Tumor aromatase expression as a prognostic factor for local control in young breast cancer patients after breast-conserving treatment

bcr2343 BCJ Research article Tumor aromatase expression as a prognostic factor for local control in young breast cancer patients after breast-conserving treatment Bollet A Marc marc.bollet@curie.net Savignoni Alexia alexia.savignoni@curie.net De Koning Leanne Leanne.De-Koning@curie.fr Tran-Perennou Carine carine.perennou@curie.net Barbaroux Catherine catherine.barbaroux@curie.net Degeorges Armelle armelle.degeorges@curie.net Sigal-Zafrani Brigitte brigitte.sigal@curie.net Almouzni Geneviève genevieve.almouzni@curie.fr Cottu Paul paul.cottu@curie.net Salmon Rémy remy.salmon@curie.net Servant Nicolas nicolas.servant@curie.fr Fourquet Alain alain.fourquet@curie.net de Cremoux Patricia patricia.de-cremoux@curie.net

Department of Radiation Oncology, Institut Curie, 26 rue d'Ulm, 75248 Paris, France

Department of Biostatistics, Institut Curie, 26 rue d'Ulm, 75248 Paris, France

Laboratory of Nuclear Dynamics and Genome Plasticity (UMR 218), Institut Curie, 26 rue d'Ulm, 75248 Paris, France

Department of Tumour Biology, Institut Curie, 26 rue d'Ulm, 75248 Paris, France

Department of Medical Oncology, Institut Curie, 26 rue d'Ulm, 75248 Paris, France

Department of Surgery, Institut Curie, 26 rue d'Ulm, 75248 Paris, France

Department of Bio-informatics, Institut Curie, 26 rue d'Ulm, 75248 Paris, France

INSERM, U900, 26 rue d'Ulm, 75248 Paris, France

Ecole des Mines de Paris, 35 rue Saint Honoré, 77300, Fontainebleau, France

Breast Cancer Research 1465-5411 2009 11 4 R54 http://breast-cancer-research.com/content/11/4/R54 19638208 10.1186/bcr2343 14 3 2009 27 4 2009 18 6 2009 28 7 2009 28 7 2009 2009 Bollet; licensee BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium provided the original work is properly cited.

Abstract

Introduction

We sought to determine whether the levels of expression of 17 candidate genes were associated with locoregional control after breast-conserving treatments of early-stage breast cancers in young, premenopausal women.

Methods

Gene expression was measured by using RT-PCR in the breast tumors of a series of 53 young (younger than 40 years), premenopausal patients. All treatments consisted of primary breast-conserving surgery followed by whole-breast radiotherapy (± regional lymph nodes) with or without systemic treatments (chemotherapy ± hormone therapy). The median follow-up was 10 years.

Results

The 10-year locoregional control rate was 70% (95% CI, 57% to 87%). In univariate analysis, no clinical/pathologic prognostic factors were found to be significantly associated with decreased locoregional control. Expression of three genes was found to be significantly associated with an increased locoregional recurrence rate: low estrogen-receptor β, low aromatase, and high GATA3. Two others were associated with only a trend (P < 0.10): low HER1 and SKP2. In multivariate analysis, only the absence of aromatase was significantly associated with an increased locoregional recurrence rate (P = 0.003; relative risk = 0.49; 95% CI 0.29 to 0.82).

Conclusions

Recent data give credit to the fact that breast cancer in young women is a distinct biologic entity driven by special oncogenic pathways. Our results highlight the role of estrogen-signaling pathways (mainly CYP19/aromatase, GATA3, and ER-β) in the risk of locoregional recurrence of breast cancer in young women. Confirmation in larger prospective studies is needed.

Introduction

Breast-conserving therapy is the preferred treatment for patients with early-stage breast cancer 1. It offers equal local control and overall survival 2, as well as superior psychosocial outcomes compared with modified radical mastectomy 34. Locoregional recurrences can be traumatizing and even fatal, despite aggressive therapies 2. Young age is generally considered to be the most important risk factor for locoregional recurrence after breast-conserving treatments 567. This higher risk, which is not yet understood despite numerous studies, could find its explanation in tumor biology.

The hormonal environment, with menopause as its archetype epitome, is the major physiological difference between younger and older patients. Estrogens not only are the main regulators of growth and differentiation in the normal mammary gland, but also play a major role in the onset and progression of breast cancer 89 (reviewed by Yager 10). Other signaling pathways, not directly related to estrogen receptors (ERs), also are involved in the growth of epithelial tissues.

In premenopausal breast cancer patients, little is known about the expression levels of genes that are directly or indirectly involved in hormone (especially ER) and growth factors signaling pathways. The aim of this study, conducted in a series of women diagnosed with invasive breast cancers before the age of 40, was therefore to determine the relation between locoregional relapse, classic biopathologic factors, and the intratumoral levels of gene expression of 17 hormone receptors, growth-factor receptors, or proliferation genes: ERα, ERβ, progesterone receptor (PR), nuclear receptor co-repressor (NCoR), nuclear receptor coactivator 3 (NCoA3/AIB1), aromatase (CYP19), GATA-binding protein 3 (GATA3), human epidermal receptor (HER) 1 to 4, insulin-like growth factor 1-receptor (IGF1R), antigen identified by monoclonal antibody ki-67 (MKI67), cyclin E1 (CCNE1), cyclin E2 (CCNE1), S-phase kinase-associated protein 2 (SKP2), and the two subunits of chromatin assembly factor 1 (CAF-1 p150 and CAF-1 p60). Quantitative reverse-transcriptase polymerase chain reaction (RT-PCR) was chosen, as it is the most precise method to measure absolute levels of expression of selected target genes within a wide range, from very high to very low transcript levels 1112.

Materials and methods

1.1 Patients and tissue specimens

Between 1988 and 1999, 257 premenopausal women, younger than 40 years, with no previous history of cancer were treated at the Institut Curie for early unilateral breast cancers. The present retrospective study was based on 53 of these 257 patients for whom frozen tumor tissue was available. Median age at diagnosis was 37 years (range, 23 to 40) with 30% of patients (16 patients) no older than 36 years. Median follow-up was 10 years (range, 2 to 18 years).

