A Genetically Encoded RNA Photosensitizer for Targeted Cell Regulation

Abstract

Genetically encoded RNA devices have emerged for various cellular applications in imaging and bio-sensing, but their functions as precise regulators in living systems are still limited. Inspired by protein photosensitizers, we propose here a Genetically encoded RNA Aptamer-based Photosensitizer (GRAP). Upon illumination, the RNA photosensitizer can controllably generate reactive oxygen species for targeted cell regulation. The GRAP system can be selectively activated by endogenous stimuli and light of different wavelengths. Compared with their protein analogs, GRAP is highly programmable and exhibit reduced off-target effect. Our results indicate that GRAP enables efficient noninvasive target cell ablation with high temporal and spatial precision. We believe this new RNA regulator system will be widely used for optogenetics, targeted cell ablation, subcellular manipulation, and advanced imaging.

Graphical Abstract

Genetically encoded RNA photosensitizer was developed for the controllable generating of ROS. These RNA regulators can be engineered into a smart platform for live-cell imaging, controlled ROS generation, and targeted cell ablation.

Fluorescent proteins and light-sensitive protein devices have transformed the field of live-cell imaging and optogenetics.^[1] These protein devices have been widely used to measure and control cellular functions. Like proteins, RNAs can be genetically encoded and perform diverse roles in living cells.^[2] RNA devices also exhibit several unique features such as the precise and predictable base pairing, the flexible design, and programmable and dynamic interactions.^[3] However, a limited number of functional RNAs are evolved in nature, thus the development and in vivo applications of RNA devices are still lagging far behind protein ones.

The emergence of functional RNA devices has attracted much attention in synthetic biology and cellular analysis. Many of these devices have been developed based on RNA aptamers, which are single-stranded oligonucleotides that can bind their target molecules with high affinity and specificity. For example, a series of genetically encoded fluorogenic RNA aptamers have been systematically evolved in vitro^[4] and been popular used for imaging various target analytes in living cells.^[5] As another example, RNA aptamers that can selectively bind a particular conformation of light-sensitive molecules have been recently identified as photo-responsive switches.^[2b,6] However, genetically encoded RNA devices that can both image target analytes and achieve controllable cell regulation remain largely underdeveloped. In this report, we will introduce a new type of RNA photosensitizer to fill this gap.

Photosensitizers are molecules that can generate reactive oxygen species (ROS) upon illumination, which have been widely used for damaging cell structures and photodynamic therapy.^[7] Traditional chemical photosensitizers suffer from low cell selectivity and off-target toxicity.^[8] In contrast, genetically encoded protein photosensitizers, such as KillerRed, MiniSOG, and FAP-TAPs,^[9] can be directly expressed in a particular cell population for the selective and localized ROS generation.^[10] As a result, these protein photosensitizers are increasingly used in manipulating cell structures and functions by cell ablation, chromophore-assisted light inactivation, and correlative light-electron microscopy.^[8a,11] However, compared with chemical photosensitizers, the ROS generation efficiency of protein photosensitizers are still limited, a high dose of light is required to achieve efficient photosensitization.^[12] In addition, most protein photosensitizers are uncontrollable and constitutively active upon light irradiation.^[13] Ideal photosensitizers should incorporate stimuli-responsive elements to enable “on-demand” activation with distinct molecular signature in target cells.^{[8c, 14]}

We hope to demonstrate in this study an alternative genetically encoded RNA photosensitizer system, which can be used to generate ROS with high target selectivity, controllability, programmability, and efficiency. We named this RNA device as Genetically encoded RNA Aptamer-based Photosensitizer (GRAP). Our idea is to exploit a selective dinitroaniline-binding RNA aptamer (DNB) to in situ activate ROS-generating chemical photosensitizers (Figure 1a). Upon binding with the RNA aptamer, the otherwise quenched photosensitizer will be activated. As a result, the ROS production can be precisely controlled by the RNA aptamer. A variety of existing chemical photosensitizers can be rationally activated using this approach. We have demonstrated that GRAP can function as a versatile platform for live-cell imaging, controlled ROS generation, and targeted cell ablation.

