Peripheral blood CD14 + monocytes in luminal breast carcinoma subtypes: in preliminary research and overview of candidate biomarker proteins

Abstract

“Surrogate” definitions of intrinsic subtypes, which imply the Ki67 proliferation marker, help to clinically distinguish luminal breast carcinomas (Lum BC). Here, mass spectrometry–based proteomics can help analyse the protein content of malignant and normal cells and eventually distinguish patient samples from healthy controls at the molecular level. In this work, peripheral blood CD14 + monocyte proteomes of the LumA, LumB-HER2 − and LumB-HER2 + subtypes and those with benign disease were compared to healthy controls (HCs). Among differentially expressed proteins (DEPs), SRSF1, CSTB, KRT2, KRT5, HEL-S-11, APOB, APOE and ITGA2B are considered as significantly changed in all Lum BC vs HCs comparisons (FC ≥ 1.4 or ≤ 0.7, adjusted P-value ≤ 0.05), while the benign vs HCs comparison showed that KRT2 and KRT5 were common DEPs for all disease groups. APOB, APOE, CSTB, HEL-S-11, SRSF1, and ITGA2B had AUC ≥ 0.6 in all Lum BC cohorts and may serve as feasible classifiers of random individuals. ENO1, KRT1, and ADIB were among the top 10 hits in LumA, LumB-HER2 − and LumB-HER2+, respectively, and can also likely be related to Lum BC. GSEA (P-value < 0.05, FDR ≤ 0.25) identified significantly enriched KEGG Gonadotropin release hormone, KEGG Leukocyte trans-endothelial migration, and KEGG Calcium signalling pathways that were all downregulated in LumA. In LumB-HER2-, KEGG Ribosome and KEGG Lysosome pathways were upregulated and downregulated, respectively, while in LumB-HER2+, the HALLMARK MYC-Targets V2 pathway was found downregulated. The suggested study represents a preliminary research and overview of feasible candidate biomarker proteins in peripheral blood CD14 + monocytes of different Lum BC subtypes. Such proteins may reflect BC-related immune responses and therefore could help to subcategorise disease cohorts using a comparative proteomic approach.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-39686-y.

Introduction

Immunohistochemical (IHC) assessment of oestrogen (ER), progesterone (PgR) hormone status, and human epidermal growth factor receptor 2 (HER2) defines the key signatures for clinical breast carcinoma (BC) subtyping. Although, more details about phenotypic diversity among BC subcategories can be revealed by the newer gene expression profiling (GEP or PAM50)^1,2. To enable molecular BC classification with classical biomarkers (ER, PgR and HER2), “surrogate” definitions of intrinsic subtypes were accepted³. Here, the abundance of Ki67 cell proliferation factor was implied to differ between Luminal A (LumA) and Luminal B (LumB-like-HER2- and LumB-like-HER2+) variants, while HER2 + was substituted with HER2-enriched BC and Triple-Negative BC (TNBC) for Basal-like form^1,3.

A mass spectrometry-based proteomics can be an option to bring even deeper information about tumour heterogeneity compared to IHC or PAM50. Here, discriminations between healthy vs. BC cells and/or benign vs. malign populations, can be achieved on a proteome-information level. Often, a quantitative shotgun proteomic approach serves as a method for monitoring changes in protein abundances².

Peripheral blood mononuclear cells (PBMCs) are an integral part of human immunity. Herein, the analysed BC-reflective candidate biomarker proteins (also signature molecules) can highly inform about the processes associated with cancer progression and tumour microenvironment (TME). Thus, immune cells might surrogate other representative samples like tissue biopsies, which are often accompanied by invasive sampling^4,5.

Jeon et al.⁶ recently utilised proteomic profiling and analysis of immune regulatory processes to categorise BC using formalin-fixed paraffin-embedded biopsies and tumour-infiltrating lymphocytes. Coronin-1 A and α-1-antitrypsin were found upregulated differential proteins in immune-inflamed and immune excluded/desert tumours, respectively.

Given the immunomodulatory role of CD14 + monocytes (fraction of PBMCs) associated with cancer growth and TME shaping^4,5,we sought to explore their proteomic profiles in patients with Luminal BC subtypes (Lum BC) and to analyse candidate biomarker proteins possibly associated with tumour progression. At this stage, we note that our laboratory has recently studied the role of CD8 + and CD19 + cells in the immune response across different BC subtypes using shotgun proteomics^7,8.

By comparative proteomic analysis of Lum BC vs. healthy controls (HCs), we found differentially expressed proteins (DEPs) significantly changed across all BC groups, while some were analysed as possible group-specific DEPs. In addition, comparison of benign vs. HCs revealed DEPs common for all studied groups. Ultimately, The Gene Set Enrichment Analysis (GSEA) showed significantly enriched pathways found in the LumA, LumB-HER-, and LumB-HER + BC subtypes vs. HCs. To our knowledge, the suggested work represents the initial effort to determine candidate biomarker proteins in peripheral blood CD14 + monocytes of different Lum BC subtypes, as reflection of immune-based responses and with the aim of subcategorise cohorts applying the comparative proteomic approach. We believe that identifying immune-associated signature molecules could enable minimally invasive diagnostics and support personalised therapies in BC management.

Materials and methods

Characteristics of patients and samples

The Human Research Ethics Committee of L. Pasteur University Hospital (LPUH) in Košice (Slovakia) approved this clinical study. After surgical biopsy, samples from women were subjected to IHC analysis done by an accredited clinical laboratory (registration number: 158/Q-044, technical norm: ISO 9001:2025, valid certification until 2026). To the freshly confirmed treatment‑naive BC patients, benign patients, and healthy controls (HCs), a paper informed consent (2020/EK/06407) was presented and voluntarily signed. Samples of BC and benign cohorts were acquired from the 1st Department of Surgery, LPUH in Košice (Slovakia), while HCs were obtained from Department of Gynaecology and Obstetrics, LPUH in Košice (Slovakia). The patients´ sample collection was performed during 2023–2024.

The patho-physiological characteristics of patients are present in Supplementary Table 1. Reasons/criteria for inclusion/exclusion of collected specimens^9,10, can be also found in Supplementary Table 1. Surrogate intrinsic definitions of BC subtyping are summarised in Supplementary Table 2.

