Article Type : Original Articles
Authors : Yoichi Robertus Fujii
Keywords : microRNA; Aging; Quantum; Alzheimer; Glioma; Network; Astrocytoma
Background: Alteration in plasma/serum microRNA (miRNAs)
profiles are associated with brain cancer and Alzheimer’s disease (AD).
Circulating miRNA profiles could be the first choice of non-invasive biomarker
candidates for the prediction, diagnosis and prognosis of human disease. To
further investigate the function of circulating miRNAs in brain tumors and AD
to build sustainable healthcare, in silico analysis was performed using a
quantum miRNA language with artificial intelligence (MIRAI) with miRNA
diagnostic panels.
Methods: Data from serum/plasma miRNA panels from patients
with AD or brain cancer was extracted from the database. The miRNA memory
package (MMP) from the miRNA diagnostic panel were selected by data mining. The
miRNA entangling target sorter (METS) bioinformatics algorithm using MIRAI was
used to analyze the etiology according to the procedure previously described.
Result: Hub miRNA/targeting protein links on glioma and
astrocytoma data were extracted as upregulation of high mobility group AT-hook
1 (HMGA1) by miR-16-5p (down regulation) and up regulation of BCL2 apoptosis
regulator (BCL2) by miR-497-5p (down regulation), and down regulation of RB
transcriptional coreceptor 1 (RB1) by miR-106a-5p (up regulation) and down
regulation of cyclin dependent kinase inhibitor 1A (CDKN1A) by miR-20a-5p (up
regulation), respectively. AD was associated with circadian rhythm,
ribosomopathy and the vascular endothelial growth factor A (VEGFA) signalling
pathway. Enhanced VEGFA expression through down regulation of miR-361-5p was a
vivid cause of AD. Elevation of the circadian rhythm-related protein, aryl
hydrocarbon receptor nuclear trans locator like (ARNTL) via down regulation of
miR-142-3p was the underlying pathogenicity of mild cognitive impairment (MCI)
and AD; therefore, no inverse correlation between common miRNA hubs and hub
miRNA levels was observed between AD and brain cancers.
Conclusion: Given the hub miRNAs in brain cancer
and AD, there were no common miRNAs from MMPs among disease. Thus, miRNA
biomarkers for brain cancer, MCI and AD could be available for pathophysiologic
diagnosis and prediction. It is the first in silico report that high levels of
VEGFA due to down regulation of miR-361-5p may be a cause of progression from
MCI to AD via angiodysplasia. As miR-361-3p and -5p were targeted to the BACE1
3’UTR, anti-VEGFA agent and/or miR-361 stem loop mimicking locked nucleic acid
(LNA) treatment may be effective approaches to MCI or early AD prevention.
Gliomas are the most prevalent malignant tumor
in brain cancers, accounting for 81%. The incidence of glioma in adults is
rare, approximate 0.1 to 0.5 in million people [1]. From World Health
Organization (WHO) classification of tumors of the central nervous system,
gliomas histologically contain the most common glioblastoma and astrocytoma in
diffuse astrocytic and oligodendroglial tumors. However, there is no consensus
definition of gliomas as a larger class of histology [2]. For example, gliomas
are classified from grade I to IV by WHO according to malignant behavior;
however, in the class of diffuse astrocytic and oligodendroglial tumors,
diffuse astrocytoma on isocitrate dehydrogenase gene (IDH)-mutant is the WHO
grade II tumor, and glioblastomas on IDH wild type and IDH-mutant are the WHO
grade IV tumor. The 5-year relative survival rate for glioma is only 0.1-5%
[1]. Temozolomide (TMZ) is a standard chemotherapeutic agent that targets
gliomas, and acts by methylation and crosslinking of DNA in glioma cells [3].
Surgery followed by concurrent radiotherapy plus TMZ chemotherapy is
recommended as current protocolized treatments for high-grade glioma [4]. It is
estimated that approximately 36 million people worldwide would suffer from
Alzheimer’s disease (AD) and other dementias in 2010, and the number of AD is
predicted to grow rapidly. For example, there will be 66 million in 2030 and
115 million in 2050 [5]. There are two types of dementia. The first is AD. AD
is characterized by a progressive, unremitting and neurodegenerative disorder.
The other is vascular cognitive impairment and dementia (VCID). VCID is caused
by hypertension and arteriosclerosis, including amyloid angiopathy, infraction
and ischemic attacks to induce cerebrovascular damage [6]. VCID comprises mild
cognitive impairment (MCI), vascular dementia (VaD), and mixed vascular and
AD-type cognitive impairment [7]. Although AD is the most common type of
dementia, AD is complex and has been diagnosed by the mini-mental state
examination (MMSE), neuropsycological tests, magnetic resonance imaging (MRI),
positron emission tomography (PET) and the single-photon emission computed
tomography (SPECT) with machine learning [8]. While the effective diagnosis of
AD, its prodromal stage-MCI including amnestic MCI (aMCI) and nonamnestic MCI
(naMCI) have recently been performed by machine learning on MRI plus MMSE, the
protocol provided an average accuracy of 92.4%, sensitivity of 84% and
specificity 96.1% for MCI versus cognitively normal (CN). However, it is
mathematically calculated that the USA model requires approximate $8 trillion
in medical and care cost to diagnose AD early and accurately [9,10]. Therefore,
for establishing sustainable health management, less expensive and simpler
diagnostic tools should need to be developed for the initial diagnosis of AD
and MCI in the early stage. It is well known that amyloid ? (A?), tau and
apolipoprotein E (APOE) are the major elements contributors to AD. Senile
plaques are consisted by A?, and amyloidopathy of the cerebrovasculature, and
loss of neurons in the temporal lobes were pathologically observed in AD [11].
A? is produced from amyloid ? precursor protein (APP) digested by ?-secretase 1
(BACE1) [12]. Progressive tau accumulation has been found in the medial and
lateral temporal cortices of APOE ?4 genotype AD patients in a PET study [13].
