Atypical R-loops in cancer: decoding molecular chaos for therapeutic gain

Abstract

R-loop is a special DNA–RNA hybrid genomic structure. Since its identification, the molecular mechanisms of physiological processes such as class switch recombination have been revealed, uncontrolled regulation of which has become the underlying cause of many diseases. With the development of molecular biology techniques, R-loops found at different sites and formed via different mechanisms have been discovered. These non-classical R-loops participate in various cellular activities via different mechanisms. A set of sophisticated regulatory mechanisms are required to control the number of various R-loops in vivo. The levels of non-classical R-loops in tumor cells differ significantly from those in normal cells; in addition, the regulatory mechanisms for establishing R-loop homeostasis differ, which may be utilized to develop breakthrough therapies for various tumors. In this review, we summarize the current state of knowledge regarding non-classical R-loops in tumor cells, the mechanisms that promote or inhibit this structure, the effects of this structure on tumor cells, and the possible therapeutic targets. By systematically elucidating the pathogenic mechanisms of atypical R-loops, we can achieve precision targeting of tumors and revolutionize clinical precision oncology.

Background

The R-loop is a DNA–RNA complex consisting of an RNA strand that forms a hybrid with its template DNA strand, accompanied by the freeing of the non-template DNA strand. Free strands often form spatial structures composed of folded consecutive guanine repeating sequences, called G-quadruplexes (G4s). R-loops are of various types, the most studied and earliest discovered of which is the genomic co-transcriptional R-loop, also known as the cis-R-loop. This structure is widely found in various genomic regions and is involved in physiological processes such as transcriptional modulation [1]. However, it also contributes to genomic instability that induces copy number variations (CNVs) [2] and is directly associated with the development of diseases such as neurodegenerative disorders and cancer [3, 4]. Regulation of co-transcriptional R-loop can intervene in the proliferation and death of tumor cells [5].

In addition to the classical co-transcriptional R-loops, other types of R-loops, mainly trans-R loops, are formed by trans-annealing of specific RNAs to DNA; however, the mechanism via which certain R-loops are formed remains unclear. Based on the differences in the formation mechanism and spatial localization, these non-classical R-loops are classified into the telomeric R-loop, formed by trans-annealing of lncRNA TERRA to the telomeric region [6]; nucleolar R-loop, formed during RNA polymerase I (Pol I)-mediated transcription of rDNA [7]; centromeric R-loop, formed by the transcription of α-satellite repeats in centromeric regions [8]; circR-loop, formed by circular RNAs (circRNAs) annealing to the relevant DNA region [9]; and mitochondrial R-loop, which is a mechanistically undefined structure [10]. Notably, these R-loops participate in critical regulatory processes at distinct genomic regions, including chromosome segregation, telomere length homeostasis, ribosome biogenesis, transcriptional regulation, and metabolism. Emerging evidence indicates differential accumulation of these R-loops in tumor cells compared with that of their normal counterparts [11–13]. Intervention in the formation of these specific R-loops may offer novel therapeutic opportunities for cancer.

Although R-loops at specific chromatin sites exhibit tumor therapeutic potential, the functional heterogeneity of genomic regions leads to divergent effects of these R-loops on tumors. Live-cell imaging further revealed distinct kinetic behaviors of R-loops at these sites [14]. Blindly manipulating non-classical R-loop formation through co-transcriptional regulatory mechanisms may paradoxically yield adverse outcomes. A recent review has outlined the molecular regulatory network of co-transcriptional R-loops and their modulation in tumors [15]; however, non-classical R-loop regulation remains inadequately explored.

In this review, we systematically analyze the roles of non-classical R-loops in tumorigenesis and progression, delineated their regulatory networks, and compile potential therapeutic targets and clinically applicable drugs. This review aims to accelerate the clinical translation of therapies targeting non-classical R-loops.

Atypical R-loops in cancer: types and functions

Telomeric R-loops

Alternative lengthening of telomeres (ALT) is a telomerase-independent telomere maintenance mechanism first discovered in yeast cells in the 1990s. The active proliferative ability of tumor cells requires special telomere extension mechanisms; 85% of the tumors activate telomerase via different mechanisms, while 15%, known as telomerase-negative tumors, use the ALT pathway [16]. A meta-analysis of neuroblastoma revealed that patients with ALT-positive tumors had a 5-year overall survival rate of 53%, significantly lower than the 77% rate among patients with telomerase-positive tumors [17]. Separately, in a clinical cohort of patients with pancreatic neuroendocrine tumors, ALT positivity was significantly associated with increased recurrence risk [18]. Telomeric R-loops cause transcription-replication conflicts in telomerase-positive cells such as HeLa cells, which tend to inhibit R-loop formation via several mechanisms [19]. Analysis of surgical meningioma samples demonstrated markedly reduced telomeric R-loops in telomerase-positive tumors versus normal tissues [20]. In contrast, in telomerase-negative cells, R-loops trigger the ALT process [21] and protect short telomeres through persistent telomere cohesion—a phenomenon where shortened telomeres induce adhesion between sister telomeres [22]. Several studies have illustrated that telomeric R-loop and TERRA transcript levels are significantly higher in telomerase-negative cells than they are in telomerase-positive cells [23, 24]. Moreover, clinical studies demonstrated that TERRA levels significantly correlate with ALT phenotype [25]. However, excessive R-loop formation is detrimental even in ALT-positive tumor cells. In contrast, excessive R-loop formation leads to increased acute replication stress and formation of excessive telomere breaks that cannot be repaired in time. At the same time, elevated ALT activity leads to increased telomere fragility and instability, ultimately compromising cell viability [26, 27].

The molecular model of ALT is similar to that of break-induced replication (BIR), relying on homologous recombination repair after the creation of double-strand breaks (DSBs) (Fig. 1) [28]. Chronic replicative stress leads to DSB formation in telomeric DNA. In such cases, exonuclease 1, DNA2 endonuclease, and Bloom syndrome protein (BLM) collaborate to perform 5′–3′ resection [29–31]. The cell then accomplishes “strand invasion” via both RAD51-dependent and RAD51-independent pathways, where EXD2 dynamically regulates pathway selection [32]. Although initially presumed essential, RAD51-independent ALT persistence established the RAD52-dependent mechanism as ALT’s predominant pathway [33]. The strand invasion results in a structure where template DNA strands separate, with the broken DNA end inserted into the gap, termed a D-loop. Following strand invasion, DNA synthesis is executed by the PCNA–RFC–DNA polymerase δ (Polδ) replisome [34]. The BLM–TOP3A–RMI (BTR) complex dissolves D-loops to facilitate telomere elongation, whereas the SLX4–SLX1–MUS81 (SMX) complex counteracts this process by promoting rapid D-loop resolution [35, 36]. This antagonism results in replication-independent crossover formation and failure of telomere length maintenance. After DNA repair, telomeric DNA may re-invade other templates, forming heterologous telomeres [37]. Although the current model involving the RAD51-dependent and non-dependent pathways is convincing, other ALT pathways exist that need to be investigated [38].

Fig. 1.

Role of telomeric R-loops in ALT progression and their regulatory network. Top: Regulatory network of telomeric R-loops. SETX and DHX9 suppress R-loop formation through helicase activity. BRCA1 and RBM14 inhibit TERRA transcription. METTL3 stabilizes TERRA via m6A methylation to promote telomeric R-loops. RNH1 and XRN2 degrade RNA to suppress R-loops. BRCA2 inhibits telomeric R-loops. TRF2 and RAD51 facilitate TERRA annealing to telomeres, promoting R-loop formation, while TRF1 functionally antagonizes TRF2. G-quadruplex (G4) structures stabilize telomeric R-loops by trapping TOP1, TOP2A, and PARP1, whereas HILDA and Mutβ suppress G4 formation. The SFPQ–NONO complex inhibits telomeric R-loops. LSD1 promotes R-loop formation by binding TERRA-associated G4 structures. RBMX and THO complexes suppress telomeric R-loops through TERRA nuclear export. ATRX–DAXX normally inhibits H3.3 deposition; however, mutation elevates H3.3 levels and activates ALT progression; HIRA–UBN1/2 counteracts excessive H3.3 deposition. The G-overhang facilitates duplex fraying at the junction between double-stranded and single-stranded DNA, promoting telomeric R-loop formation. Bottom: Telomeric R-loops contribute to multiple steps in ALT progression. They act as persistent sources of replication stress, facilitate APB assembly, and act as structural scaffolds to promote D-loop formation through both RAD51-dependent and RAD52-dependent mechanisms. R-loops also intrinsically enhance D-loop formation independent of these pathways. DSB, double-strand break; APB, ALT-associated promyelocytic leukemia bodies; RNH1, RNase H1; Met, methyl. Created using https://www.BioRender.com

