We do not know the sex of the 10 individuals who were the source of the human convalescent sera used in this study, as that information was de-identified.

HEK293T cells (ATCC: CRL-11268) were cultured in α-MEM (Gibco, cat. 12561-056), containing L-glutamine, supplemented with 5% heat-inactivated fetal bovine serum (FBS; Gibco) and 1% penicillin/streptomycin (Gibco, cat. 15140-122).

NK-92 cell line (GFP-CD16–176 V/V) is an NK cell line engineered to express the high-affinity (176V/V; GenBank: BC017865.1) FcγRIIIA (CD16A). The cell line was purchased from ATCC (item number PTA-8836) and cultured in α-MEM complete media (Gibco, cat. 12571-063) supplemented with 10% heat-inactivated horse serum (Thermo Fisher, cat. 16050122), 10% FBS, and 1% penicillin/streptomycin. The cells were passaged every 4 days in the presence of hIL-2 (Corning, cat. F354043, lot 2283004), 200 U/mL, or clarified (filtered) J558L supernatant, which served as a source of human IL-2 (29, 30).

Human IL-2-producing J558L cells are derived from a mouse myeloma cell line that has been transfected with the huIL-2 cDNA, resulting in the stable release of human IL-2 into the supernatant (31). The cells were cultured in RPMI media (Gibco, cat. 75-085) containing L-glutamine and supplemented with 10% FBS, 1% penicillin/streptomycin, 1% α-MEM, 1% β-mercaptoethanol, and 1% HEPES 1M pH 7.4 (cat. 25060CI, ThermoFisher Scientific, Waltham, MA). Hu-IL-2-producing J558L cells produced human IL-2 that served as a growth supplement for NK-92 GFP-CD16 176V cells. Cells were allowed to grow for about 2 weeks, until the media turned yellow. The culture was centrifuged for 3 minutes at 395 × g, the supernatant was filtered through a 0.22 μm filter, aliquoted, and frozen at -80°C (32).

SKOV3 cells are human ovarian cancer cells that express the HER2 receptor, making them suitable for NK cell ADCC studies when a therapeutic human IgG1 antibody to HER2 (Trastuzumab/Herceptin) is added. These cells serve as a positive control for CD107a surface expression on NK cells. SKOV3 cells were cultured in RPMI media containing L-Glutamine and supplemented with 5% FBS and 1% penicillin/streptomycin.

All cells were maintained in 5% CO₂ at 37 °C.

Protein sequences of the full-length gC2 and gD2 were derived from HSV-2 strain 333, while the full-length sequences of gE2 and gI2 were derived from HSV-2 strain 2.12 (GenBank: AQZ56445.1 for gC2; AMB66102.1 for gD2; for gE2 and gI2, the sequences and locations of the gE2 mutations are described in Supplementary Figure 1) (21, 28). The DNA sequences were cloned into the pcDNA3.1(+) vector, which is codon-optimized for expression in human cells, and sequenced to verify accuracy. The size was also checked by agarose gel electrophoresis at GenScript (Piscataway, NJ, USA).

Sub-confluent HEK293 cells were transiently transfected with gC2, gD2, and gE2, or gD2, gE2, and gI2 plasmids using 22.5 μg of total DNA (7.5 μg for each plasmid) mixed with 56.5 μg PEI (1mg/mL) (PEI transfection reagent, Polysciences, cat. 26008) in 1.5 mL of Dulbecco’s modified essential medium (DMEM) (Gibco, cat. 11965-084) without FBS. After 15 min at R.T., 6 mL of DMEM was added, and the transfection solution was incubated overnight at 37 °C in 5% CO2 on the HEK293 cell monolayer. We used DL6 mAb to detect gD2, mAb B1E6 and mAb E1 for gE2, H1196 (purchased from Virusys, now SeraCare, cat. P1124) for gC2, and rabbit polyclonal Ab UP1928 for gI2 (33–35).

Ten HCS served as the source of antibodies for the NK cell CD107a surface expression assays. The HCS were provided by Anna Wald and Christine Johnston and were obtained from subjects with genital herpes (Table 1). The individual HSC and the pooled human IgG (IgG from human sera, cat. 56834, Sigma) were pre-adsorbed with the HEK293 cells for 1 to 2 hours at R.T. The sera were centrifuged twice for 10 min at 9,500 x g, the supernatant removed, aliquoted, and stored at -20°C.

