Stability of non-canonical nucleic acid structure as a potential modulator of cell fate

Authors: Shuntaro Takahashi, Hisae Tateishi-Karimata, Naoki Sugimoto

Abstract

Cellular morphological changes occur during cell life and diseases, such as senescence and cancer. Although the cellular conditions should be varied with the morphology changes, there have been no attempts to understand the cellular morphological changes by focusing on the intracellular molecular environment and elucidating the behaviour of nucleic acids. Nucleic acids can form hierarchical secondary and higher-order structures due to intermolecular interactions and other factors. Additionally, a number of important discoveries indicate a link between the effects of intracellular cations, hydration, and metabolic products on the stability of nucleic acid structures and diseases, such as cancer. Thus, changes in gene expression by environments can trigger morphological changes in cells. To elucidate the mechanisms of intracellular gene expression governed by nucleic acid behaviour, it is extremely important to analyse the stability of nucleic acid structures in the whole cell or local cellular spaces by manipulating the actions of small molecules, such as cations, water, and metabolic products. This review article describes the research background and latest progress in controlling senescence and cancer by modulating gene expression based on the prediction of intracellular nucleic acid behaviour, with a focus on the effects of cations, hydration, and metabolites on intracellular nucleic acid structures and their stability.

Introduction

Cells have different morphologies and functions, yet they work together to maintain biological processes. Cells establish these advanced life systems by dividing and differentiating from embryonic cells into muscle, nerve, and epithelial cells. The breakdown of these systems leads to cellular senescence and cancer. Cell morphology dynamically changes during the cell cycle, from differentiation to apoptosis. The morphology of the organelles, such as the nucleus, mitochondria, and endoplasmic reticulum, also undergoes alterations. At the molecular level, changes in cell morphology involve the remodelling of the cytoskeleton components, such as F-actin, tubulin, and vimentin. Growth factors and stress activate signalling cascades that produce specific proteins mediating these structural changes. Changes in these morphologies are based on the expression of different genes in each cell type. Therefore, gene expression and cell morphology have recently become important high-dimensional biological indicators in drug discovery [1, 2]. To investigate these cellular morphological changes, transcriptomic analyses have traditionally been the cornerstone of assay technologies, evolving from Sanger sequencing and microarray analysis to more recent advancements, such as RNA-seq and its derivatives [3]. Recent developments have enabled the integration of morphological features and spatially mapped transcriptional data to characterize cells and tissue regions [4]. Additionally, proteomic approaches have been adopted to classify distinct cell states using proteomic profiles defined by known and uncharacterized proteins [5, 6]. According to these recent technological advances, the relationship between gene expressions and cell morphology has been clarified [1].

For cellular proliferation and function, certain genes must be correctly expressed spatiotemporally in response to environmental changes. All cells have identical genomic information, meaning that they form different patterns of gene expression without altering their DNA sequences. Therefore, the expression of specific genes is fundamental to changes in cellular morphology. Moreover, the regulation of these genes can be used to induce the cellular morphology of interest. For example, one representative technique is the formation of the induced pluripotent stem (iPS) cells by introducing four key Oct3/4, Sox2, Klf4, and c*-Myc* [7]. The most important regulation of gene expression occurs when deciding the genes to be transcribed from the genome. One of the mechanisms that determines the timing of gene expression is regulated by the methylation [8, 9]. Methylation along transcription promoter regions recruits methylated DNA-binding proteins and inhibits the binding of transcription factors, resulting in reduced gene transcription. Reprogramming during iPS induction involves a dramatic change in the epigenomic pattern underlying gene expression from a somatic to pluripotent stem cell type [7, 10, 11]. Furthermore, eukaryotic cell nuclei contain histone nucleosomes in which DNA is condensed and unwound before transcription [12]. Acetylation and phosphorylation of histones regulate the stability of the nucleosome and act as transcriptional regulators [13]. These gene regulation systems are known as epigenetics (Fig. 1). These chemical modifications can be inherited via cell division, which irreversibly regulates gene expression levels. Changes in cell morphology can be accompanied by the formation of a cell-specific epigenome, with dynamic changes in epigenomic patterns through cell proliferation and differentiation.

In contrast, non-epigenetic gene regulation has been widely observed in cells [14–19] (Fig. 1). Non-canonical DNA structures are important units for differentiating gene expression patterns that do not change genetic information or produce epigenetic modifications [20, 21]. Nucleic acid replication relies on the basic chemical principle of base pairing between nucleobases. In 1953, Watson and Crick described the DNA structure as a right-handed double helix with two strands, in which there was a model based on the observed abundance of DNA bases and structural requirements for an antiparallel arrangement [22]. The DNA and RNA strands form duplexes via Watson–Crick (WC) base pairs with adenine (A) and thymine (T) or uracil (U) in RNA, along with those between guanine (G) and cytosine (C) through hydrogen bonding (Fig. 2). Based on complementarity, genetic information can be accurately replicated and expressed in cells. In addition, nucleic acids are capable of forming base pairs (Fig. 2), where adenine (A) pairs with T or U and G pairs with C^+^ (protonated C), resulting in diverse configurations of nucleic acid structures [23, 24]. Triplexes are formed through Hoogsteen base pairing by the third strand of the WC-type duplex [25, 26]. Moreover, four guanine bases are stabilized via Hoogsteen hydrogen bonding, whereas i-motifs are formed from cytosine-rich sequences via intercalating hemiprotonated C–C⁺ base pairs under slightly acidic conditions (Fig. 2) [27]. These structures are formed in an inter- and intramolecular manner. In the 2000s, significant advancements were made in G4 structure research, highlighting the pivotal role of Hoogsteen base pairs [28]. After the identification of G4 in the cell genome, it was understood that the formation of G4 structures in both DNA and RNA affects the reactivity of proteins that play a role in gene expression [29]. More importantly, potential G-quadruplex-forming sequences (PQSs) are frequently found in the promoter regions of cancer- and neurodegenerative disease-related genes, which regulate the expression of these genes and cause lethal effects on cells [30, 31]. Nucleic acids are increasingly understood for their roles, which are defined by the physicochemical properties of their base canonical structures with WC base pairs store genetic information, whereas non-canonical structures formed by Hoogsteen base pairs control dynamic gene expression (Fig. 3) [21]. Non-canonical DNA structures can also regulate gene expression without changing the genetic information coding for DNA sequences or epigenetic modifications [21]. However, this regulation is reversible by the equilibrium formation of non-canonical structures in comparison with the irreversible mechanism of epigenetic modification. Therefore, non-canonical DNA structures may play an important role in the dynamic and transitional regulation of gene expression, which can determine cell fate.

Then, what dominates the equilibrium shift from duplex to non-canonical structures? Although many DNA-binding proteins are involved in formation, the chemical environment is a fundamental key to stabilizing non-canonical structures specifically. G4 structures form stably in the presence of K^+^ because they coordinate metal cations between the layers of the quadruplex [32], whereas iMs favour mildly acidic conditions because of the hemi-protonation of cytosine base pairs [33]. Moreover, the conditions within cells are both diverse and densely packed because of the intracellular components of bio(macro) biomolecules at concentrations ranging from 50 to 400 g l^−1^ [34]. This state, referred to as molecular crowding, has a profound impact on the conformation and stability of biomolecules including DNA [20]. Therefore, the chemical environment may play a key role in the regulation of gene expression via non-canonical DNA structures. Cellular conditions are highly heterogeneous, and DNAs exist in various environments. The nucleus that accommodates genomic DNA in cells consists of a liquid–liquid phase-separated (LLPS) region called the nucleolus, where the transcription of ribosomal RNA specifically occurs [35]. Mitochondria contain their own DNAs (mitochondrial DNA: mtDNA), which encode several mitochondrial machinery genes. It has been suggested that the physicochemical environment of mitochondria is different from that of the nucleus [36, 37]. In plants, chloroplasts also contain DNA [38]. Notably, PQSs are found in ribosomal DNA and mtDNA [39, 40], and cruciform structures are found in various regulatory regions of chloroplast DNA [41]. Moreover, localized regions in cells as well as the dynamics of intracellular environments, such as cell cycles and differentiation, should affect nucleic acid behaviours. One of the most drastic changes driven by cell morphology is the intracellular environment. Considering that changes in cell morphology are accompanied by environmental changes in cells, the involvement of non-canonical DNA structures in cell morphology requires investigation. Notably, the functions of ion channels, aquaporins, and metabolite transporters in senescent and cancerous cells significantly differ from those in normal cells by creating unique intracellular environments [42–45]. Moreover, it is possible that small molecules, such as water and metabolites, bind to DNA, alter its stability, and regulate gene expression. Therefore, unique environments potentially regulate gene expression via the formation of non-canonical DNA structures, depending on the spatiotemporally defined molecular environments, which lead to cellular behaviour and destiny.

From these backgrounds, one hypothesis the stability of DNA within a small window links to cell fate, including the progression of cell senescence and cancer. To prove this concept, physical chemistry contributes to elucidating the roles of non-canonical DNA structures in regulating gene expression levels and altering cell morphology by quantifying DNA stability and conformation. As chemical interactions are relatively easier to understand and regulate from a physicochemical viewpoint than epigenetic modification factors, we are now at the research stage of quantitatively elucidating how small molecules, such as ions, water, and metabolites regulate nucleic acid structures for cell morphology changes. Here, we review the recent advances in knowledge related to the formation of non-canonical DNA structures that can control cellular morphology. This review provides a unique concept of how cations, hydration, and metabolites affect nucleic acid structures, stability, and function in intracellular environments, which implies the links to the regulation of cell fate like senescence and cancer.

Conceptual: significance of the stability of nucleic acid structures

There is an equilibrium between a duplex and a non-canonical structure; therefore, the structural stability of both structures determines directly the importance of their structures. The stability of DNA structures is governed chiefly by (i) hydrogen bonds between nucleobases, (ii) stacking interactions between neighbouring base pairs, (iii) energetic costs associated with conformational entropy, (iv) condensation of counterions by cations, (v) effects of hydration, and (vi) interactions with cosolutes. Of these, (i–iii) are intramolecular bulk interactions that stabilize various DNA forms, whereas (iv–vi) represent molecular–environmental interactions that can enhance the stability of DNA, contingent on the geometry of the structures in question. There are four types of nucleobases and different base pairings, such as WC and Hoogsteen ones, indicating that the stability factors listed above depend on the sequence and secondary structure. To assess duplex stability, a method was developed to predict the thermodynamic parameters of duplexes in vitro without experimental data. This technique is known as the nearest neighbour (NN) method, and the NN parameters are commonly employed to estimate the thermal stability of duplexes. Introduced by Tinoco et al. in 1971 [46], the NN model is based on the notion that the stability of a nucleic acid duplex is substantially influenced by its neighbouring base pairs. This is because the strength of hydrogen bonds between base pairs depends on the specific base pairing. In contrast, the energy from stacking interactions, driven by London forces, diminishes in proportion to the sixth power of the distance. As a result, interactions involving base pairs beyond immediate neighbours were considered negligible. Based on this principle, the stability of a duplex is primarily determined by cumulative interactions between adjacent base pairs.

