PPARγ Ligand–induced Annexin A1 Expression Determines Chemotherapy Response via Deubiquitination of Death Domain Kinase RIP in Triple-negative Breast Cancers Free

In the past two decades, the risk of breast cancer has been increasing worldwide, as well in Asia (Japan, Korea and Taiwan; refs. 1–3). The rate of breast cancer incidence in Singapore has also become one of the highest in the world with 5.7% per year increase in premenopausal and 3.9% per year of postmenopausal women (4).

Breast cancers can be classified into several classes, with each representing unique molecular or genetic characteristics. Triple-negative breast cancer (TNBC), which is estrogen receptor (ER), progesterone receptor (PR), and C-erb B2 receptor (ERBB2R)-negative (5), is known to be heterogeneous in nature of disease. Anthracycline–taxane regimens remain the current standard for TNBC patients who tend to have a higher risk of relapse and worse overall survival rates with mixed outcomes (6, 7). However, the natural heterogeneity of TNBC has made it difficult to evaluate the true clinical value of the tested drugs, making the cohorts' selection based on specific biomarkers associated with distinct TNBC subgroups a prerequisite (8, 9). A “Genome-first Approach” concept has been introduced where patients will be prestratified and assigned to clinical trials designed to address the therapeutic hypotheses, based upon analysis of individual tumor profiles (8, 9).

PPARγ belongs to the nuclear hormone receptor superfamily of ligand-activated transcription factors. PPARγ ligands, 15-deoxy-prostaglandin J₂ (15d-PGJ₂) and the synthetic thiazolidinedione (TZD) family, are currently available in the market as therapeutic agents targeting type II diabetes mellitus (10). The role of PPARγ ligands as potential antitumor agents has been reported by many, including breast cancer (11–13). However, results from clinical trials have not been consistent among the various breast cancer subgroups, making novel biomarkers for patient selections and/or drug response for PPARγ ligand trials, a necessity.

Annexin-A1 (ANXA1), a 37-kDa calcium-dependent phospholipid binding protein, has been implicated in various biological processes including tumorigenesis and anti-inflammatory activity of glucocorticoids (14, 15). ANXA1 expression has been suggested to play roles in the progression and development of many human cancers including breast cancer (16, 17).

NFκB complex is sequestered by inhibitor of kappa B (IκB) in the cytoplasm of normal cells. Breast cancer cells with invasive and metastatic characteristics, express constitutively activated NFκB (18, 19). We have reported baseline ANXA1 expressed in the MDA-MB-231 breast cancer cells interacts with NFκB essential modulator (NEMO, also known as IKKγ) and receptor interacting Protein 1 (RIP1), driving activation of NFκB in cancer cells (20).

As our previous observations have shown that baseline ANXA1 interacts with NEMO and RIP1 to drive activation of NFκB signaling in TNBC lines (20) and having identified a putative PPARγ binding consensus within the promoter of ANXA1, we asked whether activation of PPARγ would have an effect on this multiprotein complex. Herein, we show, using a series of cell-based assays, antitumor activity of PPARγ ligands is regulated via disruption of the ANXA1–NEMO–RIP1^ubiqutinated complex “switching to” an increased ANXA1-RIP1^{deubiqutinated} interaction, activating a caspase-8–dependent death pathway.

MDA-MB-231, BT549, MCF-7, and T47D breast cancer cell lines were purchased from ATCC within one year of the study in 2016. Authentication was based on growth rate, morphology, and absence of mycoplasma. MCF-7 EV and MCF-7 V5 cells were generated in our laboratory via stable transfection, followed by antibiotic selection of the stably transfected clones. All the cell lines were maintained in RPMI1640 medium containing 10% FBS, 2 mmol/L l-glutamine, and 0.05 mg/mL gentamicin sulphate (Hyclone) in a 37°C incubator with 5% CO₂.

A stock of 15d-PGJ2 (Alexis) in solution form, with solubility of 1 mg/mL, was prepared and further diluted with treatment medium to obtain final concentrations of 5 μmol/L and 10 μmol/L. A stock concentration of 80 mmol/L of rosiglitazone (Cayman Chemical) was prepared in DMSO and further diluted with treatment medium to obtain final treatment concentrations of 50 μmol/L and 80 μmol/L. A stock concentration of 20 mmol/L GW9662 (Cayman Chemical) was further diluted using growth medium to a treatment concentration of 10 μmol/L.

