CmP signaling network unveils novel biomarkers for triple negative breast cancer in African American women

Abstract

Breast cancer is the most diagnosed cancer worldwide and remains the second leading cause of cancer death. While breast cancer mortality has steadily declined over the past decades through medical advances, an alarming disparity in breast cancer mortality has emerged between African American women (AAW) and Caucasian American women (CAW). New evidence suggests more aggressive behavior of triple-negative breast cancer (TNBC) in AAW may contribute to racial differences in tumor biology and mortality. Progesterone (PRG) can exert its cellular effects through either its classic, non-classic, or combined responses through binding to either classic nuclear PRG receptors (nPRs) or non-classic membrane PRG receptors (mPRs), warranting both pathways equally important in PRG-mediated signaling. In our previous report, we demonstrated that the CCM signaling complex (CSC) consisting of CCM1, CCM2, and CCM3 can couple both nPRs and mPRs signaling cascades to form a CSC-mPRs-PRG-nPRs (CmPn) signaling network in nPR positive( $+$ ) breast cancer cells. In this report, we furthered our research by establishing the CSC-mPRs-PRG (CmP) signaling network in nPR( $-$ ) breast cancer cells, demonstrating that a common core mechanism exists, regardless of nPR( $+/-$ ) status. This is the first report stating that inducible expression patterns exist between CCMs and major mPRs in TNBC cells. Furthermore, we firstly show mPRs in TNBC cells are localized in the nucleus and participate in nucleocytoplasmic shuttling in a coordinately synchronized fashion with CCMs under steroid actions, following the same cellular distribution as other well-defined steroid hormone receptors. Finally, for the first time, we deconvoluted the CmP signalosome by using systems biology and TNBC clinical data, which helped us understand key factors within the CmP network and identify 6 specific biomarkers with potential clinical applications associated with AAW-TNBC tumorigenesis. These novel biomarkers could have immediate clinical implications to dramatically improve health disparities among AAW-TNBCs.



Abbreviations

TNBC	Triple-Negative Breast Cancer
CSC	CCM signaling complex

PRG	progesterone
MIF	Mifepristone
mPRs/PAQRs	membrane Progesterone Receptors
	or Progestin and AdipoQ Receptors
AAW	African American Women
CAW	Caucasian American Women
CmPn	CSC-mPRs-PRG-nPRs
CmP	CSC-mPRs-PRG
DEGs	Differentially Expressed Genes
DEPs	Differentially Expressed Proteins

1. Introduction

Breast cancer is the most commonly diagnosed cancer worldwide and remains the second leading cause of cancer death in the United States [1]. While breast cancer mortality has steadily declined over the past three decades through medical advances [2], an alarming Black-White disparity in breast cancer mortality has emerged in the United States. Compared to Caucasian American Women (CAW) with breast cancer, African American Women (AAW) are less likely to be diagnosed at earlier stages [3], face more aggressive forms of the disease [4], and experience a higher mortality rate [3]. Additionally, despite the lower lifetime incidence of breast cancer, AAW diagnosed with breast cancer are 38% more likely to die from the disease than age- and prognosis-matched CAW [2]. Overall, AAW face disproportionately higher breast cancer mortality rates than that of CAW [1, 2, 4]. In addition to key social, political, and economic factors that impact the biological expression of breast cancer in AAW and CAW, mounting evidence indicates the presence of intrinsic molecular factors that contribute to racial disparities in breast cancer mortality. AAW were independently associated with poorer survival and worse prognosis than age- and clinically matched CAW, suggesting that breast cancer is a more biologically aggressive disease in AAW under equal treatment [5]. Furthermore, studies found that poorer rates of breast cancer survival are associated with an increased frequency of triple-negative breast cancer (TNBC), especially in young AAW [3, 6]. This suggests that a higher prevalence of TNBCs and a lower prevalence of luminal A tumors could contribute to the poor prognosis of young AAW with breast cancer. TNBC is a subtype of breast cancer that is defined by the absence of estrogen, nuclear progesterone, and HER-2 receptors [ER( $-$ ), nPR( $-$ ), HER2( $-$ )] [7]. While TNBC accounts for only $\sim$ 15% of all breast cancers, it exhibits the most aggressive metastatic behavior [8] and limited targeted therapies exist for TNBCs [9]. As aforementioned, AAW with TNBCs have worse clinical outcomes than that of CAW [10], suggesting that the higher prevalence and more aggressive biology of TNBCs in AAW [11] may contribute to racial differences in tumor biology [12, 13]. Overall, the biological basis of the four major disparities for breast cancer in AAW include, 1) early age of onset, 2) rapidly advancing stages, 3) aggressive histological changes, and 4) decreased survival rates, which cannot be an oversight. Although the racial disparity, in both prevalence of and risks of breast cancer between AAW and CAW may be attributable to social forces to some degree [14], decreased survival for AAW with TNBC remain the same after controlling for socioeconomic factors, treatment delay, and variations in breast cancer receptor expression [10]. Together, this provides strong evidence that after controlling for environment/treatment disparities, biological differences seem to significantly contribute to decreased survival of AAW with TNBC.

Progesterone (PRG), a sex steroid hormone, can exert its cellular effects through either its classic, non-classic, or combined responses by binding to either nuclear PRG receptors (nPRs) and/or membrane PRG receptors (mPRs), respectively, warranting both pathways equally important in PRG-mediated signaling. In our previous report, we demonstrated that the CCM signaling complex (CSC) consisting of CCM1, CCM2, and CCM3 can couple both nPRs and mPRs signaling cascades to form a CSC-mPRs-PRG-nPRs (CmPn) signaling network in nPR positive( $+$ ) breast cancer cells [15]. Additionally, our previous data support the notion that the CSC plays an important role during breast cancer tumorigenesis by coupling nPRs and mPRs (PAQRs) under PRG-induced actions [15].

Among TNBCs, over 70% are basal phenotype breast cancers, one of the most aggressive sub-types; mPRs role in tumorigenesis in this sub-type of PRG-responsive tumors is especially of great interest since these cells do not express nPRs [16]. In this report, we utilized two basal phenotype breast cancer cell lines, MDA-MB-468 (MB468) and MDA-MB-231(MB231), both only expressing mPRs [17, 18] (Suppl. Fig. 1A), to investigate the roles and relationships of key players of the newly defined CSC-mPRs-PRG (CmP) signaling network under mPR-specific PRG actions [treatment with PRG $+$ Mifepristone (MIF: nPR antagonist and mPR agonist)]. More importantly, MB468 is a well-known AAW-derived TNBC cell line, while MB231 is a CAW-derived TNBC cell line allowing us to investigate racial disparities among TNBCs. Previous observations suggested the existence of a biological basis for racial disparities in breast cancer mortality under sex steroid actions [12, 13]. Therefore, we hypothesize that there are distinct molecular regulatory mechanisms of the CmP signaling network under mPR-specific PRG actions during tumorigenesis of AAW-derived TNBCs.

The overall objective of this project is to understand the molecular regulatory mechanisms of racial disparities in breast cancer mortality concerning breast cancer tumorigenesis under mPR-specific PRG actions for the first time through molecular and cellular manipulation of cultured TNBC cell lines in-vitro. We also aim to analyze the dynamic role of the CmP network systematically and comparatively by deconvoluting the CmP signalosome utilizing systems biology approaches. Our data demonstrate that a common core mechanism exists among nPR( $-$ ) breast cancer cells, termed the CmP signaling network, regardless of nPR( $+/-$ ) status, which partially overlaps with the CmPn signaling network in nPR( $+$ ) breast cancer cells under identical PRG actions. These findings indicate a more essential role of the CSC on the stability of mPRs in nPR( $-$ ) breast cancer cells under mPR-specific PRG actions. This is the first report stating that inducible expression patterns exist between CCMs and major mPRs in multiple nPR( $-$ ) breast cancer cells. Furthermore, we demonstrate that mPRs, in TNBC cells, are localized in the nucleus and participate in nucleocytoplasmic shuttling in a coordinately synchronized fashion with CCM proteins under mPR-specific PRG actions; these findings are aligned with similar cellular distribution as other well-defined PRG receptors. Collectively, six potential biomarkers associated with altered expression in AAW-TNBC cells/tissues, consistent at both the transcriptional and translational levels, were also identified through our systems biology approach, which is quite significant for future AAW-TNBC therapeutic strategies. The knowledge gained from this project will have a great impact on deciphering the molecular and cellular mechanisms underlying racial disparities in breast cancer mortality through the CmP signaling network in AAW-TNBCs.

2. Results

2.1 Altered expression of mPRs and CCM genes across clinical tumors suggests their involvement in breast cancer tumorigenesis

Given our recent findings of the CSC potentially playing a significant role in reproductive tumorigenesis, and the important functions of mPRs for PRG signaling [19, 20, 21, 22, 23], we evaluated basal expression of key CmP players including the CSC (CCM1-3) and mPRs (PAQR5-9, PGRMC1/PGRMC2) in breast cancer tumor tissues using publicly available databases including TCGA. We first assessed expression profiles using The TCGA PANCAN database which allows for expression analysis by filtering tissues based on receptors and PAM50 classification, to evaluate CmP members expression data for two distinct breast cancer tissues, Basal [ER( $-$ ), nPR( $-$ ), HER2( $-$ )] and Her2 [ER( $-$ ), nPR( $-$ ), HER2( $+$ )] tumors, and observed significant differences in expression patterns of almost all CmP members (except CCM2/3 and PGRMC1) between the two sub-types (Fig. 1A), suggesting the impact of HER2 receptor on expression levels of major CmP players. Next, using the same database and for only Basal tumors (TNBCs), we observed significant expression differences among various histological types for CCM2/3 and PAQR5/7/9, suggesting a potential role for these key CmP players during TNBC tumorigenesis (Fig. 1B). We then utilized the TCGA BRCA database, which allows for expression analysis with enhanced filtering of tissues based on other phenotypes (such as menopause status) in combination with PAM50 classification. Interestingly, we observed differential expression patterns in CCM1 and CCM2 across various menopause stages in TNBC tumors, confirming a role of the CSC in steroid hormone signaling mechanisms (Fig. 1C). Finally, we evaluated expression levels in TNBCs, based on race, to determine if there are any expression differences between AAW and CAW with TNBCs. As expected, our analysis demonstrated significant differences in expression of CCM1, CCM2, and PAQR6 between AAW- and CAW-derived TNBCs (Fig. 1D). Together, these results validate the involvement of the CmP signaling network in TNBC tumorigenesis, illustrate the differential expression of CmP members in TNBCs, identify possible mechanisms contributing to racial disparities in TNBCs, and propose the future use of CmP members’ expression data as potential prognostic biomarkers for distinguishing between Her2 and Basal type breast cancers. Given these results, we next wanted to assess whether altered expression of these genes was associated with the worst prognosis in TNBCs.

Figure 1.

