ABSTRACT
Transcription from a minimal HIV-1 promoter containing the three Sp1 binding sites and TATA box can be
activated without Tat by template DNA replication. Here we show that this
activation can also be mediated by recombinant GAL4 fusion proteins containing
the activation domains of Sp1, VP16 or CTF (or by full-length GAL4) targeted to the HIV-1 promoter by replacing the Sp1 sites with five GAL4 binding sites.
Thus Sp1 is not unique in its ability to mediate replication activated
transcription, although the degree of processivity elicited by the different
activators varied significantly from strongly processive (GAL4-VP16) to relatively non-processive (GAL4-Sp1 or -CTF). Processive GAL4-VP16-activated transcription, but not efficient
initiation, required multiple GAL4 binding sites. In the presence of Tat,
transcription with GAL4-SP1 and GAL4-CTF was further activated (principally at the level of processivity)
but GAL4-VP16-potentiated transcription was only slightly stimulated. The Tat-dependent switch from non-processive to fully processive transcription was
particularly marked for GAL4-Sp1, an effect which may be relevant to the selection of Sp1 binding sites
by the HIV-1 promoter.
Human immunodeficiency virus type 1 (HIV-1) relies on both cellular and viral factors for regulated and efficient
gene expression.
Cis
-acting DNA targets for cellular transcription factors and transcriptional
activators, including NF-[kappa]B, Sp1, TBP and LBP-1, are located in the enhancer and promoter regions of the
HIV-1 LTR (reviewed in refs
1
and
2
). However, high-level transcription from the HIV-1 promoter is normally dependent on further activation by the viral
trans
-activator protein, Tat, which increases the efficiency of transcriptional
elongation and initiation. These activities are mediated by binding of the Tat
protein to a
cis
-acting element, TAR, located near the 5' end of each viral primary transcript (reviewed in refs
2
and
3
).
When HIV-1 infects a host cell and integrates into its genome, the virus will
initially entirely depend on cellular transcription factors, since no Tat will
have been synthesised. Clearly, Tat-independent mechanisms of transcriptional activation would be advantageous
in this phase of infection, and might subsequently play a role in further
activating Tat-mediated transcription. It is now well established that transcriptional
activation by DNA replication can potentiate efficient expression from a number
of cellular and viral promoters. For example, the human [alpha]-globin (
4
-
6
), adenovirus major late (
7
) and adenovirus (minor late) IX promoters (
8
,
9
) all exhibit replication activation in transient assays when expressed from
replicating plasmids. Similarly, Patient and co-workers (
10
,
11
) have shown that
Xenopus
and human [beta]-globin, SV40 early, adenovirus E1A and retroviral LTR promoters are
activated by replication in the presence of linked enhancer elements.
Replication-mediated transcriptional activation may have important biological
functions. It has been suggested that cellular DNA replication, coupled to
chromatin restructuring
in vivo
, may be responsible for programming patterns of eukaryotic gene expression by
controlling the accessibility of the DNA to transcription factors (
12
). It is also likely that such an activation of viral late promoters is at least
partially responsible for the observed transition in gene expression (
13
) following adenovirus DNA replication; similar early-late transitions have been observed for other DNA viruses (e.g.,
polyomaviruses-see ref.
14
).
Many of the cells capable of supporting a productive HIV-1 infection have proliferative capacity-indeed, results from a recent study (
15
) suggest that viral replication
in vivo
is likely to be restricted to dividing cells (see Discussion). Therefore,
activation of transcription from an integrated provirus by cellular DNA
replication may have an important influence on viral gene expression. Work in
this (
6
) and other (
16
-
19
) laboratories has addressed the potential relevance of replication activation
to HIV-1 gene expression using transiently transfected viral promoter constructs
containing SV40 origins of replication. Replication from such origins (reviewed
in ref.
