Molecular signatures of cortical expansion in the human fetal brain

The third trimester of human gestation is characterised by rapid increases in brain volume and cortical surface area. A growing catalogue of cells in the prenatal brain has revealed remarkable molecular diversity across cortical areas.1,2 Despite this, little is known about how this translates into the patterns of differential cortical expansion observed in humans during the latter stages of gestation. Here we present a new resource, μBrain, to facilitate knowledge translation between molecular and anatomical descriptions of the prenatal developing brain. Built using generative artificial intelligence, μBrain is a three-dimensional cellular-resolution digital atlas combining publicly-available serial sections of the postmortem human brain at 21 weeks gestation3 with bulk tissue microarray data, sampled across 29 cortical regions and 5 transient tissue zones.4 Using μBrain, we evaluate the molecular signatures of preferentially-expanded cortical regions during human gestation, quantified in utero using magnetic resonance imaging (MRI). We find that differences in the rates of expansion across cortical areas during gestation respect anatomical and evolutionary boundaries between cortical types5 and are founded upon extended periods of upper-layer cortical neuron migration that continue beyond mid-gestation. We identify a set of genes that are upregulated from mid-gestation and highly expressed in rapidly expanding neocortex, which are implicated in genetic disorders with cognitive sequelae. Our findings demonstrate a spatial coupling between areal differences in the timing of neurogenesis and rates of expansion across the neocortical sheet during the prenatal epoch. The μBrain atlas is available from: https://garedaba.github.io/micro-brain/ and provides a new tool to comprehensively map early brain development across domains, model systems and resolution scales.


GAN model architecture and training
We used the pix2pix architecture (Figure 1b) to synthesise Nissl-stained images from label annotations in order to replace artefacts within tissue sections.

Training data
We trained the pix2pix model on 1000 pairs of 256 × 256 image patches from 20μm resolution Nissl-stained sections and corresponding label annotations (Figure 1a,b).Each pair of training patches was visually inspected to ensure no tissue artefacts were present in the histological data and good alignment was observed between tissue section and anatomical labels.Pairs that failed visual inspection were rejected and replaced until 1000 pairs were selected.Pairs were automatically excluded if >75% pixels were labelled as background.Patches were randomly sampled from n=73 sections, with the remaining n=8 sections forming a validation set to evaluate model performance.Validation sections were spread evenly through the cerebral hemisphere.To increase the size and diversity of our training set and limit overfitting, we applied common data augmentation steps including random crop, jitter and flipping of images during training.

Model training
The model was trained in steps, alternating between training the generator and the discriminator with real and synthetic samples.As in Isola et al. 1 the discriminator loss was divided by 2 to slow down its learning rate compared to the generator and both D and G were trained using the Adam optimiser (β != 0.5, β " = 0.999) with a learning rate of 0.0002 and batch size of 1 for a total 100 epochs.We set the regularisation parameter, λ, to 1. Alternative parameter settings are explored in Figure S2.Finally, the 3-channel RGB label images used for training were transformed to 21-channel images (20 tissue labels + 1 background label) with each channel containing 1s in pixels belonging to a given label (0 otherwise).

Model evaluation
We retained eight full histological sections spread evenly through the cerebral hemisphere as a validation set.No image patches used for training were drawn from this dataset.To validate model performance, we split each Nissl-stained section and its corresponding anatomical label image into non-overlapping 256 ×256 patches.Labelled patches were used to generate new synthetic Nissl-stained patches by the trained generator.
To quantify visual and textural similarity between synthetic and ground truth Nissl images, we converted each image patch from RGB to HSV (Hue, Saturation, Value) format and calculated the similarity between hue and saturation values across both patches as: Where  ! and  " are the vectors of pixel hue or saturation values for ground truth and synthetic images patches, respectively, normalised to unit length.3][4] After removing the fully-connected classification layers of the pre-trained VGG19 model, each image patch (real and synthetic) was passed through the network, resulting in a 512 length vector output by the final layer to act as a high-level feature representation of each patch.
Similarity between feature vectors was calculated as above, baseline measures of hue, saturation and perceptual similarity were calculated after randomising pixels within each ground truth patch.

