| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
J. Biol. Chem., Vol. 278, Issue 43, 41691-41701, October 24, 2003
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
From the Department of Chemistry and Biochemistry and the Molecular Biology Institute, University of California at Los Angeles, Los Angeles, California 90095-1569
Received for publication, June 16, 2003 , and in revised form, July 10, 2003.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
In our work we have focused on mammalian poly(A)-dependent termination, for which a functional poly(A) signal is both necessary and sufficient to direct transcription termination in vivo (2, 3). The mammalian poly(A) signal typically consists of a core sequence, about 45 nt long, that may be flanked by diverse auxiliary sequences that serve to enhance cleavage and polyadenylation efficiency (4). The core sequence consists of a highly conserved upstream element (AAUAAA), recognized by cleavage and polyadenylation-specificity factor (CPSF),1 and a poorly defined downstream region (rich in Us or Gs and Us), bound by cleavage stimulation factor (CstF). The poly(A) cleavage site lies in between these two elements.
Although cleavage at the poly(A) site has long been thought to be a key component of the termination mechanism (1, 5, 6), this appears increasingly unlikely for several reasons. First, signaling from the poly(A) site to the polymerase to stop transcription does not require cleavage at the poly(A) site either in HeLa extracts (7) or in Xenopus oocytes (8). Second, the mammalian histone genes, which have uniquely different pre-mRNA 3'-end processing signals and which require these signals for transcription termination, do not require processing itself as part of the mechanism (9, 10). Third, the mammalian U2 small nuclear RNA genes, which are transcribed by RNA polymerase II but processed differently than mRNAs, also require the 3'-end processing signal— but not processing itself— for termination to occur (11). Fourth, in yeast, which exhibits poly(A)-dependent termination similar to mammals, cleavage at the poly(A) site is not required for termination (12). Therefore, it is likely that the mammalian poly(A) signal delivers its instructions to the polymerase and drives termination, as in these other cases, independent of the actual processing event itself.
At what point, then, does the poly(A) signal make its presence known to the polymerase, and how does it trigger termination? There are three general possibilities. First, it is possible that recognition of the poly(A) signal by the transcription apparatus occurs intrinsically, and that the poly(A) signal acts to ensure termination even before it is extruded from the polymerase. Escherichia coli RNA polymerase and its associated factors are able to monitor DNA sequence throughout the 35-bp footprint of the polymerase to receive regulatory input for all aspects of elongation—pausing, termination, and anti-termination (13–15). For RNA polymerase II the DNA contacts are even more extensive (16). Thus, recognition of the poly(A) signal could occur, in part, at the DNA level. Additional opportunities for intrinsic regulatory input come from interactions of the RNA polymerase with the DNA-RNA hybrid, and with the RNA itself in the exit tunnel (13). Moreover, many RNA processing factors, including those involved in termination, bind directly to RNA polymerase II (12, 17–20). Because these factors could alter the manner in which the polymerase interacts with the RNA, DNA, and hybrid components of the poly(A) signal within the ternary complex, their effects on termination could be more direct than merely via their role in processing. The ability of transcription factor IIS to alter fundamentally the way the polymerase interacts with its transcript while the transcript is still within the enzyme (21) illustrates how factors associated with the polymerase might communicate with the transcript prior to extrusion. This intrinsic model predicts that the poly(A) signal has already triggered the changes that lead to termination even before it has been extruded from the polymerase.
A second possibility for triggering termination is that this occurs upon extrusion of the poly(A) signal, when it encounters factors riding with the elongation complex. One model of this type is the "anti-terminator" model (22–24), which suggests that as the poly(A) signal emerges from the polymerase it triggers a response in the transcription complex that over-rides the effects of an associated anti-termination factor. Calvo and Manley (24) have suggested that one such anti-termination factor may be PC4 (positive cofactor 4). Human PC4 interacts with CstF, which, together with CPSF, forms the core of the cleavage and polyadenylation apparatus (4). CstF also binds the C-terminal domain (CTD) of the polymerase large subunit (17). Because CPSF has been shown to join the polymerase at the promoter in vitro (25), and CstF and CPSF are found complexed with each other (26) and with RNA polymerase II (27) in nuclear extracts, it is widely believed that a CstF·CPSF complex rides down the template with the elongating polymerase (see "Discussion"). Based on studies with the yeast homologs of PC4 and CstF, Calvo and Manley (24) suggested that PC4 interacts with CstF during elongation to suppress a termination activity triggered by CstF until the poly(A) signal emerges from the polymerase. The extruding poly(A) signal then binds to CstF (and CPSF) thereby triggering relief of the PC4-mediated suppression, which leads to termination (24).
