|Home | About | Journals | Submit | Contact Us | Français|
In the eukaryotic genome, the thousands of genes that encode messenger RNA are transcribed by a molecular machine called RNA polymerase II. Analysing the distribution and status of RNA polymerase II across a genome has provided crucial insights into the long-standing mysteries of transcription and its regulation. These studies identify points in the transcription cycle where RNA polymerase II accumulates after encountering a rate-limiting step. When coupled with genome-wide mapping of transcription factors, these approaches identify key regulatory steps and factors and, importantly, provide an understanding of the mechanistic generalities, as well as the rich diversities, of gene regulation.
The genetic information encoded in the DNA of eukaryotic genes is transcribed into RNA by large molecular machines called RNA polymerases. One of these machines, RNA polymerase II (Pol II), transcribes all the protein-coding genes. The control of Pol II activity is highly modulated at individual genes, and this specific regulation is critical for both the homeostasis of cells and the programmed development of multicellular organisms. The execution of this regulation is dictated by combinatorial molecular interactions of transcription factors with each other and with specific DNA sequences at each gene. Modern biochemical and molecular methods coupled with genetics and genomics have identified thousands of factors that participate in regulated transcription1. Most of these factors are proteins, but a growing number of them are RNAs. They enable Pol II to gain access to the gene's promoter, to initiate RNA synthesis at the transcription start site (TSS) of the gene and to generate a productively elongating transcription complex that produces a full-length RNA transcript.
The thousands of transcription factors involved in the transcription process may be true regulatory factors or simply critical cogs in the cycle of transcription. True regulatory factors are likely to represent only a fraction of the total number of factors that are important for gene expression. As an analogy, consider a motor vehicle: a car has numerous crucial components and processes that are required to achieve acceleration and proper speed (cylinders, spark plugs, tyres and so on), but components regulated by the driver are limited to the ignition, the steering wheel, the accelerator and brake pedals, and the gear stick. Therefore, it is important to identify the true regulatory factors and the associated biochemical processes that execute gene regulation. The status and local density of the ultimate target of regulation, the transcription machine Pol II, have proved extremely useful in assessing the steps in the transcription cycle that are rate limiting and are altered in vivo by particular transcription factors (the driver in the above analogy).
In this Review, we discuss how in-depth mechanistic analysis of individual genes coupled with large-scale analysis of transcription-factor binding over an entire genome can distinguish the key steps at which transcription is regulated, and how these steps can be accelerated in an activator-dependent manner.
The DNA sequences in and around specific gene promoters provide the code that dictates when, where and at what level specific genes are transcribed. This code comes in three parts: the core promoter, the region proximal to the core promoter, and the more distant enhancer sequences (Fig. 1). In various combinations, the elements of the core promoter sequence target the assembly of distinct preinitiation complexes (PICs) composed of the general transcription factors (GTFs)2. Promoter-proximal regions and more distant enhancer sequences direct the binding of specific transcription factors, called activators or repressors (see page 199 for a more detailed discussion of enhancers). Although activators or repressors can interact directly with components associated with the core promoter, they execute their regulation predominantly through co-regulators, which are often multiprotein complexes. Some of the co-regulators can interact directly with Pol II and GTFs and influence expression. Others can reorganize nucleosomes or covalently modify chromatin, and change the chromatin architecture of the gene. This can in turn influence transcription-factor associations and the transcriptional status of Pol II.
Although present evidence suggests that many steps in the transcription process may be rate limiting, the question remains whether these rate-limiting steps are actual points of regulation. To meet this criterion, these steps should be regulated by factors in response to particular physiological, environmental or developmental signals. Although transcription regulatory factors that act as repressors can also modulate specific steps, we focus here on activators, as they seem to predominate as critical modulators of gene expression in eukaryotes.
