Localization of RNAPII and 3′ end formation factor CstF subunits on C. elegans genes and operons

ABSTRACT

Transcription termination is mechanistically coupled to pre-mRNA 3′ end formation to prevent transcription much beyond the gene 3′ end. C. elegans, however, engages in polycistronic transcription of operons in which 3′ end formation between genes is not accompanied by termination. We have performed RNA polymerase II (RNAPII) and CstF ChIP-seq experiments to investigate at a genome-wide level how RNAPII can transcribe through multiple poly-A signals without causing termination. Our data shows that transcription proceeds in some ways as if operons were composed of multiple adjacent single genes. Total RNAPII shows a small peak at the promoter of the gene cluster and a much larger peak at 3′ ends. These 3′ peaks coincide with maximal phosphorylation of Ser2 within the C-terminal domain (CTD) of RNAPII and maximal localization of the 3′ end formation factor CstF. This pattern occurs at all 3′ ends including those at internal sites in operons where termination does not occur. Thus the normal mechanism of 3′ end formation does not always result in transcription termination. Furthermore, reduction of CstF50 by RNAi did not substantially alter the pattern of CstF64, total RNAPII, or Ser2 phosphorylation at either internal or terminal 3′ ends. However, CstF50 RNAi did result in a subtle reduction of CstF64 binding upstream of the site of 3′ cleavage, suggesting that the CstF50/CTD interaction may facilitate bringing the 3′ end machinery to the transcription complex.

Introduction

The C. elegans genome contains multi-gene transcription units, operons that contain 15% of the genes. Operon transcription is initiated at a single promoter located at the 5′ end of the cluster of as many as eight genes. During transcription, a single polycistronic pre-mRNA is co-transcriptionally processed into mature monocistronic mRNAs by splicing and 3′ end formation. The 5′ ends of downstream gene mRNAs are formed by SL2-specific trans-splicing. In contrast to single genes, in operons 3′ end formation must occur without accompanying transcription termination. The mechanism by which termination is prevented at internal sites in operons is unknown, although it seems likely that the accompanying SL2 trans-splicing, which usually occurs only ∼100 bp downstream of the 3′ end cleavage/polyadenylation site of the gene just upstream, provides a cap to the downstream mRNA, thereby preventing access to the XRN2 “torpedo” that would otherwise terminate transcription.^1,2

The largest subunit of RNAPII contains a carboxy-terminal domain (CTD) that is conserved among all eukaryotes.³ The CTD is composed of a seven amino acid repeat with the consensus sequence Y₁S₂P₃T₄S₅P₆S₇.⁴ The number of these repeats varies among species with mammals having 52 and C. elegans having 35.⁵ The CTD mediates coupling of transcription with pre-mRNA processing^6-10 by acting as a “landing pad” for recruitment of RNA processing factors to the transcription site,¹¹ thus facilitating co-transcriptional pre-mRNA processing (For a review see refs. ^12-14).

During the transcription cycle (initiation, elongation, termination) the CTD is dynamically phosphorylated and dephosphorylated,^14-16 giving the CTD specificity for recruitment of different factors. At 3′ ends of genes, Ser-5p levels decrease, presumably by phosphatase action, leaving the CTD phosphorylated at Ser2 to terminate transcription.¹⁷ The CTD phosphorylations enhance maturation of the pre-mRNA by allosteric activation of processing factors¹⁸ and by increasing the concentration of these factors at their site of action. Ser2p is involved in the recruitment of 3′ end formation factors. ChIP experiments have shown that RNAPII that is heavily phosphorylated at Ser2 generally occurs as a peak downstream of the poly-A site.^19-22 Mutant RNAPII in which the serines at position 2 within the CTD have been mutated to alanines showed a defect in 3′ end cleavage.²³ Furthermore, experiments done using the human B-globin gene have shown that inactivation of the poly A signal sequence (AAUAAA) inhibits Ser2 hyperphosphorylation 1–2 kb downstream of the gene.²⁴ Interestingly, knockdown of CPSF73 also inhibits Ser2p downstream of the B-globin gene.²⁵ Therefore, the phosphorylation of Ser2 within the CTD most likely stimulates the recruitment of 3′ end formation factors, which in a positive feedback loop strengthens Ser2 phosphorylation.²⁵

In mammals, mRNA 3′ end formation consists of an endonucleolytic cleavage, followed by polyadenylation of the free 3′ end. Transcription terminates often several kb downstream of the poly-A site^26-30 and is mechanistically linked to 3′ end formation.^31,32 The “core” of the 3′ end formation complex contains two highly conserved multi-subunit protein complexes, the cleavage and polyadenylation specificity factor (CPSF) and the cleavage stimulatory factor (CstF). CstF is a trimer of 50, 64 and 77 kD subunits in worms and mammals, required only for the cleavage step.^33-35 A dimer of CstF77 subunits is crucial for the assembly of the CstF complex, as it binds both CstF64 and CstF50, but CstF64 and CstF50 do not directly bind each other.^36-39 Importantly, CstF64 binds directly to the RNA at the U- or G/U- rich region downstream of the cleavage site, playing a key role in the location of the poly-A tail.⁴⁰ The N-terminal region of CstF50 binds to the CTD of RNAPII, and this interaction presumably facilitates recruitment of CstF to the elongating RNAPII.^7,8

