BARD1: An independent predictor of survival in ... - Wiley Online Library

IJC International Journal of Cancer

BARD1: an independent predictor of survival in non-small cell lung cancer Yong-Qiang Zhang1,2*, Andrea Bianco3*, Alvin M. Malkinson4, Vera Piera Leoni5, Gianni Frau6, Nicolina De Rosa3, Pierre-Alain Andre´1, Renato Versace7, Michel Boulvain8, Geoffrey J. Laurent9, Luigi Atzori5 and Irmgard Irminger-Finger1,2** 1

Molecular Gynecology and Obstetrics Laboratory, Department of Gynecology and Obstetrics, University Hospitals Geneva, Geneva, Switzerland Department of Genetic and Laboratory Medicine, University Hospitals Geneva, Geneva, Switzerland 3 Chair of Respiratory Diseases, Department of Health Sciences, University of Molise, Campobasso, Italy 4 Department of Pharmacological Science, University of Colorado Denver, Aurora, CO 5 Department of Toxicology, Oncology and Molecular Pathology Unit, University of Cagliari, Cagliari, Italy 6 Pathology Unit, Binaghi Hospital, Cagliari, Italy 7 Thoracic Surgery Division, Binaghi Hospital, Cagliari, Italy 8 Department of Gynecology and Obstetrics, University Hospitals Geneva, Geneva, Switzerland 9 Centre for Respiratory Research, University College London, London, United Kingdom

BRCA1 mRNA overexpression is correlated with poor survival in NSCLC. However, BRCA1 functions depend on the interaction with BARD1 for its stability, nuclear localization and ubiquitin ligase activity. Expression of alternatively spliced BARD1 isoforms that lack the BRCA1-interaction domain was found upregulated and correlated with poor prognosis in breast and ovarian cancer. These BARD1 isoforms are essential for proliferation of cancer cells in vitro. We investigated whether BARD1 isoforms are expressed in NSCLC. While in lung tissues from healthy controls BARD1 expression was undetectable on the mRNA level and protein level, we found two novel isoforms in addition to previously identified mRNAs expressed in all NSCLC samples tested. Furthermore, the pattern of BARD1 isoform expression was similar in tumor and morphologically normal peritumor tissues, and only one novel isoform p was specifically upregulated in tumors. Immunohistochemistry revealed that all 100 NSCLC cases tested expressed isoform-specific BARD1 epitopes, while BARD1 expression was undetectable in biopsies from healthy controls. Statistical analysis showed that the expression of epitopes PVC and WFS, present on isoform p, or epitope WFS alone, expressed on isoforms p, j and b, were significantly correlated with decreased patient survival. These findings were corroborated in a mouse model of chemically induced lung cancer. Immunostaining of mouse tumors showed that BARD1 epitopes PVC and WFS were specifically upregulated in invasive, but not in confined lung tumors. Thus, BARD1 isoforms might be involved in tumor initiation and invasive progression and might represent a novel prognostic marker for NSCLC.

Key words: NSCLC, BRCA1, BARD1, isoforms, differential splicing, biomarker Additional Supporting Information may be found in the online version of this article. **Irmgard Irminger-Finger is a shareholder of BARD1AG, a company dedicated to developing tests for cancer detection. *Y.-Q.Z and A.B. contributed equally to this work Grant sponsor: SNRF; Grant number: 31003A-122353; Grant sponsor: Oncosuisse; Grant number: KLS-01962-10-2006; Grant sponsor: USPHS; Grant number: CA132557 DOI: 10.1002/ijc.26346 History: Received 9 Mar 2011; Accepted 12 Jul 2011; Online 3 Aug 2011 Yong-Qiang Zhang’s present address is: Beijing Hospital, Beijing, China Correspondence to: Dr. Irmgard Irminger-Finger, Molecular Gynecology and Obstetrics Laboratory, Department of Gynecology and Obstetrics, University Hospitals Geneva, 30 Bld de la Cluse, 1211 Geneva, Tel: [þ41-22-3824-327], E-mail: irmgard.irminger@ hcuge.ch

C 2011 UICC Int. J. Cancer: 131, 83–94 (2012) V

Lung cancer is the leading cause of cancer death worldwide due to detection at an advanced stage and to inefficiency of resistance to treatment methods, other than surgery. Thus, insights into the etiology of lung cancer and its progression are urgently needed. Protein, RNA and microRNA profiles from lung tumors have been screened for potential drivers of lung cancer.1–4 TP53 is the most frequently deleted or mutated gene in lung cancer, followed by components of the p53-ARF pathway.5 Novel predictive and prognostic molecular markers in nonsmall-cell lung cancer (NSCLC) include DNA damage repair genes, such as ERCC1, RRM1 and BRCA1.6 Upregulation of mRNA of the breast cancer predisposition gene BRCA1, correlated with reduced survival of NSCLC patients, is a predictive marker for response to treatment of NSCLC.7,8 However, the mechanism behind BRCA1 mRNA expression and cancer outcome is not clear. BRCA1 is a tumor suppressor and expressed in many proliferating tissues, but its tumor suppressor functions depend on its interaction with the BRCA1-associated RING Domain 1 (BARD1) protein for stability, nuclear localization and for enhancing its ubiquitin

