Quantitative Proteomics of Breast Benign and Malignant Tumors Reveals a Malignancy Signature

doi:10.21203/rs.3.rs-3622132/v1

Download PDF

Article

Quantitative Proteomics of Breast Benign and Malignant Tumors Reveals a Malignancy Signature

https://doi.org/10.21203/rs.3.rs-3622132/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

The distinction between benign and malignant breast tumors is a challenge in clinical settings. While omic studies have contributed to discovering genetic and proteomic signatures in breast cancer, the molecular differences between benign and malignant tumors remain less studied. This pilot study aimed to investigate proteomic differences between both type of tumors to identify protein signatures indicative of malignancy. The relevance of our findings was assessed using published proteomics and transcriptomic datasets. Using SWATH-based mass spectrometry, we quantified 1,221 proteins in benign (n = 10) and malignant (n = 5) breast tumors. Protein-protein interaction (PPI)-based networks and enrichment analyses revealed dysregulation in pathways associated with extracellular matrix organization, platelet degranulation, innate immune system, and RNA metabolism. Through unsupervised analysis, a four-protein signature (OGN, LUM, DCN, and COL14A1) associated with the extracellular matrix emerged, differentiating between benign and malignant tumors. This protein dysregulation pattern was consistently verified in cancerous versus non-cancerous breast tissue across diverse proteomics and transcriptomics datasets. Notably, the dysregulation magnitude was higher in breast cancer subtypes with poor prognosis, such as Basal-Like and HER2 compared to Luminal A. These findings suggest a potential role for the identified signature in discerning malignant from non-cancerous breast tissue, offering valuable insights into enhancing diagnostic precision.

Biological sciences/Cancer/Breast cancer

Health sciences/Biomarkers/Diagnostic markers

Biological sciences/Systems biology

Breast cancer

quantitative proteomics

mass spectrometry

biomarkers

benign breast tumors

Breast cancer (BC) is the most common malignancy worldwide. Over the past three decades, the incidence and mortality rates of BC have increased, especially in low- and middle-income countries like Mexico, where it has risen by 28.2–75%^1,2. The epidemiological profile of BC in Mexico resembles that of Latin America, with patients being relatively young at the time of diagnosis and manifesting a high prevalence of risk factors including obesity, low breastfeeding rates, low physical activity, and high hormonal exposure^3,4. Limited access to qualified prevention programs and delayed early detection, diagnosis, and treatment have contributed to the increase prevalence of BC cases and high mortality rates in Mexico and worldwide^5,6.

The diagnosis of BC is complex and based on clinical assessments aided by imaging techniques and pathological tests (tissue biopsies)⁷. Once a breast tumor or lesion is detected, its grading as benign or malignant is urgently needed to determine the adequate therapy. Nonetheless, misdiagnosis (overdiagnosis, underdiagnosis, and incorrect classification) of BC occurs with current imaging screening techniques⁸ and even when experienced pathologists try to differentiate between malignant and benign tumors using cytological tests on biopsies⁹. A study assessing the diagnostic disagreement on benign and malignant breast biopsies among 115 pathologists from 8 US states compared with a consensus panel reference diagnosis reported an overall 75.3% concordance tumor classification rate¹⁰. Therefore, it is of utmost importance to study new methodologies that help in better differentiating benign and malignant breast tumors. New molecular testing based on omics techniques (e.g., transcriptomics, proteomics) has provided new insights into the BC diagnosis and prognosis. For instance, new BC subtypes have been suggested based on RNA sequencing (PAM50 classifier)¹¹ and protein sequencing (proteomics)¹². However, the characterization of the molecular differences between benign and malignant breast tumors is until now understudied. A study using fine-needle aspiration breast biopsies from benign and malignant tumors quantified 167 proteins and suggested an 11-protein signature capable of differentiating malignant tumors from benign lesions¹³. Similarly, another study employed quantitative proteomics to uncover molecular differences among benign and malignant breast fibroepithelial lesions¹⁴. Such study included the analysis of fibroadenomas and phyllodes tumors, which are considered rare breast tumors and exhibit varying in classification, spanning benign, borderline, and malignant groups. Singular proteins, such MUCL1, HTRA1, and VEGFD, emerged as potential biomarkers for distinguishing between these different breast phenotypes, specifically for fibroadenomas, benign, and malignant phyllodes tumors, respectively.

To elucidate the molecular differences between benign and malignant breast tumors, with a specific focus on the Mexican population, we conducted a proteomic and bioinformatic analysis on surgically removed breast tissue samples. Our objective was to identify potential biomarkers or protein signatures capable of distinguishing between these two tumor types. To strengthen the robustness of our study, we applied a suite of comprehensive bioinformatic tools. In addition, we performed meticulous reanalyzes of larger omics datasets, encompassing diverse ethnic populations, to validate and reinforce our discoveries.

Study design and sampling

We examined 15 breast tissue samples from Mexican women, who were diagnosed with either a benign breast tumor (n = 10) or a malignant breast tumor (n = 5). Malignant tumor diagnoses were confirmed through immunohistochemistry (IHC) testing, ultrasound imaging, and clinical assessment. For patients with malignant tumors, a neoadjuvant chemotherapy regimen was administered, consisting of 4 cycles of paclitaxel (80 mg/m²), doxorubicin (50 mg/m²), cyclophosphamide (500 mg/m²), and fluorouracil (500 mg/m²) at 3-week intervals. Patients identified as HER2/neu positive (IHC 3+) additionally received trastuzumab with a loading dose of 8 mg/kg and maintenance dose of 6 mg/kg. These samples are sourced from a prior study investigation conducted by Perez-Ortiz et al.¹⁵ at the North Zone General Hospital SSEP, Puebla, Pue., Mexico and National Cancer Institute in Mexico City (INCAN). All tissue specimens were obtained through breast-conserving surgery (tumorectomy) after chemotherapy completion and subsequently samples were preserved at -70°C (Fig. 1). The histopathological and clinical characteristics of these samples are detailed in Table 1. This research protocol has obtained approval from the Panamerican University Institutional Review Board, specifically the Research Committee and Ethical Research Committee, in accordance with the principles outlined in the Declaration of Helsinki and the Health Insurance Portability and Accountability Act of 1996 (HIPAA). All patients provided informed consent.

Chemical and reagents

Pierce Quantitative Colorimetric Peptide Assay kit, Pierce Mass Spec Sample Prep Kit for cultured Cells, and Pierce formic acid were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Acetonitrile and water were Optima LC–MS Grade and obtained from Fisher Scientific (Hampton, NH, USA). Ethanol LiChrosolv Grade was obtained from Merck KGaA (Darmstadt, Germany). All chemicals (e.g., salts for buffers), acetone, ethyl acetate, triton X-100, protease inhibitor cocktail and Bradford reagent were from Sigma Aldrich (St. Louis, MO, USA). Bovine serum albumin (BSA) was from Amresco (Solon, OH, USA).

