GAGA factor, a positive regulator of global gene expression, modulates transcriptional pausing and organization of upstream nucleosomes

In this report, we used several approaches to address the role of GAF in transcriptional regulation in broad tissue contexts. Our results support that GAF acts as a global activator and is required to maintain proper level of paused RNA-Pol and organization of upstream nucleosomes.

As expected from prevalent GAF occupancy in the genome, expression of ~20 % of coding genes (2912 genes) is affected by Gaf mutation. Based on its occupancy near the TSS, we estimate that at least 9 % of them are directly regulated by GAF. Interestingly, about 15 % of these direct targets (40 genes) are listed in FlyTF as putative transcriptional factors (Additional file 2: Table S1), of which nearly half have previously been validated experimentally and are known to carry out important developmental functions such as embryonic segment formation and cell fate determination. Conceivably, these transcriptional factors could further control the expression of a much larger spectrum of downstream genes that lack direct GAF binding. Either directly or indirectly, the vast majority (~80 %) of these genes are down-regulated in Gaf mutants. These results demonstrate that GAF functions primarily as a global transcriptional activator. It is worth noting that many of these developmental regulators were not identified in previous studies [26, 51]. Therefore, our results significantly expand the global function of GAF, particularly in the developmental processes.

For the inducible Hsp70 gene, we demonstrate that GAF affects transcription primarily at post-initiation steps. Both cytogenetic and molecular analyses of the Hsp70 gene showed that the levels of Ser-5p and Ser-2p, but not Hypo-p, are substantially affected in Gaf mutants following heat induction. Since phosphorylation of RNA-Pol occurs after transcriptional initiation, these results clearly indicate that GAF exerts its effects at post-initiation steps. Our studies also revealed reciprocal changes for Ser-5p and Ser-2p in Gaf mutants; accompanying the large decrease in promoter-proximal Ser-5p signal, Ser-2p showed a substantial increase toward the distal end of Hsp70. Given the obligatory order for the appearance of Ser-5p and Ser-2p during transcription, these observations strongly suggest that the reduction of Ser-5p signal represents the primary effect of Gaf mutation. The increase in Ser2-p seems to be at odds with the observation that the majority of genes are down-regulated in Gaf mutants. Since GAF can physically interact with NELF and since active transcription of Hsp70 occurs when NELF is dissociated from RNA-Pol [21, 68], it is plausible that less NELF is recruited to Hsp70 promoter in Gaf mutant, resulting in less efficient withholding of paused RNA-Pol. In addition, we note that Hsp70 expression can still be induced by the potent activator HSF during heat shock from a transgene lacking GAF binding sites [69]. Thus, Hsp70 may represent an unusual case in which multiple factors are critically involved in its activation. GAF knockdown in cultured cells has recently been reported to affect the accumulation of total RNA-Pol only in promoter-proximal region [40]. Since the individual contribution of Ser-5p and Ser-2p could not be distinguished in that assay and since we show that there is a reciprocal relationship between these two isoforms particularly in distal regions, it is very likely that the effects of GAF depletion were not fully evaluated.

Our results further show that the effect of GAF on Ser-5p is not unique to Hsp70. We found that Gaf mutation results in reduced Ser-5p, but not Hypo-p or Ser-2p, on polytene chromosomes, suggesting that GAF can modulate the level of Ser-5p globally. This is substantiated by our analyses of the genome-wide distribution of RNA-Pol. In WT samples, GAF occupancy in the upstream region is positively correlated only with the RNA-Pol density in the promoter-proximal region or with PI. These results clearly reveal a dose-dependent relationship between paused RNA-Pol and GAF. Importantly, this positive correlation was lost in Gaf mutants, indicating a direct role of GAF in RNA-Pol pausing. Contrary to the reduction of paused RNA-Pol in Gaf mutants, GAF over-expression results in an overall increase in Ser-5p, but not other isoforms. These complementary results further strengthen the conclusion that GAF is critical for transcription pausing.

Consistent with earlier studies [26, 68], our analyses of genome-wide nucleosome profiling revealed remarkable differences between GAF-target and non-target genes. Much lower levels of the nucleosome array were found in the transcribed region of GAF-target genes, indicating that paused genes contain less downstream nucleosomes. Although somewhat anti-intuitive, these patterns are consistent with recent studies showing a negative correlation between PI and downstream nucleosome occupancy [23]. Apparently, these observations do not support a general notion that downstream nucleosomes act directly as a barrier for RNA-Pol, triggering its pausing. Interestingly, these profiles appear to coincide closely with the SS-dinucleotide content predicted to favor nucleosome occupancy, suggesting a close relationship between them.

