Sixty-six pregnant individuals enrolled in the Massachusetts General Hospital COVID-19 biorepository between March 2020 and April 2022 were included in the study. Maternal sera were collected at the time of acute SARS-CoV-2 infection in 26 unvaccinated individuals and 21 previously vaccinated individuals. Cord sera were collected and analyzed from individuals with acute SARS-CoV-2 infection if delivery occurred during the infectious period, i.e. less than or equal to 14 days from positive test in asymptomatic individuals or from onset of symptoms in symptomatic individuals. Maternal and cord sera from 19 healthy mother-cord dyads at term without a history of SARS-CoV-2 infection during pregnancy and who tested negative for SARS-CoV-2 upon admission to the Labor and Delivery unit were analyzed as controls. Participant enrollment occurred during a period when universal SARS-CoV-2 testing was performed on admission to the Labor and Delivery unit, thus the SARS-CoV-2 status of all participants was known on admission for delivery. All participants provided informed consent. Identification of eligible individuals, participant recruitment strategies and enrollment procedures have been described in previous publications (17, 32). A flow chart describing study participants is presented in the Supplementary ( Supplementary Figure S1 ).

Pregnant individuals were eligible for inclusion if they were diagnosed with SARS-CoV-2 infection or known to be negative for SARS-CoV-2 by nasopharyngeal swab RT-PCR. Maternal SARS-CoV-2 positivity was defined by a positive nasopharyngeal swab RT-PCR. Participants negative for SARS-CoV-2 on admission to Labor and Delivery were enrolled as a convenience sample, recruited on the same days as enrolled positive cases. Demographic and clinical outcomes data were abstracted from the electronic medical record using REDCap electronic data capture tools (33). COVID-19 disease severity was defined according to National Institutes of Health criteria (34). Additional information on inclusion and exclusion criteria are provided in the Supplementary ( Supplementary Figure S1 ).

Maternal and umbilical cord blood samples were obtained by venipuncture into serum separator tubes (BD). Blood was centrifuged at 1000g for 10 minutes, aliquoted and stored at -80°C. The concentrations of 20 cytokines were measured using the ProcartaPlex Human Inflammation Panel 20-plex (ThermoFisher) ( Supplementary Table S1 ). All standards and samples were run in duplicate. The standard curve for each analyte was determined using the ProcartaPlex Analysis App (ThermoFisher). Standards with less than 80% recovery or more than 120% recovery were omitted, consistent with recommended thresholds. Analytes that were undetectable in more than 30% of samples were excluded from further analysis. For the remaining analytes, the limit of detection (LOD) of the analyte divided by √2 was substituted for levels below the LOD. For values above the upper limit of quantitation, the upper limit of quantitation was directly substituted. Outliers (defined as Z-score greater than 2.5) were excluded. All concentrations are reported as pg/mL.

Univariate differences in maternal analytes were assessed by Kruskal-Wallis test with Dunn’s posthoc testing of the three study groups: 1) participants with unvaccinated SARS-CoV-2 infection, 2) participants with vaccinated “breakthrough” SARS-CoV-2 infection, and 3) SARS-CoV-2 negative control groups. Analyte levels are reported as median [Interquartile Range (IQR)]. Spearman correlations between analytes within each group were determined and p-values adjusted for multiple comparisons using the Benjamini-Hochberg method. Spearman correlations of maternal to cord levels of each analyte were assessed in dyads who delivered during acute SARS-CoV-2 infection and in SARS-CoV-2 negative dyads. For multivariate analyses, maternal and cord analyte concentrations were log₂-transformed to improve normality. The function prcomp in the R statistical language and environment (www.r-project.org, version 4.0.2) was used to calculate principal components (PC). The top three PC were tested for associations with sample type (maternal or cord) using linear regression modeling. This analysis also included vaccination status, infant sex, and exposure to labor as possible covariates, based on a priori assumption that these factors may impact cytokine/chemokine levels in the maternal-fetal dyad (35, 36). All statistical tests were two-tailed, and significance was inferred based on p<0.05. Statistical analyses were performed in R (version 4.0.2).

