**4. Discussion**

TJs in the STB are a cornerstone of the BPB that protects the fetus from toxins and pathogen infections. Given the devastating consequences of ZIKV fetal infection, in this work, we explored the expression of E-cadherin and TJ proteins in the STB of human placentae of women infected with ZIKV. By immunofluorescence, we observed that E-cadherin, claudins -1, -3, -4, -5, and -7; JAM-B; occludin, ZO-1, and ZO-2 were present in the STB of chorionic villi from both control and ZIKV-infected women. Only JAM-A and claudins -2 and -10 (data not shown) were not detected in these tissues. In contrast, in mice, a strong induction of claudin-10 was observed in the decidua at pregnancy day 4.5, although in humans, claudin-10 was not detected in first trimester decidual cells [35].

Claudins -1, -3, and -5, JAM-B, occludin, ZO-1, and ZO-2 are present in both the apical and basal surfaces of the STB layer, whereas E-cadherin and claudins -4 and -7 concentrate at the basal surface of the STB in contact with the underlying CTB or chorionic parenchyma. This pattern of expression was not altered by ZIKV infection.

The TJ proteins that were more conspicuously expressed in the STB layer were claudins -1, -3, -4, -7, JAM-B, occluding, and ZO-2. These results agree with previous observations showing strong expression of claudin-4 in trophoblastic cells during all trimesters of human pregnancy [11] and with our previous results with occludin and claudins -1, -3, and -4 in the placentae of control women and with preeclampsia [9]. In addition, in mice placentae, a real-time PCR study revealed a higher level of expression of mRNA for claudins -1, -2, -4, and -5 in comparison to all other claudins in the family, while a Western blot revealed an increased expression of claudin-4 and -5 and a decrease in the content of claudin-2 as pregnancy advances from day 12 to 20 [36].

Here, we observed that both occludin and ZO-1 delineated the STB layer in a moderate manner in comparison to the strong staining observed in the chorionic vessels. Therefore, the decreased content of occludin detected by Western blot in the placentae from ZIKV-infected women is most likely due to a reduction in the amount of occludin present in endothelia, particularly since the immunofluorescence signal of occludin at the STB layer did not decrease in the placentae from ZIKV-infected women. Therefore, we think that the diminished expression of occludin in the endothelia of the chorionic parenchyma does not contribute to the leakiness observed in the STB layer of chorionic villi in placentae derived from ZIKV-infected women.

These observations, however, do not minimize the importance of ZO-1 and occludin for a healthy STB, particularly since in hydatidiform moles, characterized by hyperplasia of the trophoblastic tissue and distention of the chorioninc villi by fluid, the expression of ZO-1 and occludin is downregulated and their distribution in the STB changes from the cell borders to the cytoplasm [10]. In addition, ZO-1 appears to play a crucial role in the fusion of trophoblastic cells into a syncytium, as ZO-1 expression at intercellular boundaries decreases during fusion; the treatment of primary human trophoblastic cells in culture with ZO-1 siRNA blocks this process [37].

The expression of occludin in fetal vessels of human placenta is observed mainly at term [10], and occurs in the secondary chorionic villi in large and intermediate vessels but not in terminal exchange vessels, as we and others have shown [9,38]. Instead, the expression of *ZO-1* in fetal endothelia has been observed throughout gestation [10] and among the whole placental vascular tree [38]. In this respect, it is interesting to note that one of the reasons that ZO-1 knock out mice are embryonic lethal is due to abnormal angiogenesis in the yolk sac [39].

With regards to ZO-2, it is interesting to note that in contrast to *ZO-1*, this protein was abundantly expressed in the STB layer of human placentae but barely detectable in chorionic vessels. Unlike ZO-1 knock out mice, ZO-2 knock out mice do not die due to alterations in angiogenesis [39]. Instead, ZO-2 knock out mice are embryonic lethal due to defects in the development of the extraembryonic tissue. This was demonstrated when the injection of ZO-2−/− embryonic stem cells into wild type blastocysts generated viable ZO-2 chimera mice [40].

