The C. elegans COE transcription factor UNC-3 activates lineage-specific apoptosis and affects neurite growth in the RID lineage

Apoptosis plays crucial roles in development. During both embryogenesis and postnatal development many neurons and glial cells are eliminated through programmed cell death. An inappropriate level of cell death, either too low or too high, has been associated with congenital nervous system anomalies, autonomic disorders and neurodegenerative diseases (Oppenheim, 1991). Studies in C. elegans have contributed significantly to our understanding of the molecular pathways that execute programmed cell death (Ellis et al., 1991; Horvitz et al., 1994; Denning et al., 2013). Owing to a largely invariant cell lineage and the fact that a majority of apoptotic events occur shortly after the final round of cell division during C. elegans development, apoptotic events are readily observed and quantifiable (Kimble and Hirsh, 1979; Sulston et al., 1983). Genetic screens have been carried out to isolate mutants that either fail to execute apoptosis or exhibit excessive apoptosis (Sulston and Horvitz, 1981; Ellis and Horvitz, 1986). These mutants led to the identification of components of the core cell death pathway, including the apoptotic activator EGL-1 (Conradt and Horvitz, 1998), the anti-apoptotic factor CED-9 (Hengartner et al., 1992), the caspase activator CED-4 (Yuan and Horvitz, 1992) and the pro-caspase CED-3 (Ellis and Horvitz, 1986). These cell death executors are conserved in many metazoans (Danial and Korsmeyer, 2004).

A fundamental question remains as to how animals achieve cell type-specific and temporal regulation of apoptosis. Current studies suggest that regulatory pathways triggered by developmental (or damage) signals converge on the activation of the core cell death machinery to specify the fate of particular cells (Conradt, 2009). For example, in the hermaphrodite, TRA-1 (a global transcriptional regulator of somatic sexual fate) inhibits egl-1 expression in the hermaphrodite-specific neurons (HSNs) (Conradt and Horvitz, 1999), but promotes egl-1 expression in the male-specific cephalic companion neurons (CEMs) through inhibiting the transcription factor CEH-30 (Nehme et al., 2010). Asymmetrically distributed in the neurosecretory motor neuron (NSM) and its sister cell, the C2H2 zinc finger transcription factor CES-1 blocks the expression of egl-1 in the NSM cell by competing against the basic helix-loop-helix (bHLH) transcription factors HLH-2 and HLH-3, which activate egl-1 expression (Hatzold and Conradt, 2008). Diverse lineage-specific regulators of apoptosis have also been identified in other cells: the caudal-type homeodomain transcription factor PAL-1 promotes the death of the tail-spike cell by activating ced-3 expression (Maurer et al., 2007); CEH-34 (Six family homeodomain protein) and EYA-1 (Eyes absent ortholog) promote the death of the sister cell of the M4 pharyngeal neuron (Hirose et al., 2010); the Hox proteins, including MAB-5 (Liu et al., 2006) and LIN-39 (Potts et al., 2009), are involved in programmed cell death in the P blast lineage that gives rise to postembryonic motor neurons.

The COE (Collier/Olf/EBF) family transcription factors share a DNA-binding domain that contains an atypical zinc finger domain, an IPT (immunoglobulin-like/plexin/transcription factor) domain that mediates both DNA and protein interactions, an HLH dimerization domain, and an unstructured C-terminal transcription-activation region (Hagman et al., 1995). COE family transcription factors include EBF1-4 in mammals, EBF2 and EBF3 in Xenopus, Ebf2 (Zcoe2) in zebrafish, Knot (Collier) in Drosophila and UNC-3 in C. elegans (Liberg et al., 2002). COE proteins function in the differentiation and specification of neurons (Garel et al., 1999), B lymphocytes (Lin and Grosschedl, 1995), adipocytes (Akerblad et al., 2002), osteoclasts (Kieslinger et al., 2005) and muscles (Crozatier and Vincent, 1999).

C. elegans UNC-3 plays diverse, lineage-specific roles in neuronal fate specification. UNC-3 is required for the postmitotic hindgut hypodermal Y cell to redifferentiate into a PDA motor neuron (Richard et al., 2011; Zuryn et al., 2014). UNC-3 promotes the identity of the ASI chemosensory neuron by activating the expression of ASI-specific genes and repressing alternative neuronal programs (Kim et al., 2005). In a fraction of cholinergic motor neurons, UNC-3 functions as the terminal selector for cholinergic fate (Kratsios et al., 2012) and further restricts the identity of the VA and VB subclass cholinergic motor neurons by repressing the VC-type subclass fate (Prasad et al., 2008). Loss of unc-3 also leads to modest axon defasciculation and pathfinding defects (Wightman et al., 1997; Prasad et al., 1998).

