Genomic selection signals in Andean highlanders reveal adaptive placental metabolic phenotypes that are disrupted in preeclampsia

Abstract

Background

The chronic hypoxia of high-altitude residence poses challenges for tissue oxygen supply and metabolism. Exposure to high altitude during pregnancy increases the incidence of hypertensive disorders of pregnancy and fetal growth restriction (FGR) and alters placental metabolism. High-altitude ancestry protects against altitude-associated FGR, indicating hypoxia tolerance that is genetic in nature. Yet not all babies are protected and placental pathologies associated with FGR occur in some Andean highlanders.

Methods

We examined placental metabolic function in 79 Andeans (18–45y; 39 preeclamptic, 40 normotensive) living in La Paz, Bolivia (3600 – 4100m) delivered by unlabored Cesarean section. Using a selection-nominated approach, we examined links between putatively adaptive genetic variation and phenotypes related to oxygen delivery or placental metabolism.

Results

Mitochondrial oxidative capacity was associated with fetal oxygen delivery in normotensive but not preeclamptic placenta and was also suppressed in term preeclamptic pregnancy. Maternal haplotypes in or within 200kb of selection-nominated genes associated with lower placental mitochondrial respiratory capacity (PTPRD), lower maternal plasma erythropoietin (CPT2, POMC, and DNMT3), and lower vascular endothelial growth factor in umbilical venous plasma (TBX5). A fetal haplotype within 200kb of CPT2 was associated with increased placental mitochondrial complex II capacity, placental nitrotyrosine and GLUT4 protein expression.

Conclusions

Our findings reveal novel associations between putatively adaptive gene regions and phenotypes linked to oxygen delivery and placental metabolic function in highland Andeans, suggesting such effects may be of genetic origin. Our findings also demonstrate maladaptive metabolic mechanisms in the context of preeclampsia, including dysregulation of placental oxygen consumption.

Graphical Abstract

Introduction

Fetal growth is critically dependent on placental metabolism, both glycolytic and oxidative, to drive the active exchange of nutrients and production of hormones (1). These fundamental processes are severely challenged under conditions of limited oxygen availability, such as the hypobaric hypoxia of high altitudes. High-altitude gestation thus represents a compound physiological challenge of meeting the metabolic demands of pregnancy under conditions of chronic ambient hypoxia. Highlighting the clinical impact of this scenario, high altitude increases the incidence of placental disorders of pregnancy, including fetal growth restriction (FGR) and preeclampsia (2). Given that preeclampsia and FGR potently jeopardize maternal and infant survival physiological attributes protecting against these conditions at high altitudes should be subject to intense selective pressure.

Highland Andean and Tibetan ancestry protect against altitude-associated FGR and the incidence of preeclampsia among Tibetans is half that of lowland Han Chinese living at high altitudes (3), indicating adaptation to hypoxia. In Andeans, protection of fetal growth at high altitudes has been attributed, in part, to greater fetal oxygen delivery through augmented maternal uterine artery blood flow (4), and the selection of key genes encoding proteins with pleiotropic effects, including intracellular energy sensing, metabolic regulation and vascular control (5). Although maternal arterial oxygenation positively associates with birthweight at high altitude (6), the pregnancy-associated rise in maternal ventilation and arterial oxygen saturation, together with higher hemoglobin, raise uterine oxygen content to sea level values, even in cases of lower uterine artery blood flow (7). Thus, maternal hypoxia alone cannot explain altitude-associated FGR. Reports of similar altitude-associated reductions in uteroplacental oxygen delivery at term between Andean and European women (8) further support the involvement of factors other than maternal oxygen delivery for protecting fetal growth in Andeans. We and others postulate that greater efficiency in placental oxygen utilisation, nutrient transport and fetal substrate utilisation contribute to altitude and ancestry-associated differences in fetal growth (7, 8).

