Live-Cell Imaging of Guanosine Tetra- and Pentaphosphate (p)ppGpp with RNA-based Fluorescent Sensors

Abstract

Guanosine tetra- and pentaphosphate, (p)ppGpp, are important alarmone nucleotides that regulate bacterial survival in stressful environment. A direct detection of (p)ppGpp in living cells is critical for our understanding of the mechanism of bacterial stringent response. However, it is still challenging to image cellular (p)ppGpp. Here, we report RNA-based fluorescent sensors for the live-cell imaging of (p)ppGpp. Our sensors are engineered by conjugating a recently identified (p)ppGpp-specific riboswitch with a fluorogenic RNA aptamer, Broccoli. These sensors can be genetically encoded and enable direct monitoring of cellular (p)ppGpp accumulation. Unprecedented information on cell-to-cell variation and cellular dynamics of (p)ppGpp levels is now obtained under different nutritional conditions. These RNA-based sensors can be broadly adapted to study bacterial stringent response.

Graphical Abstract

Genetically encodable fluorescent RNA sensors are engineered to detect (p)ppGpp in live cells. Under nutrient starvation, cells produce and accumulate (p)ppGpp as a response, which can now be observed using these sensors.

(P)ppGpp, often referred to as the “Magic Spot”, is tightly involved in the stringent response in bacteria.^[1-3] When bacterial cells are under environmental threats, such as nutrient starvation, heat shock, and antibiotic treatment, (p)ppGpp is rapidly produced by RSH (RelA/SpoT homolog) enzymes and accumulated to high levels. As an important signaling molecule, (p)ppGpp allows the cells to redistribute resources towards amino acid synthesis and stress survival.^[4] (P)ppGpp can function by directly acting on RNA polymerases to alter their promoter selection and rate of transcription.^[5,6] It can also modulate the activity of a series of transcription factors, enzymes, and regulatory RNAs.^[7,8] Notably, (p)ppGpp was also recently identified to play an integral role in regulating biofilm formation, antibiotic resistance, and persistence.^[9,10]

One major gap in our study of (p)ppGpp’s biological functions is the lack of tools to directly detect (p)ppGpp in living cells. Traditional methods of detecting (p)ppGpp rely on radiolabeling (p)ppGpp with ³²P and measuring with thin layer chromatography^[11-13] or high-performance liquid chromatography.^[14,15] Some colorimetric and fluorescent probes have also been developed recently.^[16-21] These chemosensors enable the detection of (p)ppGpp in vitro or from cell extracts, however, not in individual living cells. Fluorescent protein-based sensors have also been developed attempting to study cellular (p)ppGpp levels.^[22,23] However, these sensors don’t directly target (p)ppGpp, but rather its translational regulation ability or enzymatic activity. Sensors that can directly measure (p)ppGpp levels in live cells are still highly demanded.

To fill this gap, we report here the development of genetically encodable RNA-based sensors for fluorescence imaging of (p)ppGpp in living cells. These RNA sensors have three major components: a (p)ppGpp-specific aptamer, a transducer sequence, and a fluorogenic RNA reporter, Broccoli.^[24] Aptamers are single-stranded oligonucleotides that can bind their target with high affinity and selectivity. Naturally existing aptamers (riboswitches) that can selectively bind with (p)ppGpp were recently discovered.^[25] In our sensor construct, when (p)ppGpp binds to the aptamer region, the transducer sequence hybridizes and facilitates the folding of the Broccoli RNA. The folded Broccoli can further bind to its cognate dye, DFHBI-1T, and activate its fluorescence (Figure 1a). As a result, the fluorescence signal can be used to detect (p)ppGpp.

Figure 1.

Design and in vitro characterization of RNA-based (p)ppGpp sensor. a) Modular design of the sensors comprises of a Broccoli (green), a (p)ppGpp-binding aptamer (blue), and a transducer sequence (gray). b) Optimization of transducer sequences. The corresponding sequences were shown in Table S1. Spectra were measured in a solution containing 1 μM RNA, 20 μM DFHBI-1T, and either 0 or 10 μM ppGpp, at 90 min after mixing. c) Selectivity of the S2 sensor. Spectra were measured with 10 mM NTPs or 10 μM other ligands in a solution containing 1 μM S2 and 20 μM DFHBI-1T. The effect of dye concentration on the S2 fluorescence is shown in Figure S3. Control was measured without adding ligands. d) Dose-response curve of the S2 sensor. The limit of detection was calculated to be 0.3 μM. Shown are the mean ± standard deviation of three independent replicates.

