Comparison of gene expression profiles in core biopsies and corresponding surgical breast cancer samples

Between June 2004 and June 2005, 22 consecutive breast cancer patients entered this study, for which both CBs and surgical tumor tissue specimens (STs) were available. This study was undertaken at the Women's University Hospital Basel, Switzerland, and approved by the local institutional review board (EKBB permission Nr. 81/04). Written informed consent was obtained from all patients.

All CBs (14-gauge needle, Magnum^® Core high speed, Bard Medica, Karlsruhe, Germany) were obtained under sonographic guidance (HDI 5'000 Sono CT^®, Philips, Zurich, Switzerland) under local anesthesia through a skin incision in a sterile field. Five biopsy specimens were taken routinely for each patient. Two biopsy specimens were divided longitudinally; two halves to be used for molecular examination were, within one minute, stored in RNAlater^®-solution (Qiagen, Basel, Switzerland), while the other two halves as well as the other three biopsies were immediately put in formaldehyde and sent for histological examination to the Institute of Pathology, University Hospital Basel.

After establishing the diagnosis of breast cancer, all patients underwent breast surgery with sentinel node lymphonodectomy or axillary lymphonodectomy if indicated. All surgical samples were examined by the same pathologist. If the tumor tissue was larger than 0.5 cm in diameter at the intra-operative frozen section, a representative piece containing more than 60% tumor cells was cryopreserved within five minutes and made available to the Stiftung Tumorbank Basel for molecular examination. The rest of the tumor tissue was embedded in paraffin for routine histological examination.

RNA expression levels of all 60 genes were detected using the same qrt-PCR method in 317 surgically excised breast cancer specimens [15] from patients undergoing primary surgery in 1992 to 1996 without previous examination by CB. All tissue samples were prepared by the pathologists as described above, all samples contained more than 60% tumor cells and they were cryoperserved within five minutes. The Stiftung Tumorbank Basel was subsequently responsible for collection, storage at -70°C and analysis.

Detailed procedures have been published elsewhere [16]. In brief, RNA was extracted using the RNeasy Mini Kit (Qiagen, Inc., Valencia, CA, USA), quantified and quality-checked on a Bioanalyzer 2100 (RNA 6000 Nano LabChip-Kit, Agilent Technologies, New Castle, DE, USA). High quality RNA samples were reverse-transcribed (10 mM DDT, 1 μg of hexamer primers, 2 U of MMLV Reverse Transcriptase (Invitrogen, Basel, Switzerland), 40 U of RNasin (Promega, Wallisellen, Switzerland), 0.5 mM of each dNTP (Promega), 1× reaction buffer). PCR primers were designed to be cDNA specific and ordered at GeneScan Europe (Freiburg, Germany). PCR was performed in 40 cycles on an ABI Prism 7000 using 2× SYBR Green I Master Mix (Applied Biosystems, Forster City, CA, USA) in a final volume of 25 μl. Relative quantities (ΔCt) were obtained by normalization against ribosomal 18S RNA, and standardization was achieved with Human Universal Standard RNA (Stratagene Europe, Amsterdam, The Netherlands). The 60 genes quantitatively assessed are listed in Table 1.

The same amount of RNA was used for the GEP analysis of each sample. For statistical analysis, ΔCt expression values of each gene were obtained by normalizing the raw gene values to 18S rRNA as a reference gene.

Cluster and TreeView programs were used to perform unsupervised hierarchical clustering of samples and genes (Spearman correlation, average linkage) [17]. Spearman correlation coefficients were calculated to compare the GEPs of all paired samples. Differentially expressed genes were identified with the paired two-sample t-test. The Mann-Whitney U-test was used to compare median expression values of genes among different subgroups of patients. All statistical analyses were carried out at 5% level of significance and performed with S-Plus software (Version 6.1, Insightful Corporation, Seattle, WA, USA).

The mean age of the 22 patients was 63 years (range: 34 to 81 years). The period between CB and final surgery ranged from 1 to 23 days. The mean tumor diameter was 2.9 cm (range: 0.9 to 9 cm). Clinicopathological characteristics of each patient are listed in Table 1. Of note, one paired tissue sample was not further evaluated due to poor RNA quality, and RNA extracted from the two halves of the CB was not pooled in three cases but analyzed separately.

As shown in Figure 1, unsupervised hierarchical clustering revealed that paired CB and ST generally clustered together; in only four cases (patients 4, 10, 11, and 12) did the GEP of the CB not agglomerate with the profile of the respective ST. Interestingly, the two separate CBs taken from patient 11 were very similar to each other, although they differed from their ST GEP. The gene dendrogram of the cluster analysis also revealed that samples agglomerated in two main groups according to their respective estrogen receptor status, reconfirming the representative value of this study population.

Subsequent analysis of paired CBs and STs confirmed the high correlation between all samples (r_s from 0.86 to 0.98, all p < 0.001) for each patient. A scatter plot of two representative examples of a paired CB/ST is displayed in Figure 2. The differences in paired GEPs does not seem to be related to the timeframe between CB and surgery or to any other clinicopathological parameters (Table 1).

The comparison of the expression levels of individual genes by means of paired t-test showed no significant difference between CB and ST with the exception of four genes. Plasminogen activator inhibitor 1 (PAI-1; also known as SERPINE1) was significantly higher expressed (lower ΔCt values) in STs compared to CBs (p < 0.001, Table 1). Similar differences, although less pronounced, were observed for cyclooxygenase 2 (COX-2; also known as PTGS2; p < 0.001), urokinase plasminogen activator receptor (uPAR; also known as PLAUR; p = 0.003) and matrix metalloproteinase 1 (MMP1; p = 0.03). The increase in the expression of these genes was not related to the timeframe between CB and surgery. All other genes were very similarly expressed in paired CB/ST as shown in Figure 3. Table 1 lists differences in RNA expression values of PAI-1, COX-2 and ERBB2 for paired samples of each patient.

To verify whether the higher expression levels of PAI-1 and COX-2 observed in STs could have been induced by the preceding CB procedure, we compared the expression levels of the same genes in an independent population of 317 primary breast cancer patients. These samples were investigated with the same qrt-PCR technique but were from patients from which no CB had been taken prior to surgery. As shown in Figure 4, the expression levels of PAI-1 and COX-2 measured in the independent STs without CB were found to be very similar to the levels detected in CBs and significantly different from those detected in STs after CB. Moreover, no variation at all was observed for the remaining genes as illustrated by the expression levels of ERBB2 as an example.