Using our 4-vessel sampling method combined with a 5000-multiplex proteomic platform, we found evidence of significant exchange of a large number of proteins between the maternal and placental compartments in vivo in healthy pregnancy near term. Furthermore, we narrowed in on proteins that exhibited enhanced release from the placenta (placenta-specific released proteins), by testing venoarterial difference in the placenta as compared to the systemic circulation. These proteins are interesting both in terms of placental and maternal physiology during pregnancy and as biomarkers of placental function since their abundance in the systemic circulation may be influenced by placental release. To the best of our knowledge, no other research groups have been able to perform such a detailed analysis of placenta-derived proteins in human pregnancy.
Due to our robust study design that combines the 4-vessel cohort and the STORK cohort, we were able to create longitudinal protein curves across gestation for the identified placenta-specific released proteins (Additional file 4: Fig. S2). These curves may serve as a first step towards a reference for the expected physiological longitudinal pattern of placenta-derived protein across gestation. However, there is large variation in abundance across gestation in the systemic circulation between the pregnant women. These inter-individual differences in protein abundance may indicate that establishment of absolute cut-off values for normal abundance of placenta-derived proteins can be difficult. Instead, we hypothesize that development of reference curves for protein abundance for a selected group of placental proteins may serve better as biomarkers of placental function. Such reference curves would allow each pregnant woman to be her own control using repeated measures of proteins, tracking protein development across gestation.
Using a machine learning algorithm, elastic net with stability selection, we identified five placenta-derived proteins (CSH, BGN, GPC3, ITIH5, and GAA) that followed a tight chronological profile across gestation when measured in the antecubital vein. The resulting placental proteomic clock, inspired by the work of Aghaeepour et al. [7], represents a novel concept where deviations from the expected gestational pattern of these placenta-derived proteins may serve as indication of placental dysfunction. The identified five placental proteomic clock proteins showed promising predictive ability when combined in a linear model, evident by prediction performance above 90% in both the training and the validation cohort. The abundance of these five placenta-derived proteins increased linearly with gestational age, possibly reflecting placental maturation or size. However, there was no correlation between placental weight and placental release (VA difference) for these five proteins (data not shown). Furthermore, even if the observed linear increase in abundance of these proteins is a reflection of placental maturation or size, the placental clock proteins may still be of value to track placental well-being and function. Interestingly, GPC3, BGN, and CSH1/2 were identified as clock proteins in both Aghaeepours and the current study [7]. Furthermore, Romero et al. also found that the proteoglycans GPC3 and BGN had a 26.04 and 2.62-fold increase across gestation, respectively [35]. Inter-alpha-trypsin inhibitor heavy chain H5 is involved in the dynamics of the extracellular matrix and has been linked to pregnancy and uterine development in animal studies [36]. Lysosomal alpha-glucosidase is an enzyme that breaks down glycogen in the lysosome with known expression in the placenta [14].
Several proteins identified in the current study as released from the placenta into the maternal circulation have ample evidence in the literature to suggest placental origin. Placental growth factor (PGF) was confirmed to be significantly released to the mother, which is consistent with our pilot study [16] and a former study from our group using ELISA for protein determination [3]. Furthermore, our longitudinal data show that the abundance of PGF in the maternal circulation increases over gestation until gestational week 30, similar to previous studies [37]. We believe that the placenta-derived protein identified by Entrez Gene Symbol FLT1 is the soluble fms-like tyrosine kinase-1 (sFlt1), based on aptamer binding to the extracellular amino acid sequence that is similar between FLT1 and sFlt1. Placental release of the soluble form of FLT1 to the maternal circulation in preeclamptic pregnancies has previously been shown by our group [3, 38] and in healthy pregnancies by others [39]. These observations validate the approach used in the current study.
Compared to our pilot study that showed 34 proteins significantly released into the maternal circulation using 4-vessel samples on a 1310 multiplex SomaScan platform [16], we now identified 256 proteins as released to the mother on the 5000 multiplex SomaScan platform. Among the 256 placenta-derived proteins released in the current study, 30 of the 34 identified placenta-derived proteins in the pilot study were confirmed (Additional file 1: Fig. S3). Thus, we identified 226 novel placenta-derived proteins. Several of the proteins listed as released by the placenta have not previously been described in humans to best of our knowledge (for example FAAH2, Heat shock 70 kDa protein 1A (HSPA1A) and Heat shock protein beta-6 (HSPB6)).
One-hundred and one proteins were taken up by the placenta according to the current study, whereas nine proteins were taken up in the pilot study [16]. Three of the nine proteins in the pilot study, including VEGFA (vascular endothelial growth factor A, isoform 121), trefoil factor 1 (TFF1), and urokinase plasminogen activator surface receptor (PLAUR), were confirmed in the current study. Two of the nine proteins identified as taken up in the pilot study were not included in the analysis based on the 5000-plex platform. Thus, the current study identified 92 novel proteins as taken up by the placenta as compared to the pilot study.