Cancer Cell Biology

2

Cancer Cell Biology

84

ligase activity,9,10 and the BRCA1-BARD1 E3 ubiquitin ligase controls the stability/turnover of key target proteins through ubiquitination.11,12 BARD1 also binds to p53 and is required for p53 signaling towards apoptosis,13,14 thus, playing a dual role in tumor suppression, with BRCA1 and p53.10 BARD1 and BRCA1 have essential functions in mitosis, in spindle formation and at cytokinesis,15,16 and hence BARD1 expression is transcriptionally and post-translationally regulated during mitosis.17,18 While BRCA1 is often mutated in breast and ovarian cancer resulting in BRCA1 deficiency, upregulated expression of aberrant forms of BARD1 was found correlated with poor prognosis in breast and ovarian cancer.19,20 The cancer-associated BARD1 isoforms lacked the regions for BRCA1 and/or p53 binding.15,19 BARD1 isoforms have pro-proliferative functions and their specific repression in cells lacking full length (FL) BARD1 leads to growth arrest of cancer cells in vitro, suggesting that BARD1 isoforms act as oncogenes.15,19 Two isoforms, BARD1b and BARD1d, displayed dominant negative functions by antagonizing the BRCA1-BARD1 E3 ligase activity.15,21 In particular, BARD1b binds the Aurora B kinase and prevents its degradation.15 Recently, SNPs in BARD1 50 untranslated and introns regions were found associated with aggressive neuroblastomas,22 suggesting that aberrant BARD1 expression might be associated with other than gynecological cancers. Lung cancer risk has been linked to hormonal factors and/or gender-specific susceptibility, as female smokers and non-smokers are more prone to develop lung cancer than males.23 Estrogen-dependent signaling might play a role in increased susceptibility, as the estrogen receptor (ER) is expressed in lung cancers.24 Interestingly, BRCA1 and BARD1 expression can be induced by estrogen via ER-a. Furthermore, ER-a stability and turnover is controlled by the BRCA1-BARD1 ubiquitin ligase, but also by cancer-associated isoform BARD1d which acts antagonistically and stabilizes ER-a.21,25 We therefore reasoned that BARD1 might also play a role in lung cancer. We performed RT-PCR and immunohistochemistry to determine structure and relative abundance of BARD1 mRNA and BARD1 epitope expression profiles in 100 NSCLC patients. Finally, we used an animal model of induced lung cancer to monitor changes of BARD1 epitope expression during the formation and progression of tumors to corroborate the conclusions drawn from the analysis of the patient data.

Material and Methods Clinical data

Totally 100 NSCLC tumors were collected at two centers (Naples, Cagliari) (Table 1). All patients were informed and approvals obtained from the local ethics committees. In addition, eight cases with benign lung diseases (five males and three females, 24–66 (median age 38) years), including pulmonary emphysema and pulmonary tuberculosis or carcinoid dysplasia, were used as control samples.

BARD1 predictor of survival in NSCLC

Follow-up records were available for 65 patients, and the median follow-up was 29 months (range 1–95). These included 17 patients treated with surgery only, four with chemotherapy post-surgery, one with chemotherapy and radiotherapy post-surgery, seven patients with chemotherapy and radiotherapy, four with chemotherapy and one with radiotherapy only and 14 were without treatment till last follow-up or death. At last follow-up (March 2010), 41 patients had died (35 from Naples, 6 from the Caliari) and 24 were still alive (13 from Naples and 11 from Cagliari). Disease-free survival (DFS) was calculated as DFS till relapse or death (DFS < overall survival (OS)) or DFS without relapse (DFS ¼ OS). RNA/reverse transcription-PCR

RNA was isolated from frozen tissue sections using Trizol extraction method.19 Concentrations were measured and D260/D280 ratios were at least 1.8. Reverse transcription (RT) and PCR conditions and primers were as previously reported.19 ERa was amplified using annealing temperature 56 C and extension times 1 min with primers 50 -ACAAGCGCCAGAGAGATGAT-30 and 50 -GATGTGGGAGAGGATGAGGA-30 . Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was amplified as internal control with primers 50 -AGCCACATC GCTCAGACACC-30 and 50 -GTATCTAGCGCCAGCATCG-30 . PCR products were analyzed on agarose/TBE (Tris/Borate/ EDTA) and visualized under UV light. The QIAEX II kit (Qiagen, Hombrechtikon, Switzerland) was used for DNA purification. Direct sequencing of amplicons was performed using internal BARD1 primers. Real-time PCR

RNA (1.5 lg) were reverse transcribed using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA). Real-Time PCR was performed in a 7300 Real-Time PCR System (Applied Biosystems). The amplification mixture contained 20 ng of cDNA, 5 ll 2 TaqMan Gene Expression PCR Master mix and 0.5 ll of specific 20 TaqMan Gene Expression Assay. Real-time PCR was performed with primers either amplifying FL BARD1 and isoforms BARD1a and p (CAATGAGCTGTCAGATTTGAAAGAA, CGAGGGCTAAA CCACATTTTAATT, and TTTAATGATGCAGGAAACAA), only isoform a (TGCTCGCGTTGTAATTGTGTAAGT, CA TGCTGTCCAGTTGTCTATTTATCTT, and TCCAGTGTGTTACACCC), or only p (GACTCCTGAGAAGGGCGACAT, TCTTTAACATTTGGATCACTTCCATT, and CCTTCTGTTG AATACCTTTT). All samples were analyzed in triplicate and b-actin gene expression was used as housekeeping control (Applied Biosystems). The relative gene expression was calculated according the 2DDCT method.26 Immunohistochemistry

Tumor and peri-tumor samples were fixed in paraffin and processed for cytopathology and immunohistochemistry after standard protocols. Immunohistochemistry was performed on C 2011 UICC Int. J. Cancer: 131, 83–94 (2012) V

Zhang et al.

85

Table 1. characteristics Napoli

Cases Gender

60 Male Female

Age

Range Median

Normal (peri-tumor)

Grade

Total 100

54

20

74

6

20

26

34–77

33–77

33–77

63

60

62

0

20

20

40

100

Adenocarcinoma

26

40

66

Squamous cell carcinoma

21

0

21

Large cell carcinoma

9

0

9

Adenosquamous carcinoma

4

0

4

2

10

12

Moderately differentiated

Well-differentiated

20

12

32

Poorly differentiated

26

18

44

1

0

1

Undifferentiated

Stage

40

60

Tumor Histology

Cagliari

Unspecified

11

0

11

IA

14

18

32

IB

19

8

27

IIA

1

5

6

IIB

9

3

12

IIIA

11

2

13

IIIB

4

1

5

IV

1

3

4

Unknown

1

0

1

adjacent 5 lm sections as reported19,20,27 with previously characterized antibodies against different regions of BARD114,15,19,20,27–30 and against BRCA1. The BRCA1 antibody and BARD1 antibodies N19 and C20 were purchased from Santa Cruz (Santa Cruz Biotechnology, Santa Cruz, CA). Antibodies PVC and WSF were generated, characterized and applied for immunohistochemistry on various tissues before.13–15,27,28,30 horseradish peroxidase (HRP)-conjugated secondary antibodies (goat anti-rabbit or rabbit anti-goat) were applied in 1:100 dilutions at room temperature for 1 hr. 30 ,30 -Diaminobenzidine (DAB) staining was for 2–15 min at room temperature. Slides were counterstained with hematoxylin before dehydration and mounting. To ascertain sensitivity and specificity, primary antibodies were omitted on control sections. BARD1 or BRCA1 expression were quantified applying Histo-score method.31 Three different fields of each tumor/staining were scored semiquantitatively from 1 (no staining) to 4 (strong staining) and percentage of stained cells in each field. The average of the different regions was calculated to define the score of each sample. Three persons performed the scoring independently and without any knowledge of clinical data. C 2011 UICC Int. J. Cancer: 131, 83–94 (2012) V