Table 1

Histopathological and clinical characteristics of benign breast tumors and breast cancer cohorts.
Variable	Benign (n = 10)	Malignant (n = 5)
Mean age	42 (22–68)	55 (40–72)
Diagnosis	Fibroadenoma	HER2/neu (n = 2) Triple negative (n = 1) Luminal B (n = 2)
Chemotherapy	None	Paclitaxel Doxorubicin Cyclophosphamide Fluorouracil Trastuzumab (HER2/neu)
Size of tumor	> 2 cm	> 2 cm

Protein extraction

Breast tissue samples were triturated using a polytron device with 150–400 uL of ice-cold lysis buffer (0.5 M of NH₄HCO₃ at pH 8, 8 M of urea, 150 mM NaCl, 3 mM MgCl₂ and 0.5% of Triton X-100) supplemented with protease inhibitors. Additionally, samples were put on an ultrasonic device for 30 minutes at 4°C, centrifuged at 14,000 rpm for 10 minutes, and the supernatant recovered (stored at -80°C until processing). The samples were kept in ice at 4°C throughout every step. The total protein content was estimated by the Bradford method using BSA as a calibrant standard.

Sample preparation for proteomics

Tryptic peptides were generated from protein extracts using the Pierce Mass Spec Sample Prep Kit for Cultured Cells according to the manufacturer’s instructions. In brief, proteins were reduced (DTT), alkylated (iodoacetamide) and enzymatically digested (trypsin and Lys-C proteases). As an additional step, samples were extracted with ethyl acetate for detergent residue removal according to the protocol by Yeung et al.,¹⁶. Peptides were dried using a SpeedVac system and resuspended in 95:5 water:acetonitrile with 0.1% formic acid. The concentration of the tryptic peptides was determined using the Pierce Quantitative Colorimetric Peptide Assay according to the manufacturer’s instructions.

LC-MS2 data acquisition

The tryptic peptides were analyzed by LC-MS2 using an Eksigent nanoLC 400 (Foster City, CA, USA) system that consists of an analytic column HALO Fused-Core C18 (0.3 × 150 mm, 2.7 µm, 90 A pore size). Mobile phases comprised 0.1% formic acid in water (solution A) and 0.1% formic acid in acetonitrile (solution B). The separation of peptides was achieved by a multistep linear gradient as follows: 5% solution B for 5 min, increase to 10% solution B during 35 min, increase to 30% solution B during 60 min, and steep increase to 100% solution B during 35 min and held constant for 15 minutes. A six-min post run with 5% solution B was applied to ensure column re-equilibration. The flow rate was 5 µL/min and the sample injection was 4 µL. Two blank samples containing 5% solution B were injected between samples to minimize potential carryover. The eluent was delivered to a TurboV source of a TripleTOF5600 + MS (AB Sciex, Foster City, CA, USA) using electrospray ionization (ESI) under positive mode. The MS parameters were as follows: IonSpray Voltage Floating, 5500 V; Source temperature, 450 ° C; Curtain gas, 30; Ion Source Gas 1, 25; Ion Source Gas 2 (GS2), 35.

To create the SWATH-MS spectral library, the tryptic peptides from all samples were pooled (5 µg each) and four technical replicates were loaded into the LC-MS. Data-dependent acquisition (DDA) with high sensitivity mode was employed. The accumulation time for the MS was 0.25 s/spectra over the m/z range 350–1200 Da and for the MS2 scan was 0.05 s/spectra over the m/z range 50-1200 Da (cycle time of 1.3 s). The DDA parameters were as follows: charge state + 2 to + 4, intensity 150 cps, exclude isotopes within 6 Da, mass tolerance 50 mDa, and a maximum number of candidate ions 20. Under Information Dependent Acquisition (IDA) settings, the “exclude former target ions” was set as 20 s after two occurrences and “dynamic background subtract” was selected. The collision energies were varied to optimize sensitivity. The instrument was automatically calibrated by the batch mode using appropriate calibration solutions before sample injection and after injection of two samples to ensure a mass accuracy of < 5 ppm for the MS1 and MS2 data.

For data-independent acquisition (DIA) or SWATH-MS, fifty-six precursor isolation variable windows were generated using the SWATH Variable Window Calculator (AB Sciex, Foster City, CA, USA) based on the precursor m/z frequencies noted in the DDA samples, with a minimum window width of 5 m/z. The accumulation time for MS was 0.15 s/spectra over the m/z range 350–1200 Da and MS2 data were obtained over the m/z range 350–1200 Da with an accumulation time of 0.05 s/spectra (cycle time of 3s). The other parameters were the same as those used in the DDA analysis.

Protein identification and quantification

The DDA-derived datasets were processed to generate the protein spectral library, as previously described by our group¹⁷, with some modifications. In brief, raw DDA files were imported into ProteinPilot software version 4.2 (AB Sciex, Foster City, CA, USA) with the Paragon algorithm. Experimental MS2 data were searched against a SwissProt homo sapiens database containing 20,360 reviewed proteins (downloaded September 2018) including common protein contaminants. The parameters were set as follows: Sample type, Identification; Cys Alkylation, Iodoacetamide; digestion, trypsin; Instrument, TripleTOF 5600; Special Factors, none; Species, homo sapiens; ID Focus, Biological modifications, and amino acid substitutions; Search Effort, Thorough ID. False discovery rate analysis was also performed.

The DIA or SWATH-MS datasets were processed for retention time calibration and peak extraction using the SWATH Acquisition MicroApp 2.0 in PeakView version 1.2 (AB Sciex, Foster City, CA, USA) along with the DDA-derived spectral library. The parameters were similar to those previously described by our group¹⁷. Two endogenous peptides (> 10,000 intensity) were selected every 10 min throughout the LC gradient for retention time calibration. For peak extraction, we set the parameters as follows: number of peptides per protein, 6; number of transitions per peptides, 6; peptide confidence threshold, 99%; false discovery rate, 1%; exclude modified peptides, checked; XIC extraction window, 10 min; and XIC width, 50 ppm. The extracted peaks were exported to Markerview software version 1.3 (AB Sciex, Foster City, CA, USA) for data cleaning, including the removal of reversed hits. Next, the filtered matrix was quantile-normalized using NormalyzerDE online version 1.3.4¹⁸.