Template construction
Following automated repair of major tissue artefacts present in the data, we aimed to develop a 3-dimensional reconstruction of the fetal brain to facilitate comparison with in vivo MR imaging data.

Slice-to-slice alignment
Repaired Nissl-stained images were converted to greyscale and padded with a 200 pixel zero-filled border to allow for large translations while retaining the section in the field-of-view during image registration.Corresponding label images were converted from RGB format to one-channel label images, with labels numbered according to a look-up table (Table S2).
For initial alignment, we implemented a graph-based, slice-to-slice registration to a chosen reference section via shortest-path transforms. 5,6The central section was chosen as the reference, and pairwise rigid transforms were estimated between each section and up to five neighbouring sections in the direction of the reference.After each transform, a weight is calculated between the aligned sections: Where the weighting, ω, between two aligned images,  and , depends upon their pixelwise correlation, , after alignment weighted by the number of intermediate sections between them (based on the absolute difference of section index, ).The hyperparameter, , was set to 0.5 and acts as a penalty on skipping slices, with a higher value penalising transforms that skip intermediate sections. 5,6Dijkstra's shortest-path algorithm was used to calculate the set of transforms with lowest cost to align a given section to the reference. 5,6The selected transforms were then composed and applied to both the image and its corresponding labels to bring all sections into approximate alignment with the central slice.
Aligned sections were stacked along the anterior-posterior axis into a 3D array with resolution 0.02 × 0.02 × 0.5mm.
Section thickness was determined by the sampling strategy of the original data (detailed in Ding et al. 7 ).Consecutive sections included in the atlas were spaced approximately 0.5 mm apart (Table S1) and were thus assigned a nominal thickness of 500m.Due to the sampling strategy employed during tissue sampling and histology, tissue sections were sparsely sampled along the length of the cerebral hemisphere.Where full sections were missing, or excluded, or the distance between adjacent slices was larger than 0.5mm, we repeated the preceding sections were up to 5 times to fill the gap.This strategy was selected to limit abrupt transitions between neighbouring sections, preserve the overall volume of the hemisphere and allow for volumetric registration in subsequent steps.To account for missing slices at the anterior and posterior poles, manual labels were drawn following the tissue contours of the adjacent slices to create a synthetic cortical label.The trained pix2pix generative model was then used to generate a Nissl-stained coronal section to append to each pole.This process resulted in a 3D NIFTI volume of voxel size 1420 × 2678 × 125 and voxel dimensions 0.02 × 0.02 × 0.5mm (Figure 1e; Figure S17).

Affine registration to a fetal brain shape reference
We employed a population-based average anatomical image: specifically the 22-week timepoint of the Gholipour et al.
spatio-temporal fetal MRI atlas as a shape prior for 3D reconstruction (Figure S3). 8 The Gholipour MRI atlas contained T2-weighted anatomical fetal MRI templates and a set of 50 anatomical brain tissue labels at 1mm isotropic voxel resolution.We downloaded the T2-weighted template image and accompanying tissue labels for the 22-week timepoint, removing extracerebral CSF, midbrain and cerebellar structures and matching tissue labels of the MRI atlas to the μBrain tissue labels.We upsampled the MRI template and label images to 50μm isotropic resolution before cropping and rotating into approximate alignment with the 3D μBrain volume.We converted the MRI volume into a Nissl-like contrast using the trained GAN model by slicing the label image coronally and passing each slice as input to the trained model (Figure S3c-d).Nissl-contrast images were re-stacked into a 3D volume and used as an anatomical prior for registration.
Using the shape prior, we performed an iterative affine registration procedure 5,6 by first estimating an affine alignment of the MRI-based anatomical prior to the histological volume.The transformed anatomical prior was then resliced, with each coronal section acting as a target for 2D registration with the corresponding histological section.The registered histological sections then form the target for the next 3D registration (Figure S17).This process was repeated for a total 5 iterations, producing a final 3D volume with aligned coronal slices and a global shape approximately matched to the in utero fetal brain (Figure 1e).A final 2D affine registration was calculated between the original and final aligned histological sections, and the Nissl-contrast images and corresponding tissue labels transformed into the 3D volume.

