Ezrin, radixin and moesin compose the family of ERM proteins. They link actin filaments and microtubules to the plasma membrane to control signaling and cell morphogenesis. Importantly, their activity promotes invasive properties of metastatic cells from different cancer origins. Therefore, a precise understanding of how these proteins are regulated is important for the understanding of the mechanism controlling cell shape, as well as providing new opportunities for the development of innovative cancer therapies. Here, we developed and characterized novel bioluminescence resonance energy transfer (BRET)-based conformational biosensors, compatible with high-throughput screening, that monitor individual ezrin, radixin or moesin activation in living cells. We showed that these biosensors faithfully monitor ERM activation and can be used to quantify the impact of small molecules, mutation of regulatory amino acids or depletion of upstream regulators on their activity. The use of these biosensors allowed us to characterize the activation process of ERMs that involves a pool of closed-inactive ERMs stably associated with the plasma membrane. Upon stimulation, we discovered that this pool serves as a cortical reserve that is rapidly activated before the recruitment of cytoplasmic ERMs.
Ezrin, radixin and moesin (ERM) are cross-linkers between the plasma membrane and the cytoskeleton. This family of proteins regulate cell morphogenesis by linking actin filaments (F-actin) and microtubules to the plasma membrane (Fehon et al., 2010; Polesello and Payre, 2004; Solinet et al., 2013). Thereby, ERMs regulate several important functions such as T cell activation, cell migration, neurogenesis, epithelial maintenance and cell division (Fehon et al., 2010). In a pathological context, overexpression and activation of ERM proteins participate in cancer progression (Clucas and Valderrama, 2014). While the role of ERM proteins in this process is not totally understood, their experimental inactivation has been shown to reduce metastasis of cells from different cancer origins (Bulut et al., 2012; Ghaffari et al., 2019). Therefore, blocking ERM activation represents a promising new therapeutic avenue against cancer metastasis.
ERMs harbor three conserved domains (Fehon et al., 2010). First, a FERM (band 4.1 ezrin radixin moesin) domain at the N-terminus that localizes ERMs at the plasma membrane by binding to phosphatidylinositol-(4,5)-bisphosphate [PtdIns(4,5)P2] (Niggli et al., 1995; Roch et al., 2010; Roubinet et al., 2011) and to membrane proteins (Tsukita et al., 1994; Wu et al., 2004). This domain also binds to microtubules (Solinet et al., 2013). Second, there is a central α-helix domain acting as a hinge region that allows switching between closed and open conformations (Hoeflich et al., 2003). Third, a C-ERMAD (C-terminal ezrin radixin moesin association domain) that interacts with F-actin (Turunen et al., 1994) and harbors a conserved regulatory threonine that is phosphorylated to sustain ERM opening (T567, T564 and T558 in ezrin, radixin and moesin, respectively). Disruption of the interaction between the FERM domain and C-ERMAD opens the molecule and unmasks the actin-binding and microtubule-binding sites on the C-ERMAD and FERM domains, respectively.
ERM activation involves multiple steps but is still not totally understood. PtdIns(4,5)P2 was shown to recruit closed-inactive ERMs at the cortex to simultaneously promote their opening, by loosening the interaction between the FERM and C-ERMAD (Fievet et al., 2004; Hamada et al., 2000). This open-active conformation is not stable, and ERMs re-localize into the cytosol in their closed-inactive conformation (Coscoy et al., 2002). A kinase stabilizes the open conformation at the plasma membrane by phosphorylating the conserved threonine in the C-ERMAD.
Recently, PtdIns(4,5)P2 and LOK (also known as STK10), a Ser/Thr kinase, were shown to act together to promote full opening and activation of ezrin. First, PtdIns(4,5)P2 primes ezrin opening at the plasma membrane. This allows LOK to wedge in between the FERM domain and the C-ERMAD to physically distance these domains to promote full opening of ezrin (Pelaseyed et al., 2017). Finally, LOK stabilizes this open-active conformation by phosphorylating the T567 regulatory threonine. These findings imply that ERMs exist under a closed conformation at the plasma membrane. Interestingly, expression of a non-phosphorylatable ezrin (T567A) mutant in epithelial cells has shown that this ERM can form inactive oligomers at the plasma membrane (Gautreau et al., 2000).