ELISA was performed to measure antibodies to gC2, gD2, and gE2 in the HCS. 100 ng of gC2, gD2, or gE2 was used, and the assay was performed as previously described (36).

HEK293 cells were transiently transfected to express gC2, gD2, and gE2, or gD2, gE2, and gI2. Twenty-four hours later, these cells served as target cells for NK cell activation (effector cells). The following day, the transfection media was removed, the target cells were washed at R.T. with PBS, detached using 0.05% trypsin-EDTA at 37°C, washed, and viable cells counted by trypan blue exclusion. 3x10⁵ target cells were incubated with pre-adsorbed pooled human IgG or HCS for 30 min. at R.T. NK effector cells were washed to remove hu-IL-2, viable cells counted using trypan blue, and 3x10⁵ cells were added to the target cells at a 1:1 ratio for two hours at 37°C in the presence of a 1:1000 dilution of monensin (BioLegend, cat. 420701), and 200 ng of anti-CD107a (LAMP1, clone H4A3, cat. 328619, BioLegend). The cells were washed twice with PBS, and Live/Dead IR stain was added for 10 min at R.T. in the dark, followed by three PBS washes, then fixed with 1% paraformaldehyde in PBS and stored at 4 °C until flow cytometry analysis the next day. Flow cytometry data were collected using a BD FACSymphony A5 SE and analyzed with FlowJo v10.8.1. Fifty thousand events were collected per sample, gated on singlets, and each condition was performed in duplicate.

The following reagents were used for flow cytometry. Monensin solution (cat. 420701, BioLegend, used at 1:1000); PMA (phorbol 12-myristate 13-acetate) cell activation cocktail (without Brefeldin A), composed of PMA and ionomycin (cat. 423301, BioLegend, used at 0.5 μl/well in a final volume of 200 μl); APC anti-human CD107a (LAMP-1); live/dead™ fixable near-ir dead dell stain kit (cat. L10119, Invitrogen); Alexa Fluor 488 anti-mouse IgG2A (Clone RMG2a-62, cat. 407121, BioLegend); Alexa Fluor 488 goat anti-mouse IgG (Clone Poly4053, cat. 405319, BioLegend); Alexa Fluor 488 goat anti-rabbit IgG (H+L) (A11034, LifeTechnology); mouse anti-human IgG Fab PE (cat. MA1 = 10377, Invitrogen). Herceptin (Trastuzumab) was obtained from the pharmacy at Fox Chase Cancer Center. A 10% aqueous paraformaldehyde solution (EM grade) was diluted to 1% in PBS (cat. 15712, Electron Microscopy Science).

HSV-1 and HSV-2 gE and gE/gI heterodimers were purified in our lab using baculovirus-produced proteins, as previously described (36). The gE1 and gE2 proteins were purified on a Ni column, while the gE1/gI1 and gE2/gI2 heterodimers were purified on a Ni column followed by an anti-FLAG column. Purification was facilitated by C-terminal His tags on gE1 and gE2, or by FLAG tags on gI1 and gI2. We evaluated whether human IgG Fc binds to gE1, gE2, gE1/gI1, or gE2/gI2 using SPR (Biacore 3000). Equivalent amounts (~110 RU) of either gE1, gE2, gE1/gI1, or gE2/gI2 were captured to an anti-HIS coupled CM5 chip via binding of their C-terminal His tags. Next, 750 nM of non-immune human IgG was injected across the chip surface for 120-240s. The RU from a control flow cell where only non-immune human IgG was injected (no gE) was subtracted from each curve.

