Chromatin‐modifying drugs and metabolites in cell fate control

Abstract For multicellular organisms, it is essential to produce a variety of specialized cells to perform a dazzling panoply of functions. Chromatin plays a vital role in determining cellular identities, and it dynamically regulates gene expression in response to changing nutrient metabolism and environmental conditions. Intermediates produced by cellular metabolic pathways are used as cofactors or substrates for chromatin modification. Drug analogues of metabolites that regulate chromatin‐modifying enzyme reactions can also regulate cell fate by adjusting chromatin organization. In recent years, there have been many studies about how chromatin‐modifying drug molecules or metabolites can interact with chromatin to regulate cell fate. In this review, we systematically discuss how DNA and histone‐modifying molecules alter cell fate by regulating chromatin conformation and propose a mechanistic model that explains the process of cell fate transitions in a concise and qualitative manner.

the conformation of chromatin, ie the 3D organization of the genome, to determine the fate of cells, through an as yet incompletely understood process.
These findings have promoted a general interest in the study of chromatin modifications and regulation. In recent years, some chromatin-modifying drugs and metabolites have been shown to possess the ability to change the fate of cells, 16,17 but there is a lack of systematic synthesis of these myriad findings. In this review, we summarize the epigenetic effects of these small molecules, discuss the mechanisms of interactions between epigenetic regulation and transcription factors during chromatin changes in cell fate determination and hypothesize the potential value of these drugs.

