Gene Transfer of ZMapp Antibodies Mediated by Recombinant Adeno-Associated Virus Protects Against Ebola Infections

Vector design and cloning

rAAV was constructed based on the prAAV-MCS promoter-less backbone with inverted terminal repeats (ITR) from rAAV2 and the human growth hormone polyadenylation site (Cell Biolabs, San Diego, CA). The CAG promoter was chosen to drive the expression of the antibodies. pCBLacZ ³⁷ was first digested by SalI and blunted, then digested by EcoRI to isolate a fragment containing the CMV enhancer, chicken β-actin promoter including exon 1 and intron 1, and splice acceptor from intron 2 of the rabbit β-globin gene. The fragment was then cloned in prAAV-MCS promoter less in the blunted-ClaI and EcoRI sites. Chimeric mouse/human antibody sequences (variable region sequences were provided by Dr. Larry Zeitlin from Mapp Biopharmaceutical, Inc., San Diego, CA), the foot-and-mouth disease virus self-cleaving peptide, and the optimized furin cleavage site ³⁸ were synthesized and codon-optimized for human expression at Genscript, Inc. (Morrisville, NC). The antibody sequences were constructed by fusing the variable regions of murine heavy and light chains (HC and LC) (protein data bank: c2G4 5KEL_H [HC], 5KEL_L [LC]; c4G7 5KEN_N [HC], 5KEN_O [LC]; c13C6 5KEL_C [HC], 5KEL_D [LC]) ³⁹ to their respective human constant regions of γ HC (CH1 domain) and κ LC (CL domain) of immunoglobulin G1 (IgG1). The codon for the lysine at the end of the HC constant region sequences was removed to prevent any interference at the protein level with the optimized furin cleavage site. ³⁸ The entire sequences were cloned in the EcoRI and BamHI site of the vector construct to give rise to prAAV-c2G4, -c4G7, and -c13C6 constructs. For negative control, the c2G4 antibody gene sequence was cloned using the same enzymes, except the fragment was blunted and cloned in the reverse orientation in the same site to give the prAAV-c2G4_INV construct. Helper sequences for E1A, E1B, E4, and E2A proteins and VA RNAs were provided separately using pXX6-80 (Dr. Jude Samulski's Lab at UNC Gene Therapy Center, Chapel Hill, NC). For serotype, the capsid proteins were expressed using another plasmid (pRep2Cap9), containing Rep from serotype 2 and capsid from serotype 9 (NCBI AY530579.1). ⁴⁰ pXX6-80 and pRep2Cap9 plasmid were used by the authors before. ⁴¹ The Rep2CapDJ plasmid was purchased at Cell Biolabs. The plasmid pVQAdCuROEBOZbGHpA contains the sequence of EBOV glycoprotein from the Zaire strain (UniProtKB/Swiss-Prot: P87666.1) in between KpnI and NotI sites. Expression was controlled by a CMV promoter/enhancer, and the gene contained a bovine growth hormone polyadenylation signal.

