Department of Chemistry and Biochemistry,
University of California at Los Angeles, Los Angeles, California
90095-1569
Received 5 February 1999/Returned for modification 24 March
1999/Accepted 13 May 1999
 |
INTRODUCTION |
Compared to splicing, the
cleavage-polyadenylation process is expected to be fairly
straightforward. The basic reaction in vertebrates involves only one
site on the RNA, not three as for splicing, and the core apparatus is
comprised entirely of proteins
no small nuclear RNAs are required
(14, 65). Spliceosome assembly occurs naturally in discrete
steps (61) both in vivo (7) and in crude extracts
in vitro (22). In contrast, no natural assembly
intermediates have been observed for the cleavage-polyadenylation apparatus in vitro with crude extracts (38). Until now,
however, there has been no information on the nature of the process in vivo. We report here that the assembly of the cleavage-polyadenylation apparatus in vivo is more complex than expected, being a gradual multistep process. Furthermore, we show that the strong simian virus 40 (SV40) late poly(A) site assembles considerably more rapidly in vivo
than other sites known to be weaker.
The core poly(A) signal in vertebrates consists of two recognition
elements flanking a cleavage-polyadenylation site. Typically, an almost
invariant AAUAAA hexamer lies 20 to 50 nucleotides (nt) upstream of a more variable element rich in U or GU residues (11, 23, 48). Cleavage between these two elements is usually on the 3'
side of an A residue (11) and, in vitro, is mediated by a
large, multicomponent protein complex that can be separated into five
distinguishable factors (14, 65). Four of these factors are
essential for cleavage in vitro at all poly(A) sites so far examined.
They are the cleavage and polyadenylation specificity factor
(CPSF), which binds the AAUAAA motif; the cleavage
stimulation factor (CstF), which binds the downstream U-rich element;
and two additional cleavage factors (CF I and CF II) that are less well
characterized (14, 65). The fifth factor is the poly(A) polymerase, which in addition to its obvious role in the
polyadenylation reaction must be present in most cases for the cleavage
step as well (13, 64).
CPSF, CstF, CF I, and CF II cofractionate as a single large complex
upon gel filtration of crude extracts (62). This has led to
the suggestion that the 3'-end processing apparatus may assemble in
vivo in a single step (46). However, multistep assembly is
also an attractive possibility (27). New data indicate that assembly in some cases may even begin with the recruitment of CPSF and
CstF to the transcription complex during initiation and elongation,
with both factors then being snared by the poly(A) signal as it emerges
from the polymerase (17, 47). By this scenario, the
remaining factors would then be recruited to the growing apparatus on
the nascent transcript. This order of assembly is consistent with in
vitro studies using purified factors (65).
Once assembled, the 3'-end processing apparatus directs cleavage and
polyadenylation. In vivo, the efficiency with which different poly(A)
sites are processed varies considerably (10, 19, 21). The
physical basis for this variation in poly(A) site "strength" is not
known. In principle, a site could be strong because it directs faster
assembly or because it assembles a more stable apparatus. Because
poly(A) site strengths are correlated with the stabilities of the
3'-end processing complexes that form in vitro, it has been suggested
that poly(A) site strength is based on complex stability (65,
66). However, it is easier to envision a kinetic explanation for
poly(A) site strength when it is recognized that the complete cleavage
and polyadenylation reaction takes place in seconds in vivo
(6) while complex stabilities in vitro are measured in terms
of minutes or hours. Preliminary experiments reported below comparing
the rates of assembly of poly(A) sites of various strength are
consistent with the kinetic view.
The fully mature 3'-end processing complex probably sequesters
considerably more RNA than just the core poly(A) signal itself. There
are numerous examples of poly(A) signals for which the flanking RNA
immediately upstream (8, 10, 26, 33, 50, 53, 60) or
downstream (5, 12, 26) of the core poly(A) signal is
essential for full activity. In some cases, specific elements in these
flanking sequences serve as binding sites for additional trans-acting factors (5, 44, 50). In other cases,
the involvement of additional factors is not apparent though the
flanking RNA is known to be important (12, 25, 26, 32, 33).
For several poly(A) sites, direct interactions of the flanking RNA with
CPSF and CstF have been demonstrated (25, 50) or are likely
(26, 33). Thus, the emerging picture is of a core poly(A)
signal at the center of a larger poly(A) signal domain.
The various elements in this domain are functionally intolerant of
significant variations in position (2, 5, 9, 11, 24, 33,
59). For example, the U/GU-rich element occupies a well-defined
position relative to the AAUAAA hexamer and the site of
cleavage (11). Function is destroyed by separating these elements with inserted RNA but is restored by sequestering the extra
RNA in secondary structure so as to bring the elements together again
(2, 9). Similarly, elements in the upstream flanking RNA
lose function if separated from the rest of the domain by unstructured
RNA but regain function if the extra RNA is sequestered in secondary
structure (24, 33). The weakness of the interactions comprising the core cleavage-polyadenylation structure and the role of
cooperativity in holding them together have been emphasized previously
(65). The similar dependence of function on proximity for
the core and flanking elements suggests that both are part of a single,
integrated structure held together cooperatively by multiple weak
interactions. Presumably the weakness of the individual interactions
within the structure accounts for the sensitivity to distance
(34). Other elements, found further upstream (56)
and downstream (42, 43) of the poly(A) site, are less
distance dependent in their function and may belong to a distinct class
of polyadenylation enhancers.
In the work presented below, we explored the assembly and function of
the poly(A) signal domain in vivo by using cis-antisense sequences that antagonize poly(A) site utilization. By moving the
antagonistic sequence gradually downstream of the poly(A) site, we were
able to determine the distance downstream at which the antagonist lost
its effect and poly(A) site function returned. An extensive series of
controls established that the return of poly(A) site function, as the
antisense element was moved downstream, corresponded to the time
required for the RNA polymerase to reach and transcribe the antisense
sequence. Using this approach, we determined for the SV40 early poly(A)
site that assembly of the cleavage-polyadenylation apparatus in vivo
begins immediately upon transcription of the poly(A) signal, covers an
extensive region of RNA, and requires about 20 s to reach
completion. The kinetics of assembly and the progressive nature of the
process add to our understanding of the structure of the 3'-end
processing apparatus and raise interesting questions regarding the
meaning of poly(A) site strength in vivo and the mechanism by which
poly(A)-dependent termination may be signaled to the elongating polymerase.
| |
MATERIALS AND METHODS |
Expression assays.
Transfections for chloramphenicol
acetyltransferase (CAT) assays were performed by the calcium phosphate
method as described previously (67). The transfected DNA
included a cotransfected luciferase-expressing plasmid, pRSVfl, in
which the HindIII-SacI fragment of Promega's
pGem-luc replaces the HindIII-BalI
segment of pRSVcat. For each 100-mm-diameter plate, 3.3 µg of
pRSVfl was used together with a threefold molar amount of
experimental plasmid and carrier (pBR322) DNA to bring the total amount
of DNA to 20 µg. Cells were rinsed twice with cold phosphate-buffered
saline at 44 to 46 h after transfection, lysed in situ by 15 min
of incubation at room temperature with 900 µl of Promega's 1×
reporter lysis buffer, transferred to a screw-cap microcentrifuge tube,
freeze-thawed (using liquid nitrogen and room-temperature water,
respectively), and centrifuged. After 200 µl of the supernatant was
removed for the luciferase assay, the rest was heat inactivated, for
the CAT assay, at 65°C for 10 min (16), followed by a
5-min spin. The supernatant was stored at 
60°C. For the luciferase
assay, 20 µl of extract was added to 80 µl of 2× luciferase assay
buffer (0.2 M K2HPO4 [pH 7.8], 10 mM ATP, 2 mM dithiothreitol) and assayed in a luminometer (LUMAT model LB 9501)
that injects 100 µl of luciferin solution (0.1 M
K2HPO4 [pH 7.8], 1 mM dithiothreitol, 943 µM luciferin) into each reaction mixture and then captures the light
emitted over a 10-s period. All samples were assayed in duplicate, and
volumes of extract containing equivalent amounts of luciferase activity
were then assayed in duplicate for CAT activity (31).
