Surprises from the crystal structure of the hepatitis C virus NS2-3 protease


  • Potential conflict of interest: nothing to report


Hepatitis C virus is a major global health problem affecting an estimated 170 million people worldwide. Chronic infection is common and can lead to cirrhosis and liver cancer. There is no vaccine available and current therapies have met with limited success. The viral RNA genome encodes a polyprotein that includes 2 proteases essential for virus replication. The NS2-3 protease mediates a single cleavage at the NS2/NS3 junction, whereas the NS3-4A protease cleaves at 4 downstream sites in the polyprotein. NS3-4A is characterized as a serine protease with a chymotrypsin-like fold, but the enzymatic mechanism of the NS2-3 protease remains unresolved. Here, we report the crystal structure of the catalytic domain of the NS2-3 protease at 2.3 Å resolution. The structure reveals a dimeric cysteine protease with 2 composite active sites. For each active site, the catalytic histidine and glutamate residues are contributed by one monomer, and the nucleophilic cysteine by the other. The carboxy-terminal residues remain coordinated in the 2 active sites, predicting an inactive postcleavage form. Proteolysis through formation of a composite active site occurs in the context of the viral polyprotein expressed in mammalian cells. These features offer unexpected insights into polyprotein processing by hepatitis C virus and new opportunities for antiviral drug design.

Lorenz IC, Marcotrigiano J, Dentzer TG, Rice CM. Structure of the catalytic domain of the hepatitis C virus NS2-3 protease. Nature 2006;442:831–835. (Reprinted by permission from Macmillan Publishers Ltd., copyright 2006;


Hepatitis C virus (HCV) is a positive-strand RNA virus belonging to the Flaviviridae family. Translation of the viral RNA yields a polyprotein precursor that is cotranslationally and posttranslationally processed into the mature structural and nonstructural proteins. The structural proteins as well as the p7 polypeptide are released from the polyprotein precursor by the endoplasmic reticulum (ER) signal peptidase. The nonstructural proteins are processed by the NS2-3 protease and the NS3-4A complex. The NS2-3 protease cleaves at the NS2/NS3 site, whereas the NS3 serine protease with its cofactor NS4A is responsible for processing of the downstream nonstructural proteins (Fig. 1). Polyprotein processing is a key step where the virus uses its own replication machinery for the first time.

Figure 1.

Schematic representation of HCV polyprotein processing at the transition between structural and nonstructural proteins. The black diamonds denote cleavages by the signal peptidase. The catalytic domain of NS2 is highlighted in gray. The NS2 active site residues His143, Glu163, and Cys184 are indicated. The N-terminal domain of NS3 which is required for a functional NS2-3 protease and at the same time harbors the distinct NS3 serine protease activity is hatched gray.

The NS2-3 protease, also designated as autoprotease, has been a difficult viral protein to study. It is dispensable for replication of subgenomic replicons and, therefore, is not essential for the formation of a functional replication complex in vitro. However, NS2-3 protease activity is essential for the complete replication cycle in vitro and in vivo.1–3 As are all HCV proteins, NS2 is associated with intracellular membranes, but the membrane topology of NS2 is still controversial.4–6 The N-terminal domain contains at least 1 or possibly 3 transmembrane segments. Recombinant proteins lacking this N-terminal membrane domain have been found to retain cleavage activity.5, 7 The catalytic activity of the NS2-3 protease resides in the C-terminal half of NS2 and the N-terminal one-third of NS3.8, 9 Site-directed mutagenesis has shown that amino acids His143, Glu163, and Cys184 are essential for proteolytic activity.8, 9 The importance of these amino acid residues is consistent with a cysteine protease catalytic mechanism but, puzzingly, NS2-3 activity was found to be stimulated by zinc or certain other divalent metal ions, leading to the hypothesis that NS2-3 might be a metalloprotease.

In the paper under discussion, Lorenz, Marcotrigiano, Dentzer, and Rice report the crystal structure of the catalytic domain of NS2 (amino acid residues 94-217, NS2pro) at 2.3 Å resolution. This work represents a major breakthrough in the field and reveals a number of surprises. The most exciting finding is that NS2pro forms a dimer with 2 composite active sites. Each monomer is composed of 2 subdomains connected by an extended linker (Fig. 2A). The N-terminal subdomain contains 2 antiparallel α-helices and the C-terminal subdomain contains a 4-stranded, antiparallel β-sheet. The 2 extended linkers cross over in the middle so that the N-terminal subdomain of one molecule interacts with the C-terminal subdomain of the other and vice versa. Overall, this domain-swapped dimer resembles a butterfly with 2-fold symmetry along the vertical axis.

Figure 2.

Structure of the NS2-3 protease. (A) Ribbon diagram showing the NS2pro dimer with one monomer in blue and the other in red. The N and C termini are indicated. The N-terminal α-helices are expected to interact with the membrane interface, placing the dimer on the cytosolic side of the phospholipid bilayer. (B) Composite active site of the NS2-3 protease. His143 and Glu163 of one monomer together with Cys184 of the other form one active site. The C-terminal Leu217 remains coordinated in the active site after cleavage. The ribbon diagrams were kindly provided by Dr. Ivo Lorenz, The Rockfeller University, New York, NY.