Our results indicate a reduced expression of claudin-4 in BeWo cells when ZIKV was added to the basolateral surface and in the STB layer of chorionic villi from ZIKV-infected women. The change observed in the chorionic villi cannot be attributed to the preeclampsia present in two of the ZIKV-infected donors, because the placentae of ZIKV-infected women without preeclampsia also exhibited a reduced claudin-4 expression. In addition, a histochemistry study reported a slight increase in *claudin-4* expression in the trophoblasts of preeclamptic placentae [11]. Finally, by Western blot, we previously observed that the amount of claudin-4 in the chorionic villi does not vary with preeclampsia [9].

The di fferences in the gestational age between the ZIKV-infected placentae and the controls arose because women infected with ZIKV were subjected to earlier cesarean sections due to their high-risk pregnancies in order to better protect fetal health, while healthy women in the control group had cesarean sections at the expected time for full-term pregnancies. Nevertheless, the di fferences in claudin expression here observed cannot be ascribed to the variation in gestational age between the groups, because in humans, there are no di fferences in the level of expression of claudin-4 in trophoblastic cells of chorionic villi, between the three trimesters of pregnancy [11].

The combination and mixing ratios of claudin species determines the barrier properties of TJ strands [41], and the alteration of a single type of claudin can significantly alter the permeability and transepithelial electrical resistance of a tissue [30,32]. Therefore, the decrease in claudin-4 expression in placentae from ZIKV-infected woman may have a big impact in the paracellular transit through the STB, particularly since this claudin functions as a cation barrier [30,32] or an anion pore [31]. Accordingly, a significant decrease in permeability and an increase in TER have been observed in MDCK cells after claudin-4 transfection [30,32].

Claudin-4 in the renal collecting duct interacts with claudin-8, and their association is required to form a paracellular chloride channel [31]. Therefore, it may be important in the future to determine whether claudin-8 is expressed together with claudin-4 in the STB layer in human placentae, and whether its expression is altered in ZIKV-infected placentae.

Claudin-4 is a critical claudin for the establishment of a permeability barrier to protect the developing embryo. When the blastocyst enters the uterus, the process of implantation and placentation starts. The first contact is established between the blastocyst trophoectoderm and the uterine epithelium. Once the blastocyst attaches, the process of decidualization is triggered, involving the stromal epithelial transition in which uterine stromal cells di fferentiate into decidual cells surrounding the implanting blastocyst. In this event, the trophoectoderm acts as a stimulus for the creation of a TJ permeability barrier in stromal cells that protects the embryo from the passage of injurious maternal immunoglobulins [42,43]. In rat stromal cells of the uterus, we have observed that claudin-1 is present in all gestational days, and that ZO-1 appears until day 6, albeit at both implantation and non-implantation sites, while claudins -3 and -4 appear until gestational day 7 and only at implantation sites [44], reinforcing the view of claudin-4 as an important TJ proteins for embryonic development. Moreover, in rat uterus, by the time of implantation of the blastocyst at gestational day 6, when the network of TJ strands increases 3-fold in depth along the lateral plasma membrane and displays more branches and interconnections with neighboring strands [45,46], claudin-4 is detected for the first time at the basolateral membrane of uterine epithelial cells [44]. Similarly, in human endometrium, an increase in claudin-4 mRNA is found during the implantation window [47–49], thus suggesting a critical role of claudin-4 during implantation.

The role of claudin-4 in human placenta is highlighted by the observation that its expression increases in hydatidiform moles and in maternal diabetes [11]. In human placentae derived from assisted reproductive technology, claudin-4 mRNA diminishes, and this change is accompanied by an increase in claudin-8 mRNA [50]; while in mice placentae, claudin-4 augments as gestation advances from days 12 to 20 and after the administration of the estrogen receptor antagonist ICI 182,780 and the progesterone receptor antagonist RU-486 [36].

