Alu RNA and their roles in human disease states

ABSTRACT

Alu RNA are implicated in the poor prognosis of several human disease states. These RNA are transcription products of primate specific transposable elements called Alu elements. These elements are extremely abundant, comprising over 10% of the human genome, and 100 to 1000 cytoplasmic copies of Alu RNA per cell. Alu RNA do not have a single universal functional role aside from selfish self-propagation. Despite this, Alu RNA have been found to operate in a diverse set of translational and transcriptional mechanisms. This review will focus on the current knowledge of Alu RNA involved in human disease states and known mechanisms of action. Examples of Alu RNA that are transcribed in a variety of contexts such as introns, mature mRNA, and non-coding transcripts will be discussed. Past and present challenges in studying Alu RNA, and the future directions of Alu RNA in basic and clinical research will also be examined.

KEYWORDS:

• Mature mRNA
• intron
• non-coding
• splicing
• A-to-I editing
• miRNA sponge
• translation
• cryptic
• inverted repeat alu rna
• circular rna

Introduction

The initial sequencing of the human genome determined that only approximately 1.5% consisted of protein-coding sequences [1]. Instead, the vast majority of the human genome encodes other types of non-coding sequences including transposable elements, DNA sequences capable of being transported and integrated into new regions of the genome. Transposable elements are highly abundant, comprising approximately 45% of the human genome [1]. These sequences can replicate and propagate throughout the genome, increasing genetic diversity and driving evolution of the host [2]. A central propagation mechanism is the process of retrotransposition, where a transposable element transcript is later reintegrated back into the genome at a new location [3]. Retrotransposons are of interest due to their dual potential of impacting human disease states through either their DNA or RNA versions [4].

Two main classes of retrotransposons are the long interspersed nuclear elements (LINEs), and short interspersed nuclear elements (SINEs). These classes compose ~20% and ~13% of the human genome, respectively [1]. The remainder of the transposable elements are composed of retrovirus-like elements ~8%, and DNA transposon fossils ~3% [1]. LINEs are autonomous, contain an internal RNA polymerase II promoter, one or two open reading frames (ORF), and a 3ʹ-terminal polyadenylation site [5]. The only active human LINE, LINE1, encodes ORF2, a multifunctional protein that possesses both reverse transcriptase and DNA endonuclease activity [5]. ORF2 is essential for the propagation of LINEs, providing its autonomy, but is interestingly also required for the functionality of the non-autonomous SINEs. The genetic composition of SINEs includes an internal RNA polymerase III promoter, internal and terminal polyadenylation tracts, an oligothymidylate termination signal, but lack ORFs [5,6]. Unsurprisingly, both LINEs and SINEs are implicated in several human diseases, often through their insertion into promoters, or coding sequences of critical genes [7].

The largest subset of SINEs are Alu elements, which are extremely abundant elements comprising over 10% of the human genome [1]. These elements were named Alu due to their initial discovery as a family of repeated sequences, of which over half were digestible by the restriction enzyme AluI [8]. Alu elements are primate specific approximately 300 base pair sequences retrotransposed through the enzymatic activity of ORF2 immediately following its synthesis (Figure 1) [6,9]. These sequences possess internal promoters (A and B box), and both internal/terminal polyadenylation tracts (Figure. 2(a)) [6]. Once Alu elements are transcribed, the RNA folds into distinct and conserved secondary and tertiary structures, regardless of minor variations in the primary sequence found in different Alu element families (Figure 2(b)). There are about 100 to 1000 cytoplasmic copies of Alu RNA per cell, with the transcription of an estimated 99% of Alu elements repressed by methylation [10,11]. In humans, it is estimated that a new Alu insertion occurs once in every ~20 births [12]. Insertions occur throughout the genome, resulting in Alu elements present in introns, protein-coding exons, as well as non-coding regions [13]. In the presence of appropriate promoter elements, Alu elements are transcribed by both RNA polymerase II and RNA polymerase III. Alu elements located within introns and exons are transcribed by RNA polymerase II in the context of a gene. Although not all pre-mRNA possess Alu sequences, on average, approximately 16 Alu sequences are found in each pre-mRNA [14]. RNA polymerase III mediated transcription of Alu elements was originally a controversial hypothesis due to difficulties in differentiating these transcripts from other abundant Alu RNA [15]. Since then, identification of sequenced RNA polymerase III Alu transcripts has been observed, but with evidence these transcripts may require upstream RNA polymerase III elements to enhance their transcription [16,17].

Figure 1. Schematic of LINE1-mediated retrotransposition of alu elements as described by Dewannieux et al. [6]. Ribosome, DNA, RNA, and protein molecules are depicted in black, green, red, and blue, respectively. a) LINE1 mRNA is translated generating the ORF2 protein which possesses reverse transcriptase and endonuclease activity; b) The alu element located in the genomic DNA is transcribed into an RNA which is bound by the SRP9/SRP14 heterodimer to localize it to the 40S ribosomal small subunit of the ribosome translating ORF2; c) positioning of the alu RNA relative to ORF2 allows for its reverse transcription and generation of an alu cDNA; d) the alu cDNA is trafficked by ORF2 back to the genomic DNA, where ORF2 both endonucleolytically cuts the genomic DNA and reintegrates the cDNA, forming a new alu DNA insert

