Summary
Terminal differentiation generates the specialized structures and functions that allow postmitotic cells to acquire their distinguishing characteristics. This process is thought to be controlled by transcription factors called “terminal selectors” that directly activate a set of effector genes. Using the mechanosensory touch receptor neurons (TRNs) in Caenorhabditis elegans, we extend this concept by identifying non-selector regulators of cell fate, which promote TRN fate indirectly by safeguarding the stability and activity of the selectors. In particular, the ZEB family transcriptional repressor ZAG-1 promotes TRN fate by inhibiting the TEA domain transcription factor EGL-44 and its binding partner EGL-46. EGL-44 and EGL-46 inhibit the TRN fate and simultaneously induce the fate of the multidendritic nociceptor FLP neurons. Since TRN and FLP share the same terminal selectors and a common ground state, the mutual inhibition between ZAG-1 and EGL-44 forms a negative feedback loop to regulate the choice between the two neuronal fates.
Introduction
The terminally differentiated state or cell fate of neurons distinguishes them from other neurons through specialized features and functions and the expression of effector genes (“terminal differentiation genes”) (Hobert, 2011). Although individual effector gene may not be expressed in a cell type-specific manner, the collective expression of a battery of terminal differentiation genes serve as a “signature” for the cell fate. For example, the cell fate of mammalian photoreceptor is defined by the expression of over 600 genes, including photoreceptor-specific retinol-binding proteins, rhodopsin, G protein Gnat1, phosphodiesterase, and others that are directly involved in light detection (Hsiau, et al., 2007; Blackshaw, et al., 2001).
The expression of the terminal differentiation genes is thought to be activated by terminal selectors (mainly transcription factors) that act alone or in combination (Hobert, 2011). Examples of such selectors are the C2H2 zinc finger transcription factor CHE-1, which determines the fate of gustatory ASE neurons in C. elegans (Etchberger, et al., 2007), the homeodomain transcription factor Crx, which controls photoreceptor fate (Corbo, et al., 2010), and the C4 zinc finger transcription factor Nurr1 and the homeodomain protein Pitx-3, which cooperatively direct the differentiation of mouse midbrain dopaminergic neurons (Martinat, et al., 2006). Terminal selectors directly up-regulate the transcription of various terminal differentiation genes through common cis-regulatory elements (Hobert, 2016; Hobert, 2011); thus, removing the selector genes leads to the loss of the expression of differentiation genes and the loss of neuronal fates.
Two problems arise in considering this model for differentiation. First, not all transcription factors that determine cell fate, however, can be classified as “terminal selectors”, because some may indirectly regulate terminal differentiation genes. Although the loss of such non-selector regulators also causes the absence of a particular cell fate, the mechanism underlying their action is not well understood. Second, the same terminal selectors are often expressed in several distinct types of neurons and are required for their differentiation into distinct fates, yet they only promote one cell fate in one particular neuron (Hobert, 2016). Such shared selectors presumably work in coordination with other fate-specific selectors. Although the overlapping expression pattern of many transcription factors could give rise to unique combinatorial code for each cell fate, the mechanism that restricts the function of shared terminal selectors during the specification of particular fate is not clear.
To address these issues, we have studied the choice of cell fate that generates either the touch receptor neurons (TRN) or the FLP neurons in C. elegans. The six TRNs are mechanosensory neurons that detect gentle mechanical stimuli along the body (Chalfie and Sulston, 1981), and the two FLP neurons are multidendritic nociceptors that sense harsh touch, noxious temperature, and humidity (Russell, et al., 2014; Chatzigeorgiou and Schafer, 2011; Chatzigeorgiou, et al., 2010; Kaplan and Horvitz, 1993). The heterodimer of the POU homeodomain transcription factor UNC-86 and the LIM homeodomain transcription factor MEC-3 acts as a terminal selector that promotes TRN cell fate (Xue, et al., 1993). Specifically, the heterodimer activates a set of TRN terminal differentiation genes (the mechanosensory channel genes mec-4 and mec-10, the tubulin genes mec-7 and mec-12, the tubulin acetyltransferase gene mec-17, and others) by binding to conserved regulatory elements in their proximal promoters (Zhang, et al., 2002; Duggan, et al., 1998).
In addition to directing the differentiation of the TRNs, MEC-3 and UNC-86 are expressed in and needed for the differentiation of two pairs of multidendritic neurons, the embryonic and anterior FLP neurons and the postembryonic and posterior PVD neurons. The expression of unc-86 and mec-3 in these cells does not lead to the normal expression of many TRN terminal differentiation genes, resulting in cells whose fate is very different from the TRNs (Finney and Ruvkun, 1990; Way and Chalfie, 1988). Our previous work suggested that the TEA domain transcription factor EGL-44 and the zinc-finger protein EGL-46 in FLP neurons prevented these cells from acquiring the TRN fate (Wu, et al., 2001). Thus, the active inhibition of TRN fate by EGL-44 and EGL-46 in FLP neurons distinguished these two types of neurons that express common selector genes. However, whether EGL-44 and EGL-46 also activate a FLP-specific genetic program is unclear.
In this paper, we extend this model of TRN and FLP neuronal determination, by asking whether other transcription factors help promote the TRN fate and cause the cells to differentiate away from the FLP fate. We screened for the effects on TRN development of the loss of 392 transcription factors using RNA interference and identified LDB-1 and ZAG-1 as positive TRN fate regulators. The LIM domain-binding protein LDB-1 binds to MEC-3 protein and stabilizes it post-transcriptionally, thus contributing to the activation of both TRN and FLP fate. The Zn-finger homeodomain protein ZAG-1 is expressed in TRNs but not FLP neurons. ZAG-1 promotes the TRN fate not by directly activating the TRN terminal differentiation genes but by preventing egl-44 and egl-46 expression, which blocks the activation of TRN genes by UNC-86/MEC-3. In FLP neurons, EGL-44/EGL-46 complex simultaneously inhibits the TRN genes and activates FLP genes. EGL-44/EGL-46 also represses the expression of zag-1 in FLP neurons, and this negative feedback loop establishes a bistable switch between TRN and FLP fates. Our work suggests that UNC-86/MEC-3 serves as a ground-state selector resulting in a common state in both TRNs and FLP neurons, and that individual fates are subsequently controlled by ZAG-1 and EGL-44/EGL-46 acting in a bistable switch. In particular, ZAG-1 regulates the choice of cell fate by inhibiting fate inhibitors and safeguarding the activation of terminal differentiation genes.
Results
An RNAi screen for transcription factors affecting TRN cell fate
The anterior ALML and ALMR neurons and the posterior PLML and PLMR neurons are two pairs of bilaterally symmetric TRNs generated in the three-fold embryo, i.e., late in embryonic development. Two other TRNs, the AVM and PVM neurons, arise post-embryonically (Chalfie and Sulston, 1981; Sulston and Horvitz, 1977). To search systematically for transcription factors specifying TRN fate, we knocked down the expression of transcription factor genes using RNAi and looked for animals that no longer expressed mec-17p::RFP, a TRN marker (see STAR Methods for details). Using RNAi against unc-86 as a positive control, we tested various genetic backgrounds that were previously found to enhance the effects of RNA interference and found that eri-1; lin-15B mutants had the highest penetrance for the loss of RFP expression. About 80% of these animals did not express RFP in either of the two ALM neurons when treated with unc-86 RNAi (Figure 1A and B). Since the posterior TRNs (PLM neurons) were less affected by the RNAi treatment (Figure 1B), we focused on the disappearance of mec-17p::RFP expression in the ALM neurons in the screen.
