Their findings, published in the journal ACS Applied Materials & Interfaces, have indicated that regular urine tests could help early detection and treatment of brain tumors, possibly leading to improved patient survival.
When brain tumors are detected by CT or MRI, in many cases, they have already grown too large to be fully removed, which could lower patients’ survival rate. From this perspective, accurate, easy, and inexpensive methods of early brain tumor detection are strongly desired.
As a diagnostic biomarker of cancerous tumors, microRNAs (tiny molecules of nucleic acid) have recently received considerable attention. MicroRNAs are secreted from various cells, and exist in a stable and undamaged condition within extracellular vesicles in biological fluids like blood and urine.
Nagoya University researchers focused on microRNAs in urine as a biomarker of brain tumors. “Urine can be collected easily without putting a burden on the human body,” says Nagoya University Associate Professor Atsushi Natsume, a corresponding author of the study.
“Urine-based liquid biopsy hadn’t been fully investigated for patients with brain tumors, because none of the conventional methodologies can extract microRNAs from urine efficiently in terms of varieties and quantities. So, we decided to develop a device capable of doing it.”
The new device they developed is equipped with 100 million zinc oxide nanowires, which can be sterilized and mass-produced, and is therefore suitable for actual medical use. The device can extract a significantly greater variety and quantity of microRNAs from only a milliliter of urine compared to conventional methods.
Their analysis of microRNAs collected using the device from the urine of patients with brain tumors and non-cancer individuals revealed that many microRNAs derived from brain tumors actually exist in urine in a stable condition.
Next, the researchers examined whether urinary microRNAs can serve as a biomarker of brain tumors, using their diagnostic model based on the expression of microRNAs in urine samples from patients with brain tumors and non-cancer individuals.
The results showed that the model can distinguish the patients from non-cancer individuals at a sensitivity of 100% and a specificity of 97%, regardless of the malignancy and size of tumors.
The researchers thus concluded that microRNAs in urine is a promising biomarker of brain tumors.
The researchers hope that their findings will contribute to early diagnosis of aggressive types of brain cancer, like glioblastomas, as well as other types of cancer. Dr. Natsume says, “In the future, by a combination of artificial intelligence and telemedicine, people will be able to know the presence of cancer, whereas doctors will be able to know the status of cancer patients just with a small amount of their daily urine.”
Assembly-Type Microfluidic Nanowire Device
We designed an assembly-type microfluidic nanowire device for extracting urinary miRNAs and acquiring miRNA expression profiles (Movie S1). We fabricated the assembly-type microfluidic nanowire device by two processes: first, we grew ZnO nanowire scaffolds from a thermally oxidized chromium layer on a silicon (Si) substrate; and second, we assembled the ZnO nanowire scaffolds, cyclo-olefin polymer (COP) resin microfluidic substrate, COP resin substrate, two stainless steel holders, and polyether ether ketone (PEEK) tubes into the device (Figure 1A).
The device was connected to PEEK tubes for introduction of urine and lysis buffer and collection of flow-through urine and miRNA-containing solution (Figure 1B). Since no bonding process was required, each component of the assembly-type microfluidic nanowire device could be sterilized, such as by autoclave treatments, ethanol treatments, and dry-heat treatments, to prevent contamination by miRNAs in saliva and sweat of persons handling the device. Furthermore, by simplifying the fabricating processes, fabrication time was shortened and the device could be mass-produced.
