CD95L Cell Surface Cleavage Triggers a Prometastatic Signaling Pathway in Triple-Negative Breast Cancer

Human breast tumors are heterogeneous, both in their clinical expression and molecular profiles. Triple-negative breast cancers (TNBC) are characterized by negative immunohistochemical staining for estrogen and progesterone receptors and HER2, and represent 10% to 20% of all breast cancers. TNBCs account for early deaths of a disproportionate number of patient due to their aggressive nature and the lack of effective therapeutic treatment options (1). Gaining insights into the molecular mechanisms that promote TNBC invasiveness could be of significant value for the design of treatment strategies aimed at preventing metastatic dissemination of these cancers. In addition, prognostic markers are needed to discriminate among those patients with TNBC who are at highest risk for metastatic relapse.

CD95 ligand (CD95L, also known as FasL) belongs to the TNF family and is the ligand for the “death receptor” CD95 (Fas/APO1). CD95L is a transmembrane “cytokine” whose extracellular domain can be cleaved by metalloproteases (2), to produce a soluble ligand. This soluble form was initially described as an inert ligand that competes with its membrane-bound counterpart for binding to CD95, thus acting as an antagonist of the death signal (3, 4). More recent findings have shown that metalloprotease-cleaved CD95L (cl-CD95L) can actively participate in aggravating inflammation in chronic inflammatory disorders, such as systemic lupus erythematosus (5, 6), and may exert pro-oncogenic functions by promoting the survival of ovarian and liver cancers (7) and chemotherapy resistance of lung cancers (8) through molecular mechanisms that remain to be elucidated.

Binding of transmembrane CD95L to CD95 leads to the recruitment of the adaptor protein Fas-associated death domain (FADD) to the intracellular region of CD95 called the death domain. In turn, FADD binds to caspases-8 and -10. This CD95/FADD/caspase complex is known as the death-inducing signaling complex (DISC; ref. 9) and plays a pivotal role in the initiation of the apoptotic signal. In contrast, cl-CD95L fails to induce DISC formation and instead promotes the formation of an atypical receptosome that we have designated motility-inducing signaling complex (MISC; ref. 6). MISC formation leads to the induction of the pro-oncogenic phosphoinositide 3-kinase (PI3K) signaling pathway (6, 10) through a molecular mechanism that remains to be elucidated.

Here, we show that the level of cl-CD95L is higher in the blood of patients with TNBC than in that of patients without TNBC and is associated with increased risk of developing distant metastases. Moreover, we demonstrate that after cleavage by a metalloprotease, the soluble CD95L promotes the motility of TNBC cells by inducing Nox3 (nicotinamide adenine dinucleotide phosphate-oxidase oxidase-3)-driven reactive oxygen species (ROS) generation, which activates Src kinase c-yes, leading to PI3K signaling through activation of the EGF receptor (EGFR) in an EGF-independent manner.

All clinical investigation was conducted in accordance with the principles outlined in the Declaration of Helsinki.

Human breast adenocarcinoma cells MDA-MB-231, MDA-MB-468, and Hs578T were maintained in Dulbecco' Modified Eagle Medium supplemented with 8% (v/v) heat-inactivated fetal calf serum and 2 mmol/L l-glutamine at 37°C in a 5% CO₂ incubator. Silencing experiments were carried out by lentiviral transduction of TNBC cells using shRNAmir-pGIPZ vectors (Open Biosystems). All reagents are described in Supplementary Materials and Methods.

NOD/SCID (nonobese diabetic/severe combined immunodeficient)/γc null mice (NSG) were obtained from Charles River Laboratory (Margate, UK). Luciferase-expressing MDA-MB-231 cells resuspended with or without cl-CD95L (100 ng/mL; 5 × 10⁵ cells in 50 μL PBS/Matrigel) were transplanted into mammary fat pads of mice (7-week-old female). Next, caudal vein injections of cl-CD95L (10 μg/kg) or control medium were repeated 5 days a week until day 34. Bioluminescence analysis was performed using PhotonIMAGER (Biospace Lab), following intraperitoneal injection of luciferin (30 mg/kg). Tumor volume was calculated using the formula |$V\, = \,0.52\,(L \times W2)$|⁠. After completion of the analysis, autopsy of mice was done and organ luminescence was assessed.