Cell culture and rAAV production

HEK293SF-3F6 cell line was maintained in serum-free SFM4Transfx-293 medium™ (HyClone, Logan, UT) supplemented with 6 mM of L-glutamine (HyClone), in vented-cap shake flasks (Corning, Corning, NY) at 37°C at 125 rpm agitation and under an atmosphere of 5% CO₂. ^42,43 For rAAV productions in disposable WAVE bioreactors (GE Healthcare, Mississauga, Canada), the cells were seeded at 2.5 × 10⁵ cells/mL 3 days before transfection. For the productions in shake flasks, the cells were diluted the day before transfection at 5 × 10⁵ cells/mL. Production of rAAV-c2G4_INV and -c13C6 in 1.3 L and 2 L format was performed in 2 L shake flasks with a cell culture volume of 660 mL each. Plasmids used for transfection were amplified in one shot TOP10™ (Invitrogen, Burlington, Canada) and purified using endotoxin-free kits (Qiagen, Toronto, Canada). The transfection was performed at ∼1 million cells/mL, as indicated in Table 1, by using 1 μg of total DNA and 2 μg of PEIpro^® (Polyplus-transfection) per mL of culture mixed in 10% of the total cell culture volume. Equal amount of each of the three plasmids were used for the transfection (ratio 1:1:1 for pRep2Cap9:pHelper:prAAV-c2G4/c4G7/c13C6). rAAV was harvested 3 days post transfection. rAAV-producing cells were centrifuged at 300 g then re-suspended in 1/10 of the starting volume with lysis buffer (2 mM of MgCl₂, 0.1% Triton X-100 in 25 mM of Tris-buffered solution, pH 7.5). In the 10 L rAAV-c2G4 production, the cells were lysed directly in the WAVE bioreactor by an addition of 1 L of 10 × lysis buffer. Benzonase (EMD Millipore, Etobicoke, Canada) was added to digest DNA and RNA at 5 IU/mL of starting cell culture volume, and incubated for 1.5 h at 37°C under agitation. Concentrated MgSO₄ was added to obtain 37.5 mM to reduce rAAV aggregation or to 75 mM to prevent further aggregation. Cellular debris were pelleted at 2,600 g for 30 min at room temperature, and the supernatant containing soluble rAAV was recovered. ^41,42,44 The rAAV lysate was then concentrated/diafiltered by tangential flow filtration to a volume corresponding to 10–12 mL of lysate/L of starting cell culture volume. Hollow fiber units (70–100 kDa MWCO; Spectrum Laboratories, Rancho Dominguez, CA) were used for tangential flow filtration (TFF). Before purification on an iodixanol-step purification, ⁴¹ the lysate was dead-end filtered using 0.8/0.45 μm filters, and 10 mL was loaded per Optiseal ultracentrifuge tube (Beckman and Coulter, Brea, CA). To reduce aggregation, a final concentration of 1.5 M of NaCl was added to the 3 L production of rAAV-c4G7 (Table 1) and in each layer of this iodixanol-step gradient (Optiprep™; Sigma–Aldrich, Oakville, Canada). ^45,46 rAAV in the 40% iodixanol layer was recovered in 5 mL per tube and pooled together. The iodixanol was removed by TFF and was replaced by 99.9% of the storage/injection buffer (phosphate-buffered saline [PBS] 2 mM of MgCl₂, 5% sucrose). The sample was further concentrated to the desired concentration by TFF. The number of rAAV vector genome (vg) copies was determined by real-time quantitative polymerase chain reaction (qRT-PCR), as described before, except a Mastercycler^® ep realplex (Eppendorf, Mississauga, Canada) instrument was used. ⁴¹ The primers and the probe were located in the CMV enhancer of the CAG promoter, having the following sequences: 5′-ACGTCAATGGGTGGAGTATTT-3′, 5′-AAGGTCATGTACTGGGCATAAT-3′, and 5′-TGCCCACTTGGCAGTACATCAAGT-3′, respectively.

Analysis of rAAV productions by SDS-PAGE and Western blot

The capsid proteins (serotypes 9 and DJ ^41,47 ) of the two 200 mL scale rAAV productions were separated by SDS-PAGE using a NuPAGE 4–12% polyacrylamide gradient gel (Thermo Fisher Scientific, Carlsbad, CA) with approximately 2.0 × 10¹⁰ vg/well then analyzed by Western blot using a specific monoclonal antibody (Clone B1) against the C-Terminal amino acids of VP1/2/3 (American Research Products, Waltham, MA) and a goat α-mouse IgG conjugated to HRP (GE Healthcare). The signal was captured and digitalized by using ImageQuant™ LAS 4000 mini biomolecular imager (GE Healthcare).

Production and characterization of the c2G4 ZMapp antibody produced in CHO cells after rAAV transduction

Twenty milliliters of Chinese Hamster Ovary (CHO) cell culture at a cell density of 1 × 10⁶ cells/mL was transduced using a multiplicity of infection (MOI) of 9,200 vg/cell of rAAV-c2G4 serotype DJ (produced as described above at 200 mL scale). One day after transduction, the cells (maintained in PowerCHO medium; Lonza) were grown for 7 days at 32°C instead of 37°C, as done previously, under an atmosphere of 5% CO₂ with 0.5 μM etoposide (Sigma–Aldrich), a topoisomerase inhibitor that can facilitate synthesis of the complementary strand for rAAV. ⁴⁸ For c2G4 purification, a protein A MabSelect SuRe column (GE Healthcare) was used according to the manufacturer's instructions. The protein concentration was determined using a Nanodrop (optical density 280 nm) instrument. To separate the HC and LC of c2G4, SDS-PAGE was performed using a Bis-Tris NuPAGE 4–12% polyacrylamide gradient gel and a BLUelf Prestained Protein Ladder (Froggabio, Toronto, Canada). The gel was stained with a 0.5% Coomassie Blue G-250, 50% methanol/10% acetic acid staining solution, and then the bands corresponding to each chain were excised and sent to the Proteomic platform of CHU of Quebec Research Center (Quebec, Canada) for liquid chromatography/tandem mass spectrometry (LC-MS/MS). The peptide sequences were blasted against the human and mouse genomes, and against the foot-and-mouth disease virus genome. The function of the purified c2G4 antibody was assessed below by confocal imagery.

