The cellular protein hnRNP A2/B1 enhances HIV-1 transcription by unfolding LTR promoter G-quadruplexes

G-quadruplexes are four-stranded conformations of nucleic acids that act as cellular epigenetic regulators. A dynamic G-quadruplex forming region in the HIV-1 LTR promoter represses HIV-1 transcription when in the folded conformation. This activity is enhanced by nucleolin, which induces and stabilizes the HIV-1 LTR G-quadruplexes. In this work by a combined pull-down/mass spectrometry approach, we consistently found hnRNP A2/B1 as an additional LTR-G-quadruplex interacting protein. Surface plasmon resonance confirmed G-quadruplex specificity over linear sequences and fluorescence resonance energy transfer analysis indicated that hnRNP A2/B1 is able to efficiently unfold the LTR G-quadruplexes. Evaluation of the thermal stability of the LTR G-quadruplexes in different-length oligonucleotides showed that the protein is fit to be most active in the LTR full-length environment. When hnRNP A2/B1 was silenced in cells, LTR activity decreased, indicating that the protein acts as a HIV-1 transcription activator. Our data highlight a tightly regulated control of transcription based on G-quadruplex folding/unfolding, which depends on interacting cellular proteins. These findings provide a deeper understanding of the viral transcription mechanism and may pave the way to the development of drugs effective against the integrated HIV-1, present both in actively and latently infected cells.

. (a) LTR-II + III + IV sequence of the G-rich LTR region comprising tracts that fold into three mutually exclusive LTR G4s, i.e. LTR-II, LTR-III and LTR-IV. These three-stacked tetrads G4s are indicated in brackets. G bases in G-tracts or involved in G-quadruplexes are shown in bold. (b) MS/MS spectrum of the precursor ion observed at m/z 1095.46 in the sample mixture from digestion of LTR-II + III + IV G4 sample (see Table 1). Only characteristic y ions are indicated 53 . The data match the sequence of peptide N326-R350 of hnRNP A2/B1, which is reported on top with the observed fragments. c) Pull-down assay of nuclear extract proteins with wt, mutant G4 LTR-II + III + IV (M) and random (R) sequences, immobilized on agarose beads. Shown is the western blot analysis with an anti-hnRNP A2/B1 antibody. Proteins complexed to the beadsbound LTRs were washed with augmented stringency by increasing the ionic strength of the wash buffer (0.2 and 1 M). The final elution was obtained in denaturing buffer at 95 °C.
Scientific RepoRts | 7:45244 | DOI: 10.1038/srep45244 mutant LTR-II + III + IV M4 + 5 was bound with a lower affinity (K D 35.5 ± 3.0 nM). Chi 2 and U-values were < 10 and < 15, respectively, indicating optimal data fitting 28,29 . In contrast, binding to a random oligonucleotide of the same length as LTR-II + III + IV was so low that in these conditions it was not possible to obtain a meaningful K D value (Fig. 2). hnRNP A2 unfolds the LTR G4s. We next set out to investigate the effect of hnRNP A2 binding to the LTR G4s. A Taq polymerase stop assay was performed on the LTR-II + III + IV template. In the presence of 100 mM K + , stop sites corresponding to formation of LTR-III and LTR-II were visible (compare lanes 1 and 2, Fig. 3a). No stop corresponding to LTR-IV was detected, as expected, since LTR-IV has been previously reported to form upon induction by G4 ligands 18 . Upon addition of hnRNP A2, both LTR-III and LTR-II stop sites showed an insignificant decrease (compare lanes 3 and 2, Fig. 3a and b). Similarly, no effect was induced by addition of K + and hnRNP A2 in the control LTR-II + III + IV M4 + 5 template lacking the possibility to form G4 (lanes 5-7, Fig. 3a).
