Phenotypic Diversity of Breast Cancer-Related Mutations in Metalloproteinase-Disintegrin ADAM12

Six different somatic missense mutations in the human ADAM12 gene have been identified so far in breast cancer. Five of these mutations involve highly conserved residues in the extracellular domain of the transmembrane ADAM12-L protein. Two of these extracellular mutations, D301H and G479E, have been previously characterized in the context of mouse ADAM12. Three other mutations, T596A, R612Q, and G668A, have been reported more recently, and their effects on ADAM12-L protein structure/function are not known. Here, we show that ADAM12-L bearing the G668A mutation is largely retained in the endoplasmic reticulum in its nascent, full-length form, with an intact N-terminal pro-domain. The T596A and R612Q mutants are efficiently trafficked to the cell surface and proteolytically processed to remove their pro-domains. However, the T596A mutant shows decreased catalytic activity at the cell surface, while the R612Q mutant is fully active and comparable to the wild-type ADAM12-L. The D301H and G479E mutants, consistent with the corresponding D299H and G477E mutants of mouse ADAM12 described earlier, are not proteolytically processed and do not exhibit catalytic activity at the cell surface. Among all six breast cancer-associated mutations in ADAM12-L, mutations that preserve the activity - R612Q and L792F - occur in triple-negative breast cancers, while loss-of-function mutations - D301H, G479E, T596A, and G668A - are found in non-triple negative cancers. This apparent association between the catalytic activity of the mutants and the type of breast cancer supports a previously postulated role of an active ADAM12-L in the triple negative breast cancer disease.


Introduction
Disintegrin and metalloproteinase domain-containing protein ADAM12 is a member of the ADAM family of proteins that mediate cleavage of substrates at the cell surface and/or modulate intracellular signaling pathways [1,2]. ADAM12 is highly upregulated in human breast tumors [3][4][5][6][7][8][9]. In triple negative breast cancers (TNBCs, lacking estrogen receptor and progesterone receptor expression and ERBB2 gene amplification), high expression of ADAM12-L, but not ADAM12-S, mRNA is associated with poor prognosis [10]. ADAM12-L and ADAM12-S are two different splice variants that encode the long, transmembrane protein isoform ADAM12-L and the short, secreted ADAM12-S, respectively [11].
Among thirteen different ADAM genes that encode catalytically active proteases [1], ADAM12 is the most frequently somatically mutated gene in human breast cancers. As of September 2013, the COSMIC database (Catalogue of Somatic Mutations in Cancer, http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/) listed 6 confirmed somatic missense mutations in the ADAM12 gene per a total of 1104 unique breast carcinoma samples analyzed. The frequencies of breast cancer-associated missense mutations in other genes encoding catalytically active ADAMs were: 1/973 in ADAM9, 1/973 in ADAM10, 1/1147 in ADAM17, 1/1010 in ADAM19, 1/973 (plus one nonsense mutation) in ADAM20, and 1/973 in ADAM30. No other missense/nonsense somatic mutations were reported for the remaining ADAM genes encoding catalytically active proteases (i.e., ADAM8, 215, 221, 228, 233, and ADAMDEC1). A relatively high frequency of mutations in the ADAM12 gene can be attributed to the fact that ADAM12 is located on human chromosome 10q26.2, in a region capable of forming highly stable secondary structures [12].
The six breast cancer-associated mutations in the ADAM12-L protein include the D301H mutation in the metalloproteinase domain, G479E in the disintegrin domain, T596A and R612Q in the cysteine-rich domain, G668A in the epidermal growth factor (EGF)-like domain, and L792F in the cytoplasmic tail [13][14][15] ( Figure 1A). We have previously shown that the D299H and G477E mutations in mouse ADAM12 (which correspond to the D301H and G479E mutations in human ADAM12) are loss-offunction mutations that inhibit the intracellular trafficking and proteolytic activation of the nascent ADAM12 protein [16]. The L792F mutation in human ADAM12-L was reported not to affect protein processing, localization, or function [16,17]. The other three mutations -T596A, R612Q, and G668A -have been identified more recently [14,15], and their effects on the structure/ function of ADAM12 are currently unknown.
Here, we show that there is functional diversity between the three recently identified mutations. While ADAM12-L containing the G668A mutation is largely retained in the endoplasmic reticulum (ER) and is not proteolytically activated in the Golgi, the T596A mutant is properly trafficked and proteolytically processed but is still catalytically inactive. The R612Q mutant is trafficked, processed, active, and is indistinguishable from the wild-type (WT) ADAM12-L. Taking into consideration all six known breast cancer-associated somatic mutations in ADAM12-L, we note an apparent association between the catalytic activity of ADAM12-L mutants and the molecular characteristic of the tumor. The two mutations that do not have any effect on ADAM12-L activity, R612Q and L792F, occurred in TNBCs. The four mutations that render ADAM12-L inactive, i.e., D301H, G479E, G668A, and T596A, were described in non-TNBCs. This observation further suggests an important role of the catalytically active ADAM12-L in the triple negative breast cancer disease.

