Fetal metabolic adaptations to cardiovascular stress in twin-twin transfusion syndrome

Summary Monochorionic-diamniotic twin pregnancies are susceptible to unique complications arising from a single placenta shared by two fetuses. Twin-twin transfusion syndrome (TTTS) is a constellation of disturbances caused by unequal blood flow within the shared placenta giving rise to a major hemodynamic imbalance between the twins. Here, we applied TTTS as a model to uncover fetal metabolic adaptations to cardiovascular stress. We compared untargeted metabolomic analyses of amniotic fluid samples from severe TTTS cases vs. singleton controls. Amniotic fluid metabolites demonstrated alterations in fatty acid, glucose, and steroid hormone metabolism in TTTS. Among TTTS cases, unsupervised principal component analysis revealed two distinct clusters of disease defined by levels of glucose metabolites, amino acids, urea, and redox status. Our results suggest that the human fetal heart can adapt to hemodynamic stress by modulating its glucose metabolism and identify potential differences in the ability of individual fetuses to respond to cardiovascular stress.

TTTS is a unique model to examine fetal metabolic adaptations to cardiovascular stress Amniotic fluid metabolites reveal significant changes in energy metabolism in TTTS Identification of distinct metabolic clusters underscores the heterogeneity of TTTS Our study highlights potential differences in individual fetal responses to stress

INTRODUCTION
Monochorionic-diamniotic twin pregnancies comprise 70% of identical or monozygotic twins and are susceptible to unique complications arising from a single placenta shared by two fetuses (monochorionicity), each within its own amniotic sac (diamnionicity). Fetal health is intertwined by vascular connections within the placenta which serve as conduits for bidirectional blood flow and control the hemodynamic equilibrium between the twins. With a significant imbalance in blood flow from one fetus to the other, the twins are at risk for a complication termed twin-twin transfusion syndrome (TTTS). In this syndrome, one fetus becomes hypovolemic (the donor), while the other becomes volume overloaded (the recipient). The massive volume shift to the recipient leads to hypoperfusion of the donor's kidneys, activation of the renin-angiotensin-aldosterone system (RAAS), and circulation of vasoactive mediators, promoting a vicious cycle. 1,2 In severe cases of TTTS the outcome is dramatic. Indeed, both fetuses may experience extreme cardiovascular stress and cardiac failure, but for different causes. Left untreated, the fetal death rate approaches 90%, and one-half of survivors are neurodevelopmentally impaired. 3 TTTS complicates approximately 10% of monochorionic-diamniotic twin pregnancies. 4 Although TTTS affects a relatively small proportion of all pregnancies, the condition offers a unique opportunity to interrogate fetal physiology and cardiovascular adaptation. 5 How the fetus responds to cardiovascular stress is of broad interest given the high prevalence of conditions that affect cardiovascular function in utero (e.g., maternal diabetes, infection, genetic conditions, and fetal anomalies). 6 Surgical treatment of TTTS involves laser ablation of the offending placental vessels during the second trimester of pregnancy, which enables the procurement of amniotic fluid for molecular analysis during a window of fetal adaptation to cardiovascular stress-information that would otherwise not be available during pregnancy. Indeed, the amniotic fluid cell-free transcriptome reflects gene expression from many organs, including the developing heart. 7,8 Furthermore, recent TTTS amniotic fluid microRNA (miRNA) studies found enrichment of miRNAs with roles in the cardiomyocyte stress response and metabolism 9 and reported differences in amniotic fluid miR-NAs of TTTS recipients with severe cardiomyopathy compared with to those with preserved cardiac function. 10 These untargeted global molecular analyses reveal the utility of amniotic fluid for monitoring fetal health and are essential for hypothesis generation, identification of prognostic biomarkers, and furthering our knowledge of conditions such as TTTS that cannot be effectively modeled in animals. 11 Variability in the natural course of TTTS raises the question of why some twins progress very rapidly to severe disease with decompensated heart failure, while others exhibit only changes in amniotic fluid volume, a proxy for fetal urine output. Although variation in the anatomy of the placental vasculature provides a mechanical explanation for TTTS, differences in fetal susceptibility to hemodynamic stress offers a potential contributing physiologic explanation. Determining the molecular underpinnings of the fetal response is critical for understanding short-and long-term effects on a child's health, particularly when considering the impact on fetal programming, also known as the Developmental Origins of Health and Disease (DO-HaD) hypothesis. Fetal programming is the concept that events or insults occurring during critical points of fetal development have long-lasting effects on the individual, especially with respect to future cardiovascular and metabolic disease risk. 12 Many mechanisms of fetal programming have already been proposed, including epigenetic modifications, effects on gene expression, and disturbances in fetal hypothalamic-pituitary-adrenal axis development. 13,14 At the same time, cell plasticity and proliferation in fetal life are required for rapid growth and development of the organism and adaptation to its environment. Notably, in contrast to the adult heart, fetal cardiomyocytes are capable of substantial regeneration. 15 This developmental flexibility appears to endow the fetus with the ability to withstand and recover from significant hemodynamic stress. 16,17 Indeed, we have proposed that reversion to the fetal genetic and metabolic programs is a protective mechanism and hallmark of adult heart failure. 18 Reconciling the predicted permanent consequences of fetal programming with the inherent flexibility of the developing fetus remains an unresolved issue of far-reaching clinical relevance. Here we explored the fetal response to cardiovascular stress using untargeted metabolite profiling of amniotic fluid from fetuses with severe TTTS to gain further insights into possible mechanisms and found heterogeneity among fetuses with severe TTTS.

