A genetically encoded BRET-based SARS-CoV-2 Mpro protease activity sensor

The main protease, Mpro, is critical for SARS-CoV-2 replication and an appealing target for designing anti-SARS-CoV-2 agents. Therefore, there is a demand for the development of improved sensors to monitor its activity. Here, we report a pair of genetically encoded, bioluminescence resonance energy transfer (BRET)-based sensors for detecting Mpro proteolytic activity in live cells as well as in vitro. The sensors were generated by sandwiching peptides containing the Mpro N-terminal autocleavage sites, either AVLQSGFR (short) or KTSAVLQSGFRKME (long), in between the mNeonGreen and NanoLuc proteins. Co-expression of the sensors with Mpro in live cells resulted in their cleavage while mutation of the critical C145 residue (C145A) in Mpro completely abrogated their cleavage. Additionally, the sensors recapitulated the inhibition of Mpro by the well-characterized pharmacological agent GC376. Further, in vitro assays with the BRET-based Mpro sensors revealed a molecular crowding-mediated increase in the rate of Mpro activity and a decrease in the inhibitory potential of GC376. The sensors developed here will find direct utility in studies related to drug discovery targeting the SARS-CoV-2 Mpro and functional genomics application to determine the effect of sequence variation in Mpro.

Anupriya et al report on the construction of a BRET-based sensor for measuring SARS-CoV-2 protease inhibition. This is a therapeutic target for Covid-19 treatments, and the creation of rapid/easy detection assays for detecting inhibition have value for the current pandemic and will likely be valuable for other present or future viruses. The paper is well presented and the authors robustly characterize their sensor. Figure 3G: The inset blocks a portion of the y-axis and its labelling. Please move it to make the axis clear.  Lines 354-428: Perhaps place this in the supplementary materials. It seems to be validating the authors construct choice, but does not add much value considering the authors rigorously tested both experimentally.
Lines 489 to 531: This experiment is unclear to me. The methods suggest an empty "filler plasmid" was also transformed. This sounds like there will be cells transformed with the filler plasmid rather than Mpro, which isn't really a dose-response of Mpro on the reporter. How do the authors know this is a dose response of Mpro rather than just having more cells transformed with the filler plasmid instead of Mpro?
Lines 33-36, 124-126,594-611, 740-741: I think the present authors did not perform enough robust experiments to make these claims of increased specificity and sensitivity. The present authors performed one experiment and showed the FlipGFP-based sensor exhibits more background than their one. Background is not that important here because the signal to noise ratio and detection limit are more important for determining specificity or sensitivity. In fact, the opposite appears to be true. The authors of the FlipGFP paper show an IC50 of 5.5 µM with GC376 (See figure 3D), whereas the present paper found a much higher IC50 of 127 and 194 µM when testing their sensor with GC376.  Graph showing a representative bioluminescence spectra of the short (C) and long (D) M pro sensor constructs either in control cells or in cells expressing the WT or C145A mutant M pro . Data were fit to a two Gaussian model reflecting mNG fluorescence and NLuc bioluminescence peaks. Note the reduction in the mNG peak (533 nm) of both the short and the long sensors when coexpressed with the wild type M pro while no reduction was observed when coexpressed with the C145A mutant M pro . (E,F) Graphs showing total mNG fluorescence (measured prior to substrate addition) in cells expressing the short (E) and the long (F) sensors. (G) Graph showing BRET ratio (ratio emission at 533 nm and 467 nm) of the short (left side) and the long (right side) M pro protease activity sensors in either control cells or when coexpressed with the wild type or the C145A mutant M pro . Note that data for both the short and the long sensor have been plotted on the same y-axis for the ease of comparison. Indicated p-values were obtained from unpaired, two-sided Student's ttest. Inset: graph showing percentage change in BRET of the short (left side) and the long (right side) when co-expressed with the wild type or the C145A mutant M pro . Data shown are mean ± S.D. from three independent experiments, each performed in triplicates. (H) Top panel: anti-His tag blot showing cleavage of the short (left side) and the long (right side) M pro sensor constructs in either control cells or in cells co-expressing the wild type or the C145A mutant M pro . Note the release of an approximately 30 kDa, His6-tagged-mNG fragment in cells expressing the wild type but not in the C145A mutant M pro . Bottom panel: anti-Strep-tag blot showing expression of the M pro in the respectively transfected cells. Comment 2: Figure 3: The authors state: "Data shown are mean ± S.D. from a representative of independent experiments performed multiple times". However, all data should be shown. The methods suggest two independent replicates each containing 3 technical replicates, and the presence of only 3 data points looks like only one set of technical replicates were graphed.
