|Home | About | Journals | Submit | Contact Us | Français|
Members of the ADAM family of proteases have been associated with mammary tumorigenesis. Gene profiling of human breast tumors identified several intrinsic subtypes of breast cancer, which differ in terms of their basic biology, response to chemotherapy/radiation, preferential sites of metastasis, and overall patient survival. Whether or not the expression of individual ADAM proteases is linked to a particular subtype of breast cancer and whether the functions of these ADAMs are relevant to the cancer subtype have not been investigated. We analyzed several transcriptomic datasets and found that ADAM12L is specifically up-regulated in claudin-low tumors. These tumors are poorly differentiated, exhibit aggressive characteristics, have molecular signatures of epithelial-tomesenchymal transition (EMT), and are rich in markers of breast tumor-initiating cells (BTICs). Consistently, we find that ADAM12L, but not the alternative splice variant ADAM12S, is a part of stromal, mammosphere, and EMT gene signatures, which are all associated with BTICs. In patients with estrogen receptor-negative tumors, high expression of ADAM12L, but not ADAM12S, is predictive of resistance to neoadjuvant chemotherapy. Using MCF10DCIS.com breast cancer cells, which express the endogenous ADAM12L and efficiently form mammospheres when plated at the density of single cell per well, we show that ADAM12L plays an important role in supporting mammosphere growth. We postulate that ADAM12L may serve as a novel marker and/or a novel therapeutic target in BTICs.
Members of the ADAM (a disintegrin and metalloproteinase) family are frequently up-regulated in breast tumors . It has been postulated that they may serve as attractive new biomarkers and/or therapeutic targets in breast cancer . However, very little is currently known about patterns of ADAM expression in individual molecular subtypes of breast cancer. Based on gene expression profiling, breast cancers can be classified into several intrinsic subtypes: luminal A and B, HER2-enriched, basal-like, claudin-low, and normal-like . Each of these subtypes is characterized by a unique gene signature, basic biology, response to endocrine and chemotherapy regimens, preferential sites of metastasis, and overall patient survival [4–7]. To consider any ADAM-based approaches, it is important to establish whether the expression of individual ADAM proteases is linked to a particular subtype of breast cancer and whether the functions of these ADAMs contribute to subtype-specific phenotypes.
ADAM12 is one of the most prominent breast cancer-associated ADAM genes. ADAM12 mRNA is alternatively spliced in human . ADAM12L (transcript variant 1, exons 1–18, 20–24) encodes the long, transmembrane protein isoform. ADAM12S (transcript variant 2, exons 1–19) gives rise to the short, secreted protein isoform. While ADAM12L levels in breast tumors have been consistently found higher than those in normal mammary tissue [9–11], the status of ADAM12S is less clear. ADAM12S was reported to be up-regulated in breast cancer based on qRT-PCR , but these results are at odds with DNA microarray profiling of ductal and lobular carcinomas versus normal ductal and lobular cells . Several other studies observed increased expression of ADAM12 in breast tumors at the mRNA or protein level, but the reagents (PCR primers or antibodies) did not distinguish between ADAM12L and ADAM12S splice variants/protein isoforms [13–15].
Due to the differences in membrane association, the biological functions of ADAM12L and ADAM12S may be different as well. Analysis of the expression levels of these two variants in individual breast cancer subtypes may provide important clues about the role of ADAM12 in mammary oncogenesis. In this work, we analyzed a number of transcriptomic datasets and found that ADAM12L is specifically up-regulated in claudin-low tumors. These tumors are poorly differentiated, exhibit aggressive characteristics, have molecular signatures of EMT, and are enriched in features of mammary stem cells  and BTICs [17–21]. Consistently, ADAM12L, but not ADAM12S, is a part of the stromal, mammosphere, and EMT gene signatures, which are all associated with BTICs. In patients with estrogen receptor (ER)-negative tumors, high expression of ADAM12L is predictive of resistance to neoadjuvant chemotherapy. Importantly, we found that ADAM12L supports tumorsphere growth of MCF10DCIS.com breast cancer cells, suggesting that ADAM12L function may actively contribute to the BTICs phenotype.
