Maintenance of neuronal identity in C. elegans and beyond: lessons from transcription and chromatin factors

Abstract

Neurons are remarkably long-lived, non-dividing cells that must maintain their functional features (e.g., electrical properties, chemical signaling) for extended periods of time – decades in humans. How neurons accomplish this incredible feat is poorly understood. Here, we review recent advances, primarily in the nematode C. elegans, that have enhanced our understanding of the molecular mechanisms that enable post-mitotic neurons to maintain their functionality across different life stages. We begin with “terminal selectors” - transcription factors necessary for the establishment and maintenance of neuronal identity. We highlight new findings on five terminal selectors (CHE-1 [Glass], UNC-3 [Collier/Ebf1-4], LIN-39 [Scr/Dfd/Hox4-5], UNC-86 [Acj6/Brn3a–c], AST-1 [Etv1/ER81]) from different transcription factor families (ZNF, COE, HOX, POU, ETS). We compare the functions of these factors in specific neuron types of C. elegans with the actions of their orthologs in other invertebrate (D. melanogaster) and vertebrate (M. musculus) systems, highlighting remarkable functional conservation. Finally, we reflect on recent findings implicating chromatin-modifying proteins, such as histone methyltransferases and Polycomb proteins, in the control of neuronal terminal identity. Altogether, these new studies on transcription factors and chromatin modifiers not only shed light on the fundamental problem of neuronal identity maintenance, but also outline mechanistic principles of gene regulation that may operate in other long-lived, post-mitotic cell types.

1. INTRODUCTION

Human neurons must acquire during development and maintain, for decades of post-natal life, their functional features: e.g., neurotransmitter (NT) biosynthesis, electrical activity, and signaling properties¹. Understanding how these non-dividing (post-mitotic) cells accomplish this remarkable feat can provide critical insights into the etiology, diagnosis, and treatment of developmental and degenerative disorders of the nervous system. Extensive research over the past few decades has made seminal contributions to our understanding of the molecular mechanisms underlying neurogenesis. However, the focus on early development has left poorly explored the molecular mechanisms underlying the last steps of neuronal differentiation, during which post-mitotic neurons acquire their functional features¹. Further, and perhaps most importantly, the mechanisms that ensure maintenance of such features throughout post-natal life remain largely unknown; these mechanisms are the focus of this review.

During the final steps of neuronal differentiation, distinct neuron types must express neuron type-specific gene batteries of terminal identity genes (often called “effector genes”). These encode proteins that determine the functional properties of a neuron throughout its lifetime², e.g., enzymes and transporters necessary for NT biosynthesis, ion channels that determine the resting potential of a neuron, neuropeptides and other signaling proteins (Figures 1–2)². Altogether, these proteins form a distinct molecular signature for each neuron type that defines its terminal identity and function throughout life. Therefore, the challenge of understanding how neurons acquire and maintain their functional features lies in understanding how the expression of these terminal identity genes is regulated over time in specific neuron types. Are the mechanisms used for initial establishment (during development) and maintenance (throughout life) of neuron type-specific terminal identity genes distinct, or do establishment and maintenance rely upon the same mechanism? Addressing this question impacts the fields of developmental neuroscience and stem cell research, where it remains an outstanding challenge to generate mature and specialized neuron types in vitro, suitable for disease modeling and drug screening.

Figure 1. Terminal selectors of neuronal identity.

Terminal selector transcription factors can operate alone, or in combinations, to establish and maintain neuronal terminal identity. They directly regulate expression of terminal identity genes and intermediary transcription factors through binding at the cis-regulatory regions of these genes. Continuous terminal selector expression, required for neuronal identity maintenance, is ensured through direct positive autoregulation.

Figure 2. Terminal selector CHE-1 9 controls ASE identity.

A. Schematic of ASE chemosensory neurons in the C. elegans head. B. CHE-1 directly promotes terminal identity gene and intermediary transcription factor expression through binding at its cognate site in the cis-regulatory regions of these genes. Continuous CHE-1 expression is ensured through autoregulation. C. ASE subtype diversification via a bistable feedback loop. CHE-1 specifies both ASEL and ASER subtype-specific fates through activation of a downstream bistable feedback loop. Initiation of ASER fate by the transcription factor COG-1 (later maintained by FOZI-1) is antagonized by DIE-1 (via the miRNA lsy-6) in the ASEL.

In this review, we primarily focus on the nervous system of the nematode Caenorhabditis elegans, where recent studies have advanced our understanding of the molecular principles that maintain neuronal terminal identity. To relate to readers outside the C. elegans community, we also compare C. elegans studies with findings reported in Drosophila and mice. Due to space constraints, we only cover C. elegans studies in a comprehensive manner.

1.1. Terminal selectors establish and maintain neuronal identity

The nematode C. elegans is a powerful system to study the problem of neuronal identity; it offers an unparalleled resource of available terminal identity markers that label with single-cell resolution each of the 118 neuron types that make up its nervous system^3,4. Studies conducted with these markers, along with powerful genetics and behavioral analysis, led to the discovery of “terminal selectors”, a distinct class of regulators of neuronal identity. Terminal selectors are transcription factors with continuous expression, from development through adulthood, in specific neuron types^4–6. They can act in combinations to establish neuronal terminal identity during development. Mechanistically, terminal selectors bind directly to the cis-regulatory regions of neuron type-specific terminal identity genes (e.g., NT receptors, ion channels, neuropeptides) and activate their transcription (Figure 1).

Terminal selectors have been described to date for 111 of the 118 C. elegans neuron types^4,7. Beyond C. elegans, terminal selectors have been identified in the nervous system of fruit flies (Drosophila), cnidarians (Nematostella vectensis), marine chordates (Ciona intestinalis), and mice (Mus musculus)⁸, suggesting a deeply conserved role for this type of transcriptional regulators. While the essential role of terminal selectors in establishing neuronal terminal identity during development is well-attested, their involvement in maintaining neuronal identity in later life remains poorly examined.

