Structures and Functions of Pestivirus Glycoproteins: Not Simply Surface Matters

Pestiviruses, which include economically important animal pathogens such as bovine viral diarrhea virus and classical swine fever virus, possess three envelope glycoproteins, namely Erns, E1, and E2. This article discusses the structures and functions of these glycoproteins and their effects on viral pathogenicity in cells in culture and in animal hosts. E2 is the most important structural protein as it interacts with cell surface receptors that determine cell tropism and induces neutralizing antibody and cytotoxic T-lymphocyte responses. All three glycoproteins are involved in virus attachment and entry into target cells. E1-E2 heterodimers are essential for viral entry and infectivity. Erns is unique because it possesses intrinsic ribonuclease (RNase) activity that can inhibit the production of type I interferons and assist in the development of persistent infections. These glycoproteins are localized to the virion surface; however, variations in amino acids and antigenic structures, disulfide bond formation, glycosylation, and RNase activity can ultimately affect the virulence of pestiviruses in animals. Along with mutations that are driven by selection pressure, antigenic differences in glycoproteins influence the efficacy of vaccines and determine the appropriateness of the vaccines that are currently being used in the field.

The E2 proteins of CSFV and BVDV have been mapped as having similar topologies. Domains DA and DB of BVDV correspond to CSFV domains B/C and D/A, respectively [29]. In another study, domain I was found to correspond to CSFV domains B/C, and domain II mapped to CSFV domains D/A. Domain III does not harbor any antibody epitopes, suggesting that it is not exposed on the viral surface [30]. Collectively, the above studies reveal that the structure of BVDV E2 is similar to that of CSFV E2 (Figure 1).
Viruses 2015, 7 4 forms a highly exposed β hairpin that protrudes into the solvent. Domain III (residues 169-343) consists of a series of three small β-sheet modules (IIIa-c), which together form an elongated domain. The E2 proteins of CSFV and BVDV have been mapped as having similar topologies. Domains DA and DB of BVDV correspond to CSFV domains B/C and D/A, respectively [29]. In another study, domain I was found to correspond to CSFV domains B/C, and domain II mapped to CSFV domains D/A. Domain III does not harbor any antibody epitopes, suggesting that it is not exposed on the viral surface [30]. Collectively, the above studies reveal that the structure of BVDV E2 is similar to that of CSFV E2 (Figure 1).  [25,26]. Domains B/C, comprising residues 1-90, are in red, and domains D/A, comprising residues 91-170, are in yellow; (b) The domains of BVDV E2 modified from El Omari et al. [29]. Domain DA, comprising residues 4-87, is in red, and domain DB, comprising residues 88-164, is in yellow. Domain DC, comprising residues 165-271, is in light blue, and domain DD, comprising residues 272-333, is in dark blue; (c) The domains of BVDV E2 modified from Li et al. [30]. Domain I, comprising residues 1-90, is in red, and domain II, comprising residues 91-168, is in yellow. Domain III, comprising residues 169-343, is in medium blue. The residue numbers are shown above the schematics. The cysteines involved in intramolecular disulfide bridges are marked by numbers sequentially according to disulfide bonds, whereas the cysteines involved in intermolecular disulfide bridges are marked by white asterisks. N-linked glycosylation (NLG) sites are denoted by black diamonds, and an O-linked glycosylation (OLG) site is denoted by a white diamond.

Intramolecular Disulfide Linkage and Dimerization of Glycoproteins
The glycoproteins form disulfide-linked complexes, including an E rns homodimer of 97 kDa, an E1-E2 heterodimer of 75 kDa, and an E2 homodimer of 100 kDa [5].
The E rns protein contains nine cysteines, eight of which form four separate intramolecular disulfide bonds [32], as confirmed by recently reported crystal structures [19]. The four intramolecular disulfide  [25,26]. Domains B/C, comprising residues 1-90, are in red, and domains D/A, comprising residues 91-170, are in yellow; (b) The domains of BVDV E2 modified from El Omari et al. [29]. Domain DA, comprising residues 4-87, is in red, and domain DB, comprising residues 88-164, is in yellow. Domain DC, comprising residues 165-271, is in light blue, and domain DD, comprising residues 272-333, is in dark blue; (c) The domains of BVDV E2 modified from Li et al. [30]. Domain I, comprising residues 1-90, is in red, and domain II, comprising residues 91-168, is in yellow. Domain III, comprising residues 169-343, is in medium blue. The residue numbers are shown above the schematics. The cysteines involved in intramolecular disulfide bridges are marked by numbers sequentially according to disulfide bonds, whereas the cysteines involved in intermolecular disulfide bridges are marked by white asterisks. N-linked glycosylation (NLG) sites are denoted by black diamonds, and an O-linked glycosylation (OLG) site is denoted by a white diamond.

