Maternal Fc-mediated non-neutralizing antibody responses correlate with protection against congenital human cytomegalovirus infection

Baseline characteristics of HCMV transmitting and non-transmitting mother-infant dyads

Our study included sera from 81 mother-infant dyads recruited as donors to the Carolinas Cord Blood Bank (CCBB), a U.S.-based public cord blood bank with 130,000 collections and 45,000 banked donors to-date (Supplementary Figure 1). Donor mothers were screened for total anti-HCMV antibodies and if present, cord blood plasma was tested by qPCR for replicating HCMV. Cord blood units that contained replicating HCMV were excluded from banking for cord blood transplantation and were retained for research. Congenital HCMV infection (cCMV) was defined based on the presence of HCMV viremia in the donated cord blood plasma. Forty-one dyads with congenital HCMV infection (“HCMV transmitting”) were matched to forty dyads with HCMV IgG seropositive mothers that gave birth to cCMV uninfected infants (“HCMV non-transmitting”). Matching criteria included infant sex, infant race, maternal age (+/- 3 years), and delivery year (+/- 3 years). Baseline demographic and clinical characteristics were comparable between HCMV transmitting and non-transmitting dyads, though transmitting dyads had a non-significantly higher rate of Cesarean section (56% vs. 40%, Fisher’s exact p = 0.22, Table 1). The majority of identified cCMV cases were male (58.5%), higher than the proportion of males in the cord blood bank overall (51.4%), which may suggest an increased susceptibility to infection for male fetuses, though studies on sex differences in cCMV infection are lacking (40).

To assess whether mothers had primary or nonprimary HCMV infection during pregnancy, we measured anti-HCMV IgG avidity and IgM responses (41). Whole virion HCMV IgG avidity indexes were comparable between HCMV transmitting (median [range]: 77.6 [46-109]) and non-transmitting (median [range]: 79.5 [65-100]) mothers. However, 11/41 (26.8%) of HCMV transmitting mothers had detectable HCMV-specific IgM responses compared to only 2/40 (5%) of non-transmitting mothers (Table 1). Thus, while the majority of mothers in both groups had high avidity IgG responses and no detectable IgM responses at the delivery timepoint, a higher proportion of mothers in the transmitting group had detectable HCMV-specific IgM, which could indicate a higher rate of primary infection or reinfection during pregnancy in the transmitting group (41, 42). To further assess comparability between groups, we quantified HCMV viral loads in maternal sera and found that a similar proportion of transmitting (26.8%) and non-transmitting (37.5%) mothers had detectable low-level HCMV DNAemia (median [range] in copies/mL: 346 [256-1052] in transmitting vs. 365 [260-719] in non-transmitting; Table 1), consistent with previous observations in healthy HCMV seropositive women (43).

Maternal and cord blood sera from HCMV transmitting dyads have high levels of anti-HCMV IgG

We previously observed that HCMV-specific IgG transfer across the placenta may be compromised in cCMV infection, a phenomenon primarily related to maternal hypergammaglobulinemia in HCMV transmitting women (44). Thus, we first measured placental HCMV-specific IgG transfer in transmitting and non-transmitting mother-infant dyads. We quantified IgG binding against 3 distinct HCMV strains including TB40E (an endotheliotropic strain expressing pentamer), AD169r (a lab-adapted strain with repaired pentamer expression), and Toledo (a low-passage clinical isolate lacking pentamer). Within mother-infant pairs, HCMV-specific IgG binding was lower in cord blood compared to maternal sera in transmitting but not non-transmitting dyads (Figure 1A). However, placental IgG transfer efficiency was only slightly lower in transmitting compared to non-transmitting dyads (Figure 1B). HCMV transmitting mothers also had higher total IgG levels (i.e., hypergammaglobulinemia), which negatively correlated with HCMV-specific IgG transfer efficiency (Supplementary Figure 2A-B). Despite slightly decreased placental IgG transfer ratios, whole virion HCMV-specific IgG titers were comparable in the cord blood of infected and uninfected infants (Figure 1C). In contrast, HCMV glycoprotein-specific IgG titers were significantly elevated (2.5-10 fold) in transmitting compared to non-transmitting dyads (Figure 1D-E). When adjusting for total IgG levels, non-transmitting women demonstrated higher relative binding to HCMV virions and cell-associated native gB, while IgG binding against pentamer, gHgLgO, and gHgL remained significantly higher in transmitting women (Supplementary Figure 2C-D).

• Download figure
• Open in new tab

Figure 1. Maternal and cord blood sera from HCMV transmitting dyads have high levels of HCMV-specific IgG despite moderately reduced transplacental IgG transfer compared to non- transmitting dyads.

