Chemoenzymatic Synthesis of 3′-Deoxy-3′,4′-didehydro-cytidine triphosphate (ddhCTP)

Abstract

3′-Deoxy-3′,4′-didehydro-cytidine triphosphate (ddhCTP) is a novel antiviral molecule produced by the enzyme viperin during the early stages of the innate immune response. ddhCTP has been shown to act as a chain terminator of flavivirus RNA-dependent RNA polymerases. To date, synthesis of ddhCTP requires complicated synthetic protocols or isolation of the enzyme viperin to catalyze the production of ddhCTP from CTP. Recombinant viperin approaches preclude the production of highly pure ddhCTP (free of contaminants such as CTP), whereas the chemical synthesis involves techniques or equipment not readily available to most laboratories. Herein, we describe the chemoenzymatic synthesis of ddhCTP, starting from commercially available ddhC. We utilize these methods to produce milligram quantities of ddhCTP, ddhCDP, and ddhCMP. Using purified semisynthetic ddhCTP and fully synthetic ddhCTP, we also show ddhCTP does not inhibit NAD⁺-dependent enzymes such as glyceraldehyde 3-phosphate dehydrogenase, malate dehydrogenase, or lactate dehydrogenase, contrary to a recent report.

Viperin (virus-inhibitory protein, endoplasmic reticulum-associated, interferon (IFN) inducible) is an interferon inducible protein that inhibits and/or is involved in the replication of a remarkable range of both RNA and DNA viruses.¹⁻³ The mechanisms by which this 42 kDa enzyme elicits such broad-spectrum activity have been uncertain since its discovery.⁴ Viperin has been proposed to interact or colocalize with six human proteins (farnesyl pyrophosphate synthase (FPPS), mitochondrial trifunctional protein (HadHB), interleuken-1 receptor-associated kinase 1 (IRAK1), TNF receptor associated factor 6 (TRAF6), and cytosolic iron–sulfur assembly component 1 (CIAO1)) and three viral proteins (Dengue fever virus nonstructural protein 3 (NS3), Hepatitis-C virus nonstructural protein 5A (NS5A,) and human cytomegalovirus viral mitochondria-localized inhibitor of apoptosis (vMIA)).^2,5−8 Though biochemical evidence for these interactions remains largely indirect, the roles of these putative viperin-interacting proteins are diverse and include metabolism, signaling, Fe/S cluster formation, and isoprenoid biosynthesis. Interestingly, in addition to antiviral activity, viperin has been implicated in other processes, such as regulation of thermogenesis in adipose tissue, dendritic cell maturation, chondrocyte development, type I interferon signaling, and macrophage polarization.⁹⁻¹²

Viperin is a member of the radical S-adenosylmethionine (RS) superfamily of enzymes. All members of this superfamily bind an oxygen-sensitive iron–sulfur (Fe/S) cluster that is indispensable for catalytic activity. Importantly, the genes encoding viperin and cytidylate monophosphate kinase 2 (CMPK2) are immediately adjacent to one another in all vertebrate genomes, are coregulated during the IFN response,¹³ and are expressed as a fusion protein in some organisms.³ These observations led to our discovery that viperin catalyzes the conversion of CTP to 3′-deoxy-3′,4′-didehydro-CTP (ddhCTP) through radical chemistry (Scheme 1). ddhCTP acts as a chain terminator in vitro against several flavivirus RNA-dependent RNA polymerases (RdRp), including those of Dengue (DNV), West Nile (WNV), and Zika viruses.³ Furthermore, we demonstrated that ddhCTP inhibited Zika virus replication in vivo.³ Our results are consistent with a subsequent report showing a ddhCTP prodrug inhibits the replication of Zika virus in green monkey Vero cells.¹⁴ The discovery of ddhCTP and its impact on flaviviral RdRps provided, for the first time, a mechanistic underpinning for viperin’s inhibition of viral replication. However, these studies do not provide a fully unified mechanism, as viperin acts as a restriction factor for picornavirus,¹⁵ although picornaviral RdRps are not sensitive to ddhCTP.³ These observations suggest that ddhCTP makes additional mechanistic contributions to viral restriction and cellular homeostasis.

Scheme 1. Proposed Reaction Catalyzed by Viperin.

Viperin abstracts the C4′ hydrogen (marked in red) using the 5′-deoxyadenosyl radical (common to all members of the radical SAM superfamily) to initiate catalysis. Elimination of water and reduction by one electron completes catalysis.