The experimental and analytical flowchart checklist referring to key steps of suggested work, followed the reporting recommendations for tumour marker prognostic studies guideline¹¹ as depicted in the Supplementary Table 3.

Regarding sampling method, 10 mL of peripheral blood was collected in the BD Vacutainer K₂EDTA tubes and stored at 4 °C tempered transporting box. Each sample preparation procedure began within 30 min since blood collection had been completed. During this interval, no sample instability was observed, thus neither subsequent storage nor preservation steps were needed.

During the sample collection window (2023–2024), a strategy to diminish feasible sample-to-sample variations and prevail consistency of results, included the strict performance of (i) same sampling protocol, (ii) flow-box conditioned sample preparation procedure with freshly-prepared reagents and solvents, and (iii) each time reconditioned nano-LC-MS/MS analysis. To adjust/correct raw MS data feasibly interfered with batch effects, a Principal Component Analysis (PCA) was used.

The HeLa digest was used to assess measurement repeatability, while the coefficient of variation was between 4.9 and 5.1% based on the 5 consecutive runs. These measurements were performed periodically throughout the entire study.

Sample preparation procedure

Blood was pipetted in a conical test tube and diluted with 20 mL saline solution (PBS with BSA and EDTA) at pH of 7.4. Sample was centrifuged (Sigma 3–18 KS, Sigma, Germany) at 600ˣg for 10 min at 4 °C. Then, the upper plasma was carefully removed. Remained fraction with blood elements (~ 6mL) was resuspended with another portion of saline solution up to 10 mL. The 250 µL of activated superparamagnetic beads of diameter = 4.5 μm (certified Dynabads™ CD14, Thermo Fisher Scientific, USA) were added to the sample and mixed (Multi Bio RS-24 multi-rotator, Biosan, Latvia) for 20 min at 4 °C. The tube with dispersed CD14 + monocytes and covalently attached beads, was inserted to a magnetic stand (DynaMag^TM-50 Magnet, Thermo Fisher, USA) for 2 min. Supernatant was removed by gently inverting the magnetic stand. Cells were 3-times rinsed with 10 mL saline solution. By combining 200 µL 50 mM NH₄HCO₃ (AppliChem GmbH, Germany), sonication (Sonorex Digiplus, Badelin electronic GmbH & CO, Germany) and vortexing (Multi-Vortex V-32, Biosan, Latvia), CD14 + monocytes were lysed. Sample was inserted again to the magnetic stand for 2 min lasting removal of the beads. Remained sample was portioned accordingly: Portion 1 of 200 µL supernatant was redissolved with 1800 µL ethanol and stored at -20 °C for overnight protein precipitation and portion 2 with rest of supernatant volume was stored overnight at 4 °C. Ultimately, pellet containing magnetic beads was resuspended with 100 µL Invitrosol™ LC/MS Protein Solubilizer (Thermo Fisher Scientific, USA) and overnight incubated (Eppendorf™ ThermoMixer™ C, Merck, Germany) at 21 °C.

BCA QuantiPro BCA Assay Kit (Sigma-Aldrich, USA) and UV-Vis 3600 spectrophotometer (Shimadzu, Japan) were used for estimation of overall protein content. Further, the 0.025 M DTT (purity ≥ 98%, Biorad, USA) was used to reduce S₂ bonds. Then, the 0.25 M IAA (purity ≥ 99%, Biorad, USA) was used to alkylate cysteine amino acid residues and 5 mM CaCl₂ (purity ≥ 98%, Merck, Germany) to enhance hydrogen bond among proteins. The trypsin/Lys-C mixture of MS purity (Promega, USA) was added in an enzyme/sample 1:40 (w/w) ratio.The overnight in-solution digestion at 21 °C followed. Formic acid (Merck, Germany) was added to quench digestion process at pH 3. Ultimately, sample volume was adjusted for LC-MS/MS analysis using SpeedVac (Labconco, USA) at 18000ˣg and 4 °C.

Nano-LC-MS/MS analysis

An Orbitrap Exploris™ 480 high-resolution mass spectrometer (Thermo Scientific™, USA) equipped with an EASY-Spray™ nano-ESI ion source (Thermo Scientific™) was connected online to a VanquishTM Neo-UHPLC (Thermo Scientific™, USA). Ten µL of 40 ng µL^− 1 protein/peptide sample was introduced to a PepMap™ Neo Trap Cartridge column (Thermo Fisher Scientific, USA) of 5 mm length and 5 μm C₁₈ particle size, using 60 µL min^− 1 injection flowrate.

After preconcentration, proteins/peptides were eluted to an analytical column with 50 cm length and 2 μm C₁₈ particle size, tempered to 40 °C with an integrated heater. Two mobile phases were used for this purpose. Mobile phase A (MPA) composed of 100 mL H₂O and 0.1 mL formic acid (V/V) and Mobile phase B (MPB) composed of 80 mL acetonitrile, 20 mL H₂O, and 0.1 mL formic acid (V/V/V). Gradient elution lasting for 120 min had the following settings: procedure started with an on-line flow mixing to get 98% MPA and 2% MPB. In the next 100 min, the composition of mixture linearly changed up to 24% MPB. During the following 20 min and 10 min, faster-made gradients reached 40% and 90% MPB, respectively.

The ionisation voltage at the nano-ESI emitter was set to 2000 V. Mass spectrometry after first ion fragmentation (MS¹) had the following settings: scan range from 350 to 1700 m/z, 120,000 Orbitrap resolution and 100 ms maximum injection time, and positive ion charge. Conditions of tandem mass spectrometry (MS²) were as follows: 30% HCD collision energy, 2 m/z isolation window, 300,000 Orbitrap resolution and 50 ms maximum injection time.