| THE REL ATI ON S HIP B E T WEEN CHROMATIN AND CELL FATE
Stem cells have the unique abilities of long-term self-renewal and multipotent differentiation, which are essential for maintaining the stem cell population and tissue integrity. Since stem cells and their differentiated progeny share the same genome and differ only in their chromatin organization, increasing evidence suggests that the unique characteristics of stem cells are largely determined by chromatin patterns. 8,18,19 The chromatin structure, dynamics and functions of stem cells are distinct from differentiated cells. [20][21][22] For example, pluripotent stem cells have more open and easily accessible chromatin, 23 which makes them highly plastic in their cell fate trajectories.
The chromatin of eukaryotes is highly complex, with different levels of assembly structure and a compression ratio of up to 10 000.
The nucleosome is the basic unit of chromosomes, which consists of two copies of two heterodimers H2A/H2B and H3/H4 to form a histone octamer (Figure 1), surrounded by double-stranded DNA of about 146 bp. 24 Histone subunits are rich in α-helices with basic Arg and Lys residues, thus endowing them with net positive charges. This allows them to interact with the acidic and negatively charged DNA molecules, via ionic and hydrogen bonding. For example, the amino acid side chains of histone residues, such as H3R42 and H3T45, form hydrogen bonds with the oxygens in the phosphodiesters of DNA. 25 The binding of DNA at the nucleosome entry/exit region (ie the head and tail of the DNA wrapped around the nucleosome) is not stable, but the internal DNA region near the bipartite axis is most tightly wrapped around the histones. 26 The structural characteristics of nucleosomes mean the DNA entry/exit regions can easily unwind from histones, thereby initiating DNA replication, transcription and repair activities.
The Arg and Lys residues in histone subunits are not only critical for interactions with DNA, but also provide side chains amenable to chemical modifications that regulate chromatin structure and gene transcription. Histone methylation usually occurs in specific Arg and Lys residues of the histone tails, 28 which have different effects on gene activity depending on the specific residues that are modified and the degree of methylation. Each Lys (K) residue has three possible methylation states: mono-, di-or tri-methylation. The di-or tri-methylation at H3K4, H3K36 and H3K79 is usually associated with transcriptional activation, [29][30][31] while H3K9 and H3K27 methylation is generally associated with transcriptional repression. 29,31 H3K9me3 is a feature of heterochromatin, [32][33][34][35] while H3K9me2 is more common in silent or near-silent genes of euchromatin. 34,35 In embryonic stem cells (ESCs), H3K9me3-marked heterochromatin domains increase with differentiation, thus contributing to lineage restriction and cell fate determination. 32 F I G U R E 1 The 3D structure of the nucleosome (PDB code 1KX5). 27 A, Top-down view of the nucleosome with acidic DNA (blue) wrapped around histones with α-helices (red) rich in basic residues. B, Top-down view of the DNA double helix (green and brown) wrapped around the histone octamer core consisting of pairs of H2A (green), H2B (yellow), H3 (purple) and H4 (red), and their respective histone tails. C, Side view of the ~ 1.75 turns of DNA wrapped around the histone octamer core, with histone H3/H4 tails flanking the DNA entry/exit regions Stem cells also have chromatin domains with a special histone modification pattern called 'bivalent domains', which consist of large regions of repressive H3K27 methylation harbouring smaller regions of activating H3K4 methylation. 36 Bivalent domains largely mark silent lineage-specific genes that are ready to be activated at any time, so they can be quickly activated or repressed during differentiation and development. 37 The bivalent domains in ESCs are regulated by BAF60 chromatin remodelling proteins, which regulate the redistribution of H3K4me3 and H3K27me3 and thus pluripotency. 38 In addition to ESCs, some adult stem cells also have bivalent domains, such as hematopoietic stem cells and muscle stem cells. 39,40 During the early stages of muscle stem cell activation, bivalent domains increase via the expansion of the repressive H3K27me3 mark. When muscle stem cells commit to the myoblast stage, most of the bivalent domains resolve, and most genes in myoblasts revert to a monovalent state. 40 Histone acetylation is often higher in undifferentiated ESCs and muscle stem cells than their differentiated progeny, and histone acetylation is known to control the dynamics of normal chromatin. [41][42][43] Meanwhile, histone deacetylase inhibitors can significantly increase histone acetylation, thereby increasing the turnover dynamics of euchromatin proteins in mouse embryonic fibroblasts. Studies have also shown that histone deacetylase inhibition can promote the reprogramming of somatic cells into pluripotent cells and also help maintain ESCs in an undifferentiated state. 44,45 Thus, histone acetylation is closely related to the highly dynamic euchromatin, cellular plasticity and cell fate determination. Besides histone acetylation and methylation, other post-translational modifications (PTMs) of histone tails may also affect the stability of the nucleosome core and its accessibility to chromatin remodelling complexes and DNA sequence-specific transcription factors. [46][47][48][49] Most of these histone modifications require metabolic intermediates as their cofactors or coenzymes and are heavily influenced by the metabolic state of cells.
In addition, alterations in DNA topology also regulate the structure and function of chromatin by affecting the binding of DNA to nucleosomes. [50][51][52] For example, DNA intercalator drugs such as doxorubicin can be inserted directly into the DNA double helix, thereby affecting the interactions between DNA and histones, resulting in changes in the 3D structure and function of chromatin. 52 We will systematically review the role of metabolites and drugs in chromatin biology and formulate a unified model of how these small molecules might regulate cell fate.

