Research ResourcesINFECTIOUS DISEASE

Multicohort analysis reveals baseline transcriptional predictors of influenza vaccination responses

See allHide authors and affiliations

Science Immunology  25 Aug 2017:
Vol. 2, Issue 14, eaal4656
DOI: 10.1126/sciimmunol.aal4656
  • Fig. 1 Overview of the data analysis strategy.

    The meta-analysis was carried out on young and older influenza vaccination cohorts. Individual gene and module signatures were validated using independent cohorts.

  • Fig. 2 Vaccination cohorts used to define and validate influenza vaccination response gene and module signatures.

    (A) The four discovery cohorts each included young and older participants. Age cutoffs are indicated by the dashed horizontal lines. In several studies, gene expression data were collected for a subset of individuals (filled circles) enriched for high and low responders, as previously described (5). Two cohorts were used to independently validate the young and older response signatures. (B) The discovery and validation cohorts spanned five vaccination seasons. Numbers indicate the total count of participants in each study. The number of participants who met the age range criteria used for the young and older groups and the subset used in the transcriptional profiling analysis are shown in fig. S1.

  • Fig. 3 The adjMFC end point is independent of baseline titers.

    An illustration of our approach for computing adjMFC. The relationship between baseline titers and (A) MFC or (B) adjMFC in SDY404. Vertical lines separate the bins used for standardization, and the inset table indicates the P value resulting from the test for correlation. Correlation strengths and P values shown were based on Spearman’s rank correlation. Note that in this example, an outlier with high day 0 titer was removed when computing the adjMFC (see Methods).

  • Fig. 4 Identification of individual genes that predict vaccination response in young individuals.

    The x axes correspond to standardized mean difference, referred to as effect size (ES), between high and low responders, computed as Hedges’ g, in log2 scale. The size of the rectangles is inversely proportional to the standard error of mean (SEM) in the individual cohort. Whiskers represent the 95% confidence interval. The diamonds represent overall mean difference for a given gene with combined support across the discovery cohorts. The width of the diamonds represents the 95% confidence interval of overall mean difference.

  • Fig. 5 Identification of gene modules that predict vaccination response in young or older individuals.

    (A) The QuSAGE activity for all gene modules that were significantly different between low and high responders in the discovery cohorts. Red indicates increased average expression of genes in the module among high vaccine responders. (B) Individual genes that comprise the three gene modules that predict vaccination response and were validated in the validation cohort (FDR ≤ 10%) in young individuals. Colors indicate the log2 gene expression fold changes comparing high responders versus low responders, with red indicating increased expression among high vaccine responders.

  • Fig. 6 Validation of gene expression signature as a baseline predictor of the influenza vaccination response in young individuals.

    (A) The geometric mean of GRB2, ACTB, MVP, DPP7, ARPC4, PLEKHB2, and ARRB1 z-scored expression values (response score) was calculated for low, moderate, and high responders in the validation cohort (SDY80). (B) ROC curve for classifiers designed to separate individual participants as high responders versus low responders or moderate responders versus low responders in the validation cohort (SDY80). CI, confidence interval. (C) Temporal behavior of response score in the validation cohort (SDY80) for low, moderate, and high responders. Each point depicts an individual participant, and each point group is summarized by a boxplot. Significant P values are indicated above the data for comparisons of low and high responders and below the data for comparison between baseline and day 1 after vaccination.

  • Fig. 7 Baseline activity of the BCR signaling gene module (M54) is associated with influenza vaccination responses in young individuals.

    QuSAGE was used to calculate the PDF for the gene module activity using baseline data in the (A) discovery cohorts (SDY63, SDY404, SDY400, SDY212, and the combination) and (B) validation cohort (SDY80). (C) Temporal behavior of gene module in the validation cohort (SDY80) for low, moderate, and high responders. Each point depicts an individual participant, and each point group is summarized by a boxplot. Significant P values are indicated above the data for comparisons of low and high responders and below the data for comparison between baseline and day 1 after vaccination.

  • Fig. 8 Inverse correlation of baseline differences between young and older participants.

    (A) Gene effect sizes and (B) module activities comparing high and low responders were calculated in young and older individuals. All values were calculated using data from the discovery cohorts. (A) Significant genes for young (squares) individuals in the discovery cohorts are highlighted in black. (B) Significant modules for young (squares) and older (triangles) individuals in the discovery cohorts are highlighted in black.

