Histopathological lesions of congenital Zika syndrome in newborn squirrel monkeys

The absence of an adequate animal model for studies has limited the understanding of congenital Zika syndrome (CZS) in humans during the outbreak in America. In this study, we used squirrel monkeys (Saimiri collinsi), a neotropical primate (which mimics the stages of human pregnancy), as a model of Zika virus (ZIKV) infection. Seven pregnant female squirrel monkeys were experimentally infected at three different gestational stages, and we were able reproduce a broad range of clinical manifestations of ZIKV lesions observed in newborn humans. Histopathological and immunohistochemical analyses of early-infected newborns (2/4) revealed damage to various areas of the brain and ZIKV antigens in the cytoplasm of neurons and glial cells, indicative of CZS. The changes caused by ZIKV infection were intrauterine developmental delay, ventriculomegaly, simplified brain gyri, vascular impairment and neuroprogenitor cell dysfunction. Our data show that the ZIKV infection outcome in squirrel monkeys is similar to that in humans, indicating that this model can be used to help answer questions about the effect of ZIKV infection on neuroembryonic development and the morphological changes induced by CZS.

www.nature.com/scientificreports/ pregnancy. In this work, we compare the neuroembryonic development of S. collinsi with that of human babies, showing that both go through critical stages of neural development in early pregnancy. Additionally, in both S. collinsi and human newborns with CZS, histopathological findings revealed altered migration of neuroprogenitor cells, neuronal apoptosis and necrosis of the periventricular grey matter [23][24][25] . Taken together, these data suggest that this species of neotropical primate is an excellent model for further study of CZS.

Results
Female squirrel monkeys infected with ZIKV during pregnancy. For this study, the four ZIKVinfected groups (G1, G2, G3, and G5; Fig. 1A) were inoculated via the intradermal (i.d.) route with a dose of 1.0 × 10 5 plaque-forming units (PFUs). Based on ultrasound (US) evaluation, embryo development and previous information about neotropical NHPs, the animals from G1 were further divided into two subgroups using a limited time point approach for chorionic vessel detection. Therefore, two animals (1A and 1B) were infected before 37 gestational days (Gd), and the other two animals (1C and 1D) were infected after 37 Gd. Of the ten infected pregnant females, seven completed gestation and gave birth to their newborns naturally. One animal (1A) had a miscarriage at 7 days post-infection (dpi), at approximately Gd 41 (Fig. 1A). Confirmation of ZIKV infection in animal 1A was shown in our previous paper 26 . Macroscopic (Fig. 2) and microscopic (Fig. 3) changes in the central nervous system (CNS) were observed in the offspring of females infected at different gestational periods and compared with those in the CNS of the newborns in the negative control group (G4) ( Fig. 2A), as well as with the information available on the lesions described in humans newborns. Our experimental model mimicked the variability of lesions caused by ZIKV during embryogenesis as well as the association between viremia and the maternal-fetal immune response 26 . Figure 1B summarizes these previous findings. Viral infection in the offspring after peripheral transmission was also a testament to the ability of ZIKV to break the physical barrier of the placenta and the blood-brain barrier during the different gestational periods of squirrel monkeys.
Macroscopic changes in the CNS. Before necropsy, the animals were weighed and measured; after necropsy, a macroscopic examination in search of morphological changes was performed, and the observations were recorded. The newborns in G1 were compared with those in G4 (uninfected/ Fig. 2A-C); G1 neonates showed several changes in brain structures, including lissencephaly, a simplified gyral pattern, blood-brain barrier damage (Fig. 2D,G), edema and vasocongestion (1C, Fig. 2H), with multifocal areas of subarachnoid haemorrhage (1D, Fig. 2F). Newborn 1D had right lateral ventriculomegaly (Fig. 2E) in the region of the third ventricle, and the occipital fontanelle was open and not fully formed (Fig. 2I), indicating a delay in development. Newborn 1B showed a few changes, such as vasocongestion and edema in the parenchyma. Macroscopic vascular changes, such as edema, congestion and haemorrhage, were also observed in newborns in G2, although the changes were more discrete than those observed in G1. The newborns in G3 did not show any macroscopic changes. The integrity of the cerebellum (CB), brain stem (BS) and cervical spinal cord was observed in all newborns from G2 and G3. The average weight of the newborns in G1 was 83.67 g, whereas newborn 1D weighed only 68 g. The average weights of newborns in G2 and G3 were 94 g, 67 g and 96 g, respectively (Fig. 2K). The brain circumferences of the G1 offspring were 12.8 cm (1B), 11.5 cm (1C) and 13.5 cm (1D) (Fig. 2K). The most remarkable finding was the ventriculomegaly seen in newborn 1D, which explains the projection of the posterior region of the brain.

