This article is mentioned in:

Introduction

Under normal conditions, hematopoietic stem cells (HSCs) can differentiate into mature blood cells of all lineages (1). Genetic stability of HSCs requires strict regulation, since genetic alterations can compromise lineage commitment (2). Myelodysplastic syndrome (MDS) is a malignant disorder that is characterized by the aberrant differentiation of HSCs (3). During MDS, genomic instability in HSCs leads to the accumulation of somatic mutations, defective differentiation and subsequent progression to leukemia (3). Genomic instability can result in DNA damage, which can be caused by multiple sources, including exposure to environmental genotoxic substances, such as polycyclic aromatic hydrocarbons or pesticides, or to endogenous generation of reactive compounds of metabolic origin, such as reactive oxygen species (4). However, the human genome harbors a plethora of potential insertional mutagens within its own architecture in the form of transposable elements (TEs). The deleterious effects of TEs primarily lie in their inherent mobile nature, where their expression can lead to insertion mutagenesis and chromosomal rearrangements, thereby causing genomic instability (5). This can in turn lead to the development of different types of cancer, including leukemia (5). During evolution, a key mechanism that functions to protect genome integrity against the potentially harmful mobilization of TEs has emerged in the form of a family of small non-coding RNAs, known as P-element-induced WImpy testis (Piwi)-interacting RNAs (piRNAs) (6).

Despite considerable advances in the understanding of MDS pathogenesis, the influence of TEs and piRNAs on genome instability in hematopoietic cells has not been defined sufficiently. Therefore, the present review summarizes the current knowledge of the transcriptional activity and function of TEs and piRNAs in hemato-oncology, with particular emphasis on the potential involvement of these non-coding molecules in the pathogenesis of myelodysplasia.

Myelodysplastic syndrome (MDS)

MDS represents a spectrum of HSC disorders that is characterized by impaired hematopoiesis, peripheral blood cytopenia and a tendency toward leukemic transformation (3). Acute myeloid leukemia (AML) with myelodysplasia-related changes (AML-MRC) gradually develops in 30-40% patients with MDS (7). In recent years, several new therapeutic agents have been approved for the treatment of MDS, where hypomethylating agents, such as azacitide or decitabine, have been found to be effective for treating MDS and AML-MRC (8,9). These hypomethylating agents can improve the overall survival, clinical outcomes and quality of life in a proportion of patients (overall response rate, 40-50%) (10,11). Although the precise mechanism of their action remains the subject of scientific investigation, DNA hypomethylation is hypothesized to reverse the inactivation of tumor-suppressor gene transcription (9).

MDS is a dynamic disorder in which clonal evolution triggers disease onset and progression (3). Clonal evolution is a multistep process that involves the successive acquisition of abnormalities in the genome of normal HSCs (12). These aberrations in turn lead to the expansion of MDS clones and subsequent transformation into AML (13). A vast body of evidence has been accumulated on the spectrum of cytogenetic abnormalities, epigenetic modifications, gene expression and signaling pathways, including apoptosis, proliferation, immune response, RNA-splicing machinery, microenvironment interactions, genome instability and DNA damage response, associated with the disease (14,15). Over the last decade, a number of breakthrough observations have been reported describing the presence of multiple somatic mutations in MDS (16–18). These mutations have been characterized in a number of key components of the spliceosome, regulator of DNA methylation, chromatin modification, transcription, signal transduction and cell cycle control machinery (19). However, the precise mechanism of MDS development remain a subject of intense research at present.

Transposable elements (TEs)

TEs are DNA sequences with the ability to move (transpose) into new locations in the genome, which frequently results in the creation of new copies of themselves (20). Since TEs were discovered in the maize crop in 1950 (21), they have become gradually recognized to be major components of eukaryotic genome. It is now estimated that they constitute ~45% of the human genome in a variety of forms and structures (5).

According to their mechanism of transposition, two major classes of TEs have been designated: Class I retrotransposons and Class II DNA transposons (22). Furthermore, each class can be divided further into several subclasses and families (22). DNA transposons move by a non-replicative mechanism from one genomic location to another (23). They encode a transposase enzyme that cleaves the TE from the genomic sequence and then mediate its reintegration into another loci in the genome in a so-called ‘cut-and-paste’ mechanism (23). DNA transposons can be divided into two subclasses based on the number of DNA strands that they can cut during transposition (24). DNA transposons also tend to be less common, where they are estimated to constitute ≥3% of the human genome (24).

By contrast, retrotransposons replicate using an RNA intermediate that is reverse transcribed into cDNA and reintegrated as an additional copy elsewhere in the genome in a ‘copy-and-paste’ retro-transposition mechanism (25). Retrotransposons are divided based on the presence or absence of long terminal repeats (LTRs) at their ends, which flank a central coding region (25). Endogenous retroviruses (ERVs) are a type of LTR retrotransposons that share similarities with exogenous retroviral proviruses and have integrated into host DNA following past infections (26). The majority of ERVs encode two proteins closely related to Gag (the original structural matrix and capsid) and Pol (original reverse transcriptase and integrase) proteins of retroviruses, but do not encode the Env (envelope) protein and therefore have lost their ability to exit the cell (27). However, ERV sequences generally only consist of solitary LTRs that are most likely generated by homologous recombination between the 5′ and 3′ LTRs (27).

