Concordance of programmed death-ligand 1 expression assessments determined via two immunohistochemical tests and the polymerase chain reaction method

Abstract

BACKGROUND

We previously demonstrated that the antibody against programmed cell death protein 1 ligand 1 (PDCD1 LG1) is a promising new marker of programmed death-ligand 1 (PD-L1) expression that correlates with both breast cancer (BC) clinicopathological characteristics and tumor sensitivity to chemotherapy. However, the concordance of PDCD1 LG1 expression scoring with immunohistochemical (IHC) tests approved for clinical use and with the polymerase chain reaction (PCR) method has not been previously studied.

AIM

To evaluate the concordance of methods for assessing PD-L1 expression, IHC tests with anti-PD-L1 (PDCD1 LG1) and anti-PD-L1 (SP142) antibodies and PCR.

METHODS

This prospective single-center observational cohort study included 148 patients with BC. PD-L1 expression in immune cells was assessed by the IHC method with anti-PD-L1 (PDCD1 LG1) and anti-PD-L1 (SP142) antibodies and by PCR. The concordance of PD-L1 scores between tests was assessed with positive percentage agreement (PPA) and negative percentage agreement (NPA). The strength of the agreement between the methods was calculated via the Cohen kappa index. P < 0.05 was considered statistically significant.

RESULTS

Regardless of the method used to assess marker expression, PD-L1 expression was significantly more often detected in patients with negative estrogen receptor status, human epidermal growth factor receptor-2-positive (HER2+) status, luminal B HER+ BC, nonluminal HER+ BC and triple-negative BC. PPA and NPA were 38.3% and 70.4%, respectively, for PD-L1 (PDCD1 LG1) and PD-L1 (SP142); 26.3% and 63.3%, respectively, for PD-L1 (PDCD1 LG1) and PD-L1 (PCR); and 36.5% and 74.4%, respectively, for PD-L1 (SP142) and PD-L1 (PCR). Cohen's kappa index for PD-L1 (PDCD1 LG1) and PD-L1 (SP142) was 0.385 (95%CI: 0.304–0.466), that for PD-L1 (PDCD1 LG1) and PD-L1 (PCR) was 0.207 (95%CI: 0.127–0.287), and that for PD-L1 (SP142) and PD-L1 (PCR) was 0.389 (95%CI: 0.309–0.469).

CONCLUSION

Thus, all three markers of PD-L1 expression are associated with the characteristics of aggressive BC, demonstrating moderate concordance between the tests.

Key Words: Breast cancer; Cohen kappa index; Negative percentage agreement; Positive percentage agreement; Programmed death-ligand 1; Programmed cell death protein 1 ligand 1

Core Tip: We determined the expression of programmed cell death ligand 1 in immune cells using immunohistochemistry with antibodies against programmed cell death protein 1 ligand 1 (PDCD1 LG1) and SP142 and polymerase chain reaction (PCR) and then examined the agreement between the results of three assays in patients with breast cancer (BC). All three tests were positively correlated with estrogen receptor-negative status, human epidermal growth factor receptor-2-positive (HER2+) status, and luminal B HER+, nonluminal HER+, and triple-negative BCs. Cohen's kappa index was 0.385 for PDCD1 LG1 and SP142 and 0.389 for SP142 and programmed death-ligand 1 (PCR), which is considered moderate agreement between the markers.

INTRODUCTION

Breast cancer (BC) remains a major medical, social and economic problem and is the leading cause of cancer mortality among the female population[1]. Analysis shows that in the coming decades, the burden of BC worldwide will only increase[2]. A feature of BC is the diversity of clinical, morphological and molecular genetic forms that require different approaches for diagnosis and treatment. Considering that BC is a systemic disease, the effectiveness of drug therapy largely determines the long-term results of treatment for this pathology. In this context, the search for new promising prognostic and predictive markers of BC has not lost its relevance.

One of the promising markers associated with the prognosis and sensitivity of BC to chemotherapy and immunotherapy is programmed death-ligand 1 (PD-L1)[3]. PD-L1 is a transmembrane protein expressed by tumor cells (TCs) and immune cells (ICs) and is involved in tumor evasion of immune surveillance[4,5]. The binding of PD-L1 to programmed cell death protein 1 (PD-1), which is located on the surface of activated T lymphocytes, natural killers, B lymphocytes, macrophages, dendritic cells and monocytes, leads to a decrease in the proliferation of PD-1-positive cells and the suppression of cytokine secretion, resulting in the inhibition of T-cell activation and eventually the induction of TC immune escape[4,5].

