Introduction
Programmed death-1 (PD-1) is an immune checkpoint that is able to inhibit the immune system by binding to its ligand programmed death-ligand 1 (PD-L1), and prohibits T cell function in this way. In literature it has been described that in different cancer types [
1], especially immunogenic tumors, including non-small cell lung cancer [
2], malignant melanoma [
3], and ovarian cancer [
4], PD-1 can be upregulated on tumor infiltrating lymphocytes (TILs) and PD-L1 on tumor and immune cells. These markers could have potential prognostic and/or predictive values [
5]. Although breast cancer is not considered very immunogenic, expression of PD-1 and PD-L1 has been described and prognostic and predictive values have been suggested, albeit with varying results [
6‐
17].
Most studies focused on the expression of PD-1 and/or PD-L1 in primary breast tumors. However, through selective metastases of subclones, or tumor progression, expression levels and prognostic values in metastases may differ from the primary tumors. For example, previous research has described receptor conversion for ER, PR, HER2 and several molecular imaging targets in a significant proportion of breast cancer metastases [
18,
19]. Additionally, it has been observed that conversion from ER/PR-positive primary breast cancers to ER/PR-negative metastases is associated with negative prognosis [
20].
Since biomarker expression status of distant metastases, rather than the primary tumor will dictate response to targeted therapy, it is important to know whether conversion from the primary tumor to distant metastases also occurs for PD-1 and PD-L1. Two previous studies described high concordance of PD-L1 expression in primary breast tumors and their matched lymph node metastases, and lower PD-L1 expression levels in non-matched distant metastases, respectively [
21,
22]. One study comparing PD-L1 expression between metastatic and non-metastatic primary breast cancers described one case where PD-L1 expression in primary tumor and metastasis was concordant, while in the second case expression was discordant [
23]. As far as we know, these previous studies have only evaluated PD-L1 expression, while PD-1 expression may be of importance as well, certainly since brain metastases, compared to their primary breast tumor, are thought to have a lower amount of TILs [
24].
Given the limited data on PD-1 expression and the anecdotal data on PD-L1 expression in primary breast cancers compared to their matched distant metastases to date, we compared immunohistochemical expression levels of PD-1 and PD-L1 in primary tumors and their matched distant metastases in a large group of metastatic breast cancer patients, and evaluated prognostic values.
Materials and methods
Patients and samples
For previous research, tissue microarrays (TMAs) with 106 primary female breast cancers and their matched distant metastases from various distant anatomical sites (bone, brain, gastrointestinal, gynecological (uterus/ovary), liver, lung/pleural and skin/subcutis) had been assembled [
25,
26]. All metastases were metachronous metastases. Clinicopathological data (age, tumor size, histology, grade (according to the modified Bloom and Richardson score) [
27], lymph node status, and surrogate molecular subtype, based on hormone receptor status—the ER/PR we used was, according to the Dutch guidelines with a cut-off of 10%, and HER2 was scored according to the ASCO/CAP guidelines) were anonymously extracted from digital reports/revised by a pathologist specialized in breast cancer (PvD). Follow-up data were anonymously obtained through the Comprehensive Cancer Center of The Netherlands (IKNL). Use of anonymous or coded left over material for scientific purposes is part of the standard treatment agreement with patients and therefore ethics approval and informed consent procedure was not required according to Dutch legislation [
28].
Immunohistochemistry
The TMAs contained three cores of 0.6 mm from representative areas in formalin-fixed paraffin-embedded (FFPE) tissue blocks of both primary tumor and metastasis. 4 µm thick tissue sections were cut and immunohistochemically stained with the Ventana autostainer (Roche, Tuscon, USA). For PD-1, we used a mouse anti-PD-1 monoclonal antibody (ab52587 (NAT105, dilution 1:50, Abcam, Cambridge, UK)) and for PD-L1, a rabbit anti-PD-L1 monoclonal antibody (741–4905 (clone sp263, dilution Ventana ready to use; Ventana Medical Systems, Tuscon, Arizona, USA)). ER (Ventana 1:100), PR (Dako 1:100) and HER2 (Neomarkers 1:100) were also stained with the Ventana autostainer.
Assessment of PD-1 and PD-L1
The immunohistochemically stained slides were scored by consensus of two experienced observers (WS or QM and PvD), in random order and blinded to other data. PD-1 was scored positive in case of any membranous staining of immune cells. For PD-L1 we scored the positivity on both tumor cells and immune cells. Concerning PD-L1 on tumor cells, we scored the percentage of positive membranous staining. To create balanced groups for analyses, all cases with scores above 0 were considered to be positive. Immune PD-L1 was considered to be positive in case of any membranous staining of immune cells. Additionally, we verified the use of TMAs by comparing the PD-L1 scores of ten patients with PD-L1 scores on whole slides. In five cases, we observed complete agreement, and for the other five we only observed minimal differences.
Statistics
Statistical analyses were performed with IBM Statistics 25. To study differences in hormone receptor expression and expression of PD-1 and PD-L1 between primary tumors and (matched) metastases, McNemar and Pearson Chi square tests were used. To compare expression of PD-1 and PD-L1 in metastases and their anatomical sites, we used Fisher’s exact test. For survival analysis, Kaplan–Meier curves were plotted and compared by log-rank test. Multivariate survival analysis was performed with Cox regression analyses. For all statistical tests, the significance level was set at p value < 0.05.
Discussion
The aim of our study was to compare expression levels of PD-1 on immune cells, and PD-L1 on tumor cells and immune cells, in primary tumors and their matched distant metastases in a large group of breast cancer patients, and to evaluate prognostic values.