Final template construction
Typically, brain templates are probabilistic estimates constructed from multiple individual datasets, representing a population-average anatomy, the principal benefit of which is to provide a common coordinate space for analysis and remove bias towards any individual's brain anatomy. 9Borrowing from this philosophy, we framed the output of the preceding steps: a 3D volume with approximately aligned coronal sections and global shape, as a single possible representation of the ground truth cerebral volume that captures various idiosyncrasies of the reconstruction pipeline including tissue sectioning frequency and selection, image repair, registration and/or potential misalignment.We employed a data augmentation technique to create multiple instances of the volume reconstruction, created nonlinear deformations by deleting and/or repeating up to 25 randomly selected slices along all 3 image dimensions and forming a population of n=50 unique representations of the original data.Matched augmentations were applied to both the histological reconstruction and accompanying anatomical labels (Figure S17).
For each augmented volume, we resampled the in-plane, coronal resolution to 150 × 150m resolution (slice thickness = 500m) and, for each coronal slice, performed a weighted nonlinear registration to neighbouring sections (symmetric normalisation [SyN] metric=cross-correlation; flow_sigma=3.0,total_sigma=1.0,grad_step=0.25).Adjacent sections (up to 8 neighbours) were weighted based on distance to the source section.After registration, the halfway transform was applied to each section.Section-to-section registration and transformation was performed once in the anteriorposterior direction, before repeating in the posterior-anterior direction.For each volume, slice-to-slice nonlinear registrations were calculated for a total of 3 iterations.
Finally, to create a smooth 3D reconstructed volume, we co-registered all 50 augmented, aligned volumes into a single probabilistic anatomical template with voxel resolution 150 × 150 × 150m using iterative, whole-brain nonlinear registration (SyN metric=cross-correlation; iterations=3; grad_step=0.25)(Figure S17).Transforms were applied to each of the corresponding anatomical label volumes and majority vote used to create a final set of brain tissue and cortical labels (Figure S4).All image registration was performed in Python 3.7 using antspyx (0.2.7). 10

Cortical reconstruction
To reconstruct the fetal cortical surface, we adapted existing protocols for ex vivo [https://freesurfer.net/fswiki/ExVivo] and non-human primate [https://prime-re.github.io/]surface reconstruction with Freesurfer. 11We first created a 'dummy' volume with 1mm voxel dimensions and created a 'white matter' mask by combining tissue labels for all subcortical structures and tissue zones excluding the cortical plate.This volume was tessellated (Freesurfer commands: mri_pretess, mri_tesselate), and the initial surface smoothed (mris_smooth) and inflated to a sphere (mris_inflate, mris_sphere).Topological errors in the initial surface estimates were detected and fixed using manual edits to the brain and white matter masks, before repeating the process.Finally, a pseudo-T2 volume was created using tissue labels, assigning all voxels in the cortical plate intensities expected in grey matter by Freesurfer.This volume was used to generate inner and outer surfaces for the volume (mris_make_surfaces). Surfaces were smoothed for 50 iterations and inspected for topological errors, before conversion to gifti format and rescaling to the original size (Figure 1f, S4).All processing was performed with Freesurfer (7.3.2).S2).S2).