Clinical characteristics of TTTS and control pregnancies
To evaluate fetal metabolic adaptations to cardiovascular stress, we collected midgestation amniotic fluid from TTTS (n = 22) and control (n = 10) pregnancies for untargeted metabolomic analysis ( Figure 1A; Table S1). We selected severe TTTS pregnancies (stage III) with evidence of cardiovascular stress based on abnormal blood flow patterns in the umbilical artery, ductus venosus, or umbilical vein-vessels that are unique to the fetal circulation ( Figure 1B). 19 These vessels are routinely interrogated in monochorionic-diamniotic twins using Doppler ultrasound to detect when there is: 1) increased placental resistance and cardiac afterload demonstrated by absent or reversed end-diastolic flow in the umbilical artery and 2) right heart overload and failure manifested by absent or reversed flow in the ductus venosus a-wave or pulsatile flow in the umbilical vein. 20 For TTTS cases, amniotic fluid was collected from the recipient twin sac at the time of fetoscopic laser surgery. 4 Control samples were obtained from singleton pregnancies with normal genetic testing and no fetal anomalies detected on prenatal ultrasound, utilizing discarded genetic amniocentesis specimens.
Maternal age and gestational age were significantly different between groups (Table 1); however, these variables had no significant impact on metabolite profiles based on correlation analysis ( Figure S1). Control patients tended to be older because the most common indication for genetic amniocentesis was advanced maternal age (35 years of age or greater). The earlier gestational age among controls is also not surprising given the timing of diagnostic genetic testing which often occurs earlier in gestation than laser surgery for TTTS. The distribution of fetal sex was similar in both groups. Of the TTTS pregnancies, 64% (14/22) of recipients and 50% (11/22) of donors had abnormal Doppler studies, indicating clinically evident cardiovascular stress in most of the TTTS twins. Detailed clinical characteristics are shown in Table S1.