Intramolecular Disulfide Linkage and Dimerization of Glycoproteins
The glycoproteins form disulfide-linked complexes, including an E rns homodimer of 97 kDa, an E1-E2 heterodimer of 75 kDa, and an E2 homodimer of 100 kDa [5].
The E rns protein contains nine cysteines, eight of which form four separate intramolecular disulfide bonds [32], as confirmed by recently reported crystal structures [19]. The four intramolecular disulfide bonds are strictly conserved across all members of the genus Pestivirus, indicating the importance of these disulfide bonds in protein folding and/or function. The ninth cysteine residue, C 171 , although not directly involved in a disulfide linkage, is essential for the homodimerization of E rns and can influence virulence [32,33]. Mutation of C 171 results in a loss of the dimeric state of E rns and reduces its binding affinity for HS, suggesting that the E rns homodimer is crucial for HS binding [34]. Because C 171 is not conserved among pestiviruses, it has been suggested that the E rns homodimerization linkage via this cysteine residue might not be essential for pestivirus viability.
The E1 protein contains six cysteines. In the formation of the E1-E2 heterodimer, substitutions of C 24 and C 94 in CSFV E1 affect the formation of E1-E2 heterodimers and alter virulence [35]. The charged residues K 174 and R 177 in the TMD of E1 and R 355 in the TMD of E2 play key roles in E1-E2 heterodimerization [36]. A recently constructed heterotetrameric model of the E1-E2 assembly suggested that residue C 171 in E1 forms a disulfide bond with residue C 295 in E2, thereby stabilizing the E1-E2 interaction that is required for virus infectivity. The model further confirmed that these charged residues are crucial for E1-E2 heterodimerization [36] because they promote disulfide bonding between E1 C 171 and E2 C 295 [22]. CSFV E2 contains 15 cysteines. An intramolecular disulfide bond forms the structural unit of domains B/C, anchored via a disulfide bond between residues C 4 and C 48 . The structural unit of domains D/A is formed by two disulfide bonds, one between residues C 103 and C 167 and the other between C 129 and C 139 [27]. In the formation of the CSFV E2 homodimer, the last three cysteines at residues C 256 , C 277 , and C 294 in the C-terminus function in mediating homodimerization [26] (Figure 1a). In BVDV E2, all 17 cysteines are involved in disulfide bridges, establishing one inter-and eight intramolecular bonds. The 17th cysteine at residue C 295 forms an intermolecular bond with the dimer partner molecule [29,30] (Figure 1b,c). Although the cysteine residues of BVDV are highly conserved, the E2 protein of BVDV, unlike that of CSFV, contains two additional cysteine residues, C 59 and C 106 . Although the exact functions of these additional cysteines remain unclear, they may be involved in alternative disulfide bonds, thus affecting the structure of the E2 protein of BVDV in a different manner than that of CSFV.

