The mechanism by which Pol II terminates transcription in metazoans is not understood. Pcf11 [mammalian homolog Pre-mRNA cleavage complex 2 protein Pcf11 (Pre-mRNA cleavage complex II protein Pcf11)] is directly involved in termination in Drosophila. Drosophila Pcf11 (referred to here is dPCF11) is concentrated at the 3' end of the hsp70 gene in cells, and depletion of dPcf11 with RNAi causes Pol II to readthrough the normal region of termination. dPcf11 also localizes to most transcribed loci on polytene chromosomes. Biochemical analysis reveals that dPcf11 dismantles elongation complexes by a CTD-dependent but nucleotide-independent mechanism and that dPcf11 forms a bridge between the CTD and RNA. This bridge appears to be crucial because an anti-CTD antibody, which also dismantles the elongation complex, is found to bridge the CTD to RNA. dPcf11 was observed to inhibit transcription at low, but not high, nucleotide levels, suggesting that dPcf11 dismantles paused elongation complexes. These results provide a biochemical basis for the dependency of termination on pausing and the CTD in metazoans (Zhang, 2006).

Termination of Pol II transcription is an essential step in gene expression, but the mechanism is poorly understood. Besides its requirement for recycling of Pol II, the choice of termination site can influence the availability of splice sites and polyadenylation sites in pre-mRNA. It has been estimated that half of the mRNAs in humans utilize alternate polyadenylation sites, and this can affect the location, stability, and coding potential of the transcripts. Pol II molecules that fail to terminate can inhibit function of downstream promoters by displacing proteins from the DNA. This so-called transcription interference can serve to regulate expression of some genes (Zhang, 2006 and references therein).

Pol II termination is coupled to polyadenylation by the polyadenylation signal in the nascent transcript. Two models have been proposed to explain this coupling. According to the torpedo model, cleavage of the nascent transcript, which precedes polyadenylation, generates an uncapped end on the residual transcript engaged with Pol II. This uncapped end is an entry point for a 5′ to 3′ exonuclease that chases down the Pol II and induces termination. The torpedo model received recent support with the finding that mutation of a 5′ to 3′ exonuclease, called Rat1, causes Pol II to readthrough terminators in yeast (Kim, 2004b). Depletion of the homologous protein Xrn2 from human cells (West, 2004) also impairs termination on a transiently transfected β globin gene (Zhang, 2006 and references therein).

An alternative model, originally called the antiterminator model but now generalized as the allosteric model (Luo, 2004), posits that the polyadenylation signal in the nascent transcript causes an allosteric change in Pol II that decreases the processivity of the elongation complex (EC). This could be due to the dissociation of an antiterminator from the EC or association of a factor that depresses processivity. Until recently, the strongest support for the allosteric model was provided by circumstances in which termination occurs in the absence of the cleavage reaction. Under these circumstances, the torpedo model for termination cannot apply, since there is no entry point for the 5′ to 3′ exonuclease (Zhang, 2006 and references therein).

Recently, a yeast protein called Pcf11 was found to dismantle a yeast Pol II EC (Zhang, 2005). This reaction depends on the CTD of Pol II, thus providing a possible reason for why deletion of the CTD impairs termination in human cells. The CTD corresponds to the unusual C-terminal domain of the largest Pol II subunit and is composed of multiple copies of a heptapeptide with the consensus YSPTSPS. yPcf11 appears to dismantle the EC by bridging the CTD to the nascent transcript (Zhang, 2005). In yeast, mutations in yPcf11 impair both termination and polyadenylation (Sadowski, 2003). yPcf11 is in a complex called CF1A, which is involved in processing the 3′ end of mRNAs (Gross, 2001a). CF1A recognizes part of the tripartite polyadenylation signal in the GAL7 gene (Gross, 2001b), thus providing a possible basis for how the polyadenylation signal might recruit or regulate the activity of yPcf11. Human Pcf11 is in a complex with at least 15 other polypeptides, and the complex is required for 3′ end processing in vitro (de Vries, 2000). The hPcf11 complex interacts with CF1m and CPSF, two proteins that recognize the polyadenylation signal in the nascent transcript (de Vries, 2000; Venkataraman, 2005). Nothing is known about the role of hPcf11 in termination (Zhang, 2006).