Production of ZMapp antibodies in HEK293SF-3F6 cells after rAAV transduction

rAAV-c2G4/c4G7/c13C6 cells were produced using the DJ serotype in HEK293SF-3F6 in six-well plates using the standard transfection conditions described above. Three days later, the cells were pelleted and re-suspended in 100 μL of cell medium and freeze/thawed three times. A fraction (15 μL) was then used to transduce 5 × 10⁵ HEK293SF-3F6 cells in 350 μL in a 24-well plate under agitation (125 rpm). The cells were co-transduced at a MOI of 2 with a first-generation adenovirus expressing luciferase titered by a standard plaque assay. ⁴⁹ After 48 h, the cells were pelleted by centrifugation at 500 g for 1 min, and the supernatant containing the ZMapp antibodies was recovered. The function of ZMapp antibodies was then assessed by flow cytometry analysis.

Confocal imagery and flow cytometry for assessing the function of ZMapp antibodies

The protocol used for cell staining was identical for confocal imagery and flow cytometry. HEK293SF-3F6 cells were seeded at 1 × 10⁶ cells/mL in a six-well plate in 2 mL of medium. They were then transfected with a plasmid expressing the Zaire Ebola glycoprotein (EboGP, pVQAdCuROEBOZbGHpA). Three days after transfection, one third of the cells were centrifuged at 400 g. The cells were re-suspended in PBS 3% bovine serum albumin (BSA) on ice for blocking. Either 2 μg of purified c2G4 produced in CHO cells or 70 μL of unpurified ZMapp antibody supernatants (one fifth of 350 μL) produced in HEK293SF-3F6 cells was incubated with the cells on ice for 1.5 h. A donkey anti-human secondary antibody conjugated with Alexa Fluor^® 488 nm AffiniPure F(ab′)₂ Fcγ specific (Jackson Laboratories, West Grove, PA) was then incubated 1 h on ice. The cells were then fixed with 2% ρ-formaldehyde. Confocal images were taken, as done previously, with purified c2G4. ⁵⁰ Labelled cells with unpurified 2G4/c4G7/c13C6 ZMapp antibodies were analyzed using a BD LSRII flow cytometer (BD Biosciences, Mississauga, Canada).

rAAV treatment and EBOV trial

All mouse experiments were performed using rAAV9. Juvenile BALB/c mice, 4–6 weeks of age, were treated using three different routes—i.v., i.m., and i.n.—as described before. ^32,34 Fifty microliters of rAAV was administered at different doses (2.7 × 10¹⁰, 7.5 × 10¹⁰, 13.0 × 10¹⁰, and 30.0 × 10¹⁰ vg). The cocktail consists of a mixture of three rAAVs, each representing 33% of the total amount of AAV cocktail. After 21 days, a dose equal to 1,000-fold the lethal dose 50 (1,000 LD₅₀) of mouse-adapted EBOV (Zaire strain, Mayinga) was injected i.p. The weight of each mouse was measured daily. Determination of individual clinical scores was based on the severity of EBOV signs of infection. The ranks allowed for one observation were: 0 = no signs; 1 = ruffled fur, slower activity, and loss of body condition; 2 = labored breathing or hunched posture; 3 = ≥25% weight loss or no movement/unresponsive/paralysis; 4 = found dead. A score of 3 resulted in immediate euthanasia. Cumulative clinical scores were calculated daily by the addition of individual scores within each group. The mouse experiments were approved by the Animal Care Committee at the Canadian Science Center for Human and Animal Health in accordance with the guidelines outlined by the Canadian Council on Animal Care. Experiments were conducted according to regulatory standards at National Microbiology Laboratory in BSL-4.