Gene expression datasets were retrieved from Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/). GSE series numbers and patient characteristic are provided in Table 1. The list of gene probe sets is provided in Online Resource Table S1.
ADAM expression levels in breast cancer subtypes were analyzed using the one-way analysis of variance (ANOVA) with Bonferroni post-test analysis for multiple pairwise comparisons. ANOVA, unpaired t test, and paired t test were performed using GraphPad Prism, version 5.0 (GraphPad Software, Inc.). To identify associations between variables and chemotherapy response, univariate and multivariate logistic regression analyses were performed using the R software . ADAM12L and ADAM12S gene expression levels were treated as continuous variables; all other explanatory variables were binary. Only the complete cases were used in the multivariate models.
HMLE and HMLE-Twist-ER cells (a gift from Dr. Robert A. Weinberg, Whitehead Institute for Biomedical Research, Cambridge, MA) were maintained as described . To induce EMT, HMLE cells were treated with 2.5 ng/ml of TGFb1 (R&D Systems), and HMLE-Twist-ER cells were incubated with 20 nM of 4-hydroxy-tamoxifen (Sigma) for 12–20 days. Control cells were treated with vehicle only. MCF10DCIS.com cells (Asterand) were grown in DMEM/F12 (1:1) with 5 % horse serum and 29 mM NaHCO3. For ADAM12 knockdown, MCF10DCIS.com cells were incubated with MISSION™ Lentiviral shADAM12 Transduction Particles (Sigma, clone ID TRCN0000047037) or with MISSION™ Non-Target shRNA Control Transduction Particles (Sigma, SHC002V), according to the manufacturer’s instructions. shADAM12 clone TRCN0000047037 was selected from five different shADAM12 clones, because preliminary analysis established that it produced the most efficient ADAM12 knockdown. After 1 day, media containing lentiviral particles were replaced with fresh media, and after additional 24 h, stably transduced cells were selected with 3 μg/ml of puromycin for 7 days. Stable overexpression of the wild-type ADAM12L and the G479E mutant in MCF10DCIS.com cells was performed by retroviral transduction, as described .
qRT-PCR was performed as described . The following primers were used: hADAM12L (F), 5′-AGCCACACCAGGATAGAGAC-3′; hADAM12L (R), 5′-CGCCTTGAGTGACACTACAG-3′; hβ-ACTIN (F), 5′-TTGCCGACAGGATGCAGAA-3′; and hβ-ACTIN (R), 5′-GCCGATCCACACGGAGTACT-3′.
Adherent cells were detached using 0.25 % trypsin and 1 mM EDTA, counted, and suspended in complete mammary epithelial basal medium (Lonza), supplemented with 20 ng/ml EGF, 20 ng/ml bFGF, 1× B27, 4 μg/mL heparin, and 1 % penicillin/streptomycin. After serial dilution to a concentration of 10 cells/ml, 100 μl of cell suspension was plated in 96-well ultralow attachment plate (Corning). One day after plating, each well was examined using a light microscope to determine the number of cells present; wells containing more than one cell were excluded from further analysis. Cells were allowed to grow undisturbed for 10 days and then each well was examined again to record the number of spheres formed.
Immunoblotting was performed using anti-ADAM12 (Ab #3394), as described .
We examined the expression levels of all catalytically active, not testis-specific, ADAM metalloproteases  in different molecular subtypes of breast tumors from the human tumor database at the University of North Carolina (the UNC337 dataset, Ref. ). Pairwise comparison of claudin-low tumors versus basal-like, HER2-enriched, luminal A, luminal B, and normal-like tumors showed that ADAM12L was the only ADAM that was significantly elevated in claudin-low tumors (Fig. 1). ADAM12S was not represented on Agilent platforms used for gene profiling in that study.
