Differential gene expression elicited by ZIKV infection in trophoblasts from congenital Zika syndrome discordant twins

Zika virus (ZIKV) causes congenital Zika syndrome (CZS), which is characterized by fetal demise, microcephaly and other abnormalities. ZIKV in the pregnant woman circulation must cross the placental barrier that includes fetal endothelial cells and trophoblasts, in order to reach the fetus. CZS occurs in ~1–40% of cases of pregnant women infected by ZIKV, suggesting that mothers’ infection by ZIKV during pregnancy is not deterministic for CZS phenotype in the fetus. Therefore, other susceptibility factors might be involved, including the host genetic background. We have previously shown that in three pairs of dizygotic twins discordant for CZS, neural progenitor cells (NPCs) from the CZS-affected twins presented differential in vitro ZIKV susceptibility compared with NPCs from the non-affected. Here, we analyzed human-induced-pluripotent-stem-cell-derived (hiPSC-derived) trophoblasts from these twins and compared by RNA-Seq the trophoblasts from CZS-affected and non-affected twins. Following in vitro exposure to a Brazilian ZIKV strain (ZIKVBR), trophoblasts from CZS-affected twins were significantly more susceptible to ZIKVBR infection when compared with trophoblasts from the non-affected. Transcriptome profiling revealed no differences in gene expression levels of ZIKV candidate attachment factors, IFN receptors and IFN in the trophoblasts, either before or after ZIKVBR infection. Most importantly, ZIKVBR infection caused, only in the trophoblasts from CZS-affected twins, the downregulation of genes related to extracellular matrix organization and to leukocyte activation, which are important for trophoblast adhesion and immune response activation. In addition, only trophoblasts from non-affected twins secreted significantly increased amounts of chemokines RANTES/CCL5 and IP10 after infection with ZIKVBR. Overall, our results showed that trophoblasts from non-affected twins have the ability to more efficiently activate genes that are known to play important roles in cell adhesion and in triggering the immune response to ZIKV infection in the placenta, and this may contribute to predict protection from ZIKV dissemination into fetuses’ tissues.

replicates in ex vivo slices from adult human cortical tissues [21]. Also, Zika virus infection 80 reprograms global transcription in the host cells [22]. ZIKV in the maternal circulation needs to cross 81 the placental barrier that includes fetal endothelial cells and trophoblasts in order to reach the fetus 82 [23,24]. Several placenta related cells have been shown to be infected by ZIKV, including placental 83 macrophages and trophoblasts [25][26][27][28][29]. 84 It has been estimated that CZS occurs in ~1-40% of cases of pregnant women infected by 85 ZIKV [3,13,30,31]. This suggests that mothers' infection by ZIKV during pregnancy is not the only 86 factor determining CZS phenotype in the fetus, and other susceptibility factors might be involved. 87 Indeed, neural progenitor cells (NPCs) from different individuals have been shown to respond 88 differently to ZIKV infection [32,33]. In this scenario, discordant twins represent a good case-control 89 sample to test for the genetic contribution determining the fetuses' discordant outcome of gestational 90 infection with ZIKV, as they are supposed to have been exposed to ZIKV under similar conditions in 91 the uterus during gestation. 92 We have previously shown that twins discordant for CZS outcome show differential neural 93 progenitor cells (NPCs) in vitro viral susceptibility to a Brazilian ZIKV strain (ZIKV BR ) [32], but it 94 remains to be determined if discordant CZS twins also show differential placental susceptibility to 95 viral infection. Here, we compared the susceptibility and molecular signatures associated with in 96 vitro ZIKV BR infection of trophoblasts from the same CZS-affected and non-affected twins, using a 97 well-established trophoblasts model that recapitulates the primitive placenta formed during 98 implantation [34,35]. We show here that hiPSC-derived trophoblasts from CZS-affected twins were 99 significantly more susceptible to in vitro ZIKV BR infection when compared with trophoblasts from 100 non-affected twins. In this context, we had previously shown that, before infection, ESC-derived 101 trophoblasts express a wide range of attachment factors for ZIKV entry and lack the components of a 102 robust antiviral response system [35]; in contrast, cells from term placentas, which resist infection, do 103 not express genes encoding attachment factors implicated in ZIKV entry and do express many genes 104 associated with antiviral defense [35]. However, no ZIKV infection assays were performed with 105 these ESC-derived trophoblasts [35]. Here, transcriptome profiling of hiPSC-derived trophoblasts 106 revealed that ZIKV BR infection elicited different responses in hiPSC-derived trophoblasts from CZS-107 affected and non-affected twins, highlighting that genes involved with extracellular matrix 108 organization as well as with immune response activation in the placental tissue may contribute to 109 modulate ZIKV infection outcome. 110

