XIAP Regulation by MNK Links MAPK and NFκB Signaling to Determine an Aggressive Breast Cancer Phenotype

The primary cause of breast cancer morbidity and mortality is metastasis to surrounding tissues and distant organs, a process dependent on hyperproliferation and hypermotility of cells derived from the primary tumor. Among all breast cancer types, inflammatory breast cancer (IBC) is a highly aggressive subtype characterized clinically by extremely motile tumor cell clusters that exhibit localized dermal invasion and frequent lymph node involvement (1–3). Despite an aggressive multimodal treatment regimen, tumor recurrence and metastatic progression are unmet challenges in IBC patients (4).

Comparative gene expression studies from preclinical models and pretreatment patient samples have attempted to define molecular profiles specific to IBC. They reveal highly activated MAPK and NFκB transcriptional profiles associated with increased proliferation in IBC compared with other locally advanced breast cancers (5–8); however, the mechanism for the linkage between these two pathways in IBC tumors has not been described. We sought to determine how EGFR-mediated MAPK activation and NFκB hyperactivity were coordinated to enhance cancer cell survival and proliferation with the goal of elucidating targets for therapeutic intervention. Our previous studies have shown that IBC tumor cells escape from various cell death and oxidative stress stimuli, including EGFR inhibition, through upregulation of the X-linked inhibitor of apoptosis protein (XIAP; refs. 9–13). XIAP, through its multiple domains, not only directly inhibits the initiation and execution phases of the caspase cascade during programmed cell death, but also regulates, in a caspase-independent manner, a range of cellular activities that enhance survival signaling, including NFκB activity (14, 15). Recent studies reveal that translational regulation of select survival proteins is regulated by MAPK signaling, protecting cancer cells during cellular stress (16). However, the understanding of mechanisms linking MAPK signaling and survival signaling remains limited.

In this study, we demonstrate that elevated XIAP expression in patient IBC tumors is associated with aggressive biology and poor clinical outcome. Using IBC cell lines derived from previously untreated primary tumors combined with modulation of XIAP expression, we found that XIAP drives activation of NFκB and its target genes, leading to enhanced tumor growth. Furthermore, we discover that MAPK interacting kinase (MNK) signaling, downstream of EGFR/HER2 activation, promotes XIAP expression and NFκB activity. Collectively, our findings indicate a role for XIAP as a central regulatory node connecting MAPK and NFκB signals, which governs IBC tumor-specific gene signatures, survival, and tumorigenesis.

Gene expression datasets previously published were used to generate a combined total of 1032 breast cancer patients [GEO datasets GSE6532, GSE9195, GSE16391, GSE16446, GSE17907, GSE20685, GSE20711, and GSE21653]. A total of 1018 patients had nonzero event-free survival time and were available for analysis for the expression of two probesets (206536_s_at and 206537_at, Affymetrix), which targeted the XIAP ORF. Patients were grouped into high or low expression of XIAP using the top quartile versus remaining patients by probe set expression values and compared using Kaplan–Meier plots with 95% confidence intervals of event-free survival (earliest event provided within each dataset) by log-rank test. For correlation of XIAP expression with lymph node involvement, two available GEO datasets, GSE6532 and GSE9195, were identified that included lymph node status for a total of 164 patients with IBC.

The TMA sections used in this study, with prior patient consent and approval from the Institutional Review Boards from each center, and their clinical characteristics are available in refs. 17–19.

Four-micron–thick paraffin sections were deparaffinized, rehydrated, and antigen retrieval performed using EDTA buffer at 95°C for 30 minutes. Slides were incubated in a 1:60 dilution of mouse anti-human XIAP (BD Biosciences) overnight at 4°C, washed, and incubated in anti-mouse secondary (Dako anti-mouse Envision kit) for 30 minutes at room temperature. Imaging was performed on a Zeiss Axio Observer A1 microscope and images analyzed with MetaMorph. Scoring of slides was carried out by a board-certified surgical pathologist in a blinded manner. Staining intensity was graded on a qualitative scale [no staining (negative), very focal or very weak staining (borderline), and positive]. For the purpose of statistical analysis, the data were dichotomized as negative or positive (including borderline).

SUM149 and SUM190 cells were obtained from Asterand, Inc. and were cultured as previously described (9). The rSUM149 cell variant is derived from SUM149 and cultured as described previously (10, 20). SUM149 cells stably expressing wtXIAP, shXIAP, and shXIAP+wtXIAP were generated using a lentiviral expression system (kindly provided by Dr. Colin Duckett, University of Michigan, Ann Arbor, MI) and previously reported (13). HeLa cells were grown in DMEM + 10% FBS. For overexpression of MNK1 T344D mutant, cells were transfected with 0.5 μg of pcDNA3.1 (empty vector) or pcDNA3.1 MNK1 T344D (previously described in ref. 21), and harvested 48 hours later for immunoblot analysis. Characterization and authentication of the purchased cell lines were done at Asterand. Additionally, short tandem repeat polymorphism analysis was performed at regular intervals on all cell lines at the Duke Sequencing facility. Cells were cultured at 37°C and 5% CO₂.

