Gas6 drives Zika virus-induced neurological complications in humans and congenital syndrome in immunocompetent mice

Zika virus (ZIKV) has the ability to cross placental and brain barriers, causing congenital malformations in neonates and neurological disorders in adults. However, the pathogenic mechanisms of ZIKV-induced neurological complications in adults and congenital malformations remain unknown. Gas6 is a soluble TAM receptor ligand able to promote flavivirus internalization and downregulation of immune responses. Here we demonstrate high Gas6 levels in the serum of patients with neurological complications which correlated with downregulation of genes associated with the type I IFN responses as consequence of Socs1 upregulation. Gas6 gamma-carboxylation is essential for ZIKV replication in monocytes, the main source of this protein. Gas6 also facilitates ZIKV replication in adult immunocompetent mice enabled susceptibility to transplacental infection and congenital malformations. Our data thus indicate that ZIKV promotes the upregulation of its ligand Gas6, which contributes to viral infectivity and drives the development of severe adverse outcomes during ZIKV infection.


INTRODUCTION
, negatively regulating type I interferon receptor (IFNAR) signalling pathway [12, 13-123 18]. However, the overall interplay of viral and host factors, such as GAS6, to orchestrate the 124 clinical course of ZIKV infection is not fully understood. 125 Here we investigated the role of the PtdSer ligand Gas6 in the pathogenesis of ZIKV 126 infection. We observed that circulating levels of Gas6 are upregulated in the serum of ZIKV-127 infected patients, including pregnant women, and a further increase occurs in patients with 128 neurological complications. Concurrently, there is a reduced transcriptional expression of 129 genes associated with type I IFN responses and other immune signatures, probably as 130 consequence of Socs1 upregulation in peripheral blood cells. ZIKV infected monocytes were 131 an important source of Gas6 production and Gas6 gamma-carboxylation was essential for 132 ZIKV replication. Conversely, we also demonstrate that Gas6 is upregulated in ZIKV-infected

164
Samples of 8 ZIKV-infected pregnant women and 6 infants born of these women were also analysed in 165 this study. Two of these infants had growth restriction and CNS alterations compatible with CZS. Due the small 166 number of patients, no clinical comparison was made between these two groups.

Increased Gas6 expression in ZIKV adult patients with neurological complications 169
To correlate Gas6 levels to the pathogenesis of ZIKV-associated neurological 170 complications, we first determined Gas6 levels in patients' serum by ELISA. Figure 1A shows 171 that circulating levels of Gas6 are significantly increased in ZIKV-infected patients compared 172 to healthy donors (Non-Neuro ZIKV : 22.56 ng/mL [25-75 interquatile range (IQ) 14.6-29.9] 173 versus 13.36 ng/mL  in HDs, p = 0.0062), whose circulating Gas6 levels 174 were comparable to previous studies [19][20][21]. Importantly, Neuro ZIKV patients showed a further 175 increase of serum Gas6 in comparison to  40.2], p = 0.0289) ( Figure 1A). To rule out the possibility of co-infections influencing Gas6 177 levels observed in these patients, we used a high-throughput screening (HTS) virus 178 metagenomic approach to identify viral co-infections that were not detected by RT-qPCR or 179 diagnosed by laboratory tests. ZIKV mono-infections in the Neuro ZIKV patients were confirmed 180 in 9 out of 10 patients tested. In one patient, the complete genome of a strain of Pegivirus C 181 (Pegivirus genus, Flaviviridae family) was sequenced (Supp Fig 1). Noteworthy, this virus has 182 not been associated with neurological complications in humans. Finally, unchanged Gas6 183 levels in the serum of 14 patients with neurological diseases non-related to acute ZIKV 184 infection (Neuro NON-ZIKV ) revealed that Gas6 upregulation is a ZIKV-specific response ( Figure  185 1A). In addition, genotyping analysis revealed no difference in GAS6 haplotypes between the 186 ZIKV-infected patient groups, both showing a predominant frequency of the c.834 + 7G>A 187 AA genotype ( Figure 1B). 188