Here we report a previously unknown function of UNC-3 in regulating apoptosis and neurite growth in specific neuronal lineages. In unc-3 mutants, Pegl-1::GFP reporter expression was lost in the RID lineage associated with the survival of the RID sister cell, which undergoes apoptosis in wild-type animals. Moreover, UNC-3 interacts with histone acetyltransferase CBP-1/P300, and cbp-1 mutants exhibit the same RID phenotype as unc-3. Our results indicate that UNC-3, together with CBP-1, regulates a lineage-specific apoptosis event that ensures the specification of a single RID neuron.

The C. elegans nervous system consists of 302 neurons of 118 subtypes, each with distinct morphology, connectivity and position. One class, consisting of a single neuron, RID, is born during embryogenesis. The RID soma, which resides in the anterior dorsal ganglion, extends a single process that initially grows ventrally to join the ventral nerve cord, then loops back to the dorsal cord, followed by a posterior outgrowth along the dorsal nerve cord (Fig. 1B). The unipolar neurite of RID is functionally compartmentalized: the ventral/dorsal loop receives synaptic inputs from the ALA, AVB and PVC interneurons, and hence functions as a dendritic region; the anterior-posterior outgrowth resembles an axonal region, making synaptic outputs to hypodermis, muscles and other motor neurons (White et al., 1986).

To identify genes that regulate RID development, we performed a genetic screen using the fluorescent marker hpIs162 [Pceh-10::wCherry], which is expressed strongly in the RID, AIY, CAN neurons, an unidentified support cell, and weakly in the ALA neuron. We isolated two loss-of-function alleles of unc-3 that change the reporter expression pattern: xd86 (R176 to opal stop) and xd106 (W280 to opal stop) (Fig. 1A). In both alleles, ∼70% of animals harbor a pair of Pceh-10::wCherry-positive neurons where the single RID normally resides in wild-type animals (Fig. 1B,C). These neurons are positioned side by side and exhibit an identical neurite projection pattern. Similar to the RID neuron in wild-type animals, the neurites first project ventrally, but when they reach the ventral cord, instead of looping back dorsally, they project posteriorly along the ventral nerve cord and stop prematurely. Hence, these neurites are missing part of the dendrite loop and the axonal region (Fig. 1B,D,E). The canonical unc-3 loss-of-function allele e151 (W309 to opal stop) exhibits the same defects of RID duplication and neurite projection (Fig. 1C,D,E). xd86, which has the earliest stop codon and highest penetrance defects, was used in the remainder of our study.

All other cells that express the Pceh-10::wCherry marker, including AIY and CAN, appear normal with respect to soma number and axon projection in unc-3 mutants (Fig. 1B). Similarly, with a pan-neuronal marker, the majority of motor neurons appear unaffected with respect to number and dorsal projections, although they exhibit defective migration and defasciculation in the ventral nerve cord as reported previously (Wightman et al., 1997; Prasad et al., 1998) (data not shown). Together, these data suggest that unc-3 functions in a lineage-restricted manner to affect the number of RID neurons and the pattern of dorsal-ventral RID neurite guidance.

We next investigated the underlying cause of RID duplication and defective neurite projection. UNC-3 is a terminal selector for a fraction of cholinergic motor neurons and sensory neurons (Kim et al., 2005; Kratsios et al., 2012). We first examined whether the RID fate is altered in unc-3 mutants. We examined the expression of several terminal differentiation markers (kal-1, zig-5, ser-2 and lim-4) that are activated in various neuronal groups that overlap in RID (Tsalik et al., 2003). All are expressed in the pair of RID-like neurons in unc-3 mutants (Fig. 2A,D), suggesting that the RID fate is probably unaltered and that this neuron represents a duplication of RID fate. The neurite morphology of the RID-like neurons in unc-3 mutants partially resembles that of ALA and AVJ(H) neurons, raising the possibility that one or both of the RID-like neurons might also acquire some traits of ALA or AVJ(H). However, neither the ALA marker Pida-1::GFP (Zahn et al., 2004) nor the AVJ(H) marker Pflp-12::GFP (Kim and Li, 2004) was misexpressed in RID neurons in unc-3 mutants (Fig. 2B-D). Therefore, the RID-like neurons in unc-3 mutants do not exhibit obvious fate transformation.

Next, we examined whether UNC-3 is sufficient to induce RID fate in other neuronal lineages. Consistent with the role of UNC-3 as a terminal selector for cholinergic motor neurons (Kratsios et al., 2012), ectopic expression of UNC-3 in glutamatergic mechanosensory neurons (PVM and PLM) was sufficient to induce the expression of cholinergic markers Pacr-14::DsRed and Pacr-2::GFP (Fig. 2E-G). By contrast, ectopic expression of UNC-3 in PVMs did not activate the RID marker Pceh-10::wCherry (Fig. 2G). Together, these results indicate that, like the LIM homeodomain protein LIM-4, which is required for some but not all aspects of RID neuron differentiation (Tsalik et al., 2003), UNC-3 is required for several aspects of RID development, including correct number and some aspects of neurite guidance, but is largely dispensable for RID terminal differentiation.