Response: We would like to thank the reviewer for raising this point. We have now included data from 3 independent experiments (mean values obtained from each experiment performed in triplicates) and used them to determine mean BRET and % change in BRET (in the inset).
The figure legend has been modified in the following way: "Data shown are mean ± S.D. from three independent experiments, each performed in triplicates." The figure has been modified in the following way:     Response: We would like to thank the Reviewer for the kind words on experimental validation of the sensor constructs and the suggestion. However, we would prefer to keep the MD simulation data in the main text.
Comment 5: Lines 489 to 531: This experiment is unclear to me. The methods suggest an empty "filler plasmid" was also transformed. This sounds like there will be cells transformed with the filler plasmid rather than Mpro, which isn't really a dose-response of Mpro on the reporter. How do the authors know this is a dose response of Mpro rather than just having more cells transformed with the filler plasmid instead of Mpro?
Response: We would like to thank the Reviewer for raising this point. We have now made changes in the text to indicate that we used a control, pcDNA3.1-based plasmid DNA (not expressing Mpro) as a transfection control to maintain the total amount of DNA used for transfection. We agree with the Reviewer in that it is not possible to control the amount of expression of the protease in individual cells in this way. However, we would like to indicate that the experiments were performed with a population of cells and the measurements were also performed at the population level, and the dose-response of Mpro is at the level of the cell population. We would like to note that the transfection of cells with the control plasmid does not affect the function of the sensor which is evident from the "No protease" control. Our data also show an inverse relation between the concentration of M pro WT and sensor BRET ratio indicating the increase in protease activity, while the BRET ratio remains unchanged in the presence of mutant M pro C145A (Fig. 4A,B).
The materials and methods section has been edited in the following way: "For dose-response experiments, a control plasmid (a pcDNA3.1-based plasmid not expressing M pro ) is also co-transfected to maintain the amount of plasmid DNA in transfection constant. In case of timecourse experiments with either the short or the long sensor, control cells (No M pro ) were transfected with the control plasmid while the wild type (WT) and the C145A mutant M pro cells were transfected with the respective M pro expressing plasmid DNA. The time-course experiments were carried out at 1:5 sensor-toprotease plasmid DNA ratio." Comment 6: Lines 33-36, 124-126,594-611, 740-741: I think the present authors did not perform enough robust experiments to make these claims of increased specificity and sensitivity. The present authors performed one experiment and showed the FlipGFP-based sensor exhibits more background than their one. Background is not that important here because the signal to noise ratio and detection limit are more important for determining specificity or sensitivity. In fact, the opposite appears to be true. The authors of the FlipGFP paper show an IC50 of 5.5 µM with GC376 (See figure 3D), whereas the present paper found a much higher IC50 of 127 and 194 µM when testing their sensor with GC376.
Response: We would like to thank the Reviewer for raising this issue. We have now included data from two independent experiments conducted with the FlipGFP-based Mpro sensor in the main text.
The text has been modified in the following way: "Epifluorescence imaging of the cells post 4 h of transfection revealed the appearance of GFP fluorescence in the transfected cells, as ascertained from mCherry fluorescence, in the presence of WT M pro after 24 h of transfection (67 ± 7%; mean ± S.D., n=2) while more cells showed GFP fluorescence after 48 h of transfection (84 ± 2%; mean ± S.D., n=2) (Supp. Fig. S7). These data indicate a delayed response of the FlipGFP sensor to M pro proteolytic activity in comparison to the BRET-based sensor. Additionally, a significant number of cells were found to be GFP positive after 48 h (11 ± 6%; mean ± S.D., n=2) of FlipGFP transfection in the presence of the C145A mutant M pro (Supp. Fig. S7). This is contrast to the observations made with the BRET-based sensor in the presence of the mutant M pro (Fig.  4C,D)." The figure containing images from one of the experiments has been moved to the supplementary information.
I think the present authors did not perform enough robust experiments to make these claims of increased specificity and sensitivity. The present authors performed one experiment and showed the FlipGFP-based sensor exhibits more background than their one. Background is not that important here because the signal to noise ratio and detection limit are more important for determining specificity or sensitivity. In fact, the opposite appears to be true. The authors of the FlipGFP paper show an IC50 of 5.5 µM with GC376 (See figure 3D), whereas the present paper found a much higher IC50 of 127 and 194 µM when testing their sensor with GC376. We agree with the Reviewer in that a high background is not as important as the signal to noise ratio. However, as shown in panel E, the FlipGFP-based sensor showed some GFP+ cells (we have included the detailed ImageJ/Fiji scrip used for the determination of GFP+ cells in the supplementary) in the presence of the C145A mutant Mpro whereas the BRET-based sensor didn't show any decrease in signal under the same condition. On the IC50 value, we are not certain as to why we are observing a higher value compared to the FlipGFP (data included in panel F). One possibility that we have mentioned in the main text is that the peptide presented to BRET sensor constructs resemble the native polyprotein better than either the bare minimum FRET-based peptides or the differently structured peptide in the FlipGFP sensor. We would like to note that we could observe a significantly difference in the IC50 value in the presence of molecular crowding agent in vitro (Fig. 7G, H). Comment 7: Figure 1: The mNG and NLuc on top of the proteins is very hard to read, and the authors might want to give a quick explanation of these in the figure legend.
