Patients ZK016 and ZK018 were selected from a cohort of ZIKV patients enrolled at the Hope Clinic of the Emory Vaccine Center in an Emory University IRB approved study of emerging infectious diseases (IRB00022371; 2016–2019). Whole blood, PBMCs and plasma samples were collected and provided to our laboratory for experimentation and analysis.

The virus strains used in this study were ZIKV PRVABC59, DENV1 WP, DENV2 Tonga/74, DENV3 Sleman/78 and DENV4 Dominica/84. All viral stocks were passaged in Vero cells as previously described [13,33], titered and stored at −80 °C. The ZIKV lysate was prepared as described previously [13] and the protein concentration was quantified using Bradford reagent. The ZIKV recombinant E and NS1 proteins were purchased from Meridian Life Science (Memphis, TN, USA; R01635, R01636). The DENV recombinant E and NS1 proteins were purchased from CTK Biotech (San Diego, CA, USA; A2301 DENV1-E, A2302 DENV2-E, A2303 DENV3-E, A2304 DENV4-E, A2311 DENV1-NS1, A2312 DENV2-NS1, A2313 DENV3-NS1, and A2314 DENV4-NS1).

The frequency of ZIKV-specific antibody secreting cells (ASCs) was determined by ELISpot assay as previously described [31]. To determine the ZIKV-specific ASC response, 96-well ELISpot filter plates (MilliporeSigma; Burlington, MA, USA; MSHAN4B50) were coated with 10 µg/well of either ZIKV E or NS1. To capture the total IgG ASC response, wells were coated with 5 µg/mL of donkey anti-human IgG antibody (Jackson ImmunoResearch; West Grove, PA, USA; 709-005-149).

PBMCs obtained from the two patients were stained with titrated amounts of CD19-FITC (BD; clone HIB19), CD3-Pacific Blue (BD; clone SP34-2), CD20-PECy7 (BD; clone L27), CD27-APC (eBiosciences; clone O323) and CD38-PE (BD; clone HIT2). The plasmablast population was defined as CD3− CD19+ CD20−/low CD27+ CD38+ lymphocytes and its frequency analyzed using FlowJo software. The single-cell sorting of plasmablasts was carried out using the FACSAriaII sorter (Becton Dickinson; Franklin Lakes, NJ, USA) at the Emory University Pediatrics Flow Cytometry Core Facility under negative air pressure. The cells were sorted into a 96-well PCR plate as described in [34,35], rapidly frozen on dry ice and stored at −80 °C for subsequent cDNA synthesis.

The mAbs were produced from the single-cell sorted plasmablasts as described [34,36], with a few modifications at the reverse transcription step. The single-cell sorted plates were thawed on ice and Triton X-100 (Sigma-Aldrich; St. Louis, MO, USA; T8787-50ML) was added at a final concentration of 0.2% to lyse cells. The synthesis of cDNA was performed using 1 µM gene specific anti-sense primer cocktail (heavy, kappa and lambda) and Sensiscript reverse transcriptase (Qiagen; Venlo, Netherlands; 205211). Ig heavy chain and light chain (kappa and lambda) variable domain were amplified by nested PCR using RT-PCR primers listed previously [34]. For sequencing of the heavy and light chain variable domains, another round of nested PCR was performed with an added 5′ M13R sequence (5′-AACAGCTATGACCATG-3′) to each sense primers. Both PCRs were performed using Hot Start Taq Plus Master Mix (Qiagen; Venlo, Netherlands; 203643) with 200 nM primer concentration. The cloning PCR was performed using Phusion Hot Start II DNA Polymerase (NEB; Ipswich, MA, USA; F-549) and V and J gene family specific primers containing appropriate restriction sites as described [34]. To protect the variable domain that has internal restriction sites from being truncated, the cloning PCR was modified using 50 µM 5-methyl-dCTP (NEB; Ipswich, MA, USA; N0356S) and 150 µM dCTP. The heavy chain variable domains, irrespective of the original isotype, were cloned into human IgG1 heavy chain expression vector, while the light chain variable domains were cloned into their respective human light chain (kappa or lambda) expression vector [34]. These vectors were then co-transfected into Expi293F cells according to the manufacturer’s protocol (Thermo Fisher Scientific; Waltham, MA, USA; A14527). Antibodies generated in the cell culture supernatants were purified using Protein A agarose beads (Pierce; Waltham, MA, USA 20333).

The variable region of heavy chain (VH) and light chain (VL) were analyzed for germline variable domain gene usage, complementarity determining region 3 (CDR3) lengths and somatic hypermutation (SHM) using ImMunoGeneTics (IMGT) database and IgBLAST tool. SHM frequencies in the VH and VL sequences were analyzed by matching their sequence to the closest germline sequence and determining the total number of silent (S) and replacement (R) mutations from the framework region 1 (FR1) till before the CDR3. The CDR3 mutations were not included in the analysis as the junction region of the reference germline sequences were estimated with low confidence. Clonality analysis was performed by sequence alignment of the rearranged VH and VL chains with matching V and J gene usage with identical junctional diversity.

