ABSTRACT
Translation of an mRNA encoding a selenoprotein requires that at least one UGA
codon in the reading frame is recoded as a site for the insertion of
selenocysteine. In eukaryotes, the termination codon recoding event is directed
by a
cis
-acting signal element located in the 3
'
untranslated region of the gene. This `selenocysteine insertion sequence'
(SECIS) comprises conserved sequences in a region of extensive base-pairing. In order to study the structure-function relationships of the SECIS structure, we have applied a
newly developed reporter gene system which allows analysis of stop codon
suppression in animal cell lines. This system obviates the need for enzymatic
or immunological estimation of selenoprotein synthesis, relying instead on the
simple quantification of translational readthrough from the
lacZ
gene into the luciferase gene. The 3'-UTR of the phospholipid hydroperoxide glutathione peroxidase (PHGPx) gene
was shown to contain a highly active SECIS element. Mutations in the base-paired sequences of other SECIS elements were used to analyse the
significance of primary structure, secondary structure and pairing stability in
the stem regions. The results demonstrate that the exact sequences of the
paired nucleotides are comparatively unimportant, provided that a consensus
combination of length and thermodynamic stability of the base-paired structures is maintained.
Selenium is found covalently associated with certain prokaryotic and eukaryotic
proteins called selenoproteins (
1
-
3
). The carrier of this group VI element is the twenty-first amino acid, selenocysteine, which can be incorporated
cotranslationally into nascent polypeptide chains. The site of selenocysteine
incorporation in both prokaryotes and eukaryotes was found to be the UGA codon
(
4
-
7
), originally identified as one of the nonsense codons in the genetic code (
8
). This alternative interpretation of the UGA codon (`recoding';
9
) requires the presence of additional signals. In
Escherichia coli
, UGA is only read as a selenocysteine incorporation site if it is immediately
followed in the reading frame by a region that has been shown to form a stem-loop structure (
10
). An element with an apparently analogous function, referred to as the
selenocysteine insertion sequence (SECIS;
11
) or selenium translation element (STE;
12
), exists in selenoprotein-encoding mRNAs of eukaryotes (
7
,
13
). However, unlike the prokaryotic signal, the eukaryotic element is found in
the 3' untranslated region (3'-UTR;
7
,
11
,
13
,
14
), and must therefore be effectively acting at a distance. The mechanism of
action of the SECIS elements is as yet unknown. In
E.coli
, recoding of UGA involves a number of specific cellular components, including a
special type of seryl-tRNA (tRNA
Sec
) and a unique translation factor (SELB) which is related to EF-Tu (
6
). Eukaryotic tRNA
Sec
species have also been described (
15
,
16
), but no eukaryotic counterpart of SELB has been characterized so far.
The SECIS element is an example of a structural element in eukaryotic mRNA that
most likely fulfils its function via interactions with an RNA-binding protein (analogously to the prokaryotic SELB). Although many
processes of posttranscriptional gene expression involve RNA-protein interactions, very little is known about the structural
properties and functional mechanisms that underlie them (
17
,
18
). An important approach to furthering our knowledge of the mechanisms of such
interactions is to obtain a deeper understanding of the structural
characteristics of mRNA binding sites that are necessary for function.
Typically for a number of known specific RNA binding sites for proteins (
17
), the SECIS type of element comprises a small number of conserved nucleotides
located in a stem-loop-like structure. The conserved sequences of the SECIS element are
three short stretches: AAA in or near the apical loop of the overall structure,
UG at an internal loop on the 3' side of the stem, and AUGA at an internal bulge site on the 5' side of the stem. The overall structure is generally large,
comprising 80-150 nucleotides. This feature is also conserved, at least as far as can
be judged from a theoretical assessment of the secondary structures likely to
be formed by the known SECIS elements (
12
-
14
).
In the present work we have applied a newly developed assay system to test the
relative levels of stop codon suppression directed by a range of SECIS
elements. This reporter gene system, which is based on the fusion of two
reporter genes, allows assessment of the functional properties of SECIS
elements independently of the selenoprotein genes with which these sequences
are normally associated. We have compared the activity of the recently
described SECIS element of the PHGPx gene (
19
) with the activities of those reported previously. Moreover, we have obtained
results that provide insight into the significance of the molecular shape and
dimensions of SECIS element structure, showing that these are more important
than simply the thermodynamic stability of its constituent parts.