Biosensor assays were also used to assess whether mAbs block human IgG Fc binding to gE2/gI2. Purified gE2/gI2 heterodimer (~200 RU) was captured on a CM5 anti-His chip via a C-terminal His-tag on gE2 with a flow rate of 10 µl/min in HBS-EP+ (10 mM HEPES, 150 mM NaCl, three mM EDTA, pH 7.4) using a Biacore 1k+ instrument (Cytiva). Murine non-immune IgG (IgG from mouse serum I5381, Sigma Aldrich) and mAbs E1 and B1E6 at concentrations of 200 µg/ml, flow rate 20 µl/min in HBS-EP+ or HBS-EP+ alone, were then injected across separate flow cells for 120 sec over the captured gE2/gI2 heterodimer. Human IgG1 isotype (InVivoMab, VWR) at a concentration of 400 µg/mL, with a flow rate of 10 µL/min in HBS-EP+, was then injected over all flow cells for 60 seconds. The RU from a control flow cell where only human IgG1 was injected (no gE2/gI2, no Abs/mAbs) was subtracted from each curve. A 100 mM glycine solution, pH 2.0, was used to regenerate the CM5 anti-His chip to baseline levels. The % blocking of each antibody was calculated near the end of the human IgG1 injection (~60 sec) using the following equation: % blocking = 100 - [(RUTest/RUBuffer alone)*100].

Analyses were done using GraphPad Prism 10. Details of the statistical method used are included in the text when P values are noted, or in the figure legends. P values <0.05 were considered statistically significant.

The use of de-identified serum samples provided by colleagues from another USA institution was considered Exempt by the University of Pennsylvania IRB.

We developed an ADCC assay to measure antibodies in human sera that target immunogens included in a trivalent gC2/gD2/gE2 mRNA genital herpes vaccine. BioNTech SE has developed a modified version of this trivalent vaccine, which is currently in human trials (ClinicalTrials.gov; NCT05432583) (36). ADCC assays require target cells expressing surface antigens. IgG antibodies bind to the surface antigens by their Fv domains, and their Fc domains bind to IgG Fc receptors (FcγR) on the NK effector cells.

We used HEK293 cells transiently transfected with HSV-2 DNA expressing full-length gC2, gD2, and gE2 as target cells. The antibody source was either human convalescent serum (HCS) obtained from HSV-2 seropositive subjects with 0–14 annual recurrences of genital herpes (provided by Anna Wald and Christine Johnston), or human IgG, pooled and purified from serum of multiple donors (Sigma). As effector cells, we used a human NK cell line, NK-92, which expresses the high-affinity homozygous (176V/V) IgG FcγRIIIA (CD16A) (29, 30). These cells were chosen because in humans, NK cells that express FcγRIIIA are the effector cells that most commonly mediate ADCC (37–39). Additionally, by using an NK cell line, we avoided donor-to-donor variability that can occur with NK cells from multiple donors. We used flow cytometry to quantify surface expression of CD107a, a marker of NK cell activation (Figure 1A) (40–42).

First, we evaluated the efficiency of HEK293 cell transfection by gC2, gD2, and gE2 DNA. The three plasmids expressing full-length glycoproteins were added in equal quantities, and surface expression of the glycoproteins was evaluated by flow cytometry. Expression of the three glycoproteins was unequal, with 32% of cells expressing gD2, 23% gE2, and 6% gC2 (ratio 5:4:1 of gD2:gE2:gC2) (Figure 1B). For this reason, we assessed the expression levels for transfected HEK cells in each series of experiments in this study. However, this ratio is comparable to the expression of these glycoproteins in infected cells (43). We chose to use transfected cells as ADCC targets rather than infected cells because we could control cell viability more effectively after transfection than after infection. In addition, transfected cells allow us to determine whether gC2, gD2, and gE2 are targets of ADCC in HCS, a question that cannot be addressed with infected cells. We used two positive controls for activating NK cells. One involved stimulating NK cells with phorbol 12-myristate 13-acetate (PMA), a potent inducer of surface CD107a expression on NK cells (Figure 1C), while the other used SKOV3 cells that constitutively express HER2 receptors that serve as a target for Trastuzumab (Herceptin) antibody, followed by adding NK-92 cells (Figure 1D) (44). Both controls demonstrated robust expression of CD107a on NK cells. To measure ADCC, we initially used human IgG pooled from multiple donors as the source of antibody. As a negative control for ADCC, we incubated gC2/gD2/gE2-transfected HEK cells with NK cells in the absence of antibody (Figure 1E, No IgG). The negative control was then used to determine an endpoint titer for the pooled human IgG, which was ≥ 1:6400 (Figure 1E, dotted line). We conclude that this assay is effective for measuring NK cell activation, and that gC2, gD2, and/or gE2 are ADCC targets for antibodies in pooled human sera.