HCMV-specific IgG levels against HCMV strains TB40E, AD169r, and Toledo were measured using whole virion enzyme-linked immunosorbent assay (ELISA). Cell-associated gB IgG binding was quantified using a flow-based assay with HEK293T cells transfected with full-length gB and HCMV glycoprotein-specific IgG binding was measured using a Luminex-based binding antibody multiplex assay. IgG binding responses in maternal (M) and cord blood (CB) sera were compared between and within HCMV transmitting (red circles, n = 41) and non-transmitting (blue squares, n = 40) mother-infant dyads. (A) Whole virion HCMV-specific IgG levels in paired maternal and cord blood sera. (B) Transplacental transfer efficiency of HCMV-specific IgG calculated as (cord blood HCMV-specific IgG concentration)/(maternal HCMV-specific IgG concentration)x100%. (C) Whole virion HCMV-specific IgG levels in transmitting versus non-transmitting dyads. (D) IgG binding to cell-associated gB via flow cytometry. (E) Mean fluorescence intensity (MFI) proportional to IgG binding to HCMV glycoproteins. gB ecto = gB ectodomain. Dotted lines indicate the lower limit of detection. Black bars denote median. FDR-corrected P values reported for Mann-Whitney U test or Wilcoxon signed-rank test. * P < 0.05, ** P < 0.01, *** P < 0.001.

Low avidity HCMV-specific IgG is enriched in the cord blood of cCMV-infected infants

Transplacental transfer of low-avidity HCMV-specific IgG complexed to HCMV virions has been posited to increase HCMV transmission risk (45) and we previously observed that low avidity HCMV-specific IgG may be preferentially transferred across the placenta in HCMV transmitting pregnancies (44). Thus, we next measured HCMV-specific IgG avidity in paired maternal and cord blood samples. Maternal sera from non-transmitting pairs had higher avidity IgG binding to AD169r and Toledo strains, but not TB40E, compared to transmitting dyads (Figure 2A). Within mother-infant pairs, whole virion HCMV IgG binding avidity was significantly lower in the cord blood compared to maternal sera in the transmitting but not the non-transmitting groups (Figure 2B). HCMV glycoprotein-specific IgG avidity was also lower in the cord blood of infected compared to uninfected infants (Figure 2C), though no differences in glycoprotein-specific IgG avidity were observed within paired sera from mother-infant dyads. In a sensitivity analysis excluding mothers with HCMV-specific IgM responses as a surrogate biomarker for recent maternal primary infection or reinfection, differences in whole virion HCMV IgG binding avidity persisted but most differences in HCMV glycoprotein-specific IgG binding avidity were no longer statistically significant (Supplementary Figure 3A-C).

• Download figure
• Open in new tab

Figure 2. Low avidity HCMV-specific IgG is enriched in the cord blood of cCMV-infected infants.

HCMV-specific IgG binding avidity against HCMV strains TB40E, AD169r, and Toledo were measured using whole virion enzyme-linked immunosorbent assay (ELISA) with an additional dissociation step using urea and relative avidity index (RAI) was calculated as (OD with urea/OD without urea)x100%. HCMV glycoprotein-specific IgG binding avidity was measured using a Luminex-based binding antibody multiplex assay with an additional dissociation step with sodium citrate and RAI was calculated as (MFI with sodium citrate/MFI with 1X PBS)x100%. IgG binding avidity in maternal (M) and cord blood (CB) sera were compared between and within HCMV transmitting (red circles, n = 41) and non-transmitting (blue squares, n = 40) mother-infant dyads. (A) Whole virion HCMV-specific IgG binding avidities in transmitting versus non-transmitting dyads. (B) Whole virion HCMV-specific IgG binding avidities in paired maternal and cord blood sera. (C) HCMV glycoprotein-specific IgG binding avidities in transmitting versus non-transmitting dyads. gB ecto = gB ectodomain. Black bars denote median. FDR- corrected P values reported for Mann-Whitney U test or Wilcoxon signed-rank test. * P < 0.05, ** P < 0.01, *** P < 0.001.

Neutralizing and non-neutralizing antibody responses differ in HCMV transmitting and non- transmitting dyads

In addition to measuring HCMV-specific IgG binding magnitude and avidity, we quantified anti-viral functions of maternal and cord blood sera from transmitting and non-transmitting dyads. Unexpectedly, HCMV neutralizing titers were approximately 1.5-4 fold higher in transmitting versus non-transmitting dyads across HCMV strains and cell types (Figure 3A-D). Within mother-infant dyads, neutralizing titers were similar in the maternal compared to cord blood sera in both groups. In contrast, ADCP, an Fc-mediated non-neutralizing antibody response, against HCMV was higher in non-transmitting compared to transmitting mothers (Figure 3E). This difference was statistically significant for the Toledo HCMV strain (p = 0.0057) and remained significant after correction for multiple comparisons (p

• Download figure
• Open in new tab

Figure 3. Higher non-neutralizing HCMV-specific ADCP responses and lower neutralizing responses in HCMV non-transmitting compared to transmitting mother-infant dyads.