Microscopic analysis of the CNS.
Semi-quantitative analysis of the histological sections of the CNS of newborns from all infected groups showed histopathological alterations, usually to a severe degree (Fig. 3A,B). Among the histopathological lesions, the most frequent were vascular changes, cell death, vascular and perivascular edema, and inflamed and healed lesions (Fig. 3C). Analysis of variance (p < 0.0001) of these lesions indicated the prevalence of vascular lesions in the cerebral epithelium of newborns from all groups analysed (G1-G3), with extensive vascular proliferation, vasocongestion, pericellular and vascular edema and cell death. In G1, newborns 1C and 1D presented lesions with greater severity and more abundance in the evaluated brain areas. Figure 3B shows how the lesions were distributed among the infected groups. The brain of newborn 1D had increased endothelial haemorrhagic changes ( Fig. 3D,J,L), neuronal death, perivascular edema, gliosis and satellitosis in the frontal cortex (FC) (Fig. 3D); vascular proliferation in the temporal cortex (TC); and gliosis and haemorrhagic lesions in the hypothalamus (HT) (Fig. 3E). In the areas of neuronal death in the occipital cortex (OC), CB and BS, cell death was frequent. Newborn 1C had lesions in several analysed areas of the CNS that ranged from medium to severe (Fig. 3B), with the most severe injuries in the OC, CB and BS, where there was severe and abundant necrosis. Interestingly, newborn 1B had minimal lesions in the focal areas, characterized by endothelial reactivity, small foci of edema, and red blood cells spilling out of the FC (Fig. 3F). The newborns in G2 also showed lesions, although to milder degrees than those seen in the newborns in G1. Newborn 2B exhibited the greatest effects, with myelin degeneration similar to that seen in severe Wallerian degeneration (Fig. 3G). The other animals had a minimal degree of injury (Fig. 3A). The only newborn in G3 showed minimal lesions (Fig. 3H) in the OC, CB and BS, with low endothelial reactivity and a few areas of edema and congestion. In newborns from G4 (Fig. 3I), few injuries (only edema and minimal vasocongestion) were observed.
ZIKV detection in fetal tissues. Using RT-qPCR, high ZIKV loads were detected in fetal brain tissues from newborn 1D (1.48 × 10 1 RNA copies/mg of tissue), while the virus genome was undetectable in the other newborns (using a standard curve analysis, the estimated limit of detection for the RT-qPCR assay was ~ 25 RNA copies/mg of tissue). However, the presence of antigens against ZIKV was confirmed in all newborns by immunohistochemistry (IHC). The areas of the brain with the highest amount of immunostaining are shown in Fig. 4B www.nature.com/scientificreports/ www.nature.com/scientificreports/ significant inflammatory infiltrate was observed; however, in G1 animals (1C and 1D), which had the highest degree of lesions; an extensive perivascular lymphomononuclear infiltrate was observed close to the affected lateral ventricle.
Immune response against ZIKV in CNS tissues. Analysis of the cytokine and inflammatory marker profile in CNS tissue samples from newborns in G1, who had the highest number of lesions, showed high IL-10 (p < 0.19) and IL-12A (p < 0.0001) expression; followed by IL-4 (p < 0.0001), with a positive correlation (ρ = 1) in infant CNS tissue; and IL-1β (p < 0.068), TGF-β (N.S.) and TNF-α (N.S.); low IFN-β expression (p < 0.0355) was detected. We observed a predominance of Th2 markers over Th1 markers, such as IL-10 and IL-4, which act in tissue reconstruction and modulation of the expression of other cytokines, such as TNF-α, accompanied by the pro-inflammatory cytokines IL12-A, TNF-α, IL-1β and IFN-β (Fig. 3B). Immunoexpression of cytokines, especially IL-4, IL-10 and IL-12A, occurred in the brain parenchyma. The expression was mainly observed in areas containing neurons, in addition to areas with dead neurons and low mononuclear inflammatory infiltrate.