  • Table 1 Validation of gene module activities that predict vaccination response in the discovery cohort for young adults.

    Shaded modules were identified in the discovery cohorts and then independently validated (FDR ≤ 10% and P ≤ 0.01).

    Gene moduleDiscovery cohortsValidation cohort
    PFDRGene module
    activity
    PFDRGene module
    activity
    BCR signaling (M54)0.0100.2750.1210.0000.0020.173
    Platelet activation
    (III) (M42)
    0.0020.2750.1420.0080.0420.140
    Inflammatory
    response (M33)
    0.0060.2750.1900.0220.0800.193
    Transmembrane
    transport (II)
    (M191)
    0.0100.275−0.0560.0820.227−0.038
    TBA (M198)0.0090.2750.1770.1470.3230.101
    TBA (M72.1)0.0080.275−0.1270.2190.4010.073
    TBA (M72.0)0.0060.275−0.1370.4340.6680.054
    E2F1 targets (Q4)
    (M10.1)
    0.0030.275−0.0720.4860.6680.035
    Enriched in T cells (II)
    (M223)
    0.0020.275−0.1620.5770.684−0.057
    Enriched in activated
    dendritic cells/
    monocytes (M64)
    0.0090.2750.2500.6220.6840.066
    E2F1 targets (Q3)
    (M10.0)
    0.0070.275−0.0510.9080.9080.005

Supplementary Materials

  • immunology.sciencemag.org/cgi/content/full/2/14/eaal4656/DC1

    Fig. S1. Distribution of low (blue), moderate (purple), and high (red) responders in the discovery and validation cohorts.

    Fig. S2. Genes that predict vaccination response in young individuals when comparing moderate responders versus low responders.

    Fig. S3. Performance of genes significantly different in young high versus low responders.

    Fig. S4. Baseline activity of the platelet activation (III) (M42) gene module is associated with influenza vaccination responses in young individuals.

    Fig. S5. Baseline activity of the inflammatory response (M33) gene module is associated with influenza vaccination responses in young individuals.

    Fig. S6. Validation of GRB2, ACTB, MVP, DPP7, ARPC4, PLEKHB2, and ARRB1 as predictors of influenza vaccination response in the validation cohort (SDY80) after correction for cell subset proportions.

    Table S1. Characteristics of the discovery and validation cohorts for young and older participants.

    Table S2. Gene module activities that are associated with vaccination response in the discovery cohorts for older participants.

    Table S3. Validation of gene modules that are associated with vaccination response in KEGG and Reactome and the modules defined in Obermoser et al. for young participants.

  • Supplementary Materials

    Supplementary Material for:

    Multicohort analysis reveals baseline transcriptional predictors of influenza vaccination responses

    HIPC-CHI Signatures Project Team* and HIPC-I Consortium

    *Corresponding authors. Email: pkhatri@stanford.edu (P.K.); rgottard@fredhutch.org (R. Gottardo); shenorr@technion.ac.il (S.S.S.-O.); john.tsang@nih.gov (J.S.T.); steven.kleinstein@yale.edu (S.H.K.);

    Published 25 August 2017, Sci. Immunol. 2, eaal4656 (2017)
    DOI: 10.1126/sciimmunol.aal4656

    This PDF file includes:

    • Fig. S1. Distribution of low (blue), moderate (purple), and high (red) responders in the discovery and validation cohorts.
    • Fig. S2. Genes that predict vaccination response in young individuals when comparing moderate responders versus low responders.
    • Fig. S3. Performance of genes significantly different in young high versus low responders.
    • Fig. S4. Baseline activity of the platelet activation (III) (M42) gene module is associated with influenza vaccination responses in young individuals.
    • Fig. S5. Baseline activity of the inflammatory response (M33) gene module is associated with influenza vaccination responses in young individuals.
    • Fig. S6. Validation of GRB2, ACTB, MVP, DPP7, ARPC4, PLEKHB2, and ARRB1 as predictors of influenza vaccination response in the validation cohort (SDY80) after correction for cell subset proportions.
    • Table S1. Characteristics of the discovery and validation cohorts for young and older participants.
    • Table S2. Gene module activities that are associated with vaccination response in the discovery cohorts for older participants.
    • Table S3. Validation of gene modules that are associated with vaccination response in KEGG and Reactome and the modules defined in Obermoser et al. for young participants.

    Download PDF

    Files in this Data Supplement:

Navigate This Article