Here, we provide an update on recent work demonstrating a continuous requirement for different terminal selectors in specific neuron types, solidifying their critical roles not only in establishing neuronal terminal identity during development, but also in maintaining it during later life stages. We focus on the best characterized examples of C. elegans terminal selectors: CHE-1 (Glass) in head sensory neurons, UNC-3 (Collier/Ebf) and LIN-39 (HOX) in nerve cord MNs, AST-1 (ETS) in dopamine neurons, and UNC-86 (POU) in sensory neurons (Table 1). Further, we highlight evolutionarily conserved functions of these terminal selectors in Drosophila and mouse neurons, complementing a previous review article on this subject¹ (Table 1). Due to space constraints, we do not cover emerging studies suggesting a broader role for terminal selectors in the adult, i.e., they are not only required to maintain neuronal identity, but also maintain axonal architecture and connectivity^11,12. Finally, we go beyond transcription factorbased mechanisms and highlight recent exciting work on how chromatin-modifying proteins control neuronal terminal identity.

Table 1.

Transcription factors and chromatin modifiers necessary for maintenance of neuron type identity

Transcription or chromatin factor	Neuron type	Literature
C. elegans
CHE-1 (C2H2 Zn finger)	ASEL/R gustatory neurons	16,17
DIE-1 (C2H2 Zn finger)	ASEL subtype identity	25
LSY-12 (MYST-type histone acetyltransferase)	ASEL subtype identity	25
UNC-3 (EBF)	ventral nerve cord cholinergic MNs	31,39,63
CFI-1 (ARID)	ventral nerve cord cholinergic MNs	39,40
BNC-1 (Zn finger)	cholinergic MN subtype (VA, VB)	42
MAB-9 (T-box)	cholinergic MN subtype (DA, DB)	42
UNC-4 (paired-type homeobox)/UNC-37 (Groucho-like?) complex	cholinergic MN subtype (DA)	115
LIN-39	midbody cholinergic MNs	32,63
UNC-86 (POU)	PLM glutamatergic touch sensory neurons
UNC-86 (POU)	PVD glutamatergic sensory neuron	88
UNC-86 (POU)	PHC glutamatergic sensory neurons	88
UNC-86 (POU)	PLN cholinergic sensory neurons	88
UNC-86 (POU)	PQR peptidergic sensory neuron	88
UNC-86 (POU)	HSN serotonergic MN	88
UNC-86 (POU)	BDU peptidergic interneurons	88
UNC-86 (POU)	URB cholinergic interneuron	88
UNC-86 (POU)	URX cholinergic sensory neuron and ring interneuron	88
UNC-86 (POU)	NSM serotonergic pharyngeal neuron	88
AST-1 (ETS), UNC-86 (POU), SEM-4 (Spalt-type Zn finger), EGL-46 (Insm-type Zn finger), EGL-18 (GATA factor)	HSN serotonergic MN	86
CEH-43 (Dlx-type homeodomain)/AST-1 (ETS)	All dopaminergic neurons	93,94
CTBP-1	AIA interneuron	116
LAG-1	ADF serotonergic chemosensory neuron	117
mig-32 and spat-3 (PRC1 component)	AIY interneurons	109
D. melanogaster
Ubx	thoracic dopaminergic neurons	80
Ubx and abdA	abdominal leucokinergic neurons	81
Dfd	feeding MNs	82
M. musculus
Hoxc8	brachial MNs	74
Hoxa5	phrenic motor column neurons	79
Hoxa5	brainstem neurons	77
BRN3A (POU)	medial habenula glutamateric and cholinergic neurons	88
CBP and p300 (KAT3)	excitatory hippocampal neurons	112,113
Ezh2 (PRC2 component)	cerebellar Purkinje cells, medium spiny neurons	106
Eed (PRC2 component)	midbrain dopaminergic neurons	105
Pet-1	Hindbrain serotonin neurons	46,47

2. Lessons from the terminal selector CHE-1⁹ in chemosensory neurons

The ASE chemosensory neurons in the C. elegans head have been a prime model to obtain mechanistic insights on the control of neuronal terminal identity. The ASE neurons are a bilateral pair, with one neuron on the left (ASEL) and another on the right (ASER), critical for sensing water-soluble chemical cues. ASEL and ASER demonstrate morphological symmetry, and share many molecular features, including terminal identity genes that code for neuropeptides, ion channels, and signaling molecules (Figure 2A). However, ASEL and ASER also show molecular and functional asymmetries¹³; while ASEL detects one set of salt cues (e.g., sodium and lithium) and expresses specific transmembrane receptors (e.g., gcy7/guanylate cyclase), ASER detects a separate set of cues (e.g., potassium and chloride) and expresses another set of transmembrane receptors (ex. gcy-5/guanylate cyclase). Hence, they are classified into separate subtypes, ASEL and ASER (Figure 2A–B).

2.1. CHE-1 is required to establish and maintain ASE neuron identity

The che-1 (chemotaxis-defective) gene encodes a C2H2 zinc-finger transcription factor selectively expressed in both ASE neurons^14,15. CHE-1 (Glass in Drosophila) is required for ASE identity; in che-1 null mutant animals, ASE neurons are normally generated but fail to acquire their terminal identity. Expression of ASE-shared and subtype-specific terminal identity genes (e.g., ion channels, guanylate cyclases, neuropeptide receptors) is lost. CHE-1 is also sufficient for ASE gene expression; expression of CHE-1 in other neuron classes causes ectopic expression of ASE terminal identity genes¹⁵. CHE-1 directly regulates these genes by binding at their cis-regulatory regions via its cognate binding site¹⁴ (Figure 2B). The expression of che-1 in post-mitotic ASE neurons is continuous, suggesting a role for CHE-1 in maintenance of ASE identity in the adult^14,16. Indeed, temporally controlled CHE-1 depletion through RNAi or auxin-inducible degradation after ASE birth showed that CHE-1 is continuously required to maintain ASE identity and chemotactic behavior^16,17. Hence, CHE-1 acts as a terminal selector of ASE terminal identity.

