Immune cell composition and immunological proles of the breast cancer microenvironment represented by histologically assessed tumor-inltrating lymphocytes and PD-L1 expression: A systematic analysis

Background. A better understanding of tumor immunology can facilitate the development of new treatment strategies for various malignancies. Histologically assessed tumor-inltrating lymphocytes (hTILs) and programmed cell death 1 ligand 1 (hPD-L1) have been established as prognostic or predictive biomarkers in certain subsets of breast cancer. However, the complexity of multiple types of immune cells is not fully understood. In this study, the immune cell fractions in breast cancer tissue and blood were evaluated to analyze their association with hTILs and hPD-L1. Methods. In total, 45 tumor and 18 blood samples were collected from breast cancer patients. The total leukocyte counts, proportions of 11 types of immune cells in the samples, and PD-L1 expression in each fraction were evaluated using multicolor ow cytometry for both the tumor and blood samples. The hTILs and hPD-L1 were evaluated with hematoxylin and eosin staining and immunohistochemistry respectively. Results. The immune cell composition of the blood showed a partial correlation with that of the tumor tissue; however, no signicant association was found between the blood immune cell compositions and hTIL or hPD-L1 expression. A higher hTIL was associated with increased leukocyte inltration as well as a higher proportion of CD4 + and CD8 + T cells and lower proportion of natural killer cells and natural killer T cells. PD-L1 was highly expressed in the monocyte/macrophage (Mo/Mφ), nonclassical monocyte (CD16 + Mo), myeloid-derived suppressor cell (MDSC), dendritic cell (DC), and myeloid dendritic cell (mDC) fractions in the tumor tissues. hPD-L1 positivity was associated with increased leukocyte inltration in the tumor tissues and PD-L1 expression in Mo/Mφ, CD16 + Mo, MDSC, DC, and mDC fractions. Conclusion. There was a partial correlation in the composition of immune cells at the tumor site and that in the peripheral blood. A high proportion of hTILs reects not only higher immune cell inltration but also


Background
Breast cancer is the most common malignancy in women worldwide, and it develops at a relatively younger age. Although there has been progress in multimodal treatment comprising surgery, systemic therapy, and radiation therapy, it is still di cult to cure advanced and recurrent disease [1]. The recent success in the clinical use of immune checkpoint inhibitors for multiple cancers has attracted attention in tumor immunology and emerging evidence has suggested that a better understanding of tumor immunology may lead to the development of new treatment strategies or the effective use of existing therapies [2].
Histologically assessed tumor-in ltrating lymphocytes (hTILs) can provide important prognostic information for diverse solid tumor types. They may also be of value in predicting the response to treatment [3]. In breast cancer, although with some controversy, hTIL has been detected more frequently in the triple-negative subtype or human epidermal growth factor receptor-2 (HER-2)-positive subtype than in the luminal subtype [3] and has also been correlated with clinicopathological factors other than subtypes [4,5]. With regard to prognosis, hTIL has been shown to be associated with a better prognosis for disease-free survival and overall survival in the triple-negative and HER-2 positive subtypes [4,6]. It has also been suggested that hTIL is a strong predictor of response to neoadjuvant chemotherapy in all molecular subtypes [7]. The programmed cell death 1 ligand 1 (PD-L1) was shown to be expressed in immune cells, including the T cells, B cells, macrophages, monocytes, and dendritic cells, as well as in tumor cells. PD-L1 can bind to the programmed cell death protein 1 (PD-1) expressed on activated T cells. The interactions with PD-L1 enable PD-1 signaling to counter the activation of T cells during the effector phase of the immune response [8,9]. Histologically assessed PD-L1 expression (hPD-L1s) have been found to be more frequent in HER-2 subtypes and triple-negative subtypes than in luminal subtypes, and have also been correlated with poor prognoses, higher histological grades, and lymphatic vessel invasions [10]. Furthermore, hPD-L1 has been established clinically, as a predictor of the e cacy of atezolizumab in triple-negative advanced breast cancer [11]. Recently, different types of immune cell subsets have been evaluated primarily by immunohistochemistry (IHC) and shown to be associated with various clinicopathological factors or prognoses. These are summarized in Table S1, and suggest their clinical signi cance in breast cancer [12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28].
