Virulence difference of five type I dengue viruses and the intrinsic molecular mechanism

Dengue virus (DENV) is the most important vector-borne virus globally. The safe and effective vaccines are still under development and there are no antiviral drugs for DENV induced diseases. In this study, we obtained five DENV1 isolates (DENV1 A to E) from the outbreak of dengue fever in 2014 of Guangzhou, China, and analyzed their replication efficiency and virulence in vitro and in vivo. The results suggested that among the five DENV1 strains, DENV1 B has the highest replication efficiency in both human and mosquito cells in vitro, also causes the highest mortality to suckling mice. Further study suggested that nonstructural proteins from DENV1B have higher capacity to suppress host interferon signaling. In addition, the NS2B3 protease from DENV1B has higher enzymatic activity compared with that from DENV1 E. Finally, we identified that the 64th amino acid of NS2A and the 55th amino acid of NS2B were two virulence determining sites for DENV1. This study provided new evidences of the molecular mechanisms of DENV virulence.


Introduction
Dengue virus (DENV) is currently the most popular mosquito-borne virus and widely spreads in tropical and subtropical regions [1]. A series of symptoms caused by DENV, such as dengue hemorrhagic fever/dengue shock syndrome (DHF/DSS), seriously threaten human health [2,3]. Dengue virus belongs to a single-stranded positive sense RNA virus and lacks of accurate replication correcting system. Virus nucleotide changes will eventually lead to stronger or weaker virus virulence during the long-term process of virus spread. Secondly, host response is induced upon virus infection and the interactions between host and virus also influence the virulence. These two aspects corporately influence the virus pathogenicity and severity of the diseases [4,5]. Therefore, it is of great significance to identify the sites within the virus genome that are associated with virulence and to investigate the interactions between virus and host.
The variations in virus genome are closely related with the virulence. Lots of studies showed that the mutations in virus E protein resulted in the significant changes in virulence of flavivirus [14]. By using the reverse genetic approaches, Prestwood and collaborators found mutations at amino acid 124 and 128 in E protein increased the virulence of DENV [15]. Three substitutions in E protein (196Met!Val, 365Val!Ile,405Thr!Ile) and one in NS3 protein (435Leu!Ser) were reported to be associated with the pathogenesis of neurotoxicity of DENV [16,17]. As a multiple function protein, NS3 has serine protease activity in its N terminal domain, which is required for the cleavage of viral polyprotein and the maturation of viral particles. The NS3 protease cleavage ability will determine the assembling efficiency of viral particles [18,19]. Furthermore, the hypervariable regions of the 3'UTR are also believed to be the virulence determination sites, although the biological relevance is remained to be elucidated [20,21].
In the past 25 years, cases of dengue infection have been reported in Guangdong province of China every year [22][23][24]. The peak of infection happened in 2014, and DENV1 appeared as the major serotype at that year [23,25]. We acquired five DENV1 isolates with different genotypes from Guangdong province among the outbreak of dengue fever of 2014. The differences between these virus strains were studied in in vitro and in vivo models. To further understand the molecular mechanism underlying the virulence difference of five variants, we investigated the ability of the variants antagonizing host innate immune response and the functions of their NS2B3 protease.