We observed that PD-1 and PD-L1 tumor and immune expression were concordant in the distant metastasis and its primary breast tumor in about half to two-third of the patients. In the other part of the patients, PD-1 or PD-L1 negative tumors had developed PD-1 or PD-L1 positive metastases or vice versa. Cimino-Mathews et al. also observed such discordance, but only in two patients and solely for the PD-L1 positive primary tumors [
23]. Other studies in breast cancer observed concordance in expression of PD-L1 for primary breast tumors and matched (nodal) metastases in 94% and 100% [
22,
29], which might be due to the relatively small size of their cohorts (15 and 17 matched cases, respectively). No previous studies that have been executed, describe PD-1 expression in breast cancer compared to matched distant metastases. Schneider et al. observed high concordance of PD-1 expression in primary head and neck squamous cell carcinoma and matched lymph node metastases (90%), while a slightly lower concordance percentage was found for PD-L1 (70%) [
30]. Additionally, we observed that if PD-1 or PD-L1 was discordant in the distant metastasis compared to the primary tumor, this was not necessarily the case for the other protein. As far as we know we are the first to observe this inconsistency. The underlying biology for this discrepancy is yet unknown.
Possible explanations for discrepancies between primary tumors and their matched metastases may be genomic evolution during tumor progression [
31], or clonal selection during metastatic process [
18]. Otherwise adjuvant (chemo)therapy might influence the tumor microenvironment in breast cancer, and within that process, PD-1 or PD-L1 expression might change. This has been observed before by Yoon et al. where PD-L1 expression levels changed in 4/14 patients treated with neoadjuvant anthracycline- and/or taxane-based chemotherapy [
32]. Additionally, Pelekanou et al. compared both TILs and PD-L1 expression in tumors of 58 patients pre- and post-neoadjuvant chemotherapy, and observed an increase of TILs, but a decrease of PD-L1 expression in the post-neoadjuvant therapy tumors [
33]. According to these results it may be interesting for future studies to look into more detail into the role of adjuvant therapy in relation to PD-1 and PD-L1 expression conversion in distant metastases.
As we compared PD-1 and PD-L1 expression in different distant anatomic sites, we observed some expression differences between those anatomic sites, in particular for PD-1 and immune PD-L1, although numbers were too small for proper statistical analysis. For example, gynaecological metastases less frequently showed PD-1 expression, while all four bone metastases were negative for PD-1. Previous studies, both in breast cancer as in other cancer types, evaluated PD-L1 expression in lymph node or clustered distant metastases [
21,
22,
30,
34], which impedes comparison of our observations. However, Ogiya et al. studied PD-L1 expression in brain metastases only, and observed comparable results with regard to PD-L1 expression frequency as we did (45% vs. 41%) [
24].
Furthermore, in subgroup analysis, we found a correlation between PD-1 positivity and higher tumor grade in primary breast tumors, which is in line with literature [
8,
14,
16]. We did not observe other associations between PD-1 expression and selected clinicopathological variables, both in primary breast tumors as in metastases. Concerning PD-L1 in primary breast tumors, we observed associations of immune PD-L1 with ER negativity and a tendency of a correlation with grade 3 tumors, and for tumor PD-L1 an association with triple negative breast cancer. These findings are in line with the literature, since PD-L1 positivity is often associated with worse clinicopathological characteristics like larger tumor size, higher tumor grade and ER and PR negativity [
7,
9‐
11,
13,
15]. The relatively small size of our cohort for subgroup analyses might explain we only observed these associations in primary tumors and not in metastases.
In survival analysis we did not observe significant differences concerning OS for PD-1 and PD-L1 on both tumor and immune cells. As far as we know, no other study has assessed prognostic value of PD-1 and PD-L1 expression in metastases. In primary breast tumors, the prognostic value of PD-1 and PD-L1 is equivocal, since previous studies described different outcomes for both PD-1 [
14,
16], as PD-L1 [
6,
7,
10,
11,
15]. Furthermore, a significantly better OS for patients with PD-L1 negative tumors that developed PD-L1 positive metastases was observed in the present study. However, subgroups were small, so we do not want to speculate on the biological background of this observation. Further evaluation in bigger groups could be interesting.
Comparison of PD-1 and PD-L1 expression between different studies remains difficult due to use of different antibodies, scoring methods and use of different cut-off values. For example, cut-off values varying from 1 to 50% were used, leading to different percentages of positive tumors [
17]. Additionally, PD-L1 is expressed on both tumor cells and immune cells. Whereas most studies evaluate PD-L1 on tumor cells, recent studies also considered PD-L1 expression on immune cells [
10,
17], like we did. Therefore, further standardization of (scoring) methods for the use of PD-1 and PD-L1 seems to be important.
Since PD-1 and PD-L1 expression are thought to be heterogeneous, it might be a limitation that we used TMAs to evaluate expression instead of full tissue slides. We therefore used three cores per tissue block that represented different tumor areas. Additionally, we verified the use of TMAs by comparing these PD-L1 scores of ten patients with PD-L1 scores on whole slides. In half of the patients complete agreement was noticed, while the other half only showed minimal differences. The strength of our study lays in the relatively large group of primary tumors and matched distant metastases of breast cancer patients.
To conclude, PD-1 and PD-L1 (on tumor and immune cells) positivity is concordant between primary tumors and distant metastases in only half to two-third of the breast cancer patients. In the other part of the patients PD-1 or PD-L1 negative tumors developed PD-1 or PD-L1 positive metastases or vice versa. Furthermore, patients with immune PD-L1 negative breast tumors that develop PD-L1 positive distant metastases seem to have the best overall survival. This illustrates the need of reassessing PD-1 and tumor and immune PD-L1 expression in biopsies of distant metastases to optimize the usefulness of these biomarkers.