RNAseq expression profiling for key CmP players utilizing multiple TCGA breast cancer databases: The TCGA PANCAN database allows for expression analysis by filtering tissue based on PAM50 classification, various phenotypic filters, and is the most comprehensive breast cancer database to date. A) We evaluated CmP members expression data for two distinct breast cancer tissues based on PAM50 classification, Basal [ER( $-$ ), nPR( $-$ ), HER2( $-$ )] and Her2 [ER( $-$ ), nPR( $-$ ), HER2( $+$ )] tumors, and observed significant differences in expression patterns of almost all CmP members (except CCM2/3 and PGRMC1) between the two sub-types ( $n=$ 283). B) When looking at only Basal (TNBC) tissues, our analysis demonstrated significant expression differences among various histological types of triple negative breast cancers (TNBC; Basal), for CCM2/3 and PAQR5/7/9 ( $n=$ 183). C) Assessing menopause status among TNBCs, we discovered differential expression patterns in CCM1/2 ( $n=$ 131), confirming a role of the CSC in steroid hormone signaling mechanisms. Legend: Pre: $<$ 6 months since Last Menstrual Period (LMP) and no prior bilateral ovariectomy and not on estrogen replacement; Post: prior bilateral ovariectomy, or $>$ 12 months since LMP with no prior hysterectomy; Peri: 6–12 months since LMP. D) Our analysis demonstrated significant differences in expression of CCM1, CCM2, and PAQR6 between African and Caucasian Americans with TNBCs ( $n=$ 184). RNAseq expression profiling for key CmP players, based on TNM status, utilizing the TCGA breast cancer database in TNBC tumors: Utilizing the TCGA database we assessed expression profiling of CmP members, only in TNBCs, to evaluate their potential involvement in the size and extent of the main tumor (T). E) Our analysis revealed significant alterations in expression of CCM2 across various tumor stages in TNBCs ( $n=$ 137). Legend: T1: tumor is $⩽$ 2 cm across; T1c: tumor is $>$ 1 cm $⩽$ 2 cm; T2: tumor is $>$ 2 cm $⩽$ 5 cm; T3: tumor is $>$ 5 cm across. F) Utilizing the same database, we assessed expression profiling of CmP members to evaluate their potential involvement in infiltration into the lymph nodes (N). We discovered significant differences in expression of PAQR6/7 and PGRMC2 across various tumor lymph node status clinical samples ( $n=$ 95). Legend: N0: No regional lymph node metastasis identified or isolated tumor cells only (ITCs); N0(I $-$ ): No regional lymph node metastases histologically, negative IHC; N0(I $+$ ): ITCs only (malignant cell clusters $⩽$ 0.2 mm) in regional lymph node; N1: Micrometastases; or metastases in 1–3 axillary lymph nodes; and/or clinically negative internal mammary nodes with micrometastases or macrometastases by sentinel lymph node biopsy; N1a: Metastases in 1–3 axillary lymph nodes, at least one metastasis $>$ 2.0 mm; N1b: Metastases in ipsilateral internal mammary sentinel nodes, excluding ITCs; N1mi: Micrometastases ( $\sim$ 200 cells, $>$ 0.2 mm, but $⩽$ 2.0 mm); N2: Metastases in 4–9 axillary lymph nodes; or positive ipsilateral internal mammary lymph nodes by imaging in the absence of axillary lymph node metastases; N2a: Metastases in 4–9 axillary lymph nodes (at least 1 tumor deposit $>$ 2.0 mm); N3a: Metastases in $⩾$ 10 axillary lymph nodes (at least 1 tumor deposit $>$ 2.0 mm). G) Utilizing the same database, we assessed expression profiling of CmP members to evaluate their potential involvement in metastases mechanisms (M). Our analysis revealed significant changes in gene expression of PAQR5 and PGRMC1 using clinical samples across various stages of metastases ( $n=$ 133). Legend: M0: there is no sign that the cancer has spread; M1: Cancer measuring more than 0.2 mm across has spread to another part of the body and confirmed by examining tissue from a biopsy, or surgery and scans. H) When assessing overall stage scores, our analysis revealed significant alterations in expression of PAQR7/9 and PGRMC2 across various tumor stages in TNBCs ( $n=$ 140). For all graphs, X axis details genes profiled, while Y axis details Log2 normalized RNAseq expression data (panels C, E-H), or Log2 batch-effect normalized RNAseq expression data (panels A, B and D). All graphs were produced using the Xena platform.

Figure 2.

Prognostic effects for key CmP players utilizing microarray data of breast cancer patients. Publicly available microarray data from breast cancer patients was analyzed using either kmplot software or Xena browser (TCGA database) to integrate gene expression and clinical data simultaneously to generate the displayed Kaplan-Meier survival curves. A) Decreased expression of CCM1 and CCM3 had significantly worst prognostic effects [Overall survival (OS)] in basal type cancers using kmplot. B) Increased expression of most major mPRs (PAQR5-7 and PGRMC2) while decreased expression of PAQR8 and PGRMC1 had significantly worst prognostic effects in basal type cancers using kmplot. C) Our analysis utilizing the TCGA database confirmed that decreased expression of PAQR8 had significantly worst prognostic effects in basal type cancers. D) Similarly, utilizing the same TCGA database, we confirmed decreased OS rates with decreased CCM1 expression in basal type cancers. Breast cancer patients were filtered to only analyze patient samples classified as basal (TNBC) subtype. Logrank $P$ values are calculated and displayed as well as hazard ratio (and 95% confidence intervals) for panels A-B. Logrank $P$ values are calculated and displayed as well as logrank test statistics for panels C-D. Red line demonstrates high gene expression (all panels), while black line (panels A-B) or blue line (panels C-D) demonstrates low gene expression.

2.2 Differential expression patterns of mPR and CCM genes across clinical tumors and their associated prognostic effects

TNM staging is a widely used standardized system, which allows for assisting in prognostic cancer assessment to evaluate tumor size (T), regional lymph node involvement (N), and metastases of the primary tumor (M). Since the TCGA database contains phenotypic filters that allow for filtering of tissues, based on TNM staging, we evaluated expression profiles for the CSC and mPRs for each category, using only Basal phenotype breast cancer tissues. Tumor size assessment revealed significant alterations in the expression of CCM2 across various tumor size categories (Fig. 1E), suggesting the involvement of the CSC in influencing the size and extent of the primary tumor in TNBCs. Next, we repeated expression profiling to evaluate genes potentially involved in infiltration mechanisms into the lymph nodes. We discovered significant differences in expression of PAQR6, PAQR7, and PGRMC2 across various tumor lymph node status clinical samples, suggesting potential involvement of these key mPRs in lymph node infiltration mechanisms in TNBCs (Fig. 1F). Next, we assessed expression profiles to evaluate genes potentially involved in metastases mechanisms. Our analysis revealed significant differences in gene expression of PAQR5 and PGRMC1 across various stages of tumor metastases (Fig. 1G), also suggesting potential involvement of mPRs in metastases mechanisms in TNBC tumorigenesis. Finally, evaluation of expression profiles based on overall tumor staging categories illustrated significant differential expression in PAQR7, PAQR9, and PGRMC2, confirming the involvement of mPRs in overall tumorigenesis in TNBCs (Fig. 1H).

To further evaluate prognostic effects for the CSC and mPRs in TNBC tissues, we utilized publicly available breast cancer tumor tissue gene expression data (microarray) [24] integrating gene expression and clinical data simultaneously to generate Kaplan-Meier survival curves. First, breast cancer patients were filtered to only analyze patient samples classified as TNBC subtypes. Our analysis revealed that decreased expression of CCM1 and CCM3 had significantly worst prognostic effects in TNBCs (Fig. 2A), re-affirming the essential role of the CSC in TNBC tumorigenesis. Our analysis also revealed that increased expression of most mPRs (PAQR5-7 and PGRMC2) but decreased expression of PAQR8 and PGRMC1 had significantly worst prognostic effects in TNBCs (Fig. 2B), confirming the essential role of mPRs in TNBC tumorigenesis. Interestingly, when we repeated our analysis utilizing a different breast cancer database (TCGA) through the Xena browser software, our new analysis confirmed that decreased expression of PAQR8 had significantly worst overall survival (OS) (Fig. 2C) validating our previous results (Fig. 2B). Finally, decreased expression of CCM1 was also confirmed to result in significantly worst OS rates in TNBCs (Fig. 2D), further validating our previous results (Fig. 2A). These results solidify the involvement of the CmP signaling network in TNBC tumorigenesis and elucidate the great potential of the CmP signaling network for prognostic applications in TNBCs.

Distinct responses to mPR-specific PRG actions through mPRs in nPR( $\mathrm{-}$ ) cells. MB468 and MB231 cells are nPR( $-$ )/mPR( $+$ ) [17, 18] (Suppl. Fig. 1A) and are therefore good nPR( $-$ ) breast cancer cell models to examine the cellular roles and relationship between the CSC and mPRs within the CmP network under mPR-specific PRG actions. Human leukemia-derived cell line, Jurkat, was defined as only bearing mPRs without any other steroid receptor [25], also making it a great model to investigate the cellular function of mPRs in a non-breast cancer cell line. A normal nPR( $-$ ) breast cell line, MCF10A (10A), was also used as a control. First, we investigated the relative expression levels of total mPRs (PAQRs) under mPR-specific PRG treatment in these nPR( $-$ ) cell lines. Interestingly, significantly increased protein expression of major mPRs was observed (Suppl. Fig. 2A), in contrast to previous nPR( $+$ ) breast cancer data (decreased mPRs in T47D breast cancer cells) [15, 26]. Increased protein expression of PAQR7 and PAQR8 induced by PRG alone has been reported in MB231 cells [17], but this is the first report stating that mPR-specific PRG treatment can also induce overexpression of all major mPRs in multiple nPR( $-$ ) cell lines. Overexpression of all major mPRs induced with mPR-specific PRG treatment in 10A, a normal breast cell line, also indicates that this treatment-induced overexpression of mPRs is likely not associated with breast cancer tumorigenesis. Hormone-induced autologous receptor down-regulation is a common property of most steroid hormones through decreasing the half-life of occupied receptors leading to their more rapid decay, which includes classic nPRs. PRG bound-nPRs will down-regulate their levels by inhibiting transcription of the nPR gene. Similar autologous down-regulation of mPRs were also observed in nPR( $+$ ) T47D cells [15], making our current findings novel and interesting in nPR( $-$ ) cells.

Common relationships among the CmP signaling network in nPR( $\mathrm{-}$ ) cells. After silencing CCM (1, 2, 3) genes respectively in both nPR( $-$ ) AAW-TNBC (MB468) and CAW-TNBC (MB231) cells, significantly decreased protein expression of major mPRs (PAQR5/7/8) was observed compared to SC controls, especially in CCM2-KD TNBC cells (Suppl. Fig. 2B, left panel). This decreased protein expression of major mPRs (PAQR5, 7, 8) caused by CCM deficiency is in concordance with our data from nPR( $+$ ) breast cancer cells [15] but with higher significance (Suppl. Fig. 2B, right panel). These data reaffirm the existence of a common core, the CmP signaling network, regardless of nPR( $+/-$ ) status. Furthermore, it also indicates a more essential role of the CSC on the stability of mPRs in nPR( $-$ ) cells under steroid actions.

New subtypes of TNBC cells are based on the role and expression patterns of the CSC and mPRs. The expression levels of all CCMs (1, 2, 3) proteins are very low in the two nPR( $-$ ) TNBC cells compared to nPR( $+$ ) cells, 293T, and various endothelial cells [15, 26], and as a result, the effects of mPR-specific PRG actions on the protein expression of CCM3 in these nPR( $-$ ) cells were not detected(Suppl. Fig. 1B). However, given the impact on mPRs expression after knocking down any member of the CSC (Suppl. Fig. 2B), suggests the effects of mPR-specific PRG actions on the CSC through mPRs is effective even with minimal CSC expression. As aforementioned, increased protein expression of major mPRs (mPR $\alpha$ / $\beta$ ) under PRG treatment has been previously reported in MB231 cells [17], and we also observed increased RNA expression levels of major mPRs under mPR-specific PRG actions in MB231 cells (CAW) and nPR( $-$ ) Jurkat cells, while decreased/unchanged RNA expression of mPR genes was observed in MB468 cells (Fig. 3A). Conversely, increased RNA expression levels of CCMs were observed under mPR-specific PRG actions in AAW-TNBC cells only, while unchanged RNA expression of CCMs genes was observed in the other two nPR( $-$ ) cells (Fig. 3B). Interestingly, the overall RNA expression patterns of mPRs and CCM genes in AAW-TNBC cells are opposite to CAW-TNBC cells under equal treatment (Fig. 3A and B). These data suggest that combined RNA profiling of both mPR and CCM genes can be used as potential biomarkers to distinguish subtypes among TNBC cells, defined here as type-2A (CAW-TNBCs) and type-2B (AAW-TNBCs) nPR( $-$ ) cells.

Figure 3.

RNA expression levels of key factors of the CmP network in nPR( $-$ ) cells are modulated by sex steroids. The expression profiles of key factors of the CmP network [all three CCMs (1, 2, 3) and major mPRs (PAQR 5, 7, 8, 9)] were investigated among different nPR( $-$ ) cell lines, MB231, MB468, and Jurkat cells 48 hrs after mPR-specific PRG treatment (PRG $+$ MIF, 20 $\mu$ M each). A. Significantly increased RNA expression of major mPR (PAQR5, 7, 8, 9) genes (except PGRMC1) was observed in MB231 while Jurkat cells displayed only significant increased expression in PAQR5/7. Conversely, decreased expression of PAQR7/8 and increased RNA expression of PGRMC1 was observed in MB468 cells under mPR-specific PRG actions. The relative RNA expression changes were measured by qPCR (Fold changes). B. The RNA expression pattern of CCM genes in MB468 are opposite to that of the other two nPR( $-$ ) cell lines in which significantly increased RNA expression of CCM genes was observed in MB468 cells while no RNA expression changes was seen in either MB231 or Jurkat cells. The relative RNA expression changes of CCM (1, 2 or 3) genes were measured by qPCR (Fold changes). For all graphs, white bars represent significantly decreased expression, light gray bars represent no change and dark gray bars represent significantly increased expression, compared to vehicle controls (VEH) (triplicates per experiment, $n=$ 3); red line is the control baseline for fold change measurements ( $-/+$ ). ${}^{**}$ , ${}^{***}$ above bar indicates $P⩽$ 0.01 or 0.001 for paired $t$ -test, respectively.