Prior studies demonstrate altered placental metabolic function at high altitudes. Placental substrate preference is reported to switch towards anaerobic glycolysis in highland-resident Europeans and Andeans (9). In contrast, others detected no change in placental glycolytic metabolites following labor in European-ancestry women (10). High altitude also affects placental mitochondrial function, as indicated by suppressed respiratory complexes I and IV alongside impaired respiratory capacity in Europeans (11), and increased mitochondrial density within chorionic villi in Tibetans (12). Increased placental AMP-activated protein kinase (AMPK) activation has also been observed at high altitudes and in FGR (13). AMPK is an intracellular energy sensor linked to the regulation of mitochondrial biogenesis (14). Recent work showing increased placental AMPK activation across a gradient of low, moderate or high altitude during pregnancy suggest that AMPK pathway activation may be involved in matching placental metabolic responses to the magnitude of maternal hypoxia (15). Single nucleotide polymorphisms located near PRKAA1, the gene encoding the α−1 catalytic subunit of AMPK, showed evidence of natural selection in highland Andeans and associated with birthweight in high-altitude resident women of Andean or European descent (5). This suggests that AMPK pathway responses may be influenced by high-altitude genetic adaptation.

Placental metabolic function is also disrupted in placental pathologies, including preeclampsia which is prevalent in Andean highland populations (2). Features of preeclampsia include placental hypoxia (16, 17), restricted uterine artery blood flow, and low birth weight at high altitude (2, 18). Placental hypoxia leads to metabolic stress, in turn disrupting placental mitochondrial function (19). This has been described in early-onset preeclampsia through activating the mitochondrial unfolded protein response (UPR^mt) (20), a protective signaling pathway that prevents the accumulation of oxidatively damaged unfolded/misfolded proteins, such as the electron transfer system complexes. Upregulation of UPR^mt components suppresses mitochondrial oxidative phosphorylation (OXPHOS) capacity in early-onset preeclampsia (20). Altered placental mitochondrial function, including differential expression of mitochondrial respiratory complexes and markers of mitochondrial fusion and fission, has also been shown to differ in term compared to preterm preeclampsia (21).

In summary, existing literature indicates that unique placental metabolic phenotypes may protect against pregnancy disorders in highland Andeans and that these effects may be of genetic origin. Yet, the metabolic features that distinguish adaptative from maladaptive placental phenotypes in highland Andean pregnancy remain unclear. Here, we first contrasted placental metabolic phenotypes in normotensive and preeclamptic Andean maternal-infant pairs residing at high altitudes (La Paz, Bolivia, 3600–4100m), hypothesizing that putatively adaptive placental phenotypes would not be present in preeclampsia. To establish the genetic origins of placental phenotypes, we subsequently tested associations between genetic regions under selection in the maternal and fetal genomes against phenotypes related to oxygen delivery and placental metabolism. We hypothesized that placental metabolic phenotypes that are putatively advantageous to placental oxidative metabolism and fetal oxygenation at high altitudes would associate with maternal or fetal genetic variation within genomic regions showing evidence of positive selection in Andeans, and that this relationship would be disrupted in preeclampsia.

Methods

Anonymized data and materials have been made publicly available at the University of Cambridge repository and can be accessed at 10.17863/CAM.97107

Study Population and Design

Included in the study were 79 mother-infant dyads (n=39 PE and n=40 normotensive controls) who received prenatal care and underwent Cesarean section delivery between August 2018 and June 2019 at the Hospital Materno-Infantil, the largest public maternal-child hospital in the La Paz-El Alto region. The study was approved for scientific rigour and ethical approval under the Colorado Multiple Institutional Review Board (Protocol #17–1529) and the Caja Nacional de Salud, Bolivia.

Statistical analysis

Control and preeclamptic parameters were contrasted using an unpaired student t test. Association between physiological parameters was assessed using simple linear regression. Genomic selection scan analysis was conducted on the full cohort using Integrated Haplotype Score (iHS), with iHS ≥ 3 considered as significant. Genotype-phenotype associations were established using regression analysis. Age and body mass index (BMI) were controlled as covariates and a false discovery rate (FDR) corrected level of p<0.05 was applied. Data analyses were conducted using R v4.2 and Graphpad Prism v9.

Detailed methods are provided in the Supplemental text.

