Dual FRET molecular beacons
Three molecular beacon FRET pairs were designed, synthesized and tested in solution. Each FRET probe pair consisted of two molecular beacons, one labeled with a donor fluorophore (donor beacon) and a second labeled with an acceptor fluorophore (acceptor beacon). These molecular beacons were designed to hybridize to adjacent regions on an mRNA target so that the two fluorophores will lie within the FRET range (~6 nm) when probe/target binding occurs for both beacons. Excitation of the donor fluorophore then results in fluorescence emission at a wavelength characteristic of the acceptor fluorophore, which serves as a positive FRET signal readily differentiable from non-FRET false positive signals due to probe degradation and non-specific probe opening (16
). As shown in Table , dual FRET molecular beacon pairs were designed in a shared-stem fashion (18
), i.e. the sequence of the fluorophore-attached arm of the stem (Fig. ) is complementary to the target so that it participates in both stem formation and target hybridization. This design was chosen to help fix the relative distance between the donor and acceptor fluorophores and improve energy transfer efficiency. For all FRET molecular beacons pairs, Cy3 (peak excitation at 545 nm) and Cy5 (peak emission at 665 nm) were used as the donor and acceptor fluorophores, respectively, and BHQ-2 and BHQ-3 were used as quenchers for the donor and acceptor molecular beacons, respectively. One pair of molecular beacons targets a segment of the wild-type K-ras gene (Table ) whose codon 12 mutations are involved in the pathogenesis of many cancers (19
). A member of the GTPase family, K-ras is involved in regulating cell growth, proliferation and differentiation (20
). As shown in Table , the target sequence for K-ras-targeting donor beacon has 17 bases, and that for the acceptor beacon has 19 bases. The other pair of molecular beacons targets the survivin gene, which is a member in the inhibitor of apoptosis protein (IAP) family (21
). The target sequences for survivin-targeting donor and acceptor molecular beacons are, respectively, 15 bases and 16 bases long. For both K-ras- and survivin-targeting molecular beacon pairs, the stem size is five bases, and the gap size between the donor and acceptor beacons on a target is four bases. For use in negative controls, we also designed Cy3- and Cy5-labeled ‘random’-sequence molecular beacons (‘random beacon’) whose specific 16-base target sequence does not match with any mammalian gene. Note that both the donor (Cy3-labeled) and acceptor (Cy5-labeled) random beacons have an identical sequence, and the ‘random sequence’ target is for a signal beacon only (Table ).
Solution studies of FRET signal and specificity
In-solution probe–target hybridization studies were carried out to determine the extent of energy transfer between, and signal-to-background ratio of, dual FRET molecular beacons, as well as the specificity of the random beacon. Shown in Figure A are five fluorescence emission spectra of molecular beacons targeting wild-type K-ras under Cy3 excitation (545 nm); they were generated by having: (i) donor beacons only in the presence of target (blue curve), representing the signal of a single beacon assay; (ii) both donor and acceptor beacons in solution with target (green curve), representing the FRET signal; (iii) donor beacons only without target (red curve), representing the background of a single-beacon assay; (iv) donor and acceptor beacons in solution without target (light blue curve), representing the background of a FRET assay and (v) acceptor beacons with target (black curve). These results imply that, even when a large amount of single (donor or acceptor) molecular beacons are degraded by nucleases or open due to non-specific interactions, the resulting fluorescence signal at the FRET detection wavelength of 665 nm (light blue curve) is still much lower than the FRET signal (green curve). The FRET signal-to-background ratio (green curve versus light blue curve at 665 nm) is approximately 9.0. The survivin-targeting molecular beacons under Cy3 excitation (545 nm) exhibited essentially the same spectroscopic features, as shown by the five curves in Figure B. When the ‘random’ beacons were mixed, respectively, with the complementary random-sequence target, or the wild-type K-ras target, or the survivin target, only targets complementary to the probe sequence of random beacons gave a strong fluorescence signal, other targets gave very low background (Fig. C), confirming the high specificity of random sequence molecular beacons.
Figure 2 Solution studies of probe–target hybridization of dual FRET molecular beacons. (A) Fluorescence emission spectra of K-ras targeting molecular beacons under Cy3 excitation (545 nm). The blue curve was generated by having donor beacons only in (more ...)
