Pathogenicity and Structural Basis of Zika Variants with Glycan Loop Deletions in the Envelope Protein

ABSTRACT

The glycan loop of Zika virus (ZIKV) envelope protein (E) contains the glycosylation site and has been well documented to be important for viral pathogenesis and transmission. In the present study, we report that deletions in the E glycan loop, which were recorded in African ZIKV strains previously, have re-emerged in their contemporary Asian lineages. Here, we generated recombinant ZIKV containing specific deletions in the E glycan loop by reverse genetics. Extensive in vitro and in vivo characterization of these deletion mutants demonstrated an attenuated phenotype in an adult A129 mouse model and reduced oral infections in mosquitoes. Surprisingly, these glycan loop deletion mutants exhibited an enhanced neurovirulence phenotype, and resulted in a more severe microcephalic brain in neonatal mouse models. Crystal structures of the ZIKV E protein and a deletion mutant at 2.5 and 2.6 Å, respectively, revealed that deletion of the glycan loop induces encephalitic flavivirus-like conformational alterations, including the appearance of perforations on the surface and a clear change in the topology of the loops. Overall, our results demonstrate that the E glycan loop deletions represent neonatal mouse neurovirulence markers of ZIKV.

IMPORTANCE Zika virus (ZIKV) has been identified as a cause of microcephaly and acquired evolutionary mutations since its discovery. Previously deletions in the E glycan loop were recorded in African ZIKV strains, which have re-emerged in the contemporary Asian lineages recently. The glycan loop deletion mutants are not glycosylated, which are attenuated in adult A129 mouse model and reduced oral infections in mosquitoes. More importantly, the glycan loop deletion mutants induce an encephalitic flavivirus-like conformational alteration in the E homodimer, resulting in a significant enhancement of neonatal mouse neurovirulence. This study underscores the critical role of glycan loop deletion mutants in ZIKV pathogenesis, highlighting a need for global virological surveillance for such ZIKV variants.

INTRODUCTION

Zika virus (ZIKV) is a member of the mosquito-borne flaviviruses. In the first 6 decades since its isolation and identification from the Zika forest of Uganda (UGA) in 1947 (1), only a few human cases with mild clinical outcomes were recorded in African and Asian regions. However, since 2007, ZIKV has caused several large outbreaks in the Pacific islands, leading to the 2015–2016 ZIKV epidemic in the Americas. The recent infections caused by ZIKV are more severe than before. Notably, human ZIKV infections have been linked to fetal microcephaly and neuropathological complications (2–4), such as Guillain-Barré Syndrome in adults (5). Although the knowledge of ZIKV biology has been greatly advanced in recent years, there are still many unanswered questions about ZIKV pathogenesis and evolution.

ZIKV is a single-stranded, positive-sense RNA virus. The approximately 11 kb genome contains a single open reading frame which encodes viral structural and nonstructural proteins (C, prM/M, E, and NS1-NS5). Similar to other closely related flaviviruses such as dengue virus, West Nile virus (WNV), yellow fever virus, Japanese encephalitis virus (JEV), and tick-borne encephalitis virus (TBEV), the structural proteins along with the genomic RNA form viral particles and participate in the attachment, as well as in the entry of the virus into host cells. The nonstructural proteins play a role in replication, assembly, and evasion of the host immune system (6–8). The E protein is the major viral surface protein, responsible for receptor binding and membrane fusion (9–13), and it contains numerous determinants of virulence, as well as neutralizing epitopes (14). The E proteins of flaviviruses consist of three distinct domains: a central β-barrel-shaped domain I (DI), an extended dimerization domain II (DII), and a C-terminal immunoglobulin-like domain III (DIII) that is connected by short flexible loops (9, 15, 16). The glycan loop (also known as the 150 loop) harboring the potential glycosylation site is located within DI, and the highly conserved fusion loop that is responsible for the fusion of the membranes of the host cell and the virus is located at the tip of DII (17–21). DIII contains the putative receptor-binding site and is thought to bind with virus receptors on the host cell surface (15, 19, 22).

The ZIKV E glycan loop contains a single N-linked glycan at position 154. Interestingly, some of the pre-epidemic ZIKV strains found in Africa did not contain the N154 glycan site. Instead, these pre-epidemic glycosylation-deficient African strains contain a 4- to 6-amino-acid deletion in the E glycan loop (23–27), resulting in the generation of nonglycosylated E proteins. These nonglycosylated ZIKV strains were initially believed to be less virulent than the epidemic glycosylated ZIKV strains. However, there is increasing evidence that the African strains are more virulent than the contemporary glycosylated Asian strains both in in vitro (28, 29) and in in vivo models (30, 31). African ZIKV strains resulted in higher mortality, increased morbidity, and higher viral loads in key tissues, such as the brain and testis, in mouse models (31–33), and were highly trophic toward primitive human placental trophoblasts, including cytotrophoblasts and syncytiotrophoblasts (28). These observations have raised questions about the function of the E glycan loop and the role it plays in ZIKV pathogenesis. The biological effects of specific glycan loop deletions and their relevance to viral virulence remain poorly understood. Of concern is the recent discovery that two ZIKV strains introduced into South Korea from the Philippines contain the same glycan loop deletions (34) as the strain MR766. As such, the biological importance of emerging ZIKV variants with specific E glycan loop deletions requires urgent investigation.

In this study, we reported the glycan loop deletions of E protein in Asian ZIKV strains and traced the origin and evolution of these unique strains by using Bayesian analysis. Then, we created recombinant Asian ZIKV mutants that carried the corresponding glycan loop deletions by standard reverse genetics approaches, and characterized these mutants using multiple established in vitro and in vivo models. Above all, the glycan loop deletion mutants exhibited enhanced neurovirulence in neonatal mouse models. Structural analysis revealed that the glycan loop deletions induced an encephalytic flavivirus-like conformational alteration. Overall, our findings identified the E glycan loop deletions as a critical marker for the enhancement of ZIKV neonatal mouse neurovirulence.

RESULTS

Characterization of E glycan loop deletions identified in ZIKV strains.

