|Home | About | Journals | Submit | Contact Us | Français|
Cholinergic motor neurons are defined by the co-expression of a battery of genes which encode proteins that act sequentially to synthesize, package and degrade acetylcholine and reuptake its breakdown product, choline. How expression of these critical motor neuron identity determinants is controlled and coordinated is not understood. We show here that in the nematode Caenorhabditis elegans all members of the cholinergic gene battery, as well as many other markers of terminal motor neuron fate, are co-regulated by a shared cis-regulatory signature and a common trans-acting factor, the phylogenetically conserved COE (Collier/Olf/EBF)-type transcription factor UNC-3. UNC-3 initiates and maintains expression of cholinergic fate markers and is sufficient to induce cholinergic fate in other neuron types. UNC-3 furthermore operates in negative feedforward loops to induce the expression of transcription factors that repress individual, UNC-3-induced terminal fate markers resulting in diversification of motor neuron differentiation programs in specific motor neuron subtypes. A chordate ortholog of UNC-3, Ciona intestinalis COE, is also both required and sufficient for inducing a cholinergic fate. Thus, UNC-3 is a terminal selector for cholinergic motor neuron differentiation whose function is conserved across phylogeny.
A core identity feature of each specific neuronal type is its neurotransmitter phenotype which is defined by the co-expression of genes encoding proteins that synthesize, package and re-uptake a specific neurotransmitter. In the case of the neurotransmitter acetylcholine (ACh), the enzyme choline acetyltransferase (ChAT) synthesizes ACh from the ubiquitous precursor choline, the vesicular ACh transporter VAChT takes up ACh into synaptic vesicles, acetylcholinesterase (AChE) breaks down ACh into choline after release in the synaptic cleft and a choline transporter (ChT) reuptakes choline to shuffle it back in the biosynthetic/release cycle (Fig. 1a)1. Moreover, cholinergic synaptic termini often contain autoreceptors to modulate synaptic signaling properties2. How the expression of these core features of cholinergic neuron identity is coordinated is not presently known. Elucidating the mechanisms of coordinated expression of neurotransmitter pathway genes also represents one key approach to understand how neuronal identity is controlled. The much anticipated goal of generating specific neuron types in vitro for basic and therapeutic studies, is a further motivation to understand how these genes are regulated.
Motor neurons in the spinal and ventral cords of vertebrates and many invertebrates generally use ACh to communicate with their target muscles. The nematode C. elegans contains 8 different classes of ventral nerve cord (VNC) motor neurons (MNs) that control locomotion of the animal3,4. Two of these 8 classes are GABAergic, the other 6 classes are cholinergic, namely the dorsal A (DA) and ventral A (VA)-class neurons (hereafter referred to as A-type MNs), the dorsal B (DB) and ventral B (VB)-class neurons (hereafter referred to as B-type MNs), the AS neurons and the ventral C (VC)-class neurons (Fig. 1b). Each class contains several members (e.g. DA1, DA2, DA3 etc.) amounting to a total of 56 ventral cord motor neurons 3,4. The cholinergic nature of these neurons is defined by the expression of unc-17/VAChT, cha-1/ChaT (both genes are organized into one common genetic locus in many animal species5), ace-2/AChE (this study), cho-1/ChT 6 and the putative ACh autoreceptor subunits acr-2, acr-5 and acr-14 7–9.
The phylogenetically conserved transcription factor unc-3, the sole C. elegans homolog of the COE (Collier/Olf/EBF) family of transcription factors, is expressed in the A-, B- and AS-type MNs, but not the VC MNs 10,11 (this study). unc-3 mutant animals were previously shown to display defects in the differentiation of the A- and B-type neurons (AS type neurons were not examined), as assessed by aberrant morphology and aberrant expression of three cell surface proteins (the ACh receptor acr-5, the DEG-type ion channel del-1 and the netrin receptor unc-5) and one transcription factor (the homeobox gene unc-4) 10,11. If, how and to what extent unc-3 affects the individual members of the cholinergic gene battery, i.e. the cholinergic phenotype of these neurons is not known. We show here that the cholinergic gene battery, as well as many other terminally differentiated features of the A-, B- and AS-type MNs are co-regulated through a common regulatory strategy, employing a shared cis-regulatory signature and trans-acting factor, UNC-3. This shared regulatory strategy defines the logic of coordinated regulation of gene expression of terminal cholinergic features. Evidence is presented for an evolutionarily conserved role of UNC-3 in the specification of cholinergic motor neurons in a chordate, the tadpole of Ciona intestinalis.
We generated a number of fluorescent reporter constructs that monitor expression of the cholinergic gene battery shown in Figure 1a. These reporters either recapitulated known expression patterns in cholinergic ventral cord motor neurons (i.e. unc-17, cho-1)6,12 or demonstrated predicted expression in cholinergic motor neurons (i.e. ace-2 )13 (Fig. 1c, Supplementary Table 1). Apart from the cholinergic gene battery, we examined the expression of 27 additional reporters of putative terminal differentiation genes for which expression in either all or individual types of ventral cord MNs was previously reported, including putative ACh metabotropic and ionotropic autoreceptors (acr-2, acr-5, acr-14, acr-15, acr-16, gar-2), other ionotropic and metabotropic neurotransmitter receptors (GABA receptor gbb-1, dopamine receptor dop-1), ion channels (DEG/ENaC channels unc-8 and del-1, TRP channel trp-1, voltage-gated Ca+ channel nca-1), gap junction proteins (inx-12), signaling proteins (TGFβ-like tig-2 and dbl-1), neuropeptides (nlp-21), axon pathfinding factors (max-1, unc-129, unc-40) and others (see Methods section for reporter strains used)(Table 1, Fig. 1c, Fig. 2a). We crossed these reporter transgenes into mutants that lack the sole homolog of the COE-type transcription factor unc-3 (mutant allele e151) and find that 87% (26/30) tested terminal differentiation genes of the A-, B- and/or AS-type motor neurons fail to be properly expressed in unc-3 mutant animals (Table 1, Fig. 1c, Fig. 2a, Supplementary Table 1, Supplementary Fig. 1). Three of the four genes whose expression is not affected by unc-3 are broadly expressed throughout the nervous system and some also outside the nervous system. Three known pan-neuronally expressed genes (rab-3, unc-119, rgef-1) also show normal expression patterns in the VNC of unc-3 mutants (Fig. 2b, Supplementary Table 1). We conclude that unc-3 controls the A-, B-, AS-type MN expression of a large number of genes that display a restricted expression in the nervous system, but does not affect expression of generic components of neuronal fate.
