Nucleoside Analogs in the Study of the Epitranscriptome

Abstract

Over 150 unique RNA modifications are now known including several nonstandard nucleotides present in the body of messenger RNAs. These modifications can alter a transcript’s function and are collectively referred to as the epitrancriptome. Chemically modified nucleoside analogs are poised to play an important role in the study of these epitranscriptomic marks. Introduced chemical features on nucleic acid strands provide unique structures or reactivity that can be used for downstream detection or quantification. Three methods are commonly used in the field to synthesize oligonucleotides containing chemically modified nucleoside analogs. Nucleoside analogs can be introduced via phosphoramidite-based chemical synthesis, via polymerases with modified nucleotide triphosphates or by metabolic labeling. In this review, these methods for incorporation of nucleoside analogs will be discussed with specific recently published examples pertaining to the study of the epitranscriptome.

1. Introduction

Modifications to the four common nucleosides present in RNA have been known for over five decades [1–4]. However, modern analytical techniques, such as mass spectrometry and next generation sequencing (NGS), have caused a renaissance in their study. Naturally occurring RNA modifications are introduced into the polynucleotide chain by RNA modifying enzymes and are ubiquitous. To date, over 150 unique and structurally distinct RNA modifications have been discovered and characterized [5]. The vast majority of these modifications are found in tRNA and rRNA, but several nonstandard nucleotides including inosine (I) [6, 7], N⁶-methyladenosine (m⁶A) [8], N⁶-2’-O-dimethyladenosine (m⁶Am) [9] and pseudouridine (Ψ) [10], have been located in mRNA as well (Figure 1). These modifications have been referred to as the epitrancriptome to highlight their modulation of transcript function in analogy to the ways epigenetic modifications can alter genome function [11]. Epitranscriptomic modifications can cause changes in the fate or function of RNA by altering base-pairing, pi-stacking, or metal-chelation [12] (among others). In this review, we will highlight strategic uses of chemically synthesized nucleoside analogs and analog-containing nucleic acids in the study of the epitranscriptome. Nucleoside analogs have been developed that facilitate the detection of specific modifications in the transcriptome, the quantification of different types of modifications and for the study of the structure of modified RNA or RNA modifying enzyme/substrate complexes.

Figure 1.

The structures of naturally occurring mRNA modifications found in mammals. For clarity, the RNA modification is highlighted in red. Abbreviations of the modification are indicated below each molecule. Inosine (I), N⁶-methyladenosine (m⁶A), pseudouridine (Ψ), N⁶-2’-O-dimethyladenosine (m⁶Am).

Nucleoside analog-containing nucleic acids can provide insight on mechanism or binding specificity which would otherwise be hard to elucidate by traditional biochemical techniques. Likewise, these analogs can be used to directly show the importance of functional group contacts at a nucleic acid-protein interface. Most commonly, nucleoside analogs are incorporated into oligonucleotides by synthesizing the corresponding phosphoramidite monomer and using it in automated chemical synthesis. Alternatively, triphosphates bearing the desired chemically modified analog can be incorporated via polymerases [13–15]. In addition to these methods, metabolic labeling can be used to prepare nucleoside analog-containing nucleic acids [16, 17]. In this method, the nucleoside analog or its precursor is taken up by metabolically active cells and incorporated into newly synthesized strands.

2. Metabolic Labeling Method

Metabolic labeling has been broadly applied to label different types of biomolecules (carbohydrates, proteins, lipids..etc) [18], but this review will focus on metabolically labeling RNA. Unnatural metabolites can enter cellular metabolic pathways where they function much like their native counterparts. Downstream detection is made possible because of unique chemical features of the unnatural metabolite. A well-established method for identifying RNA modifications is mass spectrometry. For example, supplying cellular cultures with isotopically labeled glucose results in isotopically labeled mRNA, which can be detected by mass spectrometry [19]. Furthermore, several nucleoside analogs have been developed for the purpose of labeling nascent RNA [16, 17].

A major requirement of this method is that the metabolite bearing the chemical modification must be cell permeable and non-toxic to the cells. Metabolic labeling provides a unique opportunity to probe cellular processes in vivo that would otherwise be difficult to study.