Heat-inactivated extract was added to 0.67 mM (final concentration)
acetyl coenzyme A, 0.01 µCi of [14C]chloramphenicol (54 mCi/mmol), 24.8 µM unlabeled chloramphenicol, and enough 0.25 M Tris
(pH 8.0) for a final reaction volume of 120 µl. After 20 to 60 min at
37°C, the reaction was stopped by ethyl acetate extraction and the
products were fractionated by thin-layer chromatography and then
quantitated with a Molecular Dynamics PhosphorImager. Assay results
were normalized to those for a positive control transfected in parallel.
For more recent work we have used Lipofectamine (Gibco BRL) or FuGENE 6 (Boehringer Mannheim) for transfection and a dual luciferase system
(Promega) for expression, in accordance with the manufacturers'
instructions. Cells in six-well plates were transfected with a DNA
mixture containing 35 ng of pRSVfl and 50 times this molar amount
of the experimental plasmid. Cells were harvested by use of passive
lysis buffer (Promega) 48 to 50 h after transfection, and the
resulting lysates were cleared by microcentrifugation for 30 s at
4°C. All samples were assayed in duplicate, and the averaged
Renilla luciferase values were then normalized to the
averaged firefly values. A normalized average of duplicates was
considered a single experimental point for purposes of further analysis
and was additionally normalized to data for a positive control
transfected in parallel.
RNase protection assays.
Standard procedures were used for
the RNase protection assay (4). All probes were extracted
with phenol or TRIzol (Gibco BRL) and then gel purified before use.
Cells were transfected in 100-mm-diameter plates by using FuGENE 6. Cytoplasmic RNA was prepared by using RNeasy columns (Qiagen) in
accordance with the manufacturer's protocol. For nuclear RNA, the
nuclear pellet was washed once with RLN (Qiagen) and then processed as
recommended for obtaining total animal cell RNA. Target RNA was treated
with DNase I (Ambion) (20), coprecipitated with
~106 cpm of [
-32P]CTP-labeled probe,
and then resuspended and hybridized at 56°C. RNase digestion was
carried out with RNase T1 only. The final results were
quantitated, with adjustment for C content, by using a PhosphorImager
with ImageQuant software (Molecular Dynamics).
| |
RESULTS |
cis-antisense rescue, a method for measuring the rate
of commitment to cleavage and polyadenylation in vivo.
The
rationale for cis-antisense rescue is based on
cis-antisense-mediated inhibition of cleavage and
polyadenylation (28). A copy of the poly(A) signal to be
tested is inserted in the reverse orientation downstream of the parent
poly(A) site (Fig. 1A). Although the
poly(A) site is transcribed (Fig. 1B), before processing can occur the
antisense transcript from the inverted poly(A) signal blocks cleavage
and polyadenylation by forming a duplex with the transcribed poly(A)
site (Fig. 1C and D). However, if the inverted poly(A) signal is
separated from the authentic one by a sufficient length of spacer DNA
(Fig. 1E), the additional time required for the polymerase to reach the
inverted signal (Fig. 1F and G) allows processing (or commitment to
processing) at the authentic site to proceed (Fig. 1H) and the poly(A)
site is rescued. By measuring the degree of poly(A) site rescue as a
function of DNA spacer length, the rate at which the poly(A) site
becomes committed to cleavage and polyadenylation can be inferred from
the known rate of transcription (63).
View larger version (21K):
[in this window]
[in a new window]
|
FIG. 1.
Rationale for measuring the rate of commitment to
cleavage and polyadenylation in vivo by the cis-antisense
rescue assay. See the text for details.
|
|
The above rationale assumes the following: (i) that an antisense
sequence can target an upstream poly(A) site and block polyadenylation (cis-antisense inhibition), (ii) that moving the sense and
antisense sequences apart will relieve this antisense effect
(separation-dependent rescue) and (iii) that the rescue is attributable
to a simple kinetic competition between the rate of commitment to
cleavage and polyadenylation and the rate of transcription. Below we
present experimental support for these three assumptions.
cis-antisense inhibition of polyadenylation occurs and
is rescued by separation.
In this section we address the first two
assumptions mentioned above. For our initial experiments we employed
the expression vector pRSVcat, which contains the SV40 early
poly(A) signal (Fig. 2A). We made a PCR
copy of this poly(A) signal and used it, in the reverse orientation, to
replace the BamHI-ApaI fragment of pRSVcat
(Fig. 2B). In the resulting plasmid, pE3473,
the SV40 early poly(A) signal, E, is separated by 34 bp from its
antisense-coding sequence, , of 73 bp. To assess poly(A) site
function, we then measured CAT expression in
pE3473-transfected COS cells. Other things
being equal, gene expression is proportional to the amount of poly(A)
site activity (21, 53). Figure
3 shows, in support of the first
assumption (cis-antisense inhibition), that CAT expression for pE3473 (lane 2) is drastically reduced
compared to that of a control that simply lacks the
BamHI-ApaI segment of pRSVcat (lane 1). Note
that because the antisense sequence is located downstream of the 3' end
of the canonical pRSVcat mRNA, the only way that antisense activity
can block expression is by interfering with mRNA 3'-end formation.
View larger version (23K):
[in this window]
[in a new window]
|
FIG. 2.
Construction of pE3473. (A)
The 3' end of the CAT transcription unit of pRSVcat
(30). (B) The 3' end of the CAT transcription unit of
pE3473. To obtain
pE3473 the sequence bracketed in panel C was
amplified with the primers 5'-CGGGCCCAAATAAAGCAATAGCATCACAAA
and 5'-GGGATCCGGACAAACCACAACTAGAATGCAG. The PCR
product was cut with ApaI and BamHI and then
inserted by sticky-end ligation into BamHI- and
ApaI-digested pRSVcat. (C) Sequence of the 3' end of
pRSVcat. The regions targeted by antisense sequence in
pH73 and in the pE73 series of
constructs are enclosed in parentheses and brackets, respectively. The
sequence within the brackets actually encodes two interdigitated
poly(A) signals. The upstream and downstream motifs for the dominant
signal are shown in boldface and are used exclusively in vivo to direct
cleavage as shown by the arrow (15). The sense-antisense
separation for pE3473 and its derivatives is
arbitrarily defined as the number of base pairs separating the last C
of the poly(A) signal region bracketed in the sequence and the first G
of its inverted repeat downstream.
|
|
View larger version (25K):
[in this window]
[in a new window]
|
FIG. 3.