Sixty-six pregnant individuals were included, 26 with acute SARS-CoV-2 infection who were unvaccinated, 21 with acute SARS-CoV-2 who had previously received a primary COVID-19 vaccine series, and 19 uninfected pregnant controls ( Supplementary Figure S1 ). No placentas examined had histopathologic or clinical evidence of SARS-CoV-2 placental infection and were thus not examined for SARS-CoV-2 viral RNA. Of the 26 unvaccinated individuals with SARS-CoV-2 infection, 4 (15%) had asymptomatic disease, 10 (38%) had mild disease, 2 (8%) had moderate disease, 9 (35%) had severe disease and 1 (4%) had critical disease. Of the 21 vaccinated individuals, 6 (28%) had asymptomatic disease, 12 (57%) had mild disease, 3 (14%) had moderate disease, and no individuals had severe or critical disease. Of the 21 vaccinated individuals, 14 received the BNT162b2 primary series, 6 received the mRNA-1273 primary series, and 1 individual received one dose of Ad26.COV2.S and one dose of mRNA-1273; 10 of the 21 vaccinated individuals had received a third mRNA vaccine dose prior to infection. There were no differences in maternal age, parity, race, ethnicity, body mass index (BMI), hypertensive disorders of pregnancy, neonatal sex, or labor between groups ( Table 1 ). Placental pathological examinations were performed by clinical indication per institutional protocol and reported in the birthing individual’s medical record. No placentas examined had histopathologic evidence of SARS-CoV-2 infection, although SARS-CoV-2 RNA testing was not routinely performed during the study period due to prior observations that placental infection with SARS-CoV-2 is uncommon (17). Placental pathology findings were similar between groups ( Supplementary Table S3 ). No newborns of the individuals included in this cohort tested positive for SARS-CoV-2.

Maternal sera were collected a median of 4 days [interquartile range (IQR): 1-11 days] from the onset of acute infection. Significantly higher maternal serum levels of TNFα (p=0.03) and soluble CD62P (p=0.03) were observed in vaccinated individuals with SARS-CoV-2 infection compared with SARS-CoV-2 infected but unvaccinated individuals ( Figure 1A ). Adjusting for sample collection timing in days using linear regression modeling had no effect on the significant findings of differences in CD62P and TNFα levels between vaccinated and unvaccinated groups [TNFα: effect of group, p=0.029, effect of collection time, p=0.46; CD62P: effect of group, p=0.023, effect of collection timing, p=0.81]). Maternal serum levels of IFNγ, IL4, and IL17A appeared higher in vaccinated individuals with SARS-CoV-2 infection compared to controls, however these differences did not reach statistical significance. No differences in other maternal serum cytokines/chemokines evaluated were observed between groups ( Supplementary Figure S2 ).

To further assess the maternal inflammatory response to SARS-CoV-2 we next calculated Spearman correlations between serum analytes, in SARS-CoV-2-infected individuals with and without prior vaccination ( Figures 1B, C ). After adjustment for multiple comparisons, 16 significant positive correlations were observed between analytes in unvaccinated infection and 12 positive correlations in vaccinated infection, compared with only 1 positive correlation in negative controls ( Supplementary Figure S3 ), as anticipated. Network plots depicting relationships between correlated analytes show tight correlations between interferons (IFNα and IFNγ) that are critical to the innate antiviral immune response and analytes associated with Th1 responses (IL12p70, IL1β, IL17A) in individuals without prior vaccination ( Figure 1B ). In contrast, the maternal cytokine response to vaccinated infection is not centered on interferons, but rather IL4, strongly associated with Th2 responses ( Figure 1C ). These findings suggest that SARS-CoV-2 infection occurring after vaccination may be more strongly associated with Th2-mediated responses in pregnant individuals.