2.1. SAM analog generated by metabolic labeling to identify m⁶A sites

S-Adenosylmethionine (SAM) is the primary source of methyl groups in nearly all methylation reactions within the cell [20]. The most common internal modification of eukaryotic mRNA is N⁶-methyladenosine (m⁶A) and is of particular interest because of its role in a vast array of biological processes but mostly, due to its impact on mRNA stability and abundance of mRNA transcripts [21]. Adenosine methylation is thought to be a dynamic process that is controlled by three classes of proteins, “writers”, “readers” and “erasers”. The modification is introduced by a “writer” complex comprised of a SAM-dependent heterodimer of methyltransferases, Mettl3 and Mettl14 (Mettl3–14) and prefers the consensus sequence “DR(m⁶A)CH” (D=A or G or U; R=G or A; H= A,C or U) [22]. While early studies showed both Mettl3 and Mettl14 as catalytically active, more recently it has been shown that Mettl3 is solely responsible for the methyltransferase activity [23]. The Wilms tumor-1 associated protein (WTAP), an RNA binding protein, recruits the Mettl3–14 complex and aides in its localization to nuclear speckles [24–26]. A different SAM-dependent methyltransferase, Mettl16, has recently been shown to have a distinct substrate specificity compared to the Mettl3–14 complex [27–29]. The most studied m⁶A binding proteins (“readers”) are the YTH protein family which have been shown to play roles in nearly every mRNA function including splicing, stability, translation and silencing [30–32]. “Erasers” are responsible for the removal of the m⁶A methylation and include two proteins in the AlkB family. Fat mass and obesity-associated protein (FTO) and ALKBH5 are known to demethylate m⁶A both in vitro and in vivo [33, 34]. Most reported methods for mapping m⁶A modifications require immunoprecipitation and RNA-sequencing because m⁶A is indistinguishable from A using readily available polymerases. While these methods allow whole transcriptome coverage, a lack of nucleotide resolution, scalability and antibody bias are major limitations [8, 35–38].

To overcome some of these limitations, new chemically modified nucleic acids were created for the purpose of chemically labelling m⁶A to improve direct detection [39, 40]. Shu et. al synthesized a modified SAM cofactor that replaced the methyl group with an allyl group [39] (Figure 2A). Isolated polyadenylated RNAs from HeLa cells could be incubated in the presence of Mettl3–14 and allyl-SAM to transfer the allyl group to specific adenosines. The allyl group can be treated with iodine (I₂) to promote a cyclization reaction that blocks the Watson-Crick face of the nucleobase (Figure 2A). This causes reverse transcriptases (RTs) to misread the modification site during extension (unlike m⁶A, which is read as A by RTs). The resulting cDNA is then sequenced and Mettl3–14 sites can be identified as A to T/C/G mutations, which can further be quantified. Allyl labeling was used to confirm the location of a known m⁶A site in total RNA, but at a lower level than when using the previously reported SCARLET method [8]. The difference in the two methods was suggested to be low allyl transfer or incomplete cyclization of the allyl group when treated with iodine.

Figure 2.

The study of m⁶A using SAM analogs to identify Mettl3–14 modification sites. A) Isolated mRNA from HeLa cells can be incubated in the presence of an allyl SAM cofactor and the Mettl3–14 complex to selectively label m⁶A sites with a⁶A [39]. Subsequently, mRNA can be reacted with iodine to produce a cyclized a⁶A molecule that is no longer RT silent. After mRNA is converted into cDNA and sequenced, there will be an A to T/G/C mutation which is indictive of a m⁶A site. B) Propargyl-L-selenohomocysteine can be taken up by HeLa cells and incorporated into a propargyl SAM analog [40]. The propargyl group is transferred by cellular methyltransferases to m⁶A sites. The isolated mRNA containing p⁶A can be subjected to CuAAC click reactions to install bulky groups that block RT. cDNA obtained from RT can be sequenced using NGS. As a result of installation of the bulky group, coverage drops will occur where there are m⁶A sites.