The idea that termination is triggered upon extrusion of the poly(A) signal is consistent with current thinking that processing factors ride with the polymerase for the purpose of facilitating rapid (zero order) interaction with the pre-mRNA processing signals as they emerge from the polymerase (28, 29). For 3'-end processing the case is most compelling for CPSF, which recognizes the conserved AAUAAA hexamer of the poly(A) signal. In both mammalian cells and yeast, CPSF or its homologs (but not CstF) are associated with the general transcription initiation factor TFIID (25, 30). In an in vitro mammalian system the CPSF is transferred from TFIID to the elongating polymerase where it is thought to scan the RNA, awaiting the nascent poly(A) signal (25). Just such a scanning mechanism appears to be used by vaccinia virus in the termination of early gene transcription (31–33). In vaccinia, termination is signaled by a UUUUUNU sequence in the nascent RNA. The scanning mechanism that recognizes this UUUUUNU signal is so efficient that the recognition event, as well as the trigger to terminate, and termination itself, all occur in less than 60 nt following extrusion of this signal in vivo (34). It is likely that the cellular poly(A) signal is recognized with similar efficiency (28, 29). According to the anti-terminator model this recognition event, which is temporally linked to extrusion of the poly(A) signal, triggers a chain of events that culminates in termination of transcription.
In the absence of special auxiliary elements designed to hasten termination, the process is gradual, and apparently first order (3). Why is poly(A)-dependent termination first order? One possibility is that triggering occurs immediately (a zero order process) and that the termination-prone polymerases then dissociate from the template according to first order kinetics. Alternatively the zero-order process could correspond to the activation of an enzymatic activity, and this activity may then progressively modify the transcription complex as it travels down the template. Either of these possibilities would be consistent with the two triggering scenarios discussed above, in which the poly(A) signal triggers termination either from within the ternary complex or else immediately following extrusion.
However, a third possibility is that the triggering event itself is generated according to first order kinetics. For example, the trigger to terminate could correspond to the pseudo-first order recruitment of a termination factor by the poly(A) signal. Alternatively, one or more factors engaged by the poly(A) signal following its extrusion may undergo a slow, poly(A)-dependent, conformational change.
Thus, there are at least three scenarios for triggering termination that are consistent with existing data, triggering from within the transcription complex, triggering upon extrusion of the poly(A) signal from the transcription complex, and triggering pursuant to some additional first order process in which the poly(A) signal participates. These three possibilities predict different time intervals between the moment of poly(A) signal transcription and the point at which it triggers termination. The temporal relationships of these events can be explored by use of cis-antisense inhibition (35), which involves placing an antisense sequence at various distances downstream of the poly(A) signal. Because transcription of the antisense leads to inactivation of the poly(A) signal by means of sense:antisense duplex formation (35), this approach can be used to measure the window of opportunity (the time between transcription of the sense and antisense elements) available for the poly(A) signal to act (35).
In the experiments reported below we show that the ability of the poly(A) signal to drive termination in vivo can be blocked by a downstream antisense sequence. This vulnerability to antisense shows that the poly(A) signal does not act prior to extrusion. Moreover, an antisense sequence placed several hundred bp downstream of the poly(A) signal allows gradual termination of polymerases up to the position of the antisense, but blocks further termination downstream. For those polymerases whose termination is blocked, this indicates that events surrounding the extrusion of the poly(A) signal are not sufficient to trigger a loss of processivity in the polymerase. Instead, the gradual release of polymerases over time requires active communication of each polymerase with the poly(A) signal right up until the time the polymerase terminates transcription.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
|
G-less Run-on Transcription Assays—Run-on transcription, extraction of the RNA with TRIzol, and the detection of G-less cassettes was essentially as described previously (3) with the correction that 20 units of anti-RNase (Ambion) was added to the run-on transcription reaction. For the assays shown in Fig. 1, 12 units of RNase inhibitor (Ambion) was used in place of anti-RNase, and the RNA was purified using RNeasy columns (Qiagen), which was also described previously (3). Since the experiments reported here were carried out we have also begun to include 0.5 mg/ml of heparin in the run-on transcription reaction, which improves the signal to noise ratio (data not shown).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The G-less cassette/nuclear run-on method of quantitating poly(A)-dependent termination is illustrated by constructs 1 and 2 in Fig. 1, A and B. The top diagram in Fig. 1A shows the layout of these constructs with either the mutant or wild type SV40 late poly(A) signal, L, followed by a cassette window defined by two G-less cassettes 377 and 261 bp in length. Transfection of these plasmids into COS cells followed by nuclear isolation, run-on transcription, RNase T1 digestion, and polyacrylamide gel electrophoresis gave the results shown in lanes 1 and 2 of Fig. 1B. In lane 1 it is apparent that both cassettes downstream of the mutant poly(A) signal yielded strong bands. This indicates that there were elongating polymerases located within both of these cassettes at the time of nuclear isolation. In contrast, lane 2 of Fig. 1B shows that there were few polymerases within the distal cassette when a wild type poly(A) signal was located upstream. This reflects a stochastic, distance-dependent loss of polymerases in vivo due to poly(A)-dependent termination as described previously (3).
