Optogenetic control of a GEF of RhoA uncovers a signaling switch from retraction to protrusion

Reviewer #1 (Public Review):

De Seze et al. investigated the role of guanine exchange factors (GEFs) in controlling cell protrusion and retraction. In order to causally link protein activities to the switch between the opposing cell phenotypes, they employed optogenetic versions of GEFs which can be recruited to the plasma membrane upon light exposure and activate their downstream effectors. Particularly the RhoGEF PRG could elicit both protruding and retracting phenotypes. Interestingly, the phenotype depended on the basal expression level of the optoPRG. By assessing the activity of RhoA and Cdc42, the downstream effectors of PRG, the mechanism of this switch was elucidated: at low PRG levels, RhoA is predominantly activated and leads to cell retraction, whereas at high PRG levels, both RhoA and Cdc42 are activated but PRG also sequesters the active RhoA, therefore Cdc42 dominates and triggers cell protrusion. Finally, they create a minimal model that captures the key dynamics of this protein interaction network and the switch in cell behavior.

The conclusions of this study are strongly supported by data. Perhaps the manuscript could include some further discussion to for example address the low number of cells (3 out of 90) that can be switched between protrusion and retraction by varying the frequency of the light pulses to activate opto-PRG. Also, the authors could further describe their "Cell finder" software solution that allows the identification of positive cells at low cell density, as this approach will be of interest for a wide range of applications.

Reviewer #2 (Public Review):

Summary:

This manuscript builds from the interesting observation that local recruitment of the DHPH domain of the RhoGEF PRG can induce local retraction, protrusion, or neither. The authors convincingly show that these differential responses are tied to the level of expression of the PRG transgene. This response depends on the Rho-binding activity of the recruited PH domain and is associated with and requires (co?)-activation of Cdc42. This begs the question of why this switch in response occurs. They use a computational model to predict that the timing of protein recruitment can dictate the output of the response in cells expressing intermediate levels and found that, "While the majority of cells showed mixed phenotypes irrespectively of the activation pattern, in few cells (3 out of 90) we were able to alternate the phenotype between retraction and protrusion several times at different places of the cell by changing the frequency while keeping the same total integrated intensity (Figure 6F and Supp Movie)."

Strengths:

The experiments are well-performed and nicely documented. However, the molecular mechanism underlying the shift in response is not clear (or at least clearly described). In addition, it is not clear that a prediction that is observed in ~3% of cells should be interpreted as confirming a model, though the fit to the data in 6B is impressive.

Overall, the main general biological significance of this work is that RhoGEF can have "off target effects". This finding is significant in that an orthologous GEF is widely used in optogenetic experiments in drosophila. It's possible that these findings may likewise involve phenotypes that reflect the (co-)activation of other Rho family GTPases.

Weaknesses:

The manuscript makes a number of untested assumptions and the underlying mechanism for this phenotypic shift is not clearly defined.

This manuscript is missing a direct phenotypic comparison of control cells to complement that of cells expressing RhoGEF2-DHPH at "low levels" (the cells that would respond to optogenetic stimulation by retracting); and cells expressing RhoGEF2-DHPH at "high levels" (the cells that would respond to optogenetic stimulation by protruding). In other words, the authors should examine cell area, the distribution of actin and myosin, etc in all three groups of cells (akin to the time zero data from figures 3 and 5, with a negative control). For example, does the basal expression meaningfully affect the PRG low-expressing cells before activation e.g. ectopic stress fibers? This need not be an optogenetic experiment, the authors could express RhoGEF2DHPH without SspB (as in Fig 4G).

Relatedly, the authors seem to assume ("recruitment of the same DH-PH domain of PRG at the membrane, in the same cell line, which means in the same biochemical environment." supplement) that the only difference between the high and low expressors are the level of expression. Given the chronic overexpression and the fact that the capacity for this phenotypic shift is not recruitment-dependent, this is not necessarily a safe assumption. The expression of this GEF could well induce e.g. gene expression changes.

The third paragraph of the introduction, which begins with the sentence, "Yet, a large body of works on the regulation of GTPases has revealed a much more complex picture with numerous crosstalks and feedbacks allowing the fine spatiotemporal patterning of GTPase activities" is potentially confusing to readers. This paragraph suggests that an individual GTPase may have different functions whereas the evidence in this manuscript demonstrates, instead, that *a particular GEF* can have multiple activities because it can differentially activate two different GTPases depending on expression levels. It does not show that a particular GTPase has two distinct activities. The notion that a particular GEF can impact multiple GTPases is not particularly novel, though it is novel (to my knowledge) that the different activities depend on expression levels.

These descriptions are not precise. What is the nature of the competition between RhoA and Cdc42? Is this competition for activation by the GEFs? Is it a competition between the phenotypic output resulting from the effectors of the GEFs? Is it competition from the optogenetic probe and Rho effectors and the Rho biosensors? In all likelihood, all of these effects are involved, but the authors should more precisely explain the underlying nature of this phenotypic switch. Some of these points are clarified in the supplement, but should also be explicit in the main text.