A duplex is a long, static context for storing information. Human genomic DNA consists of long DNA duplexes, containing approximately three billion base pairs. On the other hand, gene expression requires reading the information in DNA sequences, which partially unwinds the duplex. Thus, the fundamental basis of gene expression depends on the physicochemical properties of short sequence segments. This idea had been attempted since the development of NN parameters. For example, initiation, elongation, and termination of the transcription processes within genomic DNA depend on the energetics of the transcription bubble for various transcription factors [47]. The transcription bubble is generally 15–18 bp long, depending on the operative polymerase. The energy of the transcription bubble is calculated by adding the free energies of the constituting nearest-neighbour pairs and initiation factors. Thus, transcriptional sequence dependency can be generated by the thermodynamics of the DNA sequence. In Saccharomyces, the stability survey with a varying window size (<100 bp) by NN parameters revealed that the coding regions are more stable than other parts, such as introns and 3′-untranslated regions [48]. The open reading frames have more stable sense mRNA/DNA duplexes than potential antisense duplexes, which could enhance transcript formation. Furthermore, the stability of the 3′-untranslated region showed a strong negative relationship with the mRNA level. In the initiation of transcription, which is a rate-limiting step of the transcription reaction, the kinetics of processivity of the bacterial RNA polymerase from the transcription initiation site (promoter escape) depends clearly on the predicted stability of DNA/DNA and RNA/DNA of these sequences [49]. Interestingly, the transcriptome analysis in Archaea cells indicated that the promoter escape, which changed upon oxidative stress, did not correlate with the accumulation of transcription factors around various promoter regions having different strengths, but was influenced by the predicted DNA stability across the transcription start sites [50]. Moreover, the torsional stress caused by negative supercoiling destabilizes the duplex within several to tens of base pairs [51], suggesting that the unwinding process by polymerases, helicases, and topoisomerases can be affected directly by the stability of DNA.

On the other hand, the main modulators in gene regulation are proteins. Chromosomes in genomic DNA contain histone proteins that tightly pack the DNA, preventing gene expression. As protein binding bends the duplex, the local stability of the duplex may, in turn, affect binding. There have been attempts to interpret the behaviour of proteins binding to such DNA in terms of DNA stability [52]. However, comprehensively addressing actual cellular phenomena proved difficult. This is because cellular phenomena, including protein interactions, are complex, and it was not considered that stability alone could explain them. Meanwhile, recent machine learning approaches have begun utilizing DNA stability databases as a powerful resource. For instance, predicting transcription factor binding sites could be refined by incorporating predicted DNA stability into the algorithm [53]. This suggests that living systems may be utilizing DNA structural stability information as a dynamic code to maintain their dynamic biological phenomena [54]. Therefore, it can be fundamental to analyze the stability of DNA in short segments for gene expression.

Non-canonical structures, consisting of relatively short sequences of ~20 base pairs, provide a clear example where function depends on stability more than in duplexes. That is, the formation of these structures directly involves the thermodynamics governing the melting of duplexes in these short regions and the formation of non-canonical structures. Of course, the formation of non-canonical DNA structures requires the physical organization of genomic DNA. G4 formation is strongly affected by the state of chromatin-bound DNA [55]. The chromatin structure is relaxed in euchromatin regions, where transcription levels are high, thus facilitating G4 formation. In contrast, DNA is strongly condensed in heterochromatin, hindering the formation of the G4 structure. Moreover, long duplexes must partially unwind to generate single-stranded regions required for G4 or iM formation. Negative supercoiling during duplex unwinding promotes the formation of non-canonical structures. For example, RNA polymerase creates a transcription bubble to unwind the DNA duplex and read the DNA template sequence during transcription [56]. The nascent RNA hybridizes with the template DNA, and the single-stranded complementary DNA in an R-loop can form G4s. This transcription-induced supercoiling is dynamically modulated throughout the cell cycle [57]. The supercoiling magnitude is regulated by topoisomerase, which directly interacts with G4 structures [58]. Topoisomerase expression varies in response to cellular conditions during different phases of the cell cycle [59]. In the case of the human Myc gene, a major cancer-related gene, expression is regulated dynamically by DNA supercoiling, followed by non-canonical structure formation, including G4s [60, 61]. Thus, transcription-induced mechanical forces dynamically regulate gene expression by forming various DNA structures.

On the other hand, there is substantial direct evidence that the stability of G4 regions is regulated by gene expression. The first report is about the expression level of the oncogene Myc downregulated by targeting the G4 in the promoter region of the gene with the porphyrin-based TMPyP4 ligand [62]. Other oncogene promoters, such as c-Kit, Bcl-2, KRAS, and VEGF, are also reactive to downregulate the expression in response to the ligand [63]. Recently, it has been known that the stabilization of G4 by the ligand can also induce chromatin remodelling [64]. These findings indicate that G4 stability can compete with chromatin modelling, with effects comparable to the physical organization mentioned above. The physical behaviours of genomic DNAs are key determinants of non-canonical structure formation; therefore, G4 and iM formation can be controlled by the chemical factors that stabilize or destabilize both the duplex and G4/iM within these regions. Non-canonical structures are more sensitive than duplexes to the environmental factors in the solution. Therefore, the chemical environment within the cell changes with changes in cell morphology, potentially affecting the local DNA sequence and inducing the formation of non-canonical structures. To understand the behaviour of such DNA structures, current knowledge of DNA stability is based on test-tube conditions, such as 1 M NaCl. To link the behaviour of DNA from the test tube to the intracellular condition, we summarize the dependence of the solution environment on the DNA stability in the following sections.

Solution environments affecting the formations of DNA structures

The solution environments in cells are spatiotemporally variable owing to the uptake and output of cations and water, which induce changes in pH, metabolites, and molecular crowding affected by the compartment space (Fig. 4). Here, we explored how each individual contribution influences the stability of both canonical and non-canonical structures.

Cation effect

Nucleic acids are highly negatively charged polyelectrolytes that are thus surrounded by numerous cations (counterions), which neutralize the charge of the DNA via counterion condensation [65, 66]. Thus, duplex stability depends on the cation concentration [67]. Polyvalent cations stabilize the secondary structure of DNA more than monovalent cations according to counterion condensation theory. Moreover, the charge density of DNA and the surrounding ionic environment determine the physical constraints on higher-order structures of cellular DNA, such as superhelical structures and nucleosome entrapment [68]. Cation concentrations fluctuate in living cells. Intracellular concentrations and content of Na^+^ and Cl^−^ were reduced in the G0–G1 phase transition during the cell cycle, followed by an increased content of both ions in the S phase concomitant with water uptake, although K^+^ was unchanged during cell cycle progression [69, 70]. It is commonly believed that the typical intracellular concentrations of Na^+^ and K^+^ are around 10 and 140 mM, respectively [71, 72]. Additionally, the free Mg^2+^ concentration within cells fluctuates between 1 and 3 mM [72, 73]. The concentration of Ca^2+^ in cells is generally reported to lie between 3 and 10 µM [72]. These values are approximations based on the overall salt content within cells; however, the local concentrations of these cations may vary across different cellular compartments [74]. Furthermore, depending on the specific cell type and state, cation concentrations can also differ [75]. As nucleotides assemble into helices, substantial electrostatic repulsion occurs because of the negative charges surrounding the phosphate groups (Fig. 5). To reduce this repulsion, counterions are condensed around the phosphate groups, forming higher-order structures linked to cation binding. While the extent of electrostatic repulsion may vary depending on the tertiary structure, the role of cations in neutralizing negative charges remains consistent in each nucleic acid structure. Thus, the balance between the concentration of each cation is an important factor in determining the equilibrium between the canonical duplex and non-canonical structures.

Duplex stability is highly sensitive to cation concentration. Thus, basic analyses in physical chemistry have usually been performed in a solution containing 1 M NaCl [76], which shows a drastic increase in stability compared with solutions containing a range of physiological cellular concentrations of cations. Under non-crowding conditions, the contribution of Na^+^ to DNA duplex stability is almost the same as that of K^+^, whereas divalent cations, such as Mg^2+^, effectively stabilize the duplexes [77]. A 10 mM Mg^2+^ solution provides nearly the same stabilization to a six-base pair duplex as a 1 M Na^+^ solution [78], which is much greater than that expected based on the ionic strength contributed by magnesium ions alone. The actual intracellular environment contains a mixture of cations, with concentrations typically ~140 mM K^+^, 10 mM Na^+^, 0.5 mM Mg^2+^, and 0.0001 mM Ca^2+^ [79, 80]. These values were rough estimates derived from the total salt content of the cells. Consequently, local concentrations of these cations significantly vary across cellular organelles [74]. Additionally, cation levels may fluctuate depending on the condition and type of cell [75]. Mg^2+^ competes with monovalent cations, therefore, duplex stability is substantially affected by fluctuations in the concentration of Mg^2+^ in the cells. The duplex stability can be estimated by NN parameters, which have been widely used to predict the thermodynamic parameters of duplexes introduced by Tinoco et al. in 1971 [46] The original NN parameters were determined in a solution containing 1 M NaCl, and several equations were developed to calculate duplex stability in solutions with varying Na^+^ and Mg^2+^ concentrations based on extrapolation from the stability observed in 1 M NaCl [77, 81, 82]. This correction enabled the prediction of duplex stability under various solution conditions with different concentrations of Na^+^ and Mg^2+^.

In addition to metal ions, molecular ions uniquely interact with nucleic acids in the cellular environment. The 2-hydroxy-*N,N,N-*trimethylethanaminium (choline) ions and their derivatives are abundant in cells, and involved in metabolism and DNA methylation [83]. At high concentrations, choline ions, which exist as hydrated ionic liquids, make the A•T base pair more stable than the G•C base pair in DNA duplexes [84]. This is due to the strong binding of choline to the ribose in T, which stabilizes the A•T pair. In contrast, choline interacts with both G and C bases, resulting in an unfavourable G•C pair [85]. Additionally, in the case of G4, choline ions weaken stability by favouring binding to the single-stranded G sequence over a stable G4 configuration [86]. In contrast, choline ions selectively bind to bases in the loop region of iM DNA, significantly stabilizing its structure [86]. Other molecular ions, such as trimethylamine N-oxide (TMAO) and trimethylammonium, preferentially bind to the coil form and destabilize the duplex [87, 88]. However, at lower concentrations, TMAO specifically bound to the groove and enhanced the duplex stability [89]. Bulky cations exhibit unique binding to nucleic acid structures. Tetrabutylammonium and tetrapentylammonium ions stabilize hairpins with long loops [90]. This suggests that these cations stabilize G4 structures with extended loops. Consequently, larger cations appear to favour more flexible loops as they facilitate access to loop nucleotides. These findings highlight the dynamic modulation of non-canonical nucleic acid structures by cations in the cellular environment.

