GC-elements controlling HRAS transcription form i-motif structures unfolded by heterogeneous ribonucleoprotein particle A1

HRAS is regulated by two neighbouring quadruplex-forming GC-elements (hras-1 and hras-2), located upstream of the major transcription start sites (doi: 10.1093/nar/gku 5784). In this study we demonstrate that the C-rich strands of hras-1 and hras-2 fold into i-motif conformations (iMs) characterized under crowding conditions (PEG-300, 40% w/v) by semi-transitions at pH 6.3 and 6.7, respectively. Nondenaturing PAGE shows that the HRAS C-rich sequences migrate at both pH 5 and 7 as folded intramolecular structures. Chromatin immunoprecipitation shows that hnRNP A1 is associated under in vivo conditions to the GC-elements, while EMSA proves that hnRNP A1 binds tightly to the iMs. FRET and CD show that hnRNP A1 unfolds the iM structures upon binding. Furthermore, when hnRNP A1 is knocked out in T24 bladder cancer cells by a specific shRNA, the HRAS transcript level drops to 44 ± 5% of the control, suggesting that hnRNP A1 is necessary for gene activation. The sequestration by decoy oligonucleotides of the proteins (hnRNP A1 and others) binding to the HRAS iMs causes a significant inhibition of HRAS transcription. All these outcomes suggest that HRAS is regulated by a G-quadruplex/i-motif switch interacting with proteins that recognize non B-DNA conformations.


Results and Discussion
The sequence of the HRAS promoter immediately upstream of the major transcription start sites is reported in Fig. 1A. It contains two GC-rich elements, hras-1 and hras-2, composed of blocks of guanines and capable of folding into G-quadruplex structures. In previous works we have demonstrated that these sequences behave as a regulatory switch controlling gene expression 3,4 . Such a mechanism has been proposed for other relevant oncogenes including KRAS 17,18 , CKIT 19,20 , and CMYC 21,22 . A couple of comprehensive reviews on this subject have been reported 23,24 . In this work we have focused on the complementary C-rich strands hras-1 Y and hras-2 Y and have investigated if they fold into stable iMs. iM formation by the HRAS C-rich sequences. To find out if hras-1 Y and hras-2 Y can assume the iM conformation, we performed circular dichroism (CD) experiments as a function of pH, in 50 mM KCl, 50 mM Tris-acetate, 25 °C. To mimic the crowding conditions of the cell, we analysed the sequences both in the presence and absence of 40% (w/v) PEG-300 16 . Typical CD titrations are shown in Fig. 1B,C. It can be seen that the spectra of hras-1 Y and hras-2 Y change dramatically as the pH is gradually decreased from 8 to 4.5. Under acidic conditions (pH 5) both sequences exhibit the characteristic enhanced ellipticity at ~287 nm of a classical iM 10,12,[25][26][27] , while at pH 8 the sequences exhibit a much lower ellipticity, shifted at ~285 nm. By plotting the 287-nm ellipticity as a function of pH, we obtained for each sequence, in the presence or absence of PEG-300, sigmoidal curves reflecting iM formation (Fig. 1D). The crowding agent drives the folding at higher pH values: the semi-transition of hras-1 Y increases from pH 5.9 to 6.3, whereas that of hras-2 Y increases from pH 6.2 to 6.7. These plots suggest that the iMs are stable in a slightly acidic medium. However, under cellular conditions the iM can be stabilized by: (i) transcriptionally induced DNA superhelicity 28,29 ; (ii) more effective cellular crowding conditions 30,31 ; (iii) an increased intracellular acidity generated by an increase of the glucose-lactate flux 32,33 .
As the ellipticity-versus-pH curves are reversible, we evaluated the number of protons involved in the folding of hras-1 Y and hras-2 Y , by considering the following equilibria: where equilibrium 1 takes into consideration the fact that the unfolded C-rich sequence, U, may bind m protons before folding into the i-motif (as the pKa of cytosine is ∼ 4.4 34 , at pH 5 about 3 cytosines out of 12 are expected to be protonated); U*·mH + is the unfolded sequence with m bound protons. The bases in U*·mH + are assumed to be unstacked, so the formation of this species is accompanied by a negligible change in the CD spectrum 35 . In contrast, upon folding into i-motif F, U*·mH + assumes in a cooperative manner other n protons to form (n+ m) CH + :C base pairs that are stacked in the structure. A significant CD spectral change is expected for the formation of the iM 8,10 (equilibrium 2), whose equilibrium constant is  (Fig. 1E). We obtained values of n ∼ 4 and n ∼ 2 for hras-1 Y and hras-2 Y , respectively. This suggests that when hras-1 Y and hras-2 Y fold into the iM, they assume 4 and 2 protons, respectively, which agrees with the fact that the sequences are partially protonated before folding. A similar behaviour has been previously observed for the formation of the i-motif by (C 3 TA 2 ) 4 14 .
