Recent years have witnessed rapid progress in molecular imaging of living systems including the development of specific imaging probes and cross-sectional optical imaging (reviewed in19-21
). The development of imaging sensors designed for imaging of molecular processes at the intracellular level usually requires a “knock-in” of special genetic reporters22, 23
. The transformation of traditional biochemical assays that do not require such reporters into in vivo
applicable methods still represents a serious challenge. For example, biochemical detection of nucleic acid-protein binding can be performed in complex mixtures but requires separating the components for positive identification of the formation of the protein-nucleic acid complex24
. With the exception of surface plasmon resonance25, 26
, the identification is conditional to successful labeling of nucleic acids or labeling of antibodies that interact with the nucleic-acid bound proteins. The first labeling strategy is suitable for the screening of the unknown nucleic-acid binding proteins, the second one requires a priori
knowledge of the binding complex composition and depends on the availability of specific antibodies for positive identification of these proteins. The detection of protein/nucleic acid interactions under native conditions and/or using homogenous assays would require sensors that change their properties upon binding of the protein. These changes are generally limited to subtle conformational transitions. Therefore, the detection of binding based on protein-induced changes in nucleic-acid conformation is a complex task requiring carefully designed sensors. Among all potential strategies, the development of fluorescent labelled duplex sensors appears most promising. There are several known reports utilzing aptamers27
or triplex-forming oligonucleotides28
that are capable of undergoing transitions usually resulting in either a decrease, or an increase of FRET between the pair of fluorescent dyes linked to the sensor. Linking of fluorophores to ODNs for the purpose of synthesizing molecular beacons is usually limited to the modifications of ODN 3′- and 5′- ends5
, or covalent linking of the dyes to the nucleic acid bases, which in turn, could potentially interfere with base pairing. We circumvented this potential limitation by using internucleoside linkers that allowed conjugating amine-reactive dyes between any two nucleotide bases without interfering with the base pairing. We synthesized complementary oligonucleotides with hydrophilic internucleoside amino linkers that were introduced during the synthesis using two phosphoramidite synthons1
(see ). The synthesis resulted in deprotected ODNs carrying a free amino group positioned between dG and dC and any other nucleoside that underwent further covalent modification. The modification yielded ODN duplexes carrying either donor (Cy5.5) or acceptor (800CW or Cy7 ) cyanine dyes. The synthesized ODNs formed duplexes with complementary ODNs carrying either acceptor or donor dyes linked at precise positions shown in . The duplex with a “specific” NF-κB binding site contained a binding sequence derived from Ig κB -box or from IFNβ-κB binding site, with both sites exchibiting identical DNA duplex-binding affinity29
. The control duplex sensors (D4 and D9, see Supplementary Table 1S
) contained “non-specific” sequences that did not include NF-κB consensus sequence. During sensor optimization we found that by using two internucleoside amino linkers within complementary ODNs we could obtain various duplexes with identical chemical composition and either different, or identical base pair separation lengths between the donor and acceptor fluorophores (). These chemically identical or nearly identical duplexes (e.g. D1, D2, D3 ) could have completely different optical properties. If the donor and acceptor were separated by at least 10 nucleotide base pairs, the pair of donor/acceptor cyanine dyes was capable of efficient FRET as determined by measuring the relative changes of donor and acceptor fluorescence intensities (, duplex D1). However, if the separation distance was decreased to 7-8 base pairs (bp), and each one of the two dyes was closer to the 5′-end of the corresponding ODN than to the 3′-end (as shown in ) the duplexes were no longer fluorescent (, duplex D2). Instead, fluorescence intensities of the donor and acceptor linked to D2 were dramatically decreased (). Similar fluorescence quenching effect was observed when the donor- and acceptor dye linked ODNs were switched around (duplex D2A, ). Unlike the absorbance spectra of a fluorescent duplex D3 (), the spectra of quenched duplexes did not resemble the superimposed spectra of the two individual dye-linked ODNs (see ). In contrast, a new absorbance peak with a maximum of 645 nm was present (), suggesting the potential static close quenching due to formation of a fluorophore dimer. Fluorescence intensities of both donor and acceptor dyes in such duplexes was decreased regardless of whether they were excited in the regular or FRET modes (). At the same separation distance of 7-8 bp, but in the opposite reciprocal orientation along the duplex (, D3), we observed the acceptor fluorescence in emissive FRET mode () similar to that of the sensor with 10 bp separation of fluorophores (, D1).
Fig.1 a) oligodeoxynucleotide (ODN) modifiers: dC- and dG-amidite synthons used to covalently link near-infrared donor and acceptor cyanine fluorophores via the hydrophilic tri(ethylene glycol) amino linkers after ODN deprotection; b) donor and acceptor cyanine (more ...)
The dependence of fluorescence on separation and reciprocal orientation of the donor/acceptor pair in representative ODN duplex sensors.
