The DNA of mammalian cells is compacted several thousandfold by incorporation into chromatin, yet DNA must remain accessible to signaling pathways initiated at the cell membrane, as well as to mechanisms that replicate and transcribe the genome. Recent studies have revealed that the vertebrate genome contains several hundred genes involved in chromatin regulation, suggesting that chromatin regulation is more biologically specific than previously believed. Moreover, several years ago we found that one specific family of chromatin regulatory complexes (mSWI/SNF or BAF) in mammals is combinatorially assembled from the products of 23 genes encoding its 12 subunits producing several hundred possible complexes. More recent work by other groups suggests that combinatorial assembly might be a general strategy for producing biologic specificity of chromatin regulatory complexes in vertebrates. Using genetic approaches in mice, we have found that alternative assemblies of chromatin regulatory complexes have specific biologic meanings in the development of the mammalian nervous system and also in the transition from pluripotency to a multipotent neural stem cell.
One of the most distinctive steps in the development of the nervous system occurs at mitotic exit when cells lose the potential to become many types of neurons and adapt fates, morphologies, and functions that will persist for centuries (at least in the case of certain turtles). In this sense, a postmitotic neuron has perhaps the most stable epigenetic state of any cell type. Our early genetic studies indicated that this transition requires the subunits of SWI/SNF-like BAF complexes. These complexes derive energy from the alternative ATPases Brg and Brm, which are paired with a second ATPase, actin. They resemble SWI/SNF in yeast but have lost, gained, and shuffled subunits and seem to have adopted new mechanisms in vertebrates, allowing the complexes to function combinatorially and to repress transcription at a distance.
To understand the diversity and specificity of action of these chromatin-remodeling complexes, we developed rapid, efficient ways to purify and to characterize the subunits of these complexes from small, pure populations of developing cells. Our work demonstrated that each position in the complex can be occupied by the product of only one member of the family of genes encoding the subunit and that each subunit is essentially nonexchangeable in vitro. These studies suggest that there might be several hundred combinatorially assembled complexes and that they might produce biologic specificity, just as letters in a 12-letter word produce meaning by combinatorics. What was needed was some test of this combinatorial model for production of specificity.
Proteomic analysis of the complexes in the developing nervous system led to a surprising discovery: Nature had deliberately changed the spelling of the chromatin-remodeling complex. The complexes in neural progenitors or stem cells (npBAF complexes) are made of a specific composition, while complexes in neurons after mitotic exit have a second composition (nBAF complexes). The transition between these two complexes is accomplished by the exit of two subunits, BAF53a and BAF45a, and the insertion of two new homologous subunits, BAF53b and BAF45b. This occurs in the context of a specific subunit composition characteristic of neurons at all stages of their lives (Figure 1). Further genetic studies revealed that the progenitor complex is essential for self-renewal of neural stem cells, the postmitotic complex is necessary for dendritic development, and failure to switch complexes is lethal to the mice.
Andrew Yoo and Brett Staahl in our laboratory became curious about the mechanisms that control this essential switch in neural development and screened the 180-kb BAF53a gene for sequences mediating switching. Surprisingly, essentially all regulation was derived from the 3' UTR (untranslated region). The regulatory sequences within the entire 180-kb gene mapped to the binding sites for miR-124 and miR-9*, which are expressed selectively in postmitotic neurons. This suggested a simple regulatory circuit in which miR-9* and miR124 are activated, leading to the repression of BAF53a and mitotic exit. We tested this hypothesis by making strains of mice prematurely expressing both miR-9* and miR-124 along with the 180-kb BAF53a gene sequence, either lacking or containing the miR-9*- and miR-124-binding sites. Premature expression of miR-9* and miR-124 led to inhibition of mitosis, an effect that was overcome by coexpression of the BAF53a gene lacking the microRNA-binding sites. Thus it appears that the microRNAs control cell cycle exit of neural progenitors by repressing BAF53a. How then are the mircoRNAs controlled?
Prior work by Gale Mandel (HHMI, Oregon Health & Science University) had shown that NRS/REST repressed miR-124 and miR-9*. REST is a transcription factor that represses neural genes outside the nervous system. When Yoo and Staahl expressed NRSF/REST in postmitotic neurons, they found activation of BAF53a, repression of BAF53b, and a failure of normal dendritic development. Prior work in our laboratory by Jiang Wu and Julie Lessard had shown that BAF53b is essential for normal elaboration of dendritic trees, one of the most distinctive features of neurons.