Gas6 upregulation suppresses antiviral IFN response in ZIKV adult patients with 190 neurological complications 191
To determine the mechanisms by which ZIKV-induced Gas6 upregulation contributes 192 to the pathogenesis of neurological complications in adult patients, we used network analysis 193 of biomarkers and transcriptional profiling to determine how Gas6 orchestrates with a specific 194 signature of immune mediators associated with ZIKV infection, previously determined in our 195 cohort [9]. In accordance with its role as pleiotropic inhibitor of innate immune responses [12, 196 13, 16], Gas6 negatively correlates with several pro-inflammatory cytokines/chemokines, such 197 as IL-2, IL-8, IL-27, RANTES, IP-10 and TNF-α (r > 0.7; p < 0.05) in healthy donors ( Figure  198 1C, D). Interestingly, a striking change in the pattern of interactions between all biomarkers 199 appeared in Non-Neuro ZIKV patients. In these patients, Gas6 significantly and positively 200 correlates only with IFN-α (r = 0.60; p < 0.05), apart from other 2 functional clusters of 201 significant interactions between the measured biomarkers ( Figures 1E, F). In addition, the 202 network graph of Non-Neuro ZIKV patients is more heterogeneous and shows a decentralized 203 topology with lower complexity and connectivity between the immune mediators when 204 compared to the highly dense, homogenous and centralized graph of HDs (Non-Neuro ZIKV vs 205 HDs network density: 0.120 vs 0.224; network centralization: 0.107 vs 0.335; network 206 diameter: 10 vs 5) ( Figure 1D, F). Interestingly, a further increase on Gas6 levels above certain 207 threshold (estimated to be above 30ng/mL), as observed in Neuro ZIKV patients (Supp Fig 2), 208 also changes the patterns of interactions. In these patients, Gas6 displays positive correlation 209 with IL-1RA, IL-6, MCP-1 and IP-10 (r = 0.8; 0.5, 0.4 and 0.3, respectively; p<0.05) and 210 negative correlations with IFN-α, respectively;211 p<0.05) ( Figures 1G, H). In addition, as shown in the network graph ( Figure 1H), because IFN-212 γ forms a functional cluster of interactions with a variety of cytokines with key roles in the 213 antiviral response, Gas6 also displays indirect negative correlations with TNF-α, IL-1β, IL-12, 214 IL-22 and IL-27 (r between -0.3 and -0.5; p < 0.05), as shown in Figure 1G. 215     To determine whether increased Gas6 production correlates with transcriptional 274 changes induced by infection, we first verified that Gas6 mRNA is not significantly increased 275 at 24h and 48h p.i., as represented in the heatmap of log2 Fold change in Figure 3D. This is 276 consistent with the analysis ex vivo, as shown in Figure  Gas6 production by THP-1 cells is increased in a time-dependent and viral load-dependent 295 manner, with a significant increase observed after 24hp.i. (Figure 3E-G). Interestingly, as 296 shown in the heatmap of log2 Fold change in comparison to mock cells (uninfected) in the 297 Figure 3H, Gas6 mRNA is upregulated in THP-1 cells as soon as (6h p.i.) and remains 298 increased throughout the experiment (Supp Fig 4B). GGCX and VKORC1 mRNA expression 299 were not significantly altered, although it tends to decrease at higher MOI 10 ( Figure 3H, Supp 300 Fig 4B). Importantly, Axl, Mer, Tyro3 and SOCS-1 mRNA expression was increased early 301 after infection (6-12h p.i.), in particular at lower MOI 0.1 ( Figure

Inhibition of Gas6 γ-glutamic acid carboxylation (Gla domain) restores antiviral response 344
