The genome of P. falciparum clone 3D7 was searched for Putative Quadruplex Sequences (PQS) using QGRS-Mapper [33] on both the positive and negative strands. We set up the QGRS-mapper software to identify all PQS with four repeats of at least three guanines interrupted by loops of a maximum length of 11 nucleotides. As expected, most PQS were found in the telomeres (828 out of 891) due to their repetitive sequence: GGGTT(T/C)A (see Additional file 1). These telomeric G-quadruplexes of P. falciparum have recently been described by De Cian et al [34]. Here we focused on the non-telomeric PQS because of their potential role in gene transcriptional regulation. We identified 63 non-telomeric PQS (listed in full in Additional file 2). This is an average of one PQS per 380 kb, which is a much lower ratio than that seen in the human genome (1 PQS per ~8 kb) [17]. This was expected due to the extreme AT-richness (80.6%) of the P. falciparum genome [35]. 37 of the 63 PQS are in intergenic regions, and of the 26 PQS within genes, 9 are on the coding strand and 17 on the non-coding strand.

Most importantly, 16 out of the 63 PQS were found in the upstreamB regions of var genes, 1612 to 1707 bp upstream of the initiation codon (Table 1 and Additional file 2). These 16 PQS contain only three distinct sequences that thus represent three slightly different putative G-quadruplexes that we named UpsB-Q-1, UpsB-Q-2 and UpsB-Q-3 in order of their frequency (sequences shown in Table 2). As the var gene repertoire varies from one clone to another, we also searched for PQS in the upstream regions of var genes in P. falciparum clone HB3 (available from the Broad Institute website, ). Using the same parameters, we found 11 PQS in the upstream B region of var genes in HB3 (Table 1). Interestingly, UpsB-Q-1 was also the most common PQS (7 out of 11) in parasite clone HB3. One PQS was found in HB3 but not in 3D7 (named UpsB-Q-4, Table 1 and 2). These four types of PQS do not exist in any other sequenced organism to date (BLAST analysis, data not shown).

In order to confirm the formation of G-quadruplexes by these sequences, two of them (UpsB-Q-1 and 2) were analyzed by ¹H NMR in the presence of ammonium or potassium ions. Using this technique, it is well established that the presence of imino protons with a chemical shift between 10 and 12 ppm is characteristic of the formation of G-quadruplexes [36-38]. Spectra recorded on the presumably unstructured oligonucleotides in water did not show signals beyond 9 ppm, indicating that imino protons are in fast exchange with bulk water. The spectra recorded after the addition of 150 mM cation are presented in Figure 2 and show clearly the presence of imino protons for the four samples. Most importantly, for UpsB-Q1 in ammonium and potassium, 11 imino peaks are clearly distinguished. In potassium, the peak at 11.93 ppm was resolved in two different peaks at 37°C (data not shown), indicating the formation of a single stable structure containing three quartets. In ammonium, the intensity of the peak at 11.55 ppm could also indicate a superposition of two different peaks. On the UpsB-Q-2 spectra, the imino peaks are less well resolved, indicating some structural polymorphism. Moreover, in both PQS sequences, peaks were observed at higher chemical shift, indicating the possible presence of additional structures like AT-rich hairpins on these G-quadruplexes. These NMR data show that UpsB-Q1 and UpsB-Q-2 do form stable G-quadruplex structures in the presence of physiological concentrations of potassium ions.

The four types of PQS found in the upstream B regions of var genes were also examined by circular dichroism (CD), which provides information about the orientation of strands within a G-quadruplex, because the CD signal changes with the syn/anti orientation about the glycosylic bonds. In parallel G-quadruplexes, the CD spectrum typically exhibits a positive peak around 260 nm and a negative peak around 240 nm, whereas in antiparallel G-quadruplexes, the CD spectrum displays a negative peak around 260 nm and a positive peak at 295 nm [39-41]. Results of CD for each UpsB-Q are showed in Figure 3. For UpsB-Q-2, 3 and 4 in potassium and ammonium, a minimum around 243 nm and a maximum around 295 nm were observed, while for UpsB-Q-1 in potassium the minimum was around 250 nm. These kinds of spectra are generally attributed to hybrid conformations (containing a mixture of both parallel and antiparallel strand orientations). In sodium, the four sequences seem to adopt an antiparallel conformation, since a minimum near 260 nm and a maximum at 290 – 295 nm are observed. However, a shoulder was observed in the three cationic conditions at 270–275 nm for UpsB-Q-2, 3 and 4 but not for UpsB-Q-1. The absence of a shoulder in the case of UpsB-Q-1 could therefore indicate a different conformation of this G-quadruplex. Furthermore, the depth of the minimum observed for UpsB-Q-1 was systematically bigger than for the three other sequences. Together with NMR, this result suggests that the single structure adopted by UpsB-Q-1, the most frequent PQS in the UpsB regions, is different (probably closer to an antiparallel form) and that a few base mutations in UpsB-Q-1 (3 bases for UpsB-Q-2, 2 bases for UpsB-Q-3 and a single base for UpsB-Q-4) are sufficient to induce structural polymorphism.

