Terminal selector genes drive terminal neuron differentiation:

We have used three distinct neuronal cell types to genetically dissect the gene regulatory logic of neuronal diversification, the ASE gustatory neurons, the AIY interneurons (each a pair of two neurons) and the entire dopaminergic system, which is composed of 7 anatomically distinct neurons pairs (Wenick and Hobert, Dev. Cell, 2004; Etchberger et al., Genes Dev., 2007; Flames and Hobert, Nature, 2009). Through a combination of cis-regulatory, mutational reporter gene analysis and forward and reverse genetic analysis, we have uncovered strikingly simple concepts in neuronal specification. In all three neuron types examined, we found that their terminal differentiation program is driven by simple cis-regulatory motifs, present in the "nuts-and-bolts" terminal differentiation genes. These simple motifs are activated by what we term "terminal selector genes", postmitotically acting transcription factors that are absolutely required for the terminal differentiation program of individual neuron types. The terminal selector gene concept is detailed in an Essay in the journal PNAS (Hobert, PNAS, 2008).

The most recent and perhaps most intriguing terminal selector gene that we have found acts in a diverse class of neurons, the dopaminergic neurons. The terminal identity of these lineally unrelated neurons is controlled by an ETS-domain transcription factor that activates a small cis-regulatory motif that we identified in the regulatory region of terminal dopaminergic fate genes (Flames and Hobert, Nature, 2009). We have taken this specific example to the vertebrate system and through analyzing an available mouse knockout strain, found that a vertebrate homolog of this ETS factor is required for dopamine neuron development in the vertebrate CNS (Flames and Hobert, Nature, 2009). We found that in worms, this ETS factor can induce ectopic dopaminergic neuron production and we are attempting similar re-programming strategies in primary vertebrate tissue culture cells. We are also testing the role of other terminal selector genes in mice in order to examine whether such regulatory logic broadly applies to the vertebrate system. In parallel to this vertebrate work, we continue to use the C. elegans system to examine if and how terminal selector genes can reprogram other neuron types. The overall approach is to exploit the genetic amenability of C. elegans to screen for mutants that promote or inhibit the ability of terminal selector gene to reprogram neuronal fates.

Apart from actively pursuing the mechanism of action of terminal selector genes - and the conservation of this concept across phylogeny - we aim to reach deeper into the embryo, trying to understand how lineage-based cell specification events are funneled into the onset of terminal selector gene expression with exquisite specificity in individual neuron types. Apart from again using classic forward genetic screens, we have provided strong and broad support for a model put forward by Jim Priess many years ago (Lin and Priess, Cell, 1998), which postulated that the differential distribution of a Wnt-signaling component - the TCF-like factor POP-1 - creates a/p asymmetry upon each cell division in the embryo and thereby generates over time (i.e. over each successive cell division) a highly specific "code" of transcriptional states. We have found that the Wnt/POP-1 system indeed works broadly in the embryonic nervous system to control cellular identities and, specifically, we have shown that this works through a combination of transiently expressed transcription factors that cooperate with the Wnt pathway to directly turn on neuronal terminal selector genes (Bertrand and Hobert, 2009, Dev. Cell).

Model for the initiation of the terminal differentiation program of the AIY interneuron

Further diversification downstream of terminal selector genes and across the left/right axis:

