PD-1⁺ immune cell infiltration inversely correlates with survival of operable breast cancer patients

Abstract

The programmed death-1 (PD-1) molecule is mainly expressed on functionally “exhausted” CD8⁺ T cells, dampening the host antitumor immune response. We evaluated the ratio between effective and regulatory T cells (Tregs) and PD-1 expression as a prognostic factor for operable breast cancer patients. A series of 218 newly diagnosed invasive breast cancer patients who had undergone primary surgery at Ruijin Hospital were identified. The influence of CD8⁺ cytotoxic T lymphocytes, FOXP3⁺ (Treg cell marker), and PD-1⁺ immune cell counts on prognosis was analyzed utilizing immunohistochemistry. Both PD-1⁺ immune cells and FOXP3⁺ Tregs counts were significantly associated with unfavorable prognostic factors. In bivariate, but not multivariate analysis, high tumor infiltrating PD-1⁺ cell counts correlated with significantly shorter patient survival. Our results suggest a prognostic value of the PD-1⁺ immune cell population in such breast cancer patients. Targeting the PD-1 pathway may be a feasible approach to treating patients with breast cancer.

Electronic supplementary material

The online version of this article (doi:10.1007/s00262-014-1519-x) contains supplementary material, which is available to authorized users.

Introduction

Breast cancer is the most common malignancy in women worldwide, and traditionally not viewed as an immunogenic disease. However, a large body of evidence has shown the existence of immune defects in breast cancer patients, with heavy tumor infiltration of immune cells being frequently observed [1, 2]. These immune cells are mainly tumor-infiltrating lymphocytes (TILs) and are considered to represent a local tumor immune host response. The type, density, and location of TILs within the tumor have shown to be predictors of patients survival in addition to histopathological methods currently used to classify early stage breast cancer [3–5].

The majority of TILs are CD8⁺ cytotoxic T lymphocytes (CTLs), which are a major effector cell type, and prominent CTLs infiltration has been linked to a better prognosis in breast cancer [3, 6]. Various other suppressor immune cells are also recruited and infiltrated into the tumor microenvironment, probably playing an important role in antitumor immune response and tumor progression [5, 7, 8]. It has been demonstrated that, as a major immune suppressive element, regulatory T cells (Tregs) infiltrate tumors and impede the tumor immune response. FOXP3, a forkhead family transcription factor, has been considered the most specific marker to identify Treg cells [9]. Furthermore, these effector cells may become tolerant or their function “exhausted” through such mechanisms as inappropriate stimulation or chronic antigen exposure. Ample evidence exists that the coinhibitory checkpoint receptor, programmed death-1 (PD-1), represents an exhausted status of T cells [10, 11]. PD-1 is expressed on chronically activated T cells, inhibiting their cytokine production through interaction with its ligands, PD-L1 and PD-L2. This interaction also facilitates Treg-induced suppression of T effector cells, further dampening the antitumor immune response [7, 12, 13].

Encouraging early clinical results using blocking agents against components of the PD-1 pathway have validated its importance as a target for cancer immunotherapy [14, 15]. Notably, objective and durable responses were achieved in patients with non-small cell lung cancer (NSCLC), which is notoriously resistant to immunotherapy [11, 14, 16]. Of note, 36 % of NSCLC patients whose tumors expressed PD-L1 (defined as >5 %) demonstrated an objective response while none whose tumors did not express PD-L1 responded [14, 17]. Thus, PD-L1 expression has been proposed as a biomarker for selecting candidate patients for PD-1 axis inhibition [18]. Accordingly, it would be of interest to evaluate the potential of linking the expression level of PD-1 to the response to immunotherapeutic strategies in the clinical settings.

PD-1-positive T lymphocytes and Tregs have been shown to co-infiltrate tumors of high-risk breast cancer patients [19]. An inverse correlation between the expression levels of PD-1 and prognosis has been reported in renal cell carcinoma (RCC) [20], while the predictive value of PD-1 expression in other malignancies, including breast cancer, has rarely been investigated.

Therefore, our aim in this study was to analyze the influence of CTLs, Tregs, and PD-1⁺ immune cell density on breast cancer patient prognosis and determine which of the immunosuppression mechanisms seem most important [21]. Relationships between immune cells infiltrated in the microenvironment of breast cancer and patient outcome were retrospectively investigated. Both tumor-associated immune cell PD-1 expression and the ratio of CD8⁺ to PD-1⁺ cells were shown in this study to be significant prognostic indicators.