To further determine the mechanistic link between Gas6 production, ZIKV replication and 345 suppression of antiviral response, we tested whether inhibiting Vitamin-K-dependent γ-346 carboxylation of Gas6 glutamic acid residues (Gla domain) restores antiviral immune response 347 and controls viral replication. We used low-dose warfarin, which specifically block γ-348 carboxylation of Gla domain in Gas6, resulting in decreased TAM receptor activation [23, 27, 349 31]. In parallel, we used R428, a well-known inhibitor of Axl tyrosine kinase activity tested in 350 level did not change ( Figure 4B). Warfarin also decreased GGCX and VKORC1 mRNA 356 expression ( Figure 4F). Corroborating this, R428 treatment potently reduced ZIKV-induced 357 Gas6 production (Supp Fig 5A). 358 Interestingly, incubation of ZIKV -permissive Vero cell line with the supernatant from the 359 ZIKV-infected monocytes revealed that warfarin treatment, at the moment of infection or 2h 360 p.i., significantly decreases the production of viable viral particles at 48h p.i. (Figures 4C, D). 361 At 72h p.i., the drug completely blocked viral production ( Figures 4C, D). R428 did not change 362 ZIKV RNA levels but showed a modest but significant effect on decreasing viral production 363 at 72h p.i. (Supp Fig 5B, C). Accordingly, flow cytometry analysis of ZIKV NS1 protein in 364 Vero cells showed that infection was not detected when cells were exposed to supernatants 365 from monocytes treated with warfarin, regardless whether the initial virus inoculum was 366 washed-out 2h p.i. or not or whether warfarin treatment started 2h p.i. (pos-treatment approach) 367 or at the moment of infection (pre-treatment approach) ( Figure 4E). Interestingly, warfarin 368 induced a potent upregulation of IFNβ and IFIT-1 expression ( Figure 4F). Meanwhile, R428 369 increased the expression of Axl and Mer (Supp Fig 5D). Consistent with the blockade of type 370 I IFN by Axl signalling pathway, R428 allowed the increase of IFN-β and IFIT-1 (Supp Fig  371   5D), while it did not significantly affect GGCX or VKORC1 mRNA expression (Sup Fig 5D).  women revealed a significant increase during acute phase of infection (18.05 ± 1.97 ng/mL 433 versus 13.36 ± 1.23 ng/mL) (Supp Figure 6A). On the other hand, we did not detect a difference 434 in Gas6 levels in the serum of babies with (n = 2) or without (n = 4) CZS (Supp Figure 6B). 435 As the number of ZIKV-infected pregnant women with fetal growth-associated 436 malformations included in this study was limited to two patients, we decided to evaluate the 437 effect of Gas6 using a mouse model of CZS. For this, we subcutaneously injected pregnant 438 C57BL/6 mice with Gas6-coated ZIKV (ZIKV rmGas6 ) at embryonic day (E) 16 and performed 439 analysis at E18 or postnatal day (P) 4 ( Figure 6A and Supp Figure 6). Importantly, infection of 440 pregnant C57BL/6 mice with ZIKV rmGas6 rendered their offspring susceptible to fetal 441 malformations ( Figure 6B). We evidenced macroscopic changes at E18 and growth delay at 442 P4 ( Figure 6B, C). Significant differences in biparietal distance, skull length and weight of the 443 ZIKV rmGas6 newborns were observed ( Figure 6C). No difference in viral load between ZIKV 444 and ZIKV rmGas6 groups was found in the placenta and fetal spleen at E18 (Supp Figure 7A) as 445 well as in the spleen of P4 newborns (Supp Figure 7B). However, a higher viral load was 446 observed in the spleen of ZIKV rmGas6 -infected pregnant mice as long as 8 days post-infection 447 (day 0 at E16, and day 8 at P4) ( Figure 6D). 448 Since Axl is expressed in decidua, Hoffbauer, trophoblast and fibroblasts cells of 449 human placenta 2 and ZIKV has been considered a sexually transmitted disease (STD), we 450 sought to verify whether Gas6 facilitates intra-uterine ZIKV infection. For that, C57BL/6 451 pregnant mice were intravaginally infected at E10 with ZIKV or ZIKV rmGas6 and viral load was 452 determined at E18 ( Figure 6E). Contrasting with the subcutaneous infection, no difference in 453 fetal size was observed (Supp Figure 7C, D), but there is a significant increase in ZIKV load 454 in the fetal spleen and in the placenta of ZIKV rmGas6 -infected pregnant mice ( Figure 6F). 