TRIB3 supports breast cancer stemness by suppressing FOXO1 degradation and enhancing SOX2 transcription

The existence of breast cancer stem cells (BCSCs) is a major reason underlying cancer metastasis and recurrence after chemotherapy and radiotherapy. Targeting BCSCs may ameliorate breast cancer relapse and therapy resistance. Here we report that expression of the pseudokinase Tribble 3 (TRIB3) positively associates with breast cancer stemness and progression. Elevated TRIB3 expression supports BCSCs by interacting with AKT to interfere with the FOXO1-AKT interaction and suppress FOXO1 phosphorylation, ubiquitination, and degradation by E3 ligases SKP2 and NEDD4L. The accumulated FOXO1 promotes transcriptional expression of SOX2, a transcriptional factor for cancer stemness, which in turn, activates FOXO1 transcription and forms a positive regulatory loop. Disturbing the TRIB3-AKT interaction suppresses BCSCs by accelerating FOXO1 degradation and reducing SOX2 expression in mouse models of breast cancer. Our study provides insights into breast cancer development and confers a potential therapeutic strategy against TRIB3-overexpressed breast cancer.

B reast cancer is composed of heterogeneous cell populations that interact in complex networks 1 . Cancer stem cells (CSCs) play key roles in intra-and intertumoral heterogeneities, which are responsible for tumor progression, resistance to therapy, and disease relapse 2 . Breast CSCs (BCSCs) represent a dynamic subpopulation of breast cancer cells (BCCs), which have the capabilities of self-renewal, tumor initiation, and the ability to give rise to more differentiated progeny upon xenotransplantation into immunocompromised mice 3,4 . CSC markers, including pluripotency genes with normal stem cells in the tissue-of-origin, often share transcriptional profiles 4 . Recent work has indicated the plasticity of BCSCs, revealing that these cells exist in two distinct but interchangeable states: epithelial-mesenchymal transition-CSCs, a quiescent mesenchymal type marked as EpCAM − CD49f + CD44 + CD24 − , and mesenchymal-epithelial transition-CSCs, an epithelial, proliferative type identified as EpCAM + CD49f + ALDH + 2,5 . Plastic CSCs are not necessarily rare and/or quiescent, and niche signals may instruct them following neutral competition dynamics. Differentiated cells and transient-amplifying cells can be reprogrammed into CSCs in the niche via plastic mechanisms 6 . Evidence suggests that CSCs are always recreated as long as the CSC niche remains intact. Hence, modulating CSC niche functions has become an attractive alternative rather than directly targeting CSCs 6 .
The pseudokinases Tribbles homolog 3 (TRIB3, NIPK, SIKP3) is one of three mammalian homologs of tribbles in Drosophila, which inhibit mitosis in embryo and germ cell development 7 . TRIBs are fundamental regulators of cellular stress, the cell cycle, differentiation, and proliferation 8 . Cisplatin-enriching lung CSCs show activated TRIB1/HDAC activity 9 . Mice reconstituted with Trib2 by the engraftment of hematopoietic stem cells retrovirally expressing Trib2 uniformly developed fatal transplantable acute myelogenous leukemia 10,11 . TRIB3 has recently been identified as a stress sensor in response to various tumor microenvironments or niche-rich stressors, including amino acid or glucose deficiency, insulin, unfolded protein accumulation in the endoplasmic reticulum, and oxidative damage 8,12,13 . TRIB3 promotes chronic inflammation and cancer by interacting with intracellular signaling and functional proteins. The interaction of TRIB3 and the autophagic receptor p62 interferes with the degradation of autophagy and the ubiquitin-proteasome system to control the initiation and progression of cancer 14 . TRIB3 promotes PML-RARα-driven acute promyelocytic leukemia by interacting with PML-RARα and regulating PML-RARα degradation 15 . These findings are consistent with the findings that enhanced TRIB3 expression negatively associates with overall survival in colorectal cancer 16 , breast cancer 17,18 , and gastric cancer 19 . Given that TRIB3 senses a variety of stress signals, and that enhanced TRIB3 expression leads to poor prognosis for breast cancer patients, we postulated that TRIB3 contributes to the pathogenesis of breast cancer via its tumor initiation capacity. We studied the coordinative functions and mechanisms of TRIB3, Sry-related highmobility box 2 (SOX2), and Forkhead box O1 (FOXO1) in supporting BCSCs, and elucidated the implications of these findings.

