A total of 116 specimens of invasive breast carcinomas were obtained from Japanese female patients who had undergone surgical treatment from 2007 to 2008 at Tohoku University Hospital, Sendai, Japan. The clinical outcome was evaluated by disease-free and breast cancer-specific survival. Disease-free survival was defined as the time from the date of surgery to that of the first locoregional recurrence or distant metastasis within the follow-up time and the median was 59 (range 3–84) months. Breast cancer-specific survival was defined as the time from surgery to death from breast cancer and follow-up time was 61 (from 3–84) months. All specimens had been fixed with 10% formalin neutral buffer solution and embedded in paraffin wax. Experiments and analyses were performed in accordance with the Helsinki declaration and the research protocol of this study was approved by the Ethics Committee at the Tohoku University Graduate School of Medicine (approval no. 2016-1-697, 2019-1-219). This is a retrospective study and, therefore, informed consents had not been obtained.

The information regarding the antibodies is listed in Table SI. Immunohistochemistry for CD163, 5αRed1 and Ki67 was carried out using a Histofine kit (Nichirei Bio, Inc., Tokyo, Japan) as described previously (23,24). The antigen-antibody complex was visualized with 3,3′-diaminobenzidine (DAB) and hematoxylin was used for counterstaining. Immunohistochemistry for ER and progesterone receptor (PR) was carried out with Ventana Benchmark XT system (Roche Diagnostics Japan). For human epidermal growth factor receptor 2 (HER2), the HercepTest (Dako) was used.

Immunoreactivity of AR was firstly visualized with DAB as described above, and then antigen retrieval was carried out again and immunoreactivity of CD163 was visualized with nitro blue tetrazolium (NBT)/5-bromo-4-chloro-3-indolyl phosphate (BCIP).

5αRed1 immunoreactivity was detected in the cytoplasm of breast carcinoma cells and considered positive when the cases had more than 10% of positive carcinoma cells (3). Infiltration of CD163-positive macrophages was scored into 4 categories (0, no focal areas; 1, small scattered focal areas; 2, numerous larger focal areas; 3, very numerous large focal areas) according to a previous study (25), and finally dichotomized for statistical analyses (0, 1: Low infiltration and 2, 3: High infiltration). ER, PR and Ki67 immunoreactivity were detected in the nucleus of carcinoma cells and the percentage of positive cells (labeling index; LI) was calculated by evaluating in more than 1,000 cells. According to a previous report (26), cases with LI of more than 1% were considered positive for ER and PR. HER2 immunoreactivity was determined according to the grading system proposed in HercepTest (Dako).

Mouse breast cancer cell line 4T1 and mouse macrophage cell line RAW264.7 were obtained from American Type Culture Collection (ATCC) and RIKEN Bioresource Center (Japan), respectively. These cells were cultured in RPMI-1640 medium (Sigma-Aldrich; Merck KGaA) with 10% fetal bovine serum (FBS) (Biosera) and incubated at 37°C under 5% CO₂. In experiments using androgen, these cells were cultured in phenol red-free RPMI-1640 with 10% dextran-coated charcoal-stripped FBS.

The 4T1 cells were cocultured with RAW264.7 cells using ThinCerts™ (pore size 0.4 µm, Greiner bio-one). RAW264.7 cells were seeded in the bottom plate (3×10⁵ cells/well) and allowed to attach for 24 h. Then, the 4T1 cells were placed onto an upper Transwell (2×10⁵ cells/well) and cocultured with RAW264.7 cells for 3 days.

Total RNA (500 ng) was extracted using TRI Reagent (Molecular Research Center, Inc.) and cDNA was synthesized using a ReverTra Ace qPCR RT Master Mix with gDNA Remover (Toyobo Co. Ltd.). Real-time PCR was performed using the THUNDERBIRD™ SYBR^® qPCR Mix (Toyobo Co. Ltd.) and LightCycler nano system (Roche Diagnostics Japan). Thermocycling conditions were as follows: 95°C for 60 sec (initial denaturation), followed by 45 cycles at 95°C for 15 sec and 60°C for 30 sec. The sequences for the PCR primer sets are listed in Table SII. Relative mRNA levels of Arg-1, Ar, Srd5a1 and Akr1c6 were summarized as the ratio to Rpl13a.

