Breast cancer cell secretome analysis to decipher miRNA regulating the tumor microenvironment and discover potential biomarkers

MicroRNA (miRNA/miR) 526 b- and miR655-overexpressed tumor cell-free secretions regulate the breast cancer tumor microenvironment (TME) by promoting tumor-associated angiogenesis, oxidative stress, and hypoxic responses. Additionally, premature miRNA (pri-miR526b and pri-miR655) are established breast cancer blood biomarkers. However, the mechanisms of how these miRNAs regulate the TME has yet to be investigated. Mass spectrometry analysis of miRNA-overexpressed cell lines MCF7-miR526b, MCF7-miR655, and miRNA-low MCF7-Mock cell-free secretomes identified 34 differentially expressed proteins coded by eight genes. In both miRNA-high cell secretomes, four markers are upregulated: YWHAB, SFN, TXNDC12, and MYL6B, and four are downregulated: PEA15, PRDX4, PSMB6, and FN1. All upregulated marker transcripts are significantly high in both total cellular RNA pool and cell-free secretions of miRNA-high cell lines, validated with quantitative RT-PCR. Bioinformatics tools were used to investigate these markers' roles in breast cancer. These markers' top gene ontology functions are related to apoptosis, oxidative stress, membrane transport, and motility supporting oncogenic miR526b- and miR655-induced functions. Gene transcription factor analysis tools were used to show how these miRNAs regulate the expression of each secretory marker. Data extracted from the Human Protein Atlas showed that YWHAB, SFN, and TXNDC12 expression could distinguish early and late-stage breast cancer in various breast cancer subtypes and are associated with poor patient survival. Additionally, immunohistochemistry analysis showed the expression of each marker in breast tumors. A stronger correlation between miRNA clusters and upregulated secretory markers gene expression was found in the luminal A tumor subtype. YWHAB, SFN, and MYL6B are upregulated in breast cancer patient's blood, showing biomarker potential. Of these identified novel miRNA secretory markers, SFN and YWHAB successfully passed all validations and are the best candidates to further investigate their roles in miRNA associated TME regulation. Also, these markers show the potential to serve as blood-based breast cancer biomarkers, especially for luminal-A subtypes.


Introduction
In 2021, breast cancer surpassed lung cancer as the most diagnosed cancer worldwide, accounting for 11.7% of all cancer cases [1]. In addition, breast cancer-related fatalities continue to rise globally, with a 9.3% increase from 2018 to 2020 [1,2]. Early detection can improve breast cancer patient survival rate to 99% [3]. Mammograms are the most effective, affordable, and highly sensitive breast cancer screening procedure used globally. However, routine mammogram screening begins at age 50 in most countries [4]. With the incidence of breast cancer in younger populations increasing, finding accessible, minimally invasive, and sensitive early detection biomarkers is a global need. Blood-based biomarkers can serve as sensitive and specific detection tools to diagnose breast cancer in the early stages.
The complexity of breast cancer has led to the development of specialized treatment regimens that are dependent on tumor stage, grade, and the presence or absence of hormone receptors (HR; estrogen receptor (ER) and progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2)). The majority of diagnosed breast tumors (68%) are subtype luminal A (HR+/HER2-), followed by 10% that are triple-negative (HR-/HER2-), 10% are luminal B (HR+/HER2+), and 4% are HER2-enriched (HR-/HER2+) [5]. Improvement in patient survival is possible with precision medication and personalized therapy [6].
During oncogenic transformation, the tumor cell secretes various growth factors, ligands, and metabolites, to regulate diverse autocrine and paracrine regulation, which alters the tumor microenvironment (TME) to promote metastasis [7,13]. The secretome is the collection of proteins secreted by cells into extracellular space. It makes up a substantial amount (13-20%) of the human proteome and plays essential roles in cell migration, cell signaling, and cell-cell communication [14]. miRNA-overexpressed cell secretions and metabolites induced oxidative stress, tumor-associated angiogenesis, and lymphangiogenesis and enhanced hypoxic responses [10][11][12]. However, the compositions of miR526b and miR655 secretomes are unknown. Additionally, secretory proteins found in bodily fluids can be detected in blood and are potential biomarkers [14,15]. Both miR526b and miR655 expressions in human breast tumors are associated with poor breast cancer patient survival [8,9]. Premature miRNAs, pri-miR526b and pri-miR655, can be detected in blood plasma, showing a signature of highly sensitive breast cancer biomarkers, and pri-miR526b can distinguish stage one tumor compared to benign samples; hence it is an early diagnostic breast cancer biomarker [16]. Here, the secretomes of miRNA-overexpressed cell lines were systematically analyzed to identify novel secretory markers regulating the TME in breast cancer and to investigate their biomarker potential.  (BSP). Green pie chart indicates all human protein-coding genes that are listed in the BSP. Yellow pie chart shows which proteincoding genes from this study overlap with the BSP list. Gray pieces in C and D indicate listed markers not matched or found in our list.