For immunohistochemistry performed on mice with induced lung cancer, lung tissues were dissected from treated and control mice, and sections of lung and tumor tissues were analyzed by immunohistochemistry using antibodies PVC, WFS and C20. Analysis and quantification was as described for human samples. Mouse model of lung cancer

Lung tumors were chemically induced in BALB/c mice.32 Mice were sacrificed after 16, 24 and 32 weeks of treatment. Lung tissues and tumors were dissected and processed for immunochemistry. Tissues from three mice were analyzed for each time point. Statistical analysis

The Spearman’s correlation coefficient was used to assess the correlation between BARD1 and BRCA1 epitope expression. The Fisher exact test was used to compare the percentage of positivity in tumor versus peri-tumor tissues and for correlation of BARD1 expression with clinical variables. Survival differences were estimated with Kaplan–Meier method and compared by the log-rank test. Multivariate survival analysis

Cancer Cell Biology

Samples

86

was performed using the Cox proportional hazards model. For all calculations, the tests were two-sided, and a value of p < 0.05 was considered statistically significant. Analyses were performed using Statistical Package for the Social Sciences (SPSS) for Windows version 15 (SPSS Inc, Chicago, IL).

Results

Cancer Cell Biology

BARD1 mRNA expression in NSCLC

To determine the abundance and structure of BARD1 isoforms in NSCLC, we performed RT-PCR with primers amplifying the entire BARD1 coding region on RNA from tumors of 20 female and male patients. RNA was extracted from tumor and normal peri-tumor tissues. Normal lung control tissues were from patients with benign respiratory pathologies. RT-PCR from NSCLC patients showed expression of FL BARD1 and previously identified isoforms19 (Fig. 1a). In most cases, the specific expression pattern of FL BARD1 and isoforms was identical in normal peri-tumor and tumor tissues. In normal lung tissue controls BARD1 mRNA expression was absent or very low (Fig. 1b). To distinguish isoforms of similar molecular weight, we performed RT-PCR with primers amplifying exons 1–4 or 1– 6. A complete list of primers used is presented in Supporting Information Table 1. We found that FL BARD1 and previously identified isoform b, but not a,19,27 was expressed in NSCLC. Additionally, we identified two novel isoforms, j and p (Fig. 1c). Isoform j is derived from deletion of exon 3, but translation of exon 2 into exon 4 is not in-frame, and the first translation start codon may be used within exon 4 (Figs. 1c and 2). Isoform p harbors a deletion of 408 bp encoding amino acids 301–436, within exon 4 (Figs. 1c and 2a). The 30 half of exon 4 seems to be very heterogeneous, and deletions occur frequently in lung cancer (Supporting Information Fig. 1). Since BARD1 and BARD1 isoform expression can be modulated by oestrogen through ER-a,21,36 we investigated whether ER-a was expressed in these NSCLC samples by RTPCR and found ER-a expression in all lung tissues from females and males, both in normal peri-tumor and tumor tissues (Fig. 1d), which suggests that BARD1 isoform expression might be related to oestrogen signaling in NSCLC. We quantified the expression of isoforms observed in 20 female and male NSCLC cases. FL BARD1 and isoforms showed similar expression patterns in tumor and in peritumor tissues, but were upregulated in tumors (Fig. 1e). Expression of isoforms BARD1b and j was significantly upregulated in males as compared to females (p < 0.05), while expression of isoforms BARD1g, BARD1c and BARD1e was upregulated in females as compared to males (Supporting Information Fig. 2). Only, isoform p was absent or only weakly expressed in normal peri-tumor tissues and specifically expressed in tumors (Fig. 1c). To corroborate this notion, we performed PCR reactions with primers specific for p on peritumor and tumor tissues. Indeed, in the majority of cases (80 %) isoform p was upregulated in tumors. An additional iso-


form p0 , derived from deletion of exon 2, was co-expressed with p (Fig. 1f). Real time PCR was performed to quantitatively assess the upregulation of isoform p and p0 in tumor tissues (Fig. 1g), which confirmed upregulation of p and p0 in approximately 80%. These experiments suggest that FL BARD1 and spliced isoforms are expressed in tumor and normal peri-tumor tissue and might contribute to lung cancinogenesis, while isoform p, more abundant in tumors, might be involved in tumor progression. BARD1 epitopes are expressed in all NSCLC

All of the mRNA isoforms found in NSCLC could be translated (Fig. 2a), including novel isoforms j, translated from the first methionin within exon 4 (Fig. 2b), and p, translated through all exons with an in-frame deletion of 136 amino acids within exon 4 (Fig. 2c). All other isoforms are also translated and have been described before.19,27,37,38 To investigate the protein expression pattern of BARD1, we used four antibodies that map to exons 1, 3, the N-terminal part of exon 4 and exon 11 (Fig. 2a). These antibodies have been used in the past for detection of BARD1 expression in human tumor tissues or mouse tissues.14,19,20,29 FL BARD1 should express epitopes recognized by all antibodies, isoforms b and j can only be recognized by antibodies WFS and C20, while isoform p should specifically express epitopes in exons 3 and 4, reactive with antibodies PVC and WFS. Isoforms u, d, e, should only be positive for epitopes in exons 1 and 11, recognized by N19 and C20, and g only for epitopes within exon 11. We investigated BARD1 expression in sections of tumors from 100 NSCLC patients (Table 1). Positive staining for all four antibodies was observed in the majority of tumors. However, the epitopes detected with different antibodies were expressed in different regions of tumor sections and in different sub-cellular compartments (Fig. 2d), suggesting that different isoforms of BARD1 lacking certain epitopes were expressed within a single tumor. Typically, BARD1-N19 (exon 1) and BARD1-C20 (exon 11) showed granular staining in the cytoplasm and on membranes and this staining was co-localized to the same regions in a particular tumor (Fig. 2e). BARD1 PVC (exons 3) and WFS (exon 4) immunostainings were diffused in the cytoplasm. The differences in intensity and intra-tumoral localization of epitope expression suggested that expression of four epitopes did not reflect expression of FL BARD1, but rather the simultaneous expression of different isoforms. BARD1 isoforms are not co-expressed with BRCA1

Staining with antibodies recognizing four different epitopes of BARD1 could result in 16 different staining patterns, but only a few combinations were observed (Fig. 3a), suggesting that a particular set of epitopes was expressed coordinately or on the same isoform. The expression levels of N19 and C20 reactive epitopes strongly correlated with each other, so did C 2011 UICC Int. J. Cancer: 131, 83–94 (2012) V

87

Cancer Cell Biology

Zhang et al.