Next, we address how the functional loss of UNC-3 leads to the development of an ectopic RID neuron. During embryogenesis, the RID lineage precursor ABalappaa divides to give birth of ABalappaaa, which dies, and ABalappaap, which is the RID precursor. Shortly after the division of ABalappaap, the posterior daughter ABalappaapp undergoes apoptosis, whereas the anterior progeny becomes RID (Fig. 3A). HAM-1, a protein with a winged helix DNA-binding motif, is required for the proper division of the RID precursor ABalappaap and subsequently activates apoptosis in one of its two progenies (Guenther and Garriga, 1996; Frank et al., 2005). In ham-1(n1438) mutants, 48% of worms have two RID neurons and 10% have three or four RIDs (Fig. 3B). The apoptotic activator EGL-1 and the pro-caspase CED-3 are both required for RID sister cell apoptosis, as 90% of egl-1 and 80% of ced-3 mutant animals show RID duplication (Fig. 3B). Unlike in unc-3 mutants, RID-like neurons in ham-1, egl-1 and ced-3 mutants all exhibit wild-type neurite projections.

Two apoptosis events occur in the RID lineage: the sister of the RID precursor and the sister of RID. Since unc-3 mutants show duplicated, but not three or four, RID neurons, it is possible that UNC-3 restricts RID number by activating apoptosis in the RID sister cell. Consistent with this possibility, unc-3(xd86);ced-3(n717) and unc-3(xd86);egl-1(n3082) double mutants did not exhibit significant enhancement in the penetrance of RID duplication when compared with either single mutant (Fig. 3B). In unc-3(xd86);ham-1(n1438) double mutants, however, 28% of animals have three or even four RIDs, as compared with 10% in ham-1(n1438) single mutants. A similar trend of enhancement was observed in egl-1(n3082);ham-1(n1438) double mutants (Fig. 3B). These results indicate that HAM-1 may affect the division of both precursor cells, i.e. ABalappaa and ABalappaap, in the RID lineage. The extra one or two RIDs in ham-1 mutants might result from the survival of ABalappaaa (the sister of the RID precursor) or its fate transformation to that of ABalappaap (the RID precursor). Thus, it appears that ham-1 affects RID number through affecting the division pattern of the RID lineage precursor, whereas unc-3 mutations cause RID duplication due to a failure to activate apoptosis in the RID sister.

We further examined whether UNC-3 plays a ubiquitous role in apoptosis in other neuronal lineages. The birth of RMEL/R, ASEL/R, RIML/R and ASIL/R, like that of RID, involves apoptosis of their sister cells after the final round of cell division. These neurons exhibit normal number and morphology in unc-3 mutants (Fig. 3C). There is no significant difference in total cell death corpse number in the comma or 1.5-fold stage embryo in unc-3 mutants as compared with wild-type embryos (data not shown). Taken together, our results indicate that UNC-3 affects apoptosis in a lineage-specific manner.

To examine where UNC-3 functions, we monitored the embryonic and postembryonic expression of a functional unc-3 fosmid reporter that contains a bi-cistronic GFP reporter at the C-terminus of the UNC-3 coding sequence (Fig. 4A,B). This reporter fully rescued the RID phenotypes of unc-3 mutants (Fig. 4C,D).

During embryogenesis, at the pre-comma stage, the GFP signal was first observed in the RID precursor (ABalappaap) before its final round of cell division. After the division, GFP signals were observed in both progenies. After the RID sister cell had undergone programmed cell death, the RID terminal differentiation marker Pceh-10::wCherry was observed in the surviving cell. Co-expression of the ceh-10 and unc-3 reporters in the RID neuron persisted into the larval stage (Fig. 4B). Similar to what has been reported previously (Prasad et al., 1998,, 2008; Kratsios et al., 2012), the unc-3 fosmid reporter also exhibited expression in cholinergic motor neurons (data not shown).

Since the unc-3 fosmid reporter was expressed prior to the ceh-10 reporter in the RID lineage, we further compared RID development in egl-1(n3082) and unc-3(xd86) mutants using a functional unc-3 fosmid reporter and one that carries the xd86 mutation, respectively. In both mutants, the two RID-like neurons appeared at a similar developmental stage, which suggests that the ectopic RID neuron in unc-3 mutants is unlikely to be caused by a reiterative defect (Fig. 4E). These results further rule out the possibility that the ectopic RID neuron in unc-3 mutants results from fate transformation from other lineages.

The expression of the unc-3 fosmid reporter in the RID lineage is consistent with a functional requirement of UNC-3 for RID development. Indeed, restoring UNC-3 expression in unc-3 mutants in all neurons using a pan-neuronal promoter (Punc-119), or in a restricted subgroup of neuron lineages including RID (Pceh-10), restored both the number and neurite morphology of the RID neuron. As a control, expressing UNC-3 in AIY neurons did not rescue the unc-3 mutant RID phenotype (Fig. 4C,D). This suggests a cell-autonomous role of UNC-3 in the RID lineage. In two unc-3 embryos in which one ectopic RID was ablated, both animals exhibited neurite trajectory defects in the surviving RID neuron (data not shown), further supporting a cell-autonomous role of UNC-3. Moreover, loss of the ectopic RID sister in unc-3 mutants by Pceh-10-driven UNC-3 expression suggests that UNC-3 is able to promote apoptosis in the RID sister cell even past the normal developmental time window.

