Transcriptional Regulation: It Takes a Village

Abstract

A FASEB conference on “Transcriptional Regulation during Cell Growth, Differentiation and Development” met in June, 2008, just outside of Aspen in Snowmass Village, Colorado. The meeting covered a broad range of topics, including the structure of transcription factors (TFs), Preinitiation Complex (PIC) assembly, RNA polymerase II (Pol II) pausing, genome-wide patterns of histone modifications, and the role of TFs in development.

Introduction

Organized by Jessica Tyler (University of Colorado School of Medicine), Michael Carey (UCLA), and Victor Corces (Emory University), the FASEB conference brought together molecular, cellular, and developmental biologists with a shared interest in transcriptional regulation. Reflecting the different areas of expertise among the conference participants, the research presented utilized an impressive arsenal of genetic, biochemical, biophysical, and computational methods. Only by combining these techniques can one begin to understand and appreciate the coordination and complexity required to appropriately regulate transcription. Indeed, it takes an extensive array of factors to control expression of any given gene. Moreover, it is becoming clear that, despite obvious shared features of active and of inactive genes, many genes likely possess their own distinct regulatory signature. That is, their expression relies upon distinct sets of coregulators and/or subtle differences in chromatin state. Major themes discussed at the meeting are outlined in the sections below.

Paused RNA Polymerase II

A number of investigators described recent studies involving paused Pol II. One of many outstanding questions relates to how paused Pol II might potentiate gene expression. Karen Adelman (NIEHS) described one means by which this can be achieved: paused Pol II blocks nucleosome encroachment at the promoter (Gilchrist et al., 2008). The Negative Elongation Factor (NELF) can induce promoter-proximal stalling of Pol II to repress expression of numerous genes. By examining NELF-depleted Drosophila S2 cells, the Adelman lab discovered that 70% of genes affected by NELF were downregulated upon NELF depletion. Further, Pol II occupancy was concomitantly reduced at these genes. So how does the NELF-dependent stall impact Pol II recruitment? At downregulated genes, an increase in histone occupancy correlated with reduced Pol II occupancy, suggesting that paused Pol II can poise a gene for activation by maintaining a nucleosome-free region proximal to the transcription start site (TSS). Consequently, NELF depletion downregulates this subset of genes (Figure 1). Adelman also noted that many of the NELF-regulated genes respond rapidly and transiently to environmental stimuli, arguing that Pol II stalling facilitates fine-tuned control of gene induction.

Figure 1. A Positive Role for Stalled Pol II and NELF in Gene Expression.

Model summarizing how NELF/Pol II (rocket) may occlude nucleosome occupancy at a subset of promoters. Knockdown of NELF reestablishes nucleosome occupancy at these genes. Figure courtesy of Karen Adelman.

John Lis (Cornell) discussed results using a novel technique termed GRO-Seq that combines global run-on transcription followed by sequencing (after amplification) of the newly transcribed RNA. This method revealed that 40% of active genes in human fibroblasts show strong evidence of paused Pol II, suggesting that gene activity is regulated by escape from the pause. GRO-Seq also provided evidence for paused Pol II in the antisense direction (i.e., divergent transcription upstream from the TSS). Lis proposed that divergent transcription can explain why Pol II occupancy profiles resemble bell-shaped curves that peak at the TSS. Short, antisense transcripts were also highlighted by Rick Young (MIT), who surmised that antisense transcription occurs at most genes based in part upon global Pol II and histone H3 trimethylated at lysine 4 (H3K4me3) occupancy profiles. Later in the meeting, Toshi Tsukiyama (FHCRC) outlined a means by which antisense transcription can be suppressed in yeast by the ISW2 remodeling complex (see below), offering clues about how this process might be regulated. However, much remains to be resolved regarding the mechanism by which antisense transcripts are initiated, their frequency genome-wide, and their biological significance.

Examining paused Pol II in Drosophila, Mike Levine (UC Berkeley) noted that genes silenced by the repressor Snail still show paused Pol II, yet Pol II is released when Snail is absent, suggesting that Snail is indirectly responsible for “holding” Pol II in the paused state. Elucidation of the required intermediary factor(s) remains a critical question for future study. Levine concluded by outlining preliminary findings that suggest an unanticipated regulatory role for paused Pol II: coordinating gene induction within a population of cells. Comparing the MEF2 and heartless genes, Levine noted that heartless, which has paused Pol II, is coordinately activated within a cell population, whereas MEF2—which does not display paused Pol II—is induced randomly within a population. Clearly, more remains to be uncovered regarding the regulatory roles of paused Pol II.