Figure S1 :
Figure S1: Lightbox of reference atlas sections Figure S2: Model performance over hyperparameters Figure S3: Shape reference for affine registration Figure S4.µBrain volume, atlas, cortical surfaces and microarray data Figure S5: All ISH reconstructions.Figure S6: PCA analysis of each brain specimen Figure S7: Correlation between PC1 and age-related changes in expression in each tissue Figure S8: Correlation between age-related gene expression changes across tissue Figure S9: Developmental tissue enrichment of ZRT and non-ZRT genes Figure S10: Gestational age distribution of fetal MRI scans (n=195) included in analysis.Figure S11: Comparison of within-and across-subject estimates of cortical expansion.Figure S12: Tissue and cell expression of ZRTneo genes at 15 PCW.Figure S13: SATB2 ISH across cortical regions.Figure S14: Chromatin accessibility near to ZRT genes in the fetal brain.Figure S15: Expression of OPC markers in the intermediate zone.Figure S16: OLIG1 expression in prenatal bulk tissue mRNA data.Figure S17: image registration pipeline and template generation.Figure S18: Cortical surface area scaling after removing repeated scans.Figure S19: Proportion of vertex outliers in surface area data

Figure S11 :
Figure S1: Lightbox of reference atlas sections Figure S2: Model performance over hyperparameters Figure S3: Shape reference for affine registration Figure S4.µBrain volume, atlas, cortical surfaces and microarray data Figure S5: All ISH reconstructions.Figure S6: PCA analysis of each brain specimen Figure S7: Correlation between PC1 and age-related changes in expression in each tissue Figure S8: Correlation between age-related gene expression changes across tissue Figure S9: Developmental tissue enrichment of ZRT and non-ZRT genes Figure S10: Gestational age distribution of fetal MRI scans (n=195) included in analysis.Figure S11: Comparison of within-and across-subject estimates of cortical expansion.Figure S12: Tissue and cell expression of ZRTneo genes at 15 PCW.Figure S13: SATB2 ISH across cortical regions.Figure S14: Chromatin accessibility near to ZRT genes in the fetal brain.Figure S15: Expression of OPC markers in the intermediate zone.Figure S16: OLIG1 expression in prenatal bulk tissue mRNA data.Figure S17: image registration pipeline and template generation.Figure S18: Cortical surface area scaling after removing repeated scans.Figure S19: Proportion of vertex outliers in surface area data

Figure S12 :
Figure S1: Lightbox of reference atlas sections Figure S2: Model performance over hyperparameters Figure S3: Shape reference for affine registration Figure S4.µBrain volume, atlas, cortical surfaces and microarray data Figure S5: All ISH reconstructions.Figure S6: PCA analysis of each brain specimen Figure S7: Correlation between PC1 and age-related changes in expression in each tissue Figure S8: Correlation between age-related gene expression changes across tissue Figure S9: Developmental tissue enrichment of ZRT and non-ZRT genes Figure S10: Gestational age distribution of fetal MRI scans (n=195) included in analysis.Figure S11: Comparison of within-and across-subject estimates of cortical expansion.Figure S12: Tissue and cell expression of ZRTneo genes at 15 PCW.Figure S13: SATB2 ISH across cortical regions.Figure S14: Chromatin accessibility near to ZRT genes in the fetal brain.Figure S15: Expression of OPC markers in the intermediate zone.Figure S16: OLIG1 expression in prenatal bulk tissue mRNA data.Figure S17: image registration pipeline and template generation.Figure S18: Cortical surface area scaling after removing repeated scans.Figure S19: Proportion of vertex outliers in surface area data Figure S1: Lightbox of reference atlas sections Figure S2: Model performance over hyperparameters Figure S3: Shape reference for affine registration Figure S4.µBrain volume, atlas, cortical surfaces and microarray data Figure S5: All ISH reconstructions.Figure S6: PCA analysis of each brain specimen Figure S7: Correlation between PC1 and age-related changes in expression in each tissue Figure S8: Correlation between age-related gene expression changes across tissue Figure S9: Developmental tissue enrichment of ZRT and non-ZRT genes Figure S10: Gestational age distribution of fetal MRI scans (n=195) included in analysis.Figure S11: Comparison of within-and across-subject estimates of cortical expansion.Figure S12: Tissue and cell expression of ZRTneo genes at 15 PCW.Figure S13: SATB2 ISH across cortical regions.Figure S14: Chromatin accessibility near to ZRT genes in the fetal brain.Figure S15: Expression of OPC markers in the intermediate zone.Figure S16: OLIG1 expression in prenatal bulk tissue mRNA data.Figure S17: image registration pipeline and template generation.Figure S18: Cortical surface area scaling after removing repeated scans.Figure S19: Proportion of vertex outliers in surface area data