Response: We would like to thank the Reviewer for this suggestion.
We have now edited the figure (and the legend) with increased font size for mNG and NLuc:

Reviewer #2
This manuscript reported the design of Bioluminescence Resonance Energy Transfer (BRET)-based assay for the SARS-CoV-2 main protease (Mpro). Both the shorter and longer substrates were examined, which gave similar results. The ratio of reporter plasmid to the Mpro, as well as the induction time, were optimized. Flip-GFP assay was performed in parallel. It was found that the BRET assay was more sensitive and specific than the Flip-GFP assay. The optimized BRET assay was used to test Mpro inhibitor GC-376 and found to have much weaker potency than the values reported in the literature. The BRET assay was also performed in cell culture, and addition of 25% PEG 20K increased the proteolytic activity of Mpro and decreased the potency of GC-376. It was therefore claimed that Mpro is more active in the crowded environment of an infected host cell compared to in vitro conditions, thus requiring higher drug concentration for complete inhibition. Highlights of this study including the detailed assay optimization and vigorous assay calibration using the C145A dead mutants, the direct comparison with the Flip-GFP assay, and the molecular crowding experiment. To further strength the conclusions, the authors might consider the following suggestions: Comment 1: "This is especially relevant given that the binding of the peptide substrate has been reported to allosterically activate the SARS-CoV-1 Mpro dimer." Comment: reference should be given.
Response: We would like to thank the Reviewer for this suggestion. We have now included the references related to the involvement of substrate in the dimerization and activation of Mpro.
The text has been modified in the following way: "This is especially relevant given that the binding of the peptide substrate has been reported to allosterically activate the SARS-CoV-1 M pro dimer. 92,93 " Comment 2: Is there any internal control for the BRET assay to normalize the transfection efficiency? In the Flip-GFP assay, mCherry is the internal control.
Response: We would like to thank the Reviewer for the question and would like to mention that we have used mNeonGreen fluorescence (measured upon excitation with an external source of light, instead of NLuc bioluminescence; indicated as total mNG fluorescence; Fig. 3E,F) as a way to compare the expression of the sensors, in addition to the western blot analysis. More importantly, BRET is measured as a ratio of acceptor and donor emission and therefore, is internally controlled. Comment 3: "Additionally, a significant number of cells were found to be GFP positive after 24 h (9 ± 1%) and 48 h (20 ± 1%) of FlipGFP transfection in the presence of the C145A mutant Mpro (Fig. 5C,D). This is contrast to the observations made with the BRET-based sensor in the presence of the mutant Mpro (Fig.  4C,D)." Comment: the background GFP signal from the Flip-GFP assay might be a result of cleavage by the host proteases. If this the case, the BRET assay should have similar background signal as both assays contain the same substrate. Is there any explanation why the BRET assay has less leakage signal?
Response: We would like to thank the Reviewer for raising this question and much like the Reviewer, we too are perplexed as to why there is a difference between the two sensors. One possible that we could think of is that the cleavage peptide in the FlipGFP-based sensor is in a greater degree of structural constraint due to its presence in a structured protein (GFP) than that in the BRET-based sensors reported here. One may suggest that due to this, the cleavage peptide could be cleaved non-specifically by host cell proteases. Another possible reason we can think of is that while the cleavage sequence is the same in both types of sensors, additional residues spanning cleavage sequence in the sensor may result in the formation of a cleavage site for a host protease.
The text has been modified (including the new set of data) in the following way: "Additionally, a significant number of cells were found to be GFP positive after 48 h (11 ± 6%; mean ± S.D., n=2) of FlipGFP transfection in the presence of the C145A mutant M pro (Supp. Fig. S7)." Comment 4: "The lower efficacy of GC376 observed here compared to previous reports perhaps indicates a cell type-or Mpro 646 expression dependent effect." Comment: HEK 293T cells were used in both the BRET and Flip-GFP assay, so the above statement does not hold. For direct comparison, the author should also determine the EC50 of GC-376 in the Flip-GFP assay. This is to rule out the possibility of incorrect drug concentration or the different cell type used in this study from the ones reported in the literature. Another possibility might be the drug efflux pump Pgp and GC-376 is a known substrate of P-gp (ACS Infect. Dis. 2021, 7, 3, 586-597). However, this is unlikely as 293T is not known to have high levels of P-gp.