The G4 structure features a specific and vital binding site for cations in its core. These cations interact with the O6 atoms of each guanine group and are positioned between the two planes of the G-quartet (Fig. 5). Such coordination is crucial for the proper folding of G4 structures, as the absence of a coordinated cation at the quartet centre makes the structure electronically unstable and unfavourable [91, 92]. Research has examined the role of cations in stabilizing G4 structures, suggesting that their stabilizing effect follows the K^+^ > Ca^2+^ > Na^+^ > Mg^2+^ > Li^+^ [93]. While Sr²⁺ and Ba²⁺ also contribute to G4 stability, physiologically relevant cations, such as K^+^ and Na^+^, notably influence G4 stability in distinct ways. The preference of the G4 cavity for K^+^ over Na^+^ stems from the cation size and energetic penalty of dehydration upon binding. Na^+^ binds more stably to the G4 cavity than K^+^ (with Na^+^ being favoured by −1.7 kcal mol^−1^) [94]. However, the dehydration cost of Na^+^ prior to coordination with the G4 cavity was much higher than that of K^+^ (Na^+^ is unfavoured by >30 kcal mol^−1^). The net result of these two opposing factors is a free-energy change that ultimately favours K^+^ ions [95, 96].

Compared to duplexes and G4s, iMs are less sensitive to cation concentrations or show destabilization with increasing cation concentrations [97–99]. Li^+^ promoted the formation of synthetic iM structures more effectively than Na^+^ or K^+^ at lower concentrations (<200 mM). This suggests that Li^+^ is more efficient than other monovalent cations in facilitating iM formation, although at higher concentrations (>200 mM), Li^+^ strongly destabilizes the same structure. This dual effect of cations on folded DNA structures can be explained by the two opposing actions of monovalent (i) the reduction of electrostatic repulsion, which also increases the flexibility of a single stranded DNA (ssDNA), making it easier to fold, and (ii) the disruption of hydrogen bonds between cytosine bases. Li^+^ stands out in both of these effects, likely because of its smaller size. The iM formed from the human promoter region of the N-Myc gene exhibited a similar trend, with stability dependent on the ionic strength [100]. In contrast, Mg^2+^ stabilizes the iM structure [101], indicating that this divalent cation may favour binding of the unique geometry of iM via intercalated C*C^+^ pairs and interactions within the inner loop or between loops. The differing ionic responses of the iM and G4 structures suggest that, depending on intracellular ionic fluctuations, the GC-rich regions of genomic DNA can preferentially form either an iM or G4 structure, depending on the surrounding solution environment. The effects of physiological cations with cosolutes on nucleic acid stability and structure are summarized in Table 1. These findings suggest that fluctuations in the monovalent and divalent cations in cells differentially affect the stability of canonical and non-canonical structures.

pH effect

DNA stability is affected by pH depending on the pKa values of the nucleobases [102]. Cellular pH fluctuations occur in different localized regions, vary according to cell cycle stages, and differ across various cell types, including senescent and cancerous cells. The pH inside the cell (pHi) fluctuates throughout the cell decreases at G1/S, increases at mid-S, decreases at late S, increases at G2/M, and decreases during mitosis [103]. Spatiotemporal pHi dynamics are essential for proper cell cycle timing and transition. Senescent model cells with decreased expression of ATPase H^+^ transporting accessory protein 2 trigger intracellular acidification (pHi decrease) and lysosomal alkalinization [104]. Moreover, cancer progression also affects pHi, such as that in metastatic triple-negative breast cancer cells (MDA-MB-231), which was higher (7.52) compared with that in matched normal breast epithelial cells (MCF10A) (7.23). Hence, dysregulation of these dynamics may contribute to diseases, such as cancer, in which pHi is often elevated. The localized pH values in the cells were also different from each other. Within intact mitochondria, the pH is maintained at an alkaline level through the action of the electron transport chain, with the resting pH values of mitochondria ranging from 7.2 to 8.2 depending on the cell type [105]. Thus, the effect of pHi on nucleic acids must be considered in localized areas of the cells at different times and under different conditions.

The stability of canonical duplexes generally shows minimal pH dependence, except under strongly acidic or basic conditions, as protonation and deprotonation of base pairs rarely occur. This is also true for G4 structures, where the stability of Hoogsteen base pairs remains unaffected by pH. In contrast, certain non-canonical structures involving protonated cytosines exhibit a clear pH dependence. The formation of triplex and iM structures relies on C·G·C^+^ and CC^+^ base pairs (Fig. 6), with their stability being significantly influenced by pH. Cytidine has a pKa of 4.2, indicating that triplex and iM structures are more stable in acidic conditions. In the case of the triplex, the stability of the C·G·C^+^ unit increases by approximately −1 kcal/mol for each unit decrease in pH [106]. The iM structure demonstrates a linear relationship between melting temperature and pH, with a change of −23.8 kcal/mol per pH unit [107]. These structures were unstable under neutral pH conditions. However, chemical interactions can modify the pKa values and alter the pH dependence of these structures. For instance, the choline cation, as mentioned earlier, increases the pKa value, allowing iMs to form stably even at neutral pH [86]. Moreover, the surrounding molecular environment can significantly influence the protonation state of cytosine [108, 109], and certain chemical modifications of the cytosine base can shift its pK*a [110, 111]. Therefore, the effect of pH on non-canonical structures serves as a dynamic mechanism that modulates their formation within living cells. Thus, pH fluctuations in cells occur in different localized regions in the cell, different cell conditions during the cell cycle, and different cell types, such as senescent and cancer cells, which can drive the regulation of the structural formation of the triplex and iM.

Metabolite effect

In addition to its ionic contributions, the cosolute has the potential to regulate the stability and structure of DNA. Various organic molecules, such as metabolites, are present under various cellular conditions. The most-studied metabolites that interact with nucleic acids act as chemical agents to modify nucleobases. Endogenous metabolites, such as aldehyde derivatives and reactive oxygen species (ROS), cause DNA–protein cross-links that change the stability and structure of nucleic acids and hinder the biological processes of DNA [112–114]. Acetyl-CoA and S-adenosylmethionine are substrates for the modification of DNA and histone proteins for the epigenetic regulation of gene expression [115]. RNAs have a variety of covalently modified nucleotides compared to DNA, and their roles in cells are of interest. N^6^-methyladenosine (m^6^A) is the most frequently modified mRNA and is also found in lncRNAs [116]. Notably, m^6^A destabilizes duplex A–U base pairs and disrupts base stacking in certain motifs [117, 118], resulting in structural changes to the secondary structure of RNAs [119].

In addition to these metabolite effects via the covalent modification of nucleic acids, other metabolites can interact through non-covalent modifications with nucleic acids. As shown in Fig. 1, non-epigenetic control by non-canonical structures has recently been elucidated (Fig. 7) [19, 21]. Porphyrin derivatives, including N-methylmesoporphyrin IX (NMM), are G4 ligands. Thus, natural porphyrins, such as protoporphyrin IX (PplX) and hemin can bind to the G4 DNA with high affinity in the KD range of 1–3 µM at 25°C [120]. Nucleotide metabolites also interact with non-canonical structures. Among them, a guanine metabolite, cyclic guanosine mononucleotide (cGMP), fills gaps in the G-vacancy-containing G4 (vG4) structure, which is unstable and contains three guanines, to stabilize the G4 structure with 10–1000 µM KD values [121, 122] and regulate the gene expression [123]. The vG4s can be formed by the oxidation of one of the guanine bases in G4, therefore, the binding and consequent stabilization of oxidized vG4s by guanine metabolites may indicate an environment-responsive regulatory role in physiological and pathological processes [124, 125].

The intracellular concentrations of these metabolites directly affect the regulation of non-canonical DNA structure formation. PpIX and hemin are present across cell lines (PpIX: ∼0.5–1.2 μM; ∼1.5–12 μM) with the working range considered from the KD values above [120]. However, the physiological level of cGMP concentration is believed to be <5 μM, which is lower than the range of the KD values above [126]. Moreover, the basal choline concentration was reported to be 100–400 µM, showing a large working concentration range for G4 and iM stabilizations [127]. These studies suggest that spatiotemporal fluctuations in the local concentration of metabolites may trigger the formation of non-canonical structures according to stress signals. Actually, as discussed later [128], G4 formation near a liposome membrane containing phosphatidylcholines was unfavourable, which could be a possible mechanism for the role of the metabolite in the regulation of G4s by the local environment in cells, such as the surface of the cell membrane.

In addition to human cells, plant cells contain many unique secondary metabolites that are the byproducts of primary metabolism. Flavonoids are unique plant secondary metabolites (PSMs). They have a planar aromatic ring that can bind and stabilize nucleic acids, and show the binding to G4 and its stabilization. Notably, the addition of glycosidic daidzin or genistin in the presence of 100 mM K^+^ makes the melting temperature (Tm) value of human telomere G4 increased with + 2∼3°C but the Tm of the corresponding duplex decreased with –2∼4°C [129]. The notable feature of flavonoids is the sequence (structure) specificity of their binding, which depends on the difference in the modified groups, despite the similarity of their chemical structures. Quercetin favours binding to VEGF G4-DNA more than to human telomere, G4, and its duplexes [130]. In contrast, fisetin shows a strong preference for binding to human c-Myc G4-DNA, whereas naringenin exhibits a greater affinity for duplex DNA than G4 structures. This suggests that the planarity of the C-ring of flavonoids plays a pivotal role in the selective recognition of G4-DNA [131]. These findings indicate that kaempferol exhibits a strong preference for binding to VEGF G4-DNA over other G4 sequences and duplex DNA, and significantly enhances the thermal stability of VEGF G4-DNA. In contrast, morin showed a notably weaker interaction with both the duplexes and various G4-DNAs, lacking clear structural specificity. These contrasting binding profiles highlight the critical role of the 2′-OH group on the B-ring of the flavonol moiety in influencing DNA interactions [132].

Fisetin also has a unique specificity for iM DNA from VEGF, causing a structural transition to a hairpin-like structure and releasing the replication stall by the VEGF iM sequence [133]. Recent studies have delved into the selective interactions between flavonoids and iM DNA structures, which are crucial for oncogene regulation and cancer treatment. These studies highlight how subtle structural variations in flavonoids—such as fisetin, morin, quercetin, and naringenin affect their binding affinities to different iM DNA structures, particularly HRAS1 and HRAS2. Fisetin and morin specifically interact with HRAS1 and HRAS2 iM DNA, respectively, suggesting their potential as target ligands for modulating gene expression in cancer cells [134]. Quercetin demonstrates a strong preference for HRAS1 iM DNA, making it a promising natural antagonist of this structure [135]. Naringenin also preferentially binds to HRAS2 iM DNA, further supporting its potential for anticancer drug development [136]. These findings provide new insights into the mechanistic aspects of flavonoid–DNA interactions, offering promising avenues for designing therapeutics aimed at regulating gene expression and combating cancer. In other examples of PSMs, curcumin stabilizes c-Myc G4 and represses its overexpression in the human cancer cell line, MDA-MB-231 [137]. Curcumin also binds to human telomere G4 DNA with the estimated apparent binding constant value of Kb = 4.38 × 10^5^ M^−1^ at 25°C and destabilizes the structure, whereas curcumin has less affinity to thrombin-binding aptamers with Kb = 0.62 × 10^5^ M^−1^ and stabilization effect due to different topology of the G4 structure [138]. These reports suggest that plant cells use metabolites to regulate the functions of certain non-canonical structures (Table 1), and it is possible that various metabolites with potential roles in controlling non-canonical structures and their functions can be identified in human cells.