TNBC cells (3 × 10⁷ cells) were incubated with 100 ng/mL of cl-CD95L (MISC) or APO1-3 (DISC) for the indicated times. The cells were lysed, 1 μg of Apo1-3 was added to the cell lysate, and CD95 was immunoprecipitated by the addition of protein A-sepharose beads (Sigma). For the immunoprecipitation of EGFR, 1 μg of anti-EGFR (Santa Cruz Biotechnology) was added to the cell lysate and EGFR was immunoprecipitated with protein A-sepharose beads. After extensive washing, the immune complex was resolved by SDS-PAGE.

Ca²⁺ monitoring was performed as previously described (11).

Statistical analyses and Supplementary Materials and Methods are available online.

Although high CD95L levels have been detected by immunohistochemistry in breast cancer tissue (12), the role of this ligand in carcinogenesis remains unknown. To address this question, serum CD95L levels were measured in women with breast cancer. Blood dosages showed higher CD95L levels in women affected by TNBC than in patients without TNBC (98.94 ± 45.37, n = 39 vs. 52.79 ± 26.2, n = 103; P < 0.0001) and subjects with benign breast diseases (98.94 ± 45.37, n = 39 vs. 30.04 ± 28.52, n = 8; P < 0.0001; Fig. 1A). Of note, the quantity of CD95L exhibited a heterogeneous distribution in patients without TNBC (Fig. 1A); however, when this cohort of patients was subdivided on the basis of the occurrence of relapse, CD95L levels were significantly higher in relapsing patients than in those without relapse (62.72 ± 31.31, n = 39 vs. 47.68 ± 21.77, n = 64; P = 0.0053; Fig. 1A). In addition, Kaplan–Meier analyses revealed that both patients with and without TNBC with CD95L concentrations ≥80 pg/mL had significantly reduced disease-free survival (Fig. 1B) and an increased occurrence of metastases (Fig. 1C). This predisposition to metastasis was even more pronounced in patients with serum CD95L concentrations ≥120 pg/mL (Fig. 1C). To further investigate whether cl-CD95L is endowed with a prometastatic role toward TNBC cells, we orthotopically transplanted luminescent TNBC cells (MDA-MB-231) in NOD/SCID/γc mice and monitored both tumor growth and metastatic dissemination with or without repeated injections of purified cl-CD95L. Of note, injection of cl-CD95L did not modify the tumor volume (Supplementary Fig. S1A), whereas as observed by bioluminescence imaging, it augmented the spread of TNBC cells, which seemed to metastasize to the lungs (Fig. 1D). To confirm this observation, mice were sacrificed and luminescence was assessed in different organs (Fig. 1E). Although no luminescence was detected in hearts and livers of control and cl-CD95L–injected mice (Supplementary Fig. S1B), lungs- and brachial-draining lymph nodes experienced a dramatic TNBC cell invasion in mice injected with cl-CD95L as compared with control mice (Fig. 1E). Overall, these findings indicated that high amounts of soluble CD95L mainly detected in TNBC women may participate in metastatic progression.

Next, to identify cells responsible for CD95L expression in patients with breast cancer, immunohistochemistry was performed in breast tissue sections from healthy, patients with and without TNBC (Fig. 1F). CD95L staining was undetectable in healthy subjects, whereas breast tissues from patients with and without TNBC exhibited a level of CD95L expression correlated with serum concentrations (Fig. 1F). Indeed, CD95L was expressed in an increasing gradient from healthy subjects to patients with TNBC (Fig. 1F). Furthermore, CD95L was expressed in endothelial cells of blood vessels (CD31⁺) surrounding tumor masses but not in the lymphatic endothelium (D2-40⁺; Fig. 1G). These findings indicated that CD95L is ectopically expressed in tumor blood vessels and its cleavage by metalloprotease releases a poor prognostic marker associated with metastatic dissemination in patients with TNBC.

