Context dependence of proneural bHLH proteins

Summary

A key point of neural development is the commitment of progenitor cells to a specific neural fate. In all animals studied, proneural proteins — transcription factors of the basic helix-loop-helix (bHLH) family — are central to this process. The function of these factors is strongly influenced by the spatial and temporal context in which they are expressed. It is important to understand the molecular mechanisms by which developmental context interacts with and modifies the intrinsic functions and properties of the proneural proteins. Recent insights have been obtained in Drosophila and vertebrates from analysis of how bHLH proteins interact with other transcription factors to regulate target genes.

Introduction

Neuronal diversity is acquired by various spatial and temporal mechanisms that pattern the events of neural development. The defining event of neurogenesis itself is the switch from uncommitted, cycling progenitor cell to committed neural precursor cell, which has a relatively restricted cell division potential before it or its daughter cells differentiate as neurons or glia. The details of this step differ in different organisms, but a consistent feature is the involvement of basic helix-loop-helix (bHLH) transcription factors. In Drosophila, proneural bHLH factors are initially expressed in ectodermal cells, giving the cells competence to undergo neural commitment, and are then upregulated upon neural commitment, at which point they also trigger Notch-mediated lateral inhibition to suppress the competence of surrounding cells. In vertebrates, the proneural function is also closely associated with committing cycling progenitor cells to a neuronal fate, which involves activation of Notch signalling, inducing cell cycle exit, migration and terminal differentiation [1-4].

Proneural proteins are transcriptional activators that function as heterodimers with E proteins (Daughterless in Drosophila and typically E12 or E47 in vertebrates), which are generally more widely expressed and dimerise with multiple proneural factors. These bHLH heterodimers bind E box motifs (CANNTG) in the regulatory regions of their target genes (Fig. 1A). A number of observations are of major importance with regard to proneural bHLH proteins, their regulation of specific target genes, and how this relates to acquisition of neural diversity. Firstly, proneural proteins generally trigger neurogenesis, but different proneural proteins are required for different neuronal and/or glial cell types, which implies that each bHLH protein regulates both common (shared) target genes for neurogenesis and unique target genes for neuronal subtype characteristics (Fig. 1A). Thus, subtype specificity of bHLH factors underpins some of the cellular diversity in the nervous system [5-7]. Among other things, this has important implications for coaxing the production of specific neuronal subtypes from stem cells [8,9].

Figure 1.

(A) bHLH proteins regulate target gene sets by binding to E box motifs as heterodimers with an E protein (the latter generally not shown in subsequent figures). (B) A bHLH factor may regulate different subtype-specific targets in different places (or at different times) through combinatorial regulation with coexpressed transcription cofactors (CF). Arrows indicate protein interactions leading to synergistic regulation of the target.

Secondly, and somewhat paradoxically, an individual bHLH protein is required for several different neural cell types at different times or locations in development (see Table 1 for recent findings). Therefore, target gene specificity of a bHLH factor must be modulated by developmental context. A striking example is Drosophila Atonal, which is required for precursors of chordotonal proprioceptors in most of the body, olfactory receptors in the antenna, and R8 photoreceptors in the eye [5,10]. In each location, Atonal function cannot be substituted by other proneural bHLH genes [7]. bHLH proteins must therefore have different intrinsic properties that endow them with different target gene specificities, but these properties are modified strongly according to context. This review summarises recent research that begins to illuminate the molecular mechanisms by which context modifies the intrinsic functions and properties of neural bHLH factors.

Table 1.