C. elegans 3′ ends are quite similar to those in other animals, suggesting the usual 3′ end formation machinery processes them. They have a somewhat less highly conserved AAUAAA ∼30 nt upstream of their 3′ end cleavage site and a U- or GU-rich region ∼20 nt downstream.⁴¹ In addition, the C. elegans genome encodes clear homologs of all the subunits of the 3′ end formation complex.^42,43 Importantly, several 3′ end formation factors were identified in a C. elegans suppressor screen for genes that play roles in 3′ end formation.⁴⁴

We were interested in the question of how, the linkage of 3′ end formation to transcription termination is broken at internal 3′ end formation sites in operons. In mammals, transcription is terminated by the action of the XRN2 exonuclease that usually gains access to the free 5′ phosphate end of the pre-mRNA immediately downstream of the 3′ cleavage site.¹ This exonuclease digests the RNA, acting as a “torpedo” to cause termination several kb downstream, often at RNAPII pause sites. To gain some insight into this process in C. elegans, we have studied the distribution of RNAPII, its Ser2 phosphorylated form and two CstF subunits at the whole genome level, both in single genes and in operons. In addition, we reduced the levels of CstF50 to investigate its possible role in RNAPII distribution and CstF64 recruitment between regular and internal 3′ ends.

Earlier reports demonstrated large peaks of RNAPII paused at 3′ ends in C. elegans.^45,46 We have previously shown for a few genes that RNAPII at the 3′ end was phosphorylated at the Ser2 position, whether the gene was located in an operon or not. In this paper, we show that peaks of Ser2 phosphorylated RNAPII occur at 3′ ends of transcribed genes genome-wide, both in single genes and within operons, and that these peaks are accompanied by the 3′ end formation protein, CstF, irrespective of whether the 3′ end site was internal or terminal. Furthermore, we showed that when CstF50 was knocked down, CstF64 was still bound to the transcription site and Ser2p and total RNAPII levels at either regular or internal 3′ ends were unchanged.

Results

RNAPII distribution on C. elegans single genes

The unique organization of C. elegans genes provides an opportunity to investigate aspects of gene expression that cannot be studied using other systems. Worms have evolved compact operons that accommodate genes and their regulatory elements. However, around 85% of genes that are not in operons appear no different from genes in other species. Since the genome-wide localization of RNAPII and CstF has not been previously investigated, we began by exploring their localization in young adult worms with ChIP-seq experiments using antibodies specific for total RNAPII and Ser2p, which recognizes the CTD of RNAPII phosphorylated in the second serine of the heptad repeat. For RNAPII, we used the commercially available mouse monoclonal antibody 8wg16, which has been used extensively to study transcription. Our evidence suggests that in C. elegans the 8wg16 antibody likely recognizes both the unphosphorylated (IIA) and phosphorylated (IIO) form of RNAPII since it shows a ChIP pattern indistinguishable from an antibody made by the Bentley lab that recognizes both forms (Fig. S1).⁴⁷ A typical example (the soc-2 gene) taken from IGV (Fig. 1A) shows total RNAPII in the promoter region (indicated by H3K9ac), but no signal at that location with the Ser2 phospho-antibody, as expected based on ChIP results in other organisms.^19-22

Figure 1.

Localization of RNAPII and CstF subunits along isolated C. elegans genes. (A) Individual example gene. The gene model from IGV browser is aligned at the bottom with exons shown as boxes and introns as lines. The height of the ChIP peak represents the number of reads covering each location along the gene (soc-2). H3K9ac regions with greater than 90% hybridization over the genome are indicated by a box. (B–C) Metagene of 1,691 expressed genes. Metagenes were calculated by taking genes greater than 2 kb and scaling their gene body (+500 to +2500) to the same size. Five′ ends (−500 bp to +500 bp) and 3′ ends (2500 bp to 3500 bp) were not scaled. The signal was averaged at each base pair and the individual positions averaged across 50 bp windows. An arrow indicates the transcription start site and the 3′ end by a dotted vertical line. The x-axis indicates the location along the metagene in bp. The y-axis indicates normalized reads relative to an input DNA library (see Methods). (B) RNAPII and Ser2p. (C) CstF subunits. (D) Metagene of 242 non-expressed genes based on mRNA-seq data. RNAPII (red), Ser2p (purple), CstF64 (green), CstF50 (blue) and H3K9me3 (gray) were quantitated by ChIP-seq, and H3K9ac (black) by ChIP-on-chip.