ZIKV-specific IgM antibodies in patient and healthy donor plasma were measured by the Zika IgM Antibody Capture Enzyme-Linked Immunosorbent Assay (Zika MAC-ELISA) as described [37,38]. To detect ZIKV-specific IgG antibodies, the Zika MAC-ELISA was adapted from published methods [39] with slight modifications: plates were coated with anti-human IgG (KPL Seracare; Milford, MA, USA; 01-10-06) at a dilution of 1:500, followed by subsequent steps as described [37,38].

To determine plasma/mAb binding to ZIKV lysate or ZIKV/DENV E and NS1 proteins, Nunc Maxisorp ELISA plates (eBioscience; Waltham, MA, USA; 44-2404) were coated with the appropriate antigen diluted in PBS overnight at 4 °C: 50 µg/mL of ZIKV lysate, or 1 µg/mL recombinant protein respectively. Plates were then washed with PBS containing 0.05% Tween-20 (PBS-T) and blocked with PBS-T containing 10% FBS (PBS-T-FBS) for 1.5 h. Serially diluted plasma in PBS-T-FBS was added to the plates for 1.5 h. Plates were washed with PBS-T and peroxidase-conjugated anti-human IgG (Jackson ImmunoResearch; West Grove, PA, USA; 109-036-098) was added for 1.5 h. The plates were then washed with PBS-T followed by PBS and developed using an o-phenylenediamine substrate (Sigma; Burlington, MA, USA; P8787) for absorbance detection at 490 nm. The whole virus capture ELISA for ZIKV and DENV was performed as described [13]. The pan-flavivirus 4G2 antibody used to capture virions was generated using the D1-4G2-4-15 hybridoma (ATCC; Manassas, VA, USA; HB-112).

ZIKV E-specific mAbs that target the conserved FL domain in the ZIKV E protein were determined by competing them against FL-specific mouse mAbs 4G2 (ATCC; Manassas, VA, USA HB-112) [40] and E18 (BEI Resources; Manassas, VA, USA; NR-10135) [41] using an ELISA-based assay. ELISA plates were coated with 1 µg/mL of ZIKV E protein diluted in PBS overnight at 4 °C. The plates were washed and blocked as described above. E18 and 4G2 were incubated with either ZIKV E-specific mAbs (competitor mAb) or with PBS-T-FBS and added to the ELISA plate. After 1.5 h, the plates were washed and peroxidase-conjugated anti-mouse IgG (H + L) (KPL Seracare; Milford, MA, USA; 074-1806) was added. Later, the plates were developed with o-phenylenediamine substrate. Human mAbs that caused more than 50% reduction in the binding of mouse mAbs 4G2 and E18 to the ZIKV E protein compared to the no competition control were identified as FL-specific.

The virus neutralization abilities of patient plasma (heat inactivated) and mAbs were determined by the focus reduction neutralization test (FRNT) as previously described [33]. Serially diluted plasma or mAbs were incubated with 60–100 focus forming units of ZIKV, DENV1, DENV2, DENV3 or DENV4 for 1 h at 37 °C. The virus-antibody mixture was added to a Vero cell monolayer for 1 h at 37 °C. An overlay of 1.5% methylcellulose was added to the cells and followed by a 3 day incubation at 37 °C. The overlay was then removed, cells washed and fixed with a 1:1 methanol and acetone solution. The viral foci were stained using 4G2 antibody for 2 h followed by HRP conjugated anti-mouse IgG (Cell Signaling; 7076S) for 1 h and developed using TrueBlue peroxide substrate (KPL Seracare; Milford, MA, USA; 50-78-02). The foci were counted and analyzed to determine the dilution factor or concentration at which 50% reduction viral foci (FRNT50) was observed in the test samples compared to the virus only control.

To assess the infection enhancement capacity of infected plasma and mAbs, an antibody dependent enhancement (ADE) assay was performed as previously described [13,33]. Briefly, serially diluted mAbs or heat inactivated plasma were incubated with 0.5 MOI of DENV2 for 1 h at 37 °C. The immune complexes were mixed with U937 cells (ATCC; Manassas, VA, USA; CRL-1593.2) in a 96-well plate for 24 h at 37 °C. The cells were then washed and intracellularly stained with 4G2 antibody using Fixation/Permeabilization solution kit (Becton Dickinson; Franklin Lakes, NJ, USA; 554714) followed by a secondary anti-mouse IgG AF488 (Life Technologies; Carlsbad, CA, USA; A11029). The percentage of infected (4G2+) cells was determined by flow cytometry using the LSR-II (Becton Dickinson; Franklin Lakes, NJ, USA) flow cytometer followed by analysis using FlowJo software v10.3.

