Serum amyloid A predisposes inflammatory tumor microenvironment in triple negative breast cancer

Acute-phase proteins (APPs) are associated with a variety of disorders such as infection, inflammatory diseases, and cancers. The signature profile of APPs in breast cancer (BC) is poorly understood. Here, we identified serum amyloid A (SAA) for proinflammatory predisposition in BC through the signature profiles of APPs, interleukin (IL) and tumor necrosis factor (TNF) superfamily using publicly available datasets of tumor samples and cell lines. Triple-negative breast cancer (TNBC) subtype highly expressed SAA1/2 compared to HER2, luminal A (LA) and luminal B (LB) subtypes. IL1A, IL1B, IL8/CXCL8, IL32 and IL27RA in IL superfamily and CD70, TNFSF9 and TNFRSF21 in TNF superfamily were highly expressed in TNBC compared to other subtypes. SAA is restrictedly regulated by nuclear factor (NF)-κB and IL-1β, an NF-κB activator highly expressed in TNBC, increased the promoter activity of SAA1 in human TNBC MDA-MB231 cells. Interestingly, two κB-sites contained in SAA1 promoter were involved, and the proximal region (−96/−87) was more critical than the distal site (−288/−279) in regulating IL-1β-induced SAA1. Among the SAA receptors, TLR1 and TLR2 were highly expressed in TNBC. Cu-CPT22, TLR1/2 antagonist, abrogated IL-1β-induced SAA1 promoter activity. In addition, SAA1 induced IL8/CXCL8 promoter activity, which was partially reduced by Cu-CPT22. Notably, SAA1/2, TLR2 and IL8/CXCL8 were associated with a poor overall survival in mesenchymal-like TNBC. Taken together, IL-1-induced SAA via NF-κB-mediated signaling could potentiate an inflammatory burden, leading to cancer progression and high mortality in TNBC patients.


INTRODUCTION
Serum am4yloid A (SAA), an acute phase protein (APP), is mainly produced in the liver, while its extrahepatic synthesis has been reported in skin [1], in atherosclerotic lesions [2], in synovial tissues [3], in adipocytes and smooth muscle cells [4,5]. The human SAA acute phase protein family contains four different isoforms namely, SAA1, SAA2, SAA3 and SAA4. Both SAA1 and SAA2 show high degree of homology in the mRNA and the protein sequences which are hard to be identified [6]. SAA3 is a pseudogene categorized as a non-coding RNA [7], while SAA4 is constitutively expressed as a non-inducible protein [8,9]. On the other hand, the synthesis of SAA1 and SAA2 is inducible under inflammatory conditions [9], such as inflammation, trauma and infection, increasing to several hundredfold [10,11].
BC is the most frequently diagnosed cancer and the second most common cause of cancer deaths in women in the US [34]. One probable reason for this is that we lack a complete picture of the biologic heterogeneity of this disease. BC is divided into 4 intrinsic subtypes with a molecular basis as follows: luminal A (LA), luminal B (LB), human epidermal growth factor receptor 2 (HER2)enriched, and basal-like (BL) BC [35][36][37][38]. Triple negative breast cancer (TNBC), lacking estrogen receptor (ER), progesterone receptor (PR) and HER2, is known to be the most heterogeneous and comprises largely of the basallike subtype [39]. LA-BC is characterized by ER or PR positive but HER2 negative, while LB-BC presents with ER or PR and HER2 positive [40]. The HER2-enriched subtype is characterized by HER2 positive, both ER and PR negative, and high expression of proliferation-related genes [40]. The BL subtype includes plenty of TNBC and high expression of EGFR and proliferation-related genes [36,41]. Furthermore, TNBC, the most heterogeneous BC, were defined into several subtypes namely, basal-like 1 and 2 (BL1, BL2), mesenchymal-like (ML) and luminal androgen (LAR) [42]. BL-TNBC subtype is characterized by highly activated cell cycle and DNA [43]. ML-TNBC is associated with poor prognosis due to enhanced epithelialto-mesenchymal transition (EMT) [43] and elevated expression of genes involved in growth factor pathways [42,44]. LAR-TNBC expresses differentially the estrogen/ androgen metabolism pathways and is driven by androgen signaling [43,45]. BL-BC representing TNBC is of particular interest because of aggressiveness, early pattern of metastasis, greater size of tumor, and lack of welldefined therapeutic target sites due to ER-/PR-/HER2negative status [39,46]. Most patients with TNBC have experienced higher rate of distant recurrence compared to patients with other BC subtypes, requiring identification of molecular drivers for TNBC.
To date, the signature profile of APPs in BC is not well defined. Here, we analyzed the signature profiles of APPs, interleukin (IL) family and TNF superfamily using publicly available datasets of breast tumor samples and cell lines. Based on the results analyzed, we found SAA as a main factor for proinflammatory predisposition in TNBC and proposed IL-1-induced SAA via NF-κB-mediated signaling as a molecular driver for the aggressiveness of TNBC.