The levels of RNAPII fluctuate throughout the body of the soc-2 gene in a pattern roughly matching the presence of exons (Fig. 1A). A similar bias of RNAPII enrichment at exons is evident throughout the genome (data not shown), consistent with ChIP and GRO-seq results from other experimental systems.^48–52 At the 3′ end of the soc-2 gene total RNAPII and Ser2p signal accumulate just upstream of the termination site presumably due to a strong pause at the site of 3′ end formation as seen previously.^20,28

In order to determine whether these patterns are similar across C. elegans genes, we created metagenes using a list of 1,691 expressed and isolated genes greater than 2 kb in length. This includes all genes with mRNA-seq greater than 1 FPKM (fragment per kilobase per million reads) and having a location separated from any annotated transcription by at least 400 bp. This analysis shows peaks of RNAPII at the promoter and 3′ end regions (Fig. 1B). The 5′ peak presumably represents RNAPII paused or poised near the promoter, whereas the 3′ peak is likely to represent enzyme paused to facilitate 3′ end processing.²⁸ Ser2p levels are low at the promoter, but with a very large peak at the 3′ end cleavage site. As expected, the H3K9me3 heterochromatin mark is low throughout the metagene, consistent with these genes being expressed. Interestingly, when levels of Ser2p were divided by total RNAPII (Fig. 2A), it is clear that the RNAPII is phosphorylated relatively evenly as it traverses the gene. However, at the 3′ pause site there is an additional increase in Ser2 phosphorylation, consistent with the reported role of Ser2p in binding 3′ processing factors.

Figure 2.

Ratios of Ser2p and CstF subunits in expressed genes. The ratio of Ser2p/RNAPII metagenes (A), (B) CstF64/RNAPII (salmon) and CstF50/RNAPII (purple), (C) CstF64/Ser2p (green) and CstF50/Ser2p (blue) and (D) CstF64/CstF50 (gray).

CstF distribution on C. elegans single genes

Next, we wanted to investigate the localization of the core 3′ end formation component CstF on C. elegans expressed genes. ChIP can be used to map RNA-binding proteins relative to the DNA as an indirect way for revealing the site of association with the pre-mRNA during RNA processing.⁵³ Identifying sites of association for RNA processing factors using ChIP can reveal important sites of action of these proteins during transcription, because many of these factors associate with the transcription elongation complex. They are therefore, in close proximity with the DNA.⁵³ ChIP experiments done in mammals have shown that CstF binding is biased toward 3′ ends of genes, consistent with its known function in 3′ end processing.^28,53,54 The C. elegans CstF64 ortholog is much smaller than CstF64 in other organisms, because it lacks much of the C-terminal region.^42,43 but it very likely functions in 3′ end formation.⁴⁴

In order to determine if the recruitment of the CstF proteins occurs at sites of 3′ end formation, we performed ChIP-seq experiments with antibodies prepared against the worm CstF64 and CstF50 subunits. A representative example (soc-2) shows that both CstF subunits are strongly enriched at the site of 3′ end formation (Fig. 1A), as expected based on its presumed function.^28,53,55 Metagene analysis on isolated/expressed genes shows that both CstF subunits are low at the 5′ end and throughout the body of genes (Fig. 1C). However, both subunits are strongly associated with the 3′ ends of expressed genes, with peaks matching the 3′ peaks of RNAPII. Importantly, consistent with these genes being transcriptionally silent, a similar analysis on a list of non-expressed genes based on mRNA-seq data showed very low CstF ChIP levels, and only the levels of the heterochromatin mark H3K9me3 were above background (Fig. 1D).

Biochemical experiments have shown that the CstF complex is a core component of the 3′ end formation machinery.^33-35,56,57 However, very little is known about the in vivo assembly of the complex. Since both CstF subunits peak at 3′ ends of genes similar to RNAPII, we wanted to determine whether the levels of CstF64 and CstF50 are similar relative to total RNAPII. The ratios of both CstF subunits to RNAPII are very low at the 5′ end and throughout the body of the metagene suggesting very little if any association of CstF with the transcription start site or the body of the gene (Fig. 2B). However, at the 3′ end of the metagene both CstF subunits increase quite dramatically relative to total RNAPII, consistent with their functioning in 3′ end formation. Unexpectedly, CstF64 increased more dramatically than did CstF50 when normalized to RNAPII levels or to themselves (Figs. 2B–D), suggesting the possibility that not every complex containing CstF64 necessarily also contains CstF50. An alternative explanation might be that the conformation of the complex allows preferential cross-linking or IP of CstF64. This idea is not unreasonable, since CstF64 binds the pre-mRNA, whereas CstF50 binds the CTD of RNAPII. The ratios of CstF subunits to Ser2p signal are higher at the 5′ end compared to the 3′ end, presumably due to the dramatically increased level of Ser2p at the 3′ ends of genes (Fig. 2C).

RNAPII distribution on C. elegans operons

The discovery of operons was based on the observation that downstream genes in clusters of closely spaced genes were trans-spliced to the SL2 trans-spliced leader.⁵⁸ Since then a great deal of evidence from a variety of laboratories has accumulated that there are in fact more than 1,200 operons in the C. elegans genome.^{29,44,46,59,68} ChIP experiments using an antibody against RNAPII maps its in vivo association with DNA, allowing us to identify polycistronic transcription units. We present the genome-wide association of RNAPII on operons. Fig. 3 shows three operon examples of two, three and eight genes. We used H3K9ac to mark the position of promoters near the 5′ end of each operon (Fig. 3), and at a known internal promoter within the eight-gene operon, providing independent evidence for this internal initiation site ^45,60,61 (Fig. 3C). We found RNAPII present throughout operons, with notable peaks at 3′ ends (Figs. 3A–C). Ser2p signal also peaks at each 3′ end, matching the location of total RNAPII (Figs. 3A–C).

