DNA labeling using DNA methyltransferases

Abstract

DNA methyltransferases (MTases) uniquely combine the ability to recognize and covalently modify specific target sequences in DNA using the ubiquitous cofactor S-Adenosyl-L-methionine (AdoMet). Although DNA methylation plays important roles in biological signaling, the transferred methyl group is a poor reporter and is highly inert to further biocompatible derivatization. To unlock the biotechnological power of these enzymes, two major types of cofactor AdoMet analogs were developed that permit targeted MTase-directed attachment of larger moieties containing functional or reporter groups onto DNA. One such approach (named Sequence-specific Methyltransferase-Induced Labeling, SMILing) uses reactive aziridine or N-mustard mimics of the cofactor AdoMet, which render targeted coupling of a whole cofactor molecule to the target DNA. The second approach (methyltransferase-directed Transfer of Activated Groups, mTAG) uses AdoMet analogs with a sulfonium-bound extended side chain replacing the methyl group, which permits MTase-directed covalent transfer of the activated side chain alone. As the enlarged cofactors are not always compatible with the active sites of native MTases, steric engineering of the active site has been employed to optimize their alkyltransferase activity. In addition to the described cofactor analogs, recently discovered atypical reactions of DNA cytosine-5 MTases involving non-cofactor-like compounds can also be exploited for targeted derivatization and labeling of DNA. Altogether, these approaches offer new powerful tools for sequence-specific covalent DNA labeling, which not only pave the way to developing a variety of useful techniques in DNA research, diagnostics and nanotechnologies, but has already proven practical utility for optical DNA mapping and epigenome studies.

Introduction

DNA is a large linear polymer comprised of aperiodic combinations of four major types of building blocks encoding the genetic blueprint of life. Since different loci of this largely uniform biomolecule rarely contain features distinct enough to permit their chemical or physical identification among other DNA loci or other biomolecules, a key task is to furnish them with suitable reporter tags for their selective visualization and isolation from biological samples. Among the variety of enzymes involved in DNA metabolism, DNA methyltransferases (MTases) uniquely combine two useful features required for targeted labeling: recognition of a vast repertoire of specific target sequences (2-8 nt long) and covalent modification of the target site. Although targeted DNA methylation can be “read” by specific cellular proteins and thus plays important roles in biological signaling, the naturally transferred methyl group is a poor reporter and is not readily amenable for further chemical derivatization. Therefore, one strategy to unlock the biotechnological potential of these highly specific enzymes is to make them transfer “pre-derivatized” (extended) versions of the methyl group. The catalytic power of AdoMet-dependent MTases to a large extent derives from their ability to bring the two substrates, the cofactor AdoMet and a target molecule, together in the right orientation. Thus, a series of synthetic analogs of the AdoMet cofactor were developed that allowed MTases to tag DNA with extended moieties, making sequence specific MTase-directed labeling an attractive opportunity in various biotechnological applications. Two major types of cofactor analogs have been developed for MTase-catalysed DNA labelling which permit covalent deposition of either a whole cofactor molecule or its sulfonium-bound side chain. Among the three known classes of DNA methyltransferases (cytosine-5, adenine-N6 and cytosine-N4 MTases) the first two have been largely utilised for the attachment of various reactive groups, biotin or fluorophores to DNA. Due to the universal nature of the AdoMet cofactor for biological methylations, the approach also proved applicable for labeling other biomolecules, such as RNA (Motorin et al. 2011; Tomkuvienė et al. 2012; Plotnikova et al. 2014; Schulz et al. 2013; Holstein et al. 2014), proteins (Peters et al. 2010; Islam et al. 2011; Willnow et al. 2012; Wang et al. 2013; Hymbaugh Bergman and Comstock 2015) and small molecules (Zhang et al. 2006; Stecher et al. 2009; Lee et al. 2010; Winter et al. 2013) using appropriate MTases.

Another recently developed cofactor-independent DNA modification strategy is based on atypical reactions of DNA cytosine-5 MTases. Upon interaction with the target cytosine, these MTases use a covalent attack to transiently generate an activated cytosine intermediate (ACI) (see Chapter 3). In the absence of AdoMet or synthetic AdoMet analogs, the ACI can undergo a covalent addition of exogenous formaldehyde yielding 5-hydroxymethylcytosine (hmC). Moreover, hmC residues at the target site can be dehydroxymethylated to yield cytosine or can undergo further addition of thiols or selenols to yield the corresponding 5-chalcogenomethyl derivatives in DNA in a C5-MTase-dependent manner. These transformations open new possibilities for sequence-specific derivatization and analysis of epigenetic marks in mammalian DNA.

In the following sections, the DNA labeling approaches based on the two types of synthetic cofactor analogs and the reactions involving non-cofactor-like compounds are discussed in detail.