Currently, several markers, including anti-PD-L1 (SP142), anti-PD-L1 (SP263), and anti-PD-L1 (22C3) antibodies, have been approved for assessing PD-L1 expression in clinical practice. However, clinical trial data indicate low consistency in the results of PD-L1 expression assessment when different markers are used and inconsistent assessments of their prognostic and predictive significance[3,6,7]. In addition, in clinical practice, immune checkpoint (ICP) inhibitors that block the PD-L1/PD-1 pathway have unfortunately failed to meet expectations[8]. This is because PD-L1 expression in TCs and ICs is regulated by various mechanisms, including epigenetic regulation, the activation of various oncogenic pathways and transcription factors, and posttranscriptional regulation[9]. In addition, PD-L1 is present not only on the membranes of TCs but also in the cytoplasm, exosomes and nuclei of TCs, and the functions of PD-L1 are much broader than those of tumor evasion from the immune response[10-12]. On the basis of these data, it can be assumed that different types of antibodies can react with different PD-L1 “epitopes”, which are localized in different cellular components. The study of the intracellular distribution of the marker and its functions is still at the very beginning of the "path" and conceals many interesting "discoveries".

We previously demonstrated that the antibody against programmed cell death protein 1 Ligand 1 (PDCD1 LG1) is a promising new marker of PD-L1 expression that correlates with both BC clinicopathological characteristics and tumor sensitivity to chemotherapy[13-15]. However, the concordance of PD-L1 expression estimates between anti-PD-L1 (PDCD1 LG1) and markers approved for clinical use has not been studied. Therefore, the objectives of this study are (1) To determine the frequency of PD-L1 expression in ICs by immunohistochemical (IHC) using anti-PD-L1 (PDCD1 LG1) and anti-PD-L1 (SP142) antibodies and in tumor tissue by polymerase chain reaction (PCR); (2) To evaluate the frequency of PD-L1 (PDCD1 LG1), PD-L1 (SP142) and PD-L1 (PCR) expression in accordance with BC clinical and pathological characteristics; and (3) To establish the concordance of the results of PD-L1 expression assessment between these markers.

MATERIALS AND METHODS

Patient characteristics

This prospective single-center observational cohort study included 148 patients who underwent radical surgery for BC between January 15, 2023, and June 30, 2023, at the Orenburg Regional Clinical Oncology Dispensary (Orenburg, Russia). The age of the patients was 62.8 ± 26.0 years (median: 62.0 years). The study was performed in accordance with the 1964 Helsinki Declaration and internationally recognized guidelines. The study protocol was approved by Ethics Committee of Orenburg State Medical University. All patients signed an informed consent form to participate in this study. The inclusion and exclusion criteria for patients included in the study are presented in Table 1.

Table 1 Inclusion and exclusion criteria for patients in the study.

Inclusion criteria	Exclusion criteria
Stage breast cancer Tis-2N0-1M0	Chemotherapy, targeted, hormone therapy or radiation therapy before surgery
Radical surgical treatment (R0) at the first stage	Taking corticosteroids or nonsteroidal anti-inflammatory drugs
Axillary lymph node dissection at levels 1-2	Refusal to participate in the study
Voluntary informed consent to participate in a clinical trial

The BC stage was determined according to the 8^th TNM classification of malignant tumors. Data on patient age at surgery, T and N stages, tumor grade (G), estrogen receptor (ER) status, progesterone receptor (PR) status, human epidermal growth factor receptor 2 (HER2) status, the Ki67 index, the molecular biological subtype of BC, the presence of lymphovascular invasion (LVI), perineural invasion (PNI) and the intraductal component (IDC) were obtained from outpatient records and pathologists' reports. The distribution of patients included in the study according to clinicopathological characteristics and the molecular biological subtypes of BC is presented in Figure 1.

Figure 1 Distribution of patients with breast cancer. A: According to T and N stages and tumor grade; B: According to their molecular and biological characteristics. Abbreviations: ER: Estrogen receptor; HER2: Human epidermal growth factor receptor-2; PR: Progesterone receptor; TNBC: Triple negative breast cancer.