Figure S14 :
Figure S1: Lightbox of reference atlas sections Figure S2: Model performance over hyperparameters Figure S3: Shape reference for affine registration Figure S4.µBrain volume, atlas, cortical surfaces and microarray data Figure S5: All ISH reconstructions.Figure S6: PCA analysis of each brain specimen Figure S7: Correlation between PC1 and age-related changes in expression in each tissue Figure S8: Correlation between age-related gene expression changes across tissue Figure S9: Developmental tissue enrichment of ZRT and non-ZRT genes Figure S10: Gestational age distribution of fetal MRI scans (n=195) included in analysis.Figure S11: Comparison of within-and across-subject estimates of cortical expansion.Figure S12: Tissue and cell expression of ZRTneo genes at 15 PCW.Figure S13: SATB2 ISH across cortical regions.Figure S14: Chromatin accessibility near to ZRT genes in the fetal brain.Figure S15: Expression of OPC markers in the intermediate zone.Figure S16: OLIG1 expression in prenatal bulk tissue mRNA data.Figure S17: image registration pipeline and template generation.Figure S18: Cortical surface area scaling after removing repeated scans.Figure S19: Proportion of vertex outliers in surface area data

Figure S15 :
Figure S1: Lightbox of reference atlas sections Figure S2: Model performance over hyperparameters Figure S3: Shape reference for affine registration Figure S4.µBrain volume, atlas, cortical surfaces and microarray data Figure S5: All ISH reconstructions.Figure S6: PCA analysis of each brain specimen Figure S7: Correlation between PC1 and age-related changes in expression in each tissue Figure S8: Correlation between age-related gene expression changes across tissue Figure S9: Developmental tissue enrichment of ZRT and non-ZRT genes Figure S10: Gestational age distribution of fetal MRI scans (n=195) included in analysis.Figure S11: Comparison of within-and across-subject estimates of cortical expansion.Figure S12: Tissue and cell expression of ZRTneo genes at 15 PCW.Figure S13: SATB2 ISH across cortical regions.Figure S14: Chromatin accessibility near to ZRT genes in the fetal brain.Figure S15: Expression of OPC markers in the intermediate zone.Figure S16: OLIG1 expression in prenatal bulk tissue mRNA data.Figure S17: image registration pipeline and template generation.Figure S18: Cortical surface area scaling after removing repeated scans.Figure S19: Proportion of vertex outliers in surface area data

Figure S16 :
Figure S1: Lightbox of reference atlas sections Figure S2: Model performance over hyperparameters Figure S3: Shape reference for affine registration Figure S4.µBrain volume, atlas, cortical surfaces and microarray data Figure S5: All ISH reconstructions.Figure S6: PCA analysis of each brain specimen Figure S7: Correlation between PC1 and age-related changes in expression in each tissue Figure S8: Correlation between age-related gene expression changes across tissue Figure S9: Developmental tissue enrichment of ZRT and non-ZRT genes Figure S10: Gestational age distribution of fetal MRI scans (n=195) included in analysis.Figure S11: Comparison of within-and across-subject estimates of cortical expansion.Figure S12: Tissue and cell expression of ZRTneo genes at 15 PCW.Figure S13: SATB2 ISH across cortical regions.Figure S14: Chromatin accessibility near to ZRT genes in the fetal brain.Figure S15: Expression of OPC markers in the intermediate zone.Figure S16: OLIG1 expression in prenatal bulk tissue mRNA data.Figure S17: image registration pipeline and template generation.Figure S18: Cortical surface area scaling after removing repeated scans.Figure S19: Proportion of vertex outliers in surface area data Figure S1: Lightbox of reference atlas sections Figure S2: Model performance over hyperparameters Figure S3: Shape reference for affine registration Figure S4.µBrain volume, atlas, cortical surfaces and microarray data Figure S5: All ISH reconstructions.Figure S6: PCA analysis of each brain specimen Figure S7: Correlation between PC1 and age-related changes in expression in each tissue Figure S8: Correlation between age-related gene expression changes across tissue Figure S9: Developmental tissue enrichment of ZRT and non-ZRT genes Figure S10: Gestational age distribution of fetal MRI scans (n=195) included in analysis.Figure S11: Comparison of within-and across-subject estimates of cortical expansion.Figure S12: Tissue and cell expression of ZRTneo genes at 15 PCW.Figure S13: SATB2 ISH across cortical regions.Figure S14: Chromatin accessibility near to ZRT genes in the fetal brain.Figure S15: Expression of OPC markers in the intermediate zone.Figure S16: OLIG1 expression in prenatal bulk tissue mRNA data.Figure S17: image registration pipeline and template generation.Figure S18: Cortical surface area scaling after removing repeated scans.Figure S19: Proportion of vertex outliers in surface area data

