Lipid droplets-related Perilipin-3: potential immune checkpoint and oncogene in oral squamous cell carcinoma

The studies involving human participants were reviewed and approved by the Ethics Committee of Nanjing Stomatology Hospital (2019NL-009(KS)). OSCC tumors were obtained from 87 patients from 2014 to 2015 in Nanjing Stomatological Hospital. Informed consent was provided by the patients for the use of their tissues and data. All cases were reviewed by experienced pathologists. Paraffin-embedded OSCC tissue slices were obtained from the pathology department and used for immunohistochemistry study. Seventy-four blood samples from OSCC patients were obtained for flow cytometry assay before any related treatments. Additional 51 OSCC samples were used to determine the correlation between PLIN3 and B7-H2. The 51 samples tested by flow cytometry and 87 samples tested by IHC were all primary OSCC patients who had not received preoperative radiotherapy or chemotherapy.

The protocol of immunohistochemistry of formalin-fixed paraffin-embedded sections was performed as previously described [30]. The primary antibodies include PLIN3 (diluted 1:500, Abcam, ab224344), B7-H2 (diluted 1:400, Abcam, ab257321), Ki-67 (diluted 1:200, ab16667; Abcam), CD4 (ZSGB-BIO, ZM-0418), CD8 (ZSGB-BIO, ZA-0508), CD19 (ZSGB-BIO, ZM-0038), CD56 (ZSGB-BIO, ZM-0057), and CD68 (ZSGB-BIO, ZM-0464). The CD4, CD8, CD19, CD56, and CD68 antibody working solution did not need to be diluted and was used directly by dropwise addition. The assessment of antigen expression was based on the intensity of the stain and the proportion of cells that tested positive. The percentage of stained cells was defined as 0 = 0–5%; 1 = 6%–25%; 2 = 26%–50%;3 = 51%–75%; and 4 = 76–100%. Staining intensity was defined as follows: 0 = negative staining; 0 = no staining; 1 = light yellow; 2 = brownish yellow; 3 = dark brownish yellow. The IHC score was calculated by multiplying the two scores as mentioned above. The expression levels of PLIN3, B7-H2, Ki-67, and immunocyte molecules were defined as “low” when it is lower than the median value and as “high” when it is equal to or greater than the median. Two pathologists conducted all scorings without any knowledge of the patients' clinical characteristics or outcome. Define polygonal or long spindle-shaped cells as FLCs with ovoid nuclei and large, distinct nucleoli, whereas TILs have small cytoplasmic bodies and a smaller nucleoplasmic ratio and are often more regular.

We used the GEO (GSE103322, GSE164690), TISCH (http://​tisch.​comp-genomics.​org/​) to analyze the RNA expression of PLIN3 and used the Human Protein Atlas (https://​www.​proteinatlas.​org/​) database to analyze the protein levels of PLIN3 in different cell types of HNSCC and tongue tissues. We used cBioportal (https://​www.​cbioportal.​org/​) and TIMER (http://​timer.​cistrome.​org/​) database to analyze the correlation between RNA expression of PLIN3 and specific immune cell subsets markers in HNSCC.

OSCC cell lines (HN6, Cal27, and Cal33) and 293 T cells were obtained from the National Collection of Authenticated Cell Cultures and identified by Short Tandem Repeat. HN6, Cal27, Cal33, and 293 T cells were cultured in DMEM medium (Invitrogen) supplemented with 10% Fetal bovine serum (FBS) (Gibco) at 37 °C in 5% CO₂. siRNA and lentiviral plasmid vectors of PLIN3 were purchased from RiboBio and OBIO, respectively. Oleic acid (Merck, O1257) was diluted at a ratio of 1:500 and added to cells for 12 h. All steps were executed according to the manufacturer's instructed protocol.

Cells were grown on 12/24 wells plate to 50% confluence and transfected the siRNA of PLIN3 using Lipofectamine 3000 (Invitrogen) according to the manufacturer’s instructions. Cells were harvested after 24 h or 48 h for qPCR, Western blot analysis, and RNA sequence (Shanghai OE Biotech. Co., Ltd).

As for the stable transfection of PLIN3, 293 T cells were cultured in 10-cm dish and transfected with Lentiviral plasmid vector using PolyJetTM DNA in vitro transfection reagents (SignaGen, America), the virus supernatant above 293 T cells was extracted and cultured with tumor cells. The GFP-labeled Lentivirus vector expressing sh-PLIN3 (Lv-ShPLIN3) and overexpression vector containing PLIN3 (Lv-OE-PLIN3) were used, with Lv-ctrl as the matched controls (OBiO Technology Co. Ltd., Shanghai, China).