Bioinformatic analysis

The bioinformatic (overrepresentation) analysis was performed using the stringApp version 1.6.0.¹⁹ or STRING online platform²⁰. To globally describe the data, we first created a protein-protein interaction (PPI) network using the list of the dysregulated proteins (based on log2 fold change and p-value) in malignant tumors and then pathway enrichment analysis was performed against the REACTOME²¹, KEGG²², and Gene Ontology²³ Databases. Select small PPI networks were analyzed using REACTOME. A confidence (score) cutoff of 0.9 was selected and singletons or proteins with no reported interactions were filtered out from the analyses. PPI networks were visualized in Cytoscape Software version 3.8.2. (Circular layout) or STRING online platform. The hypergeometric test was used to compute p-values and the Benjamini–Hochberg procedure was used for multiple testing correction.

Data mining of published proteomics studies on breast cancer human cohorts

To validate the findings of this study, we mined published studies (that provided accessible data) using mass spectrometry to determine either proteomic differences among breast cancer tissue and adjacent-normal breast tissue or between molecular cancer subtypes (e.g., Luminal A, HER2, Basal-Like). The list of the studies and their general characteristics included are shown in Table S1. For Tang et al., data²⁴, we reprocessed the quantitative table provided in the Additional files section from the original article. We employed the non-normalized peptide spectral count table and applied the following criteria steps: 1) Proteins with duplicated IDs and missing values in ≥ 12 subjects (cohort size of 118 subjects) were removed; 2) the filtered table was analyzed with NormalyzerDE version 1.3.4 to determine the most appropriate normalization method based on qualitative and quantitative metrics¹⁸ and; 3) Missing values were imputed by the random forest algorithm using MetImp online platform²⁵ (https://metabolomics.cc.hawaii.edu/software/MetImp/). The resulting normalized protein quantification table was used for univariate and multivariate statistical analyses (Hierarchical clustering analysis/heatmap visualization and PCA). Noteworthy, this dataset contained paired and unpaired data; however, we could not find the relationship between samples in the original study. Therefore, we opt to perform an unpaired two-group comparison test, and thus the statistical results should be used cautiously. For Shenoy et al.,²⁶ data (cohort size of 72), we employed the author’s data filtered, imputed and batch normalized. All datasets used were in log2 values. This dataset contained paired data and the relationship between samples could also not be identified in the original study, therefore, a paired two-group comparison test was applied to assess for differences between groups, and thus the statistical results should be used with caution. Multivariate and univariate statistical analyses were performed using Metaboanalyst 5.0 (https://www.metaboanalyst.ca)²⁷ and PRISM 9.0 (GraphPad Software, San Diego, CA), respectively.

Data mining of published transcriptomic datasets on cancer human cohorts

Transcriptomic data mining was performed using the Gene Expression Profiling Interactive Analysis (GEPIA2) web-based tool (http://gepia2.cancer-pku.cn/#index)²⁸. GEPIA2 provides access and reanalysis to TCGA (cancer genome atlas) and GTEx (genotype-tissue expression) transcriptome data from tumor and non-tumor samples of various human tissues. The transcript abundance levels of DCN, LUM, OGN, and COL14A1 were compared among tumor (n = 1085) and non-tumor tissue (n = 291) samples. The Box plots were generated using the expression DIY tool with the following parameters: datasets, BRCA (breast, normal and tumor, n = 1376); log scale, checked; match TCGA normal and GTEx data, checked; Log2FC cutoff, 0.6 and p-value cutoff, 0.05.

Statistical analysis

For the proteomics data generated in this study, all analyses were performed on the quantile normalized data. Proteins with a log2 fold-change ≥ 0.6 or ≤ -0.6 and a p-value < 0.05 (Limma R library) were considered as differentially abundant between groups. Compared to Student’s t-test, the use of bayes-based statistics (in the Limma R library²⁹) in differential expression analysis (two groups comparison) has been shown to provide superior sensitivity and specificity, particularly when a low number of replicates per group is used (as low as 3)³⁰. Univariate statistical analysis (Limma package) was performed on the NormalyzerDE online platform. Multivariate statistical analysis was performed using MarkerView Software (PCA and PCVG) and Metaboanalyst 5.0 (PCA) (https://www.metaboanalyst.ca)²⁷. Hierarchical clustering analysis and heatmap visualization was done using Metaboanalyst 5.0. Data was log2 transformed for PCA, PCVG, and heatmap analysis. For the published proteomics reanalyzed datasets, the normality distribution of the residuals was evaluated by the D’Agostino-Pearson test and inspection of QQ plots. Parametric (two groups, Student’s t-test; three or more groups, one way ANOVA) or non-parametric (two groups, Mann-Whitney test; three or more groups, Kruskal-Wallis test) tests were accordingly applied to assess for differences between groups. Post-Hoc analysis was done using Dunn’s and Tukey’s tests. Extreme outlier samples were removed from the analysis. The software PRISM 9.0 (GraphPad Software, San Diego, CA) was used for the statistical analyses and creation of the volcano plots and column graphs. Statistical analysis was performed on the GEPIA2 platform using predetermined options for the transcriptomics datasets.

Data availability

The mass spectrometry proteomics data generated during the current study have been deposited to the ProteomeXchange Consortium via the PRIDE ³¹ partner repository with the dataset identifier PXD019334. The datasets reanalyzed during the current study are available in the original articles described in Table S1.

Proteomic quantification of malignant and benign tumor samples

Ten malignant and five benign tumor samples were analyzed, and their corresponding clinical information is provided in Table 1. Malignant tumors were obtained from subjects who had undergone chemotherapy as part of their treatment protocol. No significant differences were noted in age between both cohorts (p-value 0.1, student’s t-test). Using the spectral library, we were able to identify 1,412 proteins based on 13,096 unique peptides at a 1% false-discovery rate (FDR) (Table S2 and Table S3). A subset of 1,221 proteins were quantified using 6,673 unique peptides and used for all quantitative proteome analyses after quantile normalization (Table S4). Due to the sensitivity of DIA, no missing values across samples were noted for all the quantified proteins. Our LC-MS method was reproducible as suggested by the total ion chromatograms of the pooled spectral library samples (Fig. 2A). To determine global differences among the proteomes of benign and malignant tumors, we performed unsupervised analysis, including PCA and hierarchical clustering (heatmap visualization). PCA showed a clear distinction between both tumors (Fig. 2B) and heatmap visualization demonstrated a clustering profile according to the tumor type (Fig. 2C). We noted a cluster composed of upmodulated proteins -with associated PPI evidence- linked to the Reactome term, mRNA splicing. Quantitatively, a slightly higher number of proteins were downmodulated (n = 189) than upmodulated (n = 181) in malignant vs. benign tumors (Fig. 2D).