The ratio of the polymerase density in the post-cassette relative to that in the pre-cassette (post/pre ratio in Fig. 1B) is a measure of the efficiency of elongation across the cassette window (3). This elongation efficiency is clearly much lower downstream of the wild type poly(A) signal (0.14) than downstream of the mutant (0.96). We find it convenient to express the poly(A)-dependent decrease in elongation efficiency for the wild type as a percentage of that for the mutant (normalized post/pre ratio in Fig. 1B). Thus, of the polymerases entering the cassette window in this experiment, only 15% as many on the wild type template make it to the end of the window as on the mutant template (Fig. 1B, lanes 1 and 2). This is equivalent to 85% poly(A)-dependent termination across the cassette window.
Cis-antisense inhibition (35), which inactivates the poly(A) signal only after it leaves the polymerase, can be used to determine whether poly(A) signaling occurs before or after extrusion. To inactivate the poly(A) signal using a cis-antisense sequence, an inverted segment, 
, corresponding to the core of the poly(A) signal (AAUAAA hexamer plus G/U-rich region) was inserted into the HindIII site of construct 2 (Fig. 1A) to give construct 3. Thus, transcription of will produce an antisense that can occlude the poly(A) signal by forming a stem-loop in the RNA. The L27L51 designation in the plasmid name indicates that the antisense is directed against L (L) and that L and form a stem of 51 bp joined by a 27-nt loop. Often, as in Fig. 1C, we abbreviate this simply to LL, or even L. Because both the poly(A) signal and the downstream antisense must leave the polymerase before stem-loop formation can occur, any effect of the antisense on poly(A)-dependent termination through stem-loop formation must reflect a post-extrusion activity of the poly(A) signal.
Fig. 1B, lane 3 shows that the presence of the antisense prevents termination and restores the ability of polymerases to travel down the plasmid to the post-cassette where they once again give a strong run-on signal. Fig. 1C shows the averaged results of eleven separate experiments that demonstrate the robustness of the antisense effect. However, the antisense does not restore elongation completely to the level of the mutant (compare lanes 1 and 3). In part this may reflect the fact that during transcription the poly(A) signal precedes the antisense sequence out of the polymerase, thus providing a small window of time during which a rapidly acting poly(A) signal might function (35). In addition, as illustrated by the converging arrows in construct 3 of Fig. 1A, the antisense (leftward arrow) targets only the core of the SV40 late poly(A) signal (rightward arrow), whereas the complete poly(A) signal (gray box) includes both upstream and downstream enhancers (36, 37). These enhancers are not blocked by the antisense and may enhance the ability of factors to compete with the antisense for association with the poly(A) signal.
We decided to repeat the above experiments using a poly(A) signal that lacks enhancer sequences. We have previously observed that the SV40 early poly(A) signal drives poly(A)-dependent termination even when truncated to its minimal core (2). Therefore, we repeated the above experiments using an analogous set of constructs in which P, the truncated SV40 early poly(A) signal, replaced L (SV40 late). Fig. 1D, lanes 1 and 2 confirm that P drives termination, although somewhat less effectively than L in these constructs (compare lanes 2 in Fig. 1, C and D). Importantly, an antisense to P blocked completely the ability of P to induce termination (Fig. 1D, lane 3). This shows that the cis-antisense effect is both general and, in the absence of poly(A) signal enhancing functions, complete. Therefore, because poly(A) signaling can be blocked even after the poly(A) signal has been fully extruded, we conclude (subject to the controls described below) that signaling is a post-extrusion event.
The above conclusion depends on the assumption that inhibition by the antisense element depends on duplex formation in the RNA, and therefore on sequence complementarity between the antisense and the poly(A) signal. We have shown previously that such duplex formation occurs rapidly during transcription in a nuclear extract (7), and that cis-antisense inhibition of cleavage and polyadenylation in vivo does indeed depend on sequence complementarity between the antisense and the poly(A) signal (35). To confirm that cis-antisense inhibition of termination also depends on sequence complementarity we swapped antisense elements. Lane 4 of Fig. 1D shows that the antisense to L has little effect on the ability of P to drive termination (compare lanes 2 and 4). Conversely, Fig. 1E shows that the antisense to P has little effect on the ability of L to drive termination (compare lanes 2 and 3). The slight residual effect that may be present for the swapped antisense elements in panels D and E of Fig. 1 may reflect the fact that different poly(A) signal cores, though not identical, necessarily resemble each other. These data show that antisense sequences inhibit only termination driven by the poly(A) signals to which they are complementary. This strongly suggests that the mechanism of antisense inhibition depends on true sense-antisense duplex formation.