1. Synthetic cofactor analogs for MTase-directed modification of DNA

The first labeling strategy (named Sequence-specific Methyltransferase-Induced Labeling or SMILing) developed by the Weinhold group employed cofactor analogs in which the methionine moiety of AdoMet was synthetically replaced by an aziridine ring (N-adenosylaziridine cofactors) (Pignot et al. 1998). Upon reaction of N-adenosylaziridine with DNA in the presence of a DNA MTase, the “transfer” of an electrophilic carbon atom of the protonated aziridine ring to a nucleophilic target atom in DNA leads to ring opening, thereby turning the ring into a ethylamino linker that connects the cofactor molecule with the target nucleobase (Pignot et al. 1998) (Figure 1). Although the attached ethylaminoadenosine moiety by itself is not a good reporter group, it can serve as a carrier to which desired chemical and reporter groups are attached. To minimize interference with proper cofactor binding in the catalytic centre of a directing MTase, the selection of potential anchoring points in the adenosine moiety appears to be limited to the 6, 7 and 8 positions of the adenine ring (Pljevaljčić et al. 2004, Kunkel et al. 2015). The SMILing approach was initially developed with an aziridine AdoMet analog possessing a dansyl fluorophore attached to the C8 position of the adenine ring. This analog was shown to function as a cofactor for the adenine-N6-specific DNA methyltransferase M.TaqI from Thermus aquatiqus, resulting in the cofactor covalently attached to the exocyclic amino group (N6) of the target adenine located within sequence 5’-TCGA-3’ (Pljevaljčić et al. 2003). Subsequently, the groups of Rajsky and Comstock expanded the chemical scope of this approach by introducing 2-haloethyl N-mustard analogs, which are converted into aziridines in situ and thus are presumed to work by a similar mechanism (Weller and Rajski 2006, 2005; Townsend et al. 2009; Mai and Comstock 2011; Du et al. 2012; Ramadan et al. 2014). In the N-mustard cofactors, the N atom, which is equivalent to the sulfur atom of the sulfonium group in AdoMet, can in addition be used to attach a reactive chemical group (alkyne) (Weller and Rajski 2005) or a photocaging group (Townsend et al. 2009). Alternatively, retention of the amino acid moiety (present in AdoMet but absent in the N-adenosylaziridine analogs) renders enhanced cofactor-MTase affinity which gives a certain benefit of lower concentrations of cofactors that can be used in the labeling reactions (Weller and Rajski 2006; Du et al. 2012; Ramadan et al. 2014). Both the aziridine and N-mustard cofactors are obtained via multistep synthetic routes, and can thus only be produced in specialized chemistry laboratories. Altogether, a variety of cofactor analogs have been produced containing reporter (biotin, fluorophores) or functional reactive groups (azide, alkyne) attached to the N6 or C8, or designed C7, positions of the adenine ring (see Table I).

Figure 1. Methyltransferase-directed sequence-specific labeling of DNA using synthetic analogs of the cofactor AdoMet.

(Left) SMILing approach: covalent coupling of an aziridine (upper) or N-mustard (lower) cofactor carrying a functional or reporter group (red sphere) attached via a linker (red line) onto a target nucleobase (blue) in DNA. (Right) mTAG approach: transfer of a sulfonium-bound extended linear chain carrying an activating triple (upper) or double (lower) bond, a linker and a functional or reporter group from a double-activated AdoMet analog onto a target nucleobase in DNA. N, random nucleotide; XXXXX, recognition sequence of the directing MTase.

Table 1.

Cofactor analogs for MTase-directed derivatization and labeling of DNA

Reactive or transferable moiety	Position of linker	Linker length	Functional or reporter group	Name	MTases used	Applications	References
SMILing
N-adenosylaziridines	Ade-N6	5	Biotin	6BAz	M.BseCI, M.TaqI	Positioning of nanoparticles on bacteriophage DNA Engineering synaptic junctions in DNA Optical mapping of DNA-binding proteins	Braun et al. 2008, Wilkinson et al. 2008, Kim et al. 2012
		11	Cy3	6Cy3Az	M.TaqI	Transfection with fluorescently labeled plasmid DNA	Schmidt et al. 2008
	Ade-C7	4	Biotin		M.HhaI, M.TaqI	Biotinylation and CpG methylation detection on plasmid DNA	Kunkel et al. 2015
	Ade-C8	-	Azide		M.EcoRI, M.HhaI, M.SssI, M.TaqI	Derivatization and biotinylation of ODNsa MTase-directed DNA strand scission	Comstock and Rajski 2005a,b
		5	Azide		M.EcoRI, M.TaqI	ODN biotinylation	Comstock and Rajski 2005a
		5	Biotin		M.TaqI	Biotinylation of plasmid DNA	Pljevaljčić et al. 2004, 2007
		6	Dansyl		M.TaqI	Fluorescent labeling of plasmid DNA	Pljevaljčić et al. 2003, 2004
N-adenosyl-N-mustards	Mustard-N	1	Alkyne		M.EcoRI, M.TaqI	ODN derivatization	Weller and Rajski 2005
	Ade-N6	1; 2; 4 3; 4	Alkyne Azide		M.HhaI, M.TaqI	Derivatization of plasmid DNA	Ramadan et al. 2014
	Ade-C8	2 5	Alkyne Azide		M.HhaI, M.TaqI	Fluorescent labeling of plasmid DNA	Du et al. 2012
		-	Azide		M.TaqI	Derivatization of plasmid DNA	Mai and Comstock 2011
mTAG
S-Propargyl analogs	S	1	Alkyne	2-butynyl-SAMb	M.TaqI	DNA labeling and extraction	Artyukhin and Woo 2012
		6	Alkyne	Ado-6-ethyne	eM.HhaI	Fluorescent labeling of plasmid DNA	Lukinavičius et al. 2013
		6	Amine	Ado-6-amine	eM.HhaI, eM.SssI	Biotin labeling of DNA for epigenome profiling	Lukinavičius et al. 2013, Kriukienė et al. 2013
		9	Amine	Ado-9-amine	eM.HhaI, M.TaqI	Fluorescent labeling of plasmid DNA	Lukinavičius et al. 2007
		11	Amine	Ado-11-amine	eM2.Eco31I, eM.HhaI, eM.HpaII	Fluorescent labeling of plasmid and phage DNA Optical DNA mapping	Neely et al. 2010, Lukinavičius et al. 2012, Lukinavičius et al. 2013
		6	Azide	Ado-6-azide	eM.HhaI, eM.SssI	Fluorescent labeling of plasmid DNA ex vivo Biotin labeling of DNA for epigenome profiling	Lukinavičius et al. 2013, Kriukienė et al. 2013
		18	Biotin	Ado-18-biotin	eM.HhaI	Labeling of plasmid DNA	Urbanavičiūtė et al. UOc, Plotnikova et al. 2014
		20	TAMRA	AdoYnTAMRA	M.TaqI	Fluorescent labeling of phage DNA for optical strain typing	Grunwald et al. 2015
S-Allyl analogs	S	3	Alkyne	AdoEnYn	M.FokI, M.TaqI, M.XbaI	Fluorescent labeling of phage DNA for optical DNA mapping	Vranken et al. 2014

^aODN, oligodeoxyribonucleotide;

^bthis cofactor analogue is also called AdoButyn in Tomkuvienė et al. 2012;

^cUO, unpublisched observations.

