BIOLOGICAL OVERVIEW

A screen for modifiers of Dpp adult phenotypes led to the identification of the Drosophila homolog of the Sno oncogene (once termed snoN in FlyBase). The SnoN locus is large, transcriptionally complex and contains a recent retrotransposon insertion that may be essential for SnoN function. This is an intriguing possibility from the perspective of developmental evolution. SnoN is highly transcribed in the embryonic central nervous system and transcripts are most abundant in third instar larvae. SnoN mutant larvae have proliferation defects in the optic lobe of the brain very similar to those seen in baboon (Activin type I receptor) and Smad2 mutants. This suggests that SnoN is a mediator of Baboon signaling. SnoN binds to Medea and Medea/SnoN complexes have enhanced affinity for Smad2. Alternatively, Medea/SnoN complexes have reduced affinity for Mad such that, in the presence of SnoN, Dpp signaling is antagonized. It is proposed that SnoN functions as a switch in optic lobe development, shunting Medea from the Dpp pathway to the Activin pathway to ensure proper proliferation. Pathway switching in target cells is a previously unreported mechanism for regulating TGFß signaling and a novel function for Sno/Ski family proteins (Takaesu, 2006).

A second study has found that SnoN inhibits growth when overexpressed, indication of a oncogene role in flies. It can act in multiple tissues to selectively and cell autonomously antagonise signalling by TGF-β ligands from both the BMP and Activin sub-families. By contrast, analysis of a snoN mutant indicates that the gene does not play a global role in TGF-β-mediated functions, but specifically inhibits TGF-β-induced wing vein formation. It is proposed that SnoN normally functions redundantly with other TGF-β pathway antagonists to finely adjust signalling levels, but that it can behave as an extremely potent inhibitor of TGF-β signalling when highly expressed (Ramel, 2006),

The oncogene v-ski was originally identified in an avian Sloan-Kettering virus via its ability to transform chick embryo fibroblasts. Sno (a ski-related novel gene) shares significant amino acid identity with Ski, and Sno overexpression also causes transformation. In transfected mammalian cells Sno and Ski can form multimeric complexes and act as components of a histone deacetylase complex that represses transcription (Nomura, 1999). Sno is present in a single copy in the human genome but multiple promoters and alternative splicing generate six distinct Sno transcripts in humans (Nomura, 1989). Four isoforms of the Sno protein have been identified with the longest isoform (Pearson-White, 1997) known as SnoN (Takaesu, 2006).

Numerous studies in mammalian cells have shown that SnoN antagonizes signal transduction pathways downstream of TGFß/Activin proteins. In brief, TGFß/Activin signal transduction involves the activation of Smad2, the formation of Smad2/Smad4 complexes, and the translocation of the complex into the nucleus where it stimulates transcription. Cell culture studies show that, in the absence of TGFß/Activin proteins, Sno physically interacts with Smad2 and Smad4, repressing their transcriptional ability. Alternatively, when TGFß/Activin ligands are present, Sno is rapidly ubiquitinated and degraded, permitting these Smads to activate target gene expression, including the transcription of Sno. This subsequent round of Sno expression leads to renewed interactions with Smads and to the attenuation of Smad-mediated gene expression (Luo, 1999; Stroschein, 1999). However, two studies suggest that Sno's role in signaling is more complex (da Graca, 2004; Sarker, 2005; Takaesu, 2006 and references therein).

Sno's function in development is also uncertain. Two studies of independently derived Sno knockout mice reached different conclusions for unknown reasons. One study shows early embryonic lethality (preimplantation, day E3.5) for homozygous mutant embryos (Shinagawa, 2000). The second study (Pearson-White, 2003) reports that homozygous mutants are viable and that these mice have a defect in T-cell activation (Takaesu, 2006).

In Drosophila, as in vertebrates, two TGFß subfamilies are present. The bone morphogenetic protein (BMP) subfamily member Dpp signals through its type I receptor Thickveins to its dedicated transducer Mad (Smad1 homolog) and the Co-Smad Medea (Smad4 homolog). The TGFß/Activin subfamily member activin signals through its type I receptor Baboon to its dedicated transducer dSmad2 and Medea (Newfeld, 2006). This study shows that dSno binds Medea and then functions as a mediator of Activin signaling by enhancing the affinity of Medea for dSmad2. Antagonism for BMP signaling likely arises as a secondary consequence of dSno overexpression. Examination of dSno loss-of-function mutants shows that dSno is required in cells of the optic lobe of the brain to maintain proper rates of cell proliferation. Given that Dpp signaling is essential for neuronal differentiation in the optic lobe (Yoshida, 2005), these data suggest that dSno functions as a switch that shunts Medea from the Dpp pathway to the Activin pathway to ensure a proper balance between differentiation and proliferation in the brain (Takaesu, 2006).

