The Groucho ortholog UNC-37 interacts with the short Groucho-like protein LSY-22 to control developmental decisions in C. elegans

Drosophila Groucho, C. elegans UNC-37 and their four vertebrate orthologs Tle1-4 (Grg1-4) are WD40 domain-containing co-repressors of gene expression (Buscarlet and Stifani, 2007; Chen and Courey, 2000; Gasperowicz and Otto, 2005). Co-repressors, which are recruited to DNA by a variety of transcription factors, recruit other proteins, such as histone deacetylases (HDACs), that modify chromatin and thereby control gene expression (Privalsky, 2001). Groucho/UNC-37 and their vertebrate orthologs can interact with a plethora of different DNA-binding transcription factors and have been implicated in a number of diverse biological processes, including neuronal pattern formation, hematopoiesis, osteopoiesis, muscle, gland and eye development and tumorigenesis (Buscarlet and Stifani, 2007; Chen and Courey, 2000; Gasperowicz and Otto, 2005).

Aside from the canonical, WD40 domain-containing family members, there are several unusual and little-studied variants of Groucho/TLE proteins. These proteins, which are sometimes referred to as short Groucho proteins, are either short splice isoforms of longer Groucho/TLE family members (e.g. Grg1-S or Grg3b) or are encoded by entirely distinct genetic loci [e.g. mouse or human AES (Grg5)] (Gasperowicz and Otto, 2005; Miyasaka et al., 1993). In either case, these short Groucho proteins lack the C-terminal WD40 domains that are involved in transcription factor interactions, but contain the so-called Q-rich and GP (glycine/proline-rich) N-terminal domains. Within canonical Groucho/TLE proteins, the Q-rich domain mediates oligomerization of Groucho/TLE family members, as well as interaction with other DNA-binding transcription factors, whereas the GP domain recruits other proteins, such as HDACs (Gasperowicz and Otto, 2005; Miyasaka et al., 1993). In contrast to full-length Groucho/TLE proteins, the biological and molecular function of short Groucho proteins is not well understood. In vitro evidence suggests that they function either as antagonists or agonists of normal Groucho proteins (Buscarlet and Stifani, 2007; Gasperowicz and Otto, 2005); this contradiction illustrates the need for genetic loss-of-function studies that address the precise biological role of short Groucho proteins and their relation to normal Groucho function. We report here the isolation and functional characterization of a short Groucho protein, LSY-22, in the nematode C. elegans.

The following were used: N2 Bristol wild-type (Brenner, 1974); OH7410, lsy-5(ot37) otIs114/hT2[bli-4(e937) let-?(q782) qIs48] (I;III); OH9305, lsy-5(ot240) otIs114; OH7116, lsy-22(ot114)/hT2[bli-4(e937) let-?(q782) qIs48] (I;III); otIs3; OH7115, lsy-22(ot244) otIs114/hT2[bli-4(e937) let-?(q782) qIs48] (I;III) (Sarin et al., 2007); CB262, unc-37(e262); OH9242, unc-37(wd17wd22)/hT2[bli-4(e937) let-?(q782) qIs48] (I;III) (Miller et al., 1993); OH9289, lsy-22(tm3750)/hT2[bli-4(e937) let-?(q782) qIs48] (I;III); OH1997, otIs114; him-8; dpy-11 lsy-6(ot71); OH351, otIs114; die-1(ot26); OH1583, otIs114; cog-1(ot28), DR293, dpy-5(e61) unc-101(m1).

Transgenes to label ASEL and ASER fates included: otIs3 V and otIs4 X, Is[gcy-7^prom::gfp; lin-15 (+)]; otIs114 I, Is[lim-6^prom::gfp; rol-6(d)]; otIs220 IV, Is[gcy-5^prom::mCherry; rol-6(d)]; ntIs1 V, Is[gcy-5^prom::gfp; lin-15 (+)]; and the ASEL/ASER markers otIs151 V, Is[ceh-36^prom::DsRed2; rol-6(d)] and otIs232 V, Is[che-1^prom::mCherry; rol-6(d)]. Reporters for other cell types included: bxIs14, Is[pkd-2::gfp, pha-1] (Lints et al., 2004); wdIs3, Is[del-1::gfp] (Winnier et al., 1999); and qIs19, Is[lag-2::gfp] (Blelloch et al., 1999).

