Diana J Uribe, Yoon-Joo Shin, Eric Lau, Scot W Ebbinghaus and Daekyu Sun*
Pharmacology and Toxicology Department, Cancer Biology Graduate Interdisciplinary Program, University of Arizona, Tucson, AZ, USA
Received December 14, 2012; Accepted December 26, 2012; Published December 28, 2012
Citation: Uribe DJ, Shin YJ, Lau E, Ebbinghaus SW, Sun D (2013) Heterogeneous Nuclear Ribonucleoprotein K Binds to the Cytosine-Rich Sequence of the Hypoxia Inducible Factor 1 Alpha Proximal Promoter that forms a Stable i-motif at Neutral pH. J Phys Chem Biophys S5:001. doi: 10.4172/2161-0398.S5-001
Copyright: © 2013 Uribe DJ, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Visit for more related articles at Journal of Physical Chemistry & Biophysics
The proximal promoter of the hypoxia inducible factor 1 alpha (HIF-1α) gene contains a poly-purine/poly-pyrimidine (pPu/pPy) tract, which has been shown to affect 90% of its transcriptional control. The guanine-rich sequence of this pPu/pPy tract has been known to be structurally dynamic, easily forming a G-quadruplex structure with a 1:6:1 folding pattern. In present study, we demonstrated that the cytosine-rich (C-rich) sequence within the pPu/pPy tract of the HIF-1α promoter is able to form two major intramolecular i-motif structures with 3:3:3 or 3:4:2 folding patterns near physiological pH using circular dichroism, bromine footprinting and site-directed mutational analysis. These structures are the first known i-motifs that form at neutral pH, with a transitional pH at 6.9, which have been discovered to form within the proximal promoter sequence of oncogenes. Additionally, electrophoretic mobility shift assays (EMSA) combined with bromine footprinting revealed that heterogeneous nuclear ribonucleoprotein K (hnRNP K) is able to bind to the unfolded state of C-rich sequence in a sequence specific manner. Taken together, our results demonstrate that the i-motif structures that form within the C-rich sequence of the HIF-1α promoter can form under physiological conditions and that hnRNP K can bind to this C-rich sequence.
i-motif; HIF-1 alpha; hnRNPK
HIF-1α: Hypoxia Inducible Factor 1 alpha; pPu/ pPy: poly-purine/poly-pyrimidine; G/C-rich: Guanine/Cytosinerich; CD: Circular dichroism; hnRNP K: heterogeneous nuclear ribonucleoprotein K; CNBP: CCHC-Type Zinc Finger Nucleic Acid Binding Protein; Sp1: Specific Protein 1
Hypoxia Inducible Factor 1 (HIF1) is the most important transcription factor in the tumor response to hypoxia since the survival and adaptation of tumor cells to low oxygen levels depends upon the expression of this transcription factor and its family members [1,2]. HIF1 is known to activate genes involved in the hallmarks of cancer [3-6] that ultimately leads to increased metastasis of the tumor lesion by avoiding apoptotic events and coordinately turning on multiple cell survival mechanisms [7,8]. HIF1 is a complex composed of two subunits; hypoxia induced HIF-1α and constitutively expressed HIF- 1β [9,10]. In the presence of oxygen the α subunit is targeted for proteosomal degradation [11,12] ultimately inhibiting its dimerization to the β subunit, which is necessary for activation of the complex as a transcription factor. Accumulation of the α subunit is accomplished in extremely low oxygen levels or due to a variety of genetic abnormalities that avoid the degradation of the subunit . HIF-1α overexpression has been shown to be a predictive factor of recurrence and distant metastasis in the majority of common human solid tumors and is positively associated with increased patient mortality [2,13]. Due to its high expression in the majority of solid tumors and the dependency of these tumors on the continuous and steady protein levels of HIF-1α for nutrient and oxygen supply, the transcription factor subunit has been recognized as a molecular target for cancer therapy.
Because HIF-1α is predominantly regulated by post-translational modification, not much attention has been paid to its regulation at the transcriptional level. More so, the mechanism of action of established HIF-1α inhibitors are non-specific but rather involved in the disruption of signaling cascades upstream or downstream of HIF-1α . Due to the lack of small-molecule inhibitors selectively targeting HIF-1α, the characterization of the transcriptional regulation of HIF-1α might provide a new opportunity for tumor specific therapeutic treatment of highly angiogenic cancers.
Although the HIF-1α promoter contains multiple transcription factor binding sites, one region within this proximal promoter has been shown to regulate about 90% of the transcriptional control of the HIF-1α gene, which is called the poly-Purine/poly-Pyrimidine (pPu/ pPy) tract. Many other pPu/pPy tracts have been discovered within the proximal promoter regions of proto-oncogenes, which also maintain the majority of the transcriptional control of their respective genes . This tract is unique to other promoter domains in that it possesses the ability to form non-B-DNA structures within its guanine-/ cytosinerich (G/C-rich) sequences that, as exemplified by its overwhelming control on transcription, can act as a cis-activating element [15-18]. The non-B-DNA structures that form within the pPu/pPy tracts of these proto-oncogenes are called G-quadruplexes and i-motifs.