Enzyme-linked immunosorbent assays for detection of ZMapp and anti-ZMapp antibodies

For sandwich enzyme-linked immunosorbent assay (ELISA) detection of ZMapp antibodies, a dilution of 2 μg/mL in PBS of a whole goat anti-human IgG (H + L; Jackson Laboratories) was used to coat COSTAR flat-bottom high-binding 96-well plates (Corning). The purified standard (c2G4 antibody) was produced by a CHO stable clone generated at the National Research Council, Canada. The standard or serum samples were then deposited on the plates and incubated 2 h. The serum samples were diluted 1:100 in PBS 1% BSA. A secondary goat anti-human IgG (H + L) antibody conjugated to alkaline phosphatase (Jackson Laboratories) was used in combination with ρ-nitrophenylphosphate (Sigma–Aldrich), with the reading performed at 405 nm using a SpectraMax 340PC384 reader (Molecular Devices, Sunnyvale, CA). For detection of anti-ZMapp antibodies, the plates were coated with 2 μg/mL of c2G4. Serum samples dilutions of 1:100 were deposited on the plates, and a secondary goat anti-mouse IgG (Fc fragment) conjugated to alkaline phosphatase (Sigma–Aldrich) was used for detection.

Statistics

For statistical analyzes, one- or two-tailed Student's t-tests were used when appropriate for comparisons to verify the significance of the differences between each group. Means are presented with the standard error of the mean (SEM) as error bars or ± values. Log-rank (Mantel–Cox) tests were performed to assess the significance of survival differences between rAAV-treated groups.

AAV design and characterization of ZMapp antibodies

Design of rAAV for the expression of ZMapp antibodies (rAAV-c2G4, -c4G7, and -c13C6) combines the expression of both the HC and LC under the same expression cassette controlled by the strong ubiquitous promoter CAG (Fig. 1A). The sequence coding for the foot-and-mouth disease self-cleaving 2A peptide (F2A) was placed in between the HC and LC genes, allowing ribosome skipping and the translation of both antibody subunits from a single transcript. To obtain a blunt cleavage between HC and LC, which was associated with higher antibody stability in vivo, an optimized furin-specific cleavage site was inserted just before the F2A sequence in the rAAV genome (Fig. 1A). ^38,51 A previously established triple plasmid transfection method was used to produce rAAVs. HEK293SF-3F6 cells were transfected in suspension in serum-free conditions. Three days later, the rAAVs produced were recovered by chemical cell lysis and purified by a iodixanol-step gradient. ^41,44,52 Two different rAAV preparations (serotypes 9 and DJ) were analyzed by Western blot to verify the presence of the capsid proteins. Three bands of a proper molecular weight were observed, thus confirming the presence of VP1, VP2, and VP3 capsid proteins at 87, 72, and 62 kDa, respectively (Fig. 1B). The 62 kDa band was most intense because it is more abundant in the virion (VP1:VP2:VP3 = 5:5:50).

OPEN IN VIEWER

Figure 1.

Characterization of adeno-associated virus serotype 9 (AAV9) and ZMapp neutralizing antibodies expressed by AAV. (A) Design of recombinant AAVs (rAAVs) used to express ZMapp antibodies. L1 and L2 are secretion leader peptide sequences from the immunoglobulin G (IgG) alpha human heavy chain (HC) and the IgG kappa human light chain (LC), respectively. RKRR is an optimized furin cleavage sequence. The last lysine from HC was deleted to prevent interference with the cleavage. F2A, self-cleaving peptide from the foot-and-mouth disease virus; pA, polyadenylation signal; ITR, inverted terminal repeats from AAV2. (B) Two rAAV batches (200 mL scale; serotype 9 and DJ) were analyzed by Western blot specific to the capsid proteins VP1, VP2, and VP3. M, BluELF molecular marker. (C) Coomassie blue staining of gel after subunit separation of purified c2G4 antibody by SDS-PAGE in reducing conditions. Different quantities of purified antibody were loaded on the gel. The amino acid sequences of HC and LC were confirmed by liquid chromatography/tandem mass spectrometry. The F2A peptide was not detected suggesting a proper cleavage. (D) Untransfected HEK293SF-3F6 cells (i) and (ii) and transfected cells expressing the glycoprotein from Zaire strain of Ebola virus (EBOgp) (iii) and (iv), and a magnified image, (v), were incubated with the purified c2G4 antibody. A secondary antibody directed against human IgG and conjugated to Alexa Fluor 488 nm (AF488) was used for detection. Nuclei were stained with Hoechst 33342 (H33342), and the signal was superimposed on AF488 signal. Confocal microscopy pictures were taken with a 60 × oil-immersion objective. The right panel shows a cartoon explaining the immunodetection. (E) Quantification of EBOgp-binding and specificity of ZMapp antibodies by flow cytometry. Antibodies produced by AAV/adenovirus co-infected HEK293SF-3F6 cells were incubated on transfected cells. AF488 antibody was used to label EBOgp-expressing positive cells. AAV-c2G4 INV was used as negative control (NegCTL) because it contains the sequence of c2G4 cloned in the reverse orientation to abolish protein expression.