Transgenic arrays included: OH9296, otEx4124, Ex[ceh-36::cog-1,elt-2]; OH9297, otEx4125, Ex[lsy-22^FOS::2×FLAG::venus, elt-2::rfp]; otIs232; OH9557, lsy-22(ot244) otIs114, Ex[lsy-22^FOS, lsy-22^5′FOS, elt-2::rfp]; and OH9556, otIs288 Is[unc-37^FOS::yfp, elt-2::rfp] otIs232.

unc-37 and lsy-22 reporter genes were created utilizing λ-Red-mediated recombineering in bacteria as described (Tursun et al., 2009). Briefly, the lsy-22-containing fosmid (WRM0628cC04) and the unc-37-containing fosmid (WRM0612cG05) were electroporated into E. coli strain SW105 (Warming et al., 2005). Using FLP recombinase-removable galK-based cassettes, we inserted yfp immediately preceding the stop codon at the C-terminus of lsy-22 or unc-37, resulting in a translational fusion. Recombineered fosmids were sequenced at their recombineered junctions and correct clones were maintained in E. coli strain EPI-300 T1R (Epicentre). Fosmids were digested with SbfI and injected at 10 ng/μl, together with 2 ng/μl ScaI-digested rol-6(d) (pRF4; 2 ng/μl) or HindIII-digested elt-2::NLS-dsRed and PvuII-digested bacterial genomic DNA (150 ng/μl) to generate a complex array. The DNA was injected into wild-type N2. The resulting array is called otEx4125 (lsy-22^FOS::2×FLAG::venus) and otEx4126 (unc-37^FOS::yfp). otEx4126 subsequently spontaneously integrated to generate otIs288.

lsy-22(ot114) was mapped by standard three-factor mapping between dpy-5 (0 cM) and unc-101 (13.3 cM) using strain DR293 dpy-5(e61) unc-101(m1). Briefly, upon picking ∼100 F2s, Lsy Dpy Unc animals were never obtained, whereas both Dpy Lsy non-Unc and Unc Lsy non-Dpy recombinants were found, suggesting that lsy-22 lies between dpy-5 and unc-101. This interval was analyzed for sequence variants by whole-genome sequencing (see below).

lsy-5(ot37) and lsy-5(ot240) were placed on chromosome I because they showed linkage to the integrated transgene otIs114, with which we scored ASEL fate. Further mapping of both alleles, as well as the otIs114 reporter, using single-nucleotide polymorphisms in the CB4856 Hawaiian wild-type isolate revealed two regions of possible linkage on chromosome I, from –8.23 cM to –5.78 cM and from –1.50 cM to +3.01 cM. lsy-5(ot240) was chosen for whole-genome sequencing and both possible linkage regions were analyzed for the presence of sequence variants.

Genomic DNA from homozygous lsy-22(ot114) animals was prepared by picking 1000 progeny of heterozygous ot114 animals with a Lsy phenotype and using the Gentra GenePure DNA extraction kit. Genomic DNA was prepared from lsy-5(ot240) from a homozygous population of viable animals. Genomic DNA (5 μg) prepared from either lsy-5(ot240) or lsy-22(ot114) was sequenced as previously described (Sarin et al., 2008). ot240 was sequenced on an in-house Illumina Genome Analyzer II according to the manufacturer's specification and the resultant sequence data were analyzed with MAQGene (Bigelow et al., 2009). The following stringency criteria were used to filter variants: quality score ≥3, loci multiplicity ≤1, sequencing depth ≥3 (both orientations required). ot114 data were generated by Illumina's sequencing service and data were analyzed using MAQ (as MAQGene was not available at that time). In both cases, all detected sequence variants were analyzed for whether they were present in any one of 14 additional genomes of different genotypes that we sequenced in the laboratory; if so, they were considered background and eliminated from further analysis.

Yeast two-hybrid experiments were performed according to the Matchmaker Gold Yeast Two-Hybrid System protocol (Clontech). cDNA was amplified from a total RNA population using primers for the lsy-22, unc-37 or cog-1 locus and cloned into the yeast vectors pGBKT7 and pGADT7. Vectors were co-transformed into the yeast strain AH109 and transformations were plated on SC-Leu-Trp media. Five transformants were randomly selected from each transformation and diluted in 100 μl of water to a uniform optical density. Then 10 μl of each dilution was spotted onto the following plates: SC-Leu-Trp, SC-Leu-Trp-His + 3-amino-1,2,4-triazole (3-AT), and SC-Ade-Leu-Trp-His + X-α-Gal. Plates were incubated for 4-5 days at 30°C and scored for growth and on X-α-Gal-containing plates for blue staining of the spotted yeast culture dilutions.