Although the APOE ?4 allele status was significantly correlated with the level
of tau neurofibrillary tangle, the pathology of tau was independent of A?
plaques in PET clinical analysis [14]. Plasma analysis showed that A?+
participants had a significantly higher proportions of APOE ?4 genotype
carriers than A?- in MCI [15]. Except for hereditary AD, most AD patients are
sporadic; however, APOE ?4 genotype is involved in A?/tau accumulation in
different states of neurodegeneration and in different sample sources. On the
clinical side, the effective reduction of plaques or plasma A? by numerous drug
candidates for AD might have not provided clinical benefit, as well as the approval
of aducanumab by the U. S. Food and Drug Administration (FDA) in June 7th, 2021
could be; therefore, the fundamental question remains that A? and tau might be
‘results’ of AD but not their ‘causes’ [16-20]. On the contrary, ribosomal
deficiencies, such as the ribosomopathy have been observed in early AD and MCI
[21,22]. Therefore, impaired ribosome biogenesis and protein synthesis would be
characteristic of MCI. Further, since vascular endothelial growth factor A
(VEGFA) secretion from astrocytes has been increased by A?42, VEGF is also
implicated in AD. High levels of VEGF resulted in loss of insoluble A?42 and
increasing of blood-brain barrier (BBB) permeability [23-26]. Conversely, the
soluble form of the VEGF receptor (sVEGFR) was reduced in the plasma of AD
compared to healthy control subjects [27]. These results suggest that sVEGFR is
an antagonist of VEGFRs, VEGFR-1 (FLT1) and -2 (KDR) during the angiogenic
process, and by its decreasing, VEGF may protectively affect human vascular
endothelial cells to remove insoluble A?. But high levels of VEGFA may
inversely increase soluble A?. Subsequently, despite of these VEGF effects, the
mechanisms by which the clinical status of MCI progresses remains unclear. We
have previously investigated common neurodegeneration susceptible universal
microRNAs (miRNAs) by in silico simulation of miRNA entangling target sorting
(METS) among 105 miRNAs in Alzheimer’s disease, 44 miRNAs in Parkinson’s
disease, 28 miRNAs in frontotemporal dementia, and 93 miRNAs in Huntington’s
disease [28]. Seven miRNAs, miR-34b/c, miR-29a-3p, miR-29b-3p, miR-124-3p,
miR-9-3p, miR-132-3p and miR-138-5p were common, and was thought to be involved
in the memory mechanism of the brain. Seven protein targets, Notch receptor 1
(NOTCH1), Sirtuin 1 (SIRT1), histone deacetylase 4 (HDAC4), methyl-CpG binding
protein 2 (MeCP2), brain derived neurotrophic factor (BDNF), BACE1 and
?-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) were
extracted by computer analysis and linked to the above 7 miRNAs and 7 proteins
in the network. Our data has been corroborated by a previous report that three
miRNA panel of miR-132, miR-124 and miR-34 are associated with memory
impairment [29]. Meanwhile, in another repots, miR-30b-5p has been overlapped
among Alzheimer’s disease, Parkinson’s disease, multiple sclerosis and
amyotrophic lateral sclerosis [30]. However, it remains unclear whether the
functional universal miRNA panel can distinguish between age-related cases of
Alzheimer’s disease and human neurodegenerative disorders and human brain
tumor. In addition, by ?-secretase substrate presenilin gene (PSEN)
double-knockout mouse of a model for human AD, increasing of tumor suppressor
miRNAs has been observed, and a negative correlation between AD and cancer was
showed [31]. But clinical analysis in humans has less been reported much
comparison between miRNA-based AD and cancer etiologies [32,33]. Definitive
diagnosis of AD is challenging, and the etiology analysis of AD with diagnosis
would be the most difficult goal in neurology. It has been reported as
diagnostic markers for dementia to detect miRNAs, interleukin-10 (IL-10), IL-6
in the plasma/serum and A?, tau in cerebrospinal fluid (CSF); however, their
effectiveness and clinical discrimination among other inflammation diseases
have been disputed points except for miRNA biomarkers [11, 34-37]. Therefore,
miRNA non-invasive biomarker in the circulating system would be useful for
early diagnosis of dementia including diagnosis of the presence or absence of
cognitive impairment such as AD and MCI. We have reported that the etiology of
viral infection can be simulated in miRNA biomarker panels by using the quantum
miRNA language plus artificial intelligence (MIRAI), such as hepatitis B virus
(HBV), hepatitis C virus (HCV) and human immunodeficiency virus (HIV-1) [38].
Circulating miRNA biomarkers have also revealed the pathogenic mechanisms of
human cancer, breast, lung, colorectal, pancreatic, esophageal, gastric cancers
and liver cancers, which is in good agreement with the National Institute of
Health (NIH) definition of biomarkers. The etiology of severe acute respiratory
syndrome human coronavirus 2 (SARS-Cov-2) has not yet been known clearly, but
we have firstly revealed in silico the etiology of coronavirus disease 2019
(COVID-19) from miRNA profiles of SARS-CoV-2 infection [39-44]. In addition, by
using the same algorithm as MIRAI above, we discovered that rice MIR2097-5p
could act as anti-COVID-19 therapeutic miRNA agents. Furthermore, it has been
shown that COVID-19 vaccine (BNT162b2) has potency of a miRNA vaccine [45,46].
Thus, miRNAs have been investigated as candidates for non-invasive AD
biomarker, where we simulated neurodegeneration etiology with a quantum miRNA
language with MIRAI using miRNA biomarker panels from serum or plasma.
Circulating miRNA data in brain cancers and AD were useful to elucidate their
etiology. Particularly, in the analysis of MCI, computer simulations provided
predictive results that could distinguish the etiology of MCI and AD.
Consequently, it was showed that activation of the VEGF signaling pathway
regulated by miRNAs is a major cause of AD [47]. Then, therapeutic miRNA agents
were discussed.