Materials and methods

Patient selection

A series of 218 patients who had been diagnosed with and undergone primary surgery for unilateral invasive breast cancer at Ruijin Hospital (Shanghai, People’s Republic of China) between 2004 and 2008 were identified. Patients were selected based on availability of suitable formalin-fixed, paraffin-embedded (FFPE) tissue and complete clinicopathologic and follow-up data. None had distant metastatic disease at diagnosis or had received prior anticancer therapy. Written informed consent for research was obtained from all patients with the approval of the Ethics Committee of Ruijin Hospital.

Clinical and pathological features

The original pathological diagnosis was confirmed by breast pathologists. Pathologic features were collected by review of patient files. Clinicopathologic features including age, tumor size, histologic grade, histologic subtype, lymph node status, pathological TNM staging, ER, PR, and HER2 status were investigated. Only those showed 3+ for HER2 by immunohistochemical (IHC) staining were considered HER2-positive. Histologic grading was carried out using the modified Bloom and Richardson grading system [22].

Treatment and follow-up

All patients received standard surgical treatment of either mastectomy or wide local excision followed by radiotherapy. Standard adjuvant systemic therapies were administrated. Patients were followed up at 3-month intervals initially, then every 6 months or annually.

Overall survival (OS) and disease-free survival (DFS) data were obtained for each patient. OS was defined as time from date of surgery until death or last follow-up. DFS was defined as time elapsed from date of surgery to date of the first event (locoregional recurrence and/or distant recurrence, whichever came first). Follow-up was completed on October 15, 2012, and duration of follow-up was calculated from date of surgery to date of cancer recurrence, last follow-up, or death.

Immunohistochemistry (IHC) staining for CD8, FOXP3, and PD-1

Archived FFPE tissues were used. All cases were histologically reviewed by hematoxylin and eosin (H&E) staining. A single representative paraffin tumor block for each patient was chosen; 10 patients without available sequential sections including both tumor and tissue surrounding the tumor were thus excluded.

Sequential 5 μm specimens were collected on super defrosted slides, then baked at 60 °C for 1 h before IHC staining. IHC staining was performed by standard ABC method using the Vectastain Elite ABC kit from Vector Laboratory (cat. PK-6102).

Tissue sections were dewaxed in xylene and rehydrated through graded ethanol to distilled water. Endogenous peroxidase was quenched by 0.3 % hydrogen peroxidase for 15 min. Antigen retrieval was performed by heating slides in citrate buffer (pH 6.0) (Dako, S1699) with a regular cooking steamer for 30 min.

After serial blocking with 10 % normal horse serum, sections were incubated with primary monoclonal mouse antihuman antibody against CD8 (Dako, clone C8/144B, M7103), FOXP3 (Abcam, ab20034, clone 236A/E7), or PD-1 (Abcam, ab52587, clone NAT) overnight at 4 °C.

Biotinylated horse anti-mouse secondary antibodies (Vectastain Elite ABC kit, Vector Labs, Burlingame, CA) were added the following day, incubated for 30 min at room temperature, and tertiary reagent applied according to manufacturer’s instructions.

Antigen detection was performed by color reaction with 3,3′-diaminobenzidine tetrahydrochloride (Dako, K3468).

Finally, slides were counterstained in hematoxylin, dehydrated through ascending alcohols to xylene, and mounted.

Positive and negative controls were carried out on FFPE lymph node tissues using corresponding monoclonal antibody and mouse isotope IgG.

Evaluation of immune cell infiltration

Two breast pathologists (Xiaochun Fei and Fei Yuan) examined the H&E- and IHC-stained slides and counted the number of lymphoid infiltrates. Both were blinded as to the clinicopathologic characteristics and patient outcome.

Mononuclear immune cells infiltration was evaluated on H&E-stained sections. A standard grading system for semiquantitative scoring of lymphocytic infiltration was used to evaluate these H&E-stained sections [23]. Infiltration intensity was graded as absent (0), mild (score of 1, a few scattered lymphocytes), moderate (2, focal infiltrating immune cells), or marked (3, diffuse infiltration mimicking a lymphoid organ).

At least five independent high-powered microscopic fields (400× HPF), representing the densest lymphocytic infiltrates, were selected for each IHC-stained sample. The numbers of positively stained cells present within tumor nests and immediately adjacent stroma (in direct contact with tumor cells) were manually counted. Positive cells in distant stroma not touching tumor cells (defined as distance between positive cells and tumor nests being more than the size of one tumor cell) were disregarded. Cases in which scores differed between the two pathologists were reexamined and a consensus reached.