Functional anti-viral antibody responses in maternal (M) and cord blood (CB) sera were compared between and within HCMV transmitting (red circles, n = 41) and non-transmitting (blue squares, n = 40) mother-infant dyads. Neutralization was measured by HCMV intermediate early gene 1 (IE1) staining and antibody titers were calculated as the inhibitory dilution 50 (ID50), equivalent to the sera dilution that inhibited 50% of the max infection in virus only wells. Neutralization titers against (A) HCMV Toledo strain in fibroblasts (HFFs) and (B-D) HCMV AD169r strain in (B) fibroblasts (HFFs), (C) epithelial cells (ARPEs) and (D) differentiated macrophages (THP-1s treated with PMA). (E-F) Antibody-dependent cellular phagocytosis (ADCP) of AF647 fluorophore-conjugated HCMV virions by THP1 monocytes was quantified using a flow-based assay and calculated as percentage AF647 positive cells. (E) HCMV- specific ADCP responses in transmitting versus non-transmitting dyads. (F) HCMV-specific ADCP responses in paired maternal and cord blood sera. Black bars denote median. Dotted lines for neutralization assays indicate the starting dilution corresponding to the lower limit of detection (ID50 = 10). FDR-corrected P values reported for Mann-Whitney U test or Wilcoxon signed-rank test. * P < 0.05, ** P < 0.01, *** P < 0.001.

= 0.011) with a trend towards statistical significance for TB40E (p = 0.053) and AD169r (p = 0.068). Interestingly, within mother-infant pairs, whole virion HCMV ADCP responses were highly enriched in the cord blood sera compared to maternal sera within both transmitting and non-transmitting groups (Figure 3F), suggesting preferential transfer of ADCP-mediating IgG across the placenta. The enhanced transfer of ADCC-eliciting antibodies across the placenta has been described in healthy pregnancies (46), yet our results indicate that IgG capable of diverse Fc effector functions such as ADCP are also selectively transferred from mother-to-fetus.

Maternal ADCP and high-avidity glycoprotein-specific IgG binding correlate with decreased risk of congenital HCMV infection

For our primary immune correlate analysis, we predefined 13 maternal antibody responses (Table 2) that we hypothesized would be associated with decreased risk of placental HCMV transmission. Using univariate logistic regression, 12 out of the 13 predefined variables were significantly (p < 0.05) associated with our primary endpoint of congenital HCMV infection. Maternal HCMV glycoprotein- specific IgG binding magnitude and neutralization titers were associated with increased risk of congenital transmission whereas IgG binding avidity and ADCP were associated with decreased transmission risk (Table 2). HCMV glycoprotein-specific IgG binding magnitude and neutralization titers remained significantly associated with higher risk of transmission after adjusting for maternal total IgG levels and maternal HCMV-specific IgM status; however, glycoprotein-specific IgG binding avidity was no longer statistically significant after adjusting for these variables (Supplementary Table 1). ADCP against Toledo, an HCMV strain lacking pentamer, remained significantly associated with protection against transmission status in both adjusted univariate regression models (Supplementary Table 1).

For our multivariable immune correlate analysis, we used least absolute shrinkage and selection operator (LASSO) modeling. Many immune parameters included in our predefined primary analysis were strongly correlated with each other (Supplementary Figure 4). Due to this high degree of multicollinearity, LASSO was employed for feature selection. LASSO is an approach to minimize overfitting a regression model that can shrink the coefficients of poorly predictive variables to zero, thereby removing them from the model. First, the cohort was randomly split into a training and test dataset and a k-fold cross-validation approach was employed to train the LASSO model. LASSO- selected features included magnitude of IgG binding to pentamer, avidity of gB IgG binding, avidity of gHgLgO IgG binding and ADCP against Toledo strain (Table 2). Notably, this 4-parameter LASSO model had a 0.75 accuracy (95% CI: 0.48-0.93) in predicting risk of congenital HCMV transmission (with 1.00 equaling perfect predictive capacity and 0.50 equaling the random prediction rate with no information). Consistent with the univariate analyses, the magnitude of pentamer IgG binding was associated with increased risk of congenital transmission whereas glycoprotein-specific IgG binding avidity and ADCP were associated with decreased risk of congenital transmission.

