The Influenza Virus Polymerase Complex: An Update on Its Structure, Functions, and Significance for Antiviral Drug Design.

Influenza viruses cause seasonal epidemics and pandemic outbreaks associated with significant morbidity and mortality, and a huge cost. Since resistance to the existing anti-influenza drugs is rising, innovative inhibitors with a different mode of action are urgently needed. The influenza polymerase complex is widely recognized as a key drug target, given its critical role in virus replication and high degree of conservation among influenza A (of human or zoonotic origin) and B viruses. We here review the major progress that has been made in recent years in unravelling the structure and functions of this protein complex, enabling structure-aided drug design toward the core regions of the PA endonuclease, PB1 polymerase, or cap-binding PB2 subunit. Alternatively, inhibitors may target a protein-protein interaction site, a cellular factor involved in viral RNA synthesis, the viral RNA itself, or the nucleoprotein component of the viral ribonucleoprotein. The latest advances made for these diverse pharmacological targets have yielded agents in advanced (i.e., favipiravir and VX-787) or early clinical testing, besides several experimental inhibitors in various stages of development, which are all covered here.

INTRODUCTION

The influenza A, B, C, and D viruses belong to the Orthomyxoviridae, a family of enveloped viruses with a single‐stranded negative‐sense, and segmented RNA genome. Seasonal influenza A and B viruses affect each year approximately 5–10% of the adult and 20–30% of the pediatric population. In addition, there is permanent concern for sudden influenza pandemics, such as that of 2009 or the notorious “Spanish flu” of 1918.

A universal influenza vaccine that confers broad and long‐term protection remains the “Holy Grail” in influenza research,1 as vaccination is considered as the most effective way to prevent influenza virus infections. The existing tri‐ or quadrivalent vaccines contain the circulating A/H1N1, A/H3N2, and (one or two) B strains. They fail to provide broadly protective and long‐lasting immunity, require annual updating, and are only partially protective in some target populations.2, 3

For prevention or treatment of influenza virus infections, two classes of antiviral drugs are currently available. The M2‐blockers, amantadine and rimantadine, inhibit influenza A virus replication by occluding the M2 proton channel,4 but lack activity against influenza B virus.5 In addition, central nervous system side effects6 and the global spread of adamantane‐resistant influenza viruses7 undermine their clinical usefulness. The second group, the neuraminidase inhibitors (NAIs), prevents virus release from infected cells. Zanamivir requires administration by powder inhalation, while oseltamivir is administered orally as its prodrug oseltamivir phosphate. In the period 2007–2009, oseltamivir‐resistant influenza H1N1 viruses were disseminated worldwide.8 In contrast, resistance to zanamivir seems to be rare.9, 10 The search for novel M2‐inhibitors and NAIs with improved resistance profile and higher potency is still ongoing, as is further optimization of original lead compounds.11

In conclusion, the armamentarium to prevent and treat influenza infections is currently limited, and new influenza virus inhibitors with an entirely different mode of action are urgently required. In this review, we focus on the influenza virus polymerase that is widely recognized as a superior antiviral drug target,12 given its critical role in virus replication and high degree of sequence conservation in influenza A and B, particularly in the active sites for RNA binding, cleavage, or elongation. Before explaining the different antiviral strategies that are being explored, we provide relevant background on structure and functions of the viral polymerase. During recent years, major breakthroughs have been made in revealing the structure of this protein complex, its subunits, or the associated viral nucleoprotein (NP), creating unique opportunities for structure‐aided drug design.

STRUCTURE AND FUNCTIONS OF THE INFLUENZA VIRUS POLYMERASE COMPLEX

To avoid overlap with other recent reviews on this topic,13, 14, 15, 16 this section is limited to provide an update on new findings with direct or indirect relevance for inhibitor design. The influenza virus polymerase complex (schematically depicted in Fig. 1) is composed of three subunits: PB1, PB2 (polymerase basic protein‐1 and ‐2, respectively) and PA (polymerase acidic protein; referred to as P3 in the case of influenza C). The influenza A and B virions contain eight viral ribonucleoprotein (vRNP) segments, which have a double‐helical hairpin structure and each carry one polymerase heterotrimer. These vRNPs contain vRNA that is coated by viral NP molecules. Viral RNA synthesis occurs in the cell nucleus in two stages (Fig. 1). (i) For primary transcription of vRNA to mRNA, primers are generated by the PA‐dependent “cap‐snatching” reaction on cellular capped RNAs. The PB2 subunit initially binds the capped RNAs, and actual RNA synthesis is performed by PB1. (ii) Replication of vRNA proceeds via complementary RNA (cRNA) intermediates and occurs in a primer‐independent (de novo) manner. The nascent vRNA and cRNA replicates are immediately packaged into new vRNPs or cRNPs, respectively. The newly synthesized vRNP complexes are exported from the nucleus to the cytoplasm, for which different pathways have been proposed (reviewed in Eisfeld et al.14).

Figure 1

Role of the influenza virus polymerase complex in vRNA transcription and replication.

The influenza polymerase is a complex containing the PA, PB1, and PB2 proteins. One polymerase heterotrimer is attached to each vRNP segment inside the virion. These vRNPs have a double‐helical hairpin structure consisting of two antiparallel vRNA strands that are coated by NP molecules.17, 18

Transcription of vRNA to mRNA starts with the “cap‐snatching” reaction, in which cellular capped RNAs are bound by PB2 and cleaved, by PA, at 10–15 nucleotides from the cap, to yield primers for viral mRNA synthesis.19, 20 Termination and polyadenylation occur at a stretch of five to seven U residues near the 5ʹ end of the vRNA.21, 22 Replication of vRNA proceeds through full‐length complementary cRNAs, which are assembled as cRNP complexes. In both transcription and replication, RNA elongation is carried out by the PB1 subunit. The vRNA promoter is depicted as observed in the promoter‐bound polymerase crystal structures, in which its 5ʹ end forms a “hook” structure.23, 24 The cRNA promoter is drawn in dashed lines, since its conformation is yet to be determined. (Adapted from Portela et al.25)

The influenza virus polymerase is a major determinant of virus pathogenicity and host adaptation.26, 27, 28 One notable key factor is the PB2‐627 signature residue, which is glutamic acid in avian viruses and lysine in human viruses, and was recently linked to species differences in the host nuclear protein ANP32A.29

The PB2 Cap‐Binding Domain

The discovery that the cap‐binding activity resides in the PB2 subunit was made in 1981.30, 31 In 2008, Guilligay et al.32 reported the X‐ray structure of the PB2 cap‐binding domain (CBD; i.e., residues 318–483) of influenza A virus, complexed with the minimal cap analogue 7‐methylguanosine 5′‐triphosphate (m7GTP; Fig. 2A). A similar approach for the influenza B PB2‐CBD (Fig. 2B)33 revealed that the binding mode of m7GTP is similar in influenza A and B, indicating a conserved methylated cap recognition mechanism. Similar as in other cap‐binding proteins (such as the eukaryotic initiation factor eIF4E) in which the 7‐methylguanine moiety is sandwiched between two aromatic residues, a cation‐π packing interaction occurs between Phe404 and His357 in the influenza A PB2‐CBD, and Phe406 and Trp359 in the influenza B counterpart. However, the PB2‐CBD has a completely different protein fold compared to other cap‐binding proteins.32 It also contains a remarkable cluster of phenylalanine residues with one noticeable difference (Phe323 vs. Gln325; compare Fig. 2A and B) between the influenza A and B proteins. In two other crystal structures, that is, the influenza B PB2‐CBD in complex with GTP,33 and a Q325F mutant form of this protein in complex with m7GDP,34 an inverted conformation for the guanine and ribose moieties was seen. These data indicate structural flexibility of the influenza B PB2‐CBD, which explains its promiscuous cap recognition. In enzymatic assays with purified vRNPs35 and binding assays with isolated PB2‐CBD,33 influenza A displays strict specificity for m7G‐capped RNA, while the influenza B protein recognizes various cap structures, including unmethylated GpppG‐RNA.

Figure 2

Chemical structures of the minimal cap analogue m7GTP and reported inhibitors of cap‐binding and X‐ray structures of the PB2‐CBD in complex with the corresponding ligands. (A–D) Comparison of the crystal structures of the PB2‐CBD of influenza A32 (A; PDB: 2VQZ) or influenza B33 (B; PDB: 5EFA) PB2‐CBD in complex with m7GTP; and influenza A PB2‐CBD with bound compound “11”36 (C; PDB: 4CB6) or VX‐78737 (D; PDB: 4P1U). (E) Cap‐binding inhibitors Cap‐3 and Cap‐7.38 (F) For RO0794238,39 direct binding to the PB2‐CBD could not be demonstrated.

The PA Endonuclease Domain

Until 2009, the endonuclease activity was thought to reside in PB140 or PB2.41 A major leap forward was made when two groups independently revealed42, 43 that the endonuclease catalytic site resides in the N‐terminal domain of PA (PA‐Nter; residues 1 to ∼195) and published the first crystal structures of the holo form of PA‐Nter (Fig. 3A and B). Its structural core contains a characteristic (P)DXN(D/E)XK motif (Fig. 3) conserved in all influenza viruses, which coordinates one, two, or three manganese or magnesium ions. It is still unresolved which and how many divalent metal ions are present in the native enzyme,42, 43, 44, 45, 46, 47, 48 although this information is critical for efficient inhibitor design. The discrepancies in the crystal structures with regard to the identity of these metal ions (Fig. 3A–F) are possibly related to the crystallization conditions (i.e., soaking vs. cocrystallization or different buffer composition, for instance related to pH or which metal ion was added). More conclusive evidence requires crystal structures obtained in the presence of an RNA substrate. As of today, only cocrystal structures with mononucleotides have been determined.44, 46, 49

Figure 3

Comparison of PA‐Nter active site structures as determined by crystallography. The metal coordinating and/or catalytic residues are labeled. The metal ions and water molecules are shown as purple and red spheres, respectively. (A) The PA‐Nter crystal structure determined by Dias et al.42 (PDB: 2W69) contains two metal ions (two Mn2+ ions or one Mn2+ and one Mg2+). (B) Yuan et al.43 (PDB: 3EBJ) obtained a PA‐Nter crystal with one Mg2+ ion. (C) Zhao et al.44 (PDB: 3HW6) discerned one (Mg2+) metal (dark purple) or two (Mn2+) metal ions, depending on the crystallization conditions. (D) Dubois et al.45 (PDB: 4E5E) identified two Mn2+ ions. (E) Bauman et al.50 (PDB: 4M5Q) described two Mn2+ ions (in dark purple) in the PA‐Nter apo protein, and a third one (light purple) in the presence of a bound inhibitor (not shown; PDB: 4M5U). (F) In the crystal structure published by Tefsen et al.48 (PDB: 4NFZ), one Mn2+ ion is accommodated in the PA‐Nter active site.