Molecular crowding effect

Molecular crowding affects the stability of biomacromolecular structures via cosolutes. The molecular environment influences the diffusion rate, hybridization efficiency, folding, structural stability, and interactions of biomolecules [139–141]. According to scaled particle theory, the folding of compact structures in a crowded environment is entropically favourable (stable) because these structures occupy a smaller volume [142]. The characteristics of polymers in the cosolute, such as size, water activity, and dielectric constant, widely vary [75, 143]. Consequently, different polymers and proteins exert different effects on nucleic acids [18, 87, 144, 145].

Molecular crowding induces drastic changes in physicochemical factors in the intracellular environment. In particular, during the cell cycle, molecular crowding in the nucleus significantly increases during the M phase compared with that during the G1 and S phases [146]. This is attributed to the inflow of biomolecules, such as lipids and proteins, from the cytoplasm when the nuclear membrane dissolves during mitosis. In contrast, crowding in the cytosol remains relatively constant throughout the cell cycle, maintaining molecular density via osmotic pressure [146]. The molecular crowding effect is caused by highly concentrated biomolecules and small molecules acting as cosolutes of nucleic acids in the intracellular solution. The cosolute, a molecule inert relative to nucleic acids, can influence the physicochemical properties of the solution and affect the stability of DNA, depending on the size of the cosolute (Fig. 8). These cosolutes altered the physicochemical characteristics of the solution environment (Table 1). The stability of nucleic acid structures is modified by crowder (cosolute) effects, such as changes in water activity and the excluded volume effect [18, 145]. Under conditions of molecular crowding in vitro, canonical duplexes (such as DNA and RNA) formed with Watson–Crick base pairs and relatively long polymer chains exhibit greater stability than under non-crowded conditions, primarily because of the excluded volume effect [147]. Earlier studies have shown that the Tm of the polyinosine•poly(C) RNA duplex increases by 1.8–2.0°C with 10 wt% polyethylene glycol (PEG) 4000 (average molecular weight 4000) and PEG 20000, and 0.7°C with 10 wt% dextrans (average molecular weights 10 000 and 70 000) [148]. Likewise, PEGs stabilize long DNA duplexes as observed in the case of poly(dA)•poly(dT) showing an approximately 5°C increase of Tm with 19 wt% PEG 8000 [149]. Stabilization occurs because long base pairs occupy considerable volume. Conversely, the impact of water activity on duplex stability was opposite to that of the excluded volume effect. Molecules, such as, ethylene glycol (EG), glycerol, acetamide, and sucrose, which are small molecules and causes less excluded volume effects, reduce the Tm of poly(dA)•poly(dT) and Escherichia coli DNA by 1.0–5.0°C, which has been attributed to reduced water activity [145, 149]. These studies demonstrate that the stability of canonical duplexes under crowding conditions is primarily governed by water activity and excludes volume effects [150, 151].

For shorter DNA duplexes, the impact of water activity on duplex destabilization was particularly pronounced. The typical examples indicate that a 20 wt% concentration of EG, PEG 200, and PEG 1000 results in a reduction of Tm for DNA duplexes (8−30 mers) by between 15.9°C and 1.6°C [152]. In addition to PEGs, acetonitrile, 1,3-propanediol (PDO), and 1,2-dimethoxyethane also effectively lower water activity and destabilize duplexes [150, 152]. In contrast, larger cosolutes such as PEG 8000, dextran, and Ficoll increase duplex stability via the excluded volume effect [145, 149].

Interestingly, the effects of molecular crowding on duplex stability highlight sequence-specific interactions involved in cation binding [153]. Although the effects of Na^+^ and K^+^ on duplex stability were similar for unbiased sequences, Na^+^ stabilized the duplex to a greater extent than K^+^ in GC-biased sequences. In contrast, the AT-biased sequences showed similar stability for Na^+^ and K^+^. Moreover, the cosolute quantity affected the hydration-induced stability of biased DNAs. Thus, biased DNAs shows different dependences of duplex stability induced by groove hydration on the chemical composition of the solutions [153]. These biased sequences have the potential to form non-canonical structures, such as G4s, iMs, and triplexes, which highlight their roles in cellular processes. For instance, they are involved in cation-dependent quadruplex formation in cancer-related genes and regulation of replication initiation through fluctuations in intracellular crowding.

Unlike canonical duplexes, triplexes and tetraplexes formed by Hoogsteen base pairs are stabilized in crowded environments [149, 154]. The impact of molecular crowding on non-canonical structures was initially observed during the stabilization of triplexes, such as T28•A20•T20 [155]. This stabilization is primarily driven by a significant negative enthalpy change, which overcomes the unfavourable entropy change. G4 structures composed of Hoogsteen base pairs experience dramatic shifts in stability under molecular crowding. Human telomeric G4, a representative motif, is stabilized in the presence of PEG 200 [156]. This stabilizing effect arises from two main factors. First, the reduction in water activity caused by the crowding agent, as the human telomeric G4 undergoes substantial dehydration during folding, more so than duplexes. Second, the excluded-volume effect was amplified because of the compact folding of the G4 structure. Thermodynamic analysis suggested that the promotion of G4 formation was favoured by a positive enthalpic contribution from dehydration, which outweighed the unfavourable entropic contribution.

The effect of molecular crowding on G4 structures varies depending on both the topology of G4 and the characteristics of the crowding agents (Table 1). For example, notable topological changes have been observed, whereby PEG 200 does not alter the antiparallel topology of the human telomere G4 in the Na^+^ ion solution, but instead shifts the topology from a mixed to a parallel form in a K^+^-containing solution [157]. However, other crowding agents produce different effects because the balance between reduced water activity and the excluded volume effect plays a crucial role. In the presence of Ficoll, human telomeric G4 is stabilized, forming a mixed structure in K^+^ buffer [158]. The varying responses of G4 to different crowding agents suggest that the physiological properties of these agents, such as their size, shape, and chemical interactions, have a substantial impact on the stability and structure of G4s.

The sequence of the G4 structure introduces an additional factor to the effect of molecular crowding. The telomeric sequence of Tetrahymena closely mirrors the human telomere sequence, differing only in the guanine base of the loop that replaces adenine [159]. This sequence interacted in an intermolecular manner and formed extended and well-ordered G-wires when subjected to crowding agents and Na^+^ ions [159]. Some G4s containing short loops, which are frequently found in human promoter regions and thrombin-binding aptamers, do not undergo topological changes in response to various cations or the addition of crowding agents. However, this enhances the stability of these structures [156].

The iM sequences also exhibited dynamic behaviours when exposed to crowding agents (Table 1). Specifically, in the presence of PEG, particularly those with a larger molecular weight, iMs, including sequences from the human telomere and the c-Myc promoter, were significantly stabilized across a pH range of 4–7. This stabilization can be attributed to the effect of crowding agents on the dielectric constant of the solution, such as a shift in the pKa of cytosine by more than two units (e.g. from 4.8 to 7.0), as well as the formation of non-specific PEG/DNA complexes, both of which play a key role in stabilizing iM structures [108, 109]. A particularly noteworthy aspect is the dependence of the PEG size on the stabilization of iMs [108]. Notably, PEG 1000 and larger PEGs seemed to be crucial for stabilising the more compact iM structure over the random coil structure at higher pH levels. This stabilizing effect has also been observed in the presence of Ficoll [160]. Unlike G4, which does not show significant stabilization with different-sized crowding agents, iM structures exhibit a distinct response to crowding conditions compared to that of G4 formations.

The effect of proteins on the thermal stability of DNA structures has also been studied. Unlike PEGs, both positively charged concentrated lysozyme and cytochrome c were found to stabilize long internal and bulging loop structures, although they had no effect on fully complementary oligonucleotide duplexes [161]. These basic proteins enhance the stability of G4 DNAs, particularly those featuring long loops [162]. These stabilizing effects are thought to arise from weak electrostatic interactions between proteins and unpaired nucleotides within the loops. In contrast, negatively charged proteins such as bovine serum albumin (BSA) and α-lactalbumin did not affect the stability of short duplexes [161]. For G4 structures, BSA was found to destabilize DNA and cause a slight shift in topology from the hybrid to parallel form of the human telomere G4, which differs from the effect observed with PEG 200 [163]. BSA causes the compaction of large DNA strands, therefore, acidic proteins such as these tend to influence non-canonical structures, primarily by increasing the excluded volume effect. In more biologically relevant environments, such as extracts from Xenopus laevis oocytes, G4 structures are stabilized and adopt a mixed topology, similar to the effect observed when Ficoll is present in K^+^-containing buffers [158]. Thus, the effect of molecular crowding by proteins can be partially mimicked by synthetic polymers, suggesting that the behaviour of nucleic acids is dominated by the physicochemical properties of the solution in cells.

Compartment effect

The conditions within cells are highly varied, shaped not only by the presence of macromolecules, but also by the distinct, organized environments created by membranes and organelles (Fig. 9). The nucleolus within the nucleus is a region where the concentration of molecules is particularly high. Furthermore, organelle membranes act as critical sites for vital processes, such as replication, transcription, and translation [164, 165]. To replicate the internal and surface environments of these organelles, researchers have studied the crowding effects using amphiphilic molecules. These molecules then form micelles and liposomes. Studies investigating the stability of human telomeric G4 and iM structures within the nanosized cavities of reverse micelles made from bis(2-ethylhexyl)sulfosuccinate (AOT) have demonstrated that these structures are stable [166]. For iM, a single-molecule Förster resonance energy transfer study revealed that the AOT concentration can influence the size of the water pool (W0 = [H2O]/[AOT] within the micelles, with the relative abundance of the iM structure approaching 100% within the microemulsion at W0 = 10 (where the micelle diameter is 4.0 nm) [167]. In contrast, at W0 = 50 (micelle diameter of 5.8 nm), the random coil was the major conformer. Stabilization of both the G4 and iM structures arises from the combined effects of reduced water activity and an increase in the excluded volume due to the confined nature of the micelle structure. However, when DNA is immobilized on a membrane, human telomeric G4 undergoes destabilization when attached to a liposome composed of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), experiences destabilization [128]. This effect is specific to liposomes containing POPC because POPC directly interacts with the G4 structure on the membrane, leading to a decrease in stability. The choline cation interacts directly with the guanine base, destabilizing its structure [86]. These findings suggest that local environmental confinement can influence G4s and iMs, which differs from the effects observed in homogeneous crowding solutions, offering the potential regulation of genome maintenance and gene expression through the formation of G4s and iMs.