SAA1 and SAA2 APPs are predominantly expressed in human BL-BC subtype and TNBC cells
We investigated the APP signature in human breast tumor tissues and cell lines. We used The Human Cancer Genome Atlas (TCGA)-based dataset for human BC tissues to define the APP signature in breast tumor heterogeneity. Analysis of TCGA-based dataset by Gitools 2.3.1 revealed the following dominant APP signature: BL-BC subtype representing TNBC highly expressed SAA1, SAA2, SAA4 and TF. Both BL and HER2 subtypes highly expressed ORM1 and CP ( Figure 1A and 1B), while LA subtype highly expressed SERPINA1 and SERPINA2 ( Figure 1A and 1B). In addition, the analysis of the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) dataset on 51 human BC cell lines revealed the following signature of APP: SAA1/2 was highly expressed in both BL-and ML-TNBC; C3 and FN1 were predominantly expressed in BL-and ML-TNBC cells, respectively ( Figure 1C and Supplementary Figure 1). Based on the intersection of APP signature between human BC tissues and cell lines ( Figure 1D) to exclude the tumor heterogeneity, SAA1 and SAA2 were highly expressed both TNBC tissues and cell lines ( Figure 1D).

IL1A, IL1B, IL8/CXCL8 and IL32 are highly expressed in human BL-BC subtype and TNBC cells
We analyzed the IL superfamily signature in human breast tumor tissues and cell lines. We also used TCGAbased dataset for human BC tissues and NCBI-GEO dataset for 51 human BC cell lines. Particularly, BL-BC subtype representing TNBC highly expressed IL1A, IL1B, IL23A, IL32 and IL34 (Figure 2A and 2B). Both BL-and HER2-BC subtypes predominantly expressed IL7 and IL8, while LA-BC subtype highly expressed IL33 (Figure 2A and 2B). In addition, both BL-and ML-TNBC human cell lines highly expressed IL1A, IL1B, IL6, IL8/CXCL8 and IL32, BL-TNBC cell lines highly expressed IL18, and LA cell lines dominantly expressed IL20 ( Figure 2C and Supplementary Figure 2). Based on the intersection of interleukin superfamily signature between human BC tissues and cell lines ( Figure 2D) to exclude the tumor heterogeneity, both TNBC tissues and cell lines dominantly expressed IL1A, IL1B, IL8, and IL32 ( Figure 2D).

IL27RA is predominantly expressed in human BL-BC subtype and TNBC cells
We also assessed the IL receptor superfamily signature in human breast tumor tissues and cell lines.

CD70 and TNFSF9 are predominantly expressed in human BL-BC subtype and TNBC cells
We checked the signature of TNF superfamily in human breast tumor tissues and cells lines. According to the analysis on TCGA-based dataset for human BC tissues, CD70 and TNF are highly expressed in BL-BC representing TNBC, while TNFSF12 is highly expressed in LA-BC subtype ( Figure 4A and 4B). Both BL-and HER2-BC subtypes highly expressed FASLG, LTA and LTB ( Figure 4A and 4B). Moreover, BL-and ML-TNBC human cell lines highly expressed CD70 and TNFSF9 ( Figure 4C and Supplementary Figure 3). Based on the intersection of TNF superfamily signature between human BC tissues and cell lines ( Figure 4D) to exclude the tumor heterogeneity, both TNBC tissues and cell lines dominantly expressed CD70 and TNFSF9 ( Figure 4D).