Expression Constructs
Retroviral expression vector ADAM12-L-pBABEpuro was used for the expression of the wild-type (WT) ADAM12-L protein. The D301H, G479E, T596A, R612Q and G668A point mutations were introduced by site-directed mutagenesis using the Quick-Change kit (Stratagene). The entire lengths of all DNA inserts were sequenced to confirm that no other mutations were introduced during mutagenesis. The expression construct of Delta-like 1 (DLL1) in pIRESpuro vector was described earlier [18].

Cell Culture and Treatment
Human MCF10A mammary epithelial cells (ATCC) were cultured in DMEM/F12 (1:1) supplemented with 5% horse serum, 0.5 mg/ml hydrocortisone, 20 ng/ml human EGF, 10 mg/ml insulin, 100 ng/ml cholera toxin, and 1% penicillin/streptomycin. Retrovirus production and stable transduction of MCF10A cells with viruses encoding WT or mutant ADAM12-L, or with control viruses bearing empty pBABEpuro vector, were performed as described previously [19]. Transient transfections were performed using X-tremeGENE HP transfection reagent (Roche). For protein stability assay, cells were treated with 10 mg/ml cycloheximide (EMD Millipore) in culture medium for indicated times.

Flow Cytometry
Cells were trypsinized into a single cell suspension, washed with DPBS containing 3% BSA, and incubated with anti-ADAM12 monoclonal antibody or isotype control antibody (R&D Systems, both at 1:10 dilution) for 30 minutes on ice. Cells were then washed 3 times, incubated with allophycocyanin (APC)-conjugated anti-mouse antibodies (Jackson ImmunoResearch; 1:100) for 30 min on ice, washed again, and then incubated with 1 mg/ml propidium iodide (PI; BD Biosciences) for viability. Analysis was performed using a BD FACSCalibur flow cytometer. Only the cells negative for PI staining (viable cells) were selected for the ADAM12 analysis.

Endo H Treatment
Stably transduced cells were treated with extraction buffer and centrifuged at 21,0006g for 15 min at 4uC. Supernatants were treated with EndoH f denaturing buffer (New England BioLabs), boiled, and then treated with Endo H f (3,000 U), according to the manufacturer's instructions.

Determination of Cell Doubling Times
Stably transduced cells were seeded in 6-well plates at the density of 60,000 cells/well. After 24, 48, and 72 h, cells were detached and counted with Cellometer AutoT4 (Nexcelom Bioscience), in duplicates. Exponential growth curves were fitted to each dataset and the doubling times were calculated using a nonlinear regression function in the GraphPad Prism 5.0 software.