Membrane-less compartments in LLPS provide a local microenvironment for biomolecular reactions. The process of phase separation can be elucidated through intermolecular interactions, such as dipole–dipole, ion–ion, and hydrogen bonding, which govern the solvent characteristics of coexisting phases [168]. DNA and RNA interact with specific intrinsically disordered proteins from LLPS to regulate replication, transcription, and other DNA manipulations on the DNA, such as DNA modification and repair [169–171]. The LLPS environment is extremely condensed with proteins and nucleic acids owing to molecular crowding. Therefore, the stability of nucleic acid structures in LLPS should be investigated. Employing the nucleolar protein component, Ddx4, to replicate the conditions of the nucleus and nucleolus, the relative stability of DNA and RNA duplexes inside and outside the LLPS was investigated, revealing that the DNA and RNA duplexes were destabilized in the LLPS environment [172]. Kinetic analysis of duplex formation by NMR using the intrinsically disordered C-terminal region of CAPRIN1 and the RNA duplex suggested that LLPS decreases the association rate constant (kon) and increases the dissociation rate constant (koff), resulting in the destabilization of duplex formation due to the interaction of the protein [173]. For non-canonical structures, the G4 structure can be a platform for LLPS [174–179]. As not only G4 but also iM structure induces LLPS, in addition to electrostatic interactions between the DNA and the intrinsically disordered region of protein, π–π stacking between tetraplex DNAs could drive droplet formation, unlike in the electrostatically driven LLPS of the duplex DNA and protein [176]. LLPS is based on the direct interaction between DNA (RNA) and proteins, which is different from inert crowders of DNA and RNA. However, the trend of (de)stabilization is similar to that of PEG with a low molecular weight, such as ethylene glycol and PEG 200 mentioned above.

Chemical environments in cells with the cellular types and conditions

As mentioned in Section 3, the biological functions of nucleic acid structures, especially non-canonical structures, are driven by spatiotemporal environmental changes. One of the triggers leading to such dynamic changes is changes in the cellular types and conditions. Therefore, we describe the roles of non-canonical structures in cell types and condition changes dominated by cellular environments based on knowledge of nucleic acid behaviours in cells.

In-cell behaviours of nucleic acids

The physicochemical characteristics outlined in Section 3 strongly suggest that the chemical environment of living cells profoundly influences the stability of nucleic acid structures. Despite the challenges involved in conducting thermodynamic studies on nucleic acids in a cellular context, there have been several investigations into how the intracellular environment affects nucleic acid stability (Table 2). In-cell NMR is a powerful tool for observing reactions within the nuclei of living cells. This technique was used to study the stability of iM DNA structures in HeLa cells [180, 181]. Notably, the imino proton signals arising from the C*C^+^ base pairs were easily distinguishable from the signals of other cellular components, making it possible to track iM formation via NMR. The iM structures were formed even at physiological pH and their thermal stability was significantly higher than that observed in non-crowded solutions. This suggests that the nuclear environment in HeLa cells resembles that of solutions containing large cosolutes, such as Ficoll or dextran, which stabilize iM structures [18]. Notably, in-cell NMR experiments in living X. laevis oocytes, but not in human cells, revealed the hybrid topology of human telomere G4, a topology which could also be reproduced in solutions containing Ficoll or dextran [182]. In addition, in-cell NMR was used to explore the folding dynamics of G4 structures. The rate of proton exchange in G4 structures provides valuable insights into folding and unfolding kinetics. The results indicate that proton exchange rates within living cells are significantly higher than those detected under various in vitro crowding conditions [183]. In-cell NMR also provides evidence of fluctuating iM stability throughout the cell cycle, as indicated by G4 and iM antibody studies [181, 184, 185]. Conversely, in-cell ^19^F NMR experiments in human cells revealed not only the expected hybrid-type G4 structure but also a two-tetrad antiparallel conformation that had not been observed under crowding conditions involving PEG or polysaccharides [186]. This finding implies that the heterogeneity of intracellular conditions can profoundly influence the DNA structure. These observations underscore the dynamic nature of the DNA structure in response to varying cellular conditions.

Although quantitative thermodynamic studies on DNA structure formation in cells are limited, some important studies have been published [187–189]. Fluorescence resonance energy transfer (FRET) is an invaluable technique for analysing the structure of oligonucleotides within cells. By labelling oligonucleotides with FRET donor–acceptor pairs, various cells can be transfected and analysed using a plate reader. As FRET efficiency is governed by the equilibrium between the coil and helical forms, changes in FRET efficiency can serve as indicators of duplex stability under different cellular conditions [187]. Combining microscopy with FRET al so provides powerful insights into the nucleic acid structural behaviour within living cells. The temperature-optical oscillation method, which relies on FRET readouts affected by laser-induced temperature oscillations, allows for melting analysis of fluorescence-labelled duplexes in HeLa cells and provides kinetic data on their (un)folding [188]. The hybridization kinetics of short DNA duplexes were observed to be almost twice as fast in the nucleus as in the cytosol. While hybridization rates were found to be size-dependent and could not be replicated by large synthetic crowding agents, such as Ficoll or dextran, these findings highlight the significant thermodynamic differences in DNA structure formation arising from varying cellular environments, likely due to interactions with cytoplasmic components. The fast relaxation imaging technique, which combines laser-induced temperature jumps with FRET microscopy [190, 191] also facilitates the direct measurement of the free energy change (∆G°37) associated with the melting of nucleic acids [189]. In one study, the stability of hairpin RNA was measured in the nucleus of HeLa cells and compared to that in a diluted buffer solution. Although no significant differences in stability were observed in the nucleus, the RNA hairpin was stabilized in the cytosol. The ∆G°37 values for folding were −1.8 kcal/mol in the nucleus and −2.2 kcal/mol in the cytosol, respectively. However, RNA stability varies considerably among different regions of the cell. These thermodynamic studies provided quantitative evidence that nucleic acid structures are affected by stabilising and destabilizing forces in specific intracellular environments.

In addition to the nucleus and cytosol, LLPS in cells is an important focus for studies on nucleic acid stability. By tracking changes in the topology and stability of G4 structures in response to varying solution conditions, dual-fluorescence-labelled G4 DNA was used as a sensor to explore G4 behaviour within cellular LLPS using FRET. The sequence was designed from human telomere G4 in these studies because it changes topology in response to crowding environments. Distinct FRET signals provide a means to assess molecular crowding in different cellular compartments. After the G4 sensor DNA was delivered into living cells by microinjection, the FRET signals differed in the nucleus, nucleolus or cytosol. Notably, the nucleolus showed a FRET signal similar to that observed in solutions with high PEG 200 concentrations, suggesting that the nucleolus possesses unique crowding properties [192]. In contrast, the nucleus displays a Ficoll 70-like environment [192]. These findings are consistent with those of previous studies [172, 189]. In an in vitro experiment designed to replicate nucleolar conditions using Ddx4 to create LLPS, DNA duplexes in Ddx4 LLPS were destabilized, as observed under PEG 200 conditions, whereas RNA or DNA hairpins were slightly stabilized [172]. The similarity between G4 topology changes in the nucleolus and those in PEG200 solutions further supports the notion that LLPS conditions in cells are locally distinct from those in other intracellular regions. LLPS is also implicated in the formation of G4s. G-rich RNA transcripts can form G4 and accumulate as LLPS in the cells, a process that may be associated with neurodegenerative diseases [178]. Molecular crowding promotes LLPS formation, especially as the dielectric constant changes, suggesting an important role for the dielectric constant in cells in neurodegenerative diseases [193]. These observations highlight the importance of recognizing that crowding conditions vary across different regions of the cell; consequently, the stability and structure of nucleic acids are subject to change depending on their specific intracellular environment.

Understanding the response of nucleic acids to changes in intracellular environments will advance alongside improvements in methods of studying these environments. Chemical indicators are available for measuring pH and certain cations, such as Na^+^, Mg^2+^, and Ca^2+^, within cells. Such intracellular conditions could be more accurately analysed using FRET sensors, with fluorescent proteins linked to a peptide that folds upon binding to specific cations such as K^+^ [194, 195]. FRET sensors were used to analyse molecular crowding and the packing-induced helix formation of peptides due to intracellular crowding [196]. The magnitude of molecular crowding was also analysed to track the mobility of tandemly connected fluorescent proteins in cells and determine intracellular viscosity [36]. The advantage of these genetically encoded proteins is their applicability to the detection of spatiotemporal conditions in cells because the expression of recombinant protein can be controlled in time and location in cells. Combining these tools to quantify the chemical environments in cells would be beneficial; however, FRET sensors cannot be used with others, limiting the amount of information that can be obtained from a single experiment. Various solution conditions parameters must be quantified to understand DNA behaviour. Therefore, a novel method for obtaining this cell information involves physical chemistry (see Section 5.2 for a description and application of this method).

Cellular stress response to environmental changes

From a thermodynamic perspective, the formation of non-canonical nucleic acid structures in living cells depends on the spatiotemporal intracellular environment. The role of non-canonical nucleic acid structures in different cell types and conditions according to changes in the intracellular environment is unclear. Therefore, it is necessary to investigate how the physicochemical properties of the solution mentioned above substantially changes in cells and to determine the existence of a mechanism led by non-canonical structures. Among the changes in cell morphology, stress responses, such as hyperosmotic stress, was investigated in model studies [197]. Under hyperosmotic stress, the cell volume changes; thus, changes in the concentrations of solutes in cells, resulting in intracellular molecular crowding. In fact, hypertonic stress by microinjection of a high-salt solution (500 mM NaCl) into HeLa cells induces a decrease in the cellular volume and an increase in the concentration of biomolecules, approximately equivalent to the solution condition adjusted with 650 mg ml^−1^ Ficoll 70 [198]. Furthermore, in the potential response to hyperosmotic stress in nature, macromolecular crowding in cells is drastically increased, which the recovery of the cellular volume is initiated by the LLPS formation of the key proteins, such as WNK kinase in human cells and Arabidopsis decapping 5 (DCP5) in plant cells [199, 200]. Although the relationship between these proteins and non-canonical nucleic acid structures remains unknown, some G4 binding proteins are activated by hyperosmotic stress. Nucleolin is abundant in the nucleolus and participates in all the aspects of ribosome biogenesis [201]. However, it can also be found in cell membranes, and, upon stress stimuli, in the nucleoplasm and cytoplasm [202, 203]. As nucleolin is also known as a G4-binding protein and G4 formation is facilitated in the nucleolus, the potential G4 interaction with nucleolin may be influenced by changes in molecular crowding due to osmotic stress and cause the release of nucleolin from the nucleolus to other parts in response to stress [204]. A similar stress-responsive protein is fused in sarcoma FUS, which forms stress granules in response to various types of cellular stress, including hyperosmotic stress [205]. FUS also acts as a G4-binding protein and generates LLPS by binding to DNA or RNA G4 [206, 207]. These findings imply a profound relationship between changes in cell morphology due to hyperosmotic stress and G4 structures.