Figure 2. Structural elements of alu RNA. a) alu element and associated genomic regulatory regions. RNA polymerase III promoter elements, A- and B-boxes are found on the left arm of alu elements. an internal polyadenylation tract separates the left and right arms. the right arm is differentiated from the left arm by a 31-nucleotide insertion. b) Secondary structure of an AluY RNA found on intron 4 of the $\alpha$-fetoprotein gene. the structure forms two distinct arms with an internal polyadenylation tract separating them. potential acceptor and donor splice sites found in the alu consensus sequence as described by Sorek et al. were adapted to the presented AluY RNA [35]. darkly shaded boxes behind nucleotides indicate potential donor splice sites (5ʹ-GC and 5ʹ-GU for sense strand, and 5ʹ-GC and 5ʹ-AC for antisense strand alu RNA). LIGHTLY shaded boxes behind nucleotides indicate potential acceptor splice sites (5ʹ-AG for sense strand and 5ʹ-CU for antisense strand alu RNA). (c) secondary structure of the 7SL or signal recognition particle RNA. the structure is segmented into two regions, the alu domain and the S domain. the alu domain is recognized by SRP9 and SRP14, while the S domain interacts with SRP19, SRP54, SRP68 and SRP72 in the signal recognition particle [86]. this figure was adapted from work from Häsler et al. [33,80]

Alu elements are believed to be derived from the 7SL RNA, the RNA component of the signal recognition particle which regulates the translation of ~30% of the human proteome (Figure 2(c)) [18,19]. The human 7SL RNA is composed of a 100 nucleotide 5ʹ-end and 45 nucleotide 3ʹ-end homologous to the human right Alu monomer [19]. Only the central 155 nucleotide region of the human 7SL RNA is homologous to other non-primate 7SL RNA, supporting the hypothesis that the 7SL RNA is more ancient than Alu elements [19]. The molecular origin of Alu elements was then suggested to have occurred by the processing of 7SL RNA into fossil Alu monomers [20]. This was followed by specific deletions within the fossil Alu monomer to generate both free left and right Alu monomers [20]. Finally, a head to tail fusion between a free left Alu monomer and free right Alu monomer produced the modern dimeric Alu element [21].

The three major Alu element families from the most to least ancient, AluJ, AluS, and AluY families, possess amplification rates which were species- or lineage-specific and may have contributed to primate evolution [22–24]. In humans, these families diverged into 225 different subfamilies (as determined by a minimum spanning phylogenetic tree), some of the most populous including AluSx, AluSp, and AluY (subfamily) [25]. Four separate conclusions suggest the more ancient Alu elements were more likely to have experienced detrimental mutations that inhibited their ability to be transcribed and therefore unable to be retrotransposed. First, when comparing human AluJ, AluS, and AluY elements in a plasmid-based mobilization assay, it was evident only the AluS and AluY copies were capable of retrotransposition [26]. Second, AluS and AluY elements are transcribed in higher quantities than AluJ elements in human embryonic stem cells [27]. Third, removal of the A box of Alu elements reduced their transcription by 10–20 fold, and removal of the B box abolished all transcription [28]. Lastly, the longer and more homogeneous the terminal polyadenylation tract is on an Alu element, the better its retrotransposition capability [29]. This suggests the younger the Alu element is, the less likely it possesses any mutations that may inhibit its transcription, resulting in further propagation.

Alu elements have been found to coincide with various aspects of gene regulation; located in GC- and transcription unit-rich sequences [1,30]. Alu elements upstream of RNA polymerase II promoters are capable of upregulating the transcription of the downstream gene in a cell-type specific manner [31]. Unsurprisingly, Alu elements have also been implicated in a variety of human diseases as both insertions in the promoters of genes, and sources of homologous recombination [32]. Compared to Alu elements in DNA, Alu RNA possess different roles in gene regulation. Over 90% of all adenosine-to-inosine (A-to-I) RNA editing, a major source of post-transcriptional gene regulation, is conducted on Alu RNA sequences within pre-mRNA [33]. Alu RNA are also excellent sources of splice sites and contributors to alternative splicing, with the consensus Alu sequence providing ten 5ʹ and thirteen 3ʹ potential splice sites [34,35]. Examples of potential donor and acceptor splice sites in the context of an Alu RNA are presented in (Figure 2(b)). In an initial investigation into Alu elements in exons by a database search, the vast majority of Alu RNA sequences were found contained in the 3ʹ untranslated regions of genes, a common source of post-transcriptional regulatory elements [36]. Finally, in a computational analysis approximately 1300 human RNA polymerase III Alu transcripts were detected with close to 120 validated biologically in three different human cell lines [37]. These Alu RNA have the potential to cause human diseases using several different mechanisms and will be the focus of this review. Herein, this review will discuss the different Alu RNA that are found in human disease states, ranging from those found specifically in introns, mature mRNA, to non-coding RNA. The various challenges that exist in studying Alu RNA, and the future directions of Alu RNA basic and clinical research will also be described.

Introns

Approximately 25% and 10% of the human genome is composed of introns and Alu elements, respectively [1,38]. Although not all pre-mRNA possess Alu sequences, on average, approximately 16 Alu sequences are found in each pre-mRNA [14]. Alu elements were also thought to provide new sources of genetic diversity, with the formation of new exons from introns being one of the most common mechanisms in diversifying functional genes [39,40]. Alu RNA within introns and alterations or mutations to these sequences have consequences and functional roles in the development of human diseases.

Cryptic exon alternative splicing

In humans, alternative splicing is a mechanism that occurs to more than 95% of all multiexon genes [41]. Cryptic exons are alternatively spliced exons that generate a frameshift mutation or stop codon, impacting the resultant protein. Alu RNA sequences are often sources of cryptic exons due to the Alu sequence providing alternative 5ʹ and 3ʹ splice sites (Figure 2(b)). In general, antisense Alu sequences are more often spliced than those in the sense orientation due to the higher abundance of potential donor and acceptor splice sites [34,35]. An example of this mechanism is found in the mRNA of collagen type IV alpha 3 chain (COL4A3), in which mutations have been implicated in the development of autosomal recessive Alport syndrome [42]. Knebelmann et al. sequenced exons 4 to 6 of the COL4A3 mRNA using PCR with cDNA template from lymphocytes of 15 individuals with Alport syndrome [42]. Comparing the sequenced PCR product with non-Alport syndrome individuals revealed two alternatively spliced versions of the mRNA. In both unique versions of COL4A3 mRNA, there was a 74 base pair insertion of an antisense right arm Alu sequence originating from intron 5 of COL4A3. The Alport syndrome unique insertion was found to occur through a G to T point mutation in intron 5, upstream of the Alu sequence. This created an acceptor splice site and activated the donor splice site within the Alu sequence. The result of the Alport syndrome specific Alu RNA insertion was the truncation of the critical noncollagenous (NC1) domain and 24 missense amino acids in the COL4A3 protein. Another example of this mechanism was found in individuals with mild haemophilia A, where a 13 base pair deletion of intron 13 of the gene coagulation factor VIII (F8) led to the exonization of an antisense AluY sequence, resulting in a misfunctioning protein [43]. A compilation of Alu RNA relevant to this review and their connection to human disease states is presented in Table 1.