The ASE gustatory neurons enable us to combine our interest in neuronal fate determination with the poorly understood problem of the generation of neuronal diversity across the left/right axis. The two neurons that constitute the ASE class, the left ASE neuron, ASEL, and the right ASE neuron, ASER, are bilaterally symmetric in most regards (cell position, axon/dendrite morphology, synaptic connectivity), yet express a different spectrum of putative chemoreceptors (as initially shown by Yu et al., PNAS 1997; and recently expanded by us; Ortiz et al., Genetics, 2006), thus allowing them to broaden their chemosensory capacities (Pierce-Shimomura et al., Nature, 2001, Chang et al., Nature, 2004; Ortiz et al., in prep.). How is this L/R asymmetric specification event controlled? Using extensive genetic screening approaches, we have identified genes that disrupt the asymmetry of the ASE neurons (lsy genes, for lateral symmetry defective; Chang et al., Genes Dev., 2003, Sarin et al., Genetics, 2007) and reveal that the system is bistable in that each neuron has the potential to exist either in the left or the right state, with lsy genes controlling the state in which each neuron locks its fate. Through determining the identity and epistatic relationship of many of the lsy mutants, we found that the molecular mechanism of this bistability rests on a double-negative, bistable feedback loop composed of several lsy genes which encode transcription factors, and, intriguingly, at least two microRNAs, lsy-6 and mir-273 (Chang et al., Genes Dev., 2003, Johnston and Hobert, Nature, 2003, Chang et al., Nature, 2004, Johnston et al., PNAS, 2005). As a side-note, we have used the ability to monitor miRNA function in this physiological context to investigate various aspects of miRNA target recognition, which led us to dispute currently held tenets about the broadness of miRNA/target interactions (Didiano and Hobert, Nat. Struct. Mol. Biol., 2006, Didiano and Hobert, RNA, 2008).

The activity of the bistable loop provides a net gene regulatory output that collaborates with the terminal selector gene che-1 to regulate left/right asymmetric gene expression programs (Etchberger et al., Development, in press). But where does the asymmetry come from? Through a series of complex genetic and microsurgical manipulations, we have shown that the bistability of the miRNA-mediated regulatory loop is - surprisingly - pre-programmed very early in development and apparently memorized throughout many cell divisions (Poole and Hobert, Curr. Biol. 2006). In order to better understand this early pre-programming and ensuing memory of the predetermined state, we are continuing and scaling up our genetic screens for lsy mutants, using classic forward screening approaches (Sarin et al., Genetics, 2007) as well as performing a genome-wide RNAi screen that we have almost finished. We have identified several transcriptional regulators - including chromatin modifier - in this process (Johnston and Hobert, Development, 2005, Johnston et al., Development, 2006; unpubl. data) and deciphered relevant cis-regulatory mechanisms through a combination of mutant allele recovery and reverse genetics (Etchberger et al., Development, 2009; O'Meara et al., Genetics, 2009). We still have several uncloned lsy mutants to work through and we anticipate that these will provide us with a better understanding of the system.

Regulatory cascade

Carbohydrate biology and the Heparan Sulfate (HS) code:

Differential and combinatorial gene expression of axon guidance factors in individual neuron types have a critical impact on axonal projection programs (e.g., Boulin et al., Curr.Biol., 2006). A few years ago, we exploited the C. elegans system to proof through stringent mutant analysis that specificity in axonal wiring mechanisms is also generated through another mechanism - differential modification patterns of heparan sulfates in the extracellular environment (Bülow and Hobert, Neuron, 2004). Patterns of secondary modifications in HS chains had previously been biochemically defined and their potential to create an enormous amount of molecular complexity in the extracellular matrix had been recognized. However, technical limitations of biochemical approaches prevented the assessment of the physiological relevance of HS molecular complexity. Based on a systematic assessment of the genetic loss of individual HS modifications (2O-sulfation, 6O-sulfation, C5-epimerization) on specific axon guidance choices, we provided experimental in vivo evidence for a model in which secondary modifications in HS define a "code" involved in axonal pattern formation (Bülow and Hobert, Neuron, 2004). In this HS code model - which we detailed in a review in Ann.Rev.Cell Dev.Biol. (Bülow and Hobert, 2006) - specific patterns of HS modifications dictate the activity and specificity of individual axon guidance systems. The combinatorial activity and cross-talk of different sets of guidance factors had previously been proposed to explain how a limited number of axon guidance molecules generate a large amount of architectural complexity. However, it was completely unclear what dictates which guidance cues cooperate in specific cellular contexts. The HS code hypothesis may help to solve this problem; the regional specificity of HS modifications, established by as yet unknown means, may dictate which axon guidance cues combine to pattern axon outgrowth. In our most recent and stringent test for the HS code hypothesis, we have undertaken experiments to ask whether a defined alteration in HS modification patterns can also work instructively to misroute axons. We find this indeed to be the case. Introduction of aberrant 6O-sulfation patterns misroutes axons by ectopically activating the Slit signaling system (Bülow at al., Curr.Biol., 2008).