RESULTS
TRIB3 is associated with the stemness of breast cancer. We queried the Curtis breast dataset from the Oncomine database of BCCs for information on 1556 patients with invasive ductal breast cancer and 148 patients with invasive lobular breast cancer. These patients expressed higher levels of TRIB3 than their normal counterparts (n = 144) (Fig. 1a). Using human tumor tissue microarrays of breast adenocarcinoma, higher TRIB3 levels were observed in tumor tissues than in adjacent nontumor tissues (Fig. 1b). TRIB3 expression showed no stage differences among a variety of breast cancer patients. We then re-interrogated the TRIB3 expression in the published GSE12790 dataset and no expression differences were found among the luminal, Her2amplified, and basal BCC lines ( Supplementary Fig. 1a). TRIB3 is universally highly expressed in HMLER and eight distinct breast cancer epithelial cell lines but not in human mammary epithelial cells (HMLEs) (Fig. 1c). We interrogated The Cancer Genome Atlas (TCGA) database using online kmplot tools, to evaluate BCCs from 1117 patients with breast adenocarcinoma 20 . The patients were divided into three groups based on their relative TRIB3 expression levels in BCCs. Patients with tumors expressing TRIB3 mRNA in the upper tertile (TRIB3 H ) had a significantly shorter overall survival than patients with tumors expressing TRIB3 mRNA at the lower and intermediate tertile levels (TRIB3 L+M ) (p = 0.0002; Fig. 1d). We then queried the PubMed GEO database to evaluate BCCs from 582 patients with breast adenocarcinoma as previously described 21 . Approximately two of three of these patients (426 of 582) did not have detectable cancer in the regional lymph nodes at the time of surgery and were not administered adjuvant therapy. The remaining patients had the detectable disease in regional lymph nodes and received adjuvant therapy. Among 582 patients, 46% relapsed (n = 270) and had a median metastasis-free survival time of 22.1 months. The relative level of TRIB3 in BCCs was used to segregate patients into three groups (Supplementary Table 1). Patients with tumors expressing TRIB3 mRNA at the upper tertile level (TRIB3 H ) had a significantly shorter metastasis-free survival time than patients with tumors expressing TRIB3 at the lower (TRIB3 L ) or intermediate tertile (TRIB3 M ) levels (p < 0.0001; Fig. 1e). Moreover, patients with TRIB3 H tumors had lower rates of overall survival (TCGA-BRCA; Supplementary Fig. 1b) and metastasis-free survival (PubMed GEO Datasets; Supplementary Fig. 1c) among ER+, HER2+, and triple-negative breast cancer patients. Using a multivariate Cox regression model, high TRIB3 expression was found to have a predictive value for short metastasis-free survival (Supplementary Table 2). These data indicate that elevated TRIB3 expression positively correlates with breast cancer progression, metastasis, and relapse.
Inflammation, hypoxia, and metabolic stresses play vital roles in the progression of breast cancer by affecting key CSC phenotypes 1,2,22,23 . We found that cytokines interleukin (IL)-6, transforming growth factor-β, IL-1β, tumor necrosis factor-α, and hypoxia, as well high glucose enhanced TRIB3 expression in MCF7 cells ( Supplementary Fig. 1d), suggesting that TRIB3 acts as a stress sensor in response to a diverse range of stressors. The cytokine IL-6 was chosen for the following experiments due to its potent effects on TRIB3 induction. IL-6 treatment increased the tumorsphere-formation ability of MCF7 cells ( Supplementary  Fig. 1e). IL-6 enhanced TRIB3 expression in MCF7 and MDA-MB-231 cells in dose- (Supplementary Fig. 1f) and time- (Supplementary Fig. 1g) dependent manners. Silencing TRIB3 expression reduced mammosphere formation in MCF7 cells with or without IL-6 stimulation ( Supplementary Fig. 1h). In isolated mammary epithelial cells (MECs) from spontaneous breast cancer mice, high TRIB3 or Trib3 expression was found in the CD24 + CD29 low subpopulation (Fig. 1f, g) and CD24 + CD90 + stem-like subpopulation ( Supplementary Fig. 1i) from mice with luminaltype MMTV-PyMT breast cancer, in the CD24 + CD29 high subpopulation from mice with her2-type MMTV-ErbB2 breast cancer (Fig. 1h, i), and in the CD68 + tumor-associated macrophage (TAM) adjacent area in breast cancer patients ( Supplementary Fig. 1j). In TAMs and MECs co-culture assays ( Supplementary Fig. 1k left), silencing Trib3 in MECs had no effect on IL-6 production from MMTV-PyMT-derived TAMs ( Supplementary Fig. 1k middle), but reduced mammosphere formation ( Supplementary Fig. 1k right). These data indicate that elevated TRIB3 expression links BCSC-promoting cytokines and other stressors to breast cancer stemness.
We ranked 20,530 genes from breast cancer samples in the TCGA dataset by their relative TRIB3 expression in the top 10th percentile (TRIB3 Hi ) vs. the bottom 10th percentile (TRIB3 Low ) for gene-set enrichment analyses. TRIB3 Hi tumor samples were enriched in the expression of gene signatures associated with "ES/Stem cells" in comparison with TRIB3 Low samples (Fig. 2a). Three-dimensionally (3D) cultured CD44 bright CD24 dim (CD44 Br CD24 Di ) and non-CD44 bright CD24 dim (non-CD44 Br CD24 Di ) MCF7 cells (Fig. 2b) were isolated, and each subset was examined for TRIB3 mRNA and protein expression, and stemness-related proteins, including SOX2, NANOG, Octamer-binding transcription factor 4 (OCT4), Kruppel-like factor 4 (KLF4), and c-Myc. The mRNA expression of TRIB3, SOX2, NANOG, OCT4, KLF4, and c-Myc was higher in CD44 Br CD24 Di cells than in non-CD24 Di CD44 Br cells (Fig. 2c) Fig. 1 High TRIB3 expression is negatively associated with overall survival and metastasis-free survival in breast cancer. a Graph derived from published data available in the Oncomine database. The box charts depict the relative expression of TRIB3 in invasive ductal (top) and invasive lobular (bottom) breast cancer patients. Centre line = 50th percentiles; bounds of box = 25th and 75th percentiles; bars = 10th and 90th percentiles; whiskers = min and max values. b Formalin-fixed, paraffin-embedded tissue microarray sections of normal and breast cancer tissues were stained with an anti-TRIB3 antibody. Tissue-bound TRIB3 is shown in brown (left). The dot plots show the relative expression of TRIB3 calculated by the intensity of the brown color in tissue array immunohistological images (right). c Immunoblots of protein lysates from HMLE, HMLER, and BCCs, as indicated at the top of the left panel.
