Discovery and Synthesis of a Phosphoramidate Prodrug of a Pyrrolo[2,1-f][triazin-4-amino] Adenine C-Nucleoside (GS-5734) for the Treatment of Ebola and Emerging Viruses.

The recent Ebola virus (EBOV) outbreak in West Africa was the largest recorded in history with over 28,000 cases, resulting in >11,000 deaths including >500 healthcare workers. A focused screening and lead optimization effort identified 4b (GS-5734) with anti-EBOV EC50 = 86 nM in macrophages as the clinical candidate. Structure activity relationships established that the 1'-CN group and C-linked nucleobase were critical for optimal anti-EBOV potency and selectivity against host polymerases. A robust diastereoselective synthesis provided sufficient quantities of 4b to enable preclinical efficacy in a non-human-primate EBOV challenge model. Once-daily 10 mg/kg iv treatment on days 3-14 postinfection had a significant effect on viremia and mortality, resulting in 100% survival of infected treated animals [ Nature 2016 , 531 , 381 - 385 ]. A phase 2 study (PREVAIL IV) is currently enrolling and will evaluate the effect of 4b on viral shedding from sanctuary sites in EBOV survivors.

Introduction

Ebola virus disease (EVD) was first documented 40 years ago during an outbreak of infectious hemorrhagic fever in Northern Zaire (current Democratic Republic of Congo). More than 20 intermittent outbreaks have occurred since then, but the most recent outbreak in West Africa spanning 2013–2016 has been the largest recorded in history and presented an international public health emergency.[1] Over 28,000 cases were confirmed in Guinea, Liberia, and Sierra Leone resulting in >11,000 deaths including >500 healthcare workers, which severely strained the local medical infrastructure.[2] In survivors, the Ebola virus (EBOV) can persist in bodily fluids for months after the onset of acute infection potentially leading to EVD-related sequelae and viral recrudescence.[3] While rare, secondary transmission has been documented to occur through sexual intercourse implicating persistent virus in genital secretions.[4] Despite the end of the current outbreak, the potential for equally devastating future outbreaks together with the persistent virus observed in survivors makes the development of a safe, effective, and readily available treatment option for EVD a high priority.

EBOV, a member of the Filoviridae family, is a single-stranded, negative-sense, nonseg<span class="Species">mented RNA virus that is the causative agent of EVD. Other Filoviridae family members include Marburg, Sudan, and Bundibugyo viruses, which have all been responsible for outbreaks associated with high mortality rates in sub-Saharan Africa.[5,6] Over the course of the recent West African EVD outbreak, several direct acting anti-Ebola agents including monoclonal antibodies (ZMapp),[7] interfering-RNAs,[8−10] and small molecule nucleoside(tide) antivirals such as favipiravir (1),[11−13] and brincidofovir (2)[14] have been evaluated in early clinical trials (Figure ). More recently another nucleoside analogue, galidesivir (3, BCX4430[15]), has entered clinical development. These developments are encouraging, but to date, none of these potential therapeutics have established robust clinical efficacy for the treatment of acute infection or the viral persistence and sequelae. Several vaccines have shown strong promise for preventing EBOV infection, but the breadth and durability of protection they can afford has yet to be established.[16]

Figure 1

Structures of antiviral nucleosides and nucleoside phosphonates.

Structures of antiviral nucleosides and <span class="Chemical">nucleoside phosphonates.

Prior to the Ebola outbreak, we had embarked on a strategic initiative aimed at evaluating the potential of nucleoside analogues for the treatment of selected emerging viruses. A library of ∼1000 diverse nucleoside and nucleoside phosphonate analogues was harnessed from over 2 decades of research across multiple antiviral programs. In collaboration with the Center for Disease Control and Prevention (CDC) and the United States Army Medical Research Institute of Infectious Diseases (USAMRIID), selected compounds from the library were screened against EBOV, leading to the identification of parent 4 and a potent monophosphate prodrug mixture 4a that contained the single Sp isomer 4b (GS-5734[17]) that was selected for development. This report describes in detail the structure activity relationships (SAR) of the parent nucleoside, prodrug optimization and selection, and synthesis optimization of the development candidate 4b. Candidate compound 4b is currently in phase 2 trials to assess the effect on the chronic shedding of virus in EVD survivors following promising efficacy data established in a non-human-primate (NHP) EVD challenge model. These data have been recently reported[17] and will be summarized along with the early clinical experience with 4b.

Results and Discussion

The assembly of the ∼1000 compound nucleos(t)ide screening library was heavily focused toward ribose analogues that could target RNA viruses since this would encompass many emerging <span class="Disease">viral infections ranging from respiratory pathogens belonging to the Coronaviridae family such as severe acute respiratory syndrome (SARS) and Middle East respiratory syndrome (MERS), to mosquito-borne viruses of the Flaviviridae family such as Dengue and Zika. The majority of the library compounds were nucleosides that contained a cyclic modified ribose or “ribose-like” core. These nucleosides were also predominantly N-nucleosides. Less than 10% of the library comprised nucleoside phosphonates or acyclic analogues due to the limited success, to date, in identifying potent RNA virus inhibitors with these types of analogs. A second key factor in the library assembly was that approximately 50% of the library included monophosphate and ester prodrugs to capture analogs that may be missed in cellular screens due to either poor permeability or inefficient metabolism in the respective cell types that the different antiviral assays utilize. Nucleoside analogs require activation by intracellular nucleoside/tide kinases to generate their respective nucleoside triphosphate (NTP) metabolites in order to then compete with endogenous natural nucleotide pools for incorporation into the replicating viral RNA.[17] The first phosphorylation step to generate the nucleoside monophosphate is often rate limiting, and therefore the application of monophosphate prodrugs, especially phosphoramidates (ProTides), has been extensively explored in nucleoside analogs to bypass this initial phosphorylation step.[18] A notable example includes the phosphoramidate prodrug Sofosbuvir (5) for the treatment of HCV (Figure ).[19] Nucleoside phosphonate analogs are bioisosteres of the monophosphates but also require prodrugs to enable masking of the charged phosphonate acid thereby allowing more efficient entry into cells. A recent example of an approved drug in this class is the phosphonamidate prodrug tenofovir alafenamide (6) for the treatment of HIV.[20] In both examples the amidate prodrugs effectively deliver high levels of triphosphate (diphosphophosphonate in the case of nucleoside phosphonates) inside the target cells and demonstrate significant improvements in potency compared to their respective parent nucleos(t)ides when screened in antiviral assays.[20,21]

In the original library screening toward a panel of RNA viruses across different viral families, promising leads were identified. Subsequent to the EVD outbreak, some of these analogs were selected for EBOV testing in collaboration with the CDC and USAMRIID in a BSL-4 facility. From this screen nucleoside 4,[22] a 1′-CN modified adenosine C-nucleoside emerged with sub-micromolar activity toward EBOV in human microvascular endothelial cells (HMVEC-TERT) cells (entry 2, Table ). In addition, its phosphoramidate prodrug mixture 4a(23) (entry 3, Table ) containing ∼1:1 ratio of Sp 4b and Rp 4c diastereoisomers (Figure ) was found to be very potent toward EBOV in both HeLa and HMVEC cells. Encouraged by these data, the anti-EBOV activity for a range of nucleoside analogs and their prodrugs was evaluated and the results are reported in Table , along with activity toward respiratory syncytial virus (RSV), from the Pneumoviridae family, and HCV, from the Flaviviridae family.

Table 1

SAR of Nucleoside Parents and Selected Prodrugsa

entry	compd	EBOV EC50 HeLa cells (μM)	EBOV EC50 HMVEC cellsb (μM)	RSV EC50 HEp-2 cells (μM)	HCV 1b EC50 Huh-7 cells (μM)	CC50 HEp-2 cells (μM)	CC50 Huh-7 cells (μM)	CC50 MT4 cells (μM)
1	7				0.0035	0.038c	0.15	<0.01
2	4	>20	0.78	0.53	4.1	>100	>88	>57
3	4a	0.17	0.12	0.027	0.023d	9.2	17d	2.0
4	4b	0.10	0.053	0.015	0.057	6.1	36	1.7
5	8		>10	5.5	38	93	62	4.5
6	9			>200	>88	>200	>88	120
7	9a		3.9	1.1	6.9	>100	>44	>32
8	10		56	>100	>44	>100	>88	>53
9	11	>50	>10	7.3	12	>100	>44	>57
10	11a	>20		63	2.5	>100	>44	53
11	12			>100c	>44		>44	32c
12	13	50	>10	>100	>44	>100	>44	>57
13	13a	27	13d	>50	0.37	>50	>44	1.4
14	13b	>20	40	>20	0.31	95	51	7.8

Data reported are at least n ≥ 2 in 384 well assay format unless otherwise noted.