Some major proneural factor functions in the nervous system

Atonal	Proprioceptors (chordotonal organs) Olfactory receptors (sensilla coeloconica) Eye (R8 photoreceptor) Subset of brain neurons (differentiation)	[5]
Achaete/Scute	External sense organs (tactile/chemosensory bristles) CNS neuroblast subset	[5]
Amos	Olfactory receptors (sensilla basiconica and trichodea) Dbd neuron	[5]
Math1 (Atoh1)	Cerebellum (Granule Cells) Inner ear (sensory hair cells) Mechanoreceptors (Merkel cells)	[52]
	Dorsal spinal cord (commissural neurons)	[28]
Ngn2 (Neurog2)	Ventral neural tube (motorneurons) Forebrain (glutaminergic neurons)
	Dentate gyrus	[53]
	Midbrain dopaminergic neurons	[54]
	Retinal ganglion cells	[6]
Mash1 (Ascl1)	Dorsal neural tube (commissural neurons) Neural crest (autonomic neurons)
	Forebrain (GABAergic neurons)	[55]
	Forebrain (oligodendrocytes)	[56,57]
	Midbrain dopaminergic neurons	[31]

Only some of the more recent papers are shown here. Recent reviews on developmental roles of bHLH factors can be found in [2,10].

Combinatorial regulation – cofactors provide spatial/temporal context

One explanation of context dependence of bHLH function is that certain subtype-specific target genes also require input from other regionally expressed transcription factors (referred to here as cofactors). Thus, context-dependent specificity results from different combinations of bHLH factors and cofactors whose expression overlaps in different places or times [11,12]. A long-standing theme is combinatorial control by bHLH and homeodomain (HD) proteins, and the mouse and Xenopus retinas provide good examples of this [13-15]. For instance, a combination of Mash1 and the HD factor Chx10 are required for bipolar cell fate [14]. Another well-studied example, recently extended by Sugimori et al. [16], comes from the mouse ventral spinal cord [17,18]. Here a combinatorial code of patterning factors (Olig2, Pax6, Nkx2.2) and proneural bHLH factors (Neurogenin (Ngn), Mash1) is postulated to produce neurons, astrocytes and oligodendrocytes in different locations or times [2,16].

In the Drosophila eye, genes of the Retinal Determination Gene Network (RDGN) (particularly sine oculis and the PAX6 homologues, eyeless and twin of eyeless) are prime candidates for providing the context for neurogenesis by Atonal [19]. It is not difficult to envisage that the RDGN not only activates Atonal expression for retinal neurogenesis [20,21], but also modulates Atonal target gene regulation in an eye-specific manner [22]. However, whilst interactions within the RDGN are well characterised, much less is known about how these patterning factors interact with Atonal. Analysis of newly identified eye-specific targets of these RDGN factors may be fruitful [23,24].

Enhancers provide the molecular context for specific target gene regulation

Cellular context is achieved by the co-expression of bHLH factors and cofactors. Combinatorial control is widely expected result from the co-occurrence of binding sites for the bHLH and cofactors in the enhancers of specific target genes, with different cofactors combining with a bHLH factor to regulate different targets (Fig. 1B). Thus, enhancers act in their well-known role as integrators of diverse regulatory inputs. However, few examples are currently known in detail. One is the ‘recruitment’ enhancer of atonal, which is regulated by the combination of Atonal (for autoregulation) and the ETS factor, Pointed [25] (Fig. 3A). The proteins bind cooperatively to adjacent binding sites in the enhancer. This combination of sites also occurs in the enhancer of another Atonal target — the dacapo gene, which encodes a cdk inhibitor [26]. Thus, a subset of Atonal targets only respond in the presence of Pointed, which itself is produced in response to receptor tyrosine kinase (RTK) signalling.

Figure 3.

Examples of specific target gene regulation by bHLH factors. (A) In Drosophila Atonal and Pointed bind cooperatively to adjacent sites in an atonal autoregulatory enhancer. Scute apparently cannot utilise the E box motif; it is not clear whether it can interact with Pnt. (B) Mouse Mash1 and Brn proteins bind cooperatively to adjacent sites in the DeltaM enhancer. Ngn2 apparently cannot utilise the E box motif even though it can interact with Brn proteins. (C) Alternative E protein dimerisation partners may regulate Math1 target gene specificity in the Pontine nucleus, putatively via specific interaction with an unknown cofactor. (D) Phosphorylation of Ngn2 is required for its interaction with NLI, thereby switching its ability to activate a subset of targets with Lim-HD factors. (E) Drosophila proneural factor activity is regulated by Sens acting as a binary switch. Sens can act as DNA-binding repressor or a non-DNA binding coactivator.