N-and O-Linked Glycosylation of Glycoproteins
Glycosylation is one of the most common types of protein modification, whereby N-linked oligosaccharides are added to specific asparagine residues within the context of the consensus sequence Asn-X-Ser/Thr [37]. The CSFV E rns protein has a high degree of N-linked glycosylation (NLG) and contains seven sites, at residues 2, 7, 11, 65, 95, 143, and 158, contributing to nearly half of the molecular mass of the protein [38,39]. The CSFV E1 protein has three putative NLG sites, at residues 6, 19, and 100 [40]; CSFV E2 contains one putative O-linked glycosylation (OLG) site at residue 75 and six NLG sites at residues 116, 121, 185, 229, 260, and 297 [41] (Figure 1a). BVDV E2 has four NLG sites at residues 117, 186, 230, and 298, one of which is in domain DB, two in domain DC, and one in domain DD [29] (Figure 1b); alternatively, one is located in domain II and three in domain III [30] (Figure 1c).
N-glycosylation may play a role in the transport of E rns through the secretory pathway [38] and may also influence glycoprotein dimer formation, synthesis, and processing [42]. N-glycan moieties on E rns are essential for its ability to bind to double-stranded RNA and to block the induction of interferon (IFN) [43]. Furthermore, the glycosylation statuses of E rns , E1, and E2 affect virulence, as the abolishment of specific NLG sites leads to virus attenuation [39][40][41]. It has been suggested that glycosylation patterns play roles in viral attachment, entry, and/or exit from infected cells [41].
The glycosylation patterns of E rns and E2 affect the induction of immune responses. Neutralizing epitopes of the E rns and E2 proteins are dependent on the presence of glycosylation. Indeed, preventing proper post-translational glycosylation of E rns and E2 has been shown to lead to the synthesis of non-immunogenic proteins and to failure to induce protection against CSFV [44].

Antigenic Structure and Epitopes of Glycoproteins
Although the biochemical and functional properties of the E rns protein have been well characterized, relatively little is known about its antigenic structure, with only linear epitopes being identified thus far. Peptides comprising C-terminal residues 191-227 of E rns are immunogenic when applied as ELISA antigens [45]. By deletion analysis, three overlapping regions, at amino acid positions N 65 -S 145 , W 84 -S 160 , and E 109 -K 220 , have been identified as antigenic regions that can be recognized by pig anti-CSFV sera [46]. Five linear epitopes of E rns -31 GIWPEKIC 38 , 65 NYTCCKLQ 72 , 127 QARNRPTT 134 , 145 SFAGTVIE 152 , and 161 VEDILY 166 -have been mapped and found to contain the conserved residues W 33 , L 71 , Q 127 , N 130 , S 145 , and G 148 , which may be critical for antibody binding [47]. Additional linear epitopes, including 114 CRYDKNTDVNV 124 and 116 YDKNTDVNV 124 , have been identified [48] as containing the binding motif 117 DKN 119 [49]. The domain comprises three linear motifs, 64 TNYTCCKLQ 72 , 73 RHEWNKHGW 81 , and 88 DPWIQLMNR 96 , which are also defined, and two residues, T 102 and D 107 , are crucial for interactions with antibodies [50].
Conversely, the antigenic structure and epitopes of the E1 glycoprotein are still not resolved and remain to be investigated.
For CSFV, the antigenic structure of E2 and its epitopes have been extensively studied. Domains B/C, in which non-conserved epitopes are located, are responsible for antigen specificity among various CSFV strains, whereas domains D/A are relatively conserved [25,51]. Several important conformational epitopes have been defined within E2 [25][26][27][52][53][54][55]. To achieve the correct folding of its four antigenic domains, all of the conformational epitopes of E2 depend on the pairing of six different cysteine residues located in the N-terminal half of the protein [25][26][27] (Figure 1a). In domains B/C, the neutralizing antigenic motifs 64 RYLASLHKKALPT 76 and 82 LLFD 85 are essential for maintaining the structural integrity of conformational epitopes [25,53], with residues E 24 and D 40 being responsible for the antigenic specificities of field strains and residues D 16 and K 72 being responsible for specificity of the LPC vaccine strain [54]. Neutralizing epitopes are also present in domains D/A, and residue R 156 is responsible for the antigenic specificities of different CSFV genotypes [26]. Regarding the C-terminal half of E2, five proximal cysteines at positions 180, 188, 204, 207, and 241 are critical for the structural integrity of the C-terminal conformational epitopes [26]. A neutralizing conformational epitope with the antigenic determinant residue E 213 is also present in the C-terminal region [26].
Despite several studies that have employed epitope mapping using mAbs, no antigenic structural model is available for BVDV E2 thus far. For genotype 1, amino acid positions that are essential for neutralization have been mapped to the N-terminal half of E2 [68]. For genotype 2, three separate neutralizing antigenic domains on E2 have been defined via binding competition assays, but these domains have not yet been mapped [69]. It is presumed that the E2 proteins of both BVDV genotypes exhibit comparable antigenic structures with type-specific epitopes [70]. The immunodominant region of BVDV E2 spans residues 71-74 and includes a key residue at position 72 [68,71]. An antigenic motif in CSFV E2 has also been mapped within this region [53], and residue 72 is also an antigenic determinant [54]. Additionally, a common epitope among pestiviruses has been mapped to domain B in CSFV E2 [55].

