To the Editor: In a recent issue of Clinical Cancer Research, two articles (1, 2) addressed the use of PCR to detect micrometastases from breast cancer either in lymph nodes (1) or in peripheral blood (2). Both articles found mammaglobin (mam) to be an excellent marker for this. However, neither article, in our view, has appropriately considered the effect of the wide range of expression levels of mam in the primary tumor, and the association of mam expression with particular tumor characteristics, on these results.
Mikhitarian et al. (1) diluted the RNA of a metastasis-positive axillary lymph node with RNA from normal lymph nodes and used microarray analyses to find mam as the gene with highest sensitivity. In addition, TFF1 was found as an additional novel marker. However, for this experiment, a positive lymph node was chosen based on its 5.3 × 107–fold overexpression of mam. We have shown that in the primary breast tumor mam levels can vary over 10,000-fold (3). Thus, if Mikhitarian et al. had chosen a tumor that had a lower level of mam expression, other genes might have been found to have a better detection rate. Of note, the broad shoulder in the frequency distribution in Fig. 1 of their article is in agreement with the wide range of mam expression in primary tumors.
The authors discuss the fact that TFF1 expression in the primary tumor is positively associated with estrogen receptor and progesterone receptor status. We have described a similar, strong association of mam with estrogen receptor and progesterone receptor status. This then may account for the fact that TFF1 is found as a novel marker by these authors, as the tumor chosen for this experiment was selected based on its mam overexpression, and as both these markers are correlated with estrogen receptor status.
Reinholz et al. (2) used a PCR for mam to detect breast cancer cells in peripheral blood samples. Of the patients with breast cancer in this article, a majority (77%) had an estrogen receptor–positive tumor. This patient selection would favor the selection of mam as a sensitive marker for circulating breast cancer cells, considering the particular association of mam expression with just these tumor characteristics (3).
In Response: We thank Dr. Span for interest in our article (1) and for insightful comments. Although Dr. Span has correctly pointed out that there is wide range of mammaglobin (mam) expression levels in primary breast cancers, it still remains the single best marker for the detection of metastatic and micrometastatic breast cancer (2, 3). The frequency distribution analyses in our article confirm that the tissue specificity of mam is exquisite and there is almost no expression in normal lymph nodes (or peripheral blood; ref. 4). We have recently completed an interim analysis of a multi-institutional prospective cohort study (Minimally Invasive Molecular Staging of Breast Cancer) designed to assess the clinical relevance of molecular detection of micrometastatic disease in axillary lymph nodes (ALN) of breast cancer patients. In this study, mam was overexpressed in ALN from 79% of patients with pathology-positive ALN and in ALN from 23% of patients with pathology-negative ALN (3). Although this is higher than any other molecular marker, the heterogeneity of gene expression in primary breast cancer is a good argument for the use of multimarker panels.
Although there is evidence that mam expression is higher in estrogen receptor–positive tumors (5), mam also seems to be an excellent molecular marker for metastatic disease derived from estrogen receptor–negative tumors. Results of the Minimally Invasive Molecular Staging of Breast Cancer study reveal that mam is overexpressed in ALN from 80% of patients with pathology-positive ALN and estrogen receptor–positive tumors (n = 108) and in ALN from 72% of patients with pathology-positive ALN and estrogen receptor–negative tumors (n = 29).
With respect to Dr. Span's comment regarding tissue selection, we chose an ALN containing metastatic breast cancer that highly expressed mam as an internal positive control for the novel dilutional microarray strategy. Identification and validation of trefoil factor 1 as a marker for micrometastatic breast cancer provides additional evidence that this strategy can identify novel molecular markers. We agree with Dr. Span that selection of an ALN known to overexpress mam may have affected microarray results. However, in addition to the study described in the article, we have done dilutional microarray analyses on two pathology-positive ALN that did not overexpress mam with the goal of identifying novel molecular markers that complement mam in the detection of micrometastatic breast cancer. Although trefoil factor 1 was again one of the most overexpressed genes in one of the ALN, we also identified several other potential breast cancer markers.
In Response: I am responding to the comment made by Dr. Span in a Letter to the Editor, received on June 2, 2005. It is my understanding that Dr. Span had concerns that the patient population in my study recently published in Clinical Cancer Research (1) may have had a selection bias. He suggests that because the majority (77%) of the patients had estrogen receptor–positive tumors, the patient selection would favor the selection of mammaglobin as a sensitive marker for circulating breast cancer cells, considering the particular association of mammaglobin expression with estrogen receptor–positive tumor characteristics (2). However, our study was a prospective study that enrolled women who had an abnormality on mammography imaging with no prior history of cancer, and tumor characteristics that were not known at the time of enrollment. Thus, patients were not excluded from our study based on their tumor hormone receptor or mammaglobin status. We appreciate the interesting correlation that exists between estrogen receptor and mammaglobin expression in breast tumors observed by Span and coworkers. In general, the majority of breast cancer patients do have estrogen receptor–positive tumors and according to Span and others, these tumors may then express mammaglobin. Overall, our prospective study design precluded us from having any type of preselection bias towards tumor hormone receptor and mammaglobin status.