Telomeric R-loops form an important spatial structural basis for ALT. R-loop formation can lead to transcription-replication conflicts and is an important source of replication stress, a driver of ALT [39]. The R-loops recruit XPF and XPG, which can be utilized to generate DSBs using their endonuclease activity [40, 41]. ALT-associated promyelocytic leukemia bodies (APBs), often observed in ALT cells, have become a hallmark for ALT. Phase separation is a process whereby molecules spontaneously segregate from a homogeneous solution to form a dense phase with high molecular concentration and a dilute phase relatively depleted of molecules, serving as a crucial mechanism for cell organization and function. R-loops can increase telomeric histone H3K27me3 modification, which promotes liquid–liquid phase separation and assembly of APBs [42]. At the same time, R-loops can act as scaffolds to recruit RAD51, thereby facilitating its role in promoting DNA damage response (DDR) [43]. While the scaffolding function of classical TERRA R-loops is well-established, emerging evidence indicates that telomeric damage-induced long non-coding RNAs (dilncRNAs) exhibit analogous properties [44]. Although telomeric dilncRNA-mediated R-loop formation remains unconfirmed, their extra-telomeric R-loop generation capacity and DNA repair regulatory roles have been validated; thus, their telomeric functions warrant further investigation [45]. R-loops are also involved in the RAD51-independent repair pathway. A study of reactive oxygen species (ROS)-induced telomere disruption demonstrated that telomeric R-loops promote the localization of CSB and RAD52 at telomeres, which in turn facilitates repair via POLD3 (DNA Polδ subunit) [21]. In addition to their involvement in the two classical ALT pathways, R-loops intrinsically promote D-loop formation to engage the Polδ-mediated repair pathway by relying on opposing G4s [24]. In conclusion, R-loops play an important role in all steps of the ALT pathways, and ALT activity can be interfered by targeting R-loops.

Nucleolar R-loops

In contrast to telomeric R-loops, the functional significance of nucleolar R-loops in cancer cells has only been elucidated in the past 3 years. Pol I-driven nucleolar R-loops disrupt rDNA transcription in tumor cells, thereby impairing ribosome assembly. Precursor 47S rRNA (pre-47S rRNA), the primary transcript of rDNA, shows significant downregulation upon nuclear R-loop accumulation [46]. Clinical data demonstrated that elevated pre-47S rRNA levels serve as an independent predictor of poor prognosis in colorectal cancer patients [47]. Bioinformatics analysis revealed that patients with colorectal cancer exhibits the highest rRNA metabolism scores, followed by hepatocellular carcinoma and lung adenocarcinoma. Elevated rRNA metabolism scores significantly correlate with poor tumor prognosis [48]. Approximately 70% of global cellular transcription is driven by Pol I-mediated rDNA transcription within the nucleoli, highlighting the critical importance of nucleolar R-loop regulation [49]. Aberrant nucleolar R-loop accumulation leads to abnormal nucleolar morphology and impaired rRNA processing. These nucleolar structural and functional disruptions trigger nucleolar stress, activating the p53 pathway. Persistent p53 activation causes cell cycle arrest, whereas concurrent impairment of p53 downregulation ultimately drives cell death [46, 50]. Emerging evidence revealed that Pol II-generated R-loops coexist with Pol I-originating R-loops in the nucleoli, exhibiting functional antagonism. SincRNA is a product transcribed from the rDNA intergenic spacer (IGS). Aberrant overproduction of sincRNA impairs ribosome biogenesis and destabilizes nucleolar morphology. The antisense R-loop formed by RNA Pol II in the IGS region protects tumor cell nucleoli by counteracting RNA Pol I-mediated sincRNA transcription, a mechanism observed in Ewing sarcoma [51]. In addition to these two types of nucleolar R-loops, another R-loop is formed in the rDNA upstream control element. This R-loop is formed by either promoter RNAs (pRNA) transcribed by RNA Pol I or “promoter and pre-rRNA antisense” (PAPAS) transcribed by RNA Pol II. The accumulation of these R-loops also leads to the inhibition of rDNA transcription in tumor cells [52]. Evidence regarding this R-loop is limited and current understanding remains largely speculative (Fig. 2).

Fig. 2.

Regulatory network of nucleolar R-loops. Nucleolar R-loops form across three distinct rDNA regions. In the promoter region, pRNA and PAPAS act as primary RNAs driving R-loop formation, which suppresses rRNA production and ribosome biogenesis; this process is regulated by RPA, which binds DNA to recruit SETX for R-loop resolution. In the rRNA gene region, R-loops predominantly arise from RNA polymerase I (Pol I)-mediated transcription, with METTL8 and ROS enhancing nucleolar R-loop levels, while DDX5, DDX21, DDX47, DDX48, TOP1, and RNH1 suppress their formation. Excessive nucleolar R-loop accumulation induces nucleolar stress. In intergenic spacers, RNA Pol I transcribes sincRNA, whereas RNA Pol II inhibits this process via antisense R-loop formation. SincRNA transcription causes nucleolar morphological abnormalities and compromises ribosome biogenesis. RNH1, RNase H1; PAPAS, promoter and pre-rRNA antisense; Ac, acetyl. Created with https://www.BioRender.com

Centromeric R-loops

Centromeric RNA, transcribed by Pol II, acts as the essential structural component for centromeric R-loop formation. The centromere is the structural basis for ensuring accurate chromosomal segregation. Abnormal chromatin segregation contributes to aneuploidy. Clinical studies have revealed that tumors harbor diverse chromosome arm aneuploidies. Among tumor types, these aneuploidies demonstrated distinct prognostic implications, frequently correlating with worse patient outcomes [53]. Nevertheless, tumor cells retain surveillance mechanisms for chromatin segregation, such as CENP-A and Aurora B pathways, as concurrent multiple chromatin breaks would be catastrophic. CENP-A is a core component of centromere-specific nucleosomes. The cell cycle-dependent deposition of CENP-A at the centromere is essential for accurate chromosomal segregation. Premature or lack of CENP-A deposition results in chromosomal abnormalities, such as aneuploidy [54]. Pancancer analysis demonstrated that elevated CENP-A expression significantly correlates with inferior patient prognosis among multiple tumor types [55]. Aurora B localizes at the centromere and is required to resolve erroneous kinetochore-microtubule attachments. Activated Aurora B kinase expression significantly correlates with inferior prognosis in patients with hepatocellular carcinoma and non-small cell lung cancer (NSCLC) [56, 57]. The protective role of R-loops in centromere stability is mediated via these two molecules. In tumor cells, centromeric R-loops similarly exhibit dual roles. They activate the ATR–CHK1–Aurora B pathway in the centromeric regions of tumor cells [8, 58]. Additionally, they act as scaffolds to recruit RNF20, which promotes local H2Bub, H3K4me2, and subsequent SMARCA5 recruitment. This cascade ensures proper Aurora B kinase activation at centromeres and efficient loading of repair proteins at DNA breaks [59]. In osteosarcoma cells, binding of centromeric R-loops to EWSR1 ensures CENP-A accumulation at the centromeres [60]. However, excessive R-loop accumulation leads to chromatin instability (CIN); although CIN is a characteristic phenotype of many tumors such as leukemia, excessive CIN levels induce cell death [61, 62].

CircR-loops

Circular RNA–DNA hybrids were first reported in mitochondria in 1998 [9], and over the past two decades, they have been shown to play important roles in genome regulation in both plant and animal cells. In tumor cells, circR-loops are distinct from other R-loops as they specifically regulate single genes, enhancing their therapeutic potential. Circular dystrophin (circDMD) binds to the upstream region of the VEGFR3 promoter in tumor cells, promoting VEGFR3 expression [63]. This process depends on the NF–κB signaling pathway, uncovering a novel mechanism underlying cervical carcinogenesis associated with chronic inflammation. POLR2B represented a inferior prognostic biomarker in patients with renal cell carcinoma [64]. The R-loop formed between circPOLR2B and its parental gene inhibits POLR2B transcription. Conversely, YTHDC1 enhances POLR2B expression by mediating circPOLR2B nuclear export in a methylation-dependent manner, thereby promoting glioma development [65]. CircSMARCA5 demonstrated significant promise as a tumor biomarker. Clinical analyses confirmed that low circSMARCA5 expression significantly correlates with inferior prognosis in hepatocellular carcinoma, NSCLC, and colorectal cancer [66–68]. CircSMARCA5 exhibits a similar mechanism with circPOLR2B, except that it terminates transcription of SMARCA5 at exon 15 upon binding to its parental gene [69]. These findings demonstrate that circRNAs regulate key tumor genes with remarkable flexibility through position-dependent R-loop mechanisms, where promoter-associated R-loops enhance gene expression via transcription factor recruitment while intragenic R-loops impede transcriptional elongation.

