Background
Ductal carcinoma in situ (DCIS) is an early pathologic stage of breast cancer characterized by proliferation of tumor cells within the ductal-lobular unit. Introduction of screening mammography has resulted in increased detection of DCIS which currently accounts for 20~25% of newly diagnosed breast cancers in the USA [
1]. DCIS is a non-obligatory precursor of invasive breast cancer that progresses to invasive cancer over 10–15 years in 14–53% [
2]. The mechanism by which DCIS progresses to invasive carcinoma is not well understood, but it is thought to be a complex process driven by tumor cells (through genetic aberrations or altered expression of genes critical for invasion) and tumor microenvironment including myoepithelial cells, stromal fibroblasts, and immune infiltrates [
3].
The immune system can eliminate tumor cells or control tumor growth by immune surveillance, but interaction between tumor cells and the immune system is known to play a crucial role in tumor progression [
4]. Key players of the immune system include myeloid cells, lymphocytes, cytokines, and chemokines, of which tumor-infiltrating lymphocytes (TILs) are thought to represent tumor immunogenicity, and their composition is associated with the direction of an immune response [
5,
6]. Studies to date generally agree that CD8+ cytotoxic T lymphocytes (CTLs) and CD4+ Th1 cells are involved in effective anti-tumor immunity while FOXP3+ regulatory T cells are associated with suppression of anti-tumor immunity [
7]. Besides TILs, immune checkpoint molecules are also involved in the regulation of anti-tumor responses. Especially, programmed death-ligand 1 (PD-L1), also known as B7-H1 or CD274, is expressed on tumor cells and immune cells; it suppresses T cell migration, proliferation, and secretion of cytotoxic mediators, and it also restricts tumor cell killing through binding to programmed death-1 (PD-1) and B7.1 (CD80) [
8].
Most of the previous studies on TILs and their composition in breast cancer have focused on their predictive and prognostic significance in invasive breast cancer [
9‐
14]; their presence and significance in DCIS remain elusive. A few studies have reported that dense TILs in DCIS were associated with more aggressive clinical features and an increased risk to progression [
15,
16], and specific TIL subsets in DCIS have been linked to tumor recurrence [
17‐
19]. However, there is a lack of studies on TIL subset infiltration in DCIS and its comparison with that of invasive breast cancer. Moreover, only a few studies have evaluated PD-L1+ immune cells in DCIS with a limited number of cases [
16,
20‐
22]. Since interaction between tumor cells and its immune microenvironment play an essential role in tumor progression, evaluation of the differences in the infiltration of immune cell subsets between DCIS and invasive carcinoma may help us understand the mechanism for progression of DCIS.
Thus, in this study, we examined infiltrations of CD4+, CD8+, and FOXP3+ TILs and PD-L1+ immune cells in DCIS in relation to clinicopathologic features and compared them with those of invasive breast cancer to evaluate their changes during progression of DCIS.
Methods
Tissue samples
This retrospective study included a total of 671 cases comprising 231 cases of pure DCIS, 81 cases of DCIS with microinvasion (DCIS-M), and 359 cases of invasive breast carcinoma that had been diagnosed between 2003 and 2011 at Seoul National University Bundang Hospital. In a previous study, 377 cases of invasive breast carcinoma had been used for the evaluation of TIL subset infiltrations except for PD-L1 [
14]. After excluding 18 cases of invasive lobular carcinoma, the data of the remaining 359 invasive carcinomas were used for this study. Of the 359 invasive carcinoma cases, ninety cases which had a sufficient amount of DCIS component associated with invasive carcinoma were selected for comparative analysis of the invasive and DCIS components within the same tumor. Clinicopathologic information was obtained by reviewing medical records and hematoxylin and eosin (H&E)-stained sections. For pure DCIS and DCIS-M, clinicopathologic data including the patient age, tumor extent, nuclear grade, presence of comedo-type necrosis, architectural pattern, presence of microinvasive foci, estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2) status, Ki-67 proliferation index, and p53 overexpression were recorded. For invasive carcinomas, clinicopathologic variables had been summarized in a previous study [
14]. Baseline characteristics of pure DCIS and DCIS-M are summarized in Table
1. These two disease groups were significantly different in various clinicopathologic characteristics including DCIS extent, nuclear grade, ER, PR, and HER2 status.