Protein identification, quantitation and bioinformatics

Data acquired from LC-MS/MS analysis were processed by Proteome Discoverer (PD) version 2.5 (Thermo Scientific™, USA). A standard label-free quantitation was done by evaluation of MS1 peptide intensities with subsequent summarisation at a protein level. This enabled accurate comparison of abundances across different runs. Unique and razor peptides were used for general protein quantification. Here, razor peptides were assigned only to the protein with better identification. To other proteins, sharing peptides were not assigned. MS2 spectra confirmed peptide identities, towards ensured protein assignment and reduced false positiveness. Three protein/peptide search engines were combined available for PD such as MS Amanda 2.0, SequestHT, and Mascot with UniProt Homo sapiens database. Cleaving enzyme specificity for the search was trypsin/P, cleaving after lysine (K) and arginine except if proline (P) follows. The fixed modification was carbamidomethylating of cysteine. The variable modifications were oxidation of methionine, N-terminal acetylation, and Met-loss + acetylation of methionine. The precursor (from MS1) and fragment ions (from MS2) had mass tolerance of 10 ppm and 0.02 Da, respectively. The minimum peptide length cut-off was 6 amino acids, while the maximum permissible missed cleavage was set at 2. In PD, a Low Abundance Resampling was used for missing abundance values that were replaced with random values sampled from the lower 5% if these values were detected.

The Area Under the Curve (AUC) referring to an overall measure of performance, was used to evaluate diagnostic and discriminatory power of candidate biomarker proteins. It was calculated utilising an integrated signal intensity extracted from an ion chromatogram of the specific peptide of interest, extended to the protein amount (https://www.bioinformatics.com.cn/srplot). The Receiver Operating Characteristics (ROC) curves were used to display the plot of true positive rate (TPR = sensitivity) against false positive rate (FPR = 1 - specificity).

A MetaboAnalyst 6.0™, an on-line tool (https://www.metaboanalyst.ca/home.xhtml) was used to perform Principal Component Analysis (PCA) based on ComBat method to evaluate feasible batch effects in raw MS data, and their before and after batch correction and depiction of normalisation trends.

A SynGO™, a ID conversion tool (https://www.syngoportal.org/convert) was used to find missing gene symbols that were not assigned to protein accessions by PD. Volcano plots were done by VolcaNoseR™ (https://huygens.science.uva.nl/VolcaNoseR/). The filtering criteria for analysis of differentially expressed proteins (DEPs) were: number of identified unique peptides was ≥ 2, abundance ratio (fold change, FC) was ≥ 1.4 or ≤ 0.7, abundance ratio adjusted P-Value was < 0.05, false discovery rate (FDR) was ≤ 1%, each protein found had at least a high confidence covering at least 70% of all samples.

Gene set enrichment analysis (GSEA) was done by GSEA v. 4.4.0 software (https://www.gsea-msigdb.org/gsea/index.jsp). The searching criteria for GSEA included: Gene sets database – KEGG_Signaling_Pathway¹² (https://www.genome.jp/kegg/pathway.html) or Hallmark_Signaling_Pathway¹³, Number of permutations – 1000, Collapse/Remape to gene symbols – No_Collapse, Permutation type – phenotype (studied cohort with ≥ 7 participants)/gen_set (studied cohort with < 7 participants), Chip platform – Human_Gene_Symbol_with_Remapping, NOM P-value < 0.05, and FDR q-value ≤ 0.25. SRplot™ software (https://www.bioinformatics.com.cn/srplot) was used for on-line visualisation of bubble plots referring to the enrichment of signalling pathways. The experimental workflow used in this clinical investigation is depicted in the Fig. 1.

Fig. 1.

The experimental pipeline involving key steps such as sampling, sample preparation procedure, nano-LC-MS/MS analysis, and bioinformatic assessment.

Results

A shotgun proteomics, UHPLC-MS/MS, label-free quantitation (LFQ) and Proteome Discoverer 2.5 software enabled comparison of proteomes of LumA (n = 7), LumB-like HER2- (n = 15), LumB-like HER2+ (n = 4), benign (n = 8) and HCs (n = 10). The raw MS proteomic data used in this study have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier: PXD068744 (Project DOI: 10.6019/PXD068744).

Analysis of differentially expressed proteins

The 3251, 3011, 3195, 3049 protein accessions were analysed in LumA, LumB-like HER2-, LumB-like HER2+, and benign subgroups when compared to HCs, respectively (Supplementary Table 4). Volcano plots further revealed 70 (40 downregulated and 30 upregulated), 58 (10 downregulated and 48 upregulated), 87 (49 downregulated and 36 upregulated), and 80 (55 downregulated and 25 upregulated) DEPs present in LumA, LumB-like HER2-, LumB-like HER2+, and benign vs. HCs, respectively (Supplementary Table 5).

The SRSF1 (log₂ FC = 0.95, 0.64, 0.6 and -log₁₀ P-value = 3.846, 2.546, 2.449 in LumA, LumB-HER2-, LumB-HER2 + vs. HCs, respectively), CSTB (log₂ FC = -0.63, -0.86, -0.71 and -log₁₀ P-value = 2.522, 3.95, 3.741 in LumA, LumB-HER2-, LumB-HER2 + vs. HCs, respectively), KRT2 (log₂ FC = -0.58, 0.8, -0.75, -0.84 and -log₁₀ P-value = 1.537, 4.419, 3.507, 1.862 in LumA, LumB-HER2-, LumB-HER2+, benign respectively), KRT5 (log₂ FC = -0.74, 0.68, -0.71, -0.89 and -log₁₀ P-value = 2.659, 3.198, 3.074 in LumA, LumB-HER2-, LumB-HER2+, benign vs. HCs, respectively), HEL-S-11 (log₂ FC = 1.35, 0.95, -0.72 and -log₁₀ P-value = 8.495, 6.24, 3.278 in LumA, LumB-HER2-, LumB-HER2 + vs. HCs, respectively), APOB (log₂ FC = 0.72, 0.67, 0.91 and -log₁₀ P-value = 2.199, 2.96, 5.712 in LumA HCs, LumB-HER2-, and LumB-HER2 + vs. HCs, respectively), APOE (log₂ FC = 0.96, 0.81, 0.65 and -log₁₀ P-value of 3.048, 2.653, 2.61 in LumA, LumB-HER2-, and LumB-HER2 + vs. HCs, respectively), and ITGA2B (log₂ FC = 0.93, 0.66, 1.29 and -log₁₀ P-value = 3.601, 1.719, 11.406 in LumA, LumB-HER2-, and LumB-HER2 + vs. HCs, respectively), were found as significantly changed in all Luminal BC subtypes (Fig. 2A, B, C, D).

Fig. 2.