G-quadruplexes normally consist of square planes of four guanines, called G-tetrads, held together by Hoogsteen hydrogen bonding . The G-tetrads stack one on top of each other and are stabilized by a monocovalent cation such as potassium or sodium (Figure 1B). Our previous study revealed the formation of a propeller-type parallelstranded intramolecular G-quadruplex in the pPu/pPy tract of the HIF- 1α promoter  similar to c-myc G-rich structure that has a 1:6:1 loop configuration (Figure 1B). The G-quadruplex structure is extremely stable at physiological conditions (i.e. neutral pH, body temperatures, potassium concentration) and there is accumulating evidence of its existence in vivo in human telomeric sequences [21,22].
This article focuses on another secondary structure that has been found to form on the C-rich sequence, complimentary strand to the G-rich sequence, in the pPu/pPy tracts of proto-oncogenes called the i-motif. When hemi-protonated, this cytosine rich strand can form C-C+ intercalated base pairing making the i-motif structure stable. It was conventional wisdom that the hemi-protonation of the C-C+ base pairing was only achieved at acidic conditions. However, recent studies have proposed an increasing amount of evidence that i-motif structures can still form near neutral pH under the structural support of long stretches of intercalating C-C+ base pairs or the interaction of nucleotides along the long loops of some i-motifs , including that of the i-motif that forms within the HIF-1α promoter .
The formation of atypical secondary DNA structures, such as G-quadruplexes and i-motifs, within pPu/pPy tracts in proximal promoter regions is facilitated by negative supercoiling derived from the local unwinding of duplex DNA generated by the progression of transcriptional machinery . This negative supercoiling of the pPu/pPy tract, that is in DNA duplex conformation, accelerates the conversion of B-DNA to single-stranded DNA, which is structurally unstable and readily forms atypical DNA structures. An example of this phenomenon is the dynamic nature of the NHE III1 region, a CT-element, within the c-myc promoter that has been extensively characterized and demonstrated to form the G-quadruplex structure in supercoiled plasmids . The G-quadruplexes and i-motifs which form in the NHEIII1 region of the c-myc promoter have been the most extensively studied atypical DNA structures that form within the promoter regions of proto-oncogenes. Of interest, an array of proteins have been demonstrated to bind to the multifaceted NHE III1 region of the c-myc promoter that also play a role in the transcriptional regulation of the c-myc gene. Of interest to this paper, when the NHEIII1 region is under negative supercoiling stress the duplex DNA becomes unstable and is unwound to a single-stranded state where transcription factors, CNBP and hnRNP K, can bind to the G-rich and C-rich sequences, respectively, and turn on transcription. Although hnRNP K has been found to bind to C-rich sequences in other promoters , the exact binding sites and the structural conformation of this protein/DNA interaction to HIF-1α has not been determined.
In the current study, we have characterized the formation of i-motif structures from the C-rich strand of the HIF-1α proximal promoter, the structures were found to form at least two main loop isomers with conformations of 3:3:3 and 3:4:2, and having a transitional pH of 6.9. Furthermore we have found in this study that the transcription factor heterogeneous nuclear ribonucleoprotein K (hnRNP K) binds to the HIF-1α C-rich sequence in a non-folded single-stranded state and does so in a sequence-specific manner.
Purification of oligonucleotides
DNA oligonucleotides were purchased from Sigma Genosys (The Woodlands, TX) and gel purified prior to use. The sequences and calculated extinction coefficients of all the oligonucleotides used in this study are listed in Table 1. Oligonucleotide concentrations were determined spectrophotometrically from the molar extinction coefficients of individual oligonucleotides calculated according to the nearest neighbor method.
5’-End-labeling of oligonucleotides
The oligonucleotides were end-labeled by T4 polynucleotide kinase and the substrate γ-32P ATP. The reaction mixture was purified of unincorporated radioactive γ-32P ATP with Micro Bio-SpinTM 30 columns (Bio-Rad, Hercules, CA). The labeled oligonucleotides were further purified by a 12% denaturing polyacrylamide gel electrophoresis (PAGE) followed by elution into pure water.
Circular dichroism spectrophotometry of HC27 oligomer
The i-motif-forming oligonucleotides were prepared at a 5 μM concentration in a 50 mM Na-cacodylate buffer (pH 4.4, 5.0, 5.4, 5.9, 6.1, 6.5, 7.1, or 8.0). Due to the buffering capacity of Na-cacodylate buffer (pH 5.0-7.4), experiments at pH 4.4 were also performed using Tris-acetate, which has a buffering capacity at pH 4.4. The Tris-acetate experiments produced results similar to those with the Na-cacodylate buffer (data not shown), and for experimental consistency, only results using Na-cacodylate buffer were discussed. Circular dichroism (CD) spectra were recorded on a Jasco-810 spectropolarimeter (Jasco, Easton, MD) and Olis CD spectrometer (On-Line Instrument Systems, Athens, GA) using a quartz cell of 1-mm optical path length. The scanning rate was set at 100 nm/min, with a response time of 1 s, over a range of 200 to 350 nm. A set of three scans were averaged, smoothed, and baselinecorrected for signal contributions from buffers at 25°C for each sample.
