Asian Zika virus can acquire generic African-lineage mutations during in utero infection

ABSTRACT

The Zika virus 2015 epidemic showed an unusual phenotype for human flaviviruses, specifically fetal infection. We previously showed that in utero inoculation with the Asian Zika virus isolated from the human sample causes persistent infection in porcine fetuses. Here, we characterized the evolution of the Asian Zika virus in the fetal brain and placenta. Interestingly, the Asian Zika virus acquired generic African lineage K101R (A408G) and R1609 K (G4932A) mutations during in utero infection. Both African mutations were nonsynonymous and had a high frequency of nearly 100% in the fetal brain. Then, we synthetically generated the wild-type Asian variant and fetal brain-specific variant with generic African-lineage K101R and R1609 K mutations. In mosquito C6/36 cells, but not in human and pig cells, the fetal brain-specific variant showed higher virus loads compared to the Asian wild-type prototype. While in utero infection with both variants caused comparable virus loads in the placenta and amniotic fluids, fetuses injected with the fetal brain-specific variant had the trend to higher virus loads in lymph nodes. Also, introduced K101R and R1609 K mutations were stable and had high nearly 100% frequency at 28 days after in utero inoculation in both directly injected and trans-infected fetuses. These findings evoke concerns because Zika persists in pig herds and mosquitoes on farms in Mexico. It will be essential to identify how persistent in utero infection affects virus evolution and whether in utero-emerged Zika variants have the potential for shedding into the environment, more efficient transmission, and more aggressive infection phenotypes.

Keywords:

• Zika virus
• flavivirus
• evolution
• mutation
• fetus
• persistence
• transmission

Introduction

Zika virus is a mosquito-borne flavivirus transmitted in humans by Aedes mosquito species or sexually. This virus has been circulating among primates and mosquitoes for decades causing only sporadic human infections [1]. The first outbreaks of Zika virus occurred in 2007 in the Pacific island of Yap, Micronesia [2], and in 2013–2014 in French Polynesia and other southern Pacific islands [3, 4]. In 2015, Zika virus emerged for the first time in Brazil and spread in the Americas [5, 6]. Zika infection in humans may cause headache, myalgia, rash, a self-limiting fever, conjunctivitis, and arthralgia. In addition to these classical clinical manifestations typical to many flaviviruses, Zika virus can cross the placental barrier, infect fetuses, and cause intrauterine growth restrictions, fetal microcephaly, fetal death, neurodevelopmental pathology, and immunopathology in offspring [7–9].

Single mutations in the Zika virus genome may affect infection phenotypes. For example, the substitution of serine-to-asparagine in the precursor membrane protein increased Zika virus infectivity in human and mouse neural progenitor cells, microcephaly in mouse fetuses, and mortality in neonatal mice [10]. The substitution of alanine-to-valine in the nonstructural protein 1 (NS1) enhanced Zika virus NS1 antigenemia in infected mammalian hosts, promoting virus infectivity and prevalence in mosquitoes [11]. Moreover, the same mutation conferred NS1 to inhibit interferon-β induction [12]. Also, the threonine-to-alanine substitution of the Zika capsid protein facilitated the transmission by mosquitoes, infection in human cells and immunodeficient mice [13]. Thus, mutations acquired during Zika virus evolution may affect virulence and epidemic potential. To our knowledge, apart from our two studies in fetuses [14, 15], experimental Zika evolutionary studies in the controlled environment are only conducted in maternal samples of non-human primates and immunodeficient mice [16, 17]. And intra-host and compartment-specific Zika virus evolution in the developing fetus and placenta with immature and specialized immunity is not well studied.