455 Although we see no change in Stat1 and Socs1 mRNA expression, Ifnα4 and Ifnβ1 expression 456 were unexpectedly increased in the placenta of ZIKV rmGas6 -infected pregnant mice ( Figure 6G). reveals that elevation of circulating Gas6 levels above a certain threshold, which we estimate 521 to be higher than 30ng/mL, dampens the protective immune response that provide a strong 522 antiviral environment and a milder clinical outcome. In support, transcriptional analysis in 523 peripheral blood cells from ZIKV patients with neurological complications indicate that this 524 response is achieved by a ZIKV-Gas6-TAM receptor interaction that ultimately induces 525 downregulation of type I IFN genes modulated by SOCS1. This mechanism was corroborated 526 in cultured monocytes, where ZIKV-induced Gas6 expression also correlated with suppression 527 of type I IFN response. 528 The antiviral response also involves recruitment and coordination of specific subsets of 529 immune cells orchestrated primarily by chemokines. Gas6 can also function as an 530 inflammatory molecule by inducing leukocyte adhesion on endothelial cells surface and 531 extravasation through a P-selectin-dependent mechanism [49]. Moreover, Axl and Mer, in 532 cooperation with IFNAR signalling, have been described as key molecules for maintenance of 533 the blood-brain barrier (BBB) and protection to WNV and La Crosse virus infection [50]. 534

Accordingly, in vivo studies of ZIKV infection in immunocompromised IFNAR-KO mice 535
lacking Axl have shown protection from ZIKV neuropathogenesis and severe infection [18]. 536 In this context, suppression of IFNAR signalling due to increased amounts of circulating Gas6 537 in Neuro ZIKV patients potentially impairs BBB homeostasis and integrity. This may allow viral 538 and cellular extravasation to CNS though loose endothelial cells junctions. Noteworthy, in Neuro ZIKV patients, Gas6 positively correlates with IL-6, IP-10 and MCP-1, suggesting that 540 there might be a parallel between ZIKV and cytokine release syndrome (CRS). In addition, our 541 data reveal a unique and inappropriate inflammatory response in Neuro ZIKV patients. This 542 response is defined by a failed type IFN-I response in the periphery, juxtaposed to elevated 543 chemokines levels and high expression of IL-6, leading to recruitment and infiltration of 544 effector immune cells in deep tissues. Thus, infiltration of cells, such as CD4 and CD8 T cells, 545 along with infection of mature neurons could induce a local inflammatory response in the CNS, 546 potentially resulting in neurological complications [40]. We propose that Gas6 mediated 547 reduced innate antiviral defences coupled with exuberant inflammatory cytokine production 548 are driving features of severe clinical outcome in ZIKV adult patients. 549 It has been shown that inhibition of Axl signalling by different pharmacological or 550 genetic approaches decrease ZIKV replication in vitro in CNS cells, endothelial cells and 551 dendritic cells as well as reduce brain pathology in experimental models [16,18,[51][52][53]. TAM 552 receptors activation by Gas6 is highly dependent on its γ-carboxyglutamic acid-rich (Gla) 553 domain, required for its biological activity and to bridge enveloped viruses to bind and activate 554 TAM receptors [23][24][25][26][27]. After translation, Gas6 is activated by γ -glutamyl carboxylation via 555 the Vitamin-K cycle and cycle [23]. Vitamin-K epoxide reductase enzyme complex 1 556 (VKORC1) recycles Vitamin-K Epoxide back to Vitamin-K Hydroquinone, which in turn 557 serves as a co-factor in the