The relative TRIB3 expression quantification from three independent studies is shown (right). d, e Graph derived from TCGA (d) or published data (e) available in the PubMed GEO database (GSE2603, GSE5327, GSE2034, and GSE12276). For each analysis, 1117 patients (d) or 582 patients (e) were segregated into tertiles, with the group designated TRIB3 H representing one-third of the patients who had tumors with the highest levels of TRIB3 mRNA and the group designated TRIB3 L representing one-third of patients who had cancers with the lowest levels of TRIB3 mRNA. One-third of patients who had tumors with intermediate TRIB3 mRNA expression were designated as TRIB3 M . Overall survival (d) and metastasis-free survival (e) were determined by Kaplan-Meier analyses, and significant differences were determined by the log-rank test. The number of patients in each category, the total metastatic events, and the corresponding P-values (χ 2 -test) are shown in the embedded tables. f-i Flow cytometry of CD45 − CD31 − TER119 − (Lin − ) MECs from the mammary glands of 6-week-old MMTV-PyMT female mice (f) and 6-month-old MMTV-ErbB2 female mice (h). Relative Trib3 mRNA expression and TRIB3 protein expression are indicated in g and i. Data are presented as the mean ± SEM; P > 0.05 was considered not significant (NS); *P < 0.05, **P < 0.01, and ***P < 0.001 compared with the HMLE group in c and the normal gland group in g and i. Source data are provided as a Source Data file. SOX2, and NANOG was found in CD44 Br CD24 Di cells than that in non-CD24 Di CD44 Br cells (Fig. 2d). Moreover, Trib3-silenced MECs from MMTV-PyMT transgenic mice and MMTV-ErbB2 transgenic mice formed much fewer spheroids than the control MECs ( Supplementary Fig. 2a, b). Transfection of TRIB3negative HMLE cells with TRIB3 enhanced spheroid formation (Fig. 2e) and the CD44 Br CD24 Di subpopulation (Fig. 2g). The TRIB3-silenced cells exhibited a reduced capacity for spheroid formation ( Fig. 2f and Supplementary Fig. 2e) and a reduced number of cells in the stem-like CD44 Br CD24 Di subpopulation (Fig. 2h). In addition, silencing TRIB3 expression suppressed tumor growth ( Supplementary Fig. 2d, f) and metastasis ( Supplementary Fig. 2g). Finally, TRIB3-silenced BCCs reduced the capacity to form tumors in immunodeficient mice (Fig. 2i, j). These data suggest that elevated TRIB3 expression positively correlates with the spheroid formation in vitro and tumor engraftment efficiency in vivo.    Fig. 3c bottom). These results indicated that SOX2 is a major factor that mediates TRIB3-supporting breast cancer stemness. We thus focused on the effect of the TRIB3-SOX2 axis on breast cancer stemness.
Using human tissue microarrays of breast adenocarcinoma, SOX2 expression was determined to be higher in tumor tissues than in adjacent nontumor tissues, and the expression level of TRIB3 correlated with the expression level of SOX2 in tumor tissues (Fig. 3d). TRIB3 enhanced the SOX2 transcriptional activity in HMLE cells (Fig. 3e) and MCF7 cells ( Fig. 3f) but did not change the protein stability of SOX2 ( Supplementary Fig. 3d). Furthermore, silencing SOX2 expression reduced the capacity of HMLE cells transfected with a TRIB3-expressing vector to form spheroids (Fig. 3g). Overexpression of SOX2 rescued the spheroid formation capacity of TRIB3-silenced MCF7 (Fig. 3h) and MDA-231 cells ( Supplementary Fig. 3e). Silencing SOX2 had no additional effects in terms of reducing of CD44 Br CD24 Di subpopulation and limiting tumorsphere formation in TRIB3-shRNA MCF7 cells ( Supplementary Fig. 3f, g). These data suggest that elevated SOX2 expression is responsible for TRIB3supported breast cancer stemness.
To examine the dominant transcription factor (TF) that controls SOX2 expression in BCCs, we screened and silenced nine different SOX2-promoting TFs one by one in MCF7 cells. We found that silencing FOXO1 expression ( Fig. 4a) but not that of any other TF ( Supplementary Fig. 4a) reduced SOX2 expression in BCCs. Silencing FOXO1 expression reduced the SOX2 transcriptional activity in HMLE cells transfected with a TRIB3expressing vector (Fig. 4b). Overexpression of FOXO1 rescued the SOX2 transcription activity in TRIB3-silenced MCF7 cells (Fig. 4c). Moreover, we identified the SOX2 promoter-binding  table shows the correlation between TRIB3 and SOX2 based on their relative expression, graded by the percentage and intensity of the brown color staining in the immunohistological images (bottom). e Overexpression of TRIB3 enhanced the SOX2 transcriptional activity in HMLE cells, as determined by a luciferase reporter assay. f Silencing TRIB3 inhibited the SOX2 transcription activity in MCF7 cells, as determined by a luciferase reporter assay. g, h Tumor spheroid formation ability of HMLE cells (g) transfected with a control vector or a TRIB3-expressing vector and either CTRL-shRNA or SOX2-shRNA (as indicated at the bottom) or of MCF7 cells (h) transfected with CTRL-shRNA or TRIB3-shRNA, and either a control vector or a SOX2-expressing vector. All data are shown as the number of mammospheres per 1000 cells. Data are presented as the mean ± SEM of three independent assays; P > 0.05 was considered not significant (NS); *P < 0.05, **P < 0.01, and ***P < 0.001 compared with the vector or CTRL-shRNA group. Source data are provided as a Source Data file.  Supplementary Fig. 4c). These data indicate that FOXO1 enhances SOX2 transcriptional expression in TRIB3overexpressed BCCs.