In biochemical assays with isolated recombinant PA‐Nter51 or full‐length PA,52 the endonucleolytic activity is higher in the presence of manganese compared to magnesium. Experiments using isothermal titration calorimetry (ITC) showed a 500‐fold higher affinity for Mn2+ ions than for Mg2+ ions.51 This observation is at least partially explainable by the presence of a histidine residue in the PA‐Nter active site, as Mn2+ ions have higher affinity for histidine than Mg2+. However, these ITC measurements were performed in the absence of RNA substrate, which may be a different situation compared to the substrate‐bound enzyme state. Since the intracellular concentration of free Mg2+ is at least 1000‐fold higher than that of Mn2+, magnesium may be more biologically relevant.53, 54 Also, in the context of the heterotrimeric polymerase, efficient cleavage is seen in the presence of the less reactive Mg2+,55, 56 which can be attributed to tight RNA binding by the nearby PB2‐CBD.

In order to clarify this issue, Xiao and co‐workers performed molecular dynamics (MD) simulations to construct structural models of PA‐Nter in the presence of one or two magnesium ions and in complex with the RNA substrate.57 In Figure 4, both active site conformations and possible one‐ and two‐metal‐ion reaction mechanisms are compared and explained in full detail in the figure legend.

Figure 4

Comparison of the two‐ versus one‐metal‐ion models for the PA‐Nter active site. The metal ions are depicted as orange spheres, while the red spheres represent the water molecules coordinated to the metal ions. (A and B) Structural models obtained by molecular dynamics (MD) simulations of the PA‐Nter active site in complex with its RNA substrate, in the presence of two (A) or one Mg2+ ion(s) (B) [reprinted with permission from Xiao et al.57 Copyright 2014 American Chemical Society]. (C and D) Possible reaction mechanism of the PA‐Nter cleavage reaction, in the presence of two (C) or one (D) metal ion(s), based on the models of Xiao et al.57 In each case, the metal ion has six coordinated ligands (indicated by dotted lines). (C) In the two‐metal mechanism, both metal ions are coordinated to the scissile phosphodiester of the nucleic acid substrate, and the ribose 3ʹ‐O is coordinated apically to MB 2+. MA 2+ would lower the pK of the attacking water molecule, together with the catalytic Lys134 and the adjacent 3ʹ phosphodiester, thus activating this water molecule as a nucleophile. MA 2+ might also assist the nucleophilic attack by moving toward MB 2+, bringing the activated water molecule closer to the scissile phosphodiester. MB 2+ can stabilize the pentacovalent intermediate formed during the SN2‐type phosphodiester bond cleavage. (D) When the one‐metal mechanism is operative, this metal ion is coordinated to the scissile phosphodiester. Its role would be to stabilize the transition state of the nucleophilic attack, like MB 2+ in the two‐metal mechanism. Additionally, the single metal ion could decrease the pK of the nucleophilic water, together with the catalytic Lys137, while His41 would act as the general base that deprotonates the water molecule.

The two‐metal‐ion model is favored by the ITC data for Mn2+ binding to PA‐Nter.51 Doan et al.55 found that the endonuclease activity in vRNP complexes isolated from virions depends on metal ion concentration in a cooperative manner, with Hill coefficients close to or larger than 2. Also, synergistic activation of the cleavage activity was observed with combinations of different metal ions. This suggests that PA‐Nter requires two metal ions to perform RNA cleavage. The possibility was raised that binding of the second ion is stabilized when an RNA substrate or inhibitor is present.51, 57

The Heterotrimeric Influenza Polymerase Complex

The crystal structure of the large (∼260 kDa) polymerase complex was first resolved in 2014 by Cusack and co‐workers,23, 24 who succeeded to achieve high resolution (2.7 Å) structures of the polymerase heterotrimer, in complex with the vRNA promoter and originating from bat influenza A (FluA; Fig. 5A) or human influenza B (FluB; Fig. 5B) virus. The crystal structures of the apo influenza C polymerase complex (FluC; without bound promoter; Fig. 5C),58 and FluB polymerase in complex with a 5ʹ cRNA fragment (Fig. 5D),56 were revealed more recently.

Figure 5

Comparison of the crystal structures of the heterotrimeric influenza polymerase complex containing full‐length PA, PB1, and PB2. The models are shown in the same orientation, and the same coloring was applied for the different subdomains. (A) Bat FluA polymerase with bound vRNA promoter [PDB: 4WSB].23 (B) Superposition model of the FluB polymerase crystal structure with a template–primer (orange–green) duplex and incoming NTP (black) (taken from a poliovirus polymerase crystal structure). The yellow spheres represent the capped primer bound to PB2, after cleavage by the PA endonuclease domain. This primer is now directed toward the PB1 catalytic cavity, where primer elongation occurs. (Adapted by permission from Macmillan Publishers Ltd: Nature, Reich et al.,24 copyright 2014.) (C) Influenza C polymerase (PDB: 5D98) in apo form.58 (D) FluB polymerase structure with bound cRNA 5′ end56 (PDB: 5EPI). (E) Domain arrangement of FluB PB1, illustrating the right‐hand‐like polymerase fold. The inset shows a closeup of the PB1 catalytic residues, which coordinate two divalent metal ions (not shown). (F) Subdomain names and color scheme as applied in panels A–E, based on the FluB polymerase numbering. For clarity, the PB1 subunit is colored uniformly in cyan in panels A–D, while its different subdomains are differentiated in panel E.

A first striking observation is the complex intertwining of the three subunits, which results in vastly more extensive intersubunit interactions than previously assumed (see other review articles13, 59 for the literature until 2013). Equally intriguing are some prominent conformational differences among the different crystal structures, which most likely represent functionally different states of the enzyme complex. Third, these crystallization studies are the first to reveal the architecture of the PB1 active site. These three insights undoubtedly create several new avenues for structure‐aided drug design.

The polymerase body is formed by PB1, enclosed by the PA linker (which connects the N‐ and C‐terminal domains of PA) on one side, and the N‐terminal domain of PB2 (PB2‐N) at the other side. Akin to other RNA‐dependent RNA‐polymerase (RdRp) enzymes, the overall structure of the catalytic core of PB1 resembles a right hand (Fig. 5E). It contains fingers, fingertips, palm, and thumb domains, formed mainly by conserved RdRp motifs designated “pre‐A”/F, A, B, C, D, and E. By comparison with other polymerase structures, three conserved aspartic acids (Asp305, Asp444, and Asp445 in FluB) within the active site of PB1 were identified, which have a crucial role in catalysis60 and together coordinate two divalent metal ions (Fig. 5, inset panel E).

The promoter‐bound FluA and FluB polymerase complexes have a similar “open” conformation, a U‐shaped structure (Fig. 5A and B), in which one arm is formed by PA‐Nter and the other by the PB2‐CBD. These domains face each other in the FluA structure, suggesting that this crystal represents the cap‐snatching mode. In the FluB structure, the PB2‐CBD is rotated by 70° toward the PB1 active site where primer elongation takes place (compare Fig. 5A and B, in orange). Hence, this FluB structure is thought to represent the cap‐dependent priming step. In sharp contrast, in the crystal structures of the apo FluC polymerase (i.e., without bound promoter58; Fig. 5C) and the 5ʹ cRNA‐bound FluB polymerase (Fig. 5D),56 the polymerase adopts a “closed” conformation in which cap‐snatching is disfavored; it is not yet clear whether these structures represent an inactive or “only replication‐competent” state. Several interdomain interactions immobilize the PB2‐CBD (Fig. 5C and D, in orange) and PA‐Nter (which is named P3 in the case of influenza C) is repositioned and packed against the PB2‐NLS‐domain (Fig. 5C and D, in maroon). This pronounced difference between the open and closed forms of the polymerase complex is related to a radical in situ rearrangement of a large part of PB2. Hence, the regulatory role of the vRNA promoter, as evidenced by enzymatic studies,61, 62 may be related to triggering the transition from the closed preactivation state to the active transcription state. The crystallization studies further indicated that the influenza polymerase complex contains several hinge regions that provide high flexibility to enable its multiple roles in viral RNA synthesis.56 Stabilization of one of its conformations by an inhibitor, such that a vital rearrangement cannot take place, would be an excellent approach to inhibit the polymerase complex.

Recent Structural Insights: A First Step Toward Solving Some Unknowns

The recent crystallographic analyses also help to interpret some issues that are already debated since many years.13 A first issue with relevance for inhibitor design is related to the potential substrate or sequence specificity of the cap‐snatching reaction. The observation that the capped host RNAs is cleaved at a distance of 10–15 nucleotides from the 5ʹ‐cap can be explained by the 50 Å distance between the cap‐binding and endonuclease domains in the “open” FluA structure.23 Regarding the type of RNA substrate, the previous assumption that only host (pre‐)mRNAs are targeted, is contradicted by recent data that noncoding RNAs are the primary source of capped primers.63, 64, 65 The evidence for sequence preference is inconclusive,19, 20, 42, 52, 62, 66, 67, 68 but recent deep sequencing studies63, 69 indicate a preference for capped primers ending with a purine residue. The selection could be determined20, 70, 71, 72 (i) at the level of RNA cleavage and/or (ii) transcription initiation when requiring base‐complementarity between the 3ʹ end of the vRNA template and a given cellular RNA. To gain insight into the first selection mechanism, a cocrystal of PA‐Nter (or the entire polymerase complex) with an RNA substrate is required. This could elucidate substrate‐dependent changes in the active site of PA‐Nter, with relevance for design of optimized PA inhibitors. One indication in favor of such base preference comes from crystallization studies of PA‐Nter in complex with a ribomononucleotide, which proved successful for uridine 5ʹ‐monophosphate (UMP)44, 46, 49 (Fig. 6H–J) and adenosine 5ʹ‐monophosphate (AMP),44 but not cytidine 5ʹ‐monophosphate (CMP) and guanosine 5ʹ‐monophosphate (GMP).44, 46

Figure 6

Comparison of the predicted orientation of RNA versus that of different ligands within the PA‐Nter active site; and chemical structures of representative PA inhibitors. (A) Hypothetical disposition of the RNA substrate in PA‐Nter with two bound Mg2+ ions, based on the molecular dynamics model (kindly provided by Dr. M. Alfonso‐Prieto), as performed by Xiao et al.57 The 3ʹ end of the mRNA is at the right‐hand side, while its 5ʹ end is located at the left. (B–D) Crystal structures of PA‐Nter in complex with (B) “compound 5”45 (PDB: 4E5J); (C) “compound 7”50 (PDB: 4M5U); or (D) “compound 2”73 (PDB: 5I13).