Table 1. List of alu RNA impacting gene regulation and implicated in human diseases

Gene	Human Disease*	Alu Family**	Sequence Type	Mechanism	References
COL4A3	Alport Syndrome	Unclear	Intron	Cryptic Exon Alternative Splicing	[42]
COL4A5	Alport Syndrome	AluY	Intron	Cryptic Exon Alternative Splicing	[87]
ATM	Ataxia-Telangiectasia	AluSg	Intron	Cryptic Exon Alternative Splicing	[88]
CNTN1	Brain Cancer	AluY	Mature mRNA	Secondary Structure Directly Impacting Translation	[89]
MED13	Brain Cancer	Many	Intron	A-to-I Editing	[46]
BC200	Breast Cancer	AluJ	Non-Coding	Translation Impacting	[83,85]
BRCA1	Breast Cancer	AluSq2	Intron	A-to-I Editing	[47]
BRCA1	Breast Cancer	AluSx	Mature mRNA	Secondary Structure Directly Impacting Translation	[61,62]
MDM2	Breast Cancer	Unclear	Mature mRNA	Inverted Repeat Alu RNA	[90, 30]
MDM2	Breast Cancer	Many	Mature mRNA	Alu RNA Mediated mRNA Decay	[74]
MDM4	Breast Cancer	Many	Mature mRNA	Alu RNA Mediated mRNA Decay	[74]
PMM2	Congenital Disorders of Glycosylation-Ia Syndrome	AluJb	Intron	Cryptic Exon Alternative Splicing	[91]
DMD	Duchenne-Type Muscular Dystrophy	Unclear	Intron	Cryptic Exon Alternative Splicing	[92]
PDZK1	Gastric Cancer	AluSg	Mature mRNA	Secondary Structure Directly Impacting Translation	[93]
HIPK3	Gliomas	AluSz and AluSq2	Intron	Circular RNA	[54,55]
ZNF655	Gliomas	AluSg	Mature mRNA	Alu RNA Mediated mRNA Decay	[94]
OAT	Gyrate Atrophy	AluSx4	Intron	Cryptic Exon Alternative Splicing	[95]
F8	Haemophilia A	AluY	Intron	Cryptic Exon Alternative Splicing	[43]
PSMB2	Hepatocellular Carcinoma	AluSc and AluJb	Mature mRNA	Inverted Repeat Alu RNA	[66,67]
NOSIP	Hirschsprung’s Disease	AluSp	Mature mRNA	Secondary Structure Directly Impacting Translation	[93, 96]
PTS	Hyperphenylalaninemia	AluSq2	Intron	Cryptic Exon Alternative Splicing	[114]
LRRC56	Impaired Mucocillary Clearance	AluY	Mature mRNA	Secondary Structure Directly Impacting Translation	[93,115]
AKR1A1	Laryngeal Cancer	AluJo	Mature mRNA	Secondary Structure Directly Impacting Translation	[93,116])
XIAP	Leukaemias	Unclear	Mature mRNA	Inverted Repeat Alu RNA	[30, 117]
POLR3A	Leukodystrophy	AluSx	Intron	Cryptic Exon Alternative Splicing	[118]
1/2sbsRNA	Many	Many	Mature mRNA	Alu RNA Mediated mRNA Decay	[70]
C19MC	Many	Many	Intron	miRNA Sequence Sources	[50,52]
CYP20A1	Many	Many	Mature mRNA	miRNA Sponge	[76]
MnSOD	Melanomas	AluE	Mature mRNA	Secondary Structure Directly Impacting Translation	[63,64]
GLI1	Multiple Myeloma	AluY and AluSx	Mature mRNA	Inverted Repeat Alu RNA	[65]
TOMM40	Neurodegenerative Diseases	AluY	Intron	A-to-I Editing	97
SMN	Spinal Muscular Atrophy	AluY, AluSx1, and AluSq	Intron	Circular RNA	98
ATP7B	Wilson’s Disease	AluSq	Intron	Cryptic Exon Alternative Splicing	99

* – The mentioned human disease is only one example of those correlated with the given gene, and others not mentioned may exist. ** – Alu families were labelled as the original authors presented them or determined using the UCSC human genome browser (Kent et al., 100). If the Alu family could not be determined it was labelled as unclear in the table.

