Using NMR to Monitor TET-Dependent Methylcytosine Dioxygenase Activity and Regulation

Abstract

The active removal of DNA methylation marks is governed by the TET family of enzymes (TET1–3), which iteratively oxidize 5-methycytosine (5mC) into 5-hydroxymethycytosine (5hmC), and then 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC). TET proteins are frequently mutated in myeloid malignancies or inactivated in solid tumors. These methylcytosine dioxygenases are α-ketoglutarate (αKG)-dependent and are therefore sensitive to metabolic homeostasis. For example, TET2 is activated by vitamin C (VC) and inhibited by specific oncometabolites. However, understanding the regulation of the TET2 enzyme by different metabolites and its activity remains challenging due to limitations in the methods used to simultaneously monitor TET2 substrates, products, and co-factors during catalysis. Here, we measure TET2-dependent activity in real-time using NMR. Additionally, we demonstrate that in vitro activity of TET2 is highly dependent on the presence of VC in our system and is potently inhibited by an intermediate metabolite of the TCA cycle, oxaloacetate (OAA). Despite these opposing effects on TET2 activity, the binding sites of VC and OAA on TET2 is shared with αKG. Overall, our work suggests that NMR can be effectively used to monitor TET2 catalysis and illustrates how TET activity is regulated by metabolic and cellular conditions at each oxidation step.

Graphical Abstract

Reversible DNA methylation is a fundamental mechanism of epigenetic programming, controlling gene expression¹. In particular, cytosine methylation at the C5 position, typically present on CpG sites at gene promoters, serves as a mark for transcriptional repression^{2, 3}. Inversely, demethylation of promoter regions corresponds to active transcription. This foundational process is involved in numerous signaling pathways, including the immune response, development, and stem cell differentiation³. Unsurprisingly, the enzymes controlling these processes are commonly deregulated in nearly all types of cancer¹.

The TET (ten–eleven translocation) family of enzymes (TET1–3) drives active DNA demethylation through an iterative oxidation process, where a hydroxyl group is first added to 5-methylcytosine (5mC), creating 5-hydroxymethylcytosine (5hmC). This oxidation is repeated to form two other intermediates, 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC), before they are removed by the base-excision repair (BER) system^4–6 (Fig.1 A). TET function is frequently lost in several different types of cancer through different mechanisms. For example, in blood cancers, TET activity is reduced because of loss-of-function mutations^{4, 7 8, 9 10}. In contrast, in solid tumors, TET genes aren’t mutated, but their expression is reduced or the protein is less active^11–20. Because iron and α-ketoglutarate (αKG) are essential cofactors for TET-dependent catalysis, and oxygen is a substrate, this process is sensitive to changes in metabolic conditions and the presence of oncometabolites in the cell. Supporting this notion, stimulation of glycolysis and glutaminolysis by intraperitoneal injection of glucose and glutamine, respectively, results in rapid increase of αKG and 5hmC in mouse livers²¹. In contrast, accumulation of itaconate (ITA) in LPS-activated macrophages or 2-hydroxyglutarate (2-HG) in tumor cells with mutant IDH1 or IDH2, two metabolites that are similar in chemical structure to αKG, results in inhibition of TET activity through competition with αKG^22–24. To combat low enzyme activity of TET, treatment with vitamin C (VC)/ascorbic acid has been shown to improve both in vitro enzymatic activity with recombinant protein and in vivo TET-dependent immune response^{12, 25–31}.

Figure 1. NMR is a valuable technique to simultaneously monitor αKG, succinate, and the oxidation state of methylated DNA in TET-dependent reactions.

(A) Overview of TET2-dependent catalysis. TET2 iteratively oxidizes 5mC to ultimately form 5caC, which is removed by base excision repair. At each step, αKG is also converted to succinate. (B) Quantification of DNA intermediates during catalysis using NMR taken at 11-minute intervals. (C) Chemical structures of αKG and the TET2-dependent product succinate, indicating the protons observed in NMR. (D-E) Representative NMR spectra and quantification of the conversion of αKG to succinate over time either without (D) or with TET2 (E). Reactions were performed in at least duplicate. Errors bars: Standard Deviation (SD).

Several methods to comprehensively interrogate the basic catalytic mechanism of TET proteins and its substrate intermediates are currently available. TET activity can be monitored through ELISA-type of assays, which are typically capable of only monitoring one intermediate at a single time point^22–24. More comprehensive workflows, e.g., LC-MS/MS and bisulfite sequencing, can also be used to detect multiple substrate intermediates^32–35. However, with regards to the catalytic mechanism of TET2 with VC, it is unclear how VC improves TET function^29–31. While some reports have suggested that VC directly binds to TET^{29–31, 36}, others propose that the addition of VC reduces the free, nonenzyme-bound Fe³⁺ to Fe^{2+ 37}. Here, we set to overcome existing experimental hurdles and address some long-standing questions in the field by employing time-resolved NMR spectroscopy. Due to the distinct NMR chemical-shift signatures, we demonstrate the capability of NMR to simultaneously monitor nearly all components (i.e., substrates, cofactors, and products) of TET2-dependent catalysis in a time-dependent manner. Using this approach, we were able to directly assess the enzymatic activities of TET2, both in the presence and absence of DNA, dissect the mechanism of VC-dependent stimulation of TET2 activity, and identify an inhibitory metabolite of TET2 (oxaloacetate, OAA), providing an in-depth view of TET2-dependent methylcytosine dioxygenase activity and its regulation.