Autophagy is another important cellular response to stresses, including hyperosmotic stress is autophagy [208]. Autophagy is characterized by the sequestration and digestion of cellular components in double-membrane vesicles to maintain homeostatic balance [209]. The relevance of autophagy with G4 was surveyed by an informatics approach using a quadruplex-forming G-rich sequence (QGRS) mapper, which predicts QGRS in nucleotide sequences [210], to identify 470 G4 sequences in mammalian target of rapamycin (mTOR) [211]. mTOR, a key negative regulator of autophagy, functions as a crucial regulator of cellular homeostasis by antagonising autophagy [212, 213]. Notably, the G4 ligand stabilizes mTOR G4 DNAs and downregulates mTOR transcription and protein expression. This study indicated a possible role of the cellular environment in stabilizing mTOR G4s in the induction of autophagy by downregulating mTOR levels [211]. As the roles of G4 have been identified in the control of cancer-related genes [19], the direct finding of the role of G4 in cell homeostasis can provide important evidence that cellular environment changes trigger gene expression for the stress signal via G4 formation of the activating gene, which may be one of the causes of changes in the intracellular environment and cellular destiny.

Cell volume regulation under osmotic stress is controlled by an efflux of K^+^ from the cell, whereas shrinkage is controlled by an Na^+^ influx [214]. Moreover, aquaporins import and export water to and from inside cells and participate in the recovery of cells after osmotic stress [215]. Notably, monovalent cations and water significantly affect the stability of nucleic acids, especially non-canonical structures, as mentioned in Section 3. Although there have been no reports on the relationship between the expression of these transporters and G4, intracellular environmental changes can induce stability changes in G4 DNA (and other non-canonical structures) in these related genes and regulate the recovery from stress.

Cancer progression

During tumour development, the chemical conditions inside cells undergo substantial changes. The metastatic capabilities of cancer cells are closely related to their intracellular chemical environments. These alterations in the cellular environment, which lead to changes in cell morphology, are driven by transporter activity and cytoskeletal rearrangement. Research on the metabolome of cancer cells has shown that their metabolism significantly differs from that of normal cells [216]. Specifically, cancer cells exhibit marked shifts in the metabolism of carbohydrates, lipids, and amino acids, which is regulated by proteins involved in transport and the cytoskeleton. If the expression of genes related to these processes is regulated by non-canonical nucleic acid structures, the factors discussed in Section 3 may contribute to cancer malignancy (Table 3).

Figure 10A shows the potential changes in chemical environments during cancer progression that affect the G4 formations in cells. One of the chemical environment changes during cancer progression is K^+^ concentration because aggressive cancer cells overexpress potassium channels, which leads to a decrease in K^+^ concentrations in these cells compared with that in non-cancer cells [217–219]. Additionally, the dielectric constant (εr) in cancer cells is notably lower than in non-cancer cells [220]. G4 stability is heavily influenced by K^+^ concentration, therefore, variations in the chemical environment during tumour progression could lead to the modulation of certain oncogenes. This is further supported by the presence of numerous G4-forming sequences in the promoter regions of oncogenes [63, 221]. Recent studies have revealed the role of G4 structures in relation to changes in the chemical environment during tumour progression, particularly in live cancer cells [217–219, 222, 223]. The formation of G4 structures is significantly affected by the surrounding environment, particularly K^+^ concentration [222]. In malignant cancer cells, the overexpression of K^+^ channels results in a reduction in the intracellular K^+^ concentration compared with that in normal cells [217–219]. In normal cells, the production of RNA transcribed from DNA templates with G4s is suppressed due to the stabilization of G4 structures. However, in highly metastatic breast cancer cells, more transcripts are generated from G4-containing templates than from the less aggressive cancer cells. This suggests that in normal cells, K^+^ ions prevent transcription of specific oncogenes by stabilizing the G4 structure [223]. This study also suggests that changes in the intracellular environment induced by the malignant transformation of cancer cells induce changes in the DNA structure and accelerate the malignant transformation of cancer cells. The malignancy of cancer cells is accompanied by morphological changes that cause a marked arrangement of actin filaments. Thus, non-canonical structures may play a role in changes in cell morphology during cancer progression.

The reduced K⁺ concentrations in cells should also impact G4s on mRNAs in the cytoplasm. RNA G4s are also stabilized by K^+^ and lack complementary strands. Therefore, their role in cancer pathology must be carefully considered [224]. G4s in mRNA regulate translation of genes associated with tumourigenesis. G4s on the oncogene mRNAs, such as NRAS, MYC, MYB, and CDK6, are located in the 5′-untranslated region (UTR), ORF, and 3′-UTR regions [225]. RNA G4s can inhibit the translational machinery and serve as signals for translation initiation through the internal ribosome entry site (IRES) [226]. In contrast, the stabilization of G4s within the IRES of VEGF mRNA inhibits IRES-mediated initiation [227]. Although future research is needed to explore these mechanisms, changes in cell morphology due to cancer progression may cause the function of mRNA to alter the level of oncogene expression. LLPS in the cytosol of cancer cells also plays a considerable role in various growth factor signalling pathways involved in cancer progression, including the T-cell receptor signalling pathway [228], innate immune signalling triggered by cyclic GMP-AMP synthetase [229], and cAMP-dependent protein kinase signalling [230]. LLPS facilitates efficient substrate–enzyme interactions by creating specialized microenvironments that enhance reaction efficiency. Thus, LLPS can accelerate the signalling pathways that maintain cell proliferation and the non-canonical structures of DNA and RNA may act as scaffolds that interact with proteins. LLPS forms G4 DNA, RNA, and RNA G4 and is frequently found in phase-separated condensates [231]. As LLPS formation strongly correlates with changes in cell morphology, the trigger of morphological stress induces LLPS formation together with K^+^ concentration in the cell, potentially resulting in the development of cancer malignancy.

As mentioned above, cancer cells have varied metabolomes depending on cancer type [216]. Haeme complexes with hemopexin (HX) in vivo, which is taken up by the myeloid cell receptors, CD163 or CD91/LRP1, on macrophages that repress cancer growth and metastases [232, 233]. When the scavenging role of HX does not work, the level of poorly differentiated tumours is elevated in prostate cancer [234]. The lack of HX increases the amount of labile haeme, which promotes tumour growth and metastasis in an orthotopic murine model of prostate cancer. HX deficiency in prostate cancer results in the accumulation of labile haem in the nucleus, which promotes cancer cell growth by interacting with G4 DNAs in the nucleus to modulate the expression of specific genes. c-MYC mainly regulates haeme-driven cancer progression, with the most aggressive phenotype detected in mice lacking HX. The metabolism of ROS is also a direct and potential modulator of cancer progression via G4, because ROS causes oxidation of guanines in G4, resulting in the loss of stability and function of G4 [235, 236]. ROS increases the transcriptional activity of VEGF via its promoter, G4 [237]. Hif1a, which is clustered with PQSs (iM-forming sequences) in the promoter region and 5ʹ-UTR [238–240], plays a key role in ROS-induced carcinogenesis [241]. ROS also promotes the transcription of proto-oncogenes, such as BCL2, KRAS, and c-Kit [242], suggesting that ROS can induce various types of cancer progression. Choline metabolism is also a notable factor in cancer progression. Choline metabolism undergoes considerable changes in cancer, resulting in an increased demand for choline and its metabolites [243]. The rapid growth of cancer cells necessitates increased production of phosphatidylcholine to form new cell membranes [244]. Furthermore, altered signalling pathways in cancer cells enhance choline uptake and utilization. Changes in choline metabolism have been linked to cancer onset, tumour advancement, and resistance to treatment [245]. Although there is no direct evidence that choline modulates cancer progression by interacting with G4 and iMs, the role of these non-canonical structures remains to be clarified.

Furthermore, the oncogenic transformation of cells results in significant changes in protein expression levels and cell morphology. These changes were presumed to alter the intracellular molecular crowding environment. The cell cycle, which is the most significant change in cell morphology, is an essential process in cell division and closely regulated by a complex of cyclin-dependent kinases, CDK and cyclin D. Although cyclin D plays a central role in regulating the cell cycle, it has also been reported to act as an oncogene in many cancer cells [246]. Most human cancers are caused by abnormalities in cell-cycle regulation and growth factor-dependent pathways. Cyclin D is involved in cell cycle regulation and growth factor signalling. Thus, changes in cell morphology and cancer are closely related. The expression of the growth factors Ras, Raf, and ERK, and the activation of the transcription factor, c*-Myc*, also promote the production of cyclin D. Notably, many of these genes have sequences that can form G4, and the effect of G4 formation on the cell cycle is not yet known; however, it is possible that G4s regulate the expression of these genes and may induce carcinogenesis via cyclin D expression.

Cytoskeletons composed of protein fibres maintain cell morphology and strength. Importantly, cytoskeletons, such as tubulin and actin, are also known to be involved in cancer progression, including cell invasion and metastasis [247]. Thus, the regulation of cell morphology via the cytoskeleton may also be closely related to cancer progression. The cytoskeleton is one of the factors that induce intracellular molecular crowding, which significantly affects the formation of G4s. There may be a mechanism by which the formation of G4s regulates the expression of genes involved in cell morphology and changes in cell morphology further promote the formation of G4s, which are involved in cancer progression.

Senescence cells

Significant changes in the cell environment can occur due to tumour progression as well as senescence. During these processes, the morphology of the cells is highly heterogeneous and considerably fluctuates. Senescence can be induced in a DNA damage-dependent manner; therefore, deactivation of helicases is known to be one of the factors involved in senescence. Notably, as the G4 helicase WRN is mutated in Werner syndrome, which is characterized by accelerated aging, many G4-related proteins and enzymes responsible for DNA unwinding, chromatin remodelling, histone modification, and telomere maintenance are known to regulate senescence [248]. Moreover, stimuli responsible for senescence, such as UV irradiation, induce the accumulation of G4 in cell nuclei, which activates the UV-damage repair pathway via G4 binding proteins [249]. Otherwise, it is possible that G4 destabilization promotes senescence (Fig. 10B). Therefore, cellular environment-dependent gene expression via non-canonical DNA and RNA structures is closely related to senescence.

Senescent cells increase in size (volume), which is also an important factor influencing the chemical conditions within cells. Senescent cells are generally dilated in size and flattened in morphology compared with cells that divide and proliferate, such as cancer cells [250, 251]. Irreversible cell cycle arrest is one of the phenomena characterising cellular senescence, and decreased expressions of p16, p21, p53, and pRB (phosphorylated retinoblastoma protein) are used as indicators of cellular senescence [252]. These genes are also known as cancer suppressors. Oncogenic progression and senescence result from opposing gene actions, leading to morphological differences. Maintaining an appropriate cell morphology is critical for proper cell function, therefore, cells should sense and regulate their morphology. As mentioned in Section 4.1, changes in cell volume directly influence molecular crowding and water content in the cells. Expanded young cells exhibit the characteristic phenotypes seen in senescent cells, such as delayed cell division, increased DNA damage, reduced sensitivity to pheromones, and changes in overall transcriptional activity [253]. Importantly, that molecular crowding within the cytoplasm decreases during cellular senescence [253]. This implies that alterations in molecular crowding, resulting from changes in cell morphology, could contribute to the promotion of cellular senescence. Although it is unclear how cells exhibit the senescent phenotype through cytoplasmic dilution, the expression levels of aquaporins are closely related to the cell cycle and senescence [254, 255]. Hence, the relationship between molecular crowding and cell volume changes suggests that the modulation of gene expression of such transporters via non-canonical structures plays a key role in senescence.