In particular, an increase in Caspase 3 expression (p < 0.04) was observed in the parenchyma, especially in areas with intense neuronal damage to the cortex, where intense staining was observed in vacuolated and apoptotic neurons (Fig. 4A). Cytokine expression levels in specific brain areas are shown in Fig. 4C.

Discussion
This study showed that experimental infection of S. collinsi females in different gestational thirds (GT) with the Asian genotype of ZIKV produced results similar to those of infection in humans, demonstrating the epidemiological pattern of CZS occurrence in susceptible newborns. CNS disorders were mainly observed in newborns from females infected in the first period of pregnancy, similar to findings in babies with CZS 3,8,11,22,33 .
The first trimester of pregnancy in humans and the corresponding periods in these primates (1st GT and 2nd GT) are fundamental for proper development of the main organ systems (organogenesis), especially the brain and its associated neuroprogenitor cells 34 . Comparatively, the gestation time of NHPs is approximately 162 days, with the 1st GT of S. collinsi lasting approximately 54 days (Fig. 1B), while in humans, gestation lasts approximately 280 days. Relevantly, in this study, we were able to mimic severe lesions in newborns when maternal infection occurred after detection of chorionic vessel formation (Fig. 1B), thus demonstrating that this stage of embryonic development is an early period of organogenesis equally crucial in squirrel monkeys and humans.
Two newborns (1C and 1D) in G1, whose mothers were infected at 41 Gd, coinciding with the period after separation of fetal circulation from maternal circulation, exhibited the most serious injuries (Fig. 1B). On the other hand, two situations occurred for animals infected before complete formation of the chorionic vessel: miscarriage (1A, infected at 34 Gd) and minor injury (1B, infected at 36 Gd). The balance between virus replication and both the intensity and speed of humoral response establishment could be a determining factor for irreversible tissue damage or virus clearance.
Combining the findings reported by Imbeloni et al. 26 with the results of this study, in animal 1A, the virus began to be detected at 2 dpi, and the highest titre occurred at 7 dpi, the same time as spontaneous abortion and without any IgM detection. For animal 1B, anticipated and higher maternal IgM titres were associated with late viremia and below levels seen in the other females in this group, which could have been crucial in protecting the fetus from ZIKV infection and minor brain damage.
Robust maternal adaptive immune responses as well as the placenta form a protective barrier for the fetus in early pregnancy. However, once the formation of the placental barrier is complete, maternal immunity can be restricted, and the main defences against infection are driven by the immature immune system of the developing fetus 5,35 . Consequently, the chance of mild or severe brain damage development in the fetal brain increases, such as in newborns 1C and 1D. Furthermore, the maternal response likely contributes to the lesions in the fetus. In animal 1D, the higher and prolonged maternal viremia associated with lower and delayed IgM detection and persistent viral infection in the fetus until delivery resulted in severe damage to the fetal brain, similar to CZS. Previous studies have reported that persistent viral infection in maternal tissues can provide a reservoir that facilitates vertical transmission of ZIKV, which provides a basis for understanding the wide range of fetal developmental abnormalities caused by ZIKV 36,37 .
The frequency of CZS-like lesions in our study (67%, 2/3) was close to that found in females of other NHP species infected early 8,11 . Some studies have been able to reproduce ZIKV infection in NHPs, including rhesus, cynomolgus and even neotropical primates, such as marmosets and different species of squirrel monkeys. However, lesions consistent with CZS have not been observed in such studies 22,37 .
Biometric differences, such as differences in weight and brain circumference, were evident among newborns in G1 (Fig. 2J,K). These data suggest that in newborn 1D ZIKV infection led to an increase in brain circumference, with an increase in the occipital region due to ventriculomegaly (Fig. 2E,F), and general growth restriction, although we did not observe gross microcephaly. As organogenesis in squirrel monkeys is nearly complete at the end of the 1st GT, the malformations consistent with CZS identified in this newborn (1D) did not occur in newborns from other infected groups in the later stages of pregnancy which often occur in human babies 24,36 .
