Targeting TACE-dependent EGFR ligand shedding in breast cancer (original) (raw)
AREG and TGFA are upregulated in T4-2 cells. The nonmalignant human breast epithelial cells in this model required exogenous EGF for proliferation (Figure 1E), while their malignant derivative, T4-2, acquired self-sufficiency for this signal. The sensitivity of T4-2 cells to inhibition of EGFR (10) implies that EGFR and the downstream components of the pathway are not constitutively activated. Using direct sequencing, we showed that these cells had not sustained activating mutations in EGFR, H-Ras, K-Ras, N-Ras, or B-Raf (data not shown). As previously described (10), T4-2 cells had significantly higher levels of active EGFR than did their S1 precursors (Figure 1A). Thus, we hypothesized that T4-2 cells have escaped dependence on exogenous EGF by transcriptionally upregulating one or more ErbB ligands. Conditioned medium from T4-2 cells elicited rapid activation of MAPK in S1 cells, which was comparable to that induced by exogenously added EGF (Figure 1B). While ligands of a number of receptor tyrosine kinases activate MAPK, the observed activation was suppressed by preincubation of S1 cells with the EGFR inhibitor gefitinib (Iressa, ZD1839; AstraZeneca). Thus we suspected that T4-2 cells produce one or more soluble EGFR ligands. We tested expression of the genes encoding AREG, Betacellulin, Cripto, EGF, Epiregulin, HB-EGF, NRG1, NRG2, and TGF-α by RT-PCR. AREG and TGFA were expressed at high levels in T4-2 cells (Figure 1C). Using ELISA, we confirmed the presence of AREG and TGF-α in the conditioned medium of T4-2 cells (Figure 1D). Adding concentrations of recombinant AREG or TGF-α equimolar to that of EGF to the medium of S1 cells (860 pM) showed that these ligands can substitute for EGF to promote proliferation of the nonmalignant cells (Figure 1E).
Upregulation of an autocrine growth factor loop during a model of breast cancer progression. (A) T4-2 (malignant) cells, which grow independently of exogenous EGF, had significantly higher activity of EGFR than their phenotypically normal counterpart, S1 (nonmalignant) cells. The level of EGFR phosphorylation is consistent with activation by a soluble factor produced in these cells. Ponceau S staining was used as a loading control. (B) S1 cells were starved of EGF for 12 hours and then stimulated for 10 minutes with either T4-2 conditioned medium (CM) or 5 ng/ml EGF. A 5 minute pretreatment with Iressa (0.3 nM) abolished MAPK activation induced by the conditioned medium and by EGF. (C) RT-PCR analysis shows that AREG and TGFA were transcriptionally upregulated in T4-2 relative to S1 cells. GAPDH was used as a loading control. (D) ELISA of conditioned medium shows that T4-2 cells secreted significantly more AREG and TGF-α than did S1 cells. (E) S1 cell proliferation in the presence of all EGFR ligands (860 pM) was significantly different from control (Ctrl).
A metalloproteinase activity is critically required for mobilization of growth factors. Several growth factors, including AREG and TGF-α, are synthesized as transmembrane precursors and are processed by the members of the ADAM family of transmembrane proteases (11–13). Culture of T4-2 cells in 3D laminin-rich gels resulted in the formation of disorganized, apolar, continuously proliferating colonies (Figure 2A), a phenotype we have previously shown to be highly correlated with, and reflective of, the ability of cancer cells to form tumors in vivo (7, 16). Incubation with TNF protease inhibitor–2 (TAPI-2), a broad-spectrum inhibitor of MMPs and ADAMs, resulted in a reversion of the malignant phenotype (Figure 2C) similar to that elicited using the EGFR inhibitor AG1478 (Figure 2B), suggesting that a metalloproteinase activity is required for the proliferative phenotype of T4-2 cells. This treatment also resulted in the restoration of epithelial polarity. Vehicle-treated cells remained disorganized (Figure 2D), whereas TAPI-2–treated cells assumed a polar organization reminiscent of breast acini, here indicated by basal localization of α6 integrin (Figure 2E). The colonies formed by AG1478- or TAPI-2–treated T4-2 cells were similar in size to nonmalignant mammary acini and were significantly smaller than those formed by cells treated with vehicle alone (Figure 2F).