Database usage
Google Scholar
(https://scholar.google.co.jp) was firstly used for data extraction from miRNA
panels and miRNA profiles in plasma or serum. PubMed (pubmed.ncbi.nim.nih.gov),
in which our treatise data is not listed and not affected to AI analysis as a
false premise bias, was also used. Total information content was 502,268 in
aging, 216,289 in brain cancer, 103,170 in glioma and 44,434 in astrocytoma,
81,297 in neurodegeneration, and 165,704 in AD. Gene functions of protein and
RNA genes were searched by GeneCards (www.genecards.org). Protein ontology was
investigated by GO enrichment analysis in Geneontology (geneontology.org).
METS analysis
METS analysis was performed by a computer processing algorithm as described previously [48]. Shortly, miRNA memory package (MMP) was extracted from miRNA biomarker panels and profiles. Data mining about miRNA panels and profiles for the standard setup of MMPs was performed by: 1) data from serum or plasma, 2) cleared in expression levels of up- and down-regulation, 3) data was statistically analyzed by receiver operating characteristic (ROC) and the cut off value of an area under the curve (AUC) about biomarker profiling is 0.8. As quantum miRNA language, single miRNA quantum energy level in an MMP was calculated as single nexus score (SNS) (Table 1).
Table 1: MMPs in brain cancers and AD.
AD and tumor |
Data source |
AUC data |
miRNA |
Level |
SNS |
Source |
Reference No. |
Brain cancer |
Glioma |
>0.75 |
miR-497-5p |
down |
6 |
serum |
51-53 |
miR-125b-5p |
down |
4 |
serum |
||||
miR-16-5p |
down |
6 |
plasma |
||||
miR-21-5p |
up |
5 |
serum |
||||
miR-222-5p |
up |
5 |
serum |
||||
miR-124-3p.1 |
up |
7 |
serum |
||||
Astrocytoma |
0.9722 in a panel |
miR-15b-5p |
up |
4 |
serum |
59 |
|
miR-16-5p |
up |
6 |
serum |
||||
miR-19a-3p |
up |
4 |
serum |
||||
miR-19b-3p |
up |
4 |
serum |
||||
miR-20a-5p |
up |
7 |
serum |
||||
miR-106a-5p |
up |
7 |
serum |
||||
miR-130a-3p |
up |
6 |
serum |
||||
miR-181b-5p |
up |
7 |
serum |
||||
miR-208a-3p |
up |
5 |
serum |
||||
AD |
MCI panel |
>0.9 |
miR-483-5p |
up |
11 |
plasma |
62 |
miR-486-5p |
up |
6 |
plasma |
||||
miR-30b-5p |
down |
2 |
plasma |
||||
miR-200a-3p |
up |
5 |
plasma |
||||
miR-502-3p |
up |
6 |
plasma |
||||
miR-142-3p.1 |
down |
5 |
plasma |
||||
US & Germany cohorts |
>0.8 |
miR-17-3p |
up |
7 |
plasma |
63 |
|
miR-5010-3p |
down |
3 |
plasma |
||||
let-7d-3p |
down |
3 |
plasma |
||||
miR-26b-5p |
up |
5 |
plasma |
||||
miR-28-3p |
down |
6 |
plasma |
||||
miR-361-5p |
down |
5 |
plasma |
||||
miR-532-5p |
up |
7 |
plasma |
||||
Conventional, Meta-analysis |
>0.876 |
miR-34a-5p |
up |
8 |
plasma |
64-66 |
|
miR-34c-5p |
up |
8 |
plasma |
||||
miR-7-5p |
down |
7 |
plasma |
||||
miR-191-5p |
down |
5 |
plasma |
||||
let-7g-5p |
down |
7 |
plasma |
||||
miR-15b-5p |
down |
4 |
plasma |
||||
miR-142-3p.1 |
down |
5 |
plasma |
||||
Bold: hub miRNAs |
Next, entangling double
miRNAs’ quantum energy levels as double nexus score (DNS) were computed by
multiplication of two SNSs in a matrix based on the scalar product with
electron spin [28]. To apply quantum miRNA language to the METS algorithm, DNS
data of multi-targets to a miRNA and multi-miRNAs to a target in the higher
dimension matrix was computed from data of Target Scan Human 7.2, DIANA-TarBase
and miRTarBase Ver. 8.0 or mirtarbase data for miRTarBase release 6.1, which
was re-built up by Excel file of packages in the GitHub . The valid DNS values
of miRNA/target connections were determined by the optional cut-off value.
Protein/protein interaction search and cluster analysis were performed by using
STRING Ver. 11.0. Network among miRNAs and target proteins was constructed from
the METS algorithm according to data from MIRAI.
METS analysis with
MIRAI
MIRAI was used for
validation for METS and extraction of hub miRNAs. An AUC in ROC, accuracy, and
precision was calculated as previously described [44]. Previous METS analysis
data in pancreatic, lung, colorectal, gastric, esophageal and liver cancers was
combined with present data of METS analysis in AD and brain cancers.
Distribution of
quantum energy in brain cancers and AD
MMPs were extracted from circulating miRNA panels. MiRNA/miRNA mutual (DNS) quantum energy frequencies in each layer used in METS algorithms, named the Quantum Core Region (QCR), were computed. Frequency of quantum energies in MCI (AD MCI), AD on USA & Germany cohorts (AD USA & Germany) and on conventional AD miRNA panel in meta-analysis (AD Conventional), gliomas and astrocytoma. QCR of hub miRNAs in AD and glioma was at 21-40 QRC (brown bars), which was in a peak quantum energy frequency layer. Profile of miRNAs in AD is completely different between conventional panel and USA & Germany panel, and conventional AD panel differed from USA & German panel in that it contains a high quantum energy frequency 61-80 QCR region. On the other hand, QCR regions of brain cancers were similar in each other. USA & Germany AD hub miRNAs and conventional AD hub miRNAs were located at 21-40 QCRs (orange arrows). Glioma miRNA hub was also on the 21-40 QCR (orange arrow). MCI and astrocytoma miRNA hubs were in 41-60 QRCs (gray arrows). Data clearly suggests that quantum energy levels in brain disease are moderate. MiRNA QCR layers were cohered for METS analysis and network among miRNAs and protein/protein cluster was calculated from the data (Figure 1).