A-to-I editing

The process of A-to-I editing is performed by adenosine deaminase RNA specific (ADAR) proteins, enzymes that catalyse the adenosine deamination at the C6 position on pre-mRNA [44]. Deamination leads to the generation of an inosine that is read as a guanosine by the translation machinery and can initiate structural alterations in the pre-mRNA. Alu RNA sequences are the target for 90% of all A-to-I editing events, and essentially all adenosines on Alu RNA have evidence of being editable [33,45]. Alu RNA are popular targets for ADAR proteins to conduct A-to-I editing due to their capability of forming double stranded RNA structures, as discussed later in the review in the Inverted Repeat Alu RNA section. An interesting correlation with A-to-I editing and human cancers was found by Paz et al. studying a database of reported A-to-I editing sites in 236 cancerous and 471 non-cancerous (healthy) surgically removed brain tissue libraries [46]. They found an overall reduction in A-to-I editing levels in brain tumours relative to the healthy brain tissue, with many of the reductions occurring on Alu RNA sequences. To confirm these results, they focused on intron 9 of the gene mediator complex subunit 13 (MED13), which possesses several Alu sequences. Using direct sequencing on RNA extracted from 5 normal and 7 cancerous brain tissues from different individuals, the overall A-to-I editing events on several Alu RNA sequences on intron 9 of MED13 was reduced in the cancerous tissues. This was further validated by cloning of PCR products of 75 and 46 cDNAs of the Alu sequences from healthy and cancerous brain tissues, respectively. Using the RNA extracted from the healthy and cancerous brain tissue, and quantitative PCR with primers specific to ADAR, adenosine deaminase RNA specific B1 (ADARB1), and adenosine deaminase RNA specific B2 (ADARB2), there was an overall reduction in their mRNA levels detected. This suggests the reduction in A-to-I editing likely originates in the reduction of the enzymes that catalyse the process. Analysis of RNA sequencing data from other human cancers such as prostate, kidney, testis, and lung also corroborated the reduced A-to-I editing levels relative to healthy tissues. Although the overall trend was a reduction in A-to-I editing, the authors also observed in the same brain cancer tissues increased A-to-I editing activity on the AluSq2 sequence on intron 2 of breast cancer susceptibility gene 1 (BRCA1), a critical tumour suppressor protein involved with brain metastasis [47]. This implies there is a level of complexity in the prioritization and specification of A-to-I editing activities in disease states and warrants further investigations.

miRNA sequence sources

miRNA were originally thought to be transcribed from an miRNA gene and processed down into a functional short ~21 nucleotide miRNA [48]. This original mechanism, however, is now understood not to be comprehensive, as additional sources of miRNA have been discovered such as either arm of an Alu RNA sequence [49]. A prime example of this Alu-related miRNA biogenesis pathway involves what are thought to be several introns of the pre-mRNA of the Chr19q13.41 miRNA cluster (C19MC) [50]. This is the largest human miRNA gene cluster, with a total of 46 tandemly repeated, primate specific pre-miRNA genes transcribed. Borchet et al. were the first to attempt to elucidate the transcriptional mechanism of this miRNA cluster [51]. They proposed these miRNA were transcribed by RNA polymerase III using promoter elements within Alu elements, as they noted that within 100 base pairs upstream of 42 of these miRNA sequences there were Alu element sequences present. In an effort to validate this observation, chromatin immunoprecipitation was performed in cervical HeLa and kidney HEK293 cell lines with antibodies specific to RNA polymerase II, RNA polymerase III, and the specific RNA polymerase III subunits TFIIIB and TFIIIC. Their work demonstrated that only RNA polymerase III associated antibodies, and not RNA polymerase II, could detect the genomic sequence of the miRNA. This finding was later disputed by Bortolin-Cavaille et al. who classified this cluster as an RNA polymerase II target and intronic [50]. They determined that C19MC miRNA were primarily expressed in placental cells in an expression panel of 20 different human tissues. Using choriocarcinoma (placental) JEG3 cells, a chromatin immunoprecipitation was performed with an antibody specific to RNA polymerase II which co-immunoprecipitated the three transcribed miRNA from the C19MC cluster. In agreement with the expression panel, no enrichment of these miRNA were found in the chromatin immunoprecipitations completed in the cervical HeLa or kidney HEK293 cell lines. Finally, they utilized an $\alpha$-amanitin treatment, which specifically targeted RNA polymerase II transcripts to demonstrate that the three probed pre-miRNA were downregulated, while tRNA^Tyr and snaR-A, RNA polymerase III transcripts, were not impacted. This suggested the original findings by Borchert et al. were true for the cell lines selected, but the biologically relevant result was conducted by Bortolin-Cavaille et al., determining C19MC provided RNA polymerase II transcribed miRNA [50,51]. To be noted, one of the miRNA generated in this cluster, miR-519b, has been previously demonstrated to inhibit colorectal cancer proliferation and be downregulated in individuals with colorectal cancer [52]. Although the transcriptional mechanism has yet to be confirmed, the role of this primate specific miRNA cluster and its Alu RNA sequences are clearly important in cancer studies.

Circular RNA

The possession of several 5ʹ and 3ʹ splice sites, and capability to form inverted repeats (Fig. 3) (see Cryptic Exon Alternative Splicing and Inverted Repeat Alu RNA) enable Alu RNA to be excellent precursors for circular RNA biogenesis [34,35]. Circular RNA are the products of a back-splicing mechanism, whereby two different introns interact through complementary base pairing and circularize the exon they flank. This results in usually unpaired 5ʹ and 3ʹ splice sites brought into close proximity, allowing for splicing and generation of a circular RNA [53]. One well studied circular RNA, homeodomain interacting protein kinase 3 (HIPK3), contains an AluSz sequence in the antisense direction on intron 1, and a AluSq2 sequence in the sense direction flanking exon 2. Liang et al. demonstrated that these sequences were required for the circularization of exon 2 [54]. Plasmids with truncations of the Alu sequences transfected into HeLa cells revealed the minimum required region being nucleotides 300–331 of the AluSz sequence and nucleotides 157–188 of the AluSq2 sequence. These sequences were complementary to each other, suggesting an inverted repeat Alu RNA structure must be formed to generate the circular RNA.