Non-LTR retrotransposons do not have LTR sequences and resemble integrated RNA (28). They can be classified as autonomous or nonautonomous elements (29). Autonomous elements are also known as long interspersed elements (LINEs) and encode two proteins, Orf1 and Orf2 (30). Orf1 and Orf2 have endonuclease and reverse transcriptase activities, respectively and are able to self-mobilize (31,32). LINEs constitute the vast majority of transposable elements in the human genome (33). The LINE-1 element is one example of LINE, of which there are ~500,000 copies in the human genome, where ~99.9% of these copies are fixed and are no longer mobile (34). However, each individual is estimated to carry a set of ~100 potentially mobile LINE-1 elements (34). Non-autonomous elements, such as short interspersed nuclear elements (SINEs), depend on LINE-encoded proteins for their own cycle of retrotranscription (35). Human Arthrobacter luteus (alu) elements are ~300 bp-long SINEs that are among the most abundant SINEs observed (36). There are >1×106 copies of Alu in the human genome (37). In addition, other important members of the SINE family include mammalian-wide interspersed repeats (MIRs), which are ~260 bp in length and account for ~3% of the human genome (36).

Although they are generally assigned to their specific sub-families, the TE taxonomy and nomenclature is in a constant state of flux due to the discovery of new TEs and the necessity to introduce novel classification criteria (38). Diverse families of TEs take up a substantial proportion of the genome, where their propagation is regulated not only by their intrinsic properties but also by natural selection and genetic drift forces (39). Insertions that are potently deleterious are rapidly removed from the population, whereas insertions that exert little to no effects on the host become fixed in the genome (39). Therefore, it is not surprising that TEs are rarely randomly distributed in the genome but rather exhibit various types of insertion preferences, such that some TEs are more likely to be retained at certain genomic locations compared with others (40). For example, although de novo LINE-1 retrotransposon insertions readily occur within gene exons, these elements have rarely been observed to be fixed within the coding regions of humans (41).

The means by which TEs can lead to changes in DNA sequences are heterogeneous (20). Transposition is an imperfect mechanism that consists of repeated cycles of TE insertion and removal from the genome (42). Therefore, this process can also result in changes in the surrounding host sequences, including rearrangement of target genes or regulatory elements (39,43). Integration of TEs into exons can create frameshift mutations, leading to premature stop codons and nonsense-mediated decay or by inducing exon skipping (44,45). Furthermore, TEs, even those that have lost their mobilization capacity, can provide a large repertoire of homologous sequences scattered throughout the genome (39). This may promote nonallelic homologous recombination at distant genomic regions, resulting in large-scale deletions, duplications and/or inversions (39).

A number of TEs have been domesticated in the host genome (46). Following their integration, the vast majority of TEs become inactivated as a result of accumulating mutations that prevent further autonomous mobilization (47). However, in some cases these accumulating mutations can cause the neo-functionalization of these inserted TEs, such that they can even confer beneficial cellular effects due to the novel downstream gene product (48). Although the majority of domesticated TEs remain functionally uncharacterized, those that have been studied have been previously found to regulate a variety of cellular processes, including transcriptional regulation, proliferation, cell cycle progression and apoptosis (47).

TE silencing by P-element-induced WImpy testis-interacting RNAs (piRNAs)

The activity of TEs poses a serious threat to genome stability. Therefore, organisms have evolved a complex arsenal of mechanisms to control these potentially harmful mobilizations of TEs (49). One of the key TE-silencing mechanisms involves a large family of Krüppel-associated box (Krab) zinc-finger proteins, which bind to TEs and recruits Krab-associated protein 1 (Kap1; also known as Trim28) to form repressive chromatin complexes in association with multiple interacting partners, including the histone H3K9me3 methyltransferase, Set Domain Bifurcated 1 (SETDB1), the histone deacetylase containing NuRD complex and heterochromatin protein 1 (HP1) (50,51). Another important mechanism of TE regulation is mediated by epigenetic regulation, particularly through DNA methylation and histone modifications, which serve to maintain the viability of repressive heterochromatin (52) and small RNA molecules, called piRNAs.

piRNAs form one of the three main classes of regulatory small non-coding RNAs (sncRNAs), together with small interfering RNAs (siRNAs) and microRNAs (miRNAs) (53). These classes differ in terms of their biogenesis and mode of target regulation, but share a number of common features, including their ability to guide Argonaute (Ago) proteins to target nucleic acids in a sequence-dependent manner (54). piRNAs are single-stranded sncRNAs that are 26-31 nucleotides in length with 2′-O-methyl modification sites at the 3′ terminus (55,56). They comprise the largest class of sncRNAs expressed in animal cells (57). According to piRBase, the number of piRNA sequences reported has reached 173 million at present, with ~8.5 million unique human piRNA sequences (58).

piRNAs are expressed from repetitive intergenic elements in the genome called piRNA clusters using a Dicer-independent mechanism (Fig. 1). These clusters span wide regions of the genome and are mostly comprised of various TEs and their remnants (59). Primary piRNAs are processed from the piRNA clusters through the primary processing pathway, which then form functional complexes exclusively with Piwi proteins (60). In the germline of several organisms, primary piRNAs are subjected to an amplification system called the ‘ping-pong’ cycle to reinforce their expression (61). Within this mechanism, primary piRNAs bind to Ago3 or Aubergine proteins, recognize complementary targets to induce their cleavage, producing secondary piRNAs (61). This ping-pong mechanism has been identified in Drosophila melanogaster, zebrafish and primitive animal-like sponges, such as those in the phylum Porifera (62,63), but not in mice, suggesting that the ping-pong mechanism only exists during the early stages of evolution (64).