Figure 1: The quantum energy levels of miRNA in aging and tumor in the brain. DNSs between two miRNAs in MMP were computed.
Brain tumour
Recently, a set of
miRNAs de-regulation between the brain tumor and AD has been identified, and it
has been demonstrated that the levels of tissue expression in the set of miRNAs
are inversely associated between the tumor and AD [32]. Further, circulating miRNAs
overlapped between cancer and AD has been found and their expression levels
were usually opposite [33]. Therefore, METS analysis of brain cancer was
performed to confirm the inverse relationship between cancer and AD miRNA
expression, and as a data size control, including single panel or mixed panels,
of MIRAI in AD and MCI pathophysiological changes.
Glioma and astrocytoma
Circulating microRNAs
have been identified as a biomarker of glioma [49,50]. In our simulation, the
MMP of glioma (WHO grade II/III and IV) was selected from three circulating
miRNA panels according to the data mining rules. It contained six miRNAs,
miR-497-5p, miR-125b-5p, miR-16-5p, miR-21-5p, miR-222-5p and miR-124-3p.1
[51-53]. MiR-497-5p, miR-125b-5p and miR-16-5p have been down regulated, and
miR-21-5p, miR-222-5p and miR-124-3p.1 were up regulated (Figure 2).
Targets of miR-125b-5p had no linkage of protein/protein cluster and targets of miR-21-5p were not computed by below cut-off DNS value in METS data (data not shown). Hub miRNAs were miR-16-5p and miR-497-5p in pathogenic analysis of glioma. About targets of the miRNA, HMGA1 and BCL2 were up regulated by decreasing of miR-16-5p hub with let-7b-5p, let-7a-5p and miR-196a-5p, and with miR-34a-5p. BCL2 was also enhanced by down regulation of miR-497-5p with miR-34a-5p. HMGA1 expression has been observed in 96.7% of malignant glioma (WHO grade II/III) but not in normal brain tissues, and the expression of HMGA1 in human glioma significantly correlated with VEGFA expression [54]. MiR-361-5p was downregulated in glioma tissues and was inversely correlated with glioma grade [55]. Hence, the relation between VEGFA and miR-361-5p in glioma tissues may be analogous to that in AD. But there is no paper about miR-361-5p as a circulating miRNA marker of glioma. MiR-16-5p expression in glioma has been lower than non-cancerous brain tissues and upregulation of miR-16-5p promoted apoptosis via suppression of BCL2 [56].
Figure 2: The network scheme of METS simulation in a miRNA biomarker panel of glioma and astrocytoma. Linkages among protein clusters and miRNA/miRNA were depicted by METS algorithm in glioma (A) and astrocytoma (B). miRNAs in red: upregulation, proteins in blue: downregulation. Red circle is the hub miRNAs. Small circles in blue are the hub miRNAs.
B flavone agent,
apigenin has induced apoptosis of U87 human glioma cells via increasing of
miR-16-5p and suppression of BCL2 [57]. Further, breast cancer cell apoptosis
has been induced by up regulation of miR-497-5p through negative regulation of
BCL2 [58]. An AUC of glioma with MIRAI was 1.00 (accuracy, 1.00; precision,
1.00; recall, 1.00; F value, 1.00). Thus, it was found that the key protein targets
are HMGA1 and BCL2 in glioma. As BCL2 was down regulated in AD, the levels of
BCL2 expression was inverse between AD and glioma; however, effector miRNAs
were different in two brain diseases. The MMP of astrocytoma was determined.
Nine miRNAs, miR-15b-5p, miR-16-5p, miR-19a/b-3p, miR-20a-5p, miR-106a-5p,
miR-130a-3p, miR-181b-5p and miR-208a-3p were selected from a miRNA panel of
serum from patients with astrocytoma (WHO grade II, 30%; III 40%; IV, 30%)
[59]. All miRNAs have been up regulated. The hub miRNA was miR-106a-5p or
miR-20b-5p in the etiologic analysis of astrocytoma. RB1 was blocked by up
regulation of miR-106a-5p in combination with miR-106b-5p. Cyclin dependent
kinase inhibitor 1A (CDKN1A) was inhibited by up regulation of miR-20a-5p and
miR-106a-5p along with miR-93-5p, miR-20b-5p, miR-17-5p and miR-106b-5p, and up
regulation of miR-208-3p with miR-93-5p. Since tumor suppressor proteins, RB1
and CDKN1A were down regulated, up regulation of miR-106a/b-5p and miR-20a-5p
would be carcinogenic in the brain. Although miR-106a/b-5p and miR-20a-5p have
been up regulated in pediatric medullblastoma (WHO grade IV), ependymoma (grade
II) and pilocytic astrocytoma (grade I) tissues, there was no other reports
about the circulating miR-106a/b-5p and miR-20a-5p profiles in patients with
astrocytoma (grade II-IV) [60]. In glioblastoma CD133+ stem cell lines,
miR-106a-5p and miR-20a have been enhanced [61]. An AUC of astrocytoma with
MIRAI was 0.99 (accuracy, 0.96; precision, 0.98; recall, 0.98; F value, 0.98).
Since data of therapeutic targets was statistically significant, we found two
key protein targets, RB1 and CDKN1A on astrocytoma.
MMPs for the
etiological analysis in MCI and AD
Three MMPs of miRNAs in
AD computer simulation were prepared from circulating miRNA biomarker in
plasma. MMP of MCI contains six miRNAs of miR-483-5p, miR-486-5p, miR-30b-5p,
miR-200a-3p, miR-502-3p and miR-142-3p.1 from a panel (AUC >0.9)
(Mini-Mental State Examination, MMSE: 25.9 in MCI; 20.4 in AD) [62].