Molecular markers for the other major classes of ventral cord MNs, GABAergic neurons, are not ectopically expressed in the A-, B-, and AS-type MNs in unc-3 mutants 10. Three unidentified cells in the ventral cord of unc-3 mutants ectopically express VC motor neuron fate markers 10, but since unc-3 affects the fate of a total of 50 cholinergic MNs (A-, B- and AS-type), the unc-3 phenotype can not be generally explained by a general switch to VC motor neuron fate. Moreover, VC motor neurons are also cholinergic; therefore, if A-, B-, and/or AS-type MNs had simply switched to a VC MN fate, expression of the cholinergic gene battery should be unaffected, which is not the case (Fig. 1c, Supplementary Table 1). We also tested the possibility that cholinergic MNs in unc-3 mutants convert into a more progenitor-like state characterized by the transient expression of the NeuroD homolog cnd-1 8. Examining cnd-1 expression in unc-3 mutants, we find this not to be the case (Supplementary Fig. 2). Therefore, we conclude that in unc-3 mutants, several major classes of ventral cord cholinergic neurons fail to execute the normal, identity-defining differentiation program, yet retain generic neuronal identity.
To ask whether unc-3 is not only required but also sufficient to induce cholinergic motor neuron fate, we generated transgenic animals that misexpress unc-3 in two unrelated, non-cholinergic neuron types – the glutamatergic sensory neurons AWC and ASE, using regulatory sequences from the ceh-36 homeobox gene (an AWC and ASE fate regulator)14. In these transgenic animals, we observe ectopic expression of the cholinergic markers acr-2 and unc-17 (Fig. 3a–b).
Misexpression of unc-3 in the D-type GABAergic motor neurons, using regulatory sequences from the unc-30 gene (and inducer of GABA fate)15 also induces ectopic expression of cholinergic fate (Supplementary Fig. 3). Broader misexpression of unc-3 using an inducible heat-shock promoter results in very broad expression of cholinergic fate markers throughout the animals if unc-3 is heat shock-induced at embryonic stages (Fig. 3c–d, Supplementary Fig. 4). Taken together, these findings demonstrate that unc-3 is not only required but also sufficient to induce cholinergic fate in many spatiotemporal contexts.
Mechanistically, the effect of unc-3 on cholinergic motor neuron differentiation could be explained either by unc-3 controlling the regulation of a number of other regulatory factors that then control terminal gene expression or by unc-3 directly controlling expression of the cholinergic terminal gene battery. Such a direct, co-regulatory strategy would represent an ideal and simple way to ensure coordinated expression of the cholinergic synthesis/transport pathway genes. We tested this possibility by systematically examining members of the cholinergic gene battery, as well as all other tested UNC-3-dependent A-, B-, and AS-type motor neuron markers for the presence of putative UNC-3 binding sites. As a guide, we used the binding sites determined for vertebrate orthologs of UNC-3, which in the context of C. elegans olfactory neurons, have also been shown to be functional UNC-3 binding sites 16–19 (Fig. 4a, Supplementary Fig. 5). We refer to these cis-regulatory binding sites here as “COE motifs”. We found at least one copy of this motif in 20 of the 26 unc-3-dependent, VNC-expressed genes (Table 1); most of these motifs are phylogenetically conserved in different nematode species (Fig. 4, Supplementary Table 2, Supplementary Fig. 5). Focusing on 8 of the 26 genes, we first tested whether small regulatory regions that contain these motifs are sufficient to drive expression of a reporter gene in the A-, B-, and AS-type motor neurons and found this to be the case (Fig. 4). We mutated these binding sites in the respective reporter genes and generated transgenic animals expressing these reporters. Compared to the wild-type control, all mutated reporters failed to show correct expression in the A-, B-, and AS-type MNs (Fig. 4, Supplementary Table 2). Expression in other neuron types is unaffected. We note that several of the unc-3 and COE-motif-controlled genes contain more than one copy of the COE motif. Based on the analysis of other simple cis-regulatory motifs that control neuronal gene expression in C. elegans 20, we hypothesize that such multiplicity of cis-regulatory motifs ensures robustness in gene expression.
We next took a broader genomic view of the occurrence of COE motifs and asked how predictive their occurrence is for expression in cholinergic motor neurons. We first examined a set of genes identified through cell-type specific transcriptome profiling to be expressed in A-type cholinergic ventral cord motor neurons21. We found that 27/32 (84.4%) of these genes contain COE motifs (Supplementary Table 3). By comparison, only 5/33 (15.2%) genes whose expression cannot be detected in the A-and B-type motor neurons contain a COE motif (Supplementary Table 3).
Second, we interrogated the entire C. elegans genome for the presence of COE motifs using the CisOrtho platform which we previously developed to search genome sequences for matches to a position weight matrix of transcription factor binding sites22. We selected 8 genes with unknown expression pattern that contain phylogenetically conserved COE motifs and that encode putative neuronal function genes, focusing on putative ion channels. We generated gfp reporters for these genes and found that seven of eight genes are indeed expressed in cholinergic ventral cord motor neurons (Supplementary Fig. 6). These seven genes encode several potassium channels (twk-7, twk-13, twk-40, twk-43), a ligand-gated ion channel (lgc-55), an acetylcholine receptor subunit (acr-21) and the C. elegans ortholog of the human transmembrane channel-like gene 1 (TMC1), mutations in which cause deafness in mice and humans23. We tested whether the expression of five of these seven genes depend on unc-3 by crossing reporter genes into an unc-3 mutant background. We found that four of the five genes indeed require unc-3 for VNC MN expression (twk-13, twk-40, twk-43 and ortholog of TMC1) (Supplementary Fig. 6). With these four unc-3-dependent, COE-motif containing genes, we have identified a total of 30 direct target genes of unc-3 in various classes of VNC MNs, all of which encode for terminal differentiation (“effector”) genes (below we describe three more direct targets of unc-3 that code for transcription factors)(Table 1). We conclude that terminal cholinergic fate is controlled by a co-regulatory strategy through the UNC-3 transcription factor and its cognate cis-regulatory target site, the COE motif. The presence of phylogenetically conserved COE motifs appears to be a predictor for expression of genes in cholinergic motor neurons.