In situ extraction of urinary miRNAs within 40 min was demonstrated in two steps: first, a 1 mL urine sample was introduced into the device to capture EVs and EV-free miRNAs (20 min); and second, a 1 mL lysis buffer aliquot was introduced to extract EV-encapsulated miRNAs from the captured EVs by dissolving them and to collect EV-free miRNAs by releasing from the nanowires (20 min) (Figure 1C). The EVs and EV-free miRNAs are negatively charged. Since the ZnO nanowires have a positively charged surface, they can achieve highly efficient collection of EVs and EV-free miRNAs. Moreover, the large surface area of the nanowires and the microfluidic structure of the device contribute to the increase of capture efficiency.(14)
Performance of the Assembly-Type Microfluidic Nanowire Device
To investigate whether our assembly-type microfluidic nanowire device could extract urinary miRNAs efficiently, we extracted urinary miRNAs from noncancer individuals. The quality of miRNAs extracted from all samples was checked with an Agilent 2100 Bioanalyzer. Since a high peak was observed in the miRNA (25 nt) region but not in the 18S (1900 nt) or 28S (4700 nt) ribosomal RNA region (Figure 1D), the extracted urinary miRNAs were high-purity miRNAs. All urinary miRNAs were analyzed using a miRNA microarray and that yielded comprehensive miRNA expression profiles, which included 2565 species of miRNAs.
Compared to the ultracentrifugation method (n = 3) or ExoQuick-TC (n = 3), the assembly-type nanowire device (n = 100) showed a significantly higher number of extracted miRNAs (P = 0.01 or 0.03, respectively) and fluorescence intensity (both P < 0.01) in the miRNA microarray analysis of noncancer individuals (Figures 1E and S1A).
We also compared the number of miRNAs identified in two or more cases between our nanowire device and the two conventional methods. Although miRNAs identified by ultracentrifugation (n = 22/2388 [0.92%]) and ExoQuick-TC (n = 15/2381 [0.63%]) accounted for less than 1% of the total, miRNAs identified by the nanowire device accounted for more than 50% (Figure S1B).
Conclusively, the assembly-type nanowire device can extract a greater variety and quantity of urinary miRNAs than that by ultracentrifugation and ExoQuick-TC. The assembly-type nanowire device required less time and urine volume compared to the conventional methods in miRNA extraction (Table S1). In addition, to assess the reproducibility of the assembly-type nanowire device, we extracted miRNAs from duplicate urinary samples and performed microarray analyses, and showed that the nanowire device had high reproducibility of extracted miRNA species (R2 = 0.9849) (Figure 1F). The assembly-type microfluidic nanowire device was found to be sterilizable, mass-producible, time-saving for miRNA extraction, and efficient and reproducible for miRNA extraction.
Origin of Urinary miRNA Characteristics of Patients with CNS Tumors
To assess whether the urinary miRNAs that showed significantly higher or lower expression in patients with CNS tumors were derived from the tumor itself, we established two glioblastoma (GBM) organoids (named GBOs) from patients with GBM (GBM1 and GBM2). Since organoids are in vitro three-dimensional (3D) cell aggregates derived from the primary tissue, the organoids have been reported to maintain the histological features, gene expression, and mutational profiles of their corresponding parental tumors,(16,17) and organoid-secreted EV-encapsulated miRNAs and EV-free miRNAs could be detected in organoid culture supernatants as well as tumor cells.(18)
Targeted-capture sequencing for GBOs and corresponding parental tumors demonstrated that most gene mutations were consistent between them, 95.1% in GBM1 and 100% in GBM2 (Figure S2). Actually, the GBOs showed similar histological features as well as mutational profiles to parental tumors (Figure 2A). Our nanowire device featured the capability of extracting miRNAs from cell culture supernatants (Figure S3). A 1 mL cell culture supernatant sample was introduced into the device to capture EVs and EV-free miRNAs, and then, a 1 mL lysis buffer aliquot was introduced to extract EV-encapsulated miRNAs from the captured EVs by dissolving them and to collect EV-free miRNAs by releasing from the nanowires.
We extracted miRNAs from culture supernatants of two GBOs and immortalized human astrocytes (NHAs) (n = 2) with the nanowire device and performed miRNA microarray analyses. Differentially expressed miRNAs (DEMs) of GBO were defined as miRNAs that showed 1.5 times higher or lower expression (absolute fold-change >1.5) in culture supernatants of each GBO compared to the average expression in culture supernatants of NHAs.
