ABSTRACT
The formation of a genomic RNA dimer appears to be a critical step in the life
cycle of all retroviruses. To investigate the site and nucleotide interactions
involved in this process, a 531 bp DNA fragment encompassing sequences up- and downstream of the splice donor in human T cell leukaemia virus type 1
(HTLV-1) was inserted into a plasmid vector under the control of the SP6 promoter. RNA transcripts generated
in vitro
from this template formed dimers which could be dissociated by heating at 60-80
oC for 3 min. The physical properties of the dimeric RNA were not consistent with
either Watson-Crick base pairing or guanine tetrad formation as being solely
responsible for the interaction. Deletion mutagenesis identified a 32 nt
sequence required for dimerisation. Computer modelling was carried out in order
to identify putative RNA secondary structures within this essential region. A stem-loop structure was identified, the stem of which was conserved among different sequenced isolates of HTLV-1. This sequence also contains a 15 nt palindrome. We sought by disruptive and compensatory mutagenesis to define the possible roles of these two structures
in dimer linkage.
Retroviruses have a diploid RNA genome. In the 1970s electron microscopy under
partially denaturing conditions revealed that the two molecules appeared to be
joined at a single locus towards their 5' ends (
1
,
2
). Various structures have been suggested for the bonding between the two
molecules (
3
-
18
), but as yet there is no definitive proof of a conserved mechanism among the retroviridae.
The exact role played by dimerisation of genomic RNA in the retrovirus life
cycle is unclear, but there is evidence for its potential involvement in
recombination, packaging of the viral genome and maturation of the virion (
19
-
21
). In almost all instances, retroviral dimer linkage sites have been mapped to a
region near to or within that of the packaging signal or [psi] (
10
,
22
,
23
). Retroviral dimer formation therefore constitutes a potential antiviral
target.
HTLV-1, a type C retrovirus, is the aetiologic agent of both adult T cell lymphoma (ATL) and human T cell leukaemia virus- associated myelopathy/tropical spastic paraparesis (HAM/TSP). Due
to the high degree of sequence conservation in the HTLV-1 genome, it represents a good model for
in vitro
studies of dimerisation. In addition, since it is transmitted by cell-cell interactions, it may help elucidate the role of RNA dimers in this process.
We studied dimerisation
in vitro
using a 531 bp insert containing sequences up- and downstream of the major splice donor in HTLV-1. This region dimerised readily, the molecules dissociating at
temperatures similar to those demonstrated by virion-derived RNA in other systems (
24
).
Physical and chemical conditions required for dimer formation were studied and
the system optimised to generate reproducibly stable dimerisation. Deletion and
site-directed mutagenesis were employed to map the site necessary for RNA
interaction and the localisation was compared with secondary structure models derived from computer predictions and analysis of the identified sequence.
Plasmid pGEM11[Delta]LTR2 was constructed by insertion of nt 297-828 from HTLV-1
CH
(
25
,
26
; numbering according to Seiki
et al
.;
27
) into a pGEM11Zf+ vector (Promega) (Fig.
1
a). In this vector, sense transcripts are generated using an SP6 promoter.
For full-length sense transcripts, plasmids were linearised with
Nhe
I (Boehringer Mannheim) and the template was prepared by proteinase K digestion
followed by phenol/chloroform extraction and ethanol precipitation.
In vitro
transcription was carried out using the Promega Riboprobe system. The
transcripts were cleaned as in the manufacturer's instructions and resuspended
in diethyl pyrocarbonate (DEPC)-treated water. Labelled RNA was prepared incorporating [[alpha]-
32
P]GTP (3000 Ci/mmol; NEN, Dupont) or [[alpha]-
32
P]UTP (800 Ci/mmol; NEN, Dupont).