(E–G) Position of L‐742,001, as (E–F) determined by cocrystallization (E, PDB: 4E5H45 and F, PDB: 5CGV49) or (G) predicted by docking.74 (H–J) X‐ray structures of PA‐Nter with bound UMP, as determined by (H) Zhao et al.44 (PDB: 3HW3), (I) Kowalinski et al.46 (PDB: 4AWH), and (J) Song et al.49 (PDB: 5CL0). In the latter two crystal structures, the UMP‐binding mode fully overlaps. The binding regions in PA‐Nter are colored—in orange: metal binding and catalytic residues (His41, Glu80, Asp108, Glu119, Ile120, and Lys134); in yellow: pocket P1 (Ala37, Ile38, Leu42, and Lys34); in blue: pocket P2 (Thr40, Val122, Arg124, Tyr130, and Phe150); in purple: pocket P3 (Arg84, Trp88, Phe105, and Leu106); in red: pocket P4 (Leu16 and Gly81); in green: pocket P5 (Ala20, Tyr24, and Glu26). The bound molecules (i.e., RNA‐substrate or ligand) are shown in cyan, while the metal ions are colored dark red. (K) Chemical structures of flutimide,75 EGCG,76 N‐hydroxyimides,77 compound 16,78 Endo‐1,38 and ANA‐0.79

Substrate selection at the level of initiation is corroborated by the polymerase crystal structures, more specifically the model that was proposed24 (Fig. 5B) by superposition of the influenza B PB1 structure on the poliovirus polymerase primer–template complex.80 This model suggests that the 3ʹ end of the vRNA would pass the PB1 active site by three nucleotides, creating a template overhang that could form three base pairs with the capped primer.

A second intriguing issue is how the different conformations captured in the crystal structures of the heterotrimeric polymerase complex, relate to the cis/trans RNA synthesis model81 for which experimental support was provided by the cryo‐electron microscopic analyses of vRNPs.17, 18 Namely, the branched arrangement of the RNP complexes in which a smaller nascent RNP seems to bud from a larger full‐length RNP, indicate that genome replication is performed in trans by a second polymerase heterotrimer. On the other hand, vRNA transcription would be performed by the cis‐acting polymerase that resides in the vRNP complexes. Given the high flexibility of the polymerase heterotrimer (see above), it is plausible that structural rearrangements occur in each of its functional states which, furthermore, might also be regulated by specific host cell factors.

STRATEGIES TO INTERFERE WITH THE INFLUENZA VIRUS POLYMERASE, THE NP, OR AN ASSOCIATED HOST CELL FACTOR

General Reflections

In the following section, we describe the diverse strategies that have been explored to block influenza virus RNA synthesis, from the stage of rational design or serendipitous screening, to unravelling the presumed mechanism of action (which includes analysis of the resistance profile), and evaluation in cell culture or mouse models, or, if appropriate, in clinical trials (see Table I for a concise overview). We added in this table some published antiviral activity data, in the awareness that these values may show large variations depending on assay conditions, for instance related to virus input, incubation time, or the parameters used to monitor inhibition of virus replication or cytotoxicity of the test compounds. In our own experience, in vitro evaluation based on cytopathic effect (CPE) reduction is more stringent compared to plaque reduction or virus yield assays. In the latter approach, a reduction in virus titer of at least two‐log10 is required to conclude that a given compound has meaningful activity. For this reason, we marked with an asterisk the published compounds for which the anti‐influenza effect in cell culture has been confirmed in at least two independent studies. We also mentioned the published resistance mutations. Resistance studies help to reveal the precise interaction with the viral target and represent the gold standard to validate the proposed mechanism of action of any new class of antiviral molecules.

Table I

Overview of Published Approaches to Interfere with the Influenza Virus Polymerase, the Nucleoprotein or a Host Factor Involved in Viral RNA Synthesis

Compound name or code	Proposed mechanism of actiona	Antiviral activity in cell cultureb	Known spectrum of influenza virus activityc	Published resistance mutations	References
Inhibitors targeting PB1 (directly or indirectly)
Nucleoside and nucleobase analogues
T‐705 (in Phase 3/approved in Japan)	Alternative substrate for viral polymerase; chain termination and lethal mutagenesis	0.1–10 μM82, 83, 84[*]	A, B, and C	PB1: V43I85	82, 83, 84, 85
Ribavirin	Alternative substrate for viral polymerase; lethal mutagenesis; IMPDH inhibitor	2.5–37 μM86, 87, 88; 2–3 log10 reductions in virus titer at 10 μM89 [*]	A, B, and C	PB1: V43I85 and D27N90	85, 86, 87, 88, 89, 90
Prodrug 2	Ribavirin prodrug	0.5 μM87	A		87
Viramidine	Ribavirin prodrug	8.6–132 μM88	A and B		88
5‐Azacytidine	Alternative substrate for viral polymerase; lethal mutagenesis	2–3 log10 reductions in virus titer at 10 μM89	A		89
5‐Fluorouracil	Alternative substrate for viral polymerase; lethal mutagenesis	2–3 log10 reductions in virus titer at 80 μM89	A		89
EICAR	IMPDH inhibitor	1.5–8.6 μM91	A and B		91
ETCAR	IMPDH inhibitor	1.2–19 μM92	A and B		92
Selenazofurin	IMPDH inhibitor	0.7–1.2 μM93	A and B		93
Pyrazofurin	IMPDH inhibitor	0.24–0.88 μM94	A, B, and C		94
LY217896	IMPDH inhibitor	2.9–12 μM95	A and B		95
Compound 3c (analogue of T‐705)	Alternative substrate for viral polymerase	1.9 μM96	A		96
Compound 8a (analogue of T‐705)	?	5.6–7.4 μM97	A		97
2ʹ‐Deoxy‐2ʹ‐fluoroguanosine	Alternative substrate for viral polymerase; nonobligate chain termination	2–22 μM98, 99 [*]	A and B	?d	98, 99
2ʹ‐Substituted carba‐nucleoside analogues “6” and “10”	?	0.9–51 μM100	A and B		100
C‐3ʹ‐modified analogues	?	3.7–65 μg/ml102	A and B		102
6‐Methyl‐7‐substituted‐7‐deaza purine nucleoside analogues “5x” and “5z”	?	3.6–7.0 μM103	A		103
Protide 9j	2ʹ‐Deoxy‐2ʹ‐fluoroguanosine prodrug	EC99 = 12 μM104	A		104
2′‐Deoxy‐2′‐fluorocytidine	Alternative substrate for RNA synthesis?; immunomodulator?	0.05–8 μM105	A and B		105
2ʹ‐2ʹ‐Difluorodeoxycytidine	Interferes with pyrimidine nucleotide synthesis	<0.032–1.2 μM105	A and B		105
Protide 23a	2′‐Deoxy‐2′‐fluorouridine prodrug	EC99 = 49 μM106	A		106
Compound A3	Interferes with pyrimidine nucleotide synthesis	0.04–1 μM107	A and B		107
N10169	Interferes with pyrimidine nucleotide synthesis	3 μM108	B		108
Carbodine	Interferes with pyrimidine nucleotide synthesis	0.6–36 μM94, 109, 110 [*]	A, B, and C		94, 109, 110
Cyclopentenyl cytosine	Interferes with pyrimidine nucleotide synthesis	9.6–26 μM94	A, B, and C		94
3‐Deazaguanine	IMPDH inhibitor	14–49 μM94	A, B, and C		94
Nonnucleoside analogues
“367”	?	0.5 μM111	A	PB1: H456P111	111
ASN2	?	3‐14 μM112	A and B	PB1: Y499H112	111
Inhibitors targeting PB2
VX‐787 (in Phase 2)	Inhibits cap‐binding	0.32–2.8 nM37	A	PB2; cell culture: Q306H, S324I/N/R; F404Y and N510T113; patients: M431I11	11, 37, 113
Cap‐3 and Cap‐7	Inhibit cap‐binding	1.17–9 μM114	A		114
Inhibitors targeting PA
AL‐794 (in Phase 1)	Endonuclease inhibitor	Undisclosed
S‐033188 (in Phase 2)	Endonuclease inhibitor	Undisclosed
L‐735,882 and analogues	Metal‐chelating endonuclease inhibitor	0.18–1.6 μM115, 116	A and B		115, 116
L‐742,001	Metal‐chelating endonuclease inhibitor	0.35116 to 11117 μM [*]	A and B	PA: T20A,49, 74, 117 H41A,74 L42T,74 I79L,49 G81F/T/V,74 F105S,49 E119D,49 I120T,74 V122T74	49, 74, 116, 117
Flutimide	Metal‐chelating endonuclease inhibitor	5.9 μM75	A and B		75
Tetramic acid “36”	Metal‐chelating endonuclease inhibitor	21 μM118	A		118
Phenethylphenylphthalimide analogs PPT‐65 and PPT‐66	Metal‐chelating endonuclease inhibitor	26–48 μM119	A		119
Carboxamides “42,” “44,” “45”	Metal‐chelating endonuclease inhibitor	13–19 μM101	A		101
Pyrimidinol “26”	Metal‐chelating endonuclease inhibitor	13 μM45	A		45
Hydroxypyridinone “7”	Metal‐chelating endonuclease inhibitor	11 μM50	A		50
Hexanetetrone “4” and benzohydrazide “16”	Metal‐chelating endonuclease inhibitor	18–23 μM78	A		78
(Tri)hydroxyphenyls “1,” “2,” and “3”	Metal‐chelating endonuclease inhibitor	11–14 μM120	A		120
Dihydroxyindoles “10” and “15”	Metal‐chelating endonuclease inhibitor	EC99 = 5.7–12 μM121	A		121
THC19	?	31–45 μM122	A	PA: V44I and E165D122	122
Aptamer PAN‐2	?	About 10 nM123	A		123
Pyridopiperazinediones Endo‐1, Endo‐8, and Endo‐9	Metal‐chelating endonuclease inhibitor	0.39–2 μM114	A		114
ANA‐0	Endonuclease inhibitor	0.8–4 μM79	A		79
PPI inhibitors designed against the PA‐PB1 interface
Benzofuran derivatives	Inhibit PA‐PB1 assembly	1–60 μM124, 125	A		124, 125
Benzbromarone and diclazuril	Inhibit PA‐PB1 assembly	31–39 μM126	A		126
“1” and AL18	Inhibit PA‐PB1 assembly	8.3–23 μM127, 128	A and B		127, 128
Cycloheptathiophene‐3‐carboxamides “6” and “19”	Inhibit PA‐PB1 assembly	15–28 μM129	A and B		129
Thiophene‐3‐carboxamides “7,” “10,” “’18,” and “19”	Inhibit PA‐PB1 assembly	11–43 μM130	A and B		130
Triazolopyrimidines “16,” “31,” “’36,” and “37”	Inhibit PA‐PB1 assembly	5–51 μM131	A and B		131
4,6‐Diphenyl pyridines “1,” “11,” “15”	Inhibit PA‐PB1 assembly	7.3–26 μM132	A		132
ANA‐1	Inhibit PA‐PB1 assembly	0.09–1.2 μM133	A		133
Inhibitors targeting NP
Nucleozin	Inhibits cytoplasmic trafficking of vRNPs	0.069–0.33 μM134 [*]	A	NP: Y52H/C,111, 135 Y289H,134, 135 N309K/T,134, 135, 136	111, 134, 135, 136
PPQ‐581	Inhibits cytoplasmic trafficking of vRNPs	1 μM137	A	NP: S377G137	137
“3,” “7,” “’12,” and “23”	Disrupts NP dimerization by targeting the E339…R416‐salt bridge	1.7–118 μM138	A		138
Naproxen	Blocks the RNA‐binding groove of NP	11–25 μM139	A		139
F66	Blocks the RNA‐binding groove of NP	∼1–5 μM140	A		140
RK424	Inhibits nuclear export of NP; inhibits NP–RNA, and NP–NP interactions	0.40–0.63 μM141	A		141
Mycalamide analog “4”	Inhibits nuclear transport of NP?	59–97% reduction of plaque formation at 32 μM142	A		142
Ingavirin (approved in Russia)	Diverse effects on NP functioning?	Inconsistent143, 144, 145	A and B		143, 144, 145
Other
AVI‐7100 (in Phase 1)	Inhibition of M‐gene translation by gene silencing	10–20 μM146	A		146
KPT‐335 (in Phase 1)	CRM1 inhibition	0.01–0.42 μM147	A and B		147
BPR3P0128	Inhibitor of cap‐binding	0.05–0.19 μM148	A and B		148
DPQ	Inhibition of RNA synthesis by binding to 5′ vRNA promoter	72–276 μM149	A and B		149
Geldanamycin, 17‐AAG, and 3beta‐acetoxydeoxodihydrogedunin	Hsp90 inhibition	1–2 log10 reductions in virus titer150, 151 [*]	A		150, 151
U0126	MEK inhibition	1.2–141 μM152, 153 [*]	A		152, 153
PD‐0325901, AZD‐6244, AZD‐8330, and RDEA‐119	MEK inhibition	0.005–0.75 μM154	A		154
WV970	Kinase inhibition	0.015–0.24 μM155	A and B		155
Acetylsalicylic acid, pyrrolidine dithiocarbamate, SC75741	NF‐κB inhibition	1–4 log10 reductions in virus titer156, 157, 158	A		156, 157, 158
Bay 11–7082	NF‐κB inhibition	40–85% reduction in virus titer at 15 μM159	A		159
NSC23766	Rac1 GTPase inhibition	22 μM160	A and B		160