Inhibition of sheddase activity reverts the malignant phenotype of T4-2 cells by suppressing mobilization of growth factors and downregulating EGFR pathway activity. (A) T4-2 cells grown in 3D lrECM cultures formed continuously proliferating, disorganized, apolar colonies. (B) T4-2 cells treated with EGFR inhibitor (80 nM AG1478) underwent morphological reversion, forming small, smooth, spherical, growth-arrested colonies. (C) T4-2 cells treated with a broad-spectrum MMP/ADAM inhibitor (20 μM TAPI-2) underwent morphological reversion similar to that of EGFR inhibitor–treated cells. (D) Absence of tissue polarity as demonstrated by α6 integrin staining of vehicle-treated T4-2 cells. (E) Restoration of tissue polarity as demonstrated by α6 integrin staining of TAPI-2–treated T4-2 cells. Scale bars: 100 μm (A–C); 10 μm (D and E). (F) Analysis of cross-sectional area of T4-2 cells treated with either vehicle, AG1478, or TAPI-2 for 4 days. ***P < 0.001 versus control. (G) TAPI-2 treatment (24 hours) reduced the basal activity of kinases downstream of EGFR, but cells remained competent to respond to exogenous EGF (860 pM, 5-minute stimulation). (H) TAPI-2 treatment resulted in a dose-dependent reduction in T4-2 cell proliferation that was completely overcome by addition of soluble EGF. **P < 0.01; ***P < 0.001 compared with 0 μM TAPI-2. (I) ELISA of conditioned medium from TAPI-2–treated T4-2 cells, indicating that it suppressed the shedding of both AREG and TGF-α.
T4-2 cells exhibited a basal level of signaling kinase activity downstream of the EGFR (Figure 2G), consistent with a response to the ongoing production of an EGFR ligand by these cells. The basal activities were significantly suppressed by addition of TAPI-2, but the cells remained competent to respond to addition of exogenous EGF (Figure 2G). Furthermore, TAPI-2 caused a dose-dependent decrease in proliferation of T4-2 cells in 2D cultures, which was also overcome by addition of exogenous EGF (Figure 2H). This compound was not cytotoxic at the concentration used, nor did it interfere with the ability of S1 cells to execute normal acinar morphogenesis in the presence of soluble EGF (data not shown). Thus, the proliferative block and concomitant reversion resulting from metalloproteinase inhibition appeared to result from a defect in growth factor mobilization, which was confirmed by ELISA analysis (Figure 2I).
TACE cleaves both AREG and TGF-α in cultured mammary epithelial cells. Several lines of genetic and biochemical evidence suggest that TACE is a key regulator of cleavage of AREG as well as TGF-α (11–13, 17). TACE is expressed in both S1 and T4-2 cells (Figure 3A). To test whether TACE is the key sheddase for these endogenously produced growth factors in these cells, we used siRNA to knock down expression of TACE and measured growth factor shedding from the transfected cells (Figure 3B). The 3 siRNAs used against TACE suppressed its expression with varying degrees of efficacy. The most effective, siTACE-1, caused a dramatic decrease in the shedding of both ligands, whereas cells transfected with the less effective siRNAs retained the ability to shed ligands in proportion to the amount of TACE expressed. Introduction of the most effective siRNA against TACE (Figure 3C) had no apparent effect on the morphology of cells cultured on plastic, but resulted in a dramatic reversion of the malignant phenotype of T4-2 cells in 3D lrECM culture compared with the random siRNA–transfected control. The shedding of both EGFR ligands was significantly reduced in these cultures (Figure 3D). Thus it appears that TACE, and not another TAPI-2–sensitive protease, is the primary growth factor sheddase in the T4-2 breast cancer cell line.