20
) is very similar to cellular DNA replication, occurring in nuclear
minichromosomes and using cellular machinery with the addition of a single
viral factor, SV40 large T-antigen (Tag). In each study (
6
,
16
-
19
), activation of transcription was observed from HIV-1 promoter plasmids with functional replication origins when Tag was
provided in
trans
. This effect did not require the presence of Tat. Moreover, activation was
shown (
6
) to be dependent on three binding sites for Sp1 adjacent to the TATA box in the
viral promoter, but independent of upstream sequences including NF-[kappa]B sites and other enhancer elements. This raised the question of
whether Sp1 has a unique function in replication activation of this promoter or
is simply acting as a general transcriptional activator. To examine further the
role of Sp1 in replication activation of a minimal HIV-1 promoter, we have set up a system which is isolated from endogenous
cellular Sp1 activity by removal of the Sp1 binding sites. In place of the Sp1
sites, we have inserted binding sites for the yeast
trans
-activator protein GAL4. This protein has no functional homologues in
mammalian cells, but is capable of targeted transcriptional activation in these
cells when expressed from a suitable vector (see, for example,
21
). Since GAL4 has separable DNA-binding and activation domains (
22
), fusion proteins can be expressed from recombinant constructs (
23
) which bind GAL4 sites but contain functional activation domains from
heterologous transcriptional activators. This system has allowed us to test
directly whether replication activation mediated by Sp1 relies on the
previously characterised transcriptional activation domains of the molecule,
and if the activation domains of other activators behave similarly to those of
Sp1. The system has also enabled us to investigate co-activation of a replicating HIV-1 promoter by Tat and GAL4 fusion proteins. We have therefore been
able to assess whether Tat is specialised to activate transcription in concert
with Sp1 (and DNA replication), or whether other activators can substitute for
Sp1.
Here we show that transcriptional activation of the HIV-1 promoter by DNA replication can be mediated in
trans
by an activation domain of Sp1 targeted to a minimal viral promoter by the GAL4
system. The activation domains of CTF, VP16 and GAL4 itself are also functional
in this assay, but elicit significantly different patterns of transcription
when the number of GAL4 binding sites in the promoter construct is varied.
Moreover, Sp1, CTF and VP16 differ in their abilities to co-activate transcription from a replicating template in tandem with Tat,
although each supports Tat function. We discuss possible mechanisms for these
effects and their potential relevance to HIV-1 gene expression and replication activation in general.
p[beta]E, a construct of P. Lamb that expresses the human [beta]-globin gene, has been described previously (
6
). R[beta]SVpBR328, here called pTag, is described in ref.
24
. It expresses both SV40 large T-antigen and rabbit [beta]-globin sequences.
The HIV-1 promoter constructs containing GAL4 binding sites were derived from a
variant of HIV/[alpha] pSVod (
6
), which contains the SV40
Hin
dIII-
Eco
RII 5171-159 fragment (including the origin of replication), and the
Sca
I-
Hin
fI LTR fragment of HIV-1 (NL4-3) immediately 5' to human [alpha]-globin sequences. The variant used here also had
a
Bgl
II linker inserted in an
Nco
I site within the SV40-derived sequence that has previously been shown not to affect plasmid
replication. An oligonucleotide (5'-CCTCGAGCCCTCAGATCCTGCATATAAGCAG-3') was synthesised and inserted into the HIV/[alpha] pSVod variant cleaved at unique sites for
Xmn
I (between the SV40 origin and promoter insert) and
Pvu
II (between the Sp1 sites and cap site within the promoter insert). This
generated a construct, HIV/[alpha] pSVod*, in which the Sp1 sites and other upstream HIV-1 promoter sequences were deleted and an
Xho
I site was created at the 5' end of the remaining HIV-1 sequence. Five tandem repeats of the GAL4 binding site
CGGAGTACTGTCCTCCG, a variant on the consensus containing an internal
Sca
I site (
21
), were excised as an
Eco
RI-
Xba
I fragment from pBK103 (a construct of Bernard Khoo which contains the five
binding sites cloned into pBlueScript SK+ as a
Pst
I-
Xba
I fragment). The excised fragment was end-filled and cloned into the end-filled
Xho
I site of HIV/[alpha] pSVod*. This produced a construct, pH[alpha]SVo
+
5G, containing an intact SV40 origin and five GAL4 binding sites in place of the
three Sp1 sites and upstream sequences of the HIV-1 LTR. Partial or full digestion of pH[alpha]SVo
+
5G with
Sca
I yielded pH[alpha]SVo
+
3G (three GAL4 binding sites) and pH[alpha]SVo
+
1G (one GAL4 site), respectively. Cleavage of each of these three constructs in
the SV40 origin region with
Sfi
I, followed by end-filling and ligation with a
Bgl
II linker (GAGATCTC) produced pH[alpha]SVo
-
5G, pH[alpha]SVo
-
3G and pH[alpha]SVo
-
1G, equivalent clones lacking functional origins of replication (
6
,
7
,
25
).