Given that CD95L stems from blood vessels surrounding breast cancer cells and is associated with the risk of metastasis, we next wondered whether this ligand contributed to TNBC intravasation by promoting cell motility. Different TNBC cell lines exposed to cl-CD95L showed a dramatic increase in their motility (Fig. 2A). Exposure to amounts of CD95L equal to those measured in patients with breast cancer was sufficient to trigger cell motility. Indeed, titration of cl-CD95L showed that cl-CD95L at 100 pg/mL, a dose corresponding to the mean concentration measured in patients with TNBC (98.94 ± 45.37 pg/mL; Fig. 1A), induced migration of TNBC cells (Supplementary Fig. S1C). To confirm the effect of soluble CD95L on promoting the motility of breast cancer cells, we incubated TNBC cells in the presence of sera from healthy donors or patients with TNBC and evaluated cell migration using Boyden chambers. The results showed that the sera from women with TNBC, unlike that from healthy subjects, contained a molecule that enhanced the migration of breast cancer cells (Fig. 2B). This promotile factor was identified as soluble CD95L because preincubation of TNBC sera with a neutralizing anti-CD95L antibody abrogated their promotile effect (Fig. 2B).

Activation of the PI3K signaling pathway promotes different steps leading to metastasis, including cell motility and intravasation/extravasation (13). PI3Ks are lipid kinases generating the second-messenger phosphatidylinositol (3,4,5)-trisphosphate (PIP3), which serves as a docking site for various signaling factors. For instance, binding of the serine–threonine kinase Akt to PIP3 leads to its redistribution to the plasma membrane, where it undergoes phosphorylation and activation at Ser⁴⁷³ (for review, see ref. 14). In all tested TNBC cells, cl-CD95L induced rapid phosphorylation of Akt at Ser⁴⁷³ (Supplementary Fig. S2A). Furthermore, inhibition of the PI3K signal using the pharmacologic inhibitor LY294002 abrogated the migration of TNBC cells exposed to cl-CD95L (Supplementary Fig. S2B), indicating that after cleavage by metalloprotease, this ligand acts as a chemoattractant for malignant breast cells through activation of PI3K. Class I PI3Ks comprise four catalytic isoforms (α, β, δ, and γ). P110α and -β are ubiquitously expressed, whereas expression of p110δ and -γ is restricted to hematologic cells. To evaluate the contribution of p110α and -β in CD95-mediated cell motility in TNBC cells, we analyzed the effect of pharmacologic inhibitors of p110α and -β (Supplementary Fig. S2C). The p110β-inhibitor TGX-221 was much more efficient than a drug-targeting p110α (PIK-90) to prevent both Akt phosphorylation (Supplementary Fig. S2D) and cell migration (Supplementary Fig. S2E) in TNBC cells exposed to cl-CD95L. To confirm these observations, RNA interference was used to silence expression of p110α or -β isoform (Supplementary Fig. S2F). Although downregulation of p110α did not alter CD95-mediated PI3K activation (Fig. 2C) and cell migration (Fig. 2D), silencing of p110β inhibited Akt phosphorylation (Fig. 2C) and cell motility (Fig. 2D). These findings demonstrated that cl-CD95L selectively activates the PI3K catalytic subunit p110β, which promotes cell migration in TNBC cells.

To decipher the initial events leading to p110β activation, TNBC cell lines were stimulated with cl-CD95L followed by immunoprecipitation of CD95. An analysis of the resulting immune complex revealed that CD95 did not bind to FADD and caspase-8 but did recruit c-yes (Fig. 2E). In addition, downregulation of c-yes expression (Supplementary Fig. S2G) prevented both Akt phosphorylation (Supplementary Fig. S2H) and cell migration (Fig. 2F) in TNBC cells exposed to cl-CD95L.

