ABSTRACT
The 5
'
and 3
'
end of the HIV-1 RNA genome forms a repeat (R) element that encodes a double stem-loop
structure (the TAR and polyA hairpins). Phylogenetic analysis of the polyA
hairpin in different human and simian immunodeficiency viruses suggests that
the thermodynamic stability of the helix is fine-tuned. We demonstrated previously that mutant HIV-1 genomes with a stabilized or destabilized hairpin are severely
replication-impaired. In this study, we found that the mutant with a destabilized
polyA hairpin structure is conditionally defective. Whereas reduced replication
is measured in infections at the regular temperature (37
o
C), this mutant is more fit than the wild-type virus at reduced temperature
(33
o
C). This observation of a temperature-dependent replication defect underscores that the stability of this RNA
structure is critical for function. An extensive analysis of revertant viruses
was performed to further improve the understanding of the critical sequence and
structural features of the element under scrutiny. The virus mutants with a
stabilized or destabilized hairpin were used as a starting point in multiple,
independent selections for revertant viruses with compensatory mutations. Both
mutants reverted to hairpins with wild-type stability along various pathways by acquisition of compensatory
mutations. We identified 19 different revertant HIV-1 forms with improved replication characteristics, providing a first look
at some of the peaks in the total sequence landscape that are compatible with
virus replication. These experiments also highlight some general principles of
RNA structure building.
The rapid evolution of the human and simian immunodeficiency viruses (HIV and
SIV), combined with an enormous research effort to resolve the nucleotide
sequence of different viral isolates (reviewed in ref.
1
), provides us with a wealth of phylogenetic data that can be used for the
analysis of regulatory RNA motifs that control virus replication. We performed
previously phylogenetic analyses of several RNA elements of the untranslated
leader RNA of HIV-SIV genomes (reviewed in ref.
2
), for example, the TAR hairpin motif involved in binding of the Tat
trans
-activator protein (
3
) and the DIS hairpin involved in genome dimerization (
4
). This phylogeny shows that different RNA structures are selected by evolution
to facilitate a particular function, and that structural mimicry exists in the
RNA world. Most importantly, such a comparative sequence analysis can provide
important information on the critical sequence and structure elements within
the RNA element under scrutiny.
The natural diversity in HIV-SIV viruses, however, is likely to represent only a very limited section
of sequence space. Whereas one could argue that the area occupied by modern HIV-1 motifs represents the highest peak, reached after a careful evolutionary
walk through all of sequence space, this seems unlikely because the genetic
repertoire of natural HIV-SIV isolates is constrained by the success of a predecessor of the
contemporary viruses (
5
). In other words, the existing RNA motifs may well lie at local, not global,
optima of sequence space. A classical virological method to locate other sequence
optima is the selection of revertant viruses in long-term infections with a replication-defective mutant. When viable viruses do arise in prolonged cultures
with such HIV-1 mutants, their genomes should be altered to allow more efficient
replication. This so-called forced evolution approach has proven to be particularly valuable in
the analysis of regulatory RNA motifs of prokaryotic viruses (
6
-
8
), eukaryotic viruses (
9
-
14
) and viroids (
15
). The approach works for most RNA viruses because of their tremendous genetic
variation, which results from high frequencies of nucleotide misincorporation
without error-correcting mechanisms during replication. This genetic approach can also
be combined with the powerful SELEX approach (
16
), thereby selecting replication-competent virus from a pool of viral genomes with a short segment of
randomized sequence (
17
-
19
). Such forced evolutions can further extend the phylogeny of sequences that are
compatible with virus replication. Analysis of phenotypic revertants may allow
one to identify important features within the element under study, e.g.
invariable nucleotides or pivotal basepairs in RNA structures, and may eventually provide useful information on the biological function of the motif.
HIV-1 encodes an RNA genome with a repeat (R) element of 97 nucleotides at the
extreme 5' and 3' termini (Fig.
