Proteomics identifies EGF-like domain multiple 7 as a potential therapeutic target for epidermal growth factor receptor-positive glioma

2.1 Clinical specimens and bioinformatics

Frozen glioma tissues were collected from the Glioma Center of Xiangya Hospital (Changsha, Hunan, China). In total, 100 EGFR-positive and 100 EGFR-negative glioma samples were collected for protomic analysis. The criteria for patient selection were as follows: age of ≥ 18 and ≤ 70 years; subtype confirmed by pathology; patient did not receive glucocorticoid treatment 5 days before enrollment; no anti-EGFR treatment was received before collecting the samples. Consensus-clustering dataset was used to analyze the MS data from different batches, and SC3 software (Vladimir Kiselev, Hinxton, Cambridge, UK) was used to generate the cluster map [13]. The gene set enrichment analysis was used to select key genes in clustering results and process proteomics data [14]. In addition, clinical data was provided by two large glioma gene expression databases: 277 glioma patients from the Chinese Glioma Genome Atlas (CGGA; http://www.cgga.org.cn) and 665 glioma samples of various pathological subtypes from the Cancer Genome Atlas (TCGA; https://gdc.cancer.gov).

2.2 Isobaric tags for relative and absolute quantitation (iTRAQ) proteome analysis

We selected 20 glioma samples for proteomic characterization using iTRAQ (SCIEX, Foster City, CA, USA). Cluster sampling method, a randomization-based sampling method, was used in this selection process. Cluster sampling was performed according to different pathological subtypes. Proteomic analysis was carried out by using the iTRAQ-labeled technique on the same mass spectrometer with stable quality control. iTRAQ 8-plex was completed for all 20 tumor tissues for discovery and validation of proteomic data (Figure 1). Tumor tissue samples were kept in cryovials at -80°C until cryopulverization using a CP02 Cryprep Pulverizer (Covaris, Woburn, MA, USA). For each case, 40 mg of tissue was used to provide an expected yield for each set of 1.5-2 mg protein based on 4%-5% recovery. To obviate systematic deviation, sample processing was block randomized, with each intrinsic subtype proportionally represented in each processing tranche. The samples were mixed with the iTRAQ reagent for 1 h in dark. A volume of 0.1% trifluoroacetic acid nine times the sample volume was added into the samples. Desalted peptides were labeled with 8-plex iTRAQ reagents (SCIEX) according to the manufacturer's instructions. All peptides were separated using an online nanoflow Proxeon EASY-nLC 1000 UHPLC system (Thermo Fisher Scientific, Waltham, MA, USA) and analyzed on a Benchtop Orbitrap Q Exactive HF mass spectrometer (Thermo Fisher Scientific).

2.3 Immunohistochemistry (IHC)

Generally, specimens were fixed in formaldehyde solution, routinely processed, and then enclosed in paraffin. Five-micron-thick sections were prepared. These sections were deparaffinized and rehydrated. H₂O₂ at a concentration of 0.3% was used to block endogenous peroxidase activity. Sections were put in 10 mmol/L citrate antigen-unmasking solution (pH 6.0) heating in oven at 90°C for 15 min. All sections were incubated with 10% fetal bovine serum (FBS) (Thermo Fisher Scientific) at 37°C for 10 min. Next, sections were incubated with primary antibodies (1:100 dilution) overnight at 4°C, followed by incubation with biotin-labeled secondary antibodies (1:100 dilution) for 1 h at 37°C and then incubated with 3,3’-diaminobenzidine substrate solution (Zhongshan Bio Corp, Zhongshan, Guangdong, China), counterstained with hematoxylin (Zhongshan Bio Corp), and visualized under a light microscope. The antibodies used in IHC included antibodies against EGF-like domain multiple 7 (EGFL7) (sc-373898, dilution 1:100, Santa Cruz Biotechnology, Dallas, TX, USA), EGFR (MA5-13070, dilution 1:100, Invitrogen, Waltham, MA, USA), marker of proliferation Ki67 (KI67) (ab15580, dilution 1:200, Abcam, Cambridge, MA, USA), phosphorylated extracellular signal-regulated kinase (p-ERK)1/2 (ab214362, dilution 1:200, Abcam), and phosphorylated pyruvate kinase M1/2 (p-PKM1/2) (#3827, dilution 1:200, Cell Signaling Technology, Danvers, MA, USA). The staining intensity was scored on a scale of 0 to 3 (0, negative; 1, slightly positive; 2, moderately positive; 3, intensely positive) by two experienced pathologists without knowledge of the diagnosis. A score of 0 or 1 indicated low expression, whereas a score of 2 or 3 indicated high expression. If there was an inconsistency, the two senior pathologists simultaneously reviewed the images to achieve a consensus.