Procedures for these assays are conducted as described previously (21, 22). Detailed procedures are described in Supplementary Methods.

For the measurement of percentage of cells that undergo apoptosis, an APC-conjugated Annexin-V/propidium iodide (PI) Staining Assay kit was used (Life Technologies). Cell pellets were resuspended in 100 μL of 1× binding buffer containing APC-conjugated anti-Annexin V antibody. The samples were incubated at 4°C for minimum of 30 minutes in dark conditions. At the end of antibody incubation, each sample was diluted with 400 μL cold 1× PBS. Cells were forced through the filter caps of polystyrene tubes and then analyzed using a flow cytometry (BD FACSCalibur, BD Biosciences). Two microliters of 1× PI dye were added to each sample and gently shaken for 15 seconds before the tube was placed onto the flow cytometer.

Details for the in vivo mice xenograft work is described in ref. 23.

Use of clinical material from Canada was approved by the Institutional Review Board, Centre Hospitalier de l'Université de Montréal (CHUM) in 2005 with a renewal in 2010. Archived formalin-fixed paraffin-embedded (FFPE) tumor tissues from 164 breast cancer patients were selected from the archives of the Department of pathology, CHUM, University of Montreal at IRIC. Normal breast tissue samples were collected from healthy women undergoing plastic surgery. IHC was performed as described previously (22) and details outlined in Supplementary Methods.

These analysis were performed as described in our previous study (22). Details of these analysis are provided in Supplementary Methods.

An epithelial–mesenchymal transition (EMT) scoring method was developed using ovarian carcinoma cell lines expression profiling as described previously by our group (21, 24). A higher or lower EMT score indicates that the cell line exhibits a more mesenchymal or epithelial phenotype, respectively.

Statistical analyses for all in vitro work were performed using paired Student t test. A P value of less than 0.05 was considered significant. Statistical significance evaluation for microarray analysis were performed using Mann–Whitney test and Spearman correlation test, which were computed using Matlab R2012a (MathWorks). Dot plots were done using GraphPad Prism version 5.0 (GraphPad Software). Statistical analyses for IHC work were carried out using different packages of R language (http://www.R-project.org/). The distribution of ANXA1 among different histologic grades and molecular subtypes is depicted using bar charts. Nonparametric tests are used due to the nature of ordinal and categorical data. The overall relationship between ANXA1 scores and other characteristics (i.e., histologic grades, and molecular subtypes) was evaluated by χ² test. Correlation analysis for IHC expression levels was carried out using the Spearman ρ correlation coefficient. Statistical significance was considered, with a P value < 0.05.

Twenty-six cohorts containing 3,992 breast cancer tumors, 22 normal breast tissues collected and compiled from the public database Gene Omnibus (GEO) and ArrayExpress were analyzed for mRNA expression levels of ANXA1. ANXA1 expression is highest in Normal-like and claudin-low TNBC subtypes, followed by ERBB2⁺, luminal-A, and luminal-B subtype, while TNBC has significantly higher basal ANXA1 expression compared with luminal-A, luminal-B, and ERBB2⁺ breast cancer subtypes (Fig. 1A). ANXA1-high patients had a markedly reduced percentage of patient survival than ANXA1-low patients (log-rank test, P = 0.015; HR = 0.4815; Fig. 1B). In addition, the epithelial–mesenchymal transition (EMT) scoring system shows ANXA1 was strong and positively correlated with EMT score (Spearman correlation test, ρ = 0.6367, P < 1e−128; Supplementary Fig. S1A). To validate our ANXA1 mRNA analysis from Fig. 1A, we obtained patients' specimens from Centre hospitalier de l'Université de Montréal (CHUM) for IHC analysis. Normal breast tissue samples were collected from healthy women who underwent plastic surgery. Clinicopathologic features of these patients are shown in Table 1. IHC analyses showed higher levels of ANXA1 expression were associated with the more aggressive tumor subtypes such as TNBC and ERBB2⁺compared with the luminal subtypes (Fig. 1C). Higher ANXA1 expression was also detected in grade III than the lower grade breast tumors (Fig. 1D) with positive correlation observed between ANXA1 expression and cellular proliferation marker, Ki-67. Notably, the TNBC subtype expressed both high levels of ANXA1 and Ki-67 (Fig. 1E). Interestingly, clinical data analyzed using Spearman correlation test results showed ANXA1 gene expression is indeed positively correlated with PPARγ gene expression in patient samples (ρ = 0.29, P < 1e−128; Fig. 1F). In agreement with our mRNA correlation in Fig. 1F, ANXA1 protein levels also positively correlate to PPARγ protein expression in patient breast tumors by IHC scoring (Fig. 1G) as well as in breast cancer cell lines by relative mRNA quantification (Supplementary Fig. S1B). In vitro, baseline gene and protein expression levels of ANXA1 and PPARγ were performed in luminal subtype lines (MCF-7and T47D) and claudin-low TNBC subtype lines (MDA-MB-231 and BT549). Baseline ANXA1 mRNA and protein levels were significantly higher by 2.5 times in TNBC than luminal breast cancer cell lines (Fig. 1H). Importantly, we observed a similar corresponding pattern of baseline levels of PPARγ mRNA and protein levels were higher in the two TNBC cell lines as well (Fig. 1I). Concomittant to the differences in baseline PPARγ expression levels (Fig. 1I), lower PPARγ activity is observed in normal MCF10A and luminal type T47D compared with that in TNBC type MDA-MB-231 cells (Fig. 1J).

