Mouse Brain Organization Revealed Through Direct Genome-Scale TF Expression Analysis

See allHide authors and affiliations

Science  24 Dec 2004:
Vol. 306, Issue 5705, pp. 2255-2257
DOI: 10.1126/science.1104935


In the developing brain, transcription factors (TFs) direct the formation of a diverse array of neurons and glia. We identifed 1445 putative TFs in the mouse genome. We used in situ hybridization to map the expression of over 1000 of these TFs and TF-coregulator genes in the brains of developing mice. We found that 349 of these genes showed restricted expression patterns that were adequate to describe the anatomical organization of the brain. We provide a comprehensive inventory of murine TFs and their expression patterns in a searchable brain atlas database.

The mammalian nervous system contains thousands of distinct neuronal and glial cell types that are distinguished on the basis of morphology, projection, and marker gene expression (1). Transcription factors (TFs) play a pivotal role in brain development by directing the formation of neurons and glia from uncommitted progenitor cells (2). To determine the extent to which TFs describe the diversity of the mammalian central nervous system (CNS), we visualized the spatial and temporal expressions of TF-encoding genes on a genome-wide scale in the developing mouse CNS.

To identify unique genomic loci that encode putative TFs in the mouse genome, we analyzed and annotated existing public and private databases (36). Candidates were classified as TFs only if their predicted protein sequence included a Protein Families Database (Pfam)–defined TF-DNA binding domain (3). We identified 1445 unique transcriptional units in the mouse genome with putative TF-DNA binding domains. The nonoverlapping genes for each DNA binding family (7, 8) are summarized in Table 1 and tables S1 and S2. Recent protein prediction databases have estimated that there are over 2300 different TF proteins in the mouse genome (6, 9). This higher number is primarily due to the separate counting of each possible protein variant as a unique transcriptional unit, as well as the duplicate counting of genes with multiple DNA binding domains. The largest single class of TF proteins (∼678 members, not including nuclear hormone receptors) was the zinc-finger family. Homeo-box TFs had 227 members, and there were 50 nuclear hormone receptors and 116 basic helix-loop-helix (bHLH) TFs (Table 1 and table S1). The human genome encodes 20,000 to 25,000 genes (10). If a similar number of genes are encoded in the mouse genome, then TF genes make up more than 7% of the total.

Table 1.

Nonredundant numbers of putative TFs in the mouse genome. Columns describe the total number of unique genomic loci encoding predicted DNA binding TFs by domain and the numbers and relative percentages analyzed. The second to last column describes the number of family members available as Genetrap cell lines in the Baygenomics and/or German Gene Trap Consortium libraries. The last column describes the percentage of family members with available enhancer trap cell lines. Genes that encode multiple DNA binding domains are listed and counted in one family for clarity. Nuclear hormone receptors are not included in zinc-finger genes. Transcription cofactors and these non-TF genes we analyzed are also included. Asterisks indicate the cofactor and non-TF gene numbers we analyzed rather than the total number in the genome. HMG, high-mobility group; bZIP, basic helix-loop-helix and leucine zipper proteins; non-TFs, genes that do not encode TFs; nuclear rec, nuclear hormone receptors; ZF, zinc finger; ETS, erythroblast transformation–specific; PHD, plant homeodomain; btb/poz, broad-complex, tramtrack, and bric-a-brac/poxvirus and zinc-finger proteins.

Domain No. of genes No. cloned % cloned Genetrap available % trapped
Homeobox 227 170 74.9 12 5.3
bHLH 116 100 86.2 22 19.0
HMG 58 41 70.7 14 24.1
bZIP 57 41 71.9 16 28.1
Nuclear Rec 50 46 92.0 10 20.0
Forkhead 40 29 72.5 12 30.0
ETS 28 26 92.9 8 28.6
ZF C2H2 490 287 58.6 171 34.9
ZF PHD 60 44 73.3 42 70.0
ZF C2CH 39 18 46.2 11 28.2
ZF btb/poz 28 18 64.3 8 28.6
Other 252 163 64.7 124 49.2
TF Total 1445 983 68.0 450 31.1
Cofactors* 133 104 78.2 48 36.1
Non-TFs* 336 261 77.7 95 28.6
Total genes 1914 1348 70.4 493 25.8

We designed gene-specific polymerase chain reaction (PCR) primer pairs to produce in situ hybridization probes for 1174 TF-encoding genes. This probe set covers 91% of genes that belong to 32 out of the 33 major TF families (table S1). For the remaining (also the largest) gene family, which encodes zinc-finger proteins (divided into 12 subgroups), 71% of the genes were covered by the probe set (table S1). These primers were used to perform PCR on cDNA templates prepared from embryonic day 13.5 (E13.5) and newborn [postnatal day zero (P0)] mouse brains. A small number of additional probes were acquired from embryonic mouse kidney or testis cDNA. Among the 1174 TF-encoding genes screened, 972 (83%) showed positive PCR products. We monitored the recovery of nuclear hormone receptors as a metric for sensitivity. We cloned 46 of the 50 nuclear hormone receptors that are encoded in the mouse genome. Only one nuclear hormone receptor known to be expressed in the brain, the androgen receptor (11), was missed by our procedure. In total, 983 TF-encoding genes were subsequently cloned or acquired (Table 1 and table S1). We also cloned 104 transcription cofactor genes (Table 1 and table S1), yielding a total number of 1087 genes, which are collectively referred to as TF-encoding genes. These cloned in situ plasmids have been deposited at the American Type Culture Collection (ATCC) for open distribution.