Figure 4.

Tumorigenic assessment of TNBC cells under mPR-specific PRG treatment. AAW- and CAW-TNBC cells, were treated in triplicates in FBS-free medium containing vehicle control (EtOH, DMSO), Mifepristone (MIF, 20 $\mu$ M), progesterone (PRG, 20 $\mu$ M), or mPR-specific PRG treatment (MIF $+$ PRG, 20 $\mu$ M each). A. CAW-TNBC cells displayed no major change in cell migration with PRG, compared to vehicle controls, while MIF appeared to enhance cell migration. The combined steroid effect was in-between the individual responses (upper panel). AAW-TNBC cells displayed no effect in cell migration with PRG, compared to vehicle controls, while MIF as well as the combined steroids, enhanced cell migration. B. Similar trends in wound healing experiments were observed in which CAW-TNBC cells displayed no change in wound closure with PRG and mPR-specific PRG actions, while MIF appeared to enhance wound closure significantly, compared to vehicle controls (upper panel). AAW-TNBC cells displayed no effect in wound closure with PRG, compared to vehicle controls, while MIF as well as the combined steroids, appeared to enhance wound closure ability. C. MPR-specific PRG treatment can induce RNA expression of both CCM1 and GR (glucocorticoid receptor) genes in MB468 cells while no change was observed in the RNA expression of both PAQR7/PAQR8 genes (left panel). In contrast, in both nPR( $-$ ) MB231 and Jurkat cells, a sharp induced RNA expression of both PAQR7 and PAQR8 genes was seen under mPR-specific PRG actions but not for CCM1 or GR genes (middle and right panels), although there is a slight early induction in CCM1 gene at 24 hrs. No RNA expression was detected for either AR or nPR genes in all three nPR( $-$ ) cells (data not shown). D. The inducible expression of CCM1 protein in MB468 cells treated with mPR-specific PRG actions was discovered temporally, where green arrows demonstrate peak expression for CCM1 while no significant changes were detected for CCM3. For all panels, Relative RNA expression levels were measured through RT-qPCR (triplicates per experiment, $n=$ 3). In all bar plots, red line is the control baseline for fold change measurements ( $-/+$ ). ${}^{*}$ , ${}^{***}$ above bar indicates $P⩽$ 0.05 or 0.001, respectively, for paired $t$ -test.

Figure 5.

Modulation of key factors of the CmP network in TNBC cells under mPR-specific PRG actions. CAW and AAW-TNBC cells were treated with mPR-specific PRG treatment (MIF $+$ PRG, 20 $\mu$ M each) for 72 hrs. A-1. Immunofluorescent (IF) approaches revealed both CAW- and AAW-TNBC cells can alter localization of CCM1 proteins in and out of the nucleus and demonstrated increased relative CCM1 expression inside the nucleus at 72 hrs, although most of the staining resides in the peri-nuclear area (cytosol). The relative expression of CCM1 is also greater at 72 hrs compared to 0 hrs. A-2. IF approaches revealed significant changes in cellular localization of CCM3 in both TNBC cells, but only showed a significant increase in CCM3 intensity between 0 hrs and 72 hrs for AAW-TNBC cells.

Figure 5.

Continued. A-3. IF approaches revealed CAW-TNBC cells showed more PAQR8 staining inside the nucleus at 0 hrs followed by altering localization of PAQR8 back to the cytosol at 72 hrs. AAW-TNBC cells did not show much change in localization or expression of PAQR8 at 72 hrs compared to 0 hrs. B. Using NIKON elements analysis tools, localization of PAQR8 demonstrated the increased expression observed in total PAQR8 in both CAW-TNBC and AAW-TNBC cells appears to localize specifically inside the nucleus, and even when levels have decreased back down to those levels observed at 0 h, the ratio decreases back towards the cytosol, but is still predominantly located in the nucleus (top panels). Localization of CCM1 demonstrated increased expression in nuclear CCM1 in AAW-TNBC cells after 8 hours and the expression appears to still be predominantly located in the nucleus throughout the time course while the opposite trend was observed for CCM1 expression in CAW-TNBC cells, where CCM1 is localized in the cytosol for most timepoints (middle panels). Localization of CCM3 demonstrated expression of CCM3 in AAW-TNBC cells appears to localize specifically inside the nucleus at the earlier time points (0–12 hrs), and then localizes back into the cytosol closer to a 1:1 ratio, (except for 48 hrs) (bottom left panel). Expression of CCM3 in CAW-TNBC cells appears to localize specifically inside the nucleus at the earlier time points (0–12 hrs), shuttles back into the cytosol at 24–36 hrs, and then shuttles back into the nucleus for the remainder of the experiment. Red line indicates a 1:1 ratio of proteins in nucleus: cytosol. CCM1 was quantified through ROI intensities using wavelength 488 nm. PAQR8 and CCM3 was quantified through ROI intensities using wavelength 555 nm. Data were acquired from 8 sets of independent images and normalized against their respective internal controls using 408 nm wavelength for DAPI (nuclear) and background staining subtraction. For each replicate, Region of Interest (ROI) intensities were automatically quantified (over 1000 times/per area). In all bar plots, ${}^{**}$ and ${}^{***}$ above any bar graph indicate $P⩽$ 0.01 and $P⩽$ 0.001, respectively using One-way ANOVA ( $n=$ 8).

Figure 5.

Continued. C. Cellular fractionation and Western blot analysis show that PAQR8 (mPR $\beta$ ) is present in both the cytoplasm and the nucleus but is dominant in the nucleus. As a fractionation controls during the subcellular fractionation experiments, GAPDH, known to be present in the cytoplasm, Histone 3 (H3), known to be present in the nucleus, and $\beta$ – actin (ACTB) known to be present in both cytoplasm and nucleus (with cytoplasmic preference) are also shown. D. Utilizing our molecular works, we summarized the CmP signaling network in TNBC cells under mPR-specific PRG actions. In general, mPR-specific PRG actions (PRG $+$ MIF) have a positive effect on protein expression of the CSC in both TNBC sub-types but the inducible pattern of expression is more apparent in type 2B (AAW-TNBC) cells. Interestingly, RNA expression of CCM genes is also enhanced under mPR-specific PRG actions, while variable effects are observed in Type 2A (CAW-TNBC) cells. In general, expression levels of mPRs are enhanced at both the transcriptional and translational levels in Type 2A cells under mPR-specific PRG treatment, while mPRs expression is variable at both the transcriptional and translational levels in Type 2B cells. Overall, the CSC can stabilize expression of mPR proteins in TNBC cells. The purple line separates transcriptional and translational levels. The $+$ symbols represent enhancement, for the expression of targeted genes/proteins. Red colored symbols/lines represent positive effects of mPR-specific PRG treatment, blue colored symbols/lines represent negative effects of treatment, while yellow-colored symbols represent variable effects, temporally sub-cellular compartmentation and/or fluctuating expression along the time-course. Dark green colored letters indicate the direct supporting data generated from this report. Arrow indicates effect direction, solid line is the direct impact, dotted line for indirect effects.

2.3 TNBC cell performance is determined by the CmP network under steroid actions

TNBC cell performance is determined by the CmP signaling network. Based on our previous findings, under steroid actions, the intricate balance among key players of the CmPn signaling network is achieved through a balance between the negative effects of PRG/MIF signaling via mPRs and the positive effects of PRG signaling through nPRs in nPR( $+$ ) breast cancer cells [15, 21]. Our goal is to investigate the more fragile relationship within the CmP signaling network on the motility of nPR( $-$ ) breast cancer cells by tipping the delicate balance of the CmP network under steroid actions. CAW- and AAW-TNBC cells were treated with various steroid(s) to evaluate cell migration, wound healing, and transwell invasion in-vitro. CAW-TNBC cells displayed no change of cell migration with PRG, compared to vehicle controls, while MIF appeared to enhance cell migration, and the combined steroid effect was in-between the individual responses, both effects seen at earlier stages only (Fig. 4A, upper panel). Although PRG has been reported to decrease serum starvation-induced cell death in AAW-TNBC cells via mPR $\alpha$ [18], our data show that AAW-TNBC cells displayed no effect in cell migration with PRG, compared to vehicle controls, while MIF, as well as the combined steroid effect, appeared to enhance cell migration at earlier stages (Fig. 4A, lower panel), similar to our data obtained in CAW-TNBC cells. Similar trends in wound healing experiments were observed in which CAW-TNBC cells displayed unchanged wound closure ability with PRG and combined steroids, while MIF appeared to enhance wound closure significantly, compared to vehicle controls (Fig. 4B, upper panel, Suppl. Fig. 3A). AAW-TNBC cells displayed no effect in wound closure with PRG, compared to vehicle controls, while MIF, as well as the combined steroids, appeared to enhance wound closure ability (Fig. 4B, lower panel, Suppl. Fig. 3B). Interestingly, using cell invasion assays, trends of decreased cell invasion for both TNBC cell lines under mPR-specific PRG actions were observed, compared with vehicle controls, however, without any statistical significance (Suppl. Fig. 3C). These findings are in line with previous data that sex steroids can have direct actions on mPRs to navigate overall cell performance of nPR( $-$ ) breast cancer cells and even their metastatic potential [16, 27].

Inducible expression patterns of key factors of the CmP signaling network under mPR-specific PRG actions. The distinct RNA expression patterns between CAW- and AAW-TNBC cells are novel and interesting (Fig. 3A and B). We followed these phenomena temporally and found validating pieces of evidence for these opposite expression patterns. RNA expression of major mPR genes (PAQR7, PAQR8) in CAW-TNBC cells can be drastically induced by mPR-specific PRG actions (Fig. 4C, middle and right panels), compared to no response in AAW-TNBC cells (Fig. 4C, left panel), while RNA expression of CCM1 gene in AAW-TNBC cells can be significantly induced by sex steroids (Fig. 4C, left panel), compared to little or no response in CAW-TNBC cells (Fig. 4C, middle and right panels). These opposite temporal RNA expression patterns, seen among CAW- and AAW-TNBC cells, agree with RT-qPCR data (see Fig. 3A and B). Increased RNA expression levels of Glucocorticoid Receptor (GR) gene in AAW-TNBC cells under steroid actions is quite interesting (Fig. 4C, left panel), however, the same inducible RNA expression enhancement of GR gene under mPR-specific PRG actions was also observed in nPR( $+$ ) breast cancer cells (T47D, MCF7) with opposite cell performances [15]; these results suggest that this induction might be caused by the strong binding affinity of MIF to GR, which is unrelated to tumorigenesis. Previous studies demonstrating that the activation of GR signaling results in growth inhibition for various types of tumors, also suggest less relevance of activated GR in tumorigenesis [28]. Finally, mPR-specific PRG-inducible CCM1 expression was further confirmed at the translational level through time-course Western blots (Fig. 4D), validating our novel findings that mPR-specific PRG actions can induce distinct expression patterns of key factors (CSC) of the CmP signaling network between AAW- and CAW-TNBC cells.

