Amino-acyl tXNA as inhibitors or amino acid donors in peptide synthesis

Abstract

Xenobiotic nucleic acids (XNAs) offer tremendous potential for synthetic biology, biotechnology, and molecular medicine but their ability to mimic nucleic acids still needs to be explored. Here, to study the ability of XNA oligonucleotides to mimic tRNA, we synthesized three L-Ala-tXNAs analogs. These molecules were used in a non-ribosomal peptide synthesis involving a bacterial Fem transferase. We compared the ability of this enzyme to use amino-acyl tXNAs containing 1′,5′-anhydrohexitol (HNA), 2′-fluoro ribose (2′F-RNA) and 2′-fluoro arabinose. L-Ala-tXNA containing HNA or 2′F-RNA were substrates of the Fem enzyme. The synthesis of peptidyl-XNA and the resolution of their structures in complex with the enzyme show the impact of the XNA on protein binding. For the first time we describe functional tXNA in an in vitro assay. These results invite to test tXNA also as substitute for tRNA in translation.

INTRODUCTION

Although chemically modified nucleic acids have long been studied, xenobiotic nucleic acid (XNA) methodology has only recently been introduced, offering new applications in synthetic biology (1). XNAs are chemically modified nucleic acids, of which most have modified carbohydrate moieties and for which the chemistry can be tuned depending on the biological application. In most cases, the chemical modification increases nuclease resistance (2) and improves duplex stability of the XNA (3). In the field of synthetic biology, XNAs have been mainly tested as substitutes for DNA to store and transfer information (4), as aptamers (5,6), and to substitute RNA in biocatalysis (7).

Transfer RNAs are central compounds for the decoding of RNA and bacterial cell-wall synthesis. Here, we investigated the synthesis and function of a tXNA (transfer-XNA) in the context of a tRNA-dependent Fem transferase reaction. The use of tXNA substrates in the reaction represents a new step in the development of synthetic biology based on XNA.

The hexitol nucleic acid (HNA) system provides a noteworthy example of modified oligonucleotide analogs with improved characteristics. HNAs contain the six-membered sugar 1′,5′-anhydro-d-arabino-hexitol instead of the furanose moiety (Figure 1) (2). The 1′,5′-anhydrohexitol nucleoside mimics the natural furanose nucleoside in its 3′-endo conformation, resulting in the axial orientation of the base moiety (8). This feature makes it possible to conformationally pre-organize the single strand HNA oligonucleotide in an A-form, which has been shown to form stable self-complementary duplexes, as well as sequence-selective stable hetero-duplexes with natural DNA and RNA. The A-form character of HNA has been confirmed by NMR studies of a HNA oligomer bound to a complementary RNA oligomer (9) and by X-ray crystallography (10,11). This observation was recently confirmed by studies of a complex between HNA-modified DNA oligonucleotides and an archaeal DNA-polymerase obtained by directed evolution (12). Thus, HNA appears to be a mimic of RNA but has never been used as a functional RNA involving protein catalysis. In the present study, we aimed to evaluate its suitability for the construction of a tXNA.

Figure 1.

Nucleic acids and xeno-nucleic acids used in this study. RNA: ribonucleic acids (RNA), 2′-deoxy-2′-fluoro ribonucleic acids (2′F-RNA), 1′,5′-anhydrohexitol nucleic acids (HNA), 2′-deoxy-2′-fluoro-d-arabinonucleic acid (2′F-ANA), deoxyribonucleic acids (DNA).

Other XNAs have been designed by replacing the 2′ OH group of RNA or the 2′ H atom of DNA with a fluorine atom (13). The small size of the fluorine substitution leads to minimal steric perturbation in duplex structure formation (14–19). The fluorine can be positioned down or up to the 5-membered cycle (Figure 1), leading to 2′-deoxy-2′-fluoro-ribose (2′F-RNA) as an RNA mimic or 2′-deoxy-2′-fluoro-arabinose (2′F-ANA) as a DNA mimic. The introduction of 2′-Fluoro-RNA residues is known to stabilize duplexes by enhancing the hydrogen bond strength in Watson-Crick type interactions and promoting stacking interactions between bases (20). The high similarity in geometry to RNA (21), the lack of an immunostimulatory effect of fluoro-RNA residues (21), and the increased stability against RNAse (22) make fluoro-RNA an interesting class of XNAs for the design of new bioactive molecules.