The two patients ZK016 and ZK018 included in this study were United States citizens who acquired ZIKV infections in the summer of 2016 during their residence (ZK016) or travels (ZK018) abroad. Details including patient demographics, timing of phlebotomies as days post onset of symptoms (DPO) and prior flavivirus history are summarized (Table 1). Plasma samples obtained at 3 and 7 DPO (ZK016), and 8 DPO (ZK018) were tested by ELISA and FRNT to determine ZIKV binding and neutralization titers (Figure 1). In the case of ZK016, ZIKV-specific IgM and IgG titers increased substantially between days 3 and 7, with higher anti-ZIKV IgG titers compared to IgM as early as 3 DPO (Figure 1A). In comparison, ZK018 appeared to have a delayed IgG response as higher ZIKV-specific IgM titers were observed at 8 DPO compared to anti-ZIKV IgG titers. Overall, ZK016 had markedly higher ZIKV-specific serum IgG than ZK018 (Figure 1B). As shown in Figure 1B, the day 7 IgG endpoint titer of ZK016 to ZIKV lysate was 100-fold greater than that of ZK018 at 8 DPO.

ZIKV-specific neutralizing antibodies were detected at 7 DPO and 8 DPO for patients ZK016 and ZK018 respectively. At 3 DPO, neutralization antibody titers were detected against all the 4 serotypes of DENV, but not against ZIKV in patient ZK016 despite seroconversion. In contrast, ZK018 plasma at 8 DPO did not neutralize DENV1 and DENV2, and showed low FRNT titers against DENV3 and DENV4 (Figure 1C). The early emergence of ZIKV-specific serum IgG, and comparable ZIKV and DENV neutralization titers taken together suggest that ZIKV infection primarily induced a recall response in patient ZK016. On the other hand, the high ZIKV-specific IgM levels, delayed emergence of ZIKV IgG and low cross-neutralizing DENV titers suggest that ZK018 may have mounted a primary response to ZIKV infection.

At approximately one week after symptom onset, robust expansions of plasmablasts were apparent in circulation for both patients, amounting to 20% (ZK016) and 15% (ZK018) of their respective peripheral B cell populations (Figure 2A). As shown for ZK016 in Figure 2B, ZIKV E and NS1-specific, as well as ZIKV lysate-specific (data not shown) plasmablasts were observed at the DPO 7/8 time points for both patients by ELISpot. Patient plasmablasts were single-cell sorted to generate mAbs for B cell repertoire analysis and functional studies as outlined in the workflow chart (Figure 2A).

During a recall response, memory B cells (MBCs) generated by a previous infection or vaccination can reactivate, resulting in expansions of plasmablasts that share a clonal origin. As shown in Figure 3A, 19% of the immunoglobulin sequences obtained from patient ZK016 were clonally related, whereas zero clonal expansions were detected in ZK018 plasmablasts. Analysis of VH somatic hypermutation (SHM) frequencies revealed no striking differences between patient ZK016 (average = 17.4; median = 15) and ZK018 (average = 15.3; median = 13). Nor were significant differences observed between the ZIKV patients (average = 16.4) compared to our previously reported SHM frequencies from dengue or influenza patient plasmablasts (Figure 3B) [33]. However, of note, 6 of the 33 mAbs generated from ZK018 plasmablasts were completely un-mutated and an additional 3 had 2 mutations or fewer in the VH region. In contrast, ZK016 plasmablasts SHM frequencies ranged from 5 to 46 per sequence, none of these mAbs were un-mutated (Figure 3C, Supplementary Table S1). A majority of mAbs with higher numbers of variable region mutations were cross-reactive to both ZIKV and DENV, while the 6 germline mAbs were found to be ZIKV-specific (Figure 3C, Supplementary Table S1). We also analyzed heavy chain CDR3 lengths and found that patient ZK016 plasmablasts had a normal distribution with 5 sequences representing the peak CDR3 length of 16 aa (Figure 3D). On the other hand, ZK018 CDR3 lengths more closely resembled a multimodal distribution, with 3 sequences each at the 13 aa, 14 aa and 23 aa modes (Figure 3D).