The virulence difference among five DENV1 strains
Five different DENV1 isolates, named DENV1A to DENV1E in this study, were isolated from the DENV outbreak of 2014 in Guangdong Province, China. The nucleotide sequences of these five DENV1 strains were determined by high throughput sequencing/assembling approach and submitted to Genebank under the accession numbers MH271402 (DENV1A) to MH271406 (DENV1E).
To test the replication efficiency of these five isolates in mammalian cells, human 293T cells were infected with DENV1 A to E respectively at the same MOI of 0.5. Cells were harvested at 12, 24, 48 and 72 h post infections, and the viral replication efficiency were determined by measuring viral envelope (E) gene mRNA copies, then normalized to human β-actin gene. The results suggested that DENV1B and DENV1C had higher replication efficiency among the five virus strains in 293T cells (Fig 1A). Consistent with the intracellular viral RNA levels, the virus titers in supernatants from DENV1B and 1C infected cells were higher than those from DENV1D and 1E infected cells (Fig 1B).
To test the infection ability of these five strains to mosquito cells, C6/36 cells (origin from Aedes albopictus) were infected with DENV1 A to E. The results suggested that DENV1B also showed higher replication efficiency and produced more viral particles in mosquito cells when compared with other strains (Fig 1C and 1D).
Then we used a sucking mice infection model to test the virulence of these five DENV1 strains in vivo. Three-or four-day-old suckling mice were inoculated intracerebrally with 100 PFU of DENV1 A to E respectively, and the mortality rates were monitored daily. The results suggested that sucking mice started to die at 10 days post DENV1B infection, while this date was postponed to day 15 in case of DENV1E infection. 100% of DENV1B infected mice died at day 15 post infection, but 70% of DENV1E infected mice were survived from infection and recovered (Fig 1E). These data suggested that DENV1B is the most virulent virus, while DENV1E is the weakest. We also quantified the viral loads in infected mouse brains by qRT-PCR and plaque assays. Consistently, the viral loads in the brains of DENV1B infected mice were significantly higher than those from DENV1E infected mice (Fig 1F and 1G).
To further confirm that DENV1B is the most virulent virus while the DENV1E is the weakest, HUVEC and Huh7.0 cells were infected with DENV1B and E, respectively. The results suggested that DENV1B showed significantly higher replication levels at all time points post infection and produced more viral particles in these cells (Fig 1H to 1K). Taken together, these data suggested that DENV1B is the most virulent virus among the five DENV1 strains, while DENV1E is the weakest.

Molecular evolution analysis of five DENV1 strains
DENV-1 can be divided into 5 main genotypes. Genotype II and III have only a few early strains, and I, IV and V are the three major genotypes in circulation [26,27]. In order to understand the origin and relationships of the five DENV-1 viruses, molecular evolution analysis was performed using various bioinformatics approaches. 123 full length DENV1 sequences were downloaded from ViPR (www.viprbrc.org) database. The ORFs of these 123 viruses, together with DENV1 A to E from this study, were analyzed by Multiple alignment program Mafft (https://mafft.cbrc.jp/alignment/server/). The phylogenetic tree showed that DENV1A, B and C belong to genotype I of DENV1 (Fig 2A, and S1 Fig). DENV1A is closely related with the DENV1 viruses that are endemic in Zhongshan (China) (2013) and Shizuoka (Japan) (2014). DENV1 B and C are closely related to LC011948, which is endemic in Chiba, Japan in 2014. Then we speculated that the strains erupted in Guangdong and Japan in 2013-2015 probably came from the same ancestor. DENV1D and 1E are very similar to each other. They are located in a branch of DENV1-V genotype, share homology with DENV-1 (KX380801 and KX380796) isolated from Singapore in 2014 (Fig 2A, and S1 Fig). In addition, DENV1B and C share high similarities to a DENV1 strain (AB178040.1) isolated from Japan in 2004 (Fig 2B). Then, a DENV1 replicon, DGL2, origin from this DENV1 strain [28] was used as a reverse genetics approach to perform the single point mutations in DENV1B genomes.
Using MEGA software, the ancestral sequences for the five DENV1 viruses were analyzed and 72 amino acid variations between DENV1 A&B&C and DENV1 D&E were identified (S2 Table). At the same time, all the variations in NS2A, 2B, 4A and 4B among DENV1 A to E were showed in Fig 2C-2F.

