Breast Cancer Cell Secretome Analysis to Decipher miRNA Tumor Biology and Discover Potential Biomarkers

MicroRNA (miRNA/miR) miR526b and miR655 overexpressed tumor cell-free secretions promote breast cancer phenotypes in the tumor microenvironment (TME). However, the mechanisms of miRNA regulating TME have never been investigated. With mass spectrometry analysis of MCF7-miRNA-overexpressed versus miRNA-low MCF7-Mock tumor cell secretomes, we identified 34 novel secretory proteins coded by eight genes YWHAB , TXNDC12 , MYL6B , SFN , FN1 , PSMB6 , PRDX4 , and PEA15 those are differentially regulated. We used bioinformatic tools and systems biology approaches to identify these markers’ role in breast cancer. Gene ontology analysis showed that the top functions are related to apoptosis, oxidative stress, membrane transport, and motility, supporting miRNA-induced phenotypes. These secretory markers expression is high in breast tumors, and a strong positive correlation exists between upregulated markers’ mRNA expressions with miRNA cluster expression in luminal A breast tumors. Gene expression of secretome markers is higher in tumor tissues compared to normal samples, and immunohistochemistry data supported gene expression data. Moreover, both up and downregulated marker expressions are associated with breast cancer patient survival. miRNA regulates these marker protein expressions by targeting transcription factors of these genes. Premature miRNA (pri-miR526b and pri-miR655) are established breast cancer blood biomarkers. Here we report novel secretory markers upregulated by miR526b and miR655 ( YWHAB , MYL6B , PSMB6 , and PEA15 ) are significantly upregulated in breast cancer patients’ plasma and are potential breast cancer biomarkers.


of 31
(P = 1.24 × 10 −21 ) ( Figure 1D). This indicates these proteins have major roles in macro-molecule biogenesis. The most significantly enriched pathway is cellular response to stress (p = 1.78 × 10 −53 ) ( Figure 1E), which directly relates to the reported roles of miR526b and miR655 to promote oxidative stress in breast cancer 13 . The following most enriched pathway was cellular response to external stimuli (p = 2.96 × 10 −53 ). This result supports our earlier report that miRNA secrete angiogenesis and lymphangiogenesis stimulating factors which alter cell migration and tube formation capacity of endothelial cells 14 .  and (E) pathways of differentially expressed secretions in MCF7-miR526b and MCF7-miR655 secretomes.

Data curation and threshold determination
We established a >1.5/<-1.5 log2 fold change and >0.3 -log10 p-value as our threshold, following previous publications 17, 18 . We found 96 proteins coded by 32 genes in MCF7-miR526b secretome ( Figure 2A) and 95 proteins coded by 29 genes in MCF7-miR655 secretome ( Figure 2B) that were within our threshold. The total proteins we identified in both miRNA secretomes are 136 proteins, coded by 39 genes, which is approximately the top 92 nd percentile of proteins and genes ( Figure 2C). Of the 39 protein-coding genes, 13 were upregulated, and 26 downregulated in both miRNA-secretomes ( Figure 2C). Interestingly, many protein-coding genes came from the same protein domains, including 13 from the c-terminus of histone H2A, five from Septin, four from 14-3-3, and three from thioredoxin, a detailed list is provided in Figure S1A. Most protein domains were grouped on the agglomerative hierarchical clustering heatmap, including the c-terminus of histone H2A, septin, and 14-3-3. The total list is provided in Figure S1B. Next, differential secretory proteins within our threshold were analyzed to ensure they were found in the breast-specific proteome. There are 19670 human protein-coding genes, and 14227 are listed in the breast-specific proteome. Of our 39 genes, 33 were found in the breast-specific proteome. The six genes not in the breast-specific proteome (H2AC1, H2AC12, H2AC14, H2AC18, OBSCN, SEPT14) were excluded from further analysis.
Overall, our data curation pipeline identified eight secretome markers from the beginning 1535 secreted proteins, as summarized in Figure 3C. These eight secretome markers present in both miRNA-secretomes will be further investigated through different systems biology approaches, examining their relationships with miR526b and miR655 and evaluating their bloodbased biomarker potential.