Interactions with Cells
Cellular attachment and entry is the first step of viral infection of host cells. E rns plays a role in virus attachment [72], whereas E2 is involved in both virus attachment to and entry into target cells [72], thereby determining the cell tropism of pestiviruses [74,75]. The sequences and structures of E2 proteins are presumed to be involved in pestivirus host specificity at the level of cell entry [76]. E1 and E2 form an E1-E2 heterodimer, which is located in the viral envelope and mediates viral attachment and entry [36,73], whereas E rns appears to be dispensable to the process of cell entry. Moreover, three different positively charged residues, two in the E1 TMD and one in the E2 TMD, are essential for cell entry [36], suggesting that interactions within the E1/E2 TMD complex are essential for pestivirus entry into cells.
E rns and E2 glycoproteins interact with different cell surface receptors [72]. Cell surface GAGs, such as HS, can serve as receptors for E rns [77,78]. A recent study has demonstrated that the laminin receptor (LamR) is a cellular attachment receptor for CSFV E rns [79]. LamR operates as an alternative pathway to the HS pathway [79]. These two molecules are also associated with infection by dengue virus, which is another member of the Flaviviridae family, suggesting that CSFV and dengue virus may use similar mechanisms during viral entry [79]. The cell surface receptor of the BVDV E2 glycoprotein is bovine CD46 [80], and inhibition of BVDV infection by CSFV E2 suggests that CSFV E2 and BVDV E2 share an identical receptor [72]. Indeed, both porcine CD46 and HS were recently shown to be the primary components that drive CSFV attachment and entry [81]. The function of CD46 as a cellular receptor for BVDV is modulated by complement control protein 1 (CCP1), which subsequently promotes entry of the virus [82], and genetic and splice variants of CCP1 determine cell permissivity [83]. The BVDV receptor-binding sites of CD46 have been mapped to two peptides, 66 EQIV 69 and 82 GQVLAL 87 , which are located on antiparallel β-sheets in CCP1. These two peptides constitute a crucial region of a binding platform that interacts with BVDV [82]. The potential host cell binding sites of pestivirus are the regions 101 LAEGPPVKECAVTCRYDKDADINVVTQARN 130 of E rns and 141 AVSPTTLRTEVVKTFRRDKPFPHRMDCVTT 170 of E2 [84].
Cellular β-actin interacts with E2 and functions in both entry and the endocytic pathway [85]; it is also involved in the early replication of CSFV [86]. The domain of β-actin that interacts with E2 has been mapped to amino acids 95-188, and its counterparts on E2 have been mapped to two regions that include amino acids 182-261 and 262-341 [86]. Recently, the cellular membrane protein annexin 2 (Anx2) has been identified as a cellular binding protein for CSFV E2 and shown to impact CSFV attachment and entry, RNA replication, and virion production [87]. A study in swine cells using a yeast two-hybrid system identified several proteins that might serve as potential host binding partners, interacting with a non-linear portion of CSFV E2 [88]. Because many of the identified host proteins are also involved in interactions with other viruses, it has been suggested that these proteins might play a role in pestivirus replication or pathogenesis [88].
Cellular entry of enveloped animal viruses requires fusion between the viral and cellular membranes. E2 is characterized as a class II fusion protein that harbors a fusion peptide, 129 CPIGWTGVIEC 139 , containing a consensus sequence comprised of aromatic and hydrophobic residues between two cysteine residues [89]. As has also been recently confirmed, the fusion peptide is involved in membrane fusion activity and has a critical role in virus replication [90]. BVDV E1 contains a fusion motif, whereas E2 acts as a structural scaffold for E1 [29,30]. Pestivirus entry is dependent on clathrin-mediated endocytosis [91,92], and acidification initiates fusion [93]. During endocytosis, low pH triggers conformational changes that result in insertion of the fusion peptide into the target membrane. In fusion proteins, histidine is assumed to play a role in pH-induced conformational changes [94], and a histidine at residue 70 of BVDV E2, which is conserved among pestiviruses, is exposed on the surface of the domain at the membrane-distal end of the molecule; it may trigger order-disorder transition of this domain at low pH [29]. A recent study has proposed that two different juxtamembrane residues, H 335 and H 336 , might also participate in a pH-sensing mechanism in BVDV [22].
Recent studies have demonstrated that pestivirus infection significantly induces cell autophagy [96,97], which promotes viral replication and maturation in vitro [97]. It has been further identified that the E rns and E2 proteins serve as important regulators in autophagy pathways; conversely, E1 is not involved in this process [98].