We first designed three RNA sensor constructs by fusing Broccoli into the P0 stem region of the (p)ppGpp riboswitch through different transducer sequences. These transducers have been optimized based on an Mfold online software. However, minimal fluorescence activation was shown upon adding (p)ppGpp (Figure S1). After further examining the crystal structure of the riboswitch (PDB: 6DMC), we realized that the P2 stem region is also very close to the (p)ppGpp-binding pocket and may be sequence-independent.^[26-28] To test if (p)ppGpp sensors could be developed by inserting Broccoli into the P2 stem, we designed another five transducer sequences based on the Mfold simulation (Table S1). Indeed, three of these constructs, S1–S3, exhibited respectively, 5.9-, 7.9-, and 2.3-fold fluorescence enhancement after adding 10 μM ppGpp (Figure 1b and S2).

We chose S1 and S2 sensors for further in vitro tests due to their high fluorescence enhancement. We first tested their selectivity against several (p)ppGpp analogs (Figure 1c and S4). Both S1 and S2 fluorescence can be activated by either guanosine tetraphosphate (ppGpp) or guanosine pentaphosphate (pppGpp), and exhibited no fluorescence activation in the presence of NTPs, guanine, or phosphoribosyl pyrophosphate (PRPP). S1 and S2 are highly selective towards (p)ppGpp.

We then studied the detection range of S1 and S2. A dose-response curve was measured for each sensor after adding various concentrations of ppGpp (Figure 1d and S5). Our results indicated that S2 can effectively detect ppGpp ranging from 0.5 μM to 10 μM (EC₅₀, 2.7 μM), whereas S1, with a more stable transducer (2 bp vs. 1 bp) exhibits a broader detection range, 1–100 μM (EC₅₀, 36 μM). These two sensors can generally cover the concentration range that (p)ppGpp is known to normally exhibit in bacterial cells.^[14,29] To potentially detect even higher concentrations of (p)ppGpp, we also designed sensors based on a weak (p)ppGpp-binding D. hafniense ilvE riboswitch^[25] (Table S1). A slightly higher detection range (EC₅₀, 72 μM) was indeed observed with the new optimal sensor (named D3) (Figure S5).

We have also determined the kinetics of RNA sensor activation and deactivation. In general, S2 exhibited a faster fluorescence activation than S1. After mixing ppGpp with S2, half-maximum fluorescence level was reached in ~7 min, with 80% of maximal signal shown in ~19 min (Figure S6). By contrast, it would take ~65 or ~39 min for the S1 or D3 signal to reach half-maximum (Figure S6). The fluorescence deactivation rate of S2 is also fast, with >95% signal decrease occurs in less than 4 min after removing the free ppGpp (Figure S7). These results indicated that the fluorescence signal of the RNA sensors, especially the S2, can be used to monitor the dynamic variations of (p)ppGpp levels. Considering its relatively high fluorescence intensity and fast kinetic response, we continued with S2 for the subsequent studies.

To visualize (p)ppGpp in live cells, we first cloned S2 into a pET28c vector and transformed it into BL21 Star™ (DE3) E. coli cells (Figure S8). As controls, we transformed plasmids expressing Broccoli alone or empty vector. We also induced a high level of (p)ppGpp in the S2 reporter strain by expressing a truncated form of the E. coli RelA protein, RelA*.^[4,30] Following incubation with DFHBI-1T, as expected, Broccoli-expressing cells exhibited high fluorescence, while the empty vector control had low signal (Figure 2a). S2-expressing cells also showed minimal fluorescence, which is expected based on the low (p)ppGpp levels under rich growth conditions. Importantly, upon induction of (p)ppGpp by RelA*, strong fluorescence signal was observed in most cells (Figure 2a). As a control, Broccoli cellular fluorescence is not influenced by RelA*, and expressing an empty RelA* vector itself will not influence the S2 signal (Figure S9).

Figure 2.