Figure 1. Structure and expression of BARD1 transcripts in human lung cancer. (a–c). RT-PCR was performed with primers amplifying the entire BARD1 coding region or regions comprising exons 1–4, or 1–6, or GAPDH as control. Molecular size markers (M) are shown on the left, FL BARD1 and differentially spliced isoforms are indicated on the right. Examples of pairs of normal peri-tumor (N) and tumor (T) tissues are shown from male and female patients. (a) RT-PCR of BARD1 coding region (ATG to Stop codon) or of regions between exons 1 and 4. Normal and tumor tissues express the same pattern of isoforms. (b) RT-PCR performed on lung biopsies from individuals with benign lung diseases. No or only weak expression of individual BARD1 isoforms (c, d, g) is observed in some samples. (c) Amplification of exons 1–6 to distinguish FL BARD1, b, and novel isoforms j and p. Isoform p is specifically expressed in tumors. (d) RT-PCR for amplification of ER-a shows ER-a expression in all cases tested, with similar expression levels in normal and tumor tissues from male and female patients. (e) Quantification of isoform expression levels in peri-tumor and tumor tissues. Isoforms c, u, d, e and p are significantly upregulated in tumors. (f) Amplification of isoform p with primers at ATG and deletion junction in p. A second form, p0 , derived from additional deletion of exon 2 was identified. (g) Quantitative RT-PCR of isoforms p and p0 in 20 paired tumor and peri-tumor samples. Expression of p in tumor versus peri-tumor tissues is very variable; in comparison FL BARD1 is similar in all cases (data not shown), with similar upregulation in tumors in the majority of cases. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]


Figure 2. Structure and expression of BARD1 isoforms in NSCLC. (a) Exon structures of FL BARD1 aligned with novel (j and p) and previously identified BARD1 isoforms. FL BARD1 is presented with protein features (RING, ANK, BRCT, NLS) and positions of primers as used in experiments presented in Figure 1. Coding exons are presented green, noncoding white and alternative ORF (in b, c and g) yellow (or dotted). For isoform j deletion of exon 3 and presumed translation start (ATG) within exon 4 is shown. For isoform p deletion within exon 4 and positions of mutations and polymorphisms are indicated.33–35 Isoform designations on the left, size (amino acids) and molecular weights (MW) on the right side are indicated. Antibodies N19, PVC, WFS and C20 used for immunohistochemistry are indicated at positions of respective reactive epitopes. Pink region highlights epitopes reactive with PVC and WFS. (b) DNA and protein sequence at exon junction of isoform j. (c) DNA and protein sequence at junction of deletion within exon 4 of isoform p. (d, e) Examples of immunohistochemistry staining patterns observed with BARD1 and BRCA1 antibodies. All four BARD1 antibodies positive (d) or N19 and C20 positive (e) was most frequently observed. BARD1 N19 and C20 showed granular staining in the cytoplasm and on membranes, partially co-localized to the same cells or regions. BARD1 PVC and WFS staining was cytoplasmic or diffusely nuclear and cytoplasmic. Scale bars are indicated (upper panels ¼ 200 lm; lower panels ¼ 100 lm). Only few cases of BRCA1 co-localization with BARD1 N19 (d) and no co-localization or no staining was most frequent (e).

89

Cancer Cell Biology

Zhang et al.

Figure 3. Comparison of BARD1 expression with BRCA1 and Aurora B and with patient survival. (a) Staining pattern of NSCLC cases probed with BARD1 N19, PVC, WFS and C20 antibodies. ‘‘þ’’ indicates positive staining, ‘‘’’ indicates negative staining. All four antibodies positive was the most frequent expression pattern. (b) Pairwise comparison of N19, PVC, WFS and C20 staining scores. N19 and C20, and PVC and WFS, were strongly correlated. Other antibodies showed no correlation. (c) Comparison of N19, C20, BRCA1 and Aurora B staining in tumor and peri-tumor tissues of 10 female and 10 male NSCLC patients. All antibodies stained stronger in tumor than in peri-tumor (normal) tissues. Expression of Aurora B was analyzed in tumor and peri-tumor tissue (Supporting Information Fig. 4) and upregulation was significant. (d–f) Kaplan–Meyer analysis of disease-free survival (DFS) and (g–i) OS. DFS and OS times were related to PVC staining (d, g), WFS staining (e, h) or combined PVC and WFS staining (f, i). ‘‘’’ negative staining, ‘‘þ’’ positive staining. In (g, h), ‘‘’’ indicates either PVC and/or WFS negative staining; ‘‘þ’’ indicates the simultaneously positive staining for PVC and WFS. The p value (log-rank test) for each graph is reported. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]


Cancer Cell Biology

90

PVC and WFS (Fig. 3b), but other epitopes did not (data not shown). N19 and C20 co-expression is compatible with isoforms u, d and e, PVC and WFS co-expression only with isoform p expression. Particularly, WFS staining was upregulated in the majority of lung cancers and could be related to (co-)expression of isoforms b, j or p. C20 staining was weaker than PVC and/or WFS staining in general and the C20 epitope in exon 11 might be masked on isoforms b, j and p (Supporting Information Fig. 3). As lack of FL BARD1 expression might affect BRCA1 stability and sub-cellular localization,39,40 and function,9,41,42 we investigated the co-expression of BRCA1 with BARD1 epitopes. BRCA1 was expressed in 66.7% of NSCLC samples, but it did not correlate with the expression of any BARD1 epitopes (data not shown), and when co-expressed did not co-localized to the same intracellular compartment (Figs. 2d and 2e). Thus, without FL BARD1, the functions of the BRCA1-BARD1 heterodimer might be jeopardized in NSCLC. Since expression of BARD1 mRNA isoforms was found in tumor and peri-tumor tissue, we compared BARD1 and BRCA1 epitope expression in tumor and morphologically normal peri-tumor tissues from female and male patients (Fig. 3c). BARD1 and BRCA1 epitopes were expressed in peri-tumor tissues, but all were more elevated in tumors (Fig. 3c; Supporting Information Fig. 3). Expression was generally higher in tumors from female than from male patients (Fig. 3c). As BARD1b isoform was shown previously to inhibit the degradation of the mitotic kinase Aurora B, we tested the expression of Aurora B in peri-tumor and tumor tissues. Importantly, Aurora B expression was absent in most peri-tumor tissues, but was highly upregulated in all tumors tested (Fig. 3c; Supporting Information Fig. 3). This finding suggest that BARD1 isoforms might be involved in inhibition of Aurora B degradation, as previously shown for BARD1 b.15

