Oligonucleotides were purchased from Eurogentec (Liége, Belgium) as RP-HPLC purified products. Oligonucleotides were dissolved in milliQ H₂O to prepare 100 µM stock solutions. Before use, each sample was heated at 95°C for 7 min in the required buffer and then slowly cooled down at room temperature to equilibrate the system. The DNA sequences used in this work are listed in Table 1.

The ability of tested oligonucleotides to form intramolecular hairpins was predicted by the use of the IDT OligoAnalyzer tool (https://eu.idtdna.com/calc/analyzer). Setting parameters were 150 mM Na⁺ and 2 µM oligo concentration.

Circular dichroism spectra were recorded on a Jasco J-810 spectropolarimeter equipped with a Peltier temperature controller using 1 cm pathlength quartz cuvette. CD spectra were recorded from 230 to 330 nm. DNA samples were used at 2–4 μM final concentration in 5 mM Tris, pH 7.5. The salt concentration and the temperature used for the experiments varied according to the purpose of the assays.

For CD kinetic experiments, spectra were acquired immediately after the manual addition of the selected cation into the cuvette from a stock solution, with the following parameters: scanning speed 200 nm/min; interval scan of 60 s, bandwidth of 2 nm; data interval of 0.5 nm; response of 1 s.

For CD melting experiments, spectra were acquired between 95 and 25°C every 2°C. Each spectrum represents the average of 2 scans acquired with the following parameters: scanning speed 100 nm/min; band width of 2 nm; data interval of 0.5 nm; response of 2 s. Spectra were reported as molar ellipticity ([θ]). To analyze the variation of signals at single wavelength each data point was subtracted of the starting CD value contribution and the difference was normalized between 0 and 1 (relative molar ellipticity).

Thermal differential spectra were obtained by subtracting the oligonucleotide UV-spectrum acquired at 25°C from the one recorded at 95°C. Experiments were performed using 2–4 µM oligonucleotide solution in 5 mM Tris pH 7.5, 150 mM KCl.

The genomic sequence that corresponds to full-length human Vimentin (residues Met1-Glu466) was inserted into the pET21a + vector by Genescript. The plasmid was used to transform competent E. Coli BL21 (DE3) cells. Bacteria were grown at 37°C in LB media until an OD₆₀₀ of 0.7 was reached. At this point, the expression of Vimentin was induced by adding 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG). Vimentin was purified as previously described (Herrmann et al., 2004). Briefly, cells were collected by centrifugation at 7000 g for 20 min at 4°C, resuspended in 10 mM Tris, 2 mM PMSF, 10 mM MgCl₂, pH 7.5. After resuspension, the cells were sonicated and up to 0.2% Triton X-100 and 50 g/ml DNase I were added. The inclusion bodies were harvested by centrifugation, and then resuspended in the corresponding washing buffer (10 mM Tris, 1 mM EDTA, 1 mM DTT, 2 mM PMSF, 1.5 M KCl, pH 7.5). After centrifugation, this procedure was repeated with the same buffer without KCl. Finally, the pellets were washed with 10 mM Tris, 0.1 mM EDTA, pH 7.5. After the last washing step, inclusion bodies were resuspended in a denaturing buffer (9.5 M urea in 10 mM Tris, pH 7.5). The solubilized IF proteins were collected by centrifugation at 60,000 g for 1 h at 20°C. Vimentin was purified from the supernatant by two steps of exchange chromatography, using first an anionic (Q Sepharose FF, Cytiva) then a cationic (SP Sepharose FF, Cytiva) column. The buffer used for the exchange chromatography was 10 mM Tris, 500 mM NaCl, 1 mM DTT, pH 7.5. Purified Vimentin was stored at −80°C.