According to the presented data, the studied group predominantly included patients with T1c-T2 stages of BC (89.8%), N0-N1 stages (86.5%), tumor malignancy grades G2 and G3 (89.2%), ER-positive (ER+) status (89.2%), PR-positive (PR+) status (72.3%), HER2-negative (HER2-) tumor status (85.5%) and luminal A and luminal B HER2-BC (82.2%). LVI and PNI were detected in 56.8% and 77.7% of the tumor samples, respectively, and the IDC was noted in 41 (27.7%) BC samples. In 20 BC patients, the disease stage changed after surgery, as metastases were detected in 4 or more axillary lymph nodes.

Immunohistochemistry

IHC staining for PD-L1 was performed on 4-μm-thick sections from formalin-fixed paraffin-embedded tissue blocks according to the manufacturer’s instructions. The blocks were previously subjected to deparaffinization, dehydration with xylene, and rehydration in a graded series of alcohol solutions. Staining was performed on fully automated staining systems: (1) A BOND-MAX staining system (Leica Biosystems Melboume Pty Ltd., Mount Waverley, Australia) for antibodies against PDCD1 LG1 (1:100 dilution, Cloud-Clone Corp., Shanghai, China); and (2) A BenchMark XT IHC/ISH (Ventana Medical Systems) for antibodies against SP142 (ready to use dilution, Ventana Medical System, Oro Valley, AZ, United States). The visualization system included diaminobenzidine with hematoxylin. For the negative control sections, the primary antibodies were replaced with phosphate-buffered saline, and the samples were processed in the same way. Benign tonsil tissue was used as a positive (marker expression in lymphocytes and macrophages in the germinative centers and reticular epithelium of the crypts) and negative (no marker expression in the interfollicular areas and squamous epithelium) control.

PDCD1 LG1+ IC and SP142+ IC scoring

The expression of tumor-infiltrating ICs (lymphocytes, macrophages, dendritic cells, and granulocytes) was determined in 5 fields of view at 200 × magnification and calculated as the proportion of the tumor area (nonnecrotic and nonsclerotic areas) occupied by PDCD1 LG1-positive (PDCD1 LG1+) and SP142-positive (SP142+) ICs of any intensity. We used two assessment intervals for PDCD1 LG1+ ICs (< 10% and ≥ 10%) and two assessment intervals for SP142+ ICs (< 1% and ≥ 1%). Examples of the quantification of PDCD1 LG1+ IC and SP142+ IC scores are shown in Figure 2.

Figure 2 Assessment of tumor-infiltrating programmed cell death protein 1 ligand 1-positive and SP142+ immune cells. A: Programmed cell death protein 1 ligand 1-positive (PDCD1 LG1+) immune cell (IC) score < 10%; B: PDCD1 LG1+ IC score ≥ 10%; C: SP142 IC score < 1%; D: SP142+ IC score ≥ 1%. Staining with anti-PDCD1 LG1 antibodies (A and B), 200 ×; staining with anti-programmed death-ligand 1 antibodies (clone SP142)(C and D), 200 ×.

All samples were examined by two researchers (Senchukova MA and Saidler NV) who were blinded to the clinical and pathological data of the patients. Histological preparations were examined via light microscopy (Levenhuk D740T digital microscope connected to a 5.1 MP camera, Levenhuk, Russia).

Determination of PD-L1 expression by PCR

Tumor tissue for PCR analysis was also collected within 30–40 minutes after removal of the surgical sample, placed in Eppendorf tubes (2 mL), and stored at -40 °C until analysis. Total cellular RNA was isolated via RNA-Extran reagent (Synthol, Russia) according to the manufacturer's instructions. Next, cDNA was synthesized via an OT-1 reverse transcription kit (Synthol, Russia) containing a buffer with oligonucleotides, the MMLV-RT enzyme, and the Random-6 and Oligo-15 primers. The reverse transcription conditions were as follows: (1) 37 °C for 30 minutes; and (2) 92 °C for 5 minutes. Real-time PCR was performed on a DTLite device (DNA-Technology, Russia) with the obtained cDNA and a 2.5 × reaction mixture for RT-PCR in the presence of EVA Green dye. The RT-PCR conditions were as follows: (1) 95 °C for 5 minutes; (2) 95 °C for 15 seconds; and (3) 62 °C for 45 seconds for 55 cycles. To detect a specific product (gene expression), the following primer sequence was used: (1) Forward, 5’-GGCATTTGCTGAACGCAT-3’; and (2) Reverse, 5’- CAATTAGTGCAGCCAGGT-3’.