2.4 mPRs are nucleocytoplasmic shuttling proteins

Temporal and spatial cellular compartmentation of key factors of the CmP signaling network. Since mPRs were initially identified as cell surface and membrane-bound PRG receptors [29] with seven predicted membrane-spanning motifs [29] functioning as putative G-protein coupled receptors (GPCRs) [30], membrane-impermeable PRG-BSA conjugate techniques [31] have been widely used to measure PRG-mPRs signaling with a sole focus on its non-genomic actions [32, 33, 34]. It has been reported that there is very low expression of PAQR7 in MB231 cells [27, 32] therefore, we performed cellular localization of PAQR8 along with other key factors (CCM1/3) in the CmP network in both CAW/AAW-TNBC cells. Under our observations, initially, CCM1 is equally distributed in the nucleus in both TNBC cells, however, after mPR-specific PRG actions, both nuclear and cytoplasmic increased expression of CCM1 is visualized in AAW-TNBC cells (Fig. 5A-1, left panel), while CCM1 expression not only increases as well but concentrates along the nuclear membrane (peri-nucleus) in CAW-TNBC cells (Fig. 5A-1, right panel). In contrast, CCM3 appears to be distributed dominantly in the nucleus initially and mPR-specific PRG-enhanced expression of CCM3 was visualized more towards the cytosol (more dramatic in AAW-TNBC cells) (Fig. 5A-2). Intriguingly, PAQR8 localizes within the nucleus and remains unchanged under mPR-specific PRG actions in AAW-TNBC cells (Fig. 5A-3, left panel), while PAQR8 is dominantly distributed initially within the nucleus (likely associated with the nucleolus), and becomes more evenly distributed in both nucleus and cytosol after mPR-specific PRG actions in CAW-TNBC cells (Fig. 5A-3, right panel). It was not a total surprise for us to observe nuclear localization of PAQR8 in TNBC cells since we have observed a similar phenomenon in Luminal-A T47D cells [15]. Interestingly, there have been contradicting reports of cellular localization of mPRs in either the cytoplasmic membrane or cytoplasmic compartmentations (mainly rough ERs) in several cell types, including MB231 cells [30, 35], while cellular compartmentations of mPRs on MB468 cells have yet to be reported. Our data clearly show that PAQR8 is capable of nuclear sub-compartmentation in both TNBC cells. Furthermore, we also observed nuclear localization of PAQR5 (mPR $\gamma$ ) and PAQR7 (mPR $\alpha$ ) in both TNBC cell lines utilizing IHC techniques(Suppl. Fig. 4B). Together, these data support the notion that mPRs are nuclear proteins in TNBC cells.

Next, we examined the temporal expression patterns under mPR-specific PRG actions, demonstrating steroid-induced expression of CCM1 again in AAW-TNBC cells (Suppl. Fig. 4A-1), in line with our previous data (Figs 3B, 4D, 4E). Interestingly, there is an early surge (non-genomic) of CCM3 protein expression in AAW-TNBC cells (Suppl. Fig. 4A-2), which is unrelated to the genomic induced expression of CCM3 (increased RNA expression at earlier time points, Fig. 4C), and opposite to the data observed in luminal A T47D breast cancer cells [15]. The underlined mechanisms of CCM3 fluctuation in AAW-TNBC cells is still unknown. A visual subtle decrease of PAQR8 in AAW-TNBC cells under mPR-specific PRG actions was also observed(Suppl. Fig. 4A-3). Interestingly, mPR-specific PRG-induced expression of CCM1 was observed again in CAW-TNBC cells (Suppl. Fig. 4A-4), collaborating with our previous data (Fig. 4D). Likewise, mPR-specific PRG-induced expression of CCM3 (Suppl. Fig. 4A-5) and PAQR8 in CAW-TNBC cells (Suppl. Fig. 4A-6) were also visualized, in line with our previous data (Figs 3B and 4D). The major discovery from these observations is the temporal and spatial regulation of protein expression of key players of the CmP network under mPR-specific PRG actions in both AAW- and CAW-TNBC cells.

Cytoplasmic-nuclear trafficking of key players of the CmP network. One of the major findings associated with our data is the cytoplasmic-nuclear trafficking of mPRs. Utilizing a larger sample size, we carefully examined the temporal subcellular localization of key factors (CCM1/3, PAQR8) of the CmP network under mPR-specific PRG actions in both TNBC cells and not only found PAQR8 inside the nucleus (Suppl. Figs 4A-3, 4A-6) but also found the temporally dynamic expression patterns of PAQR8 in the nucleus and cytosol in both TNBC cells (Fig. 5B, upper panels). Cytoplasmic-nuclear trafficking of CCM1 has been well studied by our lab [36, 37, 38, 39, 40, 41], therefore, we used CCM1 as a positive control to validate the possible cytoplasmic-nuclear trafficking of PAQR8. We found that both CCM1 and CCM3 proteins are predominantly in the nucleus in AAW-TNBC cells (Fig. 5B, middle and lower left panels), while CCM1 and CCM3 show distinct nucleo-cytoplasmic shuttling capabilities with a more equal subcellular compartmentation in CAW-TNBC cells (Fig. 5B, middle and lower right panels). Intriguingly, CCM1 and PAQR8 localization, in CAW-TNBC cells, displayed the same temporal and spatial trends observed after 4 hrs, indicating perhaps coordinately synchronized cytoplasmic-nuclear trafficking activities of endogenous CCM1 and PAQR8 only in CAW-TNBC cells under mPR-specific PRG actions (Fig. 5B, middle and upper right panels). The observed phenomena suggest that PAQR8 nucleocytoplasmic shuttling might be CCM1-dependent, and recent data have shown that many proteins constantly shuttle between the cytoplasm and the nucleus without an obvious bona fide classic NLS and/or NES signaling site [42]; however, our bioinformatics simulation data suggested the possible existence of one NLS and two NES sites within PAQR8 (Suppl. Fig. 4C), further strengthening our observations of nuclear-cytoplasmic shuttling of PAQR8 in both TNBC cells. Finally, due to abundant nuclear localization of PAQR8 observed in CAW-TNBCs through IF imaging (Fig. 5A-3), cellular fractionation of CAW-TNBCs confirmed that PAQR8 (mPR $\beta$ ) is present in both the cytoplasm and the nucleus but is dominant in the nucleus (Fig. 5C), validating our previous observations of nuclear compartmentation of mPRs (Fig. 5A-3, Suppl. Figs 4A-3 and 4A-6).

Different responses to sex steroid actions in TNBCs. In the era of personalized medicine or precision medicine, the correct classification of breast cancer into molecular subtypes with distinctive gene expression signatures or profiles to provide a needed prognosis for treatment is an unmet task for this remarkably heterogeneous and deadly human condition. Although nPR( $-$ ) cells lack the appropriate nuclear receptors to enact genomic responses under PRG actions, PRG can activate downstream signaling in both nPR( $+/-$ ) cells by binding to mPRs [17, 25, 26, 43]. In light of a growing number of studies that have implicated mPRs in breast cancer [18, 44, 46, 47, 48, 49, 50], we sought to characterize the CmP signaling network in patient-derived TNBC lines, which are nPR( $-$ )/mPR( $+$ ) [17, 18]. The distinct RNA/Protein expression profiles of key factors within the CmP network under mPR-specific PRG actions have been utilized to discover potential biomarkers to classify two TNBC cells into two subgroups (Fig. 5D) based on their different responses to mPR-specific PRG actions. In general, expression levels of mPRs are enhanced at both the transcriptional and translational levels in CAW-TNBC cells under combined steroid treatment, while mPRs expression is variable at both the transcriptional and translational level in AAW-TNBC cells, making these genes great candidate biomarkers for TNBC sub-type classification (Fig. 5D). Together, our data support the notion that multiple mPRs can be co-expressed in various mammalian cell types [18, 25, 51, 52, 53] to perform multifaceted non-classic PRG signaling cascades among different nPR( $+/-$ ) cells [18]. We next sought to identify altered signaling cascades in AAW- and CAW-TNBC cell lines under mPR-specific PRG actions.

2.5 Our novel findings are further validated through Omics

Dissecting the CmP network in AAW-derived TNBC cells using omic approaches. To define key altered pathways, potential partners involved in the CmP signaling network, and potential biomarkers, we examined the expression patterns of AAW-TNBC cells under mPR-specific PRG actions, at both the transcriptional and translational levels using high-throughput Omic approaches, including RNAseq and Tandem Mass Spectrometry (LC-MS/MS). Since RNA expression of CCM1 gene in AAW-TNBC cells can be significantly induced by sex steroids (Fig. 4C, left panel), compared to no or little response in CAW-TNBC and Jurkat cells (Fig. 4C, middle and right panels), we filtered out our RNAseq and proteomics data for AAW-TNBC cells by removing similarly altered Differentially Expressed Genes/Proteins (DEGs/DEPs) shared between AAW-and CAW-TNBC cells (including Jurkat) to evaluate AAW-TNBC specific DEGs/DEPs.

Key signaling cascades identified within the CmP network in AAW-TNBC cells using RNAseq. Among the identified DEGs (Suppl. Table 4A), we were able to visualize hierarchical clustering (Suppl. Fig. 5A, top panel), and found similar temporal patterns in both intersection and union of DEGs(Suppl. Fig. 5A, bottom panels). Identification of up and down-regulated genes at each time point revealed that mPR-specific PRG actions have an overall stronger up-regulation on AAW-TNBC specific DEGs (Fig. 6A, top panel). Further stratification demonstrated 1300 $+$ genes temporally up-regulated and 700 $+$ genes temporally down-regulated under mPR-specific PRG actions, yielding a great foundation of potential biomarkers to distinguish AAW-TNBCs from other subtypes (Fig. 6A, bottom panel). Pathways functional enrichment revealed several major altered pathways including the most affected signaling pathway which was the Gonadotropin-releasing hormone (GnRH) signaling pathway (Fig. 6B, top panel), responsible for regulating mammalian reproduction, including production of hormones [54]. Amplification of EGFR members is one of the most common genetic alterations associated with breast cancer [55], and interestingly, EGF receptor signaling pathways was one of the top 10 affected pathways in AAW-TNBC cells under mPR-specific PRG actions (Fig. 6B), reaffirming the role of key CmP players in AAW-TNBC progression. The other top signaling pathways altered in AAW-TNBC cells included Inflammation mediated by chemokines and cytokines, several angiogenesis-related pathways as well as heterotrimeric G-protein pathways (Fig. 6B). Together, these results further validate the crosstalk between the CSC and mPRs in the CmP signaling network involved in AAW-TNBC tumorigenesis. Finally, several key signaling cascades in tumorigenesis were frequently observed to be altered in AAW-TNBC cells under mPR-specific PRG actions, including apoptosis, p53, MAPK, WNT, RAS, PI3k-AKT, and hormone signaling pathways, further validating the involvement of the CmP network in AAW-TNBC progression (Fig. 6B and Suppl. Fig. 6A–E).

Figure 6.

Differentially Expressed Genes/proteins (DEGs/DEPs) among AAW-TNBC cells utilizing high-throughput OMIC approaches: AAW-TNBC cells were treated with MIF $+$ PRG (20 $\mu$ M each) for 12, 24 and 48 hrs. Extracted significant DEGs were further filtered using corresponding MB231/Jurkat cells to obtain DEGs specific to AAW-TNBC cells. A) Bioinformatics analysis to identify significant DEGs, compared to vehicle controls (Top panel) that were further stratified into independent Venn diagrams for both up and down regulated genes to identify overlaps between timepoints (Bottom panels). B) Pathway functional enrichment were compiled using molecular functions, biological processes and signaling pathway data retrieved from the PANTHER (Protein ANalysis THrough Evolutionary Relationships) classification system. C) Cluster software and Euclidean distance matrixes were used for the hierarchical clustering analysis of the expressed proteins and sample program at the same time to generate the overall Heatmap of hierarchical clustering using $t$ -test statistical analysis (top panel). Further analysis was performed to identify significant DEPs, compared to vehicle controls (middle panel). Significant DEPs were further stratified into independent Venn diagrams for both up and down regulated proteins (Bottom panel). D) Overall Heatmap of hierarchical clustering using ANOVA statistical analysis (top panel); Further analysis was performed to identify significant DEPs, compared to vehicle controls (middle panel). Significant DEPs were further stratified into independent Venn diagrams for both up and down regulated proteins to identify overlaps between timepoints (Bottom panel). E) Pathway functional enrichment results were retrieved from the PANTHER classification system. For panels A, C, D, lower B, and lower E coloring indicates the log2 transformed fold change (high: red, low: blue). For upper panel B, coloring indicates the potential impact on each signaling pathway; high (more than 40 genes): red, moderate (20–39 genes): orange, medium (10–19 genes): yellow, low (5–9 genes): blue. For upper panel E, coloring indicates the potential impact on each signaling pathway; high (4–6 proteins): red, moderate (3 proteins): orange, medium (2 proteins): yellow, low (1 protein): green. For all heatmaps, x axis represents each comparing sample and Y axis represents DEGs/DEPs. For all functional enrichment graphs, X axis represents the terms of pathway while y axis represents the number of identified DEGs/DEPs involved in each pathway. For all panels, each timepoint was run with biological triplicates and technical triplicates for treated and vehicle samples.