Figure 1.

(a) Schematic of DNB-activated photosensitizers for the controllable generating of reactive oxygen species (ROS). The binding of DNB with dinitroaniline (DN) induces the separation of DN from the attached photosensitizer and restores the ROS generation ability of the photosensitizer. Fluorescence signals of the 2’,7’-dichlorodihydrofluorescein (DCFH) ROS indicator were measured in a solution containing 0 or 20 μM DNB, 0 or 10 μM HD (b) or PD (c), and 50 μM DCFH before and after a 10 min light illumination at 488 nm (for HD) or 640 nm (for PD). Shown are mean and SEM from three replicated experiments.

To develop targetable and activatable photosensitizer, we first conjugated different chemical photosensitizers with a dinitroaniline (DN) moiety. DN has been used as a general contact quencher for various fluorophores.^[15] We predicted that DN might be also useful in quenching the photo-activation of four commonly used photosensitizers including hematoporphyrin IX, pyropheophorbide A, methylene blue, and rose bengal. The corresponding DN-conjugated activatable photosensitizers were named as HD, PD, MD, and RD, respectively. As expected, a 29-fold, 7-fold, 6-fold, and 18-fold quenching in the fluorescence signal was observed for each of these conjugates (Figure S1).

We next asked if the addition of DNB could be used to restore the fluorescence and photosensitizing ability of these conjugates. In the presence of 20 μM DNB, the fluorescence signal of HD, PD, MD, and RD was enhanced by a 16-fold, 3-fold, 2-fold, and 1.4-fold, respectively (Figure S1). Our results indicated that MD cannot easily penetrate cell membranes (data not shown). Considering the efficiency of fluorescence activation and cell membrane permeability, HD and PD were chosen for the following studies.

To test if the ROS generation efficiency of HD and PD can be regulated by the DNB, a 2’,7’-dichlorodihydrofluorescein (DCFH) ROS indicator was used.^[16] After illumination for 10 min (488 nm for HD, 640 nm for PD, 50 mW/cm²), the fluorescence intensity of the DCFH indicator increased by 6.3-fold for the HD/DNB complex and 3.8-fold for the PD/DNB solution (Figure 1b,c). As a control minimal ROS generation was observed in free HD and PD solution upon the same irradiation. The DNB can indeed activate the photosensitizing ability of HD and PD.

We next tested the performance of GRAP in living cells. We first cloned the DNB sequence in a pET28c vector and transformed it into BL21 (DE3)* Escherichia coli (E. coli) cells. After incubation with HD or PD, an intense cellular fluorescence was observed. As a control, cells without expressing DNB (e.g., expressing instead, a fluorogenic RNA aptamer, Broccoli) displayed minimal fluorescence signal (Figure 2a and S2). We further evaluated the efficiency of ROS generation in these E. coli cells using a 2’,7’-dichlorodihydrofluorescein diacetate (DCFH-DA) indicator. After 30 s illumination with 488 nm light, almost all the DNB-expressing cells exhibited obvious ROS production, demonstrating the high efficiency of ROS generation. In contrast, little fluorescence or ROS generation was observed in the Broccoli-expressing cells (Figure S3), indicating the intracellular activity of photosensitizers could be controlled by the DNB.

Figure 2.

(a) Confocal fluorescence images of Broccoli- or DNB-expressing BL21 (DE3)* E. coli cells after incubation with 10 μM HD (E_x/E_m, 488/590 nm) or PD (E_x/E_m, 640/670 nm). Scale bar, 10 μm. (b), (c) The ablation of bacterial cells was calculated based on the percentage of Sytox^™ Blue (SB)-stained E. coli cells. Here, Broccoli- or DNB-expressing cells were incubated with 1 μM SB and 10 μM HD, PD, hematoporphyrin IX (HX), or pyropheophorbide A (PA) in the absence or presence of 5 min irradiation at 488 nm (for HD and HX) or 640 nm (for PD and PA). The percentage of SB-stained cells was determined from 100 cells in each case. Shown are mean and SEM from three replicated experiments. **P<0.01 in two-tailed Student’s t-test. NS, not significant. (d) Confocal fluorescence images of DNB-expressing E. coli cells after 5 min illumination and 1 h incubation with 1 μM SB and 10 μM HD (E_x/E_m, 488/590 nm) or PD (E_x/E_m, 640/670 nm). Scale bar, 10 μm.