Statistical analysis

Statistical processing of the results was performed with Statistica 12.0 software. Correlations between different indicators were estimated via nonparametric Spearman's rank correlation. χ² tests were performed to analyze differences in distribution between categorized data. To plot the receiver operating characteristic (ROC) curve and measure the area under the curve (AUC), software, which is available free, was used. The best threshold (cutoff) values were determined by the largest Youden index (J = sensitivity + specificity - 1). The effectiveness of the predictive model was assessed by the AUC. The concordance of PD-L1 scores between tests was assessed with positive percentage agreement (PPA) and negative percentage agreement (NPA). The strength of the agreement between the methods was calculated with the Cohen kappa index. P < 0.05 was considered statistically significant.

RESULTS

Specificities of PD-L1 (PDCD1 LG1) and PD-L1 (SP142) expression in benign tonsil tissue

When tonsil tissue was stained with anti-PD-L1 (PDCD1 LG1) antibodies, diffuse cytoplasmic expression of the marker was observed in most ICs. The expression of the marker in the squamous epithelium was absent. However, in the germinative centers and reticular epithelium of the crypts, cells with pronounced hyperexpression of the marker were observed. The number and localization of these cells corresponded to the number and localization of cells expressing PD-L1 (SP142) in tonsil tissue. It can be assumed that cells with PD-L1 (PDCD1 LG1) hyperexpression and cells expressing PD-L1 (SP142) have similar characteristics, but their histopathology requires clarification. Figure 3 shows the results of staining tonsil tissue with anti-PD-L1 (PDCD1 LG1) and anti-PD-L1 (SP142) antibodies.

Figure 3 Programmed death-ligand 1 expression in benign tonsil tissue. A: Staining with anti-programmed cell death protein 1 Ligand 1 antibody, 200 ×; B: Staining with anti-programmed death-ligand 1 (clone SP142) antibody, 200 ×.

ROC analysis results

ROC analysis was performed to determine the optimal cutoff point for the PD-L1 (PDCD1 LG1) score corresponding to a negative PD-L1 (SP142) status (< 1%). We used four assessment intervals for PD-L1 (PDCD1 LG1) scores in ICs: (1) < 1%; (2) ≥ 1% and < 10%; (3) ≥ 10% and < 30%; and (4) ≥ 30%. According to the obtained data, a PD-L1 (PDCD1 LG1) IC score < 10% (cutoff) corresponded to a negative status of PD-L1 expression in ICs assessed with an anti-PD-L1 (SP142) antibody. The AUC was 0.743 (95%CI: 0.648–0.843, P < 0.0001). The sensitivity and specificity of the method were 63.9% and 79.3%, respectively. Figure 4 shows the ROC curve discriminating PD-L1 (PDCD1 LG1) scores corresponding to positive and negative PD-L1 (SP142) statuses (< 1% and ≥ 1%).

Figure 4 Receiver operating characteristic curve discriminating programmed cell death protein 1 Ligand 1 scores corresponding to positive and negative programmed death-ligand 1 (clone SP142) status (< 1% and ≥ 1%). AUC: Area under the curve.

Frequency of PD-L1 expression, depending on the method of marker determination and the clinical and pathological characteristics of BC

Data on the frequency of PD-L1 expression, depending on the method of marker determination and the clinical and pathological characteristics of BC patients, are presented in Table 2.

Table 2 Frequency of programmed death-ligand 1 expression depending on marker determination methods and the clinical and pathological characteristics of breast cancer patients.