Building on the allyl modification idea, Hartstock et. al developed a method for directly detecting m⁶A sites by replacing the SAM methyl group with a propargyl group [40] (Figure 2B). The propargyl group is of particular interest because of its ability to react in the highly efficient and bio-orthogonal copper-catalyzed azide/alkyne cycloaddition (CuAAC) reaction. The authors first demonstrated that Mettl3–14 could efficiently transfer a propargyl group from a synthetic selenium-containing SAM analog. The selenium-based SAM analog has two main advantages over sulfur-containing analogs: greater stability and better activation [41]. The technique was further expanded by feeding the precursory propargyl-L-selenohomocysteine amino acid to HeLa cells. The metabolic precursor is incorporated into a SAM analog and the propargyl is donated by methyltransferases to cellular RNA. The alkyne RNA can react with an azide-biotin and be pulled down with streptavidin beads. The resulting RNA can be sequenced through next-generation sequencing (NGS) and methylation sites can be determined by termination during reverse transcription (RT) (Figure 2B). The authors note three main advantages of using propargyl over the allyl: 1) CuAAC is bio-orthogonal and highly specific towards azides, 2) allows for non-natural modifications to be introduced in vivo and 3) the propargyl was more efficiently introduced, likely due to the selenium-based SAM analog. While this method was used to validate m⁶A rRNA sites, future studies are likely to apply this method to study the m⁶A in mRNA.

3. Non-standard nucleotide triphosphate incorporation

Amplification of RNA and DNA in vitro by polymerases has transformed the landscape of biotechnology and enabled the development of many widely used techniques such as routine sequencing, library generation and new ways of detecting infectious diseases [42–45]. Nucleotide triphosphates (NTPs) are the building blocks in the synthesis of oligonucleotides using polymerases. Fortunately, NTPs are chemically very manipulatable and can be functionalized at a variety of locations. This allows creation of a diverse library of NTPs that are unique and can be used for downstream applications like labeling RNAs for pull-down assays [17]. Recognition of non-standard NTPs by DNA polymerase has been heavily researched with the goal being determining which NTPs are suitable for PCR and chemical biology applications [46–49]. The general consensus is non-standard NTPs are less efficiently incorporated than the canonical bases, however, some modifications are more tolerated than others. It has now been established that atom substitutions and major groove modifications are more efficiently incorporated compared to minor groove modifications. For adding linkers, like an alkyne functional group, the C7 and C5 are best suited for purines and pyrimidines, respectively [50]. Evolution of polymerases that have higher fidelity with non-standard NTPs is a promising strategy for increasing yields of long oligonucleotides bearing multiple non-standard bases[51]. In vitro incorporation of non-standard NTP incorporation into DNA and RNA provides a low-cost but highly informative biochemical technique that can be widely adopted by any lab.

3.1. Selenium-containing triphosphate to probe m⁶A sites

The study of the epitranscriptome frequently involves variants of next generation sequencing (NGS) approaches. For the detection of RNA modifications, samples for NGS are often enriched prior to sequencing using modification-specific antibodies or chemically modified to alter the structure of the modification. Strategies that involve immunoprecipitation for enrichment can suffer from antibody bias and do not provide nucleotide resolution in modification site determination[52, 53]. For modifications that are not distinguishable from the native nucleoside by RT, chemical conversion has been used to generate a species that causes reverse transcriptase to stop or to introduce mutations in the cDNA, allowing the precise location of the modification to be determined from the sequence reads [54]. However, the pre-treatment chemistry can involve harsh reagents and/or incomplete conversion. Chemical conversion or immunoprecipitation have been used to identify the presence of RNA modifications such as 1-methyl adenosine [55], m⁶A[53], I [56], and Ψ [10, 57], however their exact levels are highly debated [37]. The m⁶A modification is typically reverse transcribed like A, so detection requires a chemical modification to be installed, like the p⁶A, or immunoprecipitation for enrichment.

Recently, the Zhou group synthesized a C4 selenium-containing thymidine 2’-deoxytriphosphate (4SeT) to allow for the direct detection of m⁶A in RNA by RT [58] (Figure 3). Their data shows that 4SeT, when it is subjected to RT, is able to base pair with normal adenosines, but the strand will be truncated when the 4SeT is base paired across m⁶A. The atomic radius of selenium is thought to disrupt proper base-paring between 4SeT and m⁶A. As a result of the cDNA truncation during RT, there is a detectable decrease in coverage at the position of m⁶A. In order to prepare RNA for NGS, an FTO-assisted sequencing technique was developed. As mentioned in the previous section, FTO is capable of demethylating m⁶A. The m⁶A-containing RNA can be treated with or without FTO. FTO treated samples act as controls so a reduction in coverage can be detected (Figure 3). A change in coverage would then be indicative of the presence of an m⁶A modification.

Figure 3.