In response to the 2015 Zika epidemic, we developed the fetal pig model for Zika virus infection. This model reproduces aspects of in utero Zika virus infection in humans with persistent infection in the fetal brain, fetal membranes, and placenta for at least 90 days and health sequelae in offspring [14, 18–23]. We reported that an isolated in utero environment is conducive to the emergence of Zika virus variants [14]. We used infectious clones to rescue viruses with low or no genomic heterogeneity and inoculated fetuses to induce in utero infection. Then, using Next-Generation Sequencing (NGS), we identified both unique and convergent single nucleotide Zika variants in the placenta. Some variants that emerged in utero in the porcine model were reported in human and mosquito samples collected during the 2015 epidemic. These findings suggest that at least some of the Zika virus variants that caused human infections during the 2015 epidemic might have initially originated in utero in infected pregnant women or pregnant zoonotic hosts. Altogether, we concluded that in utero environment is conducive to the emergence of new Zika virus variants.

In the previous in utero study, we used the synthetic Asian-lineage Zika virus with low or no genomic heterogeneity [14]. Here, we characterize in utero evolution of the Asian Zika virus strain obtained from the human clinical sample. Afterwards, we synthetically rescued the prototype variant and fetal brain-specific Zika variant with emerged generic African-lineage mutations and characterized infection phenotypes in vitro and in utero.

Materials and methods

Animal experiments were performed following the Canadian Council on Animal Care guidelines for humane animal use and were approved by the University of Saskatchewan's Animal Research Ethics Board. Details of animal experiments, RNA extraction and reverse transcriptase quantitative polymerase chain reaction (RT-qPCR), Zika virus whole-genome NGS library construction and Illumina sequencing, NGS data processing and variant calling using iVar, designation of generic African lineage Zika virus mutations, design and production of the fetal brain-specific Zika virus variant, comparative in vitro and in utero infections kinetics, cytokine assay, and statistics are in supplementary materials and methods, and in Table S1-A, Table S2, Table S3, Table S4, File S1, and File S2.

Results

Asian Zika virus can acquire generic African-lineage mutations during in utero infection

In our previous study [18], inoculation of fetuses with the wild Asian-lineage Zika virus PRVABC59 strain caused persistent infection in porcine conceptuses. We detected Zika virus in samples from both directly inoculated and non-manipulated trans-infected adjacent and distant fetuses. In fetal samples selected for this study, viral loads varied from 4.39–5.84 log₁₀ RNA copies/g (mean 5.36 ± 0.58) in the fetal brain and 5.03-7.11 log₁₀ RNA copies/g (mean 6.10 ± 0.75) in the placenta (Table S1-A). These samples showed DNA bands of correct size after amplification with multiplex PCR primers (data not shown), 58.67–92.86% of Zika virus genome NGS coverage, and high average (8,625×) NGS depth (Table S1-A).

Using the reference Zika virus PRVABC59 strain sequence [GenBank: KU501215.1], we identified single nucleotide variants (SNVs) in the wild virus stock used for fetal inoculation, fetal brain, and placental samples (Table S3-A). The wild Zika virus stock had 0.09% of genomic SNVs as calculated from the total number of sequenced nucleotides. In the fetal brain and placenta, the percentage of Zika SNVs varied from 0.04–0.14% and 0.08–0.28% (Figure 1; Table S3-A). In the wild Zika virus stock, mean SNV frequencies ranged from 4.11–73.86%. In the fetal brain and placenta, SNV frequencies were dispersed from 6.37–100% and 3.26–99.99% (Figure 1; Table S3-A).

Figure 1. The percentage and frequency of SNVs in brain and placental samples from fetuses infected with the wild Asian-lineage Zika virus strain PRVABC59. (A) The mean percentage (from two technical replicates) of Zika genomic SNVs in the virus stock, fetal brain, and placental samples was calculated from the total number of sequenced virus genome nucleotides. (B) The mean frequency (from two technical replicates) and subtype of Zika SNVs in virus stock, fetal brain, and placental samples. Solid lines represent mean values. Raw data are shown in Table S3-A.