Terminal differentiation features of a neuron, such as neurotransmitter gene batteries, need not only be induced but also maintained throughout the life of a neuron. The ability of unc-3 to directly induce cholinergic fate suggests the possibility that unc-3 may also maintain cholinergic fate. A prerequisite for such an activity is sustained expression of unc-3 in cholinergic MNs throughout the life of the animal. While expression of unc-3 has previously been found to coincide with the generation of A- and B-type motor neurons11, its expression in these neurons has not been reported during postdevelopmental stages. To monitor unc-3 expression we generated a fosmid-based reporter that spans about 40 kb of genomic sequences, containing all intergenic regions surrounding the unc-3 locus plus several genes upstream and downstream of unc-3. Consistent with previous studies, we find that the unc-3 fosmid reporter is expressed in the A- and B-type motor neurons (both the dorsal and ventral A and B-type motor neurons DA, VA and DB, VB), but not in VC (Supplementary Fig. 7) and GABAergic D-type neurons (Fig. 5a). We also detect expression in the postembryonic AS cholinergic ventral cord motor neurons, consistent with the effect of unc-3 on expression of the cholinergic gene battery in the AS neurons. Importantly, we find that unc-3 expression persists in all these VNC motor neurons throughout the life of the animal (Fig. 5a).
To assess the physiological relevance of persistent unc-3 expression, we sought to remove unc-3 gene activity postembryonically. Since temperature-sensitive alleles for unc-3 are not available and since unc-3 gene activity could not be removed by RNAi (data not shown), we generated transgenic animals that lack endogenous unc-3 gene activity but express heat-shock inducible unc-3 from an extrachromosomal array under control of the hsp-16 promoter24. We find that the uncoordinated phenotype (data not shown) and loss of acr-2::gfp or unc-17::gfp expression of unc-3(e151); Ex[hs::unc-3] animals can be rescued through heat-shock induction of unc-3 expression in mid-larval stages (Fig. 5b, Supplementary Fig. 8). This indicates that embryonically generated MNs remain in a state that is responsive to unc-3, rather than, for example, irreversibly converting into another fate or being irreversibly damaged. If unc-3 activity is supplied in midlarval stages, but then removed through removal of the inductive heat shock stimulus, a progressive loss of acr-2::gfp or unc-17::gfp expression is observed during adulthood (Fig. 5b, Supplementary Fig. 8). This observation indicates that unc-3 gene activity is continuously required to maintain the gene expression program of cholinergic motor neurons.
Previous work has shown that a set of three phylogenetically conserved homeodomain transcription factors, unc-4 (Uncx4 in vertebrates), vab-7 (Evx in vertebrates) and ceh-12 (HB9 in vertebrates) act as transcriptional repressors to make DA become different from DB and VA different from VB fate 9,25,26. For example, the B-type specific terminal differentiation marker acr-5, coding for an acetylcholine receptor subunit, is repressed by unc-4 in the DA neurons but not in the DB neurons 9. The DB-specific vab-7 homeobox gene in turn represses unc-4 in DB, thereby allowing the acr-5 to be expressed in DB-type MNs 25. These sequential repressive regulatory events beg the previously unexplored question of how acr-5 expression is activated. We have shown above that positive induction of acr-5 in DB-type MNs is achieved by direct activation through unc-3. Interestingly, we find that the unc-4 and vab-7 homeobox genes are also both likely direct targets of unc-3. Both genes contain phylogenetically conserved COE motifs in their regulatory regions and both depend on the presence of unc-3 for their expression in the DA (unc-4) and DB (vab-7)(Fig. 6a–b). Similarly, the VB-expressed HB9 ortholog ceh-12, which is repressed in VA by unc-4 to allow for the execution of VA-specific fate 26, is also a likely direct target of unc-3 as it also contains COE motifs in its regulatory region and depends on unc-3 for correct expression in VB (Fig. 6a–b). Taken together, aside from controlling terminal features of all cholinergic MNs unc-3 also activates MN subtype-specific transcriptional repressors that act in a negative feedforward loop (a motif commonly found in bacterial gene regulatory networks and also called “incoherent feedforward loop” 27) to repress subtype-specific terminal differentiation genes, thereby generating diversity in distinct cholinergic MN types (Supplementary Fig. 13c).
unc-3 affects the cholinergic gene battery in the cholinergic A-, B-, and AS-type motor neurons. In addition to those neurons, C. elegans contains more than a dozen very distinct types of cholinergic neurons, including sensory and interneurons 1. For example, the AIY interneuron class, which is located in the ventral head ganglion and processes a variety of distinct sensory modalities, is also cholinergic, as assessed by expression of the unc-17/cha-1 locus 28. As expected, we find that AIY also expresses the choline transporter cho-1 (Supplementary Fig. 9), but it does not express unc-3 (as assessed with an unc-3 fosmid-based reporter gene fusion; data not shown). How is the cholinergic battery then induced in AIY? We have previously shown that the unc-17/cha-1 locus contains a binding site (called “AIY motif”) for the TTX-3/CEH-10 homeodomain heterodimer, which is required for unc-17/cha-1 expression 29. We find that cho-1 expression is also lost in ttx-3 mutants (Supplementary Fig. 9). Besides containing a COE motif, the cho-1 regulatory region also contains a phylogenetically conserved AIY motif (Fig. 4c). A reporter gene lacking the COE motif looses expression in A-, B-, and AS-type cholinergic neurons, but not in AIY, while a reporter that lacks the AIY motif, but contains the COE motif shows expression in A-, B-, and AS-type motor neurons, but not in AIY (Fig. 4c). These findings not only demonstrate co-regulation of cholinergic pathway genes in other cholinergic neuron types, but also suggest that the regulation of cholinergic pathway genes is conferred by a modular array of neuron-type specific cis-regulatory elements that are activated by neuron-type specific regulatory factors (Supplementary Fig. 13b).