To produce high antibody titers for characterization, CHO cells were transduced with a rAAV-DJ encoding the c2G4 antibody sequence. CHO cells are frequently used in the industry because they can produce high quantities of active antibodies during several days in culture. The serotype DJ was chosen because of its better transduction efficacy than serotype 9 for CHO cells. ⁴⁷ Secreted c2G4 antibodies from the supernatant of CHO cells were purified by a protein A column and then analyzed by SDS-PAGE under reducing conditions. After Coomassie blue staining, two bands corresponding to HC and LC were observed at the predicted molecular weights of 52 and 26 kDa, respectively (Fig. 1C). In addition, the amino acid sequences of the HC and LC were confirmed by LC-MS/MS. The presence of F2A-derived amino acids was not detected, suggesting a proper self-cleavage of LC and HC.

The function of all three ZMapp antibodies was assessed by fluorescence microscopy and flow cytometry. HEK293 cells expressing EBOV glycoprotein (EBOgp) after transient transfection of the gene were incubated with c2G4, c4G7, and c13C6 antibodies from the supernatant of AAV-transduced HEK293SF-3F6 cells. A secondary antibody recognizing the human Fc of the chimeric antibodies and conjugated to Alexa Fluor 488 nm was used to stain the cells. Positive staining was characterized by a fluorescent ring at the cell surface because the EBOgp is a transmembrane protein, as shown for the c2G4 antibody (Fig. 1D). By flow cytometry, positive staining was observed in approximately 50% of EBOgp-transfected cells but not in untransfected cells, indicating that the ZMapp antibodies produced by rAAV were specific and functional. rAAV-c2G4_INV was used as a negative control because it is unable to express c2G4 (Fig. 1E).

Scale-up of rAAV production and purification for mouse experiments

To obtain enough quantity of rAAV for in vivo experiments, the triple transfection method was scaled up to 10 L in shake flasks and WAVE bioreactors. Quantitative real-time PCR, a standard method for rAAV titration, was performed on end-purified material to determine the final rAAV titer. For this, a primer/probe set binding in the CMV enhancer of the CAG promoter was used, giving a % CV <24% (Table 1). Up to 1.3 × 10¹⁴ vg of purified rAAV material was obtained from 10 L of cell culture (Table 1). Overall, an average of 4.7 × 10¹² vg was obtained from 1 L of cell culture, with rAAV cell-specific production ranging from 6.36 × 10² to 1.34 × 10⁴ vg/cell.

