Introduction

The role of tumor microenvironment has become increasingly significant as cancer research progresses. Tumor microenvironment is the non-transformed element located in the tumor area, consisting of the immune system component (such as macrophages and lymphocytes), blood vessel cells, fibroblast, myofibroblast, mesenchymal stem cells, adipocytes, and extracellular matrix (ECM). The most important and best studied elements of tumor microenvironments are cancer-associated fibroblasts (CAF) [1–3], which surround cancer cells and are related to tumor initiation [4], tumor-stimulatory inflammation [5], metabolism [6], metastasis [7], drug response [8], and immune surveillance [9]. Despite their important effects on cancer, no specific cell origin of CAFs has been identified, and the accurate definition of CAF remains controversial [1, 10]. Many possible CAF markers have been suggested, such as α-smooth muscle actin [11], tenascin-C [12], chondroitin sulfate proteoglycan (NG2) [13], platelet-derived growth factor receptor (PDGFR) α/β [14], fibroblast activation protein (FAP) [15], podoplanin [16], prolyl 4-hydroxylase [17], and fibroblast-specific protein (FSP) 1 [13]. Therefore, CAFs are thought to have various functional subtypes, and a recent study has divided them into four main subsets as follows: FAPα, FSP1, PDGFRα, and PDGFRβ types with different features [18]. It is reported that FAP-type CAFs are activated or reactive [19] and associated with modulation of ECM [20], tumor cell invasion [20], and immunomodulatory function, while FSP1-type CAFs are associated with metastatic colonization [21], macrophage infiltration [22], and protection from carcinogens [23]. Also, it is shown that PDGFRα-type CAFs have a relationship with paracrine signaling for tumor cell growth and angiogenesis [24], as well as macrophage recruitment [25].

The cross-talk between CAFs and breast cancer cells has been actively studied, suggesting that CAFs might exert significant effects on tumors as breast cancer is representative of malignancies with a diverse amount of tumor stroma. Furthermore, CAFs have been associated with tumor progression [26], invasion or metastasis [27, 28], therapeutic resistance [29], and prognosis in breast cancer [30]. However, the correlation between breast cancer and CAF phenotype has not been well defined. The purpose of this study was to investigate the relationship between clinicopathologic factors and different breast cancer CAF phenotypes.

Materials and methods

Patient selection and histologic evaluation

The present study included patients diagnosed with invasive ductal carcinoma, not otherwise specified, at Severance Hospital from January 2000 to December 2006. Patients receiving chemotherapy, radiotherapy, or hormone treatment before surgery were excluded. In total, there were 642 cases. This study was approved by the Institutional Review Board of Yonsei University Severance Hospital, which also exempted the requirement for patients’ informed consent. All cases were reviewed by a breast pathologist (Koo JS) using hematoxylin and eosin (H&E)-stained slides. Histological grades were assessed according to the Nottingham grading system [31]. Clinicopathologic parameters evaluated in each case included patients’ age at the initial diagnosis, lymph node metastasis, tumor recurrence, distant metastasis, and survival.

The tumor stroma of all breast cancer cases was classified based on microscopic observation as follows: desmoplastic type, when the tumor stroma was composed of cellular fibroblast or myofibroblast proliferation (the number of fibroblasts or myofibroblasts/high power field (HPF): >30); sclerotic type, when the cell component was low (the number of fibroblasts or myofibroblasts/HPF: ≤30) and the tumor stroma was composed of fibrotic collagenous tissue; normal-like type, when no stromal reaction near the tumor and normal breast stroma was observed (different from desmoplastic type in the lack of stromal reaction, and different from sclerotic type in the lack of sclerotic collagenous tissue); and inflammatory type, when the tumor stroma was mostly composed of inflammatory cells such as lymphocytes (the number of inflammatory cells/HPF: >200).

