Longitudinal Analysis of the Human B Cell Response to Ebola Virus Infection

Mice

Female BALB/c mice, aged 6 to 8 weeks, were purchased from Charles River Laboratory. Upon arrival, mice were housed in microisolator cages in an animal biosafety level 4 containment area and provided chow and water ad libitum. Research was conducted under an IACUC approved protocol in compliance with the Animal Welfare Act, PHS Policy, and other Federal statutes and regulations relating to animals and experiments involving animals. The facility where this research was conducted is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care, International and adheres to principles stated in the Guide for the Care and Use of Laboratory Animals, National Research Council, 2011.

EBOV-specific ELISA

100 ng of recombinant protein (GP_ec, GP_cl, sGP, VP40, VP35, or VP30) or 50 ng NP N-terminal domain plus 50 ng NP C-terminal domain (Fusco et al., 2015, Bornholdt et al., 2013, Kirchdoerfer et al., 2015, Kirchdoerfer et al., 2016) per well was coated on ELISA plates overnight at 4 degrees in 100 μl of DPBS (pH 7.2). For peptide ELISAs, 0.5 μg of each peptide was coated per well in 100 μl of 0.1 M sodium bicarbonate, pH 9.6. After coating, plates were blocked for two hours at room temperature with 10% FCS in PBS plus 0.05% Tween-20 (block buffer). For indirect capture assays, the capture antibody was coated overnight at 100 ng per well, followed by blocking, and incubation for two hours with 100 ng/well of GP_ec. Plates were washed with PBS/0.05% Tween (PBST), and three-fold serial dilutions of serum or monoclonal antibodies were added to each well in 100 μl of block buffer for 90 minutes. After washing with PBST, bound antibodies were detected by incubation for 90 minutes with anti-human IgG, IgA, or IgM HRP detection antibodies diluted 1:1000 in block buffer. Biotinylated mAbs were made by labeling with NHS-LC-LC-biotin and detected using avidin-HRP diluted 1:5000 in block buffer. For detection of IgG subclass-specific responses, anti-human IgG1, IgG2, IgG3, or IgG4-HRP detection antibodies were used instead. Plates were washed with PBS/Tween, then PBS, followed by addition of o-phenylenediamine dihydrochloride substrate in phosphate-citrate buffer containing 0.01% hydrogen peroxide. Plates were allowed to develop for 7 minutes before adding 1 M HCl stop solution. The optical density (OD) of each well was read at 490 nm. ELISA titers are reported as the interpolated dilution of patient plasma where the cutoff OD value of 0.2 was reached.

Measurement of GP-specific IgG avidity

The concentration of plasma from each patient and time point which would give an optical density reading close to 1 was calculated from the results of the serial dilution ELISA. GP-coated plates were incubated with this plasma dilution, then washed for 5 minutes with either PBS or 8M Urea in PBS, prior to detection of bound IgG. Plates were then washed with PBS/Tween and secondary the antibody incubation and substrate addition steps were performed as usual. The avidity index was calculated for each plasma sample as the ratio of the OD on the urea wash plate divided by the OD of the same sample on the PBS wash plate. Similar results were seen when 5M Urea was used instead of 8M Urea in the wash step

Measurement of plasma CXCL13 levels

A commercial assay (BLC Human ProcartaPlex Simplex Kit; ThermoFisher) was used per the manufacturer’s instructions.

Measurement of EBOV-neutralization by patient plasma or monoclonal antibodies

Virus-specific neutralizing antibody responses were titrated essentially as previously described (Honnold et al., 2014). Briefly, plasma or antibodies were diluted serially in Minimal Essential Medium containing 5% FCS and 1X antibiotic/antimycotic (GIBCO) (MEM complete) and incubated 1 hour at 37°C with Ebola virus strain Kikwit. After incubation, the antibody-virus or plasma-virus mixture was added in duplicate to 6-well plates containing 90%–95% confluent monolayers of Vero E6 cells. Plates were incubated for 1 hour at 37°C with gentle rocking every 15 minutes. Following the incubation, wells were overlaid with 0.5% agarose in supplemented EBME media with 10% FCS and 2X Anti-Anti, and plates were incubated at 37°C, 5% CO₂ for 7 days. On day 7, cells were stained by the addition of a second overlay prepared as above containing 4%–5% neutral red. Plates were incubated for 18-24 hours at 37°C, 5% CO₂. The endpoint titer was determined to be the highest dilution with a 50% or greater or 80% or greater reduction (PRNT₅₀, PRNT₈₀) in the number of plaques observed in control wells. The assay limit of detection was calculated to be 5 plaque forming units per ml by this method.