Mitochondrial R-loops

Mitochondrial DNA (mtDNA) consists of a purine-rich heavy (H) strand and a pyrimidine-rich light (L) strand. Unlike mammalian nuclear DNA, it exists as a circular closed molecule devoid of introns, with a 1 kb control region. A unique RNA species localized at the control region forms mitochondrial R-loops. These RNAs, derived from the L-strand and mapped precisely to the mtDNA control region, have been designated as L-strand control region-associated RNAs (LC-RNAs) [10]. Mitochondrial R-loops are remarkably prevalent, accounting for > 50% of total mtDNA in cross-linked human fibroblasts. In addition, transcripts from mitochondrial genes such as CO2, CYB, and ND6 can also form R-loops at their corresponding motifs, albeit at lower abundance [70]. Mitochondrial R-loops regulate mitochondrial functions by modulating the mitochondrial genome. These mitochondrial functions, including calcium overload, oxidative stress, energy metabolism, and ferroptosis, are closely associated with tumorigenesis, metastasis, and tumor cell death [71–73]. These theoretical foundations have prompted investigations into the association between mitochondrial R-loops and tumors. One hypothesis suggested that mitochondrial R-loops influence mitochondrial DNA copy number by modulating DNA replication. Notably, mitochondrial DNA replication exhibits strand-asynchronous synthesis, and experimental evidence indicated that R-loops formed by LC-RNAs play a crucial role in initiating heavy-strand DNA synthesis [74]. However, if not promptly removed by regulatory factors such as RNase H1, accumulated R-loops can impair mitochondrial DNA polymerase γ activity, ultimately disrupting DNA replication [75, 76]. Reduced mitochondrial DNA copy number has been documented in various malignancies, including breast, gastric, renal, hepatic, and ovarian cancers [77]. Further suppression of mitochondrial DNA replication through manipulating R-loops may trigger tumor cell death via mechanisms including excessive oxidative stress and impaired energy metabolism. An alternative hypothesis proposes that mitochondrial R-loops influence tumor progression by modulating mitochondrial DNA mutations. Beyond their role in protecting DNA replication, mitochondrial R-loops facilitate the accurate separation of daughter mitochondrial DNA molecules during mitochondrial membrane fission [10]. Collectively, these mechanisms allow R-loops to minimize mitochondrial DNA variation. Furthermore, gene-specific mitochondrial R-loops may affect localized point mutations. Point mutations in mitochondrial genes ND6, CYB, and CO1 have been definitively associated with tumor progression [78]. Mitochondrial DNA CNVs correlate with prognosis in hepatocellular and gastric carcinomas, as well as the diagnostic age in prostate, colorectal, and cutaneous malignancies [79–81]. While mitochondrial DNA methylation correlates with tumor metastasis, mitochondrial R-loops are hypothesized to drive tumor progression via epigenetic mechanisms [82]. Despite these hypotheses, direct evidence demonstrating that mitochondrial R-loops regulate tumor progression by modulating mitochondrial DNA replication, mutation, or epigenetic modification remains unavailable.

Investigating the role of mitochondrial R-loops in tumors presents several technical challenges. First, mitochondrial R-loops are highly labile and prone to degradation during experimental extraction. Second, mitochondrial RNA comprises complex components. In addition to the frequent occurrence of artificial hybridization during extraction, interference may also arise from hybridization structures formed by the replication primer 7S RNA [83]. Although researchers have proposed methodological improvements, including sucrose gradient centrifugation (replacing traditional centrifugation) and psoralen plus UV crosslinking to stabilize R-loops [84], the current detection efficiency remains suboptimal. Resolution of these technical questions would revolutionize our understanding of mitochondrial R-loops. Furthermore, directly demonstrating the molecular functions of mitochondrial R-loops in tumors remains challenging. To date, no highly specific mitochondrial R-loop regulator has been identified. This limitation necessitates the exploration of alternative biological models for studying mitochondrial R-loops. The mitochondrial double-stranded DNA cytosine base editor (DdCBE) represents a potential tool for modulating R-loop formation in mitochondrial DNA [85]. Mitochondria-specific molecular inhibition requires efficient drug delivery. Dequalinium chloride (DQA), a bicationic compound composed of two symmetric molecules, exhibits highly selective mitochondrial targeting capability [86]. Alternatively, transmitochondrial cybrid technology can be utilized for generating rescue experimental models, in which experimental validation can be achieved by transferring mitochondria with elevated R-loop levels into cells with depleted R-loops [87]. These techniques are essential for establishing models with controlled mitochondrial R-loop levels, which are crucial for investigating mitochondrial R-loop biology.

Molecular regulation of atypical R-loops in cancers

Molecules with broad regulatory effects on R-loops

Certain molecules exhibit extensive regulatory functions across diverse R-loop types in various cellular contexts, and even other regulatory factors depend on them for indirect R-loop resolution. The most representative example is RNase H1, which was identified as a functional antagonist of R-loops and has been extensively utilized as a potent tool for R-loop elimination in laboratory studies. RNase H1 demonstrates remarkable versatility in suppressing all forms of R-loops [23, 46, 88, 89]. Another class of broad-spectrum R-loop regulators comprises the RNA–DNA helicases, with SETX acting as a classical prototype [52, 90]. Additional members include DDX5, DDX21, and DHX9 [91]. ILF3 acts as an indirect modulator of R-loop dynamics via its coordinated interaction with DHX9 to maintain telomeric R-loop homeostasis [92]. Notably, the RPA-mediated clearance of nucleolar R-loops via SETX recruitment represents another indirect regulatory mechanism [52].

Site-specific regulatory molecules for R-loops in tumor cells

Telomeric R-loops

G4s and R-loops have been found to coexist frequently in the genome when mapped using various techniques [93]. Owing to the single-stranded structure maintained by G4, the R-loop is able to exist more stably and can even complete the conversion to D-loop, participating in the ALT process [24]. Non-homologous end joining (NHEJ) is a template-independent DNA repair mechanism that often results in small insertions or deletions. G4s can act as a scaffold to trap TOP1, TOP2A, and PARP1 molecules, thereby increasing R-loop levels, inhibiting NHEJ pathway, and promoting the initiation of ALT [94, 95]. In contrast, HLTF and MSH2 inhibit the ALT process by resolving G4 structures [96]. Excessive G4 formation leads to over-accumulation of R-loops, resulting in cell death. The mismatch repair proteins, MSH2 and MSH3, form a heterodimer, MutSβ, which accumulates at the telomeres of ALT cells. In vitro experiments have shown that MutSβ destabilizes G4s to reduce R-loop accumulation and alleviates telomere fragility and damage [97].

Telomeres feature a G-rich 3′ single-stranded DNA overhang (composed of tandem TTAGGG repeats), which adopts the G4 configuration. The G-overhang facilitates the annealing of TERRA to DNA via unwinding at the junction between double-stranded and single-stranded telomeric regions, thereby promoting R-loop formation at telomeres [12].

The shelterin complex is a six-protein complex that binds telomeric DNA to maintain telomere integrity. TRF1 and TRF2, as its core components, perform antagonistic roles in regulating R-loop formation. TRF2 facilitates TERRA annealing to telomeric regions via its N-terminal B-domain, concurrently shielding R-loops from RNase H-mediated resolution [12]. TRF1 counteracts this process through its acidic domain, sterically hindering TRF2–TERRA interactions and thereby suppressing R-loop generation [6]. This TRF2-mediated facilitation requires G4 formation within TERRA transcripts [98]. In ALT-positive malignancies, TRF1 depletion triggers marked accumulation of telomeric DNA damage foci, whereas TRF2 inhibition induces catastrophic telomere attrition accompanied by senescence commitment [99, 100]. This functional antagonism establishes their dual targeting as a strategic vulnerability in ALT-dependent oncogenesis.

BRCA1 binds to TERRA within APBs in a R-loop-dependent manner, which is critical for telomeric DNA synthesis during ALT. Nevertheless, BRCA1 mediates the recruitment of DNMT3b to repress TERRA transcription in ALT-positive cells, thereby attenuating telomeric R-loop formation. Inhibition of BRCA1 exacerbates R-loop-mediated DNA damage while disrupting ALT-promoted scaffolding, culminating in telomere shortening in ALT cells [90].

RNA modification by the METTL3 methyltransferase at the N6 position of internal adenosine (m6A) in TERRA is essential for telomere maintenance in ALT-positive cells. Genetic ablation of METTL3-mediated m6A modification results in telomere damage. The underlying mechanism stems from the m6A-dependent interaction between hnRNPA2B1 and TERRA, which is indispensable for telomeric R-loop formation [101]. Furthermore, m6A confers stability to TERRA RNA. The m6A reader, YTHDC1, specifically recognizes and preserves m6A modifications, whereas knockdown of METTL3 or YTHDC1 induces TERRA degradation and consequently reduces telomeric R-loop levels [102].