HMVEC cells = TERT-immortalized human foreskin microvascular endothelial cells (ATCC-4025) cells.

96 well assay format.

n = 1 data only.

The presence of the 1′-CN modification in 4 was found to be critical in providing selectivity toward viral polymerases and avoiding the significant <span class="Disease">toxicity (CC50 < 0.01–0.15 μM) associated with the unmodified C-nucleoside 7 (entry 1, Table ).[24] The MT4 cell line was also used as a sensitive cell line to evaluate cytostatic effects of nucleoside analogs and confirmed the poor selectivity of 7 observed in both the HEp-2 and Huh-7 cell lines. The prodrug mixture 4a, in addition to potent anti-EBOV activity, demonstrated significant activity toward RSV and HCV, with potencies similar to or better than that for EBOV (EC50 < 120 nM). The broad and potent antiviral activity across all three viruses for 4a was further supported by the potent activity of the single Sp isomer 4b toward the same viruses and also other emerging RNA viruses such as MERS and Junin viruses, and to a lesser extent Lassa.[17] The antiviral selectivity of 4b toward EBOV was 17–32-fold compared to the MT4 cell line CC50, and higher in the other cell types reported in Table . Given the anticipated short treatment duration for EVD, this window of in vitro selectivity was considered sufficient for continued interest in 4b. The 1′-methyl analogue 8 (entry 5)[22] was less active toward EBOV and also displayed a higher degree of toxicity compared to the 1′-CN analogue 4 illustrating how small changes in the polarity and size of the 1′ substituent can impact the overall profile. The 1′-ethynyl analogue 9 (entry 6)[22] and its corresponding 2-ethylbutyl alanine prodrug 9a (entry 7) were both less active when compared to their respective 1′-CN counterparts (4 and 4a, respectively).

Compound 4 is a C-nucleoside analogue which provides chemical and enzymatic stability toward deglycosylation reactions at the anomeric center. However, alternate base modifications including N-nucleosides were also studied. Interestingly, the corresponding 1′-CN modified adenosine N-analogue 10 (entry 8)[25] was significantly less active toward all viruses, while the 1′-CN modified N-nucleoside pyrimidine 11 (entry 9) retained weak antiviral activity only for RSV and HCV.[26,27] The phosphoramidate prodrug of 1′-CN cytidine 11a (entry 10)[27] did not improve the potency toward EBOV (in HeLa cell assay) or the other viruses tested, presumably due to limitations in metabolism beyond the monophosphate. In general, the potency trends of 1′ substitution and nucleobase changes were similar across EBOV, RSV, and HCV, which was in contrast to the trends uncovered with 2′ modifications. The 2′-deoxy-2′-fluorine analogue 12 (entry 11)[23] and the 2′-β-methyl analogue 13 (entry 12)[28] both lacked significant antiviral activity. However, the 2′-β-methyl phosphoramidate prodrugs, analogs 13a and 13b (entries 13 and 14),[28] respectively, were both potent toward HCV, and only weakly active/inactive toward EBOV and RSV. This result suggests that HCV polymerase is more able to accommodate the 2′-β-methyl group compared to the EBOV and RSV polymerases. Taken together with the 1′ substitution and nucleobase SAR, the EBOV and RSV polymerases demonstrated similar activity trends, while HCV polymerase was differentiated in SAR at the 2′ position.

To interrogate the cell based SAR more rigorously the active NTP metabolite 4tp(23) was tested toward the viral polymerases (Table ). The triphosphate 4tp demonstrated a half-maximal inhibitory concentration (IC50) of 1.1 μM against the RSV RdRp and 5.0 μM against HCV RdRp. The Ebola viral polymerase has to date evaded efforts toward its isolation and expression, so the intrinsic activity of the active NTP metabolite cannot be directly evaluated. An alternate method for estimating the inhibitory properties of an NTP for its viral target is to measure the NTP levels inside cells following incubation with the parent or prodrug compound at a given concentration and then calculate the NTP levels at the EC50 measured in the same cells.[17] For example, in a continuous 72 h incubation of 1 μM 4a, the 4tp levels were measured at 2, 24, 48, and 72 h, and reached a Cmax of 300, 110, and 90 pmol/million cells in macrophages, HMVEC, and HeLa cells lines, respectively. The several-fold difference in maximum 4tp levels is not unusual and reflects the differences between cells with respect to their ability to break down the prodrug and subsequently metabolize the released monophosphate to the active 4tp. The average NTP levels over the 72 h incubation of 4a were then used along with an average cell volume of 2 pL[29] to calculate an estimated half-maximal inhibitory concentration of ∼5 μM for the intracellular inhibition of EBOV polymerase. This is comparable in potency toward RSV and HCV polymerases supporting the potent antiviral EC50 data demonstrated for the prodrug mixture 4a across the three viruses when allowing for cell differences (Table ). The selective inhibition of the viral polymerases vs host polymerases is considered a key factor in the development of a safe and effective nucleoside antiviral.[30,31] Therefore, 4tp was evaluated toward several host polymerases and was found to be a weak incorporator toward mitochondrial polymerase (POLRMT) and not a substrate for DNA polymerase γ, as would be expected given the presence of the ribose 2′ OH (Table ). Across the host RNA and DNA polymerases evaluated there was no inhibition up to 200 μM (Table ) demonstrating a high degree of selectivity of 4tp toward the viral polymerases compared to representative examples of host polymerases.

Table 2

Inhibition of RSV Polymerase, HCV Polymerase, and Human Polymerases by 4tp

enzyme	4tp IC50 (μM)	4tp SNIa rate (%)
RSV RdRp	1.1
HCV RdRp	5
POLRMT	>200	6
RNA Pol II	>200
DNA Pol α	>200
DNA Pol β	>200
DNA Pol γ	>200	0

SNI = single nucleotide incorporation.

Molecular structure information is not available for the EBOV or RSV polymerases, so modeling of the active sites was performed based on the published structures for HIV and HCV polymerases, together with an analysis of the respective sequences.[32] Within the modeled active site of EBOV polymerase the major difference between EBOV and RSV is Y636 (EBOV) compared to F704 (RSV) and the major difference between EBOV and HCV is E709 (EBOV) compared to S282 (HCV). On docking the triphosphate 4tp, the 1′-CN group occupies a pocket formed by residues that are identical between EBOV and RSV, yet very different in HCV (Figure a). Nevertheless the 1′-CN analogue retains antiviral potency across these viruses and others[17] suggesting that a pocket exists to accommodate the 1′-CN group in many viral polymerases including other filoviruses.[33] The 2′-β-H of 4tp is in close proximity to E709, and replacement of this group with a 2′-β-methyl (13tp) would be anticipated to interfere with E709 (Figure b).[34] However, the 2′-β-methyl can be accommodated by the larger pocket afforded by the smaller S282 residue of HCV (Figure c). This suggests the lack of activity for 2′-β-methyl analogs toward EBOV and RSV and retained potency toward HCV is likely due to steric constraints in the polymerase active site. Consistent with the model, EBOV and RSV both have the E709 or equivalent residue, and 13tp was found to be significantly less active (IC50 > 30 μM) toward RSV.

Figure 2

(a) Compound 4tp modeled into the EBOV polymerase active site. Residue Y636 is highlighted in green surface, sits below the ribose, and corresponds to F704 in RSV. Residue E709 is highlighted in red surface, sits in proximity to the 2′-β-H position of the ribose, and corresponds to S282 in HCV. (b) Compound 13tp modeled into the EBOV polymerase active site. The 2′-β-methyl overlaps with residue E709 highlighted in red. (c) Compound 13tp modeled into the HCV polymerase active site. Residue S282 is highlighted in the yellow surface, and the 2′-β-methyl can be accommodated.