Figure 2. GBM organoid-derived miRNAs. (A) Parental tumor and corresponding GBM organoid (GBO). Parental tumor (upper row): sample H&E staining and immunostaining for glial fibrillary acidic protein (GFAP) images. GBO (lower row): sample bright-field image of GBO, sample H&E staining, and immunostaining for GFAP images of GBO. Scale bars, 50 μm. Microvascular proliferation (yellow arrows and yellow dotted squares) and tumor cells (black arrows) were observed in both the parental tumor and GBO. (B) Schematic diagram of the analyses of miRNAs extracted from organoids and urine of the patients with parental tumors. (C) Percentages of organoid-derived DEMs in urinary DEMs of each group. Data are represented as mean. DGs, diffuse gliomas; CNS-Ts, central nervous system tumors (not including diffuse gliomas); and NCs, noncancer individuals. (D)
Number of organoid-derived DEMs that indicate the same (up- or downregulated) tendency of the expression level of urinary DEMs in each group. (E) Venn diagram of overlapped downregulated miRNAs from two groups (left panel). Results of GO analysis suggested that the suppression of these miRNAs could activate several genes associated with tumorigenesis in GBM (right panel).We also extracted urinary miRNAs from two patients for whom organoids were established and 117 patients with CNS tumors, and performed miRNA microarray analyses (Figure 2B and Table 1).
Urinary DEMs were defined in urine of patients with parental tumors, diffuse gliomas (GBMs and lower-grade gliomas (LGGs); n = 63), and other CNS tumors (not including diffuse gliomas; n = 54) and noncancer individuals who were randomly selected from 100 noncancer individuals (n = 34), as compared with the other noncancer individuals (n = 66) (absolute fold-change >1.5).
On investigating whether DEMs of each GBO corresponded to urinary DEMs of patients with parental tumors, diffuse gliomas, and other CNS tumors and noncancer individuals, DEMs of the organoid culture supernatant were much more frequently identified as DEMs in urine of patients with parental tumors (73.4%), diffuse gliomas (30.6%), and other CNS tumors (25.0%) than the urine of noncancer individuals (3.9%) (Figure 2C).
To delineate that organoid-derived miRNAs can be detected in blood as well as urine, we obtained serum miRNA data of patients with CNS tumors and noncancer individuals from the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database. Serum DEMs were defined in serum of patients with diffuse gliomas (n = 170) and other CNS tumors (not including diffuse gliomas; n = 89) and noncancer individuals who were randomly selected from 157 noncancer individuals (n = 52), as compared with the other noncancer individuals (n = 105) (absolute fold-change >1.5). DEMs of the organoid culture supernatant were much more frequently identified as DEMs in serum of patients with diffuse gliomas (62.2%) and other CNS tumors (60.9%) than serum of noncancer individuals (12.4%), the same as we found with urinary DEMs (Figure S4). These results suggested that many tumor-derived DEMs could be detected in urine as well as serum of patients with parental tumors directly.