Data curation and threshold determination
To identify the most important differentially expressed secretory proteins, a >1.5/<-1.5 log 2 fold change (FC) and >0.3 -log 10 pvalue threshold was established. This threshold determined 96 proteins coded by 32 genes in the MCF7-miR526b secretome ( Fig. 2A) and 95 proteins coded by 29 genes in the MCF7-miR655 secretome (Fig. 2B). In both miRNA secretomes, a total of 136 proteins coded by 39 genes (Fig. 2C) were identified. Of the 39 protein-coding genes, 13 are upregulated, and 26 are downregulated in both miRNAsecretomes (Fig. 2C). Interestingly, many protein-coding genes were from the same protein family, including 13 from the H2A histone family, five from Septin, four from 14-3-3, and three from thioredoxin ( Figure S1A). Most protein families are grouped on the agglomerative hierarchical clustering heatmap, including histone H2A, Septin, and 14-3-3 (Fig. 2C). This is further confirmed by these protein families' protein-protein interaction network, as they are also grouped together ( Figure S1B).
Next, differential secretory proteins were analyzed to ensure they could be found in the breast-specific proteome (BSP). There are 19,670 human protein-coding genes, of which 14,227 are listed in the BSP (Fig. 2D). From the identified 39 genes in this study, 33 were found in the BSP. The six genes absent in the BSP (H2AC1, H2AC12, H2AC14, H2AC18, OBSCN, and SEPT14) were excluded from further analysis.
Overall, our data curation pipeline identified eight secretome markers from the beginning 1535 secreted proteins, as summarized in Fig. 3C. These eight secretome markers present in both miRNA-secretomes will be further investigated through bioinformatic approaches and validated via mRNA gene expression, examining their relationships with miR526b and miR655 and evaluating their blood-based biomarker potential.

Gene ontology analysis of eight identified secretome markers
First, the general functions and individual cellular components, biological processes, and molecular functions of each secretome marker were determined using GO analysis ( Fig. 4A and Table S1). Interestingly, many secretome markers' functions and GO enrichments overlapped with miR526b-and miR655-induced breast cancer phenotypes, such as induction of oxidative stress as reported earlier [10]. For example, identified PRDX4 and TXNDC12 are associated with oxidative stress response. TXNDC12 is upregulated in both miRNA secretomes. It is a negative regulator of the endoplasmic reticulum stress-induced intrinsic apoptotic signaling pathways, allowing cells to survive while stressed. PRDX4 is downregulated in both miRNA secretomes and is an antioxidant enzyme that neutralizes oxygen species, protects cells against oxidative stress, and is a key protein regulating cellular redox homeostasis. This can be linked to both miRNAs upregulating oxidative stress in breast cancer. Previously, we have shown that treating poorly metastatic tumor cell MCF7 with cell-free conditioned media from miRNA-overexpressed cells enhanced oxidative stress [10].
Next, the eight secretome markers were analyzed to determine significantly enriched GO functions. This identified four cellular components (Fig. 4B); four biological processes (Fig. 4C); one molecular function, phosphoprotein binding (Fig. 4D); and 30 different cell regulatory pathways ( Figure S2). The top ten pathways are shown in Fig. 4E. All identified cellular components are related to the extracellular region, confirming these secretome marker roles in TME regulation. Additionally, three biological processes are all related to apoptotic regulation, and one is involved in redox reactions. These processes are altered by miR526b and miR655 in breast cancer. The most significantly enriched pathway identified is FLT3 signaling, which is involved in the cell differentiation, proliferation, and survival of dendritic and hematopoietic progenitor cells [18]. Downstream of FLT3 is the MAPK pathway, which is also a significant pathway listed on the GO function. The MAPK pathway responds to various extracellular stimuli to regulate intracellular processes such as metabolism, proliferation, and apoptosis [18]. The RAF/MAP kinase cascade is highly mutated in cancer, RAS mutations are found in ~30% of all human cancers, and the most active activator of this pathway is BRAF, which is reported to be mutated in ~7% of cancers, including breast cancer [19]. FOXO transcription factors (FOXO1, FOXO3, FOXO4) bind to 14-3-3 proteins, such as YWHAB and SFN, allowing their retention in the cytosol [20]. In this study, SFN and YWHAB are upregulated in both miRNA secretomes, indicating the involvement of both markers in cell proliferation and survival. Activation of BAD and translocation to mitochondria are sequestered by 14-3-3 proteins after Akt1 phosphorylation [20]. We reported earlier that expression and function of both miR526b and miR655 are regulated by the PI3K/Akt signaling pathway in breast cancer [8,9].

Validation of secretome markers gene expression in MCF7-miRNA-high cells
Secretome marker gene expression in MCF7-miR526b and MCF7-miR655 was analyzed to determine if changes in gene expression are similar to secretory protein expression. All upregulated secretome markers are significantly upregulated in MCF7-miR526b and MCF7-miR655 cells at the mRNA level compared to the MCF7-Mock. There is a log 2 FC of 2.53 and 3.73 in YWHAB expression (Fig. 5A), a log 2 FC of 2.78 and 3.02 in MYL6B expression (Fig. 5B), a log 2 FC of 3.77 and 3.01 in TXNDC12 expression (Fig. 5C), and a log 2 FC of 4.31 and 3.23 in SFN expression (Fig. 5D), for MCF7-miR526b and MCF7-miR655 respectively.
All markers downregulated in miRNA-high cell secretions at the protein level are also downregulated in MCF7-miRNA-high cells at the mRNA level. All markers are significantly downregulated in miRNA-overexpressed cells, except for FN1 in MCF7-miR526b, with a log 2 FC of − 0.38. However, in MCF7-miR655 cells, FN1 expression is significantly lower, with a log 2 FC of − 1.12 (Fig. 5E). In addition, there is a log 2 FC of − 1.21 and − 2.06 in PSMB6 expression (Fig. 5F), a log 2 FC of − 0.96 and − 2.42 in PRDX4 expression (Fig. 5G), and a log 2 FC of − 1.49 and − 1.00 in PEA15 expression (Fig. 5H), for MCF7-miR526b and MCF7-miR655 respectively. Overall, all markers showed the same pattern of mRNA expression as protein secretions in MCF7-miR526b and MCF7-miR655 cells.