Vertebrate genomes contain multiple orthologs of UNC-3. At least three of the four mouse UNC-3 orthologs are expressed in cholinergic motor neurons of the spinal cord 30,31, yet their role in determining the cholinergic phenotype has not been assessed. The chordate Ciona intestinalis contains cholinergic motor neurons with several organizational and developmental similarities to those of vertebrates; their axons extend along the tail of the larvae and, like motor neurons in vertebrates and C. elegans they control locomotion32. Unlike vertebrates, Ciona only contains a single EBF ortholog, called COE33. We examined its expression pattern and found it to be also expressed in cholinergic motor neurons, as assessed by a reporter that visualizes expression of the VAChT/ChAT locus (Fig. 7a–b)34. The 5′ upstream regulatory region of the VAChT/ChAT locus included in the reporter construct contains a copy of the COE recognition motif, raising the possibility that COE is essential for specifying cholinergic neuronal identity. To investigate this possibility we selectively expressed a dominant-negative form of COE (COE-WRPW) in the visceral ganglion (e.g., the Ciona spinal cord) 35. The resulting transgenic tadpoles fail to induce cholinergic motor neuron differentiation in the spinal cord (Fig. 7c and Supplementary Fig. 10). To ask whether C. intestinalis COE is not only required but also sufficient to induce cholinergic fate, we misexpressed the normal versions of the COE protein under the control of regulatory sequences from the Msx gene. These regulatory elements drive expression in mainly non-cholinergic territories, including future palp sensory neurons, which are glutamatergic 36. Ciona larvae expressing the COE in these cells show ectopic induction of the VAChT reporter, indicating that palp neurons now have adopted cholinergic features (Fig. 7d).
To further explore the functional conservation of Ciona COE and nematode UNC-3, we expressed the Ciona COE gene in a C. elegans unc-3 mutant background. As assessed by analyzing the expression of two unc-3-dependent markers (acr-2 and unc-17), we found that Ciona COE can functionally compensate for the lack of unc-3 (Fig. 7e, Supplementary Fig. 12). We conclude that the role of UNC-3 in generating cholinergic motor neurons is conserved from invertebrates to chordates.
We have provided insights into the coordinated regulation of the core determinants of the cholinergic phenotype of C. elegans motor neurons. While mechanisms of coordinated gene expression are well characterized in, for example, metabolic pathways in bacteria and yeast 37,38, much less is known about how such coordination is achieved within specific cell types in metazoan organisms. We have shown here that cholinergic pathway genes are co-regulated through a common cis-regulatory motif and documented that the cognate trans-acting factor for this binding site is both required and sufficient for the expression of the cholinergic phenotype (schematically summarized in Supplementary Fig. 13a). While cholinergic motor neuron fate is affected by various regulatory factors in mice 39, it is not clear whether such effects are a reflection of true co-regulation through the same trans-acting factor, or a reflection of an indirect effect in which an upstream regulator controls the expression of several trans-acting factors that independently regulate cholinergic pathway genes.
UNC-3 controls the cholinergic phenotype in all but one class of C. elegans ventral cord MNs (VC class) and has no apparent impact on the cholinergic phenotype of non-MN cholinergic neurons. Other cholinergic neuron classes may use similar co-regulatory strategies to control their cholinergic phenotype. Indeed, we find that the TTX-3/CEH-10 homeodomain heterodimer coregulates the cholinergic pathway genes in the head interneuron AIY 29. Together these findings demonstrate that members of the cholinergic gene battery contain a modular array of cis-regulatory motifs that respond to distinct cholinergic selector genes in different neuronal cell types (schematically summarized in Supplementary Fig. 13b). This notion is consistent with reporter gene assays that monitor expression of the vertebrate VAChT/ChAT locus in transgenic mice. A longer and a shorter 5′ upstream regulatory region shows that the 5′ regulatory region contains separable regulatory elements 40, whose detailed cis-regulatory composition is unknown.
Our results go beyond the demonstration that neurotransmitter-specific pathway genes are co-regulated. Results mainly obtained in C. elegans and a few isolated cases in vertebrates have begun to suggest that many features of terminally differentiated neurons, i.e. nuts-and-bolts genes defining the structural and functional properties of a mature neuron (e.g. ion channels, neuropeptides, adhesion molecules), are co-regulated through a common regulatory strategy 41,42. Unbiased analysis of cis-regulatory control regions of terminal gene batteries have revealed shared cis-regulatory motifs and trans-acting factors that act through these motifs. Such transacting factors have been termed “terminal selectors”, as they control the terminal features of a given neuron 41,42. Terminal selectors are continually expressed in mature neurons, thereby ensuring the maintenance of key identity genes. Terminal selectors are, at least in specific cellular contexts, also sufficient to induce specific fates. UNC-3 fulfills all the criteria to be classified as a terminal selector, thereby providing critical support for the as yet unproven hypothesis that terminal selector-based gene regulatory strategies are abundantly employed throughout the nervous system to drive terminal neuronal fates. In total, we have identified 33 genes whose expression depends on unc-3, most of which are likely direct targets of unc-3.
As seen for other terminal selectors, unc-3 may act differently in distinct neuronal cell types. It functions as a terminal selector that activates gene expression in A-, B-, and AS-type MNs, but represses alternative odorsensory fates in the ASI sensory neuron16. This differential activity is presumably conferred by as yet unknown, cell-type specific interaction partners, whose activity may also explain why a subset of terminal markers still show some residual MN expression in unc-3 null mutants. Cell-type specific cofactors are also the likely explanation for the diverse function of unc-3 orthologs in distinct cellular contexts in vertebrates, one being the coregulation of scores of genes involved in B cell terminal differentiation 17.