Table 1Clinicopathologic characteristics of pure ductal carcinoma in situ (DCIS) and DCIS with microinvasion
Age (years) | Median (range) | 47 (25–88) | 48 (26–76) | 0.556 |
Extent of DCIS (cm) | Median (range) | 2.5 (0.4–12.2) | 4.2 (0.9–14.5) | < 0.001 |
Nuclear grade | Low | 16 (6.9) | 2 (2.5) | < 0.001 |
Intermediate | 129 (55.8) | 20 (24.7) | |
High | 86 (37.2) | 59 (72.8) | |
Comedo-type necrosis | Absent | 181 (78.4) | 30 (37.0) | < 0.001 |
Present | 50 (21.6) | 51 (63.0) | |
ER | Negative | 30 (13.0) | 42 (51.9) | < 0.001 |
Positive | 201 (87.0) | 39 (48.1) | |
PR | Negative | 44 (19.0) | 47 (58.0) | < 0.001 |
Positive | 187 (81.0) | 34 (42.0) | |
HER2 | Negative | 196 (84.8) | 43 (53.1) | < 0.001 |
Positive | 35 (15.2) | 38 (46.9) | |
Ki-67 index | < 10% | 171 (74.0) | 30 (37.0) | < 0.001 |
≥ 10% | 60 (26.0) | 51 (63.0) | |
P53 | Negative | 201 (87.0) | 51 (63.0) | < 0.001 |
Positive | 30 (13.0) | 30 (37.0) | |
Tissue microarrays
H&E-stained slides from formalin-fixed, paraffin-embedded tissue blocks were reviewed, and representative sections were selected in each case for construction of tissue microarrays (TMAs). For pure DCIS and DCIS-M, one to three tissue columns of 4-mm-diameter circles (depending on the extent of the tumor) were arranged in TMA using a trephine apparatus (Superbiochips Laboratories, Seoul, Korea). For invasive carcinoma, three sets of TMAs (2 mm in diameter) that had been constructed for the previous study [
14] were used. For comparative analysis between invasive and DCIS components of the same tumor, one tissue column of DCIS associated with invasive carcinoma (DCIS-INV) (4 mm in diameter) was selected and was made into TMAs.
Evaluation of basic biomarkers
The expression of the basic biomarkers including ER, PR, HER2, p53, and Ki-67 was evaluated from the surgical specimens at the time of diagnosis. As for those with missing data, immunohistochemical staining on representative tissue sections was carried out using the following antibodies: ER (clone SP1; 1:100 dilution; LabVision, Fremont, CA), PR (clone PgR 636; 1:70 dilution; Dako, Carpinteria, CA), HER2 (clone 4B5; ready to use; Ventana Medical Systems, Tuscon, AZ), p53 (clone D07; 1:600 dilution; Dako), and Ki-67 (clone MIB-1; 1:250 dilution; Dako).
ER and PR were regarded as positive if at least 1% of the tumor cells were stained. HER2 positivity was defined as an immunohistochemical score of 3+ or the presence of gene amplification on fluorescence/silver in situ hybridization. For p53, staining in 10% or more of the tumor cells was considered positive. High Ki-67 proliferation index was defined as staining in 10% or more of the tumor cells.
Immunohistochemistry for immune cells and counting
Immunohistochemical staining was performed with a BenchMark XT autostainer (Ventana Medical Systems) using an UltraView detection kit (Ventana Medical Systems) for CD4, CD8, FOXP3, and PD-L1. The following antibodies were used: CD4 (Clone SP35; ready to use; Dako), CD8 (Clone C8/144B; ready to use; Dako), FOXP3 (Clone 236A/E7; 1:100 dilution; Abcam, Cambridge, UK), and PD-L1 (clone E1L3N; 1:100 dilution; Cell Signaling, Danvers, MA).