2.2. Positive autoregulation of che-1 is required to establish and maintain ASE neuron identity

A key feature of terminal selectors is their maintained expression in specific neuron types, often achieved through positive autoregulation^5,6. Expression of che-1 in ASE neurons is initiated, but not maintained, by the nuclear receptor NHR-67/Tailless/TLX¹⁸. Two pieces of evidence suggest CHE-1 employs direct positive autoregulation after initiation. First, che-1 reporter expression is lost in ASE neurons of che-1 mutant animals¹⁴. Second, CRISPR/Cas9-mediated mutation of the CHE-1 binding site in the endogenous che-1 locus led to a striking decrease of che-1 expression in adult ASE neurons¹⁹. Importantly, this manipulation impacted the expression of ASE terminal identity genes and led to chemosensory defects in adults. Altogether, direct positive autoregulation of che-1 is required to maintain ASE terminal identity during late developmental and adult stages. Interestingly, Glass (CHE-1 ortholog) also acts as a terminal selector in sensory neurons (photoreceptors) of the Drosophila retina. Like che-1, Glass positively autoregulates and controls the expression of terminal identity genes in photoreceptors^20,21. Hence, the terminal selector function of che-1 appears conserved from C. elegans to Drosophila, albeit in a different neuron type.

Positively autoregulating genes must be continuously expressed above a critical threshold to ensure maintenance of their autoregulation^17,19. In the case of CHE-1, transient depletion of CHE-1 protein caused a permanent loss of CHE-1 expression, ASE identity, and chemotactic behavior¹⁷. This suggests additional mechanisms must be in place to maintain CHE-1 positive autoregulation. One possible mechanism is preferential recruitment of CHE-1 to its own promoter relative to its target gene promoters. In silico modeling, informed by in vivo quantification of mRNA levels and protein half-life, demonstrated that during depletion of CHE-1, levels of target gene transcripts decreased, whereas che-1 transcript levels did not. Taken together, these data suggest a model where CHE-1 binding at target gene promoters provides a reservoir of CHE-1 protein that can then be preferentially recruited to the che-1 locus upon a transient drop in CHE-1 levels. This “target gene reservoir buffering model” may constitute a general mechanism for stable maintenance of cell identity¹⁷.

2.3. A bistable feedback loop downstream of CHE-1 diversifies ASE neurons

Since che-1 acts as a terminal selector in both ASE neuron subtypes (ASEL/R), how are the unique identities of ASEL and ASER established and maintained? A gene regulatory network downstream of CHE-1 composed of miRNAs and transcription factors is responsible for ASE subtype diversification (Figure 2C)^13,22–24. One key transcription factor within this network is the C2H2 zinc-finger transcription factor DIE-1¹³. Initially identified in a genetic screen, DIE-1 is required for appropriate ASE subtype specification. DIE-1 functions to simultaneously promote ASEL and repress ASER identity; in die-1 genetic null mutant animals, ASEL is transformed to an ASER identity. Importantly, post-developmental removal of die-1 via RNAi also led to an ASEL-to-ASER identity conversion, indicating that die-1 is required continuously to maintain ASEL identity^22,25.

Restriction of die-1 expression to ASEL is initially ensured by the homeobox gene cog-1/GTX/Nkx6, and maintained by the zinc finger transcription factor fozi-1²². In turn, die-1 functions through the miRNA lsy-6 to repress cog-1 from being expressed in the ASEL^13,24. Altogether, die-1, lsy-6, and cog-1 form a double-negative feedback loop downstream of the terminal selector CHE-1. This loop ensures ASE neurons adopt one of two possible identities (ASEL or ASER), thereby ensuring their diversification (Figure 2C). Feedback loops like the one described here are referred to as “bistable” because they determine that a system (e.g., ASE neurons) stably exists in one of two possible states^26,27.

2.4. The MYST histone acetyltransferase complex is required to initiate and maintain ASE subtype identity

Intriguingly, a forward genetic screen for genes involved in ASEL/ASER specification identified lsy-12 (laterally symmetric), which encodes a chromatin-modifying enzyme²⁸. Specifically, lsy-12 is an MYST-type histone acetyltransferase (HAT) of the MOZ/MORF subfamily²⁵. Animals carrying a mutant allele of lsy-12 display the ASEL-to-ASER conversion phenotype, and post-developmental elimination of LSY-12 function causes a similar phenotype, demonstrating that LSY-12 is continuously required to maintain ASEL identity²⁵. The similarity between DIE-1 and LSY-12 in the maintenance of ASEL neuron identity suggests a model in which DIE-1, a sequence-specific transcription factor, recruits LSY-12, a MYST-type HAT, to the cis-regulatory regions of ASEL-specific terminal identity genes, thereby ensuring local histone acetylation, and thus transcriptional activation, of ASEL genes²⁵. Taken together, the continuous requirement of DIE-1 and LSY-12 exemplifies the need of both transcription factors and chromatin-modifying proteins for the control of neuronal identity maintenance.

3. Lessons from UNC-3 (Collier/Ebf) in cholinergic motor neurons

In the C. elegans ventral nerve cord, a structure analogous to the vertebrate spinal cord, there are 8 different classes (or subtypes) of motor neurons (MNs)²⁹. Five subtypes of cholinergic (DA, DB, VA, VB, AS) and two subtypes of GABAergic (DD, VD) MNs control locomotion, whereas one cholinergic subtype³⁰ is essential for egg laying (Figure 3A). At least one terminal selector has been identified to date for each of these subtypes: UNC-3 (Collier/Ebf) controls DA, DB, VA, VB, and AS identity³¹, LIN-39 (Scr/Dfd/Hox4-5) controls VC identity³², and UNC-30 (PITX) controls DD and VD identity^33,34. Below, we highlight recent studies on UNC-3 that provide mechanistic insights into how terminal selectors maintain neuronal identity throughout life.

Figure 3: UNC-3 is a terminal selector of cholinergic MN identity.

A. Five (colored) cholinergic MN subtypes used for C. elegans locomotion. The 50 cells that comprise these subtypes intermingle along the ventral nerve cord. B. UNC-3/Ebf activates, via direct binding at its cognate site (COE motif) in cis-regulatory regions, a diverse suite of genes essential for cholinergic MN function. Schematic modified from¹¹⁴ C. UNC-3 broadly activates both shared and subtype-specific terminal identity genes for cholinergic MNs. Subtype diversity is established and maintained by various subtype-specific repressor proteins (R1, R2) that antagonize UNC-3’s ability to activate terminal identity genes.