Patient and tumor characteristics are reported in Table 1. Clinical stage 13 was either T1 or T2, N0 or N1. All specimens were reviewed by the same pathologist (BSZ). Histologic classification of the infiltrating carcinomas was reported according to the World Health Organization criteria, and histologic grade was reported according to Ellis and Elston 14. The median number of mitoses, calculated per 10 high-power fields 15, was 13 (2 to 120). Hormone receptors (HRs) were positive when nuclear staining for either estradiol receptors (ERs) or progesterone receptors (PRs) was observed in at least 10% of invasive cells with immunohistochemistry 16. No pathologic axillary lymph node involvement was observed in 31 (58%) patients.

Table 1

Patient characteristics

RT-PCR series

Whole series

Total

Family history of breast cancer (MD = 0 and 1)^a

0.77

Without

198

With

Age (MD = 0 and 0)^a

Median (min-max) in years

37 (23–40)

0.71

≤ 35 years old

176

0.85

> 35 years old

Clinical T stage (MD = 0 and 0)^a

0.0496

cT0–1

186

cT2

Clinical N stage (MD = 0 and 0)^a

0.70

246

Type of invasive carcinoma (MD = 0 and 0)^a

0.18^b

Ductal

223

Lobular

Other

Histologic grade (MD = 0 and 1)^a

0.03^b

1–2

110

120

Unclassifiable

Estrogen receptor (ER) (MD = 6 and 37)^a

0.59

ER^-

ER⁺

165

Progesterone receptor (PR) (MD = 6 and 39)^a

0.84

PR^-

PR⁺

179

Hormone receptors (HRs) (MD = 6 and 37)^a

0.87

HR^-

HR⁺

185

Lymphovascular involvement (MD = 3 and 43)^a

0.0004

Absent

188

Present

Histologic T stage (MD = 5 and 12)^a

0.89

pT1

176

pT2

Histologic N stage (MD = 0 and 34)^a

0.06

pN0

182

pN1–3

Surgical margins (MD = 4 and 10)^a

0.10

Satisfactory (≥ 3 mm)

129

Unsatisfactory

118

Systemic therapy (MD = 0 and 0)^a

0.0001

None

151

Hormone therapy (HT) only

101

Chemotherapy ± HT

Total RT dose (MD = 0)^a

Median (min-max) in Gy

66 (50–75)

64 (0–80)

0.14

< 60 Gy

101

0.043

≥ 60 Gy

156

Characteristics of the rt-PCR series (53 patients) in comparison to the whole population of consecutive patients with the same selection criteria treated over the same period (1988–1999) at the Institut Curie (257 patients). MD = missing data (both in the rt-PCR and in the whole series); total RT dose = total radiotherapy dose to the tumor bed (whole-breast radiotherapy dose + boost dose). ^a0.37 when the comparison excluded histologic types other than ductal or lobular. ^b0.05 when the comparison included all histologic grades.

This study was approved by the Institutional Review Board and Ethics committee. Patients were informed that their biologic samples could be used for research purposes and that they had the right to refuse if they so wished.

1.2 Treatments

Surgery consisted of breast-conserving procedures as first-line treatment in all cases. The quality of the surgical margins was classified as wide (more than 3 mm) in 32 (65%) patients, close (3 mm or less) in 11 (22%) patients, involved with ductal carcinoma in situ in 2 (4%) patients, involved with invasive carcinoma in 4 (8%) patients, and unknown in 4 patients. Axillary lymph node dissection was performed in all patients.

Patients received posttumorectomy radiotherapy with a median dose of 51 Gy (range, 45 to 54 Gy) to the breast. A boost to the tumor bed was performed for 40 (75%) of patients with a median dose of 16 Gy (range, 7 to 25 Gy). The median total dose to the tumor bed was 66 Gy (range, 50 to 75 Gy). Supraclavicular irradiation was performed in 27 (51%) of patients. Internal mammary irradiation was performed in 38 (72%) of patients. In the case of lymph node involvement, the internal mammary chain and the supraclavicular area were both irradiated. In the absence of lymph node involvement, irradiation of the internal mammary chain, with or without the supraclavicular area, was indicated for centrally located tumors with histopathologic features of aggressiveness. Axillary irradiation was added in the presence of extensive axillary involvement or in the absence of axillary lymph node dissection.

No protocol to boost all young patients with negative surgical margins was available at that time, and some patients reported in this series were accrued in the EORTC boost trial that randomized between boost and no boost from 1989 to 1996 51718. In the case of positive surgical margins, a radiotherapy boost of generally 20 to 28 Gy was added to the whole-breast irradiation. For patients not participating in the EORTC randomized trial, a boost of 10 to 16 Gy was added in the case of aggressive histopathologic features (unsatisfactory margins, high histopathologic grade, high proliferation index, absence of hormone receptors).

The reasons for absence of repeated excision were not always specified, but when they were specified, the reasons were the patient's choice not to undergo a new surgical procedure that could have been mastectomy.

No systemic therapy was administered in 16 (30%) patients. Chemotherapy only was given to 23 (43%) patients, consisting of anthracycline-based combination chemotherapy in most cases (usually six cycles of 5-fluorouracil, doxorubicin, and cyclophosphamide). A combination of chemotherapy and hormone therapy (tamoxifen for 5 years) was administered to 14 (26%) patients.

1.3 RNA isolation, RT-PCR

Total RNA was extracted from frozen tumor samples and first-strand cDNA synthesis was performed, as previously described from 1 μg of total RNA by using Superscript II RT (Invitrogen, Carlsbad, CA, USA) 19. ERα, ERβ, PR, HER1 to HER4, MKi67, Cyclin E1 and E2, GATA3, IGF1-R, NCor, NCoA3, CYP19 (aromatase gene), SKP2 and CAF-1 p150, and CAF-1 p60 transcripts were quantified by using real-time quantitative reverse transcription-PCR assays. The nucleotide and probe sequences and the conditions of PCR were described previously for ERα, ERβ, PR, HER1 to HER4, MKi67, cyclin E1 and E2, NCor, NCoA3, CYP19 [see Additional data file 1] 1920. GATA3, IGF1R were quantified by Assays-on-Demand from Applied Biosystems (Applied Biosystems, Inc., Foster City, CA). The nucleotide and probe sequences SKP2 and CAF-1 were chosen with the help of Primer express software (Applied Biosystems). TBP (TATA box binding protein) was used as the endogenous reference gene for immediate quantification of transcripts. Searches were conducted in dbEST, htgs, and nr data bases 21 to confirm the total gene specificity of nucleotide sequences. Primers were placed at the junction between two exons. All PCRs were performed in duplicate by using ABI Prism 7700 Sequence Detection System (Perkin Elmer Applied Biosystems) and the Core reagent kit (Eurogentec). A 5-μl diluted sample of cDNA (12.5 ng) was added to 20 μl of the PCR mix. The thermal-cycling conditions comprised an initial denaturation step at 95°C for 10 min, and 45 cycles at 95°C for 15 s and either 60°C or 65°C, depending on the target, for 1 min.

