Zika virus co-opts miRNA networks to persist in placental niches detected by spatial transcriptomics

Abstract

BACKGROUND:

Zika virus (ZIKV) congenital infection evades double-stranded RNA detection and may persist in the placenta for the duration of pregnancy, without accompanying overt histopathologic inflammation. Understanding how viruses can persist and replicate in the placenta, without causing overt cellular or tissue damage, is fundamental to deciphering mechanisms of maternal-fetal vertical transmission.

OBJECTIVE:

Placenta-specific microRNAs (miRNAs) are believed to be a tenet of viral resistance at the maternal-fetal interface. We aimed to test the hypothesis that ZIKV functionally disrupts placental miRNAs, enabling viral persistence and fetal pathogenesis.

STUDY DESIGN:

To test this hypothesis, we utilized orthogonal approaches in human and murine experimental models. In primary human trophoblast cultures (n=5 donor placentae), we performed high-throughput sequencing crosslinking and immunoprecipitation (AGO-HITS-CLIP) to identify any significant alterations in the functional loading of miRNAs and their targets onto the RNA-induced silencing complex (RISC). Trophoblasts from same-donors were split, and infected with a contemporary first-passage ZIKV strain HN16 (MOI=1 plaque forming unit per cell) or mock infected. To functionally cross-validate miRNA-mRNA interactions, we compared our AGO-HITS-CLIP results with an independent analysis of published (Lum et al., 2019) bulk RNA-seq data from human placental disc specimens (n=3 subjects; ZIKV-positive in 1^st, 2^nd, or 3^rd-trimester, CD45⁻ cells sorted by flow-cytometry) compared to uninfected controls (n=2 subjects). To investigate the importance of these miRNA and RNAi networks in ZIKV pathogenesis, we utilized a gnotobiotic mouse model uniquely susceptible to ZIKV infection. We evaluate if small-molecule enhancement of miRNA and RNAi pathways with enoxacin influenced ZIKV pathogenesis (n=20 dams total yielding 187 fetal specimens). Lastly, placentae (n=14 total) from this mouse model were analyzed with Visium spatial transcriptomics (9,743 spatial transcriptomes) to identify potential ZIKV-associated alterations in immune microenvironments.

RESULTS:

We found that ZIKV infection of primary human trophoblast cells led to an unexpected disruption of placental miRNA regulation networks. Compared to uninfected controls, ZIKV-infected placentae had significantly altered SLC12A8, SDK1, and VLDLR RISC-loading and transcript levels (−2<Log₂(fold-change)>2; adjusted p-value<0.05, Wald-test with FDR correction q<0.05). In silico miRNA target analyses revealed 26 of 119 transcripts (22%) in the TGF-β Signaling Pathway were targeted by miRNAs which were found to be dysregulated following ZIKV infection in trophoblasts. In gnotobiotic mice, relative to mock controls, ZIKV-associated fetal pathogenesis included fetal growth restriction (p=0.036) and viral persistence in placental tissue (p=0.011). Moreover, spatial transcriptomics of murine placentae revealed that ZIKV-specific placental niches were defined by significant upregulation of complement cascade components and coordinated changes in TGF-β gene expression. Finally, treatment of ZIKV-infected mice with enoxacin abolished placental ZIKV persistence, rescued the associated fetal growth restriction, and ZIKV-associated transcriptional changes in placental immune microenvironments were no longer observed.

CONCLUSION:

These results collectively suggest that (i) ZIKV infection and persistence is associated with functionally perturbed miRNA and RNAi pathways specifically related to immune regulation in placental microenvironments, and (ii) enhancement of placental miRNA and RNAi pathways in mice rescued ZIKV-associated pathogenesis, specifically persistence of viral transcripts in placental microenvironments and fetal growth restriction.

Tweetable statement :

Zika virus co-opts miRNA networks to persist in placental niches detected by spatial transcriptomics [hyperlink] @barrozophd @norsketexsci and colleagues

Introduction

Zika virus (ZIKV) is an enveloped, non-segmented, positive-sense single-stranded arbovirus in the Flaviviridae family related to Dengue virus, West Nile virus, and Japanese Encephalitis virus.^{1, 2} Initially isolated in the Zika forest of Uganda in 1947, outbreaks of ZIKV were largely sporadic across Southeast Asia and African equatorial belts, but later spread east to Yap Island in 2007, followed by epidemics in French Polynesia, New Caledonia, the Cook Islands, and Easter Island in 2013 and 2014.^2,3 ZIKV infection leads to relatively mild symptoms in only 20–25% of infected subjects. As local epidemics progressed, rare but concerning cases of (i) Guillain-Barre syndrome were observed in adults (260 per 100,000 cases), and (ii) transplacental vertical transmission was estimated to be as high as 1%.^3–5 However, as ZIKV spread to the Americas (1.3 million cases in 2015), an approximately 20-fold increase in congenital cases of microcephaly with brain and ocular malformations was reported throughout Brazil^{6, 7}, and recapitulated in murine ZIKV pathogenesis models^{8, 9}. Case reports identified ZIKV RNA in amniotic fluid, the placenta, and fetal neural tissues persisting weeks to months after the initial maternal infection.¹⁰ Based on ZIKV studies in human tissue and primate models, we now appreciate the broad cellular tropism and months-long persistence in the placenta, vaginal epithelium, and testes.^{2, 10–17} This broad tropism and persistence with limited inflammation suggest ZIKV evades or co-opts double-stranded RNA (dsRNA)-sensing pathways. Therefore, based on the roles of dsRNA in viral sensing and immune evasion, the RNA-interference (RNAi) and microRNA (miRNA) pathways were of interest.

The miRNA regulatory pathway is estimated to regulate up to 60% of all human mRNAs, and designed to sense dsRNAs and target them for silencing.¹⁸ Endogenous primary miRNAs are processed by DROSHA and diced by DICER to form 21–22 base mature miRNAs that are loaded into the RNA-induced silencing complex (RISC). RISC functions as a post-transcription regulation mechanism that scans transcripts throughout the cell for target sequences complementary to the mature miRNA, leading to translation inhibition and mRNA decay.¹⁸ DICER and RISC are also involved in RNAi, a distinct antiviral pathway that dices exogenous dsRNA into small-interfering RNA (siRNA).¹⁹ Not surprisingly, viruses have evolved mechanisms to evade RNAi and co-opt the miRNA biogenesis pathway.²⁰ ZIKV was found to inhibit DICER by a protein-protein interaction of viral anchoring protein C, influencing both the RNAi and miRNA pathways.²¹ In human neuronal progenitor cells, DICER knockdown with synthetic siRNAs led to increased ZIKV replication, and treatment with DICER agonist enoxacin reduced viral replication.²² Dang and colleagues (2019) assessed miRNA accumulation, RISC-loading, and RNA levels in neurons and observed disrupted miR-124–3p targeting transferrin receptor protein 1 (TFRC).²³

While these studies have identified ZIKV-associated miRNA interactions in neuronal cells, the impact of ZIKV infection on miRNAs has not been systematically assessed in placental cells.^{13, 24} Given the role of the placenta as a reservoir and mediator of fetal infection, understanding how ZIKV evades RNAi and alters miRNA regulatory pathways in trophoblasts potentially represents a key first step in developing targeted therapeutic strategies. To assess the biological significance of miRNAs and RNAi pathways in ZIKV pathogenesis, we considered the impact of perturbations of miRNA biogenesis pathways on ZIKV replication in the placenta. Prior studies had shown that several host-viral miRNA-mRNA interactions found to be significant in in vitro culture or luciferase assays, were not significant in animal viral pathogenesis models^{25, 26} and vice-versa^27–31. Thus, we sought to undertake rigorous and complementary in vitro and in vivo approaches to test the hypothesis that ZIKV functionally disrupts placental miRNAs, enabling viral persistence and fetal pathogenesis.