Signaling pathways dysregulated in malignant tumors

We performed a network-based enrichment analysis using Reactome and Gene Ontology databases to gain insight into the pathways dysregulated in malignant tumors. Integrating both upregulated and downmodulated proteins, we identified several enriched pathway terms (FDR < 0.05) To enhance clarity, we visually represented the number of proteins and their directional changes (up or down) within the top 4 Reactome terms (Fig. 3). Our findings evidenced the extracellular matrix organization pathway as the most significantly enriched term (FDR, 3.2E-19), closely followed by platelet degranulation, innate immune system, and metabolism of RNA. These pathways exhibited a mix of down and upregulated proteins, with the metabolism of RNA primarily enriched in upregulated proteins. Noteworthy is the presence of several proteins with the largest reduced fold-changes (malignant vs. benign) within the extracellular matrix organization pathway, such as OGN, DCN, COL14A1. For a comprehensive view, the complete list of enriched pathways can be found in Table S4.

Identification of a tumor malignancy signature by multivariate analysis

Although a single protein could be used as a diagnostic biomarker, the combination of several proteins (i.e., signature) could yield stronger results. Therefore, we performed a PCVG analysis, a technique examining PCA loading values to intuitively identify proteins with shared abundance patterns across samples³². Our analysis unveiled a four-protein signature, comprising decorin (DCN, P07585), mimecan (OGN, P20774), lumican (LUM, P51884), and collagen alpha-1 chain (COL14A1, Q05707) (Fig. 4A). Significantly, these proteins exhibited substantial interactions according to PPI analysis, evident from a notable enrichment p-value of 6.33E-11. Remarkably, three of the four proteins belong to the category of small leucine-rich repeat proteoglycans³³, all intricately associated with the extracellular matrix³⁴. We also retrieved top enriched terms from KEGG, Reactome, and GO databases when using the four proteins (Fig. 4B).

Evaluation of the tumor malignancy signature using published proteomics and transcriptomic databases

In our reanalysis of Tang et al., dataset²⁴, we methodically refined the protein quantification table, reducing it to 6409 proteins after eliminating duplicates and proteins detected in fewer than 12 samples, in contrast to the original 7141 quantified proteins. Among eight normalization methods scrutinized, cyclic loess outperformed others, including the author’s designated method, regularized-logarithm transformation (rlog). Cyclic loess normalization yielded better-centered and distributed samples, enhancing suitability compared to all other methods, as demonstrated in Fig. S1A-C. Therefore, our subsequent analysis and data mining of Tang et al., dataset were based on the cyclic loess-normalized data, that enhanced the separation between cancerous and non-cancerous breast tissue in PCA, with PC1 explaining 25.6% variance, a notable improvement from 17.7%. While we were unable to assess the most appropriate normalization method for Shenoy et al.,²⁶ proteomics data due to unavailable quantification tables (in the articles’ supplementary section), we utilized the authors’ filtered and normalized protein abundances for our analysis.

To validate our findings, we compared the abundance profiles of the signatures’ proteins across cancerous and non-cancerous breast tissues from Tang et al. and Shenoy et al.’s datasets and transcriptomics (TCGA/GTEx²⁸) datasets (Fig. 5A). In line with our results, Shenoy et al.’s data presented downregulation of all proteins in malignant breast tissue (Fig. 5B). Similarly, using Tang et al.’s data, OGN, LUM, and DCN were downregulated in malignant breast tissue (Fig. 5C). Transcriptomic analysis further confirmed dysregulation of all four proteins in malignant breast tissue, with LUM being the only upregulated protein (Fig. 5D).

Abundance levels of the signature-associated proteins among breast cancer molecular subtypes

In our comprehensive reanalysis of diverse and extensive proteomics datasets spanning multiple breast cancer molecular subtypes, we meticulously extracted quantitative data related to OGN, DCN, LUM, and COL14A1 protein levels. Our focus was on discerning the abundance differences among HER2, Luminal A, and Basal-Like breast cancer subtypes given that HER2 and Basal-Like are associated with a higher risk of mortality compared to Luminal A subtype³⁵. Our findings highlighted a significant trend: Basal-Like cohorts consistently exhibited markedly lower protein abundance levels compared to other subtypes within the same dataset, except for Mertins et al.’s dataset ⁴⁵, where although no statistically significant differences were observed, a clear trend toward lower levels was evident. In contrast, Luminal A subtype consistently demonstrated the highest protein levels, underscoring the molecular heterogeneity inherent in these breast cancer subtypes (Fig. S2). This variability in protein abundance levels was visually represented in the box plots, wherein dispersion across datasets and groups is apparent.

In this study, we conducted a comprehensive characterization of proteomic differences between benign and malignant breast tumors within a small cohort of Mexican women. Aided by multivariant statistical analysis, we exhibited a distinctive four-protein signature capable of differentiating between these two tumor types. Subsequent bioinformatic analysis revealed an intricate involvement within the extracellular matrix of these signature proteins. To validate the significance of our findings, we undertook a reanalysis of published proteomic and transcriptomic databases (encompassing diverse ethnic groups, larger sample sizes, and stratified by molecular breast cancer subtypes) and this enabled us to assess the capacity of our identified signature in discriminating malignant breast tumors from nonmalignant breast tissues. Notably, our investigations revealed consistent dysregulated (i.e., downmodulation) of all four signature proteins in breast cancer cases. Interestingly, the signature proteins presented a more pronounced downmodulation in highly aggressive breast cancer subtypes (such as Basal-like and HER2) compared to less aggressive subtypes (Luminal A).

Mammography has traditionally served as the primary screening tool for early detection of breast tumors. Breast ultrasound and magnetic resonance imaging (MRI) are used when mammograms prove insufficient or as an additional diagnostic test⁷. Molecular testing has become invaluable for cancer diagnosis, prognosis, and therapy guide selection. However, these tests are cancer type-specific, leading to disparities in research focus. Immunohistochemical biomarkers, such as estrogen receptor, progesterone receptor, HER2, and ki67, are commonly employed for breast cancer molecular subtyping³⁶. New diagnostic approaches have emerged to address the limitations of immunohistochemical classification associated with the heterogenetic biology of breast cancer. In 2000, Perou et al., proposed novel molecular subtypes, including Luminal A, HER2-enriched, and Basal-like, based on gene expression profiling using hierarchical clustering analysis³⁷. Subsequently, a 50-gene subtype predictor (denominated as PAM50 classifier) emerged based on microarray and quantitative PCR¹¹. This supervised risk predictor demonstrated a high agreement in classification with that reported by Perou et al. Over the past decade, the PAM50 classifier has proven value in predicting disease prognosis and response to drug treatments, surpassing clinical factors and immunohistochemical classification^38–44. Recent studies have employed quantitative proteomics to gain insights into the subtyping of breast cancer, considering that genetic and RNA-level alterations do not always translate into protein synthesis. These studies, while aligning with PAM50 subtyping, have identified new subtypes within PAM50 groups^12,45–48, some of which have been correlated with distinct clinical outcomes¹². However, despite extensive research on breast cancer subtyping, distinguishing between benign and malignant tumors at the molecular level remains understudied. Therefore, our study involved a rigorous quantitative proteomic analysis of surgically derived benign and malignant breast tumors aimed at identifying markers of malignancy. Employing SWATH-based quantification, renowned for its high reproducibility^49,50, and following stringent filtering criteria we retained 1,221 proteins without missing values for all bioinformatic analyses. It is worth mentioning that we opted against sample subfractioning, a method that could have expanded the proteome coverage. Our approach was to maintain simplicity in data generation, focusing on a straightforward strategy for the study.