The absorbance versus heating and cooling curves were obtained measuring molar ellipticity at 288 nm (the λ of the maximum molar ellipticity) over a temperature range of 20-95°C, at a scanning rate of 7 min/2°C at which the heating and cooling hysteresis become superimposable , and then by plotting against temperatures for melting temperature (Tm) determination. Tm values were calculated using Boltzmann sigmoidal equation to determine the best-fit curve and V50 as the melting temperature as resulted on Prism (GraphPad Software, Inc).
Bromine footprinting experiments
Bromine footprinting experiments were performed according to the previously published procedures [28,29]. In brief, 5’-endlabeled oligonucleotides with 32P were incubated with a molecular bromine solution, consisting of equal molar concentrations (50 mM) of KHSO5 and KBr, for 30 minutes at room temperature in 10 mM Na-acetate buffer (pH 6.5) containing 140 mM KCl, 50 mM NaCl. The reaction was terminated by adding the stop solution containing 5 μg of calf thymus DNA and 300 mM Na-acetate followed by an ethanol precipitation to purify DNA. The precipitated DNA was dried, dissolved in 1 M piperidine solution, and cleaved with thermal treatment at 95°C for 30 minutes. After piperidine treatment, each sample was dried, resuspended in alkaline dye, and was loaded on a 20% denaturing polyacrylamide sequencing gel. The gel was dried, exposed to a phosphor screen (Bio-Rad), and scanned on a STORM 860 Scanner (GE Healthcare Life Sciences, Pittsburgh, PA) for analysis.
Chromatin Immunoprecipitation assay (ChIP)
The ChiP assays were carried out as described previously with slight modifications . In brief, the cultured A498 cells were crosslinked with formaldehyde solution and then lysed. The cross-linked chromatin was sheared by sonication and the resulting supernatant was purified. The protein–DNA complexes are selectively immunoprecipitated using specific antibodies against RNA Polymerase II (Pol II), Sp1, Nucleolin, hnRNP K, CNBP, and Immunoglobulin G (IgG) (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). The immunoprecipitated complexes were reverse cross-linked and proteins were removed by digestion with proteinase K. The DNA associated with the complex was purified using a Qiagen (Valencia, CA) PCR purification kit and the HIF-1α proximal promoter region was amplified using the PCR prim-ers (forward primer: 5’-TCGCTCGCCATTGGATCTCGAGGA-3’; reverse primer, 5’-ACTGGCCGAAGCGACGAAGAGGGTGA- 3’). The reaction condition for the PCR amplification was of the forty cycles of 95°C for 30 seconds, 52°C for 30 seconds, and 72°C for 30 seconds.
Electrophoretic mobility shift assay of HC27 oligomer
About 1,000 cpm of radio labed oligomer was incubated with increasing amounts of recombinant hnRNP K (Amprox, Carlsbad, CA) in buffer containing 140 mM KCl, 50 mM NaCl, and 10 mM Tris-HCl on ice for 15 minutes to form a protein-DNA complex. The reaction mixtures were then immediately loaded on a native polyacrylamide gel electrophoresis (PAGE) to separate the protein-DNA complex from free DNA by the difference in the electrophoretic mobility. The dissociation constant (Kd) was calculated by quantifying the intensity of each band by densitometry and substituting [A][B] for the values obtained for the DNA/protein complex band and [AB] for the free probe in the Kd formula, Kd=[A][B]/[AB] plotted by the protein molar concentration.
EMSA/Footprint of HC27 and CmidT oligomer in complex with hnRNP K
About 50,000 cpm (1 μM) of the 5’-end labeled HC27 or CmidT oligonucleotide with 32P was incubated with hnRNP K protein (400 nM) in a Tris-HCl (pH 6.5) buffer containing 140 mM KCl and 50 mM NaCl for 15 minutes on ice. The reaction mixtures were then subject to treatment with a solution of bromine (40 μM KHSO5 and 2 mM KBr) for 30 minutes at room temperature. The bromination reaction was terminated by adding 1.5 volume of non-denaturing dye (50% glycerol, 1 mg/mL xylene cyanol bromophenol blue) and each sample was immediately loaded on a 12% non-denaturing polyacrylamide gel to separate the DNA-protein complex from free DNA. The gel fragments corresponding to the DNA-protein complex or and free DNA were cut from the gel and DNA fragments were extracted in a solution containing 0.4 M NH4OAc, 0.2% SDS, 1 mM EDTA, and 1 mM MgCl2 by incubation overnight at 37°C. The DNA solutions were subject to ethanol precipitation/purification and the rest of the experiment was performed according to the non-protein bromine footprint protocol.