Volcano plots depict DEPs analysed in CD14 + monocytes in (A) LumA vs. HCs, (B) LumB-HER2- vs. HCs, (C) LumB-HER2+ vs. HCs, and (D) benign vs. HCs . Settings: Criterion for ranking hits – Manhattan distance, FC ≥ 1.4 or ≤ 0.7, -log₁₀ significance threshold = 1.5. The VolcaNoseR™ (https://huygens.science.uva.nl/VolcaNoseR/) enabled the on-line assessment – 18.03.2025.

Among the top 10 hits of DEPs found as significantly changed in a particular BC cohort, ENO1 (log₂ FC = -1.53 and -log₁₀ P-value = 12.531) downregulated in LumA vs. HCs, KRT1 (log₂ FC = 0.79 and -log₁₀ P-value = 4.296) upregulated in LumB-HER2- vs. HCs, and ADIB (log₂ FC = 1.11 and -log₁₀ P-value = 8.926) upregulated in LumB-HER2 + vs. HCs, were also assessed as feasibly BC-related proteins (Fig. 2A, B, C).

ROC curve and referring AUC showed DEPs with fail-to-poor = 0.5–0.6, poor-to-acceptable = 0.6–0.7, fairly acceptable = 0.7–0.8, and considerable = 0.8–0.9 discriminatory power. Herein, APOB, APOE, CSTB, HEL-S-11, SRSF1, and ITGA2B had AUC ≥ 0.6 in all Lum BC groups and were chosen as possible classifiers of random individuals (Supplementary Fig. 1A, B, C). However, in the future, it is our belief that analysed DEPs can gain better discriminatory power (AUC score) using larger disease cohorts.

Assessing potential batch effects of collected samples of patients in 2023 and 2024, the PCA scatter plots (Supplementary file for PCA) indicate a greater influence of batch corrections from LumB-HER2 + across LumA and LumB-HER2-. However, rather intermixing along principal components then distinct clustering can be seen in all tested groups. Raw MS dataset was used to be evaluated, considered as usable for further analysis.

Analysis of molecular pathways

GSEA analysis revealed signalling pathways significantly enriched, passing the criteria of NOM P-value < 0.05 and FDR q-value ≤ 0.25 (Supplementary Table 6). In LumA, KEGG_GNRH (NES = -1.937, FDR q-value = 0.071, and nominal P-value = 0.002), KEGG_LTM (NES = -1.9, FDR q-value 0.055, and nominal P-value = 0.004), and KEGG_CALCIUM (NES = -1.727, FDR q-value = 0.174, and nominal P-value = 0.016) pathways were found downregulated. In LumB-HER2- cohort, KEGG_RIBOSOME (NES = 1.987, FDR q-value = 0.019, nominal P-value = 0.0001) pathway was found upregulated, while KEGG_LYSOSOME (NES = -2.143, FDR q-value = 0.005, and nominal P-value = 0.0001) pathway was downregulated. Ultimately, LumB-HER2 + had HALLMARK_MYC-TARGETS_V2 (NES = -1.691, FDR q-value = 0.137, nominal P-value = 0.009) pathway significantly downregulated (Fig. 3A, B, C, D, E, F and Supplementary material with pathways). Further, by using multi-group GO (a pathway enrichment bubble plot module), we combined pathways to show their feasible overlap (Fig. 4). We found two identical pathways such as VEGF and Phosphatidylinositol signalling system, repeatedly present when comparing LumA and LumB-HER2-. However, these were assessed as not significantly enriched (P-value < 0.05 and FDR ≤ 0.25).

Fig. 3.

GSEA of enrichment plots of KEGG¹² and HALLMARK¹³ molecular pathways (NOM P-value < 0.05 and FDR q-value ≤ 0.25). Studied cohorts represent comparisons of LumA vs. HCs (A, B, C), LumB-HER2- vs. HCs (D, E), and LumB-HER2 + vs. HCs (F). GSEA™ software (https://www.gsea-msigdb.org/gsea/index.jsp) enabled the pathway analysis.

Fig. 4.

Enrichment bubble plot analysis from KEGG and HALLMARK databases assessed for Lum BC subtypes vs. HCs, to show feasible overlaps of enriched pathways NOM (P-value < 0.05 and FDR q-value ≤ 0.25). The colour of each bubble represents the P-value in the context from the lowest assigned by green to the highest marked by red. The size of each bubble refers to the number of involved genes. The rich factor corresponds to a proportion of the total number of genes. SRplot™ software (https://www.bioinformatics.com.cn/srplot) enabled on-line visualisation of bubble plots.

Discussion

Human peripheral blood monocytes, which arise from myeloid lineage, are an integral part of innate immunity. They are classified based on surface marker expression (CD14/CD16) to CD14 + + CD16- cells (classical), CD14 + + CD16+ (intermediate), and CD14 + CD16 + + cells (non-classical). Monocyte cells can have opposing functionalities during immune reaction based on cancer type, growth stage, TME character, etc. The importance of monocytes in cancer understanding, diagnostics, and therapy strategies has been discussed in several recent review papers^4,14,15. However, to our knowledge, there is still a poor of evidence devoted to comparative proteomic analysis of monocytes, particularly between luminal BC subtypes. The main effort to distinguish between LumA (higher endocrine sensitivity, lower aggressiveness, and better prognosis) and LumB (less endocrine sensitivity, higher aggressiveness, and worse prognosis) refers mainly to the different implications associated with adjuvant cytotoxic therapy. Therefore, differentiation between LumA vs. LumB remains important due to therapy decisions³.

DEPs found in CD14 + monocytes of luminal BC subtypes

In this section, we discuss the significant DEPs found in the LumA, LumB-HER2- and LumB-HER2 + groups in the context of the immune response of peripheral blood CD14 + monocytes of newly diagnosed patients with treatment‑naive BC vs. HCs.