BARD1 epitope expression as predictor of survival

We compared the expression of BARD1 epitopes with patients’ gender and age, and tumor grade, stage and histological type. We found that expression of specific BARD1 epitopes (mapping to exons 3 and 4) was less frequent in adenocarcinomas than in non-adenocarcinomas (including squamous cell carcinoma and large cell carcinoma) (Supporting Information Fig. 4). However, we found no correlation of any BARD1 epitope expression with tumor grade and stage or patients’ gender and age (data not shown). We also compared the expression of individual BARD1 epitopes with DFS and OS for 65 patients with follow-up data. We found that patients with positive staining for either PVC (exon 3) or WFS (exon 4) had significantly shorter DFS (Figs. 3d and 3e) and shorter OS (Figs. 3g and 3h) times than those with negative staining. Patients showing simultaneous positive staining for PVC and WFS also had shorter survival times (Figs. 3f and 3i). No correlations with either


DFS or OS were found for N19, C20 or combinations of other staining patterns. Univariate and multivariate analysis using Cox’s proportional hazards model were performed to evaluate whether expression of BARD1 PVC and WFS epitopes had prognostic significance independently of other prognostic factors. Multivariate analysis included the pathological stage (which showed prediction of DFS and OS in univariate analysis) and two other possible prognostic factors, namely histological type of tumors and patient gender. The multivariate analysis showed that individual PVC or WFS positive staining have significance as independent prognostic factors for DFS, but simultaneous PVC or WFS positive stainings have significance for DFS and OS (Table 2). Sequential expression of BARD1 epitopes during tumorigenesis

To investigate a correlation of BARD1 isoform expression with initiation and progression of lung cancer, we monitored BARD1 expression in experimentally induced (urethane treatment) lung cancer from primary tumors to progression into adenocarcinomas.32 This treatment reproducibly leads to macroscopic tumors after 16 weeks, larger tumors at 24 weeks and invasive lung cancer after 32 weeks. We selected normal tissues and tumor regions at all stages for immunohistochemistry and scored the antibody staining (Figs. 4a–4d). Consistently, expression of PVC and WFS was weak, and expression of C20 was strong in normal tissue. This pattern was similar in 16 weeks tumors. However, in 24 weeks tumors, PVC and WFS expressions increased in comparison to C20 and were highly upregulated in 32 weeks tumors. These experiments demonstrate that the BARD1 epitope expression pattern changes during different stages of tumorigenesis and suggest that epitopes mapping to exons 3 and 4 may be involved in tumor promotion and progression towards an invasive stage. Similar observations are made with mouse cell lines with premalignant and malignant phenotypes (Supporting Information Fig. 5).

Discussion We report that epitopes reflecting BARD1 isoforms derived from differential splicing are expressed in each of 100 NSCLC samples tested, while in healthy tissue expression is undetectable. Alternative splicing is specifically frequent in lung cancer, and observed for many regulatory proteins.43 All NSCLC express several BARD1 mRNA isoforms previously described in breast and ovarian cancers, and in addition, two novel ones. All were expressed in peri-tumoral and tumoral tissues alike, and only isoform p, was significantly upregulated in tumors. The expression of BARD1 epitopes in peritumor tissues seems specific for lung cancer and was not observed in a similar study on colon cancer cases (Zhang and Irminger-Finger, unpublished). In vitro studies have shown that BARD1 isoforms, but not FL BARD1, are essential for cell proliferation and that their C 2011 UICC Int. J. Cancer: 131, 83–94 (2012) V

Zhang et al.

91

Table 2. Univariate and multivariate analysis of survival in 65 NSCLC patients with follow-up data DFS