Statistical analysis

Statistical analyses were conducted with SPSS 16.0 statistical software (SPSS, Chicago, IL). The relationship among immunoreactive markers and the association between (high or low) numbers of tumor-infiltrating cells and clinicopathologic characteristics were analyzed by chi-square tests, Spearman’s rank correlation or Mann–Whitney U test where appropriate. T test was used to compare FOXP3⁺ cells numbers in PD-1 expression high and low groups. The log-rank test was used for univariate analyses, and cumulative survival curves were estimated by the Kaplan–Meier method. The influence of PD-1⁺ immune cells on survival was evaluated in multivariate analyses by Cox proportional hazards regression model. Two-sided p values <0.05 were considered statistically significant.

Results

Clinical profiles of patients

Clinicopathologic characteristics are summarized in Supplemental Table 1. The average patient age was 57.6 years [range 31–85; standard deviation (SD) 12.3 years), and mean tumor size was 2.4 cm (range 0.8–6.0; SD 0.9 cm). The 208 patients were categorized into two groups for further analysis, using median value of age (56 years) and tumor size (2.1 cm) as cutoffs, respectively. There were 76 patients (36.5 %) with stage I, 101 (48.6 %) with stage II, and 31 (14.9 %) with stage III disease. Histological subtypes were 184 with invasive ductal carcinoma (IDC), 16 with invasive lobular carcinoma (ILC), and 8 with others.

Clinical outcome

Patients were followed for 0.67–8.52 years, with a median of 6 years. Mean follow-up was 5.63 years, during which 32 (15.4 %) patients had local, regional-, and/or distant recurrences, of whom 6 were alive at the end of the study.

At last follow-up, 32 patients (15.4 %) had died, including 26 (12.5 %) from breast cancer at a median of 2.98 years following surgery (range 0.67–5.61). Among the 176 (84.6 %) patients still alive at last follow-up, median duration of follow-up was 6.71 years (range 4.29–8.52).

OS and DFS rates [standard error (SE), patient number still at risk) of patients at 3, 5, and 7 years following surgery are shown in Supplemental Table 2.

Immunohistochemical characteristics

A total of 208 tumors were examined after exclusion of uninformative slides. Of these, mononuclear immune cell tumor infiltration was absent in 10 (4.8 %) specimens, focally present in 62 (29.8 %), moderately in 104 (50.0 %), and markedly in 32 (15.4 %). Lymphocytes infiltrating within tumor beds presented in a diffuse pattern, while those in surrounding tissues were significantly more abundant and tended to form lymphoid aggregates.

Infiltration pattern, frequency, and localization of CD8⁺, PD-1⁺ and FOXP3⁺ cells

There were more IHC-stained cells in peritumoral areas when compared to those within tumor nests (Fig. 1). Ratios of CD8⁺, PD-1⁺, and FOXP3⁺ cells varied substantially among tumor samples.

Fig. 1.

Representative tumor tissue infiltrated by CD8⁺, PD-1⁺, and FOXP3⁺ cells depicting the localization pattern: intratumoral (red arrow), adjacent, and distant stromal (black arrow head). CD8⁺ (a, b) and PD-1⁺ (c, d) cells displayed membranous staining pattern, FOXP3⁺ (e, f) lymphocytes exhibited distinct nuclear staining, a, c, e ×100 and b, d, f ×200

CD8⁺ and PD-1⁺ lymphocytes displayed a membranous staining pattern, while FOXP3⁺ cells exhibited distinct nuclear staining, consistent with its expected transcription function (Fig. 1). No tumor cell expression of FOXP3 was demonstrated. Expression levels of CD8, PD-1, and FOXP3 in tumor-associated immune cells ranged from weak to intense.

Results were expressed as mean, median, and interquartile numbers of positive cells for one computerized 400× microscopic field (0.0768 mm²/field). Representative images (Fig. 1) and statistics of IHC variables are shown (Table 1).

Table 1.

