Background
Mammographic density (MD) refers to the percentage of radio-opaque fibroglandular breast tissue on a mammogram [
1]. In 1969, Wolfe et al. first proposed that an increased proportion of dense breast tissue might be associated with increased breast cancer (BC) risk [
2]; however, evidence was scarce at that time to support this hypothesis. Furthermore, dense breast tissue could present a masking effect for small tumors on mammography, making early cancer detection on mammograms challenging [
3]. Hence, it was widely believed that the increased BC risk was in fact secondary to the masking effect, rather than any real difference in cancer development due to MD [
4]. However, over the past 20 years, well-powered case-control and cohort studies have consistently shown that increased MD is a strong risk factor for BC, independent of any potential masking effect [
5‐
8]. In particular, McCormack and dos Santos Silva performed a meta-analysis of 14,000 cases and 226,000 noncases from 42 studies and found that the proportion of dense area, or percent MD (PMD), was consistently associated with BC risk [
9]. Subsequently, important questions raised were whether BC preferentially arose from tissue of HMD areas, and if so, what are the characteristics of HMD origin BC compared with LMD. Ursin and colleagues showed in a retrospective study that ductal carcinoma in situ (DCIS) lesions were more likely to develop from HMD than from LMD areas of the breast in 28 women, by comparing mammograms at BC diagnosis with the women’s previous mammograms [
10]. Other studies also found that BCs arising in HMD regions are more likely to demonstrate features that suggest poor prognosis than those that arise in LMD areas [
11‐
13].
The significance of MD-associated BC risk was highlighted by the fact that in 1993, the American College of Radiology developed the Breast Imaging Reporting and Data System (BI-RADS) system, which divides density qualitatively into four categories [
14,
15]. More recently, the Density Education National Survivors’ Effort (
www.areyou
dense.org) in the United States led a high-profile campaign that encouraged women to ask for additional investigations if their breast tissues were reported as mammographically dense [
16]. This subsequently led to bold legislation changes in 32 U.S. states to mandate physicians to inform their patients of their MD categories [
17,
18].
Although the association of HMD with increased BC risk has now been well established for years, the underlying biological mechanism of this association continues to perplex researchers. Many biological and molecular studies are beginning to unravel the complexities of the biology behind HMD-associated BC risk [
19‐
24]. Using paired HMD and LMD breast tissues from women undergoing prophylactic mastectomy, we and others have found that HMD breast tissue was associated with increased epithelium, stroma, and collagen and decreased fat percentages compared with LMD tissue; furthermore, HMD regions showed increased number of CD45
+ immune cells in the epithelium [
25,
26]. To date, there is little data on the association of MD with immune cell infiltration; however, immune cell infiltration is observed in early-stage BC (proliferative benign disease and DCIS) as well as invasive BC, where numbers can predict prognosis [
27]. In this study, we further investigated the innate and adaptive immune cell infiltration and their functional polarization in HMD and LMD normal breast tissue.
Methods
Patient accrual
This study was approved by the Peter MacCallum Human Research Ethics Committee (number 08/21) and St. Vincent’s Hospital Animal Ethics Committee (number 049/09). It was conducted in accordance with the Australian National Statement on Ethical Conduct in Human Research. Between 2008 and 2015, 54 women undergoing unilateral or bilateral prophylactic mastectomy at St. Vincent’s Hospital and the Peter MacCallum Cancer Centre consented to tissue collection through the Victorian Cancer Biobank (VCB 10010). The reasons for their mastectomy procedures were confirmed BRCA1/2 mutation carrier status, other confirmation mutations such as PTEN gene mutation, and a strong family or personal past history of BC. Women were excluded from the study if there was suspicion of malignant lesions on radiological investigations or if the breast that the mastectomy was performed on had been diagnosed with BC or DCIS in the past.
Selection of HMD and LMD regions within the same breast
Upon the completion of mastectomy, the resected breast was immediately sent to the pathology department on ice. Using sterile techniques, 1-cm-thick, craniocaudal breast slices were resected (breadboarding), palpated for suspicious stiffness, and a slice surplus to diagnostic needs was chosen. The breast slice was then X-rayed against a calibration strip using consistent radiological parameters by trained breast radiographers, followed by selection of high and low MD tissue regions for comparative study. The method was detailed in previous studies [
19‐
21,
26,
28,
29]. If the breast slice imaging did not result in clear black and white regions (allowing density differences to be observed) the woman was excluded from this study. The paired HMD and LMD tissues were then fixed in neutral buffered formalin, before undergoing tissue processing, embedding and sectioning at 4 μm thickness for subsequent staining.