Infected pig fetuses are from two pregnant pigs – #295 and #296 [18] (Table S1-A). In pig #295, fetuses were inoculated intracerebrally with 4 log₁₀ TCID₅₀ (in 25 μl) of Zika virus. In pig #296, fetuses were inoculated intraperitoneally + intra-amniotic (in 100 μl + 100 μl) with 5 log₁₀ TCID₅₀. Despite different inoculation routes and doses, both directly inoculated and trans-infected fetuses developed distinct but comparable Zika virus loads in the brain and placenta (Table S1-A). Thus, having comparable virus loads and a relatively small number of samples suitable for NGS, we did not subdivide the SNV data into subgroups to specifically analyze evolutionary differences between fetuses inoculated intracerebrally and intraperitoneally + intra-amniotic, and between directly inoculated and trans-infected fetuses. We identified different types of SNVs – nonsynonymous, synonymous, convergent nonsynonymous, and convergent synonymous, in the brain and placenta of fetuses from both pregnant pigs #295 and #296 (Figure 1B).

We identified iSNVs – single nucleotide variants which emerged in utero, by comparing genomic positions of SNVs between sequences from the Zika virus stock used for fetal inoculation and sequences from in vivo samples. Tissue-specific SNVs not present in the Zika virus stock were considered iSNVs. We identified unique patterns of iSNVs in the brain and placental tissues of different fetuses (Figure 2). In utero-emerged mutations located in the Zika genomic regions encoding structural and nonstructural proteins. In addition to unique iSNVs, we identified convergent iSNVs that emerged in different fetuses, suggesting the independent in utero evolution of similar Zika virus genomic features and possibly the biological relevance of tissue-specific iSNVs. Specifically, in three placental samples (295-F14-PL, 295-F15-PL, and 295-F16-PL), we identified convergent nonsynonymous iSNV A4337G (Figure 2; Table S3-A); in the iSNV dataset from these fetuses, the degree of convergent evolution was 9.7%. In the other three placental samples (296-F10-PL, 296-F11-PL, and 296-F16-PL), we identified convergent nonsynonymous iSNV A4962G (Figure 2; Table S3-A); in this dataset, the degree of convergent evolution was 9.7%. Interestingly, another convergent iSNV A4362G was identified in fetuses from different pregnant pigs – 295-F14-PL, 295-F15-PL, and 296-F11-PL, and the degree of convergent evolution was 7.9%. All convergent iSNVs were nonsynonymous.

Figure 2. Zika virus iSNVs in the fetal brain and placenta (the wild Asian-lineage Zika virus strain PRVABC59). Patterns of Zika virus iSNVs in brain and placental tissue samples from individual fetuses. C: Zika virus genomic region encoding capsid protein; prM: precursor membrane protein; E: envelope protein; NS1: nonstructural protein 1; NS2A: nonstructural protein 2A; NS2B: nonstructural protein 2B; NS3: nonstructural protein 3; NS4A: nonstructural protein 4A; NS4B: nonstructural protein 4B; NS5: nonstructural protein 5. UTR: untranslated region. Raw data are shown in Table S3-A.

For comparative in utero and field Zika virus evolution, we compared iSNVs (Figure 2; Table S3-A) to sequences available in public databases. Nineteen iSNVs were reported in 2015–2020 in field Zika virus sequences from Colombia, Brazil, Panama, Puerto Rico, Mexico, Nicaragua, Dominican Republic, Guatemala, Virgin Islands, Guadeloupe, Thailand, Cuba, and Singapore (Table S5-A). Thus, in utero infection with the wild Asian-lineage Zika virus PRVABC59 strain resulted in the emergence of virus genomic variants that at least partially resemble natural Zika evolution during the 2015 epidemic.

Interestingly, two iSNVs – A408G (K101R) and G4932A (R1609 K) – emerged in the brain of the fetus (Wild-296-F15-BR; Table S3-A, highlighted in green; Figure 1B; Figure 2), which was inoculated intraperitoneally + intra-amniotic. These iSNVs had a high frequency of 99.74% and 99.83%. Both iSNVs were nonsynonymous – K101R and R1609 K – and located in Zika genomic regions encoding C and NS3 proteins (Table S3-A; Figure 1B; Figure 2). We classified these two iSNVs as generic African lineage mutations because alignments of C and NS3 protein sequences showed that iSNV K101R is found in all (n = 17) African-lineage Zika virus variants/strains (Figure 3, Table S3-A) and in none (n = 613) of Asian-lineage variants/strains. The R1609 K iSNV is in 6% of African-lineage Zika virus variants/strains (Figure 3, Table S3-A) and in none of Asian-lineage variants/strains.

