ABSTRACT
Previous work from this laboratory [Dompenciel,R.E., Garnepudi,V.R. and
Schoenberg,D.R. (1995)
J. Biol. Chem.
270, 6108-6118] described the purification and properties of an estrogen-regulated endonuclease isolated from
Xenopus
liver polysomes that is involved in the destabilization of albumin mRNA. The present study mapped cleavages made by this
enzyme onto the secondary structure of the portion of albumin mRNA bearing the
major cleavage sites. The predominant cleavages occur in the overlapping
APyrUGA sequence AUUGACUGA present in a single-stranded loop region, and in AUUGA located within a bulged AU-rich stem. A structural mutation which converted the major loop
cleavage site to a hairpin bearing one APyrUGA element eliminated cleavage at
the intact site. This confirms that the polysomal RNase is specific for single-stranded RNA. Additional point mutations in the major loop characterized
the nucleoside sequence requirements for cleavage. Finally, snake venom
exonuclease was used to demonstrate the polysomal RNase generates products with
a 3
'
hydroxyl. Binding of an estrogen-induced protein to a portion of the 3
'
UTR of vitellogenin mRNA may be involved in its stabilization by estrogen
[Dodson,R.E. and Shapiro,D.J. (1994)
Mol. Cell. Biol.
14, 3130-3138]. The core binding site for this protein bears the sequence APyrUGA,
suggesting that stabilization may be accomplished by occlusion of a cleavage site for the polysomal RNase.
The digestion of a given mRNA
in vivo
by a ribonuclease may be thought of as either an initial rate-limiting step in the process of its turnover, or as the ultimate endpoint
of a series of molecular interactions that culminate in the degradation of the
molecule. The nucleolytic processes that have been implicated to date in the
turnover of eukaryotic mRNAs include almost every conceivable mechanism
(reviewed in
1
). 3'-5' Exonucleases appear to be important in the turnover of a wide
variety of polyadenylated mRNAs and the non-polyadenylated histone mRNA (
2
). In yeast, the process of mRNA destabilization begins with the loss of the 3' poly(A) tail, probably by the action of poly(A) nuclease (or PAN;
3
,
4
). Subsequent to poly(A) shortening, yeast mRNAs undergo decapping by a
pyrophosphatase followed by 5'-3' degradation by the exonuclease encoded by the
XRN1
gene (
5
). Parker and co-workers (
6
) have found that, at least in yeast, nonsense-mediated decay proceeds through decapping and 5'-3' degradation without prior deadenylation, indicating that
these are separable processes.
While neither PAN nor a decapping pyrophosphatase have been definitively
identified in higher eukaryotes, a number of unstable mammalian mRNAs which
have AU-rich elements in their 3' UTR (e.g. cytokines, oncogenes and transcription factors) also
undergo deadenylation prior to decay (see review by Ross,
7
). While there is no
in vivo
evidence for decapping and 5'-3' degradation as seen in yeast, a 5' exonuclease has been identified that is capable of
removing the cap and degrading the body of the mRNA (
8
). Endonucleolytic cleavage is an important, and likely rate-limiting, step in the decay of a number of vertebrate mRNAs. Examples of
mRNAs that appear to be degraded through endonucleolytic processes include
avian apo-very low density lipoprotein II (
9
), insulin-like growth factor II (IGF II) (
10
), maternal homeobox-containing mRNAs (
11
,
12
), the cytokine
gro
[alpha] (
13
) and transferrin receptor (
14
). In the case of the
Xenopus
maternal homeobox mRNAs, a ribonuclease activity was identified in oocyte
extracts that acts like an RNA restriction endonuclease, cleaving only at a
specific sequence element present within the 3' UTR (
12
). Structural features do not appear to be important for the specificity of this
enzyme, since a 19 base sequence is sufficient to impart sensitivity to
endonucleolytic cleavage to a heterologous mRNA. In contrast, IGF II mRNA is
highly structured, and it is likely that structural constraints play a major
role in targeting endonucleolytic cleavage to a specific site in the 3' UTR (
10
).