Among the 443 bacterial clones expressing dsRNA against 392 transcription factors and associated proteins (Table S1), we identified 14 genes that were required for the expression of TRN markers (Figure 1). Four genes (unc-86, mec-3, ldb-1, and ceh-20) were previously known to affect the expression of TRN terminal differentiation genes. We examined null mutants for all the remaining ten genes but could only confirm the loss of mec-17p::RFP expression in zag-1 mutants.
We were surprised that RNAi against nine genes affected TRN expression through six rounds of testing but loss-of-function mutations in these genes did not. These false positive results are unlikely to result from the specific genetic background of the RNAi strain, since mutation of several of the genes (zip-4, hmbx-1, and nhr-119) in eri-1; lin-15B animals did not affect TRN fate. Activating the RNAi pathway non-specifically by RNAi against GFP in those triple mutants (e.g. zip-1; eri-1; lin-15B mutants) did not cause the loss of mec-17p::RFP expression either. Therefore, we reason that the discrepancy between the RNAi and mutant phenotypes was likely due to mistargeting of the dsRNAs.
The LIM domain-binding protein LDB-1 promotes TRN fate by stabilizing MEC-3
ldb-1 encodes the only C. elegans homolog of nuclear LIM domain-binding protein NLI/Ldb1/CLIM2 and binds to LIN-11 and MEC-3, two of the founding members of the LIM homeodomain domain protein family (Gupta, et al., 2003; Cassata, et al., 2000; Freyd, et al., 1990; Way and Chalfie, 1988). Cassata et al.(2000) found that RNAi against ldb-1 eliminated the expression of TRN gene mec-2 without affecting mec-3 expression. We found that an ldb-1 null mutation, which caused early larval arrest, eliminated the expression of several TRN markers (mec-4p::GFP, mec-7p::GFP, mec-17p::RFP, and mec-18p::GFP) in both ALM and PLM neurons in those arrested L1 animals (Figure 2A and B). The loss of ldb-1 also eliminated the expression of FLP fate marker sto-5p::GFP in FLP neurons (Figure S1A). Thus, ldb-1 is required for the specification of TRN and FLP fates, both of which depend on MEC-3 as a terminal selector. We found that the expression of mec-3p::RFP reporter persisted in ldb-1 mutants (Figure 2C), but quantitative measurements of mec-3 mRNA levels using single molecule fluorescent in situ hybridization (smFISH) revealed that mec-3 transcription was moderately reduced in ldb-1 mutants compared to the wild type animals (Figure 2D). Therefore, our data suggest that LDB-1 is not only required for the activation of MEC-3 target genes but is also needed for optimal mec-3 expression. This result is consistent with the known mec-3 autoregulatory activity (Topalidou, et al., 2011; Xue, et al., 1992).
The physical interaction between LDB-1 and MEC-3 was confirmed using a yeast two-hybrid assay (Figure S1B-D). This interaction presumably enables MEC-3 to activate downstream TRN differentiation genes in a manner similar to the function of the mouse LDB-1 homologs NLI and CLIMs, which potentiate the transactivation by LIM-homeodomain proteins, like Lhx3 and Isl1 (Bach, et al., 1997; Jurata, et al., 1996). Although LDB-1 is not needed for the DNA binding capacity of MEC-3 at least in vitro (Xue, et al., 1993), the activation of downstream genes by MEC-3 may depend on LDB-1 in vivo.
Interestingly, the binding of MEC-3 to UNC-86 was very weak compared to the MEC-3/LDB-1 interaction in our yeast two-hybrid system (Figure S1D), which was unexpected given earlier results of electrophoretic mobility shift assay (EMSA) that found stable association of UNC-86/MEC-3/DNA complex (Xue, et al., 1993). Our results suggest that the UNC-86/MEC-3 heterodimer is not stable without the target DNA. Formation of this heterodimer may mostly occur on the promoter of TRN genes, where the binding sites for UNC-86 and MEC-3 are adjacent.
We also found that the strong association of LDB-1 with MEC-3 is required for the protein stability of MEC-3; unlike the mec-3 promoter reporter, the expression of mec-3::GFP translational fusions was strongly reduced in ldb-1 mutants (Figure 2E). These results suggest that the MEC-3 protein is highly unstable in the absence of LDB-1. In fact, treatment with the proteasome inhibitor MG-132, which reduces the degradation of ubiquitin-conjugated proteins, significantly increased the level of MEC-3::GFP protein in ldb-1 mutants (Figure 2F), supporting the idea that LDB-1 stabilizes MEC-3 and prevents its degradation. Similarly, the Drosophila LDB-1 homolog Chip binds to and stabilizes the LIM-homeodomain protein Apterous, and the dissociation of Chip from Apterous triggers its degradation (Weihe, et al., 2001). These results suggest a common function for the nuclear LIM interactor (Ldb/NLI/Clim) in regulating LIM-homeodomain protein homeostasis.
We confirmed the results of Cassata et al. (2000) that ldb-1 is expressed in many neurons, vulval cells, and some muscle cells, but also found that it was expressed in the TRNs in a mec-3-dependent fashion (Figure 2G). This latter result suggests that MEC-3 positively regulates its own activity and stability by activating the expression of its binding partner. Thus, MEC-3 autoregulation occurs at both the post-translational and functional levels.
ZAG-1 is required for the expression of TRN fate markers independently of MEC-3
zag-1 encodes the sole C. elegans ZEB transcription factor. Human ZEB transcription factors, which like ZAG-1 have a homeodomain flanked by clusters of C2H2-type zinc fingers, induce the epithelial to mesenchymal transition (EMT) and are essential for normal embryonic development (Vandewalle, et al., 2009). Mutations in human ZEB genes cause defects in neural crest development and are linked to malignant tumor progression (Vandewalle, et al., 2009). C. elegans zag-1 prevents the posterior post-embryonic TRN neuron PVM from adopting the morphology of the multidendritic nociceptor neuron PVD, since a hypomorphic zag-1 allele (zd86) led to the development of dendritic arbors in PVM (Smith, et al., 2013). Here, we found that a zag-1 null mutation (hd16) led to developmental arrest at the L1 stage (AVM and PVM do not arise until late in the L1 stage) and the loss of expression of all the tested TRN fate markers in ALM and PLM neurons (Figure 3A-C). Thus, the role of ZAG-1 is important for general TRN fate specification.
We next tested the genetic interaction between zag-1 and mec-3 and found that the zag-1 (hd16) null allele did not affect the expression of the mec-3 using a transcriptional reporter (Figure 3A), smFISH (Figure 3B), or MEC-3::GFP translational fusions (Figure 3C). In addition, expression of zag-1(+) from the mec-3 promoter restored the expression of the TRN fate marker mec-17p::RFP in zag-1 mutants, suggesting that ZAG-1 induces the expression of TRN terminal differentiation genes cell-autonomously. Since mec-3 expression is completely dependent on unc-86, we expected and found that the expression of unc-86 was not changed in zag-1 mutants.