Regulatory Elements within the Genome

In an unexpected finding from ChIP-chip analysis in Drosophila, Levine and coworkers discovered secondary or “shadow” enhancers that augment regulation from the primary enhancer. Levine noted that shadow enhancers appear widespread in the fly and are sufficient to direct developmentally correct expression upon mutation of the primary enhancer. Levine postulated that these secondary elements help ensure fidelity of developmental programs and might also drive evolutionary changes, as these elements appear to evolve more rapidly than the principal enhancer. Later in the meeting, Ken Zaret (Fox Chase) briefly described similar observations of “cryptic enhancer elements” in the mouse.

Several talks related some intriguing clues about the complexities of TF binding sequences and how distinct sequence elements play key regulatory roles. Keith Yamamoto (UCSF) provided evidence that different Glucocorticoid Response Elements (GREs) direct distinct transcriptional programs (So et al., 2007). The role of sequences flanking the GRE varies among GREs that regulate different subsets of genes, yet the flanking sequences for any given GRE are highly conserved in mammals, suggesting a critical regulatory role. Indeed, Yamamoto outlined a series of studies indicating that alterations in the GRE that do not impact Glucocorticoid Receptor (GR) binding affinity nevertheless trigger changes in gene expression. Structural analysis showed that a loop region adjacent to the GR DNA-binding domain altered its conformation when bound to distinct GR-binding sequences. Together, this provided a straightforward mechanism by which gene-specific regulatory information, encoded within the GRE, might direct gene-specific regulatory programs. Also focusing on nuclear receptors, Mair Churchill (University of Colorado School of Medicine) discussed transcriptional regulation by the progesterone receptor (PR) and how its activity is modulated by a short unstructured C-terminal extension motif that is located between the DNA-binding and the ligand-binding domains. Through structural and functional studies, Churchill outlined how this motif regulates PR binding to auxiliary factors, which in turn might direct the activation of specific subsets of PR target genes (Roemer et al., 2008).

Barbara Graves (University of Utah) described the genome-wide localization of all 27 Ets family TFs in Jurkat cells (Hollenhorst et al., 2007). This analysis revealed that 30% of Ets targets are “specific,” whereas 70% are “redundant”; that is, they bind multiple Ets family members. Redundant sites typically occurred at housekeeping genes and had strong consensus sequences, whereas specific targets lacked a consensus site. Interestingly, no particular class of genes appeared to be represented by the “specific” target sites. In a talk focused on core promoter elements, Jer-Yuan Hsu from the Kadonaga lab (UCSD) noted that the Drosophila TF Caudal activates genes with a specific core promoter architecture, providing a clear means by which the composition of the core promoter can help control developmental processes.

Several talks focused on insulators, which block the action of enhancers and silencers when placed between a promoter and these regulatory elements. In Drosophila, the 12 Zinc Finger (ZnF) DNA-binding protein Su(Hw) is necessary for insulator function of the gypsy retrotransposon. Pam Geyer (University of Iowa) asked whether endogenous Su(Hw) sites function as insulators by deleting the 1A2 cluster of Su(Hw) sites. Based on transgenic studies, 1A2 was predicted to be an insulator that promotes the regulatory independence of the yellow and achaete genes. Yet deletion of 1A2 did not affect expression of either gene and, instead, altered expression of a previously unknown noncoding RNA that lies between them. Thus, Su(Hw) binding sites show context-dependent activity. This finding raises the interesting question of how different configurations of Su(Hw) binding sites and other cis-regulatory elements are read out to direct insulator or transcriptional activation function.

The only mammalian protein implicated in insulator function is an 11 ZnF protein, CTCF, which can promote interaction or looping of distant sequences. In the β-globin locus control region (LCR), one of the five DNase hypersensitive sites (HS), HS5, binds CTCF and acts as an insulator in an enhancer blocking assay. Ann Dean (NIDDK, NIH) presented analyses of mice with a human β-globin locus transgene that contained a second copy of HS5 placed between the LCR and downstream genes. Whereas HS5 normally loops with 3′HS1 downstream of the locus, two loops formed in the transgenic line: HS5 to 3′HS1 and also to ectopic HS5. This new looping pattern correlates with impaired TF recruitment at HS2 and HS3 (which normally activates β-globin) and with impaired interaction of these LCR sites with the gene. Thus, ectopic HS5 appears to reconfigure chromatin loops to block enhancer-gene interaction. Also focusing on CTCF, JJ Miranda (UCSF) put forth a model for the molecular architecture of CTCF based upon biochemical and biophysical analysis of its distinct domains and fragments. Notably, this model postulates that CTCF might direct local alterations in DNA topology via potential “combinatorial” binding patterns of its 11 different ZnF domains. Also, Andreas Reik (Sangamo Bio-Sciences, Inc.) described how zinc finger nucleases can be used for targeted gene disruption or insertion with no detectable off-target cleavage within the genome. He further outlined how these powerful tools for genome manipulation could hold promise for clinical applications and cell engineering.

