The prognostic value of a combined immune score in tumor and immune cells assessed by immunohistochemistry in triple-negative breast cancer

A total number of 306 TNBC patients who received surgical resection after being diagnosed with invasive ductal carcinoma at Seoul National University Hospital between 2003 and 2006 were included in this study. Histologic grading was based on the Nottingham grading system [30]. Selected clinicopathologic parameters and patient survival data were retrieved from the electronic medical records system. These included the following: age, tumor size, nuclear grade, histologic grade, presence/absence of a ductal carcinoma in situ component, lymphovascular invasion or lymph node metastasis, in addition to each patient’s anatomical (pTNM) and prognostic stage, according to the 8th edition of the American Joint Committee on Cancer (AJCC) [31]. Patients who had received neoadjuvant chemo- or radiotherapy were excluded from further analysis.

All enrolled TNBC cases had previously been subjected to estrogen receptor (ER) and progesterone receptor (PR) immunohistochemical analysis, also to Human epidermal growth factor receptor 2 (HER2) immunohistochemical and/or in situ hybridization testing, and had been evaluated based on the 13th St. Galen International Breast Cancer Conference and ASCO/CAP guidelines [20, 32, 33]. ER (1:100, 1D5; Novocastra Laboratories, Newcastle, UK) and PR (1:200, PgR636; Dako, Glostrup, Denmark) expression was considered as positive when ≥ 1% of staining was observed in tumor cells, according to the 2010 ASCO/CAP guidelines [32]. HER2 immunohistochemistry (1:1, 4B5; Ventana, Medical System, Tucson, AZ, USA) was regarded as positive (3+) when ≥ 10% of tumor cells showed complete and intense circumferential membranous staining, based on 2018 ASCO/CAP guidelines [33]. In equivocal cases (2+), fluorescence in situ hybridization (FISH) using the PathVysion assay (Abbott Molecular, Downers Grove, IL) was performed, and the test was considered as positive when either a ≥ 2.0 ratio of the HER2 to chromosome 17 gene copy number or ≥ 6.0 of average HER2 signals per tumor cell were observed [33]. This study was approved by the Institutional Review Board of Seoul National University Hospital (IRB No. 1511-085-720).

Immunohistochemical staining was performed on 227 TNBC cases, arranged in tissue microarrays consisting of 2 mm cores from the aforementioned cases (Superbiochips Laboratories, Seoul, Korea). A 4-μm section from each block was subjected to immunohistochemistry, using the Benchmark automatic immunostaining device (Ventana, Arizona, USA). The following eight primary antibodies were used: PD-1 (PDCD1) (1:20; Cell Marque, California, USA), PD-L1 (B7-H1) (1:100; Cell Signaling, Massachusetts, USA), PD-L2 (B7-DC) (1:500; Sigma-Aldrich, Missouri, USA), IDO (IDO1) (1:30, 1L30; Millipore-Sigma, Massachusetts, USA), TIM3 (HAVCR2) (1:550; Abbexa Ltd, Cambridge, UK), OX40 (TNFRSF4) (1:125; Novus Biologicals, Colorado, USA), OX40L (TNFSF4) (1:30; Millipore-Sigma, Massachusetts, USA), and B7-H2 (ICOSLG) (1:300; Novus Biologicals, Colorado, USA). Interpretation of all eight biomarkers was based on both intensity and proportion of the positively stained TCs and ICs; the latter included lymphocytes, macrophages, dendritic cells, and granulocytes [34]. For each case, the intensity score (IS) of each immune marker was graded as follows: 0 (negative), 1 (weak), 2 (moderate), and 3 (strong). Furthermore, the proportion score (PS) was graded as follows: 0 (stain under 1%), 1 (1–5%), 2 (5–10%), 3 (10–25%), 4 (25–50%), and 5 (> 50%). Subsequently, the IS and PS of each biomarker were multiplied to generate a final score, ranging from 0 to 15. To determine its clinicopathologic and prognostic significance, each marker’s expression score of 1 or greater (IS × PS ≥ 1) was regarded as positive when evaluating either TCs or ICs, similar to previous studies that considered as positive the expression of at least 1% of cells with any intensity [35, 36]. All immunohistochemical stainings were interpreted independently by three experienced breast pathologists (H.S.R., M.S.J., and J.E.C.) to enhance accuracy, while any discordance among them was resolved with a consensus.

