TET-dioxygenase deficiency in oncogenesis and its targeting for tumor-selective therapeutics

Abstract

TET2 is one of the most frequently mutated genes in myeloid neoplasms. TET2 loss-of-function perturbs myeloid differentiation and causes clonal expansion. Despite extensive knowledge regarding biochemical mechanisms underlying distorted myeloid differentiation, targeted therapies are lagging. Here we review known biochemical mechanisms and candidate therapies that emerge from this. Specifically, we discuss the potential utility of vitamin C to compensate for TET-dioxygenase deficiency, to thereby restore the biochemical function. An alternative approach exploits the TET-deficient state for synthetic lethality, exploiting the fact that a minimum level of TET-dioxygenase activity is required for cell survival, rendering TET2-mutant malignant cells selectively vulnerable to inhibitors of TET-function.

Cytosine, one of the 4 nucleotides in our DNA, when present in the genome in a CpG configuration (C preceding G in cis-), is often modified by methylation of its C-5 position. This modification of cytosine contributes critically to the structure and assembly of chromatin (DNA and the histone proteins around which it is organized) and to transcription regulation [1]. The cytosine methylation is catalyzed by DNA cytosine-5-methyltransferases (DNMTs) which transfer a methyl group from S-adenosyl methionine (SAM) to the cytosine to form methylcytosine (5mC). Methylation of DNA in this way is associated with silencing of gene transcription. Thus, DNA methylation is an epigenetic process central to gene regulation. In humans, 60–80% of CpGs are methylated in somatic cells as a default state [2]. What about demethylation? It was recognized several decades ago that removal of methyl groups from cytosines in DNA was a passive process, that is, demethylation was by failure to recapitulate methylation marks on a parental DNA strand onto the newly synthesized DNA strand during S-phase. Only last decade, however, the mechanism was identified by which specific cytosine residues were identified for demethylation [3, 4]. Ten eleven translocation genes TET1, TET2, and TET3 encode for iron(II) and α-ketoglutarate (α-KG) dependent DNA dioxygenases that sequentially oxidize 5mC ➔ 5hydroxy methyl cytosine (5hmC)➔ 5-formyl cytosine (5fC)➔5-carboxyl cytosine (5caC), and facilitate the removal either via thymine DNA glycosylase (TDG)-mediated base excision repair of 5fC and 5caC, or inability of DNMTs to recognize 5hmC during replication. In this way, therefore, TET-dioxygenase activity mediates mCpG demethylation. Loss-of-function mutations in TET2 are amongst the most frequently observed lesions in hematologic malignancies, including myelodysplastic syndromes and acute myeloid leukemia [5], and as expected, the TET2 loss-of-function is associated with CpG hypermethylation compared to normal myeloid precursors [6].

Structure-Function of TET-dioxygenases

Each member of the TET family possesses a highly conserved and well characterized C-terminal catalytic domain. The catalytic domain consist of a conserved cysteine-rich and a double-stranded β-helix (DSBH) regions along with a stretch of so called low complexity structure dispensable for catalytic activity [4, 7–13]. The iron(II), α-KG and substrate binding sites are imbedded in these cysteine-rich and DSBH regions and are critical for dioxygenase function [8–10, 14]. Although the enzymatic domain is highly conserved across species, including human, mouse, Xenopus, and zebrafish [13, 15, 16], the N-terminal regulatory domains of TET1, TET2 and TET3 greatly differ between one another and between species, with a maximum amino acid similarity of less than 13% among TET-dioxygenases (Table 1). TET1 and TET3 have a cysteine-X-X-cysteine (CXXC) zinc finger domain at the N-termini, however, this zinc finger motif is absent in TET2 [16–18]. The CXXC motif has been shown to be critical for binding to DNA segments containing modified (e.g., 5mC, 5hmC, 5fC, and 5caC) as well as unmodified cytosines [16, 18]. Interestingly, due to the lack of the CXXC domain, it has been reported that TET2 may target DNA via other DNA binding adaptor proteins, such as SNIP1 [19, 20]. In short, how TET2 is targeted to specific genetic loci is not fully understood, and likely involves adaptor proteins.

Table 1.