SRSF1, found to be upregulated in all luminal BC subgroups, is considered a prototypical serine/arginine-rich protein with a key role in cellular alternative splicing (AS), which can be active during differentiation and activation of PBMCs (e.g., monocytes). The endogenous SRSF1 can regulate alternative splicing events to increase cell proliferation and decrease apoptosis. Its overexpression may further lead to a promoted expression of genes enriched in carcinogenesis-associated pathways¹⁶. A recent clinical correlation study¹⁷ reported that triggering/upregulation of SRSF1 would lead to deregulation of AS events in BC model samples (MCF7 with SRSF1 knockdown). The AS of PTPMT1 exon-3 was found to feasibly mediate the oncogenic role of SRSF1 by activating the P-AKT/C-MYC signalling pathway. In other work¹⁸, SRSF1 was found to possibly enhance immunotherapy by acting on both PBMCs (CD8 + cells) and tumour cells in mice. The inhibition of SRSF1 was suspected to enhance/boost antitumour immune response. Due to above information, we infer that SRSF1 overexpression across Lum BC subtypes vs. HCs could represent a link to sustained AS associated with cancer processes.

CSTB (also STFB), an endogenous cysteine-cathepsin inhibitor whose expression might reflect monocyte-to-macrophage differentiation, was found to be downregulated across all Lum BC subtypes vs. HCs. The proteolytic events can be influenced by inhibiting cysteine-cathepsin leakage caused by cell damage¹⁹. In a recent work on BC mouse model lacking CSTB but susceptible to mammary cancer (PyMT transgenic mice)²⁰, CSTB depletion was found to feasibly reduce tumour growth, but did not affect metastasis. In our case, we also assume that CSTB underproduction in malignant cohorts could likely come from cell damage and cysteine-cathepsin leakage.

The next significantly downregulated in LumA and LumB-HER2+, while significantly upregulated in LumB-HER2-, were KRT2 and KRT5. With relation to immune cells, keratins (KRTs) can promote alternatively activated polarisation of macrophages derived from circulating monocytes. They can deregulate processes in pathogenic cells and mediate cell-to-cell crosstalk. By contributing to cancer cell invasion and metastasis, KRTs may serve as useful disease-related prognostic and diagnostic markers²¹. Apart from other keratins (e.g., K18, K19, etc.), KRT2 and KRT5 are not commonly found as differentially expressed in Lum BC cells. The differential expression patterns would require further investigation to reveal their complex roles in Lum BC.

HEL-S-11, which was found significantly upregulated in LumA and downregulated in LumB-HER2-, and LumB-HER2+, belongs to a family of carbonic anhydrases (CAs). The expression of CAs in e.g., myeloid cells (including macrophages) can be associated with acidosis exhibited during most of solid tumours. It is known that CAs can reduce acidity of the TME by eliminating acidic metabolites, thereby decelerating of tumour growth²². Therefore, in our cases, we assume that differential regulation of HEL-S-11 might be referred to imbalances in pH homoeostasis as well.

Circulating lipoproteins and apolipoproteins may contribute to tumour development by influencing tumour cell migration and invasion²³. The monocyte inflammatory profile can refer to the altered lipid levels in blood. During our investigation, we found upregulation of APOB, which could be responsible for the transport of lipids, in all BC groups. Regarding this protein, Liu et al.[24] reported APOB to be an independent risk factor associated with intraocular metastasis. In other work[25], the studied link between BC and serum lipid levels was reported to be rather unclear and requiring further investigation. As for APOE, its polymorphisms and BC risk was studied, identifying the ε allele as a low penetration factor in BC development ²⁶.

ITGA2B, found significantly upregulated across all BC groups, belongs to a protein family of integrins, which are generally associated with a cell-surface mediation of molecular signalling. Leukocyte β-integrins can be important for monocyte/macrophage migration during inflammatory responses²⁷. Recently, Fredolini et al.²⁸ reported a broad expression of ITGA2B in diverse tissues, particularly in platelets, feasibly marking tumour-associated cell-induced platelet aggregation. ITGA2B was also assigned as one of the candidate breast biomarkers. Along with above information, we assume that upregulation of ITGA2B in BC groups vs. HCs comparison, could be explained by feasible tumour-cell induced platelet aggregation.

Among top ten hits of DEPs analysed in a particular BC cohort vs. HCs, (i.) ENO1 was found significantly downregulated in LumA. ENO1 is a glycolytic enzyme present on the cell surface and cytoplasm but can be also overproduced on the surface of monocytes/macrophages during the immune reaction. Highly expressed during the progression of different types of cancer, it is a well-accessible onco-therapeutic target. Recently, Shi et al.²⁹ reported its overexpression in early-stage BC (I/II) and a positive correlation with its immune-related functions and immune cell infiltration. Further Giannoudis et al.³⁰, studied in silico-based prognostic and predictive effects of ENO1 expression in BC. The highly expressed TNBC-associated pathways were found in invasive ductal BC with a high-grade, however yet not been detected in other BC subtypes. Even though, we assume that the observed ENO1 downregulation in LumA vs. HCs, might be possibly linked to the less aggressive nature of LumA when compared with other breast carcinomas. KRT1, which we found to be significantly upregulated in LumB-HER2-, is considered a molecule providing structural support and protection of epithelial cells from various stressors (mechanic). Soundy et al.³¹ reported that during BC progression, the peptide p160 (VPWMEPAYQRFL) and its enzymatically stable analogue 18 − 4 can bind to KRT1. Links of KRT1 to elevated cell growth, proliferation and cell stress, could be also factors connecting its upregulation during cancer progression in our study. Ultimately, ADIB, adipokine derived from adipose tissue was found significantly upregulated in the LumB-HER2 + cohort. It can modulate innate immunity by regulating macrophage proliferation and polarisation, thus influencing inflammatory responses. Its ability to suppress BC growth by reducing fatty acid synthesis through downregulation of SREBP-1 and FAS-related enzymes was recently reported³². In contrast, reduced plasma ADIB in obesity-related adipose expansion may amplify oestrogen signalling and promote BC development³³.

Signalling pathways in CD14 + monocytes of Lum BC subtypes

Luminal A

GSEA analysis (NOM P-value < 0.05 and FDR q-value ≤ 0.25) showed that the pathways for KEGG gonadotropin-releasing hormone (GNRH), KEGG Leukocyte trans-endothelial migration (LTM) and KEGG Calcium (Ca²⁺) were significantly downregulated in the LumA vs. HCs comparison (See Compressed Folder with Pathways). The results of the KEGG database are displayed with the permission (252173) of Kanehisa Laboratories (https://www.kanehisa.jp/) and in accordance with the Open Access Policy of the Scientific Reports journal. These pathways had 25, 22, and 35 analysed proteins, of which 13, 7, and 21, respectively, were assigned with positive core enrichment (See Compressed Folder with Pathways). In addition to this context, we primarily focused on proteins with positive core enrichment that may be associated with BC processes.