K-ras mRNA detection in normally-growing and stimulated HDF cells
In order to demonstrate the advantages of the dual FRET molecular beacon approach, living cell assays were performed with both normally growing and stimulated HDF cells. To increase the K-ras mRNA expression level, HDF cells were first starved for 24 h and then stimulated with serum for 8 h before molecular beacon delivery (22
). Cells were permeabilized using streptolysin O (SLO) (17
) and were exposed to either Cy3-labeled random-sequence or K-ras-targeting single molecular beacons, or Cy3- and Cy5-labeled random-sequence or K-ras-targeting donor and acceptor molecular beacon pairs. The resulting fluorescence signal was observed under Cy3 excitation (545 nm) 1 h after beacon delivery with the Cy3 emission channel (570 nm) for single beacon assays and Cy5 emission channel (665 nm) for dual beacon assays (Fig. ). Signals from the single random beacon negative controls (Fig. A and C) should be due entirely to beacon degradation, non-specific interactions and possibly other backgrounds such as autofluorescence of the cell. Signals from single Cy3-labeled (i.e. unpaired donor) K-ras-targeting molecular beacons (Fig. B) were not appreciably greater than those found with random beacons (Fig. A), clearly demonstrating the limitation of using single molecular beacons in RNA detection in living cells, especially when the expression level is relatively low. With stimulated HDF cells, the fluorescence signal level in single beacon K-ras detection increased (Fig. D), but it was still not much higher than the corresponding background signal (Fig. C).
Figure 3 Detection of K-ras mRNA expression in normally-growing and stimulated HDF cells using single donor molecular beacons only (A–D), or dual FRET molecular beacons (E–H). (A and C) Fluorescence signal of single ‘random’ sequence (more ...)
When random beacon FRET pairs were delivered into either normally growing or stimulated HDF cells, the resulting FRET signals detected in the Cy5 channel were very low (Fig. E and G). The expression levels of K-ras mRNA in normally growing and stimulated HDF cells were detected using the dual FRET molecular beacons, and the resulting fluorescence signals are shown in Figure F and H, respectively. Even with unstimulated HDF cells, the fluorescence signal as a result of K-ras mRNA detection (Fig. F) was much higher than the background (Fig. E), indicating that dual FRET molecular beacons are capable of distinguishing true and false positive signals, which is in sharp contrast with the results of single beacon detection shown in Figure A and B. With stimulated HDF cells, the fluorescence signal level as a result of the dual FRET molecular beacon detection of K-ras mRNA had a significant increase, as shown in Figure H. Quantitative analysis of the average signal intensities of the images in Figure F and H gave a factor of 2.25 increase in K-ras expression in stimulated HDF compared with that in normally growing cells, consistent with the result of the RT–PCR assay, which indicated a factor of 1.95 increase in K-ras expression after stimulation. In contrast, single-beacon detection assays yielded an increase of only a factor of 1.2 (Fig. I). This clearly demonstrates that the dual FRET molecular beacons approach has much better detection sensitivity and allows for a more quantitative measurement of relative changes in mRNA level in living cells upon stimulation. This capability is important for both basic biological studies and drug discovery research in which it is critical to quantify changes of gene expression in living cells in response to stimuli including hormones, growth factors, candidate drug molecules and other chemical/mechanical insults.
In comparing the ratios of K-ras mRNA expression in stimulated and unstimulated HDF cells, we made the assumption that the K-ras expression level in different cells imaged simultaneously is roughly the same, which is supported by Figure B, D, F and H. Consequently, although the quantitative analysis of fluorescence intensity was performed based on, and averaged over, only a few cells, the results should remain valid when a large number of cells are examined. Therefore, we believe that the ratios of 1.2 (single-beacon) and 2.25 (dual-beacon) given in Figure I represent fairly accurately the difference in K-ras mRNA detection using single and dual molecular beacon approaches. Further, we believe that the result of dual-beacon FRET detection is more accurate than that of the single-beacon approach, since it yielded a much lower background signal as demonstrated by Figure A, C, E and G. However, when specific mRNA expression has large variations from cell to cell, analyses of fluorescence intensity of only a few cells may not be statistically significant and the results may not be reproducible. In this case, the fluorescence intensity of a large number of cells (say, at least a few hundred cells) must be analyzed in order to obtain accurate quantification of mRNA expression.