pGEMlLTR (constructed by Jan Eggermont) contains the
Ava
I-
Hin
fI HIV-1 (NL4-3) LTR fragment, end-filled and cloned into pGEM-7Zf(+) at the end-filled
Xba
I site, in the antisense orientation with respect to the T7 promoter. pGEMsLTR effectively contains the
Ava
I-
Sac
I LTR fragment of HIV-1 in the antisense orientation relative to the T7 promoter of pGEM-3. It was constructed by excision of a
Sac
I fragment from a variant of HIV-1 LTR HApSP65 (described in ref.
6
), in which the
Ava
I-
Hin
fI HIV-1 fragment was inserted in the
Sma
I site of the pSP65 polylinker in the sense orientation with respect to the Sp6
promoter. The
Sac
I fragment was cloned into the
Sac
I site of pGEM-3 and screened for orientation.
pSG4 expresses full-length GAL4 sequences (amino acids 1-881) and employs an SV40 early promoter and polyadenylation signals
(as in pSG424, ref.
23
). The other GAL4 fusion expression vectors were constructed and kindly provided
by Bernard Khoo. They are based on pHKG4, a derivative of pSG424 (
23
) constructed by C. Hagemeier and T. Kouzarides. In this clone the pBR322
plasmid backbone is replaced by ping14 vector sequence (with a
Bgl
II-
Eco
RI deletion). Both pHKG4 and pSG424 express the 1-147 aa domain of GAL4 and carry identical polylinkers, SV40 early
promoters/origins and polyadenylation signals. pHKG-CTF, pHKG-VP16 and pHKG-Sp1 (expressing the proline-rich activation domain of CTF, the acidic activation
domain of VP16, and the 516C[Delta]int112 serine/threonine-rich and glutamine-rich domains of Sp1-see ref.
26
) were derived from equivalent pSG424 constructs (kindly provided by N. Mermod,
ref.
27
) by excising the
Xba
I-
Xho
I fragment from each and cloning into similarly cut pHKG4.
pOGS213, which expresses a synthetic
tat
gene from the human cytomegalovirus immediate early promoter, is described in
ref.
28
.
The pH[alpha]SVo plasmids (10 [mu]g in Fig.
3
, 5 [mu]g in Figs
4
,
5
and
6
) were transfected with the appropriate GAL4 fusion expression vector (30 [mu]g in Fig.
3
, 15 [mu]g in Figs
4
,
5
and
6
) and p[beta]E (5 [mu]g, co-transfection control) and/or pTag (5 [mu]g, source of T-antigen/co-transfection control) and pOGS213 (3 [mu]g Tat expression vector, Fig.
6
only) into subconfluent HeLa cells by calcium phosphate co-precipitation without glycerol shock. HeLa cells were harvested after 48 h
and cytoplasmic RNA was purified as described previously (
5
). RNAs expressed from the co-transfection control plasmids p[beta]E and pTag were S1 mapped with
Eco
RI-cleaved, end-labelled probes made from these constructs. RNAs expressed from the
pH[alpha]SVo constructs were mapped by RNase protection assays (
29
) using T7 antisense riboprobes expressed from pGEMlLTR (linearised with
Eco
RI) or pGEMSLTR (linearised with
Hin
dIII) labelled with [[alpha]-
32
P]UTP or [[alpha]-
32
P]CTP and purified on Stratagene NucTrap columns. Hybridisations were carried
out at 56oC for 12-16 h and RNase digestion at 22oC for 30 min. Reproducibility of the patterns of expression
observed was confirmed in at least three independent RNase protection assays.
Results from a representative assay, with accompanying phosphorimage
quantitations (Molecular Dynamics PhosphorImager SF/ImageQuaNT), are presented
for each experiment.