Excessive amniotic fluid in TTTS recipients dilutes amniotic fluid components
A critical clinical characteristic is aberrant amniotic fluid volume, given that a fluid imbalance is the first diagnostic sign of TTTS in monochorionic-diamniotic twin pregnancies. We used amniotic fluid from the recipient twin who has excess amniotic fluid (polyhydramnios) as a result of volume overload and increased fetal urine output. (Donor twin amniotic fluid is by definition low (oligohydramnios) and is not acquired during laser surgery.) To assess the extent to which volume affected the concentration of amniotic fluid metabolites, we measured amniotic fluid protein concentration ( Figure 1C). The mean protein concentration was significantly lower in TTTS vs. control amniotic fluid (1. iScience Article greater than 2-fold dilution of amniotic fluid components in TTTS, in agreement with our prior work. 21,22 We therefore normalized metabolite data by protein concentration for each sample to adjust for dilution. Amniotic fluid metabolites reveal changes in fatty acid, glucose, and steroid hormone metabolism in TTTS Untargeted liquid chromatography/mass spectrometry (LC/MS)-based profiling of amniotic fluid metabolites recognized >400 metabolites. TTTS cases grouped separately from controls based on principal component analysis (PCA), with controls clustered more closely together ( Figure 1D). Among 133 differentially abundant metabolites (absolute fold change [FC] > 2, FDR %0.001), more than half exhibited low retention times upon chromatography employing an aqueous normal-phase matrix (e.g., fatty acids and lipids); a subset of these metabolites was effectively resolved and identified by targeted reverse-phase liquid chromatography (Table S2). iScience Article Capric acid, a saturated, medium-chain fatty acid, was the most abundant metabolite enriched in TTTS vs. control (log 2 FC 17.6, adjusted p = 8.32e À20 ; Figure 1E). Relative accumulation of several other saturated free fatty acids in TTTS amniotic fluid was also observed, including myristic and palmitic acids, which are known to promote adaptive cardiac growth and hypertrophy via AMP-activated protein kinase (AMPK) signaling. 23-25 AICA (5-aminoimidazole-4-carboxamide) riboside (AICAR), an AMPK activator, had the second highest FC in TTTS vs. control amniotic fluid (log 2 FC 11.7, adjusted p = 3.89e À6 ). AICAR positively regulates glucose uptake in muscle 26,27 and inhibits gluconeogenesis, 28 as well as glycogen synthesis, in rodents. 29 Notably, AICAR has been studied as a cardioprotective agent in animal models of cardiac ischemia. [30][31][32] Amniotic fluid pyruvate was depleted in TTTS amniotic fluid, consistent with other changes in glucose metabolites. Also depleted was retinoic acid, an essential transcriptional regulator of genes involved in lipid and energy metabolism 33,34 as well as the development of multiple organs, including the heart. 35,36 Retinoic acid levels were previously shown to be reduced in human adult heart failure. 37 Arachidonic acid and its derivatives (prostaglandins, thromboxane B2) were found to be enriched in TTTS. These polyunsaturated fatty acids (PUFAs; components of membrane phospholipids) are major mediators of inflammation, contribute to the regulation of vascular tone (reviewed in ref. 38,39 ), and have been associated with hypertension. 40 Cell damage from oxygen-derived free radicals and lipid peroxidation might also affect levels of amniotic fluid PUFAs.
We also noted an increase in adrenal steroid hormones with glucocorticoid and mineralocorticoid activity, including aldosterone precursors, deoxycorticosterone, corticosterone, and cortexolone (also known as 11-deoxycortisol, the precursor of cortisol). Umbilical cord blood corticosterone has been associated with fetal stress in labor, 41 and deoxycorticosterone is known to be produced in the fetal kidney. 42 Other metabolites of interest enriched in TTTS amniotic fluid reflected potential effects on fetal organ systems other than the heart. Endocannabinoids, such as 2-arachidonoylglycerol and anandamide, are arachidonic acid-derived neuroregulators that bind CB1 cannabinoid receptors and play important roles in brain development, specifically in guidance of axon growth and synapse formation. 43,44 Metabolites that accumulate in the setting of hepatobiliary injury or stasis, including bile acids (e.g., taurolithocholic acid) and pipecolinic acid (also known as pipecolic acid), 45,46 were also increased in TTTS amniotic fluid.
Taken together, metabolic findings suggest that TTTS recipient twins are exposed to major shifts in metabolism in response to hemodynamic overload. These changes include fatty acid and glucose utilization, as well as activation of the fetal hypothalamic-pituitary-adrenal axis. Compared to the control group, amniotic fluid metabolite profiles were highly variable within the TTTS group, suggesting significant heterogeneity within the disease group. To further investigate this heterogeneity, we reclustered the samples using the most variable metabolites in TTTS amniotic fluid (top 10%). Two subgroups emerged, which we labeled as TTTS cluster 1 (n = 8) and TTTS cluster 2 (n = 14; Figure 2A). The clusters were not clearly distinguishable by clinical characteristics (Table 2; case details are provided in Table S3). Maternal age, maternal body mass index, gestational age, fetal sex, and ultrasound parameters iScience Article were similar between clusters, including the frequency of abnormal Doppler studies that are reflective of cardiac strain, although three recipients from cluster 2 had evidence of mitral or tricuspid regurgitation compared with none from cluster 1. Of note, none of the pregnancies was complicated by maternal pregestational or gestational diabetes or hypertension (potential confounders). One pregnancy from each cluster was complicated by twin anemia-polycythemia sequence (TAPS). 