As suggested by Xodo 30 , the apparent lack of effect in the presence of an unfolding protein could be due to the protein binding to the template sequence which would stimulate polymerase stop and mask G4 release at the same binding site. We thus switched to a dual-labelled system where oligonucleotide folding could be monitored by changes in fluorescence. Fluorescence resonance energy transfer (FRET) is a spectroscopic technique that provides information about structure and dynamics of nucleic acids folding. It involves a donor fluorophore in an excited state, the excitation energy of which can be transferred to a proximal acceptor chromophore. Given that the major determinant of FRET efficiency is the distance between the acceptor and the donor, fluorescence intensities depict nucleic acids folding states. Consequently, when annealed to its complementary sequence to form a double-stranded structure, the tested oligonucleotide would yield the maximum fluorescence intensity, while the G4 folded conformation would be the least fluorescent. In these conditions the measured fluorescence intensity allows to calculate the energy transfer (E) and the end-to-end distance (R) between the two fluorophores, and therefore the unfolding degree. The G4 folded LTR-II + III + IV sequence was characterized by R of ~38Ǻ, with E ~0.84. When the G4 structure was converted into the duplex conformation by addition of the complementary C-rich strand, the fluorophores were separated by ~150 Ǻ, and FRET was approximately 0. Treatment of LTR-II + III + IV G4 with hnRNP A2 increased fluorescence by 3-folds with respect to the free G4, and E decreased to 0.51 ( Fig. 4a and b, Supplementary Table S2). Considering that the complete unfolding of the G4 structure required Δ E = 0.72, the Δ E = 0.30 induced by hnRNP A2 reflected a 42% unfolding (Fig. 4c). The negative control protein BSA showed negligible unfolding activity in these conditions ( Fig. 4c and Supplementary Table S2). To test the statistical significance of the effect observed in the presence of the protein, we applied a model comparison approach, which allows to test the statistical role of an independent variable (in our case the hnRNP A2 protein) through the comparison of two models differing only in that variable. The test yielded F (2, 189) = 36.27 and P < 0.05 31 , indicating that the difference observed in the presence and absence of the protein was statistically significant. Analysis of the activity of hnRNP A2 was extended to shorter LTR sequences, i.e. LTR-III + IV, LTR-III and LTR-IV G4s, folded in 100 mM K + . The unfolding of LTR-III + IV and LTR-IV was similar but lower than that on the full-length sequence (31%); unfolding of LTR-III was very low (8.5%) (Fig. 4c 51 is the probability that the observed match is not a random event. The score is reported as − 10x log10(P) where P is the absolute probability. The mass of the putative peptides and of the 50 most intense fragment ions were used for the data base search. Significant Mascot hits were accepted as positive matches and their ion score reported. The highest score obtained in the three experiments is reported.
unfolding up to 83% was obtained in the LTR-II + III + IV oligonucleotide (Fig. 4c, Supplementary Table S1 and Supplementary Fig. S1). Because G4s are less stable at lower K + concentrations, we proposed that the diverse unfolding efficiency towards the different-length LTR G4s depended on the stability of the LTR G4s. To test this hypothesis, melting temperatures (T m ) of the dual-labelled oligonucleotides were thus measured by FRET-monitored thermal unfolding ( Table 2). The most stable sequence was LTR-III (T m 62.1 ± 0.1 °C), followed by LTR-IV and LTR-III + IV (T m 59.1 ± 0.1 °C and 55.2 ± 0.1 °C, respectively); LTR-II + III + IV was the least stable sequence (T m 49.1 ± 0.1), confirming that the observed unfolding scale depended on the oligonucleotide stability in the FRET system. To note that in the unlabelled oligonucleotides measured by circular dichroism (CD), LTR-IV was the least stable sequence (Table 2), bias which likely derives by the absence of fluorophores at the oligonucleotides' 5′ -and 3′ -end, which affect T m values, as previously noted 32 .