c2G4 gene transfer protects against Ebola infections in vivo

The objectives of the following experiment were first to evaluate the efficacy of the i.v. route in addition to the i.m. and i.n. routes examined in vivo elsewhere ³⁴ for the vector delivery of ZMapp antibodies, and to determine a therapeutic dose. For simplicity, only the rAAV encoding c2G4 was administered because it has been previously demonstrated that the murine version (2G4) was able (after its direct injection) to confer protection against an EBOV challenge in mice. ⁵¹ Thus, rAAV-c2G4 was administered at doses of 2.7 × 10¹⁰ and 13.0 × 10¹⁰ vg via the i.v. route, and the i.m. and i.n. routes were also re-examined following the protocol depicted in Fig. 2A. The maximal dose (13.0 × 10¹⁰ vg) was selected because a dose in the range of 10.0 × 10¹⁰ vg was safely and successfully tested in mice in previous publications using the i.n., i.v., and i.m. routes of injection. ^{33,34,53 –55} The lower dose (2.7 × 10¹⁰ vg) used is also comparable to the one used (3.0 × 10¹⁰ vg) in a previous study. ²⁸ A period of 20 days was allowed for expression of antibody c2G4, and its serum concentration was measured. Expression of antibodies at 20 days post rAAV administration was significantly higher than the expression at 10 days. This was also the case with c4G7 and c13C6 ZMapp antibodies (Supplementary Fig. S1 and statistics in Supplementary Table S1; Supplementary Data are available online at www.liebertpub.com/hum). Altogether, this indicated that the chosen time for the challenge was appropriate. On day 0, the EBOV challenge was performed, and the mice were kept under observation for the following 21 days. The results indicated that the i.v. route was the most effective, with a maximal protective efficacy of 80%, while the i.m. route allowed up to 50% protection. Surprisingly, the i.n. route failed to confer any protection (Fig. 2B). Interestingly, mouse sera collected 1 day before the EBOV challenge demonstrated higher concentrations of the c2G4 antibody in the three groups, showing the highest survivability (i.v.Hi: 38 μg/mL; i.v.Lo: 5.3 μg/mL; i.m.Hi: 26 μg/mL; statistics in Supplementary Table S1). Furthermore, the number of survivors and concentration of c2G4 antibody correlated positively (Spearman's correlation, r = 0.81, p < 0.01), indicating that the more the antibody was produced, the higher was the survival probability. Mice treated i.n. had no detectable serum concentration of c2G4, thus explaining the absence of a therapeutic effect in that group. The results indicated that the i.v. and i.m. routes were effective at delivering rAAV, as shown by the higher antibody concentration and survivability.

OPEN IN VIEWER

Figure 2.

Gene transfer of the c2G4 antibody mediated by AAV protects against Ebola infections in a mouse model. (A) rAAV-c2G4 was injected at the high dose of 13.0 × 10¹⁰ vg (Hi) and low dose of 2.7 × 10¹⁰ vg (Lo) at day −21. Intravenous (i.v.), intramuscular (i.m.), and intranasal (i.n.) routes were used for administration of rAAV-c2G4. The negative control group (NegCTL) received a high dose i.v. of the non-functional rAAV-c2G4_INV. Bleeding was performed 20 days after rAAV-c2G4 administration (day −1) to analyze mouse serum (samples were diluted 100-fold for enzyme-linked immunosorbent assay [ELISA]). (B) Concentration of c2G4 in the serum of mice by ELISA. The challenge was performed on day 0 with a 1,000 LD₅₀ dose of Zaire Ebola strain (Mayinga) by intraperitoneal injection. The experiment ended on day +21 after the survivors displayed no sign of Ebola infection. (C) Normalized weights in percentage (based on day −1) for each group. (D) Cumulative clinical scores. A cumulative clinical score (CCS) is the sum of individual mouse scores within a group. (E) Kaplan–Meier survival curves of different experimental groups. N = 10 mice per group. Note: Unexpected and unique survivor in the 2G4_INV negative control group. n.d., not detected: values were under the background level.

Signs of EVD were significantly attenuated in comparison to negative control (2G4_INV) after the treatment for the following groups: i.v.Hi, i.m.Hi, and i.v.Lo. The i.v. administration of rAAV-c2G4 at a high dose fully prevented weight loss, while in other groups the weight loss was significantly reduced (day 6, Fig. 2C). Group cumulative clinical scores consisting in the sum of individual scores within each group were improved in comparison to the negative control between 3 and 9 days post challenge, the period in which the mouse's physical condition deteriorated (p < 0.05; Fig. 2D). A log-rank test on Kaplan–Meier plotted values supported an increased survivability for the i.v.Hi group (p < 0.05; Fig. 2E and statistics in Supplementary Table S2). For the first time, it was shown that a treatment with a single rAAV (rAAV-c2G4) could effectively suppress signs of EVD in mice.