An analogous example in mouse is provided by combinatorial regulation of Delta1 by Mash1 and Brn factors, acting co-ordinately by binding cooperatively to adjacent E box and POU protein binding sites in the DeltaM enhancer [27] (Fig. 3B). The two binding sites form a characteristic motif that is also found in other likely downstream targets, including Delta3, Insm1 (Zn finger factor involved in differentiation) and Fbw7 (involved in cell cycle arrest and Notch degradation). In both these examples, the cofactor interaction is required only for a subset of targets, supporting a model in which bHLH factors interact with different cofactors to activate different subprograms of neurogenesis (Fig. 1B) [28].

Another example of combinatorial regulation is the cooperation between Ngn2/NeuroM and the LIM-HD factors, Lhx3 and Isl1, to specify motor neurons in the chick neural tube. Part of their joint function is to activate the Hb9 motor neuron-specifying factor [18]. This is achieved via two E boxes and two LIM-HD sites in the Hb9 motor neuron enhancer (Fig. 3C). Interestingly, in this case the sites are not directly adjacent, and instead a bridging cofactor, NLI, mediates the interaction between the DNA binding factors. This arrangement mirrors a Drosophila interaction between Scute and the GATA factor, Pannier, which is bridged by the NLI homologue, Chip [29].

The structure of target enhancers therefore provides the important molecular context for bHLH protein function. Collocation of binding sites within enhancers of targets determines whether a particular target gene responds to bHLH factors in a particular cellular context.

bHLH selectivity – DNA or protein interactions?

Whilst these are powerful examples of combinatorial control, an important but largely unanswered question is what provides the specificity of co-regulation by different bHLH proteins and different cofactors. Multiple bHLH factors and multiple potential coregulators may be present in the same cells, but targets respond to a specific combination. Is specificity for a particular bHLH factor driven by specificity of protein-protein interactions or DNA binding specificity (Fig. 2)? Unfortunately, in vivo E box occupancy by different bHLH factors (as measured by chromatin immunoprecipitation (ChIP)) may not clearly answer the question since occupancy may depend on protein interactions as much as on DNA binding affinity. In structure-function experiments, functional differences between bHLH proteins have been mapped to non-DNA contacting residues within the bHLH domain, whilst DNA-contacting residues tend to be highly conserved between different bHLH factors [7,10,30-33]. It is attractive to think therefore that these ‘specificity residues’ contact different cofactors thereby providing specificity of target gene regulation (Fig. 2A). However, it is also clear that different proneural bHLH factors do have different E box binding site preferences in vivo if not in vitro. For instance, Scute and Atonal-specific target genes have functionally distinct Scute- and Atonal-specific variant E box motifs [34]. Moreover, these E box motifs are differentially used by Scute and Atonal even in a non-neural cell culture system, suggesting that specialised cofactors may not always be required for bHLH specificity [35]. How can these observations be reconciled? Non-DNA-contacting ‘specificity’ residues in the bHLH domain may well cause conformation effects on DNA-contacting residues, thereby changing DNA interaction properties, perhaps in a manner induced or modified by cofactor interactions.

Figure 2.

Mechanisms for specificity between bHLH factors. (A) Multiple bHLH factors can utilise an E box, but specific cofactor interactions determine that only bHLH factor ‘b’ results in target gene activation. (B) Multiple bHLH factors can interact with the cofactor, but specific DNA-protein interactions determine that only bHLH factor ‘b’ utilises that particular E box motif and results in target gene activation.

An elegant example of specifity in bHLH-cofactor interaction is provided by the DeltaM enhancer mentioned above. Whilst bHLH proteins Mash1 and Ngn2 are both generally capable of synergising with the Brn proteins, it appears that selective E box utilisation determines that only Mash1 synergises at the DeltaM enhancer [27] (Fig. 3B). A different explanation has been suggested for the mouse Hb9 promoter. Whilst NeuroM and Mash1 can both bind an E box in the Hb9 promoter (as measured by ChIP), only NeuroM can synergise with LIM-HD cofactors bound to the same promoter (via the NLI adaptor protein). Although not proven, it is suggested that this selective protein interaction provides the Hb9 promoter’s specificity [18] (Fig. 3C). Conceivably, expression of different adaptor proteins may be an important mechanism for determining target gene specificity in other contexts [36].