Crystal violet staining assay was performed on the breast cancer cells using two PPARγ ligands, 15d-PGJ₂, or rosiglitazone. Dose-dependent decrease in cell viability was observed in MDA-MB-231 and BT549 cells after 48-hour treatments (Fig. 2A), while marginal decrease in viability of MCF-7 and T47D cells was observed (Fig. 2B). These findings were further validated by microscopic examinations in TNBC (Supplementary Fig. S1C and S1D) versus luminal breast cancer (Supplementary Fig. S1E and S1F). Colony-forming abilities of MDA-MB-231 and BT549 cells were also greatly affected by PPARγ ligand treatment (Fig. 2C). In contrast, MCF-7 and T47D cells were relatively resistant to PPARγ ligand treatment (Fig. 2D). Interestingly, in both MDA-MB-231 (Fig. 2E) and BT549 (Supplementary Fig. S2A and S2B), the significant decrease in cell viability after exposure to PPARγ ligands was abrogated when ANXA1 depleted in these cells. MDA-MB-231 cells transfected with scrambled siRNA showed 15.3% and 21.8% increase in total apoptosis after exposure of 15d-PGJ₂ (10 μmol/L) and rosiglitazone (80 μmol/L), respectively (Fig. 2F and G). However, less than 5% increase in apoptosis was observed in cells with depleted ANXA1 after PPARγ ligand treatments in MDA-MB-231 cells (Fig. 2H). Conversely, in the MCF-7-V5 cells (an MCF-7 cell line stably overexpressing ANXA1 ectopically), treated with the same concentrations of 15d-PGJ₂ or rosiglitazone resulted in up to 40% reduction in cell viability compared with MCF-7-EV cells (stably transfected with empty vector; Fig. 2I), with a 17.7% and 33.8% increase in total apoptosis when treated with 15d-PGJ₂ or rosiglitazone, respectively. However, only 5% increase in total apoptosis was observed in MCF-7-EV cells after PPARγ ligand treatments (Fig. 2J).

As we observed manipulation of ANXA1 expression influences the effect of PPARγ ligands, we next investigated whether PPARγ ligand activity could influence ANXA1 expression. ANXA1 promoter activity increased 2.5-fold and 1.5-fold after treatment with 15d-PGJ₂ and rosiglitazone, respectively (Fig. 3A). A time-dependent increase in ANXA1 mRNA (Fig. 3B and C) and corresponding protein expression (Fig. 3D and E) were observed in MDA-MB-231 cells after treatment with 10 μmol/L of 15d-PGJ₂ or 80 μmol/L of rosiglitazone. These in vitro observations were further confirmed in vivo using a MDA-MB-231 xenograft mouse model previously reported by our group. Mice injected with 15d-PGJ₂ showed no significant increase in tumor volume throughout the experimental period, while mice treated with vehicle control exhibited an increase in tumor volume with time (23). Importantly, in agreement with our cell line data, tissue collected from these 15d-PGJ₂–treated mice exhibited more positive staining for ANXA1 than the PBS-treated mice (Fig. 3F). On the other hand, the luminal MCF-7 cells that are insensitive to PPARγ ligand (Fig. 2A and B), did not show an increase in ANXA1 gene expression (Supplementary Fig. S2C).