We first compared all available ZIKV E protein sequences (n = 646) deposited in GenBank, and we identified a total of 12 ZIKV sequences that contained specific amino acid deletions in the glycan loop of the E protein (Fig. 1A). Of the 12 sequences, the origin of the first 6 can be traced to Uganda. Among these, the first five isolates were isolated from Macaca mulatta and derived from the ancient prototypical MR766 strain, while the sixth was isolated from an unknown source from Uganda in 1963. There were four strains isolated from different African countries, including Nigeria, Central African Republic, and Senegal. The last two strains were recently isolated from Asia.

FIG 1.

Deletions of amino acids in the E glycan loop emerged independently in Asian ZIKV strains. (A) Alignment of amino acid sequences in the E glycan loop region from different ZIKV strains. All ZIKV E sequences were aligned using ClustalW, resulting in the identification of 12 strains that had deletions in the E glycan loop. (B) Maximum-likelihood analysis of the Asian and African lineages of ZIKV. The consensus tree was summarized after 1,000 nonparametric bootstrap replicates using MEGA7.0.26. The glycan loop deletion identified in ZIKV strains of African and Asian lineages are highlighted in green and orange, respectively. (C) A Bayesian phylogenetic analysis of the glycan loop deletion mutants among ZIKV Asian lineages. Maximum clade credibility (MCC) tree of 494 full-length ZIKV genomes derived from the Asian lineage. Bars at the tree node of the South Korean cluster (orange) represent the median and 95% highest posterior density of the height. The verified neurovirulence marker S139N is highlighted in blue box.

Among all the ZIKV strains with deletions, four MR766 isolates contain four amino acid deletions (153-VNDT-156) in the glycan loop, while two MR766 isolates contain six amino acid deletions (155-DTGHET-160). The same six amino acid deletion was also present in the Nigeria strain isolated in 1968 and the Central African Republic strain isolated in 1980. Interestingly, an epidemic strain isolated in Senegal contained almost the same four-amino-acid deletion as the MR766 isolates, except for one additional 153I amino acid deletion (Fig. 1A). Most importantly, the two epidemic strains isolated in South Korea in 2016, contained the same four-amino-acid deletion (153-VNDT-156) in the glycan loop as the African MR766 strain. The five MR766 strains isolated from Macaca mulatta have a long passage history in mouse brain and cell culture. The strain isolated from Uganda in 1963 has an unknown host and passage history. While the four African strains isolated from human (Nigeria) or mosquitoes (Central African Republic and Senegal) had a short passage history, the two Asian strains were directly isolated from human samples without any passage history. The independent emergence of epidemic strains with specific glycan loop deletions in multiple continents suggest a continual evolution of ZIKV during global transmission.

We further performed phylogenetic analysis to trace the origins of these deletion mutants of ZIKV. Clearly, the glycan loop deletion mutants were not clustered together within the African strains, suggesting independent emergence of the deletions (Fig. 1B). Most strikingly, the two South Korean strains with deletions in the glycan loop were clustered together with an isolate from the Philippines and fell within the Asian lineage (Fig. 1B). A further Bayesian analysis dated the emergence of the Korean cluster with glycan loop deletion to around December 2015 (Fig. 1C), which was about 2 years after the emergence of the S139N mutation in the prM protein (35), suggesting a recent independent origin of this cluster.

The E glycan loop deletions generated nonglycosylated ZIKV.

Although the emergence of deletions in the E glycan loop has been suggested to influence the virulence of the African ZIKV strain MR766 (24), there is no experimental evidence to support this hypothesis. Considering the evolutionary importance of these glycan loop deletions, we sought to determine the biological function of these glycan loop deletions by reverse genetics. Recombinant ZIKV containing different deletions in the ZIKV E protein, including the deletions 153-VNDT-156 (DEL4) and 153-VNDTGH-158 (DEL6), were designed and constructed based on the infectious clone of ZIKV strain FSS13025 (36) (Fig. 2A). Both glycan loop mutants were successfully recovered in BHK-21 cells. These mutants exhibited similar plaque morphologies to the wild-type (WT) virus (Fig. 2B). Immunostaining of BHK-21 cells infected with the WT, DEL4, and DEL6 viruses using a ZIKV E-specific antibody (Fig. 2C) and growth kinetic analysis in multiple cell lines (Fig. 2D) showed that there was no obvious difference between the DEL4 and DEL6 mutants and the WT virus, suggesting these deletions did not affect viral replication in vitro.

FIG 2.

Generation and characterization of E glycan loop deletion mutants. (A) Schematic representation of the genome of the WT and E glycan loop deletion mutants. (B) Plaque morphologies of the WT, DEL4, and DEL6 viruses. (C) Immunostaining of the ZIKV E protein in BHK-21 cells infected with the WT, DEL4, or DEL6 viruses. (D) Growth curves of the WT, DEL4, and DEL6 viruses in BHK-21, Vero, and C6/36 cells infected at an MOI of 0.01. Viral titers were measured by plaque assays on BHK-21 cells. All data are shown as means ± the SD. (E and F) Glycosidase analysis of the ZIKV E protein. Cell lysates infected with the WT and mutant viruses were treated with PNGase F (E) or Endo H (F) and subjected to Western blotting with an anti-ZIKV E protein MAb.

Since the two mutant viruses did not contain the N154 glycosylation site, the respective E proteins should not be glycosylated. To confirm this experimentally, we conducted deglycosylation tests using the mutant viruses. As shown in Fig. 2E and F, PNGase F and Endo H treatment of lysates from cells infected with the WT virus significantly reduced the molecular weight of the E protein from approximately 53 to 52 kDa. However, the molecular weight of the E proteins encoded by the DEL4 and DEL6 mutants was not affected by either PNGase F or Endo H treatment. These results clearly demonstrated that unlike the E protein of the WT virus, the E proteins of the DEL4 and DEL6 viruses are not glycosylated.

The E glycan loop deletions reduced oral infection of mosquitoes with ZIKV.