We synthesized digoxigenin-labeled probes for the TF-encoding genes. To visualize TF expression at an early developmental stage, we analyzed the expression of 1013 TF-encoding genes using whole-mount in situ hybridization on E10.5 mouse embryos. Of these 1013, at least 142 were clearly expressed in a spatially restricted manner (table S6 and fig. S7). We also performed in situ hybridization for 1034 TF genes on transverse sections through the E13.5 and P0 head and trunk, as well as on sections through the postnatal cerebellum at P7, P15, and P21. Of 1034 genes examined in the E13.5 and/or P0 nervous system, 349 showed spatially restricted expression patterns (table S3). For TFs that belong to non–zinc-finger families, ∼38% of them showed restricted expression patterns in the CNS. However, only ∼10% of zinc-finger genes showed restricted patterns. Collectively, ∼27% of all the TFs exhibited spatially restricted patterns (table S3).

We divided the CNS into seven general areas for annotation: the cortex, striatum (and other basal ganglia), thalamus, hypothalamus, midbrain, hindbrain, and spinal cord. Very few of the 349 TFs with spatially restricted expression patterns were expressed in only one region of the brain (table S4). Nearly all TF-encoding genes expressed in the neonatal brain were also detected in postmitotic neurons at E13.5. Digital images of the entire in situ hybridization set have been deposited in the Mahoney Transcription Factor Atlas (12), as well as in the Jackson Laboratory's Gene Expression Database (13).

The in situ hybridization data show that postmitotic anatomical diversity within the CNS can be described by TF expression. For example, the neocortex is a highly laminated and regionally organized anatomical structure (14). We found that several dozen TF-encoding genes occupy different dorsoventral positions or different laminae of the neocortex (Fig. 1). In the striatum, a major basal ganglia component crucial for movement control is organized in a somatotopic fashion (15). This somatotopic organization is echoed by the gradient expression of many TF-encoding genes in this area (fig. S1). The thalamus and hypothalamus are organized into discrete anatomical and functional nuclei that are marked by the overlapping expression pattern of specific TFs within these regions (figs. S2 and S3). In the retina, a large number of TF genes are expressed with distinct density within the retinal ganglion and amacrine cell layers (figs. S4 and S5), which may correlate with the morphological diversity of these cell types (1). In the postnatal cerebellum, laminar-specific TF expression marks the immature and mature granule cells and Purkinje cells (fig. S6).

Fig. 1.

Diversity of transcription factor expression in the P0 mouse cerebral cortex. Nonradioactive in situ hybridization patterns for 20 representative TFs or TF cofactors on sections through the forebrain of P0 mice are shown. Labels at the bottoms of individual panels indicate Locuslink gene names. (A) Over a dozen TF-encoding genes occupy different dorsoventral positions of the hippocampus (top row) and neocortex (middle and bottom rows). (B) A dozen genes show laminar specific expression in the neocortex. HC, hippocampus; BG, basal ganglia; SC, superior colliculus; TH, thalamus; HY, hypothalamus; DG, hippocampal dentate gyrus; CA1 to 3, hippocampal CA1, CA2, and CA3 regions. Scale bar in (A), 1 mm; all images in this section show the same magnification. Scale bar in (B), 0.2 mm; all images in this section show the same magnification.

The genome-scale whole-mount and section in situ hybridizations also identified over 100 TF genes expressed in spatially restricted patterns within non-neuronal tissues such as the nose, oral cavity, teeth, salivary gland, inner ear bone, mandibular bone, eye muscle, facial muscle, skin, and others (fig. S8 and table S4). Although not the explicit focus of this study, the open-source TF data set generated in this work provides detailed information, enabling comprehensive study of developing cranial facial tissues. The in situ plasmid set provided here can be used to define TF expression patterns in other tissues and at other developmental times in neural tissues.

Our atlas of TF expression provides a visual filter for functional analysis of the TF gene set in brain development. Over 7% of the murine genome may encode TFs. However, our studies suggest that at a given developmental stage, fewer than 27% of these TF-encoding genes are expressed in spatially restricted patterns consistent with a direct role in the formation of specific neural types. The regulatory elements for these spatially restricted TFs might be useful reagents for driving the expression of reporter genes that will mark specific neural cell types, as described recently by Gong et al. (16). The TF expression profile in postmitotic neurons will be useful to study the expression of neural-specific genes encoding ion channels, neurotransmitter receptors, and synaptic proteins, whose molecular control remains largely unknown. On a clinical front, TF mutations have already been shown to underlie certain disorders in speech (17), appetite control (18), breathing patterns (19), and autism (20) in humans. The TF atlas presented here may have practical overtones for understanding additional neurological or behavioral disorders in children and adults.

Supporting Online Material

Materials and Methods

Figs. S1 to S8

Tables S1 to S7


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article