Differential prolonged multiomic responses to mRNA and inactivated virus COVID-19 vaccines

Technical Summary: Differential Prolonged Multiomic Responses to mRNA and Inactivated Virus COVID-19 Vaccines

Problem Statement
While SARS-CoV-2 vaccines have been pivotal in controlling the pandemic, a comprehensive, longitudinal understanding of host responses at the systems biology level remains limited. Existing literature often focuses on immunogenicity (antibody titers) or short-term side effects, lacking a multi-omic perspective that integrates transcriptomics, metabolomics, lipidomics, and cytokine profiling over time. Furthermore, direct head-to-head comparisons of the molecular mechanisms underlying different vaccine platforms—specifically the mRNA-based BNT162b2 versus the inactivated virus CoronaVac—are scarce. This study addresses the gap in understanding how these distinct modalities differentially alter the host's multi-omic landscape and how these alterations correlate with the durability and intensity of immune responses.

Methodology
The study employed a prospective cohort design involving 553 adult participants in Hong Kong (274 receiving BNT162b2 and 279 receiving CoronaVac). Participants were seronegative for SARS-CoV-2 (confirmed via ORF8 antigen) and had no prior vaccination history.

• Longitudinal Sampling: Blood samples were collected at three timepoints: baseline (pre-vaccination), one month post-second dose, and six months post-second dose.
• Multi-omic Profiling:
  • Transcriptomics: RNA sequencing of Peripheral Blood Mononuclear Cells (PBMCs) to analyze gene expression.
  • Metabolomics & Lipidomics: Untargeted analysis of plasma using LC-MS/MS platforms to quantify metabolites and lipids.
  • Cytokine Profiling: Quantification of inflammatory cytokines via flow cytometry-based immunoassay.
  • Immunogenicity: Neutralizing antibody levels were measured using surrogate virus neutralization tests (sVNT) and ELISA against the SARS-CoV-2 spike receptor-binding domain (RBD).
• Statistical & Computational Analysis:
  • Differential Analysis: Linear mixed-effect models (Limma/Voom for RNA-seq; GLM for metabolites/lipids) were used to assess longitudinal changes, adjusting for age, sex, and batch effects.
  • Pathway Analysis: Gene Set Enrichment Analysis (GSEA) using Blood Transcription Modules (BTMs) and Hallmark gene sets; KEGG pathway analysis for metabolites.
  • Network Analysis: STRING database and Spearman correlation networks (visualized via Gephi) were used to map gene-metabolite-lipid-cytokine interactions.
  • Predictive Modeling: Machine learning (Random Forest with Boruta feature selection) was applied to baseline multi-omic data to predict post-vaccination immunogenicity (antibody levels) and durability (rate of antibody decline).

Key Results

1. Divergent Multi-omic Signatures:

   • Magnitude of Change: BNT162b2 induced significantly more profound and widespread changes across all omics layers compared to CoronaVac. At one month, BNT162b2 altered 563 transcripts, 50 metabolites, and 48 lipids, whereas CoronaVac altered only 45 transcripts, 25 metabolites, and 14 lipids.
   • Immune Polarization: BNT162b2 strongly upregulated pathways associated with adaptive immunity (T cell activation, B cell enrichment, NK cell signatures) and antiviral responses (Type I interferon, TLR signaling). In contrast, CoronaVac elicited a modest elevation in innate immunity pathways (monocyte and dendritic cell signatures).
   • Neural Function Suppression: A distinct finding was the downregulation of neural function-related genes (e.g., FOSL1, AREG, LHX3, EP300) and metabolites (e.g., D-Sphingosine, Sphingomyelins) specifically in the BNT162b2 group. These changes were not observed in the CoronaVac group and persisted to some extent at six months.

2. Durability and Persistence:

   • Antibody Durability: BNT162b2 recipients maintained significantly higher neutralizing antibody levels at six months compared to CoronaVac recipients, whose levels largely returned to baseline.
   • Omics Persistence: The multi-omic perturbations induced by BNT162b2 were more durable. At six months, 52 of the 563 altered transcripts and a majority of altered metabolites/lipids remained significantly different from baseline in the BNT162b2 group, compared to only 21 transcripts and fewer metabolic changes in the CoronaVac group.
   • Correlation with Decline: Persistently downregulated transcriptomic markers (e.g., PGLYRP1, LHX3) in the BNT162b2 group were positively correlated with the relative decline in antibody levels, suggesting a link between specific molecular signatures and waning immunity.

3. Multi-omic Crosstalk:

   • Network Complexity: The BNT162b2 group exhibited a denser and more complex network of interactions between genes, lipids, and metabolites, particularly involving Lyso-phosphatidylcholine (LPCs) and Sphingomyelins (SMs). This crosstalk was associated with both higher immunogenicity and longer durability.
   • Lipid Mediators: Specific lipids, such as LPC (18:2), were identified as unique features correlating with both heightened immunogenicity and durability in the BNT162b2 cohort. In contrast, CoronaVac responses showed a narrower spectrum of metabolic associations.

4. Predictive Modeling:

   • Baseline multi-omic profiles could predict post-vaccination outcomes. Machine learning models achieved Area Under the Curve (AUC) values ranging from 0.65 to 0.84 for predicting immunogenicity and durability.
   • Transcriptomic data was the most frequently selected feature type across models, indicating its dominant role in forecasting vaccine response.

Significance and Claims
The authors claim that this study provides pivotal insights into the mechanistic distinctions between mRNA and inactivated virus vaccines. The primary contributions are:

• Mechanistic Differentiation: The study delineates that mRNA vaccines (BNT162b2) drive a robust, durable adaptive immune response accompanied by significant metabolic reprogramming and transient neural pathway modulation, whereas inactivated vaccines (CoronaVac) primarily trigger innate immune pathways with less systemic omic perturbation.
• Durability Drivers: The research suggests that the superior durability of the mRNA vaccine is linked to sustained immune gene modifications and specific lipid-metabolite crosstalk (particularly involving sphingolipids and LPCs).
• Personalized Vaccination: The ability of baseline multi-omic profiles to predict individual vaccine performance (AUC 0.65–0.84) supports the integration of multi-omics data to tailor vaccination strategies. The authors advocate for using these profiles to optimize public health outcomes, particularly for vulnerable populations, by aligning vaccine modalities with individual immune landscapes.

The paper maintains a modest tone regarding clinical implications, noting that while neural-related molecular changes were observed in peripheral blood, current evidence does not support a direct clinical risk of neurological disease from the vaccine, and further experimental validation is required to understand the long-term subclinical impacts.