, 46 as well as Stat5b and Stat3 expression were investigated.
MATERIAL AND METHODS
Tumor specimens and clinical data
Eighty-three randomly selected primary breast adenocarcinomas were obtained from the Lombardi Cancer Center Histopathology and Tissue Shared Resource under Georgetown University IRB approval. All patient identifiers were removed, but clinical information recorded at the time of specimen acquisition was available (Table I). DNA flow cytometry analysis was carried out using paraffin-embedded tissues as described previously;49 the median cut-off points for classifying S-phase fractions were 5.9% for diploid DNA tumors and 11% for DNA aneuploid tumors.47, 48 Data was not available for either the response to treatment or course of disease. None of the patients were known to have distant metastases at the time the samples were acquired.
|Age at the time of surgery (yrs.)||83|
|Mean 55 ± 12|
|Premenopause (≤49 years||28||34|
|Postmenopause (≥50 years||55||66|
|Tumor size (cm)||74|
|Differentiation status (Elston grade)||73|
|S-phase fraction values2||47|
|Lymph node metastases||49|
Immunohistochemistry and scoring system
Immunohistochemistry (IHC) was carried out on individual tumor samples (Stat5a, Stat5b, Stat3, p21 and p27: n = 30; phospho-Stat5: n = 8) and a breast cancer tissue array (Stat5a: n = 53), as well as normal non-pregnant breast tissue samples from reduction mammoplasty (Stat5a: n = 7; phospho-Stat5: n = 2). The breast cancer tissue array was developed by one of the investigators (B.S.) and validated previously.50 Five-micrometer sections were cut from paraffin-embedded specimens or the tissue array. The first section was stained with hematoxylin and eosin (H&E) and read by a board certified pathologist (B.S.) to confirm the original pathological diagnosis before any immunohistochemical reactions were carried out. Five of the 53 sections on the array contained no tumor and were excluded from the analysis. Both the array and the individual sections were deparaffinized in xylene and rehydrated through a graded series of ethanol and water washes. The sections were subjected to antigen retrieval by either boiling in sodium citrate buffer (pH 6) for 1 min (for Stat5a)/10 min (for Stat5b, Stat3, p21 and p27) or by using the DAKO high pH retrieval solution (for phospho-Stat5). Endogenous peroxidase activity was quenched by incubating the sections with 3% H2O2 for 10 min at room temperature. Immunohistochemical staining was carried out with a 1:125 dilution of anti-Stat5a antibody (sc-1081; Santa Cruz Biotechnology, Santa Cruz, CA), 1:40 dilution of anti-phospho-Stat5 antibody (71-6900; Zymed Laboratories, San Francisco, CA) or 1:25 dilution of anti-phospho-Stat5 antibody (9351; Cell Signaling, Beverly, MA), 1:50 dilution of anti-Stat5b antibody (71-2500; Zymed Laboratories), 1:35 dilution of anti-Stat3 antibody (sc-483, Santa Cruz Biotechnology), 1:50 dilution of anti-p21 antibody (sc-6246; Santa Cruz Biotechnology) and 1:50 dilution of anti-p27 antibody (sc-1641; Santa Cruz Biotechnology) using the HistoMouse-SP Kit (Zymed Laboratories) according to the manufacturer's protocol. All sections were counterstained with hematoxylin. Both phospho-specific Stat5 antibodies recognize Stat5a as well as Stat5b, when phosphorylated at the tyrosine residue required for activation.6 Staining with non-immune serum was used as a negative control. The same system was used for all IHC reactions to score the extent of nuclear localization. The staining intensity of the nuclei was not considered to be a relevant interpretive parameter. The samples were ranked into 5 different groups based on the percentage of breast adenocarcinoma cells exhibiting nuclear staining: no nuclear staining, Group 0; 1–25% breast cancer nuclei staining, Group 1; 26–50% breast cancer nuclei staining, Group 2; 51–75% breast cancer nuclei staining, Group 3; 76–100% breast cancer nuclei staining, Group 4. Two independent readers, one board certified pathologist (B.S.) and one research pathologist (P.A.F.), interpreted the samples in a double blind manner. The independently derived scores for each sample were combined to derive a consensus score for each sample.
