Synthesis of Structural ADP-Ribose Analogues as Inhibitors for SARS-CoV-2 Macrodomain 1

Protein adenosine diphosphate (ADP)-ribosylation is crucial for a proper immune response. Accordingly, viruses have evolved ADP-ribosyl hydrolases to remove these modifications, a prominent example being the SARS-CoV-2 NSP3 macrodomain, “Mac1”. Consequently, inhibitors are developed by testing large libraries of small molecule candidates, with considerable success. However, a relatively underexplored angle in design pertains to the synthesis of structural substrate mimics. Here, we present the synthesis and biophysical activity of novel adenosine diphosphate ribose (ADPr) analogues as SARS-CoV-2 NSP3 Mac1 inhibitors.

T he COVID-19 pandemic caused by SARS-CoV-2 has prompted extensive research on different treatment possibilities of this disease. 1−8 Mac1 not only binds ADPr but also hydrolyses the glycosidic bond between the distal ribose of ADPr and the side chain of the amino acid thus removing the PTM. 7,9The capability of Mac1 to cleave ADPr from the protein has evolved by the virus to counteract the host immune responses.Specifically, mammalian antiviral ADPribosyl-transferases such as PARP14 modify protein targets by attaching a single ADPr residue to an amino acid side chain in the protein.Such mono-ADP-ribosylation (MARylation) results in the induction of interferon and initiation of antiviral immune responses. 10,11ecent studies suggest that Mac1 reverses the PARP14catalyzed MARylation by removing ADPr from the signaling proteins and thus abolishes the antiviral response of the host. 7herefore, Mac1 inhibition could lead to reinstating the protective PARP-mediated immunologic function after infection. 12−17 A number of small molecule inhibitors of micromolar potency 13 demonstrated encouraging Mac1 inhibitory properties.Notably, mimicking the natural substrate with ADPr analogues is a relatively underexplored approach for the development of Mac1 inhibitors but has recently led to highly potent Mac1 binders. 18Here, we report on such an approach to Mac1 inhibition by synthesizing advanced ADPr derivatives and evaluating their potential as inhibitors for Mac1.−26 In this work, we first coupled different 5′-O-phosphoryl ribosides as the P(V) components to a 5′-phosphoramidite of a protected adenosine as the P(III) component, generating a focused library of ADPrmimics (Figure 1).After evaluating the binding potency to Mac1 of these first-generation mimics, we substituted the adenosine nucleoside for remdesivir using the P(III)−P(V) chemistry (Scheme 1) in two of the best binders from the ADPr library and obtained NDPr derivatives that demonstrate binding potency in the low nanomolar range.
We set out to synthesize the ADPr derivatives with the distal ribose modified at three sites, the 1″ (anomeric), 2″, and 3″ positions, while keeping the ADP-part unchanged (Figure 1B, Table 1).The general synthetic strategy and the key steps toward the ADPr-mimics are in Scheme 1, while the synthesis is detailed in Schemes S1−S6.We started with the synthesis and evaluation of α-O-methyl-ADPr 1 (Table 1, Scheme S1) that we compared with previously described 23 β-O-methyl-ADPr derivative 2 (Table 1) since they are known to inhibit other macrodomains.As expected, the β-oriented ADPr 2, in which the O-methyl would sterically interfere with the glycine-rich loop region, is a notably worse inhibitor for Mac1.Interestingly, installation of an α-azide at the anomeric center of the distal ribose 25 increased the Mac1 inhibitory activity to the high nanomolar range (3, IC 50 = 0.49 μM), a higher potency than native ADPr, which we used as reference (IC 50 = 1.2 μM).Notably, the binding of compound 3 to Mac1 with submicromolar potency has been reported by Lin and coworkers and is in line with our findings. 18It is likely that the polarity of the azide group may allow additional interactions with the glycine-rich loop region (G46-G48) of Mac1 surrounding the 1″ position of the distal ribose, consequently increasing its affinity for Mac1 binding.The previously reported 25 β-azide-ADPr 4 is a notably worse inhibitor (IC 50 = 127 μM).
Because of the possible hydrolytic sensitivity of αribosylazide, a stable isostere bearing a linear substituent that mimics the anomeric azide was considered.To this end, we prepared ADPr-mimic 5 (Schemes 1 and S2).A key step in the synthesis of 5 was the establishment of the C-glycosidic bond in 13 that was installed through a stereoselective Grignard reaction on the C-1 aldehyde in the open-chain form of Dribose derivative 12 (d.r.= 95:5).Interestingly, the protecting group choice is known to be a decisive factor in shaping the configuration of C-glycoside 13.It has been reported that Felkin-Anh control gives the β-alkyne which occurs in the case of 4-methoxybenzyl (PMB) protection groups, while anti-Felkin-Anh control gives the corresponding α-alkyne in the case of iso-propylidene protection. 27Alkyne 13 was subsequently converted into the required 5′-phosphate, and then, the corresponding ADPr analogue 5 was produced via the P(III)−(PV) pyrophosphate coupling method applying 5′phosphoramidite of adenosine as a P(III) component. 28otably, Mac1 affinity for the C-glycoside 5 did not significantly diminish compared to the N-or O-linked glycosides 1 or 3 (2.4 μM), indicating that the glycosidic linkage is dispensable for binding.
