Abstract

   Alzheimer’s Disease (AD) is a progressive neurodegenerative disorder
   and the most common cause of dementia; however, early diagnosis of the
   disease is challenging. Research suggests that biomarkers found in
   blood, such as microRNAs (miRNA), may be promising for AD diagnostics.
   Experimental data on miRNA–target interactions (MTI) associated with AD
   are scattered across databases and publications, thus making the
   identification of promising miRNA biomarkers for AD difficult. In
   response to this, a list of experimentally validated AD-associated MTIs
   was obtained from miRTarBase. Cytoscape was used to create a visual MTI
   network. STRING software was used for protein–protein interaction
   analysis and mirPath was used for pathway enrichment analysis. Several
   targets regulated by multiple miRNAs were identified, including: BACE1,
   APP, NCSTN, SP1, SIRT1, and PTEN. The miRNA with the highest numbers of
   interactions in the network were: miR-9, miR-16, miR-34a, miR-106a,
   miR-107, miR-125b, miR-146, and miR-181c. The analysis revealed seven
   subnetworks, representing disease modules which have a potential for
   further biomarker development. The obtained MTI network is not yet
   complete, and additional studies are needed for the comprehensive
   understanding of the AD-associated miRNA targetome.

   Keywords: Alzheimer’s disease, protein–protein interaction (PPI),
   biomarker, microRNA (miRNA), miRNA–target interaction (MTI)

1. Introduction

   Alzheimer’s disease (AD) is a complex, multifactorial, progressive
   neurodegenerative disorder afflicting the central nervous system (CNS)
   and is the most common cause of dementia. The disease’s clinical
   progression is variable with several contributing factors, is
   irreversible and inevitably fatal [[28]1]. The cause of the disease is
   mostly still unknown. It has been associated with the accumulation of
   misfolded amyloid beta (Aβ) proteins, hyperphosphorylation of tau
   proteins, inflammation, the formation of neurofibrillary tangles, and
   single-nucleotide polymorphisms (SNPs) in certain AD-associated genes,
   such as the APOE gene [[29]2]. AD is characterized by the loss of
   neurons and synapses in the brain, leading to a gradual loss of
   cognitive function. Disease progression is divided into three clinical
   stages: preclinical, prodromal, and dementia stages. In the disease’s
   early stages, this manifests through episodes of forgetfulness, such as
   forgetting the names of family members and friends and confusion in
   unfamiliar situations. As the disease progresses, more regions of the
   brain are affected, resulting in severe difficulties with speech,
   thought, motor control, and other functions. Late-stage AD outcomes
   include irreversible disruptions to visual and visuospatial perception,
   behavioral alterations, losing one’s ability to care for oneself, and
   progressively worsening cognitive and memory faculties. The formation
   of new memories becomes highly impaired, though older memories are
   often retained [[30]1]. In 2015, it was estimated that 29.8 million
   people worldwide were living with AD [[31]3].

   Diagnosing AD is often carried out by interviewing relatives about the
   person’s overall health, medical history, drug use, and other relevant
   information. Cognitive tests can also be performed along with blood and
   urine tests. Brain scans may be used to rule out other causes of
   dementia; these include computed tomography (CT), magnetic resonance
   imaging (MRI), and positron emission tomography (PET) scans. Modern
   diagnostic methods are based on IWG-2 criteria, which rely on both
   biomarker and clinical phenotypes [[32]4]. Despite these diagnostic
   methods, a definitive diagnosis can only be made after death with the
   examination of brain tissue. This is changing, however, as advancements
   in biomarker research are allowing more accurate assessment of the
   presence of AD. Three biomarkers have been established and examined in
   depth: Aβ proteins, tau protein, and phosphorylated tau proteins. The
   current AD biomarker panel is categorized into three types of biomarker
   evidence for pathology, known together as the ATN classification
   system. This system allows individuals to be analyzed for three
   parameters: alterations of Aβ proteins (in CSF or detected with PET
   scans; A), the hyperphosphorylation of tau proteins (in CSF or PET
   scans; T), and neurodegeneration levels (PET, MRI, and others; N)
   [[33]5]. Accumulating evidence suggests that biomarkers found in blood
   (circulatory biomarkers) may be promising in identifying cases of AD.
   These include microRNAs (miRNAs), inflammatory markers, blood-based Aβ
   markers, and biomarkers for oxidative stress [[34]2].

