Coronavirus disease 2019 (COVID-19) poses a significant global health threat. As of October 2022, over 600 million individuals had been infected with SARS-CoV-2, resulting in nearly 6.5 million deaths. Growing evidence indicates that pregnant women face a heightened risk of adverse COVID-19 outcomes, including increased ICU admission, need for mechanical ventilation, and mortality compared to non-pregnant women. Furthermore, pregnant women with COVID-19 experience more obstetrical complications such as preeclampsia, preterm birth, and stillbirth. These adverse effects not only impact the mother but also negatively affect the short- and long-term well-being of the offspring, potentially causing respiratory distress syndrome, increased neonatal ICU admission, and impaired cognitive development. Understanding the biological mechanisms underlying this increased susceptibility is crucial. Numerous studies have explored the effects of SARS-CoV-2 infection on maternal, fetal/placental, and neonatal physiology, including immune responses. Researchers have characterized changes in systemic parameters such as cellular immune responses, virus-specific immunoglobulins, and inflammatory mediators in maternal and cord blood to profile maternal-fetal immune responses against SARS-CoV-2 infection. Comparative studies of pregnant and non-pregnant COVID-19 patients revealed that pregnant women are more likely to experience a cytokine storm, characterized by specific mediators, elevated neutrophil counts, and lymphopenia. Recent longitudinal and multi-omics approaches have identified specific processes contributing to COVID-19 progression and severity in the general population. However, a comprehensive understanding of the proteomic dysregulation distinguishing pregnant from non-pregnant COVID-19 patients remains lacking. Aptamer-based technologies have been used to profile the human proteome during normal pregnancy and its complications in maternal plasma and amniotic fluid; however, the SOMAScan platform's expanded version (4.1), capable of measuring over 7000 analytes, had not been previously used to study obstetrical pathology. This study aimed to utilize this high-throughput platform to characterize the plasma proteome of pregnant and non-pregnant women with COVID-19, focusing on identifying differentially affected proteins and understanding the distinct responses.

Prior research has established the increased risk of adverse outcomes for pregnant women with COVID-19, including severe illness and obstetrical complications. Studies have investigated the maternal and fetal immune responses to SARS-CoV-2, characterizing changes in systemic parameters like cellular immune responses, virus-specific immunoglobulins, and inflammatory mediators. These studies highlight the likelihood of cytokine storms in pregnant women with COVID-19, marked by specific mediators, increased neutrophil counts, and lymphopenia. Multi-omics approaches have been applied to understand COVID-19 progression and severity in the general population. However, a detailed proteomic analysis comparing pregnant and non-pregnant COVID-19 patients remained absent, necessitating the present study to address this gap using high-throughput proteomics. While aptamer-based technologies have been utilized in studying the proteome in normal pregnancy and complications, the more comprehensive SOMAScan v4.1 platform, capable of analyzing a wider array of protein targets, offered a unique opportunity for a more in-depth understanding of the specific proteomic changes in COVID-19.

This study employed a case-control design, enrolling 101 pregnant women and 93 non-pregnant individuals (men and women) of Hispanic ethnicity from Colombia. Pregnant participants were enrolled upon admission to the labor and delivery unit or during clinic visits. Non-pregnant participants were enrolled upon admission for various medical reasons. All participants underwent COVID-19 screening and testing. COVID-19 cases were categorized into asymptomatic, mild, moderate, severe, or critically ill groups using NIH classifications. Blood samples were collected in EDTA tubes, and plasma was separated via centrifugation. Plasma samples were stored at -80°C until proteomic analysis using the SOMAmer (SOMAScan v4.1) platform, which allowed for the quantification of over 7000 analytes corresponding to approximately 6596 unique protein targets. Data analysis involved log2 transformation of relative fluorescence units (RFU), principal component analysis (PCA) to identify major sources of variability in the proteome, and differential abundance analysis using the limma package in R. Gene ontology enrichment analysis using GOstats and MSigDB pathway analysis were performed to interpret differentially abundant proteins. Finally, random forest models were developed to assess the ability of the plasma proteome to discriminate COVID-19 patients from controls, employing leave-one-out cross-validation (LOOCV) and evaluating model accuracy via receiver operating characteristic (ROC) curves. Statistical significance was set at p<0.05. The study accounted for confounding variables such as age, sex, BMI, gestational age, and chronic hypertension in the analysis.