Figure 1. Mechanism of the piRNA pathway. In the nucleus, the piRNA precursor is transcribed from a piRNA cluster by Pol-II. After its export to the cytoplasm, the precursor is cleaved by the Zuc ribonuclease, generating a 5′-phosphate residue. The 3′ fragment of the transcript is incorporated into Piwi (gray ovals), forming the piRISC complex and trimmed to its final length. The 2′-hydroxyl group at the 3′ end is then methylated by the Hen1 2′-O-methyltransferase, rendering piRNA more stable. This piRISC complex can migrate back into the nucleus, where it can mediate co-transcriptional TE silencing. piRNA recruits DNMT or HDAC to its complementary, nascent TE transcript to repress its expression by making DNA/histone modifications. In the cytoplasm, piRISC can mediate the post-transcriptional gene regulation of complementary TEs or mRNAs. piRNAs typically target the complementary RNA sequence in the 3′-untranslated region, resulting in RNA degradation, translational inhibition or intracellular localization changes. piRNA molecules can also enter the ‘ping-pong’ cycle, where the piRNA-Aub complex recognizes and cleaves the complementary TE transcript or a transcript derived from the opposite strand of the same piRNA cluster. This cleavage produces a new piRNA that is loaded into the Ago3 protein and in turn induce the cleavage of its complementary RNA, generating a new copy of piRNA that is identical in sequence to the piRNA that initiated the cycle. PIWI, P-element-induced WImpy testis; piRNA, PIWI-interacting RNA; RISC, RNA-induced silencing complex; Pol-II, DNA polymerase II; TE, transposable element; DMNT, DNA methyltransferase; HDAC, histone deacetylase; Zuc, Zucchini ribonuclease; Aub, aubergine; Ago, argonaute.

Piwi proteins are a class of Argonaute proteins that are expressed in several stem cell populations across various organisms and tissues (62). However, they are most robustly expressed in the male germline of adult mammals (62). These proteins form specific RNA-induced silencing complexes (RISCs) with piRNAs called piRISCs (65). There are four known human Piwi proteins to date: Hiwi (also known as Piwil1), Hili (Piwil2), Hiwi2 (Piwil4) and Hiwi3 (Piwil3) (66). Piwi proteins are predominantly localized to the nucleus, where they colocalize with known epigenetic modifiers, such as polycomb group proteins (67). They enhance DNA methylation, thereby operating at the chromatin level and can modify the histone code to repress the transcription of TEs (68).

After its formation, the piRISC enters the nucleus, where it targets an actively transcribed nascent TE by recognizing its sequence, which is complementary to the piRNA molecule (69). The piRISC then prevents further transcription of the TE by recruiting histone deacetylases and DNA methyltransferases, leading to the enrichment of repressive signals, such as H3K9me3, and formation of the transcriptionally-silent heterochromatin (70). In addition to chromatin silencing, piRNAs can control TE and mRNA expression in the cytoplasm by mediating their degradation, translational inhibition and regulation of intracellular localization (62). In addition, >25% human mRNAs possess a retrotransposon in their 3′ untranslated region, rendering them potential targets for piRISCs (71).

Methods for the exploration of TEs and piRNAs

Recent progress in genomic sequencing, coupled with the rapid generation of large sequencing datasets and continuously improving bioinformatic technologies, has sparked interest in the previously unexplored, non-coding parts of the human genome. This has enabled the association between their structure, transcription and functional relevance for human health and disease to be accurately studied. Although standard next-generation sequencing (NGS) methods are frequently applied for the identification of TEs and piRNAs, their successful detection in genome-wide data requires specific tools, for which experimental design brings a number of novel challenges. Since the genome harbors multiple copies of similar TEs and piRNA sequences, degree of uniqueness and/or repetitiveness are important parameters that need to be considered, especially when aligning NGS reads that could be originated from multiple genomic locations (72). In this context, novel technologies based on long-read sequencing are providing new opportunities for reducing the complexity of TE detection (73).

Goerner-Potvin et al (74) previously published a comprehensive review of the available bioinformatics resources, databases and classification tools developed to detect and analyze TEs. Therefore, the present study only briefly describes the current bioinformatics approaches applied to analyze TEs. Discovery and annotation of TEs in NGS reads can be performed with or without a genome assembly, where various tools exist to perform such tasks. Currently, the ‘gold-standard’ strategy is by using a repository-based annotation, whereby sequences are queried against known TE consensus sequences or TE motifs (74). RepeatMasker (75) is the most widely accepted database query tool that is used to identify, classify and mask repetitive elements, including low-complexity sequences and interspersed repeats. RepeatMasker searches for repetitive sequences by aligning the input sequences against TE repositories (75). The repositories can be classified into the following three specific types: TE-centric (consensus sequences associated with each TE family); genome-centric (all individual TE instances within a reference genome); and polymorphism-centric (polymorphic insertions in individuals diverging from the annotated reference genome) (Table I) (74). Furthermore, RepBase (76) is a popular source for probing consensus sequences of eukaryotic TEs. Dfam (77) is another TE database where TE families are defined by multiple sequence alignments through hidden Markov models. Dfam is the only TE repository for individual TE instances annotated in mammalian genomes, where the TE sequences and their genomic locations are also available through the RepeatMasker tracks in the genome browsers (77). An alternative method of TE discovery and annotation that does not require a genome assembly is de novo annotation, which offers the potential to identify novel TE families (74). The most commonly used de novo annotation tools include RepeatModeler (78) or RepeatExplorer (79). Focusing on TE transcription, TEtranscripts (80) is a differential expression analysis software tool that has been shown to be the most accurate for analyzing TE expression in RNA-sequencing data.

Table I. Public resources specialized in human TE and piRNA data.

Table I.

Public resources specialized in human TE and piRNA data.