To delineate cellular partners sharing the same inducible RNA expression patterns as CCM1 (Fig. 4C, left panel), we temporally analyzed RNAseq data to search for significantly inducible up-regulated genes (Suppl. Table 4B), and identified 14 genes under mPR-specific PRG actions (Suppl. Fig. 5B, left panel). These findings suggest that mPR-specific PRG actions can induce RNA expression for a subset of genes, identical to the expression patterns seen for CCM1 in AAW-TNBC cells. Among the identified 14 AAW-TNBC-specific DEGs, was IL6, a well-known player in key cancer pathways. Additionally, identified inducible expression of PLK3 has been suggested as a novel independent prognostic marker in breast cancer [56]. Alterations in several key pathways, resulting from these 14 DEGS, included cell population proliferation, reproduction, and immune system processes(Suppl. Fig. 5D). Additionally, these results validated altered pathways seen in the global RNAseq profiling of AAW-TNBC cells under mPR-specific PRG actions (compare Fig. 6B, and Suppl. Fig. 5D). Next, we also searched for significantly inducible down-regulated genes (Suppl. Table 4B) and identified 8 genes under mPR-specific PRG actions (Suppl. Fig. 5C). Altered signaling pathways, resulting from these 8 DEGS, included angiogenesis, WNT, and apoptosis (Suppl. Fig. 5E).

Key signaling cascades within the CmP network in AAW-TNBC cells using LC-MS/MS approaches. Through Euclidean mapping of DEPs (Suppl. Table 5A), we were able to visualize proteomic regulation (Fig. 6C–D, top panels), and upon identifying up and down-regulated genes temporally, revealed that there is a greater propensity of upregulated proteins at 24 and 72 hrs when analyzed using $t$ -test (Fig. 6C, middle panel) while there is a greater propensity of upregulated proteins at all three-time points when analyzed using ANOVA methods (Fig. 6D, middle panel). Two statistical methods were used to determine changes between two specific time points ( $t$ -test) and overall temporal changes (ANOVA). Hormone-treated AAW-TNBC cells showed a total of 66 upregulated proteins using $t$ -test analysis (Fig. 6C, bottom left panel) while 342 proteins were identified using ANOVA methods (Fig. 6D, bottom left panel), demonstrating more dramatic changes temporally affecting increased/induced protein expression. In contrast, 70 proteins were observed down-regulated using $t$ -test analysis (Fig. 6C, bottom right panel) and only 27 proteins using ANOVA statistics (Fig. 6D, bottom right panel). In complementation to potential DEG biomarkers identified (Fig. 6A and Suppl. Fig. 5), we now have an additional index of potential AAW-TNBC specific DEP biomarkers (Fig. 6C and D and Suppl. Fig. 7) that may help limit the scope of commonly shared biomarkers in different sub-types of TNBCs. When proteins were functionally enriched, we noted that some pathways overlaid with our RNA enrichment, reinforcing the concept that these pathways may play potential key roles in hormone signaling in AAW-TNBCs (Fig. 6E both panels compare with Fig. 6B both panels, and compare Suppl. Figs 6A-E & 8A-D). Pathways functional enrichment from these DEPs revealed several major altered pathways (Fig. 6E, top panel) including the most affected signaling pathway, which was again the GnRH signaling pathway, similar to our RNA-seq data (Fig. 6B, top panel). Notably, modulation of the GnRH pathway is also seen in breast, prostate, and endometrial cancers [57], all sex steroid-responsive cancers, and is dysregulated in our proteomic analysis of AAW-TNBC cells under mPR-specific PRG actions.

We again delineated cellular partners sharing the same/reverse inducible expression patterns as CCM1 (Fig. 4D), as we did with our transcriptomic data, at the translational level. We identified 50 consistently up-regulated proteins (Suppl. Table 5B and Suppl. Fig. 7A) with similar expression patterns as CCM1. Intriguingly, 30 of the 50 consistently up-regulated proteins are differentially expressed in numerous cancers including breast, cervical, liver, lung, and brain cancer, contributing to tumor growth, metastasis, and high invasive activities(Suppl. Table 5C). We observed several of the previously altered signaling pathways (detected using DEGs) for these up-regulated DEPs(Suppl. Fig. 7A), including Integrin, Inflammation, toll receptor, and interleukin signaling pathways (Suppl. Fig. 7C). Finally, we identified only 4 DEPs with inverse expression patterns to CCM1 (Suppl. Fig. 7B, Suppl. Table 5B) and observed similar regulatory effects (seen at the transcriptional level) in the analysis of signaling pathways for these down-regulated DEPs(Suppl. Fig. 7B) including biological regulation, response to stimuli and localization pathways (Suppl. Fig. 7D).

Together, these results solidify the existence of a cellular relationship between the CSC and mPRs signaling at the translational level by establishing the CmP signaling network in AAW-TNBC cells. Our extensive omics analysis has also provided several new candidate biomarkers, that can be further investigated to determine their potential clinical applications for AAW-TNBCs.

Key signaling cascades within the CmP network in AAW-TNBC cells identified using systems biology approaches. To look at the inter-relationship between RNA and protein expression from AAW-TNBC cells under mPR-specific PRG actions, we overlaid the regulation patterns and extrapolated the overlaps between our two OMIC datasets(Suppl. Table 6A). When looking at the enriched annotations we see pathways involved in the regulation of cancer and tumor factors still consistently present from the individual analyses, including Integrin, cytokine-mediated responses, GnRH, WNT, and angiogenesis among the top 10 affected pathways (Fig. 7A and Suppl. Table 6B). These results further validate our independent observations at both the transcriptional and translational levels (Fig. 6 and Suppl. Figs 5-8). When reviewing DEG/DEP expression patterns, it should be noted that DEGs/DEPs with the same regulation trends (averaged log2 Fold Changes) seen at both levels (Bolded blue and red ID’s) are key AAW-TNBC specific potential biomarker candidates (Fig. 7B), while DEGs/DEPs displaying opposite expression patterns (black-colored ID’s) commonly reflect genes that are capable of undergoing potential feedback auto-regulation (Fig. 7B). Up-regulated DEGs/DEPs illustrate a significant impact on essential cancer pathways including Integrin, Rho GTPase, angiogenesis, VEGF, GNRH, G-protein, EGF, PDGF, and Cadherin signaling pathways (Fig. 7B).

Figure 7.

Systems biology analysis utilizing high-throughput omics data to identify novel biomarkers unique to tumorigenesis in AAW-TNBC cells under mPR-specific PRG actions: Extracted significant DEGs/DEPs were further filtered using corresponding MB231/Jurkat cells to obtain DEGs/DEPs specific to AAW-TNBC cells. A. Pathway functional enrichment results for corresponding DEGs/DEPs were compiled using the PANTHER classification system. Coloring indicates the potential impact on each signaling pathway; high (5–6 genes/proteins): red, moderate (3–4 genes/proteins): orange, medium (2 genes/proteins): yellow, low (1 gene/protein): green. B. A corresponding table is provided for each signaling pathway that details the genes involved, and the statistical analysis used with the proteomics data (A, ANOVA; T, $t$ -test). Genes/Proteins with similar regulation at both the transcriptional and translational levels are color coded and bolded with red DEGs/DEPs (up-regulated) or blue DEGs/DEPs (down-regulated) while Genes/Proteins in black display opposite expression. C–F. Cluster software and Euclidean distance matrixes were used for the hierarchical clustering analysis of DEGs/DEPs along with overlaps in CHIP-seq databases to generate the overall Heatmap of hierarchical clustering (panels C and E); x axis represents each comparing sample and Y axis represents DEGs/DEPs. Further bioinformatics analysis was performed to stratify significant DEGs/DEPs, compared to vehicle controls that overlapped with CHIP-seq databases in hormone treated breast cancer cells (panels D and F). G. A corresponding table is provided with affected signaling pathways that details DEGs/DEPs from our systems biology analysis. DEGs/DEPs with similar regulation at both the transcriptional and translational levels are color coded and bolded with red DEGs/DEPs (up-regulated), while Genes in black display opposite expression at the transcriptional and translational levels. Pathway functional enrichment results were performed using signaling pathway data retrieved from the PANTHER classification system. For all functional pathway’s analysis, X axis represents the terms of pathway while the Y axis represents the number of identified DEGs/DEPs. For panels B-G, coloring scheme indicates the log2 transformed fold change (high: red, low: blue). For all panels, each timepoint was run with biological duplicates (RNA) or triplicates (Protein) and technical triplicates (both) for treated and vehicle samples.

We then performed a meta-analysis using a working database (Suppl. Table 6C) with CHIP-seq data from various breast cancer cells (containing AAW-TNBC cells) treated with either PRG/MIF. This task was accomplished by using the NCBI Batch Web CD-Search tool [58, 59]. The list of overlapped genes (Suppl. Table 6C) was graphed through Euclidian heat maps for both RNA and protein as well as their regulation patterns (Fig. 7C–F). In general, we observed that most of the genes identified in our systems-wide analysis with combined RNA-seq/Proteomics/CHIP-seq data appear to be capable of being auto-regulated through feedback mechanisms in AAW-TNBC cells under mPR-specific PRG actions. However, we still identified two steroid-inducible AAW-TNBC specific candidate biomarkers namely EGFR and MYO1B (Fig. 7G).

2.6 Differential expression of our candidate biomarkers in breast cancer tissues and their associated prognostic effects

We set out to filter our identified candidate biomarkers for AAW-TNBCs from our omics approaches to obtain a more clinically relevant set of prognostic biomarkers. To accomplish this, 11 publicly available breast cancer datasets from the University of California-Santa Cruz (UCSC) Xena Browser [60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71] as well as two TCGA-BRCA databases available in Xena Browser (TCGA-BRCA and GDC-TCGA-BRCA) were compiled to validate differential expression among breast cancer subtypes for our Identified candidate biomarkers(Suppl. Table 6D). Additionally, we evaluated prognostic effects for our candidate biomarkers [24] integrating clinical and gene expression data simultaneously to generate Kaplan-Meier (KM) survival curves. Breast cancer patient samples were filtered to only analyze patient samples classified as TNBC subtype. For the first group of 12 biomarkers, a worst prognostic effect was observed with higher expression(Suppl. Fig. 9A1–A12). We also observed up-regulation for 10/12 biomarkers (CNN3 and LSS were down-regulated) with a disrupted CmP signaling network, under mPR-specific PRG actions, in TNBC cells(Suppl. Table 6D), demonstrating that mPR-specific PRG actions could lead to worst prognostic outcomes in AAW-TNBCs. Alternatively, our second group of 9 biomarkers demonstrated a better prognostic effect with higher expression(Suppl. Fig. 9B1–B9). Interestingly, we observed up-regulation of all 9 genes/proteins with a disrupted CmP signaling network, under mPR-specific PRG actions, in TNBC cells(Suppl. Table 6D). Together these results demonstrate the heterogeneity response of TNBCs under mPR-specific PRG treatments. Utilizing our KM survival curve data, we further assessed our biomarkers by only proceeding with biomarkers that had significant KM survival curves in TNBC patients to assess their basal expression patterns in clinical samples.

Utilizing publicly available breast cancer tumor tissue gene expression data (microarray) [24] we explored expression patterns of our identified candidate biomarkers in 925 nPR( $+/-$ ) breast cancer tumors (ER and HER2 receptor status not available). 10 of our candidate biomarkers were found to be significantly up-regulated in nPR( $-$ ) tumor tissues(Suppl. Fig. 9C1–C10). We observed the same up-regulated trends for 9/10 biomarkers, except for LSS, with a disrupted CmP signaling network under mPR-specific PRG actions in AAW-TNBC cells(Suppl. Table 6D). Alternatively, 5 of our candidate biomarkers were found to be significantly down-regulated in nPR( $-$ ) tumor tissues (Suppl. Fig. 9C11–C15). Interestingly, we observed opposing up-regulation of these 5 biomarkers with a disrupted CmP signaling network. Furthermore, 6 of our biomarkers were found to have similar expression patterns in nPR( $+/-$ ) breast cancer tumor tissues(Suppl. Fig. 10A). However, disruption of the CmP signaling network, with mPR-specific PRG actions, resulted in significant up-regulation of 5/6 biomarkers, apart from CNN3(Suppl. Table 6D).