47 Differential metabolite analyses revealed the enrichment of several amino acids in TTTS cluster 2 compared with cluster 1, including alanine, aspartate, methionine, and tyrosine ( Figure 2B; Table S4). Carnitine and its derivatives, important cofactors for the metabolism of long-chain fatty acids, were also increased in cluster 2, suggesting potential pressure to increase the transport capacity of fatty acids into mitochondria for b-oxidation. A change in energy metabolism in cluster 2 was also supported by the enrichment of ATP precursors or building blocks (creatine and ribose) and metabolites involved in the production of NAD+ (1-methylnicotinamide, quinolinic acid), an essential substrate for redox reactions. 48 Accumulation of the isoprostane 8-epi-prostaglandin F2a (8-epi-PGF 2a ) in cluster 2 was also notable as 8-epi-PGF 2a is a product of lipid peroxidation and arachidonic acid metabolism and an in vivo biomarker of oxidative stress 49,50 that increases in response to angiotensin II-and norepinephrine-induced hypertension. 51 TTTS clusters are defined by glycolytic metabolites, redox stress, and altered nitrogen metabolism To construct a coherent model of the metabolic changes observed in TTTS recipients, we next considered the metabolic remodeling that occurs in the failing postnatal heart. With increased oxygen availability after birth, the heart switches from using glucose and lactate as the primary energy substrates in fetal life to fatty acid oxidation for mitochondrial ATP production postnatally. Under conditions of stress, the postnatal heart reverts to fetal-like metabolic programs, 52,53 switching back to the primary utilization of glucose as fuel. This metabolic flexibility endows the postnatal heart with the ability to compensate and restore energy homeostasis in order to maintain function. The fetal heart relies on glucose at baseline, and therefore, iScience Article cannot activate the same ''switch'' in metabolism as the postnatal heart. The degree to which the baseline preference of the fetal heart for glucose affects metabolic flexibility and cardiac adaptation to stress is not known.
To evaluate the effects of cardiac stress on glycolysis, we first examined glycolytic metabolites ( Figures 2C-2F, S2A, and S2B) by cluster. Compared with controls, normalized amniotic fluid glucose levels were lower in TTTS clusters 1 and 2 ( Figure 2D), while pyruvate and acetyl-coenzyme A (CoA) levels were similar between all groups (Figures S2A and S2B). Lactate and alanine, the other products of anaerobic glycolysis, 54 were increased in TTTS cluster 2 compared with cluster 1 and control ( Figures 2E and 2F). Although stable isotope flux studies would be needed for definitive information regarding differential glucose metabolism, these results suggest a selective acceleration of glycolysis in TTTS cluster 2 cases.
To assess the redox state associated with these apparent shifts in energy substrate metabolism, we evaluated relative levels of glutathione in its reduced (GSH) and oxidized forms (GSSG). The GSH:GSSG ratio was significantly reduced in TTTS cluster 2 compared with cluster 1 and control amniotic fluid ( Figure 2G), indicating a more profound redox imbalance in cluster 2. Notably, these differences appeared to be driven by depletion of GSH ( Figures S2C and S2D). Cluster 2 also exhibited an increased lactate:pyruvate ratio ( Figure S2E). Elevations in the extracellular lactate:pyruvate ratio are correlated with reduced tissue oxygen tension 55,56 and reflective of the intracellular NADH:NAD+ ratio, 57,58 an indicator of redox potential. Accumulation of antioxidants, such as taurine and ergocalciferol (vitamin D2), was also observed in cluster 2 ( Figures S2F and S2G), potentially reflecting compensatory mechanisms. These data, along with the accumulation of amniotic fluid 8-epi-PGF 2a , are consistent with enhanced oxidative stress in TTTS cluster 2 vs. cluster 1.
Disruptions in cellular homeostasis and energy metabolism are tightly linked to disruptions in protein synthesis and degradation. 59 We therefore investigated amino acid and nitrogen metabolism by considering relative levels of urea cycle intermediates ( Figures 2H-2J and S2H-S2J). The urea cycle ultimately converts ammonia generated by amino acid metabolism to urea ( Figure S2H). Compared to controls and cluster 1, TTTS cluster 2 exhibited a significant depletion of argininosuccinate, urea, and ornithine ( Figures 2H-2J), while no difference was observed in arginine levels ( Figure S2I). Of the urea cycle intermediates, citrulline was uniquely increased in cluster 2 ( Figure S2J), suggesting impaired urea cycle flux. Alternatively, citrulline is produced during the conversion of arginine to nitric oxide ( Figure S2H); thus elevated citrulline levels might reflect an increased generation of nitric oxide, a pivotal mediator of vasodilation in the setting of fetal hypertension.
Collectively, the fluid metabolite pattern associated with TTTS cluster 2 suggests increased glycolytic activity in the setting of redox stress and reduced amino acid utilization and metabolism compared with cluster 1.