Macroscopic examination of the brains of G1 newborns showed severe vascular lesions, such as vasocongestion, subarachnoid haemorrhage and edema (Fig. 2E-H), which were similar to brain lesions caused by other encephalitic flaviviruses [38][39][40] . In newborn 1D, the right occipital lobe in the region of the 3 rd and lateral ventricles was enlarged (Fig. 2E). Lissencephaly and reduced cerebral gyri in the occipital region of the right hemisphere were seen in newborns 1D and 1C (Fig. 2D,G). Newborn 1D presented lambda malformation without closure of the sagittal and lambdoid sutures (Fig. 2I). Lesions such as lissencephaly, reduced cerebral gyri in the occipital region and ventriculomegaly have been previously described in rhesus monkeys by Martinot et al. 8 and Adams Waldorf et al. 11 . In human children exposed to ZIKV during pregnancy, magnetic resonance imaging (MRI) revealed ventriculomegaly, simplified gyral pattern and lissencephaly 23,[40][41][42] .
The results of our macroscopic and histopathological analyses, integrated with the immunohistochemistry and immunoexpression data and the evidence of viral expression in the tissues, demonstrated the possibility of neurological persistence of ZIKV infection in the CNS until birth, as indicated by the expression of proinflammatory mediators 33,34 . We detected low IFN-β expression, mainly in the areas of the OC, PC/TC and HR, in the nervous tissues of newborns in G1 (Fig. 4B,C), likely related to development of a classic antiviral response (Th1) against ZIKV, which is crucial for inhibiting viral replication [36][37][38][39] . Based on previous studies, ZIKV appears to selectively inhibit translation of IFN-I mRNA, while translation of other host viral proteins remains intact. www.nature.com/scientificreports/ Blocking IFN-I translation appears to be an exclusive feature of ZIKV because other viruses do not block translation or secretion of this cytokine 50 . In animals, low IFN-I and IFN-II levels promote ZIKV replication and spread of ZIKV infection to the CNS 51 . We believe that the low expression of this cytokine in the nervous tissue of newborns indicates the specific nature of the immune avoidance mechanism that likely contributes to the viability of ZIKV until the birth of the fetus. Although replication inhibition mechanisms become active to a small extent, ZIKV remains in the escape sites of the CNS 33,36,37 .
The production of different cytokines has a strong influence on the infection environment and modifies the immune response mechanisms. The Th1 cell profile mainly produces IFN-γ, IL-12 and TNF-α, and malfunction of this response has been correlated with severe viral infections 41 . Our data demonstrated an increase in IL-12A and IL-4 expression. The balance between IL-12, favouring Th1 responses, and IL-4, favouring Th2 responses, promotes a balanced immune response in situ 42 . An increase in IFN expression is supported by the presence of IL-12A, a cytokine that classically induces IFN-γ production, which in turn can modulate microglial activity in response to the presence of ZIKV 11,46,52 .
Sousa et al. 47 and Azevedo et al. 46 linked the Th1 lymphocyte response to positive immunostaining for IFNγ, IFN-α, IFN-β, IL-6, IL-12A, IL-1β and TNF-a. Our panel showed positive IFN-β, IL-12A, IL-1β and TNF-α immunostaining in the tissues of squirrel monkeys, demonstrating a similar profile in the nervous tissues of the newborns in G1, who were born after maternal infection. The increase in the level of pro-inflammatory cytokines suggests a coordinated protein response that may be associated with induction of cellular damage mechanisms. In fact, cytokines are the main mediators that trigger neurotoxicity and the production of reactive oxygen and nitrogen intermediates directly involved in cell death. In addition, IL-1β and TNF-α may also be involved in the apoptotic cell death mechanism because they are associated with activation of death receptors and inflammasome triggers, respectively 49,53,54 .
Positive immunostaining for IL-4, IL-10, IL-33, IL-37 and TGF-1β, indicative of the presence of a Th2 response, has already been reported in other viral encephalitis cases and may be related to simultaneous development of suppressive agents and an antiviral response, which are necessary for elimination of the infectious agent without causing major inflammatory damage 46,47,49,54,55 . The involvement of the Th2 profile indicates cell death via necrosis and apoptosis, and the induction of the M2 microglial response indicates that involvement of the Th2 profile is closely related to the pathogenesis of ZIKV. The Th2 profile is related to cell death, giving rise to a more favorable environment for viral replication and activating M2 macrophages 11,46 .