Live-cell imaging of (p)ppGpp with the S2 sensor. a) Confocal fluorescence imaging in live BL21 Star™ (DE3) cells. Images were taken 1 h after adding 200 μM DFHBI-1T. The effect of dye concentration on the cellular fluorescence is shown in Figure S10. Both the cell morphology and S2 fluorescence were shown. Scale bar, 10 μm. b) Distributions of the cellular fluorescence. In each condition, a total of >100 cells were measured from three experimental replicates. The width of violin represents the percentage of cells with the same fluorescence intensity. The white box region indicates the interquartile range, i.e., middle 50% of cellular fluorescence. The dot and short horizontal line in the box indicate the mean and median value of each data set.

We have quantified the fluorescence intensities from hundreds of individual cells in each condition. Our results indicated that in strains with elevated (p)ppGpp, S2 fluorescence was approximately 70% of that of Broccoli (Figure 2b). A ~7-fold fluorescence enhancement was observed if compared with that in the absence of RelA* expression. These results are highly consistent with in vitro data (Figure 1b).

To further confirm that these RNA-based sensors can indeed detect intracellular (p)ppGpp, we conducted another series of experiments under different nutritional conditions. We first cultured S2-expressing E. coli cells in M9 minimal medium, where high level of (p)ppGpp exists due to the low nutrient content.^[31] As predicted, strong cellular S2 fluorescence was observed (Figure 3a). After supplementing the M9 medium with nutrients casamino acids (CAA) and glucose, a dramatic and rapid fluorescence decrease was observed (Figure 3a). As a control, fluorescence signals of Broccoli-expressing cells did not change during the same nutrient supplementation process (Figure S11). We also tested reversibly by first incubating cells in M9 medium containing both CAA and glucose, and then deprived the nutrients by switching to the M9 minimum medium. Indeed, a significant fluorescence enhancement was observed (Figure 3a). An HPLC assay was also used to validate the changes in the cellular ppGpp levels under these nutrient conditions (Figure S12). Meanwhile, the expression of S2 will not affect the cellular amounts of ppGpp (Figure S12). All these data indicate that S2 can be used to detect changes in the cellular (p)ppGpp levels.

Figure 3.

Imaging of (p)ppGpp biosynthesis under nutritional stress. a) Confocal fluorescence imaging of S2-expressing BL21 Star™ (DE3) cells in either M9 minimal medium or M9 supplemented with 0.2% casamino acids (CAA) and 0.4% glucose (Glu). Shown are images before or 15 min after changing the medium. Scale bar, 10 μm. b) Serine hydroxamate (SHX)-induced (p)ppGpp biosynthesis. These E. coli cells were first incubated in M9 medium supplemented with 0.2% CAA, and then 1% final concentration of SHX was added 15 min before imaging. c) Methyl-α-glucose (α-MG)-induced (p)ppGpp biosynthesis. These bacterial cells were first incubated in M9 medium supplemented with 0.4% glucose, and then 2.5% final concentration of α-MG was added 15 min before imaging.

We have further studied the effect of some chemical inducers in the biosynthesis of (p)ppGpp. Serine hydroxamate (SHX) and methyl-α-glucose (α-MG) are two commonly used chemicals to trigger the intracellular accumulation of (p)ppGpp, which functions by inducing the starvation of amino acids and glucose, respectively.^[32-34] Indeed, after adding SHX or α-MG, significant increase in cellular fluorescence was observed (Figure 3, S13, and S14). Compared to the effect of SHX, the addition of α-MG induced less fluorescence enhancement. This result is consistent with some previous findings that glucose starvation has less influence on the biosynthesis of (p)ppGpp than amino acids starvation does.^[33,35]

We also measured the kinetics of cellular (p)ppGpp accumulation in response to the medium exchange. Cells were first grown in nutrient-rich medium and then switched into a nutrient-limited M9 medium. By imaging every 10 min over 1 h, a time-dependent increase in the cellular fluorescence was observed. In most cells, a clear fluorescence activation exhibited within the first 20 min and reached plateau (~7-fold increase) at ~60 min (Figure 4a and S15). As a control, minimal photobleaching was noticed when Broccoli fluorescence was imaged under the same condition (Figure S16). By comparing to the in vitro S2 kinetics (Figure S6), our results indicated that the cells started to gradually produce (p)ppGpp after the medium swap.

Figure 4.