Indeed, two recent studies have expanded the role of ddhCTP in cellular processes. A recent paper by Hsu et al. demonstrated that doxycycline-induced viperin expression or treatment with the ddhCTP prodrug, ddhC resulted in global inhibition of protein translation.¹⁶ Importantly, both these treatments did not affect overall mRNA transcription or DNA synthesis, indicating a targeted effect on protein translation, which was shown to be elicited through enhanced ribosome collisions and phosphorylation of serine 51 of eukaryotic initiation factor 2α (eIF2α) by serine/threonine kinase, general control nondeprepressible 2 (GCN2)—a well-studied regulator of the integrated stress response during amino acid deprivation.¹⁶ How ddhCTP increases ribosome collisions is currently unclear. Second, a recent report by Ebrahimi et al. showed that crude ddhCTP produced by the viperin-like enzyme from the fungal species Thielavia terrestris (TtVip) was capable of inhibiting the NAD⁺-dependent enzymes human glyceraldehyde-3-phosphate dehydrogenase (GAPDH), human lactate dehydrogenase (LLDH), and human malate dehydrogenase (MDH).¹⁷ Based on computational modeling, the mechanism by which ddhCTP inhibits NAD⁺-dependent activities was predicted to involve direct engagement with the NAD⁺-binding pocket of these enzymes. The ability of different nucleotide binding domains to either bind or discriminate against ddhCTP requires further study. To this point, an elegant study showed that the presence of viperin negatively affected the expression of GFP when its gene was placed behind a T7 bacteriophage promotor, while YFP-CFP fusion protein expression was unaffected when its transcription was driven by a RNA polymerase II dependent CMV promotor.¹⁸ This study demonstrated that less sophisticated polymerases struggle to differentiate ddhCTP from CTP. Together, these studies exemplify the need for highly pure ddhCTP to study the mechanism of how ddhCTP functions against the many purported targets of viperin/ddhCTP.

Based on the need for substantial quantities of highly purified ddhCTP to perform in vitro and in vivo studies, and given the challenges associated with the chemical synthesis of nucleoside triphosphates, we developed a chemoenzymatic strategy to produce ddhCTP from commercially available ddhC at the milligram scale. We compared this semisynthetic ddhCTP to the recently described synthetic ddhCTP¹⁹ and found them to be identical. In addition, we tested the ability of both ddhCTP preparations to inhibit the previously reported NAD⁺-dependent enzymes. Contrary to the previous report, we found that ddhCTP does not inhibit these NAD⁺-dependent enzymes. Instead, we find that the apparent inhibition of NAD⁺-dependent enzymes is an artifact arising from the presence of contaminating dithionite in the viperin-prepared ddhCTP utilized in those studies. These observations underscore the need for highly pure ddhCTP to study its emerging roles in cellular metabolism.

Results and Discussion

Detailed in vitro and in vivo analyses of ddhCTP in viral restriction and cellular function require the synthesis and purification of large quantities of ddhC nucleoside phosphates. The facile, high-yielding chemical synthesis of ddhCTP has recently been reported;¹⁹ however, most biochemistry laboratories lack the expertise and/or infrastructure to carry out this synthetic strategy. Therefore, we aimed to develop a facile, enzymatic method to produce ddhCMP, ddhCDP, and ddhCTP from commercially available ddhC. Because ddhC can be converted to ddhCTP by HEK293T cells when supplied exogenously, we hypothesized that the endogenous kinases were capable of performing the required transformations (Scheme 2). We cloned the human genes for uridine/cytidine kinase 2 (UCK2), cytidine monophosphate kinase 1 (CMPK1), and nucleotide diphosphate kinase (NDK) into a pSGC-His construct, which allowed us to produce these proteins as N-terminal hexa-histidine tag fusions for facile Ni-NTA purification (see Supporting Information). Using these methods, >100 mg quantities of stable, soluble, and active enzymes can be produced (Figure S1).

Scheme 2. Proposed Sequential Enzymatic Phosphorylation of ddhC to ddhCMP, ddhCDP, and ddhCTP Using Recombinantly Expressed and Purified UCK2, CMPK1, and NDK, Respectively, Using ATP as the Phosphate Donor.