The screening and modeling efforts established 4 as the best lead for prodrug optimization. The ability to evaluate prodrugs, especially in vivo, required an efficient synthesis route for both the nucleoside 4 and, preferably, a single prodrug diastereoisomer. Neither was available at the outset, so significant chemistry resources were applied to improve the robustness and scalability of the route along with generation of single prodrug diastereoisomers. The first generation synthesis of 4 and the single Sp phosphoramidate prodrug 4b commenced with a glycosylation reaction via metal–halogen exchange of the bromo-base 15 followed by addition into the ribolactone 14 (Scheme ). Two conditions were identified to render this desired C–C bond formation. The first condition (a) proceeded through addition of excess n-BuLi to a mixture of TMSCl and 15, which was designed to result in lithium–halogen exchange after removal of the acidic 6N protons by silyl protection. Addition of this in situ generated reagent to the ribolactone 14 then afforded 16 in 25% yield.[23,35,36] The alternative conditions (b) employed sodium hydride and 1,2-bis(chlorodimethylsilyl)ethane for the 6N protection step, followed by lithium–halogen exchange, and addition to the lactone to afford 16 in 60% yield.[22,36]

Scheme 1

First Generation Synthesis of 4b

Reagents and conditions: (a) n-BuLi, (TMS)Cl, THF, – 78 °C, 25%; (b) 1,2-bis(chlorodimethylsilyl)ethane, NaH, n-BuLi, THF, – 78 °C, 60%; (c) (TMS)CN, BF3·Et2O, CH2Cl2, – 78 °C, 58% (89:11β-17/α); (d) BCl3, CH2Cl2, – 78 °C, 74%; (e) 19, NMI, OP(OMe)3, 21%; (f) OP(OPh)Cl2, Et3N, CH2Cl2, 0 °C, 23%.

The efficiency of both conditions was suboptimal as the yields were capricious and highly dependent on the cryogenic temperatures and the rate of n-BuLi addition required for the transformation. Furthermore, premature quenching and reduction of <span class="Chemical">lithio base was observed, which was rationalized to be a consequence of deprotonation α to the lactone under the highly basic conditions. Compound 16 was isolated as a mixture of 1′-isomers, which were taken into the subsequent 1′-cyanation reaction to isolate the major product, β-anomer 17, by chromatography.[37] Following removal of the three benzyl protecting groups to afford 4, the diastereomeric mixture of the phosphoramidoyl chloridate prodrug moiety 19 was then coupled to provide 4a in 21% yield, as an ∼1:1 diastereomeric mixture.[23] The two diastereomers were resolved using chiral HPLC to afford the Sp isomer 4b and Rp isomer 4c, respectively.[38] While this route initially provided quantities of 4b, the variability in yields, suboptimal selectivity, frequent use of cryogenic temperatures, and chiral chromatography hindered this route from being suited to larger scales.

The second generation route enabled the diastereoselective synthesis of the single Sp isomer 4b on scales suitable to advance the compound into preclinical efficacy and toxicity studies (Scheme ).[17] The glycosylation step employed the iodo-base 20 instead of the bromo base, which enabled a more facile metal–halogen exchange compatible with i-PrMgCl·LiCl complex.[39] Treatment with PhMgCl and TMSCl provided 6N protection to remove the acidic protons with a higher degree of control, and addition of i-PrMgCl·LiCl followed by the ribolactone 14 at −20 °C afforded the glycosylation product 16 in a 40% yield. The milder reagents and temperature enabled large-scale batches to be carried out with consistent yields. Treatment of 16 with TMSCN, TMSOTf, and TfOH at −78 °C afforded 17 in 85% yield in >95:5 anomeric ratio. The inclusion of TfOH was key to promote the high yield and high selectivity favoring the desired β-anomer. Benzyl deprotection was effected through treatment with BCl3, and 4 was readily isolated through crystallization. Acetonide protection of the 2′,3′-hydroxyl moieties with 2,2-dimethoxypropane in the presence of H2SO4 afforded 21 in 90% yield. Utilizing the 2′,3′-acetonide protection was found to be optimal as the yield of the coupling reaction with the p-nitrophenolate 2-ethylbutyl-l-alaninate prodrug mixture 22a was dramatically improved compared to directly coupling to the unprotected nucleoside 4 (70% vs 40%). In the event, reaction of 21 with the single Sp isomer of the p-nitrophenolate prodrug precursor 22b in the presence of MgCl2 and Hünig’s base efficiently appended the prodrug group in 70% yield as a single Sp isomer. Final deprotection of the acetonide with concentrated HCl in THF afforded 4b in 69% yield.

Scheme 2

Second Generation Synthesis of 4b

Reagents and conditions: (a) TMSCl, PhMgCl, i-PrMgCl·LiCl, THF, – 20 °C, 40%; (b) TMSCN, TfOH, TMSOTf, CH2Cl2, – 78 °C, 85%; (c) BCl3, CH2Cl2, – 20 °C, 86%; (d) 2,2-dimethoxypropane, H2SO4, acetone, rt, 90%; (e) 22b, MgCl2, (i-Pr)2NEt, MeCN, 50 °C, 70%; (f) 37% HCl, THF, rt, 69%; (g) OP(OPh)Cl2, Et3N, CH2Cl2, – 78 °C, then 4-nitrophenol, Et3N, 0 °C, 80%; (h) i-Pr2O, 39%.

The single Sp isomer 22b of the p-nitrophenolate 2-ethylbutyl-l-alaninate prodrug precursor 22a was prepared through a sequence beginning with exposure of 2-ethylbutyl-l-alanine 18a to OP(OPh)Cl2, followed by 4-nitrophenol, to afford 22a as a diastereomeric mixture at phosphorus. Importantly, the single Sp isomer 22b was readily resolved from the mixture in 39% yield through crystallization in diisopropyl ether, a discovery that was paramount for the success of the diastereoselective synthesis of the 4b.[40] Thus, utilizing the p-nitrophenolate 2-ethylbutyl-l-alaninate prodrug coupling partner 22b offered a significant advantage over the chloridate 19 in the first generation sequence. Overall the second generation synthesis of 4b offered the following improvements: (1) milder glycosylation conditions at higher temperature to allow for consistent yields and scalability, (2) higher selectivity and yield for the 1′-cyanation reaction, and (3) a highly efficient coupling sequence of a single Sp prodrug moiety for the diastereoselective synthesis of 4b. Through this second generation route >200 g was rapidly prepared to support preclinical efficacy and toxicity studies.

The stereochemistry of the p-nitrophenolate 2-ethylbutyl-l-alaninate prodrug 22b and candidate compound 4b were unambiguously assigned by small molecule X-ray crystallography (Figure ). In both cases the Sp isomer was established and suggests that the coupling with the nucleoside and reagent follows a SN2 type inversion of the phosphorus stereocenter.

Figure 3

Thermal ellipsoid representations of (a) 22b and (b) 4b.

The improved method for preparing 4 enabled many monophosphoramidate and bisphosphoramide prodrug analogues to be synthesized, the results of which are summarized in Scheme . A number of conditions were identified to affect the coupling of the prodrug moieties to 4 or the 2′,3′-acetonide protected analogue 21. The reactions employing 4 proceeded under either Brønsted basic conditions utilizing t-BuMgCl or Lewis acidic conditions with MgCl2 in polar aprotic solvents to afford the desired prodrugs in yields ranging from 10 to 43%. The coupling reaction of the 2′,3′-acetonide protected analogue 21 under Lewis acidic conditions followed by in situ acetonide deprotection in general afforded much higher yields ranging from 60 to 70% (analogues 4b and 4n). Both the p-nitrophenol (PNP) and pentafluorophenol (PFP) prodrug electrophiles were compatible in the coupling reactions and typically achieved comparable yields. The reactions utilizing diastereomeric mixtures of the prodrug electrophiles, 22a,d–h,j,l,n provided the prodrug products in 1.5–2.6 to 1 diastereomeric ratios (unassigned) at phosphorus. In addition to the monophosphoramidate prodrugs, two bisphosphoramide prodrugs, 4o and 4p, were synthesized and evaluated since they avoided the preparation of chiral phosphorus reagents.

Scheme 3

Prodrug Synthesis

Prodrug is an undetermined mixture of diastereoisomers unless otherwise indicated.

aa = amino acid, Ala = alanine, Phe = phenylalanine, AIB = 2-aminoisobutyrate, c-Bu = cyclobutyl, c-Pent = cyclopentyl, Pent = pentyl, Neopent = neopentyl, 2-EtBu = 2-ethylbutyl, PNP = p-nitrophenolate, and PFP = pentafluorophenolate.

LG = leaving group.

Single Sp isomer.

Reagent was a single unassigned isomer at phosphorus.

NA = not applicable.