Calcium ions (Ca²⁺) participate in cell signaling as a second messenger that relies on intensity (cytosolic concentration), temporal parameters (i.e., duration and frequency), and spatial localization to trigger a variety of cellular responses. Because Ca²⁺ plays a pivotal role in cell motility (15), we analyzed its impact on the signaling pathway induced by cl-CD95L in breast cancer cells. In nonexcitable cells, Ca²⁺ responses mainly occur through a biphasic signal caused by activation of inositol 1,4,5-trisphosphate (IP₃) receptors and the release of Ca²⁺ from the endoplasmic reticulum followed by a sustained Ca²⁺ entry across the plasma membrane (16). Recently, STIM1 was identified as the endoplasmic reticulum–located Ca²⁺ sensor that links endoplasmic reticulum depletion to activation of the plasma membrane Ca²⁺ channel formed by Orai1 subunits, allowing Ca²⁺ to enter the cell (17). This store-operated calcium (Ca²⁺) entry (SOCE) plays pivotal roles in both the replenishment of the endoplasmic reticulum store and cell signaling (18). Pretreatment of breast cancer cells with the Ca²⁺ chelator, BAPTA-AM or the pharmacologic inhibitor of SOC channels BTP2 inhibited both CD95-mediated PI3K activation and cell migration (Fig. 3A and B). To determine whether Orai1 participated in CD95 signaling in TNBC cells, Orai1 expression was blocked using shRNAmir. As shown in Fig. 3C, downregulation of Orai1 did not alter the CD95-mediated Ca²⁺ mobilization from the endoplasmic reticulum, but it abolished the subsequent Ca²⁺ entry observed in TNBC cells stimulated with cl-CD95L. Of note, Orai1-driven Ca²⁺ entry was essential for inducing PI3K activation and cell migration (Fig. 3D and E). To further characterize the mechanistic link between CD95 engagement by cl-CD95 and Ca²⁺ signaling in breast tumor cells, the contribution of c-yes and p110β to the CD95-mediated Ca²⁺ response was examined. Silencing of c-yes abrogated CD95-mediated Ca²⁺ signaling (Fig. 3F). On the other hand, although Ca²⁺ signaling contributed to PI3K activation (Fig. 3A), the opposite was not true, as p110β knockdown did not affect the CD95-mediated Ca²⁺ response (Fig. 3F). These results indicated that c-yes occupies a proximal position in the sequence of events, leading to the activation of the Ca²⁺/PI3K signaling pathway in breast tumor cells exposed to cl-CD95L.