Whole cell protein extracts were prepared in parallel with cytoplasmic RNA. Half
the volume of harvested cell suspension in PBS was pelleted and the pellet
washed in 40 mM Tris-HCl pH 7.4; 1 mM EDTA; 150 mM NaCl. The cells were re-pelleted and resuspended in 150 [mu]l of extract buffer (40 mM HEPES pH 7.9, 0.4 M KCl, 1 mM DTT,
10% glycerol; 0.1 mM phenylmethylsulphonyl fluoride, 0.1% Approtinin). The
extract was freeze-thawed three times, the debris was pelleted in a microcentrifuge for 5 min
at 14 000 r.p.m., 4oC, and the supernatant was stored at -70oC until required. Protein interactions were assayed by gel
mobility shift analysis as previously described (
30
), using an end-labelled GAL4 binding site DNA fragment.
To investigate replication activation of the HIV-1 promoter, we have employed a modified version of the HIV-1/[alpha]-globin gene system described previously (
6
). As shown in Figure
1
, the basic construct used, pH[alpha]SVo
+
5G, contains the SV40 origin of DNA replication (without enhancer) separated
from the HIV-1 promoter by 400 bp of vector sequence. The promoter is deleted up to -46 bp from the CAP site which removes the HIV-1 enhancer and three Sp1 binding sites but still retains the
TATA box. Five binding sites for the yeast transcription factor GAL4 were then
inserted at this deletion end point, adjacent to the TATA box. HIV-1 sequence downstream of the Cap site extends through the TAR region and
is then joined to the human [alpha]-globin gene from its first intron through to the [alpha]-globin poly(A) site. The 5' ends of transcripts expressed from this
construct were detected by RNase protection analysis using two different
riboprobes which detect either the first 37 nt of the HIV-1 transcript (sLTR probe) or all of the HIV-1 transcript up to the [alpha]-globin gene sequence (lLTR probe). DNA replication of the pH[alpha]SVo
+
5G plasmid following its transfection into HeLa Cells requires SV40 large T-antigen (Tag) (see ref.
6
). This was supplied in
trans
by co-transfection with pTag, a construct that expresses both Tag and rabbit [beta]-globin sequences. Transfection efficiencies were monitored by
measuring levels of rabbit [beta]-globin transcripts (as in Figs
4
and
5
), or (as in Fig.
3
) by co-transfection with a third plasmid, p[beta]E, which expresses the human [beta]-globin gene (transcript levels of which were similarly
measured).
An unusual feature of the HIV-1 promoter is that it is capable of synthesising two types of transcript.
In the presence of the viral encoded transactivator Tat, HIV-1 transcription is highly processive, reading through the whole viral
genome. However, in its absence transcripts are much less processive, resulting
in the formation of heterogeneous steady state RNAs extending over only the
first 100-200 nt of the coding sequence (see ref.
2
for review). These RNAs are further degraded to a short (~60 nt) RNA product corresponding to the nuclease resistant TAR hairpin loop
at the beginning of the viral transcript. As shown in Figure
3
b, we observe this same heterogeneous transcription pattern from our HIV-1/[alpha]-globin gene construct using the longer lLTR riboprobe. However we also note that the ratio of
transcripts that read through the HIV-1 5'-terminal region to yield full length RNAs (breaking up into
three major bands due to variable digestion with the RNases used in the RNase
protection experiment) as compared to shorter RNAs that accumulate at ~60 nt (over the TAR region) significantly changes with the different
activation domains. Quantitation of the ratio of full to short length RNAs
reveals that Sp1 stimulates non-processive transcription while VP16 produces mainly full length
transcripts. These results are consistent with recent studies (
33
,
34
) which show that different transcription factors may have differential effects
on transcriptional initiation and elongation. We also show that both the CTF
activation domain and the residual activation properties of the GAL4 binding
domain alone stimulate non-processive transcription. In Figure
4
b we confirm these results although the degree of ratio change between the Sp1,
CTF and VP16 activation domains is slightly reduced. The strongly activating
full length GAL4 transcription factor (SG4) gives an intermediate ratio of
processive to non-processive HIV-1 transcription.
