The recent publication by Sung Jeep Kim et al. [1] entitled "Prognostic impact and clinicopathological correlation of CD133 and ALDH1 expression in invasive breast cancer" and the Commentary by Ieni and Tuccari [2] along with the Author's reply prompted us to share several considerations on the immunodetection of CD133 (Prominin-1) and data regarding its expression in mammary epithelial cells. When assessing the predictive role of CD133 in breast cancer, it is important to consider these data. The authors of both publications fairly discuss the importance of the scoring methods used and the origin of the surgical samples in the observed differences of CD133 immunopositivity rates, and the association of CD133 with breast cancer subtypes or other predictive parameters in several recent studies (Table 1).

However, differences may also be attributed to the antibodies used (Table 1). The immunodetection of the pentaspan membrane glycoprotein CD133 has generated discrepancies on numerous occasions. For example, Hermansen et al. [3] have reported inconsistent immunohistochemical patterns on replicates of glioblastoma samples using different anti-CD133 antibodies. Moreover, beyond the fact that variability in the reactivity of polyclonal antibodies is inherent to their production, a given company may market (or may have) several rabbit polyclonal anti-CD133 with distinct specificities. Unfortunately, the description of the antibodies used in these studies is sometimes minimal or incomplete, creating ambiguity about their nature or making it difficult to track information about their precise specificity and to compare the different results [456]. In addition, although most studies are very recent, several anti-CD133 antibodies are not commercially available any longer, impeding any further investigation. In some cases, the antibody simply cannot be found with the sellers [7] or is actually specific for a different molecule, E-cadherin [8] (Table 1). Surprisingly, this latter publication by Aomatsu et al. [8] is repeatedly quoted by newer publications, including Kim et al. [1] and Ieni and Tuccari [2], in support of the prognostic role of CD133 in breast cancer [5910].

The specificity of the anti-CD133 antibody is essential for the interpretation of data since the targeted portions of the molecule are in certain cases absent from some splicing variant isoforms [11]. A notable example is the cytoplasmic Cterminal domain of CD133, which constitutes a splicing cassette where alternative and facultative exons are expressed [12]. Therefore, caveats apply to observations made with rabbit antisera against the last 18 amino acids of CD133 as long as the nature of the CD133 isoforms examined is not ascertained (Table 1). Moreover, the glycosylation status of CD133 is known to interfere with the accessibility of certain epitopes, particularly CD133/1, which is recognized by the AC133 monoclonal antibody (Table 1) [1314]. Therefore, depending on the antibody, one might or not be looking at different entities with potentially different biological and/or pathological roles [15].

When analyzing the expression of CD133 in healthy and pathological samples, two other aspects need to be considered. First, the expression of CD133 is not limited to stem and cancer stem cells. Providing adequate antibodies and immunological techniques (antigen retrieval) are applied, CD133 can be detected in various differentiated epithelia found in kidney, prostate, liver, and pancreas [151617], and in mature glial cells [18]. In glandular epithelia, CD133⁺ cells might constitute facultative stem cells acting during regeneration [19]. In good concordance with the strong expression of PROM1 mRNA in the mammary glands, CD133 was detected by immunohistochemistry at the apical side of the lactiferous ducts in normal human mammary glands using rabbit antiserum αhE2 directed against its first extracellular loop (residues 240–388 of splice variant s2) but not its AC133 (CD133/1) epitope [13]. Such linear staining along the luminal surface in benign mammary lobules or ductules has been used as a positive control for CD133 detection in a recent study by Lin et al. [10] showing (with an unspecified antibody) a differential expression of CD133 between benign and malignant papillary lesions, with strong expression in benign papillomas and most atypical intraductal papillomas and loss of expression in malignant papillary carcinoma. Moreover, apical/endoluminal membrane staining with AC133 monoclonal antibody and antigen retrieval was reported in normal mammary epithelia and carcinoma [17]. In the hypothesis that breast cancer subtypes are linked to an epithelial differentiation hierarchy, the cellular origin of CD133⁺ cancer cells and its impact on cancer progression and metastases would be particularly relevant [20]. It is also important to distinguish CD133 expression in hematopoietic progenitor cells within lymph nodes from that in epithelial cells or tumor masses to evaluate the meaning of CD133 in breast cancer [4].

Second, because CD133 is associated with plasma membrane protrusions (e.g., microvilli, primary cilium) and extracellular vesicles budding thereof [2122], differing subcellular localization of the CD133 staining is likely to reflect modified cellular activity. Nevertheless, intracellular localization of CD133 has been observed under normal conditions, with accumulation in multivesicular bodies that are released as CD133⁺ exosomes upon fusion with the plasma membrane [1623]. The presence of CD133 in various physiological body fluids or glandular lumina in cancer samples might be considered to be a biomarker [2224], and the characterization of exosomes in plasma serum might be instructive regarding cancer origin [25]. A cytoplasmic localization of CD133 might reflect deficient cell polarization related to epithelial/mesenchymal transition [26], although the latter issue remains to be demonstrated. In addition, it seems important, in the presence of cytoplasmic expression of CD133, to consider that some anti-CD133 monoclonal antibodies have shown cross-reactivity with cytokeratin 18 [27].

In conclusion, to reach a standard in scoring systems for CD133 it is highly important to ascertain the identity of the antigenic molecule being detected and to use clearly defined and characterized reagents. However, an exhaustive expression profile of this biomarker in both normal breast and cancer tissues still needs to be completed. CD133, which was proposed to promote mammary gland branching in a mouse model [28], clearly appears to be of significance both for development and for cancer progression and prognosis.