To validate our findings, the TCGA breast cancer dataset was used to investigate the mRNA expression of the ICI-related markers exhibiting prognostic relevance in our immunohistochemical analysis. The expression of each biomarker was considered as positive or negative, using the 25th percentile as a cut-off. We also classified all breast cancer and TNBC cases into two groups, based on the gene expression levels of a combination of immune markers, similar to our immunohistochemical analysis. Kaplan–Meier curves and log-rank tests were used to determine whether there was a significant difference in survival. The TCGA data was obtained from cBioPortal and the survival analysis was processed using the R software, version 4.1.0 (R Foundation for Statistical Computing, Vienna, Austria).

The chi-square and Pearson’s correlation coefficient tests were utilized to investigate the potential association of the expression of the abovementioned eight immune markers in TCs and ICs with the extracted clinicopathological variables and between each other, respectively. To investigate the effect of each immune marker expression on TNBC patients’ progression-free survival (PFS)—measured from the date of surgery to the date of local recurrence and/or distant metastasis—and overall survival (OS), the Kaplan–Meier analysis with log-rank test was performed. The Cox proportional hazard model was further adapted to conduct multivariate analysis and identify independent prognostic predictors. Clinicopathologic characteristics were adjusted using a backward stepwise model, including covariates with a prognostic role. Statistical significance was considered at the level of p < 0.05. To construct each CIS, we only selected immune markers exhibiting a P-value of less than 0.1 (p < 0.1) in the PFS Kaplan–Meier analysis [37]. Statistical analysis was performed with the SPSS software for Windows, version 22.0 (SPSS, Chicago, IL, USA).

The clinicopathologic characteristics of the enrolled patients are shown in Table 1. All patients were female with a median age of 48 (range: 21–76) years. Most of the patients were under 60 years old (86.3%) at the time of diagnosis. There were 145 (63.9%) patients with a tumor diameter greater than 2.0 cm, while 73 (32.2%) patients showed lymph node metastasis. Most TNBCs had nuclear score 3 (80.2%), also histologic grade 3 (82.4%), according to the Nottingham histologic grade. The anatomic stage was I in 61 (26.9%), II in 128 (56.4%), and III in 38 (16.7%) patients. The prognostic stage was II in 62 (27.3%) and III in 165 (62.7%) patients. The median follow-up duration was 103 months, and 217 (95.6%) of the tested TNBC patients had received adjuvant chemotherapy.

Expression of each immune marker was evaluated separately in TCs and ICs (Fig. 1). In TCs, positive expression was identified in the following markers: PD-L1 (n = 46, 20.3%), PD-L2 (n = 197, 86.8%), IDO (n = 129, 56.8%), TIM3 (n = 109, 48.0%), OX40 (n = 82, 36.1%), OX40L (n = 86, 37.9%), and B7-H2 (n = 193, 85.0%). PD-1 was not expressed in TCs in all TNBC cases examined. Positive expression in ICs was observed in the following markers: PD-1 (n = 126, 55.5%), PD-L1 (n = 91, 40.1%), PD-L2 (n = 108, 47.6%), IDO (n = 154, 67.8%), TIM3 (n = 104, 45.8%), OX40 (n = 84, 37.0%), OX40L (n = 103, 45.4%), and B7-H2 (n = 150, 66.1%). The expression of the eight immune markers were compared between one another, as shown in Fig. 2. In ICs, the expression of each biomarker was positively correlated with the rest at a significant level, except between PD-L1 and OX40, also TIM3 and OX40. On the other hand, in TCs, a significant positive correlation was also shown—especially between PD-L2 and TIM3 (p < 0.01, rho = 0.42), TIM3 and B7-H2 (p < 0.01, rho = 0.43), and OX40 and OX40L (p < 0.01, rho = 0.53)—albeit less often than in ICs. Notably, IDO expression in TCs was negatively correlated with the OX40 expression, at a significant level (p = 0.04, rho =− 0.14).

