Catalytic Space Engineering as a Strategy to Activate C-H Oxidation on 5-Methylcytosine in Mammalian Genome

Abstract

Conditional remodeling of enzyme catalysis is a formidable challenge in protein engineering. Herein, we have undertaken a unique active site engineering tactic to command catalytic outcome. With ten-eleven translocation (TET) enzyme as a paradigm, we show that variants with an expanded active site significantly enhance multi-step C-H oxidation in 5-methylcytosine (5mC) whereas a crowded cavity leads to a single-step catalytic apparatus. We further identify an evolutionarily conserved residue in the TET family with a remarkable catalysis-directing ability. The activating variant demonstrated its prowess to oxidize 5mC in chromosomal DNA for potentiating expression of genes including tumor suppressors silenced in human neoplasia.

Graphical Abstract

Enzymes that modify post-replicative genetic material are crucial for safeguarding and accessing the information encoded therein. DNA methyltransferase, which methylates carbon 5 on cytosine to form 5-methylcytosine (5mC) to drive mammalian gene expression, is a prime example.¹ 5mC has been shown to undergo iterative C-H oxidation by the Ten-eleven translocation (TET) enzymes to 5-hydroxymethylcytosine (5hmC), 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC) in presence of 2-ketoglutarate (2KG). Subsequently, 5fC and 5caC are both excised by thymine DNA glycosylase (TDG) for repair, constituting an intriguing mechanism for DNA demethylation (Figure 1A).^2–6 Due to their distinct stereo-electronic features, the 5mC oxidative intermediates (5mC_ox) interact with a range of chromatin-associated proteins and regulate gene expression.^7–9

Figure 1.

Catalytic cycle of TET-mediated 5mC oxidation. (A) Cytidine demethylation by TET and TDG. (B) In a catalytic cycle, proton abstraction by Fe^IV(O) constitutes rate-determining transition state. (C) Engineering active site space of TET enzymes to control 5mC_ox distribution in DNA.

Much is unknown about the context-dependent functions of the intermediates as demethylation mediators and transcriptional regulators. Assigning a distinct and compensatory role to 5mC_ox species, particularly 5fC and 5caC, in mammalian physiology has remained a challenge due to their intrinsic low abundance and fleeting existence. Limited efforts to modulate 5mC_ox distribution in DNA include ascorbic acid mediated TET activation.¹⁰ Ascorbic acid, however, activates several 2KG-dependent dioxygenases, obscuring its exclusive role in controlling TET activity. Other approaches focus on conditional reconstitution of wildtype TET2 that are less likely to regulate the 5mC_ox distribution.^11–12

We envisioned an active site remodeling tactic to stimulate C-H oxidation for enriching the human genome with the low abundant intermediates for functional analysis. In TET-mediated C-H oxidation, hydrogen abstraction by oxo-ferryl species is the rate-limiting step with relative activation barrier following the order 5fC > 5hmC > 5mC (Figure 1B).¹³ Computational analyses indicate that restrained conformations of 5hmC and 5fC within the active site likely prevent the abstractable hydrogens from adopting a favorable orientation for C-H oxidation; in contrast, all the rotational conformers of 5mC lead to productive hydrogen abstraction.^14–16 We reasoned that systematic alteration of the stereo-electronic component at the TET-2KG-5mC interface could modulate the transition state barrier towards a more controlled product distribution (Figure 1C).

To test our hypothesis, we generated a panel of TET2 mutants by replacing several hydrophobic and polar residues in the active site with alanine, one at a time (Figure 2A).¹⁷ Employing a MALDI-MS based assay, we examined the oxygenase activity of these mutants towards a short dsDNA substrate CAC5mCGGTG.^18–19 To analyze the product distribution in a quantitative manner, we first examined if the DNA segments carrying various 5mC_ox intermediates displayed comparable ionization potentials. For this, we synthesized three additional DNAs (CACXGGTG; X = 5hmC, 5fC and 5caC) (Figure S1, Scheme S1–3). 5mC DNA was mixed with each synthesized 5mC_ox DNA separately in different ratios and subjected to MALDI-MS. Peak intensity of the individual DNA reflected their known ratio in the mixture, indicating equivalent degrees of ionization in MALDI analyses (Figure 2B, S2–4).

Figure 2.

Active site engineering towards novel TET2 variants. (A) Structure of TET2 bound with 5mC-DNA (PDB: 4NM6).¹⁷ (B) % Intensities of MALDI-MS signals for a range of mixtures of 5mC DNA with either of 5hmC, 5fC, and 5caC. (C) Heat-map representation of % 5mC_ox furnished by TET2 mutants. (D) MALDI-MS spectrum showing activating effect of V1395A to generate 5caC.