Another possibility is that the assembly and disassembly of membrane-less organelles are regulated by alterations in the physicochemical environment of the cells. It has been proposed that compartmentalization via LLPS is driven by macromolecular crowding in the cytoplasm [256]. More recently, it has been suggested that entropy-driven LLPS is enhanced by molecular crowding [257]. Therefore, a decrease in molecular crowding could prevent the formation of membrane-less organelles or lead to the disintegration of existing organelles, thereby contributing to the expression of senescent phenotypes. During the early stages of senescence, various cellular stressors induce the formation of stress granules in the cytoplasm. However, as senescence progresses, the number of stress granules decreases. This suggests that the loss of the ability to form stress granules due to cytoplasmic dilution may accelerate senescence. In addition to stress granules, the number of processing bodies (p-bodies) in the cytoplasm increases in response to specific stresses, such as glucose starvation or osmotic stress [258]. P-bodies may also play a role in protecting cells from various stressors during the early stages of senescence. RNA G4s can accelerate phase separation in cells, therefore, the assembly of stress granules and p-bodies [193, 231] may be attributed to the formation of G4s in senescent cells.

Moreover, the nucleolus of nuclei LLPS assembles during the G1 phase and disassembles in the mitotic (M) phase throughout the cell cycle [35]. Importantly, molecular crowding in the nucleus increases from the G1-to-M phase [259], suggesting that this increased crowding during cell cycle progression may promote the formation of LLPS. Additionally, the nucleolus expands during senescence [260]. The nucleolus is responsible for ribosome biogenesis, including the transcription of rDNA into rRNA and rRNA processing [35]. Oncogenic stress increases rRNA transcription, whereas replicative stress delays rRNA processing, leading to the accumulation of nucleolar RNA and the induction of senescence [260]. Increased pre-rRNA levels may promote nucleolar expansion via LLPS stimulation. Elevated pre-rRNA levels could enhance transcription and the newly transcribed pre-rRNA may further increase molecular crowding in the nucleus. Molecular crowding promotes transcription by increasing the affinity between promoter DNA and RNA polymerase through an excluded volume [261, 262]. Notably, pre-rRNA, which contributes to molecular crowding in the nucleus, contains a GC-rich sequence that is prone to forming G4s. G4s are present in rDNA and large subunit rRNA of the human genome [263]. Recent studies identifying PQSs have revealed 107 such sequences in the human 45S pre-rRNA, indicating eight G4s per 1 kb. This density of G4s in the 45S pre-rRNA contrasts with the general frequency of 1.3 G4s per 1 kb in non-coding RNAs. RNA G4s can be recognized by specific proteins, such as nucleophosmin (NPM1), a ribonucleoprotein in the nucleus [264]. Moreover, under conditions of molecular crowding, water activity decreases compared with that in dilute solutions, favouring phenomena such as G4 formation and interactions between RNA and proteins. Therefore, the interaction between pre-rRNAs and nucleolar proteins during LLPS and transcription should be considered. Both the increase in pre-rRNA and the formation of G4s within it may contribute to the nucleolar expansion. This could be an example of how non-canonical nucleic acid structures promote the assembly of membrane-less organelles during increased molecular crowding. Future research on the role of nucleic acids as sensors of cell morphology through changes in molecular crowding is required. G4s in rRNA and nucleolar proteins can alter molecular crowding conditions in the nucleolus, potentially influencing transcriptional activities. The morphology of molecular crowding reagents does not uniformly affect transcription reactions [265]; therefore, the varying morphologies and shapes of the complexes formed between 45S pre-rRNA and associated proteins should be further explored.

Cellular senescence is strongly related to the mitochondrial condition. Mitochondria are intracellular organelles responsible for energy (ATP) production and contain their own DNA (mtDNA). mtDNA is more susceptible to oxidative damage by ROS, and its capacity to repair ROS-induced damage is weaker than that of nuclear DNA [266, 267]. Point mutations and deletions accumulate in mtDNA with age, which weakens mitochondrial function, leading to energy deficiency and cell death [268]. Interestingly, mtDNA contains PQS with a higher density than the genomic DNA [269]. These G4s may play a role in mtDNA replication and transcription [270–272]. The G4 located in non-coding regulatory regions may hinder access to replication machinery, modulating the initiation and progression of mtDNA replication [273]. Stabilizing the G4 on mtDNA considerably reduces the expression level of polycistronic mRNAs during transcription and suppresses ATP synthesis [274], which are vital for mitochondrial function. In addition, mitochondrial G4 is associated with mtDNA deletions, which lead to ageing [270, 275]. These G4 structures are implicated in DNA deletions that contribute to inherited disorders and the aging process. A deeper understanding of the influence of the mitochondrial environment on mtDNA is essential for elucidating mitochondrial functions and for advancing treatments and technologies targeting mitochondrial-related diseases.

A growing interest focuses on molecular crowding within mitochondria and its role in forming G4. The protein concentration in the mitochondrial matrix is high, suggesting a specialized biochemical setting [276, 277]. The macromolecule density within the mitochondrial matrix of human HeLa cells far exceeds that of the nucleus or cytosol, indicating a uniquely crowded intracellular environment [278]. More stable G4 structures form from the same sequence in mitochondria than in the nucleus, which causes efficient transcription stall of mitochondrial RNA polymerases [279]. mtDNA mutations are also related to cancer. Cancer occurs and progresses due to the accumulation of somatic mtDNA mutations with aging [280], which can be triggered by the G4 stabilized in various mitochondrial environments. Thus, the modulation of the mitochondrial environment may induce cell senescence via forming G4.

Application to regulate the intracellular environment via control of non-canonical structures

The aforementioned section summarizes the relationship between changes in the cellular environment and non-canonical nucleic acid structures, which can cause changes in cell morphology. To regulate cell destiny related to these phenomena, the classical concept is the development of ligands targeting non-canonical structures of interest, such as oncogenes. However, we propose that the concept of gene regulation in these phenomena is different. Changes in the intracellular environment causing changes in cell behaviour can be induced by a balance or dysfunction in the expression of membrane transporters, altering the structure and stability of nucleic acids, which regulate the onset and progression of senescence and cancer. Normalization of the levels of non-canonical structures on genes by modulating crowding environments in cells can activate (or inactivate) allosteric gene expression to prevent senescence and cancer. In addition to the conventional analysis of the stability of the structure using energy diagrams, the effect of the solution environment on in-cell nucleic acids should be analysed as a physicochemical parameter.

Prediction of nucleic acid behaviours in cells depending on different cellular environments

Thermodynamic data on sequence- and environment-dependent folding in cells can be used as a database for the prediction of nucleic acid structures in various cells. From a physical chemistry perspective, it is possible to predict the formation of non-canonical structures, such as G4s and iMs, competitive with duplexes, if we obtain the relationship between the stability and the sequence in solution defined by the various physicochemical properties. Currently, NN parameters are the most reliable platform for predicting duplex stability. The classical NN parameters were developed from the thermodynamic parameters obtained under test tube conditions, including 1 M NaCl, which is far from physiological conditions [46, 281–285]. To establish the application of NN parameters to the in-cell stability of duplexes, improvements in the NN parameters have been made for crowded environments. The initial improvements in the NN parameters under crowding conditions were determined under 40 wt% PEG conditions with 0.1 M NaCl for the DNA/DNA duplex [286] and 20 wt% PEG 200 conditions with 1 M NaCl for the RNA/RNA duplex [287]. The recent improvement in NN parameters for various cellular conditions was aimed at being available for any cation and crowding condition. The free energy of each NN parameter (∆G°37,NN) for the duplex was considered simply as the sum of cation and crowder factors as ∆G°37 NN = ∆G°37 NN [cation] + ∆G°37 NN [crowder]. The ∆G°37 NN[cation] parameters at any concentration of NaCl can be calculated from those in 1 M NaCl using the dependence of [Na^+^] for each NN base pair [288]. The ∆G°37 NN[crowder] parameters can be obtained from the linear function of changes in water activity ∆aw, because duplex destabilization in the presence of the crowder (∆∆G°37) correlated linearly with changes in water activity under crowded conditions [152]. Based on this strategy, improved NN parameters were obtained for any molecular environment [289]. This approach to develop the NN parameters in any cation and crowder conditions was applied to RNA/RNA and RNA/DNA duplexes [290, 291].

These universal parameters can be used to predict the in-cell stability of the duplexes. For example, the stability of the DNA duplex in Ddx4 LLPS [172] was successfully predicted using the “universal” parameters defined from 50 wt% PEG 200 with 0.1 M NaCl conditions. In RNA duplexes, the addition of 30 wt% PDO as a cosolute with 100 mM NaCl and 40 wt% PDO with 100 mM NaCl reproduced RNA duplex stability directly measured in the nucleus and cytosol, respectively [290]. This method illustrates that in-cell conditions can be replicated through a simple, yet distinct, crowding environment, facilitating the study of different nucleic acid behaviours within various cellular compartments under these solution conditions. These parameters have been used to predict the efficiency of gene editing using CRISPR/Cas9 [291]. Furthermore, these parameters can be utilized to predict G4 and iM formation because duplex dissociation is necessary to form G4 and iM in genomic DNA. Although the prediction was for in vitro conditions, the improved parameters for GC-biased sequences of DNA duplexes could predict the formation of G4 and iMs under various cation and molecular crowding conditions [153], indicating that stability prediction could work for the formation of G4s and iMs in cells, although this prediction did not consider the stability of G4s and iMs.

Unlike duplexes, intricate chemical interactions in non-canonical structures pose considerable challenges in predicting their stability. For example, the thermodynamic profile for the formation of a G-quartet stack was derived by subtracting the energetic parameters, resulting in ∆G°20 = −2.2 kcal mol^−1^, ∆H° = −14.6 kcal mol^−1^, and T∆S = −12.4 kcal mol^−1^ at 20°C from the difference of the thermodynamic data of two and three G-quartets [292]. However, this scenario was ruled out based on the effects of the loop and G4 topology on G4 stability. The G4 stability depending on the loop length was adopted to the simple logarithmic trend to obtain the temperature-dependent energy function as ∆G = (n − 1)∆Gq + (m − 2)∆GL, where n is the number of G-quartets, m is the sum of the length of three loops in the G4 structure, ∆Gq is the free energy contribution of the G4 core, and ∆GL is the free energy contribution of the G4 loops [293]. This rule supports the prediction of RNA secondary structures using the Vienna RNA Platform [294]. However, these predictions can only be used in limited situations because cation conditions and molecular crowding significantly affect the stability of G4s and iMs. Systematic analyses of the thermodynamic parameters of various G4 and iM formations under molecular crowding conditions have been previously reported [107, 295]. Fundamental physicochemical analyses and data acquisition are required to establish more reliable and useful prediction methods for G4s and iMs.