Immunoprecipitation and Western blot analysis
Frozen breast cancer samples (n = 12) and normal non-pregnant breast tissue samples from reduction mammoplasty (n = 5) were obtained from the same Lombardi Cancer Center Histopathology and Tissue Shared Resource that the formalin-fixed samples were obtained from under Georgetown University IRB approval after all patient identifiers were removed. All tumor samples contained >90% breast cancer cells. Proteins were extracted from the samples and Western blot (WB) and immunoprecipitation (IP) followed by WB assays carried out as described previously.51 The following antibodies were used for IP: anti-Stat5a (sc-1081, Santa Cruz Biotechnology,), anti-Stat5b (13-5300, Zymed Laboratories,) and anti-Stat3 (sc-482, Santa Cruz Biotechnology,). Western blot analyses were carried out with the same antibodies: anti-Stat5a 1:3,000 dilution, anti-Stat5b 1:500 dilution, anti-Stat3 1:400 dilution and anti-phospho-tyrosine antibody (05-321; Upstate Biotechnology, Lake Placid, NY) 1:4,000 dilution. Horseradish peroxidase-conjugated goat anti-rabbit antibody (sc-2004; Santa Cruz Biotechnology) 1:7,500 dilution or anti-mouse antibody (sc-2005; Santa Cruz Biotechnology) 1:10,000 dilution were used as secondary antibodies, depending on the source of the primary antibody. Lactating mouse mammary gland tissue samples were used as positive controls for total Stat5a, Stat5b and Stat3, as well as tyrosine phosphorylated Stat5a/Stat5b.26 Involution Day 2 mouse mammary gland tissue samples served as positive controls for tyrosine phosphorylated Stat3.26 The mouse mammary epithelial HC11 cell system was utilized as described previously52 to provide both negative and positive controls for tyrosine phosphorylated Stat5a.
Stat5a nuclear staining status was classified as negative/low (Groups 0, 1) or positive/high (Groups 2–4) before statistical evaluation. For the purposes of statistical analysis, tumor samples that had 0–25% cells exhibiting nuclear Stat5a (Groups 0, 1) were combined because Stat5a transcriptional activity53 and its effects on apoptosis30 are dose-dependent and therefore, the biologically significant activity of Stat5a was not anticipated to be present in the cancers with low percentages of cells demonstrating nuclear localization. The clinical data was classified according to established prognostic criteria (Table I). Different statistical tests were carried out depending on the type of clinical data. For categorized data (e.g., menopausal status, lymph node status, etc.), the Fisher's exact test was employed to examine whether or not the percentage of positive/high samples differed according to clinical category. For data that was divided into ordered groups (e.g., tumor size, differentiation status, etc.), the Spearman's rank correlation coefficients were calculated and the tests of the correlation equal to zero carried out. The Spearman's rank correlation coefficient ranked the clinical data (e.g., tumor size ≤2 cm = 1, 2–5 cm = 2, >5 cm = 3) and calculated the rank correlation between those samples having negative/low or positive/high Stat5a nuclear staining and the corresponding clinical parameter. Statistical analyses were carried out for each clinical marker using only the Stat5a stained samples with available data for that particular clinical marker (Table I). In consequence, no special methods were needed to account for missing data points.