1
). This RNA element folds two stem-loop structures, the TAR and polyA hairpins,
and both are evolutionarily conserved and critical for viral replication
(reviewed in ref.
2
). The TAR stem-loop binds the viral Tat protein and is involved in Tat-mediated transcriptional activation of the HIV-1 long terminal repeat (LTR) promoter. The second stem-loop
structure is termed the polyA-hairpin because it encompasses the AAUAAA signal for polyadenylation.
Obviously, polyadenylation should be strictly controlled such that the 5' signal is ignored and the 3' signal is efficiently recognized during virus replication.
Several mechanistic models have been proposed to explain this differential
polyadenylation of the HIV-1 genome (
20
-
28
), and the presentation of the polyadenylation signal in a hairpin structure may
provide additional regulatory possibilities. Alternatively, the hairpin at the
5' and/or 3' end of the HIV-1 RNA genome may play a role in other steps of the viral
replication cycle. In this study, we performed an extensive reversion analysis
with replication-incompetent HIV-1 mutants containing a stabilized or destabilized polyA hairpin
structure. The different evolution pathways observed in this study highlight
the genetic plasticity of retroviral RNA genomes.
The T cell line SupT1 was used throughout this study for transfection with wild-type and mutant HIV-1 molecular clones and subsequent infection with the corresponding
viruses. SupT1 cells were maintained in RPMI1640 medium with 10% fetal calf
serum and transfected by electroporation as described previously (
29
). HIV-1 plasmids were derived from the full-length molecular HIV-1 clone pLAI (
30
). Introduction of the polyA hairpin mutations into the 5'LTR region of the pLAI infectious clone was performed in two steps as
described previously (
31
).
SupT1 cells (15 * 10
6
) were electroporated with 15 [mu]g of the individual HIV-1 clones and subsequently resuspended in a volume of 15 ml RPMI medium
with 10% fetal calf serum. The culture was split directly into a 24-well tissue culture plate (0.6 ml/well) and incubated at 37oC and 5% CO
2
. The samples were maintained in culture for 57-89 days (A mutant) and 188-278 days (B mutant). Cells were split 1 in 10 approximately once a
week. When virus-induced syncytia were visible, cell-free virus was passaged onto a fresh SupT1 culture in 24-well plates. Initially 5 [mu]l culture supernatant was used per passage, but this
amount was gradually reduced to 0.1 [mu]l for mutant B, and eventually to 0.01 [mu]l for mutant A. The revertant sequences of mutant A (Fig.
3
) represent samples taken at different times: no. 1, day 86; no. 2, day 75; no.
3, day 66; no. 4, day 62; no. 5, day 59; no. 7, day 62; no. 8, day 82; no. 9,
day 89; no. 12a/b, day 59; no. 15, day 57; no. 16, day 60; no. 18, day 60; no.
19, day 86; no. 20, day 59. The B revertants (Fig.
4
) were sampled as follows: no. 1, day 198; no. 8, day 258; no. 9, day 248; no.
10, day 199; no. 16, day 209; no. 23, day 188 and 278; no. 24, day 261. For
each mutant we also included one revertant sequence that was not obtained in
the 24-well format, but in a larger culture volume (10 ml), the A200 and B127
samples were taken at day 200 and 127 post-transfection, respectively.
HIV-1 infected cells were collected by centrifugation (4 min, 4000 r.p.m.),
washed once with PBS, and solubilized in 10 mM Tris-HCl (pH 8.0), 1 mM EDTA,
0.5% Tween 20. The lysate was treated with proteinase K (200 [mu]g/ml) for 30 min at 56oC and 10 min at 95oC. The complete LTR leader region of proviral HIV-1 DNA was PCR amplified with the primer pair LAI5'X and AD-GAG as will be described in detail elsewhere
(Das
et al.