For the evaluation of CD4+, CD8+, and FOXP3+ T cell infiltration, the number of each TIL subset was counted by two pathologists (MK and YRC) who were blinded to the clinicopathologic features of the tumors. From the TMA cores, three areas with the highest infiltration were chosen under the high-power field (400X), and the number of TIL subset was counted in both intra-tumoral and stromal compartments either manually or using a digital image analyzer, ScanScope CS system (Aperio, Vista, CA). Then, the average numbers of CD4+, CD8+, and FOXP3+ T cells from the selected areas were calculated. In cases of DCIS, the stromal compartment was defined as the area of the specialized stroma surrounding the involved ducts, or when it is not clear, as an area surrounding ducts within 2 high-power fields according to a proposal from the International Immuno-Oncology Biomarker Working Group [
23,
24]. As for PD-L1, PD-L1+ immune cells were considered to be present when at least 1% of the tumor stromal area was occupied by PD-L1+ immune cells, as previously described [
16].
Statistical analysis
Statistical analysis was performed using Statistical package, SPSS version 25.0 for Windows (IBM Corp., ARMONK, NY). CD4+, CD8+, and FOXP3+ TIL counts did not show normal distributions, and thus, non-parametric tests were used for statistical analyses. Spearman’s rank correlation tests were used to assess the associations among infiltrations of CD4+, CD8+, and FOXP3+ TILs and PD-L1+ immune cells. The difference in the infiltration of CD4+, CD8+, and FOXP3+ TILs was analyzed by the Mann-Whitney U test between two groups and by the Kruskal-Wallis test among three groups. For the detection of the predominant TIL subset in the same disease group or for comparison of TIL subset infiltration between invasive and in situ components in the same tumor, the Wilcoxon signed rank test was used. The presence or absence of PD-L1+ immune cells was analyzed by chi-square test between groups, and its comparison between invasive and in situ components in the same tumor was performed using the McNemar test. In pure DCIS, a receiver operating characteristic (ROC) curve analysis was performed to identify the cut-off values for CD4+, CD8+, and FOXP3+ TILs and ratios of TIL subsets (FOXP3+/CD8+ TIL, FOXP3+/CD4+ TIL, and CD4+/CD8+ TIL) that maximized the sum of sensitivity and specificity in predicting ipsilateral breast cancer recurrence using recurrence as a dichotomous outcome. Recurrence-free survivals were analyzed by drawing Kaplan-Meier curves, and differences were determined with the log-rank test. When comparing immune cell subset infiltration among pure DCIS, DCIS-M, and DCIS-INV, corrections for multiple testing were made by the Bonferroni method, and adjusted (adj.) p values were calculated. p values less than 0.05 were considered significant with all reported p values being two-sided.
Discussion
In this study, we have shown that the high infiltration of CD4+, CD8+, and FOXP3+ T cells and the presence of PD-L1+ immune cells were generally associated with aggressive features of DCIS including high nuclear grade, comedo-type necrosis, HR negativity, and high Ki-67 proliferation index. Increased TIL density in DCIS has been associated with high-risk features including large tumor size, high nuclear grade, comedo-type necrosis, ER negativity, and HER2 positivity in previous studies [
15,
16,
26]. As TIL subset infiltration is higher in TIL-rich DCIS, it is reasonable to assume that high TIL subset infiltration is associated with aggressive features of DCIS. As for the relationship between TIL subset infiltration and characteristics of DCIS, a few studies have reported a positive correlation between TIL subset infiltration such as CD4+, CD8+, FOXP3+, and CD20+ cells and nuclear grade of DCIS [
17,
27]. Similarly, the presence of PD-L1+ immune cells in DCIS has been reported to be associated with high TIL infiltration, younger patient age, ER negativity, and HER2 positivity in previous studies [
16,
20‐
22].