In-depth analyses of individual genes, or sets of co-regulated genes, have revealed critical mechanistic insights into transcription factors that take part in regulation in response to specific cellular signals. This information, when coupled with more recent large-scale analyses of the associations of such factors over the entire genome (which have been carried out by individual laboratories, as well as by the Encyclopedia of DNA Elements (ENCODE) and modENCODE consortia), allows the generality of particular regulatory mechanisms to be assessed. These genome-wide efforts efficiently appraise the collections of genes that associate with particular transcription factors and thereby define many potential participants in any regulatory mechanism. They also reveal the regulatory circuitry of gene expression networks and how these networks respond to cellular signalling3. Interpreting how the transcription factors and gene circuitry respond to signals and lead to transcriptional regulation requires that we identify not only the factors that respond to signals but also the rate-limiting steps in transcription.
The transcription cycle consists of at least eight distinct major steps at which transcription could be rate limiting and activators could potentially act to increase the rate of transcription (Fig. 2). The transcription cycle begins with Pol II gaining access to the promoter, which in some cases requires the promoter being cleared of nucleosomes that obscure access to Pol II and the GTFs (step 1). A PIC assembles on the core promoter (step 2). The DNA is then unwound, and Pol II initiates transcription (step 3). Early-elongating Pol II gets a stable grip on both the DNA and the growing RNA chain, escapes/clears the core promoter and proceeds to the promoter-proximal pause region (step 4). The paused Pol II complex is then hyperphosphorylated and escapes from the pause region in an unknown manner, either terminating or entering productive elongation (step 5). If it has not terminated, Pol II must then productively elongate through the entire body of the gene (step 6). After this, Pol II undergoes termination (step 7), and it can reinitiate to start a new round of transcription (step 8).
Any of these major steps could, in principle, be rate limiting, and the distribution of Pol II across a gene can suggest which steps are rate limiting for that gene. The Pol II density across many genes has been determined in a plethora of individual gene studies4; moreover, a wealth of data has been obtained in recent genome-wide chromatin immunoprecipitation (ChIP) studies examining Pol II distribution across the genomes of several organisms: Saccharomyces cerevisiae5, Drosophila melanogaster6,7 and Homo sapiens8. In each organism, these studies have identified different classes of gene on the basis of their Pol II distribution: no Pol II, Pol II evenly distributed and Pol II enrichment at the 5′ end. Genes without Pol II are in an ‘off’ state, and are limited by the step at which the promoter is cleared of nucleosomes (step 1) or the step at which a PIC assembles (step 2). An even distribution of Pol II suggests that Pol II recruitment (step 2) is the rate-limiting step: none of the downstream steps leads to an accumulation of Pol II in other regions of the gene9. An enrichment in Pol II at the 5′ end suggests that steps downstream of Pol II recruitment (steps 3–5) are rate limiting. Because ChIP localization with a single Pol-II-specific antibody cannot distinguish between steps 3, 4 and 5, more experiments pinpointing the exact rate-limiting step need to be performed. The transition between PIC formation (step 2) and promoter escape (step 4) is marked by the unwinding of DNA, formation of a transcription bubble with a stable RNA–DNA duplex and lengthening of the nascent transcripts associated with Pol II. Transcription-bubble formation and RNA length can be distinguished by permanganate mapping of unpaired thymidines in the transcription bubble10 and run-on assays11,12, respectively. In addition, the transition between initiation and pausing (step 4) is marked by phosphorylation of the Pol II carboxy-terminal domain (CTD) repeats on Ser 5 by the kinase subunit of the GTF TFIIH (CDK7 in Drosophila), and productive elongation (step 6) is generally marked by phosphorylation of Pol II CTD repeats on Ser 2 by the kinase complex positive transcription elongation factor b (P-TEFb; CDK9–cyclin T in Drosophila). Therefore, using specific antibodies to examine these phosphorylation marks on genes with 5′-end Pol II peaks can help distinguish the rate-limiting step for those genes13.