   In recent years, progress has been made in the field of biomarker
   development for AD. Analyzing Aβ proteins in plasma together with other
   blood biomarkers can accurately detect cerebral Aβ. This method yields
   even more accurate results when combined with APOE genotyping and thus
   reduces the cost of PET scans and need for lumbar punctures [[35]6].
   CSF and plasma p-tau181 and p-tau217 levels have also been studied as
   potential early biomarkers of AD [[36]7]. These biomarkers were
   identified to be present in significantly higher concentrations in
   patients with early or mild AD. PET scans for tau proteins, on the
   other hand, presented more accurate results when gauging disease
   progression [[37]8]. It has also been shown that plasma biomarkers,
   such as p-tau217, can be used to reduce the necessary sample size for
   future clinical trials [[38]9]. Blood assay for p-tau181 has been
   identified as a promising biomarker for AD pathology. P-tau 181 has
   also been identified as a promising AD biomarker and has been shown to
   provide high diagnostic accuracy and the ability to differentiate AD
   from several other neurodegenerative diseases [[39]10].

   miRNAs are a class of short, non-coding RNA consisting of about 22
   nucleotides. They play a role in the post-transcriptional regulation of
   gene expression as they, along with their associated proteins, bind to
   mRNA. This protein–miRNA complex then mediates either the degradation,
   destabilization, or repression of the mRNA. miRNA have been shown to
   have a wide array of targets, and one miRNA can target multiple mRNAs.
   It has also been shown that over 60% of protein-coding genes in humans
   have been under selective pressure to maintain sequences that would
   allow miRNA binding. Over 2000 human miRNA have so far been identified,
   and they are involved in numerous physiological processes as well as
   disease development [[40]11,[41]12,[42]13].

   Multiple miRNAs have thus far been associated with AD [[43]2,[44]14].
   These miRNAs were identified as downregulated or upregulated in AD
   patients compared to healthy controls, and some have been shown to
   regulate AD-associated genes, such as APP and tau protein genes
   [[45]14]. Plasma concentrations of miR-15b and miR-125b have been
   proposed as biomarkers of AD pathophysiology, and the study proposes a
   pathway-based approach to therapies for AD [[46]15]. miRNA expression
   in AD is, similar to other contributing factors to the disease, highly
   heterogenous. A precision medicine approach to AD diagnostics has been
   proposed that includes: biomarker testing, and PET and MRI scans during
   the prodromal period of AD. This approach would allow more accurate
   disease trajectory predictions and treatment based on the individual
   patient’s genetic, epigenetic, and neuroimaging profile [[47]1]. miRNAs
   are also involved in both of the leading hypotheses for AD
   development—the amyloid cascade hypothesis and the tau hypothesis
   [[48]16]. Nineteen miRNAs have been identified as having diagnostic
   potential in human AD studies. Among them, miR-206, miR-181a, and
   miR-146a from blood samples have shown the ability to predict whether
   mild cognitive impairment would progress to AD [[49]17,[50]18].
   Anti-miRNA (AM) treatments have also been proposed- using miRNA
   complementary strands of RNA. The AM approach has shown promise in
   murine models and cell cultures [[51]19,[52]20]. AM strategies for AD
   patients will likely require a more individual-focused approach to
   disease treatment, tailored to the individual’s needs due to AD
   heterogeneity [[53]21]. Potential therapeutic approaches are, however,
   not limited to AM strategies. Studies on cell cultures and animal
   models have also identified compounds that affect miRNA expression
   [[54]22]. Current commonly used diagnostic methods are primarily based
   on CSF biomarkers. As circulatory miRNAs can be assessed in blood and
   do not require a lumbar puncture, their usage as biomarkers for
   diagnosis or potential treatment may be advantageous. While miRNAs may
   prove to be the preferable AD biomarker, data on their interactions
   with targets are fragmented, making it difficult for researchers to
   find a comprehensive overview of MTIs.