Evaluation of ADAM12-L-mediated Shedding of EGFR Ligands
MCF10A cells stably overexpressing WT or mutant ADAM12-L proteins, or cells stably transduced with empty pBABEpuro vector, were incubated for 16 h in serum-free media. Conditioned media were then collected, pre-cleared by centrifugation, and added to the duplicate wells containing ''reporter'' empty vectortransduced MCF10A cells that were pre-incubated for 16 h in serum-free medium. After 30 min, cells were washed with DPBS, treated with extraction buffer containing phosphatase inhibitors  [20,21]). The metalloproteinase, disintegrin, cysteine-rich, and EGF-like domains are shown in purple, blue, yellow, and green, respectively. Positions of the five amino acids mutated in breast cancers (red spheres) and the side chain of the catalytic residue E351 (cyan sticks) are indicated. (C) Sequence alignment of the cysteine-rich and EGF-like domains of ADAM12 from different species. NCBI RefSeq numbers are: Homo_sapiens, NP_003465; Mus_musculus, NP_031426; 50 mM NaF, 2 mM Na 3 VO 4 , and 10 mM Na 4 P 2 O 7 , and analyzed by SDS-PAGE and Western blotting using antiphospho-EGFR (pY1173) and anti-EGFR antibodies.

Cell Migration Assay
MCF10A cells with stable overexpression of WT or mutant ADAM12-L proteins, or control empty vector-transduced cells, were suspended in MCF10A medium containing 0.1% BSA instead of horse serum, and seeded in the upper chambers of Transwell inserts with a 8-mm pore size polyethylene terephthalate membrane (BD Biosciences), at 2.5610 4 cells/chamber. The lower chambers contained the full culture medium supplemented with 5% horse serum and 20 ng/ml EGF. After incubation at 37uC for 18 h, cells were fixed with 3% glutaraldehyde in DPBS for 20 min, washed twice with DPBS, and stained with 0.5% crystal violet in 20% methanol for 10 min. Cells at the upper face of the membrane were removed with cotton swabs, and cells at the lower face were examined with an inverted microscope using a 106 magnification, and photographed. Numbers of cells in five random fields were counted, and the mean number of migrated cells for each insert was determined.

ADAM12-L Structure Prediction
The I-TASSER software (http://zhanglab.ccmb.med.umich. edu/I-TASSER/, refs. [20,21]) was used to predict the structure of the extracellular domain of the mature form of human ADAM12-L (amino acids 208-708). The ADAM22 template (PDB:3G5C) was excluded from the I-TASSER template library due to a compact packing of the metalloproteinase domain against the cysteine-rich domain and steric hindrance of the pseudo-catalytic site [22], a feature that may be characteristic for catalytically inactive ADAMs.

Statistical Analysis
Fisher's exact test, unpaired t test, and linear regression were performed using GraphPad Prism 5.0.