To directly compare ADAM12L and ADAM12S levels in different subtypes of breast cancer, we used three other datasets, EMC286, EMC192, and MSK82 [27–29], profiled on Affymetrix platforms. Tumor classification into basal-like, claudin-low, HER2-enriched, combined luminal, and normal-like subtypes was adopted after the study of Harrell et al. . While ADAM12L was significantly up-regulated in claudin-low tumors, no differences in the expression level of ADAM12S were observed between different tumor subtypes (Fig. 2).
Claudin-low signatures are enriched in residual tumors remaining after endocrine therapy for ER-positive disease . We analyzed ADAM12L and ADAM12S levels before and after letrozole treatment in two different datasets [20,30]. In both datasets, ADAM12L, but not ADAM12S, was significantly increased in tumors after letrozole therapy (Fig. 3).
Claudin-low subtype is characterized by high expression of stromal/mesenchymal genes, a high abundance of CD44 and a low amount of CD24, and is associated with the mammosphere gene signature [17, 18]. Therefore, we analyzed ADAM12L and ADAM12S levels in the relevant, publicly available gene expression studies. In a study where breast tumor epithelium and stroma were isolated using laser capture microdissection , ADAM12L expression was significantly higher in stromal than in epithelial cells (Fig. 4a). Strikingly, ADAM12S showed an opposite pattern of expression, and it was significantly enriched in epithelial cells (Fig. 4a). When primary breast tumor cells were sorted by flow cytometry into CD44+/CD24−/low cells and “other” populations , ADAM12L was enriched in the CD44+/CD24−/low population. Although the difference in ADAM12L expression did not reach statistical significance, it was clearly more pronounced than the difference in ADAM12S expression (Fig. 4b). Finally, ADAM12L, but not ADAM12S, was significantly higher in cancer mammospheres grown in suspension than in primary bulk tumors  (Fig. 4c). These observations are consistent with the previously reported up-regulation of ADAM12L in mammospheres formed by normal mammary epithelial cells .
Analysis of the results of several gene expression studies indicates that ADAM12L is significantly up-regulated during EMT. First, in human mammary epithelial (HMLE) cells induced to undergo EMT by overexpressing TGF-β1, transcription factors Twist, Gsc, or Snail , or by knocking down expression of E-cadherin , ADAM12L was significantly up-regulated during EMT (Fig. 4d). Second, ADAM12L expression was higher in spontaneously arising mesenchymal subpopulations of cells (MSPs) than in CD24hi epithelial subpopulation of HMLE cells  (Fig. 4e). Third, in Ras-transformed HMLE cells treated in culture with either paclitaxel, a commonly used breast cancer chemotherapeutic drug, or salinomycin, a compound that selectively targets cells after EMT , ADAM12L was significantly lower in salinomycin-treated cells (Fig. 4f). In contrast, ADAM12S was not associated with EMT in any of the datasets analyzed in Fig. 4d–f.
EMT and mammosphere gene signatures are important characteristics of BTICs [37–40]. BTICs are largely resistant to chemotherapy [41–43]. Consistently, a stromal metagene, a group of highly co-expressed 50 genes with a pattern of expression similar to that of EMT and mammospheres, was reported to predict resistance to neoadjuvant chemotherapy in ER-negative breast tumors . While ADAM12 was listed as one of the 50 genes of the stromal metagene, the relative contribution of ADAM12L and ADAM12S variants was not addressed. Here, we asked whether there is a difference in the predictive value of ADAM12L and ADAM12S to chemotherapy and whether ADAM12L is an independent predictor of chemoresistance.
Among patients with ER-negative tumors undergoing preoperative anthracycline-based chemotherapy (the EORTC10994 dataset, [Ref. 44]), ADAM12L was significantly lower in cases with pathologic complete response (pCR, n = 28) than in patients with near pCR (n = 37) (Fig. 5a). In contrast, there was no difference in the level of ADAM12S expression between these two groups. In another study of the response to anthracycline/taxane-based neoadjuvant therapy (the MAQCII/MDACC dataset, [refs. 45, 46]), ADAM12L was significantly lower in ER-negative patients with pCR (n = 40) compared to patients with residual disease (n = 50) (Fig. 5b). ADAM12S again did not significantly differ between the two groups.