An important feature of the SMILing reaction is that the directing MTase remains tightly (although non-covalently) bound to the coupling product, which represents a chemically linked bisubstrate derivative entangling the enzyme. Therefore, stoichiometric amounts of an enzyme with respect to its target sites on target DNA are required for quantitative conversion, and additional steps may be necessary if the bound enzyme is to be removed from the DNA. The attached cofactor represents a relatively bulky group, which may be a useful or inferior feature depending on downstream applications.

The second DNA labeling approach is based on AdoMet analogs in which the sulfonium-bound methyl group of AdoMet is replaced with an extended side chain, and only this part of the cofactor is transferred to the target nucleotide (Figure 1). In AdoMet, the transferable methyl group is activated by the adjacent sulfonium center, and AdoHcy serves as the leaving group during the MTase-catalyzed S_N2 reaction. Replacement of the methyl group in AdoMet with larger aliphatic carbon chains had previously been attempted by Schlenk and Dainko (1975), who found that even short groups such as ethyl or propyl led to a drastic decline of transfer rates by MTases. A strongly decreased reaction rate observed with the saturated alkyl groups predominantly results from unfavorable steric effects within the transition state. In a joint effort, the Klimašauskas and Weinhold groups found that the efficiency of the reaction can be enhanced by placing π-orbitals near the reaction center (Dalhoff et al. 2006a). This activation was observed with synthetic AdoMet analogs carrying a double bond (allylic system) or a triple bond (propargylic system) next to the reactive carbon in the extended side chain (Figure 1). Mechanistic considerations suggest that the π-orbitals in the unsaturated bond lower the energy barrier of the reaction via conjugative stabilization of a pentacoordinated S_N2 transition state. The discovery of the double-activated AdoMet analogs paved the way to a rapid development of a new approach termed methyltransferase-directed Transfer of Activated Groups (mTAG).

Synthetic access to the mTAG cofactors appears somewhat easier as compared to the aziridine and N-mustard analogs, since they can be produced in a single step by chemical “recharging” the cofactor product AdoHcy via regiospecific alkylation of its sulfur atom with a desired linear side chain. Suitable electrophilic side chains can sometimes be obtained directly from commercial sources, but certain cases may require advanced synthetic skill (Lukinavičius et al. 2007; Lukinavičius et al. 2013; Dalhoff et al. 2006b; Masevičius et al. 2016). Chemical synthesis typically yields the cofactor analogs as diasteromeric mixtures of R,S- and S,S-isomers, which can be chromatographically enriched in the enzymatically active S,S-isomer by reversed phase chromatography (Lukinavičius et al. 2013). Recently, a chemo-enzymatic synthesis of enantiomerically pure mTAG cofactors from corresponding methionine analogs and ATP using engineered methionine adenosyltransferases has been demonstrated (Singh et al. 2014), which can in principle be performed even in living cells (Wang et al. 2013).

Since only the extended sulfonium-bound side chain is transferred from the cofactor analog to DNA, these AdoMet analogs circumvent the problem of catalytic product release, which is unavoidable for the SMILing reactions. A variety of both allyl-based and propargyl-based analogs have been designed that carry unique chemical groups such as primary amine, alkyne and azide, or reporter groups (biotin, fluorophores) (see Table I). Notably, although many MTases accept well both types of mTAG cofactors, some exhibit certain preferences with respect to the activating unsaturated bond (double or triple) or the side chain length. In particular, allylic cofactors have gained significant popularity with protein labeling (Peters et al. 2010; Islam et al. 2011; Wang et al. 2011; Islam et al. 2012; Islam et al. 2013; Blum et al. 2013; Wang et al. 2013; Bothwell and Luo 2014; Guo et al. 2014), whereas propargylic side chains are more preferably transferred by the C5-DNA MTases (Table II). Unexpectedly, some of the propargyl cofactors containing an electronegative group (amino or amido) at position 4 of the side chain were found to undergo a rapid loss of activity under physiological conditions. Further studies indicated that a close proximity of electron withdrawing groups makes the triple C-C bond highly susceptible to base-promoted addition of a water molecule. This problem was resolved by synthesis of a series of hex-2-ynyl cofactor analogs in which the separation between the electronegative group and the triple bond is increased from one to three carbon units (Lukinavičius et al. 2013). A similar mechanism has also been proposed for the fast inactivation of the AdoMet analog carrying a short unsubstituted prop-2-yn-1-yl side chain. In this case, the undesirable chemical reactivity of the triple bond has been diminished by synthetically replacing the sulfur atom in the highly electronegative sulfonium center with selenium (Bothwell et al. 2012; Willnow et al. 2012).

Table 2.