Packed worms [1.5 ml; either otEx4125 (lsy-22::FLAG::venus) worms that express FLAG-tagged lsy-22 or wild-type N2] were washed with M9 buffer and resuspended in 2 ml IP buffer [50 mM Hepes-KOH pH 7.6, 100 mM NaCl, 1 mM EDTA, 0.25 M LiCl, 1% sodium deoxycholate, 0.5% NP40, 10% glycerol and protease inhibitors (Sigma S8820)] and homogenized in a glass homogenizer on ice. After a 30-minute incubation on ice, lysates were centrifuged for 20 minutes at 15,000 g at 4°C to clear of debris. Each cleared lysate was incubated at 4°C for 4 hours with either 100 μl of a 50% slurry of anti-FLAG M2-agarose beads (Sigma A2220) in IP buffer or anti-HA HA-7-agarose beads (Sigma A2095) as negative control. By volume, 2.5% of each lysate was kept as input. The beads were washed three times with IP buffer and precipitates and lysates were subjected to immunoblotting using anti-UNC-37 [R-4677, a gift from D. Miller (Winnier et al., 1999)] or anti-FLAG (Sigma A8592) antibody.

The ASE neurons of the nematode C. elegans are a pair of morphologically symmetric gustatory neurons that are functionally lateralized and express distinct chemoreceptors in the left (ASEL) versus right (ASER) neuron (Fig. 1A) (Hobert et al., 2002; Ortiz et al., 2009). Genetic screens have revealed a complex gene regulatory network composed of transcription factors and miRNAs that control ASE laterality (Chang et al., 2004; Chang et al., 2003; Hobert, 2006; Johnston and Hobert, 2003; Johnston et al., 2005; Sarin et al., 2007) (Fig. 1A). Two distinct, previously uncloned loci retrieved by these screens, lsy-5 and lsy-22, displayed identical defects in their ability to execute the ASER fate, which instead converted to ASEL fate (Fig. 1B,C; note that lsy-5 is referred to as unc-37 in Fig. 1C, for reasons explained below).

The ASER fate-inducing activity of the lsy-5 and lsy-22 genes appears intimately linked as they not only share similar ASE phenotypes but also show similar epistatic relationships to previously described members of the regulatory network that control ASE laterality. That is, both mutants suppressed the phenotypic consequence (i.e. ASEL to ASER conversion) of the loss of the ASEL fate-inducing miRNA encoded by lsy-6 (Fig. 1C). Both genes also genetically interact in a similar manner with cog-1, which encodes an Nkx6.1-like, EH1 domain-containing homeodomain protein that is thought to directly bind UNC-37 (Chang et al., 2003). First, both mutants failed to complement cog-1 (Fig. 1C); such failure to complement is often an indication that the encoded gene products act in the same complex (e.g. Hays et al., 1989). Second, cog-1 requires both lsy-5 and lsy-22 to specify ASER fate. Forced bilateral expression of cog-1 in both ASE cells resulted in ectopic ASER fate induction in ASEL (Fig. 1C). This ability of cog-1 to induce ASER fate was abrogated in lsy-5 and lsy-22 mutants (Fig. 1C), demonstrating that COG-1 requires both genes to fulfil its function as an ASER fate inducer. Lastly, both lsy-5 and lsy-22 act upstream of the ASEL fate inducer die-1, as the `two ASEL fate' mutant phenotype observed in lsy-5 and lsy-22 mutants was suppressed in die-1 mutants (Fig. 1C). We conclude that lsy-5 and lsy-22 affect the bistable ASE cell fate decision in a similar manner.