PTK2B (PYK2), a non-receptor focal adhesin tyrosine kinase, had the most negative rank metric scores in all mentioned pathways. Generally, it can regulate cell growth and proliferation to feasibly participate in cell migration in various cancers³⁴. Recently, possible activation of GNRH during cancerous processes (also BC) was reported, where PTK2B may inhibit tumour cell expansion³⁵. In other work, Okitsu-Sakurayama et al.³⁶ observed binding of PTK2B to GRB2 after GNRH treatment. However, there is still lack of evidence to explain relation between PTK2B, GNRH activation and BC processes. In this manner, the context of immune responses that participate in GNRH activation during tumorigenesis deserves attention. In terms of LTM, this process can be associated with the mobility of immune cells from blood to the site-of-infection, activated during pathological stimuli (e.g., BC metastasis). PTK2B can have influence on monocyte activation and migration, while the triggered LTM pathway could participate on adherence of leukocytes to endothelial cells and cross blood vessel walls, reaching eventually the tumour site. Regarding Ca²⁺ pathway, it is known that deregulation of its homeostasis can be associated with start of malignant processes. PTK2B can be stimulated by elevated intracellular calcium concentration, leading Ca²⁺-mediated production of reactive oxygen species (ROS).

Further, ITPR1, GNAS, and PLCB2 were found with a positive enrichment core in the GNRH and Ca²⁺ signalling pathways. ITPR1, a membrane channel type, can participate in muscle contraction, cell proliferation, and other essential cell functions. Han et al.³⁷ recently studied the role of ITPR1 in correlation with tumour-infiltration immune cells in BC samples, reporting lower production in patients compared to HCs, while increased expression was manifested in longer overall survival, disease-specific survival, and relapse free survival. Their GO and KEGG analyses indicated that ITPR1 can be involved in Ca²⁺ signalling pathway, activating calcium release from endoplasmic reticulum (ER) membrane into the cytosol. As for the relation of ITPR1 with GNRH signalling pathway, the participation in hormonal signalling has been assumed. The GNRH receptors were found in BC cells³⁸, while detail mechanism of relation between ITPR1 in GNRH pathway has not been reported so far. Regarding GNAS protein, its dysregulation could be associated with uncontrolled cell proliferation and intracellular signal transduction. Recently, Jin et al.³⁹ reported that elevated GNAS expression can promote BC cell proliferation and migration via the PI3K/AKT/Snail1/E-cadherin axis, leading to reduced overall survival and more frequent distant metastasis. Unfortunately, in our study, we did not find significant enrichment of the PI3K/AKT/Snail1/E-cadherin signalling pathway in BC samples, through which GNAS can promote EMT. Regarding the PLCB2 protein, Bertangolo et al.⁴⁰ reported that PLCB2 can promote mitosis and migration of human BC-derived cells. Its elevated levels correlated with worsening of malignant BC as it can improve motility and capability of tumour cells to invade other organs. Regarding Ca²⁺ signalling, PLCB2 can release intracellular calcium and thus influence overall homeostasis.

PLCG2, which was found to be enriched in the Ca2⁺ and LTM signalling pathways, is a signalling molecule activated by cell-surface receptors containing a tyrosine-based activation motif. The activated protein can generate various second messengers that drive cell proliferation and Ca²⁺ flux. PLCG2 can also drive circumstances related to multitude immunological diseases and was also found to be involved in a few cancers⁴⁰. PLCG2 can also drive circumstances related to multitude immunological diseases and was also found to be involved in a few cancers⁴¹. More recently, Rudolph et al.⁴² reported association of PLCG2 in menopausal hormone therapy (MHT) and BC risk.

CDC42, a small GTPase of the Rho protein family, was found to be enriched in the GNRH and LTM pathways. It is considered an essential factor for directed cell migration, promoting the epithelial-mesenchymal transition (EMT), and cell proliferation. During cancer processes, it can be activated by oncogenic cell receptors and guanine nucleotide exchange factors. Very recently, Torrez-Sanchez et al.⁴³ reported Rho GTPases RAC and CDC42 as ideal molecular targets regulating cancer cells and immune cell migration, studying a BC preclinical model where inhibition of activity was accompanied with forming of antitumour environment.

Luminal B-HER2-

GSEA analysis (NOM P-value < 0.05 and FDR q-value ≤ 0.25) showed KEGG Ribosome (RSP) and KEGG Lysosome (LSP) signalling pathways upregulated and downregulated in LumB-HER2- vs. HCs comparison, respectively. While LSP had 56 analysed proteins, RSP had 44 analysed proteins. Of these, 37 and 20 proteins were found to have a positive core enrichment in LSP and RSP, respectively (See Compressed folder with pathways). The KEGG database outcomes are displayed with permission of Kanehisa Laboratories (https://www.kanehisa.jp/) and with accordance of Open Access Policy of Scientific Reports journal).

Different ribosomal proteins are known to be also involved in BC and some of them are mediators of the ribosomal stress response. Disruption in the production of these proteins can affect the protein synthesis machinery and tumour-associated pathways. Ribosome disfunction can further lead to altered cell cycle progression, resistance to apoptosis, and enhanced migration ability of BC cells. Recently, Lin et al.⁴⁴ identified differentially expressed ribosomal proteins in TNBC and KEGG analysis highlighted the ribosome pathway as significantly enriched, suggesting an important role in the development of TNBC. It is known that ribosomal biogenesis is an emerging target for BC therapy, as altered regulation of ribosomal proteins involved in mRNA translation can cause cell death instead of survival. In our study, we found proteins enriched in the ribosome signalling pathway belonging to a family of RPL and RPS. As reported by Ebright et al.⁴⁵, differential expression of these proteins can increase metastatic growth in organs and selectively enhance translation of other ribosomal proteins and cell cycle regulators. Korte de Azevedo et al.⁴⁶ recently conducted a comprehensive analysis of large and small ribosomal proteins in BC subtypes. Authors highlighted RPS27A, RPS25, and RPL22 as prognostic markers for the LumB subtype. However, in our study, none of these proteins showed positive core enrichment in the ribosome signalling pathway. However, we highlight other small (RPS16, RPS18, RPS3A, RPS4X, RPSA, RPS21, RPS27, RPS28) and large (RPL5, RPL7, RPL23, RPL23A, RPL32, RPL35, RPL35A) ribosomal proteins that may be potentially linked to BC processes.