Activity of DNA methyltransferases with AdoMet analogs

Enzyme	Target sequence 5‘ – 3‘		SMILing		mTAG
		Reactions performed*	References	Reactions performed*	References
m6A-MTases
M.TaqI wt	TCGA	M F L	Pignot et al. 1998, Weller and Rajski 2006 Weller and Rajski 2005, Comstock and Rajski 2005a,b, Du et al. 2012, Mai and Comstock 2011, Ramadan et al. 2014 Pljevaljčić et al. 2003, 2004, 2007, Braun et al. 2008, Schmidt et al. 2008, Wilkinson et al. 2008, Kunkel et al. 2015	M F L	Dalhoff et al. 2006a Lukinavičius et al. 2007, Vranken et al. 2014, Artyukhin and Woo 2012, Grunwald et al. 2015
M.BseCI	ATCGAT	L	Braun et al. 2008, Wilkinson et al. 2008, Kim et al. 2012
M.EcoRI	GAATTC	M	Weller and Rajski 2006	N	Vranken et al. 2014
		F	Comstock and Rajski 2005a, Weller and Rajski 2005
M.FokI	GGATG/CATCC			F	Vranken et al. 2014
M.XbaI	TCTAGA			F	Vranken et al. 2014
M.EcoDam	GATC			N	Vranken et al. 2014
M.PstI	CTGCAG			N	Vranken et al. 2014
m5C-MTases
M.HhaI wt	GCGC	M F	Weller and Rajski 2006, Pljevaljčić et al. 2004, Comstock and Rajski 2005a,b, Ramadan et al. 2014, Du et al. 2012	M	Dalhoff et al. 2006a
		L	Kunkel et al. 2015
Q82A/N304A				F	Lukinavičius et al. 2007, 2012, 2013
Y254S/N304A				F	Lukinavičius et al. 2012
Q82A/Y254S/N304A				F	Neely et al. 2010, Lukinavičius et al. 2012, 2013
				L	Urbanavičiūtė et al., unpublished
M.HhaI ΔL2–14	GCG			M	Gerasimaitė 2009
				F	Gerasimaitė et al., unpublished
Q82A/N304A				L	Urbanavičiūtė et al., unublished
M.SssI wt	CG	M	Weller and Rajski 2006,	N	Vranken et al. 2014
		F	Comstock and Rajski 2005b
Q142A/N370A				F	Kriukienė et al. 2013
M.HpaII wt	CCGG	M	Comstock and Rajski 2005a	M	Lukinavičius et al. 2012
Q1042A/N335A				F	Lukinavičius et al. 2012
M2.Eco31I wt	GGTCTC			F	Lukinavičius et al. 2012
N127A/Q233A				F	Lukinavičius et al. 2012
M.BsaHI wt	GRCGYC			N	Vranken et al. 2014
Q75A/N274A/V220S				N	Vranken et al. 2014
4mC-MTases
M.BcnIB	CCSGG	M	Pljevaljčić et al. 2004	M	Dalhoff et al. 2006a
M.BamHI	GGATCC	M	Du et al. 2012
M.PvuII	CAGCTG			N	Vranken et al. 2014

^*N - none or low alkyltransferase activity, M- modification by transfer of a core unit (SMILing) or a short non-functional moiety (mTAG), F- derivatization with a functional group (2-step labeling possible), L- labeling with a reporter group in one step

Both the SMILing and mTAG cofactors can be used for two-step or one-step labeling. A key advantage of the two-step labeling approach is the flexibility in manipulating the chemical parameters of the labeling reaction (linker length, conjugation chemistry, reporter group) by simply combining different cofactors and chemoselective reporter compounds. Alternatively, single-step labeling by direct attachment of a desired reporter group may be beneficial in situations when minimal sample manipulations, simplicity, and speed are required. However, beside this potential advantage, the one-step approach entails an added synthetic complexity to the cofactor analog, as reporter groups are typically larger and more complex than functional groups. Moreover, an increased steric bulk of the transferable side chain may also lead to a partial or complete impairment of the directing MTase (Table II).

2. MTase activity with the synthetic cofactor analogs

Bacterial and archaeal DNA MTases generally exhibit a clearly defined sequence and base specificity. Bacterial type-II DNA MTases (typically, single polypeptides of 250–400 residues) seem to be better suited for DNA labeling purposes as compared to the type I and III enzymes or mammalian DNA MTases, mostly due to their compact size and better enzymatic parameters (turnover rate, cofactor affinity, sequence fidelity, protein stability etc.), although this general assumption does not preclude the existence of useful MTases derived from other than type-II cohorts. Current listings of type-II DNA MTases (REBASE, http://rebase.neb.com) count over 350 distinct recognition sequences ranging from two to eight base pairs in length. Therefore, a wide repertoire of DNA sequences can potentially be targeted, which is in par with that of the widely used restriction endonucleases.

Naturally, DNA MTases have evolved for optimal performance with the natural cofactor AdoMet. The use of extended AdoMet analogs raises the question of steric limitations that may be imposed by the architectures of the active sites and cofactor binding pockets of MTases. As mentioned above, the SMILing cofactors offer several potential anchoring points in the adenosine moiety (6, 7 and 8 positions of the adenine ring for the aziridines and additionally 5’-N for the N-mustards) that can be used for building a suitable extension carrying a desired functionality. This thus offers several chemical options for designing suitable cofactor for particular MTases. In the mTAG cofactors, there is only one attachment point, and the chemical variability of the side chain is basically limited to either the allyl or propargyl moieties, which demand quite distinct geometries of the cofactor pocket in both the ground and transition states. Of course, the length of the side chains and other chemical features can also influence the reaction to some extent, but such effects decline with increasing distance from the active site.