The day before use, Vimentin was renatured following the protocol developed by Herrmann and others to avoid polymerization into filaments (Herrmann et al., 2004). The protein was dialyzed at room temperature against dialysis buffer (5 mM Tris, 1 mM EDTA, 0.1 mM EGTA, 1 mM dithiothreitol, pH 8.4) containing a progressively reduced urea concentration (6, 4, 2, and 1 M urea). Dialysis against a large volume of buffer with 1M urea proceeded overnight at 4°C. The next day, dialysis was continued against tetramer buffer (5 mM Tris, pH 8.4) for 1 h at room temperature. After dialysis, the protein concentration was determined by measuring the absorption at 280 nm with ε = 24,900 cm⁻¹M⁻¹.

For each sample, a solution of oligonucleotide (2–4 μM strand concentration) in 5 mM Tris, pH 7.5 was heated at 95°C for 7 min and allowed to cool down to room temperature overnight. Afterwards, variable concentrations of KCl (0, 10 and 150 mM) were added to the samples and they were led to equilibrate overnight at room temperature. The samples were then loaded on a native 15% polyacrylamide (19:1 acrylamide: bis-acrylamide) gel in 1 × TBE (89 mM Tris, 89 mM boric acid, 0.02M EDTA) implemented with 20 mM KCl. The gel was stained with Sybr Green II and the resolved bands were visualized on an image acquisition system (Geliance 600 Imaging system, Perkin-Elmer).

Electrophoretic mobility shift assays (EMSA) were performed on 1.5% agarose gels in 1 × TBE buffer and 10 mM KCl. Each reaction sample contained 500 nM DNA and 8 μM Vimentin, in 5 mM Tris, 150 mM KCl, pH 8.4. Vimentin was added to oligonucleotide solutions and incubation was performed for 1 h at room temperature. Immediately before loading, 3 μl of gel loading buffer (50% glycerol, 50% water) were added to the samples. Electrophoresis proceeded for 2 h at 6 V/cm at 4°C. Gels were stained with Sybr Green II to visualize nucleic acids and with Coomassie Brilliant Blue G250 for protein detection. Gel images were acquired with a Geliance 600 apparatus.

In a previous study (Ceschi et al., 2022) the Q-PARSE bioinformatics tool was used to search sequences containing two contiguous G4-forming elements in the region covering 100 nucleotides upstream the transcription starting site of human genes. The search was extended to both, the forward and reverse strands. The searched pattern comprised islands of 3–4 guanines with connecting loops of maximum 5 nucleotides. The software provided an output of 1477 genes that contain at least one putative double G4-repeat. Among the retrieved sequences, we chose to characterize a sequence, TEAD4-full (Table 1), located 31 nucleotides upstream the TSS of the TEAD4 oncogene.

Besides the functional roles of the protein encoded by this gene, the selected sequence represented an attracting system thanks to its guanines number and distribution. The study of any G4-repeat system might be very complex due to the potential coexistence of multiple structural equilibria (Monsen et al., 2022). However, it was possible to subdivide TEAD4-full into two shorter domains, each of them comprising 4 runs of at least 3 guanines (thus each of them putatively able to fold into a stable G4), separated by a short linker (Figure 1). Thus, we took advantage of this base composition of TEAD4-full and we started its characterization by analyzing the conformational features of the two short sequences located at the 5′- and 3′-end, addressed as TEAD4-near and TEAD4-far, respectively.

To address the conformational features of the selected sequences, we performed CD spectroscopy that allows to follow the folding process of oligonucleotides and provides useful insights into the topology of the nucleic acid arrangements in solution. Since it is known that G-quadruplex formation is promoted by the presence of some cations, in particular potassium, we started our analyses by monitoring the modifications of the CD spectrum of TEAD4-near in 5 mM Tris, pH 7.5 upon addition of 10 and 150 mM KCl (Largy et al., 2016). Under both conditions, we observed an increase in molar ellipticity at 290 nm and a decrease at 260 nm that occurred in the same time scale at the two tested KCl concentrations (Figure 2).

