Goosecoid (Gsc) is a homeodomain protein expressed in the organizer region of vertebrate embryos. Its Drosophila homologue, D-Gsc, has been implicated in the formation of the Stomatogastric Nervous System. Although there are no apparent similarities between the phenotypes of mutations in the gsc gene in flies and mice, all known Gsc proteins can rescue dorsoanterior structures in ventralized Xenopus embryos. This paper describes how D-Gsc behaves as a transcriptional repressor in Drosophila cells, acting through specific palindromic HD binding sites (P3K). D-Gsc is a 'passive repressor' of activator homeoproteins binding to the same sites and an 'active repressor' of activators binding to distinct sites. In addition, D-Gsc is able to strongly repress transcription activated by Paired-class homeoproteins through P3K, via specific protein-protein interactions in what is defined as 'interactive repression'. This form of repression requires the short conserved GEH/eh-1 domain, also present in the Engrailed repressor. Although the GEH/eh-1 domain is necessary for rescue of UV-ventralized Xenopus embryos, it is dispensable for ectopic induction of Xlim-1 expression, demonstrating that this domain is not required for all Gsc functions in vivo. Interactive repression may represent specific interactions among Prd-class homeoproteins, several of which act early during development of invertebrate and vertebrate embryos (Mailhos, 1998).

Gsc proteins are homeodomain transcription factors belonging to the Prd-class. This class of homeoproteins is unique, since their HDs are able to cooperatively homo- and hetero-dimerize on P3 palindromic sequence motifs composed of two characteristic TAAT HD core sequences arranged as a palindrome with a 3 bp spacing. Depending on which residue occupies position 50 (Q50, S50 or K50), the Prd-class HDs recognize P3 sites with different levels of cooperativity and with different spacer sequences between the TAATs. For K50 HDs such as the Gsc or Otd proteins, the optimal sequence is TAAT CCG ATTA and is called P3K. Although Bcd also bears a K50 HD, its HD does not belong to the Prd-class and binds DNA as a monomer. The optimal binding sequence defined for Bcd is the non-palindromic TAATCCC site, which corresponds to a P3K half site. In order to test whether D-Gsc could modulate transcription, co-transfection experiments were carried out in S2 Drosophila Schneider cells. Initially, a reporter construct containing a single copy of the monomeric Bcd binding site (BcdBS: TAATCCC) through which Bcd could drive CAT expression was used as the simplest site to measure D-Gsc activity. Following co-transfection of the 1xBcdBSCAT reporter and an expression plasmid driving D-gsc (pPAC-Dgsc), CAT activity levels remained indistinguishable from background. These data suggest that D-Gsc was not an activator of transcription, providing the protein was expressed and could bind to the Bcd site. To test whether D-Gsc acts as a repressor of transcription expression plasmids were used for the transcriptional activators Bcd and Otd. Both proteins could activate transcription through the Bcd site in co-transfection assays. An amount of activator expression plasmid was used that resulted in near saturation levels of transcription, as measured by reporter gene activity. This amount was used in all subsequent repression assays in which only the amount of pPAC-Dgsc was varied. When both pPAC-bcd and pPAC-Dgsc were co-transfected, Bcd-driven CAT activity was down-regulated and brought down to basal levels with increasing amounts of pPAC-Dgsc. The same situation was observed when pPAC-Dgsc was co-transfected with pPAC-otd. When equimolar amounts of D-gsc and either otd or bcd expression plasmids were co-transfected, CAT activity decreased by approximately 50%. These data, together with gel retardation assays using nuclear extracts from cells transfected with either D-Gsc, Otd or Bcd producer plasmids and P3K as a probe, suggest that the amount of each protein present in transfected cells correlates with the amounts of expression plasmid transfected. Thus, when using the monomeric Bcd site, the down-regulation of Otd and Bcd-activated transcription by D-Gsc appears to occur through passive competition for binding (passive repression) (Mailhos, 1998).

Transcriptional repressors have become the focus of intense research as a consequence of their important regulatory function. Homeoprotein repressors are of particular interest since they are key developmental regulators that must achieve functional specificity despite their highly related DNA binding specificity. This paper has shown that the D-Gsc homeoprotein acts as a strong repressor of transcription by three distinct mechanisms: passive, active and interactive repression. Passive and active repression by homeoproteins have been previously described, in particular with reference to En. Repression of Bcd-driven activation by DGsc through the monomeric Bcd site (1xBcdBS) is likely to involve passive repression i.e. competition for binding sites. Repression of Bcd-driven activation thorough multiple copies of the monomeric Bcd site (3xBcdBs) could be due to active or passive repression depending on how DGsc binds to this site. Active repression by noncooperative D-Gsc binding to 3xBcdBS is favored based on previous analysis of Prd-class HD binding. When using the dimeric palindromic P3K site, cooperative dimerization of D-Gsc on this site would render it a very effective competitor of the Bcd monomer, resulting in efficient repression of Bcd-driven activation. Consistent with this model, repression of Bcd activation is significantly reduced when using the monomeric Bcd binding sites or a D-Gsc molecule unable to dimerize. Further evidence that D-Gsc could act as an active repressor is its ability to strongly repress Otd-driven transcription through multiple copies of P3K. Considering both homeoproteins can cooperatively dimerize on P3K, competition for binding is unlikely to yield effective repression when D-Gsc is present at very low levels. The strong repression observed could be explained by D-Gsc binding to a single copy of P3K (either as a homodimer or in a Otd:D-Gsc heterodimer when present at low levels, see below) and repressing transcription from the remaining copies of P3K (Mailhos, 1998).