OS

HR

95% CI

p

HR

95% CI

p

N19 pos vs., neg

1.08

0.50-2.32

0.849

1.11

0.46-2.67

0.811

PVC pos vs., neg

3.84

1.50-9.87

0.005

4.79

1.41-16.2

0.012

WFS pos vs., neg

3.39

1.50-7.67

0.003

2.47

1.01-6.00

0.047

C20 pos vs., neg

1.26

0.53-2.98

0.602

1.45

0.51-4.12

0.481

N19 and C20 pos vs., others

1.08

0.50-2.32

0.849

1.11

0.46-2.67

0.811

PVC and WFS pos vs., others

3.22

1.49-6.96

0.003

3.02

1.22-7.46

0.017

4 Abs pos vs., others

2.04

1.11-3.77

0.123

1.86

0.91-3.81

0.091

Stage III, IV vs., I, II

3.04

1.61-5.72

0.001

3.60

1.85-7.02

0.000

AC vs., non-AC

0.72

0.41-1.24

0.232

0.62

0.33-1.14

0.124

Grade 3, 4 vs., 1, 2

1.50

0.84-2.66

0.167

1.87

0.96-3.66

0.068

Predictors

Age (years) < 60 vs., > 60

1.06

0.61-1.83

0.837

1.21

0.64-2.28

0.557

Male vs., female

1.35

0.65-2.80

0.414

0.89

0.35-2.27

0.804

Surgery vs., Surgery þ adjuvant therapy

0.61

0,33-1,13

0.114

0.76

0,38-1,52

0.440

PVC pos vs., neg*

2.85

1,03-7,89

0.043

3.08

0,85-11,17

0.088


2.48

1,21-5,10

0.013

3.22

1,49-6,95

0.003

AC vs., non-AC

1.32

0,71-2,45

0.381

1.47

0,73-2,94

0.278

Male vs., female

1.46

0,67-3,18

0.340

1.13

0,42-3,07

0.808

Age (years)60

1.08

0,58-2,00

0.810

1.32

0,65-2,68

0.449

Surgery vs., surgery þ adjuvant therapy

0.92

0,45-1,87

0.817

1.05

0,48-2,28

0.912

WFS pos vs., neg*

3.98

1,61-9,86

0.003

2.48

0,92-6,69

0.072


3.97

1,88-8,38

0.000

4.43

2,02-9,74

0.000

AC vs., non-AC

1.14

0,62-2,12

0.669

1.39

0,69-2,80

0.361

Male vs., female

1.53

0,71-3,33

0.278

1.25

0,46-3,39

0.659

Age (years)60

1.00

0,55-1,83

0.991

1.32

0,66-2,65

0.434


1.20

0,57-2,52

0.641

1.26

0,55-2,84

0.586

PVC and WFS pos vs., others

3.71

1,55-8,91

0.003

3.05

1,10-8,43

0.032


3.79

1,81-7,93

0.000

4.41

2,01-9,65

0.000

Multivariate analysis

1

AC vs., non-AC

1.12

0,60-2,08

0.721

1.30

0,64-2,63

0.462

Male vs., female

1.46

0,67-3,18

0.340

1.21

0,45-3,28

0.710

Age (years)60

0.97

0,53-1,77

0.918

1.22

0,60-2,46

0.581


1.20

0,57-2,53

0.628

1.30

0,58-2,93

0.526

Abbreviations: HR, hazard ratio; 95% CI, 95% confidence interval; p, p-value; pos, positive staining; neg, negative staining. AC, adenocarcinoma; including squamous cell carcinoma and large cell carcinoma; others, all other combinations. 1 HR adjusted for all the other predictors in the model (pathology, stage , sex, age and surgery þ/ adjuvant).

repression leads to proliferation arrest.15,19 BARD1 isoforms play oncogenic roles and are associated with poor prognosis in breast and ovarian cancer.19,20 One might assume that BARD1 isoforms have similar functions in NSCLC. The novel BARD1 isoforms, j and p, either lack the BRCA1 interaction domain (j) or nuclear localization signals (NLS) (p) (Fig. 2a). Both features are essential for BARD1 and BRCA1 sub-cellular localization and function.30,33,34,39,44–47 Deletion of isoform p harbors mutations found in breast and C 2011 UICC Int. J. Cancer: 131, 83–94 (2012) V

ovarian cancers,20,33–35 and isoform p seems particularly important for lung cancer as it is the only isoform that is significantly upregulated in tumor tissue and absent or only weakly expressed in peri-tumor tissue (Figs. 1c and 1e–1g). Epitopes PVC and WFS (exons 3 and 4) presumably reflect isoform p expression. This combination of epitopes was not observed for breast and ovarian cancer,19 consistent with the specific expression of isoform p in lung cancer. Lack of NLS causing aberrant intracellular localization could affect

Cancer Cell Biology

Univariate analysis


Cancer Cell Biology

92

Figure 4. Sequential expression of BARD1 epitopes in induced mouse lung cancer. (a) BARD1 expression in morphologically normal lung tissue (Normal) of urethane treated mice. PVC and WFS epitopes were detected in some type II pneumocytes, but not in type I pneumocytes at 16 weeks (wk). PVC and WFS were expressed and upregulated in type II and type I pneumocytes at 24 weeks and 32 weeks. C20 staining was strong at 16 weeks, weak at 24 weeks and almost negative at 32 weeks. (b) BARD1 expression in tumors. In tumor regions, PVC and specifically WFS expression were upregulated from 16 to 32 weeks, while C20 expression was downregulated from 16 to 32 weeks. (c, d) Expression pattern of BARD1 epitopes in normal (c) and tumor (d) tissues of three mice is summarized. The staining scores, as percentage of positive cells, are indicated.

protein–protein interactions and posttranslational modifications. This might explain why staining with C-terminal C20 antibody is weak although exon 11, present on all isoforms, is expressed (Supporting Information Fig. 3). Epitopes PVC and WFS are also increasingly expressed in aggressive lung tumors of the mouse model (Fig. 4), and their expression correlates with decreased patient survival times, suggesting that PVC and WFS epitopes are predictive biomarker for NSCLC. None of the BARD1 epitopes expressed in NSCLC showed correlated expression or co-localization with BRCA1. Upregulation of BRCA1 mRNA is a predictive marker of response to chemotherapy and correlates with poor prognosis in NSCLC.48 However, BRCA1 protein expression does not correlate with outcome, and the mechanistic link behind BRCA1 mRNA expression and patient prognostics is not clear. Our

data show BRCA1 is mis-located to the cytoplasm, due to the lack of FL BARD1,39,46 in all NSCLC samples. The observed high BRCA1 mRNA levels might reflect a previously described compensatory autoregulatory transcriptional feedback loop of BRCA1 expression.49 In addition to lack of FL BARD1 in cancer cells, BARD1 isoforms act antagonistically to FL BARD1 and stabilize targets of the BRCA1-BARD1 ubiquitin ligase, such as Aurora B and ER-a.15,21 Aurora B expression is indeed upregulated in NSCLC, while it is absent in normal tissue (Fig. 3c and Supporting Information Fig. 3). It is likely that NSCLC-specific isoforms of BARD1 are involved in this upregulation. Our data also provide a possible link between lung cancer susceptibility and estrogen signaling. BARD1 isoforms are upregulated by estrogen via ER-a.21,36 Isoforms g, previously C 2011 UICC Int. J. Cancer: 131, 83–94 (2012) V

Zhang et al.

93

associated with invasive human cytotrophoblasts,27 is significantly upregulated in tissues from female NSCLC patients (Supporting Information Figs. 2a and 2b), suggesting that estrogen signaling induces proproliferative and oncogenic BARD1 isoforms in NSCLC. In conclusion, several BARD1 isoforms seem to be associated with carcinogenesis in lung cancer, and epitopes expressed on these isoforms are correlated with decreased DFS and OS of NSCLC patients.

A role of spliced BARD1 isoforms as tumor promoters is now also emerging in other cancers (Zhang and IrmingerFinger, unpublished).

Acknowledgements Authors are grateful to Drs. J.Y. Wu, S. Picciau and M. Pilyugin for experimental expertise and critical discussions. This work was financed by SNRF grant 31003A-122353 and Oncosuisse grant KLS-01962-10-2006 to I.I.-F and USPHS grant CA132557 to A.M.M.