Perinatal outcomes associated with TTTS metabolic clusters
We sought to correlate the metabolomics results with clinical outcomes for TTTS clusters 1 and 2. Frequencies of perinatal death, early preterm birth (<34 weeks of gestation), and preterm premature rupture of membranes were similar between cluster 1 and 2 ( Figures 3A-3C; Table S5). Median gestational age at birth and the time interval between laser therapy and birth were also similar ( Figures 3D and 3E; Table S5). We further compared amniotic fluid protein and N-terminal pro-brain natriuretic peptide (NT-proBNP; released by cardiomyocytes in response to stretch 60 ) between clusters to assess potential subclinical differences in volume overload; these were similar between groups ( Figures 3F and 3G). An important consideration is that all TTTS cases were treated with fetoscopic laser ablation of placental vessels to effectively divide the fetal circulations and treat the underlying cause of the disease. Therefore, similarities in clinical outcomes might be attributed to the success of laser therapy rather than a lack of biologically relevant difference between clusters. Collectively, our data demonstrate significant metabolic remodeling in the setting of recipient volume overload and cardiovascular stress in TTTS ( Figure 3H) with still uncertain implications for future childhood health.

DISCUSSION
We leveraged severe TTTS as a model to explore potential metabolic changes in fetuses under cardiovascular stress from volume overload. Metabolomic analyses of amniotic fluid revealed accumulation of free iScience Article fatty acids, adrenal stress hormones, biomarkers of oxidative stress, and concomitant reduction in pyruvate, suggesting significant changes in energy metabolism in TTTS. Using PCA of the top differential metabolites, we uncovered two distinct clusters of TTTS (cluster 1 and cluster 2). These clusters were characterized by differences in markers of glycolysis, redox state, and amino acid metabolism.
TTTS cluster 2 showed more pronounced differences in amniotic fluid metabolites than TTTS cluster 1 when compared with control. The suggested reliance on glycolysis in TTTS cluster 2 is a hallmark of metabolic remodeling of the stressed postnatal heart. 61 Our data also suggest that fetuses, which rely on glycolysis at baseline, are able to further augment glycolytic flux in response to cardiovascular stress. This adaptive response was associated with an imbalance in the redox state and evidence of activation of compensatory mechanisms to restore homeostasis based on accumulation of NAD+ metabolites and antioxidants in the amniotic fluid. We speculate that multiple factors may contribute to the notable accumulation of amino acids and reduction in urea in cluster 2, including differences in the fetal response to: 1) hepatocyte injury and mitochondrial dysfunction due to right heart failure leading to urea cycle defects and 2) cellular stress response mechanisms that alter amino acid metabolism and reduce overall protein synthesis. 62,63 Although significant differences in signs of cardiac failure (NT-proBNP, mitral or tricuspid regurgitation) were not found, our small sample size limited our ability to detect differences in certain variables.  Table S5.
(D and E) Similar median gestational age at birth and interval between laser surgery and birth (latency in days). See also Table S5.
(F) Amniotic fluid protein levels stratified by cluster (data also shown in Figure 1C). iScience Article One insight gleaned from this study is the variability in fetal metabolic adaptation to cardiovascular stress. Whether this heterogeneity in TTTS represents fetuses at different points along the same spectrum (i.e., early vs. late or milder vs. more severe disease physiology) or innate differences in the ability of individual fetuses to adapt is uncertain. Although changes in amniotic fluid metabolites for cluster 2 were more prominent than cluster 1, cluster 2 changes are not necessarily indicative of worse disease. On the contrary, cluster 2 fetuses may be exhibiting greater flexibility and compensation than cluster 1 fetuses. In other words, cluster 2 may represent an adaptive response, whereas cluster 1 may be maladaptive. Nevertheless, we propose that recognition of disease subtypes is an important first step toward understanding variations in clinical presentation and factors associated with perinatal and longer-term adverse health outcomes. Although we did not find differences in perinatal outcomes between the two TTTS clusters, we propose that differences in intermediary metabolism might have more subtle clinical significance. For instance, the adequacy of the fetal adaptive response may be related to the gestational age at diagnosis and the unpredictable rate of disease progression. Long-term impacts on cardiovascular conditions cannot be ruled out either, due to our current limited collection of TTTS samples.
Our work integrates and builds upon previous studies that have examined metabolic remodeling in postnatal heart failure and molecular alterations in TTTS amniotic fluid. A recent study identified mitochondrial dysfunction, endoplasmic reticulum stress, and oxidative stress as defining features of early heart failure in patients and mice with hypoplastic left heart syndrome, 64 suggesting similarities between fetal and postnatal metabolic features associated with cardiovascular stress. Relatively few studies have reported results from human amniotic fluid metabolomics, and those studies have focused on preterm birth. [65][66][67][68][69] One study that used LC/MS-based metabolomics to assess TTTS recipient amniotic fluid investigated differences in metabolite profiles according to cardiac function and pre-vs. post-laser sampling. 70 Notably, prior to laser treatment, metabolites that negatively correlated with cardiac function (myocardial performance index) included acyl carnitines, fatty acids, ceramides, lipids, and hormones. Our data showing accumulation of similar metabolites in TTTS vs. control validate these observations. We did not, however, observe significant differences in these metabolites between TTTS clusters, perhaps reflecting the clinical similarity between samples in our cohort. Our use of an unsupervised bioinformatic approach to cluster cases revealed that stratification using clinical variables may not adequately reflect the underlying biology.
We and others have demonstrated increases in heart failure biomarkers 21,71 and steroid hormone levels 22 in TTTS recipient amniotic fluid compared with controls. We have also previously observed elongation of the umbilical artery in recipient twins associated with cardiac dysfunction, suggesting potential fetal vascular remodeling. 21 These data, together with the results from the present study, raise the question of the longterm impacts of in utero stress on TTTS survivors. This question requires further investigation.