A, TE databases
Database	Description	First author/s, year	Web page	(Refs.)
RepBase	This largest collection of eukaryotic transposons and repetitive sequences includes >44,000 sequences (mostly family consensus). It functions as a standard reference for annotating repetitive DNA in genomic data.	Bao et al, 2015	www.girinst.org/repbase	(76)
Dfam	Database of repetitive DNA elements organized around multiple sequence alignments of TE families. This seed alignment provides a model-neutral representation of a sequence family from which both a traditional consensus sequence and a hidden Markov model profile can be built.	Wheeler et al, 2013	www.dfam.org	(77)
TranspoGene	Collection of TEs that are located in the protein-coding genes of seven eukaryotic organisms.	Levy et al, 2008	www.transpogene.tau.ac.il	(153)
Line Fusion	Database of human LINEs with information on LINE structure, expression pattern, tissue distribution and chromosomal location of the host genes, in addition to the domain structure altered by LINE integration.	Kim et al, 2006	www.primate.or.kr/line	(154)
RCPedia	Database of retrocopies (processed pseudogenes and retrogenes) from six primate genomes.	Navarro and Galante, 2013	www.bioinfo.mochsl.org.br/rcpedia	(155)
dbVar	NCBI database of human genomic structural variations, including mobile element variations.	Lappalainen et al, 2013	www.ncbi.nlm.nih.gov/dbvar	(156)
dbRip	Database of human retrotransposon insertion polymorphisms that contains all currently known Alu, L1 and short interspersed nuclear elements-variable number tandem repeat-Alu polymorphic insertion loci in the human genome.	Wang et al, 2006	www.dbrip.brocku.ca	(157)
euL1db	A curated database of human-specific L1 insertion polymorphisms identified in healthy or pathological human samples. It facilitates the understanding of the link between L1 retrotransposon insertion polymorphisms and phenotype or disease.	Mir et al, 2015	www.eul1db.ircan.org	(158)
B, piRNA databases
Database	Description	First author/s, year	Web page	(Refs.)
piRNAdb	Storage and search system that includes information on the alignments, clusters, datasets and targets of piRNAs, providing information on >27,000 piRNAs from six model	Piuco and Galante, 2021	www.pirnadb.org	(159)
piRBase	Organisms.	Wang et al, 2019	www.regulatoryrna.org/database/piRNA	(58)
piRNABank	Information on piRNAs in human, mouse, rat and Drosophila that compiles piRNA clusters and depicts piRNAs along with the associated genomic elements.	Lakshmi and Agrawal, 2008	www.pirnabank.ibab.ac.in	(160)
piRNAQuest	Annotation of human, mouse and rat piRNAs based on their genomic location. Information on all possible piRNA clusters along with significant motifs, piRNA expression and several analysis tools (Homology search, Dynamic piRNA Clusters, Pattern Search and AT-GC% calculator). The results can be browsed using piBROWSE and piSynBrowse.	Sarkar et al, 2014	www.bicresources.jcbose.ac.in/zhumur/pirnaquest	(161)
piRNA cluster database	Data on piRNA clusters in multiple species, tissues and developmental stages based on small RNA sequence data deposited in the NCBI Sequence Read Archive.	Rosenkranz, 2016	www.smallrnagroup-mainz.de/piRNAclusterDB.html	(162)

[i] PIWI, P-element-induced WImpy testis; piRNA, PIWI-interacting RNA; TE, transposable element; LINEs, long interspersed elements; Alu, Arthrobacter luteus; L1, putative monooxygenase locus 2.

Detection of piRNAs requires additional specialized methodology that is different from optimized techniques used for the analysis of miRNAs (81). To profile piRNAs on a genome-wide level, an NGS variant modified for the detection of small RNAs, such as small RNA-seq, is utilized (82). However, 2′-O-methylation of the 3′-terminal nucleotide within piRNA molecules may hinder adaptor ligation during library preparation, causing the underrepresentation of piRNA species in the data output (83). This was previously confirmed by Leshkowitz et al (84), who modified the 3′ end of synthetic miRNAs with an O-methyl group and compared the results of the analysis with their unmodified counterparts, which revealed that this modification resulted in a reduction in sequence detection (84). A number of solutions have been proposed to avoid this bias and improve library preparation, where several ‘low bias’ kits have been designed (85). These include the use of randomized adapters (86) and polyethylene glycol for ligation reactions (87) or avoiding the ligation step altogether (85). After library sequencing, small RNAs that are 24-32 nucleotides long were extracted from the raw reads and computationally processed (88,89). Bioinformatic analysis of piRNAs within the sequencing data can also be challenging, since a large number of piRNAs may be rare or expressed at low levels, such that single counts can result in technical artifacts (90). Therefore, the presence of small RNAs should be checked in public databases (Table I), where a huge number of piRNA sequences discovered by previous large-scale sequencing studies have been deposited.

As new TE and piRNA molecules accumulate, there remains to be insufficient understanding in their activities, interactome with other molecules or functions in cellular processes. Therefore, physiological studies of these newly discovered molecules are essential. In particular, the expression of individual TE/piRNA transcripts should be validated by reverse transcription-quantitative PCR (RT-qPCR), in situ hybridization or Northern blotting. Western blotting and immunofluorescence detection of the TE-derived proteins, such as Orf2 protein from LINE-1, can help to validate the expression of the TE mobilization machinery itself. Similarly, purification or direct visualization (such as by using electron microscopy) of replication complexes, such as ribonucleoprotein particles or virus-like particles, can also facilitate the detection of the assembled replication intermediates. This can consequently unravel information on the functionality of TEs (91). Furthermore, the ability of TEs to induce double-stranded breaks can be monitored using the immunofluorescence imaging of the γ-H2AX foci (92) or by using single-cell gel electrophoresis analysis (93). In addition, co-immunoprecipitation of the piRNAs with Piwi proteins is an important method for the verification of any small RNA detected as a true piRNA. Since piRNAs are defined as Piwi-binding RNAs (55), the Piwi complex should be isolated using a Piwi-specific antibody by immunoprecipitation. Subsequently, RNAs associated with Piwi protein should be extracted and sequenced or quantified by RT-qPCR to validate their involvement in the piRNA/Piwi pathway (94). The mechanism underlying the function of individual piRNAs or TEs can be evaluated by modulating their expression levels, by either overexpression using their corresponding mimics or suppression of their transcription using RNA interference or CRISPR-Cas9 approaches, before assessing changes in cell physiology. However, it is also important to note that the repetitive nature of TE sequences must be considered during experimental design and analysis. In particular, short DNA oligonucleotide sequences for PCR, short-hairpin RNAs or CRISPR-Cas9 need to be carefully designed and validated to ensure their specificity for a unique target.