Descriptive statistics of IHC variables

Variable	Mean	Lower IQ	Median	Higher IQ	Rangea
CD8+	13.98	4.40	7.80	15.35	0–173
FOXP3+	5.56	0.73	2.70	6.28	0–71
PD-1+	7.59	0.45	2.10	7.00	0–118

^a Cell count per field (cells/0.0768 mm²) IQ interquartile

Correlation between number of CD8⁺ CTLs, FOXP3⁺ Tregs, and PD-1⁺ immune cells

Next investigated was the correlation between CD8⁺ TIL, PD-1⁺ immune cells, and FOXP3⁺ Tregs infiltration in tumors, and vise versa. Highly significant correlations were shown between expressions of PD-1 in tumor-associated immune cells and FOXP3⁺ Tregs and CD8⁺ CTLs (r ², 0.385 and 0.385, respectively, both p < 0.001) as shown by Spearman’s rank correlation.

Notably, the majority of patients within the high PD-1 expression group (72/104, 69.2 %) also possessed more FOXP3⁺ cells (mean ± SD: 8.79 ± 12.13 vs 2.32 ± 2.99, p < 0.001). In addition, there was a significant correlation between FOXP3⁺ Tregs infiltration and CD8 expression in TIL (r ² = 0.154; p = 0.027).

Correlation of PD-1, CD8, and FOXP3 expressions in TIL with clinicopathologic features

Infiltrating densities of CTLs, Tregs, and PD-1⁺ cells were categorized as high or low in relation to their medians. Comparison of clinical and pathologic features by FOXP3 and PD-1 expression for the 208 patients is shown in Table 2. Both PD-1⁺ cells and FOXP3⁺ Tregs correlated positively with high histologic grade and later pathologic stage (both p < 0.001 and p = 0.006, respectively). Tumor-associated FOXP3⁺ Tregs were found also to be associated with progesterone receptor (PR) status and ER/PR/HER2 (triple)-negativity, although this did not reach statistical significance (p = 0.061 and 0.081, respectively) (Table 2).

Table 2.

Correlation of Foxp3⁺ Treg and PD-1⁺ immune cells with the clinicopathologic parameters

Variable	Foxp3			PD-1
	Low (%)	High (%)	p	Low (%)	High (%)	p
Age (years)
≤55	53 (51.0)	51 (49.0)	0.890	48 (46.2)	56 (53.8)	0.332
>55	51 (49.0)	53 (51.0)		56 (53.8)	48 (46.2)
Tumor size (cm)
≤2	51 (49.0)	53 (51.0)	0.890	53 (51.0)	51 (49.0)	0.890
>2	53 (51.0)	51 (49.0)		51 (49.0)	53 (51.0)
Histological grade
I, II	96 (61.5)	60 (38.5)	<0.001	94(81.9)	62 (50.0)	<0.001
III	8 (15.4)	44 (84.6)		10 (18.1)	42 (50.0)
TNM stage
I, II	96 (54.2)	81 (45.8)	0.006	96 (54.2)	81 (45.8)	0.006
III	8 (25.8)	23 (74.2)		8 (25.8)	23 (74.2)
Lymph node status
Negative	75 (52.1)	69 (47.9)	0.453	81 (56.2)	63 (43.8)	0.010
Positive	29 (45.3)	35 (54.7)		23 (35.9)	41 (64.1)
ER status
Negative	19 (40.4)	28 (59.6)	0.184	13 (27.7)	34 (72.3)	<0.001
Positive	85 (52.8)	76 (47.2)		91 (56.5)	70 (43.5)
PR status
Negative	31 (40.8)	45 (59.2)	0.061	23 (30.3)	53 (69.7)	<0.001
Positive	73 (55.3)	59 (44.7)		81 (61.4)	51 (38.6)
HER2 status
Negative	97 (51.6)	91 (48.4)	0.239	98 (52.1)	90 (47.9)	0.098
Positive	7 (35.0)	13 (65.0)		6 (30.0)	14 (70.0)
TNBC
No	89 (53.3)	78 (46.7)	0.081	95 (56.9)	72 (43.1)	<0.001
Yes	15 (36.6)	26 (63.4)		9 (22.0)	32 (78.0)
TIL grade
Low (0–1)	52 (72.2)	20 (27.8)	<0.001	54 (75.0)	18 (25.0)	<0.001
High (2–3)	52 (38.2)	84 (61.8)		50 (36.8)	86 (63.2)

Pearson’s chi square-test

Bold values are statistically significant (p <0.05)

Italic values are p values between 0.05 and 0.1, indicating a trend, though not reaching statistical significance

ER estrogen receptor, PR progesterone receptor, HER2 human epidermal growth factor receptor-2, TNBC triple-negative breast cancer, TIL tumor-infiltrating lymphocytes

On the other hand, numbers of PD-1⁺ cells were significantly inversely correlated with both ER and PR status (p < 0.001). In addition, higher density infiltration of PD-1⁺ immune cells also correlated significantly with triple-negative disease (p < 0.001). Furthermore, there was a moderately positive correlation between number of PD-1⁺ cells and lymph node status (p = 0.010). Finally, patients with greater infiltrating PD-1⁺ immune cells seemed more likely to have HER2-positive tumors (p = 0.098) (Table 2).