These studies indicate that a triple-negative genetic circuit regulates mitotic exit in neural development and initiation of dendritic morphogenesis (Figure 2). However, for several reasons we believe that this circuit is not complete. First, it does not intersect with other important regulatory mechanisms mediated by Notch, Hedgehog, FGF4, or neurogenic genes, all of which have major roles in the development of the vertebrate nervous system. Second, the switch in protein expression happens too quickly to be mediated solely by mRNA degradation and may also involve BAF53a protein degradation. Finally, this circuit does not explain the rapid switch from BAF45a/d-based complexes to BAF45b/c-based complexes that also coincides with mitotic exit. We hope to address these questions in the future.
The Switch from Pluripotency to Multipotency Requires Subunit Exchange
Leno Ho, a graduate student in our laboratory, set out to determine whether SWI/SNF-like BAF complexes found in pluripotent cells, such as embryonic stem (ES) cells (ESCs) and cells of the inner cell mass, are distinctive. Previous work in collaboration with Scott Bultman and Terry Magnuson had shown that Brg, but not the highly related ATPase Brm, is essential for formation of ESCs and appears to have a genetically dominant role. Ho found that ESCs have a distinctive family of complexes that we had not seen in other cell types. To determine if this family of complexes, which we call esBAF, is required for pluripotency, Ho "misspelled" the esBAF chromatin-remodeling word by expressing subunits not found in ESCs. This interfered with the ability of ESCs to proliferate in a state that could give rise to the different germ layers. This raised the question of how esBAF is required for the pluripotent state.
Genome-wide Studies Reveal that esBAF Complexes Are Programmatic Regulators of the Pluripotent State
We and others found that deletion of Brg led to a loss of expression of the pluripotency genes Oct4, Sox2, and Nanog and reduced proliferation. These findings led us to collaborate with Keji Zhao, a former fellow in our lab who had developed the CHIP-seq and CHIP-sage techniques after establishing his own laboratory. In the first genome-wide analysis of an ATP-dependent remodeling complex, we found that the 300,000 esBAF complexes per ESC bind to about 10,000 sites per genome, each about 6 kb in length. We found remarkable co-occupancy of esBAF with Oct4, Sox2, and Nanog over the ESC genome (Table 1). In addition, esBAF bound with STAT3 and SMAD1, which transmit signals induced by LIF and BMP that are essential to maintain the pluripotent state. Interestingly, esBAF appears to avoid cobinding with polycomb complexes. Coanalysis with mRNA microarray studies on ESC in which the Brg gene was conditionally deleted revealed that Brg is remarkably programmatic in controlling genes that support the pluripotent state, dispelling the old notion that ATP-dependent remodeling might function randomly on all genes. Instead our studies indicate that the esBAF complexes provide robustness and stability to the pluripotent state by (1) enabling the core transcriptional circuitry of pluripotency, (2) facilitating signaling by LIF and BMP, which prevent differentiation of ESC, and (3) interacting directly with pluripotency factors.
Calcineurin-NFAT Signaling in the Diversification of Neural Crest
Many growth factors, cytokines, and ion channels activate a signaling pathway employing Ca2+, calcineurin, and NFATc proteins that activates genes involved in cell-cell interactions, development, morphogenesis, and stress responses (Figure 3). We defined this pathway in collaboration with Stuart Schreiber (HHMI, Harvard University) and demonstrated that it is the target of the immunosuppressive drugs, cyclosporine A and FK506. The NFATc family of genes first appeared about 500 million years ago near the emergence of vertebrates. Genetic studies in our lab and others have led to the realization that this pathway is essential for many aspects of vertebrate development. These observations led us to speculate that the creation of the NFATc family of genes by recombination of a Ca2+-sensing domain with a DNA-binding domain about 500 million years ago gave the raw material needed for aspects of vertebrate organogenesis. This hypothesis predicts that NFAT signaling should be essential for vertebrate-specific structures in present-day vertebrates.
Work in our lab and others has shown that calcineurin/NFAT signaling is essential for the development and function of the recombinational immune system, heart valve morphogenesis, bone development, lung development, vessel assembly and patterning, epithelial stem cell maintenance, pancreatic development, and vertebrate-specific aspects of neural development. A classic vertebrate-specific structure is the neural crest, which gives rise to myelinating Schwann cells. Myelination increases conduction speed by 10- to 100-fold and may have allowed the emergence of the large vertebrate body plan. Using conditional mutant mice, Shih-Chu Kao in our lab found that calcineurin-NFAT signaling is essential for Schwann cell differentiation and myelin formation. Further studies revealed that neuregulin binding to ErbB receptors leads to Ca2+ entry, the activation of the phosphatase calcineurin, and nuclear localization of NFATc3 and -c4 and their assembly with Sox10 into NFAT complexes (Figure 4). This pathway activates genes at the top of a transcriptional hierarchy, such as Krox20 (Egr2), leading to myelination, and fills an essential gap in our understanding of neural crest diversification. The results of these studies make it possible to imagine a systematic approach to many aspects of vertebrate development, using conditional null mice for NFATc and calcineurin genes.