Preparation of libraries and repertoire sequencing

RNA extracted from PBMCs at the peak of the B cell response was purified using TRIZOL (ThermoFisher). Infectious samples were handled under BSL-4 conditions until inactivated. For convalescent time points, RNA was extracted from PBMCs using the RNEasy Plus mini kit (QIAGEN). In the case of patient EVD9 at the discharge time point, RNA from purified antibody secreting cells (ASCs) was used instead of PBMC RNA. IGH libraries for high-throughput DNA sequencing were prepared from cDNA reverse transcribed from purified total RNA as previously described (Ellebedy et al., 2016). Briefly, cDNA was synthesized and rearranged IGH transcripts were amplified by PCR using BIOMED2 framework region 1 (FR1) IGHV forward primers (Langerak et al., 2012) and isotype-specific reverse primers for IgG, IgA and IgM located in the first constant region exon (Roskin et al., 2015). IGHV primers were multiplexed in single reactions for each isotype for each sample. Sample identity was encoded using 8-mer barcodes within the forward and reverse PCR primers. 2x300 paired-end sequencing of libraries was performed on the Illumina MiSeq instrument, using 600 cycle kits.

Processing of repertoire sequencing

Paired-end reads were merged using FLASH (Magoč and Salzberg, 2011) and samples were de-multiplexed based on 100%, full-length matches to the forward and reverse 8-mer barcodes. Primers were identified and trimmed allowing for a single mismatch within the primer (sequences without identifiable primers were discarded). Isotype subclasses were determined based on 100%, full length matches to the CH1 exon sequence upstream of primers (sequences without confirmed subclasses were discarded). Primer trimmed sequences were aligned against the human germline V, D, and J reference repertoires from IMGT/GENE-DB (Giudicelli et al., 2005) using a local installation of IgBLAST (Ye et al., 2013). IgBLAST output was used to determine germline gene utilization, to delimit complementarity determining regions (CDRs) and FR regions (IMGT definitions), define mutational changes and to determine rearrangement productivity. Non-productive (VDJs harboring stop codons or with out-of-frame IGHJ gene segments) and artificial or artifactual PCR amplicons were excluded from analysis. Artificial amplicons were identified based on the pattern of CDR3 nucleotide sequence association across reads utilizing disparate IGHVs.

Clonal relationships between rearranged genes in the repertoire sequencing were inferred using single linkage clustering. For each subject, all IGH that utilized the same IGHV and IGHJ and which had the same CDR3 lengths were clustered based on their CDR3 nucleotide sequence at a 90% identity threshold.

Inference of mAb lineages in repertoire sequencing

The membership of an mAb IGH to a clonal lineage from the repertoire sequencing was inferred using cd-hit (Fu et al., 2012) clustering of the CDR3 nucleotide sequences from the mAb to all IGH combined across the subjects that utilized the same IGHV, IGHJ and had same CDR3 length. Clustering was nucleotide sequence based at a 90% identity threshold, without gaps. Unlike clonal lineage inferences, the clustering of the mAbs was not restricted to within a subject’s individual repertoire but was extended across all subjects to test for potential convergent responses that may have arisen independently in the different subjects.

Repertoire sequencing data analysis

Analysis was performed and visualized in R (R Development Core Team, 2017) using R base packages for statistical analysis and the ggplot2 package for visualization (Wickham and Chang, 2008). Sequence logos were produced using Skylign (Wheeler et al., 2014) and multiple sequence alignment was performed using Clustal W version 2 (Larkin et al., 2007).

Analysis of ASC, ABC, and GP-binding B cell numbers:

For samples analyzed during the acute phase, ABC and ASC numbers were measured as described in a BSL-4 laboratory using an Accuri cytometer (McElroy et al., 2015). For convalescent time points, numbers of ASCs and ABCs were measured in whole blood from fresh samples. ASCs were gated as singlet lymphocytes that were CD3, CD19+, IgD, CD20, CD38^hi, CD27^hi singlets. ABCs were gated as singlet lymphocytes that were CD3, CD19+, IgD, CD20+, CD71+ singlets. Erythrocytes were removed by treatment BD FACS lyse prior to running samples on a BD LSR II instrument. For staining GP-specific cells, GP_cl and GP_ec were labeled with Alexa Fluor 488 and Alexa Fluor 647, respectively, using amine-reactive protein labeling kits. The gate used for GP-binding cells was CD3, CD19+, IgD, CD20+, GP_cl+ live singlets.

