The 2011 maternal influenza (MatFlu) vaccination trials (approval numbers: 101106 and 101107) and this sub-study (approval number: M200444) were approved by the Human Research Ethics Committee of the University of the Witwatersrand. All study participants provided written informed consent to have their stored samples used for future studies. All healthy donors in this study provided written informed consent to obtain plasma, isolate IgG and have their samples stored for future use.

The two randomized, double-blind, placebo-controlled MatFlu vaccination trials conducted in Soweto, South Africa have been previously described (6). Briefly, women, all of whom were pregnant, between the ages of 18-38 years were stratified according to their HIV status and randomly assigned (1:1) to the placebo or vaccine groups. The trivalent inactivated vaccine (TIV) used in the trials was Vaxigrip, which contained 15 µg each of A/California/7/2009 (A/H1N1/pdm09), A/Victoria/210/2009 (A/H3N2), and a B/Brisbane/60/2008-like virus (B/Victoria) as recommended by the World Health Organization for the Southern Hemisphere in 2011. Active surveillance for respiratory illness and PCR-confirmed influenza-illness was performed. In this study, plasma samples from a sub-set of HIV-uninfected women (74 placebo-recipients and 67 vaccinees) and WLWH (53 placebo-recipients and 66 vaccinees), collected prior to vaccination and one-month post-vaccination, were tested. Data from participants with PCR-confirmed influenza-illness were excluded from this study.

A chimeric recombinant hemagglutinin (cHA) protein, composed of an H6 head and an H1 stalk (cH6/1) was produced as previously described (33). For use as a positive control in all the Fc assays, plasma was pooled from 5 healthy donors with high HA stalk IgG titers. From this pooled plasma IgG was isolated using Protein G (Pierce Biotechnology), according to the manufacturer’s instructions and confirmed by IgG enzyme linked immunosorbent assay (ELISA). A cross-reactive HA stalk antibody CR9114 that mediates potent Fc effector function was expressed as a positive control (34). For antibody expression, plasmids encoding heavy or light chain genes were co-transfected into HEK293F cells with PEI-MAX 40,000 (Polysciences) head-to-head. Cells were cultured for six days in 293F Freestyle media at 37°C, 10% CO₂, then harvested supernatants were filtered and purified using Protein G (Thermo Scientific). Antibody concentration was quantified by nanodrop using sequence-specific extinction coefficients as determined by ProtParam (ExPASy) and confirmed by ELISA.

Modified HAI assays were previously performed (6, 35). Briefly, plasma samples instead of serum were treated with receptor-destroying enzyme (RDE) from Vibrio cholera (Denka-Seiken). Then these were diluted 1:10 in saline and subsequent serial 2-fold dilutions of the plasma were used in a standard HAI assay using 4 hemagglutinating units of the antigen and 0.75% turkey red blood cells. Plasma samples with titers ≥10 were considered indicative of immune responses. The antigen used in the assays was A/H1N1/pdm09.

H1/stalk IgG ELISAs were performed as previously described (32). The cH6/1 recombinant protein described above was utilized and binding against this protein measures antibodies against the H1 stalk domain. Plates were coated with cH6/1 diluted to 2 µg/ml overnight at 4°C. Plates were washed and subsequently blocked with 3% fetal bovine serum (Biowest), 0.5% non-fat dry milk powder (Bio-Rad) for 2 hours at room temperature (RT). After washing, plasma samples were added to the plate diluted to a starting concentration of 1:40, followed by two-fold serial dilution and incubated for 2 hours at RT. Following a wash, anti-human IgG (Fab specific)-peroxidase antibody (Sigma, USA) diluted to 1:3000 in blocking solution was added to the plate. After 1-hour incubation at RT, plates were washed and developed using SigmaFast o-phenylenediamine dichloride (OPD) (Sigma, USA) for 10 min at RT. The reaction was stopped by adding 3M HCl and the optical density (OD) was measured at 490 nm. The antibody concentration was quantified against the standard curve included on each plate that consisted of polyvalent human normal immunoglobulin (Polygam) (National Bioproducts Institute, South Africa) with an assigned arbitrary value of 1000 arbitrary units (AU)/ml.