Each active site is composed by critical residues from the 2 monomers, i.e., residues His143 and Glu163 are contributed by one monomer and Cys184 by the other (Fig. 2B). Strikingly, superimposition of the NS2pro active site with the active sites from cysteine and serine proteases of a number of organisms, including viruses (Sindbis virus capsid and poliovirus 3C protease), bacteria (subtilisin), and plants (papain), demonstrated a conserved spatial orientation of the active site residues. This nicely illustrates how enzyme active site geometries remain conserved during evolution. However, none of these related proteases has a composite active site.

The authors point out that their structure represents the postcleavage conformation of the enzyme. The C-terminal residue Leu217 is located near the active site and adopts a similar orientation as inhibitors of previously characterized cysteine and serine proteases. This indicates that the C-terminus of NS2 may exert an inhibitory function after cleavage. Based on this important observation, the authors propose that each NS2 active site catalyzes a single NS2-3 cleavage event and then remains locked, allowing tightly regulated processing. These observations render the perspectives for the design of specific inhibitors of this essential viral enzyme promising (because inhibitors of related proteases are known) and difficult (because the enzyme is locked after a single cleavage event) at the same time.

The NS2pro structure did not reveal any zinc coordination motif. Why then is the activity of the enzyme stimulated by zinc? This is where the NS3 portion of the NS2-3 protease comes into play. NS3 contains in its N-terminal one-third a zinc atom that is essential to stabilize its structure. NS2-3 protease activity is independent of the serine protease activity present in the N-terminal one-third of NS3. Thus, the zinc in NS3 may influence an interaction with NS2 that may be required for the proper conformation of NS2 and correct positioning of the scissile bond. Clearly, a structure of the NS2-3 precursor will be required to resolve this interesting issue.

The dimeric structure with composite active sites was a great surprise with major implications for the processing of the polyprotein precursor and the formation of a viral replication complex. Because dimers may form artificially during the crystallization process, the authors corroborated their unexpected finding by elegant and convincing molecular surface analyses, biochemical studies, and transfection experiments. Indeed, the authors could demonstrate that mixing 2 polypeptides, each containing a single mutation of an NS2 active site residue (His143 to Ala and Cys184 to Ala, respectively), reduced but did not abolish NS2-3 function in cultured cells. This can only be explained by the fact that NS2-3 can form composite active sites in cells.

The work by Lorenz et al. changes the current view of HCV polyprotein processing. Until now, NS2-3 cleavage was thought to occur as a unimolecular reaction in cis, hence the designation autoprotease. However, the requirement for dimerization to form an active NS2-3 protease suggests that NS2-3 cleavage and formation of an active replication complex may be regulated by the concentration of NS2. The authors put forward the interesting hypothesis that such a requirement for a critical concentration of NS2-3 for subsequent processing could delay the initiation of HCV RNA replication until sufficient amounts of active NS3-4A protease are generated. It has recently been shown that the NS3-4A protease cleaves 2 crucial adaptor proteins in innate immune sensing, namely TRIF and Cardif (also known as MAVS, IPS-1, or VISA).10, 11 Thus, a sufficient amount of NS3-4A may be required to antagonize double-strand RNA-induced type I interferon production in an infected cell.

From a structural point of view, it is challenging to conceive the mechanism of NS2 domain swapping in the context of the entire polyprotein. Because signal peptidase cleavage between p7 and NS2 liberates the nonstructural region, NS2 dimer formation would have to occur in the presence of the downstream nonstructural proteins. However, an initial NS2-NS2 recognition may already occur cotranslationally. The N-terminal membrane domain of NS2 may play an important role in this homotypic interaction. In a second step, the NS2 domain swapping would occur to allow NS2-3 active site formation and complete processing of the polyprotein.

As discussed above, NS2 is associated with intracellular membranes. Even though the major determinants for membrane association are located in the N-terminal half of the protein, the authors predict that the catalytic domain itself may associate with membranes through the 2 α-helices in the N-terminal subdomain. This prediction is based on the presence of detergent molecules that were co-crystallized with the protein and that mapped to these helices. Recent data from our laboratories confirm a peripheral membrane association of the NS2 catalytic domain (Pantxika Bellecave, J.G., F.P., and D.M., unpublished data).

To date, little is known about the function of the N-terminal part of NS2 responsible for membrane association. Elucidation of the structure of this segment may yield new insights into additional functions of NS2 which was recently found to be essential for the production of infectious virus, possibly affecting a late step (assembly or release) of the viral life cycle.3 Interestingly, efficient HCV chimeras could be constructed using a fusion site in NS2, and natural intragenotypic and intergenotypic recombinants with cross-over sites in NS2 have been identified in patients.12, 13 Clearly, the availability of complete cell culture systems for HCV will allow to further investigate the different functions of NS2 in the viral life cycle.

In conclusion, the work by Lorenz, Marcotrigiano, Dentzer, and Rice represents a milestone with major implications for our understanding of HCV polyprotein processing and replication complex formation. Aside from these fundamental issues, the 3-dimensional structure of the NS2-3 protease will boost research on this difficult-to-study protein and hitherto neglected target for antiviral intervention.