Prediction of treatment responses to neoadjuvant chemotherapy in triple-negative breast cancer by analysis of immune checkpoint protein expression

A total of 177 patients with resectable, early-stage breast cancer diagnosed as stage IIA (T1, N1, M0 or T2, N0, M0), IIB (T2, N1, M0 or T3, N0, M0), or IIIA (T1-2, N2, M0 or T3, N1-2, M0) were treated with NAC between 2007 and 2013. Tumor stage and T and N factors were stratified based on the TNM Classification of Malignant Tumors, UICC Sixth Edition [19]. Breast cancer was confirmed histologically by core needle biopsy and staged by systemic imaging studies using computed tomography (CT), ultrasonography (US), and bone scintigraphy. Breast cancer was classified into subtypes according to the immunohistochemical expression of ER, PgR, HER2, and Ki67. Based on their immunohistochemical expression, the tumours are categorized into the immunophenotypes luminal A (ER+ and/or PgR+, HER2−, Ki67-low), luminal B (ER+ and/or PgR+, HER2+) (ER+ and/or PgR+, HER2−, Ki67-high), HER2-enriched (ER−, PgR−, and HER2+), and TNBC (negative for ER, PgR and HER2) [20]. In this study, HER2-enriched and luminal B (ER+ and/or PgR+, HER2+) were considered as HER2-positive breast cancer (HER2⁺BC).

All patients received a standardized protocol of NAC consisting of four courses of FEC100 (500 mg/m² fluorouracil, 100 mg/m² epirubicin, and 500 mg/m² cyclophosphamide) every 3 weeks, followed by 12 courses of 80 mg/m² paclitaxel administered weekly [21, 22]. Forty-five patients had HER2⁺BC and were additionally administered weekly (2 mg/kg) or tri-weekly (6 mg/kg) trastuzumab during paclitaxel treatment [23]. All patients underwent chemotherapy as outpatients. Therapeutic anti-tumor effects were assessed according to the response evaluation criteria in solid tumors (RECIST) criteria [24]. Pathological complete response (pCR) was defined as the complete disappearance of the invasive compartment of the lesion with or without intraductal components, including in the lymph nodes. Patients underwent mastectomy or breast-conserving surgery after NAC. All patients who underwent breast-conserving surgery were administered postoperative radiotherapy to the remnant breast. Overall survival (OS) time was the period from the initiation of NAC to the time of death from any cause. Disease-free survival (DFS) was defined as freedom from all local, loco-regional, and distant recurrences. All patients were followed up by physical examination every 3 months, US every 6 months, and CT and bone scintigraphy annually. The median follow-up period for the assessment of OS was 3.4 years (range 0.6–6.0 years), and for DFS it was 3.1 years (range 0.1–6.0 years). This study has been conducted in our institution, Osaka City University Graduate School of Medicine, Osaka, Japan, according to the reporting recommendations for tumor marker prognostic studies (REMARK) guidelines and a retrospectively written research, pathological evaluation, and statistical plan [25].

All patients underwent a core needle biopsy prior to NAC, and they underwent curative surgery involving a mastectomy or conservative surgery with axillary lymph node dissection after NAC at the Osaka City University. Immunohistochemical studies were performed as previously described on core needle biopsy specimens [26]. Tumour specimens were fixed in 10% formaldehyde solution and embedded in paraffin (FFPE), and 4-µm-thick sections were mounted onto glass slides. Slides were deparaffinized in xylene and heated for 20 min (105 °C, 0.4 kg/m²) in an autoclave in Target Retrieval Solution (Dako, Carpinteria, CA, USA). Specimens were then incubated with 3% hydrogen peroxide in methanol for 15 min to block the endogenous peroxidase activity, then incubated in 10% normal goat or rabbit serum to block non-specific reactions.

Primary monoclonal antibodies directed against ER (clone 1D5, dilution 1:80; Dako, Cambridge, UK), PgR (clone PgR636, dilution 1:100; Dako), HER2 (HercepTest™; Dako), Ki67 (clone MIB-1, dilution 1:00; Dako), PD-1 (clone NAT105, dilution 1:100; Abcam, Cambridge, UK), PD-L1 (clone 27A2, dilution 1:100; MBL, Nagoya, Japan), and PD-L2 (clone #176611, dilution 1:100; R&D Systems, Minneapolis, MN) were used. Tissue sections were incubated with each antibody for 70 min at room temperature or overnight at 4 °C, then incubated with horseradish peroxidase-conjugated anti-rabbit or anti-mouse Ig secondary antibodies (HISTOFINE (PO)™ kit; Nichirei, Tokyo, Japan). Slides were subsequently treated with streptavidin-peroxidase reagent and incubated in phosphate-buffered saline-diaminobenzidine and 1% hydrogen peroxide (v/v), followed by counterstaining with Mayer’s haematoxylin. Positive and negative controls were carried out on FFPE lymph node tissues using corresponding monoclonal antibody and mouse isotype IgG.

Immunohistochemical scoring was performed by two pathologists specialized in mammary gland pathology, using the blind method to confirm the objectivity and reproducibility of diagnosis. The cut-off value for ER and PgR positivity was set at ≥ 1% in accordance with previous studies [27]. HER2 expression was scored according to the accepted grading system (0, no reactivity, or membranous reactivity in less than 10% of cells; 1 +, faint/barely perceptible membranous reactivity in ≥ 10% of cells or reactivity in only part of the cell membrane; 2 +, weak to moderate complete or basolateral membranous reactivity in ≥ 10% of tumour cells; or 3 +, strong complete or basolateral membranous reactivity in ≥ 10% of tumour cells). HER2 expression was considered positive if the immunostaining score was 3 +, or in cases where the score was 2 + and included gene amplification via fluorescent in situ hybridization (FISH). For FISH analyses, each copy of the HER2 gene and its centromere 17 (CEP17) reference were counted. The interpretation followed the criteria of the ASCO/CAP guidelines for HER2 IHC classification for breast cancer: positive if the HER2/CEP17 ratio was higher than 2.0 [28]. A Ki67-labeling index ≥ 14% was classified as positive [20].