We next evaluated NK cell CD107a expression using 10 sera obtained from HSV-2-infected individuals as the source of antibody. The transfection efficiency was determined (Figure 2A). It varied little from the prior experiment (Figure 1B). The positive controls for NK cell activation, using PMA and SKOV3 cells (Supplementary Table 1), also varied little from those shown in Figures 1C, D. The negative control had gC2, gD2, gE2-transfected HEK cells plus NK-92 cells with no antibody. In contrast, the positive control used pooled human IgG at 1:100 dilution (Figure 2B). The 10 human convalescent sera were evaluated at dilutions ranging from 1:200 to 1:3200. The endpoint titer for the 10 HCS was ≥1:3200 (Figure 2B). We conclude that HCS contains high titers of antibodies to one or more of the vaccine antigens, gC2, gD2, and gE2, and that target cells coated with these antibodies induce CD107a expression on NK cells.

As an additional negative control, we evaluated CD107a expression using three HSV-1/-2 double seronegative human sera as a source of antibodies. We expected that serial dilutions (1:200 to 1:6400) of the seronegative sera would not stimulate NK cell CD107a expression beyond background levels noted when NK cells were exposed to transfected cells without antibody. The transfection efficiency for gC2/gD2/gE2 showed the same general pattern as in prior experiments (Figure 2C), as did the controls for NK cell CD107a expression (Supplementary Table 1). As expected, the human seronegative sera did not induce NK cell CD107a expression above background levels of the negative control (Figure 2D, No IgG). We conclude that HSV-2 seropositive HCS stimulates NK cell CD107a expression, while seronegative serum does not.

We compiled the results of the 10 HCS samples and the pooled human IgG to display the number of annual genital recurrences, years since the onset of genital disease, gC2, gD2, and gE2 IgG binding (ELISA) endpoint titers, as well as ADCC endpoint titers, as determined by NK cell CD107a expression (Table 1). All 10 subjects produced IgG binding antibodies to gC2, gD2, and gE2, and all had ADCC titers ranging from 1:3,200 to 1:12,800. We did not detect a correlation between the number of genital recurrences per year and ADCC endpoint titers (r = 0.1144, P value not significant); however, only one of the 10 HCS samples was from a subject with fewer than six annual recurrences. Perhaps the narrow range of recurrent genital infection episodes accounts for the lack of correlation. We conclude that gC2, gD2, and/or gE2 serve as targets of ADCC.

Prior reports indicated that HSV-2 gE2 differs from HSV-1 gE1 in that gE1 binds human IgG Fc as a monomer or as a heterodimer with HSV-1 gI1. In contrast, HSV-2 gE2 only binds human IgG Fc as a heterodimer with HSV-2 gI2 (28). We used surface plasmon resonance (SPR) to confirm these observations. HSV gE1 or gE2 was captured on the SPR chip. Human IgG purified from an HSV-1/2 double-seronegative donor serum was then flowed over the chip. Human IgG bound to gE1 (Figure 3A) but not to gE2 (Figure 3B). Next, we captured the HSV-1 gE1/gI1 complex or the HSV-2 gE2/gI2 complex on the SPR chip and added human IgG. Both gE1/gI1 and gE2/gI2 bound non-immune human IgG, although the kinetics of binding differed, with gE2/gI2 showing faster on and off rates than gE1/gI1 (Figures 3C, D). These results confirm those of Galli et al. and indicate that to assess the importance of gE2 in immune evasion, cells need to be transfected with both gE2 and gI2, as performed in the ensuing experiments (28).