Using a fosmid-based GFP translational fusion, we found that zag-1 was expressed in the six TRNs but not FLP and PVD neurons (Figure 3D), and the expression of zag-1 in TRNs was not affected by mutations in mec-3 (Figure 3E). Therefore, zag-1 and mec-3 are transcriptionally independent of each other. We also failed to find any physical interaction between ZAG-1 and MEC-3 in yeast two-hybrid assays (Figure S1B-D). Together, our results suggest that ZAG-1 promotes TRN fate independently of the previously identified TRN fate determinants UNC-86 and MEC-3; the expression of the three transcription factors only overlap in the TRNs and thus can form a unique combinatorial code for TRN fate.
Smith et al. (2013) found that another conserved transcription factor AHR-1 (aryl hydrocarbon receptor) controls the differentiation of the anterior post-embryonic TRN neuron AVM; in ahr-1 null mutants, AVM cells adopted a PVD-like multidendritic shape. We found that although expressed in all six TRNs, ahr-1 is only required for the expression of TRN markers in AVM neurons but not in the other TRNs (ALMs, PLMs, and PVM), suggesting that AHR-1 plays a subtype-specific role in TRN fate specification (Figure S2A-C). Thus, our findings disagree with the hypothesis presented by Smith et al. (2013) that AHR-1 and ZAG-1 function in parallel to specify TRN morphology in the AVM and PVM, respectively. Instead, ZAG-1 is required for TRN fate adoption in general, whereas AHR-1 functions only in AVM. Moreover, Smith et al. (2013) found that AHR-1 specified TRN fate in AVM by regulating mec-3, whereas we found that ZAG-1 promoted TRN fate independently of mec-3.
ZAG-1 safeguards TRN fate by suppressing TRN fate inhibitors EGL-44/EGL-46
The TEA domain transcription factor EGL-44 and the Zn-finger protein EGL-46 inhibited the TRN fate in the FLP neurons (Wu, et al., 2001). These proteins appeared to work together to regulate gene expression and physically interacted in yeast two-hybrid assays (Figure S3; the EGL-44 and EGL-46 complex is hereafter abbreviated as EGL-44/EGL-46). Both genes were normally expressed in the FLP neurons, and the expression of egl-46 was dependent on egl-44 (Figure 4A and B). Our lab previously found and we confirmed that mutations in egl-44 and egl-46 caused the ectopic expression of mec-17p::GFP and other TRN reporters in FLP neurons (Figure 4A and C; Table S2). egl-44 and egl-46 were not expressed in differentiated TRNs (Figure 4A and B).
Since zag-1 was selectively expressed in TRNs but not FLP neurons, we tested if ZAG-1 promoted the TRN fate by preventing the activation of EGL-44/EGL-46 in these cells. Consistent with this hypothesis, GFP reporters for both egl-44 and egl-46 were ectopically expressed in the TRNs in zag-1 mutants (Figure 4D). In addition, egl-44 mRNA, as measured by smFISH, was increased in zag-1 TRNs, an indication that ZAG-1 transcriptionally repressed egl-44 (Figure 4E). Importantly, egl-44 and egl-46 are epistatic to zag-1, since mutations in them restored the expression of TRN fate markers in zag-1-deficient animals (Figure 4C and Table S2). Thus, the TRN cell fate did not require ZAG-1 in the absence of EGL-44/EGL-46. Instead, ZAG-1 promoted the TRN fate through a double inhibition mechanism by preventing the expression of the TRN fate inhibitors egl-44 and egl-46.
We also found that ZAG-1 loss affected general neurite growth and guidance in TRNs, since egl-44; zag-1 and zag-1; egl-46 mutants showed shortened and misguided TRN neurites. These results extend those of Smith et al. (2013), who found that ZAG-1 inhibited dendritic growth in PVM neurons, and are consistent with the function of ZAG-1 in regulating axon guidance in other neurons (Clark and Chiu, 2003; Wacker, et al., 2003).
Misexpression of ZAG-1 converts FLP neurons into TRN-like cells
To address whether zag-1 could affect FLP differentiation, we misexpressed zag-1 in FLP neurons using the mec-3 promoter. This misexpression led to the expression of mec-17p::GFP and other TRN markers (Figure 5A and Table S2) and diminished expression of egl-44 and egl-46 (Figure 5A and B) in FLP neurons. These data suggest that misexpressed ZAG-1 converts FLP neurons to a TRN-like fate by inhibiting egl-44 and egl-46. The expression of mec-3p::zag-1 transgene also activated TRN markers in PVD neurons (Figure 5A), which is consistent with previous findings that ZAG-1 prevents PVM neurons from adopting a PVD-like morphology (Smith, et al., 2013). Thus, ZAG-1 not only prevents PVM neurons from taking on the PVD fate but can also turn PVD neurons into TRN-like cells. Since egl-44 was not expressed in PVD neurons, the misexpressed ZAG-1 presumably directs this conversion of cell fate by inhibiting some unidentified factor(s) that normally suppresses the TRN fate in PVD neurons.
We also expressed ahr-1 from the mec-3 promoter and found that misexpressed AHR-1 could activate TRN fate markers in PVD but not FLP neurons (Figure S2D-E). Overexpression of AHR-1 in TRNs also caused morphological defects, such as the growth of an ectopic ALM posterior neurite (Figure S2D), which does not occur when ZAG-1 was overexpressed. These results further support the hypothesis that ZAG-1 and AHR-1 have different functions.
EGL-44/EGL-46 prevents zag-1 expression in FLP neurons
Given the mutually exclusive patterns of zag-1 and egl-44/egl-46 expression in FLP neurons and TRNs, we next tested whether EGL-44/EGL-46 regulated zag-1 expression.
Mutations in egl-44 and egl-46 resulted in ectopic expression of a zag-1::EGFP reporter in FLP neurons. Misexpression of egl-44 from the mec-3 promoter is sufficient to suppress the TRN fate, because it activates the endogenous egl-46, a cofactor that is required for EGL-44 functions (Figure 5B and D; Wu, et al., 2001). We found that EGL-44 misexpression reduced, but did not completely eliminate, the expression of zag-1 in TRNs (Figure 5B and C). Therefore, the positive TRN fate regulator ZAG-1 and the negative regulator EGL-44/EGL-46 reciprocally inhibit each other’s expression, and their expression must be regulated so one or the other predominates.
Simultaneous misexpression of both EGL-44 and ZAG-1 from the mec-3 promoter, however, led to the activation of TRN marker mec-17p::GFP in all mec-3-expressing neurons (TRNs, FLP, and PVD; Figure 5E), because ZAG-1-mediated repression of endogenous egl-46 blocked the effects of EGL-44. In fact, supplying EGL-46 using the mec-3 promoter removed the dominance of ZAG-1 and turned off the expression of TRN markers again (Figure 5F). Moreover, expression of an egl-46::GFP reporter was reduced by ZAG-1 despite the presence of misexpressed EGL-44 (Figure 5B), suggesting that ZAG-1 can repress egl-46 transcription both through egl-44 and independently of egl-44. Functionally, animals carrying the uIs211[mec-3p::egl-44] transgene were completely insensitive to gentle touch; co-expression of zag-1 from mec-3 promoter could restore the sensitivity in animals where the TRN markers were reactivated but failed to do so when mec-3p::egl-46 was expressed (Figure S4A). Thus, the activity of EGL-44/EGL-46 complex predominates over ZAG-1 when both expressed from the same heterologous promoter in the same cell, leading to the suppression of TRN fate. The action of these competing inhibitors is not likely through direct interaction, since we did not observe any physical interaction of ZAG-1 with either EGL-44 or EGL-46 or of MEC-3 with EGL-44, EGL-46, and ZAG-1 in yeast two-hybrid assays (Figure S3).