Additional file 1

Table listing primer and probe sequences, PCR efficiencies, regression coefficient and references.

Click here for file

Results were expressed as N-fold differences in target gene expression relative to a reference gene defined as "N target" (arbitrary units) and was determined as follows:

where E is the efficiency of PCR measured by using the slope of the calibration curve, and Ct is the cycle threshold.

1.4 Statistical analysis

Survival rates, defined from the date of surgery to the occurrence of the event, were estimated by using the Kaplan-Meier estimate; groups were compared by using the log-rank test. Event-free patients were censored at the date of their last known contact or death. Local relapses were defined as the occurrence of breast carcinoma (either invasive or ductal carcinoma in situ) in the treated breast. Locoregional relapses were our primary end point and were defined as either a local relapse or a recurrence in the ipsilateral lymph node areas (axillary, internal mammary, supraclavicular). Contralateral breast cancers could be either ductal in situ or infiltrating carcinoma. Distant disease was defined as disease occurring elsewhere than in the contralateral breast or locoregional site. Disease-free was defined as the absence of locoregional or distant relapses or both. Survival rates and relative risks (RRs) are presented with their 95% confidence intervals (CIs). Annual risks were calculated and plotted. The RT-PCR series of 53 patients was compared with the whole series of all consecutive patients treated over the same period (1988 through 1999) at the Institut Curie (257 patients) after exclusion of the RT-PCR patients. All gene expressions were analyzed as continuous variables to avoid having to search for cut-off values. Fisher's Exact or χ²tests were used to compare percentages, as appropriate. Univariate analyses were performed to identify prognostic factors of outcome and to estimate crude relative risks (RRs). The influence of each factor, adjusted to the others, was assessed in a multivariate analysis by the Cox proportional hazards model 22. A stepwise modeling algorithm was used, with a limit of significance of 0.10 for entering and 0.05 for removing risk factors. The limit of significance was 0.05. Analyses were performed by using R software, 2.5.0.

Results

The comparison of our series of patients with unselected, consecutive patients is summarized in Table 1. The patients analyzed with RT-PCR had tumors of higher clinical stage, higher histologic grade, and more lymphovascular invasion, and more often received chemotherapy and a total dose to the tumor bed higher than 60 Gy. All other patient, tumor, and treatment characteristics were similar.

Fifteen locoregional recurrences occurred (13 only local, 2 local and regional). Five-year and 10-year locoregional control rates were 82% (95% CI, 72% to 93%) and 70% (95% CI, 57% to 87%), respectively. All local recurrences occurred in the same quadrant as the primary tumor. Treatment of the local recurrence included salvage mastectomy with or without systemic therapy in 11 (73%) patients. In three (6%) patients, contralateral breast cancer developed. In 15 patients, distant metastases developed. Five-year and 10-year distant disease-free survival rates were 82% (95% CI, 72% to 93%) and 69% (95% CI, 56% to 85%), respectively. Five-year and 10-year overall survival rates were 85% (95% CI, 76% to 95%) and 78% (95% CI, 65% to 92%), respectively. In no patient did a second non-breast cancer develop.

Because regional recurrence was always accompanied by local recurrence, risk factors for local and locoregional recurrence were the same (data not shown). In univariate analysis, no clinicopathologic prognostic factor was found to be significantly associated with decreased locoregional control (Table 2). In univariate analysis, three gene-expression levels were found to be significantly associated with an increased locoregional recurrence rate: low ERβ, low aromatase, high GATA3 (Table 3). Two other factors were associated only with a trend (P < 0.10): low HER1 and SKP2. Multivariate analysis was performed, taking the factors associated (P < 0.10) with locoregional recurrence into account in a Cox model (that is, 5 RT-PCR features (ERβ, HER1, GATA3, SKP2 and aromatase). Only low tumor expression of aromatase was significantly associated with an increased locoregional recurrence rate (P = 0.003; relative risk = 0.49; 95% CI, 0.29 to 0.82). Locoregional-free recurrence interval according to the level of intratumoral expression of aromatase with a cut-off at 5 is displayed in Figure 1. GATA-3 was associated with a trend toward an increased locoregional recurrence rate (P = 0.06; relative risk = 1.49; 95% CI, 0.02 to 2.39).

Figure 1

Locoregional-free recurrence interval according to the level of intratumoral expression of aromatase

Locoregional-free recurrence interval according to the level of intratumoral expression of aromatase.

Table 2

Univariate analysis of clinical and histopathologic prognostic factors for locoregional control

rt-PCR series

10-year LRC (%)

[95% CI]

P ^a

RR [95% CI]

Family history of breast cancer

1.0

70 [56–87]

Yes

1 [0.22–4.59]

Age (continuous variable)

0.38

0.94 [0.83–1.07]

Age (years; dummy variable)

0.57

> 35

73 [59–90]

≤ 35

60 [32–100]

1.37 [0.47–4.0]

Clinical tumor stage

0.71

cT1

70 [53–91]

cT2

74 [56–96]

0.81 [0.28–2.39]

Histologic T stage

0.55

pT1

70 [54–91]

pT2

70 [49–100]

0.71 [0.24–2.13]

Surgical margin

0.70

≥ 3 mm

61 [43–86]

< 3 mm

81 [64–100]

0.81 [0.27–2.37]

Lymphovascular invasion

0.78

Absent

71 [56–91]

Present

71 [50–100]

1.18 [0.36–3.87]

Histologic N stage

0.54

pN0

63 [47–84]

pN1

78 [53–100]

0.49 [0.13–1.77]

pN2

80 [52–100]

0.77 [0.17–3.52]

Histologic type

0.14

Ductal

75 [63–90]

Lobular

33 [7–100]

2.61 [0.72–9.53]

Estrogen receptors (ERs)

0.14

ER^-

90 [73–100]

ER⁺

63 [47–85]

4.28 [0.55–33.27]

Progesterone receptors (PR)