3.1. The terminal selector UNC-3 (Collier/Ebf) establishes cholinergic MN identity during development

The unc-3 gene, named after the uncoordinated locomotion phenotype its mutation causes, encodes the sole C. elegans orthologue of the COE (Collier/Olf/Ebf) transcription factor family^35,36. UNC-3 is expressed in neurons of the head, tail, and ventral nerve cord^31,37. In five subtypes of cholinergic MNs (DA, DB, VA, VB, AS), UNC-3 promotes the expression of dozens of terminal identity genes (e.g., NT biosynthesis components, NT receptors, ion channels, neuropeptides)^31,38,39 (Figure 3B). Notably, these MNs are normally generated in unc-3 mutants, but fail to acquire their terminal identity. UNC-3 is also sufficient to promote cholinergic MN identity, as its forced expression in non-cholinergic neurons induced the ectopic expression of MN markers³¹. Two major pieces of evidence suggest UNC-3 binds directly to activate its target genes. First, ChIP-sequencing for UNC-3 revealed hundreds of terminal identity genes as putative direct targets, with UNC-3 binding occurring in their promoters and/or enhancers³⁹. Second, mutagenesis experiments of UNC-3’s cognate DNA binding sites (COE motifs) in target genes resulted in unc-3 mutant-like phenotypes^31,39,40. In summary, UNC-3 acts directly to establish cholinergic MN identity.

3.2. UNC-3 (Collier/Ebf) maintains cholinergic MN identity

The continuous expression of UNC-3 in cholinergic MNs throughout adulthood suggests UNC-3 is required at both early and later stages of life to control MN identity³¹. While UNC-3’s importance in promoting terminal identity gene expression in early stages has been demonstrated in animals carrying unc-3 null alleles^31,38, its continuous requirement in maintaining that expression was only recently shown. Using an auxin-inducible degradation⁴¹ method for UNC-3 depletion selectively in adult cholinergic MNs, two recent studies reported adult-specific loss of expression of various UNC-3 targets genes^39,42. Altogether, the terminal selector UNC-3 (Collier/Ebf) initiates during development and maintains throughout life the expression of numerous terminal identity genes in cholinergic MNs (Figure 3B), thus ensuring locomotory function in C. elegans.

3.3. Transcriptional repressors antagonize UNC-3 to generate and maintain MN diversity

Neurons are classified into types and subtypes based on molecular, morphological and functional criteria. Like the ASEL/R neurons discussed above (Figure 2C), the five cholinergic subtypes of nerve cord MNs (DA, DB, VA, VB, and AS) also provide an ideal model to study how neuronal subtype identities are established and maintained (Figure 3C)^31,42. The terminal selector UNC-3 not only initiates and maintains the expression of terminal identity genes shared by all five subtypes (e.g., genes coding for acetylcholine biosynthesis components), but also of genes specific to certain subtypes (e.g., ion channels, NT receptors, neuropeptides) (Figure 3B). What prevents UNC-3 from activating these subtype-specific terminal identity genes more broadly, i.e., in all five MN subtypes?

Two recent studies pointed out that the answer lies in transcriptional repressor proteins, present in unique combinations in each MN subtype^40,42. These repressors bind directly to the cis-regulatory regions of UNC-3 target genes and prevent their broad activation by UNC-3^40,42 (Figure 3C). For example, the combinatorial repressor activity of C2H2 Zinc finger transcription factor BNC-1 (Disco/BNC1–2) and Prd-type homeodomain protein UNC-4 (UNCX) defines VA subtype identity (Figure 3C). Similarly, the transcription factors MAB-9 (TBX20), VAB-7 (Eve/EVX), CFI-1 (Dead ringer/ARID3) and UNC-55 (COUP) also repress UNC-3 target gene expression in specific MN subtypes^40,42. These repressor proteins are also continuously expressed in adult MNs, so they may be continuously required to antagonize UNC-3. Indeed, post-embryonic depletion of BNC-1 or MAB-9 with the auxin-inducible degradation system shows that repressors are continuously required to maintain MN subtype terminal identity^40,42. Altogether, MN subtype identities are shaped by unique combinations of repressor proteins that continuously antagonize the terminal selector UNC-3. This strategy may constitute a general principle of neuron subtype diversification, as suggested by studies in C. elegans sensory neurons⁴³, as well as in mouse midbrain dopaminergic neurons⁴⁴ and retina photoreceptors⁴⁵.

3.4. UNC-3 function in MNs is modular across different life stages

The continuous requirement of transcriptional repressors in C. elegans MNs prompts the question of how repressor expression is maintained throughout life. Unlike terminal selectors, these repressors do not positively autoregulate^39,42. Instead, the terminal selector UNC-3 is specifically required to maintain repressor expression in MNs during late larval and adult stages. This maintenance-only requirement contrasts with UNC-3’s critical role in both initiating and maintaining the expression of terminal identity genes (Figure 3B). Hence, the suite of UNC-3 target genes in C. elegans MNs is partially modified over time because UNC-3 gains new targets (e.g., transcriptional repressors) specifically during late larval and adult stages. This phenomenon was recently described as “temporal modularity” in terminal selector function³⁹.

Current evidence suggests that such modularity is beneficial for generating and maintaining neuronal subtype diversity. By expanding the pool of its target genes in late development to include transcriptional repressors, UNC-3 has an additional means of regulating individual subtype identities. Besides acting directly to activate and maintain expression of subtype-specific terminal identity genes, UNC-3 can also indirectly repress alternate identity genes by maintaining repressor gene expression (Figure 3C).

Temporal modularity has also been observed in mouse serotonergic neurons, where the terminal selector Pet-1 changes its transcriptional targets over time⁴⁶. Pet-1 is also required to maintain chromatin accessibility during serotonergic neuron maturation, suggesting open chromatin in post-mitotic neurons requires continuous regulatory input by a terminal selector⁴⁷. The cases of UNC-3 and Pet-1 showcase how the same terminal selector controls different target genes over time to establish and maintain neuron type identity. A recent study in embryonic stem cell (ESC)-derived MNs reported an additional mechanism: maintenance of expression of the same terminal identity genes is achieved via transient enhancers activated by stage-specific combinations of TFs⁴⁸. Altogether, these studies point out that the transcriptional programs that maintain neuronal identity are more dynamic than previously thought and require further investigation across different life stages.

