Plasmatic exosome-derived circRNAs panel act as fingerprint for glioblastoma

Subject characteristics

After analyzing the clinical information, we found no difference in the age and gender distribution for the GBM comparing with the healthy controls. Of the 120 GBM patients, 71 patients were proved with tumor size ≤3cm(diameter), the rest 49 was above 3cm. The tumor was infratentorial in 31 patients, while 89 patients' tumors were supratentorial. According to the WHO grade of GBM, 45 patients were labelled with I-II grade and 75 patients were labelled with III-IV grade (Table 1).

Exosome identification and circRNA expression profile in GBM and control group

Exosome specific markers such as CD9, CD63 and TSG101 were used as positive markers and calnexin was used as negative markers to detect human plasma samples.

We first confirmed the abundant levels of exosomes in plasma samples. In addition, we also confirmed the morphology of exosomes as irregular planets with a diameter span of 30-100nm under projection electron microscopy and with a clear outer membrane structure. (Figure 1A). After extracting the total RNA from the exosome, the circRNA microarray was further employed for high-throughput detection. Three samples from each group were randomly selected for microarray detection. As shown in Figure 1B, we screened the differentially expressed circRNAs with 2/0.25 as the fold change cut-off, and found that there were significantly different circRNA expression profiles in the plasma exosomes of patients in the GBM group compared with the healthy controls. Furthermore, P<0.05, CT value <35 and >75% of the sample detection efficiency were selected as the screening parameters. Finally, 102 circRNAs with abnormally high expression and 67 circRNAs with abnormally low expression were obtained as the targets for subsequent studies (Figure 1C).

Figure 1. The expression landscape of circulating exosome-derived circRNAs in GBM patients. (A) Western blot was used to detect the expression of CD9, CD63, TSG101 and calnexin in exosome isolated from plasma samples of GBM patients (upper panel). Plasma exosomes were observed using TEM (lower panel). (B) Cluster analysis of the different expression of the circRNAs in different groups. (C) Volcano plot presented the dysregulated circRNAs in GBM comparing with control group. Red plots indicated with significant difference. (D) The expression of circRNAs were confirmed by RT-PCR in training set. * indicated P<0.05, n.s. indicated no significance.

Abnormal elevation of the index is usually selected as the traditional diagnostic marker. In this study, in order to further screen potential biological markers for GBM, we selected the top 20 differential multiples in the above abnormally high expression as the targets for further verification.

Multiple staged validation of differentially expressed circRNAs in the training group and validation group

Multistage validation was performed for the 20 circRNAs included above. 20 randomly selected pairs of GBM patients and 20 pairs of healthy control plasma exosomes were selected as the test group for the first batch of screening, followed by further validation of the remaining 100 pairs of samples as the validation group. We found statistically significant differences in 10 of the 20 circRNAs in the test group. Further analysis showed that 5 circRNAs were expressed in a state consistent with the microarray results, while the remaining 5 circRNAs were presented with lower expression in GBM group (Figure 1D).

According to the data of the above training group, the rest 100 pairs of GBM and healthy control group samples were used for further verification. We detected the above circRNAs with upregulated expression and found that hsa_circ_0055202, hsa_circ_0074920 and hsa_circ_0043722 were still expressed in an elevated state in GBM. The expression of hsa_circ_0009760 was high in GBM, but there was no statistical difference. There was no significant difference in the expression of hsa_circ_0010178 (Figure 2).

Figure 2. Validation of three candidate circRNAs. Plasma from 100 paired GBM and healthy controls were enrolled. Data were presented as plot of the median and range of log-transformed relative expression level. * indicated p<0.05; ** indicated p < 0.01, n.s. indicated no significance.

Risk score analysis and diagnosis ability investigation

To further analyze the potential predicting ability of GBM from healthy controls. Receiver operating characteristic curve (ROC) was used to calculate the specific sensitivity and specificity. Firstly, the isolated circRNAs and three combined factors were analyzed in the training set (20 samples in each group as described above). As presented in Figure 3A, the areas under the curve (AUC) of hsa_circ_0055202, hsa_circ_0074920, hsa_circ_0043722 and combination of circRNAs (Merged) were 0.810, 0.670. 0.938 and 0.988, respectively. We continued the analysis using the same method in the validation group. As presented in Figure 3B, in validation group, the AUC of the four circRNAs and merged were 0.850, 0.625. 0.750 and 0.925, respectively.

