Background
Mammographic breast density is one of the strongest risk factors for breast cancer. Women with high breast density have 4–6-fold increased risk of breast cancer as compared to women with low breast density [
1‐
4]. Reflecting the composition of fibroglandular and fat tissue in the breast, mammographic breast density is inversely related to age and higher body mass index (BMI). Radiologically dense tissue, such as stromal and epithelial tissue, appears white on a mammogram, whereas the radiologically lucent fat tissue appears dark [
5]. Several breast cancer risk factors are known to influence breast density [
6]. It has been shown that body weight and reproductive and lifestyle factors explain an estimated 20–30% of the difference in density between women [
7]. Through twin studies, we and others have estimated the heritability of percent density to be around 65% [
7‐
9].
Despite the strong and independent association between mammographic breast density and breast cancer risk, little is known about the biological mechanisms behind this risk factor. Identifying determinants of density may provide insights into the aetiology of breast cancer. It may also be useful for better identifying women at increased risk of developing breast cancer.
Considerable effort has been made to identify biomarkers for early detection and/or monitoring of breast cancer. Although a few potential plasma protein targets have been identified [
10], validation and reproducibility have thus far not been satisfactory for clinical implementation. Prior investigations of plasma markers associated with breast density have mainly focused on endogenous hormones and inflammatory markers with inconsistent or negative results [
6]. No putative independent markers of mammographic density have so far been identified after adjustment for BMI and other confounding factors.
Blood plasma is well-suited for expanded affinity proteomic analysis as it enables a direct but less invasive view into the health status compared to biopsy sampling. Affinity proteomics assays using antibodies with suspension bead arrays (SBA) have been utilised for plasma protein profiling within the context of various diseases including cancer [
11]. The approach allows for many proteins to be screened in small plasma volumes of a large number of samples [
12], thus enabling large-scale proteomic investigations of body fluids like plasma.
In this study, we used a multiplexed affinity proteomics assay with antibodies from the Human Protein Atlas (HPA) [
13] to screen proteins in plasma of women without any prior history of breast cancer, and who were enrolled in a unique prospective population-based cohort in Sweden, the Karolinska mammography project for risk prediction for breast cancer (KARMA) cohort [
14,
15]. The aim of this exploratory approach was to identify density-associated proteins, to improve our still limited understanding of mammographic breast density as a risk factor for breast cancer.
Discussion
We used antibodies to profile proteins in plasma from healthy women with high and low breast density. Proteins were selected based on their possible linkage to mammographic breast density, cancer development and/or progression or tissue composition and remodelling based on literature review. We identified 20 protein profiles in plasma that were linearly associated with AD in both of the studied sample sets. To our knowledge, this is the first study in which plasma proteins were correlated to AD.
Our study provided indications for eleven candidate proteins for which expression was identified in breast tissue (see Table
2) by analyses of omics data through HPA expression [
13] and transcriptome data [
25]. Four of these candidates were positively and seven negatively associated with AD. We present a refined description of these proteins and their relation to AD and breast cancer in Additional file
2. There we also explain our perspective on plasma protein associations with age and BMI.
Mammographic density is predominantly associated with higher extracellular matrix (ECM)-rich stromal tissues and epithelial composition, and lower proportion of adipose tissue [
17,
26,
27]. High collagen levels in the mouse mammary gland increase tumour formation and invasive behaviour [
28], suggesting that dense tissue areas may be tumour promoting. In fact, carcinomas largely arise in the dense region of the breast, supporting the link between tumour formation and mammographic density [
6]. Genetic profiles of extra-tumoural stromal microenvironments have identified a so-called “inactive signature”, comprising higher levels of cell adhesion and cell-cell contact genes, associated with higher mammographic density [
29,
30]. Collagen-rich stromal tissues are also mechanically stiffer [
31,
32], and stiffening of the existing stromal collagen microarchitecture promotes high mammographic density within the breast [
33]. Cells sense force and stiffening through mechanoreceptors such as cell-cell junctions and cell-matrix adhesions mediated by integrins, and respond by activating downstream signalling pathways to maintain tissue homeostasis [
34,
35]. Consistently, we identified positive association between AD and the epithelial cell-cell adhesion molecule F11R. We also identified negative association with AD and the integrin ITGB6. Elevation of F11R and decrease of ITGB6 in plasma from women with high AD emphasise the complexity of maintaining tissue homeostasis to prevent malignant transformation.