unc-3 does not represent the regulatory endpoint in cholinergic MN specification. The phylogenetically conserved unc-4, vab-7 and ceh-12 homeodomain repressor proteins are known to diversify the fate of embryonic and postembryonic A- and B-type MNs through negative regulation of terminal differentiation genes, such as acr-5 9,25,26. Previous work did not address, however, how those repressor targets are activated in those cells in which the repressor is not present. We find that these terminal differentiation genes are direct targets of unc-3 (e.g. acr-5) and that the unc-4, vab-7 and ceh-12 transcription factors therefore negatively modulate unc-3-induced transcriptional programs. Intriguingly, each of these transcription factors are also likely direct targets unc-3, as evidenced by their dependence on unc-3 activity, as well as the presence of conserved COE sites in their regulatory regions. Therefore, unc-3 acts in a negative feedforward configuration to ensure that some of its terminal targets are only expressed in a subset of unc-3 expressing neurons (schematically summarized in Supplementary Fig. 13c). Other terminal selectors work through similar negative feedforward loops. For example, the ASE terminal selector che-1 controls the expression of terminal differentiation genes in ASE neurons, but also directly controls regulatory factors that diversify the ASEL and ASER neuronal subtypes (schematically summarized in Supplementary Fig. 13d) 42. Apart from conferring specific kinetic properties to a network 27, negative (also called “incoherent”) feedforward loops may be a reflection of a sequential recruitment of target genes for a transcription factor in an evolutionary context, enabling the system to progress from an ancient state of generic activation (unc-3 activating acr-5 in all MNs) to finer scaled differential activation in distinct cell types (unc-3 able to activate acr-5 only in a subset of MNs).
Our findings in the chordate Ciona intestinalis indicate that the role of UNC-3 as a key determinant of cholinergic MN identity is deeply conserved in animal evolution. Mice express several COE family members in postmitotic, cholinergic spinal cord MNs 30,31 and well-conserved COE motifs can be found in the cis-regulatory region of the cholinergic genes VAChT/ChaT, CHT1, and AChE (Supplementary Fig. 14). Future loss of function studies in mice may further corroborate the notion of deep conservation of the control mechanisms of terminal MN differentiation. We propose that the simple terminal selector regulatory logic may lie at the evolutionary base of neuronal diversity.
Details on the reporter strains used in Fig. 1 and and22 are provided in Supplementary Table 1. The following reporter strains were used: wdEx290 (acr-15::gfp), wdEx419 (acr-16::gfp), ctIs43 (dbl-1::gfp), wdEx346 (tig-2::gfp), rtEx330 ( nlp-21::gfp), vsIs28 (dop-1::gfp), wdEx351 (tsp-7::gfp), otEx223 (rig-4::gfp), zwEx112 (inx-12::gfp), wdEx345(F55C12.4::gfp), wdEx457 (F39B2.8::gfp), wdEx359 (F29G6.2::gfp), wdIs4 (unc-4::gfp), stIs10140 (vab-7::mCherry), stIs10055 (cnd-1::mCherry), wdIs62 (ceh-12::gfp), inIs179 (ida-1::gfp), otIs264 (ceh-36::TagRFP), mgIs18 (ttx-3::gfp), ses28 (kvs-1::gfp), nuEx1072 (gar-2::gfp), nuEx1066 (gbb-1::gfp).
Fosmid recombineering was done as previously described 45, using the following fosmids: cho-1 - WRM0613dC12, ace-2 - WRM0641bD11 and unc-3 - WRM0622bH08. For all fosmid reporters, an SL2 spliced, nuclear localized gfp (mcherry for unc-3 reporter) coding sequence was engineered at the C-terminus of the respective locus, as previously described 45. Fosmids were injected as complex arrays 45.
Reporter gene fusions for cis-regulatory analysis were done using a PCR fusion approach 46. Genomic fragments were fused to a nuclear localized DsRed2 coding sequence, which was followed by the unc-54 3′ UTR. Mutagenesis was performed using the Quickchange II XL Site-Directed Mutagenesis Kit (Stratagene). PCR fusion DNA fragments were injected into young adult pha-1(e2123) hermaphrodites at 50ng/μl using pha-1 (pBX plasmid) as co-injection marker (50ng/μl).
The ceh-36::unc-3_cDNA and hsp-16.2::unc-3_cDNA constructs were a generous gift from Piali Sengupta 16. The ceh-36::unc-3_cDNA construct was injected as simple array at 10 ng/μL. The hsp-16.2::unc-3_cDNA construct was injected into young adult hermaphrodites as a complex array, using 2 ng/μL linearized plasmid DNA, 150 ng/μL PvuII digested bacterial genomic DNA and 2 ng/μL injection marker. The injection markers were pRF4 (rol-6d) for ceh-36prom::unc-3 and ttx-3prom::mCherry for hsp::unc-3. The unc-3 cDNA was fused to 2.8 kb of promoter sequence of the unc-30 gene and the resulting PCR fragment was injected into young adult hermaphrodites at 30 ng/μL using myo-2prom::gfp as co-injection marker (8 ng/μL). The Ciona intestinalis COE cDNA was cloned into the pPD49.78 heat-shock vector using EcoRI. The hsp-16.2::COE construct was injected into young adult hermaphrodites as a complex array, using 2 ng/μL linearized plasmid DNA, 150 ng/μL PvuII digested bacterial genomic DNA and 2 ng/μL injection marker.
Two transgenic lines for hsp-16.2::unc-3 (otEx4536, otEx4441) were used for the heat-shock experiments. For the ectopic induction experiment shown in Fig. 3 and Supplementary Fig. 4, embryos at 2- and 3-fold stage were heat shocked at 37°C three times for 30 minutes with one hour incubation at 20°C between each heat shock to let worms recover. After heat shock worms were kept at 25°C over-night and scored at the indicated times. The same heat-shock protocol was followed for the rescue experiment shown in Fig. 7e and Supplementary Fig. 12 but the heat shock occurred at L1. For the maintenance experiment shown in Fig. 5, third larval stage (L3) worms were heat shocked at 37°C three times for 30 minutes with one hour incubation at 20°C between each heat shock to let worms recover. After heat shock worms were kept at 25°C over-night and then transferred at 15°C for 5 days.
Worms were mounted on 5% agarose on glass slides and images were taken using an automated fluorescence microscope (Zeiss, AXIO Imager Z1 Stand). ). Snapshots were taken for Figures 1 and and2,2, and Supplementary Figures 1, 2, 6 and 11 using the Micro-Manager software (Version 3.1)47. The rest of the images (except Fig. 6) in the remaining figures were created after acquisition of several z-stack images (~1μm thick), and subsequent max-projection using the maximum intensity projection type.
