TLR4-Endothelin Axis Controls Syncytiotrophoblast Motility and Confers Fetal Protection in Placental Malaria

ABSTRACT

Pregnancy-associated malaria is often associated with adverse pregnancy outcomes. Placental circulatory impairments are an intriguing and unsolved component of malaria pathophysiology. Here, we uncovered a Toll-like receptor 4 (TLR4)-TRIF-endothelin axis that controls trophoblast motility and is linked to fetal protection during Plasmodium infection. In a cohort of 401 pregnancies from northern Brazil, we found that infection during pregnancy reduced expression of endothelin receptor B in syncytiotrophoblasts, while endothelin expression was only affected during acute infection. We further show that quantitative expression of placental endothelin and endothelin receptor B proteins are differentially controlled by maternal and fetal TLR4 alleles. Using murine malaria models, we identified placental autonomous responses to malaria infection mediated by fetally encoded TLR4 that not only controlled placental endothelin gene expression but also correlated with fetal viability protection. In vitro assays showed that control of endothelin expression in fetal syncytiotrophoblasts exposed to Plasmodium-infected erythrocytes was dependent on TLR4 via the TRIF pathway but not MyD88 signaling. Time-lapse microscopy in syncytiotrophoblast primary cultures and cell invasion assays demonstrated that ablation of TLR4 or endothelin receptor blockade abrogates trophoblast collective motility and cell migration responses to infected erythrocytes. These results cohesively substantiate the hypothesis that fetal innate immune sensing, namely, the TRL4-TRIF pathway, exerts a fetal protective role during malaria infection by mediating syncytiotrophoblast vasoregulatory responses that counteract placental insufficiency.

INTRODUCTION

Up to 120 million pregnancies occur in regions where women are exposed to stable or low Plasmodium transmission rates (1), and mounting evidence has revealed that both Plasmodium falciparum and Plasmodium vivax parasites significantly impact outcomes of pregnancy (2). Pregnancy renders women more susceptible to malaria infection (3), and severe manifestations include increased parasite density, anemia, premature delivery, intrauterine growth restriction, stillbirth, and perinatal death (2, 4). These outcomes may occur several weeks after clearance of the parasite (5), demonstrating that treatment of the infection itself does not reverse the underlying pathological conditions.

During infections occurring after the first trimester, when maternal placental circulation has been fully established (6), placental tissue of fetal origin is perilously exposed to circulating infected erythrocytes (IEs) as maternal proinflammatory factors engage in antiparasite responses (7). A key event in the pathogenesis of placental malaria (PM) in P. falciparum infections is the sequestration of IEs expressing the parasite antigen, VAR2CSA. These bind to chondroitin sulfate proteoglycans (8) in the extracellular matrix of placental villous syncytiotrophoblasts. Recent studies have shown that P. vivax IEs also possess the ability to adhere to placental tissue (9) and provoke disturbances in placental physiology (10), suggesting that additional IE-trophoblast interactions are involved in placental responses. The accumulation of IEs in the placenta typically induces an intense infiltration of maternal inflammatory cells into the placental intervillous spaces and significant increases in placental levels of proinflammatory cytokines, including tumor necrosis factor alpha (TNF-α) and interferon gamma (IFN-γ) (7, 11). Intervillositis is associated with alterations of nutrient transport pathways (12), and also with morphological alterations of placental villi (13) that likely affect placental function and fetal growth. However, a refined description of the mechanisms leading to placental insufficiency is still needed.

Placentas from women infected with malaria show several pathological conditions strongly resembling those observed in preeclampsia, including reduced placental perfusion (14) and impaired trophoblast invasion (15), resulting in similar outcomes, including low birth weight and increased rates of preterm delivery (16). These similarities extend to soluble receptors for angiogenic and vasoregulatory pathways, which are dysregulated in the placenta during preeclampsia (17) and have an adverse effect on the outcomes of pregnancy, most likely due to their role in regulating placental perfusion (18). Experimental and epidemiological evidence supports the notion that angiogenic and vasoactive pathways are altered in PM (19, 20).

Recently, a prospective study on a cohort of women with PM revealed that reduced levels of l-arginine, a precursor to the potent vasodilator nitric oxide, were strongly correlated with poorer pregnancy outcomes (21). Further, in a murine model of PM, it was demonstrated that dietary l-arginine supplementation improved fetal viability and weight (21), implying that interventions on vasoregulatory systems may help to prevent and recover placental dysfunction. Impaired placental perfusion in murine PM can also be inferred from observations of reduced blood sinusoid spaces (22), altered microcirculatory dynamics (23), impaired fetal-placental angiogenesis (24), dysregulated expression of angiopoietins (25), reduced bradykinin receptor B2 expression, and reduced endothelial nitric oxide synthase expression (26). Initial evidence on the impact of innate responses upon placental angiogenic responses in PM was provided by the improved branching of placental vasculature during infection seen in mice with a deletion of complement 5a receptor (24). Nonetheless, mechanisms by which malaria leads to impairments of vasoregulatory systems in the placenta have been less explored.

Innate sensing of pathogen- and damage-associated molecular patterns mediated by Toll-like receptors (TLRs) plays critical roles in the inflammatory response to malaria parasites (27) and have been repeatedly highlighted in malaria genetic association studies (28). In mice, the major TLR adaptor protein, MyD88, and TLR4 take part in host-pathogen interactions, both in the fetal and maternal compartments and impact the outcomes of PM (29–32). Genetic crosses showed that fetal TLR4 opposes the deleterious effects of maternal TLR4 and protects fetal viability and takes part in interactions between IEs and fetal trophoblast in vitro (32). We have hypothesized that the conflict of fetal and maternal innate responses is a determinant of placental insufficiency during malaria infection (33). Specifically, we propose that placental trophoblasts, which are fetally derived but in direct contact with maternal blood, use TLRs to sense infection and mediate protective responses by modulating vasoregulatory pathways that regulate placental perfusion. TLR4 signaling has been linked to expression of several vasoactivators, including endothelin (34), which is produced by fetal-placental cells (35) and is increased in the placentas of mice with PM (36). Here, we investigated the action of the TLR4-endothelin axis in PM and its role in governing fetal trophoblast effector functions and in protecting the fetus.