Two independent studies have shown that removal of the N154 glycosylation site in the ZIKV E protein prevents oral, but not intrathoracic infection, in mosquitoes (37, 38). Here, we sought to confirm whether the glycan loop deletions conferred a similar phenotype. As shown in Fig. 3A, high levels of ZIKV genomic RNAs were detected in the midguts and carcasses of A. aegypti mosquitoes that received intrathoracic microinjection with either the WT or glycan loop mutant viruses. There was also no significant difference in the relative amounts of viral RNA in the midguts or carcasses of mosquitoes inoculated with the three viruses. In contrast, when the mosquitoes were inoculated via the oral route, the infection rates of the glycan loop mutant viruses DEL4 and DEL6 (6/43 [13.95%] and 17/41 [41.46%], respectively) were significantly lower than those of the WT virus (34/39 [87.18%]) (Fig. 3B). The relative viral RNA loads in mosquitoes infected with the DEL4 and DEL6 viruses were also significantly lower than those infected with the WT virus. In addition, the glycosylation deficient mutant T156I (38) was set as a negative control for these experiments, and T156I totally lost infectivity upon oral infection as expected (Fig. 3B). These results indicated that the glycan loop deletions of ZIKV retained intrathoracic infectivity but reduced oral infectivity to mosquitoes.

FIG 3.

The E glycan loop deletions reduced oral infection of mosquitoes with ZIKV. (A) Female A. aegypti mosquitoes were injected with 60 PFU of the WT, DEL4, or DEL6 viruses by intrathoracic injection. Total RNA from the midguts or the carcass (mosquito body without the midgut) were extracted at 7 days postinfection (n = 20). The viral RNA abundance in these samples was quantitated by qRT-PCR. (B) Female A. aegypti mosquitoes were fed with blood containing 5 × 10⁵ PFU of the WT, DEL4, or DEL6 viruses. The total RNA from each individual mosquito was extracted, and the relative viral RNA abundance was determined by qRT-PCR for the WT (n = 39), DEL4 (n = 43), and DEL6 (n = 41) viruses. n.s., P > 0.05; ****, P < 0.0001.

The E glycan loop deletions reduced ZIKV virulence in adult mice.

Previously, ZIKV encoding a nonglycosylated E protein has been demonstrated to be attenuated and defective in mouse neuroinvasion (39, 40). We first examined the virulence phenotype of the WT and deletion mutant viruses in A129 mice that genetically lack the IFN-α/β receptor. Upon subcutaneous injection of A129 mice, all three viruses readily induced robust viremia. However, the viral loads in the DEL4- or DEL6-infected mice were lower than those observed in WT virus-infected mice at the time points assessed (Fig. 4A). Remarkably, survival curves showed that the DEL4 and DEL6 viruses caused less mortality than the WT virus in A129 mice (Fig. 4B). Compared to the WT virus, DEL6 in particular, showed a highly attenuated phenotype. Few animals developed clinical signs upon intraperitoneal challenge with DEL6 at a dose of 10⁶ PFU. The average survival times of the mice infected with DEL4 and DEL6 were 22.8 and 25.4 days, respectively, which was significantly longer than for mice infected with the WT virus (15.3 days).

FIG 4.

The ZIKV glycan loop deletion mutants DEL4 and DEL6 are attenuated in adult A129 mouse. (A) Viremia profile in WT-, DEL4-, and DEL6-infected A129 mice with 10⁵ PFU viruses via the subcutaneous route (n = 10). Sera were collected on days 1, 3, 5, and 7 postinfection, and viral RNA amounts were quantified using qRT-PCR. Two-way ANOVA (Tukey’s) was used for statistical analysis. (B) The A129 mice were injected with 10⁶ PFU of WT, DEL4, or DEL6 viruses via the intraperitoneal route (n = 10). The survival rate was recorded daily for 30 days. A log-rank test was performed for the survival curve. Tissue samples from the brain (C), liver (D), testis (E), and rectum (F) were collected from the A129 mice infected with 10⁵ PFU viruses via the subcutaneous route at 7 days postinfection (n = 7 or 8). ZIKV genome copies in tissues were determined by qRT-PCR. Dotted lines indicate the detection limit. (G and H) Histopathological examination of brains (G) and livers (H) of A129 mice infected with the WT or DEL4 and DEL6 glycan loop deletion mutant viruses. A129 mice were inoculated subcutaneously with 10⁵ PFU of the WT or mutant viruses, and tissue samples were collected at 7 days postinfection. Hematoxylin and eosin (H&E) staining of brain and liver sections from animals infected with WT, DEL4, and DEL6 viruses or PBS as a control was performed; representative images from the experiment are shown. Scale bars: 100 μm in brain, 250 μm in liver (n = 3). Multiple t tests were used for statistical analysis. The data are presented as means ± the SD (n.s., P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001).

Most importantly, the WT ZIKV efficiently invaded the brain of A129 mice after subcutaneous infection, with the viral RNA loads in the brains reaching 2.6 × 10⁹ RNA copies/g at 7 days postinfection, while those for DEL4 and DEL6 were 1.5 × 10⁸ and 2.6 × 10⁵ RNA copies/g, respectively (Fig. 4C), suggesting that the replication of both mutant viruses was restricted in mouse brains. Histopathological examination of brain sections showed that WT ZIKV infection caused diffuse lymphocytic infiltration and neuronal death, while no obvious lesions and inflammation was seen in the brains of DEL4- and DEL6-infected mice (Fig. 4G). In addition, organ tropism assays using tissue from the testis, liver, and rectum showed that viral RNA loads in DEL6-infected mice were significantly lower than those in the WT-infected group (Fig. 4D to F). Meanwhile, severe lytic necrosis was observed in liver sections from the WT-infected mice, while only slight and scattered lytic necrosis were seen in the DEL4- and DEL6-infected mice (Fig. 4H). Taken together, these results demonstrated that the glycan loop mutant viruses had a highly attenuated phenotype in A129 mice, which was similar to the previously described phenotype of ZIKV encoding a nonglycosylated E protein (39).

The E glycan loop deletions enhanced ZIKV neurovirulence in neonatal mice.