Figure S18 :
Figure S1: Lightbox of reference atlas sections Figure S2: Model performance over hyperparameters Figure S3: Shape reference for affine registration Figure S4.µBrain volume, atlas, cortical surfaces and microarray data Figure S5: All ISH reconstructions.Figure S6: PCA analysis of each brain specimen Figure S7: Correlation between PC1 and age-related changes in expression in each tissue Figure S8: Correlation between age-related gene expression changes across tissue Figure S9: Developmental tissue enrichment of ZRT and non-ZRT genes Figure S10: Gestational age distribution of fetal MRI scans (n=195) included in analysis.Figure S11: Comparison of within-and across-subject estimates of cortical expansion.Figure S12: Tissue and cell expression of ZRTneo genes at 15 PCW.Figure S13: SATB2 ISH across cortical regions.Figure S14: Chromatin accessibility near to ZRT genes in the fetal brain.Figure S15: Expression of OPC markers in the intermediate zone.Figure S16: OLIG1 expression in prenatal bulk tissue mRNA data.Figure S17: image registration pipeline and template generation.Figure S18: Cortical surface area scaling after removing repeated scans.Figure S19: Proportion of vertex outliers in surface area data Supplemental references

Figure S1 :
Figure S1: Lightbox of reference atlas sections Serial Nissl-stained sections and corresponding anatomical annotations used to construct the Brain atlas.Annotations represent a simplified label set (brain-labels) based on the hierarchical ontology of the BrainSpan reference atlas (TableS2).

Figure S2 :
Figure S2: Model performance over hyperparameters Mean similarity of hue (left) and saturation (middle) of patch predictions compared to ground truth during model training with different parameters (patch resolution in m; loss weight; number of patchGAN layers and input channels).Perceptual similarity (right) was calculated based on the similarity of outputs from a pretrained VGG19 model.Baseline measures of similarity were generated by randomising pixels within each patch (dashed lines).

Figure S3 :
Figure S3: Shape reference for affine registration a.A fetal brain atlas at 22 gestational weeks was used as a shape reference for affine registration.b. manual tissue labels were matched to Brain anatomical labels (c) and used to generate synthetic 'Nissl-contrast' sections (d) to act as a target for registration.
Figure S1.c. lateral and medial cortical surface reconstruction labelled by cortical area (cortex-labels;

Figure S5 :
Figure S5: All ISH reconstructions.Axial slices from partial 3D reconstruction of ISH expression data from n=41 neurodevelopmental genes.

Figure S6 :
Figure S6: PCA analysis of each specimen PCA was performed separately on each specimen's micrparray data.The first two components are shown for each specimen.

Figure S7 :
Figure S7: Correlation between PC1 and age-related changes in expression in each tissueFor each tissue zone, we calculated the correlation between age-related changes (between 15 and 21PCW) and the first principal component.Age related change explained the majority of variance in each tissue zone.