The lineage-specific involvement of UNC-3 in apoptosis along with the restricted expression pattern of UNC-3 raise the question of whether UNC-3 could induce ectopic apoptosis in other lineages. The PLMs, which are a pair of mechanosensory neurons derived embryonically, are widely used to analyze ectopic apoptotic activity (Wang et al., 2002). When UNC-3 was ectopically expressed in these mechanosensory neurons, 95% of the transgenic animals lost Pmec-4::GFP reporter expression in one or both PLMs (Fig. 4F). The loss of reporter expression might be due to ectopic cell death or altered PLM fate. PLM marker expression was partially restored by the egl-1 or ced-3 mutation (Fig. 4F), suggesting that ectopic expression of UNC-3 was able to induce cell death. The incomplete suppression by egl-1 or ced-3 might reflect an altered fate of the surviving PLM neurons. Indeed, in some transgenic animals we noted the presence of ectopic Pacr-2::GFP-positive neurons where PLMs normally reside, consistent with their transformation to a cholinergic fate (Fig. 2F).

Several transcription factors promote or suppress cell type-specific apoptosis through regulating expression of the apoptotic gene egl-1 (Nehme and Conradt, 2008). We examined whether UNC-3 regulates apoptosis of the RID sister cell by affecting egl-1 expression. In 30 out of 32 wild-type embryos examined, a Pegl-1::GFP reporter (smIs89) was expressed in the RID sister cell shortly after the division of its precursor (Fig. 5A), which coincided with its apoptosis. By contrast, 28 out of 30 unc-3 embryos failed to activate the Pegl-1::GFP reporter in the RID lineage (Fig. 5A). The gross expression pattern of Pegl-1::GFP was similar in wild-type and unc-3 embryos, suggesting that unc-3 mutation does not globally block reporter expression. Together, these results indicate that UNC-3 is required for egl-1 expression in, and subsequent apoptosis of, the RID sister.

Since UNC-3 is a transcription factor, we explored the possibility that it directly activates egl-1 expression. We first sought to identify the minimal cis DNA element required for egl-1 rescue activity in RID sister cells. Because the Pegl-1::GFP reporter has a short promoter (∼1 kb) and long 3′ downstream region (∼6 kb), we generated egl-1 genomic fragments with various lengths of 3′ downstream region (Fig. 5B) and tested their ability to restore RID number in egl-1(n3082) mutants. The most significant reduction in rescuing activity occurred when a 1.2 kb 3′ downstream region was removed (Fig. 5B,D, compare constructs 1# and 5#).

Both mammalian COE proteins and C. elegans UNC-3 bind to a conserved COE binding motif (TCCCNN^G/_A^G/_A^G/_A^G/_A) to regulate target gene transcription (Wang et al., 1993; Kim et al., 2005; Kratsios et al., 2012). Within this 1.2 kb egl-1 fragment, we identified two putative COE binding sites (sites 1 and 2) and a less stringent binding site (site 3), which are conserved among the C. elegans, C. remaneii, C. briggsae and C. brenneri genomes (Fig. 5B,C). Deleting site 3 (construct 2#) did not affect rescue activity (Fig. 5B,D), whereas further removing a region containing site 2 (construct 3#) partially impairs rescue activity; moreover, point mutations that disrupt site 1 (construct 4#) further reduced rescue activity (Fig. 5B,D). Together, this suggests that putative COE binding sites 1 and 2 might act redundantly in mediating UNC-3-dependent egl-1 activation in the RID sister. To examine whether UNC-3 can physically bind these sites, we performed ChIP analysis with a FLAG-tagged full-length UNC-3 fusion protein. Compared with other regions tested, a strong binding signal was detected around sites 1 and 2 (Fig. 5E). Together, these results suggest that UNC-3 may regulate the expression of egl-1 in RID sister cells through multiple binding sites. We note that all our functional egl-1 constructs, including construct 1#, required multi-copy transgenic array for rescue (Fig. 5B; data not shown). Therefore, there might be other elements, regulated by UNC-3 or other transcription factors, that control egl-1 expression in the RID lineage.

If UNC-3 promotes apoptosis of the RID sister by activating egl-1 expression, then forced reactivation of egl-1 expression should bypass the requirement for unc-3 in the RID lineage. HLH-2 and HLH-3 promote egl-1 expression in the NSM neuron and its sister (Thellmann et al., 2003). Ectopic HLH-2 and HLH-3 expression in the RID lineage (Pceh-10::HLH-2 and Pceh-10::HLH-3) resulted in ∼6% RID loss, presumably due to ectopic RID cell death (Fig. 5F). Introducing the same transgene into unc-3 mutants decreased the two RID neuron phenotype from ∼60% to 40% (Fig. 5F). These results support the hypothesis that UNC-3 promotes egl-1 expression in the RID sister to promote its death.