Gene ontology analysis of secretome markers
First, we determined the secretome markers' general functions and individual cellular components, biological processes, and molecular functions with GO analysis ( Figure 4A and Table S1). Interestingly, many of the secretome markers' functions and GO enrichments overlapped with miR526b and miR655 induced breast cancer phenotypes. For example, PRDX4 and TXNDC12 have contrasting roles in oxidative stress. TXNDC12 was found 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 an antioxidant enzyme that neutralizes oxygen species and protects cells against oxidative stress. It is a key protein that regulates cell redox homeostasis. Since miRNA overexpression has been shown to increase oxidative stress in breast cancer, decreased expression of PRDX4 and increased expression of TXNDC12 in miRNA-secretomes shows the complexity of the breast cancer TME and the tumor regulatory roles of miRNA. FN1 modulates angiogenesis and is downregulated in miRNA secretomes. Since miRNAs promote angiogenesis, FN1 expression in miRNA-high cells might be regulated by miRNA targeting transcription factors of FN1.
Next, the eight secretome markers were analyzed further to find any significantly enriched function. This found four cellular components ( Figure 4B), four biological processes ( Figure 4C), one molecular function, which is phosphoprotein binding ( Figure 4D), and 30 different pathways ( Figure S2) linked to these markers. The top ten pathways are shown in Figure 4E. Interestingly, all cellular components were related to the extracellular region, further confirming that these secretome markers are secreted by miRNA-overexpressed breast tumor cells, suggesting they might play a role in TME regulation. Additionally, three biological processes were related to apoptotic regulation, and one was involved in redox reaction, and both processes are altered by miR526b and miR655 in breast cancer. The most significantly enriched pathway was FLT3 signaling. This pathway is involved in the differentiation, proliferation, and survival of dendritic and hematopoietic progenitor cells 20 . When adapter and signaling molecules bind with FLT3's active receptor, activation of downstream pathways such as PI3K/Akt and MAPK cascades occur. MAPK signaling is also a significant pathway listed. This pathway responds to various extracellular stimuli to activate intracellular processes such as gene expression, metabolism, proliferation, and apoptosis 21 . 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, BRAF, is mutated in ~7% of cancers, including breast cancer 22 . FOXO transcription factors (FOXO1, FOXO3, FOXO4) bind to 14-3-3 proteins, allowing their retention in the cytosol 21 , and 14-3-3 proteins are upregulated in the secretome. Activation of BAD and translocation to mitochondria is sequestered by 14-3-3 proteins after Akt1 phosphorylation 21 . Most of these pathways are known regulatory functions of miR526b and miR655 involvement.

In silico analysis of miRNA regulating secretome markers gene expressions
To establish a regulatory connection between miRNA and secretome markers, we crossreferenced known transcription factors (TFs) of the eight markers with predicted targets of both miRNAs. Altogether, miR526b showed 4133 predicted targets, and miR655 had 3264 predicted targets, of which 1252 were predicted common targets of both miRNAs ( Figure 5A). TXNDC12 was an indirect target of miR526b and a predicted direct target of miR655. However, TXNDC12 was found upregulated in both miRNA secretomes. Therefore, we found TFs that negatively regulate TXNDC12 and are common targets of both miRNAs ( Figure 5B). miRNA downregulates negative regulator TFs NANOG and KLF10, which upregulates TXNDC12 expression. Upregulated secretome marker MYL6B has two positive regulation TFs, MYC and SP3, and one negative regulation TF MECP2. Since MYL6B is upregulated in both miRNAs' secretomes, we predict that MECP2 is completely downregulated by miRNAs, leading to MYL6B's increased secretome expression. Alternatively, downregulated secretome marker PRDX4 has three positive regulation TFs, FOXP1, MYC, and NANOG, and two negative regulation TFs, FOXP1 and ZNF148. Since PRDX4 is downregulated in miRNA-high secretomes, we anticipate that the three positive regulation TFs are targeted by miRNAs resulting in PRDX4 downregulation. We could not find TFs of SFN that are common miRNA targets. However, seven TFs have predicted miR526b targets ( Figure S3A), and six have predicted targets of miR655 ( Figure S3B). Therefore, in miR526b's secretome, SFN could be upregulated through one or more of its negative regulation TFs ETS2, SOX4, THRB, ESR1, and POU5F1 downregulation. In miR655's secretome, SFN could be upregulated via targeting negative regulator TFs THRA, SRF, and HNF4A. Interestingly, two negative regulators of SFN, THRB, and THRA (thyroid hormone receptors) are targets of miR526b and miR655, respectively.