This study provides evidence that dPcf11 is directly involved in Pol II termination in Drosophila. Immunofluorescence microscopy and ChIP indicated that dPcf11 is concentrated at the 3′ end of the hsp70 gene, and depletion of dPcf11 from Drosophila cells increases the level of Pol II normally detected downstream from the polyadenylation signal of hsp70. In addition, it was found that the N-terminal region of dPcf11 completely dismantles an elongation complex. This last result sets dPcf11 apart from all other proteins that have been implicated in Pol II termination and is strong evidence that dPcf11 is directly involved in termination. The detection of dPcf11 at most highly transcribed loci in polytene chromosomes suggests that dPcf11 is involved in termination at many genes. dPcf11 provides a basis for connecting three key aspects of termination: the CTD, the polyadenylation signal, and pausing (Zhang, 2006).

A crucial step in the termination reaction mediated by dPcf11 appears to be the formation of a bridge between the CTD and the nascent transcript, as this is the only thing common to the CTD antibody and dPcf11, both of which dismantled the EC. Additional support for the importance of the bridge comes from analysis of yeast Pcf11: mutations impairing RNA binding or CTD binding each inhibit the dismantling reaction (Zhang, 2005). In addition, the dismantling reaction can be inhibited by hybridizing a DNA oligonucleotide to the nascent transcript in the region just outside from where RNA exits Pol II (Zhang, 2005). Presumably, the oligonucleotide blocks formation of the bridge by interfering with the Pcf11-RNA interaction (Zhang, 2006).

Because the CTD antibody and Pcf11 are structurally unrelated, it is unlikely that the dismantling reaction involves Pcf11 directly recognizing part of the body of Pol II. How the formation of the bridge disrupts the elongation complex is a mystery. One possibility is that constraining the CTD or the RNA causes either of these or Pcf11 itself to contact the RNA exit channel in a way that destabilizes the EC. RNA-protein interactions in the RNA exit channel of bacterial RNA polymerase contribute to pausing and termination (Toulokhonov, 2003). The molecular contacts at the RNA exit channel of the Pol II EC may be uniquely suited for allosteric control of the EC, because it was observed that Rho, which normally functions in termination in bacteria, disrupts Pol II ECs, but not Pol I or Pol III ECs (Lang, 1998). Rho moves along RNA in a 5′ to 3′ direction, so it probably collides with the region of Pol II at the RNA exit channel (Zhang, 2006).

dPcf11 seems to interact with a relatively small region of the Drosophila CTD. This is in contrast with the yeast and human CTDs where Pcf11 could in principal coat almost all of the yeast CTD and half of the human CTD. The results from Drosophila suggest that the bridge does not have to form close to the body of the Pol II molecule to dismantle the EC. The binding of dPcf11 to the Drosophila CTD may not be dictated by the heptad per se but by a slightly larger motif that appears four times in the region where dPcf11 binds the CTD. This motif, PSYSPTSP, corresponds to the region of a peptide composed of two consensus heptads that was contacted by yPcf11 (Meinhart, 2004) in a crystallized complex (Zhang, 2006).