Secretome markers gene expression in cell-free secretions
Next, RNA was extracted from the cell-free secretions of all cell lines, and secretome markers were measured at the mRNA level in MCF7-Mock compared to miR526b-and miR655-overexpressing MCF7 cell lines. All upregulated secretome markers are significantly upregulated in cell-free secretions except TXNDC12 in MCF7-miR655 and SFN in MCF7-miR526b cells. There is a log 2 FC of 2.51 and 0.83 in YWHAB expression (  where an antibody has been bound to its corresponding antigen, and blue represents counterstaining. miR655 respectively. Among the downregulated secretory protein markers, PRDX4 and PEA15 are significantly downregulated in MCF7-miRNA-high cell-free secretions. In MCF7-miR526b, there is a log 2 FC of − 7.29 log 2 in PRDX4 expression, and a log 2 FC of − 1.77 in PEA15 expression ( Fig. 6E-F). In MCF7-miR655, there is a log 2 FC of − 1.55 in PRDX4 expression, and a log 2 FC of − 1.20 in PEA15 expression. PSMB6 is only marginally downregulated in MCF7-miR526b, with a log 2 FC of − 0.41 (Fig. 6G). FN1 is upregulated in both MCF7- miR526b and MCF7-miR655 cell-free secretions with a log 2 FC of 5.69 and 3.15 respectively (Fig. 6H). Overall, all identified upregulated secretory markers and two downregulated markers, PRDX4 and PEA15, showed the same expression pattern at the transcript level, as their respective protein abundance in miRNA-high cell-free secretions.

In silico analysis of miRNA regulating secretome markers gene expressions
To establish a regulatory connection between miRNA and differential secretion of peptides, known transcription factors (TFs) of the eight markers were cross-referenced with predicted targets of miRNAs. Altogether, miR526b has 4133 predicted targets, and miR655 has 3264 predicted targets, of which 1252 are common targets of both miRNAs (Fig. 7A). Of all targets, TXNDC12 is a predicted direct target of miR655, and a predicted indirect target of miR526b. However, TXNDC12 is upregulated in both miRNA secretomes and at the transcript level. This indicates a post-transcriptional regulation of this gene. Therefore, transcription factors negatively regulating TXNDC12 were investigated, identifying common targets of both miRNAs are NANOG and KLF10 (Fig. 7B). It is plausible that miRNA downregulates negative regulator TFs NANOG and KLF10, and the absence of these TFs resulted in the upregulation of TXNDC12 expression. To establish a regulatory connection between miRNA and secretory markers, the indirect relationship between miRNA and marker's TFs was evaluated. For example, upregulated secretome marker MYL6B has two positive regulation TFs, MYC and SP3, and one negative regulation TF, MECP2, and all these TFs are targets of miR526b and miR655. Since MYL6B is upregulated in both miRNAs' secretomes, we predict that MECP2 is targeted by both miRNAs and is downregulated. The absence of a negative regulator leads to MYL6B's increased gene and protein expressions. Alternatively, downregulated secretome marker PRDX4 has three positive regulation TFs, FOXP1, MYC, and NANOG, and two negative regulation TFs, FOXP1 and ZNF148, that are predicted targets of both miRNAs. Since PRDX4 is downregulated in miRNA-high secretomes, hypothetically, all three positive regulation TFs could be targeted by miR526b and miR655, resulting in PRDX4 downregulation. There are no TFs of SFN that are common predicted targets of both miRNAs. Individually however, seven predicted miR526b targets ( Figure S3A), and six predicted miR655 targets ( Figure S3B) are TFs regulating SFN. In MCF7-miR526b, SFN could be upregulated through the downregulation of one or more negative regulation TFs, ETS2, SOX4, THRB, ESR1, and POU5F1. Similarly, in MCF7-miR655, SFN could be upregulated via miR655 targeting negative regulator TFs, THRA, SRF, and HNF4A. These predicted targets need to be further validated.

Immunohistochemistry analysis of secretome markers
We wanted to further investigate the presence of these secretory markers within human breast tissue, so an immunohistochemistry analysis was conducted. This also confirms the protein expression of secretome markers in breast tissue compared to normal tissue. One to three normal tissues and 10-12 tumor tissue data were available in the Human Protein Atlas (HPA). For SFN, which is upregulated in the secretome, five of 12 breast tumor tissues show high or low positive intensity staining, compared to normal tissue, which did not show any staining (Fig. 8A). All breast tumor samples stained positive for YWHAB (at high or medium levels) compared to medium staining in normal tissue (Fig. 8B), thus confirming secretome results. For PRDX4, eight of 11 samples show low staining in tumor tissue, compared to medium intensity in all normal samples (Fig. 8C). PEA15 protein is downregulated in both miRNA secretomes, interestingly, in HPA data, no breast cancer tissues showed PEA15 expression, but all three control samples showed low PEA15 expression (Fig. 8D), supporting PEA15 secretome expression. Another downregulated secretome marker, FN1, showed no staining in breast tumor tissue, while normal tissues showed low-intensity staining (Fig. 8E). PSMB6 and MYL6B have the same expression in breast tumors and normal tissues ( Fig. 8F-G). TXNDC12 has no immunohistochemistry data available. Overall, SFN, YWHAB, PRDX4, FN1, and PEA15 immunohistochemistry data supported the trends seen in the secretome data.