TRIB3 enhances FOXO1 expression. We next investigated how TRIB3 upregulates FOXO1 expression in BCCs. TRIB3 markedly enhanced the half-life of FOXO1 degradation from 3 to 6 h in HMLE cells (Fig. 5a). Silencing TRIB3 reduced the half-life of FOXO1 degradation from 3.1 to 0.8 h (Fig. 5b). The degradation of FOXO1 was inhibited by the proteasome inhibitor MG132 but not by the autophagy inhibitor bafilomycin (BAF) in MCF7 cells (Fig. 5c). Silencing TRIB3 reduced FOXO1 expression in MCF7 and MDA-MB-231 cells, which was inhibited by MG132 but not by BAF (Fig. 5d). These data suggest that TRIB3 interferes with FOXO1 degradation by compromising the ubiquitin-proteasome system. The activation of various protein kinases often leads to the phosphorylation and ubiquitination of FOXO proteins 24 . Overexpression of TRIB3 reduced the S256, T24, and S319 phosphorylation of FOXO1 in HMLE cells (Fig. 5e). Silencing TRIB3 enhanced FOXO1 phosphorylation at S256, S319, but not at T24 in MCF7 cells (Fig. 5f). Indeed, ectopic expression of g Formalin-fixed, paraffin-embedded tissue microarray sections of normal or breast cancer tissues were stained with an anti-FOXO1 antibody. Tissuebound FOXO1 is shown in brown (top). A 3 × 2 contingency table shows the correlation between TRIB3 and FOXO1 based on their relative expression, graded by the percentage and intensity of the brown color staining in the immunohistological images (bottom). h, i Tumor spheroid formation of HMLE cells (h) transfected with a control vector or a TRIB3-expressing vector and either CTRL-shRNA or FOXO1-shRNA (as indicated at the bottom), or of MCF7 cells (i) transfected with CTRL-shRNA or TRIB3-shRNA and either a control vector or a FOXO1-expressing vector. All data are shown as the number of mammospheres per 1000 cells. Data are shown as the mean ± SEM; P > 0.05 was considered not significant (NS); *P < 0.05, **P < 0.01, and ***P < 0.001 compared with the vector or CTRL-shRNA group. Source data are provided as a Source Data file.
TRIB3 suppressed FOXO1 ubiquitination but did not reduce the ubiquitination of the S256D FOXO1 mutant in HMLE cells (Fig. 5g). Silencing TRIB3 enhanced FOXO1 ubiquitination but not S256A-mutated FOXO1 ubiquitination in MCF7 cells (Fig. 5h). These data suggest that phosphorylation of FOXO1 at S256 is a key signal for FOXO1 ubiquitination and degradation.
On the other hand, overexpression of TRIB3 in HMLE cells enhanced the mRNA expression of FOXO1 ( Supplementary  Fig. 8a) and FOXO1 luciferase activity (Supplementary Fig. 8b). Silencing TRIB3 attenuated the mRNA expression of FOXO1 and FOXO1 luciferase activity in MCF7 cells ( Supplementary Fig. 8a,  b). Interestingly, silencing SOX2 reduced the transcriptional activity of FOXO1 in HMLE cells transfected with a TRIB3expressing vector ( Supplementary Fig. 8c), but overexpression of SOX2 rescued the transcriptional activity of FOXO1 in TRIB3silenced MCF7 cells (Supplementary Fig. 8c). SOX2 binds DNA in a sequence-specific manner via its high-mobility-group domain 27 . We analyzed a 5 kb locus upstream of the FOXO1 gene transcription start sites (TSS). Using ChIP, we found that the −246~−1460 base pair (bp) region of the FOXO1 promoter upstream of the TSS was occupied by the SOX2 protein ( Supplementary Fig. 8d). We evaluated whether FOXO1 transcriptional activity was promoted by SOX2 using luciferase reporter constructs. Based on these assays, a −1032~−793 bp fragment of the FOXO1 promoter was identified as an enhancer region in which SOX2 was activated ( Supplementary Fig. 8e-g). We then mapped the binding locus on the FOXO1 promoter using electrophoretic mobility shift assays (EMSAs) with different oligos from the −1032~−753 bp fragment. The −882~−843 bp segment of the FOXO1 promoter was found to contain the SOX2binding sequence ( Supplementary Fig. 8h-j). These data suggest that a positive feedback mechanism of the TRIB3-FOXO1-SOX2 axis supports breast cancer stemness. TRIB3 enhances FOXO1 expression by impairing its degradation, the elevated FOXO1 induces SOX2 transcriptional expression, and the latter conversely activates FOXO1 transcriptional expression.