Stat5a was nuclear localized and tyrosine phosphorylated in over 75% of human breast cancers
Stat5a was present and nuclear localized in 59 of 78 (76%) invasive breast adenocarcinomas examined. The percentage of breast cancer cells exhibiting nuclear localized Stat5a ranged from 0–100% among the cancers (Fig. 1a, 2). Thirty-eight of 78 (49%) breast cancers demonstrated relatively high levels of Stat5a nuclear positivity (Groups 2–4). To verify that nuclear localized Stat5a was phosphorylated, an immunohistochemical analysis employing 2 different commercially available phospho-specific Stat5 antibodies was carried out in a subset of the breast cancers. All samples demonstrating nuclear localized Stat5a by the Stat5a-specific antibody staining also demonstrated nuclear localization of Stat5 detected by the 2 phospho-specific Stat5 antibodies (Fig. 1b). Stat5a was nuclear localized and tyrosine phosphorylated in the surrounding normal adjacent breast epithelium to the same extent as in normal non-pregnant breast tissue samples obtained from reduction mammoplasty (Fig. 1c and data not shown). Immunoprecipitation/Western blotting experiments on frozen breast cancer tissues and normal non-pregnant human breast tissues confirmed Stat5a tyrosine phosphorylation in the samples (Fig. 4b).
Nuclear localization of Stat5a was associated positively with increased levels of histologic differentiation
Statistical analyses were carried out to characterize the relationship between nuclear localized Stat5a and available clinical markers. A statistically significant positive association between nuclear localized Stat5a and increased levels of histologic differentiation was identified by Spearman's rank order correlation (R = 0.24, p = 0.03; Table II). As shown in Table II, the majority of breast cancers with increased levels of differentiation (moderate and well differentiated cancers) demonstrated Stat5a nuclear positivity (59%; 30/51), whereas poorly differentiated breast cancers were mainly Stat5a negative (73%; 16/22) (Fisher's exact test, p = 0.02). Notably, the relationship was not dependent upon a chosen cut point given that similar results were obtained when Stat5a nuclear staining status was classified as positive/high vs. negative/low using the 50% or 75% cut points (Spearman's rank order correlation, p < 0.05; data not shown).
|Stat5a nuclear staining||Differentiation status||Total samples|
|Negative/low (groups 0,1)||16||13||8||37|
|Positive/high (groups 2,3,4)||6||19||11||36|
No relationship was found between the presence and absence of nuclear localized Stat5a and menopausal status, ER status, presence or absence of lymph node metastases at the time of surgery, ploidy or percentage of cells in S-phase by Fisher's exact test (p > 0.05, data not shown). No relationship was found with tumor size or ErbB2 expression by Spearman's rank order correlation (p > 0.05, data not shown).
Nuclear localized Stat5a was associated positively with nuclear localized p27, but not p21
Because we identified a statistically significant positive association between nuclear localization of Stat5a and differentiation status, and p21 and p27 have been linked previously to differentiation of human breast cancers,44, 45, 46 we tested whether or not there was any evidence to support an association between nuclear localization of Stat5a and nuclear localization of either of these two cyclin-dependent kinase inhibitors. The percentage of breast cancer cells demonstrating nuclear localization of p21 and p27 was determined after immunohistochemical detection in a subset of 30 samples selected from the 83 breast cancers (Fig. 3). There was no significant difference in the distribution of differentiation status between the 30 samples and the other 53 samples (Kruskal-Wallis test; p = 0.34). A statistically significant positive association between nuclear localized Stat5a and nuclear localized p27 was identified by Spearman's rank order correlation (Table III, Fig. 3). The relationship was not dependent upon a chosen cut point because similar results were obtained when Stat5a nuclear staining status was classified as positive/high vs. negative/low using the 50% or 75% cut points (Spearman's rank order correlation, p < 0.05; data not shown). No association between nuclear localized Stat5a and presence or absence of nuclear localized p21 was found (Spearman's rank order correlation, p > 0.05; data not shown).
|Stat5a nuclear staining||p27 nuclear staining (% breast cancer cells)||Total samples|
|Negative/low (groups 0,1)||5||2||3||4||1||15|
|Positive/high (groups 2,3,4)||1||1||4||2||4||12|
Stat5a, Stat5b and Stat3 were expressed, tyrosine phosphorylated and nuclear localized coordinately in some but not all human breast cancers
Expression levels, tyrosine phosphorylation status and nuclear localization of Stat5b and Stat3, 2 other STAT family members implicated in breast cancer,7, 25, 30, 35 were examined and compared to Stat5a. Western blot analysis indicated that all three STAT family members could be expressed concomitantly in the same cancer (Fig. 4a). Immunoprecipitation/Western blot analysis (Fig. 4b–d) and immunohistochemical studies (data not shown) demonstrated that Stat5a, Stat5b and Stat3 could all be tyrosine phosphorylated and nuclear localized simultaneously in the same tumor samples. No direct correlations were found, however, between Stat5a nuclear localization and either Stat5b or Stat3 nuclear localization by Spearman's rank order correlation (p > 0.05, data not shown).