The highly-efficient synthesis of the yolk protein precursor vitellogenin in response
to estrogen stimulation of
Xenopus
liver results from both transcriptional and post-transcriptional processes. Estrogen, acting through its nuclear receptor,
causes the transcriptional induction of the four vitellogenin genes (
15
,
16
). In addition, vitellogenin mRNA is also stabilized by estrogen through an
estrogen receptor-dependent process (
17
). While these pathways lead to significant accumulation of vitellogenin mRNA,
one might predict that the very short (13 nucleotide) 5' UTR present on this mRNA would make it a poor competitor for translation
initiation in the presence of high levels of serum protein-coding mRNAs normally found in liver polysomes. However, such competition
does not occur
in vivo
because estrogen also causes the coordinate destabilization of the mRNAs
encoding the major serum proteins, including albumin, transferrin and the
fibrinogens (
18
-
20
). The process of coordinate mRNA destabilization is restricted to mRNAs for serum proteins; ferritin,
poly(A)-binding protein and actin mRNA levels are unaffected by estrogen.
In earlier studies we identified an estrogen-induced ribonuclease activity on
Xenopus
liver polysomes whose properties suggested it may be involved in the
destabilization of the serum protein-coding mRNAs (
21
,
22
). A hallmark of this activity is its selectivity; whereas it degrades albumin mRNA present either free in solution, as mRNP or on
polysomes, it does not degrade ferritin mRNA (
22
). Furthermore, this activity is a component of neither ribosomal subunit, but
requires both subunits be together in an 80S complex for its binding to
polysomes (
21
). The purified RNase consists of two isoforms of molecular mass 62 000 and 64
000 (
23
), which most likely result from the duplication of the
Xenopus
genome. The purified RNase has many of the same properties reported earlier for
the activity present in crude polysome extracts (
22
). These include specificity for albumin versus ferritin mRNA, resistance to
inhibition by EDTA or placental ribonuclease inhibitor, and a lack of
dependence on divalent cations for activity. Two dimensional gel
electrophoresis showed the RNase purified from liver of estrogen-treated
Xenopus
consists of three isoforms with isoelectric points of 9.6, 9.7 and 9.8. These
are likely the result of post-translational modification. While the significance of this modification is
unknown at present, we suspect these changes are involved in either the
activation of the RNase by estrogen or its ability to target to polysomes.
This putative messenger RNase was purified based on its ability to generate a
doublet cleavage product of ~194 nucleotides (nt) using as substrate a uniformly labeled 500 nt
transcript from the 5' end of albumin mRNA. The cleavage that generated this doublet was mapped
to the sequence AUUGACUGA, which can be thought of as two overlapping copies of
the pentamer APyrUGA. This element is present multiple copies in mRNAs that are
destabilized in response to estrogen (14 in albumin, nine in transferrin and
seven in [gamma]-fibrinogen), but it is absent in ferritin mRNA. Furthermore, this
sequence is not cleaved when the radiolabeled substrate transcript is converted
to a double-stranded RNA by hybridization to an excess of antisense transcript. These
results suggest that the structural context of the APyrUGA element may be
important for cleavage by the polysomal RNase. The purpose of the present study
was to solve the secondary structure surrounding the major cleavage site on
albumin mRNA, and use this information to determine the role both structure and
primary sequence play in cleavage by the polysomal RNase. We show here that the
polysomal RNase preferentially cleaves at APyrUGA located in a single-stranded region of a stem-loop structure, but can also cleave at adjacent sites. Furthermore,
collapsing this loop to a hairpin bearing a single APyrUGA element in double-stranded RNA blocked cleavage, confirming the importance of substrate
structure. In addition, we demonstrate that cleavage generates products with a
free 3' hydroxyl which may be susceptible to exonucleolytic degradation
in vivo
.
The purification of the polysomal RNase, preparation of plasmids, assay for
ribonuclease activity, preparation of transcripts and gel analyses were
performed as described by Dompenciel
et al
. (
23
). The handling of experimental animals and preparation of total RNA were
performed as described by Schoenberg
et al
. (
20
). The experiments reported herein were conducted according to the principles
set forth in the Guide for the Care and Use of Laboratory Animals, Institute of
Animal Resources, National Research Council, Department of Health and Human
Services Pub. No. (NIH) 78-23.