Results
The five breast cancer-associated mutations mapping to the extracellular portion of ADAM12-L are located in different domains and spread over a region spanning more than 300 amino acids ( Figure 1A). We asked whether these mutations might be clustered in a particular region in the three-dimensional structure of the protein. Since the X-ray structure of human ADAM12-L is not available, we used the I-TASSER protein structure prediction tool [20,21] to build a 3D model of the extracellular domain of the active form of human ADAM12-L comprising the metalloproteinase, disintegrin, cysteine-rich, and EGF-like domains. From the model, it is apparent that the cancerassociated mutations are scattered over the entire structure and they do not cluster in a particular region of the protein ( Figure 1B).
The three recently identified mutations -T596A, G668A, and R612Q -involve amino acid residues in ADAM12 that are highly conserved between species ( Figure 1C). The G668 residue is also conserved between human ADAM12 and the most closely related human ADAMs, namely ADAM19, ADAM33, and ADAM15 [1]. In contrast, the T596 and R612 residues are not conserved between ADAM12-L and the other three ADAMs ( Figure 1D). To study the effects of the T596A, G668A, and R612Q mutations on the functionality of ADAM12-L, we stably expressed WT and mutated ADAM12-L proteins in human MCF10A mammary epithelial cells. Western blotting of total cell lysates with an antibody specific for the cytoplasmic tail of ADAM12-L demonstrated that the proteolytic processing of WT ADAM12-L and the T596A and R612Q mutants was virtually indistinguishable (Figure 2A). In all three cases, the mature, processed form of ,90-kDa was easily detected, and its abundance was similar to that of the nascent, full-length form of ,120-kDa. In contrast, processing of the G668A mutant was significantly inhibited. Staining of live cells with an antibody recognizing the extracellular domain of ADAM12-L and analysis by flow cytometry demonstrated that WT ADAM12-L and the T596A and R612Q mutants were readily detected at the cell surface, whereas the G668A mutant showed much weaker cell surface staining ( Figure 2B). To further explore the intracellular localization of WT ADAM12-L and the mutants, we performed immunofluorescence staining of permeabilized cells with anti-ADAM12 and anti-KDEL antibody, a marker of the ER. We observed that substantial amounts of WT ADAM12-L and the T596A and R612Q mutants were present in post-ER compartments located at the cell periphery ( Figure 2C), consistent with cell surface localization of these proteins detected by flow cytometry. In contrast, anti-ADAM12 staining in G668Aexpressing cells largely coincided with anti-KDEL staining ( Figure 2C), suggesting that this mutant was retained, at least partially, in the ER and explaining why it was poorly detected at the cell surface ( Figure 2B).
The pronounced effect of the G668A mutation on the processing and localization of ADAM12-L is unexpected, given the very conservative nature of the Gly-to-Ala substitution. Thus, the maturation of the G668A mutant in the secretory pathway was further probed with endoglycosidase H (Endo H) and compared to the WT ADAM12-L. When the lysate of WT-expressing cells was treated with Endo H, the mobility of the nascent ADAM12-L increased, while the mature form was more resistant to the Endo H treatment ( Figure 3A). This result indicated that the mature form, but not the nascent form, progressed through the Golgi compartment, where the resistance of N-linked oligosaccharides to Endo H is acquired. The G668A mutant, represented predominantly by the full-length form, was sensitive to Endo H, further suggesting that this mutant did not efficiently progress beyond the ER. Finally, we performed cell surface biotinylation of intact cells expressing the WT or G668A mutant protein. In WT ADAM12-L-expressing cells, the mature form was efficiently biotinylated, whereas the nascent form was resistant to the modification ( Figure 3B). This result indicated that at the cell surface of WT ADAM12-L-expressing cells, the mature ADAM12-L form was much more abundant than the nascent form. In G668A mutantexpressing cells, the extent of cell surface biotinylation of ADAM12-L was considerably lower than in WT ADAM12expressing cells, further indicating that the G668A mutation impaired trafficking of the ADAM12-L protein to the cell surface.