The cytosine-rich sequence within the pPu/pPy tract in the HIF-1α promoter forms stable i-motif structures at pH 6.9
Our previous work has demonstrated that G-quadruplex structures, with a 1:6:1 loop configuration, could form in the G-rich sequence that is within the pPu/pPy tract of the HIF-1α proximal promoter . In this study, we determined whether the C-rich sequence of the HIF-1α pPu/pPy tract, complementary strand of the G-rich sequence, was able to form stable i-motif structures near physiological conditions using circular dichroism (CD) spectral analysis. Consistent with signature peaks of previously characterized i-motif structures [28,31-33], the HIF-1α C-rich sequence that is represented by HC27 oligonucleotide (Table 1), forms an i-motif that has a positive peak at 288 nm and a negative peak at 262 nm at pH 7 or lower (Figure 2A). The positive peak begins to shift to a lower wavelength and the negative peak is lost at pH 7.5, indicative of a more unstructured DNA oligonucleotide . The transitional pH was determined at pH 6.9 by plotting the molar ellipticity, of the i-motif positive peak at 288 nm, of each sample at their corresponding pH buffers (Figure 2B).
In addition, we tested the dissociation and association of the i-motif structure formed by the HC27 oligonucleotide by plotting melting and cooling curves done at the 288 nm i-motif signature peak at a rate where the association and dissociation hysteresis were able to be nearly superimposed at different pHs (Figure 2C). This analysis revealed the melting temperature (Tm) of HC27 oligonucleotide to be 33°C at pH 6.5 and 25°C at pH 7 (Figure 2C). As in the CD study described in figures 2D and 2E, the melting temperatures followed by a large decrease in Tm between Tm of 70°C and 65°C at pH 4.5 and 5 to 45°C at pH 6, followed by a slow transition from pH 6.5 and 7, from 33°C to 25°C, respectively. It should be noted that the transitional pH and the melting temperatures of the i-motif formed by the HC27 oligonucleotide, at neutral pH, are higher than previously observed i-motif structures . An independent study was able to demonstrate and correlate the Tm for the HIF-1α i-motif at 25°C or 27°C, depending on the amount of adjacent nucleotides flanking the C-rich run .
We determined if the i-motif structure was forming as an intermolecular or intramolecular species by running the HC27 and CmidT oligonucleotides in a non-denaturing gels at pH 4.5 as well as measuring the Tm of the HC27 oligonucleotide at a high and low concentrations. The non-denaturing gel analysis of HC27 and CmidT oligonucleotides in figure 2F show only one band for each corresponding oligonucleotide and the gel shifting is independent of concentration, 0.2 to 20 uM. As shown in figure 2G, the Tm value of HC27 is independent of the oligonucleotide concentration, thus the oligo forms an intramolecular i-motif species .
The HIF-1α C-rich sequence forms 3:3:3 and 3:4:2 loop isomers at pH 6.5
In order to determine the structural composition of the HIF-1α i-motif we used bromine footprinting, a method developed in our lab, that takes advantage of the preferential reactivity of bromine to cytosines not involved in a steric or electrostatic interaction [25,29]. With this method, we deciphered which cytosines were involved in the C-C+ base pairing and in turn resolved the loop structures of the HIF-1α i-motifs represented by the HC35 and HC27 oligonucleotides (Figures 3A and 3B). We first determined if the fifth run of cytosines (V) on the 3’ end of the pPu/pPy tract (Figure 1A) was involved in the i-motif formation by running a bromine footprint experiment using the HC35 oligonucleotide (Table 1). As shown in figure 3A, the bromine footprinting pattern revealed that the fifth run (V) is completely unprotected from bromine reactivity, thus ruling out its role in the i-motif folding conformation. As shown in figures 3A and 3B, the cytosines involved in runs I and III show a strong protection pattern where C5-C7 in run I and C17-C19 in run III are completely protected from bromine reactivity. In contrast, the cytosines involved in the second (II) and fourth (IV) cytosine runs, consisting of cytosines C9-C13 and C23-C25, show a more complex protection pattern. The high bromine reactivity of C9 indicates that it is involved in the loop region, however C10 and C13 both show partial bromination which can suggest the shifting of these cytosines from the loop regions to the C-C+ base paired regions. C23, and to a lesser extent C25, in run IV also show some bromination when compared to C22 that is in the same loop area. This suggests these cytosines are involved in the shifting of C-C+ interactions, however, since the bromination is not comparable to that of C12 and C10 in run II (Figures 3A and 3B), we interpret the bromination is due to the on and off C-C+ interactions that are shifting between the two cytosine runs. Following these bromine reactivity patterns there are likely two main forms of i-motif structures forming in equilibrium consisting of the following loop conformations 3:3:3 and 3:4:2, with a transitional i-motif structure having a 4:4:3 loop conformation (Figure 3E). Although the predominant structure and the intercalating topology must be determined by NMR mutational studies, we believe the 3:3:3 conformation of the three loop isomers is the predominant structure since cytosine run IV in HC35 (Figure 3A) shows a clear protection of three cytosines (C23-C25) and the three less brominated cytosines in run II are C11, C12 and C13. The high conservation of protection patterns within HC35 and HC27 suggest that the i-motif structures that form in HC35 oligonucleotide also form in the HC27 oligonucleotide. Brazier et al., published preliminary NMR experiments with the same oligonucleotide sequence and have determined 3:3:3 and 3:4:2 loop conformations are amongst the possible loop conformers for the i-motif formation within the HIF-1α pPu/pPy-tract .