Kinetics of (p)ppGpp biosynthesis in E. coli cells. a) Representative confocal fluorescence images of S2-expressing BL21 Star™ (DE3) cells after switching at 0 min from M9/0.2% casamino acids/0.4% glucose medium to M9 minimal medium. b) Distribution of S2 cellular fluorescence levels as measured from 102 individual cells. Shown are the changes in the fluorescence signal of each individual cell (gray) and the averaged signal (red). Scale bar, 2 μm.

To study if there is cell-to-cell variation in (p)ppGpp biosynthesis, we monitored the fluorescence signal change of 102 individual cells during this medium switching process (Figure 4b). On average, half-maximum fluorescence could be reached in ~20 min. By defining fast kinetics as cellular fluorescence doubled within 20 min, the majority (84%) of cells exhibited fast (p)ppGpp accumulation rates, while 16% of cells showed a slow rate of (p)ppGpp biosynthesis. These data indicate that (p)ppGpp is rapidly accumulated in E. coli cells upon nutritional stress.

We next wanted to measure the cellular distributions of (p)ppGpp. To normalize the variations of plasmid expression level among different cells, we first cloned the S2 sensor into a pETDuet vector that contains a separate promoter expressing eqFP670 far-red fluorescent protein as a reference (Figure S8). After transforming these plasmids into BL21 Star™ (DE3) cells, we studied the effect of nutritional stress on the (p)ppGpp accumulation. We incubated these bacterial cells in either M9/CAA/glucose medium or M9 minimal medium for 90 min. As expected, significantly higher S2 fluorescence exhibited in the M9 medium, while similar level of eqFP670 fluorescence was observed in both conditions (Figure S17).

To better normalize the cell-to-cell variation in the RNA sensor expression level, we cloned S2 in a pET28c vector together with a DNB RNA aptamer tag (Table S1). Here, the DNB/TMR-DN signal can be used as a reference to calibrate the cellular expression levels of Broccoli-based sensors.^[36] To test this new ratiometric RNA sensor, we transformed the S2-DNB conjugate into one batch of BL21 cells and S2-DNB with RelA* gene into another. Indeed, clear (p)ppGpp-induced S2 signal was observed in the RelA* strain, and the DNB fluorescence was quite similar in both strains (Figure 5a).

Figure 5.

Cellular distributions of (p)ppGpp. a) Confocal fluorescence imaging of BL21 Star™ (DE3) cells that express S2 (green) and a DNB reference (red). These cells were incubated in either M9/0.2% casamino acids/0.4% glucose medium or M9 minimal medium for 90 min before imaging in the presence of 200 μM DFHBI-1T and 1 μM TMR-DN. Pseudo-color ratio images indicated the distributions of green-to-red fluorescence intensity ratios, in the range of 0 to 1. Scale bar, 10 μm. b) Statistical distributions of green-to-red fluorescence intensity ratios as measured from >1,000 cells. Individual cells were binned according to the fluorescence intensity ratio and the percentage of cells in each bin was plotted.

We further determined the S2 fluorescence at a single-cell level and normalized it to the corresponding reference signal (Figure 5b and S17). In the nutrient rich M9/CAA/glucose medium, over 90% of cells exhibited minimum (p)ppGpp levels (Green/Red ratio (G/R)< 0.10), and ~3% of cells exhibited moderate (p)ppGpp biosynthesis (G/R> 0.10). By contrast, when nutrients were removed or with the RelA* expression, cellualr (p)ppGpp levels were largely up-regulated. Significantly more cells (~43%) showed moderate (p)ppGpp levels (G/R, 0.10–0.60), and a small portion of cells (~4.5%) even exhibited quite high (p)ppGpp concentrations (G/R> 0.60). These data suggested that bacterial cells can indeed respond to nutritional stress by rapidly accumulating (p)ppGpp. The highly accumulated (p)ppGpp in a small portion of cells may be potentially an interesting target for the study of biofilm formation and persistence.

In summary, we reported here the development of RNA-based fluorescent sensors for the selective and sensitive detection of (p)ppGpp. For the first time, these genetically encoded sensors allowed (p)ppGpp to be directly imaged in living cells, which provided a groundbreaking approach to study the cell-to-cell variations and dynamics of (p)ppGpp biosynthesis and accumulation. Considering the importance of (p)ppGpp in the stringent response of bacteria, we envision that these sensors can be broadly used to study the detailed mechanism of the physiological regulation of bacterial survival during stress, virulence, antibiotic resistance and persistence.