In contrast to the downstream region, GAF-target genes contain more nucleosomes around the TSS and immediately upstream region than non-target genes. Overall, these patterns are consistent with the SS-dinucleotide content of the upstream region. However, considering the strong enrichment of SS-dinucleotide content in GAF targets, the increased nucleosome levels seem to be somewhat modest. It is also puzzling that nucleosomal profiles of different ranks of target genes appear to be largely indistinguishable, despite there being more extended SS-dinucleotide sequences in higher ranks. Nevertheless, the levels of nucleosomes were selectively increased in upstream regions from the TSS up to ?500 for rank III and more so for rank IV in Gaf mutants, resulting in closer matches to predicted nucleosome profiles. Thus, the upstream sequences of these genes may intrinsically favor higher levels of nucleosomes, potentially producing stronger physical constraints for the loading and assembly of transcriptional machinery. Apparently, the presence of high levels of GAF relieves such constraints by reducing the amount of nucleosomes and consequently makes promoters more accessible. Like RNA-Pol pausing, the modulation of nucleosome profiles by GAF is highly dependent upon the degree of occupancy. This dosage-dependent effect strongly supports a direct role of GAF. Thus, caution should be taken when making inferences about GAF’s role in regulation based on the GAF motif alone.

Since GAF can recruit NURF to remodel nucleosome organization in vitro [38] and since our genetic studies showed that NURF is involved in GAF-mediated regulation, we suggest that GAF may act through NURF to remodel upstream nucleosomes and subsequently facilitate the loading and assembly of factors involved in regulation and general transcription. Based on the observation that a reduction of paused RNA-Pol is correlated with elevated upstream nucleosomes [23, 59], an alternative might suggest a different sequence of events. However, this possibility is not supported by our observations that nucleosome occupancy is not affected in non-target genes with significantly reduced RNA-Pol and the nucleosome occupancy is elevated in intergenic regions in Gaf mutants.

The role of GAF in RNA-Pol pausing and nucleosome remodeling has recently been reported, based on similar studies of a cultured cell line [40]. While our results on RNA-Pol pausing are quite similar, the effects on nucleosomal remodeling are somewhat different. In our study, nucleosome perturbation is found only in the upstream region where maximal GAF occupancy is located, while the earlier study has shown that the effect is exerted on more extended region including TSS and further downstream. Since our collections of GAF targets differ by ~50 %, this difference may reflect the selection of different gene pools. It is also worth noting that multiple tissues are used in our study. Thus, the effect we have observed may represent a more general scheme in living organisms.

In addition to NURF, our study shows that NELF and BAB1 are physiologically relevant to GAF’s function. Since NELF can physically interact with GAF and RNA-Pol, GAF may also regulate pausing through recruitment of NELF. However, comparison of the available data suggests that GAF targets overlap only partially with those of NELF or NURF [30, 70] (Additional file 1: Fig. S11). Therefore, GAF may act cooperatively with individual or multiple factors, depending upon specific targets. Furthermore, earlier studies suggest that BAB1 and GAF share a common interacting protein, TAF3, a component of TFIID essential for the assembly of the pre-initiation complex [71, 72]. Our findings that BAB1 binding sites are enriched specifically in rank IV genes and that there is a strong genetic interaction between bab1 and Gaf mutants support cooperation between GAF and BAB1 in controlling transcriptional pausing. These results clearly show that the upstream region of GAF targets is intrinsically prone to nucleosome assembly, resulting in promoter occlusion. By its ability to trigger nucleosome re-organization, GAF may facilitate the promoter accessibility, assembly of the basal transcriptional machinery and its transition into initiation (Fig. 7). Subsequently, the activity of pausing factors recruited by GAF may arrest RNA-Pol at the pausing state, awaiting developmental or environmental cues. Thus, GAF appears to act as a central player of this regulatory circuitry.

Proposed model for GAF-mediated transcriptional pausing. Genes with high SS-dinucleotide contents around and upstream of the TSS are occupied by much higher levels of upstream nucleosomes (dark circle) than others. This organization prevents the loading and assembly of the transcriptional complex. The occupancy of GAF results in the recruitment and cooperative interactions with NURF, NELF and BAB1 to facilitate local nucleosome remodeling and provide accessible space for the assembly of the transcriptional complex and subsequently triggers its transition to the paused state