The MatInspector program from Genomatix (http://www.genomatix.de) was used to predict the binding sites for the COE transcription factor family in the cis-regulatory region of all the tested C. elegans genes, as well as the Ciona intestinalis and mouse cholinergic pathway genes. The logo for the position weight matrix that describes the COE binding site in 4 nematode species was created using enoLOGOS 48. Clustalw2 (www.ebi.ac.uk/Tools/msa/clustalw2/) was used to align the COE binding sites in different vertebrate species (Supplementary Fig. 14).
The VAChT -4315/-823 genomic DNA fragment was amplified by PCR using the following primers: VAChT −4315 fwd (CCCTACTGTAACACAGTAAC) and VAChT −823 rev (CTTTCTATTGAATCGTACACCTAAG). This fragment was cloned upstream of unc-76-tagged Venus (YFP). COE>mCherry and FGF8/17/18>H2B::mCherry have been previously described 44. HA-tagged COE::WRPW 35 and Ets::WRPW 43 coding sequences were subcloned downstream of FGF8/17/18 driver in place of H2B::mCherry to make FGF8/17/18>COE::WRPW and FGF8/17/18>Ets::WRPW, respectively. A fragment spanning from −2304 to +145 bp of the Msx gene relative to the start codon was amplified by PCR using the following primers, Msx −2304 fwd (CTACGCATTGATGTCGCAATC) and Msx +145 rev (AGAGGATTGAATGCGATCGG) and cloned upstream of H2B::mCherry. A slightly smaller fragment from −2304 to −74 bp was subcloned upstream of the COE or lacZ coding sequences to produce Msx>COE and Msx>lacZ, respectively. All constructs were made using the pCESA plasmid backbone.
Embryo fertilization, dechorionation, electroporation, and fixation were performed according to established methods 49. Embryos and larvae were mounted in 50% glycerol on glass slides and imaged using an Zeiss LSM 700/Axio Observer. Z1m SP inverted confocal setup (Fig. 7b) or Zeiss AxioImager A.2 compound microscope (Fig. 7c–d). Larvae destined for imaging with the confocal microscope were counterstained with phalloidin-AlexaFluor647 conjugate (Invitrogen).
Statistical analysis was performed using the Student’s t-test (tail 2, type 2). Values are expressed as mean ± standard deviation (s.d). Differences with p value < 0.001 were considered significant.
Supplementary Figure 1. Genes regulated by unc-3 in the ventral nerve cord.
Genes with previously described expression in cholinergic ventral nerve cord (VNC) motor neurons were tested for unc-3 dependence by crossing gfp reporters of these genes into the unc-3(e151) mutant background. At least 20 animals at larval stage 4 (L4) were scored for each reporter. Representative images are shown. Anterior is to the left. gfp expression in motor neurons of the VNC (indicated with an arrow) was severely affected (acr-15, gar-2, tig-2, dop-1, tsp-7, gbb-1, F29G6.2, F55C12.4) or completely lost (acr-16, inx-12, dbl-1, kvs-1, nlp-21, rig-4) in at least 80% of the unc-3(e151) mutant animals. Colored dots indicate the types of motor neurons expressing the reporter, with the color code taken from Fig. 1b. Ach, acetylcholine; GPCR, G protein-coupled receptor; IgSF, immunoglobulin superfamily. Scale bars 30 μm.
Supplementary Figure 2. Motor neurons in the ventral cord of unc-3 mutant animals do not convert into a progenitor-like state.
Animals carrying the cnd-1prom::mCherry transgene were crossed into unc-3(e151) mutants and the number of cnd-1prom::mCherry positive motor neurons in the ventral nerve cord (VNC) was quantified at L2 (a) and L4 (b) stage. c: No differences were found in the number of cnd-1prom::mCherry positive motor neurons between the wild-type and mutant animals at L2 and L4 stages. The cnd-1prom::mCherry positive motor neurons located in the retrovesicular ganglion were not considered. Scale bar 20 μm (a) and 30 μm (b).
Supplementary Figure 3. unc-3 is sufficient to induce cholinergic fate in GABAergic motor neurons.
a: Representative image of an acr-2prom::gfp transgenic animal (upper panel) showing expression in 30 cholinergic motor neurons (MNs), which are located in the ventral nerve cord between the retrovesicular ganglion (RVG) and the pre-anal ganglion (PAG). acr-2prom::gfp; Ex[unc-30prom::unc-3] transgenic animals (lower panel) display increased numbers of gfp positive MNs (36 in the image shown) in the ventral nerve cord. Pharynx is gfp positive because myo-2prom::gfp was used as a co-injection marker for the extrachromosomal array Ex[unc-30prom::unc-3]. Arrows indicate RVG and PAG. gfp positive motor neurons in the RVG and PAG were not included in this analysis due to the super-imposition of their cell bodies. Anterior is to the left. Hermaphrodites were scored at larval stage 4 (L4). Scale bar 30 μm.
b: Quantification of the percentage of animals showing extra acr-2prom::gfp positive MNs (minimum, 1 extra MN; maximum, 7 extra MNs) in the ventral nerve cord in acr-2prom::gfp; Ex[unc-30prom::unc-3] transgenic animals. The unc-30 promoter fragment chosen to drive unc-3 cDNA is active in 13 out of 19 D-type GABAergic MNs. Thirty animals were scored for each transgenic line.
c: The ectopic expression of unc-3 in GABA MNs results in partial loss of the GABAergic marker unc-47prom::mCherry in the ventral nerve cord. The number of mCherry positive MNs was quantified in wild-type (n = 24) and Ex[unc-30prom::unc-3] transgenic (n = 32) animals. Error bars represent standard error of the mean (s.e.m). ***: p value < 0.001.
Supplementary Figure 4. Misexpression of unc-3 results in ectopic expression of the cholinergic marker unc-17::gfp.
The inducible heat-shock promoter (hsp-16.2) was used to achieve broad misexpression of UNC-3. Ectopic expression of the cholinergic fate marker unc-17::gfp along the body of unc-17::gfp; Ex[hsp16.2:unc-3] animals was observed one day after heat shock. Heat-shock induction was performed at the embryonic stage (2–3 fold embryo). Magnified images of the numbered boxes (1 and 2) are shown below. Scale bar 25 μm.
