R Loops and Links to Human Disease

Abstract

Aberrant R-loop structures are increasingly being realized as an important contributor to human disease. R loops, which are mainly co-transcriptional, abundant RNA/DNA hybrids, form naturally and can indeed be beneficial for transcription regulation at certain loci. However, their unwanted persistence elsewhere or in particular situations can lead to DNA double-strand breaks, chromosome rearrangements, and hypermutation, which are all sources of genomic instability. Mutations in genes involved in R-loop resolution or mutations leading to R-loop formation at specific genes affect the normal physiology of the cell. We discuss here the examples of diseases for which a link with R loops has been described, as well as how disease-causing mutations might participate in the development and/or progression of diseases that include repeat-associated conditions, other neurological disorders, and cancers.

Introduction

R loops are nucleic acid structures in which a nascent G-rich transcript hybridizes with the DNA template strand, leaving the non-template DNA single stranded [1]. R loops were first described in prokaryotes more than 20 years ago [2]. For a long time, they were considered largely as a by-product of transcription that did not have significant consequences and did not generate much interest. However, during the past decade or so, an increasing number of studies have revealed very important functions of R loops in transcription that when disturbed can be linked to a variety of diseases. Genome-wide mapping studies of R loops in human, mouse, and yeast cells by DNA/RNA immunoprecipitation followed by high-throughput sequencing (DRIP-seq) have tremendously helped to understand how and where R loops form [3–7]. Recent high-resolution DRIP-seq experiments suggest that R loops are abundant structures and are estimated to occupy up to 5% of mammalian genomes and 8% of the budding yeast genome [6,7].

R loops occur naturally during transcription and can have important functions. For example, they are important for class switch recombination of immunoglobulin (Ig) genes in activated B cells [8] and for mitochondrial replication [9,10]. R loops are also found frequently at GC-rich regions such as typify many promoters and 3′end regions, where they appear to play significant roles in transcription [3,4,7,11,12]. However, in some cases, persistence of R loops can have deleterious effects. For example, they can result in the accumulation of DNA double-strand breaks (DSBs) [13], leading to DNA rearrangements and genome instability [1,14]. It is thus not surprising that R loops have been linked to disease, including some cancers and several neurodegenerative disorders.

Several reviews describing how R loops form and how they can be resolved or lead to genome instability have appeared in the last few years [1,15–18]. The role of R loops in disease has also been recently reviewed [16,17,19]. However, examples of R loops in pathological conditions have continued to bloom, and here, we highlight recent advances in understanding how R loops are connected to disease.

R Loops and Genome Stability

Before delving into the interconnections between R loops and disease, we briefly highlight the many connections between these structures and genome instability, as this is central to their role in many pathological conditions. The first strong links to genome instability were discovered over 10 years ago, in yeast and chicken cells. In yeast, deletion of hpr1, which encodes a component of the THO complex that couples transcription with mRNA export, was found to induce the inhibition of transcription elongation, DNA damage, and hyper-recombination through the formation of R loops [20]. The same connection was shown later with the mammalian THO complex [21]. In chicken DT40 cells, depletion of the splicing factor SRSF1 was shown to cause DSBs and DNA hypermutation, again due to the formation of R loops during transcription [22,23]. In fact, it is now known that many factors involved in RNA metabolism and processing play roles in preventing R-loop formation by packaging nascent RNAs into ribonucleoproteins (RNPs) that prevent its hybridization to the DNA template [20,23–26]. In the presence of R loops, the displaced ssDNA becomes sensitive to nucleases such as activation-induced cytidine deaminase, which leads to DSBs necessary to create hypermutations and diversity at the Ig locus in B cells [27]. XPF and XPG, two endonucleases of the nucleotide excision repair machinery, have also been found to process R loops into DSBs [13]. Such DSB-inducing mechanisms could participate in R-loop-dependent genome instability. Loss of topoisomerase I, which plays roles in replication and transcription, favors R-loop formation through increased levels of negative supercoiling. This can drive the stalling of both RNA polymerase (RNAP) and replication forks, leading to chromosome breaks [2,28,29]. Stalled replication forks are unstable structures prone to recombination and genome instability [30]. R loops have been found to cause replication impairments in various conditions and organisms, from bacteria to humans [31–34]. Collision between the transcription and replication machineries at common fragile sites in higher eukaryotes, the hot spot of chromosomal rearrangements, has been found responsible for R-loop formation and genomic instability [35]. However, it is unclear what comes first. Does the collision between RNAP and the replisome cause RNAP stalling and lead to R-loop formation? Or does R-loop formation cause RNAP stalling and then collision that arrests the replication fork? While evidence point to both possibilities, genomic instability appears to be the common outcome in the development of several diseases.