BC characteristics	Programmed cell death protein 1 Ligand 1 score				SP142 score				Polymerase chain reaction test				P value or χ² test
	< 10%		≥ 10%		< 1%		≥ 1%		Negative		Positive
	n	%	n	%	n	%	n	%	n	%	n	%
T stage
Тin situ	2	50.0	2	50.0	4	100	0	0	4	100	0	0	0.0211, 0.262, 0.0053
Т1в	4	36.4	7	63.6	6	54.6	5	45.4	4	36.4	7	63.6
Т1с	59	66.3	30	33.7	68	76.4	21	23.6	76	85.4	13	14.6
Т2	36	81.8	8	18.2	34	77.3	10	22.7	29	65.9	15	34.1
N stage
N0	58	71.6	23	28.4	60	74.1	21	25.9	61	75.3	20	24.7	0.701, 0.432, 0.733
N1	31	66.0	16	34.0	36	76.6	11	23.4	36	76.6	11	23.4
N2-3	12	60.0	8	40.0	16	80.0	4	20.0	16	80.0	4	20.0
Tumor grade
G1	10	62.5	6	37.5	14	87.5	2	12.5	12	75.0	4	25.0	0.251, 0.000152, 0.823
G2	64	73.6	23	26.4	74	85.1	13	14.9	68	78.2	19	21.8
G3	27	60.0	18	40.0	24	53.3	21	46.7	33	73.3	12	26.7
Lymphovascular invasion
Absent	60	71.4	24	28.6	66	78.6	18	21.4	64	76.2	20	23.8	0.341, 0.352, 0.963
Present	41	64.1	23	35.9	46	71.9	18	28.1	49	76.6	15	23.4
Perineural invasion
Absent	79	68.7	36	31.3	84	73.0	31	27.0	84	73.0	31	27.0	0.821, 0.162, 0.083
Present	22	66.7	11	33.3	28	84.9	5	15.1	29	87.9	4	12.1
Intraductal component
Absent	73	68.2	34	31.8	78	72.9	29	27.1	85	79.4	22	20.6	0.991, 0.202, 0.153
Present	28	68.3	13	31.7	34	82.9	7	17.1	28	68.3	13	31.7
Estrogen receptor status
Negative	5	31.3	11	68.7	0	0	16	100	10	62.5	6	37.5	0.00081, 0.00002, 0.173
Positive	96	72.7	36	27.3	112	84.9	20	15.1	103	78.0	29	22.0
Progesterone receptor status
Negative	25	61.0	16	39.0	16	39.0	25	61.0	27	65.8	14	34.2	0.241, 0.00002, 0.0633
Positive	76	71.0	31	29.0	96	89.7	11	10.3	86	80.4	21	19.6
HER2 status
Negative	99	77.3	29	22.7	112	87.5	16	12.5	103	80.5	25	19.5	0.00001, 0.00002, 0.0033
Positive	2	10.0	18	90.0	0	0	20	100	10	50.0	10	50.0
BC subtypes
Luminal A	42	75.0	14	25.0	50	89.3	6	10.7	50	89.3	6	10.7	0.00001, 0.00002, 0.00053
Luminal B HER2-	54	81.8	12	18.2	62	93.9	4	6.1	50	75.8	16	24.2
Luminal B HER2+	0	0	10	100	0	0	10	100	3	30.0	7	70.0
Nonluminal HER2+	2	20.0	8	80.0	0	0	10	100	7	70.0	3	30.0
Triple negative BC	3	50.0	3	50.0	0	0	6	100	3	50.0	3	50.0

¹Programmed cell death protein 1 Ligand 1 score.

²SP142 score.

³Polymerase chain reaction test.

BC: Breast cancer; HER2: Human epidermal growth factor receptor-2.

According to the data presented, regardless of the method of marker determination used, PD-L1 expression in ICs was significantly more often detected in patients with negative ER status, HER2+ status, luminal B HER+ BC, nonluminal HER+ BC and triple negative BC (TNBC). Interestingly, marker expression was also detected more often in the T1b stage of invasive BC than in the T1c and T2 stages. In ductal carcinoma in situ, PD-L1 (PDCD1 LG1) expression was positive in two of the four BC samples, and PD-L1 (SP142) and PD-L1 (PCR) expression was negative in all samples. However, the small number of patients in the Tin situ (4 patients) and T1b (11 patients) groups does not allow us to consider these conclusions sufficiently reliable. The frequency of PD-L1 (SP142) expression was greater in patients with a tumor grade 3 and negative PR status, but the frequency of PD-L1 (PDCD1 LG1) and PD-L1 (PCR) expression did not depend on these BC characteristics. The Ki67 indices for negative and positive PD-L1 statuses were 23.6% ± 13.0% and 51.7% ± 29.0%, respectively, for PD-L1 (SP142) (P < 0.00001); 25.1% ± 14.8% and 41.9% ± 29.1χ%, respectively, for PD-L1 (PDCD1 LG1) (P = 0.002); and 25.1% ± 15.2% and 47.7% ± 29.9%, respectively, for PD-L1 (PCR) (P = 0.00005). No differences were found in the frequency of PD-L1 expression of all three markers depending on N stage, the presence of LVI and PNI, or the IDC.