RESULTS

Placental infection reduces birth weight and expression of EDN and EDNRB.

We examined the placental TLR4-endothelin (EDN) axis in a sample collection from a region endemic for both P. vivax and P. falciparum in Acre, Brazil (37), yielding a final data set of 401 pregnancies (Fig. S1 in the supplemental material). We noted that, in female newborns, birth weight and placental weight were reduced as a result of malaria during pregnancy, an effect more pronounced in P. vivax infections (Fig. 1A and B). These influences of maternal infection were not apparent in male newborns (Fig. S2A and B), suggesting that sex-linked factors may contribute to response to infection. We conducted a semiautomized quantitative immunohistochemistry analysis in placental sections and found that the area positively stained for endothelin receptor B (EDNRB) was reduced among P. vivax-infected placentas (Fig. 1C), a change which, once again, was not observed in pregnancies with male newborns (Fig. S2C). No significant differences in EDN expression were detected in infected placentas compared to uninfected placentas (Fig. 1D and Fig. S2D). Nevertheless, when subjects were stratified by infection history, we found that a striking correlation between EDN and EDNRB expression in uninfected placentas was maintained in past and chronic infections, irrespective of sex and infecting parasite, but was abolished in placentas with active, acute infection (Fig. 1E to I). Likewise, active infection increased the risk of low birth weight (Fig. 1J). Furthermore, qualitative analysis of all placentas, including male and P. falciparum infected, revealed an increase in the number of placentas which were negative for EDN and EDNRB (Fig. S3). Multivariate analysis of variance (MANOVA) on all samples revealed that infection during pregnancy, irrespective of PM status, significantly influenced birth weight (P = 0.013) and EDNRB-positive area (P = 0.039). Further analysis using multivariate multiple regression revealed that EDN was reduced in placentas with acute, active PM (P = 0.023). These results strongly suggest that, in addition to negatively impacting birth weight, placental malaria causes deregulation in expression of the EDN/EDNRB system.

FIG 1.

Malaria infection during pregnancy reduces birth weight, placental weight, and placental expression of endothelin (EDN) and endothelin receptor B (EDNRB) proteins. Data from 199 female newborns (A) and their placentas (B, C, and D) are plotted according to maternal infection status during pregnancy. NI, uninfected (n = 87); I, infected (n = 112); PV, with P. vivax infection (n = 58); PF, with P. falciparum infection (n = 29); M, with mixed P. vivax/P. falciparum infection (n = 25). (C and D) Percentage of the area of the tissue which stained positively for EDN and EDNRB in immunohistochemistry (IHC)-stained placental sections was quantified using QuPath software. For panels A to D, ***, P < 0.001; **, P < 0.01; *, P < 0.05; ns, nonsignificant (adjusted P value of Dunnett’s multiple-comparisons test). (E to H) Correlation of EDN and EDNRB measurements in women who did not have a malaria infection during pregnancy (n = 150) (E), who were infected but had no evidence of PM at the time of delivery (n = 147) (F), who had signs of previous PM (n = 60) (G), and who had active, chronic PM at the time of delivery (n = 10) (H). (I) Correlation was not found in women showing acute PM infection (n = 16). (E to I) Samples were classified according to maternal infection and histopathology criteria as described in Fig. S1 in the supplemental material. P values and Pearson’s correlation coefficient are shown. (J) Ratio of odds of low birth weight in the presence of maternal or placental infection with 95% confidence interval (CI) (acute, P = 0.011; chronic, P = 0.017).

TLR4 gene variants control placental expression of EDN and EDNRB.

Next, we asked whether TLR4 controlled EDN and EDNRB expression in placentas of female newborns and searched for genetic association using 17 single-nucleotide polymorphisms (SNPs) covering the TLR4 gene region. Association of maternal and fetal TLR4 alleles was addressed in separate analyses. EDN expression was quantitatively controlled by several TLR4 SNPs of maternal origin (e.g., RS11536878; P = 4.6 × 10⁻⁴) but not by fetal-derived TLR4 SNPs (Fig. 2). On the other hand, EDNRB expression was reduced by maternal alleles (e.g., RS11536879; P = 7.8 × 10⁻⁴) but was increased by the fetally derived RS4986790 alleles (P = 9.4 × 10⁻⁴). This RS4986790 allele is a missense mutation (Asp299Gly) resulting in a TLR4 hypomorph (38) and maps in an LD block distinct from the other mentioned associated SNPs (Fig. S4), suggesting it represents an independent association signal. Interestingly, TLR4 maternal alleles controlled placental expression of EDN and EDNRB irrespective of infection (Fig. 2B and C). These results indicate that EDN/EDNRB expression in the placenta is controlled by maternal and fetal TLR4 genotypes and suggest that functional fetally derived TLR4 alleles (e.g., RS4986790 AA genotype) are associated with downregulation of endothelin vasoregulatory system in the placenta (Fig. 2D).

FIG 2.

Endothelin (EDN) and endothelin receptor B (EDNRB) protein expression in placentas of female newborns is under TLR4 genetic control. (A) Diagram with relative positioning of 17 SNPs tested in the TLR4 gene. The table shows results of quantitative trait locus analysis testing association of fetal or maternal minor alleles with area of staining of EDN or EDNRB proteins in placental IHC sections of 199 female newborns. MAF, minor allele frequency; P-emp, empirical P value (adjusted with 1 × 10⁶ label permutations). Significant association results are highlighted in bold. (B to D) Genetic effects of maternal alleles at RS11536878 (B) and RS11536879 (C) and RS4986790 (D) fetal alleles (not tested in infected placentas due to low allele frequency). Mann-Whitney P value (D) and Dunn’s posttest P value (B and C) for individual genotype class comparisons.

Autonomous placental TLR4 protects fetal viability in experimental placental malaria.