The well-established ZIKV neonatal mouse model has been used to characterize ZIKV neurovirulence and its capability to cause microcephaly (35, 41–43). Here, we further evaluated the in vivo neurovirulence phenotype of the WT and the glycan loop mutant viruses using 1-day-old BALB/c neonatal mice. As expected, upon intracerebral inoculation with the three viruses, all groups of mice showed clinical disease with typical neurological manifestations, including inactivity, motor weakness, and bilateral hind limb paralysis. Remarkably, all DEL4- and DEL6-infected mice succumbed to infection within 14 and 20 days, respectively, while only 54.55% of the WT-infected mice died (Fig. 5A). Significantly, the 50% lethal doses (LD₅₀) of the DEL4 and DEL6 viruses were calculated to be 55 and 28 PFU in BALB/c neonatal mice, values approximately 7- and 14-fold lower than that for the WT virus, respectively (Table 1). In addition, the average survival times of the mice that were inoculated with the mutant viruses were all shorter than that of the WT virus at the same doses (Table 1). Most importantly, the DEL4 and DEL6 viruses caused a more severe microcephalic brain compared to the WT virus (Fig. 5B). Furthermore, the glycan loop mutant viruses replicated more efficiently than the WT virus in mouse brain at 3 and 6 days postinfection (Fig. 5C). Immunostaining of the brain sections also demonstrated more robust viral protein production in the cerebral cortices of mice infected with the glycan loop deletion viruses compared to those infected with the WT virus (Fig. 5D). In addition, histopathological examination of brain sections revealed that the glycan loop deletion mutant virus-infected mice developed severe lymphocytic infiltration and inflammation, with a large amount of degeneration and necrosis of neurons and more microglia appeared to phagocytose the necrotic neurons (Fig. 5E).

FIG 5.

The DEL4 and DEL6 mutant viruses exhibited enhanced neurovirulence in neonatal mice. (A) Survival curves in BALB/c neonatal mice intracerebrally injected with WT, DEL4, and DEL6 viruses at a dose of 1,000 PFU. Analysis of survival was performed by using log-rank tests. (B) Representative images of mouse brains from the infected animals on day 9 postinfection. (Right panel) The ratios of the brain width and the cerebral cortex length were calculated and analyzed. Scale bars, 1 cm. (C) Infectious viral loads in the infected brains (at 3 and 6 days postinfection) were quantified by using a plaque assay. (D) Immunostaining of brain sections taken from the infected animals with an anti-ZIKV E protein MAb. (Right panel) Quantification of ZIKV-positive cells in the cerebral cortex (n = 3). Scale bars, 25 μm. (E) Histopathological examination of the brains of BALB/c suckling mice infected with WT or the glycan loop deletion mutants. P1 BALB/c mice were intracerebrally injected with 500 PFU of virus, and brains were collected at 9 days postinfection (n = 3). Images of H&E staining of brains infected with WT, DEL4, or DEL6 viruses or given PBS as a control are shown. Scale bars: 100 μm in brain. (F) Survival curves in BALB/c neonatal mice intracerebrally injected with WT and N154Q viruses at a dose of 1,000 PFU. (G) Survival curves in BALB/c neonatal mice intracerebrally injected with MR766 and T156I viruses at a dose of 0.01 PFU. The data are presented as means ± the SD (n.s., P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001).

TABLE 1.

Survival analysis of BALB/c suckling mice infected with the WT or the DEL4 or DEL6 mutant^a

Virus	Dose (PFU)	No. of mice		Death rate (%)	AST (days)	LD50 (PFU)
		Survive (n)	Death (n)
WT	10,000	0	7	100.00	12.57	398
	1,000	5	6	54.55	17.18	398
	100	7	3	30.00	20.20	398
	10	9	1	10.00	21.00	398
DEL4	10,000	0	9	100.00	9.67	55
	1,000	0	10	100.00	14.10	55
	100	7	10	60.00	15.17	55
	10	8	2	20.00	19.30	55
DEL6	10,000	0	9	100.00	8.11	28
	1,000	0	8	100.00	9.88	28
	100	6	12	66.67	14.89	28
	10	5	5	50.00	16.00	28

^aGroups of neonatal BALB/c mice were intracerebrally injected with the WT virus or the DEL4 and DEL6 mutant viruses, and mortality was observed for 21 days. The death rates, average survival times (AST), and LD₅₀ values were calculated.

To further clarify the function of glycan loop deletion, we also compared the neurovirulence of two known glycosylation mutants (N154Q based on Asian ZIKV and T156I based on African ZIKV) (37, 38) using the neonatal mouse model described above. As shown in Fig. 5F, the mortality rate of the mutant N154Q was similar to that for the WT at a dose of 1,000 PFU, which is consistent with previous studies (37). Moreover, mice infected with the mutant MR766-T156I showed similar mortality rate to MR766 at a dose of 0.01 PFU (Fig. 5G). The results indicated that the glycosylation mutant itself did not lead to enhanced ZIKV neurovirulence in our model. Taken together, these results demonstrated that the four- and six-amino-acid deletions in the glycan loop of the ZIKV E protein enhanced its neurovirulence and led to a more severe microcephalic brain in neonatal mice.

The E glycan loop deletions induce a “JEV-like” conformational alteration in the ZIKV E protein structure.