Concordance between the results of PD-L1 expression assessment using immunohistochemistry and PCR

The expression of PD-L1 (PDCD1 LG1), PD-L1 (SP142) and PD-L1 (PCR) was detected in 47 (31.8%), 35 (23.6%) and 36 (24.3%) BC samples, respectively. Figure 5 demonstrate the frequency of expression of the markers and their combinations.

Figure 5 Frequencies of programmed death-ligand 1 expression in immune cells according to marker determination methods. A: Frequencies of programmed death-ligand 1 (PD-L1) expression for all three markers and their combinations; B: Frequencies of programmed cell death protein 1 Ligand 1 (PDCD1 LG1) and PD-L1 (clone SP142) expression and their combination; C: Frequencies of PDCD1 LG1 and PD-L1 [polymerase chain reaction (PCR)] expression and their combination; D: Frequencies of PD-L1 (clone SP142) and PD-L1 (PCR) expression and their combination. PCR: Polymerase chain reaction; PDCD1 LG1: Programmed cell death protein 1 Ligand 1; PD-L1: Programmed death-ligand 1.

The Spearman rank order correlation between PD-L1 (PDCD1 LG1) and PD-L1 (SP142) was 0.391 (P < 0.0001), that between PD-L1 (PDCD1 LG1) and PD-L1 (PCR) was 0.201 (P = 0.014), and that between PD-L1 (SP142) and PD-L1 (PCR) was 0.389 (P < 0.0001). PPA and NPA for PD-L1 (PDCD1 LG1) and PD-L1 (SP142), PD-L1 (PDCD1 LG1) and PD-L1 (PCR) and PD-L1 (SP142) and PD-L1 (PCR) were 18.1% and 56.3%, 38.3% and 70.4%, 26.3% and 63.3%, and 36.5% and 74.4%, respectively. Cohen's kappa index for PD-L1 (PDCD1 LG1) and PD-L1 (SP142) was 0.385 (CI: 0.304–0.466); for PD-L1 (PDCD1 LG1) and PD-L1 (PCR), it was 0.207 (95%CI: 0.127–0.287); and for PD-L1 (SP142) and PD-L1 (PCR), it was 0.389 (95%CI: 0.309–0.469).

DISCUSSION

The effectiveness of BC treatment depends on the accuracy of the disease prognosis and the effectiveness of the therapy. Ineffective treatment involves not only the absence of a result but also the negative consequences for the health of patients. BC is characterized by a variety of clinical, morphological and molecular genetic forms that differ not only in the prognosis of the disease but also in sensitivity to drugs and radiotherapy. In this context, the search for new prognostic and predictive markers of BC has not lost its relevance.

The immune system plays a decisive role in the occurrence and progression of malignant neoplasms. In addition, the severity and cellular composition of the immune infiltrate of the tumor stroma significantly affect the effectiveness of antitumor radiation and drug therapy and, consequently, the prognosis of the disease[16-18]. However, the cellular composition of tumor-associated ICs is very diverse[18,19]. Some ICs are involved in the activation of tumor progression factors, whereas others, on the contrary, have antitumor activity[20-22]. A special place in the IC population is occupied by cells expressing PD-L1. PD-L1 expression on TCs is associated with immune escape mechanisms, but the role of PD-L1 expression in ICs is not clear. For example, a high density of PD-L1-positive ICs in the tumor stroma is associated with more aggressive BC characteristics, including negative ER and PR statuses, HER+ BC, TNBC and a high Ki67 index[23,24]. However, high infiltration of the tumor stroma with ICs, including PD-L1+ ICs[22,25,26], and a high density of PD-L1+ intratumoral macrophages[27] are associated with a better prognosis of the disease and response to chemotherapy and chemoimmunotherapy in patients with TNBC.

The discovery of ICP inhibitors has given hope for significant breakthroughs in the treatment of the most aggressive malignancies, including TNBC. Unfortunately, the results of the use of ICP inhibitors in BC patients have not been convincing enough, and the advisability of their use in this disease continues to be debated[8,28]. The lack of effect on treatment with ICP inhibitors may be due to insufficient understanding of the role of PD-L1 in tumor progression and the lack of reliable markers that allow stratification of patients into groups for whom therapy with ICP inhibitors is indicated and those who will not benefit from this treatment. Accordingly, the search for new reliable markers of PD-L1 expression has not lost its relevance.