IHC staining
Conventional H&E staining was performed on all paired high and low MD tissues. On the basis of H&E-stained slides, 15 paired high and low MD tissues were further selected for a focused examination of immune cell influx because they showed abundance of epithelial-stromal areas that represent the characteristics of mammary specimens.
These 15 paired high and low MD tissues underwent successive IHC staining for immune cell analyses. Manual staining was performed for pan-macrophage marker CD68 (1:200, Dako clone PG-M1, code M0876; Agilent Technologies, Santa Clara, CA, USA) and B lymphocyte marker CD20cy (1:100, Dako clone L26, code M0755; Agilent Technologies). A diaminobenzidine (DAB) IHC autostainer (Ventana Medical Systems, Tucson, AZ, USA) was used for dendritic cell (DC) marker CD11c (1:25, clone 5D11; Cell Marque, Rocklin, CA, USA); programmed cell death protein 1 (PD-1) (1:100, clone NAT105; Abcam, Cambridge, MA, USA); helper CD4+ T cell (1:50, AP20210 PU-N; Novus Biologicals, Littleton, CO, USA); cytotoxic CD8+ T cells (1:50, orb 10325; Biorbyt, San Francisco, CA, USA); cytokines IL-4 (1:500, AB9622; Abcam) IL-6 (1:800, AB6672; Abcam), and interferon (IFN)-γ (1:500, AB25101; Abcam), and natural killer cell (NK) marker CD56 (MRQ-42, Ventana® catalogue no. 760-2625; Cell Marque). For each staining, human benign tonsil tissue was used as a positive control, and no primary antibody was used as a technical negative control. In the case of IL-6, only nine paired samples were used because the other six pairs did not have enough tissue left following staining for the other immune cell markers.
Five-micrometer paraffin-mounted tissue sections were dewaxed and underwent antigen retrieval using either citrate buffer (CD68, CD11c, CD56, CD20cy, IL-4, and IL-6) or ethylenediaminetetraacetic acid (EDTA) (CD4, CD8, PD-1, and IFN-γ) and then incubated with the appropriate primary antibodies followed by biotinylated secondary antibodies. The VECTASTAIN Elite ABC kit (Vector Laboratories, Burlingame, CA, USA) with Dako DAB peroxidase (Agilent Technologies) used as chromogen.
For CD3 and PD1 double staining, we performed Opal multiplex imaging (PerkinElmer, Waltham, MA, USA) according to the manufacturer’s instructions. In brief, 3-μm formalin-fixed, paraffin-embedded sections were deparaffinized and then stained with rabbit monoclonal anti-CD3 (clone SP7; Spring Bioscience, Pleasanton, CA, USA) and a rabbit monoclonal anti-PD1 (Bio SB, Santa Barbara, CA, USA). Antigen retrieval was performed in high-pH EDTA in a pressure cooker for the first antibody and then EDTA in the microwave for subsequent antibodies. Endogenous peroxidases were blocked using hydrogen peroxidase, and following incubation with anti-rabbit secondary antibody, the immunofluorescent signal was visualized using TSA dye 570 or 650 from the Opal™ 7 color fIHC kit (PerkinElmer). All sections were counterstained with Spectral DAPI. Slides were imaged on the Vectra® 3.0 Automated Quantitative Pathology Imaging System (PerkinElmer). Color separation, tissue and cell segmentation, and phenotyping were performed using inForm® software version 2.2 (Perkin Elmer) to extract image data.
Histological review and digital image analysis
All slides were examined using digital microscopy (AxioVision photomicroscope; Carl Zeiss Microscopy, Thornwood, NY, USA) for punctate brown cytoplasmic staining of each individual immune cell marker. Four random glandular-stromal areas within each tissue sample were photographed at × 40 magnification. The number of positively stained immune cells and the total number of cells within each histological compartment (epithelial and stromal regions) were manually counted for each image. Thus, for the immune cell counts in the epithelial area, the total number of epithelial cell nuclei was counted on the selected region, and then the number of positively stained cells was counted. The number of positive cells was then presented as a percentage of total epithelial cells. Scoring was performed in a blinded manner. The percentage of positive staining was calculated as an average value derived from results of all four images for each tissue sample. For stroma, only cells with elongated nuclei and positive cytoplasmic brown staining were counted as positive stains.
Statistical analysis
For each immune cell marker staining, the data were first assessed for equal variance using a normality test. If data passed the normality test, a paired t test was used to analyze the percentage of positive immune cell staining in the epithelium and the stroma, respectively. When data were not normally distributed, a nonparametric Wilcoxon matched-pairs rank test was used. Outliers were identified using Grubbs’ test. A conventional two-tailed alpha level of 0.05 was used to define statistical significance.