| CHROMATIN -MODIF YING ME TABOLITE S IN CELL FATE CONTROL
Histone tails play an important role in nucleosome stability, 53 nucleosome localization 54 and the binding of transcription factors to DNA, 55,56 In addition, probably because of their accessibility, histone tails are the most heavily modified domains of histones and exert the greatest impact on chromatin structure. For example, the tails of histones H3 and H4 are located near the DNA entry/exit region ( Figure 1). PTMs are more frequently found on the H3/H4 tails than the H2A/H2B tails, and many of them are associated with gene transcription and replication. 57 As mentioned above, most of these histone modifications require metabolic intermediates as their substrates and are therefore affected by the metabolic state of the cell.
S-adenosylmethionine (SAM) is a universal methyl donor for histone methylation. Methyltransferases transfer methyl groups from SAM to proteins and DNA, to produce S-adenosyl-L-homocysteine (SAH) and methylated biomolecules. 16,58,59 SAM is a metabolite derived from one-carbon (1C) metabolism involving the folate and methionine cycles. Serine, glycine and threonine are the primary metabolic sources of 1C units. Serine is broken down to methyl-tetrahydrofolate (THF) and glycine by serine hydroxymethyltransferase (SHMT). Glycine can be broken down again by the glycine cleavage system (GCS) to synthesize additional methyl-THF. Threonine also supplies methyl-THF, glycine and acetyl-CoA to cells through a similar reaction mechanism. 58,60 In pluripotent mouse ESCs, threonine dehydrogenase (TDH)-mediated threonine catabolism plays a key role in the regulation of histone methylation, as both threonine deprivation and TDH inhibition can reduce the SAM content in mESCs, thereby causing ESC differentiation. 60,61 Acetyl-CoA is the two-carbon (2C) metabolic substrate used to fuel histone acetylation. 16,62 During cellular metabolism, acetyl-CoA is synthesized from pyruvate, citrate, acetate and β-ketoacyl-CoA, which are catabolism products of glucose, fatty acids and amino acids, respectively. 63 Under culture conditions rich in carbohydrates, the mitochondrial pyruvate dehydrogenase (PDH) complex converts pyruvate to acetyl-CoA. 64,65 In the mitochondria, citrate synthase combines the 2C portion of acetyl-CoA with oxaloacetate to form citrate, which is then oxidized during the TCA cycle. Alternatively, citrate is transported into the cytoplasm, where ATP citrate lyase (ACL) cleaves citrate to regenerate acetyl-CoA and oxaloacetate in the cytosol. 66 Accumulation of cytosolic acetyl-CoA can either promote the synthesis of lipids or promote the transportation of acetyl-CoA into the nucleus. In the nucleus, acetyl-CoA acetylates histones, resulting in gene activation. 67 Therefore, the basal level of histone acetylation is highly dependent on the state of cellular catabolism. 67 Studies have found that the early differentiation of pluripotent stem cells is accompanied by a decrease in acetyl-CoA produced by glycolysis. 68 In fact, it has been shown that glycolysis can improve the reprogramming efficiency of human and mouse fibroblasts. 69,70 In addition, studies have shown that ACL and cytosolic acetyl-CoA play a crucial role in the acetylation of histones, with important implications for stem cell differentiation. [71][72][73][74] Thus, acetyl-CoA availability can directly affect histone acetylation and serve as a metabolic signal to regulate cell fate decisions.
Histone (de)acetylation and (de)methylation are highly dynamic processes regulated not only by metabolic substrate availability, but also the metabolic enzymes themselves. Histone acetyltransferases (HATs) and histone deacetylases (HDACs) are two families of enzymes with opposing effects. 75 HATs catalyse the transfer of acetyl groups from acetyl-CoA to histone lysine residues, while HDACs remove acetyl groups from histone lysine residues. To date, 18 genes have been found to encode for HDACs in the mammalian genome.
They are divided into four classes, each with different subcellular localizations and specificities. Class I HDACs include HDAC1, HDAC2, HDAC3 and HDAC8, which are mostly located in the nucleus. Class I HDACs often interact with different cofactors to form multiple repressive complexes such as Sin3A, NuRD and CoREST. 76 Class II HDACs include HDAC4, HDAC5, HDAC6, HDAC7, HDAC9 and HDAC10, and they can shuttle between the nucleus and cytoplasm in response to different cell signals. Class III HDACs are comprised of the prominent SIRT1-7 sirtuin proteins. Class IV has only one member, HDAC11. 77 Class I, II and IV HDACs rely on Zn 2+ to function, whereas Class III HDACs require NAD + as a necessary cofactor.
Amongst them, SIRT1, SIRT2, SIRT6 and SIRT7 are located in the nucleus and can deacetylate specific histone residues.
Similarly, histone methyltransferases (HMTs) and histone demethylases (HDMs) are also a pair of enzyme families with opposing effects.
HMTs are a class of enzymes that mediate the methylation of histone lysine or arginine residues. These enzymes are highly selective for the histone residues they target and are divided into two types: arginine methyltransferases (PRMTs) and lysine methyltransferases (KMTs). To date, more than 50 human KMTs have been reported. Based on the catalytic domain, KMTs are further divided into two families: SET domain-containing KMTs, which include the Su(var)3-9, Polycomb and Trithorax proteins, and the non-SET domain-containing KMTs, such as DOT1-like proteins. [78][79][80] In contrast, lysine demethylases (KDMs) generally consist of two families: the KDM1 family of FAD-dependent amine oxidases and the JmjC domain containing family of dioxygenases, which are enzymes that depend on Fe 2+ , ascorbate (vitamin C) and 2-oxoglutarate (α-ketoglutarate, α-KG) as cofactors. 81 Ascorbate and 2-oxoglutarate have strong demethylation effects, which are used as metabolic agonists or cofactors of HDMs. [82][83][84][85] For example, in iPS reprogramming, the addition of ascorbate can enhance the expression of pluripotency genes, by removing H3K9me3 and H3K36me3 marks. 86,87 Studies in neural stem cells (NSCs) indicated that ascorbate can promote the removal of histone H3K9me3, H3K27me3 marks by JmjC domain-containing HDMs, to upregulate a series of dopaminergic neuron-specific genes. This increased the production of midbrain dopaminergic (mDA) neurons. 83 Studies of 2-oxoglutarate show that it can promote the self-renewal of naive ESCs through increased histone H3K27me3 demethylation. 84 Studies in mesenchymal stem cells (MSCs) derived from hESCs also showed that ascorbate and Fe 2+ can act synergistically as cofactors to promote histone demethylation by JmjC demethylases, including the removal of H3K9me3, H3K36me3 and H3K27me1 marks, thereby promoting MSC fate specification, long-term self-renewal and senescence resistance. 88