Figure 3. The ROC curve based on risk score analysis for predicting GBM from healthy population. (A) ROC curve analysis was conducted for discrimination between GBM patients and control group in training set. Each group including 20 samples. (B) ROC curve analysis was conducted for discrimination between GBM patients and control group in validation set. Each group including 100 samples.

We defined the cut-off value as the maximum of sensitivity and specificity and calculated the positive predictive value (PPV) and negative predictive value (NPV). To distinguish between GBM and the control group, PPV and NPV in the training set were 90% and 90%, respectively. By using the same method in the validation set, we obtained the PPV and NPV were 84% and 86%, respectively (Table 2).

Stability detection of circRNAs in plasma samples

To further verify the stability of circRNA, we randomly selected 5 healthy control plasma samples and subjected the above samples to different treatments, including storage at -80° C for 7 days, storage at room temperature for 12 hours and 24 hours, repeated freeze-thaw treatment for 5 times and digestion with RNase enzyme. After different treatments, the exosome RNA was extracted and amplified by RT-PCR, and the changes of circRNA expression among the groups were compared. The results showed that the expression of circ_0055202, circ_0074920 and circ_0043722 did not significantly change after the above different treatments, suggesting that the expression of circ_0055202, circ_0074920 and circ_0043722 were relatively stable and detectable in plasma (Figure 4).

Figure 4. Stability detection of the four circRNAs in human plasma. RT-qPCR was applied for detecting the expression level of the three circRNAs. Data was presented as mean ± SEM with log-transformed. No significant difference was observed in each group.

The hsa_circ_0055202, hsa_circ_0074920 and hsa_circ_0043722 could promote the proliferation of GBM cells in vitro

The detailed function of the three circRNA including circ_0055202, circ_0074920 and circ_0043722 were further investigated. Firstly, as presented in Figure 5A, the expression abundance of the there circRNA were examined in GBM cells including U87, U251, Ln229, T98, and A172. Based on this, we chose the U87 cells for further function assay with a relatively high abundance level. We knocked down the three circRNAs through the shRNA technology and confirmed the suppression of the three circRNA in U87 (Figure 5B). The CCK8 assay and transwell assay was employed. We found that either suppression of circ_0055202, circ_0074920 or circ_0043722 could reduce the proliferation of GBM cells (Figure 5C); however, no significance was found in cell invasion (Supplementary Figure 1).

Figure 5. The hsa_circ_0055202, hsa_circ_0074920 and hsa_circ_0043722 promoted cell proliferation of GBM cell lines. (A) Relative expression of hsa_circ_0055202, hsa_circ_0074920 and hsa_circ_0043722 in GBM cell lines. (B) Relative expression of hsa_circ_0055202, hsa_circ_0074920 and hsa_circ_0043722 in cells treated with shRNA. (C) CCK8 assay in cells treated with circRNA shRNA. * indicated P < 0.05, ** indicated P < 0.01.

Subjects

We conducted the retrospective case-control study by enrolling total 120 GBM patients and paired 120 healthy controls. All the patients were diagnosed histologically in Haimen People’s Hospital range from June 2015 to May 2019. This study was approved by an institutional review board of the affiliated Haimen People’s Hospital (HMPH_IRB_01152). Written informed consents were obtained from all enrolled subjects. All plasma samples were collected using EDTA anticoagulant tubes. Half an hour after sampling, the samples were centrifuged at 3000 rpm for 10 minutes. The upper plasma was absorbed and stored at -80° C. All patients signed informed consent before sample collection. The relevant clinical data of all patients was summarized in Table 1. All experiments were performed in compliance with government policies and the Helsinki Declaration. All subjects were informed about the study and provided their consent prior to specimen collection.

Exosome isolation and identification

Exoquick Exosome precipitate solution (System Biosciences, Mountain View, CA, USA) was used to extract exosomes from plasma samples. Exosomes were efficiently isolated using SBI’s ExoQuick/ExoQuick-TC solutions, resuspend exosome pellet in 350 μl of Lysis Buffer. Further, biomarker detection and morphological identification were carried out for the extracted exosomes. Exosome biomarkers including CD9, CD63 and TSG101 were first used to detect their expression levels. In addition, calnexin was used as negative control. Then the morphology of exosomes extracted was observed by transmission electron microscopy (TEM).