Differences in proteins identified as released and taken up in the current study as compared to our pilot study originates both from differences on the platform (more proteins included in 5000-plex platform) and more statistical power in the current study due to larger sample size, as well as data preprocessing steps. In the current study, we used “median normalization” [30] on the venoarterial differences to adjust for movement of water across the placenta and other potential biases influencing the data, whereas a fixed global factor was used to correct for any water shifts in the pilot study as described by Holm et al. [16, 19].
It is well known that the maternal plasma proteome changes across gestation [7, 35]. When interpreting our findings, it is important to take into consideration the cross-sectional nature of the 4-vessel sampling. Thus, the proteins we have defined as being released or taken up by the placenta are a reflection of physiology at the time of the cesarean section near term gestation. Our longitudinal analysis of the antecubital vein levels of placenta-specific released proteins across gestation showed two clusters with distinct patterns of change across gestation. We hypothesize that proteins with high abundance in early pregnancy have the most impact on physiology at an early stage, whereas the proteins that increase in abundance towards term play a larger role at the end of pregnancy. Some changes in abundance may reflect placental mass or maturation. Among the placenta-specific released proteins, the Wnt regulation was prominent (Fig. 2). This finding is in accordance with data showing that Wnt signaling plays important roles in normal physiology and abnormal trophoblast function [40]. Interestingly, GO biological function labyrinthine layer morphogenesis was enriched. Rinkenberger and Werb stated “the labyrinth in the mouse and the floating chorionic villi in human are homologous structures, both characterized by extensive branching” [41]. Thus, this GO term is also used in humans, and our finding may indicate active morphogenesis in the human placenta close to term.
This study has some strengths and limitations. We have used the term “placenta” being well aware that the uterine vein, from which samples were taken, also drains non-placental tissues like decidua, myometrium, and chorion. Furthermore, we use the radial artery as a proxy for the uterine artery. The reason we use “placenta” is partly that the conceptual aim is placental functions, partly for the sake of simplicity. We acknowledge that proteins may enter the circulation in different ways, including exocytosis, or by random shedding or apoptosis of cells. Thus, some proteins found to be released by the placenta could be due to non-secretory processes. Within the placenta, there are a variety of cells that may contribute to the proteins released, including syncytiotrophoblasts, cytotrophoblasts, mesenchymal stem cells, fibroblasts, and immune cells. Twenty-three percent (3008 /13074) of proteins expressed in the placenta according to the Human Protein Atlas are measured by the SomaLogic 5000 multiplex platform (Additional file 1: Fig. S4A). Furthermore, 37.5% (108/288) proteins that have been shown to have elevated expression in the placenta as compared to other tissues in the Protein Atlas database are measured by the 5000-mulitplex platform. Additional file 1: Fig. S4B shows that most of the placenta-derived and placenta-specific proteins overlap with the 13074 proteins expressed in the placenta from Human Protein Atlas. Additionally, 19 of our 256 placenta-derived proteins overlap with the 108 elevated expressed proteins according to the Human Protein Atlas (data not shown). Mapping of placenta-derived proteins and placenta-specific released proteins in our study to genes with elevated expression in syncytiotrophoblasts, cytotrophoblasts, and extravillous trophoblasts according to the Proteins Atlas show considerable overlap (Fig. 6).
Traditional proteomics has shortcomings in terms of the need to remove highly abundant proteins as well as limited capacity to identify a broad range of proteins. The SOMAscan protein-binding technology offers the possibility to investigate protein expression on a platform with high sensitivity and dynamic range for a large number of specific proteins, although the quantification is in relative fluorescence levels. Absolute quantification of the proteins could have provided additional information, but the use of the relative protein levels is sufficient when comparing vessels. However, absolute quantification may improve the clinical relevance of longitudinal reference curves for placenta-derived proteins. The proteins have been selected to include human proteins associated with disease, and the aptamer technology therefore lacks the unbiased discovery approach character.
We performed a median normalization [30, 31] of venoarterial differences in the placenta (uterine vein-radial artery) based upon the assumption that most proteins are not different between the vein and artery. This assumption takes advantage of the power of the nearly 5000 proteins measured and adjusts for any movement of water across the placenta and other possible biases that may affect venoarterial differences (see methods). However, there is a risk that this method resulted in too many venoarterial protein differences set to zero, excluding biologically relevant proteins from our findings.
The SomaScan was performed on the antecubital vein samples in the 4-vessel cohort with the intention of using the venoarterial differences in the arm (antecubital vein vs radial artery) as a negative control of venoarterial differences in “the placenta” (radial artery vs uterine vein). We used this negative control to define proteins especially released from the placenta as “placenta-specific proteins.” Interestingly, we found that a large number of proteins had different levels in the antecubital vein and radial artery, possibly reflecting protein exchange in musculoskeletal tissues or systemic differences between veins and arteries. We are well aware that strictly all venoarterial differences in any maternal tissues or organs should have been tested as negative controls before concluding that proteins released by placenta are specific [3]. However, our method is, to the best of our knowledge, the only feasible one. Additionally, our definition of proteins released by the placenta as “placenta-specific” combined with a false discovery rate of 5% mean that our analysis followed strict criteria.