Disturbing the TRIB3/AKT interaction reduces cancer stemness. We examined whether TRIB3 reduced AKT1 phosphorylation by interacting with AKT1. The endogenous or overexpressed AKT1 was co-immunoprecipitated (IP) with TRIB3. The deletion mutants of Myc-tagged AKT1 were subjected to IP with TRIB3-HA. TRIB3 interacted with the region from residues 149 to 214 in the catalytic domain of AKT1 (Fig. 7a-c). Furthermore, the deletion mutants of green fluorescent protein (GFP)-tagged TRIB3 were used to map the AKT1 interaction region (Fig. 7d). AKT1 interacted with the C terminus of the TRIB3 KD domain (Fig. 7e).
To verify the critical role of the TRIB3/AKT1 interaction in the regulation of FOXO1 expression and tumor stemness, we tried to interrupt this interaction by screening for a short α-helical peptide able to inhibit this protein-protein interaction (PPI) 14,15 . We first docked the TRIB3 protein, created by homology models, with the AKT1 protein from Protein Data Bank (PDB, 4GV1) by Discovery Studio. The α-helical region at residue 191-204, which is in the AKT1 catalytic domain, is the closest α-helical region in the TRIB3-AKT1 interaction range (Fig. 7f, yellow α-helix). We further screened the abilities of a series of α-helical peptides to disturb the TRIB3/AKT1 interaction based on the crystallized secondary structure of the AKT1 catalytic domain available in the PDB (Fig. 7f, AKT1 red regions). We found that peptide Ae, which mimics the AKT1 191-204 α-helical region (Fig. 7f, yellow α-helix), displayed the best binding affinity with TRIB3 among all the α-helical peptides analyzed (Fig. 7g). The AKT1 230-325 region (WX region) reportedly mediates the TRIB3/AKT1 interaction. We further identified that deletion of the Ae mutant (△Ae) but not deletion of the WX mutant (△WX, 230-315AA) inhibited the binding of AKT1 to TRIB3 (Fig. 7h, i). The TRIB3binding regions of peptide Ae were predicted to reside inside the TRIB3 KD domain (Fig. 7j). These data indicate that the catalytic region of AKT1 (residues 149−214, especially 191-204) interacts with the C terminus of the TRIB3 KD domain to cause downstream effects. To assess the contribution of the peptide Ae amino-acid sequence to the TRIB3/AKT1 binding, each amino-acid residue of Ae was substituted with alanine. The residues 1V(M1), 5L(M5), 6T(M6), 11L(M11), 12Q(M12), and 13N(M13) of peptide Ae were critical for the binding of Ae to TRIB3, because the alanine mutations abolished Ae/TRB3 binding ( Supplementary Fig. 9a).

DISSCUSION
The double threat of CSCs includes therapy resistance and the ability to regenerate a tumor from original and distant sites once therapy is halted 4 . Many tumors, including breast cancer, harbor CSCs in dedicated niches 6,30 . The CSC niches or microenvironments determine not only the tumor progression molecular heterogeneity but also the fate of CSCs 31 . Both intrinsic and extrinsic stressors, including hypoxia, metabolism, reactive oxygen species, and inflammation, upregulate the CSC stress signaling pathway to enhance cancer cell survival and maintain cancer cell stemness 32 . However, the molecular mechanisms by which the stress signaling pathway sustains cancer cell stemness remain unclear. TRIB3 can act as a stress sensor in response to all of these stressors and thus participates in the pathogenesis of chronic inflammatory and malignant diseases by interacting with intracellular signaling and functional proteins 14,[33][34][35] . In this study, we demonstrated that elevated expression of TRIB3 is positively associated with the initiation and progression of breast cancer as well as with the poor prognosis of breast cancer patients by enhancing breast cancer stemness. TRIB3 protein acts as a stress sensor in response to a diverse range of stressors, including inflammation, hypoxia, and metabolic stress in BCCs. Enhanced TRIB3 activates the AKT1-FOXO1-SOX2 axis in TRIB3overexpressed BCCs. Importantly, interrupting the TRIB3/ AKT1 interaction by a modified α-helix peptide, Pep2-Ae, accelerated FOXO1 degradation, reduced SOX2 accumulation, and decreased the tumor-initiating capacity in MMTV-PyMT and PDX mice. Our work indicates that TRIB3 links stress signals to breast cancer stemness through the coordination of FOXO1 and SOX2 in breast cancer with high TRIB3 expression.