RNA was obtained by using AG SteadyPure RNA Extraction Kit following the manufacturer’s procedure and then reversed into cDNA using 5X Evo M-MLVRT Master Mix (Accurate Biotechnology, AG11706). The relevant expression of the genes was determined via 2X SYBR Green Pro Taq HS Premix (Accurate Biotechnology, AG11718).

The protocol of WB was performed as previously described [5]. Protein samples were isolated by 4–20% SDS-PAGE (smart-Lifesciences). The primary antibodies include PLIN3 (1:1000, Abcam, ab224344), B7-H2(1:1000, Abcam, ab257321), PD-L1(1:1000, CST, #13684), and ACTIN (1:2000, Servicebio, GB12001).

GSEA includes GO enrichment and KEGG pathway assays, employed to unveil the fundamental biological process, cellular composition, and functional molecules. Each gene set was assigned a distinct gene set based on a 'hallmark'.

The working Oil Red O solution was obtained by diluting saturated Oil Red O solution (Servicebio technology) at 6:4 with distilled water. The cells which were cultured in 6-well plate were washed twice using PBS and fixed in 4% paraformaldehyde for 10 min before staining, and then the cells were incubated in 60% isopropanol for 15 s and incubated in Oil Red O staining solution avoiding light for 10 min, and finally, washed twice, photographed and counted. Using image J to measure the length of LDs. Oil Red O staining of frozen sections was also carried out according to the instructions [21].

Cultured cells were fixed in 4% paraformaldehyde solution on the 6-well plates and washed in 1xPBS prior to staining with 10 μg/ml BODIPY 493/503 (Invitrogen, D3922) for 10 min. Samples were then washed with PBS twice. Similarly, cells were incubated with 5 μg/ml Nile red solution (Invitrogen, N1142) for 15 min at room temperature to stain neutral lipids, and the images were visualized by immunofluorescence microscopy.

The constructed stable-transformation cell lines and control cells were seeded in 6-well plates until 100% fusion. The wounds were scored with a micropipette tip and then washed with PBS and cultured in serum-free DMEM medium. The same area of the wound was photographed at 0 and 24 h to analyze the wound closure of the cells. ImageJ software was used to measure the trace widths of different treatment groups for analysis.

To assess cell migration capacity, transwell assays were done as described previously [21]. In brief, 2 × 10⁴ cells were seeded in the top chamber of the insert (200 μL/well) with serum-free medium, followed by adding media containing 10% FBS in the lower chamber as a chemoattractant (600 μL/well). Five fields were randomly chosen, and the migration ability of tumor cells was assessed by ImageJ software.

To assess the proliferative capacity of tumor cells, we used 3D spheroid models. Spheroid formation was achieved by using ultra-low attachment (ULA) culture plates (96-well spheroid microplates with ULA surface). HN6 cells were grown in DMEM supplemented with 10% FBS and 1% penicillin–streptomycin. A total of 8000 cells were seeded per well and cultured in an ultra-low adhesive environment in a humidified 5% CO₂ atmosphere at 37 °C, and the medium was refreshed once a day. The diameter of tumor spheres was observed daily for the following eight days.

PBMC was prepared as previously reported [5]. All study participants provided informed consent and approved by the ethical committee of Nanjing Stomatology Hospital, affiliated Hospital of Medical School, Nanjing University, Nanjing, China.

The experimental group and the control group of tumor cells were plated in 12-well plates, and the count of cells in each well was 1 × 10⁵. According to the ratio of PBMC: tumor cells = 5:1, PBMC was added into the 12-well plates and cultured for 48 h, and then flow cytometry was performed.

The PBMC samples were collected from patient's peripheral blood. Percentage and absolute numbers of lymphocytes were assessed by flow cytometry. Cells were collected and then washed and suspended in 200μL PBS. Cells were stained with CD45-PerCP, CD3-FITC, CD4-APC, CD8-PE, CD19-APC, CD16-PE, CD56-PE (BD Multitest™) for 30 min to enumerate the CD3⁺CD4⁺ T cells, CD3⁺CD8⁺ T cells, CD3⁻CD19⁺ B cells, and CD3⁻CD16⁺CD56⁺ NK cells, respectively. Flow cytometry was finally performed using FACSCalibur instrument and analyzed using FlowJo software, and the graphs were generated by GraphPad Prism. For the cell subtypes of PBMC analysis, details of this protocol were the same as the previous description [31].