For compounds with broader antiviral activity, only the mechanism of action proposed for influenza virus is given. Note that for many of the experimental compounds, the antiviral mode of action in cell culture is still unproven (e.g., because no resistance data are thus far available).

Unless stated otherwise, the EC50 value is given. The [*] annotation indicates that the antiviral activity was reported in two or more independent publications.

The information in this column indicates only the clearly defined activity spectrum. For most compounds, only a limited number of influenza viruses was tested, and hence, the activity spectrum was not (yet) specified.

Rollins et al.99 described the selection of partially (fivefold) resistant influenza virus obtained by serial passaging; and Tisdale et al.161 reported a partially (tenfold) less susceptible polymerase obtained from resistant virus; however, no sequence analysis of these viruses was reported.

During compound testing, influenza B virus is sometimes ignored, although it accounts for about 25% of seasonal influenza cases. The protein sequences of PA, PB1, and PB2 are, overall, highly conserved among influenza A and B viruses, but even subtle amino acid differences in a binding pocket may render an inhibitor inactive against either of the two virus types. Several inhibitor classes described below were designed to interact with critical functional residues (for instance crucial for binding the capped RNA or NTP substrates). Still, high inhibitor binding affinity often requires involvement of adjacent and less conserved sites, which can give influenza A or B specific activity, or a reduced barrier for the virus to acquire resistance. In other words, although the viral polymerase complex is, generally speaking, an excellent target for developing broad influenza A and B inhibitors that do not readily select for resistance, this theoretical assumption requires biological verification for every single class of new inhibitors.

Besides application in seasonal influenza infections, antiviral drugs are essential to combat serious zoonotic infections (in particular, by highly pathogenic avian influenza A viruses). This seems less of a concern for inhibitors that directly interact with active domains in PA, PB1, or PB2, since these regions are highly conserved among human and avian influenza A viruses. On the other hand, there are species‐dependent amino acid differences in some parts of the protein complex, supposedly related to a regulatory role for host cell factors such as ANP32A,29, 162 RNA polymerase II,163 or nuclear import or export proteins.14 The insights into the influenza virus host interactome are rapidly growing164 and will hopefully rationalize the concept of host cell based antiviral therapeutics. As of today, some approaches (described in Section 3.,H) have already been proposed, although their mechanistic details remain to be established.

Nucleoside and Nucleobase Analogue Inhibitors

Ribavirin and Structurally Related Carboxamide Analogues

Ribavirin [1‐(β‐d‐ribofuranosyl)‐1,2,4‐triazole‐3‐carboxamide; Fig. 7A] was already discovered in 1972.86 This nucleoside analogue inhibits many diverse RNA viruses including influenza virus. Ribavirin has been a first‐line therapeutic for hepatitis C virus infections, but its clinical utility in the management of influenza infections seems limited. In clinical trials performed in the 1970s and 1980s, ribavirin was found to be ineffective against experimentally induced influenza,165 though some benefit was provided at higher drug doses,166 and in vivo data support its potential usefulness in combination therapy.167 Although it has severe side effects (e.g., hemolytic anemia) and teratogenic properties,168 ribavirin may be a last resort for clinicians encountering rare cases of multidrug‐resistant influenza viruses.169

Figure 7

Chemical structures of proven or tentatively proposed polymerase inhibitors. (A) Nucleoside or nucleobase analogue inhibitors ribavirin,86 favipiravir82 (T‐705), 2ʹ‐deoxy‐2ʹ‐fluoroguanosine98 (2ʹ‐FdG), “8a,”97 and “3c.”96 The rotating carboxamide moiety (encircled) in ribavirin and favipiravir explains their ambiguous base pairing, since their base part mimics guanine as well as adenine. (B) Compounds for which the detailed mechanism of action at the level of the polymerase has not yet been revealed. (C) Compounds targeting a host cell factor that has been linked to viral polymerase function. See Table I for references on the individual compounds.

The mode of action of ribavirin is rather complex, since it appears to inhibit virus replication through a combination of different mechanisms. After phosphorylation by adenosine kinase170 or cytosolic 5ʹ‐nucleotidase II,171 ribavirin 5ʹ‐monophosphate inhibits the cellular enzyme inosine monophosphate dehydrogenase (IMPDH), resulting in a decrease in the intracellular GTP pool and, hence, indirect inhibition of viral RNA synthesis.172, 173, 174 Other proposed mechanisms include direct inhibition of the viral polymerase by ribavirin 5ʹ‐triphosphate,175, 176 immunosuppression,177 and lethal virus mutagenesis.85 It was recently demonstrated that ribavirin induces mutagenesis in the influenza virus genome by acting as an ambiguous purine analogue and increasing G‐to‐A and C‐to‐U mutations.85 Serial virus passaging in the presence of the compound gave rise to a mutant virus with twofold resistance to ribavirin and a lower degree of resistance to favipiravir (described below). This mutant virus carried mutation V43I in PB1, which was associated with increased polymerase fidelity and reduced pathogenicity in mice.85 This Val43 residue lies in close proximity of the putative NTP entrance channel, as predicted in the polymerase crystal structure of Pflug et al.,23 which could explain how the V43I mutation may affect the activity of ribavirin, for instance by modifying the interaction of the polymerase with GTP or ribavirin 5ʹ‐triphosphate. The last analysis is difficult to reconcile with the observation that ribavirin 5ʹ‐triphosphate is a rather weak inhibitor of the viral polymerase in enzymatic assays, with reported IC50 values of 70 μM173 or ∼100 μM.176 By using random mutagenesis, Binh et al.90 identified PB1 mutation D27N, which conferred a 1.8‐fold resistance to ribavirin. This mutation was also picked up by Pauly et al. after serial passaging of influenza virus under ribavirin.89 By inspecting the published crystal structure,23 we noticed that Asp27 is located outside the catalytic site, in the vicinity of the 5ʹ vRNA promoter binding site. Pauly et al. further confirmed that ribavirin and two other nucleoside analogues, 5‐azacytidine and 5‐fluorouracil, are lethal mutagens for influenza virus in vitro.89

With the aim to improve the efficacy or safety, Dong et al. obtained an alkoxyalkylphosphodiester prodrug of ribavirin, prodrug “2,” which still requires in vivo validation.87 Viramidine, the 3‐carboxamidine prodrug of ribavirin,178 was found to be slightly less active against influenza virus in vitro and in vivo, but also showed less toxicity.88 In addition to ribavirin, the following carboxamide‐containing nucleoside analogues exhibit broad antiviral activity, presumably by inhibition of IMPDH: 5‐fluoro‐1‐β‐d‐ribofuranosylimidazole‐4‐carboxamide (FICAR),179 5‐ethynyl‐1‐β‐d‐ribofuranosylimidazole‐4‐carboxamide (EICAR),91 5‐ethynyl‐1‐β‐d‐ribofuranosyl‐1H‐[1-3]triazole‐4‐carboxylic acid amide (ETCAR),92 selenazofurin,93 and pyrazofurin.94 All these are cytostatic molecules mainly explored as anticancer agents.180 The substituted thiadiazole compound LY217896 was shown to display broad anti‐influenza virus activity95 via IMPDH inhibition,181 but proved ineffective in a placebo‐controlled clinical trial.182

Favipiravir

Favipiravir (6‐fluoro‐3‐hydroxy‐2‐pyrazinecarboxamide; Fig. 7A), also known as T‐705, was approved in Japan in March 2014 for restricted use in uncomplicated influenza virus infections, and is currently in Phase 3 clinical trials in the USA and Europe. Unpublished clinical data suggest that the antiviral effects of favipiravir are similar to those of oseltamivir.183 In one reported Phase 2 study,11 favipiravir significantly reduced the time to resolution of symptoms. No signs of drug resistance were seen in more than 700 samples tested.