The ADCP assay was performed as previously described (36). Briefly The EZ-Link Sulfo-NHS-LC-Biotin kit (Thermo Scientific) was used to biotinylate cH6/1, which was coated onto fluorescent neutravidin beads (Invitrogen). The coated beads were incubated for 2 hours with mAbs at a final concentration of 50 µg/ml or a 1:100 dilution of plasma sample, prior to overnight incubation with a monocytic cell line, THP-1 cells. The THP-1 cells were obtained from the NIH AIDS Reagent Program and cultured at 37°C, 5% CO₂ in Roswell Park Memorial Institute (RPMI) 1640 media supplemented with 10% fetal bovine serum (FBS), 100 units/ml Penicillin and 100 ug/ml Streptomycin (Gibco), referred to as R10, and not allowed to exceed 4 x 10⁵ cells/ml. This assay was completed on a FACS Aria II (BD Biosciences). Phagocytic scores were calculated as the geometric mean fluorescent intensity (MFI) of the beads multiplied by the percentage bead uptake minus the no antibody control as background. For this and all functional Fc assays, pooled plasma from 5 healthy donors with previous exposures to influenza, screened for H1 stalk antibodies and CR9114 were used as positive controls and Palivizumab (RSV mAb; MedImmune, LLC) and VRC01 (In house HIV mAb) were used as negative controls. In order to normalize across plates and runs, the pooled plasma positive scores were averaged, divided by the pooled plasma score per plate and this normalizing factor multiplied across the scores.

The ability of plasma antibodies to cross-link cH6/1 HA stalk antigen and activate FcγRIIIa on Jurkat-Lucia™ NFAT-CD16 cells (In vivogen) was measured as a proxy for ADCC or cell lysis. These cells In vivo were cultured according to the manufacturer’s instructions. Adapted from elsewhere (37), high-binding 96 well plates were coated with 1 μg/mL cH6/1 and incubated at 4°C overnight. Plates were then washed with phosphate buffered saline (PBS) and blocked at room temperature for 1 hour with 2.5% bovine serum albumin (BSA)/PBS. After washing, mAbs at a starting concentration of 1 mg/ml or a 1:10 dilution of plasma sample was added and incubated for 1 hour at 37°C. Subsequently, 2x 10⁵ cells/well in R10 were added and incubated for 24 hours at 37°C, 5% CO₂. Then 25 µl of supernatant was transferred to a white 96-well plate with 75 µl of reconstituted QUANTI-Luc secreted luciferase and read immediately on a Victor 3 luminometer (PerkinElmer) with 1s integration time. Relative light units (RLU) of a no antibody control were subtracted as background. The pooled plasma and CR9114 were used as positive controls and Palivizumab and VRC01 were used as negative controls. The reported values are the mean of three kinetic reads taken at 0, 2.5, and 5 min. To induce the transgene 1x cell stimulation cocktail (Thermo Scientific) and 2 μg/ml ionomycin in R10 was added as a positive control to confirm sufficient expression of the CD16 Fc receptor. ADCC RLUs were normalised across plates and runs, by averaging the pooled plasma RLUs and multiplying the RLUs across the samples with the normalizing factor.

ADCD was measured using a previously described high-throughput bead-based assay (38). Biotinylated cH6/1 was coated 1:1 onto fluorescent neutravidin beads (Invitrogen) for 2 hours at 37°C. The antigen-coated beads were incubated with a 1:10 plasma sample dilution or mAbs at a starting concentration of 100 μg/ml for 2 hours and incubated with guinea pig complement diluted 1 in 50 with gelatin/veronal buffer for 15 minutes at 37°C. Beads were washed in PBS and stained with anti-guinea pig C3b-FITC, fixed and interrogated on a FACS Aria II (BD Biosciences). Complement deposition scores were calculated as the percentage of C3b-FITC positive beads multiplied by the geometric MFI of FITC in this population minus the no antibody or heat inactivated controls. The pooled plasma and CR9114 were used as positive controls and Palivizumab and VRC01 were used as negative controls. ADCD scores were normalised between plates and runs, by dividing the ADCD score of the pooled plasma by the average across the plates and multiplying the scores with this normalizing factor.

High-binding 96 well ELISA plates were coated with 1 ug/ml cH6/1 in PBS overnight at 4°C. Three wells on each plate were directly coated with 5 ug/ml IgG, isolated from healthy donors, signals from these wells were used to normalize the FcR activity of the plasma samples and pooled plasma was used as a positive control. Plates were washed with PBS and blocked with PBS/1 mM ethylenediaminetetraacetic acid (EDTA)/1% BSA for 1 hour at 37°C. Plates were then washed and incubated with 1:10 diluted plasma for 1 hour at 37°C and then with 0.2 ug/ml of biotinylated FcγRIIa dimer or 0.1 ug/ml of biotinylated FcγRIIIa dimer for 1 hour at 37°C. The dimeric FcγRs were provided by Prof. Mark Hogarth from the Burnet Institute, Melbourne, Australia (39). Subsequently, a 1:10000 dilution of Pierce high-sensitivity streptavidin-horseradish peroxidase (Thermo Scientific) was added for a final incubation of 1 hour at 37°C. Lastly, TMB (3,3,5,5-tetramethylbenzidine) substrate (Sigma-Aldrich) was added, colour development was stopped with 1 M sulfuric acid and absorbance read at 450 nm.