RNA was diluted 1:5 in buffer D (250 mM cacodylic acid, 40 mM KCl, 5 mM MgCl
2
). The dissociation temperature was determined by heating aliquots of the RNA at a range of temperatures for 3 min and running the products on a 2% non-denaturing agarose gel. The gels were fixed, subjected to autoradiography
and the radioactivity counted by real time analysis with an Instant Imager
(Packard). To test the stability of the dimer, RNA was heated at each
temperature for 3 min, before allowing time for dimers to reform by incubating
at 37oC for 60 min.
Plasmid pST8+ was used to produce heterologous RNA for control experiments. It
contains a cDNA copy of the influenza virus strain A/PR8/34 segment 8
downstream of an SP6 promoter and flanked by the 5' and 3' untranslated regions from the
Xenopus
[beta]-globin gene (
28
). The plasmid was linearised using
Xba
I (Boehringer Mannheim).
In vitro
transcription using this template and/or the HTLV-1 pGEM11[Delta]LTR2 template was performed as described before using SP6
polymerase. The RNA was cleaned as described previously. Plasmid pST 8+
generates an 850 nt transcript and any heterodimers that formed would be
expected to run as a band of ~1381 nt. RNA was diluted 1:5 in dimer buffer and run out on a 1% TBE-agarose gel.
In order to investigate the effect of cations on dimerisation, increasing
concentrations of the cation to be investigated were incorporated into buffer D
in place of the 40 mM KCl. The RNA was dissociated at 80oC for 3 min and allowed to reform dimers as before. Stability of the dimer
under denaturing conditions was investigated by running the RNA out on a 2%
denaturing gel (2% agarose, 20% formaldehyde). Antisense transcripts were generated by linearising the parent plasmid with
Hin
dIII (Boehringer Mannheim), preparing the template as before and carrying out
the
in vitro
transcription using a T7 promoter.
Restriction enzymes were used to generate 3' end deletions.
Bgl
I (New England Biolabs) generated a 363 nt transcript,
Rsa
I (Boehringer Mannheim) generated a 384 nt transcript and
Avr
II (New England Biolabs) generated a 498 nt transcript. Transcripts were diluted
as described above and run out on 2% gels heated (80oC, 3 min) or unheated to determine whether or not they formed dimers.
Internal deletion and site-directed mutagenesis was performed by the Kunkel methodology (
29
). RNA was prepared as described previously and dissociation temperature
experiments carried out in order to assess the relative stability of the dimers
formed. Mutants were sequenced for verification using the Sequenase Version 2.0 DNA sequencing system (United States Biochemical) on 6% acrylamide gels (National Diagnostics).
Free energy minimisation studies of the 531 nt transcript were carried out by
folding overlapping windows of 200 bases. The programs employed were MFOLD,
adapted for GCG (Genetics Computer Group, University of Wisconsin, WI) (
30
,
31
), with the graphical presentation of Squiggles (
32
) in the GCG program PLOTFOLD, as well as STAR (
33
).
The full-length 531 nt RNA transcript formed dimers and monomers following
in vitro
transcription (Fig.
1
b, lane 1), which could be fully dissociated by heating at 80oC for 3 min (Fig.
1
b, lane 5). Incubation of the HTLV-1 transcript with a heterologous transcript was carried out to test the
specificity of the interaction. The heterologous influenza virus transcript did not form homodimers (Fig.
1
c, lanes 2 and 5), neither did it form heterodimers with the HTLV-1 transcript (Fig.
1
c, lanes 3 and 6), testifying to the specificity of the interaction we observed.
Following dissociation at 80oC for 3 min, the dimers would reform only after incubation at 37oC for 60 min in the presence of high sodium or potassium ion
concentrations (data not shown). However, the ionic concentration required (>1 M) was not consistent with the guanine tetrad model as proposed for HIV-1 (
3
,
15
).
On 2% agarose gels containing formaldehyde, the RNA maintained dimeric mobility,
indicating that Watson-Crick base pairing is not the sole interaction involved in the formation
of HTLV-1 RNA dimers (data not shown). This is in agreement with investigators studying dimerisation of HIV-1 RNA (
15
). Antisense transcripts were generated by linearising the plasmid with
Hin
dIII and using T7 polymerase. The 531 nt antisense transcript formed dimers to a
greatly reduced extent (Fig.