3.5. The terminal selector function of UNC-3 is conserved across species

Accumulating evidence suggests that the terminal selector function of UNC-3 is conserved across species. In the chordate Ciona intestinalis, UNC-3 ortholog COE is both necessary and sufficient for cholinergic MN identity³¹. In mice, UNC-3 orthologs Ebf1 and Ebf2 are respectively expressed in MNs of the hypaxial and medial columns of the developing spinal cord. Ebf2 is required for proper differentiation of a subset of these neurons⁴⁹. While Ebf1’s function remains unknown, it remains highly expressed in spinal MNs of adult mice⁵⁰. Mouse Ebf1 and Ebf2, and C. intestinalis COE, demonstrate functional equivalence to UNC-3, as each is able to compensate for loss of unc-3 in C. elegans^31,49.

UNC-3 orthologs also control neuronal terminal identity beyond MNs. In C. intestinalis, COE controls the terminal identity of dorsal interneurons at the motor ganglion (analogous to vertebrate spinal cord)⁵¹, and in mice, EBF transcription factors (along with Lhx2 and LDB1) are required for the terminal differentiation of olfactory neurons^52,53,54,55. Importantly, mutations in the human UNC-3 ortholog EBF3 cause a neurodevelopmental disorder characterized by developmental delay, ataxia and intellectual disability^56,57. Future studies are needed to determine whether these clinical manifestations are due to loss of EBF3 terminal selector function in human neurons.

4. Hox proteins function as terminal selectors in C. elegans MNs

The highly conserved family of Hox transcription factors has been primarily associated with early development roles, such as anterior-posterior patterning of the embryo and the nervous system^58,59. An overwhelming body of work in invertebrate and vertebrate models firmly attests that Hox gene function is essential for cell survival, proliferation, specification and migration. However, recent studies in the C. elegans nervous system have uncovered non-canonical roles for Hox genes during later stages of nervous system development and adult life, broadening the functional repertoire of this fundamental family of transcription factors beyond early patterning. Below, we highlight findings demonstrating a continuous requirement for Hox gene activity in the context of post-mitotic neurons in C. elegans, Drosophila and mice.

4.1. C. elegans Hox genes are continuously expressed in MNs

Like in other organisms, Hox genes in C. elegans are chromosomally clustered. In total, there are six Hox genes (ceh-13/Lab/Hox1, lin-39/Scr/Dfd/Hox4-5, mab-5/Antp/Hox6-8, egl-5/AbdB/Hox9-13, nob-1/Abd-B/Hox9-13, and php-3/Abd-B/Hox9-13)⁶⁰, five of which exhibit spatial collinearity, i.e., the order of Hox expression domains along the rostro-caudal axis corresponds to their order on the chromosome (only ceh-13 breaks the collinearity rule)^61,62. Hox genes in C. elegans are expressed in various cell types of a specific region along the rostro-caudal axis of embryos and larvae⁶⁰. In addition to their critical roles during early patterning, recent C. elegans studies in touch and MNs uncovered that Hox genes are necessary for the final steps of neuronal development, i.e., the control of terminal identity^7,32,63–69. To date, Hox involvement in maintaining neuron identity in the adult has only been demonstrated in MNs; therefore, we focus our discussion on this cell type.

The C. elegans ventral nerve cord is composed of various subtypes of cholinergic MNs (DA, DB, VA, VB, VC, AS) (Figure 3A). Each subtype contains multiple neurons (DA = 9 cells, DB = 7, VA = 12, VB = 11, AS = 11, VC = 6), and their cell bodies intermingle along the A-P axis (Figure 3A). Four of the six C. elegans Hox genes are expressed continuously, from development through adulthood, in specific groups of MNs. The anterior Hox ceh-13 (Lab/Hox1) is not spatially restricted, and is expressed in most MNs along the rostro-caudal axis of the nerve cord (Figure 4A). On the other hand, the midbody Hox gene lin-39 (Scr/Dfd/Hox4-5) is selectively expressed in midbody MNs, mab-5 (Antp/Hox6-8) in MNs of the posterior half of the cord, and the posterior Hox gene egl-5 (Abd-B/Hox9-13) in the most posterior members of the DA and VA subtypes (DA9, VA12) (Figure 4A). The continuous expression of these four Hox genes, from embryonic through adult stages^32,63,66, suggests they may function as terminal selectors to establish MN terminal identity during development and maintain it in the adult.

Figure 4: Hox factors collaborate with terminal selectors to control MN identity along the rostro-caudal axis in C. elegans and mice.

A. Four of the six C. elegans Hox genes are expressed continuously from development through adulthood in specific groups of MNs, depending on their positions in the ventral nerve cord. Hox factors thus intersect with UNC-3 in a region-specific manner. B. Distinct Hox genes and UNC-3 collaborate in a region-specific manner to establish and maintain the spatial identities of MNs along the rostro-caudal axis of the C. elegans nerve cord. C. Hox factors and UNC-3 collaborate to regulate terminal identity genes in a feed-forward-loop (FFL): both Hox and UNC-3 directly activate terminal identity genes; Hox directly activates UNC-3 expression; Hox expression is amplified by positive autoregulation and limited by negative feedback from UNC-3. This interaction may serve to buffer terminal identity gene expression against fluctuations in the levels of either activator. Panels A-C modified from⁶⁰ D. Hoxc8 is required to maintain MN terminal identity in the mouse spinal cord. E. Hoxa5 is needed in the mouse brainstem to maintain neuronal terminal identity and function.

4.2. Hox genes collaborate with UNC-3 (Collier/Ebf) to establish MN terminal identity

The terminal selector unc-3 (Collier/Ebf) is expressed in all neurons of the DA, DB, VA, VB, and AS subtypes, but the expression of Hox genes (lin-39, mab-5 and egl-5) is spatially restricted, intersecting with unc-3 in a region-specific manner (Figure 4A). Genetic analyses revealed that lin-39 and mab-5 are required, like unc-3, in midbody MNs to establish the expression of various terminal identity genes (e.g., unc-17/VAChT, cho-1/ChT, del-1/ENAC)^32,63,66. Similarly, egl-5 is required in posterior (rostral) MNs, activating itr-1/ITPR, glr-4/GRIK4, mig-13/LRP12, and flp-18/FMRF⁶⁶. These findings point to an intersectional strategy controlling cholinergic MN terminal identity: distinct Hox genes and UNC-3 collaborate in a region-specific manner to establish the spatial (regional) identity of MNs along the A-P axis of the C. elegans nerve cord (Figure 4B).