Cell sorting for mAb cloning

PBMC samples from the following time points were sorted at Emory University for mAb cloning: EVD2 1 month, 6 months, and 10 months (MPER sort); EVD5 1 month and 6 months; EVD9 6 months. The following B cell subsets were single cell sorted at Emory for antibody cloning: i) GP-binding B cells (gate: live, singlets, CD3, CD20+, CD19+, IgD, GP_cl+ or GP_ec+). ii) GP-negative activated B cells (gate: live, singlets, CD3, CD20+, CD19+, IgD, GP_cl and GP_ec, CD71+). iii) total ASCs (gate: live, singlets, CD3, CD20, CD19+, CD38^hi, CD71+, IgD). GP-binding B cells were sorted from all time points; GP-negative ABCs were sorted from EVD5 1 month; total ASCs were sorted from EVD2 1 month and EVD5 1 month. Cells were sorted into lysis buffer in 96-well plates for cloning of antibody heavy and light chain genes as described (Smith et al., 2009). For sorting MPER-peptide specific B cells a peptide spanning EBOV GP residues 630-649 was synthesized with a C-terminal cysteine and conjugated to APC using a commercial kit. MPER-binding B cells were sorted from the 10 month sample from EVD2 using this probe.

PBMC samples from the following time points were sorted at Atreca for next generation sequencing of paired heavy and light chain sequences: EVD5 6 months, EVD9 1 month, EVD15 discharge. Cell sorts at Atreca were performed as at Emory except that only GP_cl was used as the sort antigen. Bulk-sorted GP-binding B cells were cultured for 4 days in IMDM medium (Invitrogen) in the presence of FBS, Normocin, IL-2 (PeproTech), IL-21 (PeproTech), rCD40 ligand (R&D Systems), and His-Tag antibodies (R&D Systems) prior to single cell sorting. Paired chain antibody sequencing was carried out on cells sorted into 384-well plates at one cell per well by applying Immune Repertoire Capture technology as previously described (DeFalco et al., 2018).

Cloning of antibody heavy and light chain genes by PCR

For sorts done at Emory, mAb heavy and light chain genes were amplified by multiplex RT-PCR from individual sorted B cells or plasmablasts and cloned into human Igγ1, human Igκ, or human Igλ expression vectors as described (Smith et al., 2009), with minor modifications as follows: i) to amplify antibodies from IgA and IgM expressing cells, separate nested PCRs were run using constant region primers for IgM (Smith et al., 2009) and IgA (Benckert et al., 2011) instead of IgG. ii) an M13 Reverse primer (CAGGAAACAGCTATGACC) was added to the 5′ terminus of all 5′ primers used in the nested PCR in order non-specifically increase primer annealing to the template and increase the PCR efficiency (Regier and Shi, 2005). iii) to increase miniprep plasmid yields and reduce satellite colonies after transformations, the AgeI-HindIII fragment of vectors Abvec-hIgG1, AbVec-hIgKappa, and AbVec-hIgLambda containing the promoter and ORF regions were subcloned into the vector backbone of pcDNA6.2-GW/EmGFP-miR (ThermoFisher), which contains a pUC origin of replication and spectinomycin resistance. A WPRE element was cloned from vector pNL(CMV)EGFP/CMV/WPREΔU3 (Ou et al., 2012), a gift from Jakob Reiser, and placed downstream of the antibody ORF region for increased antibody yield in transfections. The new vectors with the pcDNA6.2 backbone and WPRE elements are referred to as AbVec6W vectors.

Cloning from paired heavy and light chain sequences

A total of 486 unique antibody sequences were identified by next generation sequencing, including 425 from EVD5 day 263, 30 from EVD9 day 71, and 31 from EVD15 day 14. Antibody heavy and light chains were initially synthesized for 10 mAbs chosen at random from each donor/time point to confirm that the majority of mAbs bound to GP. An additional 10 randomly selected mAbs were produced from EVD15 in order to ensure sampling of mAbs across the wide range of SHM levels observed in this donor. An additional 90 mAbs were produced from EVD5 because the initial set of 10 yielded one therapeutically-promising antibody. In total, we produced antibody for 100 out of 425 mAb sequences from EVD5 (24%), 10 out of 30 from EVD9 (33%), and 20 out of 31 sequences from EVD15 (65%). Codon-optimized DNA sequences encoding the heavy and light chain variable regions (FR1 through FR4) were synthesized with flanking restriction enzyme sites for subcloning. Codon optimization and DNA synthesis were performed by Genscript and Gen9. The synthesized heavy chain variable region genes were subcloned into AbVec6W-hIgG1 vectors and light chain genes were cloned into AbVec6W-hkappa or AbVec6W-hlambda.