Antagonizing of IFN signaling by NS proteins from five DENV1 strains
DENV NS proteins were reported to have the ability to suppress IFN signaling, and this activity will contribute to its virulence in mammalian host. To test whether NS proteins from DENV1 A-E have different capacity against IFN signaling, NS2A, 2B, 4A and 4B from all these five strains were cloned and expressed in 293T cells. IFN-β-Luciferase reporter assay suggested that NS2A and 2B from DENV1B showed the highest inhibitory activity against RIG-I directed IFN activation, compared with NS2A/2B from other strains. NS4A and NS4B from DENV1 A&B&C also have higher inhibitory capacity to IFN signaling compared to those from DENV1 D&E (Fig 3A-3D). Consistent with these results, RIG-I induced IFNβ mRNA expression was also dramatically decreased in cells expressing NS proteins from DENV1B when compared with those from other strains (Fig 3E). Type one IFNs binds to interferon receptor and activates the transcription of genes containing an ISRE responsive element in their promoters. We also tested whether NS proteins from different DENV1 strains showed variable capacities to modulate the ISRE activation. The results suggested that NS proteins from DENV1B also showed the highest inhibitory activity on ISRE-Luc activity during RIG-I-N or IFNα stimulation (Fig 3F and 3G). Consistently, RIG-I-N or IFNα induced transcriptions of typical ISG genes, such as IFIT1 and Cig5, were significantly inhibited in DENV1B NS protein expressing cells (Fig 3H and 3I). In line with this, we also confirmed that DENV1B showed a better replication than DENV1E in 293T cells if we treated the cells with IFNα.
To confirm the functions of DENV1B NS2A, we constructed two mutant DENV1 replicon plasmids (NS2A A64V and F159L) based on the DGL2 replicon (Fig 3J) (The NS2A proteins of DGL2 are 100% identical with that of DENV1B). The 64 th Ala (A) and 159 th Phe (F) of NS2A from DENV1B were changed to Val (V) and Leu (L) (from DENV1E), respectively. DNA sequencing results indicated that the point mutation was successfully introduced into DGL2 replicon (Fig 3K). After transfecting these replicons into 293T cells, we found that NS2A A64V mutation significantly impaired the replication efficiency of DENV1 replicon, but F159L substitution only slightly influenced the replication (Fig 3L). These data suggested that NS2A amino acid position 64 is one of the important virulence determinants for DENV1.

DENV1B shows higher replication efficiency in IFNAR1 -/mice and cells
Since NS proteins from DENV1B have higher capacity to inhibit IFN signaling, we wondered whether this is a major factor that determining the virulence of these five DENV1 strains. IFNAR1 -/mice were introduced to study the replication efficiency of DENV1 B and E in IFN non-responsive system. MEF cells from wild type and IFNAR1 -/mice were obtained and infected with DENV1 B and E, respectively. Surprisingly, DENV1B still has higher infection efficiency than DENV1E in IFNAR1 -/-MEFs, just like what it does in wild type MEFs (Fig  4A-4D). To further confirm this, IFNAR1 -/mice were challenged with DENV1B and E via intraperitoneal infection for 3 days, then the viral loads in blood and spleens were tested by qRT-PCR and plaque assay. The results suggested that the viral load in blood and spleen samples from DENV1B infected IFNAR1 -/deficient mice were significantly higher than that from  (Fig 4E-4H). These data suggested that the difference in antagonizing IFN signaling is not the only determination factor for the virulence of these DENV1 strains.

