CC chemokines are differentially expressed in Breast Cancer and are associated with disparity in overall survival

Despite recent advances, breast cancer (BrCa) still affects many women and the impact is disproportional in African Americans (AA) compared to European Americans (EA). Addressing socioeconomic and behavioral status has not been enough to reduce disparity, suggesting contribution of biological differences in BrCa disparity. Our laboratory was first to show involvement of CC chemokines in BrCa. In this study, using ONCOMINE, TCGA, bc-GenExMiner and KMplotter, we examined the association of CC chemokines in BrCa outcomes and disparity. We show over-expression of CCL5, -7, -11, -17, -20, -22 and -25 in BrCa tissues. High mRNA levels of CCL7, -8, -17, -20 and -25 predicted a decrease in overall survival (OS). CCL7 and CCL8 were associated with decreased relapse-free survival. Expression of CCL17 and CCL25 was associated with decreased OS in AA. In EA, CCL8 was associated with decreased OS. Expression of CCL5, -7, -8, -17, -20 and -25 was highest in TNBC. Expression of CCL11 and CCL22 was associated with HER2. CCL7, -8, -17, -20 and -25 were elevated in AAs. In conclusion, our analysis suggests significant association of CC-chemokines in BrCa progression, OS and disparate disease outcome in AA compared to EA patients.

CC Chemokine expression affects BrCa patient prognosis. To elucidate the prognostic significance of all CC chemokines in BrCa in regards to metastatic relapse the Bc-GenExMiner-v4.1 online database was used (Table 3). Significance (p < 0.04) was seen in CCL4 (p = 0.004, HR = 0.91), CCL5 (p = 0.034, HR = 0.94), CCL8 (p = 0.0017, HR = 1.1), CCL19 (p = 0.001, HR = 0.9), CCL21 (p = < 0.0001, HR = 0.9), CCL22 (p = < 0.0001, HR = 0.9) and CCL23 (p = 0.0014, HR = 0.9). CCL8 was the only one significantly associated with an increase in metastatic relapse (HR > 1.0). Next, we determined the association of CC chemokines, which were higher in  Table 1. Significant changes of CC-Chemokines among BrCa tissues in comparison to normal tissues using ONCOMINE. Fold Change denotes the folds difference in CC chemokine mRNA expression in BrCa in comparison to normal tissues. The p value represents the statistical significance of the difference in the mean of the two groups as analyzed using t-test. Bold represents significant RNA expression with a cut off of: fold change ≥1.4 and p-value ≤ 0.04 using t-test.