First, we sought to test our hypothesis in vitro and utilized high-throughput sequencing crosslinking and immunoprecipitation (AGO-HITS-CLIP, a transcriptome-wide approach to identifying functional AGO-bound miRNA and targets) and analyzed RISC-loading in primary human trophoblast cultures infected by ZIKV. Second, to formally determine the biological significance of disrupting miRNA and RNAi functional activity during ZIKV infection in vivo, we leveraged previous studies demonstrating that mice with intact microbiomes (specific-pathogen-free; SPF) are endogenously resistant to ZIKV pathogenesis, and require IFNαR-antibody treatment for ZIKV pathogenesis.^{16, 32, 33} However, we found that gnotobiotic mice are susceptible to ZIKV pathogenesis, resulting in significant fetal growth restriction compared to mock controls.^{34, 35} In mice, conditional deletion of Dicer, an essential component of the miRNA biogenesis and RNAi pathways, resulted in a congenital microcephaly phenotype.³⁶ Since microcephaly was a potential phenotype of ZIKV, and ZIKV already inhibits DICER²¹, we aimed to investigate the impact of miRNA and RNAi enhancement on ZIKV pathogenesis. Enoxacin, a formerly FDA-approved fluoroquinolone, harbors a unique DICER agonist small-molecule activity that results in the enhancement of both miRNA and RNAi pathways³⁷ (reviewed in references^{38, 39}). Subsequent studies have shown enoxacin has broad antiviral activity, including a reduction of ZIKV replication in mosquito cells⁴⁰ and human neuronal progenitor cell cultures²². Importantly, since enoxacin retains both a quinolone-mediated antibiotic function and a unique miRNA enhancement function, the assessment of enoxacin in a gnotobiotic ZIKV mouse model devoid of bacteria enables us to attribute any functional impact to its miRNA enhancement activity and not its inherent quinolone-antibacterial function. Third and lastly, we utilized spatial transcriptomics to understand how ZIKV infection alters murine placental immune microenvironments, and persist in placental niches. We demonstrate through these three orthogonal approaches that ZIKV co-opts miRNA regulatory networks and establishes niche placental reservoirs, which presumptively then serve as permissive conduits for transplacental vertical transmission and fetal pathogenesis.

Materials and Methods

Statement of compliance.

Animal study protocols were approved by the Baylor College of Medicine Institutional Animal Care and Use Committee (AN6697). Protocols utilizing human tissue were approved by the Baylor College of Medicine Institutional Review Board (H26364). Subjects were screened for inclusion and exclusion criteria by chart review, and informed consent was obtained before de-identification and utilization in this study. Inclusion and exclusion criteria were set a priori as follows: For subjects to be eligible for participation, they must have been pregnant, at least 18 years of age, and able to provide informed consent to provide a placental sample at delivery. Parental permission was required for eligible subjects under the age of 18. Subjects would be excluded if they were less than 13 years of age, unable to provide signed and dated informed consent, or unwilling to provide biological samples. In this study, all participants were over 18 years of age and relevant clinical metadata are reported in Supplementary Table 1A.

Primary human trophoblast isolation and ZIKV infection.

Placentas from n=5 healthy term pregnancies (fetal sex: 1 male and 4 female) were collected from recruited participants with informed consent at Harris County Ben Taub Hospital in Houston, Texas. All pregnancies were healthy term pregnancies free of major obstetrical or other health complications and vaginally delivered with spontaneous labor. Placentas were subject to cytotrophoblast purification as described previously.¹³ Cytotrophoblasts were seeded at 105–90 million cells/replicate in 60 cm dishes and maintained with DMEM:F12 media (Gibco, 11320–033) supplemented with + 10% heat-inactivated fetal bovine serum (Gibco, 16140–071) + 1% penicillin-streptomycin (Gibco, 15140–122). Media was changed daily for 5 days until cells differentiated into syncytiotrophoblasts, confirmed by β-human chorionic gonadotropin expression¹³. At 5 days post-isolation, differentiated primary human trophoblasts cell plates from each patient were either mock-infected with PBS or infected with a contemporary first-passage ZIKV strain HN16 (GenBank accession KY328289.1)^{12, 16} at an MOI of 1 plaque-forming unit (PFU) and harvested 24 hours post-infection.

Argonaut High-throughput Sequencing UV-crosslinking and Immunoprecipitation (AGO-HITS-CLIP).

AGO-HITS-CLIP was done as described previously^41–43, excluding photoactivatable ribonucleoside analog thiouridine. Briefly, RNA were crosslinked to the RNA-binding proteins by two rounds of UV-crosslinking on ice (1000 J/cm² at 265 nm). Cells were then lysed, and RNAs not protected by protein were digested with RNase T1. Argonaut 2, the main component of RISC was then immunoprecipitated along with the crosslinked RNAs with Protein G Dynabeads (Invitrogen, 10003D) coupled to 0.125 μg/mL anti-Argonaut 2 antibody clone 11A9 (Millipore, MABE253). After immunoprecipitation, the bead complexes were washed and subjected to a second RNase T1 treatment, followed by several stringent high-salt washes. To purify the RISC-loaded RNAs, we first radiolabeled the RNAs and then performed gel electrophoresis. The excised band was proteinase K (Roche, EO0491) treated and RNA was precipitated, followed by acidic phenol-chloroform (with isoamyl alcohol, 125:24:1, Invitrogen, AM9720) RNA extraction. RNAs were then eluted and libraries were prepared for sequencing using the TruSeq small RNA library Set A (Illumina RS-200–0012), and cDNA was single-end sequenced using the Illumina HiSeq 2500 platform (50–51 bp reads) at the Baylor College of Medicine Human Genome Sequencing Center.

Ribonomics analyses.