All animal experiments were performed in accordance with protocols approved by the Duke University Institutional Animal Care and Use Committee. IBC cells (5 × 10⁶) were suspended in 50-μL PBS and 50-μL Matrigel and injected orthotopically into the fourth mammary fat pad of female SCID mice. Mice were monitored twice weekly and tumor volume measured using the formula V = (L × W²)/2, where L is the longer measurement. Tumor doubling time was found by fitting a nonlinear regression model to the tumor volumes. Mice were euthanized when tumors reached a humane endpoint of ∼1,500 mm³, at the first sign of morbidity, or at end of study. Tumors were removed and tissue harvested for RNA and Western immunoblot analysis.

A dorsal skin-fold window chamber was implanted in Nu/Nu mice as described previously (22). Briefly, the dorsal skin was tented and sutured to a c-frame to hold it in position. Three ∼1 mm diameter holes were cut, through which the window frame could be secured, and a 12-mm diameter full-thickness skin punch was removed from the superior skin fold. The titanium window frame (Small Dorsal Kit, APJ Trading) was sutured in place and a total of 1 × 10⁵ tumor cells were injected into the fascia in the center of the window in a 20-μL volume, using a 30-gauge needle. Sterile saline was used to fill the window, over which a coverslip was placed and affixed by a retaining ring.

Female nude mice (around 10 weeks old) were randomized into two groups (vehicle and CGP57380 group, n = 2 for each group) after the installment of dorsal window chamber and implantation of GFP-tagged SUM149 IBC cells. Tumor-bearing mice were treated with either CGP57380 (25 mg/kg, i.p.) or vehicle for 4 doses on days 0, 2, 4, 6 after the surgery. A stock solution of CGP57380 in DMSO was made up and further diluted into PBS for administration; vehicle solution contained the same percentage of DMSO in PBS as the CGP57380 solution. The experiment was repeated twice.

Mice were anesthetized and mounted to a microscope stage with a custom-made mouse window chamber slide mount. The Zeiss Axio Observer Z1 microscope was used for all imaging with a 5× objective and Apotome (Carl Zeiss AG). GFP fluorescence was excited and acquired with a 488 nm/509 nm excitation/emission filter set, as well as a bright-field transmission image, all recorded by a CMOS camera (C11440, HAMAMATSU photonics K.K). A whole window chamber image was acquired using the tiling function within Zen Pro software. Tumor growth was monitored with fluorescent imaging at designated time points.

Total RNA isolation from adherent cells was completed using the Ambion mirVana miRNA isolation kit (Invitrogen) following manufacturer's instructions. Tissue samples were homogenized in the provided lysis buffer (mirVana kit) and total RNA isolated following instructions.

RNA quality was assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies) and total RNA profiled using the U133A 2.0 Human Gene microarrays at the Duke Institute for Genome Sciences & Policy Microarray facility. Expression data were quantile-normalized and summarized using GCRMA express (23). Probe sets with a fluorescence intensity above log₂(100) in at least two samples were considered informative. Expression levels were compared using generalized linear models on log₂ expression data and probe sets with nominal P values less than 0.05 were considered significant. Differentially expressed genes were translated into pathways using Expression2Kinases (24). Transcription factors and kinases with a combined enrichment score of at least 10 were included for the protein–protein interaction (PPI) network construction. The PPI network was then analyzed for enriched pathways using the Reactome FI plugin from Cytoscape. Pathways with a prior probability of less than 1% were considered relevant. Unsupervised hierarchical cluster analysis of the gene expression data was performed using the Manhattan distance as the dissimilarity metric and the Ward linkage as the dendrogram drawing method. Application of the IBC signature to the expression data of SUM149, wtXIAP, and shXIAP tumors was done as described before (7).

Expression data of wtXIAP (XIAP high) cells was compared with the combined parental and shXIAP cell lines (XIAP low) using default parameters with gene-set level permutations and signal2noise used to rank genes. Gene-set enrichment visualization was performed using Cytoscape 2.8.3 and a P < 0.001, Q-value cutoff 0.006, similarity cutoff of 0.5, and false discovery rate of 0.1. Gene sets examined were from the current molecular signature (MSigDB) versions 4.0.

Cells were seeded and allowed to reach approximately 80% confluence. Cell viability was determined by Trypan blue exclusion assay as described previously (9). Caspase-3/7 activity was determined in cells untreated and treated with 50 ng/mL TRAIL, using the Caspase-Glo Assay (Promega) as per manufacturer's instructions.