Gene expression analysis in breast cancer tissue
Next, the expression of the eight secretome markers were analyzed (data extracted from GEPIA2) in breast tumor tissue (n = 1085) compared to normal tissues (n = 291) to investigate whether their secretory protein expression matches mRNA expression in breast tumor tissue. mRNA expression was measured at the level of Transcripts Per Million (TPM), log 2 (TPM+1). YWHAB is found at 7.51 log 2 (TPM+1) in breast tumor tissue compared to 6.80 log 2 (TPM+1) in normal tissue samples (Fig. 9A). In tumor tissue, TXNDC12 is 5.40 log 2 (TPM+1) in comparison to 4.83 log 2 (TPM+1) in normal tissue (Fig. 9B). MYL6B is 5.74 log 2 (TPM+1) in tumor tissue in contrast to 5.25 log 2 (TPM+1) in normal tissue (Fig. 9C). In tumor tissue, SFN expression is measured at 5.47 log 2 (TPM+1) compared to 4.21 log 2 (TPM+1) in normal tissue (Fig. 9D). However, downregulated secretome markers FN1, PSMB6, PRDX4, and PEA15 are also upregulated at the mRNA level in breast tumors (Fig. 9E-H). The observed downregulation of secretome markers in miRNA-high secretomes could be due to miRNA targeting positive regulator TFs of markers. All upregulated secretory protein markers show the same expression trends in breast cancer tumor tissue at the mRNA level.

Correlation between secretome markers gene expression and miRNAs cluster expressions in different breast cancer subtypes
To determine the level of correlation between miRNA and secretome markers in different breast cancer subtypes, miR526b and miR655 cluster expressions and secretome markers mRNA expression data were extracted from cBioportal. 283 matched tumor tissue samples were available for both miRNA clusters expression and secretome markers mRNA expression (Fig. 9I-J). The Pearson correlation coefficient between miRNA cluster expression and marker mRNA expression was measured in non-stratified tumor samples, the luminal A subtype, and all remaining tumor subtypes.
In non-stratified samples, miR526b cluster expression showed a positive correlation with SFN and a negative correlation with FN1 (Fig. 9I), confirming secretome results. However, in the luminal A subtype, all four upregulated secretome markers, YWHAB, TXNDC12, MYL6B, and SFN, have a significant positive correlation with miR526bs cluster and downregulated secretome marker PEA15 showed a significant negative correlation, which supports the secretome results.
For miR655 cluster, in non-stratified samples, TXNDC12 and FN1 are positively correlated with the miR655 cluster (Fig. 9J). In the luminal A subtype, YWHAB, MYL6B, TXNDC12, and SFN showed a significant positive correlation to the miR655 cluster, and PSMB6, PRDX4, and PEA15 showed a significant negative correlation, which supports secretome results. Overall, four upregulated secretome markers (YWHAB, SFN, TXNDC12, MYL6B) showed a statistically significant positive correlation and one downregulated secretome marker (PEA15) showed a statistically significant negative correlation with both miRNA clusters in luminal A tumors. Hence, these markers play a key role in miRNA-induced tumor metastasis in luminal A breast cancer.

Secretome markers associated with breast cancer patient survival
Kaplan-Meier survival analysis was conducted using HPA data for secretome markers in non-stratified (all stages I-IV), early (stages I & II), and late stages (III & IV) tumor samples to examine if the expression of any secretome marker is associated with poor patient survival (Fig. 10). Decreased SFN expression led to poor survival in non-stratified samples (p = 0.056) (Fig. 10A) and significantly reduced survival in early stages (p = 0.018) (Fig. 10B). In contrast, in late tumor stages, high SFN expression led to significantly reduced survival (p = 0.0039) (Fig. 10C), indicating the association of SFN with tumor progression. High YWHAB expression showed marginally reduced survival in non-stratified samples (p = 0.092) (Fig. 10D) but significantly reduced survival in early-stage tumors (p = 0.020) (Fig. 10E). Hence, it is a marker associated with the early onset of disease. TXNDC12 expression is not correlated with breast cancer patient survival in non-stratified samples (p = 0.95) (Fig. 10F). However, high TXNDC12 expression in late-stage tumors is significantly associated with poor survival (p = 0.00087) (Fig. 10G), confirming its role in tumor metastasis. Low MYL6B expression showed slightly reduced survival in non-stratified samples (p = 0.09) (Fig. 10H) and significantly reduced survival in early stages (p = 0.044) (Fig. 10I), considering this is a marker associated with cell adhesion, low expression of MYL6B in early stages might promote metastasis. High PRDX4 expression led to significantly decreased survival in non-stratified samples (p = 0.018) (Fig. 10J), and also in late stages (p = 0.00049) (Fig. 10K), this marker is associated with oxidative stress, conforming its associated with breast cancer progression and metastasis. Overall, SFN, YWHAB, TXNDC12, MYL6B, and PRDX4 expressions are significantly associated with breast cancer patient survival. Detailed survival analysis for all markers in all stages can be found in Figure S4.