R-loop formation at trinucleotide repeat-associated diseases

More than 40 genetic disorders are caused by gene-specific repeat expansions. These include Huntington's disease [CAG repeats in huntingtin (HTT)], myotonic dystrophy type 1 [CTG repeats in dystrophia myotonica protein kinase (DMPK)], spinocerebellar ataxia type 1 [CAG repeats in ataxin 1 (ATXN1)], fragile X mental retardation or fragile X syndrome (FXS) [CGG repeats in fragile X mental retardation 1 (FMR1)], and Friedreich ataxia [GAA repeats in frataxin (FXN)] [36,37]. Those extended repeats can lead to various outcomes, such as transcription inhibition and expression of toxic RNAs or toxic polyglutamine proteins. A significant number of these have been linked to R loops (Table 1). For example, such repeats often have high GC content [36] and can potentially form R loops [38]. Indeed, in vitro transcription of CAG, CTG, GAA (not so high GC content), CGG, and CCG repeats can form R loops in Escherichia coli [39]. R loops formed in vitro and in bacteria at GAA repeats have been linked to RNA polymerase arrest [40]. Using a similar system, CTG repeats lead to the formation of R loops, promoting repeat instability in bacteria and human cells carrying transgenic repeats [41]. Repeat instability leading to expansion or deletion of the repeats has been shown in other disease models in bacteria [42], flies [43], and human cells [44]. Very often, the repeats are transcribed in both directions, which is believed to stimulate repeat instability [39,44–47]. This increased instability has been proposed to result from the formation of R loops on both strands, referred to as double R loops [39,48]. Interestingly, R-loop composition affects instability by leading to repeat expansion or deletion [48].

Table 1. List of diseases and associated genes for which a connection to R loops has been shown or is highly suspected.

Disease	R-loop-associated genes (repeats)n	References [In vivo detection of R loops at expanded repeats (#)]
DM1 (myotonic dystrophy type 1)		[39,41(#)]
HD (Huntington's disease)		[39]
FRAXE (fragile X syndrome type E)		[39]
FXS or FRAXA (fragile X syndrome type A)		[39,49(#),55(#),56(#)]
FXTAS (fragile X tremor ataxia syndrome)		[39,55(#)]
SCAs (spinocerebellar ataxias)		[39]
FRDA (Friedreich ataxia)		[39,40,49(#)]
ALS/FTD (frontotemporal dementia)		[48,63]
Breast and ovarian cancer		[75, 76, 77]
Eosinophilic leukemia		[25]
Seminoma		[79]
Colon cancer		[65]
FA (Fanconi anemia)		[87,88]
ALS (amyotrophic lateral sclerosis)		[134] (yeast)
AOA2 (ataxia with oculomotor apraxia)		[11,149,151,154]
AGS (Aicardi–Goutières syndrome)		[162]
PWS (Prade–Willi syndrome)		[64]
Burkitt's lymphoma		[82]
MM (Multiple myeloma)		[82]
AIDS-associated malignancies		[92]

In blue, genes where R loops form (*shown at endogenous locus) or may form (lack of in vivo evidence at endogenous locus). In gray, genes encoding proteins involved in R-loop prevention or removal.

R loops have been well documented in the FXN and FMR1 genes. Using transformed lymphoblastoid cell lines derived from FRDA and FXS patient cells, Groh et al. confirmed that highly stable R loops form at the expanded repeats of both genes [49]. FRDA is the most common inherited ataxia, caused by the GAA expansion in the first intron of FXN, leading to transcriptional silencing of FXN [50]. Importantly, R-loop enrichment has been found to be associated with the repressive H3K9me2 chromatin mark at FXN in Friedreich ataxia cells [49]. Interestingly, H3K9me2 is also found to be associated with R loops forming at gene 3′ ends, perhaps contributing to efficient transcription termination by promoting RNAPII pausing and release from the DNA template [11,12]. While a decrease in H3K9me2 does not affect R-loop or mRNA levels, increased R-loop formation, by the addition of the topoisomerase I inhibitor camptothecin, leads to increased H3K9me2 and transcription inhibition, suggesting that histone modification is a consequence of R-loop formation [49]. Expanded repeats at FXN appear to be another example of R-loop-dependent transcription elongation inhibition [20,51].