PAGE and analyses of melting curves. The intramolecular iM, being a folded structure, migrates in a polyacrylamide gel faster than an unfolded oligonucleotide of the same length. We analysed hras-1 Y and hras-2 Y by PAGE under different pH conditions. The mobility of the two C-rich sequences was compared with that of hras-1 Y variants: ODN-1 (unable to form any structure); ODN-2 (forming an iM with 4 CH + :C) and ODN-3 (forming a stable W.C. hairpin) ( Fig. 2A,B) (Methods). Under denaturing conditions (7 M urea), the 27-mer oligonucleotides (hras-1 Y and variants) exhibited the same mobility, with the exception of ODN-3 that forms a hairpin even in the presence of 7 M urea. Sequence hras-2 Y , being embedded in a 36-mer oligonucleotide, migrates slowly (Methods). In contrast, under native conditions at pH 5, hras-1 Y , that assumes the iM according to CD, migrates with a sharp band faster than that of ODN-1. Interestingly, when the hras-1 Y sequence is modified to fold into a hairpin stabilized by a stem of 8 W.C. bps (ODN-3) (S 1 ), it migrates even quicker than the iM. This is because the 6 positive charges of iM reduce its negative charge density. The folding of hras-1 Y into the iM is quite fast, as the mobility does not change when the sample is heated before loading. Variant ODN-2, forming an iM with 4 positive charges (S 1 ), migrates slightly faster than hras-1 Y , as expected. At pH 5, the 36-mer oligonucleotide containing sequence hras-2 Y migrates quicker than unstructured 27-mer ODN-1, as it folds into the iM (see CD). At pH 7, both hras-1 Y and hras-2 Y still migrate faster than ODN-1, indicating that even at neutral pH the sequences are folded. However, they migrate either with a smeared band (hras-1 Y ) or with two bands (hras-2 Y ), suggesting that more than one folded structure is formed: most likely iM-like structures stabilized by C:C and CH + :C bps. It is also possible that at pH 7, the iM-like structure is in equilibrium with a flexible hairpin (Fig. 2C, S 1 ). We have also examined the thermal stability of the iMs by CD-and FRET-melting experiments. Typical CD-melting profiles for hras-1 Y and hras-2 Y are reported in Fig. 3A,B. By heating (20 → 85 °C) and then cooling (85 → 20 °C) the DNA solutions in 50 mM sodium cacodylate, pH 5, 50 mM KCl, the folded/unfolded transitions showed to be cooperative and reversible, as previously found for C-rich oligonucleotides under similar experimental conditions 10,[12][13][14] . The melting in the pH range between 5 and 7 was examined by FRET experiments, using oligonucleotides end-labelled with ATTO-488 (5′ -end) and TAMRA (3′ -end) (Fig. 3C,D). The T M of both hras-1 Y and hras-2 Y decreased with increasing pH: hras-1 Y , 55.9, 52.1, 47.6, 44.8, 41.3 °C at pH 5, 5.5, 6, 6.5, 7, respectively; hras-2 Y , 65.5, 58.7, 55.1, 53.8, 53.1 °C at pH 5, 5.5, 6, 6.5, 7. At pH values > 5, the T M of the iMs decreases as the structure is probably stabilized by a number of CH + :C < 6. It is indeed reasonable to assume that when the medium is not sufficiently acidic, the iM is stabilized by both C:C and CH + :C base pairs 36,37 . iM structures of different protonation levels may coexist in solution. Mechanical stability experiments with the ILPR C-rich sequence showed that partially folded iM-like species are in equilibrium with fully folded iM at neutral pH 36 . The higher T M of hras-2 Y is likely due to the additional CH + :C bp that stabilizes the iM. The two structures have similar rupture forces and sufficient stability to stall RNA polymerase 36 . Yang and Rodgers have reported that the energy of C:C is about 1/3 of that of CH + :C 38 . Sequence hras-2 Y shows a behaviour similar to hras-1 Y , with the difference that at pH 5 it shows a higher T M , 65.5 °C (Fig. 3D).