In addition to wet laboratory methods, identification of TE/piRNA feature, functions and exploration of underlying molecular mechanisms have become important areas of research in computational biology (95,96). A number of bioinformatics methods have been proposed for the deeper characterization of these molecules, where a number of analytical software programs are readily available to the scientific community in the form of the databases aforementioned (Table I). Furthermore, integrative data analysis protocols that can simultaneously process multiple heterogeneous datasets facilitate the discovery of complex relationships between different types of piRNAs and TEs by assembling various piRNA-TE-mRNA coexpression networks (97). Based on the knowledge obtained from target prediction algorithms and information on protein-protein interactions within these networks, computational analysis can be used to predict novel TE/piRNA regulatory functions (98). This approach has already been applied, for example, by Liu et al (99). Based on network construction, they predicted cancer-associated piRNA-mRNA and piRNA-lncRNA interactions as well as piRNA regulatory functions in breast, head and neck, kidney and lung cancer (99). However, functional analyzes can also be enabled by using different freely accessible tools. For example, the piPipes (88) software includes various independent tools to analyze different types of datasets, including small RNA sequencing, RNA sequencing, degradome- and CAGE-sequencing, chromatin immunoprecipitation- sequencing and genomic DNA sequencing, to provide information on TE expression, TE insertions, structural variation and piRNA-induced cleavage product data for comprehensively studying piRNA-TE interactions. Similarly, TEtools (100) maps unassembled reads of RNA sequencing data obtained from previous classical mRNA and small RNA analyzes, which has been used by Lerat et al (100) to show the association in expression between TEs and piRNA precursor genes. In conclusion, computational biology has become an integral part of large-scale, big-data studies due to the enhanced supercomputing and data storage capabilities, coupled with the continuously optimized algorithms.

TEs and piRNAs in somatic and cancer cells

Accumulating evidence has documented the important roles of TEs and piRNAs in human carcinogenesis (6,66,101,102). Mobile TEs are considered to be highly mutagenic, where their transposition can cause mutational events that are associated with cancer development and progression (102). Transposition can induce the rearrangement of host genes or their regulatory elements by creating frameshifts and premature stop codons or by exon skipping. For example, TE insertions into DNA repair-associated genes, such as breast cancer type 1/2 susceptibility protein (103) or retinoblastoma 1 (104), can cause genomic instability and activate a number of oncogenic pathways. Increased transcription of TEs has also been reported in various of types of cancers, including lymphoma (105), colorectal cancer (106) and bladder cancer (107). Therefore, it is generally considered to be detrimental. Oncogenic processes associated with TE reactivation include the expression of TE-derived endonucleases, which can induce DNA breaks and genomic instability (108), whereas the activation of TE promoters that can also lead to oncogene activation (109). For example, reactivation of ancient LTRs has been associated with the aberrant oncogenic transcription of the Erb-B2 receptor tyrosine kinase 4 gene in anaplastic large-cell lymphoma (110). In addition, previous studies have revealed that the accumulation of RNA transcripts and extrachromosomal DNA copies derived from TEs may trigger an innate immune response leading to autoimmune diseases and inflammation (111–113). Chiappinelli et al (111) showed that the presence of TE copies in the cytoplasm can activate cellular viral defense responses and activate interferon signaling in cancer cells. This activation could induce lymphocyte recruitent, reversing immune tolerance in tumors (111). An important consequence of this process is that the increased transcription of TEs can mediate not only deleterious but also beneficial effects, because this increased transcription can also lead to the activation of the viral recognition pathway and immune system-mediated cancer cell death (111,112).

Although piRNAs were originally described in germline cells as suppressors of TEs that function to preserve genome integrity during development, the roles of piRNAs in somatic and cancer cells have also been previously documented (6,114,115). Somatic functions of the piRNA pathway are associated not only with TE silencing but also with genome rearrangement and epigenetic programming, with biological roles in stem cell functions, differentiation and malignant transformation. Despite the high number of piRNAs encoded in the human genome, Martinez et al (116) demonstrated that only a small fraction of piRNAs are consistently expressed in both somatic non-malignant and tumor tissues derived from 10 different anatomical sites: Bladder, breast, colon, head and neck, kidney, lung, prostate, stomach, thyroid and uterine corpus available from The Cancer Genome Atlas (TCGA) consortium. However, the expression patterns of these piRNAs differed sufficiently in that they can be applied to identify their corresponding tissue of origin and differentiate tumors from non-malignant tissues in a cancer-type specific manner (116). A number of studies have shown that differential expression of piRNAs can be detected in various human cancers, including multiple myeloma (MM), breast, lung and gastric cancers (6,90,117,118). At present, the role of piRNAs in cancer has been extensively reviewed elsewhere (6,90,117,118). However, piRNAs have been defined to be either oncogenes or tumor suppressors. For example, piRNA-823 is among the most extensively studied of piRNA molecules in the context of human cancer (119). Reduced piRNA-823 expression was observed in gastric cancer (118) or renal cell carcinoma (120). Overexpression of this piRNA using piRNA mimics in gastric cancer cell lines was shown to inhibit cell proliferation, the results of which were supported by similar inhibitory effects of tumor growth in a xenograft nude mouse model (118). However, piR-823 expression was found to be increased in patients with breast cancer (120) and MM (121,122), where it contributes to tumorigenesis. This suggests that the role of piR-823 is rather disease-specific (119). The expression profile of particular piRNAs can be applied to delineate clinical features, including histological subgroups, disease stages or survival, which was also shown by Martinez et al (116) in 12 different tumor types. The findings that piRNA expression can associate with poorer clinical outcomes, particularly in breast cancer, suggest that these molecules can be utilized as a novel class of cancer biomarkers (116).