The scenario at FMR1 appears quite different. The FMR1 promoter and repeat region (in the 5′UTR) are hypermethylated in FXS patient cells containing more than 200 repeats (full mutation), resulting in the repression of FMR1 transcription [52,53]. The absence of the FMR1 protein is the leading cause of the intellectual disability of the disease. In contrast, premutation alleles, which contain between 55 and 200 repeats, trigger higher transcriptional activity, as seen in fragile X-associated tremor/ataxia syndrome (FXTAS) for which a toxic gain-of-function of the protein is suspected to cause disease [53,54]. However, R loops form at FMR1 in both FXS and FXTAS [49,55,56]. Indeed, after transcription reactivation by methylation inhibition, R loops accumulate at the FMR1 locus in FXS cells [49]. A study monitoring the differentiation of FXS embryonic stem cells into neurons [56,57] showed that the formation of an R loop leads to epigenetic silencing of the FMR1 promoter [56]. Altogether, these data indicate that R-loop formation may play a role in switching off FMR1 expression early during development in FXS patients through epigenetic silencing. Furthermore, R loops forming at premutation loci trigger the formation of a hairpin-like structure in the displaced ssDNA that might promote the transcriptional regulation seen in such alleles [55]. R loops are suspected to protect CpG island promoters from methylation and epigenetic silencing [3], and perhaps, a similar mechanism is used at premutation loci to stimulate transcription. However, it is difficult to understand how R-loop formation at the same gene can have two different outcomes. Is it possible that R-loop formation at shorter repeats triggers different epigenetic changes or different DNA structures, such as hairpins [55,58] or G-quadruplexes (G-Qs), which have the potential to stabilize R loops [59]? G-Qs are highly stable stacks of planar tetramers of four guanines connected by Hoogsteen hydrogen bonds [60]. Similar to R loops, they are enriched in 5′UTRs, first introns, and 3′UTRs [61,62]. Additionally, G-Q and hairpin structures have different protein binding specificity that might trigger distinct processes [63].

Trinucleotide repeat-associated diseases make a clear connection between R loops and epigenetic changes. R loops have been shown to associate with particular chromatin states, which can be open [7,64,65] or condensed [12,66]. It will be important to understand further how R-loop formation can trigger those differences not only in normal and disease cells but also in different cell types or physiological states.

R loops and cancer

Genome instability is a hallmark of almost all cancer cells [67–69]. Replication stress and genome instability at common fragile sites have been linked to cancer development [70]. As we discuss in the following paragraphs, inappropriate accumulation of R loops and the resultant genomic instability appear to play a role in a number of human cancers.

Mutations in the tumor suppressor genes BRCA1 and BRCA2 increase the risk of breast and ovarian cancers and some other cancers [71–73]. BRCA1 and BRCA2, which function in DSB repair by homologous recombination, have both been shown to prevent R-loop formation. Indeed, BRCA1 and BRCA2 knockdown (KD) leads to the formation of γH2AX foci, indicative of DNA damage, and this correlates with an increase in R-loop foci [74-76]. Later, Hatchi et al. showed that BRCA1 accumulates at certain RNAPII termination regions that appear to depend on R-loop formation, where it prevents the accumulation of R loops and resultant DNA damage [77]. Interestingly, mutational profiling of 21 breast cancers showed that BRCA1-deficient tumors have significantly more insertions-deletions near R-loop-dependent terminators than BRCA2-dependent or BRCA1/2-independent breast cancers [77]. This data indicate that while BRCA1 and BRCA2 both play a role in R-loop prevention, they might use different mechanisms and perhaps have different locus specificities.

Mutations in the mRNA polyadenylation machinery can also result in R-loop formation and DNA damage. For example, mutations affecting the polyadenylation factor FIP1 result in increasing DNA damage foci and R loops in yeast [25]. Interestingly, 10–20% of eosinophilic leukemias express a truncated version of FIP1, resulting from an oncogenic fusion between FIP1 and PDGFRα [78]. While truncated FIP1 induces R-loop-mediated DNA damage, FIP1 KD increases 53BP1 foci, an indicator of DNA damage [25]. Altogether, those data indicate that partial loss of function of FIP1 and a subsequent increase in R-loop formation could be responsible for genomic instability in eosinophilic leukemia.