The monophosphate prodrugs can improve the potency of the parent nucleosides substantially by delivering the monophosphate into cells and effectively bypassing a rate limiting first phosphorylation step. The phenol and amino acid esters mask the negative charge of the monophosphate group enabling facile passive penetration into the cell. The prodrug breakdown is initiated by intracellular esterases (e.g., carboxy esterase 1 and cathepsin A) that cleave the ester unraveling the carboxylate moiety, which then continues to breakdown to the monophosphate that serves as the prescursor to synthesis of the intracellular NTP.[17]

Prodrugs 4a–p were evaluated toward EBOV in three cell lines and for human plasma stability (Table ). In general the antiviral activity trends for EBOV across all three cell lines were similar supporting the efficient conversion of these prodrugs across multiple different cell types. A series of ethyl esters with differing amino acids (entries 2–5) established that the phenylalanine and alanine amino acids were the most promising (Table ). Given the intended route of administration was intravenous, increasing lipophiliity beyond log D ∼ 2 was considered a potential issue due to solubility concerns. Therefore, to improve potency the emphasis was placed on the less lipophilic and more commonly used alanine amino acid, with subsequent modification of the ester. Nonproximally branched esters of alanine with increasing log D ranging from 0.6 to 2.1 (entries 5, 11, and 12) demonstrated increased potency. Esters that contained proximal branching without cyclic motifs e.g. i-Pr, t-Bu, and 3-pentyl (entries 7, 8, and 10, respectively) were generally less active, consistent with the increased steric hindrance that likely slows the cleavage rate by esterases. For example, the proximally branched 3-pentyl ester 4l (entry 10) has comparable log D to that of the neopentyl analogue 4m (entry 11) yet much lower potency. In contrast, cyclic butyl and pentyl esters (entries 6 and 9, respectively) are more potent than 4l despite the proximal branching and lower log D, although the potency in the HeLa cell assay was reduced. The d-Ala 2-EtBu mixture (entry 15) was less potent that the corresponding l-Ala analogue mixture (entry 12), and the two bisphosphoramide prodrugs (entries 16 and 17) also had reduced activity compared to their monoamidate counterparts (entries 5 and 12, respectively). Thus, based on antiviral properties across HeLa and HMVEC cells, the most promising monophosphoramidate prodrugs were the neopentyl ester 4m (single undefined isomer) and 2-EtBu ester mixture 4a. The Sp and Rp isomers of 4a (entries 14 and 13, respectively) were separated and found to be similar in potency, but both were marginally more potent than 4m. For the intended iv route of administration, plasma stability was not deemed critical in the selection process provided sufficient stability (t1/2 > 60 min) was maintained to allow loading of target cells harboring the virus during drug infusion. The selection of 4b was made based on the high potency across multiple cell lines and the crystalline nature of the Sp prodrug reagent 22b that allowed rapid scale up for efficacy and IND enabling studies of the single Sp isomer 4b.

Table 3

Antiviral Activity of Prodrugs 4a–p.a

entry	compd	ester (R)	aab	EBOV EC50 HeLa (nM)	EBOV EC50 HMVECc (nM)	EBOV EC50 macrod (nM)	CC50 MT4 (μM)	human plasma t1/2 (min)	log D
1	4			>20000	780	>20000	>57		0.3
2	4d	Et	l-Phe	4380	587	270	15	1584	1.6
3	4e	Et	l-Val	7040	3151		>100	1584	1.2
4	4f	Et	AIB	8470	1585		>100	1584	0.9
5	4g	Et	l-Ala	2425	636		13	1584	0.6
6	4h	c-Bu	l-Ala	420	88		6	815	1.1
7	4i	i-Pr	l-Ala	1845	367	297	21	1561	1.1
8	4j	t-Bu	l-Ala	30410	3790		>100	1584	1.3
9	4k	c-Pent	l-Ala	633	160	120	8.8	1578	1.3
10	4l	3-Pent	l-Ala	1810	845		>100	860	1.6
11	4m	Neopent	l-Ala	168	92		3	700	1.7
12	4a	2-EtBu	l-Ala	170	121	100	2	195	2.1
13	4c	2-EtBu	l-Ala	80	53	111	3	234	2.0
14	4b	2-EtBu	l-Ala	100	53	86	1.7	69	2.1
15	4n	2-EtBu	d-Ala	550	518	729	42	1584	2.1
16	4o	Et	l-Ala	20420	9102		>53		<0.3
17	4p	2-EtBu	l-Ala	970	678		9	507	2.7

Data is at least n ≥ 2 unless otherwise reported.

aa = Amino Acid, Ala = Alanine, Phe = Phenylalanine, AIB = 2-aminoisobutyrate, c-Bu = cyclobutyl, c-Pent = cyclopentyl, Pent = pentyl, 2-EtBu = 2-ethylbutyl.

HMVEC = TERT-immortalized human foreskin microvascular endothelial cells (ATCC-4025).

Macro = human macrophages.

aa = Amino Acid, Ala = <span class="Chemical">Alanine, Phe = Phenylalanine, AIB = 2-aminoisobutyrate, c-Bu = cyclobutyl, c-Pent = cyclopentyl, Pent = pentyl, 2-EtBu = 2-ethylbutyl.

In vivo efficacy evaluation of 4b was conducted in monkeys since this represented the most relevant animal model of EVD with similar pathophysiology to the actual human disease. In addition, phosphoramidate esters are highly prone to plasma metabolism in rodents on account of high expression of plasma carboxylesterases,[41] thereby excluding pilot efficacy studies in small animal models. Due to the high first pass hepatic extraction of phosphoramidates, oral administration was also not explored in favor of injectable routes of administration. Moreover, oral delivery in patients acutely infected with EBOV that are demonstrating symptoms of the disease may not be ideal because gastrointestinal symptoms may limit the dose that is effectively absorbed. Intravenous administration of 4b in rhesus monkeys demonstrated rapid elimination of prodrug and appearance of parent nucleoside in systemic circulation (Figure a). However, 4b also rapidly distributed to peripheral blood mononuclear cells (PBMCs) and triphosphate levels in PBMCs were elevated to a maximum within 2 h (Figure b). A dose of 10 mg/kg resulted in an estimated PBMC triphosphate level at 24 h that was several-fold higher than the estimated IC50 of 5 μM for EBOV, with a half-life of 14 h similar to that measured in vitro in human macrophages. The long intracellular half-life of 4tp supported once-daily iv administration in the rhesus efficacy study and in the clinical program.

Figure 4

Concentration–time profiles following 10 mg/kg iv single dose slow bolus administration of 4b in Rhesus (mean ± SD, n = 3 per time point). (a) Plasma profile of prodrug 4b (black circle) and parent nucleoside 4 (blue triangle). (b) Intracellular concentration of active metabolite 4tp in PBMCs (green diamond) and estimated 4tp EBOV IC50 = 5 μM (dashed black line).

Preclinical in vivo efficacy for 4b was conducted in an EBOV-infected rhesus challenge model. The overall strategy was to determine a maximally efficacious dose of 4b that could be safely administered and to understand whether delayed time of treatment would be effective, a property that was considered critical for the successful clinical application of 4b. These studies have recently been published in full and will only be summarized here.[17] The starting dose of 3 mg/kg iv was modeled based on the 4tp levels in rhesus PBMCs following 10 mg/kg iv dosing to have the potential for inhibiting Ebola viral replication in vivo. A correlation of the in vitro antiviral activity and intracellular metabolism described earlier suggested the 3 mg/kg dose would produce 4tp levels that exceed the estimated IC50 for most of the dosing interval. Compound 4b was dosed iv at 3 mg/kg on day 0 or day 2 relative to EBOV inoculation, and continued once daily for 12 days. Systemic viremia was reduced and survival out to day 28 postinfection was improved. Animals administered 3 mg/kg 4b, starting at day 0, had a survival rate of 33% while those initiated on day 3 had a 66% survival at 28 days. These encouraging data were then followed with a second study in which one arm explored dosing initiation on day 3 with 10 mg/kg daily for 12 days to assess whether increased dose compared to the 3 mg/kg result could be more efficacious. Two other arms explored an initial loading dose of 10 mg/kg on day 2 or day 3, followed by 11 3 mg/kg daily maintenance doses. All of the animals in the two arms in which 4b treatments were initiated on day 3 (n = 12 total) survived through day 28, the end of study. In the daily 10 mg/kg dosed group the effect on viremia was consistently greater than in any of the other groups and was below the limit of quantitation (8 × 104 RNA copies mL–1) in four of the six animals on days 5 and 7 relative to the vehicle-treated control in which the geometric mean exceeded 109 copies mL–1 at these time points. These data established the effectiveness of 4b in treatment of EVD in NHPs and accelerated its progress into clinical development. In addition to the efficacy studies, distribution studies in cynomolgus monkeys using [14C]4b at the same effective dose of 10 mg/kg established the presence of drug-related products in the potential sanctuary sites for the virus including testes, epididymis, eyes, and brain.[17] At 4 h postdose, the drug levels in the testes and epididymis exceeded those observed in plasma, and even at 168 h postdose detectable levels of drug-related products were still observed in the testes. Exposure levels in the brain were lower than other tissues, including plasma at 4 h, but were detectable above the plasma levels at 168 h indicating a long half-life of exposure in brain relative to plasma. These data supported the potential that 4b treatment may also reduce persistence of virus in these sanctuary sites.