A simple model for bHLH function suggests that common (shared) neurogenesis targets are regulated via non-specific E boxes (Fig. 1B). Indirect evidence suggests this is true for the senseless gene, which has a single E box responding both to Scute and Atonal [37], and Prokineticin 2, with an E box that binds both Ngn1 and Mash1 [38]. However, other shared targets respond to different bHLH factors via separate subtype-specific enhancers. This includes Brd (regulated by two enhancers containing Atonal- and Scute-specific E boxes respectively)[34] and mouse Delta1 (two enhancers for Mash1 and Ngn1/2 respectively) [27]. Why should some shared targets be regulated in this way? Possibly it allows such targets to be regulated with different dynamics in different neural subtypes.

Mechanisms providing signalling context

Signalling context (the signals a cell is receiving) is clearly important for modulating bHLH activity in both space and time. The Atonal/Pointed interaction is one obvious way in which this is achieved [25]. Recently, the motor neuron Hb9 paradigm has been extended to show how bHLH specificity may be affected by signalling context via direct post-translational modification. Phosphorylation of Ngn2 has previously been shown to have several roles, including regulation of its stability [39]. Ma et al. [40] have now shown that serine phosphorylation by GSK3 is required for the subtype determination function of Ngn2, but apparently not its neurogenesis function. For the Hb9 motor neuron enhancer, this phosphorylation is necessary for Ngn2’s interaction with the NLI adaptor (Fig. 3C). Thus, a post-translational modification modulates the target gene specificity of a bHLH factor via controlling its capacity for protein-protein interactions. Tyrosine phosphorylation was previously shown to be required for Ngn2 function in migration and dendrite morphology of cortical pyramidal neurons [41], and it is possible that this modification also allows regulation of a specific subset of targets.

Expression of a coactivator protein may also underlie a temporal switch in activity rather than specificity. In Drosophila neurogenesis, proneural factors are thought to regulate only a few targets in neurally competent ectodermal cells (notably concerned with lateral inhibition) but many more upon commitment of the neural precursors. What promotes this switch in activity? A compelling series of papers shows that the Zn finger GPS factor, Senseless, is part of this switch: it can act as both a positive and negative modulator of proneural activity (Fig. 3D) [37,42-44]. There is a strong possibility that some aspects of this interaction are conserved [45,46]. It is not clearly known what switches the activity of Senseless, although a concentration-dependent switch is suggested [37]. Senseless provides temporal context for proneural bHLH activity rather than subtype specificity. However, recent evidence suggests that Senseless interaction may enhance the E box selectivity of different proneural factors, perhaps by strengthening bHLH recognition of different E box motifs [35].

Finally, the role of bHLH dimerisation partners should not be overlooked. In vertebrates there are several E protein partners [47], opening the possibility that choice of E protein may also provide a way of modulating proneural bHLH function. Flora et al. [47] found a specific requirement for Math1/Tcf4 heterodimers (as opposed to heterodimers of Math1 with E47, E12 or HEB) in the generation of the pontine nucleus neuronal structure (derived from the rhombic lip of the hindbrain [48]). They propose that this may reflect a unique ability of Math1/Tcf4 heterodimers to interact with tissue-specific cofactors [47] (Fig. 3), although choice of dimerisation partner might also affect DNA binding site preferences [49].

Conclusions

Although exciting progress is being made, we still have a very patchy understanding of the molecular mechanisms of bHLH specificity in relation to context. One message is that enhancers provide the all-important molecular context that enable different protein interactions to occur at different target genes even within the same cell [11]. A major bottleneck in progressing is a dearth of subtype-specific enhancers to be analysed in structure-function studies. Genome-wide target gene identification, bioinformatic and gene network approaches promise to add much more in the future [50,51].