Pregnancy-specific responses to COVID-19 are revealed by high-throughput proteomics of human plasma

Pregnant women are at greater risk of adverse outcomes, including mortality, as well as obstetrical complications resulting from COVID-19. However, pregnancy-specific changes that underlie such worsened outcomes remain unclear. Herein, we profiled the plasma proteome of pregnant and non-pregnant COVID-19 patients and controls and showed alterations that display a dose-response relationship with disease severity; yet, such proteomic perturbations are dampened during pregnancy. In both pregnant and non-pregnant state, the proteome response induced by COVID-19 showed enrichment of mediators implicated in cytokine storm, endothelial dysfunction, and angiogenesis. Shared and pregnancy-specific proteomic changes were identified: pregnant women display a tailored response that may protect the conceptus from heightened inflammation, while non-pregnant individuals display a stronger response to repel infection. Furthermore, the plasma proteome can accurately identify COVID-19 patients, even when asymptomatic or with mild symptoms. This study represents the most comprehensive characterization of the plasma proteome of pregnant and non-pregnant COVID-19 patients.

, which was considered in 116 the data analysis. Among the pregnant COVID-19 cases, 6 (8%) were asymptomatic, 20 (28%) 117 were mild, 13 (18%) were moderate, 12 (17%) were severe, and 21 (29%) were critically ill 118 according to NIH classification 53 . 119 Non-pregnant individuals: Plasma samples were also collected from 93 non-pregnant 120 individuals, which included 52 COVID-19 cases and 41 controls ( Fig. 1a and Table 1). Among 121 the non-pregnant COVID-19 cases, 1 (2%) was mild, 4 (8%) were moderate, 12 (23%) were 122 severe, and 35 (67%) were critically ill.  Fig. 1d). Yet, the host response to COVID-19 represented the primary 139 source of variability in the proteome, as PC1 and PC3 (PC1, 27% of variance explained; PC3, 140 6% of variance explained) were significantly different between COVID-19 cases and controls, 141 regardless of pregnancy status (p < 0.001 for both, Fig. 1e). The proteomic changes with 142 COVID-19 were larger for non-pregnant than for pregnant women based on both PC1 and PC3 143 (interaction p < 0.005) (Fig. 1e), which is partly explained by the greater proportions of severe 144 and critically ill cases in the non-pregnant than in the pregnant population. Moreover, we 145 observed a dose-response relationship between PC3 and disease severity, regardless of 146 pregnancy status (p < 0.001 for both linear and quadratic trends, Fig. 1f). Together, these data 147 provide an overview of the plasma proteome in pregnant and non-pregnant individuals infected 148 with SARS-CoV-2, and suggest dramatic changes with infection in a dose-response relationship 149 with disease severity. In addition, these data hint that the host response to SARS-CoV-2 includes 150 shared and unique processes between pregnant and non-pregnant individuals, which we further 151 explore below.

153
The plasma proteome response to COVID-19 follows a dose-response relationship with 154 disease severity in pregnant and non-pregnant individuals, yet such a response is 155 dampened in pregnancy 156 Pregnant women have been reported to display heightened susceptibility to severe 157 COVID-19 4-6 . Therefore, we first explored the differential effects of COVID-19 on the maternal 158 proteome compared to control pregnancies according to disease severity (Fig. 2a). When 159 comparing pregnant COVID-19 cases to controls after adjustment for maternal age, BMI, and 160 gestational age at sampling, we identified 68, 81, 242, 144, and 1072 differentially abundant 161 proteins in asymptomatic, mild, moderate, severe, and critically ill cases, respectively ( Fig. 2b-f). 162 Given that both disease severity and sample size may affect the number of differentially 163 abundant proteins in specific groups, we next used the protein changes between critically ill 164 patients and controls (1072 proteins) as a reference to compare with the changes observed in the 165 less severe COVID-19 groups (Fig. 2g). The log2-transformed fold change of protein abundance 166 between COVID-19 subgroups (i.e., asymptomatic, mild, moderate, and severe) and controls 167 were more attenuated than those between critically ill patients and controls (regression slopes < 168 1.0) (Fig. 2g). Yet, the magnitude of correlation and the correlation slope followed a dose-169 response relationship with disease severity, and even asymptomatic patients showed plasma 170 proteomic changes that were significantly correlated to those observed in critically ill patients (r  Control; r = 0.88 for Severe vs. Control; p < 0.001 for all) (Fig. 2g). We then performed the same analysis of differential protein abundance in non-pregnant 174 patients (Fig. 3a), and identified 21, 1961, and 2966 differentially abundant proteins in moderate, 175 severe and critically ill cases, respectively ( Fig. 3b-d), after adjusting for relevant covariates. 176 Similar to the analysis in pregnant women, the log2-transformed fold changes of protein 177 abundance between COVID-19 subgroups and controls were more attenuated than those found 178 between critically ill cases and controls, and followed a dose response with disease severity (r = 179 0.84 for Moderate vs. Controls; r = 0.94 for Severe vs. Controls; p < 0.001 for both) (Fig. 3e).

