Fig 1.
A principal axis of the neonatal cortex indexed by multimodal MRI.
(A) Average cortical neuroimaging metrics in a cohort of healthy, term-born neonates (n = 292). Metrics derived from structural MRI (T1/T2 contrast and cortical thickness) and diffusion MRI model parameters using DTI (FA and MD) and NODDI (fICVF and ODI). Right: cortical ROI based on anatomical references with corresponding developmental transcriptomic data (S1 Fig). (B) Z-scored cortical metrics are shown for each participant grouped within each cortical ROI. (C) Top: cortical representations of the first 2 principal components (PC1 and PC2) derived from the PCA of the regional MRI metric data in B. Bottom: position of each cortical ROI in PCA state space; the position of each region is dictated by its component score (D) for the first 2 principal components. Regions are labelled and coloured by PC1 score. (D) PCA scores of each metric for the first 2 principal components, coloured by PC1 score. See https://github.com/garedaba/baby-brains/tree/master/figures. A1, primary auditory cortex; DLPFC, dorsolateral prefrontal cortex; DTI, diffusion tensor imaging; FA, fractional anisotropy; fICVF, intracellular volume fraction; IPC, inferior parietal cortex; ITC, inferior temporal cortex; M1, primary motor cortex; MD, mean diffusivity; MFC, medial frontal cortex; MRI, magnetic resonance imaging; NODDI, neurite orientation dispersion and density imaging; ODI, orientation dispersion index; OFC, orbitofrontal cortex; PCA, principal component analysis; ROI, regions of interest; S1, primary sensory cortex; STC, superior temporal cortex; V1, primary visual cortex; VLPFC, ventrolateral prefrontal cortex.
Fig 2.
Genes associated with neuronal differentiation are differentially expressed along the principal imaging axis.
(A) Volcano plot showing enrichment of GO terms (biological processes) in genes with age-corrected expression levels positively correlated with PC1. Significantly enriched terms (FDR < 0.05, reference: protein-coding genes) are labelled. (B) Gene co-expression analysis of all PC+ genes revealed 2 modules). Intra-modular connections are shown with node size and colour indicating strength and edge thickness and colour indicating weight. (C) Differential expression of PC+ genes across cortical regions (top) measured using LMD microarrays of the cortical plate in two 21-pcw fetal samples (https://www.brainspan.org/lcm/). Heatmap shows relative expression of all 71 PC+ genes in the inner and outer cortical plate of each labelled region. (D) Total expression (in TPM) of PC+ (top) and PC− (bottom) genes in single cells (n = 572) extracted from cortical regions in an independent single-cell RNA-seq survey of the mid-gestational fetal cortex [51]. Scatterplots show mean TPM averaged over cells in each region, correlated with each region’s PC1 score. See https://github.com/garedaba/baby-brains/tree/master/figures and https://github.com/garedaba/baby-brains/tree/master/results/wgcna. CNS, central nervous system; DLPFC, dorsolateral prefrontal cortex; FDR, false discovery rate; GO, Gene Ontology; IPC, inferior parietal cortex; ITC, inferior temporal cortex; LMD, laser microdissection; M1, primary motor cortex; OFC, orbitofrontal cortex; RNA-seq, RNA sequencing; S1, primary sensory cortex; STC, superior temporal cortex; TPM, transcripts per million; VLPFC, ventrolateral prefrontal cortex.
Fig 3.
Cell-specific gene expression is associated with cortical morphology at birth.
(A) UMAP embedding of 86 cell types based on trajectories of relative gene expression over time recovers annotated cell classes. Subplots reflect enrichment ratios of cell classes in PC+ and PC− gene sets (darker colour represents higher enrichment ratio). (B) Enrichment ratio for fetal marker genes expressed by each cell class is shown for PC+ (left) and PC− (right) gene sets. See https://github.com/garedaba/baby-brains/tree/master/figures. OPC, oligodendrocyte precursor cell; UMAP, Uniform Manifold Approximation and Projection.
Fig 4.
Tissue maturity correlates with regional variation in cortical morphometry at birth.
(A) Developmental patterns of mean cortical gene expression illustrated in each specimen for all 120 regionally variant genes (PC+ and PC−), ordered by age. (B) The relationship between predicted and true sample age for all regional samples (n = 198 samples from n = 21 brains) in the PsychENCODE dataset aged between 50 and 400 postconceptional days (8 pcw to 4 postnatal months), estimated using SVR and LOO cross-validation. The SVR model was validated using additional samples from the BrainCloud dataset (n = 46 samples). Shaded area indicates 95% CI. (C) Left: The correlation between regional PC1 score and predicted tissue maturity is shown for each sample during gestation. Error bars show 95% CI for regional age predictions over 1,000 bootstrapped gene samples. Right: PC1 correlation is plotted against specimen age for each brain. Shaded area indicates 95% CI for linear model fit over bootstrap samples. See https://github.com/garedaba/baby-brains/tree/master/figures. CI, confidence interval; LOO, leave-one-out; SVR, support vector regression.
Fig 5.
Disruption of cortical development in preterm-born infants.
(A) Left: group difference in individual variance across multiple neuroimaging metrics explained by the principal imaging axis in term (blue) and preterm (green) infants (left). Right: the relationship between age at scan and variance explained by PC1 across all cortical metrics (right). Regression lines are shown for term (blue) and preterm (green) infants with 95% CI. (B) Group differences in regional T1w/T2w contrast ordered by position along PC1 (linear regression shown with 95% CI). (C) Enrichment of gene sets from 10 fetal cortical cell classes (top: neuronal; middle: nonneuronal; bottom: precursor) based on genes significantly associated (FDR p < 0.05) with group differences in T1w/T2w contrast at 10 time points in the preterm period. (D) Enrichment of genes both expressed by each cell type and significantly correlated with T1/T2 in at least 1 age window in DEGs measured in an experimental mouse model of preterm brain injury [57]. See https://github.com/garedaba/baby-brains/tree/master/figures. CI, confidence interval; DEG, differentially expressed gene; FDR, false discovery rate.
Fig 6.
Cellular pathways associated with genes expressed by oligodendrocytes and developmental alterations in the preterm cortex.
Left: Genes expressed by oligodendrocytes in the fetal cortex and significantly associated with group differences in T1w/T2w contrasts across at least 3 age windows are shown. Dark green indicates periods where gene expression and T1w/T2w contrast were significantly correlated for each gene (FDR p < 0.05) across the preterm period. Right: PPI networks derived using STRING. Top functional enrichments of molecular pathways are shown where applicable. Genes associated with listed enriched pathway and genes differentially expressed in an animal model of preterm brain injury are highlighted. See https://github.com/garedaba/baby-brains/tree/master/data/gene_lists. FDR, false discovery rate; PPI, protein–protein interaction; STRING, Search Tool for the Retrieval of Interacting Genes/Proteins.
Fig 7.
Cellular pathways associated with genes expressed by microglia and developmental alterations in the preterm cortex.
Left: Genes expressed by microglia in the fetal cortex and significantly associated with group differences in T1w/T2w contrasts across at least 3 age windows are shown. Dark blue indicates periods where gene expression and T1w/T2w contrast were significantly correlated for each gene (FDR p < 0.05) across the preterm period. Right: PPI networks derived using STRING. Top functional enrichments of molecular pathways are shown where applicable. Genes associated with listed enriched pathway and genes differentially expressed in an animal model of preterm brain injury are highlighted. See https://github.com/garedaba/baby-brains/tree/master/data/gene_lists. FDR, false discovery rate; PPI, protein–protein interaction; STRING, Search Tool for the Retrieval of Interacting Genes/Proteins.