, in preparation). PCR fragments were ligated into the TA cloning vector pCRII
(Invitrogen) and sequence analysis was performed on an Applied Biosystems 373
DNA sequencer with the
Taq
DyeDeoxy Terminator cycle sequencing protocol.
We reported previously the construction and initial characterization of two HIV-1 mutants, designated A and B, that stabilize and destabilize the polyA
hairpin structure (
31
). Mutant A is stabilized by deletion of two bulging bases on the right-hand side of the stem and by modification of one G-U basepair into G-C (Fig.
1
). Mutant B contains four nucleotide substitutions on the left hand side that
were designed to open the central and lower stem segments, resulting in a
hairpin with a slightly rearranged basepairing scheme and a stability of -11.4 kcal/mol (Fig.
1
). To minimize effects because of changes in the nucleotide sequence, the
mutations did mimic natural sequence variation seen in other HIV-SIV isolates. For instance, three substitutions in mutant B mimic the
nucleotide sequence of the HIV-1 ANT-70 isolate (
1
).
To test the replication capacity of these HIV-1 variants, the SupT1 T cell line was transfected with the wild-type and mutant HIV-1 molecular clones. We measured that stabilization as well as
destabilization of the polyA hairpin inhibited virus replication (Fig.
2
, upper panel). Virus replication was delayed by at least 10 days for the B
mutant compared with the wild-type virus, consistent with a profound replication defect. Around day 20
we observed a rapid spread through the culture of this B mutant. However, we
should note that the slope of the replication curve is not an appropriate
measure of the replication capacity since virus production is limited by the
culture system at this stage of the infection (there simply are not enough
cells to support maximal replication). In other words, the replication assay is
only linear up to low levels of virus production, and it is better to use the
time of virus breakthrough as a relative measure of the replication capacity.
The replication defect of the A mutant is even more pronounced, but we
emphasize that the defect is not absolute because delayed virus production was
measured in cell cultures transfected with more plasmid DNA (e.g. 5 [mu]g instead of 1 [mu]g per 5 * 10
6
cells). Thus, both HIV-1 variants seem to be
quasi
-infectious, a term that has been introduced in poliovirus research to
describe an RNA genome that can replicate in transfected cells at greatly
reduced levels (
32
). These smouldering virus mutants were used for the genesis of an extensive set
of revertant viruses (see below).
To provide further evidence for the importance of the thermodynamic stability of
the polyA hairpin structure, we tested whether the replication capacity of the
destabilized mutant B was selectively improved at reduced temperature.
Obviously, this experiment is limited by the adverse effects of low temperature
on the host cell metabolism. We measured considerable cell growth and
reasonable replication kinetics of the wild-type HIV-1 virus at 33oC (Fig.
2
, lower panel). Whereas a severe replication defect of mutant B was apparent at
37oC, this mutant replicated significantly better than the wild-type virus at 33oC. This result indicates that the replication potential of
mutant B is temperature-sensitive. As expected, mutant A did not demonstrate improved replication
at lower temperature.
To generate a considerable number of revertant viruses, a large size
transfection of the SupT1 T cell line was performed with the mutant molecular
clones A and B (15 [mu]g DNA per 15 * 10
6
cells). The transfected cells were split in a 24-well tissue culture plate and cultured for an extended period. Replicating
virus appeared in several wells after a varying lag phase. For instance,
massive syncytia were observed in 15 wells of the A mutant between 3 and 4
weeks after transfection. The putative revertant viruses were passaged for
several months and cells were harvested for analysis of the proviral DNA
sequences. In general, it was much easier to isolate revertants of the
replication-impaired mutant A in comparison with the partially defective mutant B. In
order to obtain B revertants, we had to continue the cultures for up to 10
months. The HIV-1 LTR-leader region was PCR amplified from total cellular DNA, the PCR
fragment was cloned and at least two clones were sequenced for each well. The
results are summarized in Figures
3
and
4
for mutant A and B, respectively. Shown is the predicted RNA structure for the
wild-type/mutant/revertant sequences, with the calculated thermodynamic
stability indicated below the hairpins. The mutations observed in the revertant
viruses are marked by a black box. These mutations were observed in more than
one clone of an individual tissue culture well, indicating that the mutation
was fixed in the virus population and suggesting that it may provide a
selective advantage to the virus. In addition to these accumulating mutations,
there were also sporadic changes in individual clones. These base changes may
have resulted from PCR/sequencing errors and were ignored in this analysis.