Discussion
Chemokines are predominantly known for their roles in immune cell trafficking and shaping the immune system. The role of chemokines and their corresponding receptors are well appreciated in cancer progression and metastasis 28 . Among all known chemokines, CXC chemokines are highly appreciated in cancer compared to CC chemokines. However, the CC chemokines are the largest family of chemokines and play an important role in inflammation 29 . Our laboratory was first to show association of CC chemokine receptor-9 in prostate, breast, ovarian and lung cancer 10,18-20 .    www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ microenvironment and supporting BrCa progression. In addition, production of CCL5 from mesenchymal stem cells (MSCs) results in BrCa production of colony-stimulating factor 1 (CSF1) under hypoxic conditions 32 . This causes the recruitment of myeloid-derived suppressor cells (MDSC) and tumor associated macrophages (TAM) to the tumor microenvironment 32 . Studies have shown that higher CCL5 promotes BrCa metastasis 33 . Our analysis also shows higher levels of CCL5 in medullary and ductal carcinoma and in TNBC compared to non-TNBC cases. Higher CCL5 may be involved in developing tumor tolerance resulting in the poor TNBC prognosis. As observed in our analysis, others have shown higher CCL5 in TNBC 34 . Non-remissive and later stage BrCa was reported to be correlated with CCL5 expression 35,36 , possibly due to its ability to promote pro-invasive factor MMP9 and monocyte migration to the BrCa tumor site, in which they undergo polarization allowing them to support tumor progression through angiogenesis 37 . This is not to our surprise as CCL5 was elevated in AA patients who often develop aggressive forms of BrCa. Furthermore, antagonizing CCR5 decreases CCL5 induced angiogenesis of BrCa cells 38 . CCL5 has been shown to promote BrCa progression in a p53 dependent manner through CCR5 39 . Interestingly, endothelial cells have been shown to increase metastasis of TNBC cells through secretion of PAI-1 and CCL5 40 . Hence our analysis showing higher CCL5 in BrCa (medullary and ductal carcinoma) compared to normal, higher CCL5 in TNBC compared to non-TNBC and higher CCL5 in AA compared to EA, suggests its potential contribution in shaping the tumor favoring microenvironment.
Our analysis also shows higher CCL7 in ductal breast carcinoma compared to normal tissues and other histological tissue types. It was also elevated in AA in comparison to EA patients. CCL7, that can bind and activate CCR1, CCR2 and CCR3, has been shown to promote metastasis by activating the MAPK cascade, promoting epithelial-mesenchymal transition (EMT) and CCR2+ TAM recruitment (enhancing vascular permeability 41 ). Cancer associated fibroblast (CAF) derived CCL7 promotes BrCa proliferation 42 . Interestingly, CCL8 also known as monocyte chemo-attractant protein-2 (MCP-2), a natural ligand shared by CCR2, CCR3 and CCR5 was not elevated in BrCa tissues when compared to normal tissues, but was associated with poorer OS and RFS. It has been shown to drive BrCa metastasis. More specifically CCL8 stimulates fibroblasts generating a pro-tumor environment in the TNBC stroma, which was not seen in non-TNBC 43 . This was not to our surprise, as we found significantly higher CCL8 in TNBC tissues compared to non-TNBC tissues and it was associated with increased metastatic relapse. More importantly, when compared between AA and EA women, CCL8 was found higher in AA patients, who are frequently diagnosed with TNBC compared to EA. In contrast to its classical contribution on biology and outcome of BrCa, CCL8 was associated with poor OS in EA, suggesting race specific differences in CCL8 biology and immunity against BrCa.
Using tissue data we found that CCL11 (Eotaxin), which has a high affinity for CCR3, was higher in ductal breast carcinoma tissues when compared to healthy controls and other histological tissue types. Furthermore, a supporting study found that CCL11 was higher in the serum of BrCa patients when compared to serum of healthy individuals 44 . Bone colonization of BrCa cells is promoted by CCL11 45 , this is important to note as nearly 70% of BrCa cases show bone metastasis 46 . Co-culture studies showed an increase in CCL11 secretion by fibroblasts that enhanced chemoresistance and metastasis of BrCa cells 47 . In an allergic inflammation study, CCR3 antagonism prohibited chemotaxis of basophils and eosinophils 48 . Eosinophil degranulation has been reported in BrCa 49 , however this anti-tumor response is not quite understood. In pancreatic cancer basophil recruitment into tumor draining lymph nodes correlates with inflammation and poor survival 50 . Moreover, in hypoxic conditions tumor cells have been shown to secrete CCL11, recruiting CD206 expressing macrophages to the tumor, which in turn polarizes the classical macrophages to a M2 pro-angiogenic phenotype 51 . Clearly these facts and our data emphasize the need to study CCR3-CCL11 axis with respect to BrCa.
Furthermore CCR4 (receptor for CCL17 and CCL22) antagonists have been shown to decrease the tumor-promoting environment 52 . CCL17 recruits CCR4 positive regulatory T cells (Tregs) and promotes lung metastasis of BrCa by elimination of NK cells 53,54 . This coincided with our findings that higher CCL17 is expressed in BrCa and associated with decreased OS. Our data also shows CCL17 is overexpressed and is associated with poor prognosis in AA patients unlike EA. Its expression is higher in TNBC patients compared to non-TNBC patients. Moreover, CCR4 positively correlates with HER2 expression is BrCa cells 54 , supporting our result of higher expression of CCL22 in HER2 expressing tissues, when compared to basal and luminal types. We also found CCL22 to be significantly higher in BrCa (lobular) tissues compared to normal tissues, which agrees with a study reporting circulating CCL22 levels to be significantly higher in BrCa patients when compared to healthy controls 55 . As a prognostic factor in BrCa, tumor derived CCL22 has also been shown to activate and recruit CCR4 expressing Tregs 56,57 . Moreover, in prostate cancer, TAM promote tumor cell migration by activating CCL22-CCR4 signaling 58 . Additionally this axis has been shown to promote bone metastasis of lung cancer 59 . In contrast, our results showed that high CCL22 expression was associated with increased OS and RFS in BrCa.
The ligand for CCR6, CCL20 was higher in ductal and medullary breast carcinoma when compared to normal controls and other histological tissue types. Moreover, BrCa cells secrete CCL20, which recruits CCR6 expressing immune cells in the tumor vicinity 60 . Tumor cell produced CCL20/MIP3α allows them to attract CCR6 positive dendritic cell precursors 61 . Furthermore, CCL20 promotes migration and proliferation of surrounding breast cells 62 and promotes BrCa initiation 63 . This could explain our finding showing higher expression of CCL20 in earlier stage I and stage II. Through paracrine signaling CCL20 promotes EMT in BrCa cells 64 . Studies have shown involvement of CCL20 in bone 65 and lung 8 metastasis of BrCa. We saw that higher expression of CCL20 associates with decreased OS. Moreover, expression of CCL20 was higher in AA with BrCa when compared to EA and in TNBC when compared to non-TNBC status. Chemotherapy has been shown to trigger higher production of CCL20 which in turn contributes to chemoresistance through ABCB1 drug efflux pump 66 , further suggesting involvement of CCL20 in disparities in therapeutic outcome.
CCR9 is selectively expressed on T cells and it's only natural ligand (CCL25) is expressed in the thymus and small intestines 67 . CCR9 is often expressed on immature double (CD4+ and CD8+) positive thymocytes 68,69 , but has been reported to be expressed on thymocytes of all stages 70 . Circulating CCR9/CD4+ T cells exhibit a www.nature.com/scientificreports www.nature.com/scientificreports/ Th1 cytokine profile 71 . We found that CCL25 expression was associated with increased RFS when looking at all BrCa patients. We previously reported that CCL25 promotes migration/invasion 20 and chemo-resistance 21 in BrCa. Another study confirmed its role in BrCa and liver cancer migration and invasion 72 . Our current analysis further confirms our previous studies showing higher expression of CCL25 in BrCa compared to normal controls. Additionally, our data shows higher levels of CCL25 in TNBC compared to non-TNBC status, AA women compared to EA women and that it is associated with decreased OS in AA. Additionally, there was a trend of increased expression of CCL25 from stage I to stage III. Hence our current analysis explains a potential role of CCR9-CCL25 in disparities associated with poor overall outcome of BrCa in AA women.
It has been shown that chemokine receptors alone do not play a role in disease free or OS in BrCa 8 (which we confirmed using Kmplotter, data not shown), suggesting the importance of their ligands. To our knowledge, this manuscript is the first to consolidate patient tissue data on CC chemokine ligands in BrCa and show that CC chemokine ligands are associated with disease free and OS. This suggests that chemokines expressed in the BrCa tissues not only support cancer cells expressing their corresponding receptors, but also shapes tumor microenvironment to overcome immune attack. However, these results should be confirmed and used with caution due to the fact that there are many factors downstream that could alter the effects of the variations seen at the mRNA level. In conclusion, identifying the role of chemokines in BrCa patients showing prominent expression could provide patient or race specific immunotherapeutic targets.