Secretome markers correlation with miRNAs cluster expressions in breast cancer
To determine a correlation between miR526b and miR655 with secretome markers in breast tumors, miRNA cluster expressions and secretome markers mRNA expression data were extracted, and the Pearson correlation coefficient was measured. For all breast cancer subtypes, there were a total of 283 tumor tissue samples available for both miRNA clusters and secretome marker mRNA expression ( Figure 5C-D). In non-stratified samples, upregulated secretome markers SFN and TXNDC12 showed significant positive correlations with miR526b and miR655 clusters, respectively. At the same time, downregulated secretome marker FN1 had a significant negative correlation to miR526b's cluster. Tumor samples were then stratified into luminal A or luminal B, HER2-enriched, and triple-negative tumor subtypes.
In luminal A samples, miRNA cluster expressions showed a stronger correlation to secretome markers in luminal A breast cancer for all markers in miR655's cluster, and seven correlations were significant ( Figure 5C). This is denoted by the increased Pearson correlation coefficient in total tumor samples compared to luminal A samples for upregulated markers and decreased Pearson correlation coefficient between the two categories for downregulated markers. In miR526b's cluster, all markers except FN1 had improved correlations in luminal A breast cancer, and five markers correlation coefficients were significant ( Figure 5D). When comparing total samples to luminal B, HER2-enriched, and triple-negative subtypes, only one enhanced correlation was found, as FN1 had a stronger negative correlation with miR526b's cluster compared to all samples, but this was not significant.

Secretome markers gene expression in breast cancer cell lines
Gene expression of secretome markers in poorly metastatic ER-positive MCF7, T47D, HER2 positive SKBR3, and highly metastatic triple-negative HS578T and MDA-MB-231 breast cancer cell lines was compared to immortalized mammary epithelial cell line MCF710A. Beginning with upregulated secretome markers, YWHAB showed significant upregulation in both poorly metastatic T47D, SKBR3 cell lines, and highly aggressive MDA-MB-231 cells ( Figure  6A); TXNDC12 was significantly upregulated in poorly metastatic T47D and SKBR3 cells ( Figure 6B); MYL6B showed significant upregulation in T47D cells ( Figure 6B), and SFN was significantly downregulated in all breast cancer cell lines ( Figure 6D). So, all upregulated markers except SFN validated secretome results.
Among downregulated secretome markers, FN1 was significantly downregulated in MCF7 and T47D cells and was significantly upregulated in HS578T cells ( Figure 6E); PSMB6 was high in all cell lines but was significantly upregulated in MDA-MB-231 cells ( Figure 6F); PRDX4 showed significant upregulation only in aggressive cell lines HS578T and MDA-MB-231 ( Figure 6G); and PEA15 is significantly downregulated only in MCF7 cells, while was significantly upregulated in all other cell lines ( Figure 6H). Secretome is protein data, and these are gene expression data. However, we found YWHAB, TXNDC12, MYL6B, FN1, and PEA15 marker expression is similar in secretome and cell lines. miRNAs play a major role in gene expression and regulation; hence these observations need further validation in miRNA upregulated cells.