EGL-44/EGL-46 and ZAG-1 regulate a switch between FLP and TRN fates
We next investigated how EGL-44/EGL-46 and ZAG-1 regulated FLP fate using sto-5 (Topalidou and Chalfie, 2011) and several other genes (dma-1, bicd-1, and flp-4) that were expressed in FLP neurons (Aguirre-Chen, et al., 2011; Liu and Shen, 2011; Kim and Li, 2004) but not the TRNs (Figure 6A). Expression of all four genes in the FLP neurons depended on EGL-44 and EGL-46 (Figure 6B and C). Moreover, the expression of the FLP and TRN markers were mutually exclusive; FLP neurons in egl-44 and egl-46 mutants or animals carrying mec-3p::zag-1 transgene never showed mixed expression of the FLP- and TRN-specific genes (Figure 6E). Mutations in zag-1 or the misexpression of egl-44 in TRNs activates the FLP markers in addition to turning off the TRN markers (Figure 6B and S5), and we did not observe any mixed expression of both types of markers under those conditions (Figure 6E). Since FLP markers were not expressed in egl-44; zag-1 double mutants, our results suggest that EGL-44/EGL-46 not only repressed the TRN fate but also promoted the FLP fate. Thus, the action of EGL-44/EGL-46 and ZAG-1 results in two mutually exclusive fates.
This bistable switch affected neuronal morphology as well as transcriptional reporters. For example, ALM neurons that assumed a FLP-like fate because of egl-44 misexpression grew an ectopic posterior neurite (Figure S5A), and zag-1-expressing and thus TRN-like FLP and PVD neurons had far fewer dendritic branches than the wild-type cells (Figure S4B). This latter result is consistent with ZAG-1’s role in preventing PVM from adopting a PVD-like morphology (Smith, et al., 2013).
dma-1, bicd-1, and flp-4 were also expressed in PVD neurons, suggesting that FLP and PVD share a common genetic program. The expression of these PVD markers was blocked by the misexpression of ZAG-1 in PVD neurons, suggesting that ZAG-1 represses the unknown PVD fate selector. On the other hand, the sto-5 reporter was expressed in FLP but not PVD neurons, indicating difference in the transcriptome of the two cell types (Figure S5E).
The activation of the FLP and PVD markers, like the TRN markers, required UNC-86 and MEC-3, which are the common terminal selectors for the three fates (Figure 6E and S5C). We identified one common terminal differentiation gene, mec-19, which was expressed in TRN, FLP, and PVD neurons at similar levels (and no other cells); its expression depended on UNC-86 and MEC-3 but was not affected by the loss of EGL-44/EGL-46 and ZAG-1 (Figure 6D-E). Thus, unc-86 and mec-3 serve as the “ground-state selectors” that promote a common ground states shared by the three types of neurons, and mec-19 is a marker for this ground state. Subsequently, zag-1 and egl-44 act as “modulators” that shift the ground state towards specific fates. Therefore, the development of a particular fate requires the activities of both the ground-state selectors and the fate-restricting modulators.
EGL-44/EGL-46 inhibits the expression of TRN genes by competing with UNC-86/MEC-3 for DNA binding and by suppressing ALR-1
We next investigated the mechanism, by which EGL-44/EGL-46 prevented the expression of TRN terminal differentiation genes. Although we found two UNC-86/MEC-3 binding sites required for the expression of a minimal TRN-specific promoter (a 184-bp mec-18 promoter) in TRNs, we failed to identify any additional cis-regulatory element that mediated its repression in FLP neurons (Figure S6A). Thus, EGL-44/EGL-46 seems unlikely to suppress TRN markers via distinct, repressive elements. We then tested the possibility that EGL-44/EGL-46 acts through the UNC-86/MEC-3 binding site, since EGL-44 belongs to the TEA domain class transcription factors, which recognize DNA sequences similar to the UNC-86/MEC-3 binding site (Figure 7A) (Zhang, et al., 2002; Jiang, et al., 2000). EGL-44 and the EGL-44/EGL-46 complex bound to previously identified UNC-86/MEC-3 motifs in the mec-4, mec-7, mec-17, and mec-18 promoters in electrophoretic mobility shift assays (Figure 7B and C). These results suggest that EGL-44 mediates the direct contact with the TRN promoters and EGL-46 acts as a corepressor. More importantly, EGL-44/EGL-46 outcompeted UNC-86/MEC-3 for the binding to these same sequences (Figure 7C), suggesting that the cis-regulatory sites normally bound by UNC-86/MEC-3 biochemically prefer EGL-44/EGL-46. Therefore, EGL-44/EGL-46 in the FLP neurons may prevent the activation of TRN genes by occluding the UNC-86/MEC-3 binding sites essential for the expression of TRN fate.
However, how EGL-44/EGL-46 avoids inhibiting UNC-86/MEC-3 targets that are commonly expressed in FLP and TRNs is unclear. One such target is mec-3 itself, whose activation and maintenance depend on two UNC-86/MEC-3 binding sites (Xue, et al., 1992). We found that the two sites were not bound by EGL-44 (Figure 7B), which explains why mec-3 expression is not affected by the presence of EGL-44/EGL-46 in FLP neurons. Comparing the cis-regulatory motif sequences on mec-3 promoter with those on TRN-specific promoters, we found that the four nucleotides (positions 1 to 4 in Figure 7A) following the UNC-86/MEC-3 binding sequence showed significant divergence between EGL-44-binding and nonbinding sites. In particular, the EGL-44-binding sites all contain adenine at the fourth position and the nonbinding sites do not (Figure 7A). Changing this adenine to thymine eliminated the binding of EGL-44 to the site on mec-4 promoter (Figure 7B). However, converting other nucleotides to adenine at the fourth position was not sufficient to enable the binding of EGL-44 to mec-3 promoter motifs and coordinated change of the nucleotides on the first and second positions are needed to evoke EGL-44 binding (Figure 7B). Similar results were also obtained for the UNC-86/MEC-3 target gene mec-10, which is expressed in both TRN and FLP cells (Huang and Chalfie, 1994). The UNC-86/MEC-3 site on mec-10 promoter was not bound by EGL-44 and changing the nucleotides on the first and fourth position enabled EGL-44 binding. Thus, our results suggest that EGL-44/EGL-46 can differentiate TRN/FLP common genes from TRN-specific genes via the cis-regulatory sequences in their promoters.