Axon maintenance:

We had previously discovered that the complex architectural composition of a nervous system is actively maintained post-developmentally by a set of dedicated Ig domain proteins, which serve to counteract various disruptive forces such as mechanical movements (Aurelio et al., Science, 2002). Different sets of maintenance factors appear to be required for different sets of neurons to maintain their position postdevelopmentally. Thus, axons do not appear to require unselective "glues" to maintain their relative position, but sample cell-specific information from their environment. We had initially described a number of secreted, 2-Ig domain-containing, dedicated maintenance factors, the zig genes (Aurelio et al., Science, 2002). Over the past few years, we have identified additional factors involved in axon maintenance in the ventral nerve cord, including a specific splice-form of the FGF receptor (Bülow et al., Neuron, 2004), the transmembrane Ig-domain protein sax-7 (Pocock et al., Mol.Cell.Neurosci., 2008) and the gigantic basement membrane protein dig-1 (Benard et al., Development, 2006). More recently, we have found that various zig genes act in redundant and antagonistic configurations. For example, zig-5 and zig-8 reveal strong mutant phenotypes only in a double mutant configuration and this phenotype is suppressed by either zig-1 alone or a combination of zig-2 and zig-3. Suppression can also be achieved by overexpression of the L1-like Ig protein sax-7. Our current working model is that nervous system architecture is maintained through competing and finely balanced adhesive and anti-adhesive forces; tilting the system in either way will have profound disrupting effects on nervous system structure. In an ongoing collaboration with the Shapiro lab here at Columbia, we are currently investigating the biophysical interactions of these diverse Ig domain proteins in order to get at the mechanistic basis of this phenomenon.

Axon maintenance at the ventral midline in C. elegans. The PVT neuron, schematically shown in panel A and B, expresses a set of zig genes (panel C), which are required to maintain axon position.

For a review, see Hobert and Bülow, 2003.

Plasticity of axonal architecture:

We found that other, transient environmental cues can impart on axonal architecture and also nervous system function. A drop in ambient oxygen concentrations can cause axon mistargeting through misregulation of Ephrin-receptor signaling pathways (Pocock and Hobert, Nature Neurosci., 2008). Moreover, lowering oxygen levels below a certain threshold results in alterations in neurotransmitter expression levels, resulting in specific behavioral defects. We are currently trying to tie these observations of behavioral plasticity together with molecular mechanisms of rapid plasticity in gene expression patterns through examining the role of oxygen levels on the expression of miRNAs, which may have broad roles in rapid, adaptive gene regulatory responses to the environment (Hobert, Science, 2008).

One goal that we have very actively pursued over the past few years is the development of technology that allows to better exploit the specific experimental advantages of the worm model system to study neuronal pattern formation. We have developed an automated screening protocol for the isolation of neuronal cell fate mutants ("Worm sorter" technology) (Doitsidou et al., Nature Methods, 2008), which we are currently actively "milking" to isolate - on saturation scale level - neuronal cell fate mutants (L/R asymmetry mutants in the ASE neurons; dopaminergic cell fate mutants). It is this extensive genetic analysis that we expect to provide major insights into the mechanisms that generate neuronal diversity.

Moreover, we have sought to circumvent another major stumbling block of genetic analysis, that of mutant cloning. We have applied whole genome sequencing to achieve this goal (Sarin et al., Nature Methods, 2008; Shen et al., PloS One, 2008), which allows us to comprehensively reap the fruits of our extensive, automated screening efforts. Since 2009 we have set up our lab owned deep sequencing facility (Illumina Genome Analyzer) to improve the throughput of our mutant identification.

We have continued to develop tools for gene expression analysis, which is of particular importance for our aim to venture into earlier stages of neuronal fate specification during embryonic development, which requires accurate reporter tools. We have developed a robust fosmid recombineering pipeline to generate reliable reporter gene constructs (Tursun et al., PloS One, 2009) and are exploiting it to investigate embryonic transcription factor expression profiles. And lastly, we have developed an improved and updated second generation version of a bioinformatical tool, CisOrtho (Bigelow et al., BMC Bioinformatics, 2004), which we use to identify transcription factor binding sites.