Translational validation of secretome markers in breast cancer: Gene expression analysis in breast cancer tissue and blood
Next, the eight secretome marker expressions were analyzed in breast tumor tissue (n=1085) and normal tissues (n=291). All markers were upregulated in breast cancer tumor tissues compared to control ( Figure 7A). This supports upregulated secretome markers YWHAB, TXNDC12, MYL6B, and SFN. However, downregulated secretome markers FN1, PSMB6, PRDX4, and PEA15 were also upregulated at the mRNA level in breast tumors. The observed downregulation of secretome markers in miRNA-high secretomes could be due to miRNA targeting the positive regulator TFs of the identified markers as shown in miRNA target analysis.
Using the Human Proteome Organizations (HUPO) Human Proteome and Plasma Proteome Projects data, we identified that all 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). Additionally, secretome marker mRNA expression was explored in the blood exosomes of breast cancer patients (n=140) and healthy controls (n=118) ( Figure 7B-I). In breast cancer patients' blood, YWHAB, MYL6B, PSMB6, and PEA15 expressions were significantly higher, and SFN and PRDX4 expressions were also higher but not significant compared to control blood samples. On the other hand, TXNDC12 and FN1 showed no difference in mRNA expression in blood plasma.

Immunohistochemistry analysis of identified markers in breast tumor and normal tissue
Immunohistochemistry staining data of breast cancer tumors and normal tissues were examined to determine secretome markers protein expression and localization. SFN had five of 12 breast tumor tissues showing high or low positive intensity staining, compared to normal tissue not showing any expression ( Figure 8A). YWHAB had all breast tumor samples stained positive at high or medium levels compared to YWHAB found only at medium expression in normal tissue ( Figure 8B). MYL6B showed the same expression in both tumor and control tissue ( Figure 8C). TXNDC12 had no immunohistochemistry data. PRDX4 had eight of 11 samples showing low staining in tumor tissue, compared to normal samples, all having medium intensity ( Figure 8D). FN1 had most samples (eight of 12 samples) showing no expression in breast tumor tissue, and normal tissues showed low-intensity staining ( Figure 8E). No breast cancer tissues showed PEA15 expression, but all three control samples showed low PEA15 expression ( Figure  8F). PSMB6 in both breast and normal tissues showed medium or high-intensity staining ( Figure  8G). Overall, SFN, YWHAB, PRDX4, FN1, and PEA15 immunohistochemistry data reflected secretome expression.