Similarity among TET Proteins compared to human TET2

Proteins*	Full Length	N-terminus	C-terminus
Homo Sapiens	100	100	100
Pan troglodytes	99.2	99.1	99.3
Macaca mulatta	96.8	95.4	98.4
Bos Taurus	84.6	79.4	91.5
Gallus gallus	64.4	49.9	82.7
Rattus norvegicus	63.7	60.6	67
Mus musculus	62.7	60.4	65.3
Xenopus tropicalis	51	31.7	70
Danio rerio	38.5	21.7	55.7
Homo Sapiens TET3	29.9	12.7	47
Homo Sapiens TET1	25	9.8	46.4

^*The protein sequences were obtained from the NCBI database with protein accessions: NP_001120680.1, XP_003310448.3, XP_014994413.2, XP_010804239.1, XP_015131709.1, XP_006224326.1, NP_001035490.2, XP_002934823.2, XP_005159960.1, NP_001274420.1 and NP_085128.2 for indicated species. The sequences were aligned against human TET2 using COBALT (https://www.ncbi.nlm.nih.gov/tools/cobalt/re_cobalt.cgi) to identify the corresponding N-terminus and C-terminus for each protein. Identity of fragments were calculated by DNASTAR MegAlign Version 7.1 using Clustla W.

Structural determinants of TET-dioxygenase activity have been extensively studied via several high resolution crystal structures [7, 9, 10]. Studies of the structure of the engineered catalytic domain of human TET2 showed that the cysteine-rich and the DSBH regions are stabilized by three zinc atoms that allow these structures to form a compact fold in which DNA substrate is embedded above the DSBH. TET2 structures in complex with DNA substrate demonstrated that the methylated cytosine is located in the catalytic cavity with orientation towards catalytic iron(II) and pseudo substrate N-Oxalylglycine (NOG) (Figure 1) [9, 10]. TET-dioxygenase activity is a two-step reaction that utilizes iron(II) and α-KG. The first step is the activation of molecular oxygen that involves Fe (II) and α-KG to convert the dioxygen into a highly active iron(IV)-oxo intermediate transition state complex which then oxidizes the C-H bond to C-OH in the second step [9, 17, 21]. A key aspect of all dioxygenases is the conserved nature of residues involved in binding to the α-KG cofactor and to the metal ion (Table 2). A detailed ligand plot analysis of different classes of DNA [9], RNA [22], protein [23], and small molecule [24] dioxygenases demonstrate that cofactor binding sites are similar in disparate classes of α-KG/iron(II) dependent enzymes (Figure 1 and Table 2). Interestingly, two histidine, one arginine, or lysine and one glutamic acid or aspartic acid are always conserved and seem essential for binding of α-KG and the divalent metal ions. However, the mechanism of substrate recruitment and binding are significantly different from one class of dioxygenase, e.g., TET-dioxygenases, to another, e.g., histone demethylases [21].

Figure 1. The catalytic site residues in a-KG/iron(II) dependent dioxygenases are conserved.

Ligand plot of the catalytic sites of DNA dioxygenase TET2, RNA dioxygenase FTO, protein dioxygenases PHD2, P4H and JMJD2A and small molecule dioxygenase BBOX1. Analysis was performed using LIGPLOT v.4.5.3 (https://www.ebi.ac.uk/thornton-srv/software/LIGPLOT/). Three dimensional structures with PDB ID 4NM6 (TET2), 5L9B(PHD2), 4IW3 (P4H), 2OQ6 (JMJD2A) and 3O2G (BBOX1) were used for the analysis.

Table 2.

Residues essential for substrate, metal and the cofactor αKG binding in the catalytic domain

Protein*	Substrate	α-KG binding site		Metal binding site
TET2	5mC	R1896		H1382	H1881	D1384
PHD2	Proline	R383		H313	H374	D315
P4H	Proline	R192	Y141	H124	H183	D126
JMJD2A	Lysine(Me)3	Y132	K206	H188	H276	E190
BBOX1	Small Molecule	R360		H202	D204

^*The crystal structure of proteins with cofactors were obtained from the RCSB protein data bank (https://www.rcsb.org/) with ID 4NM6 (TET2), 5L9B(PHD2), 4IW3 (P4H), 2OQ6 (JMJD2A) and 3O2G (BBOX1).