Tissue microarray

A representative area of tumor and tumor stroma was selected on an H&E-stained slide, and a corresponding spot was marked on the surface of the paraffin block. The selected area was obtained using a biopsy needle, and a 3-mm tissue core was transferred to a 6 × 5 recipient block. Two such tissue cores were extracted from invasive tumors to minimize extraction bias. Each tissue core was assigned a unique tissue microarray location number that was linked to a database containing other clinicopathologic data.

Immunohistochemistry

All antibodies used for immunohistochemistry are listed in Table 1. Immunohistochemical studies were performed using formalin-fixed, paraffin-embedded tissue sections. Briefly, 5-μm-thick sections were obtained using a microtome, transferred onto adhesive slides, and dried at 62 °C for 30 min. After incubation with primary antibodies, immunodetection was performed with biotinylated anti-mouse or anti-rabbit immunoglobulin, followed by peroxidase-labeled streptavidin using a labeled streptavidin biotin kit with 3,3′-diaminobenzidine chromogen as the substrate. The previous antibody incubation stem was omitted in negative controls. Tissue samples for positive controls were obtained as per the manufacturer’s recommendation. Slides were counterstained with Harris hematoxylin.

Table 1 Source, clone, and dilution of the antibodies used
Full size table

Interpretation of immunohistochemical staining results

All immunohistochemical staining was evaluated by light microscopy. A cut-off value of 1 % or more positively stained nuclei was used to define estrogen receptor (ER) and progesterone receptor (PR) positivity [32]. Human epidermal growth factor receptor 2 (HER-2) staining was analyzed according to the American Society of Clinical Oncology/College of American Pathologists guidelines using the following categories: 0, no immunostaining; 1+, weak incomplete membranous staining, less than 10 % of tumor cells; 2+, complete membranous staining, either uniform or weak in at least 10 % of tumor cells; and 3+, uniform intense membranous staining in at least 30 % of tumor cells [33]. HER-2 immunostaining was considered positive when strong (3+) membranous staining was observed, whereas cases with 0 to 1+ staining were regarded as negative, and those showing 2+ expression were further evaluated for HER-2 amplification by fluorescent in situ hybridization (FISH). The HER-2 gene copy number on the slides was evaluated using an epifluorescence microscope (Olympus, Tokyo, Japan). At least 60 tumor cell nuclei in three separate regions were investigated for HER-2 and chromosome 17 signals. HER-2 gene amplification was determined using the ASCO/CAP guidelines [33]. An absolute HER-2 gene copy number lower than 4 or an HER-2 gene-to-chromosome 17 ratio (HER-2/Chr17 ratio) less than 1.8 was considered HER-2 negative. An absolute HER-2 copy number between 4 and 6 or an HER-2/Chr17 ratio between 1.8 and 2.2 was considered HER-2 equivocal. Also, an absolute HER-2 copy number greater than 6 or an HER-2/Chr17 ratio higher than 2.2 was considered HER-2 positive.

Immunohistochemical staining for podoplanin, prolyl 4-hydroxylase, FAPα, S100A4, PDGFα, PDGFβ, and NG2 was accessed via light microscopy. Semi-quantitative evaluation was performed on stained slides using previously reported methods [34]. Tumor and stromal cell staining was assessed as follows: 0, negative, or weak immunostaining in <1 % of the tumor/stroma; 1, focal expression in 1–10 % of the tumor/stroma; 2, positive in 11–50 % of the tumor/stroma, and 3, positive in 51–100 % of the tumor/stroma. The whole area of the tumor was evaluated and defined as positive when the score exceeded 2. Immunostaining only in the ECM was discounted.

Tumor phenotype classification

In this study, we classified breast cancer phenotypes according to the immunohistochemistry results for ER, PR, HER-2, and Ki-67 as well as FISH results for HER-2 as follows [35]: luminal A type, ER or/and PR positive, HER-2 negative and Ki-67 labeling index (LI) <14 %; luminal B type (HER-2 negative), ER or/and PR positive, HER-2 negative, and Ki-67 LI ≥14 %; luminal B type (HER-2 positive), ER or/and PR positive, and overexpressed or/and amplified HER-2; HER-2 overexpression type, ER and PR negative, and overexpressed or/and amplified HER-2; and triple-negative breast cancer (TNBC) type, ER, PR, and HER-2 negative.