To test the ability of UCK2 to phosphorylate ddhC to ddhCMP using ATP as the phosphate donor, we employed a coupled phosphoenolpyruvate/lactate dehydrogenase assay to detect the formation of ADP via a continuous spectrophotometric assay (see Supporting Information).^20,21 Though the activity of UCK2 with ddhC is reduced by ∼130-fold relative to UCK2’s formation of CMP (Figure 1A), ddhCMP is produced in a stoichiometric rate with ADP formation, as confirmed by high performance liquid chromatography coupled to an Agilent 6490 triple quadrupole mass spectrometer (LCMS, Figure 1B). Knowing that UCK2 can catalyze phosphorylation of ddhC to ddhCMP in vitro, we tested whether a combination of UCK2, CMPK1, and NDK can phosphorylate ddhC to ddhCMP, ddhCDP, and ddhCTP. We mixed UCK2, CMPK1, and NDK each at 10 μM with 1 mM ddhC, 150 μM ATP, 3 mM PEP, and 1.2 U of pyruvate kinase (PK) and 1.8 U of l-lactate dehydrogenase (LLDH) per mL in a 50 mL reaction (see Methods in the Supporting Information) to produce milligram quantities of ddhCMP, ddhCDP, and ddhCTP. We followed the reaction by LCMS, and at 30 min, the reaction was quenched by addition of concentrated HCl to pH 2.0. At this point, the reaction is represented by an ∼1:1:1 stoichiometry between ddhCMP, ddhCDP, and ddhCTP. The reaction mixture was then neutralized by adding Tris base, centrifuged for 30 min at 5000g to remove the precipitated proteins, filtered through a 0.22 μm filter, and diluted to 500 mL (10× dilution) with 10 mM triethylammonium bicarbonate (TEAB) at pH 9.5. The mixture was loaded on a MonoQ 10/100 GL column previously equilibrated in the above buffer, and the column was washed with 8 column volumes of loading buffer, before applying a stepwise gradient from 0% to 30% 1 M TEAB pH 9.5 (Buffer B). The stepwise gradient was increased at 2.5% Buffer B increments and held at each concentration for 2 column volumes. As can be seen in Figure 1C, all three nucleotides can be readily separated using this method with ddhCMP eluting at 5% B, ddhCDP eluting at 12.5% B, and ddhCTP eluting at 27.5%–30% B. Using this method, we were able to purify 16 mg of each nucleotide to >98% purity based on LCMS (Figure 1D, Figures S2 and S3). With ddhCMP and ddhCDP in hand, we were able to demonstrate that, similar to UCK2, CMPK1 and NDK were ∼82-fold and >1000-fold less efficient at phosphorylating ddhCMP and ddhCDP than their unmodified cytidine nucleoside phosphate substrates (Figure 1E). Despite sharing identical pyrimidine bases and equivalent phosphorylation states, these results suggest that the planar configuration of the ddhC ribose sugar engenders native kinases with the ability to discriminate against this modification, likely similar to how native polymerases differentiate ddhCTP from other nucleotides.¹⁶ To generate milligram quantities of only ddhCTP, we used the same method as above, but increased the reaction time to 360 min. This protocol allows all of the ddhC to be consumed and converted to ddhCTP, resulting in a final yield of ∼78% starting from 10 mg of ddhC (see Supporting Information). Importantly, ddhCTP produced by these methods has identical LCMS properties to that of the previously reported synthetic ddhCTP (Figures S2 and S3).¹⁹

Figure 1.

Phosphorylation and purification of ddhC nucleoside phosphates. (A) Reaction of ddhC with UCK2. The activity of UCK2 with cytidine (C, left bar) versus its activity with ddhC (right bar), producing their respective monophosphates. The activity of UCK2 is reduced by ∼130-fold with ddhC as a substrate. (B) HPLC trace of UCK2 mixed with ddhC and ATP at 0 (green) and after 32 min (red) at 275 nm. Peak showed a m/z of ddhCMP of 304.9 Da as shown in Supporting Information Figure S2. (C) Separation of ddhCMP, ddhCDP, and ddhCTP using anion exchange chromatography from a reaction with UCK2, CMPK1, and NDK (blue trace, 260 nm). ddhC nucleotide peaks are labeled. (D) Purified ddhCDP (blue trace) and ddhCTP (orange trace) were characterized by LCMS (UV absorbance at 260 nm). HPLC analysis shows that nucleotides were >98% pure (Figure S3). Peaks produced a m/z of ddhCDP of 384.1 and ddhCTP of 464.1 as shown in Supporting Information Figure S2. (E) CMPK1 and NDK activity was tested with purified ddhCMP and ddhCDP. Similar to UCK2, the activities were reduced ∼82× and >1000× when ddhCMP and ddhCDP were used as substrates, respectively.

