Rational Zika vaccine design via the modulation of antigen membrane anchors in chimpanzee adenoviral vectors

Zika virus (ZIKV) emerged on a global scale and no licensed vaccine ensures long-lasting anti-ZIKV immunity. Here we report the design and comparative evaluation of four replication-deficient chimpanzee adenoviral (ChAdOx1) ZIKV vaccine candidates comprising the addition or deletion of precursor membrane (prM) and envelope, with or without its transmembrane domain (TM). A single, non-adjuvanted vaccination of ChAdOx1 ZIKV vaccines elicits suitable levels of protective responses in mice challenged with ZIKV. ChAdOx1 prME ∆TM encoding prM and envelope without TM provides 100% protection, as well as long-lasting anti-envelope immune responses and no evidence of in vitro antibody-dependent enhancement to dengue virus. Deletion of prM and addition of TM reduces protective efficacy and yields lower anti-envelope responses. Our finding that immunity against ZIKV can be enhanced by modulating antigen membrane anchoring highlights important parameters in the design of viral vectored ZIKV vaccines to support further clinical assessments.

Numerous Zika vaccines have recently been reported demonstrating protective efficacy in mice and/or non-human primates. Most of these vaccines elicited neutralizing antibody responses after one or more immunization(s) without any adverse events. As Zika virus and dengue virus often co-circulate, it is critical to test preclinical Zika (and dengue) vaccines for antibody-dependent enhancement (ADE) that can induce dengue hemorrhagic fever, a serious disease with poor outcome and potentially worsen the disease course or manifestations of Zika. Lopez-Camacho et al designed 5 DNA and 6 chimpanzee adenoviral (ChAdOx1) vectored Zika vaccines using various portions of the Zika virus genome. Of those, several ChAdOx1 vaccines quickly induced longlasting T and B cell immune responses and one of them provided sterilizing immunity against Zika virus after a single non adjuvanted immunization. Most importantly, (1) ChAdOx1 vaccines elicit antibodies that can bind to dengue virus envelope raising the possibility that the present vaccine can protect from both Zika and dengue virus infections, and (2) none of the presented vaccines induced in vitro ADE to DENV2 that highly increases the safety profile of those vaccines.
Main criticisms: There are major problems with the experimental design that need to be addressed: • The authors do not show NAb responses in Balb/c mice or E-specific IgG in CD1 mice, which creates a problem for the interpretation of the data given that Ext Data Figure 3C shows marked differences in the antibody responses in these two mouse strains. Both analyses should be performed in both mouse strains.
• The authors offer a new and interesting take on the neutralization assay, which may help in making this analysis more high-throughout. However, they do not compare it to any of the more standard assays conducted in this field, such as a traditional plaque reduction neutralization test, which is necessary to compare this vaccine with any of the other promising Zika vaccine candidates. It is known that very large differences in titer is observed between different neutralization assays. A comparison of their new assay to standard and published assays would greatly improve the paper. • Many of the figures and assays are not described in enough detail for the reader to understand what was done, and the discussion could be much improved, as outlined below.
Other important concerns: • Figure 1A: Something seems wrong with this tree -there should be much greater distance between "Other Flaviviruses" and Zika than between different Zika strains. We also can't read the text. • Figure 1B: There are no serum dilutions stated for these ELISAs, so this data cannot be interpreted.
• Figure 3B: A linear, three-segment y-axis on these graphs is very hard to read. A log scale would be much clearer.