We developed a mutant oligonucleotide, CmidT, modeled after the C-C+ base pairs in the 3:3:3 i-motif conformation that cannot form an i-motif structure and was used as a non-i-motif control in these studies. The CmidT oligonucleotide has a mutation of a thymine in place of each middle cytosine involved in the C-C+ base paired cytosine runs of the 3:3:3 i-motif isomer depicted by asterisks in figures 3C, 3D and 3E. The bromine footprinting analysis of CmidT shows complete bromination of all cytosine runs in a non-specific manner (Figure 3C). Although bromine can react to pyrimidine residues thymine is less susceptible to reactivity with bromine when compared to cytosine thus the band intensity of each pyrimidine is not equally distributed throughout the footprinting gel for the CmidT oligonucleotide.
Mutational analysis of the HIF-1α C-rich sequence
To confirm the loop isomers developed from our bromine footprinting studies, we conducted a series of single cytosine-tothymine (C to T) base mutational studies and compared their thermal stability by melting temperature (Tm) measurements at 288 nm, the wavelength at which the positive peak of the i-motif signature is observed. As shown in figure 4, when the cytosines in the shortest cytosine run (I) are mutated the results display the lowest melting temperatures collectively, when compared to the C to T mutations in the other three cytosine runs, suggesting the importance of these cytosines in their involvement on the i-motif structure. As observed in the bromine footprinting data (Figure 3), the cytosines in the core of the i-motif structure, C6, C11, C18-19 and C23-24, have the lowest Tms, meaning the i-motif will unfold at a lower temperature when these cytosines are mutated to thymines. The outer most cytosines tend to have a higher melting temperature, when compared to the wild-type and the inner most cytosines, which remain at least 10°C lower than the wild-type Tm. This swing pattern, of low Tms in the core cytosines and high Tms in the outer cytosines, exemplifies the flexibility of the C-C+ base pairing configuration to neighboring cytosines when runs of four or more cytosines are involved. Unexpectedly, when the only cytosine in the loop region (C15) was mutated, it produced a much higher Tm, 40°C, than that of the wild-type, 33°C, oligonucleotide. The mutated cytosine-to-thymidine nucleotide might be interacting with other bases in the loop-capping region of the i-motif structure, therefore increasing its stability. A thymidine-triad in the loop-capping region of other secondary structures have been shown to be of importance to the stability of G-quadruplexes .
HnRNP K and nucleolin bind to the HIF-1α promoter in vivo
Since the identification of specific-protein 1 (Sp1) as a major transcription factor regulating HIF-1α activity  very little efforts have been made to characterize any additional transcription factors interacting within the pPu/pPy tract of the HIF-1α promoter that regulate its transcription. This pPu/pPy tract of the HIF-1α promoter (-85 to -65 bp relative to the starting site) has been shown to be a major contributor to transcriptional activity and have found to contain other putative binding sites for transcription factors such as AP-1, AP-2, and NFκB . Therefore, we investigated whether hnRNP K and nucleolin could bind to the multi-faceted DNA structures within this tract in vitro as well as in vivo thus supporting our recent evidence of these two transcription factors binding to the C- and G-rich strands, respectively, of the pPu/pPy tract in the HIF-1α promoter . We also investigated the interaction of CCHC-type zinc finger nucleic acid binding protein (CNBP) with the pPu/pPy tract of the HIF-1α promoter since it was known to bind to single-stranded G-rich sequences and act as a transcription factor for the c-myc promoter  and has been recently reported to promote G-quadruplex formation . Since, Sp1 has been described as a major transcription factor for the HIF-1α promoter  and the pPu/pPy tract contains one Sp1 binding motif in its sequence, 5’-(G/T)GGGCGG(G/A)(G/A)(C/T)-3’ (GC box element) (Figure 1A), we used Sp1 as a positive control in our experiments.
As shown in figures 5A and 5B, Sp1, hnRNP K and nucleolin were identified as major transcription factor that interact with the HIF-1α proximal promoter region in vivo. However, we ruled out CNBP as a potential transcription factor candidate interacting with the promoter of HIF-1α gene in vivo since the CNBP sample showed very weak amplification product corresponding to the pPu/pPy tract. However, both Sp1 and hnRNP K samples showed a strong and specific amplification products corresponding to the HIF-1α proximal promoter region (Figure 5A). Furthermore, the result of this experiment revealed that nucleolin specifically binds to the HIF-1α G-rich region in vivo.