| Histone-modifying drugs
Small molecule drugs that regulate the activity of histone-modifying enzymes, often as modified analogues of enzyme-binding metabolites, also affect cell fate decisions. In particular, HDAC inhibitors (HDACi), HAT inhibitors (HATi), HMT inhibitors (HMTi) and HDM inhibitors (HDMi) have been found to play an important role in the regulation of cellular epigenetics and cell fate decisions. Below, we summarize some of the known effects of HDACi, HATi, HMTi and HDMi on cell fate determination.

| HDAC inhibitors
The most widely used HDACi can be generally divided into two categories based on their chemical structures and enzymatic activities: hydroxamates and fatty acids.
Hydroxamates were the first HDACi to be discovered. Because of their simple structures and powerful effects on HDACs as Zn 2+ chelators, they were widely used and studied. 89  genes. 92 In terms of somatic cell nuclear transfer (SCNT), TSA also shows great potential. It improves the efficiency of development to full-term state and favors the establishment of pluripotency of SCNT embryos by influencing histone acetylation status. 93,94 Panobinostat (LBH589) is another hydroxamate-based HDACi.
It acts on all Class I, Class II and Class IV HDACs, but it mainly acts on HDAC1/2/3/6. 89 The drug has been shown to increase the levels of CDKN1A (p21) and induce excessive acetylation of H3 and H4. 95,96 By promoting the accumulation of acetylated histones, it can induce cell cycle arrest and apoptosis. 96,97 However, studies have also found that low doses of LBH589 can induce terminal differentiation and irreversible mitotic arrest, but not cell death, in committed osteogenic progenitors. 98 Treatment with panobinostat upregulates osteogenic differentiation genes, including RUNX2, ALPL, BMP4 and SPP1.
Belinostat (Beleodaq or PXD101) belongs to a new class of hydroxamate-type HDACi, but it acts on the same targets as panobinostat. 89 In MCF-7 epithelial cells, belinostat inhibits cell proliferation by targeting the Wnt/β-catenin and PKC pathways. 99 Suberoylanilide hydroxamic acid (SAHA) is the first HDACi to be approved for clinical use on the market. It binds to the active site of Class I and Class II HDACs, with a predominant preference for Class I HDACs. 89 Like TSA, it was found that SAHA treatment could reduce senescence and improve self-renewal in MSCs. 100 SAHA can also affect the differentiation potential of MSCs by regulating the inflammatory response. 101 In synovium-derived MSCs of the temporomandibular joint, IL-1b-mediated upregulation of IL-6 and IL-8 inhibits MSC cartilage formation potential. SAHA can inhibit IL-1b-mediated upregulation of IL-6 and IL-8 and regulate the repair function of MSCs. In human neural progenitor cells, SAHA treatment can activate brain-derived neurotrophic factor (BDNF) mRNA expression, thereby promoting neural development and neurogenesis. 102 Valproic acid (VPA) is a branched-chain saturated fatty acid that comprises of a propyl substituent on a pentanoic acid stem. It is an inhibitor of Class I and IIa HDACs that has shown potent anti-tumour effects. VPA functions as an HDACi most likely by binding to the catalytic centre of its target HDACs, thus blocking substrate access. 89 Studies have found that low-dose VPA treatment can promote pluripotency in ESCs. 103 VPA induces a genome-wide acetylation of histone H3K9 in ESCs, thereby changing the chromatin state and promoting pluripotency. In C2C12 myogenic progenitors or myoblasts, VPA has long-term protective effects on myoblast survival, proliferation and differentiation by increasing histone acetylation. 104 In addition, VPA can also induce neurogenic differentiation of human adipose tissue-derived MSCs by activating canonical Wnt or non-canonical Wnt signalling pathways. 105 Sodium butyrate (NaBu or NaBt), a 4-carbon (4C) fatty acid, can be synthesized and absorbed naturally after microbial metabolism in the colon. NaBu is a non-competitive inhibitor of HDACs which acts mainly on HDAC2 and does not associate with the substrate-binding site. 106,107 NaBu has an important effect on the maturation of oocytes and the expression of developmental genes.