Figure
3
shows that there are at least 13 escape routes for the stabilized mutant A. We
organized the different solutions into arbitrary groups on the basis of the
number and position of the acquired mutations. For instance, variants with a
single mutation on the right hand side of the stem are boxed in the upper right
corner and double mutants affected on both sides of the stem are grouped in the
lower left corner. Contrasting with this enormous variation in repair
strategies, we observed an invariable drift towards hairpin structures with a
reduced thermodynamic stability. Mutant hairpin A ([Delta]G = -25.7 kcal/mol) was consistently remodelled into a hairpin with less basepairing potential ([Delta]G = -14.8/-23.1 kcal/mol) and it is likely that some
revertants did not yet attain the most optimal configuration. For instance,
revertant no. 4 acquired only one mutation ([Delta]G = -23.1 kcal mol), but we think that this represents a sub-optimal intermediate because this particular mutation was
found combined with a second helix-destabilizing mutation in other revertants (no. 7, [Delta]G = -19.3 kcal/mol and no. A200, [Delta]G = -17.1 kcal/mol).
Weakening of the A hairpin can be accomplished by mutation of any basepair along
the helix. Mutation near the end of the helix segment may affect multiple
basepairs. For instance, revertant no. 12b opens the third basepair from the
top and consequently disrupts all three top basepairs, thereby creating a
larger single-stranded loop. An analogous situation is observed at the bottom of the
stem for revertant A200. The most frequent change was from a C-G basepair to a C
.
Revertant no. 1 exhibits a rather unusual pathway towards viability in that two
nucleotide insertions are observed. This is remarkable because insertions, as
well as deletions, are not frequently observed in this type of experiment. In
this study, variant no. 1 represents the only variant with a base insertion,
and it is even more striking that a second insertion was introduced in the same
genome. It is rather amazing that both insertions coincide with the two bulge
residues that were deleted from the wild-type sequence in mutant A. A similar phenomenon of precise multiple
reversions has been observed in studies with poliovirus mutants (
44
). We can rule out contamination by the wild-type virus because the revertant has two U-bulges, whereas the wild-type has one U- and one C-bulge. Furthermore, the third change in mutant A,
the G-U to G-C basepair conversion in the lower stem, is maintained in this
peculiar revertant no. 1.
The tendency of the A revertants to lower the basepairing potential contrasts
with the evolutionary drift observed for the revertants of the destabilized
helix mutant B (Fig.
4
). All B revertants acquired mutations that restore the stability of the polyA
helix. The wild-type hairpin has a thermodynamic stability of -15.3 kcal/mol, which was reduced to -11.4 kcal/mol in mutant B, but subsequently restored to -12.3/-16.9 values for the revertants. Thus,
additional mutations that restored the predicted basepairing of mutant B
rescued viral replication.
Two strategies are seen in repair of the stabilized B hairpin. One group
acquired one or multiple hits that lead to rearrangement of the central stem
domain (Fig.
4
; nos. 1, 10, 16, 24). In particular, the B hairpin structure with two proximate
1-nucleotide bulges is rearranged to form one bulge element consisting of
two nucleotides. Thermodynamically speaking this is an improvement of hairpin
stability by -2.6/-3.6 kcal/mol. The alternative strategy is observed in another
group consisting of three revertants (Fig.