RNase T1 and A digestion of total liver RNA from male
Xenopus
was done under non-denaturing conditions in buffer containing 20 mM Tris-HCl, pH 8.0, 5 mM MgCl
2
and 50 mM KCl. RNase U2 digestion was performed in buffer containing 15 mM
sodium citrate, pH 3.6, 50 mM MgCl
2
and 50 mM KCl. Prior to enzymatic digestion 10 [mu]g samples of total liver RNA were heated in the above buffers at 50oC for 5 min and slowly cooled to 25oC to insure proper folding. Digestion was performed with 10 U/[mu]l RNase T1, 1 U/[mu]l RNase U2 and 0.1 ng/[mu]l RNase A at 25oC. Equal portions were removed after 5, 10 and 15
min of digestion. The reactions were stopped by addition of 9 vol 0.3 M sodium
acetate, pH 5.0, 10 mM EDTA and extraction with an equal volume of phenol:HCCl
3
:isoamyl alcohol (25:24:1). RNA was recovered by ethanol precipitation and
analyzed by primer extension.
Primer extension was performed as described previously (
24
). Ten [mu]g total liver RNA or 20 ng
in vitro
transcript (plus 10 [mu]g yeast tRNA) were ethanol precipitated with 1-2 * 10
5
d.p.m. of 5' end-labeled primer. The primer used for secondary structure analysis in
Figure
1
consisted of the sequence CACTCAGGAGTTTTGTCATTAA (DAN25), which is
complementary to nucleotides 280-301 of the 74 kDa albumin mRNA. The precipitated RNA and primer were
dissolved in 10 [mu]l annealing buffer (50 mM Tris-HCl, pH 8.7, 0.54 M KCl, 1 mM EDTA), heated at 65oC for 10 min and slowly cooled to 25oC. To each tube was added a mixture of 0.9 mM dATP, dCTP,
dGTP, TTP, 50 mM Tris-HCl, pH 8.3, 13 mM MgCl
2
, 7 mM dithiothreitol and 200 U M-MLV reverse transcriptase to a total volume of 40 [mu]l. The reaction mixture was incubated for 1.5 h at 42oC and stopped by the addition of 260 [mu]l 0.3 M sodium acetate, 10 mM Tris-HCl, pH 8.0, 1 mM EDTA followed by the addition of 600
[mu]l cold ethanol. The final pellets were dissolved in 6 [mu]l formamide loading buffer and electrophoresed on a 6% acrylamide/urea
gel. The position of primer extension stops and nuclease cleavage sites was
determined relative to a sequencing ladder prepared from the cloned cDNA and
the same primer.
Site-directed mutagenesis was performed using the pBluescript plasmid
containing the 5' 470 bp of albumin cDNA that bears the site which yields the
characteristic doublet cleavage product following digestion with the mRNase.
Mutagenesis was done by the method of Jones and Howard (
25
). The following mutations were prepared. Lower case letters represent the
change from the wild-type sequence. The HP160 mutation collapsed the loop containing the major
cleavage sites. The albumin plasmid pXa470 was amplified with pair EC4
(GTATACACCGCtcaGACTGAGCGGA, corresponding to position 147-171) plus EC6 (ATCAGCAATATGCTTGTGATGGT, corresponding to position 146-124) and pair EC5 (TCCGCTCAGTCtgaGCGGTGTATAC, corresponding to
position 171-147) and EC7 (CCTTCAAAGGACTTACATTGGCT, corresponding to position 171-193). PCR products were purified on agarose gels, mixed, annealed
and used to transform competent
Escherichia coli
. Positive clones were selected by colony hybridization to 5'-labeled primer EC5. The U164A mutation was prepared as above with
primers EC8 (GTATACACCGCATTGACaGAGCGGA, corresponding to position 147-171) plus EC6, and EC9 (TCCGCTCtGTCAATGCGGTGTATAC, corresponding to
position 171-147) plus EC7. Selection was done by hybridization to radiolabeled EC8.