In addition, the longest sequences, i.e. LTR-II + III + IV and LTR-III + IV, where multiple G4s may form, display T m values that represent the average stability, which depends on the effective formation of the possible G4s. We thus performed a Taq polymerase stop assay to assess the G4 mutual formation and stability at increasing K + concentration in the full-length LTR. LTR-III was used as a control sequence where only one G4 could form. We observed that in the full-length LTR-II + III + IV sequence, both LTR-II and LTR-III G4s formed upon addition of K + (lanes 2-3, Fig. 5a); of these, LTR-III G4 induced a slightly more intense stop than LTR-II, suggesting a preferred formation of LTR-III in the full-length sequence (Fig. 5b). LTR-III was also the only G4 forming in the LTR-III + IV template (lanes 5-7, Fig. 5a). Interestingly, the LTR-III that formed in the full-length sequence was less intense than LTR-III forming in the shortest oligonucleotides (compare lanes 2-3, with 6-7 and 9-10, Fig. 5a and b). These data indicate that i) LTR-III is the most prominent G4 among the mutually exclusive LTR G4s and ii)  when LTR-III forms in the full-length LTR promoter, its stability maintains a level that is fully susceptible to hnRNP A2 processing.

Depletion of hnRNP A2/B1 decreases viral transcription.
We have shown that LTR G4 folding inhibits viral transcription, whereas when point mutations that disrupt G4s are introduced in the LTR promoter, an increase in promoter activity is evidenced 18 . Silencing of the LTR G4 folding/stabilizing protein nucleolin significantly increased promoter activity, indicating the inhibitory effect of nucleolin on LTR-driven transcription 22 .
To assess the effect of hnRNP A2/B1 binding/unfolding of the full-length LTR G4s, a set of anti-hnRNP A2/B1 siRNAs was used, which effectively silenced hnRNP A2/B1 expression in 293T cells (Fig. 6a). In these conditions, LTR-driven transcription decreased by 45% at the highest siRNA concentration in the wt LTR sequence (Fig. 6b). In contrast, the LTR sequence with two point mutations that disrupt G4 folding was only marginally affected by hnRNP A2/B1 depletion (Fig. 6b), indicating that the effect was G4-specific. In these conditions, hnRNP A2/ B1-silenced 293T cells showed no reduction in cell viability at 48 h with respect to the controls, thus confirming the consistency of the inhibitory effects on LTR promoter activity. Silencing of hnRNP A2/B1 was also performed in the TZM-bl reporter cell line, which contains stably integrated copies of the luciferase gene under control of the LTR wt promoter and thus closely resembles the integrated provirus. The LTR-driven reporter transcription decreased up to 33% in hnRNP A2/B1-silenced cells ( Supplementary Fig. S5). These data further confirm the unfolding activity of hnRNP A2/B1 on the LTR G4s.

Discussion
We have shown that hnRNP A2/B1 selectively binds the HIV-1 LTR G4s and unfolds them. While hnRNPs are RNA/protein complexes that bind newly synthesized RNA (pre-mRNA) in the cell nucleus during gene transcription and post-transcriptional modifications 33 , two of the most abundant of them, hnRNP A1 and hnRNP A2/B1, have been reported to bind also DNA in the quadruplex conformation. Initially hnRNP A1 was found to unfold the telomeric G4s and promote telomerase activity [34][35][36] , to unfold the G4 in the promoter of the KRAS oncogene and regulate transcription 30 and to destabilize the tetraplex forms of the fragile X expanded sequence d(CGG)n 37 . HnRNP A2 has been shown to unfold the fragile X repeats 38 and the shorter hnRNP A2 splice variant to promote telomere extension in mammalian cells 39 .
Our finding that hnRNP A2 binds and unfolds the G4s in the promoter of HIV-1 not only confirms the DNA G4 unfolding activity of this protein, but also indicates for the first time that this type of activity is exploited by a virus.
In HIV-1, hnRNP A2/B1 has been reported to play with Rev an important accessory role in promoting nuclear viral RNA retention and nucleocytoplasmic viral RNA transport 40 . In this context, the lower virus production observed upon siRNA-mediated depletion of hnRNP A2/B1 in infected cells was ascribed to accumulation of viral genomic RNA in cytoplasmic compartments or in the nucleus 41 . Our results showing decreased transcriptional activity upon depletion of hnRNP A2/B1 indicate a possible additional mechanism of inhibition of virus production mediated by increased G4 folding in the HIV-1 LTR promoter in the absence of hnRNP A2/B1. Surprisingly, overexpression of this protein and other hnRNP members induced similar effects, i.e. inhibition of transcription and reduction of virus production 42 . We suggest that upon stimuli that boost viral transcription (i.e. integral release of LTR G4s upon hnRNP A2/B1 massive overexpression) the virus (or the cell) activates counteracting mechanisms to avoid excessive exploitation of the cell by the virus that would lead to fast cell death, in the end impairing virus production. Therefore, the reported transcription inhibition might be mediated by other factors triggered by hnRNP A2/B1 overexpression.