Studies using knock-out models have demonstrated that TET-dioxygenase actions on CpG methylation control patterns and efficiencies of gene transcription in a manner that dictates lineage-commitment, maturation within a lineage, and accordingly, also proliferation [25–27]. Specifically TET2 haploinsufficiency or knock-out from the myeloid compartment did create mutator phenotype and accelerated neoplastic evolution [5, 28–30].

Configuration of TET2 mutations

Myeloid oncogenesis selects for truncating frame shift deletions or stop gain missense mutations in TET2, that account for more than 80% of all TET2 mutations (TET2^MT) [28, 31]. Most missense mutations are localized in or near the catalytic site (Figure 2A). TET2 mutations are found in myeloid malignancies in all possible configurations, mono allelic, bi-allelic, heterozygous, hemizygous, and homozygous.

Figure 2. Distribution of missense and truncating mutations in TET2.

A. Configuration of truncating and missense mutation in TET2 among 200 myeloid neoplasia patients (data source: Hirsch C., et al 2018). B. Distribution missense and truncating mutations among 11664 patients across 28 studies in hematologic malignancies and 67743 patients across 206 studies of solid tumors (data source: www.cbioportal.org). Percentage of different mutations were plotted.

Consistent with the experimentally demonstrated role of TET2 in hematopoiesis, the loss-of-function mutations in myeloid malignancies are linked with a differentiation block and proliferative advantage, hence increasing fitness of mutant over normal cells [5, 6, 32–37]. Studies of variant allele frequencies in patient bone marrow and blood indicate that TET2 loss-of-function mutations increase ectopic replications in several hematopoietic lineages and at different maturation stages within these lineages [6, 28, 31, 38–43]. Diminished TET activity has also been reported in melanoma, glioblastoma, and breast cancer, wherein, it seems to correlate with advanced disease and poor survival [44]. However, the frequency of loss-of-function truncating mutations in solid tumors is far less than in hematologic malignancies (Figure 2B). In chronic myelomonocytic leukemia (CMML), the prevalence of TET2^MT approaches 50% with almost half being biallelic alterations [28, 31, 35]. TET2^MT are ancestral (the first driver mutation in the malignant clone) in nearly 50% of cases and their incidence increases with age. TET2^MT prevalence is >70% in MDS patients over the age of 80 [28]. TET2^MT are also found in aging “healthy controls” in whom their detection in otherwise asymptomatic patients is referred to as ‘clonal hematopoiesis of indeterminate potential (CHIP)’. The prevalence of CHIP in healthy adults strongly associates with an increased risk of subsequently developing hematological malignancies [28, 45, 46]. Altogether, these findings demonstrate that TET2 is a key player in myeloid oncogenesis [47].

Although the TET2 gene has been widely sequenced in large cohorts of patients in leukemia and somatic mutations frequently associated with disease occurrence and severity, evidence of inherited TET2 mutations have only been reported recently [48–50]. A Finnish family with multiple cases of lymphoma segregating with a heterozygous germline TET2 frameshift variants were reported [49, 50]. In these case reports, neither unusual predisposition to atherosclerosis nor abnormal pro-inflammatory cytokine or chemokine expressions were observed in these patients. These observations were in cells from three additional unrelated TET2 germline mutation carriers. The TET2 defect elevates blood DNA methylation levels, especially at active enhancers and cell-type specific regulatory regions with binding sequences of master transcription factors involved in hematopoiesis. Interestingly, a recent case report of a French pedigree harboring germline TET2 frameshift variant found that the three sibling developed myeloid malignancies at early age [48].