0.84

PR^-

71 [43–100]

PR⁺

69 [53–88]

1.18 [0.26–5.27]

Hormone receptors (HRs)

0.48

ER^-and PR^-

86 [63–100]

ER⁺or PR⁺

67 [52–86]

2.1 [0.06–3.72]

Histopathologic index

0.59

Grade 1–2

64 [35–100]

Grade 3

73 [59–91]

0.70 [0.39–5.19]

Systemic therapy

0.35

None

54 [33–86]

ChT

85 [70–100]

0.47 [0.15–1.49]

ChT + HT

43 [11–100]

0.48 [0.12–1.90]

Total RT dose^b

0.27

≥ 60 Gy

76 [59–97]

< 60 Gy

50 [28–88]

1.8 [0.63–5.15]

NA = not applicable; CI = confidence interval; RR = relative risk for locoregional recurrence. ^aP values pertain to the relative risks. ^bTotal dose of radiotherapy to the tumor bed (whole breast dose + boost dose).

Table 3

Univariate and multivariate analyses of gene-expression prognostic factors (as continuous variables) for locoregional recurrences

Gene expression by quantitative rt-PCR (continuous variables)

Univariate analysis

Multivariate analysis

HR (RR)

[95% CI]

HR (RR)

[95% CI]

Estrogen receptor-α

1.22 [0.90–1.66]

0.15

Estrogen receptor-β

0.61 [0.39–0.95]

0.04

0.54

Progesterone receptor

1.13 [0.88–1.45]

0.32

HER1

0.65 [0.37–1.13]

0.10

0.72

HER2

0.89 [0.64–1.25]

0.49

HER3

1.06 [0.55–2.02]

0.87

HER4

1.22 [0.91–1.63]

0.14

MKI 67

0.89 [0.61–1.30]

0.56

Cyclin E1

0.81 [0.48–1.36]

0.43

Cyclin E2

0.75 [0.49–1.14]

0.18

GATA 3

1.61 [0.95–2.73]

0.04

1.49 [0.92–2.39]

0.06

IGF1R

1.41 [0.93–2.14]

0.13

NCoR

0.90 [0.38–2.15]

0.82

NCoA3/AIB1

0.51 [0.21–1.22]

0.14

CYP19 (aromatase)

0.48 [0.28–0.80]

0.003

0.49 [0.29–0.82]

0.003

Skp2

0.52 [0.23–1.17]

0.10

0.18

CAF-1 p150

1.00 [0.48–2.12]

1.00

CAF-1 p60

1.00 [0.97–1.03]

0.94

RR = relative risk for locoregional recurrence calculated with the nonparametric Cox model; CI = confidence interval; HER = human epidermal receptor. Because of lack of prognostic values of the clinical variables in univariate analysis, none was entered in the multivariate analysis.

Associations between gene expressions and locoregional recurrences were looked for in two publicly available microarray studies by Kreike and colleagues 23 and Nuyten and associates 24 that gave information regarding the age of the young women (48 and 35 patients with 18 and 7 local recurrences, respectively) treated with breast-conserving treatments with whole-breast radiotherapy. They both failed to show an association between the gene expression of either aromatase or gata3 and local recurrences [see Additional data file 2].

Additional file 2

Provides information from publicly available gene-expression studies of women, younger than 40 years old, treated with breast conserving treatments and whole breast radiotherapy.

Click here for file

Discussion

This study was based on a series of 53 premenopausal women younger than 40 years with a long (10 years) follow-up treated at the Institut Curie for invasive breast cancer with primary breast-conserving surgery. Given the specificity of the hormonal environment in young patients, genes directly or indirectly involved in hormone and growth-factor signaling pathways were analyzed. In addition to the usual clinical and histopathologic features, the levels of expression of 17 candidate genes for an association with locoregional control also were examined.

To verify that these findings could be extrapolated, we compared the characteristics of patients and tumors of the 53 patients included in this study with the 257 other patients treated over the same period (1988 through 1999) at the Institut Curie who met the same inclusion criteria (age, medical history, therapeutic sequence) (Table 1). This comparison showed that this RT-PCR series consisted of patients with tumors with a more advanced clinical stage, higher histologic grade, and more lymphovascular invasion and who therefore more frequently received chemotherapy and a total dose to the tumor bed higher than 60 Gy. All other patient, tumor, and treatment characteristics were comparable. It could be hypothesized that these differences were due, at the time of this study (1988 to 1999), to technical parameters: tissue samples for frozen storage would be easier to obtain from larger tumors.

In this series of 53 patients, neither clinical nor pathologic features were associated with increased locoregional recurrence rates. Very young age, in particular, was not significantly associated with increased locoregional recurrences, in contrast with a previous study of unselected young patients from the Institut Curie 6.

RNA expressions of 17 target genes were studied here with RT-PCR for their association with locoregional control. They are involved in hormone signaling, either directly (ERα, ERβ, PR, NCoA3/AIB1, NCoR, GATA3, CYP19/aromatase, SKP2) or indirectly (HER family, IGF1-R) or in proliferation (MKi67, cyclins E, and CAF-1). Classic hormone receptors (ERα, PR) have been extensively explored in breast cancers 10. In contrast with previously published studies 25, a high frequency of young women in this series had positive hormone receptors (ERα and PR in 79% and 81% of cases with immunohistochemistry, and in 75% and 82% of cases with RT-PCR).

Among the genes activated in response to ER activation (NCoA3/AIB1, NCoR, SKP2, GATA3, CYP19/aromatase), only the absence of CYP19/aromatase was significantly associated with an increased locoregional recurrence rate (P = 0.003; relative risk = 0.49; 95% CI, 0.29 to 0.82). Given the small size of this series, it also is noteworthy that the level of GATA-3 was associated with a trend toward an increased locoregional recurrence rate (P = 0.06; relative risk = 1.49; 95% CI, 0.02 to 2.39). CYP19/aromatase plays a critical role in breast cancer development by converting androgen into estrogens. CYP19/aromatase mRNA and protein have previously been detected in both tumor stroma and parenchymal cells in breast cancer tissue 26. Previous data demonstrated that CYP19/aromatase mRNA is correlated with CYP19/aromatase enzymatic activity in cultured breast tumor fibroblasts 27. The importance of in situ estrogen production has been demonstrated in breast carcinoma 28, where it is higher than that in normal breast tissue 2930. Inhibition of the aromatase pathway is considered to be clinically useful to reduce progression of breast tumors in postmenopausal women 31. An apparently paradoxic finding was that a low level of CYP19/aromatase transcripts in this population of young, premenopausal patients was significantly associated with an increased locoregional recurrence rate. However, preclinical studies demonstrated that estrogens or estrogenic compounds repress the transcriptional control of CYP19/aromatase 3233. It can be hypothesized that, in premenopausal women, who already have high plasma estradiol levels, the level of aromatase transcripts may be inversely correlated with the plasma estradiol level 34. Low tumor expression of CYP19/aromatase would reflect a high level of circulating estrogen and would therefore be associated with poor outcome. This finding is supported by a study by Zhang and colleagues 35 in a series of 162 invasive ductal breast carcinomas that showed that aromatase mRNA levels were lower in patients younger than 50 years, with tumors larger than 2 cm and with axillary lymph node involvement 35.

