Fig. 7: Module 20 activity may pattern FEZF2-expressing deep layer subtypes during development.

a, Module 20 was identified in an analysis for modules with dynamic, cell-type-specific activities. Module 20 is enriched in neuronal subtypes (UMAP), increasing in activity in deep layer neurons (box plot). Box plot data represent cells in each developmental stage: Deep Layer, Stage (1) = 6; (2) = 114; (3) = 1,012; (4) = 443; (5) = 13,590; (6) = 2,450; (7) = 3; (9) = 1,835; EN.SERPINI1.NEGR1, Stage (1) = 135; (2) = 20; (3) = 10; (4) = 52; (5) = 1,717; (6) = 1,215; (7) = 12; (9) = 6,485; Subplate, Stage (1) = 10; (2) = 178; (3) = 3,611; (4) = 663; (5) = 1,836; (6) = 1,968; (7) = 2; (9) = 461; EN.PCSK1N, Stage (1) = 33; (2) = 7; (3) = 72; (4) = 110; (5) = 11,745; (6) = 11,829; (7) = 32; (9) = 83; Upper.CTTNBP2, Stage (1) = 2; (2) = 33; (3) = 114; (4) = 272; (5) = 7,670; (6) = 16,279; (7) = 38; (9) = 2. b, In the Jorstad et al.³⁶ adult cortex dataset, module 20 is most active among deep layer subtypes (UMAP, cells with the top 10th percentile of module 20 activity in blue). Module 20-positive adult cells are enriched for deep layer and white matter cells (bar plot; dashed line, expected distribution). The average module 20 activity of layer V/VI subtypes annotated as FEZF2 expressing by Jorstad et al.³⁶ is significantly higher than all other subtypes in that layer (box plots represent average module 20 activity of layer V/VI subtypes; dots show average activity of each individual subtype). Layer V: n = 19 FEZF2⁺ subtypes, 90 FEZF2⁻ subtypes; Layer VI: n = 18 FEZF2⁺ subtypes, 85 FEZF2⁻ subtypes. c, Sparse FEZF2 expression in our developing meta-atlas (left UMAP), peaks in deep layer neurons at developmental stages preceding peak module 20 activity. FEZF2 expression (log₁₀-normalized CPM; green) and module 20 activity (purple) in deep layer neurons are shown overlaid in the right box plot for each developmental stage. Number of deep layer neurons per stage described in a. d, Venn diagram shows that 35 of 74 module 20 genes are putative Fezf2 targets in mouse cortical progenitors⁷⁵, including the TF TSHZ3 (UMAP, expression in developing meta-atlas). Box plot shows TSHZ3 expression (log₁₀-normalized CPM; green) spikes within human deep layer neurons transiently in the middle of peak FEZF2 expression and module 20 activity (purple). Number of deep layer neurons per stage described in a. e, Validation of FEZF2 (red) and TSHZ3 (green) expression throughout the CP via immunofluorescent staining of GW 16 primary human cortical tissues (3×-zoom insets highlight FEZF2⁺/TSHZ3⁺ cells). Lines demarcate cortical layers based on cell density patterns⁸⁵. Box plots show the significantly enriched co-expression of FEZF2 and TSHZ3 in the CP (including upper layers (UL) and deep layers (DL)) compared with ventricular zones (VZ, including inner subventricular zone (iSVZ), outer subventricular zones (oSVZ) and intermediate zone (IZ)), IZ and CP). Stains from two GW 16 slices (dots) were quantified. f, GW 20 primary human cortical sections immunostained for FEZF2 (red), TSHZ3 (green) and deep layer marker CTIP2 (magenta) are consistent with our meta-atlas data. Cells co-expressing CTIP2 and TSHZ3 are highlighted in the 4×-zoom insets and white arrows. Box plot shows FEZF2 and TSHZ3 expression quantification in these two images (dots), confirming TSHZ3 enrichment in deep layers. All significance values were calculated two-sided with Welch’s t-tests. For all immunofluorescent images, scale bar is 100 µm (main panels) and 20 µm (insets). For all box plots, whiskers extend from minimum to maximum values of data; box extends from the 25th to the 75th percentile with line at the median. ChIP-seq, chromatin immunoprecipitation sequencing; DL, deep layer; mo., months; pctl, percentile; UL upper layer.