• Ext Data Figure 3C: No dilutions given for ELISA. • Figure 3E: Please provide a clearer definition for "viremia peak ablation" in either the figure legend or Methods. • Figure 4A: If I am reading this correctly, you have averaged together the neutralization curves for all mice in each group. Better would be to show the individual curves for each mouse (with error bars if possible), and/or to plot the MN50 value for each mouse, so readers can see the full variation and complexity of the data. This is needed given your statement on line 143 that neutralization was only seen in a few mice in some groups. • Figure 4B: What is the capture antigen (and its source) in these ELISAs-DENV E protein, whole DENV virus, or a synthetic peptide of residues 147-166? • Intro/Discussion: The discussion of this report in the context of the Zika vaccine field could be much improved. Please reference the other Zika vaccine papers, including Muthumani et al, npj Vaccines 2016;Kim et al, EBioMedicine 2016;Pardi et al, Nature 2017;and especially Richner et al, Cell 2017, which directly addresses the issue of ADE, a major topic of your manuscript. The latter two papers additionally discuss mRNA-based vaccines, a platform which is not mentioned in your intro or discussion. Please discuss how your vaccine compares in neutralizing titer and other characteristics (e.g. species specificity) with these other vaccines, in particular RhAd52. • Ext Figure 3A: The x-axis here is not labeled clearly with respect to the mouse strains (the middle groups fall under neither label). And are prM and E the antigens used for stimulation? Please also provide the source of the peptide pools used. • Lines 108-109 and throughout: "Sterile" protection might not be the most accurate term when serum viremia is undetectable, in the absence of testing other highly infected tissues (such as spleen) or looking at anamnestic antibody responses in response to viral challenge. Better to be more specific, e.g. "protected from detectable viremia" • Line 129-130: I am not sure what you mean by "complete, intermediate, and low sterile protection." Which adjectives correspond to which of the six vaccine conditions listed? • Line 336: Do you mean "neutralized 50% of infection" rather than ">50%"? • Line 138: "signal-to-noise ratio of >10" -It is not clear how this was calculated. This may be better expressed in terms of a standard deviation or as error bars or a plot of all data points in Figure 4A. • In the challenge experiment, no information about the injected vaccine dose can be found in the text. This should be added to Fig. 3 legends.

Reviewers' comments:
Reviewer #1 (Zika)(Remarks to the Author): Lopez-Camacho and colleagues report the efficacy results of chimpanzee adenovirus vector-based Zika vaccine. They showed prevention of Zika virus infection in mice that were pre-immunized with the vaccine. Although the topic is interesting and important, this reviewer is concerned about several important gaps that are missing in the current manuscript. 1. The authors should provide traditional PRNT values for the immunized mice from the immunized BALB/C mice. This is critical to evaluate the correlation between efficacy and functional neutralization titers.
We aimed to establish a high-throughput methodology to allow a rapid assessment of ZIKV vaccine efficacy in vitro; especially when assessing many vaccine candidates using dilutions and various time points with 3 replicates. Although we acknowledge PRNT is a traditional assay, we have seen that this assay displays lab-to-lab variability, especially when using different ZIKV strains. We hope the reviewer appreciates the vaccine efficacy achieved with our candidates, using the same ZIKV challenge model as Larocca et al., which is now being used in several papers. Moreover, we have reached similar endpoint titers as the ones reached by Larocca et al., which mirrored protection (viral loads) observed in our study.
Although PR is a traditional method, Larocca et al. have reported a neutralisation assay in Nature (Nature 536(7617):474-8. 2016), using a Polymerase Chain Reaction-based technique. Our method is highly similar to the methodology reported by Larocca et al. (MN50) with a difference consisting of assessing the Envelope ZIKV protein expression, rather than assessing mRNA levels. The technique was used for BALB/c sera and it is reported in our manuscript (Line 132-150 and Figure 4).
2. The authors should provide a cohesive data set from the same experimental system. In the current manuscript, the authors performed efficacy experiments in BALB/c, but provided the non-traditional neutralization titers of sera collected from immunized CD1 mice. These are key experiment results and should not be disjointed by piecing the results from two types of mice.
We agree with this comment and we now provide neutralisation (NT) results both, in BALB/c and CD-1 (Lines 151-161, highlighted). Importantly we have carried out the assay shortly and long after vaccination using two time points of 4 and 16 weeks to assess memory responses. CD1 NT is provided as an extended legend and we believe that the NT assay from outbred mice such as CD1, which is not limited to specific HLA genes, is biologically meaningful.