2.4 Cell lines and culture conditions

EGFR-active human GBM cell lines, U87-MG and U251-MG, were obtained from Institute of Biochemistry and Cell Biology, Chinese Academy of Science (Shanghai, China). Both cell lines were grown and cultured in Dulbecco's modified Eagle's medium (DMEM) nutrient mixture (Gibco, Grand Island, NY, USA) supplemented with 10% FBS, 2 mmol/L glutamine (Sigma, St. Louis, MO, USA), 100 units/mL of penicillin (Sigma), and 100 mg/mL of streptomycin (Sigma) at 37°C with 5% CO₂.

2.5 Drug treatments and lentiviral infection

Selumetinib (Selleck Chemicals, Houston, TX, USA) was used to inhibit MEK1/2 phosphorylation. GBM cells were treated with 0.195, 0.39, 0.78, 1.56, 3.125, 6.25, 12.5, 25, 50, and 100 µmol/L selumetinib dissolved in dimethyl sulfoxide (DMSO, Sigma-Aldrich, St. Louis, MO, USA) for 72 h.

Lentiviruses containing an EGFL7 inhibitor sequence (shEGFL7) or negative control sequences were purchased from GeneChem (Shanghai, China) (Supplementary Table S1). The transfection process follows GeneChem's instructions. U87-MG and U251-MG cells were infected with lentiviruses at 70% confluence. The infected cells (named U87-EGFL7kd and U251-EGFL7kd) were harvested for further use after 48 h. A universal negative control lentiviral vector was used as control.

2.6 qRT-PCR

Total RNA was extracted using the TRIzol reagent (Life Technologies, Rockville, MD, USA). To detect EGFL7 mRNA, quantitative real-time reverse transcription-polymerase chain reaction (qRT-PCR) was performed using a Reverse Transcription System (Promega, Madison, WI, USA) according to the manufacturer's instructions. All EGFL7 expression data were normalized to that of glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The primers for EGFL7 and GAPDH are shown in Supplementary Table S2.

2.7 Cell growth assay

Cells were plated at 10⁴ cells per well in 96-well plates with three replicate wells per sample. Then, 20 µL of Cell Counting Kit-8 (CCK-8) reagent (Beyotime, Shanghai, China) was added into each well for 4 days after plating, and the cells were incubated for additional 1 h. At last, 200 µL of DMSO was added to each well to dissolve the precipitate. The optical density (OD) was measured at a wavelength of 450 nm. The results are presented as the mean ± standard deviation (SD), which were derived from triplicate samples of at least three independent experiments.

2.8 Immunofluorescence analysis

Cultured U87-MG and U251-MG cells were fixed using 4% paraformaldehyde in phosphate-buffered saline (PBS) (Santa Cruz Biotechnology) for 20 min at room temperature. Thereafter, the cells were permeabilized using 0.2% Triton X 100 (Solarbio, Beijing, China) in PBS for 15 min. Non-specific binding was blocked by incubation with 5% goat serum (Thermo Fisher Scientific) in PBS for 30 min. The primary anti-EGFL7 antibodies (sc-373898, dilution 1:100, Santa Cruz Biotechnology) and anti-EGFR antibodies (MA5-13070, dilution 1:100, Invitrogen) were used for immunofluorescence. The slides were then incubated with appropriate antibodies in antibody dilution buffer overnight at 4°C, followed by a further incubation at room temperature for 1 h with Alexa Fluor^® 488 donkey anti-mouse IgG antibody (A-21202, dilution 1:500, Invitrogen), anti-rabbit IgG Fab2 Alexa Fluor^® 488 antibody (ab181346, dilution 1:500, Cell Signaling Technology) or Alexa Fluor^® 594 donkey anti-rabbit IgG antibody (R37119, dilution 1:500, Invitrogen). The samples were washed and embedded into mounting medium with 4,6-diamino-2-phenylindole (DAPI; nuclear DNA was labeled in blue; VECTOR, Burlingame, CA, USA). Microscopy analysis was performed using a confocal laser scanner microscope (Olympus, Tokyo, Japan).