With lead compounds 1 and 3 in hand, the 2″ and 3″ positions were modified next.These positions have been reported to participate in hydrogen bonding within the enzymatic pocket. 4However, a phenylalanine residue of Mac1 (Phe132) near the distal ribose could potentially interact with apolar functionalities like benzyl groups.Such beneficial hydrophobic interactions could possibly offset the loss of hydrogen bonding. 29To this end, monobenzylated mimics 6 and 7 were prepared (Schemes S3 and S4).However, both compounds showed poor affinity for Mac1 with IC 50 values far above of those of the parent α-O-methyl-ADPr 1 and α-azide-ADPr 3, indicating that loss of hydrogen bonding, steric bulk, or a combination of the two counteracted any favorable hydrophobic interactions.Additionally, analogue 8 with a fluorine at the 3″ position was prepared by using diethylaminosulfur trifluoride (DAST) to displace a secondary alcohol from D-lyxose derivative 14 to afford 15 (75%, for the α-anomer), which was then further transformed into its Scheme 1. Key Steps Towards ADPr Analogues 5, 8, and 9 and General Synthetic Strategy Table 1.IC 50 Values for ADPr Analogues Organic Letters corresponding ADPr analogue 8 (Schemes 1 and S5).Interestingly, the introduction of the fluorine also drastically diminished binding affinity, indicating that not necessarily group size alone but also lack of hydrogen bonding was one of the issues at hand.Finally, to probe the binding pocket flexibility toward changes in geometry at the 3″ position, an Omethyl oxime was proposed.To this end, the 2″ position of Dribose analogue 16 was regioselectivity protected with a PMB group (63%), whereafter oxidation of the remaining alcohol using Dess-Martin Periodinane (DMP) to its corresponding ketone (68%) and subsequent transformation into the Omethyl oxime gave intermediate 17 with 82% yield (Schemes 1 and S6).This intermediate was then further developed into its corresponding ADPr analogue 9.However, biophysical activity testing showed that the oxime was detrimental to binding affinity to Mac1, thus resulting in an IC 50 of >200 μM.
Since modifications on the 2″ and 3″ positions did not yield any desirable results in terms of potency, we did not pursue the modification of these positions any further.Instead, we thought to capitalize on the enhanced potency of the adenosine mimic known as the remdesivir metabolite GS-441524 for binding Mac1. 30To this end, compounds 10 and 11 were synthesized, possessing either the α-O-methyl riboside or α-azide riboside moiety, respectively (Scheme 2, Table 1).
Conveniently the corresponding ribosyl-5-O-phosphate (20 and 21) building blocks were readily available from the precursors used in the synthesis of ADPr-mimics 1 and 3.
We therefore focused on preparation of phosphoramidite 22 (Scheme 2), starting from GS-441524, the synthesis of which has been reported as part of efforts in the development of remdesivir. 31To this end, we selectively protected the 5′position with a tert-butyldiphenylsilyl group, and the remaining positions were protected with Boc-groups using an excess of ditert-butyl dicarbonate and 4-dimethylaminopyridine (DMAP).This afforded the fully protected intermediate in 80% yield, with the exocyclic amine being double protected.Subsequent removal of the silyl protecting group using HF in pyridine (quantitative) and installing the phosphoramidite gave the required P(III)-building block 22 in 94% yield.Coupling of this phosphoramidite with phosphates 20 and 21 and subsequent acidic global deprotection and HPLC purification delivered the desired pyrophosphates 10 and 11 in 17% and 4.3% yield over five steps.
Interestingly, both analogues demonstrated excellent IC 50 values for Mac1 inhibition upon biophysical activity testing: 63 nM and 30 nM for α-O-methyl riboside 10 and α-ribosylazide 11, respectively, which is in line with the work of Lin and coworkers, 32 who also described the development of GS-441524based ADPr analogues, which was reported as this manuscript was in preparation.The slightly higher inhibitory activity of compound 11 compared to compound 10 mirrors the higher binding affinity of the azide ADPr analogue 3 in comparison to O-methyl-ADPr 1 (Figure S2).
To shed light on the binding interactions, we docked inhibitors 10 and 11 into the binding pocket of Mac1 (PDB 7KQP) (Figure 2).Both analogues demonstrated an additional hydrogen bond with Asp157 through the cyano group, which is not possible for ADPr 1 and 3.The docking study also revealed a new hydrogen bond for the anomeric azide in 11 to Gly47, which is not observed for 10 (Figure S1).The binding energies derived from the molecular docking study (ΔG = −11.55,−12.04, and −10.08 kcal/mol for 10, 11, and ADPr, respectively) are in good agreement with the inhibitory potency of the two binders with respect to ADPr.Further analysis by determining hydrogen bond occupancy values from the docked structures (Figure S1) indicated that the most potent binder interacts the strongest with Val49, Ile131, and Phe132.Finally, molecular dynamics computations delivered binding enthalpies that were in good agreement with the IC 50 values and docking energies obtained for ADPr and the two analogues.
In conclusion, we report the synthesis and biophysical evaluation of a focused array of ADPr-mimics to reveal that modification of the 2″ and 3″ alcohols on the distal ribose (as in 6−9) is detrimental for inhibitor potency but that installing an α-azido group on the ribose significantly enhances binding.The modular approach, in which the phosphate of the modified ribose was coupled to a phosphoramidite of GS-441524, resulted in the development of a nanomolar Mac1 inhibitor.Our molecular docking studies identified several new hydrogen bonding interactions established by both the cyano group of the GS-441524 moiety as well as the anomeric azide on the distal ribose, explaining/rationalizing the high binding affinity of the inhibitor to Mac1.Furthermore, the pyrophosphate 11 is a promising lead compound for the development of more potent and more stable analogues as future therapeutic antivirals against SARS-CoV-2 NSP3Mac1 and other macrodomain-containing viruses.

Figure 2 .
Figure 2. Docking of compound 11 in the Mac1 active site.