The Flavivirus genus of the Flaviviridae family includes more than 70 related arthropod-borne viruses ^[1][2][3][4][1,2,3,4]. The most common and representative members are the dengue virus (DENV) with four serotypes (DENV-1, -2, -3, and -4), Zika (ZIKV), West Nile (WNV), Yellow Fever (YFV), Japanese-encephalitis (JEV), and tick-borne encephalitis (TBEV) viruses ^[5][6][7][5,6,7]. These are the causative agents for viral hemorrhagic fever and encephalitis in human beings.

Susceptible hosts are normally infected by arthropod vectors carrying flaviviruses ^{[1][8][9][10]}[1,8,9,10]. Many flaviviruses are restricted to a particular region, though occasional spillovers of viruses can occur, which pose a significant challenge for international health care. For examples, WNV jumped from the Middle East into the Americas ^[11][12][11,12]; and ZIKV spread from Africa to Southeast Asia, the islands of Polynesia, and later to Brazil in 2015–2016, causing an epidemic ^[3][13][14][3,13,14].

Specific antiviral therapeutics or vaccines are not available to combat most flavivirus infections, except for YFV, DENV, JEV, Kyasanur forest disease virus, and TBEV [10]. In the case of DENV, vaccine design and development have been significantly impeded because of an antibody-dependent enhancement effect due to the presence of four DENV serotypes, leading to a more severe disease phenotype in subsequent natural infections ^[10][15][10,15]. It remains a challenge for the scientific community to develop a DENV vaccine that simultaneously elicits a balanced tetravalent immunity against all four DENV serotypes. In 2019, the United States Food and Drug Administration (FDA) approved the live-attenuated, tetravalent vaccine, Dengvaxia (from Sanofi Pasteur), but only for individuals living in endemic areas between 9–16 years of age with prior DENV infection ^[15][16][15,16]. These challenges increase the demand to develop therapeutics and vaccines that target different flaviviruses.

2. Flavivirus Genome Organization

The Flavivirus genome consists of a positive-sense single-stranded RNA which is ∼11 kb in length, consisting of a single long open reading frame (ORF) flanked by a 5′ untranslated region (UTR) and a 3′ UTR ^{[3][4][17][18]}[3,4,19,20]. The viral genome is not polyadenylated but capped as in cellular mRNA. The single ORF encodes a PP that is further processed by proteases from the host and the viral NS2B-NS3 complex. The flavivirus protease is highly conserved and essential for virus replication. The viral PP is processed to three structural proteins (Capsid, pr-Membrane, and Envelope) and seven non-structural (NS) proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) ^[4][18][4,20]. Three structural proteins form the virus shell, whereas seven NS proteins participate in the membrane-bound replication complex. Among these NS proteins, only NS3 and NS5 bear enzymatic functions ^[19][20][21][18,21,22] (Figure 1).

Figure 1. Schematic diagram of flavivirus PP organization, processing, and predicted membrane topology of a mature protein. Top: the representative flavivirus genome. c, RNA cap. Middle: schematic representation of the structural and nonstructural proteins within PP. Black arrows denote cleavage by the viral NS2B-NS3 protease complex, whereas the blue arrow indicates cleavage by the Golgi protease and brown arrows denote cleavage by signal peptidase. White blank arrow indicates unknown protease. Bottom: putative membrane topology of PP predicted by biochemical and cellular analyses, and protease cleavage sites (indicated by arrows).

3. Flavivirus Protease

NS3 carries multiple enzymatic functions, including protease, helicase, and triphosphatase activities ^[22][23][23,24]. The NS3 N-terminal domain (amino acids (aa) 1–180) is a trypsin-like serine protease with a classic catalytic triad such as Ser135-His51-Asp75 for DENV-2, JEV, WNV and ZIKV, and Ser138-His53-Asp77 for YFV ^{[3][24][25][26][27][28]}[3,25,26,27,28,29]. A small hydrophilic proportion of NS2B is required as a co-factor for activating the NS3 protease enzymatic function ^[29][30][31][30,31,32]. The NS3 C-terminus domain (aa 180–618) is a superfamily 2 ATPase/helicase that binds to the 3′ end of transient dsRNA and unwinds it in a 3′→5′ direction ^[32][33]. In addition, the C-terminal domain of NS3 also has an RNA 5′-triphosphatase activity to cap viral RNA ^[21][33][22,34].