MiR-483-5p, miR-486-5p, miR-200a-3p and miR-502-3p have been up regulated, and
miR-30b-5p and miR-142-3p.1 were down regulated. In the case of AD on USA &
Germany cohort data (MMSE: 20.0 in AD; 29.1 in healthy control), seven miRNAs,
miR-17-3p, miR-5010-3p, let-7d-3p, miR-26b-5p, miR-28-3p, miR-361-5p and
miR-532-5p were prepared as MMP (AUC >0.8) [62,63]. MiR-17-3p, miR-26b-5p
and miR-532-5p have been up regulated, and miR-5010-3p, let-7d-3p, miR-28-3p
and miR-361-5p were down regulated. A panel of conventional case from data of
meta-analysis (AUC >0.876) contained seven miRNAs of MMP, miR-34a-5p,
miR-34c-5p, miR-7-5p, miR-191-5p, let-7g-5p, miR-15b-5p and miR-142-3p.1
[64-66]. MiR-34a/c-5p has been up regulated and 5 miRNAs remained were down
regulated. There was no linkage of protein/protein cluster to miR-30b-5p and
miR-502-3p in MCI data, miR-28-3p and 361-5p in USA & Germany cohort data,
and miR-191-5p and miR-15b-5p in conventional meta-analysis data (data not
shown). Targets against let-7-3p and miR-5010-3p in USA & Germany were not
computed by below the cut-off DNS value in METS data (data not shown).
Ribosomopathy and
circadian rhythm dysregulation upon MCI
About target proteins against circulating biomarker miRNAs of MCI, have determined that the center of network is the mitochondrial respiratory chain pathway involving in oxidative stress [62]. When our METS analysis was performed by MCI microRNA data, ribosomal protein L31 (RPL31) was suppressed by upregulation of miR-483-5p along with miR-409-5p and miR-653-5p (Figure 3).
Figure 3: Network diagram of METS simulation in a biomarker panel of the MCI stage and AD USA & Germany cohorts. Interaction among protein and miRNA/miRNA was simulated by METS analysis in MCI (A) and AD (B). miRNAs and proteins in red: upregulation, in blue: downregulation. Small circles in red and blue are the hub miRNAs. Large circles in red (circadian rhythm), in blue (ribosomopathy or ribosomopathy in the mitochondria) and in yellow (VEGF-receptor signaling pathway) are presented as GO functions.
RPL31 is a component of
the 60S RNA subunit with ribosomal protein. Decreasing of protein synthesis via
impairment of ribosomal function has been found in MCI and expression of RPL31
has been reduced in AD, which is a ribosomopathy [67,68]. While we have
previously shown that human ribosomal RNA derived miRNAs would be implicated in
carcinogenesis as a ribosomopathy, and ribosomopathies lead to oxidative stress
via the mitochondria, present data and suggests that ribosomal malfunction by
upregulation of miR-483-5p would induce the stress against neural cells. It is
well known that disturbed circadian rhythms with sleep problem is implicated in
AD, and circadian locomotor output cycles gone kaput (CLOCK) gene and external
Zeitgeber are related with circadian rhythmicity [69-71]. Disturbance of
circadian rhythm occurs in an early AD as well [72]. However, molecular targets
of circadian rhythmicity for early AD therapeutic inventions have not yet been
provided. In our simulation, expression of aryl hydrocarbon receptor nuclear
translocator like (ARNTL, BMAL1) was augmented by downregulation of miR-142-3p
in combination with miR-206 and miR-1-3p. ARNTL is a core component of the
circadian clock, which forms a heterodimer with clock circadian regulator
(CLOCK). Although in AD mouse model experiments, loss of ARNTL in the brain
parenchyma increased expression of APOE and A? accumulation, protein expression
levels of human ARNTL and CLOCK have significantly elevated in impaired
astrocytes of cerebral cortex from patients with AD [73,74]. Therefore,
enhancement of ARNTL and CLOCK may promote apoptotic cytotoxicity via metabolic
dysfunction in human astrocytes. Imaging analysis by fluorodeoxyglucose
(FDG)-positron emission tomography (PET) have showed hypometabolism and
reducing utilization of glucose in the brain with AD [75]. These results
strongly support that aberration of circadian rhythm would be associated with
neurodegeneration, and downregulation of miR-142-3p participated in mechanisms
of MCI [76,77]. Citron Rho-interacting serine/threonine kinase (CIT) and
mitogen-activated kinase 3 (MAPK3) were inhibited by upregulation of miR-486-5p
and upregulation of miR-483-5p with miR-205-5p. MAPK overlaps in a report [62].
Further, Rac family small GTPase 1 (RAC1) and Rho associated coiled-coil
containing protein kinase 2 (ROCK2) were increased by downregulation of
miR-142-3p with miR-122-5p and with miR-138-5p, respectively. From GO analysis,
these four proteins in a cluster were dominantly implicated in RhoA/ROCK2
pathway (GO: 0007266), which was also related with upstream VEGF receptor (GO:
0005021). VEGF signaling pathways is the major process of angiogenesis and
vasculogenesis, and ROCK 1 and 2 inhibition has activated angiogenesis via
VEGFA/Ras homolog family member A (RhoA) [78]. Therefore, inhibition of
VEGF/VEGFR/RhoA/ROCK/mitogen-activated kinase (MAPK) axis may progress MCI
condition. However, VEGF/VEGFR-1 or VEGF/VEGFR-2 signaling directly or
indirectly can transduce to angiogenesis without affecting through the
RhoA/ROCK/MAPK pathways because RhoA controls diverse pathways, such as
RhoA/actin-regulating kinase PRK1 (PRK1) / phosphatidylinositol-4-phosphate5-kinase
(PIP5K)/phospholipase D (PLD) pathway [79]. These data suggest that suppression
of MAPK3 and augment of ROCK2 may partly be anti-angiogenesis effects via VEGF
signalling pathway. But the altered expression of VEGF has not been observed in
MCI [79]. Finally, by MIRAI, it was resulted that an AUC about the etiology of
MCI in AD subjects was 0.53 (accuracy, 0.86; precision, 0.92; recall, 0.92; F
value, 0.95), in contrast, an AUC of a diagnostic miRNA panel in MCI has been
over 0.9 as described above. It is showed that the miRNA panel is useful for
differential diagnosis to distinguish MCI in neurodegenerative disease in
precious medicine. Further, another plasma miRNA biomarker panel and validation
study have been documented in MCI [80]. Six miRNA biomarkers, miR-132-3p,
miR-128-3p, miR-874-3p, miR-134-5p, miR-382-5p and miR-323b-5p (AUC, >0.9)
have been up regulated in plasma of patients with MCI. The etiologcal analysis
of this panel was performed according to the METS simulation with MIRAI
procedures (data not shown). Although GO analysis of miRNA-target proteins
showed just the cell cycle-related protein cluster alone (GO: 0061575), an AUC
about the etiology in AD was 0.64 (accuracy, 0.80; precision, 0.94; recall,
0.85; F value, 0.89). The results of validation data were similar pattern to
data [62]. Cell cycle dysregulation may not dominantly be the primary cause of
MCI. If in MCI the validated results of circulating miRNA biomarker were
distinct from those as AD of the etiology analysis as below, diagnostic miRNA
biomarker would be available for precious medicine of neurodegeneration.