   The aim of the present study was to: 1. review published data on the
   currently known miRNAs associated with AD, 2. present this information
   as the miRNA–target network to identify central molecules with
   potential for biomarker and therapeutic target development, and 3.
   conduct a pathway enrichment analysis and protein interaction analysis.

2. Methods

   AD-associated miRNAs were retrieved from the online database
   miRTarBase, release 7.0
   ([55]http://mirtarbase.mbc.nctu.edu.tw/php/index.php) (accessed on 21
   September 2021) [[56]23]. The database contains experimentally
   validated MTIs. These MTIs were validated by using various
   methodologies, including reporter assay, Western blot, quantitative PCR
   (qPCR), microarrays, next-generation sequencing (NGS), and pSILAC. All
   data obtained from the database were manually reviewed. Each entry was
   also manually verified and only MTIs reported to be associated with AD
   were included in the analysis. The visual representation of the MTI
   network was made using the Cytoscape tool ([57]https://cytoscape.org)
   (accessed on 24 September 2021) [[58]24]. miRNA target genes were
   investigated for known protein–protein interactions (PPI) using the
   STRING database ([59]https://string-db.org/) (accessed on 24 September
   2021) [[60]25]. AD-associated miRNAs were also analyzed for enrichment
   in biological pathways using mirPath v.3
   ([61]http://diana.imis.athena-innovation.gr/DianaTools/index.php)
   (accessed on 2 November 2021) [[62]26]. MirPath is a prediction-based
   bioinformatics tool that enables the identification of biological
   pathways in which the query miRNAs’ target genes are enriched. The
   MirPath’s KEGG analysis tool was utilized using the following
   parameters: p-value threshold of 0.05 and conservative estimates
   applied to the MicroT-CDS search algorithm. The obtained enriched
   biological pathways were manually reviewed for association with AD in
   previously published literature.

3. Results

   A total of 37 unique miRNAs associated with AD were extracted from
   miRTarBase. The network consists of 37 miRNAs and 43 target genes,
   which are connected through 66 MTIs and experimentally validated by 45
   unique articles. A list of MTIs is presented in [63]Table 1
   [[64]27,[65]28,[66]29,[67]30,[68]31,[69]32,[70]33,[71]34,[72]35,[73]36,
   [74]37,[75]38,[76]39,[77]40,[78]41,[79]42,[80]43,[81]44,[82]45,[83]46,[
   84]47,[85]48,[86]49,[87]50,[88]51,[89]52,[90]53,[91]54,[92]55,[93]56,[9
   4]57,[95]58,[96]59,[97]60,[98]61,[99]62,[100]63,[101]64,[102]65,[103]66
   ,[104]67,[105]68,[106]69,[107]70,[108]71]. All MTIs presented in these
   results have been previously experimentally validated. The MTI network
   was visualized using Cytoscape software and is shown in [109]Figure 1.
   miR-9, miR-107, miR-125b, and miR-146a were among miRNAs with the
   highest number of interactions, with four MTIs each. miR-16, miR-34a,
   miR-106a, and miR-181c were also highly connected, with three MTIs
   each. The most prominent miRNA targets were BACE1, APP, and NCSTN,
   which were the target of seven, seven, and four miRNAs, respectively.
   miRNAs and targets connected by multiple edges represent interactions
   confirmed by multiple independent experiments, such as the connections
   between BACE1 and hsa-miR-107. The results revealed a larger subnetwork
   consisting of 18 miRNA and 15 targets. Additionally, six smaller
   subnetworks of up to four MTIs were also identified. Twelve reported
   MTIs were between pairs of a single miRNA and target and were not part
   of a larger network. The targets of these 12 MTIs were: ATG5, BAX,
   BDNF, FKBP5, FOXO1, LRP1, MFN2, RARA, RCOR1, SNX6, SORL1, and UBE2A
   ([110]Figure 1). The complete table of MTIs, which includes miRTarBase
   IDs and experimental validation methods for each MTI, is available in
   [111]Supplementary Data (Supplementary Table S1).