Many genes regulated by the recruitment of Pol II have promoters covered with nucleosomes. Activators at these genes recruit nucleosome remodellers and nucleosome-modifying enzymes to allow GTFs and Pol II access to the promoter (Fig. 2, step 1) (see page 193 for details on nucleosome remodellers). PHO5 in S. cerevisiae is one of the best studied of the genes that are regulated in this manner (Box 1). In other examples, it has been shown that both human and yeast activators interact with the SWI/SNF remodelling complexes (Swi/Snf complex in yeast) and positively stimulate transcription from nucleosome-containing templates14. In addition, recruitment of histone-modifying enzymes (for example recruitment of the histone acetyltransferase Gcn5 to galactose-inducible genes by the yeast activator Gal4 (ref. 15)) provides another means by which activators influence and modulate the outcome of transcription by modifying promoter chromatin state.
Transcription from the Saccharomyces cerevisiae acid-phosphatase gene PHO5 (see figure, panel a) is regulated at the level of activator recruitment and eviction of four positioned nucleosomes (brown, −1 to −4) from the upstream regulatory and promoter region51,52. Pho2, a homeodomain-containing activator, and the histone acetyltransferase complex NuA4, which acetylates histones H4 and H2A (purple Ac) before induction, are both present at the promoter. Phosphate (Pi) starvation (see figure, panel b) induces PHO5 by activating the cyclin-dependent-kinase inhibitor Pho81 (not shown), which inhibits the Pho80–Pho85 kinase complex (also not shown) and allows accumulation of the active unphosphorylated form of the basic helix–loop–helix activator Pho4 in the nucleus53. Pho4 binds mainly to the low-affinity UASp1 within the hypersensitive site that is flanked by two positioned nucleosomes on each side, and cooperatively interacts with Pho2. This Pho4–Pho2 complex triggers disruption of the positioned nucleosomes, and this event is concurrent with Pho4 binding to the high-affinity UASp2 and induction of transcriptional activation in a manner that depends on the acidic transactivation domain of Pho4 and on NuA4 (refs 54–56).
After Pho4 binding, the positioned nucleosomes become hyperacetylated (light mauve Ac) through the histone-acetyltransferase activity of the SAGA subunit Gcn5 and then undergo remodelling (see figure, panel b) before being evicted (see figure, panel c, green arrow) from the promoter. Both Swi/Snf and Ino80 complexes have been implicated in chromatin remodelling at PHO5 (refs 56, 57). The H3–H4 histone chaperone Asf1 has also been shown to play a part in the eviction process58,59. Although Gcn5, Asf1 and chromatin remodellers are not essential for PHO5 induction, their deletion results in a kinetic delay in the loss of nucleosomes and gene activation. These observations indicate that multiple mechanisms are in place for remodelling and eviction of the positioned nucleosomes at PHO5. The co-regulated PHO8 gene is dependent on Gcn5 and Swi/Snf60, indicating that these nucleosome modifications and remodelling events can have a range of effects on the Pho4-mediated activation of this co-regulated gene family.
Other phosphate-responsive genes are also induced during Pi starvation. But the degree of sensitivity to environmental Pi and the extent of expression on induction vary greatly among these genes. A recent study showed that variabilities in the activation threshold and transcription range of phosphate-system genes are governed, respectively, by the accessibility of high-affinity Pho4-binding sites before induction and the affinity and number of these Pho4-binding sites61, highlighting the role of activator binding-site accessibility and nucleosome positioning on the dynamic range of transcriptional output.
