Background
Sarcomas are malignant mesenchymal tumors that account for approximately 1% of adult solid cancers [
1]. Sarcomas can be divided into more than 50 distinct histological subtypes, and many of these subtypes are not limited to a specific location of the body [
2]. Due to its rarity and morphological variability, the clinical and pathological study of sarcomas have been limited, and the mainstay of sarcoma treatment has not changed for decades [
3,
4]. Surgical resection with enough safety margins remains as the only curative therapeutic option despite its limited indication and several complications. In cases with inoperable tumors, doxorubicin and ifosfamide have been used for more than 30 years and remain the mainstay for treatments. However, these cytotoxic agents are known to provide overall response rates of only about 25% in the first-line setting and are currently used for palliative, but not curative, purposes [
5]. There is still a need for new treatment methods that surpass previous therapies.
After the discovery of programmed cell death protein 1 (PD-1) in 1992, PD-1 and PD-L1 have been revealed to have a fundamental role in cancer immune surveillance [
6,
7]. Anti-PD-1 therapies were approved for melanoma, non-small-cell lung cancer, and various solid tumors worldwide [
8‐
10]. Also in cases of sarcoma, several previous studies have reported that more than 50% of sarcomas, including leiomyosarcoma, dedifferentiated liposarcoma (DDLPS), undifferentiated pleomorphic sarcoma (UPS), osteosarcoma, epithelioid sarcoma, and other sarcomas, showed PD-L1 expression in tumor cells [
11‐
13]. However, recently the SARC028 trial report described that only 4% (3/70) of the sarcoma biopsy samples (all three were from patients with UPS) were immunopositive for PD-L1 in more than 1% of tumor cells [
14]. This study also reported that 11% (9/80) of the patients with sarcomas showed an objective response, especially in patients with undifferentiated sarcomas (4/10) and liposarcomas (2/10). This result is promising but also resulted in several new questions regarding the PD-L1 immunohistochemical expression rate and its role in practice.
To gain insight into the PD-L1 expression pattern in various patients with sarcomas, we examined the PD-L1 expression using various cell lines and patient tissues including both TMA and whole sections and evaluated the association between PD-L1 expression and clinicopathological features in patients with sarcomas.
Methods
Patient tissue specimens
A total of 230 archival formalin-fixed paraffin-embedded (FFPE) soft tissue sarcoma tissue samples, each from a different patient, were collected at the Samsung Medical Center in Seoul, Korea. Ten myxoid liposarcomas, 33 DDLPSs, and 100 UPSs were collected as tissue microarrays (TMAs), while 87 samples were analyzed as whole section from FFPE tissue blocks and comprised 32 DDLPSs, 24 myxoid liposarcomas, and 31 osteosarcomas. This study was approved by the Institutional Review Board of Samsung Medical Center in Seoul, Korea (IRB file No. 2018-03-143). Informed consents were waived by the board.
Cell lines, reagents, and IFN-γ treatment
Human soft tissue sarcoma cell lines were obtained from American Type Culture Collection (ATCC), Korean Cell Line Bank (KCLB), and other laboratories detailed in Additional file
1: Table S1. Each cell line was grown in appropriate culture medium (Additional file
1: Table S2) with 10% fetal bovine serum (Gibco, 16000-044) and 1% antibiotic–antimycotic 100× (Gibco, 15240-112). Cell lines were tested and validated for mycoplasma detection and human cell line authentication (STR DNA profiling) using AmpFLSTR™ Identifiler PCR Amplification Kit (Thermo Fisher Scientific, 4322288). For IFN-γ treatment, each cell line was seeded into six-well plates and treated with IFN-γ (R&D systems, 285-IF-100; 50 or 100 ng/ml) or BSA (Thermo Fisher Scientific, 23209; 50 or 100 ng/ml) as controls and incubated at 37 °C for 48 h.
