Ryuta
Shioi‡
a,
Lu
Xiao‡
a,
Sayantan
Chatterjee
a and
Eric T.
Kool
*ab
aDepartment of Chemistry, Stanford University, Stanford, CA 94305, USA. E-mail: kool@stanford.edu
bSarafan ChEM-H, Stanford University, Stanford, CA 94305, USA
First published on 7th November 2023
The reactivity of RNA 2′-OH groups with acylating agents has recently been investigated for high-yield conjugation of RNA strands. To date, only achiral molecules have been studied for this reaction, despite the complex chiral structure of RNA. Here we prepare a set of chiral acylimidazoles and study their stereoselectivity in RNA reactions. Reactions performed with unfolded and folded RNAs reveal that positional selectivity and reactivity vary widely with local RNA macro-chirality. We further document remarkable effects of chirality on reagent reactivity, identifying an asymmetric reagent with 1000-fold greater reactivity than prior achiral reagents. In addition, we identify a chiral compound with higher RNA structural selectivity than any previously reported RNA-mapping species. Further, azide-containing homologs of a chiral dimethylalanine reagent were synthesized and applied to local RNA labeling, displaying 92% yield and 16:
1 diastereoselectivity. The results establish that reagent stereochemistry and chiral RNA structure are critical elements of small molecule-RNA reactions, and demonstrate new chemical strategies for selective RNA modification and probing.
![]() | ||
Fig. 1 Chiral reagents exhibit stereoselective acylation of 2′-OH hydroxyl groups in RNA. (a) Recently reported reagents reacting with RNA 2′-OH groups have been achiral and relatively flat. New reagents possess asymmetric centers, conferring stereoselectivity (Table 1). Heteroatoms (X) are included to enhance solubility and carbonyl electrophilicity. R denotes varied alkyl and aryl substitution. (b) Structures of chiral acylimidazole enantiopairs screened for diastereoselective reaction with RNA. Asterisks mark asymmetric centers. (c) MALDI-TOF mass spectra of the products of reaction of (R)-2 and (S)-2 (100 μM, 2 h, 0 °C) with a single-stranded RNA, showing up to five substitutions and high conversion for the (R) enantiomer, but only one substitution and low conversion for the (S) isomer. Red numerals denote number of adducts on the RNA; “+0” indicates unmodified starting RNA mass peak. |
RNA can fold into complex three-dimensional shapes, and is inherently chiral due to the (D)-ribose sugars that comprise the phosphodiester backbone. Chirality in nucleic acids is widely discussed and has been leveraged in numerous studies for selective recognition of small molecules with DNA;17–21 for example, chiral metal intercalator complexes have long been studied for their selectivity in associating with right-handed B-DNA helices versus left-handed structures.17 However, for RNA much less attention has been paid to chiral recognition by small molecules.22 A chief example has been the testing of RNA on solid support as a stationary phase for chiral chromatography.23,24 Chiral fluorinated diamines have also been tested as NMR probes of RNA.25 RNA aptamers selected for binding a naturally occurring small molecule have in a few cases also been shown to be selective for that enantiomer over the opposite one.26,27 While the number of small molecule ligands known to bind RNA is growing rapidly, the R-BIND database28 lists no enantiomer pairs among 131 ligands. However, in a recent example of chiral recognition by synthetic compounds, ligands for a bacterial RNA riboswitch have been shown to be stereoselective in binding.29
Although large oligonucleotide conjugates have been investigated more recently,30–32 a small number of early studies described reactions of nucleotides with chiral small molecules. Prebiotic chemistry experiments several decades ago tested the reactions of activated amino acids with dinucleotides in DMF solutions, and documented yields of up to 12% at internal 2′-OH and stereoselectivity up to 2:
1.33 One report also tested reactions with homopolymer RNAs, noting yields of 7% in reaction of polyU (for example) with an alanine derivative.34,35 As solid-phase RNA/DNA synthesis was not yet available, no defined RNA sequences or folded contexts were tested. Thus it remained unknown whether preparatively useful yields could be achieved with chiral compounds, whether higher levels of selectivity are possible, and whether folded RNA structure might affect stereoselectivity. Given this lack of information, multiple questions arise: to what degree can reagent stereochemistry influence reactions at 2′-OH? Second, does the addition of steric bulk that confers asymmetry near the reactive carbonyl hinder RNA reaction? Finally, and importantly, how does RNA folded structure locally influence diastereoselectivity?