Finally, Tom Sexton from Peter Fraser’s lab (Babraham Institute) presented work suggesting that sequence-specific TFs might help target genes to transcription factories, discrete nuclear foci that are the sites of expression of most, if not all, genes. Using chromosome conformation capture-on-chip, they identified sequences associated with the transcribed β-globin locus. Notably, genes containing binding sites for the EKLF TF were corecruited to specific transcription factories whereas mutation of EKLF altered the colocalization of these genes. These data suggest that coordinate expression of specific subsets of genes could be orchestrated in part by TF-dependent recruitment of these genes to transcription factories.

Chromatin Remodelers

The ISWI and SWI/SNF complexes both remodel chromatin in an ATP-dependent manner, yet their mechanism of action and their biological roles are distinct. Several investigators provided further insight into the structure and function of these essential factors. Geeta Narlikar (UCSF) probed the molecular mechanism of the ACF remodeler, which contains the ISWI/SNF2h ATPase. Unlike SWI/SNF, which disrupts ordered arrays and promotes a dynamic architecture, ACF yields highly ordered chromatin structures. Narlikar provided biophysical studies suggesting the structural basis for the fundamentally different properties of these two families of remodelers.

To investigate the biological role of ISWI, John Tamkun (UCSC) and coworkers previously generated ISWI mutant flies and noted that loss of ISWI correlated with loss of histone H1 at the male X chromosome (Corona et al., 2007). Following up on this work, Tamkun presented data that further delineated the regulatory interplay between ISWI and H1. Focusing on the yeast ortholog of ISWI, ISW2, Toshi Tsukiyama (FHCRC) demonstrated that ISW2 targets more than 1000 sites in the budding yeast genome with a roughly equal distribution between 5′ and 3′ ends of genes (Whitehouse et al., 2007). Further analysis revealed that ISW2 likely maintains chromatin integrity around nucleosome-free regions. One striking consequence is that ISW2 activity suppresses antisense transcription at these sites (see above).

Transitioning to the SWI/SNF family of remodelers, Trevor Archer (NIEHS) described recent studies involving the BRG1 ATPase within human SWI/SNF. The helicase/SANT-associated (HSA) domain within BRG-1 was shown to be critical for SWI/SNF-dependent activation (Trotter et al., 2008). This was not an effect of remodeling per se, but rather, the HSA domain stabilized the association of SWI/SNF subunit BAF250a with BRG1. This finding suggests that BAF250a might stabilize SWI/SNF at target promoters and/or augment the activity of the BRG1 ATPase via interactions with the BRG1 HSA domain.

Outlining an expanded role for the INO80 chromatin remodeling complex, Joan Conaway (Stowers Institute) discussed recent findings regarding its Uch37 subunit, which possesses deubiquitinase activity. A population of Uch37 exists associated with either INO80 or the 19S proteasome regulatory complex, yet only when associated with 19S is the Uch37 deubiquitinase activated. However, INO80-associated Uch37 can be activated by transient interaction of the INO80 complex with the proteasome, suggesting that the proteasome and the INO80 complex might cooperate to regulate transcription or DNA repair. This mechanism involves dynamic histone exchange, which was a theme at this meeting.

Indeed, given the large number of chromatin remodelers within the nucleus (perhaps as many as one for every two to three nucleosomes), it is likely that remodelers, together with histone chaperones, contribute significantly to histone turnover. Steve Henikoff (FHCRC) presented data supporting histone turnover as a prominent feature of active loci. He started by noting that the genome-wide distribution pattern of histone H3.3 has peaks that correspond with reduced histone occupancy. Such an observation (and others presented) is most consistent with continuous histone exchange at these regions. As active turnover would leave the DNA template more accessible to TFs (and would reflect DNase I hypersensitive sites), Henikoff postulated that rebinding of such factors after remodeling/exchange would help maintain the active state. By similar means, factors could rebind upon DNA replication, suggesting that nucleosome dynamics might also contribute to maintenance of gene expression patterns (Henikoff, 2008). It will be interesting to see in future studies how histone turnover models will be reconciled with dynamic regulation of histone modifications. Also highlighting nucleosome dynamics, Woojin An (University of Southern California) discussed mechanisms of H2AZ exchange in human cells. An described the isolation of two distinct H2AZ exchange complexes and demonstrated—through an impressive series of biochemical experiments—that Tip48 and Tip49 are critical for exchange and also that exchange may be regulated by acetylation of nucleosomal histones.

Rohinton Kamakaka (UCSC) described the role of Rsc2, a component of the RSC chromatin remodeling complex in boundary function in budding yeast. A tRNA gene marks a histone-depleted region (HDR) and acts as a boundary to block spread of heterochromatin from the silent mating type locus into flanking sequences. Boundary function depends on binding of TFIIIB and TFIIIC. Using ChIP, Kamakaka showed that Rsc2 is located at the boundary tRNA gene and that in Rsc2 mutants, the HDR is not present and TFIIIB binding is reduced, leading to a model in which Rsc2 displaces a nucleosome and promotes binding of TFIIIB. Rtt109, which acetylates histone H3 at lysine 56 (H3K56ac), was also necessary for formation of the HDR and TFIIIB binding, thus suggesting an interplay between acetylation and remodeling.