Roles of secretome markers in breast cancer patient survival
We have conducted Kaplan-Meier survival plot analysis for secretome markers in nonstratified (all stages I-IV) and in stratified early (stages I & II) and late (stages III & IV) tumor stages to examine if secretome markers expression is associated with poor patient survival (Figure 9). High YWHAB expression showed reduced survival in non-stratified samples, but that was not significant (p = 0.092) ( Figure 9A). However, YWHAB significantly reduced survival in early-stage tumors (p = 0.020) ( Figure 9B). But in late-stage tumors, YWHAB expression was not associated with patient survival ( Figure 9C). TXNDC12 expression was not correlated with breast cancer patient survival in non-stratified samples (p = 0.95) ( Figure 9D), and in the early stages, low TXNDC12 expression showed slightly reduced survival, but that was not significant (p = 0.23) ( Figure 9E). But in late-stage tumors, high TXNDC12 expression is significantly associated with poor patient survival (p = 0.00087) ( Figure 9F). Low MYL6B expression showed reduced survival in non-stratified samples but was not significant (p = 0.09) ( Figure 9G); however, low MYL6B expression significantly reduced survival in early stages (p = 0.044) ( Figure 9H). Alternatively, high MYL6B expression led to slightly decreased survival in late stages, but that was not significant (p = 0.58) ( Figure 9I). Decreased SFN expression led to poor patient survival in non-stratified samples (p = 0.056) ( Figure 9J) and significantly reduced survival in early-stages (p = 0.018) ( Figure 9K). On the other hand, in late stages, high SFN expression led to significantly reduced survival (p = 0.0039) ( Figure 9L). High FN1 expression led to slightly decreased survival in non-stratified samples (p = 0.21) ( Figure 9M). In the early tumor stages, FN1 was not associated with breast cancer patient survival ( Figure 9N). Increased FN1 expression showed marginally reduced survival in late stages but not significantly ( Figure  9O). PSMB6 expression was not connected to patient survival in non-stratified samples (p = 0.94) ( Figure 9P). However, low PSMB6 expression led to slightly decreased survival in early stages (p = 0.16) (Figure 9Q), and high PSMB6 expression showed moderately reduced survival in late stages (p = 0.21) (Figure 9R). High PRDX4 expression led to significantly decreased survival in non-stratified samples (p = 0.018) ( Figure 9S). Low PRDX4 expression led to reduced survival in early-stage tumors but was not significant (p = 0.72) ( Figure 9T). In late stages, high PRDX4 expression showed significantly decreased survival (p = 0.00049) ( Figure 9U). PEA15 had no survival data available. Overall, the most important relationships in breast cancer patient survival were found with upregulated secretome marker expressions.