Next, we utilized GW9662, an irreversible and selective PPARγ antagonist (25) to ascertain whether the induction of ANXA1 expression was dependent on PPARγ activity. ANXA1 promoter activity induced by PPARγ activation was decreased in MDA-MB-231 cells preincubated with GW9662 (Fig. 4A and B). Preincubation with GW9662 abrogated the induction of ANXA1 mRNA (Fig. 4C and D) and protein expression (Fig. 4E and F) after treatment with PPARγ ligands. Consistent with these observations, preincubation with GW9662 also exerted its inhibitory effect on PPARγ-induced upregulation of ANXA1 gene expression in BT549 cells (Supplementary Fig. S3A–S3C). In addition, inhibition of PPARγ activity by GW9662 effectively blocked PPARγ-induced apoptosis in both MDA-MB-231 and BT549 cells (Supplementary Fig. S3D and S3E).

Alternatively, the dependence of ANXA1 gene induction on PPARγ activity was examined by using a dominant-negative (DN) mutant of PPARγ (Supplementary Fig. S3F). Overexpression of DN-PPARγ significantly inhibited the induction of peroxisome proliferator response element (PPRE) luciferase activity by PPARγ ligands (Supplementary Fig. S3G). Importantly, DN-PPARγ effectively prevented the drug-induced increase in ANXA1 promoter activity (Fig. 4G and H). In addition, DN-PPARγ also significantly suppressed the PPARγ-induced ANXA1 mRNA (Fig. 4I and J) and protein expression (Fig. 4K and L).

We next assessed whether activation of PPARγ influences protein–protein interactions within ANXA1–NEMO–RIP1 complex, as we previously reported (20). Treatment of MDA-MB-231 cells with 10 μmol/L of 15d-PGJ₂ leads to induction of both ANXA1 and RIP1 protein expression, but reduction in NEMO protein expression (Supplementary Fig. S4A). A 4-fold increase in RIP1 mRNA and a 30% decrease in NEMO mRNA was observed after treatment with 15d-PGJ₂ (Supplementary Fig. S4B and S4C). Moreover, these changes in RIP1 and NEMO mRNA were found to be dependent on PPARγ activity, as preincubation with GW9662 attenuated these changes (Supplementary Fig. S4B and S4C). Activation of PPARγ also resulted in a receptor-dependent reduction of NFκB activation (Supplementary Fig. S4D and S4E).

To gain deeper insight into the mechanism for our observed results in Supplementary Fig. S4A–S4E, immunoprecipitation (IP) was performed for each of the proteins in the tricomplex upon PPARγ activation. MDA-MB-231 cells treated with 10 μmol/L of 15d-PGJ₂, exhibited an increase in ANXA1–RIP1 interaction (Fig. 5A), together with a decrease in NEMO–RIP1 interaction (Fig. 5A) and ANXA1–NEMO interaction (Fig. 5A). The role of RIP1 on the cell viability after treatment with PPARγ ligands was also investigated. A 30% to 40% decrease in cell viability was observed in control MDA-MB-231 cells transfected with scrambled siRNA after exposure to 15d-PGJ₂ or rosiglitazone (Fig. 5C). Importantly, this reduction in cell viability was attenuated upon RIP1 depletion. Moreover, silencing RIP1 expression also blocked the increase in apoptotic cells upon treatments with PPARγ ligands (Supplementary Fig. S5A). These results suggest that the loss of RIP1 expression prevented cell death stimulated by PPARγ activation. Interestingly, the loss of RIP1 not only reduced the baseline expression of ANXA1, but also prevented the induction of ANXA1 protein expression to the same level of expression by PPARγ activation. However, NEMO protein expression was not downregulated upon PPARγ activation in RIP1-depleted cells (Fig. 5D). In addition to changes in ANXA1 and NEMO protein expression levels, depletion of RIP1 did not cause significant changes in ANXA1–NEMO protein interactions in the presence of 10 μmol/L of 15d-PGJ₂ (Supplementary Fig. S5B–S5D).

