Prognostic significance of PD-L1-positive cancer-associated fibroblasts in patients with triple-negative breast cancer

Background Cancer-associated fibroblasts (CAFs) are some of the most abundant components of the tumour microenvironment. A recent study suggested that in some cancers, CAFs express programmed death ligand 1 (PD-L1), which can act as a prognostic marker. The aim of this study was to investigate the clinicopathological significance of CAF PD-L1 expression in patients with triple-negative breast cancer (TNBC) and to identify the most suitable primary antibody for immunostaining for CAF PD-L1. Methods Immunohistochemical staining (primary antibodies of 73–10, SP142, and E1L3N) and tissue microarrays were used to analyse the expression profiles of PD-L1 in CAF in 61 patients with TNBC who underwent surgery. Overall survival (OS) was compared based on CAF PD-L1 expression, and the risk factors for OS were analysed. The relationship between clinicopathological parameters and survival was also examined. Results Thirty-four (55.7%) patients were positive for CAF PD-L1 (73–10) expression. Compared with CAF PD-L1 negativity, there was a significant correlation between CAF PD-L1 positivity and better OS (p = 0.029). CAF PD-L1 expression, evaluated using SP-142 or E1L3N, did not correlate with OS. CAF PD-L1-positivity (73–10) correlated significantly with better prognosis in multivariate analyses (hazard ratio: 0.198; 95% confidence interval: 0.044–0.891; p = 0.035). Conclusions CAF PD-L1 expression is a novel marker for a better prognosis of patients with TNBC, and the 73–10 assay may be suitable for immunostaining CAF PD-L1.


Background
Breast cancers usually influence the immune system; however, the level of immune infiltration reportedly differs among breast cancer molecular subtypes [1]. Triplenegative breast cancer (TNBC), characterised by the lack of oestrogen and progesterone receptors and human epidermal growth factor receptor 2 (HER2) expression, is a more aggressive breast cancer than other subtypes that correlate with poor prognosis [2,3]. Most patients with TNBC are categorised as having a robust tumour lymphocytic infiltrate compared to other subtypes. Therefore, TNBC is considered the most immunogenic subtype [4]. Recently, the expression of programmed death ligand 1 (PD-L1) in tumour cells and/or immune cells in breast cancer tissues has received much attention because of its effectiveness in anti-PD-L1/PD1 targeted therapy [5,6]. It has been reported that PD-L1 expression in tumour cells was associated with higher histological grade, hormone receptor-negative phenotype, poor prognostic outcome, and lymph node status in breast cancer [7][8][9]. Moreover, previous studies revealed that PD-L1 expression in both tumour and immune cells was the highest in TNBC compared to other subtypes [10][11][12].
Cancer-associated fibroblasts (CAFs) are considered to produce a pro-tumourigenic microenvironment and are some of its dominant components [13]. CAFs have been known to play important roles in cancer growth, invasion, metastasis, and therapeutic resistance through the secretion of various soluble factors, including chemokines, growth factors, and exosomes in some types of carcinomas, including breast cancer [4,[13][14][15][16][17]. Although PD-L1 expression has been demonstrated in subsets of tumourinfiltrating lymphocytes and macrophages, its expression in CAFs has not yet been analysed in detail. Recently, PD-L1 expression in CAFs of non-small cell lung carcinoma tissues has been demonstrated to correlate with good patient prognosis [18]. However, PD-L1 expression in CAFs has never been analysed in breast cancer tissues, and its prognostic significance remains to be clarified. Thus, the aim of this study was to determine PD-L1 expression in CAFs of TNBC tissues and its prognostic significance in patients with TNBC.

Patient selection
We enrolled 165 consecutive patients with TNBC who underwent surgical resection from January 2006 to December 2018 at the Department of Surgery of the Kansai Medical University Hospital. Patients who received neoadjuvant chemotherapy were excluded from the study because this chemotherapy may influence PD-L1 expression and may also have an impact on the patient prognosis. Patients who were diagnosed with invasive carcinoma of no special type according to the recent World Health Organization Classification of Breast Tumours [19] were selected. Patients with special types of invasive carcinoma were excluded from the study, because each special type of carcinoma has unique clinicopathological features; hence, 61 patients with TNBC comprised this study cohort. The present cohort was fundamentally the same as that of our previous study [20]. In the previous study, we analysed the relationship between adipophilin expression, a lipid dropletassociated protein, and the clinicopathological features of TNBC patients. The content of this study did not overlap with that of our previous one [20].
This study was conducted in accordance with the Declaration of Helsinki, and the study protocol was approved by the Institutional Review Board of the Kansai Medical University Hospital (Approval #2019041).