Upon initial analysis, a distinct separation between benign and malignant tumors became evident through unsupervised analysis (PCA and hierarchical clustering analyses). To comprehensively understand these differences at a systems level, we focused on the sets of dysregulated proteins in malignant tissue. Their statistical enrichment (or overrepresentation) was assessed against diverse pathway systems databases and the GO database. The biological interpretation of the molecular differences revealed that upregulated proteins in malignant tumors were primarily linked to proliferation, specifically the metabolism of DNA/RNA, a phenomenon extensively studied⁵¹ and associated with the Warburg effect in cancer cells⁵². However, for a deeper understanding of the altered biochemistry in malignancy, we broadened the analysis to include both upregulated and downregulated proteins in malignant tumors. This expanded approach was employed in the enrichment analysis against REACTOME database²¹. Remarkably, pathways such as extracellular matrix organization, platelet degranulation, and innate immune system emerged as top enriched terms, integrating both upmodulated and downmodulated proteins. On the contrary, the metabolism of RNA was predominantly represented by upregulated proteins.

Upon examining the specific protein members within each pathway term, we predominantly observed a non-redundant composition (i.e., different proteins per term) per pathway, as illustrated by PPI network visualizations. Notably, we observed significant enrichment in the platelet degranulation pathway, which is thought to contribute to tumor progression and metastasis⁵³. Previous studies by Brunoro et al.⁵⁴, Braakman et al.⁵⁵, and Tang et al.²⁴ had independently identified dysregulation of proteins in this pathway in various breast cancer samples, including nipple aspirate fluid, laser capture microdissection, and fresh-frozen tissue, in comparison to non-cancerous breast tissue. In our comparative analysis, the dysregulation of specific proteins - KNG1, CFL1, ALDOA, HRG, TF, and A2M- remained consistent across all studies, regardless of the sample type or collection technique.

Furthermore, we observed enrichment of the innate immune system pathway, with a majority of dysregulated proteins falling within this category. According to findings from Brunoro et al.,⁵⁴, Braakman et al.,⁵⁵ and Tang et al.,²⁴, our study identified dysregulation of proteins associated with the immune system in malignant breast tissue compared to non-cancerous tissue. Specifically, proteins like ANPEP, C3, HSP90AA1, and TTR exhibited consistent dysregulation across all studies, indicating their potential significance in breast cancer pathology. Additionally, we found an increased abundance of complement system proteins (C2, C6, C9, C8A, C8G, among others) in malignant tissue, a pattern corroborated by previous research highlighting elevated levels of complement system proteins in malignant breast tumors⁵⁶. Activation of the complement system is theorized to promote tumor growth and metastasis according to animal models⁵⁷.

Another relevant enriched pathway was the extracellular matrix organization, which was primarily composed of downmodulated proteins, including collagen and laminin subunits, along with cell adhesion proteins. These results aligned with previous studies by Tang et al.,²⁴, Cha et al.,⁵⁸ and Braakman et al.,⁵⁵ evidencing dysregulation of extracellular matrix-associated proteins in malignant breast tissue compared to non-cancerous counterparts. The extracellular matrix, known to regulate critical cellular processes such as proliferation, differentiation, and migration, has established connections with tumorigenesis and metastasis⁵⁹.

In our pursuit to move beyond mere molecular or pathway descriptions associated with malignant tumors, we mined our data deeper, aiming to discern a distinct signature, or set of proteins capable of differentiating between malignant and benign tumors. Employing unsupervised and intuitive analyses, we identified a cluster of proteins linked to the extracellular matrix, namely DCN, LUM, OGN, and COL14A1, all of which exhibited significant downregulation in malignant tumors.

To validate our findings’ relevance and comprehend the roles of these proteins in breast malignancy, we conducted a rigorous reanalysis of published proteomics^24,26 and transcriptomics²⁸ datasets. Across these datasets, we consistently observed the dysregulation of all identified proteins and through manual scrutiny of reported differential abundance analyses in comparable proteomic studies^55,58, we further confirmed the dysregulation of these proteins in malignant breast tissue, reinforcing their potential significance in breast malignancy. Moreover, inspecting the results from other studies, we observed the dysregulation of LUM and DCN (OGN and DCN were not measured) in malignant breast tissue compared to their non-cancerous counterparts, as demonstrated through RT-PCR analyses⁶⁰. It is important to note that these compelling results emerged from the analysis of diverse types of breast tissues, including fresh-frozen, laser capture microdissection, and formalin-fixed paraffin-embedded tissue samples. This diversity underscores the robustness of our findings, encouraging that these dysregulations persist regardless of the sample type, enhancing the reliability and generalizability of our conclusions.

Considering the well-documented variation in mortality rates among different breast cancer molecular subtypes^35,61,62 we aimed to explore potential differences in the abundance of signature-associated proteins across these subtypes. Although the limited size of our cohorts precluded an in-depth investigation, we conducted a reanalysis of larger proteomic studies encompassing breast cancer subtypes according to the PAM50 classification¹¹. Interestingly, in two studies^12,24, we observed consistently lower abundance levels of the four identified proteins in Basal-Like subtype (with a trend toward lower levels in the HER2 subtype) compared to Luminal A subtype. Another study, while not statistically significant, also indicated a trend toward lower abundance levels⁴⁵. Additionally, a recent study reported lower mRNA abundance levels of COL14A1 (DCN, OGN, and LUM were not quantified) in HER2 and Basal-Like subtypes compared to Luminal subtype ⁶³. It is noteworthy that Basal-Like and HER2 subtypes are associated with worst survival rates compared to Luminal A subtype^24,61,62.