Post-translational modifications in TET2

Enzyme activity depends on the three dimensional structure of the catalytic site, with contributions also from non-catalytic domains that maintain stability in a cellular milieu. Apart from mutations, post-translational modifications, such as phosphorylation [51–58] and acetylation, have been reported to impact TET2 stability and activity [59–69]. The consequences of posttranslational modifications depend on the type of modification and the residue modified [52, 59, 63]. Acetylation of the C-terminal catalytic domain lysine residues cause loss of TET-dioxygenase activity [59, 61, 62], due to the changes in the conformation of the catalytic site that prevent substrate recognition or cofactor binding [10, 11, 59]. N-terminal regulatory domain lysine acetylation prevents ubiquitination and protects TET2 from proteasome mediated degradation and increases its half-life in cells [59, 63]. Interestingly, several oncogenic mutations target the post-translationally modified residues to structurally mimic function suppressing acetylation, e.g., by causing lysine to glutamic acid or glutamine substitution [59] (Figure 3).

Figure 3. Residues target for post translational modifications.

A. Schematics of human TET2 and residues target for post translational modification and their effect on the activity. B. Sequence alignment of TET dioxygenases showing the conserved nature of these residues.

Similar to N-terminal lysine acetylation, phosphorylation of N-terminal serine (S99) by AMP-activated kinase (AMPK), or phosphorylation of tyrosines (Y1939 and Y1964) by cytokine receptor-associated kinase JAK2, prolong TET2 protein half-life and in this way increase dioxygenase activity [53, 58]. AMPK is a nutrient sensor usually inactivated in diabetic patients due to high glucose levels. The phosphorylation of S99 in TET2 by AMPK inversely correlated with glucose concentration [58]. Accordingly, activation of AMPK by the anti-diabetic drug metformin augmented TET2 half-life resulting in genome wide increased 5hmC levels even under high glucose conditions [58]. These findings indicated that low levels of 5hmC observed in high glucose conditions are a consequence of the inhibition of AMPK-mediated TET2 phosphorylation and hence TET2 degradation [58].

Similarly, it is reported that hematopoietic cytokines can phosphorylate TET2, leading to its activation in erythroid progenitors in vitro. Specifically, cytokine receptor-associated kinase JAK2 phosphorylated TET2 tyrosines (Y1939 and Y1964), which promoted the interaction between TET2 and the erythroid lineage master transcription factor KLF1 [53]. This interaction between TET2 and KLF1 was further increased by exposure to erythropoietin. Contrary to the loss-of-function mutations in TET2 that reduce the global 5hmC in myeloid malignancies, JAK2V617F mutation seen in myeloproliferative neoplasms (MPN) increases TET-activity and decreases genome-wide cytosine methylation [53]. These findings suggest avenues, that is, kinases, by which TET-dioxygenase function may be regulated [53].

Therapeutic targeting of TET-dioxygenases in myeloid neoplasia.

Thus, TET2 is a bona fide tumor suppressor gene (TSG), with its partial or complete loss-of-function playing a driver role in myeloid oncogenesis [70–72]. In considering how to turn this understanding into therapy, it is worth bearing in mind that the malignant cells, despite partial or complete loss-of-function of TET2, contain and express two other TET-dioxygenases, TET1 and TET3, that provide residual TET-dioxygenase activity likely critical to cell survival [73]. Thus, two approaches for preferential targeting of TET-dioxygenase deficient cells can be conceptualized. The first, to compensate for the TET2 loss by augmenting the activity of remaining TETs, by utilizing either a ‘super’ substrate or co-substrate, e.g., vitamin C and/or α-KG analogs that are known to increase TET-dioxygenase activity. The second, to drive already reduced net TET-dioxygenase activity below minimum thresholds needed for cell survival (‘synthetic lethality’).