Table 1. Participants’ Characteristicsa
| characteristics | training set (n = 134) | validation set (n = 68) | P | exploratory set (n = 15) | organoid generation (n = 2) |
|---|---|---|---|---|---|
| Patients with CNS Tumor | |||||
| total | 68 | 34 | 15 | 2 | |
| age, median (range), y | 53 (21–86) | 56 (14–81) | 0.24 | 34 (2–63) | 51 (44–58) |
| gender | |||||
| male | 32 (47.1) | 21 (61.8) | 0.21 | 7 (46.7) | 0 |
| female | 36 (52.9) | 13 (38.2) | 8 (53.3) | 2 (100) | |
| miRNA concentrations, average | 0.40 ng/μL | 0.58 ng/μL | 0.27 | 0.75 ng/μL | 0.23 ng/μL |
| histologic findings | 0.67 | ||||
| diffuse glioma | 40 (58.8) | 23 (67.7) | |||
| glioblastoma | 16 | 12 | N/A | 2 (100) | |
| glioblastoma, IDH-mutant | 4 | 3 | N/A | 1 | |
| glioblastoma, IDH-wild type | 12 | 9 | N/A | 1 | |
| lower-grade glioma | 24 | 11 | N/A | N/A | |
| diffuse astrocytoma, IDH-mutant | 5 | 3 | N/A | N/A | |
| Diffuse astrocytoma, IDH-wild type | 3 | 3 | N/A | N/A | |
| Anaplastic astrocytoma, IDH-mutant | 2 | 1 | N/A | N/A | |
| Anaplastic astrocytoma, IDH-wild type | 3 | 1 | N/A | N/A | |
| oligodendroglioma, IDH-mutant | |||||
| and 1p/19q-codeleted | 9 | 3 | N/A | N/A | |
| anaplastic oligodendroglioma, | |||||
| IDH-mutant and 1p/19q-codeleted | 2 | 0 | N/A | N/A | |
| meningioma | 21 (30.9) | 7 (20.6) | N/A | N/A | |
| WHO grade I | 20 | 6 | N/A | N/A | |
| grade II | 0 | 1 | N/A | N/A | |
| grade III | 1 | 0 | N/A | N/A | |
| Schwannoma | 5 (7.4) | 2 (5.9) | N/A | N/A | |
| metastatic tumor | 2 (2.9) | 2 (5.9) | N/A | N/A | |
| CNS neuroblastoma | N/A | N/A | 3 | N/A | |
| pilocytic astrocytoma | N/A | N/A | 2 | N/A | |
| CNS embryonal tumor, NOS | N/A | N/A | 1 | N/A | |
| DNT | N/A | N/A | 1 | N/A | |
| ependymoma | N/A | N/A | 1 | N/A | |
| epidermoid cyst | N/A | N/A | 1 | N/A | |
| gangliocytoma | N/A | N/A | 1 | N/A | |
| ganglioglioma | N/A | N/A | 1 | N/A | |
| MVNT | N/A | N/A | 1 | N/A | |
| pituitary adenoma | N/A | N/A | 1 | N/A | |
| primary CNS lymphoma | N/A | N/A | 1 | N/A | |
| SFT/HPC | N/A | N/A | 1 | N/A | |
| Noncancer Individuals | |||||
| total | 66 | 34 | N/A | N/A | |
| age, median (range), y | 41.5 (20–70) | 40 (20–75) | 0.81 | N/A | N/A |
| gender | |||||
| male | 36 (54.5) | 22 (64.7) | 0.4 | N/A | N/A |
| female | 30 (45.5) | 12 (35.3) | N/A | N/A | |
| miRNA concentrations, average | 0.28 ng/μL | 0.25 ng/μL | 0.55 | N/A | N/A |
Unless otherwise specified, data are presented in number (percentage).The numbers of DEMs in both organoid culture supernatant and urine of each patient were 1299 and 603 (Figure 2D), respectively, and among them, 527 miRNAs (7 upregulated miRNAs and 520 downregulated miRNAs) were overlapped in both cases (Figures 2E and S5). Since it has been recognized that miRNAs regulate gene expression by binding some messenger RNAs (mRNAs),(5) we searched in ENCORI (starBase v3.0), which is an open-source platform for studying miRNA–mRNA interactions, to identify the target genes of DEMs that overlapped in two GBOs. Using ENCORI, we were able to obtain biological targets of miRNAs by searching in seven databases (microT, miRanda, miRmap, PicTar, PITA, RNA22, and TargetScan).
Four out of seven upregulated DEMs that overlapped in two GBOs were listed in ENCORI, while 117 out of 520 downregulated DEMs were listed as of November 15, 2020; however, the specific target genes of most miRNAs remain unknown. We performed Gene Ontology (GO) enrichment analysis using Metascape software to investigate the functions of upregulated and downregulated miRNAs in GBOs compared to NHAs (Figures 2E and S5).