Finally, thermal difference spectra (TDS) were recorded [42]. Similarly to CD, this technique differentiates between the various potential structures adopted by DNA. It was shown previously by Mergny et al [42] that G-quadruplexes exhibit two positive peaks at 243 and 273 nm and one negative peak at 295 nm, while other DNA structures show different combinations of maxima and minima (see [42] for details of maxima and minima characterizing other DNA structures). The spectra obtained for the PQS from the var gene upstream B regions match expectations for G-quadruplex structures (Figure 4). There is a negative peak at 295 nm for all sequences in the three cations. Positive peaks were observed at 246 (only in potassium), 256 (except in ammonium) and 267 nm. These values are in good agreement with previously described results [42] and thus confirm the ability of these sequences to form G-quadruplexes. There is a small wavelength shift for positive peaks, which may be because of the presence of a long loop (10 bases) in the PQS of the var gene upstream regions (all the G-quadruplexes previously tested contained no more than four bases in their loops).

The stoichiometries of the G-quadruplexes formed by the UpsB-Q were also examined to determine whether these structures are likely to form intra-molecular bonds (unimolecular structures) or inter-molecular bonds (multimolecular structures) [43]. Mass spectrometry showed only monomeric DNA (Additional file 3), indicating that the UpsB-Q form intra-molecular G quadruplex structures (inter-molecular structures would have been indicated by the presence of multimers by mass spectrometry).

When the UV absorbance is monitored at 295 nm [44], thermal denaturation experiments allow determination of the T_m of G-quadruplexes, i.e. the temperature at which the half of the signal is lost and thus indicate the relative stability of structures adopted by the oligonucleotides. Because G-quadruplexes with long loops are usually less stable that those with small loops [43,45,46], it was necessary to check if the UpsB-Q are able to form G-quadruplexes that are stable under physiological conditions (37°C and [K⁺] ≈ 150 mM). These results are shown in Figure 4 and the T_m's are listed in Table 3. In potassium, the four PQS sequences from the var gene upstream B regions have a T_m about 50°C, and the transition curves (Figure 5) show that, at 37°C, the proportion of folded G-quadruplexes is higher than 85% for the four sequences, confirming their potential to form G-quadruplexes in living cells. As expected, T_m's observed in sodium and ammonium are lower than in potassium. This is generally attributed to the weaker stabilization capacity of these two cations, due to their smaller ionic radius [47-49]. The T_m ranking K⁺ > Na⁺ > NH₄⁺ is also characteristic of antiparallel G-quadruplexes (parallel G-quadruplexes have T_m ranking K⁺ > NH₄⁺ > Na⁺) [43].

In addition to the potential transcriptional repressor activity of G-quadruplex sequences themselves [27], it has been shown previously that G-quadruplex ligands can further suppress transcription of genes containing potential G-quadruplexes in their promoters, by impeding the binding of proteins needed for initiation of transcriptional activity on DNA [50,51]. Moreover, these molecules are also able to interfere with telomere structure and to indirectly induce their shortening [52-55]. These molecules are thus promising weapons in the fight against cancer, since this disease needs both a high expression of oncogenes and stable telomere length to develop and survive [56-62]. With the discovery of G-quadruplex forming sequences in the genome of P. falciparum, it can be hypothesized that these ligands may also have the potential to affect parasite gene expression by stabilizing G-quadruplexes located in gene promoter regions.

It was thus decided to evaluate the equilibrium dissociation constant of each UpsB-Q with the G-quadruplex ligand BOQ1 (Figure 6A). BOQ1 is a synthetic compound that exhibits a good selectivity for G-quadruplex versus duplex DNA [63,64]. A relatively high association constant with the UpsB-Q sequences would therefore provide additional evidence for the presence of G-quadruplex structures. Binding constants were determined by electrospray mass spectrometry. A typical spectrum of a G-quadruplex-ligand mixture is shown in Figure 6B. The peaks correspond to the free DNA sequence, and to 1:1 complexes (1 ligand per DNA sequence), with charge states of 6-, 7- and 8-. The binding of a second ligand molecule was not observed, revealing that a single binding site is present.

The equilibrium dissociation constants of all UpsB-Q with BOQ 1 were deduced from the relative intensity of peaks of free DNA and complexes, as described previously [65]. For the four sequences, the values are around 2 μM (Table 3). They are lower than those obtained by mass spectrometry for the binding of BOQ1 to telomeric G-quadruplex (5.7 μM) or to model duplex sequences (57 μM) (unpublished data). These results confirm the ability of the PQS in the upstream B regions of the var genes to fold in G-quadruplexes, and show that G-quadruplex ligands are likely to bind to these structures within the P. falciparum genome, and could therefore be tested for biological activity against the parasite.