Histopathological analysis
Histopathological features were independently evaluated by more than two experienced pathologists. We used the TNM Classification of Malignant Tumours, Eighth Edition. Histopathological grading was based on the Nottingham histological grade [21]. The Ki-67 labelling index (LI) was considered high when ≥40% of the neoplastic cells were labelled [22]. Stromal tumourinfiltrating lymphocytes (TILs) were identified using haematoxylin and eosin staining, and were considered lymphocyte-predominant breast cancer (LPBC) at ≥60% and non-LPBC at less than 60%, according to the TIL Working Group recommendation [23,24].

Tissue microarray
The most morphologically representative carcinoma regions were selected on haematoxylin and eosin-stained slides, and three tissue cores (2 mm in diameter) were punched out from the paraffin-embedded blocks for each patient. These tissue cores were arrayed in a paraffin blocks.
Spindle-shaped non-neoplastic cells in tumour stroma were morphologically recognized as CAFs, and membranous and/or cytoplasmic expression of PD-L1 in these cells was considered positive. PD-L1 expression scores of CAFs were determined based on the staining intensity and were classified into three levels (0, negative; + 1, weak; + 2, strong). As previously reported, CAF PD-L1 positivity was defined as the presence of CAFs with staining intensities of + 1 and + 2 in more than 1% of a section and positive immunoreactivity of ≥1 from the same patient [18]. In addition, PD-L1 expression in stromal TILs was defined as expression in more than 5% of TILs (TIL PD-L1-positive) [25,26].

Statistical analyses
SPSS Statistics 25.0 (IBM, Armonk, NY, USA) was used to perform the statistical analyses. Correlations between two groups were calculated using Fisher's exact test for categorical variables and the Mann-Whitney U test for continuous variables. The rates of relapse-free survival (RFS) and overall survival (OS) were evaluated using Kaplan-Meier method. The Cox proportional hazards model was used to examine the relationship between clinicopathological parameters and survival. A multivariate analysis was performed using a step-wise method. A p-value of < 0.05 was considered to be significant.

Clinicopathological features
This study comprised 61 female patients whose clinical characteristics are summarised in Table 1

Double immunofluorescence staining
Immunofluorescence staining revealed spindle-shaped cells around the tumour cells, which co-expressed α-SMA and PD-L1 (Fig. 4a-c). These cells were recognised as CAFs.
Correlation between CAF PD-L1 expression and clinicopathological factors Table 2 shows the correlation between CAF PD-L1 expression (73-10) and clinicopathological factors. CAF PD-L1 expression did not correlate with any clinical factors, including age, menopausal status, or presence of adjuvant chemotherapy. Only TIL PD-L1 LPBC lymphocyte predominant breast cancer expression was significantly correlated with CAF PD-L1 expression (p < 0.001). CAF PD-L1 expression according to the SP-142 and E1L3N assays was also significantly associated with TIL PD-L1 expression (p < 0.001 for both), similar to the 73-10 assay (Tables 3, 4).

Prognostic significance of CAF PD-L1 expression (73-10)
Based on the univariate analysis, presence of lymph node metastasis and no adjuvant chemotherapy correlated with a poor OS (p = 0.001 and 0.016, respectively), and CAF PD-L1 expression significantly correlated with a better OS (p = 0.040) (