No significant associations were identified between PD-1⁺ immune cells or FOXP3⁺ Treg infiltration and patient age or tumor size. Also, CTLs did not associate with clinicopathologic features except for a negative correlation with patient age (p = 0.008) (Supplemental Table 3).

Survival analysis

Association with patient outcome by univariate assessment

On univariate analysis, age, tumor size, histologic grade, lymph node involvement, and ER, PR, HER2 status showed prognostic significance for OS; age, tumor size, histologic grade, and lymph node involvement were also negatively associated with DFS. Increased pathological stage was associated with decline in OS and DFS, but this did not reach significance (p = 0.055 and 0.054, respectively) (Table 3).

Table 3.

Univariate analyses of factors associated with survival and recurrence

Variable	OS			DFS
	HR	95 % CI	p	HR	95 % CI	p
Age, years (>55 vs ≤55)	2.49	1.18–5.26	0.017	3.37	1.51–7.50	0.003
Tumor size, cm (>2 vs ≤2)	4.97	2.05–12.10	<0.001	3.42	1.56–7.76	0.002
Histological grade (III vs I–II)	3.69	1.84–7.38	<0.001	2.81	1.40–5.66	0.004
TNM stage (III vs I–II)	2.19	0.98–4.88	0.055	2.20	0.99–4.89	0.054
LN involvement (yes vs no)	5.95	2.81–12.58	<0.001	6.15	2.90–13.04	<0.001
ER status (pos vs neg)	0.42	0.20–0.85	0.016	0.58	0.28–1.23	0.159
PR status (pos vs neg)	0.49	0.25–0.99	0.046	0.63	0.31–1.27	0.193
HER2/neu status (pos vs neg)	2.71	1.11–6.61	0.028	1.69	0.59–4.84	0.326
CD8+ CTL (high vs low)	0.97	0.48–1.94	0.925	0.71	0.35–1.42	0.329
FOXP3+ Treg (high vs low)	1.74	0.85–3.56	0.130	1.76	0.86–3.61	0.121
PD-1+ cells (high vs low)	3.29	1.48–7.32	0.002	2.29	1.08–4.84	0.024

Cox proportional hazards regression model

Bold values are statistically significant (p <0.05)

OS overall surviva, DFS disease free survival, HR hazard ratio, CI confidence interva, neg negative, pos positive

Prognostic significance of PD-1 immune cells, CTLs, and Tregs infiltration

Patients with fewer tumor-associated PD-1⁺ immune cells had longer OS (median 70 months) and DFS (median 69 months) than those with abundant PD-1⁺ immune cells (median 51 and 34 months, respectively). High tumor infiltrating PD-1⁺ cell counts were associated with significantly shorter OS (p = 0.002, log rank = 9.55, Fig. 2a), with a hazard ratio (HR) of 3.29 (95 % CI 1.48–7.32) (Table 3). Furthermore, there was significant difference in DFS between patients with high and low PD-1⁺ cell numbers (p = 0.024, log rank = 5.13, Fig. 2b).

Fig. 2.

Kaplan–Meier analysis of overall survival and disease-free survival for tumor-infiltrating PD-1⁺ immune cells (a, b), tumor-infiltrating FOXP3⁺ regulatory T cells (Tregs) (c, d) and balance between tumor-infiltrating CD8⁺ cytotoxic cells (CTLs) and PD-1⁺ immune cells (e, f). High PD-1⁺ immune cells infiltration associated with both shortened survival and increased recurrence

Extent of lymphocytic infiltration correlated with OS but not for DFS (Supplemental Fig. 1A and 1B). Neither CD8⁺ TILs nor FOXP3⁺ Tregs were significantly associated with patient survival (Supplemental Fig. 1C and 1D, Fig. 2C and 2D, respectively).