Western immunoblots were carried out as described previously (10). Cells were harvested after indicated treatments and times. Tissue samples were homogenized in lysis buffer in a Bullet Blender Storm 24 (Next Advance). Membranes were incubated at 4°C overnight with antibodies against NFκB (P65), p-NFκB (P65), ERK, p-ERK, p-eIF4E, eIF4E, p-p38, p38, survivin, MNK1 (Cell Signaling Technology, all 1:1,000), SOD2 (1:1,000), Bcl-2, (1:1,000), XIAP (1:2,000; BD Biosciences), c-Myc (Sigma-Aldrich), or GAPDH (Santa Cruz Biotechnology, 1:4,000). Densitometric analysis was performed using the NIH ImageJ software (25); for western quantitation measuring effects of MNK modulation on XIAP and NFκB signaling a LI-COR Odyssey FC imager with Image Studio software (LI-COR) was used.

The NRAGE peptide was purchased from NeoBioLab and used as previously described by our lab (13). For all experiments, unpurified NRAGE peptide was added to cells for 24 hours with 6 μmol/L EndoPorter delivery reagent (GeneTools LLC).

Anchorage-independent growth assay was performed as described previously (26). Indicated treatments were applied for 24 hours, after which cells were harvested and counted. Once visible colonies had formed, they were counted under a microscope, and colony counts were normalized to the untreated sample. Images of representative fields were taken with 5× magnification using a Zeiss Axio Observer A1 microscope (Zeiss), Hamamatsu Orca ER digital camera, and MetaMorph software (Molecular Devices).

ALDH enzymatic activity was evaluated using the ALDEFLUOR kit (Stem Cell Technologies) according to the manufacturer's instructions. Cells were incubated with provided ALDH substrate for 35 minutes at 37 °C. The specific ALDH inhibitor diethylaminobenzaldehyde (DEAB) was used as a negative control. Sorting gates were established using 7-AAD for viability DEAB-treated, ALDEFLUOR-stained cells as negative controls.

Total RNA was subjected to reverse transcription using the iScript Reverse Transcription SuperMix Kit (Bio-Rad) and oligo d(T) primers as per manufacturer's instructions. cDNA and SYBR Green were added to a custom PrimePCR plate (Bio-Rad) containing primer pairs for the indicated NFκB target genes and β-actin as a loading control. Further information on primers can be found in Supplementary Table S1. For MNK qRT-PCR studies, qRT-PCR was performed using Applied Biosystems MYC, BIRC5, IL1B, and BCL2 primers and probes using the RNA to CT one step RT-PCR reagent kit (Invitrogen), following the manufacturer's instructions. The PCR was conducted on an iCycler instrument (Bio-Rad) using the following conditions: [95°C × 2 minutes, (95°C × 5 seconds, 60°C × 30 seconds) × 40 cycles] and fold changes calculated by the 2^(-ΔΔC_t) method, except for MNK qRT-PCR studies where primer efficiency correction was deemed necessary and the 2^(-ΔΔCP) method was used (27).

Simulations were performed in MATLAB (Mathworks Inc.), and bifurcation diagrams were drawn using MATCONT (28). The model formulation for the interactions of MNK, XIAP, and NFκB based on this study and previous work is given by:

where N, M, and X denote NFκB, MNK, and XIAP levels, respectively. g_N and g_X are the respective production rates for NFκB and XIAP, and k_N and k_X are their respective degradation rates. Shifted Hill functions, denoting the effect of X on Y, are defined as:

where H⁻ (X) is the inhibitory Hill function, H⁺ (X) is the excitatory Hill function, and |${\lambda _{X,Y}}$| denote the equivalent of fold-change in production of Y due to X (29). The parameters used here are: |${k_N} = 1.0,{\rm{ }}{k_{X = 1.2}},{\rm{ }}{g_N} = 120,{\rm{ }}{g_X} = 150$|⁠; |${\lambda _{N,X}}{\rm{ = 4,}}\,{\lambda _{X,N}}{\rm{ = 5}}$|⁠, and |${\lambda _{M,X}}{\rm{ = 2}}$|⁠. For different Hill functions used in the model, |$N_X^0 = X_N^0 = M_X^0 = 400,{\rm{ }}{n_{N,X}} = {n_{X,N}} = 4,{\rm{ }}{n_{M,X}} = 2$|⁠, where |${n_{X,Y}}\ {\rm{ }}{n_{X,Y}}$| denotes the Hill coefficient for the Hill function corresponding to the effect of X on Y, and |$X_Y^0$| represents the respective half-maximal concentration. Degradation rates (represented in per unit hour) for XIAP and NFκB are estimated from experimental data on their half-lives (30, 31). The fold-change for effect of MNK on XIAP has been estimated from our results, while the effect of XIAP on NFκB and vice versa were gathered from existing data (15, 32). Production rate (represented in 1,000 molecules per hour) estimation is based on total number of protein molecules per cell reported for signaling molecules (∼100,000; ref. 33). For considering the effect of NRAGE mimic, a shifted Hill function |${H^s}( {Nr,{\rm{ }}{\lambda _{Nr,N}}} )$| is included in the first term of equation for NFκB levels, since NRAGE is not supposed to alter the levels of XIAP, but its interaction with NFκB. Parameters used are: |${\lambda _{Nr,N}} = 0.5$| (estimated from our data), and |$Nr_N^0 = 500,{\rm{ }}{n_{Nr,N}} = 2$|⁠.