HCMV glycoprotein-specific IgG binding to Fc receptors differs in HCMV transmitting and non- transmitting dyads

In an exploratory analysis, we quantified FcR engagement by HCMV-specific IgG in maternal and cord blood sera to examine the role of Fc-mediated antibody responses and placental IgG transfer in congenital HCMV transmission. Since FcRs can transfer immune complexes across the placenta, we explored if differential HCMV-specific IgG binding to FcRs in maternal versus cord blood sera was associated with placental transmission (47). Using a multiplex FcR binding assay, we measured HCMV glycoprotein-specific IgG binding to the neonatal Fc receptor (FcRn), which regulates IgG half-life and placental IgG transfer, and the Fcγ receptors FcγRI, FcγRIIA, FcγRIIB and FcγRIIIA, which mediate Fc effector functions such as ADCP (47–49). HCMV-specific IgG binding to FcRs was significantly higher in transmitting compared to non-transmitting dyads across most antigens (Supplementary Figure 5A-G). Since transmitting dyads had higher binding to gB-, pentamer-, gHgLgO- and gHgL-coated multiplex beads at baseline (Figure 1E), we normalized FcR binding to total antigen-specific IgG binding for each HCMV glycoprotein measured. After normalization, transmitting and non-transmitting dyads had similar magnitude HCMV-specific IgG binding to FcRn, FcγRIIB and FcγRIIIA between and within mother- infant pairs (Supplementary Figure 6A, C-E). However, HCMV glycoprotein-specific IgG binding to FcγRI was significantly higher in non-transmitting compared to transmitting dyads (Figure 4A). Within mother-infant pairs, HCMV glycoprotein-specific IgG binding to FcγRI was also highly enriched in the cord blood of uninfected infants (Figure 4B). HCMV viral loads in the cord blood of infected infants were significantly negatively correlated with FcγRI binding to maternal gHgLgO- (Spearman ρ = -0.3, p = 0.019), pentamer- (Spearman ρ = -0.36, p = 0.020), and gB- (Spearman ρ = -0.32, p = 0.037) specific IgG, indicating that HCMV-specific IgG engagement of FcγRI may help control HCMV viremia. Conversely, glycoprotein-specific IgG binding to FcγRIIA was enhanced in the cord blood of infected infants with preferential transfer across the placenta within mother-infant dyads, though this was only observed for the high-affinity variant FcγRIIAH131 (Figure 4C-D, Supplementary Figure 6B).

• Download figure
• Open in new tab

Figure 4. HCMV glycoprotein-specific IgG binding to FcγRI and FcγRIIA differs in HCMV transmitting and non-transmitting dyads.

HCMV glycoprotein-specific IgG binding to Fcγ receptors (FcγRs) was measured using a Luminex-based binding antibody multiplex assay with a biotinylated FcγR and streptavidin-PE detection antibody. Normalized antigen-specific IgG binding to host FcγRs was compared between and within HCMV transmitting (red circles, n = 41) and non-transmitting (blue squares, n = 40) mother-infant dyads. (A-B) HCMV-specific IgG binding to FcγRI in (A) transmitting versus non-transmitting dyads and (B) paired maternal and cord blood sera. (C-D) HCMV-specific IgG binding to FcγRIIA high affinity H131 variant in (C) transmitting versus non-transmitting dyads and (D) paired maternal and cord blood sera. gB ecto = gB ectodomain. Black bars denote median. FDR- corrected P values reported for Mann-Whitney U test or Wilcoxon signed-rank test. * P < 0.05, ** P < 0.01, *** P < 0.001.

Systems serology analysis reveals distinct anti-HCMV antibody profile in HCMV transmitting compared to non-transmitting dyads

We next employed a principal components analysis (PCA) across all 50 humoral immune variables tested to define differences in HCMV-specific antibody responses in transmitting versus non- transmitting dyads. In the maternal and cord blood sera, PC1 accounted for 57 and 59% of the variance, respectively; however, PC2, which accounted for 16 and 17% of the variance, was superior at delineating between transmitting and non-transmitting dyads (Figure 5A-B). The top 12 contributors to PC2 were the same in the PCA of maternal and cord blood sera and included HCMV-specific IgG binding avidity, IgG binding magnitude, and ADCP (Figure 5C-D). These PCA results demonstrate that high avidity HCMV-specific IgG and ADCP responses were enriched in the maternal and cord blood sera of non-transmitting compared to transmitting dyads, further validating the findings in our primary analysis.

• Download figure
• Open in new tab

Figure 5. Principal components analysis (PCA) reveals distinct HCMV-specific antibody responses in HCMV transmitting compared to non-transmitting dyads.

Principal components analysis (PCA) across all 50 antibody responses measured for HCMV transmitting (red, n = 41) and non-transmitting (blue, n = 40) mother-infant dyads. Triangles (n=14) indicate dyads where mothers screened positive for HCMV-specific IgM responses and circles (n=67) indicate dyads where mothers had no detectable HCMV-specific IgM responses. Scatterplot of PC1 and PC2 for (A) maternal and (B) cord blood sera. Bar plot shows the magnitude of the top twelve antibody responses contributing to PC2 in (C) maternal and (D) cord blood sera.

