Prior to any therapy, between 10 and 20 ml of bone marrow were aspirated from the anterior iliac crest of 254 primary breast cancer patients undergoing surgical treatment from 2005 to 2007 at the Department of Gynecology and Obstetrics, University Hospital Tuebingen, Germany.

The characteristics of the patients are shown in Table 1. All specimens were obtained after written informed consent was given and were collected using protocols approved by the institutional review board (114/2006A). Tumour cell isolation and detection was performed based on the recommendations for standardised tumour cell detection 10. BM samples were separated by density centrifugation over Ficoll with a density of 1.077 g/ml (Biochrom, Germany). If necessary red blood cells were lysed with lysis buffer (155 mM NH₄Cl, 10 mM KHC0₃, 0.1 mM EDTA pH 7.2). Using a cytocentrifuge (Hettich, Tuttlingen, Germany), 10⁶ mononuclear cells were spun onto a glass slide. The slides were air-dried overnight at room temperature. For detection and characterisation of DTCs, slides were fixed in a 0.5% neutral buffered formalin solution for 10 minutes. Control cytospins with ERα-positive MCF-7 cells were prepared, stored and fixed in the same way to ensure that ERα negativity of a patient's sample was not due to a handling error. Two slides per patient was analysed for the presence of DTCs (2 × 10⁶ cells per patient).

For establishing the ERα staining procedure, preparations of breast cancer cell lines MCF-7 and SKBR3 mixed with either BM or peripheral blood mononuclear cells (PBMCs) from a healthy volunteer were used (Figure 1). To optimise the staining procedure, all relevant parameters of the protocol were evaluated as follows: types of primary ERα antibodies used were monoclonal mouse antibodies (NCL-L-ER-6F11, Novocastra Laboratories, UK), polyclonal rabbit antibodies (H-184, Santa Cruz Biotechnology, Inc., CA) and monoclonal rabbit antibodies (SP1, Lab Vision, CA); antibody dilutions used were 1:200, 1:100, 1:50 and 1:25 made with DAKO Antibody Diluent (1% BSA in PBS, 0.1% Tween 20); incubation times for primary and secondary antibodies were 30, 45 and 60 minutes; selection of secondary antibodies was with Tex-Red labelled horse anti-mouse AB (Vector Laboratories, Inc., CA), Tex-red labelled goat anti-rabbit AB (CB 11, Biogenex, CA) and Alexa Fluor 594 labelled goat anti rabbit AB (Molecular Probes, Invitrogen, CA); cell fixation was 10 minutes of acetone at 4°C, 100% ethanol for 10 minutes or 0.5% neutral buffered formalin solution for 10 minutes, all three fixations at room temperature. The optimal ERα staining (low background, strong nuclear staining, no cytoplasmic staining) was determined to be as indicated below.

After an initial washing step with PBS (Sigma, Munich, Germany), cells were blocked for 30 minutes with normal goat serum (Dako, Glostrup, Denmark) at a 1:10 dilution. The automated double immunofluorescence staining procedure was performed on the DAKO Autostainer using the monoclonal rabbit ERα-antibody SP1 (dilution 1:25, Lab Vision, Fremont, CA, USA) for 60 minutes and secondary detection with a goat anti-rabbit antibody, labelled with Alexa Fluor 594 (1:100, Invitrogen Molecular Probes, Carlsbad, CA, USA) for 30 minutes. Cytospins were then incubated with a pan-cytokeratin (CK) antibody (C11) directly conjugated to fluorescein isothiocyanate (FITC) (1:100, Sigma, Munich, Germany) for 30 minutes. This monoclonal antibody recognises human CKs 4, 5, 6, 8, 10, 13 and 18. Counterstaining was performed with 4'6-diamidino-2-phenylindole (DAPI) in mounting media (Vector Laboratories, Burlingame, CA, USA). Preparations of the breast cancer cell line MCF-7 mixed with PBMCs from a healthy volunteer served as a positive control for CK and ERα staining. ERα negative control slides of SKBR-3/PBMC mixtures were also included with each batch of samples. Cytospins of PBMCs with no added tumour cells served as a negative control for both.