Overexpression of pluripotency genes, such as SOX2, OCT4, NANOG, KLF4, and c-Myc, can induce somatic cells to acquire pluripotency 36 . These proteins also show functional differences in CSCs [37][38][39] . Evidences showed that the expression of a single stemness-contributing protein might have a relatively broad range 40,41 . In this study, we found that mammary CD24 + CD29 lo cells from MMTV-PyMT mice and CD24 + CD29 hi cells from MMTV-ErbB2 mice express the higher TRIB3 than their agematched normal counterparts. Moreover, isolated MCF7 CD44 Br CD24 Di cells had higher TRIB3 and SOX2 expression levels than their counterparts. SOX2 plays an essential role, by not only regulating pluripotency but also mediating self-renewal and differentiation. SOX2 is aberrantly expressed in several types of cancers, such as breast, lung, ovarian and prostate cancers 41-45 .  Fig. 7 Mapping of the TRIB3-AKT1 interaction. a Mapping AKT1 regions binding to TRIB3. Schematic diagram of full-length AKT1 and deletion mutants. b, c HEK293T cells were co-transfected with the indicated AKT1-Myc and TRIB3-HA constructs. Cell extracts were IP with an anti-HA Ab. d Mapping TRIB3 regions binding to AKT1. Schematic diagram of full-length TRIB3 and deletion mutants. e HEK293T cells were co-transfected with the indicated AKT1 and TRIB3-GFP constructs. Cell extracts were IP with an anti-Myc Ab. f Prediction of the AKT1 and TRIB3 interaction using Discovery Studio. g Kinetic interactions of α-helical peptides and the TRIB3 protein were determined by SPR analyses. h Schematic diagram of full-length AKT1 and deletion mutants. However, the mechanisms by which SOX2 and other tumorinitiating cell markers are overexpressed in cancer remain unclear. Recent studies demonstrated that SOX2-expressing cells are the founding CSC population driving tumor initiation and therapy resistance 46,47 . SOX2 expression has been shown to positively correlate with the cancer cell stemness of solid tumors, including breast cancer, and knockdown of SOX2 decreases invasiveness and cancer cell stemness [48][49][50][51] . The expression of SOX2 and SOX9 is essential for the survival and metastasisinitiating properties of latency-competent cancer cells in multiple host tissues 52 . In the current study, TRIB3-promoted tumorsphere ability is reduced in SOX2-but not in c-MYC-or KLF4silenced BCCs. Moreover, SOX2 but not c-MYC, NANOG, or KLF4 is consistently upregulated by TRIB3 in several different types of BCCs. We proposed that SOX2 rather than other pluripotency factors plays a vital role in supporting the breast cancer stemness in breast cancer with high TRIB3 expression. Indeed, nuclear accumulation of FOXO1, but not that of other TFs, promotes SOX2 gene expression in breast cancer. FOXO1dependent genes are implicated in stem cell renewal, migration and invasion, differentiation, and oxidative stress 53,54 . FOXO1, as a recently recognized pluripotency factor, coordinates with SOX2, to support stemness 54 . Elevated FOXO1 usually degrades quickly through ubiquitination pathway after phosphorylation in physiological conditions. However, TRIB3-enhanced SOX2 can bind to the FOXO1 promoter, to further enhance FOXO1 gene expression. This positive feedback loop causes FOXO1 nuclear accumulation and promotes breast cancer stemness. Moreover, inhibition of S256 FOXO1 phosphorylation by TRIB3 is responsible for FOXO1 accumulation. Further studies need to clarify potential additional pathways for CSC regulation other than SOX2 in TRIB3-elevated BCCs. For example, it remains unclear whether c-MYC, NANOG, or KLF4 play a role in maintaining stemness in a certain type of BCCs by using isolated CSCs based on multiple parameters.
Current CSC therapy includes the inhibition of key CSC pathways such as the WNT and NOTCH pathways, CSC ablation using antibody-drug conjugates, epigenetic therapy, and quiescent CSC eradication 43 . However, recent evidence, especially from research on CSC plasticity, suggests a need for rethinking that the removal of resident CSCs can cure cancer 43 . The approach to specifically target the interaction between CSCs and their niche or microenvironment rather than pursuing therapies based on intrinsic CSC features is important for improving patient outcomes. In the current study, the activity of AKT1 was universally regulated by TRIB3 in BCCs. The fact that the TRIB3-AKT1 interaction prohibits the AKT1-FOXO1 interaction provides an opportunity to modulate the effects of the niche-enriched stress Cell extracts were prepared, and the levels of the indicated proteins were detected by immunoblotting. g Schematic diagram illustrates the elevated TRIB3 expression promoted BCSCs and breast cancer development. Data are shown as the mean ± SEM; P > 0.05 was considered not significant (NS); *P < 0.05, **P < 0.01, and ***P < 0.001 compared with Pep2-con. Source data are provided as a Source Data file.
protein TRIB3, which indeed shows enhanced expression adjacent to CD68 + TAMs in breast cancer patients. Several PPI modulators that inhibit the interactions among MDM2/p53, XIAP/caspase-9, and BCL2/beclin1 are being tested in clinical trials for cancer patients. The TRIB3/AKT1 interaction is reported to play roles in metabolic disease 55 . Here we demonstrated that Pep2-Ae, an α-helix from the catalytic region of AKT1, disrupts the TRIB3/AKT1 interaction and reduces the cancer initiation capacity of luminal and basal BCCs, and primary BCCs from transgenic mice and PDX mice via the activation of FOXO1 phosphorylation, ubiquitination, and degradation, which leads to the inhibition of SOX2 expression and breast cancer stemness. These results not only verify that the TRIB3-FOXO1-SOX2 signaling axis plays a key role in maintaining tumor initiation capacity, but also provide therapeutic options to target the interaction between the CSCs and their niche. Notably, we did not find that this AKT activation by TRIB3 blocking leads to tumor growth, which is consistent with the observation that AKT activation does not always lead to tumor proliferation 56 . However, further studies are needed to clarify the potential negative physiological signals for TRIB3-AKT-FOXO1-SOX2 loop and the exact role of AKT1 activation and upstream regulators such as PI3K by silencing TRIB3 expression. In summary, our study indicates that the elevated stress protein TRIB3 links stress signals to induce breast cancer initiation and progression by supporting breast cancer stemness, which is coordinated with elevated FOXO1 and SOX2, and triggered by activation of the TRIB3-AKT1-FOXO1-SOX2 axis in TRIB3overexpressed BCCs. Thus, this work provides a proof-of-concept for directly targeting BCSCs against breast cancer through inhibition of the TRIB3/AKT1 interaction.