The proteins COL14A1, DCN, OGN, and LUM play crucial roles in the extracellular matrix organization. DCN, OGN, and LUM belong to the group of small leucine-rich proteoglycans⁶⁴ and have been linked to tumor growth suppression in both in-vitro and in-vivo models^65–68. These proteoglycans are thought to exert their effects by modulating intracellular signaling pathways, extending beyond their conventional role in extracellular matrix organization. For instance, DCN is believed to act as a defense mechanism, countering abnormal cancer cell growth by downregulating EGFR activity and inducing the expression of p21 and p27, inhibitors of cyclin-dependent kinases that lead to cell cycle arrest^65,69. Similarly, in-vitro and in-vivo studies have hinted at the involvement of TGF-β2⁷⁰ and p21⁷¹, as well as the PI3K/Akt/mTOR⁶⁸ and EGFR⁷² pathways in the tumor suppression effects exerted by LUM and OGN, respectively. Considering this evidence, it is proposed that these proteins act as oncosuppressive proteoglycans. Their reduced abundance in breast tumors raises the hypothesis that this decrease may facilitate tumor growth and metastasis and lead to higher rates of mortality. Supporting this notion, Muraoka et al., demonstrated that the protein levels of LUM, OGN, and COL14A1 were lower in subjects with breast cancer classified as low-risk compared to high-risk, as determined by the MammaPrint diagnostic test⁷³. Likewise, other studies have associated reduced levels of DCN and LUM with worse prognosis⁷⁴. However, the dynamic behavior of these proteins in breast cancer remains unknown. Further research is needed to determine whether breast cancer actively promotes the dysregulation of these proteins or if breast cancer develops because of factors influencing their expression or abundance levels.

A limitation of our study is that subjects underwent chemotherapy before the collection of breast samples. This may have affected the molecular integrity of malignant tumors, potentially masking molecular distinctions compared to benign tumors. Ideally, samples should have been obtained prior to chemotherapy. However, despite this limitation, we observed clear differences between tumors, and our findings were corroborated using diverse datasets, thereby enhancing the robustness and confidence in our results.

In conclusion, our study focused on identifying a distinctive protein signature for differentiating malignant and benign breast tumors through quantitative proteomics and comprehensive bioinformatics (Fig. 6). Despite our limitation in sample size, our meticulous reanalysis of published datasets provided robust validation for our findings. The consistent downregulation of COL14A1, DCN, OGN, and LUM (with intrinsic associations as members of the extracellular matrix) in malignant tumors, especially in aggressive Basal-Like and HER2 subtypes, was evident across diverse datasets. This innovative methodology not only highlighted the potential of these proteins as reliable markers for malignancy but also hinted at their role in breast tumor progression. Our study demonstrates the power of integrative analyses, transcending sample size constraints, and offers valuable insights for precise diagnostics and targeted therapeutic interventions in breast cancer.

Acknowledgement

This work was derived from the Thesis Project of V.L.Z.C. at the Posgrado en Ciencias de la Vida, CICESE. We thank Alan G. Hernandez-Melgar for his important technical assistance using the NormalyzerDE software. We declare the use of ChatGPT only to improve the readability and proofreading of the manuscript.

Funding

This work was supported by CICESE (Internal grant No. 685-109, 685-604) to A.M.U and CONAHCyT (Grant No. SALUD-2013-01-201387) to F.J.E.M.

Conflict of interest

The authors have no conflict of interest to declare.

Author contributions

Study conception and design: A.M.U., F. J. E.M. and B.E.M.; data collection: A.M.U., F. J. E.M., V.L.Z.C., I.R.S., J.C.C.L., A.P.O., C.V.G., J.D.C., B.E.M., B.A.A., X.P.L.A., E.L.R.; methodology, formal analysis, and interpretation of results: A.M.U., F. J. E.M., V.L.Z.C., I.R.S and A.P.O.; project administration, and funding acquisition: A.M.U., F. J. E.M. All authors contributed to data interpretation, critical revision of the manuscript, drafting, and approval of the final version of the manuscript.