Data were analyzed in Prism (v9; GraphPad Software Inc., San Diego, CA, USA). Non-parametric tests were used for all comparisons. The Mann-Whitney and Wilcoxon tests were used for unmatched and paired samples, respectively. Fc polyfunctionality Z-scores were calculated by subtracting the mean of the Fc function from the individual value and divided by the standard deviation of the mean and then adding all the Z-scores for each function per individual. The proportions of responders and non-responders in each group and the proportions of participants with high or low Fc polyfunctionality Z-scores were compared by Fisher’s exact tests. All correlations reported are non-parametric Spearman’s correlations. P values less than 0.05 were considered statistically significant.

In our previous study, HAI titers were significantly boosted by vaccination regardless of HIV status, but WLWH showed significantly lower HAI titers post-vaccination compared to HIV-uninfected women (6). We confirmed this finding in a subset of plasma samples from pregnant vaccinated participants from the MatFlu cohort, including 67 HIV-uninfected women and 66 WLWH, matched in age with a median of 25 years (range 18-39) and 27 years (range 18-38) respectively, and excluded all PCR-confirmed influenza virus infection cases. Plasma samples were collected pre-vaccination and post-vaccination at a median of 31 days (range 28-33) for HIV-uninfected and 30 days (range 28-31) for WLWH ( Supplementary Table 1 ). Post-vaccination HAI titers were significantly lower in WLWH compared with HIV-uninfected women (medians 80 vs. 320 A/H1N1 HAI titer) ( Figure 1A ), coupled with decreased boosting levels following vaccination (median fold change 1.5 vs. 4) ( Supplementary Figure 1A ). In these samples, we also confirmed that post-vaccination H1 stalk antibody titers were significantly lower in WLWH compared with HIV-uninfected women (medians 259.3 vs. 433.5 AU/ml) ( Figure 1B ). This was despite vaccine boosting H1 stalk antibody titers in both groups, which was consistent with the parent study (32). Furthermore, we also confirmed that in contrast to HAI titers, the HIV-uninfected women and WLWH had more comparable fold increases in the H1 stalk antibody titer boosting (median fold change 1.9 vs. 1.7) ( Supplementary Figure 1B ). As expected, no changes in HAI and H1 stalk antibody titers amongst placebo recipients living with or without HIV were observed ( Supplementary Figure 2 ).

We first assessed whether WLWH differ in terms of pre- and post-vaccination HA stalk antibody Fc-mediated effector functions. To do this, we measured HA stalk-specific ADCP, ADCC reporter activity and ADCD using a chimeric recombinant HA protein with an H6 head, to which humans are generally naïve, and an H1 stalk. TIV administration resulted in significant boosting of HA stalk-specific ADCP in HIV-uninfected women but not in WLWH ( Figure 2A ). Despite this, post-vaccination ADCP activity between the two groups was similar, a consequence of pre-vaccination HA stalk-specific ADCP being significantly higher in WLWH (medians 348.6 vs. 189.7). No ADCP boosting was observed in the placebo groups indicating that these responses were TIV specific ( Supplementary Figure 2 ). For ADCC, which was measured using a high throughput FcγRIIIa activation assay, previously shown to correlate with NK degranulation assays (40), there was no significant boosting in either HIV-uninfected women or WLWH ( Figure 2B ). However, like ADCP, pre-vaccination ADCC was significantly higher in the WLWH (medians 154.5 vs. 73.5), and therefore also post-vaccination (medians 170 vs. 83.2). The ADCC activity amongst the vaccinated and placebo groups were similar and remained unchanged since there was no vaccine-mediated boosting of this function ( Supplementary Figure 2 ).

To assess Fc-FcγR engagement required for ADCP and ADCC activity, plasma samples were tested for the capacity to bind cH6/1 antigen and cross-link recombinant soluble dimeric FcγRIIa or FcγRIIIa, respectively (39). No boosting in FcγRIIa and FcγRIIIa binding was observed in placebo recipients ( Supplementary Figure 2 ). However, as observed in the cell-based functional and reporter assays, FcγRIIa binding was boosted by TIV in HIV-uninfected women but not WLWH, whereas for FcγRIIIa binding, no boosting was observed in either group ( Supplementary Figures 3A, B ). Furthermore, strong positive Spearman’s correlations were noted between FcγRIIa binding and ADCP scores (r=0.44; p<0.0001), and between FcγRIIIa binding and ADCC RLUs (r=0.55; p<0.0001) ( Supplementary Figures 3C, D ).