TACE cleaves AREG and TGF-α and is necessary for T4-2 cell proliferation. (A) RT-PCR analysis showing TACE expression in S1 and T4-2 cells. GAPDH was used as a loading control. (B) ELISA analysis of EGFR ligand shedding in T4-2 cells transfected with 3 siRNA oligos, either individually or as a pool. Ligand shedding was proportional to the level of TACE expression. (C) Reversion of the malignant phenotype of T4-2 cells in 3D lrECM culture following transfection of siRNA against TACE. Left insets: Phase-contrast micrographs of transfected cells grown on plastic. Right insets: α6 integrin immunostaining of representative colonies. Original magnification, ×100; ×600 (right insets). (D) ELISA analysis of the conditioned medium from the experiment shown in C.
AREG and TGF-α are the key substrates of TACE in T4-2 cells. In addition to shedding growth factors, TACE has been implicated in the shedding of several other cell surface molecules, the inhibition of which might also contribute to the observed reversion of the T4-2 cell phenotype. Characterized substrates of TACE include TNF-α (18, 19), L selectin and TNFRII (20), β-APP (21), collagen XVII (22), growth hormone receptor (23), TrkA (24), ErbB4 (25), and GPIbα (26). To test whether modulation of growth factor cleavage is the key role of TACE and whether overexpression of these substrates leads to a genetic rescue of the TAPI-2–imposed reversion, we generated soluble secreted mutants of AREG and TGF-α lacking both the transmembrane and the cytosolic domains (AREGΔTM and TGF-αΔTM, respectively; Figure 4A). Whereas each stably infected T4-2 cell line was susceptible to reversion by EGFR inhibition (Figure 4B), those cells that produced soluble growth factors were completely resistant to TAPI-2 by criteria of colony size and morphology (Figure 4, B and C). Much like the parental cells, they continued to proliferate and formed disorganized, nonpolarized colonies (Figure 4D). Thus, despite the number of TACE substrates expressed by these cells, it is the suppression of growth factor mobilization that results in the reversion of the malignant phenotype.
TAPI-2–induced reversion of T4-2 cells is a direct result of inhibition of growth factor ectodomain shedding. (A) Schematic representation of full-length (pro-) and deletion mutants (ΔTM) of AREG and TGF-α. Deletion mutants lack both the transmembrane and the cytoplasmic domains and can thus be secreted without requiring TACE activity. (B) T4-2 cells overexpressing full-length or deletion growth factor constructs were susceptible to reversion by the EGFR inhibitors, but cells expressing either soluble AREG or soluble TGF-α escaped the TAPI-2–induced reversion. Scale bar: 100 μm. (C) Analysis of cross-sectional area of T4-2 cells and derivatives in response to pharmacological inhibition of EGFR and TACE. Horizontal bars represent median values. (D) Higher-magnification (×600) analysis of representative colonies from B. Colonies expressing the soluble mutants of AREG and TGF-α remain disorganized and apolar in the presence of TAPI-2.
TACE inhibition reduces EGFR ligand shedding in several breast cancer cell lines. To test whether our observations were generalizable, we screened several additional breast cancer cell lines to identify those that secrete either AREG or TGF-α. AREG was secreted by MCF-7, HCC1500, and ZR75B cells, while TGF-α was secreted by HCC1500 and MDA-MB-468 cells. In each case, 20 μM TAPI-2 significantly reduced ligand shedding from these cells (Figure 5, A and B). Transfection of MCF-7 cells with the siRNA against TACE led to a reduction in AREG shedding by 90%, indicating that, as with T4-2 cells, TACE is the key sheddase in this cell line (data not shown). We examined the activity of the EGFR pathway in these cell lines treated with TAPI-2 for 1 or 5 hours and found significant downregulation of MAPK activity in those cells expressing high levels of EGFR (Figure 5C). These data indicate that TACE-dependent growth factor shedding is common, at least in established breast cancer cell lines, and that it is amenable to therapeutic intervention.
Suppression of growth factor shedding by TAPI-2 in a panel of breast cancer cell lines. (A) Three AREG-expressing breast cancer lines were treated with 20 μM TAPI-2 or vehicle for 90 minutes, and AREG shedding was quantified by ELISA. (B) Two breast cancer cell lines expressing TGF-α were identified and treated as in A, and TGF-α shedding was quantified by ELISA. TAPI-2 suppressed TGF-α shedding. (C) Each cell line was treated with TAPI-2 for either 1 or 5 hours. Downregulation of MAPK activity was detected in those cell lines expressing EGFR.