The statistical analyses were conducted using GraphPad Prism (GraphPad Software, Inc.) Student two-tailed t test, Fisher exact test (IHC), and Mantel–Cox log-rank test (survival). Differences were considered significant at P < 0.05. When comparing multiple groups, ANOVA protected Tukey HSD test was performed. Comparison of XIAP mRNA expression using the two probesets were analyzed with respect to their expression distribution using the rank-based Mann–Whitney test.

To explore whether XIAP expression correlates with survival outcomes in breast cancer, we queried expression data from a collection of 1,032 breast cancers, using microarray probe sets targeting the open reading frame (ORF) of XIAP (Supplementary Fig. S1). Breast cancer patients with elevated expression (top quartile) of XIAP mRNA had decreased event-free survival compared with patients with lower XIAP mRNA levels (Fig. 1A). We further examined whether expression of XIAP was independently associated with disease-free survival adjusting for available covariates, which included PAM50-based molecular subtype and lymph node status. High XIAP expression was associated with increased HR of recurrence (HR 1.68, P < 0.001) after adjusting for PAM50 subtype and lymph node status, while expected trends were observed with increased recurrence noted for nonluminal molecular subtypes and decreased recurrence risk noted for lymph node–negative disease. Given the role of IAP proteins in metastasis (34), we hypothesized that XIAP expression would be increased in breast cancer patients with lymph node–positive disease (LN+). We observed increased expression of XIAP among patient tumors with lymph node involvement versus no lymph node involvement at diagnosis (Fig. 1B and C).

Lymph node involvement and the presence of tumor cell clusters (tumor emboli) in the dermal lymphatic vessels is a classic feature of IBC presentation at diagnosis. Therefore, we conducted IHC analysis for XIAP protein expression in breast tissue microarrays, which included benign and malignant samples (n = 198) of different stages and grades from non-IBC and IBC patients (Table 1). Overall, positive cytoplasmic staining of XIAP was only observed in invasive breast tumors and triple-negative samples. IBC samples characterized by tumor emboli showed strong staining for XIAP in >90% cells along with positive staining in the identified tumor emboli. Representative images are shown in Fig. 2A–F. We performed a multivariate analysis to determine whether XIAP overexpression in IBC is related to other clinicopathologic features [e.g., histologic grade, hormone receptor status, HER2 status, and triple-negative breast cancer (TNBC) status - n = 158]. In univariate analysis, XIAP expression was significantly associated with high histologic grade (grade 3 vs. grade 1; HR = 1.305; P < 0.001), ER status (ER⁺ vs. ER⁻; HR = 0.887; P = 0.036), PR status (PR⁺ vs. PR⁻; HR = 0.887; P = 0.010), TNBC status (TNBC⁺ vs. TNBC⁻; HR = 1.178; P = 0.012) and tumor phenotype status (IBC vs. non-IBC; HR = 1.299; P = 0.010). In multivariate analysis, including all parameters associated with XIAP expression in univariate analysis, only histologic grade (grade 3 vs. grade 1; HR = 1.232; P = 0.016) and tumor phenotype (IBC vs. non-IBC; HR = 1.230; P = 0.049) remained significant. These data suggest that XIAP expression is correlated with breast cancer of higher histologic grade and that XIAP overexpression is specifically associated with IBC, independent of the classical clinicopathologic determinants of IBC.

We next investigated whether XIAP upregulation contributes to the hyperproliferative phenotype in IBC. To explore this, we used SUM149 (basal-like, constitutively activated EGFR) and SUM190 (luminal-like, HER2-overexpressing) tumor cells, which are considered true IBC-like models derived from primary tumors of IBC patients before treatment (35, 36). To assess the global effects of modulating XIAP expression in IBC cells we conducted transcriptome profiling of genetically modified derivatives with XIAP overexpression (wtXIAP), depletion (shXIAP), and reconstitution (shXIAP+XIAP), along with appropriate vector controls. XIAP expression and function in these variants was validated by immunoblot (Fig. 2G) and measurement of caspase activation and cell viability after treatment with TRAIL (Fig. 2H). GSEA analysis and GNF expression atlas ontologies revealed a network of related gene sets enriched in XIAP-overexpressing cells (Supplementary Fig. S2) reported to be associated with cell-cycle regulation and proliferation, response to cell stress and stem cell maintenance, and resistance to hypoxic and oxidative stress. The list of the top 100 genes that are differentially expressed in XIAP high versus low samples are provided in Supplementary Table S2A and S2B. Of interest was a strong and positive correlation between XIAP overexpression and gene sets (37) enriched for high-grade breast cancer (Fig. 2I), corroborating the observed increased XIAP staining in IBC specimens, in particular IBC, which are already at stage III or higher at diagnosis.