We next tested whether converting the EGL-44 binding site to a nonbinding site in a TRN promoter could prevent the suppression by EGL-44/EGL-46 in FLP neurons in vivo. Contrary to our expectation, a mec-4 promoter reporter harboring the adenine-to-thymine change at the fourth position was not activated in FLP neurons, although its TRN expression was preserved. In the 184-bp mec-18 promoter, changing the four nucleotides following the UNC-86/MEC-3 binding sequence from ACCA to CTAT completely abolished EGL-44 binding in vitro; but the mutant reporter was only weakly expressed in ∼30% of FLP neurons and remained silenced in the rest ∼70% (Figure S6B-D). In addition, creating an EGL-44-binding site in the mec-3 promoter by mutating the regulatory motif did not suppress mec-3 expression in FLP neurons (Figure S6B). The discrepancy between the in vitro and in vivo results suggests that the lack of EGL-44 binding to TRN differentiation genes per se is not sufficient to distinguish TRN fate from the TRN/FLP ground state. In addition to directly binding to the cis-regulatory elements of TRN promoters, EGL-44/EGL-46 may also inhibit TRN genes by activating or repressing the expression of other trans-acting factors.
One such factor may be ALR-1, which is an ortholog of human Arx and Drosophila aristaless and is needed for the robust differentiation of TRN fate (Topalidou, et al., 2011). Loss of alr-1 variably reduced but did not eliminate the expression of TRN markers in the TRNs in wild type and strongly eliminated the ectopic expression of TRN markers in FLP neurons in egl-44 and egl-46 mutants (Figure 7D and S6E). This difference between reduction and elimination may result from the fact that FLP neurons have lower mec-3 expression than the TRNs (Topalidou, et al., 2011) and thus a stronger need for ALR-1 to activate TRN markers. The observation that alr-1 is epistatic to egl-44 and egl-46 suggested that ALR-1 is a downstream effector suppressed by EGL-44/EGL-46. Indeed, a fosmid-based alr-1 translational reporter, which was normally expressed in TRNs but not FLP neurons, became de-repressed in FLP cells in egl-44 and egl-46 mutants (Figure 7D and S6E).
The lack of ALR-1 and the lower level of mec-3 in wild-type FLP neurons may explain the inactivation of the mutated TRN fate reporters that cannot be bound by EGL-44/EGL-46. Supporting this hypothesis, forced expression of ALR-1 in FLP neurons ectopically activated even the wild-type TRN fate reporters (Figure 7D) (Topalidou, et al., 2011), suggesting that ALR-1 not only promotes TRN fate but can also overcome the direct suppression from EGL-44/EGL-46 on the TRN genes. This ability of ALR-1 may result from ALR-1 upregulating mec-3 (Topalidou, et al., 2011) and directly interacting with TRN promoters (ChIP-seq data; Figure S6F). Therefore, EGL-44/EGL-46 inhibits TRN-specific genes by both occupying the UNC-86/MEC-3 site in the TRN promoters and suppressing TRN fate-promoting transcription factor ALR-1. Consistent with this model, overexpression of MEC-3 in FLP neurons could overcome EGL-44/EGL-46-mediated inhibition and activate the TRN program (Topalidou and Chalfie, 2011) presumably by both retaking the UNC-86/MEC-3 sites and by activating alr-1, which is a mec-3-dependent gene (Topalidou, et al., 2011).
Spatial and temporal expression of zag-1 and egl-44 are mutually exclusive
Given the mutual inhibition of zag-1 and egl-44 in the FLP neurons and the TRNs, we asked whether the two genes were generally expressed in different cells in the nervous system. Using the neurotransmitter maps for glutamatergic, cholinergic, and GABAergic neurons (Gendrel, et al., 2016; Pereira, et al., 2015; Serrano-Saiz, et al., 2013), we found that, in addition to TRNs, zag-1 was expressed in the AIB, AIM, AIN, AIZ, AVA, AVB, AVD, AVE, AVG, AVK, AVL, M4, M5, RIA, RIB, RIF, RIG, RIM, RIV, RMD, RME, RMF, RMH, SIA, and SMD neurons in the head, all the DD, VD, and VC neurons in the ventral cord, and the DVA, DVB, LUA, PDA, PVC, PVP, PVQ, PVR, and PVT neurons in the tail. zag-1 is also expressed in the serotonergic HSN neurons. In comparison, egl-44 expression was much more restricted, being found only in the FLP, ADL, and SAB neurons, as well as a few VA and VB motor neurons. Moreover, egl-44 was widely expressed in hypodermis, pharynx, and intestine, whereas zag-1 expression was absent in these tissues. We also constructed an egl-44::RFP reporter and cross it with zag-1:EGFP and did not observe overlapping expression in any cell (Figure S7A). The above data support the hypothesis that zag-1 and egl-44 expression are mutually exclusive in the nervous system and throughout all tissues of the animal. Given the mutual inhibition of EGL-44 and ZAG-1 in the TRNs/FLPs, we expected that the loss of zag-1 might affect egl-44 expression more broadly. We did not, however, observe a systematic upregulation of EGL-44::GFP expression in the nervous system in either the zag-1(zd86) hypomorphic mutants or the arrested L1 animals of zag-1(hd16) null mutants (Figure S7B and C). Similarly, tissues like hypodermis, pharynx and intestine did not gain zag-1 expression upon the loss of egl-44 either, suggesting that, in addition to the reciprocal inhibition, other activating signals are needed to create the expression pattern of the two transcription factors.
A few neurons expressed both zag-1 and egl-44 but did so at different times. egl-44 and egl-46 were expressed in the precursors of the postembryonic TRNs (the AVM and PVM neurons) and persisted in these cells for a few hours after their generation to promote cell cycle exit (Feng, et al., 2013; Wu, et al., 2001), but they were not expressed in terminally differentiated AVM and PVM. These differentiated cells expressed zag-1, which promoted the TRN fate. Similarly, egl-44 and egl-46 were transiently expressed in the early embryos in HSN neurons before they migrated and differentiated (Wu, et al., 2001) (Figure S7D), whereas zag-1 was expressed in the terminally differentiated HSN neurons in adults and was required for the activation of the HSN fate marker tph-1p::GFP (Figure S7E and F).
Discussion
RNAi screen is a systematic method to identify genes involved in cell fate determination
We demonstrate here that a systematic RNAi screen of a library of transcription factors can be used to identify neuronal cell fate regulators, particularly genes whose mutations lead to lethality or sterility. Our previous forward genetic screens, which searched for viable mutants with touch-sensing defects, despite reaching saturation, only identified unc-86 and mec-3 as the TRN fate determinants (Chalfie and Au, 1989; Chalfie and Sulston, 1981). Using the RNAi screen, we not only recovered unc-86 and mec-3 blindly, but also identified ceh-20, ldb-1, and zag-1 as genes required for the expression of TRN fate. These latter genes would not have been identified in our previous screens, because null mutations in them lead to early larval arrest. Moreover, our previous neuronal RNAi screen of nonviable genes would not have identified these genes either, because it only looked for postembryonic effects on touch sensitivity after TRN differentiation was completed (Chen, et al., 2015). In contrast, the current RNAi screen knocked down gene expression in the embryos and could identify genes whose silencing perturbed TRN fate specification.