γ-carboxylation of Gas6 induced by Vitamin-K-dependent enzyme 558 γ-glutamyl carboxylase (GGCX). Low-dose of warfarin functions as a direct VKORC1 559 inhibitor, preventing γ-carboxylation of Gas6 and TAM receptor activation [23,31]. Thus, we 560 used low-dose warfarin to further determine the mechanistic link between Gas6 production, 561 ZIKV replication and suppression of antiviral response. It is important to highlight that one 562 could argue that decreased Gas6 production after warfarin treatment could be a result of 563 decreased binding of anti-Gas6 capture antibody in the ELISA due to restricted recognition of 564 γ-carboxylated residues (amino acids 53-92) in Gas6 protein. However, the capture antibody 565 recognizes the residues 118-678, which are not in the Gla-domain. In addition, Gas6 can be 566 transcriptionally upregulated by Axl-mediated autocrine mechanisms [23], which could 567 explain warfarin-induced downregulation of Gas6 expression. In our experiments, this 568 mechanism was confirmed by restoration of antiviral response and complete blockage of 569 production of infective viral particles when monocytes were treated with warfarin. These 570 findings implicate that, by tethering ZIKV to TAM receptors, Gas6 mediates the pivotal 571 suppression of type I interferon receptor (IFNAR) signalling, thereby favouring ZIKV evasion from antiviral immunity and sustained replication, pointing out how interactions with 573 membrane receptors go beyond attachment and internalization of viral particles [54]. 574 We demonstrate a correlation between Gas6 protein expression, ZIKV RNA load and 575 production of infecting particles associated with suppression of type I IFN response in human 576 cells infected in vitro. Although we did not detect difference in ZIKV RNA load in the 577 peripheral blood specimens in our cross-sectional cohort, this could probably be due to the 578 moment blood sampling was performed. When immunocompetent adult C57BL/6 and SJL 579 mice were infected with Gas6-coated ZIKV (ZIKV rmGas6 ), we observed that upregulation of 580 Gas6 increases viral load very early after infection, at 1 d.p.i. After 3 and 5 days of infection, 581 differences were no longer detected. This might explain our observation in patients, as the 582 median interval between illness onset and sampling was 4 days, ranging from 1 to 6 days. arteries apoptosis, foetal hypoxia and prominent growth restriction. In the same study, 610 intraperitoneal injection of poly: IC, a major TLR3 agonist which is also activated by ZIKV 611 [61], led to resorption of all foetuses in an IFNAR-dependent manner [58]. Accordingly, the 612 hypothesis for our findings would be that the immunocompetence described in C57BL/6 613 animals may be beneficial while low amounts of pathogen are present. However, if viral load 614 is increased, as it happens due to infection with ZIKV rmGas6 , the exacerbated immune response 615 in the pregnant mother could result in detrimental changes to foetal development probably 616 resulting from increased placental damage.   symptoms. ZIKV infection was confirmed by qRT-PCR and/or specific IgM detection, neurologic complications as described above but were negative for ZIKV by qRT-PCR and/or 677 specific IgM detection. Although three of these patients were positive for DENV by NS1/IgM 678 Rapid immunochromatographic tests, the pathological origin of these neurological symptoms 679 was undetermined at the moment of sample collection. All these patients were negative for 680 DENV, CHIKV and OROV by qRT-PCR. 681

MATERIALS AND METHODS