In other genes, the promoter is free from nucleosomes, but Pol II recruitment is still rate limiting (step 2). During activated transcription, recruited Pol II quickly progresses into productive elongation and becomes relatively uniformly distributed across the gene16. At these genes, PIC assembly must be upregulated by activators. Extensive in vitro studies have shown activators can interact with many GTFs: TATA-binding protein (TBP), TFIID, TFIIA and TFIIB17. Activators also recruit the coactivator Mediator, which can interact with GTFs and increase expression18,19. These interactions might increase the binding of GTFs to the promoter or stabilize the PIC, allowing more efficient recruitment of Pol II. Additionally, activator-dependent recruitment of chromatin-modifying enzymes results in distinctive chromatin marks on promoters. Domains associated with GTFs can bind to these marks20,21, and these interactions can further aid in stabilizing PIC formation.
In vivo Pol II distributions have also indicated that post-recruitment steps can be rate limiting. Enrichment in Pol II at the 5′ ends of genes suggests that steps between recruitment and productive elongation (steps 3–5) are rate limiting in these genes. Some of them may be regulated at initiation (step 3) or promoter escape (step 4). In the case of regulation at initiation, the Pol II associated with the 5′ end of the gene is contained within a PIC, and activators may regulate open-complex formation by recruiting or stimulating factors important for this step. For example, Mediator can interact with two GTFs crucial for unwinding DNA and forming open complexes: TFIIE and TFIIH18,22. Therefore, activators recruiting Mediator may increase the rate of open-complex formation. In the case of regulation at promoter escape/clearance, the Pol II associated with the 5′ end of the gene has initiated transcription but cannot transcribe to the promoter-proximal pause region owing to the instability of the RNA–DNA duplex and the inability of Pol II to break contacts with factors establishing the PIC. This can lead to abortive initiation4. Activators may mitigate these problems, but results on the extent of regulation at step 4 or how this happens in vivo are limited so far. TFIIH is again important for this step, not only for further unwinding of downstream DNA but also for the TFIIH-dependent Ser 5 CTD phosphorylation that occurs around this step, which may aid in breaking Pol II contacts with some promoter-bound factors23. Indeed, an activator can promote this phosphorylation in vitro24, and Mediator enhances the TFIIH-dependent phosphorylation of the CTD19.
Assays other than ChIP have shown that the Pol II that is enriched on the 5′ ends of many genes is already engaged in transcription but is held paused12. Directed studies of specific genes in the 1980s showed that Pol II was at high density on the 5′ ends of some genes, and this Pol II was extensively characterized in focused studies of Drosophila Hsp70 and other heat-shock genes (Box 2; reviewed in ref. 25). Upon activation, the paused Pol II on Hsp70 is released into productive elongation, and Pol II becomes evenly distributed across the gene. This indicates that the activator is regulating the transition from the paused state to productive elongation (step 5). P-TEFb is a major switch that has a critical role in facilitating the transition of Pol II from promoter-proximal pause sites into productive elongation26 at most (if not all) genes; inhibition of P-TEFb dramatically decreases global transcription27. P-TEFb interacts directly with some activators28–30, but others rely on different mechanisms to recruit P-TEFb indirectly (reviewed in ref. 31). Although P-TEFb is important for pause escape, Pol II still elongates many dozens of base pairs from the canonical Hsp70 pause sites when P-TEFb is inhibited during heat shock26. Therefore, there may be other P-TEFb-independent mechanisms for releasing paused Pol II. In addition, elongation requires nucleosome loss or remodelling to occur, and it has been proposed that nucleosomes block the escape from pausing32.