The method of using a DNA triphosphate analog to distinguish m⁶A sites during RT [58]. A C4 selenium containing thymidine DNA triphosphate can be synthesized in several steps from thymidine. During RT, 4SeT can successfully be incorporated across from adenosine but will fail to be incorporated across from m⁶A, resulting in truncated cDNA. The resulting cDNA can be sequenced through NGS or specific targets can be evaluated by gel analysis. To aide in the determination of truncations, RNA can be treated with or without FTO prior to RT. Appearance of a truncated band by denaturing gel analysis is evidence of a m⁶A site. In NGS data, this would be determined by a coverage drop due to truncation of cDNA.

While this method was capable of probing m⁶A sites in a small RNA construct, secondary structure of longer mRNA may interfere with the demethylation reaction and m⁶A sites might not be detected. Also, the seven-step synthesis of 4SeT may hinder this technique from being widely adopted by labs who lack equipment and expertise to carry out the synthesis. Despite these downsides, this method could provide single nucleotide resolution probing of m⁶A sites and not require immunoprecipitation, potentially avoiding antibody bias.

4. Solid phase oligonucleotide synthesis to incorporate nucleoside analogs

Starting in the 1950’s, Merrifield began developing solid phase synthesis for polypeptides [59]. By anchoring the first amino acid to an insoluble solid support, the protected amino acid monomers could be added one at a time for an iterative, chemical approach to peptide synthesis. Thirty years later, an efficient solid phase, automated method was developed to generate oligonucleotides by utilizing phosphoramidite monomers as building blocks [60]. Now the phosphoramidite-based solid phase synthesis strategy is one of the most commonly used for the generation of RNA strands bearing nucleoside analogs. Because of the coupling efficiencies of a typical RNA solid phase synthesizer, strand length is restricted to less than ~ 120 nucleotides and is usually much shorter, in the ~ 50 nucleotide range. While oligonucleotide length can be a shortcoming, the phosphoramidite method is more compatible with heavily functionalized nucleoside analogs compared to the other methods discussed above since recognition by metabolic enzymes or polymerases is not required. This technique does rely on having a synthetic chemistry skill-set. However, many chemically modified phosphoramidite monomers are currently commercially available.

4.1. An adenosine deamination transition state analog to trap ADAR-RNA complexes

Recently, our lab applied the phosphoramidite method to generate nucleoside analog-containing RNAs useful for the study of adenosine deaminases that act on RNA (ADARs), the enzymes responsible for the incorporation of inosine in mRNA [61–63]. Since conversion of adenosine to inosine changes the base pairing properties of the nucleobase, this reaction is a form of RNA editing [64]. Due to the change in Watson-Crick pairing, this conversion can have profound downstream consequences such as codon changes, alternative splicing and alteration of protein-RNA interactions [65–67]. In humans, two active RNA editing deaminases exist, (ADAR1 and ADAR2) and one that is catalytically inactive (ADAR3) [68]. It is known that ADARs play essential roles in proper cellular functions and aberrant RNA editing is linked to disease [69]. Indeed, mutations in the ADAR1 gene are known to be a cause of Acardi Goutieres Syndrome [70]. A variety of methods exist for transcriptome-wide sequencing of A-to-I editing events [6, 7, 56]. However, these methods are insufficient for determining ADAR’s biochemical preferences for binding and catalytic activity.

To address these knowledge gaps, high resolution structures of ADAR bound to RNA were needed. These were recently provided in the form of crystal structures of the human ADAR2 deaminase domain bound to RNA [71]. In order to solve such structures, two advances were required. The first was a method to trap ADAR bound to RNA in a complex relevant to the editing reaction. For this purpose, the purine analog 8-azanebularine (8-azaN) was used [72]. The covalent hydrate of 8-azaN is an excellent mimic of the high energy tetrahedral intermediate expected for the adenosine deamination reaction (Figure 4A, 4B). The other breakthrough was the identification of RNA sequences that were efficiently edited by an ADAR catalytic domain in vitro and formed a tight and specific binding complex with the protein when modified with 8-azaN [73]. The combination of 8-azaN and highly efficient ADAR2 substrate sequences enabled structural characterization of ADAR2-RNA complexes (Figure 4C).

Figure 4.

Method for crystallization of ADAR2 bound to dsRNA containing a transition state analog [71]. A) A-to-I editing reaction pathway showing the hydrated intermediate that leads to the formation of inosine. B) 8-Aza-nebularine when subjected to the ADAR reactions becomes a trapped hydrated intermediate. C) 8-Azanebularine phosphoramidite can be synthesized from the starting materials of 8-azaadenosine and a tetraacetate protected ribofuranose. Subsequently, the exocyclic amine can be removed, and the sugar’s protecting groups removed. The standard phosphoramidite protecting groups are installed prior to RNA solid phase synthesis. Finally, crystal structures can be solved of ADAR2 with dsRNA containing 8-azanebularine. This method afforded four high resolution crystal structures of ADAR2 bound to its substrate.