As described above, previous work in this laboratory has demonstrated that three
upstream Sp1 sites potentiate replication activation of the HIV-1 promoter (
6
). Here it was also shown that two Sp1 sites, or even a single site, were
sufficient to stimulate HIV-1 transcription to some degree. Levels of stimulation, however, were
significantly higher when three binding sites were employed. To determine
whether the HIV-1/[alpha]-globin gene construct used here behaves similarly, we
constructed variants of pH[alpha]SVo
+
5G and pH[alpha]SVo
-
5G that contained three GAL4 binding sites (pH[alpha]SVo
+
3G and pH[alpha]SVo
-
3G) or a single site (pH[alpha]SVo
+
1G and pH[alpha]SVo
-
1G). Transfections and RNA analyses were carried out as described above. Results
obtained using pH[alpha]SVo
+
3G and pH[alpha]SVo
-
3G were similar to those from the experiments with pH[alpha]SVo
+
5G and pH[alpha]SVo
-
5G (data not shown). However, the results obtained with pH[alpha]SVo
+
1G and pH[alpha]SVo
-
1G were markedly different. Figure
5
shows data from a 1-GAL4 binding site experiment conducted in parallel with the 5-GAL4 binding site experiment shown in Figure
4
. Use of the single GAL4 site in place of the five sites led to a general
reduction in the degree of replication activation, with some notable
differential effects between activators. Most strikingly, the level of
replication activation mediated by full-length GAL4 was reduced to lower levels than with GAL4-Sp1; similarly GAL4-CTF activation was reduced to very low levels. Significantly,
although GAL4-VP16 was still capable of mediating replication activation to high levels,
the ratio of full-length to short transcripts detected by the lLTR probe was markedly lower than observed before, suggesting that
multiple GAL4 binding sites are required for effects on transcriptional
elongation but are less important for initiation (see Discussion).
Although DNA replication may have an important role in activating early viral
transcription before significant amounts of Tat are synthesised, it is equally
possible that such activation continues to function at a later stage in
combination with
trans
-activation by Tat. We therefore wished to investigate how the two
activation mechanisms interact. In particular we were interested in determining
whether Sp1 is uniquely suitable for co-operation with Tat, since previous studies in non-replicating systems (
35
-
39
) have reached different conclusions on its necessity.
As shown in Figure
6
, the minimal HIV-1 promoter construct requires a functional origin of replication for
significant expression even in the presence of Tat and each of the three
activators tested. Transcription was detectable (using a longer radioautograph
exposure than shown) from a non-replicating template when activated by Tat plus GAL4-Sp1, -VP16 or -CTF, but levels were >10-fold lower than those obtained with a replicating
promoter, reflecting the large effect that replication activation has in this
system. It should be noted, however, that intermediate levels of transcription
(see ref.
6
) can be obtained from a non-replicating promoter carrying wild-type Sp1 sites which allow more efficient activation by endogenous
non-chimaeric Sp1.
When the levels of transcription from a replicating template with and without
Tat are compared, there are clear differences between the activators, although
each supports Tat function. Transcription activated by Sp1, or to a lesser
extent by CTF, becomes markedly more processive in the presence of Tat (8.7- or 2.7-fold increase, respectively). However, transcription activated by
VP16, in which elongation is already highly efficient, becomes only marginally
(1.2-fold) more processive in the presence of Tat. Similarly, total levels of
transcription, and hence levels of initiation, are significantly increased by
Tat when Sp1 or CTF is the activator (4-6-fold total increase), whereas in the case of VP16 this increase is
slight. In general, the pattern of transcription activated by Sp1 or CTF in the
presence of Tat is similar to that obtained with VP16 with or without Tat.
In this paper we have shown that the chimaeric transcription factors GAL4-Sp1, GAL4-VP16 and GAL4-CTF as well as intact GAL4, when targeted to GAL4 binding
sites upstream of the minimal HIV-1 promoter, potentiate transcriptional activation by DNA replication in
transient assays. The precise patterns of transcription obtained with each of
the four activators varied significantly, especially when the number of GAL4
binding sites was reduced from five to one. GAL4-Sp1, GAL4-VP16 and GAL4-CTF were also tested for their ability to co-operate with Tat in
trans
-activating a replicating promoter; each was able to support Tat function
to some extent, but only transcription activated by GAL4-Sp1 and GAL4-CTF underwent significant further activation when Tat was added. In
particular, Tat induced a dramatic shift from non-processive to processive transcription from a promoter activated by GAL4-Sp1.