Dense TIL infiltration in DCIS has been linked to a high risk of progression [
15,
16]. As for the composition of TILs, low CD8+ T cells, low CD8+/FOX3+ T cell ratio, and low CD8+HLA-DR+ cells have been reported to be associated with ipsilateral recurrence [
17,
19]; high numbers of B cells were associated with shorter recurrence-free interval in DCIS [
18]. In this study, we showed that high infiltration of FOXP3+ TIL, high FOXP3+/CD8+ TIL ratio, and high FOXP3+/CD4+ TIL ratio were associated with decreased recurrence-free survival. Moreover, we also demonstrated that the presence of PD-L1+ immune cell was associated with poor recurrence-free survival. These results suggest that suppression of anti-tumor immunity by FOXP3+ TILs and PD-L1+ immune cells plays an important role during progression of DCIS. Thus, pure DCIS with high FOXP3+ TIL infiltration or PD-L1+ immune cells could be a target for active surveillance or aggressive treatment. However, the analyses about tumor recurrence have a limitation in that only a small number (
n = 6) of cases revealed ipsilateral breast recurrence. Moreover, clinicopathologic variables which were known to be associated with recurrence in pure DCIS were not associated with tumor recurrence in this study. Thus, the results should be interpreted cautiously and further large studies are warranted to confirm these findings.
The evaluation of difference in immune cell infiltration between DCIS and invasive carcinoma and also among pure DCIS, DCIS-M, and DCIS-INV may provide some insight into its role during DCIS progression. We showed that all TIL subset infiltration and the presence of PD-L1+ immune cells were higher in invasive carcinoma than in pure DCIS irrespective of the HR status as in previous studies which reported a gradual increase in the number of immune cells during progression of breast cancer [
28,
29]. Interestingly, in this study, HR-negative breast cancers revealed high CD4+ TIL infiltration in both in situ and invasive components of the same tumors with no statistical difference. Moreover, in HR-negative tumors, CD4+ TIL showed a gradual increase from pure DCIS to DCIS-M and DCIS-INV. These findings suggest that CD4+ T cells increase at an early stage of DCIS progression in HR-negative tumors and may play a crucial role during in situ to invasive transition in HR-negative tumors. However, we did not evaluate the CD4+ TILs by subset except for FOXP3+ regulatory T cells. The infiltration of FOXP3+ TILs did not differ between pure DCIS, DCIS-M, and DCIS-INV in the HR-negative group. It is well known that CD4+ TILs display a large degree of plasticity and the ability to differentiate into multiple sublineages in response to environmental cues [
30]. CD4+ Th1 cells, CD4+ CTLs, and follicular helper T cells exert potent anti-tumor activity, whereas regulatory T cells or, under certain circumstances, CD4+ Th2 cells and CD4+ Th17 cells show tumor-promoting activity [
30,
31]. Thus, in further studies, analyses of CD4+ TIL subsets would be crucial to find the overall effect of heavy CD4+ TIL infiltration around HR-negative DCIS.
In HR-positive breast cancers, despite the fact that FOXP3+ TIL was significantly higher in DCIS-INV than in pure DCIS, the other TIL subsets seemed to increase in number at a late stage of DCIS progression, or seemed unlikely to be associated with in situ to invasive transition. Growing evidence supports a role of host immune surveillance in influencing response to therapy and prognosis in HER2+ and triple-negative breast cancer, but not in HR-positive breast cancer which appears to be less immunogenic than HER2+ and triple-negative breast cancer [
7]. However, it has been reported that FOXP3+ TIL infiltration is strongly associated with adverse clinical outcome in HR-positive breast cancer [
14,
32,
33]. Furthermore, in the present study, we observed that infiltration of FOX3+ TIL was associated with tumor recurrence in HR-positive pure DCIS. Thus, treatment or prevention of tumor progression in HR-positive breast cancer would have to be focused on FOXP3+ regulatory T cells.