In this study, we explored the impact of the XNA modification in aminoacyl-tRNA (aa-tRNA) analogs using HNA or fluoro-RNAs. Aa-tRNAs are substrates of the ribosomal machinery in the synthesis of proteins and are also used in non-ribosomal protein synthesis (NRPS) such as the synthesis of the bacterial cell wall of gram-positive bacteria by Fem transferases (Supplementary Scheme 1) (23), protein degradation by L/F transferases (24), and the modification of bacterial membranes by phosphatidylglycerol synthase (25). Aa-tRNAs are also involved in the synthesis of secondary metabolites, such as cyclodipeptides (26) and natural antibiotics (27).

In our study, we use the Fem transferase enzyme as a model of non-ribosomal peptide synthesis to explore the potential of tXNA analogs. Fem transferases use aa-tRNAs to form branched peptides of the bacterial cell wall peptidoglycan synthesis (see supplementary section 1 and Scheme S1 for details).

We focused on the impact of the incorporation of XNA into the acceptor arm of Ala-tRNA^Ala, the substrate of the transferase FemX (Figure 2A). In previous studies, we synthesized analogs of L-Ala-tRNA^Ala as substrates of FemX (28) following a strategy described by the group of Hecht (29), which involves the synthesis of aminoacyl-dinucleotides, an enzymatic ligation, and a final deprotection step of the alanine residue (Figure 2B and C). More recently, we synthesized peptidyl-RNA conjugates to obtain high-affinity inhibitors. These peptidyl-RNA conjugates are bi-substrate type analogs in which a peptidoglycan fragment and a RNA moiety are covalently linked by a 1,4-triazole ring. The key chemical step to obtain these molecules is a copper-catalyzed alkyne-azide cycloaddition (CuAAC) between a 2′-azido oligonucleotide (24-nt RNA) and an alkyne-UDP-MurNac-pentapeptide precursor (Figure 3A). These peptidyl-RNA conjugates have shown high inhibitory activity when tested on FemX from W. viridescens (30) or FmhB from S. aureus (31). Modification of these bi-substrate analogs containing a shorter oligonucleotide (8-nt RNA, Figure 3B) allowed us to solve the first crystallographic structure of a FemX-bi-substrate complex and to identify the protein residues involved in RNA binding and enzymatic transfer (32). Here, we investigated which ribose modifications are tolerated by the NRPS process catalyzed by FemX and we studied the structural features involved in the recognition of XNA fragments by the enzyme.

Figure 2.

(A) Schematic representation of the reaction catalyzed by the alanyl synthetase (AlaRS) which incorporate the L-Ala onto the tRNA^Ala and the reaction catalyzed by FemX which transfer the L-Alanyl residue from tRNA^Ala to UDP-MurNAc-pentapeptide (UM5K). (B) Schematic representation of the hemi-synthesis of L-Ala-Helix^Ala used in this study (C) Structure of tRNA^Ala and the 24- and 8-nt helices that mimic the acceptor arm of tRNA^Ala. Acceptor arm positions are highlighted in gray.

Figure 3.

(A) Chemical strategy for the synthesis of the 24-nt peptidyl-oligonucleotide bi-substrate analog. (B) Chemical strategy for the synthesis of the 8-nt peptidyl-oligonucleotide bi-substrate analogs. (C) Helices mimicking the acceptor arm of tRNA^Ala used in this study. (D) 8-nt helices mimicking the terminal part of the acceptor arm of tRNA^Ala. A, G, C and U are RNA (black). A^F, G^F, C^F and U^F are 2′F-RNA (red). A^FA, G^FA, C^FA and U^FA are 2′F-ANA (orange). A^H, G^H, C^H and U^H are HNA (blue). A^D, G^D, C^D and T^D are DNA (green).

MATERIALS AND METHODS

Enzyme production and purification

FemX, alanyl-tRNA synthetase, T4 RNA ligase, MurC, MurD, MurE and MurF were produced and purified as previously described (31,32). Protein concentration was determined by the Bradford assay (Biorad) using BSA as a standard.

tRNA production and purification

tRNA^Ala(5′-GGGGCCUUAGCUCAGCUGGGAGAGCGCCUGCUUUGCACGCAGGAGGUCAGCGGUUCGAUCCGCUAGGCUCCACCA-3′), corresponding to 76 nt of tRNA^Ala of E. faecalis strain V583, was obtained by in vitro transcription using T7 RNA polymerase (35).