Figure 3. Generic African-lineage Zika virus mutations. Amino acid motif sequences representing the generic African lineage mutations identified in the fetal brain infected with the Asian Zika virus. X-axis – amino acid position in the Zika virus polyprotein.

We conclude that Asian Zika virus can acquire generic African-lineage mutations during in utero infection.

The fetal brain-specific Zika virus variant shows a more aggressive infection phenotype in C6/36 mosquito cells

To design and produce the fetal brain-specific Zika variant with generic African-lineage mutations, we introduced K101R mutation in the Asian Zika PRVABC59 strain genomic region encoding C protein (A408G in the reference sequence [GeneBank: KU501215.1]) and R1609K mutation in the NS3 genomic region (G4932A). We decided to introduce K101R and R1609K because both mutations had nearly 100% frequency in the Zika virus genome in the fetal brain representing the dominant haplotype. We transfected several 6-well plates containing C6/36 Aedes albopictus mosquito cells with 1, 2, and 3 µg (5 wells for each concentration) of three infectious subgenomic amplicons (ISA) DNA fragments representing whole genomes of the wild-type (WT) Zika variant or fetal brain-specific variant with generic African-lineage mutations (AFM). Interestingly, the recovery of infectious viruses differed between WT and AFM variants: At 1 µg transfection, we did not recover infectious variants. At 2 µg, only one well (20%) out of five transfected wells showed the infectious WT Zika variant; in contrast, three wells (60%) out of five showed the infectious AFM Zika variant. At 3 µg, only three wells (60%) out of five transfected showed the infectious WT Zika variant; in contrast, all five wells (100%) showed the infectious AFM Zika variant. Thus, during the production of synthetic infectious viruses, two variants showed different recovery rates after transfection, and the AFM variant had higher recovery in C6/36 mosquito cells.

Next-generation sequencing showed that two fetal brain-specific African iSNVs are introduced in the AFM stock with nearly 100% frequency (Table S3-C, highlighted in green). Wild-type and AFM stocks had five common SNVs with a low frequency of 4.70-5.21% (Table S3-B, C). The AFM stock also had one unique low-frequency SNV (3.33%) at the genomic position 6,562. These low-frequency SNVs unlikely may affect in vitro and particularly in vivo infection phenotypes because all stock-specific mutations disappeared during in utero infection (Table S3-B, C; also see in utero infection results below). Thus, we synthetically generated WT and AFM virus stocks suitable for comparative in vitro and in vivo studies.

We compared infection phenotypes of WT and AFM Zika variants in three cell lines representing pig, human, and mosquito hosts – Dulac (porcine epithelial kidney cells), HTR-8/SVneo (human trophoblast), and C6/36 (cells isolated from the larva of Aedes albopictus mosquito). Cells were infected at MOI 0.01. In pig Dulac cells, WT and AFM variants had similar infection phenotype (Figure 4A-B) with no difference in the number of virus-positive cells and RNA loads in cell culture supernatant. In human trophoblast HTR-8/SVneo cells, the AFM variant infected a higher number of cells (on average 1.16 times more infected cells) than the WT variant, but the difference was not statistically significant (P = 0.1508). However, AFM RNA loads were significantly higher than WT loads in HTR-8/SVneo cell culture supernatant (P = 0.0079) (Figure 4B). In mosquito C6/36 cells, the AFM variant infected significantly more cells (P = 0.0079) and produced significantly higher virus RNA loads in supernatant (P = 0.0371) (Figure 4A and B).