Results

Abnormalities in placental mitochondrial respiratory function in preeclampsia

Our study cohort comprised 79 Andean maternal-infant pairs, including 40 normotensive controls and 39 preeclamptic cases, residing at high altitudes in La Paz-El Alto, Bolivia (3600 – 4100m). Normotensive or preeclampsia status classification strictly adhered to guidelines established by the American College of Obstetrics and Gynecology (22), including new-onset hypertension after 20 weeks of gestation accompanied by proteinuria, or laboratory values or symptoms indicative of renal, liver, or central nervous system dysfunction. Maternal and newborn characteristics are presented in Table 1. Compared to controls, women with preeclampsia had a higher incidence of preterm births (76%) and delivered infants of lower birth weight, with 47% born small for gestational age (Table 1). Preeclamptic pregnancies showed evidence of maternal and fetal hypoxemia. This was indicated in previously published work in this cohort demonstrating reduced umbilical venous PO₂ and PO₂ venous-arterial difference, increased circulating levels of antiangiogenic factor sFlt1 in preeclampsia, elevated concentrations of umbilical cord blood and maternal hemoglobin, and elevated concentrations of umbilical cord blood and erythropoietin (EPO) (23). However, placental expression of the EPO receptor was similar between control and preeclampsia (Figure S1A).

Table 1.

Maternal and newborn characteristics

A. Maternal characteristics
Variables	Control	Preeclampsia	P-value
Maternal age, yrs	32.4 ± 5.6	33.5 ± 5.1	.37
Altitude of birth, m	3341 ± 981	3687 ± 233	.04
Altitude of childhood, m	3414 ± 931	3775 ± 259	.03
Duration HA residence, yrs	29.9 ± 7.9	31.7 ± 6.7	.29
Weight (prepregnant), kg	59.6 ± 10.9	63.9 ± 9.6	.09
Height, m	1.56 ± .08	1.55 ± .06	.59
Body mass index, kg/m2	24.6 ± 4.3	26.7 ± 4.0	.05
SpO2, %	92.6 ± 3.8	89.5 ± 15.9	.25
Hemoglobin, g/dL	13.7 ± 2.7	14.6 ± 1.4	.08
Gravidity	2.1 ± .8	2.2 ± 1.1	.78
Parity	.97 ± .78	.86 ± .87	.55
Prenatal visits, #	5.7 ± 1.7	4.7 ± 1.9	.05
First prenatal visit, week	19.6 ± 9.2	15.8 ± 8.7	.08
B. Newborn characteristics
Variables	Control	Preeclampsia	P-value
Gestational age, weeks	38.1 ± 2.1	33.9 ± 3.9	<.001
Preterm, %	15 [5, 32]	76 [59, 89]	<.001
Birth weight, g	3158 ± 511	2081 ± 761	<.001
Birth weight* (adjusted), g	2825 ± 79	2438 ± 77	.003
SGA, %	0 [0, 11]	47 [29, 65]	<.001
Infant sex (male), %	54 [37, 71]	62 [44, 78]	.50
Apgar (5-min)	8.9 ± .3	8.7 ± .7	.04

Shown are means ± standard deviation for continuous variables or percentage and 95% CI for proportions.

^*Adjusted for gestational age at delivery and infant sex (for this variable only the estimated marginal mean and standard error of the mean are shown).

Abbreviations: HA, high altitude; SpO₂, peripheral oxygen saturation; SGA, small for gestational age.

Placental mitochondrial oxidative capacity was assessed using high-resolution respirometry and a substrate-uncoupler-inhibitor titration on cryopreserved placental villous tissue (24). Term preeclamptic pregnancies showed an 18% decrease in placental maximal electron transfer system capacity (GMS_E) and a 42% decrease in complex II driven respiration (S_E) compared to term, normotensive controls (Figure 1A). No effect of preeclampsia on mitochondrial respiratory capacity was observed in preterm pregnancy in comparison to pre-term controls (Figure S1B). The capacity of respiratory complexes I and II relative to maximal OXPHOS capacity was examined through the calculation of the flux control ratio (FCR). No difference was observed in Complex I FCR, yet Complex II FCR was suppressed by 35% in term preeclampsia cases compared to control (Figure 1B). Again, no difference between preeclampsia control placenta was observed in Complex II FCR in preterm pregnancy (Figures S1C). Complex IV capacity also displayed no difference between control and preeclampsia, either in term or preterm (Figure S1D–E). Placental protein expression of citrate synthase, a putative marker of mitochondrial content, was similar between control and preeclampsia (Figure 1C). Taken together, these findings suggest a specific impairment of complex-II supported respiration in preeclamptic placentas at term, with no overt change in overall mitochondrial content or capacity.