The next question might be: Why is the G668A mutant inefficiently trafficked to the cell surface and predominantly retained in the ER? We reasoned that this mutant might be misfolded and retained by the protein quality control system Rattus_norvegicus, XP_001054670; Bos_taurus, NP_001001156, Equus_caballus, XP_001490097; Gallus_gallus, NP_001136322, and Xenopus_tropicalis, NP_001035103. (D) Sequence alignment of the cysteine-rich and EGF-like domains of human ADAM12 and the most closely related human ADAMs. NCBI RefSeq numbers are: ADAM19, XP_005266060, ADAM33, NP_079496.1, and ADAM15, NP_997080. In C and D, asterisks indicate three novel mutations in human ADAM12 found in breast tumors [14,15]. Clustal X color scheme was applied. doi:10.1371/journal.pone.0092536.g001 ADAM12 Mutations in Breast Cancer PLOS ONE | www.plosone.org operating in the ER. However, using co-immunoprecipitation, we did not detect interaction between the G668A mutant (or WT ADAM12-L) with ER chaperones BiP, Grp94, and calnexin, or with ER stress proteins that assist in proper disulfide formation ERp44, ERp57, ERp72, Ero1, PDI (results not shown). Also, cellular levels of these chaperones/stress proteins were not  Figure 3C). Collectively, these results suggested that although the G668A mutant was retained in the ER, most likely it was not misfolded and it was not subject to rapid degradation.
Next, we asked about the catalytic activity of ADAM12-L mutants. While a mutation causing impaired intracellular trafficking of ADAM12-L is naturally expected to cause a decrease in the enzyme activity at the cell surface (unless it exerts an indirect effect, such as enhancing cell surface expression of a different proteolytic enzyme), a mutation that does not affect intracellular trafficking and processing can still have an impact on the catalysis or substrate recognition. For comparison, we also included the D301H and G479E mutants in the current analysis. The corresponding D299H and G477E mutations in mouse ADAM12 were shown previously to block the intracellular trafficking and processing of the protein, as well as ADAM12-mediated cleavage of the substrate protein DLL1 [16]. Trafficking, processing, and catalytic activities of human D301H or G479E mutants have not been examined before. The L792F mutation in the cytoplasmic tail of human ADAM12-L was reported to have no effect on ADAM12-L trafficking, processing, or catalytic activity [16,17].
We used three different approaches to evaluate the catalytic activity of the D301H, G479E, T596A, R612Q, and G668A mutants at the surface of MCF10A cells ( Figure 4A). In the first approach, cells with stable overexpression of WT or mutant ADAM12-L were transfected with a plasmid encoding DLL1. The full-length (FL) DLL1 and the C-terminal fragment (CTF) generated by the proteolytic cleavage were detected by Western blotting, as described previously [16,18]. We observed that the CTF/FL ratio of DLL1 was higher in cells expressing WT ADAM12-L or the R612Q mutant than in control cells, indicative of the catalytic activity of these ADAM12 proteins toward the DLL1 substrate ( Figure 4B). In contrast, the CTF/FL ratio was not increased in D301H, G479E, G668A, or T596A mutantexpressing cells compared to control cells, suggesting that these four mutations significantly reduced the ability of ADAM12-L to cleave DLL1.
In the second approach, we focused on ADAM12-L-mediated shedding of EGFR ligands, as this function of ADAM12-L has been recently shown to be important in the biology of TNBC [10]. Among different EGFR ligands, ADAM12-L was previously shown to cleave EGF [23,24] and heparin-binding(HB)-EGF [25,26]. Here, the amount of endogenous EGFR ligands shed to the media by MCF10A cells stably overexpressing WT or mutant ADAM12-L proteins was evaluated. Cells were incubated for 16 h in serum-free media, the conditioned media were then transferred to starved ''reporter'' MCF10A cells, incubated for 30 min, and then the extent of EGFR phosphorylation at Tyr1173, one of the major autophosphorylation sites in response to ligand binding [27], was examined. Conditioned media from WT ADAM12-L or R612Q mutant-expressing cells increased EGFR phosphorylation in reporter cells ( Figure 4C). In contrast, conditioned media from D301H, G479E, G668A, or T596A mutant-expressing cells did not cause elevation in EGFR phosphorylation ( Figure 4C), suggesting that these mutants most likely did not efficiently shed EGFR ligands.