Sung, H. et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 71, 209-249, doi:10.3322/caac.21660 (2021).
Chavarri-Guerra, Y. et al. Breast cancer in Mexico: a growing challenge to health and the health system. Lancet Oncol 13, e335-343, doi:10.1016/S1470-2045(12)70246-2 (2012).
Reynoso-Noveron, N. et al. Clinical and Epidemiological Profile of Breast Cancer in Mexico: Results of the Seguro Popular. J Glob Oncol 3, 757-764, doi:10.1200/JGO.2016.007377 (2017).
Amadou, A., Torres-Mejia, G., Hainaut, P. & Romieu, I. Breast cancer in Latin America: global burden, patterns, and risk factors. Salud Publica Mex 56, 547-554, doi:10.21149/spm.v56i5.7379 (2014).
Villarreal-Garza, C. et al. Breast cancer in young women in Latin America: an unmet, growing burden. Oncologist 18 Suppl, 26-34, doi:10.1634/theoncologist.18-S2-26 (2013).
Hanna, T. P. et al. Mortality due to cancer treatment delay: systematic review and meta-analysis. BMJ 371, m4087, doi:10.1136/bmj.m4087 (2020).
Cardoso, F. et al. Early breast cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-updagger. Ann Oncol 30, 1194-1220, doi:10.1093/annonc/mdz173 (2019).
Ryser, M. D. et al. Estimation of Breast Cancer Overdiagnosis in a U.S. Breast Screening Cohort. Ann Intern Med 175, 471-478, doi:10.7326/M21-3577 (2022).
Quinn, C., Maguire, A. & Rakha, E. Pitfalls in breast pathology. Histopathology 82, 140-161, doi:10.1111/his.14799 (2023).
Elmore, J. G. et al. Diagnostic concordance among pathologists interpreting breast biopsy specimens. JAMA 313, 1122-1132, doi:10.1001/jama.2015.1405 (2015).
Parker, J. S. et al. Supervised risk predictor of breast cancer based on intrinsic subtypes. J Clin Oncol 27, 1160-1167, doi:10.1200/JCO.2008.18.1370 (2009).
Asleh, K. et al. Proteomic analysis of archival breast cancer clinical specimens identifies biological subtypes with distinct survival outcomes. Nat Commun 13, 896, doi:10.1038/s41467-022-28524-0 (2022).
Franzen, B. et al. A fine-needle aspiration-based protein signature discriminates benign from malignant breast lesions. Mol Oncol 12, 1415-1428, doi:10.1002/1878-0261.12350 (2018).
Kumar, A. et al. Quantitative proteome profiling stratifies fibroepithelial lesions of the breast. Oncotarget 12, 507-518, doi:10.18632/oncotarget.27889 (2021).
Perez-Ortiz, A. C. et al. Pharmacogenetics of response to neoadjuvant paclitaxel treatment for locally advanced breast cancer. Oncotarget 8, 106454-106467, doi:10.18632/oncotarget.22461 (2017).
Yeung, Y. G., Nieves, E., Angeletti, R. H. & Stanley, E. R. Removal of detergents from protein digests for mass spectrometry analysis. Anal Biochem 382, 135-137, doi:10.1016/j.ab.2008.07.034 (2008).
Hilda Carolina Delgado De la Herrán, L. D.-M., Carolina Álvarez-Delgado, Francisco Villarreal, Aldo Moreno-Ulloa. Formation of multinucleated variant endothelial cells with altered mitochondrial function in cultured coronary endothelium under simulated diabetes. bioRxiv, doi:https://doi.org/10.1101/622407 (2019).
Willforss, J., Chawade, A. & Levander, F. NormalyzerDE: Online Tool for Improved Normalization of Omics Expression Data and High-Sensitivity Differential Expression Analysis. J Proteome Res 18, 732-740, doi:10.1021/acs.jproteome.8b00523 (2019).
Doncheva, N. T., Morris, J. H., Gorodkin, J. & Jensen, L. J. Cytoscape StringApp: Network Analysis and Visualization of Proteomics Data. J Proteome Res 18, 623-632, doi:10.1021/acs.jproteome.8b00702 (2019).
Szklarczyk, D. et al. The STRING database in 2023: protein-protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic Acids Res 51, D638-D646, doi:10.1093/nar/gkac1000 (2023).
Fabregat, A. et al. The Reactome Pathway Knowledgebase. Nucleic Acids Res 46, D649-D655, doi:10.1093/nar/gkx1132 (2018).
Kanehisa, M. & Goto, S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res 28, 27-30, doi:10.1093/nar/28.1.27 (2000).
Harris, M. A. et al. The Gene Ontology (GO) database and informatics resource. Nucleic Acids Res 32, D258-261, doi:10.1093/nar/gkh036 (2004).
Tang, W. et al. Integrated proteotranscriptomics of breast cancer reveals globally increased protein-mRNA concordance associated with subtypes and survival. Genome Med 10, 94, doi:10.1186/s13073-018-0602-x (2018).
Wei, R. et al. Missing Value Imputation Approach for Mass Spectrometry-based Metabolomics Data. Sci Rep 8, 663, doi:10.1038/s41598-017-19120-0 (2018).
Shenoy, A. et al. Proteomic patterns associated with response to breast cancer neoadjuvant treatment. Mol Syst Biol 16, e9443, doi:10.15252/msb.20209443 (2020).
Pang, Z. et al. Using MetaboAnalyst 5.0 for LC-HRMS spectra processing, multi-omics integration and covariate adjustment of global metabolomics data. Nat Protoc 17, 1735-1761, doi:10.1038/s41596-022-00710-w (2022).
Tang, Z., Kang, B., Li, C., Chen, T. & Zhang, Z. GEPIA2: an enhanced web server for large-scale expression profiling and interactive analysis. Nucleic Acids Res 47, W556-W560, doi:10.1093/nar/gkz430 (2019).
Ritchie, M. E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 43, e47, doi:10.1093/nar/gkv007 (2015).
van Ooijen, M. P. et al. Identification of differentially expressed peptides in high-throughput proteomics data. Brief Bioinform 19, 971-981, doi:10.1093/bib/bbx031 (2018).
Perez-Riverol, Y. et al. The PRIDE database and related tools and resources in 2019: improving support for quantification data. Nucleic Acids Res 47, D442-D450, doi:10.1093/nar/gky1106 (2019).
Ivosev, G., Burton, L. & Bonner, R. Dimensionality reduction and visualization in principal component analysis. Anal Chem 80, 4933-4944, doi:10.1021/ac800110w (2008).
Hultgardh-Nilsson, A., Boren, J. & Chakravarti, S. The small leucine-rich repeat proteoglycans in tissue repair and atherosclerosis. J Intern Med 278, 447-461, doi:10.1111/joim.12400 (2015).
Byron, A., Humphries, J. D. & Humphries, M. J. Defining the extracellular matrix using proteomics. Int J Exp Pathol 94, 75-92, doi:10.1111/iep.12011 (2013).
Haque, R. et al. Impact of breast cancer subtypes and treatment on survival: an analysis spanning two decades. Cancer Epidemiol Biomarkers Prev 21, 1848-1855, doi:10.1158/1055-9965.EPI-12-0474 (2012).
Viale, G. The current state of breast cancer classification. Ann Oncol 23 Suppl 10, x207-210, doi:10.1093/annonc/mds326 (2012).
Perou, C. M. et al. Molecular portraits of human breast tumours. Nature 406, 747-752, doi:10.1038/35021093 (2000).
Gnant, M. et al. Identifying clinically relevant prognostic subgroups of postmenopausal women with node-positive hormone receptor-positive early-stage breast cancer treated with endocrine therapy: a combined analysis of ABCSG-8 and ATAC using the PAM50 risk of recurrence score and intrinsic subtype. Ann Oncol 26, 1685-1691, doi:10.1093/annonc/mdv215 (2015).
Liu, S. et al. Prognostic and predictive investigation of PAM50 intrinsic subtypes in the NCIC CTG MA.21 phase III chemotherapy trial. Breast Cancer Res Treat 149, 439-448, doi:10.1007/s10549-014-3259-1 (2015).
Prat, A. et al. Response and survival of breast cancer intrinsic subtypes following multi-agent neoadjuvant chemotherapy. BMC Med 13, 303, doi:10.1186/s12916-015-0540-z (2015).
Chia, S. K. et al. A 50-gene intrinsic subtype classifier for prognosis and prediction of benefit from adjuvant tamoxifen. Clin Cancer Res 18, 4465-4472, doi:10.1158/1078-0432.CCR-12-0286 (2012).