Figure 1: Placental mitochondrial phenotype disruption in preeclampsia.

A. Mitochondrial respiratory capacity in placenta delivered at term using a substrate/uncoupler/inhibitor assay as oxygen flux (JO₂) corrected to placental wet mass, comprising: LEAK state respiration with malate and pyruvate (PM_L), OXPHOS state respiration via complex I with malate and pyruvate (PM_P) and with the addition of glutamate (GM_P), OXPHOS state respiration via complex I and II with the addition of succinate (GMS_P) and maximal respiratory capacity via the addition of the mitochondrial uncoupler FCCP (GMS_E). Complex I was then inhibited through the addition of rotenone restricting electron flux to the S-pathway via complex II (S_E). Electron flux was then blocked through the addition of complex III inhibitor antimycin A, leading to residual oxygen consumption (R_OX). Data are represented as an average value obtained from duplicate oxygraph runs on the same cryopreserved placental sample in both control (white=28) and preeclamptic (blue, n=13) term pregnancy, * p≤0.05.

B. Flux control ratios of complex I and II driven respiratory capacity in term pregnancy.

C. Protein expression of citrate synthase in the placenta corrected to vinculin in control (n=37) and preeclampsia (n=31).

D. Simple linear regression analysing the relationship between GMS_P umbilical cord venous PO₂ in control (n=22) and preeclampsia (n=16).

E. Expression of placental proteins involved in the UPR^mt corrected to β-actin in control (n=6) and preeclampsia (n=13).

F. Protein expression of placental superoxide disumutase 2 (SOD2) in the placenta corrected to vinculin in control (n=34) and preeclampsia (n=32).

Data presented as ±SD and analysed using an unpaired student t test, *p<0.05, ** p<0.01.

To examine whether placental mitochondrial respiratory function was related to fetal oxygen delivery we used linear regression analysis, which revealed a negative association between umbilical venous PO₂ and maximal respiratory capacity in controls (Figure 1D), suggesting a lower placental respiratory capacity translated to increased oxygen delivery to the fetus. This relationship was not present in preeclampsia (Figure 1D).

We assessed indicators of metabolic stress in the placenta and demonstrated greater protein expression of UPR^mt components in preeclampsia versus control. Specifically, a 77% increase in the expression of the mitochondrial chaperone tumorous imaginal disc 1 (TID1) in preeclampsia. This occurred alongside a 96% increase in the regulator of mitochondrial protein degradation machinery caseinolytic mitochondrial matrix peptidase proteolytic subunit (CLPP) and a 31% increase in activating transcription factor 5 (ATF5), a regulator of TID1. Heat shock protein 60 (HSP60) did not differ by preeclampsia status (Figure 1E). We also observed a 16% decrease protein expression of the antioxidant enzyme superoxide dismutase 2 (SOD2, also known as manganese-dependent dismutase [MnSOD]) in preeclamptic placenta compared to controls (Figure 1F). No change was found in the placental protein expression of nitrotyrosine, a product of tyrosine nitration by reactive nitrogen species and an indicator of oxidative stress (Figure S1F). Original Western blot images for these analyses are presented in Figure S2.

Population genetic ancestry and identification of putatively adaptive haplotypes

Maternal-infant pairs were genotyped using the 1.8 million SNP Illumnia MultiEthnic Genotyping Array, which provides broad marker representation in ancestrally-diverse populations (25). Global admixture coefficients were estimated using ADMIXTURE (26) with k=4 unsupervised clusters. Our Bolivian cohort was primarily of highland Amerindian origin (88%) with minimal European, African, and lowland Amerindian admixture (6%, 2% and 5%, respectively) (Figure 2).