Drosophila Hsp70 was one of the first genes discovered to have promoter-proximal paused Pol II, and has been extensively studied. As a result, Hsp70 has served as the model for genes regulated at the step of early elongation. Its promoter resides in a nucleosome-free region extending to about 250 bases downstream of the TSS36,62 (see figure, panel a). This open promoter is bound by GAGA factor (GAF, orange circles) and GTFs (blue rectangle)63. Studies have suggested that the GAGA elements are crucial for setting up the paused Pol II (red rocket)64–66, which is partially phosphorylated (red P). And in vitro evidence suggests that GAF bound to the promoter can recruit nucleosome remodellers to maintain this nucleosome-free state67. This open promoter allows Pol II to initiate and transcribe 20–40 bases downstream of the TSS, where it is held paused. This pausing is, at least partially, mediated by the SPT4–SPT5 complex (pink pentagon) and the NELF complex (purple circle). In vivo, NELF is present on uninduced Hsp70, and it is still present, but at lower levels, after heat shock68 (see figure, panel c). Furthermore, NELF depletion in vivo reduces the amount of engaged Pol II on uninduced Hsp70 (ref. 69). Additionally, the downstream sequence may also be important for pausing. When the sequence within 30 bases downstream of the Hsp70 TSS is switched with the sequence from another gene, the amount of pausing markedly decreases64. This may indicate that either the factors binding to downstream elements or the intrinsic pause-inducing characteristics of the transcribed sequence, or both, have a role in pausing. The paused Pol II is phosphorylated by the TFIIH subunit CDK7 on Ser 5 of the CTD repeats. This phosphorylation may be involved in pausing. A temperature-sensitive mutant of CDK7 decreases the amount of paused polymerase on Hsp70 at non-permissive temperatures70; whether this affects pausing directly or at an earlier step remains to be resolved.
Heat shock (see figure, panel b) causes the transcriptional activator HSF (yellow diamonds) to trimerize and stably bind upstream of Hsp70 (ref. 71). Such a temperature shift also activates HSF, resulting in the recruitment of coactivators (green hexagon), a rapid general loss of nucleosome protection across the gene36 and release of the paused Pol II into productive elongation. Upon heat shock, P-TEFb (blue triangle) is recruited to the gene13,72 and phosphorylates (blue P) the CTD, SPT5 and NELF subunits; the NELF complex dissociates from the Pol II complex; and Pol II releases from the pause sites, allowing rapid recruitment of new Pol II to the gene (see figure, panel c). Although Pol II still resides in the canonical pause sites under these conditions, it is estimated that the pause is of much shorter duration, with Pol II escaping every 4 s rather than once every 10 min before heat-shock induction63.
Several studies have demonstrated that P-TEFb is important for releasing the paused polymerase upon induction of Hsp70. In vitro assays show that P-TEFb relieves the inhibitory effects of SPT4-SPT5 and NELF73. Depletion or inhibition of P-TEFb severely reduces Hsp70 RNA expression70,74, and P-TEFb inhibition, either before or after heat shock, blocks Pol II escape from the 5′ end of the gene26. Additionally, TFIIS is important for Pol II escape from the pause sites through its maintenance of paused Pol II in an elongation-competent state75. Depletion of TFIIS impedes the release of Pol II from the pause and reduces the rate of Hsp70 mRNA production.
At present, the case for regulation at later stages (steps 6–8) in the transcription cycle is hard to make, but hints of such regulation exist33. It seems probable, for some genes, that cells have evolved means of at least modest regulation at these stages in response to cellular signals. Activator-dependent loss of nucleosomes aids in elongation (step 6). Additionally, the activator-dependent GTF-stabilizing interactions discussed earlier are important for recycling and reinitiation of Pol II (step 8). Some GTFs can remain associated with the promoter after the Pol II has escaped, and they form a scaffold that allows Pol II to initiate efficiently in successive rounds of transcription34.
As suggested from this discussion, activators can act during distinct steps in transcription in vivo. Certain activators, such as Sp1 in mammals, target early steps in the cycle, whereas others, such as those with an acidic activation domain, can target early elongation/escape from pausing. Studies suggest that the distinct sets of targets may be independent of one another. The very strong viral acidic activator VP16 seems to act at both early and pausing escape steps35. The effect of Drosophila activator HSF on nucleosome removal could be separated from its effects in stimulating transcription on the Hsp70 gene36. The ability of activators to stimulate multiple slow steps can lead to a much more rapid and robust activation through a kinetic synergism (reviewed in ref. 37).