In an in vitro assay, wildtype TET2 converted 5mC DNA to 5hmC, 5fC and 5caC. In contrast, the mutants oxidized substrate DNA with varied degrees of specificity (Figure 2C, S5). T1372A, T1393A and T1883A mutants showed remarkable specificity for generating primarily 5hmC. It has been shown that a H-bonding network involving T1372 is critical for rotating 5hmC to switch from product to substrate orientation for further oxidation; lack of such interaction in T1372A locks 5hmC in the product conformation.²⁰ We suspect a similar mechanism is operating for T1393A and T1883A mutants as well to impart the degree specificity in C-H oxidation. Among the hydrophobic residues screened, the most notable difference was observed for the V1395A mutant as it generated 5caC as the dominant product (>90%) (Figure 2C, D). To the best of our knowledge, such an activating TET mutant is yet to be reported. Collectively, our systematic screening experiment identified TET2 variants that could significantly modulate 5mC_ox distribution by either stalling the oxidation pathway at the first step to exclusively generate 5hmC or accelerating to the fully oxidized product. The V1395A mutant is particularly important given that 5caC is rare in mammalian genome.

We further analyzed 5caC formation using base-resolution sequencing. It has been shown that 5caC is deaminated upon treatment with bisulfite and read as T during sequencing (Figure 3A).²¹ In contrast, 5mC and 5hmC remain as C due to their inertness towards the reagent. We subjected a duplex 76-mer DNA carrying a central 5mCpG unit to TET-mediated oxidation followed by bisulfite treatment, PCR amplification and Sanger sequencing. While 5mC in samples exposed to either no protein or wildtype TET2 read as C, the equivalent site in the V1395A-treated sample emerged as T (Figure 3B, S6). These results demonstrate the ability of the mutant to predominantly generate 5caC.

Figure 3.

Analysis of 5mC_ox using Sanger sequencing. (A) Scheme showing bisulfite-mediated 5caC deamination and decarboxylation. (B) V1395A-catalyzed 5mC oxidation followed by bisulfite treatment and PCR confirming 5caC as major product based on Sanger sequencing data. BS=bisulfite

We next compared the activity of V1395A with that of wildtype TET2 on three individual substrates (Figure 4A–C, S7–12). With respect to the wildtype enzyme, V1395A led to a significantly higher production of 5hmC from 5mC, while both enzymes showed comparable activity on 5hmC DNA. The mutant displayed further improved activity on 5fC leading to higher 5caC formation, demonstrating that it is indeed acting as a superior dioxygenase in two discrete cycles involving conversions of 5mC to 5hmC and 5fC to 5caC in a three-step catalytic process. To further gauge the relative activity, we measured catalytic efficiencies of V1395A and wildtype TET2 on 5mC with varying concentrations of 2KG (Figure S13). While the mutant had improved turnover (k_cat), both enzymes showed comparable K_M values, ruling out the possibility that enhanced activity of V1395A could be from improved cofactor binding.

Figure 4.

Characterization of the TET variants. (A, B, C) Time-dependent activity of wildtype TET2 and V1395A towards 5mC (A), 5hmC (B) and 5fC (C) substrate DNAs. (D) Heat-map representation of % 5mC_ox generated by V1395A congeners. (E) Sequence alignments of eukaryotic TETs highlighting the gatekeeper residue. (F) Activity of wildtype TET1 and V1685A mutant on 5mC, 5hmC and 5fC substrate DNAs.

To further examine the influence of active site space on product distribution, we generated a panel of V1395 mutants differing in amino acid sidechains (Figure 4D, S14). As expected, the glycine variant generated 5caC as the major product. Remarkably, the mutants with a crowded active site furnished primarily 5hmC without affecting oxidation efficiency, demonstrating a clear relationship between catalytic pocket size and 5mC_ox distribution. This switch from three-step catalysis (5mC→5hmC→5fC→5caC) by V1395A to just one (5mC→5hmC) by its congeners thus must be a result of crowding the active site.

Intrigued by the unique catalysis-directing feature of V1395 in TET2, we analyzed the sequences of representative TET homologues and found that a bulky hydrophobic residue at equivalent sites is strictly conserved among eukaryotes, indicating its gatekeeper-like role in controlling 5mC_ox derivatives in the genome (Figure 4E).^22–26 To access a similar activating variant of human TET1, we generated V1685A mutation which indeed showed remarkably higher enzymatic activity compared to wildtype TET1 (Figure 4F, S15–17). Our work thus identifies a conserved gatekeeper residue in eukaryotic TETs that can be engineered to regulate 5mC_ox content in DNA. Such mutants are valuable for probing member-specific activity of TET1-3 with identical catalytic pockets but distinct expression patterns and domain organizations.²⁷

We next examined the ability of V1395A mutant to oxidize 5mC on genomic DNA isolated from HEK293T cells. The product distribution was analyzed in dot-blot assay using 5mC_ox-specific antibodies. The mutant furnished 5fC and 5caC more than 3-fold in excess compared to wild type, a result consistent with the variant having a higher enzymatic activity (Figure 5A, S18). It is notable that the variant with an expanded active site is capable of generating an excess of 5caC on the entire human DNA carrying millions of 5mC moieties. Remarkably, V1685A also enhanced 5mC_ox content on genomic DNA in comparison to wildtype TET1, consistent with the activating nature of the expanded catalytic site (Figure 19).