Finally, to decipher the possible reasons underlying the enhancement of neurovirulence by the glycan loop deletions, we determined the crystal structures of the soluble ectodomains of the WT and DEL6 ZIKV E proteins (sE) at 2.5 and 2.6 Å, respectively. Structures of both sE proteins have four protein molecules in the asymmetric unit (Table 2 and Fig. 6E). As shown in Fig. 6A, two molecules of E lie antiparallel to form a homodimer and each monomer consists of three domains (DI, DII, and DIII), as expected. The overall structure of the E protein ectodomain from DEL6 shares a high structural similarity with its counterpart from the WT ZIKV with a root mean square deviation (RMSD) of only 0.7 Å between the main chain atoms of the two structures. But, a number of loops, such as the glycan loop in DI, the i-j and k-l loops in DII, and several loops in DIII, assume different conformations in the two structures (Fig. 6A). In line with previous sequence analysis (18, 44, 45), the structures of the DI glycan loop are the most divergent, and this region happens to be variable in other flaviviruses as well. In the DEL6 mutant, the glycan loop, which is shorter by six residues than its counterpart, is observed protruding on the dimer surface, while the glycan loop of the WT E protein made up of a short α-helix and a relaxed loop lies on the dimer surface (Fig. 6A and B). Of note, although ZIKV DEL6 is diverse in amino acid sequence from JEV, the structural conformation of the glycan loop in the DEL6 mutant is more similar to the one observed in JEV than to WT ZIKV, which has the same amino acid sequence of the glycan loop with DEL6 mutant except for six residues deletions (Fig. 6C and D).

TABLE 2.

Data collection and refinement statistics of X-ray crystallographic studies^a

Name	DEL6	WT
Data collection
Resolution (Å)	50.00–2.60 (2.69–2.60)	50.00–2.50 (2.59–2.50)
Unique reflections	55,340 (5,408)	58,187 (5,414)
Space group	P212121	P212121
Cell dimensions
a, b, c (Å)	60.3, 134.1, 215.6	60.6, 134.5, 215.0
α, β, γ (°)	90.00, 90.00, 90.00	90.00, 90.00, 90.00
Redundancy	6.4 (6.2)	3.3 (2.9)
Completeness (%)	99.9 (99.8)	93.9 (88.9)
Rmerge	0.20 (0.79)	0.12 (0.37)
I/σ〈I〉	9.71 (1.60)	8.17 (2.25)
Refinement
Resolution (Å)	2.60	2.50
No. of reflections	55,252	58,078
Rwork/Rfree	0.233/0.288	0.235/0.273
No. of non-H atoms
Protein	12,293	11,913
Mean B-factor (Å2)	39.8	70.0
Ramachandran statistics (%)
Most favored	96.46	91.84
Allowed	3.41	7.92
Outliers	0.13	0.87
RMSD
Bond lengths (Å)	0.010	0.011
Bond angles (°)	1.146	1.280

^aValues in parentheses are for the highest-resolution shell. R_merge = Σ_hklΣ_i|I(hkl)_i − 〈I(hkl)〉|/Σ_hklΣ_iI(hkl)_i. R_work = Σ_hkl |Fo(hkl) − Fc(hkl)|/Σ_hklFo(hkl). R_free was calculated for a test set of reflections (5%) omitted from the refinement. RMSD, root mean square deviation.

FIG 6.

The E glycan loop deletion (DEL6) induces JEV-like conformational alteration. (A) The sE dimer of DEL6 is shown in ribbon representation. DI, DII, and DIII are colored in red, yellow, and blue, respectively. The fusion peptide, i-j, E0-F0, and k-l loops are highlighted in green, magenta, cyan, and purple, respectively. Insets show the enlargements of the glycan loop and the i-j and k-l loops with the structure of the WT (in gray) superimposed. (B) Surface representation of the dimers of sE from the DEL6 (left) and WT (right) viruses. The color scheme is the same as in panel A. The glycan loop is colored in cyan and marked by cyan lines. The perforations are highlighted by black dots. (C) Superimposition of the dimer of sE from WT ZIKV (gray, left) and the dimer of sE from JEV (gray, right) onto the sE dimer of the DEL6. The color scheme for the ribbon presentation of the DEL6 sE dimer is the same as in panel A. The loops that show distinct configurations and contribute to the formation of the perforations on the dimer surface are labeled. The perforations are marked by black dots. (D) Superimposition of the dimer of sE from JEV (gray) onto the dimer of sE from WT ZIKV. The color scheme for the ribbon presentation of the WT sE dimer is the same as in panel A. (E) Superimposition of four chains from DEL6. The structures are refined without NCS to avoid possible bias induced by NCS averaging refinement. A, B, C, and D are colored in green, blue, rose red, and yellow. (F) Space-filling representation of E dimer. The conformational change of i-j loop region between the structures of the WT E protein and its DEL6 mutant led to two symmetry-related perforations on the homodimer surface of the DEL6 mutant.

Another major conformational change between the structures of the WT E protein and its DEL6 mutant was observed around the i-j loop region. This loop, followed by the αB helix has shifted ~7 Å from its original position in the WT protein (Fig. 6A to C), leading to two symmetry-related perforations on the homodimer surface of the DEL6 mutant (Fig. 6F). These perforations are formed due to slight receding of the B₀-C₀, and k-l loops, coupled with the conformational changes in the i-j loop (Fig. 6A to C). Again, these structural features of the sE of the DEL6 mutant are also identified in JEV, yielding similar encephalitis flavivirus-specific perforations on the surface (Fig. 6C), where the putative receptors of neuronal cells might attach (46). These results indicate that the deletions in the E glycan loop induce a “JEV-like” conformational alteration, including the formation of the perforations and an altered conformation of the glycan loop, which probably contribute to the enhanced neurovirulence and microcephaly phenotype in neonatal mice.

DISCUSSION

In the present study, we generated two ZIKV mutants with specific glycan loop deletions in E protein using an infectious clone of an Asian ZIKV strain. Our biological analysis showed that the deletion of the amino acids of the E glycan loop does not affect ZIKV propagation in vitro but results in the production of a nonglycosylated form of the E protein. Further in vivo characterization demonstrated that the DEL4 and DEL6 mutant viruses with reduced oral infectivity to mosquitoes and attenuated phenotype in A129 mice. Previously, using an infectious clone of an African ZIKV strain, Annamalai et al. showed that the deletion of the glycan loop did not affect viral infectivity and replication in vitro, but induced an attenuated phenotype in A129 mice (39), as well as in Ifnar1^−/− mice (40), which is in agreement with our results. However, the DEL4 and DEL6 mutants constructed in our study showed significantly enhanced neurovirulence (Table 1) and resulted in a more severe microcephalic phenotype in neonatal mice (Fig. 5). This inconsistency in neuroinvasiveness and neurovirulence phenotypes may be attributed to the difference in inoculation route and target cells, as well as to the induced immune response in animals. In addition, the glycan loop deletion mutants reduced oral infections in laboratory-raised A. aegypti (UGAL/Rockefeller strain) mosquitoes, which was also used in the glycosylation-specific mutants (37, 38), whether this finding can be directly translated to field mosquitoes in real world remains to be determined. Whatever, ZIKV variants with glycan loop deletion have been isolated from Aedes africanus, Aedes opok, and Aedes luteocephalus mosquitoes in Africa (Fig. 1A). These results highlight a complex and critical role for the E glycan loop deletion during ZIKV infection and pathogenesis.