Although previous studies of individual genes identified ceh-20 (Zheng, et al., 2015) and ldb-1 (Cassata, et al., 2000) as regulators of TRN fate, RNAi screens, such as we have done, provide a systematic, complementary approach to identify genes controlling neuronal differentiation. ceh-13 was also known to affect the TRN fate in ALM neurons (Zheng, et al., 2015) but was not identified from the RNAi screen, suggesting the existence of false negatives. Importantly, through the screen, we identified ZAG-1 as a new TRN fate determinant and the third piece in a combinatorial code (UNC-86, MEC-3, and ZAG-1) that defines TRN fate. Although previous studies suggested that ZAG-1 regulates PVM morphological differentiation using a viable hypomorphic zag-1 allele (Smith, et al., 2013), our RNAi screen and the subsequent studies using a null allele found that ZAG-1 is essential for the acquisition of TRN fate more generally.
Our RNAi results, however, raise two concerns about the reliability and applicability of such screens. First, only 5 of the 14 genes we identified from the screen were confirmed with mutants. The remaining 9 genes (65%) appeared to be false positives, since their null mutations did not result in TRN differentiation defects, although RNAi treatment against these genes all repeatedly showed the loss of TRN marker expression. These false positives probably resulted from mistargeting of the dsRNA. A similar problem was encountered in a genome-wide RNAi screen for genes involved in the specification of ASE neuron in C. elegans (Poole, et al., 2011), since a significant fraction of the genes identified from the screen could not be verified using mutants (Oliver Hobert, personal communication). Second, although both ALM and PLM neurons required similar factors for cell fate determination, RNAi against unc-86 and the other fate regulators caused much stronger effects in the ALM neurons than in PLMs. This bias towards the anterior was previously observed in our enhanced neuronal RNAi system (Calixto, et al., 2010), suggesting that neurons with the cell body located in the tail region, especially positioned posteriorly to the anus, may be less affected by dsRNA introduced by feeding than those located in the anterior half of the animal. This bias led to the failure of recovering egl-5, whose loss was known to only affected PLM differentiation (Zheng, et al., 2015).
Non-selector type of cell fate regulators promotes fate acquisition indirectly
The key concept of “terminal selectors” is that transcription factors controlling neuronal cell fate directly activate terminal differentiation genes associated with the identity and function of the neuron through common cis-regulatory elements (Hobert, 2011; Hobert, et al., 2010). POU homeodomain protein UNC-86 and LIM homeodomain protein MEC-3 are examples of such terminal selectors, since they directly activate TRN genes via the UNC-86/MEC-3 binding sites in the promoter of TRN genes (Duggan, et al., 1998).
Here, however, we found that at least two differentiation regulators that are required for TRN fate indirectly promote the expression of TRN genes. The LIM domain-binding protein LDB-1 binds to MEC-3, which protects MEC-3 from degradation and enables it to activate downstream TRN genes. ZAG-1 promotes the expression of TRN markers by repressing the cell fate inhibitors EGL-44 and EGL-46. Although both LDB-1 and ZAG-1 are absolutely required for the adoption of TRN fate, neither of them directly associated with the DNA elements of TRN genes and control the activity of TRN promoters. ZAG-1 belongs to the ZEB family of transcription factors, which mostly function as transcriptional repressors (Vandewalle, et al., 2009); so, it could only indirectly induce the expression of TRN genes. Moreover, unlike the highly confined expression of classical terminal selectors (e.g. mec-3 is only expressed in ten neurons), LDB-1 and ZAG-1 were expressed broadly. ldb-1 is widely expressed in the nervous system, the reproductive system, and the muscles (Cassata, et al., 2000), and zag-1 is expressed in many neurons in the head and tail ganglia and ventral cord motor neurons, as well as various muscles (Wacker, et al., 2003). These features suggest that LDB-1 and ZAG-1 may represent a non-selector type of cell fate regulators that is phenotypically identical to but mechanistically different from the conventional terminal selectors.
A bistable negative feedback loop controls binary cell fate decisions
A central question for neuronal differentiation is how the extraordinary diversity of neuron types are generated and maintained. The “combinatorial code” hypothesis suggests that the identity of a neuron type or subtype is determined by a combination of transcription factors (Allan, et al., 2003; Jessell, 2000; Mitani, et al., 1993), but how their actions are coordinated mechanistically in a regulatory network is not well understood. One way to delineate the activities of a complex code is to investigate how binary fate decisions are made to distinguish one neuron type from another. In this study, we found that within the three-factor code for the TRN fate, two factors (UNC-86 and MEC-3) form a selector for a ground state shared by FLP and TRNs, whereas the third factor (ZAG-1) controls a binary choice between the two fates.
Although FLP and TRN share the same selectors, FLP neurons fail to express the TRN genetic program because EGL-44/EGL-46 blocks the activation of TRN genes by occupying the cis-regulatory elements bound by UNC-86/MEC-3 and by repressing the TRN fate regulator ALR-1 (Figure 8). This inhibition of TRN fate is very robust, since EGL-44/EGL-46 outcompete UNC-86/MEC-3 for the binding of the same cis-regulatory element in TRN promoters (Figure 7C) and misexpressing egl-44 in TRNs can completely shut off the expression of TRN genes (Figure 5D). As a consequence, TRNs utilize a protective mechanism that uses ZAG-1 to strongly repress both egl-44 and egl-46 and thus ensures their absence in the TRNs. Reciprocally, EGL-44/EGL-46 also silences zag-1 in FLP neurons, which secured the expression of egl-44 and egl-46 and prevented the adoption of TRN fate. Such mutual repression enables a bistable switch that locks the cell fate in one of the two possible states (Figure 8). In the absence of such a switch (e.g. in egl-44; zag-1 double mutants), both TRNs and FLP neurons expressed the TRN genetic program (Figure 4), suggesting that the ground state is a TRN-like fate and the selectors UNC-86/MEC-3 activate TRN genes by default. EGL-44/EGL-46-induced modification of the default genetic program gives rise to the FLP fate.
Interestingly, the Drosophila homolog of ZAG-1, Zfh1, also regulates binary cell fate choices. Zfh1 promotes a GW motor neuron fate over an EW interneuron fate in the 7-3 neuroblast lineage, and its expression is suppressed by the steroid receptor family transcription factor Eagle in the EW interneuron (Lee and Lundell, 2007); whether Zfh1 also suppressed Eagle in GW motor neuron is unclear. In another example, the mutual antagonism between Zfh1 and another zinc-finger transcription factor, Lame duck (Lmd), regulates the decision between pericardial cell and fusion competent myoblast fates in the mesoderm (Sellin, et al., 2009). Although a genetic interaction between Zfh1 and the Drosophila EGL-44, Scalloped, has not been reported, these results suggest that Zfh1 can form regulatory switches with other transcription factors. Mouse homologs of ZAG-1, ZEB1 and ZEB2, repress tissue differentiation during early embryogenesis, induce epithelial mesenchymal transition (EMT), and are essential for neural tube development (Vandewalle, et al., 2009). Their roles in the specifying terminal cell fates, however, are unclear, because their knockout leads to embryonic lethality (Miyoshi, et al., 2006).