To further study if the GRAP system can be used for the targeted cell ablation, we chose to use a Sytox^™ Blue dye to stain the dead/dying cells. In the presence of HD, after 5 min irradiation with 488 nm light, 63% of DNB-expressing cells were Sytox^™ Blue stained, while only 6% of Broccoli-expressing cells were lighted up (Figure 2b and S4). Under the same illumination condition, the photoactive form of hematoporphyrin IX (HX) induced a 65% cell damage, similar as that of the HD/DNB complex. As another control, without light irradiation, almost no cytotoxicity was induced by the HD or the HD/DNB complex, similar as those without HD and HX (Figure 2b and S4).

We also studied the cell damage selectivity of another photosensitizer, PD. Upon 5 min irradiation with 640 nm light, 11% of Broccoli-expressing cells and 46% of DNB-expressing cells were photo-damaged, the latter of which is comparable with that of the photoactive form of pyropheophorbide A (PA), 53% (Figure 2c and S5). The photo-toxicity of both HD and PD can be effectively activated by the DNB for the targeted cell injury. As shown in Figure 2d, a good colocalization between the fluorescence signal of the photosensitizer and the Sytox^™ Blue dye, with Pearson correlation coefficient of 0.80 for HD/DNB and 0.88 for PD/DNB, was observed, which indicated that the targeted cell damage is indeed induced by the cellular activation of photosensitizers. We also applied pulsed illumination (repetitive 1.0 s light and 1.0 s dark) to enhance the photo-toxicity of the GRAP system.^[10b] After 5 min pulsed light exposure, HD/DNB complex induced a 91% cell damage (Figure S6).

To enable stimuli-regulated ROS generation, we designed a modular RNA regulator system that contains three components: a stimuli-recognition sequence, a transducer sequence, and a DNB (Figure 3a). The specific recognition of stimuli will induce the folding and activation of the DNB, which can then restore the photosensitizing ability of HD and PD.

Figure 3.

(a) Schematic of the target RNA-responsive GRAP system. The target RNA hybridizes with the recognition sequence (purple) of the RNA regulator and re-assembles the DNB (blue), which further activates the HD or PD photosensitizer. (b) Confocal fluorescence images of Broccoli- and S4-expressing BL21 (DE3)* E. coli cells after 3 h incubation in the presence of 0 or 20 g/L glucose. Images were taken after 5 min irradiation at 488 nm and 1 h incubation in the presence of 1 μM Sytox ^™Blue (SB) and 10 μM HD. Scale bar, 10 μm. (c) Cellular fluorescence intensity distributions in the experiments as shown in the panel (b). Shown are mean and SEM from 35 individual cells in each case. ***P<0.001 (two-tailed Student’s t-test). (d) Confocal fluorescence images of Broccoli-, C-bDNB-, and bDNB-expressing BL21 (DE3)* E. coli cells after 5 min irradiation at 488 nm and 1 h incubation in the presence of 1 μM SB and 10 μM HD. Scale bar, 10 μm. (e) Cellular fluorescence intensity distributions in the experiments as shown in the panel (d). Shown are mean and SEM from 35 individual cells in each case. ***P<0.001 (two-tailed Student’s t-test).

We first chose a small RNA stimulus, SgrS, which is known to be involved in the bacterial glucose transportation pathway.^[18] Based on in silico simulation, four RNA regulator strands of different transducer sequences, named S1–S4 (Table S1), were designed and in vitro synthesized. With a 2.7-fold activation of the HD fluorescence in the presence of 10 μM SgrS, an optimal regulator S4 was identified (Figure S7a,b). We further cloned S4 into a pET28c vector and transformed it into BL21 (DE3)* cells. It is known that the cellular level of SgrS can be regulated by the addition of glucose.^[19] Indeed, after adding 20 g/L glucose and upon a 5 min illumination, an obvious activation of the HD fluorescence and cell damage (68%) was shown in S4-expressing cell. As a control, under the same condition, Broccoli-expressing cells exhibited minimal fluorescence activation, which was even below that from endogenous SgrS (0 g/L glucose)-induced S4 activation (Figure 3b,c and S7c).