Strikingly, our high-resolution structural analysis revealed that the specific deletions in the glycan loop of the ZIKV E protein induced a clear change to the topology of the glycan loop and yielded “JEV-like” perforations on the surface of the E homodimer (Fig. 6). Previously, we have demonstrated that encephalitic flaviviruses, e.g., JEV, WNV, TBEV, etc., contain unique structural features that distinguish them from nonencephalic flaviviruses (46). Specifically, mature JEV particles have unusual perforations on the surface, which are surrounded by multiple encephalitic flavivirus-specific motifs implicated in receptor binding (46). The conformational changes arising out of deletion of the glycan loop in the E protein of ZIKV endow ZIKV with additional features similar to the encephalitic form of JEV, which explains the observed biological phenotype alteration in neonatal mouse neurovirulence. This inference is supported by the crystal structure of ZIKV E protein, and the cryo-EM structure of ZIKV mutants is likely to further strengthen the support for our hypothesis. In addition, the clinical outcomes caused by these ZIKV strains with glycan loop deletion warrants further investigation.

Another interesting finding of note in this study is the evolutionary implications for the emergence of ZIKV mutants with deletions in the glycan loop. Although the first glycan loop deletion mutant of ZIKV strain MR766 arose from numerous serial passages in cell cultures and animals, the subsequent deletion mutants were isolated from mosquitoes and patients. Therefore, these mutants should be deemed as a new subtype in nature, occurring during stages of the virus life cycle alternating between mosquitoes and humans. The sudden emergence of ZIKV glycan loop mutants in Asia should arouse special interest about the persistent evolution of ZIKV neurovirulence. Indeed, contemporary ZIKV epidemic strains have accumulated multiple adaptive mutations from their Asian ancestor during their global transmission from Asia to the Americas (47). Previously, we have shown that a single S139N mutation in the prM protein substantially increased ZIKV replication in both human and mouse neural progenitor cells, which leads to a more severe form of microcephaly and higher mortality rates in mice (35). Liu et al. have demonstrated that ZIKV has evolved to acquire a spontaneous mutation at residue 982V in the NS1 protein, which increases the secretion of NS1, enhancing the transmissibility of ZIKV from mouse to mosquitoes (48). Although these glycan loop deletion mutants showed reduced oral infectivity to mosquitoes, the enhanced neurovirulence phenotype warrants extensive investigation and surveillance.

In summary, our study characterizes the biological phenotype of ZIKV mutants with glycan loop deletion and highlight the importance of E glycan loop on ZIKV pathogenesis. Meanwhile, the conformational change in ZIKV E protein induced by glycan loop deletion was potentially related to the observed enhanced neurovirulence phenotype in neonatal mice. Our study highlights the extensive surveillance and urgent investigation are required to monitor emerging ZIKV strains with E glycan loop deletions in Asia and study their effects on human health.

MATERIALS AND METHODS

Cell lines.

Baby hamster kidney fibroblast cell line BHK-21 (ATCC CCL-10), African green monkey kidney cell line Vero (ATCC CCL-81), and Aedes albopictus cell line C6/36 (ATCC CRL-1660) were cultured in Dulbecco modified Eagle medium (DMEM; Gibco/Thermo Fisher Scientific), Minimum Essential Medium α (Gibco/Thermo Fisher Scientific) and RPMI 1640 (Gibco/Thermo Fisher Scientific), containing 10% fetal bovine serum (Gibco/Thermo Fisher Scientific) and 1% penicillin/streptomycin (Thermo Fisher Scientific), respectively. BHK-21 and Vero cell lines were grown at 37°C in 5% CO₂, whereas C6/36 cells were maintained at 28°C. Spodoptera frugiperda (Sf9) cells (catalog no. 11496015; Thermo Fisher Scientific) were cultured in serum-free medium at 27°C on a shaker maintained shaking at 100 rpm.

ZIKV strains and mutant construction.

The WT ZIKV strain FSS13025 (GenBank KU955593.1) was originally isolated from a patient in Cambodia in 2010 and recovered from an infectious cDNA clone (pFLZIKV) (36) and was the backbone for the introduction of the specific amino acid deletions using overlapping PCR methods. In one mutant the amino acids 153-VNDT-156 were deleted from the E glycan loop (DEL4), and in the other were the amino acids 153-VNDTGH-158 (DEL6). In addition, the N154Q mutation was introduced as the glycosylation mutant, as previous described (37). The T156I mutation was introduced into the ZIKV MR766 infectious clone (GenBank LC002520) by site-directed mutagenesis (38). All of the mutants were confirmed by restriction endonuclease digestion and DNA sequencing. The infectious clone plasmids that had been sequence verified were linearized with ClaI (New England Biolabs) and purified by phenol-chloroform extraction. In vitro-transcribed viral RNA was prepared by using a RiboMAX T7 large-scale RNA production kit (Promega) and purified by using a PureLink RNA minikit (Thermo Fisher Scientific). The RNA was then transfected into BHK-21 cells using Lipofectamine 3000 reagent (Thermo Fisher Scientific) according to the manufacturer’s instructions, and the culture supernatants were collected at 48 to 72 h posttransfection. ZIKV stocks were prepared in C6/36 cells, the titers were measured by using a standard plaque-forming assay on BHK-21 cells, and the mutations were confirmed by RT-PCR and DNA sequencing. All viral stocks were stored in aliquots at −80°C until use. The primers used for plasmid construction and verification of the clones are listed in Table 3.