To begin to address these questions, we report here the study of enantiopairs of aqueous-soluble acylimidazole reagents having varied asymmetric substitution near their reactive carbonyl groups. Our findings show that with appropriate substitution, reagents can demonstrate high diastereoselectivity and preparatively useful yields in RNA reaction. The results establish new strategies for selective RNA recognition, with applications in conjugation, modification, and mapping of the biopolymer.
Compound | Conversionb (%) | Diastereoselectivityc (R/S) |
---|---|---|
a Conditions: RNA acylation yields for the chiral acylating reagents with a single-stranded RNA. 10 μM RNA was treated with 200 mM reagent, 20% DMSO in water for 2 h at 0 °C. b Conversion of starting RNA to acylated products (at 2 h) as measured by MALDI-TOF MS. c This is an aggregate ratio of all reactive sites in the test RNA. d 80 mM reagent was used. e 100 μM reagent. f 50 mM reagent. g 100 mM reagent. h 200 mM reagent, 6 h. i 1 mM reagent. j 20 mM reagent. | ||
(R) 1d | 57 | 0.61 |
(S) 1d | 94 | |
(R) 2e | 98 | 3.92 |
(S) 2e | 25 | |
(R) 3 | 55 | 1.34 |
(S) 3 | 41 | |
(R) 4f | 56 | 1.16 |
(S) 4f | 48 | |
(R) 5g | 77 | 0.86 |
(S) 5g | 90 | |
(R) 6 | 57 | 0.83 |
(S) 6 | 69 | |
(R) 7h | 20 | 1.11 |
(S) 7h | 18 | |
(R) 8 | 38 | 2.00 |
(S) 8 | 19 | |
(R) 9i | 67 | 1.63 |
(S) 9i | 41 | |
(R) 10j | 72 | 0.78 |
(S) 10j | 92 | |
(R) 11 | 41 | 0.51 |
(S) 11 | 80 |
Among these chiral scaffolds, we observed that several yielded significant levels of diastereoselectivity in reaction with RNA (Table 1, Fig. S1 and S2†). Note that the reaction sites in the single-stranded RNA were not controlled, and thus conversions reflect aggregate yields, and diastereoselectivities represent aggregate selectivity in esterification reactions across these multiple 2′-OH sites. Highest diastereoselectivity in this initial screen was observed for compounds 2, 8, and 11, with selectivities of up to 3.9:
1 (R
:
S) for alanine derivative 2, and 2
:
1 for structurally related proline derivative 8. Interestingly, the stereochemical preference varied, with α-amino species showing preference for the R isomer (e.g.2, 3, 8), while others (primarily α-alkoxy compounds such as 1, 10, 11) reacted selectively as the S enantiomer.
Cations in solution can strongly affect RNA dynamics;36 however, we found that the presence of 100 mM Na+ or 6 mM Mg2+ in solution had no effect on reaction conversions or stereoselectivity of 2 with single-stranded RNA (Fig. 2c). We also investigated the effect of temperature on reactions (Fig. 2d), and observed increasing diastereoselectivity but decreasing reaction conversions with higher temperature.
Comparisons of reactions with RNA and DNA of the same sequence (Fig. 2e and S5†) ruled out significant reaction at exocyclic amines of the nucleobases.37 The results indicate that, as with most acylimidazoles tested to date, reaction occurs essentially exclusively at 2′-OH groups, yielding ester linkages. Even the primary 5′-alcohol of DNA does not appear to react substantially, consistent with the observation that the increased acidity of 2′-OH groups is important for promoting reaction with RNA.38 As an additional control for observations of stereoselectivity, we confirmed that the two enantiomers of the most selective scaffold (R,S)-2 react equally with an achiral alcohol (Fig. S6†).
For an initial test of whether RNA folded structure can influence diastereoselectivity of 2′-OH acylation, we compared reactions of acylimidazole scaffolds 1 and 2 with three RNAs: single-stranded RNA, a hairpin containing a 3 nt loop, and a hairpin containing a small 1-nt bulge loop (Fig. S7 and Table S1†). Diastereoselectivity (as measured in aggregate over multiple reaction sites) was not markedly different over the three contexts; again, the methoxy species displayed the opposite chiral preference ((S) favored) as compared with the dimethylamino compounds ((R) favored).
