Rational Design of a Chimpanzee Adenoviral-Vector Vaccine Against Yellow Fever Through the Modification of Antigen Transmembrane Domains

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript is well aligned with the scope of Vaccines and addresses a timely and relevant topic: rational antigen design for a ChAdOx1‑vectored yellow fever vaccine and its comparison—functionally and conceptually—with the licensed YF17D vaccine. The experimental design is coherent, the data are generally convincing, and the discussion nicely connects antigen biology, vector platform properties, and outbreak needs.

Specific questions and suggestions

1. The data clearly show that deletion of the C‑terminal TM and/or prM alters secretion and subcellular localization of E, and this correlates with reduced or more variable immunogenicity. It would strengthen the paper if the authors could more explicitly articulate a mechanistic model—at least at the hypothesis level—linking these trafficking/secretory changes to the quality and magnitude of B‑ and T‑cell responses (e.g., antigen form, persistence, or presentation pathways).
2. A key and interesting finding is that robust protection is achieved in the absence of detectable nAbs in this mouse model, both for ChAdOx1‑YFprME and YF17D. The discussion already touches on non‑neutralizing antibodies and T‑cell responses, but this could be made more explicit and structured:

• Can the authors better reconcile their findings with the classical view of nAbs as the main correlation of protection in humans?
• It would be helpful to more clearly distinguish between “correlate” and “mechanism” of protection and to highlight how their data fit with recent work on T‑cell–mediated protection in the absence of nAbs.

3. The intracranial YF17D challenge model is stringent but not physiological. The authors briefly acknowledge limitations, but it would be useful to:

• Provide a short rationale for choosing this model over peripheral challenge in this study.
• Comment more explicitly on how protection in this model may or may not translate to protection against natural, mosquito‑borne infection in humans.

4. Only the full‑length ChAdOx1‑YFprME construct is taken forward into the challenge experiment, although the earlier sections carefully compare four designs. While this is understandable from a practical standpoint, it would be helpful if the authors could:

• Explicitly acknowledge this as a limitation.
• Briefly justify why the other constructions were not included in the challenge (e.g., ethical considerations, clear inferiority in immunogenicity, resource constraints).

This will help readers understand how broadly the conclusions can be generalized across the different antigen designs.

5. The Methods describe the statistical tests in detail, but in several figures the exact n per group and whether data represent one or multiple pooled experiments are only partially clear. I suggest:

• Ensuring that each figure legend explicitly states the number of animals per group, whether data are from a single or combined experiment, and the exact test used.
• Where possible, indicating effect sizes or confidence intervals in addition to p‑values, especially for key comparisons (e.g., viral loads, ELISA titres).

6. The manuscript is generally well written, but there are a few typographical and consistency issues that should be corrected in a final version (e.g., “hardly scalable” →“hardly scalable”could be rephrased as“difficult to scale”;“stablished”→“established”; consistent use of “pre‑membrane” vs “pre-membrane”, “IFN‑γ” vs “IFN‑Y”, etc.). A careful language and style check will improve readability and align with Vaccines formatting standards.
7. The discussion nicely highlights manufacturing advantages of ChAdOx1, but the translational path could be framed a bit more concretely:

• Could the authors briefly comment on how a ChAdOx1‑YFprME vaccine might be positioned relative to existing YF17D vaccines (e.g., as a stockpile back‑up, for contraindicated populations, or for outbreak surge capacity)?

A short paragraph outlining key next steps (e.g., dose‑finding, assessment in non‑human primates, evaluation of pre‑existing anti‑adenovirus immunity) would help readers appreciate the practical implications of this work.

Comments for author File: Comments.pdf

Author Response

Comments 1: The data clearly show that deletion of the C‑terminal TM and/or prM alters secretion and subcellular localization of E, and this correlates with reduced or more variable immunogenicity. It would strengthen the paper if the authors could more explicitly articulate a mechanistic model—at least at the hypothesis level—linking these trafficking/secretory changes to the quality and magnitude of B‑ and T‑cell responses (e.g., antigen form, persistence, or presentation pathways).