Vectored ZMapp antibody delivery with a rAAV-ZMapp cocktail in vivo

The rAAV-ZMapp cocktail containing three rAAVs expressing the ZMapp antibodies (c2G4, c4G7, and c13C6) was evaluated in mice to determine its protective efficacy. rAAVs from the cocktail were also administered separately to determine the protective efficacy they provide individually and for comparison with the cocktail (Fig. 3A). As mentioned before, the cocktail consists of a mixture of three rAAVs, each representing 33% of total amount of AAV cocktail. For this experiment, only the i.m. and i.v. routes were employed because the i.n. route failed to protect the mice. The dose administered was also increased to 30.0 × 10¹⁰ vg to obtain a higher protective efficacy. Furthermore, the same maximal dose was used in a previous study with rAAV-ZMapp cocktail. ³⁴ A lower dose (7.5 × 10¹⁰ vg) in comparison to 13.0 10¹⁰ vg (Fig. 2) was also administrated to study the treatment response following the dose used. Robust protective efficacies of 90% and 60% were obtained with the cocktail after i.v. and i.m. administration, respectively, at the highest dose administered (Fig. 3B). Reducing the dose resulted in less protection, thus indicating that the dose is an important factor for protective efficacy (Fig. 3B). After i.v. administration, the cocktail at the highest dose was more effective than single rAAV-c4G7 (70%) and rAAV-c13C6 (30%), but not rAAV-c2G4. Administration of rAAV-c2G4 by the i.v. and i.m. routes achieved complete protection (100% efficacy) and 90% protective efficacy, respectively (Fig. 3B).

OPEN IN VIEWER

Figure 3.

ZMapp antibody gene transfer mediated by AAV protects against Ebola infections in a mouse model. (A) rAAV-c2G4/c4G7/c13C6 was injected separately or in a cocktail at high (30.0 × 10¹⁰ vg; Hi) and low (7.5 × 10^{10 VG}; Lo) doses at day −21. Intravenous (i.v.) and intramuscular (i.m.) routes were used for administration. The negative control group (NegCTL) received a high i.v. dose of the non-functional AAV-c2G4_INV. Bleeding was performed 20 days later (day −1) to analyze mouse serum (serum samples were diluted by 100-fold for ELISA quantification). (B) Concentration of ZMapp antibodies in the serum of mice by ELISA. The mice were then challenged on day 0 with a 1,000 LD₅₀ dose of Zaire Ebola strain (Mayinga) by intraperitoneal injection. The experiment ended on day +21 after the survivors displayed no sign of Ebola infection. (C) Normalized weights in percentage (based on day −1) for each group. (D) Cumulative clinical scores. A cumulative clinical score (CCS) is the sum of individual mouse scores within a group. (E) Kaplan–Meier survival curves of different experimental groups. N = 10 mice per group.

Protective efficacies obtained after rAAV treatments correlated positively (Spearman's correlation, r = 0.88, p < 0.01) with the serum concentrations of ZMapp antibodies in samples collected 1 day before the EBOV challenge. The minimal effective concentration for each ZMapp antibody was thus determined based on a protective efficacy of 90%. Minimum effective concentrations were estimated to 6 μg/mL for the rAAV-ZMapp cocktail, 1.8 μg/mL for the c2G4 antibody, and 33.3 μg/mL for the c13C6 antibody. The minimal concentration could not be determined for c4G7 because the number of survivors was too low (n = 3) and because the variability in the data set was too high. Effective rAAV doses determined above and concentration of ZMapp antibodies could be used to establish initial conditions for testing in larger animal models.

The rAAV treatment was again well tolerated, despite the 2.2-fold higher dose administered. In challenged mice, weight loss was fully prevented, irrespective of the dose and route of administration, by using rAAV-ZMapp cocktail and rAAV-c2G4, except for the low dose i.m. with the cocktail (an average weight loss of 15% was observed at day 6; Fig. 3C). In addition, the treatments with rAAV-c4G7 and rAAV-c13C6 partially prevented weight loss in comparison to the negative control (Fig. 3C). The clinical score was steadily maintained at zero for the mice treated i.v. at the highest dose with rAAV-c2G4, indicating that the treatment completely prevented the appearance of EVD clinical signs (Fig. 3D). The clinical scores between 3 and 8 days post challenge, a period accounting for the highest scores, was significantly lower than the negative control group in all groups except for c4G7 and cocktail injected i.m. at 2.7 × 10¹⁰ vg. Survivability was also increased in all treated mice, except for the cocktail at low i.m. dose (Fig. 3E). In summary, the results show that the rAAV treatments suppress signs of EVD. Statistics for Fig. 3B and E can be found in Supplementary Tables S1 and S2.