NS2B K55R is a virulent determinant for DENV1
The results above remind us that the amino acid variations may not only contribute to the difference in virus-host interaction, but also determine the replication ability of the virus itself. In DENV's life cycle, NS2B forms a complex with NS3, and plays an important role in viral polyprotein procession. We then try to address whether amino acid changes in NS2B will directly influence viral protease function as well as viral replication. The mature forms of NS2B3 protease, which has a 48 amino acids NS2B co-factor domain (48-95 aa) and 180 amino acids NS3 protease domain, have been cloned and expressed in recombinant GST prokaryotic expression system (Fig 5A and 5B). The enzymatic assay suggested that NS2B3 protease from DENV1B showed higher substrate cleavage efficiency than NS2B3 from DENV1E (Fig 5C and 5D). Using site-directed single point mutation technology, we made a K55R mutation in NS2B3 of DENV1B, in which the 55 th amino acid Lys (K) was changed to Arg (R) (which is from were normalized as "100"). The expressions of indicated NS proteins were shown by Western Blots in the lower panels. (E) Co-expressing of four NS proteins (2A+2B+4A+4B) from DENV1A to E suppressed IFNβ mRNA production in RIG-I-N transfected 293T cells. (F-G) Co-expressing NS proteins (2A+2B+4A +4B) from DENV1A to E suppressed RIG-I-N (F) or IFNα (G) induced ISRE promoter activation. (H-I) Co-expressing NS proteins from DENV1A to E impaired ISG mRNA production from RIG-I-N (F) or IFNα (G) stimulated 293T cells. The relative mRNA expression levels of typical ISG genes IFIT1 and  DENV1E NS2B), as well as a R55K mutation to NS2B3 of DENV1E (Fig 5E and 5F). The enzymatic test showed that DENV1B NS2B3-K55R protein had lower enzymatic activity than wild type NS2B3 from DENV1B, while DENV1E NS2B3-R55K showed higher cleavage activity than wild type DENV1E NS2B3 (Fig 5G). We then made a NS2B K55R mutation to DENV1 replicon DGL2 (Fig 5H), and tested its replication efficiency. NS2B-K55R DGL2 replicon showed significantly lower replication efficiency than WT replicon (Fig 5I). These results suggested that NS2B K55 is a virulence determinant that important for DENV1 NS2B3 protease activity and viral replication. While, there are also several amino acid differences in the NS3 1-180 protease domain between DENV1B and DENV1E. Our preliminary data suggested that these mutations may also slightly influence the activity of NS2B3 activity. Further study need be performed to characterize other potential virulence determinants in NS3 protease domain.

Discussion
Viruses are small infectious agents that replicates only inside the living cells of other organisms. The viral genome only encodes a limited number of proteins which are necessary for viral structure and replication. Instead, viruses use the machinery and metabolism of a host cell to complete their life cycles. At the same time, viral infections provoke an immune response that usually eliminates the infecting virus. To counteract the host defense mechanism, many viruses have evolved suppressor proteins to overcome the antiviral responses. So that, the virulence of a virus will be determined by two aspects: one is the ability of virus utilizing or antagonizing host responses, the other is the essential functions of those viral proteins.
A number of studies have reported single mutations in flavivirus protein influence the viral-host interactions, thereby determining the virulence of distinct virus. Yuan L et al. reported that S139N mutation in preM protein significantly increased the neurovirulence of Zika virus (ZIKV), and this could be the reason why ZIKV caused more microcephaly since the outbreak of 2010s [29]. Xia H et al. identified the A188V substitution in ZIKV NS1, which enhancing its IFN antagonizing activity [30]. The N124D and K128E mutations in DENV2 E protein reduced its heparin sulfate binding activity, and weakened the infectivity of mutated viruses [15]. In our current study, we also noticed that DENV1B NS proteins have stronger inhibitory ability against host IFN signaling. The A64V mutation in NS2A impaired the replication of DENV1, suggesting that A64 NS2A is a novel virulence determinant that may influence virus-host interaction.
At the same time, other studies suggested that variations in viral proteins may directly influence the viral protein functions. Some of DENV virulence determinants have been described, most of which locate at the E protein [14]. For example, the N67Q mutation in DENV2 E protein decreased virus growth, and N67 was identified as an important N-glycosylation site for this protein which is critical for viral assemble and budding [31]. Some substitutions in NS1, NS4B, and NS5 proteins were evidenced to increase viral replicative fitness in native mosquitoes [32]. In this study, we also identified two novel virulence determinants in the genomes of DENV1. The K55R substitution in NS2B results in an impaired protease activity of NS2B3, thereby compromised the viral replication efficiency. The 64 th amino acid of NS2A was also important for DENV1 replicon replication. The transmembrane topological structure of NS2A and NS2B [33][34][35], as well as the 3D crystal structure of NS2B3 [36], were reported previously by several groups. By sequence alignment analysis, we found that 55 th amino acid residue of NS2B was located near the end of the first β-strand structure of the NS2B cofactor domain, which could be critical for stabilization of NS3 protease. Mutations made to the 63-65 th amino acids of NS2A displayed a lethal phenotype to DENV2 virus [35]. The 64 th amino acid residue was located in the top of the hinge area of the third transmembrane helix domain which may influence the topology of NS2A. Beside of this, molecular evolution analysis also suggested an A9G variation in Capsid protein between DENV1 B and DENV1 E (S2 Table). In the ViPR database, almost all of the Capsid proteins from DENV1 strains are A9, and only 52 strains which are G9. This suggests that the 9 th Alanine residue may be the dominant virulence loci. There are also a 20nt deletion in 3'UTR region of DENV1 D and E compared with DENV1 A-C, and this deletion may interfere with the stability of a SL1 loop in the 3'UTR, which is critical for the sfRNA generation [21]. Further study will be required to explore those potential virulence determinants for DENV1. We should also mention that even though we have confirmed that NS2A A64V and NS2B K55R mutations in replicons of DENV1B backbone have defect replication efficiencies than wild type replicon, the reverse mutations of these amino acids in a replicon with DENV1E backbone should also be important to support these findings. Further experiments will be performed to address this question.
Taken together, we compared the difference of replication efficiency and virulence of five DENV1 variants. We found that DENV1B is the most virulent virus, and DENV1E is the weakest. We further suggested that the 64 th amino acid of NS2A and 55 th amino acid of NS2B were potential virulence determinants of DENV1, which provided a theoretical basis for better understanding the molecular mechanisms of DENV virulence. It also provides new ideas for investigation of DENV protein function, pathogenic mechanism and novel attenuated vaccine.