Figure 3. Schematic of an inverted repeat alu RNA. these structures arise when two different alu RNA sequences on a single pre-mRNA are transcribed anti-parallel to each other. due to their homology and orientation, these sequences can complementary base pair to form an inverted repeat. the formation of an inverted repeat alu RNA can also occur in a trans-acting mechanism, whereby alu RNA from two separate molecules interact to generate the double stranded structure [70]. inverted repeat alu RNA are targets for gene regulation as targets for STAU1 in mRNA decay, A-to-I editing by ADAR proteins, as well as precursors to back-splicing to allow for circular RNA formation [44,54,70]

Two separate studies have identified this HIPK3 circular RNA to function as an miRNA sponge, an RNA molecule that binds to miRNA to minimize their accessibility to their target RNA. The first, shown by Hu et al., found HIPK3 circular RNA interacting and preventing the normal function of miR-124-3p, which is an miRNA critical to glioma progression [55]. Using luciferase assays in U87 and U251 (derived from glioblastoma tumours) cells, the authors demonstrated an interaction between the 3ʹ UTR of signal transducer and activator of transcription 3 (STAT3) mRNA and miR-124-3p, as well as the HIPK3 circular RNA with miR-124-3p. Further validation of the downregulation of STAT3 under either HIPK3 circular RNA siRNA knockdown or transfection with an miR-124-3p mimic suggested the HIPK3 circular RNA sponged miR-124-3p and led to subsequent glioma progression. A similar sponging mechanism by HIPK3 circular RNA for miR-30a-3p was elucidated by Shan et al. and indicated its role in diabetes mellitus induced retinal vascular dysfunction [56].

Mature mRNA

Mature protein-coding transcripts possess multiple exons within open reading frames as well as flanking 5ʹ and 3ʹ untranslated regions (UTRs). UTRs are a source of post-transcriptional regulation, prone to translational inhibition, modulation of transcript stability, and a variety of other regulatory processes [57,58]. Although less common, exons in open reading frames can also be the source of post-transcriptional regulation including targets for miRNA [59]. Alu RNA sequences can be found in mature mRNA, in mRNA 5ʹ UTR (14%), coding sequence (4%), and 3ʹ UTR (82%), but emerging evidence suggests an underestimation of Alu RNA in coding sequences [36,60]. Not surprisingly, the roles of Alu RNA in mature mRNA are important in understanding several human disease states.

Secondary structure directly impacting translation

The translation of specific transcripts can be inhibited due to the presence of an Alu RNA secondary structure on mRNA UTR. An example of this involves BRCA1, a DNA repair protein which often has reduced mRNA levels in breast tumours, suggesting its translation is critical for healthy cell function [61]. A breast cancer tissue specific BRCA1 transcript variant possesses an AluSx sequence in its 5ʹ UTR. Sobczak et al. utilized luciferase assays to demonstrate the reduced translation efficiency of the BRCA1 transcript variant with the 5ʹ UTR AluSx sequence compared to the variant lacking the Alu sequence [62]. In-vitro translation assays were also performed in competition with a methylated guanosine cap analogue to confirm BRCA1 was translated in a cap-dependent mechanism. From these results, the authors hypothesized a model in which the AluSx sequence on the 5ʹ UTR of the transcript could inhibit BRCA1 translation by preventing the 40S small ribosomal subunit from scanning the mRNA for the start codon.

Another similar example involves the mRNA of superoxide dismutase 2 (MnSOD), an antioxidant enzyme that protects cells from oxidative damage and believed to be a tumour suppressor gene [63]. This mRNA holds an AluE sequence in its 3ʹ UTR and was thought to potentially hybridize with a secondary, small ~300 nucleotide trans-acting Alu RNA [64]. Again, using luciferase assays in-vitro, Stuart et al. demonstrated that the AluE RNA on the 3ʹ UTR of MnSOD inhibited translation significantly [64]. Their polysome profiling results showed MnSOD mRNA migrating alongside actively translating ribosomes. Consequently, they proposed the mechanism of action involved a hybridization with a small cytoplasmic Alu RNA which then prevented translation of MnSOD. Whether this led to decreased stability of the transcript or inhibition of translation elongation was unclear, but the polysome profiling results suggest the effect was not isolated to translation initiation. Although intriguing, this conclusion should be considered with caution, as the hybridized Alu was not validated with a secondary method but relied on a single detection probe.

Inverted repeat alu RNA

The high level of homology of Alu RNA sequences are exploited in the complementary base pairing of inverted repeat Alu RNA structures. These structures consist of two different transcribed Alu RNA, one in the sense, and one in the antisense direction (Figure 3). Alu RNA in general are targets for 90% of all A-to-I editing events, favouring the double stranded RNA structure of an inverted repeat Alu RNA structure over the stand-alone Alu RNA [33,44]. One example of an inverted repeat Alu RNA leading to a disease state is through the glioma-associating oncogene homologue 1 (GLI1) mRNA, a transcription factor involved in the hedgehog signalling pathway and multiple myeloma relapse [65]. The mRNA of GLI1 possesses an inverted repeat Alu RNA (AluY and AluSx) on exon 12 which corresponds to its coding sequence. Lazzari et al. utilized RNA editing site-specific quantitative real time PCR to quantify A-to-I editing levels at this GLI1 inverted repeat Alu RNA [65]. It was concluded that A-to-I editing at this GLI1 inverted repeat Alu RNA occurred at significantly higher rates in samples from individuals with multiple myeloma and plasma cell leukaemia compared to age-matched controls. This was mechanistically important due to the R701G amino acid mutation generated by the A-to-I editing event. The R701G transcript variant was also previously reported to have higher stability and demonstrated with luciferase assays to translate more efficiently.