To further investigate the clinical relevance of XIAP in IBC, we evaluated how the gene expression profile of the XIAP-overexpressing versus knockdown cells compared with a published IBC-patient derived gene signature from a comparative analysis of untreated primary tumors from stage- and subtype-matched IBC and non-IBC patients, which includes the largest collection of IBC tumors from the World IBC Consortium (7). As expected, the patient-derived SUM149 cell line is IBC-like with an average posterior probability (similarity) of 44.7%; likewise, the XIAP-overexpressing (wtXIAP) cells also show a significant posterior IBC probability (i.e. 51.5%). In contrast, we observed that knockdown of XIAP (shXIAP) abolished the IBC-specific patient gene expression profile (i.e., posterior IBC probability of 0.05%; Fig. 2J). Taken together, these results indicate that XIAP maintains an IBC-like phenotype and associated gene expression signatures, which are dominated by proliferative and prosurvival gene network.

Despite the correlation of XIAP overexpression with proliferative genes, modulating XIAP expression alone in vitro does not have a significant effect on cellular proliferation (10). Therefore, to determine a possible in vivo role for XIAP overexpression, we characterized the tumor growth kinetics of the XIAP modulated (wtXIAP, shXIAP, and vector control) cells. Tumor cells were implanted in the mammary fat pad of nude mice and tumor growth measured over time (Fig. 3A). Initially all mice formed tumors with similar kinetics; however, the growth in mice bearing wtXIAP-overexpressing tumors was significantly increased compared with vector control tumors (doubling time of 6.9 days, wtXIAP; 10.3 days, vector controls). In contrast, although shXIAP tumors grew to palpability, most regressed (10/12) or plateaued in size compared with vector control tumors in the study period. Vector controls are shown combined in Fig. 3A as they had similar growth kinetics (separate in Supplementary Fig. S3A). Demonstrating specificity of these phenotypes to XIAP, robust tumor growth similar to wtXIAP was observed in tumors expressing an shRNA-resistant XIAP construct (shXIAP+wtXIAP; Fig. 3A).

In addition to our exogenously XIAP-modulated cell lines, we also compared parental SUM149 cells to rSUM149, a highly apoptotic-resistant cell line with endogenous XIAP overexpression dependent on XIAP IRES-mediated translation (10). High XIAP levels in rSUM149 correlate with multidrug resistance to chemo-, immuno-, and targeted therapy–mediated apoptosis (20); however, the in vivo growth characteristics of this cell line has not been studied. Similar to wtXIAP tumors, mice bearing rSUM149 tumors showed an aggressive growth pattern with formation of multiple tumor cell clusters (Fig. 3B). Representative pictures from tumor-bearing mice and related statistical analysis for SUM149 and rSUM149 are shown in Fig. 3C and D, respectively; additional images are shown in Supplementary Fig. S3.

Collectively, these results reveal that XIAP overexpression (endogenous or exogenous) enhances tumor growth, possibly explaining why XIAP expression is associated with aggressive features in patient tumors.

A pathologic hallmark of IBC tumors is the formation of tumor emboli enriched with ALDH⁺ cells [enzymatic marker of cancer stem-like cells reported to be high in IBC cells and patient tumors (38) and postulated to reflect collective tumor cell migration of IBC cells (39)]. On the basis of our evidence of a proliferative signature corresponding with enhanced tumorigenicity in vivo, we investigated whether XIAP expression modulates cancer stem-like characteristics. Indeed, both wtXIAP and rSUM149 cells exhibited increased proportions of ALDH⁺ cells compared with significantly reduced ALDH positivity in XIAP-silenced (shXIAP) cells (Fig. 3E). This effect was specific to XIAP, as reconstitution of XIAP in shXIAP cells enabled reemergence of ALDH⁺ cells.

To test whether XIAP inhibition affects IBC tumor cell motility, we employed a dorsal skin fold window chamber model in nude mice that allows for intravital imaging of local tumor growth and migration (Fig. 3F and G). GFP-SUM149 (vector ctr) IBC tumors grew as multiple tumor cell clusters inside the window chamber (Fig. 3H), similar to that observed in IBC patients (39, 40), as opposed to single solid masses observed with other breast cancer lines (22). Using this approach, we compared the short-term (0–120 hours) growth and migration pattern of the SUM149-derived vector control and shXIAP-implanted cells, which revealed significant inhibition of motility in the tumors arising from XIAP-depleted cells (Fig. 3I).