We further investigated the TLR4-endothelin axis using a mouse model of acute PM (39) that revealed TLR4-mediated fetal protection effects (32). We crossed Tlr₄^+/− males with Tlr4^−/− females to yield heterogenic litters with sibling feto-placental units that expressed or did not express TLR4 and shared the same maternal environment, enabling the investigation of placental autonomous effects of fetal TLR4 (Fig. 3A). While the absence of maternal TLR4 in these crosses did not significantly affect peripheral parasitemia (Fig. 3B), these litters showed a reduction in premature delivery or loss of the litter prior to G₁₈ (Fig. 3C). The impact of infection on reducing fetal viability was observed in wild-type mice, as expected (Fig. 3D). This impact was notably reduced in heterogenic litters carried by Tlr4^−/− females, supporting a deleterious role for maternal TLR4 (Fig. 3E) (39). Furthermore, we found that litter viability rate positively correlated with the proportion of fetuses expressing TLR4 (Fig. 3F). To discern the effects of infection and fetal TLR4 on viability rate, we used a generalized linear model incorporating infection and fetal genotype as fixed effects and the litter size, genotype proportion, and the effects from individual pregnancies as random effects. We found that fetal Tlr4 significantly increased viability (P = 0.037), whereas maternal infection reduced it (P = 0.049), demonstrating a beneficial effect of fetal TLR4.

FIG 3.

Placental TLR4 protect fetal viability during placental malaria. (A) Breeding scheme showing Tlr4^−/− females crossed with Tlr4^+/− males to generate heterogenic pregnancies with Tlr₄^+/− and Tlr4^−/− fetuses. Females were infected on the 13th day of gestation, and fetuses and placentas were collected on the 18th day. (B) Maternal TLR4 ablation did not impact on peripheral parasitemia at G₁₈. (C) Premature delivery (prior to G₁₈) in uninfected and infected Tlr₄^+/+ and Tlr4^−/− females (χ² tests; *, P < 0.05; **, P < 0.01). (D) Survival rate of Tlr4^+/+ fetuses in uninfected and infected mothers. Generalized linear mixed-effects model, with infection as fixed effects and litter-wise random effects for fetuses from the same litter; P (infection) < 0.001 (E) Survival rate of Tlr4^+/− fetuses is compared to Tlr4^−/− counterparts. Generalized linear mixed-effects model, with infection and fetal Tlr4 genotype as fixed effects and litter-wise random effects for fetuses from the same litter; P (Tlr₄^+/−) = 0.037; P (infection) = 0.049. (F) Correlation of litter viability rate with proportion of Tlr₄^+/− fetuses in litter (each data point represents one heterogenic litter; n = 10).

In contrast, we found that low fetal weight imposed by infection correlates with maternal parasite burden irrespective of maternal or fetal TLR4 genotype and is not influenced by the proportion of fetuses expressing TLR4 in the litter (Fig. S5A to E). Using linear modeling, we were able to confirm that infection during pregnancy reduces fetal weight at G₁₈ (P = 0.035) independently of fetal TLR4 (P = 0.63), indicating that distinct mechanisms control the impact of PM on fetal viability and fetal weight. These results corroborate our previous finding that fetal TLR4 acts to protect fetal viability during PM (32) and further indicate that this effect operates at the level of the feto-placental unit in an autonomous manner with respect to the maternal environment.

Malaria infection modulates placental Edn1 gene expression in a TLR4-dependent manner.

We analyzed the RNA expression of the endothelin gene (Edn1) and its receptors (Ednra and Ednrb) comparing placentas differing in TLR4 genotype (Tlr₄^+/− versus Tlr4^−/−) from heterogenic litters. We used a discordant sibling pair approach which compared a pair of placentas that differed in the TLR4 genotype (Tlr₄^+/− versus Tlr4^−/−) from each litter. The expression of Edn1 was not altered by infection in Tlr4^−/− placentas but in Tlr₄^+/− placentas infection provoked a significant increase in gene expression (Fig. 4A and B). In complement, sibling pairwise analysis showed that, in infected mothers, the expression of Edn1 was increased in Tlr₄^+/− placentas compared to their Tlr4^−/− siblings (Fig. 4E and F). To further explore alterations of the placental endothelin pathway, we analyzed the gene expression of Ednra and Ednrb. As expected, we did not detect expression of Ednra in the placenta (40). Expression of Ednrb was mildly increased during infection in Tlr₄^+/− placentas (Fig. 4C, D, G, and H). Immunohistochemistry staining of both EDN and its receptor, EDNRB, showed diffuse protein expression throughout the labyrinth region, hampering delimitation of stained cells and quantification by imaging (Fig. 4I and J). Importantly, placental parasite load quantification did not show significant differences between TLR4-discordant siblings (Fig. 4K). These results indicate that placental tissue responses to infection include enhancement of Edn1 gene expression that is favored by fetally encoded TLR4 but did not attribute this effect to specific placenta cell types.

FIG 4.

Fetal TLR4 modulates placental endothelin 1 (Edn1) gene expression during malaria infection. Placental expression of Edn1 and Ednrb using one Tlr4-discordant sibling pair from each heterogenic litter. (A to D) Grouping by fetal genotype to evaluate the effect of infection represented by fold change relative to the average of noninfected placentas. (E to H) Grouping by infection status to evaluate the effect of placental Tlr4 genotype represented by fold change relative to the average of Tlr4^−/− (pairwise analysis). (I and J) Images of EDN (I) and EDNRB (J) immunohistochemistry staining showing the pattern of protein expression throughout the placental labyrinth. Scale bar, 100 μm. (K) Parasite load in Tlr₄^+/− and Tlr4^−/− placentas (pairwise analysis) from infected females. (A to D) Wilcoxon rank-sum (Mann-Whitney) tests. n = 6 per group; **, P < 0.01. (E to H and K) Wilcoxon matched-pairs signed-rank tests, n (noninfected) = 8 sibling pairs, n (infected) = 10 sibling pairs; **, P < 0.01.

TLR4-TRIF pathway mediates downregulation of trophoblast Edn1 expression in response to infected erythrocytes.

We tested whether control of placental endothelin expression by TLR4 could result from interactions of infected erythrocyte (IE) with trophoblast using a coculture assay. Murine primary wild-type trophoblasts collected on the G₁₈ cultured in subconfluent density and exposed to IE for 4 h revealed reduced Edn1 RNA expression. Interestingly, this effect was absent in trophoblasts isolated from Tlr4^−/− and Trif^−/− placentas but not in Myd88^−/− trophoblasts (Fig. 5A to D). These results indicate that downregulation of Edn1 induced by IE operates in trophoblasts through a TLR4-TRIF pathway, suggesting that this TLR4 signaling pathway is required in modulation of placental Edn1 expression provoked by infection.