111
Infection with ZIKV BR of hiPSC-derived trophoblasts from discordant twins 112 We obtained blood from three pairs of dizygotic twins discordant for CZS (non-affected: 113 #10608-4, #10763-4, and #10788-4; CZS-affected: #10608-1, #10763-1, and #10788-1) for 114 generation of hiPSC-derived trophoblasts and for phenotypic and gene expression analysis after in 115 vitro infection with ZIKV BR (Fig 1A). Erythroblasts from the three pairs of DZ twins were 116 reprogrammed towards hiPSCs. All hiPSC lines were previously shown by immunofluorescence 117 staining to express markers of pluripotency (TRA-1-60 and OCT4) and by RT-qPCR to express 118 endogenous pluripotent transcription factors including NANOG and OCT4 [32]. 119 Then, the hiPSCs originated from the three pairs of twins were differentiated into primitive 120 trophoblasts using an established protocol [34,35] and further characterized to confirm their 121 differentiation in vitro. Under these conditions, we found that all hiPSC-derived trophoblast lines 122 robustly expressed chorionic gonadotropin subunit beta 3 (β-CG/ hCGB3) and keratin 7 (CK7/KRT7) 123 (Fig 1B), two of the most commonly used trophoblast markers [36]. 124 Subsequently, we infected hiPSC-derived trophoblasts with ZIKV BR using multiplicity of 125 infection (MOI) of 0.3 and 3, and at 96 hpi (hours post-infection) we investigated the viral titer by 126 measuring the number of plaque-forming units (PFU) in cell culture supernatants (Fig 1C). Virus 127 titers were significantly higher (2.8-fold) in the supernatant of CZS-affected twins' trophoblasts 128 infected with MOI 0.3 (Fig 1C and 1D), indicating that trophoblasts from CZS-affected twins were 129 significantly more susceptible to ZIKV infection or at least more virus productive when compared 130 with trophoblasts from non-affected twins. Infection with the higher MOI (MOI = 3) obliterated this 131 difference due to the high susceptibility of trophoblasts in general to the Zika infection. In order to 132 have a better resolution of the differences between cells derived from CZS-affected and non-affected 133 twins, subsequent studies were performed with MOI of 0.3. 134 Potential flavivirus attachment factors and IFN receptor genes were not differentially 135 expressed between trophoblasts from discordant twins 136 To evaluate the possible differences in molecular signatures associated with ZIKV BR infection 137 in trophoblasts from CZS-affected and non-affected twins, we performed RNA-Seq analysis in 138 hiPSC-derived trophoblasts before and after in vitro infection with ZIKV BR for 96 h (MOI = 0.3). 139 With this approach, we could analyze possible differences in gene expression of potential attachment 140 factors for ZIKV and of genes related to antiviral response. 141 We first confirmed the efficiency of hiPSC differentiation into trophoblasts by looking at the 142 expression levels of a set of over100 genes which have been associated with the trophoblast lineage 143 of mammals [35,36]. When compared with hiPSCs, most of these genes showed significant 144 upregulation in trophoblasts after differentiation (S1 Table). We validated by RT-qPCR the 145 differential expression of NANOG, a hiPSC marker, and of HCGA, HCGB and KRT7, three of the 146 genes upregulated in the trophoblasts upon differentiation, as compared with the hiPSCs (S1   Edge R exact test) (Fig 2A). These genes (S2 Table) include interferon-stimulated genes (ISG) and 210 genes related to cytokine secretion. We found a significant (FDR < 0.05, cumulative hypergeometric 211 distribution) enrichment of different Gene Ontology (GO) terms among these 471 upregulated DEGs 212 detected in the RNA-Seq experiment, being the top categories "response to type I interferon", 213 "defense response to virus" and "negative regulation of viral process" (Fig 2B, S3 Table). Consistent 214 with the GO analysis, Ingenuity Pathway Analysis (IPA) also pointed to an enriched network of 215 interferon-stimulated genes upregulated in hiPSC-derived trophoblasts after ZIKV infection 216 (upregulated ISGs are colored in pink in Fig 3A). 217 We also quantified in the hiPSC-derived trophoblasts the secreted levels of type I (IFNA2), 218 type II (IFNG) and type III (IFNL1) IFNs produced by these cells in the absence of virus or after 219 infection with ZIKV BR and culture for 48 h (Fig 3B, upper panel) or 96 h (Fig 3B, lower panel). 220 After 48 h in culture without infection, hiPSC-derived trophoblasts from both CZS-affected and non-221 affected twins were able to secrete IFNA2 and IFNG, whereas IFNL1 was not detectable. There was 222 no statistically significant increase in any IFN secretion at 48 h after ZIKV BR infection of 223 trophoblasts (Fig 3B, upper panel). After 96 h in culture, hiPSC-derived trophoblasts from both 224 CZS-affected and non-affected twins secreted higher levels of IFNA2, IFNG and IFNL1 compared 225 with those at 48 h, although high variability in the secreted IFN levels was observed among the 226 trophoblasts from different individuals (Fig 3B, lower panel). More importantly, a statistically 227 significant increase in the secretion of IFNL1 by trophoblasts from non-affected twins was observed 228 at 96 h after infection with ZIKV BR , but not by trophoblasts from CZS-affected twins (Fig 3B, lower  229