Similar to piRNAs, a number of studies have also documented the aberrant expression of Piwi proteins in a variety of somatic cancers. Hiwi was demonstrated to be overexpressed in testicular tumors (121) and gastric cancer (122,123), whereas Hili expression was found to be increased in breast cancer, colon cancer, gastrointestinal stromal tumors, renal cell carcinoma and endometrial carcinoma (124). In particular, as diagnostic procedures transition from biopsies towards less invasive methods, extracellular piRNAs may serve to be an attractive alternative biomarker of cancer. The majority of studies in this field have been mainly focused on the application of measuring miRNAs in the blood circulation. Similar to miRNAs, piRNAs remain largely stable in the blood plasma and have the ability to survive a wide range of incubation and storage conditions regularly used in the laboratory (125). A previous study of piRNAs in gastric cancer found that, compared with existing miRNA-based biomarkers (miR-106a and miR-17), piRNAs showed higher sensitivity and specificity, particularly piR-823 and piR-651 (126).

Although the number of studies on dysregulated expression of TEs and piRNAs in cancer is rapidly accumulating, conclusions remain preliminary and specific functions of these molecules in cancer physiology remain poorly understood. In addition, the majority of the reported findings is only correlative, whereby altered expression does not necessarily mean that they have a causative role in oncogenesis. Instead, it may merely be a consequence of other processes associated with oncogenesis.

TEs in leukemia and MDS

The role of TEs in hematopoiesis is slowly becoming unraveled (113,127). Although evidence in support of a role for TE in leukemia is gradually increasing, its functional consequence in MDS remains scarce and existing studies generally focused on AML rather than MDS.

The expression of TEs in leukemia has been recently measured by Colombo et al (113,127), who investigated the mobilization of TEs in AML. They first measured TE expression in pre-leukemic HSCs, leukemic stem cells (LSCs) and leukemic blasts from seven patients with AML in freely-available sequencing data produced by Corces et al (128). From RNA sequencing data obtained from Wang et al (129), significant downregulation of a number of TEs, including the SINE Alu and LTRs ERV1, ERVL and ERVK) retrotransposons, was reported in LSCs compared with that in preleukemic HSCs and blasts (130). Furthermore, they also showed that the expression of TEs, especially that of LTR retrotransposons ERV3 and ERVL, is significantly reduced in high-risk MDS (n=6) compared with that in patients with low-risk MDS (n=6) (130). Since this suppression of TEs was associated with the significant reduction in the activities of several immune system, interferon and inflammation-related pathways, induction of TE expression in low-risk cases was proposed to be a potential mechanism for immune system-mediated clearance of cancer cells through the viral recognition pathway (130). By contrast, TE suppression in high-risk cases of AML may enable the AML cells to escape immune surveillance (113).

In a subsequent study, Colombo et al (127) explored the transcriptome data of 178 patients with AML from The Cancer Genome Atlas (TCGA) with respect to the mutation-specific dysregulation of TEs and correlated the TE expression profile to the respective transcription networks. However, since the sequencing libraries in TCGA were prepared using the polyA selection method, a number of TEs, which are mainly comprised of SINEs without polyA tails, remained undetectable from the dataset (127). The highest numbers of altered TE transcripts were associated with promyelocytic leukemia/retinoic acid receptor-α (PML-RARα) translocation and mutations in mitochondrially encoded cytochrome c oxidase II (MTCO2), nucleophosmin 1 (NMP1), structural maintenance of chromosomes 1a and runt-related transcription factor 2, with only minimal overlap (127). PML-RARα, MTCO2 and NMP1 were generally associated with the upregulation of TEs, whereas FMS-like tyrosine kinase 3, isocitrate dehydrogenase 1, TP53 and neuroblastoma-RAS were predominantly or uniquely associated with the downregulation of TEs (127). Although CpG methylation has been shown to regulate TE expression (111,112,130), DNA methyltransferase (DNMT)3A and tet methylcytosine dioxygenase 2 mutations were associated with only minimal TE dysregulation (127). In addition, a TE expression signature was found to accurately predict AML prognosis independent of mutation-based and coding gene expression-based risk stratification (127). In total, 14 candidate prognostic TE transcripts were identified, namely TEs from Alu (AluJo and AluSq2), ERV1 (LTR24C, LTR53, MER101B, MER31-1 and LTR45B), ERVK (MER11A and LTR14A), ERVL (MER77), L1 (putative monooxygenase locus 2) and Tc/mariner (MER44C, Tigger5A and Tigger9b) subfamilies, all of which were associated with AML prognosis (127). These 14 TEs constituted a diverse group of transcripts, some of which were classified as low risk and some as high risk, likely due to their naturally diverse functions (127). The validity of their prognostic power was successfully demonstrated in two independently downloaded datasets consisting of cohorts of 284 pediatric patients with AML (131) and 19 adult patients with relapsed AML (132). From this analysis, an improved prognostic algorithm utilizing a comprehensive model that includes the mutational status, cytogenetic status, coding gene expression and expression profiles of the 14 TEs was proposed for the prognosis of AML (127).