Purification of nucleotide precursors from bacterial extracts

UM5K 5 was purified from vancomycin-susceptible S. aureus strain RN4220 as previously described with minor modifications (32) (Supplementary information).

Ligation of N-pentenoyl-l-alanyl-dinucleotide (pdCpA-2′-l-Ala-pentenoyl) to 22-nt helices and iodine mediated l-alanyl deprotection

Modified dinucleotide pdCpA-l-Ala-pentenoyl was synthesized, ligated to helices 6a to 11a using purified T4 RNA ligase, and deprotected with iodine (36,37) (for details, see supplementary information section 7).

FemX activity in the presence of Ala-tXNAs

The efficiency of the l-alanyl transfer catalyzed by FemX using our tXNA analogs (containing RNA, 2′F-RNA, HNA, 2′F-ANA or DNA nucleotides) was determined at 21°C in 20 μl of 50 mM ammonium acetate (pH 7.2). After 0, 15, 30, 60, 120, 300 and 1100 min, an aliquot of each reaction mixture was quenched by the addition of an acetic acid solution (0.1%) containing an internal standard of the reaction product, [¹³C-¹⁵N]-l-Ala-UDP-MurNAc-pentapeptide for quantification of the transfer. The velocity of the transfer of the L-alanyl residue from 6b, 7b, 8b, 9b, 10b and 11b to UDP-MurNAc-pentapeptide was calculated using the linear portion of the curves between 0 and 30 min (Insets in Figure 4). Transfer efficiencies (TE) are expressed in min⁻¹ and were calculated by dividing the velocity of each reaction by the final product concentration determined at 1100 min (Supplementary information section 8) to avoid bias arising from differences in substrate concentrations from the steps of ligation, deprotection, and precipitation of the amino-acylated helices.

Figure 4.

(A) Semi-synthesis of l-Ala-tRNA^Ala analogs. Kinetics of the transfer of l-Ala from l-Ala-helices and quantification by mass spectroscopy (Supplementary information). Transfer from RNA (B) 2′F-RNA (C), HNA (D and E), 2′F-ANA (F), or DNA (G). (H) Transfer efficacy of L-Ala from helices 6b-11b to UM5K by FemX. ND: not detected.

Synthesis of the alkyne-containing UDP-MurNAc-peptide analogue

Semi-synthesis of the UDP-MurNAc-peptide analogue 4 was performed as previously described (30) (See Supplementary information section 7 and scheme S2 for details).

Synthesis of 24-nt tXNA bi-substrate inhibitors

The convergent synthesis of XNA-containing bi-substrates 17, 18, 19, 20 and 21 (Figure 3A) was based on a CuAAC reaction between alkyne UDP-MurNAc-pentapeptide 4 and 2′-azido XNA-containing oligonucleotide helices 12, 13, 14, 15 and 16. The 2′-Azido-dinucleotide pdCpA-2′-N₃ was synthesized and ligated to helices 7a, 8a, 9a, 10a and 11a using purified T4 RNA ligase (30), with minor modifications (Supplementary information). The resulting 2′-azido helices 12, 13, 14, 15 and 16 were purified by anionic-HPLC and analyzed by HPLC, mass spectrometry (Supplementary information), and denaturing polyacrylamide gel electrophoresis (Supplementary Figure S1 in Supplementary information). Alkyne-containing UDP-MurNAc-pentapeptide 4 was obtained by semi-synthesis using Mur enzymes from E. coli (30). The expected bi-substrates 17, 18, 19, 20 and 21 were purified and analyzed by anionic-HPLC and mass spectrometry (Supplementary information).

Synthesis of 8-nt tXNA bi-substrate inhibitors

The 2′-azido-adenosine (A⁷⁶) at the 3′-extremity of each 8-nt oligonucleotide was introduced by solid-phase synthesis. The 2′-azido-adenosine moiety was used to react with the alkyne containing UDP-MurNAc-pentapeptide 4, affording compounds 26 (2′F-RNA), 27 (HNA), 28 (2′F-ANA) and 29 (DNA) (Figure 6A). Production of the resin substituted by the 2′-azido nucleotide, the coupling reactions, and the final deprotection step were performed as previously described (31). Compounds 26, 27, 28 and 29 were purified and analyzed by mass spectrometry (Supplementary information section 11) and anionic-HPLC (Supplementary information section 12).