Having established that sequence complementarity is essential to the antisense mechanism, it then became necessary to show that the expected stem-loop per se is not responsible for the adverse effects on termination. For example, termination may be signaled prior to extrusion but then blocked nonspecifically by secondary structure formation in the RNA following extrusion. One way to test this is to place an additional poly(A) signal in front of a poly(A) signal-antisense stem-loop and then ask if termination is restored despite the continued presence of the stem-loop. To carry out this control we first modified the three constructs of Fig. 1A by placing a BglII site upstream of L to give the first three constructs of Fig. 2A. This allowed us to place a second L upstream of the potential stem-loop in construct 3 of Fig. 2A to give construct 4. Constructs 1, 2, and 3 served as controls to indicate that the insertion of the BglII site did not itself have any effect in our assay. Accordingly, Fig. 2B shows, in agreement with Fig. 1C, that termination occurs (lanes 1 and 2) and is largely blocked by antisense (lane 3) in these constructs. However, inserting the additional L upstream reversed the effect of the antisense (lane 4), thereby demonstrating that the continued presence of the stem-loop did not block termination. From these experiments we conclude that the antisense blocks termination, not merely because it forms a secondary structure, but because of a direct effect on the poly(A) signal. Therefore, these controls validate the overall conclusion that poly(A) signaling is a post-extrusion event that can be blocked by antisense after the poly(A) signal leaves the polymerase.
|
The cis-Antisense Element Acts Truly in cis—Having determined that the trigger to terminate requires an extruded poly(A) signal (an accessibility requirement) we next wished to ask whether this triggering event is temporally coupled to the extrusion event (a timing issue). If the zero-order interaction of the poly(A) signal with factors riding on the polymerase (28, 29) provides the trigger to terminate, then the polymerase would quickly become immune to antisense inhibition after poly(A) signal extrusion. This can be tested by placing the antisense at increasing distances downstream from the poly(A) signal to determine at what distance the antisense loses its ability to block termination.
However, to use cis-antisense inhibition to investigate the coupling between extrusion of the poly(A) signal and other activities, it is imperative that cis-antisense inhibition itself be tightly coupled to extrusion of the antisense. For experiments in vitro this has been experimentally confirmed (7), but for experiments in vivo the possibility remains that at steady state some part of the inhibition occurs in trans. For example, viral nuclear RNA levels can be down-regulated by antisense RNA expressed from a plasmid in trans (38), and a variety of nuclear regulatory phenomena that resemble RNAi accompany the transcription of double-stranded RNA or RNA stem-loops in the nucleus (39–41). Here we test for the possibility of a trans-antisense effect, and rule it out, showing that inactivation of the poly(A) signal by antisense does indeed occur in cis and must therefore be coupled to the extrusion of the antisense.
We showed in lane 4 of Fig. 2B that adding a second poly(A) signal (L) to the sense-antisense pair, L, to give LL, rescued termination. Which L in LL drives the rescued termination? According to a cis-inhibition mechanism, it would be the upstream L, because the downstream L is sequestered in a hairpin by the adjacent antisense. According to a trans-inhibition mechanism, both L sequences would be expected to contribute similarly to the observed rescue because adjacency to the antisense would be irrelevant (inhibition being mediated by previously transcribed antisense sequences accumulated in the nucleus). To determine which L is active we introduced hexamer mutations into each L of LL individually. For cis-inhibition, all-or-none results are expected. Mutation of the downstream L should have no effect because the downstream L is already inactivated by antisense, whereas mutation of the upstream L should abolish all termination activity. However, for trans-inhibition, mutation of either L should have a partial effect. The results of Fig. 2C match the predictions for cis-inhibition. Mutation of the second L (Fig. 2C, lane 1) gave a result similar to that of unmutated LL (Fig. 2B, lane 4), whereas mutation of the first L (Fig. 2C, lane 2) gave a result similar to that of L (Fig. 2B, lane 3), which entirely lacks an upstream L. We conclude that antisense inhibition in these constructs is a cis effect, an effect, therefore, that is coupled to the extrusion of the antisense from the polymerase.
The Signal to Terminate Can Still be Blocked Long After Extrusion of the Poly(A) Signal—Knowing that inhibition by a cis-antisense element is genuinely a cis and not a trans effect, allowed us to return to the question of whether signaling the polymerase to terminate is coupled temporally to extrusion of the poly(A) signal. The experimental rational was to grant the poly(A) signal a limited window of time during which signaling to terminate might occur, and then to block further poly(A) signal function with antisense. This was accomplished by introducing spacer segments of DNA between the poly(A) signal and its antisense so that the poly(A) signal would remain uninhibited during the time it takes for the polymerase to reach and transcribe the antisense segment (35).