However, the association between tumor aromatase expression and outcome remains controversial; no consistent correlation between aromatase immunoreactivity, activity, or mRNA level and known clinicopathologic factors or outcome has been conclusively reported 353637383940. Some authors have even reported, in univariate analysis, a correlation between high tumor aromatase activity and poor outcome 3740.

The association between tumor aromatase activity or expression and estrogen receptors is also very controversial, as some authors have reported a positive correlation 353640, whereas others have reported no correlation 37, or even an inverse correlation 39. In the present series, no correlation was observed between aromatase expression and either age, hormone receptors, or histologic grade (data not shown).

GATA3 is a transcription factor involved in human growth and differentiation. Gene-expression profiling has shown that GATA3 is highly expressed in luminal A and B subtypes of cancer and closely related to ERα 414243. This study, the first to our knowledge to use quantitative RT-PCR in a population of young patients, showed a trend toward an association between a high level of GATA3 expression and higher locoregional recurrence rates. The favorable outcome in very young patients with ER-positive breast cancers is a controversial issue. Aebi and associates 44 reported a series of 3,500 premenopausal women treated in four randomized trials from the International Breast Cancer Study Group (IBCSG). They found that younger patients with ER-positive tumors had a significantly poorer disease-free survival than did younger patients with ER-negative tumors.

In addition, conflicting data have been published regarding the independent prognostic value of GATA3 in unselected breast cancer. Some studies found GATA3 to be associated with good prognosis 4546. The study by Voduc and colleagues 47 found no independent prognostic value of GATA3 in 3,119 breast cancer patients with immunohistochemistry on tissue microarrays.

ERβ expression in breast (normal and tumor) and the relation between ERβ and other clinicopathologic features and its role in hormone therapy have been extensively investigated (recently reviewed by Zhao and others 48). A consensus seems to have been reached regarding the protective role of ERβ against breast cancer development. No clinically validated cut-off has been defined for ERβ ; transcript levels are generally lower in tumor tissue than in normal tissue. The loss of ERβ expression by promoter methylation, frequently observed in breast tumors, has led to the hypothesis that ERβ is a possible tumor-suppressor gene 49. The ERβ transcript level in breast cancers was analyzed in a recent study by Anders and colleagues 50. With clinically annotated microarray data from 200 early-stage breast cancers in women, they observed that the ERβ transcript level was statistically lower in young women (younger than 45 years) than in older women (older than 65 years; P = 0.02). As expected from these data, the present study confirmed the protective effect of ERβ in terms of locoregional control.

In younger women with breast cancer, a higher incidence of growth-factor receptors, namely HER2, was observed both in terms of protein expression, associated with a more-aggressive phenotype 5152, and of HER2 gene expression with no predictive value for DFS 50.

Interestingly, the HER family and IGF1R have been shown to be intertwined with the estrogen-mediated signaling pathway 53.

Hormone receptors and growth-factor receptors act as mitogens, promoting cell proliferation in normal tissue and in breast carcinomas. Strong evidence suggests that IGF1, HER, and estrogen-mediated signaling are closely connected 54. HER1 (or EGFR) has been described as both a prognostic marker and a predictor of hormone-therapy resistance in breast cancer 55. However, downregulation of EGFR has never been explored. The present study showed a trend toward an impact of low EGFR expression on locoregional recurrence in young women. No other effect was observed for any of the other growth-factor receptors.

Proliferation markers, included in the definition of histologic grading 14, are also prognostic factors associated with poor outcome, high locoregional recurrence rates 556, and predictive factors of the response to chemotherapy 1557. The levels of gene expression of two subunits of CAF-1 (p150 and p60) 58 and other proliferation markers (the classic KI67, cyclins E1 and E2, and SKP2) were well correlated in the present study (data not shown) but were not associated with a higher risk of locoregional recurrence. One explanation could be a lack of statistical power, as most patients (68%) in this series had grade 3 tumors.

The search for a signature associated with local recurrence for premenopausal women treated with breast-conserving therapies by using high-throughput gene-expression analyses is ongoing 232459. The fact that gene-expression levels of neither CYP19/aromatase nor GATA3 were found to be associated with risk of local recurrences in the two studies 2324 that gave information about the patients' age could reflect either the lower sensitivity of microarray data compared with RT-PCR or a difference in the studied populations. Correlations between quantitative RT-PCR measures and microarray data have been deceiving. Koscielny and associates 60 investigated the correlation between gene expressions evaluated by the qRT-PCR and microarrays in 42 colon tumors. Of 39 genes randomly selected for analysis by qRT-PCR, seven showed a correlation between microarrays and quantitative RT-PCR of 0.5 or less. The authors concluded that microarray and qRT-PCR data are not deterministically related and therefore not interchangeable.

Conclusions

The present results highlight the role of estrogen-signaling pathways, mainly CYP19/aromatase, GATA3, and ERβ, in the risk of recurrence in young women with breast cancer. These preliminary results must be confirmed in a larger prospective study. One hypothesis would be that the higher the level of circulatory estrogen, the higher the risk of locoregional recurrence, whereas low tumor expression of aromatase and high tumor expression of GATA3 would reflect only the high plasma estrogen levels. However, recent data also give credit to the fact that breast cancer arising in young women is a distinct biologic entity driven by specific oncogenic pathways 5061. In conclusion, our results demonstrate that the hormone status is of extreme importance for the prognosis of premenopausal breast cancer patients. Although these findings must be confirmed in a larger prospective study on patients younger than 40 years, assessing the expression levels of hormone-related genes, among which aromatase, might contribute to predict disease outcome in young breast cancer patients.