Table 1

6 identified prognostic candidate biomarkers in AAW-TNBCs

BIOMARKERS (ALTERNATE ID)	DEGs TNBC	DEPs TNBC	TNBC-AAW $N=$ 23 patients (filter 1)	TNBC-CAW $N=$ 19 patients (filter 1)	K-M Survival curve TNBC PATIENTS	Expression differences $N=$ 925 nPR( $-$ ) patients; $N=$ 926 nPR( $+$ ) patients	TNBC-AAW $N=$ 64 patients (filter 2)	TNBC-CAW $N=$ 113 patients (filter 2)
EGFR					I.S. $N=$ 176	N.S.
GARS					D.S. $N=$ 392	HIGH IN nPR( $-$ )
EML2			N/A	N/A	D.S. $N=$ 392	LOW IN nPR( $-$ )
LANCL2					D.S. $N=$ 176	N.S.
LSS					I.S. $N=$ 392	HIGH IN nPR( $-$ )
COPE					I.S. $N=$ 392	N.S.

Candidate biomarkers shared at both RNA/Protein levels under mPR-specific PRG actions (columns 1, 2, 3), were preliminarily validated in the first set of AAW- ( $n=$ 23) and CAW-TNBC ( $n=$ 19) tumor tissue data sets (columns 4, 5). These biomarkers were also analyzed by a set of publicly available microarray data (22,277 probes) from 1,809 breast cancer samples to generate the displayed Kaplan-Meier survival curve results (column 6). Candidate biomarkers were further assessed with two groups of breast cancer tissue samples, based on nPR status determined by Immunohistochemistry (IHC), to evaluate clinical expression levels focusing on 925 nPR( $-$ ) breast cancer samples (column 7). Finally, candidate biomarkers were further validated in an independent and larger cohort of AAW- ( $n=$ 64) and CAW-TNBCs ( $n=$ 113) tumor tissues utilizing the TCGA database (columns 8, 9). Red colored background indicates increased/higher expression levels, while blue colored background indicates decreased/lower expression levels. Colored biomarkers (red in column 1) indicate intrinsic biomarkers, while black colored biomarkers are inducible biomarkers, similar to CCM1. Abbreviations: I.S., Increased Survival; D.S., Decreased Survival; N.S., No statistical significance; HIGH IN nPR( $-$ ), Higher expression levels in nPR( $-$ ) breast cancer tissues; LOW IN nPR( $-$ ), Lower expression levels in nPR( $-$ ) breast cancer tissues; N/A, Not Available.

Dual validation of our candidate biomarkers between CAW- and AAW-TNBC tissues for their prognostic effects. We next sought to enhance our analysis to investigate differential expression patterns between two cohorts of comparative expression data from AAW- and CAW-TNBC tumor tissues. To accomplish this task, first, we utilized publicly available RNA-seq data comparing expression levels in 23 AAW-TNBC and 19 CAW-TNBC samples (filter 1) [72] that were used to obtain expression data for the candidate biomarkers associated with significant KM curves identified in this study. 10 of our markers displayed similar trends with a disrupted CmP signaling network, as was uniquely observed in the 23 AAW-TNBC patient samples(Suppl. Table 6D, red and blue colored, bolded biomarkers, column 1). This allowed us to preliminarily define these genes as intrinsic biomarkers for AAW-TNBCs. Alternatively, the remainder of our biomarkers displayed opposite or inducible trends with a disrupted CmP signaling network in AAW-TNBC cells, as was naturally observed in the 23 AAW-TNBC patient samples, preliminarily allowing us to define these genes as inducible biomarkers (CmP disruption) for AAW-TNBCs(Suppl. Table 6D, black-colored, bolded biomarkers). Finally, we wanted to confirm our initial 21 identified biomarkers with differential expression patterns between AAW and CAW-TNBCs(Suppl. Table 6D, bolded markers column 1) using a separate and larger cohort (filter 2) of publicly available microarray data containing 64 AAW-TNBC and 113 CAW-TNBC samples through the TCGA database. Our analysis confirmed the identification of EGFR and LSS as inducible biomarkers (Table 1, black-colored biomarkers) as well as identification of COPE as an intrinsic biomarker (Table 1, red-colored biomarker) for AAW-TNBCs(Suppl. Fig. 10B). Interestingly, this larger cohort instead identified GARS and LANCL2 as inducible biomarkers (Table 1, black colored biomarkers) for AAW-TNBCs, while EML2 emerged as a newly identified intrinsic biomarker(Suppl. Fig. 10B and Table 1, red-colored biomarker). Together, these results reinforce the existence of both intrinsic and inducible biomarkers in AAW-TNBCs, a pioneering first step to tackle the currently alarming Black-White disparity in breast cancer mortality.

Utilizing all the expression and prognostic data, we filtered out our initial 47 biomarker hits (Suppl. Table 6D) to a preliminary list of 21 intrinsic and inducible biomarkers for AAW-TNBCs(Suppl. Table 6D, bolded biomarkers), that were even further condensed to a clinically relevant list of 2 intrinsic and 4 inducible biomarkers (Table 1) confirmed to have statistically significant differential expression among AAW and CAW-derived TNBC tumors.

3. Discussion

There has been sufficient evidence for the existence of PRG-mPRs signaling cascade in either nPR( $+/-$ ) cells [16, 17, 25, 43, 52], suggesting that PRG signaling in nPR( $-$ ) cell lines can be mediated solely through mPRs (PAQRs) [16]. In this current study, we further defined the novel CSC-mPRs-PRG (CmP) signaling network in nPR( $-$ ) breast cancer cells (Fig. 5D) which overlaps with our previously defined CmPn (CSC-mPRs-PRG-nPRs) network in nPR( $+$ ) breast cancer cells [15]. In the CmP signaling network, the CSC can stabilize mPRs under mPR-specific PRG actions in a forward fashion (CSCmPRs), indicating a more essential role of the CSC on the stability of mPRs in nPR( $-$ ) breast cancer cells. Overall, the CSC can stabilize the expression of mPRs in TNBC cells (Suppl. Fig. 2B), in concordance with our observations in nPR( $+$ ) breast cancer cells, indicating the consistent function of the CSC on the stability of mPRs under mPR-specific PRG actions [15]. In contrast to previously observed nPR( $+$ ) breast cancer and nPR( $-$ ) endothelial cell data [15, 26], mPR-specific PRG actions have a positive effect on protein expression on the CSC in both TNBC sub-types with inducible expression patterns more apparent in AAW-TNBC cells (Figs 4D and 5A and Suppl. Fig. 4A1-6). Interestingly, RNA expression of CCM genes is also enhanced with mPR-specific PRG treatment in AAW-TNBCs, while variable effects are observed in CAW-TNBC cells (Figs 3B and 4C). In general, expression levels of mPRs are enhanced at both the transcriptional and translational levels in CAW-TNBC cells under mPR-specific PRG treatment, while mPRs expression is variable at both the transcriptional and translational level in AAW-TNBC cells, making these genes great candidate biomarkers for TNBC sub-type classification (Fig. 5D). Another major finding is that mPRs are localized within the nucleus in both TNBC cells (Fig. 5A–C and Suppl. Fig. 4A1-6), and exhibit capabilities of shuttling between the nucleus and cytosol (Fig. 5B), identical to the cellular compartmentations of other steroid receptors, making mPRs a new type of PRG receptors functionally similar to nPRs. Utilizing multiple AAW- and CAW-TNBC tumor datasets and a systemic biology pipeline, we identified 6 new AAW-TNBC candidate biomarkers (Table 1) that will have future potential clinical applications for AAW-TNBCs. Furthermore, this project provides new insights into the CmP signaling network in TNBC tumorigenesis, which may revolutionize the current concepts and molecular mechanisms of tumorigenesis in AAW-TNBC, and reproductive cancers in general, leading to new therapeutic strategies.

3.1 Novel discoveries in the CmP signaling network with potential revolutionary impacts

The CmP signaling network and mPRs subcellular compartmentation. mPR expression patterns in many tissues and cell lines, and the possible involvement of mPRs in various biological processes have been extensively studied, however, the precise functions of mPRs remain controversial [73], especially regarding their subcellular compartmentation [74]. Up to date, mPRs have been agreed upon as a new type of PRG receptor, with only confirmed non-genomic PRG-actions, and while their stimulation varies drastically in a tissue- and cell-dependent fashion, debates on their subcellular compartmentation continue [74]. As mentioned earlier, mPRs were initially identified as a new family of membrane-bound PRG receptors [29], and PRG-mPRs signaling cascades, as well as their mediated cellular functions, have been investigated in various organisms and cell types [32, 33, 34, 75, 76]. However, recently, evidence of cytoplasmic localization or cytoplasmic compartmentation of mPRs in various cell types, including MB231 cells, has emerged [30, 35]. Initial pieces of evidence of cytoplasmic vesicles of mPRs [30, 35] were speculated as to the result of either technical issues/cytotoxic events for recombinant mPRs from transiently transfected constructs [30] or being found in the membrane of the endoplasmic reticulum (ER) [35]. Recently, contradicting evidence has emerged of localized recombinant mPR $\alpha$ into the ER of CHO cells, while others reported mPR $\alpha$ in the nuclear membrane of small and large luteal cells [77]. Additionally, reports showed cytoplasmic (vesicles) and nuclear membrane localization of recombinant mPR $\alpha$ in COS1 cells [78], while cytoplasmic (vesicles, ER) localization of various forms of recombinant mPRs in MB231 [35], HEK293, LLC-PK and MDCK cells [79] were also reported. The question of whether the cytoplasmic localization of mPRs is due to artifacts of transient transfection [30], clathrin-mediated endocytosis to cytoplasmic endosomes [80], lack and/or deficiency of PGRMC1 [81], or misclassification of mPRs as membrane proteins [74, 82] are still under debate. In this paper, we have discovered that mPRs are predominantly nuclear proteins and are capable of shuttling in and out of the nucleus as necessary under mPR-specific PRG actions.

Our novel finding of cytoplasmic-nuclear trafficking of mPRs will certainly revolutionize the current view of the functions and subcellular compartmentation of mPRs. Since only endogenous mPR $\beta$ was examined for this revolutionary discovery, questions about the quality and specificity of primary antibodies will certainly be raised despite their excellent and consistent outcomes in multiple experiments. After extensive examination, we found previous data from recombinant mPRs in transient transfection experiments, in various cells, that support our observed results. The nuclear compartmentation with strongly condensed localization within the nucleolus of recombinant NFlag-mPR $\alpha$ protein in HEK293 cells [79], and in MDCK cells [74] match our data of endogenous mPR $\beta$ in MB231 cells (Fig. 5A-3); furthermore, the nuclear localization of endogenous mPR $\alpha$ in M11 cells [80] and COS-1 cells [35] provide more supportive evidence.

The most common and significant characteristic shared by all steroid hormone receptors is the ability for cytoplasmic-nuclear trafficking [83]. The molecular mechanisms of nucleocytoplasmic shuttling have been well defined for AR [84], GR [85], ERs [86], and nPRs [87]. In this study, our finding that nucleocytoplasmic shuttling of PAQR8 (mPR $\beta$ ) is a dynamic event under mPR-specific PRG actions in TNBC cells is almost identical to the behaviors of other steroid hormone receptors. These findings indicate that the currently classified membrane-bound mPRs are actually a novel group of PRG receptors with similar functions as classic nPRs that can perform both genomic and non-genomic PRG-actions, with possibly more tissue- and cell-specific intracellular performance. Lastly, another convincing piece of evidence is that PGRMC1, a close relative to mPRs, has been localized to the nucleus and participates in nucleocytoplasmic shuttling after treatment with human chorionic gonadotropin (hCG) in rat granulosa cells [88], behaving the same as mPRs. Since then, PGRMC1 has been widely accepted as localizing to both the nucleus and cytosol. So far, PGRMC1 cytoplasmic-nuclear trafficking has been observed in spontaneously immortalized granulosa cells (SIGCs) [89, 90, 91], human granulosa cells [92], and human ovarian cancer cells [89, 90]. Nucleocytoplasmic shuttling was also observed [89, 90, 93] in various cells, further validating the ability of membrane PRG receptors for nucleocytoplasmic shuttling.

3.2 CmP signaling network and breast cancer

Potential involvement of mPRs in breast cancer has been suggested, however, due to the existence of both nPR( $+/-$ ) breast tumors and breast cancer cell lines, it is extremely challenging to define the role of PRG in breast cancer [94]. Non-classic mPRs are widely expressed in breast cancer cell lines and biopsies, regardless of nPR( $+/-$ ) status [16, 17, 18, 27]. Overexpression of major mPRs (PAQR5/7/8) has been associated with breast cancer development and progression [18], as well as other reproductive tumors, such as cervical and ovarian cancer [48, 95, 96] and has also been detected in diverse epithelial human ovarian cancer biopsies [95] and the HeLa cervical cancer cell line [96], suggesting a potential role for mPRs in reproductive tumor biology [21]. Our systems biology (Figs 1B, F–H, and 2B and C) and in-vitro data (Fig. 5D) certainly provide more evidence in supporting these views.