The ability of the GAL4-Sp1 construct to mediate replication activation is consistent with and
extends the finding (
6
) that native and foreign Sp1 sites confer responsiveness to replication
activation on HIV-1 and [beta]-globin promoters. The present results strongly suggest that
the previously observed requirement for Sp1 sites is a consequence of the need
to recruit endogenous Sp1 to the promoter to elicit a functional response,
rather than a direct
cis
-acting effect of the GC-rich binding sites. It is also clear from our assays with GAL4-VP16, GAL4-CTF and GAL4 that Sp1 has no unique activity required
for potentiating replication activation. Indeed, the simplest interpretation of
our results is that conventional mechanisms of transcriptional activation,
mediated by well-defined activation domains, are operating in tandem with DNA replication
to induce relatively high levels of expression from the HIV-1 promoter in the absence of Tat. Although this does not invalidate the
previous suggestion (
6
) that the (Sp1)
n
-TATA region is important for activating transcription from replicating
templates, it is clear that promoters containing binding sites for other
transcriptional activators may also be responsive to replication activation.
Patient and co-workers, whose initial work demonstrated a replication activation effect
on [beta]-globin gene transcription that required the SV40 enhancer (
10
), have more recently shown that a similar effect can be obtained using
multimerised binding sites for specific factors (e.g., octamer motifs) in place
of the enhancer (
11
). Taken together with these results, our data may reflect a general mechanism
by which DNA replication can activate transcription in concert with a variety
of
trans
-acting factors which bind either at distal (enhancer) or proximal
(promoter) sites. Interestingly, transcriptional activators including GAL4 and
GAL4-VP16 have been shown to be functional when bound at remote locations (
40
).
The pattern of GAL4-VP16 activated transcription from the replicating HIV-1 promoter construct containing five GAL4 sites is consistently
different from that obtained with the other activators. The ratio of full-length to short, truncated transcripts is markedly higher with GAL4-VP16, suggesting that GAL4-VP16 action involves effects on both transcriptional
initiation and elongation. Indeed, Yankulov
et al
. (
33
) have found that GAL4-VP16, when targeted to various promoters by multiple GAL4 binding sites,
stimulates elongation as well as initiation in
Xenopus
nuclei and 293 cells. Comparable results for the c-
myc
promoter have recently been published by Krumm
et al
. (
34
). Surprisingly, we observed a different pattern of expression when GAL4-VP16 was used to activate the promoter construct containing only one GAL4
binding site. Although high levels of transcriptional activation were again
observed, the ratio of full-length to short transcripts dropped markedly. This suggests that
interaction with one site is sufficient to stimulate initiation but not
elongation. It is possible that some form of synergistic interaction between
activators that enhances elongation occurs when multiple binding sites are
available (perhaps by allowing multiple contacts with both initiation and
elongation factors). The synergistic activity of GAL4-VP16 when targeted to multiple binding sites has been reported previously
(
32
,
41
-
43
), although in this case it was reduced by replication (
43
). Similarly, it has been shown (
32
) that a GAL4 derivative bearing a single GAL4 activation domain activates
transcription >300-fold more efficiently when targeted to five binding sites as it does when
targeted to one. The CAT and luciferase reporter gene assays used in these
experiments did not distinguish between activator effects on initiation and
elongation. It is therefore possible that differences in processivity underlie
the observed synergistic effects.
The mechanism by which DNA replication stimulates transcriptional activation is
unclear. Previous work in this laboratory (
6
) has established that SV40 origin-containing plasmids do not replicate efficiently in HeLa cells under
identical experimental conditions to those used here, ruling out a simple copy
number effect (others have reached similar conclusions using HeLa and COS cells-see refs
7
-
11
,
18
). Nevertheless, low level replication appears to produce a transcriptional
template that is more responsive to activators. For reasons discussed
previously (
6
) we do not consider a mechanism based on differences in DNA methylation to be
likely, although it remains possible that the subnuclear position or
conformation of newly-replicated plasmid DNA favours its transcription if activators are
available. The unwinding of the DNA duplex which occurs in response to Tag
binding (
25
) may also be a factor. However, an alternative explanation is possible.
Plasmids containing SV40 origins, when transfected into mammalian cells, are
known to be assembled into chromatin structures which are restructured when the
DNA is replicated (
44
). Nucleosomes compete with transcription factors for access to DNA (reviewed in
ref.