(i). Augmenting the activity of residual TET-dioxygenase activity

Ascorbic acid (AA, vitamin C) is a water-soluble ketolactone with two ionizable hydroxyl groups. It has two pKa values, pK1 is 4.2 and pK2 is 11.6; thus, the ascorbate monoanion is the dominant form at physiological pH. Ascorbate is an excellent reducing agent and readily undergoes two consecutive, one-electron oxidations to form ascorbate radical and dehydroascorbic acid (DHA). The ascorbate is relatively unreactive due to resonance stabilization of the unpaired electron; it readily dismutes to ascorbate and DHA (k_obs = 2×10⁵ M⁻¹ s⁻¹, pH 7.0)[74]. One of the main physiological function of AA is to maintain physiological redox balance [75]. Thus, AA acts as a specific co-factor for a large family of enzymes known as the iron(II)- and 2-oxoglutarate-dependent dioxygenases [76]. In recent years, several high profile studies reported that known as vitamin C enhances TET-dioxygenase activity by reducing the Fe moiety in the catalytic site from iron(III) to Irion(II) [14, 59, 77]. Several studies suggested that vitamin C acts as a broad spectrum augmenter of the activity of all α-KG/Irion(II)-dependent dioxygenases, including TETs, and therefore may have therapeutic utility in TET2-mutated neoplasms[75, 76]. Recent reports using in vitro cell culture and in vivo murine model systems have postulated that the treatment of TET2 deficient cells with vitamin C can restore TET-dioxygenase activity, block aberrant self-renewal, and prevent neoplastic evolution [14, 59, 78–80]. In interpreting these studies, there are two important considerations, first, that vitamin C has anti-cancer properties independent of its effects on TET-dioxygenase, e.g., by perturbing the redox balance in cancer cells [73–76]. The second is that vitamin C mediated augmentation of TET-dioxygenase activity (EC50 ~14μM) is saturated at vitamin C concentrations of >100 μM [14, 59]. However, the intracellular levels of vitamin C remains above 1 mM in most of the circulating cells (Table 3). For example, ~500 mg/daily dietary ascorbic acid supplementation routinely achieve 3.5 mM in monocytes and 4.0 mM in lymphocytes[81]. Therefore, the antileukemic effect of the supra physiological doses of vitamin C humans [82] and in mice leukemia model [78] remains unclear. A comprehensive analysis of the cellular concentration of vitmin C suggested that the intracellular levels of ascorbic acid remains in the milimolar range[81], well above the saturating concentration required for activation of majority of αKG/iron(II) dependent dioxygenases [21, 76]. The most recent National Health and Nutrition Examination Survey (NHANES)[83] of healthy adults living in the United States suggested that the overall age-adjusted mean from the square-root transformed (SM) concentration was 51.4 μM (95% CI: 48.4, 54.6). Interestingly, the highest concentrations were found in children (age <18) and older persons (age >60)[83] (Table 4). Taking these two factors together, it is possible that anti-cancer effects observed in mice and humans, using very high doses of vitamin C, may not actually be mediated via augmentation of TET-dioxygenase activity. Interestingly, mice manufacture vitamin C endogenously (vitamin C is not a ‘vitamin’ for mice), in this way differing from humans. A recent case report [82] indicated that an acute supraphysiological dose of vitamin C, as a single agent, benefitted a patient with chemorefractory TET2 mutant AML. In this case report the intravenous vitamin C dose was gradually increased from 35 g/day to 95 g/day 2X/week over a period of ~4 weeks. However, for the reasons mentioned above, it cannot necessarily be concluded that the mechanism of action was by augmentation of TET-dioxygenase activity, and TET independent effects of vitamin C may have also been in play [82].

Table 3.

Intracellular concentration of ascorbic acid (AA) in circulating cells achieve at 500 mg daily supplement

Circulating Cells	Ascorbic acid (mM)
Lymphocytes	4.0
Monocytes	3.0
Platelets	3.5
Neutrophils	1.2

EC50 for majority of aKG and Iron (II) dependent dioxygenases are < 0.1 mM [75, 76]

Table 4.

Serum vitamin C concentration in healthy individuals from 2003–2004 National Health and Nutrition Examination Survey [83]

Age, year	Ascorbic acid, median, (95% CI), μM
	Male	Female
>6	48.0 (44.9, 51.2)	54.8 (51.6, 58.0)
6–11	73.5 (69.8, 77.3)	68.9 (64.7, 73.3)
12–19	50.7 (47.2, 54.2)	54.88 (50.6, 59.2)
20–39	42.0 (37.6, 46.6)	48.88 (44.5, 53.3)
40–59	43.2 (38.9, 47.6)	52.08 (47.5, 56.7)
>60	52.7 (48.8, 56.4)	62.98 (60.5, 65.4)

The ability of vitamin C to augment TET-dioxygenase activity also depends on the acetylation status of catalytic domain lysine residues [61–63]. That is, the suppression of TET-dioxygenase activity that is produced by acetylation of catalytic domain lysines is not restored by increasing vitamin C concentrations [59, 62]. From a translational perspective, these observations are important because several TET2 mutations mimic these post-translational modifications, and vitamin C is not expected to increase the TET-dioxygenase activity of these mutants [59, 62]. The conditions or triggers for TET family acetylation or deacetylation have not as yet been characterized, although in our studies of leukemia cell lines, we found that acetylation appears to be the default setting.