First, to unambiguously identify chemical-shift signatures of the DNA substrate and products, double-stranded DNA oligos with modified cytosines, based on the intrinsic sequence preference of TET2, were examined^{38, 39}. Indeed, the NMR peaks corresponding to 5mC, 5hmC, 5fC, and 5caC could be clearly and differentially observed in 1D ¹H NMR spectra (Fig. S1A). After purifying the catalytic domain of TET2 as previously described^{32, 33}, we used assay conditions containing equimolar Fe²⁺ and excess VC. Excitingly, we found TET2-dependent changes in the 1D ¹H NMR spectra representing all four DNA methylation states, as well as the concentrations of αKG and succinate in a time-dependent manner (Fig. 1B, Fig. S1B–C). Specifically, the 5mC was completely consumed by the 35-minute time point. The 5hmC and 5fmC, iteratively accumulated and reached their peak at the 25- and 100-minute time points, respectively. These intermediates reached a maximum of ~60% of the total DNA content before they began depleting, followed by a build-up of 5caC. Simultaneously, the six peaks (2.9–2.94 ppm representing H11/12 and 2.34–2.37 ppm representing H13/14) of the cofactor αKG and the single peak of the product succinate (2.31 ppm from H10/11/12/13) were readily observed, and their depletion and accumulation were TET2 dependent (Fig. 1C–E).

Having established a robust NMR-based TET2-dependent DNA oxidation assay, we set out to examine how the combination of cofactors (Fe²⁺ and αKG) and VC impact TET2 activity. By monitoring the oxidation of 5mC and αKG to succinate, we found that TET2 activity was the greatest when equimolar labile Fe²⁺ and excess VC were present (Figs. 2A, B), as evident by the almost complete oxidation of 5mC to 5hmC by the first time point. Interestingly, while the removal of both Fe²⁺ and VC from the reaction markedly reduced TET2 function, it is surprising that TET2 activity was effectively nil in reactions with only Fe²⁺ added without VC. On the other hand, the presence of VC in the absence of free Fe²⁺ was sufficient to promote the conversion of all 5mC to 5hmC and subsequently to 5fC and 5caC during a longer time course. Our data suggest that VC, at least in this albeit different workflow which initiates the reaction with TET2 instead of Fe²⁺, is imperative to our system by reducing TET2-bound iron during the catalytic cycle.

Figure 2. Vitamin C is imperative to activate recombinant TET2 in our system. (.

A) Similar to Fig. 1b, bar graphs indicating the oxidation state of methylated DNA over a time course, revealing that vitamin C is crucial to TET2 function. Each reaction was supplemented with Fe²⁺ and/or VC as indicated. (B) Similar to Fig. 1e, line graphs of αKG and succinate levels over time during reactions indicated in (A). (C-D) DNA increases the rate of conversion of αKG to succinate, compared to the uncoupled reactions without DNA, monitored by NMR. Reactions were monitored without DNA (left), with unmethylated DNA (center), or with methylated DNA (right) in the absence (C) or presence (D) of VC. Experiments were performed in at least duplicate. Errors bars: SD.

To interrogate the mechanism of VC-dependent stimulation of TET2 activity, we then examined how the reducing agent VC impacted αKG consumption in the presence of DNA substrates with and without 5mC modifications, where we specifically monitored the conversion of αKG to succinate in these reactions. Because our assays allow us to interrogate multiple cofactors at a single time in lieu of a substrate, we found that VC alone could significantly improve the conversion of αKG to succinate in a DNA-uncoupled reaction (Fig. 2C–D, Fig. S2). Furthermore, these reactions were noticeably enhanced by the presence of methylated or unmethylated DNA. Collectively, these data suggest that VC improves the DNA-uncoupled and coupled reactions and serves to reduce TET-bound Fe³⁺ to Fe²⁺ since these reactions lack labile Fe²⁺ in the solution.

Next, we expanded the application of our approach to examine the impact of small molecule inhibitors on TET2 activity. Here, we focused on OAA, an intermediate of the TCA cycle, like other TET2 inhibitors, which has been implicated in cancer ^{40, 41}. Since OAA is similar in chemical structure to αKG and has already been shown to inhibit other αKG-dependent enzymes⁴², we hypothesized OAA might exhibit certain inhibitory effects on TET2 (Fig. 3A). After incubating the reaction components (OAA, VC, and Fe²⁺) together with the substrate, TET2 was added right before data collection. Because TET2 was not preincubated with the other components, including OAA, the only activity (5mC to 5hmC) detected was immediately following TET2 addition but quickly ceased without performing further oxidation steps or the conversion of αKG to succinate (Fig. 3B–C, Fig. S3A–B). This inhibition was further confirmed by a traditional dot-blot assay where either genomic DNA or methylated DNA oligos were used as the substrate, and TET2 activity was monitored using the anti-5hmC antibody. In a dose-dependent manner, OAA was found to strongly inhibit TET2 activity (Fig. S3C)^{43, 44}. Hence, OAA might outcompete VC and αKG, restricting TET2 from completing its later reaction steps.