Figure 3.

Representative snapshots of C. elegans operons from ChIP experiments. RNAPII, Ser2p, CstF64 and CstF50 are ChIP-seq experiments and H3K9ac (black) is a ChIP-on-chip experiment. Dotted lines mark genes boundaries. Transcription is from left to right. (A) A two-gene operon, CEOP1452. (B) A three-gene operon, CEOP5100. (C) An eight-gene operon, CEOP1484, which contains an internal promoter (red arrow).

In order to study the localization of RNAPII on operons genome-wide, we created “meta-operons,” consisting of metagenes created using lists of first, internal and terminal genes. Total RNAPII accumulates at the operon promoters, which can be distinguished from 3′ end signals in the first gene of each operon (Fig. 4A). Ser2p signal is strongly associated with each 3′ end in the operon, coinciding with a peak of total RNAPII. This supports and extends to a genome-wide level our previous study⁴⁵ showing that each gene in the operon is indistinguishable from individual genes with respect to RNAPII and Ser2p distribution.

Figure 4.

Localization of RNAPII and CstF subunits in expressed C. elegans operons. Metagene of operon genes divided by first gene (n = 184), internal genes (n = 111) and terminal genes (n = 705). (A) Metaoperon showing RNAPII, Ser2p and H3K9me3. (B) Metaoperon showing CstF64, CstF50 and H3K9me3.

CstF distribution on C. elegans operons

It is now well established that 3′ end formation is needed for correct transcription termination. In fact, several mRNA cleavage and polyadenylation factors required for 3′ end formation are also required for transcription termination.^27,62,63 Moreover, both 3′ end formation and termination depend on the same RNA sequences.^31,64 However, genes arranged in operons contain 3′ end formation sites not accompanied by termination. The mechanism by which RNAPII is able to transcribe functional poly-A sites without triggering termination is not known. One possibility might be that internal cleavage and polyadenylation occurs by a different mechanism than 3′ end formation at terminal sites. To test whether CstF functions in 3′ end formation at internal sites as well as termination sites, we performed anti-CstF ChIP-seq experiments. Individual examples of CstF64 and CstF50 taken from IGV in two, three and eight gene operons are shown in Fig. 3. In all cases both CstF subunits are strongly associated with internal and terminal 3′ ends, consistent with 3′ end formation occurring at these sites by the usual CstF-dependent mechanism. In addition, the same bias toward 3′ ends is seen with meta-operons (Fig. 4B), indicating that this pattern is representative of operons at a genome-wide level. The CstF binding seen at 3′ ends co-localizes with elevated Ser2p and with paused RNAPII (Figs. 3 and 4). The presence of paused Ser2 phosphorylated RNAPII, accompanied by the 3′ end formation machinery at the 3′ ends of genes at internal operon positions suggests that a mechanism must exist to prevent transcription termination when standard cleavage and polyadenylation occurs at these sites. A likely possibility is that the SL2 trans-splicing that co-occurs with 3′ end formation provides a 5′ cap to the downstream RNA, so the XRN2 exonuclease cannot act to cause termination by the usual torpedo mechanism.⁶⁵

CstF50 plays a role in CstF64 localization upstream of the cleavage site

The nascent pre-mRNA is rapidly processed into mature mRNA at the site of transcription. In this process the CTD of RNAPII acts as a landing pad for the recruitment of protein complexes, thereby facilitating co-transcriptional processing. In vitro experiments have shown that the amino-terminal region of CstF50 interacts with the CTD of RNAPII and this interaction is important for 3′ end formation.⁸ Therefore, one possibility is that the CstF complex could be recruited to RNAPII at the 3′ ends of genes by CstF50 binding to the CTD. Whether CstF50 is needed for the recruitment of the CstF complex to the 3′ formation site has never been investigated. We asked whether CstF64 recruitment was affected when CstF50 was knocked-down by RNAi. We reduced CstF50 levels by more than 70% (CstF64 levels were unchanged under these conditions) (Fig. S2A). Both 3′ end formation and termination were reduced, as expected based on mRNA-seq and ribosomal-depleted RNA-seq data, respectively (Figs. S2B–C). Levels of RNAPII and Ser2p throughout the gene were unaffected by this treatment (Figs. S3 and 4).

Interestingly, CstF64 was still recruited to 3′ ends of expressed genes when CstF50 levels were reduced, although to a lower extent (Fig. 5A), suggesting the CstF50/RNAPII interaction may facilitate recruitment of the CstF complex. Similarly, on meta-operons we found that on average CstF64 was recruited at both internal and terminal 3′ ends (Fig. 5B). However, we cannot exclude the possibility that the residual CstF50 is sufficient for most CstF64 recruitment.

Figure 5.

CstF64 localization to C. elegans genes and operons is altered by CstF50 knockdown. ChIP-seq experiments of CstF64 in untreated (green) and CstF50 RNAi'ed (light green) samples. (A) Metagene analysis of expressed and isolated genes based on mRNA-seq data (n = 1,691). (B) Metaoperon of first (n = 184), internal (n = 111) and terminal (n = 705) genes.