The E3 ligase Bre1 (RNF20/40) monoubiquitinates histone H2B and plays a role in transcription elongation control and DNA damage repair. Bre1-depleted cells show increased γH2AX levels that are RNAse H1 sensitive, suggesting that the DSBs and genomic rearrangements associated with Bre1 KD are R-loop-dependent [79]. Importantly, Bre1, which is normally highly expressed in testis, is downregulated in a type of testicular germ cell cancer (called seminoma) known to accumulate high levels of genome instability. It is thus possible that R-loop formation plays a role in this cancer. Notably, the CpG island-rich RNF20 promoter is often hypermethylated in breast and prostate cancers [80]. RNF20 expression is often found down regulated in metastatic prostate cancer, and DNA methylation inhibition leads to an increase in RNF20 mRNA in prostate cancer cells [80].

CpG islands are characteristic of promoter regions of many human genes, especially the so-called “housekeeping” genes. These regions are subject to DNA methylation, but when hypomethylated, they have a tendency to form R loops [3]. This can be relevant to cancer, as the following two examples illustrate.

The vimentin (VIM) gene encodes an intermediate filament protein, and its CpG island promoter region is hypermethylated, and expression is silenced in several cancers, including colon cancer [81]. Interestingly, in normal cells, an antisense RNA hybridizes with the VIM promoter region to form R loops, which help maintain an open chromatin state and increase VIM transcription [65]. It will be important in the future to explore the relationship between the observed DNA hypermethylation and R-loop formation in VIM transcriptional regulation and its relevance to cancer.

Tudor domain-containing protein 3 (TDRD3) is a transcriptional co-activator that facilitates gene expression by recognizing specific methyl-histone marks at promoters. TDRD3 forms a complex with the DNA topoisomerase TOP3B, which is recruited to the c-MYC CpG island promoter, where it prevents R-loop formation, likely by relaxing negative super-coiling, and facilitates c-MYC transcription [82]. TDRD3-null mice show increased DSBs and R-loop formation at c-MYC and at the Ig heavy chain (Igh) locus, which correlates with increased c-MYC-Igh translocations [82]. This suggests that R loops may have a role in the oncogenic c-MYC-Igh translocation commonly seen in Burkitt's lymphoma and multiple myeloma [83,84]. Consistent with this, Igh and c-MYC transgenes have been found to be DSB and recombination hot spots in an hpr1Δ yeast strain that accumulates R loops [20,85].

Proteins from the Fanconi anemia (FA) pathway are important to ensure proper DNA replication. Loss of any FA genes leads to FA, a condition characterized by bone marrow failure and an increased risk of cancers [86]. Interestingly, R loops accumulate in KD and knockout (KO) cells for proteins from the FA pathway and in FA patients cells and result in an increase in replication fork stalling and DNA breaks [87,88]. Significantly, BRCA1 and BRCA2 are also part of the FA pathway. Mutations affecting another FA pathway protein that interacts with BRCA1, the FANCJ helicase, are linked to breast cancer and FA. Interestingly, stabilization of G-Qs that can normally be unwound by FANCJ leads to DNA damage in FANCJ-depleted cells [89].

Kaposi's sarcoma-associated herpesvirus is an oncogenic virus associated with several AIDS-associated malignancies. The virus has been shown to cause DSBs and chromosome instability in infected cells [90,91]. We noted above that the THO/TREX complex prevents R-loop accumulation in human cells [21]. Intriguingly, Kaposi's sarcoma-associated herpesvirus-infected cells express a protein, ORF57, implicated in viral mRNA processing, which binds and inactivates the hTREX complex, making it no longer available for cellular mRNA processing and leading to R-loop formation and genome instability [92]. Expression of the THO component THOC1 is decreased in testis and skin tumors [93], although the significance of this remains to be determined.

All the above examples establish a connection between R-loop accumulation and genomic instability in cancer. The actual steps that lead to instability are still largely unknown.