To prepare these constructs 56 or 141 bp spacer segments were inserted at the HindIII site of construct 3 in Fig. 1A between the poly(A) signal and its inverted repeat. Because the sense and antisense elements in construct 3 are already separated by 27 nt, this increased the separation to a final distance of 83 and 168 nt, respectively. As controls the same spacer segments were inserted into the HindIII site of construct 2 of Fig. 1A, which has an active poly(A) signal but no antisense. Fig. 3A shows that the antisense was still able to substantially inhibit the ability of the poly(A) signal to drive termination even when separated from the poly(A) signal by 83 or 168 nt (compare lanes 1 and 3 with lanes 2 and 4). We conclude that a significant fraction of the polymerases have not yet committed to termination even after 168 bp of transcription beyond the poly(A) signal. This suggests that the trigger to terminate is not temporally linked to poly(A) signal extrusion.
|
It is apparent that as the separation between sense and antisense increases, the efficiency with which the antisense blocks termination decreases (compare lanes 2 and 4 in Fig. 3A with lanes 3 in Figs. 1C and 2B). One possible explanation is that this reflects the rate of signaling by the poly(A) signal and that a significant proportion of the polymerases have been successfully signaled to terminate by the time they have traveled 168 bp past the poly(A) signal. However, our previous work with cis-antisense inhibition (35) suggests a different possibility. In those experiments we found, for antisense inhibition of cleavage and polyadenylation, that a similar distance-dependent rescue from antisense (Fig. 8A of Ref. 35) did not correspond to completion of processing. Instead, it reflected a gradual assembly of the cleavage and polyadenylation apparatus, which protected the poly(A) signal from the antisense in advance of processing itself (Fig. 9C of Ref. 35). By using a much longer antisense sequence we were able to overcome this protection of the poly(A) signal and extend the cis-antisense inhibition out to much longer sense-antisense separations (Fig. 8B of Ref. 35). We therefore decided to repeat the termination experiments described above using a longer antisense sequence. This would enable us to determine whether the diminishing effectiveness of the antisense in blocking termination reflects protection of the poly(A) signal from the antisense or commitment to termination.
The constructs used for this next part of the study are diagrammed in Fig. 3B. The SV40 late poly(A) signal, including the enhancers, is represented, as before, by a gray box. The black rectangles inside the boxes represent the poly(A) signal core elements that are targeted by the short (51 nt) antisense in the constructs discussed above (Figs. 1, 2, and 3A). The leftward arrows depict the much longer antisense sequences, targeting the entire SV40 late poly(A) signal (Fig. 3B, even-numbered constructs), used in the experiments to be described below. For example, the antisense in construct 2 targets the entire 158 nt SV40 late poly(A) signal (including enhancers) to yield a predicted 160 bp stem (that includes 2 bp of flanking sequence) and a 55-nt loop. Lanes 1 and 2 of Fig. 3C show that the longer antisense sequence in construct 2 (Fig. 3B) does, indeed, appear to be more effective at inhibiting termination than the shorter 51 nt antisense of previous figures (compare with lanes 2 and 3 of Figs. 1C and 2B). Note that this is so, despite the fact that the polymerase must travel considerably farther on construct 2 of Fig. 3B to fully transcribe the longer antisense (the distance separating sense and antisense is twice as much, and the antisense itself is three times as long).
The above results for constructs 1 and 2 of Fig. 3, B and C, show that the longer antisense effectively blocks termination despite the fact that the polymerase does not complete transcription of the antisense until 215 bp downstream of the 3'-end of the 3' enhancer of L. This result supports the inference from Fig. 3A that the trigger to terminate is not temporally linked to extrusion of the poly(A) signal. Moreover, with this longer, more effective antisense it was now possible to explore the effectiveness of cis-antisense inhibition out to much greater distances beyond the poly(A) signal. Accordingly we made constructs 4, 6, and 8 of Fig. 3B in which spacer sequences of increasing length were inserted between the poly(A) signal and its antisense. As controls for each antisense-containing construct the same spacer sequences were inserted into the non-antisense-containing parents. As an additional control for construct 8, the antisense sequence was replaced by an unrelated sequence of similar length to give construct 9.
Lanes 4, 6, and 8 of Fig. 3C show that cis-antisense inhibition of poly(A)-dependent termination persists even as the antisense is moved as far as 442 bp downstream. Therefore, there are polymerases that still require poly(A) signal input to terminate, even 442 bp downstream of the poly(A) signal. Moreover, 442 bp is merely the distance separating the sense from the antisense sequences. The functional separation of the sense from the antisense is actually larger than this. We do not know how far into the antisense sequence the polymerase must travel to produce a length of antisense sufficient to be effective. However, in our experiments the effective length must certainly be greater than 51 nt (see above) or 73 nt (see Fig. 8A of Ref. 35) because antisense sequences of these lengths are not effective over separations of several hundred bp. Thus, it is apparent that the polymerases whose termination can still be blocked by the antisense of construct 8 (Fig. 3, B and C) have traveled more than 500 bp (e.g. 442 + 73 = 515) downstream of the distal edge of the downstream enhancer of the poly(A) signal without yet having received the necessary input from the poly(A) signal to terminate transcription.