2.9 Nude mouse tumor xenograft model and treatment

For intracranial xenograft experiment, 4-week-old Bagg albino (BALB/c) nude mice (Animal Center of the Cancer Institute at the Chinese Academy of Medical Science, Beijing, China) were assigned to three groups with different conditions: 1) U87-MG cells only, 2) U87-EGFL7kd cells, and 3) U87-MG with selumetinib treatment. Each group contains 9 mice. Tumor cells were intracranially injected into mice. All experimental procedures, including animal feeding, intracranial tumor implantation, and specimen collection, were performed according to Xiangya Hospital Medical Research Used Animal Principle policies. U87-MG (5 × 10⁵) cells expressing luciferase were intracranially injected into athymic nude mice. After 3 days, 50 mg/mL selumetinib dissolved in 5 µL of DMSO were intratumorally injected every 3 days. Tumor volume was monitored by using in vivo bioluminescence imaging spectrum system (PerkinElmer, Waltham, MA, USA), which was carried out at day 14 after the injection to measure tumor volume using the formula: volume = (length × width²) / 2. When the mice hunched posture for more than 48 h and appeared labored breathing and cyanosis (skin or mucous membranes in blue), it is considered a humane endpoint. The mice were euthanized, and their brains were dissected for further pathological examination. Paraffin-embedded tissue sections were used for IHC analysis [15].

2.10 Statistical analysis

All statistical calculations were running on GraphPad Prism 6 (GraphPad Software, San Diego, CA, USA). Differences between two groups were determined using Student's t-test. Gene profiling data were analyzed using either BRB-Array Tools (Richard Simon, Bethesda, MD, USA) [16] or Gene Pattern software (Broad Institute, Cambridge, MA, USA) [17]. Proteomics data were processed using MaxQuant software (Jurgen Cox, Martinsried, Bavaria, Germany) [18] at 1% false discovery rate. For the screening of significantly differentially expressed proteins, we set fold change greater than 1.3 and P value < 0.1 [19].

Kaplan-Meier analysis and log-rank tests were used to evaluate the prognostic significance of EGFL7 expression level for patient overall survival (OS), which was defined as the duration from diagnosis to death of any cause. The patients survived at the end of the study period were censored. One-way analysis of variance (ANOVA) was used to test for differences among three or more groups, and the least significant difference post-hoc test was used to obtain individual P values. P values < 0.05 were considered significant. The Pearson product moment correlation test was used to evaluate the relationships between EGFL7 and p-ERK and p-PKM2 with semi-quantitative scores (using a scale from 0 to 10).

3.1 Demographic data of patients and protein identification in glioma tissues

We selected 100 EGFR-positive and 100 EGFR-negative tumor tissues from 200 patients with glioma diagnosed at Xiangya hospital (Table 1). In World Health Organization (WHO) grade IV glioma, classic-like and mesenchymal-like classification glioma, EGFR showed a higher positive rate (P < 0.001). OS was shorter in EGFR-positive patients than in EGFR-negative patients (Figure 2A). Compared with patients with other glioma pathological subtypes, patients with GBM had the shortest OS (Figure 2A).

iTRAQ proteomic analysis on the 20 selected cases (10 EGFR-positive cases and 10 EGFR-negative cases) showed that more proteins were identified from EGFR-positive tumors than from EGFR-negative tissues (Figure 2B). In particular, 4531 proteins per EGFR-positive glioma and 4324 proteins per EGFR-negative glioma were identified (Figure 2C). We identified 322 significantly different proteins (approximately 7% of the 4743 spots) using the data obtained from the proteomic analysis. Among them, 181 proteins were up-regulated and 141 proteins were down-regulated in EGFR-positive tumors compared with those in EGFR-negative tumors (Figure 2D).

3.2 Signature proteins and pathways in EGFR-positive glioma

Consensus-clustering analysis demonstrated dramatically different proteomic profiles between EGFR-positive and -negative tumor samples, highlighting the high heterogeneity among tumor samples (Figure 3A).