Favipiravir has broad activity against influenza A, B, and C viruses,82 including the 2009 pandemic A/H1N1 virus,184 highly pathogenic avian influenza H5N1185 and H7N9186 viruses, and virus strains with resistance to M2 blockers or NAIs.187 Among all nucleobase/nucleoside inhibitors of influenza virus reported thus far, favipiravir is absolutely unique in having a clearly superior therapeutic window (i.e., antiviral EC50 value in the range of 0.1–10 μM82, 83, 84 and no cytotoxicity at 6400 μM82).

Mechanistically, favipiravir is a nucleobase mimetic that undergoes intracellular conversion to its ribofuranosyl 5ʹ‐triphosphate metabolite. The cellular hypoxanthine guanine phosphoribosyltransferase (HGPRT) converts favipiravir into its ribose‐5′‐monophosphate (RMP), which is further metabolized, by cellular kinases, to favipiravir‐ribosyl‐5′‐triphosphate (favipiravir‐RTP).83 Its requirement for very high dosing (in the order of 1600–2400 mg per day in some clinical trials) may, at least partially, be related to the low efficiency of its metabolic activation.83 This limitation could be solved by designing prodrug forms of favipiravir that bypass one or more of its activation steps.

Favipiravir‐RTP is recognized by influenza virus RNA polymerase as an alternative for the natural substrates GTP and, to a lesser degree, ATP.188, 189, 190 The precise mechanism of action of favipiravir remains to be fully explained, and two nonmutually exclusive hypotheses have been proposed. Similarly to what is described above for ribavirin, favipiravir can cause lethal virus mutagenesis by inducing a high rate of mutations and generating a nonviable viral phenotype. In virus grown under favipiravir, Baranovich et al.84 observed a reduction in virus infectivity without a corresponding decrease in the number of viral RNA copies, together with a dose‐dependent increase in mutation frequency in the influenza virus genome (primarily G‐to‐A and C‐to‐U). The biochemical basis for this mutagenic effect was revealed in the enzymatic studies by Jin et al.,189 since the influenza polymerase was shown to efficiently incorporate favipiravir‐RTP both opposite to C and U, meaning that favipiravir‐RTP mimics GTP as well as ATP. This ambiguous base‐pairing behavior is related to the rotating carboxamide moiety (encircled in Fig. 7A). The second hypothesis, “non‐obligate chain termination,” is supported by the biochemical observation that incorporation of a single molecule of favipiravir‐RMP (which carries a normal 3ʹ‐hydroxyl group) into a nascent influenza RNA strand causes inhibition of viral RNA extension.190 However, another study did not confirm these results and concluded that at least two consecutive incorporation events of favipiravir‐RMP are needed to terminate viral RNA elongation.189 Hence, the RNA chain‐terminating effect may prevail at higher favipiravir concentrations, while at lower compound concentrations, the mutagenic effect may become apparent. Interestingly, the compound appears to have an exceptionally high barrier for selecting resistance, since no favipiravir‐resistant influenza virus was obtained in cell culture after up to 30 serial virus passages in the presence of favipiravir.84, 191 From a drug development perspective, this quality is of course clearly advantageous. On the other hand, it obstructs experiments aimed at providing conclusive evidence on favipiravir's mechanism of action.

Next to its activity against influenza virus,82 favipiravir inhibits the replication of various other RNA viruses.192 Until now, few structural analogues of favipiravir have been reported. The 6‐fluoro substituent is not required for antiviral activity,83, 191 whereas the 3‐hydroxyl group is indispensable.83 An example of a base‐modified analogue is “3c” (Fig. 7A), which displayed comparable antiviral activity as T‐705 in cell culture and influenza polymerase enzymatic assays.96 A recent example of a sugar‐modified analogue is the 2ʹ,4ʹ‐bridged analogue “8a” (Fig. 7A), which has anti‐influenza virus activity comparable to that of favipiravir.97 This compound has the 3ʹ‐hydroxyl of the pentose in the inverted xylo position. Evaluation of “8a” (or, rather, its 5ʹ‐triphosphate) at the level of the viral polymerase was not yet reported.

2ʹ‐Deoxy‐2ʹ‐Fluoroguanosine and Other Nucleoside Analogues

2ʹ‐Deoxy‐2ʹ‐fluoroguanosine (2ʹ‐FdG; Fig. 7A) was reported years ago as a broad inhibitor of influenza A and B viruses in cell culture.98 Its analogues 2ʹ‐deoxy‐2ʹ‐fluoroadenosine and 2ʹ‐deoxy‐2ʹ‐fluoroinosine were significantly less active.99 In vivo, 2ʹ‐FdG was shown to reduce influenza virus titers in the respiratory tract of mice and ferrets.193 Mechanistically, 2ʹ‐FdG‐triphosphate was found to inhibit the influenza polymerase complex by nonobligate chain termination. This agrees with the observation that a virus selected for resistance to 2ʹ‐FdG, contained a polymerase with tenfold lower susceptibility to 2ʹ‐FdG‐triphosphate in an enzymatic assay.161 The identity of the amino acid changes in this mutant polymerase was not disclosed, which is unfortunate since this insight could be very helpful to explain the role of specific residues in the catalytic or other functional domain of PB1, as identified in the recent crystallographic studies.23, 24, 56, 58

During more recent years, 2ʹ‐substituted carba‐nucleoside analogues,100 C‐3ʹ‐modified analogues,102 and 6‐methyl‐7‐substituted‐7‐deaza purine nucleoside analogues 103 were reported to have anti‐influenza activity comparable to 2ʹ‐FdG. The antiviral activity of 2ʹ‐FdG and analogues is dependent on intracellular conversion to the active nucleoside 5ʹ‐triphosphate form. ProTide prodrugs were employed to overcome the first (rate‐limiting) phosphorylation step, and enable intracellular delivery of the nucleoside 5ʹ‐monophosphate species.194 This ProTide concept was successfully applied to 2ʹ‐FdG and its uridine analogue 2ʹ‐FdU.104, 106

The pyrimidine analogue 2ʹ‐deoxy‐2ʹ‐fluorocytidine (2ʹ‐FdC) seems more potent than 2ʹ‐FdG, with in vitro and in vivo activity against various strains of influenza A or B.105 It remains to be demonstrated whether the 5ʹ‐triphosphate of 2ʹ‐FdC acts as an alternative substrate for the influenza polymerase complex, as is the case for 2ʹ‐FdG. Alternatively, since 2ʹ‐FdC was shown to be cytostatic in cells,195 it could act as an immunomodulator in vivo, or inhibit influenza virus through inhibition of cellular enzymes which are involved in de novo pyrimidine biosynthesis. Similar to what is explained above for ribavirin and other inhibitors of IMPDH, nucleoside analogues that interfere with purine or pyrimidine nucleotide synthesis can give strong inhibition of influenza virus replication in cell culture. Unfortunately, the therapeutic window of this approach is too narrow to allow broad application in influenza virus therapy. A few examples are 2ʹ‐2ʹ‐difluorodeoxycytidine 196 (known as the anticancer drug gemcitabine), compound “A3,”107 and N10169,108 which lower the intracellular pyrimidine levels by inhibiting ribonucleotide reductase, dihydroorotate dehydrogenase, and orotidylate decarboxylase, respectively. Also, the anti‐influenza activity of the carbocyclic nucleoside analogues carbodine and cyclopentenyl cytosine has been linked to pyrimidine depletion through CTP‐synthetase inhibition.94, 197, 198, 199 For a few carbocyclic purine nucleoside analogues with borderline to modest activity against influenza virus, no mechanistic data are available.200 For the broad antiviral agent 3‐deazaguanine, a mechanism involving IMPDH inhibition was proposed.94, 201

Time to Revisit Nucleoside Inhibitors for Influenza?

That nucleoside analogues merit more attention in influenza drug development is nicely illustrated by the successful progression of T‐705. The unprecedented bonuses of this agent are: high resistance barrier, low cytotoxicity, and broad coverage of diverse RNA viruses. As for 2ʹ‐FdG, the relevance of the 2ʹ‐fluoro modification to achieve nonobligate chain terminators of RNA virus polymerases202 is underscored by the fact that this substitution is present in successful drugs for hepatitis C203 or drug candidates for respiratory syncytial virus therapy.204 An important aspect is the avoidance of inhibitory effects on cellular polymerases or enzymes of the purine or pyrimidine pathways, since this will de facto reduce the therapeutic index. Besides this relevance for drug development, nucleoside analogues (or their 5ʹ‐triphosphate forms) are unique tools to study the kinetics or fidelity of the influenza polymerase, and the role of specific PB1 residues.

Inhibitors of the Cap‐Snatching Reaction

Inhibitors of Cap‐Binding by PB2

As explained in Section 2,A, the first part of the cap‐snatching reaction by the influenza virus polymerase involves cap‐binding by PB2. Initially, the fact that PB2 recognizes capped RNA alike eukaryotic cap‐binding proteins such as eIF4E,205 raised doubts about the target drugability of the PB2‐CBD and the possibility to design influenza virus‐selective cap‐binding inhibitors. However, Hooker et al.39 used quantitative UV crosslinking to analyze the elements in cap analogues that contribute to their recognition by isolated vRNPs and eIF4E, and identified some notable differences. Based on this, compound RO0794238 (Fig. 2F) was designed as a cap analogue lacking a negative charge and containing an acyclic moiety instead of a ribose, with the aim to achieve superior potency and selectivity for PB2. This molecule was indeed able to inhibit cap‐binding by influenza virus vRNPs in a dose‐dependent manner, with no effect on eIF4E.39

The revelation of the cocrystal structure of the PB2‐CBD in complex with m7GTP32 (Fig. 2A and Section 2,A) finally enabled the rational design of selective inhibitors, given its unique protein fold and cluster of hydrophobic residues. Four compounds (e.g., compound “11”; Fig. 2C) displayed potent activity in a binding assay with the isolated PB2‐CBD but, unfortunately, were devoid of antiviral activity in influenza virus‐infected cells (possibly due to the presence of multiple negative charges, resulting in poor cellular uptake).36 The previously identified compound RO079423839 however appeared inactive in this binding assay. The authors speculated that RO0794238 might inhibit cap‐binding by PB2 in an indirect manner through its interaction with another part of the viral RNP complex. In addition, given its long substitution at the N‐7 position, it seems implausible that RO0794238 would be able to bind in the m7GTP pocket of the PB2‐CBD.

