Figures and data
Gut microbiome community changes in SCD patients (n=98) compared to healthy controls (n=46).
Sample whole community sequencing was profiled using the MetaPhlAn taxa marker database 20. (a) Distribution of sample alpha diversity measured as Shannon diversity for patients and controls. (b) A generalized linear model was used to determine the effect size and association of bacteria taxa abundance with sample metadata including SCD status, age, race, ethnicity, and gender, all modeled as fixed effects. Taxa with q-value < 0.05 for any effect are shown. (c) Distribution of Firmicutes to Bacteroidetes (F:B) ratio in the gut microbiome of patients and controls. (d) Scatter plot of health and disease indicator scores per sample. (e) Sample whole community sequencing was profiled using the HUMAnN pathway marker database 20. A generalized linear model was used to determine the effect size and association of pathway abundance with sample metadata including SCD status, age, race, ethnicity, and gender, all modeled as fixed effects. Additionally, the F:B was also included as fixed effect. Pathways with q-value < 0.05 for any effect are shown. In a and c, samples with values +/- 3 s.d. from mean of all samples were removed. M.W.W.- Mann-Whitney-Wilcoxon test. F:B, it was inversely correlated with health indicators and positively correlated with disease indicators (Supplemental Figure 1a), providing evidence that provirus enrichment mirrors changes in the bacterial species present in the sample.
Gut microbiome viral population is altered in SCD.
Sample whole community sequencing was assembled and viral sequences were identified using marker genes and sequence features. The number of viral sequences (a) and the fraction of provirus sequences (b) were compared between patients and controls. M.W.W= Mann-Whitney-Wilcoxon test.
Aged neutrophil percentage (n=57) and gut microbiome and virome feature (n=98) correlations with clinical measures and blood molecular immune markers.
Correlations were measured with Spearman ρ and significance was measured using permutation testing. Rows are grouped as community metrics, taxa, functional pathways, and viral features. Columns are grouped as hematologic/inflammatory markers, hemolysis markers, liver marker, kidney marker, acute care burden, and cytokines/chemokines. Boxes denote nominal p < 0.05; * denotes FDR q < 0.05; † denotes FDR q < 0.10. FDR correction was applied within each feature-group by marker-block family. For the number of samples assayed for each clinical and molecular measure, see Supplemental Tables 4-5.
Highly conserved prophage sequences are shared by gut microbiomes of SCD patients.
(a) Patient microbiomes sharing a highly conserved prophage sequence visualized as fully connected networks. Conserved prophages had 99% sequence identity over at least 70% of bidirectional sequence coverage. Node color indicates predicted prophage activity: active (blue), dormant (orange), and undetermined (gray). (b) Prophage sequence distribution of high-level viral protein functions compared between clustered (n=1,249) and singleton (n=3,713) prophages. Significance tested with a Mann-Whitney-Wilcoxon test: **** = p < 0.0001.
A model of interactions between the gut microbiome and markers of inflammation in the blood of SCD patients.
Using multiple data modalities, we begin to understand the complex interaction of gut microbiome changes and immune system activation in the pathology of SCD. Clinical and molecular features are shown as boxes, microbiome and virome features as ovals. Edges represent associations and do not represent causal or mechanistic interactions. Edge style indicates significance: dotted line (nominal p < 0.05), dashed line (FDR q < 0.10), solid line (FDR q < 0.05)
Further comparison of virome measures.
(a) Comparison of viral population and bacterial population community-level metrics. (b) Histogram of active provirus fraction for SCD samples. Value of correlations is shown for comparisons with p < 0.05 otherwise comparison cell is gray.
Enrichment in provirus or lysogenic virus prediction across multiple labeling methods.
The determination of a lysogenic virus can be made by either looking for the virus sequence integrated into a host genome or by predicting that a viral sequence has lysogenic potential. We utilized multiple methods that rely on different approaches to determine that the provirus enrichment observed is robust to metho
Correlation of flow cytometry and EHR measures of patient white blood cell populations.
Correlation measured with Pearson correlation coefficient. Abbreviations: WBC-white blood cells, ANC-absolute neutrophil count.
Provirus length distribution with kernel density estimation.
Prophage cluster sequence homology with bacterial hosts.
Representative cluster sequences with blast sequence homology to bacterial species sequences were aggregated to the genus level. Count represents the number of clusters homologous to species in the genus.
Genome content comparison between predicted active (n=136) and dormant (n=4739) prophages in the SCD gut microbiome.
Significance tested with a Mann-Whitney-Wilcoxon test: * = 0.01 <= p < 0.05; ** = 0.001 < p <= 0.01.
Study participant demographics by cohort.
SCD patient (n=98) medical history.
SCD patient (n=98) treatment history.
Clinical measures for SCD patients (n=98).
Molecular inflammatory measures for SCD patients (n=98).