TABLE 3.

Primers used in this study

Primer	Genome positionsa	Sequence (5′–3′)b	Purpose
DEL4/6-1F	14354–14376	TGAGCGAGGAAGCGGAATATATC	Amplification of fragment DEL4/6
DEL4-1R	1416–1460	ATCAGTTTCATGTCCTGTATCATTAACGATCATCCCACTGTGCTG	VNDT deletion
DEL4-2F	1434–1467	GTGGGATGATCGTTAATGATACAGGACATGAAACTGATGAGAATA	VNDT deletion
DEL4/6-2R	2090–2110	TCCAGTTCCAGCATCATCTTG	Amplification of fragment DEL4/6
DEL6-1R	1417–1465	TTCTCATCAGTTTCATGTCCTGTATCATTAACGATCATCCCACTGTGCT	VNDTGH deletion
DEL6-2F	1423–1473	GTGGGATGATCGTTAATGATACAGGACATGAAACTGATGAGAATAGAGCGA	VNDTGH deletion
F2	727–744	AAGGTGAAGCACGGAGAT	Detection of deletion
ZIKV ASF	1193–1209	CCGCTGCCCAACACAAG	Detection of viral RNA
ZIKV ASR	1246–1269	CCACTAACGTTCTTTTGCAGACAT	Detection of viral RNA
ZIKV FAM	1214–1244	AGCCTACCTTGACAAGCAGTCAGACACTCAA	Detection of viral RNA

^aBased on ZIKV FSS13025 GenBank accession no. KU955593.1.

^bDeletion sites are underlined.

Sequence alignment and phylogenetic analysis.

All ZIKV sequences were retrieved from GenBank. A total of 646 ZIKV E protein sequences were used for sequence alignment by MEGA7, and 68 nearly complete genome sequences were selected for phylogenetic analysis. The phylogenetic tree was generated by using the maximum likelihood with MEGA7 (49), a substitution model based on a general time reversible (GTR) model with gamma-distributed rate variation (+G, parameter = 0.3035) and a proportion of invariant sites ([+I], 37.30% sites) was used (BIC = 80516, AICc = 78954, InL = –39340, Invariant = 0.00001, Gamma = 0.2761, R = 7.9626, Freq A = 0.2747, Freq T = 0.2150, Freq C = 0.2183, Freq G = 0.2920), and 1,000 bootstrap replications were performed. All positions containing gaps and missing data were eliminated. In addition, the Bayesian analysis was performed using an uncorrelated log-normal distributed (UCLD) relaxed molecular clock model with a Bayesian skyline coalescent applied in BEAST 1.8.4 (50). The GTR+Γ nucleotide model was used, with four categories for the Γ distribution. A Markov chain Monte Carlo with one hundred million steps was run, with sampling frequency of every 10,000 steps and removal of the first 10% as the burn-in.

Growth curves.

BHK-21, Vero or C6/36 cells were seeded into 24-well plates and infected with viruses at a multiplicity of infection (MOI) of 0.01. At the indicated time points after infection, culture supernatants were collected, and viral titers were determined by plaque assay.

Deglycosylation assay and Western blotting.

BHK-21 cells inoculated with viruses at an MOI of 1 were harvested at 48 h postinfection. Lysate aliquots were treated with peptide N-glycosidase F (PNGase F) or endoglycosidase H (Endo H) or RNase-free water as a control in accordance with the manufacturer’s instructions (New England Biolabs). Proteins were analyzed using 10% denaturing SDS-PAGE and transferred using a TransBlot Turbo Blotting System (Thermo Fisher Scientific) onto a polyvinylidene difluoride membrane. Western blotting was then performed with a mouse anti-ZIKV envelope protein monoclonal antibody (MAb; 1:1,500 dilution, catalog no. BF-1176-56; BioFront Technologies), and finally developed using an Immobilon western chemiluminescent HRP substrate (1:10,000 dilution, catalog no. WBKLS0500; Millipore) according to the manufacturer’s recommendation.

Indirect immunofluorescence assay.

BHK-21 cells grown in 24-well plates containing a 1-cm² cover slip were infected with viruses at an MOI of 0.01. At 24, 48, or 72 h postinfection, the coverslips containing infected cells were removed and directly used for indirect immunofluorescence assay. The cells on the coverslips were fixed in acetone-methanol (3/7 [vol/vol]) at −20°C. The fixed cells were incubated with mouse anti-ZIKV E protein MAb (1:1,500 dilution, catalog no. BF-1176-56; BioFront Technologies) at 37°C for 1 h and later washed three times with phosphate-buffered saline (PBS). They were then incubated with secondary antibodies conjugated to Alexa Fluor 488 (anti-mouse IgG, 1:200 dilution; Zsbio) in PBS for 1 h at 37°C and washed again as described above. For cell nuclei staining, DAPI (4′,6′-diamidino-2-phenylindole; 0.5 ng/μL) was added to the cells, followed by incubation for 10 min at room temperature.

Mosquito infection experiments.

A. aegypti (UGAL/Rockefeller strain) mosquitoes were maintained in the laboratory as described previously (51). Adults were fed a diet consisting of water and 10% (wt/vol) sugar solution. For oral infections, 5- to 6-day-old female mosquitoes were starved for about 24 h before ZIKV oral infection. Viruses were diluted to a final concentration of 5 × 10⁵ PFU/mL in 50% mouse blood and 50% DMEM before feeding. The mosquitoes were then allowed to suck the viral mixtures at 37°C through a thin Parafilm for 30 min. Subsequently, the mosquitoes were chilled at 4°C for 10 min 7 days postinfection, total RNA from the whole body of a single mosquito was extracted using the PureLink RNA minikit (Thermo Fisher Scientific) according to the manufacturer’s instructions, and the viral RNA level was detected by qRT-PCR (52) (see Table 3 for a primer and probe list). The qRT-PCR was performed using a One Step PrimeScript RT-PCR kit (TaKaRa) and a LightCycler 480 System (Roche). The viral RNA level was normalized against the reference gene ribosomal protein S7 (RPS7). Totals of 39 to 43 mosquitoes were tested from each group.