Gene expression profiles identified 933 differentially expressed genes between control and wtXIAP tumors (n = 3 each genotype, Supplementary Table S3). Those genes were enriched for biological processes of transcription, RNA biosynthesis, and protein metabolism among others. As IBC patient tumor profiles are dominated by NFκB target genes, we investigated whether IBC tumors generated with XIAP-overexpressing cells have increased NFκB activity. Expression2Kinases (X2K) analysis revealed that this gene list was enriched for target genes of two transcription factors in the NFκB family (RELA and NFκB1). In addition to target gene enrichment analysis, X2K also builds a protein–protein interaction (PPI) network that provides signal transduction pathways capable of explaining observed gene expression differences. Analysis of this PPI network identified a subnetwork regulated by NFκB activity (Fig. 4A). These data were further confirmed with qRT-PCR analysis: increased expression of several known NFκB target genes (NFKB1, MYC, and TNFAIP3) was observed in wtXIAP and shXIAP+XIAP tumors, while their expression was reduced in shXIAP tumors (Fig. 4B). Immunoblot analysis of both wtXIAP and shXIAP+XIAP tumors confirmed enhanced activation of the nuclear transcription factor NFκB (phospho-p65; Supplementary Fig. S3).

To further investigate the role of XIAP-mediated NFκB activity in IBC cells, we employed a small peptide mimetic modeled after the NRAGE repeat domain (41), which blocks the XIAP–BIR1 domain interaction with TAB1, interrupting XIAP-NFκB signaling (15). Treatment of cells with NRAGE peptide, delivered using Endoporter, led to decreased transcriptional activity of NFκB as measured by target gene expression (Fig. 4C). wtXIAP cells exhibited a modest increase in anchorage-independent growth relative to control cell lines (Fig. 4D and E). Treatment with NRAGE peptide reduced anchorage-independent growth in control and wtXIAP cells in a dose-dependent manner (Fig. 4D and E). Collectively, these results reveal that XIAP drives activation of NFκB and its target genes, demonstrating a functional interplay in advanced breast cancers like IBC, which are characterized by an increased proliferative state. Furthermore, use of the NRAGE peptide highlights the potential for developing BIR1 domain antagonists that can target the XIAP–NFκB interaction and/or signaling to potentiate therapeutic apoptosis in IBC cells.

Our results demonstrate a role for XIAP in the proliferative phenotype of IBC through a functional partnership with NFκB, which we have shown to be caspase-independent (13). Although previous reports indicate pervasive NFκB activity in IBC and suggest that NFκB activity may be downstream of the EGFR/HER2–MAPK signaling (6), the link between the two remains ill-defined. MAPK signaling is a critical regulator of stress response, including control of protein synthesis machinery driving cancer cell survival (42). To evaluate the effects of ERK1/2 signaling on XIAP expression, we treated SUM149 and SUM190 IBC cells with the MEK1/2 inhibitor UO126, which effectively reduced ERK1/2 MAPK phosphorylation, but led to only modest decreases in XIAP levels (Fig. 5A). As expected, p38 MAPK phosphorylation, which is downstream of MKK3/4/6, was relatively unchanged by UO126. Both p38 and ERK1/2 MAPK signaling intersect translation machinery through the MNK, which phosphorylates the cap-binding protein, eIF4E (43). Likely due to the maintenance of p38 MAPK signaling during UO126 treatment, MNK signaling to eIF4E was also only modestly reduced (Fig. 5A). Given MNK's known roles in oncogenesis and survival signaling (44), we next investigated whether MNK functions downstream of MAPK signaling to promote XIAP–NFκB signaling. Treatment with the prototypical MNK inhibitor CGP57380, abolished eIF4E (S209) phosphorylation, indicating MNK signaling interruption. Strikingly, MNK inhibition led to a significant decrease in XIAP protein levels in both cell lines, an effect more robust than that of UO126 (Fig. 5B). The intensity of XIAP reduction by MNK inhibition, relative to ERK1/2 inhibition, suggests that MNK more directly controls XIAP than ERK1/2 MAPK. Confirming this effect on XIAP expression, MNK depletion by RNAi also led to a decline in XIAP protein levels (Fig. 5C), and overexpression of a constitutively active MNK1 mutant (T344D) in HeLa cells induced XIAP expression concomitant with increased eIF4E (S209) phosphorylation (Fig. 5D). Quantitation of the effects of MNK modulation on XIAP protein levels (Fig. 5A–D) is shown in Fig. 5E. Taken together, these results suggest that MNK activity controls XIAP expression.