To unveil the spatial distribution of PLIN3 in vitro, immunohistochemical (IHC) staining was performed in a cohort of OSCC patients (n = 87). Fifty-five patients were male, and 32 patients were female. Median patient age was 62 years. The site of occurrence was predominantly the tongue in 87 patients, accounting for 34.5% of all sites. Median follow-up was 30 months for overall survival time (OS), 26 for metastasis-free survival time (MFS), and 21 months for disease-free survival time (DFS). Most patients (n = 58) were in advanced stages of OSCC (cTNM III or IV), and 23 patients exhibited postoperative distant metastasis.

Histological sections showed that spotted lipid droplets-like PLIN3 staining was dominantly enriched in tumor cells (TCs) than other cell types including fibroblast-like cells (FLCs) and tumor-infiltrating lymphocytes (TILs) (Fig. 1a, b). Similarly, single-cell RNA-seq data revealed that in head and neck squamous cell cancer (HNSCC), PLIN3 was enriched in epithelial cells (Fig. 1c, d) (GSE103322, GSE164690), and a similar expression pattern was also exhibited in tongue tissues (Fig. 1e) (https://​www.​proteinatlas.​org). Interestingly, OSCC patients with PLIN3^high tumor cells exhibited a higher Ki-67 positivity (Fig. 1f) contrary to immunocytes (Fig. 1G), and we speculate that PLIN3 was as important as LDs for the growth of tumor cells, but also indispensable for immune cells in the activation of CD8⁺ T lymphocytes [32, 33]. There was no significant link between PLIN3 and Ki-67 in OSCC fibroblasts (Fig. 1h).

Correlation between protein expression levels of PLIN3 and clinicopathological variables was evaluated in patients with OSCC, such as postoperative recurrence, distant metastasis, sex, age of patients, as well as tumor node metastasis (TNM) stage, T stage, N stage, and differentiation. Our data showed that patients with increased PLIN3^TCs, instead of PLIN3^TILs and PLIN3^FLCs, had a higher risk of postoperative distant metastasis, but the recurrence incidence was comparable in total OSCC cohort (Fig. 1i, j). Notably, we also observed that patients with PLIN3^high tumor cells were strongly correlated with poor differentiation and high PLIN3⁺ fibroblasts showed advanced T stage (Table 1), suggesting that PLIN3 in OSCC might promote malignant progression.

Intricate crosstalk between cancer cells and host immune system fuels tumor immune evasion and ultimately results in tumor overgrowth and metastatic dissemination [34]. To explore the immunocytes in situ tissue of PLIN3⁺ HNSCC and OSCC, we utilized cBioPortal database and IHC staining to evaluate the relativity of PLIN3 and CD8, CD4, CD56, CD68, or CD19. In HNSCC, PLIN3 correlated inversely with CD8, CD4, and CD19 densities (Fig. 2a, b, e), while was in direct ratio with CD56 and CD68 (Fig. 2c, d). In OSCC, PLIN3^High tumors harbored obviously less infiltrated CD8⁺ T lymphocytes (Fig. 2f, g, l) and CD4⁺ T lymphocytes (Fig. 2h, m), whereas CD56⁺ NK cells, CD68⁺ tumor-associated macrophages (TAMs), and CD19⁺ B lymphocytes were essentially comparable (Fig. 2i–k). Noteworthy, CD4/CD8 ratio < 1 at the tumor–host interface has prognostic value in triple-negative breast cancer [35]; we likewise found the mean CD4/CD8 ratio to be at 0.47, which was less than 1 and negatively associated with PLIN3 (Fig. 2n), indicating that patients with PLIN3^high tumor had a poor outcome.

In addition, through the correlation analysis of immune cells markers in HNSCC patients by TIMER database (http://​timer.​cistrome.​org/​), we found that patients with high PLIN3 showed increased TGF-β⁺ regulatory T cells and CD68⁺ TAMs, while diminishing CD8⁺ T cells and NOS2⁺ M1 macrophage (Fig. 2o), indicating that PLIN3 favored the establishment of a unique and highly immunosuppressive tumor microenvironment (TME). Furthermore, cBioPortal database showed that PLIN3 enhanced immune checkpoint molecules such as CD244 (2B4), CD274 (PD-L1), interleukin 10 (IL-10), NECTIN2 (CD112), PDCD1LG2 (PD-L2), transforming growth factor beta 1 (TGFβ1), and transforming growth factor beta receptor 1 (TGFβR1) (Fig. 2p).