Figure 5.

Activity of TET2 and V1395A on genomic DNA. (A) Dot-blot assay showing time-dependent increase in 5caC by V1395A. Bar diagram provides relative fold changes in 5caC level. (B) Chemical strategy to selectively derivatize 5caC on genomic DNA. (C) Dot-blot assay with 5aC and biotin antibodies before and after derivatization, respectively. Bar representation of biotinylated 5caC on genomic DNA. Non-linearity in biotin signal is likely due to multi-step functionalization of 5caC.

This result prompted us to apply the TET2 variant in combination with chemo-functionalization strategy to enrich 5caC-containing DNA (Figure 5B).^{21, 28} To selectively derivatize 5caC, 5fC group in TET-treated DNA was reduced using NaBH₄ followed by coupling with biotinylated hydroxylamine on an activated carboxylate substrate. Subsequently, the samples were analyzed in dot-blot assay using 5caC- and biotin-specific antibodies to gauge pre- and post-functionalized 5caC (Figure 5C). Consistent with the mutant’s enhanced prowess to generate 5caC, we observed a 5-fold stronger signal of biotinylated 5caC for the mutant over wildtype TET2, implying that the activating variant can be exploited for improved isolation of the low abundant intermediates.

To examine if the TET2 variant could activate 5mC oxidation on the DNA within human cells, we expressed the catalytic domain of wildtype TET2 or the mutant in HEK293T cells (Figure 6A). Dot-blot analysis of the isolated DNA revealed the presence of ~2-fold higher 5hmC and 5caC DNA in V1395A expressing cells compared to those expressing the wildtype enzyme (Figure 6B, S20), demonstrating that the activating effect of V1395A is well preserved in the complex environment of mammalian cells.

Figure 6.

In-cellulo activity of TET2 variant. (A) Schematic displaying activity of V1395A on chromosomal DNA. (B) Dot-blots confirming higher level of 5hmC and 5caC generated by the mutant. (C) Steps involved in DIP-qPCR. (D) Fold change (mutant/wild-type) in enrichment of selected genes using 5caC antibody as revealed by DIP-qPCR. (E) TET2 interacts with WT1 to regulate DACT1. (F) Expression of DACT1 is enhanced by V1395A.

We then assessed the ability of V1395A to oxidize 5mC at individual promoters. We selected a panel of eight genes including microRNA-200 and EPCAM as TET-mediated oxidation of promoter 5mC leads to reactivation of the corresponding transcripts.²⁹ We performed DNA immunoprecipitation (DIP) from HEK293T cells expressing wildtype or the variant enzymes with 5mC_ox-specific antibodies followed by quantitative polymerase chain reaction (qPCR) using gene-specific primers (Figure 6C, S20).³⁰ Results from three independent replicates suggested that the mutant indeed boosted 5mC oxidation at these loci compared to wild type TET2 as revealed by >2-fold enrichment of the genes in DIP-qPCR assay (Figure 6D).

Finally, we examined if V1395A is functionally relevant to potentiate the expression of tumor suppressors. This is particularly important given that inactivating TET2 mutations are known to downregulate tumor suppressors in hematological malignancies.³¹ We focused on DACT1, a tumor suppressor that is frequently silenced in cancers via promoter methylation.³² It has been shown that the transcription factor WT1 interacts with TET2 to regulate expression of DACT1 in normal hematopoietic stem cells via 5mC oxidation (Figure 6E).³³ In western blot, we noted that wildtype TET2 marginally increased DACT1 level in cells compared to a control vector. Remarkably, the mutant catalyzed DACT1 expression up to 3-fold increase over TET2, consistent with its higher dioxygenase activity towards 5mC (Figure 6F, S21). Such results imply that the activating mutant is capable of enhancing the expression of a tumor suppressor that is aberrantly silenced in cancers.

In this work, we explored active site space engineering as a novel tactic to direct catalytic outcome. Mutations at an evolutionarily conserved residue led to TET variants with tunable oxygenase activity: an enlarged cavity augmented multi-step oxidation of 5mC to 5caC as the dominant product, whereas reduced catalytic space led to primarily 5hmC via a single-step oxidation. Remarkably, the spacious variant exhibited superior activity to potentiate expression of anti-metastatic genes in human cells. We anticipate the engineered TET in combination with genome editing strategies would constitute a powerful tool for establishing locus-specific 5mC_ox pattern to reprogram downstream processes. V1395A mutant will find further application in the enrichment and analysis of 5caC content in the genome using TET-assisted bisulfite (TAB)³⁴ and related sequencing apporaches. Furthermore, our work provides a blueprint for remodeling the catalytic apparatus of 2KG-dependent dioxygenases, which are emerging as critical regulators of mammalian gene expression.