We evaluated the thermodynamics of the folding transitions according to a two-state model. This was done at pH 5, where the two sequences fold into only one structure, as shown by PAGE. From the CD-and FRET-melting curves at pH 5, we obtained the following average thermodynamic data (± 10%): Δ H = 252 kJ/mol and Δ S = − 770 J/mol K and Δ G = − 17 kJ/mol for hras-1 Y ; Δ H = 323 kJ/mol and Δ S = − 950 J/mol K and Δ G = − 27 kJ/mol for hras-2 Y . Assuming that the breaking of a CH + :C bp needs approximately 46 ± 4 kJ/mol 12 , the number of CH + :C broken by the thermal disruption of hras-1 Y is ~6, of hras-2 Y is ~7, in accord with the number of expected protons that should bind to the sequences at pH near pKa 10,12 . At pH 5, nearly half of the cytosines is protonated and the sequences are completely folded into iM, showing the highest stability.
The HRAS iMs are recognized by hnRNP A1. As the iM-forming sequences overlap critical GC-elements immediately upstream of TSS, we interrogated if these unusual structures are recognized by nuclear proteins. Previous studies have reported that DNA sequences composed by runs of cytosines such as the C-repeats in the telomeres and in the CMYC promoter are recognized by proteins of the heterogeneous nuclear riboproteins family (hnRNP) 39,40 Moreover, Hurley and co-workers recently reported that the iM formed in the BCL2 promoter interacts with hnRNP LL 6 . Specific binding of heterogeneous ribonuclear proteins to C-rich DNA sequences is also supported by previous work, according to which hnRNP A1 binds to the GC-element of KRAS 17 , which shows a high sequence/functional homology with the HRAS GC-elements. HnRNP A1 is one of the most abundant nuclear proteins of eukaryotic cells that regulates several aspects of mRNA biogenesis 41 . As it is over-expressed in a variety of cancers 41,42 , we wondered if this protein plays a role in the promoter of the HRAS oncogene, in the region where the iM can potentially be formed. To address this question, we first investigated by chromatin immunoprecipitation (ChIP) if in HRAS-mutant T24 bladder cancer cells, hnRNP A1 is associated to the GC-elements under in vivo conditions. The occupancy of hras-1 and hras-2 (located 6 bp upstream of first TSS) by hnRNP A1 was compared with the occupancy of a reference GC-rich sequence unable to fold into a non-B DNA structure (located 870 bp downstream from first TSS). A typical ChIP is shown in Fig. 4A. We found that the occupancy of hras-1 and hras-2 by hnRNP A1 was, respectively, ~6-and ~5-fold higher than the occupancy by IgG (negative control). As hras-1 and hras-2 are located in the region of the major transcription start sites, they show a significant occupancy by RNA Pol II: ~4-fold higher than the IgG signal. In contrast, the reference sequence showed almost no occupancy by any of the proteins considered. The ChIP data provided strong evidence that hnRNP A1, under in vivo conditions, is indeed associated to the critical GC-elements of the HRAS promoter. However, ChIP data do not provide information about the conformation of the GC-elements interacting with hnRNP A1. To know if the nuclear factor recognizes the iM, we performed EMSA at pH 5.5 and 20 °C of mixtures composed by hras-1 Y or hras-2 Y and recombinant hnRNP A1, which was produced with a high degree of purity (Fig. 4B,C). It can be seen that at pH 5.5 hras-1 Y and hras-2 Y (which are in the folded iM conformation) form with hnRNP A1 a retarded band due to a 1:1 DNA-protein complex. In the presence of 4 μ g hnRNP A1, the iM is completely bound to the protein (Fig. 4B). With higher protein amounts, a second retarded band of much weaker intensity, probably a 1:2 complex, can be seen in the gel. When hnRNP A1 was thermally denatured before being added to the iM, the DNA-protein complex was abrogated and the iM migrated as a free molecule. As a further control, we used an unspecific protein like BSA and we found that it did not bind to the iM, as expected (Fig. 4C). When the iMs was destabilized by replacing 4 C with 4 T (hras-1 Y (m)) (see Methods), the binding was significantly attenuated, suggesting that the iM conformation is essential for optimal hnRNP A1 binding. hnRNP A1 unfolds the HRAS i-motif. In a previous work we have found that the HRAS promoter is highly active when the GC-elements are unfolded in the double-stranded conformation 3,4 . We now asked if the binding of hnRNP A1 to the iM involves the unfolding of this non B-DNA structure. To this purpose, we performed FRET experiments with hras-1 Y tagged with ATTO-488 (donor) and TAMRA (acceptor) in 50 mM sodium cacodylate, pH 5.5, 50 mM KCl. By exciting the donor at 480 nm, both donor (520 nm) and acceptor (580 nm) emit fluorescence, as a result of FRET between the fluorophores (Fig. 5A). When hras-1 Y is folded into the iM, the energy transfer, E T , between the two fluorophores is 0.77, and their end-to-end distance is ∼ 40 Å (S 2 ). The effect of hnRNP A1 on the iM was investigated by incubating protein and DNA for 1.5 h and measuring the fluorescence between 500 and 650 nm, upon donor excitation at 480 nm. It can be seen that hnRNP A1 causes a dramatic increase of the donor emission, accompanied by a decrease of E T as a function of r (r = [protein]/[iM]), in a dose-dependent manner, from 0.77 (r = 0) to 0.18 (r = 7) (inset). This means that the end-to-end distance in the iM increases from ~40 to ~64 Å, suggesting that upon binding to the protein, hras-1 Y goes through a conformational change. When hras-1 Y is in the duplex conformation, the fluorophores are separated by ~86 Å (26 × 0.33 = 86 Å, where 0.33 Å is the vertical rise per bp), as a 27-mer duplex behaves as an extended rigid rod. It follows that the iM bound to hnRNP A1 is not fully extended as in the duplex. As a control we used an unspecific protein as BSA and heated hnRNP A1 before incubation with the iM (20 min, 95 °C). In both cases the fluorescence of the donor did not increase, as expected (Fig. 5A).
To further support the finding that hnRNP A1 disrupts the iM, we carried out FRET-melting experiments, reasoning that the iM would not give its typical melting profile when bound to hnRNP A1. Fig. 5B shows that hras-1 Y iM in 50 mM sodium cacodylate, pH 5.5, 50 mM KCl has a T M ~ 50 °C. In the presence of 4 equivalents (r = 4) of BSA or denatured hnRNP A1, the melting profile of hras-1 Y did not change, as expected. In contrast, when the hras-1 Y iM was incubated with increasing amounts of native hnRNP A1 (r = 1-7), its melting profile strongly changed, in keeping with the fact that hras-1 Y bound to the protein is not in the iM conformation. It is worth noting that free hras-1 Y melts with an increasing sigmoidal curve, whereas hras-1 Y bound to hnRNP A1 melts with a decreasing sigmoidal curve (Fig. 5B). A melting profile similar to that of the hras-1 Y :hnRNP A1 complex is obtained with hras-1 Y in the duplex conformation with its complementary strand, where the two fluorophores are separated by ~86 Å. Upon melting, the duplex releases the hras-1 Y strand which, thanks to its flexibility, will have an end-to-end distance < 86 Å (Fig. 5C). This results in a decreasing sigmoidal melting curve, and thus in a -dRFU/dT versus T curve marked by a positive peak. In the same way, hras-1 Y bound to hnRNP A1 is more extended than when it is free. Therefore, also the complex gives a decreasing melting curve and a first derivative curve with a positive peak at ∼ 50 °C. The melting of free hras-1 Y folded in the iM gives an increasing melting curve and a -dRFU/dT versus T curve with a negative peak (Fig. 5D). Summing up, both FRET titrations and melting provide strong evidence that hnRNP A1 unfolds the HRAS iM.
We have also analysed the effect of hnRNP A1 on the hras-2 Y iM. We found that the protein unfolds the iM of hras-2 Y at higher r values, as the iM formed by hras-2 Y has a higher stability than the hras-1 Y iM (58.7 versus 52.1 °C, at pH 5.5) (S 3 ).