3. The "new" neutralization assay is not clearly described.
We describe the methodology and results in lines 132-161 and include a detailed description in Material and Methods in line 365-376 (highlighted).
4. The authors are over "marketing" the study. This is a mouse efficacy study and the presented results do not conclude "Phase I clinical trial is scheduled to start in 2017." The next logical step is to test its efficacy in NHP model.
We appreciate this concern and we have decided to exclude this statement. Pointing out that our platform is suitable for further pre-clinical and clinical testing. Importantly, our development has received funding from the British Government to be taken to clinical trials in 2018 and the vaccine is currently being produced in GMP. Although assessment in macaques would be desirable, this has not been a requisite to obtain funding for a phase I trial in Oxford but we agree it is a logical step to take and we are considering this and applying for funds to test the most efficacious vaccine reported in this paper.
5. It is interesting that no ADE was observed in the U937 cell-based in vitro assay. However, the author should temper the relevance of the cell culture results.
We appreciate this feedback. We rephrased the conclusions and tempered the description of our results (Line 173-177).
Reviewer #2 (Virus, RNA vaccine)(Remarks to the Author): Numerous Zika vaccines have recently been reported demonstrating protective efficacy in mice and/or non-human primates. Most of these vaccines elicited neutralizing antibody responses after one or more immunization(s) without any adverse events. As Zika virus and dengue virus often cocirculate, it is critical to test preclinical Zika (and dengue) vaccines for antibody-dependent enhancement (ADE) that can induce dengue hemorrhagic fever, a serious disease with poor outcome and potentially worsen the disease course or manifestations of Zika. Lopez-Camacho et al designed 5 DNA and 6 chimpanzee adenoviral (ChAdOx1) vectored Zika vaccines using various portions of the Zika virus genome. Of those, several ChAdOx1 vaccines quickly induced long-lasting T and B cell immune responses and one of them provided sterilizing immunity against Zika virus after a single non adjuvanted immunization. Most importantly, (1) ChAdOx1 vaccines elicit antibodies that can bind to dengue virus envelope raising the possibility that the present vaccine can protect from both Zika and dengue virus infections, and (2) none of the presented vaccines induced in vitro ADE to DENV2 that highly increases the safety profile of those vaccines.

Main criticisms:
There are major problems with the experimental design that need to be addressed: • The authors do not show NAb responses in Balb/c mice or E-specific IgG in CD1 mice, which creates a problem for the interpretation of the data given that Ext Data Figure 3C shows marked differences in the antibody responses in these two mouse strains. Both analyses should be performed in both mouse strains.
We agree with this comment, neutralisation in BALB/c is now provided and importantly, we have extended the analysis to include results in BALB/c on weeks 4 and 16 post immunisation (Lines 132-161). CD1 neutralisation is also provided in the results (Lines 157-161) and as an extended legend (Extended Fig 3B). We believe that the results in outbred mice such as CD1 is biologically meaningful as it is not limited to a few HLA genes, • The authors offer a new and interesting take on the neutralization assay, which may help in making this analysis more high-throughout. However, they do not compare it to any of the more standard assays conducted in this field, such as a traditional plaque reduction neutralization test, which is necessary to compare this vaccine with any of the other promising Zika vaccine candidates. It is known that very large differences in titer is observed between different neutralization assays. A comparison of their new assay to standard and published assays would greatly improve the paper.
We aimed to establish a high-throughput methodology to allow a rapid assessment of ZIKV vaccine efficacy in vitro; especially when assessing many vaccine candidates using dilutions and various time points with 3 replicates. Although we acknowledge PRNT is a traditional assay, we have seen that this assay displays lab-to-lab variability, especially when using different ZIKV strains. We hope the reviewer appreciates the vaccine efficacy achieved with our candidates, using the same ZIKV challenge model as Larocca et al., which is now being used in several papers. Moreover, we have reached similar endpoint titers as the ones reached by Larocca et al., which mirrored protection (viral loads) observed in our study.