IL-4 and IL-10 immunoexpression was increased in the nervous tissue of G1 newborns, in which necrosis of neurons in the cortical layer of the OC and BS/CB was much more intense. Apoptosis is likely the main type of cell death involved in the development of CNS damage induced by ZIKV. We observed increased Caspase 3 expression in the parenchyma, including in areas of the cerebral cortex, with a prevalence of immunostaining in the OC, PC/TC and FC (Fig. 4B,C). This finding is consistent with results from in vitro studies suggesting that apoptosis is the main factor in neuronal cell death and reduced brain mass in humans 55,56 .
Our data also demonstrate that the in situ immune response against ZIKV in the CNS of newborns of this species is complex; the expression of Th2 and Th1 cytokines is predominant, but other cytokine profiles are involved to a lesser extent. Mechanisms of neuronal cell death include necrosis and apoptosis, which are induced by different immunological factors and likely by direct action of the virus itself, which has already been reported  Table 1). The scoring system is defined in Table 1. (B) Maximum score for neuropathological lesions in newborns of G1, G2, G3, and G4 (uninfected newborns); statistically significant difference between groups were determined by ANOVA one way (p < 0.0001) with Tukey's posttest for multiple comparisons; red bars indicate the values that are statistically significant, G1 had ρ = 0,0003, whereas G2 and G3 presented no significant differences. (C) Most frequent pathological lesions found in the CNS of G1 ZIKV-infected newborns. The p values reflect ANOVA one way test (p < 0.0001), vascular injuries were the most frequent in the CNS of infected newborns, followed by scarring, neuron death, and less frequent inflammatory process. (D) Ventriculomegaly right side (3rd ventricle) from newborn 1D (G1), it is observed hemorrhage (red arrows), slight mononuclear inflammatory infiltrate, perivascular necrosis/apoptosis (black arrows) (magnification ×100 and scale bar 25 µm), perivascular edema, and vasocongestion (box, magnification ×400 scale bar 50 µm). (E) Histopathological image of 3rd ventricle from 1C newborn brain showing vascular congestion (large black arrow), perivascular necrosis/apoptosis (thin black arrows), and satellitosis (black circle (magnification ×100 and scale bar 25 µm). (F) Histopathological image of OC lobe 1B newborn from G1 (magnification ×100 and scale bar 25 µm), few injuries compared to the other two newborns in the same group (Fig. 2D, E)  www.nature.com/scientificreports/ in humans 46,47 . However, studies in experimental models elucidate the pathogenic mechanisms and immunological factors involved in the host response against ZIKV infection in the CNS. Although our findings help clarify some aspects of CZS, a better understanding of the role of placental permissiveness is still necessary and would improve understanding of the mechanisms involved in the different tissue responses observed in newborns in G1, whose mothers were infected during the first GT. However, despite the need for further studies in this animal model, our data suggest strong vascular involvement and dysfunction of neuroprogenitor cells during fetal development, with involvement of pathological mechanisms that occur early in pregnancy. Thus, in addition to assisting in chronological comparison of the neuroembryonic development of S. collinsi with that of humans, these findings corroborate what has already been observed in other species of NHPs and increase knowledge of the pathogenesis of CZS.

ZIKV strain BeH815744. The isolate was provided by the Section for Arbovirology and Haemorrhagic
Fevers (SAARB) of the IEC. The ZIKV BeH815744 strain (GenBank KU365780) was isolated, sequenced and genetically characterized as previously described 4 . The viral isolate was amplified and titrated using the Vero cell line; cells were grown in 199 medium supplemented with 2% fetal bovine serum (FBS) and antibiotics 10 . Females were infected via the intradermal (i.d.) route with 1.0 × 10 5 PFUs of virus inoculum, which was injected into the right intercostal space above the nipple. Study design. Ten female squirrel monkeys (S. collinsi) from a colony in the CENP were naturally mated.
Pregnancy was confirmed by US, and for the purposes of this study the gestational time was considered to be approximately 162 days, as previously published 26 . According to the GT at the time point of i.d. ZIKV inocula- www.nature.com/scientificreports/ tion, the females were divided into three groups: G1 (n = 4), G2 (n = 3) and G3 (n = 1). One group (G4) with two non-infected pregnant monkeys was used as a control, and their newborns were sacrificed (4A and 4B) for histopathological comparisons (Figs. 1A, 2B).