Figure 4. In vitro infection kinetics of WT and AFM Zika virus strains. All cell lines were inoculated with an MOI 0.01. (A) The 24-well plates with cell monolayers were fixed at 72 h after virus inoculation and stained with Zika virus-specific antibodies, and infected cells were counted in the whole 5 wells with a bright-field microscope. Dots represent the number of cells in the individual well. Columns and whisks represent the mean ± SD. (B) Cell culture supernatants were collected at 72 h after virus inoculation, and Zika RNA loads were measured using RT-qPCR. (C) In vitro infection kinetics of WT and AFM Zika virus strains in porcine Dulac cells at 37°C and 39°C. The dotted horizontal line represents the limit of detection. WT: The wild-type Asian Zika virus variant synthetically rescued with Infectious Subgenomic Amplicons. AFM: The fetal brain-specific Zika virus variant synthetically rescued with Infectious Subgenomic Amplicons. *: P < 0.05; **: P < 0.01.

Pigs, like cynomolgus macaques (Macaca fascicularis) [24], have higher body temperature than humans. The physiological body temperature range in pigs is 38.7–40°C. Thus, we tested whether the generic African-lineage Zika mutations that emerged in the Asian virus genome during infection in pig fetuses convey fitness advantages at higher temperature. Wild-type and AFM Zika variants infected a significantly lower (P = 0.0068 and 0.0014) number of porcine Dulac cells at 39°C than at 37°C (Figure 4C), but the difference between variants was insignificant. Viral RNA levels in cell culture supernatants were similar at different temperatures and between variants (Figure 4C).

Altogether, the fetal brain-specific Zika virus variant shows a more aggressive infection phenotype in C6/36 mosquito cells. But the generic African-lineage Zika mutations that emerged in the Asian virus genome during infection in the pig fetal brain do not convey fitness advantages in porcine cells at higher temperature.

The acquired fetal brain-specific generic African mutations are stable during persistent in utero infection

We inoculated four conceptuses in two pregnant pigs with Zika virus WT or AFM variants to induce in utero infection. In utero inoculation did not cause maternal infection because maternal lymph nodes were negative for Zika virus, and maternal blood did not have virus-specific antibodies. Our previous studies show that Zika virus has specific tropism and persists in the fetal brain, fetal lymph nodes, and placenta with virus shedding to amniotic fluids. In the present study, all fetal brains were negative for the virus. As in previous studies, we found viral RNA loads in the placental tissues, amniotic fluid, and lymph nodes from directly injected and trans-infected fetuses (Figure 5). Fetuses from WT and AFM groups had similar viral loads in the placenta and amniotic fluid. In contrast, fetuses from the AFM group had a trend of high viral loads in lymph nodes; the difference, however, was not statistically significant. But when we removed one outlier lymph node sample with the highest RNA load (5.04 log₁₀ RNA copies/g) in the WT group (Figure 5), the difference was statistically significant (P = 0.0399 with unpaired t-test; P = 0.0398 with Mann–Whitney test). Also, the number of transinfected fetuses with virus-positive lymph nodes is two-fold higher for the AFM group than for the WT group (73% vs. 27%, p = 0.086 for Fisher's exact test; Figure 5).

Figure 5. Zika virus loads in the placenta, amniotic fluid, and fetal lymph nodes. Dotted lines represent the limit of detection. Solid lines represent mean values. Solid green and red circles indicate samples from fetuses directly injected with Zika virus. WT: The wild-type Asian Zika virus variant synthetically rescued with Infectious Subgenomic Amplicons. AFM: The fetal brain-specific Zika virus variant synthetically rescued with Infectious Subgenomic Amplicons.

We also quantified and compared cytokines in the fetal blood plasma. Levels of IFN-γ, IL-1β, and IL-10 were below the detection limits. Fetuses infected with the WT and AFM Zika variants showed significantly higher levels of IFN-α, IL-6, IL-8, IL-12, IL-13, and IL-17A than mock-inoculated fetuses (Figure 6). Fetuses from WT and AFM groups had similar IFN-α, IL-6, IL-8, and IL-13 loads. Fetuses from the AFM group had a trend to high IL-12 and IL-17A loads; the difference, however, was not statistically significant.

