Quantitative proteomics analysis of glioblastoma cell lines after lncRNA HULC silencing

Glioblastoma multiforme (GBM) is a life-threatening brain tumor. This study aimed to identify potential targets of the long noncoding RNA (lncRNA) HULC that promoted the progression of GBM. Two U87 cell lines were constructed: HULC-siRNA and negative control (NC). Quantitative real-time PCR (qRT-PCR) was performed to validate the transfection efficiency of HULC silencing vector. Mass spectrometry (MS) was used to generate proteomic profiles for the two cell lines. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed to distinguish HULC-related genes and pathway mapping. Colony formation, Transwell, and wound-healing assays were used to investigate the functional effects of HULC knockdown on GBM. We identified 112 up-regulated proteins and 24 down-regulated proteins from a total of 4360 quantified proteins. GO enrichment illustrated that these proteins were mainly involved in organelle structure, catalysis, cell movement, and material metabolism. KEGG pathway analysis indicated that some of these proteins were significantly enriched in tight junction, metabolic pathways, and arachidonic acid metabolism. In vitro experiments demonstrated that HULC knockdown inhibited GBM cell proliferation, invasion, and migration. Our KEGG analyses revealed that PLA2G4A was a shared protein in several enriched pathways. HULC silencing significantly down-regulated the expression of PLA2G4A. Knockdown of HULC changed the proteomic characteristics of GBM and altered the behaviors of GBM cells. Specifically, we identified PLA2G4A as an HULC target in GBM. This study provides a new perspective on the mechanisms and potential drug targets of GBM treatment.


Identification of proteins exhibiting altered expression in HULC-siRNA cells. We performed
LC-MS/MS to identify proteins exhibiting altered expression in HULC-siRNA cells. Comparisons of the quantitative values of protein expression were made between the mean and standard error of the readings of the HULC-siRNA and NC cells. The data were filtered as statistically significant when the P value was < 0.05, and a fold change in protein expression > 1.2 was regarded as up-regulation. Conversely, a fold change in protein expression < 1/1.2 was regarded as down-regulation. A total of 112 up-regulated proteins and 24 down-regulated proteins was detected ( Table 1). The top five up-regulated proteins were APOC3, CCDC146, MPZ, CRYAB, and RNF7, and the top five down-regulated proteins were CCDC159, SASH1, ANXA8L1, PLA2G4A, and CYP51A1. A volcano plot shows the log2 (fold change) as the abscissa, and -log10 of the P value as the ordinate (Fig. 1G).
Our results indicate that HULC knockdown alters the protein profile of GBM cells, which likely contributed to tumor pathogenesis.

Functional classification of identified proteins.
To determine the functional characteristics of the identified proteins, three primary annotations were first obtained from the GO analysis: biological process, cellular component, and molecular function. In the GO secondary classification, the differentially expressed proteins were related to some important biological processes, including cells (87.5%), organelles (75.7%), and biological regulation processes (70.6%). These proteins participate in the composition of multiple cellular components (75.5%) and play a pivotal role in molecular binding (94.9%) and catalytic activity (26.5%). Moreover, this functional annotation appeared in both up-regulated and down-regulated proteins ( Fig. 2A,B). The Fisher's exact test was further applied to the GO functional enrichment analysis of the identified proteins. As shown in Fig. 2C,D, when HULC was silenced, the proteins involved in the formation of the extracellular region were most significantly down-regulated, while proteins forming actin filament bundles were most obviously up-regulated. Demethylase activity and calcium-dependent phospholipid binding were significantly down-regulated, while proteins involved in actin binding were notably up-regulated. Moreover, various lipid metabolism pathways were significantly enriched in biological process (Fig. 2C,D). Directed acyclic graphs ( Supplementary Fig. S1) not only intuitively reflect the enrichment differences of each GO classification, but also present the upper and lower hierarchical relationships of GO functions, indicating that GO function enrichment provided a deeper level of classification. For example, actin-dependent ATPase activity was significantly up-regulated at level 10 and calcium-dependent phospholipid binding was enriched in down-regulation at level 6.
KEGG pathway annotation and enrichment. To understand the regulatory network associated with HULC knockdown, KEGG pathway analysis was performed with all differentially expressed proteins. We used the Fisher's exact test to further reveal the significantly enriched proteins in the annotated KEGG pathways.
The P values are presented as -log10 conversion. Our results indicate that tight junction was the most enriched pathway and that there was a 3.4-fold up-regulation in this pathway following HULC knockdown. The downregulated KEGG pathways were distributed in metabolic pathway, arachidonic acid metabolism, terpenoid backbone biosynthesis, and platelet activation (Fig. 3).