Figure S8 :
Figure S8: Correlation between age-related gene expression changes across tissueCorrelation between age-related changes in each gene across tissue zones.Age-related changes in gene expression were most similar between neighbouring tissue zones.

Figure S9 :
Figure S9: Developmental tissue enrichment of ZRT and non-ZRT genesOutput from Cell-Specific Analysis 12 of ZRT and non-ZRT genes.Significant, cell-specific enrichment of genes are illustrated by hexagon size and colour.

Figure S10 :
Figure S10: Gestational age distribution of fetal MRI scans (n=195) included in analysis.Four fetuses were scanned more than once during gestation

Figure S11 :
Figure S11: Comparison of within-and across-subject estimates of cortical expansion.a.A single fetus was scanned three times during gestation After MRI processing, cortical surfaces were extracted from each scan and vertex area was calculated after resampling native topology onto the dHCP template surface.b.Differences in proportional area, calculated as the area at each vertex divided by total cortical surface area, between the earliest and latest timepoints.Warm colours indicate a preferential increase in regional surface area over time.c.Whole-group, cross-sectional expansion map from Figure 3.

Figure S12 :
Figure S12: Tissue and cell expression of ZRTneo genes at 15 PCW.a. normalised (Z-score) expression profiles for genes correlated with areal scaling in each tissue zone at 15PCW.Negative associations (higher relative expression in hypoallometric regions) shown in blue, positive associations are in red.Lighter colours indicate higher relative expression.b. mid-gestation cell clusters 13 significantly enriched for genes associated with areal scaling in at 15PCW.Territories of three canonical cell types are shown.

Figure S13 :
Figure S13: SATB2 ISH across cortical regions.Example ISH staining of SATB2 + cells across the cortical anlage in five cortical regions.

Figure S14 :
Figure S14: Chromatin accessibility near to ZRT genes in the fetal brain.a. enrichment of different ZRT genesets in all genes located near to OCRs in the prenatal brain.b. enrichment in laminar-specific OCRs.c. number of transcription factor motif annotations located near to ZRT genes.

Figure S15 :
Figure S15: Expression of OPC markers in the intermediate zone.Top: regional expression of oligodendrocyte-lineage markers (S100B, NKX2-2, EGFR) in the intermediate zone of two mid-gestation brains is associated with cortical expansion (scaling coefficient).Bottom: Mean normalised expression of each marker in the IZ calculated within each Brain label and projected onto the 35w surface template.

Figure S17 :
Figure S17: image registration pipeline and template generation.a.clockwise: i) Repaired Nissl sections were first aligned to the mid-section using a graph-based rigid registration algorithm and stacked into a 3D volume.ii) An MRI-based anatomical template was used as a shape reference to guide affine alignment between sections.After transforming each slice of the template to a Nissl-like contrast using the trained pix2pix model (FigureS3), we aligned the reference to the Nissl volume using a 3D affine registration.The aligned volume was resliced and 2D section-to-slice registration to the shape reference performed for each Nissl section.This process was repeated for 5 iterations to form an 'affine volume' (iii).iv)We employed a data augmentation to induce 3D deformations in the affine volume and create a population (n=50) of alternative representations of the cerebral volume, resampled to 150m resolution.For each volume, each section was registered to its neighbours using nonlinear registration.Registrations were performed slice-by-slice, once forward and once backwards along the volume for a total of 3 iterations.v) All 50 volumes were co-registered and averaged to create a final 3D template using symmetric nonlinear registration.b.A two-step registration process was used to align the Brain cortical surface to the dHCP fetal template.Nonlinear surface registrations were performed between spherical representations of each surface using MSM driven by sulcal depth (Stage 1).This alignment was used to initialise a second registration driven by a set of matched cortical regions-of-interest (n=11; Stage 2).Only the occipital (yellow) and motor (blue) regions are shown for simplicity.Each cortical region was passed to the algorithm as a separate metric with alignment jointly optimised over all regions.c. cortical surface alignment after Stage 1 (sulcal depth only) and Stage 2 (sulcal depth +