Previous reports suggested that mPR $\alpha$ is not produced in response to PRG induction [16] and there are undetectable levels of mPRα in MB231 cells [32], however, our current data demonstrated that the expression of mPRα can be induced by mPR-specific PRG actions in MB231 cells, at both the transcriptional and translational levels (Suppl. Fig. 2A and Figs 3A and 4C), suggesting that very subtle changes in either the CmP network or cell types could lead to very different outcomes. Similarly, the expression of CCM1 can also be induced by mPR-specific PRG actions in MB468 cells, at both the transcriptional and translational levels (Fig. 4C and D). These data suggest mPRα might contribute to PRG induced EMT suppression in both TNBC cells.

Potential implication for cancer therapy: PRG regulates specific functions in nPR( $-$ ) cancer cells that implicate mPRs in cancer initiation and progression [16, 27]. Data showed that PRG promotes cell proliferation through induced activation of MAPK signaling in both nPR( $+/-$ ) breast cancer cell lines, independent from nPRs [97], suggesting mPRs may play some role in this signaling pathway.

As a well-known antagonist of PRG, MIF has been regarded as a great candidate for the treatment of reproductive cancers [98] and is currently on many active interventional clinical trials [99, 100]. Some data demonstrated that elevated levels of MIF act as a growth inhibitory agent enhancing induction of apoptosis triggered by high doses of PRG in either nPR( $+/-$ ) cancer cells [101, 102]. Other reports showed that MIF significantly improves treatment efficacy of chemotherapy regimens for human ovarian carcinoma cell lines [103] and was able to inhibit the growth of nPR( $-$ ) MB231 breast cancer cells [104], suggesting MIF has anti-tumor effects on gynecological-related tumor cells [105].

Although the anti-proliferative activity of MIF in cancer cells is independent of nPRs [106], the cellular effects of MIF on proliferation are controversial, being reported as both a growth inhibitor [107] as well as a stimulator [108], depending on cell types [107, 109], dosage, and perhaps even existence or ratio of mPRs/nPRs [110]. Interestingly, MIF works synergistically with high dose PRG to inhibit endometrial cancer cell growth [101] and anticancer properties of MIF are known as an endocrine-related phenomenon [102]. In summary, the previous contradicting data with PRG, MIF, or a combination of both (mPR-specific PRG actions) might all reflect one thing in common, the final cellular effects and outcomes are dependent on cell types, dosage, and existence or ratio of mPRs/nPRs in breast cancer cells.

Our data of mPR-specific PRG actions on two nPR( $-$ ) TNBC cells are unique and interesting. PRG displayed no major change in cell migration in both AAW- and CAW-TNBC cells, while MIF appeared to enhance cell migration in both TNBC cells at early stages (Fig. 4A). The effects from mPR-specific PRG actions were in-between the individual responses only in AAW-TNBC cells also at earlier stages (Fig. 4A, bottom panel). Similarly, mPR-specific PRG actions appeared to enhance wound closure ability only in AAW-TNBC cells (Fig. 4B, bottom panel), suggesting that PRG/MIF might not be good anti-cancer candidate drugs to be used in AAW-TNBC patients. With newly emerging mPR specific agonists with higher binding affinities [111], mechanistic studies to uncover the key roles and cellular relationship of key players in the CmP network among various breast cancer cell types will be timely and exciting [106].

Potential prognostic biomarkers for AAW-TNBCs. Biomarkers have been widely used for breast cancer therapies and offer precise treatment based on tumor and patient characteristics by utilizing almost all available current cutting-edge biomedical technologies (genetics, epigenetics, genomics, proteomics, etc.) [112]. Biomarkers can be used as a prognostic tool to predict the future course of the disease and patient response to the designated therapy [113]. The treatment of metastatic breast cancer has become more complicated due to the increasing number of new therapies necessary to treat the elevated number of various histological, molecular, and cellular subgroups. More biomarkers are urgently needed to assist these new therapies, such as the treatment of patients with/without BRCA mutations [114]. The difference among breast cancer subtypes is quite large in terms of clinical relevance, patterns of gene expression, selection of therapeutic strategies, responses to treatment, and prognosis [115, 116, 117, 118, 119]. Therefore, the identification of new biomarkers for specific breast cancer subtypes is important in guiding treatment decisions and predicting prognosis [120].

With our experience in multi-omics and systems biology analysis, we utilized AAW-derived TNBC cells, to discover a total of 47 potential biomarkers synchronously regulated at both the transcriptional and translational levels, associated with tumorigenesis in AAW-derived TNBCs (Suppl. Table 6D). Utilizing publicly available clinical data of expression and prognosis, we filtered these 47 biomarkers down to 6 clinically relevant AAW-TNBC specific biomarkers with differential expression and significant survival curves in multiple TNBC databases (Table 1), resulting in two groups of biomarkers; intrinsic and inducible for AAW-TNBCs (Table 1). Despite previous reports of differentially expressed cellular markers between TNBC subtypes [24, 64, 121, 122], our data show that mPR-specific PRG-induced differential expression patterns among AAW-TNBCs is not overlapped with any previous data and is unrelated to the reported differentially expressed cellular markers in TNBC subtypes, making our newly identified biomarkers great candidates for AAW-TNBC clinical applications. The identified potential AAW-TNBC specific biomarkers (Table 1) will be the first important tool for breast cancer therapy/prevention in AAW-TNBC patients, a pioneering first step to tackle the currently alarming ethnic disparity in breast cancer mortality.

4. Conclusion

With very limited knowledge, there has been much speculation regarding the relationship between classic PRG actions via nPRs and non-classic PRG actions via mPRs. It has been reported that increased expression of mPRs often leads to activation of nuclear response elements and transcriptional changes, suggesting an integrated model of sex steroid receptor signaling where early steroid-dependent (non-genomic) effects can contribute to late nuclear (genomic) actions [123]. However, this integrated model is only viable in nPR( $+$ ) cells regarding mPRs-PRG-nPRs (mPn) signaling. In our previous study, we provided strong evidence that the CSC couples both classic PRG receptors (nPRs) and non-classic PRG receptors (mPRs/PAQRs) to form the CmPn signaling network in nPR( $+$ ) breast cancer cells [15]. In this report, we extended our findings that the CSC coordinately works with non-classic PRG receptors (mPRs/PAQRs) to form the CmP signaling network in nPR( $-$ ) breast cancer cells (Fig. 5D). Our discovery that mPRß is predominantly a nuclear protein and capable of shuttling between the nucleus and cytosol under mPR-specific PRG actions suggests that mPRs are a new type of PRG receptors, behaving the same as classic nPRs that can perform both genomic and non-genomic actions. Based on these novel findings, we provide proof of principle that the CmP signaling network is a core component of our previously defined CmPn signaling network [15]. To emphasize one of the most exciting findings in this study, utilizing comprehensive Omic approaches and clinical TNBC databases, we identified a total of 6 AAW-TNBC specific biomarkers (Table 1). Future implementation of these biomarkers among AAW-TNBC patients would have a tremendous positive impact on cancer therapies and health disparities in this highly aggressive breast cancer group in African American women.

5. Materials and methods

5.1 Cell culture, treatments, and performance assays

Cell culture and treatments. Multiple nPR( $-$ ) cell lines (MDA-MB-231, MDA-MB-468, MCF10A, and Jurkat) were cultured in RPMI1640 medium following manufacturer’s instructions (ATCC); when cells reached 80% confluency, cells were starved with FBS-free RPMI1640 medium for 3 hrs, then treated with either vehicle control (ethanol/DMSO, VEH), mifepristone (MIF, 20 $\mu$ M), progesterone (PRG, 20 $\mu$ M), or mPR-specific PRG treatment (MIF $+$ PRG, 20 $\mu$ M each). For RNA knockdown experiments, 80% confluent selected cells were transfected with a set of siRNAs, targeting specific genes(Suppl. Table 1), by RNAiMAX (Life Technologies) as described before [19, 21]. After cells were treated in various conditions, they were harvested for RNA and protein extraction to measure the expression levels of CCMs and mPRs genes as described before [19, 21].

Cell migration Assay. Cell migration was assessed using scratch assays. Briefly, a cell scratch was pressed through the confluent TNBC (triple negative breast cancer cells, MB231, MB468) cell monolayer in the plate to mark the starting line. The cells were swept away on one side of that line. Vehicle control (EtOH, DMSO), Mifepristone (MIF, 20 $\mu$ M), progesterone (PRG, 20 $\mu$ M), or mPR-specific PRG treatment (MIF $+$ PRG, 20 $\mu$ M each) were added to start the experiments. Migration was monitored temporally. The migration area was visualized using a Nikon Biostation and recorded with a high-resolution digital camera. The migration area was measured from four different fields under 20 $\times$ magnifications for each condition, time-point, and cell type. Each experiment was repeated three times. We calculated the rate of cell migration ( $R_M$ ) according to the equation: $R_M=(W_i-W_f)/t$ , where $W_i$ is the average of the initial scratch width, $W_f$ is the average of the final scratch width and t is the time span of the assay.

Wound healing assay. MB231, MB468 breast cancer cells were cultured in 24-well plates (5 $\times$ 10 ${}^4$ cells/well) to reach confluency. Then, cells were starved in FBS-free medium for 3 hrs, then a 100 $\mu$ l pipette tip was used to scratch the monolayer of confluent cells to create a middle wound groove. Cells were cultured in triplicates in FBS-free medium in the vehicle control (EtOH, DMSO), mifepristone (MIF, 20 $\mu$ M), progesterone (PRG, 20 $\mu$ M), or mPR-specific PRG treatment (MIF $+$ PRG, 20 $\mu$ M each). The wound areas were visualized using a Nikon Biostation and recorded as aforementioned. Each experiment was repeated three times. The wound area, wound coverage of total area, as well as average and standard deviation of the scratch width was performed with the aid of an imageJ plugin [124]. We calculated the percentage of wound closure with the following equation: Wound Closure % $=$ ( $A_{t=0}-A_{t=\mathrm{\Delta }t}/At=0$ )*100, where $A_{t=0}$ is the initial wound area, $A_{t=\mathrm{\Delta }t}$ is the wound area after n hours of the initial scratch, both in $\mu$ m ${}^2$ .

Cell Invasion assay. The migration of MB231 and MB468 cells were performed using 24-well transwell plates (8 $\mu$ m pore size; Millipore, Billerica, MA, USA). The cells were plated in triplicates into the upper chamber (1 $\times$ 10 ${}^5$ cells/well) with 1640 (no FBS) in the presence of mPR-specific PRG treatment (MIF $+$ PRG, 20 $\mu$ M each) or vehicle alone for 48 hrs at 37 ${}^{\circ }$ C, 5% CO ${}_2$ conditions. The bottom chambers were supplemented with 1640 containing 10% FBS. Non-invading cells remaining on the upper surface of the membrane were removed using a cotton swab, and the invading cells on the bottom surface were stained with Diff-Quick stain kit (Fisher). Invading cells were visualized using a 40 $\times$ magnification Nikon microscope and photographed with a high-resolution digital camera. The invasion value was quantified from the mean number of five different fields per condition. Each experiment was repeated three times.

5.2 Immunofluorescence (IF) of CAW-/AAW-TNBC cells under mPR-specific PRG actions

Growth of TNBC cells using chamber slides for IF applications: IF staining methods were performed as previously described [15]. Briefly, cells were grown to 80% confluency and then treated with MIF $+$ PRG (20 $\mu$ M each) over time in glass chamber slides (Nunc-Lab-Tek II) and fixed using 4% (w/v) Paraformaldehyde (PFA). Slides were washed 3X with PBT (0.2% Triton X-100) before proceeding with antigen retrieval.

Staining/mounting/sealing: Slides were treated and stained as previously described [15]. Briefly, slides were blocked with Pierce fast blocking buffer (Fisher) for 2 hours at RT. CCM1-Alexafluor® 488 (1:50), CCM3-Alexafluor® 546 (1:50), and PAQR8 (1:50) antibodies to incubate for 4 hrs at RT in the dark. Secondary antibody was added for PAQR8 stained slides only using mouse anti-rabbit IgG-CFL 555 (1:100). All antibodies used are detailed in Suppl. Table 2. Slides are DAPI stained during the mounting/sealing process (Suppl. Table 2).