Western blot
Cells were lysed in RIPA buffer (0.1% SDS, 0.5% sodium deoxycholate, 250 Mm NaCl, 1% Triton X-100 and 50 Mm pH 8.0 Tris) containing a phosphatase inhibitor and protease inhibitor cocktail tablets (Roche) and quantified using Pierce™ BCA Protein Assay kit (Thermo Fisher Scientific, 23227) according to the manufacturer’s instructions. One hundred micrograms of total protein from cells were separated by 10% SDS-PAGE and transferred to nitrocellulose membrane (Pall Corporation), then the membranes were blocked with 5% nonfat milk in 1× TBST (Tris-buffered saline with Tween 20). Proteins were probed with the following primary antibodies: monoclonal anti-PD-L1 (Cell Signaling Technology, E1L3N, 1:3000) and anti-β-actin antibodies (Santa Cruz, sc-47778, 1:1000), and washed three times with 1× TBST. Goat anti-rabbit IgG HRP (Abcam, ab6721) and goat anti-mouse IgG HRP antibodies (Abcam, ab6789) were used as secondary antibodies. Proteins were detected using ECL western blotting substrate (Promega, W1015).
Immunocytochemistry (ICC) and immunohistochemistry (IHC)
Cells were fixed in 95% ethanol and embedded in paraffin. Egg albumin was used for cell aggregation. For ICC and IHC, 4 μm thick sections from FFPE tissue blocks were cut using a microtome and routinely deparaffinized. The sections were incubated with 0.3% hydrogen peroxide to block endogenous peroxidase activity. Antigen retrieval was performed in 0.01 M of citrate buffer (pH 6.0) or Tris–EDTA buffer (10 mM Tris at pH 9.0, 1 mM EDTA, 0.03% Tween 20) at 95 °C. Three different PD-L1 antibodies (DAKO 22C3, 1:50; VENTANA SP142, 1:50; and VENTANA SP263, 1:50) were used for immunocytochemical and immunohistochemical staining. For SP142 IHC amplification and DAB development, the Biotin-Free Catalyzed Amplification System (DAKO, K1497) was used. For SP263 IHC amplification and DAB development, the OptiView DAB IHC Detection Kit (VENTANA, 760-700) was used according to the manufacturer’s instructions. Each slide was counterstained with hematoxylin and then mounted.
To evaluate the IHC results of tissue samples including both whole sections and TMAs, each case was separated into groups with < 1% (negative), 1–49% (low), or ≥ 50% (high) positive tumor cells. A tumor cell with membranous staining, at least weak and partial, counted as a positive tumor cell.
Flow cytometry
Cells were washed with fluorescence-activated cell sorting (FACS) buffer (filtered 0.1% BSA in PBS) and stained with phycoerythrin (PE)-conjugated monoclonal antibody specific for PD-L1 (eBioscience, MIH1) or IgG (Miltenyi Biotec, 130-092-212). Cells were filtered using a Falcon 5 ml round bottom tube with a cell strainer snap cap (Corning, 352235). Flow cytometric analysis was performed with FACSVerse and FACSuite (BD Biosciences).
Total RNA was isolated using RNeasy Mini Kit (Qiagen, 74106) according to the manufacturer’s instructions and quantified using Nanodrop™ 2000 spectrophotometer (Thermo Fisher Scientific, ND-2000). One microgram of total RNA was used for the synthesis of cDNA. The cDNA was synthesized using the SuperScript III First-Strand Synthesis System (Invitrogen). To determine mRNA levels of
PD-
L1 and
STAT1, qRT-PCR was performed using SYBR Green PCR Master Mix (Applied Biosystems, 4367659) and specific primer sets (Additional file
1: Table S3). Relative mRNA expression levels were normalized to the expression level of
CTBP1 using 2
−ΔΔCt (mean fold change).
Statistical analysis
Continuous variables were tested for normality of distribution using the Kolmogorov–Smirnov test and Shapiro–Wilk test. Unpaired t-test was used for the continuous variables fitting a normal distribution. Mann–Whitney U-test was used for the continuous variables showing a skewed distribution. Categorical variables were compared using the Chi square test or Fisher’s exact test. RFS was defined as the time interval between initial resection and tumor recurrence or last follow-up. OS was defined as the time interval between the initial diagnosis and death or last follow-up. Survival analysis was performed using the Kaplan–Meier method with the log-rank test. P-values ≤ 0.05 (2-tailed) was considered statistically significant. Statistical analyses were performed using Prism v.7 (GraphPad) and SPSS version 17.0 (SPSS Inc.).