Recently, Zeng et al (133) explored the presence of TEs in the open chromatin regions of genomes of patients with AML produced using assay for transposase-accessible chromatin with HTS data performed by Corces et al (128). A cluster of ~21,000 TEs, consisting primarily of MIRs and LINE-2, were found to be more abundant in AML cells compared with those in normal blood cells (133). Subsequent pathway enrichment analysis revealed that nearby genes, such as FBJ Murine Osteosarcoma Viral Oncogene Homolog B (FOSB), FBJ Murine Osteosarcoma Viral Oncogene Homolog (FOS) and V-Jun Avian Sarcoma Virus 17 Oncogene Homolog (JUN), in the open chromatin regions of these MIRs were involved in leukemogenesis (133). Using the AML cell line Kasumi-1, several MIRs (particularly MIR9 and MIR18) were validated to be functional enhancers (133). Although MIRs are potentially involved in myeloid leukemogenesis, the specific mechanism for MIR involvement in AML remains unknown. However, it was speculated that MIRs can function as independent enhancers through interaction with other important upstream regulators to alter gene expression in AML cells (133).

A recent study by Deniz et al (134) revealed ERVs to be potentially oncogenic enhancers and regulators in AML. They identified six ERV families (LTR2B, LTR2C, LTR5B, LTR5_Hs, LTR12C and LTR13A) with AML-associated enhancer chromatin signatures which are enriched in the binding of key regulators of hematopoiesis and AML pathogenesis (134). Furthermore, correlations were found between differential chromatin accessibility at 20 ERVs and the expression of nearby genes, some of which were linked to AML prognosis (134). CRISPR-based experiments additionally identified five ERVs that can function as enhancers in leukemia cells and 13 different elements, the silencing of which led to the downregulation of nearby genes (134). Similar to observations by Colombo et al (127), variations in ERV activity were observed among patients with AML, which appeared to be partly driven by their unique mutational profiles (134).

Because genomic regions that contain TEs are highly methylated and silenced by heterochromatin in somatic cells (135), hypomethylating agents can increase the expression of TEs in cancer cells. Since hypomethylating compounds are used for the treatment of AML and high-risk MDS (8,9), these compounds may at least partially operate via the demethylation of TE promoters, which may in turn induce ‘viral mimicry’ and interferon signaling to mediate leukemic cell clearance (111,112). In addition, there has even been speculation that the ‘viral mimicry’ pathways are likely to be activated endogenously without hypomethylation therapy in low-risk MDS to enable immune system-mediated control of low-risk AML (113). Despite this clinically proven efficacy of hypomethylating agents against MDS, only ~50% patients benefit from this type of therapy (9). The majority of patients eventually develop therapy resistance or disease relapse whilst still under treatment (10,11). Therefore, understanding the regulatory mechanism of TEs in MDS pathogenesis and during disease treatment would facilitate the discovery of novel biomarkers for response prediction and novel strategies for preventing relapse in patients treated with hypomethylating agents.

piRNAs in leukemia and MDS

Germ cells, stem cells and cancer cells share key biological characteristics, including the ability to rapidly proliferate and self-renew (136). Because the piRNA pathway maintains germline stem cells by preserving the self-renewal mechanism, the same pathway may also serve similar roles in the self-renewal of rapidly-dividing hematopoietic stem and leukemic cells. However, reports on the transcription and exact role of piRNAs in blood cells remain scarce.

The first study on the function of the piRNA pathway in hematopoiesis and leukemia was reported in 2001 by Sharma et al (137), who demonstrated that the Piwi protein is expressed in human CD34+ hematopoietic stem and progenitor cells but not in differentiated hematopoietic cell populations. In addition, it was shown that transient overexpression of Piwi in human leukemia cells resulted in a potent reduction in cell proliferation and activation of programmed cell death, suggesting that Piwi functions as an important negative developmental regulator, where they regulate cell ‘stemness’ (137). However, another previous study where all three mouse Piwi genes were knocked out revealed no detectable effects on hematopoiesis (138). This lead to the conclusion that Piwi expression in HSCs is not required for normal adult hematopoiesis or that Piwi function is redundant with other proteins, possibly those in the Ago subfamily (138).

Over the past decade, piRNA dysregulation has been repeatedly reported in various hematological malignancies, including MM (139–141) and classical Hodgkin lymphoma (cHL) (142). Yan et al (139) previously measured the expression of piRNA-823 in MM and showed that it was upregulated in bone marrow biopsies from patients with MM and four MM-derived cell lines compared to samples from healthy individuals, which was in turn positively associated with clinical stages. The silencing of piRNA-823 also dysregulated the expression of cell cycle regulators and apoptosis-related proteins, which was accompanied by the inhibition of cell tumorigenicity (139). In addition, suppression of piRNA-823 expression in MM cells resulted in the reduction of de novo expression of DNA methyltransferases DNMT3A and 3B at both mRNA and protein levels, which in turn led to a decrease in global DNA methylation and reactivation of the tumor suppressor p16 INK4A, which was previously inhibited by methylation (139). piRNA-823 accumulation was also found in MM-derived extracellular vesicles, which transported piRNA-823 to endothelial cells effectively and promoted their malignant transformation (140). Cordeiro et al (142) previously investigated the Piwi/piRNA pathway in cHL by measuring the expression of Piwi proteins and three selected piRNAs piR-651, piR-20365 and piR-20582. Piwi proteins were found to be expressed mainly in cells derived from patients with cHL and cell lines, suggesting that this pathway was active (142). At diagnosis, piR-651 expression was found to be reduced in the serum of patients with cHL, whilst after complete remission, piR-651 levels were increased to levels similar to those of healthy individuals (142). Furthermore, lower levels of piR-651 were found to be associated with the lack of complete response to first line treatment, shorter disease-free and overall survival (142).