Fig.2 Spectra of duplex sensors: a) superimposed fluorescence spectra of D2-Cy5.5 (black) D2-800CW (green); D2-Cy5.5/800CW (purple) demonstrating the close quenching of cyanines excited at 675 nm. Red traced spectrum shows fluorescence of 800CW excited at 774 (more ...)
The models of duplexes with two covalently linked cyanine dyes obtained using Molecular Operating Environment (MOE) suggested that due to the double helical conformation of the fluorophore-labeled duplex, a pair of covalently linked fluorophores separated by 7-8 bp on antiparallel DNA strands will be positioned across from each other in the major groove of the duplex only if each one of the fluorophores is linked closer to the 5′-end of ODN than the to the 3′-end (). In this configuration the fluorophores can form dimers in close proximity to each other in a non-parallel orientation (). The analysis of the available X-ray crystallography data for NF-κB/DNA complexes revealed that amino acid residues within DNA-binding polypeptide chains of both p50 and p65 NF-κB proteins are localized within the major groove of duplex and can potentially perturb the formation of the fluorophore dimers (). We further performed dual wavelength imaging of EMSA gels to verify that both single dye- and double fluorophore-labelled duplexes are capable of binding NF-κB proteins (). Upon the addition of a mixture of recombinant NF-κB proteins p50 and p65 a characteristic shift was observed indicating the formation of a complex with the duplex sensors (lanes 1-4). Similar shifts were observed if HeLa nuclear lysates were used instead of p50/p65 mixture. Due to fluorescence quenching, the formation of the complex was mostly obscured (lanes 5-8) and had to be verified by using fluorescence emitting duplexes (lanes 9-12). HeLa nuclear lysates were clearly an adequate model protein mixture for comparing fluorescent properties of complexes and free duplexes. There was also evidence of more efficient binding of HeLa proteins to the specific duplex containing Ig κB box sequence than to the control duplex. However, the overall ODN sequence-independent level of protein binding in those lysates was very high (compare lanes 10 and 12) suggesting cautious interpretation of the lysate experiments.
Fig. 3 Representative molecular models built in MOE as described in Materials and Methods: a) - closely interacting dyes in the non-parallel aggregate of Cy5.5 and 800CW dyes within the major groove of ODN duplex resulting in the lack of fluorescence emission (more ...)
Fig. 4 Fluorescent analysis of protein binding to the duplex sensors: a) electrophoretic mobility shift assay in the (+) presence or (−) absence of p50/p65 or HeLa cell nuclear lysate and the following sensors: D3-Cy5.5 (lanes 1,2); D3-Cy5.5/800CW (lanes (more ...)
The above results indirectly suggested that the binding of transcription factor proteins to fluorescent duplex sensors can be followed by measuring changes in fluorescent properties of the cyanine dye pairs. This was accomplished initially by tracking fluorescence intensity increase after the addition of recombinant p50 and p65 NF-κB proteins to duplex sensors in solution at 1:6 molar ratios (ODN duplex/protein, mol/mol). Initially EMSA was used for testing of the binding of p50/p65 recombinant proteins to both fluorescent and quenched duplex sensors. We investigated two quenched duplex sensors carrying the Cy5.5/800CW pair at two different locations along the duplex (D2 and D10, see Supplemental information) and a total of three sensors with different structures resulting in radiative FRET sensors (D6, D7, D8, ). We excited Cy5.5 fluorescence at 675 nm and measured: 1) relative change of the donor fluorescence at 700 nm and; 2) relative acceptor fluorescence change at 800 nm. Both fluorescent and quenched duplex sensors showed Cy5.5 donor and 800CW acceptor fluorescence intensity increase after the addition of p50/p65 NF-κB protein mixture. The overall changes in fluorescence intensity were nearly instantaneous and stable over a period of 40-60 min (the total time of observation). The overall relative increase of 800CW fluorescence intensity was greater in quenched duplex sensors () if compared to fluorescence emitting duplexes. The relative 800CW fluorescence intensity change in FRET mode () was greater than the dequenching of the donor dye, ). In both cases quenched duplexes showed higher ratios of fluorescence intensities measured prior to and after the addition of NF-κB proteins than fluorescence emitting duplexes (). The latter showed only marginally higher increases of fluorescence intensities if compared to D2 carrying only one fluorophore instead of a dye pair. The lack of measurable changes of single fluorophore fluorescence were observed in both FRET () and Cy5.5 relative increase of Cy5.5 fluorescence signal (i.e. the donor fluorescence () modes.
The results of our experiments suggested that the observed differences in fluorescence intensities should be detectable using a time-resolved fluorescence lifetime (FL-TD) imaging setup. It should be noted that there are multiple advantages of FL-TD strategy over other fluorescent assays: (1) unlike continuous waveform (CW) excitation of fluorescence, the FL-TD approach is more robust and quantitative and is not affected by the sample concentration and excitation intensity, which are in many cases difficult to control; (2) FL-TD is also generally independent on the excitation wavelength; (3) FL-TD is not affected by excitation leakage into the fluorescence filter, whereas in CW excitation this can be a major source of light contamination that is difficult to eliminate (the presence of excitation leakage is readily identified in FL-TD as a sharp peak in the initial rise of temporal decay profile (see ) whereas the long time-decay is unaffected by the initial rise and can be used for the lifetime fitting; (4) the lifetime assays can potentially be performed in multiple formats: on a chip or in solution by simple mixing of the sensor with the sample in a plate.