We next investigated whether MNK inhibition could target NFκB (p65) activity. Indeed, using either CGP57380 (Fig. 5B) or direct MNK depletion (Fig. 5C) reduced the levels of both p-p65 and p65 in IBC cells, coincident with a decline in XIAP. Analysis of RNA expression by qRT-PCR following MNK1 depletion revealed a decrease in RNA levels of several select NFκB target genes including MYC and survivin (Fig. 5F). Together, these results indicate that modulating MNK signaling regulates XIAP protein levels and downstream NFκB target genes. Thus, MNK controls XIAP–NFκB signaling and can be targeted to restrain the oncogenic effects of XIAP–NFκB activity.

MNK inhibition has been shown to reduce in vivo tumor growth in several cancer types (45, 46). However, the role of MNK signaling in mediating tumor cell invasion, a characteristic of IBC we show is reliant upon XIAP expression (Fig. 3), has not been tested. Using the window chamber model with GFP-tagged SUM149, we tested infiltration of SUM149 cells within the window chamber in vehicle or CGP57380-treated mice. MNK inhibition with CGP57380 led to a marked reduction in cell dissemination within the window chamber, suggesting that MNK signaling enables IBC migration (Fig. 5G and H). Thus, MNK inhibition reduces XIAP levels and NFκB activity and mirrors the effects of XIAP depletion in IBC cells in vivo (Figs. 3 and 5).

Integrating our quantitative immunoblot and qRT-PCR data indicating XIAP regulation of NFκB in a MNK-responsive manner combined with earlier reports that also propose XIAP as a transcriptional target of NFκB, we constructed a quantitative mathematical model to decode the dynamics of the MNK/XIAP/NFκB axis. Mutual activation between NFκB and XIAP leads to bistability in the system: cells display either low XIAP/low NFκB, or high XIAP/high NFκB protein levels (shown by two solid green circles in Fig. 6A). To switch between these two states/phenotypes, cells must cross a “tipping point” or threshold (as shown by hollow green circle in Fig. 6A). Once cells have attained such a threshold (high XIAP, high NFκB), this mutual activation would maintain that state (Fig. 6B and C). Only when a “brake” is significantly applied on either NFκB or XIAP levels (e.g., using an inhibitor like NRAGE peptide), can cells be postulated to exit that state and eventually attain a low XIAP, low NFκB state (Fig. 6D). These findings indicate a homeostatic relationship between XIAP and NFκB and suggest that mutual activation of XIAP/NFκB stabilizes a hyperproliferative phenotype in cancer.

Collectively, our findings define XIAP as a signaling intermediate linking the MAPK mitogenic cascade to NFκB prosurvival signaling (schema in Fig. 6E). In doing so, XIAP confers a proliferative signature and phenotype to IBC cells, enabling aggressive tumor growth in one of the most lethal subtypes of breast cancer.

The current study uncovers XIAP as an oncogenic signaling intermediate, linking the MAPK and NFκB signaling pathways, with significant implications for locally advanced breast cancer tumor growth. XIAP mRNA levels correlated with lymph node involvement and decreased event-free survival among patients with IBC, and XIAP overexpression was observed in high-grade breast cancers and IBC patient tumors, substantiating previous reports of XIAP overexpression in breast cancer tissue (47) and correlation of XIAP expression with tumor recurrence in basal-like breast cancer patients (48). We show that XIAP is necessary for the constitutive activation of the NFκB pathway in IBC, and demonstrate that the XIAP–NFκB axis directly correlates with the tumor growth rate in vivo. These findings reveal a functional necessity for XIAP expression in the progression of aggressive, locally advanced breast cancers like IBC. Finally, we defined a critical role of XIAP in transducing MAPK signals to NFκB downstream of MNK, possibly explaining the survival and oncogenic phenotypes associated with MNK signaling (44).

It has been postulated that in aggressive tumors like IBC, a delicate balance exists between unabridged cellular proliferation, the requirement for cancer stem cell self-renewal, and the ability of cancer stem cell progeny to “self-metastasize” and migrate away, freeing up space for continued tumor expansion (39, 49). Our results showing that XIAP expression directly correlates with the number of ALDH⁺ cells and cell motility in IBC cells warrants further investigation of the role of XIAP in IBC metastatic progression. Studies overwhelmingly show that XIAP antagonism in established tumors or in cell lines can sensitize tumor cells to therapy-mediated cell death, thereby implicating XIAP as a chemoresistance factor (50). However, other reports suggest that XIAP expression correlates with favorable clinical outcome (51). Perhaps, contributing to these contradictions are the upstream signals regulating XIAP expression and broader cellular context of the signaling landscape.