Induction of an antibody-specific humoral response

As observed above, i.n. administration of rAAV-c2G4 conferred no protection because the quantity of c2G4 antibody in the mouse's system was insufficient (Fig. 2B). One possible explanation for this is provided by the analysis of the immune response generated against the chimeric antibody. Chimeric antibodies were used for these experiments because (1) they can easily be monitored by ELISA in serum, and (2) they are more likely to be used as an active drug in humans in comparison to more immunogenic murine antibodies. Despite these advantages, neoepitopes harbored by non-native therapeutic proteins have been held responsible for triggering a deleterious immune responses that can reduce the efficacy of treatment. ^{15,34,56 –59} To investigate this, the presence of a humoral response against the chimeric antibodies was monitored in mouse sera 1 day before the challenge (i.e., 20 days after rAAV administration; Fig. 4A). Interestingly, at a dose of 13.0 × 10¹⁰ vg, the humoral response was stronger in the i.n. group in comparison to the i.v. and i.m. groups (p < 0.001), but the difference was no more significant at a lower dose (2.7 × 10¹⁰ vg). The presence of anti-c2G4 antibodies in the i.n. group demonstrated that c2G4 was expressed at some point in these animals. This was also confirmed by ELISA 10 days after rAAV administration, although the c2G4 antibody was detected at a low concentration (∼200 ng/mL; Supplementary Fig. S1).

OPEN IN VIEWER

Figure 4.

Induction of a humoral response reduces the efficacy of AAV-ZMapp treatment. (A) and (B) Serum samples (diluted by 100-fold) from mice previously treated with ZMapp rAAVs were analyzed by ELISA (day −1) to determine the anti-ZMapp antibody titer. The number of survivors per group is indicated on top. The negative control group (NegCTL) received a i.v. high dose of the non-functional AAV-c2G4_INV. (C) Anti-ZMapp antibody titers and ZMapp antibody concentrations were plotted. Data from (A) and (B) were pulled for the analysis, except for i.n. and 2G4_INV groups that were excluded. Spearman's ρ = −0.75 and p < 0.0001 (n = 120), indicating a strong negative correlation between the concentrations of ZMapp and anti-ZMapp antibodies. The graph was divided into quadrants to show the serum levels of ZMapp and anti-Zmapp of survivors and mice that have succumbed. n.d., not detected, values were under the background level.

The humoral response was then evaluated in mice injected with the rAAV-ZMapp cocktail and single rAAVs, although it was not significantly different between the groups, except for the negative control 2G4_INV (i.e., not detected) and the mice treated i.m. with the low dose of rAAV-ZMapp cocktail (Fig. 4B). Statistics for Fig. 4 can be found in Supplementary Table S3.

ZMapp antibodies and anti-ZMapp antibody concentrations from both in vivo experiments (Figs. 2 and 3) excluding those from the i.n. and 2G4_INV groups were then plotted to determine whether there was a correlation (Fig. 4C). A very strong negative correlation (Spearman's ρ = −0.75, p < 0.00001, n = 120) was found between the concentration of ZMapp antibodies and anti-ZMapp antibodies, indicating an inverse proportional relationship. For example, a low concentration of anti-ZMapp antibodies generally corresponds to a high concentration of ZMapp antibodies and thus to better survival, as shown above (Figs. 2B and 3B). Furthermore, the higher the expression of ZMapp antibodies was, the higher was the rate of survival. All of the survivors had >1 μg/mL of ZMapp antibodies circulating in their serum. An anti-ZMapp antibody titer <0.24 O.D. was another characteristic of survivors. Together, the data agree with previous studies demonstrating that neoepitopes harbored by non-native therapeutic proteins may trigger a deleterious immune response that can reduce the therapeutic efficacy. ^{15,34,56 –59}

In the experimental set-up, all groups contained the same number of males and females. Thus, any gender effect was mitigated by averaging the data for comparison. Interestingly, it was found that, irrespective of the groups analyzed, the concentration of antibodies was always higher in males (fivefold, p = 0.0019, except in i.n. groups where no expression was detected) and anti-antibody titers were always higher in females (twofold, p = 0.00061). Therefore, it is possible that females have developed a stronger humoral response that could have prevented the treatment of been effective by suppressing the expression of therapeutic antibodies.