The G165C mutation was prepared with primers EC10 (GTATACACCGCATTGACTcAGCGGA)
plus EC6, and EC11 (TCCGCTgAGTCAATGCGGTGTATAC) plus EC7. Selection was done by
hybridization to radiolabeled primer EC10. The U159A mutation was prepared with
primers EC16 (GTATACACCGCAaTGACTGAGCGGA) plus EC6, and EC17
(TCCGCTTCAGTCAtTGCGGTGTATAC) plus EC7. Selection was done by hybridization to
radiolabeled primer EC16. The T160A mutation was prepared with primers EC14
(GTATACACCGCATaGACTGAGCGGA) plus EC6, and primers EC15
(TCCGCTCAGTCtATGCGGTGTATAC) plus EC7. Selection was done by hybridization to
radiolabeled primer EC14. The A162U mutation was prepared with primers EC18
(GTATACACCGCATTGtCTGAGCGGA) plus EC6, and EC19 (TCCGCTCAGaCAATGCGGTGTATAC) plus EC7. Selection was done by hybridization to
radiolabeled primer EC18. All mutations were sequenced to insure that no
additional changes were introduced during PCR. Furthermore, the structure of
each of the RNAs prepared from these mutations was confirmed by digestion with
RNase T1. The cleavage of wild-type and mutated albumin RNAs by the mRNase was assayed using RNAs that
were end-labeled by
in vitro
transcription in a mixture containing [[gamma]-
32
P]GTP. A sequencing ladder was run as a size standard with each reaction.
Prior to digestion, 5' end-labeled transcripts were heated in buffer for each nuclease at 50oC for 5 min, and cooled slowly to 25oC to insure proper folding. RNase T
1
digestion buffer contained 2.5 mM Tris-HCl pH 7.2, 2.5 mM MgCl
2
and 25 mM KCl. S1 nuclease buffer contained 30 mM sodium acetate pH 4.6, 50 mM
NaCl, 10 mM zinc acetate and 5% glycerol. Buffer for the polysomal RNase
contained 40 mM Tris-HCl pH 7.2, 2 mM dithiothreitol and 5 mM MgCl
2
. All reactions also contained 0.25 [mu]g/[mu]l yeast tRNA. Digestions were performed at 25oC in a 20 [mu]l volume for 10 min for RNase T
1
and S1 nuclease, and 30 min for the polysomal nuclease. The amounts of nuclease
used were 0.05 U/[mu]l RNase T
1
, 0.05 U/[mu]l S1 nuclease and 10 ng/[mu]l of the polysomal RNase. The reactions were stopped by addition of 15 [mu]l 7.5 M ammonium acetate plus 115 [mu]l H
2
O, followed by extraction with phenol/HCCl
3
/isoamyl alcohol (25:24:1) and ethanol precipitation. The resultant pellets were
dissolved in 15 [mu]l phosphodiesterase buffer, which consisted of 100 mM Tris-HCl, pH 8.7, 100 mM NaCl and 14 mM MgCl
2
. Three [mu]l portions were mixed with 1 [mu]l snake venom phosphodiesterase (Boehringer) containing 3 * 10
-2
U/[mu]l. Reactions were incubated for 10 min at 25oC and were stopped by addition of 4 [mu]l formamide loading buffer followed by heating for 5 min at 70oC. Digestion products were analyzed by electrophoresis on a 6%
urea/acrylamide gel.
We previously reported that the cleavage which generates the doublet at position
194 (relative to the 5' end of a T3 RNA polymerase-generated transcript of albumin mRNA) mapped to the sequence
AUUGACUGA. This sequence, which can be thought of as two overlapping copies of
the sequence APyrUGA, lies between nucleotides 158-166 on albumin mRNA, the difference in numbering resulting from the
presence of polylinker sequence on the
in vitro
transcript. The numbering used here corresponds to the sequence of albumin
mRNA, with position 1 being the cap site.
The 194 nt doublet cleavage product does not accumulate stoichiometrically, even
with highly purified enzyme. We suspected that, unlike the homeobox-specific RNase identified in
Xenopus
oocytes (
11
,
12
), the liver polysomal RNase cleaves its substrate RNA at multiple sites, with
the major 194 nt product also serving as the substrate for additional
cleavages. To test this, reactions were performed with substrate that was
labeled on the 5' end by inclusion of [[gamma]-
32
P]GTP in the transcription reaction. The reasoning behind this was that the
relative signal intensity of digestion products from end-labeled substrate is independent of length, whereas signal intensity of
products from uniformly-labeled substrate is proportional to length. Furthermore, the use of end-labeled substrate allows an approximation of the relative activity
at a particular site independent of product size. Digestion of wild-type albumin transcript (Fig.