The following hypothesis is proposed to resolve what appear to be contradictory roles for dSno in TGFß signaling (antagonism and mediation): dSno is a BMP-to-Activin pathway switch in brain development. This hypothesis is consistent with the data, with other reports examining neural development in Drosophila and with studies of mammalian neural stem cells (Takaesu, 2006).

Three studies of TGFß signaling in optic lobe development support this hypothesis. A study by Yoshida (2005) showed that Dpp signaling via its receptors Thickveins and Medea is essential for the differentiation of optic lobe neuroblasts into lamina glia and neurons. This study shows that the Activin receptor Baboon via its signal transducers dSmad2 and dSno is required for the maintenance of neuroblast proliferation. Clearly, during optic lobe development neuroblasts must balance self-renewal via proliferation and the generation of neurons and glia via differentiation. Given its ability to interact with both the Dpp pathway and the Activin receptor pathway, it seems logical that in neuroblasts dSno functions as a pathway switch, shunting Medea away from forming complexes with Mad in the BMP pathway that directs differentiation and toward forming complexes with dSmad2 in the Activin pathway that directs proliferation (Takaesu, 2006).

This hypothesis is also consistent with a microarray study of neuroblast differentiation in Drosophila (Egger, 2002). In this study, widespread misexpression of glial cells missing, a transcription factor that specifies the fate of differentiating neuroblasts, significantly reduced the expression of dSno. Reduction of dSno expression in differentiated cells fits well with two of the results: (1) Sno mutants have no effect on the expression of EcR-1B because this gene is a Baboon target only in mature neurons; (2) dSno transcript accumulation is reduced in the differentiated cells behind the morphogenetic furrow in the eye disk. Further, given the analogous roles played by Dpp in the differentiation of optic lobe neuroblasts and the role played by Dpp homologs (BMPs; Bonaguidi, 2005) in the differentiation of mammalian neuronal stem cells, it is possible that Sno is also a pathway switch in mammals (Takaesu, 2006).

The current model for dSno function in optic lobe development is as follows: dSno expression is activated by an unidentified factor in a subset of optic lobe neuroblasts in third instar larvae. In these cells, cytoplasmic dSno (studies of mammalian Sno show several cell types with predominantly cytoplasmic localization) forms complexes with Medea. Subsequently, Dpp signaling from the lamina glia precursor region, capable of reaching all optic lobe neuroblasts, is incapable of inducing differentiation in neuroblasts expressing dSno. Alternatively, Activin functions as a secreted hormone (T. Haerry, personal communication to Takaesu, 2006) also capable of reaching all optic lobe neuroblasts. Those expressing dSno are capable of responding to proliferative Activin signals. Thus a balance between proliferation and differentiation is achieved in the optic lobe neuroblast population (Takaesu, 2006).

The dSno locus clearly illustrates the complexity of developmental evolution. In this case significant conservation at the protein level (isoform number, structure, and sequence) coexists with a recent transposon insertion that appears to be involved in dSno transcriptional regulation (Takaesu, 2006).

Regarding conservation, multiple protein isoforms are rare among growth factor signal transduction pathway components. Sno is the first example of a TGFß signal transducer with four conserved isoforms. The role that each alternative isoform plays is unknown, but experience with Df(2L)BSC41 suggests that the shortest (dSnoI) may be sufficient to fulfill nearly all roles, at least under laboratory conditions. Amino acid conservation in the Sno homology domain extends to a large multigene family. The topology of the Sno/CORL/Dachsund tree is similar to that of the Smad family. In both families, fly and human genes cluster together while worm genes are present in a subset of clusters and also fall between clusters. Thus, the Sno/CORL/Dachsund family tree argues against the existence of an Ecdysozoan phylum containing worms and flies. It is consistent with the traditional Coelomate classification that places flies closer to vertebrates than to worms (Takaesu, 2006).

The presence of a 297TE in the mist of the putative dSno promoter region is the only example currently known in which a transposable element may have co-opted regulatory functions for a nearby gene and thereby rendered itself indispensable. There are 57 297TE elements in the D. melanogaster genome but only 18 are full length like the 297{323} insertion in dSno. To determine if the 297TE family is an ancient one in Drosophila, its species distribution was examined by BLAST. It was found that 297TE family members are present only in species of the melanogaster group (divergence 44 million years). 297TE sequences are absent from other Drosophila species and from other insect phyla. Within the melanogaster group, the relationship between 297TE sequences mirrors the relationship between the species. Thus, the 297TE family entered Drosophila via the common ancestor of the melanogaster group (Takaesu, 2006).