We do note one intriguing change: CstF64 binding immediately upstream of the cleavage site is reduced when levels of CstF50 are lower (Fig. 5, arrows). We are confident that this represents a significant reduction, since it is seen with single genes and all three classes of operon genes, all representing independent datasets. Perhaps under normal conditions CstF50 brings CstF64 to RNAPII at the transcription site before the 3′ end signal has been synthesized. However, under low levels of CstF50, CstF64 is not recruited until the pre-mRNA beyond the 3′ end is synthesized, since CstF64 is the RNA binding subunit. We also wanted to investigate CstF50 recruitment in the absence of CstF64, but CstF64 knockdown was lethal, so we were unable to perform this experiment.

Discussion

Our data have implications for how transcription termination may be prevented at internal 3′ ends in operons. We used ChIP-seq to map the in vivo localization of RNAPII at a genome-wide level, which showed that all 3′ ends contain paused RNAPII phosphorylated at Ser2, including internal 3′ ends in operons, which are not accompanied by termination. This indicates that the 3′ pause is a feature of 3′ end formation, rather than of transcription termination. Furthermore, we found that this peak of RNAPII at 3′ ends of genes co-localizes with maximal recruitment of the CstF complex, suggesting that RNAPII molecules at internal 3′ end in operons should be competent for termination. Thus, the hypothesis that RNAPII is prevented from terminating at internal 3′ ends in operons by restriction of XRN2 access by the 5′ cap provided by SL2 trans-splicing seems likely.

RNAPII pauses at each 3′ end within operons, thus treating these multi-gene loci as if they were independent genes. C. elegans contains clear homologs of all CPSF subunits⁴⁴ and has been shown to contain the canonical 3′ end formation sequences at both internal and terminal 3′ ends.⁴¹ RNAPII pausing is a known characteristic of poly-A signal transcription. Indeed, RNAPII pausing at 3′ ends of mammalian core histone genes, which code for non-polyadenylated transcripts, is brief or absent.⁶⁶ The paused RNAPII at the 3′ end of all operon genes co-localizes with Ser2p. This is consistent with the finding that RNAPII pausing downstream of the 3′ end of genes is sufficient for Ser2 phosphorylation.²⁵ Mutational analysis performed in human cells using an RNAPII lacking serine 2 resulted in impaired 3′ end formation as well as loss in the binding of the 3′ end factor PCF11.²³ Moreover, RNAPII levels that are maximally phosphorylated on serine 2 tend to peak downstream of the site of polyadenylation.^19–22 The fact that the peak of 3′ pausing and Ser2 phosphorylation occurs right at the site of 3′ end cleavage in worms is an interesting difference from mammals. Therefore, Ser2p at 3′ ends of internal genes in operons most probably functions in guiding 3′ end formation, similar to terminal and single gene 3′ ends. Since the ratio of Ser2p to total RNAPII does not change appreciably as the operon is traversed, it seems likely that RNAPII remains phosphorylated on Ser2 throughout the operon (Fig. 4). It appears the peaks at each 3′ end represent sites of pausing, rather than of new Ser2 phosphorylation. The one possible exception is the 3′ end of the terminal gene in the operon, where Ser2p increases more than total RNAPII. This may imply a special role for Ser2p in transcription termination.

ChIP experiments done in C. elegans and mammals have shown that RNAPII continues transcribing downstream of the poly-A site for another 0.5 kb1–.5 kb.^28,29 Prior to termination, RNAPII is in a paused conformation, co-localizing with maximal Ser2p and bound by high levels of CstF64 and 77.²⁸ In a similar situation, we have shown that internal 3′ ends in operons are both maximally phosphorylated at Ser2 and contain high levels of CstF50 and 64. In contrast to mammals, the peak of paused RNAPII containing Ser2p and CstF in worms occurs much closer to the poly-A site, which may simply reflect a more compact genome, or perhaps a somewhat different relationship between 3′ end pausing and termination. Overall, these results provide evidence that RNAPII is pausing at all 3′ ends in operons, suggesting that internal 3′ ends in operons is primed for termination but is prevented by a mechanism other than affecting RNAPII or CstF recruitment. Liu et al.,⁶⁵ proposed that RNAPII is prevented from terminating at internal 3′ ends by a factor bound to the Ur element, a sequence between operon genes required for SL2 trans-splicing. This factor could temporarily block the passage of the XRN-2 exonuclease until SL2 trans-splicing provides a cap to the downstream RNA.^1,2

Several pre-mRNA processing factors have been shown to bind the phosphorylated form of RNAPII CTD, leading to the proposal that these phosphorylations may guide the co-transcriptional processing of the pre-mRNA (for a review see ref. ¹⁶). However, for the case of Ser2p the possibility remains that these phosphorylation events are not needed for guiding pre-mRNA processing, but instead for events following pre-mRNA processing, such as transcription termination. We found that Ser2p is associated with all 3′ ends in operons, even those not associated with transcription termination. This result provides novel and independent evidence that, again, this phosphorylation functions in 3′ end formation rather than transcription termination. Furthermore, this indicates that Ser2p cannot be sufficient for causing transcription termination, since Ser2p is present at 3′ ends in which termination does not occur following 3′ end formation.⁶⁹

How is CstF recruited? The ability of CstF50 to bind the CTD⁸ may be important, as may the ability of CstF64 to bind RNA. Interestingly, the CstF50 ChIP profile along genes does not match that of RNAPII, except at 3′ ends, suggesting the possibility that the CstF50/CTD interaction only occurs at 3′ ends of genes when RNAPII is paused. CstF64 was still recruited to gene 3′ ends when CstF50 was reduced by RNAi, albeit to a lesser extent. These results suggest that CstF64 could be recruited to the transcription site directly by the RNA and independent of CstF50. Perhaps both the protein/protein and protein/RNA interactions act together to ensure CstF recruitment.