Combined influence of ratio between CTLs and Tregs or PD-1⁺ immune cells

On the basis of low and high expression groups for each IHC characteristic, patients were further classified into 4 groups to reflect CD8⁺/FOXP3⁺ and CD8⁺/PD-1⁺ ratios. CD8⁺/FOXP3⁺ ratio was not significantly associated with patient outcome in the univariate Cox model (Supplemental Fig. 1E and 1F).

Interestingly, patients with a high CD8⁺/PD-1⁺ ratio had relatively longer OS, and this was statistically significant in univariate analysis (HR 0.46; 95 % CI 0.25–0.84; p = 0.012). Furthermore, patients with CD8⁺ low/PD-1⁺ high infiltration showed a significantly poorer outcome compared with other groups in both OS and DFS (Fig. 2e, F).

PD-1 expression not an independent prognostic factor for overall survival

Well-recognized clinical variables, including tumor size, histological grade, and lymph node involvement, were selected as covariates for Cox proportional hazards analyses. Multivariate analysis showed that higher PD-1⁺ cell counts were not independently associated with shorter OS (HR = 2.06; 95 % CI 0.89–4.77; p = 0.091, Table 4).

Table 4.

Multivariate analyses of factors associated with overall survival

Variable	HR	95 % CI	p
Tumor size, cm
≤2	1
>2	3.92	1.61–9.58	0.003
Histological grade
I–II	1
III	2.73	1.33–5.64	0.007
LN involvement
No	1
Yes	4.48	2.10–9.57	<0.001
PD-1+ cell numbers
Low	1
High	2.06	0.89–4.77	0.091

Bold values are statistically significant (p <0.05)

HR hazard ratio, C confidence interval

Discussion

The immune system is envisaged to have a homeostatic role in preventing cancer development through immune surveillance. However, immunologic clearance of macroscopic tumors is rare, although it has been well established that all cancer cells express specific antigens which are derived from genetic alterations and epigenetic dysregulation and are potentially recognizable by T immune cells. The recently refined concept of “cancer immunoediting” integrates the immune system’s dual host-protective and tumor-promoting roles [7]. Tumors may escape the host immune attack by a variety of complementary mechanisms of immunosuppression. The ability to evade destruction by the host immune system, also called immune escape from the immunosurveillance, is now viewed as one of the emerging “hallmarks” of cancer [24]. This is felt to occur through cancer cells “hijacking” multiple mechanisms of immunosuppression, which have evolved to control the stimulation of “immunity to self” [25]. The functional profile of so-called “effector” T cells, which may lead to the death of target cells, is dramatically impacted in cancer by both cellular and molecular mediators [5, 21].

Various immunosuppressive cells are recruited to the tumor microenvironment by suppressive cytokines secreted by cancer-associated fibroblasts (CAFs), among other cells. Of these immunosuppressive cells, Treg cells are crucial mediators in cancer. They inhibit activation and function of effector T cells through immediate cell–cell contact and paracrine production of cytokines such as TGF- and IL-10 [26].

In addition, tumor cells may upregulate surface ligands, including B7 homolog 1 (B7-H1, PD-L1) [27, 28] and other ligands [29–31], which mediate resistance to CTL-induced tumor cell apoptosis by methylation of suppressor of cytokine signaling 1 (SOCS1) [32, 33]. Inducing PD-1 expression on infiltrated immune cells, mainly CD8⁺ TILs, leading to their “exhausted” state, may explain one way in which tumor cells adapt to escape from immune surveillance [31, 34, 35]. It has recently been suggested that PD-1 may also contribute to anergy of TILs merely by its expression, independent of interaction with its ligands [36]. Hence, heavy infiltration of both FOXP3⁺ Tregs and PD-1⁺ anergic immune cells reflects the induced dysfunctional status of the adaptive immune system in the tumor microenvironment. The intratumoral balance between functional effector T cells and regulatory T cells reflects the host’s anticancer ability and thus may be an independent prognostic indicator [37].

In this study, tumor cell-associated immune cells (CD8⁺ CTLs, FOXP3⁺ Tregs, and PD-1⁺ cells) including their density and types were retrospectively investigated, with the aim of determining a correlation with patient outcome. These analyses were performed by IHC staining using sequential tumor sections.