A similar strategy may control the identity of spinal MNs in mice. Isl1, responsible for establishing spinal MN terminal identity⁷⁰, is not spatially restricted in its expression. Distinct Hox genes, however, are expressed in different MN subtypes located at different regions along the rostro-caudal axis of the mouse spinal cord⁵⁹. Future studies are needed to determine whether Isl1 and distinct Hox genes collaborate, in a region-specific manner, to establish and maintain MN terminal identity.

4.3. The midbody Hox gene lin-39 (Scr/Dfd/Hox4-5) functions as a terminal selector of cholinergic MN identity

Although mab-5 and egl-5 are involved in the establishment of terminal identity of midbody and posterior MNs, respectively⁶⁶, whether they are continuously required remains untested (Figure 4B). However, accumulating evidence strongly suggests that lin-39 (Scr/Dfd/Hox4-5) acts as a terminal selector of cholinergic MN identity in the midbody region of the C. elegans ventral nerve cord. First, LIN-39 ChIP-sequencing revealed that LIN-39, like UNC-3, binds directly at its terminal identity gene targets. Mutational analysis of LIN-39 binding sequences upstream of these loci confirmed they are functionally important^32,63, indicating LIN-39 regulates terminal identity genes via direct binding at their cis-regulatory regions. Second, like UNC-3, LIN-39 is continuously required to maintain the terminal identity of midbody MNs. Temporally controlled depletion of LIN-39 in adult C. elegans led to reduced expression of various MN terminal identity markers, indicating LIN-39 is not only necessary for establishment, but also maintenance, of MN identity^32,63. Altogether, LIN-39 acts as a bona fide terminal selector that works with UNC-3 to establish and maintain cholinergic MN identity in the midbody region (Figure 4B–C). The function of LIN-39 as a terminal selector extends to other MNs that do not express unc-3, i.e., the cholinergic VC neurons that control egg-laying³², suggesting widespread employment by other MN subtypes of a Hox-based strategy for the control of their terminal identity.

4.4. Homeostatic control of Hox gene expression in adult MNs

The terminal selector lin-39 positively autoregulates its expression. Transgenic reporters of lin-39 fail to be expressed in MNs of lin-39 null mutant animals³², and LIN-39 binds directly to its own locus, suggesting it autoregulates at the transcriptional level in MNs (Figure 4C). Consistent with this idea, CRISPR/Cas9-mediated disruption of LIN-39 binding sites in the lin-39 locus decreased lin-39 mRNA and protein levels. This manipulation also reduced expression of terminal identity genes (e.g., cho-1/ChT) in MNs. Importantly, inducible degradation of LIN-39 in adult MNs led to decreased lin-39 expression, suggesting that this Hox gene is continuously required to maintain its own expression³².

Hox expression levels must be tightly controlled to avoid detrimental effects during tissue patterning^71,72. In the C. elegans nerve cord, low levels of LIN-39 lead to a failure to maintain expression of terminal identity genes in MNs, whereas high LIN-39 levels result in locomotion defects⁶³. Hence, lin-39 expression levels must be tightly controlled in MNs. This is achieved through a two-component principle where Hox positive autoregulation is counterbalanced by negative UNC-3 feedback³²: UNC-3 direct binding to the lin-39 locus limits lin-39 mRNA levels, suggesting a direct negative effect on lin-39 transcription.

Thus, Hox gene expression in adult C. elegans MNs is under homeostatic control, with positive autoregulation counterbalanced by negative UNC-3 feedback (Figure 4C). Because Hox genes are expressed in adult neurons of the fly, mouse and human, homoeostatic mechanisms may be a conserved strategy to control Hox expression and therefore neuronal identity across metazoan nervous systems.

4.5. Hox and UNC-3 operate in a feed-forward loop to ensure maintenance of cholinergic MN identity

Recent work revealed an additional role for Hox in the control of C. elegans MN identity: lin-39 controls expression of the terminal selector unc-3 (Figure 4C). LIN-39 binds extensively at the unc-3 locus, and mutational analysis combined with quantification of unc-3 mRNA levels in lin-39 mutants strongly suggest that LIN-39 acts directly to activate unc-3 transcription³². In summary, the Hox gene lin-39 acts directly in MNs to control transcription both of terminal identity genes and of unc-3, which in turn also activates terminal identity genes, creating a feed-forward loop (FFL) (Figure 4C). This FFL may be viewed as a form of “redundancy engineering” to ensure robust and continuous expression of terminal identity genes, as their protein products are critical for neuronal function (e.g., neurotransmitter receptors, ion channels). For example, if fluctuations in unc-3 expression levels occur, terminal identity gene expression in MNs will not be significantly affected due to the presence of LIN-39. Similarly, a transient decrease in the expression of the upstream activator (LIN-39) will not immediately cause a corresponding decrease in the transcription of its target genes due to the presence of the intermediate activator (UNC-3). Thus, the LIN-39-UNC-3 FFL may operate in C. elegans MNs throughout life to ensure robust levels of terminal identity gene expression.

4.6. Continuous requirement for Hox gene activity in fly and mouse neurons

Like in C. elegans, Drosophila and mouse Hox genes are also expressed in the nervous system during late developmental and adult stages^73–79. Emerging evidence indicates that sustained Hox expression is required for maintaining neuron terminal identity in these systems as well, dovetailing C. elegans Hox studies.

One of the Drosophila Hox genes, Ultrabithorax (Ubx), remains expressed in adult fly cholinergic, glutamatergic, and dopaminergic neurons. RNAi-mediated knockdown of Ubx in adult Drosophila causes a decrease in tyrosine hydroxylase (TH) and SLC ion channel expression in dopaminergic neurons and also impacted flight muscle activity⁸⁰, revealing a post-developmental role for Ubx in maintaining dopaminergic neuron identity. Other known examples of Hox involvement in maintaining neuronal identity include: (a) Ubx, abdominal-A (abd-A) and Abdominal-B (Abd-B), which are continuously required to maintain the identities of peptidergic abdominal neurons⁸¹, and (b) Deformed (Dfd), which is required for feeding motor unit formation from its initial specification to the establishment of active synapses⁸².