Figure 3. The inhibitor OAA and activator VC have overlapping binding sites with αKG.

(A) Chemical structures of VC, αKG, VC, OAA, and L-2HG. (B-C) OAA inhibits TET2-dependent DNA oxidation (B) and the conversion of αKG to succinate (C) as monitored by NMR, similar to Fig. 1b and 1e. (D-F) Saturation transfer difference (STD) NMR spectra, revealing αKG (d), VC (e) and OAA (f) all bind to TET2 based on the comparison between the difference reference (top) and the difference spectra (bottom). Binding is observed by the presence of peak signals in the difference spectra. (G-H) OAA prevents αKG (g) and VC (h) from binding to TET2. Reactions were performed in at least duplicate. (I) Curve fits to determine the IC₅₀ for OAA and L-2HG in the NMR-based TET2 activity assay. Open circles at endpoints are control experiments where positive control is the data from Figure 1B and negative control is the data from Figure 1D. Errors bars: SD.

To further examine whether VC and OAA directly interact with TET2, we performed saturation transfer difference (STD) NMR spectroscopy, a widely used technique for identifying weak small molecule-protein interactions. The STD spectra of αKG, VC, and OAA in the presence of TET2 strongly support their direct interactions (Fig. 3D–F). Using competition experiments, we showed that OAA and αKG compete for the same binding site on TET2 (Fig. 3G), similar to what we have observed previously for TET2-inhibitor ITA²⁴ and as expected based on other αKG-dependent enzymes. To test if OAA could also impede VC from binding to TET2, we performed a similar experiment monitoring the TET2-VC interaction in the presence or absence of OAA. Indeed, the TET2-VC interaction was lost because of the addition of OAA (Fig. 3H). Interestingly, we observed NMR STD signals of both αKG and VC in their direct competition assay, suggesting that αKG and VC may have comparable affinities for TET2. Next, we used NMR to determine the IC₅₀ of OAA and compare it to a well-established inhibitor 2-HG. In our experiments, we found that OAA inhibited TET2 ~25-fold better than 2-HG (Fig. 3I). Taken together, our results indicate that OAA potently binds and inhibits TET2 activity by competing with αKG, whereas VC binds the TET2 active site to promote its activity by reducing inactive Fe³⁺ back to active Fe²⁺. The relatively weak binding of VC and αKG could serve to enable them to alternatively enter TET2 active site for function, while the tight-binding nature of OAA prevents this cycling during catalysis.

In summary, we establish a method for interrogating TET-dependent mechanisms of catalysis, revealing that VC and OAA directly bind to the TET2 active site. Even though VC and OAA both compete with αKG for binding to the TET2 active site, as expected based on prior work^{42,45, 46,}, VC promotes both DNA oxidation and the uncoupled reaction (the conversion of αKG to succinate in the absence of 5mC-containing DNA). Our data thus suggest that VC directly reduces the enzyme-bound Fe³⁺ to promote TET function, and this mechanism would ultimately be more favorable than indiscriminately reducing free iron as cellular unbound iron is scarce⁴⁷. Similar to other TCA intermediates (e.g., succinate and fumarate) and related αKG-dependent enzymes^42–44, we also found TET2 is inhibited by OAA. To date, both the enzyme that generates OAA, pyruvate carboxylase, and OAA itself have been implicated in cancer progression^{40, 41}. While their roles have been traditionally ascribed to energy utilization of cancer cells, our findings provide an additional potential role for OAA in cancer by inhibiting TET2 function, and therefore promoting tumor evasion of the immune system and downregulating tumor suppressors.

The ability to observe the reactants, cofactors, and products of TET-dependent catalysis is an important step in understanding its regulation. As TET2 protein is present in solid tumors at wild-type levels, its function is regulated at the protein level by post-translational modifications, metabolites, and protein-protein interactions, e.g., TET2 interacts with transcription factors to promote sequence specific demethylation^{11, 12, 48–54}. Current methods use relative short snapshots using antibody-based methods or detailed expertise, e.g., bisulfite sequencing or LC-MS/MS, to identify TET reaction intermediates^{32–34, 55}. While our method has limitations, such as high enzyme and DNA substrate concentrations, it permits the dissection of the TET2 mechanism in the absence of DNA entirely by monitoring the uncoupled reaction. Therefore, we expect this method can also be used to interrogate how these different modes of regulation impact the multifaceted, iterative steps of DNA oxidation, determine the dynamics of DNA that improve or disrupt catalysis, and improve the assessment of TET and other enzymes involved in epigenetic modifications.