![]() | ||
Fig. 3 The effect of varied RNA folded structure on local reactivity and stereoselectivity of chiral reagents. (a and b) Relative reactivity of (R) and (S) enantiomers of 1 and 2 at each position within a library of folded RNAs12 (native (D) configuration), measured by deep sequencing. 884 nucleotides of sequence and 20 loop motifs are represented in the library. Loops are marked by dark bars and denoted by size, and stems fall in between. Sequences are as published;12 H = hairpin loop; B = bulge loop; I = internal loop; T = three-way junction loop, with numbers indicating loop size. Conditions: 50 mM 1 or 100 μM 2, 0 °C, 2 h. (c) Dynamic range of reactivity as measured by loop/stem (L/S) reactivity ratios for two enantiopairs of reagents. (d) Diastereoselectivity of reagents (R,S)-1 and (R,S)-2 with three single-stranded RNAs (SS-polyA, SS-polyU, mixed-sequence SS-combo). (e) Diastereoselectivity of reagents (R,S)-1 and (R,S)-2 averaged over all loops, all double-stranded stems, and for a single-stranded RNA (ss-combo) containing all bases. (f and g) Plots of diastereoselectivity at each position of loop nucleotides in the 43 RNAs, as measured by differential R/S reactivity of reagent 1 (f) and 2 (g). |
Interestingly, nucleotides in small bulge loops (e.g., B4, B5) and internal loops (I21, I22, I32) are the most reactive contexts for the chiral reagents in general, and indeed are more reactive than nucleotides in single-stranded RNAs (e.g. ss-polyA and ss-mixed sequence RNA (“ss-combo”) (Fig. 3a)). Unlike previous achiral compounds,12,39 certain small loops (B2, B3) that were highly reactive for achiral reagents proved to be relatively unreactive for the current chiral reagents. Also noteworthy for the chiral compounds is that small uridine loops are generally more reactive than those composed of adenosine nucleotides (Fig. S9†), which is the opposite of what was seen for previous achiral aryl species.12,39
Finally, we measured overall (averaged) selectivity of each compound ((R,S)-1, (R,S)-2) for acylation of unpaired loop residues over paired ones (stems) in the folded RNA library. This ratio has been used as a marker to evaluate acylating reagents for structure mapping; a high loop/stem (L/S) reactivity ratio indicates elevated signal over background. The data (Fig. 3c) show that each enantiomer has distinct loop/stem ratios, with compounds (R,S)-1 showing higher L/S values than (R,S)-2. The data for (R,S)-1 are particularly remarkable, with a higher value for (S)-1 than its mirror image; indeed, the L/S value of 8.5 for (S)-1 is considerably higher than that of all structure mapping acyl reagents currently in use.12,40
Positional diastereoselectivity in the folded RNAs varied both among distinct loop types and sizes, as well as along the different positions within a single loop, apparently reflecting distinct chiral structural contexts. For example, very small hairpin loops (H2, H3) exhibit relatively low diastereoselectivity in reactions with scaffold 1 (Fig. 3f), and a three-way junction with a 3-nucleotide loop (T3) displays the opposite stereochemical preference relative to most other loops. We also evaluated diastereoselectivity with three single-stranded RNAs in the library (ssRNA): polyU, polyA, and a mixed A,C,G,U sequence (ss-combo). There was very little reactivity with polyA, as seen previously for achiral reagents, presumably reflecting the high helicity of this sequence due to the strong stacking of adenines.41 As with folded sequences, scaffolds 1 and 2 showed opposite diastereoselectivity that was significant especially for single-stranded polyU (Fig. 3e and S10†). Overall, stereoselectivity is comparatively low for these relatively unconstrained RNAs, suggesting that the more rigid structural contexts in small loops enable greater discrimination between mirror-image reagents. The data also establish that sequence effects appear to be small for stereoselectivity as compared with folded structure, which has a significantly larger influence (Fig. S11†).
To further explore the role of reagent stereochemistry in local nucleotide reactivity across this RNA library, we examined nucleotide-by-nucleotide positional correlations of reactivity of enantiomers of 1 and 2 (Fig. S12†). Interestingly, enantiomers of 2 (the more stereoselective scaffold) show very little correlation with each other, while the enantiomers of 1 show a substantial positive correlation. This is also the case for prior relatively flat aryl reagents NAIN3 and 1M7, which also correlate well in their positional reactivity.12 The results suggest that compounds with greater asymmetry can better sense differences in local chiral environment around each 2′-OH.