Representatives of all three classes of bacterial DNA MTases (m6A, m4C and m5C forming enzymes) showed activity with certain types of extended cofactor analogs (see Table II). On one end of the spectrum is M.TaqI, which demonstrated high tolerance with respect to a wide range of SMILing and mTAG cofactors examined. More typically though, the efficiency of mTAG transalkylations with wild-type enzymes is insufficiently high for routine applications. For C5-MTases, this issue was approached by engineering of the cofactor pocket of a well characterized representative of the class, M.HhaI (Lukinavičius et al. 2012). The engineering effort was guided by a structure-based model of a M.HhaI-DNA-butynyl cofactor complex (Figure 2), which suggested that the side chains of residues Gln82 and Asn304 (located in conserved sequence motives IV and X), and Tyr254 (located in the so-called variable region) might sterically interfere with the extended transferable side chain, precluding cofactor binding or its proper orientation for catalysis. These three positions were therefore selected for steric engineering (Ala or Ser replacements). It turned out that double and triple replacements conferred substantial improvements of the transalkylation activity and a reduction of the methyltransferase activity in M.HhaI. The achieved turnover rates permit complete derivatization of DNA in 15-30 min, which makes the reaction suitable for routine laboratory applications. Detailed studies of the mutants showed that these replacements substantially enhance the rate of alkyl transfer and also reduce the enzyme affinity toward the natural cofactor AdoMet and its product AdoHcy. The catalytic transfer of butyn-2-yl and pentyn-2-yl groups by the triple mutant was faster than its methyltransfer activity, indicating that the engineered enzyme (eM.HhaI) was turned into an alkynyltransferase. Importantly, eM.HhaI can efficiently utilize extended synthetic analogs even in the presence of AdoMet (which is naturally abundant in cells and cell lysates), opening new ways for targeted covalent deposition of reporter groups onto DNA for a variety of ex vivo and in vivo applications (Lukinavičius et al. 2013). In line with these findings, a substantial improvement of the transalkylation activity was also observed in another engineered version of M.HhaI, which was designed to target non-symmetrical GCG sites. Directed evolution of the MTase aimed at enforcing the new sequence specificity resulted in the above described Tyr254Ser mutation and additional deletions in the vicinity of the cofactor binding pocket (Gerasimaitė et al. 2009).

Figure 2. Structure-guided engineering of DNA cytosine-C5 methyltransferases for the mTAG transalkylation reactions.

(a) Model of an extended propargylic cofactor analogue (AdoButyn, shown in ball and stick) bound in the active site of the HhaI MTase (based on M.HhaI-DNA-AdoMet ternary complex X-ray structure, PDB code 6mht, shown as space-fill). An arrow points at the transferable carbon atom. (b) Sequence alignment of regions corresponding to IV and X conserved motifs of sterically-engineered prokaryotic cytosine-C5 MTases. Arrows indicate positions corresponding to Gln82 and Asn304 of M.HhaI. (c) Permutation of conserved motifs and the variable region (vr) in M.HhaI, M.HpaII, M.SssI, M.BsaHI (top) and M2.Eco31I (lower) DNA methyltransferases. Adapted from Lukinavičius et al. 2012.

The high structural conservation of C5-MTases suggested that other orthologs can be similarly engineered based on sequence alignment even in the absence of crystal structures. Indeed, the double alanine mutants involving conserved motifs IV and X led to a significant improvement of the transalkylation activity with a wide range of propargyl-based cofactor analogs by M2.Eco31I and M.HpaII, which recognize hexanucleotide and tetranucleotide target sites, respectively (Lukinavičius et al. 2012), as well as by M.SssI acting on the 5’-CG-3’ dinucleotide (Kriukienė et al. 2013). On the other hand, analogous replacements in M.BsaHI showed no significant improvement in the transfer of allyl-based extended groups onto DNA (Vranken et al. 2014). This appears to agree with the observed weaker acceptance of double-bond cofactors by the engineered M.HhaI variants (Lukinavičius, Lapinaitė, Klimašauskas, unpublished observations), suggesting that the triple-bond cofactors are generally better compatible with the C5-MTases. For more on MTase design, see Chapter X.

3. Implementation of MTase-directed labeling of DNA

Sequence-specific covalent derivatization and labeling of DNA has potentially opened new avenues in DNA research, diagnostics and bionanotechnology. However, along with methodological developments of the MTase-directed labeling reactions, the properties and practical value of such covalently modified DNA that suddenly became available needed to be assessed. Many experimental demonstrations involving various covalently tethered reporter and reactive groups have been performed at the level of oligonucleotides, PCR fragments and then plasmid DNA. These studies can be grouped into those that exploited covalent derivatizations for: 1) general covalent labeling of DNA; 2) analysis of particular DNA sites or sequences.

3.1. General covalent labeling of DNA

Soon after convincing demonstrations that both SMILing and mTAG techniques can achieve high sequence specificity of label incorporation into plasmid DNA (Pljevaljčić et al. 2007; Lukinavičius et al. 2007), the behavior of covalently labeled plasmid DNA was examined in transfected cells. For example, an aziridine-based cofactor with a Cy3 fluorophore was used for labeling of pUC19 and pBR322 plasmids with M.TaqI; the plasmids were successfully transfected and optically tracked in mammalian cells (Schmidt et al. 2008). Independently, mTAG-derivatized plasmids were shown to have transformation efficiencies similar to unmodified plasmid controls in E.coli cells (Lukinavičius et al. 2012). Moreover, sequence-specific mTAG click-labeling of endogenous plasmid DNA using eM.HhaI and Ado-6-azide cofactor followed by strain-promoted azide-alkyne cycloaddition (SPAAC) of a cyclooctyne probe was demonstrated in bacterial cell extracts (Lukinavičius et al. 2013). Altogether these experiments demonstrated a high biological tolerance (bioorthogonality) of both types of covalent modifications pointing at potential suitability of this approach for in vivo studies. An exceptional selectivity of DNA MTases towards DNA can also be used for covalent capture and extraction of DNA from complex mixtures (Artyukhin and Woo 2012). Modification of DNA with alkynyl groups using mTAG technique and further covalent immobilization through copper(I)-catalyzed azide-alkyne cycloaddition (CuAAC) reaction on azide-coated surfaces permits its further manipulations and compatibility with downstream reactions. The approach of DNA separation from other biomolecules, including RNA, showed sensitivity and selectivity unprecedented in other DNA extraction methods.