Statistical analysis

Data were analyzed using SPSS for Windows, version 12.0 (SPSS Inc., Chicago, IL, USA). For determination of statistical significance, Student’s t and Chi-square tests were used for continuous and categorical variables, respectively. In cases of multiple comparisons, a corrected p-value with the application of the Bonferroni multiple comparison procedure was used. Statistical significance was set to p < 0.05. Kaplan–Meier survival curves and log-rank statistics were employed to evaluate time to tumor recurrence and overall survival. Multivariate regression analysis was performed using the Cox proportional hazards model.

Results

Characteristics of breast cancer cases

In this study of 642 cases, 275 were of luminal A type (42.8 %), 152 of luminal B type (23.7 %), 58 of HER-2 type (9.0 %), and 157 were TNBC (24.5 %). TNBC patients had a higher histologic grade (p < 0.001), T stage (p = 0.004), Ki-67 LI (p < 0.001), tumor recurrence (p = 0.013), and death rate (p = 0.008) than other subtypes, whereas those with HER-2 type had a higher rate of older age (p = 0.014) (Table 2).

Table 2 Clinicopathologic characteristics of patients according to breast cancer molecular subtype
Full size table

Clinicopathologic features according to stromal phenotype

When the studied cases were categorized based on the tumor stromal histology, 181 were of desmoplastic type (28.1 %), 62 of inflammatory type (9.7 %), 25 of normal-like type (3.9 %), and 374 of sclerotic type (58.3 %). There were significant differences in histologic grade (p = 0.006), ER status (p < 0.001), PR status (p < 0.001), HER-2 status (p = 0.020), Ki-67 LI (p < 0.001), and molecular subtype (p < 0.001) according to tumor stromal phenotype. The inflammatory type exhibited higher histologic grade, ER negativity, PR negativity, HER-2 positivity, and higher Ki-67 LI, whereas the sclerotic type showed lower histologic grade, ER positivity, PR positivity, and lower Ki-67 LI. Moreover, TNBC was more frequently observed among tumors of inflammatory and normal-like types, whereas luminal A was prominent in desmoplastic and sclerotic types (Table 3).

Table 3 Clinicopathologic characteristics of patients according to breast cancer stromal histologic phenotype
Full size table

Expression of CAF-related proteins according to molecular subtype

The expression of all CAF-related proteins, except PDGFRβ and NG2, was observed in not only the stromal component but also cancer cells. Cancer cell expression of podoplanin and S100A4 was more commonly observed in HER-2 type (p = 0.001 and p < 0.001, respectively), whereas expression of FAPα and PDGFRα was more commonly seen in TNBC (p < 0.001). On the other hand, luminal types less commonly expressed FAPα, S100A4, and PDGFRα (p < 0.001).

All CAF-related proteins were more commonly expressed in the stromal component of HER-2 type (p < 0.05), whereas luminal A type less commonly expressed FAPα, PDGFRα, PDGFRβ, and NG2. In addition, the expression of podoplanin, prolyl 4-hydroxylase, and S100A4 was less commonly observed in TNBC (Table 4; Fig. 1).