The ubiquitination status of RIP1 determines its role as either a prosurvival or a prodeath agent (26). Interestingly, reduction in global polyubiquitination was observed in both MDA-MB-231 and BT549 cells treated with 10 μmol/L of 15d-PGJ₂ in a time-dependent manner (Fig. 5E; Supplementary Fig. S5E). More importantly, treatment with the PPARγ ligands also led to a time-dependent reduction in RIP1-specific ubiquitination in TNBC cell lines (Fig. 5F). To investigate the role of ANXA1 in RIP1 ubiquitination, endogenous ANXA1 was silenced in MDA-MB-231, followed by treatment with 10 μmol/L of 15d-PGJ₂. Depletion of ANXA1 prevented a decrease in global protein polyubiquitination levels in MDA-MB-231 cells (Fig. 5G).

RIP1 needs to be in a deubiquitinated state to enable its interaction with caspase-8 to promote apoptosis (27). A moderate decrease in procaspase-8, together with a marked increase in caspase-8 cleavage after treatment with 15d-PGJ₂ was observed in MDA-MB-231 cells. However, these observations were not seen in ANXA1-depleted cells (Fig. 5H). Moreover, knocking down expression of ANXA1 in MDA-MB-231 cells also blocked the ligand-induced increase in caspase-3 cleavage as well as inhibiting the downregulation of prosurvival proteins such as Bcl-2 and survivin (Fig. 5I).

cIAP1 has been defined as the direct E3 ligase that promotes RIP1 ubiquitination in cancer cells (28, 29). On the basis of our findings in Fig. 5G, we were motivated to explore whether treatment with PPARγ ligands had an effect on expression of cIAP1. Exposure of MDA-MB-231 cells to 10 μmol/L of 15d-PGJ₂ not only induced expression of ANXA1 and RIP1 but also downregulated cIAP1 protein expression in a time-dependent manner while preincubation with GW9662 attenuated these effects (Fig. 6A). As activation of PPARγ is known to induce proteasomal degradation of various proteins (30, 31), we proceeded to demonstrate that preincubation of MDA-MB-231 cells with MG132 a proteasomal inhibitor, followed by exposure to 10 μmol/L of 15d-PGJ₂ for 24 hours, prevented the degradation of cIAP1 protein (Fig. 6B). On the other hand, we observed 15d-PGJ₂ reduced the degradation of cIAP1 with a much delayed and weaker when the endogenous ANXA1 was silenced (Fig. 6C). Importantly, we also observed a marked increase in ubiquitinated cIAP in MDA-MB-231 cells after exposure to 15d-PGJ₂, which was not observed when ANXA1 was silenced (Fig. 6D). Exposure to 15d-PGJ₂ also reduced NEMO-specific ubiquitin levels, which was also not observed when ANXA1 was silenced (Supplementary Fig. S5F). Indeed, we observed an interaction between RIP1 and cIAP1, which was followed by a gradual time-dependent decrease in cIAP1–RIP1 interaction upon treatment with 10 μmol/L of 15d-PGJ₂. However, when ANXA1 was depleted, the time-dependent dissociation of cIAP1 from RIP1 after exposure to 15d-PGJ₂ was not observed (Fig. 6E). Importantly, the RIP1-specific ubiquitin level remained unchanged with 15d-PGJ₂ treatment in cells silenced with ANXA1 (Fig. 6F). Finally, PPARγ ligand treatment induced an increase in RIP1-procaspase-8 association in a time-dependent manner (Fig. 6G).

To date, several in vitro studies have shown that PPARγ activation causes cell death and induces apoptosis in various cancer cell lines, including breast cancer. Nevertheless, applications of PPARγ activators as anticancer options in clinical trials have not yet shown promising results. Particularly, in the context of breast cancer, inconsistent outcomes of PPARγ activators have been reported across different subgroups of breast cancer tissues or cells (32–34).In parallel, the antiproliferative role of ANXA1 has been reported (35, 36); however, there has not been a suggested a role for ANXA1 as a biomarker in any cancer types (17, 37). Our study aimed to address the question of whether the presence of ANXA1 could influence the effectiveness of PPARγ activators in breast cancer.

