ABSTRACT
A RNA fragment which is protected from degradation by ribosome pausing at a stop
codon has been identified in growing
Escherichia coli
. The fragment is 261 nt long and corresponds to the 3
'
-end of the mRNA expressed from a semi-synthetic model gene. The 5
'
-end of the RNA fragment, denoted rpRNA (ribosomal pause RNA), is located
13 bases upstream of the stop codon.
In vivo
decay of the complete mRNA and accumulation of rpRNA are dependent on the
nature of the stop codon and its codon context. The data indicate that the
rpRNA fragment arises from interrupted decay of the S3A'
mRNA in the 5' ->
3'
direction, in connection with a ribosomal pause at the stop codon. RF-2 decoding of UGA is less efficient than RF-1 decoding of UAG in identical codon contexts, as judged from rpRNA
steady-state levels. The half-life of UGA-containing rpRNAs is at least 5 min, indicating that ribosomal
pausing can be a major factor in stabilising downstream regions of messenger
RNAs.
The stop codons UAA (ochre), UAG (amber) and UGA (opal) are protein synthesis
termination signals of mRNA reading frames (
1
,
2
). The UAA and UAG codons are recognised by release factor one (RF-1) and the UAA and UGA codons by release factor two (RF-2).
mRNA levels are highly dependent on translation, since stop codons in coding
regions often induce transcription polarity (
3
). mRNA that escapes premature termination of transcription is degraded by endo- and/or exonucleolytic RNases (
4
). A protein complex containing both ribonuclease E (RNase E) and polynucleotide
phosphorylase (PNPase) has been characterised in
Escherichia coli
. This complex can processively degrade mRNAs, after an initial RNase E
endonucleolytic cleavage, which occurs frequently in the 5'-part of messages (
5
,
6
). Interestingly, the formation of such complexes and their activity seems to be
dependent on both 5'-phosphorylation and 3'-polyadenylylation (
7
,
8
).
Stop codons that are placed early in coding regions have been shown to influence
mRNA stability, in contrast to stop codons which are introduced further
downstream (
9
-
11
). For instance, a study of
bla
transcripts in
E.coli
showed that UAA codons at positions 4 and 26 decrease mRNA stability, whereas
stop codons introduced at positions 56 (UAG), 103 (UAG) and 192 (UAA) had no
influence (
10
). However, recent reports show that there are exceptions to this rule, since
UAA at codon position 15 in the mRNA for ribosomal protein S20 does not
influence mRNA stability (
12
). Furthermore, UAA at codon position 10 in
lacZ
mRNA increases the stability of the entire transcript (
13
). This effect probably results from ribosomal pausing at the UAA stop codon,
which protects the mRNA from RNase E cleavage at its 5'-end (
14
).
Ribosomal pausing has been shown to be important for expression of several genes
coding for resistance to antibiotics that affect translation (
15
,
16
). The resistance of
Bascillus subtilis
to erythromycin is caused by elevation of
ermC
gene expression. The antibiotic induces a ribosomal pause in translation of the
ermC
leader peptide, which stabilises the downstream part of the mRNA, which encodes
the
ermC
gene methylase (
16
).
In this article we describe an mRNA degradation intermediate (rpRNA; ribosomal pause RNA) of S3A' mRNAs with an internal stop codon (
17
). rpRNA seems to be the result of interrupted mRNA degradation on the 5'-side of the ribosome, i.e. pausing at the stop codon.
[[alpha]-
32
P]ATP was purchased from Amersham and restriction enzymes, T4 polynucleotide
kinase and T4 DNA ligase from Promega. The DNA Sequenase kit was from United
States Biochemical Corporation. Deoxyoligonucleotides were made on a Gene
Assembler Plus from Pharmacia.
LB and M9 based media were formulated according to Miller (
18
). The M9 medium was supplemented with glucose, thiamine and all amino acids at
recommended concentrations (
19
). Ampicillin was used at 100 [mu]g/ml in plates and 200 [mu]g/ml in liquid medium. The
E.coli
strain MC1061 was used in cloning experiments (
20
). Translation assays were in XAc [wild-type,
trpT
+
(Su
-
), with respect to tRNA genes] and CDJ64 [containing the
trpT
(Su9) suppressor gene] (
21
). The
E.coli
strain UY211, a rifampicin-sensitive derivative of XAc, was used for mRNA half-life determinations (
22
).
Escherichia coli
strains with the
prfB2
mutation are P1 transduction derivatives of XAc, containing the
zgc21
::Tn
10
marker with either the
prfB2
mutation (UY2687) or the wild-type RF-2 (
prfB
+
) allele (UY2688).