Averaging diastereoselectivity of reagent scaffolds 1 and 2 over all loops (Fig. 3e) revealed remarkable diastereoselectivity in loops, but almost no asymmetric discrimination in duplexes (stems). Taken together, our RNA library data confirm that 2′-OH acylation by these chiral acylimidazole reagents occurs primarily at unpaired loop nucleotides, and that reactivity and diastereoselectivity, at least at low reaction conversions, vary greatly with local RNA loop structure. Small loops (especially bulge and internal loops) tend to show elevated reactivity and diastereoselectivity at discrete positions.
![]() | ||
Scheme 1 Synthesis of a chiral azide-functionalized analog of (R)-2 for local RNA labeling. Conditions: (a) azidoacetaldehyde, NaBH(OAc)3, DCM, 16 h, r.t. (73%); (c) HCl in 1,4-dioxane, DCM, 17 h, r.t. (87%); (d) CDI, DMSO, r.t., 2 h, quant. (as HCl salt). See ESI file† for synthesis of the (S) enantiomer and analytical data. |
We then explored the possibility of acylation reactions of these chiral reagents with a bulged ribonucleotide in an RNA/DNA hybrid structure (Fig. 5a), to test the possibility of site-specific conjugation via induced loop formation.4 The results showed conversion yields at 1 h of 83% (R) and 5% (S) for the two enantiomers, yielding diastereoselectivity of >16:
1 (Fig. 5b). We reacted the products with a commercial strained alkyne-functionalized Cy5 dye to prepare fluorescent-labeled adducts. Analysis by gel electrophoresis documents the much higher yield of the preferred enantiomer as seen by its strong labeling intensity (Fig. 5c), whereas the less-preferred enantiomer yields little visible fluorescence. This clearly demonstrates that chirality can have large influences in high-yield RNA labeling.
While most applications of modified RNAs involve strands of natural chirality (composed of D-ribose phosphodiester backbone), recent studies have investigated the properties and applications of enantiomeric (“mirror image” or “spiegelmer”) RNAs, with L-ribose as the source of chirality.42–47 Given the favorable properties of enantio-RNAs in applications such as aptameric ligands in biological settings, we investigated a further test of chiral selectivity in conjugation and labeling of enantiomeric RNA. We carried out reactions of D- and L-enantiomers of a short RNA having a reactive bulge loop structure (sequence B4A from the above library) using each enantiomer of the dimethylalanine reagent 2. MS data (Fig. 5d) documented conversions at 1 h of ∼90% for each “matched” pair of reagents with RNA ((R) with D-RNA; (S) with L-RNA), and considerably lower 5–8% conversions for the “mismatched” pairs. Diastereoselectivity was over 100:
1 at early time points and averaged 24
:
1 at ∼70% conversion. Thus, highly enantioselective reaction was seen for both native and mirror-image RNAs using reagents of appropriately matched handedness. Extending this to a functionalizable reagent, we reacted the enantiomers of azidoethyl-2 with the D- and L-RNAs, resulting in preparative conversions of 85% and diastereoselectivities averaging 8
:
1 (Fig. 5e). The experiments confirm that both antipodes of the synthesized azidoethyl-alanine acylimidazole reagent do indeed function normally in acylation, allowing us to rule out degradation or impurity in one enantiomer leading to misinterpretation of selectivity. Taken together, the new findings enable selective conjugation of natural RNA or its mirror image, which was previously not possible.
It is clear from our initial screening data that reagent structure can have large effects on selectivity, as most of the reagents tested here show modest stereoselectivity. However, for selected scaffolds, we observe near-complete diastereoselectivity at low conversions at specific sites in folded RNA, and up to 24:
1 diastereoselectivity in preparative reactions. These findings add a new physicochemical dimension to the study of RNA recognition and reaction. Our findings confirm chirality as a significant and overlooked structural mechanism for selectivity control in covalent RNA recognition and conjugation.
Our new results add the dimension of asymmetry as an important influencing factor in reagent design, going beyond the effects of alkyl/aryl substitution and steric bulk.12,39,40 For example, our experiments have identified a chiral acylating species, (S)-1, that exhibits exceptionally high loop/stem selectivity relative to its enantiomer, and indeed considerably higher than that of any reported agent to date. No previous chiral compound has been employed in mapping of folded RNA structure. In our preliminary tests, this resulted in markedly improved signal over background in intracellular RNA mapping. Further exploration of this finding seems warranted, with the aim of developing future high-performing generations of transcriptome-probing reagents.