Table 1.

   Experimentally validated MTIs associated with AD obtained from
   miRTarBase and literature
   [[112]27,[113]28,[114]29,[115]30,[116]31,[117]32,[118]33,[119]34,[120]3
   5,[121]36,[122]37,[123]38,[124]39,[125]40,[126]41,[127]42,[128]43,[129]
   44,[130]45,[131]46,[132]47,[133]48,[134]49,[135]50,[136]51,[137]52,[138
   ]53,[139]54,[140]55,[141]56,[142]57,[143]58,[144]59,[145]60,[146]61,[14
   7]62,[148]63,[149]64,[150]65,[151]66,[152]67,[153]68,[154]69,[155]70,[1
   56]71].
   miRNA Target Gene Symbol Target Gene (Entrez Gene ID) Reference (PMID)
   hsa-miR-20a-5p E2F1 1869 19110058 [[157]39]
   hsa-miR-146a-5p CFH 3075 18801740 [[158]51]
   hsa-miR-106b-5p APP 351 19110058 [[159]39]
   hsa-miR-101-3p APP 351 20395292 [[160]59]
   hsa-miR-101-3p APP 351 21172309 [[161]67]
   hsa-miR-29b-3p SP1 6667 23435408 [[162]32]
   hsa-miR-146a-5p IRAK1 3654 23952003 [[163]43]
   hsa-miR-107 BACE1 23621 18234899 [[164]57]
   hsa-miR-107 BACE1 23621 20489155 [[165]42]
   hsa-miR-29b-3p BACE1 23621 18434550 [[166]60]
   hsa-miR-29b-3p BACE1 23621 26818210 [[167]62]
   hsa-miR-146a-5p ROCK1 6093 27221467 [[168]48]
   hsa-miR-205-5p LRP1 4035 19665999 [[169]40]
   hsa-miR-9-5p BACE1 23621 18434550 [[170]60]
   hsa-miR-29a-3p BACE1 23621 18434550 [[171]60]
   hsa-miR-520c-3p APP 351 18684319 [[172]63]
   hsa-miR-106a-5p APP 351 19110058 [[173]39]
   hsa-miR-106a-5p APP 351 18684319 [[174]63]
   hsa-miR-34a-5p BCL2 596 19683563 [[175]49]
   hsa-miR-20a-5p APP 351 19110058 [[176]39]
   hsa-miR-17-5p APP 351 19110058 [[177]39]
   hsa-miR-125b-5p CDKN2A 1029 20347935 [[178]50]
   hsa-miR-107 GRN 2896 20489155 [[179]42]
   hsa-miR-125b-5p PPP1CA 5499 25001178 [[180]34]
   hsa-miR-34a-5p SYT1 6857 22160687 [[181]54]
   hsa-miR-34a-5p STX1A 6804 22160687 [[182]54]
   hsa-miR-29a-3p NAV3 89795 20202123 [[183]27]
   hsa-miR-375 SP1 6667 23435408 [[184]32]
   hsa-miR-181c-5p TRIM2 23321 21720722 [[185]37]
   hsa-miR-181c-5p SIRT1 23411 21720722 [[186]37]
   hsa-miR-9-5p SIRT1 23411 21720722 [[187]37]
   hsa-miR-9-5p TGFBI 7045 21720722 [[188]37]
   hsa-miR-181c-5p BTBD3 22903 21720722 [[189]37]
   hsa-miR-22-3p RCOR1 23186 23349832 [[190]47]
   hsa-miR-29c-3p BACE1 23621 21565331 [[191]70]
   hsa-miR-29c-3p BACE1 23621 25973041 [[192]29]
   hsa-miR-16-5p APP 351 26440600 [[193]46]
   hsa-miR-138-5p RARA 5914 25680531 [[194]64]
   hsa-miR-125b-5p BCL2L2 599 25001178 [[195]34]
   hsa-miR-106a-5p STAT3 6774 23399684 [[196]31]
   hsa-miR-132-5p FOXO1 2308 24014289 [[197]53]
   hsa-miR-96-5p SLC1A1 6505 24304186 [[198]56]
   hsa-miR-96-5p SLC6A6 6533 24304186 [[199]56]
   hsa-miR-195-3p MFN2 9927 27693395 [[200]71]
   hsa-miR-26b-5p RB1 5925 24027266 [[201]33]
   hsa-miR-339-5p BACE1 23621 24352696 [[202]69]
   hsa-miR-214-3p BAX 581 23408966 [[203]35]
   hsa-miR-455-5p NCSTN 23385 25100943 [[204]30]
   hsa-miR-186-5p NCSTN 23385 25100943 [[205]30]
   hsa-miR-24-3p NCSTN 23385 25100943 [[206]30]
   hsa-miR-125b-5p DUSP6 1848 25001178 [[207]34]
   hsa-miR-107 SH3GL2 6456 27038654 [[208]65]
   hsa-miR-511-5p FKBP5 2289 27334923 [[209]66]
   hsa-miR-299-5p ATG5 9474 27080144 [[210]52]
   hsa-miR-98-5p SNX6 58533 27541017 [[211]28]
   hsa-miR-16-5p BACE1 23621 26440600 [[212]46]
   hsa-miR-16-5p NCSTN 23385 26440600 [[213]46]
   hsa-miR-106b-5p FYN 2534 27520374 [[214]58]
   hsa-miR-26b-5p IGF1 3479 26847596 [[215]55]
   hsa-miR-302a-3p PTEN 5728 26890744 [[216]41]
   hsa-miR-9-5p CAMKK2 10645 27394443 [[217]45]
   hsa-miR-200c-3p PTEN 5728 28008308 [[218]68]
   hsa-miR-146a-5p LRP2 4036 27241555 [[219]38]
   hsa-miR-7-5p UBE2A 7319 27929395 [[220]61]
   hsa-miR-613 BDNF 627 27545218 [[221]44]
   hsa-miR-1229-3p SORL1 6653 27328823 [[222]36]
   [223]Open in a new tab