Secretome marker expression in blood
The mRNA expression of each secretome marker was explored in the blood exosomes of breast cancer patients (n = 140) compared to healthy controls (n = 118) to examine their potential as blood-based biomarkers (Fig. 11). Data was extracted from ExoRBase2.0.
MYL6B expression is significantly upregulated in breast cancer patient blood at 3.42 log 2 (TPM+1) compared to 3.06 log 2 (TPM+1) in healthy participants' blood samples (Fig. 11A). In breast cancer patient blood, YWHAB has significantly higher expression at 7.91 log 2 (TPM+1) compared to 7.84 log 2 (TPM+1) in healthy samples (Fig. 11B). SFN is determined to be 0.61 log 2 (TPM+1) in cancer samples blood in contrast to 0.47 log 2 (TPM+1) in healthy samples (Fig. 11C). PEA15 and PSMB6 are significantly upregulated in cancer samples compared to healthy controls, but this is opposite to their secretory protein expression (Fig. 11D-E). TXNDC12, FN1, and PRDX4 have conflicting expressions in breast cancer patient blood compared to control samples in relation to secretory protein expression (Fig. 11F-H). Thus, only MYL6B, YWHAB, and SFN expression in breast cancer blood reflected secretory protein expression.
Additionally, using the Human Proteome Organizations (HUPO) and Human Proteome and Plasma Proteome Projects data, all eight secretome markers could be found in human blood at the highest level of protein evidence (protein level) and the highest level of certainty (canonical) (Table S2). This indicates all these markers are potential blood biomarkers. A summary of secretome markers validated in vitro and passed bioinformatics assays is shown in Table 1. These results show evidence of identified markers' roles as breast cancer tumor markers associated with TME regulation, and selective secretome markers should be further investigated in various tumor stages and grades in blood samples to confirm their potential as biomarkers.

Discussion
We have identified that cell-free miRNAs miR526b and miR655, and miRNA-overexpressed tumor cell secretory proteins change cellular phenotypes in the TME. Therefore, analysis of miR526b-and miR655-overexpressing cell secretomes in the luminal A MCF7 breast cancer cell line might help decipher the mechanisms of miRNA regulating the TME and identify breast cancer biomarker candidates. However, extensive secretome analysis can impose difficulties, as many extracellular proteins are signaling molecules found at low levels, and thus, some molecules of interest may be lost during data curation and analysis. Our platform combining nanohigh-performance liquid chromatography with large-sensitivity mass spectrometry ensured an in-depth, sensitive secretome analysis. After systematic data curation, four upregulated (YWHAB, TXNDC12, MYL6B, SFN) and four downregulated (FN1, PSMB6, PRDX4, PEA15) markers in miRNA-overexpressed tumor secretomes were identified. miR526b and miR655 have been shown to induce oxidative stress by overproduction of ROS and promoting ROS levels during hypoxia in miRNA-overexpressed cells [10,12]. Overproduction of ROS disrupts tissue homeostasis, causes DNA damage, and often triggers apoptosis. The hypoxic core in a growing tumor influences the apoptotic pathways; however, in miR526b and miR655-overexpressed breast cancer cells, hypoxia further promotes oxidative stress, cell migration, and tube formation [12]. Here, we show that the most enriched biological processes regulated by the eight secretome markers are related to apoptosis regulation and cell redox homeostasis. Secretome markers may regulate ROS levels, cellular response to hypoxia, and apoptosis in favor of tumor cell survival, supporting miR526b and miR655 as mediators of these processes in breast cancer.
In our study, YWHAB and SFN are upregulated in both miRNA secretomes. YWHAB and SFN are both members of the 14-3-3 protein family. These proteins regulate cell cycle machinery and signaling pathways and modulate the activity of their binding partners [21]. YWHAB has an oncogenic role in cancer [22], and SFN (14-3-3σ) is typically documented as a tumor suppressor [22], although the roles of SFN in breast cancer are not so clear [23][24][25]. YWHAB is a non-classically secreted protein, and SFN is classically secreted. These 14-3-3 proteins could be secreted in MCF7-miRNA-high cells by altering cell machinery and signaling pathways, as 14-3-3 proteins have an active role in protein transport and can activate the Wnt signaling pathway [26,27]. This, in turn, promotes cell survival during oxidative stress and hypoxia by inhibiting pro-apoptotic pathways stabilizing hypoxia-inducible factor 1 alpha (HIF-1α), allowing the expression of genes that promote EMT and metastasis [20]. Thus, 14-3-3 proteins may promote miR526b-and miR655-induced cell proliferation and EMT mechanisms in the TME.