Table 4 Expression of CAF-related proteins according to breast cancer subtype
Full size table
Fig. 1
figure 1

Expression of CAF-related proteins according to molecular subtype. Tumoral expression of podoplanin and S100A4 was more commonly seen in HER-2 type, whereas FAPα and PDGFRα expressions were more commonly seen in triple-negative breast cancer cells. The expression of CAF-related proteins in the stromal component was as high as their overall expression in HER-2 type. PDGFRβ and NG2 were not expressed in cancer cells

Full size image

Expression of CAF-related proteins according to stromal phenotype

The expression of prolyl 4-hydroxylase and PDGFRα in cancer cells differed depending on stromal phenotype (p = 0.003 and p < 0.001, respectively). Tumoral expression of prolyl 4-hydroxylase was more commonly seen in the desmoplastic type yet observed less in the inflammatory type, whereas PDGFRα expression in tumor cells was more commonly seen in the normal-like type and less commonly seen in the sclerotic type.

Furthermore, the expression of all CAF-related proteins in the stromal component also differed according to stromal phenotype (p < 0.001). The desmoplastic type more commonly expressed podoplanin, prolyl 4-hydroxylase, S100A4, PDGFRα, and PDGFRβ in the stromal component, whereas FAPα, PDGFRα, PDGFRβ, and NG2 expression in the sclerotic type was less commonly seen. The inflammatory type exhibited FAPα and NG2 expressions more commonly and podoplanin expression less commonly in the stromal component, whereas the normal-like type less commonly expressed prolyl 4-hydroxylase and S100A4 (Fig. 2; Table 5).

Fig. 2
figure 2

Expression of CAF-related proteins according to stromal histologic phenotype. Tumor cell expression of prolyl 4-hydroxylase was high in the desmoplastic type, whereas that of PDGFRα was high in the normal-like type. Desmoplastic type had high expression of podoplanin, prolyl 4-hydroxylase, S100A4, PDGFRα, and PDGFRβ in the stromal component, while expression of FAPα and NG2 was high in the inflammatory type. Both PDGFRβ and NG2 were not expressed in cancer cells

Full size image
Table 5 Expression of CAF-related proteins according to stromal phenotype
Full size table

Correlation between the expression of CAF-related proteins and clinicopathologic factors

Stromal FAPα positivity (p < 0.001), tumor S100A4 positivity (p = 0.001), and tumor PDGFRα positivity (p < 0.001) were related to higher histologic grade. Stromal NG2 positivity (p < 0.001), stromal PDGFRβ positivity (p = 0.003), tumor FAPα positivity (p < 0.001), tumor S100A4 positivity (p < 0.001), and tumor PDGFRα positivity (p < 0.001) were associated with ER negativity. PR negativity was related to stromal NG2 positivity (p < 0.001), tumor FAPα positivity (p < 0.001), tumor S100A4 positivity (p < 0.001), and tumor PDGFRα positivity (p = 0.002). On the other hand, HER-2 positivity was associated with stromal PDGFβ positivity (p < 0.001), tumor podoplanin positivity (p = 0.001), stromal prolyl 4-hydroxylase positivity (p < 0.001), and stromal S100A4 positivity (p < 0.001). In addition, stromal NG2 positivity (p < 0.001), stromal PDGFRβ positivity (p < 0.001), tumor prolyl 4-hydroxylase negativity (p < 0.001), stromal FAPα positivity (p < 0.001), tumor S100A4 positivity (p = 0.001), tumor PDGFRα positivity (p < 0.001), and stromal PDGFRα positivity (p = 0.001) were related to higher Ki-67 LI (Fig. 3).

Fig. 3
figure 3

Correlation between expression of CAF-related proteins and clinicopathologic factors. Y-axis represents the number of cases, T tumor, S stroma, TG tumor grade, ER estrogen receptor, PR progesterone receptor, LI labeling index

Full size image

Impact of CAF-related protein expression on patient prognosis

Univariate analysis revealed that tumor S100A4 negativity and S100A4 positivity was related to shorter disease-free survival (DFS, p = 0.015) and overall survival (OS, p = 0.016), respectively (Table 6; Fig. 4). Furthermore, analysis by molecular subtype showed that tumor podoplanin positivity was associated with shorter OS in luminal A (p = 0.002), while tumor FAPα positivity with shorter OS in luminal B (p = 0.018). Additionally, stromal phenotype was associated with varying DFS in TNBC, with normal-like type having shorter DFS (p = 0.008) (Fig. 5).