Escherichia coli
strain YN3231 was the source of the
prfB2
mutation and the
zgc21
::Tn
10
marker (
23
).
New codon context variants (pAB15, pMF18, pAB29 and pAB52) of the S3A' gene were made as previously described (Fig.
1
) (
17
). The A' domain has earlier been referred to as Z or B' (
24
-
26
).
Escherichia coli
cultures (20 ml) were inoculated using the M9-based medium described above. Cells were harvested in the mid log growth
phase, in SS-34 centrifugation tubes containing crushed ice and chloramphenicol (200 [mu]g/ml). Total RNA was prepared using the hot phenol method (
27
). Chloramphenicol was used, although later control experiments indicated that
it gave neither improvement of the RNA preparation nor a change in the obtained
results (data not shown). Rather, it is the rapid cooling on ice that gives
good quality RNA. Samples containing 5 [mu]g total RNA were first fractionated on 5 or 6% Sanger gels in a minigel
system at 15 mA/gel for 30 min (Midget system; Pharmacia) and then blotted
overnight in a Midget minigel blotting unit at 20 V. The
32
P-labelled oligonucleotide probe ABS01 (5'-CGTTGTTCTTCGTTTAAGTTAGG-3'), complementary to a sequence within each A' domain of S3A' mRNA (
17
,
24
), was used in hybridisation (20 pmol/filter).
Primer extension analysis of S3A' mRNA was made using
32
P-labelled primer ABP02 (5'-CTTACTTAAGCTTGGCTGCAG-3'), which is complementary to a sequence at the 3'-end of S3A' mRNA. The ABP02 primer was
annealed to total
E.coli
RNA (2.5 [mu]g) overnight at 30oC. Extensions were made using 50 U M-MuLV reverse transcriptase and reaction products were analysed on
a 5% Sanger gel. Products from DNA sequencing reactions primed with the same
primer were used as size markers.
mRNA half-life was measured using a Northern blotting technique. A 250 ml culture of
UY841 (UY211/pAB7) or UY931 (UY211/pAB11) was grown to the desired cell density
(OD
540
0.5) and rifampicin was added (600 [mu]g/ml) to stop transcription initiation. Samples were taken at regular
intervals and RNA was prepared, blotted and probed as described above for
Northern blots. The amount of probe bound to the filter was monitored with a
PhosphorImager (Molecular Dynamics Inc). RNA half-lives were calculated by least squares analysis of a semi-logarithmic plot of mRNA concentration as a function of time.
An artificial gene, S3A', has been used previously for characterisation of ribosomal read-through of nonsense codons in
E.coli
strains during balanced growth (Fig.
1
;
17
,
25
). The S3A' gene has three identical A' domain sequences, which are based on the antibody binding B
domain of the protein A gene from
Staphylococcus aureus
(
17
,
24
). In this report we analysed turnover of mRNA expressed from different S3A' alleles. These alleles have an internal test codon at position 162, in
the linker region between the second and third A' coding domains. The test codon was either a stop codon (UGA or UAG) or a
sense codon (UGG or UUA), with a variable codon context defined by the two
flanking codons at positions 161 and 163 (
17
). The tight U7 (AGC UGA UGU) and the leaky U4 (CCA UGA AGU) codon contexts were
characterised by 1 and 27% ribosomal read-through, respectively, in the tRNA suppressor-free
E.coli
strain XAc used here (
17
,
21
). We have reason to believe that the U4 context is exceptionally inefficient
with respect to RF-2 interaction (
28
). This increased the time available for tRNA selection, which resulted in the
observed high translation read-through of this UGA codon context, even in a strain lacking any opal tRNA
suppressor.
mRNA was isolated from exponentially growing XAc cells harbouring plasmids with
the different S3A' alleles and analysed by Northern blotting using a probe (ABS01)
complementary to a sequence within the three A' coding domains. The results are shown in Figure
2
. The S3A' mRNA is 787 nt in length, as indicated. The amount of mRNA associated
with the different alleles varied, even though equal amounts of total RNA were
applied in each sample. The observed differences in steady-state mRNA levels were in direct correlation with the protein expression
levels reported previously (
17
). This figure also shows the presence of an additional small RNA fragment which
was expressed to a varying degree by strains with some of the different UGA
context alleles of S3A'. This small RNA, denoted rpRNA for reasons which will be explained
below, was not observed when the S3A' gene had a sense codon as the test codon at position 162. Figure
2
also shows mRNA from S3A' alleles with the UAG contexts 122 or A24. These codon contexts are
identical to two of the UGA codon contexts used (Figs
1
and
2
). It can be seen that the amount of rpRNA was much lower in the case of UAG
than was found for the corresponding UGA codon contexts.