Genetic damage to proliferating cells has been postulated to partake in the increased risk of breast cancer associated with extensive mammographic density [
6]. It was recently shown that epithelial cells from high mammographic density tissue have elevated activity in DNA damage signalling, shorter telomeres, and altered DNA damage response compared with epithelial cells from low-density tissues [
36]. The authors hypothesise that elevated basal DNA damage in high-density epithelial cells can result in subsequent induction of the desmoplastic-like phenotypes observed in high-density tissues. Therefore, a breast with more DNA-damaged epithelial cells would exhibit more mammographically dense areas, leading to overall high mammographic density. Supporting this hypothesis, we identified two other proteins expressed in breast tissue, namely FANCD2 and RASSF1, which are both related to DNA integrity and were inversely associated with AD. The p53 target gene TNFRSF10D inhibits apoptosis induction and was positively associated with AD in our sample sets. We also observed a negative association with AD and the CASP8 and FADD-like apoptosis regulator CFLAR. Hence, the association of TNFRSF10D and CFLAR plasma levels with high-density tissues could be indicative of mechanisms by which high-density cells avoid apoptosis induced by DNA damage.
The association between endogenous sex hormones and breast cancer risk is widely described; nonetheless, the mechanisms through which sex hormones contribute to mammographic density are complex and incompletely understood. We identified a positive association between the oestrogen receptor (ER)-related nuclear factor ERRF and AD, emphasising the link between oestrogen-mediated signalling and mammographic density.
Both the RAS pathway related protein SHC1, which transmits signalling of cell surface receptors to activate downstream pathways, and the homeobox protein IRX5, involved in cell differentiation and cell cycle regulation, were negatively associated with AD, ss was the acyl-coenzyme A oxidase ACOX2, part of the degradation of long branched fatty acids. AD was also positively associated with the membrane transport-protein ABCC11. Association between AD and proteins involved in cellular proliferation and control of metabolic functions is indicative of the complex dynamic control to maintain an internal steady state in high-density tissue.
Our study has also some limitations. Although we initially selected participants based on volumetric mammographic density, we performed the statistical analyses using the absolute area-based measurement of mammographic density. Current research has led us to believe that area is a better representation of the true dense tissue in the breast and thus the best measurement of mammographic density for analyses of plasma markers of density [
23,
37‐
41]. We also analysed the data in relation to VD in accordance with the initial strategy. When comparing women with different VD, we identified significantly elevated levels of forkhead box P3 (FOXP3) in the high VD compared to the low VD group (Fig.
4). AD and VD are differently associated with age and BMI, which may partly explain this discrepancy (Additional file
2: Table S8). Other limitations are that all exposure data, such as BMI, are self-reported, which may result in some misreporting. However, both exposure data and mammograms were collected at the same time at KARMA study entry. Noticeable is that we used plasma to identify proteins associated with mammographic density. It remains to be ascertained how well blood plasma protein concentrations reflect the protein expression in the breast tissue. Nonetheless, the identified epithelial and stromal cell-specific proteins support protein leakage, shedding or elevated turnaround in breast tissue leading to the detection of these proteins in the circulation. The strengths of our study reside in the large number of samples and the use of two independent sample sets from the KARMA study. This included the centralised collection of mammograms and blood samples, the quantitative assessment of mammographic density by STRATUS, and collection of background information on all participants [
15].
The affinity-based assay used in this study provides opportunities for high-throughput screening for novel proteins associated with disease or selected phenotypes. The design allows the combination of different protein assays in one multiplexed approach and it is attractive due to consumption of only minimal volumes of samples. We have taken great care in generating and assessing the data prior to statistical analysis (see Fig.
1) and the candidates presented provide leads for further studies. The method identifies relative protein quantities in plasma and would require the development of assays such as sandwich ELISA for the determination of actual protein concentrations. We have used four different assays to validate the antibodies (see Additional file
2: Figures S4-S6 and Additional file
1: Table S7). This demonstrates the challenge when working with antibodies in exploratory analyses: Depending on the assay sensitivity, sample preparation and target concentration, the performance of the antibody may differ between assays and cannot yet be predicted. Further investigations that preferentially use multiplexed sandwich ELISAs with the shortlisted targets will then allow us to quantify the proteins in abundance to monitor and compare alterations in these in relation to mammographic density in different study sets.
Acknowledgements
We thank all the participants in the KARMA study, the study personnel for their devoted work during data collection. We also thank the Märit and Hans Rausings Initiative Against Breast Cancer, the Swedish Research Council and the Kamprad Family Foundation for Entrepreneurship, Research & Charity. Also, we thank the Human Protein Atlas and its funders, the Knut and Alice Wallenberg Foundation. The KTH Center for Applied Precision Medicine funded by the Erling-Persson Family Foundation is acknowledged for financial support. This work was also supported by grants for Science for Life Laboratory, by the SRA grant from the Swedish Government (CancerUU), and grants from the Swedish Research Council for Health, Working Life and Welfare (FORTE), and the Swedish Cancer Society.