Response: We would like to thank the Reviewer for raising this concern. We have now performed live cell Mpro inhibition assay using the FlipGFP sensor and find that the GC376 is IC50 similar to that reported previously 2 . We have now edited Fig. 5 to include this. We have edited the main text in the following way: "These data showed that the IC50 values for the short sensor is 127.4 ± 23.33 μM and that for long sensor is 194.7 ± 7.49 μM while the IC50 value obtained using the FlipGFP sensor was 5.453 ± 1.03 μM (Fig. S8). The lower efficacy of GC376 observed here with the BRET-based sensor compared to the FlipGFPbased sensor 2 (and previous reports 2-6,27,58 ) perhaps indicates that the peptide substrate presented sandwiched between the mNG and NLuc proteins serves as a better substrate than either in the FRETbased peptides or in the FlipGFP sensor construct." Comment 5: "Together, these data indicate that Mpro could be more active in the crowded environment of an infected host cell compared to in vitro conditions, and may require higher concentrations of pharmacological inhibitors for effective inhibitions of its catalytic activity than those determined from in vitro assays." Comment: To provide additional evidence for this conclusion, the authors should repeat the FRET assay with and without 25% PEG 20K. In addition, the lack of direct correlation between the results of in vitro assay and the cell-based assay might due to many factors including cell membrane permeability, drug efflux, protein binding, off-target effect, metabolism and etc.
Response: We would like to thank the Reviewer for this excellent suggestion and would like to note that such activation of Mpro has been previously shown using the FRET assay 7 . We agree with the Reviewer that molecular crowding in cells may not entirely explain the differences we observed in in vitro and live cell assays. However, the difference in the IC50 values obtained using the BRET-based sensors and the FlipGFP-based sensor (Fig. S8) likely indicates that the BRET-based sensors serve as better substrates for Mpro and thus, one may need to use higher concentrations of the Mpro inhibitors, especially under the crowded environment of living cells. We have edited the main text in the following way: "While a number of factors including cell membrane permeability, drug efflux, interaction with proteins, offtarget effects and metabolism of the drug can impact efficacy of a drug, results from live cell assays with the BRET-based sensors and the FlipGFP-based sensor combined with those from in vitro assays suggest that the molecular crowding-mediated M pro activation and the better presentation of the substrate peptide in the BRET-based sensor constructs could be the probable reasons for the increased IC50 value of GC376 for SARS-CoV-2 M pro determined using the BRET-based sensors. This then posits that higher concentrations of pharmacological inhibitors may be required for effective inhibitions of M pro catalytic activity in living cells than those determined from in vitro assays." The authors addressed most of the comments, however I still have concerns around two points.
Response to Reviewer 1, comment 6: The authors seem confused on the meaning of specificity and sensitivity. Below are the definitions of these terms.
-Sensitivity (true positive rate) refers to the probability of a positive test, conditioned on truly being positive.
-Specificity (true negative rate) refers to the probability of a negative test, conditioned on truly being negative.
The authors show no experiments looking at the chance of true positives or false negatives. The reduced background the authors discuss again does not affect the ability to distinguish true positives and negatives (and is therefore unrelated to specificity and sensitivity). Any assay has background and that is why researchers use negative controls.
If anything, the authors seem to show the FlipGFP assay to work better by being able to detect GC376 at a 24X lower concentration (5.453 ± 1.03 μM vs. 127.4 ± 23.33 μM). In the absence of proper experiments, my guess is that this would translate to better sensitivity and specificity for the FlipGFP assay.
Response to Reviewer 1, comment 5: Thank you for clarifying the experiment. The authors did not answer how this is a dose response of Mpro rather than just having more cells transformed with the filler plasmid instead of Mpro?
The setup of the experiment means that when lower Mpro plasmid concentrations are used there will be more cells containing only the BRET sensor and not Mpro. It follows that these cells cannot undergo the BRET reaction and would be expected to create the result seen. I find the description of this as a dose-dependency of Mpro to be incorrect when all they do is create less cells with Mpro. The authors should make it clear that what they are doing is altering the proportion of cells that contain both Mpro and BRET, and not refer to this as a potency of Mpro.

Reviewer #2 (Remarks to the Author):
Comments from the previous round of review were properly addressed. I therefore recommend acceptance.