Figure 6. Cytokine levels in the fetal blood plasma. IFN-α, IL-6, IL-8, IL-12, IL-13, and IL-17A in blood plasma were detected in the Bioplex cytokine assay. Solid lines represent means. The dotted line represents the limit of quantification. Solid green and red circles indicate samples from fetuses directly injected with Zika virus. WT: The wild-type Asian Zika virus variant synthetically rescued with Infectious Subgenomic Amplicons. AFM: The fetal brain-specific Zika virus variant synthetically rescued with Infectious Subgenomic Amplicons. *: p < 0.05; **: p < 0.01; ***: p < 0.001; **** p < 0.0001. Control blood plasma samples were from the fetuses of two healthy pigs sampled at 78 days of pregnancy (28 days after in utero inoculation with virus-free media) in our previous study [18].

Next, we implemented nearly the entire Zika virus genome NGS in placental samples from three fetuses infected with WT or AFM variants (Table S1-B, C) to compare variant evolution and identify whether the acquired fetal brain-specific generic African mutations are stable during persistent in utero infection. Placental samples showed DNA bands of correct size after amplification with multiplex PCR primers (data not shown), 83.34–92.93% of Zika virus genome NGS coverage, and high average (11,997–13,160×) NGS depth (Table S1-B, C).

Using the reference Zika virus PRVABC59 strain sequence [GenBank: KU501215.1], we identified SNVs in the WT and AFM virus stocks used for fetal inoculation and placental samples (Table S3-B, C; for the AFM variant, two introduced generic African mutations were not considered in this SNV analysis). The WT and AFM Zika virus stocks had 0.05% and 0.08% of SNVs, as calculated from the total number of sequenced nucleotides (Figure 7A). In the placenta from WT and AFM fetuses, the percentage of Zika SNVs varied from 0.04–0.10% and 0.06–0.14% (Figure 7A; Table S3-B, C). In the WT and AFM Zika virus stocks, mean SNV frequencies were low and varied from 4.7% to 5%; stock-specific SNVs were mainly located in the genomic regions encoding NS5 protein (Table S3-B, C). In the placenta from WT and AFM fetuses, mean SNV frequencies varied from 4.05–99.93% and 3.24–99.97% (Figure 7B; Table S3-B, C). We identified different types of SNVs – nonsynonymous, synonymous, convergent nonsynonymous, and convergent synonymous, in placental tissues of WT and AFM fetuses (Figure 7B).

Figure 7. Zika virus SNVs and iSNVs in placental samples from fetuses infected with the wild-type Asian (WT) Zika variant or fetal brain-specific variant with generic African-lineage mutations (AFM). (A) The mean percentage (from two technical replicates) of Zika genomic SNVs in the virus stock and placental samples was calculated from the total number of sequenced virus genome nucleotides. (B) The mean frequency (from two technical replicates) and subtype of Zika SNVs in virus stock and placenta. Solid lines represent mean values. Raw data are shown in Table S3-B, C. (C) Zika virus iSNVs in the fetal placenta. C: Zika virus genomic region encoding capsid protein; prM: precursor membrane protein; E: envelope protein; NS1: nonstructural protein 1; NS2A: nonstructural protein 2A; NS2B: nonstructural protein 2B; NS3: nonstructural protein 3; NS4A: nonstructural protein 4A; NS4B: nonstructural protein 4B; NS5: nonstructural protein 5. UTR: untranslated region. Raw data are shown in Table S3-B, C.

We identified iSNVs – single nucleotide variants which emerged in utero, by comparing the genomic positions of SNVs between sequences from the Zika virus stock used for fetal inoculation and sequences from placental samples. Interestingly, the in-utero environment selected against stock-specific SNVs because they were not identified in placental samples (Table S3-B, C). In utero-emerged mutations located in the Zika genomic regions encoding structural and nonstructural proteins (Figure 7C).

In addition to unique iSNVs, we identified convergent iSNVs that emerged in the placenta of different fetuses, suggesting the independent in utero evolution of similar Zika virus genomic features and possibly the biological relevance of tissue-specific iSNVs. Specifically, in two placental samples from WT fetuses (WT-F13-PL and WT-F16-PL), we identified convergent synonymous iSNV G2503A (Figure 7C; Table S3-B); in this dataset, the degree of convergent evolution was 11.8%.