Numerous RNA-binding proteins regulate telomeric R-loop levels via TERRA binding. The hnRNP protein family performs critical regulatory functions in modulating TERRA abundance. RBMX facilitates TERRA nuclear export [103], while RBM14 not only suppresses TERRA transcription but also inhibits its capacity to form R-loops [104]. Endonucleases such as XRN2 mediate TERRA degradation, thereby attenuating telomeric R-loop formation and alleviating replication stress [105]. RTEL1, a DNA helicase, maintains telomere stability in telomerase-positive cells by preserving G-overhang structures or resolving t-loops [106]. It downregulates TERRA expression while paradoxically enhancing TERRA localization at the telomeres, thereby increasing telomeric R-loop levels [107]. RTEL1 participates in mitotic DNA synthesis by recruiting SLX4, RAD52, and POLD3—a process mechanistically analogous to ALT-associated mechanisms. Although no direct evidence currently links RTEL1 to ALT activity regulation, its functional parallels in telomere maintenance warrant further exploration of this target [108]. FANCM, functioning as an RNA–DNA helicase, specifically clears telomeric R-loops (validated experimentally in vitro) to sustain ALT cell viability [109]. RAD51 functions as a trans-acting factor that recruits TERRA to telomeric regions, enhancing TERRA's annealing capacity to telomeric DNA [12, 19]. In vitro reconstitution assays have demonstrated that LSD1 binds to TERRA-associated G4 structures, stimulating telomeric R-loop formation. Mechanistic analysis shows that LSD1-driven phase separation at telomeres recruits RAD51AP1, a factor with significantly higher R-loop-promoting activity than that of RAD51. LSD1 depletion results in significant impairment of ALT progression [24, 110]. THOC, a hexameric complex essential for RNA nuclear export, was recently shown to suppress TERRA-telomere annealing through TERRA binding, thereby restricting R-loop formation. This R-loop limitation suppresses homologous recombination and attenuates telomeric fragility [111]. The NONO/SFPQ heterodimer, an ALT regulator, inhibits excessive telomeric R-loop formation via TERRA binding, preventing telomere fragility. Genetic ablation of NONO/SFPQ induces 35% telomere elongation in ALT cells (Fig. 1) [112].

Nucleolar R-loops

Among DDX family members, DDX47 and DDX48 perform specialized functions within the nucleolus. DDX48 is highly expressed in various tumor cells and is responsible for R-loop processing within the small subunit processome [50]. DDX47 utilizes its DNA–RNA-unwinding activity to inhibit both RNA pol I and RNA pol II R-loops, although the latter effect is weaker [113]. Although DDX21 itself lacks nucleolar specificity, NAT10 specifically regulates nucleolar R-loops by acetylating DDX21 at K236 and K573 [114]. In addition to these inhibitory factors, METTL8 is the only molecule conclusively shown to maintain nucleolar R-loop levels. METTL8 knockdown nearly eliminates nucleolar R-loops yet paradoxically exerts antitumor effects [115]. This discrepancy suggests two hypotheses: first, METTL8's involvement in mt-tRNA m3C modification may mediate its antitumor role independent of R-loop regulation [116]; second, basal nucleolar R-loop levels might be crucial for tumor growth, potentially via R-loop-associated DNA repair mechanisms. Further studies are required to resolve this (Fig. 2).

Regarding Pol II-induced antisense R-loops, TBPL1 localizes to the TCT motif and coordinates Pol II and Pol I activities to maintain baseline sincRNA levels. PAF1 facilitates Pol II elongation while suppressing unscheduled R-loop formation and maintains transcriptional homeostasis in the IGS region. However, the pathological relevance of this regulatory axis remains unvalidated in tumor cell models [117].

Centromeric R-loops

Centromeric R-loops have only recently been recognized, and the molecular networks regulating them in tumor cells remain poorly understood. Molecules that specifically regulate centromeric R-loops include BRCA1, DNMT3b, EHMT2, and DAXX (Fig. 3). Binding of BRCA1 to the centromere prevents the latter from being disrupted by R-loop overaccumulation and promotes Rad52-mediated homologous recombination [118]. In pediatric gliomas and pancreatic neuroendocrine tumors, DAXX forms a complex with SETDB1 in an ATRX-independent manner to resolve R-loops and reduce centromeric genomic damage [61]. Among these molecules, EHMT2 uniquely promotes centromeric R-loop accumulation in tumor cells. CUL3 ubiquitinates EHMT2 for degradation, resulting in impaired chromosome segregation and suppressed tumor cell proliferation, a phenomenon particularly prominent in prostate cancer [58]. DNMT3b uniquely protects centromeric R-loops from XPG-mediated cleavage, thereby preventing centromeric DNA damage and aberrant NHEJ activation [119].

Fig. 3.

Regulatory network of centromeric R-loops. Top: Centromeric R-loops primarily ensure accurate chromosome segregation via the ATR–CHK1–Aurora B pathway and CENP-A. Bottom: BRCA1 suppresses R-loops by recruiting SETX. DNMT3b protects centromeric stability by blocking XPG’s access to R-loops. The DAXX-SETDB1 complex inhibits centromeric R-loop formation, a process dependent on DAXX–H3.3 interaction. EHMT2 elevates centromeric R-loop levels, whereas CUL3 counteracts this by ubiquitinating EHMT2. Mutant SF3B1 causes aberrant splicing of SERBP1, generating isoform 3 that promotes centromeric R-loop accumulation. Ub, ubiquitin; iso, isoform. Created using https://www.BioRender.com

CircR-loops and mitochondrial R-loops

No consensus has been reached regarding circR-loop regulation owing to substantial heterogeneity among its subtypes. A mechanistic study on circRNA nuclear export revealed XPO4 as a key regulator suppressing circR-loop formation [120], thereby elucidating its tumor-suppressive function. However, experimental validation in tumor cell models remains lacking. Although the role of mitochondrial R-loops in tumor cells remains hypothetical, investigating their regulatory mechanisms in tumors may not directly inform therapeutic strategies but can nevertheless facilitate future research advancements. Beyond RNase H1, the human mitochondrial degradosome (mtEXO), composed of SUV3 and PNPase, reduces mitochondrial R-loop levels by promoting RNA–DNA hybrid dissociation and RNA degradation [121]. Additionally, mitochondrial matrix-localized p53 suppresses mitochondrial R-loop accumulation via its 3′ → 5′ exoribonuclease activity [122]. These findings indicated that the mitochondrial RNA metabolic system plays a crucial role in regulating mitochondrial R-loops. However, because of the aforementioned technical challenges, the current understanding of mitochondrial R-loop regulatory mechanisms remains incomplete, warranting further investigation.

Mutation-driven R-loop regulatory mechanisms in tumor cells

The ATRX–DAXX complex suppresses telomeric R-loop formation via H3.3 histone deposition, mutations in which constitute a hallmark of ALT [123]. Recent studies have identified H3.3 as a master regulator of R-loop homeostasis. Upon ATRX–DAXX dysfunction, failed H3.3 incorporation triggers R-loop elevation and ALT pathway activation. Concurrently, the HIRA–UBN complex mediates chromatin recruitment of CHK1-phosphorylated H3.3, thereby suppressing pathological R-loop accumulation. Synthetic lethality between HIRA and ATRX–DAXX corroborates this mechanism [124]. Notably, HIRA recruitment requires PARylation—a process potently blocked by PARPi [125].

Clinical analyses have demonstrated that approximately half of BRCA2-mutated human cancers exhibit APBs, whereas no BRCA2 wild-type breast cancer specimens display APBs, establishing an essential role of BRCA2 in APB biogenesis. BRCA2 suppresses telomeric R-loops, thereby inhibiting the recruitment of polycomb repressive complex 2 (PRC2). PRC2 is required for H3K9me3-mediated epigenetic modification of telomeric DNA, which is essential for APBs assembly [42].

Under hypoxic conditions, nucleolar RNA pol I-driven R-loop accumulation promotes EHMT2-mediated H3K9me2 deposition. This modification reduces transcriptional efficiency, triggers nucleolar remodeling, and ultimately fosters tumorigenesis [13]. Consequently, tumors acquire mutations such as PTEN loss and MYC amplification to boost compensatory rDNA transcription [126]. Although derived from tumorigenesis research, targeting of R-loops via ROS modulation holds significant therapeutic potential.

SF3B1, a key co-transcriptional R-loop regulatory protein, has recently been shown to drive centromeric R-loop accumulation in a mutation-dependent manner [62]. SF3B1 mutations cause aberrant alternative splicing of SERBP1, generating a non-canonical splice isoform 3. This isoform compromises cellular RNA processing capacity, leading to transcriptional stalling and impaired R-loop resolution, which results in R-loop accumulation. Inhibition of SERBP1 recapitulates the effects of SF3B1 mutation (Table 1).

Table 1.