1.

Han MK, Oh YH, Kang J, Kim YP, Seo S, Kim J, Park K, Kim HS. Protein profiling in human sera for identification of potential lung cancer biomarkers using antibody microarray. Proteomics 2009;9:5544–52. 2. Kohno T, Otsuka A, Girard L, Sato M, Iwakawa R, Ogiwara H, Sanchez-Cespedes M, Minna JD, Yokota J. A catalog of genes homozygously deleted in human lung cancer and the candidacy of PTPRD as a tumor suppressor gene. Genes Chromosomes Cancer 2010;49:342–52. 3. Goto Y, Shinjo K, Kondo Y, Shen L, Toyota M, Suzuki H, Gao W, An B, Fujii M, Murakami H, Osada H, Taniguchi T, et al. Epigenetic profiles distinguish malignant pleural mesothelioma from lung adenocarcinoma. Cancer Res 2009;69: 9073–82. 4. Bishop JA, Benjamin H, Cholakh H, Chajut A, Clark DP, Westra WH. Accurate classification of non-small cell lung carcinoma using a novel microRNA-based approach. Clin Cancer Res 2010;16:610–19. 5. Bastide K, Guilly MN, Bernaudin JF, Joubert C, Lectard B, Levalois C, Malfoy B Chevillard S. Molecular analysis of the Ink4a/Rb1-Arf/Tp53 pathways in radon-induced rat lung tumors. Lung Cancer 2009;63:348–353. 6. Bartolucci R, Wei J, Sanchez JJ, Perez-Roca L, Chaib I, Puma F, Farabi R, Mendez P, Roila F, Okamoto T, Taron M, Rosell R. XPG mRNA expression levels modulate prognosis in resected non-small-cell lung cancer in conjunction with BRCA1 and ERCC1 expression. Clin Lung Cancer 2009;10:47–52. 7. Rosell R, Skrzypski M, Jassem E, Taron M, Bartolucci R, Sanchez JJ, Mendez P, Chaib I, Perez-Roca L, Szymanowska A, Rzyman W, Puma F, et al. BRCA1: a novel prognostic factor in resected non-small-cell lung cancer. PLoS One2007;2:e1129. 8. Reguart N, Cardona AF, Carrasco E, Gomez P, Taron M, Rosell R. BRCA1: a new genomic marker for non-small-cell lung cancer. Clin Lung Cancer 2008;9:331–339. 9. Baer R, Ludwig T. The BRCA1/BARD1 heterodimer, a tumor suppressor complex with ubiquitin E3 ligase activity. Curr Opin Genet Dev 2002;12:86–91. 10. Irminger-Finger I, Jefford CE. Is there more to BARD1 than BRCA1? Nat Rev Cancer2006;6: 382–91. 11. Hashizume R, Fukuda M, Maeda I, Nishikawa H, Oyake D, Yabuki Y, Ogata H, Ohta T. The RING


12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

heterodimer BRCA1-BARD1 is a ubiquitin ligase inactivated by a breast cancer-derived mutation. J Biol Chem 2001;276:14537–40. Xia Y, Pao GM, Chen HW, Verma IM, Hunter T. Enhancement of BRCA1 E3 ubiquitin ligase activity through direct interaction with the BARD1 protein. J Biol Chem 2003;278:5255–63. Feki A, Jefford CE, Berardi P, Wu JY, Cartier L, Krause KH, Irminger-Finger I. BARD1 induces apoptosis by catalysing phosphorylation of p53 by DNA-damage response kinase. Oncogene 2005; 24:3726–36. Irminger-Finger I, Leung WC, Li J, DuboisDauphin M, Harb J, Feki A, Jefford CE, Soriano JV, Jaconi M, Montesano R, Krause KH. Identification of BARD1 as mediator between proapoptotic stress and p53-dependent apoptosis. Mol Cell 2001;8:1255–66. Ryser SE, Dizin C, Jefford B, Delaval A, Christodoulidou N, Gagos S, Krause KH, Birnbaum D, Irminger-Finger I. Distinct roles of BARD1 isoforms in mitosis: full length BARD1 mediates Aurora B degradation, cancer-associated BARD1b scaffolds Aurora B and BRCA2. Cancer Res 2009;69:1125–34. Joukov V, Groen AC, Prokhorova T, Gerson R, White E, Rodriguez A, Walter JC, Livingston DM. The BRCA1/BARD1 heterodimer modulates ran-dependent mitotic spindle assembly. Cell 2006;127:539–52. Ren B, Cam H, Takahashi Y, Volkert T, Terragni J, Young RA, Dynlacht BD. E2F integrates cell cycle progression with DNA repair, replication, and G(2)/M checkpoints. Genes Dev 2002;16: 245–56. Hayami R, Sato K, Wu W, Nishikawa T, Hiroi J, Ohtani-Kaneko R, Fukuda M, Ohta T. Downregulation of BRCA1-BARD1 ubiquitin ligase by CDK2. Cancer Res 2005;65:6–10. Li L, Ryser S, Dizin E, Pils D, Krainer M, Jefford CE, Bertoni F, Zeillinger R, Irminger-Finger I. Oncogenic BARD1 isoforms expressed in gynecological cancers. Cancer Res 2007;67: 11876–85. Wu JY, Vlastos AT, Pelte MF, Caligo MA, Bianco A, Krause KH, Laurent GJ, Irminger-Finger I. Aberrant expression of BARD1 in breast and ovarian cancers with poor prognosis. Int J Cancer 2006;118:1215–26. Dizin E, Irminger-Finger I. Negative feedback loop of BRCA1-BARD1 ubiquitin ligase on estrogen receptor alpha stability and activity