Thus, hTIL and hPD-L1 have been established as biomarkers for breast cancer, and recent studies have shifted their attention to the various immune cell subsets that make up hTILs. However, the complexity of the multiple types of immune cells in TIL-or PD-L1 expressing cells has not yet been fully understood, because of the technical di culties in detecting multiple types of immune cells in the tissues using conventional IHC. To address this issue, we rst used multicolor ow cytometry (FCM) to assess the multiple immune cell fractions in breast cancer tissue and blood and then we analyzed the association between hTIL and hPD-L1. By performing a systematic analysis of the immune cell composition, we aimed to show that speci c immunological pro les of the breast cancer microenvironment were represented histologically by hTIL and hPD-L1.

Patients
Forty-seven tumor samples from the primary site and 19 matched blood samples were obtained from the breast cancer patients regardless of the clinicopathological factors or treatment histories, except for the patients with distant metastases or a clinical complete responses to neoadjuvant chemotherapy. None of the patients in this study received irradiation or endocrine therapy prior to their surgeries. Since the amendment of the study protocol in July 2016, blood sample collection commenced, thus there were cases with no blood samples. Clinicopathological data including menopausal statuses, histories of preoperative chemotherapy, histological types, invasive tumor sizes, lymph node statuses, lymphatic involvement, vascular involvement, histological grades, estrogen receptor (ER) statuses, progesterone receptor (PgR) statuses, HER-2 statuses, and Ki67 labeling indexes were collected by reviewing the patients' case records. The histological grades were evaluated according to the method described by Robbins et al. [29]. The ER, PgR, and HER-2 statuses were evaluated using IHC staining. The cut-off value for ER and PgR positivity was set at ≥ 10% [30]. The HER-2 status was determined according to the ASCO CAP guideline 2013 [31]. For the reasons mentioned later, 45 tumor samples and 18 blood samples were included in the analysis. The clinicopathological characteristics of the whole cohort and the cohort with blood samples are shown in Tables S2 and, S3 respectively. TIL/PBMC preparation Tumor and blood samples were collected simultaneously during surgery and processed. Tumor samples were mechanically dissociated on ice with 10% FBS-PBS, ltered using a 70-micron strainer, and washed with 10% FBS-PBS. The whole blood was collected using a collection tube containing EDTA-2Na. For both the tumor and the blood samples, the mononuclear cell components were separated using densitygradient centrifugation with a Ficoll-Paque PLUS (Cytiva Inc., Tokyo, Japan). In brief, the samples diluted with PBS were layered on Ficoll-Paque and centrifuged at 1,100 × g for 20 min at 18-20℃. Mononuclear cell components were collected, washed twice with PBS, and stored in liquid nitrogen in CELLBANKER I (Takara Bio Inc. Shiga, Japan).