Functional effects of HULC knockdown on U87 cells.
To assess the functional effects of HULC silencing on GBM cells, we first analyzed the effect of HULC knockdown on cell proliferation by colony formation assay in U87 cell lines. Proliferating colonies were scored as the 12 days after seeding. Compared to the negative control, the siRNA-mediated knockdown of HULC showed a 3.39-fold decrease in the number of www.nature.com/scientificreports/ clusters (P = 0.0002), indicating that cell proliferation was significantly inhibited (Fig. 4A). We next used the Transwell assay to determine whether HULC knockdown affected cell invasion. We found that HULC knockdown decreased cell invasion capability by 2.45-fold (P = 0.0003) compared to the negative control (Fig. 4B). The wound-healing assay showed that cell migration was also suppressed following HULC knockdown. Migration was reduced by 1.84-fold at 24 h (P = 0.0002), and 1.62-fold at 48 h (P = 0.0003) (Fig. 4C). These data indicate that HULC promotes GBM cell proliferation, invasion, and mobility in vitro.  To identify key proteins regulated by HULC, we analyzed the common proteins of several significantly different signaling pathways (arachidonic acid metabolism, platelet activation, etc.). As a result, we found that the protein encoded by PLA2G4A plays a pivotal role in these pathways. Therefore, we used Western blot analysis to verify differences in PLA2G4A protein expression. Our results showed that knockdown of HULC significantly reduced the protein abundance of PLA2G4A (Fig. 4D).