   Legend: APP: amyloid beta precursor protein, ATG5: autophagy related 5,
   BACE1: beta-secretase 1, BAX: BCL2 associated X, apoptosis regulator,
   BCL2: BCL2 apoptosis regulator, BCL2L2: BCL2 like 2, BDNF: brain
   derived neurotrophic factor, BTBD3: BTB domain containing 3, CAMKK2:
   calcium/calmodulin dependent protein kinase 2, CDKN2A: cyclin dependent
   kinase inhibitor 2A, CFH: complement factor H, DUSP6: dual specificity
   phosphatase 6, E2F1: E2F transcription factor 2, FKBP5: FKBP prolyl
   isomerase 5, FOXO1: forkhead box O1, FYN: FYN proto-oncogene, Src
   family tyrosine kinase, GRN: granulin precursor. IGF1: insulin like
   growth factor 1. IRAK1: interleukin 1 receptor associated kinase 1.
   LRP1: LDL receptor related protein 1, LRP2: LDL receptor related
   protein 2, MFN2: mitofusin 2, NAV3: neuron navigator 3, NCSTN:
   nicastrin, PPP1CA: protein phosphatase 1 catalytic subunit alpha, PTEN:
   phosphatase and tensin homolog, RARA: retinoic acid receptor alpha,
   RB1: RB transcriptional corepressor 1, RCOR1: REST corepressor 1,
   ROCK1: Rho associated coiled-coil containing protein kinase 1, SH3GL2:
   SH3 domain containing GRB2 like 2, endophilin A1, SIRT1: sirtuin 1,
   SLC1A1: solute carrier family 1 member 1, SLC6A6: solute carrier family
   6 member 6, SNX6: sorting nexin 6, SORL1: sortilin related receptor 1,
   SP1: Sp1 transcription factor, STAT3: signal transducer and activator
   of transcription 3, STX1A; syntaxin 1A, SYT1: synaptotagmin 1, TGFBI:
   transforming growth factor beta induced, TRIM2: tripartite motif
   containing 2, UBE2A: ubiquitin conjugating enzyme E2 A.