In two placental samples from AFM fetuses (AFM-F14-PL and AFM-F15-PL) we identified convergent nonsynonymous iSNV A2177G (Figure 7C; Table S3-C); in this dataset, the degree of convergent evolution was 9.5%.

Interestingly, in placental samples of two WT (WT-F13-PL and WT-F17-PL, Table S3-B) and three AFM fetuses (AFM-F12-PL, AFM-F14-PL, and AFM-F15-PL, Table S3-C), we identified convergent nonsynonymous iSNV A3281G; in this dataset from two different pregnant pigs, the degree of convergent evolution was 10.6%.

For comparative in utero and field Zika virus evolution, we compared iSNVs (Figure 7C; Table S3-B, C) to sequences available in public databases. For the WT Zika variant (Table S5-B), ten iSNVs were reported in 2015–2021 in field Zika virus sequences from different countries. For the AFM Zika variant (Table S5-C), ten iSNVs were reported in 2006–2016 in field Zika virus sequences from different countries.

Finally, we identified fetal brain-specific generic African mutations A408G (K101R) and G4932A (R1609 K) introduced into the Asian Zika virus variant in both AFM stock and placental samples from all AFM-inoculated fetuses (Table S3-C, highlighted in green). These generic African mutations had a high frequency of nearly 100% in the stock and tissue samples.

Altogether, the acquired fetal brain-specific generic African mutations are stable during persistent in utero infection, at least for 28 days after inoculation.

Discussion

We studied intra-host compartment-specific Zika virus evolution in fetuses and placenta with developing and specialized immunity. The key finding in the study – the Asian Zika virus can acquire generic African-lineage mutations during in utero infection. And the fetal brain-specific African mutations may convey more aggressive infection phenotypes to the Asian virus. Remarkably, acquired African mutations were stable and had nearly 100% frequency within 28 days of in utero infection.

Zika virus evolution during transplacental infection is complex, with the evolution in maternal tissues, placental, and fetal tissues. The animal models to capture the whole complexity of this evolution were not reported yet. In the present study, we used direct fetal injection with the virus, which may be considered a limitation. However, we dissected and studied specific in utero Zika evolution in the placenta and fetal brains. Previous studies with clinical and field Zika strains were focused on only virus evolution in maternal tissues [16, 17]. Here, we extended evolutionary studies to the fetal compartment. Also, the fetal pig model provides relevant data because pigs and humans have comparable immune responses and fetal development [25, 26].

There are at least two distinct Zika virus lineages – Asian and African. While the Asian-lineage viruses caused the 2015 epidemic, recent studies identified the African Zika sequences in non-human primates and mosquitoes in Brazil [27, 28]. During the 2015 epidemic, it has been suggested that the causative Asian Zika virus strains acquired mutations distinct from the parental African strains, and this acquisition enhanced its epidemic potential. Afterward, studies showed that viruses from both Asian and African lineages cause disease in animal fetuses [29]. Also, there is evidence that the African-lineage strains may be more virulent [23, 29–32]. In the present study, we also showed that the fetal brain-specific variant with generic African-lineage mutations has the potential for more aggressive infection in vitro – particularly in mosquito cells, and in vivo – in fetal lymph nodes. It has been shown that African Zika viruses caused higher infection in mosquitoes than Asian virus [30]. In the present study, we were limited to C6/36 mosquito cells, and it will be interesting to compare whether the fetal brain-specific variant causes more aggressive infection and transmission in live mosquitoes. We and others showed that African strains cause more aggressive infection, inflammation, pathogenicity, and transmission in chicken, mouse, and porcine fetuses and embryos than Asian virus strains [23, 29, 31–33]. Here, we observed that the variant with acquired generic African-lineage mutations has a trend of higher infection in fetal lymph nodes but not in the placenta. It is possible that the placenta, with the immunity tuned to maintain maternal tolerance to the semi-allogeneic fetus, lacks immune components that are present in fetal lymph nodes. And fetal lymph nodes are more suitable tissues to compare infection phenotypes caused but different Zika virus variants. In support, in our recent study, CpG-enriched Zika virus variants which show attenuated infection phenotypes in different mammalian cell lines, adult mice, and porcine fetal lymph nodes, did not show attenuation in the placenta [15].