This type of mechanism is not exclusive to coding sequences but can also occur in UTR of mRNA, with presented examples found in the 3ʹ UTR. One example is found with proteasome subunit 20S subunit beta 2 (PSMB2), which has been implicated as critical in hepatocellular carcinoma proliferation [66]. The mRNA of PSMB2 possesses an inverted repeat Alu RNA (AluSc and AluJb) in its 3ʹ UTR. Capshew et al. sought to compare the capability of the 3ʹ UTR of PSMB2 to regulate translation between the endogenously oriented Alu RNA sequences compared to mutant forms [67]. Utilizing luciferase assays in both HEK293 and HeLa cells, PSMB2 translation efficiency was ~2.5 fold higher in the endogenous 3ʹ UTR compared to when one or the other Alu sequence were oriented in the opposite direction. They were also able to demonstrate this effect was independent of A-to-I editing. In HeLa cells with ADAR1 stably knocked down, very similar luciferase assay results were obtained. Together, these examples suggest the regulation of gene expression through inverted repeat Alu RNA may require or be independent of A-to-I editing. An additional 280 3ʹ UTR inverted repeat Alu RNA were detected by Chen et al., including many genes implicated in human disease states [68].

Alu RNA Mediated mRNA decay

The decay of mRNA to prevent their translation is a well-established biological process to regulate transcripts post-transcriptionally. These mechanisms can involve miRNA, small ~21 nucleotide RNA that target 3ʹ UTR of specific mRNA by partial complementary base pairing and recruitment of the miRNA induced silencing complex [48]. Another mechanism for mRNA decay (not involved in innate immunity) involves Staufen 1 (STAU1), binding specific double stranded RNA structures after the termination signal on the 3ʹ UTR of mRNA and recruiting mRNA decay protein partners like UPF1 [69]. Gong et al. discovered a set of long non-coding RNA possessing an AluSx or AluJb sequence, capable of duplexing with Alu sequences on the 3ʹ UTR of non-sense mediated decay targets like serpin family E member 1 (SERPINE1) and ankyrin repeat domain-containing protein 57 (FLJ21870) [70]. These set of long non-coding RNA were labelled half-STAU1-binding site RNAs (1/2sbsRNA) due to their possession of only half the sequence required for a STAU1 binding site. The authors found these RNA through a few different corroborating methods, one of which included a calculation of the free energy change ($\mathrm{\Delta }G$) according to predicted duplexing. They also discovered co-immunoprecipitation of these mRNA with STAU1 and tagged 1/2sbsRNA, and an approximately 2-3-fold upregulation of the mRNA upon STAU1 or UPF1 siRNA knockdowns. Tumour protein p53 (TP53) is a well studied tumour suppressor gene and transcription factor that has a high prevalence of mutations in a variety of cancers [71]. The $\mathrm{\Delta }G$ calculation predicted that 1/2sbsRNA1 would be able to duplex with the AluJb sequence on the 3ʹ UTR of TP53. Also, upon STAU1 siRNA knockdown, TP53 mRNA levels increased by 2.2-fold according to a microarray analysis. Together, the interplay between TP53 and 1/2sbsRNA1 are an attractive target for future TP53 studies.

The miRNA method of targeted decay is also found to often occur through Alu sequences, bolstered by evidence of coevolution between miRNA and Alu RNA [72,73]. One example of this was shown by Hoffman et al., who utilized a streptavidin pull-down of a biotinylated miR-661 miRNA, to demonstrate its direct interaction with mouse double minute 2 (MDM2) and mouse double minute 4 (MDM4) mRNA [74]. The three predicted miR-661 target sites in the MDM2 3ʹ UTR overlapped with three different Alu RNA sequences, and eight of the nine predicted sites in the MDM4 3ʹ UTR overlapped with Alu sequences. It was also shown that miR-661 is positively correlated with good cancer prognosis in patients, and with patients with a mutation in the TP53 gene. Considering these data, and the fact that MDM2 and MDM4 are negative regulators of p53 suggests that Alu RNA have a role in regulating p53-related pathways in cancers.

Alu-miRNA sponge (mature mRNA)

miRNA are important biologically to prohibit the translation of a specific mRNA, but there are also RNA molecules that work to prevent miRNA from reaching their target. This class of RNA are called competitive endogenous RNA (ceRNA) or miRNA sponges and are found on non-coding RNA, and expressed on 3ʹ UTR of mRNA [75]. Similarly to those described in the Circular RNA section, another example of an miRNA sponge exists on the 3ʹ UTR of the cytochrome P450 family 20 subfamily A member 1 (CYP20A1) mRNA. This gene has 9 transcript variants, one of which possesses an ~9 kb 3ʹ UTR, with 65% of the sequence transcribing 23 Alu RNA sequences of different subfamilies including AluSx, AluSp, AluSc, AluSz6, AluSq2, AluSx3, AluSc8, AluSx1, AluSz, AluSg AluJo, AluJb, and AluJr [76]. Bhattacharya et al. searched the Alu-rich 3ʹ UTR of CYP20A1 through the miRTarBase software and found 169 different predicted miRNA target sites [76]. 46 of these predicted miRNA were classified as functional miRNA according to the FuncMir database and approximately 50% of these miRNA were primate or human specific [76]. They performed further bioinformatic analyses to elucidate a potential group of 140 miRNA with affinity to the 3ʹ UTR of CYP20A1 and followed by validation using small RNA-sequencing. This created a list of candidate miRNA which had more than 10 miRNA recognition elements (4–10 are usually found for each miRNA), and at a minimum 50 transcripts per million in MCF-7 and primary neurons. There were 7 miRNA targets that fit these criteria and were found in both MCF-7 and primary neurons, one of which included miR-1304. This miRNA has been previously reported to promote gastric cancer progression by the upregulation of MYC, a well-studied oncogenic protein [77]. This field of research is expanding, and further investigations are required to elucidate novel miRNA sponges and their downstream impacts on human diseases.

Non-coding RNA

The human genome is composed of approximately 98.5% non-coding sequences [1]. The final functional product of a non-coding sequence is the transcriptional product rather than the translational one. These non-coding RNA are transcribed by both RNA polymerase II and RNA polymerase III [78,79]. Alu RNA possess internal RNA polymerase III promoters, but often require upstream elements to be efficiently transcribed [16]. Although initially not thought to be transcribed, the detection of nearly 1300 human RNA polymerase III Alu transcripts and subsequent validation of ~120 transcripts in three different human cell lines reaffirmed the presence of these non-coding RNA [15,37]. These Alu RNA are less studied than those related to mature mRNA and introns, but their impact on translation, sponging miRNA, and general accumulation indeed lead to a variety of human disease states.