Discussion
GBM is a grade IV glioma and is the most aggressive malignant type of brain tumor. Increasing evidence demonstrates that many lncRNAs play various roles in a series of biological processes associated with the occurrence and development of GBM. For example, high expression of PVT1 in the nucleus can accelerate glioma cell proliferation, invasion, and aerobic glycolysis by inhibiting the expression of miR-140-5p 17 . GAS5-AS1 is another lncRNA expressed in glioma tissues. One study showed that GAS5-AS1 binded to miR-106b-5p to promote expression of downstream genes that play a role in inhibiting cell proliferation, migration, and invasion of glioma cells 18 . Emerging studies have invested the mechanisms by which lncRNAs influence other tumor behavior 19,20 .
Much of these efforts have been focused on identifying highly specific and sensitive biomarkers to promote early diagnosis, predict prognosis, and provide potential therapeutic targets for different cancers. The lncRNA HULC has been shown to be highly expressed in GBM cells compared to normal cells, as well as to promote the proliferation of GBM cells in vitro 11  www.nature.com/scientificreports/ suppression 12 . However, the molecular mechanisms responsible for HULC's regulation in GBM tumorigenesis have only begun to be scrutinized. Our study provides insight into this mechanism by identifying the potential targets of HULC in glioma cells. Proteomics research has gained much attention in tumor biology studies. Farhadul et al. analyzed differences in the total proteome between esophageal squamous cell carcinoma and non-tumor cells using label-free shotgun proteomics combined with MS 13 . Zhao et al. screened tumor-specific antigens for high-grade serous ovarian cancer with MS, and found potential targets for ovarian cancer immunotherapy 14 . In this study, we obtained Among the differentially expressed proteins, we selected the top 5 up-regulated and 5 down-regulated proteins to further analyze. Based on a search of the PubMed database, none of these 10 proteins was previously reported to be related to HULC. Only two of the proteins, CRYAB and SASH1, have been studied in glioma 21,22 .    21 . Methylation of SASH1 gene has been shown to inhibit cell adhesion and promote migration of astrocytes 22 . The remaining 8 proteins have not been previously reported to have any association with glioma. Although we did not further analyze these 8 proteins in the current study, we believe future in-depth analysis of these proteins will be helpful to better understand the underlying molecular mechanisms in GBM. However, the GO findings were unexpected in that we identified some up-regulated proteins in cell activity, such as actin filament bundles and actin binding after HULC knockdown, that indicate that HULC suppression can promote tumor migration and invasion, which contradicts our functional results. We speculate that this discrepancy correlates with the complex characteristics of glial cells. In addition to participating in the formation of actin frameworks, glial cells can contract and phagocytose cell fragments, as well as repair and replenish neurons. We also acknowledge that there are likely differences between the MS data and actual verification results 23 . The biosynthesis pathway of the terpenoid backbone was significantly down-regulated, indicating that HULC knockdown exhibited a suppressive effect on cell proliferation. In addition, the strong down-regulation of the platelet activation pathway suggested that HULC was associated with GBM complications, such as thrombosis, to a certain extent 24 . DNA methylation is known to be an early event of tumorigenesis. MGMT (O6-methylguanine-DNA methyltransferase) is a DNA repair enzyme. It was reported that the methylation of the MGMT gene promoter is associated with glioma prognosis and recurrence 25 . Our proteomics analysis demonstrated that demethylase activity was decreased after HULC knockdown. Previous studies have also illustrated that demethylation behavior could promote tumorigenesis and progression 26,27 . Thus, the methylation or demethylation of HULC's target gene should be investigated in future studies.
We found that the PLA2G4A encoded protein appeared in several notable KEGG pathways. Therefore, we hypothesized that PLA2G4A might be a potential downstream target of HULC. PLA2G4A is the most abundant subtype in the family of phospholipase A2. Phospholipase hydrolyzes membrane phospholipids and releases arachidonic acid, which is further involved in many pathophysiological processes, including inflammation, signal transmission, and cell growth 28 . Although one study showed that reducing PLA2G4A expression could promote the migration and invasion of esophageal squamous cell carcinoma 29 , others proposed that PLA2G4A was an oncogene [30][31][32] . For example, PLA2G4A has been shown to facilitate the metastasis of osteosarcoma by promoting epithelial-mesenchymal transition (EMT) 32 . Our proteomics data supported PLA2G4A's role as an oncogene. Our independent Western blot assay also confirmed that the HULC knockdown significantly reduced PLA2G4A protein expression, suggesting that PLA2G4A might be a key protein that was regulated by HULC in GBM. Our enrichment analysis showed that PLA2G4A was involved in many important processes, including positive regulation of cell proliferation, calcium-dependent phospholipid binding, and the arachidonic acid metabolism pathway. Since the concept of tumor-promoting inflammation was proposed in 2011, tumor-associated inflammation has been considered a trigger point for cancer progression 33 . We hypothesize that PLA2G4A may also play an important role in the formation of tumor-related inflammation. Thus, targeting PLA2G4A might provide a promising therapy to GBM. Moreover, Tsuji S, et al. put forward that temozolomide might affect cPLA2 34 , which inspired us that targeting PLA2G4A might reverse temozolomide resistance.
This study has some limitations that should be noted. As we were limited to studying HULC with one cell line, and the validation of the LC-MS/MS data was performed only with one down-regulated protein. It remains for future experiments to further confirm the proteomic analysis results and to determine whether additional targets of HULC can be identified.