Open reading frame of Ar and FLAG-tagged aldo-keto reductase family 1 member C6 (Akr1c6, mouse testosterone-producing enzyme) were amplified via PCR using KOD FX Neo (Toyobo Co. Ltd.) and cloned into pIRES2-AcGFP1 vector (Clontech Laboratories) and pcDNA3.1 (−) vector (Invitrogen; Thermo Fisher Scientific, Inc.), respectively, using DNA Ligation Kit <Mighty Mix> (Takara Bio) (pIRES-mAR and pAKR1C6-FLAG). Thermocycling conditions were as follows: 94°C for 2 min, followed by 35 cycles at 98°C for 10 sec, 62°C for 30 sec and 68°C for 90 sec. The sequences for the PCR primer sets are listed in Table SII.

Since 4T1 and RAW264,7 cells do not express Akr1c6 and Ar, respectively, we tried to establish their sublines which stably express these genes in order to investigate the roles of androgens in intratumoral macrophages. According to the manufacturer's instructions for transfection, RAW264.7 cells were transfected with pIRES-mAR or pIRES2-AcGFP1 using GenomONE™-Neo (Ishihara Sangyo Kaisya, Ltd.) and 4T1 cells were transfected with pAKR1C6-FLAG (4T1-AKR1C6) or pcDNA3.1(−) (4T1-CT) using Avalanche^®-Everyday Transfection Reagent (APRO Science). These cells were cultured in media containing 400 µg/ml G418 (Wako), and stable clones expressing AR (RAW-AR), AKR1C6 (4T1-AKR1C6) and corresponding negative control clones (RAW-CT and 4T1-CT) were established.

Western blotting was performed according to previous reports (27,28) and the information regarding the primary antibodies is listed in Table SI. The cells were lysed using M-PER Mammalian Protein Extraction Reagent (Pierce Biotechnology) containing Halt Protease Inhibitor Cocktail (Sigma-Aldrich; Merck KGaA). The protein extracts (10 µg) were separated by SDS-PAGE (10% acrylamide gel) and then transferred to Hybond P polyvinylidene difluoride membrane (GE Healthcare). The membrane was blocked in 5% non-fat milk in TBS-T for 1 h at room temperature and then incubated with primary antibodies at 4°C overnight. The blots were incubated with the appropriate HRP-conjugated secondary antibody (anti-mouse; 1:10,000, cat. no. NA931 or anti-rabbit; 1:10,000, cat. no. NA934, GE Healthcare) for 1 h at room temperature. Detection of antibody-protein complexes on the membrane was visualized using ECL-Prime Western blotting detection reagents (GE Healthcare) and LAS-4000 image analyzer (Fuji Photo Film Co.).

RAW-CT and RAW-AR cells (1×10⁵ cells/well) were cultured in a 12-well plate with or without synthetic androgen R1881 (10 nM) for 48 h.

A sphere-forming assay was performed using EZ-BindShut^®II micro plate (IWAKI). The 4T1 cells (1×10⁵ cells/well) were cocultured in a well with RAW-CT or RAW-AR cells (3.33×10⁴ cells/well) with or without R1881 (10 nM). After 3 days, spheres were photographed under a microscope and the size of 100 spheres was measured using ImageJ 1.53e (https://imagej.nih.gov/ij/).

The 4T1-AKR1C6 or 4T1-CT cells were seeded in a 96-well plate (3,000 cells/well) and the cell proliferation was measured by the WST-8 method using a Cell Counting Kit-8 (Dojindo Molecular Technologies) for 4 days. Absorbance was determined using a Bio-Rad iMark plate reader (Bio-Rad Laboratories Inc.).