Two features of 297{323} suggest that the insertion of a 297TE family member in the putative dSno promoter region is a very recent event. The 412-bp direct repeats in 297{323} are absolutely identical, indicating little time for mutations to accumulate. Second, there is no 297TE upstream of dSno in D. melanogaster's closest relatives, D. simulans and D. sechellia (divergence 5.4 million years). Thus, variation in the splice junction between exon 1 and exon 2 in transcripts generated from the proposed dSno promoter 1 and variation in start sites for transcripts generated from the potential promoter 3 likely result from natural selection for sites that bypass or incorporate the 297TE. The specific developmental processes affected by dSno transcripts generated from within the transposon are unknown, but if they are identified, they will likely be the first example of a transposon positively influencing a specific phenotype (Takaesu, 2006).

In summary, these data support the view that dSno functions as a switch in optic lobe development, shunting Medea from the Dpp pathway to the Activin pathway to ensure a proper balance between differentiation (Dpp) and cell proliferation (Activin). Pathway switching in target cells is a previously unreported mechanism for regulating TGFß signaling and a novel function for Sno/CORL/Dachsund proteins (Takaesu, 2006).

GENE STRUCTURE

Sequencing the 12 distinct dSno cDNAs (9 encode SnoN and 1 each encoding dSnoA, dSnoI, and dSnoN2) and mapping their 5'- and 3'-ends to genomic sequence by BLAST revealed that the minimum size of the dSno locus is 92 kb (see genomic structure of dSno). In addition to the alternative 3' exons that create four dSno isoforms, cDNA sequences suggest three discrete promoters that together account for six classes of dSno transcripts (Takaesu, 2006).

Unexpectedly, the dSno locus contains an insertion of a 297 family transposable element (long-terminal-repeat-bearing retrotransposon) in the midst of the putative promoter region. As a result, the 297 transposable element (297TE) is predicted to have a significant effect on dSno transcription. Further, a genetic analysis of dSno suggests that deletion of the 297TE is lethal (Takaesu, 2006).

One cDNA sequence suggests that a possible dSno promoter (promoter 2) sits inside of the 297TE element. This cDNA splices around the downstream 412-bp terminal repeat to an exon that contains the dSno initiator methionine. Four cDNAs span the 297TE and begin at the most 5' dSno exon and potentially represent transcripts initiating at promoter 1. Each of these contains distinct variations in their exon 1/2 splice junction. Two splice acceptor sites are utilized: one is 76 bp upstream of the 297TE and the other is adjacent to the 297TE terminal repeat. There are four splice donor sites: one is inside the 297TE direct repeat and three are downstream. All splice junctions joining exons 1 and 2 for cDNAs beginning at the putative promoter 1 and promoter 2 conform to the consensus. Five cDNAs begin within or immediately downstream of the 297TE element, suggesting the presence of a third promoter (promoter 3). Two of these begin within the 297TE and contain the 412-bp repeat, two begin within the 412-bp repeat, and one begins downstream of the 297TE. All five of these cDNAs read directly through to the dSno initiator methionine in exon 1 (Takaesu, 2006).

cDNA clone length - 7223 bp

Bases in 5' UTR - 1773

Exons - 3 (see genomic structure of dSno)

Bases in 3' UTR - 1778

PROTEIN STRUCTURE

Amino Acids - 1223

Structural Domains

The 297{}323 retrotransposon present in the FlyBase genomic annotation and Oregon R wild-type flies is absent from all the insertion lines analysed and from wild-type Canton S flies. Thus, the EP(2)2004 and EP(2)2510 elements are located only about 1 kb upstream of the snoN ORF, while GS18054 is located about 0.7 kb upstream (Ramel, 2006).

The annotated Drosophila SnoN protein is 338 amino acids long with a molecular weight predicted to be about 38 kDa. Blast homology search and protein-protein alignment showed that fly SnoN is orthologous to Ski/Sno proteins. It contains the Ski/Sno family domain, characterized by conserved CLPQ residues, which is known to be essential for the transforming and differentiation activities of Ski/Sno proteins. It also includes the conserved SAND domain, which, in mammals, is required to mediate binding of Ski to Smad4. However, it lacks the C-terminal region, present in chicken and human Ski/SnoN (but not mouse) that is involved in dimerization of Ski/Sno proteins. Thus, Drosophila SnoN displays most, but not all, of the conserved domains present in vertebrate Ski/Sno proteins (Ramel, 2006).