The shape of the CD spectrum is known to be different for a G4 of parallel or antiparallel topology (Balagurumoorthy et al., 1992). Parallel G4s typically show a positive signal at 265 nm and a negative one at 240 nm, whereas antiparallel quadruplexes exhibit a positive signal at 290 nm and a negative one around 260 nm. Thus, the CD spectra of TEAD4-near suggested that the addition of potassium drove the formation an antiparallel G4. The presence of a G4 structure was further confirmed by the UV thermal differential spectrum (TDS) (Supplementary Figure S1A), that exhibited the characteristic pattern associated to this non-canonical structure, with two positive peaks at 240 and 277 nm, a shoulder at 255 nm and a negative minimum at 296 nm (Mergny 2005).

The melting and annealing profiles of TEAD4-near were acquired at the physiologically relevant KCl concentration (150 mM) and they were perfectly reversible (Figure 3). The melting profile showed a progressive reduction of the CD signal at 290 nm. This decrease was paired with an increment of the signal at 265 nm that reached the maximal intensity at about 70°C and that was lost at higher temperatures (Figure 3A). This transition can be attributed to an intermediate state occurring along the folding/unfolding of the antiparallel G4 but which presence was negligible at physiological temperature.

To verify whether some of the detected species might correspond to intermolecular arrangements, native polyacrylamide gel electrophoresis (PAGE) experiments were performed. TEAD4-near samples were analyzed before and after an annealing step carried out in the presence of KCl. All of them run as a single band with an electrophoretic mobility higher compared to the one of an unfolded 22-nts scrambled oligonucleotide (Supplementary Figure S2A), thus confirming that TEAD4-near folds only into intramolecular structures under our experimental conditions.

Consistently with the above described experimental approach, we started the characterization of TEAD4-far structural equilibria by monitoring in time the CD spectra modifications upon addition of increasing KCl concentrations up to 150 mM (Figure 4A). At all tested conditions, the CD spectra of TEAD4-far progressively evolved towards the formation of two main positive peaks located at 290 and 265 nm that, as supported by TDS, were associated to its folding into G4 (Supplementary Figure S1B).

CD spectra with maxima at both 265 and 290 nm are typical of G4 with hybrid topology (Balagurumoorthy et al., 1992). Alternatively, they can be observed when a sequence in solution distributes between parallel and antiparallel conformations. Here, by comparing the molar ellipticity of the oligonucleotide solutions after 1 h of incubation, it emerged that their relative intensities were a function of the metal ion concentration (Figure 4B). As far it concerns the peak at 265 nm, it reached the maximum of molar ellipticity at 100 mM KCl. Conversely, the intensity of the peak at 290 nm increased moving from 1 to 10 mM KCl while further increments of the metal ion concentration reduced it to converge towards a constant value of molar ellipticity. The differential response at these two wavelengths was further highlighted by comparing the folding kinetic of the oligonucleotide at these two wavelengths, the variation of the signal at 265 nm being much slower than the one occurring at 290 nm at all tested KCl concentrations (Supplementary Figure S3). These data indicated a competition between two distinct oligonucleotide arrangements that leads to their different relative abundance as a function of KCl concentration. At high KCl concentration parallel G4 arrangement was favored. Interestingly, melting/annealing profiles acquired in 10 mM KCl nicely overlapped (Supplementary Figure S4). Conversely, in 150 mM KCl, by heating of the solution up to 65°C, a modest but significant rearrangement occurred, which corresponded to an increase of the molar ellipticity at 265 nm associated to a reduction of the signal intensity at 290 nm (Figures 4C,D). Noteworthy, this species was the one preserved at r.t. after the annealing.

Overall, data suggested that TEAD4-far distributes between a kinetically favored antiparallel structure and a thermodynamic favored parallel conformation, the annealing shifting the system towards the last one at physiologically relevant KCl concentrations. PAGE experiments confirmed that none of them is related to intermolecular folded structures (Supplementary Figure S2B).