The protocols of all animal experiments were submitted, reviewed and approved in advance by the Institutional Animal Care and Use Committee at Tohoku University (2019MdA-172). The experiments were carried out according to ARRIVE guidelines and Guidelines for Animal Experiments at Tohoku University. A total of 11 4-week-old wild-type female BALB/c mice were obtained from CLEA Japan (Tokyo, Japan) and randomly assigned to two types of tumor-bearing models as described below with no inclusion criteria. They were housed in a specific pathogen-free facility with free access to water and MR stock food and environmental enrichment. In the present study, 4T1 and RAW264.7 sublines were subcutaneously injected in both the left and right side of mice in different combinations in order to avoid individual differences. In the first experiment (n=6), 4T1-CT or 4T1-AKR1C6 cells (5×10⁵ cells) were subcutaneously injected together with RAW-AR cells (1.25×10⁵ cells) into the left or right side of the back of mice, respectively, using 150 µl Matrigel (Corning, Inc.). In the mirror experiment (n=5), 4T1-AKR1C6 cells (5×10⁵ cells) were injected together with either RAW-CT or RAW-AR cells (1.25×10⁵ cells) into the left or right side of the back of mice, respectively. Mice were monitored for body weight and tumor volume changes twice a week. Tumor volume was measured using a digital caliper at least twice a week and calculated by modified ellipsoid formula 1/2 (length × width²). Tumors were allowed to grow in mice until they reached the volume of 2,000 mm³ or 10 weeks from injection had elapsed, whichever came first. No adverse events were observed and mice were euthanized 4 weeks after injection by cervical dislocation under anesthesia using isoflurane (4%) and thereafter tumors were resected, weighted, and then fixed in 10% formalin for immunohistochemistry.

Statistical analyses were performed using JMP pro 14.0.0 (SAS Institute). χ² test or Mann-Whitney U test were applied for correlation between the infiltration of CD163-positive macrophages and clinicopathological factors. Breast cancer-specific and disease-free survival curves were generated by Kaplan-Meier method and statistical significance was evaluated by a log-rank test. In the in vitro and in vivo experiments, the statistical analyses were performed using paired or unpaired t-test. The data are presented as the mean ± SD. In this study, P<0.05 was considered as indicative of statistical significance.

We first examined whether macrophages in breast carcinoma tissues expressed AR or not. By double immunohistochemistry for AR and CD163, a marker of M2 macrophages, the macrophages with immunoreactivity for both CD163 (blue) in the cell membrane and AR (brown) in the nuclei were sporadically observed in the stroma (Fig. 1A).

We next immunolocalized CD163 as well as 5αRed1 in 116 breast carcinoma tissues in order to address the clinical significance of androgen action on macrophages in breast cancer. CD163 immunoreactivity was observed in the macrophages (Fig. 1B and C) and 34% (39 out of 116 cases) were categorized as high infiltration. 5αRed1 immunoreactivity was observed in the cytoplasm of breast carcinoma cells (Fig. 1D and E) and 56% (65 out of 116 cases) were considered positive for 5αRed1. Correlation between macrophage infiltration and clinicopathological parameters is presented in Table I. Macrophage infiltration was positively correlated with lymph node metastasis (P=0.0057), histological grade (P<0.0001) and Ki67 labeling index (LI) (P<0.0001), while it was negatively correlated with ER (P=0.0020) and PR (P=0.0014). No significant correlation was detected between macrophage infiltration and 5αRed1 immunoreactivity (P=0.74). When we compared the correlation between macrophage infiltration and clinicopathological parameters in the 5αRed1-negative group (n=51) and the 5αRed1-positive group (n=65) (Table II), macrophage infiltration was correlated with lymph node metastasis (P=0.011), PR (P=0.0061) and HER2 (P=0.015) only in the 5αRed1-positive group. On the other hand, it was negatively correlated with ER only in the 5αRed1-negative group (P=0.001), while it was correlated with histological grade and Ki67 LI in both the 5αRed1-positive and -negative group (histological grade; P=0.0013 in the 5αRed1-negative group and P=0.0069 in the 5αRed1-positive group, Ki67 LI; P=0.0017 in the 5αRed1-negative group and P=0.0020 in the 5αRed1-positive group). Furthermore, it is of interest that a significant correlation between macrophage infiltration and these parameters was detected only in the 5αRed1-positive group when the cases were limited to the ER-positive group.