The activity of hnRNP A2/B1 is fit to unfold the G4s that actually form in the HIV-1 LTR region. In fact, G4s folded in short oligonucleotides, such as LTR-III and LTR-IV, which displayed higher thermal stability, were only partially unfolded by the protein. In contrast, the same G4 structures folding in a longer context (LTR-II + III + IV) were less stable and were effectively processed by hnRNP A2/B1. Obviously, the longer oligonucleotide better mimes the actual G4 condition in the full-length LTR, as also proved by transcription inhibition obtained by hnRNP A2/B1 depletion in cells that contained the entire HIV-1 LTR. The LTR G4 system displays regulating features similar to those described for the c-myc 43,44 , HRAS 26 and KRAS oncogene-promoters 30,[45][46][47] . As in these eukaryotic G4-modulated promoters, the HIV-1 LTR promoter is processed by G4 stabilizing (nucleolin) 22 and destabilizing proteins (hnRNP A2/B1). Because the LTR promoter has been suggested to be the region where viral latency is regulated 48,49 , the G4 switch may play a role not only in activation of effective viral transcription, but also in its shift to latency. Currently, only the actively transcribed virus is targeted by antiviral drugs, while eradication of the HIV-1 infection has been made impossible by the existence of reservoirs of the latent virus 50 . Therefore, our findings not only advance our understanding on the mechanism of viral transcription but may also constitute a progress from a therapeutic point of view.   , where R 0 (Förster distance) is the distance at which energy transfer is 50% of the maximum value. Between FAM and TAMRA fluorophores, R 0 is assumed to be 50Ǻ 26 . F-test (F) and the probability (P) values were calculated using R statistical environment (v. 3.3.2) 31 . A conventional alpha = 0.05 was considered to evaluate the test significance.
Taq Polymerase Stop Assay. Taq polymerase stop assay was carried out as previously described 18,22 .
Briefly, the 5′ -end labelled primer was annealed to its template (Table S1) in lithium cacodylate buffer in the presence or absence of KCl (50-100 mM) and by heating at 95 °C for 5 min and gradually cooling to room temperature. Where specified, samples were incubated with 29 ng of human recombinant hnRNP A2 (Origene Technologies, Rockville, USA) at 37 °C for 15 min. Primer extension was conducted with 2 U of AmpliTaq Gold DNA polymerase (Applied Biosystem, Carlsbad, California, USA) at 37 °C for 30 min. Reactions were stopped by ethanol precipitation, primer extension products were separated on a 15% denaturing gel, and finally visualized by phosphorimaging (Typhoon FLA 9000). Circular Dichroism Spectroscopy. DNA oligonucleotides were diluted to a final concentration (4 μ M) in lithium cacodylate buffer (10 mM, pH 7.4) and KCl 100 mM. Samples were annealed by heating at 95 °C for 5 min and gradually cooled to room temperature overnight. CD spectra were recorded on a Chirascan-Plus (Applied Photophysisics, Leatherhead, UK) equipped with a Peltier temperature controller using a quartz cell of 5 mm optical path length, over a wavelength range of 230-320 nm. For the determination of T m , spectra were recorded over a temperature range of 20-90 °C, with temperature increase of 5 °C. The reported spectra are baseline-corrected for signal contributions due to the buffer. Observed ellipticities were converted to mean residue ellipticity (θ ) = deg × cm 2 × dmol −1 (mol ellip). T m values were calculated according to the van't Hoff equation, applied for a two-state transition from a folded to unfolded state, assuming that the heat capacity of the folded and unfolded states are equal 52 .