CAT assay of cis-antisense inhibition and
rescue for the SV40 early poly(A) site. Diagrammed on the right for
each construct are the poly(A) signal (solid arrowhead), the inverted
poly(A) signal (open arrowhead), and the separation between them, drawn
to scale.
|
|
Thus, a downstream antisense sequence can block processing, but is it
by an antisense mechanism? To test this, we asked first whether the
mere insertion of DNA into the BamHI site of pRSVcat (Fig. 2A) is deleterious to expression. Accordingly, we inserted various randomly chosen DNA sequences into pRSVcat to yield
constructs pCat1 to pCat4 (Table 1),
which we then assayed for the ability to direct CAT expression in
transfected COS cells. These constructs did not differ appreciably from
each other or from pRSVcat in their ability to express CAT (data
not shown). Since pRSVcat depends on the presence of the SV40 early
poly(A) site for significant levels of CAT expression (54),
we conclude that the mere insertion of DNA into the pRSVcat
BamHI site does not significantly affect cleavage and
polyadenylation at the SV40 early poly(A) site.
To verify that it is the downstream insert that is responsible for
extinguishing CAT expression in pE3473 (Fig.
3, lane 2), rather than second-site damage to the plasmid, we removed
the inverted poly(A) signal by BamHI-ApaI
digestion. Full potential to express CAT was restored (data not shown).
This showed that inhibition of CAT expression in
pE3473 is a specific effect of the
inverted poly(A) signal.
Finally, we sought to confirm explicitly that this is an effect
directed at the upstream poly(A) signal and that it is mediated by
sequence complementarity. To check the role of sequence
complementarity, we replaced the SV40 early poly(A) signal of
pE3473 with the chicken
H-globin poly(A) signal. The resulting construct,
pG90E73, recovered 91% ± 12% (mean ± standard deviation) of the original ability to express CAT,
demonstrating that the globin poly(A) signal does not serve as a
significant target for the inverted SV40 sequence, with which it
shares little sequence homology. To confirm that sequence
complementarity specifically to a poly(A) signal, rather than generally
to the 3' end of the message, is principally responsible for the
inhibition of CAT expression in pE3473, we
prepared a final control construct, pH73.
This construct is like pE3473 except that
the antisense sequence is directed to a region surrounding the
HpaI site just upstream of the poly(A) signal (Fig. 2C)
rather than to the poly(A) signal itself. After transfection of
pH73 into COS cells, we found that it
expressed substantial CAT activity (58% ± 22% of that of
pRSVcat). This demonstrates that the highly effective
inhibition of CAT activity by the inverted poly(A) signal in
pE3473 (13% ± 4% of that of pRSVcat)
(Table 2) is a consequence of targeting
the poly(A) signal specifically and is not simply a result of antisense
activity generally.
The controls described above showed that the inverted poly(A) signal in
pE3473 blocks CAT activity by means of
authentic cis-antisense inhibition, thereby satisfying the
first assumption on which the rationale for measurement of the rate
of commitment to cleavage and polyadenylation in vivo is
based. The second assumption requires that the antisense
inhibition be dependent on adequate proximity of the sense and
antisense sequences so that moving them apart will progressively reduce
the effect. For example, a nonstoichiometric, catalytic type of
antisense interference (49) would not give rise to
separation-dependent rescue. The data from lanes 3 and 4 of Fig. 3 show
that when the poly(A) signal and its antisense sequence are moved
apart, CAT expression is indeed rescued. This confirms the second
assumption on which the rationale is based.
Do cis-antisense inhibition and separation-dependent rescue
occur for poly(A) signals other than the SV40 early poly(A) signal? Looking first at cis-antisense inhibition, we repeated the
above-described control experiments with a series of antisense
constructs based on the synthetic poly(A) signal (SPA) of Levitt et al.
(39) together with a variation of this signal (SPV) bearing
53% sequence identity to the original SPA. Figure
4 summarizes these additional controls.
In the reference constructs (pS and pSv) (Fig. 4, lines 1 and 2) the SPA and SPV were placed so as to support expression of a
gene encoding Renilla luciferase. A control plasmid with no
inserted poly(A) site was used to set the baseline (p
S) (Fig. 4,
line 3). An inverted SPA downstream of the authentic SPA reduced luciferase expression to background levels (Fig. 4, line 4), in agreement with the results for pE3473 (Fig.
3, lane 2). Expression was largely rescued when the sequence
complementarity between sense and antisense elements was reduced to
59% (G-U pairs were counted as complementary) by replacing the
upstream SPA with the related SPV (Fig. 4, line 5). This confirms that
a high level of sequence complementarity and, presumably, duplex
formation are required for antisense-mediated inhibition and that the
antisense sequence is not a nonspecific "poison" to poly(A) site
function. Expression was also rescued by placement of additional
poly(A) sites downstream of a sense-antisense pair (Fig. 4, lines 6 and 7), showing that the mere presence of secondary structure in this region of the RNA, or pausing or termination of transcription, cannot
be responsible for the poor expression that characterizes our
cis-antisense constructs.
View larger version (15K):
[in this window]
[in a new window]
|
FIG. 4.
Dual luciferase assay of cis-antisense
inhibition and rescue for the synthetic poly(A) site (SPA). All
transfections and assays of the plasmids in lines 2 to 7 were performed
in parallel with those of the positive control, pS, of line 1. Construct pS, which contains no added poly(A) site, served as the
negative control. The background expression exhibited by pS
(mean ± standard deviation, 34% ± 5% of that of pS) was high,
due in part to cryptic poly(A) sites elsewhere in the plasmid
(52). The basic antisense construct,
pS2359, exhibited a level of expression
(30% ± 10% of that of pS) not significantly different from that for
pS, so the average of these two constructs was taken as the baseline
for calculating the expression levels of the rest. The expression
levels of the remaining constructs (lines 5 to 7) were taken to be that
which exceeds the pS:pS2359 average,
expressed as a percentage of the range between the negative (lines 3 and 4) and the positive (line 1 or 2) controls. Construct pS (line 1)
was used as the 100% control for the SPA-based plasmids (lines 4, 6, and 7), and construct pSv (line 2), which expressed 43%
more than pS, was used as the 100% control for the SPV-based plasmid
(line 5).
|
|
The control experiments described above confirm that the simplest
interpretation of the CAT rescue results in Fig. 3 is one based on the
rationale presented in Fig. 1. Sense-antisense duplex formation
prevents polyadenylation when a poly(A) signal is followed closely by
an inverted copy of itself. As a result, only small amounts of mature
mRNA reach the cytoplasm, and the level of expression is low (Fig. 3,
lane 2). However, polyadenylation is rescued if the inverted copy is
moved downstream, allowing the polyadenylation apparatus time to
assemble. A clear prediction of this scenario is that cells transfected
with cis-antisense-inhibited plasmids not only should lack
mRNA in the cytoplasm but also should accumulate uncleaved pre-mRNA in
their nuclei (15).
cis-antisense inhibition causes accumulation of
uncleaved nuclear pre-mRNA.