Response 1: We thank the reviewer for this suggestion. We have expanded the Discussion section to describe a potential mechanistic link between the observed alterations in antigen trafficking/secretion and the resulting immune responses. Specifically, we now discuss how deletions of prM and/or the C-terminal transmembrane region of E may impair the formation and secretion of subviral particles (SVPs) (ref 34 in ms). As SVPs are known to efficiently present E protein in a particulate, repetitive form that promotes B-cell activation and neutralizing antibody responses, impaired secretion may reduce B-cell stimulation. Conversely, intracellular retention of the antigen may help processing through MHC class I and II pathways, potentially promoting T-cell responses. With this in mind, constructs retaining both transmembrane regions are expected to remain membrane associated and may assemble into SVPs. In contrast, constructs lacking c-terminal transmembrane region appear to produce primarily soluble E protein, which might get processed differently and fail to induce humoral responses. The corresponding text has been added to the Discussion (lines 521-531).

Comments 2: A key and interesting finding is that robust protection is achieved in the absence of detectable nAbs in this mouse model, both for ChAdOx1‑YFprME and YF17D. The discussion already touches on non‑neutralizing antibodies and T‑cell responses, but this could be made more explicit and structured:

• Can the authors better reconcile their findings with the classical view of nAbs as the main correlation of protection in humans?
• It would be helpful to more clearly distinguish between “correlate” and “mechanism” of protection and to highlight how their data fit with recent work on T‑cell–mediated protection in the absence of nAbs.

Response 2: We thank the reviewer for highlighting this important point. We agree that neutralizing antibodies (nAbs) are widely considered the main serological correlate of protection against yellow fever in humans, and we have revised the text to emphasize it. Historically, nAb titres have been used as a serological marker of immunity rather than as direct evidence of the mechanism mediating protection. The work of Theiler and Smith (1937) (ref. 40 in MS), who described immunity “as measured by the antibody titer developed”, established nAbs as a correlate of protection rather than proof of antibody-mediated protection per se. In the revised Discussion, we now more clearly distinguish between a correlate and a mechanism of protection. While nAb titres remain an important correlate in humans, accumulating evidence suggests that additional immune mechanisms may contribute to protection. In particular, previous studies in mice (ref 21; ref 43; ref 44 in MS) demonstrated that protection can occur in the absence of detectable nAbs. These findings are supported by recent human data (ref 41 in MS) and have been further discussed previously (ref 42 in MS). Together, these studies indicate that vaccine-induced virus-specific T cells may play an important role in protection when neutralizing antibodies are absent or below detectable levels. The Discussion section has been revised (lines 533-550).

Comments 3: The intracranial YF17D challenge model is stringent but not physiological. The authors briefly acknowledge limitations, but it would be useful to:

Provide a short rationale for choosing this model over peripheral challenge in this study.

Comment more explicitly on how protection in this model may or may not translate to protection against natural, mosquito‑borne infection in humans.

Response 3: We thank the reviewer for highlighting the difference of the physiological way of the YFV infection compared to the challenge model used in this manuscript.

The intracranial YF17D challenge is lethal but we agree that does not fully replicate the physiological route of natural infection. We chose this model because it is very well-established and validated model in our laboratory (ref 21 in MS), allowing a clear readout of vaccine-mediated protection, and avoids the use of BSL3 facilities required for wild-type YFV. Importantly, peripheral challenge with YF17D is not lethal in immunocompetent mice, except in mice lacking a functional innate antiviral Type I and II interferon response (ref 48 in MS) which may alter and skew the response to vaccination.

We agree that protection in this model may not fully predict protection against natural, mosquito borne infection in humans. Nonetheless, we observed comparable levels of protection with  the current licensed YF17D vaccine, supporting the relevance of the model. We used a challenge dose of 3x10³ PFU of YF17D which is substantially higher than the minimal lethal dose of 1-10 PFU i.c. in mice. While the model may not replicate the natural course of infection in humans, the level of protection demonstrated represents a very stringent surrogate for a severe YF pathology. Likewise, peripherally inoculated YFV is also neurotropic in non-human primates and humans with high sequelae, in addition i.c. inoculated YF17D is viscerotropic in mice and readily detected in liver, spleen and kidneys (see Fig. 4c); this compound virological evidence further supports the utility of this i.c. mouse model as a measure of vaccine efficacy.

Our data may justify assessing the preclinical efficacy of our vaccine candidate in additional, step-up animal models such as Syrian hamsters or non-human primates that are naturally susceptible to wild-type YFV infection and disease, before progression in human clinical trials.