Reads were pre-processed and quality-filtered using FastqQC (v0.11.9). Barcodes and adapters were trimmed using Trimmomatic⁴⁴ (v0.33). Reads were aligned to custom human+ZIKV+hsa-miRbase transcriptome using STAR⁴⁵ (v2.7.8). PCR duplicate reads were removed using Picard⁴⁶ (v2.24.0). Unique reads were counted using HTseq⁴⁷ (v0.11.1). Counts from the 5 mock and 4 infected biological replicates were analyzed by differential expression (DE) in R (v4.0.2) using DEseq2⁴⁸ (v3.12). AGO-HITS-CLIP peaks were called with Piranha⁴⁹ (v1.2.1) or PureCLIP⁵⁰ (v1.3.1). Positive or negative-sense reads-per-million-scaled (RPM) reads, and CLIP peaks were converted to bigwig files for visualization on the UCSC genome browser⁵¹ (hg38). Host miRNAs with different accumulation levels were analyzed by mirPath⁵² (v3, DIANA Tools) using the MicroT-CDS database⁵³ (v5) and visualized using the Kanehisa Laboratories Pathway Viewer. Full differential RISC-loading results are available in Supplementary Table 1. The unique reads and accompanying AGO-HITS-CLIP peaks can be visualized with over 50 fully interactive UCSC strand-specific tracks (see https://genome.ucsc.edu/s/ebarrozo/barrozo%2Dzikv). We acknowledge experimental differences between the datasets and each ZIKV dataset was analyzed relative to its uninfected controls to account for these differences.

Gnotobiotic ZIKV pregnancy mouse model.

Following timed mattings of Swiss/webster mice (>8 weeks old) drawn from Baylor College of Medicine Gnotobiotics Core colonies, plugged dams were weighed and transferred to separate sterile insulators. Intraperitoneal injections of enoxacin (Sigma-Aldrich, 557304) at 10 mg/kg were previously shown to be safe and effective in a model of obesity⁵⁴ and were done daily from embryonic days E1.5-E18.5. Dams either received mock United States Pharmacopeia-grade PBS (VWR, 76081–514) subcutaneous injections or ZIKV infection with HN16 strain (first passage in Vero-E6 cells) at a dose of 1×10⁵ PFU on E7.5, E8.5, and E9.5. On day E18.5, dams were humanely sacrificed, and uterine horns were dissected. Images of the uterine horns were taken to record resorption events. Individual fetuses were weighed. Placental tissues were stored at −80°C in Trizol. All mouse experiments were conducted under the supervision of Baylor College of Medicine Institutional Animal Care and Use Committee approved protocols.

RNA extractions and reverse transcription quantitative PCR (RT-qPCR).

Murine placental tissues were dissected and finely minced for lysis in 1 milliliter of Trizol per 50–100 milligrams of tissue (Invitrogen, 15596–026). RNA was extracted using a Direct-zol RNA Miniprep Kit (Zymo, R20520) following the manufacturer’s instructions including in-column DNase treatment and further DNase treated using TURBO DNase (Invitrogen, AM2239) in the presence of RNAse-inhibitor (Promega, N2611). RNA concentrations were measured, and purity was confirmed by spectrophotometer. RNA input was standardized to 1 microgram, denatured at 65°C for 5 minutes before TaqMan Fast Virus 1-step Master Mix (ThermoFisher/Applied Biosystems, 4444432) following the manufacturer’s instructions. TaqMan qPCR was done using the Applied Biosystems StepOnePlus probing for a custom primers/probe set for the ZIKV envelope gene (Forward Primer 5p-TGCCCAACACAAGGTGAA, Reverse Primer 5p-CCTCTGTCCACTAACGTTCTTT, Probe 5p-AGTGTCTGATTGCTTGTCAAGGTAGGC). As a positive control for ZIKV, viral RNA was extracted from purified virus aliquots and serially diluted to generate a standard curve.

Spatial transcriptomics in murine placentae.

At E18.5, placentae were dissected and preserved in RNA-later at −25°C for 18 hours. Tissues were dehydrated by incubation in 30% sucrose at 4°C for 6 hours. At the Baylor College of Medicine Pathology and Histology Core placentas were fresh-frozen in optimal cutting temperature compound (OCT) with a liquid nitrogen and isopentane bath. Tissue blocks were cryosectioned at 10 μm for quality control H&E staining and 10x Genomics Visium Tissue Optimization. The Tissue Optimization slide revealed optimal permeabilization was observed after 12 minutes. Finally, the tissue blocks were cryosectioned and placed on Visium Gene Expression slides (v2) for H&E staining and Visium spatial transcriptomics library preparation following the manufacturer’s instructions. Visium libraries were sequenced (Illumina, NovaSeq 6000) at the Baylor College of Medicine Genomic and RNA Profiling Core.

Spatial transcriptomics analysis.

Spatial transcriptomes were aligned to a custom human and ZIKV transcriptome and transcripts counts were quantified using SpaceRanger (v1.3.0) and bash scripts. Downstream analyses were done in R (v4.0.5) using the Seurat (v4.0.6) scRNA-seq analysis package.⁵⁵ H&E slides were annotated by a trained pathologist and transferred to spatial barcodes using the Loupe Browser (v5), followed by annotation based on PCA dimension reduction and clustering analysis to identify broad transcriptome niches. Spatial transcriptomes were then integrated with a published murine single-nuclei RNA-seq (snRNA-seq) dataset⁵⁶ to assess transcriptome annotations (GEO accession GSE152248). The Marsh & Blelloch, eLife. (2020) snRNA-seq dataset included murine placentas sampled at E9.5, E10.5, E12.5, and E14.5 across 2,933 independently filtered (nFeature_RNA >500 & nFeature_RNA <4000 & nCount_RNA >500 & nCount_RNA <10000 & percent.mt <10 & percent.ribo <20) high-quality transcriptomes. Full differential expression results are available in Supplementary Table 2, along with fetal sex as determined by Y chromosome gene expression. The niche annotations of the spatial transcriptomes were predicted based on integration with a published murine placenta single-nuclei RNA-seq (snRNA-seq) dataset⁵⁶, dissected at embryonic days E9.5, E10.5, E12.5, and E14.5 (Supplementary Fig. 3; Supplementary Table 2B). Using the snRNA-seq dataset as an anchor, the snRNA-seq cell-type annotations were transferred (anchor-based transfer of annotations⁵⁵) to the spatial transcriptomes niches. Broad niche profiles (e.g. trophoblast, endothelial; Fig. 4C) were denoted as “Type.Profiles” and subtypes (e.g. trophoblasts subdivided into several classes; Fig. 4D) were denoted as “Subtype.Profiles”. The 9,743 spatial transcriptomes were subject to PCA dimension reduction and visualization in Unique Manifold Projection and Approximation (UMAP) space labeled by treatment, cluster, Type.Profiles, and Subtype.Profiles (Fig. 4E; n=7 male and n=7 female placentae as determined in Supplementary Fig. 1D; Supplementary Tables 2C–F). Notably, there were no significant differences in Type.Profiles or Subtype.Profiles proportions according to treatment group (Fig. 4F, p>0.999) or among individual samples (Supplementary Fig. 3B, p>0.999). While the integration and overlap between these two technologies (Supplementary Fig. 4) reflected the current technical differences between 10x Genomics Visium (v2) and single-nuclei resolution, the sensitivity and specificity of transcript marker overlap between spatial transcriptome and snRNA-seq reference datasets were promising.