TNFRSF21 is mainly expressed in human BL-BC subtype and TNBC cells
We further assessed the TNF receptor superfamily signature in human breast tumor tissues and cells lines. Analysis on TCGA-based dataset for human BC tissues showed that BL-BC subtype representing TNBC highly expressed FAS, LTBR, TNFRSF10B, TNFRSF10D, TNFRSF11A, TNFRSF13B, TNFRSF21, and TNFRSF25 ( Figure 5A and 5B). LA-BC subtype highly expressed Oncotarget 514 www.oncotarget.com NGFR, TNFRSF10C, TNFRSF14 and TNFRSF19 ( Figure 5A and 5B). Both BL-and HER2-BC subtypes highly expressed TNFRSF4, TNFRSF8, TNFRSF9, TNFRSF10A, TNFRSF12A and TNFRSF17 ( Figure 5A and 5B). However, our analysis of NCBI GEO dataset on 51 human BC cell lines showed that only TNFRSF21 is highly expressed in BL-and ML-TNBC cells ( Figure 5C and Supplementary Figure 3). Based on the intersection of TNF receptor superfamily signature between human BC tissues and cell lines ( Figure 5D) to exclude the tumor heterogeneity, both TNBC tissues and cell lines dominantly expressed TNFRSF21 ( Figure 5D).

TLR1 and TLR2 are predominantly expressed in human BL-BC subtype and TNBC cells
We investigated the SAA receptor and TLR family signature in human breast tumor tissues and cell lines.
SAA has multiple receptors, including the FPR2, the TLRs TLR2 and TLR4, the scavenger receptor SR-BI, and the ATP receptor P2X7 [47]. Analysis on TCGA-based dataset for human BC tissues revealed that the BL-BC subtype representing TNBC highly expressed SCARB1, TLR1, TLR2, and TLR6, while LA-BC subtype dominantly expressed TLR3 ( Figure 6A and 6B). Both BL and HER2subtypes highly expressed TLR8 and TLR9 ( Figure 6A and 6B). In addition, the analysis of NCBI GEO dataset on human BC cell lines revealed that TLR1 and TLR3 were highly expressed in BL-TNBC, while TLR2 was highly expressed in BL-and ML-TNBC and HER2 subtype ( Figure 6C and Supplementary Figure 3). Based on the intersection of SAA receptors and TLR superfamily signature between human BC tissues and cell lines ( Figure  6D) to exclude the tumor heterogeneity, TLR1 and TLR2 were highly expressed in both TNBC tissues and cell lines ( Figure 6D).