R-loop connections to neurological disorders

ALS/FTD

Amyotrophic lateral sclerosis (ALS), also known as Lou Gehrig's disease, is a very complex disease for which many genes have been implicated in pathogenesis [94]. ALS is the most common motor neuron disease, characterized by a loss of motor neurons in the brain and spinal cord that leads to a rapid loss of control muscle movement and death typically 3 to 5 years following diagnosis [95]. About 90% of ALS cases are sporadic, and less than 10% are familial [96]. While the exact molecular mecha-nism(s) leading to disease are unknown, and indeed multiple mechanisms undoubtedly exist, evidence supporting a possible contribution from R loops in certain cases emerged [97]. We discuss below several genes with ALS mutations in which R loops have been implicated.

C9ORF72

The most common known cause of both familial and sporadic ALS is a hexanucleotide expansion in the C9ORF72 (C9) gene. In fact, the expansion, GGGGCC (G4C2; located in the first intron of C9), is frequent not only in ALS but also in the related syndrome frontotemporal dementia (FTD) [98–100], which is characterized by the degeneration of the frontal and temporal lobes of the brain and is a common cause of early onset dementia [101]. While normal individuals have a number of G4C2 repeats below 30 [102], ALS patients can have up to several thousand repeats. As would be expected from the extreme G-richness of this sequence, R loops form at in vitro-transcribed repeat expansions, and this occurs concomitantly with RNAP II stalling and the accumulation of abortive RNAs containing G-Qs, coupled with a decrease in full-length mRNAs [48,63,103–105]. G-Qs also form on the displaced DNA strand and are believed to stabilize R loops [63].

How might R-loop formation affect C9 transcription? C9 patients exhibit repressive chromatin marks such as H3K9me3 or H3K26me3 on the C9 gene [106], although as noted above, the relationship between R loops and chromatin modification is poorly understood. Again, consistent with the sequence of the C9 hexanucleotide, DNA methylation has also been found to be associated with the repeat expansion [107–111]. Since bidirectional transcription also occurs at the expansion [112–116] and that R loops are enriched at genes producing antisense RNA [5,12], it is possible that R loops reduce C9 expression through the regulation of bidirectional transcription and chromatin modification (reviewed in Ref. [117]). While reduced C9 protein levels may contribute to disease [104,118,119], other mechanisms, such as the sequestration of one or more RNA binding protein (e.g., Ref. [63]) and/or the accumulation of toxic dipeptide repeat proteins generated by Repeat Associated Non-AUG (RAN) translation [112,120,121], may also be important. Thus, considerable work is required to understand how the C9 repeats cause ALS/FTD and how R-loop formation at the locus may be involved.

G4C2 repeats have not been found at other genes implicated in ALS or FTD [122]. While this could indicate that the formation of R loops and/or G-Qs is unique to the C9 expansion, many other G/C combination repeats can lead to these structures [39], and it is conceivable that they may be found associated with some of the ∼50% of ALS cases for which no mutation is known. Notably, the C9 expansion has also been linked to other neurological disorders, such as Alzheimer's [123–126] and Huntington's [127] diseases.

Ataxin-2

PolyQ expansions in Ataxin-2 (ATXN2) have been linked to ALS [128,129] and spinocerebellar ataxia type 2 (SCA2) [130–132]. SCA2 results from the expansion of CAG repeats in ATXN2, which normally contains 22 repeats. It has recently been shown that intermediate-length repeats (27–33 repeats) might play a role in the development of ALS in C9 patients and others [128,129,133]. The ortholog of ATXN2 in yeast, Pbp1, maintains ribosomal DNA stability through the inhibition of the deleterious formation of R loops at G-Q DNA sites in the intergenic spacer regions of the ribosomal DNA repeats [134]. Pbp1 also suppresses R-loop formation at telomeres [134]. It is possible that ATXN2 inhibits the formation of R loops at some loci, although ATXN2 is predominantly a cytoplasmic protein [128]. It will be interesting to determine whether SCA2 cells accumulate R loops. In fact, ATXN2 interacts with TAR DNA-binding protein 43 (TDP-43), and the longer the ATXN2 polyQs, the stronger the interaction [129]. TDP-43 is associated with ALS and FTD [135], and while normally nuclear, mutated TDP-43 aggregates in the cytoplasm and associates with stress granules in ALS neurons, a hallmark of ALS [135,136]. Interestingly, ATXN2 and TDP-43 both relocalize to cytoplasmic stress granules after stress induction [136,137]. It is then possible that intermediate-length polyQ expansions in ATXN2 help in sequestering TDP-43 into cytoplasmic aggregates and thus contribute to the disease. This scenario implies that TDP-43 could have a role in R-loop formation/resolution in the nucleus. Similarly, ATXN2 also co-localizes with Fused in Sarcoma (FUS), which is also mutated in ALS, in cytoplasmic inclusions [133]. Determining whether TDP-43 and FUS have a direct role in R-loop regulation might bring some light on the link between ATXN2 polyQ expansions and ALS.