45
) but DNA replication causes a transient disruption in chromatin structure
followed by a staged reassembly process. This may allow ready access of
transcription factors to their binding sites (even after core histones have
bound); nucleosome maturation progressively blocks such interactions (
12
). Furthermore, Kamakaka
et al
. (
46
) have found
in vitro
evidence for a replication-dependent chromatin reconfiguration mechanism for transcriptional
activation by GAL4-VP16. Such an explanation is plausible for the episomal replication
activation described here, and may also be applicable to activation of
integrated viral DNA.
The final experiments described in this study indicate that replication
activation mediated by each of the three activators tested can co-operate with the viral
trans
-activator, Tat. Differences in the patterns of co-activation observed may have relevance to the question of why Sp1
was `selected' as a transcriptional activator by HIV-1. Previous analyses in non-replicating systems have reached varying conclusions on whether Sp1
is an essential co-activator for Tat. Jeang and co-workers (
35
,
36
) have found that activators such as Oct-1 and NF-[kappa]B are poor substitutes for Sp1 in co-operating with Tat function. Co-activation ability was found to correlate with
the strength of physical interaction between the activator and Tat. One recent
study, using a heterologous chimaeric promoter (
37
) has defined further activators that co-operate poorly with Tat (Bel-1, E1a and Tax) but has also concluded that others, including VP16
and CTF, synergise with Tat at least as well as Sp1. In contrast, two other
groups (
38
,
39
) have found that synergy is much greater between Sp1 and Tat than it is between
VP16 and Tat; adding Tat did not increase levels of expression much beyond
those obtained with VP16 alone. The latter results are consistent with those
obtained in our replicating system. Moreover, unlike the previous studies which
relied on chloramphenicol acetyl transferase reporter gene assays (
35
-
39
) we have made direct comparisons between the abilities of Sp1, VP16 and CTF to
co-operate with Tat by RNA analysis. From this analysis, we can conclude that
a major reason for the inability of VP16 to synergise with Tat to the same
extent as Sp1 is its different mechanism of action (although we also note that
VP16-activated transcription appears to be less sensitive to stimulation of
transcriptional initiation by Tat than Sp1 or CTF-activated transcription). In our system, transcription activated by VP16
even in the absence of Tat is already highly processive. Since a major function
of Tat is to increase elongation efficiency, further activation by Tat has only
a limited effect. In contrast, RNA transcribed in the presence of Sp1
(especially) or CTF elongates relatively inefficiently, so additional effects
of Tat are more significant. Indeed, the Sp1+Tat and CTF+Tat phenotypes are
similar to that obtained with VP16 (irrespective of Tat). We suggest that the
selection of a `two-stage' activation mechanism involving a cellular activator (Sp1) that
operates primarily on transcriptional initiation and synergises with a viral
regulator of transcript elongation (Tat) may have regulatory advantages for the
virus. A two-component regulatory system may allow more flexible control of gene
expression than activation by a factor such as VP16 that strongly affects both
initiation and elongation. Sp1-mediated replication activation may initially induce an `early' pattern of
transcription balanced between full-length and short transcripts which progresses to a `late' processive
pattern when adequate amounts of Tat are synthesised. Alternatively, the virus
may simply have evolved Tat as a regulatory protein that complements an
existing cellular activator, rather than duplicating all its functions.
To what extent DNA replication influences HIV-1 transcription
in vivo
is as yet unknown. From the present results it is only possible to say that the
viral promoter is certainly responsive to replication activation, and that this
process could potentially activate HIV-1 gene expression in infections of proliferating cells. Since HIV-1 is capable of productively infecting macrophages (usually regarded
as non-dividing) and growth-arrested cells (
47
), replication activation may only be relevant to a subgroup of cellular
infections. However, in a recent study (
15
) it was concluded that productive infection of macrophages requires cellular
factors only present in a small proportion of these cells with proliferative
capacity, suggesting that HIV-1 infection
in vivo
is restricted to dividing cells. Thus replication activation could have a more
general role in stimulating viral gene expression. In either case, studies of
integrated viral constructs or provirus should provide a clearer picture of the
relevance of this type of activation to HIV-1, while analyses of chromatin structure may provide more insight into its
molecular mechanism.
R.D.W, B.A.L. and N.J.P. were supported by Medical Research Council grant
9313072. S.P.J. was supported by grant SP 2143/0101 from the Cancer Research
Campaign (UK).
REFERENCES
Return