How ZIKV alters the Cl-4 expression of STB without actively replicating in these cells remains an open question. It has been observed that a wide array of viruses use integral proteins located at the apical junctional complex (AJC) of epithelial cells (for review see [51]) as cellular receptors, including for example hepatitis C virus, that associates to claudins -1, -6, and -9 [52,53]. The use of such proteins is important for the entry of viruses into epithelial cells, but also implies the disruption of the AJC that compromises the integrity of the epithelial barrier in consequence. Thus, West Nile virus specifically induces the endocytosis of claudin-1 and JAM-A [54], and in rotavirus, the VP8 protein, generated from the proteolytic cleavage of spike protein VP4, opens the TJs of epithelial cells in a reversible and dose-dependent manner [55]. VP8 has several segments with high similarity to domains present in the extracellular loops of claudins and occludin; hence, it was proposed that VP8 opened the TJs by competing with the homotypic interactions established among the extracellular domains of certain claudins and occludin. Another interesting protein in this respect, derived from a microorganism, is *Clostridium perfringens* enterotoxin (CPE). This toxin opens the paracellular barrier due to the selective removal of claudins -4 and -3 from the TJ [56,57]. In the case of ZIKV, the e ffect of structural E and M proteins on TJ proteins has not ye<sup>t</sup> been explored.

The observation that only the basolateral exposure of ZIKV reduced TER and claudin-4 expression of BeWo monolayers is not surprising, as a similar situation had been observed with several viruses. Thus, the adenovirus fiber protein can only access its CAR receptor at the apical junctional complex and open the TJs when added to the basolateral surface [20]; HSV-1 only infected epithelial cells if added to the basolateral surface or if depletion of extracellular calcium had weakened the strength of the AJC to allow the virus to access its nectin receptor [58,59]; and hepatitis C virus first localizes with the epidermal growth factor receptor at the basolateral membrane and then accumulates at the TJ and associates to claudin-1 and occludin [60,61]. These results sugges<sup>t</sup> that ZIKV passed from the maternal basal decidua to the fetal invading CTB and the cell columns of CTB, or through the transport system facilitated by the neonatal Fc receptor to transcytose across the STB layer, could open the paracellular pathway of the STB layer due to its presence in the parenchyma of chorionic villi that faces the basolateral surface of STB cells. Hence, the opening of the TJ in the STB could occur not as an initial step in the vertical transmission of ZIKV, but as a consequence of chorionic villi infection.

The heterogeneous maturation of chorionic villi, Hofbauer cell hyperplasia, and intravillous calcifications that we observed in ZIKV-infected placentae have also been reported in other studies [15,62–64]. In this respect, alterations in Hofbauer cells homeostasis are known to be associated with placental pathologies involving infection, inflammation, and inadequate placental development [65]. With regards to the diameter of chorionic villi, as the third trimester of pregnancy advances, stem villi branch into distal villi; in consequence, the diameter of chorionic villi decreases [66]. Thus, the higher diameter of chorionic villi observed in placentae from ZIKV-infected women suggests villous maldevelopment, although an e ffect due to the di fferent ages of ZIKV and control placenta cannot be disregarded. Nevertheless, it should also be mentioned that this e ffect could be related to the altered expression of claudin-4, since in both mice and humans, during placental development, frizzled 5 induces the disassociation of cell junctions for chorion branching initiation through the downregulation of ZO-1, claudin-4 and claudin-7 in trophoblast cells [67]. Therefore, another important aspect to study in the future in ZIKV-infected placentae could be the expression of frizzled 5.

In summary, our results indicate that the chorionic villi of placentae from women infected with ZIKV display Hofbauer cell hyperplasia, an increased diameter of microvilli and intravillous calcifications, while the study of the STB layer of these placentae shows a decreased expression of claudin-4 and ruthenium red permeability, suggesting that these placentae are leakier than the normal, control ones (Figure 6). These observations allowed us to propose the paracellular pathway of the STB layer as a route of vertical transmission of ZIKV. However, the observation that ZIKV only reduced the TER of a trophoblast cell line when added to the basolateral surface raises the possibility of seeing the opening of TJs in the STB as a consequence of ZIKV infection of the chorionic villi.