Figure 6.

Inhibition of FemX by 8-nt peptidyl-oligonucleotide conjugates. (A) General structure of the 8-nt peptidyl-oligonucleotides (in gray: RNA or XNA). (B) Inhibition curves of compounds 26 (2′F-RNA), 27 (HNA), 28 (2′F-ANA) and 29 (DNA).

Ki determination

Inhibition of FemX was tested in a radioactive coupled assay involving [¹⁴C]-l-Ala acylation of tRNA^Ala by alanyl tRNA synthetase (AlaRS) and transfer of [¹⁴C]-l-Ala from the resulting [¹⁴C]-l-Ala-tRNA^Ala to UDP-MurNAc-l-Ala-γ-d-Glu-l-Lys-d-Ala-d-Ala (UM5K) to form UDP-MurNAc-l-Ala-γ-d-Glu-l-Lys([¹⁴C]-l-Ala)-d-Ala-d-Ala ([¹⁴C]-l-Ala-UM5K) (Supplementary information section 8) (32).

FemX crystallization, data collection, structure determination and refinement

FemX (255 μM) in complex with peptidyl-XNA 26 to 29 (306 μM) was co-crystallized by sitting-drop vapor diffusion (Supplementary information). The structures of the complexes were solved by the molecular replacement method at 1.77 Å (HNA), 1.80 Å (2′F-RNA), 1.34 Å (2′F-ANA) and 1.49 Å (DNA) resolution. All complexes include the 337 residues of the protein (see Supplementary information and Supplementary Table S1 for crystal structure determination, data collection and refinement statistics).

RESULTS AND DISCUSSION

Chemical synthesis of Ala-tXNA^Ala as an analog of Ala-tRNA^Ala

The study of the impact of XNAs on biological systems has been hampered by the difficulties of their synthesis. Although great progress has been made in recent years to obtain XNA-specific polymerases by directed evolution, the enzymatic synthesis of XNAs in significant quantities is still challenging (33,34). Solid-phase synthesis techniques have improved the synthesis of XNA but do not allow the synthesis of large oligonucleotides. Here, we show that an enzymatic ligation catalyzed by T4 RNA ligase between an l-Ala-acylated dinucleotide and an XNA can be used to obtain tXNA^Ala analogs. As FemX interacts mainly with the acceptor arm of tRNA^Ala, we restricted the size of our XNA probes to 24-nt aminoacylated helices as mimics of the tRNA^Ala (35). For this study, we synthesized six helices of 24-nt mimicking the acceptor arm of tRNA^Ala (Figure 3C) containing RNA, 2′F-RNA, HNA, 2′F-ANA or DNA, nucleotides in the double stranded portion of the helices (compounds 6a, 7a, 8a, 10a and 11a) or HNA in the full helix (compound 9a). The RNA and DNA helices (6a and 11a, respectively) were synthesized by the phosphoramidite method and purified by anion-exchange chromatography (DNAPac). The XNA helices containing 2′F-RNA, HNA, and 2′F-ANA nucleotides (7a, 8a, 9a, 10a) were obtained by solid phase synthesis on an Expedite® 8909 DNA synthesizer (Supplementary information). The acylated dinucleotide (pdCpA-L-alanyl-pentenoyl, Figure 2B) was synthesized (36) and ligated to helices 6a to 11a using purified T4 RNA ligase and precipitated using cold ethanol (37). Iodine-mediated L-alanyl deprotection was performed to obtain substrate analogues 6b to 11b (29) (Figure 2B).