Stat5a was nuclear localized in 76% of the human breast cancers examined. Forty-nine percent of the breast cancer specimens demonstrated relatively high percentages of cancer cells exhibiting nuclear localized Stat5a. Tyrosine phosphorylation of nuclear localized Stat5a was confirmed by the combination of immunohistochemistry and immunoprecipitation followed by Western blotting assays employing various phospho-specific antibodies. Statistically significant positive associations between nuclear localized Stat5a and increased levels of differentiation as well as p27 nuclear localization were identified.
The fact that a previous study failed to detect tyrosine phosphorylation of Stat5 in primary human breast cancers24 may be secondary to important differences in the sensitivity of the specific techniques and antibodies employed.
The positive association with nuclear localized p27, a recently identified marker of differentiation for breast cancers,45 supported the positive association between nuclear localized and tyrosine phosphorylated Stat5a and increased levels of histologic differentiation. It is noteworthy that this is consistent with what is known about the role of Stat5a in normal mammary gland development. Activated Stat5a is required for the functional differentiation of mammary epithelial cells and lactogenesis during pregnancy and lactation.3 Moreover, activated Stat5a promotes the differentiation of various hematopoietic progenitor cells of both myeloid and lymphoid lineages13, 14, 16 at least in part through p21 transcription activation.13 Nonetheless, the mechanisms by which activated Stat5a controls the differentiation of either normal or malignant mammary epithelial cells remain under investigation.
The next fundamental question will be to define the role of nuclear localized Stat5a in human breast cancer progression and response to therapy. Although mouse model studies of mammary cancer and investigations of Stat5a activity in hematopoietic cancer cells indicate that nuclear localized Stat5a promotes survival of cancer cells and disease progression, our study found a positive association between nuclear localized Stat5a and increased levels of histologic differentiation. It is plausible that Stat5a nuclear presence and activity in breast cancers is independent from but parallels that found for ERα, a factor that is expressed in normal breast epithelium, contributes to breast cancer progression, and is associated with a better prognosis.43, 54
The role of nuclear localized Stat5a as an independent prognostic factor for breast cancer patients, as well as a marker of response to various current therapeutic regimens can be addressed best using large sets of human breast cancer samples in which outcome data is available, controlling for differentiation status and then examining the survival data and response to treatment data, respectively in patients with and without nuclear localized Stat5a.
In summary, the data presented in our study demonstrates that Stat5a was expressed, tyrosine phosphorylated and nuclear localized in a significant proportion of breast cancers. Nuclear localization of Stat5a was correlated positively with increased levels of histologic differentiation and p27 nuclear localization.
This work was supported by US Army Medical Research grants DAMD17-01-1-0310 (P.A.F.) and DAMD 17-03-1-0185 (I.C.). The authors would like to acknowledge the assistance of the Lombardi Cancer Center Histopathology and Tissue Shared Resource and Lombardi Cancer Center Biostatistics Shared Resource, which are supported in part by a Cancer Center Support Grant from the NCI (P30-CA51008).
- 43Breast Cancer. In: LenhardREJr, OsteenRT, GanslerT. Clinical oncology. Atlanta: American Cancer Society, 2001. 251–68..
- 47Prognostic value of DNA cytometry in 281 premenopausal patients with lymph node negative breast carcinoma randomized in a control trial: multivariate analysis with Ki-67 index, mitotic count, and microvessel density. Cancer 2000; 89: 1748–57., , , , , , , , , .
- 50Representation of the heterogeneity of breast cancer and hormone receptor expression in a tissue microarray. Mod Pathol 2000; 13: 225A., , , , .