Because recent reports have shown that Src kinases can behave as redox sensors promoting cell migration in leukocytes (19), we investigated whether CD95 activates c-yes through the generation of ROS. Different TNBC cells exposed to cl-CD95L experienced a rapid increase in intracellular ROS (Supplementary Fig. S3A), which was blocked by pretreatment with reduced nicotinamide adenine dinucleotide phosphate oxidase (NADPHox) inhibitors, such as DPI and apocynin (Fig. 4A). To determine the effect of ROS on c-yes activation, c-yes was immunoprecipitated and its autophosphorylation status (hallmark of its activation) was evaluated by immunoblotting using a phospho-Src family antibody, recognizing Src when phosphorylated at tyrosine 416 and c-yes at an equivalent site (tyrosine 426). In the presence of cl-CD95L, TNBC cells undergo rapid activation of c-yes that is inhibited by NADPHox inhibitors (Fig. 4B and Supplementary Fig. S3B). Also, pretreatment of mammary cancer cells with DPI or apocynin inhibited CD95-mediated activation of Akt (Supplementary Fig. S3C and S3D) and cell migration (Supplementary Fig. S3E). The Nox family is defined by seven distinct catalytic isoforms, namely Nox1 to -5 and Duox1 and Duox2. Although Nox1, -2, -3, and -4 are associated with p22^Phox, which is involved in their proper membrane targeting and activity, Nox5, Duox1, and Duox2 do not require p22^phox for their activity (20). MISC analysis showed that p22^Phox was recruited after stimulation with cl-CD95L (Fig. 4C). In addition, downregulation of p22^Phox (Supplementary Fig. S3F) prevented activation of c-yes and Akt (Fig. 4D) and cell migration in TNBC cells exposed to cl-CD95L (Fig. 4E and Supplementary S3G), supporting that association of p22^Phox with Nox1, -2, -3, or -4 orchestrated the CD95-mediated ROS production. To identify the Nox subunit responsible for ROS generation in TNBC cells stimulated with cl-CD95L, we next analyzed the recruitment of Nox1, -2, -3, or -4 to the MISC. As shown in Fig. 4F, only Nox3 was recruited in the MISC of TNBC cells stimulated with cl-CD95L. In addition, although Nox2 downregulation (Supplementary Fig. S3H) did not alter cell migration in cl-CD95L–stimulated TNBC cells (Supplementary Fig. S3I), Nox3 silencing (Supplementary Fig. S3H) completely abrogated it (Supplementary Fig. S3I). Finally, pharmacologic and genetic inhibitions of NADPH oxidase completely abrogated the CD95-mediated Ca²⁺ response (Supplementary Fig. S3J), confirming that CD95-triggered ROS generation is a proximal event in the signaling cascade induced by metalloprotease-cleaved CD95L in TNBC cells. These findings highlighted that TNBC cells exposed to cl-CD95L undergo a rapid Nox3-driven ROS production, resulting in the activation of c-yes and its downstream signaling pathway.

Taken together, these findings raise the question of how PI3K can be activated in TNBC cells in the presence of cl-CD95L. An in silico analysis of short linear motifs (21) in the intracellular region of CD95 did not reveal any consensus sequences that could account for recruitment of p110β or its regulatory subunit p85 after “death receptor” phosphorylation by c-yes, suggesting that at least one additional factor connects CD95 to PI3K signaling. The Src kinase family has been reported to phosphorylate EGFR at Tyr845, a modification that stabilizes the activation loop and maintains the receptor in an active state (22). Moreover, Src kinase can also be associated with EGFR to form a heterocomplex in TNBC cells (23). Because 72% of TNBC cells express EGFR (24) and its expression is correlated with cell migration (25), we examined the potential contribution of EGFR to CD95 signaling. EGFR was expressed predominantly in TNBC cell lines compared with non-TNBC cells (Supplementary Fig. S4A and S4B). On the basis of our immunoprecipitation experiments showing an interaction between CD95 and c-yes (Fig. 2E), we hypothesized that recruitment of c-yes to MISC may lead to EGFR activation, which in turn may serve as a molecular platform to elicit the PI3K (p110β) signaling pathway. To explore this hypothesis, the phosphorylation status of EGFR at Tyr845 was monitored. As shown in Fig. 5A, exposure of MDA-MB-468 breast cancer cells to cl-CD95L resulted in phosphorylation of EGFR at Tyr845, which reached a peak at 2 minutes and disappeared after 10 minutes of stimulation. This EGFR phosphorylation preceded Akt activation (Fig. 5A). EGFR activation relied on CD95 stimulation because preincubation of TNBC cells with neutralizing anti-CD95L antibody (Nok-1) totally abrogated EGFR phosphorylation (Supplementary Fig. S4C). In addition, EGFR phosphorylation was dependent on c-yes, as shown by the loss of EGFR phosphorylation in TNBC cells in which c-yes expression was suppressed (Fig. 5B). Next, we used biochemical and imaging approaches to elucidate whether EGFR was recruited to the MISC in TNBC cells exposed to cl-CD95L. Immunoprecipitation of CD95 or EGFR from TNBC cells exposed to cl-CD95L revealed the rapid formation of an immune complex containing CD95/c-yes/EGFR/p110β (Fig. 5C and D). These biochemical observations were confirmed by the colocalization of EGFR and CD95 at the leading edge of emitted pseudopodia in TNBC cells (Supplementary Fig. S4D and S4E). Overall, these findings demonstrated that EGFR is recruited to the MISC in TNBC cells exposed to cl-CD95L. To further investigate the role of EGFR in the CD95-mediated cell motility, we used the EGFR inhibitor erlotinib. Erlotinib prevented CD95-mediated Akt phosphorylation in TNBC cells (Supplementary Fig. S5A and S5B). In addition, erlotinib abolished migration of TNBC cells exposed to cl-CD95L (Supplementary Fig. S5C). To prove that EGFR was instrumental in CD95-mediated cell signaling, the tyrosine kinase receptor was next silenced (Supplementary Fig. S5D). Downregulation of EGFR expression inhibited both CD95-mediated PI3K activation (Fig. 5E) and cell migration in TNBC cells (Fig. 5F and Supplementary Fig. S5E). Examination of the role of EGFR in the cl-CD95L–induced Ca²⁺ response showed that although the binding of EGF to EGFR evoked a Ca²⁺ response, which was blocked by erlotinib (Supplementary Fig. S5F), treatment with neither erlotinib nor EGFR silencing altered the CD95-mediated Ca²⁺ response (Supplementary Fig. S5F). This latter observation questioned the involvement of EGF in the recruitment and activation of EGFR in breast cancer cells stimulated with cl-CD95L. To address the possible involvement of EGF in this process, cl-CD95L–mediated signaling was examined in TNBC cells preincubated with cetuximab, an antibody that binds to the EGFR ectodomain and prevents its interaction with EGF (26). Although this blocking antibody abrogated the EGF-induced Ca²⁺ response, it did not affect CD95-mediated Akt activation (Supplementary Fig. S5G) and cell motility (Supplementary Fig. S5H), supporting the notion that CD95-driven recruitment of EGFR occurs through an EGF-independent mechanism in TNBC cells.