The different steps in transcription provide multiple targets for the evolution of regulatory mechanisms. A block at early stages of promoter accessibility provides a means of placing a gene under tight control. An activator that stimulates nucleosome removal to unmask the promoter would allow that first step to occur; however, the gene could then require additional activators to stimulate later steps that eventually produce a messenger RNA. Thus, the activation of a gene could be regulated by a combination of signals that each acts on particular activators and their targeted steps, resulting in tight control; an example of such a gene is PHO5 (Box 1).
The promoter-proximal paused Pol II seems to provide a means of achieving a rapid, and perhaps synchronous, activation of gene expression38. The paused Pol II has already progressed through multiple processes that can be slow and stochastic, and a transcriptional activator, acting on a preloaded paused Pol II, allows a rapid transition into productive elongation. Genes with paused Pol II seem not to be in a completely transcriptionally ‘off’ state12. Therefore, regulation of pausing may sacrifice tight control of RNA production in favour of the uniform and rapid response of a gene. The heat-shock genes are a classic example of this regulation: their rapid induction seems critical in responding to a stress that is normally lethal (Box 2). Other stress-response genes, such as those responsible for DNA-damage, unfolded-protein and immune-response pathways, are also enriched in paused Pol II6,12. In the early embryo, narrow bands of cells must respond rapidly and uniformly to developmental signals, and genes that respond to these signals are also highly enriched in paused Pol II at the developmental stage at which they must be turned on7.
Although many powerful methodologies have been developed for investigating mechanisms of gene regulation in vivo, there follows a wish list of key tools and approaches for the future. This list is not meant to be comprehensive, and the approaches described benefit both from the interplay with in vitro studies, which provide critical tests of mechanisms and quantification of binding and rate constants for factor interactions, and from structural studies, which provide insight into the precise molecular architectures of proteins and larger macromolecular complexes.
First in the list are protein–DNA crosslinking technologies (for molecular imaging), which produce snapshots of transcription-factor interactions on specific genes in vivo and thereby set critical limits in evaluating models of transcriptional regulation. This approach is especially powerful when applied at high temporal and spatial resolution to track the recruitment kinetics and location of specific factors on specific genes during a time course of their activation13,39. These crosslinking methods can be used at individual genes, as well as at the whole genome level. As the resolution of these assays improves, so too does the power of this data in evaluating mechanistic models of transcriptional regulation. Ultimately, the utility of these assays would benefit from the development of this technology to allow mapping of contacts at single-nucleotide resolution and at sufficient kinetic resolution to resolve known major steps in the transcription cycle (Fig. 2), as well as steps yet to be discovered.
Second are highly sensitive microscopy methods, which should provide a strong complement to biochemical methods for examining protein–DNA interactions by allowing observation of the recruitment and dynamics of proteins in real time. The tracking of factors during a time course of the rapid and synchronous activation of a regulated gene will be greatly enhanced when microscopic imaging technology is sufficiently sensitive to examine the recruitment and dynamics of individual proteins on a single chromatid in vivo. Tracking proteins at specific loci is now possible on polytene chromosomes or in diploid cells, where genes are amplified in tandem, but single-chromatid tracking of Pol II and particular transcription factors would offer a comprehensive and ordered view of the process and provide the detail that is often masked in measurements that rely on averaging events at many gene copies in a single cell or biochemical measurements of genes in a population of cells.
Third are methodologies that evaluate the catalytic and modification state of the key proteins, which, along with the tracking of protein– DNA interactions in vivo, are also critical. The antibodies that detect the phosphorylation status of Pol II have been crucial in assessing the activity state of Pol II at various positions along a gene and during the time course of gene activation. Additional antibodies, or other detection reagents, that can evaluate the modification status of transcription factors could certainly provide valuable insights into the way in which different modifications influence each other, and how the final modification code influences the mechanisms of activation. Ultimately, the development of highly effective chromatin purification schemes and highly sensitive mass spectrometry should allow the examination of the complete range of proteins and protein modifications in a particular region and under any condition. There has already been some success in such an examination of a repetitive region of the genome40, and taking this to the level of specific genes would be extremely powerful.