Maternal HCMV-specific IgG signaling through FcγRI and FcγRIIA is associated with protection against congenital HCMV infection

Since our predefined immune correlate and exploratory analyses both implicated Fc-mediated effector functions as protective against congenital HCMV transmission, we sought to further validate this novel finding by using a cell-culture based assay to evaluate anti-viral IgG activation of host FcγRs (50, 51). This approach leverages mouse thymoma BW cell lines stably transfected to express chimeric FcγRs with an extracellular human FcγR domain fused to an intracellular mouse CD3ζ signaling domain (Figure 6A). Using this previously validated model system with mouse IL-2 as a quantitative read-out for FcγR activation, we quantified HCMV-specific IgG signaling through FcγRI and FcγRIIA. We measured HCMV-specific IgG signaling through FcγRI and FcγRIIA because 1) there were differences in IgG binding to these FcγRs in transmitting versus non-transmitting dyads (Figure 4) and 2) FcγRI and FcγRIIA mediate ADCP, which we identified as associated with protection against congenital HCMV transmission. We first used flow cytometry to confirm that each previously validated cell line was exclusively expressing the FcγR of interest (Figure 6B). HCMV-specific IgG signaling through FcγRI was 3-4 fold higher when reporter cells were stimulated with maternal or cord blood sera from non- transmitting versus transmitting dyads (Figure 6C). HCMV-specific IgG signaling through FcγRIIA was similar in transmitting and non-transmitting mothers, yet there was a trend towards significantly increased FcγRIIA activation by HCMV-specific IgG in cord blood from non-transmitting pairs (Figure 6D). Moreover, a sensitivity analysis excluding outliers from both groups (ROUT method, Q = 1%) revealed that HCMV-specific IgG from non-transmitting dyads was significantly better at signaling through both FcγRI and FcγRIIA (Supplementary Figure 7A-B). Within mother-infant pairs, HCMV- specific IgG activation of FcγRI and FcγRIIA was significantly lower in cord blood versus maternal sera in both groups (Figure 6E-F). While HCMV glycoprotein-specific IgG binding to FcγRI and FcγRIIA correlated weakly with ADCP (Supplementary Figure 4), HCMV-specific IgG signaling through FcγRI and FcγRIIA correlated strongly with ADCP responses in maternal and cord blood sera (Figure 6G-H). These results imply that robust HCMV-specific IgG engagement of host FcγRI/IIA mediates effective monocyte/macrophage virion phagocytosis and protection against placental HCMV transmission.

• Download figure
• Open in new tab

Figure 6. HCMV-specific IgG signaling through FcγRI and FcγRIIA is increased in HCMV non- transmitting compared to transmitting mother-infant dyads and strongly correlated with ADCP.

HCMV-specific IgG activation of FcγRs was measured using maternal (M) and cord blood (CB) sera from HCMV transmitting (red circles, n = 41) and non-transmitting (blue squares, n = 40) mother-infant dyads. (A) To quantify HCMV-specific IgG activation of FcγRs, mouse BW cell lines stably expressing chimeric human FcγRs fused to a mouse CD3ζ signaling domain were co-cultivated with virus:sera immune complexes for 20 hours. Activation of FcγRs by immune complexes triggered CD3ζ signaling and mouse IL-2 (mIL-2) secretion, which was measured by ELISA as a quantitative read-out of HCMV-specific IgG signaling through host FcγRs. (B) Flow cytometry of BW cell lines stained with anti-FcγRI/CD64 (light green), anti-FcγRII/CD32 (orange), anti-FcγRIII/CD16 (blue) and isotype control (red) PE-conjugated antibodies to confirm FcγR expression. HCMV-specific IgG activation of (C) FcγRIA and (D) FcγRIIA in transmitting and non-transmitting dyads. HCMV-specific IgG activation of (E) FcγRIA and (F) FcγRIIA within paired maternal and cord blood sera. (G-H) Spearman correlations between HCMV-specific IgG signaling through FcγR and ADCP. Black bars denote median. P values for Mann-Whitney U test or Wilcoxon signed-rank test. * P < 0.05, ** P < 0.01, *** P < 0.001.

Study population

We analyzed maternal and cord blood sera from 81 mother-infant dyads recruited from 2008-2017 as donors to the Carolinas Cord Blood Bank (CCBB). The CCBB collects maternal and cord blood biospecimens at delivery and mother-infant dyads were identified from over 29,000 CCBB donor records (Supplementary Figure 1). Maternal donors underwent infectious diseases screening for HCMV, hepatitis B virus, syphilis, hepatitis C virus, HIV-1/2, HTLV I and II, Chagas Disease, and West Nile virus. All mothers in our study were HCMV IgG seropositive and negative for other infectious diseases. Cord blood plasma was screened by the CCBB for HCMV infection with a Real-Time PCR COBAS AmpliPrep/TaqMan nucleic acid test (WHO HCMV reference standard, limit of quantification 147 IU/mL (75)), and all cord blood units positive for HCMV underwent a second confirmatory PCR test.