2
, lane 3) yields the characteristic doublet products, plus three additional
cleavages within the mapped structure domain. These are labeled 1-5 on the figure and indicated on the adjacent structure diagram. Sites 1
and 2 (black arrows) correspond to the cleavages at 159 and 163 which generate
the characteristic doublet product. Several weaker cleavage sites (3 and 4) are
also seen, plus an unexpected strong cleavage site was found further upstream
at position 5. The identification of additional cleavages on the substrate
transcript is in agreement with the observed absence of stoichiometry in
accumulation of the doublet cleavage product. With the exception of site 5, all
of the additional cleavage sites fall within single stranded regions. Site 5
contains the core sequence AUUGA in an AU-rich stem adjacent to a bulge, so it is conceivable that localized
unpairing of this structure might make the sequence accessible for cleavage. It
is noteworthy that the strongest cleavages all contain the sequence APyrUGA,
suggesting that this sequence is a preferred site for the RNase. Densitometry
was used to determine the relative signal intensities of the five bands
resulting from cleavage of wild-type transcript (lane 3). By arbitrarily setting the signal of the weakest
band (site 4) to 1, the strongest signal intensities were seen at site 1 (8.6),
site 2 (5.2) and site 5 (3.0), all of which have the APyrUGA pentamer. Stated
another way, cleavages at positions 1, 2 and 5 accounted for 43, 27 and 15%,
respectively of the total products in this portion of albumin mRNA. Since
cleavages at sites 3-5 will each reduce the amount of the doublet product seen with uniformly-labeled transcript, we conclude that these additional cleavages are
responsible for the lack of stoichiometry observed in our assays using the
purified RNase.
A series of point mutations were prepared in various positions within the 158-166 loop to characterize the sequences involved in cleavage of this
region by the polysomal RNase. We could not perform saturation mutagenesis
because many of these changes altered the secondary structure of the substrate.
Rather, only those mutations that did not alter the secondary structure of the
RNA were selected, a result which was confirmed by enzymatic secondary
structure analysis as in Figure
1
(data not shown). Lanes 6-9 in Figure
2
shown the effect of two single base mutations made in the 3' portion of the loop which converted U164 to A and G165 to C. Both
mutations abolished cleavage at site 2 (lanes 7 and 9), but had no effect on
cleavage at site 3 on the opposite side of the loop or any of the other
cleavages.
The effect of mutations at the top and in the 5' half of the 158-166 loop are shown in Figure
3
. The A162U mutation at the top of the loop had no effect on cleavage at any
site (compare lanes 2 and 4). Similarly, converting U159 to A had no effect
(lane 6). However, mutating the second pyrimidine (U160) in this sequence to A
abolished cleavage at site 3 with no effect on other sites within the molecule.
We conclude from the data in Figures
2
and
3
that, while the APyrUGA sequence element is a major site for cleavage by the
mRNase, it is by no means the only site. Furthermore, cleavage at the APyrUGA
element requires that this sequence be in a single-stranded context, and the most critical residues affecting cleavage are
the adjacent U and G nucleotides.
A number of important insights were gained by studying sequence and structural
elements involved in cleavage by the polysomal RNase. As noted above, the
cleavage sites responsible for generating the characteristic doublet product
from albumin mRNA (sites 1 and 2) lie in a single-stranded loop which bears the sequence AUUGACUGA. We previously noted that
this can be viewed as an overlapping repeat of the sequence APyrUGA (
23
). Another cleavage site (site 5) also contains the sequence AUUGA in a bulged
stem structure that could conceivably become single stranded by localized
unpairing. Of the five sites analyzed here, cleavage at these three sites
comprise 85% of the products mapped to this region of albumin mRNA. The fact
that cleavage occurs at sites other than APyrUGA indicates that the polysomal
RNase is sequence selective (as opposed to being sequence specific). This
contrasts with the oocyte RNase that cleaves only a specific sequence within
maternal homeobox mRNAs (
11
,
12
).
Data obtained with the hp160 mutation, in which the 158-166 loop was collapsed to a hairpin, demonstrate that the nuclease is
selective for single-stranded RNA and cannot cleave even though the APyrUGA element was
present. Thus, while this element could occur with relatively high frequency in
cellular mRNA, it must be in a single-stranded portion of the molecule to be accessible to the polysomal RNase.