Slides were manually analysed for the presence of tumour cells using a computerised fluorescence microscope Axiophot (×40 oil immersion objectives, Carl Zeiss Micro Imaging GmbH, Göttingen, Germany). To screen for ERα-positive tumour cells, a single-pass filter for individual fluorochromes, FITC, Texas Red or DAPI, and a dual-pass filter for FITC/Texas Red were used. Criteria for evaluation of immunostained cells were based on the criteria of the International Society of Hematotherapy and Graft Engineering Working group for standardisation of tumour cell detection and the consensus statements 1011. Criteria for ERα positivity were either moderate or intense staining of the entire nucleus. Slides were evaluated by two, or in doubtful cases three, independent investigators (TF, NK and ES).

Immunohistochemical analysis was performed either on core biopsies or surgical resection specimens. The tissue was fixed in 4.5% buffered formalin (pH 7.0) and embedded in paraffin. Immunohistochemical staining was performed on 3 to 5 μm thick sections using a commercially available ABC kit (Vectastain, Vector Laboratories, Burlingame, CA, USA). The ERα antibody (clone SP1) was diluted 1:200 in Tris-HCl (pH 7.5) and applied according to the manufacturer's instruction (DCS, Hamburg, Germany). 3,3'diaminobenzidine (DAB) was used as a chromogen. Finally, the slides were counterstained with haematoxylin and mounted for examination. For assessment of the ERα status, the percentage of cells with nuclear reactivity (score 0: none, 1: > 10%, 2: 10 to 50%, 3: 51 to 80%, 4: > 80%) and the intensity of ER staining (score 0: none, 1: weak, 2: moderate, 3: strong) was determined. ERα expression was scored semi-quantitatively using the Remmele-score (score nuclear staining × score intensity of ER staining). Tumours with a score of 2 or more were considered ERα positive.

A chi-squared test or Fisher's exact test was used to evaluate the relation between ERα-positive DTCs and clinicopathological factors. Statistical analysis was performed by SPSS, version 11.5 (SPSS Inc., Chicago, IL, USA). p < 0.05 was considered statistically significant.

A total of 254 patients were included in the study. Clinical data are shown in detail in Table 1. Of patients, 82% had ERα-positive primary tumours and DTCs were observed in 107 (42%) of them. Figure 2 shows the cytomorphology and immunophenotype of a representative DTC of a patient with breast cancer. As can be seen, the nuclear to cytoplasmic ratio is high, the nucleus has irregularities and the CK stains the cytoplasm at the periphery of the cell causing a ring-like appearance. These are all accepted morphological criteria for malignant cells. The number of DTCs ranged from 1 to 55 cells/patient (2 × 10⁶ mononuclear cells). In 38 of the 107 (35%) BM-positive patients, more than one DTC could be detected. No correlation was observed between positive BM status and any of the established prognostic markers including the ERα status of the primary tumour (Table 1).

ERα status of DTCs was simultaneously evaluated using a double immunofluorescence staining procedure. The majority of patients (88%) had ERα-negative tumour cells in BM (Table 2). ERα-positive tumour cells could only be detected in 13 of 107 (12%) patients with BM involvement. ERα-positive but CK-negative cells were not observed. Figure 3 shows ERα-positive tumour cells from different patients. As can be seen, the nuclei are strongly stained with the ER antibody.

Of the 107 patients, 38 had more than one DTC in the BM. Of these 38 patients, 28 had only ERα-negative tumour cells (Table 3). Heterogeneity of ERα expression could be detected in the remaining 10 (26%) patients (Figure 3i).

The ERα status of the primary tumour could be determined in all 107 patients with detectable DTCs in the BM. ERα positivity of the primary tumour was demonstrated in 88 (82%) of these patients. The concordance rate between ERα status of DTCs and primary tumour was 28%. Only 12 of the 88 (14%) patients with ERα-positive primary tumour had ERα-positive DTCs in the BM. In contrast, 18 of 19 (95%) patients with ERα-negative primary tumours also had ERα-negative DTCs (Table 2). The extent of ERα expression (negative, low, moderate or strong) of the primary tumour was not correlated to the ERα status of DTCs. The comparison of ERα expression between primary tumours and DTCs is summarised in Tables 2 and 3.