Transfer efficiency of the amino-acid from Ala-tXNAs

Compounds 6b to 11b were used as substrates in an in vitro assay to compare the transfer efficiency (TE) of the alanyl residue catalyzed by the enzyme (Figure 4A). We observed similar alanyl transfer in a time-dependent manner with compound 6b containing natural RNA residues, compound 7b (2′F-RNA) and compound 8b (HNA) (TE = 0.015 ± 0.007, 0.015 ± 0.003 and 0.016 ± 0.003 min⁻¹, respectively, Figure 4B–D). The introduction of HNA nucleotides in both the double strand and loop, compound 9b (Figure 4E) had also no effect on the efficiency of the alanyl transfer compared to the natural RNA (TE = 0.018 ± 0.002 min⁻¹). On the contrary, alanyl transfer from compounds 10b (2′F-ANA) and 11b (DNA) was abolished, as we did not detect any product of the reaction by mass spectrometry, even after 1,100 min (Figure 4F and G). These results (Figure 4H) indicates that helices containing HNA or 2′F-RNA are substrates of FemX and 2′F-ANA or DNA in the double strand are unable to sustain the activity of Fem transferase, even if the single-strand terminal part (ACdCA) of the substrates is unmodified.

Synthesis and biological evaluation of helix-based tXNA bi-substrates

Our next objective was to synthesize peptidyl–XNA conjugates as FemX bi-substrate analogs and measure the impact of XNA incorporation in peptidyl–RNA conjugates on the binding of FemX. The peptidyl-XNAs (17 to 21) were synthesized in two steps starting from 2′-azido-2′-deoxy-dinucleotides (pdCpA-N₃) (30). First, the 2′-azido-dinucleotide was ligated to helices 7a to 11a to afford azido-tXNA 12 to 16 (Figure 3A). The second step used the CuAAC between the azido-tXNA and the alkyne-UDP-MurNAC (compound 4) to obtain peptidyl-XNAs 17 to 21 (Figure 3A). We have previously shown that the peptidyl UDP-MurNAc-peptide (UM5K) and the RNA helix mimicking the acceptor arm of the tRNA^Ala both contribute to high-affinity binding to FemX (30). Inhibition experiments were performed using a coupled enzymatic assay that relies on the aminoacylation of tRNA^Ala by the alanyl-tRNA synthetase (AlaRS) and the transfer of the L-alanyl residue from the resulting Ala-tRNA^Ala to UM5K by FemX (Figure 5A) (38,39). Compound 17, which contains 2′F-RNA nucleotides in the double stranded portion of the helix, inhibited FemX, with a K_i of 397 ± 107 pM (Figure 5B), which is in the same order of magnitude to the Ki observed for the inhibition of FemX by peptidyl-RNA (89 ± 10 pM) (30). The introduction of HNA residues in the double strand (compound 18, Figure 5C) also resulted in inhibition of the same order of magnitude, with a K_i of 133 ± 33 pM. The addition of HNA residues in the helix loop (compounds 19, Figure 5D) decreased the inhibition of FemX, with a K_i of 484 ± 156 pM. These results show that the introduction of HNA residues in the double strand of the helix did not impact inhibition but that the introduction of HNA in both the double strand and the loop of the helix slightly decreases the inhibition of FemX. Finally, compounds 20 and 21 (Figure 5E and F), containing 2′F-ANA and DNA nucleotides in the double strand of the helix, showed a 2,500-fold (K_i = 221 ± 10 nM) and 300-fold (K_i = 26 ± 1 nM) reduction of inhibition, respectively. These results unambiguously show that the introduction of 2′F-ANA or DNA nucleotides has a deleterious effect on the formation of a stable complex between FemX and the peptidyl-XNA inhibitors.

Figure 5.

(A) Inhibition of FemX by peptidyl-oligonucleotide conjugates containing 2′F-RNA residues in the helix stem (B), HNA residues in the helix stem (C), HNA residues in the helix stem and loop (D), 2′F-ANA residues in the helix stem (E) and DNA residues in the helix stem (F). K_i values were calculated by fitting the Morrison equation to the data.

Synthesis and biological evaluation of minimal peptidyl-tXNA bi-substrate inhibitors for co-crystallization