The evaluation of Pol II activity state is enhanced by nuclear run-on assays that measure transcriptionally engaged RNA polymerase complexes. RNA polymerases that are in an elongation state or simply associated with DNA can be detected by ChIP assays, whereas only the former are detected by nuclear run-on assays. An approach called GRO-seq, which uses massively parallel sequencing to measure nascent run-on transcripts, has greatly enhanced the sensitivity of nuclear run-on assays and provides a genome-wide analysis of all transcriptionally engaged polymerases12. Fourth in our wish list is further development of the GRO-seq assay and the continued examination of short RNAs41,42 and RNAs associated with chromatin43, which may allow various states of elongating Pol II to be distinguished (for example the promoter-proximal paused, arrested, abortively initiating and productively elongating states) and further enhance our understanding of the transcription cycle. The utility of these GRO-seq and derivative assays will be enhanced by the development of strategies that allow single-nucleotide resolution and thereby enable the location of Pol II to be precisely defined relative to sequence elements and particular transcription factors.
Although the mapping of protein–DNA interactions in vivo at specific genes is well developed, the determination of protein–protein interactions (which are equally important) is much less so and is the fifth item in our wish list. High-resolution microscopy methods that provide subwavelength resolution, for example fluorescence resonance energy transfer44 and stochastic optical reconstruction microscopy/photo-activated localization microscopy45,46, have the resolution to assess whether proteins are separated by tens of nanometres or less and thus evaluate whether these proteins are close enough to be in contact. Other optical techniques such as fluorescence cross-correlation spectroscopy47 make it possible to assess whether pairs of proteins are part of the same complex. The use of biological amplification provided by polytene chromosomes47 or tandem polymers of genes48 has provided a glimpse of the potential of optical methods in viewing transcription at specific genes. With improvements in fluorescent labels and detection methods, these approaches should in principle be applicable to factors associated with specific genes, allowing factor–factor associations to be tracked in real time. In addition, recent studies in S. cerevisiae have used photoreactive amino acids to provide detailed protein–protein contacts during initiation in vivo49. Extending this analysis to other steps and other organisms will allow in vivo protein–protein interactions to be examined during the transcription cycle in unprecedented detail.
Last, pronounced augmentation of the optical and molecular imaging described here can be achieved by depleting or inhibiting specific factor interactions. Re-examining (re-imaging) the consequences of such experimental treatments can provide critical tests of proposed mechanisms. Although depleting factors with RNA interference is convenient and can generally be used to disrupt factors, sorting primary effects from secondary effects is difficult. Drugs that target specific transcription-factor kinases have been particularly useful, especially when effects are examined immediately after cells have been treated23,26. Ultimately, cell-permeable drugs26, or RNA-aptamer-based drugs50 synthesized in cells, that target protein–protein interactions during transcription will be extremely useful for assessing the primary effects of such perturbations.
The rapid advancement of techniques in biochemistry and microscopy is providing powerful methods to examine the molecular details of biological processes in living cells. These techniques, when coupled with sophisticated approaches to genetically alter and chemically inhibit transcription factors, will provide a new understanding of the transcription cycle, including more detailed knowledge of the known steps in transcription and perhaps identification of new steps. Analysis of individual genes will continue to reveal important mechanistic information about transcription-factor function. Such studies of single genes will be complemented with genome-wide assays to investigate the generality of discoveries and identify specific mechanisms used by individual and co-regulated genes. The next decade will undoubtedly yield exciting insights into the mechanisms of transcription and its regulation.
We thank members of the Lis laboratory for discussions and critical reading of the manuscript. Work in our laboratory is supported by US National Institutes of Health grant GM25232.
The authors declare no competing financial interests.