DISCUSSION
A main finding in this study is that interaction of SAA and proinflammatory cytokines is potentiated in TNBC compared to other BC subtypes, probably Oncotarget 517 www.oncotarget.com leading to inflammatory tumor environment followed by cancer progression and a high mortality. Among expression profiles of APPs, SAA1/2 is emerging as a novel biomarker in TNBC. SAA is an acute-phase protein known to mediate proinflammatory response and generated primarily in the liver. Serum of healthy donors expresses SAA at relatively low levels [49]. SAA protein levels were higher in patients with ER-negative breast tumors compared to those with ER-positive [50], supporting our findings of higher levels of SAA1/2 in TNBC (Figure 1 and Supplementary Figure 1). High levels of SAA were related to survival time of patients less than one year with breast invasive ductal carcinoma as a useful marker in BC recurrence [51]. Elevated SAA was associated with reduced OS in BC [52]. On the other hand, SAA was linked to poor recurrence-free survival in BC but not OS [53]. Our results indicate that SAA is associated with poor OS in ML-TNBC subtype ( Figure 8C) but not all of BC. Interestingly, elevated SAA in tumor-associated macrophage and breast tumor cells was associated with both lymphovascular invasion and lymph node metastasis [53]. In a BC mouse model, ectopic expression of SAA1 or SAA3 in tumor cells potently promoted widespread metastasis [54]. SAA in stages II, III and IV BC patients had a higher value compared to those of the healthy, benign and stage I groups. Also, BC patients with lymph node metastasis or distant metastasis were found to have significantly higher SAA levels [55]. These findings indicate a metastatic effect of SAA in BC, probably contributing to aggressiveness of TNBC which highly expresses SAA (Figure 1 and Supplementary  Figure 1). The expression of SAA is restrictedly regulated by proinflammatory cytokines, such as IL-1, IL-6 and TNF [21][22][23]56]. Because IL1A, IL1B, IL8/CXCL8, and IL32 are highly expressed in TNBC (Figure 2 and Supplementary Figure 2), these cytokines can be directly Oncotarget 518 www.oncotarget.com involved in upregulating SAA in TNBC. TNBC tumor irradiation in a mouse model significantly increased the plasma level of IL-1β, which was associated with lung metastases [57]. The basal levels of IL-1β were higher in TNBC cells compared with non-TNBC cells and IL-1β increased invasiveness of TNBC cells [58]. IL-1β also significantly increases IL-8/CXCL8 in TNBC cells [59]. IL-1β was found to induce SAA1 promoter activity in an NF-κB-dependent manner in TNBC cells (Figure 7) and SAA was able to induce IL8/CXCL8 promoter activity ( Figure 8B), indicating that interaction of SAA and proinflammatory cytokines could enhance inflammatory burden in TNBC. IL-8/CXCL8 is a blood biomarker for TNBC, significantly increased in TNBC cells compared to non-TNBC, and increase the invasiveness and growth of TNBC cells [60][61][62]. IL-8/CXCL8 was associated with short disease-free survival and OS in TNBC [60,63], supporting poor OS to high levels of IL8 in ML-TNBC subtype ( Figure 8C). IL-32 increased migration and invasion capacities of TNBC cells [64] and promoted TNBC cell growth [65]. In addition, IL-32 is a potential immunotherapy target antigen in HLA-A2-positive TNBC [66]. Based on these results, a high level of IL32 in TNBC (Figure 2) may be involved in aggressiveness of TNBC despite unclear direct relationship between SAA and IL32. Although IL27RA is found to be predominantly expressed in BL-TNBC (Figure 3 and Supplementary  Figure 2), a functional role of IL27RA in BC is poorly understood at this point. CD70 expression levels were significantly higher in BL-BC compared to LA [67], supporting our findings of high levels of CD70 in TNBC (Figure 4 and Supplementary Figure 3). A functional role of CD70 in BC is poorly understood at this point. TNFSF9 (4-1BB/CD137) antibody favored the propagation of CD8+ tumor-infiltrating lymphocytes (TILs) from TNBC tumors, being capable of cytotoxic functions [68]. This indicates that high levels of TNFSF9 in TNBC (Figure 4 and Supplementary Figure 3) might inhibit CD8+TILs to reduce cytotoxicity, leading to enhanced aggressiveness of TNBC. TNFRSF21 (DR6) level is known to be increased in a grade-dependent manner in BC [69], although the functional role of TNFRSF21 in TNBC is still unclear.
Based on the correlations of APPs with both dominant IL and TNF superfamily in TNBC found by the present studies and other studies, the interrelationship between SAA, IL-1 and IL8/CXCL8 appears to be a main driver in aggressiveness of TNBC. IL-1β-induced SAA1 promoter activity is critically involved in NF-κB-mediated Oncotarget 519 www.oncotarget.com pathway ( Figure 7A and 7B) which upregulates IL8/ CXCL8 [48]. IL-8/CXCL8 secretion by SAA correlates with NF-κB activation through FPRL1/LXA4R, a G protein-coupled receptor, in neutrophils [70]. SAA induced IL-8 in monocytes and dendritic cells [71] and peripheral blood mononuclear cells (PBMCs) [72]. SAA1 induced IL-8/CXCL8 via TLR4-mediated NF-κB signaling in human bone marrow-derived mesenchymal stem cells (hMSCs) [73]. In addition, SAA released TNF, IL-1β and IL-8 in neutrophils [16]. Although TLR2 among multiple receptors of SAA cells are highly expressed in TNBC (Figure 6), the interrelation between TLR2 and TNBC is poorly understood. We demonstrate that IL-1β-induced SAA1 and SAA1-induced CXCL8 promoter activities are partially mediated by TLR2 signaling in MDA-MB231 cells ( Figure 8A and 8B). Other studies have also shown that SAA could activate NF-κB via TLR2-mediated pathway in HeLa cells and mouse macrophages [72,73]. TLR2 is expressed in normal mammary epithelia and inhibition of TLR2 reduces growth of human BC cells [74]. Expression of TLR2 was increased in circulating tumor cell-positive patients [75]. TLR2 was highly expressed by two-fold in mammary cancer stem cells and inhibition of TLR2 signaling impaired in vitro mammosphere generation in MDA-MB-231 cells [76]. TLR2 mediates invasion by activating NF-κB in MDA- TNBC cells overnight, a luciferase promoter activity assay was performed at post-treatment of IL-1β (10 ng/ml) for 6 h. (C) Effects of IL-1β on luciferase activity in NF-κB mutated constructs of the SAA1 promoter. Site-directed mutants were generated from the SAA1-P319LUC using primers with mutant κB-like sites (-287/-278) and κB-consensus site (-95/-86). After transfection with SAA1-P319LUC and its mutant κB-site luciferase vectors in MDA-MB231 TNBC cells overnight, a luciferase promoter activity assay was performed at post-treatment of IL-1β (10 ng/ml) for 6 h. Results were normalized to the protein level and expressed as a fold increase compared to non-treated control. Gray circles indicate κB site mutants. * , ** indicate significant (p < 0.05) increase compared to each control, when a Student's-t test was analyzed. Also, significant (p < 0.05) change exists between * and ** groups. Representative results are shown from triplicated experiments.
In conclusion, IL-1-induced SAA could predispose proinflammatory tumor microenvironment in TNBC, leading to aggressiveness of TNBC followed by a higher mortality.