Figure 2: Population genetic ancestry.

Genetic cluster analysis of the Bolivian cohort (BOL) in our study, and six reference panels: European (Utah residents from northern and western Europe [CEU]), African (Yoruba from Ibadan, Nigeria [YR]), Peruvian (Lima, Peru [PEL]), Mexican (Los Angeles, CA [MXL]), Colombian [Medellin, Colombia [CLM]), and Puerto Rican [Puerto Rico [PUR]) individuals from the 1000 Genomes phase 3 panel for genetic cluster formation. Colors correspond to the proportion of ancestry from each of four presumed ancestral populations (K=4). Vertical bars represent individuals.

Within-population integrated Haplotype Scores (iHS) were used as an index of positive selection in the maternal and fetal genetic data. Putatively adaptive haplotypes (iHS ≥3) with functional physiological relevance were defined as those overlapping with an a priori list of candidate genes implicated in hypoxic cellular signaling, oxygen transport, and metabolic function (Table S1). These haplotypes were prioritized for genotype-phenotype association analysis.

Putatively adaptive maternal and fetal haplotypes associate with phenotypes related to oxygen delivery and placental mitochondrial respiratory capacity.

Phenotypes assessed in the genotype-phenotype association analysis included: birthweight, preeclampsia and FGR status, maternal and newborn haemoglobin, umbilical venous and arterial blood gases, and measures of maternal and umbilical venous plasma angiogenic (vascular endothelial growth factor, free VEGF) and antiangiogenic (soluble fms-like tyrosine kinase, sFlt1) factors (23). Additional phenotypes included placental mitochondrial respiratory capacity (Figure 1A), placental expression of SOD2 (Figure 1F), the EPO receptor and nitrotyrosine (Figure S1A,F).

Maternal genotype-phenotype associations revealed associations between prioritized haplotypes and key outcome measures. In the maternal genome, haplotypes within 200kb of CPT2 and both POMC and DNMT3 associated with lower maternal plasma EPO (Figures 3A–B). A haplotype within 200kb of TBX5 associated with lower protein levels of the angiogenic factor VEGF in umbilical venous plasma (Figure 3C). Finally, a haplotype within PTPRD associated with lower placental respiratory capacity, specifically maximal OXPHOS capacity (GMS_P, Figure 3D) and maximal electron transfer system capacity (GMS_E, Figure 3E).

Figure 3: Maternal genotype – phenotype association analyses reveal novel links between putatively adaptive haplotypes and phenotypes related to oxygen delivery and placental mitochondrial respiration.

A–E. Association of phenotypes with a priori gene candidates showing evidence of positive selection (iHS score ≥ 3) in the maternal genome. Significant associations were identified with: maternal plasma erythropoietin (EPO) levels, vascular endothelial growth factor (VEGF) in umbilical cord plasma and placental respiratory capacity. This was measured as oxygen flux (JO₂) corrected to placental wet mass in the following states: maximal OXPHOPS (GMS_P) and maximal electron transfer chain flux (GMS_E). Linear regression analysis (adjusted for maternal age and BMI) followed by false discovery rate (FDR) correction identified phenotypes significantly related to a priori genes of interest. All associations presented are FDR corrected p value <0.05.

F. Maternal haplotype location, including chromosome (Chr) and top marker position, and iHS score.

In the fetal genome, a haplotype within 200kb of CPT2 associated with three phenotypes: lower expression of placental nitrotyrosine and GLUT4 (Figure 4 A–B), and an increased placental mitochondrial complex II capacity (Complex II FCR, Figure 4C).

Figure 4: Fetal genotype – phenotype association analyses reveal novel links between putatively adaptive haplotypes and placental metabolic phenotypes.

A–C. Association of placental phenotypes with CPT2, an a priori gene candidate showing evidence of positive selection (iHS score ≥ 3) in the fetal genome. Negative associations were identified for placental nitrotyrosine and GLUT4 protein expression normalized to vinculin, and a positive association for mitochondrial respiratory complex II flux control ratio (FCR). Linear regression analysis (adjusted for maternal age and BMI) followed by FDR correction identified phenotypes significantly related to a priori genes of interest. All associations presented are FDR corrected p value <0.05.