In both egl-1 and ced-3 mutants, the duplicated RID neurons exhibit normal axon trajectories. This is in contrast to unc-3 mutants, in which the pair of RID neurons exhibit both dorsal-ventral and anterior-posterior neurite outgrowth defects (Fig. 1B). This suggests that, in addition to triggering EGL-1-mediated apoptosis in the RID lineage, UNC-3 plays roles in neuronal differentiation, probably through different effectors.

Several guidance signals, including Netrin/UNC-6 and its receptors DCC/UNC-40 and UNC-5, and UNC-129/TGFβ, are known to contribute to dorsal-ventral neurite guidance and cell migration (Hedgecock et al., 1990; Colavita et al., 1998). The RID neurite exhibited varying degrees of dorsal-ventral guidance defects in unc-6(ev400), unc-5(e53), unc-40(n324) and unc-129(ev554) single mutants, the combinatorial receptor mutant unc-5(e53);unc-40(n324) and the compound guidance cue mutant unc-6(ev400);unc-129(ev554), but none of these defects was as severe as those of unc-3(xd86) mutants (Fig. 6A,B). Therefore, additional mechanisms are involved in dorsal-ventral guidance of the RID neurite.

Ectopic expression of UNC-3 could alter the neurite projection pattern in a specific neuronal context. In wild-type animals the PVM axon migrates ventrally and, upon reaching the ventral nerve cord, it turns and migrates anteriorly. In more than 30% of transgenic animals ectopically expressing UNC-3 in PVM, the PVM axon first extended anteriorly or posteriorly, prior to growing ventrally. Upon reaching the ventral nerve cord, it often sent a branch posteriorly (Fig. 6C,D). Together, these data suggest that in the RID lineage UNC-3 not only determines the survival states during neurogenesis, but also regulates several aspects of neurite projection during neuronal differentiation.

Several UNC-3 co-factors have been identified in various developmental processes, including the histone demethylase JMJD-3.1 for Y cell to PDA redifferentiation and the C2H2 zinc finger protein PAG-3 for VA and VB motor neuron terminal differentiation (Prasad et al., 2008; Zuryn et al., 2014). RID number and neurite pattern were normal in pag-3(ok488), pag-3(n3098), jmjd-3.1(gk384) and jmjd-3.1(gk387) mutants (data not shown), indicating that UNC-3 may interact with other factors to regulate apoptosis and neurite outgrowth.

We performed a yeast two-hybrid screen to identify additional UNC-3 partners in apoptosis and/or neurite outgrowth. We uncovered UNC-3, consistent with a previous report that UNC-3 forms homodimers (Prasad et al., 2008). In addition, we identified CBP-1, a C. elegans histone acetyltransferase P300, which is required for DNA damage-induced apoptosis (Yang et al., 2009). CBP-1 is a large protein; the recovered cbp-1 clone encodes amino acids 1319-1635, including its two zinc finger domains (Fig. 7A). Using the GST pulldown assay, we confirmed that UNC-3 and the CBP-1 zinc finger region interact directly in vitro (Fig. 7B).

cbp-1 is essential for C. elegans survival, precluding an analysis of the RID phenotype in null mutants. To explore the role of cbp-1 in the RID lineage, we took two alternative approaches. First, we reasoned that if binding to UNC-3 is important for CBP-1 function, then overexpression of the UNC-3-binding region of CBP-1 may compete with wild-type CBP-1, causing a dominant-negative effect in wild-type animals. Indeed, overexpression of the CBP-1 zinc finger region using the Pceh-10 promoter in wild-type animals led to a low penetrance unc-3-like RID phenotype: ∼30% of transgenic animals displayed two RID-like neurons, with either ventral or lateral growing neurites (Fig. 7C,D). Second, cbp-1(ku258), which encodes a CBP-1 variant with two amino acid changes that may compete with wild-type CBP-1 for binding partners (Eastburn and Han, 2005), exhibits a similar RID phenotype to that caused by overexpression of the CBP-1 zinc finger region (Fig. 7C,D). By contrast, expression of the cholinergic marker Punc-17::GFP is unchanged in cbp-1(ku258) mutants (data not shown). Taken together, these results indicate that UNC-3 partners with an alternative C2H2 zinc finger protein, CBP-1, to promote lineage-specific apoptosis and neurite outgrowth in the RID lineage.

COE transcription factors have been implicated in the development and differentiation of multiple tissues. UNC-3, the single C. elegans COE family transcription factor, specifies the identity of cholinergic motor neurons and ASI sensory neurons, and promotes postembryonic epithelial to neuron transition in the Y cell (Kim et al., 2005; Prasad et al., 2008; Richard et al., 2011; Kratsios et al., 2012). Here we show that UNC-3, in partner with alternate co-factors, acts as a lineage-specific activator of apoptosis in the RID neuron lineage.

