**3. Discussion**

During pregnancy, HIV-infected mothers are treated with two NRTI and one PI to prevent the viral transmission to the fetus [1]. According to several studies those treatment, notably the PI, have secondary effects such as pre-term birth, pre-eclampsia, and intra uterine growth restriction [3,4]. Moreover, PI impair steroidogenesis, inducing a decreased P4 level in maternal blood and impaired adrenal function in neonates exposed in utero [3,5,6]. However, very little is known about the placenta, which produces P4, a steroid hormone mandatory for the maintenance of pregnancy [7,8,35]. Our aim was to investigate in vitro whether PI such as RTV and LPV disturb human placental steroidogenesis, focusing on the P4, mitochondria and main intracellular pathways potentially targeted.

For many years, the syncytiotrophoblast (ST) has been considered as the main endocrine tissue of the placenta as it is in direct contact with maternal blood in the intervilli chamber [36]. As in the chorionic villi, the ST arises from the differentiation of the villous cytotrophoblast (VCT), we used our in vitro model to extensively characterize P4 production in human villous placenta. In our previous study [11], we established that placental steroidogenesis is not restricted to ST only but starts early in the VCT, which expresses the cholesterol transporter MLN64 and the key enzymes P450SCC and HSD3B1 required for P4 synthesis and secretes significant levels of P4. We confirmed these findings in this work, pointing out that the whole trophoblast is able to produce placental steroid and protein hormones. This is in agreement with previous studies showing that VCT and also extravillous cytotrophoblast (EVCT) produce significant amounts of hCG and its subunits [37]. We confirmed in this in vitro study that the endocrine production increases with the morphological differentiation of VCT into ST, hCG secretion increase being associated with ST formation. We also established that this increase in hCG secretion is correlated with an increase in P4 synthesis and secretion [11].

The mitochondria is a key organelle involved both in steroidogenesis and in trophoblast differentiation [16,27,30]. We have previously demonstrated that the differentiation of VCT into ST is associated with a decrease in mitochondrial transmembrane potential. This functional change in mitochondria could explain the observed changes in P4 synthesis [11]. Indeed, a previous study on placental tissue demonstrated that a decrease in transmembrane potential in mitochondrial fraction is associated with a high steroidogenesis activity [38]. In addition, we showed that in vitro differentiation of VCT into ST is associated with morphological changes in mitochondria using transmission electronic microscopy. Indeed, in our in vitro model, VCT are predominant at 24 h of culture, whereas ST is formed at 72 h of culture. In VCT, mitochondria are larger than in ST. Moreover, mitochondria in VCT present cristae with typical structure, whereas they present in ST an atypical structure of cristae and a denser matrix than in VCT, in agreement with previous histological studies [20]. These structural modifications are known to be related to fusion/fission dynamics of mitochondria [21]. However, in our model, we observed no modification in the expression of two main fusion proteins Mfn2 and OPA1 during differentiation of VCT into ST. These results are not consistent with a previous study in BeWo cells that demonstrated a decrease in Mfn2 and OPA1 expression but, under forskolin, induced differentiation [27]. This suggests that the observed changes in mitochondria depend on the model used, i.e., primary trophoblast culture in our model compared to choriocarcinoma cell lines and on experimental conditions, i.e., basal or stimulated by forskolin. Mitochondrial dynamics are known to be involved in the regulation of steroidogenesis. Some studies demonstrated that fission process is necessary for steroidogenesis [22,27], while others reported on the contrary that steroidogenesis relies on fusion process [30,39]. In our model, it is likely that fusion process is not necessary for P4 synthesis as we found no change in Mfn2 and OPA1 expression during the ST formation. It would thus be of interest to evaluate the expression of other factors such as fission factors Drp1 and Fiss1. An increase in Peroxisome proliferator-activated receptor Gamma Coactivator 1-α (PGC1-α) expression induced by P4 could lead to an increase in mitochondria biogenesis allowing an increase in P4 synthesis itself [40].