The study revealed several key findings: 1. **Dose-response relationship:** COVID-19 induced changes in the plasma proteome with a dose-response relationship to disease severity in both pregnant and non-pregnant individuals. The number of differentially abundant proteins increased with disease severity. 2. **Dampened response in pregnancy:** The magnitude of proteomic changes in response to COVID-19 was significantly attenuated during pregnancy, regardless of disease severity. This suggests a dampened immune response in pregnant women. 3. **Shared and unique proteomic changes:** Both shared and pregnancy-specific proteomic changes were identified. Pregnant women exhibited a tailored response potentially protecting the fetus from heightened inflammation, while non-pregnant individuals displayed a stronger response against infection. Shared changes were enriched for processes like cytokine storm, endothelial dysfunction, and angiogenesis. 4. **Accurate COVID-19 identification:** Machine learning models using the plasma proteome accurately identified COVID-19 patients, even those with asymptomatic or mild symptoms, demonstrating the potential use of proteomics for early disease detection. The AUC was 0.978 for the full analysis set, 0.974 for pregnant women and 0.985 for non-pregnant individuals. The accuracy to distinguish most severe cases (severe or critical COVID-19) from controls was higher (AUC=0.99) than the one obtained for discriminating between controls and moderate cases (AUC=0.94). Similarly high accuracy was obtained also for distinguishing asymptomatic or mild cases from uninfected controls (AUC=0.95). 5. **Dampened cytokine response:** Pregnant women exhibited a dampened systemic cytokine response to COVID-19 compared to non-pregnant individuals. While pro-inflammatory cytokines like IL-6, IL-1β, and IL-18 were elevated in both groups, the increases were less pronounced in pregnant women. 6. **Angiogenic and inflammatory changes:** Distinct angiogenic and inflammatory proteomic changes were observed between pregnant and non-pregnant women with COVID-19. Proteins related to angiogenesis and wound healing, alarmins, cytokines, and growth factors showed opposing dysregulation between the two groups. 7. **Predictive biomarkers:** Several proteins, such as ISG15, MX1, ZBP1, and IFNL1, were identified as significant predictors in the machine learning models for discriminating COVID-19 cases from controls across all severity levels.

The findings highlight a dampened systemic immune response in pregnant women with COVID-19, potentially due to physiological changes during pregnancy, like reversible thymic involution affecting T-cell development, or as a protective mechanism to prevent aberrant immune activation that could harm the pregnancy. The concept of selective immune tolerance in pregnancy is supported by studies showing that the suppression of key immune pathways, such as IFN, could underlie the higher risk of severe viral infection during pregnancy. The study's findings regarding the dampened cytokine response in pregnant women are consistent with previous research indicating a reduced release of IFN-β and IL-8 in response to SARS-CoV-2. The protective functions of the placenta, which exhibits strong antiviral properties, may partially offset the diminished maternal systemic response, preventing a cytokine storm harmful to the fetus while preventing viral transmission. The study's observation of a perturbed proteomic profile with enhanced release of cytokines and mediators associated with inflammation, endothelial dysfunction, and angiogenesis aligns with the understanding of severe COVID-19 characterized by cytokine storms. The partial overlap between COVID-19 and preeclampsia biomarker profiles suggests shared pathways influenced by placental inflammatory responses to maternal SARS-CoV-2 infection, even in asymptomatic cases. The potential for using proteomic profiles to classify and monitor COVID-19 outcomes is significant for patient management, especially during pregnancy. The identified predictive biomarkers (ISG15, MX1, ZBP1, and IFNL1) hold potential for development of diagnostic tools and therapeutic strategies.

This study provides the most comprehensive characterization to date of the plasma proteome in pregnant and non-pregnant individuals with COVID-19. The distinct immune modulation observed between pregnant and non-pregnant states offers insights into the pathogenesis of COVID-19 in pregnancy and explains the observed more severe outcomes in pregnant women. The unique proteomic profiles in pregnant women suggest that preeclampsia-like disease in this population may have a different pathogenesis from canonical preeclampsia pathways. Further research should explore the molecular mechanisms underlying SARS-CoV-2 induced maternal cytokine storms and their effects on offspring.

The study's limitations include the lack of information on the interval from infection or symptom onset to sample collection, which prevents evaluation of potential differences in the kinetics of the proteomic response between pregnant and non-pregnant patients. The absence of a truly healthy control group representing an unperturbed proteomic state is another limitation. While age and comorbidity matching was done for the non-pregnant controls, a truly healthy group would be ideal for comparison. Finally, the practical application of comprehensive proteomic profiling for routine COVID-19 diagnosis may be challenging; however, the identified proteomic signature could be used to generate risk scores for potential COVID-19 cases in studies with other primary endpoints.