Previously, we examined the expression of PD-L1 (PDCD1 LG1), a new marker of PD-L1 expression, and showed that its expression in TCs and ICs is associated with BC clinical and pathological characteristics and tumor sensitivity to chemotherapy[13,14]. Unlike antibodies approved for clinical use, including SP142, 22C3, SP263, 22C3, 28-8 and other antibodies, which are characterized by membrane expression of the marker in TCs, PD-L1 (PDCD1 LG1) is expressed in the cytoplasm of TCs. We explain the differences in the expression of PD-L1 markers by the fact that different antibodies can visualize different PD-L1 epitopes localized in different cellular structures. This assumption is supported by data from recent studies indicating that PD-L1 can be localized not only on cell membranes but also in the cytoplasm, exosomes and even cell nuclei and that the functions of PD-L1 Localized in different parts of the cell can differ significantly[10,11].

Despite the potential prognostic and predictive value of this new marker, the comparability of PD-L1 (PDCD1 LG1) expression estimates and markers approved for clinical use has not been practically studied. Therefore, we compared the results of PD-L1 (PDCD1 LG1), PD-L1 (SP142) and PD-L1 (PCR) expression assessments. Given that antibodies against PD-L1 (PDCD1 LG1) stain the cytoplasm of TCs and that antibodies against PD-L1 (SP142) stain the membrane of TCs, comparisons of the expression of markers in TCs would be incorrect. For this reason, we did not compare the results of PD-L1 expression assessment in TCs and focused on ICs.

We first compared the results of PD-L1 (PDCD1 LG1) and PD-L1 (SP142) expression in benign tonsil tissue, which was used as a positive control. We found that the PDCD1 LG1 antibody diffusely stains the cytoplasm of most ICs. However, in the background of diffuse staining, in the germinal centers and crypt reticular epithelium, we found that cells overexpressing PD-L1 (PDCD1 LG1) were similar to cells expressing PD-L1 (SP142) in the tonsils. We hypothesize that cells overexpressing PD-L1 (PDCD1 LG1) and cells expressing PD-L1 (SP142) in the tonsils may have similar characteristics and perform the same functions. However, this hypothesis needs to be verified with double IHC staining to exclude or confirm the coexpression of markers.

We subsequently performed ROC analysis and determined that PD-L1 (PDCD1 LG1) scores in ICs < 10% and ≥ 10% corresponded to PD-L1 (SP142) negative (< 1%) and positive (≥ 1%) statuses in ICs. We compared the frequency of PD-L1 (PDCD1 LG1), PD-L1 (SP142) and PD-L1 (PCR) expression with the clinicopathological characteristics of BC and found that the PD-L1-positive status for all three markers was significantly more often observed in the negative ER status, HER2+ status, and in Luminal B HER+ BC, nonluminal HER+ BC and TNBC. In addition, the Ki67 index was significantly greater in patients with positive PD-L1 status for all three markers. These results are consistent with the results of most researchers who have used other antibodies to assess PD-L1 expression. Thus, in the study of Ni et al[23], a positive status of PD-L1 (SP142) expression was associated with negative ER and PR statuses and HER2+ BC and positively correlated with the Ki67 index. PD-L1 (SP142) expression was also associated with better disease-free survival and BC-specific survival[23]. In another study, the Ki67 index was positively correlated with PD-L1 expression, but this dependence was noted only for clones 28-8 and E1 L3N but not for SP142. In multivariate analysis, among the three markers (SP142, 28-28 and E1 L3N), only negative PD-L1 status in the ICs for clones 28-8 was associated with decreased overall survival[29]. Cha et al[3] reported that negative PD-L1 (22C3) status was significantly associated with decreased recurrence-free survival and distant metastasis-free survival in patients with BC. However, PD-L1 (SP142) expression and PD-L1 (SP263) expression were not associated with disease prognosis.

In the study by Yazaki et al[30], the PD-L1 status in ICs was examined using SP142, SP263, and 22C3 antibodies. A positive PD-L1 status in early BC was significantly more often observed in patients with a histological grade 3 and HER2+ BC. However, this dependence was observed only with a high density of tumor-infiltrating lymphocytes (TILs). With a low TIL density, no differences were observed between the patient groups. Notably, in this study, a positive PD-L1 status in ICs was associated with decreased survival in patients with early BC[30].

Some researchers noted an association between positive PD-L1 (SP142) status in ICs and the absence of metastases in the lymph nodes but did not find an association with tumor grade[6]. In our study, only the PD-L1 (SP142)-positive status was correlated with high-grade tumors. The correlations of the studied markers with the risk of regional metastasis have not been established. However, our study predominantly involved patients with early-stage BC, which, of course, must be taken into account when the results are interpreted.