SETX

Senataxin (SETX) is an RNA/DNA helicase that is mutated in two distinct neurological disorders, ataxia with oculomotor apraxia type 2 (AOA2) [138] and a juvenile form of ALS known as ALS4 [139]. Interestingly, AOA2 mutations are recessive, while ALS4 mutations all appear to be dominant. Mutations in SETX have also been linked to atypical forms of AOA2 or ALS4 that have variant phenotypes [140–145]. Sen1, the SETX homolog in yeast, plays a role in transcription termination of non-coding RNAs and some mRNAs [146,147] and, importantly, was shown to suppress R-loop formation and transcription-associated recombination [148]. SETX was also shown to resolve R loops that form normally at the transcription termination pause site of the human β-actin gene [11]. Recently, SETX association with R-loop-dependent 3′ pause sites was extended to several other genes and was shown to be BRCA1-dependent [77]. SETX and BRCA1 co-localize on meiotic chromosomes and have been found to interact genetically and physically [75,77,149]. SETX/BRCA1 recruitment at 3′ pause site prevents the accumulation of deleterious R loops that would otherwise lead to DNA damage [77]. Interestingly, SETX maps to deleted regions in ovarian cancer, which might suggest a tumor suppressor role for SETX [75]. While this seems possible, AOA2 patients do not show a higher susceptibility to cancer than the rest of the population.

How do mutations in SETX lead to disease? Both sen1 and SETX prevent genome instability by suppressing DSBs arising from R loops, reflecting the collisions between the transcription and replication machineries [150,151]. In response to such replication stress, SETX relocalizes to R-loop-containing nuclear foci [151,152], where it likely resolves the RNA:DNA moiety of the R loop. Unexpectedly, the nuclear exosome, a 3′ to 5′ exoribonuclease complex that functions in RNA quality control, was found to relocalize with SETX, mediated by direct interaction with an exosome subunit, Rrp45 [152]. Importantly, the SETX–Rrp45 interaction is dependent on SETX sumoylation, which is disrupted by AOA2, but not ALS4, mutations [152,153]. Thus, it seems that after SETX unwinds the R loops, the exosome is poised to degrade the released RNA, preventing rehybridization or other possible toxic effects. How and if this is critical to AOA2 specifically remain to be determined.

Perhaps not unexpectedly, SETX KO mice accumulate R loops. However, SETX KO does not lead to any neurological characteristics of AOA2, but it does cause male sterility [149,154]. It is still unclear at what level R-loop accumulation is related to disease. Interestingly, AOA2 cells have reduced telomere length, and since R loops and G-Qs form at telomeres, it is possible that SETX also plays a role in telomere stability [155–157]. Finally, most data show an accumulation of R loops after SETX KO or KD, which relates more to AOA2. It will then be of great interest in the future to explore the R-loop pattern in cells containing ALS4 mutations.

The 5′-3′ exonuclease XRN2 also participates in transcription termination by degrading nascent RNA downstream from the 3′ cleavage site [158,159]. Interestingly, XRN2 KD was recently found to lead to the accumulation of R loops and DSBs, which are in turn associated with increased replication stress and genomic instability [160]. Similar to SETX, XRN2 prevents DSB accumulation at 3′ pause sites of a subset of genes. Thus, while R loops are necessary at certain transcriptional terminators, their persistence appears deleterious to the cell.