To study the impact of XNA in the 3′-terminal ACCA part of tRNA^Ala we synthesized minimal bi-substrate analogs fully modified by XNA nucleotides (compounds 26 to 29, Figure 3B). To mimic the terminal part of the acceptor arm of tRNA^Ala we chose 8-nt XNA oligonucleotides (Figure 3D), as similar compounds were successful to obtain the crystallographic structure of Fem transferase enzymes in the presence of peptidyl–RNA (32). The synthesis of the 8-nt helices (compounds 22 to 25) was performed by SPS, starting from the 2′-azido 2′-deoxyadenosine attached to a solid support (Figure 3B). A polyethylene glycol linker was incorporated into each oligonucleotide to stabilize the base pairing necessary to conserve good recognition with the enzyme (32) (Figure 3D). Finally, azido-oligonucleotides 22 to 25 were reacted with alkyne containing UDP-MurNAc-pentapeptide 4 through a CuAAC reaction, affording compounds 26 (2′F-RNA), 27 (HNA), 28 (2′F-ANA) and 29 (DNA) (Figure 6A). These short peptidyl-XNAs were tested as inhibitors of FemX (Figure 6B and Table 1). HNA-based molecule 27 inhibited FemX 19-fold less than the peptidyl-RNA analog (K_i = 17 ± 1 μM and 0.93 ± 0.07 μM, respectively). Compounds 28 (2′F-ANA) and 29 (DNA) inhibited FemX moderately less than the RNA analog: 2.6-fold (K_i = 2.4 ± 0.1 μM) and 8.7-fold (K_i = 8.0 ± 0.6 μM), respectively. Interestingly, 2′F-RNA based molecule 26 is the best inhibitor of FemX (K_i = 0.057 ± 0.002 μM) compared to 0.93 ± 0.07 μM obtained for the peptidyl-RNA. These results show that the XNA residues present in the 3′ terminal single-strand portion of the bi-substrate affect ligand binding in the catalytic site of the enzyme. We thus undertook to co-crystallize the four bi-substrate analogs 26 to 29 with the protein to investigate structural determinants that could explain these inhibition variations.

Table 1.

Calculated K_i values deduced by fitting the Morrison equation to the data. Gibbs free enthalpies ΔG° were calculated using the equation ΔG° = –RTln(K_i), with R = 1.987 cal mol⁻¹ K⁻¹ and T = 303.15 K

Compound composition	K i (μM)	ΔG° (kcal mol−1)	ΔΔG° (kcal mol−1)
RNA(a)	0.93 ± 0.07	−8.37	NA
2′F-RNA (26)	0.056 ± 0.002	−10.05	−1.68
HNA (27)	17 ± 1	−6.60	+1.77
2′F-ANA (28)	2.4 ± 0.1	−7.80	+0.57
DNA (29)	8.0 ± 0.6	−7.07	+1.30

(a) 8-nt peptidyl-RNA (32).

Structural studies of short peptidyl–tXNA bi-substrates in complex with FemX

We determined the crystal structure of FemX bound to compounds 26, 27, 28 and 29 (Figure 7). Complex crystals were grown under conditions similar to those used to obtain the complex with our first 8-nt peptidyl–RNA (PDB: 4II9).

Figure 7.

Structures of peptidyl–RNA and peptidyl–XNA bi-substrate analogs in complex with FemX. Protein in complex with bi-substrate containing RNA (A), HNA (B, compound 27), 2′F-RNA (C, compound 26), 2′F-ANA (D, compound 28) and DNA (E, compound 29).

Analysis reveals that in the four complexes, the enzyme structure is almost identical to that in the complex with a peptidyl–RNA. R.M.S.D. values between the RNA-based complex and the complexes with compound 26 (2′F-RNA), 27 (HNA), 28 (2′F-ANA) and 29 (DNA) were 0.204, 0.131, 0.144 and 0.116 Å, respectively (including all residues, Supplementary Figure S2). The only minor conformational change observed consisted of a 1.8 Å shift of the Met²⁶⁰–Thr²⁶⁴ loop (Supplementary Figure S2) for the two complexes containing the bi-substrates 28 and 29 containing 2′F-ANA and DNA nucleic acids, respectively. This shift could be explained by the loss of the interaction between this loop and the C⁷⁵ and A⁷⁶ nucleotides, which are not correctly positioned to interact with the protein.

The UDP–MurNAc-pentapeptides and the triazole rings in the four complexes were well defined and positioned as in the peptidyl-RNA complex, except for the peptidyl-2′F-RNA 26, for which the D-iGlu² is repositioned (Supplementary Figure S3).

In all four structures, the N₃ of the triazole ring is in the vicinity of the catalytic residue Lys³⁰⁵, which has been shown to directly participate in the chemical step of the transfer reaction (32).