Safety and pharmacokinetics of 4b administered as once-daily iv infusion were evaluated in single and multiple dose phase 1 clinical trials. No serious adverse effects of the drug were observed. During the course of the phase 1 studies two requests for a compassionate use of 4b were received. The first case involved a healthcare worker who had survived acute infection but had relapsed with symptoms of acute meningoencephalytis.[42] Ebola virus was detected both systemically in plasma and in cerebrospinal fluid. When treated with monoclonal antibodies, the patient developed adverse reaction and was subsequently treated with supportive therapy and 4b for a period of 14 days beginning with a dose of 150 mg and then increasing to 225 mg after two daily infusions. No serious adverse effects related to drug were observed. The patient recovered and cleared the virus from both plasma and CNS although without proper control or natural history data to compare, it is not clear whether the antiviral therapy was effective. The second case involved a newborn infant congenitally infected with EBOV and treated with monoclonal antibodies, blood transfusion, and subsequently 4b.[43] The infant recovered and was eventually declared free of Ebola virus after repeated testing failed to detect viremia. The unprecedented scale of the West African epidemic and the ability to reduce mortality rates through supportive therapy has resulted in many survivors of EVD. The persistence of virus sequelae in survivors in multiple body compartments has now been documented in addition to secondary sexual transmission via virus in genital secretions.[4] This has prompted the initiation of a randomized, blinded, placebo-controlled phase 2 study (PREVAIL IV) which plans to enroll at least 60 adult male survivors to receive either 100 mg of 4b or placebo once-daily over 5 days to assess the effect of 4b therapy on the viral shedding in semen.[44] The results of this study could provide the first evidence as to the potential of 4b to reduce virus replication in humans.

Conclusion

The recent EBOV outbreak prompted the urgent need for antiviral therapeutics for the treat<span class="Species">ment of EVD. We identified a promising nucleotide therapeutic 4b through initial screening and subsequent optimization of the prodrug moiety for iv administration. The partnership with government organizations, including CDC and USAMRIID, that generated the screening data and conducted the rhesus efficacy studies was critical to the successful identification of 4b. Also of importance was the significant chemistry effort that rapidly identified a more efficient route to the parent compound 4 and the ability to prepare, through crystallization of a key reagent, the single Sp phosphorus diastereoisomer 4b for in vivo model studies. The active triphosphate delivered by the prodrug has low micromolar polymerase activity toward EBOV, high selectivity for the viral polymerase compared to host polymerases, and a long intracellular half-life supporting once-daily administration. Parenteral treatment with 4b in EBOV infected NHPs at 10 mg/kg over 12 days demonstrated a substantial antiviral effect along with 100% survival. Based on its promising potential, and preliminary safety data from phase 1 studies, regulatory authorities approved the compassionate use of 4b in two cases including an newborn infant with EVD. Further clinical data on 4b is being collected in the phase 2 PREVAIL IV study that aims to assess the ability of 4b to reduce persistence of EBOV in sanctuary sites of survivors.

Experimental Section

All organic compounds were synthesized at Gilead Sciences, Inc. (Foster City, CA, USA) unless otherwise noted. Commercially available solvents and reagents were used as received without further purification. Nuclear magnetic resonance (NMR) spectra were recorded on a Varian Mercury Plus 400 MHz instrument at room temperature, with tetramethylsilane as an internal standard. Proton nuclear magnetic resonance spectra are reported in parts per million (ppm) on the δ scale and are referenced from the residual protium in the NMR solvent (chloroform-d1, δ 7.26; methanol-d4, δ 3.31; DMSO-d6, δ 2.50). Data are reported as follows: chemical shift [multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, p = pentet, sep = septet, m = multiplet, br = broad, app = apparent); coupling constants (J) in hertz; integration. Carbon-13 nuclear magnetic resonance spectra are reported in parts per million on the δ scale and are referenced from the carbon resonances of the solvent (chloroform-d1, δ 77.16, methanol-d4, δ 49.15; DMSO-d6, δ 39.52). Data are reported as follows: chemical shift. No special nomenclature is used for equivalent carbons. Phosphorus-31 nuclear magnetic resonance spectra are reported in parts per million on the δ scale. Data are reported as follows: chemical shift [multiplicity (s = singlet, d = doublet, t = triplet); coupling constants (J) in hertz. No special nomenclature is used for equivalent phosphorus resonances. Analytical thin-layer chromatography was performed using Merck KGaA silica gel 60 F254 glass plates with UV visualization. Preparative normal phase silica gel chromatography was carried out using a Teledyne ISCO CombiFlash Companion instrument with silica gel cartridges. Purities of the final compounds were determined by high-performance liquid chromatography (HPLC) and were greater than 95% unless otherwise noted. HPLC conditions to assess purity were as follows: Agilent 1100 Series HPLC, Phenominex Kinetex C18; 2.6 μm,100 Å, 100 × 4.6 mm2 column, 2–98% gradient of 0.1% trifluoroacetic acid in water, and 0.1% trifluoroacetic acid in acetonitrile; flow rate, 1.5 mL/min; acquisition time, 8.5 min; wavelength, UV 214 and 254 nm. High-resolution mass spectrometry (HRMS) was performed on an Agilent model 6230 accurate mass time of flight mass spectrometer featuring Agilent Jet Stream Thermal Focusing Technology, with an Agilent 1200 Rapid Resolution HPLC. HRMS chromatography was performed using an Agilent Zorbax Eclipse Plus C18 RRHD 1.8 μm, 2.1 × 50 mm2 column at 30 °C, with a 10–90% gradient of 0.05% trifluoroacetic acid in water and 0.05% trifluoroacetic acid in acetonitrile. LC-MS (MS) was conducted on a Thermo Finnigan MSQ Std. using electrospray positive and negative [M + 1]+ and [M – 1]−, and a Dionex Summit HPLC System (model P680A HPG) equipped with a Gemini 5 μ C18 110A column (30 mm × 4.60 mm), eluting with 0.05% formic acid in 1% acetonitrile/water and 0.05% formic acid in 99% acetonitrile/water. Optical rotations were recorded on a Jasco P-2000 polarimeter.

The synthesis, characterization data, and associated references for the following compounds are provided in the Supporting Information: 4, 4a, 7–11, 9a, 11a, 12–13, 13a, 13b, 4tp, <span class="Chemical">13tp, 16–17, 18a, 19, 21, and 22a–n.

(S)-2-Ethylbutyl 2-(((S)-(((2R,3S,4R,5R)-5-(4-Aminopyrrolo[2,1-f][1,2,4]triazin-7-yl)-5-cyano-3,4-dihydroxytetrahydrofuran-2-yl)methoxy)(phenoxy)phosphoryl)amino)propanoate (4b)

Compound 4b was prepared from 4 and 22b as described previously.[17]1H NMR (400 MHz, methanol-d4): δ 7.86 (s, 1H), 7.33–7.26 (m, 2H), 7.21–7.12 (m, 3H), 6.91 (d, J = 4.6 Hz, 1H), 6.87 (d, J = 4.6 Hz, 1H), 4.79 (d, J = 5.4 Hz, 1H), 4.43–4.34 (m, 2H), 4.28 (ddd, J = 10.3, 5.9, 4.2 Hz, 1H), 4.17 (t, J = 5.6 Hz, 1H), 4.02 (dd, J = 10.9, 5.8 Hz, 1H), 3.96–3.85 (m, 2H), 1.49–1.41 (m, 1H), 1.35–1.27 (m, 8H), 0.85 (t, J = 7.4 Hz, 6H). 13C NMR (100 MHz, methanol-d4): δ 174.98, 174.92, 157.18, 152.14, 152.07, 148.27, 130.68, 126.04, 125.51, 121.33, 121.28, 117.90, 117.58, 112.29, 102.60, 84.31, 84.22, 81.26, 75.63, 71.63, 68.10, 67.17, 67.12, 51.46, 41.65, 24.19, 20.56, 20.50, 11.33, 11.28. 31P NMR (162 MHz, methanol-d4): δ 3.66 (s). HRMS (m/z): [M]+ calcd for C27H35N6O8P, 602.2254; found, 602.2274. [α]21D – 21 (c 1.0, MeOH).