Fig. 4. Benchmarking spatial transcriptomics in murine placentae reveals distinct microenvironments.

A, Schematic of distinct zones in murine placentae. Image created with BioRender. B, Placentae were fresh-frozen in OCT, cryosectioned, and placed on Visium Gene Expression Slides for hematoxylin and eosin (H&E) staining and spatial transcriptomics. Using the 6mm-by-6mm fiducial frame, the 55-micron spatial transcriptome spots are overlayed on the H&E images and annotated by a placental pathologist (Sample Mock-Enoxacin-2). C-D, Spatial transcriptomes were annotated utilizing a mouse placenta snRNA-seq dataset (Marsh & Blelloch. (2020))⁵⁶ as a reference to identify cellular type and subtype profiles (denoted as Type.Profile and Subtype.Profile). E, Dimension reduction and visualization of 9,743 spatial transcriptomes in Unique Manifold Projection and Approximation (UMAP) space labeled by treatment, cluster, and type or subtype profiles (n=8 dams, 1–2 placentae each, 7 male and 7 female placentae). F, Proportions of type and subtype profiles were not statistically different. The color annotations for Type.Profile and Subtype.Profiles align for panels E and F.

Data & code availability.

The AGO-HITS-CLIP and spatial transcriptomics data have been deposited to the Gene Expression Omnibus (accessions GSE205609 and GSE205631). The custom human+ZIKV+hsa-miRbase reference (GRCh38.p13_and_NC_012532.1_and_hsamiRbase22.1.gtf) and scripts used for AGO-HITS-CLIP and spatial transcriptomics analyses are available at https://github.com/Aagaardlab/ZIKV_AGOHITSCLIP and https://github.com/Aagaardlab/murineplacenta_spatialtranscriptomics respectively. We independently analyzed published datasets including bulk RNA-seq data from term placentae of subjects that tested positive for ZIKV during pregnancy (PRJNA578707)⁵⁷, and a murine single-nuclei RNA-seq data which included murine placentae sampled at embryonic days E9.5, E10.5, E12.5, and E14.5 (GSE152248)⁵⁶. The scripts used for independent analyses of these published datasets are included in the respective GitHub repositories above.

Statistical analyses.

In differential expression analyses, significance was determined with Wald tests with False Discovery Rate (FDR) multiple test correction where significance was defined as a q<0.05 in R with DEseq2. Transcripts with differential expression q<0.05 and −2<Log₂(fold-change)>2 were subject to Reactome 2022⁵⁸ pathway analysis using EnrichR⁵⁹. Additional statistical tests were done in Prism (v8.3.0). Significant differences from animal experiments were determined by two-tailed Mann-Whitney tests with significance defined as p<0.05.

Results

ZIKV disrupts miRNA regulatory pathways in human placental trophoblast cultures.

To determine if ZIKV infection altered the composition of transcripts functionally loaded into the RISC (Fig. 1), we performed AGO-HITS-CLIP in primary human placental trophoblasts cultures derived from term placentae. Clinical characteristics of study participants are reported in Supplementary Table 1A; n=5 donor placentae; 4 female and 1 male fetal sex. In a paired fashion to control for donor variability, primary trophoblast cultures were either infected with a contemporary first-passage ZIKV strain HN16 (GenBank accession KY328289.1)^{12, 16} (n=4, MOI=1 PFU/cell) or mock infected with PBS for negative controls (n=5; an additional unpaired pilot mock sample was included) over 24 hours. Based on Log₂ fold-change, the top 95^th percentile of miRNAs most perturbed by ZIKV infection in placental cells were identified (Fig. 1A; Supplementary Table 1B) and included miRNAs known to interact with other viruses⁶⁰ (miR-132 and miR-34B) and regulate tumor progression⁶¹ (miR-32, miR653, miR-200b, miR-16–1, and miR-489). Analyses of RISC-loaded mRNAs identified CYB561D2 and ARMCX5-GPRASP2 as being significantly upregulated following ZIKV infection (Fig. 1B).

Fig. 1. ZIKV disrupts miRNA regulatory pathways in human placentae.

A-B, Paired primary human trophoblast cultures were infected with ZIKV strain HN16 at an MOI of 1 (n=4 placentae, replicates paired with mock samples) or mock (n=5) for 24 hours and subjected to Argonaut High-throughput Sequencing UV-crosslinking and Immunoprecipitation (AGO-HITS-CLIP) analysis. A, The top 95th percentile of miRNAs with changes in RNA-induced silencing complex (RISC)-loading as determined by Log₂(fold-change) of ZIKV infected samples compared to mock. B, Volcano plot with cutoffs at −2<Log₂(fold-change)>2 on the x-axis and p<0.00001 significance on the y-axis for transcripts loaded into RISC comparing ZIKV infected samples compared to uninfected controls. C, Independent analysis of Lum et al., (2019)⁵⁷ bulk RNA-seq data from the placental disc of human placentae infected by ZIKV (n=3 subjects; infected in either the 1^st, 2^nd, or 3^rd trimester, enriched for CD45⁻ cells by flow-cytometry, 1–3 replicates each) and compared to uninfected controls (n=2 subjects, 1–3 replicates each). D, Functional cross-validation of AGO-HITS-CLIP and bulk RNA-seq transcripts significantly changed following ZIKV infection. Transcripts with −2<Log₂(fold-change)>2 and an adjusted p-value<0.05 (Wald-test, with False Discovery Rate (FDR) multiple test correction) were filtered as significant.

To our surprise, even with trophoblasts in culture with experimental evidence of ZIKV replication, we did not observe ZIKV reads in the AGO-HITS-CLIP datasets, nor did we detect virus-derived siRNAs loaded into RISC or host miRNAs loaded into RISC with ZIKV RNA. We next compared our trophoblast AGO-HITS-CLIP results to an independent analysis of Lum et al., (2019) bulk RNA-seq data⁵⁷ from human placental disc samples naturally infected by ZIKV (n=3 subjects; maternal ZIKV-positive in either the 1^st, 2^nd, or 3^rd-trimester, enriched for CD45⁻ cells by flow-cytometry, 1–3 technical replicates per subject) and compared to uninfected controls (n=2 subjects, 1–3 replicates each). Among the top 90^th percentile of differentially expressed transcripts associated with ZIKV infection, the 20-fold upregulation of TMSB15B stood out (Fig. 1C), along with altered immune response transcripts found in CD45⁻ cells, including macrophage markers CD14 and HLA-DRB5, proinflammatory cytokine CSF3, and leukocyte marker VCAM1. We assessed the functional cross-validation (Fig. 1D) of the primary AGO-HITS-CLIP culture and bulk RNA-seq data. Based on fold-change and significance (−2<Log₂(fold-change)>2 and an adjusted p-value<0.05, Wald-test, with FDR correction), we identified three high-confidence transcripts, including SLC12A8, SDK1, and VLDLR that had altered RISC-loading and transcript levels that surpassed these stringent filters. Given these two lines of evidence demonstrating ZIKV disrupted miRNA-mRNA regulation, next, we assessed the impact on biological pathways in placental cells.