To dissect the plasmablast response induced by ZIKV infection, we generated a panel of 59 mAbs from the two patients combined: 26 mAbs from ZK016 (derived from 18 IgG and 8 IgA plasmablasts) and 33 mAbs from ZK018 (derived from 23 IgG and 10 IgA plasmablasts). Of the 26 mAbs generated from ZK016, 20 (77%) bound to ZIKV E and/or whole virus. All ZK016 mAbs that bound ZIKV antigens also cross-reacted with DENV, largely with equal or greater potency (Figure 4). By contrast, a smaller frequency of ZK018 mAbs, 22 of the 33 mAb total (67%) were ZIKV E or whole virus-reactive, of which 6 exhibited poor binding (minimum effective concentrations = 1–10 µg/mL) to these antigens. Additionally, only 9 mAbs (27%) from ZK018 were cross-reactive to DENV. A majority of cross-reactive mAbs from both patients were mapped to the highly conserved fusion loop (FL) (Supplementary Figure S2). Taken together, these data show that the primary response generated by ZK018 comprised non-affinity matured, de novo plasmablasts that were largely weakly-reactive and ZIKV-specific. In contrast, for ZK016 the selection and reactivation of pre-existing cross-reactive MBCs may have resulted in a predominantly high affinity, cross-reactive plasmablast repertoire.

All ZK016 and ZK018 mAbs were tested against ZIKV and DENV1-4 for in vitro neutralization capacity (Figure 5, Supplementary Figure S1). For ZK016, only one mAb sufficiently neutralized ZIKV to yield an FRNT50 value within the range of concentrations tested. This mAb, P1-G3, poorly neutralized ZIKV with an FRNT50 value of 8.74 µg/mL. Interestingly, despite the limited ZIKV neutralization capacity of ZK016 mAbs, an additional 11 mAbs besides P1-G3 from this donor cross-neutralized DENV2. These mAbs neutralized DENV with >10-fold greater potency, with FRNT50 values ranging between 50 and 800 ng/mL. This neutralization bias towards DENV was not observed in mAbs from ZK018; 14 of the 33 mAbs neutralized ZIKV, and only 4 of these ZIKV-neutralizers also neutralized DENV. ZK018 mAbs neutralized ZIKV at comparable or slightly higher potency than P1-G3, with FRNT50 values ranging from 2.1 to 6.66 µg/mL. The 4 mAbs that neutralized DENV1-4 were all FL-specific (Figure 5, Supplementary Figure S1). These findings suggest that a number of DENV-biased, ZIKV-reactive B cell clones were selected for reactivation during the ZK016 recall response to ZIKV, resulting in an original antigenic sin (OAS) neutralization phenotype. In the absence of pre-existing dengue immunity, ZK018 generated a de novo plasmablast response selected primarily on the basis of reactivity towards ZIKV. Therefore, a greater number of ZK018 clones neutralized ZIKV overall, including a handful that also cross-neutralized DENV.

ADE is one of several hypothesized causes for the increased disease severity associated with secondary heterotypic dengue infections [43]. Recently, several groups have studied ADE in the context of ZIKV infection in efforts to understand whether pre-existing cross-reactive antibodies from a prior DENV exposure can affect ZIKV infection, and vice-versa [13,23,44,45]. Given that both patients in our study had recently experienced ZIKV infections, we first examined whether plasma collected early after ZIKV exposure could enhance DENV2 infection through ADE. We tested ADE using an Fc-gamma receptor-bearing monocytic cell line, U937, and determined percent infection in the presence or absence of antibodies based on 4G2+ staining (Figure 6A). As shown in Figure 6B, 7 DPO plasma from ZK016 caused considerable ADE of DENV2, infecting over 30% of cells at the peak ADE titer. By contrast, 8 DPO plasma from patient ZK018 did not enhance DENV2 infection. To determine which types of B cell clones were contributing to the plasma ADE titers, we tested the ability of select mAbs from both patients to cause ADE. As expected, mAbs that strongly bound ZIKV E and whole virus caused ADE of DENV2, while mAbs that showed little to no reactivity to these antigens largely failed to enhance infection (Figure 6C). In general, FL-specific mAbs caused greater enhancement of DENV2 infection compared to non FL-specific mAbs, and a larger number of mAbs from ZK016 enhanced DENV2 infection compared to ZK018.

While a majority of mAbs generated from the two donors were either E-protein or whole virus-reactive, we also observed a few that bound NS1 protein. Of the 5 ZIKV NS1-specifc mAbs, two came from ZK016 and three from ZK018 (Figure 7A). Both ZK016 mAbs cross-reacted to ZIKV and DENV NS1 proteins, binding DENV NS1 at equal or higher potency compared to ZIKV NS1 (Figure 7B). Surprisingly, ZK018 NS1 mAbs were ZIKV-specific and did not bind DENV NS1, even though NS1 is known to be highly conserved in flaviviruses (Figure 7). Additionally, the ZIKV NS1-specific mAbs from ZK018 had little to no SHM (Supplementary Table S1). Given the difficulties in discerning ZIKV infections from DENV based on serology, these ZIKV-NS1 specific antibodies could be pursued as potential diagnostic tools for use early after ZIKV infection.