Flow cytometry analysis
The cryopreserved TILs and PBMCs were thawed and washed twice with PBS containing 10% FBS and PBS. The cell suspensions were then processed for surface staining with an antibody cocktail (Table S4) for 20 min at 4°C. The cells were then washed with PBS containing 2% FBS and resuspended in CellFix (BD Biosciences). The stained cells were detected using an LSR II Fortessa with FACS Diva software (BD Biosciences). All the analyses were performed using FlowJo software (BD Biosciences, Franklin Lakes, NJ, USA). The gating strategy is shown in Figure S1. The immune cell fractions are classi ed into the following fractions according to the de nitions shown in Table S5: Leucocytes, Total T cells (Total T), CD4 + T cells (CD4 + T), CD8 + T cells (CD8 + T), B cells (B), monocytes/macrophages (Mo/Mφ), nonclassical monocytes (CD16 + Mo), myeloid-derived suppressor cells (MDSC), dendritic cells (DCs), myeloid dendritic cells (mDCs), natural killer cells (NK), minor NK cells, and natural killer T cells (NKT). Two cases with a low number of living cells (count < 1000) in the CFM analysis of the tumor tissue were excluded. A blood sample that was also identical to one of the two cases, was also excluded. As mentioned previously, 45 tumor samples and 18 blood samples were included in the analysis. The leucocyte density based on the weight of the tissue fragment and the number of viable CD45 + cells was determined using the same method as a previous study [32]. For the immune cell fraction, we determined both the percentages of each fraction in the leukocytes (% in leukocytes) and the densities based on the weights of the tissue fragments (count/g).

Histological evaluation of tumor immunity-related biomarkers
We evaluated the histological tumor immunity-related biomarkers, as described in a previous paper [33][34][35][36]. Brie y, the percentages of stromal TILs were evaluated using 4-µm sections from formalin-xed specimens stained with hematoxylin and eosin using a light microscope at ×200-400 magni cation. Stromal TILs were de ned as mononuclear cells localized in the stromal tissue of the breast cancer. The stromal TIL count was categorized, according to the International TILs Working Group guideline, into three grades: low (0-10%), intermediate (10-40%), or high (40-90%); and were scored from 0 to 2. The denominator used to determine the TIL grade was the stromal tissue area. PD-L1 expression was assessed using IHC with rabbit monoclonal anti-PD-L1 clone SP142 (prediluted; Spring Bioscience, USA) and the Ventana Benchmark ULTRA auto staining machine. Tumors with ≥ 1% immune cells with cytoplasmic and/or membrane PD-L1 staining were determined to be PD-L1 positive. The evaluation of the stromal TIL counts and PD-L1 expression was performed by two evaluators (CH and SK). In this study, the histologically assessed TIL and PD-L1 were described as hTIL and hPD-L1, respectively, to distinguish them from the CFM-assessed data. In one case, the hTIL and hPD-L1 could not be evaluated because of insu cient tumor tissue remaining. Thus, hTIL and hPD-L1 were analyzed in the remaining 44 patients.

Statistical analysis
All the statistical analyses and graph drawings were performed using GraphPad Prism ver. 9.1.0 software. All the analyses for the correlation between the two groups were performed using the Spearman's rank correlation coe cient. For the comparison of the paired samples of the tumor and blood, the Wilcoxon test was used. For the comparison of the two unpaired groups, the Mann-Whitney U test was used. A Chi-square test was used to compare clinicopathological factors between the hTILs and hPD-L1. A Fisher's exact test was used when the Chi-square test indicated a signi cant P-value (P < 0.05) and there were cells with a sample size of ve or less.

Results
Distribution of leukocyte, and immune cell fraction determined by FCM.
Leucocyte densities based on the weights of the tissue fragments and the number of viable CD45 + cells was determined in the entire cohort (n = 45). The mean density and interquartile range of the tumorin ltrating leukocytes were 226 × 10 3 cells/g and 80× 10 3 to 574 × 10 3 cells/g, respectively (Fig. 1a). For the immune cell fraction in the leukocytes of the tumor tissues (TILs), the main population consisted of CD8 + T, CD4 + T, Mo/Mφ, and B ( Fig. 1b and d). There was a similar trend in the leukocyte composition in blood (PBMC), showing that the main population consisted of CD8 + T and CD4 + T cells ( Fig. 1c and d).
Immune cell composition of blood is partly correlated with that of tumor tissue but the abundance ratio of each fraction itself is different between the two.