Cases of congenital HCMV infection (cCMV) were defined as mother-infant dyads with cord blood that screened positive for HCMV viremia at birth per PCR testing. “HCMV transmitting” cases with cCMV infection (n=41) were matched to a target of 1 “HCMV non-transmitting” seropositive mother (n=40) with no HCMV viremia detected in the cord blood (Supplementary Figure 1). Maternal HCMV IgG seropositivity and avidity were confirmed by an in-house whole virion HCMV ELISA and HCMV IgM seropositivity was determined using a clinical diagnostic ELISA (Bio-Rad). Maternal sera was screened for HCMV DNAemia via quantitative PCR (qPCR). For qPCR, 150-200ul of maternal sera was ultracentrifuged then DNA was extracted using the DNA QIAamp Kit (Qiagen) and tested in duplicate using SybrSelect and 300nM of forward and reverse primers designed to amplify the HCMV immediate early-1 (IE1) gene (24). HCMV viral loads were then interpolated from an IE1 plasmid standard curve. Maternal sera was also screened for hypergammaglobulinemia (total IgG concentration >15,000 mg/dL) via ELISA as previously described (44).

Cell culture and HCMV virus growth

Human retinal pigment epithelial cells (ARPEs), human foreskin fibroblasts (HFFs), human embryonic kidney (HEK)-293T cells, and human monocytes (THP-1s) were acquired from ATCC and cultured according to ATCC protocols. For differentiation, THP-1 cells were cultured in R10 media with 100nM PMA and incubated for 48 hours. HCMV strains TB40/E, AD169r, and Toledo were propagated as previously described (76). HFF or ARPE cells were infected (MOI = 0.01) and incubated for ∼14 days until 90-95% of cells showed cytopathic effect. Cells and supernatant were harvested, filtered (0.45-μm), and then concentrated on a 20% sucrose cushion (20,000 rpm).

Whole virion HCMV IgG binding and avidity

384-well ELISA plates were coated with 33 PFU/well of TB40/E, 2700 PFU/well of AD169r, or 1000 PFU/well of Toledo virus diluted in 0.1 M sodium bicarbonate buffer then incubated overnight before blocking. Maternal and cord blood sera serial dilutions were tested starting at 1:30, plated in duplicate, then incubated before adding horseradish peroxidase (HRP)-conjugated goat anti-human IgG Fc (Jackson ImmunoResearch). Plates were developed with tetramethylbenzidine (TMB) and peroxidase substrate (KPL) then optical density (OD) at 450nm was measured via SpectroMax. HCMV-specific IgG concentrations were interpolated from the linear range of a 5-parameter HCMV-hyperimmuneglobulin (Cytogam) standard curve. For IgG avidity, duplicate wells were treated with 7M urea or 1X PBS between the primary and secondary incubation steps and relative avidity index (RAI) was calculated as (OD with urea)/(OD with PBS)x100%. Placental HCMV-specific IgG transfer was calculated as (cord blood IgG)/(maternal IgG)x100%. Duplicates with coefficients of variance (CVs) >20% were repeated.

HCMV glycoprotein-specific IgG binding and avidity

A binding antibody multiplex assay (BAMA) was used to quantify HCMV glycoprotein-specific IgG binding and avidity (76, 77). HCMV gB ectodomain, pentamer complex, gH/gL/gO and gH/gL antigens were covalently coupled to intrinsically fluorescent beads (Bio-Rad). Maternal and cord blood sera were diluted 1:500 in assay diluent, plated in duplicate, then co-incubated with antigen-coupled beads. Antigen-specific IgG binding was detected with mouse anti-human IgG-PE (Southern Biotech) and mean fluorescent intensity (MFI) was acquired on a Bio- Plex 200 (Luminex). To quantify avidity, duplicate wells were incubated with sodium citric acid (pH = 4.0) or 1X PBS (pH = 7.4) between the primary and secondary incubations and RAI was calculated as (MFI with sodium citric acid)/(MFI with PBS)x100%. A serial dilution of HCMV-hyperimmuneglobulin (Cytogam) was included as a positive control and the cut-off for positivity was determined by calculating the mean MFI of seronegative sera binding plus 3 standard deviations. Blank beads and wells were included to account for background signal. Duplicates with CVs >25% were repeated.

Fc receptor (FcR) binding by HCMV glycoprotein-specific IgG

FcR binding by HCMV glycoprotein- specific IgG was measured using a modified BAMA (36, 78). Purified human FcRn, FcR1a, FcR2a (clones H131 and R131), FcR2b, and FcR3a (clones V158 and F158) were produced by the DHVI Protein Production Facility and biotinylated in-house. Maternal and cord blood sera were diluted 1:500 in assay diluent before incubation with HCMV glycoprotein-coated beads, as above. Next, biotinylated human FcRs were complexed with streptavidin-PE (BD Biosciences) then co-incubated with antibody- bound beads after washing. For FcRn, the assay buffer pH was adjusted to 6.5 to facilitate FcRn-IgG binding. MFI was acquired on a Bio-Plex 200 and duplicates with CVs >25% were repeated.