The validity of the PB2‐CBD as a drug target is best demonstrated by the clinical candidate VX‐787 (JNJ‐63623872 or JNJ‐872; Fig. 2D). In a Phase 2a challenge study, VX‐787 yielded a dose‐dependent decrease in symptom scores and duration of symptoms.11 Additional clinical trials with this promising compound are ongoing. Its discovery process was reported in 2014 by Clark et al.37 VX‐787 is an azaindole derivative that resulted from extensive iterative synthesis to develop PB2 inhibitors, which optimally occupy the m7GTP binding pocket, as demonstrated by cocrystallizing VX‐787 with the PB2‐CBD of influenza A (Fig. 2D). In particular, optimal interactions with the hydrophobic as well as basic residues were achieved by introducing the bicyclooctane‐carboxylate moiety. A further structure‐activity relationship (SAR) exploration with isosteric replacements of the carboxylic group of VX‐787 showed that the pKa value and orientation of the negative charge significantly affect both anti‐influenza potency and selectivity for unwanted protein kinase targets (the latter being related to compound binding to the ATP site of cellular kinases).206 VX‐787 possesses strong antiviral activity (i.e., EC50 values in the nanomolar range) in cellular assays with a broad range of influenza A virus strains, including NAI‐ and amantadine‐resistant isolates, 2009 pandemic H1N1, and circulating avian H5N1 strains.37, 207 Its activity against influenza B virus is negligible, which can be attributed to amino acid differences in the PB2‐CBD.208 In particular, the π‐stacking interaction between VX‐787 and Phe323 in the influenza A PB2‐CBD cannot be formed in the influenza B protein, which contains a glutamine (Q325) at the corresponding position (Fig. 2A and B). The apparent flexibility in the PB2‐CBD of influenza B (see Section 2,A) may be another factor complicating the development of cap‐binding inhibitors that cover both influenza A and B virus.

Resistance studies with VX‐787 in cell culture delivered six resistance mutations in PB2 (Q306H, S324I, S324N, S324R, F404Y, and N510T), which yielded at least 60‐fold reduction in VX‐787 sensitivity.207 In the Phase 2a clinical study, an M431I mutant virus was detected in a minority of the patients, which, in cell culture, displayed 57‐fold lower sensitivity to VX‐787, yet reduced viral fitness.11 In mice, prophylactic use of VX‐787 (which is orally bioavailable) provided 100% protection against the APR8 laboratory strain, a pandemic 2009 strain, or a highly pathogenic avian H5N1 virus. In addition, this inhibitor provided 100% survival when treatment was initiated up to 4 days after challenge with influenza virus.207

Recently, Roch et al.38 reported two compounds, Cap‐3 and Cap‐7 (Fig. 2E), which bind to the PB2‐CBD and inhibit transcription in an enzymatic assay with the polymerase complex. In addition, they inhibit virus replication in cell culture, albeit considerably less potently than VX‐787 (i.e., EC50 values in the range of 1–9 μM). Hitherto, mechanistic validation was not yet reported.

Metal‐Chelating Inhibitors of PA‐Nter Endonuclease Activity

The conserved nature of several amino acid residues inside the catalytic site of PA‐Nter implies that this domain is highly relevant to achieve inhibitors with broad activity against influenza A and B. The strategy explored thus far consists of metal‐chelating scaffolds containing coplanar oxygens to bind the divalent metal ion(s), which resembles the action principle of approved HIV integrase inhibitors.209, 210 The challenge is to achieve inhibitors, which optimally occupy the PA‐Nter catalytic site and surrounding regions, yet do not interact with metal‐dependent proteins of the host cell. The concept seems now validated since the recent introduction of two PA inhibitors (AL‐794 and S‐033188) into clinical trials.211

The literature contains diverse PA inhibitors with inhibitory activity in enzymatic assays with the viral polymerase or isolated PA‐Nter. It is important to note that for many of these agents, cell culture data on antiviral activity and, in particular, mechanism of action are not elaborated. Some compounds have a high anionic charge that impedes their cellular uptake. For example, we demonstrated that for EGCG (Fig. 6K), the activity in virus‐infected cells is related to inhibition of virus entry.74 To definitely prove that an assumed PA inhibitor acts upon PA in a virus/cell context, resistance studies should be carried out. As of today, this decisive mechanistic evidence is only available for L‐742,001 (Fig. 6E–G), since three laboratories including ours49, 74, 117 demonstrated that specific mutations in PA confer moderate (3‐) to high (>10‐fold) viral resistance to this agent. The substitutions were present in the catalytic core of PA or surrounding hydrophobic pockets (colored purple, green, red, and orange in Fig. 6), confirming that these PA regions are critical for the antiviral mode of action of L‐742,001. This molecule is the prototype of the first class of influenza endonuclease inhibitors, already discovered at Merck about two decades ago.115, 116 Among this series of 4‐substituted‐2,4‐dioxobutanoic acids with a characteristic β‐diketo acid motif, compound L‐742,001 was identified as a particularly potent inhibitor of the influenza virus endonuclease reaction in an enzymatic assay, and of virus replication in cells.49, 74, 116, 117 In a mouse model, the compound provided up to 4‐log10 reduction in virus lung titers.116 Strong in vivo activity was also reported for the recently discovered inhibitor ANA‐0 proposed to act upon PA (Fig. 6K).79 Other classes of reported endonuclease inhibitors (awaiting mechanistic validation in cell culture) include: flutimide 75 (Fig. 6K) and a series of more potent aromatic analogues,212 and (Fig. 6K) compounds,77 tetramic acid derivatives,118 polyphenolic catechins,76 trihydroxyphenyl‐bearing compounds73, 78, 120 (Fig. 6D and K, compound 273 and compound 1678), phenethylphenylphthalimide analogues derived from thalidomide,119 macrocyclic bisbibenzyls,213 fullerenes,214 hydroxyquinolinones,215 hydroxypyridinones,47, 50 hydroxypyridazinones,216 hydroxypyrimidinones,216 β‐diketo acid (DKA), and DKA‐bioisosteric compounds,68 bis‐dihydroxyindolecarboxamides,121 2‐hydroxybenzamides,217 thiosemicarbazones,218 pyrimidinoles,45 pyridopiperazinediones 38 (Fig. 6K, Endo‐1), and miscellaneous compounds bearing distinct pharmacophoric fragments.45, 78, 101 All these different chemotypes have in common that they bear chelating motifs able to bind the bivalent metal ion(s) in the catalytic core of PA‐Nter.219

As explained in Section 2,B, the different crystal structures for PA‐Nter that have become available since 2009 converge on the overall structure of the active site, yet do not agree on whether the enzyme contains one or two (or possibly three) divalent metal ions in its catalytic site, nor on which metal ions (i.e., Mn2+ or Mg2+) are present in the native enzyme.42, 43, 44, 45, 46, 47 Knowledge of the precise number and identity of these metal ions is critical to design optimized metal‐chelating PA inhibitors. Similar to what was observed for HIV integrase,220 we found that the potency of metal‐chelating inhibitors in enzymatic assays with PA‐Nter can show large variations depending on which bivalent metal ion (Mg2+ or Mn2+) is used.68 This phenomenon can be explained by the different ligand preferences of these two metal ions. For example, magnesium generally binds oxygen atoms rather than nitrogen, while manganese has slightly greater affinity for nitrogen.221 Moreover, two tautomeric forms of an inhibitor can display a difference in metal preference. For DKA compounds, it was shown that Mn2+ preferentially binds to the diketo form, while Mg2+ predominantly recognizes the β‐keto‐enol form.222 Dependent on which metal ion is physiologically relevant, the compounds’ activity may increase by stabilization of the preferred tautomeric form, for instance by substituting the flexible β‐diketo acid (DKA) moiety by a “locked” motif having optimal geometry for bidentate metal coordination.

Crystallographic analyses of PA‐Nter in complex with diverse PA inhibitors have revealed that the catalytic center is surrounded by different hydrophobic pockets, which are amenable to inhibitor design.45, 46, 47, 49, 50, 120, 215, 216 Overall, the active site of PA‐Nter is quite spacious and flexible, and in silico design of optimized PA inhibitors is further complicated by the fact that inhibitors bind via an induced‐fit mechanism. Assuming that metal‐chelating PA inhibitors act as substrate or product mimics, precise insight into the binding mode of the RNA substrate or cleavage product is crucial to identify which of the hydrophobic pockets have the highest relevance for PA inhibitor design. A cocrystal structure for PA‐Nter in complex with an RNA substrate is highly needed. Three X‐ray structures are available of PA‐Nter in complex with UMP (Fig. 6H–J), which can be regarded as a mimic of one of the two cleavage products.44, 46, 49 Two distinct UMP‐binding poses were observed, particularly related to the position of the ribose and base, which underlines the impact of crystallization procedures (i.e., cocrystallization vs. soaking; addition of metal ions or not). In the context of PA, cocrystallization seems more reliable since it enables the binding pocket to adapt to the ligand. One particularly flexible region is the helix containing residue Tyr24, which, in two cocrystal structures, stacks with the uracil base46, 49 of UMP (Fig. 6I and J, in green) and also shows hydrophobic interactions with PA inhibitors like those shown in Figures 6C and E.45, 46, 49, 50 To accommodate these ligands, this Tyr24 moves or rotates compared to its pose in the apo form. Since, upon crystal soaking, this conformational change is unlikely to occur, the ligand may be forced into an alternative and biologically less relevant binding pose.

Another interaction site that seems of high relevance is Arg124. A role for this residue in RNA substrate binding was suggested by Yuan et al.43 This was corroborated by Xiao et al.,57 whose MD simulations (Fig. 6A) pointed to Arg124, Arg125, and Arg192 as the residues responsible for interaction with the 3ʹ end of the RNA substrate. Likewise, the relevance of this Arg124 pocket (in blue in Fig. 6) was suggested by our cell‐based antiviral studies in which we introduced mutations at different sites within or around the PA‐Nter active center, and analyzed the resistance of the mutant viruses to the prototype PA inhibitor L‐742,001.74 The data agreed to a docking model74 for L‐742,001 in PA‐Nter (Fig. 6G), but only partially agreed to the cocrystallization results45, 49 (Fig. 6E and F). In addition, an analogous orientation as for the docked L‐742,001 molecule was seen in the X‐ray structures of PA‐Nter in complex with the dihydropyrimidine‐based “compound 5”45 (Fig. 6B), and the particularly potent hydroxypyridinone‐based PA inhibitor “compound 7”50 (Fig. 6C). For both these compounds, the authors reported a role for Arg124 in binding of the inhibitor.