Abbreviations

CAF-1: chromatin assembly factor 1; CCNE1: cyclin E1; CCNE1: cyclin E2; CI: confidence interval; CYP19: aromatase; EORTC: European Organisation for Research and Treatment of Cancer; ER: estrogen receptor; GATA3: GATA-binding protein 3; Gy: Gray; HER: human epidermal receptor; HR: hormone receptor; IGF1R: insulin-like growth factor 1-receptor; MKI67: antigen identified by monoclonal antibody ki-67; NCoA3/AIB1: nuclear receptor co-activator 3; NCoR: nuclear receptor co-repressor; PR: progesterone receptor; Pts: patients; RR: relative risk; RT-PCR: reverse transcriptase polymerase chain reaction; SKP2: S-phase kinase-associated protein 2; TBP: TATA box-binding protein.

Competing interests

The authors declare that they have no competing interests.

Authors' contributions

MAB acquired clinical data and obtained funding for the study. MAB and PDC designed the study and drafted the manuscript. BSZ performed all histopathologic reviews. LDK, CTR, and CB performed all RT-PCR experiments and analyzed the data. AS and NS performed all statistical analyses. MAB, LDK, PDC, AD, NS, and AS analyzed and interpreted the data. MAB, LDK, PDC, AS, BSZ, AD, GA, PC, RS, NS, and AF helped in acquiring the data and critically revised the manuscript for important intellectual content. All authors read and approved the final manuscript.

Acknowledgements

The authors thank the patients who participated in this study and all members of the Institut Curie Breast Cancer Study Group who contributed to this study. They also thank Chantal Gautier for her assistance in patient data management and Bas Kreike, Adam Cook, and Anthony Saul for their helpful suggestions.

This research was financed by the Institut Curie, the "Courir pour la vie, Courir pour Curie" association, the "Odyssea" association, and the PHRC 2006 (AOM 06 149).