To study if GRAP can be modularly engineered towards other stimuli, we developed another activatable ROS generation system targeting the bglF mRNA. The optimal bglF-responsive RNA regulator (named bDNB) exhibited a 2.6-fold activation of the HD fluorescence in the presence of 10 μM bglF (Figure S8a). After transforming the bDNB-expressing vector in BL21 (DE3)* cells, endogenous bglF-induced HD fluorescence signal can be clearly observed nearby the cell membranes. Indeed, it is known that bglF mRNA are predominantly membrane-bound.^[20] As expected, cells expressing the control RNA regulator (C-bDNB) with a random recognition sequence (Table S1) or Broccoli exhibited no HD fluorescence or photo-toxicity upon illumination (Figure 3d,e and S8b). All these data suggested that GRAP could be programmably activated by different endogenous stimuli.

We next asked whether GRAP could regulate ROS generation in eukaryotic cells. To increase the stability and expression level of RNA regulators in mammalian cells, we first designed a circular DNB (circDNB) system using a recently developed Tornado platform.^[21] After demonstrating the successful RtcB ligase-mediated DNB circularization in a native gel electrophoresis assay (Figure S9a), we wanted to test if circDNB could still activate the photosensitizer HD and PD. Interestingly, circDNB was even more efficient than DNB in activating the fluorescence of HD and PD by 2.3-fold and 2.1-fold, respectively (Figure S9b,c). This result could likely be explained by the stabilized and enhanced folding of DNB after circularization.^[21]

We next cloned the circDNB sequence in a pAVU6+27-Tornado vector, and then transformed into HeLa cells. As expected, bright fluorescence signal of HD and PD was observed in these circDNB-expressing cells, while little fluorescence was shown in the control HeLa cells that express circular Broccoli (circBroccoli) (Figure S10a,b). To test if circDNB could regulate cellular ROS generation, we added DCFH-DA. In the presence of 10 μM HD, after 30 s of 488 nm light irradiation, markedly enhanced DCFH-DA fluorescence was shown in circDNB-expressing cells, but not in circBroccoli-expressing cells (Figure 4a). As another control, negligible ROS generation was observed before irradiation (Figure S11). All these data suggested that circDNB would activate the ROS generation in the mammalian cells upon irradiation.

Figure 4.

(a) Confocal fluorescence images of circBroccoli- or circDNB-expressing HeLa cells after incubation with 10 μM DCFH-DA and 10 pM HD or HX for 30 min. Shown are images taken before and after 30 s light irradiation at 488 nm. Scale bar, 20 μm. (b) Confocal fluorescence images of circBroccoli-or circDNB-expressing HeLa cells after 5 min irradiation at 488 nm and then 1 h incubation with 1 μM Sytox^™ Blue (SB) and 10 μM HD or HX. Scale bar, 20 μm. (c) Cellular SBfluorescence intensity distributions were measured in circBroccoli- or circDNB-expressing HeLa cells after 1 h incubation with 1 μM SB and 10 μM HD or PD, in the absence or presence of 5 min irradiation (488 nm for HD, 640 nm for PD). Shown are mean and SEM from ten individual cells in each case. ***P<0.001 in two-tailed Student’s t-test.

To further study if circDNB can be used to induce targeted cell ablation, we elongated the laser irradiation to 5 min. Indeed, the circDNB-expressing HeLa cells incubated with HD began to lose their normal morphology by swelling and blebbing,^[9c] whereas the circBroccoli-expressing cells or circDNB-expressing cells without HD remained alive and preserved their native shapes (Figure S12). The efficiency of cell damage could be temporally regulated by the illumination time. Most circDNB-expressing cells were damaged after 5 min illumination (Figure S13).