The results of the present study show that an elevated serum concentration of CD95L in women diagnosed with breast cancer is associated with poor prognosis and high risk of distant metastasis. Furthermore, the distribution of CD95L on endothelial cells of blood vessels suggests that its shedding by a yet unknown metalloprotease may create the concentration gradient required to promote intravasation of breast tumor cells and thus distant cancer metastases.

Further analysis of the molecular pathway connecting the “death receptor” CD95 to downstream PI3K signaling in TNBC cells exposed to cl-CD95L showed the recruitment and activation of EGFR. Although EGFR was necessary for CD95-mediated activation of PI3K, it was dispensable in the Ca²⁺ response, indicating that the two pathways diverge early in the unconventional pathway of cl-CD95L–triggered signaling. This divergence must occur downstream of ROS generation and subsequent c-yes activation and upstream of EGFR activation, as inhibition of NADPH oxidase and silencing of c-yes expression abolished both PI3K activation and Ca²⁺ signaling in TNBC cells exposed to cl-CD95L. In the presence of EGF, TNBC cells evoked a Ca²⁺ response that was abolished by pretreatment with the tyrosine kinase inhibitor erlotinib. Given that erlotinib and EGFR downregulation did not alter the Ca²⁺ signal observed in breast cancer cells exposed to cl-CD95L, we surmise that recruitment and activation of EGFR by CD95 occurs through an atypical mechanism independent of EGF. Confirming this assumption, the EGFR neutralizing antibody cetuximab did not impair CD95-driven EGFR activation in breast cancer cells. Conventionally, binding of EGF ligands to EGFR leads to receptor dimerization, activation of its intrinsic tyrosine kinase activity and subsequent phosphorylation of downstream signaling molecules. However, this dogma has been challenged by the discovery that the activation of EGFR-mediated signaling can occur in a ligand-independent manner in the presence of ROS (27). In addition, G protein–coupled receptor activation can mediate EGFR transactivation through Src family tyrosine kinases (28), and c-Src itself is able to facilitate EGFR activation by phosphorylation of Tyr⁸⁴⁵ (22). Our results bring to light a novel mechanism for EGFR activation in TNBC cells stimulated with metalloprotease-cleaved CD95L. From a mechanistic standpoint, CD95-driven EGFR activation is dependent on the generation of ROS by Nox3. ROS activate c-yes, which in turn recruits and activates EGFR.