For intrathoracic microinjection, 5- to 6-day-old female mosquitoes were intrathoracically injected with 60 PFU of virus diluted in 200 nL of DMEM. Then mosquitoes were fed with water and 10% sugar solution. At 7 days postinfection, total RNA from 20 midguts or carcasses (mosquito bodies without the midguts) were extracted individually and subjected to qRT-PCR assay as described above.

Mouse experiments.

All experimental procedures were approved by the Animal Experiment Committee of Laboratory Animal Center, Academy of Military Medical Science (AMMS). All animal experiments were performed in strict accordance with the guidelines of the Chinese Regulations of Laboratory Animals (Ministry of Science and Technology of People’s Republic of China) and Laboratory Animal-Requirements of Environment and Housing Facilities (GB 14925-2010; National Laboratory Animal Standardization Technical Committee). Mice were maintained at a constant ambient temperature using a 12-h day/night cycle and fed ad libitum in a specific-pathogen-free facility. The BALB/c and A129 mice (129/SvEv mice deficient in IFN-α/β receptor) used in this study were raised in the Animal Laboratory Animal Center, AMMS.

(i) Adult A129 mouse model. Groups of 4-week-old A129 mice were injected with 10⁵ or 10⁶ PFU of WT or mutant viruses via a subcutaneous or intraperitoneal route. The number of surviving mice was recorded daily for 30 days. Blood was collected at different time points postinfection for determination of viral RNA loads by qRT-PCR. Mouse brain, testis, liver, and rectum samples were collected 7 days postinfection and subjected to qRT-PCR and histopathological analysis.

(ii) Neonatal mouse model. To testing the neurovirulence of ZIKV in neonatal mice, groups of 1-day-old suckling BALB/c mice were inoculated with serial dilutions of the virus via the intracerebral route, and the survival curves were monitored and recorded for 21 days. In the microcephaly model, 500 PFU of ZIKV strains were inoculated in 1-day-old suckling BALB/c mice intracerebrally. The intact brain tissues were collected on day 3, 6, and 9 postinfection and divided into two equal parts at the median sagittal plane: one half was used to detect viral load and virus titers, and the other half was used for immunofluorescence analysis.

Protein production and purification.

The E ectodomain from the WT and DEL6 viruses (residues 1 to 400) were inserted into the pFastBac_Bee vector for transformation into Sf9 cells. The pFastBac_Bee is a vector that has been derived from pFastBac-one. This vector can express recombinant protein with a melittin tag at the N terminus and a 6×His tag at the C terminus. Positive clones were identified by PCR and DNA sequencing and were transformed into Escherichia coli DH5α cells to obtain large amounts of recombinant plasmid DNA. E ectodomains of the WT and DEL6 viruses were then produced using Sf9 cells by infecting 1 L of cells (2 × 10⁶/mL confluence) with 10 mL of P3 viral stock that was obtained according to the manufacturer’s instructions (Invitrogen). The medium containing the secreted protein was collected after 3 to 4 days of incubation and centrifuged for 30 min at 4,000 × g. The supernatant was filtered through a 0.22-μm-pore size filter and then dialyzed against a 100-fold excess of PBS (pH 7.4). The sample was then mixed with pre-equilibrated Ni-NTA agarose (Qiagen) beads at a ratio of 5 mL settled beads per L of culture and stirred at 4°C for 2 h. The target protein was eluted with the elution buffer (20 mM Tris-HCl [pH 8.0], 150 mM NaCl). The protein sample was further purified by ion-exchange chromatography using a HiTrap Q column (GE Healthcare), followed by gel filtration using a Superdex 200 column (GE Healthcare). The purified samples (~95% purity as judged on a Coomassie blue-stained SDS-PAGE gel) were concentrated to ~8 mg/mL for crystallization and stored at −20°C until further use.

Crystallization and data collection.

Crystals of E ectodomain (WT and DEL6) were grown at 16°C using the hanging drop vapor diffusion method (53). Next, 1-μL crystallization drops containing protein mixed with mother liquor (1:1 ratio) were equilibrated over a reservoir solution made up of 0.1 M ammonium citrate tribasic (pH 7.0) and 12% (wt/vol) polyethylene glycol 3350 (WT and DEL6). Crystals were cryoprotected by soaking them in a reservoir solution containing 20% (vol/vol) glycerol prior to flash-freezing in liquid nitrogen. The same crystal conditions of WT and DEL6 used here can eliminate the possibility of flexible E ectodomain adopt various conformations in different crystals forms (54).

Structure determination.

All diffraction data sets were collected at beam line BL17U (wavelength, 1.0 Å) of the Shanghai synchrotron facility with the resolution being 2.5 to 2.6 Å. Crystals belonged to the P2₁2₁2₁ space group. Data sets were processed and scaled using the HKL2000 package (55). Four protein molecules were found in an asymmetric unit with a solvent content of ~55%. The initial structure solution of E was obtained by molecular replacement using the program Phaser v2.1 (56) with the cryo-EM structure of ZIKV (25) as a search template. Manual model building and refinement were performed using COOT and PHENIX (57). The RMSD between the four NCS-related subunits of E from the WT and its DEL6 mutant is <0.2 Å. Structural figures were drawn with the PyMOL Molecular Graphics System (v1.5.0.4; Schrödinger, LLC [DeLano, 2002]). Data collection and refinement statistics are summarized in Table 2.

Statistical analysis.

All data were analyzed using the GraphPad Prism software. Results are expressed as the means ± the standard deviations (SD) or as described in the corresponding legends. Log-rank tests were performed for the survival analysis. For other results, statistical analysis was performed by using a Student unpaired t test or two-way analysis of variance (ANOVA).

Data availability.

The atomic coordinates of the E ectodomain of the WT and DEL6 have been deposited in the Protein Data Bank (PDB) under accession codes 7YW8 and 7YW7, respectively. Other data that support the findings of this study are available from the corresponding author upon request.