To date, the function of the piRNA/Piwi axis in leukemia has only been partially explored, with the majority of the emphasis on Piwi proteins rather than piRNAs. Bamezai et al (143) showed that Piwil4 showed aberrantly high expression in >72% patients with AML. Depletion of Piwil4 in mixed-lineage leukemia (MLL) rearranged AML cell lines was found to impair cell proliferation and clonogenic growth whilst delaying the onset of leukemia in mice (143). This depletion was found to be associated with a global reduction in the levels of repressive H3K9me3 signatures and an increase in the levels activating H3K4me3 signatures (143). In addition, it was demonstrated that Piwil4 depletion altered the piRNA expression profile in AML cells with rearranged MLL (143). The role of Piwi has also been previously studied in chronic myelogenous leukemia (CML) (144). A lentiviral expression vector was utilized for the overexpression of Hiwi in a CML cell line, which was shown to inhibit CML cell proliferation and migration, suggesting that Hiwi is a negative regulator of leukemogenesis (144).

Unlike miRNAs, for which there is a large body of information regarding their role in MDS and has been comprehensively reviewed elsewhere (145,146), information on the role of other types of sncRNAs remain limited. In terms of piRNA expression in MDS, only preliminary data have been reported to date (147,148). Using sequencing of the small RNAome, Beck et al (147) demonstrated the enrichment of piRNAs in unsorted bone marrow cells from patients with low-risk MDS, refractory anemia (RA). Because this previous study was focusing on miRNAs, piRNA expression was reported to be increased in patients with the RA status and accounted for ~9% of total small RNA counts, compared with ~2% in those with high-risk MDS and 1% in healthy individuals, without studying which piRNA further (147). Additionally, transcription of Piwil1 and Piwil2 was significantly upregulated in patients with RA compared with that in healthy individuals and patients with high-risk MDS (147). It was concluded that piRNA enrichment may potentially protect DNA from mutations in low-risk MDS cells, a mechanism that is not observed in cells of high-risk MDS (147). In a more recent study (148), the extracellular small RNAome was explored by sequencing in the plasma samples of patients with MDS. Although miRNAs was focused upon, a number of piRNAs were also found to be dysregulated in MDS, including the upregulation of hsa_piR_019914 and hsa_piR_020450, and those associated with specific characteristics of the disease, namely the upregulation of hsa_piR_000805 and downregulation of hsa_piR_019420 in low-risk MDS (148). These early findings suggest that piRNAs are dysregulated in MDS and might be informative as MDS biomarkers. However, further studies of their transcription profile and roles in MDS pathogenesis are warranted.

Conclusions

TE mobilization and piRNA expression are generally increased in advanced malignancies and are considered deleterious. Analogous to other types of cancers, these pathways were anticipated to be endogenously activated in MDS and may contribute to genome instability, where they in turn induce harmful changes during disease progression and leukemic transformation. However, early studies by Colombo et al (113) and Beck et al (147) reported clear enrichment of TEs and piRNAs during early MDS and subsequent suppression as the disease progressed (Fig. 2). These findings led to an opposing hypothesis, which assumes that the induction of TEs is a potential mechanism for the immune system-mediated elimination of leukemic cells during early phases of the disease.

Figure 2. TEs and piRNAs in MDS. A somatic mutation in the genome of a healthy cell leads to the development of myelodysplasia. In low-risk MDS, activated transcription of TEs and piRNAs has been described by Colombo et al (113) and Beck et al (147), respectively. Increased TE transcription induces the viral defense pathway, which may facilitate the elimination of leukemic blasts and maintain the non-proliferative status of the disease. However, subsequent somatic mutations can mediate clonal evolution and the development of high-risk MDS. In this stage, TE/piRNA expression is suppressed, which leads to immune escape, leukemic blast proliferation and further progression of the disease. PIWI, P-element-induced WImpy testis; piRNA, PIWI-interacting RNA; TE, transposable element; MDS, myelodysplastic syndrome.

MDS is a largely heterogeneous disease that is mainly classified on the clinicopathological criteria (149). Therefore, aberrant transcription of novel classes of molecules, such as TEs and piRNAs, may be potentially exploited as novel molecular markers of disease progression or to individualize patient management. In addition, because hypomethylating therapy has been associated with the induction of TE expression in colorectal cancer cells (112), understanding the regulation of TEs during MDS treatment by this type of compound may assist in the definition of novel predictive markers for therapy response and prevention of relapse.

Beyond their possible role as biomarkers, a number of early studies have also demonstrated the potential benefits of using piRNA molecules for therapy (90). Since piRNA dysregulation can alter the biological features of cancer cells through various regulatory mechanisms, modulating their expression may reverse certain cancer phenotypes. Several therapeutic strategies can be designed, the most appealing of which involves the use of synthetic piRNAs capable of blocking the synthesis of cancer-related proteins by binding to mRNAs. Previously, artificial piRNAs introduced to male embryonic germ cells were reported to be sufficient to induce piRNA-dependent gene silencing (150). Although enthusiasm for sncRNA-based cancer therapy is on the rise, it is necessary to unravel the precise molecular mechanisms and specific downstream targets of these agents. The current state of knowledge, evolving principles and challenges associated with the delivery and mitigation of the off-target effects of sncRNA-based therapeutic approaches have been previously summarized elsewhere (151,152).

Whilst investigations of TEs and piRNAs in leukemia remain in its infancy, further progress in the field of disease classification and monitoring, in addition to the area of therapeutic intervention is expected in the future. This is due to the continuous improvements in the understanding of the mechanisms and functions underlying TE/piRNA-associated processes in normal and leukemic cells.

Acknowledgements

Not applicable.

Funding

The present study was supported by AZV CR (grant no. NU20-03-00412) and MH CZ-DRO (UHKT; grant no. 00023736).

Availability of data and materials

Not applicable.

Authors' contributions

MDM: conceptualization, writing of original draft, revision and funding acquisition. ZK: writing of original draft.

Ethics approval and consent to participate

Not applicable.

Patient consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

References