In the peptidyl–RNA complex, only the electron density of the C⁷⁴C⁷⁵A⁷⁶ extremity of the oligonucleotides was well defined (Figure 7A). We were able to observe an additional nucleotide (A⁷³) for the HNA, 2′F-ANA and DNA conjugate complexes (Figure 7B–E). Even more interestingly, for the 2′F-RNA:FemX complex, the electron density was sufficiently well defined for positioning all nucleotides (Figure 7C), making it possible to observe for the first time a full peptidyl-helix in complex with a Fem transferase enzyme.

These four complexes allow us to compare the structural features of the different tXNA analogs in interaction with a protein.

For the peptidyl-2′F–RNA (26) and peptidyl-HNA (27), the nucleotides show productive π-stacking interactions between both A⁷³ and C⁷⁴ bases and between C⁷⁴ and C⁷⁵ bases (Supplementary Figure S4A). For the 2′F–RNA complex, we also observe the Watson–Crick base pairing of G¹–C⁷² as well as G²–C⁷¹ (Supplementary Figure S4A). We further observed enzyme residues in the vicinity of the C⁷¹ and C⁷² bases that we had previously identified as major determinants of FemX specificity towards tRNA^Ala (Supplementary Figure S4B) (35). For the peptidyl–DNA (29) and peptidyl–2′F–ANA (28), C⁷⁵ is not well positioned to be in a π-stacking interaction with C⁷⁴.

We also noted that the positioning of nucleotide A⁷⁶ is different, depending on whether the bi-substrate is an RNA or DNA analogue (Figure 8). The A⁷⁶ adenine is rotated by 159° with 2′F-ANA (28) and DNA (29) and thus prevents this base from interacting with Ile²⁰⁸ and Leu³⁰¹ (Figure 8C and D), as observed for the RNA (PDB 4II9), 2′F-RNA (26), and HNA (27) complexes (Figure 8A to B). We previously showed that hydrophobic interactions between the A⁷⁶ terminal adenine and Ile²⁰⁸ side chain are needed for efficient activity of the Fem enzyme, as substitution of Ile²⁰⁸ by an alanine residue led to a 50-fold decrease in turnover number (32). The mispositioning observed for the A⁷⁶ adenine, which carries the l-alanine in the aminoacyl tRNA^Ala, might explain why compounds containing 2′F-ANA (10b) and DNA (11b) are not substrates of FemX (Figure 4) and why the corresponding compounds 20, 21, 28 and 29 are not inhibitors of the enzyme (Figures 5 and 6).

Figure 8.

Superposition of the terminal nucleotides C⁷⁵ and A⁷⁶ of the peptidyl-RNA bi–substrate³² (gray) with the bi-substrates analogues 26 to 29 containing (A) 2′F-RNA (red), (B) HNA (blue), (C) 2′F-ANA (orange) and (D) DNA (green), respectively. Red arrows represent rotation or movement of the terminal C⁷⁵ and A⁷⁶ bases.

Concerning the 1′,5′-anhydrohexitol moiety in compound 27, the 1.77 Å resolution of the structure allowed us to define, for the first time, the conformation of the HNA sugar and bases in a catalytic site (Supplementary Figure S5). A single structure of a complex between HNA containing oligonucleotide and an archaeal polymerase enzyme has been described but the conformation of the base and the sugar moieties could not be modeled because of insufficient resolution (12). In our case, the HNA sugar moiety could be perfectly superposed with the ribose of the RNA, confirming that the 1′,5′-anhydrohexitol of HNA mimics the ribose in its C3′-endo conformation (Figure 8B).