We also found morphological changes in nuclei and ER during VCT differentiation into ST. In ST, we observed that the nuclei present a denser chromatin compared to VCT, confirming previous placental histological findings [36]. The ER is also thinner in VCT than in ST cells. The presence of large ER in ST is associated with an intense protein synthesis [41,42], consistent with the observed increase in hCG secretion. It is also in agreement with the UPR activation during VCT differentiation into ST [17]. To conclude, the use of transmission electronic microscopy allowed us to fully characterize the differentiation of VCT into ST at cellular level, pointing out a membrane fusion, a nuclear high-condensed chromatin, small and dense mitochondria and large ER. This is in agreement with in vivo findings showing a release of syncytial knots containing ST DNA and UPR activation and an increased hormonal production.

We tested the impact of PI, notably the LPV, at a concentration of 10 µM, which corresponds to maternal blood concentration. Higher concentrations are likely to induce apoptosis as observed in HEK293 cells [43]. LPV at 10 µM induced an early decrease in P4 production in VCT that could be in part responsible for the decrease in P4 level in treated mother serum [3,6]. However, it would be of interest to investigate the effect of LPV/RTV as RTV is often associated to LPV as a booster [3]. Surprisingly, the decrease in P4 secretion under LPV was not associated with a decrease in expression of P450SCC and HSD3B1 enzymes in VCT. LPV could act rather indirectly by modulating P4 catabolism or transport. Indeed, LPV could activate Cytochrome P450 (CYP) family enzymes [44] and could modulate Adenosine-Tri-Phosphate (ATP)-Binding Cassette transporter activity [45]. Moreover, in BeWo cells, Papp et al. have demonstrated that PI could act on 20-α Hydroxysteroid Dehydrogenase, an enzyme involved in progesterone catabolism [46]. Consequently, in short incubation times, PI could decrease progesterone secretion in inhibiting enzymes involved in its synthesis and in activating enzymes involved in its catabolism. In ST, after 48 h of incubation with LPV, we also noted a discrepancy between the decreased expression of P450SCC and HSD3B1 enzymes, and the level of P4 secretion. The decrease in enzyme expression could be due to a global alteration in protein synthesis involving the eucaryotic Elongation Factor-2 (eEF2) translation factor. Indeed, it has previously been shown in the myocyte that LPV induces a decrease in protein synthesis [47]. The absence of P4 decrease after 48 h of incubation with lopinavir, while P450SCC and HSD3B1 expression is reduced, could be explained by the high stability of steroid hormones in our media culture. Consequently, we are not able to detect significant changes in P4 concentration after 48 h of incubation. So, it would be interesting to measure P4 concentration at shorter time or by changing our media culture.

LPV did not only induce an early decrease in P4 but also in hCG secretion in VCT. However, the decrease in hCG also persisted in ST after 48 h of incubation. This could be due to the fact that LPV affects rather ER and protein pathway than mitochondria and steroidogenesis. During ST formation, dynamic changes in some mitochondria could overlap the initial effect of LPV observed on VCT. On the contrary, as far as hormone production is concerned, we observed that LPV leads to a decrease in plasma membrane fusion as demonstrated by the decrease in fusion index associated with a decreased nuclei chromatin condensation and a thinner ER. Consequently, the 48 h incubation with LPV may prevent the morphological changes physiologically associated with functional differentiation. However, it is possible that the early decrease in P4 and hCG secretion induced by LPV from 6 h of incubation in VCT is responsible for the reduced morphological differentiation as both hormones are involved in autocrine an paracrine regulation of the trophoblast differentiation. As a matter of fact, hCG receptors are more expressed in VCT cells than in ST [13] and could be regulated by P4 [48] in order to control plasma membrane fusion [14]. Consequently, the decrease in P4 observed in VCT cells could impair hCG receptor expression. The decrease in hCG receptors expression and hCG secretion would lead to a decrease in plasma membrane fusion, preventing the global differentiation of VCT into ST.

Several intracellular mechanisms are involved in trophoblast differentiation [13,14] such as mitochondria dynamics and UPR pathways [16,17]. It is known that PI, notably LPV, alter both mitochondria DNA content and dynamics [49]. We analyzed the protein expression of two mitochondrial pro-fusion factors Mfn2 and OPA1. We demonstrated that LPV induced a decrease in Mfn2 expression after 48 h of incubation while OPA1 expression remained unchanged. The decrease in at least one fusion factor could result in a decrease in fusion process. These results are consistent with previous studies demonstrating that mitochondrial fusion is necessary for steroidogenesis [30,39]. However, they are inconsistent with the observed P4 secretion in ST after 48 h of incubation. This discrepancy could be explained by a global decrease in protein synthesis including Mfn2. It could be induced by a global cellular stress targeting rather the ER and the proteins than the mitochondria and the steroidogenesis [47].