Since ADCD has been shown to confer protection from influenza virus infection (16, 21), we examined this function in WLWH and HIV-uninfected women, before and after TIV vaccination. ADCD, measured using a bead-based flow cytometry assay, was significantly boosted by TIV in all participants irrespective of HIV status ( Figure 2C ). However, ADCD was substantially impaired in WLWH prior to vaccination and post-vaccination remained significantly lower in comparison to the HIV-uninfected women after TIV boosting (medians 16.2 vs. 225). Placebo recipients showed no increase in ADCD activity ( Supplementary Figure 2 ).

When we compared Fc-mediated functions, we observed differential boosting across the HIV-uninfected women and WLWH for ADCP and ADCD following vaccination ( Figure 2D ). For ADCP the WLWH were not boosted, and the proportion of high responders (individuals that exceeded a 2-fold increase in Fc activity) was significantly lower than in the HIV-uninfected group (17% vs. 54%). Since there was no boosting of ADCC potential, the proportion of high responders was similar between HIV-uninfected women and WLWH (31% vs. 27%). However, for ADCD, in the WLWH the TIV boosting and the proportion of high responders was significantly lower (median fold change 1.8 and high responders 48%) than in the HIV-uninfected women (median fold change 2.7 and high responders 75%). Thus, while ADCD activity was boosted in both groups in response to TIV vaccination, this function was substantially impaired in WLWH prior to vaccination.

Co-ordinated antibody responses elicited by vaccines against other diseases have been associated with increased efficacy (41). We therefore assessed the correlations between HAI titers, stalk-specific responses and Fc-mediated effector functions at baseline and after TIV vaccination and compared these in HIV-uninfected women and WLWH. Prior to vaccination, there was a significant correlation between HAI and HA stalk titers in HIV-uninfected women, but the significant boosting of HAI antibodies resulted in loss of this correlation post-vaccination ( Figures 3A, B ). In contrast in WLWH, while there was a strong correlation between HA stalk and HAI pre-vaccination ( Figure 3C ), HA stalk titers were boosted simultaneously with HAI titers, as previously reported (32) and remained significantly correlated ( Figure 3D ). In the HIV-uninfected women, ADCP, ADCC, ADCD and FcγRIIa and FcγRIIIa binding were significantly associated with the HA stalk titers before vaccination, and these associations improved slightly following vaccination ( Figures 3A, B ). In WLWH, ADCP, ADCD and FcγRIIa and FcγRIIIa binding were associated with HA stalk titers pre-vaccination, with the FcγRIIIa binding been the only association to not improve, after TIV vaccination ( Figures 3C, D ). We also observed varying relationships amongst the Fc-mediated effector functions. In the HIV-uninfected women, we observed a significant correlation between ADCP and ADCD pre-vaccination, but not post-vaccination. Instead, after vaccination we observed a correlation between ADCC potential and both ADCP and ADCD. The FcγRIIa and FcγRIIIa binding correlated with all the Fc functions both pre- and post-vaccination in the HIV-uninfected women. In contrast, in WLWH the existing association between ADCD and ADCP prior to vaccination was strengthened post-vaccination. There was also an in increase in the FcγRIIa and FcγRIIIa binding associations with all the Fc functions, following vaccination in this group. Therefore, vaccination resulted in nuanced improvements of particular functional relationships that differed between WLWH and HIV-uninfected women, but overall co-ordination of antibody responses and Fc-mediated functions was not significantly improved.

Fc polyfunctionality has been shown to be associated with protective immunity and broader responses in other diseases (41–44). We calculated HA stalk-specific Fc polyfunctionality scores by summing the z-scores of ADCP, ADCC and ADCD for each of the participants. Prior to vaccination, WLWH showed significantly higher Fc polyfunctionality directed at the HA stalk than HIV-uninfected women, as a result of their baseline higher ADCC and ADCP responses (medians 0.89 vs. -0.26) ( Figure 4A ). The proportion of participants with high HA stalk-specific Fc polyfunctionality (the sum of z-scores greater than 0) was also higher for WLWH (76%) than for HIV-uninfected group (45%). Following vaccination, HA stalk-specific Fc polyfunctionality in the HIV-uninfected women significantly increased, reaching the same levels observed in WLWH at baseline. In contrast, WLWH did not show improvement of overall HA stalk-specific Fc-mediated functionality following vaccination.

We next assessed the influence of H1 stalk titer on stalk- specific Fc polyfunctionality after vaccination. In the HIV-uninfected group there was a strong correlation between the HA stalk antibody titers and HA stalk- specific Fc polyfunctionality (r=0,71; p<0.0001) ( Figure 4B ). However, in WLWH we observed no correlation between H1 stalk titers and HA stalk-specific Fc polyfunctionality (r=0,19; p=0.26) ( Figure 4C ). The lack of association in WLWH suggests that in this group, the potency of the Fc–mediated functions is mediated through qualitatively different HA stalk-specific antibodies to those occurring in HIV-uninfected women. Overall, these data indicate a fundamentally different humoral response in WLWH in response to vaccination.