In mice, RNA-sequencing of isolated spinal MNs at early post-natal stages revealed sustained expression of multiple Hox genes in brachial (Hoxc4, Hoxa5, Hoxc5, Hoxa6, Hoxc6, Hoxa7, Hoxc8), thoracic (Hoxd9), and lumbar (Hoxa10, Hoxc10, Hoxa11) MNs⁷⁴. MN-specific removal of Hoxc8 gene activity at early and late developmental stages resulted in significant downregulation of various terminal identity genes (Nrg1, Mcam, Pappa) and motor defects⁷⁴, indicating that Hoxc8 is continuously required for MN terminal identity and function (Figure 4D). A similar continuous requirement has been shown for Hox5 genes in mouse spinal MNs responsible for breathing⁷⁹. Lastly, tamoxifen-inducible depletion of Hoxa5 in the brainstem at postnatal stages resulted in downregulation of multiples terminal identity genes (e.g., Slc17a7/VGLUT1, and glutamate receptor subunits Grm4, Grin2c, Grid2) required for synaptic transmission and plasticity⁷⁷, demonstrating a continuous requirement for Hoxa5 in the postnatal brainstem (Figure 4E). Thus, accumulating evidence, from C. elegans to flies and mammals, support the idea that Hox genes are continuously required to maintain the identity and functionality of post-mitotic neurons.

5. POU homeodomain transcription factors establish and maintain neuronal identity in worms, flies, and mice

The POU family of transcription factors, named after its founding members (Pit1, Oct1, UNC-86), all share a homeodomain and a distinct POU domain, both essential for DNA binding⁸³. Here, we highlight recent evidence for POU-family proteins as important players in maintaining neuronal identity. For a comprehensive review on POU proteins in nervous system development, we refer the reader to a recent article by Leyva-Díaz et al.⁸⁴.

The C. elegans POU ortholog UNC-86 is known to prevent lineage reiterations during early development⁸⁵. In addition, UNC-86 acts as a terminal selector in diverse classes of neurons (13 classes of sensory neurons, 7 of interneurons, and 4 of MNs)⁸⁴. In each class, UNC-86 activates a distinct battery of terminal identity genes by cooperating with class-specific transcription factors. For example, UNC-86 collaborates with at least five transcription factors from different families to establish the identity of the HSN MN in the midbody⁸⁶, whereas it partners with CFI-1 (ARID3) to establish sensory neuron (IL2) identity in the head^40,87. unc-86 expression is also post-developmentally maintained via autoregulation. Post-embryonic removal of UNC-86 does not affect neuronal survival, but does cause loss of neurotransmitter identity in specific neuron classes: glutamatergic touch neurons (PLM) and two other sensory neuron classes (PVD, PHC), cholinergic neuron classes (URB, URX), serotonergic neuron classes (NSM, HSN), and peptidergic sensory neurons (PQR, BDU)⁸⁸.

Beyond C. elegans, the function of UNC-86 in controlling neuron terminal identity is conserved in flies and mammals. The fly ortholog of UNC-86, Acj6, is predominantly expressed in sensory neurons, and remains expressed in those neurons throughout life⁸⁹. Its function is well studied in olfactory neurons, where a single olfactory receptor (OR) gene must be expressed for odor discrimination. Acj6 is necessary for OR gene activation, but also for preventing ectopic OR expression⁹⁰. In the sea anemone Nematostella vectensis, the UNC-86 ortholog NvPOU4 is expressed in neural cell types, such as cnidocytes and sensory neurons, where it is also required for terminal differentiation⁹¹.

In mice, the orthologs of UNC-86, Brn3a, Brn3b, and Brn3c, are mostly expressed in sensory neurons (e.g., auditory, somatosensory)⁹². Brn3a function has been examined in mouse medial habenular (MH) neurons, where its expression persists throughout adulthood⁸⁸. When Brn3a is specifically depleted in adults, 50% of all MH neurons lose their glutamatergic identity (assessed by loss of VGluT1), and 50% of cholinergic ventral MH neurons lose their cholinergic identity (loss of ChAT). Other terminal identity features, such as ion channels and neuropeptides, are also lost, indicating that Brn3a is required to maintain various terminal identity features in adult MH neurons⁸⁸.

6. Control of dopaminergic neuron identity in C. elegans and mice

The eight dopaminergic neurons of C. elegans (ADEL/R, CEPDL/R, CEPV/R, PDEL/R) are used for mechano-sensation. They are defined by the co-expression of five terminal identity genes encoding proteins necessary for dopamine biosynthesis (cat-2/TH, cat-4/GTPCH, bas-1/AAAD, cat-1/VMAT, and dat-1/DAT). Through a combination of unbiased genetic screens and candidate gene approaches, the ETS-type factor AST-1, the Dlx/Distalless ortholog CEH-43, and the Pbx ortholog CEH-20 were identified as terminal selectors^93,94. They bind directly to the cis-regulatory regions of dopamine biosynthesis genes and co-regulate their expression. AST-1 acts in all eight dopaminergic neurons, whereas CEH-43 acts primarily in ADE and CEP neurons, and CEH-20 acts in PDE neurons. Hence, the activity of AST-1 intersects with CEH-43 and CEH-20 in distinct types of dopaminergic neurons, forming neuron type-specific combinations of terminal selectors. All three factors are also continuously expressed in adult C. elegans dopaminergic neurons^93,94. Transient rescue assays showed that ast-1 gene activity is continuously required to maintain the expression of dopamine biosynthesis genes⁹⁴. Similarly, animals carrying a hypomorphic allele of ceh-43 showed a progressive loss of dopaminergic marker expression⁹³.

Interestingly, the mouse orthologs of AST-1 (Etv1/ER81), CEH-43 (DLX2) and CEH-20 (PBX1) are also critical for the terminal differentiation of dopamine neurons in the mouse olfactory bulb^94–98. Both Etv1/ER81 and PBX1 are known to directly bind at least one dopamine pathway gene (Th) to promote its expression^96,98, and are also continuously expressed in olfactory bulb dopaminergic neurons throughout life, suggesting a conserved role for the C. elegans terminal selectors AST-1 and CEH-20 in maintenance of dopamine neuron identity.