Discussion
Recent advances from molecular analyses have revealed that many sarcomas are not only morphologically but also genetically distinct neoplasms. According to genomic profiles, soft tissue tumors can be broadly divided into two groups [
4,
21]. The first group comprises tumors associated with specific genetic alterations and relatively simple karyotypes. Sarcoma subtypes belonging to this group include translocation-associated tumors, such as synovial sarcoma and myxoid liposarcoma, and mutation-associated tumors such as a gastrointestinal stromal tumor. Because there is a specific genomic alteration in each subtype, gene-targeted therapies are currently under investigation. On the other hand, the second group of sarcomas is those with complex karyotypes [
3]. Sarcoma subtypes belong to this group includes UPS, DDLPS, pleomorphic rhabdomyosarcoma, myxofibrosarcoma, and malignant peripheral nerve sheath tumor [
21,
22]. Because most of these entities have no disease-specific genomic alterations, gene-targeted therapies are of limited value. However, several recent studies have revealed that the tumors with complex karyotypes emit signals that increase their immunogenicity, and evasion of local immune surveillance plays an important role in their tumorigenesis [
23]. Although the mechanisms of immune evasion are still unknown, in our opinion, the immune evasion via PD-L1 expression could be an important component.
In this study, we firstly examined the prevalence of PD-L1 expression in in a 16 sarcoma cell lines comprising 12 different subtypes. Pleomorphic rhabdomyosarcoma (HS-RMS-1), fibrosarcoma (HT1080), and DDLPS (LP6) cell lines showed a consistently increased PD-L1 protein expression via western blot, FACS, and ICC by 22C3 antibody. In a review of their genetic characteristics, both pleomorphic rhabdomyosarcoma and fibrosarcoma have been known to have complex karyotypes without specific genomic alterations [
22,
24]. In cases of DDLPS, although it is defined by 12q13~15 amplifications, it is currently classified as a complex karyotype group because of additional genomic alterations during dedifferentiation [
21]. In summary, cell lines that showed high PD-L1 expression all belonged to the complex karyotype group, which was consistent with our theory and a previous report [
25]. On the other hand, UPS (GBS-1) and myxofibrosarcoma (NMFH-1) cell lines did not show PD-L1 expression despite their complex karyotype [
21]. However, our IHC results revealed a 20% (12/60) expression rate in the TMAs of UPS. Therefore, we suggest that high PD-L1 expression tend to match with a complex karyotype group, although not all tumors with a complex karyotype express PD-L1. Screening methods such as IHC would be required in practice for the patient selection, which is similar for non-small cell carcinoma.
Although there have been several previous studies that focused on the PD-L1 immunohistochemical expression in sarcomas, most previous studies are based on a small number of specimens and showed controversial results. For example, the PD-L1 expression rates of leiomyosarcoma have been reported as 0% (0/4), 11% (1/9), and 70% (14/20) [
11,
13,
26]. In this study, we used 230 sarcoma tissue samples, comprised of 87 whole sections and 143 TMAs, and three different anti-PD-L1 antibodies to solve this controversy. Our overall expression rate was 10.9% (20/184), which was lower than that of several previous studies (43–58%) [
11,
13]. However, several other studies have also reported overall expression rates similar to our study (5–12%) [
14,
27]. In comparison with our study and previous studies, we suggest that the overall expression rates could be related to the anti-PD-L1 antibody used for IHC. More recent studies and our study used the 22C3 clone of anti-PD-L1 antibody, which is currently used in practice, and showed similar overall expression rates. It is a well-known problem that the expression rates of PD-L1 could vary according to the antibodies used in the IHC. The Blueprint Project, which included four anti-PD-L1 antibodies used in clinical trials, showed that the staining proportion of tumor cells could vary according to the antibody clone that was used [
28]. These discordances were most significant with the SP142 clone compared to the 22C3, 28-8, and SP263 clones [
28]. Our results showed a strong correlation (Pearson’s r = 0.882) between 22C3 and SP263, but SP142 showed only moderate correlation with 22C3 (Pearson’s r = 0.551) and SP263 (Pearson’s r = 0.503), which was consistent with the Blueprint Project.