One of our goals was to determine whether gE2/gI2 inhibits ADCC by binding the IgG Fc domain of antibodies that target HSV-2 antigens expressed on transfected cells. We postulated that wild-type gE2/gI2 (gE2_WT) would reduce NK cell CD107a expression compared with gE2 mutant (gE2_MUT)/gI2 because of antibody bipolar bridging, where gE2_WT binds the IgG Fc domain of the antibody that is bound to HSV-2 antigens by the IgG Fv domain (antibody bipolar bridging) (Figure 4A) (17). We prepared three gE2 mutants, two of which were based on published reports describing gE2 mutations that disrupt IgG Fc binding to gE2/gI2, and one gE2 mutation was at a location homologous to a site in HSV-1 gE1 that abolished IgG Fc binding (28, 45) (Supplementary Figures 1A, B). We confirmed that all three mutants were expressed at the cell surface when co-transfected with gD2 and gI2 (Supplementary Figure 2A), and that the three gE2 mutants failed to bind human non-immune IgG Fc (Supplementary Figure 2B). We selected gE2_MUT#2 for further studies of NK cell CD107a expression.

HEK cells were transfected with gD2, gI2, and gE2_WT or gE2_MUT#2 DNA. We omitted gC2 DNA due to difficulties in achieving good surface expression of gC2 when used in combination with the other three DNA plasmids. Transfection efficiency was assessed using antibodies to detect gD2, gI2, and gE2_WT or gE2_MUT#2 expression. Two different mouse monoclonal Abs (mAbs) were used for gE2 detection: B1E6, which targets the Fc binding domain, and E1, which recognizes an epitope on gE2 not involved in IgG Fc binding. The binding of either mouse mAb to gE2 reflects binding by their IgG Fv domain because the IgG Fc domain of mouse IgG does not bind to gE or gE/gI of HSV-1 or HSV-2 (28, 46). Both mAbs recognized gE2_WT, but only E1 bound to gE2_MUT#2 (Figure 4B). The same 10 HSV-2 HCS as in Figure 2 were used to evaluate gE2/gI2 inhibition of NK cell CD107a expression by comparing ADCC endpoint titers when gD2/gE2_WT/gI2 was the target cell compared with gD2/gE2_MUT/gI2. NK cell CD107a expression was higher in cells transfected with the gE2_MUT#2 than gE2_WT (Figure 4C, Table 2). The geometric mean endpoint ADCC titer was 1:985 for cells transfected with gE2_WT, compared to 1:6400 for cells transfected with gE2_MUT#2, representing a 6.5-fold difference (P<0.01, by two-tailed Mann-Whitney test). This increase in titer occurred despite high titers of gE2/gI2 IgG binding antibodies (as measured by ELISA) in the HCS (Table 2). This result suggests that gE2/gI2 antibodies produced by infection do not prevent gE2_WT/gI2 from inhibiting ADCC.

We evaluated NK cell CD107a expression using HSV-1/2 double-seronegative human sera as the source of antibody, with the expectation that the seronegative human sera would not stimulate NK cell CD107a expression. We confirmed the transfection efficiency (Figure 4D). As expected, the seronegative human sera did not trigger expression of NK cell CD107a using either gD2/gE2_WT/gI2, or gD2/gE2_MUT#2/gI2 transfected cells as targets (Figure 4E). We conclude that a mutation that abrogates explicitly IgG Fc binding leads to higher NK cell CD107a expression, supporting our hypothesis that gE2_WT/gI2 inhibits ADCC.

Another goal was to determine whether we can prevent gE2_WT/gI2 from inhibiting ADCC by using an antibody that targets the IgG Fc binding domain on gE2 to block its immune evasion activity. We reasoned that a positive result would support efforts to produce blocking antibodies by gE2 immunization. These experiments required two gE2 mAbs with differing properties, one that blocks IgG Fc binding to gE2 and another that does not. We postulated that the mAb that binds to gE2 and blocks IgG Fc binding would enhance the level of ADCC, while the mAb that binds gE2 but does not block would have no effect.