Note on antibody nomenclature:

Antibody names contain the patient ID, the time point (in months after hospital discharge) when the PBMC sample was obtained that was used for B cell sorting, and the clone ID. For example, mAb 2.1.1B2 is from EVD2, from a B cell sorted from 1 month post-discharge, and was cloned from sort plate 1, well B2. mAbs sequenced at Atreca have clone IDs beginning with “c” (e.g., 5.1.c2618).

Analysis of monoclonal antibody sequences

For antibodies cloned by multiplex PCR directly from cells, the nested PCR sequence was analyzed to determine the original heavy chain constant region and IgG or IgA subclass, if applicable. mAb variable region sequences were determined from the consensus expression clones in vector AbVec6W. Portions of the sequence which were derived from PCR primers were removed before analysis. Antibody V(D)J usage and somatic hypermutation were analyzed using IMGT V-quest (Brochet et al., 2008). For antibodies cloned by next generation sequencing at Atreca, V(D)J assignment and mutation identification was performed using an implementation of SoDA (Volpe et al., 2006). IgG subclass was determined by top-scoring BLAST alignment to constant region nucleotide sequence (Camacho et al., 2009).

Screening neutralization assay using biologically contained EBOV

To identify neutralizing antibodies, a biologically contained EBOV, EbolaΔVP30 virus, containing the Renilla luciferase reporter gene (RenLuc) instead of the VP30 gene and expressing the Makona glycoprotein (GP; H.sapiens-tc/GIN/2014/Gueckedou-C07) instead of the Mayinga GP, was used as previously described (Halfmann et al., 2008). Briefly, in a 96-well format, EbolaΔVP30-RenLuc virus (10,000 focus-forming units) was incubated with 10 μg/mL of monoclonal antibody (mAb) for 2 hours at 37°C. The virus/mAb mixture was then inoculated onto Vero cells that stably express VP30 (VeroVP30 cells) to facilitate EbolaΔVP30-RenLuc virus replication. Three days later, luciferase expression, given as relative light units (a readout for virus infection and replication), was measured using live cell luciferase substrate, EnduRen (Promega), on a Tecan M1000 Pro plate reader. A neutralizing monoclonal antibody was determined as positive hit if the RLU was reduced by 90% compared to the negative control (EBOV VP35-specific mAb 5-69.3.2) (Takada et al., 2007). No antibodies in this study exhibited partial neutralization, which we consider to be a > 50% reduction in luciferase activity without reaching the 90% threshold.

Identification of viral escape mutants

The generation of antibody escape mutant viruses was previously described (Halfmann et al., 2008). Up to 15 individual plaque-picked escape mutant viruses were isolated and the GP was sequenced.

Monoclonal antibody production

The Expi293 expression system was used for antibody production according to the manufacturer’s instructions. A 2:1 ratio of light chain to heavy chain plasmid was used in transfections. mAb-containing supernatants from transfected expi293 cells were clarified by centrifugation and then incubated with protein A agarose resin (Genscript) in batch format overnight, followed by washing, elution, and buffer exchange into DPBS as described (Smith et al., 2009). Antibodies used in vivo were verified endotoxin-free using a commercial kit (ThermoFisher). Note that all antibodies produced in this study were expressed as human IgG1.

Generation of Jurkat cell lines stably expressing EBOV GP

A plasmid encoding EBOV GP (Makona strain) was obtained from Sino Biological. The GP gene was subcloned into cloned into previously described lentivirus vector pNL-sGPC (Kulkarni et al., 2014). The resulting plasmid is termed pNL-GP. Lentiviral particles were produced by transfection of expi293 cells with pNL-GP and packaging plasmids pMD2.G and psPax2 (gifts from Didier Trono). The viral stock was concentrated using PEG-it virus precipitation solution (System Biosciences). Jurkat cells were transduced at a multiplicity of infection of 10, and a stable expressing clone (Jurkat-GP) was obtained after three rounds of single cell sorting (FACSAria) into 96 well plates. Alexa 488 labeled mAb 5.6.1A2 (this study) was used to sort high expressing cells. 50% Jurkat-conditioned medium was added to wells to promote cell outgrowth after single cell sorting.

Cell-based antibody binding assays

For testing mAb binding to native GP, Jurkat-GP cells were incubated with mAbs at 4°C at a concentration of 5 ug/ml in DPBS supplemented with 2% fetal calf serum and 2 mM EDTA (FACS buffer). For testing plasma IgG binding, cells were uncubated with at 1:5000 dilution of plasma in FACS buffer. Cells were washed in FACS buffer and incubated with anti-human IgG-PE, washed again, and analyzed on a FACSCanto instrument.