(S)-2-Ethylbutyl 2-(((R)-(((2R,3S,4R,5R)-5-(4-Aminopyrrolo[2,1-f][1,2,4]triazin-7-yl)-5-cyano-3,4-dihydroxytetrahydrofuran-2-yl)methoxy)(phenoxy)phosphoryl)amino)propanoate (4c)

Compound 4c was prepared from 4a by chiral chromatography. Compound 4a was dissolved in acetonitrile. The resulting solution was loaded onto a Lux Cellulose-2 chiral column, equilibrated in acetonitrile, and eluted with isocratic acetonitrile/methanol (95:5 (v/v)). The first eluting compound was the Rp diastereomer 4c, and the second eluting compound was the Sp diastereomer 4b.1H NMR (400 MHz, methanol-d4): δ 8.05 (s, 1H), 7.36 (d, J = 4.8 Hz, 1H), 7.29 (br t, J = 7.8 Hz, 2H), 7.19–7.13 (m, 3H), 7.11 (d, J = 4.8 Hz, 1H), 4.73 (d, J = 5.2 Hz, 1H), 4.48–4.38 (m, 2H), 4.37–4.28 (m, 1H), 4.17 (t, J = 5.6 Hz, 1H), 4.08–3.94 (m, 2H), 3.94–3.80 (m, 1H), 1.48 (sep, J = 12.0, 6.1 Hz, 1H), 1.34 (p, J = 7.3 Hz, 4H), 1.29 (d, J = 7.2 Hz, 3H), 0.87 (t, J = 7.4 Hz, 6H). 31P NMR (162 MHz, methanol-d4): δ 3.71 (s). MS m/z: 603.1 [M + 1]. [α]21D – 46 (c 1.0, MeOH).

(2S)-Ethyl-2-(((((2R,3S,4R,5R)-5-(4-aminopyrrolo[2,1-f][1,2,4]triazin-7-yl)-5-cyano-3,4-dihydroxytetrahydrofuran-2-yl)methoxy)(phenoxy)phosphoryl)amino)-3-phenylpropanoate (4d)

Compound 4 (0.030 g, 0.103 mmol) was dissolved in DMF (1 mL), and then THF (0.5 mL) was added. t-BuMgCl (1 M/THF, 154.5 μL, 0.154 μmol) was added to the reaction in a dropwise manner with vigorous stirring. The resulting white slurry was stirred at rt for about 30 min. A solution of compound 22d (0.058 g, 0.124 mmol) in THF (1 mL) was added in a dropwise manner to the reaction at rt. The reaction progress was monitored by LC-MS. When the reaction progressed to 50% conversion, the reaction was cooled in an ice bath and quenched with glacial acetic acid (70 μL). The reaction was concentrated and the crude residue was purified by reverse phase preparatory HPLC to afford compound 4d (22 mg, 34%, as a 2.6:1 mixture of diastereomers at phosphorus). 1H NMR (400 MHz, DMSO-d6): δ 7.91 (d, J = 4 Hz, 1H), 7.90 (br s, 2H), 7.09–7.30 (m, 8H), 7.01, (t, J = 8.2 Hz, 2H), 6.89 (d, J = 4.4 Hz, 1H), 6.82 (t, J = 4.4 Hz, 1H), 6.27 (m, 1H), 6.14 (m, 1H), 5.34 (m, 1H), 4.62 (t, J = 5.6 Hz, 1H), 4.15 (m, 1H), 4.01−3.78 (m, 6H), 2.92 (m, 1H), 2.78 (m, 1H), 1.04 (m, 3H). 31P NMR (162 MHz, DMSO-d6) δ 3.69 (s), 3.34 (s). MS m/z = 623.0 [M+1].

(2S)-Ethyl 2-(((((2R,3S,4R,5R)-5-(4-Aminopyrrolo[2,1-f][1,2,4]triazin-7-yl)-5-cyano-3,4-dihydroxytetrahydrofuran-2-yl)methoxy)(phenoxy)phosphoryl)amino)-3-methylbutanoate (4e)

Compound 4 (0.040 g, 0.14 mmol) was dissolved in NMP (1.5 mL) and then THF (0.25 mL) was added. This solution was cooled in an ice bath and t-BuMgCl (1M/THF, 425.7 μL, 0.426 μmol) was added in a dropwise manner with vigorous stirring. The ice bath was removed, and the resulting white slurry was stirred at rt for about 15 min. A solution of compound 22e (0.081 g, 0.192 mmol) in THF (0.5 mL) was added in a dropwise manner to the reaction at rt. The reaction progress was monitored by LC-MS. When the reaction progressed to 50% conversion, the reaction was cooled in an ice bath and quenched with glacial acetic acid (70 μL). The reaction was concentrated and crude residue was semipurified from the residue by reverse phase HPLC. The semipure material was further purified by silica gel column chromatography (eluent, 100% EtOAc ramping to 10% MeOH in EtOAc) to afford compound 4e (0.034 g, 43% as a 1.8:1 mixture of diastereomers). 1H NMR (400 MHz, DMSO-d6): δ 7.91 (d, J = 1.6 Hz, 1H), 7.88 (br s, 2H), 7.32 (m, 2H), 7.15 (m, 3H), 6.90 (t, J = 4.2 Hz, 1H), 6.84 (d, J = 4.8 Hz, 1H), 6.26 (dd, J = 13.4, 6.2 Hz, 1H), 5.87 (q, J = 11.2 Hz, 1H), 5.35 (m, 1H), 4.64 (m, 1H), 4.25 (m, 2H), 4.15−3.93 (m, 4H), 3.45 (m, 1H), 1.87 (m, 1H), 1.16−1.09 (m, 3H), 0.83−0.70 (m, 6H). 31P NMR (162 MHz, DMSO-d6): δ 4.59 (s), 4.47 (s). MS m/z = 575.02 [M + 1].

Ethyl 2-(((((2R,3S,4R,5R)-5-(4-Aminopyrrolo[2,1-f][1,2,4]triazin-7-yl)-5-cyano-3,4-dihydroxytetrahydrofuran-2-yl)methoxy)(phenoxy)phosphoryl)amino)-2-methylpropanoate (4f)

Compound 4 (66 mg, 0.23 mmol) was dissolved in NMP (2.0 mL) and the mixture was cooled to about 0 °C. t-BuMgCl (1.0 M in THF, 0.34 mL, 0.34 mmol) was then added and the resulting mixture was stirred at 0 °C for about 30 min. A solution of compound 22f (139 mg, 0.34 mmol) in THF (1.0 mL) was then added, and the reaction mixture was heated to about 50 °C. After about 2 h, the reaction was cooled to rt and quenched with acetic acid and methanol. The resulting mixture was concentrated under reduced pressure and purified by preparatory reverse phase HPLC to afford compound 4f (32 mg, 25% as a 1.5:1 mixture of diastereomers). 1H NMR (400 MHz, DMSO-d6): δ 7.89 (m, 3H), 7.31 (q, J = 8.1 Hz, 2H), 7.22–7.05 (m, 3H), 6.87 (d, J = 4.5, 1H), 6.80 (d, J = 4.5 Hz, 1H), 6.27 (d, J = 11.7, 1H), 5.81 (d, J = 9.7, 1H), 5.35 (d, J = 5.6 Hz, 1H), 4.64 (dt, J = 9.0, 5.6 Hz, 1H), 4.24 (m, 2H), 4.11 (m, 1H), 4.04–3.90 (m, 3H), 1.39–1.23 (m, 6H), 1.10 (t, J = 7.1, 3H). 31P NMR (162 MHz, DMSO-d6): δ 2.45, 2.41. MS m/z = 561.03 [M + 1].

Ethyl ((((2R,3S,4R,5R)-5-(4-Aminopyrrolo[2,1-f][1,2,4]triazin-7-yl)-5-cyano-3,4-dihydroxytetrahydrofuran-2-yl)methoxy)(phenoxy)phosphoryl)-l-alaninate (4g)

Compound 4 (50 mg, 0.17 mmol) was dissolved in NMP–THF (1:1 mL)) and cooled with an ice bath. t-BuMgCl (1.0 M in THF, 0.257 mL, 0.257 mmol) was then added over about 5 min. The resulting mixture was allowed to warm to rt and was stirred for about 30 min. Then a solution of compound 22g (74.6 mg, 0.189 mmol) in THF (2 mL) was added. After about 30 min, the reaction mixture was purified by preparatory reverse phase HPLC. Fractions containing the desired product were further purified with silica gel chromatography (eluent: 0–20% methanol in dichloromethane) to afford compound 4g (23 mg, 24% as a 2.5:1 mixture of diastereomers). 1H NMR (400 MHz, methanol-d4): δ 7.76 (d, J = 6.0 Hz, 1H), 7.25–7.14 (m, 2H), 7.11–6.99 (m, 3H), 6.87–6.72 (m, 2H), 4.70 (d, J = 5.4 Hz, 1H), 4.39–4.24 (m, 2H), 4.20 (dddd, J = 9.7, 7.9, 5.1, 2.8 Hz, 1H), 4.10 (dt, J = 12.8, 5.5 Hz, 1H), 4.06–3.91 (m, 2H), 3.72 (ddq, J = 14.3, 9.3, 7.1 Hz, 1H), 1.17 (dd, J = 7.1, 1.0 Hz, 1H), 1.14–1.06 (m, 5H). 31P NMR (162 MHz, methanol-d4): δ 3.73, 3.68. MS m/z = 547 [M + 1].