VEGF signaling,
ribosomopathy and circadian rhythm dysregulation upon AD
AD USA & Germany
panel have already extracted a GO enrichment analytic data from the miRTarBase,
which included ATP binding cassette subfamily A member 1 (ABCA1), Death
associated protein kinase 1 (DAPK1), Insulin-like growth factor 1 receptor
(IGF1R) and VEGFA in aging, and cyclin D1 (CCND1), cyclin E1 (CCNE1), cyclin E2
(CCNE2), cyclin dependent kinase 6 (CDK6), MYC proto-oncogene (MYC), RAD51
recombinase (RAD51) and RB transcriptional corepressor 1 (RB1) in DNA damage
response; however, they were unable to exclude some biases about selection of miRNA
small sets and targets. Whereas, in our METS analysis, VEGFA was upregulated by
downregulation of miR-361-5p along with miR-378a-3p, miR-126-3p, miR-16-5p and
miR-15a-5p. Further, enhancer of guanylate binding protein 1 (GBP1) was blocked
by upregulation of miR-532-5p with miR-23a/b-5p. Human GBP1 has reduced VEGF
expression in mouse cancer model [81]. Therefore, reduction of GBP1 may
contribute VEGFA secretion. VEGFA secretion from astrocytes is increased by
A?42; therefore, VEGFA increasing would be deeply implicated in AD as described
above. Further, VEGF genotypes or haplotypes was not associated with AD in the
case-control study, suggesting that high levels of VEGFA by downregulation of
miR-361-5p may be the major cause of AD, which is acquired but not be
genetical. BH3 interacting domain death agonist (BID) was suppressed by
upregulation of miR-26b-5p with miR-4465. BID is a death agonist and antagonist
of BCL2 apoptosis regulator (BCL2), and BID is a mediator of mitochondrial
damage by caspase-8; therefore, BID is proapoptotic protein [82,83]. Since
myeloid cell leukemia-1 (MCL-1) was induced by VEGF and MCL-1 inhibited cell
apoptosis by BCL2 associated X (BAX), the VEGF/MCL-1/BAX axis was associated
with prostate cancer metastasis. Accordingly, BID suppression and VEGF
upregulation may be implicated in anti-apoptotic and indirectly migrating
effects to A? accumulated neural cells. RUNX family transcription factor 3
(RUNX3) was inhibited by upregulation of miR-532-3p with miR-106a/b-5p and
miR-130a-3p. RUNX3 knockdown has enhanced endothelial progenitor cell function
through VEGF/MAPK/AKT serine-threonine kinase (AKT) pathway [84]. Therefore,
decline of RUNX3 may induce excessive angiogenesis. From GO analysis, these
four proteins in a cluster were implicated in VEGFR signaling pathway (GO:
0048010). Altogether, it is suggested that VEGFA-autocrine/paracrine would be
increased by A? accumulation and it may induce to inhibit apoptosis of A?
accumulated neural cells. White matter hyperintensities (WMH) in magnetic
resonance imaging (MRI) vascular origin have increased the risk of dementia
[85]. In a rat experiment, the mitochondrial ribosomal protein 18 (MRPL18) gene
has been related with WMH [86]. As described above, ribosomopathy including the
mitochondrial functions is implicated in AD [67]. Since in METS analysis,
MRPL18 was reduced by increasing of miR-532-5p, mitochondrial ribosomopathies
with oxidative stress could be implicated in AD. MIRAI analysis showed that an
AUC of AD USA & Germany was 1.0 (accuracy, 0.93; precision, 1.0; recall,
0.93; F value, 0.96) (Figure 4).
In AD conventional data from meta-analysis, ARNTL up regulation was observed by down regulation of miR-142-3p along with miR-206 (data not shown), which was similar case to the up regulation for the CLOCK/ARNTL axis in MCI as described. Although ARNTL gene rs2278749 and CLOCK gene rs1554483 polymorphisms have been susceptible to AD in a Chinese population, circadian rhythm gene polymorphisms including CLOCK, Period circadian regulator 2 (PER2), PER3 and hypocretin receptor 2 (OX2R) genes were not associated with AD in a Brazilian population [87-89]. It suggests that the inheritable genetic factor of circadian rhythm genes is not the basic cause of AD. On the contrary, rs2526377 (A>G) at the 17q22 locus annotated to MIR124 and BZRAP1-AS long non-coding RNAs (lncRNAs) was significantly associated with a reduced risk of AD and decreased the expression of miR-142-3p [90].
Figure 4: Validation of METS analysis with MIRAI. METS analysis in AD and brain cancers were validated by MIRAI. An AUC (blue line), accuracy (amber line), precision (gray line) and F value (yellow line) were calculated and presented with radar charts.