Furthermore, the immunohistochemical expression of the eight immune markers in both TCs and ICs was compared with the extracted clinicopathologic parameters (Table 2). Regarding TCs, PD-L1 positive expression was significantly associated with an anatomic stage I and II, compared to stage III (p = 0.038). OX40 positive expression was associated with a higher rate of lymph node metastasis (p = 0.011) and an advanced anatomic stage (p < 0.001), at significant levels. OX40L positivity was also associated with an advanced anatomic stage (p = 0.016). Regarding ICs, PD-L1 positive expression was associated with an anatomic stage I and II, compared to stage III (p = 0.024), which was a concordant finding with the result of TCs. TIM3 and OX40L positive expression was significantly associated with lymph node metastasis (p = 0.031 and p = 0.040). OX40L positivity was also significantly associated with an advanced anatomic stage (p = 0.003). In our analysis, OX40 was the only immune marker that was significantly associated with an aggressive tumor behavior, such as a higher risk of lymph node metastasis and an advanced anatomic stage both in TCs and ICs (in TCs: p = 0.011 and p < 0.001; in ICs: p = 0.040 and p = 0.003, respectively). In addition, when positive expression in both TCs and ICs was considered to assign the expression of an immune marker as positive, OX40 positive expression was significantly associated with the presence of lymph node metastasis and advanced anatomic stage (p = 0.005 and p < 0.001, respectively), in contrast to other immune markers (Additional file 1: Table S1).

During the median follow-up period of 102.6 months, 33 recurrent or metastatic events occurred and 23 patients died with disease. The median PFS and OS were 100.9 and 102.6 months, respectively. The prognostic significance of each individual immune-related marker was analyzed, based on its expression on TCs and ICs (Table 3; Additional file 3: Fig. S1; Additional file 4: Fig. S2). Regarding TCs, OX40 overexpression was associated with a shorter PFS (78.0% vs. 89.7%, p = 0.025), while an increased B7-H2 expression had a tendency for shorter PFS (83.4% vs. 97.1%, p = 0.054). In contrast, positive expression of PD-L1 in TCs had a tendency of longer PFS (93.5% vs 83.4%, p = 0.081). Regarding ICs, positive expression of PD-1 (89.7% vs. 80.2%, p = 0.040), PD-L1 (91.3% vs. 80.9%, p = 0.020), and OX40L (91.3% vs. 80.6%, p = 0.020) were significantly associated with a longer PFS. Additionally, the overexpression of PD-L2 in ICs had a tendency for longer PFS (89.8% vs 81.5%, p = 0.086). When positive expression in both TCs and ICs was considered to assign the expression of an immune marker as positive, the positive expression of all eight markers did not show a statistical significance with either PFS or OS (Additional file 2: Table S2).

To comprehensively analyze the prognostic significance of combined immune marker expression in TNBC, we evaluated “combined immune scores (CIS)” expressed in both TCs and ICs. The definition of CIS and its scoring method in our study are summarized in Fig. 3. As the positive expression of all immune markers in both TCs and ICs failed to exhibit prognostic value (p < 0.1) regarding PFS and OS, CIS were constructed separately for TCs and ICs. First, we selected the three immune markers (PD-L1, OX40, and B7-H2) that showed prognostic significance (p < 0.1) in TCs, according to our previous results (Table 3). The expression of the three markers was counted as follows: one positive number (+ 1) for PD-L1 overexpression, as it was linked with a favorable prognosis, and one negative number (− 1) for OX40 and B7-H2 overexpression, both linked with unfavorable clinical outcomes. Then, the sum of the three immune markers was combined, and each TNBC patient was classified into a “tumor cells combined immune score—high (TCIS-H)” or a “tumor cells combined immune score—low (TCIS-L)” group, based on the cut-off of the relevant combined score (TCIS-H: ≥ − 1 and TCIS-L: <  − 1). In ICs, four immune markers (PD1, PD-L1, PD-L2, and OX40L) were included, based on our previous results (Table 3). The positive expression of all four immune markers was associated with a favorable PFS, thus each was counted as + 1. The sum of four immune markers was then calculated and each TNBC patient was classified within an “immune cells combined immune score—high (ICIS-H)” or an “immune cells combined immune score—low (ICIS-L)” group, based on the cut-off of the relevant combined score (ICIS-H: ≥ 1 and ICIS-L: < 1).