All DNA manipulations were carried out according to standard procedures (
20
). Synthetic oligonucleotides used in the plasmid constructions are listed in
Table
1
. An overview of the constructed plasmids is given in Table
2
. In the first step towards the construction of pBPLUGA, the plasmid pBgalluc-1 (
21
) was cleaved with
Hin
dIII and partially with
Bam
HI. The annealed oligonucleotide pair GLBAM1/2 was inserted between the
resulting ends in the plasmid DNA leading to elimination of the
Bam
HI site, whereby a silent mutation in the N-terminal amino-acid sequence of the [beta]-galactosidase was introduced. The resulting plasmid was
called pBGLB. After digestion with
Sal
I and
Bam
HI, pBGLB was religated with the oligonucleotide pair GLUGA1/2, creating an in-frame stop codon in the intergenic region between [beta]-galactosidase and luciferase. In parallel, an oligonucleotide
pair (GLPLCYS1/2) containing a cysteine codon instead of the stop codon was
introduced into the same site. The polylinker oligonucleotide pair PL3U1/2 was
inserted into the
Kpn
I site of the 3'-UTR. The resulting plasmids were called pBPLUGA and pBGLPLCys.
Insertion of the polylinker sequence in the reversed orientation resulted in
the plasmid pBLPUGA. Insertion of the oligonucleotides GLSTOP1/2 (
Sal
I/
Bam
HI) between the two reporter genes fused the luciferase gene in the -2 frame relative to the
lacZ
gene (pBSTOP). pBPLUGA (Fig.
1
A) was the master plasmid used for the insertion of different SECIS elements,
which were ligated into the polylinker in the 3'-UTR. The plasmids pBPHGPx3U and pBPHGPxU3 contain the 3'-UTR of the pig heart PHGPx gene (
19
) in the natural and reversed orientation, respectively. This 3'-UTR was obtained by digesting the plasmid pMM7 using
Bgl
II and
Nsi
I. The resulting fragment was ligated with the
Bgl
II/
Pst
I-digested plasmids pBPLUGA and pBLPUGA. pMM7 is derived from pHST7 (
22
) and contains the 3'-UTR of the PHGPx gene (nucleotides 509-734 in ref.
23
), synthesized by PCR using the primers RPHGX and LPHGX and the vector pMM5 (
23
) digested with
Bgl
II and
Eco
RI. The other SECIS elements were cloned as synthetic oligonucleotides bearing
Bgl
II (5') and
Bam
HI (3') ends. The plasmid with the minimal PHGPx SECIS element (pBPHGPx
min
) comprises nucleotides 570-626 of the PHGPx 3'-UTR (Fig.
2
B). The SECIS element of the rat 5' deiodinase (pBDI, Fig.
2
B) comprises nucleotides 1519-1596 of the gene sequence (
11
). The mutant forms of this structure (pBDIM1 and pBDIM2) are indicated in
Figure
2
B. The construct pBSELP1
min
includes the nucleotides 1472-1519 (Fig.
2
B;
24
) of the first SECIS element of the rat selenoprotein P gene. The sequence of
the mutated form of this SECIS element (pBSELP1M1) is indicated in Figure
2
B. The plasmid pBSECIS contains a semi-synthetic SECIS element, cloned by integration of a
Bgl
II/
Afl
II oligonucleotide (SECIS1/2) in pBPLUGA. The mutation of the SECIS element of
the plasmid pBSECISM is marked in Figure
4
B. Further extensions (E) of the SECIS element were created using
oligonucleotides comprising naturally occurring sequences or other synthetic
extension sequences. The 5' extensions of the pBSELP1
min
SECIS elements were constructed using
Bgl
II oligonucleotide pairs (NE151/2, NE251/2, E51/2 or SE51/2) while the 3' extensions were created using
Bam
HI/
Pst
I oligonucleotide pairs (NE131/2, NE231/2, E31/2 or SE31/2). The plasmid
pBSELP1SEI contains the 5' extension in inversed orientation, whereby pBSELP1E3 contains only the 3' extension. Further stabilisation of pBSELP1E with additional base
pairs was achieved using a
Sna
BI/
Sac
II oligonucleotide pair (EE51/2) in the 5' arm and with a
Pst
I oligonucleotide pair (EE31/2) in the 3' arm after cleaving the plasmid with the respective restriction enzymes.