We have included the corresponding observations to the Discussion of the manuscript as well as included more clarifying sentences with regards the challenge model in section 3.4 of the results section.

Comments 4: Only the full‑length ChAdOx1‑YFprME construct is taken forward into the challenge experiment, although the earlier sections carefully compare four designs. While this is understandable from a practical standpoint, it would be helpful if the authors could:

Explicitly acknowledge this as a limitation.

Briefly justify why the other constructions were not included in the challenge (e.g., ethical considerations, clear inferiority in immunogenicity, resource constraints).

This will help readers understand how broadly the conclusions can be generalized across the different antigen designs.

Response 4: We thank the reviewer for this suggestion. We have revised the Discussion to explicitly acknowledge that only the full-length ChAdOx1 YFprME construct was evaluated in the challenge experiment, which represents a limitation in generalizing the results across all antigen designs. Considering 3Rs guidelines of animal use, mostly supported by the principle of Reduction, the other constructs were not included due to a combination of inferiority in humoral immunogenicity in earlier experiments, and resource constraints, which made it impractical and unethical to test all variants in the stringent intracranial challenge. This explanation has been added to the Discussion (lines 571-578)

Comments 5: The Methods describe the statistical tests in detail, but in several figures the exact n per group and whether data represent one or multiple pooled experiments are only partially clear. I suggest:

• Ensuring that each figure legend explicitly states the number of animals per group, whether data are from a single or combined experiment, and the exact test used
• Where possible, indicating effect sizes or confidence intervals in addition to p‑values, especially for key comparisons (e.g., viral loads, ELISA titres).

Response 5: Thank you for pointing out this omission. We have updated all figure legends to explicitly state the number of animals per group, whether the data are from a single experiment or pooled from multiple experiments, and the exact statistical tests used. For key comparisons, we have also included effect sizes and confidence intervals in addition to p values. Figure 4 has been updated to include full statistical analysis, and the Statistical Analysis section of the Materials and Methods has been revised to comprehensively describe all tests performed.

Comments 6: The manuscript is generally well written, but there are a few typographical and consistency issues that should be corrected in a final version (e.g., “hardly scalable” →“hardly scalable”could be rephrased as“difficult to scale”;“stablished”→“established”; consistent use of “pre‑membrane” vs “pre-membrane”, “IFN‑γ” vs “IFN‑Y”, etc.). A careful language and style check will improve readability and align with Vaccines formatting standards.

Response 6: Thank you for your detailed revision. I have carefully gone through the manuscript and edited a few typographical words, language check and readability.

Comments 7: The discussion nicely highlights manufacturing advantages of ChAdOx1, but the translational path could be framed a bit more concretely:

Could the authors briefly comment on how a ChAdOx1‑YFprME vaccine might be positioned relative to existing YF17D vaccines (e.g., as a stockpile back‑up, for contraindicated populations, or for outbreak surge capacity)?.

A short paragraph outlining key next steps (e.g., dose‑finding, assessment in non‑human primates, evaluation of pre‑existing anti‑adenovirus immunity) would help readers appreciate the practical implications of this work.

Response 7: We thank the reviewer for this suggestion. We have revised the Discussion to include a paragraph highlighting the potential positioning of ChAdOx1 YFprME relative to existing YF17D vaccines, including use as a stockpile backup, for populations with contraindications, and for rapid surge capacity during outbreaks. We also added the next steps, such as the evaluation of efficacy in Syrian Hamsters or non-human primates, to guide the design of future clinical trials. This paragraph has been added to the Discussion (lines 602-608).

Reviewer 2 Report

Comments and Suggestions for Authors

Four ChAdOx1 YF vaccines were constructed based on the pre-membrane (prM) and envelope (E) proteins, with or without the transmembrane domain using the ChAdOx1 adenoviral vector. Immunogenicity of four vaccines were evaluated in Balb/c mice against intracranial challenge with the YF17D virus. This research has reference value for vaccine development.

Figure 1.b lacks the indication of molecular weight and a negative control. Most importantly, according to the design scheme of Figure 1a, the full-length protein is consistent with the PrME and the truncated form of the E protein, and is larger than PrME△TM and E△TM. However, it appears smaller in Figure 1.b, which is illogical. It is suggested that the start and end amino acid positions of the protein be marked in Figure 1a.