S3A' mRNAs with contexts U7 and pwtU7 were analysed by a primer extension
experiment (Fig.
3
). The U7 and pwtU7 contexts were used since the former, but not latter, is
associated with a significant amount of rpRNA. The primer used (ABP02) was
homologous to a sequence at the 3'-end of S3A' mRNA. Only one primer extension product was observed in the
case of pwtU7. This cDNA species corresponds to full-length S3A' mRNA, as determined from a gel that was run for a longer time
(data not shown). However, two cDNA products were observed in the primer
extension reaction from mRNA with the U7 context. The long cDNA, which was also
found in the pwtU7 mRNA-primed reaction, corresponds to full-length S3A' mRNA. The small cDNA species was only present in the U7 mRNA-primed reaction, indicating that it corresponds to the
rpRNA. The deduced 5'-end sequence of rpRNA is shown in Figure
3
. It can be seen that the 5'-end of rpRNA is located 13 nt upstream of the internal stop codon.
This finding indicates that the small RNA (rpRNA) is a mRNA decay intermediate
which accumulates as a result of a pausing ribosome at the stop codon. The size
of rpRNA, starting 13 nt upstream of UGA, is 261 nt, since its 3'-end should be identical to the 3'-end of full-length S3A' mRNA.
The fates of S3A' mRNA and the rpRNA were monitored after addition of rifampicin to
exponentially growing UY211 bacteria; this strain carries a plasmid with an S3A' allele (UGA codon context U7 or U4). Samples were taken at regular time
intervals and the RNA analysed by Northern blotting (Fig.
4
). The antibiotic acts immediately, as can be seen in the expression levels of
both species of S3A' mRNAs, which showed exponential decay from the very beginning of the
experiment. The initial steady-state level of the U7 S3A' mRNA was three times higher than the level of U4 S3A' mRNA. The amount of U7 S3A' mRNA decreased during the course of the experiment,
with a half-life of 1.5 min, whereas decay of the U4 S3A' mRNA was faster, with a half-life of 0.6 min. The difference in steady-state levels of U7 and U4 mRNAs was therefore probably
caused by different degradation rates.
Figure
5
shows a Northern blot analysis of S3A' mRNAs expressed in some
E.coli
strains. The UGA codon in the U7 context was decoded by a wild-type tRNA
Trp
at 1% efficiency in the suppressor-free XAc strain and at 6% efficiency in CDJ64 by a
trpT
(Su9) suppressor form of the same tRNA (
17
). It can be seen that the relative steady-state level of rpRNA was lower in the
trpT
(Su9) strain compared with strain XAc. Thus increased read-through of the stop codon by a suppressor tRNA caused less protection from
degradation of the fragment. On the other hand, the relative amount of rpRNA
was only slightly increased in a
prfB2
(RF-2) mutant strain as compared with its
prfB
+
counterpart (Fig.
5
).
The small RNA species described here is a degradation intermediate of the mRNA
expressed from the semi-synthetic model gene S3A' (Fig.
4
;
17
) and appears only when S3A' mRNA has an internal stop codon (Fig.
2
). The small RNA carries the UGA codon close to its 5'-end (Fig.
3
) and the relative level of this RNA species, compared with full-length mRNA, is decreased in a strain with the
trpT
(Su9) UGA suppressor tRNA (Fig.
5
). These findings suggest that the small RNA is formed in connection with
ribosomal pausing at UGA, because of slow binding of RF-2; this small RNA is therefore referred to as rpRNA (ribosomal pause RNA).
The UAG codon contexts give lower steady-state levels of rpRNA than the UGA contexts. This suggests less efficient
protection against RNA decay by ribosome pausing at UAG than at UGA. Therefore,
UAG decoding by RF-1 appears to be faster than UGA decoding by RF-2.
The 5'-end of rpRNA is located 13 bases upstream of the stop codon (Fig.
3
). In another study, using mRNA coding for bovine preprolactin in an
in vitro
translation system, it was deduced from nuclease treatment of the translation
reactions that ribosomes tend to pause at a UAA codon and cover 12-13 bases upstream of this stop codon (
29
). This finding using extracts from eukaryotic cells is in line with our results
in growing bacteria and supports the idea that rpRNA is a decay intermediate of
full-length S3A' mRNA, caused by ribosomal pausing at a stop codon, which delays
further 5' -> 3' degradation of the mRNA.
Our RNA decay study shows that the U4 mRNA decays two to three times faster than
the U7 context mRNA (Fig.