The choosing of one of two alternative fates appears to be a common way to diversify neuronal types and subtypes during development, as shown by several examples in C. elegans and Drosophila (Sarin, et al., 2007; Mikeladze-Dvali, et al., 2005). In vertebrates, neuronal fate, especially neurotransmitter identity, is also subjected to such binary regulation, but the underlying mechanism is largely unknown. For example, Lbx1 and Helt promotes GABAergic identity and suppresses glutamatergic differentiation in the dorsal spinal cord (Cheng, et al., 2005) and the mesencephalon (Nakatani, et al., 2007), respectively. The selector gene Fezf2 induces the opposite choice in mouse corticospinal motor neurons by activating Vglut1 to produce glutamatergic cells and by repressing Gad1 to inhibit a GABAergic fate (Lodato, et al., 2014). However, whether the two fates are exclusive to each other because of a bistable switch is unknown. One important feature of those binary fate choices is that the selector often simultaneously induces one fate and suppresses the other. Whether there two activities can be uncoupled is unclear. Our studies of the FLP versus TRN fate choice suggests that uncoupling is possible. Although EGL-44/EGL-46 simultaneously induces FLP genes and suppresses TRN genes, ZAG-1 appears to be solely devoted to inhibiting FLP fate (and possibly PVD fate) and is not involved in directly activating TRN genes. The uncoupling between UNC-86/MEC-3 activating TRN fate and ZAG-1 preventing alternative fates suggest that the two modules may be independently evolved.
Binary fate switches may enable the generation of neuronal diversity
Because the introduction of self-reinforcing, bistable switches can generate diversity within pre-existing neuronal fates, we envision that the extraordinary variety of neuronal types in the nervous system evolved from a few primitive neuronal fates through stepwise addition of binary switches. For example, since TRN, FLP, and PVD fates all require the same selectors, UNC-86 and MEC-3, they may be derived from a common ancestral fate. In fact, the existence of UNC-86/MEC-3 target genes like mec-19, which is expressed in all the FLP, PVD, and TRNs but not any other neuron, suggests such an ancestral fate.
One possible evolutionary derivation of these cells is that some of the ancestral TRN-like cells acquired the ability to express egl-44 and egl-46. This expression led to the suppression of TRN genes, the activation of FLP genes and the emergence of this new cell type. The fact that the ground state is a TRN-like state seems to support this hypothesis. Moreover, since egl-44 is mostly expressed in non-neuronal tissues, mutations in the regulatory elements of egl-44 might have allowed expression in some neurons. Thus, EGL-44 may have been co-opted to induce divergence among neurons that share a common fate, and this cell fate divergence is subsequently stabilized by the establishment of a negative feedback loop between EGL-44 and ZAG-1.
Alternatively, a regulatory element in the zag-1 gene, which was already present in the ancestral TRNs to repress egl-44 and egl-46, may have been mutated to prevent its expression in some ancestral TRN cells. This loss led to the de-repression of egl-44 and egl-46 and the subsequent acquisition of the FLP fate. The widespread expression of zag-1 in neurons suggests that it may be in general required for the differentiation of many neuronal fates, and selective loss of zag-1 expression in some cells through changes in cis-regulatory sequences may prevent these cells from committing to particular fates and allowing them to adopt alternative ones.
Overall, we imagine that a limited number of ground-state selectors may first define a handful of shared, and subsequently a series of binary fate switches carry out further differentiation that modifies the ground state to generate a diverse array of terminal neuronal fates. The broad expression of ZAG-1 in the nervous system and the lack of increased EGL-44 expression in zag-1 mutants suggests that ZAG-1 may form bistable regulatory loops with many different inhibitors. In fact, ZAG-1 is required for the fate specification of at least TRN, HSN, and PVQ neurons (this study; Clark and Chiu, 2003), although whether ZAG-1 also regulates HSN and PVQ fates by repressing alternative fates is unclear. Moreover, the fact that ZAG-1 expression is only found in a subset of neurons that express the same selector supports the hypothesis that ZAG-1 may ensure diversification from a shared ground state through down-regulation of inhibitors; e.g. zag-1 is expressed in 7/17, 6/15, and 3/11 classes of neurons that use UNC-86, UNC-3, and CEH-14 as a terminal selector, respectively (classification according to Hobert, 2016). Finally, the broad expression of ZEB family transcription factors in the nervous system of both Drosophila (Lai, et al., 1991) and mice (Vandewalle, et al., 2009) suggests possibly a conserved role for them in regulating binary fate choices and the generation of neuronal diversity.
Author Contributions
C.Z., F.Q.J., and B.L.T. performed experiments and analyzed the data. C.Z. and M.C. conceived the study and wrote the manuscript. M.C. supervised the work and provided funding. F.Q.J. and B.L.T. contributed equally to the work.
STAR Methods
❖ Strains, Constructs, and transgenes
C. elegans strains and mutant alleles
C. elegans wild type (N2) and mutant strains were maintained as previously described (Brenner, 1974). Most strains were provided by the Caenorhabditis Genetics Center, which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440), or the National BioResource Project of Japan. VH514, zag-1 (hd16) /unc-17 (e113) dpy-13 (e184) and VC639, ldb-1 (ok896)/szT1; +/szT1[lon-2 (e678)] were used as the balanced null alleles for zag-1 and ldb-1, respectively. zag-1 (zd86) was used as a hypomorphic allele for zag-1. For other genes identified from the RNAi screen (see below), we tested zip-4 (tm1359), nhr-119 (gk136908), nhr-166 (gk613), nhr-159 (tm2323), egl-38 (ok3510)/nT1[qIs51], hmbx-1 (ok3467), fkh-2 (ok683), lin-40 (ku285), lin-40 (s1506) unc-46 (e177)/eT1, elt-6 (gk723), and elt-6 (gk754). Other mutant alleles used in this study include ahr-1(ju145), egl-44(n1080), egl-46(n1127), mec-3(u184), and unc-86(u5).
Constructs
A 2.4 kb mec-3 promoter, a 5.9 kb ldb-1 promoter, a 2.2 kb zag-1 promoter, a 2.2 kb bicd-1 promoter, 4.9 kb dma-1 promoter, 3.1 kb flp-4 promoter, 2.3 kb sto-5 promoter, and 1.3 kb mec-19 promoter were cloned from wild type (N2) genomic DNA into the Gateway pDONR221 P4-P1r vector. The genomic coding region of GFP, ldb-1a, zag-1, egl-44, egl-46, alr-1, and ahr-1 were cloned into Gateway pDONR221. The resulted entry vectors, together with pENTR-unc-54-3’ UTR and the destination vector pDEST-R4-R3 were used in the LR reaction to create the final expression vectors. Gateway cloning was performed according to the manual provided by Life Technologies (Grand Island, NY).
DNA constructs TU#625 and TU#626 (Wu, et al., 2001) contain translational GFP fusion of egl-44 and egl-46, respectively, and were injected into animals to form reporters for the two genes, named as uIs215[egl-44::GFP] and uEx927[egl-46::GFP]. TU#924 contains a 400 bp mec-18 promoter inserted into pPD95.75 between HindIII and BamHI sites, and this mec-18p::GFP construct was used as a template to create a series promoter variants shown in Figure S10 using the Q5 site-directed mutagenesis kit from New England Biolabs (Ipswich, MA).