Other possible R-loop-related neurological disorders

Aicardi–Goutières syndrome (AGS) is a severe childhood inflammatory disorder causing neurological damage [161]. Genome-wide analysis of AGS patient fibroblasts (characterized by mutations in TREX1, which encodes a 3′-5′ ssDNA exonuclease or RNASEH2, which encodes a second RNase H enzyme) revealed a global increase in R loops [162]. Interestingly, however, the overall pattern at the expected GC-skewed regions was similar between control and AGS patent cells [4]. Thus, the excess of R loops specific to AGS cells was not at the 5′ or 3′ end of genes but instead at the intergenic or gene body regions [162]. This suggests that R loops formed in AGS patient cells are probably of a different type than the more naturally forming R loops. AGS patient cells also show a global DNA hypomethylation pattern, which corresponds to the AGS-specific R-loop loci, particularly in cells carrying an RNASEH2 mutation [162]. These data strongly suggest that R-loop formation in AGS patient cells is linked to epigenetic changes that might trigger the immune response stimulation that is typical of the disease. Interestingly, mutations in TREX1 and RNASEH2 are also linked to systemic lupus erythematosus [163,164], which might indicate an R-loop contribution to another autoimmune disease.

Prader–Willi syndrome (PWS) leads to obesity in children, is associated with motor and mental retardation, [165] and involves genes subject to imprinting in the brain [166,167]. The locus includes the maternally expressed gene encoding the E3 ubiquitin-protein ligase Ube3a and a long paternally expressed transcript, which includes the antisense transcript to Ube3a, Ube3a-ATS, and the GC-skewed SNORD116 locus that encodes a cluster of non-coding RNAs, including small nucleolar RNAs (snoRNAs). Paternally inherited deletions of SNORD116 and the loss of snoRNAs are linked to PWS [168]. Normally, R loops form at the SNORD116 intronic region in neurons and have been linked to chromatin decondensation that allows the expression of Ube3a-ATS, which in turn silences Ube3a on the paternal allele [64]. However, treatment with Topotecan, a topoisomerase inhibitor, leads to excess levels of R loops over the SNORD116 locus that inhibit transcription through Ube3a-ATS, allowing the expression of Ube3a [64]. This example thus reveals a versatile and subtle form of transcriptional regulation mediated by R loops and is consistent with the possibility that the loss of R-loop formation at SNORD116 may contribute to PWS.

Conclusions

R loops have gone from being a “ghost structure” that did not receive much attention to a master player of transcription regulation, chromatin modification, and genomic stability. While certain cancers and a considerable number of neurological disorders have been linked in one way or another to R loops, it is still very difficult to determine whether R loops are the cause of the disease or a secondary effect due to other issues. Nevertheless, it seems that resolving R loops or, in some cases, inducing them might have a therapeutic value. However, considerably more research to obtain a clear understanding of how and under what circumstances R loops form and to distinguish a “good” R loop from a “bad” R loop is needed. How can some R loops be deleterious to the cell while some are necessary and protective? It also seems that many loci prone to R-loop formation might also be prone to DNA and RNA G-Qs. Many proteins have been suggested to bind G-Qs, and it will be interesting to determine whether any of these proteins also bind R loops. And then, what is the deleterious structure for the cell? The R loops or the G-Qs or both? Is it the actual change in chromatin accessibility, resulting from R loops and/or G-Qs, that is responsible for genomic instability? Could chromatin modification and compaction generated through R-loop formation be the obstacle leading to instability and to replication fork progression [34,169]?

Many intriguing properties of R loops remain to be investigated. While essentially all described R loops are co-transcriptional, and thus form in cis, R loops forming in trans, far from their site of transcription, have been reported in yeast [170]. The possibility that R loops can form in trans in the human genome would potentially extend both their regulatory functions and deleterious effects to a considerable yet unknown locus and might be linked to more diseases than the ones described here. The identification of R-loop-binding proteins will provide important knowledge on how they form, are maintained, and impact genome stability. Finally, many diseases are caused by gene-specific repeat expansions for which a possible role for R loops has not yet been investigated [36]. It is thus likely that more examples of R loops playing important and unexpected roles in disease remain to be uncovered.

Abbreviations

DSB

DNA double-strand break

RNAP

RNA polymerase

FXS

fragile X syndrome

FXTAS

fragile X-associated tremor/ataxia syndrome

ssDNA

single-stranded DNA

G-Q

G-quadruplex

KD

knockdown

TDRD3

tudor domain-containing protein 3

FA

Fanconi anemia

ALS

amyotrophic lateral sclerosis

FTD

frontotemporal dementia

ATXN2

Ataxin-2

SCA2

spinocerebellar ataxia type 2

TDP-43

TAR DNA-binding protein 43

SETX

senataxin

AOA2

ataxia with oculomotor apraxia type 2

AGS

Aicardi–Goutières syndrome

PWS

Prader–Willi syndrome