Although PR is a traditional method, Larocca et al. have reported a neutralisation assay in Nature (Nature 536 (7617):474-8. 2016), using a Polymerase Chain Reaction-based technique. Our method is highly similar to the methodology reported by Larocca et al. (MN50) with a difference consisting of assessing the Envelope ZIKV protein expression, rather than assessing mRNA levels. The technique was used for BALB/c sera and it is reported in our manuscript (Line 132-150 and Figure 4).
• Many of the figures and assays are not described in enough detail for the reader to understand what was done, and the discussion could be much improved, as outlined below.
• Figure 1A: Something seems wrong with this tree -there should be much greater distance between "Other Flaviviruses" and Zika than between different Zika strains. We also can't read the text.
We have made the diagrams bigger and also replace it with a simpler version, we have relied on advice from a phylogenetics expert and threre seems to be no error, as it can be seen that "other Flaviviruses group" is branched out and separated from ZIKV lineages. • Figure 1B: There are no serum dilutions stated for these ELISAs, so this data cannot be interpreted.
We are sorry but this is not an ELISA, Figure 1B is a comparison of identity of our vaccine antigen vs Asian lineage sequences. However, we have made sure all the ELISAs presented in the study are clearly explained and described. We have also made sure it is properly described in Material and Methods.
• Figure 3B: A linear, three-segment y-axis on these graphs is very hard to read. A log scale would be much clearer.
We believe that keeping a three-segment y-axis permits a good comparison of the vaccinated groups with the naïve control group. To help with this, we enlarged the figures and make sure these read properly by allowing more space between axis intervals.
• Ext Data Figure 3C: No dilutions given for ELISA.
We have now included more informative graphs in this figure, including serial dilution graphs for the pre-challenge ELISAs.
• Figure 3E: Please provide a clearer definition for "viremia peak ablation" in either the figure legend or Methods.
We now have change it as: "Avoided 3 day viraemia peak", hoping this will be clearer.
• Figure 4A: If I am reading this correctly, you have averaged together the neutralization curves for all mice in each group. Better would be to show the individual curves for each mouse (with error bars if possible), and/or to plot the MN50 value for each mouse, so readers can see the full variation and complexity of the data. This is needed given your statement on line 143 that neutralization was only seen in a few mice in some groups.
We agree with this, now we have graphed the Microneutralisation (MT 50 titre) instead of the dilutions used for the NT assay. This will allow a better appreciation of the variation of each mouse serum sample for every mouse.
• Figure 4B: What is the capture antigen (and its source) in these ELISAs-DENV E protein, whole DENV virus, or a synthetic peptide of residues 147-166?
Information has been added now in Material and Methods (Line 385-392), these are full DENV1-4 Env proteins from Dundee Cell Products as indicated in the text. which directly addresses the issue of ADE, a major topic of your manuscript. The latter two papers additionally discuss mRNA-based vaccines, a platform which is not mentioned in your intro or discussion. Please discuss how your vaccine compares in neutralizing titer and other characteristics (e.g. species specificity) with these other vaccines, in particular RhAd52.
We agree and have now referenced these papers. Various of these were published while our manuscript was submitted and during a 4-month review by Nat Comm. We have now included all this papers on the discussion. We appreciate the feedback.

Fixed
• Lines 108-109 and throughout: "Sterile" protection might not be the most accurate term when serum viremia is undetectable, in the absence of testing other highly infected tissues (such as spleen) or looking at anamnestic antibody responses in response to viral challenge. Better to be more specific, e.g. "protected from detectable viremia" Fixed, we have added this important comment throughout the manuscript. We appreciate the feedback.
• Line 129-130: I am not sure what you mean by "complete, intermediate, and low sterile protection." Which adjectives correspond to which of the six vaccine conditions listed? Fixed, we have deleted this and changed to Parameters of protection after challenge. • Line 138: "signal-to-noise ratio of >10" -It is not clear how this was calculated. This may be better expressed in terms of a standard deviation or as error bars or a plot of all data points in Figure 4A. This info is shown in material and methods and included in the legend.