Survivin mRNA expression in normal and cancerous cells
To detect survivin mRNA expression in HDF and MIAPaCa-2 cells, survivin-targeting donor beacon alone, survivin-targeting dual FRET molecular beacons, and the random beacons (Table ) were delivered, respectively, into these cells using SLO. The resulting fluorescence signal was visualized 1 h after delivery. Figure A and C displays the fluorescence signal of random beacons in HDF and MIAPaCa-2 cells, respectively, representing the background signal level in each cell type with single beacon detection. When the fluorescence of single (unpaired) survivin-targeting donor beacons were imaged under Cy3 excitation (545 nm) and Cy3 emission detection (570 nm), the signal level in MIAPaCa-2 cells (Fig. B) was similar to that of random beacon (Fig. A), indicating that the signal was mainly due to false positive events. The fluorescence signal of single survivin-targeting molecular beacons in HDF cells was essentially the same as that of the random beacon. Therefore, using single beacons, the true signal of survivin mRNA detection cannot be distinguished from false positive signals.
Figure 4 Detection of survivin mRNA expression in MIApaCa-2 and normal HDF cells using single donor molecular beacons only (A–D), or dual FRET molecular beacons (E–H). (A and C) Fluorescence signal of single ‘random’ sequence molecular (more ...)
Using FRET detection, i.e. with Cy3 excitation and Cy5 emission detection, the signal generated by random beacons was very low in both HDF and MIAPaCa-2 cells, as can be seen from Figure E and G. This implies that the false positive signals due to beacon degradation and non-specific opening can be dramatically reduced using FRET optics. We found that, using dual FRET molecular beacons targeting survivin (Table ), and with the FRET optics (i.e. 545 nm excitation and 570 nm emission detection), the fluorescence signal level in MIAPaCa-2 cells (Fig. F) was much higher than that in HDF cells (Fig. H), with a 250% increase in average signal intensity. This demonstrates that dual FRET molecular beacons have the ability to quantify specific mRNA expression in different cell populations.
Localization of K-ras and survivin mRNA in living cells
We found that the dual FRET molecular beacons approach can give a clear and detailed picture of mRNA localization in living cells which may reveal important information on mRNA processing, transport and protein production. To demonstrate, in Figure A and B, fluorescence images of K-ras mRNA in stimulated HDF cells are shown, indicating an intriguing localization pattern. Evidently, the K-ras mRNA molecules were not randomly distributed but rather well organized and localized in the cytoplasm. It is clear from the fluorescence images that the K-ras mRNA molecules were well distributed in the cytoplasm and followed the cell morphology as indicated by the cable-like portion of an elongated cell in Figure B. When the fluorescence image of a small peripheral region of a cell is expanded, the K-ras mRNAs seem to be localized along a cytoskeletal filament system, possibly the microtubule system. Indeed the co-localization of mRNA with cytoskeletal filaments has been suggested (23
). A similar feature is shown in Figure B in which the image of a small region of a different HDF cell is expanded. We believe that this is the first direct visualization of K-ras mRNA localization in living cells. Surprisingly, we found that survivin mRNA localized in MIAPaCa-2 cell very differently. As shown in Figure C in which the fluorescence image was superimposed with a white light image of the cells, survivin mRNAs seemed to localize in a non-symmetrical pattern within MIAPaCa-2 cells, often to one side of the nucleus of the cell. The intriguing differences in mRNA localization is likely a consequence of the association of mRNA with the cell cytoskeleton (24
) or mitochondria (25
Figure 5 mRNA localization in HDF and MIAPaCa-2 cells. (A and B) Fluorescence images of K-ras mRNA in stimulated HDF cells. Note the filamentous K-ras mRNA localization pattern. (C) A fluorescence image of survivin mRNA localization in MIAPaCa-2 cells. Note that (more ...)
Control study using in situ hybridization
As a control, we performed a fluorescence in situ hybridization (FISH) assay detecting K-ras mRNA in fixed HDF cells. We used a fluorescently labeled linear probe (5′-Cy5-CCTACGCCACCAGCTCC-3′) that has the same probe sequence as the K-ras-targeting donor molecular beacon (Table ). As demonstrated in Figure A, the fluorescence image obtained in the FISH assay of K-ras mRNA detection in HDF cells gave a filamentous localization pattern as well, especially in the cell peripheral region, similar to that shown in Figure A and B, confirming that the mRNA localization revealed in this study is true. However, in the region near the cell nucleus, the fluorescence image as a result of FISH has a high background compared with that of living cell assays using dual FRET molecular beacons. Since the probes entered both the cell cytoplasm and nucleus during FISH, a strong and diffused fluorescence signal appeared in the fixed HDF cell nuclei (Fig. A). As a negative control, we performed a FISH assay with fluorescently labeled linear Poly-A probes (5′-Cy5-AAAAAAAAAAAAAAAAAA-3′) and the resulting background signal was very low, as can be seen from Figure B. This further confirmed that the fluorescence signal observed in our live cell and fixed cell studies of specific mRNA detection was truly due to probe/target hybridization.