Author Response

Reviewer #1 (Public Review):

De Seze et al. investigated the role of guanine exchange factors (GEFs) in controlling cell protrusion and retraction. In order to causally link protein activities to the switch between the opposing cell phenotypes, they employed optogenetic versions of GEFs which can be recruited to the plasma membrane upon light exposure and activate their downstream effectors. Particularly the RhoGEF PRG could elicit both protruding and retracting phenotypes. Interestingly, the phenotype depended on the basal expression level of the optoPRG. By assessing the activity of RhoA and Cdc42, the downstream effectors of PRG, the mechanism of this switch was elucidated: at low PRG levels, RhoA is predominantly activated and leads to cell retraction, whereas at high PRG levels, both RhoA and Cdc42 are activated but PRG also sequesters the active RhoA, therefore Cdc42 dominates and triggers cell protrusion. Finally, they create a minimal model that captures the key dynamics of this protein interaction network and the switch in cell behavior.

We thank reviewer #1 for this assessment of our work.

The conclusions of this study are strongly supported by data. Perhaps the manuscript could include some further discussion to for example address the low number of cells (3 out of 90) that can be switched between protrusion and retraction by varying the frequency of the light pulses to activate opto-PRG.

The low number of cells being able to switch can be explained by two different reasons:

1. first, we were looking for clear inversions of the phenotype, where we could see clear ruffles in the case of the protrusion, and clear retractions in the other case. Thus, we discarded cells that would show in-between phenotypes, because we had no quantitative parameter to compare how protrusive or retractile they were. This reduced the number of switching cells

2. second, we had a limitation due to the dynamic of the optogenetic dimer used here. Indeed, the control of the frequency was limited by the dynamic of unbinding of the optogenetic dimer. This dynamic of recruitment (~20s) is comparable to the dynamics of the deactivation of RhoA and Cdc42. Thus, the differences in frequency are smoothed and we could not vary enough the frequency to increase the number of switches. Thanks to the model, we can predict that decreasing the unbinding rate of the optogenetic tool should allow us to increase the number of switching cells.

We will add further discussion of this aspect to the manuscript.

Also, the authors could further describe their "Cell finder" software solution that allows the identification of positive cells at low cell density, as this approach will be of interest for a wide range of applications.

There is a detailed explanation of the ‘Cell finder’ in the method sections. It is also available on github at https://github.com/jdeseze/cellfinder and currently in development to be more user-friendly and properly commented.

Reviewer #2 (Public Review):

Summary:

This manuscript builds from the interesting observation that local recruitment of the DHPH domain of the RhoGEF PRG can induce local retraction, protrusion, or neither. The authors convincingly show that these differential responses are tied to the level of expression of the PRG transgene. This response depends on the Rho-binding activity of the recruited PH domain and is associated with and requires (co?)-activation of Cdc42. This begs the question of why this switch in response occurs. They use a computational model to predict that the timing of protein recruitment can dictate the output of the response in cells expressing intermediate levels and found that, "While the majority of cells showed mixed phenotypes irrespectively of the activation pattern, in few cells (3 out of 90) we were able to alternate the phenotype between retraction and protrusion several times at different places of the cell by changing the frequency while keeping the same total integrated intensity (Figure 6F and Supp Movie)."

Strengths:

The experiments are well-performed and nicely documented. However, the molecular mechanism underlying the shift in response is not clear (or at least clearly described). In addition, it is not clear that a prediction that is observed in ~3% of cells should be interpreted as confirming a model, though the fit to the data in 6B is impressive.

Overall, the main general biological significance of this work is that RhoGEF can have "off target effects". This finding is significant in that an orthologous GEF is widely used in optogenetic experiments in drosophila. It's possible that these findings may likewise involve phenotypes that reflect the (co-)activation of other Rho family GTPases.

We thank reviewer #2 for having assessed our work. Indeed, the main finding of this work is the change in the GEF function upon its change in concentration, which could be explained with a simple model supported by quantitative data. We think that the mechanism of the switch is quite clear, supported by the data showing the double effect of the PH domain and the activation of Cdc42. The few cells that are able to switch phenotype have to be seen as an honest data confirming that 1) concentration is indeed the main determinant of the protein’s function, and the switch is hard to obtain (which is also predicted by the model) 2) the two underlying networks are being activated at different timescales, which leaves some space for differential activation in the same cell. We are here limited by the dynamic of the optogenetic tool, as explained in the response to reviewer #1, and the intrinsic cell-to-cell variability.

Regarding the interpretation of our results as RhoGEF “off target effects”, we think that it might be too reductive. As said in the discussion, we proposed that the dual role of the RhoGEF could have physiological implications on the induction of front protrusions and rear retractions. While we do not demonstrate it here, it opens the door for further investigation.

Weaknesses:

The manuscript makes a number of untested assumptions and the underlying mechanism for this phenotypic shift is not clearly defined.