To determine if ADAM12L or ADAM12S provides predictive information for resistance of ER-negative tumors to chemotherapy in the two datasets, univariate and multivariate logistic regression analyses were performed with the nodal status and T stage (for the EORTC10994 dataset), and the nodal status, T stage, patient age, and tumor grade (for the MAQCII/MDACC dataset) as categorical variables, and with ADAM12L and ADAM12S gene expression levels as continuous variables. The information on patient age and tumor grade was not available for the EORTC10994 dataset.
ADAM12L was the only significant variable in univariate analysis in both datasets (Table 2). ADAM12L was also a predictive factor of pCR in the multivariate analysis of the EORTC10994 dataset [odds ratio (OR) 0.37; 95 % confidence interval (CI) 0.16–0.77; P = 0.0129], with the nodal status, T stage, and ADAM12S included in the model. In multivariate analysis of the MAQCII/MDACC dataset, ADAM12L was borderline significantly associated with pCR (OR 0.76, CI 0.54–1.04, P = 0.0936). Similar borderline significance of ADAM12L was achieved when age and grade were omitted from the model to match the number of variables with the EORTC10994 dataset (results not shown). ADAM12S was not predictive of pCR in any model (Table 2).
To begin to characterize possible roles of ADAM12L protein in BTICs, we investigated the relationship between ADAM12L, EMT, and sphere growth. First, we validated the results of gene expression profiling by qRT-PCR. HMLE cells were induced to undergo EMT with TGFβ, and EMT was confirmed by the loss of expression of E-cadherin. ADAM12L level increased by ~fivefold and ~tenfold after 12 or 20 days of treatment with TGFβ (Fig. 6a). Similarly, when HMLE cells with tamoxifen-inducible expression of TWIST1, HMLE-TWIST-ER were induced to undergo EMT by 12- or 20-day treatment with 4-hydroxy-tamoxifen, ADAM12L was up-regulated by ~twofold and ~threefold, respectively (Fig. 6b). For mammosphere experiments, we chose MCF10DCIS.com cells, an ER-negative, CD44-positive breast cancer cell line that forms spheres with high efficiencies when plated in ultralow attachment plates in serum-free media supplemented with growth factors . Comparing to the monolayer conditions, the level of ADAM12L was ~fivefold higher in MCF10DCIS.com cells grown as spheres (Fig. 6c). Furthermore, among all ADAM mRNA encoding catalytically active, transmembrane, non-testis-specific ADAMs (i.e., ADAM8, 9, 10, 12, 15, 17, 19, 28, and 33), the only other ADAM that was up-regulated in spheres was ADAM28 (result not shown). We calculated the mRNA copy numbers for ADAM12L and ADAM28 in MCF10DCIS.com spheres and found that the abundance of ADAM12L was at least two orders of magnitude higher than that of ADAM28 (result not shown), suggesting that ADAM12 is the main ADAM up-regulated in cells grown as spheres.
An important question is whether ADAM12L plays an active role in the process of EMT or sphere growth. By up-regulating or down-regulating ADAM12L in mammary MCF10A or HMLE cells, we did not observe a significant effect on the time course of EMT induced by limiting doses of TGFβ (results not shown). In contrast, ADAM12L appeared to support sphere formation in MCF10DCIS.com cells. To ensure that spheres contain the progeny of single cells truly capable of proliferation in suspension (Fig. 7a), and not simply clusters of originally plated cells, MCF10DCIS.com cells were seeded at the density of 1 cell/well. Down-regulation of ADAM12 by shRNA significantly decreased the sphere formation efficiency (Fig. 7b). In addition, cells with down-regulated ADAM12L level were efficiently excluded from spheres (Fig. 7c), further suggesting that ADAM12L protein may have an important function during sphere formation.