Figure 1.

   [224]Figure 1
   [225]Open in a new tab

   Network of experimentally validated MTIs associated with AD visualized
   using the Cytoscape software. Orange-colored nodes represent target
   genes while blue-colored nodes represent miRNAs. Each edge represents
   an experimentally validated interaction between a miRNA and its target.
   (a) The largest subnetwork identified in the data set, consisting of 18
   miRNAs and 15 target genes, with a total of 37 MTIs. (b) Smaller
   subnetworks and miRNA–target pairs identified in the data set.

   Target genes (n = 43) were then explored for PPIs using the STRING tool
   ([226]Figure 2). A total of 41 of the 43 targets were part of a large
   PPI network; only two targets, UBE2A and CFH, had no known interactions
   with the rest of the proteins in the network. The network includes 43
   nodes with 110 interactions, which is significantly more interactions
   than expected (PPI enrichment p-value: <1.0^−16). Some proteins had
   notably more interactions than others, representing hubs. The five
   proteins with the most PPIs were: PTEN, SIRT1, APP, STAT3, BACE1, with
   19, 16, 16, 9, and 7 PPIs, respectively. These five proteins are
   therefore present in 67 of the 110 PPIs in the network, thus
   representing central hub proteins.

Figure 2.

   [227]Figure 2
   [228]Open in a new tab

   Protein–protein interaction network of 43 AD-associated miRNA targets
   using STRING software. The colors of connections between nodes
   represent the type interaction between the two proteins. The most
   reliable of these connections are “known interactions” from curated
   databases and experiments. Common gene neighborhoods, fusions, and
   co-occurrences as well as textmining, co-expression, and protein
   homology are considered less reliable connections and tend to require
   independent verification.

   To identify pathways in which 37 AD-associated miRNAs were enriched, we
   conducted an analysis with the mirPath tool. The analysis conducted
   using mirPath identified enrichments of the 37 unique AD-associated
   miRNAs in 68 biological pathways. Biological pathways were manually
   reviewed for association with AD based on literature. A total of 44 of
   the 68 biological pathways were associated with AD in previously
   published literature ([229]Table 2)
   [[230]69,[231]70,[232]71,[233]72,[234]73,[235]74,[236]75,[237]76,[238]7
   7,[239]78,[240]79,[241]80,[242]81,[243]82,[244]83,[245]84,[246]85,[247]
   86,[248]87,[249]88,[250]89,[251]90,[252]91,[253]92,[254]93,[255]94,[256
   ]95,[257]96,[258]97,[259]98,[260]99,[261]100,[262]101,[263]102,[264]103
   ,[265]104,[266]105,[267]106,[268]107,[269]108,[270]109,[271]110,[272]11
   1,[273]112]. Thus, a total of 87 articles were reviewed for MTIs and
   enriched biological pathways. [274]Supplementary Data (Supplementary
   Table S2) includes the full results of the pathway enrichment analysis.

Table 2.

   Results of the pathway enrichment analysis using mirPath tool. The
   table includes pathways associated with AD in previously published
   literature. A total of 37 AD-associated miRNAs were enriched in 68
   pathways. This table includes 44 pathways, which were associated with
   AD in previously published literature. The PMID (Reference) column
   includes references to publications associating the KEGG pathway with