Thus, the effect of vitamin C is dependent on whether baseline concentrations are below levels that saturate TET-dioxygenase activity, and the complex interplay of different types of post translational modifications that impact dioxygenase activity and the ability of vitamin C to affect this. One implication is that pharmacologic modulation of acetyltransferases and histone deacetylases may regulate TET dioxygenase-dependent effects of vitamin C, creating a rationale for vitamin C combinations with class I/II HDAC inhibitors [59, 63] or sirtuin activators [59, 62]. Since the overall pharmacodynamic goal is to decrease aberrant DNA methylation, combinations with DNA hypomethylating agents, e.g., azacytidine [64, 84] and decitabine [85], can also be proposed.

(ii). Inhibiting TET-dioxygenases in TET deficient cells

Basal TET-dioxygenase activity is essential for the activation of several 5mC-sensitive transcription factors (reviewed in [86]), including Myc [20], and Runx1 [87]. Thus, TET2-deficient leukemia cells rely on the remaining TET-dioxygenase activity provided by TET1 and TET3, as evidenced by the persistence of hydroxymethylation in cells with biallelic inactivation of TET2 in human leukemias and in Tet2^−/− mice [88]. In addition, TET3 has been shown to have overlapping function with TET2 in hematopoietic progenitors [73]. Knockout of all three Tet genes is embryonically lethal in mice [73, 89–91]. In zebrafish, conditional knockout of all three TET-dioxygenases in HSC abrogated hematopoiesis [89]. In humans with myeloid malignancies, mutations in IDH1 or IDH2, that produce the oncometabolite R-2-hydroxyglutarate (R-2HG) that antagonizes α-KG, are notably mutually exclusive with TET2 mutations. There are two potential reasons for this. One is that by inhibiting TET-dioxygenase activity (via the oncometabolite that competes with α-KG that is a mandatory cofactor for TET-dioxygenases), IDH mutations and TET mutations are in the same pathway and thus redundant. Another potential reason is that 2HG producing neomorphic IDH mutations and TET mutations are synthetic lethal, by reducing TET-dioxygenase activity beneath minimum thresholds needed for cell survival[92]. Our own data supports the latter reason, since we have found that introducing neomorphic IDH mutations into TET2-mutant cells causes cell death[93]. Altogether, these observations suggest that TET2 mutant malignant cells may be more vulnerable to inhibition of residual TET-dioxygenase activity, essential for cell survival, than normal dividing cells containing the usual TET family complement (therapeutic index). We accordingly are developing TET-specific inhibitors (Figure 1, Table 2) - an extensive analysis of the structural elements of the substrate recognition motif demonstrate a large structural diversity among different class of dioxygenases that can be exploited to developed inhibitors specific for TET-dioxygenases [21], with potential application to simultaneously expand normal hematopoietic progenitors and suppress TET2-mutant malignant cells [94].

Conclusion and future prospect

A central challenge in developing novel strategy for oncotherapy is to terminate malignant but not normal replication needed for health and life. Increasingly detailed understanding of TET-dioxygenase biochemistry and roles in cell physiology has created opportunities to meet this challenge for TET2-mutated myeloid malignancies. TET2 mutation in many healthy elderly significantly increase the risk of developing myeloid neoplasms. The strategies that either circumvent the loss in neoplastic cells and induce terminal differentiation as well as the one that terminate the deficient clone without impacting normal hematopoiesis can be employed for a successfully to develop novel therapeutics[93]. For example, naturally existing neomorphic mutations in IDH1/2 causes TET-dioxygenase inhibition which in turn restrict the growth of TET2^MT hematopoietic clones in human body. Hence, establishing TET-dioxygenase deficient cells can be therapeutically targeted.