We used RNase protection to assay for
uncleaved pre-mRNA. Figure 5, lane 2n,
confirms that most nuclear SV40 early RNA produced by
pE3473 is uncleaved. Little of this RNA
reaches the cytoplasm (lane 2c). The small amount of uncleaved RNA that
does appear in the cytoplasm presumably results from cryptic poly(A)
site activity downstream (52). Lanes 3 to 6 of Fig. 5 show
the results of RNase protection assays for constructs in which the
poly(A) site was separated from its antisense sequence by increasing
lengths of spacer DNA. The results show that the proportion of
uncleaved RNA in both the nucleus and the cytoplasm decreases as the
spacing between sense and antisense sequences increases, as expected. Thus, at least for the SV40 early poly(A) site,
cis-antisense-mediated inhibition of polyadenylation appears
to operate as envisioned (Fig. 1), simply by occluding the upstream
poly(A) signal and allowing uncleaved pre-mRNA to accumulate. The
effect is relieved, and the poly(A) site is rescued, by increasing the
separation between sense and antisense sequences.
View larger version (51K):
[in this window]
[in a new window]
|
FIG. 5.
RNase protection assay of cis-antisense
inhibition and rescue for the SV40 early poly(A) site. The probe used
was a T7 RNA polymerase transcript of DraI-digested pCat1.
The RNA lengths given in the figure refer to T1 digestion
at susceptible G's. The probe lengths corresponding to uncleaved
pre-mRNA varied from 196 to 198 nt depending on the construct. For the
six numbered lanes, 24 to 31 µg of nuclear RNA (n) and 60 to 62 µg
of cytoplasmic RNA (c) were analyzed. For the mock lane, 30 µg was
used. The faint band at the cleaved position for RNA from
mock-transfected cells reflects the low level of SV40 early mRNA
produced by the COS cells themselves (29). After subtracting
the estimated cellular contribution of cleaved RNA, the molar
percentages of uncleaved nuclear RNA for lanes 2 to 6 were 88, 73, 49, 40, and 9, respectively.
|
|
To determine whether the situation is similar for the unrelated SPA,
RNase protection was used to assay nuclear RNA from cells transfected
with a series of SPA-containing constructs. When the SPA is followed
closely by an inverted SPA (Fig. 6A,
construct 2), the nuclei of transfected cells accumulate mostly
uncleaved RNA (Fig. 6B, lane 2). When the inverted SPA is moved
progressively downstream (Fig. 6A, constructs 3 to 5), cleavage at the
authentic SPA is progressively rescued (Fig. 6B, lanes 3 to 5). Thus,
the SPA exhibits antisense inhibition and separation-dependent rescue just as does the SV40 early poly(A) site.
View larger version (45K):
[in this window]
[in a new window]
|
FIG. 6.
Distance-dependent rescue occurs for downstream but not
upstream antisense sequences. (A) Diagrammatic representation, drawn to
scale, of the plasmids and probes used for panel B. The downward arrows
indicate the positions of poly(A) site cleavage. (B) RNase protection
assays of nuclear RNA. The amounts of RNA used were as follows: 4 to 15 µg for lanes 2 to 5, 20 µg for lane 1, and 40 µg for lanes 6 and
7. The molar percentages of uncleaved RNA for lanes 2 to 7 were 73, 31, 25, 19, 73, and 5, respectively. The S probe was a T3 RNA polymerase
transcript of a DraI-digested plasmid obtained by inserting
the PvuII fragment of pBluescript SK(+) into the
HpaI site of pS2359. The S' probe
was a T7 RNA polymerase transcript of a plasmid obtained by inserting
the ApaI fragment of p59465S' into the
SacI site of pBluescript SK(+). An NdeI site was
inserted into this plasmid by in vitro mutagenesis (QuikChange;
Stratagene) to allow production of a 541-nt truncated transcript. The
RNA lengths given in the figure refer to T1 digestion at
susceptible G's. The probe has a variable tendency to be cleaved at an
internal position downstream of the poly(A) signal to yield a ~210-nt
fragment. (C) Sequence of the SPA-anti-SPA region in
pS2359. The SPA and its antisense sequence
are capitalized, and the poly(A) hexamer and its antisense sequences
are shown in boldface type. The probable poly(A) cleavage site
(3) is indicated by an arrow.
|
|
cis-antisense rescue is attributable to 3'-end
processing events that occur before the polymerase transcribes the
antisense sequence.
In this section we address the third
assumption on which the cis-antisense rescue rationale is
based, namely, that rescue occurs because the time required for the
polymerase to reach the inverted poly(A) signal exceeds the time
required to commit to processing (Fig. 1). This assumption could fail
in either of two ways. On the one hand, separation-dependent rescue of
a poly(A) signal from cis-antisense inhibition could reflect
a nonspecific transcript packaging process which precedes the
recruitment of processing factors to the poly(A) signal. In this case,
the rate of rescue would reflect the rate of the nonspecific event, not the rate of commitment to cleavage and polyadenylation. On the other
hand, increasing the separation of the sense sequence from the
antisense sequence could conceivably lead to rescue because of the
increased time required for their mutual search through three-dimensional space.
To test the third assumption, we prepared construct 6 of Fig. 6A, in
which the antisense sequence lies upstream rather than downstream of
the poly(A) site. If separation-dependent rescue reflects a generic
transcript packaging process, or the time during which sense and
antisense sequences search for each other through three-dimensional
space, then construct 6 (Fig. 6A), whose sense and antisense sequences
are well separated from each other, should exhibit less inhibition of
processing than construct 5, in which the sense and antisense sequences
are closer together. The opposite is observed, however: the
cleavage-polyadenylation process for construct 6 is mostly inhibited,
whereas for construct 5 it is mostly rescued (Fig. 6B, lanes 5 and 6).
A comparison of the results for construct 6 with those for construct 2 is even more revealing. After adjusting for the C contents of the two
different probes, quantitation of the results reveals equal molar
proportions (73%) of uncleaved RNA for both constructs. Thus, even
when moved 465 nt upstream of the poly(A) site (construct 6), the
antisense sequence remains just as effective as when it is almost
immediately adjacent (construct 2). Clearly, for construct 6, the
antisense sequence is neither shielded from the poly(A) site by
packaging nor delayed in its assault by search time. Therefore, the
rescue of cleavage and polyadenylation observed for construct 5 is
attributable to the processing (or commitment to processing) that
occurs during the time it takes the polymerase to reach the antisense sequence.
Careful attention was paid to the design of the constructs used for
Fig. 6. First, the spacer inserts for constructs 3 to 5 were excerpts
of the same sequence used for construct 6. Second, this sequence,
shared by all of the constructs, is a G-free cassette (see below), a
relatively monotonous sequence whose transcript is likely to have
little or no secondary structure. Therefore, the differences between
construct 6 on the one hand and constructs 3 to 5 on the other are not
likely to reflect sequence-specific effects on RNA structure or
transcription time. These data support the conclusion that the
separation dependence of cis-antisense rescue is a measure
of the rate of commitment to cleavage and polyadenylation.
Efficiency of rescue varies with transcript abundance.