Similar to EGFR, c-Met is a transmembrane receptor tyrosine kinase for hepatocyte growth factor. Recent reports have shown that the CD95 signaling pathway can be modulated by c-Met (29). Indeed, the amino-acid residues YLGA in c-Met can bind CD95 (29) and prevent its homotrimeric self-aggregation, a pivotal step in the implementation of the apoptotic signal (30). It is noteworthy that the YLGA motif is not detected in the EGFR sequence, and although the CD95/c-Met association is lost when CD95L interacts with its receptor (31), binding of cl-CD95 is mandatory for recruitment of EGFR by CD95. Taken together, these findings strongly suggest that receptor tyrosine kinases such as c-Met and EGFR can interact with CD95 by different molecular mechanisms. More importantly, these results indicate that implementation of CD95-mediated nonapoptotic signals may occur through the recruitment of tyrosine kinase receptors that behave as docking sites for several proteins, including PI3K.

Because serum CD95L level is associated with relapse in breast cancers, exhaustive identification of the components of the MISC and characterization of the molecular pathway leading to the induction of CD95-mediated nonapoptotic signaling are critical to understand how this “death receptor” can transmit non-death signals depending on the ligand with which it interacts. Our findings point to the activation of the Src kinase c-yes, which implements Orai1-mediated SOCE, recruits EGFR, and induces the downstream activation of the p110β catalytic isoform of PI3K (summarized in Fig. 6).

In summary, this study identified serum CD95L level as a new prognostic marker of metastatic dissemination in women with breast cancer. This finding may help to guide clinicians in the selection of the most appropriate treatment regimen for patients with high levels of cl-CD95L who should undergo intensive therapy, whereas patients with low cl-CD95L should be spared. In addition, the elucidation of this atypical CD95-mediated signaling pathway provides new therapeutic targets for preventing metastatic dissemination in TNBC. Although inhibitors of EGFR kinase activity may be attractive therapeutic agents, especially because gefitinib and erlotinib are already in use for the treatment of non–small cell lung cancer, most patients on prolonged gefitinib and erlotinib treatment develop secondary mutations in the EGFR kinase domain that block drug binding, leading to clinical resistance (32, 33). Therefore, inhibition of the CD95/CD95L interaction may be a more appropriate treatment approach, especially because such an inhibitor already exists and is well tolerated by patients (34).

M. Campone has honoraria from speakers' bureau and is a consultant/advisory board member of Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Malleter, S. Tauzin, R. Castellano, Y. Collette, P. Vacher, P. Legembre

Development of methodology: M. Malleter, S. Tauzin, A. Bessede, R. Castellano, A. Goubard, L. Debure, P. Vacher

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Malleter, S. Tauzin, R. Castellano, A. Goubard, F. Godey, J. Levêque, P. Jézéquel, M. Campone, T. Ducret, G. MacGrogan, P. Vacher

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Malleter, S. Tauzin, R. Castellano, L. Campion, M. Campone, P. Vacher, P. Legembre

Writing, review, and/or revision of the manuscript: M. Malleter, S. Tauzin, L. Campion, M. Campone, Y. Collette, P. Legembre

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Bessede, F. Godey, P. Jézéquel

Study supervision: Y. Collette, P. Legembre

The authors thank A.M. Vacher and V. Velasco (Institut Bergonié, Bordeaux), O. Cabaud, TrGET and animal facilities (CRCM, Marseille), and L3 and microscopy facilities (SFR Biosit, Rennes) for their technical assistance.

This work was supported by INCa, Ligue Contre le Cancer (Comités d'Ille-et-Vilaine/du Morbihan/des Côtes d'Armor/du Maine et Loire et des Landes), ARC, Cancéropole GO, Région Bretagne, Région Pays de la Loire, and Rennes Métropole.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.