FIG 5.

Edn1 expression is downregulated in vitro upon trophoblast-IE interactions in a TLR4- and TRIF-dependent manner. (A to D) Edn1 expression in primary murine trophoblasts from wild-type (A), Tlr4^−/− (B), Myd88^−/− (C), and Trif^−/− (D) mice exposed to infected erythrocytes (IEs) for 4 h. Data represented as fold change relative to exposure to noninfected erythrocytes (NIEs). Ednrb expression in wild-type (E) or Tlr4^−/− (F) primary murine trophoblasts exposed to IEs or NIEs for 4 h. Each data point represents the average value of experimental triplicates. Mann-Whitney tests; *, P < 0.05; **, P < 0.01. Plots representative of multiple experiments.

In contrast to our previous observations, Ednrb expression in trophoblasts was not altered upon exposure to IEs (Fig. 5E and F). Likewise, gene expression of Nos3 coding for endothelial NOS, Kdr coding for VEGFR2 and Bdkrb2 coding for bradykinin receptor B2 that have been previously implicated in PM were not altered by trophoblast exposure to IE and/or ablation of TLR4 (Fig. S6A to F). Furthermore, we used the IE-trophoblast interaction assay to analyze the expression of two nutrient transporters which have been implicated in PM, namely, the glucose transporter Slc2a1 (12) and the amino acid transporter Slc38a1 (41). Slc2a1 exhibited increased expression in trophoblasts exposed to IEs irrespective of their Tlr4 genotype (Fig. S6G and H), while expression of Slc38a1 was unchanged (Fig. S6I and J), showing that TLR4 is not involved in genetic control of these nutrient transporters. Together, these results support the possibility that TLR4 acts in trophoblasts to modulate vasoactivation functions via the EDN/EDNRB system without affecting other vasoactivation mechanisms involved in PM or influencing impairments of nutrient pathways caused by IE.

Trophoblast motility is controlled by the TLR4-endothelin axis.

We have previously demonstrated that in murine PM trophoblast, conformational changes impact on microcirculatory dynamics of maternal blood spaces (23), possibly impairing local placental perfusion. This led us to hypothesize that trophoblast conformational changes are controlled by vasoregulatory mechanisms and are responsive to interactions with IEs. To determine if trophoblast motion was affected by exposure to IEs, we developed a time-lapse microscopy imaging assay that uses fluorescent protein-labeled trophoblasts obtained from placentas of C57BL/6 females crossed with C57BL/6-GFP or C57BL/6-CFP males. We quantified trophoblast conformational activity in the presence of IEs by measuring the change between consecutive imaging frames (Movies S1 and S2) represented by the integration of image intensity change (ΔRID) in 9 microscopic fields of one cell culture well. We found that exposure to IEs consistently and significantly decreased trophoblast activity (representative experiment shown in Fig. 6A). Quantification of total activity throughout the imaging acquisition period (4 h) in five independent experiments shows that exposure to Plasmodium berghei IEs caused a significant reduction in the dynamics of trophoblast activity (Fig. 6B). These data reveal an intrinsic conformation activity of primary murine trophoblasts which is reduced upon exposure to IEs. To further examine the impact of infection on trophoblast motility, we evaluated trophoblast migration across an extracellular matrix (ECM) gel cushion (Fig. 6C). This assay revealed that 48 h exposure to IEs significantly reduces trophoblast migration (Fig. 6D) in a Tlr4-dependent manner (Fig. 6E). Importantly, we found that inhibiting endothelin signaling with bosentan, an endothelin receptor antagonist, also led to reduced trophoblast migration (Fig. 6F). These results indicate that IEs reduce the dynamics of trophoblast motility in a TLR4-dependent manner and strongly suggest that endothelin signaling is a critical mechanism in modulating trophoblast topological changes.

FIG 6.

Exposure to infected erythrocytes reduces trophoblast motile activity in a TLR4-dependent manner. Time lapse imaging (Movies S1 and S2 in the supplemental material) of primary trophoblasts was carried out in 9 adjacent microscope fields over a 4-h period (10-min intervals), and overall trophoblast conformation activity was measured between consecutive frames (ΔRID). (A) Representative plot of the ΔRID data obtained from one experiment, showing the ARID values of three replicates exposed to infected erythrocytes (IEs) compared to noninfected erythrocytes (NIEs), indicating reduced collective motility. (B) Area under the curve (AUC), representing the total amount of movement over the 4-h period, was calculated from the ΔRID data of each culture well across 5 independent experiments, and each point was then standardized to the average of the uninfected controls. Ratio-paired t test was applied to the mean AUCs of each experiment, **, P < 0.01. (C) Trophoblast migration assays were carried out with primary trophoblasts seeded on a ECM gel cushion in a Transwell. (D and E) Number of cells that crossed the layer was quantified and is reduced in wild-type trophoblasts which are exposed to IEs (D), but not in exposed Tlr4^−/− trophoblasts (E). (F) Addition of the endothelin receptor antagonist, bosentan, also reduces migration in primary wild-type trophoblasts. One-sample t tests; *, P < 0.05; **, P < 0.01.

DISCUSSION

Infection by the malaria parasite during pregnancy impairs organ function, resulting in an array of detrimental clinical outcomes for the developing fetus. Evidence collected in recent years has unveiled placental dysfunctions caused by Plasmodium infection, including alterations in the uptake of amino acids (41), glucose (12), other nutrients, and hormones (42). PM is associated with impaired placental perfusion (14, 16) and defective trophoblast invasion into the uterine decidua (15). Reportedly, TLRs (29–32), the complement system (24), and other innate immunity mediators (22) are implicated in the outcomes of acute murine PM. Furthermore, it has been shown that ablating the C5 complement receptor results in increased placental angiogenesis (24), which may act as a compensatory mechanism during infection (43). These data emphasize the role of inflammatory factors in placental angiogenesis, providing a rationale for how placental vasoregulatory pathways are controlled by innate immune responses.