In LSP pathway, lysosome can act as a signalling hub. Identifying lysosomal functional alterations can help clarify their role in cancer development and progression, for example, changes in lysosomal pathways can affect autophagy in ways that promote cancer[47]. However, identifying lysosome-related genes/proteins that yield prognostic lysosomal signatures for cancer patients remains ongoing. In our study, several proteins showed positive assignment in LSP core enrichment. Among them, (i) HEXA, a lysosomal enzyme can influence the behaviour of immune cells responding to tumour cells; (ii) CD68, a marker that can be used for the common identification of tumour-associated macrophages and their density can correspond to the TME status. Previously, it was reported that with relation to BC, the identification of CD68 + macrophages can be associated with tumour progression⁴⁸; (iii) TCIRG1 (V-ATPase-a3), a proton pump protein can regulate cellular pH and can contribute to immune cell activation and endocytosis. Overexpression of the V-ATPase a3 isoform can greatly increase the invasiveness of MCF10a BC cells⁴⁹; (iv) FUCA1, an enzyme involved in the hydrolysis of fucose-containing glycoconjugates can be associated with tumour aggressiveness and metastatic rate. Recently, a longer cancer-specific survival in LumB LN + was found to be associated with differential regulation of FUCA1 mRNA, which can be is inversely related to BC⁵⁰; (v) LAMP1, a heavily glycosylated lysosomal protein protects the lysosomal membrane from intracellular proteolysis. Its differential expression across tumour types can play a crucial role in tumour development. Wang et al.⁵¹ analysed upregulation of LAMP1 in BC samples when compared to HCs. Its expression was associated with histological grade, expression of ER and PgR, lymph node metastasis, and tumour node metastasis; (vi) IGF2R is a membrane glycoprotein involved in lysosome transport, cell growth, and survival, but can be also linked to malignancies as tumour suppressor. It can play a critical role in the survival of immune cells (e.g. CD8 + cells) and their activation and differentiation. High IGF1R-alpha expression in combination with low expression of IGF2R in LumA and LumB subtypes, can be associated with significantly longer survival⁵².

Luminal B-HER2+

GSEA analysis showed HALLMARK MYC TARGETS V2 signalling pathway downregulated in LumB-HER2 + vs. HCs comparison. MYC TARGETS V2 had 21 analysed proteins, of which 11 were assigned to have a positive core enrichment, (See Compressed Folder with Pathways). MYC is a transcription factor that can contribute to an increased and uncontrolled proliferation of cells. It can be associated with differential expression of MKI67, a gene coding Ki67 proliferation factor⁵³.

Among proteins found with a positive core enrichment, MCM4, a component of MCM complex is responsible for initiation of DNA replication and elongation in eukaryotic cells, while its dysregulation can lead to genomic instability. MCM4 can be frequently overexpressed during BC when compared to normal tissue. Differential expression could correlate with a higher tumour grade, lymph node metastasis, and poor survival. Recently, MCMs (also MCM4) were studied as prognostic markers in the aggressive TNBC form and for discrimination of LumA and B subtypes⁵⁴. It was found that both luminal BCs can be differentiated based on MCMs.

NOLC1, a regulator of RNA polymerase I is responsible for ribosomal processing and modification. Recently, NOLC1 was found differentially expressed in MDA-MD-231 TNBC cells when compared to HCs. Its knocking down by siRNA decreased protein levels of key stemness regulator MYC and ALDH. Ultimately, NOLC1 was reported to be an independent risk factor for overall survival in BC, while its differential expression can relate to poor BC prognosis⁵⁵.

HSPD1, a molecular chaperone is involved in folding and maintenance of proteins in mitochondria. Recently, it was found that HSPD1 could induce cellular plasticity in TNBC by regulation of HSP60, a mitochondrial protein which stabilises and refold proteins. Furthermore, authors reported that differential expression of HSPD1 can be associated with poor cancer-specific survival⁵⁶.

Summary on enriched molecular pathways

GSEA analysis comparing LumA samples with HCs identified proteins implicated in the regulation of Ca²⁺ homeostasis, whose dysregulation may contribute to oncogenic processes. Proteins in the GnRH signalling pathway, which were also analysed in the LumA vs. HCs comparison, are postulated to be functionally interconnected with immune responses and hormonal regulation. Dysregulation or imbalance of these processes can contribute to the development and progression of pathological conditions. The proteins found to be enriched in the LTM pathway in the comparison between LumA and HCs may be associated with regulating the motility of immune cells in response to pathological stimuli.

The enrichment of ribosome and lysosome signalling pathways observed in the LumB-HER2- vs. HCs comparison, is indicative of enhanced protein biosynthesis and intracellular signalling processes, respectively, both of which may contribute to the modulation of the immune response.

The enrichment of the MYC Targets V2 pathway observed in LumB-HER2 + tumour may contribute to cellular processes such as cell growth and proliferation during cancer progression.

Conclusions, limitations of the study and future outlooks

It is known that breast carcinomas (BC) vary in morphology, clinical behaviour and response to therapy. Although tumour heterogeneity can hinder the diagnostic process, searching for specific identification factors can help with BC subtyping. The IHC and GEP are well-established protocols to analyse BC subtypes.