Cell-associated HCMV glycoprotein B (gB) IgG binding

A gB-transfected cell binding assay was used to measure IgG binding to cell-associated gB (36). HEK-293T cells were co-transfected with DNA plasmids expressing green fluorescent protein (GFP) and full-length gB (Sino Biological) using the Effectene Transfection Kit (Qiagen). After incubation at 37°C for 48 hours, 200,000 live cells/well were plated into 96-well U-bottom plates then centrifuged and washed before a 5 minute incubation in Human TruStain Fc Block (1:1000 dilution; BioLegend). In duplicate, cells were then co-incubated with maternal and cord blood sera diluted 1:500 or with controls for 2 hours at 37°C. A serial dilution of HCMV-hyperimmuneglobulin (Cytogam) and a gB-specific monoclonal antibody (TRL-345) were included as positive controls and seronegative sera samples were included as negative controls. Following incubation, cells were stained with Live/Dead Near-IR (1:1000; Invitrogen) then washed and stained with PE-conjugated goat anti-human IgG-Fc (1:200; Southern Biotech). Lastly, cells were washed and fixed with 10% Formalin. Events were acquired on an LSRII flow cytometer, and the percentage of PE-positive cells was calculated from the live, GFP-positive cell parent population using FlowJo (36). Unstained cells and single color-stained cells were included for setting gates and compensation. The cut-off for positivity was the mean signal of HCMV seronegative samples plus 3 standard deviations. Duplicates with CVs >50% were repeated.

Neutralization

HCMV neutralization was measured by high-throughput fluorescence bioimaging (36). Epithelial cells (ARPEs), fibroblasts (HFFs) or differentiated monocytes/macrophages (THP-1s) were plated in 384-well clear, flat-bottom plates and then incubated at 37°C overnight. To quantify neutralization, maternal and cord blood sera samples were diluted 1:10 followed by an 8-point serial dilution and then co-incubated with HCMV strains AD169r (MOI = 2) or Toledo (MOI = 1) for 2 hours at 37°C to allow immune complex formation. Virus-only wells and an 8-point HCMV-hyperimmuneglobulin (Cytogam) serial dilution were included as positive controls while seronegative samples and no virus wells were included as negative controls. After addition of the virus:sera mixture, cells were incubated at 37°C for 24-48 hours (depending on optimized conditions for each virus strain and cell type) then fixed in 10% Formalin. To quantify HCMV infection, plates were stained with mouse anti-HCMV immediate-early 1 (IE1) gene (1:500; MAB810 Millipore) followed by goat anti-mouse IgG-AF488 (1:500; Millipore) and cell nuclei were stained with DAPI (1:10,000; Thermo Fisher). After staining, plates were visualized with a Cellomics fluorescent plate reader (Supplementary Figure 8A) to enumerate total cell count, infected cell count, and percent infected cells in each well. Following image acquisition, neutralization titers corresponding to the dilution that resulting in a 50% reduction in percent infected cells (ID50) were calculated using interpolation in GraphPad Prism (Supplementary Figure 8B). Samples containing wells with low cell counts and with duplicates with CVs >50% were repeated.

Whole HCMV virion antibody-dependent cellular phagocytosis (ADCP)

HCMV strains TB40/E, AD169r, and Toledo were conjugated to fluorochrome AF647 (Invitrogen) to measure ADCP (36). In brief, 2.0×10⁶ PFU of TB40/E, 1.0×10⁷ PFU of AD169r, or 1.0×10⁷ PFU of Toledo virions were buffer- exchanged with 1X PBS using a 100,000-kDa Amicon filter (Millipore) and conjugated to 10µg of AF647 N-hydroxysuccinimide ester reconstituted in DMSO during a 1-hour incubation with constant agitation. The conjugation reaction was quenched with 80ul of 1M tris-HCl (pH = 8.0) and fluorescently-labeled virus was diluted 25X in wash buffer. A serial dilution of HCMV-hyperimmuneglobulin (Cytogam) was included as a positive control while seronegative sera samples and an anti-RSV monoclonal antibody (Synagis) were included as negative controls. In a 96-well plate, fluorescently-labeled virus was co- incubated with maternal sera, cord blood sera (1:10) or controls at 37°C for 2 hours to allow immune complex formation before adding 50,000 THP-1 cells per well. Plates were then centrifuged (1200g) at 4°C for 1 hour in a spinoculation step before a 1-hour incubation at 37°C to allow for phagocytosis. Next, cells were transferred to a 96-well U-bottom plate, washed and fixed with 1% Formalin. Events were acquired on an LSRII flow cytometer, and the percentage of AF647-positive cells was calculated from the live THP-1 monocyte cell parent population using FlowJo (gating strategy in Supplementary Figure 9C-D). Unstained cells and single color-stained cells were included as controls for setting gates and compensation. The cut-off for positivity was the mean signal of HCMV seronegative samples plus 3 standard deviations and duplicates with CVs >50% were repeated.