The shRNA construct used in the experiments shown in Fig. 7 was more potent in down-regulating ADAM12L than several other shRNA tested, but it targeted both the ADAM12L and ADAM12S variants. As another approach to specifically limit the function of the ADAM12L variant in MCF10DCIS.com cells, we overexpressed the dominant-negative mutant of ADAM12L, G479E . While overexpression of the wild-type (WT) ADAM12L did not significantly increase the efficiency of sphere formation, the number of spheres formed by cells overexpressing the G479E mutant was significantly lower than the number of spheres formed by cells with overexpression of the WT ADAM12L (Fig. 8a). The level of ADAM12L in spheres was further compared to the level of ADAM12L in cells grown as monolayers by Western blotting (Fig. 8b). We observed a weak signal representing the endogenous ADAM12L and strong signals for the WT and G479E proteins. The WT protein existed as the full-length, nascent form of ~120 kDa and the processed, active form of ~90 kDa. As expected, the G479E mutant was not processed and lacked the active form . Importantly, the ratio of the amount of G479E mutant protein in spheres versus monolayers was significantly lower than the ratio of the WT ADAM12L protein (nascent and processed forms combined) (Fig. 8c). This result indicated that the G479E mutant might have been selectively excluded from the spheres, further suggesting an important role of the endogenous, active ADAM12L in sphere formation.
Our analysis of a number of transcriptomic datasets for primary breast tumors demonstrates that ADAM12L and ADAM12S transcripts show distinct patterns of expression in breast cancer. This finding is remarkable, because the two corresponding protein isoforms, ADAM12L and ADAM12S, have different biochemical properties and unique intracellular localizations. Since ADAM12L, and not ADAM12S, is up-regulated in claudin-low tumors that are enriched for BTIC features , and since it is a part of the EMT and mammosphere gene signatures, both of which are also linked to BTICs [20, 21, 42], we postulate that any potential function of ADAM12 in BTICs requires membrane anchoring present in ADAM12L, and not in ADAM12S. While the mechanisms responsible for the differential regulation of ADAM12L and ADAM12S in breast cancer are not known, the most likely scenario involves post-transcriptional up-regulation of ADAM12L via specific elements in its unique 3′-UTR, a hypothesis that is currently being tested in our laboratory.
Our analysis also reveals that among different molecular subtypes of breast cancer, ADAM12L is specifically up-regulated in the claudin-low subtype. Interestingly, recent profiling of histologically benign tissue adjacent to breast tumors identified two distinct subtypes of tumor microenvironment, named “Active” and “Inactive” . The “Active” cancer-adjacent tissue showed features of EMT and expressed claudin-low-related genes, including ADAM12L. These claudin-low gene expression features were more common in the extratumoral than intratumoral locations and were independent of whether the tumor itself was claudin-low or not. Furthermore, claudin-low features in normal breast tissues from non-diseased women have been shown to be associated with high risk of developing breast cancer . ADAM12L was also significantly increased in this healthy claudin-low group, reinforcing the inherent link between ADAM12L and the claudin-low phenotype.
ADAM12 was reported earlier to be a part of stromal gene signatures [44, 51, 52], and here we show that this conclusion applies selectively to the ADAM12L variant. While it is accepted that the stromal gene cluster represents reactive tumor stroma, the cellular origin of the stromal gene signature is less clear. Since the pattern of gene expression within the stromal cluster is similar to the EMT and mammosphere signatures, it has been argued that EMT of tumor epithelial cells could account for increased stromal gene expression . The EMT program directly contributes to the resistance of chemotherapeutic agents [36, 53–56]. Consistently, we found ADAM12L to be an independent predictor of chemoresistance to neoadjuvant anthracycline-based therapy in ER-negative breast tumors from the EORTC10994 dataset. In the MAQCII/MDACC dataset, the association between the level of ADAM12L expression in ER-negative tumors and resistance to anthracycline/taxane-based neoadjuvant therapy was borderline significant. Since the pathological response of claudin-low tumors to anthracyclines and taxanes are similar , we believe that a somewhat lesser power of ADAM12L to predict response to chemotherapy in the MAQCII/MDACC datasets may be related to a different method of tissue harvesting [fine needle aspiration (FNA) versus core biopsy in the EORTC10994 dataset]. As FNA samples typically contain little stromal component [58, 59], contribution of stromal/mesenchymal genes to gene signatures in the MAQCII/MDACC dataset must have been naturally underestimated.