Altogether, for the purpose of general DNA labeling, the MTase-directed methods offer important advantages over random chemical labeling or other commonly used methods:

• Elimination of uncertainties related to loss of non-covalently bound labels in cells or in vitro experiments;

• Control of the labeling density and positioning of reporter groups around (or away from) functional sites by selecting appropriate MTases;

• High flexibility in selecting functional and reporter groups;

• Covalent integrity of DNA strands (preserved supercoiling of plasmid DNA);

• Biological orthogonality of the underlying modifications.

3.2. DNA labeling for analysis of particular DNA sites or sequences

Another layer of utility of the MTase-directed labeling is related to exploitation of individual labeled sites in DNA. One such area is the construction of DNA-based nanostructures. Braun et al. (2008) used biotinylated aziridine cofactor together with M.TaqI and M.BseCI (recognizing tetranucleotide and hexanucleotide sequences, respectively) for biotin labeling and subsequent targeted deposition of gold nanoparticles on model kilobase-sized DNA fragments via biotin-streptavidin interaction. Wilkinson et al. (Wilkinson et al. 2008) used similar tools to engineer synaptic three- and four-way junctions in PCR-derived DNA, which were unequivocally visualized using single-molecule AFM imaging. These examples demonstrated the capacity of MTase-directed labeling for controlled manipulation of nanoparticles on DNA scaffolds and directing the bottom-up assembly of nanomaterials, which await their further technological implementation in many fields related to molecular electronics, biosensors, optical waveguides, etc.

3.2.1. Optical mapping of DNA sequences

AFM (Wilkinson et al. 2008) or EM (Kunkel et al. 2015) visualization of several bulky nano-objects along the irregular contour of a DNA molecule spotted on a Mica surface gives a nice qualitative illustration, but is poorly suited for fast parallel analysis of DNA molecules containing a large number of target sites. Direct determination of physical distances (positioning) between the specific sites becomes possible on stretched-out DNA molecules, leading to a visual pattern characteristic of that particular DNA. Such a linear representation of a DNA sequence, called optical map, can be read as a barcode and analyzed with a high degree of automation (Figure 3a). Direct single-molecule analysis of large DNA fragments, which far exceed the read length of widely used sequencing technologies, provides valuable genomic information for the identification of structural or copy number variations and assists with DNA sequence assembly or rapid strain typing (reviewed in Levy-Sakin and Ebenstein 2013). Several partially or fully automated optical DNA mapping platforms (BioNano Genomics, GenomicVision, PathoGenetix, OpGen) are already available. However, implementation of various known methods for optical DNA map generation is dependent on many technical parameters related to the degree and accuracy of label incorporation, repertoire of available target sites, covalent continuity of labeled DNA strands, inhomogeneous stretching, chemical and physical stability of the fluorophores, resolution and speed of signal readout, etc. Existing methods for specific visual pattern generation in optical DNA mapping include restriction map generation (Teague et al. 2010), nick-translation (Lam et al. 2012) or probe hybridization (Weier et al. 1995), all of which suffer from one or more of the above listed limitations (discussed in Zohar and Muller 2011 and Levy-Sakin and Ebenstein 2013).

Figure 3. Major applications of methyltransferase-directed labeling in genome studies.

(a) Optical DNA mapping using fluorescent mTAG labeling (Neely et al. 2010). A two-step mTAG reaction involving an engineered version of HhaI methyltransferase (eM.HhaI) was used for fluorescent labeling of GCGC sites in bacteriophage lambda DNA (a1). The labeled DNA molecules were stretched out by combing and positions of the fluorophores on individual DNA molecules were determined using super-resolution imaging (a2). Illustration in the upper right corner shows: 1 and 2 – experimental consensus fluorocodes derived by using different processing parameters; 3 - in silico generated (theoretical) reference map (adapted from Neely et al. 2010). (b) DNA “unmethylome” profiling by covalent mTAG labeling of unmodified CG sites (Kriukienė et al. 2013). Unmodified CG sites in fragmented genomic DNA from the human brain (50–300 bp fragments) were biotin-tagged in a two-step mTAG labeling reaction involving an engineered variant of the SssI methyltransferase (eM.SssI) (b1). Biotin-tagged fragments were affinity-enriched and sequenced to produce a genome-wide profile of unmodified CG sites (b2). Illustration in the lower right corner shows a genome browser view of mTAG-seq data over a part of SHANK3 gene including its promoter, overlapping with a predicted CpG island region (adapted from Kriukienė et al. 2013). C^m refers to naturally modified cytosine (mC/hmC/fC/caC).

The MTase-based approaches, owing to their unique combination of high specificity, covalent bonding and DNA strand preservation, appear particularly suited for this purpose. In a proof of principle study, two-step mTAG labeling was employed to attach fluorophores on 215 HhaI sites in bacteriophage lambda DNA (48.5 kb, see Figure 3a) (Neely et al. 2010). The labeling employed engineered M.HhaI and a cofactor bearing a transferable linear side chain with a terminal amino group followed by a chemoselective attachment of an Atto647N dye. The DNA molecules were stretched by combing onto polymer-coated coverslips using an evaporating droplet technique. Positions of fluorophores along individual DNA strands were recorded at sub-diffraction resolution (10 nm, or just 20 bp) using dSTORM imaging, which utilized photobleaching of the fluorophores to ensure that single emitters are isolated and their positions accurately determined. Whilst an average density of localized sites of approximately one per 650 bases represented a significant advance compared to other DNA mapping technologies, this density was achieved by assigning only 34% of the 215 available HhaI sites on the DNA. To further increase the number of fluorophores in the experimentally derived map, a consensus “fluorocode” encompassing nearly 90% of the sites (density 1/270 bp) was generated from twenty automatically aligned molecules.