Supplementary Figure 5. Alignment of the UNC-3 binding site (COE motif) in the cis-regulatory region of unc-17, cho-1, ace-2, acr-2 and acr-14 in 4 nematode species.
The genomic location of each UNC-3 site (COE motif) is shown relative to ATG. The logo demonstrates the position weight matrix for the UNC-3 site found in the regulatory region of the aforementioned genes in 4 nematode species. The logo was created using the program enoLOGOS.
Supplementary Figure 6. Identification of novel genes that are expressed in cholinergic MNs of the ventral nerve cord (VNC).
The C. elegans genome was searched for the presence of COE motifs using CisOrtho. DsRed2 reporters for genes with COE motif sites were generated. 7 out of 8 reporters are expressed in cholinergic motor neurons of the VNC. Images are shown for T13G4.3, twk-7, twk-13, twk-40, and twk-43. lgc-55 and acr-21 are also expressed at low levels in the VNC, while acc-2 is only expressed in head neurons (data not shown). Transgenic animals carrying the T13G4.3, twk-7, twk-13, twk-40, and twk-43 reporters were crossed into the unc-3(e151) mutant background. The expression of T13G4.3, twk-13, twk-40, and twk-43 in the ventral cord motor neurons was severely affected in unc-3(e151) animals. Representative images of L4 animals are shown (n = 30). The expression of twk-7 remained unaffected in the unc-3(e151) mutant background, suggesting that UNC-3 co-factors are able to induce expression of a subset of terminal differentiation markers in the ventral nerve cord. Scale bar 30 μm.
Supplementary Figure 7. unc-3 is not expressed in the VC class of cholinergic ventral cord motor neurons.
The unc-3fos::NLS-mCherry reporter shows no co-localization with the VC motor neuron marker ida-1prom::gfp. Arrows indicate the six VC motor neurons at the ventral nerve cord. VC4 and VC5 surround the vulva (indicated at the top panel). The expression of the unc-3fos::NLS-mCherry transgene is shown in white for better contrast (third panel). a; anterior, p;posterior, d; dorsal, v; ventral, NLS; nuclear localization signal. Scale bar 15 μm.
Supplementary Figure 8. unc-3 is required to maintain cholinergic gene expression.
Heat-shock induction of unc-3 expression at the larval stage L3 restores unc-17::gfp expression in the VNC of unc-3(e151) mutant animals (note that unlike heat-shock induction in the embryo shown in Fig. 3, this postembryonic induction does not result in the generation of ectopic cholinergic cells). The number of unc-17::gfp positive VNC neurons was assessed one day after heat-shock (L3 + 1 day). However, no increase in the number of unc-17::gfp positive neurons is observed in unc-3(e151); unc-17::gfp animals that carry the extrachromosomal array Ex[hsp-16.2::unc-3] in the absence of heat shock. A significant decrease (p value = 1.10498E-18) in the number of unc-17::gfp positive VNC neurons was observed in unc-3(e151); unc-17::gfp;Ex[hsp-16.2::unc-3] animals six days after heat shock (L3 + 6 days) when compared to one day after heat shock (L3 + 1 day). No differences were found in the number of unc-17::gfp positive VNC neurons in the wild-type and unc-3(e151) mutant background at the indicated time points. Two different hsp-16.2::unc-3 extrachromosomal lines (otEx4536, otEx4441) were analyzed for this experiment and showed similar effects. 30 – 50 animals were used per time point per genotype. See Methods section for details on heat-shock protocol. Quantification of the number of unc-17::gfp positive VNC neurons is shown at the indicated time points at the bottom left. Error bars represent standard error of the mean (s.e.m). ***: p value < 0.001. A schematic describing the heat-shock protocol is shown at the bottom right. Scale bars: 25 μm for L3 stage, 30 μm for L3 + 1d, and 50 μm for L3 + 6d.
Supplementary Figure 9. cho-1 is expressed in the cholinergic interneuron AIY and its expression is controlled by the terminal selector ttx-3.
The expression of the AIY marker (ttx-3::gfp) overlaps with cho-1 reporter expression, in which 3.7 kb of cho-1 regulatory region is fused to DsREd2. The gfp/DsRed2 overlap was observed in 29 out of 32 animals examined (90.6%). Strains used: ttx-3::gfp, mgIs18; cho-1 3.7kb prom:DsRed2, otEx5435. The cho-1 expression in AIY is lost in ttx-3(ot22) mutants (27 out of 34 animals analyzed, 79.4%). The right AIY interneuron (AIYR) was identified by the stereotypic position of its nucleus, using DIC microscopy. Scale bars 10 μm.
Supplementary Figure 10. COE::WRPW overexpression converts the motor neuron lineage into ependymal-like cells.
Mid-tailbud Ciona embryos electroporated with VAChT>YFP (yellow) and FGF8/17/18>H2B::CFP (cyan) reporters and either a) FGF8/17/18>lacZ (control) or b) FGF8/17/18>COE::WRPW. Embryos were counterstained with phalloidin-AF647 (magenta) to reveal cell outlines. White dashed line: apical side of the neural tube epithelium. Red dashed line: outline of A9.30 lineage-derived cholinergic MN delaminated from the neural tube. COE::WRPW overexpression converts the lineage into ependymal cells that do not activate VAChT reporter nor delaminate from the neuroepithelium.
Supplementary Figure 11. The expression of unc-17, acr-2, acr-5, acr-14, and unc-8 is affected in unc-3 (n3435) mutant animals.
Similar effects in the expression of unc-17, acr-2, acr-5, acr-14, and unc-8 were observed in the unc-3(e151) and unc-3(n3435) mutant background (See also Fig. 1 and Fig. 2). The e151 mutation generates a premature stop codon, while the n3435 mutation introduces a deletion, which removes the DNA binding domain of UNC-3. Animals were photographed and scored at the larval stage L4. Arrows indicate MNs in the VNC expressing the cholinergic reporters. Colored dots indicate the types of MNs expressing the reporter, with the color code taken from Figure 1b. Quantification of the number of MNs expressing each reporter in wild-type and unc-3(n3435) mutant animals is shown on the right. Error bars represent standard deviation (s.d). ***: p value <0.001. Scale bars 30 μm.