(2S)-Cyclobutyl 2-(((((2R,3S,4R,5R)-5-(4-Aminopyrrolo[2,1-f][1,2,4]triazin-7-yl)-5-cyano-3,4-dihydroxytetrahydrofuran-2-yl)methoxy)(phenoxy)phosphoryl)amino)propanoate (4h)

Compound 4 (58 mg, 0.20 mmol) was mixed with compound 22h (101 mg, 0.240 mmol) in 2 mL of anhydrous DMF. Magnesium chloride (42 mg, 0.44 mmol) was added in one portion. The reaction mixture was heated to about 50 °C. N,N-Diisopropylethylamine (87 μL, 0.5 mmol) was added, and the reaction was stirred for about 2 h at about 50 °C. The reaction mixture was cooled to room temperature, was diluted with ethyl acetate, and was washed with 5% aqueous citric acid solution followed by saturated aqueous sodium chloride solution. The organic layer was then dried over anhydrous sodium sulfate and concentrated under reduced pressure. The crude residue was purified with silica gel column (eluent, 0–5% methanol in dichloromethane) to afford compound 4h (42 mg, 37% yield, as a 3:2 mixture of diastereomers). 1H NMR (400 MHz, methanol-d4): δ 7.85 (m, 1H), 7.34–7.22 (m, 2H), 7.22–7.08 (m, 3H), 6.94–6.84 (m, 2H), 4.95–4.85 (m, 1H), 4.79 (m, 1H), 4.46–4.34 (m, 2H), 4.34–4.24 (m, 1H), 4.19 (m, 1H), 3.81 (m, 1H), 2.27 (m, 2H), 2.01 (m, 2H), 1.84–1.68 (m, 1H), 1.62 (m, 1H), 1.30–1.16 (m, 3H). 31P NMR (162 MHz, methanol-d4): δ 3.70, 3.65. MS m/z = 573.0 [M + 1].

(S)-Isopropyl 2-(((R)-(((2R,3S,4R,5R)-5-(4-Aminopyrrolo[2,1-f][1,2,4]triazin-7-yl)-5-cyano-3,4-dihydroxytetrahydrofuran-2-yl)methoxy)(phenoxy)phosphoryl)amino)propanoate (4i)

Compound 4 (60.0 mg, 206 μmol) was dissolved in NMP (0.28 mL). THF (0.2 mL) was added followed by tert-butyl magnesium chloride (1.0 M solution in tetrahydrofuran, 0.309 mL) at rt under an argon atmosphere. After 20 min, a solution of compound 22i (81 mg, 206 μmol) in THF (0.2 mL) was added, and the resulting mixture was warmed to about 50 °C. After 3 h, the reaction mixture was allowed to cool to rt and was purified directly by preparatory HPLC to afford compound 4i (44 mg, 38% as a single diastereomer). 1H NMR (400 MHz, methanol-d4): δ 7.86 (s, 1H), 7.34–7.26 (m, 2H), 7.21–7.12 (m, 3H), 6.91 (d, J = 4.6 Hz, 1H), 6.87 (d, J = 4.6 Hz, 1H), 4.92 (sep, J = 6.3 Hz, 1H), 4.80 (d, J = 5.4 Hz, 1H), 4.43–4.34 (m, 1H), 4.33–4.24 (m, 1H), 4.18 (t, J = 5.6 Hz, 1H), 3.82 (dq, J = 9.7, 7.1 Hz, 2H), 1.27 (dd, J = 7.1, 1.0 Hz, 3H), 1.18 (dd, J = 6.3, 4.8 Hz, 6H). 31P NMR (162 MHz, methanol-d4): δ 3.72 (s). MS m/z = 561.11 [M + 1].

tert-Butyl ((((2R,3S,4R,5R)-5-(4-Aminopyrrolo[2,1-f][1,2,4]triazin-7-yl)-5-cyano-3,4-dihydroxytetrahydrofuran-2-yl)methoxy)(phenoxy)phosphoryl)-l-alaninate (4j)

To a mixture of intermediate 4 (80 mg, 0.28 mmol), intermediate 22j (174 mg, 0.41 mmol), and MgCl2 (39 mg, 0.41 mmol) in DMF (4 mL) was added N,N-diisopropylethylamine (0.12 mL, 0.69 mmol) dropwise at room temperature. The reaction mixture was stirred at 50 °C for 1 h and was cooled to rt. The resulting mixture was concentrated under reduced pressure to approximately 2 mL volume and was purified by reverse phase preparative HPLC. Fractions containing the desired product were combined and further purified by silica gel column chromatography (eluent, 0–20% methanol in methylene chloride) to afford compound 4j (51 mg, 32%, 1.5:1 diastereomeric mixture). 1H NMR (400 MHz, methanol-d4): δ 7.86 (s, 0.4H), 7.84 (s, 0.6H), 7.28 (m, 2H), 7.21–7.10 (m, 3H), 6.96–6.83 (m, 2H), 4.79 (m, 1H), 4.46–4.34 (m, 2H), 4.28 (m, 1H), 4.22–4.13 (m, 1H), 3.81–3.64 (m, 1H), 1.40 (m, 9H), 1.22 (m, 3H). 31P NMR (162 MHz, methanol-d4): δ 3.79 (s). MS m/z = 575 [M + 1].

(2S)-Cyclopentyl 2-(((((2R,3S,4R,5R)-5-(4-Aminopyrrolo[2,1-f][1,2,4]triazin-7-yl)-5-cyano-3,4-dihydroxytetrahydrofuran-2-yl)methoxy)(phenoxy)phosphoryl)amino)propanoate (4k)

Compound 4 (100 mg, 0.34 mmol) was dissolved in THF (2 mL) and cooled under an ice water bath. Then 1 M t-BuMgCl (0.52 mL, 0.77 mmol) was added dropwise slowly. The resulting mixture was stirred for about 30 min at rt. Then compound 22k (247 mg, 0.52 mmol) in THF (2 mL) was added over about 5 min and the resulting mixture was stirred for about 24 h at rt. The resulting mixture was diluted with ethyl acetate, cooled under ice–water bath, treated with aqueous NaHCO3 (2 mL), washed with brine, dried with sodium sulfate, and concentrated under reduced pressure. The resulting mixture was purified by silica gel column chromatography (eluent, 0–20% methanol in dichloromethane) followed by reverse phase preparatory HPLC to afford compound 4k (47 mg, 23% as a 27:1 mixture of diastereomers). 1H NMR (400 MHz, methanol-d4): δ 7.85 (s, 1H), 7.33–7.22 (m, 2H), 7.14 (tdd, J = 7.6, 2.1, 1.1 Hz, 3H), 6.95–6.87 (m, 2H), 5.13–5.00 (m, 1H), 4.78 (d, J = 5.4 Hz, 1H), 4.48–4.35 (m, 2H), 4.30 (ddd, J = 10.6, 5.7, 3.6 Hz, 1H), 4.19 (t, J = 5.4 Hz, 1H), 3.78 (dq, J = 9.2, 7.1 Hz, 1H), 1.81 (dtd, J = 12.5, 5.9, 2.4 Hz, 2H), 1.74–1.49 (m, 6H), 1.21 (dd, J = 7.1, 1.2 Hz, 3H). MS m/z = 587 [M + 1].