HnRNP K binds to the C-rich sequence of HIF-1α promoter in a sequence specific single-stranded state
Although the pPu/pPy tract of the HIF-1α proximal promoter accounts for 90% of its transcriptional control , there is scant information that exists on transcription factors that bind to this region. hnRNP K is a known C-rich binding protein  and can act as a transcription factor as in the case of c-myc and VEGF [26,41], genes whose promoters also contain atypical secondary structures such as G-quadruplexes and i-motifs within their pPu/pPy tract [28,42]. Therefore, we studied the interaction of hnRNP K to the C-rich sequence in the pPu/pPy tract of the HIF-1α promoter, specifically, we characterized its interaction to cytosines within oligonucleotides which could form the wild-type HIF-1α i-motif (HC27) and its noni- motif forming mutant (CmidT). We first conducted an EMSA that showed the shift analysis of hnRNP K complexed to HC27 alone and in the presence of the complementary G-rich sequence (HG27) or R23 oligonucleotide, which contains a random sequence of 23 bp in length. As shown by figure 6, a shift is present when 400 nM of hnRNP K is added to the sample containing HC27 oligonucleotide at pH 6.5, this shift is depleted when HG27 is added to the sample and a second shift corresponding to the duplexed probe appears. The depletion of the shift corresponding to the DNA/protein complex by the addition of the HG27 cold oligonucleotide suggests the preferential binding of hnRNP K to HC27 in a single-stranded state, however, whether folded or un-folded was undetermined in this experiment. Evidence of the specificity of hnRNP K binding to the HIF1-α C-rich sequence is seen when the DNA/protein shift was unchanged in the addition of increasing amounts of R23 oligonucleotide to the EMSA sample.
To have a better understanding of the state in which hnRNP K binds to HC27, we performed EMSA gels of hnRNP K and HC27 at pH 6.5 and 8 (Figure 7A). The intensity of the bands in the HC27/hnRNP K EMSA were quantified using densitometry to calculate the dissociation constant (Kd) of hnRNP K to HC27 oligonucleotide at pH 6.5 and 8 (Figure 7B). The binding activity of hnRNP K to single-stranded DNA at pH 7.9 has been previously demonstrated through EMSA analysis , however we cannot rule out the effect of the difference in pH from 6.5 to 8 could have on the protein activity of hnRNP K. The Kd of hnRNP K to HC27 was estimated to approximately 225 nM at pH 6.5 and 110 nM at pH 8 using the graph form of Kd=[A][B]/[AB] where the Kd is the half point that HC27 oligonucleotide has complexed with hnRNP K. Although the accuracy of the Kd value can be improved with higher number of data points, we are able to determine that hnRNP K binds with higher affinity to HC27 at pH 8, when compared to pH 6.5. This suggests the preference of hnRNP K binding to the C-rich sequence at a higher pH, which is shown by CD spectra to be in a nonstructural conformation (Figure 2).
EMSA/Bromine footprint reveals the molecular interactions of hnRNP K to the HIF-1α C-rich sequence
We characterized the binding motif of hnRNP K to the HIF1-α C-rich sequence, represented by HC27 oligonucleotide, as well as the non-i-motif forming CmidT mutant oligonucleotide. As in the i-motif structural studies, we used bromine footprinting to understand which cytosines where protected by bromine reactivity due to the DNA/protein complex interactions. Our study focused on the interaction of hnRNP K binding to the HIF-1α C-rich sequence in the i-motif structural conformation or the non-folded linear state. First, we incubated the radiolabelled HC27 and CmidT probes with or without 400 nM of recombinant hnRNP K protein in the presence of 2 mM bromine solution at pH 6.5. The samples were then run on an EMSA gel in order for the bands corresponding to the DNA/protein complexes and free probes to be separated (Figure 8A), excised from the gel and precipitated to be analyzed on a DNA sequencing gel. As the oligonucleotide becomes more brominated, an increase in steric hinderance of hnRNP K binding to the oligonucleotide is observed. Figure 8A exemplifies the decreasing “bound” band from 0 to 2 mM Br2 and increasing “free” probe at the bottom of the EMSA gel. We focused our efforts on the highest concentration of bromine (2 mM) since it gave is roughly half bound to free probe species that were both able to be separated and further tested for base-pair analysis in a sequencing gel (Figure 8B), although the bound and free probe species at the lower concentrations of bromine were tested with similar results.