Fcγ receptor signaling assay

HCMV-specific IgG signaling through FcγRs was measured using a previously validated approach (50). Briefly, we used mouse BW thymoma cells stably expressing chimeric FcR-CD3ζ, which contains an extracellular human FcR and intracellular CD3ζ signaling domain, to quantify anti-viral IgG activation of host FcγRs. To confirm FcγR expression, 0.5×10⁶ BW cells were added to each well in a 96-well V-bottom plate. The non-transfected parental BW cell line and BW cell lines expressing either human CD64 (FcR1a) or CD32a (FcR2a) were stained for surface expression of human FcγRs with 5µl of anti-human CD64-PE (clone 10.1, eBioscience), 5µl anti-human CD32 (clone 6C4, eBioscience), and 5µl anti-human Ig-PE isotype control (clone P3.6.2.8.1, eBioscience) then cells were fixed with 2% paraformaldehyde. Events were acquired on Fortessa LSR flow cytometer and analyzed using FlowJo v10.7.2.

To quantify FcγR activation, 96-well plates were coated with 20,000 PFU/well of HCMV strain AD169r and incubated at 4°C overnight. After coating, plates were washed with assay buffer (1X PBS + 1% FBS) then blocked with buffer at room temperature for 1 hour. After blocking, HCMV-coated plates were co-incubated with maternal or cord blood sera diluted 1:10 in BW cell media at 37°C for 1 hour. HCMV-hyperimmuneglobulin (Cytogam), seronegative and no antibody conditions were included as positive and negative controls. Following immune complex formation, plates were washed with BW media before adding 100,000 FcR1a-expressing or FcR2a-expressing BW cells per well. In separate wells, the parental (non-transfected) BW cells were added as a negative control. Cells were incubated at 37°C with immune complex coated plates for 20 hours before being transferred to V-bottom plates and pelleted (1200 rpm). Cell supernatants were harvested and mouse IL-2 (mIL-2) levels in culture supernatants were measured using ELISA. For the mIL-2 ELISA, 384-well plates were coated with 3µg/mL of purified rat anti-mouse IL-2 (BD biosciences) and incubated at 4°C overnight before blocking. A serial dilution of purified mIL2 and BW cell culture supernatants were added in duplicate and incubated for 1 hour at room temperature. After primary incubation, plates were incubated with rat anti-mIL2 conjugated to biotin (BD biosciences; 1:2000) followed by streptavidin-HRP (1:8000) for 1 hour and 30 minutes respectively. Plates were developed with TMB/KPL then OD at 450nm was measured via SpectroMax. Mouse IL-2 concentrations were interpolated from a 5-parameter mIL-2 standard curve using GraphPad Prism. Duplicates with CVs >30% were repeated.

Statistics

All primary raw and analyzed data underwent independent data quality control (QC) by a second lab member prior to statistical analysis and inclusion in the study. For all analyses, interpolated antibody or cytokine concentrations, MFI values and neutralization titers were log-transformed to normalize the data distribution. Antibody responses below the limit of detection were set equal to the limit of detection for statistical analyses. To assess differences between HCMV transmitting and non- transmitting mother-infant dyads, immune variables were compared between groups using Mann- Whitney U/Wilcoxon rank-sum and within dyads using Wilcoxon signed rank tests. Statistical significance was defined a priori as p < 0.05 with a two-tailed test and a Benjamini-Hochberg correction for multiple comparisons. For the primary immune correlate analysis, 13 predefined maternal humoral immune variables were included for univariate logistic regression analysis. Additional sensitivity analyses included multivariable models adjusted for 1) maternal hypergammaglobulinemia and 2) maternal HCMV-specific IgM status. Using the 13 predefined immune variables from our primary immune correlate analysis, LASSO regression analysis was performed using the caret R package and a k-fold cross-validation framework. For the LASSO regression, the cohort was randomly split into two independent datasets, which were used for training the LASSO model and testing the predictive performance of the model respectively. All statistical analyses were completed in R v4.1.1 and GraphPad Prism v9.1. The principal components analysis (PCA) plots were rendered using ggplot2 and correlation matrices were plotted using the corrplot package in R. All other figures were generated using GraphPad Prism or created in Biorender.com.

Study Approval

Approval was obtained from Duke’s Institutional Review Board (Pro00089256) to use de-identified clinical data and biospecimens provided by the CCBB. No patients were prospectively recruited for this study and all samples were acquired retrospectively from the CCBB biorepository from donors who had previously provided written consent for banked biospecimens to be used for research.