Reviewers' comments:
Reviewer #1 (Remarks to the Author): This is a revised manuscript. Although the authors have made some efforts to improve the study, they have not adequately addressed this reviewers previous concerns. 1. Traditional neutralizing assay was not performed, as requested by both reviewers. These are essential data to evaluate the vaccine and to compare with other vaccine candidates in pipeline. 2. The assay described in lines 142-148 is not clear to this reviewer. 3. For construct Env ∆TM, the sentence from lines 126-129 is not clearly supported by the data shown in Figure 3F.
Reviewer #2 (Remarks to the Author): The authors addressed most of the previous concerns. The following minor issues were not adequately addressed.
Minor issues: The authors should provide statistics for Figure 4B and Extended Data Figure 3. The best would be to add a section about Statistical Analyses to the Methods, too.
There are some typos on Figures 3F and 4A. These should be fixed.

Reviewer #3 (Remarks to the Author):
This revised manuscript by Lopez-Camacho and colleagues has addressed most of the concerns raised by reviewers on the original submission. The manuscript now shows that a recombinant chimpanzee adenoviral vector vaccine for Zika virus (ZIKV) is both immunogenic and protective against wild-type ZIKV challenge in BALB/c mice. Specifically, they also show that deletion of the transmembrane region of the prME construct produced better vaccination outcome than full prME gene constructs. The authors suggest that the ChAdOx1 ZIKV vaccine without the transmembrane anchors could be further explored in preclinical and clinical trials.
A couple of areas require additional attention: 1. The explanation on why ChAdOx1 prME and E constructs did not perform as well as ChAdOx1 prME ∆TM and E ∆TM, respectively, is incomplete. The authors showed that the cellular localization of the ZIKV structural proteins are different, depending on the presence or absence of the transmembrane region of the proteins. However, there is no data to show that these differences in cellular localization resulted in differences in antigen secretion or HLA presentation and thus contribute to the observed differences in immunogenicity and vaccine efficacy. Measuring antigen abundance in the culture supernatant of transfected cells would be an informative data to include in this paper.
2. The suggestion that the vaccine-induced antibodies do not enhance DENV infection is problematic. The authors reacted mouse antibodies with ZIKV and then inoculated this mixture onto human U937 cells. It is well established now that there is some degree of Fc-Fc receptor incompatibility between mouse Fc and human Fc receptors that reduces the degree of ADE, in vitro. What is also missing in this experiment is a group of control mice inoculated with a heterologous serotype of DENV before challenge with DENV-2. The evidence to support the notion that vaccination of ChAdOx1 would not produce in DENV enhancing antibodies is thus weak. The 1. Traditional neutralizing assay was not performed, as requested by both reviewers.
These are essential data to evaluate the vaccine and to compare with other vaccine candidates in pipeline.
We thank the reviewer for the feedback. For this round of review, it is 1 out of 3 reviewers that requested a comparison between MN (microneutralisation) and PRNT (Plaque Reduction Neutralisation titre). All the authors in this manuscript agree that our ZIKV challenge efficacy data plus our MN50 assay are strong evidence to support that our vaccine development is effective, and that performing an experimental comparison between other pipeline vaccines is beyond the scope of this manuscript.