Interestingly, in some studies, the prognostic significance of PD-L1 expression in BC was assessed via a combined score[31,32]. We believe that this approach is not optimal because PD-L1 expression in TCs is more often associated with decreased patient survival[33-35], whereas PD-L1 expression in ICs is associated with better survival of BC patients[23,36,37].

An important stage of our study was the assessment of the consistency of the results of determining PD-L1 expression in ICs with antibodies against PD-L1 (PDCD1 LG1) and PD-L1 (SP142) and in tumor tissue by PCR. Although all three markers correlated with each other, the consistency of the PD-L1 expression estimates between the markers was quite low. The highest consistency in the assessment of markers was noted between PD-L1 (PDCD1 LG1) and PD-L1 (SP142) and between PD-L1 (SP142) and PD-L1 (PCR). The NPA and PPA were 70.4% and 38.3% for PD-L1 (PDCD1 LG1) and PD-L1 (SP142) and 74.4% and 36.5% for PD-L1 (SP142) and PD-L1 (PCR), respectively. Cohen's kappa index was 0.385 (95%CI: 0.304–0.466) for PD-L1 (PDCD1 LG1) and PD-L1 (SP142) and 0.389 (95%CI: 0.309–0.469) for PD-L1 (SP142) and PD-L1 (PCR), which is considered moderate agreement between these markers.

Our results are consistent with other studies showing low agreement in PD-L1 expression assessments using different antibody clones[38,39]. Sun et al[29] examined PD-L1 expression in patients with TNBC using three antibodies (E1 L3N, 28-8, and SP142). The authors used four cutoff values for TCs (1%, 5%, 10%, and 50%) and three cutoff values for marker expression in ICs [negative; low-positive, stained ICs ≤ 30/high power field (HPF); and high-positive, stained ICs > 30/HPF]. The kappa values were highest for clones 28-8 and E1 L3N: For TCs, the kappa values were 0.660 and 0.997 for the 10% cutoff and 50% cutoff, respectively; for ICs, the kappa value was 0.607. For clones 28-8 and SP142 and for SP142 and E1 L3N, the kappa values were significantly lower, especially for ICs. For antibodies 28-8 and SP142, the kappa value was 0.309, and for antibodies SP142 and E1 L3N, the kappa value was 0.305[29].

A study of the concordance of PD-L1 expression assessments in ICs between the SP142, 22C3 and SP263 assays revealed that the concordance between the SP142 and SP263 assays was better than that between the SP142 and 22C assays. The kappa value for the SP142 and SP263 assays was 0.831, and for the SP142 and 22C3 assays, it was 0.114[40]. Other studies have reported low concordance of PD-L1 expression estimates between SP142, 22C3 and SP263[6,7,24]; between SP263 and SP142[41]; and between SP142 and 22C3[42].

CONCLUSION

Thus, the study of PD-L1 expression in ICs by IHC with antibodies against PD-L1 (PDCD1 LG1) and PD-L1 (SP142) and in tumor tissue by PCR revealed that, regardless of the method used for marker determination, PD-L1 expression is significantly more often detected in patients with an ER-negative status, HER2+ status, and in Luminal B HER+ BC, nonluminal HER+ BC and TNBC. The highest agreement in the assessment of PD-L1 expression was noted between PD-L1 (PDCD1 LG1) and PD-L1 (SP142) and between PD-L1 (SP142) and PD-L1 (PCR), and taking into account Cohen's kappa index, the agreement between these markers was assessed as moderate. The differences in the frequency of PD-L1 expression between markers may be because different antibodies and methods can visualize and detect different PD-L1 epitopes localized in different cell structures. The features of the intracellular distribution of PD-L1 and its role in tumor progression are of undoubted interest for future research. This study has limitations related to the small sample size, single-center nature of the study, predominance of patients with early BC and luminal BC subtypes, and insufficient number of patients with TNBC and HER2+ BC. We believe that further study of the features of PD-L1 (PDCD1 LG1) expression in TCs and ICs and the relationships of this marker with BC clinicopathological characteristics and disease prognosis will contribute to a better understanding of the role of PD-L1 in tumor progression. Studies performed in different scientific and clinical centers and on larger cohorts of patients will help improve the accuracy and reliability of the method and justify the possibility of its use in clinical practice.