Key regulatory molecules of atypical R-loops in cancer

R-loop	Cancer	Cell line	Animal	Key regulators	Effect on R-loop	Phenotype	Therapeutic strategy	Refs.
Telomere R-loop	Cancer with ATRX mutation	U2OS (osteosarcoma), ATRX KO IMR90 (fibroblast) and VA13 (fibroblast)	NA	HIRA-UBN-CHK1/H3.3	Downregulation	Telomere instability	Synthetic lethality	[124]
Telomere R-loop	ALT-positive cancers	U2OS (osteosarcoma), SAOS2 (osteosarcoma) and SKLU1 (lung adenocarcinoma)	NA	RAD51AP1, G4s	Upregulation	ALT	Targeted therapy	[24]
Telomere R-loop	Cancer with ATRX mutation	U2OS (osteosarcoma), Hela LT (cervical carcinoma), SW26 (ovarian adenocarcinoma), SW39 (fibroblast), MGBM1 (glioblastoma)	NA	G4s	Upregulation	ALT	Targeted therapy	[94]
Telomere R-loop	ALT-positive cancers	U2OS (osteosarcoma)	NA	HLTF, MSH2	Downregulation	ALT	Targeted therapy	[96]
Telomere R-loop	ALT-positive cancers	NA	NA	TRF2	Upregulation	NA	Targeted therapy	[12]
Telomere R-loop	Cancers	U2OS (osteosarcoma), Hela (cervical carcinoma)	NA	TRF1	Downregulation	Telomere instability	Targeted therapy	[6]
Telomere R-loop	Cancers	U2OS (osteosarcoma), Hela (cervical carcinoma)	NA	BRCA1, DNMT3b, XRN2, SETX	Downregulation	Telomere instability	Targeted therapy	[90]
Telomere R-loop	Cancers	HeLa LT (cervical carcinoma), BJ (fibroblast)	NA	BRCA2, PRC2	Downregulation	ALT	Synthetic lethality	[42]
Telomere R-loop	ALT-positive cancers	U2OS (osteosarcoma), SK-N-FI (neuroblastoma)	Mouse xenograft model	METTL3, hnRNPA2B1	Upregulation	ALT	Targeted therapy	[101]
Telomere R-loop	ALT-positive cancers	U2OS (osteosarcoma), VA13 (fibroblast), CAL27 (oral squamous cell carcinoma)	NA	METTL3, YTHDC1	Upregulation	ALT	Targeted therapy	[102]
Telomere R-loop	ALT-positive cancers	U2OS (osteosarcoma), VA13 (fibroblast), GM847 (fibroblast)	NA	RNaseH1	Downregulation	ALT	Targeted therapy	[23]
Telomere R-loop	Cancers	U2OS (osteosarcoma), HeLa (cervical carcinoma)	NA	RBMX	Downregulation	Telomere instability	Targeted therapy	[103]
Telomere R-loop	Cancers	U2OS (osteosarcoma), Hela (cervical carcinoma)	NA	RBM14	Downregulation	Telomere instability	Targeted therapy	[104]
Telomere R-loop	ALT-positive cancers	SaOS2 (osteosarcoma)	NA	XRN2	Downregulation	ALT	Targeted therapy	[105]
Telomere R-loop	ALT-positive cancers	U2OS (osteosarcoma), VA13 (fibroblast)	NA	ILF3, DHX9	Downregulation	Telomere instability	Targeted therapy	[92]
Telomere R-loop	ALT-positive cancers	U2OS (osteosarcoma), SaOS2 (osteosarcoma)	NA	LSD1	Upregulation	ALT	Targeted therapy	[110]
Telomere R-loop	Cancers	HeLa (cervical carcinoma), U2OS (osteosarcoma) and Saos2 (osteosarcoma)	NA	THO complex	Downregulation	ALT	Targeted therapy	[111]
Telomere R-loop	Cancers	HeLa (cervical carcinoma), U2OS (osteosarcoma) and Saos2 (osteosarcoma)	NA	MutSβ	Downregulation	ALT	Synthetic lethality	[97]
Telomere R-loop	Cancers	U2OS (osteosarcoma), Hela (cervical carcinoma)	NA	NONO, SFPQ	Downregulation	ALT	Targeted therapy	[112]
Telomere R-loop	ALT-positive cancers	U2OS (osteosarcoma), HuO9 (osteosarcoma) and Saos2 (osteosarcoma)	NA	FANCM	Downregulation	Telomere instability	Targeted therapy	[109]
Nucleolar R-loop	Cancers	HeLa (cervical carcinoma)	NA	RNaseH1, TOP1	Downregulation	Ribosome biogenesis	Targeted therapy	[46]
Nucleolar R-loop	Cancers	HeLa (cervical carcinoma)	NA	RPA, SETX	Downregulation	Ribosome biogenesis	Targeted therapy	[52]
Nucleolar R-loop	Cancers	HeLa (cervical carcinoma), HCT116 (colon cancer)	Mouse xenograft model	METTL8	Upregulation	Cell proliferation	Targeted therapy	[115]
Nucleolar R-loop	Cancers	U2OS (osteosarcoma), A549 (lung carcinoma), HeLa (cervical carcinoma), MCF7 (breast cancer), HCT116 (colon cancer)	NA	DDX48	Downregulation	rRNA processing	Targeted therapy	[50]
Nucleolar R-loop	Cancers	U2OS (osteosarcoma), Hela (cervical carcinoma)	NA	DDX47	Downregulation	Ribosome biogenesis	Targeted therapy	[113]
Nucleolar R-loop	Cancers	HeLa (cervical carcinoma), HCT116 (colon cancer)	NA	NAT10, DDX21	Downregulation	Ribosome biogenesis	Targeted therapy	[114]
Nucleolar R-loop	Cancers	A549 (lung carcinoma), HCT116 (colon cancer), RKO (colon cancer)	NA	ROS	Upregulation	Ribosome biogenesis	Synthetic lethality	[13]
Nucleolar R-loop	Cancers	HeLa (cervical carcinoma), U2OS (osteosarcoma)	NA	RNA polymerase II	Upregulation	Ribosome biogenesis	Targeted therapy	[51]
Centromeric R-loop	Cancers	HT1080 (fibrosarcoma), U2OS (osteosarcoma), HCC1937 (breast cancer)	NA	BRCA1, SETX, RNase H1	Downregulation	Centromeric instability	Synthetic lethality	[118]
Centromeric R-loop	Cancers	HCT116 (colon cancer), pGM08714 (lymphoblastoid cell line)	NA	DNMT3b, XPG/XPF	Upregulation	Centromeric instability	Targeted therapy	[119]
Centromeric R-loop	Pancreatic neuroendocrine tumor	SF188 (glioblastoma), BON-1 (pancreatic neuroendocrine tumor)	NA	DAXX, SETDB1	Downregulation	Centromeric instability	Synthetic lethality	[61]
Centromeric R-loop	Chronic lymphocytic leukemia	K562 (myelogenous leukemia), Nalm-6 (lymphoblastic leukemia)	Double mutant mice model	SF3B1, SERBP1	Downregulation	Centromeric instability	Synthetic lethality	[62]
Centromeric R-loop	Prostate cancer	DU145 (prostate cancer), 22Rv1 (prostate cancer)	Mouse xenograft model	EHMT2, CUL3	Upregulation	Centromeric instability	Targeted therapy	[58]
circR-loop	Ovarian cancer	PA1 (ovary Teratocarcinoma)	NA	RNase H1, DDX5, DDX21, DHX9	Downregulation	Transcription elongation	Targeted therapy	[88]

Therapeutic targeting of atypical R-loops

Strategies for telomeric R-loops

The distinct telomere elongation mechanisms suggest that R-loop-targeted therapies may require tumor-specific profiling. Current data indicated ALT tumor frequencies of 63% in osteosarcoma, 60% in leiomyosarcoma, 32% in pancreatic neuroendocrine tumors, and 20% in CNS tumors [16, 127–129]. Telomeric R-loop levels are essential for the ALT process and are significantly elevated in ALT tumors, a tumor subtype that intrinsically exhibits poorer prognosis [130]. Furthermore, even within ALT-positive tumor populations, more telomeric R-loops correlate with poorer clinical outcomes [101]. Stratifying patients by telomeric R-loop levels effectively differentiates ALT-positive from telomerase-positive tumors and serves as a prognostic indicator in the aforementioned tumor types. Tumors with high telomeric R-loop levels demonstrate greater sensitivity to R-loop-increasing therapies while remaining responsive to R-loop-reducing treatments. In contrast, tumors exhibiting low R-loop levels derive better therapeutic benefit from telomerase inhibitors than from R-loop-targeted approaches.

Telomeric R-loop-induced intratumoral telomere instability drives resistance to conventional chemotherapy, telomerase inhibitors, and targeted therapies. Activation of the ALT pathway leading to therapy resistance is commonly observed in multiple tumor types treated with telomerase inhibitors [131]. ALT-positive neuroblastomas with high telomeric R-loops depend on downstream ATM activity for temozolomide chemoresistance [132]. Moreover, ALT-induced telomere breaks and telomere fusions are associated with drug targets such as BRAF^V600E amplification in tumor cells [133, 134]. This indicates that they may represent a common mechanism underlying resistance to targeted drugs, such as methotrexate and vemurafenib. These drug resistance mechanisms illustrate another layer of significance in utilizing telomeric R-loops for patient stratification in oncology. That is, patients with high telomeric R-loops exhibit resistance to conventional chemotherapy. Monitoring telomeric R-loop levels in drug-resistant patients could enable timely adjuvant therapies targeting R-loop/ALT pathways to overcome resistance.