High concentrations of NaBu will hinder the meiosis of oocytes, but at low concentrations it can change the mRNA expression of developmental genes such as Sox2 and Oct4, thereby improving the embryo quality. 106 Previous studies have shown that NaBu can increase the expression of target genes by enhancing the acetylation of H3K9 and H4, thus promoting the differentiation of rat bone marrow-derived MSCs into smooth muscle cells. 107 NaBu can also promote the differentiation of satellite cells into myoblasts. This 4C fatty acid promotes the acetylation of genes that are conducive to muscle differentiation, such as Mef2 and MyoD, thereby promoting myogenesis. 40 In terms of iPS reprogramming, NaBu is more effective than TSA and VPA, and it can also promote the self-renewal of ESCs and reduce their differentiation. However, ESCs are very sensitive to NaBu, and it will only promote self-renewal within a narrow concentration range, inducing differentiation at higher concentrations. 108 The physiochemical principles underlying these preferences and quantitative relationships remain unclear.

| HAT inhibitors
HATi are divided into three classes. Class I HATi are bisubstrate inhibitors, but they are not commonly used at present. Class II HATi are natural compounds, such as curcumin and garcinol. Class III HATi are synthetic compounds, which are more specific than natural compounds, such as C646. 109 Garcinol is a natural compound isolated from Garcinia indica, a plant in the mangosteen family, and has been reported to inhibit p300 HAT via its binding to a non-active site region of p300, resulting in a conformational change that decreases the binding affinities of p300 to acetyl-CoA (uncompetitive inhibition) and histones (competitive inhibition). 110,111 Early studies have shown that garcinol can increase the ability of hematopoietic stem cells (HSCs) to expand in vitro and enhance their potential for homing to bone marrow by reducing the level of p53 acetylation. 112,113 It is worth noting that garcinol is also an effective neuroprotector, which enhances neuronal survival through the ERK signalling pathway. 114 Curcumin is a natural compound isolated from different Curcuma species of plants and inhibits p300 HAT activity in the same way as garcinol. 109 Studies have found that the addition of curcumin can inhibit the specificity of the cardiac lineage and the expression of cardiac muscle regulators in the early stages of cardiac differentiation.
Early administration of curcumin inhibits ~ 94% of cardiomyogenesis by inhibiting the transcription and expression of GATA4 and MEF2C. 115 Curcumin also regulates NSC fate. It can induce neurogenesis, synapse generation and cell migration in adult brain-derived NSCs in vitro. 116 C646 is a synthetic p300 inhibitor, which shows the highest potency by competitively inhibiting acetyl-CoA binding and non-competitively inhibiting the binding of H4-15 peptide substrate. 109 In MSCs, C646 blocks p300 HAT activity effectively and delays aging by inhibiting the p53-p21 signalling pathway. 117 During the differentiation of hematopoietic cells, because C646 inhibits the interaction between p300 and GATA1 and reduces GATA1 acetylation and transcription activity, it significantly inhibits the erythroid differentiation mediated by EDAG. 118  can also enhance the myocardial differentiation potential of BM-MSCs. 125 In this process, BIX01294 can enhance the proliferative capacity of myocardial progenitor cells without compromising their ability to function as myocardial progenitor cells during myocardial repair. 126 Compared with BIX01294, UNC0638 is an inhibitor with lower toxicity and higher efficacy and specificity for G9a. UNC0638