We next examined the correlation between macrophage infiltration and clinical outcomes of breast cancer patients. As shown in Fig. 2A and B, macrophage infiltration was significantly correlated with an increased risk of recurrence (P=0.0049) and worse prognosis (P=0.016). In contrast, when we analyzed according to 5αRed1 status, macrophage infiltration was significantly correlated with increased risk of recurrence (P=0.01, Fig. 2C) and worse prognosis (P=0.045, Fig. 2E) only in the 5αRed1-positive group and no significant correlation was detected in the 5αRed1-negative group (Fig. 2D and F). This tendency was still observed when the cases were limited to the ER-positive group (Fig. 2G-I), although breast cancer-specific survival curve could not be generated because no patients had died in the ER-positive/5αRed1-negative group.

We suggested that androgens may have important roles in macrophage-induced breast cancer progression by immunohistochemical analyses for CD163 and 5αRed1 as described above. We then tested this hypothesis in in vitro experiments using mouse breast cancer and macrophage cell lines, 4T1 and RAW264.7.

When RAW264.7 cells interacted with 4T1 cells in a coculture chamber, mRNA and protein expression of Arg-1, an M2 macrophage marker was significantly increased when compared to the control (monoculture group), indicating that RAW264.7 cells were successfully polarized into an M2 phenotype (Fig. 3A). We next investigated mRNA expression, in 4T1 and RAW264.7 cells, of mouse AR (Ar) as well as 5αRed1 (Srd5a1) and Akr1c6, the ortholog of human AKR1C3 and involved in androgen synthesis in the mouse (29). Real-time PCR analysis showed that Ar (Fig. 3B) and Akr1c6 (Fig. 3C) mRNA was almost negligible in both cell lines, while Srd5a1 mRNA was detected in both (Fig. 3D). In addition, the mRNA level of these genes was almost comparable in coculture condition (Fig. 3E and F). To investigate the effects of androgens on macrophages, we generated RAW264.7 sublines expressing AR (RAW-AR) and the corresponding negative control (RAW-CT) (Fig. 3G). When we treated RAW-CT and RAW-AR cells with synthetic androgen R1881, mRNA expression of M2 macrophage marker Arg-1 was not affected by R1881 (Fig. 3H). We confirmed that AR protein was not detected in 4T1 cells (Fig. 3G), and 4T1 cells were cocultured with RAW-CT or RAW-AR cells and cell proliferation was investigated by sphere-forming assay. As shown in Fig. 3I, sphere size was significantly increased by R1881 when the cells were cocultured with RAW-AR (P<0.01), while R1881 had no effect on monocultured 4T1 cells or those cocultured with RAW-CT.

In order to further address whether androgen acts on macrophages to promote breast cancer progression, we performed in vivo experiments using tumor-bearing mice. We generated 4T1 cells stably expressing AKR1C6 (4T1-AKR1C6) using pAKR1C6-FLAG vector or control vector (Fig. 4A). When we compared in vitro cell proliferation of these cells, it was comparable in 4T1-AKR1C6 and 4T1-CT cells (Fig. 4A). 4T1-CT or 4T1-AKR1C6 cells were subcutaneously injected together with RAW-AR cells on the left or right side of the back of mice, respectively (Fig. 4B; left image). The tumors on the right side (4T1-AKR1C6+RAW-AR) grew more rapidly compared with those on the left side (4T1-CT+RAW-AR) (Fig. 4C and D). In addition, immunohistochemistry demonstrated that Ki67 LI was higher in the tumor on the right side (70%) compared to those on the left side (30%) (Fig. 4E). We then performed the mirror experiment, where 4T1-AKR1C6 cells were subcutaneously injected together with RAW-CT and RAW-AR cells on the left or right side of the back of mice, respectively (Fig. 4B; right image). The tumors on the right side (4T1-AKR1C6+RAW-AR) grew more rapidly compared with those on the left side (4T1-AKR1C6+RAW-CT) (Fig. 4F and G). Ki67 LI was higher in the tumors on the right side (80%) compared to those on the left side (30%) (Fig. 4H).