Jessica Tyler (University of Colorado School of Medicine) also presented evidence for Rtt109 in remodeling, in this instance at the PHO5 promoter. Using H3 mutated for K56, Tyler showed that H3K56ac, mediated by Rtt109 and Asf1, is required for the rapid promoter chromatin disassembly and transcriptional activation of the PHO5 gene (Williams et al., 2008). Her lab examined chromatin assembly within the Asf1-Rtt109- H3K56Ac framework. To do this, they examined chromatin reassembly at a single, defined double-strand break (DSB) in yeast cells and noted that mutations in either Asf1, Rtt109, or H3K56 had the same effect, highlighting a clear role for these factors in chromatin assembly after DSB repair (Chen et al., 2008). Significantly, the Tyler lab also noted that reassembly of chromatin around the DSB was essential for reversing Rad53-induced cell-cycle arrest.

Histone Modifications

Several talks focused on the role of histone modifications in transcriptional regulation. The Set1-COMPASS complex mediates H3K4 methylation (H3K4me), and this modification is dependent upon prior monoubiquitination of histone H2B. Ali Shilatifard’s lab (Stowers Institute) used a complete library of alanine substitutions within each of the four core histones in Saccharomyces cerevisiae to determine whether other histone residues are also required for H3K4me (Nakanishi et al., 2008). They identified several residues within the H3 tail as well as a patch of residues within the core of histone H2A and H2B that are required for COMPASS-mediated H3K4me (Figure 2). Interestingly, some identified residues appear to function independently of H2Bub, suggesting a sophisticated network of interactions between nucleosomes and the Set1-COMPASS complex. Shilatifard also showed that depletion of WDR82, a component of the human SET1 complex, results in global loss of H3K4me3, whereas H3K4me2 was unaffected. Accordingly, they observed that the human MLL complexes—unlike the SET1 complex—contributed little to bulk H3K4me3 in cells. This observation offers some insight into potential alterations in biochemical function of the “COMPASS-like” complexes in humans, yet much work remains to delineate their different physiological roles.

Figure 2. Histone Residues that Impact H3K4 Methylation in Budding Yeast.

Scheme showing residues within the H3 tail, including H3K14ac, and core residues within H2A and H2B that are required for proper H3K4me by COMPASS. Image provided by Ali Shilatifard.

Both Ramin Shiekhattar (Center for Genomic Regulation, Barcelona) and Peter Verrijzer (Erasmus University) presented evidence of coordinate regulation of histone modifications. Methylation of histone H3 lysine 27 (H3K27me) is a posttranslational modification that is correlated with silencing. H3K27me recruits PRC1, which contains an ubiquitin ligase that monoubiquitinates H2A (H2Aub) (Lee et al., 2007). Shiekhattar showed that human UTX is an H3K27me2/me3 demethylase. At the promoters of HOX genes, UTX modulates the recruitment of PRC1 and H2Aub. In addition, UTX associates with the MLL complexes, and during retinoic acid signaling events, the recruitment of the UTX complex to HOX genes results in H3K27 demethylation and a concomitant methylation of H3K4. The coupling of H3K4me by MLL2/3 with the demethylation of H3K27 through UTX might promote transcription. Peter Verrijzer described isolation of a new Drosophila Polycomb complex, dRAF. This complex, via its KDM2 subunit, couples two repressive marks: H2Aub and H3K36 demethylation. In fact, dRAF—not PRC1—was shown to be primarily responsible for the bulk of H2Aub in cells. Verrijzer further noted that dRAF shares many target genes with PRC1, raising interesting questions regarding possible cooperative biochemical roles for these complexes.

Brian Strahl (University of North Carolina School of Medicine) presented studies on the role of Set2 in regulating the methylation state of H3K36, which is necessary to recruit the deacetylase Rpd3, which in turn represses cryptic internal transcriptional initiation events. Although Set2 is the only H3K36 methyltransferase in budding yeast, it is not known whether the different methylation states at H3K36 are functionally distinct. Strahl showed that the N-terminal 261 residues of Set2, which contain the methyltransferase domain, are sufficient for H3K36me2, histone deacetylation, and repression of cryptic promoters (Youdell et al., 2008). Although Set2-catalyzed H3K36me2 did not require Pol II phosphorylation or the Set2 phospho-CTD interaction domain (SRI), H3K36me3 required the SRI domain and Pol II phosphorylation. This finding suggests a model in which in the association with the phosphorylated Pol II CTD influences Set2 methyltransferase activity to promote H3K36me3 during transcription elongation. What is the functional significance of this switch in methyltransferase activity? Strahl noted that H3K36me2 is generally repressive for gene expression. Perhaps the switch to H3K36me3 relieves this repression, thus ensuring efficient gene expression.