Methods

Worm culture and growing conditions

Mixed stage and synchronized Bristol (N2) worms were maintained and grown as described by Sulston and Brenner.⁶⁷ Worms for ChIP experiments were synchronized and grown in liquid culture until the young adult stage with few eggs. For other assays, worms were grown in NGM plates. The worms were then washed three times in water and the bacteria were cleared by sucrose flotation.

Formaldehyde in vivo cross-linking

Worms were frozen in liquid nitrogen and grounded to a powder with a mortar and pestle. The resulting worm powder was transferred to cross-linking buffer (1 mM PMSF, 1 mM EDTA, 1 mM EGTA, 1% formaldehyde, PBS) for 10 minutes at room temperature. The reaction was quenched for 5 minutes at room temperature by addition of glycine to a final concentration of 125 mM, and the mixture was sedimented at 4,000 g. Pellets were washed three times with FA buffer + 0.1% SDS (50 mM HEPES pH 7.5, 1 mM EDTA, 1% Triton, 0.1% deoxycholic acid, 150 mM NaCl, 0.1% SDS) containing one protease inhibitor cocktail tablet (Roche). Each wash was sedimented at 4,000 g. Then pellets were divided into 500 uL aliquots and stored at −80°C.

Chromatin immunoprecipitation (ChIP)

Each aliquot was resuspended in 1.5 mL of FA buffer + 0.3% SDS containing protease (Roche complete cocktail tablets 11697498001) and phosphatase inhibitors (GBiosciences 786-450). The samples were sonicated with a Virsonic digital 600 sonicator using a microtip (20 pulses of 11 seconds each at 30% amplitude with bursts of 0.9 seconds on and 0.5 seconds off), to generate DNA fragments of approximately 500 bp, as determined experimentally. Samples were sedimented at 13,000 g for 15 minutes at 4°C, and the supernatant transferred to a new tube, which was then diluted to 4.5 mL with FA buffer containing protein and phosphatase inhibitors. The extract was divided into four 1 mL samples.

Immunoprecipitation and elution were performed according to Lee et al. ⁷⁰ with some modifications: For each 1 mL extract, 100 uL of protein A Dynabeads (Invitrogen), conjugated with antibody, were added. After a 4°C overnight incubation, the beads were washed five times with RIPA buffer (50 mM HEPES pH 7.5, 1 mM EDTA, 1% NP-40, 0.7% deoxycholic acid, 0.5 M LiCl) and once with TE + 50 mM NaCl. The beads containing the protein–DNA complex were transfer using cold TE to a clean eppendorf tube before elution. After removal of any TE buffer, 210 uL of elution buffer (50 mM Tris-HCl pH 8, 10 mM EDTA, 1% SDS) was added and incubated at 65°C for 30 minutes, with intense mixing briefly every 5 minutes. The elution was separated from the beads by applying the magnet. 10 uL of proteinase K (20 mg/mL) was added to the eluted fraction and incubated at 55°C for 2 to 3 hours. Alternatively, in recent versions of the protocol the RNA was degraded from the IP DNA to improve the recovery from the Qiagen PCR purification columns. In these cases, 30ug of affinity purified RNAse A (Ambion) was added to each tube and incubated at 37°C for ∼2 hours. Then, 10 uL of proteinase K (20 mg/mL) was added to each eluted and incubates at 55°C for 2 hours. The tubes in either version were transferred to 65°C overnight to reverse the crosslinks.

DNA isolation and purification

For ChIP-on-chip the DNA was purified using the Affymetrix cDNA cleanup kit eluted twice with 20 uL of elution buffer. For ChIP-seq the DNA was purified using a Qiagen column and eluted twice with 30 uL of water.

Affymetrix tiling arrays

We used the GeneChip C. elegans Tiling 1.0R array from Affymetrix. Each array is composed of 3.2 million probes spanning the whole non-repetitive worm genome. The probes are tiled at an average resolution of 25 base pairs, measured from the center of the adjacent probe. The preparation of the IP DNA for hybridization was done according to the manufacturer's protocol. Briefly, PCR was performed on the purified IP DNA for dUTP incorporation; cycle number is dependent on the quality of the antibody used and maintained of the enrichment post-amplification was determined experimentally by qPCR. Next, the amplified IP DNA was enzymatically fragmented (uracil DNA glycosylase and APE-1) into smaller pieces (< 100 bp) and labeled with biotin at their 3′ ends with terminal deoxynucleotidyl transferase. Finally, the fragmented and labeled DNA was hybridized to the array.