Interactions with Other Proteins within Virions
Pestivirus glycoproteins interact with each other by forming disulfide-linked complexes, namely, the E rns homodimer, E1-E2 heterodimer, and E2 homodimer [5]. E1-E2 heterodimers are thought to be a major complex in mature virions [8]. During virus assembly, E2 homodimers are formed early, and E1-E2 heterodimers are formed later, after the release of E1 from the ER chaperone calnexin [99]. The dimerization of pestivirus glycoproteins indicates that intact disulfide bonds are critical for acquiring a stable conformation of E2 monomers [122]. Forcing E2 to adopt a reduced conformation during the process of virus maturation results in protein misfolding and proteasome degradation. In contrast, dimerization of E2 results in a conformation that is resistant to reducing agents and degradation. Furthermore, E1-E2 heterodimers are essential for viral entry and infectivity [36].

Interactions with the Host
Pestivirus glycoproteins can elicit both humoral and cellular immune responses in a host. As discussed above, E2 functions as a major antigen that can elicit neutralizing antibody production that confers protection to the host [8], whereas E rns functions as a secondary antigen during infection [9,10]. Thus, serological diagnoses of CSFV-infected animals are primarily based on the detection of E2-or E rns -specific antibodies [123][124][125]. E2 has also been identified as a target for T-cell activation, which is important for targeting cytotoxic T-lymphocyte (CTL) responses [100][101][102]. As they contain several defined T-cell epitopes, both E rns and E1 proteins have been identified as targets for the cellular immune response [100].
In the family Flaviviridae, the structural E rns protein is unique to pestiviruses [2] in that it harbors an RNase active domain of the T2 RNase superfamily [12]. Monoclonal antibodies (mAbs) that inhibit the RNase activity of E rns tend to neutralize virus infectivity, which suggests that the RNase activity of E rns plays a role in the CSFV life cycle [126]. This RNase activity of E rns can induce apoptosis in lymphocytes [103] and can block the synthesis of type I IFN, which is induced by viral single-stranded and double-stranded RNA [104,105]. As the blockage of IFN induction occurs during the initial step of Toll-like receptor 3 signaling [43], pestivirus E rns plays a central role in evading the host's IFN response and favors the establishment and maintenance of persistent infection [106]. Furthermore, a previous study has demonstrated that the RNase activity of E rns can prevent the activation of plasmacytoid dendritic cells (pDCs) by CSFV-infected cells [107]. As pDCs are the most potent source of type I IFN during the early phases of viral infection, this important finding identifies a novel pathway by which viruses can escape the IFN system.
Recently, the localization of E rns and the mechanism leading to evasion of host innate immunity have been further examined. E rns is taken up by a cell within minutes via clathrin-mediated endocytosis, and this uptake is largely dependent on its C-terminus, which binds to cell surface GAGs. The inhibitory activity of E rns remains for several days, indicating its potent and prolonged effect as a viral IFN antagonist [95].
It is likely that E2 serves as the major pestivirus protein that interacts with cell receptors, whereas E rns serves as an accessory protein that interacts with other cell surface molecules, which in a sense resemble co-stimulatory molecules, to deliver appropriate signaling for viral entry or endocytosis. Inappropriate signaling would likely alter viral replication and govern whether a productive or a persistent infection will ensue.