The defining characteristics of human breast cancer subtypes, including the claudin-low subtype, are conserved among human cancer cell lines and among mouse models of the disease [17, 18]. Because ADAM12L is up-regulated in human claudin-low tumors, we believe that claudin-low cell lines and mouse models are best suited to study the role of ADAM12L in mammary tumorigenesis. Of note, the role of ADAM12 in breast cancer was previously addressed using luminal MCF7 and T47D cell lines [12, 60] or TgMMTV-PyMT mice [13, 61], which are models of the luminal tumors [17, 18]. Application of claudin-low cell lines or mouse models in the future may be necessary to obtain new insights into the biological role of ADAM12L in the breast cancer disease.
In conclusion, up-regulation of ADAM12L in claudinlow tumors, during EMT, and in mammospheres, as well as poor chemotherapy response of tumors with high expression of ADAM12L, all link ADAM12L to a BTIC phenotype. Whether or not ADAM12L may serve as a marker for BTICs, or at least a certain subpopulation of BTICs, is currently being investigated in our laboratory. In particular, we are performing orthotropic cell transplantation assays to directly compare the tumorigenic potential of cells with high versus low expression of the endogenous ADAM12L. Sphere formation assay, while representing a simple and frequently used tool for evaluation of stem-like properties of cells and a surrogate assay for tumorigenicity, has important limitations. As it has been demonstrated previously, not all spheres are necessarily derived from stem-like cells. In fact, for primary mouse mammary epithelial cells, the proportion of spheres containing the regenerative stem cells can be as low as 15–33 % . Therefore, sphere formation assays alone may not provide an accurate measure of the frequency of BTICs, and limiting dilution transplantation assays are needed to directly evaluate the tumorigenic potential of ADAM12L-expressing cells.
Finally, the fact that down-regulation of ADAM12L compromised sphere formation of MCF10DCIS.com cells suggests that ADAM12L protein may play an active role in sphere growth and/or may increase survival of cells deprived of attachment to the matrix. Molecular mechanisms by which ADAM12L would enhance sphere growth are currently not known, but they may involve modulation of critical signaling pathways such as Notch, EGFR, or TGFβ [24, 63, 64]. If an active role of ADAM12L in maintaining the BTICs phenotype is confirmed, ADAM12L may be considered as a candidate for a therapeutic target in these cells.
This work was supported by NIH grants R15CA151065 and R01CA172222, and Innovative Research Award from Terry C. Johnson Center for Basic Cancer Research at KSU. This is contribution 13-289-J from Kansas Agricultural Experiment Station.
Conflict of interest The authors declare that they have no conflict of interest.
Electronic supplementary material The online version of this article (doi:10.1007/s10549-013-2602-2) contains supplementary material, which is available to authorized users.
Hui Li, Department of Biochemistry and Molecular Biophysics, Kansas State University, 141 Chalmers Hall, Manhattan, KS 66506, USA.
Sara Duhachek-Muggy, Department of Biochemistry and Molecular Biophysics, Kansas State University, 141 Chalmers Hall, Manhattan, KS 66506, USA.
Suzanne Dubnicka, Department of Statistics, Kansas State University, Manhattan, KS 66506, USA.
Anna Zolkiewska, Department of Biochemistry and Molecular Biophysics, Kansas State University, 141 Chalmers Hall, Manhattan, KS 66506, USA.