Healthy donors (HDs): 13 age-matched individuals without signs of infection within 682 30 days prior to sample collection. They were included and pre-screened for presence of ZIKV 683 RNA and ZIKV-specific antibodies. 684 Additionally, not included in the 90 described patients, we analysed acute-phase serum 685 samples of 8 ZIKV-infected pregnant women and 6 infants born of these women. Of these 686 infants, two had CNS abnormalities associated to Congenital Zika Syndrome (CZS), while 4 687 infants were healthy, without congenital zika syndrome (Non CZS). All participants were 688 tested for a series of arboviruses by qRT-PCR. All patients and healthy donors were negative 689 to Dengue (DENV), Chikungunya viruses (CHKV) and Oropouche were negative as 690 determined by RT-qPCR. Samples were collected after consent of the patients. 691 692 Serum Collection and Processing: Peripheral blood and/or urine specimens were collected at a 693 median of 3 days post-illness onset. Serum was obtained from 10 mL of peripheral blood 694 collected in a dry heparinized tube after peripheral venepuncture. All samples were transported 695 and processed as previously described [9]. Positivity ZIKV RNA in the blood, serum and urine 696 samples was verified by real-time quantitative RT-PCR (qRT-PCR) [9,67]. In addition, the 697 presence of ZIKV-specific IgM and IgG antibodies in the serum was determined by enzyme-698 linked immunosorbent assay (ELISA), as previously described [9,67]. penicillin/streptomycin. Human brain microvascular endothelial cells (hBMEC) were cultured 707 in DMEM high glucose, supplemented with 1% L-glutamine, 1% non-essential aminoacids, 708 and 10% FBS [19]. Peripheral blood mononuclear cells (PBMC) from healthy donors were obtained after centrifugation of buff y coat samples over ficoll-hypaque gradient. Subsequently, 710 the mononuclear cell ring was collected and washed with 1x PBS, followed by centrifugation. 711 Then, the pellet was resuspended in complete RPMI-1640 (supplemented with 10% FBS and 712 1% antibiotics). Cells were then counted for determination of cell viability by Trypan Blue. 713 PBMCs and THP-1 human monocytic cell line were maintained in a complete RPMI-1640 714 medium supplemented with 10% FBS, 2mM L-Glutamine (Corning), and 1x Penicillin-715 Streptomycin Solution at 37°C in a fully humidified atmosphere containing 5% CO 2 . 716 717 Virus strains: For in vitro studies in cultured human cells, Brazilian ZIKV strain (BeH823339, 718 GenBank KU729217), originally isolated from a patient in Ceará, Brazil in 2015, was provided 719 by Professor Edison Durigon (Biomedical Sciences Institute, University of São Paulo, Brazil). 720 Virus stocks were produced by inoculating Vero CCL81 cells (ATCC) with ZIKV in minimum 721 essential medium (MEM) for 2h at 37°C and 5% CO 2 . Further, the supernatant was removed, 722 and MEM supplemented with 2% FBS, 1% penicillin and streptomycin was added. The cells 723 were incubated for 4 days until 70% of cytopathic effect. Supernatant was then collected, 724 centrifuged for 5 min at 10,000g, 4°C and snap-frozen at -80°C until use. were infected for 1h at 27ᵒC in the absence of CO 2 . Further, 45mL of complete medium was 729 added (2% of FBS + 1% of Pen/Strep) and cultures were followed until reaching cytopathic 730 effect. At this time, supernatants were harvested and centrifuged at 3200rpm for 10min at 4°C 731 to remove any detached cell. ZIKV culture supernatants were further precipitated with 50% of 732 PEG (polyethylene glycol) for 18h at 4°C. Precipitated virus supernatants were centrifuged 733 (30', 3200 g, 4°C), and the pellet was diluted in DMEM with 25Mm HEPES quantified by 734 PFU assay in VERO cells and used as necessary. 735 736 Mice: SJL and C57BL/6 mice were bred under specific-pathogen-free conditions at University 737 of São Paulo animal facility of the Department of Immunology -ICB. 8-week-old non-pregnant 738 or 11-week-old pregnant female mice were used. All animals were maintained in accordance 739 with institutional guidelines for animal welfare after approval by the Institutional Animal Care 740 and Use Committee at University of São Paulo, as described above. 741