To further explore the technicalities of the labeling reaction for improved optical mapping of DNA, a cofactor carrying a short allylic side chain with a terminal alkyne group (AdoEnYn) was used along with CuAAC mediated attachment of a fluorophore (Vranken et al. 2014). Eleven MTases were screened for activity with this cofactor, of which three adenine-specific enzymes were found to be active. The CuAAC-based approach generated bacteriophage T7 fluorocodes with labeling efficiency reaching 70%, however the authors noted substantial degradation of DNA in the presence of Cu(I), which precluded generation of full-length labeled DNA molecules. Given that a wide selection of mTAG cofactors along with a proven set of engineered and wt MTases is now available, a successful implementation of this approach (for instance, via a copper-free AAC reaction or one-step labeling) seems just around the corner.

To this end, one-step mTAG labeling with M.TaqI and a propargylic cofactor carrying a linker-bound TAMRA fluorophore was used to label over 200 target sites on lambda DNA (Grunwald et al. 2015) followed by physical stretching of the DNA molecules in commercial microchip-based nanochannel arrays. A fluorescent signature of DNA was generated by “conventional resolution” imaging, i.e. measuring the amplitude modulations of fluorescence intensity along its length rather than isolated fluorescent spots. This resulted in lower resolution density profiles characteristic of an underlying DNA molecule. Although nearly quantitative labeling was presumably achieved, images of individual molecules did not appear identical, pointing to the stochastic nature of single molecule measurements. Nevertheless, the generated consensus profiles permitted a clear distinction between the two types of bacteriophage DNA used. The generality, rapidness and high-throughput make this concept promising for routine applications in strain typing assays.

The SMILing technique proved instrumental for sparse fluorescent labeling of the T7 bacteriophage genome (3 sites per ~40 kb) using M.BseCI and a biotinylated aziridine-based cofactor in a two color-experiment. Streptavidin-coated quantum dots were attached at the biotinylated target sites as genomic reference tags to aid mapping the locations of non-covalently bound RNA polymerase molecules labeled with differently colored probes (Kim et al. 2012). Introduction of such reference tags allowed a higher precision in the assignment of the RNAP binding sites relative to localization based on their distance from DNA ends.

Altogether, the above examples demonstrate that MTase-directed approaches provide a valuable addition to the toolbox of sequence-specific labeling techniques, which will accelerate the development of automated high throughput technologies for optical DNA mapping.

3.2.2. Applications of MTase-directed labeling in epigenomics.

Yet another emerging direction of practical utility for MTase-directed sequence-specific labeling is the analysis of modified target sites in natural DNA. As discussed in the previous chapters, the prevalent covalent modification of the genome is sequence-specific methylation of cytosine and adenine residues. In higher eukaryotes including mammals, DNA cytosine-5 methylation predominantly occurs at CpG dinucleotides and acts as a stably inherited modification affecting gene regulation and cellular differentiation. Aberrant DNA methylation is an early and fundamental event in pathogenesis of many human diseases, including cancer (reviewed in Jones 2012). Beside 5mC, other DNA modifications have been discovered recently, giving rise to extensive discussions of their potential roles as epigenetic marks (reviewed in Kriukienė et al., 2012). To gain mechanistic insights into the dynamics and function of DNA methylation, genome-wide analyses of DNA modification patterns have been performed in different organisms and cell types employing a variety of profiling techniques (Weber et al., 2005; Schumacher et al., 2006; Bock 2008; Harris et al., 2010). During the past few years, chemical tagging of modification sites has been adapted for in vitro epigenome studies. Covalent derivatization of modified residues permitted incorporation of reactive azide, keto or primary amine groups followed by chemo-selective conjugation of biotin (Song et al. 2011; Pastor et al. 2011; Zhang et al. 2013).

The key concept of using MTase-directed labeling for analysis of mammalian genomic DNA lies in selective covalent tagging of the unmodified fraction of CG sites whereas the naturally modified sites will remain untagged due to pre-existing modification of the target residue. As unmodified cytosines represent a smaller proportion of CG sites compared with methylated ones (depending on the tissue, 65–80% of cytosines in the human genome are methylated (reviewed in Suzuki and Bird 2008)), analysis of this smaller, unmethylated DNA fraction may reduce the number of statistical comparisons and is more sensitive for detecting subtle epigenetic changes. An early attempt to analyze DNA methylation sites through targeted DNA scission (Comstock and Rajski 2005b) used derivatization of model oligodeoxynucleotide substrates with M.TaqI or M.HhaI and an azide-bearing aziridine cofactor, which was further subjected to the Staudinger ligation with triarylphosphines derivatized with phenanthroline. Presentation of these duplexes with Cu(II) promoted strand scission at the vicinity of the base modified by the enzyme. However, this chemistry leads to extensive DNA damage, and the remaining DNA fragments are not readily analyzed by modern sequencing techniques.

A more recent demonstration of chemo-enzymatic profiling of the unmodified fraction of the genome (named ‘unmethylome’) was based on selective covalent capture of CG sites (Kriukienė et al., 2013) (Figure 3b). Covalent tagging of DNA was performed using the engineered version of M.SssI (see section 2) and a synthetic mTAG type analog of AdoMet cofactor carrying a terminal amine or azide group. In the next step, conventional chemoselective coupling of the amine group with a NHS-biotin probe or, alternatively, SPAAC of the attached azide group with a dibenzocyclooctyne biotin reagent was employed. Biotin-labeled DNA fragments were then enriched on streptavidin beads and analyzed on tiling DNA microarrays (mTAG-chip) or by next-generation sequencing (mTAG-seq). Pilot profiling studies of human DNA samples from cultured cells and tissues demonstrated that this approach offers nanogram sensitivity and permits identification of unmethylated CG sites genome-wide with high precision and reproducibility. Moreover, mTAG-seq can be considered not only as a powerful and economical alternative, but also as a complementary technique to 5mC-specific methods such as affinity-based method MeDIP (Weber et al., 2005) and, likely, to TAmC-seq, a method for covalent derivatization and analysis of methylated cytosines (Zhang et al. 2013). Most recent modification of the approach includes covalent tethering of a priming oligonucleotide to the tagged nucleotides, which offers a particularly cost-effective direct genomic mapping of each unmodified CG site at near single-base resolution (Kriukienė et al., manuscript in preparation).