The effect of hnRNP A1 on the HRAS iMs was also investigated by CD (Fig. 6). At pH 5.5, hras-1 Y and hras-2 Y show the typical strong ellipticity at 287 nm of the iM conformation. When the iMs are thermally denatured, the positive 287 nm ellipticity drops dramatically. This is a hallmark of the transformation of iM into ssDNA. A similar spectral change was obtained when we added increasing amounts of hnRNP A1 (r = 1, 2, 3, 4) to the iMs. HnRNP A1 causes a progressive reduction of the 287 nm ellipticity, indicating that the iM structures are unfolded by the protein. As already observed with the FRET experiments, the unfolding effect is stronger with hras-1 Y than with hras-2 Y , owing to the different stability of the two iMs.
Insights into the binding of hnRNP A1 to the hras-1 Y iM. Clues to the binding mode of hnRNP A1 to the iM can be obtained from the co-crystal structure of UP1 (the N-terminus of hnRNP A1 with DNA binding activity) and the human telomeric sequence d(TTAGGG) 2 43 . The co-crystal shows that a protein dimer binds to two single-stranded strands, in the antiparallel orientation. As the two binding domains (RRM1 and RRM2) within each protein molecule are also antiparallel, the 5′ → 3′ polarity of ssDNA with respect to the RRM orientation is the same for each RRM. The two lateral loops of the iM may (after a minor adjustment) provide suitable binding sites for hnRNP A1, as they are antiparallel and separated by ~15-20 Å. So, the iM's main function should be to provide a rigid chemical frame displaying two lateral loops with the precise nucleic acid directionality with respect to the RRM orientation. In other words, the iM structure should offer a kinetic advantage to the binding of hnRNP A1 (indeed, when the iM is disrupted, the binding is strongly attenuated, Fig. 4C). If we assume that the protein binds to the lateral loops, it can form either a 1:1 or a 1:2 complex, depending whether one or two protein molecules bind to the iM. The equilibria occurring in solution are: P + iM = P•iM (4); P + P•iM = (P) 2 •iM (5), where P is hnRNP A1. In the presence of a large excess of P compared to iM (1:100), both equilibria shift to the right forming complex 1:2. In contrast, with less P (ratio 1:50) equilibrium (5) does not shift to the right and only complex 1:1 is formed (Fig. 7A, lanes 1-3). In addition, as hnRNP A1 unfolds the iM, its binding depends on temperature: at 0 °C only complex 1:1 is formed probably because it requires a partial unfolding of iM, at 37 °C complex 1:2 is favoured as it requires a complete opening of the iM (S 4 ). To support the binding of the protein to the lateral loops we performed the following competition experiment. As the two lateral loops of the hras-1 Y iM are separated by 10 nt, complex 1:2 should be competed by an oligonucleotide containing the two lateral-loop binding sites separated by a spacer of 10 nt (with a 10 nt spacer the competitor assumes a U-shape so that two antiparallel binding sites can interact with the protein RRMs). Moreover, if the spacer of the competitor is reduced to 8, 6, 4, and 2 nt, its capacity to compete with the formation of the 1:2 complex should gradually become weaker. To test this hypothesis we designed the competitors shown in Fig. 7C. When the competitors (150-fold in excess over iM) were incubated with the iM and hnRNP A1 (100-fold over iM), we found that the best competitor was the oligomer containing a spacer of 10 nt, which corresponds exactly to the distance of the lateral-loop binding sites in the wild-type hras-1 Y iM. These data support the binding of hnRNP A1 to the lateral loops of the iM, as observed for hnRNP LL and BCL2 iM 6 . As stated above, the iM facilitates the initial binding to hnRNP A1. Then, after the iM is unfolded, the protein should bind more stably to the iM sequence.

HnRNP A1 knockout results in downregulation of HRAS transcription in human bladder cancer cells.