References3, 20, 40, 41 The first letter of each word in the title is capitalized differently from the first letter of the first word in other references.

The figure2d does not clearly indicate the time of serum collection. If the serum was collected on the day of the prime test, then all the P values would be 0. However, in the d figure, the P group is clearly not 0.

It described the changes in body weight and clinical symptoms after the poisoning, but Figure 4b only showed the weight change values, without clinical scoring data.

There were no data in the form of text indicating the survival rates of the mice in each experimental group after the poisoning, nor were there any lesion diagrams of tissue damage.

Author Response

Comments 1: Four ChAdOx1 YF vaccines were constructed based on the pre-membrane (prM) and envelope (E) proteins, with or without the transmembrane domain using the ChAdOx1 adenoviral vector. Immunogenicity of four vaccines were evaluated in Balb/c mice against intracranial challenge with the YF17D virus. This research has reference value for vaccine development.

Figure 1.b lacks the indication of molecular weight and a negative control. Most importantly, according to the design scheme of Figure 1a, the full-length protein is consistent with the PrME and the truncated form of the E protein, and is larger than PrME△TM and E△TM. However, it appears smaller in Figure 1.b, which is illogical. It is suggested that the start and end amino acid positions of the protein be marked in Figure 1a.

Response 1: We thank the reviewer for noticing these differences. The molecular weight markers and the relevant controls have been added to Figure 1b and the whole blots can be found in the Supplementary Figure A1.

Proteins entering ER initially acquire high-mannose N-glycans. During their transit to the Golgi, these mannose residues are replaced by additional sugars generating complex N-glycans, which can substantially affect the electrophoretic mobility on SDS-PAGE. Our results show that the deletion of the TM region enhance secretion, suggesting that a larger fraction of the protein is able to process through Golgi. Consistent with this localisation, they are expected to predominantly carry high-mannose glycans. We wanted to test this hypothesis and we treated the samples with PNGase F, which confirmed this. Therefore, the observed mobility differences between TM-deleted and TM-containing constructs are consistent with differential trafficking and glycan maturation. Additionally, the C-terminal transmembrane element of the envelope protein is arranged in an antiparallel coiled-coil hairpin structure (Fritz et al. 2011 Journal of virology) which might induce differential migration rates on SDS-PAGE depending on detergent binding ratios (Rath et al. 2009 Proc Natl Acad Sci USA). Consequently, the hairpins of the TM-constructs might show fastest migration as compared to ΔTM constructs.

We have included in the Results section 3.1 the information described as well as added in figure legend 1 the amino acid number relative to the YFV polyprotein sequence.

Comments 2: References 3, 20, 40, 41 The first letter of each word in the title is capitalized differently from the first letter of the first word in other references.

Response 2: Thank you to the reviewer for spotting these typos. We have updated and corrected them accordingly along with additional references: references edited are: 3,20, 40, 43, 50, 51.

Comments 3: The figure 2d does not clearly indicate the time of serum collection. If the serum was collected on the day of the prime test, then all the P values would be 0. However, in the d figure, the P group is clearly not 0.

Response 3: Thank you to the reviewer to point out the need of clarification of Figure 2d. Blood samples were withdrawn four weeks after immunisation and we have updated the figure legend with the following sentence: “IgG anti-YF envelope protein (E) titres from sera of mice immunized with ChAdOx1 vaccines, withdrawn four weeks after prime (P) or Prime-boost (B) immunization”

Comments 4: It described the changes in body weight and clinical symptoms after the poisoning, but Figure 4b only showed the weight change values, without clinical scoring data.

Response 4: In this intracranial model, the deterioration and weight loss occur very rapidly (Fig-4b). Therefore, we did not include quantitative or cumulative scoring to assess disease. Additional criteria for humanly end-point apart from weight loss, are described in the Material and Methods and include hind limb paralysis, weakness, and ruffled fur. These criteria were approved by our institutional ethical review board and were strictly followed during the experiments.

Comments 5: There were no data in the form of text indicating the survival rates of the mice in each experimental group after the poisoning, nor were there any lesion diagrams of tissue damage.

Response 5: Thank you for this suggestion. None of the vaccinated groups required euthanasia due to apparent neurological involvement, and all animals survived corresponding to 100% survival. This have been emphasized in the Results (lines 456-462).