With pure ddhCTP in hand, we wanted to confirm the effect of ddhCTP on NAD⁺-dependent enzymes, as previously reported.¹⁷ Human glyceraldehyde-3-phosphate dehydrogenase (GAPDH), human lactate dehydrogenase (LLDH), and human malate dehydrogenase (MDH) and their respective substrates were purchased from Sigma-Aldrich and used as received. Ebrahimi et al. reported that ddhCTP produced by the TtVip-mediated conversion of CTP was capable of reducing the GAPDH-catalyzed conversion of G3P to 1,3BPG by ∼3-fold at a concentration of ∼180–200 μM ddhCTP, with an IC₅₀ of 55.8 ± 0.2 μM.¹⁷ In addition, they reported that the TtVip-generated ddhCTP was also capable of inhibiting the reactions of LLDH and MDH by ∼5-fold and 4-fold, respectively. Ebrahimi et al. used molecular docking simulations with the SwissDock server to predict the binding energy of ddhCTP to the crystal structures of GAPDH (PDB: 5C7O) and LLDH (PDB: 1T2F).¹⁷ These docking experiments were interpreted to predict direct binding of ddhCTP to the substrate pockets of GAPDH and LLDH, with calculated K_D values of 66 and 17 nM, respectively. Using our highly purified ddhCTP, we worked to define the putative mechanism by which ddhCTP inhibited these important NAD⁺-dependent enzymes. However, using reaction conditions similar to Ebrahimi et al.,¹⁷ we found that 100 μM ddhCTP did not inhibit either GAPDH, LLDH, or MDH at substrate concentrations between 0.1 and 1 mM (Figure S4A–C). In addition, we tested inhibition of these enzymes with higher concentrations of ddhCTP (Figure 2A). Using our purified ddhCTP, we saw no inhibition of GAPDH, LLDH, and MDH even with 1.0 mM ddhCTP. Similar results were obtained when using fully synthetic ddhCTP. The fully synthetic ddhCTP was compared with ddhCTP produced via enzymatic phosphorylation of isotopically labeled ddhC by LCMS—the chemically synthesized ddhCTP coeluted with isotopically labeled chemoenzymatic ddhCTP (Figure S5).

Figure 2.

Dithionite contamination reduces apparent activity of NAD⁺-dependent enzymes. (A) ddhCTP from 0.2 to 1 mM did not effectively inhibit NAD⁺-dependent enzymes GAPDH, MDH, and LLDH. (B) Recapitulation of the assay conditions from the Ebrahimi et al. paper.¹⁷ GAPDH activity was assayed in the presence of the uninitiated reaction containing viperin-like enzyme TtVip (GAPDH+TtVip), filtrate from an overnight reaction containing TtVip reacted with CTP, SAM and dithionite, or 5 mM dithionite (GAPDH+TtVip+DT). Similar to a previous publication, the reaction containing dithionite inhibited the reaction. We then tested if pure ddhCTP (GAPDH+ddhCTP) produced would inhibit GAPDH and saw no statistically significant inhibition. However, pure dithionite (GAPDH+DT) gave comparable inhibition to the GAPDH+TtVip+DT condition. Dithionite reduction of NAD+ gives the appearance of inhibition. NS demarcates no statistical significance. ∗∗ indicated significance to the p < 0.05 level.