Tissue microarray and immunohistochemistry. Formalin-fixed, paraffinembedded human breast cancer tissue and non-cancer tissue microarrays (BR1921c) were purchased from Alenabio (China). The human breast cancer microarray contained 80 cases of invasive ductal carcinoma, 80 cases of invasive lobular breast carcinoma, 24 cases of cancer adjacent normal breast tissue, and 7 cases of normal breast tissue.
Quantitative PCR. Total RNA was extracted using Trizol (TransGen Biotech, China) according to the manufacturer's instructions. cDNA synthesis was performed with 0.5~1 μg of total RNA at 42°C for 30 min by using TransScript II One-Step RT-PCR SuperMix (TransGen Biotech). mRNA levels were measured with gene-specific primers listed in Supplementary Table 4 by using the KAPA SYBR FAST qRT-PCR Kit (Kapa Biosystems, KK4601). The results were normalized to β-actin or GAPDH. The primers used of the indicated genes are shown in Supplementary Table 4.
Chromatin immunoprecipitation. EZ-Zyme chromatin preparation kits  and ChIP kits (17-371) were obtained from Merck Millipore (Billerica, MA, USA). Briefly, cells were cross-linked with 1% formaldehyde for 10 min at room temperature and redundant formaldehyde was inactivated by the addition of glycine 7 . The chromatin extracts containing DNA fragments were IP using anti-RNA Polymerase, normal mouse IgG, and FOXO1 or SOX2 antibody. The ChIPenriched DNA was then decrosslinked and analyzed by real-time PCR. The primers used to amplify specific regions of the indicated genes are shown in Supplementary  Table 4.
Coimmunoprecipitation. Cell pellets were lysed with coimmunoprecipitation buffer (25 mM Tri-cl (pH 7.4), 150 mM NaCl, 0.5% NP-40, 2.5 mM MgCl, 0.5 mM EDTA, 5% Glycerol), incubated with IP antibodies overnight at 4°C with shaking, and then incubated with Protein A/G Plus-Agarose (Santa Cruz Biotechnology) for 2 h at 4°C 15 . Interaction complexes were separated from the beads by boiling for 10 min and then subjected to SDS-PAGE and detected with immunoblotting.
Mass spectrometry analysis. Whole-cell lysates of MCF7 cells were IP with anti-FOXO1 antibody (CST, 14952) or anti-IgG1 isotype control antibody (CST, 5415) overnight at 4°C, then incubated with Protein A/G Plus-Agarose (Santa Cruz Biotechnology) for 2 hr at 4°C. Interaction complexes were separated from the beads by heating at 98°C for 10 minutes. Mass spectrometric data analysis was performed by Beijing Qinglian Biotech, Co., Ltd.
In vivo ubiquitination was carried out as described previously 63 . Briefly, HMLE or MCF7 cells were transfected with 6× His-ubiquitin and indicated vectors for 24 h. Cells were incubated with 20 μM MG132 for 6 h at 37°C before harvest cells. One-tenth of whole cells were lysed with RIPA buffer containing protease inhibitors (Roche). Remaining cells were lysed in urea buffer ( Luciferase reporter assay. HMLE or MCF7 cells were seeded 24 h before transfection. The FOXO1-GLuc (HPRM22487-PG04) and SOX2-GLuc (HPRM15202-PG04) vectors were obtained from GeneCopoeia, Inc. At 48-72 h post transfection, the cell culture medium was collected for GLuc and SEAP luminescent assays by using the Secrete-Pair Dual Luminescence Assay Kit according to the manufacturer's instruction (GeneCopoeia).
Electrophoretic mobility shift assay. Nuclear extracts of MCF7-SOX2-Myc cells were obtained by using NE-PER Nuclear and Cytoplasmic Extraction Reagents (Pierce). EMSAs were performed by using LightShift Chemiluminescent EMSA Kits (Pierce, 20148) as previously described 27 . Briefly, nuclear extracts were incubated with an anti-SOX2 antibody for supershift for 10 min at room temperaure (RT) in a reaction buffer. Then, biotin-labeled probes with or without unlabeled probes for the competitive reaction were added into the reaction system and incubated for 20 min at RT. Samples were separated in a 6% polyacrylamide gel and transferred onto a nylon membrane (Pierce), cross-linked by UV for 20 min, blocked with blocking buffer for 15 min, probed with streptavidin-HRP conjugates and incubated with the detection substrates. The probe sequences used for detection are shown in Supplementary Table 4.
Homology modeling of the TRIB3 protein. The amino-acid sequences of TRIB3 (NP_066981.2) were retrieved from the NCBI database and used as targets for homology modeling using Discovery Studio 2016 (BIOVIA). Create homology models modules were used to perform target-template sequence alignment after searching the putative X-ray template proteins in the NCBI server for generating the 3D models 64 . The structures of the top ten hits, including 5CEM, 5CEK, 4IXP, 4BL1, 4D2P, 4UMT, 4B6L, 2YZA, 2VN9, and 2WEL, were loaded and their sequences were aligned to the TRIB3 protein template sequences. The Build homology models module was then used to create the TRIB3 protein structure. The interactions between TRIB3 and AKT1 (PDB: 4GV1), peptide, and TRIB3 were predicted by the Dock Proteins (ZDOCK) module.