For cases with matched samples of blood and tumor tissue, a correlation analysis was performed for the percentages of each of the immune cell fractions (Fig. 2a-l). The proportions of DC, mDC, NK, and minor NK in tumor tissue showed a positive correlation with those of blood ( Fig. 2h-k). The percentages of each immune cell fraction of tumor tissue (TIL) and that of blood (PBMC) were compared, the same cases ( Fig. 3a-l). The proportions of DC, mDC, and minor NK were signi cantly higher in the tumor tissues compared with the proportions in blood (Fig. 3h, i, and k). On the other hand, the proportions of CD4 + T and NK cells were signi cantly lower in tumor tissues compared with the proportions in blood ( Fig. 3b and j). Although we investigated the relationship between the immune cell composition in blood and hTIL or hPD-L1, there were no signi cant associations found between them ( Figure S2 and S3). Similarly, no signi cant associations were found between hPD-L1 and the FCM assessed PD-L1 positive ratios in the immune cell fraction of blood ( Figure S4).
Histologically assessed TIL (hTIL) was associated with not only the degree of leukocyte in ltration in tumor tissue but also its composition.
For cases with tumor tissue samples, a correlation analysis was performed for the hTIL scores and the leukocyte densities (count/g) in the tumor tissues, showing a strong positive correlation with each other (Fig. 4a). Furthermore, the hTIL scores showed a positive correlation with the densities (count/g) of almost all of the immune cell fractions in the tumor tissues, except for NK ( Figure S5). A Correlation analysis was also performed for the hTIL scores and percentages of each immune cell fraction in the tumor tissues. We found that there were positive correlations between the hTIL scores and the percentages of total T, CD4 + T, and CD8 + T (Fig. 4b-d), but negative correlations between the hTIL scores and NK and NKT (Fig. 4k, m), showing that hTIL was associated with not only the degree of leukocyte in ltration in tumor tissue but also its composition.
Histologically assessed PD-L1 (hPD-L1) expressions are associated with leukocyte in ltration in tumor tissue and re ect PD-L1 expression in certain immune cell fraction.
The percentages of the PD-L1 positive cells in each immune cell fraction was determined for the tumor tissues (TILs) and blood (PBMC) using FCM. Percentages of the PD-L1 positive cells of each immune cell fraction in the tumor tissues are shown in Fig. 5a, indicating that PD-L1 is preferentially expressed in non-B-cell antigen-presenting cell fractions such as Mo/Mφ, CD16 + Mo, MDSC, DC, and mDC. For cases with matched samples of blood and tumor tissue, the percentages of the PD-L1 positive cells in each immune cell fraction were compared between the tumor tissues and blood. The PD-L1 positive ratios were signi cantly higher in the tumor tissues than in blood for all lineages except for the lymphoid fractions ( Fig. 5b-l). Next, we investigated the relationship between hPD-L1 and the immunological pro les of the tumor tissues. Comparisons of the leukocyte densities (count/g) in the tumor tissues between the hPD-L1-negative and -positive cases showed that hPD-L1 positivity was associated with increased leukocyte in ltrations in the tumor tissues (Fig. 6a). Similarly, hPD-L1 showed a positive correlation with the densities (count/g) of almost all the immune cell fractions in the tumor tissues, except B and NK ( Figure  S6). With regard to the immune cell composition in the tumor tissues, although hPD-L1 positivity was associated with a lower percentage of NK and NKT, it was not correlated with the percentages of other lineages (Fig. 6b-m). For tumor tissue samples, PD-L1 positive ratios in each immune cell fraction were compared between the hPD-L1 positive and-negative cases. We found that hPD-L1 positivity showed a positive association with the PD-L1 positive ratio in some of the immune cell fractions, including Mo/Mφ, CD16 + Mo, DC, and mDC, but not with the other lineages ( Fig. 7a-k). These data suggest that the histologically assessed PD-L1 expression re ects leukocyte in ltration in the tumor tissues and PD-L1 expression in certain immune cell fractions.