The budding yeast Paf1 complex, which also associates with elongating Pol II, is another complex that regulates H3K36me, as well as H3K4me. Karen Arndt (University of Pittsburgh) showed that discrete nonoverlapping segments of Paf1 subunit Rtf1 are required for interaction with the remodeler Chd1, histone modification, recruitment to active ORFs, and association with other components of the Paf1 complex (Warner et al., 2007). She also described rtf1 mutations that separate its H2B ubiquitylation and H3K4me functions. Also focusing on H3K36me in yeast, Joe Reese (Penn State University) showed that H3 tail deletion mutants (residues 1–28) exhibited defects in elongation. Ultimately, this led to the discovery that H3K36me3 was blocked in the H3 tail mutants and that lysine residues within the H3 tail played a major role in H3K36me. Although the molecular mechanism behind these observations remains to be elucidated, Reese noted that Set2 and Paf1 recruitment was normal in H3 mutant cells. He proposed that modifications to the tail alter Set2 activity during elongation.

Staying on the topic of histone methylation, Johnathan Whetstine (MGH Cancer Center) showed an unanticipated role for the demethylase JMJD2 in chromosomal stability in C. elegans. A mutation that deleted the Jmjc domain caused an increase in H3K9me3 and H3K36me3 levels, DNA double strand breaks, and apoptosis in the germline. This p53-dependent increase in DNA damage occurred in meiotic and mitotic regions, suggesting that JMJD-2 plays a role in genomic stability of the germ line. Also, Ke Zhang for Shiv Grewal’s lab (NIH) discussed how the Clr4 methyltransferase complex (ClrC) nucleates and spreads heterochromatin in fission yeast centromeres (Zhang et al., 2008). Zhang showed that nucleation involved RNAi-dependent loading of Rik1, a WD domain-containing subunit of ClrC, onto the transcribed centromeric repeats during S phase. Heterochromatin spreading depended on the Clr4 chromodomain, which binds specifically to histone H3 methylated on lysine 9 (H3K9me), the product of Clr4. Thus, the ability of Clr4 to both “write” and “read” H3K9me facilitates heterochromatin maintenance through successive cell divisions.

Finally, Paul Brindle (St. Jude Children’s Research Hospital) talked about experiments with mouse embryonic fibroblasts deficient in both p300 and CBP. Remarkably, p300 and CBP are not required for survival; however, such cells cannot proliferate. Interestingly, Brindle observed that CREB target genes were differentially affected in these cells, with some genes even exhibiting hyperactivation. Further, p300/CBP null cells showed only a modest reduction in H3 acetylation, whereas H4 acetylation was dramatically reduced in response to cAMP. This work raises many further questions, such as why different effects are observed at different genes and how target gene expression is activated in the absence of these major coregulatory enzymes.

Transcription Regulation and Cell Signaling

Signaling pathways often converge on the transcriptional machinery, and numerous talks elaborated on this theme. For example, Richard Treisman (London Research Institute) discussed how actin might coordinately regulate cell morphology and transcription through its interactions with the TF MAL (Vartiainen et al., 2007). He outlined how monomeric actin forms a complex with MAL that helps control not only the cellular localization of MAL but also its coactivator function when in the nucleus. Michael Green (UMass) presented evidence that epigenetic silencing of the proapoptotic gene Fas mediated by the Ras oncogene occurs through a specific pathway that involves signaling molecules, TFs, and chromatin regulatory proteins. An elegant RNAi screen identified 28 genes necessary for Fas silencing, and these were also involved in epigenetic silencing of five other unrelated genes (Gazin et al., 2007). These results suggest that Ras employs the same pathway to silence a group of genes.

Shelly Berger (Wistar Institute) and David Auble (University of Virginia) spoke about cellular responses to UV and other stresses. Berger provided evidence that coordinate phosphorylation of p53 and core histones occurs in response to stress, such as DNA damage or nutrient deprivation. Noting other examples of factors that modify both p53 and histone substrates (e.g., p300 acetylation), Berger suggested that such links might help coordinate the signaling cascade and transcriptional response. Auble showed that approximately 200 UV-response genes require both Snf1 and Rad23, yet these factors act upon different pathways. Snf1 indirectly activates UV-response genes via phosphorylation of the Mig3 repressor, whereas Rad23 activates through proteasome recruitment (19S). Together, these talks provided new insight into how cellular responses to stress are orchestrated.