Using TLR4 genetic variants, we analyzed the expression of placental EDN/EDNRB system in human and mouse placentas and uncovered that TLR4 controls modulation of EDN/EDNRB expression during active infection. However, TLR4 impacts differently in expression of endothelin and endothelin receptor B in infected placentas and in isolated trophoblasts. TLR4 signaling has been linked to increased production of endothelin in dendritic cells (34) but leads to decreased endothelin secretion in microvascular endothelial cells (44), suggesting differential regulation dependent on cell type or stimuli involved. Our in vitro results pinpoint trophoblast autonomous responses to TLR4 stimulation as a trigger for Edn1 gene downregulation. Specifically, we found that, in mouse trophoblast, these effects on Edn1 expression were mechanistically governed by the TLR4-TRIF signaling pathway recently described to control bacteria phagocytosis and signaling from phagosomes, leading to regulation of interferon beta (IFN-β) production (45). Coincidentally, we previously showed that IE uptake by trophoblasts is impaired in the absence of TLR4 (32). In concert, these data raise the possibility that endothelin modulation in trophoblasts is linked to TLR4-TRIF-mediated phagocytosis of parasite components. Nevertheless, it is plausible that other cell types and inflammatory factors use other triggering mechanisms and also contribute to the observed TLR4-dependent increase in Edn1 expression in the inflamed placenta.

On the other hand, mouse trophoblast TLR4 responses do not appear to control alterations in nutrient transport pathways during PM and do not impact pathogen burden or fetal weight (Fig. S5 and S6 in the supplemental material) (32, 46). We also found that risk of birth weight reduction was increased in active human infection, was only detectable in female newborns, and was not controlled by TLR4 genetic variants. This sex-related effect was also reported in an African cohort where female newborn gender increased the risk of low birth weight in late-gestation infections (47). These findings suggest that responses to infection that compromise fetal nourishment and growth at late stages of pregnancy are, at least in part, distinct from TLR4-mediated mechanisms that protect fetal viability and sustain pregnancy. These data corroborate that, in human PM, distinct pathophysiological mechanisms operate in various combinations, contributing to the wide range of observed clinical outcomes (2, 4).

Our findings support the notion that placental innate immunity sensing is molecularly wired with vasoregulatory mechanisms, a possible hallmark of placental responses to infection. In the placenta, endothelin is predominantly produced in the fetal-derived syncytiotrophoblast layer on the surface of chorionic villi (35), which also express endothelin receptor B (40) both in humans (Fig. S3) and in mice (Fig. 4J). This suggests that endothelin would act locally to influence circulation (48) or influence the synthesis of TNF-α in macrophages and monocytes (49), which are major constituents of the maternal inflammatory infiltrate found in the placenta during PM (4). Remarkably, we found that human PM impacted on EDN/EDNRB protein expression; however, these effects were observed in placentas of female but not male newborns, indicating that these responses are subjected to sex-related modifiers, an observation which has also been made in other inflammatory conditions affecting the placenta (50).

Using motility of trophoblasts in culture as a functional correlate of endothelin action, we found that interactions with IEs reduced trophoblast motility in a TLR4-dependent manner. This effect was mimicked by endothelin receptor inhibition (Fig. 6), indicating that endothelin signaling is required for inherent trophoblast motility and acts in an auto-paracrine fashion. Time-lapse microscopy revealed cell-autonomous trophoblast conformation activity with dramatic changes in trophoblast motility, supporting the compelling hypothesis that altered dynamics of trophoblast conformational changes in infected placenta may be an unappreciated component of microcirculatory regulation in the placenta. Furthermore, it has been established that trophoblast invasion is critical for placental perfusion, with impaired invasion contributing to preeclampsia and intrauterine growth restriction. Previous work has demonstrated that trophoblast invasion can be impaired by exposure to serum from Plasmodium-infected women (15), and our finding that primary trophoblast motility is impaired by exposure to Plasmodium-infected erythrocytes corroborates (Fig. 6) corroborates that trophoblasts at the placental barrier are equipped to detect maternal blood infections and can be critical in ensuing placental vasoactive responses.

The mouse model of acute PM has limitations, as the mice are always primigravid and immunologically naive to infection and develop high-grade parasitemia. In this context, it is plausible that severity of fetal outcomes represented by low viability rates, low birth weights, and high rates of premature delivery result from strong inflammatory maternal responses that determine severe placental insufficiency. Disease severity is much lower in human PM cases described in health services settings. The human sample collection analyzed here does not contain any severe PM cases entailing fetal loss; this is most likely due to clinical interventions. Nevertheless and in accordance with other studies (51), odds ratio calculations demonstrated that active PM occurring in later stages of pregnancy conferred a higher risk of reduced birth weight (Fig. 1).

This work interrogated the TLR4-endothelin axis in PM at multiples levels, including human disease, experimental infection, placental molecular pathology, and trophoblast assays. Our strategy of discerning maternal and fetal genetic effects in human infection and in mouse models allowed us to identify fetal placental responses that would otherwise be masked by maternal responses (32). Intriguingly, we found that opposing to maternal TLR4, fetal TLR4 expression in mouse placenta correlated with fetal viability protection in a placenta-autonomous manner (Fig. 3). Evidence obtained from other experimental models of pregnancy disorders has highlighted the contribution of maternal TLR4 responses in adverse outcomes of pregnancy caused by bacterial infections (46, 52), lipopolysaccharide exposure (53), and uterine ischemia (54). In humans, increased TLR4 activity on maternal monocytes is correlated with spontaneous preterm labor (55), and increased TLR4 expression in the maternal decidua has been linked to recurrent miscarriages (56). In contrast, decreased placental TLR4 activity has been implicated in miscarriages and preeclampsia (57). Furthermore, the fetal hyporesponsive TLR4 Asp299Gly polymorphism (38), which was associated with increased ENDRB expression in human placenta (Fig. 1), is associated with severe prematurity (58), corroborating the notion that maternal and fetal TLR4 responses may be acting in opposition to one another. Examples of these conflicting responses during infection are not restricted to malaria or TLR4. Placental CCl22 production is increased upon exposure to Toxoplasma gondii (59), whereas it is usually found in maternal decidua only during unhealthy pregnancies (60). It has been demonstrated that vascular endothelial growth factor (VEGF) levels are increased in maternal cells within the placenta during malaria, whereas fetal trophoblasts produce larger amounts of sFlt1, a soluble VEGF receptor which blocks VEGF action (19). Interestingly, polymorphisms of the sFlt1 gene have been shown to influence the outcomes of placental malaria and may, consequently, be under natural selection (61).