Therefore, reduced
levels of miR-142-3p would be implicated in the AD risk. Further, although
alpha rhythm during wake is known to be down regulated in early AD, it has been
implicated in suppression of hyperpolarization activated cyclic nucleotide
gated potassium channel (HCN) [91]. HCN3 was reduced by up regulation of
miR-34a/c-5p with miR-449a and miR-449b-5p (data not shown). Since reduction of
HCN and accumulation of A? have been correlated in AD, down regulation of HCN3
would be related with AD. Vesicle associated membrane protein 2 (VAMP2) was
inhibited by up regulation of miR-34a/c-5p with miR-373-3p, miR-372-3p and
miR-520d-3p (data not shown) [92]. VAMP2 is a component of the protein complex
with the 25-kd synaptosomal-associated protein (SNAP25) and syntaxin 3 (STX3)
involved in the docking and fusion of synaptic vesicles on the presynaptic
membrane. VAMP2 has showed decreasing expression in aging and AD [93]. Thus,
VAMP2 down regulation would be implicated in AD. These results indicate that
dysregulation of circadian rhythm-dependent sleep/wake might be a common
hallmark of aging including MCI and AD. NOTCH is highly expressed in the
hippocampus of the human brain and plays an important role for learning and
memory; therefore, reduction of NOTCH signalling is deeply involved into AD
[94]. While NOTCH, human brain memory-associated protein has already been
documented in previous our book, NOTCH2 was reduced by up regulation of
miR-34a-5p (data not shown). So, it suggests that suppression of NOTCH2 by
miR-34a-5p might be associated with both AD and brain memory. Simultaneously,
miR-34a-5p up regulation inhibited MDM4 expression. MDM4 regulator of P53
(MDM4) has pro-survival role for neurons; therefore, MDM degradation is
associated with neuronal death. Progressive loss of MDM4 has been shown to be a
cause of A?-induced neural cell death during AD progression [95]. Further,
anti-apoptotic protein, BCL2 was down regulated by up regulation of
miR-34a/c-5p. A? protein down regulated BCL2 in human neuron [96]. High
mobility group AT-hook 1 (HMGA1) levels were increased in brain tissues from
sporadic AD [97]. HMGA1 increasing was also found in our simulation by down
regulation of miR-142-3p with let-7b-5p (data not shown). HMGA1 may be implicated
in AD. Thus, it is suggested that NOTCH2, MDM4 and BCL2 down regulation would
be implicated in AD. Finally, an AUC of AD conventional was decreased to 0.87
(accuracy, 0.97; precision, 0.97; recall, 1.0; F value, 0.98) compared with
that of AD USA & Germany in MIRAI (AUC, 1.0). AI analysis indicates that up
regulation of a VEGFA signalling via miR-361-5p down regulation is more
important one than circadian rhythm upon AD diagnosis and etiology. Increased
VEGFA has been associated with an age-related eye disease, neovascular
age-related macular degeneration (AMD), and pegaptanib, bevacizumab,
ranibizumab and aflibercept can use for treatment of neovascular AMD as
anti-VEGFA agents [98,99]. With respect to age-related disease,
inflammation-related atherosclerotic peripheral vascular disease contains type
2 diabetes mellitus and its complication. Higher VEGF in exosome has been
observed in diabetes and it was associated with diabetes status [100]. Further,
chronically increasing VEFGA has been associated with cardiac hypertrophy
[101]. Downregulation of miR-361-5p has been found in the plasma of neovascular
AMD patients [102]. Additionally, serum VEGF reduction by treatment with
cerebrolysin plus donepezil has improved functioning and cognition in patients
with AD [103]. In integrative bioinformatics, VEGFA gene together with several
other protein genes has been related with AD [104,105]. Our AI analysis shows
that high levels of VEGFA are responsible for the progression of MCI to AD
through angiodysplasia, even if accumulation of A? and/or tau protein is
associated with the effect of angiogenic changes for example [106,107]. In
mouse AD model, vatalanib, tyrosine kinase inhibitor inhibits angiogenesis
through inhibition of VEGFR signaling and reduced the number and area of A?
plaques in the cortex [108]. It suggests that anti-VEGF agents and/or
miR-361-5p mimicking locked nucleic acid (LNA) treatment has the potential to
suppress A? in plaques or plasma, so it is an approach to address future
challenges as a treatment of MCI or early AD. Interestingly, miR-361-3p has
also been reduced in AD patients’ brains and suppression of miR-361-3p induced
high expression of its target BACE1 protein; therefore, miR-361-3p mimic
administration inhibited A? plaques in the brain of mouse AD model [109]. By
deep METS analysis, miR-316-3p was targeted to the BACE1 3’UTR (position,
3538-3545) and miR-316-5p was also targeted to the BACE1 3’UTR (position,
1658-1664) with context++ score at -0.12 and -0.04, respectively. In addition,
miR-361-3p was weakly targeted to the VEGFA 3’UTR (position, 36-42) with
context++ score at -0.08. As an lncRNA, small nucleolar RNA host gene 1 (SNHG1)
has sponged miR-361-3p and miR-361-3p overexpression reversed the promotion
effect of SNHG1 in vitro, that positively regulated A? [110]. Thus, the stem
loop of miR-361 mimicking agents would be more effective than miR-361-5p alone.
Even with the FDA’s approval of the expensive aducanumab, which costs at
$56000/person/year, clinical trials of large number of anti-AD drug candidates
with A? suppression statistically failed to show effective benefits [18]. It
will not be sustainable medical care for MCI and AD, even for low-income and
perhaps high-income countries. Currently, one of the strongest proofs of the
mRNA vaccines against the COVID-19 pandemic is the ability of nucleic acid
agents, quickly delivery and cheaper, more abundant and effective worldwide.
Therefore, miR-361 stem loop mimic, as an anti-VEGF nucleic acid agent may be
an effective and sustainable tool for the prevention and treatment of
increasing MCI or early AD incidents.