We have done an updated revision of the zika vaccines in the pipeline, as the reviewer has kindly suggested, with special attention to vaccines that have completed Phase I clinical trials in humans. For all the trials cited below, the methodology of choice to assess in vitro neutralisation is a microneutralisation assay, rather than a PRNT: 1) a Phase 1 trial performed at UPENN-Inovio Pharma, 2) a Phase 1 trial completed at Walter Reed Army Institute of Research, Silver Spring, MD, USA, 3) a Phase 1 trial done at Saint Louis University, MO, USA, and 4) a Phase 1 clinical trial performed at Beth Israel Deaconess Medical Center, Boston, MA., USA. Furthermore, another 5) Phase 1 trial, performed at National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD, used a neutralisation based on a Zika Reporter Virus Particle Assay (GFP) rather than a PRNT. Below, the references: Due to the limited availability of serum samples, budget and time constraints, and taking into consideration the reviewer suggestion, we have performed a fair head-to-head comparison between our MN assay vs a traditional PRNT assay, using a series of 2-fold dilutions of a highly potent monoclonal neutralising antibody to ZIKV, EDE1 C8 (Characterised in Gavin Screaton's Lab). Our results demonstrate that our MN assay is more stringent at detecting ZIKV infection than PRNT (see graph below, now included in Extended Figure 2). Therefore, the MN50 values in our study are moderate when compared to PRNT. Interestingly, the data gathered from this experiment, mimics exactly the NT comparisons from Dowd et al. in her Science paper (2016), in which the MN50 titers are actually lower than those of that FRNT (Focus Reduction Neutralisation Titre) 50 assay, which is basically a PRNT50 (see Despite PRNT50 is a traditional method, there are numerous vaccine developments that are not using PRNT but VLP-ZIKV; and using many types of reporters, such as luciferase, mCherry or GFP. Some of them used flow cytometry and others have used immunofluorescence to assess infected cells. These methodologies, in addition to Focal Reduction NT and PRNT show great variability between published papers as they used different ZIKV strains, different ZIKV purification methodologies and different cell lines (i.e Vero, BHK-21). Therefore, it will be challenging to reach a meaningful comparison to all vaccines in the pipeline but to compare vaccine developments head to head, which is even more challenging. We have included most of these considerations as a new paragraph in the discussion section. Thank you very much for highlighting this issue, we have now corrected it and made it much clearer. We also have updated the Material and Methods accordingly.
Thank you for this, we have rephrased now the sentence. In an effort to connect the subcellular localisation with antigen expression, we are now including a thorough analysis of expression in both intra-and extracellular compartments for all the immunogens used in this study, which in turn can support antigen secretion and availability after vaccination. The following diagram has been added to Fig 3G and will complement the immunofluorescence data Fig 3F. Also, we have assessed the retention within the cells of the antigens in early expression times, which now has been included as Extended Figure 6 (see below). We are confident that addition of these data will positively improve our manuscript.
Thank you for raising this concern. We are using a human cell line with Fc receptor, and we are testing mice sera. We have included Extended figure 5 showing the ADE in vitro assay by using an anti-DENV2 mouse monoclonal (4G2) antibody as a positive control, in U937 cells. This control provides evidence that that Fc-Fc receptor incompatibility is not an issue in our assay (see graph below). However, we agree with the reviewer it will be a good idea to try a group control mouse inoculated with a heterologous serotype of DENV before challenge with DENV-2. However, this experiment is exceptionally time-consuming and technically difficult as our lab in Oxford is not equipped to handle Category 3 pathogens. We have therefore included this concern and temper the ADE data in the discussion. Thank for your feedback.
3. The use of BALB/c mice for ZIKV infection studies is somewhat limited by the lack of disease outcome in these mice. Critically, any vaccine against ZIKV must be able to prevent infection in key organs where the virus can either persist, such as the testes, or more rarely result in lifethreatening disease, such as the CNS. Most groups investigating Zika have used the A129 mice, which is deficient in type-I interferon receptor. Although how well prevention of disease or persistent ZIKV infection in A129 extrapolates to similar protection in humans is unclear, demonstration of the effectiveness of ChAdOx1 in preventing such outcomes would likely still be necessary for subsequent clinical translation. A discussion on this limitation in data should be considered for this manuscript. This is a very interesting and important point. We acknowledge that our current model may be limited by the fact that we are looking only at viremia, and it is now added into the discussion, as requested. Thank you very much for the feedback, coincidentally we are working now in collaboration with Public Health England to assess efficacy of our vaccine candidates in a lethal challenge model, using A129 mice. As the reviewer suggest, we will look for viral loads in several organs, including the brain as well as testing a vaccine specific to African lineage. This will be another paper we expect to publish in the near future.