A proteomic clock of human pregnancy

Early detection of maladaptive processes underlying pregnancy-related pathologies is desirable because it will enable targeted interventions ahead of clinical manifestations. The quantitative analysis of plasma proteins features prominently among molecular approaches used to detect deviations from normal pregnancy. However, derivation of proteomic signatures sufficiently predictive of pregnancy-related outcomes has been challenging. An important obstacle hindering such efforts were limitations in assay technology, which prevented the broad examination of the plasma proteome. The recent availability of a highly multiplexed platform affording the simultaneous measurement of 1310 plasma proteins opens the door for a more explorative approach. The major aim of this study was to examine whether analysis of plasma collected during gestation of term pregnancy would allow identifying a set of proteins that tightly track gestational age. Establishing precisely timed plasma proteomic changes during term pregnancy is a critical step in identifying deviations from regular patterns caused by fetal and maternal maladaptations. A second aim was to gain insight into functional attributes of identified proteins and link such attributes to relevant immunological changes. Pregnant women participated in this longitudinal study. In 2 subsequent sets of 21 (training cohort) and 10 (validation cohort) women, specific blood specimens were collected during the first (7–14 weeks), second (15–20 weeks), and third (24–32 weeks) trimesters and 6 weeks postpartum for analysis with a highly multiplexed aptamer-based platform. An elastic net algorithm was applied to infer a proteomic model predicting gestational age. A bootstrapping procedure and piecewise regression analysis was used to extract the minimum number of proteins required for predicting gestational age without compromising predictive power. Gene ontology analysis was applied to infer enrichment of molecular functions among proteins included in the proteomic model. Changes in abundance of proteins with such functions were linked to immune features predictive of gestational age at the time of sampling in pregnancies delivering at term. An independently validated model consisting of 74 proteins strongly predicted gestational age (P = 3.8 × 10–14, R = 0.97). The model could be reduced to 8 proteins without losing its predictive power (P = 1.7 × 10–3, R = 0.91). The 3 top ranked proteins were glypican 3, chorionic somatomammotrop