Generation of 8-azaN-modifed RNA requires synthesis of an 8-azaN phosphoramidite monomer [74]. The synthesis of this compound starts with a Vorbruggen glycosylation reaction between the modified base 8-azaadenine and a tetraacetate protected ribofuranose. Subsequent steps remove the C6 exocyclic amine and deprotection of the sugar affording 8-azanebularine free nucleoside. The next two synthetic steps are to install the protecting groups required for automated synthesis, 5’-O-DMTr and 2’-O-TBDMS. Lastly, the nucleoside analog is subjected to the phosphoramidite reaction. This reaction must be anhydrous and anaerobic because oxidation of the phosphoramidite nucleoside will result in a poor coupling efficiency in the solid phase synthesizer. The purified phosphoramidite monomer bearing the nucleoside modification is analyzed by ³¹P NMR to assess the oxidation level. The phosphoramidite monomer is then incorporated into a specific location within the RNA via the solid phase synthesis method. The RNA bearing the nucleoside analog is purified away from failure sequences by denaturing urea poly-acrylamide gel electrophoresis and confirmed by mass spectroscopy. The desired RNA is hybridized with its complementary strand to form the duplex recognized by ADAR. The ADAR2 deaminase domain is overexpressed in Saccharomyces cerevisiae and purified. S. cerevisiae is a preferred expression system for producing large amounts of ADARs, primarily because E. coli does not produce inositol hexakisphosphate (IP₆), a co-factor essential for proper folding and activity of ADAR [63]. To obtain protein of suitable quality for crystallography, it must be purified over Ni-NTA, heparin and finally size-exclusion chromatography. Lastly, crystal trays can be set up with crystallography quality ADAR2 deaminase and the RNA duplex containing 8-azanebularine.

The ADAR2-RNA structural studies enabled by this method helped explain editing site selectivity and revealed previously unknown contacts made between protein and RNA [71]. In each of the four crystal structures obtained of ADAR2-bound to dsRNA, the 8-azaN hydrate was observed in the active site of the protein. While some of the important residues were previously identified by sequence alignment of ADARs and subsequent alanine scan [75], the new structures provided atomic-resolution detail for their roles and revealed previously uncharacterized protein loops important for RNA recognition.

ADARs have preferences for specific nucleotides neighboring editing sites [76], however the reasons these preferences were unknown. ADAR2-RNA crystal structures revealed an interaction between the 2-amino group of the 3’ nearest-neighbor G and the carbonyl backbone of S486. Because this was not an amino acid side chain interaction to the RNA, a point mutation in ADAR2 would not be expected to have a substantial effect. Instead, to further validate the observations in the crystal structure, RNA substrates containing the guanosine analogs inosine (I), N²-methylguanosine (N²MeG) and 2-aminopurine (2AP) adjacent to the editing site were prepared by solid phase chemical synthesis, allowing for an investigation of the importance of the guanosine 2-amino group. Deamination reactions between these model substrates and ADAR2 were then performed. Together, the observed results highlighted the importance of this contact and its likely role in determining ADAR2’s 3’ nearest neighbor preference. Thus, nucleoside analog-containing RNA proved to be critical both in the formation of a stable complex for crystallization (e.g. 8-azaN RNA) and in testing the importance of specific contacts observed in the ADAR2-RNA crystal structures.

5. Concluding Remarks

In this review, we have highlighted recent examples of the use of nucleoside analog containing nucleic acids in the study of the epitranscriptome. Nucleoside analogs are introduced into nucleic acid strands by three commonly used methods; metabolic labeling, via polymerases and modified NTPs or by phosphoramidite-based solid phase synthesis. These methods allow for the introduction of unique chemical features into the strand imparted by the structure of the nucleoside analog. These features can then enable novel approaches to the study of the epitranscriptome such as new detection methods for m⁶A or mechanism-based trapping of an inosine-generating ADAR enzyme.

Highlights.

• Three commonly used methods for generating nucleoside analog-containing nucleic acids are described

• Examples of the use of each method in the study of the epitranscriptome are examined

• Strengths and limitations to each method discussed