By interacting with conserved COE binding sites, COE family transcription factors activate or repress the expression of target genes. During B cell development, EBF1 regulates the expression of transcription factors Pax5, Pou2af1 and Foxo1, which are required for B cell differentiation (Zandi et al., 2008). EBF1 can also directly regulate the expression of a B cell differentiation marker, mb-1 (Hagman et al., 1991). Similarly, in C. elegans, UNC-3 directly activates terminal differentiation genes (e.g. unc-17, cho-1), as well as other transcription factors that indirectly inhibit the expression of terminal differentiation genes in a subset of motor neurons (Kratsios et al., 2012).

In the RID lineage, with the exception of several aspects of neurite trajectory, UNC-3 is largely dispensable for terminal differentiation. Its dominant role is to ensure apoptosis of the RID sister in order to prevent neuron duplication. An EGL-1 reporter that is transiently activated in the RID sister in wild-type animals fails to be activated in unc-3 mutants. This finding supports a notion that EGL-1, a shared pro-apoptotic protein, can be activated by multiple lineage-specific transcription factors, instead of by a shared factor in multiple lineages. In this regard, apoptosis might simply represent another lineage specification event that is subject to regulation by lineage-specific transcriptional codes.

During development, transcription factors can be recruited to multiple complexes, where they function with diverse co-factors to activate or repress the transcription of different sets of targets. This allows them to perform distinct functions in multiple cell types. These co-factors include chromatin remodelers, such as histone acetylases, deacetylases, kinases and methylases (Latchman, 1997). The COE transcription factor EBF1, in the presence of the zinc finger protein OAZ/Zfp423 or ZNF521/EHZF, recruits the nucleosome remodeling and histone deacetylase (NuRD) complex to genomic DNA. This leads to reduced accessibility to transcription factors and to the downregulation of target transcription in neuronal precursors and hematopoietic progenitors (Tsai and Reed, 1997; Liao, 2009; Mega et al., 2011). By contrast, in the absence of these co-factors free EBF1 binds the histone acetyltransferase (HAT) p300/CBP and represses its HAT activity, while DNA-bound EBF1 releases p300/CBP, resulting in upregulation of target gene transcription (Zhao et al., 2003). Similar to the partnership between mammalian CBP/p300 and EBF1, we found that CBP-1 is a partner of UNC-3 during RID development. The observation that the UNC-3 co-factors PAG-3 and JMJD-3.1 do not affect RID fate suggests that the lineage-specific function of UNC-3 might be specified through its partnership with different co-factors (Fig. 7E).

UNC-3 expression is observed in both progenies of the RID precursor, but only the RID sister undergoes apoptosis. Other UNC-3-expressing embryonic cells do not undergo apoptosis either. Therefore, although UNC-3 is required for apoptosis, it alone cannot dictate the life-death decision. Thus, there must be another factor(s) that is asymmetrically distributed or that functions differentially between the two progenies to coordinate the RID lineage-specific cell death. This factor either promotes apoptosis in the RID sister with UNC-3, or protects the RID neuron from activating apoptosis by inhibiting egl-1 expression. CBP-1 is unlikely to be this factor because, like unc-3, cbp-1 affects both RID and the RID sister.

UNC-3 and mammalian COE transcription factors share over 60% amino acid sequence identity at the DNA-binding (DBD), HLH and IPT domains. A mammalian COE gene, EBF3, is either deleted or transcriptionally silenced through methylation in some brain tumors. EBF3 overexpression results in apoptosis by repressing expression of the anti-apoptotic Bcl-2 family protein MCL1 and activating the cleavage of caspase 3 (Zhao et al., 2006). By contrast, another COE factor, EBF1, is required for B cell survival (Gyory et al., 2012). These findings suggest that mammalian COE transcription factors are likely to be involved in the process that determines whether cells survive or die.

Although the UNC-3 target genes responsible for RID neurite guidance/outgrowth remain to be determined, the role of COE genes in neurite guidance/outgrowth is likely to be conserved. Neurite defects have been associated with mutations in COE genes in other organisms. EBF1 mutation affects the response of facial branchiomotor neurons to the rhombomeric environment and leads to migration defects in these neurons (Garel et al., 2000). EBF1 inactivation also affects the pathfinding of thalamic and retinal ganglion neurons (Garel et al., 2002; Jin and Xiang, 2011). Inactivation of EBF2 and EBF3 leads to the failure of olfactory neurons (ORNs) to project to the dorsal olfactory bulb (Wang et al., 2004). The Drosophila COE transcription factor Knot is expressed in class IV multi-dendritic sensory neurons and is essential for the development of their elaborative dendritic arbors (Kieslinger et al., 2005; Hattori et al., 2007). C. intestinalis COE rescues the cholinergic marker expression defects of unc-3 mutants (Kratsios et al., 2012). Consistently, we found that in unc-3 mutants the anterior-posterior neurite outgrowth defect of RID (but not the soma number or dorsal-ventral guidance defect) was rescued by C. intestinalis COE (data not shown). In summary, these findings suggest that COE family protein function is conserved in regulating neurite outgrowth/guidance.