Beyond conventional chemotherapy, two therapeutic approaches are currently being developed for ALT tumors: one involves direct inhibition of proteins in the ALT-associated BIR pathway using agents such as BTR inhibitors [135], while the other targets key mediators in DDR following telomeric dysfunction, such as ATR inhibitors (ATRi) [136], ATM inhibitors [137], WEE1 inhibitors [138], and p53 inhibitors [139]. These strategies have been under investigation for years. Despite encouraging preclinical data, clinical efficacy remains suboptimal [140], with ATRi and CHK1 inhibitors demonstrating comparable efficacy in both ALT-positive and -negative tumors [141]. Considering the critical role of R-loops in the ALT pathway, R-loop-targeted therapies may demonstrate enhanced specificity and efficacy.

G4 stabilizers are considered potential inhibitors of telomerase activity in telomerase-dependent cancers; yet over a decade ago, multiple studies demonstrated that G4 stabilizers, including BMVC4, RHPS4, and tetra-Pt, also exhibit therapeutic efficacy against ALT-positive malignancies [142–144]. The mechanism via which these drugs combat ALT-positive tumors involves elevation of R-loop levels and enhancement of ALT activity, which ultimately leads to increased telomere fragility and induction of cell death [94]. As the most clinically promising G4 stabilizer to date, CX5461 has completed Phase I clinical trials for both hematologic malignancies and solid tumors [145, 146]. Despite phototoxicity, it demonstrates efficacy against various tumors, particularly those with BRCA1/2 mutations, with some cases achieving long-term remission. Although the results of Phase I clinical trial have been reported and recent studies demonstrate the efficacy of CX-5461 in murine xenograft models of ATRX-deficient malignant gliomas [147], clinical trials have not yet validated the use of G4 stabilizers in ALT-positive tumors. Further investigation regarding these agents in clinical practice is highly warranted, and the outcomes of Phase III clinical trials are eagerly awaited.

PARP inhibitors (PARPi) have become well-established therapeutics in ovarian cancer, with their safety and efficacy profiles bolstering confidence in small-molecule targeted agents. Expanding their clinical utility remains an ongoing pursuit. Notably, ALT-positive cells exhibit intrinsic sensitivity to TOP1 inhibitors such as irinotecan, positioning these agents as cornerstone therapies for ALT-driven malignancies [148]. G4s mediate PARP trapping, thereby suggesting that G4 stabilizers mimic the therapeutic effects of PARP trappers [94]. Concomitantly, PARPi exhibits histone chaperone inhibitory activity [125]. A recent case report demonstrates that a combination of PARP-trapping inhibitors (talazoparib and niraparib) with irinotecan potently suppresses ALT-positive tumors [148]. In a Phase I clinical trial investigating the effects of first-line treatment of glioblastoma with a combination of olipudase alfa with temozolomide and radiotherapy, a median progression-free survival of 6.2 months and an overall survival of 19.8 months were observed [149]. Niraparib leads PARP inhibitor development with completed Phase II glioma/glioblastoma trials and ongoing Phase III glioblastoma trials. Olaparib is under broader Phase II investigation in osteosarcoma, neuroendocrine tumors, and brain tumors, whereas talazoparib and pamiparib remain in Phase I evaluation (Table 2).

Table 2.

Clinical trials targeting telomeric R-loops in ALT-positive tumors

Drug type	Inhibitors	Clinical trial	Tumor
PARP inhibitor	Niraparib	NCT04544995 I (ongoing)	Neoplasms
		NCT05297864 II (completed)	Glioblastoma Astrocytoma Oligodendroglioma Glioma
		NCT06258018 II (ongoing)	Malignant glioma
		NCT01294735 I (completed)	Glioblastoma
		NCT04221503 II (ongoing)	Glioblastoma
		NCT06388733 III (ongoing)	Glioblastoma
	Talazoparib	NCT05053854 I (ongoing)	Neuroendocrine Tumors
	Pamiparib	NCT03150862 Ib/II (completed)	Central Nervous System Tumors
		NCT03914742 I/II (completed)	Glioma Glioblastoma
	Olaparib	NCT04086485 I/II (ongoing)	Neuroendocrine tumors
		NCT05870423 I (ongoing)	Neuroendocrine tumors
		NCT04417062 II (ongoing)	Osteosarcoma
		NCT05188508 II (ongoing)	Glioma
		NCT03991832 II (ongoing)	Glioma
		NCT03561870 II (completed)	Glioma
		NCT06607692 II (ongoing)	Neuroblastoma Glioma Neuroendocrine tumors
		NCT03212274 II (ongoing)	Glioblastoma Glioma
		NCT05463848 II (ongoing)	Glioblastoma
		NCT04614909 I (ongoing)	Glioblastoma
		NCT04375267 I (ongoing)	Neuroendocrine Tumors
		NCT03212742 I/IIa (ongoing)	Gliomas
RNA Pol II inhibitor	Lurbinectedin	NCT02611024 I/II (ongoing)	Glioblastoma
DNMT inhibitor	Decitabine	NCT02332889 I/II (completed)	Gliomas Medulloblastoma Neuroectodermal Tumors
		NCT01241162 I (completed)	Neuroblastoma Osteosarcoma
		NCT05178693 I/II (ongoing)	Neuroendocrine tumors
	Azacitidine	NCT03628209 Ib/II (ongoing)	Osteosarcoma
		NCT03666559 II (ongoing)	Glioma
		NCT03206021 I/Ib (ongoing)	Ependymoma
		NCT06896110 I (ongoing)	Glioblastoma Glioma Astrocytoma
		NCT03684811 I/II (completed)	Glioma

The TRF1 inhibitors, ETP-47228, ETP-47037, and ETP-50946, suppress the growth of ALT-positive glioma stem cells potently, the efficacy of which is enhanced when combined with temozolomide or radiation therapy [99]. These drugs promote telomeric R-loop accumulation and significantly increase telomere dysfunction-induced foci, a validated marker of telomere damage. Furthermore, ETP-47228 and ETP-47037 inhibit the PI3K/AKT axis, collectively amplifying their antitumor significance [150]. Currently, no TRF1 inhibitor has entered clinical trials. Blood–brain barrier penetration remains a key challenge for neurological tumor applications, necessitating further drug optimization. Conversely, TRF2 inhibitors suppress ALT activation by reducing telomeric R-loop accumulation and shortening telomeres to induce senescence. FKB04, a leading TRF2 inhibitor currently under preclinical investigation, exhibits anti-proliferative effects in a TRF2-overexpressing hepatocellular carcinoma model via telomere shortening mediated by T-loop formation [151]. This demonstrated its capacity to overcome tumor heterogeneity and induce senescence among tumors with diverse telomere maintenance mechanisms. Structure-guided optimization of FKB04 has yielded novel TRF2-targeting compounds that show promising anti-osteosarcoma activity in preclinical models, laying the groundwork for clinical translation [152].

EZH2 acts as a catalytic core protein of PRC2. Phase II trial data regarding the EZH2 inhibitor, tazemetostat, has garnered the approval of the Food and Drug Agency for epithelioid sarcoma treatment [153]. Although EZH2 does not directly regulate R-loop levels and monotherapy with EZH2 inhibitors exerts minimal cytotoxicity, co-administration with G4-stabilizing agents enables dose-reduced G4 stabilizers to achieve enhanced therapeutic efficacy, highlighting their combinatorial therapeutic potential [42].

METTL3 inhibitors represent the most promising class of RNA methyltransferase inhibitors. By suppressing telomeric R-loop levels, they induce telomere shortening. METTL3 inhibitors, including STM2457, UZH1A, and UZH1B, significantly suppress the growth of ALT-positive neuroblastoma [101]. Furthermore, STM2457 inhibits the pentose phosphate pathway (PPP), modulating tumor metabolism [154]. In addition, it simultaneously enhances interferon-gamma (IFNγ) and granzyme B (GzmB) expression, potentiating T-cell cytotoxicity [155]. These findings demonstrated its therapeutic potential in hepatocellular carcinoma and small-cell lung cancer [154, 156]. STC-15, a derivative of STM2457, is the most advanced METTL3 inhibitor. It has exhibited an excellent safety profile with good patient tolerance in a global Phase I trial (NCT05584111).

Several DNMT3b inhibitors under development show therapeutic potential [157]. Nanaomycin A, the most selective DNMT3b inhibitor identified to date, demonstrates therapeutic potential in pediatric neuroblastoma preclinical models [158]. However, its clinical translation was limited by poor bioavailability and significant cytotoxicity [159]. Although no DNMT3b-specific inhibitors have reached clinical trials, the DNMT inhibitors decitabine and azacitidine are clinically approved. However, in a Phase I clinical trial evaluating the efficacy of decitabine in neuroblastoma treatment, objective responses were not observed, with the optimal dosing regimen achieving only a median disease stabilization period of 7 months [160]. These findings clearly fail to demonstrate the full therapeutic potential of DNMT3b inhibitors in ALT-associated tumors, underscoring the need for more precise drug development and translational strategies.