Disruption of TGF-β signaling networks in placentae infected by ZIKV.

The top miRNAs differentially loaded into RISC (AGO-HITS-CLIP analysis passing the thresholds of −2<Log₂(fold-change)>2 and significance q<0.05, listed in Supplementary Table 1E) were analyzed with mirPath (v3, DIANA Tools) utilizing the MicroT-CDS database (v5)⁵³ of known miRNA-mRNA interactions.⁵² Of the 119 transcripts in the TGF-β Signaling Pathway, 26 were targeted by miRNAs significantly perturbed by ZIKV in trophoblasts (Fig. 2; Supplementary Table 1E). Since the TGF-β signaling pathway has numerous immunoregulatory functions^62–65, especially in the placenta (see Comment section), we hypothesized the impact of these miRNA-regulatory networks on placental pathophysiology and ZIKV disease would be most apparent in vivo.

Fig. 2. Disruption of TGF-β signaling networks in placentae infected by ZIKV.

The impact of ZIKV-associated miRNA changes on placenta miRNA-regulatory pathways were analyzed by uploading the top 98^th percentile of miRNAs (−2<Log₂(fold-change)>2; q<0.05) from the AGO-HITS-CLIP analysis into mirPath (v3, DIANA Tools) using the MicroT-CDS database of verified miRNA-target interactions. In the TGF-β Signaling Pathway (hsa04350), 26 of 119 transcripts were targeted by these host miRNAs significantly perturbed by ZIKV infection.

Small-molecule enhancement of miRNA and RNAi pathways by enoxacin rescues ZIKV pathogenesis phenotypes in a gnotobiotic mouse model.

In our gnotobiotic mouse model (Fig. 3A) we injected a non-toxic dose⁵⁴ of enoxacin (10 mg/kg) intraperitoneal injections daily from E1.5-E18.5 or a phosphate-buffered saline (PBS) vehicle control. Dams were infected with ZIKV strain HN16 (first passage) at an MOI of 1×10⁵ PFU (or mock-PBS) via subcutaneous injections mid-gestation on embryonic days 7.5 (E7.5), E8.5, and E9.5. Fetal ZIKV pathogenesis was assessed at E18.5 from a significant number of dams (n=4 gnotobiotic dams in the mock+PBS vehicle control group, 5 mock+enoxacin, n=5 ZIKV+vehicle, and n=6 ZIKV+enoxacin). Fetuses were dissected and weighed (n=56 fetuses mock+vehicle, n=30 mock+enoxacin, n=58 ZIKV+vehicle, n=43 ZIKV+enoxacin) to determine the occurrence of fetal growth restriction. There were no significant effects of daily enoxacin treatment on fetal growth nor survival in the absence of ZIKV infection when comparing mock+ vehicle control dams and mock+enoxacin treatments (Fig. 3B, p=0.82; two-tailed Mann-Whitney tests). However, we did observe significant fetal growth restriction in the ZIKV+PBS vehicle gnotobiotic group compared to uninfected controls (Fig. 3B;*p=0.0363). In the daily enoxacin-treated dams, ZIKV-associated fetal growth restriction was not seen (Brown-Forsythe and Welch ANOVA with Dunnett’s multiple comparisons mock vs. ZIKV+enoxacin adj. p=0.968; ZIKV+vehicle vs. ZIKV+enoxacin adj.*p=0.0371).

Fig. 3. Small-molecule enhancement of miRNA and RNAi pathways by enoxacin inhibits the establishment of the ZIKV placental niche and rescues ZIKV-associated fetal growth restriction in a gnotobiotic mouse model of perinatal infection.

A, Experimental workflow: Timed mated gnotobiotic Swiss/Webster mice (n=4 dams mock+PBS vehicle control, n=5 mock+enoxacin, n=5 ZIKV+vehicle, n=6 ZIKV+enoxacin) received daily intraperitoneal injections with enoxacin (10 mg/kg) or a PBS vehicle control. Consistent with our previous gestational ZIKV exposure model, mice were inoculated with subcutaneous injections of ZIKV (strain HN16, 1×10⁵ PFU) or mock-infected with PBS at E7.5, E8.5, and E9.5. Experiments ended on E18.5 when Cesareans were performed. B, Fetuses were immediately weighed to determine fetal growth restriction (n=56 fetuses mock+vehicle, n=30 mock+enoxacin, n=58 ZIKV+vehicle, n=43 ZIKV+enoxacin). C, RNA was extracted from placental tissue and viral loads were determined by 1-step RT-qPCR probing for the ZIKV envelope gene (ZIKV-ENV; n=6 placentae from 3 dams mock, n=6 placentae from 3 dams ZIKV+vehicle, n=12 placentae from 6 dams ZIKV+enoxacin). ZIKV was not detected in mock-infected controls or ZIKV+enoxacin placentae, but positive in 4 of 6 ZIKV+vehicle placentae (Ct-threshold <35). Significant differences were determined by two-tailed Mann-Whitney tests with significance defined as p<0.05. Error bars represent the standard error of the mean.

To assess differences in ZIKV persistence in these murine placental tissues, RNA was extracted and viral loads were determined by 1-step RT-qPCR probing for the ZIKV envelope gene (Fig. 3C; n=6 mock placentae from 3 dams, 6 ZIKV+vehicle placentae from 3 dams, 12 ZIKV+enoxacin placentae from 6 dams). Viral RNA was detected in 4 of 6 ZIKV+vehicle placentae (66.6%), and no viral RNA was detected after enoxacin treatment or in the uninfected controls (Ct-threshold<35; Kruskal-Wallis test with Dunn’s multiple comparisons test mock+vehicle vs. ZIKV+vehicle *p=0.011; mock+vehicle vs ZIKV+enoxacin p>0.99; ZIKV+vehicle vs. ZIKV+enoxacin *p=0.0035). Since both enoxacin and ZIKV were administered systemically, we cannot say if the antiviral effects were in the maternal milieu or at the maternal-fetal interface. We did not find significant differences in (i) fetal demise/resorption events (p=0.77), (ii) viral loads by RT-qPCR in the maternal brain (p>0.99), maternal spleen (p=0.46), and fetal brain tissues (p=0.22), or (iii) fetal histopathology after whole fetuses were bisected, formalin-fixed, paraffin-embed, and serial sections were stained for H&E, Ki-67, and cleaved Parp-1 (data not shown). In turn, we turned to high-resolution spatial transcriptomics in the murine placenta to investigate how ZIKV altered placental immune microenvironments with or without enoxacin.