All strains were maintained on NGM plates at 22°C using standard methods (Brenner, 1974). The wild-type strain was Bristol (N2). Other strains used were as follows. LGI: unc-40(n324), hpIs162 [Pceh-10::wCherry+pRF4(rol-6)], zdIs5 [Pmec-4::GFP+lin-15(+)]; LGII: inIs179 [Pida-1::GFP], juIs76 [Punc-25::GFP]; LGIII: otIs264 [Pceh-36::tagRFP], ynIs82 [Pflp-12::GFP]; LGIV: cbp-1(ku258), ced-3(n717), ham-1(n1438), unc-5(e53), unc-129(ev554), hpIs202 [Pceh-10::GFP+lin-15(+)], otIs33 [Pkal-1::GFP]; LGV: egl-1(n3082), ntIs1 [Pgcy-5::GFP+lin-15(+)], sraIs49 [Pnmr-1::G-CaMP+unc-119(+)]; LGX: unc-3(e151), unc-3(xd86), unc-3(xd106), unc-6(ev400). Ex lines and strains without exact chromosome information were: hpEx2819 [Plim-4::LIM-4::GFP], hpEx3205 [Punc-3^fos::wCherry+Pmyo-3::RFP], hpEx3206 [Punc-3(m)^fos::wCherry+Pmyo-3::RFP], jxEx30[Pins-6::mCherry+Pofm-1::GFP], otEx536 [Pser2prom2::GFP], otEx4844 [Pacr-14(300bp promoter)::DsRed2+pha-1(+)], otIs11 [Pzig-5::GFP+pRF4(rol-6)], smIs89[Pegl-1::NLS::GFP], xdEx902 [Pmec-7::UNC-3+pRF4(rol-6)], xdEx1056 [Punc-119::UNC-3+Psur-5::GFP], xdEx1087 [Pceh-10::UNC-3+Psur-5::GFP], xdEx1091 [Punc-3^fos::GFP+Psur-5::RFP], xdEx1149[Pceh-10::CBP-1zinc], xdIs61[Punc-119::3xFLAG::UNC-3].

For the EMS screen, hpIs162 animals were treated with 48 nM EMS, and mutants with an abnormal RID marker expression pattern were isolated in the F2 generation. Approximately 6000 mutagenized haploid genomes were screened and 12 mutants were recovered. xd86 and xd106 were mapped to chromosome X. Since both showed uncoordinated and weak coiler phenotypes resembling unc-3 mutants, we sequenced the entire coding region of the unc-3 gene by amplifying PCR products from xd86 and xd106 animals. A C-to-T mutation at 4036 bp (R176stop) and a G-to-A mutation at 7958 bp (W280stop) downstream of the unc-3 ATG were found in xd86 and xd106, respectively (Fig. 1A).

To generate rescue constructs, the wild-type unc-3 cDNA fragment was RT-PCR amplified from N2-derived total RNA and inserted into the KpnI/EcoRI sites of pPD95.77 (kindly provided by Dr Andrew Fire, Stanford University, CA, USA). unc-119, ceh-10 and ttx-3 promoters were amplified from wild-type genomic DNA and inserted into the multiple cloning sites of pPD95.77. To generate FLAG-tagged UNC-3, 3xFLAG sequence was amplified and inserted into the KpnI site of the Punc-119::UNC-3 vector. The unc-3 cDNA was also inserted into pPD52.102 (kindly provided by Dr Andrew Fire) to generate plasmid Pmec-7::UNC-3. The fosmid-based reporter Punc-3^fos::GFP was constructed following standard protocols (Tursun et al., 2009). For the xd86-mimic mutant form fosmid reporter Punc-3(m)^fos::wCherry, the galK gene flanked by 50 bp of homologous sequence was used to replace 240 bp of the unc-3 locus surrounding the unc-3(xd86) mutation site by recombination. An appropriate PCR product was then amplified from xd86 genomic DNA and used to replace the galK sequence by recombination. The cDNA of HLH-2, HLH-3 or the CBP-1 zinc finger region was amplified from N2-derived cDNA and inserted into the ceh-10 promoter-containing vector to generate Pceh-10::HLH-2, Pceh-10::HLH-3 and Pceh-10::CBP-1zinc constructs, respectively. For egl-1 rescue fragments, various lengths of egl-1 genomic DNA were amplified from N2. Site-directed mutagenesis was performed to mutate specific nucleotides within the putative UNC-3 binding sites.