panel). 230
Taken together, these results indicate that hiPSC-derived trophoblasts from both CZS-231 affected and non-affected twins were able to respond to ZIKV BR infection by secreting IFNA2, IFNG 232 and IFNL1, which induced the upregulation of a set of ISG genes potentially involved in the response 233 to ZIKV BR infection. In addition, only trophoblasts from non-affected twins showed a significant 234 increase in secreted IFNL1 at 96 hpi with ZIKV BR (Fig 3B, lower panel). 235

A set of genes involved with extracellular matrix organization and leukocyte activation was 236 downregulated in trophoblasts from CZS-affected twins after ZIKV BR infection 237
We next investigated if there were differences in gene expression between trophoblasts from 238 CZS-affected and non-affected twins after ZIKV BR infection. In total, 44 genes were downregulated 239 after ZIKV BR infection in trophoblasts from CZS-affected when compared with non-affected twins 240 (FDR < 0.05, Edge R exact test) (Fig 4A and S4 Table). Different Gene Ontology (GO) terms were 241 found to be significantly (FDR < 0.05, cumulative hypergeometric distribution) enriched among 242 these 44 downregulated DEGs, including "extracellular matrix" and "regulation of leukocyte 243 activation" (Fig 4B, S5 Table). Significantly enriched Gene Ontology (GO) terms that were found 244 among the upregulated DEGs are shown in S6 Fig and S5 Table, and the most significantly enriched  245 categories are "amino acid biosynthetic process" and "acid secretion". 246 The levels of a panel of cytokines and chemokines secreted by the hiPSC-derived 247 trophoblasts were quantified in the supernatants of 48 h or 96 h cell cultures in the absence of virus 248 or after infection with ZIKV BR . Interestingly, from all the analytes tested, the chemokines 249 RANTES/CCL5 and IP10 showed a consistent (both at 48 h and 96 h post-infection) and significant 250 increase in secretion by trophoblasts from non-affected twins after infection with ZIKV BR , but not by 251 trophoblasts from CZS-affected twins (Fig 4C). RANTES/CCL5 secretion levels by trophoblasts 252 from non-affected twins increased 2.4-and 4.6-fold at 48 h or 96 h after infection with ZIKV BR , 253 respectively, while IP10 secretion levels increased 16-and 96-fold (Fig 4C). The other tested 254 cytokines and chemokines did not show statistically significant differences in the levels produced by 255 infected trophoblasts from CZS-affected or non-affected twins. 256