If the
efficiency of cis-antisense rescue is related to the
processing rate and not simply to the generic packaging properties of
the RNA surrounding the poly(A) site, then any change in the rate
of poly(A) site processing should alter the efficiency of cis-antisense rescue. For example, at any given separation
of sense sequence from antisense sequence, a decrease in the processing rate is predicted to result in less rescue because less cleavage and
polyadenylation will occur before the antisense sequence is transcribed. The efficiency of the cleavage-polyadenylation process is
known to be affected by processing factor abundance in vivo (14). We therefore reasoned that the rate at which
individual pre-mRNAs are processed would be lower, and rescue would be
less, for a replicating plasmid whose many transcripts must compete for
a fixed pool of processing factors. To test this, we replaced the Rous
sarcoma virus promoter of pS16859 (Fig. 6A,
construct 4) with the SV40 early promoter, which contains the SV40
origin of replication (18). This generated a replicating
version of the plasmid, which we called
pS16859ori. The RNase protection assay
results of Fig. 7 confirm that there is
less rescue (less processing) for the replicating plasmid (lane 2) than
for its nonreplicating parent (lane 1). Rescue was similarly impaired by plasmid copy number for an entirely unrelated poly(A) site (Fig. 7,
lanes 6 and 7). It is formally possible that the SV40 promoter, rather
than the origin of replication, in these plasmids is responsible for
the altered processing (17). However, the important point is
that the replicating and nonreplicating plasmids are identical for
several kilobases surrounding their poly(A) sites (e.g., more than 2 kb
upstream and 1.5 kb downstream for pS16859
and pS16859ori), thus confirming that even
remote changes that affect processing rates are reflected in the
efficiency of cis-antisense rescue.
View larger version (34K):
[in this window]
[in a new window]
|
FIG. 7.
Effect of plasmid multiplicity and poly(A) site strength
on rate of commitment to cleavage and polyadenylation. (A) RNase
protection assays of nuclear RNA. The probe lengths for uncleaved S, L,
and E RNAs are 227, 197, and 239 nt, respectively. The corresponding
cleaved lengths are 182, 146, and 140 nt. For ease of comparison, the
bands are arranged side by side in the figure. Lanes 1 and 4 are taken
from lanes 4 and 2, respectively, of Fig. 6B. Minor bands corresponding
to uncleaved RNA were produced in the assays of lanes 4, 6, and 7. These bands were included in the calculation of mole percent uncleaved
RNA but are not shown in the figure. The L plasmids are based on a
derivative of pCat1 in which the Renilla luciferase gene was
inserted as for pRSVrl and an EcoRV site was introduced
100 bp upstream of the HpaI site by an in vitro mutagenic
conversion of G to T. The EcoRV-HindIII
fragment of this plasmid was then replaced with the DNA fragment shown
in panel B by sticky-end ligation to give pL. The bracketed sequence in
panel B, extended with AGCT at the 5' end and with A at the 3' end, was
then synthesized, as was its complement, extended with AGCTT at its 5'
end. These two oligonucleotides were annealed and ligated into
HindIII-cut pL to give pL2751
(in which the insert is in the antisense orientation with a unique
HindIII site between the sense and antisense sequences).
The PCR fragment used for pS8359 was then
inserted into HindIII-cut
pL2751 to yield
pL9151. To replace the RSV promoter with the
SV40 origin, the latter was obtained as an
NdeI-BstBI fragment from pRL-SV40 (which had been
modified to contain an NdeI site by converting a C to T 11 bp upstream of the BglII site) and then used to replace the
NdeI-BstBI fragment of the appropriate RSV
plasmid. For completeness we point out that each ori plasmid also
possessed (for reasons unrelated to this work) a
SmaI-EcoNI G-free cassette fragment from
pORgf3·2 (67) inserted at the AatII site 1 to 2 kb downstream of the poly(A) site in a nonfunctional part of the
plasmid. Also the L ori plasmids contained a
HindIII-BamHI G-free cassette insert from
pORgf3·2 at their KpnI sites downstream of the
transcription unit, making them nearly identical to the S plasmids, all
of which already have this fragment at a similar position. For RNase
protection, the E and S probes of Fig. 5 and 6 were used. The L probe
was a T7 RNA polymerase transcript of SspI-digested pL.
Although this probe traverses the HindIII site into
which insertions were made for the derivatives of pL, most uncleaved
cellular RNA nevertheless protected the full length of the probe by
looping out the inserted sequences. A small amount of probe was
protected only up to the HindIII site. This band is not
shown in the figure but is included as uncleaved RNA in the mole
percent calculation. (B) Top strand of a
StuI-HindIII-trimmed SV40 late poly(A) site
PCR fragment. The bottom strand is 4 nt longer because of the
HindIII sticky end. The poly(A) hexamer is shown in
boldface, and the cleavage site (45) is indicated with an
arrow. The sequence corresponding to the 51-nt antisense target in the
RNA is enclosed in brackets. (C) L- duplex formation in vitro.
Linearized plasmid DNAs were transcribed with T7 RNA polymerase at
37°C for 1 h in a solution containing 6 mM MgCl2, 2 mM spermidine, 40 mM Tris (pH 7.9), and 10 mM dithiothreitol. Following
transcription, DNaseI was added to 20 U/ml, KCl was added to a final
concentration of 0.1 M, and RNase A and RNase T1 (RNase
Cocktail; Ambion) were added to final concentrations of 6 and 250 U/ml,
respectively. After digestion at 25°C for 30 min, the samples were
analyzed as for RNase protection. Similar results were obtained using
four times as much RNase Cocktail. The marker is a 56-nt piece of
G-free RNA prepared by T1 RNase digestion. The plasmid
templates were constructed as follows: pLT7, as for pL in panel A
except with insertion into HpaI- and
HindIII-cut pAP<cat> (67);
pL2751T7, as for
pL2751 in panel A except starting with pLT7;
and pL16851T7, as for
pL9151ori in panel A except that the PCR
fragment used for pS16859 was inserted into
HindIII-cut pL2751 by
sticky-end ligation. The transcripts had the following lengths and were
produced from templates linearized at the following restriction sites:
lane 1, 1,384 nt, DraI; lane 2, 1,440 nt, DraI;
lane 3, 808 nt, HindIII; lane 4, 2,299 nt,
EcoNI; and lane 5, 943 nt, DraI. The central
region surrounding L and is shown drawn to scale in the figure. In
all cases, the T7 promoter (not shown) was 731 nt upstream of L.
|
|
A strong poly(A) site commits to processing faster than
weaker poly(A) sites.
It is thought that the strength of a
poly(A) signal reflects the stability of its processing complex
(65, 66). However, the data in Fig. 5 and 6 suggest that
after the poly(A) signal is transcribed, commitment to cleavage and
polyadenylation occurs during the time it takes the polymerase to
travel the next few hundred base pairs. Since this time period is very
short compared to the off rates measured in vitro (66), it
is unlikely that complex stabilities measured in vitro are relevant to
poly(A) site strengths expressed in vivo. It is more likely that, in
vivo, kinetic parameters are determinative of poly(A) site strength. To
determine whether poly(A) site strength is correlated with the rate of
processing in vivo, we carried out some preliminary comparisons of
processing rates for the SV40 late, the SV40 early, and the synthetic
poly(A) sites (abbreviated L, E, and S, respectively, in our plasmid
nomenclature). Carswell and Alwine have shown that L is five- to
sixfold stronger than E in vivo (10), and we have found,
in similar experiments, that E and S are of equivalent strength
(data not shown).
Figure 7 shows, for a replicating plasmid, that S is slow to act, with
most of its nuclear RNA remaining vulnerable to an antisense sequence
located 168 nt downstream, which sequesters it and prevents cleavage
(lane 2). In contrast, L acts rapidly and rescues itself almost
completely from an antisense sequence which is only 91 nt downstream
(lane 3). Similar, though less dramatic, results were obtained in a
comparison of L to both S and E in a nonreplicating background (lanes 4 to 6). It is not surprising that L enjoys a greater advantage when
factors are limiting (i.e., for a replicating plasmid). These results
show that L, one of the strongest poly(A) sites known, commits to
processing considerably faster than the weaker poly(A) sites, S
and E.