ChIP-seq library preparation

Preparation of the DNA library was performed according to the manufacturer's protocol (Illumina). Approximately, 10 ng of enriched IP DNA was used for the library preparation. The overhangs resulting from the ChIP experiment were converted into blunt ends using the end repair mix. Single “A” nucleotides were added to 3′ ends of the blunt fragments to prevent them from self-ligating. Different sequence adapters were ligated to the 5′ and 3′ ends, follow by gel extraction of ligation products between 250 and 300 bp. PCR of 20 cycles was used to enrich DNA fragments containing the adapters, allowing the DNA to bind the flow cell. The library was first validated by fluorescence Qubit (Invitrogen) and Agilent Technologies 2100 Bioanalyzer, and then sequenced using the Illumina Hiseq 2000 platform, multiplexing two samples in a single lane (Table 1).

Table 1.

ChIP-seq and RNA-seq library information. All libraries were sequenced with an Illumina Hiseq2000 machine. The number of mapped reads was calculated in each case using Samtools flagstat function on Bam files.

ChIP-seq libraries	# map reads	Raw data	Processed data
CstF64	55,054,346	cstf64.fq	cstf64_WT.wig.gz
CstF64 rep1	42,066,080	CstF64_TCA_L001_R1_001.fastq	CstF64_repeat_WT.wig.gz
CstF50	71,610,607	Cstf50_OLD_TTCAGC_L004_R1_001.fastq	cstf50.wig.gz
CstF50 rep1	56,192,990	CstF50_AB_GACG_L002_R1_001.fastq	cstf50-rep.wig.gz
Total RNAPII (8wg16)	38,121,201	8wg16_ACG_L008_R1_001.fastq	8wg16.wig.gz
Total RNAPII (Bentley) rep1	23,395,444	Bentley_CTD_rep.fastq	BentleyCTD_rep.wig.gz
Total RNAPII (Bentley) rep2	37,584,291	Bentley_CTD.fastq	BentleyCTD.wig.gz
Ser2p	80,681,498	Ser2p.fastq	ser2p.wig.gz
Ser2p rep1	49,376,809	SER2p_GGA_L001_R1_001.fastq	Ser2p_repeat.wig.gz
CstF64 in RNAi	51,618,437	CstF64_RNAi.fq	CstF64_RNAi.wig.gz
CstF64 in RNAi rep1	29,001,792	CstF64_RNAi_rep_TCA_L001_R1_001.fastq	cstf64-RNAi-rep1.wig.gz
CstF64 in RNAi rep2	22,194,982	CstF64_RNAi_1_GGA_L001_R1_001.fastq	CstF64_RNAi__rep2.wig.gz
8wg16 in RNAi	74,005,112	8wg16_RNAi_CCTCGG_L006_R1_001.fastq	8wg16_RNAi.wig.gz
8wg16 in RNAi rep1	72,988,938	8wg16_RNAi_Rep1_GGACCC_L006_R1_001.fastq	8wg16_RNAi_rep.wig.gz
Ser2p in RNAi	46,298,353	Ser2p_RNAi_1_CCC_L008_R1_001.fastq	Ser2p_RNAi.wig.gz
Ser2p in RNAi rep1	47,296,297	Ser2p_RNAi_2_AGC_L008_R1_001.fastq	ser2p_RNAi_rep.wig.gz
H3K4me3	117,399,626	N2-1_GGACCC_L005_R1_001.fastq.gz	H3K4me3.wig.gz
Input DNA	68,623,378	N2_r1-suffix.fastq, N2_r2-suffix.fastq	InputDNA.wig.gz
	k
RNA-seq libraries	# map reads	Raw data	Processed data
Empty vector PolyA	9,941,393	SK412_13_NoIndex_L005_R1_001.fastq	cuffdiffs_mRNA.txt
Empty vector PolyA rep1	6,756,942	SK412_14_NoIndex_L006_R1_001.fastq	cuffdiffs_mRNA.txt
CstF50 RNAi PolyA	9,537,397	SK412_15_NoIndex_L007_R1_001.fastq	cuffdiffs_mRNA.txt
CstF50 RNAi PolyA rep1	8,786,963	SK412_16_NoIndex_L008_R1_001.fastq	cuffdiffs_mRNA.txt
Empty vector total	184,994,326	N2_totalRNA_file1.fastq, N2_totalRNA_file2.fastq	Deseq_totalRNA.txt
Empty vector total rep1	93,764,790	N2_totalRNA_file3.fastq, N2_totalRNA_file4.fastq	Deseq_totalRNA.txt
CstF50 RNAi total	326,670,681	cpf1_RNAi-1_GCCAAT_L008_R1_001.fastq, cpf1_RNAi-1_GCCAAT_L008_R2_001.fastq	Deseq_totalRNA.txt
CstF50 RNAi total rep1	268,564,434	cpf1_RNAi-2_CTTGTA_L008_R1_001.fastq, cpf1_RNAi-2_CTTGTA_L008_R2_001.fastq	Deseq_totalRNA.txt

RNA-seq library preparation

RNA was extracted using Trizol reagent (Invitrogen) following manufacture recommendations. For poly-A selected RNA, total RNA was shipped to Dr Scott Kuersten PhD at Epicenter for poly-A selection, library construction and sequencing. For total RNA samples, Ribozero (Illumina) was used to remove rRNA contamination followed by ScriptSeq RNA-Seq Library Preparation Kit (Illumina) (Table 1).