D. Table detailing the fetal haplotype location, including chromosome (Chr) and top marker position, and iHS score.

To consider the effects of preeclampsia on the relationship between the putatively adaptive haplotypes and phenotypes, we examined these interactions in control and preeclamptic subjects separately. Putatively adaptive genetic variation in PTPRD within the maternal genome was associated with lower respiratory function in controls (Figure 3D–E). In preeclampsia, this association was more extreme, with a more profound drop in maximal respiratory capacity and maximal electron transfer system function (control vs. preeclampsia difference for GMS_P p=0.02, for GMS_E p=0.03). This was the only significant effect of preeclampsia on the genotype-phenotype interaction. Preeclampsia status did not affect the relationship between haplotypes under positive selection and phenotypes in the fetal genome.

Discussion

High-altitude pregnancy increases the incidence of placental pathologies and suppresses fetal growth. Highland Andeans exhibit relative protection from altitude-associated growth restriction, yet this does not appear to be the case for placental pathologies such as preeclampsia (2). Here, we explore the placental metabolic phenotype of highland-resident Andeans with and without preeclampsia. We demonstrate the relationship between low placental oxygen consumption and high oxygen delivery to the fetus is lost in preeclampsia, and occurs alongside suppressed placental mitochondrial respiratory capacity at term. Examining maternal and fetal genomes revealed haplotypes under selective pressure in this Andean population. Association analysis revealed novel links between selected haplotypes and phenotypes indicative of altered placental metabolic function and uteroplacental oxygenation. Our findings indicate a role for putatively adaptive genomic variants in regulating placental metabolism, and thereby fetal oxygen and nutrient delivery, at high altitudes.

Evidence suggests that high-altitude exposure during pregnancy induces placental metabolic alterations, including suppressed placental mitochondrial respiratory function (11), which may enable greater allocation of oxygen to the fetus (7). Disruption of placental metabolic function due to placental hypoxia is also central to the pathology of preeclampsia and may manifest during trophoblast development or result from underlying maternal mitochondrial defects (27). In lowland pregnancy, placental mitochondrial oxidative capacity is reportedly suppressed in preeclampsia (<34 weeks) versus normotensive controls (20). Placental mitochondrial function in preeclampsia is impacted by pregnancy duration at sea level. In preeclamptic placentas that progress to term, reports demonstrate that tissue function is preserved through salvaging of dysfunctional mitochondria (21). In comparison to control term, preeclamptic term placenta had higher mitochondrial LEAK respiration, suppressed reserve capacity and alterations to mitochondrial complex levels including higher complex II and III protein (21). Our results in high altitude pregnancies indicate lower mitochondrial respiratory capacity, including the suppression of complex II respiration, in preeclamptic versus normotensive term placenta. This was not seen in preterm placenta. Of note, preterm controls may not be a perfect control comparator for preterm preeclamptic placental function. Although normotensive, the preterm nature of the delivery indicates a complication.

Our experimental approach for assessing mitochondrial respiratory function does not distinguish between subpopulations of mitochondria from cytotrophoblasts and syncytiotrophoblasts, which have been shown to exhibit differences in respiration and morphology (28). It was not feasible to carry out subcellular fractionations at the point of delivery in our study, but we acknowledge this possible limitation. Our technique does, however, result in the retention of all mitochondriawithin an integrated cellular context, including damaged and dysfunctional mitochondria which might otherwise be lost during fractionation, and this is a strength of our approach (29).

In healthy highland pregnancy, low mitochondrial respiratory capacity is associated with high fetal oxygen delivery, indicated by greater umbilical venous PO₂. This suggests that under conditions of maternal hypoxia, reduced placental mitochondrial respiration may serve an adaptive function by preserving fetal oxygenation. This presumed adaptive function is lost in preeclampsia, suggesting that a failure to reduce placental respiration may contribute to impaired fetoplacental oxygen delivery in high-altitude preeclampsia. Multi-modal magnetic resonance images in a limited number of placentas from normotensive pregnancies indicate that most fetal oxygen uptake occurs within a short distance of the center of placental lobules (30), therefore further studies are needed to establish whether and how changes in placental mitochondrial respiration affect fetal oxygenation.