Highlighting a different example of signaling and transcriptional regulation, Tong Zhang from the Kraus lab (Cornell) discussed recent findings regarding the metabolic coenzyme NAD+ and its role in regulating SIRT1, an NAD+-dependent deacetylase. Interestingly, two NAD+ biosynthetic enzymes, NAMPT and NMNAT-1, regulate SIRT1-dependent histone deacetylation and transcriptional outcomes at target gene promoters, supporting a model in which these enzymes act locally to control SIRT1 enzymatic activity. Joe Lipsick (Stanford University) showed that the opposing activities of two TFs, the Myb oncoprotein and the E2F2 transcriptional repressor, mediate epigenetic regulation of gene expression in Drosophila (Wen et al., 2008). His results suggested that the mutation of some oncogenes and tumor suppressor genes, like Myb, E2Fs, and Rb family members, might cause cancer by a global disruption of epigenetic homeostasis rather than by activating or repressing a small number of critical target genes.

Turning to the p53 tumor suppressor, Joaquin Espinosa (University of Colorado, Boulder) presented work supporting a “selective context model” controlling physiological responses to p53 activation. Espinosa described how the cellular response to Nutlin-3, a small-molecule inducer of p53, varies greatly with cell type (Paris et al., 2008). By example, Nutlin-3 induces arrest in HCT116 cells and apoptosis in BV173 cells. Upon further examination, the Espinosa lab uncovered a number of mechanisms that contribute to the distinct cellular responses. Interestingly, the arrest genes p21 and 14-3-3σ were strongly induced in HCT116 cells, but not in apoptotic BV173 cells. Surprisingly, p53 effectively activated p21 transcription in both cell types, but the p21 mRNA was rapidly degraded in BV173 cells. An alternative mechanism worked to downregulate 14-3-3σ in a cell type-specific manner: its promoter was heavily methylated in cells undergoing Nutlin-3-induced apoptosis. These results offer a glimpse into the elaborate network of factors that orchestrate the p53 transcriptional program.

Arnold Berk (UCLA) described a study that dealt with p53 sumoylation and its sequestration in PML bodies. Adenovirus E1B-55K protein, a repressor of p53-dependent transcription, is known to function by binding the p53 activation domain, but not simply by sterically blocking coactivator recruitment. Berk showed evidence that E1B is a SUMO-E3 ligase for p53, and sumoylation triggers p53 transport to and tethering in PML bodies. A model in which p53 tetramers and E1B dimers form a network within PML bodies was presented. Notably, this model postulated that transient sumoylation was sufficient for loading p53 and E1B into the network, tethered by sumoylation of a small fraction of molecules, providing a suitable rationale for how low steady-state p53 sumoylation levels can correlate with nearcomplete inhibition of p53 function by E1B.

In collaboration with Berk, Siavash Kurdistani (UCLA) presented studies showing that interactions between adenovirus small e1a protein, Rb proteins, and p300/CBP regulates histone H3 lysine 18 acetylation (H3K18ac) and induces quiescent human cells to replicate. p300 and CBP were required for H3K18ac in uninfected cells, and the e1a-p300/CBP interaction caused a 3-fold decrease in global H3K18ac, similar to cancers with poor prognosis. This decrease reflected global relocalization of Rb proteins and p300/CBP, thus restricting H3K18ac to genes that promote cell growth and cycling and inhibiting antiviral responses and differentiation. A precise temporal order of e1a binding and the induced relocalization of Rb proteins and p300/CBP were required to direct an epigenetic program contributing to oncogenic transformation.

Expanded Roles for SAGA

Noting substantial differences between Gcn5 null mice versus mice with mutations only in its AT domain, Sharon Dent (M.D. Anderson) and coworkers uncovered a link between SAGA and telomere maintenance. Although both Gcn5 mutations caused embryonic lethality, only cells from Gcn5 null mice had significant numbers of telomeric fusions. This led the Dent lab to investigate the effects of Gcn5 loss on the telomere-associated shelterin complex. They noted that although mRNA levels of shelterin components were unaffected in Gcn5 null cells, the relative levels of protein were dramatically reduced, suggesting a change in protein stability. Dent went on to describe links between the SAGA complex, which contains Gcn5 and the ubiquitin protease Usp22, and shelterin protein turnover.

Steven Hahn (FHCRC) described a newly discovered link between acetylation of yeast SAGA and activation of transcription. They identified 10 acetylation sites within the SAGA subunit Spt7 and confirmed that Gcn5 mediated acetylation at these sites. In a striking result, they observed a growth phenotype only upon mutation of all ten Spt7 acetylation sites. Biochemical analysis of this acetylation mutant indicated that it bound TBP with 20-fold greater affinity, and this correlated with a 2-fold increase in transcriptional activity. One major unanswered question regards four phosphorylation sites also uncovered within Spt7; Hahn noted that phosphorylation at these sites required prior acetylation of Spt7. The regulatory role of Spt7 phosphorylation and the identity of the Spt7 kinase remains to be determined. The SAGA complex was also discussed by Vikki Weake from the Workman lab (Stowers Institute), whose work centered on Usp22-dependent deubiquitination of H2Bub. Nonstop, the fly ortholog of Usp22, is required for appropriate axon guidance within the developing visual system (Weake et al., 2008). Because SAGA is present in all tissues, the Workman lab is now trying to understand why mutation of Nonstop has specific effects in eye development.