Figure 6 Fluorescence in situ hybridization (FISH) studies. (A) Detection of K-ras mRNA in fixed HDF cells using fluorescently labeled linear probes. Note the filamentous localization pattern near the cell peripheral region. (B) A negative control study of the (more ...)
Intracellular distribution of probe/target binding sites
It should be noted that in all of the images shown in Figures –, very little mRNA expression was detected in the cell nucleus. Although our results seemingly contradict the observation that antisense oligonucleotide probes rapidly accumulate in the nucleus (26
), it remains to be seen if such nuclear localization is due to any fundamental biological reason. In fact, when molecular beacons were microinjected into cells, considerably more fluorescence signal was observed in the cytoplasm than nucleus (6
). It is also possible that the intracellular distribution of signal is related to the specific delivery method used. For example, when delivered via the endocytic pathway (e.g. liposome-based transfection), antisense oligonucleotide probes tend to be trapped inside endocytic vesicles and degraded in the endosomes and lysosomes (29
). In this study, we used a toxin-based delivery method so that the probes do not go through endosomes or lysosomes after entering cells through membrane permeabilization. We believe that after internalization, the probes bind to their mRNA targets in the cytoplasm before reaching the nucleus. The probe concentrations we used are significantly lower (at least a factor of 10 lower) than in most of the antisense work (31
); therefore, the probability of driving the probes into the nucleus by a high concentration gradient is much smaller. As a control, we have delivered fluorescently labeled linear oligonucleotides using SLO and most of the signal observed was in the cytoplasm as well (data not shown).
It has been revealed over the last few years that RNA molecules have a much wider range of functions, from physically conveying and interpreting the genetic information of living cells, to essential catalytic roles, to providing structural support for molecular machines, to gene silencing. These activities are controlled by the dynamics of both the expression levels of specific RNAs and their spatial distributions. Although in vitro assays such as DNA microarrays and northern blotting can reflect the relative changes in RNA expression level of a cell population, imaging of specific RNA (including mRNA and microRNA) in living cells in real time can provide essential information on how external stimuli alter the gene expression level in cells, what are the processes of RNA localization, transport and interference, and how RNAs interact with proteins or protein complexes. As demonstrated here, the dual FRET molecular beacons technique can detect endogenous mRNA in living cells rapidly with high specificity, sensitivity and signal-to-background ratio, thus providing a powerful tool to address all these issues. For example, in drug discovery, this method can be used in high-throughput assays to quantify and monitor the dose-dependent changes of specific mRNA expression in response to different candidate drug molecules. In basic biological studies, this method will allow researchers to visualize the dynamics and localization of specific RNAs.
A very intriguing observation in this study is the localization of mRNA in living cells. As demonstrated in Figure , K-ras mRNAs displayed an interesting filament-like localization pattern in HDF cells, with a clear indication of spreading out in the cytoplasm and following the cell morphology. On the other hand, survivin mRNA was localized on one side of the cell nucleus. But why are K-ras and survivin mRNAs localized in such a peculiar way? What are the biological implications of such localization? RNA localization is an evolutionarily conserved phenomenon and may be the first step in targeting proteins to specific locations to facilitate protein–protein interactions (32
). For example, mRNA localization and translation has been found in dendrites and axons in neural cells (33
). It has been indicated that intracellular mRNA localization may involve interactions between targeting signals within the RNA, motor molecules and the cytoskeleton (34
). Although the association of mRNA with microtubules has been suggested based on the results of biochemical assays, there is still a lack of direct confirmation of mRNA localization to cytoskeletal filaments. One difficulty is that the cytoskeleton is a dynamic network and the mechanism of transport for mRNA is largely unknown. Studies have been performed using either GFP fusion proteins that bind to RNA (35
) or by injecting fluorescently labeled mRNAs into cells and tracking them (36
). However, by imaging endogenous mRNAs directly we will be able to gain insights into the rates of RNA synthesis and processing, the dynamics of RNA transport and localization, and RNA–protein interactions.