Active repression is the bona fide mechanism by which DGsc represses glucocorticoid receptor (GR) driven activation since GR and D-Gsc bind to different sites. There are two general views of how active repression might work. One is that the repressor contacts the transcriptional machinery directly or via a general cofactor, and thus blocks transcription irrespective of the activator (direct repression). Alternatively, the repressor could interact with the activator, either directly or indirectly through a cofactor, and thus block the effect of that particular activator (quenching). In both cases, the repressor does not act by competing for binding sites. D-Gsc effectively represses GR-driven activation when both activator and repressor sites are in close proximity to each other and the promoter, suggesting that D-Gsc could act both through quenching and/or through direct contact with the transcriptional machinery. Consistent with short-range repression by D-Gsc, its repressor activity is impaired once P3K is 700 bp away from the GRE. However, significant repression of GR-driven activation is observed with high levels of D-Gsc when using the (P3K)3-700bpGRE reporter indicating that, at high enough levels, D-Gsc will interfere with transcription irrespective of the position of its binding sites (Mailhos, 1998).

Interactive repression is referred to as a form of repression that involves protein-protein interactions between repressor and activator homeoproteins. This type of repression is illustrated by the repression of Otd activation by D-Gsc, which depends on the presence of the HD and GEH/eh-1 domain in D-Gsc, and on the presence of a Prd-class HD in the activator protein. The conservation of the eh-1 domain which is similar to the GEH/eh-1 domain but is found in other classes of homeoproteins such as En, has led to the proposal that it mediates protein-protein interactions (Mailhos, 1998). Subsequent studies showed that the eh-1 domain interacts with Groucho (see Tolkunova, 1998).

The eh-1 domain has been shown to be an important determinant of the ability of En to repress transcription in vivo. The En eh-1 is neither the only region that contributes to En repressor activity, nor is it required for all En functions. Similarly, the interactions mediated by the GEH/eh-1 domain in D-Gsc are unable to account for active repression: ΔGEH is as efficient as wt D-Gsc in repressing GR and Bcd-driven activation and requires a general repression domain present between aa98 and 218 in D-Gsc. The fact that ΔGEH is unable to rescue UV-ventralized Xenopus embryos but can still induce Xlim 1 expression, reinforces the notion that the GEH/eh-1 domain is only required for a subset of Gsc functions in vivo. The GEH/eh-1 domain is present in divergent classes of homeoproteins (e.g., En, Msh, NK), suggesting that this domain is ancient, and must have been present before the duplication events that resulted in the various HD classes in which eh-1 has been identified. The presence of this domain in a subset of homeoproteins may be a necessary additional feature they possess to achieve increased specificity. This might also be true for another short amino acid stretch, the conserved heptapeptide that mediates interactions between the Hox and En homeoproteins and the Pbx/Exd family of homeoproteins. Despite mediating interactions with the same family of proteins (Pbx/Exd), the heptapeptides of Hox and En proteins are only distantly related. Thus, it is possible that the GEH/eh-1 domain of En and D-Gsc (which are different from the hexapeptide), may have distinct functions, having evolved independently from each other, but in parallel with the factor with which they interact. Thus, the D-Gsc GEH/eh-1 domain may have evolved to interact with Prd-class homeoproteins as shown by the inability of ΔGEH to repress activation brought about by Otd, PrdK50 and BcdOtdHD (Mailhos, 1998).

Interestingly, both Gsc and Otd/Otx are co-expressed in flies and vertebrates and are thought to belong to a conserved regulatory network. There is indirect evidence that Gsc may act as both an activator and repressor of Otd in vivo. In Drosophila, although otd regulates the early anterior expression pattern of D-gsc, any effect of D-gsc on otd targets must be redundant, since D-gsc mutant embryos, unlike otd mutant embryos, develop normal head structures. Other Prd-class homeoproteins may also be expressed early in development and their activator function be affected by Gsc repression (Mailhos, 1998).

The HD and the GEH/eh-1 domains are the only common features among Gsc proteins. All forms of repression by DGsc, however, require a general repression domain outside these conserved regions which shares no homology in primary sequence with other Gsc molecules. The repressor domain in D-Gsc has been narrowed to 120 aa between the GEH/eh-1 domain and the HD. The repressor function of D-Gsc appears to be conserved amongst Gsc proteins, despite the lack of sequence homology of the repressor domain, since X-gsc is also a transcriptional repressor behaving similarly to D-gsc in (Mailhos, 1998).

Differences in homeoproteins potency to activate or repress transcription may partially account for their functional specificity. Although the HD itself does not appear to mediate repressor function, it substantially contributes to the targeting and protein-protein interactions which potentially define homeoprotein function. The Prd-class HDs can interact with other HDs of the same class to cooperatively dimerize. It can also interact with other protein domains to mediate repression or activation of gene expression. Heterodimeric transcription factor complexes offer several advantages. First, they allow for more elaborate regulation of transcription as production of each partner can be independently regulated. Second, their interactions allow a small number of proteins to generate a large number of combinations of transcription factors with different binding specificities and functions. This strategy would be particularly useful during development (Mailhos, 1998).