Prior work suggests that a fetal immune response may be observed in the setting of maternal infection with SARS-CoV-2 that is linked to the degree of maternal immune activation and clinical disease severity (15, 21, 37) In contrast to what we observed in the maternal circulation in the setting of SARS-CoV-2 infection, univariate analyses of cord analytes demonstrated no significant differences in cord serum analyte levels between individuals with and without SARS-CoV-2 infection at the time of delivery, regardless of prior COVID-19 vaccination history, and negative controls, with the exception of MIP1α which was significantly lower in unvaccinated infection compared to negative controls ( Supplementary Figure S4 ). Additionally, no significant correlations were noted between cord analytes within groups. These data support the lack of a specific, coordinated fetal immune response to maternal SARS-CoV-2 infection in either vaccinated or unvaccinated individuals.

To further assess the potential connection between maternal and cord cytokine responses to maternal SARS-CoV-2 infection, we next assessed Spearman correlations of analytes between paired maternal and cord samples obtained at delivery in individuals with acute SARS-CoV-2 infection (delivery occurring within 14 days of SARS-CoV-2 diagnosis), and uninfected controls. The results of these analyses are depicted in Figure 2 . We identified that while maternal and cord levels of IL13, TNFα, and MCP1 were significantly positively correlated to each other in acute SARS-CoV-2 cases at delivery, these analytes were also correlated in healthy, SARS-CoV-2 negative maternal: cord dyads. Maternal and cord levels of CD62E and CD62P – leukocyte adhesion molecules that are expressed by activated endothelium and known to be elevated in some inflammatory pathologies – were positively correlated in SARS-CoV-2 cases but not controls, indicating that maternal endothelial activation may be related to fetal endothelial activation in SARS-CoV-2 cases, an effect not observed in healthy controls (38–40).

To build an integrated view of the serum inflammatory profile of the maternal-fetal dyad in the setting of acute maternal SARS-CoV-2 infection and to simultaneously handle the multiple correlated analytes, we used Principal Component Analysis (PCA) to resolve covariation between analytes into independent components ( Figure 3 ). Separation of maternal and cord sera was observed along the first two PCs, which accounted for 43.1% and 14.8% of sample variation, respectively, indicative of overall distinct immune profiles between maternal and cord samples at the time of acute maternal infection ( Figure 3A ). The contribution of individual cytokines to the first 3 principal components (PC1-3) of the model are shown in Figure 3B . The top 5 cytokines contributing to the construction of PC1 are IL1β, IL17A, IL13, TNFα, and CD62E, whereas IP10 (CXCL10), IL4, IL1α, IFNα and MCP1 define PC2 ( Figure 3C ). Compared to maternal samples, cord samples clustered positively in PC1 – defined by increases in markers of both classic pro-inflammatory (IL1β, IL17A, TNFα) and anti-inflammatory (IL13) responses. Maternal samples were positive in PC2, defined by significant contribution from IP10 (CXCL10) and IL4, factors that have been previously identified as serum biomarkers of COVID-19 severity (41, 42).

Linear regression models of the first 3 PCs, which describe 69% of the variation in the dataset, were built to assess the contribution of clinical covariates to sample variation: sample source (maternal or cord), history of vaccination prior to infection, infant sex, and presence of labor. PC1 and PC2 were positively associated with cord sample (p<0.001 and p<0.01, respectively); presence of labor was associated with PC3 (p<0.05), driven by chemokines IL8, MIP1α and MIP1β. Prior vaccination and infant sex were not correlated with the first 3 PCs of the model. Timing of infection to sample collection had no impact on PC1 or 2 but it did have a small but significant impact on PC3 (p<0.05). Taken together these findings suggest that maternal and cord cytokine/chemokine profiles in the setting of SARS-CoV-2 are distinct and not significantly impacted by maternal vaccination status or infant sex; exposure to labor had a minor impact on chemokines represented by PC3, which itself represented only 11% of variation in the dataset. Details of models are available in Supplementary Table S2 .