TACE and TGFA predict poor prognosis in human breast cancer patients. Having thus established that TACE-dependent growth factor shedding plays a role that is both critically important and therapeutically tractable in the HMT3522 model of breast cancer progression and demonstrated its prevalence in other breast cancer models, we sought to determine the extent to which these factors play a role in human breast cancer. We interrogated a comprehensive gene expression microarray data set of 295 primary human breast tumors prepared by van de Vijver and colleagues, who used it to identify gene expression signatures predictive of outcome (15). The detailed clinical characteristics of these tumors have been reported but briefly: all were either stage I or II, less than 5 cm in diameter at excision, and derived from 295 consecutively treated patients less than 53 years old. Approximately three-quarters of the tumors were estrogen receptor α (ERα) positive, and half were associated with positive lymph nodes. The median time for which follow-up information is available is 10.2 years (range, 0.05–21.6 years).
Our analysis of this data set revealed a statistically significant positive correlation between expression levels of TGFA and TACE (P < 0.001; Table 1). EGFR expression also correlated with both TGFA and TACE, although not at the level of statistical significance (P = 0.053 and P = 0.061, respectively). Unexpectedly, AREG expression in this patient population was inversely correlated with expression of EGFR and TGFA (P < 0.05 and P < 0.001, respectively). AREG and TACE levels tended to be inversely correlated as well, although not reaching statistical significance (P = 0.053). The explanation appears to involve the subtype of breast cancers examined. Tumors positive for TGFA, TACE, and EGFR were essentially _ER_α negative (P < 0.0001, P < 0.005, and P < 0.0001, respectively). Conversely, _ER_α-positive tumors had higher levels of AREG (P < 0.0001). We stratified the tumors into the standard subtypes (Basal, ErbB2, Luminal A, Luminal B, and Normal breast–like; ref. 27) to determine whether tumors expressing high levels of either growth factor were overrepresented in a particular subtype (Supplemental Table 1; supplemental material available online with this article; doi:10.1172/JCI29518DS1). These data indicate that the tumors with the highest levels of TGFA and TACE were overrepresented in the basal subtype, while tumors expressing high levels of AREG were predominantly found in the normal breast–like subgroup. The data also suggest that TACE and TGF-α may be the more important protease/growth factor pair for EGFR activation in human breast tumors (see Discussion).
Pearson’s correlation analysis of markers in 295 primary human breast tumors
To analyze the contribution of AREG, TGFA, and TACE expression to survival of human breast cancer patients, tumors were divided in quartiles by expression level of each marker, and survival curves were computed for the upper and lower quartiles (74 samples each) and the interquartile range (147 samples). Survival was evaluated at 5 and 10 years after surgery. High levels of TACE expression were associated with poor survival (Figure 6A), as were high levels of TGFA (Figure 6B). Tumors that expressed high levels of AREG, however, had a better outcome than those with lower AREG expression (P < 0.005, Figure 6C), as expected based on the high correlation between AREG and _ER_α expression levels. Of the 74 tumors with the highest levels of AREG expression, 73 were _ER_α-positive, i.e., only 1 _AREG_-high tumor was ERα negative. While a positive association between an EGFR ligand and good outcome may at first appear counterintuitive, it is important to note that these survival figures do not represent the natural course of the disease: they represent the course of the disease following treatment. Because of the success of tamoxifen in patients with ERα-positive tumors, it is unsurprising that AREG correlates with survival — it appears to be a marker for ERα-positive breast tumors, which have a relatively good prognosis because of the use of this drug (Figure 6D).
Kaplan-Meier survival analysis of 295 human breast tumors stratified by marker expression level. High levels of (A) TACE and (B) TGFA predict poor survival. High levels of (C) AREG or (D) _ER_α are correlated with good outcome (AREG and _ER_α are related; see Discussion). P values represent the log-rank comparison between the upper and lower quartiles of marker expression evaluated at 5 and 10 years after surgery.
These data shed light at a molecular level on the steps by which tumor cells may become independent of extrinsic proliferative signals and suggest that targeting ADAM family members may be a useful strategy for treating EGFR-dependent tumors of the breast and other tissues.