Genetic Basis of Pestivirus Virulence
The molecular determinants of pestivirus virulence have been defined by reverse genetic technology [112]. It appears that seven proteins of pestiviruses, including its three glycoproteins, play roles in virulence [113]. As will be reviewed below, the majority of events that occur on the surface of viable virions, such as the RNase activity of E rns , variations in amino acids and antigenic structures, and altered patterns of glycosylation and dimerization, can profoundly affect virulence in animals or cells.
Mutations in E rns that abrogate RNase activity in CSFV lead to virus attenuation [21,114]. CSFV becomes attenuated after abolishing the NLG site in E rns by residue N 2 substitution [39]. Additionally, mutation of C 171 in E rns prevents homodimerization and also leads to the attenuation of CSFV [33].
When 19 amino acids are inserted into the C-terminus of E1, the highly virulent CSFV strain becomes completely attenuated [115]. Additionally, when amino acids N 6 , N 19 , and N 100 of E1 are substituted, thus abolishing the NLG sites, the highly virulent CSFV strain loses its infectivity and becomes attenuated [40]. Substitution of cysteine residues at positions 24 and 94 in CSFV E1 affects E1-E2 heterodimerization and alters virulence [35].
The highly virulent CSFV strain is attenuated when the E2 gene is replaced, suggesting that E2 plays a major role in virulence [116]. Mutations in its C-terminal region can also influence its virulence [117]. The conserved epitope 140 TAVSPTTLR 148 in domain A plays an important role in virulence [118], and the misposition of T 140 (T 141 in CSFV vaccine strain GPE-) in this conserved epitope reduces CSFV virulence by influencing both virus replication efficiency in vitro and viral pathogenicity in pigs [119]. Mutations in two residues in E2, S 74 L and P 279 H, result in attenuation of the virus [120]; however, the E2 L 21 H mutation only attenuates the virus if there are additional mutations in residues R 9 , R 209 , and I 210 in E rns [121]. These mutations likely affect the interaction between E2 and E rns during membrane anchoring. Glycosylation of E2 also influences virulence in swine, with residue N 116 being involved in attenuation of the virulent parental virus and residue N 185 being critical for virus viability [41].
Considering the types of cells that are susceptible to CSFV, it is unlikely that the attenuation in pathogenicity that is discussed above involves a change in cell tropism (i.e., "altered" tropism). Rather, it is more likely the result of "reduced" tropism, meaning that CSFV tropism of cell types remains unchanged with its attenuation, whereas its affinity or intensity is reduced. A similar phenomenon exists for CSFV live attenuated vaccine virus, which presents within the animal body in similar cell types but at a reduced intensity and for a shorter period of time post-infection. However, it is anticipated that the live attenuated vaccine virus contains alterations in glycoproteins as well as in other proteins.

Antigenic Differences Influence the Efficacy of E2 Subunit Vaccines
Because the genotypes of vaccine viruses (mostly genotypes 1.1 and 1.2) are different from those of the currently prevalent field viruses (genotype 2.1) [127], it is critical to clarify how antigenic differences influence cross-protection between vaccines and field isolates. Recombinant E2 proteins are effective against challenge with genotypically homologous CSFV strains [9,[128][129][130][131][132] but do not offer complete protection against heterologous strains [9,133,134]. CSFV genotype-specific pig antisera bind to heterologous E2 proteins less efficiently than to homologous E2 proteins [135,136]. Additionally, antibodies that are induced by recombinant E2 proteins neutralize genotypically homologous strains better than heterologous strains [28]. All previous studies have indicated that the antigenic variation of E2 among CSFVs is crucial to cross-neutralization, which may explain the incomplete E2 vaccine protection with respect to heterologous strains [133].