Pentan-3-yl ((((2R,3S,4R,5R)-5-(4-Aminopyrrolo[2,1-f][1,2,4]triazin-7-yl)-5-cyano-3,4-dihydroxytetrahydrofuran-2-yl)methoxy)(phenoxy)phosphoryl)-l-alaninate (4l)

To a mixture of compound 4 (80 mg, 0.28 mmol), 22l (170 mg, 0.39 mmol), and MgCl2 (39 mg, 0.41 mmol) in DMF (4 mL) was added N,N-diisopropylethylamine (0.12 mL, 0.69 mmol) dropwise at rt. The resulting mixture was stirred at 50 °C for 1 h, concentrated to approximately 2 mL volume, and purified by reverse phase preparative HPLC to afford compound 4l (68 mg, 42%, 1.5:1 diastereomeric mixture). 1H NMR (400 MHz, methanol-d4): δ 7.86 (s, 0.4 H), 7.85 (s, 0.6 H), 7.33–7.23 (m, 2H), 7.21–7.08 (m, 3H), 6.95–6.84 (m, 2H), 4.79 (m, 1H), 4.69 (m, 1H), 4.47–4.34 (m, 2H), 4.34–4.24 (m, 1H), 4.19 (m, 1H), 3.85 (m, 1H), 1.64–1.42 (m, 4H), 1.29 (dd, J = 7.0, 1.1 Hz, 1.1 H), 1.23 (dd, J = 7.2, 1.3 Hz, 1.9H), 0.91–0.76 (m, 6H). 31P NMR (162 MHz, methanol-d4): δ 3.71, 3.69. MS m/z = 589 [M + 1].

(S)-Neopentyl 2-(((S)-(((2R,3S,4R,5R)-5-(4-Aminopyrrolo[2,1-f][1,2,4]triazin-7-yl)-5-cyano-3,4-dihydroxytetrahydrofuran-2-yl)methoxy)(phenoxy)phosphoryl)amino)propanoate (4m)

Compound 4 (100 mg, 0.34 mmol) was dissolved in THF (2 mL) and cooled under ice water bath. Then 1 M t-BuMgCl (0.52 mL, 0.77 mmol) was added dropwise slowly. The resulting mixture was stirred for 30 min at room temperature. Then compound 22m (248 mg, 0.52 mmol) was added over 5 min, and the resulting mixture was stirred for 24h at room temperature, diluted with EtOAc, cooled under ice–water bath, treated with aqueous NaHCO3 (2 mL), washed with brine, dried with sodium sulfate, and concentrated in vacuo. The resulting mixture was purified by silica gel column chromatography (MeOH 0 to 20% in DCM) and prep-HPLC (acetonitrile 10 to 80% in water) to give compound 4m (12 mg, 10% as a single diastereomer). 1H NMR (400 MHz, methanol-d4): δ 7.86 (s, 1H), 7.36–7.24 (m, 2H), 7.23–7.10 (m, 3H), 6.96–6.85 (m, 2H), 4.78 (d, J = 5.4 Hz, 1H), 4.38 (tdd, J = 10.0, 4.9, 2.5 Hz, 2H), 4.32–4.24 (m, 1H), 4.17 (t, J = 5.6 Hz, 1H), 3.91 (dq, J = 9.8, 7.1 Hz, 1H), 3.81 (d, J = 10.5 Hz, 1H), 3.69 (d, J = 10.5 Hz, 1H), 1.31 (dd, J = 7.2, 1.1 Hz, 3H), 0.89 (s, 9H). MS m/z = 589 [M + 1]+

2-Ethylbutyl ((((2R,3S,4R,5R)-5-(4-Aminopyrrolo[2,1-f][1,2,4]triazin-7-yl)-5-cyano-3,4-dihydroxytetrahydrofuran-2-yl)methoxy)(phenoxy)phosphoryl)-d-alaninate (4n)

Compound 21 (50 mg, 0.15 mmol) was dissolved in anhydrous tetrahydrofuran (5 mL) and stirred under atmos<span class="Chemical">pheric argon. Compound 22n (75 mg, 0.17 mmol) was added followed by magnesium chloride (21 mg, 0.23 mmol), and the reaction was warmed to 50 °C and stirred for 30 min. N,N-diisopropylethylamine (65.0 μL, 0.375 mmol) was added dropwise, and the reaction mixture was stirred for 3 h at 50 °C. The reaction mixture was then cooled in an ice bath, and 12 N HCl(aq) (175 μL) was added dropwise. The ice bath was removed, and the reaction mixture was stirred at rt for 4 h. The reaction mixture was diluted with ethyl acetate (15 mL) and cooled in an ice bath. Aqueous 1 N NaOH solution was added slowly to give pH of 10. The organic layer was then washed with 5% aqueous sodium carbonate solution and then saturated aqueous sodium chloride solution. The organic layer was then dried over anhydrous sodium sulfate and concentrated under reduced pressure. The crude residue was purified with silica gel column chromatography (eluent, 0–10% methanol in dichloromethane) to afford compound 4n (60 mg, 66% yield as a 1.1:1 diastereomeric mixture). 1H NMR (400 MHz, methanol-d4): δ 7.87–7.83 (m, 1H), 7.37–7.22 (m, 2H), 7.22–7.04 (m, 3H), 6.96–6.79 (m, 2H), 4.82–4.75 (m, 1H), 4.45–4.23 (m, 3H), 4.18 (m, 1H), 4.06–3.85 (m, 3H), 1.52–1.38 (m, 1H), 1.38–1.24 (m, 7H), 0.85 (m, 6H). 31P NMR (162 MHz, methanol-d4): δ 3.87, 3.55. MS m/z = 603.1 [M + 1].

(2S,2′S)-Diethyl 2,2′-((((2R,3S,4R,5R)-5-(4-Aminopyrrolo[1,2-f][1,2,4]triazin-7-yl)-5-cyano-3,4-dihydroxytetrahydrofuran-2-yl)methoxy)phosphoryl)bis(azanediyl)dipropanoate (4o)

Compound 4 (14.6 mg, 0.05 mmol) was dissolved in anhydrous trimethyl phosphate (0.5 mL) and stirred under N2(g) at rt. POCl3 (9.2 μL, 0.1 mmol) was added, and the mixture stirred for about 60 min. Alanine ethyl ester hydrochloride 18b (61 mg, 0.4 mmol, Aldrich, CAS No. 1115-59-9), and then Et3N (70 μL, 0.5 mmol) was added. The resultant mixture was stirred for about 15 min, and then additional Et3N (70 μL, 0.5 mmol) was added to give a solution pH of 9–10. The mixture was stirred for about 2 h and then diluted with ethyl acetate, washed with saturated aqueous NaHCO3 solution, followed by saturated aqueous NaCl solution. The organic layer was dried over anhydrous sodium sulfate and concentrated under reduced pressure. The residue was purified by reverse phase preparative HPLC to afford compound 4o (5.5 mg, 16%). 1H NMR (400 MHz, methanol-d4): δ 8.13 (s, 1H), 7.41 (d, J = 4.8 Hz, 1H), 7.18 (d, J = 4.8 Hz, 1H), 4.78 (d, J = 5.6 Hz, 1H), 4.36 (m, 1H), 4.25–4.08 (m, 7H), 3.83 (m, 2H), 1.33–1.23 (m, 12H). 31P NMR (162 MHz, methanol-d4): δ 13.8. MS m/z = 570.0 [M + 1].

(2S,2′S)-Bis(2-ethylbutyl) 2,2′-((((2R,3S,4R,5R)-5-(4-Aminopyrrolo[1,2-f][1,2,4]triazin-7-yl)-5-cyano-3,4-dihydroxytetrahydrofuran-2-yl)methoxy)phosphoryl)bis(azanediyl)dipropanoate (4p)

To a suspension of compound 4 (52 mg, 0.18 mmol) and solid sodium bicarbonate (53 mg) in trimethyl phosphate (1.5 mL) at 0 °C was added POCl3 (120 mg, 0.783 mmol). The mixture was stirred at 0 °C for 3 h, at which point a solution of 18a (790 mg, 4.56 mmol) in MeCN (1 mL) was then added. The reaction mixture was stirred at 0 °C for 0.5 h, then triethylamine (0.1 mL) was added and stirred at rt for 0.5 h. The reaction mixture was diluted with ethyl acetate (10 mL), washed with water (10 mL), and concentrated under reduced pressure. The residue was purified by silica gel column chromatography (eluent, 50–100% ethyl acetate in hexanes gradient followed by 0–10% methanol in ethyl acetate gradient) to afford compound 4p (71 mg, 58%). 1H NMR (400 MHz, methanol-d4): δ 7.88 (s, 1H), 6.95 (d, J = 4.8 Hz, 1H), 6.89 (d, J = 4.4 Hz, 1H), 4.85 (d, J = 5.6 Hz, 1H), 4.35−4.32 (m, 1H), 4.26−4.12 (m, 3H), 4.09−4.04 (m, 2H), 3.98−3.94 (m, 2H), 3.89−3.79 (m, 2H), 1.54−1.44 (m, 2H), 1.39−1.27 (m, 14H), 0.89 (t, J = 7.2 Hz, 12H). 31P NMR (162 MHz, methanol-d4): δ 13.83. MS m/z = 682.1 [M + 1].