PD-1⁺ cell infiltration was observed in most of the breast tumors. Similar to previous studies investigating TILs, the median value was chosen as the cutoff in order to compare the role of high numbers of PD-1⁺ immune cells. In data from RCC patients [20], where 29.2 % (78/267) of patients demonstrated moderate to marked infiltration of mononuclear immune cells, 28.8 % (77/267) of tumors presented PD-1⁺ cells [20]. In a recent similar study in breast cancer [38], the ratio (15.8 %, 104/660) was much less than the percent of patients with PD-1⁺ cells (91.3 %, 190/208) in our study. Of note, we used whole sections, which could provide more information than tissue microarray, and high-powered microscopic fields representing the densest lymphocytic infiltrates were selected and counted. Thus, we obtained far more positive PD-1⁺ cells than Muenst et al. [38].

In accordance with a previous breast cancer report [19], both PD-1⁺ cells and FOXP3⁺ Tregs were significantly associated with poor tumor differentiation, more advanced stage and other unfavorable prognostic factors. Significantly, more PD-1⁺ immune cell infiltration was observed in tumors with unfavorable molecular features, including negative ER/PR expression and triple negativity. Furthermore, patients with HER-2 positive tumors and lymph node involvement were prone (p = 0.098 and 0.010, respectively) to have more infiltrating PD-1⁺ cells (Table 2).

No correlation between CD8⁺ cells and clinicopathologic factors except for patient age was found (Supplemental Table 3). Furthermore, CTLs (CD8⁺ cells), which are primarily responsible for cancer control through the adaptive immune system and reported to be prognostic in various cancer types [3, 4, 37, 39], did not correlate with improved OS or DFS. Notably, in our study, we attempted to distinguish between stromal lymphocytes and TILs. Only those lymphocytes in direct contact with tumor cells were considered to represent an in situ host antitumor immune response and thus counted. This criterion was based on recent studies in this field [3, 4, 40]. The present results are consistent with the findings that CD8⁺ TILs in remote stroma, rather than tumor cell-associated CD8⁺ T cells, seem to predict breast cancer survival [3, 4].

FOXP3⁺ Tregs infiltration also did not demonstrate an association with the patient survival, consistent with previously published breast cancer data [40], suggesting that it is not possible to predict clinical outcome using Tregs alone. In that study, intratumoural and tumor-adjacent stromal FOXP3⁺ cells were associated with a worse prognosis with long-term follow-up. However, in multivariate analysis, number of FOXP3⁺ cells was not found to be an independent prognostic factor. It was thus concluded that FOXP3⁺ infiltrating cells did not have a dominant role in prognosis [40].

By univariate analysis, we observed a positive association between high tumor infiltrating PD-1 counts and shorter OS (p = 0.002, log rank = 9.55, Fig. 2a). HR was 3.29 (95 % CI 1.48–7.32), compared to the HR, 2.24, of PD-1 expression in RCC patients [20]. Thus, it could be that induction of PD-1 expression on infiltrating immune cells allows tumor cells to escape immune surveillance [28].

While absolute numbers of TILs reflect general activation of the host immune system, the ratio between effector and regulatory T cells (Tregs) may serve as a better independent prognostic factor for breast cancer survival [4, 9]. In one breast cancer study, the peritumoral CTL/Treg ratio was shown to be an independent prognostic factor [4]. However, the CD8⁺/FOXP3⁺ effect on OS and PFS was not found in our study to be significant, possibly because of the small patient number. However, the high CD8⁺/PD-1⁺ ratio seemingly improved survival (Fig. 2e, f), suggesting its fundamental role in reflecting the functional immune host status, thereby predicting outcome.

In contrast to the study by Muenst et al. [38], high PD-1 expression was not an independent prognostic factor in multivariate analysis when adjusted for well-recognized clinical variables. This may be a result of low statistical power of this study due to the relatively small patient number. However, multiple layers of immune suppression mechanisms in tumor evasion [41] and the complexity of the tumor microenvironment [26] could also explain this. Other suppressive immune cells besides Tregs such as myeloid-derived suppressor cells (MDSCs) and tumor-associated macrophages (TAMs) [42] might also promote indirect immune escape [21]. Clearly, there are other co-stimulatory checkpoints expressed on immune cells besides PD-1, including 2B4 (CD244), BTLA, CTLA-4, CD160, LAG-3, and Tim-3, each of which may play a part in the immune response [8, 10, 43].