257
Here, we were able to analyze, for the first time, the in vitro viral susceptibility and the gene 258 expression patterns after in vitro ZIKV BR infection of hiPSC-induced trophoblasts from dizygotic 259 (DZ) twins discordant for the presence of microcephaly. Because these are DZ twins whose mothers 260 were infected with ZIKV BR during pregnancy, the two fetuses in each of the three twin pairs were 261 exposed to the virus at the same time, thus representing a rare and unique cohort to test whether the 262 host genetic background plays any role in determining CZS outcome. Indeed, we have previously 263 shown that hiPSC-induced NPCs from the same subjects exhibit differential in vitro ZIKV BR 264 susceptibility, as NPCs from CZS-affected twins had significantly higher ZIKV BR  Here, we show that hiPSC-induced trophoblasts from DZ twins discordant for CZS were 286 differentially susceptible to infection with ZIKV BR . Interestingly, 96 hpi with ZIKV BR , virus titers in 287 culture supernatants were significantly higher in the CZS-affected twins' trophoblasts when 288 compared with non-affected (Fig 1C), indicating that ZIKV BR can replicate more efficiently in the 289 trophoblasts from CZS-affected twins, potentially facilitating virus dissemination into fetal tissues. 290 Noteworthy, our RNA-seq results indicate that hiPSC-derived trophoblasts express the ZIKV 291 candidate receptor genes (HSPG2 and TAM receptor genes) and the IFN receptor genes, and at lower 292 levels the IFN genes; however none of these genes were differentially expressed between the 293 trophoblasts from CZS-affected and non-affected twins either before or after ZIKV BR infection. 294 Interestingly, ZIKV BR infection caused a significant increase in IFNL1 secretion by 295 trophoblasts from non-affected twins (Fig 3B), whereas in trophoblasts from CZS-affected twins no 296 significant increase was observed (Fig 3B). IFNL1 is a type III IFN constitutively released by 297 primary human trophoblasts from full-term placentas [25], which are known to be refractory to ZIKV with these downregulated differentially expressed genes is "extracellular matrix". It is known that for 317 successful fetus development, one of the critical steps is proper invasion of the maternal decidua by 318 trophoblasts [74] and that many molecules, including galectins, are involved in this process [75,76]. 319 One of the downregulated genes in the trophoblasts from CZS-affected twins after ZIKV BR 320 infection was COL3A1 (Fig 4A).

Measurement of viral burden 419
For ZIKV titration, plaque assay was performed with the supernatants of cell cultures. For 420 plaque assay, an amount of 6 × 10 4 VERO cells/well were seeded in 24-well plates 48 h before the 421 assay. Samples were serially diluted in DMEM culture medium from 10 −1 to 10 −6 , applied in 422 duplicates of 100 µL to each well, and incubated for 30 min at 37 °C. After virus adsorption, wells 423 were overlaid with culture medium containing carboxymethyl cellulose (1 %) and incubated at 37 °C. 424 After 5 days, plates were drained, washed with PBS, and stained with 0.1 % naphthol blue-black, 1.6 425 % sodium acetate in 6 % glacial acetic acid for 30 min. Plaque formation units were visually 426 determined in the most appropriate viral dilution and expressed as PFU/mL. 427

RNA-Seq assay 428
Total RNA from hiPSC-derived trophoblasts was extracted using the RNeasy Micro Kit [93]. To call differentially expressed genes a general linear models (glm) was fitted, and likelihood 446 ratio tests (lrt) were performed using twins' covariates as blocking groups in all cases. P-values were 447 adjusted using FDR, all genes with FDR lower than 0.05 were considered differentially expressed 448 genes. To identify up-regulated or down-regulated genes, the logCPM data from edgeR were used. 449 Heatmaps were plotted using the R package pheatmap. Enriched gene ontology analysis was perform 450 using the bioconductor package clusterProfiler [94], based on the annotation of the bioconductor 451 package org.Hs.eg.db. All plots were generated with the R packages gg plot2, cowplot and the 452 bioconductor package DOSE. 453

Reverse transcription -quantitative PCR (RT-qPCR) 454
Total RNA from hiPSCs and hiPSC-derived trophoblasts was extracted as described above.