Results from our human cohort suggested that reductions in placental EDNRB are a consequence of malaria during pregnancy and that EDN expression is deregulated during active acute infection (Fig. 1). EDNRB is a driver of nitric oxide production (62), allowing the speculation that reduced expression of this receptor may be related to the observed reduction of nitric oxide levels during malaria in pregnancy (21). Overall, we propose that trophoblast TLR4 and endothelin-dependent alterations in motility may regulate placental blood flow in a manner beneficial to the fetus. Further avenues of this proposal which warrant investigation include unveiling the precise nature and impact of this modulation and unraveling the intracellular signaling mechanism linking TLR4 to Edn1 expression via TRIF, with a view to determining if it is also effective in other pregnancy-related conditions.

MATERIALS AND METHODS

Ethical permits.

The samples used for this study were collected as part of an earlier study which is registered as RBR-3yrqfq in the Brazilian Clinical Trials Registry and with Plataforma Brasil, CAAE 86696718.5.0000.5467, with more details provided by Dombrowski et al. (37). Animal housing and all procedures with live animals were in accordance with Portuguese national regulations on animal experimentation and welfare and were approved by both the national animal welfare authorities and the Instituto Gulbenkian de Ciência ethics committee.

Collection of human samples.

The human cohort study was originally conducted in the Amazonian region of the Alto do Juruá Valley in Acre, Brazil, between January 2013 and April 2015. Six hundred pregnant women with P. falciparum and/or P. vivax or uninfected pregnant women were enrolled by volunteer sampling and followed until delivery. The women were recruited during their first visit to the antenatal care clinic and monitored by a trained nurse, which included two additional domiciliary visits at the second and third trimesters to monitor their clinical state. As recommended by the Brazilian Ministry of Health, when it is not possible to use ultrasound, the gestational age was estimated using the last menstrual period method. At the time of recruitment, data were collected on several clinical and obstetric variables, and molecular diagnosis of maternal infection during pregnancy was performed by reverse transcriptase PCR (RT-PCR) with identification of parasite species. At the time of delivery, clinical data were collected from the mother and the newborn, birth weight and placental weight were registered, and a placental biopsy specimen and blood samples were also collected. The placental samples were fixed in 10% neutral buffered formalin and stored at 4°C before paraffin embedding. Histopathological diagnosis of PM was carried blindly by two trained experts, and, in case of disagreement, a third evaluation was performed in order to achieve a consensus (37). Placentas were classified according the Bulmer scale for PM (63). The final data set for phenotypic analysis included 401 pregnancies.

Immunohistochemistry analysis of human samples.

The tissue microarray technique, conducted at the AC Camargo Hospital (São Paulo, Brazil), was used to prepare slides from samples of human placental samples (20). Slides were deparaffinized and rehydrated before heat-induced antigen retrieval using citrate buffer, pH 6.0 (catalog no. CBB500; ScyTek Laboratories) in a water bath at 100°C for 20 min. Slides were cooled in phosphate-buffered saline (PBS) and then rinsed with PBS containing 0.05% Tween 20 (PBS-T). The slides were then incubated with blocking solution (2.5% bovine serum albumin [BSA] and FcBlock in PBS-T) for 1 h and 15 min and washed with PBS-T. The slides were incubated with the primary antibodies for endothelin (catalog no. sc-98727; Santa Cruz Biotechnology) at 3 μg/ml or endothelin receptor B (catalog no. pb-9554; Boster Biological Technology) at 1.5 μg/ml overnight. Following incubation, the slides were washed with PBS-T and incubated with 3% hydrogen peroxide in PBS for 15 min. Antibody detection was carried out with the Reveal polyvalent horseradish peroxidase (HRP)-DAB (3,3′diaminobenzidine) detection system (Spring Bioscience) following the manufacturer’s directions. Counterstaining was carried out with Harris hematoxylin and washing with saturated lithium carbonate solution and tap water. Slides were dehydrated before mounting with Tissue-Tek mounting medium (product code 6419; Sakura FineTek). A Hamamatsu NanoZoomer-SQ digital slide scanner and NDP.scan software (Hamamatsu) were used to digitalize the tissue section images, and then they were analyzed with NDP.view2 free software. Quantitative analysis to determine positively stained area as a fraction of total tissue area in each section was performed using QuPath, an open-source digital pathology and whole-slide analysis software. In parallel, tissue images were scored as positively or negatively stained by 2 examiners under blinded conditions.

Genotyping and SNP association analysis.

The peripheral and placental blood was collected in heparin tubes and then separated into plasma and whole blood cells using a centrifuge. Total DNA was obtained from whole blood cells using a commercially available extraction kit (QIAamp DNA minikit; Qiagen) following the manufacturer’s instructions. All samples were kept at −80°C. SNPs covering the TLR4 region were selected and genotyped in all samples using the Sequenom's iPlex assay (San Diego, USA) and the Sequenom MassArray K2 platform at the Genomics Unit of the Instituto Gulbenkian de Ciência. Extensive quality control was performed using eight HapMap controls of diverse ethnicity. Analyzed SNPs passed the exclusion criteria, with >90% call rate and Hardy-Weinberg equilibrium (HWE) with a P value of >0.05, with exception of RS4986790 and RS4986791, known to be low frequency. Samples with <90% call rate and duplicates were excluded from analysis. Genetic analysis was performed in samples from maternal and newborn samples from female births in a data set consisting of 199 mother-female child pairs. Quantitative trait locus analysis testing TLR4 minor-frequency alleles was performed using the PLINK package (64) implemented in the BCGene platform. Genetic effects were evaluated by Mann-Whitney or Dunn’s posttest for individual genotype class comparisons.

Murine pregnancy monitoring.