Of the 227 TNBC patients, 169 (74.4%) and 167 (73.6%) were classified as TCIS-H and ICIS-H, respectively (Additional file 5: Fig. S3). The clinicopathologic parameters according to the CIS groups are shown in Table 4. The TCIS-H group was significantly associated with a patient age 60 or older (p = 0.029), nuclear score 3 (p = 0.001), histologic grade 3 (p = 0.002), absence of ductal carcinoma in situ (p < 0.001), absence of lymphovascular invasion and lymph node metastasis (p = 0.003 and p = 0.017, respectively), and anatomic stage I & II (p < 0.001). The ICIS-H group was associated with a tumor size 2 cm or less (p = 0.037), histologic grade 3 (p = 0.011), absence of ductal carcinoma in situ (p = 0.011), absence of lymphovascular invasion (p = 0.019), and prognostic stage I & II (p = 0.031).

Of interest, both TCIS-H and ICIS-H were significantly associated with a longer PFS [(TCIS-H, 90.5% (mean 128.8 months) vs. TCIS-L, 70.7% (mean 110.8 months); p < 0.001) and ICIS-H group (ICIS-H, 90.4% (mean 130.0 months) vs. ICIS-L, 71.7% (mean 108.4 months); p < 0.001)] (Fig. 4A, B). On the other hand, only the ICIS-H group was associated with a longer OS compared to the ICIS-L group [ICIS-H, 95.8% (mean 135.4 months) vs. ICIS-L, 83.3% (mean 121.5 months); p = 0.001] (Fig. 4C, D).

As both TCIS and ICIS exhibited prognostic significance, we performed univariate and multivariable analysis along with other clinicopathological variables, using a Cox proportional hazard model, to determine whether TCIS and/or ICIS were independent prognostic predictors (Table 5; Fig. 5). Multivariate analysis identified both TCIS and ICIS as independent prognostic factors linked with PFS. More specifically, both TCIS-H and ICIS-H groups showed a significantly lower risk of progression compared with TCIS-L and ICIS-L groups [hazard ratio (HR) for TCIS = 0.381; 95% confidence interval (CI), 0.178–0.812; p = 0.012; HR for ICIS = 0.303; 95% CI 0.150–0.615; p = 0.001]. ICIS-H was also shown to be an independent factor predicting a lower risk of death compared with ICIS-L (HR = 0.229; 95% CI 0.086–0.609; p = 0.003).

To further validate our findings regarding the prognostic role of TCIS, we extracted and tested the mRNA expression levels of the relevant immune markers from the TCGA database. In total, data from 1,082 breast cancers, including 171 TNBC samples, were extracted. The samples were classified into TCIS-H and TCIS-L groups according to the sum of PD-L1, B7-H2, and OX40 gene expression scores, following our immunohistochemical analysis. In total, 942 breast cancers were grouped as TCIS-H and 140 as TCIS-L, while the TCIS-H group was significantly associated with a longer 5-year PFS (p < 0.001) and 5-year OS (p = 0.0018) (Fig. 6a, b). In the TNBC cohort, the TCIS-H group also showed a significant association with a longer 5-year PFS (p = 0.0033) and 5-year OS (p < 0.001) (Fig. 6c, d), being concordant with the findings derived from our immunohistochemical analysis. In contrast to TNBC, the TCIS-H group was not significantly associated with PFS and OS in the other breast cancer subtypes (Additional file 6: Fig. S4).