G quadruplex DNA structures may exist
G-quadruplex DNA structures may exist in either positive or negative strand of a promoter to regulate gene expression. The positive strand of the HOXC10 promoter is overall C-rich in the 1000 bp upstream of the TSS (36.6% of C versus 24.1% of G) and contains many C-runs; thus, G-quadruplex structures should be mostly in its negative strand, which was proven by the analyses using a G-quadruplex prediction algorithm . While most G-quadruplexes can stack 3 or 4 planar quartets, the region between −295 and −273 in the HOXC10 promoter can potentially form 5 layers of quartets (Fig. 2A) and is presumably stronger than other G-quadruplex structures with 4 or less quartets. We also noticed several C-runs between −380 and −350 of the HOXC10 promoter, which also have the potential of forming G-quadruplex structures in its complementary strand and thus may play a role in regulating HOXC10 expression. However, compared with the region analyzed in this study, their relative scores for G-quadruplex structures (Fig. 1C) are fairly low and the position is relatively distant from the transcription start site, suggesting that they are likely less important. It is noteworthy that any mutation in a promoter sequence may either abolish or create TAME mg of transcription factor(s) and thus alter gene expression. In the reporter assays, we focused on the seven G-tracts between −297 and −243 in the HOXC10 promoter and differentially mutated or deleted them, followed by evaluating their relative activity in driving Gluc expression. All mutations or deletion of the four 5G-tracts (or G-tracts 4–7 in Fig. 4A or Fig. 3A) with the presence of intact G-tracts 1–3 led to increased Gluc activity, suggesting that G-quadruplex structures play a negative role in HOXC10 gene expression. The G-to-A or G-to-T substitutions showed no difference in reporter activity. It is certainly possible that G-tracts 4–6 may contain a binding site(s) of transcription repressor(s) and thus their absence led to enhanced gene transcription. Interestingly, disruptions of either G-tracts 1–3 or G-tracts 1–6 positively impacted HOXC10 promoter activity. The data suggested that G-tracts 1–3 may contain binding sites of transactivators indispensable to HOXC10 promoter-mediated transcription, while four 5G-tracts negatively regulate HOXC10 gene expression by forming G-quadruplex structures. The presence of multiple G-tracts in the negative strands of the HOXC10 promoters from other species just suggests G-quadruplex structure-mediated HOXC10 expression may be evolutionarily conserved. Whether G-quadruplex structure is indeed formed in these promoters and regulates HOXC10 expression in these species needs further investigation. Both WT and M2 had intact four 5G-tracts between −295 and −273, and showed molar ellipticity at both 262 nm and 295 nm. However, they displayed significant difference in thermo melting stability (Table 1). It is possible that the presence of adjacent G-tracts between −267 and −243 could increase the multiformity of G-quadruplex structures and consequently enhance thermo melting stability. The position of G-quadruplex structure relative to single strand oligonucleotides in native PAGE is controversial among different reports [35,39,46]. In our study, WT oligonucleotide in NC (no cation) solution did not show a signature peak of G-quadruplex structure (Fig. 3B), suggesting its low capacity in forming G-quadruplex structure at this condition. In native PAGE, this sample displayed only one major band at the most front position, and WT at K+ and Li+, which both displayed absorbance at 262 nm, did not show any band migrating faster than it (Fig. 3D). This strongly suggested that oligonucleotides with G-quadruplex structure migrated slower than single strand DNA in native PAGE in our experimental setting. In addition, mutant oligonucleotide M3, which had most G-tracts disrupted by thymidine replacement and showed no G-quadruplex feature in circular dichroism spectra, gave a major band at a similar level to that of WT in NC condition. Importantly, our gel-shift study clearly identified the position of intramolecular G-quadruplexes formed by WT oligonucleotide, which was not the fastest-migrating band (Fig. 3E). We also noticed that WT in NC condition migrated faster than M1 and M3 oligonucleotides, which could be due to difference of undetermined structures among these molecules. For the slowly migrated oligonucleotides, especially WT and M2 at K+ and Li+ conditions, we believe they were mostly intra- and intermolecular G-quadruplex structures. When we introduced these oligonucleotides into cells followed by immunostaining using a G-quadruplex antibody, we could clearly detect signal in cells transfected by WT, while M1 and M3 showed staining with markedly reduced intensity. It is possible that WT formed diversified G-quadruplexes among the seven G-tracts, which could be well-detected by the antibody, while less G-tracts, including M2 with intact four 5G-tracts, may not be preferentially recognized. We also noticed that the staining of WT was mostly cytoplasmic. We predict that either transfected oligonucleotides could not enter the nucleus, or WT oligonucleotides, if forming any G-quadruplex structure, were unwound inside the nucleus by DNA helicases. Nevertheless, the staining of WT in cells by the G-quadruplex antibody provided strong proof of G-quadruplex structure formation in the HOXC10 promoter.