Phosphorylation of the CTD could influence the activity of Pcf11. Phosphorylation of serine 2 in the CTD appears to increase the affinity of yPcf11 for the CTD (Barilla, 2001; Licatalosi, 2002; Sadowski, 2003). Importantly though, yPcf11 binds the unphosphorylated CTD (Barilla, 2001; Licatalosi, 2002; Sadowski, 2003), and there is evidence in yeast indicating that the CTD of Pol II is dephosphorylated just prior to termination (Buratowski, 2005). ChIP data indicate that the level of serine 2 phosphorylation increases as Pol II moves from the 5′ to the 3′ end of the hsp70 gene (Boehm, 2003), and the same occurs on several yeast genes (Ahn, 2004). This rising level of serine 2 phosphorylation could contribute to the recruitment of Pcf11 near the 3′ end of the gene. However, the phosphates on the CTD might also antagonize the ability of Pcf11 to form a bridge with the nascent transcript due to electrostatic repulsion. The CTD phosphatase Ssu72 has been implicated in termination (Krishnamurthy, 2004; Steinmetz, 2003). Ssu72 might participate in termination by removing phosphates from the CTD so the bridge can form between the CTD and RNA (Zhang, 2006).

dPcf11 is concentrated near the polyadenylation signal of hsp70, similar to what was observed for several genes in yeast (Kim, 2004a). Though Pcf11 binds RNA, it seems unlikely that Pcf11 alone recognizes the polyadenylation signal in the nascent transcript. First, equivalent crosslinking is observed to two unrelated RNAs, neither of which contained a polyadenylation signal. Second, amino acids 1–149 of Pcf11 lack any known RNA recognition motifs. Nevertheless, Pcf11 appears to have a surface that interacts specifically with RNA, because mutating one amino acid in yeast Pcf11 impaired RNA binding without affecting CTD binding (Zhang, 2005; Zhang, 2006).

Yeast Pcf11 is part of a complex called CF1A, which contains three other subunits (Gross, 2001a). One subunit, Rna15, recognizes part of the polyadenylation signal, thus providing a way to recruit yPcf11 to the end of the gene after the polyadenylation signal has been transcribed (Gross, 2001b). Human Pcf11 is part of a complex called CFIIA, which itself does not appear to recognize the polyadenylation signal (de Vries, 2000). CFIIA, however, interacts with CPSF and CFI_m, two proteins that recognize different parts of the polyadenylation signal in humans and that are involved in pre-mRNA 3′ end processing (Venkataraman, 2005; Zhao, 1999). If CPSF and CFI_m are involved in recruiting Pcf11 to the 3′ end of genes in metazoans, regulation is needed to prevent Pcf11 from prematurely terminating transcription. ChIP detects both CPSF and CFI_m well upstream of the polyadenylation site in the human G6PD gene (Venkataraman, 2005), and earlier studies indicated that CPSF is recruited to the 5′ end of genes (Dantonel, 1997) through association with TFIID (Zhang, 2006).

The location of pause sites will be a key parameter in dictating where Pcf11 dismantles the elongation complex. As long as the EC is moving, it resists the action of Pcf11. This resistance is thought to arise because the RNA reeling out of an actively moving EC interferes with physical interactions that might be required for the dismantling reaction. There are ample data to indicate that pause sites are involved in selection of termination sites (see Plant, 2005 and references therein). Diverse mechanisms could be used by the cell to cause the EC to pause. These include the presence of pause sites that are intrinsic to the DNA sequence. Intrinsic pauses are found scattered throughout almost any stretch of DNA, so this could account for the stochastic selection of termination sites downstream from a polyadenylation signal. Specific proteins bound to the DNA could cause pausing. Finally, nucleosomes cause ECs to pause. This could explain why chromatin remodeling factors appear to act as terminators (Zhang, 2006 and references therein).

In conclusion, the dependence of the Pcf11 dismantling reaction on pausing and the CTD provide possible explanations for why these two things are important for termination. The specificity of termination probably arises from the combinatorial actions of factors that control pausing, the association of Pcf11 with the CTD, and the association of Pcf11 with the nascent transcript. The results provide direct support for an allosteric model of termination but certainly do not preclude possible contributions from an RNA exonuclease after cleavage of the nascent transcript. One possibility is that the exonuclease shortens the residual nascent transcript, forcing Pcf11 to bind close to the RNA exit channel (Zhang, 2005).

Drosophila Pcf11 has 37% amino acid sequence identity with the CTD interacting domain of yPcf11 (Meinhart, 2004; Zhang, 2005)