For competition binding assays, unlabeled competitor antibody was preincubated with Jurkat-GP cells for 30 minutes at 100 ug/ml. Incubations were done on ice in FACS buffer containing 0.05% sodium azide to prevent GP internalization or shedding. Alexa 488 labeled test antibody was then added to a final concentration of 2 ug/ml without removing the competitor mAb. After 30 minutes, cells were washed and FITC channel fluorescence was assessed on a BD FACSCanto. The specific fluorescence in each condition was calculated by subtracting the geometric mean fluorescence in the FITC channel of cells incubated without Alexa 488 labeled antibody. The percentage of binding in the presence of each competitor was calculated as the specific FITC fluorescence in the presence of the competitor, divided by the specific fluorescence without competitor, times 100%.

Treatment of cells with thermolysin or dithiothreitol (DTT):

In some experiments, parental Jurkat E6-1 cells or Jurkat-GP cells were treated with thermolysin to remove the glycan cap and mucin domains of GP (Kaletsky et al., 2007). Cells were washed in DPBS containing calcium and magnesium (DPBS++), then resuspended at one million cells per ml in DPBS++ containing 0.5 mg/ml thermolysin. Cells were incubated 30 minutes at 37°C, then the reaction was stopped by washing with FACS buffer prior to staining.

In other experiments, cells were treated with the reducing agent DTT (Promega), which has been shown to cause partial release of GP1 from the cell surface and increase exposure of epitopes on the GP base (Francica et al., 2010). Parental Jurkat cells or Jurkat-GP cells were washed in FACS buffer, incubated for 30 minutes in FACS buffer containing 1 mM DTT at 37°C, then washed in FACS buffer prior to staining.

Jurkat cell display of EBOV-GP mutants

Error prone rolling circle amplification was used to mutagenize the pNL-GP using the TempliPhi kit and random hexamers (Fujii et al., 2006). An extra 0.5 mM of dNTPs was added to the reaction to enhance the amplification efficiency.(Rector et al., 2004) The amplified product was digested with BamHI to separate individual plasmid copies and then self-ligated. NEB Stable cells were transformed with the ligated plasmid. A plasmid library was generated from 80,000 clones and used to generate lentiviral particles with pMD2.G and psPax2 as described above. The lentivirus stock was to transduce Jurkat cells at a multiplicity of infection of 0.05. Five to ten days post-transduction, Jurkat cells expressing GP mutants which lacked binding to a specific mAb but retained binding to a panel of 2-3 control mAbs from separate competition groups were enriched by multi-color sorting with mAbs directly labeled with Alexa-405, −488, −647, or PE. Sorted cells were then single cell cloned using Clona-cell-TCS medium. Individual Jurkat-GP mutant clones were rescreened to confirm that they lacked binding to the mAb of interest and maintained binding to the panel of control mAbs. RNA was extracted from cell lines and amplified with primers flanking the GP insert to determine the specific amino acid changes present in GP. Cell lines expressing GP molecules with multiple amino acid changes were excluded from analysis.

For determining mAb binding to GP mutants, Jurkat cells expressing wild-type or mutant GPs were incubated with unlabeled mAbs at 5 ug/ml, followed by staining with anti-human IgG-PE and detection of fluorescence by flow cytometry as above. The binding of a control mAb (2.1.6F2 or KZ52) was used to control for variable GP expression across cell lines. An irrelevant human IgG1 was used as a negative control to assess background binding. The percentage binding of a test mAb to a mutant GP was defined as: 100% × [(binding of test mAb to mutant) ÷ (binding of test mAb to wild-type GP)] × [[(binding of control mAb to mutant) ÷ (binding of control mAb to wild-type GP)].

Visualization of GP

Pymol software version 1.8 was used to visualize mutations on the Ebola GP structure (PDB ID 5JQ3) (Zhao et al., 2016).

Author Contributions

C.W.D., K.J.L.J., A.J.G., P.H., I.C., P.J.G., G.C., Y.K., C.F.S., S.D.B., and R.A. designed and analyzed experiments. A.K. McElroy, A.K. Mehta, and R.A. designed the clinical study. A.K. Mehta ran the clinical study. C.W.D., K.J.L.J., A.K. McElroy, P.H., J.H., C.C., A.E.P., Y.L., C.G.A., and A.H.E. performed experiments. J.S., A.S., A.H.E., G.C., and E.O.S. provided novel reagents and methods and gave experimental suggestions. C.W.D., K.J.L.J., S.D.B., and R.A. wrote the paper.