If the G-quadruplex structures in the P. falciparum genome can inhibit gene transcription in vivo, one would expect a mechanism to regulate these structures. It has been shown previously in vitro that some helicases specifically unwind G-quadruplexes in telomeres [66,67]. Two helicases of the conserved RecQ family, Sgs1p in yeast and BLM in human, show a much higher affinity for G-quadruplex than duplex DNA [68,69]. Moreover, a recent microarray study in a yeast mutant suggests that genes with a PQS in their upstream region tend to be downregulated when Sgs1p is knocked out [70]. The proposed explanation is that, in the absence of the helicase, G-quadruplexes in promoters cannot be unwound and they therefore act as a steric block to the transcriptional machinery [71]. A BLAST search with BLM and Sgs1p in P. falciparum brings up PFI0910w, annotated as a putative RecQ helicase on PlasmoDB. A ClustalW alignment with these 3 sequences shows that the 7 domains found in all RecQ helicases are also conserved in PFI0910w (Figure 7). PFI0910w is comparatively shorter and is lacking the RecQ-C-terminal and a HRDC (Helicase-and-RNaseD-like-C-terminal) domain present in BLM or Sgs1p. However, RecQ4, another human helicase part of the RecQ family, is also lacking these 2 domains [72]. Therefore it appears that P. falciparum does encode a helicase that may have the capacity to regulate G-quadruplex structures, allowing us to hypothesize that the RecQ helicase (PFI0910w) could be involved in UpsB var gene regulation. This suggestion that members of a particular var gene group could have a unique regulatory mechanism is not unprecedented, as it has been suggested previously that Group A and E var gene transcription, but not groups B and C were influenced by the histone deacetylase Sir2 [11].

Both strands of each chromosome of the P. falciparum 3D7 clone (PlasmoDB_5.4 [73]) were analyzed using QGRS-Mapper [33]. The parameters used were: Max length: 33; Min G-group: 3; loop size: 0 to 11. The P. falciparum HB3 genome was downloaded from the Broad Institute . Upstream sequences of var genes were analyzed using QGRS-Mapper with the same parameters.

All oligonucleotides were ordered from Eurogentec (Seraing, Belgium) with Oligold quality. The oligonucleotide sequences used are shown in the Table 2. Oligonucleotides were received lyophilized and stock solutions were prepared in bi-distilled water with 300 μM total strand concentration. For all experiments, the stock solution was heated at 80°C for 5 minutes, diluted using a cold aqueous solution containing either KCl, NaCl or NH₄OAc to reach the desired DNA concentration in 150 mM cation, and then cooled rapidly on ice. 10 mM lithium cacodylate, pH 7.4 was added in thermal denaturation and circular dichroism experiments. The molecule BOQ1 was synthesized as described previously [74]

Experiments were performed on a Jasco J-810 spectropolarimeter using 1-cm path length cells (Hellma, type No. 114-QS, France). The final concentration of oligonucleotide was 5 μM in a buffer containing 150 mM salt and 10 mM lithium cacodylate, pH 7.4. For each sample, five spectra were recorded from 220 nm to 350 nm with a scan rate of 100 nm/min.

NMR samples were prepared by dissolving the oligonucleotides in H₂O/D₂O 90/10, lithium cacodylate 10 mM, pH 7.4 to get a oligonucleotide final concentration of 270 μM. Ammonium acetate or potassium chloride were progressively titrated in to a final cation concentration of 150 mM. NMR data were collected at 500 MHz on a Bruker Avance spectrometer (fitted with a TXI triple resonance probe with z-axis gradient). 1D ¹H spectra were recorded at a temperature of 25°C using a WATERGATE sequence with a water flip-back pulse [75,76].

Thermal denaturation experiments were carried out on a Uvikon XS spectrophotometer (Secomam), using 1-cm path length quartz cells (Hellma, type No. 115B-QS, France). The final oligonucleotide concentration was 5 μM in 150 mM salt and 10 mM lithium cacodylate, pH = 7.4. Absorbance was monitored as a function of the temperature at 295, 240, 260 nm for the determination of the melting temperature (T_m) [44] and at 405 nm as control wavelength. Gradient was 0.2°C/min between 10 and 90°C. Melting temperatures were determined using the method described by Marky and Breslauer [77]. Before heating and after the cooling, spectra were recorded from 220 to 440 nm, to allow thermal difference spectra (TDS) to be obtained. TDS were obtained by subtracting the low temperature curve from the high temperature curve and normalization, as described previously by Mergny et al. [42].

All measurements were carried out on a Q-TOF Ultima Global mass spectrometer (Micromass, now Waters, Manchester, U.K.), using the electrospray ionization (ESI) source in negative mode, as described previously [78]. Source conditions were optimized to avoid in-source fragmentation: capillary voltage = -2.2 kV, cone voltage = 50 V, RF, source block temperature = 80°C, and desolvation gas temperature = 100°C. Source backing pressure was set to 3.5 mbar. Oligonucleotide samples were first prepared at 50 μM final concentration in NH₄OAc 150 mM. Just before injection in the mass spectrometer, they were further diluted to 10 μM in 150 mM NH₄OAc and 20% methanol. The role of methanol is to increase ion signals.