New Insights into the Dynamic Nature of the PIC

Previous work by Mike Carey and coworkers (UCLA) identified p300 autoacetylation as a catalytic switch that triggers p300 dissociation from promoter-bound VP16-Mediator and subsequent TFIID recruitment to the PIC (Black et al., 2006). Building upon this observation, Carey noted that the p300 AT domain is heavily acetylated, and this acetylation is dynamic in cells. Significantly, they identified SIRT2 as the sole p300 deacetylase, implicating SIRT2 as a modulator of PIC assembly. This SIRT2-dependent deacetylation was also demonstrated in vivo, as was the dependence on p300 deacetylation for appropriate PIC assembly. Thus, an interplay between p300 autoacetylation and SIRT2-dependent deacetylation provides a means to control PIC assembly and transcription initiation in human cells. Also focusing on cofactor dynamics within the PIC, Dylan Taatjes (University of Colorado) discussed how a 600 kDa “cdk8 submodule” can function to switch transcription off by reversibly associating with Mediator to control its interaction with Pol II.

David Stillman (University of Utah) outlined recent work that clearly defined time-dependent factor recruitment at the yeast HO promoter. Using ChIP assays across the promoter with synchronized cells, Stillman observed that Mediator, SWI/SNF, and SAGA coordinately migrated from a distal regulatory site (URS1) to a proximal site (URS2) as the cell cycle progressed. Significantly, this “migration” was dependent upon FACT, and ChIP analysis indicated FACT recruitment to the HO promoter. Although the molecular mechanism underlying this coordinated factor migration remains to be elucidated, these compelling results suggest an expanded role for FACT in transcription initiation.

Transcriptional Regulation during Differentiation and Reprogramming

Many developmental signals culminate in the deployment of TFs that direct specific alterations in gene expression and trigger differentiation. In addition to programming normal developmental pathways, TFs can also be used to reprogram cells, most notably, reprogramming differentiated cells into induced pluripotent stem (iPS) cells. Numerous talks focused on the role of TFs and chromatin in programming and reprogramming. For example, Kathrin Plath (UCLA) used microarray analysis to study how genes are activated during the reprogramming process. Plath described recent data that compared binding profiles of key reprogramming TFs in ESCs, partially reprogrammed cells, and iPS cells.

In his keynote address, Rick Young discussed the role of chromatin in reprogramming differentiated cells into iPS cells by exogenous expression of the pluripotency TFs, e.g., Oct4, Sox2, Klf4, and c-Myc. The bottleneck in reprogramming appears to be the loss of CpG methylation at endogenous copies of the pluripotency genes, which must be expressed to generate stable iPS cell lines. At the meeting, it was hypothesized that inhibitors of DNA methylation might therefore facilitate reprogramming, and this has indeed been recently demonstrated (Mikkelsen et al., 2008). Young outlined another means to improve reprogramming that involved eliminating c-Myc, which, despite improving the efficiency of iPS cell production, causes tumors in animals derived from these cells. Young and coworkers discovered that Wnt-conditioned medium could compensate for c-Myc, enabling efficient reprogramming in its absence (Marson et al., 2008). Such improvements in reprogramming are clearly necessary before potential clinical applications can be realized.

What are the molecular mechanisms by which tissue-specific TFs direct gene expression? Ken Zaret (Fox Chase) addressed how initial developmental programs are set by early acting TFs, such as the FOX proteins, which are important for endoderm-specific gene expression. FOXA is an example of a pioneer factor, DNA-binding factors that are among the earliest to appear at a gene during development and that are sufficient to access and open local chromatin structure. Zaret found that another FOX family TF, FOXD3, binds the same recognition sequence and functions upstream of FOXA (Xu et al., 2007). Whereas FOXA is expressed in endodermal tissues, FOXD3 is expressed in ESCs, which will give rise to endoderm. In collaboration with Jian Xu and Stephen Smale (UCLA), they found that FOXD3 is necessary for the ESC-specific CpG demethylation in an enhancer that is subsequently bound and activated by FOXA in endodermal cells. These data suggest that the loss of CpG methylation may mark genes in ESCs for subsequent tissue-specific expression. The endodermal enhancer in ESCs is not enriched for H3K27me3 and H3K4me2, the combination of activating and repressive histone marks that are thought to keep other developmentally important genes silent, yet poised for eventual expression. So it appears that ESCs have several ways of regulating developmentally important genes. A better understanding of these mechanisms will undoubtedly provide important insights into differentiation and reprogramming.