Discussion
In this study, we evaluated multiple immune cell fractions in both breast cancer tissues and matched blood samples, then observed that there was a partial correlation in the composition of immune cells at the tumor site and in the peripheral blood. In the comprehensive analysis of the association between each immune cell lineage fraction with hTIL and hPD-L1, we found for the rst time that clinically diagnosed breast cancers, especially in cases with higher hTIL or positive hPD-L1 were already in a state in which innate immune responses were poorly involved or had escaped, and acquired immune responses were active. We also found that Non-B-cell antigen-presenting cell fractions were involved primarily in the PD-L1 pathway in breast cancer microenvironments.
To the best of our knowledge, most of the analyses of the immune cell compositions of breast cancer tissues using a multicolor FCM had 10 colors or less [32,37], and that there were only two studies with more than 11 colors [38,39]. Although the reactivities of the labeled antibodies were not always the same, and a direct comparison was not possible, there was a similar distribution of the leukocyte in ltrations in the tumor tissue in our study and a previous study with a distribution median of 218 CD45 + TIL/mg of tumor tissue (interquartile range: 85-445 CD45 + TIL/mg) [32]. Although there are very few reports of systematic examinations of leukocyte compositions in breast cancer tissue, some studies have reported a ratio of total T being 86% (mean) [32] or 75% (median) [38] of the leukocytes (CD45 + cells) in breast cancer tissues, suggesting that T cells accounted for the majority of TILs [40]. In our study, the proportion of total T cells in the leukocytes in the tumors was 57.3% (mean), which was slightly lower than the proportion in previous reports probably due to the difference in the antibody used and the gating strategy. But consistent with the fact that T cells account for the majority of leukocytes in the tumor tissues. In three previous studies, the proportions of CD19 + B cells in CD45 + TIL were found to be 8% (mean), 4.58% (median), and approximately 10% (mean), respectively [32,37,38], which was similar to the results of our study. With regard to other lineages, the ndings of a previous study showed that CD14 + /CD40 + /CD163 + M2 macrophages were 0.06% (median), CD11b + /CD15 + /HLA-DR-MDSCs were 1.19% (median), and CD56 + NK was 2.33% (median) [38]. However, the number of studies were small, and the de nitions of each lineage did not match those in our study, therefore valid comparisons could not be made. No comparable reports were found for the remaining lineages. This is the rst study to show an association between the immune cell composition of blood and that of breast cancer tissues. The immune cell composition of blood showed a partial correlation with the tumor tissues (Fig. 2) and the percentages of the immune cell fractions showed some differences between the tumor tissues and blood (Fig. 3). While this suggested that the composition of tumor-in ltrating immune cells may be estimated using blood samples, there were signi cant differences between them. Each also has a different function or clinical signi cance.
Although there was a signi cant difference between the subtypes, hTIL was shown to correlate with some clinicopathological factors including subtypes [3][4][5], prognoses, and responses to chemotherapy [4,6,7] in breast cancer. In our study, higher hTIL scores were associated with high-grade tumors, ER-negativity, higher Ki67 positive ratios, and hPD-L1 positivity (Table S6). Many studies have also reported that ERpositive breast cancer is the least immune-in ltrated subtype, which was consistent with our results [2,5].
However, there are some controversies with regard to other clinicopathological factors, and results differ from study to study [2,4,5,41]. To date, no studies have systematically assessed the relationship between hTIL and the immune cell fraction using FCM. In this study, we demonstrated that hTIL was associated with not only the degree of leukocyte in ltration in the tumor tissues but also the composition (Fig. 4). We found that while there were positive correlations between the hTIL scores and percentages of total T, CD4 + T, and CD8 + T (Fig. 4b-d), there were negative correlations between the hTIL scores and NK and NKT (Fig. 4k, m). Therefore, we speculate that a higher hTIL not only re ects the amount of immune cell in ltration, but also re ects the state in which acquired immunity is activated, relative to innate immunity in clinically diagnosed breast cancers. Furthermore, they may suggest that there is still room for the development of cancer immunotherapy to promote acquired immune responses in breast cancers other than triple negative breast cancer.