LSD1 inhibitors are being extensively investigated in oncology, with more than a dozen candidates having entered clinical development. Current LSD1 inhibitors can be categorized into two mechanistic classes: irreversible covalent inhibitors (e.g., IMG-7289) and reversible inhibitors (e.g., SP-2577). However, LSD1-mediated promotion of R-loop formation depends on its RNA-binding capacity rather than on enzymatic activity [110]. Therefore, existing LSD1 inhibitors are not recommended for ALT-associated tumor treatment. Novel LSD1 inhibitors capable of disrupting LSD1–TERRA interactions may represent viable therapeutic candidates for ALT tumors.

Another promising RNA-targeting therapeutic strategy involves Pol II inhibitors, which comprise two mechanistically distinct categories. Previous studies have demonstrated telomeric R-loop suppression in cells treated with 5,6-dichlorobenzimidazole 1-β-D-ribofuranoside, an RNA Pol II inhibitor [24]. The first class modulates RNA Pol II transcription through selective inhibition of CDK family kinases, with CDK9 and CDK7 being the most clinically relevant targets [161]. Notably, treatment with the CDK7 inhibitor, THZ2, induced significant tumor growth suppression in osteosarcoma models [162]. Nevertheless, no such agents have yet demonstrated efficacy against ALT-positive tumors in clinical trials. The second class consists of direct-acting small-molecule Pol II inhibitors, exemplified by lurbinectedin, which has demonstrated acceptable safety profiles in Phase II trials [163]. Beyond telomeric alterations, extensive DNA damage was observed in lurbinectedin-treated cells [164], along with a reduction in tumor-associated macrophages in the tumor microenvironment [165]. In a Phase III trial, combining Tecentriq with lurbinectedin significantly improved survival in patients with small-cell lung cancer [166]. A Phase I trial evaluating the efficacy of lurbinectedin in glioma treatment has been initiated (Table 2). Thus, lurbinectedin currently represents the most promising RNA Pol II inhibitor for ALT tumors. We further anticipate developing a novel telomere-targeted RNA Pol II inhibitor that could provide more specific antitumor effects with enhanced safety.

Nucleolar R-loop modulation

The high transcriptional activity of RNA polymerase I makes R-loops in nucleolar rDNA regions critical targets for cancer therapy. Tumor cells with low nucleolar R-loop levels maintain efficient rRNA transcription, whereas pharmacological or mutational induction of R-loop accumulation significantly suppresses pre-47S rRNA expression. Progressive R-loop accumulation ultimately triggers nucleolar stress and tumor cell death. Notably, some tumors with low ribosome biogenesis exhibit progression dependent on "onco-ribosomes," which are mutant ribosomes that selectively translate oncogenes and metastasis-related transcripts [167]. Nucleolar R-loop levels serve as biomarkers to stratify ribosome-dependent tumors, and this stratification can predict patient prognosis especially in colorectal cancer. Nucleolar R-loop-targeted therapies could be particularly effective against these malignancies. Tumors with elevated nucleolar R-loop levels should be investigated for “onco-ribosomes” to identify sensitive therapeutic targets. Furthermore, Increased rRNA transcription has been identified as a key mechanism underlying resistance to two classical chemotherapeutic agents, cisplatin and paclitaxel [168]. Tumor cells counteract chemotherapy toxicity by enlarging nucleoli and elevating rRNA levels to achieve enhanced ribosome biogenesis [169]. Combination therapies targeting nucleolar R-loops represent a promising strategy to overcome such chemoresistance.

Ribosome biogenesis acts as an essential lifeline for tumor cells. Induction of nucleolar stress has emerged as a novel therapeutic strategy, currently implemented via two primary approaches. RNA polymerase I inhibitors (e.g., BMH-21) suppress nucleolar biogenesis by inhibiting rDNA transcription [170]. However, such inhibitors exhibit chromosomal toxicity and suboptimal clinical efficacy. Activation of downstream p53 pathways demonstrates clinical potential, with recent evidence revealing that BRIX1 inhibition enhances 5-fluorouracil sensitivity and suppresses proliferation [171]. Notably, targeted accumulation of nucleolar R-loops via novel molecular targets represents a mechanistically analogous approach to p53 activation, showing substantial promise for future cancer therapeutics.

Therapeutic targeting of nucleolar R-loops in tumors remains nascent. The current therapeutic paradigm focuses on enhancing RNA Pol I-driven R-loops while suppressing Pol II-associated R-loops. Specific inhibitors for the major targets that modulate nucleolar R-loops—including METTL8, DDX47, and DDX48—have not been reported, warranting further drug development. TOP1 inhibitors demonstrate dual-phase nucleolar R-loop modulation: transient treatment elevates nucleolar R-loop levels, while prolonged exposure induces nucleolar disintegration, establishing them as effective nucleolar R-loop modulators [46]. Specific inhibitors for several targets already exist. For instance, while RPA inhibitors (e.g., HAMNO, TDRL-551) remain in preclinical development, they demonstrate efficacy in the treatment of nasopharyngeal carcinoma and NSCLC models [172, 173]. The heightened sensitivity of RPA-depleted tumor cells to TOP1 inhibitors further validates this approach [52]. NAT10 depletion specifically induced nucleolar R-loop accumulation, leading to nucleolar stress. Additionally, remodelin, a early NAT10 inhibitor, suppressed tumor angiogenesis by inhibiting HIF expression [174] NAT10 inhibitors have demonstrated effects in suppressing osteosarcoma and improving doxorubicin resistance in hepatocellular carcinomas [175, 176]. However, no clinical trial data currently support their use in tumors. The newly identified NAT10 inhibitors paliperidone and AG-401 demonstrate superior therapeutic efficacy compared with remodelin. Accelerating clinical evaluation of these compounds, coupled with structure-based drug design utilizing the resolved NAT10 structure, will facilitate the translation of NAT10 inhibitors [177]. Notably, the causal relationship between nucleolar R-loop modulation and the observed clinical efficacy of RPA/NAT10-targeting agents remains unverified, necessitating dedicated mechanistic studies. Considering tumor hypoxia, the combination of R-loop modulators with PTEN/MYC-targeted agents (e.g., mTOR/AKT/PI3K inhibitors) should show synergistic potential. A Phase Ib trial of rapamycin–dasatinib–irinotecan–temozolomide combination therapy revealed superior neuroblastoma control with mTOR inhibitor/irinotecan co-administration [178]. This approach warrants systematic investigation of various combination strategies.

Centromeric R-loop targeting

CIN is a hallmark of cancer that can both promote tumorigenesis and induce tumor cell death. To survive, cancer cells must establish mitotic homeostasis. Since the late twentieth century, when scientists first proposed the spindle assembly checkpoint (SAC) concept, the development of CIN-exploiting therapies has remained a prime objective in oncology [179]. Guided by the rationale that CIN-phenotype tumor cells exhibit heightened vulnerability to further CIN induction, current therapeutic investigations are focusing primarily on lymphoma, acute myeloid leukemia, and colorectal cancer. Among these, Aurora kinase (AURK) family inhibitors represent the most extensively clinically investigated agents [180, 181]. However, direct inhibition of AURK members (e.g., Aurora B) raises concerns regarding potential malignant transformation in normal cells during chronic administration, necessitating the development of more precise targeting strategies. Centromeric R-loop modulation emerges as a promising approach to amplify therapeutic specificity by selectively destabilizing tumor cells, potentially redefining the therapeutic window for CIN-exploiting paradigms. Centromeric R-loop regulation strategies differ among tumor types. In centromeric R-loop-dependent tumors exemplified by prostate cancer, elevated centromeric R-loop levels correlate with poor prognosis [58]. Therapeutically reducing R-loop levels inhibits tumor progression. Conversely, in centromeric R-loop-independent tumors, inhibiting R-loop accumulation is essential for their survival. These tumors maintain mitotic homeostasis through the regulation of SAC components or other AURK members [182]. For these tumors, artificially elevating centromeric R-loop levels and concurrently targeting corresponding mechanisms represent correct strategies to induce cell death. Notably, elevated centromeric R-loops can drive resistance to enzalutamide in prostate cancer. Moreover, the ATR-CHK1-Aurora B pathway activation correlates with osimertinib resistance in lung cancer [183]. It remains unclear whether centromeric R-loops serve as upstream regulators in this process, warranting further investigation. Combined with centromeric R-loop suppression may prevent such resistance mechanisms [58].