The Signal to Terminate Is Not Temporally Coupled to Extrusion of the Poly(A) Signal but Instead Is Generated Progressively over Time—As discussed in a previous section it is essential to the interpretation of these experiments that the antisense inhibition genuinely reflect a cis-interaction of the poly(A) signal with the downstream antisense. Although the data in Fig. 2C ruled out any significant effect in trans for the short antisense sequence used in those experiments, it remained possible that the longer antisense sequence used for Fig. 3C may trigger a trans effect. This could account for the unusual effectiveness of inhibition by the antisense even at considerable distances downstream. Fortunately, antisense inhibition via trans mechanisms makes different predictions for what happens in the region between the poly(A) signal and the downstream antisense than does inhibition in cis, so inhibition via cis and trans mechanisms can be distinguished experimentally.
We have shown previously that poly(A)-dependent termination occurs stochastically with respect to distance downstream of the poly(A) signal, exhibiting a simple first order decay in polymerase density over large distances (3). Thus, when the window is far downstream of the poly(A) signal (e.g. construct 9 in Fig. 3B), a considerable amount of termination must occur in the region between the poly(A) signal and the cassette window (but in terms of percentage, termination across the window itself should remain constant, see Ref. 3). If inhibition acts in cis then one would predict that an antisense sequence placed adjacent to a downstream window (e.g. Fig. 3B, construct 8) should have no effect on this upstream pre-window termination because the antisense has not yet been transcribed. On the other hand, to the extent that inhibition acts in trans, the inhibition of poly(A) signaling should begin as soon as the poly(A) signal has been transcribed regardless of the location of the antisense, because trans inhibition would be mediated (directly or indirectly) by antisense sequences in other RNA molecules produced in earlier rounds of transcription. Indeed, one would expect inhibition by a trans mechanism to be most effective close to the poly(A) signal, before assembly of the cleavage and polyadenylation apparatus begins to protect the poly(A) signal from the antisense (see above). Thus, a cis mechanism predicts no inhibition upstream but substantial inhibition downstream of the location of the antisense sequence, whereas a trans mechanism predicts equal or better inhibition close to the poly(A) signal (i.e. upstream of the antisense) than farther downstream (e.g. following the antisense).
To test these predictions it was necessary to monitor polymerase density both upstream and downstream of the antisense sequence, especially for constructs in which the antisense is located far downstream of the poly(A) signal. For technical reasons the constructs in Fig. 3B were not ideally suited to such an analysis (see legend to Fig. 4) so new constructs, illustrated in Fig. 4A, were prepared. In these constructs G-less cassettes allowed monitoring of polymerase density between the poly(A) signal and the antisense, whereas additional cassettes farther downstream allowed for monitoring of polymerase density as far as 2.7 kb beyond the antisense.
|
First we wished to confirm that the effect of antisense on termination for these new constructs was the same as for the constructs in Fig. 3B. Recall that the elongation efficiency downstream of the poly(A) signal, for all of the constructs discussed previously (Figs. 1, 2, 3), was estimated by reference to a window bounded on the upstream side by a 377-bp cassette and on the downstream side by a 261-bp cassette. For the new constructs (Fig. 4A) we defined the reference window as the region bounded by the 120-bp cassette on the upstream side and the 377-bp cassette on the downstream side. The post/pre ratio for the window so defined is thus the polymerase density in the 377-bp cassette divided by the polymerase density in the 120-bp cassette. Fig. 4B shows, just as did Fig. 3C, that a long antisense sequence placed as far as 442 bp downstream of the poly(A) signal can inhibit poly(A)-dependent termination in the cassette window (Fig. 4B, compare lanes 3 and 4). Note that different constructs with completely unrelated DNA in their cassette windows (bacterial cat DNA in Fig. 3 versus chicken C9 DNA in Fig. 4) and with the cassettes themselves in a different order (377 first in one case, last in the other) yield the same results. Thus the antisense inhibition is not construct specific.