Inversion of the 3' extension gave the plasmid pBSELP1EEI. Extensions of the SECIS elements
of pBSECIS and pBSECISM, yielding the plasmids pBSECISE and pBSECISME were
introduced using the same strategy. All plasmids were characterized by
restriction endonuclease analysis and DNA sequencing.
Table 1
Table 2
BHK-21 cells (baby hamster kidney cells; ATCC CC110) were cultivated in
Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum.
Transient transfections with plasmids were carried out by means of calcium
phosphate coprecipitation (
25
). Determination of activities of [beta]-galactosidase and luciferase was performed in cells harvested 48 h
after transfection. Cells were detached from culture plates by incubation with
TEN (40 mM Tris-HCl, pH 7.5; 1 mM EDTA, 150 mM NaCl) after rinsing the cells twice with
PBS (137 mM NaCl, 2.7 mM KCl, 8 mM Na
2
HPO
4
,
1.47 mM KH
2
PO
4
; pH 7.0). After centrifugation, the cell pellet was resuspended and subjected
to three cycles of freeze-thaw in 250 [mu]l 250 mM Tris-HCl (pH 7.5). For measurement of the luciferase activity, 20-60 [mu]l of the supernatant was mixed with 350 [mu]l reaction buffer (25 mM glycylglycerine, pH
7.8; 5 mM ATP, 15 mM MgSO
4
). The light emission (light units/10 s) of the reaction mixture was measured in
a Berthold Biolumat (LB9501) after injection of 100 [mu]l luciferin (0.2 mM; Sigma). For determination of [beta]-galactosidase activity, 20-60 [mu]l cell extract were incubated with 1 [mu]l reaction buffer (60 mM Na
2
PO
4
, 40 mM NaH
2
PO
4
, 10 mM KCl, 1 mM MgCl
2
, 50 mM [beta]-mercaptoethanol) and 200 [mu]l substrate (2 mg/ml
o
-nitrophenyl- [beta]-D-galactopyranoside in 60 mM Na
2
PO
4
).The reaction was stopped by addition of 500 [mu]l 1 M NaCO
3
. The colorimetric change was measured at 405 nm.
The luciferase activity of cells transfected with the construct pBGLPLCys
(harbouring a cysteine codon between the coding regions of [beta]-galactosidase and luciferase) was taken as the reference value
(100%). Stop codon suppression efficiency was calculated by relating the
enzymatic activities of cells containing the listed constructs (Table
2
) to the values obtained with this control. The specific activity of the [beta]-galactosidase enzyme segment was shown previously to be insensitive
to the fusion of a further protein to its C-terminus (
21
). Thus the ratio of the respective enzyme activities can be taken as a reliable
indicator of the relative levels of SECIS-dependent readthrough.
We developed an assay system designed to facilitate analysis of elements in the
3'-UTR that influence stop codon suppression in mammalian cell lines.
It is based on the reporter genes encoding [beta]-galactosidase and luciferase, which are fused in frame via a TGA
stop codon (pBPLUGA, Fig.
1
A). After transfection of the respective plasmids into mammalian cell lines, the
DNA is transcribed under the control of the SV40 promoter and translation leads
to the synthesis of the reporter enzymes. Translation can either terminate at
the C-terminus of
lacZ
or, upon suppression of the TGA codon, continue through the linker region to
generate a fusion protein of [beta]-galactosidase and luciferase (GAL-LUC) (Fig.
1
B). The sequence context of the TGA codon can be altered by using the
Sal
I and
Bam
HI restriction sites either side of it. The polylinker region in the 3'-UTR serves as a site for insertion of known or putative SECIS
elements, or alternatively derivatives of them, using the restriction sites
shown (Fig.