4. Cofactor-independent MTase-directed labeling

In addition to their well characterized catalytic activity, DNA C5-MTases were found to catalyze atypical reactions involving non-cofactor-like substrates. As mentioned above (Chapter 3), the C5-MTases use a covalent mechanism for nucleophilic activation of their target cytosine residues. The transiently generated activated cytosine intermediate is not only active toward AdoMet or its synthetic analogs, but can also attack other exogenous electrophiles such as aliphatic aldehydes, yielding corresponding 5-α-hydroxyalkylcytosines (Liutkevičiūtė et al. 2009) (Figure 4a). The reactions occur under fairly mild conditions and retain the high sequence and base specificity characteristic of bacterial DNA MTases. The coupling with formaldehyde yields 5-hydroxymethylcytosine (hmC), which is a naturally occurring cytosine modification in mammalian DNA (Kriaucionis and Heintz 2009; Tahiliani et al. 2009). Although the hydroxymethyl groups themselves are not good chemical reporters, they add a unique functionality to DNA (analogous to benzylic hydroxyl) that can be exploited for chemical or enzymatic derivatization. For example, a mild oxidation to formyl or keto groups would enable a further conjugation with compounds carrying hydrazine or hydroxylamine functions (Prescher and Bertozzi 2005). Alternatively, hmC residues can be enzymatically glucosylated (Gommers-Ampt and Borst 1995), thereby permitting selective DNA labeling through application of glycan modification/recognition techniques (Chittaboina et al. 2005; Song et al. 2011).

Figure 4. Cofactor-independent methyltransferase-directed sequence-specific derivatization of DNA.

(a) Transformations of a target cytosine catalyzed by DNA C5-MTases. Biological methylation by C5-MTases occur via an S_N2 reaction between an activated cytosine intermediate (ACI) and cofactor AdoMet, yielding 5-methylcytosine (5mC) (Biological C5-methylation). The ACI can undergo nucleophilic addition reaction with short aliphatic exogenous aldehydes, which in the case of formaldehyde yields hmC (C5-hydroxymethylation). In the reverse reaction, hmC residues can be converted to unmodified cytosines in DNA by the enzyme (Dehydroxymethylation of hmC). Similarly, 5-carboxycytosine (caC) can be converted to cytosine (Decarboxylation of caC). hmC residues, including those naturally occurring in DNA, can undergo further methyltransferase-directed condensation with thiol or selenol reagents to give stable 5-alkyl chalcogenomethyl derivatives (5-Thioalkylation of hmC). Modifying reagents are shown in red and green (thiol), C5-MTase and its catalytic moieties are shown in blue. (b) MTase-directed covalent amino-derivatization and labeling of hmCG dinucleotides in DNA with biotin (shown as a ball).

Curiously, it was also found that the covalent activation of 5-substituted cytosine residues present at the target position of a C5-MTase, can lead to their conversion into unmodified cytosine (Figure 4a). This reaction does occur with hmC and 5-carboxylcytosine, but was not observed with 5-formylcytosine (Liutkevičiūtė et al. 2009; Liutkevičiūtė et al. 2014). The MTase-activated hmC in DNA can also undergo condensation with exogenous aliphatic thiols and selenols yielding corresponding 5-alkylchalcogenomethyl derivatives (Liutkevičiūtė et al. 2011) (Figure 4a). Since this MTase-directed derivatization reaction is not possible at 5-methylated and unmodified cytosine residues, it appears well-suited for selective covalent capture of 5-hydroxymethylated-CG sites in mammalian genomic DNA. As a proof of concept, C5-MTase-directed derivatization of hmC with cysteamine and subsequent amine-selective biotin-labeling (Figure 4b) was demonstrated on plasmid DNA and model DNA fragments (Liutkevičiūtė et al. 2011), and was subsequently implemented in a commercial analytical tool (EpiJet 5-hmC Enrichment Kit). Moreover, M.HhaI and M.SssI have recently been shown to render sequence-specific conjugation of short Cys-containing peptides, to hmC-containing DNA (Serva and Lagunavičius 2015).

Altogether, the presented variety of atypical reactions demonstrate a high catalytic plasticity of DNA C5-MTases and offer additional ways for sequence-specific derivatization of canonical and modified bases within DNA. As compared to the cofactor-based reactions, these reactions typically require simpler and less expensive compounds thereby avoiding multistep syntheses of AdoMet analogs.

Conclusions/outlook

MTase-directed labeling of DNA is an enabling technology with many unique demonstrated applications. Due to its relative simplicity, robustness and wide-range applicability this approach is becoming a method of choice where targeted covalent derivatization of DNA is required. Although certain technical questions still require attention, the rapidly growing popularity indicates that the field is approaching its maturity stage. The two most developed applications of the method are optical DNA mapping and analysis of epigenetic states in mammalian DNA; both methods are now entering the phases of automation and commercial exploitation, and no doubt will soon become commonly used technologies. Another important area of research that is poised to see a rapid bloom in the near future is DNA labeling in biological systems and in living cells. Currently, two main obstacles can be envisioned: 1) entrance/delivery of cofactor analogs into cells; 2) design of highly orthogonal cofactor-MTases pairs for allele-specific labeling. The first issue can be addressed by harnessing cell delivery systems, which are widely used to cargo a variety of other molecules across the cell membrane (Janib et al. 2010; Falanga et al. 2015), or by enzymatic production of cofactor analogs in situ from corresponding methionine analogs, which show superior wall-penetration properties (Wang et al. 2013; Singh et al. 2014).