Gnant, M. et al. Predicting distant recurrence in receptor-positive breast cancer patients with limited clinicopathological risk: using the PAM50 Risk of Recurrence score in 1478 postmenopausal patients of the ABCSG-8 trial treated with adjuvant endocrine therapy alone. Ann Oncol 25, 339-345, doi:10.1093/annonc/mdt494 (2014).
Nielsen, T. O. et al. A comparison of PAM50 intrinsic subtyping with immunohistochemistry and clinical prognostic factors in tamoxifen-treated estrogen receptor-positive breast cancer. Clin Cancer Res 16, 5222-5232, doi:10.1158/1078-0432.CCR-10-1282 (2010).
Dowsett, M. et al. Comparison of PAM50 risk of recurrence score with oncotype DX and IHC4 for predicting risk of distant recurrence after endocrine therapy. J Clin Oncol 31, 2783-2790, doi:10.1200/JCO.2012.46.1558 (2013).
Cancer Genome Atlas, N. Comprehensive molecular portraits of human breast tumours. Nature 490, 61-70, doi:10.1038/nature11412 (2012).
Mertins, P. et al. Proteogenomics connects somatic mutations to signalling in breast cancer. Nature 534, 55-62, doi:10.1038/nature18003 (2016).
Krug, K. et al. Proteogenomic Landscape of Breast Cancer Tumorigenesis and Targeted Therapy. Cell 183, 1436-1456 e1431, doi:10.1016/j.cell.2020.10.036 (2020).
Johansson, H. J. et al. Breast cancer quantitative proteome and proteogenomic landscape. Nat Commun 10, 1600, doi:10.1038/s41467-019-09018-y (2019).
Muntel, J. et al. Comparison of Protein Quantification in a Complex Background by DIA and TMT Workflows with Fixed Instrument Time. J Proteome Res 18, 1340-1351, doi:10.1021/acs.jproteome.8b00898 (2019).
Collins, B. C. et al. Multi-laboratory assessment of reproducibility, qualitative and quantitative performance of SWATH-mass spectrometry. Nat Commun 8, 291, doi:10.1038/s41467-017-00249-5 (2017).
Hanahan, D. & Weinberg, R. A. Hallmarks of cancer: the next generation. Cell 144, 646-674, doi:10.1016/j.cell.2011.02.013 (2011).
Warburg, O., Wind, F. & Negelein, E. The Metabolism of Tumors in the Body. J Gen Physiol 8, 519-530, doi:10.1085/jgp.8.6.519 (1927).
Lal, I., Dittus, K. & Holmes, C. E. Platelets, coagulation and fibrinolysis in breast cancer progression. Breast Cancer Res 15, 207, doi:10.1186/bcr3425 (2013).
Brunoro, G. V. F. et al. Differential proteomic comparison of breast cancer secretome using a quantitative paired analysis workflow. BMC Cancer 19, 365, doi:10.1186/s12885-019-5547-y (2019).
Braakman, R. B. et al. Proteomic characterization of microdissected breast tissue environment provides a protein-level overview of malignant transformation. Proteomics 17, doi:10.1002/pmic.201600213 (2017).
Niculescu, F., Rus, H. G., Retegan, M. & Vlaicu, R. Persistent complement activation on tumor cells in breast cancer. Am J Pathol 140, 1039-1043 (1992).
Afshar-Kharghan, V. The role of the complement system in cancer. J Clin Invest 127, 780-789, doi:10.1172/JCI90962 (2017).
Cha, S. et al. In situ proteomic analysis of human breast cancer epithelial cells using laser capture microdissection: annotation by protein set enrichment analysis and gene ontology. Mol Cell Proteomics 9, 2529-2544, doi:10.1074/mcp.M110.000398 (2010).
Winkler, J., Abisoye-Ogunniyan, A., Metcalf, K. J. & Werb, Z. Concepts of extracellular matrix remodelling in tumour progression and metastasis. Nat Commun 11, 5120, doi:10.1038/s41467-020-18794-x (2020).
Eshchenko, T. Y., Rykova, V. I., Chernakov, A. E., Sidorov, S. V. & Grigorieva, E. V. Expression of different proteoglycans in human breast tumors. Biochemistry (Mosc) 72, 1016-1020, doi:10.1134/s0006297907090143 (2007).
Cheang, M. C. et al. Basal-like breast cancer defined by five biomarkers has superior prognostic value than triple-negative phenotype. Clin Cancer Res 14, 1368-1376, doi:10.1158/1078-0432.CCR-07-1658 (2008).
Carey, L. A. et al. Race, breast cancer subtypes, and survival in the Carolina Breast Cancer Study. JAMA 295, 2492-2502, doi:10.1001/jama.295.21.2492 (2006).
Malvia, S. et al. ABERRANT EXPRESSION OF COL14A1, CELRS3, and CTHRC1 IN BREAST CANCER capital ES, CyrillicELLS. Exp Oncol 45, 28-43, doi:10.15407/exp-oncology.2023.01.028 (2023).
Iozzo, R. V. & Schaefer, L. Proteoglycan form and function: A comprehensive nomenclature of proteoglycans. Matrix Biol 42, 11-55, doi:10.1016/j.matbio.2015.02.003 (2015).
Moscatello, D. K. et al. Decorin suppresses tumor cell growth by activating the epidermal growth factor receptor. J Clin Invest 101, 406-412, doi:10.1172/JCI846 (1998).
Hu, X. et al. Decorin-mediated suppression of tumorigenesis, invasion, and metastasis in inflammatory breast cancer. Commun Biol 4, 72, doi:10.1038/s42003-020-01590-0 (2021).
Karamanou, K. et al. Lumican effectively regulates the estrogen receptors-associated functional properties of breast cancer cells, expression of matrix effectors and epithelial-to-mesenchymal transition. Sci Rep 7, 45138, doi:10.1038/srep45138 (2017).
Xu, T. et al. Osteoglycin (OGN) Inhibits Cell Proliferation and Invasiveness in Breast Cancer via PI3K/Akt/mTOR Signaling Pathway. Onco Targets Ther 12, 10639-10650, doi:10.2147/OTT.S222967 (2019).
Xaus, J., Comalada, M., Cardo, M., Valledor, A. F. & Celada, A. Decorin inhibits macrophage colony-stimulating factor proliferation of macrophages and enhances cell survival through induction of p27(Kip1) and p21(Waf1). Blood 98, 2124-2133, doi:10.1182/blood.v98.7.2124 (2001).
Brezillon, S., Pietraszek, K., Maquart, F. X. & Wegrowski, Y. Lumican effects in the control of tumour progression and their links with metalloproteinases and integrins. FEBS J 280, 2369-2381, doi:10.1111/febs.12210 (2013).
Li, Y. et al. Cleavage of lumican by membrane-type matrix metalloproteinase-1 abrogates this proteoglycan-mediated suppression of tumor cell colony formation in soft agar. Cancer Res 64, 7058-7064, doi:10.1158/0008-5472.CAN-04-1038 (2004).
Hu, X. et al. Osteoglycin (OGN) reverses epithelial to mesenchymal transition and invasiveness in colorectal cancer via EGFR/Akt pathway. J Exp Clin Cancer Res 37, 41, doi:10.1186/s13046-018-0718-2 (2018).
Muraoka, S. et al. Strategy for SRM-based verification of biomarker candidates discovered by iTRAQ method in limited breast cancer tissue samples. J Proteome Res 11, 4201-4210, doi:10.1021/pr300322q (2012).
Troup, S. et al. Reduced expression of the small leucine-rich proteoglycans, lumican, and decorin is associated with poor outcome in node-negative invasive breast cancer. Clin Cancer Res 9, 207-214 (2003).

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Quantitative Proteomics of Breast Benign and Malignant Tumors Reveals a Malignancy Signature

Status:

Version 1

Abstract

Figures

Introduction

Materials and Methods

Study design and sampling

Chemical and reagents

Protein extraction

Sample preparation for proteomics

LC-MS2 data acquisition

Protein identification and quantification

Bioinformatic analysis

Data mining of published proteomics studies on breast cancer human cohorts

Data mining of published transcriptomic datasets on cancer human cohorts

Statistical analysis

Data availability

Results

Proteomic quantification of malignant and benign tumor samples

Signaling pathways dysregulated in malignant tumors

Identification of a tumor malignancy signature by multivariate analysis

Evaluation of the tumor malignancy signature using published proteomics and transcriptomic databases

Abundance levels of the signature-associated proteins among breast cancer molecular subtypes

Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1