Preliminary analysis across 3,992 breast cancer tumors showed ANXA1 mRNA expression was higher in basal and claudin-low breast cancer (grouped as TNBC) subtypes correlating to poor patient survival. In addition, higher level of ANXA1expression was positively correlated with mesenchymal phenotype. In agreement with our mRNA analysis, IHC staining of various breast cancer tissues showed that TNBC subtype exhibited high level of ANXA1 expression that was correlated with very high histologic grading scores. These findings strongly support the observation of in vitro screening of breast cancer cells, that the highly invasive TNBC cells express more baseline ANXA1 than the poorly invasive ER⁺ breast cancer cells. Baseline PPARγ expression and therefore activity also showed a similar trend across the panel. To our understanding, this is the first report on the positive correlation between ANXA1 and PPARγ in breast cancer.

Despite the continuous reporting of ANXA1 expression in many human cancers, the results have been conflicting across different studies and in different types of cancers; thus, the exact role of ANXA1 in cancer development and progression remains undefined. Our IHC staining for ANXA1 expression in tumors of various breast cancer subtypes shows that ANXA1 expression is not evenly distributed in all breast tumor tissues. High ANXA1 expression is detected only in TNBC tumors and histologic grade III tumors. However, most of the earlier studies on ANXA1 expression in breast cancer did not include tumors of all the different molecular or histologic subtypes or histologic grade of breast cancer (16, 17, 38). This may be the reason for the inconsistency in reporting ANXA1 expression in breast cancers conducted in the earlier studies.

IHC staining results demonstrated that TNBC tumors, which express a high level of ANXA1, also demonstrate a higher level of Ki-67 expression. Additional supporting evidence on high ANXA1 expression as an indicator of cancer metastasis and invasion is found in more recent literature (37, 39, 40). Particularly, de Graauw and colleagues reported that ANXA1 expression is functionally related to the progression of breast cancer, and is associated with the highly invasive BLBC subtype (41). Linking back to our in vitro result on the basal gene expression of ANXA1 among breast cancer cell lines, these observations suggest that the ANXA1 expression level is potentially associated with the degree of malignancy of breast cancer tumors.

Importantly, depletion of endogenous ANXA1 (or rather not allowing for PPARγ-induced increased expression) conferred resistance in two TNBC cell lines with exposure to PPARγ agonists. On the other hand, stable clones of MCF-7 cells with ectopic overexpression of ANXA1 showed enhanced responsiveness to PPARγ drugs as compared with the unresponsive parental empty vector control clones. These results proved our hypothesis that high ANXA1 expression in breast cancer cells acts as an indicator of PPARγ agonist response. Interestingly, activation of PPARγ failed to induce baseline ANXA1 expression in these PPARγ-unresponsive cells. It has been reported that ERα could bind to PPRE elements and represses PPARγ transactivation in MCF-7 cells (42). Our group has previously reported ANXA1 as a putative ERα target (43). It is therefore probable that an active ERα may have negated PPARγ activity at the promoter of ANXA1 in MCF-7 cells.

PPARγ activation has been reported to be involved in suppression of tumor proliferation and growth in various types of cancer tissues (13, 23, 32, 44). The functional role of PPARγ was firstly elucidated by us when MDA-MB-231 cells were injected subcutaneously into mice mammary fat pad, followed by treatment with 15d-PGJ₂. Mice exposed to 15d-PGJ₂ showed inhibited growth in tumor volume as compared with mice treated with control vehicle (23). More interestingly, ANXA1 expression in tumor tissues obtained from 15d-PGJ₂–treated mice was significantly higher than that obtained from the untreated animals. Using our published PPRE search engine (45), we located a putative PPRE in the promoter of ANXA1. To verify our hypothesis that perhaps ANXA1 is a novel target of PPARγ activity, we performed in vitro evaluation on changes in ANXA1 gene expression levels upon PPARγ ligand exposure in two TNBC models. We observed induction of ANXA1 expression by PPARγ ligands, from promoter activity to protein expression, suggesting that ANXA1 expression is indeed regulated by PPARγ.

The irreversible PPARγ inhibitor, GW9662, and DN-PPARγ, which has no transcriptional activity, was able to block the induction of ANXA1 expression and attenuated the apoptotic effects of PPARγ ligands. These findings further confirm that the upregulation of ANXA1 is mediated by the activity of ligand-activated PPARγ, and the apoptotic effects observed upon incubation with PPARγ ligands are also due to the intrinsic activity of the receptor.