Translation impacting

Alu RNA have been shown in vitro to be able to regulate translation initiation as individual Alu RNA, or through a ribonucleoprotein complex with the signal recognition particle 9 (SRP9)/SRP14 heterodimer [80]. There are also arguments that Alu elements lack any selective pressure to exist except for elevating translational levels [81]. One prominent example of a non-coding Alu RNA implicated in disease states and translation regulation is brain cytoplasmic RNA 1 (BC200). This RNA is 200 nucleotides long, with 119 nucleotides homologous to the left arm of an AluJ sequence [82]. BC200 is expressed minimally in all healthy human tissues except for brain, but are highly expressed in breast, lung, ovarian, and other tumours [83]. Unlike most Alu RNA, there is a unique 80 nucleotide sequence that could be targeted for degradation, without impacting other genes or Alu RNA levels. The knockdown of BC200 was demonstrated by Booy et al. with the use of locked nucleic acid (LNA) GapmeRs in 10 different cell lines (combination of tumour and primary tissues) and shown to induce apoptosis in all 10 cell lines [83]. In follow-up articles, Booy et al. identified protein binding partners to BC200 by mass spectrometry which unsurprisingly (given its AluJ RNA secondary structure) included SRP9 and SRP14 [84]. They also conducted puromycin incorporation assays and polysome profiling experiments on stably overexpressing or LNA knockdowns of BC200. These experiments revealed a positive correlation between BC200 RNA levels and translational initiation and elongation rates [85]. The exact mechanism of function of BC200 has yet to be elucidated, but two of its binding partners, SRP9 and SRP14, are critical proteins in the signal recognition particle which is responsible in regulating nearly a third of all protein translation [86]. Consequently, it would not be surprising if the translational regulation role of BC200 and its pathogenesis was related to the alteration of the normal signal recognition particle function.

Alu-miRNA sponge (non-coding RNA)

As described earlier, Alu RNA derived UTR sequences can be implicated as miRNA sponges, and non-coding forms also exist. The RNase DICER1, a component of the miRNA-induced silencing complex, is responsible for the processing of Alu RNA [101]. The downregulation of DICER1 had been reported to be associated with poor prognosis in several cancers [102]. Di Ruocco et al. showed using quantitative PCR, that DICER1 deficits lead to an accumulation of AluY RNA previously detected from a cDNA sequencing [clone TS 103; 17] in two different colorectal cancer cell lines (SW480 and SW620) [102]. The reduction of DICER1 has also been known to induce epithelial to mesenchymal transitions by eliminating the abundance of specific miRNA, like miR-566 which is downregulated in colorectal cancer progression. By transfecting antisense oligonucleotides specific to both Alu RNA and DICER1 mRNA, cells were rescued from epithelial to mesenchymal transitions compared to the cells with only DICER1 mRNA antisense oligonucleotides transfected. Also, using luciferase assays with the Alu RNA conjugated to the luciferase transcripts, SW480 cells co-transfected with a miR-566 mimic had decreased translational levels of the luciferase reporter compared to an miRNA negative control. This suggested the Alu RNA was acting as an miRNA sponge for miR-566 corroborating the impact of DICER1 reductions in this mechanism.

Alu RNA accumulation

Cell stress has been documented to induce the accumulation of all SINEs, including Alu RNA [103]. Furthermore, Herpes simplex virus infection in humans induces RNA polymerase III mediated transcription of Alu RNA [104,105]. One example of Alu RNA accumulation leading to pathogenesis in humans involves hyperglycaemia in human umbilical vein endothelial cells. This was presented by Wang et al., where hyperglycaemia induced by oxidative stress accumulated AluSc subfamily transcripts [106]. When an AluSc RNA sequence was overexpressed in human umbilical vein endothelial cells three main results were observed. First, levels of reactive oxygen species and the pro-inflammatory interleukin 1$\beta$ were both increased. Second, the mRNA levels of nitric oxide synthase 3 (eNOS) and MnSOD, two proteins critical in nitrogen oxide biosynthesis, were downregulated, and MnSOD translation was also downregulated. Lastly, the AluSc RNA induced nuclear factor kappa B (NF$\kappa$B) signalling leading to endothelial oxidative stress. Kaneko et al. exemplified the accumulation of the AluY RNA from the cDNA clone TS 103 being cytotoxic to retinal pigment epithelial cells and consequently caused blindness through geographic atrophy [101]. Together, these authors revealed the impact of the accumulation of Alu RNA igniting a diverse set of pathogenesis in hyperglycaemia, a main characteristic in diabetes, and geographic atrophy with blindness.

Challenges and future directions

Challenges

The study of Alu RNA in human disease states is an attractive field for understanding novel aspects of gene regulation and development of potential therapeutics but come with challenges. The most prominent challenge in studying Alu RNA is their extremely high abundance (over 10% of the genome) and homology (225 different subfamilies) [1,25]. Probing a single Alu RNA species often utilizes either northern blots or primer-based amplifications. In both instances, non-specific binding to other Alu RNA can easily mask the identity and quantification of the Alu RNA of interest. This can be visualized in northern blots where probes cross hybridize with the 7SL RNA [10]. As a similar concern, identification, and quantification of specific Alu RNA sequences is dependent on the precision of cellular RNA extractions. Any contaminating genomic DNA would likely dominate the signal of northern blots and primer-based amplifications. These challenges address the non-specific binding of probes, but the processing of Alu RNA by DICER1 should also be noted. As described in this review, specific Alu RNA can be sources of miRNA. In a recent study, endogenous DICER1 and ADAR1 were identified to process intronic inverted repeat Alu RNA into an miRNA-like product [107]. Consequently, probes for northern blots and primers for PCR oriented work can underestimate the levels of the full-length Alu RNA if they target a processed region of the Alu RNA. Together, the identification and quantification of specific Alu RNA can be quite difficult and probes require a rigorous set of experiments or direct sequencing to validate results.