Benchmarking spatial transcriptomics in murine placentae reveals distinct microenvironments permissive to ZIKV replication.

Spatial transcriptomics enables measurement of in situ gene expression, while retaining the structural microarchitecture essential to the heterogeneous maternal and fetal cells in the placenta. A subset of the E18.5, placentae from the murine experiments above were subject to H&E staining directly on Visium Gene Expression slides (v2), followed by permeabilization, and spatial transcriptomics library preparation (n=2 mock+vehicle, n=4 mock+enoxacin, n=4 ZIKV+vehicle, and n=4 ZIKV+enoxacin). Spatial transcriptomes from each spot (55 μm in diameter, representing 1–5 cells cut 10 μm in thickness) were sequenced and aligned to a custom human and ZIKV transcriptome. Conservative quality-control filtering (Supplementary Fig. 1) yielded 9,743 high-quality spatial transcriptome spots. We undertook rigorous benchmarking, and assessed placental spatial marker transcripts in uninfected controls (Fig. 4, Supplementary Figs. 1–4, Supplementary Table 2A–F). Since the junctional zone separates the fetal labyrinth zone and the maternal decidua (Fig. 4A; Sample Mock-Enoxacin-2 shown), spatial transcriptomes were aligned to the spots on the H&E images using unique spatial barcodes matching the slide serial number and the fiducial frame (Fig. 4B). Unbiased differential expression of the spatial transcriptomes by cluster (Supplementary Fig. 2; Supplementary Table 2A) correctly identified the labyrinth zone (significant increases in carbonic anhydrase 4 (Car4) expression (Log₂(fold-change)>2, q<0.05 Wilcoxon Rank Sum test), decidua (marked by carboxypeptidase E (Cpe) expression), and the junctional zone (marked by prolactin family 3a1 (Prl3a1)). Additional unbiased spatial marker analysis identified 107 significant spatial markers (Supplementary Fig. 2; Supplementary Table 2A).⁶⁶ We considered the multinucleated nature of syncytiotrophoblasts as demarcations of the maternal-fetal interface and compared them to single-nuclei profiles. Utilizing a published mouse placenta single-nuclei RNA-seq dataset as a reference⁵⁶, we then verified the identities of these niches with probabilistic transfer of annotations based on anchoring features. The spatial annotations of Type.Profiles, and Subtype. Profiles, were aligned to the H&E stain images and visualized in Supplementary Figure 3. Next, we focused in on spatial niches with immune cell gene expression.

Interrogation of immune cell markers identifies a subset of macrophage microenvironments of informed interest.

Based on our findings from our AGO-HITS-CLIP analysis, we were particularly interested in the disruption of TGF-β and related immune pathways. Thus, in an approach akin to flow cytometry in silico, we cast a refined net and utilized density plots to identify clusters where canonical myeloid and lymphocyte markers^{67, 68} were expressed individually, or co-expressed (Fig. 5A–C). Myeloid transcriptomes were subclassified into monocytes, placenta-associated maternal macrophages (PAMMs), and fetal Hofbauer cell niches (Fig. 5B). Lymphoid transcriptomes were divided into natural killer T (NKT), CD8⁺ cytotoxic T lymphocytes (CTL), and regulatory T (Treg) cell niches (Fig. 5C). We then annotated the immune niches most representative of each cluster (Fig. 5D). Cluster 7 was of interest (Fig. 5E), because it resembled PAMMs and fetal Hofbauer macrophages, aligning with top markers from the human ZIKV placenta transcriptomics analysis (Fig. 1). These 667 macrophage spatial transcriptome profiles were mapped to the H&E images in Fig. 5E, to visualize potential ZIKV-associated immune microenvironments.

Fig. 5. Interrogation of immune cell markers identifies a subset of macrophage microenvironments in murine placentae.

A, UMAP of Visium spatial transcriptomes split by cluster. B-C, Expression density plots of (B) myeloid or (C) lymphoid markers expressed individually, or co-expressed. D, Annotation of clusters based on canonical immunophenotype markers. E, Spots within cluster 7 for each sample were selected for further analysis. The spatial locations of each spot were overlayed on top of the H&E images and colored differently for each sample.

Analysis of macrophage immune microenvironments determines spatial markers associated with ZIKV infection.

The 667 spatial transcriptome profiles were parsed by treatment group for further analysis (Figure 6). Differential expression based on treatment (Supplementary Table 2G) identified 33 transcripts significantly upregulated in the ZIKV+vehicle group, 1 in the Mock+vehicle group (Hba-a1 and Gm42418), 1 in the ZIKV+enoxacin (Lor), and 1 in the Mock+enoxacin group (Eln). Visualization of the ZIKV+vehicle gene signature (Fig. 6A), in each cluster revealed that these markers were not differentially expressed in trophoblast-like clusters, but were highly expressed in a subset of spatial transcriptomes found in clusters 5 and 7. With enoxacin treatment, the ZIKV-associated gene signature returned to normal (Fig. 6B). The ZIKV+vehicle gene signature was associated with significant upregulation of key immunomodulators, including complement cascade components C3, C1qa, Cfh, and Serping1, CTL marker Ctla2a, cytokines Ccl8 and Ccl21a, and MAVS antiviral innate immune signaling molecules⁶⁹ Lyz2 and Pla1a. Pla1a aligned with the top bulk RNA-seq markers noted in Figure 1, however, the remaining 33 markers were only revealed with spatial transcriptomics technology (Fig. 6C). The spatial gene expression of these markers was visualized in their accompanying placental niches (Fig. 6D–E, Supplementary Fig. 5). EnrichR⁵⁹ pathway analysis of the 65 transcripts significantly upregulated in ZIKV+vehicle samples (Supplementary Table 2H–I) identified 6 significant Reactome pathways⁵⁸, including the Innate Immune System (R-HSA-168249), and 3 pathways of Complement Cascade Regulation (R-HSA-977606, R-HSA-166658, R-HSA-166663). Next, we aimed to determine if TGF-β signaling, the top ZIKV-associated pathway from the AGO-HITS-CLIP analysis in human cells (Fig. 2), was also altered in these ZIKV placental niches.

Fig. 6. Analysis of macrophage immune microenvironments identifies spatial markers associated with ZIKV infection.