Targets of 117 downregulated miRNAs were significantly enriched in the pathways, which are known to be associated with tumorigenesis and activated in GBMs (i.e., cell cycle, cell division, and Wnt signaling pathway).(19,20) It has been generally recognized that miRNAs negatively regulate gene expression at the post-transcriptional level.(5) These results suggested that the suppression of tumor-derived miRNAs could activate several tumor-associated pathways, and these miRNAs could be detected in urine of patients with GBMs using our nanowire device.
Identification of the Best Combination of miRNAs for Detection of CNS Tumors
To evaluate the usefulness of the assembly-type microfluidic nanowire device for CNS tumor screening, we developed a diagnostic model for CNS tumors using the miRNA expression profiles. To construct a discriminant function as a diagnostic model between patients with CNS tumors and noncancer individuals, we randomly divided 100 noncancer individuals and 102 patients with CNS tumors (i.e., diffuse gliomas, meningiomas, schwannomas, and metastatic brain tumors) into two groups designated as the training set and the validation set (Figure 3).
The training set had twice as many members as the validation set. No statistically significant differences were found in age, gender, miRNA concentration, and histologic findings between patients with CNS tumors and noncancer individuals in the training set or validation set (Table 1). The remaining CNS tumor patients (n = 15) were allocated to an exploratory set, which was intended to evaluate whether the model could also detect other kinds of CNS tumors: CNS neuroblastoma (n = 3), pilocytic astrocytoma (n = 2), CNS embryonal tumor, NOS (n = 1), dysembryoplastic neuroepithelial tumor (DNT) (n = 1), ependymoma (n = 1), epidermoid cyst (n = 1), gangliocytoma (n = 1), ganglioglioma (n = 1), multinodular and vacuolating neuronal tumor of the cerebrum (MVNT) (n = 1), pituitary adenoma (n = 1), primary CNS lymphoma (n = 1), and solitary fibrous tumor/hemangiopericytoma (SFT/HPC) (n = 1). On the basis of the microarray results, we identified a total of 57 miRNAs that showed significantly higher or lower expression (absolute fold-change >1.5 and P < 0.05) in the patients with CNS tumors compared to noncancer individuals in the training set; 22 miRNAs were upregulated and 35 miRNAs were downregulated in patients with CNS tumors (Figure S6).
Figure 3. Workflow for developing the diagnostic model. Urine samples were obtained from 119 patients with CNS tumors and 100 noncancer individuals. The sample set was divided into three groups: the training set, validation set, and exploratory set.
Next, logistic least absolute shrinkage and selection operator regression analysis (logistic LASSO regression analysis) was performed to select the most useful biomarkers among the 57 differentially expressed miRNAs (Figure S7); as a result, a 23-miRNA classifier was developed with data from the training set (miR-6070, miR-22-3p, miR-4538, miR-1285-3p, miR-372-5p, miR-4525, miR-5698, miR-204-3p, miR-6763-5p, miR-101-5p, miR-208a-5p, miR-371a-3p, miR-378a-5p, miR-216a-5p, miR-6864-3p, miR-450b-3p, miR-640, miR-4426, miR-17-3p, miR-450a-2-3p, miR-1248, miR-100-5p, and miR-16-5p). The diagnostic index was calculated based on the individualized values of 23 miRNAs to differentiate between patients with CNS tumors and noncancer individuals (Table S2):
Diagnostic index = −0.2333 + miR-6070 × (0.3359) + miR-22-3p × (0.3087) + miR-4538 × (0.1244) + miR-1285-3p × (0.1097) + miR-372-5p × (0.0246) + miR-4525 × (0.0194) + miR-5698 × (0.0059) + miR-204-3p × (−0.0003) + miR-6763-5p × (−0.0021) + miR-101-5p × (−0.0106) + miR-208a-5p × (−0.0142) + miR-371a-3p × (−0.0591) + miR-378a-5p × (−0.0873) + miR-216a-5p × (−0.1004) + miR-6864-3p × (−0.1024) + miR-450b-3p × (−0.1565) + miR-640 × (−0.1607) + miR-4426 × (−0.3587) + miR-17-3p × (−0.3744) + miR-450a-2-3p × (−0.3744) + miR-1248 × (−0.3981) + miR-100-5p × (−0.4780) + miR-16-5p × (−0.5498).