Imaging and Quantification: Imaging was performed using a Nikon Eclipse Ti confocal microscope using a 60X objective lens. Quantification was done automatically using Elements Analysis software provided with the Nikon microscope. Thresholding (used for quantification) was defined and maintained throughout all images for each application to ensure no bias was applied to data and to exclude low and high outliers. Using Nikon elements analysis tools, localization of CCM1/CCM3/PAQR8 was performed using binary operations to identify the relative ratio of expressed proteins that localized inside the nucleus (overlaps of DAPI/GFP or mCherry) compared to expressed proteins that localized in the cytosol (unique GFP or mCherry without overlapping DAPI signal). Fluorescent images are quantified for CCM1 using wavelength channel 488 nm while CCM3 and PAQR8 were quantified using wavelength channel 555 nm.

5.3 RNA extraction, RT-qPCR, and RNA-seq for TNBC cells

Total RNAs were extracted with TRIZOL reagent (Invitrogen) following the manufacturer’s protocol. For cultured cells, monolayer was rinsed with ice cold PBS. Cells were lysed directly in a culture dish by adding 1 ml of TRIZOL reagent per flask and scraping with a cell scraper. The cell lysate was passed several times through a pipette and vortexed thoroughly. The quality (purity and integrity) of each RNA sample was assessed using a Bioanalyzer (Agilent) before downstream applications. All RNA-seq data were produced using Illumina HiSeq 2000; clean reads for all samples were over 99.5%; 60-80% of reads were mapped to respective reference genomes.

Real Time Quantitative PCR analysis (RT-qPCR). Real-time quantitative PCR (RT-qPCR) assays were designed using primer sets (Suppl. Table 3) and applied to quantify the RNA levels of the endogenously expressed CCMs (1, 2, 3) and mPRs (PAQR5, 6, 7, 8, 9) using Power SYBR Green Master Mix with ViiA 7 Real-Time PCR System (Applied Biosystems). RT-qPCR plates with various cell-lines were prepared using an epMotion 5075 automated liquid handling system (Eppendorf). The RT-qPCR data were analyzed with DataAssist (ABI) and Rest 2009 software (Qiagen). The relative expression level (2- $\mathrm{\Delta }$ CT) was calculated from all samples and normalized to a reference gene ( $\beta$ -actin); fold change (2- $\mathrm{\Delta }\mathrm{\Delta }$ CT) comparison was performed by further normalizing to control groups [125]. All experiments were performed with triplicates.

RNA-seq processing of files to assemble interactomes for TNBC cells: The data consisted of 8 FASTQ files, 2 PE FASTQ files for each of the four groups: 468_Veh, 468_MP treated 12 hrs, 468_MP treated 24 hrs and 468_MP treated 48 hrs. All cohorts consisted of two samples and the RNA-seq files were processed and analyzed as previously described [23]. The identified Differentially expressed Genes (DEGs) were inputted into the PANTHER [126] classification system (GeneONTOLOGY) as well as iDEP [127] (Integrated Differential Expression and Pathway Analysis) program to build signaling networks with altered KEGG pathways, GO cellular components, GO molecular functions and GO biological processes.

5.4 Protein extraction, Western blots, and proteomics analysis for TNBC cells

Protein Extraction and quality assessment: TNBC cell lines were harvested and lysed using a digital sonifier cell disruptor (Branson model 450 with model 102C converter and double step microtip) in ice cold lysis buffer containing 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.5% NP-40 (Sigma), 50 mM sodium fluoride (Sigma), 1 mM PMSF (Sigma), 1 mM dithiothreitol (Invitrogen) and 1 EDTA-free complete protease inhibitor tablet (Roche). The concentration of protein lysates was measure by Qubit assay (Invitrogen) before analysis. For proteomics analysis, proteins were prepped using the filter-aided sample preparation (FASP) method (Expedeon, San Diego, CA). Samples were reduced with 10 mM DTT for 30 min at RT and centrifuged on a spin filter at 14,000 $\times$ g for 15 min at RT and washed twice with 8 M urea in 50 mM Tris-HCl buffer. Samples were then washed 3 $\times$ with 50 mM ammonium bicarbonate. Finally, samples were digested with trypsin (Sigma-Aldrich, St. Louis, MO) and peptides were eluted using 0.1% formic acid.

Cell fractionation and Western blots. Subcellular fractionation of breast cancer cells was performed as previously described [40]. The relative expression levels of candidate proteins were measured with Western blots (WB) with equal amounts of protein lysates from different treatments or cellular fractionations loaded into Criterion precast gels for SDS-PAGE gel electrophoresis and transferred onto PVDF membranes at 4 ${}^{\circ }$ C, then probed with confirmed antibodies(Suppl. Table 2) as described before [15, 19, 21, 26, 89].

Liquid Chromatography-Tandem Mass Spectrometry (LC-MS/MS): The cell lysates were generated from the four cohorts, 468_Veh, 468_MP treated 12 hrs, 468_MP treated 24 hrs and 468_MP treated 72 hrs. All cohorts consisted of three replicates and were processed and analyzed as previously described [23]. Proteomic samples were analyzed via two statistical methods, Students $t$ -test and ANOVA. This was done to determine changes between two specific time points ( $t$ -test) and changes between all-time points (ANOVA). A cutoff of $p⩽$ 0.05 was executed to determine significance in all comparisons. A scoring system of proteins identified in at least 3 groups were used to improve data validation. Sample regulation over 12 hrs, 24 hrs and 72 hrs were processed to determine temporal expression patterns. Expression patterns were grouped to their respective fold change categories. The overlaps were inputted into the PANTHER [126] classification system (GeneONTOLOGY) as well as iDEP [127] program to build signaling networks.

Omics analysis to assemble interactomes for TNBC cells at both the transcriptional and translational levels. A Python script was created to identify shared differentially expressed genes/proteins between cohorts. The overlaps were inputted into the PANTHER [126] classification system (GeneONTOLOGY) as well as iDEP [127] as previously aforementioned. Identified overlaps were compared between proteomics and RNA-seq data. The agreements in differential expression were noted using the Python comparison script.

Meta-analysis of CHIP-seq data overlaps with AAW-TNBC cells under mPR-specific PRG actions. Four databases were used [128], from various breast cancer cells (containing AAW-TNBCs) treated with either PRG/MIF. Additionally, 11 publicly available breast cancer datasets [60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71] as well as two TCGA-BRCA databases (TCGA-BRCA and GDC-TCGA-BRCA) were compiled to validate differential expression among breast cancer subtypes for key CmP members as well as our Identified candidate biomarkers. Proteomic and RNA matching to the curated datasets was carried out using general python scripts. Gene names were used as identifiers when processing overlaps between RNA and Proteomic Analysis. Expression patterns of up or down regulation were integrated to determine transcriptional and proteomic relationships.

5.5 Prognostic effects for identified candidate biomarkers associated with a perturbed CmP signaling network

Differential expression of candidate biomarkers and key CmP members using Microarray expression data. Expression analysis was performed as previously described [15]. Briefly, microarray data containing breast cancer tumors, with nPR status determined by IHC, were analyzed using kmplot software [24]. This resulted in 925 nPR( $-$ ) and 926 nPR( $+$ ) breast cancer samples that were used to obtain expression data for the candidate biomarkers identified in this study. Utilizing publicly available microarray data (normalized) from TCGA, we also assessed expression profiles of our biomarkers as well as key CmP members using various datasets to include phenotypic filters including tissue type, sample type, receptor status, PAM50 classification, as well as other diagnostic parameters [69].

Differential expression of candidate biomarkers using RNA-seq data for AAW- and CAW-TNBCs. To preliminarily evaluate basal expression of our identified candidate biomarkers in AAW- and CAW- TNBCs, we utilized publicly available RNA-seq data comparing expression levels in 23 AAW- and 19 CAW-TNBC samples that were used as filter 1 to obtain expression data for the candidate biomarkers identified in this study [72]. Next, we utilized a larger cohort of AAW and CAW TNBCs with any biomarkers that passed filter 1, to evaluate basal expression from TCGA which contained 64 AAW- and 113 CAW-TNBC samples [69].

Construction of Kaplan-Meier (KM) survival curves for identified biomarkers. KM survival curves were generated as previously described [24]. Additionally, publicly available microarray data was also assessed from TCGA to integrate gene expression and clinical data simultaneously [69] to confirm the initial analysis performed using kmplot. To ensure the patients in the database reflected cohorts seen in the everyday clinical practice, we only selected cohort data similar to SEER published prevalence’s [24]. Breast cancer patients were filtered to only analyze patient samples classified as TNBC (BASAL) subtype, which reduced our initial 1,809 patients to 176/392 patient samples (depending on biomarker analyzed). Logrank $P$ values were calculated by the software [24] as well as hazard ratio (and 95% confidence intervals).

5.6 Statistical analysis

For RT-qPCR analysis, the relative RNA expression changes of CmP network genes were measured by qPCR (Fold changes). All pairwise multiple comparison procedures were analyzed using Tukey and Student’s $t$ -test. For Western blots, all pairwise multiple comparison procedures were analyzed using Tukey and Student’s $t$ -test. For cell migration/invasion and wound healing assays, two-way analysis of variance (ANOVA) was used to detect the differences in the mean temporal values among the treatment groups in migration and wound healing assays. For invasion assays, all pairwise multiple comparison procedures were analyzed using Tukey and Student’s $t$ -test. For immunofluorescence analysis, both one-way (analysis of a single time-point compared to control) and two-way (analysis of multiple time-points compared to controls) ANOVA was used to detect the differences in the mean values among the treatment groups. For transcriptomics analysis, all pairwise multiple comparison procedures were analyzed using Tukey and Student’s $t$ -test. For proteomics analysis, one-way ANOVA was used to detect the differences in the mean values among the treatment groups. All pairwise multiple comparison procedures were analyzed using Tukey and Student’s $t$ -test. For microarray/RNA-seq analysis, statistical significance was performed with students $t$ -test. All graphs/plots were constructed and produced by SigmaPlot 12.0 (Systat Software, Inc.) or GraphPad Prism 8.

Author contributions

JZ: Conceptualization, Methodology, Investigation, Writing-Original draft preparation, Writing-Reviewing and Editing; JAF: Investigation, Visualization, Software, Data curation, Validation, Writing-Reviewing and Editing, XTJ: Investigation, Visualization; BG: Software, Data curation, Validation; AP: Investigation, Writing-Original draft preparation; CCE: Data curation, Validation; EF and AMDCDLO: Investigation, Visualization.

Funding

1R21NS061191 (NINDS/NIH) and the Coldwell foundation (JZ).

Supplementary data

The supplementary files are available to download from http://dx.doi.org/10.3233/CBM-210351.

sj-pdf-1-cbm-10.3233_CBM-210351.pdf - Supplemental material

Supplemental material, sj-pdf-1-cbm-10.3233_CBM-210351.pdf

sj-xlsx-1-cbm-10.3233_CBM-210351.xlsx - Supplemental material

Supplemental material, sj-xlsx-1-cbm-10.3233_CBM-210351.xlsx

sj-xlsx-2-cbm-10.3233_CBM-210351.xlsx - Supplemental material

Supplemental material, sj-xlsx-2-cbm-10.3233_CBM-210351.xlsx

sj-xlsx-3-cbm-10.3233_CBM-210351.xlsx - Supplemental material

Supplemental material, sj-xlsx-3-cbm-10.3233_CBM-210351.xlsx

sj-xlsx-4-cbm-10.3233_CBM-210351.xlsx - Supplemental material

Supplemental material, sj-xlsx-4-cbm-10.3233_CBM-210351.xlsx

sj-xlsx-5-cbm-10.3233_CBM-210351.xlsx - Supplemental material

Supplemental material, sj-xlsx-5-cbm-10.3233_CBM-210351.xlsx

sj-xlsx-6-cbm-10.3233_CBM-210351.xlsx - Supplemental material

Supplemental material, sj-xlsx-6-cbm-10.3233_CBM-210351.xlsx

sj-xlsx-7-cbm-10.3233_CBM-210351.xlsx - Supplemental material

Supplemental material, sj-xlsx-7-cbm-10.3233_CBM-210351.xlsx

sj-xlsx-8-cbm-10.3233_CBM-210351.xlsx - Supplemental material

Supplemental material, sj-xlsx-8-cbm-10.3233_CBM-210351.xlsx

sj-xlsx-9-cbm-10.3233_CBM-210351.xlsx - Supplemental material

Supplemental material, sj-xlsx-9-cbm-10.3233_CBM-210351.xlsx