1
A). The encoded enzyme activities are readily measured and afford a high level
of sensitivity. Calculation of suppression activity is based on the parallel
transfection of the reference plasmid pBGLPLCys. In this plasmid, the coding
regions of [beta]-galactosidase and luciferase are fused via an in-frame cysteine codon. Translation of the mRNA encoded by this
plasmid leads exclusively to synthesis of the fusion protein GAL-LUC. The efficiency of any given SECIS element can be calculated by
comparing the enzyme activities of pBGLPLCys-transfected cells, which are taken as 100% reference values, to those of
cells transfected with the appropriate derivative of pBPLUGA. Further
experiments revealed that variation of the sequences in the 3'-UTR region does not have any influence on either the enzyme
activities or the stability of the mRNA (data not shown).
We found the fusion gene reporter system to be suited for the analysis of wild-type and mutated forms of natural SECIS elements. Stop codon suppression
activities were measured in transiently transfected BHK-21 cells. We inserted the complete 3'-UTR of the recently described pig heart phospholipid
hydroperoxide glutathione peroxidase (PHGPx) gene (
19
). This supported a stop codon suppression activity of 2.8% (pBPHGPx3U, Fig.
2
A). A minimal SECIS element only harbouring the upper part of the stem-loop structure, including the conserved nucleotides (pBPHGPx
min
; Fig.
2
B), supports a 10-fold reduced activity (0.3%). In contrast, a construct lacking a SECIS
structure (pBPLUGA), or with the 3'-UTR inserted in the inverse orientation (pBPHGPxU3), mediates
suppression with an efficiency of 0.1%. This value reflects the naturally
occurring, non SECIS-dependent, stop codon suppression activity observed in mammalian cell
lines. A further control construct, in which the luciferase gene is fused in
the reading frame that is -2 with respect to
lacZ
(pBSTOP, referred as `control' in Fig.
2
A), yields the background value of the test system (0.05%).
We compared other known SECIS elements to that of the PHGPx gene. The SECIS
element of the gene encoding rat 5' deiodinase (nucleotides 1519-1596 in ref.
11
; pBDI) led to a relative stop codon suppression efficiency of 1.1% (Fig.
2
A). We also examined a loop mutant, previously tested by Berry and colleagues (
13
), which converts one of the conserved As into a G (pBDIM1; Fig.
2
B). It was found to reduce drastically the efficiency, although a residual
activity remained (0.16%). This contrasts with the complete elimination of
activity caused by a combination of this mutation with a further change in the
3' bulge which converts the conserved UG to UA (pBDIM2; Fig.
2
B). Finally, we also investigated whether a minimal SECIS element, comprising
the upper part of the first stem-loop of the selenoprotein P 3'-UTR (pBSELP1
min
; Fig.
2
B), is functional, finding that it supports stop codon suppression at a level of
0.2%. A deletion mutant form of this minimal element lacking the three
conserved As in the loop region (pBSELP1M1; Fig.
2
) showed no activity.
The above experiments demonstrate that a minimal SECIS element retaining only
the part with the conserved regions is sufficient to mediate stop codon
suppression above the level of the naturally occurring suppression activity of
the cell. However, activities typical of the complete wild-type sequences are evidently not attainable with such minimal structures.
Inspection of the different stabilities of the respective SECIS elements
analysed (Fig.
2
B) leads to the suggestion that there might be a correlation between activity
and structural stability. In order to examine this possibility we made use of
the synthetic minimal SELP1 element as the starting point for a further series
of constructs (Fig.
3
B). The stem was extended by introducing additional nucleotides (pBSELP1E). The
synthetic extension sequences were chosen arbitrarily and are not derived from
known SECIS elements. A 2-fold increase in the stability of this element resulted in a doubling of
the efficiency (0.2-0.4%). A control in which only the 3' arm was extended (pBSELP1E3) had the same activity as the minimal
selenoprotein P SECIS element. This approach left open the question as to the
significance of the specific sequences that had been inserted. We therefore
added back wild-type sequences to the minimal element that are present in the bottom part
of the natural selenoprotein P SECIS element. The first extension (pBSELP1NE1;
Fig.