RIP1 exists as a double-edged sword in controlling the cellular fate. Together with NEMO, they cooperate to promote cell survival in response to excessive DNA damage (46). Our earlier study reported that basal ANXA1 in MDA-MB-231 cells interacts with NEMO and RIP1, and contributes to the constitutively active NFκB in these cells (20). To take these findings further, we propose a new mechanism by which activated PPARγ promotes apoptosis in breast cancer cells in ANXA1-dependent manner. MDA-MB-231 cells treated with 15d-PGJ₂ exhibit increased ANXA1 and RIP1 expression after treatment, while NEMO expression was downregulated. We also found increases in levels of cleaved caspase-8 and caspase-3 after treatment, suggesting the drug treatment leads to apoptosis via activation of caspase-8. These observations indicate that 15d-PGJ₂ treatments have shifted the balance of RIP1 activity toward proapoptosis. Interestingly, 15d-PGJ₂ exposure for 24 hours led to stronger interaction between RIP1 and ANXA1 but important dissociation of NEMO from both RIP1 and ANXA1.

A previous study suggested the proapoptotic activity of RIP1 requires its dissociation from NEMO (47). This is in line with our expectation of decreased RIP1–NEMO interactions upon drug treatment, due to the fact that PPARγ drugs are able to induce apoptosis. One possible explanation to the inhibitory effect of NEMO on RIP1-dependent cell death is that ubiquitination of RIP1 at lysine 377 recruits ubiquitin-binding proteins, such as NEMO, and this binding sterically hinders RIP1 from associating with caspase-8 and other downstream death mediators (29). On the other hand, ubiquitination of RIP1 may also prevent RIP1 from oligomerization. The oligomerization of the death domain of RIP1 has been implicated to contribute to its cell death–triggering ability.

The increased RIP1–ANXA1 association attracted our attention and led us to investigate the role of ANXA1 on RIP1-mediated cell death. Depletion of ANXA1 reduced apoptosis upon PPARγ ligand treatment and leads to a reduction in caspase-8 and caspase-3 activation and inhibition in the downregulation of prosurvival molecules such as Bcl-2 and survivin. More importantly, the loss of ANXA1 caused a reduction in RIP1 protein and mRNA expression in MDA-MB-231 cells. This indicates the importance of ANXA1 in maintaining both mRNA and protein expression of RIP1 and its role in the activation of caspase-8 and the initiation of apoptosis. Furthermore, activation of PPARγ failed in suppressing NEMO transcription when ANXA1 was absent in MDA-MB-231 cells. This provides possible alternative mechanism for the higher resistance to PPARγ ligands found in TNBC cells with low ANXA1 expression.

A time-kinetic study with treatment of 15d-PGJ₂ showed deubiquitination of RIP1 in MDA-MB-231 cells. Without hindrance of the polyubiquitin chain, RIP1 is free to interact with the pro-caspase-8 and causes the subsequent initiation of apoptotic events. This provides a possible mechanism where activation of PPARγ induces apoptosis in MDA-MB-231 cells via the action of deubiquitinated RIP1. In agreement with our findings, activation of caspase-8 is known to be dependent on deubiquitination of RIP1 (48). The essential role of RIP1 in PPARγ-induced apoptosis was granted with the observation that depleting RIP1 was able to block the apoptosis rate in treated MDA-MB-231 cells. Furthermore, the absence of RIP1 restrains the drug-induced dissociation of NEMO from ANXA1. Interestingly, the expression of ANXA1 exerts influence on the ubiquitination status of RIP1. We observed that depletion of ANXA1 has an inhibitory effect on the reduction of total protein ubiquitin expression after the treatment, while the RIP1-specific deubiquitination is also hindered in MDA-MB-231 cells with depleted ANXA1. Linking back to the earlier observations that knocking down ANXA1 promotes resistance to PPARγ drugs in TNBC cells, this confirms that ANXA1 is an important intermediate molecule that ensures the prodeath activity of ligand-activated PPARγ.