Histologic characteristics also can increase interobserver variability in the evaluation of PD-L1 expression of sarcomas. Under a light microscope, many sarcomas, especially UPS and DDLPS, show tumor cells intermixed with inflammatory cells. According to a recent molecular analysis report, UPS and DDLPS showed the highest median number of macrophages among sarcomas [
21]. Considering that the inflammatory cells including macrophages can show PD-L1 immunoreactivity, the interpretation of PD-L1 immunoreactivity in sarcomas is not straightforward and can result in high interobserver variability. Additionally, as we described above, the recent molecular analysis revealed that many histologic sarcoma subtypes have distinct molecular characteristics. Considering that different sarcoma subtype shows different morphologies and genetics, we hypothesize that different sarcoma subtypes would show different overall PD-L1 expression rates.
DDLPS showed no PD-L1 expression in the TMAs but showed a 21.9% (7/32) PD-L1 expression rate in whole sections. This result may have been mainly due to low tumor cell proportion with PD-L1 positivity. In this study, only 3 cases showed PD-L1 expression in more than 50% of tumor cells and are usually limited to a dedifferentiated area. In a review of previous studies, there were no positive cases in the SARC028 study, which was based on biopsy samples [
14]. Torabi et al. [
29] reported that only one weak positive case in 64 liposarcoma cases, but they did not include DDLPS cases in their study. Osteosarcomas had a PD-L1 expression rate of 3.2% (1/31). The only case with PD-L1 immunoreactivity was conventional osteosarcoma with high-grade spindle cell morphology. Similar to our results, Torabi et al. [
29] reported no (0/26) PD-L1 expression in osteosarcomas. There was also a previous study that reported that 24% (9/38) cases of osteosarcoma showed high PD-L1 RNA expression, but they did not report the immunoreactivity of PD-L1 protein in their study [
30]. UPS had a 20% (12/60) PD-L1 expression rate in our study. This result was consistent with the SARC028 study which reported that only UPS cases (3/10) showed PD-L1 immunoreactivity [
14]. However, both the SARC028 study and our study are based on biopsied specimens. Considering the difference in expression rates between TMAs and whole sections of DDLPS in our study, there is a possibility of underestimating the overall expression rate in UPS. In summary, in this study, we confirmed positive PD-L1 expression with the 22C3 clone in DDLPS and UPS. Our finding supports a result where pembrolizumab have a specific activity in patients with DDLPS and UPS among seven subtypes of sarcomas, which was comprised of 84 patients with bone or soft tissue sarcomas in the SARC028 study [
14].
The association between PD-L1 expression and poor prognosis had been reported in many previous studies [
13,
25,
30]. The cases of DDLPS in our study also showed significantly worse outcomes, whereas cases of UPS in our study showed no significant differences in outcome. Considering that UPS is a sarcoma with very poor prognosis, the contribution of PD-L1 expression to prognosis could be masked by the aggressive of the UPS.
IFN-γ has been reported to induce PD-L1 expression in several cell lines including chordoma, angiosarcoma, and osteosarcoma cell lines [
17,
19,
30]. Noteworthy, PD-L1 expression was induced following vascular-targeted photodynamic treatment or ionizing radiotherapy through an increase in IFN-γ, then a mono- or combination treatment with systemic PD-1/PD-L1 pathway blockade inhibits the generation of potent local and systemic tumors in mouse models using human renal cells or murine colon cancer cells [
31,
32]. These previous reports suggest the possible effectiveness of antitumor immunotherapy through a dual treatment with IFN-γ and PD-L1. We also observed the possibility of dual therapy using IFN-γ and PD-L1 in pleomorphic rhabdomyosarcoma, DDLPS, myxoid liposarcoma, and rhabdomyosarcoma. However, the effectiveness of dual therapy with IFN-γ and PD-L1 will need to be validated in various sarcoma subtypes in vivo.
Authors’ contributions
HKP, YJK, and YLC designed the study. HKP, MGK, MJS, and YJK wrote the manuscript. HKP and YLC reviewed the histopathology and IHC results. HKP, MGK, MJS, SEL, and YJK performed the clinical analysis and experiments. HKP, YJK, and YLC provided critical comments on the manuscript. All authors read and approved the final manuscript.