To validate the observed circDNB-induced photo-toxicity, we added the Sytox^™ Blue to stain dying/dead cells. As shown in Figure 4b and S14, the specific cell death was observed with those expressing circDNB, but not circBroccoli-expressing cells. Similarly, targeted cell damage was also observed in the presence of PD. After 640 nm light irradiation for 5 min, selective morphological changes and Sytox^™ Blue fluorescence signal was observed only in the circDNB-expressing cells (Figure S15). The cellular Sytox^™ Blue fluorescence were directly correlated with the HD or PD fluorescence (Figure 4c). Together, these results indicated that GRAP could be used to induce targeted cell toxicity.

We also wondered if the GRAP system could be used for wavelength-selective photosensitizing. We expected that by incubating the circDNB-expressing cells with HD, PD, or a mixture of HD and PD, an orthogonal and multi-color photosensitizer system could be developed. Indeed, in the presence of only HD or PD, photo-induced cell toxicity was observed exclusively after the corresponding 488 nm or 640 nm light irradiation. In another case, when both HD and PD were added, efficient cell damage was shown under either 488 nm or 640 nm light irradiation (Figure 5a). The wavelength selectivity of GRAP could be thus easily tuned without changing the genetically encoded RNA aptamers.

Figure 5.

(a) Confocal fluorescence images of circDNB-expressing HeLa cells after 1 min irradiation at 488 nm or 640 nm and 1 h incubation with 1 μM Sytox^™ Blue (blue) and 10 μM HD (green), 10 μM PD (red), or both. Scale bar, 20 μm. (b) Confocal fluorescence images of miniSOG-, KillerRed-, and circDNB-expressing HeLa cells after 2 min irradiation (at 488 nm, 561 nm, and 488 nm, respectively) and 1 h incubation with 1 μM Sytox^™ Blue. 10 μM HD was added for circDNB-expressing cells. Scale bar, 20 μm.

Genetically encoded protein photosensitizers such as KillerRed and miniSOG have also been widely used for the targeted cell ablation, our next goal is to study if GRAP exhibits similar cell damage efficiency as its protein analogs. Under similar luminous intensity, circDNB-expressing HeLa cells started to exhibit cell damage and morphological changes after 2 min irradiation (Figure 5b and S16), while the KillerRed-expressing cells began to change their shapes after 10 min illumination. In contrast, little cell damage was observed in the miniSOG-expressing cells during this whole process (Figure S17). These results indicated that GRAP is more powerful in mediating cell ablation than the KillerRed and miniSOG system.

Lastly, we asked if stimuli-responsive GRAP system could be constructed in mammalian cells. To design a modular circDNB-based regulator, we inserted a molecular beacon moiety within the circDNB structure (Figure S18a). The formation of molecular beacon prevented the folding of DNB pocket. The loop region of molecular beacon was modularly designed to hybridize with different RNA stimuli. In the presence of target stimuli, the opening of molecular beacon induced the refolding of the DNB. To test this idea, we chose to target survivin mRNA, a cancer biomarker.^[22] After optimizing the molecular beacon sequence, a 2.2-fold activation of the HD fluorescence signal was observed using a circS6 regulator in the presence of 3.0 μM survivin mRNA (Figure S18b).

We further cloned circS6 into the pAVU6+27-Tornado vector, and then transfected into HeLa cells. Endogenous survivin mRNA-induced HD fluorescence activation and photo-toxicity could be clearly observed in the circS6-expressing cells. To confirm that the cell damage was indeed controlled by the survivin mRNA, we treated these cells with YM155, a known survivin inhibitor in HeLa cells.^[23] As expected, the cellular fluorescence signal and damage disappeared after adding YM155 (Figure S18c,d). These results indicated that GRAP could be activated by intracellular RNA stimuli and used for targeted eukaryotic cell ablation.

In sum, we envision that the modular, selective, and highly controllable GRAP system can be broadly applied for the targeted cell regulations. A number of existing chemical photosensitizers can be easily modified with a dinitroaniline moiety and used in this activatable genetically encoded system. This new versatile RNA regulator can dramatically extend the function of RNA devices in the field of chemical and synthetic biology. So far, most genetically encoded RNA devices have not been used in vivo, we are interested in testing the in vivo function of GRAP in the future.