**Figure 6.** Schematic representation of the changes observed in chorionic villi of ZIKV-infected women. Chorionic villi derived from women infected with ZIKV during pregnancy, displayed several alterations including Hofbauer cell hyperplasia, increased diameter of microvilli, intravillous calcifications, and a STB layer with a diminished expression of claudin-4 and permeable to ruthenium red passage though the paracellular pathway.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4409/8/10/1174/s1, Figure S1. E-cadherin presence in the basolateral membrane of the STB was not altered in placental tissue from ZIKV-infected women, Figure S2. The expression of claudin-1 in the STB of chorionic villi in human placenta increased with ZIKV infection, Figure S3: Claudin-3 was strongly expressed in vessels in the chorionic parenchyma, Figure S4. Claudin-5 was strongly expressed in chorionic villi vessels and faintly stained the STB cell layer, Figure S5: Claudin-7 is present in the basolateral surface of STB cells and infection with ZIKV did not alter its expression, Figure S6. ZO-1 strongly stains the chorionic vessels, Figure S7. ZO-2 stained the STB layer of placentae, Figure S8. JAM-B was present in the STB cell layer and its expression was not affected by ZIKV infection, Figure S9: JAM-C expression was abundant in chorionic vessels but scarce in the STB layer, Figure S10: Chorionic villi stained with haematoxylin and eosin in D1 placenta from a ZIKV-infected woman, Figure S11: Chorionic villi stained with haematoxylin and eosin in D5 placenta from a ZIKV-infected woman, Figure S12: Chorionic villi stained with haematoxylin and eosin in D8 placenta from a ZIKV-infected woman, Figure S13: Chorionic villi stained with haematoxylin and eosin in D10 placenta from a ZIKV-infected woman, Figure S14: Chorionic villi stained with haematoxylin and eosin in D9 placenta from the control group, Figure S15: Chorionic villi stained with haematoxylin and eosin in placenta D17 from the control group, Figure S16: Chorionic villi stained with haematoxylin and eosin in placenta D18 from the control group, Figure S17: Chorionic villi stained with haematoxylin and eosin in placenta D19 from the control group, Figure S18: Chorionic villi stained with Masson's trichrome stain in D1 placenta from a ZIKV-infected woman.

**Author Contributions:** Conceptualization, L.G.-M.; methodology, L.G.-M. and J.M.; formal analysis, J.M. and L.G.-M.; investigation, J.M., Y.V.-V., D.M.-T., L.A., A.E.-N., M.G.-H., J.E.M.-M.; M.S., B.C.-M., J.E.M.-M.; resources, L.G.-M., G.E.-G. and S.L.; data curation, J.M.; writing—original draft preparation, L.G.-M.; writing—review and editing, L.G.-M., J.E.L., G.E.-G., M.S.; validation, L.G.-M.; visualization, L.G.-M. and J.M.; project administration, L.G.-M.; funding acquisition, L.G.-M., J.E.L. and G.E.-G.

**Funding:** This work was supported by gran<sup>t</sup> C-623/2016 of the Mexican National Council of Science and Technology (Conacyt) to L.G-M., by gran<sup>t</sup> CB-254461 of Conacyt to J.E.L.; by a gran<sup>t</sup> from the Miguel Alemán Valdés Foundation to L.G-M.; by a SEP-Cinvestav gran<sup>t</sup> to L.G-M. and by INPer project 212250-1000-10107-01-16. Jael Miranda was a recipient of a doctoral fellowship from Conacyt (262817) and the Mexiquense Council of Science and Technology (Comecyt, 2018AD0035-1).

**Acknowledgments:** We would like to thank the technical support of Angélica Silva-Olivares. **Conflicts of Interest:** The authors declare no conflict of interest.