4
; nos. 8, 9, 23). These isolates choose to improve the bottom part of the helix,
either by a mutation on the left or right hand side of the stem. The revertant
B127 combines the two strategies to arrive at the most stable RNA structure
(no. B127, [Delta]G = -16.9 kcal/mol). Longitudinal sampling of the no. B127 reversion
indicated that the central part of the hairpin was restored first (
36
). Stabilization of the bottom stem can apparently be achieved by several
routes. The terminal G-U basepair can be replaced by either
The combined results obtained for the A and B revertants convincingly
demonstrate that a hairpin with a wild-type stability does optimally serve virus replication. In general, the
selections were rich in terms of the large number of different revertants
isolated. The existence of so many possibilities for reversion is a further
argument that the sequence of this motif is not of primary importance. It may
be noted that all revertants acquired at least one mutation in the polyA
hairpin region, suggesting that remote second-site mutations, which could be indicative of long-range RNA tertiary structure interactions, cannot restore the
function of this hairpin. Furthermore, no `true revertants' or wild-type viruses were recovered, although some positions reverted back to the
original sequence (e.g. no. 1 for mutant A, nos. 8 and 9 for mutant B). Some
sequence stretches within the polyA hairpin did not change in any of the
revertant genomes and such constant motifs are likely to contain critical
sequence information (Fig.
5
: e.g. the GCUUAAGC palindrome motif at position 62-69). These signals may be used in a sequence-specific interaction with an RNA-binding protein. We also observed some hotspots for the
accumulation of reversion-based mutations (Fig.
5
). For instance, a G to A transition at one of two consecutive G residues in the
mutant A helix (position 90 and 92 in the wild-type sequence) was present in 10 revertants, and both Gs were mutated in
revertant no. 18 (Fig.
3
). These preferential reversion sites may represent mutational hot-spots. In addition, we observed a typical mutation bias. We scored 30
transitions and only five transversions. Among the transitions, the G to A
substitution (15*) was found to be dominant, consistent with other
in vitro
and
in vivo
HIV-1 replication studies (
33
,
34
).
One way to isolate structurally informative mutants is as second-site revertants of a replication-defective virus mutant. This approach was used previously by us to
study known RNA motifs in the HIV-1 leader that control several steps in the viral replication cycle (
4
,
12
,
29
,
35
). In this study, we used this approach to analyze the sequence and structure
requirements of a novel RNA hairpin motif that is present at both the 5' and 3' ends of the retroviral RNA genome. The results of this extensive
mutant/revertant analysis, combined with the phylogenetic analysis of natural
HIV-SIV sequences (
31
), indicates that the polyA hairpin structure is critical for efficient virus
replication. These combined analyses also reveal that the stability of the
polyA hairpin is confined to a narrow range around -15 kcal/mol. The sequence at many positions along the basepaired stem is
less critical. Other features of the structure (position and sequence of
bulges/internal loops, size of loop etc.) are not of primary importance as
well. For instance, natural HIV-SIV isolates seem to prefer to have the destabilizing bulged nucleotide(s)
on the right hand side of the polyA stem (
19
). This forced evolution analysis indicates that basepair mismatches can also be
used to destabilize a helix (see multiple revertants in Fig.
3
). We therefore think that these motifs are primarily there to restrain the
thermodynamic stability of the helix, which otherwise would exceed the
allowable margin. Consistent with this idea, we reported a correlation between
the helix length and the number of bulges in naturally occurring polyA
structures (
31
).
The finding that the replication defect of mutant B with a destabilized hairpin
is not apparent at 33oC is consistent with the notion that altered stability of the polyA hairpin is the primary cause of the
impaired replication at 37oC. If an important sequence motif was inactivated in mutant B, one would
not expect such a temperature-sensitive replication phenotype. Theoretically, two possibilities can explain the improved relative replication of mutant
B at 33oC. The stable wild-type hairpin may become inhibitory at reduced temperature, perhaps by
occluding the polyA signal for recognition by cellular polyadenylation factors.