As mentioned previously, PD-L1 plays a signi cant role in the immune tolerance mechanisms that suppress T-cell activation [8,9], and its expression is suggested to re ect ongoing (or active) immune responses or in addition to immunosuppression via the PD-1/PD-L1 pathway [39]. The hPD-L1 was shown to correlate with some clinicopathological factors, including subtypes [10]. It is also a clinically approved predictive marker for atezolizumab in triple-negative advanced breast cancer [11]. In the present study, hPD-L1 positivity was associated with ER-negative diseases and higher hTIL scores but no signi cant association with the other factors, probably due to the small cohort size (Table S7). Although [8,9], there is no consensus as to which immune cell fraction is responsible for the substantial function of the PD-L1 pathway in breast cancer. The ndings of only one report that evaluated the PD-L1 expression in CD4 + T and CD8 + T, B showed that the overall proportion of the PD-L1 positive TILs was very low and could only be detected in a small number of tumors [39]. In the present study, we found that a substantial proportion of PD-L1 positive immune cells were non-B-cell antigen-presenting cell fractions such as Mo/Mφ, CD16 + Mo, MDSC, DC, and mDC fractions (Fig. 5a), and that the PD-L1 positive ratios were signi cantly higher in tumor tissues than in blood (Fig. 5b-l), suggesting that these fractions were involved primarily in the PD-L1 pathway in breast cancer tissue. In addition, we found that hPD-L1 positive tumors exhibited increased leukocyte in ltration in tumor tissues (Fig. 6a), and that the hPD-L1 re ected PD-L1 expression in Mo/Mφ, CD16 + Mo, DC, and mDCs ( Fig. 7a-k). These results suggested that hPD-L1 expression indicates the activation status of the immune tolerance mechanism that occurs in non-B-cell antigen-presenting cells in response to an increased immune cell in ltration, mainly effector cells which secrets IFN-gamma to induce PD-L1 expression on various cells, into the breast cancer microenvironment.

PD-L1 expression in multiple types of immune cells or tumor cells has been reported
The FCM analysis ndings will be useful in the exploration of new immune-related factors in breast cancer. Brie y, by evaluating the expression of candidate proteins related to tumor immunity by IHC and analyzing them together with these data, function of candidate proteins may be veri ed. Currently, we are focusing on some candidate proteins as immuno-regulatory factors in breast cancer and further analyzing FCM data of this study to validate their immunological functions.
This study had several limitations. A relatively small number of patients were enrolled in this study. In addition, in a pilot study, it was found empirically that the number of cells required for FCM was not su cient in cases of ER-positive breast cancer, especially in cases with lower Ki67s. In addition, cases of small tumor sizes and post-NAC with pathological CRs were excluded due to technical problems in the collection of the tumor tissues. Therefore, it should be noted that there was an inevitable bias in the enrollment of the cases; it differed from the general breast cancer cohort in terms of larger invasive tumor sizes, more ER-negative cases, and higher Ki67 cases (Table S2). Although, as mentioned above, the signi cance of the TIL is suggested to be different between subtypes, a subgroup analysis could not be performed because of the small sample size. It is recommended that in future studies, more samples be collected and more detailed analyses be performed.

Conclusions
In this study, there was a partial correlation in the composition of immune cells at the tumor site and that in the peripheral blood. A comprehensive analysis of the immune cell fractions revealed the immunological pro les of breast cancer tissue represented by hTIL or hPD-L1. Our ndings indicated that hTIL not only re ected the amount of immune cell in ltration, but also the state in which acquired immunity is activated relative to innate immunity. Non-B-cell antigen-presenting cell fractions such as Mo/Mφ, CD16 + Mo, MDSC, DC, and mDCs were involved primarily in the PD-L1 pathway in breast cancer microenvironments. In addition, hPD-L1 re ects PD-L1 expression in these immune cell fractions. Our