Consequently, we speculate that a PA inhibitor with a binding orientation that allows interactions with these RNA‐binding residues, mimics the 3ʹ end of the capped RNA substrate and is able to efficiently compete with the substrate. Presumably, the interactions with the 3ʹ end are essential for correct positioning and cleavage of the substrate. In contrast, the interactions between the 5ʹ end of the capped RNA and PA‐Nter (i.e., at the pocket colored in purple in Fig. 6) may be weaker, since the cap‐binding site in the PB2 subunit already ensures strong binding of the 5ʹ end. As a result, the pocket in PA‐Nter, which binds the 5ʹ end could possibly be more flexible. An additional argument is the fact that, once cleavage has occurred, the PA active site has to release the capped 5ʹ end in order to allow its relocation into the PB1 catalytic site for primer elongation and transcription of the viral mRNA. Hence, tight binding affinity of PA to the 5ʹ end of the capped RNA substrate would substantially decrease the rate of product release, which would be detrimental to the overall process of transcription.

PA Targeting Aptamers

Recently, a DNA aptamer library screening performed by Yuan et al.123 yielded a number of aptamers with binding affinity to either entire PA or PA‐Nter. Four aptamers were able to inhibit the endonuclease reaction and one of them showed broad influenza A inhibition when transfected into virus‐infected cells. At present, aptamers still require significant improvements in terms of pharmacokinetics and—dynamics before having broad clinical applicability. Nevertheless, this study demonstrates that aptamers can be excellent tools to identify new relevant binding pockets for inhibitors of the PA endonuclease or other viral target proteins.

Protein–Protein Interaction Inhibitors of the Influenza Polymerase Complex

The assembly of the three subunits (PB1, PB2, and PA) into a functional viral polymerase complex is essential for influenza RNA synthesis and virus replication. Thus, interference with its correct assembly through inhibition of a crucial protein–protein interaction (PPI) is currently explored as an innovative antiviral strategy. Until 2014, tailored design of assembly inhibitors was based on at that time available crystal structures for two specific interfaces in the polymerase heterotrimer (Fig. 8). The revelation of the full polymerase structure now creates the possibility to explore other inter‐ or intrasubunit interfaces, yet two important issues are: (i) the size of a specific interface, to allow fitting of small‐molecule inhibitors; and (ii) the conservation of its residues when envisaging both influenza A and B.

Figure 8

The two thus far explored protein–protein interaction (PPI) domains and chemical structures of PAC‐PB1N PPI inhibitors. (A) Location in the FluA polymerase crystal structure23 (PDB: 4WSB) of the two PPI domains, which have been targeted by peptides or small molecules. (B) Closeup showing a superposition of the crystal structures of the PAC‐PB1N interface223 (PDB: 3CM8) on that of the FluA polymerase (light gray) and the apo form of PAC 224 (light blue; PDB: 4IUJ). (C) Closeup showing a superposition of the crystal structure of the PB1C‐PB2N interface225 (PDB: 3A1G) on the FluA polymerase complex, in light gray. (D) Chemical structures of representative PAC‐PB1N‐PPI inhibitors “compound 1,”127 diclazuril,126 benzbromarone,126 “compound 7e,”124 “compound 36,”131 and ANA‐1.133

The structural details for one important PA‐PB1 interaction domain were revealed in 2008, based on cocrystallization of the PA C‐terminal domain (PAC, residues 239 or 256–716) with a short N‐terminal PB1 peptide (PB1N).223, 226 This PAC domain resembles a dragon's head that clamps the PB1N peptide. As shown in Figure 8B, the base of this pocket overlaps between the crystal structures of the PB1N‐bound PAC domain223 (in dark blue), the full polymerase heterotrimer23 (in gray), and the apo PAC domain224 (in light blue). In contrast, there is less overlap at the periphery, meaning that the binding groove is slightly more narrow in the structure of the PB1N‐bound PAC compared to the two other crystal structures. The less mobile base of the pocket appears relevant as the initial interaction point for PPI inhibitors; to these scaffolds, structural elements should be added to target also the more flexible protein parts on the periphery of the interface.

Three features explain the drugability of this PAC‐PB1N interface: several residues are conserved (among influenza A, B, and C), and the interface is hydrophobic and relatively small, implying that it can be targeted by small molecules.227 Initially, the concept to disrupt the PAC‐PB1N interface was explored with PB1‐derived peptides.228, 229, 230 Among a first series of small‐molecule inhibitors identified by in silico screening followed by a PAC‐PB1N biochemical interaction assay, “compound 1” emerged as particularly relevant given its broad anti‐influenza A and B virus activity in cell culture (Fig. 8D).127 Other early lead molecules with anti‐influenza A activity in cell culture are benzbromarone and diclazuril 126 and “compound 7e”124 (Fig. 8D). This was followed by rational development of diverse lead compounds,125, 129, 130, 131, 132, 231 some of which were used to generate a pharmacophore model for PAC‐PB1N interaction inhibitors.130 For the recently identified inhibitor ANA‐1 (Fig. 8D), the anti‐influenza activity was confirmed in a mouse model.133 On the basis of docking, the binding site of ANA‐1 in PAC was predicted to lie in an allosteric site adjacent to the PB1 interacting domain. Yet, for all the PAC‐PB1N interaction inhibitors reported thus far, mechanistic studies using cocrystallization or resistance selection remain to be performed to verify their antiviral mode of action and precise binding mode. This could also aid to design PAC‐PB1N assembly inhibitors, which establish hydrophilic besides hydrophobic interactions, thereby leading to better solubility and potentially higher antiviral potency than the current lead compounds.232

Another protein–protein interface that was validated with peptide inhibitors233, 234 but is as yet unexplored with small molecules, is the interaction domain between PB1C (residues 678–757; Fig. 8C, in light green) and PB2N (residues 1–37; in dark green). The crystal structure of this isolated domain was published in 2009225 and nicely overlaps with that in the full polymerase complex23 (in gray). Despite the fact that this PB1C‐PB2N interface carries conserved amino acids, the inhibition by a PB2N‐derived peptide seemed prone to strain dependency.235 One proposed hypothesis relates this observation to a conformational change in PB1, when it is bound in the PB1‐PA dimer intermediate, thereby rendering the PB2N‐binding part in PB1 inaccessible. Hence, this PB1‐PB2 interface appears a more challenging target for development of broad influenza virus inhibitors.234, 235

Gene Silencing Approaches

Pharmaceutical development of oligonucleotide inhibitors faces some common obstacles, particularly related to their low in vivo stability and inefficient cellular delivery. Antisense oligonucleotides function as a single strand and block mRNA processing or translation by binding to the mRNA to which they are complementary. One influenza virus inhibitor currently in Phase 1 clinical trials,236 is the antisense oligonucleotide AVI‐7100 (Radavirsen), a phosphorodiamidate morpholino oligomer (PMO). AVI‐7100 contains nonionic morpholino rings (instead of ribose rings) and three phosphorodiamidate intersubunit linkages, and was designed to interfere with translation and splicing of mRNA derived from the M‐gene. In preclinical studies, AVI‐7100 proved effective against influenza A virus infection in mice and ferrets, even when administered after virus challenge.146

The possible application of RNA interference (RNAi) for influenza therapy has been the subject of a series of studies.237 This includes siRNAs, small RNA duplexes, which trigger the destruction of specific mRNAs by associating with the RNA‐induced silencing complex. This concept can be valuable for PA, PB1, PB2, and NP since some regions in their nucleotide sequences are conserved among influenza A subtypes. The siRNAs for NP and PA were shown to prevent accumulation of their mRNAs, and reduce the mRNA, vRNA, and cRNA levels for other viral genes.238 siRNAs against NP, PA, and PB1 were shown to have antiviral activity in cell culture and mice.238, 239 5ʹ‐Triphosphate modification of siRNA (for NP) provided a dual effect by combining gene silencing with RIG‐I activation.240 Alternatively, to circumvent the poor delivery of siRNA, a lentiviral241 or Escherichia coli vector242 was explored to express NP‐ and PB1‐targeting, or NP‐ and PA‐targeting RNA interfering sequences, respectively.

Small‐Molecules Targeting the Viral NP

The viral NP directly interacts with the viral polymerase complex to support viral RNA synthesis.243, 244, 245 However, it primarily has a structural function since it forms the protein scaffold of the vRNP complexes through its homo‐oligomerizing246, 247, 248 and RNA‐binding properties.246, 249 The NP protein is crescent shaped with head, body, and tail domains246, 247 and a putative RNA‐binding groove (Fig. 9A and B).246 NP–NP oligomerization is the result of one tail loop inserting into the body of a neighboring monomer. This insertion is stabilized by intermolecular β‐sheets, hydrophobic interactions, and salt bridges.246

Figure 9

Structure of the viral nucleoprotein and chemical structures of proposed or proven NP inhibitors. The NP structures are based on the H1N1 NP X‐ray structure246 (PDB: 2IQH). The residues lining the RNA‐binding groove are shown in darker colors. (A) Head, body, and tail domains in NP. (B) As in panel A, but rotated 120° around the x‐axis. (C) Predicted binding sites for compound 4,142 naproxen139 and naproxen C0,250 RK424,141 and F66.140 (D) Binding site for nucleozin111, 134, 135 (proven by cocrystallization) and that predicted for PPQ‐581.137 The overlap of the nucleozin‐binding site with that of compound 4 or PPQ‐581 is colored in maroon and orange, respectively. (E) The model in (C) is rotated 90° around the y‐axis to show the internal tail loop binding cavity (in cyan), which is the predicted binding pocket for compound 3.138 (F) Chemical structure of Ingavirin®, which was reported to impair biogenesis and oligomerization of NP in vitro.144

The NP protein contains different sites that are amenable to inhibitor design and are briefly described below (Fig. 9C–E).136 A first series are NP‐aggregating agents such as the aryl piperazine amide compound called nucleozin (Fig. 9D).111, 134, 135, 251 This molecule forms bridges between NP subunits to give higher order NP oligomers.135 Depending on the assay, nucleozin exerts its activity prior to nuclear import of the vRNPs; at the stage of viral RNA synthesis; or during cytoplasmic trafficking of progeny vRNPs.252 It exhibits robust activity against influenza A virus in vitro and in vivo, yet is inactive against influenza B. Viral resistance to nucleozin emerged after only five cell culture passages.111, 134, 135 This agent is inherently inactive against some influenza virus strains (such as the 2009 pandemic H1N1 virus) that already carry one of its resistance mutations.111, 134 Another NP‐aggregating compound, PPQ‐581, shows no cross‐resistance to nucleozin and selects for a resistance mutation adjacent to the nucleozin‐binding site (Fig. 9D).137

A second type of NP inhibitors target the highly conserved Glu339‐Arg416 salt bridge among NP neighbor monomers.246 Shen et al.138 identified four small‐molecule influenza inhibitors, which are able to disrupt NP oligomerization, with compound “3” (Fig. 9E) being the most potent one in cell culture.