Necropsy of animals.
The necropsy of the mothers and their newborns was performed 4 days after delivery. Macroscopic examination of the CNS was performed and noted on a specific form. Samples from each area of the brain, as well as from the cervical spinal cord, were fixed in 10% buffered formaldehyde and processed after 48 h.
Histological analysis of CNS tissues. The assessment of the CNS was performed as described by Mlakar et al. 14 . Sections (5 μm) of different anatomical regions of the CNS, the frontal (FC), parietal (PC), temporal (CT) and occipital (OC) cortices, and the thalamus (TH), hippocampus (HC), brain stem (BS), cerebellum (CB) and cervical spinal cord, were included in paraffin blocks and stained with hematoxylin and eosin (H&E). The sections were hydrated in a decreasing alcohol gradient (100%, 80%, 70% and 50% ethyl alcohol), washed, stained with hematoxylin dye, washed again and stained with eosin dye, then washed, sequentially dehydrated in an increasing sequence of alcohol (ethyl alcohol 50%, 70%, 80% and 2 repetitions at 100% concentration) and finally two fixations for the conservation of the material (xylol I and II).The stained sections were analysed by optical microscopy (Zeiss, Oberkochen, Germany), with magnifications of 100× and 400×, by two independent pathologists, and the slides were viewed proximal to the meninges, parenchyma and perivascular region. Quantitative analysis was performed (Table 1)  Immunohistochemistry. The streptavidin-alkaline phosphatase method was adapted to detect the viral antigen using a polyclonal anti-ZIKV antibody produced in mice at the IEC. Based on previous studies 27,28,57 , the peroxidase method was used for tissue immunostaining with specific monoclonal antibodies to track cytokines and markers of apoptotic cell death pathways to evaluate a small panel designed for this study and containing the following markers: IL-10, TGF-β, IFN-β, IL-12A, IL-1β, Caspase 3 and TNF-α (see Supplementary Data 2 for specifications and dilutions) 29 .
Detection and quantification of the ZIKV genome. Infected tissues were mixed with PBS (pH 7.4) on Tissuelyser (Qiagen, Carlsbad, USA) and subjected to viral RNA extraction using a TRIzol Plus RNA Purification kit (Thermo Fisher Scientific, Carlsbad, USA). The RT-qPCR protocol 16 was adapted to use QuantiTect Probe PCR kits (Qiagen, Carlsbad, USA), and RT-qPCR was performed on a 7500 Real-Time Fast PCR system (Thermo Fisher Scientific) with the standard curve and synthetic amplification target included in the PUC plasmid for RNA quantification, following the methods described by Nunes et al. 31 . The RNAase P gene was used as an internal endogenous control 32 . The limit of detection was ~ 25 RNA copies/mg of tissue, as reported by Lanciotti et al. 30 . Data on viremia (RT-qPCR) and the immune response (ELISA IgM) of mothers are available in Supplementary Data 1.

Statistical analyses.
For univariate analysis, frequencies and measures of central location and dispersion were obtained. The hypotheses were tested using one-way ANOVA with a Kruskal-Wallis test. The level of significance adopted was α = 0.01 (p ≥ 0.01, not statistically significant (N.S.); p < 0.01, significant; p < 0.001, very significant; and p < 0.0001, extremely significant). We used Pearson's correlation (ρ), assuming only values between − 1 and 1, to indicate positive or negative correlations. All analyses were performed with Prism 6.0 software (GraphPad, San Diego, USA). Multifocal small lesions in a single brain section Endothelial reactivity, perivascular necrosis, large foci of extravasated red blood cells outside a vessel (hemorrhage) 3 Moderate Multifocal larger lesions in a single brain section Large aggregate of persistent neuroprogenitor cells in unusual patterns in the brain with or without vascular changes/vasculitis in brain or spinal cord 4 Severe Multifocal larger lesions in a single brain section with similar lesions in other brain sections; or fetal loss associated with tissue viremia Necrosis/apoptosis associated with vascular change/vasculitis, increased numbers of neuroprogenitor cells (gliosis, satellitosis), 'red-dead' neurons, congestion, and neuron clustering