Discussion
miR526b and miR655 are oncogenic miRNAs that promote aggressive breast cancer phenotypes and alter cells within TME. We have identified that cell-free miRNA and miRNAhigh tumor cell secretory proteins change the phenotypes of cells present in TME. Analysis of miR526b and miR655 cell secretomes in ER-positive breast cancer cell lines might decipher the mechanisms of miRNA regulating TME and identify biomarker candidates. Extensive secretome analysis can impose difficulties, as many extracellular proteins are signaling molecules found at low levels hence might not be selected due to higher threshold 17 . However, our platform of combining nano-high-performance liquid chromatography with large sensitivity mass spectrometry ensured in-depth, sensitive secretome analysis. After systematic data curation, we identified four upregulated (YWHAB, TXNDC12, MYL6B, SFN) and four downregulated (FN1, PSMB6, PRDX4, PEA15) markers in miRNA-overexpressed tumor secretomes. miR526b and miR655 have been shown to induce oxidative stress by overproduction of ROS, and ROS levels were further enhanced during hypoxia in miRNA overexpressed cells 13,15 . These miRNA-induced functions are also regulated by secretory marker expression by tumor cells. ROS production causes DNA damage so often triggers apoptosis, and in growing aggressive tumors, hypoxia influences the apoptotic pathways. Here we identified that the most enriched biological processes regulated by the eight secretome markers were related to apoptosis regulation and cell redox homeostasis. In miRNA-high breast cancer cells, instead of induction of apoptosis, hypoxia promotes oxidative stress, cell migration, tube formation 15 . Secretome markers regulate ROS levels, apoptosis, and hypoxic response, supporting miR526b and miR655 regulating these processes in breast cancer.
Strong relationships between upregulated secretome markers and breast cancer were identified individually when analyzing each secretome marker. YWHAB is one of seven 14-3-3 family proteins, and all family members, except SFN, have oncogenic properties in epithelial carcinomas and are known to potentiate tumor growth and progression 23 . In luminal A breast cancer, YWHAB overexpression led to a worse breast cancer prognosis 24 and showed higher expressions in non-triple-negative breast cancer tissues and cell lines 25 . In our study, YWHAB was upregulated in both miRNA secretomes by miRNA targeting negative regulator TFs of YWHAB and showed strong positive correlations to miRNA cluster expression in luminal A breast cancer. In addition, YWHAB was significantly upregulated in luminal A breast cancer cell T47D, HER2-enriched SKBR3, and triple-negative MDA-MB-231 cells. YWHAB was also significantly upregulated in breast cancer blood samples and tissues and showed high expression 21 of 31 in immunohistochemistry staining in breast tumors compared to controls. High YWHAB expression could significantly predict a worse prognosis in early-stage breast cancer. Overall, YWHAB promotes oncogenic breast cancer functions and shows the potential to serve as a biomarker for luminal A breast cancer.
TXNDC12 inhibits cell death, regulates intrinsic apoptotic signaling pathways, and modulates oxidoreductase activity during oxidative stress 26 . Upregulation of TXNDC12 inhibits apoptosis by endoplasmic reticulum stress-inducing agents and promotes EMT and metastasis in many epithelial cancers 27,28 . However, TXNDC12 has not been studied in breast cancer. Two of its gene family members, ARG2 and ARG3, are known serum-based breast cancer biomarkers 29 . In our study, TXNDC12 was upregulated in both miRNA secretomes; however, TXNDC12 is a predicted target of miR655. Negative regulator TFs of TXNDC12, NANOG, and KLF10, are targets of both miR526b and miR655, which may explain TXNDC12 upregulation in miRNA secretomes. In our study, TXNDC12 had significant positive correlations with both miRNA clusters in luminal A breast cancer and was also significantly higher in T47D and SKBR3 cell lines. TXNDC12 was highly expressed in breast cancer tissue, and high TXNDC12 expression leads to significantly reduced survival in late-stage breast cancer. Thus, TXNDC12 shows potential as a prognostic biomarker for metastatic breast cancer. Since both miRNAs and TXNDC12 promote EMT, TXNDC12 likely collaborates with miRNAs to promote breast cancer cell migration, invasion, and metastasis.
MYL6B is an essential light chain subunit for myosin motor proteins and regulates cell mobility functions 30 . Although no studies have investigated MYL6B in breast cancer, in other epithelial cancers, such as rectal adenocarcinoma, MYL6B overexpression promoted EMT, and high MYL6B expression was a predictor of poor patient survival 31 . In our study, MYL6B is upregulated in miRNA secretions potentially through both miRNAs targeting MYL6B's negative regulator TF MECP2. Since miRNA secretions enhance mesenchymal phenotypes in primary human endothelial and MCF7 cells 14 , MYL6B could contribute towards miRNA regulating EMT. Furthermore, MYL6B had a significant positive correlation with miRNA cluster expression in luminal A breast cancer. Additionally, MYL6B has significantly increased expression in luminal A T47D cells and in the blood and tissue of breast cancer patients; hence, MYL6B can serve as a blood biomarker. However, MYL6B mRNA expression is not associated with breast cancer patient survival. So, MYL6B might be a diagnostic biomarker to add to the list along with pri-miRNAs.