We may not have been clear in our manuscript, but we think that the underlying mechanism for this phenotypic shift is clearly explained and backed up by the data and the literature. It relies on 1) the ability of PRG to activate both RhoA and Cdc42 and 2) the ability of the PH domain to directly bind to active RhoA (which is, as shown in the manuscript, necessary but not sufficient for protrusions to happen). The model succeeds in reproducing the data of RhoA with only one free parameter and two independently fitted ones. The fact that activation of RhoA and Cdc42 lead to retraction and protrusion respectively is known since a long time. Thus, we think that the switch is clearly and quantitatively explained.

This manuscript is missing a direct phenotypic comparison of control cells to complement that of cells expressing RhoGEF2-DHPH at "low levels" (the cells that would respond to optogenetic stimulation by retracting); and cells expressing RhoGEF2-DHPH at "high levels" (the cells that would respond to optogenetic stimulation by protruding). In other words, the authors should examine cell area, the distribution of actin and myosin, etc in all three groups of cells (akin to the time zero data from figures 3 and 5, with a negative control). For example, does the basal expression meaningfully affect the PRG low-expressing cells before activation e.g. ectopic stress fibers? This need not be an optogenetic experiment, the authors could express RhoGEF2DHPH without SspB (as in Fig 4G).

We thank reviewer #2 for this suggestion. PRG-DHPH is known to affect the phenotype of the cell as shown in Valon et al., 2017. Thus, we really focused on the change implied by the change in optoPRG expression, to understand the phenotype difference. However, we agree that this could be an interesting data to add and will do the experiments for the revised version of the manuscript.

Relatedly, the authors seem to assume ("recruitment of the same DH-PH domain of PRG at the membrane, in the same cell line, which means in the same biochemical environment." supplement) that the only difference between the high and low expressors are the level of expression. Given the chronic overexpression and the fact that the capacity for this phenotypic shift is not recruitment-dependent, this is not necessarily a safe assumption. The expression of this GEF could well induce e.g. gene expression changes.

We agree with reviewer #2 that there could be changes in gene expression. In the next point of this supplementary note, we had specified it, by saying « that overexpression has an influence on cell state, defined as protein basal activity or concentration before activation. » We are sorry if it was not clear and will change this sentence for the new version.

One of the interests of the model is that it does not require any change in absolute concentrations, beside the GEF. The model is thought to be minimal and fits well and explains the data with very few parameters. We don’t show that there is no change in concentration but we show that it is not required to invoke it.

We will add in the revised version of the manuscript a paragraph discussing this question.

The third paragraph of the introduction, which begins with the sentence, "Yet, a large body of works on the regulation of GTPases has revealed a much more complex picture with numerous crosstalks and feedbacks allowing the fine spatiotemporal patterning of GTPase activities" is potentially confusing to readers. This paragraph suggests that an individual GTPase may have different functions whereas the evidence in this manuscript demonstrates, instead, that a particular GEF can have multiple activities because it can differentially activate two different GTPases depending on expression levels. It does not show that a particular GTPase has two distinct activities. The notion that a particular GEF can impact multiple GTPases is not particularly novel, though it is novel (to my knowledge) that the different activities depend on expression levels.

We thank the reviewer for this remark and didn’t intended to confuse the readers. Indeed, we think that this manuscript confirms the canonical view on the GTPases (as most optogenetic experiments did in the past years). We show here that it is more complicated at the level of the GEF. We agree that this is not particularly novel. However, to our knowledge, there is no example of such clear phenotypic control, explained solely by the change in concentration.

We think that the last paragraph of the introduction is quite clear in the fact that it is the GEF itself that switches its function, and not the Rho-GTPases, but we will reconsider the phrasing of this paragraph for the revised version.

Concerning the overall model summarizing the authors' observations, they "hypothesized that the activity of RhoA was in competition with the activity of Cdc42"; "At low concentration of the GEF, both RhoA and Cdc42 are activated by optogenetic recruitment of optoPRG, but RhoA takes over. At high GEF concentration, recruitment of optoPRG lead to both activation of Cdc42 and inhibition of already present activated RhoA, which pushes the balance towards Cdc42."

These descriptions are not precise. What is the nature of the competition between RhoA and Cdc42? Is this competition for activation by the GEFs? Is it a competition between the phenotypic output resulting from the effectors of the GEFs? Is it competition from the optogenetic probe and Rho effectors and the Rho biosensors? In all likelihood, all of these effects are involved, but the authors should more precisely explain the underlying nature of this phenotypic switch. Some of these points are clarified in the supplement, but should also be explicit in the main text.

We are going to precise these descriptions for the revised version of the manuscript. The competition between RhoA and Cdc42 was thought as a competition between retraction due to the protein network triggered by RhoA (through ROCK-Myosin and mDia-bundled actin) and the protrusion triggered by Cdc42 (through PAK-Rac-ARP2/3-branched Actin). We will make it explicit in the main text.