H3K9 methyltransferase inhibitors represent leading candidates for targeting centromeric R-loops. EHMT2 regulates both nucleolar and centromeric R-loops: its activation promotes tumorigenesis, while inhibition induces cell death. Treatment with the EHMT2 inhibitor, BIX-01294, significantly reduces centromeric R-loop levels, activates CIN phenotype, and suppresses cell growth [58]. Currently, the most specific and potent EHMT2 inhibitors are BIX-01294 and UNC0642. BIX-01294 additionally suppresses EGFR signaling [184], whereas UNC0642 induces oxidative stress [185]. EHMT2 inhibitors overcome resistance to carfilzomib and PARPi [186, 187], while inducing apoptosis in bladder cancer and NSCLC cells [188–190]. However, the cytotoxicity of BIX-01294 and the poor bioavailability of UNC0642 have hindered their clinical translation. The novel EHMT2 inhibitor, MS152 (derived from UNC0642), has demonstrated superior bioavailability in vivo, positioning it as a top oral therapeutic candidate [191]. Notably, the classic antifolate drug raltitrexed demonstrates EHMT2 inhibitory activity. Structural modifications of this compound could expand its clinical indications [192]. Owing to its critical role in shaping the tumor immune microenvironment, SETDB1 has emerged as a promising therapeutic target in recent years [193]. However, selective inhibitors have not been developed to date. A structure-based study proposed a specific inhibitor targeting the SETDB1 tudor domain, further development of which remains highly anticipated [194].

CircR-loop precision therapy

Targeted therapies against circR-loop remain in their infancy, with small-molecule inhibitors targeting XPO4 and YTHDC1 currently limited to preclinical development. Significant advances have been achieved in augmenting tumor immunotherapy through the innate immunogenicity of circRNAs, while high-efficiency circRNA delivery via lipid nanoparticles has been established successfully [195]. Identification of circRNAs capable of broadly inducing R-loop formation in tumor tissues represents a pivotal challenge for therapeutic application. Rational design of sequence-specific circRNAs to spatially engineer R-loops for gene modulation, or development of entirely novel high-precision therapeutics based on this mechanism warrants in-depth investigation.

Challenges and opportunities

While these therapeutic strategies show initial promise, significant challenges remain to be addressed. First, safe and effective specific inhibitors for many targets, such as SETDB1 and SERBP1, are currently lacking. Although structural resolutions have been achieved for several targets, further drug development is necessary. Challenges in biomarker development persist for R-loop-targeted therapies, with current ALT tumor identification relying on empirical classification of tumor types exhibiting high ALT prevalence. Validated biomarkers for nucleolar or centromeric R-loop-targeted therapies do not exist. Potential direct biomarkers include tumor R-loop levels, quantified using DNA–RNA immunoprecipitation sequencing (DRIP-seq), while indirect biomarkers include telomere length heterogeneity, assessed using hybrid immunohistochemistry/quantitative fluorescence in situ, along with C-circle levels, APBs, and ATRX mutation status. Genetic approaches can be used to quantify CIN, and nuclear stress prediction models might be developed using whole-genome sequencing. Developing R-loop-specific antibodies or advanced imaging techniques may enable patient stratification for R-loop-targeted therapies. At the same time, these detection technologies should be based on the clinical methods of obtaining tissue, such as endoscopic ultrasound biopsy or intraoperative pathology [196–198]. Furthermore, the intricate R-loop regulatory network predisposes monotherapies to resistance. Elevated R-loop levels may trigger compensatory resolution mechanisms (e.g., upregulated RNase H1/SETX) or adaptive DNA repair. A clinical example involves a patient with pancreatic cancer treated with CX-5461 developing secondary mutations in BRCA2 that restored homologous recombination repair capability [145]. Therefore, R-loop-targeted therapies should be combined with other therapies to circumvent resistance. These challenges, compounded by incomplete understanding of R-loop biology in tumors, have severely limited clinical investigations into atypical R-loop targeting. We strongly advocate multidisciplinary collaboration among molecular biologists, clinical researchers, and pharmacologists to advance research on the targeting of non-canonical R-loop pathways, ultimately contributing to overcoming of these challenges.

Conclusions and future directions

This review summarizes the impact of atypical R-loops within the genome on tumor cell biology, along with their dynamic regulation in malignant cells. Atypical R-loops orchestrate critical oncogenic processes in tumor cells by modulating telomere length maintenance, nucleolar stress responses, and chromosomal segregation fidelity, thereby regulating the fundamental phenotypes associated with tumor cell survival and proliferation. Targeting of this regulatory nexus offers a strategic avenue for overcoming the inherent limitations of emerging therapies including telomerase inhibitors, nucleolar stress inducers, CIN-exploiting agents, and circRNA-based approaches, with particular therapeutic promise for ALT-positive malignancies, CIN-phenotype cancers, and other neoplasms with unmet clinical needs. In the future, improvements in the following areas remain highly desirable. First, developing novel and robust atypical R-loop biomarkers is essential for clinical translation. Integrating deep learning-based predictions with single-cell R-loop sequencing of clinical samples validation will significantly accelerate the discovery of new biomarkers. Second, developing novel atypical R-loop detection technologies is crucial. Emerging biomarkers such as RNA m6A modification could enable the development of liquid biopsy kits and biosensors for detecting pathological R-loop levels using blood, cerebrospinal fluid and other biofluid samples, both ex vivo and potentially in vivo. Moreover, combining novel agents targeting R-loops with current tumor therapies, either concurrently or sequentially, may help overcome resistance driven by R-loop plasticity. Clinical trial results are highly anticipated. Advancements in biomarkers, detection technologies, and therapeutic strategies will enable comprehensive integration of atypical R-loops into precision oncology practice. For diagnostic applications, DRIP-seq techniques can be employed to define cancer subtypes. Regarding therapeutic applications, R-loop intervention strategies can be implemented, with real-time monitoring of R-loop level variations in patients to assess therapeutic response. The systematic resolution of these scientific and technical barriers will catalyze the clinical translation of atypical R-loop-targeted strategies, ultimately redefining precision oncology through spatiotemporal control of R-loops.

Abbreviations

lncRNA

Long non-coding RNA

DNA2

DNA replication helicase/nuclease 2

EXD2

Exonuclease 3′-5′ domain containing 2

PCNA

Proliferating cell nuclear antigen

RFC

Replication factor C

TOP3A

DNA topoisomerase III alpha

RMI

RecQ mediated genome instability

FANCM

FA complementation group M

H3K27me3

Histone H3 lysine 27 trimethylation

CENP-A

Centromere protein A

CHK1

Checkpoint kinase 1

RNF-20

Ring finger protein 20

H2Bub

Histone H2B monoubiquitination

H3K4me2

Histone H3 lysine 4 dimethylation

SMARCA5

SNF2 related chromatin remodeling ATPase 5

EWSR1

EWS RNA binding protein 1

POLR2B

RNA polymerase II subunit B

RNase H1

Ribonuclease H1

SETX

Senataxin

DDX5

DEAD-box helicase 5

DDX17

DEAD-Box helicase 17

DHX9

DExH-box helicase 9

ILF3

Interleukin enhancer binding factor 3

TOP1

DNA topoisomerase I

TOP2A

DNA topoisomerase II alpha

PARP1

Poly(ADP-ribose) polymerase 1

HLTF

Helicase like transcription factor

TRF1

Telomeric repeat binding factor 1

TRF2

Telomeric repeat binding factor 2

DNMT3b

DNA methyltransferase 3 beta

METTL3

Methyltransferase 3, N6-adenosine-methyltransferase complex catalytic subunit

hnRNPA2B1

Heterogeneous nuclear ribonucleoprotein A2/B1

YTHDC1

YTH N6-methyladenosine RNA binding protein C1

RBMX

RNA binding motif protein X-linked

RBM14

RNA binding motif protein 14

XRN2

5′–3′ Exoribonuclease 2

RTEL1

Regulator of telomere elongation helicase 1

LSD1

Lysine demethylase 1A

RAD51AP1

RAD51 associated protein 1

THOC

THO complex

NONO

Non-POU domain containing octamer binding

SFPQ

Splicing factor proline and glutamine rich

NAT10:

N-acetyltransferase 10

TBPL1

TATA-box binding protein like 1

EHMT2

Euchromatic histone lysine methyltransferase 2

DAXX

Death domain associated protein

CUL3

Cullin 3

XPO4

Exportin 4

HIRA

Histone cell cycle regulator

UBN

Ubinuclein

SF3B1

Splicing factor 3b subunit 1

Author contributions

SSY and YF are the co-corresponding authors responsible for the overall conception, design, and revision of the review. SY is the first author responsible for collecting and analyzing information and drafting first manuscript. WS, GN and GJT participate in reviewing and editing the language. WGX is mainly responsible for drawing the charts and graphs in the review. All authors read and approved the final manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (No. 82470702 to Siyu S; No. 82100700 to Fan Yang), the High-quality Development Fund Project from the Science and Technology of Liaoning Province (No. 2023JH2/2020 0063 to Fan Yang), and 345 Talent Project of Shengjing Hospital (No. 52-30B to Fan Yang).

Data availability

Not applicable to this review as no datasets were generated or analyzed in this review.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.