As demonstrated in this study, the dual FRET molecular beacons approach has the advantage that false positive signals in living cell gene detection due to nucleases and nucleic acid binding proteins can be significantly reduced or even eliminated, leading to high detection sensitivity. This approach is based on simultaneous hybridization of two probes to the same mRNA target so that a FRET signal can be emitted upon proper excitation. Although dual FRET linear probes have been used for living cell mRNA detection (37
), they typically provide a higher background signal than molecular beacons as a result of the fluorescence emission of unbound probes, including donor emission at the detection wavelength and direct acceptor excitation. Further, hairpin probes can provide better specificity, as analyzed in Tsourkas et al
). While the use of dual FRET probes may further increase detection specificity, compared with single molecular beacon assays, the dual-probe approach does require twice as many probes to be delivered into cells and a longer time for probe targeting and hybridization. Therefore, for specific applications where fast detection (~30 min) is crucial, single molecular beacon assays such as that based on peptide-linked molecular beacons (30
) may be more suitable.
As demonstrated in previous studies, the use of 2′-O
-methyl modified molecular beacons has certain potential advantages including improved intracellular stability and faster probe/target hybridization kinetics (39
). However, there are issues and potential drawbacks as well. One issue is that the 2′-O
-methyl modified molecular beacons may be more prone to open in living cells due to nucleic acid binding proteins (12
), therefore generating a higher background. Further, 2′-O
-methyl modified beacons are RNA-like, and therefore may form double-stranded RNA upon probe–target binding and trigger unwanted RNAi in living cells. A major advantage of using 2′-O
-methyl molecular beacons is their enhanced nuclease resistance. However, most of the nucleases are in the endosomes, lysosomes and nucleus (40
). If the probes are delivered directly into the cytoplasm and the endocytic pathway can be avoided, nuclease resistance may not be an essential issue. Although we chose to use unmodified molecular beacons in this study, it is important to compare the performance of molecular beacons with different backbone modifications in living cell gene detection (P.J.Santangelo and G.Bao, unpublished work).
A critical issue in living cell mRNA detection is target accessibility, which is largely controlled by mRNA secondary and tertiary structures and RNA-binding proteins. Specifically, mRNA in a cell almost always has proteins bound to it, which may alter mRNA structure and prevent probe hybridization. Therefore, in selecting the probe sequences of the dual FRET molecular beacons, it is important to avoid targeting sequences that are ‘buried’ inside the tertiary structure or where double-stranded RNA is formed. Although predictions of mRNA secondary structure can be made using existing software (e.g. mfold), they may not be accurate due to limitations of the biophysical models used. Further, we only have very limited knowledge of the sequences occupied dynamically by RNA-binding proteins. Therefore, for each gene to target, it is often necessary to select multiple unique sequences along the target RNA, and have corresponding molecular beacons designed, synthesized and tested in cells to achieve high signal-to-background ratio.
Another important issue in living cell gene detection is the quantification of mRNA expression in single cells. There are many challenges in obtaining an accurate measure of the number of mRNA molecules per cell using molecular beacons (or any imaging method). For example, it is necessary to distinguish true and background signals, determine the fraction of mRNA molecules hybridized with probes, and quantify the possible self-quenching effect of the reporter, especially when mRNA is highly localized. Since the fluorescence intensity of the reporter may be altered by the intracellular environment, it is also necessary to create an internal control by, for example, injecting fluorescently labeled oligonucleotides with known quantity into the same cells and obtaining the corresponding fluorescence intensity (37
A related issue in molecular beacon based gene quantification in living cells is the detection sensitivity, which is dictated not only by probe/target hybridization but also by the properties of fluorophore, fluorescence detection method, optical imaging instrumentation (microscope and camera) and background signal. As a result of the low background in the dual FRET molecular beacons approach, improved detection sensitivity can be achieved compared with FISH, as demonstrated by Figures and . Using microinjection of probe/target duplexes, Sokol et al
. found that the molecular beacon based approach could detect 10 mRNA molecules per cell (6
) when they are concentrated at the same spot. It is estimated that the current approach can reliably detect a few hundred copies of endogenous mRNA per cell in single cell gene detection. With the use of advanced fluorophores such as quantum dots (41
) or lanthanide chelates (16
), fluorescence detection methods such as time-resolved FRET, and more sensitive optical imaging instruments such as that with photon counting, it is possible to detect as low as 1–5 copies of mRNA molecules in a single cell. An alternative approach is to obtain the average mRNA expression over a large number of cells (say one million), similar to that in RT–PCR studies. In this case, a FRET-based flow cytometry assay using Fluorescence Activated Cell Sorter (FACS) could be performed to detect the fluorescence signal level in living cells. While single cell mRNA detection can be used to study more accurately the variation of gene expression in a cell population with different stages in the cell cycle or different disease states, the determination of the average mRNA expression level over a large number of cells may be advantageous in that, when the cell–cell variation of mRNA expression is large, it is statistically more significant and thus more reproducible compared with the mRNA level detected using a relatively small number of cells.