The mechanism by which immune cells influence tumor cell invasion in DCIS is unclear. Degradation of the basement membrane, a prerequisite for tumor invasion, has been attributed primarily to overproduction of proteolytic enzymes by the tumor or the surrounding stromal cells [
34]. However, cumulating data support a hypothesis that myoepithelial cells act as “natural tumor suppressors” and lose such property during tumor progression [
35]. Man et al. reported that leukocyte infiltration was increased at focal myoepithelial cell disruption sites in DCIS, suggesting that a localized death of myoepithelial cells and subsequent immune reactions are a trigger for myoepithelial cell layer disruptions, basement membrane degradation, and tumor invasion [
36]. Conversely, it can be postulated that when tumor cells with an increased invasive property invade the stroma, it activates the immune system leading to more immune cell infiltration around DCIS with invasion.
We showed that CD4+ and FOXP3+ TIL infiltration was significantly higher in DCIS-M and DCIS-INV compared to pure DCIS in the whole group. CD4+ and CD8+ TIL infiltration was significantly higher in DCIS-INV than in pure DCIS in the HR-negative group, and FOXP3+ TIL infiltration was significantly higher in DCIS-INV than in pure DCIS in the HR-positive group. Beguinot et al. reported that microinvasive carcinoma showed a significantly higher TIL density with more CD8+, CD4+, and CD38+ cell infiltration than pure DCIS. Toss et al. also showed that DCIS-INV showed denser TILs as opposed to pure DCIS [
15]. Although these authors did not show the difference in TIL infiltration according to HR status, their findings are consistent with the results of our study. It is known that in situ and invasive components of the same tumor exhibit similar patterns of genetic alterations [
37]. Thus, tumor cells in DCIS-INV may show higher immunogenicity than tumor cells in pure DCIS eliciting stronger immune responses compared to pure DCIS.
In our study, there were more CD4+ T cells than CD8+ T cells infiltrating pure DCIS regardless of the HR status. In invasive carcinoma, CD4+ and CD8+ T cell infiltration showed no difference in HR-positive tumors, whereas CD8+ T cells were predominant in HR-negative tumors. In line with our study, Thompson et al. found slightly more CD4+ T cells than CD8+ T cells in DCIS [
20], and Sheu et al. showed that CD8+ T cells significantly increased with stage progression of invasive breast cancer [
38]. On the contrary, Gil Del Alcazar et al. [
39] reported a decrease in CD8+ signatures in invasive breast cancer and fewer activated GZMB+CD8+ T cells in invasive breast cancer compared to DCIS. However, these studies were conducted using a small series, and we observed a greater infiltration of CD8+ T cells than CD4+ T cells in a minority of DCIS, and vice versa in some cases of invasive carcinoma. Thus, further large-series studies investigating the composition of TIL subsets in DCIS and their change during tumor progression are required to understand the role of individual TIL subsets during tumor progression.
The current study includes a relatively large number of cases that can provide a general idea on changes in immune cell infiltration during in situ to invasive transition. However, this study has some limitations. First, even though the International Immuno-Oncology Biomarker Working Group recently published a proposal for evaluation of TILs on hematoxylin- and eosin-stained sections in DCIS [
23,
24], the scoring system for TIL subsets by immunohistochemistry in DCIS has not been optimized yet. It is recommended that only stromal TILs be evaluated as mean values in DCIS [
23,
24]. However, we evaluated both intra-tumoral and stromal TIL subsets in hot spots in DCIS to compare with our previous study data of TIL subset infiltration in invasive carcinoma even though most of the TIL subsets were found in the stromal compartment. Second, the evaluation of immune cell subsets was confined to CD4+, CD8+, and FOXP3+ TILs and PD-L1+ immune cells, and their subtypes or activation status was not assessed. In addition, although TIL subset and PD-L1+ immune cell infiltration in DCIS may be heterogeneous from one area to another, the cells were counted using the TMA platform in order to evaluate a large number of samples which may have resulted in selection bias.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.