Table 1
Summary of secretome markers in vitro and bioinformatic translational assays in breast cancer that abided by protein expression from this study. YWHAB and SFN passed all in vitro and bioinformatic analysis.
YWHAB and SFN are upregulated in MCF7-miRNA-high cells and cell-free secretions at the mRNA level. These two markers show higher mRNA and protein expression in breast tissue compared to non-cancerous tissue. The gene expression of both has positive correlations to miRNA cluster expression in luminal A breast cancer. High YWHAB expression could significantly predict a worse prognosis in early-stage breast cancer, and its overexpression in luminal A breast cancer has led to a worse prognosis [28]. So, YWHAB is a marker associated with breast cancer metastasis. SFN has a tumor suppressor role, so the loss of SFN is required to initiate metastasis, and at a progressive stage, mutated SFN overexpression might behave more like an oncogene; hence higher expression of SFN promotes metastasis. We found low expression of SFN is associated with early-stage patient overall survival, whereas higher expression is associated with late-stage tumors. Both YWHAB and SFN are upregulated in breast cancer blood samples and could be found in blood plasma at the highest level of protein certainty. Hence, YWHAB and SFN are strong blood-based breast cancer biomarker candidates and tools for understanding TME regulation.
TXNDC12 is a classically secreted protein that is upregulated in MCF7-miRNA-high secretomes. Transcripts of TXNDC12 are significantly upregulated in MCF7-miRNA-high cells and cell-free secretions. This may be attributed to miR526b and miR655 targeting negative regulator TFs NANOG and KLF10 of TXNDC12. TXNDC12 has roles in redox regulations, defense against oxidative stress, and regulation of transcription factors TFs [29]. Upregulation of TXNDC12 inhibits apoptosis by endoplasmic reticulum stress-inducing agents and promotes EMT and metastasis in many epithelial cancers [30,31]. Additionally, since both miRNAs and TXNDC12 promote oxidative stress and EMT, TXNDC12 likely collaborates with miRNAs to promote breast cancer cell migration, invasion, and metastasis. While TXNDC12 has not been studied in breast cancer, two gene family members of TXNDC12, ARG2 and ARG3, are known serum-based breast cancer biomarkers [32]. In breast cancer tissue, TXNDC12 mRNA expression is higher compared to normal samples and is significantly correlated with both miRNA clusters in luminal A breast cancer. High TXNDC12 expression could differentiate late-stage metastatic tumors. Thus, TXNDC12 shows potential as a prognostic biomarker for metastatic breast cancer and a major regulator of miRNA functions.
MYL6B is a classically secreted protein found upregulated in MCF7-miRNA-high cell secretomes. In miRNA-high cells and cell-free secretions, MYL6B is significantly upregulated, potentially from both miRNAs targeting negative regulator TF MECP2. MYL6B is an essential light chain subunit for myosin motor proteins and regulates cell mobility functions [33]. Both miRNA-overexpression enhanced cell migration and invasion of tumor cells and cell migration and angiogenesis properties of HUVEC cells in the TME, so we speculate that MYL6B might be driving these phenotypes for miR526b and miR655. Although no studies have investigated MYL6B in breast cancer, MYL6B overexpression promotes EMT in rectal adenocarcinoma, and in other epithelial cancers, high MYL6B expression is reported as a predictor of poor survival [34]. MYL6B mRNA expression is higher in breast tumor tissue than in normal tissue, and in luminal A breast cancer, MYL6B expression shows a significant positive correlation with miRNA cluster expression. Similarly, we reported both miRNA expression promotes metastasis in luminal-A breast cancer subtype. MYL6B is significantly upregulated in breast cancer patient blood and can be identified at the highest level of protein certainty in human plasma. Thus, the potential of MYL6B as a blood-based breast cancer biomarker should be further explored.
MYC is a regulator of seven secretome markers except for SFN. MYC regulates ~15% of human genes, and in breast cancer, MYC targets genes that participate in cancer stem cell (CSC) regulation, angiogenesis, cell growth, and transformation [35]. Both miR526b and miR655 enhanced CSC phenotypes in both luminal A and HER2 positive breast cancer [8,9], and MYC expression in MCF7-miR655 cells is marginally upregulated [9]; however, MYC is a predicted target of miR655. We speculate that other downstream effector molecules may compensate for MYC expression in miRNA-overexpressed cells. MYC has been deemed both a positive and negative regulator of YWHAB and FN1, hence depending on tissue and tumor type, the regulation of MYC varies. Thus, the function of MYC in the breast cancer secretome in miRNA-high breast tumor cells requires further investigation.
All downregulated secretory markers have decreased cellular gene expression in all miRNA-high cells. One of them, PEA15, is downregulated at both nuclear and cell-free secretion mRNA expression levels. This could be due to miRNAs indirectly targeting TFs of respective genes. Three downregulated markers, PEA15, PRDX4, and FN1, also showed low protein expression in breast tumors, as confirmed with immunohistochemistry staining. Low expression of PEA15 is strongly correlated to both miRNA clusters expression in luminal A breast cancer. Therefore, PEA15 is a negative regulator of breast cancer; hence it is downregulated in the secretions of MCF7-miR526b and MCF7-miR655.
PRDX4 protects cells during oxidative stress [36], and hence decreased expression of PRDX4 in miRNA-high tumors secretome could contribute to increased oxidative stress in the TME. Furthermore, FN1 fragments are known to inhibit angiogenesis [37]. Thus reduced FN1 expression in miRNA-high cells may add to the angiogenesis-promoting phenotype of miRNA-high cells. The downregulation of four markers in miRNA-secretomes could be related to miRNA epigenetically regulating these marker expressions in breast cancer. Further investigation of downregulated marker expression in breast cancer with miRNA context is needed. All identified secretome markers are prognostic markers in other epithelial cancers and can be detected in the blood. Therefore, identified upregulated secretory markers YWHAB, SFN, TXNDC12, and MYL6B show strong potential in establishing a battery of breast cancer biomarkers alongside pri-miR526b and pri-miR655. To our knowledge, this is the first time secretory proteins from miR526b-and miR655-overexpressing ER-positive luminal A cells have been identified and investigated to decipher the mechanisms of miRNA TME regulation. In addition, most secretome marker expressions correlate with miRNA cluster expressions in luminal A breast cancer, strengthening our hypothesis that upregulated secretome markers might be potential blood-based biomarkers for ER-positive luminal A breast cancer.
Many low abundance but vital secretory proteins were excluded during data curation to find common secretory proteins in both miRNA-secretomes. In the future, each MCF7-miRNA-secretome will be investigated separately to identify specific secretory proteins regulated by each miRNA. Additionally, we will further investigate the roles of the eight identified secretome markers in breast cancer progression and metastasis to establish their roles as potential drug targets. Each of the novel secretory markers will be tested in matched breast cancer patient plasma and tissue samples to determine their potential as blood-based breast cancer biomarkers. These secretome markers might increase the sensitivity and specificity of breast cancer early detection in combination with pri-miRNAs, allowing us to develop a battery of blood-based breast cancer biomarkers for luminal A or ER-positive breast cancers.