3
B), which was predicted to lead to only a slight increase in stability compared
to the minimal element, did not stimulate the rate of selenocysteine
incorporation. In contrast, a further stabilization using additional wild-type sequences (pBSELP1NE2), yielding a predicted increase in stability ([Delta]G = -46.8 kcal.mol
-1
), led to an enhancement of stop codon suppression activity (0.85%). Overall,
this set of results indicated that extension of the minimal SECIS element to
allow formation of a more stable structure enhances activity.
We examined more closely the sequence requirements of the SelP SECIS minimal
element by exchanging the base-paired stem region between the apical loop and the bulges with base pairs
equal in base composition but different in sequence. This manipulation had only
a minimal effect on the predicted stability and did not change the overall
shape of the SECIS element (compare pBSELP1
min
, Fig.
3
B with pBSECIS, Fig.
4
B). Despite the complete substitution of all the nucleotides in each arm of the
stem region, there was little change in efficiency. Moreover, stabilisation of
this rearranged minimal element led also to an enhancement of stop codon
suppression efficiency (pBSECISE, Fig.
4
B; compare pSELP1E in Fig.
3
B). In contrast, the exchange of only one conserved base located in the apical
loop of the RNA structure eliminated activity (pBSECISM and pBSECISME, Fig.
4
).
We have described a novel reporter gene system for the analysis of
selenocysteine incorporation in eukaryotes. This system can be used for
structure-function studies of the mRNA elements involved in selenocysteine
incorporation as well as for investigations of the mechanism of this process.
Clearly, selenocysteine incorporation can be directed into a heterologous
protein that normally does not contain this amino acid provided the appropriate
3'-UTR sequences are present. This principle could be of use in
generating selenocysteine-containing derivatives of a range of proteins. Assuming that these
derivatives could be produced in sufficiently large amounts, this would provide
an alternative approach to solving the well-known phase problem encountered in X-ray diffraction analysis of protein crystals. The use of a reporter
gene fusion for the estimation of stop codon suppression obviates the necessity
to determine selenocysteine incorporation efficiency mediated by different
SECIS elements on the basis of measurements of the activities of specific
selenoproteins, radioactive labelling with
75
Se, or immunochemical methods (
11
,
12
,
14
). The system could also be utilized to analyse the significance of the codon
context of the suppressed TGA codon.
The new system has allowed us to assess the activity of the recently described 3'-UTR of the PHGPx gene (
19
) relative to the activities of previously identified SECIS elements. Our data
clearly show that the PHGPx 3'-UTR contains an active SECIS element. Indeed, this element
functions highly effectively (pBPHGPx3U, Fig.
2
) compared with the SECIS elements of both selenoprotein P and 5' deiodinase. This observation is relevant to the physiological regulation
of selenocysteine incorporation into different proteins in mammalian tissues.
Under conditions of limited selenium availability, there is apparently
selective incorporation of selenium into PHGPx in preference over GPx (
26
). The data presented here provide a potential explanation for this selective
incorporation phenomenon. The enhanced ability of the PHGPx SECIS element to
direct recoding of the UGA codon may allow selenocysteine incorporation into
this protein to function relatively efficiently at limiting selenium
concentrations. The same principle may also play a role in the organ-specific expression hierarchies that have been described for
selenoproteins (
27
,
28
). It should be noted that our results indicate similar efficiencies for the 5' deiodinase and SelP SECIS elements, whereas Berry
et al
. (
13
) reported that the SelP element is the more effective of the two. This might be
at least partially attributable to the different cell types used in the
respective studies. Moreover, Berry and colleagues (
13
) used large parts of the 3'-UTRs of the respective selenoprotein-encoding genes, whereas we have focused on the sequences
defining the SECIS element itself.