A particularly specialized cell type is sperm, which must tightly package its DNA and yet make it accessible to the transcriptional machinery at the onset of zygotic transcription. Brad Cairns (University of Utah) analyzed the distribution of histones, protamines, and DNA methylation in sperm chromatin. The majority of sperm DNA (~95%) is packaged with protamines, but a small amount is packaged with canonical nucleosomes or nucleosomes containing tH2B, a testis-specific H2B variant. Cairns reported that histones are not randomly distributed on sperm DNA; rather, canonical or tH2B-containing nucleosomes were distributed at specific subsets of genes. In fact, Cairns identified clear differences in nucleosome composition at genes involved in differentiation, suggesting a molecular means to preset gene expression patterns.

Jerry Crabtree (Stanford University) showed that developmentally regulated alterations in the subunit composition of a chromatin regulator was important to direct the transition of self-renewing mammalian neural stem cells into neuronal and glial cells of the adult nervous system. A switch in subunit composition of neural, ATP-dependent SWI/SNF-like chromatin remodeling complexes (nBAFs) accompanies and is necessary for this developmental transition (Lessard et al., 2007). Thus, the SWI/SNF-like complexes in vertebrates achieve biological specificity by combinatorial assembly of their subunits, providing another mechanism that can be employed to confer developmentally regulated gene expression. Interesting areas for future work involve identifying the mechanism that triggers the subunit exchange and differential targeting of the two related nBAF and npBAF complexes. Also, Barbara Panning presented evidence that a different chromatin remodeling complex, the Tip60-p400 complex, is necessary for self-renewal of embryonic stem cells, further highlighting the role of remodeling activities in stem cell-fate decisions (Fazzio et al., 2008).

Large Data Sets: Bigger Is Better

The histone code hypothesis suggests that histone modifications act in a combinatorial fashion to specify distinct chromatin states that in turn affect gene expression. Two speakers, Bing Ren (UCSD) and Keji Zhao (NIH), used genome-wide location analysis to determine the distribution of modified histones in large regions of the human genome. Ren presented a high resolution map of chromatin modifications along 30 Mb of the genome and found that active promoters are marked by H3K4me3 and H3K9Ac, whereas enhancers are marked by monomethylation of H3K4 (Heintzman et al., 2007). He also confirmed the H3K4me3 mark was preserved at promoters across different human cell types but also noted that non-CpG island promoters display cell-type specific modification patterns. Zhao reported the patterns derived from the genome-wide analysis of 39 histone modifications in human CD4(+) T cells (Wang et al., 2008). Consistent with great complexity, a large number of different combinations of histone modifications are associated with promoters and enhancers, suggesting enormous diversity in histone marks and associated coregulatory factors. For example, 1,200 unique modification patterns were detected among 4,200 enhancers screened, and similar diversity was observed at promoters. Zhao’s group also identified a module consisting of 17 modifications that was detected at 3,286 promoters. These data suggest that histone modifications may act cooperatively during transcriptional activation. Yet much remains to be uncovered regarding the combinatorial nature of histone modifications genome-wide: because of inherent limitations of modification-specific antibodies (e.g., one recognizing H3K14Ac may not bind H3 also phosphorylated at S10), histone tails possessing multiple modifications may not be represented in current screens.

Oliver Rando (UMass Medical Center) used genome-wide location analysis to monitor the changes in chromatin structure in response to transcriptional activation or genomic replication in budding yeast. By analyzing nucleosome occupancy and histone modifications over time, Rando showed that genes fall into several classes based upon sequential changes in nucleosome occupancy and modification patterns. Although the data generated from this work is still being mined, it is clear that such studies will reveal critical information about how much information within chromatin is truly heritable and what (if any) early instructive states may signal the Pol II machinery.

Nevan Krogan presented large-scale genetic epistasis mapping of chromatin proteins in S. pombe, comparing the connectivity between complexes in S. cerevisiae. Although S. cerevisiae lacks the genes encoding several complexes that together mediate RNAi (the RNAi module), this module is present in S. pombe. The presence of the RNAi module in S. pombe correlated with changes in connectivity of other complexes relative to their organization in S. cerevisiae. These results have significant evolutionary implications: as genomes lose or gain modules, the wiring diagrams connecting different complexes may change. Finally, Tim Harkins (Roche, in collaboration with the Virginia Biotechnology Institute) demonstrated the power of next generation sequencing techniques, which provided quantitative accuracy that was comparable to DNA microarrays and qRT-PCR. In addition, the length of reads allowed discovery of many new splice variants and single nucleotide polymorphisms, shedding new light on the complexity of the human transcriptome.

Concluding Remarks

Transcriptional regulation is in itself an enormous topic, and this group of researchers with diverse interests showed many ways of achieving mechanistic insight into this fundamentally important area. The meeting generated considerable excitement about the progress made and the tantalizing new questions to pursue. We anticipate that all of these questions will be resolved and discussed at the next FASEB conference on this topic, which will convene in 2 years. So get to work everyone, and we’ll see you back in Snowmass in 2010.