Table 6 Univariate analysis of the impact of CAF-related proteins expression in breast cancer on disease-free and overall survival by the log-rank test
Full size table
Fig. 4
figure 4

Impact of CAF-related protein expression on patient prognosis in breast cancer

Full size image
Fig. 5
figure 5

Impact of CAF-related protein expression on patient prognosis according to the molecular subtype of breast cancer

Full size image

Discussion

Our results from the present study indicated differential expression of CAF-related proteins in the stromal component of breast cancer according to the molecular subtype. Different molecular subtypes of breast cancer exhibit various clinical, histological, molecular, and therapeutic features, possibly owing to different cancer cell characteristics [36–39]. Consequently, the cross-talk between breast cancer cells and the tumor stroma might differ, affecting tumor microenvironment [40]. In the present study, we tested the hypothesis that the tumor stroma exhibited different characteristics depending on breast cancer subtype. Differential CAF gene expression according to breast cancer molecular subtype has been reported previously [41], with the expression of genes associated with cancer cell migration being higher in CAFs from HER-2 type than those from TNBC and ER-positive type. Our results also showed high expression of all CAF-related proteins in HER-2 type (p < 0.05). Since CAF-related proteins are reportedly associated with tumor invasiveness [18], such results were in agreement with those from prior studies.

In this study, the expression of all CAF-related proteins in the stromal component also differed according to stromal histologic phenotype, with the desmoplastic type showing high expression of podoplanin, prolyl 4-hydroxylase, S100A4, PDGFRα, and PDGFRβ, and the inflammatory type having high FAPα and NG2 expressions. Although the tumor stroma of breast cancer has been histologically categorized in previous studies [42], the differences in CAF-related protein expression according to the tumor stroma histology has not been investigated. Nonetheless, CAF markers are known to possess distinct features, despite difficult comparison. Our results were in agreement with previous reports that the desmoplastic responses of PDGFRα-type CAFs occurred via PDGF-AA signal in breast cancer [43]. Similarly, we observed high FAPα expression in the inflammatory type, which was in agreement with reports that FAPα was associated with immunomodulatory function [18]. Therefore, our findings suggested a relationship between CAF features and tumor stromal histology, which warrants further investigation.

Furthermore, we found that tumor expression of CAF-related proteins, but not their expression in the stromal component was related to patient prognosis, and that tumor S100A4 positivity was associated with poor prognosis in breast cancer. Previous studies have demonstrated such a relationship between tumor expression of S100A4 and poor prognosis in hepatocellular carcinoma [44], endometrial carcinoma [45], lung cancer [46], and pancreatic cancer [47]. In addition, our results on the association between poor prognosis and tumor podoplanin positivity in luminal A type were also in agreement with other reports on podoplanin expression in esophageal cancer [48] and oropharyngeal cancer [49]. Similar results on FAPα were also described in osteosarcoma [50] and pancreatic cancer [51]. Therefore, CAF-related proteins could be used as prognostic markers in breast cancer. This study shows that normal-like stroma are associated with shorter DFS in TNBC; however, the previous study suggested that stroma-high-type TNBC tumors (those with a high rate of stroma) have a poor prognosis compared to stroma-low-type tumors, which indicates that stroma rate is associated with prognosis in TNBC [52]. However, this study did not include stroma rates and is only categorized within a histological view; therefore, further study is required.

The clinical implication of our findings was that CAFs might serve as a potential target for cancer therapy. However, a specified agent might be required for each CAF phenotype to achieve efficacy. Therefore, future research on such targeted agents depending on CAF phenotype is needed.

In conclusion, our results indicated differential expression of CAF-related proteins according to breast cancer molecular subtype and stromal histologic type. We also observed high expression of CAF-related proteins in the tumor stroma of HER-2 and desmoplastic types.