Collection of conditioned media for secretome analysis
Once cell confluency reached 90%, all cells were washed with 1x phosphate-buffered saline (PBS) (Gibco, ON, Canada) to remove any trace of complete media. Cells were then serum-starved with basal media for 24 h, and the conditioned media, which contains all the secretory proteins and cell metabolites, was collected.

Preparation of proteins to be analyzed by nanoHPLC-MS
Proteins within the conditioned media were precipitated overnight with 35% ethanol, followed by acidification with sodium acetate and the addition of a digestion buffer (1% sodium deoxycholate and 50 mM NH 4 HCO 3 ). Precipitated proteins were quantified by BCA Protein Assay Kit (Pierce, Rockford, IL) with at least 100 μg protein per cell line. Peptides were isolated by stage tip purification before analysis by nanoHPLC-MS (Agilent 6530 Accurate-Mass Q-TOF LC/MS, Santa Clara, CA). In a single run, there were two experimental replicates of each sample. So, with three biological replicates, n = 6 sets of data for each cell line was generated.

Identification of secreted protein IDs
NanoHPLC-MS used Mascot Server (version 2.6) to identify peptides [38]. A full scan of peptides was quantified by MS1 filtering, extracted ion chromatogram, and verified by spectral matching (Uniprot human protein reference data file) and amino acid database search (<1% false discovery rate (FDR)). Mass spectrometry raw data files and peptide masses were analyzed using Skyline (version 20.1.0.155), which allowed us to acquire a list of protein IDs [39]. MCF7-miR526b and MCF7-miR655 cells (case) were normalized to MCF7-Mock (control), and Skyline gave their protein IDs with corresponding FCs and p-values.

GO analysis of all differentially expressed miRNA-high proteins
The IDs of all differentially expressed proteins were entered into The Gene Ontology Resource for Homo sapiens (release 2021-09-01), and data was extracted [40,41].

Threshold determination and data curation
Volcano plots for differentially expressed secreted protein from Skyline were extracted (version 20.1.0.155) [39]. We established a >1.5/<-1.5 log 2 FC and >0.3 -log 10 p-value as the threshold following similar studies with secretome analysis in other diseases [42,43]. There were no statistically significant protein IDs (p < 0.05), so the top 92nd percentile of data was considered, which roughly translates to 0.3 -log 10 p-value. Next, we submitted protein IDs to Uniprot and extracted protein names, primary gene names, and synonyms [44]. One Skyline protein ID was unmapped (no peptide ID found), and 13 Skyline protein IDs had no gene names in Uniprot. Therefore, these Skyline protein IDs were excluded from this study. Gene names were used for all differentially secreted protein IDs. If a protein ID corresponded to the same gene name, and one or both log2 FC and -log 10 p-value differed, average or mean of all IDs with the same gene name were considered. The log 2 FC and -log 10 p-value of gene names were further analyzed to identify proteins that pass through our threshold in at least one miRNA-secretome. R were used to produce the agglomerative hierarchical clustered heatmap.

Breast-specific proteome
The 14,227 human protein-coding genes within the breast-specific proteome were extracted from the HPA (version 20.1) [45] and compared to genes within our secretome threshold.

Secretome prediction methods
Classical secretome prediction method data was retrieved from the HPA (version 20.1) [45] for HPA, MDSEC, Phobius, SignalP, and SPOCTOPUS and compared genes within each method to our list of genes that were within our set threshold.
For the non-classical secretome prediction method, SecretomeP, FASTA sequences of the identified threshold protein IDs were obtained via Uniprot and submitted in SecretomeP 2.0 (December 2020) for mammalian sequences [17,44]. Genes were considered non-classically secreted following the previously established guideline of a neural network score >0.6 and odds >3 [17,46]. Secretome markers were considered classically secreted if found in classical and non-classical secretion methods [17].

Secretome marker functions and GO analysis
Secretome marker general functions were obtained by using www.GeneCards.org version 5.6.0 Build 515 [47]. Individual GO functions of each secretome marker were obtained through Uniprot (Last modified: February 2, 2021) [44] and QuickGo (GO version 2021-11-08) [48]. Shared GO of the eight secretome markers was found by analyzing all protein IDs as one quarry into the Gene Ontology Resource for Homo sapiens (release 2021-09-01) [40,41]. This obtained cellular component and Reactome pathway results. Additionally, secretome markers were analyzed with STRING database (version 11.0 b) [49] to identify biological processes and molecular functions GO.

Cellular RNA extraction
Cells were grown to 90% confluency, and RNA extraction was performed with the Qiagen miRNeasy Mini Kit (Qiagen, ON, Canada) following the manufacturer's protocols.