Histopathologic analysis of TILs was evaluated on a single full-face hematoxylin and eosin (HE)-stained tumor section using criteria described [29‐31]. To evaluate PD-1 expression, five stained areas were selected, and the number of TILs in stroma surrounding the stained cancer cells was quantitatively measured in each field under 400-times magnification (Fig. 1a). The median value of the average each field was determined, and that number was set as a cutoff value. To evaluate PD-L1 and PD-L2 expression, 3 fields of view (FOVs) in darkly stained areas were selected, and the percentage of cancer cells stained with anti-PD-L1 antibody and anti-PD-L2 antibody in each FOVs was measured under 400-times magnified microscopy (Fig. 1b, c). Based on previous reports, ≥ 10% was defined as high expression, and < 10% was defined as low expression [12, 32].

Statistical analysis was performed using the SPSS version 19.0 statistical software package (IBM, Armonk, NY). The associations between PD-1, PD-L1 and PD-L2 and clinicopathological variables were evaluated using χ² tests. Multivariate analysis of pCR was carried out using a binary logistic regression model. The Kaplan–Meier method was used to estimate DFS and OS, and the results were compared between groups using log-rank tests. Multivariate analysis of prognostic factors was carried out using a Cox regression model. A p value < 0.05 was considered significant. Cutoff values for different biomarkers included in this study were chosen before the statistical analysis.

The subtype in 177 patients who received NAC was TNBC in 61 (34.5%) patients and HER2⁺BC in 45 (25.4%) patients. Regarding treatment response, 67 (37.9%) patients had a pCR, and 110 (62.1%) patients had a non-pCR. According to subtype, 28 (45.9%) TNBC patients and 18 (40.0%) HER2⁺BC patients had a pCR.

TIL PD-1 expression ranged from 0 to 68. The median value of the average in 3 FOVs was 6. There were 37 (20.9%) patients with high PD-1 expression (≥ 6) and 140 (79.1%) patients with low PD-1 expression (< 6). In addition, 42 (23.7%) patients had high PD-L1 expression, and 52 (29.4%) patients had high PD-L2 expression.

Evaluation based on clinicopathologic features showed that patients with high PD-1 and PD-L1 expressions had a significantly higher rate of TNBC (p = 0.041) (p < 0.001) and HER2⁺BC (p = 0.004) (p = 0.004). Patients with PD-L1 expression had a significantly higher rate of non-pCR (p < 0.001), and PD-1 and PD-L2 expressions were greater (p < 0.001) (p < 0.001) (Table 1). There were no significant differences for other clinicopathologic features.

DFS was significantly longer in patients with low, compared to high, PD-1 and PD-L1 expressions (p = 0.006, log-rank) (p = 0.001, log-rank) (Fig. 2a, b). OS was also significantly longer in patients with low, compared to high, PD-1 and PD-L1 expressions (p = 0.048, log-rank) (p = 0.022, log-rank) (Additional file 1: Fig. S1A, B). DFS and OS did not differ significantly between patients with low vs high PD-L2 expression (p = 0.657, log-rank) (p = 0.615, log-rank) (Fig. 2c) (Additional file 1: Fig. S1C).

Univariate analysis showed that PD-1 and PD-L1 expressions were associated with significantly shorter DFS (p = 0.008, HR = 2.752) (p = 0.002, HR = 3.194). However, although multivariate analysis showed that lymph node metastases were an independent poor prognostic factor (p = 0.046, HR = 4.330), PD-1 and PD-L1 expressions were not independent prognostic factors (p = 0.492, HR = 1.415) (p = 0.084, HR = 2.613) (Table 2).

Among the 61 patients with TNBC, 18 (29.5%) had high PD-1 expression, 24 (39.3%) had high PD-L1 expression, and 20 (32.8%) had high PD-L2 expression. Analysis of clinicopathologic features showed that the high PD-1 expression group was significantly older (p = 0.016), and that patients with high PD-L1 expression had a significantly lower Ki67 index (p = 0.005). Patients with high PD-1 and PD-L1 expressions had significantly higher rates of non-pCR (p = 0.003) (p < 0.001). PD-1 expression was significantly correlated with PD-L1 expression (p < 0.001), it but was not correlated with PD-L2 expression (Table 3).

Analysis of outcomes showed that DFS was significantly longer in patients with low, compared to patients with high, PD-1 and PD-L1 expressions (p = 0.036, log-rank) (p = 0.001, log-rank) (Fig. 3a, b). OS was also significantly longer in patients with low, compared to patients with high, PD-1 expression (p = 0.021, log-rank), but OS was not significantly different based on PD-L1 expression (p = 0.155, log-rank) (Additional file 1: Fig. S1D, E). DFS and OS were also not significantly different based on PD-L2 expression (p = 0.665, log-rank) (p = 0.595, log-rank) (Fig. 3c) (Additional file 1: Fig. S1F).

Univariate analysis showed that PD-1 and PD-L1 expressions also significantly shortened DFS in TNBC (p = 0.048, HR = 3.318) (p = 0.007, HR = 8.375). However, multivariate analysis found that only PD-L1 expression was an independent prognostic factor (p = 0.041, HR = 9.479) (Table 4).