A third possibility is to target the RNA‐binding groove in NP. In silico screening139 led to the identification of naproxen (Fig. 9C), which was found to inhibit RNA binding by NP in biochemical assays, and display anti‐influenza virus properties in cell culture and mouse models. The latter activity is likely potentiated by naproxen's well‐known anti‐inflammatory effect. The analogue naproxen C0,250 which carries a 1,3‐dicarboxylated phenyl group instead of the propionic acid in naproxen, has higher NP binding affinity in vitro, but its activity in cell culture was not yet reported. A different virtual screening hit encoded F66 showed antiviral activity in cell culture and mouse models,140 yet its RNA‐NP disrupting activity and binding mode in NP remain to be established.

Compound RK424 141 was reported to inhibit influenza A virus in cell culture and mouse models, and interfere with RNA binding to NP, the NP–NP interaction, and nuclear export of NP. Although this complex mode of action remains to be verified, it seems to agree with the docking analysis for RK424, which predicted its binding in an NP pocket surrounded by the RNA‐binding groove, NP dimer interface, and nuclear export signal.141 Other proposed NP inhibitors are compound “4” (Fig. 9C) and Ingavirin® (Fig. 9F), a controversial143 influenza inhibitor that is commercially available in Russia. This agent was reported to impair biogenesis and oligomerization of NP in vitro144 and suppress influenza A and B infections in animal models.145, 253 However, since its broad antiviral activity encompasses unrelated respiratory viruses,145 Ingavirin® most likely targets a cellular rather than a virus‐specific component.

Taken together, besides nucleozin, several lead molecules to target NP have been proposed during recent years. Among these, nucleozin is the only inhibitor for which the mechanism has been firmly proven by cocrystallization and resistance analysis.

Less Defined Approaches to Inhibit the Influenza Polymerase Complex

For a number of compounds suggested to act on influenza virus polymerase activity in cell culture, the detailed mechanism of action has not yet been revealed. Compound “367”111 (Fig. 7B) was proposed to target PB1 since mutation H456P in PB1 rendered the virus resistant to “367” and its close analogue “715.” Compound “367” inhibited viral RNA synthesis in the vRNP reconstitution (minigenome) assay, an effect that possibly contributes to its antiviral activity in influenza virus‐infected cell cultures. Another small molecule that may target PB1 is “ASN2” (Fig. 7B). This compound inhibits influenza A virus polymerase function while simultaneously activating the innate immune system by inducing type I interferon expression.112 ASN2 inhibited viral transcription but not all genes were equally affected. It induced preferential downregulation of mRNAs encoded by the smallest genome segments, that is M1, M2, NS1, and nuclear‐export protein (NEP). Loss of NS1 expression resulted in induction of interferon‐I, which may prove beneficial in an in vivo setting. Resistance selection revealed that the antiviral activity of ASN2 is linked to the serine or tyrosine residue at position 499 in PB1.

Before the three‐dimensional structure of PB1 was revealed, it was impossible to analyze whether the identified resistance mutations for ASN2 and “367” (at position 499 and 456 of PB1, respectively) are located close to each other. Using the crystal structures published by Pflug et al.,23 we were now able to establish that both residues are located at the outside of PB1, close to the PB2‐627 domain (Fig. 5, in maroon) and at a distance of 15–20 Å from each other. Both molecules appear relevant not only in terms of drug development, but also as tools to unravel as yet unknown functions of particular regions in PB1.

Compound THC19 (Fig. 7B) was reported to inhibit influenza RNA synthesis and virus replication by targeting the PA subunit, yet its mode of action remains to be established.122 The quinoline‐based compound BPR3P0128 (Fig. 7B) was shown to strongly inhibit viral mRNA synthesis.148 This molecule inhibited binding and cleavage of capped RNA fragments by influenza polymerase complexes in nuclear extracts, yet did not compete with cap structures in an in vitro cap‐binding assay using isolated PB2. Based on this, it was hypothesized148 that BPR3P0128 might target the interaction between PB2 and a host factor, such as the heat shock proteins Hsp70 or Hsp90. Based on NMR‐based fragment screening, Lee et al.149 identified the small molecule DPQ (Fig. 7B) that appears to interact with the influenza vRNA promoter. This molecule was found to inhibit virus replication in a plaque reduction assay.149

Strategies Directed Toward Associated Host Cell Factors

The concept of targeting a host factor to indirectly inhibit influenza virus RNA synthesis is still in its infancy. Like all other viral pathogens, influenza viruses strictly rely on a variety of host factors to support their replication.254 Thus, targeting such a cellular component represents an alternative to traditional drugs that are directed toward a viral factor.255, 256, 257 This approach could diminish the emergence of resistance since the mutation rate of the host cells is intrinsically lower compared to that of the virus, and the virus most probably does not readily adapt itself to use an alternative cellular factor. Obviously, the inherent challenge of all host‐directed drug discovery campaigns is to ensure that inhibition of the cellular protein is not detrimental to the host. Hence, the pharmaceutical industry has been rather hesitant when considering host cell targets to treat virus infections.

As explained below, miscellaneous host targeting approaches for influenza have been published, which, at least conceptually, may be linked to indirect inhibition of the viral polymerase complex. The interpretation is complicated by the fact that many of these molecules have multiple pharmacological (such as anti‐inflammatory) effects. Hence, their therapeutic activity in mouse influenza models may be related to immune‐related besides antiviral effects. Although this deviates from the traditional view on antiviral therapy, it fits with an ongoing call for broader acting or immunomodulatory antiviral agents.258

One explored host‐targeting strategy is to interfere with nuclear transport. All details on the mechanisms for vRNP nuclear import and export can be found in a recent review.14 Inhibition of the cellular CRM1 protein by the cytotoxic agent leptomycin B was already shown several years ago to give nuclear retention of the vRNPs, presumably by disrupting the interaction between NP and the CRM1 nuclear export receptor.259 In 2014, the more selective CRM1‐antagonist KPT‐335 (Verdinexor; Fig. 7C), which is currently in Phase 1 clinical trials,260 was proven to increase survival in influenza virus‐infected mice.147

An alternative approach relates to the heat shock protein Hsp90, which is involved in assembly and nuclear transport of the viral RNA polymerase, possibly as a molecular chaperone for its protein subunits prior to the formation of a mature ternary polymerase complex.261 It was demonstrated that Hsp90 inhibitors (Fig. 7C) such as geldanamycin,150 its analogue 17‐allylamino‐17‐demethoxygeldanamycin150 (17‐AAG), and 3beta‐acetoxydeoxodihydrogedunin,151 impair influenza viral growth by reducing the levels of ribonucleoprotein complexes. Recently, Swale et al.262 revealed how the PA‐PB1 heterodimer assembles with host RanBP5 into a complex, probably being the PA‐PB1 import complex. Targeting the assembly or disintegration of this complex with inhibitors, could be a valuable antiviral strategy.

Another pharmacological class is protein kinase inhibitors to inhibit phosphorylation of a viral protein or crucial cellular factor. The MEK (mitogen‐activated protein kinase kinase) inhibitor U0126 (Fig. 7C) was shown to cause nuclear retention of viral RNPs, impaired function of the NEP, and concomitant inhibition of virus production.263 U0126 proved able to suppress propagation of influenza viruses, including highly pathogenic avian influenza viruses, and was reported to have low cytotoxicity in vitro and in vivo.152, 153 Four other MEK inhibitors (Fig. 7C), PD‐0325901, AZD‐6244, AZD‐8330, and RDEA‐119 (which are all orally available and at least in Phase 1 clinical trials for cancer therapy) demonstrated antiviral activity in vitro as single agents or in combination with oseltamivir.154 Also, the multiple kinase inhibitor WV970 (Fig. 7C) was reported to suppress influenza virus in cell culture through vRNP inhibition.155

The NF‐κB signaling pathway is critical for efficient replication of influenza virus, and its inhibition results in reduced virus titers. Mazur et al.156 reported that acetylsalicylic acid (aspirin; Fig. 7C) can efficiently diminish influenza virus replication in vitro and in vivo through NF‐κB inhibition, which blocks caspase‐mediated nuclear export of VRNPs. Three other NF‐κB inhibitors, pyrrolidine dithiocarbamate, SC75741, and Bay 11–7082 (Fig. 7C), were reported to significantly decrease influenza virus‐induced disease in mice.157, 158, 159, 264

Another host cell factor with diverse regulatory functions, the Rac1 GTPase, was demonstrated to have both virus‐supportive as antiviral effects. The Rac1 GTPase inhibitor NSC23766 (Fig. 7C) was found to possess anti‐influenza virus properties by affecting viral polymerase activity in a cell‐based (mini‐genome) assay. NSC23766 was able to reduce virus replication in mice and prolong the survival rate of infected mice.160

CONCLUSION AND PERSPECTIVES

For many viruses such as HIV, hepatitis B virus, hepatitis C virus, and herpesviruses, inhibitors of viral DNA or RNA polymerases or hereto related activities are the cornerstone of current antiviral interventions. For influenza virus, the development of viral polymerase inhibitors has been lagging behind, but presently seems to be at a turning point. During the past years, major progress was made in the structural elucidation of different influenza virus polymerase subdomains, and very recently of the entire heterotrimeric complex. In parallel, significant advances were achieved in unravelling the reaction mechanisms and functioning of the influenza replication machinery. These accomplishments fuel the ongoing search for unique influenza polymerase inhibitors, which has evolved from serendipitous discovery to rational structure‐aided drug design. It can be anticipated that, in the near future, some of these inhibitors will enter the clinic to fundamentally reshape the field of influenza therapy and prevention.