As shown in figure 8B, the free HC27 probe in lane I displayed an i-motif footprinting pattern similar to that of the i-motif footprinting pattern displayed at pH 6.5 (Figure 3B). Whereas the HC27 probe complexed to hnRNP K in lane II displayed a distinct pattern to that of the i-motif pattern as compared by the band intensity analysis. For example, cytosines C22, C20 and C12 showed protection from bromination in lane II, which is complexed to hnRNP K, but are completely unprotected in lane I, corresponding to the free probe. It’s important to note that hnRNP K preferentially binds to cytosines that are amongst long stretches of cytosines and not neighbored by noncytosine nucleotides such as C15. Knowing that C15 is in the loop region of the i-motif and is highly brominated when folded in an i-motif and that hnRNP K does not bind preferentially to cytosines in such a microenvironment, we believe the partial protection of C15 in Lane II can be regarded as non-specific bromine protection when in a protein/ DNA complex. In addition C15 seems to be less affected by hnRNP K binding to the HC27 oligonucleotide in terms of its propensity to bromine reactivity when compared to C20. A similar protection pattern was observed in the pPu/pPy tract of the VEGF promoter when complexed to hnRNP K as previously published by our group .
Although hnRNP K bound to the CmidT probe in the EMSA gel (Figure 8A) there did not seem to be a distinct bromine protection pattern as compared to the HC27 probe but rather demonstrated a non-specific overall protection of all cytosines within the CmidT oligonucleotide (Figure 8B, lane IV), very similar to the non-bromine control lane. This suggests the preferential binding of hnRNP K to the wild-type C-rich sequence in a site-specific manner that, we believe, is in a non-folded state.
As the field of atypical secondary DNA structures evolves from its first initiation in the 1970’s, growing evidence of the presence of these structures in the promoters of oncogenes and other important housekeeping genes has warranted the studies uncovering their role in the transcriptional control of their respective promoters. Now, G-quadruplexes are regarded as major types of secondary DNA structures, which are able to form under physiological conditions. There is accumulating evidence of their existence in vivo in human telomeric sequences [21,22] and well as their function as cis-acting elements . We have previously characterized the G-rich sequence of the pPu/pPy tract within the HIF-1α proximal promoter to form a G-quadruplex structure with a 1:6:1 loop conformation. In this study we have shown that the C-rich sequence of this tract can form stable i-motif structures at neutral pH and that the structures are stable above body temperatures at this pH.
The formation and stabilization of the HIF-1α i-motif at physiological conditions raises the question as to the mechanism of action of the stability of C-C+ base pairs independent of low pH levels to allow the hemi-protonation of cytosines at the N3 position. The site mutational analysis demonstrates the ability of the HIF-1α i-motif to “breathe” by shifting the C-C+ predominant base pairs to neighboring cytosines within its long C-rich runs if the preferred cytosines are not available (Figure 4). For example, the mutated cytosines within the shortest run (I) and its corresponding run which it base pairs to (IV) had the lowest melting temperatures, meaning the C to T mutation of these cytosines had the highest impact in the formation of the i-motif structure. On the other hand, cytosine runs II and III had the highest amount of cytosines within their runs and displayed very little change in melting temperature which compared to the wild-type sequence. This pattern displays a unique form of stability when compared to previously characterized C-rich sequences and i-motif structures of known promoters.
Figure 9 demonstrates the folding patterns of previously characterized i-motifs that form within the pPu/pPy tracts of the VEGF, RET, Rb, c-myc and bcl-2 proximal promoters, which have been recent reviewed and identified to belong in two classes of i-motif structures . The authors of this review describe these two classes of i-motif structures as either having short-loop sizes, class I, or long-loop sizes, class II. Class I i-motif structures include the VEGF, RET and Rb i-motifs, which have loop size conformations of 5’-(2:3/4:2)-3’ and transitional pH (or the pH at which half the structural species consists of i-motif and the other half of non-i-motif folded species) of 5.8, 6.4 and 5.9, respectively. As suggested by their transitional pH range, the formation of these types of i-motif structures are believed to be heavily dependent on the presence of long stretches of C-C+ base-pairs. Whereas class II i-motif structures that include the c-myc and bcl-2 i-motifs have higher transitional pH at 6.6 which is believed to be due to the structural conformations of their respective pPu/pPy tracts. For example, the i-motif that forms within the c-myc promoter has been shown to form a 6:2:6 loop conformation that is highly dependent on the negative superhelicity of the NHE III1 region, rather than acidic pH in supercoiled plasmid experiments . The bcl-2 i-motif structure has been shown to form an unusually long loop structure of 8:5:7 conformation that is believed to have capping structures within the bases of its long loops that contribute to its stability . The class II i-motifs have introduced a new level of stability to the intercalated C-C+ base-pair composition through their dependency on other nucleoside interactions and compositions.