| HDM inhibitors
Overexpression of HDMs and histone demethylation is a common theme in cancers, but loss of function occurs less frequently, thus motivating the search for HDM inhibitors to treat cancer cells. 81 Studies have shown that neomorphic IDH1/2 mutations can reduce α-KG production and increase production of the α-KG analogue 2-hydroxyglutarate (2-HG) instead, resulting in the inhibition of JmjC demethylases and genome-wide changes in histone methylation. 85 Many researchers have used the α-KG dependence of the JmjC demethylases to design HDM inhibitors. There are two main skeletons for inhibitors of α-KG-dependent enzymes: N-oxalylglycine (NOG), an α-KG mimic that binds to the Fe 2+ cofactor but is resistant to superoxide attack, and para-2,4-dicarboxylic acid (2,4-PDCA), another α-KG mimic that occupies the α-KG binding site but which cannot complete catalysis. 130

| DNA-MODIF YING DRUG S
Beyond histones, the most critical part of chromatin would be the DNA double helix. According to the manner in which they bind to DNA, drugs that regulate DNA conformation can be divided into three types: intercalation, groove binding and covalent binding.

| Intercalators
Intercalation refers to the insertion of molecules between DNA bases. Since intercalators do not break the DNA, they have a limited impact on DNA damage. According to their different molecular skeletons, intercalator drugs can be roughly divided into carbazole drugs, anthracycline drugs and acridine drugs.
Carbazole-based drugs: curaxins are well-known carbazole-based drugs. The curaxin skeleton consists of electron-withdrawing groups at positions 3 and 6 of the carbazole core, such as nitrosyl or carbonyl functional groups, and the aminoalkyl chain bound to N9 (Table 1).