Conservation surgery for breast cancer as the preferred choice: a prospective analysis Temple WJ Russell ML Parsons LL Huber SM Jones CA Bankes J Eliasziw M J Clin Oncol 2006 24 3367 3373 10.1200/JCO.2005.02.7771 16849750 Effects of radiotherapy and of differences in the extent of surgery for early breast cancer on local recurrence and 15-year survival: an overview of the randomised trials Clarke M Collins R Darby S Davies C Elphinstone P Evans E Godwin J Gray R Hicks C James S MacKinnon E McGale P McHugh T Peto R Taylor C Wang Y Lancet 2005 366 2087 2106 16360786 Quality of life following breast-conserving therapy or mastectomy: results of a 5-year prospective study Engel J Kerr J Schlesinger-Raab A Sauer H Holzel D Breast J 2004 10 223 231 10.1111/j.1075-122X.2004.21323.x 15125749 Psychosocial outcomes of breast-conserving surgery versus mastectomy: a meta-analytic review Moyer A Health Psychol 1997 16 284 298 10.1037/0278-6133.16.3.284 9152708 Can patient-, treatment- and pathology-related characteristics explain the high local recurrence rate following breast-conserving therapy in young patients? Vrieling C Collette L Fourquet A Hoogenraad WJ Horiot JC Jager JJ Bing Oei S Peterse HL Pierart M Poortmans PM Struikmans H Bogaert Van den W Bartelink H Eur J Cancer 2003 39 932 944 10.1016/S0959-8049(03)00123-0 12706362 Age remains the first prognostic factor for loco-regional breast cancer recurrence in young (< 40 years) women treated with breast conserving surgery first Bollet MA Sigal-Zafrani B Mazeau V Savignoni A de la Rochefordiere A Vincent-Salomon A Salmon R Campana F Kirova YM Dendale R Fourquet A Radiother Oncol 2007 82 272 280 10.1016/j.radonc.2007.01.001 17287037 The impact of young age on locoregional recurrence after doxorubicin-based breast conservation therapy in patients 40 years old or younger: how young is "young"? Oh JL Bonnen M Outlaw ED Schechter NR Perkins GH Strom EA Babiera G Oswald MJ Allen PK Thames HD Buchholz TA Int J Radiat Oncol Biol Phys 2006 65 1345 52 16750315 Estrogens, progestogens, normal breast cell proliferation, and breast cancer risk Pike MC Spicer DV Dahmoush L Press MF Epidemiol Rev 1993 15 17 35 8405201 Mechanisms of hormone carcinogenesis: evolution of views, role of mitochondria Chen JQ Brown TR Yager JD Adv Exp Med Biol 2008 630 1 18 18637481 Estrogen carcinogenesis in breast cancer Yager JD Davidson NE N Engl J Med 2006 354 270 282 10.1056/NEJMra050776 16421368 Quantification of mRNA using real-time RT-PCR Nolan T Hands RE Bustin SA Nat Protoc 2006 1 1559 1582 10.1038/nprot.2006.236 17406449 A standard curve based method for relative real time PCR data processing Larionov A Krause A Miller W BMC Bioinformatics 2005 6 62 1274258 15780134 10.1186/1471-2105-6-62 TNM Classification of Malignant Tumours Sobin LH Wittekind C New York: Wiley-Liss 2002 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 Elston CW Ellis IO Histopathology 1991 19 403 410 10.1111/j.1365-2559.1991.tb00229.x 1757079 Proliferation markers predictive of the pathological response and disease outcome of patients with breast carcinomas treated by anthracycline-based preoperative chemotherapy Vincent-Salomon A Rousseau A Jouve M Beuzeboc P Sigal-Zafrani B Freneaux P Rosty C Nos C Campana F Klijanienko J Al Ghuzlan A Sastre-Garau X Eur J Cancer 2004 40 1502 1508 10.1016/j.ejca.2004.03.014 15196533 [Recommendations for the immunohistochemical evaluation of hormone receptors on paraffin sections of breast cancer: Study Group on Hormone Receptors using Immunohistochemistry FNCLCC/AFAQAP: National Federation of Centres to Combat Cancer/French Association for Quality Assurance in Pathology] Balaton AL Coindre JM Collin F Ettore F Fiche M Jacquemier J Le Doussal V Mandard AM Mathieu MC Migeon C Simony-Lafontaine J Treilleux I Verriele V Vilain MO Zafrani B Ann Pathol 1996 16 144 148 8767687 Effect of age and radiation dose on local control after breast conserving treatment: EORTC trial 22881-10882 Antonini N Jones H Horiot JC Poortmans P Struikmans H Van den Bogaert W Barillot I Fourquet A Jager J Hoogenraad W Collette L Pierart M Hart G Bartelink H Radiother Oncol 2007 82 265 271 17126434 10.1016/j.radonc.2006.09.014 Recurrence rates after treatment of breast cancer with standard radiotherapy with or without additional radiation Bartelink H Horiot JC Poortmans P Struikmans H Bogaert Van den W Barillot I Fourquet A Borger J Jager J Hoogenraad W Collette L Pierart M N Engl J Med 2001 345 1378 1387 10.1056/NEJMoa010874 11794170 Inter-laboratory quality control for hormone-dependent gene expression in human breast tumors using real-time reverse transcription-polymerase chain reaction de Cremoux P Bieche I Tran-Perennou C Vignaud S Boudou E Asselain B Lidereau R Magdelenat H Becette V Sigal-Zafrani B Spyratos F Endocr Relat Cancer 2004 11 489 495 10.1677/erc.1.00808 15369450 Role of chemotherapy resistance genes in outcome of neuroblastoma de Cremoux P Jourdan-Da-Silva N Couturier J Tran-Perennou C Schleiermacher G Fehlbaum P Doz F Mosseri V Delattre O Klijanienko J Vielh P Michon J Pediatr Blood Cancer 2007 48 311 317 10.1002/pbc.20853 16609945 dbEST, htgs and nr data bases http://blast.ncbi.nlm.nih.gov Analysis of Survival Data Cox DR Oakes D London: Chapman & Hall 1984 Local recurrence after breast-conserving therapy in relation to gene expression patterns in a large series of patients Kreike B Halfwerk H Armstrong N Bult P Foekens JA Veltkamp SC Nuyten DS Bartelink H Vijver van de MJ Clin Cancer Res 2009 15 4181 90 10.1158/1078-0432.CCR-08-2644 19470741 Predicting a local recurrence after breast-conserving therapy by gene expression profiling Nuyten DS Kreike B Hart AA Chi JT Sneddon JB Wessels LF Peterse HJ Bartelink H Brown PO Chang HY Vijver van de MJ Breast Cancer Res 2006 8 R62 1779489 17069664 10.1186/bcr1614 Effect of age on the survival of breast cancer patients Holli K Isola J Eur J Cancer 1997 33 425 428 10.1016/S0959-8049(97)89017-X 9155527 Aromatase localization in human breast cancer tissues: possible interactions between intratumoral stromal and parenchymal cells Miki Y Suzuki T Tazawa C Yamaguchi Y Kitada K Honma S Moriya T Hirakawa H Evans DB Hayashi S Ohuchi N Sasano H Cancer Res 2007 67 3945 3954 10.1158/0008-5472.CAN-06-3105 17440110 Aromatase activity and expression in breast cancer and benign breast tissue stromal cells Santner SJ Pauley RJ Tait L Kaseta J Santen RJ J Clin Endocrinol Metab 1997 82 200 208 10.1210/jc.82.1.200 8989259 In situ aromatization enhances breast tumor estradiol levels and cellular proliferation Yue W Wang JP Hamilton CJ Demers LM Santen RJ Cancer Res 1998 58 927 932 9500452 Breast tumor aromatase: functional role and transcriptional regulation Chen S Zhou D Okubo T Kao YC Yang C Endocr Relat Cancer 1999 6 149 156 10.1677/erc.0.0060149 10731103 Comparison of estrogen concentrations, estrone sulfatase and aromatase activities in normal, and in cancerous human breast tissues Chetrite GS Cortes-Prieto J Philippe JC Wright F Pasqualini JR J Steroid Biochem Mol Biol 2000 72 23 27 10.1016/S0960-0760(00)00040-6 10731634 Aromatase inhibitors in breast cancer Smith IE Dowsett M N Engl J Med 2003 348 2431 2442 10.1056/NEJMra023246 