Based on the above presented results, it was unexpected to observe that the addition of KCl up to 150 mM to TEAD4-full produced only modest changes of the dichroic signal. They comprised a fast increment at 290 nm that was paired with a very slow rearrangement at lower wavelengths that occurred on the time scale of hours (Figures 5A; Supplementary Figure S5). This chiroptical profile markedly changed as a function of potassium concentration when we analyzed the annealed oligonucleotide solutions (Figure 5B) with a positive band centered at 265 nm progressively emerging above 50 mM KCl. In particular, at 150 mM KCl, this contribution was largely dominant. UV thermal differential scans confirmed that it corresponded to TEAD4-full preferentially folded into G4, since the negative signal at 296 nm, characteristic of the G4 structure, was remarkably more intense when compared to the one acquired before the annealing step (Supplementary Figure S1C–D). These data supported the preferential formation of a parallel G4 at physiological KCl concentration.

Worth of note, once annealed, at all tested salt concentrations the system was at the thermodynamic equilibrium and, subsequent melting/annealing cycles were fully reversible (Figure 6). Their profiles further supported that at the lower tested KCl concentration the presence of multiple components was preserved (Supplementary Figure S6). Also for this sequence, PAGE confirmed the conserved presence of only intramolecular structures (Supplementary Figure S2C).

According to these data, it is possible to propose that the addition of metal ion to TEAD4-full stabilized an intramolecular arrangement that prevents the formation of the G4-repeat, the thermal denaturation of this starting arrangement(s) allowing the formation of the more thermodynamic stable parallel G4. Based on the intensity of the signal, it appeared that at physiological conditions it covered both the two G4 domains. This is compatible with the formation of a G4-repeat.

The described structural folding landscape of TEAD4-full is complex. To better describe it, we designed a sequence with a reduced propensity to adopt G4-competing structures. By analyzing the TEAD4-full sequence with the IDT OligoAnalizer prediction tool, we retrieved several hairpin structures, as expected for a sequence with such a remarkable high GC content. Noteworthy, the most stable hairpins involved constantly the central four cytosines paired with different guanines that, based on our data on the short TEAD4 domains, should be involved in G-tetrads formation (Supplementary Figure S7). To validate the potential impact of these secondary structures on the G4 formation, we considered to reduce them by converting the central cytosines into thymines (Figure 1). Thus, we characterized the mutated sequence TEAD4-full-2TT reported in Table 1.

The CD spectra of TEAD4-full-2TT after the addition of 150 mM KCl showed a weak peak at 290 nm similar to the one of TEAD4-full. However, differently from the wild-type sequence, it comprised also a weak shoulder at 265 nm, likely due to the presence of a small fraction of parallel G4 (Figure 6A). The annealing of this solution readily drove a strong increment of this component, comparable to TEAD-full although slightly more intense (Figure 6B). These data indicated that the introduced mutations did not fully prevent the formation of potentially G4-repeat competing structures, but they reduced them thus making more efficient the structural rearrangement towards the thermodynamically favored parallel G4-repeat.

To address the binding of Vimentin to TEAD4, we performed electrophoretic mobility shift assays with the purified recombinant human protein. Binding reactions were carried out at pH 8.4 to avoid Vimentin polymerization into filaments (Herrmann et al., 2004). Due to the complex folding landscape of the herein tested sequences, oligonucleotides were either equilibrated or annealed in 150 mM KCl to promote G-quadruplex formation before protein addition.

As expected from the high selectivity of Vimentin for G4-repeats, no protein-DNA complexes were observed when Vimentin was incubated with the isolated TEAD4-near and TEAD4-far (Figure 7A). Conversely, as shown in Figure 7B, Vimentin interacted with TEAD4-full. Remarkably, the complex formation occurred only when the oligonucleotide was previously annealed in 150 mM KCl. This result further confirmed that the parallel G4 arrangement induced in these conditions actually corresponds to a G4-repeat.

It was interesting to observe that Vimentin did bind to TEAD4-full-2TT either before and after the annealing. In the first case, the interaction was weak, while no free DNA was observed when Vimentin was added to the previously annealed TEAD4-full-2TT (Figure 7C). The higher affinity of the protein for this full-length mutated sequence further indicated that the selected mutations efficiently reduced the formation of structural motives that compete with the complete folding of TEAD4-full into a G4-repeat.