The 667 FetalMesenchyme/Immune/Endothelial spatial transcriptome profiles were subset and reanalyzed. A, Differential expression based on treatment identified 33 transcripts significantly upregulated in ZIKV-Vehicle samples. The spatial expression of these genes relative to each Subtype.Profile. B, Expression of top markers for the immune profile subset relative to treatment groups. C, Cross-validation of ZIKV-associated transcription changes (Log₂(fold-change)>2, q<0.05 Wilcoxon Rank Sum test) between the Visium immune subset, AGO-HITS-CLIP, and bulk RNA-seq datasets. D-E, Spatial gene expression of PLA1A or C3 zoomed in on the 667 immune microenvironments.

Significant coordinated changes in TGF-β pathway gene expression observed in ZIKV microenvironments.

Pearson’s correlation analysis on the co-expression of genes from the TGF-β Molecular Signature Database (GSEA MM3865) in all 9,743 spatial transcriptomes or in the subset of 667 ZIKV immune microenvironments (Figures 5 and 6) demonstrated limited coordinated expression of TGF-β genes (Fig. 7A). However, when the 667 ZIKV-associated immune microenvironments were parsed, we identified two significantly correlated TGF-β gene expression programs (Fig. 7B, blue blocks indicating a significant positive correlation between 16 TGF-β genes, “ZIKV TGF-β microenvironment expression program 1”). In addition, 5 genes were significantly positively correlated with each other, but anti-correlated with another set of 16 genes (Fig. 7B, red blocks indicating “ZIKV TGF-β microenvironment expression program 2”). In other words, Fkbp1a, Junb, Id1, Id2, and Id3 in program 2 may be negative regulators of program 1, or vice-versa. In these two gene expression programs of TGF-β genes perturbed in ZIKV immune microenvironments, the transcripts Junb, Ltbp2, Bmpr2, Wwtr1, Smad7, Serpine1, Sptbn1, Cdh1, and Bcar3 were significantly differentially expressed in the bulk RNA-seq data from human placentae infected with ZIKV during pregnancy (Fig. 1; Supplementary Table 2J), suggesting these distinct transcription programs identified in ZIKV susceptible mouse placentae may be translatable to human placental infection.

Fig. 7. ZIKV-specific placental niches are defined by coordinated TGF-β gene expression.

A-B, Pearson’s correlation analysis of co-expression of TGF-β genes from the HALLMARK_TGF_BETA_SIGNALING Molecular Signature Database (GSEA MM3865) for (A) all 9,743 spatial transcriptomes and (B) the 667 ZIKV-Vehicle transcriptomes from the macrophage immune subset. Insignificant correlations (p>0.05) were colored white.

Comment

Principal Findings.

We and others have previously demonstrated that the placenta serves as a reservoir for ZIKV replication and persistence throughout pregnancy in clinical studies, experimentally engineered susceptible murine models, and non-human primate models which faithfully recapitulate human disease.^{13, 70–73} However, prior to the current study, how ZIKV develops a permissive replication niche in the placenta devoid of evident histopathology had been poorly understood.

Based on established evidence that placenta-specific miRNAs (such as C19MC) serve as potent placental post-transcription regulators^{24, 74–77} and maintain trophoblast cell-type characteristics⁷⁸, we tested the hypothesis that ZIKV functionally disrupts placental miRNAs, enabling viral persistence and fetal pathogenesis. Employing orthogonal and unbiased multi-omics approaches, we identified differences in RISC-loading and mRNA levels between ZIKV-infected and uninfected placental cells, and pathway analysis revealed that ZIKV disrupts placental miRNA regulation networks known to target TGF-β signaling. We then sought to understand the importance of miRNA and RNAi pathways in ZIKV pathogenesis with enoxacin. In a gnotobiotic ZIKV pathogenesis model, we found significant fetal growth restriction and ZIKV vRNA persistence in placentae were rescued with enoxacin treatment. We then applied spatial transcriptomics to murine placentae and found a distinct ZIKV-associated gene signature in placental immune microenvironments marked by increased complement cascade activation and coordinated changes in TGF-β signaling. Finally, we demonstrated that treatment of ZIKV-infected mice with enoxacin, the ZIKV-associated gene signature and placental immune niches returned to normal. Collectively, these results highlight the importance of miRNA regulation in the placenta, and reveal how ZIKV generates a niche permissive to viral persistence, vertical transmission, and resultant fetal pathogenesis.

Results in the context of what is known.

We and others have demonstrated rather conclusively that ZIKV infection is not associated with generalized placental dysfunction nor ubiquitous placental inflammation. A notable review published in the NEJM in 2019 warned that recurrent ZIKV outbreaks are inevitable, given that a large proportion of the world’s population lives in arboviral endemic regions, and most are naïve to ZIKV.⁷⁹ Thus, the applicability and significance of understanding how congenital arboviruses, such as ZIKV, can persist and replicate in the placenta without causing overt cellular or tissue damage, is fundamental to elucidating mechanisms of vertical transmission and developing efficacious interventions. Here, we implicate the disruption of host factors, specifically miRNAs which normally modulate immune processes, in ZIKV fetal pathogenesis. Moreover, in a series of in vivo mouse experiments, we faithfully recapitulate these findings and demonstrate that small-molecule enhancement of miRNA and RNAi pathways with enoxacin abolishes placental ZIKV persistence and rescues the associated fetal growth restriction. Moreover, ZIKV-associated transcriptional changes in placental immune microenvironments were no longer observed.

Our findings meaningfully expand those of prior studies by providing mechanistic insights previously not observed. A recent study utilizing immortalized human trophoblasts found TGF-β treatment significantly increased ZIKV receptor expression, virus uptake, and budding.⁶⁵ During early pregnancy, serum and placental levels of TGF-β increase and are thought to promote placental invasion of the maternal decidua, fetal immune tolerance, angiogenesis, and vascularization.⁶² TGF-β can act as a suppressive cytokine by recruiting Tregs to the placenta.⁸⁰ Co-opting these anti-inflammatory functions of TGF-β would be ideal for immune evasion. Here, ZIKV AGO-HITS-CLIP highlighted perturbed RISC-loading of several host miRNAs that target numerous transcripts within the TGF-β signaling pathway. RNA-seq highlighted a 20-fold increase in Thymosin-β isoform TMSB15B, which can be directly regulated by TGF-β.⁸¹ In our spatial transcriptomics analysis, ZIKV-associated immune microenvironments exhibited two distinct TGF-β coordinated programs. These were not observed in the uninfected controls nor ZIKV-infected placentae treated with enoxacin. While other studies have detailed cross-talk between TGF-β and the complement cascade in epithelial-mesenchymal transition, lung fibrosis, and retina pigmentosa^82–84, the complement cascade is primarily associated with responses to bacteria and presumably not a well-known effector of antiviral responses. In our study, spatial transcriptomics resolution of immune microenvironments within the placenta was critical to detail the localized TGF-β and complement alterations in the microarchitecture of placentae of ZIKV-infected dams (Fig. 7) and reveal that ZIKV-specific placental niches were defined by significant and specific upregulation of complement cascade components and coordinated changes in TGF-β expression. These results collectively suggest that (i) ZIKV infection functionally perturbs specific miRNA and RNAi pathways in placental microenvironments, and (ii) enhancement of placental miRNA and RNAi pathways in mice rescued ZIKV-associated pathogenesis, specifically persistence of viral transcripts in placental microenvironments and fetal growth restriction.