The predictive values of Tregs and CTLs in breast cancer remain fairly controversial. FOXP3⁺ infiltrating cells have been found inconsistently to be an independent prognostic factor in patients with ER-positive-only tumors [44], both ER-positive and ER-negative groups [45], or neither of them [40]. Surprisingly, one study found that high numbers of FOXP3 TIL were associated with a favorable prognosis in ER-negative patients [46]. Similarly, CD8⁺ cells were a favorable independent prognostic value only in ER-negative [47] or in basal-like breast cancers, but not in other intrinsic molecular subtypes [48]. Thus, it seems likely that different immune mechanisms are operational in different types of breast cancer [6]. Due to the relatively small number of patients, 111 in all, we made no attempt to determine the predictive value of the lymphocytic variables in various molecular subtypes. Further studies are warranted to verify these findings and further explore the underlying mechanisms.

Passive transfer of the monoclonal antibody, trastuzumab, has been well-established HER2⁺ breast cancer treatment [49]. Although not always considered as ‘immunotherapy’, its success reflects the ability of inducing an immediate immune reaction against the non-immunogenic cancer [5]. At this time, strategies for inducing antitumor immune responses by the patients’ own immune system to undermine the cancer-induced immunosuppressive mechanisms are actively under investigation.

Molecular and cellular mediators of cancer-induced immunosuppression, such as CTLA-4, PD-1, and Treg cells, have been studied in breast cancer [32, 35, 50], as well as the best-known escape mechanisms such as loss of tumor antigen expression [5, 7]. One of the generally accepted reasons for which the immune system has a low anticancer efficacy is that TILs are dysfunctional through expression of inhibitory receptors, like PD-1 [12, 50, 51]. Dissecting the functional role of different subsets of TILs in the tumor suppressive microenvironment is thus important.

Treatment for cancer patients by restoration of effective antitumor immune responses may require approaches aimed at protecting antitumor immune cells from the effects of regulatory cells and/or prolonging survival and function of effector cells by inhibiting negative stimulation and metabolic constraints [25]. Interestingly, the PD-L1/PD-1 axis has been reported to be involved in the development of induced Treg cells [33]. Therefore, targeting this signaling pathway may also reduce Treg development and function [52].

Cancer immunotherapy seems finally to have come of age after decades of failure and disappointment [5, 8]. Following the success of ipilimumab (anti-CTLA4) in metastatic melanoma, several groups have conducted trials aiming at subverting other immune checkpoints [11]. Blocking the PD-1 pathway has achieved objective responses and durable remissions of several late stage solid malignancies, including the poorly immunogenic NSCLC, with an 18 % overall response rate [14]. These preliminary results may shed light on a curative treatment for breast cancer, as well, although only 4 patients were recruited in the anti-PD-L1 trial, with no responses observed [15]. Still, this approach needs to be further explored in breast cancer [11, 15, 53]. As suggested, such immunotherapeutic strategies may be more rationally applied as adjuvant therapy in the early stage, low-volume settings [21].

Conclusions

In conclusion, our results provide some evidence of the prognostic importance of the PD-1⁺ immune cell population in breast cancer. Tumor progression was associated with increased Tregs and the immune check point PD-1 expression on CTLs and other immune cells at the tumor site [51]. Expression of PD-1 is associated with anergy and an inactive or apoptotic status of activated tumor infiltrating effector cells [54], thus weakening the host immune response and promoting tumor cell growth, ultimately affecting the outcome. Targeting T cell dysfunctional mechanisms thus seems to be a promising approach to treating breast cancer patients.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

PD-1

Programmed death-1

Treg

Regulatory T cell

CTL

Cytotoxic T lymphocyte

NSCLC

Non-small cell lung cancer

OS

Overall survival

DFS

Disease-free survival

BCSS

Breast cancer-specific survival

TIL

Tumor-infiltrating lymphocyte

PD-L1

Programmed death-ligand 1

IHC

Immunohistochemical

FFPE

Formalin-fixed paraffin-embedded

H&E

Hematoxylin and eosin

HPF

High-powered field

SD

Standard deviation

IDC

Invasive ductal carcinoma

ILC

Invasive lobular carcinoma

SE

Standard error

ER

Estrogen receptor

PR

Progesterone receptor

HER2

Human epidermal growth factor receptor-2

TNBC

Triple-negative breast cancer

CAF

Cancer-associated fibroblast

SOCS1

Suppressor of cytokine signaling 1

RCC

Renal cell carcinoma

HR

Hazard ratio

CI

Confidence interval

MDSC

Myeloid-derived suppressor cell

TAM

Tumor-associated macrophage

ORR

Overall response rate