Transgenes
Transgenes zdIs5[mec-4p::GFP] I, muIs32[mec-7p::GFP] II, uIs31[mec-17p::GFP] III, uIs115[mec-17p::RFP] IV, uIs134[mec-17p::RFP]V, and uIs72[mec-18p::mec-18::GFP]were used as fluorescent markers for the TRN cell fate. uEx1104[Pbicd-1::GFP], uEx1105[dma-1p::GFP], uEx1106[flp-4p::GFP], and uIs232[sto-5p::GFP] served as FLP fate markers. zdIs13[tph-1p::GFP] and vsIs97[tph-1p::DsRed2] were used as HSN fate marker. uIs22[mec-3p::GFP] and uIs152[mec-3p::RFP] were used as mec-3 transcriptional reporter; uEx1007[mec-3p::mec-3::GFP] and wgIs55[mec-3::TY1::EGFP::3xFLAG] were used as mec-3 translational reporter. wgIs83[zag-1::TY1::EGFP::3xFLAG], wgIs476[unc-86::TY1::EGFP::3xFLAG], uEx1006[ldb-1p::GFP], wgIs200[alr-1::TY1::EGFP::3xFLAG], leEx1709[ahr-1::GFP] and uEx1107[mec-19p::GFP] served as the reporters for zag-1, unc-86, ldb-1, alr-1, ahr-1, and mec-19 respectively. uIs211[mec-3p::egl-44], uEx926[mec-3p::zag-1], and uEx1027[mec-3p::alr-1] were used for misexpression.
To perform cell identification, we crossed wgIs83[zag-1::TY1::EGFP::3xFLAG] and uIs215[egl-44::GFP] into the otIs518[eat-4::SL2::mCherry::H2B], otIs544[cho-1::SL2::mCherry::H2B], and otIs564 [unc-47::SL2::H2B::mChopti], labeling glutamatergic, cholinergic, and GABAergic neurons (Gendrel, et al., 2016; Pereira, et al., 2015; Serrano-Saiz, et al., 2013), respectively. We identified zag-1 and egl-44 expressing neurons, based on the position and neurotransmitter identity of the cells expressing GFP.
❖ RNAi screen
RNAi screen was performed using a modified bacteria-feeding protocol previously reported (Poole, et al., 2011; Kamath, et al., 2003). We used the Ahringer RNAi library from Source Bioscience (http://www.lifesciences.sourcebioscience.com/) and the list of 392 RNAi clones targeting transcription factors were generated by searching WormBase WS238 using Gene Ontology terms related to “DNA binding” and “transcription factor activity” (see the complete list in Table S1). To perform the RNAi experiments, we seeded bacteria expressing dsRNA on NGM agar plates containing 6 mM IPTG and 100 µg/ml ampicillin. One day later, eggs from TU4429, eri-1 (mg366); lin-15B (n744); uIs134[mec-17p::RFP] animals were placed onto these plates; the eggs hatched and grew to adults at 20°C. The F1 progeny of these worms were scored for the expression of TRN markers at the second larval stage. RNAi clones were considered to be positive if more than 15% (n > 20) of the treated animals failed to show RFP expression in the ALM neurons in at least two of the three replicate plates. Three initial rounds of screens were performed on all the 392 clones, and 14 clones were found positive in all the three rounds. Two more screening rounds were then conducted on these 14 RNAi clones, which were all confirmed to be positive. We sequenced the inserts of all positive clones to confirm the identity of the target genes.
❖ Yeast two-hybrid assay
Yeast media and plates were prepared according to recipes from Clontech (Mountain View, CA) and yeasts were grown at 30°C. The yeast strain PJ69-4a (provided by Songtao Jia at Columbia University) used for the two-hybrid assays contains GAL1-HIS3, GAL2-ADE2, and GAL7-lacZ reporters. Vectors pGAD424 and pGBT9 (Clontech) were used to express proteins fused to the yeast activating domain (AD) and binding domain (BD), respectively. cDNA fragments of mec-3, unc-86, zag-1, ldb-1, egl-44, and egl-46 were cloned into the two-hybrid vectors either using restriction enzymes or with Gibson Assembly (NEB).
Combinations of the AD or BD vectors were co-transformed into yeast using the Frozen-EZ II kit from Zymo Research (Irvine, CA) and using empty vectors as negative controls. Growth assays were performed by growing individual colonies overnight in selective media lacking tryptophan and leucine. Cultures were then diluted to let OD600 become 0.5, and 10 μl of a further 1:10 diluted culture were spotted onto plates lacking histidine to test the expression of the HIS3 reporter. Plates were imaged after 2 days of growth. Liquid β-galactosidase assays were performed using the Yeast β-Galactosidase Assay Kit (Thermo Scientific, Rockford, IL).
Electrophoretic mobility shift assay (EMSA)
Recombinant GST::EGL-44 proteins were produced in E. coli BL21 (DE3) using the expression vector pGEX-6p-1 (Amersham Pharmacia Biotech, UK) and purified using affinity chromatography columns filled with Glutathione Sepharose 4B beads (Amersham). EGL-44 was cleaved off the column using PreScission Protease (Amersham). EGL-46 was expressed using the pET32a vector (Novagen, Madison, WI) and purified using the S-Tag rEK purification kit (Novagen). UNC-86 and MEC-3 were produced according to previous methods(Xue, et al., 1993).
Gel mobility shift assays were performed using a modified protocol previously reported (Xue, et al., 1993). DNA probes were labeled with Biotin by Biotin 3’ End labeling kit (Pierce, Rockford, IL) and then annealed into double strands. 100 ng of proteins were incubated with 0.005 ng probe at room temperature for 30 min, and the mixture was loaded onto a 10% TBE polyacrylamide mini gel (Bio-Rad, Hercules, CA). The gel was transferred to a 0.2 mm nylon membrane, which was then treated with UV light to crosslink DNA and proteins. Biotin-labeled DNA was detected using LightShift chemiluminescence EMSA kit (Pierce). The sequences of the probes are listed in Table S3.
❖ smFISH, phenotypic scoring, and statistical analysis
smFISH
Single-molecule fluorescence in situ hybridization (smFISH) was performed as described previously (Topalidou, et al., 2011). Imaging was conducted on a Zeiss Axio Observer Z1 inverted microscope with a CoolSNAP HQ2-FW camera (Photometrics, Tucson, AZ).
Phenotypic scoring
To examine the expression pattern of TRN markers, we grew animals at 20° C, examined them using the same microscope and recorded the percentages of TRN cells that express the fluorescent reporter in three independent experiments. The results are presented as aggregates; no significant differences were seen between replicates. To test transgenic animals, we injected DNA constructs (5 ng/μl for each expression vector) into the animals to establish stable lines carrying the extrachromosomal array; at least three independent lines were tested. In some cases, the transgene was integrated into the genome using γ-irradiation (Mello, et al., 1991), and at least three integrant lines were outcrossed and examined.
Statistical analysis
Statistical significance was determined using the Student’s t-test for the majority of comparisons of two sets of data. For multiple comparisons, the Holm-Bonferroni method was used to correct the p values.
Acknowledgement
We thank Songtao Jia and Elizabeth Miller for sharing materials and reagents. We also thank Oliver Hobert, Richard Mann, and the members of our laboratory for helpful discussions and comments. This work was supported by NIH grants GM30997 and GM122522 to MC.