C57BL/6 (wild-type [WT]), C57BL/6.GFP, C57BL/6.CFP, Tlr4^−/−, Trif^−/−, and Myd88^−/− mice were obtained from the animal facility at Instituto Gulbenkian de Ciência. Tlr₄^+/− mice were generated by crossing Tlr4^−/− females with a C57BL/6 male. All mice were bred and maintained under specific-pathogen-free (SPF) and Helicobacter-free conditions, housed in individually ventilated cages, and females used for infection between 8 and 12 weeks of age. Mating schemes for evaluation of pregnancy outcomes, placenta collection, and preparation of trophoblast cultures included C57BL/6.Tlr₄^+/− or C57BL/6.Tlr4^−/− males with C57BL/6.Tlr4^−/− females, WT males with WT females, and C57BL/6.GFP or C57BL/6.CFP males with WT females. Females were examined for a vaginal plug the morning after being placed with a male, and if detected, this was deemed day 0.5 (G_0.5). In the absence of plug detection after an additional 24 h, females were removed, and this time point was considered gestational day G_0.5 in resulting pregnancies. Pregnancy was diagnosed by weighing the females. Successful gestation was confirmed at G₁₃ when females exhibited an increase of 3 to 4 g in body weight. Abrupt weight loss after this point was an indicator of unsuccessful pregnancy.

Parasites and infection.

The parasite lines Plasmodium berghei NK65, originally at New York University, and P. berghei ANKA mCherry were kindly provided by Maria Mota (Instituto de Medicina Molecular, Lisbon, Portugal). Infections for the expansion of parasite from frozen aliquots were performed by intraperitoneal (i.p.) injection of 10⁵ IEs. For other experiments, pregnant mice were intravenously (i.v.) injected with 10⁵ IEs. Peripheral blood samples were stained with DRAQ5 (BioStatus Limited) and assessed by flow cytometry. Samples were mixed by inversion to allow incorporation of DRAQ5 into the parasite DNA and were immediately analyzed, with parasitemia expressed as the percentage of stained cells within the erythrocyte morphological gate. In pregnant females, the time of infection (G₁₃), route (i.v.), and dose of IEs (106) are in accordance with the previously characterized model of pregnancy-associated malaria in C57BL/6 females infected with P. berghei NK65 parasite (32, 39).

Outcomes of pregnancy and fetal survival.

Infected pregnant mice were killed by CO₂ narcosis on the 18th day of gestation, and a cesarean section was carried out. Fetuses were extracted from their amniotic sac, and viability was evaluated by reaction to touch using forceps. A lack of prompt movement indicated that the fetus had recently died. Fetal resorptions were identified as small implants with no discernible fetus and placenta and recorded as nonviable. Viability rates in individual females were calculated, taking into account the total number of stillbirths and resorptions within each litter. All fetuses were weighed, and in the case of mixed Tlr₄^+/− and Tlr4^−/− litters, a sample of tissue was collected for genotyping.

Gene expression analysis of placentas.

Placentas from infected and noninfected females were collected on the 18th day of gestation and snap-frozen using liquid nitrogen. Samples were transferred to lysis buffer (RNeasy minikit; Qiagen) containing 1% β-mercaptoethanol for RNA extraction and homogenized with tungsten beads on a Qiagen TissueLyzer II. Total RNA was extracted from the lysate using an RNeasy minikit (Qiagen) according to the manufacturer’s instructions and quantified using a NanoDrop UV-visible (UV-Vis) spectrophotometer (Thermo Fisher Scientific). Transcriptor First Strand cDNA synthesis kit (Roche) was used to prepare cDNA from the RNA preparations prior to qPCR using an ABI Prism 7900HT system. P. berghei parasite loads were quantified using 18S RNA TaqMan assays with the following specific primers and probe: forward primer, 5′-CCG ATA ACG AAC GAG ATC TTA ACC T-3′; reverse primer, 5′-CGT CAA AAC CAA TCT CCC AAT AAA GG-3′; and probe, 5′-ACT CGC TAA TTA G-3′ (6-carboxyfluorescein [FAM]/MGB). Commercial primer/probe sets for qPCR analysis of mouse genes are listed in Table S1 in the supplemental material. The endogenous control gene Gapdh (mouse GAPD endogenous control, catalog no. 4352339E; ABI) was used in all PCR assays. ACT was calculated by subtracting the cycle threshold (CT) of the target gene from the Gapdh CT. Gene expression results are presented as fold change (2^−ΔΔCT) relative to the average of control values (65). Placental parasite burden is presented as P. berghei 18S RNA levels normalized to Gapdh levels and plotted as 2^−ΔΔCT.

Isolation of primary murine trophoblasts.

Pregnant C57BL/6 and C57BL/6.Tlr4^−/− females were sacrificed by carbon dioxide narcosis on the 18th day of gestation and the placentas retrieved by caesarian section. Placentas from each individual mother were pooled and cells dispersed in digestion medium containing 1 mg/ml collagenase type 1 (catalog no. C9891; Sigma), 20 μg/ml DNase I (catalog no. 11284932001; Roche), 10 mM HEPES, 350 μg/ml sodium bicarbonate, and RPMI 1640 medium (with stable glutamine) (Gibco) incubated at 37°C for 1 h (66). The digestion was then passed through a 70-μm cell strainer and centrifuged at 500 relative centrifugal force (RCF) for 5 min to pellet the cells. The cells were resuspended in 4 ml of 25% Percoll (catalog no. 17-0891-01; GE Healthcare) in RPMI medium and layered onto 4 ml of 40% Percoll with a 2-ml layer of PBS. The density gradient was centrifuged at 800 RCF for 20 min with no brake, and the interface between the Percoll layers, containing trophoblasts, was collected. The cells were washed and resuspended in 1 to 2 ml of RPMI complete medium (RPMI with stable glutamine, 10% fetal bovine serum, 1× HEPES, 1 mM sodium pyruvate, 1× antibiotic/antimycotic, and 2-mercaptoethanol), counted, plated, and incubated at 37°C for 7 to 10 days. Purity of the trophoblast cultures was assessed by expression of the trophoblast marker cytokeratin 7 detected in fluorescence-activated cell sorter (FACS) analysis. A sample of cells was taken from the cells prior to plating and again before experimental use. The cells were washed with FACS buffer and processed for intracellular staining with anti-cytokeratin 7 antibodies (RCK105; catalog no. sc-23876; Santa Cruz Biotechnology Inc.). Preparations of cultured cells used in experiments were >85% KRT7 positive (KRT7⁺) (26, 32).