Computer simulations, circular dichroism and DNAse I footprinting all
show that the carbazole group of CBL0137 is inserted between the bases of DNA, 135,136 which greatly increases the distance between the base pairs and unwinds the DNA double helix. The symmetrical side chains of the carbonyl group containing C3 and C6 extend into the major groove of DNA, while the N9 side chain is inserted into the minor groove of DNA. By using different concentrations of CBL0137, the dissociation constant (K d ) of CBL0137-DNA binding was found to be 40 ± 20 μM, where each nucleosome bound 2-5 CBL0137 molecules. In addition, the insertion of CBL0137 into DNA changes it from B-DNA to Z-DNA. 136 At the same time, it also destabilizes the nucleosome, separates the H2A-H2B dimer 136 and evicts histone H1 from the nucleosome. 51 Hi-C analysis further confirmed that CBL0137 can remodel the 3D structure of the genome, causing the TAD (topologically associating domain) boundary to be partially disrupted and the chromatin circle to disappear. 50 The resultant decrease in distances between promoters and enhancers activates gene expression. 137 CBL0137-induced binding of FACT (facilitates chromatin transcription) and dissociation of CTCF (CCCTC-binding factor) from insulator sites may be other reasons for these observed changes in 3D genome organization. 50 In conclusion, CBL0137 neither inhibits topoisomerase II nor causes damage to DNA, but causes 3D chromatin remodelling after DNA intercalation. These changes result in a variety of cell signalling cascades, such as p53 and NF-κB, leading to mitotic arrest or apoptosis of cancer cells. 50,51 Anthracycline drugs: The basic structure of anthracycline drugs consists of a four-ring unit connected to a sugar (see Table 2), which with histone H4-Arg residues for binding of the DNA minor groove. 139 Another study showed that Dox can evict H2A from chromatin and evict H2B from the nucleus to the cytoplasm. 140  Experiments using a viscometer system that measures changes in PM2 DNA hypervolume confirmed the hypothesis that the planar ternary acridine ring is inserted between the bases of DNA, 144 which increases the distance between bases and changes DNA topology. Taking quinacrine as an example (see Table 3), besides the acridine ring inserted between the bases, its C9 nitrogen-containing side chain can interact with the DNA minor groove to stabilize the binding of quinacrine to DNA. 145 According to assays for the phosphorylation of histone H2A.X, quinacrine does not induce DNA double-stranded breaks. However, quinacrine will activate p53, not via the phosphorylation of p53, 146 but likely by loosening chromatin and capturing FACT to inhibit NF-κB activity. 147

| Groove binders
Groove binders non-covalently bind to the DNA major and minor grooves through electrostatic interaction forces, van der Waals forces and hydrogen bonding. They are another class of drugs that do not damage DNA. Major grooves have multiple interaction sites to provide a relatively strong possibility of binding to drugs, together with an easily accessible channel for large molecules. 148 Minor grooves have fewer joints and smaller sizes, but they are usually tension-free, so they are also suitable targets of small molecule drugs. 149 To match the structure of their binding site, the minor groove drugs are generally crescent-shaped. The more typical minor groove drug distamycin A is rich in amide bonds, which easily form hydrogen bonds with bases. At the same time, its positive charges can interact with DNA phosphates through electrostatic forces, while the remaining structure is in a small trench. 148

| Covalent binders
Covalent binding of small molecules to DNA is often irreversible and therefore usually highly toxic. The alkylating agent is one of the uni-  [150][151][152] and they are all highly toxic.

| Mechanistic Model
Whether the targets of these small molecules are DNA or histones, the end result is often a change in chromatin conformation ( Figure 2). Therefore, it is essential to understand the detailed mechanism of how changes in chromatin conformation are effected to regulate cell fate determination, to clarify and improve the effects of these drugs.
Normally, chromatin is in a highly compressed state, 24 Figure 3B), eventually leading to stable changes in specific cell fates. [154][155][156] Although this process has been confirmed in myogenesis and neurogenesis mediated by MyoD and Ascl1 respectively, 157,158 there is still no consensus on the mechanism of how pioneer factors achieve initial binding, while the DNA is still hidden amongst nucleosomes.
There are currently two plausible explanations that are distinct but For example, the structure of FoxA's winged-helix domain is highly similar to linker histone H1, and the eviction of H1 can facilitate the opening of condensed chromatin. 161 Interestingly another metabolic modification-ADP-ribosylation-can also reduce the affinity of H1 for DNA, to evict H1 and ready the nucleosomes for loosening. 162 HMG proteins such as HMGA and HMGB likely play a role in the eviction of H1 linker histones after ADP-ribosylation. 163 Regardless of the specific mechanisms, after nucleosome loosen- recruits HATs to assist SWI/SNF localization and achieve regional remodelling of nucleosomal DNA, 164 and further investigation showed that histone H4 hyperacetylation is necessary for this process. 165 This mechanistic model provides a concise and qualitative expla-

AUTH O R CO NTR I B UTI O N S
ZY, YC, WC and N.S-C. analysed, interpreted and wrote the manuscript. All authors read and approved the final manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
None available.