Graveley et al. have shown that RNA secondary structure is an important
determinant of poly(A) site strength in vitro (32). To
determine whether the ability of L to rescue itself so rapidly from
antisense attack is attributable to rapid folding into a secondary
structure that resists duplex formation with the antisense sequence, we
carried out nuclease digestions in vitro. RNA produced by T7 RNA
polymerase was RNase digested in the presence of 0.1 M KCl and 6 mM
Mg2+ immediately following transcription (Fig. 7C). Lane 1 of Fig. 7C shows that only a collection of short RNAs survived RNase
digestion of transcripts that contained L but no antisense sequence.
Lane 2 shows that transcripts of a plasmid with the inverted repeat 27 nt downstream of L yielded a prominent RNase-resistant band close to
the length expected for RNase digestion of an L- hairpin (53 to 54 nt). This band was still present for a plasmid in which had been
moved 141 bp farther downstream (lane 4). If either plasmid was cut
between L and before transcription, no protected band appeared
(lanes 3 and 5), confirming that was responsible for the RNase
resistance. These results indicate that the SV40 late poly(A) site is
not intrinsically resistant to formation of a duplex with an antisense sequence.
Commitment to cleavage and polyadenylation is a multistep
process.
To estimate quantitatively the rate of commitment to
cleavage and polyadenylation for the SV40 early poly(A) site, the
spacer constructs of Fig. 5, and several additional constructs in the same series, were transfected into COS cells and assayed for CAT expression as described for Fig. 3. After normalization for
transfection efficiency and adjustment for the actual fraction of
cytoplasmic RNA polyadenylated at the SV40 early poly(A) site, the
resulting CAT activity values were plotted in Fig.
8A as percentages of the activity of the
positive control (i.e., percent rescue) versus the distance separating
sense sequence from antisense sequence.
View larger version (29K):
[in this window]
[in a new window]
|
FIG. 8.
Distance dependence of cis-antisense rescue
for short and long antisense sequences targeted to the SV40 early
poly(A) site. (A) Rescue from the 73-nt antisense element. The CAT
expression data in Table 2 were corrected for the presence of extended
mRNAs polyadenylated at cryptic sites downstream (estimated from the
cytoplasmic lanes of Fig. 5). The corrected data were then plotted and
fitted to the following version of the first-order rate equation:
(c0 c)/c0 = A[1 e k(s + x)], where (c0 c)/c0 is the fraction of poly(A) sites rescued as
the polymerase travels a distance s + x down the
template, c0 is the starting number of
unprocessed poly(A) sites, c is the number of unprocessed
poly(A) sites remaining after the polymerase has traveled a distance
s + x, s is the separation between sense and antisense
sequences as defined in the legend to Fig. 2C, x is a
correction factor to be evaluated by curve fitting, and A is
the maximum fraction of CAT activity that can be rescued in our
experimental system. Notice that s + x is used here as
a proxy for time t, which can be determined by dividing
s + x by 40 nt/s, the known rate of RNA polymerase II
elongation in mammalian cells (63). The curve in the figure
returns the following values for the variables floated during the fit:
A = 81%, k = 4.9 × 103
bp1, and x = 18 bp. The inset
summarizes the RNase protection data of Fig. 5 fit to the same
first-order equation. The symbols for the sequence types inserted into
the BamHI site of pE3473 are as
follows:  , no inserted sequence;  , 50 bp forward;  , 50 bp
reverse;  , G free; and  , chicken globin. (B) Rescue from the
1,387-nt antisense element, with separation defined exactly as for
pE3473. Data were obtained and corrected as
for panel A, but curve fitting was by eye.
|
|
We wanted to be able to convert the separation distance into time based
on the average rate of transcription (63). We therefore needed to ensure that the DNA spacers used were representative of
average DNA sequences. In addition, at the RNA level, we needed to
ensure that the spacers did not introduce unanticipated biases due to
RNA folding. Accordingly, we chose carefully the types of DNA to be
used as spacers in the construction of the various plasmids employed in
the experiment shown in Fig. 8A. In one approach to generating spacers,
we inserted tandem multiples of a 50-bp sequence to ensure that long
spacers were simply longer and did not include new sequence variables.
We made plasmids containing both forward and reverse orientations of
the basic sequence, thus obtaining two spacer series with what are
essentially different sequences. A second approach to minimizing the
effects of sequence made use of spacers whose DNA sequence is
relatively monotonous and whose transcripts are expected to be devoid
of secondary structure. The inserts for these plasmids were all derived
from the original G-free cassette of Sawadogo and Roeder
(58). In the G-free orientation, these cassettes yield
transcripts devoid of G's, for which MulFold secondary-structure
prediction (37) indicates that no significant secondary
structure should exist. In addition to the carefully chosen sequences
just described, we also included a plasmid whose spacer was taken from
the 3'-flanking region of the chicken H-globin gene.
The CAT rescue data in Fig. 8A show that as the antisense sequence is
moved progressively downstream from the poly(A) signal, the CAT
activities of the various constructs steadily recover from inhibition
and are 50% rescued by about 200 bp. The relationship between
separation and CAT rescue fits first-order kinetics, as shown by the
curve in the figure (see the legend to Fig. 8). Candidate rate-limiting first-order (or pseudo-first-order) processes include factor binding and RNA folding steps that would sequester the poly(A)
signal and make it inaccessible to the antisense sequence. The curve in
Fig. 8A shows that all of the plasmid constructs used yielded data that
scatter about the same first-order curve. Indeed, each of the three
plasmid series (50-bp forward, 50-bp reverse, and G free) gave points
that scattered on both sides of the curve. We therefore conclude that
the results of this cis-antisense rescue assay are not
significantly compromised by sequence effects and that they provide a
reliable measure of commitment to 3'-end processing as a function of
transcription time [assuming there is no poly(A)-dependent
change in the transcription rate].
The asymptote for the curve in Fig. 8A is less than 100%, suggesting
that complete rescue is not achieved. Perhaps this reflects a residual
trans-antisense effect from neighboring transcripts. For the
large sense-antisense separations at the asymptote, the antisense
segment of each transcript has no cis target because the
poly(A) site has already been sequestered by factors or processed (Fig.
1H). However, before the antisense segment is degraded, perhaps it can
assault nearby nascent transcripts in trans as previously
described by Liu et al. (40). Such a trans effect would presumably be minimal at short separations in which the antisense
sequence is engaged in the cis interaction (Fig. 1C and D).
The rate of poly(A) site rescue (Fig. 8A) was analyzed not only by CAT
expression but, as shown in the inset to Fig. 8A, also by RNase
protection, using the data of Fig. 5. The percentage of cleaved
pre-mRNA in the nucleus was determined and then analyzed by plotting
and curve fitting as for the main part of Fig. 8A. The curve in the
inset confirms that cleavage is 50% rescued by about 200 bp for the
SV40 early poly(A) site. Thus, there is good agreement between
conclusions based on measurements of cytoplasmic mRNA expression
relative to a control (Fig. 8A) and measurements of cleaved-RNA levels
in the nucleus as a percentage of the total (Fig. 8A, inset). Note,
however, that the latter approach is affected by the unknown half-lives
of the cleaved and uncleaved RNAs in the nucleus and by the unknown
extent to which the cis-antisense sequences in the uncleaved
RNA compete with the radioactive probe during hybridization.