Isolation and synchronization of infected erythrocytes.

P. berghei NK65 was used for all experiments except imaging, where P. berghei ANKA mCherry was used; the process to isolate IEs was the same. We intraperitoneally injected 2 × 10⁶ IEs from frozen stocks into C57BL/6 females. When parasitemia reached between 2 and 5%, the mice were sacrificed by carbon dioxide narcosis, and their blood was collected by cardiac puncture into a syringe containing a small amount of heparin to prevent clotting. The blood was resuspended in 10 ml synchronization medium (RPMI containing 20% fetal bovine serum [FBS] and 0.01 mg/ml neomycin) and centrifuged for 5 min at 800 RCF. After discarding the supernatant, the erythrocytes resuspended in synchronization medium and incubated for 18 h in 75-cm² tissue culture flasks in a total volume of 50 ml. Schizont-stage IEs were isolated using magnetically activated cell sorting (MACS) 25LS columns (Miltenyi Biotec) according to the manufacturer’s instructions and resuspended in RPMI complete medium.

Gene expression analysis of primary trophoblasts exposed to erythrocytes.

Trophoblasts plated at a concentration of 10⁶ cells/well in 96-well plates with RPMI complete medium had their medium removed and were briefly rinsed with PBS prior to the addition of 10⁶ IEs in fresh medium RPMI complete. Noninfected erythrocytes (NIEs) at the same concentration were used for controls. cDNA from trophoblast cultures was prepared using a Cells-to-Ct kit (Invitrogen) according to the manufacturer’s instructions, and quantitative PCR (qPCR) was executed in the same manner as for placental samples. Commercial primer/probe sets for qPCR analysis of mouse genes are listed in Table S1.

Time-lapse imaging of primary trophoblasts.

Trophoblasts expressing green fluorescent protein (GFP) or cyan fluorescent protein (CFP) were isolated as described above and placed onto 8-well Lab-Tek chambered cover glass slides (catalog no. 155411; Thermo Fisher Scientific) at a concentration of 175,000 cells/well and maintained with RPMI complete for 7 days. Prior to imaging, medium was replaced with RPMI complete without phenol red, which contained either P. berghei ANKA mCherry IESoR NIEs. The final ratio of erythrocytes to trophoblasts was 1:1. Time-lapse images were acquired on an Applied Precision DeltaVision Core system, mounted on an Olympus inverted microscope, and equipped with a Cascade II 2014 electron-multiplying charge-coupled device (EM-CCD) camera, using the 10× 0.30 numerical aperture (NA) objective and mCherry plus CFP fluorescence filter sets. The cells were maintained at 37°C and 5% CO₂ during the imaging process. Images were acquired at 10-min intervals over a 4-h period using a tile-scan method, with 9 fields acquired in each well. Consecutive frames were superimposed, and the absolute difference calculated using the difference operator (in the image calculator). Raw integrated density was calculated from the resulting frames. We refer to this measure as (ΔRID). All image analysis was done using functions of Fiji software.

Trophoblast migration assay.

Individual Millicell 24-well hanging inserts with a pore size of 5 μm (catalog no. MCMP24H48; EMD Millipore Corporation) were placed into a 24-well plate and the upper chamber incubated with ECM gel (catalog no. E1270; Sigma-Aldrich) for 2 h at 37°C. The excess liquid was aspirated, and the inserts were left to air dry for 10 min before the addition of 5 × 10⁶ primary trophoblasts and an equal number of NIEs or IEs to the upper chamber and lower chamber in a final volume of 500 μl. Endothelin signaling inhibition experiments used bosentan at a final concentration of 80 μM in the medium, with a corresponding concentration of its vehicle (dimethyl sulfoxide [DMSO]) in the controls. Trophoblasts were exposed to the stimuli for 48 h at 37°C. The medium was removed and the inserts gently rinsed with PBS prior to fixation of the cells on both surfaces of the membrane with 4% paraformaldehyde for 5 min. After removing the fixative, cells on the upper layer were removed with cotton swabs, and the cells on the lower surface were stained with 0.5% crystal violet solution for 10 min. The number of cells which had migrated through the insert was quantified by dissolving the cells in 200 μl of 10% acetic acid, and the absorbance at 595 nm was then read using a GloMax Explorer multimode microplate reader (Promega) (67).

Statistical analysis.

Human birth weight and placental weight, and area of positive staining of EDNRB and EDN were examined using one-way ANOVA with post hoc Dunnett correction for multiple comparisons. Two-sided χ² tests were also used to interrogate differences between the proportion of positive and negative samples in EDNRB and EDN staining. Risks for adverse outcomes were assessed using odds ratios and represented as the odds ratio and 95% confidence intervals. Multivariate analysis of variance (MANOVA), using infection as the dependent variable, and multivariate multiple regression, using PM status, nested within infection, as the dependent variable were performed in R.

Maternal peripheral parasitemia was compared using two-sample t tests. Fetal weight in the mice was assessed with a linear mixed-model approach, which incorporated fetal genotype and maternal infection status as fixed effects alongside a litter-wise random effect for fetuses from the same litter; this accounts for variation within each litter and the size of the litter, and analysis of variance (ANOVA) with Satterthwaite approximation for degrees of freedom was performed. Fetal viability in mixed litters was assessed using a generalized mixed-effects model approach, incorporating the same fixed and random effects as for the weight. The fixed effects were considered significant when the P value was <0.05. Linear mixed-model analysis was performed using the lme4 and lmerTest packages available for R. Premature delivery in the mice was tested using the two-sided χ² test. Correlation analysis for mouse and human data was performed using GraphPad Prism 6, and P values and Pearson correlation coefficient values are reported. Gene expression and protein quantifications from placentas were tested with either paired or unpaired Wilcoxon rank (Mann-Whitney) tests, as indicated in the individual captions of each figure. Trophoblast gene expression assays were tested using unpaired Wilcoxon rank (Mann-Whitney) tests. Trophoblast gene expression data in each case is representative of one experiment, which was repeated several times. Differences in ΔRID from imaging experiments were interrogated using a ratio-paired t test, and migration assays were tested using a one-sample t test with the hypothetical mean of 1. In all cases, results were considered significant when the P value was <0.05.