Since the receiver operating characteristic (ROC) analysis is a tool used to describe the discrimination accuracy of a diagnostic model,(21) we produced the ROC curve by plotting sensitivity (true-positive rate) on the y-axis against 1-specificity (false-positive rate) on the x-axis for the various diagnostic index values in the training set (Figure 4A). Each (x, y) coordinate on the plot represented the true-positive rate and the false-positive rate associated with a cutoff value for defining positive and negative. The cutoff value of the diagnostic model was determined to be 0.4 based on the Youden index(22) (sensitivity + specificity – 1) in the training set (Figure 4B); a diagnostic index ≥0.4 indicated a CNS tumor, and a diagnostic index <0.4 indicated its absence. The area under the ROC curve (AUC) was used to discriminate whether or not the CNS tumor was present, and we followed the established criteria that an AUC of 0.5 represents a test with no discriminating ability, while an AUC of 1.0 represents a test with perfect discrimination.(23) This diagnostic model provided the best discrimination in the training set; and the following values were obtained: AUC, 1.00 (95% confidence interval (CI), 1.00–1.00); sensitivity, 1.00; and specificity, 1.00 (Figure 4A,B). The performance of the diagnostic model was confirmed using the validation set, and the model was very accurate: AUC, 1.00 (95% CI, 1.00–1.00); sensitivity, 1.00; and specificity, 0.97 (Figure 4A).
Figure 4. CNS tumor diagnostic model based on the 23-urinary miRNA classifier. Left panel, training set; middle panel, validation set; and right panel, exploratory set. (A) ROC curves for detecting patients with CNS tumors using 23 miRNAs selected for the diagnostic model. (B) Diagnostic index in each set. Each diagnostic accuracy (%) is indicated. The horizontal axis represents histological diagnosis of the CNS tumors. The gray lines represent the median value of each subgroup and the dotted lines represent the diagnostic index of 0.4. (C) Association between the tumor volume (cm3) and diagnostic index. Pearson’s correlation R and P values (Kruskal–Wallis rank sum test) are indicated. The blue lines represent the regression lines.
Accuracy of the 23-miRNA-Based Diagnostic Model for CNS Tumor Screening
Our diagnostic model was found to be able to accurately discriminate CNS tumors from noncancer samples irrespective of the tumor grade (Figure 4B, left and middle panels); the World Health Organization’s (WHO) classification of CNS tumors comprises a histological grading into four distinct grades of malignancy: WHO grades I–IV. In addition, the model also successfully classified all CNS tumors (n = 15) in the exploratory set as positive: sensitivity, 1.00 (Figure 4B, right panel). Next, we investigated whether our model could identify small CNS tumors.
The diagnostic index of our model did not show a statistically significant association with tumor volume in the training set (P = 0.14), validation set (P = 0.99), and exploratory set (P = 0.22) (Figure 4C). In particular, CNS tumors with a volume of 1 cm3 or less were observed in 2, 3, and 1 cases in the training set, validation set, and exploratory set, respectively. Even in these cases, the diagnostic index was extremely high: median (range), 0.98 (0.76–1.00). Accordingly, our diagnostic model was found to be able to accurately discriminate CNS tumors from noncancer samples irrespective of the tumor grade and size. Thus, we judged this combination of 23 miRNAs as a promising biomarker for CNS tumor screening.
reference link : https://pubs.acs.org/doi/10.1021/acsami.1c01754
More information: Yotaro Kitano et al, Urinary MicroRNA-Based Diagnostic Model for Central Nervous System Tumors Using Nanowire Scaffolds, ACS Applied Materials & Interfaces (2021). DOI: 10.1021/acsami.1c01754