Data analysis from GEO and the TCGA datasets
Data analysis was performed using publicly available microarray data sets deposited in NCBI-GEO (http://www.ncbi.nlm.nih.gov/geo/) database under accession number GSE12777. Raw microarray data for APPs, IL family and TNF superfamily were RNA expression levels prepared from 51 human breast cancer cell lines. The basal acute-phase proteins, cytokines and their receptors expression levels were determined by global gene expression profiling of BC cell lines, while molecular subtyping was determined using gene expression and HER2 status by fluorescent in situ hybridization. We utilized Gitools 2.3.1 (http://www. gitools.org) based on Oracle Java 7, an open-source tool to perform Genomic Data Analysis and Visualization as interactive heat-maps [86]. Breast invasive carcinoma dataset for TCGA individual projects was used for BC subtypes as follows: n = 140 for BL-BC, n = 67 for HER2-BC, n = 419 for LA-BC and n = 192 for LB-BC patients (http://www.gitools.org/datasets/tcga).
We found two κB sites in SAA1 promoter and generated mutant constructs by mutating each κB site. Primers for mutation of κB-consensus site (lowercase) were designed as follows: 5′-TGA CCT GCA aGG ACT TTC tCC AGG GAC CAC-3′ for -104/-74 and κBlike site; 5′-TGC CGC CAT CAC aGG GCT CCt ACT CTC AAC-3′ for -299/-269. The mutation of κB sites was performed by PCR-based mutagenesis using a sitedirected mutagenesis kit according to manufacturer's instructions (Stratagene, La Jolla, CA). The mutant constructs of the SAA1 promoter were confirmed by DNA sequencing analysis. Human CXCL8 promoter (-322/+10) was generated as described previously [48].

Transient transfection and luciferase assay
TNBC MDA-MB231 cells at approximately 50% confluency in 24-well plates were washed once with fresh media without additives and then transiently transfected with SAA1 constructs and κB-site mutants for 24 h at 37°C using Lipofectamine solution. Transfected cells were treated as outlined in Results and incubated for 6 h. After rinsing cells with cold 1X PBS and adding lysis buffer (Promega, Madison, WI), cell lysates mixed with Bright-Glo TM Assay Reagent (Promega, Madison, WI) were used for determination of luciferase activity using a microplate luminometer. Luciferase activity, expressed as relative light units, was normalized to measured protein levels.

Overall survival analysis
Kaplan-Meier plotter database was utilized to assess overall survival (OS) using proportional hazards regression to estimate Hazard Ratio (HR) and 95% Confidence Interval (CI) with auto computed cutoff value based on the gene expression levels of SAA1/2, TLR2 and IL8/CXCL8 from 73 ML-TNBC patients downloaded from GEO (Affymetrix HGU133A and HGU133+2 microarrays) [87].

Statistics
Data were analyzed by coefficient of determination (R 2 ), the paired Student's t-test and one-way analysis of variance (ANOVA) as appropriate. If statistical significance (p ≤ 0.05) was determined by ANOVA, the data were further analyzed by Tukey's pairwise comparisons to detect specific differences between treatments.