Animal studies. NOD-SCID mice (Vital River Lab Animal Technology, Beijing, China) and NPG (NOD-Prkdc scid Il2rg null ) mice (Beijing Vitalstar Biotechnology) were housed in maximum barrier facilities, with individually ventilated cages, sterilized food, and water. FVB mice (Vital River Lab Animal Technology, China), MMTV-PyMT transgenic mice and MMTV-ErbB2 transgenic mice (Model Animal Research Center of Nanjing University, Jackson Laboratory) were maintained in the animal facility at the Institute of Materia Medica under specific-pathogenfree conditions. All mice were used at 6-8 weeks of age. Female mice were hosted for all breast cancer models. All experiments using animals were performed following protocols approved by the Animal Experimentation Ethics Committee of the Chinese Academy of Medical Sciences, and all procedures were conducted following with the guidelines of the Institutional Animal Care and Use Committees of the Chinese Academy of Medical Sciences. All animal procedures were consistent with the ARRIVE guidelines 65 .
Tumor cells were suspended in PBS with 50% Matrigel (354230, Corning) on ice before xenograft. The indicated numbers of MCF7 or MDA-MB-231 cells were transplanted into the 4th mammary fat pads of 6-week-old female NOD/SCID mice. Tumorigenesis was analyzed at 4 or 6 weeks after transplantation. The day before MCF7 cell transplantation, mice were subcutaneously implanted with 17βestradiol control release pellets (SE-121, Innovative Research of America). The indicated numbers of primary mouse BCCs from MMTV-PyMT transgenic mice were injected into the 4 th mammary fat pads of 6-week-old female FVB mice. PDX MECs were isolated from NPG mice by using the Tumor Dissociation Kit human (Miltenyi Biotec, 130-095-929) and injected into the 4th mammary fat pad of 6-week-old female NPG mice. One day after xenograft, the mice were intraperitoneally treated with 2 mg/kg Pep2-con or 2 mg/kg Pep2-Ae twice a week for 4 or 5 weeks. Tumorigenesis was analyzed at 4 or 5 weeks after transplantation. The last peptide treatments were provided one hour before primary BCCs isolation. The cell lysates of the isolated primary BCCs were IP with anti-AKT (1:100, CST, 2938) antibody or probed with the indicated antibodies for protein expression analysis by Western blotting.
Human subjects. Human breast cancer tissues were obtained from Anyang Cancer Hospital, Henan University of Science and Technology. Primary breast cancer fragments were mechanically minced before implantation 66 . The clinical features of the patients are listed in Supplementary Table 3. All protocols using human specimens were approved by the Institutional Review Board of the Chinese Academy of Medical Sciences and Peking Union Medical College. Informed consent was obtained from all subjects. The study conforms to the principles outlined in the Declaration of Helsinki.
Generation of PDX animal models. Fresh breast cancer tissues were spliced into small fragments (1-3 mm 3 ) in the medium. The tissue fragments were suspended in diluted Matrigel (Corning, 354248) 1:1 with PBS, and subcutaneously implanted into NPG mice (NOD-Prkdc scid Il2rg null , Beijing Vitalstar Biotechnology). Early passages (1-5) of primary tumor tissues from these PDX models were mechanically minced and dissociated using gentleMACS TM Dissociator (Miltenyi Biotec) in accordance to the manufacturer's protocols 2 .
Mammosphere assays. For mammosphere assays, PDX cells were seeded on 96well ultralow attachment plates (Corning) at a density of 500 cells/well in StemXVivo Serum-Free Tumorsphere Media (R&D, CCM012) in the presence of the indicated peptides. Images were acquired by using a phase contrast microscope (Olympus Microsystems), and spheres were counted 5-7 days later. Mammosphere assays with MMTV-PyMT or MMTV-ErbB2 MECs were performed as described previously 67 . Single cells were plated in 24-well ultralow attachment plates (Corning) with sphere culture liquid medium. DMEM and Ham's F12 (50/50 Mix) media were used as the sphere culture liquid medium, and supplemented with B27 (Gibco), 20 ng/ml epidermal growth factor, 20 ng/ml basic fibroblast growth factor (Peprotech), and 4 mg/ml heparin. The spheres were counted 5-7 days later. MCF7 cells were cultured in 24-well ultralow attachment plates with liquid sphere culture medium for 5 days. The spheres were collected, and protein analysis was conducted after stimulation with 10 ng/ml IL-6 for 12 h, 50 mM glucose (HG) for 24 h, or 200 μM CoCl 2 for 24 h. Mammosphere assays with indicated cells were performed as previously described 68 . Resuspended cell solutions (10 4 cells/ml) were mixed with the same volume of Salmon fibrinogen (Sea Run Holdings, SEA-133), and the cell mixtures were seeded into 24-well plates that were pre-treated 5 μl of salmon thrombin (0.1 U/μl) (SEA-135, Sea Run Holdings). Thirty minutes after incubation at 37°C in a cell culture incubator, 1 ml of complete medium was then added. The tumor spheres were counted at 4-7 days later.
Isolated tumor MECs from MMTV-PyMT transgenic mice were co-cultured with sorted mouse TAMs to conduct a mammosphere assay. The mouse mammary TAM cells were spin infected with GFP-adenovirus for 2 h at 1000 × g at 4°C 69 . A total of 5000 MMTV-PyMT MECs were mixed with 20,000 TAMs and grown in low-adherence plates in sphere culture liquid medium 70 . Spheres were counted 5-7 days later. IL-6 expression in the co-cultured medium was analyzed by ELISA (LEGEND MAX™ Mouse IL-6 ELISA Kit, BioLegend, 431307) according to the manufacturer's instructions.