In the third approach, we investigated the effect of WT and mutant ADAM12-L expression on cell migration. We showed previously that overexpression of mouse ADAM12 in NIH3T3 cells increased cell migration using scratch wound assay, and that the D299H and G477E mouse ADAM12 mutants were inactive [16]. ADAM12-L has been also found to potentiate the migration of head and neck squamous cell carcinoma cells [28]. However, overexpression of ADAM12-L in breast cancer MCF-7 cells did not affect cell migration [29], and mouse ADAM12 was reported to inhibit keratinocyte migration or integrin a4b1-mediated CHO cell migration cells [30,31]. Thus, the effect of ADAM12 on cell migration appears to be highly context-dependent and may involve distinct mechanisms. In the current study, we used a Transwell assay to assess the effect of mutations in human ADAM12-L on the migration of MCF10A cells. While overexpression of WT ADAM12-L or the mutants did not affect cell growth ( Figure 5A), the WT and the R612Q mutant ADAM12-L significantly increased cell migration (Figure 5B and C). This upregulation of cell migration required the catalytic activity of ADAM12-L, because the E351Q mutant-expressing cells migrated at a rate similar to control cells. Importantly, overexpression of the D301H, G479E, G668A, or T596A mutants did not increase cell migration, further indicating that these mutants were either not efficiently targeted to the cell surface or had activities significantly lower than the WT ADAM12-L and the R612Q mutant.
A summary of the functional characterization of breast-cancer associated ADAM12-L mutations, as well as the properties of breast tumors in which each mutation was identified, is provided in Table 1. Included in Table 1 are also SIFT scores and PolyPhen scores obtained from the Ensembl Genome Browser (www. ensembl.org). SIFT is a sequence homology-based tool that sorts intolerant from tolerant amino acid substitutions and predicts whether an amino acid substitution in a protein will have a phenotypic effect (http://sift.jcvi.org, ref. [32]). PolyPhen-2 is a tool which predicts the variation effect on protein function based on physical and comparative considerations (http://genetics.bwh. harvard.edu/pph2; ref. [33]). SIFT scores #0.05 and PolyPhen scores $0.95 designate amino acid substitutions that are predicted to be damaging to the structure/function of a protein. As summarized in Table 1, with the exception of the T596A mutation, there is an agreement between the predicted effect of each mutation and the experimentally determined activity of Conditioned media were then transferred to reporter MCF10A cells that were also pre-starved for 16 h, and incubation continued for 30 min. The level of phosphorylation of EGFR in reporter cells was evaluated by Western blotting using anti-phosho-Y1173 antibody, band intensities were quantified by densitometry, and the extent of phosphorylation of EGFR (pEGFR) normalized to the total EGFR protein was evaluated. Conditioned media from EV-transduced cells supplemented with 15 ng/ml EGF served as positive control. The experiment was repeated three times, mean values 6 SEM are shown. *, P,0.05. Notice that EGFR phosphorylation in reporter cells was increased upon adding conditioned media from WT-or R612Q mutant ADAM12-L-expressing cells. doi:10.1371/journal.pone.0092536.g004 ADAM12-L. Importantly, mutations that do not have impact on the ADAM12-L activity -R612Q and L792F -were found in TNBCs. Mutations that inhibit the ADAM12-L activity -D301H, G479E, G668A, and T596A -occurred in non-TNBCs. Thus, there is an apparent association between the intact activity of ADAM12-L and the triple-negative status of tumors, although this association is only borderline significant due to the small sample size (P = 0.0667, Fisher's exact test).  Predicted effect of each mutation on ADAM12-L function according to the SIFT algorithm [32]. SIFT scores #0.05 indicate amino acid substitutions that are predicted to be damaging (shown in bold). b Predicted effect of each mutation on ADAM12-L function according to the PolyPhen-2 tool [33]. Scores represent prediction confidence that a given mutation changes the protein function; scores $0.95 are indicated in bold. doi:10.1371/journal.pone.0092536.t001