As hnRNP A1 binds to the critical GC-elements of the HRAS promoter, we asked if the protein plays a role in transcription. We therefore evaluated the effect on transcription of knocking out hnRNP A1 by shRNA. First, we determined by quantitative real-time PCR the efficiency of shRNA to knock out hnRNP A1, finding that hnRNP A1 mRNA (normalized by the transcripts of β 2-microglobulin and hypoxanthine guanine phosphoribosyltransferase, HPRT) was reduced to 29 ± 8% of the control (cells untreated or treated with a non-specific shRNA C ), 48 h after treatment (Fig. 8A). In the same cells we also measured the level of HRAS mRNA finding that when hnRNP A1 is knocked out, the HRAS transcription is also down-regulated to 44 ± 5% of control. This suggests that hnRNP A1 is an essential factor for transcription, as previously reported 41 . Further evidence that the proteins recognizing the iMs of HRAS are essential for transcription was obtained with decoy oligonucleotides mimicking hras-1 Y iM. These molecules, once introduced in the cells, should sequester the proteins (hnRNP A1 included) recognizing the C-rich strand of the HRAS GC-elements. To increase their nuclease resistance, the decoy oligonucleotides, namely 5291-5294 (see Methods), have been designed with unlocked nucleic acid (UNA) modifications (Fig. 8B) (S 5 ) 44,45 . The capacity of the UNA-modified oligonucleotides to inhibit HRAS transcription was investigated by quantitative real-time PCR. T24 bladder cancer cells were transfected with the decoy oligonucleotides as well as with wild-type hras-1 Y , using as transfecting agent jet-PEI 46 . After an incubation of 24 h, the total cellular RNA was extracted and the amount of HRAS mRNA relative to the housekeeping HPRT mRNA was evaluated by qRT-PCR (Fig. 8C). The results showed that oligonucleotides 5292 and 5293 reduced HRAS mRNA to ~50% of the control (untreated cells). We also examined by electrophoresis the nuclease resistance of the decoy oligonucleotides (Fig. 8D). The oligonucleotides were incubated in cell cultured medium containing 10% fetal bovine serum at 37 °C, pH 5.5 for 0, 18, 24 and 48 h. While hras-1 Y was quickly degraded, the UNA-modified oligonucleotides, in particular 5292 and 5293, showed a remarkable stability, as the fraction of unbroken oligonucleotide was > 0.5, after 48 h of incubation. Interestingly, the enhanced activity shown by these compounds correlates nicely with their higher stability in serum.

Conclusion
We have demonstrated that two neighbouring GC-rich elements controlling HRAS expression can form non B-DNA iM structures, which are stable under near-physiological conditions. These unusual DNA structures are recognized by hnRNP A1, one of the most abundant nuclear proteins involved in the biogenesis of RNA. We have discovered that hnRNP A1 has a clear unfolding activity against the iM. As the knockout of hnRNP A1 by shRNA in T24 bladder cancer cells results in the inhibition of HRAS, hnRNP A1 behaves as an activating transcription factor. Our data, together with those of Hurley and co-workers, who showed that hnRNP LL binds to the iM of the BCL2 promoter and activates transcription 6 , provide the first evidence that non B-DNA iM structures are recognized by nuclear proteins.
The proteins of the hnRNP family have been associated with the promoter of several genes where they are supposed to participate in the transcription regulation mechanisms, although their exact role is not yet fully understood. Some of them recognize C-rich sequences in the promoters of CMYC (hnRNP K) 39,40 , BCL2 (hnRNP LL) 6 and HRAS (hnRNP A1) (present study). These proteins seem to have a complex binding capacity, as hnRNP A1 is also able to bind to G-quadruplex DNA structures in KRAS 17 and telomeres 47 . Remarkably, this type of binding is also associated with the disruption of G-quadruplex structures 17 .
Recent mechanical folding/unfolding experiments showed that G-quadruplex and iM are mutually exclusive within the same double-stranded tract 48 . However, whether this also holds under in vivo conditions, where double stranded DNA is exposed to negative superhelicity and located in a molecular crowding environment, has not yet been demonstrated. It is possible that both G-quadruplex and iM are extruded from each double-stranded GC-element, in the same way as two opposing hairpins (a cruciform) are extruded from a palindromic sequence. HRAS could therefore be regulated by a G-quadruplex/iM switch that represses transcription when the structural elements are in the folded conformation. Transcription will be activated when hnRNP A1 and MAZ, which recognize the HRAS G-quadruplexes 3 , bind to the iM and G-quadruplex, respectively, and then to other proteins of the transcriptional activator complex. These non B-DNA structures provide a mechanism for the control of gene expression at a different level than duplex, involving proteins recognizing these unusual structures that play a central role in gene regulation.