2
).
A series of 3' end deletions defined a 114 nt region essential for the RNA to form
dimers (Fig.
3
a).
Bgl
I generated a 363 nt transcript and
Rsa
I generated a 384 nt transcript, both of which failed to form dimers (Fig.
3
b, lanes 1 and 8 respectively).
Avr
II generated a 498 nt transcript which did dimerise (Fig.
3
b, lane 10), thus localising the dimer linkage site to a region between nt 681
and 795. A series of sequential deletions of ~30 nt was designed and introduced into the full-length template (Fig.
4
a). DM1, DM3 and DM4 formed dimers to the same extent as the unmutated HTLV-1 transcript, however, the DM2 deletion virtually abolished dimer formation (Fig.
4
b, lane 2). A key sequence was therefore identified between nt 718 and 749, upon which dimer formation
appeared to depend.
We went on to examine the sequence we had located by both computer aided
prediction of potential secondary structures and by introduction of point
mutations to test these predictions, as well as models suggested by our own
sequence analysis.
Nucleic acid folding programs MFOLD and STAR predicted a small stem-loop structure between bases 732 and 744. The stem-loop spanned the region identified as essential for RNA dimerisation by mutagenesis. This is a small but conserved
structure, the stem of which was conserved among all HTLV-1 sequences we examined (
25
,
26
,
34
) (Fig.
5
a).
Observation also revealed the presence of a palindromic sequence (nt 730-744; see Fig.
5
b) which would have the potential to form a stable base paired link between two
RNA molecules, analogous to the model proposed for dimer initiation in HIV-1 (
12
).
A series of mutants (L1-L12) which would disrupt the stem-loop structure and additional compensatory mutants (L5C-L7C) were made which probed within the region deleted by
DM2 (Table
1
). The percentage of dimer formed initially and the stability of the dimers at
various temperatures was examined for each of the mutants and compared with the
unmutated HTLV-1 template. Mutants L1 and L3, which changed bases within the loop of the
structure identified by computer modelling, had little or no effect on dimer
formation, whereas mutants L2 and L4-L8 all showed some reduction. This appeared to support the stem-loop structure. However, the compensatory mutants L5C, 6C, 7C, 9 and 10 also
reduced dimer formation, implying that the stem structure alone was not
important for dimer formation. Mutants L11 and 12 were designed to introduce
compensatory changes with adenine and uracil bases, but also reduced dimer
formation.
Table 1
Site-directed mutagenesis of the 32 bp sequence
The number of base pairing changes which would be introduced into the palindrome model by the mutagenesis are detailed in Table
1
, column 3. Mutant L6C, which introduces C:C and G:A mismatches in place of G:C
and A:U pairings, showed a decrease in dimer formation of 89% compared with the
native transcript, suggesting that base pairing is important. However, L3 (for example) also introduces four base pair mismatches, incorporating unfavourable C:C and U:U bonds, and this mutant dimerises nearly as well as the
unmutated transcript. The data shows a tendency to more pronounced inhibition
of dimerisation when putative Watson-Crick pairing in the outer half of the palindrome is disrupted. However,
even this is not consistent, as L7 shows a severe phenotype but does not affect
a Watson-Crick pair. Even so, disruption of the palindrome does appear to affect dimerisation,
since most of the mutations reduced dimer formation.
A 32 nt sequence has been identified which is essential for dimerisation of HTLV-1 RNA transcripts in an
in vitro
assay system. This sequence lies downstream of the major splice donor and just
upstream of the primer binding site. In the closely related retrovirus bovine
leukaemia virus (BLV), the dimer linkage site has been defined as a 129 nt
sequence lying downstream of the splice donor and including the primer binding
site (
22
,
23
,
35
). In addition, this sequence overlaps with the primary packaging signal of BLV
(
23
). As yet the packaging signal of HTLV-1 has not been mapped.