Discussion
The current study, together with two other previous reports [16,17], provides an insight into the structural/functional aspects of the currently known breast cancer-associated mutations in ADAM12. These mutations are scattered along the entire length of the protein, and they do not appear to cluster within a specific region in the three-dimensional model of ADAM12-L. Among three novel mutations characterized for the first time in this study, only one -G668A -has a severe impact on the intracellular trafficking and proteolytic processing of the protein. Considering the very conservative nature of the Gly-to-Ala substitution, this finding is somewhat unexpected. However, as Gly is highly flexible and can adopt conformations that are forbidden for other amino acids, it is possible that the replacing Gly with Ala at position 668 is not compatible with the native structure of ADAM12-L. High conservation of Gly668 between ADAM12-L and other related ADAMs ( Figure 1D) further suggests the importance of a Gly residue at this position.
The G668A mutation adds to the other two mutations that were previously found to cause ER retention and a lack of proteolytic processing of mouse ADAM12, and are now confirmed to inhibit the processing of human ADAM12-L, i.e., D301H and G479E. While we do not find any evidence that these mutants are unfolded and rapidly degraded, they must assume a significantly different conformation from the WT ADAM12-L to be retained in the ER. Clearly, since these three mutants exist mostly as the ,120-kDa precursors, with the inhibitory pro-domain intact, they are expected to be catalytically inactive. The lack of catalytic activity of the D301H, G479E, and G668A mutants has been confirmed here by the DLL1 cleavage assay, by the EGFR activation assay, and by the Transwell migration assay.
The effects of D301H, G479E, and G668A mutations on the structure/function of ADAM12-L agree well with the effects predicted by the SIFT and PolyPhen tools (see Table 1). Interestingly, the T596A mutation is predicted to be tolerated by SIFT, but harmful by the PolyPhen-2 algorithm. We find that the T596A substitution does not affect intracellular processing of ADAM12, and this result is consistent with the fact that with Thr596 is poorly conserved between ADAM12-L and other closely related human ADAMs ( Figure 1D). However, we find that the T596A mutation renders ADAM12-L inactive at the cell surface. Thr596 is not located at the active site and most likely it is positioned distantly from the metalloproteinase domain in the three-dimensional structure of ADAM12 ( Figure 1B). It is currently unclear whether the T596A substitution exerts a longrange inhibitory effect on the catalytic site or whether it blocks the interaction of ADAM12-L with its substrates. Overall, we conclude that the D301H, G479E, T596A, and G668A mutations should be classified as loss-of-function mutations.
The most striking observation emerging from this study is the apparent association between the catalytic activity of ADAM12-L mutants and the type of breast cancer. Two mutations that did not have any impact on the catalytic activity of ADAM12-L, R612Q and L792F, were found in triple-negative tumors. In contrast, all four loss-of-function mutations, D301H, G479E, T596A, and G668A, occurred in non-triple-negative tumors. Thus, it appears that loss-of-function mutations in ADAM12-L tend to be excluded from TNBCs. However, a larger patient population and possibly a broader mutation spectrum are needed to test whether there is indeed a significant association between the type of ADAM12-L mutations and the triple-negative status of breast tumors. Interestingly, the ADAM12 gene is located in the genomic region that has been recently found to be significantly deleted in Luminal B tumors (cytoband 10q26.11, wide peak boundaries chr10:104674916-135534747, q value 0.0056373) [34]. Low qvalues associated with gene amplifications/deletions (typically below 0.25) suggest that amplifications/deletions at a particular locus are enriched by selective pressures [35]. The same study found that the ADAM12 gene was hypermethylated and showed lower expression in Luminal B tumors than in other types of breast cancer [35]. As Luminal B tumors are estrogen receptor-positive [36,37], these results, together with our functional analysis of ADAM12 breast cancer-related mutants presented here, collectively suggest that ADAM12 may play fundamentally different roles in TNBCs and in non-TNBC.
We have recently postulated that ADAM12-L may be the primary protease responsible for the activation of EGFR in early stage, lymph node-negative TNBCs [10]. This conclusion was supported by decreased distant metastasis-free survival times of patients with high expression levels of ADAM12-L, increased EGFR phosphorylation in a mouse xenograft model of breast cancer, and a strong correlation between the level of anti-ADAM12-L and anti-phospho-EGFR immunostaining in human breast tumor samples. We have also noticed a positive correlation between ADAM12-L and HB-EGF and EGFR in TNBCs, but not in receptor-negative non-TNBCs [10]. In estrogen receptorpositive MCF-7 breast cancer cells, overexpression of ADAM12-L promoted estrogen-independent proliferation, and this effect of ADAM12-L was linked to elevated EGFR activation [9]. Furthermore, a recent study demonstrated cancer cells under hypoxia up-regulate ADAM12-L expression, leading to increased HB-EGF shedding, EGFR activation, formation of invadopodia, and cancer invasion [26]. We believe that these results, together with the analysis of breast cancer-associated ADAM12-L mutants described here, point to an important role of ADAM12-L in the pathology of triple negative breast cancer.