We mapped lsy-5 and lsy-22, each represented by two independently isolated alleles, to genetic intervals on chromosome I and then used the recently introduced `gene cloning' technology of whole-genome sequencing (Sarin et al., 2008) for one mutant allele of each to identify the molecular nature of the affected genes. The genome sequence data are summarized in Fig. 2A. This approach identified a small number of protein-coding sequence variants in the respective genetic intervals, including a premature stop codon in F27D4.2 of lsy-22(ot114) animals and a missense mutation in W02D3.9 of lsy-5(ot240) animals (Fig. 2A,B). These are the phenotype-causing variants based on the following evidence: (1) the second allele of each locus showed a mutation in the same, respective gene (Fig. 2B); (2) additional, independently isolated alleles of F27D4.2 and W02D3.9 resulted in similar ASE laterality defects (Fig. 1C); (3) RNAi against F27D4.2 and W02D3.9 reproduced both the lethality and ASE phenotypes of the strongest alleles of each locus (see Table S1 in the supplementary material); and (4) mutant alleles of either locus could be rescued by wild-type genomic copies of the respective genes (see Table S1 in the supplementary material). We conclude that F27D4.2 is lsy-22 and W02D3.9 is lsy-5.

W02D3.9/lsy-5 encodes the previously described Groucho ortholog UNC-37 (from here on we refer to lsy-5 as unc-37) (Fig. 2B). unc-37 was previously implicated in controlling ASE laterality through its interaction with the EH1 domain-containing homeobox factor COG-1 (Chang et al., 2003); we had not considered unc-37 as a candidate for lsy-5 as the ot37 allele that we isolated complemented the canonical unc-37 allele e262 (see Table S2 in the supplementary material). Allelic complementation is a relatively rare phenomenon, but is known to occur in multimeric multidomain proteins (e.g. Griffin and Chan, 2006).

F27D4.2/lsy-22 encodes a small, previously uncharacterized protein. Close sequence inspection reveals similarity to the N-terminus of UNC-37 (see Fig. S1 in the supplementary material); both proteins contain a Q-rich domain and a GP domain, yet only UNC-37, not LSY-22, contains C-terminal WD40 domains (Fig. 2B). Like other Groucho-related proteins, LSY-22 is predicted to contain two amphipathic alpha-helices within the Q-rich domain, the first of which contains a coiled-coil region (see Fig. S1 in the supplementary material). The Q-rich domain of long and short Groucho proteins has been shown to mediate their multimerization, whereas the GP domain is thought to mediate interactions with HDACs (Chen et al., 1998; Pinto and Lobe, 1996). Overall, LSY-22 is therefore similar to a small number of invertebrate and vertebrate Groucho-related proteins that also share similarity to the N-terminus of Groucho, but lack the C-terminal WD40 domains (AES proteins and shorter splice forms of Groucho/TLE family members) (Buscarlet and Stifani, 2007). No other obvious AES-related proteins were identified by sequence similarity in the C. elegans genome.

To analyze their expression, we recombineered a fluorescent reporter gene into ∼40 kb fosmids that contain the lsy-22 and unc-37 genomic loci (Fig. 2C). Each fosmid reporter rescued several of the mutant phenotypes of lsy-22 and unc-37 (data not shown). lsy-22 was very broadly expressed in the nucleus of most, if not all, cells tested, starting in early embryogenesis and persisting throughout larval and adult life (Fig. 2C). Consistent with previously described antibody staining (Pflugrad et al., 1997), a fosmid-recombineered unc-37 reporter showed a similar, broad expression pattern in apparently all cells at all times (Fig. 2C). Such broad expression is characteristic of Groucho/TLE family members in other organisms as well, including short Groucho proteins such as AES (Mallo et al., 1993).

In addition to their similar ASE laterality defect and indistinguishable epistatic relationship to other ASE fate regulators (Fig. 1C), unc-37 and lsy-22 null mutants showed a striking set of phenotypic similarities. First, lsy-22 and unc-37 null mutants displayed a maternally rescued sterility (Ste) and embryonic lethality (Emb) phenotype and partial removal of both genes together revealed a genetic interaction (Fig. 3A). Second, viable homozygous offspring of lsy-22 and unc-37 heterozygous parents not only displayed a Lsy phenotype, but also uncoordinated locomotion (Fig. 3B). As in unc-37 mutants (Miller et al., 1993; Pflugrad et al., 1997), the lsy-22 phenotype is caused by the misspecification of a specific class of ventral cord motoneurons (Fig. 3C). Third, both unc-37 and lsy-22 mutants display morphological defects in the vulva (Pvl phenotype) (Fig. 3D). Fourth, removal of maternal and zygotic unc-37 through RNAi results in ectopic expression of the Notch ligand lag-2 during embryogenesis (Neves and Priess, 2005); the same phenotype was observed upon removal of lsy-22 function (Fig. 3E). Finally, in both unc-37 and lsy-22 mutants, sensory ray structures in the male tail were transformed into hypodermal cells (Pal phenotype) (Zhang and Emmons, 2002) (Fig. 3F). In summary, in addition to their shared ASE laterality defects and broad nuclear co-expression, unc-37 and lsy-22 null mutants share a significant series of phenotypes and therefore function in similar biological processes.