Conclusions
In the era of big data, it is important to identify molecules that can guide the direction of disease research through in-depth analysis of gene and protein profiles. Our study indicates that HULC significantly changes the proteomic characteristics of U87 cell line, and that PLA2G4A is negatively regulated by HULC knockdown in GBM cells. This study provides a new perspective on the pathogenesis of GBM, and also provides a potential target for GBM treatment.

Methods
Cell culture and transfection. The human GBM glioma cell line, U87, was obtained from China Center for Type Culture Collection (Wuhan, China) and maintained in Dulbecco's modified eagle's medium (DMEM) (BD, USA) supplemented with 10% fetal bovine serum (FBS) (BD, USA). The cells were grown at 37 °C in a 5% CO 2 atmosphere.
To generate lentivirus stable cell lines, cells were digested, resuspended, and plated in six-well dishes (Nest, China) at a density of approximately 10 × 10 5 cells per well, and then cultured under the same conditions for 24 h. The lentiviral vectors (LV3-shNC and LV3-shHULC) and lentiviral packaging were purchased from GenePharma (Shanghai, China). The overall transfection procedure was in accordance with the recommendations of the manufacturer. A 200 μl lentivirus stock solution was diluted 5 times with DMEM containing 10% FBS according to the manufacture's protocol. Infection enhancer polybrene (Sigma, USA) was added to a final concentration of 5 μg/ml. Stably-transfected cells were selected by puromycin (1 μg/ml, Sangon, Shanghai, China) and the green fluorescent protein (GFP) was observed under a fluorescence microscope (Olympus, Japan). After a 96 h in culture, the cells were harvested and stored at − 80 °C for subsequent experiments. Thus, two stable siRNA expressing cell lines were constructed, including HULC-siRNA and the negative control (NC). The sequence of shRNA targeting HULC was 5′-GAA CTC TGA TCG TGG ACA TTT-3′. (qRT-PCR). In a week, RNA was extracted from two samples using a total RNA extraction kit (QIAGEN, Germany). cDNA was synthesized according Tandem mass tags (TMT) labeling. The digested peptides were desalted using a Strata X C18 SPE column (Phenomenex) and freeze-dried in the vacuum environment. Peptides were reconstituted in 0.5 M NH 4 HCO 3 (Sigma, USA) and labeled using a TMT kit (Thermo Fisher Scientific, USA) according to the manufacturer's protocol.

High performance liquid chromatography (HPLC) fractionation. The Agilent 300 Extend C18
reversed-phase column (5 μm particles, 4.6 mm inner diameter, 250 mm length) was used to fractionate 0.2 mg peptides into 60 fractions with a gradient of 8% to 32% acetonitrile (Fisher Chemical, USA) under the condition of pH 9 over 60 min. The peptides were then combined into 9 components and freeze-dried by vacuum centrifuging.