12802030 A positive feedback pathway of estrogen biosynthesis in breast cancer cells is contained by resveratrol Wang Y Ye L Leung LK Toxicology 2008 248 130 135 10.1016/j.tox.2008.03.017 18462857 Aromatase gene expression and regulation in the female rat pituitary Galmiche G Corvaisier S Kottler ML Ann N Y Acad Sci 2006 1070 286 292 10.1196/annals.1317.028 16888180 Regulated CYP19 aromatase transcription in breast stromal fibroblasts Pauley RJ Santner SJ Tait LR Bright RK Santen RJ J Clin Endocrinol Metab 2000 85 837 846 10.1210/jc.85.2.837 10690899 Quantitative determination, by real-time reverse transcription polymerase chain reaction, of aromatase mRNA in invasive ductal carcinoma of the breast Zhang Z Yamashita H Toyama T Omoto Y Sugiura H Hara Y Wu X Kobayashi S Iwase H Breast Cancer Res 2003 5 R250 R256 314416 14580261 10.1186/bcr657 Significance of aromatase activity in human breast cancer Miller WR Hawkins RA Forrest AP Cancer Res 1982 42 3365s 3368s 7083201 Intratumoral aromatase as a prognostic factor in human breast carcinoma Silva MC Rowlands MG Dowsett M Gusterson B McKinna JA Fryatt I Coombes RC Cancer Res 1989 49 2588 2591 2713843 Prognostic significance of aromatase and estrone sulfatase enzymes in human breast cancer Evans TR Rowlands MG Silva MC Law M Coombes RC J Steroid Biochem Mol Biol 1993 44 583 587 10.1016/0960-0760(93)90263-V 8476770 Aromatase and 17 beta-hydroxysteroid dehydrogenase type 1 in human breast carcinoma Sasano H Frost AR Saitoh R Harada N Poutanen M Vihko R Bulun SE Silverberg SG Nagura H J Clin Endocrinol Metab 1996 81 4042 4046 10.1210/jc.81.11.4042 8923858 The role of aromatase and 17-beta-hydroxysteroid dehydrogenase type 1 mRNA expression in predicting the clinical outcome of human breast cancer Salhab M Reed MJ Al Sarakbi W Jiang WG Mokbel K Breast Cancer Res Treat 2006 99 155 162 10.1007/s10549-006-9198-8 16541304 Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications Sorlie T Perou CM Tibshirani R Aas T Geisler S Johnsen H Hastie T Eisen MB Rijn van de M Jeffrey SS Thorsen T Quist H Matese JC Brown PO Botstein D Eystein Lonning P Borresen-Dale AL Proc Natl Acad Sci USA 2001 98 10869 10874 58566 11553815 10.1073/pnas.191367098 Repeated observation of breast tumor subtypes in independent gene expression data sets Sorlie T Tibshirani R Parker J Hastie T Marron JS Nobel A Deng S Johnsen H Pesich R Geisler S Demeter J Perou CM Lonning PE Brown PO Borresen-Dale AL Botstein D Proc Natl Acad Sci USA 2003 100 8418 8423 166244 12829800 10.1073/pnas.0932692100 Gene expression profiling of primary breast carcinomas using arrays of candidate genes Bertucci F Houlgatte R Benziane A Granjeaud S Adelaide J Tagett R Loriod B Jacquemier J Viens P Jordan B Birnbaum D Nguyen C Hum Mol Genet 2000 9 2981 2991 10.1093/hmg/9.20.2981 11115842 Is chemotherapy alone adequate for young women with oestrogen-receptor-positive breast cancer? Aebi S Gelber S Castiglione-Gertsch M Gelber RD Collins J Thurlimann B Rudenstam CM Lindtner J Crivellari D Cortes-Funes H Simoncini E Werner ID Coates AS Goldhirsch A Lancet 2000 355 1869 1874 10.1016/S0140-6736(00)02292-3 10866443 Identification of GATA3 as a breast cancer prognostic marker by global gene expression meta-analysis Mehra R Varambally S Ding L Shen R Sabel MS Ghosh D Chinnaiyan AM Kleer CG Cancer Res 2005 65 11259 11264 10.1158/0008-5472.CAN-05-2495 16357129 Estrogen-regulated genes predict survival in hormone receptor-positive breast cancers Oh DS Troester MA Usary J Hu Z He X Fan C Wu J Carey LA Perou CM J Clin Oncol 2006 24 1656 1664 10.1200/JCO.2005.03.2755 16505416 GATA-3 expression in breast cancer has a strong association with estrogen receptor but lacks independent prognostic value Voduc D Cheang M Nielsen T Cancer Epidemiol Biomarkers Prev 2008 17 365 373 10.1158/1055-9965.EPI-06-1090 18268121 Estrogen receptor beta: an overview and update Zhao C Dahlman-Wright K Gustafsson JA Nucl Recept Signal 2008 6 e003 2254331 18301783 DNA hypermethylation: when tumour suppressor genes go silent Garinis GA Patrinos GP Spanakis NE Menounos PG Hum Genet 2002 111 115 127 10.1007/s00439-002-0783-6 12189484 Young age at diagnosis correlates with worse prognosis and defines a subset of breast cancers with shared patterns of gene expression Anders CK Hsu DS Broadwater G Acharya CR Foekens JA Zhang Y Wang Y Marcom PK Marks JR Febbo PG Nevins JR Potti A Blackwell KL J Clin Oncol 2008 26 3324 3330 10.1200/JCO.2007.14.2471 18612148 Differential expression of prognostic factors and effect on survival in young (< or = 40) breast cancer patients: a case-control study Hartley MC McKinley BP Rogers EA Kalbaugh CA Messich HS Blackhurst DW Lokey JS Trocha SD Am Surg 2006 72 1189 1194 discussion 1194–1185. 17216817 C-erbB-2 overexpression and survival in early onset breast cancer Agrup M Stal O Olsen K Wingren S Breast Cancer Res Treat 2000 63 23 29 10.1023/A:1006498721508 11079156 Interactions between estrogen and insulin-like growth factor signaling pathways in human breast tumor cells Hamelers IH Steenbergh PH Endocr Relat Cancer 2003 10 331 345 Review. 12790794 10.1677/erc.0.0100331 Estrogen and growth factor signaling pathway: basic approaches for clinical application Hayashi S Sakamoto T Inoue A Yoshida N Omoto Y Yamaguchi Y J Steroid Biochem Mol Biol 2003 86 433 442 10.1016/S0960-0760(03)00354-6 14623541 The biology of antihormone failure in breast cancer Nicholson RI Gee JM Knowlden J McClelland R Madden TA Barrow D Hutcheson I Breast Cancer Res Treat 2003 80 suppl 1 S29 S34 discussion S35. 10.1023/A:1025467500433 14535532 Risk factors for local recurrence after breast-conserving therapy for invasive carcinomas: a case-control study of histological factors and alterations in oncogene expression Elkhuizen PH Voogd AC Broek van den LC Tan IT van Houwelingen HC Leer JW Vijver van de MJ Int J Radiat Oncol Biol Phys 1999 45 73 83 10477009 Mitotic anthracycline-based chemotherapy in patients with early breast cancer Andre F Khalil A Slimane K Massard C Mathieu MC Vignot S Assi H Delaloge S Spielmann M J Clin Oncol 2005 23 2996 3000 10.1200/JCO.2005.08.046 15860855 Chromatin assembly factor-1, a marker of clinical value to distinguish quiescent from proliferating cells Polo SE Theocharis SE Klijanienko J Savignoni A Asselain B Vielh P Almouzni G Cancer Res 2004 64 2371 2381 10.1158/0008-5472.CAN-03-2893 15059888 Gene expression profilers and conventional clinical markers to predict distant recurrences for premenopausal breast cancer patients after adjuvant chemotherapy Nimeus-Malmstrom E Ritz C Eden P Johnsson A Ohlsson M Strand C Ostberg G Ferno M Peterson C Eur J Cancer 2006 42 2729 2737 10.1016/j.ejca.2006.06.031 17023159 Validation of microarray data by quantitative reverse-transcriptase polymerase chain reaction Koscielny S Michiels S Boige V Hill C J Clin Oncol 2005 23 9439 9440 author reply 9440. 10.1200/JCO.2005.04.1004 16361648 Age-specific differences in oncogenic pathway deregulation seen in human breast tumors Anders CK Acharya CR Hsu DS Broadwater G Garman K Foekens JA Zhang Y Wang Y Marcom K Marks JR Mukherjee S Nevins JR Blackwell KL Potti A PLoS ONE 2008 3 e1373 2148101 18167534 10.1371/journal.pone.0001373