Limitations of the study
In spite of the exciting new observations, there are several important caveats to consider. First, TTTS twins have a shared circulatory system. Therefore, donor twin physiology, such as upregulation of RAAS factors due to hypovolemia, 72,73 may affect recipient twin biology in a manner specific to TTTS fetuses. Secondly, we have interpreted our results through the lens of the recipient's cardiovascular biology which includes volume overload, hypertrophy, and heart failure. In this regard, we acknowledge that the recipient fluid has inputs from many sources, including organ systems other than the heart, such as the placental membranes and the maternal circulation ( Figure S3). We are also aware that amniotic fluid analysis provides only a snapshot of a complex and dynamic system of intracellular metabolic fluxes reflected in the extracellular environment. Complete amniotic fluid turnover occurs every 48 h; 74 thus, our ''snapshots'' represent metabolites that have recently accumulated. Third, we recognize the limitations of the sample size, the use of unmatched singleton controls (amniotic fluid from healthy twin pregnancies was unavailable for investigation), and the potential for confounding. Lastly, because every TTTS case was treated with laser ablation to separate the circulations of the donor and the recipient, the use of amniotic fluid metabolite signatures to predict perinatal outcomes is limited. In the future, a longitudinal cohort needs to be enrolled to collect amniotic fluid samples from two time points, pre-laser and at birth, to assess whether metabolic differences persist after definitive laser therapy.
In summary, our study presents an attempt to assess metabolic reprogramming in the human fetus experiencing cardiovascular stress. Notwithstanding some limitations, amniotic fluid provides an invaluable opportunity to assess molecular aspects of human fetal development and disease. This applies especially to ll OPEN ACCESS iScience 26, 107424, August 18, 2023 9 iScience Article diseases for which animal models are not available. Advancing knowledge of fetal plasticity and adaptability is critical for understanding how in utero insults impact future health. Clarifying the thresholds at which fetal adaptive responses become maladaptive and injurious is equally important. We suspect innate differences in fetal adaptability exist given the wide-ranging clinical presentations observed. Lastly, a more comprehensive picture of the molecular pathophysiology of TTTS is the first step toward changing our current inability to predict disease progression and complications after laser therapy. Future research efforts correlating fetal molecular studies with longer-term patient outcomes are needed to determine the extent to which fetal programming and the DOHaD hypothesis apply to TTTS and other fetal conditions.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