Earlier work on RNA dimerisation in HIV-1 (
3
,
15
) suggested that guanine tetrads might be involved in dimer formation in a
manner comparable with that detected in telomeres. This was based largely on
the findings that dimer formation of
in vitro
transcripts of HIV-1 RNA were stabilised by potassium ions and that dimer formation was
disrupted by substitution of guanine residues by other bases (
15
). The finding in our study that only very high, non-physiological concentrations of cations stabilised HTLV-1 dimer formation does not support this mechanism in this
retrovirus. In addition, deletion of the primer binding site, which contains a
sequence of five consecutive guanines, had no effect on dimer formation.
We also examined the possibility that Watson-Crick base pairing might play a role in the formation of RNA dimers. In
agreement with other reports (
15
), we found little evidence for this, since our transcripts were remarkably stable under denaturing conditions and failed to form significant amounts of dimer in antisense.
However, as discussed below, a palindromic sequence is contained within the
region we have identified as essential for dimerisation in our system and so we
cannot rule out the involvement of Watson-Crick base pairing altogether.
Deletions from the 3' end of the insert identified a 114 nt region which was essential for
dimerisation of the RNA transcripts. Of the sequential 30 bp deletions which
were made across this portion of the sequence, three of the deletions had no
effect on dimer formation. The DM2 deletion, however, which deletes a sequence
just upstream of the primer binding site, almost completely prevented dimer
formation.
Computer modelling of the entire 531 nt sequence identified a small stem-loop structure conserved amongst all the predicted foldings, spanning the
32 nt region determined by mutagenesis to be essential for RNA dimer formation.
Minor variations in the loop did not affect dimer formation, however, the
importance of the structure in dimer formation was not supported by results of
site-directed mutagenesis directed at changing bases on either side of the
stem.
The whole 531 bp HTLV-1 insert was searched using FASTA and BLAST searches for homologies to
palindromic sequences described recently in the literature as being required
for RNA dimer formation in HIV-1 and MoMuLV (
7
-
9
,
12
,
14
). No homologies were found to these sequences. However, the 32 nt sequence
identified is highly palindromic and it is possible that sequence, rather than
secondary structure, is crucial to the interaction observed here. In addition
to the secondary structure identified by the computer predictions, we detected
a 15 base palindromic sequence spanning from nt 730 to 744. This interaction
was disrupted by site-directed mutagenesis, however, the results did not support interstrand
complementarity from base pairs in this palindrome as the major RNA linkage.
Some mutations caused major disruption of dimerisation while others had little
effect. There was a tendency for mutations disrupting base pairing at the outer
halves of the palindrome to have a more pronounced effect.
In conclusion, a 32 nt sequence located at the 5' end of the HTLV-1 genome has been identified which appears to be necessary for
dimerisation of RNA transcripts. We cannot conclude from our data that this
sequence would be sufficient for dimerisation of full-length genomic RNA, however, we intend to address this question. Although
the mechanism by which dimeric transcripts form has yet to be fully elucidated,
it is clear that guanine tetrads are not involved and that Watson-Crick base pairing is not the sole interaction. Predicted secondary
structures, though formed by conserved sequences, also appear unlikely to be
involved. We have observed a palindromic sequence which might be involved in
the dimerisation process, although not as the sole interaction. The nature of
the dimer linkage structure remains unclear, although a site for the
interaction has been defined in HTLV-1. As no current nucleotide interaction model fits the physical
characterisation of this interstrand pairing it must be assumed that the
linkage, which forms in the complete absence of protein, is a novel structure.
Further work is under way to distinguish the precise nucleotides in the dimer
linkage and to examine the affect of the DM2 deletion in the context of an
infectious molecular clone.
We thank Dr Paul Digard for the gift of plasmid pST8+, Dr Edwin Ten Dam for
helpful suggestions and Teresa Barnes for secretarial assistance.
REFERENCES
Return