The relation between
AD, MCI and brain cancer
Taken together, given the hub miRNAs in the
brain cancer and AD, there was no common miRNAs from MMPs among disease. As shown
in Table 2, hub miRNAs of AD MCI, AD USA & Germany, AD conventional, glioma
and astrocytoma are miR-483-5p (up) and miR-142-3p (down), miR-532-5p (up) and
miR-361-5p (down), miR-34a-5p (up) and miR-142-3p (down), miR-16-5p (down) and
miR-497-5p (down), and miR-106a-5p (up) and miR-20a-5p (up), respectively. DNSs
of hub miRNAs in AD MCI, AD USA & Germany, AD conventional, glioma and
astrocytoma were 55, 35, 40, 36 and 49, respectively. As DNSs of AD USA &
Germany, AD conventional and MCI were 35, 40 and 55 at the AI validated value
as an AUC of 1.0, 0.87 and 0.53, respectively, and the lower quantum energy
levels among two miRNAs as DNS would increase the risk of AD. However, the
inverse relationship of miRNA expression levels between the brain cancer and AD
was not observed in our METS analysis. Although AUC values of MIRAI in the
brain cancers was statistically significance, the miRNA prediction value for
MCI was not statistically enough to extract AD pathology; therefore, the MCI
stage at this time MMP may be still reversible to normal condition. On the
other hand, the MCI data will be some limitation via the small open data
because only two miRNA panels of MCI were used in this study. So, to confirm AD
prediction from MCI using the current data, much more VEGFA data in MCI should
be necessary. Further clinical data was needed to in silico our simulation for
building up sustainable healthcare.
Since the hub miRNAs were found in brain cancer and AD, there were no common miRNAs from MMPs among disease, see Table 2.
Table 2: Summary of major hub miRNAs and their targets in AD and the brain tumors.
Data source |
Hub miRNA |
Level |
Target |
Function |
Level |
DNS |
AI (AUC) |
|
Brain tumor |
Glioma |
miR-16-5p |
down |
HMGA1 |
Metastatic
progression |
up |
36 |
1.00 |
miR-497-5p |
down |
BCL2 |
Anti-apoptosis |
up |
||||
Astrocytoma |
miR-106a-5p |
up |
RB1 |
Tumor suppression |
down |
49 |
0.99 |
|
miR-20a-5p |
up |
CDKN1A |
Cyclin inhibition |
down |
||||
AD |
MCI |
miR-483-5p |
up |
ROCK2 |
RhoA/ROCK2 pathway |
down |
55 |
0.53 |
RP31 |
Ribosomopathy |
down |
||||||
miR-142-3p.1 |
down |
ARNTL |
Circadian rhythm |
up |
||||
US & Germany |
miR-532-5p |
up |
MRPL18 |
Ribosomopathy |
down |
35 |
1.00 |
|
miR-361-5p |
down |
VEGFA |
VEGFA pathway |
up |
||||
Conventional |
miR-34a-5p |
up |
NOTCH2 |
Brain memory |
down |
40 |
0.87 |
|
MDM4 |
Neural cell alive |
down |
||||||
BCL2 |
Anti-apoptosis |
down |
||||||
miR-142-3p.1 |
down |
ARNTL |
Circadian rhythm |
up |
Thus, miRNA biomarkers
for brain cancer, MCI and AD could be available for pathophysiologic diagnosis
and prediction. It is mathematically calculated that the USA model requires
approximate $8 trillion in medical and care expenses for an early and accurate
diagnosis of AD; therefore, cost issues are an urgent one for establishing
sustainable health management in AD. For brain cancer, early diagnosis of AD
and MCI, cheaper and simpler diagnostic tools need to be developed. METS using
MIRAI was used for etiologic analysis of data from the serum/plasma miRNA panel
of patients with AD and cancers. Consequently, it was found that the
enhancement of VEGFA via miR-361-5p down regulation is the most important cause
of AD (Figure 5).
Up regulation of the circadian rhythm-related protein, aryl hydrocarbon receptor nuclear translocator like (ARNTL) via miR-142-3p down regulation was the basic pathogenicity of MCI and AD. The causative factors directly associated with A? were not outcome; however, in deep METS analysis, miR-361-5p and miR-361-3p were targeted to BACE1 3’UTR. Our findings upon the cause of MCI-to-AD progression through angiodysplasia could contribute to the establishment of baseline references in the diagnosis, prognosis and treatment of aging diseases in the human brain. Further, analysis of hub miRNA/target proteins in glioma and astrocytoma showed up regulation of HMGA1 by miR-16-5p down regulation, up regulation of BCL2 by miR-497-5p down regulation, and down regulation of RB1 by miR-106a-5p up regulation and down regulation of CDKN1A by miR-20a-5p up regulation, respectively; therefore, no inverse correlation of common miRNA and hub miRNA levels was observed, and there was no relationship between AD and the cause of brain cancer.
Figure 5: Etiologic causes from MCI and AD onset. Molecular mechanisms in progression of MCI and AD were depicted in the middle panel and the left panel, respectively. Upregulated proteins and pathways were in red, downregulated ones were in blue. miR-361 stem loop mimic agents could block both angiodysplasia and A? accumulation in MCI and AD according to prediction of AD onset (AUC from 0.5 in MCI to 1.0 in AD).
Thus, AD, MCI, glioma
and astrocytoma can be completely differentially diagnosed by circulating miRNA
biomarkers. There is one of the strongest proofs that nucleic acid agents, such
as the mRNA vaccines against the COVID-19 pandemic has the ability of cheaper,
more abundant and effective worldwide. Since miR-361-3p and -5p were targeted to
the BACE1 3’UTR, anti-VEGFA agents and/or miR-361 stem loop mimicking locked
nucleic acid (LNA) treatment may help sustainable approaches to prevent
progression from MCI or early AD to AD or late AD.
The authors declare
that there are no conflicts of interest.