Transgenic lines were generated by microinjection of the indicated constructs or PCR products at 5-50 ng/μl together with transformation markers at 50-100 ng/μl. Co-injection markers were plasmid pRF4 (kindly provided by Dr Andrew Chisholm, University of California, San Diego, USA) containing the dominant marker rol-6(su1006), or one of the following fluorescent markers: Psur-5::RFP, Psur-5::GFP, Pmyo-2::RFP or Pmyo-3::RFP.

Young adult worms and embryos were mounted on dried 2% agarose pads and immersed in M9 buffer with or without 30 mM NaN₃. Images were taken using a Zeiss compound microscope or a Nikon Eclipse confocal microscope. To quantify the RID duplication defects, neurite guidance defects and fluorescent marker expression levels, young adult worms were mounted and then examined by fluorescence microscopy. For the dorsal-ventral guidance defects, we categorized worms as defective if either one of the two RIDs had a dorsal-ventral defect. For the anterior-posterior defects, we quantified the RID neuron with the longer axon.

The yeast transformation procedure was conducted using standard techniques. A full-length unc-3 cDNA was inserted into pBTM116 (containing the LexA DNA-binding domain for Leu selection; kindly provided by Dr Andrew Chisholm). An UNC-3 fusion protein with the DNA-binding domain was used as bait to screen a C. elegans mixed stage cDNA library. Yeast bearing both UNC-3 and the potential interactors were seeded onto –Leu –Trp –His plates containing 2.5 mM 3-amino-1,2,4-triazole (3AT). Clones grown on –Leu –Trp –His plates were further tested for their ability to produce β-galactosidase. Positive clones were analyzed by DNA sequencing.

The FLAG-tagged UNC-3 transgenic strain (xdIs61) was grown in a liquid culture at 20°C. Adult worms were bleached [1.5 M NaOH, 12% NaOCl (v/v)] to collect embryos. These embryos were treated with 2% formaldehyde in PBS for 30 min at room temperature to allow cross-linking, before the reaction was quenched with 125 mM glycine for 15 min. Treated embryos were washed four times with PBS containing 1 mM PMSF. The embryos were homogenized with a cordless motor and then resuspended in FA buffer [50 mM HEPES/KOH pH 7.5, 1% Triton X-100, 1 mM EDTA, 0.1% sodium deoxycholate, 150 mM NaCl, 1 mM PMSF, 20 mM DTT, Protease Inhibitor Cocktail (Roche)] and the chromatin was sheared to 100-1000 bp range by sonication (30% amplitude, 10 s on, 59 s off, repeat 35 times). Cellular debris was removed by centrifugation at 13,000 rpm (15,700 g) for 15 min at 4°C, and the supernatant was used for chromatin immunoprecipitation (ChIP). The samples were pre-cleared with protein A- and protein G-agarose beads (Millipore), then incubated with anti-FLAG antibody (Sigma) at 4°C overnight prior to a 2 h incubation with protein A- and protein G-agarose beads. Bound DNA in the beads was eluted and purified using a DNA purification kit (Dongsheng Biotech), then quantified by real-time PCR using the indicated egl-1 primer sets. Results were plotted as fold-enrichment of anti-FLAG immunoprecipitation versus input. The following primers were used: p1f, 5'-cggtagcctaaggaatggaa-3′; p1r, 5′-acgtggatggcattgttaga-3′; p2f, 5′-tggggagacaagtcaagaca-3′; p2r, 5′-aattgtgcgtacccaatctg-3′; p6f, 5′-gtaacgggcaaatagcgggc-3′; and p6r, 5′-accttttgaacgtcgtttcc-3′. Sequences of other primer sets can be found in Table S1 in the supplementary material. Primer set p6 was chosen as an internal control.

The cbp-1zinc cDNA corresponding to amino acids 1320-1666 of CBP-1 was cloned into pET28c vector (Novagen) to generate an N-terminal His tag fusion. Full-length unc-3 was cloned into pGEX-4T-1 vector (GE Healthcare) to generate an N-terminal GST tag fusion. These constructs were expressed in E. coli BL21(DE3) cells with 0.5 mM IPTG at 18°C. The cells were collected and resuspended in lysis buffer [1× PBS, 1% NP40, 1 mM PMSF, 20% glycerol (v/v)], and then lysed by sonication. His-tagged CBP-1zinc protein with GST-tagged UNC-3 or GST alone was incubated with glutathione-Sepharose beads (GE Healthcare) in lysis buffer at 4°C overnight. The beads were washed with lysis buffer three times, and then boiled for 10 min in SDS-PAGE sample loading buffer. The CBP-1zinc protein was detected by His antibody (1:4000, Abcam).

We thank Drs Oliver Hobert, Xiaochen Wang, Chonglin Yang and the Caenorhabditis Genetics Center for strains and reagents; Drs Guangshuo Ou and Yongping Chai for live image analysis; Dr Helen McNeill for use of the image facility; and Dr Chonglin Yang for critical reading of the manuscript.