Since the metastasis of tumor requires cancer cells to circulate in the bloodstream, endure pressure in blood vessels, and escape deadly combat with immune cells [36, 37], we next analyzed the ratio and absolute number of key immunocytes in peripheral blood of OSCC patients (n = 74) according to PLIN3 expression by flow cytometry. Human CD3⁺ T cells, CD3⁺CD4⁺ helper/inducer T cells, CD3⁺CD8⁺ cytotoxic T cells, CD3⁻CD19⁺ B cells, and CD3⁻CD16⁺, and/or CD56⁺ NK cells were analyzed in PLIN3^high and PLIN3^low groups, and the strategy for gating lymphocytes is shown in Fig. 3a. The results indicated that patients with enhanced PLIN3⁺ tumor cells had relatively low ratio and numbers of CD3⁺ T cells and CD3⁺CD4⁺ helper/inducer T cells in blood (Fig. 3b, e), whereas significance was lost within immunocytes (Fig. 3c, f) and fibroblasts (Fig. 3d, g). In addition, the percent and absolute count of NK cells and B lymphocytes were comparable in different cell populations, in agreement with the lack of differences in situ tissue. Patients with PLIN3^high tumor cells showed a trend toward lower absolute count of CD8⁺T cell infiltration in TME (Fig. 3e). Collectively, high PLIN3⁺ tumor cells might result in the exhaustion of T cells in peripheral blood.

To further clarify the prognostic value of PLIN3, we conducted 5-year survival curve assays. Results showed that patients with PLIN3^high tumor cells or fibroblasts had shorter postoperative OS (Fig. 4a, c), MFS (Fig. 4d, f), and DFS (Fig. 4g, i). Conversely, we found that patients with high PLIN3⁺ immunocytes showed longer OS (Fig. 4b), MFS (Fig. 4e), and DFS (Fig. 4h). We hypothesized that high intracellular PLIN3 expression promoted tumor cell growth; however, for immune cells, it enhanced the antitumor immune efficacy of CD8⁺ T lymphocytes. Similar results were observed in cervical squamous cell carcinoma (Fig. 4j), esophageal squamous cell carcinoma (Fig. 4k), and liver hepatocellular carcinoma (Fig. 4l) by Kaplan–Meier plotter database (http://​kmplot.​com/​analysis/​index.​php?​p=​service), and patients with enhanced PLIN3 expression in tumor cells had shorter OS.

To analyze the prognostic value of clinicopathological features, we used univariate and multivariate Cox regression analyses. Our data confirmed that the presence of nodal and PLIN3^high expression in OSCC was associated with shorter OS and DFS in univariable analysis (Fig. 4m). Moreover, lymph node metastasis (yes vs no, HR 2.403, 95% CI 1.149–5.026, P < 0.05, Fig. 4m) and PLIN3 in fibroblasts (high vs low, HR 2.615, 95% CI 1.260–5.428, P < 0.05, Fig. 4m) could be used as independent prognostic factors for OS and DFS of OSCC in the multivariable models. Altogether these data indicated that PLIN3 was upregulated in OSCC tissues and high PLIN3⁺ tumor cells may lead to immunosuppression through T cell exhaustion, accompanied by infiltration of PLIN3⁺ fibroblasts, thus collaboratively promoting tumor progression and predicting inferior patient outcome.

Considering the metastasis- and immune-regulatory role of PLIN3 in OSCC patients, we next investigated the role of PLIN3 in OSCC cells in vitro. Firstly, we evaluated the expression level of PLIN3 in normal epithelial cells and different tumor cells and found it was higher in HN6 and Cal27, but lower in Cal33 (Fig. 5a). We thus knocked down PLIN3 (PLIN3^kd) in HN6 cell lines with three independent siRNA (Fig. 5b, c) and then established stable PLIN3 silencing HN6 cells and PLIN3 overexpressed (PLIN3^oe) Cal33 cell lines with lentivirus plasmid vector, and the efficiency was verified by qPCR and WB (Fig. 5d, e).