Fig. 5 Sensitivity of FL-TD measurements. a) Cy5.5 fluorescence decay curves showing a bi-exponential behaviour due to varying proportions of D3-Cy5.5 and D3-Cy5.5/Cy7; b) the comparison of the true fractional content (in %) of a single Cy5.5 dye duplex sensor (more ...)
Since FL is exquisitely sensitive to the changes of microenvironment, we tested whether the sensors based on close positioning of donor/acceptor pair of fluorophores would “sense” the protein binding events. Using a time-domain setup30
we were able to measure FL in solutions containing between 50 and 100 nM of Cy5.5 and/or Cy7 and 800CW (2.5-5 pmol fluorophore/sample). Initially we measured FL of the fluorophores that were used for linking to ODNs. Both fluorophores, as expected, showed some FL lengthening after linking them to the duplexes due to the decrease of rotational freedom and the increase of rotational correlation time31
(Supplementary Table 2S
). Furthermore, the FL increased to the baseline average of 1.19 ns in the case of Cy5.5 after hybridizing Cy5.5-labeled ODN to a complementary ODN. This property was not shared by acceptor fluorophores. However, their FL values also increased slightly after the conjugation to ODNs. We further determined the sensitivity of FL-TD to the presence of “dequenched” duplex sensors in mixtures that contained sensor populations with long and short FL. Towards this goal we used model mixtures of fluorescence emitting FRET (Cy5.5/Cy7) and emitting non-FRET (Cy5.5 only) duplex sensors (D3, ). In these mixtures, FL of Cy5.5 was always shorter in the case of FRET (D3-Cy5.5/Cy7, ). Biexponential data fit of fluorescence decay curves enabled fractional analysis of the mixtures containing various amounts of long-lifetime components (D3-Cy5.5) in mixtures with short lifetime components (D3-Cy5.5/Cy7, ). The FL-TD approach showed a sensitivity limit of approximately 10% of non-FRET duplex in 50 μl samples of mixtures with a total concentration of 100 nM of duplex sensor. Under the above conditions, the accurate two-component lifetime unmixing was possible at 25% duplex-linked fluorophore in the absence of FRET, i.e. D3-Cy5.5.
The initial FL-TD experiments with recombinant NF-κB proteins showed that the addition of 3-fold molar excess of p50 and p65 per mole of D3 duplex sensor carrying Cy5.5 dye and FRET acceptor dye resulted in a measurable increase of Cy5.5 dye FL (). Fluorescence lifetime of the donor fluorophore Cy5.5 increased by 0.08 ns in the case of Cy7, and by 0.15 ns in the case of 800CW acceptors, respectively. The 35 p50/p65 binding resulted in higher FL changes if we used non-fluorescent quenched sensors (e.g. duplex sensor D2, ). The difference between the baseline FL and the FL of the donor Cy5.5 dye measured after adding p50/p65 to D2 increased to 0.21 ns (0.19 ns in the case of Cy5.5/Cy7 pair) compared to FL values measured using emitting FRET sensors (e.g. D3). The addition of the excess of NF-κB positive HeLa nuclear extract resulted in FL differences exceeding 0.4 ns in duplex sensors carrying either Cy5.5/800CW or Cy5.5/Cy7 pairs (). The analysis of electrophoretic migration shift assay results showed that unlike recombinant NF-κB proteins, HeLa cell nuclear extract caused a 100% shifting of the ODN duplex band under the identical electrophoresis conditions (). The binding of the proteins comprising HeLa nuclear extract to duplex sensors was very efficient but non-specific since we observed a shift in migration of control sequence D9 that contained no NF-κB binding sequence. FL measurements in FRET mode (i.e. at the excitation of 650 nm and emission at >800 nm) showed higher specificity of p50/p65 interaction with the NF-κB binding sensor D2 than the control sensor D4 (). The average non-specific change of FL (0.03 ns) was 4-times lower than specific (0.12 ns) change of FL in the presence of p50/p65 protein mixture. The observed differences in average FL values were further visualized by scanning the surface of the samples and representing the FL data in pseudo-colour maps ().
Table 2 Fluorescence lifetimes measured at 650/716 band pass (donor dye Cy5.5) and 650/800 long pass (acceptor dye, 800CW and Cy7) in duplex sensors in the absence and in the presence of DNA binding proteins. The results of 2-3 independent measurements shown (more ...)
Fig. 6 Pseudo-colour maps showing FL distribution in experimental and control ODN sensor samples in the absence (−) or in the presence (+) of p50/p65 mixture measured in FRET mode. The sample FL scanning results obtained using the experimental NF-κB (more ...)