Ethics statements
The HUVEC (Human Umbilical Vascular Endothelium Cells) and PBMC (human Peripheral Blood Mononuclear Cells) were obtained from BeNa Culture Collection (Bejing, China). The projects using of human biological specimens were approved by an institutional review board (IRB) of Soochow University.
Animal experiments were conducted according to the Guide for the Care and Use of Medical Laboratory Animals (Ministry of Health, People's Republic of China) and approved by the Animal Care and Use Committee as well as the Ethical Committee of Soochow University (No. SYSK-(S2012-0062)). Virulence difference of five type I dengue viruses were prepared from the mouse embryo using standard protocols [37]. Cells were infected with DENV at a multiplicity of infection (MOI) of 0.5, unless otherwise stated.

Virus, cells and mice
BABL/C and C57BL/6J mice were obtained from Shanghai Laboratory Animal Center (Shanghai, China). IFNAR1 -/mice (in a C57BL/6J background) were prepared by Institute of medical laboratory animal research, Chinese Academy of Medical Sciences (Beijing, China). All the animals were maintained in a biosafety level 2 animal facilities.

RNA isolation and qRT-PCR
Total RNA from DENV infected cells were extracted using the total RNA kit I (OMEGA, USA) and reverse-transcribed using the PrimeScript Master Mix kit (TaKaRa, Japan). cDNAs were mixed with RT-PCR primers and SYBR Premix Ex Taq II (TaKaRa, Japan) and amplified for 40 cycles (95˚C 15 s, 60˚C 30 s, and 72˚C 15s). The intracellular viral loads, in terms of transcript levels of the specific viral genes, were quantified through qRT-PCR and normalized to β-actin gene. The mRNA expression levels of human IFNβ1, IFIT1, and Cig5 genes were also determined via qRT-PCR. (Oligo-primer sequences for qRT-PCR of this study were shown in S1 Table).