4
). This finding is correlated with a 3-fold lower steady-state level of the U4 mRNA species. The observed difference in
stability between U7 and U4 mRNAs is remarkable, in view of the fact that the
only differences between these two mRNAs are the codon on the 5'-side and the first base on the 3'-side of the internal UGA codon. We also find that the
kinetics of rpRNA production is less complex for U4 rpRNA than for U7 rpRNA.
The level of U4 rpRNA decreases exponentially during the sampling period. In
contrast, the U7 rpRNA level seems to decrease during the first minute, then
increase until most S3A' mRNA has been broken down, and finally decrease exponentially.
An initial lag period is observed in the functional decay of many mRNA species
in
E.coli
when rifampicin is used to stop transcription initiation (
30
-
32
). Addition of rifampicin stops the production of mRNA, causing increased levels
of free ribosomes and translation factors as mRNA decays. This leads to increased translation
initiation at mRNAs with under-saturated ribosome binding sites, since protein synthesis in
E.coli
is normally limited by the number of free ribosomes (
23
). It is possible that the difference in accumulation observed for rpRNAs with
the U7 or U4 codon contexts originates from such secondary effects after
rifampicin addition.
The half-life of UGA codon-containing rpRNAs is at least 5 min. This time corresponds to 0.003
decoding events/s, which is many orders of magnitude slower than the normal
rate of translation
in vivo
, ~4-22 events/s (
33
). Even if most natural stop codons are decoded faster by the release factors
than found for the UGA contexts analysed here, a considerable portion of
ribosomes should be trapped at stop codons in growing bacteria. Such temporary
inactivation of ribosomes by slow termination is unlikely, since protein
synthesis in
E.coli
seems to be limited by the number of free ribosomes (
34
). The efficiency of translation termination would in such a case have a strong
influence on gene expression and be a bottleneck for cellular growth. However,
we see no specific effects on cellular growth, induced by UGA codon context
versions of the S3A' gene used here (data not shown). Thus even if ribosomal pausing at the
stop codon is significant and could be of importance for gene expression, it is
expected to be considerably shorter than the estimated 5 min half-life of rpRNA.
A ribosomal pause at the stop codon would allow other ribosomes to initiate and
stack behind the first pausing ribosome. The result of this would be a ladder-like pattern of several fragments on the gels used for RNA separation,
arising from protection against degradation by two or more ribosomes. This is
not seen, and it is therefore an intriguing observation that rpRNA seems to
result from protection by pausing of a single ribosome at the stop codon. The
apparent precursor-product relationship between S3A' mRNA and rpRNA (Fig.
4
) indicates that rpRNA is produced by 5'-end-initiated decay of S3A' mRNA. A protein complex containing RNase E and PNPase
may, after an initial 5'-end cleavage, give a 5' -> 3' directed decay of many mRNAs (
5
,
6
). Cleavage by this, or by some other RNase complex, could trim the mRNA on the
5'-side of the last translating ribosome. This RNase activity might
track the mRNA until it reaches a pausing ribosome, whereby further decay will
be delayed. Alternatively, ribosome pausing at a stop codon could be a signal
to some RNase to bind and cleave the unprotected mRNA 13 nt upstream of the
stop codon. Both of the indicated models predict that rpRNA production is
directly correlated with the duration of ribosome pausing at the stop codon.
Thus fast termination at UAG should give a lower level of rpRNA compared with
slow termination at UGA, as indeed is observed. Release of the presumed RNase
and a successful termination event would give a free, relatively stable rpRNA,
with a half-life that is independent of stop codon efficiency, as is observed for U7
and U4 rpRNAs. These rpRNAs both have a half-life of ~5 min.
The relative level of rpRNA should be higher in a
prfB2
than in a
prfB
+
strain (Fig.
5
), because of much slower termination and a longer pause in the former case. We
observe only a minor increase in the
prfB2
strain. It appears that extended ribosomal pausing is ineffective in producing
more rpRNA, suggesting that the relevant RNA decay pathway is already saturated
at some step. Altogether, we favour a S3A' mRNA decay model with 5' -> 3' processivity, where decay can be temporarily halted
by ribosome pausing at stop codons. Complete degradation of the fragment
containing the stop codon (rpRNA) will then be achieved by other RNases after
release of the ribosome from the stop codon in connection with termination of
translation.
We thank Mitch Dushay and Steve Muir for comments on the manuscript, Margareta
Faxén for the pMF18 plasmid and Yoshikazu Nakamura for strain YN3231. This
work was supported by grants from the Swedish Natural Science Foundation (NFR).
REFERENCES
Return