The proteins shown in Figure 3A are typical examples of signaling molecules involved in the EGFR network. In total, more than 100 interacting proteins and more than 200 EGFR-related substrates are described in the literature. Using a non-negative matrix factorization consensus-clustering analysis, we identified 9 major cell function groups in the EGFR-positive tumor cohort. Signaling pathways including angiogenesis, differentiation, apoptosis, survival, degradation, transcription, and endocytosis were significantly enhanced in the EGFR-positive tumor cohort (Figure 3B).

3.3 Associations of EGFL7 expression with patient prognosis and glioma cell growth in vitro

Focusing on proteins that directly interact with EGFR noted in the previous subsection, we identified a member of the EGF protein family, EGFL7, of which the expression intensity was higher in EGFR-positive tumors than in EGFR-negative tumors. By analyzing the genome-wide sequencing data of human glioma cases recorded in CGGA and TCGA, we found that the EGFL7 gene was highly expressed in 469 of 942 patients with glioma. EGFL7 expression level was significantly associated with OS in our proteomic analysis cohort, the CGGA primary glioma cohort, and the TCGA GBM cohort (Figure 4A). Kaplan-Meier analysis indicated that an elevated level of EGFL7 was a risk factor for poor prognosis in patients with glioma in the three cohorts (Table 2).

To confirm the role of EGFL7 in glioma, we specifically knocked down EGFL7 in U87-MG and U251-MG cells. The immunofluorescence results showed EGFL7 protein signals were weak in cell lines silenced for EGFL7 (Figure 4B). A CCK-8 assay showed that cell proliferation was significantly suppressed in U87-EGFL7kd and U251-EGFL7kd cells (Figure 4C). We found the survival rate was decreased in both U87-MG and U251-MG cells after treatment with selumetinib (Figure 4D).

3.4 Associations of EGFL7 expression with xenograft tumor growth and mouse survival

We explored the effects of EGFL7 in an intracranial xenograft mouse model. To better understand the heterogeneity of GBM, nude mice received intracranial injection of U87-MG cells, intracranial injection of U87-EGFL7kd cells, or intracranial injection of U87-MG cells with selumetinib treatment.

The intracranial signal intensity of the EGFL7 knockdown group and the selumetinib treatment group indicated differences in tumor volume on the 14th day after tumor cell injection (Figure 5A and B, and Supplementary Fig. S1). The mouse survival analysis (Figure 5C) and IHC staining for KI67 (Figure 5D) indicated that selumetinib effectively inhibited the growth of U87 glioma in mice and, to a certain extent, prolonged the survival of mice.

Quantification of the staining intensity showed that both ERK and PKM2 expression were positively correlated with EGFL7 expression among intracranial tumor specimens from mice (Figure 5E). Furthermore, the local treatment of selumetinib inhibited tumor growth (Figure 5A) and reduced the phosphorylation of ERK and PKM2 (Figure 5E).

ETHICS APPROVAL AND CONSENT TO PARTICIPATE

This study was approved by the Research Ethics Committee of Xiangya Hospital. Written informed consent was obtained from all patients.

CONSENT FOR PUBLICATION

Not applicable.

AVAILABILITY OF DATA AND MATERIALS

The CGGA and TCGA datasets analyzed for this study are available as indicated in the Materials and Method section. The proteomic cohort data are available from the corresponding author upon reasonable request.

CONFLICT OF INTEREST

We declare that no conflict of interest exits in the submission of this manuscript. Additionally, the manuscript is approved by all authors for publication. I would like to declare on behalf of my co-authors that the work described was original research that has not been published previously.

AUTHORS’ CONTRIBUTIONS

Dr. Fei-yi-fan Wang conducted the major molecular biology and animal experiments; Dr. Si-yi Wang-gou designed the experiment and processed the data; Dr. Hang Cao and Dr. Nian Jiang conducted the cell culture and animal experiments; Dr. Qi Yang and Dr. Qi Huang participated in the experimental design and paper writing; Prof. Chun-hai Huang participated in the paper writing; Prof. Xue-jun Li managed the research, designed the experiments, and provided financial support.

FUNDING

This work was supported by grants from the National Natural Science Foundation of China (81770781 to XJL), Hunan Provincial Innovation Foundation for Postgraduate (CX2018B114 to FYFW), the Fundamental Research Funds for the Central Universities of Central South University (2018zzts043 to FYFW), and China Scholarship Council (201806370130 to FYFW).