We have demonstrated a highly possible mechanism by which PPARγ activation lead to caspase-8–initiated apoptosis in TNBC cells, through a process of ANXA1-dependent deubiquitination of RIP1. Several earlier studies reported that cIAP1 functioning as an E3 ligase is involved in catalyzing the polyubiquitination of RIP1, while loss of cIAP1/2 blocks ubiquitination of RIP1 in response to TNF (49, 50). More recently, cIAP1 has been identified as the direct E3 ligase that conjugates ubiquitin chains to RIP1 (28). Treatment with 15d-PGJ₂ showed a time-dependent decrease in cIAP1 protein expression level and preincubation with GW9662 was able to block the drug-induced reduction in cIAP1 gene expression. It suggests that cIAP1 expression could be PPARγ regulated. However, the downregulation of cIAP1 in MDA-MB-231 cells may not be due to transcriptional regulation by PPARγ. The activity of PPARγ to cause degradation of cIAP1 was uncovered; we also demonstrated that inhibiting proteasome activity by MG132 was able to attenuate PPARγ-induced degradation of cIAP1 protein expression. Interestingly, silencing of ANXA1 in MDA-MB-231 cells slows down this degradation process on cIAP1 protein upon exposure to the PPARγ ligands. Moreover, the PPARγ ligands caused increased autoubiquitination of cIAP1 and turn its degradation in ANXA1-dependent manner. These findings suggest that drug-induced ANXA1 expression is crucial for the promotion of proteasome-dependent degradation of cIAP1 by PPARγ ligands.

In summary, our study proposes a novel schematic model to describe the mechanism of how PPARγ activation leads to apoptosis in TNBC through the actions of ANXA1 and RIP1 (Fig. 7). Under basal conditions, NEMO binds to polyubiquitinated RIP1 and ANXA1, inducing constitutive activation of NFκB to promote cancer cell survival (20). Agonists that activate PPARγ induce expression of RIP1 and ANXA1. Meanwhile, deubiquitinated RIP1 dissociates from NEMO, while it remains associated with ANXA1. RIP1 then binds procaspase-8 and leads to initiation of caspase-8–dependent apoptosis. ANXA1 expression required for the stabilization of RIP1 protein to ensure function as a mediator of the proapoptotic action of PPARγ ligands. Taken together, we show that drug-induced ANXA1 expression mediates actions of PPARγ ligands in promoting cell apoptosis and that ANXA1 could serve as a clinically useful surrogate biomarker for such therapies in the future trials.

No potential conflicts of interest were disclosed.

Conception and design: L. Chen, P.F. Koh, L. Lim, A.P. Kumar

Development of methodology: L. Chen, Y. Yuan, S. Arora, P.F. Koh, H.-M. Shen, G. Sethi, L. Lim

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Chen, Y. Yuan, P.F. Koh, E. Yousef, M.K. Shanmugam, S.-W. Shin, J.-I. Park, L. Gaboury, A.P. Kumar

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Chen, S. Kar, P.F. Koh, E. Yousef, T.Z. Tan, H. Yang, L. Gaboury, P.E. Lobie, G. Sethi, L. Lim, A.P. Kumar

Writing, review, and/or revision of the manuscript: L. Chen, Y. Yuan, S. Kar, J. Kim, P.F. Koh, F. Arfuso, J.-I. Park, L. Gaboury, A.P. Kumar

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Chen, M. Kanchi, E. Yousef, B.C. Goh, P.E. Lobie, A.P. Kumar

Study supervision: H.-M. Shen, P.E. Lobie, L. Lim, A.P. Kumar

Other (fine tuning and revision of the manuscript): R. Samy

This work was supported by grants from the National Medical Research Council of Singapore and by the NCIS Yong Siew Yoon Research Grant through donations from the Yong Loo Lin Trust (to A.P. Kumar). A.P. Kumar and P.E. Lobie were supported by grants from the Singapore NMRC Clinician Scientist IRG and the National Research Foundation Singapore and the Singapore Ministry of Education under its Research Centers of Excellence Initiative to Cancer Science Institute of Singapore, National University of Singapore. This work was also supported by the NUHS Basic Seed Funding and Ministry of Education Tier 1 grant (to G. Sethi). J.I. Park was supported by the Korea Science and Engineering Foundation (KOSEF) grant funded by the Korea government (MEST). L.H.K. Lim was supported by a grant from Singapore Ministry of Education Tier 2. This work is supported in part by National Medical Research Council Singapore (NMRC) grants (to H.M. Shen).

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.