We also explored the interactions of the penultimate C⁷⁵ sugar unit with the Phe³⁰⁴ residue conserved in all Fem transferases. This residue was shown to be important for activity, as its substitution by Ala or Leu decreased the FemX turnover number by 16- and 29-fold, respectively (32,39). The distance between 2′F and the phenyl group of Phe³⁰⁴ is 3.2 Å in the peptidyl-2′F-RNA compound 26 (Figure 8A), which corresponds to the sum of the Van der Waals radii of the two atoms (3.17Å). The fluorine in compound 26 is at a distance from Phe³⁰⁴ that can lead to a lone-pair π stacking interaction (lp–π) (40,41), which was calculated to result in a −1.5 kcal mol⁻¹ energy gain (42). This value is compatible with the free energy difference of binding of the RNA and 2′F-RNA compounds, as calculated from their inhibition constants: K_i^RNA = 0.93 μM (ΔG° = −8.37 kcal mol⁻¹) and K_i2′^F-RNA = 0.056 μM (ΔG° = −10.05 kcal mol⁻¹); ΔΔG° = −1.68 kcal mol⁻¹. For the HNA and DNA molecules, the distances between the C2′ position of the C⁷⁵ sugar and the Phe³⁰⁴ are 4.7 and 3.6 Å, respectively (Figure 8B and D), and the absence of an electron-rich atom (O or F) on the sugar excludes an lp-π interaction. For compound 28 (2′F-ANA), the distance between the 2′F and Phe³⁰⁴ is 4.1 Å (Figure 8C), resulting in a less productive lp–π interaction between the fluorine and phenyl side chain of Phe³⁰⁴. These observations could explain the loss of inhibitory potency of bi-substrates 27, 28 and 29, which possess HNA, 2′F-ANA, and DNA at position C⁷⁵ relative to the peptidyl-RNA and the stronger inhibition for compound 26, which could be due to a stabilizing lp-π interaction between the fluorine pointing towards Phe³⁰⁴.

CONCLUSION

In this study, we designed the first tXNA molecules as substrates or inhibitors of an amino acyl tRNA-dependent enzyme. We show that tXNA can be synthesized using enzymatic ligation with T4 RNA ligase or by solid-phase synthesis. We were able to obtain Ala-tXNA^Ala and peptidyl-XNA analogs of different sizes containing XNA in various positions.

We used these tXNA molecules to study the impact of XNA nucleotides on the activity and inhibition of a bacterial Fem transferase enzyme (FemX), as a model of a non-ribosomal peptide synthesis process. We show that the impact of introducing XNA into the tRNA substrate depends on the nature of the XNA, both at the levels of the double-strand and single stranded parts of the tRNA^Ala. HNA and 2′F-RNA motifs did not significantly modify substrate binding or catalysis, whereas the presence of DNA or 2′F-ANA nucleotides abolished both catalysis and substrate binding.

In terms of modifications in the 3′-ACCA portion of the tRNA, we observed that the 2′F-RNA modification had a more positive impact on recognition by FemX than that with HNA nucleotides. The observed differences rely on the fact that, peptidyl-RNA, -HNA and −2′F-RNA are well positioned in the active site and form more-or-less productive lp-π stacking interactions with the Phe³⁰⁴ residue of the protein. The 2′F-ribose of the C⁷⁵ moiety appears to form a very strong interaction with Phe³⁰⁴, resulting in a 16-fold lower inhibition constant than the inhibitor-containing ribonucleotides (K_i = 0.056 μM versus 0.93 μM). Our structural data also show for the first time that the geometry of HNA nucleotides is virtually indistinguishable from that of ribonucleotides and that the 1′,5′-anhydrohexitol of HNA can mimic a ribose in its C3′-endo sugar puckering. In contrary, introduction of 2′F-ANA and DNA residues in the single strand induced a loss of one π-stacking interaction, leading to poor positioning of the ACCA terminal moiety.

In conclusion, we show here that tXNA molecules can be employed in a non-ribosomal peptidyl enzymatic process and this is the first example of the use of XNA in a functional in vitro assay involving protein catalysis. This result paves the way to test tXNA for RNA decoding, which could represent a major step forward in the development of an enzymatically stable translation system. In addition, the observation that tXNA can inhibit enzymes involved in bacterial cell-wall synthesis holds promise for the development of a new class of antibiotics and other tXNAs of therapeutic importance.

DATA AVAILABILITY

The atomic coordinates and structure factors for the reported crystal structures of FemX in complex with compounds 26, 27, 28 and 29 and the structural factors reported in this paper have been deposited in the Protein Data Bank, www.pdb.org (accession nos. 7Z6A, 7Z5Y, 7Z6K and 7Z5Z, respectively).

Notes

Present address: Matthieu Fonvielle, Institute for Integrative Biology of the Cell (I2BC), CEA, CNRS, Université Paris-Saclay, 91198 Gif-sur-Yvette, France.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

FUNDING

French Infrastructure for Integrated Structural Biology (FRISBI) [ANR-10-INBS-05]; Hercules Foundation of the Flemish Government [20100225-7]. Funding for open access charge: INSERM.

Conflict of interest statement. None declared.