Alternatively, the relatively instable RNA structure of mutant B could be more
active at lower temperature if an optimal hairpin stability was reached. We
cannot currently discriminate between these two possibilities, but some insight
is gained from calculation of the thermodynamic stabilities ([Delta]G) of the two hairpins at 33 and 37oC with the Zuker algorithm (
37
). Lowering the temperature by 4oC increases the stability of the wild-type structure from -15.3 to -17.1 kcal/mol, and the B mutant changes from -11.4 to -13.1 kcal/mol. Apparently, HIV-1 replication is better served by a
hairpin of [Delta]G = -15.3 than -11.4, and the 33oC assay suggests that -13.1 is more optimal than -17.1.
The polyA mutations introduced into the 5' end of the HIV-1 genome have no effect on the production of virion particles in
transfected cells (
36
), indicating that the 5' hairpin structure does not influence processes like transcription, mRNA
splicing and translation. Detailed analysis of these virions indicates that
opening of the hairpin structure leads to reduced packaging of HIV-1 RNA genomes in these particles (
36
). Obviously, the reversion experiments presented in this study deal with virus
genomes having both the 5' and 3' polyA mutation, and the repair pathways observed may be because
of a defect in function of the 5', 3' or both motifs. We are currently investigating the precise
contribution of the 5' and 3' motifs in the viral replication cycle.
A replication defect was also apparent upon further stabilization of the wild-type polyA hairpin, indicating that a too stable RNA stem in the HIV-1 leader transcript interferes with a replicative step. First,
scanning ribosomes may be blocked during translation of HIV-1 mRNAs (
38
,
39
). Second, the hairpin may also interfere with the elongating RT enzyme during
reverse transcription (
40
,
41
). Third, the stem structure may occlude the polyadenylation signal that
overlaps the stem domain, thereby preventing recognition of the AAUAAA signal
by the protein factors involved in mRNA polyadenylation. Consistent with the
latter idea, preliminary experiments with reporter gene constructs indicate
that stabilization of this HIV-1 hairpin reduces the efficiency of the polyA signal contained within this
structure (Klasens and Berkhout, in preparation).
A large variety of revertant genomes were obtained in this study. This gives us
the opportunity to investigate the plasticity of the primary sequence in a
functional RNA structure. One question is how far the nucleotide sequence can
be changed without destroying the function of this hairpin motif. The phylogeny
of this motif in natural HIV-SIV isolates suggested that many alternate solutions are compatible with
virus replication, and the phylogeny generated in this forced evolution study
underscores this flexibility. These results are generally consistent with a
theoretical study showing that RNA stem-loop structures can pervade all regions
of sequence space (
42
). On the other hand, it is obvious that many subtle restrictions apply to the
evolution of a native sequence/structure motif such as the HIV-1 polyA hairpin. For instance, it is obvious that the polyadenylation
signal AAUAAA cannot be modified. Thus, it seems possible to make many altered
forms of the polyA hairpin that support virus replication. By going via the
less fit mutants A and B to a revertant, we could have entered domains of
sequence space that were never accessed and tried out by the existing HIV-SIV viruses. In principle, it should be possible to use these revertants
in another selection round by introduction of other deleterious mutations and
selection for fast virus growth. In this way we can force the evolution of HIV-1 sequences progressively away from the wild-type towards novel forms that are distant in nucleotide sequence,
but near in terms of function. The
in vitro
analysis presented in this study may suggest that the possibilities for
evolving new functionally successful HIV-1 variants are immense.
We thank Alje van Dam for help in the culture experiments, Jan van Duin for
critical reading of the manuscript, Keith Peden for providing us with the HIV-1 molecular clone pLAI and members of the Berkhout laboratory for
constructive comments during the course of this work. This research was
supported by the Dutch Cancer Society (KWF) and the Dutch AIDS Foundation.
REFERENCES
Return