DECLARATION OF INTERESTS
Dr. Papanna reported receiving personal fees from UpToDate.

INCLUSION AND DIVERSITY
We support inclusive, diverse, and equitable conduct of research.

Study design
We set out to perform an exploratory study that compared second-trimester amniotic fluid from monochorionic-diamniotic twin pregnancies complicated by stage III TTTS and singleton controls. The study cohort comprised a convenience sample of cryopreserved specimens. Included TTTS cases were selected from a well-characterized subset with: 1) sonographic findings consistent with stage III disease: polyhydramnios in the recipient sac (maximum vertical pocket of fluid >8 cm); oligohydramnios in the donor sac (maximum vertical pocket <2 cm); bladder not visible in the donor; and abnormal Doppler evaluation in either twin (absent or reversed end-diastolic flow in the umbilical artery, reversed ductus venosus a-wave flow, or pulsatile umbilical vein flow), 19 and 2) complete clinical and outcomes data. Controls with normal genetic testing (karyotype or chromosomal microarray) and normal anatomical ultrasound were included. All TTTS patients underwent comprehensive ultrasound examination with Doppler studies prior to surgery for diagnosis and staging per protocol; however, genetic testing results were not available for all cases (testing is not routinely performed for non-anomalous twins with TTTS). Pregnancies with apparent genetic or structural abnormalities were excluded.

Amniotic fluid and clinical data collection
Amniotic fluid was previously collected and banked from pregnant individuals with monochorionic-diamniotic twins complicated by TTTS undergoing fetoscopic laser ablation of placental anastomoses at the UTHealth Houston Fetal Center at Children's Memorial Hermann Hospital. Samples were collected from the recipient twin sac immediately upon entry with the operative cannula and prior to placental laser ablation or amnioinfusion for improving the visualization. The amniotic fluid was centrifuged, and the supernatant was stored at À80 C for future use. Frozen genetic amniocentesis samples, discarded from further analyses, served as controls. For TTTS cases, demographics, clinical characteristics, and outcomes were abstracted from the Fetal Center research database which is maintained by trained research staff. For controls, available clinical variables were limited to maternal age, gestational age, indication for amniocentesis, and genetic testing results.