We next asked whether the antisense inhibition documented in Fig. 4B affects termination immediately downstream of the poly(A) signal, consistent with a trans mechanism, or whether it takes effect only downstream of the antisense sequence as required by a cis mechanism. Fig. 4C is an example of a G-less cassette run-on transcription experiment involving constructs 3 and 4. Construct 4 differs from construct 3 only in that construct 4 contains the antisense sequence between the 174- and 120-bp cassettes (see Fig. 4A). From Fig. 4C it can be seen, using the 5'-most cassette (the 131-bp cassette) as a point of reference, that the polymerase densities within the 174-bp cassettes, upstream of the antisense, are essentially the same for the two constructs. In contrast, far downstream, there is little or no signal for the 261- and 377-bp cassettes of construct 3 because of termination, whereas significant signal remains for these cassettes in the case of construct 4 because termination has been inhibited. This inhibition of termination for construct 4 is all the more significant because the length of the antisense itself places the 261- and 377-bp cassettes farther downstream in this construct than in construct 3. Such an increase in distance would normally lead to more, not less, termination (3). Thus, termination in construct 4 is blocked at downstream locations, whereas up to the position of the antisense, constructs 3 and 4 behave the same. This implies that termination in construct 4 is normal (i.e. like construct 3) until after the antisense has been transcribed, as predicted by the cis-inhibition mechanism.
Importantly, although the 261- and 377-nt cassettes give distinctly stronger signals for construct 4 than for construct 3 because termination has been blocked, the reverse is true for the 120-nt cassette, for which the signal is less from construct 4 than from construct 3 (Fig. 4C). This illustrates explicitly the occurrence of termination in construct 4 before, but not after, the antisense has been transcribed. The antisense lies immediately upstream of the 120-bp cassette in construct 4 thus increasing the distance between the 174- and 120-bp cassettes (Fig. 4A). This increase in distance would be expected to result in more termination across this region for construct 4 than for construct 3. Accordingly, the 120-bp cassette signal is reduced for construct 4 in Fig. 4C, indicating that termination occurred across the antisense sequence as it was being transcribed but before it could act as an antisense. Conversely, for polymerases that successfully traversed the antisense, further termination downstream was inhibited.
The results of Fig. 4C illustrate clearly that termination occurs on construct 4 prior to completion of transcription of the antisense, and that termination is then inhibited at points farther downstream. Constructs 1 and 2 gave similar results (data not shown). From these data we conclude that the antisense inhibition observed in these experiments is genuinely a cis-inhibition phenomenon and, therefore, that polymerases even beyond 500 bp downstream of the poly(A) signal still require input from the poly(A) signal to terminate (see above).
A quantitative representation of these results is given in Fig. 4D. Constructs 3 and 4 were run in parallel with their respective poly(A) signal mutants (not shown in Fig. 4C) and then, following the same logic as for constructs 1 and 2 in Fig. 1B, the signal for each cassette was normalized to the most upstream cassette in the plasmid (i.e. the 131-bp cassette for constructs 3 and 4). Finally, we expressed the normalized intensity of each cassette for the wild type as a percent of the normalized intensity for that same cassette in the mutant (akin to the normalized post/pre ratio of Fig. 1B). Thus, any value different from 100% represents an effect that is dependent on a functional (wild type) poly(A) signal.
Fig. 4D shows, as expected, that transcription termination downstream of L is very efficient for construct 3, which lacks any antisense (dotted line). By 3 kb downstream of the poly(A) site polymerase density on the wild type template has decreased to 6% of that for the mutant. A transient increase in polymerase density proximal to the poly(A) site due to poly(A)-dependent pausing (3) is also evident. Importantly, both construct 3 (without antisense) and construct 4 (with antisense) exhibited comparable levels of poly(A)-dependent pausing and poly(A)-dependent termination for the first few hundred bp downstream of the poly(A) site. Most significantly, both templates exhibited comparable poly(A) signal function until a point on the template corresponding roughly to the 3'-end of the antisense element in construct 4. Then, as predicted by the cis mechanism of antisense action, termination at points farther downstream was rapidly inhibited. Thus, the simplest interpretation for all of these results is that the poly(A) signal delivers its final instructions to the polymerase stochastically with respect to time and distance, and this communication is not temporally coupled to the appearance of the poly(A) signal during transcription.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The currently favored anti-terminator model for poly(A)-dependent termination, originally proposed by McCracken et al. (23) and recently extended by Calvo and Manley (24), suggests that the polymerase is triggered to become termination-prone when CPSF and CstF, riding with the polymerase, capture the emerging poly(A) signal. However, we have shown here that an active poly(A) signal continues to be required for termination long after it has been transcribed. Thus, CPSF and CstF, if they do capture the poly(A) signal as it is extruded, do not at the same time trigger a change in state of the polymerase sufficient to induce termination. Further input from the poly(A) signal is required downstream.
The downstream input from the poly(A) signal that ultimately suffices for termination is apparently stochastic. We have shown previously that termination itself is stochastic across the first kb or so following the poly(A) site (3), a result consistent with Fig. 4D above (dashed line). As envisioned by the CPSF-CstF capture model, such stochasticity would involve the stochastic release of polymerases following a zero order signaling event. However, as discussed above, our data rule out models in which the trigger to terminate is delivered during a zero order CPSF-CstF capture event, because an antisense to the poly(A) signal, placed far downstream, aborts the first order termination process (Fig. 4D, solid line). Thus, we are driven to the view that the final triggering event itself is stochastic. Apparently, continued access of the poly(A) signal to the polymerase for long distances downstream is required to provide the opportunity, which arises stochastically, to trigger the polymerase to release.
