The neocortical infrastructure for language involves region-specific patterns of laminar gene expression

Significance Excitatory connections between different regions of the cerebral cortex arise and innervate in layer-specific patterns. We measured gene expression across layers of the postmortem cortex in core regions of the left hemisphere that support language, using a technique called spatial transcriptomics. Fifty-six genes showed laminar patterns of expression that differed between the frontal and temporal cortex, as well as high expression in excitatory neurons of layers II/III and/or layers V/VI. In large-scale data from the general population, variants in these genes showed a significant association with white matter connectivity between the frontal and temporal language cortex, and with the language-related conditions dyslexia and schizophrenia. Region-specific patterns of laminar gene expression are therefore a feature of the brain’s language network.

Supplementary Figure 1.Schematic of the left-hemisphere regions sampled in this study.Two tissue blocks were taken from the inferior frontal gyrus.Blocks were approximately centred on the yellow spots indicated (block 'gfi1' more anterior/inferior, block 'gfi3' more posterior/superior).Two tissue blocks were also taken from the superior temporal sulcus.Again, blocks were approximately centred on the yellow spots indicated (block 'gts4' more anterior/inferior, block 'gts5' more posterior/superior).The broader coloured regions around the yellow spots represent four areas defined in the SENtence Supramodal Areas AtlaS (SENSAAS) (Labache et al. ( 2019) -see reference in the main manuscript).
Supplementary Figure 3. Spots excluded during spot-level quality control.Spots coloured red were excluded according to the process described in the main text (Methods).The labels of the 48 cortical tissue sections are given as Donor_Block_Section, where the three donors were anw082, anw797 and anw945, the blocks were gfi1 & gf13 from the inferior frontal gyrus and gts4 & gts5 from the superior temporal sulcus, and the tissue sections from each block are called a1 & a2 (the first adjacent pair of sections) and a3 & a4 (second pair of adjacent sections).
Supplementary Figure 4. Spatial distributions of gene expression data-driven clusters of spots across 48 tissue sections.The labels of the 48 cortical tissue sections are given as Donor_Block_Section, where the three donors were anw082, anw797 and anw945, the blocks were gfi1 & gf13 from the inferior frontal gyrus and gts4 & gts5 from the superior temporal sulcus, and the tissue sections from each block are called a1 & a2 (the first adjacent pair of sections) and a3 & a4 (second pair of adjacent sections).
Supplementary Figure 5.The expression levels of layer marker genes in data-driven clusters.Each panel shows the expression of a single marker gene, with the x-axis showing the data-driven clusters and the y-axis showing the normalized gene expression across 48 tissue sections.Clusters are ordered from left-to-right according to their laminar spatial locations from upper to lower.These data support the following correspondence: layerI=cluster9, layerII=cluster10, layerIII=clusters11&2, layerIV=cluster 6, layerV=cluster 3, layerVI=cluster1, with other clusters corresponding to white matter or with sporadic spatial distributions.Supplementary Figure 7. Significant layer*lobe interaction effects for 56 genes that also showed upregulation in Layer II/III excitatory neurons and/or Layer V/VI cortico-cortical projection neurons.Each panel shows the expression of a single gene, with the x-axis showing four pseudo-bulked clusters (see main text) and the y-axis showing the normalized gene expression across 48 tissue sections.Supplementary Figure 8. White matter connectivity between four regions of the core left-hemisphere language network.(A) The four cortical regions are shown in different shades of blue, as defined in the Automated Anatomical Labelling brain atlas (Tzourio-Mazoyer et al. 2002), and applied in the genome-wide association study of Sha et al. (2023) in 30,810 adults (see reference list in the main manuscript).Also shown is an abstract network representation where the four cortical regions are nodes and the structural connections between them are edges.In the present study we were interested in the four frontal-temporal network edges, i.e. pars opercularis -superior temporal cortex; pars triangularis -superior temporal cortex; pars opercularis -middle temporal cortex; and pars triangularis -middle temporal cortex.(B) Visualization of the white matter connections between the four cortical regions in an example individual, with gold representing connections between the pars opercularis and middle temporal cortex, blue representing connections between the pars opercularis and superior temporal cortex, purple representing connections between the pars triangularis and middle temporal cortex, and yellow representing connections between the pars triangularis and superior temporal cortex.Also shown are connections between the pars opercularis and pars triangularis (red), and connections between the middle temporal cortex and superior temporal cortex (green), but these within-lobe connections were not considered in the present study.
Figure reproduced from Sha et al, (2023) under an open access Creative Commons Attribution License 4.0 (CC BY) (see reference list in the main manuscript).

Table 3 :
Supplementary Table2: Gene Ontology enrichment analysis of 72 genes that showed significant layer-by-lobe interaction effects.Sets with adjusted enrichment p values <0.01 are shown.Gene Ontology enrichment analysis of 56 genes upregulated in LII/III and/or LV/VI excitatory neurons that showed significant layer-by-lobe interaction effects.Gene ontology sets with FDR-adjusted enrichment p values <0.01 are shown.

Table 4 .
Association results for 56 genes in relation to white matter connectivity in 30,810 adults Supplementary Table 5. Association results for 56 genes in relation to word reading ability in 33,959 individuals, and dyslexia in 51,800 adults who reported having a diagnosis versus 1,087,070 controls