In highland Andean preeclamptic pregnancies, we demonstrate suppressed respiratory capacity versus normotensive controls with concomitant activation of the UPR^mt pathway, including upregulation of CLPP, TID1 and ATF5. The presence of UPR^mt may provide a mechanistic explanation for the reduction of complex II activity, as activation of UPR^mt has been associated with suppressed complex II respiratory capacity and expression of a complex II subunit in trophoblastic cells exposed to hypoxia-reoxygenation (20). The changes to UPR^mt components reported previously align with our findings of upregulated TID1, yet were distinct in demonstrating downregulation of CLPP (20). We report decreased expression of SOD2 in highland preeclamptic pregnancy, which is in line with previous reports of a decreased SOD expression in placenta from lowland preeclamptic pregnancies (31). This occurred alongside no change in placental expression of the oxidative stress marker nitrotyrosine. Our findings suggest subtle changes in abundance of mitochondrial quality-control and antioxidant mechanisms in highland preeclamptic placenta. Further examination of redox balance is warranted to establish the functional importance of UPR^mt and oxidative stress in high-altitude preeclampsia and how these factors affect placental metabolism in highland Andeans; such studies are underway.

Using an integrative approach, our selection-nominated association analysis was designed to identify links between putatively adaptive genetic variation and phenotypes related to oxygen delivery or placental metabolism. Our analyses identified phenotype associations for genes that showed evidence of positive selection, as assessed by iHS, and belonged to our a priori gene list of functionally relevant genetic targets and were proximal to the selection signal. Whilst the direct mechanism(s) of action by which such genotypes affect phenotypes related to fetal development remain unclear, our analyses identified potential mechanisms and pathways of interest. We demonstrate association between a maternal haplotype containing PTPRD and suppressed placental mitochondrial respiratory capacity. PTPRD encodes a transmembrane receptor that dephosphorylates STAT3 at tyrosine 705 (32). Metabolic effects of PTPRD have been demonstrated in the context of fetal growth in sheep, with hepatic glucose production and hepatic PTPRD expression decreased by over 80% in FGR (33). PTPRD expression is also linked to development of metabolic diseases, including non-alcoholic fatty liver disease (34). A variant within PTPRD (rs17584499) is associated with onset of type 2 diabetes in Han Chinese (35, 36) acting via suppressed PTPRD expression serving to down-regulate insulin receptors (35). Therefore, whilst the mechanistic link between PTPRD and mitochondrial respiration remains unclear, the selected-for PTPRD gene variants might be acting to suppress expression and thereby glucose uptake. While additional studies are needed, the more extreme observation that the relationship between PTPRD and suppressed respiratory capacity in preeclampsia compared to control may indicate a novel pathway by which preeclampsia impairs placental metabolism and fetal growth.

At the level of oxygen carriage, lower maternal plasma EPO concentration was associated with a maternal haplotype within 200kb of CPT2. As in our study, CPT2 has been shown previously to be under strong selective pressure in Andeans (37). CPT2 functions as a shuttle for L-carnitine into the mitochondria where it is required for fatty acid oxidation. We describe a novel role for CPT2 in relation to EPO concentration that may relate to the function of L-carnitine in maintenance of red blood cell stability, as oral L-carnitine treatment can increase haematocrit in response to EPO treatment in haemodialysis patients (38). Lower plasma EPO was also associated with a maternal haplotype including POMC and DNMT3. DNMT3A affects hypoxic cellular gene expression by inducing the de novo methylation of CpG motifs located within the EPAS1 promoter, thereby prohibiting HIF-2α activation (39), which is required for the induction of renal and hepatic EPO production. Further, the expression of EPO is dependent upon methylation events within the EPO promoter and 5’-untranslated region (40). DNMT3 is also the most commonly mutated gene in clonal hematopoiesis (41). Notably, the action of EPO extends beyond haematopoietic tissue as EPO receptors are expressed at high levels in hypothalamic POMC neuronal cells (42), where EPO function is linked to metabolic homeostasis (42). Together, these results suggest strong selective pressure on the low plasma EPO phenotype, with potentially widespread mechanistic implications. This may be particularly relevant in preeclampsia as the overproduction of red blood cells, or polycythemia, increases risk of developing hypertensive pregnancy complications (43).