MNK is well known for its role in regulating IRES-mediated translation (21) and in our study, interruption of MNK signaling led to a significant reduction in XIAP protein expression. Intriguingly, XIAP mRNA contains an IRES (52) and MNK regulation of eIF4G and eIF4E may function to facilitate XIAP translation in IBC, or MNK may act on XIAP through one of its more recently described eIF4E-independent contexts (53). Interestingly, MNK1/2–null (44) like XIAP-null (54) animals are viable and do not exhibit any major defects in growth and ability to undergo apoptosis. MNK1/2–null mice were reported to exhibit delays in tumor development, suggesting a role for MNK1/2 and downstream effectors in tumorigenesis (44). Our current study showing a strong correlation between XIAP expression and tumor growth in vivo in two MAPK-hyperactivated IBC models provides support for XIAP as a possible downstream effector in potentiating the mitogenic effects of MAPK/MNK signaling. Indeed, we found that MNK inhibition restricted IBC tumor invasion/migration, suggesting therapeutic potential in targeting MNK in IBC as a means to target the XIAP–NFκB axis in cancer.

IBC cells overexpressing XIAP were shown to be resistant to immunotherapy-mediated cell death (13). This work revealed a caspase-independent ability of XIAP to activate NFκB and demonstrated that direct targeting of caspase-binding domains may not reverse resistance (unpublished data). The efficacy of the NRAGE peptide, which prevents the XIAP–BIR1 domain from activating NFκB, in inhibiting anchorage-independent growth in IBC cells underscores the recent mounting evidence for a nonapoptotic function of XIAP as a signaling intermediate in tumor growth (14, 55). However, the practical hurdles of delivering a peptide to tumor cells renders this approach clinically difficult, particularly as NRAGE peptide has a short half-life. Our findings demonstrate that inhibition of MNK signaling represents another mechanism to target XIAP–NFκB signaling in IBC. The finding that MNK inhibition disrupts tumor dissemination is especially relevant for IBC and other subtypes of cancer that are highly metastatic (Fig. 5). Importantly, MNK inhibitors are being developed and pursued clinically, making MNK a more practical target for this pathway. This work presents a new druggable pathway consisting of MNK, XIAP, and NFκB (Fig. 6) that can be used to enhance the efficacy of therapeutic agents by pushing the cells below the tipping point (Fig. 6A) and consequently constraining the proliferative advantage. Thus, XIAP serves as a link between MAPK and NFκB signaling to control IBC proliferation and tumor aggression.

No potential conflicts of interest were disclosed.

Conception and design: M.K. Evans, M.C. Brown, M.A. Morse, S. Van Laere, G.R. Devi

Development of methodology: M.K. Evans, M.C. Brown, J. Geradts, X. Bao, T.J. Robinson, M.K. Jolly, G.M. Palmer, G.R. Devi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.K. Evans, M.C. Brown, J. Geradts, X. Bao, P. Vermeulen, G.M. Palmer, S. Van Laere, G.R. Devi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.K. Evans, M.C. Brown, J. Geradts, X. Bao, T.J. Robinson, P. Vermeulen, G.M. Palmer, H. Levine, M.A. Morse, S. Van Laere, G.R. Devi

Writing, review, and/or revision of the manuscript: M.K. Evans, M.C. Brown, J. Geradts, X. Bao, T.J. Robinson, M.K. Jolly, P. Vermeulen, M. Gromeier, H. Levine, M.A. Morse, S. Van Laere, G.R. Devi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Gromeier, H. Levine, S.J. Van Laere, G.R. Devi

Study supervision: G.R. Devi

This work was supported by American Cancer Society Research Scholar Grant (to G.R. Devi), P30 Cancer Center Support Grant NIH CA014236 developmental funds (to G.R. Devi), Department of Defense W81XWH-13-1-0047 and -0046 (to G.R. Devi and M.A. Morse); Department of Defense W81XWH-17-1-0297 (to G.R. Devi); Duke University Diversity Enhancement Fellowship (to M.K. Evans), and the Duke School of Medicine Interdisciplinary Inflammatory Breast Cancer Colloquium Funds (to G.R. Devi); Computational Cancer Biology fellowship from Gulf Coast Consortia - CPRIT RP170593 (to M.K. Jolly); and NSF PHY-1427654 (to H. Levine). The authors would like to thank Amy Aldrich, Courtney Edwards, Adrian Ramirez, Arianna Price, Ronnie Shammas Jr, Larissa M. Gearhart-Serna for technical assistance; Drs. Hengtao Zhang, Yulin Zhao, and Scott Sauer for technical support in the window chamber studies. We thank Tao Wang for flow cytometry support, members of the Devi Lab, Drs. Mark Dewhirst, Ashley Chi, Jeffrey Marks, and Sally Kornbluth, for helpful discussions during manuscript preparation, Dr. Donna Crabtree for editorial assistance, and Duke University Core facilities (Light Microscopy, Cancer Center Isolation Facility, Center for Genomic and Computational Biology, Flow Cytometry, Optical Molecular Imaging and Analysis, Preclinical Translational Research Unit).

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.