One model that can account for the stochasticity of the final trigger to terminate is that termination depends on the pseudo-first order recruitment of a pivotal factor. CstF itself fits the profile for such a factor. CstF binds both the poly(A) signal in the RNA (42) and the CTD of the polymerase (17) and is likely to be intimately involved in the termination mechanism (20). Moreover, at least in B cells (43, 44), recruitment of CstF can be rate-limiting for cleavage and polyadenylation. Most importantly, the particular effects of cellular CstF levels on processing suggest that recruitment of CstF to the processing apparatus occurs directly from a pool of free factors rather than in a separate step via the polymerase (see below). The stochastic recruitment of a key component such as CstF in vivo is consistent with the observation that assembly of the cleavage and polyadenylation apparatus in cultured mammalian cells follows first order kinetics (35).
Although it is commonly assumed that CstF, as part of a CPSF·CstF complex, is handed over from the polymerase to the extruding poly(A) signal (a zero order reaction), the B cell data referred to above (43, 44) are more consistent with recruitment of CstF to the processing apparatus directly from the cellular pool (pseudo-first order). Takagaki et al. (43, 44) manipulated CstF concentrations in vivo and monitored the effects on IgM heavy chain pre-mRNA processing. They observed that decreasing the CstF concentration caused a decrease in both the total amount of heavy chain pre-mRNA processing as well as in the relative amount of processing at the first of two poly(A) sites in tandem. Although a polymerase hand-over mechanism can explain the first effect (decreasing the CstF concentration decreases the proportion of transcription elongation complexes bearing CstF), it cannot explain why a change in the CstF concentration should alter the relative efficiency with which two poly(A) sites in tandem are processed. If the first of two sites has a characteristic probability (less than 1) of being processed by a CstF on a polymerase, this probability will not change simply by altering the number of polymerases carrying a CstF. On the other hand, if, as suggested by Takagaki et al. (43), recruitment to the cleavage and polyadenylation apparatus is directly from the cellular pool of CstF then the rate of cleavage and polyadenylation would be sensitive to the concentration of CstF in this pool. This would affect most directly the first poly(A) site to appear during transcription, and therefore its chance of being processed compared with any poly(A) sites that follow. Thus, the results of Takagaki et al. (43, 44) are consistent with pseudo-first order recruitment of CstF, but are difficult to reconcile with zero order hand-over.
Of course, it is possible to devise more complicated models that retain the zero order capture feature of a CPSF·CstF complex riding with the polymerase. However, to accommodate our data, the ultimate termination event would additionally require further communication of the poly(A) signal with the polymerase in a final stochastic phase.
It remains a formal possibility that the stochastic step leading to termination is poly(A) site cleavage itself. However, we consider this possibility unlikely because cleavage probably does not lie on the path to termination. Although it is believed by some to be "clear that transcript cleavage at the poly(A) site represents a key determinant in transcriptional termination" (1), we cannot agree. Poly(A) signal function has been recognized for some time to be complex and multifaceted (45), and we are aware of no data indicating that the particular aspect of function that drives termination is cleavage. Indeed, as pointed out previously (2, 46), the correlation between processing and termination is not completely general. Therefore, a role for cleavage in termination is not established, and for the reasons outlined in the introduction may even be unlikely.
Our finding that the poly(A) signal remains functionally engaged with the transcriptional apparatus, long after it has been extruded from the polymerase, is consistent with current notions of an mRNA factory in which components required for both processing and transcription function as constituents of a single large supramolecular complex (1, 23, 28, 29, 47). Nothing is known for mammals about what factors in this complex contribute to transcription termination except that termination may depend on the presence of the polymerase CTD (23). However, in yeast, several components of the cleavage and polyadenylation apparatus, with clear mammalian homologs, have been shown to be involved in termination, and some of these interact directly with the CTD (12, 18–20). Therefore, our results support the view that the signal to terminate is generated over time by the cleavage and polyadenylation apparatus as it matures while riding on the CTD of the polymerase.
|
FOOTNOTES |
|---|
To whom correspondence should be addressed. Tel.: 310-825-3767; Fax: 310-206-4038; E-mail: hgm{at}chem.ucla.edu.
1 The abbreviations used are: CPSF, cleavage-polyadenylation-specificity factor; CTD, C-terminal domain; PC4, positive cofactor 4; CstF, cleavage stimulation factor; L, SV40 late poly(A) signal; P, SV40 early poly(A) signal core; nt, nucleotide.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