Our results indicate a further functional role for CPT2, as a fetal haplotype including CPT2 was associated with lower placental nitrotyrosine and GLUT4 expression and increased complex II FCR. These diverse metabolic parameters are linked to oxidative stress, glucose uptake and respiratory chain function, respectively. In relation to oxidative stress, CPT2 mediates mitochondrial fatty acid oxidation, which can potentiate oxidative stress-induced inflammation (44). Increased complex II FCR alongside lower GLUT4 may indicate an adaptive phenotype whereby oxidative metabolism is maintained and appropriately matched to fetal oxygen delivery, whilst glucose is conserved.

Finally, our data demonstrate an association between a maternal haplotype within 200kb of TBX5 and decreased umbilical venous plasma VEGF concentration. TBX5 is essential in regulating developmental gene expression controlled, at least in part through direct interaction with HIF-1α (45). Strengthening our observations, whole genome sequencing identifiedTBX5 as being subject to recent positive selection in highland Andeans, and found strong associations between the Andean genotype and cardiovascular phenotypes (46). Moreover, the Genotype-Tissue Expression portal demonstrates high expression levels of TBX5 in the human aorta and left ventricle (46). The mechanistic link between TBX5 and VEGF has been attributed to TBX5 mediated decrease in expression of Sox7, which was postulated to interact with the VEGF/Notch signaling systems impacting vascular branching in zebrafish (47). Further investigation is required to determine whether the Sox7-mediated mechanism is responsible for the link between TBX5 and umbilical cord VEGF, but it may indicate potential effects on vascular branching impacting fetoplacental oxygenation.

To conclude, we provide evidence of distinct placental metabolic phenotypes that differ between normotensive vs. preeclamptic pregnancy in Andean highlanders, revealing disruption in the relationship between oxygen delivery to the fetus and placental oxygen consumption. Further, we demonstrate novel links between putatively adaptive phenotypic traits linked to oxygen carriage and mitochondrial oxygen consumption in the placenta that are under selective pressure and may act to affect fetal oxygenation. Together, this indicates a delicate balance between adaptation of the oxygen cascade enabling successful pregnancy outcomes in highland Andeans and maladaptation leading to placental pathology.

Non-standard abbreviations and acronyms

AMPK

AMP-activated protein kinase

ATF5

Activating transcription factor 5

BMI

Body mass index

CLPP

Caseinolytic mitochondrial matrix peptidase proteolytic subunit

CPT2

Carnitine palmitoyl transferase 2

DNMT3

DNA methyltransferase 3

EPO

Erythropoietin

FCCP

Carbonyl cyanide-p-trifluoromethoxyphenylhydrazone

FGR

Fetal growth restriction

FDR

False discovery rate

FCR

Flux control ratio

GM_P

OXPHOS state respiration via complex I with the addition of glutamate

GMS_P

OXPHOS state respiration via complex I and II with the addition of succinate

GMS_E

Maximal respiratory capacity via the addition of the mitochondrial uncoupler FCCP

HIF

Hypoxia inducible factor

HSP60

Heat shock protein 60

iHS

Integrated Haplotype Scores

JO₂

Oxygen flux

OXPHOS

Oxidative phosphorylation

PM_L

LEAK state respiration with malate and pyruvate

PM_P

OXPHOS state respiration via complex I with malate and pyruvate

POMC

Proopiomelanocortin

PTPRD

Protein tyrosine phosphatase receptor-δ

R_OX

Residual oxygen consumption

S_E

Electron flux restricted to the S-pathway via complex II through addition of rotenone

SOD2

Superoxide dismutase 2

TID1

Tumorous imaginal disc 1

UPR^mt

Mitochondrial unfolded protein response

VEGF

Vascular endothelial growth factor