However, global identification of G4s in genomes has been investigated to predict G4 formation in cells. Initial informatics approaches for predicting G4s were based on biophysical principles, suggesting that a DNA motif with four runs of at least three guanines separated by loops of one to seven nucleotides was likely to form a G4 structure. The first search of such patterns in the human genome identified ~376 000 potential G4s in the human genome [296]. QGRS Mapper [210] and QuadFinder [297] performed broader searches for G4s, focusing on short loops and uninterrupted G runs, both of which enhance G4 stability. The use of these algorithms revealed that gene promoters (defined as the 1 kb region upstream of the transcription start site) contain a higher number of quadruplex motifs than other genomic regions, with one or more G4 motifs present in over 40% of human gene promoters [298]. As the understanding of G4 diversity expands to include G4s with long loops, bulges, or missing guanines, algorithms for G4 search and prediction have evolved to accommodate these variations [299–303]. Recent advancements in whole-genome search methods include the detection of intermolecular G4s formed between two DNA strands, as well as slightly mismatched sequences [304, 305]. Moreover, improved prediction strategies have integrated contextual factors, such as the continuity of G-runs, loop size, and nucleotide content biases. The G4 hunter defines the G-richness and G-skewness of a specific sequence, providing a propensity score for possible G4 formations as output [306].

Novel experimental techniques for G4 detection have recently been developed. A polymerase stop assay using Illumina sequencing, termed G4-seq, enabled the detection of over 525 000 G4s in purified nuclear DNA in vitro, with >710 000 G4s detected in the presence of a G4-stabilizing ligand [307]. In cell-based G4 detection, methods, such as G4 chromatin immunoprecipitation sequencing, have been used with structure-specific antibodies to identify G4s in cellular cultures [55, 308, 309] and tumour xenografts [310]. G4 detection using cleavage under targets and tagmentation (CUT&Tag) techniques has also been used [311, 312]. iM detection was also developed using an iM-targeting antibody [313]. The iM CUT&Tag technique revealed 53 000 iMs in three human cell lines (MCF7, U2OS, and HEK293T), which were generally less abundant than G4s in human cells [185, 314]. The number of G4s and iMs reported in cells varies considerably depending on the experimental method and cell type used.

Established databases and prediction tools are currently undergoing significant progress. However, information on the effects of cellular environments is limited. Predicting G4 and iM stability using sequence and environmental information remains challenging. Thus, the combination of genome-wide experimental data with machine/deep learning is a powerful tool for providing information about the effects of intracellular conditions on non-canonical structures in cells. The recently developed Quadron algorithm serves as a predictive model for G4 formation, using tree-based gradient boosting machines trained on the G4-seq data. It generates a score that evaluates the stability of G4 structures formed within the genome, based on sequence information [315]. PENGUINN was also trained by G4 forming data based on convolutional neural networks, which learn the characteristics of G4 sequences and accurately predict G4s, outperforming state-of-the-art methods [316]. These approaches use in vitro G4 data from G4-seq. Following the extension of G4-seq to produce G4 maps for the whole genome of 12 species, including well-studied organisms and pathogens [317], these methods offer a powerful approach for investigating G4 formation in the genome under different solution conditions because G4 formation is highly sensitive to these conditions. Recently, DeepG4 was developed based on in-cell G4 data from G4 CHIP-seq [318]. DeepG4 is highly accurate for predicting active G4 regions in different cell types. Moreover, DeepG4 identified key DNA motifs predictive of G4 activity. However, although DeepG4 first implemented the actual cellular context for G4 formation prediction, only the G4s that were formed both in vitro and in cells were selected, which suggests that G4 formation in the test tube condition is not reliable to support G4 formation in cells. Another approach, epiG4NN, comprises a hybrid deep neural network which uses cellular epigenetic features and DNA sequences for G4 prediction in genomic DNA has been developed [319]. Notably, when trained on A549 cell data, epiG4NN was able to predict G4 phase formation in HEK293T and K562 cell lines. These findings suggest that epigenetic modifications and chromatin accessibility are important factors for the formation of cellular G4s. The aforementioned programs for predicting the stability of nucleic acid structures under various conditions are summarized in Table 4. Although predicting the formation of non-canonical structures, particularly G4s, is becoming more accurate, the progress of the methodology should be completely based on the physicochemical properties that directly affect G4 formation, such as cations, pH, metabolites, molecular crowding, and compartment effects. These prediction methods can incorporate information on G4 and iM structures formed via physical cellular processes, such as negative supercoiling by polymerases and topoisomerases, as well as chromatin dynamics. The organization of DNA in the nucleus affects cancer and cellular senescence. Therefore, the physical factors affecting the formation of G4 and iMs in cells must be simultaneously considered for the prediction methods of G4/iM formations. For example, the Hi-C technique has been used to reveal how the structure of G4 affects the 3D structure of chromatin [320, 321]. This chemical and physical information could be integrated to predict the formation and function of non-canonical DNA structures. However, the cells are difficult to handle in physical chemistry.

Physicochemical techniques for understanding the behaviour of nucleic acids inside cells

Ligands that regulate non-canonical structures are conventionally considered drugs. However, as mentioned above, the intracellular environment is an important factor in senescence and cancer triggered by changes in cellular morphology. These morphological changes induce conformational changes in nucleic acids, which can regulate the expression of genes involved in aging and cancer. This idea leads to a novel concept that ‘living cells act as living drugs’. The significance of intracellular conditions must be understood more quantitatively and systematically than before.

The types and amounts of proteins expressed in cells change significantly with changes in the cellular status, such as cell cycle, ageing, and disease onset and progression. Furthermore, the expression level and activity of ion channels in the plasma membrane change according to the cellular state [322]. These changes are responsible for considerable fluctuations in intracellular molecular crowding and ionic environments. Importantly, these changes are influenced by the physical properties of the solution, such as water activity and dielectric constant, resulting in different behaviours of nucleic acids depending on the cellular state. Therefore, it is important to quantify the changes in the cell that affect the structure and stability of nucleic acids; however, the complexity of the intracellular environment has made it difficult to quantify these effects.

To quantitatively and systematically understand the effects of such environments in the cell, a pseudo-cellular system, which is the system for highlighting the environments inside the cell (SHELL), has been developed [323]. SHELL is a process in which small molecules are removed from the cell and a crowded intracellular environment is maintained. SHELL offers two prominent (i) precise quantitative biochemical analysis of a specific factor, and (ii) the study of any cell, such as normal, breast cancer, and nerve cells, thereby facilitating the study of target molecule effects in various cellular environments. SHELL allows the solution environment to change and experiments to be performed using physical and biochemical methods (Table 2).

SHELL is an experimental system that preserves the intracellular environment inside the cell. SHELL was prepared using an extension of the immunostaining method (Fig. 11). Immunostaining, reagents, such as alcohols and cross-linking agents, are commonly used to preserve and visualize the morphology of biomolecules within cells [323, 324]. These reagents hinder the diffusion of water-soluble biomolecules and facilitate their deposition or immobilization, thereby enhancing their physical stability. Immunostaining was used to develop SHELL to replicate the intracellular environment [323].

Using SHELL, research has shown the importance of changes in the cellular environment during cancer progression. SHELL was used with mild and aggressive cancer cells. DNA with G4-forming sequences was then introduced into the SHELLs and thermal stability was analysed. The results showed that G4 was more stable in SHELL than that in vitro and stabilization resulted from a favourable enthalpy change. Notably, the degree of stabilization depended on the cell type; that is, G4s were more stable in mild cancer cells than in malignant cancer cells. The formation of G4s depending on the degree of cancer progression, is consistent with the results obtained by transcription assays, which show that more RNA is transcribed from template DNA with G4 in malignant cancer cells than in mild cancers [223]. Detailed physicochemical analyses have shown that these differences in stabilization depend on the differential affinities of G4s and potassium ions within the cell. These changes in transcript levels via G4 phase formation are important for determining gene expression levels during cancer progression.

SHELL has great potential for investigating environmental factors, such as pH and ROS, and their impact on nucleic acids from a physicochemical perspective. The use of SHELL to understand the physicochemical behaviour of biomolecules will provide a better understanding of important biological reactions, molecular mechanisms of diseases, and drug design.

Conclusion and perspective

The regulation of gene expression can occur through epigenetic control due to the chemical modification of nucleic acids and related proteins, and non-epigenetic control depending on the non-canonical structures of nucleic acids. The former is a constant change because the chemical modification is linked to covalent bonds, whereas the latter is a dynamic change based on the equilibrium affected by the solution environment. This implies that epigenetic modification is useful for fixing the condition of cells, such as certain cell types, whereas non-canonical structures are more suitable for the changes in the state of the cells. This review focuses on the formation of non-canonical nucleic acid structures that can be governed by the solution environment and their biological role in cell behaviour, such as cell type and morphology changes. Based on the knowledge, it can be considered that there are two different roles of non-canonical structures for these cell behaviours. The first involves the regulation of gene expression in various cell types. This concept is only partially acceptable because G4-based gene expression is different for each cell condition. Oncogene expression is a representative example of this. Although the relationship between G4-based gene expression and unique intracellular environments, such as cations and molecular crowding, is still unclear, the mechanism of environment-dependent gene expression will be elucidated. Second, the dynamic formation of non-canonical structures triggers changes in the cell behaviour. Similar to the stress response, non-canonical structures may be used to sense external or internal signals to regulate the cellular environment. This might be induced in the non-canonical structures of genes related to the architecture of cells, including transporters and the cytoskeleton. If the regulation of these gene expressions occurred transiently, we might have overlooked the importance of these gene expressions based on non-canonical structures. Therefore, it is important to predict the formation of non-canonical structures from the sequence and molecular environment information, and identify new roles of non-canonical structures in cell destiny. Moreover, the detection of non-canonical structures should be improved to identify dynamic structural changes in cells. Drugs targeting G4/iM have been developed for static structures and, therefore, cannot regulate the expression level of a certain gene. Thus, the transient and intermediate structures of the target G4/iM can be a potential novel platform to regulate efficiently the target gene expressions. Gene expression could also be modulated by the cellular environment and thus regulate cell senescence and cancer. After finding novel motifs of non-canonical structures that transiently and dynamically form for cell behavioural changes, we will be able to design and obtain drugs targeting these motifs and demonstrate the prevention of senescence and cancer. DNA stability and function in a cell can be parameterized to quantitatively assess cell fate. Information on DNA thermodynamics and the physicochemical properties of the intracellular environment can be integrated using a pseudo-cellular system, such as SHELL, to create a database for predicting gene expression in cells with specific morphologies. For example, a series of databases representing cells in different malignancies may reveal trends in gene expression across various cellular environments as cancer progresses.

Currently, the thermodynamics of simple nucleic acid model systems measured in vitro under a two-state assumption cannot directly address what is happening in the cell. These thermodynamic principles are a critical piece, a tiny step forward, of a far more complex set of coupled equilibria that occur in the cell. However, recent machine learning and AI-driven data analysis identified a connection between stability and gene expression and protein binding, which are directly linked to cell functions [53]. Therefore, the concept that information about DNA stability in the DNA sequence serves as a dynamic code for the genome’s function will be highlighted. Consequently, genomic DNA sequences and cellular environment data can serve as inputs for such databases to evaluate senescence and predict cancer occurrence for certain cells.

Data availability

All data are available in the main text.