180
To contrast the magnitude of COVID-19-driven changes in the proteome between 181 pregnant and non-pregnant patients, we then performed correlation analysis based on a core set 182 of 486 proteins with significant and consistent changes in both pregnant and non-pregnant 183 patients (see more details below) (Fig. 4a). By comparing the magnitude of changes between the 184 pregnant and non-pregnant groups, we showed that the magnitude of changes for this set of core 185 proteins were diminished during pregnancy for the same disease severity group, as indicated by    195 We then sought to further unravel pregnancy-driven differences in the plasma proteomic 196 response to COVID-19 as well as changes that are shared between pregnant and non-pregnant 197 states. First, we identified all proteins that were differentially abundant with COVID-19, which 198 resulted in 708 differentially abundant proteins for pregnant women ( Fig. 5a Table 2).

202
Next, we explored the biological processes that were enriched among the entire set of were characterized by cell adhesion and immune responses as well as response to wounding and 212 blood coagulation (Fig. 5f).

213
In addition to biological processes, we also evaluated the enrichment of pathways and 214 gene sets derived from the C2 collection of the MSigDB database (Fig. 5g). Similar to biological  Table 6). 220 Together, these data further demonstrate that, although there is a set of common 221 responses to COVID-19 in both pregnant and non-pregnant state, pregnancy-specific changes  to angiogenesis and wound healing as well as alarmins, cytokines, and growth factors (Table 2).

233
Proteins that decreased with COVID-19 in pregnancy but were increased in non-pregnant cases 234 included vascular endothelial growth factor receptor 1 (VEGF-sR1 or sFLT1) and 235 angiotensinogen (AGT); yet, this could potentially be explained by their already elevated 236 baseline among pregnant patients ( Fig. 6b&c and Table 2). Consistent with these findings, 237 proteins that underwent pregnancy-specific regulation with COVID-19 were enriched for 238 biological processes and pathways related to vasodilation, angiogenesis, and regulation of 239 inflammatory response (Supplementary Table 7). A previous report indicated that COVID-19 240 during pregnancy is characterized by a profile of proteomic factors that is distinct from but 241 overlaps with that observed in preeclampsia 7 , an obstetric syndrome characterized by 242 intravascular inflammation 54 . Therefore, we further evaluated changes in angiogenic or 243 endothelial factors between pregnant and non-pregnant COVID-19 patients. Several factors such 244 as soluble TNF receptor II (TNFRSF1B) and von Willebrand factor (VWF) were found to 245 increase with COVID-19 regardless of pregnancy status (Fig. 6d&e). Notably, neutrophil  slight tendency to increase during pregnancy (Fig. 7e&f). The alarmin IL-1α was found to be 265 downregulated only in pregnant COVID-19 cases, although a tendency towards the same 266 reduction was observed in non-pregnant patients (Fig. 7g). By contrast, IFNγ was reduced with 267 COVID-19 in non-pregnant individuals but not pregnant patients (Fig. 7h). The anti-268 inflammatory cytokine IL-10 was downregulated in pregnant and non-pregnant COVID-19 cases 269 (Fig. 7i), whereas TGFβ1 was upregulated in both groups (Fig. 7j). Several chemokines were 270 also found to exhibit differential regulation with COVID-19 in the pregnant and non-pregnant   In this study, we utilized the SOMAScan v4.1 platform to profile over 7,000 protein 303 targets in the peripheral blood of pregnant women and non-pregnant individuals diagnosed with 304 COVID-19, and found that this disease drives changes in their plasma proteomes in a dose-305 response relation with disease severity. Importantly, we showed that the response to COVID-19 306 is dampened during pregnancy, regardless of disease severity. Distinct and overlapping  Finally, we utilized machine learning to demonstrate that the plasma proteome can be used to 314 discriminate COVID-19 patients from controls, even those who were asymptomatic or had mild shown that monocytes from pregnant women appear to undergo accelerated transition and 353 activation in response to SARS-CoV-2 exposure 30 , which is in line with a previous report 39 . 354 Notably, we found that the cytokine profile of peripheral leukocytes was also impacted by  born to infected mothers 35 . Therefore, it is imperative to follow and evaluate these infants for 426 eventual adverse outcomes, as has been suggested by recent evidence demonstrating 427 neurodevelopmental sequelae at one year of life in children exposed to SARS-CoV-2 in utero 18 .

428
The establishment of biomarkers that allow for the classification and monitoring of 429 COVID-19 outcomes is essential to guide patient management, particularly during pregnancy. In

488
After the labeling reaction, the beads were exposed to an anionic competitor solution that 489 prevents non-specific interactions from reforming after disruption. Pure cognate-SOMAmer 490 complexes and unbound (free) SOMAmer reagents were then released from the streptavidin 491 beads using ultraviolet light that cleaves the photo-cleavable linker used to quantitate proteins.

492
The photo-cleavage eluate, which contains all SOMAmer reagents (some bound to a biotin-493 labeled protein and some free), was separated from the beads and then incubated with a second  The proteomic data preprocessing, including an adaptive normalization by maximum 524 likelihood (ANML) step and a calibration step, were performed by SomaLogic, Inc. The goal of 525 these steps was to make data comparable across samples by calculating plate-specific and 526 analyte-specific scale factors. After log (base 2) transformation, data were compared between 527 pooled COVID-19 cases and controls or compared separately between each disease severity 528 group against controls. When analyzing data from pregnant women, maternal age, body mass 529 index (BMI), and linear and quadratic terms of gestational age were included as co-variates.