NS2B is an integral membrane protein of 14 kDa in size, with three domains: a central hydrophilic domain flanked by two transmembrane segments. The central hydrophilic region of 47 amino acids (spanning aa 49–96) interacts with the NS3 protease as a co-factor (Figure 1) ^[31][34][32,35]. It has been reported that flavivirus NS3 protein is catalytically inactive as a protease in the absence of NS2B in either linked ^[35][36][37][36,37,38] or unlinked formats ^{[38][39][40][41][42]}[39,40,41,42,43]. Moreover, the NS3 protease domain expressed in E. coli is not very soluble without its co-factor NS2B. Proteases among flaviviruses have significant similarity on both structural and sequence levels ^[43][44].

The flavivirus NS2B-NS3 protease is one of the most attractive and validated targets for developing a pan flavivirus treatment. Viral proteases in general are highly promising drug targets. Examples include ten HIV-1 protease inhibitors (PIs) ^[44][45][46][45,46,47] and two HCV PIs ^[47][48][49][48,49,50]. Thus, it is plausible that a protease inhibitor for flaviviruses will be efficacious in the clinic. This is also evident by the fact that the NS2B-NS3 proteases from different flaviviruses, such as DENV, ZIKV, and WNV, show a high degree of similarity in their sequences and structures. However, lessons from HIV and HCV drug development suggest that PIs have certain drawbacks. Viral variants resistant to PIs were rapidly developed, because both HIV and HCV cause chronic infections ^[50][51]. In contrast, most flaviviruses cause acute infections, and drug resistance may be less of a concern.

The NS3 active site has been the main focus of developing flavivirus protease inhibitors. However, only limited success has been achieved, possibly due to the flat and featureless active site and requirement of a charged substrate to bind with P1 and P2 sites which denote the first and second positions of amino acids outward from the protease cleavage site to the N-terminus of the protease substrate ^[51][52]. Many of the active site inhibitors are only effective in biochemical assays and show low cellular antiviral activity and poor bioavailability in vivo ^[50][51].

4. Structural Insight of NS2B-NS3 Protease

Structural study of NS2B–NS3 led to better screening assays to identify PIs ^[52][53][53,54]. The NS2B and NS3 protein complex adopts two different conformations [14]. An “open” inactive conformation is present for the NS2B C-terminal portion when substrate or an active-site inhibitor is absent (Figure 2A) ^[14][54][14,55]. However, binding of an inhibitor or substrate triggered a “closed” conformation for the C-terminal portion of NS2B ^[55][56] (Figure 2B). The sequences of the DENV-2 and DENV-3 proteases in Figure 2A and Figure 2B, respectively, are 68% identical. Binding of NS2B to NS3 is required for NS3 function; mutations that disrupt NS2B-NS3 interactions greatly diminish the proteolytic activity of the complex ^[14][56][57][14,57,58].

Figure 2. Crystal structures of the NS2B-NS3 proteases of DENV-2 (A) and DENV-3 (B) in the absence (A) and presence (B) of a substrate analog. (A) The open inactive conformation of the DENV-3 NS2B-NS3 protease in the unbound state (PDB ID of 2FOM). The NS2B cofactor is colored in purple and NS3 protease in yellow. (B) The closed active conformation of the DENV-3 NS2B-NS3 protease in complex with a substrate peptide analog (magenta sticks) (PDB ID of 3U1I). The NS2B cofactor is colored in blue and NS3 protease in orange. Superimposed structure of inactive and active conformation. The catalytic triad His51-Asp75-Ser135 is displayed in stick representation.