SFN is another member of the 14-3-3 protein family and is the only member with a tumor suppressive role 32 . Many studies highlight the importance of the loss of SFN expression in breast cancer development 33 . In this study, SFN was upregulated in both miRNA secretomes. However, there were no common SFN TFs regulated by both miRNAs. However, the beta and alpha thyroid hormone receptors (THRB and THRA), two paralogous genes of the same family, are negative regulators of SFN and are targets of miR526b and miR655, respectively. Therefore, SFN could be upregulated in miRNA secretomes through these negative regulator TFs. Furthermore, SFN expression in human tumor tissues could significantly differentiate between early-and late-stage breast cancer, as higher SFN expression significantly decreased breast cancer patient survival in late stages, while lower SFN expression significantly reduced survival in early stages. This might be due to SFN having a tumor suppressor role, and loss of SFN is required to initiate metastasis, and at a progressive stage, mutated SFN behaves more like an oncogene. Additionally, SFN expression correlates with miRNA cluster expression, specifically in the luminal A tumor subtype, and is highly expressed in breast cancer tissue and blood. miRNA-overexpression in MCF7 cells promotes aggressive tumor phenotypes, and high SFN expression is significantly associated with poor patient survival in progressive tumor stages; hence SFN has more oncogenic functions in advanced stages of breast cancer.
The TF MYC is a regulator of seven secretome markers except SFN. MYC had been deemed both a positive and negative regulator of YWHAB and FN1. MYC regulates ~15% of human genes, and in breast cancer, MYC target genes participate in the CSC phenotype, angiogenesis, cell growth, and transformation 34 . We have previously observed that miRNA enhanced CSC phenotypes in breast cancer 11,12 , and MYC expression in MCF7-miR655 cells is marginally upregulated 12 ; however, MYC is a miR655 target. We speculate that if MYC is a miR655 target, its expression may be compensated for by other downstream effects. Thus, the underlying relationship between MYC in the secretome and the CSC phenotype of miRNA-high breast tumor cells requires further investigation.
All secretome markers are deemed prognostic markers in at least one other epithelial cancer and can be identified in the blood 35 . Downregulated secretome markers PSMB6, PRDX4, and PEA15 showed significant negative correlations with miRNA clusters expression in the luminal A subtype tumors. However, they were upregulated in breast cancer tumor tissue, PSMB6 and PEA15 were significantly upregulated in the blood of breast cancer patients, and high expression of PRDX4 had significantly reduced survival in late-stage breast cancer. FN1 mRNA was downregulated in all luminal breast cancer cell lines, supporting secretome data. And FN1, PRDX4, and PEA15 immunohistochemistry staining showed lower expression in tumoral samples than the normal controls, indicating these markers might have tumorsuppressor-like functions. The observed downregulation of these four markers in miRNAsecretomes could be related to miRNA epigenetically regulating these marker expressions in breast cancer. Upregulated secretome markers YWHAB, TXNDC12, MYL6B, and SFN might be more beneficial in establishing a battery of breast cancer biomarkers alongside pri-miR526b and pri-miR655 since their translational bioinformatic analysis supports our secretome findings.
To our knowledge, this is the first time secretory proteins from miR526b and miR655 overexpressing ER-positive cells have been identified and investigated to know the roles of miRNA in TME regulation. Upregulated 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. Although we used only ER-positive breast cancer cells for our secretome analysis, we 23 of 31 validated secretome markers in various breast cancer cell lines and tumor tissues. In addition, the majority of secretome marker expressions were correlated 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 breast cancer. Many low abundances but vital secretory proteins were excluded during data curation to find common secretory proteins present in both miRNA-secretomes. In the future, each miRNA secretome will be investigated separately to identify specific miRNA secretory proteins regulating individual miRNA TMEs. Additionally, we will further investigate the roles of upregulated secretome markers in breast cancer progression and metastasis. Some of the identified markers are showing potential as drug targets. Novel secretory proteins identified in miRNA-high tumor secretions will be tested in breast cancer patient plasma and tissue to test their blood-based breast cancer biomarker potential.