LC-MS/MS analysis.
Two types of liquid chromatography mobile phases were first prepared. Phase A: an aqueous solution containing 0.1% formic acid (Fluka, USA) and 2% acetonitrile; Phase B: an aqueous solution containing 0.1% formic acid and 90% acetonitrile. Peptides were dissolved in phase A and separated using the EASY-nLC 1000 UPLC system (Thermo Fisher Scientific, USA) at a constant flow rate of 400 nL/min. The separation gradient was set to increase from 8 to 16% in phase B within 30 min, then increased to 30% within 25 min and 80% within 2 min, which was maintained for 3 min. The peptides were injected into the nanospray ionization source for ionization at a voltage of 2.0 kV. The precursor ions and the secondary fragments of the peptides were detected and analyzed using the Orbitrap Fusion Lumos high-resolution mass spectrometer (Thermo Fisher Scientific, USA). According to the data dependent acquisition (DDA) mode, the precursor ions with top 20 signal intensities after primary scan were fragmented with 32% fragmentation energy in the HCD collision cell. The secondary MS/MS scan then followed. The MS scan parameters are shown in Table 2.
The quantitative values of each sample in three replicates were obtained. The Pearson correlation coefficient was calculated between two pairs to assess whether the results of replicate samples were statistically consistent. Fold change was defined as the ratio of the average values of HULC-siRNA to NC. The relative quantitative value of each sample was log2 transformed to conform the data for normal distribution. Quantified data between the two groups were evaluated using a two-tailed test. Differentially expressed proteins were filtrated based on the following criteria: fold change was equal to or greater than 1.2 and less than 0.83, and the P value was less than 0.05. The protein ID was converted to UniProt ID, the corresponding Gene Ontology (GO) ID was obtained by searching the UniProt-GOA (www. http:// www. ebi. ac. uk/ GOA/) database, and GO was performed on differential protein annotations. For proteins that were not annotated, an algorithm software InterProScan (v.5.14-53.0, http:// www. ebi. ac. uk/ inter pro/) was used to predict their GO functions. Kyoto Encyclopedia of Genes and Genomes (KEGG) annotation was realized using KAAS (v.2.0, http:// www. genome. jp/ kaas-bin/ kaas_ main). KEGG Mapper (v2.5, http:// www. kegg. jp/ kegg/ mapper. html) was used to match the gene with the pathway in the database. The two-tailed Fisher's exact test was employed to evaluate the GO or KEGG pathway enrichment.
Colony formation, Transwell, and wound-healing assays. Cells were seeded at 200 cells per well in a 6-well plate and cultured for 12 days during which DMEM was renewed every 4 days. The cells were then fixed with formaldehyde (ZhanWang Chemical, China) for 30 min and stained with crystal violet (Beyotime, China) for 10 min. An inverted microscope (Olympus, Japan) was used to count the number of clones with more than 50 cells at 100 × magnification.
Transwell chambers (Corning, USA) were coated with 10% Matrigel (BD, USA). Cells were first starved with serum-free DMEM for 12 h, and 1 × 10 5 cells were then diluted with serum-free medium and seeded in the upper chamber. Complete medium was added to the lower chamber. After a 48-h incubation, cells remaining in the upper chamber were discarded. Chambers were fixed with formaldehyde for 30 min and stained with crystal violet for 10 min. Stained cells were photographed under a microscope with 200 × magnification.
We plated 3 × 10 5 cells/well in a 6-well plate and allowed the cells to grow to a density of approximately 70%. A 10 ul pipette tip was used to draw a straight line in the center of each well. Scraped cells were washed off 3 times with phosphate buffered saline (PBS). Cells were then cultured and photographed at 0 h, 24 h, and 48 h under a microscope with 100 × magnification.

Western blot analysis.
A total of 40 μg of cell lysates was electrophoresed using 10% SDS-PAGE (Beyotime, China) and transferred to PVDF membranes (Millipore, USA). The membranes were blocked with 5% skimmed milk powder for 2 h and then incubated with the primary antibody at 4 °C overnight. The membranes were then incubated for 1.2 h at room temperature with the secondary antibody conjugated to a horseradish peroxidaselabeled anti-mouse IgG (1:20,000) (Zsbio, ZB-2305). Protein bands were detected using an ECL kit (Thermo, USA) according to the manufacturer's protocol. Primary antibodies included mouse anti-PLA2G4A (1:500) (sc-376618, Santa Cruz, USA) and the internal control mouse anti-β-actin (1:1000) (TA-09, Zsbio, China).

Statistical analysis.
All experiments were performed in triplicate, and the data are expressed as mean ± standard error of the mean (SEM). Image J (National Institutes of Health, USA) was used to calculate cell numbers, scratch area, and band intensity. A two-tailed t-test was conducted using Graphpad Prism 7 software (Graphpad, USA), and the Fisher's exact test was carried out using the Perl module (v.1.31, https:// metac pan. org/ pod/ Text:: NSP:: Measu res:: 2D:: Fisher). P-values < 0.05 were considered statistically significant.

Data availability
The data used to support the findings of this study are available from the corresponding author upon reasonable request.