Virus titration
The titers of DENV in cell-free supernatants or tissue extracts were determined with a median tissue culture infective dose (TCID50) assay and plaque assay according to protocols previously described [38,39], with slight modifications. Briefly, samples were serially diluted and inoculated into Vero cells in 96-well plates. After 5-day incubation, cells were fixed with 4% paraformaldehyde, stained with 10% crystal violet buffer, and examined for cytopathic effects (CPE) and plaque formation under a light microscope. The virus titer (TCID50/ml) was calculated using the Reed-Muench method. 1 TCID50/ml was equivalent to 0.69 pfu/ml [40].
For the Gaussia luciferase assay, 50 ng of DGL2 replicon plasmid was transfected into 293T cells in 96-well plates. Culture supernatants were collected at different time points and luciferase was measured using BioLux Gaussia Luciferase Assay Kit (New England Biolabs, UK) according to manufacturer's instructions.

DENV infection of newborn BABL/C mice and adult IFNAR1 -/mice
Each 3-or 4-day-old BABL/C suckling mouse was inoculated intracerebrally with 100 PFU of DENV1 A-E respectively as previously described [41]. Animals were monitored for 21 days to evaluate the morbidity and mortality. The DENV replication levels in cerebrum at day 4 were measured by qRT-PCR method described above.
For infection of IFNAR1 -/mice, 4-6 week-old IFNAR1 -/mice were infected with 1×10 7 PFU of DENV1B or DENV1E respectively by intraperitoneal injection. At day 3 post infection, mice were euthanized and the viral loads in whole blood cells and spleens were determined by qRT-PCR and plaque assays as described above.

Plasmid construction, protein purification and western blot
The ORFs of NS2A, 2B, 4A and 4B from DENV1 A-E were subcloned into a eukaryotic expression vector pcDNA3.1A-His/Myc individually. The expressions of NS proteins were confirmed by western blot using anti-His Antibody (Sigma, USA). Prokaryotic expression plasmids for protease NS2B3 of DENV1 B and E (and NS2B3 mutants) were constructed as described previously [11,18,42]. Briefly, the coding region of NS2B enzyme co-factor domain (48-95 aa) and NS3 protease domain (1-180 aa) were amplified by nested PCR using the primers listed in the S1 Table, and cloned into the pGEX-6p2 bacterial expression vector. Then the pGEX-NS2B3 constructs were transformed into Escherichia coli strain BL21 (DE3) for protein expression. The GST-tagged recombinant proteins were induced with 0.1mM IPTG at 30˚C for 4 h and purified using GST affinity agarose (GE Healthcare, Sweden). The GST tag was removed by Prescission Protease (Sigma, USA).

Luciferase reporter assays
IFNβ-or ISRE-Luciferase reporter assay was performed as described previously [39,43]. Briefly, 293T cells were transfected with IFNβ-Luc (or ISRE-Luc) (Firefly luciferase, experimental reporter, 100 ng/well) and pRL-TK reporter (Renilla luciferase, internal control, 5 ng/well) plasmids (Clontech, USA), IFNβ activator RIG-I-N (the active caspase recruitment domain (CARD) containing form of RIG-I), together with individual NS proteins from DENV1 A-E or vector control. (For ISRE-Luc assay, cells were also treated with IFNα (1000 U/ml) for 6h instead of stimulation with RIG-I-N transfection.) 24 h post-transfection, cells were lysed and the luciferase activity was measured using a Dual Glow kit according to the manufacturer's instructions (Promega, USA).

Statistical analysis
Prism 7 software (GraphPad Software) was used for survival curves, charts and statistical analyses. The significance of results was analyzed using ANOVA followed by Tukey's test for multiple comparisons, Student's t-test (for comparisons between two groups) and Log-rank (Mantel-Cox) Test (for survival data), with a cutoff P value of 0.05.
Supporting information S1 Fig. Phylogenetic tree of 128 DENV1 virus strains based on the sequences of viral genome. Sequences for additional 123 DENV1 strains were obtained from ViPR database (www.viprbrc.org). The sequence alignment was performed by Mafft 7.394 software (https:// mafft.cbrc.jp/alignment/server). The whole genome phylogenetic tree was constructed by Neighbor-Joining method. The viruses belonging to genotype I, IV and V of DENV1 were indicated in the phylogenetic tree. (TIF) S1