In this study we have shown that the i-motifs that form within the pPu/pPy tract of the HIF-1α proximal promoter has the most physiologically relevant transitional pH to date, whose stability cannot be explained by pH dependency or structural conformation that either class I or class II i-motif structures ascribe to. A recently published study has also determined that the unusually stable i-motif structure that arises from the HIF-1α pPu/pPy tract cannot be classified in either Class I or II i-motif structures . Therefore, we believe the HIF-1α i-motifs belong to a third class of i-motif structures which contain medium-loop size conformations. This class of i-motifs generally has 3 or 4 cytosines in each run and its loop sizes are of medium length 2-4 base-pairs long, and tend to be associated with higher transitional pHs. The key to a stable i-motif with medium loops may lie in the flexibility of the structure to use long stretches of cytosine sequences and be able to switch C-C+ base-pair partners at an equilibrium state. Another example is the human telomeric C-rich sequence, which does not have the long cytosine runs like the HIF-1α C-rich sequence but it still follows the 3:3:3 loop size and has a transitional pH of 7, although achieving this transitional pH is only observed at 4°C . Therefore, we believe the i-motif formed by the C-rich sequence in the pPu/pPy tract of the HIF-1α proximal promoter forms the most stable i-motif structures at physiological conditions, that is, at neutral pH and above body temperatures.
Although the techniques used in this study determined the HC27 oligonucleotide to form stable i-motif structures at neutral pH, mutational and NMR studies would be needed to further characterize the conditions at which these i-motif structures are stable. In addition, we cannot completely rule out that other intramolecular and intermolecular structures or multiple interactions within various segments of oligonucleotides are not at play and are responsible for the increased stability of the i-motif structure.
The discovery of an i-motif that is able to form at neutral pH suggests the existence of this structure is physiologically relevant and can be available to recruit transcription factors to the multifaceted pPu/ pPy tract of the HIF-1α proximal promoter. Therefore, we investigated the possibility of hnRNP K, a known C-rich binding protein, binding to the C-rich sequence in the pPu/pPy tract of the HIF-1α promoter. Here, we show that the four cytosine runs in the HIF1-α pPu/pPy tract that are involved in the i-motif formation also serve as binding motifs for hnRNP K and its three KH domains, and to a lesser extent, its RGG domain (Figure 8B).
EMSA analysis done at pH 6.5 and 8 suggest that hnRNP K binds to the C-rich sequence in an unfolded state (Figure 7). Specifically, the graphs which plot the [A][B]/[AB] versus the concentration of hnRNP K (Figure 7B) do not follow a true linear plot indicative of a classical Michaelis–Menten equation but rather of a sigmoidal plot, which suggests cooperative binding between the KH domains and the RGG domain to each cytosine run. The production of a sigmoidal plot of hnRNP K binding to the C-rich sequence may also indicate that hnRNP K unfolds the i-motif structure since binding of one KH domain to one cytosine run increases binding of the other KH domains. Availability of one cytosine run binding to a KH domain in hnRNP K may be achieved between the switching of C13 and C10 in C-C+ interactions in cytosine run II from the 3:3:3 i-motif isomer to the 3:4:2 i-motif isomer.
The detailed binding motif, along with the competitive EMSA analysis suggests the involvement of hnRNP K in the transcriptional control of HIF1-α via the i-motif structure. While the levels of HIF-1α mRNA should be analyzed while under the overexpression of hnRNP K protein, this will be the subject of future studies in order to explain the extent of control of hnRNP K on HIF-1α transcription. The dynamic nature of the HIF-1α pPu/pPy tract, demonstrated by the i-motif and G-quadruplex structures, suggests the possibility of drug targeting HIF-1α at a transcriptional level. By designing small molecules that interact with these atypical DNA structures and ultimately inhibit their interaction with transcription factors such as hnRNP K, these can serve as lead models in drug discovery for novel anti-angiogenic compounds.
In summary, we examined the C-rich sequence within the HIF- 1α pPu/pPy tract and determined that, despite the dependence of hemi-protonation of the C-C+ base pairing on low pH environments, the C-rich sequence forms stable i-motif structures at pH 7 with 3:3:3, 4:4:3 and 3:4:2 loop conformations. In addition, we determined hnRNP K binds to the HIF-1α C-rich sequence in a non-folded single stranded state. The exact binding sequence was determined using EMSA coupled with bromine footprinting analysis which depicted a protection pattern that is distinct from that of the i-motif footprint. Interestingly, the CmidT mutant oligonucleotide also bound to hnRNP K, although it could not form an i-motif structure, and also seemed to be protected against bromine reactivity in the EMSA/footprinting analysis. However, the CmidT oligonucleotide did not exhibit the same protection patterns as that of the HC27 complexed to the hnRNP K protein and rather displayed non-specific protection throughout the oligonucleotide.
We thank Dr. David Bishop in his contribution to the preparation of the final figures that are included in this article. The National Institutes of Health (CA109069) funded this work. Dr. Diana J. Uribe was a recipient of a Minority Supplement Grant (CA109069S) provided by the Research Supplements to Promote Diversity in Health Related Research program at the National Institutes of Health.
|Figure 1||Figure 2||Figure 3||Figure 4||Figure 5|
|Figure 6||Figure 7||Figure 8||Figure 9|