Untargeted metabolite profiling Metabolite extraction
Amniotic fluid metabolites were extracted by addition of 1 part amniotic fluid to 15 parts 70% acetonitrile in ddH2O (vol:vol). The mixture was briefly vortexed and then centrifuged for 5 min at 16,000 3 g to pellet precipitated proteins. The protein pellet was solubilized in 0.2M NaOH and quantified by DC Protein Assay (Bio-Rad). The volume of metabolite extract was normalized by amniotic protein content. An aliquot of the resulting extract (3 mL) was subjected to LC/MS untargeted metabolite profiling in positive and negative ion modes as described previously described, 76

Reversed-phase LC/MS
We performed reversed-phase LC/MS to confirm targeted hydrophobic metabolites that eluted early in the ANP column and showed a significant difference between groups. Extraction of lipids and fatty acids was carried out using cold methanol, methyl tert-butyl ether (MTBE), and water. Specifically, 10 mL amnionic fluid was added to 225 mL cold methanol, vortexed for 10 seconds, then 750 mL of cold MTBE was added.
The mixture was vortexed for 10 seconds and shaken for 6 minutes at 4 C. Phase separation was induced by adding 188 mL of LC/MS-grade water followed by centrifugation at 14,000 rpm for two minutes. The lipid phase was collected and dried under vacuum. Dried lipid extracts were normalized to protein content and resuspended in methanol/toluene (9:1, v/v) mixture for positive and negative detection by LC/MS.
Lipid extracts were also analyzed by LC/MS, using a platform comprised of an Agilent Model 1290 Infinity II liquid chromatography system coupled to an Agilent 6550 iFunnel time-of-flight MS analyzer. Lipids were separated by reversed phase chromatography on a ZORBAX EclipsePlus C18 (100 x 2.1 mm, 1.8 mm, Agilent Technologies). For positive ion mode, mobile phases consisted of: (A) 60:40 (v/v) acetonitrile: H2O containing 10 mM ammonium formate and 0.1% formic acid, and (B) 90:10 (v/v) isopropanol:acetonitrile containing 10mM ammonium formate and 0.1% formic acid. Column temperature was 60 C and sample injection volume was 2 mL. The following gradient was applied: 0 min, 15% B; 0-2.0 min, to 30% B; 2.0 to 2.5 min, to 48% B; 2.5-8.5 min, to 72% B; 8.5 to 11.5 min, to 99% B; 11.5 to 12 min, 99% B; 12.1 to 15 min, 15% B. The flow rate was 0.6 mL/min. LC conditions for negative ion mode are the same as positive mode, except that 10 mM ammonium acetate was replaced by 10 mM ammonium acetate.
The following mass spectrometer parameters were used for both positive and negative ion modes: drying gas temperature was set at 200 C with a flow rate of 14 L/min. Nebulizer pressure was at 35 psi. Sheath gas temperature was set at 350 C with a flow rate of 11 L/min. Capillary and nozzle voltage were at 3500v and 1000v, respectively. Mass spectra were acquired at 2 spectra/sec over the range of 100 to 1700 m/z.

Metabolite identification
Raw LC/MS data were analyzed using MassHunter Profinder 8.0 and MassProfiler Professional (MPP) 15.1 software (Agilent Technologies). To ascertain the identities of metabolites, LC/MS data were searched against an in-house annotated personal metabolite database created using MassHunter PCDL manager 8.0 (Agilent) based on monoisotopic neutral mass (<5 ppm mass accuracy) and chromatographic retention times of pure standards. A molecular formula generator (MFG) algorithm in MPP was used to generate and score empirical molecular formulae, based on a weighted consideration of monoisotopic mass accuracy, isotope abundance ratios, and spacing between isotope peaks. A tentative compound ID was assigned when the PCDL database and MFG scores concurred for a given candidate molecule. Tentatively assigned molecules were confirmed based on a match of LC retention times and/or MS/MS fragmentation spectra for pure molecular standards.

Amniotic fluid biochemical analysis
Amniotic fluid protein concentration was determined using the Pierce BCA protein assay (Thermo Fisher Scientific). Amniotic fluid NT-proBNP concentration was measured using the Alpco NT-Pro BNP ELISA (Fisher Scientific 04BI20852W) and normalized to amniotic fluid protein levels.

Metabolomics data analysis
Normalized concentrations for metabolites were first log 2 -transformed. R package limma 75 was then applied to fit a linear regression model to identify differential metabolites using empirical Bayes statistics. Differential metabolites were thus calculated based on a log 2 fold-change threshold of R 2 and FDR of % 0.001. Principal ll OPEN ACCESS iScience Article