To discover the potential biological functions and signaling pathways of PLIN3, we obtained KEGG, GO, and HALLMARK annotations in PLIN3^kd OSCC cells based on functional enrichment analysis of RNA sequence results. The results showed that PLIN3 played a critical role in multiple lipid metabolism-related pathways including adipocytokine signaling pathway, fatty acid transport, cholesterol homeostasis, and adipogenesis (Fig. 5f). To further investigate whether PLIN3 contribute to LDs biosynthesis, an Oil Red O staining assay was conducted as a visual indicator of intracellular LDs in OSCC. LD deposits were obviously decreased in PLIN3^kd HN6 cell lines, but increased in PLIN3^oe Cal33 cell lines (Fig. 5g). Consistently, immunofluorescence and Oil Red O staining were performed to identify intracellular vacuole-like structures as being LDs; OA was reported to promote lipid synthesis in HepG2 cancer cells [38] and here also increased the extent of LDs in OSCC cells (Fig. 5H, I). Moreover, the intracellular LDs were smaller and less abundant in the PLIN3-knockdown HN6 cells than in the controls, but the addition of exogenous OA restored LD biogenesis (Fig. 5j, k). Multiple pieces of evidence demonstrated LDs were responsible for incusing cancer cells proliferation and migration [39]. As expected, knocking-down PLIN3 led to a significant reduction in the migration of HN6 cells in wound healing and transwell assays (Fig. 5l, m), but failed to influence proliferation when cells were cultured continuously for 8 days by 3D spheroid models (Fig. 5n).

LDs are linked to the regulation of immune by acting as specialized hubs of signaling with lipid mediator formation in cancer [26, 39]. Consistently, differential changed genes by PLIN3 knockdown were significantly enriched in pathways of immune system (Fig. 6a), and enrichment plot revealed that PLIN3 could enrich the migration of lymphocytes and myeloid leukocytes (Fig. 6b). To further estimate the role of PLIIN3⁺ tumor cells in immune escape, we established human tumor/immunocytes co-culture system using peripheral blood mononuclear cells (PBMCs) (Fig. 6c) and performed flow cytometry to detect CD8⁺ T cell activation ratio. The results showed that the proportion of CD3⁺CD8⁺ T cells was significantly upregulated in PLIN3^kd HN6 cells compared to controls (Fig. 6d). As reported previously, increased LDs associated with reduced recruitment of IFNγ-secreting CD8⁺ T cells to tumor site could consequently avoid exposure of PD-L1 and PD-1 on tumor and CD8⁺ T cells [40]. We indeed observed a significant downregulation of B7-H1 (PD-L1) and B7-H2 (ICOSLG) mRNA and protein expression in PLIN3^kd HN6 cells (Fig. 6e, f). Patients with B7-H2^high tumor cells showed advanced TNM stage and N⁺ stage with decreased OS, MFS, or DFS, and this sub-cohort was featured with a trend of diminished CD4⁺ T cells and CD8⁺ cells in tumor center in situ, which positively correlated with PD-L1 and IL10 [41]. In this study, we focused on PLIN3/B7-H2 signals in OSCC and correlation between PLIN3 and PD-L1 was investigated in our other unpublished research.

Given that PLIN3 promoted B7-H2 protein expression, we reasoned that B7-H2 was required for PLIN3-mediated tumor promotion. To test our hypothesis, we analyzed the expression of PLIN3 and B7-H2 in the same tissue sections of OSCC by IHC staining (n = 51), and the results showed that the protein levels of B7-H2 were largely upregulated in PLIN3^High tumors (n = 51), which was positively correlated (Fig. 6g, h). Thus, we divided OSCC patients into different subpopulations based on PLIN3 and B7-H2 expression, and OSCC patients with high PLIN3⁺B7-H2⁺ tumor cells had significantly shorter OS and DFS (Fig. 6i, j) and harbored less absolute count of CD3⁺CD8⁺ and CD3⁺CD4⁺ T cell infiltration in peripheral blood (Fig. 6k, l). Compelling evidence suggests that LDs accumulation led to failure in dendritic cells (DCs) maturation, which limited recruitment/activation of naïve CD8⁺ T cells via co-stimulatory factors CD80/86/MHC-I and CD28/TCR [40]. These data indicated that LDs-related PLIN3 promoted CD8⁺T cells exhaustion in OSCC patients, which might be related to B7-H2-mediated immune homeostasis (Fig. 7).