7. Emerging roles for chromatin-modifying proteins in neuronal terminal identity

In addition to transcription factors, post-translational modifications of histones are critical for gene expression because they affect chromatin structure⁹⁹. Polycomb group (PcG) proteins, as well as histone methyltransferases and acetyltransferases, are key families of evolutionarily conserved proteins, whose activity affects chromatin structure. We highlight here recent findings implicating these proteins in the establishment and/or maintenance of neuronal identity in C. elegans and mice.

7.1. Polycomb proteins control neuronal terminal identity

Polycomb group (PcG) proteins play evolutionarily conserved roles in gene regulation during development^100,101. Polycomb-mediated gene regulation is provided by two multi-protein complexes, known as Polycomb repressive complex 1 (PRC1) and 2 (PRC2). PRC2 methylates histone H3 at lysine 27 (H3K27me3), a histone mark that leads to chromatin compaction at genes targeted for repression^102–104. In mice, conditional deletion of the PRC2 component Embryonic ectoderm development (Eed) in postmitotic midbrain dopaminergic neurons resulted in a loss of H3K27me3, followed by a progressive activation of genes associated with alternate fates (e.g., Gata6, Foxg1, Dlx), as well as a progressive loss of terminal identity gene expression¹⁰⁵. Importantly, this genetic manipulation did not affect dopaminergic neuron survival, but it did lead to behavioral deficits characteristic of Parkinson’s disease. Eed deletion in mouse serotonergic neurons yielded similar results¹⁰⁵. Interestingly, a separate study showed that PRC2 is required to maintain silencing of death-promoting genes in medium spiny neurons and Purkinje cells, uncovering a key role for PRC2 in protecting neurons against degeneration¹⁰⁶. Although these two studies did not employ adult-specific PRC2 inactivation, the findings nevertheless strongly suggest that PRC2 is required to maintain the identity and function of dopaminergic and serotonergic neurons in the adult mouse brain.

PRC1 can control gene expression positively or negatively. During mouse spinal MN development, genetic removal of a PRC1 component (Ring1) leads to increased chromatin accessibility and ectopic expression of various genes, suggesting MN development relies on PRC1-mediated gene repression¹⁰⁷. However, recent data suggest that PRC1 can also play a role in gene activation^103,108. In C. elegans, PRC1 affects the terminal identity of head interneurons (AIY) by controlling the expression levels of two terminal selector genes (ttx-3, ceh-10)¹⁰⁹. Genetic removal of PRC1 factors has no effect on the initiation of these genes, but causes a progressive loss of expression in late larval and adult stages. PRC1 plays a similar role in C. elegans dopaminergic neurons (ADE)¹⁰⁹. Altogether, this indicates that PRC1 factors are needed to ensure maintenance of terminal selector gene expression, a key step for maintenance of neuronal identity.

7.2. Histone methyltransferases and acetyltransferases involved in neuronal terminal identity

Accumulating evidence in C. elegans suggests that histone methylation is necessary for establishment of neuronal terminal identity. In the context of nerve cord MNs, mutations in H3K9 histone methyltransferases met-2 (SETDB1) and set-25 (SUV39H2) result in ectopic expression of terminal identity genes and transcription factors^40,110. These studies provided genetic evidence that H3K9 methylation, the epigenetic hallmark of gene repression¹¹¹, is necessary for MN terminal identity. On the other hand, recent work in the adult mouse brain showed that acetylation of lysine 27 in H3 (H3K27), a mark associated with active transcription, is necessary to maintain the terminal identity of forebrain and hippocampal neurons¹¹². Combined elimination in adult forebrain neurons of CBP and p300, two proteins with a lysine acetyltransferase catalytic domain, led to decreased H3K27 acetylation and downregulation of neuronal identity genes, as well as a rapidly progressing ataxia phenotype. Similarly, combined depletion of CBP and p300 in excitatory hippocampal neurons resulted in reduced expression of terminal identity genes^112,113. These findings demonstrate that CBP and p300 are required in mice to maintain excitatory neuron identity by preserving acetylation levels at neuron type-specific genes. Taken together, these recent studies in C. elegans and mice provide early evidence that post-translational modifications of histones (H3K9, H3K27) are essential for neuronal terminal identity.

8. OPEN QUESTIONS

This review highlights the functions of terminal selector-type transcription factors and chromatin modifiers in the control of neuronal terminal identity in C. elegans. Studies on fly and mouse orthologs of these C. elegans proteins uncovered deep functional conservation. Overwhelming evidence supports the idea that terminal selectors are not only required to establish during development, but also maintain throughout adult life, the identity of specific neuron types. Hence, initial establishment (during development) and maintenance (throughout life) of neuron type-specific identity relies upon the activity of the same transcription factors. However, a handful of studies showed that terminal selectors can also switch target genes – in the same neuron type – across different life stages^12,39,46. Overall, the transcriptional mechanisms that maintain neuronal identity are likely more dynamic than previously thought, and future studies on chromatin accessibility are needed.

More recent studies indicate that chromatin modifiers are also necessary for the establishment of neuronal terminal identity, but whether they are actively required for its maintenance remains poorly understood. An unresolved issue relates to the classic problem of specificity. Since chromatin modifiers are expressed broadly (often ubiquitously) and act on DNA in a non-sequence-specific manner, how do they exert such specific effects on gene expression in specific neuron types? One to-be-tested hypothesis is that specificity arises through interactions with terminal selectors, which do bind specific DNA sequences and can potentially recruit chromatin factors to the cis-regulatory region of terminal identity genes, enabling their activation or repression in a neuron type-specific manner. Further work is also needed to test another clinically relevant hypothesis, originally formulated by Deneris and Hobert¹, i.e., disruption of transcriptional programs that maintain neuronal identity can cause neurological conditions. Indeed, human mutations in terminal selectors, chromatin factors and their effector target genes are linked to neuropsychiatric and neurodegenerative conditions¹. Exciting new findings in mammalian serotonergic neurons further support this hypothesis; terminal selectors Pet-1 and Lmxb1 are required, in the adult, to protect long-distance axons and synapses against slow degeneration^11,12.