During the course of these studies, when different lengths of antisense
sequence were used, we noticed that the poly(A) site required more time
to become immune to a longer antisense sequence than to a shorter one.
Figure 8B shows, for example, that when a very long antisense sequence
is used (1,387 nt), 50% rescue is not attained until the antisense
sequence is moved 550 bp downstream of the sense element. Moreover,
there is a distinct lag of about 300 bp before rescue itself actually
begins, indicating a requirement for one or more preparatory steps
prior to the event that finally achieves rescue. The simplest
explanation for the antisense length-dependent lag is that the 3'-end
processing complex is assembled progressively and that early stages of
assembly are sufficient to protect against the shorter antisense
sequences but only a more mature complex can resist the disruptive
influence of a long antisense sequence.
The above explanation assumes that the rescue following the lag in Fig.
8B results from events taking place at the SV40 early poly(A) site.
Although it is formally possible that transcription termination between
sense and antisense sequences contributes to the rescue, this is not
likely because poly(A) sites do not drive termination in the plasmid
background we have used (67). We should also point out that
although the 1,387-nt antisense sequence happens to contain the SV40
late poly(A) site (10), the presence of the complementary
sense sequence upstream ensures that this downstream site never becomes
accessible for processing (Fig. 6B, lane 6). We have used RNase
protection assays to confirm this and to further establish that rescue
is due exclusively to polyadenylation at the early site (data not shown).
To estimate the amount of time required for complete assembly of the
SV40 early 3'-end processing complex, we prepared additional constructs
with various separations and with antisense sequences targeted to
either 51 or 136 nt surrounding the poly(A) site (Fig. 9A, constructs 1 and 3). The rescue data
for these constructs, together with the data of Fig. 8 (i.e., for
constructs in series 2 and 4 of Fig. 9A), provided us with the
relationship shown in Fig. 9B, in which the length of the antisense
target on the RNA is plotted against the lag preceding rescue. Because
the lag is difficult to estimate for the shorter antisense sequences,
we have, for the purposes of Fig. 9B, operationally defined the lag as
the separation required for 25% rescue. Thus, in Fig. 8B, for example,
435 bp are required for sufficient poly(A) site maturation to occur so
that rescue can begin and proceed to the extent of 25%. Figure 9B
shows that the lag increases with antisense target size until a maximum
target of about 200 nt is reached, for which a lag corresponding to
about 400 bp of transcription is required to reach 25% rescue. From
this we infer an assembly time of about 10 s for the first poly(A)
sites to become completely resistant to antisense sequences and an
assembly domain of about 200 nt for the SV40 early poly(A) site.
Because of the asynchrony of the process, it takes a further 10 s
(i.e., 400 bp more) for all rescue to reach completion (Fig. 8B).
View larger version (28K):
[in this window]
[in a new window]
|
FIG. 9.
Progressive assembly of the cleavage-polyadenylation
processing complex. (A) The antisense targets on pRSVcat of the
four E series of plasmids used to obtain the data for panel B, as
well as the targets of three control plasmids. For the control
plasmids, the sense-antisense separation (172, 34, and 33 nt for
plasmids 5 to 7, respectively) is defined as the number of base pairs
separating the last C of the poly(A) signal region bracketed in the
sequence of Fig. 2C and the beginning of the downstream antisense
sequence. This definition maintains the poly(A) signal as the reference
point even though the antisense sequences are targeted to non-poly(A)
signal regions of the transcript. This is because rescue from a
downstream antisense element for these constructs cannot begin until
the tether between the transcript and its antisense sequence is broken
by cleavage and polyadenylation. (B) The separation of sense from
antisense sequences (the lag) required to achieve 25% rescue from
antisense elements of increasing length. The circles represent
measurements made on the pRSVcat-based antisense constructs
indicated. The triangle is an estimated value based on the
relationships between the pE51rl and the
pE73rl classes of constructs, which are based on
pRSVrl. The E51rl measurements could not be used
directly, we discovered, because polyadenylation rates differ
significantly for mRNAs containing different coding
sequences. (C) Model depicting progressive assembly of the
cleavage-polyadenylation apparatus. Short antisense elements (e.g.,
51) interfere only with early stages of assembly. Longer
antisense elements (e.g., 136) can interfere also with
partially assembled complexes, but mature complexes are immune to
antisense sequences. Although assembly is clearly a multistep process,
we show here only a single step.
|
|
Since increasing the antisense target size beyond 200 nt requires
little additional assembly time to effect rescue (Fig. 9B), it appears
that most of the upstream target of the long antisense sequence in the
E1387 series of constructs lies outside of the assembly
domain of the 3'-end processing complex. Results obtained using clones
5 to 7 of Fig. 9A are consistent with the notion that the poly(A) site
assembly domain does not cover much more than 200 nt. Clone 5 (B1254) expresses CAT at 42% ± 10% of control levels
for a separation of 172 bp. This is over 10-fold more expression than
for the clone 4 series (E1387) at a comparable
separation (<4% expression compared to controls [Fig. 8B]). Yet the
antisense sequence of clone 5 targets 90% of the same sequence as does
the antisense sequence in the clones of series 4 (Fig. 9A). Thus, clone
5 lacks antisense sequence to 133 nt surrounding the poly(A) site, and
the remaining 1,254 nt of antisense sequence are largely ineffective at
inhibiting expression. Clone 6 of Fig. 9A,
H73, is targeted to a region at the edge but
still within the 200-nt assembly domain of the poly(A) site. As may be
expected, and as described at the beginning of Results, it has a mild
inhibitory effect on polyadenylation. In contrast, the much longer
antisense sequence in clone 7 (U517), which is targeted
to an upstream portion of the transcript, has no effect on expression
(100% ± 25% of control). These results strengthen the interpretation
that Fig. 9B identifies an assembly domain surrounding the poly(A) site
that encompasses about 200 nt and has an assembly time of about 10 s (Fig. 9C).
| |
DISCUSSION |
cis-antisense rescuemeasuring the kinetics of
commitment to 3'-end processing in vivo.
We have described a
method for using cis-antisense sequences to interrogate
poly(A) signals following their synthesis in vivo. By progressively
moving an antisense element downstream from its target poly(A) site in
a series of constructs, we provided measured increases in potential
assembly time to the poly(A) signal before the inhibitory antisense
sequence was transcribed. When the time provided in this way
became equal to the time required for commitment to cleavage and
polyadenylation, we witnessed restored poly(A) site function (rescue).
We studied the SV40 early poly(A) site and an artificial poly(A) site,
the SPA (39), using this assay. Numerous controls were
carried out to confirm that a simple antisense mechanism was involved
in inhibition and that relief of inhibition upon moving the antisense
sequence downstream reflected simply the travel time required for the
RNA polymerase to reach the antisense element. Thus, inhibition
depended on the presence of adequate sequence complementarity between
the antisense element and the upstream poly(A) signal to which it was
targeted; inhibition gave rise to large amounts of uncleaved pre-mRNA
in the nuclei of transfected cells, and inhibition was relieved only by
moving the antisense sequence downstream of the poly(A) site so that
its appearance was transcriptionally delayedan antisense moved
upstream was just as inhibitory as one immediately adjacent to the
poly(A) site.
Cleavage-polyadenylation complex assembly and folding.