Discussion
Following our work identifying an increase in the immune cells (CD45
+vimentin
+cytokeratin
− cells) within the epithelial region of dense breast tissue [
26], in the present study we show dense breast tissue has increased parenchymal macrophages, DC, and B cells, but not NK cells. HMD tissue also has a trend for increased CD4
+ T cells and a significant increase in expression of the checkpoint inhibitor PD-1, confirming that the T cells present are activated. IL-6 was increased in the epithelial regions, and both IL-6 and IL4 were increased in the stromal regions of HMD samples, but not IFN-γ. This suggests that the preneoplastic breast tissue of women with high breast density has increased numbers of innate and adaptive immune cells, with a proinflammatory functional polarization consistent with the increased risk of developing BC that has been associated with HMD.
BC has largely not been considered immunogenic, because incidence is not increased in patients who are immune-suppressed; however, there are now irrefutable data demonstrating that the immune cell infiltrate of a breast tumor affects its growth and metastasis [
27]. In addition, progression from normal to preinvasive and invasive BC has been associated with increases in T cells, B cells, and macrophages [
32,
33]. Degnim and colleagues assessed 11 normal breast samples and found increased T- and B-cell numbers in breast lobules with lobulitis and found that immune cells were present mainly in breast lobules rather than in the stroma and that cytotoxic T cells and DC were integrated within the epithelium [
34]. We also found macrophages, DC, NK cells, B cells, and T cells in normal breast tissue. Furthermore, we showed the expression of PD-1 and secretion of IL-6, IL-4, and IFN-γ in glands and stroma of normal breast, suggesting that there is a dynamic breast immune surveillance system. Similarly, Degnim and colleagues assessed benign breast disease tissue and found that this tissue had higher densities of multiple immune cell types, especially macrophages and DC, compared with normal breast tissues [
31]. Although our HMD tissue did not show signs of benign breast disease, we also found increased levels of these two innate immune cells, indicating that they may be important for the early immune response to breast changes that may, in some cases, develop into lesions.
We found that the number of innate immune cells was increased in HMD samples compared with LMD samples. Macrophages are phagocytic cells that act to maintain immune surveillance within tissues and constantly survey their surroundings for signs of tissue damage or invading organisms. They stimulate lymphocytes and other immune cells to respond when danger signals are phagocytosed and/or detected by cell surface receptors [
35]. In the present study, we show that macrophages were increased in the epithelial regions of HMD samples. Previously, we reported that epithelial macrophage numbers were not altered with density [
26]; however, only nine women and three random areas of each section were assessed. The larger sample sizes and increased areas used in the present study allowed significant macrophage changes to be revealed. The increased epithelial macrophages support our recent finding that chemokine ligand 2 (CCL2) or monocyte chemoattractant protein 1 (MCP-1) is significantly increased in HMD epithelium [
36]. CCL2 recruits macrophages but also stimulates protumorigenic M2 macrophage polarization [
36,
37], which is supported by the increased epithelial IL-6 expression that we report here, as well as its role in M2 orientation [
38].
DC are powerful antigen-presenting cells [
39] with molecular sensors enabling them to sense danger and to transmit this to lymphocytes to initiate the T-cell immune response [
40] and aid in tumor cell death. We found increased DC in HMD epithelial and stromal compartments compared with LMD, suggesting increased antigen presentation and hence potentially enhanced immune surveillance. Because the generation of tumor-specific T cells relies on the ability of mature DCs to cross-present tumor antigens, future studies will need to assess the functional status of DC to fully understand the implications of increased DC. NK cells are tumor cell- and virus-killing innate immune cells that do not need to match with a major histocompatibility complex (MHC) subclass, the way CD8
+ T cells do [
41]. NK cell dysfunction has been associated with BC progression [
42]. We found no significant difference in the percentage of NK cells between high and low MD, suggesting that NK cells do not play a key role in the inflammatory microenvironment of high breast density.
In addition to the changes in the innate immune cells, we found changes in the adaptive immune cells within high-density breast tissue. The percentage of B lymphocytes was significantly increased in HMD epithelium. B cells secrete antibodies and inflammatory cytokines and can recognize antigens, regulate antigen processing and presentation, and mount and modulate T-cell and innate immune responses. B-cell infiltration was been associated with worse outcome in patients with metastatic ovarian cancer and progression of orthotopic tumors in mice [
43], and their numbers increase with the progression of normal breast and benign proliferative disease through to DCIS and invasive ductal carcinoma [
33]. We postulate that the increased number of B cells in HMD may reflect changes in the breast immune surveillance and possibly increased differentiation of B regulatory cells, which themselves can drive T-regulatory differentiation of CD4
+ T cells.