In the present study, brains from all fetuses infected with WT and AFM variants were negative for the Zika virus. Here we used the synthetic virus variants with very low genomic heterogeneity (Figure 7B). It is possible that to breach the fetal blood–brain barrier more complex initial genomic diversity in the inoculum virus is needed; more complex in terms of higher SNV frequency and not SNV numbers (Figure 1B). In support, in utero inoculation of 8 pregnant pigs with the wild human or mosquito Zika strains isolated from clinical and field samples resulted in infection in the brains of their fetuses at 28 days after inoculation [18, 23]. In contrast, in utero inoculation of 4 pregnant pigs with synthetic viruses in the present and previous studies [15] did not result in infection in fetal brains at the same sampling time. These data also suggest that in utero emerged fetal brain-specific African mutations are not paramount for breaching the fetal brain barrier. We cannot identify, however, whether African mutations first emerged in the placenta or internal fetal organs and then breached the fetal brain barrier. Or African mutations emerged in the fetal brain after the Asian virus breached the barrier. We cannot identify this because fetal samples were collected only at 28 days after in utero injection, and multiple frequent samplings of many pregnant pigs are needed to answer this question.

Our findings evoke epidemiological concerns. In the study where the neonatal pig model for Zika infection was developed [34], we discussed the potential threat of Zika virus to establish endemic infections in swine herds and a sylvatic cycle involving pigs. We suggested this threat because pigs are native host and reservoir for other flaviviruses like classical swine fever virus, atypical porcine pestivirus, and zoonotic Japanese encephalitis virus. Afterward, studies from other groups showed evidence of Zika virus infection in pigs on farms in Mexico and in mosquitoes collected inside pigpens [35, 36]. Experimental intravenous injection of two pregnant pigs with Zika virus did not result in transplacental and fetal infections. But pregnant pigs developed virus-specific antibodies, and the virus RNA was identified in the placenta [37]. Thus, it will be essential to monitor pig herds in geographical regions endemic to the Zika virus and determine herd potential as an emerging zoonotic reservoir. And whether Zika virus adaptation may lead to transplacental infection, emergence of in utero-specific variants with higher virulence, and spread of these variants to the environment. It is important because here (Figure 6) and in previous Zika studies [20–23], we showed that infected fetuses have increased levels of IFN-α and other cytokines in their blood. Despite vigorous IFN-α and probably other antiviral fetal responses, Zika virus persists in the fetal brain, lymph nodes, and placenta for 28–90 days (Figure 5) [20–23]. Similar findings were reported in our study with another zoonotic flavivirus related to Zika virus – Japanese encephalitis virus, where infected porcine fetuses showed increased IFN-α levels and virus persistence in the fetal blood, brain, and placenta [38]. Thus, flaviviruses persist in fetuses under the selective pressure of vigorous but underdeveloped fetal type I IFN and probably other antiviral responses that may hypothetically result in the emergence of response-resistant and more virulent virus variants.

Altogether, the Asian Zika virus can acquire stable generic African-lineage mutations during in utero infection in pig fetuses. These findings evoke concerns because Zika virus may persist in farm pigs and mosquitoes. Further studies are needed to identify how persistent in utero infection affects virus evolution and whether in utero-emerged Zika variants have the potential for shedding into the environment, more efficient transmission, and more aggressive infection phenotypes.

Author contributions

Conceptualization: UK. Investigation: AJS, PPS, IT, NPKL, UK. Data analysis: AJS, PPS, IT, NPKL, UK. Funding: UK. Writing – original draft preparation: AJS, PPS, UK. Writing – review and editing: AJS, PPS, IT, NPKL, UK.

Acknowledgments

We thank Vaccine and Infectious Disease Organization (VIDO) animal care technicians and veterinarians for their help with animal experiments.

Data availability

An accession number for raw NGS data is PRJNA837638 in NCBI BioProject.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by grants to UK from New Frontier in Research Fund (NFRF) #421586. The funders had no role in study design, data collection and analysis, the decision to publish, or preparation of the manuscript.