Research implications.

An inherent challenge of miRNA-targetomics is the varying degree in determining the biological significance of miRNA interactions. Individual miRNA-mRNA interactions with subtle phenotypes in cell culture systems may not be biologically significant for pathogenesis, and vice-versa. In turn, animal pathogenesis models reveal the sum effects of compounding webs of miRNA regulatory networks. Here, we did not focus on specific miRNA-mRNA interactions, since we have previously identified ZIKV-associated miRNA perturbations¹³, and the identification of individual miRNA-mRNA interactions is better suited for higher-resolution AGO-HITS-CLIP variants. Instead, we focused on determining the sum results of functional RISC-loading of miRNAs and their targets during ZIKV infection in placental cells with high-resolution transcriptomics and a murine model of fetal viral pathogenesis. Recently, spatial transcriptomics has been applied to human first trimester^{85, 86} and term⁸⁷ placentae. Here, the benchmarking and application of this technology in murine placentae aided understanding the link between miRNA pathways, ZIKV fetal pathogenesis phenotypes, and viral persistence in the placenta. This is an exciting and translationally significant application of the AGO-HITS-CLIP technology, with implications for use in other studies which aim to identify the role of individual miRNA-mRNA interactions in perinatal biology.

Clinical implications.

While ZIKV-related microcephaly has not been reported as extensively in recent years, it is erroneous to assume that endemic ZIKV infections are not clinically problematic, nor that the pathophysiology of perinatal infections has changed. Rather, human behavior, naturally acquired immunity, and implementation of public health prevention and screening protocols are likely contributors to the noted reduced prevalence of severe neonatal and infant congenital Zika syndrome disease. We and others have published on increasingly more common adverse outcomes of congenital Zika syndrome independent of microcephaly, all of which are suffered most frequently in vulnerable populations with social disparities.^88–92 The persistence of ZIKV and the disproportionate burden born by vulnerable populations has led to multiple vaccine and therapeutic trials in the last 2 years,^{89, 90, 93–95} the net findings of which support ours and others earlier work showing that the most common sequelae of ZIKV infection in pregnancy are fetal loss or growth restriction, with and without developmental sequelae and microcephaly.^{10, 14–16, 96, 97}

Strengths and limitations.

There are notable limitations to our study. We did not examine ZIKV strain differences, dose dependence, or longer-term outcomes of our in vivo modeling. The impact of enoxacin on ZIKV pathogenesis may not have been limited to the maternal-fetal interface because enoxacin treatment potentially influenced ZIKV replication in the dam. Nonetheless, the use of the gnotobiotic (bacteria-free) ZIKV mouse model in this study allowed us to separate the antibiotic function of enoxacin from its enhancement of miRNA and RNAi pathways and enabled us to identify fetal ZIKV pathogenesis phenotypes and demonstrated their rescue by enoxacin treatment at the phenotypic and molecular levels. Previously, we rigorously evaluated sexually dimorphic responses to ZIKV in mice at term and over 6 weeks in the offspring following early (E4.5), mid-gestation (E8.5), and late (E12.5) ZIKV infections.¹⁶ The only sexually dimorphic response we found was a consistent reduction in biparietal diameters in male offspring over the first 5 weeks of age, but no significant differences were observed in brain weights after dissection at 6 weeks. Therefore, this study focused on the fetal specimen and was not powered to contrast differences in fetal sex, exposure across gestation, or in offspring in response to ZIKV.

There are several additional strengths to our study, including our use of multiple orthogonal approaches with in vitro and in vivo faithful recapitulation of findings. Additionally, we benefit from the use of highly novel, state-of-the-science technologies. There are limited publications on the use of spatial transcriptomics in placenta tissues.⁹⁸ In the current study, we utilized the 10x Genomics Visium (version 2) spatial transcriptomics platform for the first time in murine placentae. The current version uses 55-micron spots that contain 1 to 5 cells, and does not yield single-cell resolution. We predicted the cellular profiles based on cluster label transfers using top markers from murine placenta snRNA-seq data to address this. While Visium was able to separate different types and subtypes of trophoblasts (Fig. 6A), we could not differentiate between fetal Hofbauer macrophages and PAMMs. Fetal Hofbauer macrophages, the site of ZIKV persistence in human placentae^{13, 32, 70}, reside within blood vessels in murine placentae⁹⁹, which have structural differences compared to the human placenta. As anticipated, we did not observe ZIKV transcripts in the spatial transcriptomes of the ZIKV-infected placentae because the fresh-frozen solution of Visium relies on poly(d)T reverse transcription, and flaviviruses lack poly(A) tails.^{100, 101} Yet, we did observe transcription and spatial niche alterations in the placenta associated with maternal ZIKV infection, which may have an impact on fetal immune development, even without vertical transmission in utero.

Conclusions.

The mechanisms governing the sparsity of microbes (including viruses) in the placenta, and how microbes may establish a placental niche, remain poorly understood. In placental tissue, low-biomass microbes including bacterial 16S ribosomal RNA, and viruses such as ZIKV and SARS-CoV-2, have been identified and visualized in the absence of inflammation.^{10, 14, 16, 35, 57, 102–113} Collectively, our current study results in complementary human trophoblast and murine pathogenesis models demonstrate that placental miRNA-mediated antiviral responses are fundamental and rapid molecular means of sequestering inflammatory responses to placental microenvironments, and avoiding widespread and potentially catastrophic placental inflammation. We speculate that similar mechanisms likely play a key role in enabling sparse commensal microbial communities to exist without evidence of inflammation, potentiating a change in fetal immune development, even in the absence of vertical transmission.^{10, 14, 16, 35, 57, 98, 102–118} Enhancing miRNA and RNAi pathways warrants further investigation as promising therapeutic targets for combating congenital viruses.

AJOG at a Glance:

Why was this study conducted?

Molecular mechanisms underlying congenital Zika virus (ZIKV) pathogenesis and viral persistence in the placenta remain poorly understood.

Key findings.

Placental ZIKV infection disrupts miRNA regulatory networks, leading to altered gene expression of key immunoregulatory pathways in distinct placental niches. ZIKV pathogenesis in susceptible gnotobiotic mice manifests as fetal growth restriction and viral persistence in placental tissue, both of which were rescued by treatment with the miRNA agonist enoxacin.

What does this add to what is known?

Together, these studies suggest that ZIKV evades normal RNA interference (RNAi) and miRNA-regulation mechanisms to persist in placental niches and render fetal pathogenesis.