Selection Pressure-Driven Mutations Determine the Appropriateness of Vaccines
Vaccination may affect strain diversity and immune escape through recombination events between vaccine strains and wild strains and through point mutations. Additionally, vaccination may influence the population dynamics, evolutionary rate, and adaptive evolution of CSFV [137].
The positive selection pressure that acts on the E rns , E1, and E2 envelope protein genes of CSFV has been studied to identify specific codons that are subjected to diversification. The selection for diversity likely occurs via two mechanisms, one of which is cell independent and the other of which is cell dependent. Selections of random mutations (e.g., 2.1 × 10 −2 nucleotide substitution/site/year [138]) occur when extracellular virions are confronted with immunity, such as antibody-or cell-mediated immunity. Conversely, the selection of mutants in intracellular or cell membrane-associated virions or viral proteins would likely occur at an enhanced rate and via a more complicated mechanism because intracellular signaling would be involved in driving viral mutations, requiring at least one cycle of replication to correct lesions (mutations) and perpetuate viral diversification. No evidence for positive selection has been observed to date in E1.
The positively selected site at residue 209 of E rns corresponds to an amino acid substitution from Ser to Arg that has been found in an HS-binding CSFV variant [108]. Four positively selected sites in E2 at residues 49, 72, 75, and 200 have been identified [108,109]. The mutant at residue 72 is responsible for antigenic specificity [54], and residue 75 is located within an O-glycosylation motif that alters the predicted glycosylation pattern of the protein [108]. Additionally, positive selective pressure has defined six residues at 34, 36, 49, 72, 87, and 88 in domains B/C of E2 [110]. As variation in a single amino acid mutation could substantially affect the antigenicity of E2 [26,54,135], positive selective pressure may influence the cross-neutralization activities of vaccines (see above). Because important antigen-specific residues contribute to neutralization and because the positively selected sites were primarily identified as being located within the highly variable E2 B/C domains [26,54,[108][109][110]135], these domains should represent main targets that are amenable to antigenic evolution under selection pressure imposed by vaccine immunity. These domains are also associated with strong reductions in neutralizing titers of the heterologous virus.
A previous study on the BVDV E2 gene identified five positively selected sites, at residues 194, 196, 213, 252, and 254 [111], all located at the C-terminus of E2. These sites that are found in BVDV are opposite to those found in CSFV, in which positive selected sites are defined at the N-terminus, are surface-exposed, and are therefore prime targets for host antibody responses. These contradictory results suggest that selection to avoid antibody recognition has not been a major factor in BVDV.

Concluding Remarks
As reviewed above, the glycoproteins of pestiviruses clearly possess a limited number of domains, three to four (Figure 1), and it is interesting to note how these glycoproteins can adapt and increase in complexity to fulfill a great number of different functions (Table 1). Pestiviruses contain a maximum of three glycoproteins, including several major proteins, such as E2, and others that serve as accessory proteins, such as E rns and/or E1. Disulfide bonds are important structural components of pestivirus that are involved in the following processes: (1) maintaining the stable intramolecular conformation of E2; (2) dimerization, such as E rns and E2 homodimerization and E1-E2 heterodimerization, which are important during viral entry and infectivity; and (3) increasing the number of epitopes, particularly conformational epitopes, on glycoproteins, such as those on E2. Glycosylation is another important aspect of the pestivirus life cycle, and N-glycosylation is important for the following processes: (1) transport and secretion; (2) the synthesis and processing of glycoproteins; (3) blocking of the induction of IFN, thereby evading innate immunity; and (4) induction of protective neutralizing antibodies. Variations in amino acids and antigenic structures also serve as strategies for increasing pestivirus complexity, either by random or selection pressure-driven point mutations or by recombination. The substitution of amino acids at key positions in pestivirus glycoproteins can profoundly affect (1) antigenic structures that interact with antibodies; (2) abrogation of disulfide bonds, the importance of which is described above; (3) glycosylation; and (4) virion viability. Collectively, pestiviruses' glycoproteins possess multiple functions and play critical roles in virus replication and pathogenicity.