One of
the principal conclusions from this study derives from the observation
that the time required for poly(A) site assembly and rescue (for
example, the lag evident in Fig. 8B) increased with the size of the
antisense target region containing the poly(A) site (Fig. 9B). This
indicates that the poly(A) site is progressively assembled into
structures that are resistant to inhibition by ever-more-potent (i.e.,
longer) segments of antisense sequence. A model which can explain this
result is presented in Fig. 9C. Our data do not suggest the nature of
the progressive steps in maturation, but for the sake of discussion, we
consider a scenario involving stepwise addition of factors to the
growing complex (27). As shown in Fig. 9C (for simplicity,
only a single intermediate step is illustrated), a short antisense
sequence could inhibit processing by forming a duplex with the naked
poly(A) signal, thereby interfering with the first step of factor
binding. However, if the factors initiate binding before the
poly(A) signal encounters the antisense element, the bound
factors will rescue the poly(A) site from the short antisense
sequence. Nevertheless, the partially assembled poly(A) site
remains susceptible to inhibition by longer antisense sequences. Thus,
a short antisense element is effective only if transcribed very soon
after the poly(A) signal itself, before factor binding gets
seriously under way. A longer antisense sequence, on the other hand,
remains effective even when its appearance is delayed somewhat by the
time required for transcription, because it can bind to and, according
to this scenario, inactivate partially assembled complexes. A fully
mature complex is resistant to antisense sequence of any length,
possibly because cleavage immediately ensues, severing the
cis relationship between the sense and the antisense
sequences. Other scenarios are also possible, especially the
simultaneous binding to the RNA of all factors in the form of a
holocomplex followed by an ordered series of conformational transitions
that increase the footprint of the complex on the RNA. Note, however,
that the progressively greater effectiveness of longer and longer
antisense sequences cannot be explained as trivial kinetic or stability
effects related to length since even a short antisense element (59 nt)
is capable of acting with undiminished effectiveness across at least
465 nt of upstream RNA (Fig. 6B, lane 6).
Assuming that the progressivity of complex maturation reflects stepwise
factor binding, CPSF and CstF would presumably be the first to bind,
perhaps by direct transfer from the polymerase (14, 17, 47,
65). Our data are consistent with an early occurrence of the
initiating event, as implied by such a scenario. Lane 2 of Fig. 6B
(same as Fig. 7A, lane 4) shows that rescue of 3'-end cleavage from the
relatively short (59-nt) anti-SPA sequence is already under way by the
time the antisense element, only 23 nt downstream, is extruded
from the polymerase. Rescue is even faster (i.e., more cleavage) for
the strong SV40 late poly(A) site (Fig. 7A, lane 6). Thus, the
poly(A) signal begins recruiting factors while it is still in the
vicinity of the transcription complex. Indeed, the converse may occur:
it may be the polymerase-bound CPSF and CstF that recruit the poly(A)
signal (35), with subsequent assembly steps then taking
place in association with the transcriptional apparatus.
Figure 9C, which presents a physical interpretation of the present
work, also accommodates our functional understanding of the 3'-end
processing domain (see the introduction). Thus, RNA sequences
immediately flanking the core poly(A) signal are important for function
(5, 8, 10, 12, 26, 33, 50, 53, 60), and in at least one in
vitro study, the presence of flanking RNA per se was important,
regardless of the sequence (36). Also, all elements in the
poly(A) signal domain (upstream flanking, AAUAAA hexamer,
cleavage site, U/GU rich, and downstream flanking) occupy narrowly
prescribed positions relative to each other (2, 5, 9, 11, 24, 33,
59), suggesting that they, and the factors that bind to them,
comprise a single, well-defined structure. The fact that elements in
the flanking RNA, when examined, are found directly associated
with CPSF or CstF (25, 26, 33, 50) supports this
interpretation. We suggest that it is this structure whose assembly we
have observed in vivo and which is shown schematically in Fig. 9C.
Our estimate of 10 s for the minimum time to maturation of an SV40
early 3'-end processing complex (and 20 s for the completion of
all processing) is consistent with previous estimates of the time
required for cleavage and polyadenylation (1, 6, 51, 57). In
the early work, time was measured directly, using pulse-labeling, and
rough estimates of approximately 1 min for the completion of both
cleavage and polyadenylation were obtained (1, 51, 57).
Very recently, Baurén et al. (6) used reverse
transcription-PCR to determine that cleavage of nascent Balbiani
ring 1 transcripts in Chironomus occurs about 600 bp
downstream of the poly(A) site. This result is remarkably similar to
our results showing that assembly of the cleavage-polyadenylation
complex is complete at about the same position on the template (Fig.
8B). Although it is unlikely that cleavage occurs at the same
distance downstream for all poly(A) sites, this coincidence does
strengthen the possibility that fully mature complexes, as defined by
cis-antisense rescue, proceed immediately to cleavage.
Does assembly rate govern poly(A) site strength?
Assembly of
the cleavage-polyadenylation apparatus is a multistep process that
takes a significant length of time. One may therefore expect
strong poly(A) sites to be those that can assemble quickly in vivo so as to minimize the opportunity for
interference with the process. The results of Fig. 7 show that the rate
of processing complex assembly in vivo is indeed correlated with poly(A) site strength, with the strong SV40 late poly(A) site committing to cleavage more rapidly than the weaker synthetic or SV40
early poly(A) sites placed in comparable plasmid backgrounds (compare
lane 3 with lane 2 and lane 6 with lanes 4 and 5). Moreover, the
cis-antisense rescue assay itself illustrates that rapid
assembly leads to poly(A) site strength. The synthetic poly(A) site,
with an antisense sequence located 168 nt downstream, behaves like an
extremely weak poly(A) site (Fig. 7, lane 2) because it is unable
to process an appreciable fraction of the pre-mRNA molecules before the
downstream antisense sequence interferes. In contrast, the SV40 late
site processes very quickly and is scarcely interfered with at all by
an antisense sequence located a similar distance downstream (Fig. 7,
lane 3). [The SV40 late antisense sequence is effective, however, if
located sufficiently close to the poly(A) site (Fig. 7, lane 7).]
Thus, at least with these experimental constructs, the strength of a
poly(A) site is a direct consequence of its rate of assembly.
The realizations that weak sites assemble slowly and that the assembly
process can be interfered with at any point in the pathway have
regulatory implications. A particularly intriguing possibility in this
regard concerns poly(A) sites that are subject to downstream
regulation. These sites delay the decision to process until they
encounter elements that may lie from several hundred base pairs
(43) to more than 1.5 kb (3) downstream. Perhaps such sites have strategic impediments to assembly that maintain them in
a partially assembled but arrested state while awaiting regulatory input.
The gradual nature of processing complex assembly may also be relevant
to the mechanism by which the poly(A) signal instructs RNA
polymerase II to terminate. Current models invoke events at the two
extremes of the assembly continuum, either the initial binding of
factors (47) or the final cleavage of the RNA (15, 55). To these we can now add the intermediate steps of the
assembly process as possible triggers of termination, particularly if
assembly occurs astride the polymerase.
L.C.C. and A.J. contributed equally to this work.
We thank D. H. Liu for assistance with plasmid construction
and E. Landaw for advice on data analysis.
![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
Top
Abstract
Introduction
Present address: Directorate of Research, University of
Agriculture, Faisalabad, Pakistan.