In addition, given the rapid development of RNAs as therapeutic agents, establishing methods to locally substitute RNA for conjugation is also of substantial interest;5,45,48 however, it is significantly challenging, due to the large number of chemically similar groups in the biopolymer. In this light, our observation of the highly selective stoichiometric labeling reaction of (R)-azidoethyl-2 seems promising (Fig. 4b and e). Taken together, the current studies establish the use of reagent chirality coupled with local RNA macro-chirality as a potentially promising strategy for enhancing local selectivity in RNA reaction; more work will be needed to establish ideal sequences and to test applications in longer biologically relevant RNAs.
It is interesting and noteworthy that α-amino compounds in this study (e.g.2, 8) exhibit the opposite diastereoselectivity for RNA, for the most part, as the structurally analogous α-alkoxy compounds (e.g.1, 10). This suggests that the two classes of compounds adopt distinct conformations that favor approach of 2′-OH to the opposite faces of the reactive carbonyls. As a possible explanation (Fig. S16†), we hypothesize that steric effects of the larger dialkylamino group relative to the alkoxy group play a role in this difference. More studies will be needed to shed light on this divergence, and a better understanding could possibly lead to future reagent designs with altered or improved selectivity.
Also remarkable and unexpected is our finding of exceptionally high reactivity by compounds (R,S)-2, and the (R) enantiomer in particular, in acylation of RNA. Indeed, (R)-2 is the most highly RNA-reactive compound yet described in the literature to our knowledge. An experiment benchmarking the reactivity of enantiomers of 2 in comparison to the most commonly applied previous acylation scaffolds employed in RNA research (1M7 and NAIN3) underscores this marked difference (Fig. S17†). It seems likely that relative to non-basic alkoxy scaffolds, the protonation of the amine group in these N,N-dimethylaminoalanine derivatives at the pH of our reactions plays a role, by inductively activating the carbonyl carbon to greater electrophilicity, as supported by the high reactivity of 2 relative to non-protonated compound 1. Similarly, proline derivative 8 also shows higher reactivity (by a factor of >50-fold) than structurally related tetrahydrofuran scaffold 10. Comparisons of reactivity of (S)-2 and dimethylglycine (DMG) acylimidazoles (which differ by only a methyl group) reveal similar RNA reactivity and half-lives for hydrolysis (Fig. S3†), indicating that the methyl substitution in (S) has little effect on carbonyl reactivity with water in this isomer. However, the (R) enantiomer of 2 shows higher reactivity for RNA than the (S) isomer (by a factor of roughly ten), and acylates the biopolymer with >90% conversions even at the remarkably low concentration of 100 μM, one thousand-fold lower than common concentrations of RNA acylating agents. This suggests that the methyl group substitution in the (R) configuration confers specially elevated reactivity with RNA.
In an effort to better understand the high reactivity of (R)-2, we considered whether the compound might bind RNA noncovalently. We followed up by performing a titration of concentration vs. conversion, and by testing the effects of competition by a chiral nonreactive analogue of (R)-2 (Fig. S18 and S19†). The competitor did modestly reduce RNA yields at high concentrations, suggestive of binding, but had similar effects on the less reactive achiral compound DMG as well (Fig. S19†). Given the similar reactivity of (R)-2 in multiple loop types (Fig. 3b), any RNA binding by this compound appears to be nonselective. More work is needed to fully understand the origin of this favorable chiral methyl effect, although the special effects of methyl substitution conferring potency have been noted widely for protein-targeted drugs.49
Taken together, these data establish new ways to achieve selectivity in RNA conjugation reactions, by taking advantage of stereoselective recognition between asymmetric small-molecule reagents and the local chiral environment of RNA 2′-OH groups. The chiral reagent scaffolds studied here are in several cases commercially available as carboxylic acids and are easily activated to acylimidazole form in dry DMSO in one step without purification to yield stock solutions. Thus our findings, and extensions to new chiral molecular scaffolds, should be readily accessible to other researchers. We envision multiple outcomes that may result in the long term. For example, it may prove possible to develop chiral probes that are specific for certain folded RNA motifs. Second, further study may lead to the identification of folded structural contexts that may be functionalized selectively over more common motifs (such as duplex regions) in RNAs, enabling improved site selectivity in localized conjugation and labeling of larger RNAs. Future work will be required to explore these possibilities.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3sc03067a |
‡ These authors contributed equally. |
This journal is © The Royal Society of Chemistry 2023 |