Recombinant hnRNP A1 tagged to GST was obtained with a high degree of purity as previously described 49  CD and FRET experiments. CD spectra have been obtained with a JASCO J-600 spectropolarimeter equipped with a thermostatted cell holder. CD experiments were carried out with oligonucleotides hras-1 Y and hras-2 Y (3 μ M) in 50 mM Tris-acetate, pH from 4.5 to 8, 50 mM KCl. Spectra were recorded in 1 or 0.5 cm quartz cuvette. A thermometer inserted in the cuvette holder allowed a precise measurement of the sample temperature. The spectra have been calculated with J-700 Standard Analysis software (Japan Spectroscopic Co, Ltd) and reported as ellipticity (mdeg) versus wavelength (nm). Each spectrum was recorded three times, smoothed and the baseline subtracted. CD spectra of 3 μ M hras-1 Y and hras-2 Y have been obtained also at various temperatures (20-85 °C), by both heating and cooling the sample solutions (in 50 mM sodium cacodylate pH 5, 50 mM KCl). By plotting the 287 nm ellipticity versus temperature, sigmoidal denaturing and renaturing curves were obtained, which were practically overlapping.
FRET with oligonucleotides hras-1 Y and hras-2 Y , tagged at the 5′ and 3′ ends with ATTO-488 and TAMRA (as donor we used ATTO-488 because its pH dependence is weaker than that of FAM), were carried out on a Microplate Spectrofluorometer System (Perkin Elmer 2300 Enspire, USA). Each sample contained 50 μ l dual-labelled oligonucleotide (200 nM) in 50 mM Tris-acetate buffer, pH from 4.5 to 8, 50 mM KCl and an amount of hnRNP A1 as specified in the figure captions. The samples were incubated at 37 °C as specified in the text. Emission spectra were obtained setting the excitation wavelength at 480 nm and recording the emission from 500 to 650 nm. FRET-melting experiments of hras-1 Y and hras-2 Y have been performed on a real-time PCR apparatus (CFX96, BioRad, Hercules, CA) in 50 mM sodium cacodylate at pH 5, 5.5, 6, 6.5 and 7, 50 mM KCl. FRET-melting experiments were obtained by increasing the temperature from 20 °C to 95 °C (0.3 °C/min). From the melting data we obtained curves reporting the fraction of folded iM against temperature. These curves were reversible (denaturing and renaturing curves overlapping). The energy transfer (E T ) was calculated from the fluorescence intensity of the donor D in the presence (I DA ) and absence (I D ) of the acceptor as: I DA and I D were measured in same buffer under identical concentrations (I D was obtained by transforming the dual-labeled oligonucleotide into the corresponding duplex in which the fluorophores are at a distance for which FRET = 0). The FRET efficiency values were converted to distances between donor and acceptor by using: where R is the distance (Å) and R 0 is the Föster distance [defined as the distance at which energy transfer is 50% of the maximum value, assumed to be 50 Å 50 ].
Thermodynamic analysis of reversible iM melting curves. The thermodynamic parameters for the folding of C-rich sequence into the iM conformation were obtained from the melting curves. From Δ G° = − RT ln K = Δ H°-TΔ S° (8) it is obtained lnK = (Δ H°/R)(1/T) + Δ S°/R (9). The equilibrium constant K as a function of T is given by K = f/(1-f), where f, the fraction of sequence folded in the iM conformation, is obtained form the melting curves. The plot of ln K versus 1/T gives a straight line whose slope (-Δ H/R) and y-intercept (-Δ S/R) provide the thermodynamic parameters (S 9 ).
Cell culture and transfections. T24 human urinary bladder cancer cells were maintained in exponential growth in Dulbecco's Modified Eagle's Medium (DMEM) containing 100 U/ml penicillin, 100 mg/ml streptomycin, 20 mM L-glutamine and 10% fetal bovine serum (Euroclone, Milan, Italy). For transfection we plated 10000 cells for each well in a 96 well plate and transfected using Jet PEI (Polyplus Illkirch FRANCE) following manufacturers in vitro protocol for DNA oligonucleotides transfection with 400 nM oligonucleotide (48 pmol) and N/P = 3.