Supplementary Figure 12. Ciona COE rescues the gfp expression defects of unc-3 mutant animals.
a: Animals carrying the heat-shock inducible extrachromosomal array Ex[hsp::COE] were crossed into unc-3(e151); unc-17::gfp mutants. COE expression was induced at the first larval stage and animals were scored one day later. In the absence of heat shock, only the VC class of MNs are expressing unc-17 in the VNC of unc-3(e151);unc-17::gfp; Ex[hsp::COE]. Following heat-shock induction of Ciona COE, MNs that belong in A-, B-, or AS- class are expressing unc-17 (apart from VCs, 3–25 extra gfp positive MNs were counted in the VNC of unc-3(e151);unc-17::gfp; Ex[hsp::COE] animals). Three independent transgenic lines were scored. b: In the absence of heat shock, the unc-3(e151);acr-2::gfp; Ex[hsp::COE] animals display residual acr-2 expression due to partial penetrance. Heat-shock induction of Ciona COE results in partial restoration of acr-2 expression in A- or B-type MNs (1–20 extra gfp positive MNs were counted in the VNC of unc-3(e151);acr-2::gfp; Ex[hsp::COE] animals). Three out of four independent transgenic lines showed this effect. The number of animals that showed partial rescue of unc-17 or acr-2 expression is shown at the top of each bar.
Supplementary Figure 13. Diagrams summarizing the major findings of this paper.
a: Summary of coregulatory strategy through which unc-3 controls cholinergic MN differentiation. Parallel regulatory routines ensure the expression of panneuronal markers through as yet unknown mechanisms (indicated with question marks).
b: Terminal differentiation genes that are expressed in diverse cell types, such as the unc-17/VAChT gene which is expressed in the complete set of cholinergic neurons in C. elegans, are regulated through a modular, piece-meal assembly of cis-regulatory motifs that are activated by neuron-type specific terminal selectors, such as TTX-3/CEH-10 (cholinergic AIY interneurons) and UNC-3 (cholinergic A-, B-, and AS-type MNs).
c: Negative feedforward loops diversify motor neuron differentiation programs. unc-3 directly regulates terminal differentiation genes (e.g. acr-5) as well as transcription factors that inhibit the expression of terminal differentiation genes in a subset of motor neurons. This network configuration is conceptually similar to diversification of the gustatory neuron fates across the left right axis (panel D). Grey indicates that the gene is not expressed. “VA input genes” refer to genes that determine synaptic inputs into the VA neurons.
Supplementary Figure 14. COE motifs in mouse cholinergic genes.
The genomic loci of the mouse cholineric gene battery (panel a: VAChT/ChaT, Panel b: AChE, Panel c: ChT1) as they appear in UCSC browser (http://genome.ucsc.edu/). Phylogenetic conservation of the loci is shown at the bottom of each image. Schematics (below each image) show the COE motifs (red vertical line) in the regulatory region of the respective mouse gene. Blue rectangles span from ATG to STOP codon of each cholinergic gene. The relative to ATG position of each COE motif is indicated in brackets. Alignment of each COE motif is shown on the right. The position weight matrix for the COE family of transcription factors (presented also in Fig. 4a) is shown at the bottom right for comparison.
Supplementary Figure 15. UNC-3 is sufficient to induce the motor neuron-specific marker acr-2 in the AIY interneurons.
The inducible heat-shock promoter (hsp-16.2) was used to achieve broad misexpression of unc-3. The motor neuron-specific marker acr-2 labels cholinergic motor neurons in the ventral nerve cord (arrowheads), whereas ttx-3prom::mCherry labels the AIY interneurons (dashed circles). As expected, in non-heat shocked animals no overlap is observed for acr-2prom::gfp and ttx-3prom::mCherry because the AIY interneurons do not express acr-2 (upper panel; line #1, 0 out of 20 animals showed overlap of acr-2prom::gfp and ttx-3prom::mCherry; line #2, 1 out of 25 animals showed overlap of acr-2prom::gfp and ttx-3prom::mCherry). Following heat shock-induction of unc-3 in the embryo (2–3 fold stage), ectopic expression of the cholinergic marker acr-2prom::gfp is observed in the head and along the body of the worm, scored one day after heat shock (see also Figure 3). In the head, expression of the AIY specific marker ttx-3prom::mCherry is reduced and co-localizes with the cholinergic motor neuron marker acr-2prom::gfp after heat-shock induction of unc-3 (lower panel). Two independent transgenic hsp16.2::unc-3 lines were used and showed similar effects (line #1, 14 out of 20 animals showed overlap of acr-2prom::gfp and ttx-3prom::mCherry; line #2, 13 out of 25 animals showed overlap of acr-2prom::gfp and ttx-3prom::mCherry). Representative images of acr-2prom::gfp; ttx-3prom::mCherry; hsp16.2::unc-3 transgenic animals are shown. Anterior is to the left. Scale bars 10 μm.
Supplementary Table 1. Quantification of terminal differentiation defects in A-, B-, and AS-type motor neurons in unc-3(e151) animals.
Supplementary Table 2. Quantification of promoter deletion analysis on UNC-3 target genes.
Supplementary Table 3. UNC-3 motifs are enriched in DA/VA-type transcriptome set.
Supplementary Table 4: Expression of the Ciona VAChT −4315/−823>YFP construct
We thank Q. Chen for expert assistance in generating transgenic strains, P. Sengupta for providing reagents, D. Wu and G. Minevich for bioinformatic analysis, J. Rand and members of the Hobert lab for comments on the manuscript. We are grateful to Caenorhabditis Genetics Center (University of Minnesota) for providing strains. This work was funded by the Muscle Dystrophy Association and the NIH (R01NS039996-05; R01NS050266-03). O.H. is an Investigator of the Howard Hughes Medical Institute.
AUTHOR CONTRIBUTIONSP.K. performed the C.elegans experiments under supervision of O.H., A.S. performed the Ciona intestinalis experiments under supervision of M.L. All authors participated in the writing of this paper.