Cell-free RNA extraction
Once cell confluency reached 90%, all cells were washed with 1x phosphate-buffered saline (PBS) (Gibco, ON, Canada) to remove any trace of complete media. Cells were then serum-starved with basal media (FBS and penicillin-streptomycin free) for 24 h, and the conditioned media, which contained cell secretion and metabolites, was collected. The media was centrifuged for 5 min at 25 • C and 4000RPM. The supernatant was removed and placed in a new falcon tube; the remaining pellet was discarded. TRIzol was added at the same volume of sample used, and chloroform was added at half the amount of sample used. The falcon tube was centrifuged at 4000 RPM at 4 • C for 15 min. The aqueous layer was collected and combined with 1.5 times the amount of ethanol. RNA was collected using the miRNeasy Mini Kit (Qiagen, ON, Canada), following the manufacturer's protocols for cell-free RNA extraction.

cDNA synthesis
The RNA extracted from cells, and cell-free secretions were reverse transcribed into complementary DNA (cDNA) using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, MA, USA).

Quantitative real-time PCR
For quantitative real-time PCR (qRT-PCR), TaqMan Gene Expression Assays were used, with TaqMan Universal PCR Master Mix (Applied Biosystems, MA, USA) and specific control and experimental probes. All probes and primers are designed following Build GRCh38 ( 160,205,319-160,215,376), to determine gene expression in total cellular RNA level and cell-free secretions of different breast cancer cell lines. We established that RPL5 is a suitable control gene for cell-free secretions [16]. RPL5 showed consistent expression in all cell lines compared to Beta-actin (data not shown), thus, RPL5 served as a control gene for cell-free secretion qRT-PCR.

miRNA target genes and TF analysis
Predicted miR526b (hsa-mir-526 b) and miR655 (hsa-mir-655) targets were downloaded for both mature five-prime sequences from TargetScanVert (Release 7.1) [50]. Only TXNDC12 was found to be a predicted direct target of miR655. Therefore, a combined list was created, which included common targets of both miRNAs. The Enrichr database [51] was used to identify the TFs for each secretome marker. Each marker and its TFs were matched against miR526b and miR655 common targets.

Immunohistochemistry analysis
Normal and breast cancer tissues immunohistochemistry staining data were obtained from the HPA (version 20.1) [45]. For each normal sample, adipocyte, glandular cell, and myoepithelial cell parameters were combined (staining, intensity, and quantity), and the median intensity value was considered. TXNDC12 had no immunohistochemistry data available in the HPA. For most secretome markers, data for several antibodies were available. For each secretome marker, the antibody giving the strongest staining signal was selected for analysis. The antibody that gave positive staining signals for most of the tumor samples, and showed signal intensity, was selected for further analysis.

Breast tissue mRNA expression
mRNA expression of the secretome markers in the breast cancer tumor tissue and controls were obtained from GEPIA2 (2018 version 9) [52], with all parameters set as default.

Kaplan-Meier survival analysis
Kaplan-Meier survival data were retrieved from the HPA (version 20.1) [45]. Analysis was done on non-stratified (all stages), early-stage (stages I & II; I, IA, IB, II, IIA,

Breast cancer blood plasma analysis
Secretome marker expression in breast cancer blood exosomes versus healthy controls was retrieved from ExoRBase2.0 (version 2.0) [55].

Statistical analysis
False discovery rate (FDR) corrected p-values were calculated by Skyline using the MSstats R package (version 3.13.6) [39]. GO statistics from The Gene Ontology Resource and STRING database used FDR corrected p-values [40,41,49]. Gene expression results were compared using unpaired, two-tailed t-tests in GraphPad Prism (Version 9.2.0). Gene expression in normal and breast cancer tissue was measured using the absolute value of fold change cutoff of log 2 1 and q-value cutoff of 0.01 (ANOVA) [52]. miRNA cluster and secretome marker correlation were measured using the Pearson Correlation coefficient. Significant differences between Kaplan-Meier survival curves were determined with the log-rank test [45]. Gene expression in blood exosomes between two groups was compared using an unpaired, two-tailed t-test [55].

Data availability
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD031771 [57]. Breast-specific proteome data was extracted from the HPA (version 20.1) [45]. Classical secretome prediction method data was extracted from the HPA (version 20.1). Non-classical secretome prediction method data was completed in SecretomeP 2.0 (December 2020) [17]. miR526b and miR655 targets were identified through TargetScanVert (Release 7.1) [50]. Enrichr identified the TFs of secretome markers [51]. miR526b and miR655 clusters were identified through miRBase (release 22.1) [53]. Breast cancer and normal immunohistochemistry staining data were obtained from the HPA (version 20.1) [45]. mRNA expression of secretome markers in breast tumors and controls was collected from GEPIA2 (2018 version) [52]. miRNA cluster expression and secretome marker mRNA expression were obtained through cBioPortal (v 3.7.15) [54]. Kaplan-Meier survival data were found from the HPA (version 20.1) [45]. Secretome marker expression in breast cancer blood exosomes and healthy controls were collected from ExoRBase2.0 (version 2.0) [55]. Computer code and all other data supporting the findings of this study are available from the corresponding author upon request.

Author contribution statement
Riley Feser: Performed the experiments; Analyzed and interpreted the data; Wrote the paper. Reid M Opperman; Braydon Nault: Analyzed and interpreted the data. Sujit Maiti: Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data. Vincent C Chen: Contributed reagents, materials, analysis tools or data. Mousumi Majumder: Conceived and designed the experiments; Wrote the paper.

Data availability statement
Data associated with this study has been deposited at The mass spectrometry proteomics data have been deposited to the Pro-teomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD031771 (Username: reviewer_ pxd031771@ebi.ac.uk and Password: Pa5Hr3Fk).

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.