The sequence APyrUGA is present 14 times in albumin, nine times in transferrin
and seven times in [gamma]-fibrinogen mRNA, all of which are destabilized following estrogen (
18
). Ferritin mRNA, which is not degraded by the polysomal nuclease (
18
,
23
), lacks the sequence APyrUGA. Cleavage sites 3 and 4 do not contain this
pentamer, so it is not clear how they are selected. Both are cleaved to the
same degree in the hp160 mutation, so selection at these sites must be
independent of cleavage in the loop. The hp160 construct retained APyrUGA at
position 5, so it is conceivable that a single site is sufficient for substrate
selectivity, with additional cleavages resulting from those sequences adopting
a particular secondary or tertiary structure. Alternatively, selectivity may
come from the overall secondary or tertiary structure, with APyrUGA serving as
a preferred cleavage site when present in a single-stranded portion of the molecule.
A major pathway for mRNA degradation in prokaryotes consists of endonucleolytic
cleavage by RNase E in a single-stranded region adjacent to a stable structural element, followed by 3'-5' degradation of the resultant product by polynucleotide
phosphorylase (reviewed in
26
). These two steps in mRNA turnover appear to be coupled, as both enzymes are
found in a single macromolecular complex (
27
-
30
). Since the prokaryotic and eukaryotic 3'-5' exonucleases that have been identified require a substrate with a
free 3' hydroxyl for maximal activity, one would anticipate that vertebrate
endonucleases involved in mRNA turnover would generate products like those
resulting from RNase E cleavage. Results in Figure
4
show that cleavage by the polysomal RNase generates products with free 3' hydroxyls. This contrasts with our earlier finding that cleavage
products had 5' ends that could be phosphorylated by T4 kinase. The latter finding most
likely resulted from dephosphorylation of a sub-population of RNase cleavage products. Subsequent incubation with kinase
and [[gamma]-
32
P]ATP would give the erroneous impression that the entire population became
phosphorylated. Since the starting substrate was labeled in Figure
4
, the approach used here more accurately reflects the properties of the entire
population of degradation fragments.
In addition to its action to destabilize albumin and other mRNAs encoding serum
proteins, estrogen causes the transcriptional induction and stabilization of
the mRNA for the yolk protein precursor vitellogenin (
31
,
32
). Recently Dodson and Shapiro (
33
) described an estrogen-regulated protein which selectively binds to a region in the 3' UTR of vitellogenin mRNA and may be involved in the regulation of
its stability. The core sequence for this binding contained either one or two
copies of the element ACUGAU, depending on whether they examined the B2 or B1
vitellogenin mRNA. This portion of both vitellogenin mRNAs as well as the
region of albumin mRNA examined here are shown in Figure
5
. The similarity to polysomal RNase cleavage sites characterized in the present
study is striking. Based on this observation, a potential mechanism for the
stabilization of vitellogenin mRNA by estrogen might be the induction of a
protein that blocks access to this sequence by a messenger RNase. Binder
et al
. (
14
) have proposed a similar mechanism for the stabilization of transferrin
receptor mRNA, in which an iron-regulated protein (IRP1) occludes an endonuclease cleavage site by binding
to iron response elements in the 3' UTR. If this turns out to be the case for the stabilization of
vitellogenin, it will be important to determine why the same protein does not
prevent the destabilization of albumin and other serum protein-coding mRNAs bearing the APyrUGA element. The answers to these questions
may provide important insights into the molecular interactions involved in the
differential regulation of mRNA stability.
We wish to thank Murray Deutscher for his suggestion about using snake venom
exonuclease for mapping the 3' ends of the cleavage products. This work was supported by Public Health
Service Grant GM38277 from the National Institutes of Health. R.E.D. was
supported in part by a Research Supplement for Underrepresented Minorities from
the National Institute of General Medical Sciences and by a Ford Foundation
Minority Postdoctoral Fellowship.
*To whom correspondence should be addressed. Tel: +1 614 688 3012; Fax: +1 614
688 3012; Email: schoenberg.3@osu.edu
+
Present address: Abbott Diagnostics, PO Box 278, Barceloneta 00617, Puerto Rico
REFERENCES
Return