RNA extraction and quantitative real-time PCR (qRT-PCR)

We used Trizol (Invitrogen, CA, USA) method for RNA extraction, and the specific method was carried out according to its instructions. After RNA extraction, its concentration and purity were determined by ND-1000 (Thermo, CA, USA), and RNA was clarified by 1% agarose electrophoresis to ensure RNA integrity. In order to clarify the stability of circRNA expression, we used the combination of internal and external parameters for correction. The internal reference was GAPDH, and the external reference was cel-miR-39 (ABI, CA, USA). The circRNA was amplified by random primer method and amplified in ABI 7900 (ABI, CA, USA). The relative expression of circRNA was calculated using the 2^-ΔΔCq method.

circRNA microarray

The microarray-based circRNA expression profiling was conducted by Capitalbio (Beijing, China) using plasma RNA samples. We conducted the sample preparation and microarray hybridization in accordance with instrument maufacture's manual (Arraystar, Inc). Briefly, samples were enriched for circRNAs by treating them with RNase R to degrade linear RNA, after which random primers were used to reverse transcribe enriched circRNAs into cDNA. This reduces the initial 37,681 input genes in each sample to about 15,000 detected genes. A bilateral Mann-Whitney U test performed by Wilcox was employed. Finally, we use the Benjamin-Hochberg correction (R. P. Adojust function) to explain the multiple tests.

Multiple phase validation

The screening phase was divided into training group and validation group. 20 samples each group were enrolled in training group while the validation group contained another 100 samples in each group. Groups were named as healthy volunteers and GBM patients.

Groups were named as healthy volunteers and GBM patients. Risk scores analysis was used to analyze the predictive power of certain circRNAs. In brief, the upper 95% reference interval (95%CI) of each circRNA value in the control group was used as the cut-off value for a certain circRNA expression level. If circRNA expression was higher than 95% CI in this sample, we rated it as 1, and if it was lower than 95% CI, we rated it as 0. The risk score was defined as a linear combination of the expression levels of each circRNA. The risk score for the circRNAs was calculated using weight regression coefficients derived from univariate logistic regression analyses for each circRNAs. The samples were ranked by RSF and then divided into a high-risk group (representing the GBM group) and a low-risk group (representing the predictive control individuals).

Cell lines and cell culture

GBM cell lines, including U87, U251, Ln229, T98, and A172 cells, were purchased from the Cell Bank of Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China), and were cultured and stored according to the guidelines of the cell bank. The culture medium used was Dulbecco’s modified Eagle’s medium (DMEM; Winsent, Quebec, Canada) containing 10% fetal bovine serum (FBS), 100 U/ml penicillin, and 100 μg/ml streptomycin. All the cell lines were incubated in a 5% CO2 humidified incubator at 37° C.

Cell proliferation assay

In each well of a 96-well plate, 1000 cells were plated in total. The CCK-8 reagent (Dojindo Seed, Japan) was added directly to the culture medium at the stated time (24 h, 48 h, 72 h, 96 h). Cells were then incubated at 37° C for 2 hours, and the optional density (OD) value was measured at 450 nm by microplate reader (BioTek Instruments, USA). These studies were carried out three times.

Transwell assay

Cells were seeded in upper chambers with 200ul of serum-free medium for Transwell assays. The transwell chamber (Corning, USA) was paved with matrigel mix (BD Biosciences, USA) for invasion assays. The bottom chamber medium contained 10% FBS to attract upper cells. The upper chambers were set and stained for 20 minutes with a 0.5% crystal violet solution after a 48-hour incubation period. The cell lines were photographed and counted in five distinct regions.

Statistical analysis

SPSS 20.0 (IBM, SPSS, Chicago, USA) and GraphPad Prism 5 were used for statistical analysis. To explore whether two or more classes were statistically significant, we used the Student's t-test and one-way ANOVA. The data was interpreted as Mean±SEM. For all studies, P<0.05 between groups were considered statistically significant.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.