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

antagonized by cancer-associated isoform of BARD1. Int J Biochem Cell Biol 2010;42:693–700. Capasso M, Devoto M, Hou C, Asgharzadeh S, Glessner JT, Attiyeh EF, Mosse YP, Kim C, Diskin SJ, Cole KA, Bosse K, Diamond M, et al. Common variations in BARD1 influence susceptibility to high-risk neuroblastoma. Nat Genet 2009;41:718–23. Planchard D, Loriot Y, Goubar A, Commo F, Soria JC. Differential expression of biomarkers in men and women. Semin Oncol 2009;36:553–65. Chen GG, Zeng Q, Tse GM. Estrogen and its receptors in cancer. Med Res Rev 2008;28: 954–974. Eakin CM, Maccoss MJ, Finney GL, Klevit RE. Estrogen receptor alpha is a putative substrate for the BRCA1 ubiquitin ligase. Proc Natl Acad Sci USA 2007;104:5794–9. Skern R, Frost P, Nilsen F. Relative transcript quantification by quantitative PCR: roughly right or precisely wrong? BMC Mol Biol2005;6:10. Li L, Cohen M, Wu J, Sow MH, Nikolic B, Bischof P, Irminger-Finger I. Identification of BARD1 splice-isoforms involved in human trophoblast invasion. Int J Biochem Cell Biol 2007;39:1659–72. Irminger-Finger I, Soriano JV, Vaudan G, Montesano R, Sappino AP. In vitro repression of Brca1-associated RING domain gene. Bard1, induces phenotypic changes in mammary epithelial cells. J Cell Biol 1998;143:1329–39. Feki A, Jefford CE, Durand P, Harb J, Lucas H, Krause KH, Irminger-Finger I. BARD1 expression during spermatogenesis is associated with apoptosis and hormonally regulated. Biol Reprod 2004;71:1614–24. Jefford CE, Feki A, Harb J, Krause KH, IrmingerFinger I. Nuclear-cytoplasmic translocation of BARD1 is linked to its apoptotic activity. Oncogene 2004;23:3509–20. Bosman FT, de Goeij AF, Rousch M. Quality control in immunocytochemistry: experiences with the oestrogen receptor assay. J Clin Pathol 1992;45:120–4. Redente EF, Dwyer-Nield LD, Barrett BS, Riches DW, Malkinson AM. Lung tumor growth is stimulated in IFN-c/ mice and inhibited in IL-4Ra/ mice. Anticancer Res 2009;29: 5095–101. Ishitobi M, Miyoshi Y, Hasegawa S, Egawa C, Tamaki Y, Monden M, Noguchi S. Mutational analysis of BARD1 in familial breast cancer

Cancer Cell Biology

References

94

34.

35.

36.

Cancer Cell Biology

37.

38.


patients in Japan. Cancer Lett2003; 200:1–7. Huo X, Hu Z, Zhai X, Wang Y, Wang S, Wang X, Qin J, Chen W, Jin G, Liu J, Gao J, Wei Q, Shen H. Common non-synonymous polymorphisms in the BRCA1 Associated RING Domain (BARD1) gene are associated with breast cancer susceptibility: a case–control analysis. Breast Cancer Res Treat 2007;102: 329–37. Thai TH, Du F, Tsan JT, Jin Y, Phung A, Spillman MA, Massa HF, Muller CY, Ashfaq R, Mathis JM, Miller DS, Trask BJ, et al. Mutations in the BRCA1-associated RING domain (BARD1) gene in primary breast, ovarian and uterine cancers. Hum Mol Genet 1998;7:195–202. Creekmore AL, Ziegler YS, Boney JL, Nardulli AM. Estrogen receptor a regulates expression of the breast cancer 1 associated ring domain 1 (BARD1) gene through intronic DNA sequence. Mol Cell Endocrinol 2007;267:106–15. Tsuzuki M, Wu W, Nishikawa H, Hayami R, Oyake D, Yabuki Y, Fukuda M, Ohta T. A truncated splice variant of human BARD1 that lacks the RING finger and ankyrin repeats. Cancer Lett 2006;233:108–16. Tembe V, Henderson BR. BARD1 translocation to mitochondria correlates with Bax

39.

40.

41.

42.

43. 44.

oligomerization, loss of mitochondrial membrane potential and apoptosis. J Biol Chem 2007;282: 20513–22. Fabbro M, Rodriguez JA, Baer R, Henderson BR. BARD1 induces BRCA1 intranuclear foci formation by increasing RING-dependent BRCA1 nuclear import and inhibiting BRCA1 nuclear export. J Biol Chem 2002;277: 21315–24. Joukov V, Chen J, Fox EA, Green JB, Livingston DM. Functional communication between endogenous BRCA1 and its partner, BARD1, during Xenopus laevis development. Proc Natl Acad Sci USA 2001;98:12078–83. Thakar A, Parvin JD, Zlatanova J. BRCA1/ BARD1 E3 ubiquitin ligase can modify histones H2A and H2B in the nucleosome particle. J Biomol Struct Dyn 2010;27:399–406. Starita LM, Parvin JD. Substrates of the BRCA1dependent ubiquitin ligase. Cancer Biol Ther 2006;5:137–41. Pio R, Montuenga LM. Alternative splicing in lung cancer. J Thorac Oncol 2009;4:674–78. Schuchner S, Tembe V, Rodriguez JA, Henderson BR. Nuclear targeting and cell cycle regulatory function of human BARD1. J Biol Chem 2005;280: 8855–61.

45. Fabbro M, Schuechner S, Au WW, Henderson BR. BARD1 regulates BRCA1 apoptotic function by a mechanism involving nuclear retention. Exp Cell Res 2004;298:661–73. 46. Henderson BR. Regulation of BRCA1, BRCA2 and BARD1 intracellular trafficking. Bioessays 2005;27:884–93. 47. Rebbeck TR, Mitra N, Domchek SM, Wan F, Chuai S, Friebel TM, Panossian S, Spurdle A, Chenevix-Trench G, Singer CF, Pfeiler G, Neuhausen SL, et al. Modification of ovarian cancer risk by BRCA1/2-interacting genes in a multicenter cohort of BRCA1/2 mutation carriers. Cancer Res 2009; 69:5801–10. 48. Boukovinas I, Papadaki C, Mendez P, Taron M, Mavroudis D, Koutsopoulos A, Sanchez-Ronco M, Sanchez JJ, Trypaki M, Staphopoulos E, Georgoulias V, Rosell R, et al. Tumor BRCA1. RRM1 and RRM2 mRNA expression levels and clinical response to first-line gemcitabine plus docetaxel in non-small-cell lung cancer patients. PLoS One2008;3:e3695. 49. De Siervi A, De Luca P, Byun JS, Di LJ, Fufa T, Haggerty CM, Vazquez E, Moiola C, Longo DL, Gardner K. Transcriptional autoregulation by BRCA1. Cancer Res 2010;70: 532–42.