Another factor to take into consideration for studying the impact of Alu RNA are the difficulties in attempts to perturb their expression in cells. Often to overexpress a specific RNA in mammalian cells, the sequence of interest is inserted downstream of a promoter within a plasmid and these plasmids are transfected into the cells. With Alu RNA, this can be difficult to achieve since they possess internal RNA polymerase III promoters, and their transcription is often modulated by upstream elements in the genomic context [16]. Furthermore, with other RNA, tuning of the expression is often conducted by a variety of truncations of the promoter within the plasmid upstream of the target RNA. Alu elements possessing internal promoters complicates this process, as the element cannot be truncated or mutated without potentially disrupting the structure and function of the RNA. The accumulation of Alu RNA can also be toxic to cells and could limit experiments probing the impact of overexpressed Alu RNA [101]. It should also be considered that epigenetic factors can regulate the transcription of Alu RNA. The methylation of Alu elements is a dynamic process that changes throughout the development of humans, with aberrant methylations in some cases being early events in tumourigenesis [108]. Consequently, in studying human disease states, it may be difficult to mimic biological processes that require epigenetic factors when overexpressing an Alu RNA by transient transfections.

The knockdown of specific Alu RNA can also be difficult to achieve. Since Alu RNA are highly homologous and abundant, using RNA interference technology to knockdown a specific Alu RNA through its Alu domain would have wide ranging impacts and confound any obtained result. This complicates the study of specifically non-coding Alu RNA unless they possess unique sequences that are targetable. One example of this is with BC200, where Booy et al. utilized locked nucleic acid (LNA) GapmeRs targeted to a sequence within the unique ~30 nucleotide region to knockdown the left arm of AluJ containing long non-coding RNA [83]. LNA GapmeRs are used as an RNA interference technology as the LNA can hybridize with specific RNA sequences, and the LNA-RNA hybrid has a structure that mimics a DNA-RNA hybrid [109]. This LNA-RNA hybrid is then degraded by an RNase H mediated mechanism. Often multiple siRNA or LNA GapmeRs are utilized to rule out off-target effects of a knockdown. When the targetable region is short, the choices for regions to target are limited already, but not all siRNA-like sequences are able to anneal to their targeted region. This is one of the largest limitations in studying these Alu RNA as conventional experiments must be discarded for more complicated procedures.

Future directions

One of the prevailing goals of understanding the molecular biology of Alu RNA is to eventually incorporate their regulatory processes into therapeutics. Non-coding Alu RNA are attractive targets for therapeutics since they are already solely expressed as RNA, and RNA interference technologies are often LNA/RNA molecules. In both instances, the required components for the treatment can be easily delivered into cells with current technologies, and cross-reactions between the delivered molecule and cellular components could be limited [110]. An example of an Alu RNA that could be used for this type of treatment includes the delivery of LNA GapmeRs to knockdown BC200 in tumour cells. Also, two AluJb containing non-coding RNA have treatment potential when upregulated. The first, NDM29, was found to slow down cell proliferation in a neuroblastoma cell line (SK-N-BE(2)) [111]. The second is 21A, an RNA whose levels negatively correlate with tumour proliferation rates [112–113]. One issue that may arise from the study of these Alu RNA used as therapeutics are the fact that Alu elements are primate specific sequences. There are analogous sequences in mice, B1 elements, which are also SINEs but are unique compared to Alu elements due to their shorter length (approximately 130 nucleotides long) [99]. This should be accounted for in biological studies in mouse-models as the mechanism of action may not overlap in the two species.

In terms of preventative measures, there needs to be a concerted effort in identifying novel Alu RNA. There is a new Alu insertion for every ~20 births, and over time it should be expected that these insertions would be eventually transcribed [12]. As discussed throughout the review, Alu RNA can both induce and prevent human diseases, and discovery and research of these molecules has the potential for large impacts on human health. As one example, a computational analysis of possible human RNA polymerase III Alu RNA transcripts found ~1300 and ~120 were validated to be expressed in three different cell lines [37]. Any of these Alu RNA when overexpressed could have the same beneficial impact as NDM29 and 21A or detrimental impact of BC200. On the other hand, Alu RNA with their preserved splice sites, as discussed previously, can be dangerous for the induction of a myriad of diseases. The development of a system for predicting potential point mutations that could cause the exonization of known Alu RNA sequences and development of cryptic exons would have a great benefit in detecting potential diseases.

Conclusion

Alu RNA, the transcripts of an element that comprises over 10% of the human genome provide a variety of mechanisms by which human disease states are introduced or prevented [1]. In humans, on average ~16 Alu elements are found on each pre-mRNA, and ~1300 non-coding Alu RNA have been identified [14,37]. Categorization of Alu RNA found in introns, mature mRNA, non-coding RNA, and the challenges and future directions in studying them were discussed. The mechanisms of pathogenesis of these Alu RNA were diverse, with some examples including miRNA sponging, cryptic exon alternative splicing, and regulation of translation. Overall, the importance of studying Alu RNA and their roles in human disease states was demonstrated.

Acknowledgments

The authors want to acknowledge the help of Dr. Ayush Kumar and Dr. Pavel Dibrov for their support and guidance in the writing of this review. The authors also want to acknowledge the editing and proof-reading conducted by Dr. Evan Booy.

Disclosure statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Additional information

Funding

This work was supported by a CIHR Project Grant [427781]. D.G. was supported by an NSERC CGS-M.