Collection of conditioned media
MCF7-Mock, MCF7-miR526b, and MCF7-miR655 cell lines were grown in complete RPMI 1640 media (Gibco, ON, Canada) with 10% fetal bovine serum (VWR, ON, Canada) and 1% Pen-Strep. Stable miRNA-overexpressed cells and mock (empty vector-transfected) cells received Geneticin (G418) at a concentration of 200μg/ml (Biobasic, ON, Canada), and all cell lines were maintained in cell culture at 37°C, 5% CO2. 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. All cells were followed by being serum-starved with basal media for 16 hours before collection of the cell secretion in the media. The conditioned media, which contains all the secretory proteins and cell metabolites, was collected from all cell lines.

Preparation of proteins to be analyzed by nanoHPLC-MS
Conditioned media proteins were precipitated overnight with 35% ethanol, followed by acidification with sodium acetate and the addition of a digestion buffer (1% sodium deoxycholate and 50mM NH4HCO3). Precipitated proteins were quantified by BCA Protein Assay Kit (Pierce, Rockford, IL) with at least 100μg protein used per cell line. The protein underwent reduction and alkylation (dithiothreitol and iodoacetamide) before digestion with trypsin. 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, we had two experimental replicates of each sample. So, with four biological replicates, we generated n=8 sets of data for each cell line.

of 31
NanoHPLC-MS used Mascot Server (version 2.6) to identify peptides 36 . 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 37 . MCF7-miR526b and MCF7-miR655 cells (case) were normalized to MCF7-Mock (control), and Skyline gave their protein IDs with corresponding fold changes and p-values.

GO analysis of all differentially expressed miRNA-high proteins
All protein IDs differentially expressed between miRNA-high versus miRNA-low cells were entered into The Gene Ontology Resource for Homo sapiens (release 2021-09-01), and data was extracted 38,39 .

Threshold determination and data curation
We extracted the volcano plots for differentially expressed secreted protein IDs for MCF7-miR526b and MCF7-miR655 from Skyline (version 20.1.0.155) 37 . First, we converted Skyline's original fold change, and p-value data of differential secreted protein IDs to log2 fold change and -log10 p-value, respectively, since Skyline gave volcano plots with this output and variables. Next, we submitted protein IDs to Uniprot and extracted protein names, primary gene names, and synonyms 40 . 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 our 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 fold change and -log10 pvalue differed, we took the average of all IDs with the same gene name. The log2 fold change and -log10 p-value of gene names were further analyzed to identify proteins that abided by our threshold in at least one miRNA-secretome. We had no statistically significant protein IDs (p <0.05), so we considered the top 92nd percentile of data, which roughly translates to 0.3 -log10 pvalue.

Generation of heatmaps
We used R-studio version 1.4.1103 and R version 4.0.3 to make gene heatmaps. We used the gene name and corresponding mean log2 fold change values for MCF7-miR526b and MCF7-miR655-secretomes as an input file for R to make heatmaps. Heatmap functions with default arguments in R were used to produce the agglomerative hierarchical clustered heatmap.

Breast-specific proteome
The 14227 human protein-coding genes within the breast-specific proteome were extracted from the HPA (version 20.1) 35 and compared to genes within our secretome threshold.

Secretome prediction methods
We downloaded the classical secretome prediction method data from the HPA (version 20.1) 35 for HPA, MDSEC, Phobius, SignalP, and SPOCTOPUS and compared genes within each method to our list of genes that followed our threshold.
For the non-classical secretome prediction method, SecretomeP, FASTA sequences of our threshold protein IDs were obtained via Uniprot and submitted in SecretomeP 2.0 (December 2020) for mammalian sequences 19,40 . Genes were considered non-classically secreted following the previously established guideline of a neural network score >0.6 and odds >3 19,41 . Secretome markers were considered classically secreted if found in both classical and non-classical secretion methods 19 .

Secretome marker functions and GO analysis
Secretome marker general functions were obtained by using www.GeneCards.org version 5.6.0 Build 515 42 . Individual GO functions of each secretome marker were obtained through Uniprot (Last modified: February 2, 2021) 40 and QuickGo (GO version 2021-11-08) 26 . 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) 38,39 . This obtained cellular component and Reactome pathway results. Additionally, secretome markers were analyzed with STRING database (version 11.0b) 43 to identify biological processes and molecular functions GO.

miRNA target genes and TF analysis
miR526b (hsa-mir-526b) and miR655 (hsa-mir-655) targets were downloaded for both mature five-prime sequences from TargetScanVert (Release 7.1) 44 . 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 and Enrichr, then identified the TFs for our secretome markers 45 . Each marker and their TFs were matched against miR526b and miR655 common targets. If the miRNA target gene matches the TF of the gene, and if that TF is a positive regulator of the gene, then gene expression will be downregulated. In contrast, if miRNA targets the negative regulator TF of a gene, gene expression will be upregulated.