An interesting mechanistic problem arises in the case of a natural gene, such as
SelP, which has 10 TGA sites at which selenocysteine has to be incorporated. If
incorporation were to occur independently at each site with an efficiency as
low as reported here, the yield of complete SelP protein would be negligible,
since it would be a function of the multiplicative sum of all 10 insertion
efficiencies. This clearly cannot be true, and we assume that suppression of
translational termination in selenoprotein mRNAs follows a processive
mechanism. One explanation of this would be that the presence of a SECIS
element in the 3'-UTR leads to a `reprogramming' of a small percentage of the
ribosomes translating that particular mRNA. The degree of reprogramming of the
translational apparatus would be equivalent to the suppression efficiencies
measured in an assay system of the type we have presented. The `reprogrammed'
ribosomes would then interpret all UGAs on the mRNA that they are translating
as selenocysteine incorporation sites. It should also be noted in this context
that the low efficiency of SECIS-dependent UGA suppression is an intrinsic property of the incorporation
mechanism. It was shown previously that translation of the complete reading
frame of the 5' deiodinase is 20- to 400-fold more efficient if the internal UGA is substituted by an
UGU cysteine codon (
29
).
We have studied the components of SECIS elements directly involved in the
formation of the secondary structure of this type of element. In particular, we
have investigated the nature of the boundaries of an efficiently functioning
SECIS element and the minimal structural and sequence requirements for
activity. We find that the stability of the base-paired regions in the SECIS structure is, in itself, an inadequate
criterion for the structural and functional significance of these regions
(compare pBSELP1SE to pBSELP1EE and pBSELP1EEI; Fig.
3
B). Equally stable base-paired stem regions can be substituted into the SECIS element in which a
smaller number of G[middot]C base pairs achieves the same stability as that of the equivalent
region in a natural SECIS element. Yet these shorter stem regions support only
greatly reduced activity. This indicates that the length, and thus overall
topology of the SECIS element, is essential for optimal function. Moreover, the
less than maximal density of highly stable base pairing may allow a degree of
flexibility that could be required for the SECIS to be able to perform its
function, for example in a `looping back mechanism', as proposed previously by
Berry
et al
. (
13
). The content of distorted helical sections and A[middot]U base pairs presumably contributes to specific properties of SECIS
elements that cannot be achieved in more compact structures. For example, one
functional requirement of the SECIS structure may be that at least sections of
it need to be partially unwound or restructured.
While the correct balance between length and stability of base-paired stretches in the SECIS element is essential for optimal function,
the exact nucleotide sequence in the respective arms can be varied with little
effect. This principle even applies to the base-paired arms immediately adjacent to the apical loop (compare pBSELP1
min
, Fig.
3
with pBSECIS, Fig.
4
). This suggests that apart from the nucleotides that are conserved between
SECIS elements, the major criterion for identification of a SECIS element is
the ability of the component regions to assume a specific shape and size.
Examination of other known protein-binding sites in mRNA reveals that a number of them, such as the iron
responsive element (IRE;
30
), the Rev-responsive element (RRE;
31
), the
trans
-activation response element (TAR;
32
) and the
U1A 3'-UTR protein-binding site (
33
), comprise partly base-paired structures. The specifically recognized and conserved nucleotides
generally lie in single-stranded regions, or in parts of the structures unlikely to form stable,
undistorted helices (
17
). An informative comparison of sequence and structure can be made with the results of structure-function studies of the IRE. It has been shown that a minimal IRE
structure comprising the conserved apical loop and a 10 bp A/U-rich stem containing a single-base bulge is sufficient to mediate strong translational repression when present in the 5'-UTR of mRNAs in animal cells (
34
,
35
) and yeast (
36
). This minimal structure lacks the base-paired `flanking regions' that are otherwise present in ferritin mRNAs (
37
), and which would normally be expected to stabilize the IRE structure
considerably. Yet binding of the iron regulatory protein (IRP) to the minimal
IRE is evidently still sufficiently tight to lead to regulatable translational
inhibition. This contrasts with the drastic loss of SECIS function observed upon size reduction. The minimal IRE required for binding of IRP, and thus for
tight translational regulation via the 5'-UTR, is evidently considerably smaller than the SECIS structure
necessary in the 3'-UTR for mediation of selenocysteine incorporation. One possible
explanation for the drastic loss of function associated with reductions in
overall size of the SECIS element might be the consequent changes in its
binding behaviour towards protein factor(s) essential for selenium
incorporation.
We thank Dr Joop van den Heuvel and Dr Matilde Maiorino for providing pMM7, as
well as Dr Hansjörg Hauser for pBgalluc-1.
REFERENCES
Return