The proteomic method can help to bring new information about tumour heterogeneity by analysing candidate biomarker proteins. However, for unequivocal differentiation among BC subtypes, the use of only a single protein signature would not suffice. For this reason, established biomarker panels can better reflect the character of disease. In clinical investigation, peripheral blood samples benefit over tissue specimen as their sampling causes minimal damage to biological structures compared to solid biopsies². In this manner, we performed a comparative proteomic analysis of CD14 + monocytes from peripheral blood of BC-confirmed patients vs. HCs and benign cases vs. HCs. The results revealed DEPs found in all BC cohorts such as SRSF1, CSTB, KRT2, KRT5, HEL-S-11, APOB, APOE, and ITGA2B. Among them, APOB, APOE, CSTB, HEL-S-11, SRSF1 and ITGA2B had AUC ≥ 0.6 and are assumed to be potential classifiers of random individuals.Comparison of benign vs. HCs revealed KRT2 and KRT5, which were found as common DEPs for all disease groups. ENO1 in LumA, KRT1 in LumB-HER2- and ADIB in LumB-HER2+, were found interesting possibly group-specific DEPs analysed among the top 10 hits. GSEA analysis of LumA vs. HCs revealed significant enrichment of Ca²⁺, GNRH and LTM signalling pathways. The Ribosome and Lysosome signalling pathways were also found to be enriched in the LumB-HER2- vs. HCs comparison and the MYC Targets V2 pathway was found to be enriched in LumB-HER2+.

The suggested study has been subject to several uncertainties that render its findings rather as hypothetical and preliminary in nature. These can be summarised as follows:

1. Nearly the entire 10 mL sample was used for a single analytical run, leaving no material for additional biological replicates.

2. With the reduced and unbalanced sample set (LumA = 7, LumB-HER2- = 15, LumB-HER2 + = 4, benign = 8, HCs = 10), differences in protein abundance between Lum BC subtypes may be less significant and robust than in larger cohorts – for example, proteomic profiles of circulating monocytes obtained from small cohorts are highly susceptible to modulation by co-morbid conditions present in the organism, including subclinical states such as asymptomatic inflammatory processes.

3. Validation of these preliminary findings in an independent cohort is required to further substantiate the robustness and generalisability of the results.

Due to above, suggested work is meant to be a primary study providing important background for a consecutive clinical investigation.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ADIB

Adiponectin B

APOB

Apolipoprotein B

APOE

Apolipoprotein E

CD68

Macrosialin

CDC42

Cell division control protein 42 homolog

CSTB

Cystatin-B

ENO1

Enolase 1

FUCA1

Alpha-L-Fucosidase 1

GNAS

Guanine nucleotide-binding protein

GOLM2

Protein GOLM2

GRB2

Growth factor receptor-bound protein 2

HEL-S-11

Carbonic anhydrase

HEXA

Beta-hexosaminidase subunit alpha

HSPD1

60 kDa heat shock protein, mitochondrial

IGF2R

Cation-independent mannose-6-phosphate receptor

ITGA2B

Integrin alpha-IIb

ITPR1

Inositol 1,4,5-trisphosphate-gated calcium channel ITPR1

KRT1

Keratin, type II cytoskeletal 1

KRT2

Keratin, type II cytoskeletal 2 epidermal

KRT5

Keratin, type II cytoskeletal 5

LAMP1

Lysosome-associated membrane glycoprotein 1

MCM4

DNA replication licensing factor MCM4

NOLC1

Nucleolar and coiled-body phosphoprotein 1

PLCB2

1-phosphatidylinositol 4,5-bisphosphate phosphodiesterase beta-2

PLCG2

1-phosphatidylinositol 4,5-bisphosphate phosphodiesterase gamma-2

PTK2B

Protein-tyrosine kinase 2-beta

PTPMT1

Protein tyrosine phosphatase mitochondrial 1 protein

RPL5

Large ribosomal subunit protein uL18

RPL7

Large ribosomal subunit protein uL30

RPL22

Large ribosomal subunit protein eL22

RPL23

Large ribosomal subunit protein uL14

RPL23A

Large ribosomal subunit protein uL23

RPL32

Large ribosomal subunit protein eL32

RPL35

Large ribosomal subunit protein uL29

RPL35A

Large ribosomal subunit protein eL33

RPS3A

Small ribosomal subunit protein eS1

RPS4X

Small ribosomal subunit protein eS4, X isoform

RPS16

Small ribosomal subunit protein uS9

RPS18

Small ribosomal subunit protein uS13

RPS21

Small ribosomal subunit protein eS21

RPS25

Small ribosomal subunit protein eS25

RPS27

Small ribosomal subunit protein eS27

RPS28

Small ribosomal subunit protein eS28

RPS27A

Ubiquitin-ribosomal protein eS31 fusion protein

RPSA

Small ribosomal subunit protein

SREBP-1

Sterol regulatory element-binding protein 1

SRSF1

Serine/arginine-rich splicing factor 1

TCIRG1

V-type proton ATPase 116 kDa subunit a 3

Author contributions

Michal Alexovič: conceptualisation, sample preparation, investigation, software analysis, writing original draft, visualisation, references, project administration; Peter Bober: sample preparation, investigation, software analysis, review and editing; Miroslav Marcin: sample preparation, MS measurements, investigation, software analysis, Jozef Parnica: sample preparation; Michal Marcin: sample preparation; Marek Lenárt: sample collection; Dávid Tóth: sample collection; Jozef Radoňak: selection of patients; Peter Urdzík: selection of healthy controls; Ján Sabo: supervision, review and editing, project leading and administration.

Funding

Financial support was provided by the Research Agency of Ministry of Education, Research, Development and Youth of the Slovak Republic (Project-ID: ITMS2014+: 313011V446), Slovak Research and Development Agency of Ministry of Education, Research, Development and Youth of the Slovak Republic (APVV-19-0476), and Scientific Grant Agency of Ministry of Education, Research, Development and Youth of the Slovak Republic and Slovak Academy of Sciences (VEGA 1/0582/25).

Data availability

All data generated or analysed during this study are included in this published article and its supplementary information files (Supplementary Tables 4, 5, and 6). The raw MS proteomic data used in this study have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD068744 (Project DOI: 10.6019/PXD068744).

Declarations

Competing interests

The authors declare no competing interests.

Informed consent

Informed consent was obtained from all subjects involved in the study.Statement.

Institutional review board statement

The study was conducted in accordance with the Declaration of Helsinki, and approved by Ethics Committee of Medical Faculty UPJŠ, protocol code 9 N/2021, date of approval 17 June 2021, Ethics Committee of L. Pasteur University Hospital code 2020/EK/06047, date of approval 25 June 2020, and Ethics Committee of East Slovakian Institute of Oncology, protocol code EK/1/06/2020, date of approval 1 June 2020.