The CD4
+ T lymphocytes were markedly increased in the HMD stroma, whereas no significant difference was observed in terms of CD8
+ T cells in either compartment. CD4
+ T cells carry out multiple functions, including activation of innate immune cells, B lymphocytes, and cytotoxic T cells, as well as nonimmune cells. They also play a critical role in the suppression of immune reaction. CD4
+ T cells can be either T-helper cells (Th) or regulatory T cells (Treg), and they are activated through two signals: T-cell receptor on the T cell and an antigenic peptide presented by MHC class II on the antigen-presenting cell requiring a second costimulatory signal [
44,
45]. Within the Th cell population, there are a number of subtypes, including Th1, Th2, Th9, Th17, Th22, and ThFH (Follicular T helper), that differ in their functions, signature cytokine profiles, and cell targets [
46]. Due to the increased levels of IL-6 and IL-4 in HMD breast tissue (but steady state of IFN-γ), we postulate that the increased CD4
+ T cells are Th2-oriented because IL-6 drives Th2 differentiation [
47,
48] and IL-4 production is enhanced by IL-6. IL-6 has already been associated with breast density in genome-wide association studies on cancer-free breast tissue, where genetic variations in nine tagging single-nucleotide polymorphisms in the IL-6 gene were significantly associated with HMD [
49]. Higher transcript expression of IL-6 has also been reported in HMD epithelial areas compared with LMD in BC [
50].
The PD-1/PD-L1 pathway is an inhibitory immune checkpoint pathway that is upregulated within the tumor microenvironment [
51,
52], and PD-1/PDL-1 checkpoint inhibitors are showing unprecedented clinical efficacy in certain cancer types, especially melanoma and lung [
53]. Its normal biological role lies in preventing overstimulation of the immune responses and helping maintain immune tolerance to self-antigens [
54,
55]. PD-1 is expressed on activated T cells but also on other immune cells, including activated B cells and NK cells [
56] and Tregs [
57]. We show that PD-1 expression was increased within the epithelial and stromal regions of HMD samples compared with LMD, suggesting an increased immune self-tolerance in the HMD. The levels of PD-1 in normal human breast (0.29–2.79%) (
see Table
2) are much lower than in BC, where expression ranges from 19% to 59% [
58,
59]. In murine BC studies, tumor-induced CD8
+ T cells that express a high level of PD-1 were found to be ineffective in controlling tumor growth [
60]. Thus, the increased level of PD-1 protein in HMD compared with LMD regions suggests that the function of T cells may be impaired, which may contribute to the increased BC risk that occurs in HMD.
Table 2
Overview of all immune cell analyses between high (HMD) and low (LMD) density breast samples
CD68 | 5.61 (0.61) | 2.56 (0.75) |
p
=0.004
| 4.72 (0.9) | 3.14 (0.98) | p=0.15 |
CD11c | 1.53 (0.36) | 0.5 (0.22) |
p
=0.02
| 5.30 (1.06) | 2.16 (0.80) |
p
=0.03
|
CD56 | 5.30 (1.41) | 2.29 (0.73) | p=0.07 | 1.81 (0.74) | 0.92 (0.36) | p=0.31 |
CD20 | 1.93 (0.52) | 0.31 (0.15) |
p
=0.0039
| 5.35 (2.03) | 1.50 (0.59) | p=0.28 |
CD4 | 20.47 (2.69) | 15.07 (2.55) | p=0.05 | 21.42 (2.63) | 15.95 (2.64) |
p
=0.04
|
CD8 | 6.69 (1.43) | 4.82 (1.75) | p=0.22 | 3.49 (0.83) | 2.18 (0.66) | p=0.15 |
PD1 | 2.79 (0.52) | 1.25 (0.35) |
p
=0.02
| 1.45 (0.46) | 0.29 (0.20) |
p
=0.04
|
IL-6 | 24.68 (4.13) | 18.34 (5.80) |
p
=0.03
| 21.07 (5.53) | 11.24 (3.98) |
p
=0.006
|
IL-4 | 23.94 (5.19) | 16.04 (3.60) | p=0.09 | 4.41 (1.29) | 1.28 (0.59) |
p
=0.04
|
IFNγ | 4.66 (0.88) | 2.22 (0.80) | p=0.06 | 1.41 (0.69) | 0.67 (0.36) | p=0.56 |