Galaxy: reads manipulations

Galaxy is an open, web-based platform for accessible, reproducible and transparent computational biomedical research. We used Galaxy for processed and mapped all sequencing reads. Bases from reads were trimmed from either end that had a quality score less than 31 (scale 0–40), then the entire read length was filtered by a quality score no less than 20 and allowed 0 bases outside this range. Adapters and barcodes were cleaved from the ends of each read, and mapped to the C. elegans genome (ce10) using Bowtie (default settings).

Peak calling algorithm

The model-based analysis of ChIP-seq (MACS 1.0.1) algorithm was used from the galaxy web browser to find statistically significant peaks. Bam files containing aligned reads were used as input files for MACS, which outputs a wiggle file at 1 bp resolution (used for python script) and a Bed file. Based on the parameters used on galaxy, MACS slides 500 bp windows across the genome to find regions containing reads 2-fold enriched relative to a background level. For each candidate peak a background level is calculated by counting the number of reads in a 5 kb, or 10 kb window centered from the peak location. Using a p value cutoff of 1e–5, the candidate peaks are selected and reported as fold enrichment over the local background level.

Python script

Will Kruesi, from the B. Meyer lab at the University of California at Berkley, wrote the script on python called “average profiles.” It is used to create metagenes around a feature (poly A site), or across a gene by scaling them to the same size. MACS created all wiggle files and they were normalized using the java-genomics-toolkit package against a DNA input control. The script calculates average fold enrichment by adding the level of signal at each base pair and divides it by the number of genes. If a gene 3′ end is not present in the input wiggle file then nothing will be added to the total, but it will still be counted toward the average. All metagene analysis was done using the transcription start site based on Kruesi et al.⁴⁶

Meta-operons

Expressed operons (first, internal and terminal) were obtained using a recently published list of operons (69) requiring an FPKM > 20 in the RNA-seq data. Only operons containing an intercistronic space between 90 and 120 bp, high levels (>80%) of SL2 trans-splicing to the downstream gene and greater than 2 kb in length were considered in the analysis (n = 1,000).

CstF50 knock-down

Strain containing a balanced deletion allele for CstF50 (tm 4163) was obtained from the Caenorhabditis Genetic Center (CGC). We used RNAi by feeding to significantly knocked-down CstF50 levels in this balance deletion strain. Bacteria expressing dsRNA corresponding to part of the CstF50 locus were used to feed synchronized starved L1s either on plates or liquid culture. For the RNAi done in plates, the bacteria was induced by adding log-phase grown bacteria to NGM plates containing ampicillin and IPTG and left at room temperature for at least 2 hours before adding the starved L1s. For the RNAi done in liquid, the bacteria was previously induced before being fed to the starved L1s. Briefly, a single colony of the CstF50 RNAi bacteria was used to inoculate a 3 mL overnight LB broth culture. Then, 100 uL of this overnight was used to inoculate 1 L of LB broth and grown at 37°C until log-phase (∼3 hours). Next, IPTG was added to a final concentration of 1 mM and induced at 37°C overnight. Finally, the bacteria were sedimented at 4,000 g and fed to the starved L1s. In both plates and liquid the worms were allowed to grown on the RNAi bacteria for 3 to 4 days at 20°C.

Measuring transcription termination in CstF50 RNAi worms

In order to calculate transcription termination, we deep sequenced ribosomal depleted RNA from control and CstF50 RNAi samples. To determine termination defects, we quantify the number of reads in a 50 bp window downstream of the cleavage site. Then these reads were normalized to the total number of mapped reads in each data set and compared to each other.

Antibodies

α-Ser2p antibodies were from Bethyl Laboratories (A300-654A). The “total RNAPII” antibody is either a rabbit polyclonal antibody raised against the recombinant mouse CTD (52 repeats) protein, which was a gift from David Bentley, or the 8wg16 antibody that is commercially available from Millipore (05-952). α-CstF64 antibody is a rabbit polyclonal antibody raised against the recombinant full-length C. elegans protein. α-CstF50 antibody is a polyclonal antibody raised against a synthetic peptide encoded in the N-terminal region of the protein. α-H3K9ac is a commercially available rabbit polyclonal antibody from Abcam (4441).

Abbreviations

ChIP

chromatin immunoprecipitation

CPSF

cleavage and polyadenylation specificity factor

CstF

cleavage stimulatory factor

CTD

C-terminal domain

RNAi

RNA interference

RNAPII

RNA Polymerase II

Ser2p

serine 2 phosphorylation

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgments

We gratefully acknowledge the technical help of Chad Cockrum. We thank the sequencing facilities at the University of Colorado, Boulder and the Anschutz Medical Campus for Illumina sequencing, D. Bentley for comments on the manuscript and for the RNAPII CTD antibody. We are also grateful to Chris Link for reagents. We thank Will Kreusi and Barbara Meyer for communication and help with data analysis. In addition, we thank Scott Kuersten at Epicenter for the mRNA-seq.

Funding

This research was supported by research grant R01 GM42432 from the National Institute of General Medical Sciences.

Supplemental Material

Supplemental Material may be downloaded here: publisher's website