Marie Cohen et al. also previously demonstrated that trophoblast differentiation and hCG secretion are under the control of UPR pathway [17]. Interestingly, several studies showed an activation of UPR, especially the IRE1α pathways under LPV exposition [43,50–53]. In collaboration with Dr Marie Cohen, we firstly demonstrated that UPR pathway regulates the expression of enzymes involved in P4 synthesis in placenta. We established in our model that LPV induces an early increase in sXBP1 and GRP78 expression in VCT pointing out an activation of IRE1α pathway in agreement with previous studies [43,50–53]. The activation of IRE1α is not maintained in ST, where GRP78 expression is on the contrary decreased. All these findings under LPV exposition, i.e., UPR activation, and decrease in GRP78, P450SCC, HSD3B1 and Mfn2, are likely to reflect an uncontrolled global cellular stress. Physiologically, UPR activation allows the maintenance of cellular homeostasis in case of stress. During a stress which cannot be controlled by UPR pathways activation, other processes take place, including particularly a general decrease in protein synthesis [32]. Thus, LPV could induce a stress activating IRE1α pathways. This stress could become uncontrollable after 48 h incubation with LPV ending in a global cell alteration and decreased protein synthesis as previously demonstrated in myocytes cells [47]. In summary, LPV have dual effects, an early effect (after 6 h of incubation) on enzyme and transporter activities to reduce P4 secretion and a later effect (after 48 h of incubation) on protein synthesis due to an ER stress.

In comparison to other PI, we demonstrated that RTV has no effect either on the ST formation or on hCG and P4 secretion in vitro, whatever the used concentration.

LPV exposition could not only disrupt VCT and its functional differentiation into ST, but it could also directly damage the ST covering the villi in direct contact with maternal blood and thus with PI. Indeed, on preexisting ST, we showed that LPV induces a decrease in both hCG and P4 secretion. In ST, on the contrary of VCT, this decreased secretion is associated with a decrease in P450SCC and HSD3B1 expression. LPV induces an increase in Mfn2 expression in ST. Consequently, the mitochondria in ST are more fused under LPV exposition. The increase in mitochondrial fusion induced by LPV could explain the decrease in P4 secretion by ST. We also investigated the effect on UPR pathways. In the preexisting ST, LPV induces an increase in sXBP1 expression demonstrating the activation of IRE1α pathway. To investigate whether this activation is responsible for the disruption in hCG and P4 secretion, we used an IRE1α inhibitor. We showed that the IRE1α inhibitor does not prevent either the decrease in hCG and P4 secretion or the decrease in P450SCC and HSD3B1 expression induced by LPV. The absence of effect of the IRE1α inhibitor may be explained by compensatory mechanisms of other UPR pathways. Indeed, the use of chemical UPR inhibitors is known to lead to compensatory mechanisms via the activation of the other UPR pathways [54]. It will thus be of interest to perform additional experiments combining different chemical inhibitors or siRNA targeting the 3 UPR pathways. Further investigation would also be necessary to check whether LPV directly activates UPR pathways or not. As a matter of fact, in HEK293 cells, Taura et al. [43] demonstrated that LPV is able to produce Reactive Oxygen Species (ROS), leading to an

activation of JUNK kinase pathway, further activating UPR pathways. Consequently, ROS production could be involved in the indirect effect of LPV on UPR activation in trophoblast.

The treatment of HIV infection uses a combination of two NRTI combined with a PI. In our study, we have demonstrated for the first time that RTV has no significant effects on endocrine function in trophoblast cells. On the contrary, LPV induces alteration of trophoblast morphological differentiation related to a decrease in plasma membrane fusion and in nuclei chromatin condensation, larger mitochondria and a thinner ER. These morphological alterations are associated with a decrease in P4 and hCG secretions. LPV also impairs mitochondrial dynamics as attested by an increase in Mfn2 expression and an induction of ER stress. ER stress leads to IRE1α activation marked by an increase in sXBP1 and GRP78 expression. These findings give the beginning of an explanation for the PI in vivo toxicity as they are known to induce preterm birth especially when they are "boosted" with RTV.