A screen for modifiers of dpp adult mutant phenotypes identified a mutant, E(dpp)46.3, that has a chromosomal inversion breakpoint in the region 28D. CG7233 is a predicted intronless open reading frame of 338 amino acids that is referred to as SnoN in FlyBase. Unfortunately, E(dpp)46.3 is no longer available. Screening of cDNA libraries with CG7233 identified 17 clones (15 from a pupal library and 2 from an embryonic library). After eliminating duplicates, 12 distinct cDNAs remained. They ranged in size from 4.1 to 7.2 kb and most included the adjacent but 70-kb distant gene prediction CG7093. While it was clear that several cDNAs were not full length, they could nevertheless be sorted into six distinct classes. From these it was determined that four protein isoforms were encoded and that these corresponded to the four human Sno proteins: SnoA, SnoI, SnoN, and SnoN2 (Pearson-White, 1997). Like their mammalian counterparts, dSno proteins share a common N-terminal region (the Sno homology domain) and have distinct C termini. Further, each C terminus has motifs such as coiled-coils that are conserved in both species (Takaesu, 2006).

Fifteen sequences from humans, mice, flies, and worms were identified with significant sequence similarity to the longest dSno isoform, dSnoN. A phylogenetic analysis of this multigene family revealed the presence of three major subfamilies (Sno/Ski, CORL, and Dachsund). This tree has considerable statistical support and is topologically similar to those of other TGFß pathway components. In all subfamilies, vertebrate and invertebrate genes cluster together, suggesting that they are homologs descended from a common ancestor like Dpp and BMP2/BMP4 (Takaesu, 2006).

The Dachsund (Dac) subfamily appears to be the oldest, since it is found in worms, flies, and mammals. These are Dach1 and Dach2 in mammals, Dac in D. melanogaster, and Dac-1 in Caenorhabditis elegans. Within a 66-amino-acid region near their N termini (referred to as the Dac-box or DS domain) human Dach1 and SnoN show 29% identity. In transfected mammalian cells, human Dach1 was shown to antagonize TGFß/Activin signaling (Wu, 2003). In Drosophila, interaction between Dac and TGFß family members is indirect: during eye development Dpp activates dac expression via the Eyes absent protein. C. elegans Dac-1 was identified in a screen for genes affecting DNA stability and has not been connected to TGFß signaling (Takaesu, 2006 and references therein).

Note that Daf-5, which does not belong to any subfamily, is not the C. elegans Ski homolog as previously reported (da Graca, 2004). Daf-5 has equal similarity to both the CORL and Sno/Ski subfamilies and may approximate their common ancestor. The CORL subfamily is not present in C. elegans and to date only CORL1 has been studied (Mizuhara, 2005). Within a stretch of 161 amino acids (completely encompassing the Dac box) mouse CORL1 and SnoN have 34% identity. CORL1 functions as a corepressor for a homeodomain protein in neurons but has not been linked to TGFß signaling (Takaesu, 2006).

The Sno/Ski subfamily is also absent from C. elegans and alignments of dSnoN with hSno and hSki indicate that dSnoN and hSnoN are more similar. Thus Ski is the newest member of this subfamily and is unique to mammals. Alignment of SnoN isoforms from human and fly shows significant homology in a region near the N terminus that is common to all Sno proteins in both species. This Sno homology domain is part of the CG7233 prediction and falls between amino acids 92 and 296 in dSnoN. Within this domain there are a number of conserved features. These include the Dac box (dSnoN 138-201), CORL domain (dSnoN 119-270), the nine-amino-acid destruction box/APC recognition site, and the three amino acids that interact with Smad4 (Wu, 2002). One difference between the species is that dSnoN is missing the Smad2/3 interaction domain found in hSnoN. Downstream of this region is the splice to CG7093 and the rest of the protein corresponds to that predicted gene (Takaesu, 2006).

In the central domain of the protein, dSnoN is missing the ubiquitinated lysines important in the regulation of hSnoN activity (Stroschein, 2001). Interestingly, these lysines are also missing in hSnoN2 (the internally deleted form of SnoN). In both cases, it is possible that other lysines are targeted for ubiquitination. A 164-amino-acid stretch at the C terminus of hSnoN and hSnoN2 forms a coiled-coil domain that is important for dimerization (Wu, 2002). In dSnoN, there are two coiled-coil domains that together contain 99 residues separated by 495 amino acids. The first domain is present in all dSno isoforms but the second is seen only in dSnoN and dSnoN2. In dSnoN2 the distance between the domains is 230 amino acids. An alignment of the distinct 3'-ends of hSnoA and dSnoA shows this region is roughly similar in size (52 vs. 62 amino acids) and moderately conserved (31% amino acid similarity). The most highly conserved region is the C-terminal 9 amino acids that are 33% identical and 66% similar (Takaesu, 2006).

Overall, proteins of the Sno/CORL/Dachsund family show substantial evolutionary conservation: the Sno homology domain is easily discernible across species while fly and human Sno have the same number of structurally similar isoforms (Takaesu, 2006).

date revised: 15 March 2007

Home page: The Interactive Fly © 2006 Thomas Brody, Ph.D.

The Interactive Fly resides on the
Society for Developmental Biology's Web server.