We chose mAb B1E6 based on it not binding to gE2_MUT#2, suggesting that B1E6 recognizes an epitope involved in IgG Fc binding. A prior publication noted that the B1E6 epitope was detected in all but 2 of 2,400 HSV-2 clinical isolates, indicating that this epitope is highly conserved (47). We chose mAb E1 based on it binding to each of the gE2_MUTs, indicating that E1 does not interact with any of the epitopes modified by these mutations (Figure 4B; Supplementary Figure 2A). As a further approach, we evaluated whether these mAbs block IgG Fc binding to gE2/gI2 in a biosensor (surface plasmon resonance) assay. The gE2/gI2 complex was added to a nickel-coated chip via a gE2 His-tag, followed by the flow of non-immune mouse IgG, mAb B1E6, or mAb E1 over the chip, and then non-immune human IgG. B1E6 totally blocked non-immune human IgG Fc binding to gE2_MUT#2/gI2, while E1 and non-immune mouse IgG had little or no effect, respectively (Figure 5A; Supplementary Figures 2A, B). B1E6 and E1 meet our goal of having one mAb that blocks human IgG Fc binding and another, as a control, that fails to block (experimental model, Figure 5B).

We transfected HEK cells with gD2, gI2, and gE2_WT or gE2_MUT#2 and confirmed transfection efficiency (Figure 6A). B1E6 and E1 both bound to gE2_WT, while only E1 bound to gE2_MUT#2 (Figure 6A). Next, we spiked pooled human IgG with B1E6 or E1 (Figure 6B). We hypothesized that B1E6 would block IgG Fc binding to gE2_WT but not gE2_MUT#2 and would increase NK cell CD107a expression of cells transfected with gE2_WT but not gE2_MUT#2. In addition, we postulated that mAb E1 would not affect NK cell CD107a expression in cells transfected with gE2_WT or gE2_MUT#2. The first three sets of bars in Figure 6B (both left and right panels) are controls: going from left to right, the first set of bars (labeled No IgG) shows NK cell CD107a expression when HEK cells are transfected with gD2/gE2_WT/gI2, or gD2/gE2_MUT#2/gI2 in the absence of antibody. These controls establish the background NK cell CD107a expression in the absence of antibody. The second and third sets of bars (labeled B1E6 6400ng and E1 6400ng, respectively) show that mAbs B1E6 and E1, used at the highest concentration, do not induce NK cell CD107a expression, because mouse IgG Fc does not engage with the human FcγRIIIA on NK cells (48). The remaining sets of bars involve pooled human IgG as the source of antibodies. The first of these, labeled IgG + E1 6400ng, shows that mAb E1, even at the high concentration of 6400ng, when added to pooled human IgG, does not affect NK cell CD107a expression of cells transfected with gD2/gE2_WT/gI2 or gD2/gE2_MUT#2/gI2 (compare the fourth bar with the one immediately to the right labeled IgG alone). The remaining six bars show the result for NK cell CD107a expression when pooled human IgG is spiked with increasing concentrations of B1E6. As the B1E6 concentration increases from 200ng to 6400ng, NK cell CD107a expression increases for cells transfected with gD2/gE2_WT/gI2, while B1E6 does not affect NK cell CD107a expression of cells transfected with gD2/gE2_MUT#2/gI2 DNA. We conclude that gE2 mAb B1E6, which blocks IgG Fc binding, increases NK cell CD107a expression when added to pooled human IgG (P<0.05 compared to no B1E6), and that gE2 mAb E1, which does not block IgG Fc binding, does not increase NK cell CD107a expression. These results support the hypothesis that antibodies that antibodies that target the Fc binding domain on gE2 can bind and block its immune evasion properties.

We next evaluated the effect of spiking each of the 10 HSV-2 HCS with B1E6, instead of pooled human IgG. Transfection efficiency of the HEK cells was confirmed (Figure 6C). As with pooled human IgG, B1E6 increased NK cell CD107a expression in cells transfected with gE2_WT DNA (P<0.0001) (Figure 6D, left panel). In contrast, B1E6 had no effect when cells were transfected with gE2_MUT#2 DNA (Figure 6D, right panel). We conclude that gE2_WT/gI2 blocks ADCC, and that mAb B1E6 can overcome this block to enhance ADCC responses. These 10 HCS have high titers of gE2/gI2 binding antibodies (Table 2). However, the HCS is only capable of preventing gE2/gI2 from inhibiting ADCC when B1E6 is added. These results indicate that blocking antibodies targeting the IgG Fc binding domains on gE2/gI2 are not present in HCS sera at sufficient titer to block IgG Fc binding, perhaps because the immune evasion domain(s) on gE2/gI2 are not very immunogenic.