The spectrum of phenotypic similarities of unc-37 and lsy-22, their genetic interactions and their nuclear co-expression in many cell types suggest that both proteins might physically interact. Groucho/TLE family proteins are thought to homomultimerize via the N-terminal Q-rich domain and it has been proposed that the short Groucho family members might engage in heteromeric interactions with full-length Groucho/TLE proteins (Pinto and Lobe, 1996; Ren et al., 1999). We tested for an UNC-37–LSY-22 interaction in two different settings. First, we undertook a yeast two-hybrid analysis and indeed found that the two proteins interact with one another (Fig. 4A). In addition to interacting with UNC-37, our yeast two-hybrid experiment indicated that LSY-22 is able to interact with itself. The two-hybrid data also revealed a COG-1–UNC-37 interaction (Fig. 4A), corroborating their genetic interaction data (Fig. 1C) (Chang et al., 2003). However, we did not find two-hybrid evidence for UNC-37 homodimers; this is not because the UNC-37 prey or bait was not functional, as both interacted with LSY-22. This raises the intriguing possibility that C. elegans UNC-37 does not homomultimerize, but rather requires LSY-22 to form a functional heteromeric complex. Second, we investigated the interaction between LSY-22 and UNC-37 in vivo by co-immunoprecipitation from worm total lysates and found that endogenous UNC-37 protein is in a complex with LSY-22 (Fig. 4B). Taken together, our data suggest that COG-1, UNC-37 and LSY-22 exist in a complex that controls ASE cell fate (Fig. 4C).

Our genetic and physical interaction tests indicate that LSY-22 and UNC-37 act together to control a number of distinct developmental processes. We propose that UNC-37 does not exist in a homomultimeric state, but rather in a heteromultimeric complex with LSY-22, possibly, although not necessarily, mediated through Q-rich domain interactions. Our study therefore provides a novel perspective on the small Groucho proteins that were previously thought to antagonize canonical Groucho proteins. Such antagonistic relationships were exclusively inferred from overexpression studies that often required significant amounts of exogenously added short Groucho proteins (Buscarlet and Stifani, 2007). We also find that overexpression of lsy-22 can mimic certain loss-of-function phenotypes of unc-37. lsy-22-overexpressing animals display a mild ASE laterality defect, similar to that observed in lsy-22 mutants (see Table S1 in the supplementary material). Yet, as our loss-of-function analysis clearly shows, this mild overexpression effect is not indicative of the endogenous function of lsy-22. Rather, LSY-22 appears to promote, not antagonize, UNC-37 function, and it might do so in what we propose to be a novel co-repressor complex. The stoichiometry of this complex might be tightly regulated, such that overexpression of LSY-22 disrupts its formation. Taken together, these findings should prompt a careful examination of the composition of vertebrate Groucho/TLE-containing co-repressor complexes to better understand the important roles that they play in development and disease.

We thank Q. Chen for expert DNA injection; A. Boyanov for expert handling of the Illumina Genome Analyzer; Illumina for providing the ot114 genome sequence data; S. Mitani at Tokyo Women's Medical University School of Medicine for the lsy-22(tm3750) allele; D. Miller for advice and reagents; and members of the O.H. laboratory for discussion and comments on the manuscript. We acknowledge funding by the NIH (R01NS039996-05; R01NS050266-03 to O.H.) and the NSF (NSF-CISE 0829882 to I.P.). B.T. is funded by a Francis Goelet Postdoctoral Fellowship. R.J.P. is a Charles H. Revson Senior Fellow in Biomedical Science. O.H. is an Investigator of the HHMI. Deposited in PMC for release after 6 months.