To determine the origin of this discrepancy, we reproduced the precise procedure reported by Ebrahimi et al.¹⁷ (see Supporting Information). The viperin-like enzyme TtVip prefers UTP as a substrate but can also utilize CTP to produce ddhCTP.²² Therefore, we cloned, expressed, and purified TtVip and set up a reaction to produce ddhCTP as previously described by Ebrahimi et al. (see Supporting Information).¹⁷ The reaction was incubated for ∼16 h, then applied to a 10 kDa MWCO spin concentrator to separate TtVip from the reaction products. The filtrate was collected, and the percent conversion of CTP to ddhCTP was determined via LCMS. After the reaction, the filtrate was calculated to contain ∼1.5 mM ddhCTP with 0.5 mM CTP remaining, similar to Ebrahimi et al.¹⁷ We next performed the assays with GAPDH under conditions reported by Ebrahimi et al. The GAPDH reaction with the addition of 200 μM of TtVip-produced ddhCTP (GAPDH+TtVip+DT) showed a ∼2.5-fold reduction of GAPDH activity relative to GAPDH alone (GAPDH) or GAPDH with the uninitiated reaction (GAPDH+TtVip). In addition to ddhCTP, the initiated reaction contained 5 mM dithionite relative to the GAPDH control and GAPDH with the uninitiated TtVip reaction. To test whether ddhCTP or dithionite inhibited the rate of reaction of NAD⁺-dependent enzymes, we then tested GAPDH in the presence of either pure ddhCTP (GAPDH+ddhCTP) or pure dithionite (GAPDH+DT). Pure ddhCTP did not recapitulate the decreased rate of NADH formation, but the addition of pure dithionite showed comparable levels of inhibition with GAPDH+TtVip+DT reaction (Figure 2B). These results indicate that ddhCTP did not inhibit GAPDH, but that some component in the TtVip reaction initiated with dithionite was inhibiting the assayed GAPDH activity. Because it is known that dithionite can readily reduce NAD⁺ to NADH,²⁴ we reasoned that the observed inhibition was due to the presence of residual dithionite in the filtrate, which could appear to inhibit GAPDH by reducing NAD⁺ back to NADH or by another unknown mechanism. Indeed, when using pure dithionite in a GAPDH reaction, we observed a comparable level of inhibition of activity to that of the GAPDH+TtVip+DT condition (Figure 2B). We therefore conclude that ddhCTP does not inhibit these NAD⁺-dependent enzymes.

ddhCTP and the other ddhC nucleoside phosphates represent a previously unexplored family of small molecules, generated by enzymes encoded by the host genome, with undetermined functions in the cell. Understanding the role of ddhCTP in the cell physiology will depend upon the rigorous biochemical characterization of protein interactions with ddhCTP and metabolomics characterizing the breakdown and metabolic transformation of the various ddhC nucleoside phosphates. Herein, we present a robust and effective method of producing ddhC nucleoside phosphates and their purification. Our chemoenzymatic synthesis had 78% yield, which was comparable to the chemical synthesis of ddhCTP from a protected ddhC nucleoside published in Wood et al. (60.5% yield starting from a protected ddhC nucleoside).¹⁹ The chemical method is better equipped to produce ddhC nucleoside phosphates on the gram scale, but our chemoenzymatic method produces pure ddhC nucleoside phosphates with high efficiency at the milligram scale making it an effective method for the purposes of most biochemical laboratories. Using these methods, we have tested and rebuked previous claims of a functional interaction between ddhCTP and NAD⁺-dependent enzymes. Although ddhCTP does not appear to inhibit NAD⁺-dependent enzymes in our hands, there are still myriad potential protein targets that must be tested to better understand the mechanisms by which the human body combats a broad array of viral infections using ddhCTP. With this semisynthetic method in hand, the ability to interrogate these interactions and discern the role of ddhCTP in both the innate immune response and metabolism is possible.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsbiomedchemau.3c00014.

• Figures S1–S5 and Methods. Recombinant expression and purification of nucleotide kinases (Figure S1). Mass to charge ratio peaks for chemoenzymatically synthesized ddhC nucleoside phosphates (Figure S2). UV traces of ddhCTP elution via liquid chromatography for determination of purity (Figure S3). Assay of NAD⁺-dependent activity in the presence of purified ddhCTP (Figure S4). LCMS coinjection of isotopically labeled chemoenzymatically synthesized ddhCTP with chemically synthesized ddhCTP (Figure S5). (PDF)

Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

This work was supported by NIH Grants P01 GM118303 (S.C.A.), R21-AI133329 (T.L.G. and S.C.A.), the Wollowick Family Foundation (S.C.A.), the Price Family Foundation (S.C.A.), contributions to the Montefiore Einstein Cancer Center Program for Experimental Therapeutics by Pamela and Edward S. Pantzer, the Einstein-Rockefeller-CUNY Center for AIDS Research (P30AI124414), and the Searle Scholars Program (T.L.G.). We acknowledge the Albert Einstein Anaerobic Structural and Functional Genomics Resource (http://www.nysgxrc.org/psi3/anaerobic.html) within the Macromolecular Therapeutics Development Facility. J.M.W. and L.D.H. thank the Ministry of Business Innovation & Employment for support of this work (Endeavor Fund, Contract UOOX1904 (N.Z.))

The authors declare no competing financial interest.