Discussion
Recent studies on various types of carcinomas have highlighted the important roles of the tumour microenvironment components, including TILs, in cancer growth, invasion, metastasis, and therapeutic resistance [13]. CAFs are some of the dominant components of the tumour microenvironment [13]. CAFs have also been increasing interests in several types of carcinomas, such as head and neck, lung, and rectal carcinomas, because they have been considered to have essential functions in cancer growth and prognosis [18,[27][28][29]. Recently, PD-L1 expression in CAFs reportedly demonstrated a significantly better prognostic value in patients with non-small cell lung carcinoma [18]. However, the significance of CAFs in breast cancer has not received enough attention.
In the present study, we examined the clinicopathological significance of CAF PD-L1 expression in patients with TNBC and had two main findings: (1) CAF PD-L1  (red line) and -negative (blue line) patients evaluated using the SP142 assay. c OS curves in CAF PD-L1-positive (red line) and -negative (blue line) patients evaluated using the E1L3N assay expression was an independent prognostic factor in patients with TNBC and (2) the 73-10 PD-L1 assay would be more suitable for evaluating CAF PD-L1 expression in TNBC, compared to the other two assays. The histopathological grade, lymph node status, tumour size, and Ki-67 LI have been identified as the prognostic factors in patients with TNBC [30]. Furthermore, a recent study revealed that upregulation of PD-L1 was correlated with a good prognosis in patients with TNBC [12]. In our study, a multivariate analysis of OS showed that lymph node status and adjuvant chemotherapy were significant prognostic factors, but histological grade and Ki-67 LI were not. Notably, CAF PD-L1 expression was also established as a significantly better prognostic factor (p = 0.035). These results indicate that CAF PD-L1 expression is a novel and useful prognostic factor for OS in patients with TNBC.
It is well known that high PD-L1 expression in the tumour microenvironment is a poor prognostic factor [31]. Stromal PD-L1 was also reported to inhibit the immune responses of CD8-positive T lymphocytes in colorectal cancer [32]. In addition, the expression of PD-L1 on CD8-positive T-cells was a poor prognostic factor in patients with TNBC [31]. These findings suggest that high PD-L1 expression on CAFs could also suppress anti-tumour immune responses through the exhaustion of PD-1-positive lymphocytes. However, in our study, high CAF PD-L1 expression significantly correlated with better prognosis. Moreover, a recent study demonstrated that patients with PD-L1 expression on CAFs had a significantly better prognosis in non-small cell lung cancer, similar to the results of the present cohort analysis. It was also shown that interferon gamma (IFN-γ) activated PD-L1 expression on CAFs [18]. PD-L1 on CAFs was upregulated through interaction with IFN-γ, hence releasing activated lymphocytes; furthermore, PD-L1 expression on CAFs indicated abundant infiltration of TILs in the tumour microenvironment [18]. Nevertheless, there was no significant association between CAF PD-L1 expression and TILs in non-small cell lung cancer [18] and TNBC assessed with the 73-10 assay in our study (stromal TILs were significantly associated with CAF PD-L1 positivity when assessed using SP142 and E1L3N assays in the present cohort). Thus, immune cells other than TILs such as macrophages might have played a role in this discrepancy for specific sources of IFN-γ. Further studies are hence needed to clarify the underlying molecular mechanism.
Several immunohistochemical assays for PD-L1 have been independently developed for companion diagnostics to determine the indication for anti-PD-L1/PD1 targeted therapy. Interestingly, only CAF PD-L1 expression determined using the 73-10 assay was significantly correlated with better OS (results of the SP-142 and E1L3N assays did not) in our cohort. The 73-10 antibody is known to bind to the region of the C-terminal cytoplasmic domain of PD-L1 [33]. Although SP-142 and E1L3N are also known to bind to the C-terminal cytoplasmic domain of PD-L1, it has been reported that they have slightly different binding sites [34,35]. SP142 binds to an epitope in the cytoplasmic domain at the extreme Cterminus, and several mutations lead to lack of immunostaining for SP142 [35]. Although the specific binding site of 73-10 has not been reported, it is speculated that 73-10 binds to a different intracytoplasmic domain of PD-L1, from those of SP142 and E1L3N, resulting in the difference in staining properties in non-small cell lung cancer [33,35,36]. CAF PD-L1 expression was detected the most by the 73-10 assay, compared to the SP142 and E1L3N assays; this might be a reflection of the difference of binding sites among primary antibodies. The 73-10 antibody would therefore be more suitable for studying PD-L1 expression in CAFs of patients with TNBC. Furthermore, in this study, we demonstrated that spindle cells around the tumour cells co-expressed α-SMA and PD-L1 with immunofluorescence staining, and these cells were considered as PD-L1-expressing CAFs, because α-SMA is one of the most common markers of CAFs in breast cancer [14]. Nevertheless, it is important to note that some limitations were present in our study. This was a retrospective study with a small sample size that could have led to selection bias. Because tissue microarrays were used to determine CAF PD-L1 expression, cancer tissue may have shown heterogeneous expression, despite that we have selected the morphologically most representative regions of the cancer tissue. Finally, this study was focussed on the expression of CAF PD-L1 in TNBC. PD-L1 expression in stromal cells differs among molecular subtypes of breast cancer [37], hence CAF PD-L1 expression might also be different in luminal and HER2 subtypes. Further analyses are needed to clarify the prognostic value of CAF PD-L1 expression in patients with breast cancer subtypes other than the TNBC.

Conclusions
This study demonstrates that CAF PD-L1 expression is an independent better prognostic factor in patients with TNBC, which could have implications in diagnosis, disease management, and the development of targeted therapeutics. Nevertheless, additional studies are needed to elucidate the molecular mechanisms involved in CAF PD-L1 expression in TNBC and to develop therapeutic interventions for patients with CAF PD-L1-positive TNBC. Moreover, the 73-10 assay may be the most suitable for immunostaining of CAF PD-L1 in TNBC.