Materials and Methods
Oncomine. ONCOMINE 73 , online database (www.oncomine.org) of RNA/DNA sequencing data was used to screen all 24-CC chemokines' RNA expression in BrCa tissues in comparison to normal tissues using the TCGA dataset (Table 1). Significant overexpression in BrCa tissues compared to normal tissues was determined using a t-test, utilizing a cut off of: fold change ≥1.4 and p-value ≤ 0.04. Additionally, these cut offs were used to classify significance for all downstream analysis. Further elucidation of histological type of chemokines significantly expressed in BrCa (compared to normal tissues) was conducted with ONCOMINE utilizing all available datasets ( Table 2). Using the referenced cutoffs, t-test were conducted comparing BrCa histological tissue types within each dataset.
Bc-GenExMiner-v4.1. Microarray DNA expression results from BrCa Gene-Expression Miner (bc-GenExMiner) 74 were used to classify prognostic association of metastatic relapse in BrCa of all 24 CC chemokines using univariate cox analysis (Table 3).
Kaplan Meier Plotter-Breast Cancer. Association of CC chemokines on overall (Fig. 1) and relapse free ( Fig. 2) survival was determined using Kaplan Meier (KM) Plotter (http://kmplot.com/analysis/) 75,76 . The two groups for plotting were high (red line) or low (black line) based on median mRNA gene chip expression. cBioPortal and Analysis. The cBio Cancer Genomics Portal (http://cbioportal.org) 77 open-access database was used to download The Cancer Genome Atlas (TCGA) Pan Cancer Atlas dataset (RNASeq V2 RSEM) of 1084 BrCa patients (normalization and batch correction description here: https://www.synapse.org/#!Synapse:syn4976363). Graphpad Prism was used to produce Kaplan-Meir OS curves based on race (AA (Fig. 3a) or EA (Fig. 3b)), whereas red represents high expression and black low expression based on expression median. For AA vs. EA (Fig. 3c) survival curves red represents high expression in AA, black is high expression in EA, blue is low expression in AA and gray is low expression in EA. Log-rank (Mantel-Cox) test was used to determine p-values and hazard ratio. Expression based on clinical parameters was graphed as Log 10 axis scale using box and whisker plots with whiskers at the minimum and maximum values and the middle line as the median (Fig. 4). Statistical analysis was conducted with non-parametric Mann-Whitney for t-test and global significance for ANOVA with Kruskal Wallis (Graphpad Prism7, La Jolla, CA). All statistics were thoroughly confirmed and verified by a statistician using SAS.

Data Availability
All data generated or analyzed during this study are included in this published article.