Presently, available tools do not allow to precisely dissect the sequence of events linking the translocation to the plasma membrane and the opening and activation of ERMs. Up to now, only the last step of ERM activation can be assessed, using an anti-phospho antibody that targets the phosphorylated regulatory threonine residue. Since the surrounding region of this threonine is 100% identical in all ERM proteins, this antibody is not specific to the activation state of individual ERMs. Here, we aimed to develop new tools to monitor ERM activation. We elected to develop ERM biosensors based on the principle of bioluminescence resonance energy transfer (BRET) (Angers et al., 2000; Xu et al., 1999) and enhanced-bystander BRET (ebBRET) (Namkung et al., 2016), since such sensors are easy to use in living cells and are compatible with high-throughput screening (Avet et al., 2020; Benredjem et al., 2019; Namkung et al., 2018; Schönegge et al., 2017). BRET is similar to fluorescence resonance energy transfer (FRET) except that the energy donor is a luminescent and not a fluorescent molecule. FRET requires the external fluorescence excitation to initiate energy transfer, which may cause background noise due to autofluorescence and photobleaching. BRET, on the other hand, results from the emission of photons by a luminescent enzyme upon oxidation of cell permeable substrates. Therefore, BRET avoids some of the drawbacks associated with fluorescence excitation seen in FRET.
We developed a collection of ebBRET-based biosensors that monitor the conformation and activation of individual ERM proteins (i.e. ezrin, radixin and moesin) in living cells and validated their use for high-throughput screening, demonstrating that they could ultimately allow the identification of novel small compounds blocking the metastatic activity of ERMs.
Finally using these biosensors, we discovered an intermediate step in the mechanism of activation of ERM proteins. We confirmed the existence of a pool of closed-inactive ERMs that stably associates with the plasma membrane. Moreover, upon stimulation, we discovered that this intermediate pool allows a fast and local activation of ERMs that precedes the cortical recruitment and activation of cytoplasmic ERMs.
Development of ebBRET-based biosensors to monitor the regulation of individual ERMs
As a proxy for ERM activation, we aimed to design BRET-based biosensors that can quantify ezrin, radixin or moesin recruitment at the plasma membrane. To do this, we constructed the first component of the ebBRET biosensors by fusing the bioluminescent donor Renilla luciferase (rLucII) at the C-terminus of individual ERMs (hereafter referred to collectively as individual E,R,M-rLucII). The other component consists of the BRET fluorescent acceptor Renilla GFP (rGFP) targeted to the plasma membrane through addition of the prenylation CAAX box of KRAS (rGFP-CAAX; Namkung et al., 2016) (Fig. 1A). Thus, when ERMs are recruited at the plasma membrane, we expected a rise in BRET signals resulting from an increased proximity between the individual ezrin-, radixin-, moesin-rLucII donors (E,R,M-rLucII) and rGFP-CAAX acceptor.
To validate that the fusion of rLucII at E,R,M C-terminus did not affect their regulation, their phosphorylation state in response to Ser/Thr kinase and phosphatase inhibitors were tested in HEK293T cells. As shown in Fig. 1, both calyculin A [which inhibits Ser/Thr phosphatases and increases ERM phosphorylation (Kondo et al., 1997)] and staurosporine [which inhibits Ser/Thr kinases and prevents ERM phosphorylation (Tachibana et al., 2015)] modulated the phosphorylation of the regulatory threonine of individual E,R,M-rLucII to a similar extent to endogenous ERMs (Fig. 1B–D). This suggests that rLucII fusion does not dramatically affect the overall regulation of the individual ERM biosensors.
The ability of the biosensors to detect the association of ERM with the plasma membrane was then characterized by measuring BRET in HEK293T cells co-expressing individual E,R,M-rLucII and rGFP-CAAX. Upon addition of the rLucII substrate, we found that the BRET signals generated by these three biosensors were saturable in titration experiments with increasing amounts of the rGFP-CAAX BRET acceptor (Fig. 1E–G). This progressive increase of BRET signals following a hyperbole illustrates the specificity of the signals detected and indicates that a detectable fraction of the E,R,M-rLucII donors are interacting with the plasma membrane under basal condition. This confirms that ERMs are constitutively associated with the plasma membrane, which is consistent with previous reports and with their critical role in cortical regulation (Berryman et al., 1995; Fievet et al., 2004; Gautreau et al., 2000).
Chemical activation of ERMs confirmed that a pool of closed-inactive ERMs associates with the plasma membrane
To determine whether the BRET signals detected under basal condition reflect the presence of closed-inactive or open-active forms at the plasma membrane, we assessed the effect of the phosphatase inhibitor on the signal. Unexpectedly, addition of calyculin A promoted a net drop of each individual ERM biosensor BRET signals compared to the vehicle control (Fig. 2A–C). This decrease suggests the existence of a significant fraction of ERMs that is already localized at the plasma membrane in a closed-inactive conformation. In their closed conformation, ERMs are ∼10 nm in length whereas they can span up to ∼40 nm when open (Liu et al., 2007). Given that BRET efficiency varies to the 6th power of the distance and the BRET R0 is ∼5 nm (Dacres et al., 2012), if a pool of closed ERMs stably associates with the plasma membrane, ERM opening and activation could increase by four times the distance between the rLucII donor and rGFP-CAAX acceptor, leading to an important decrease of BRET signals (Fig. 2F) as observed upon calyculin A activation.
Concentration response curves of phosphatase inhibitor revealed that the respective IC50 and EC50 of calyculin A assessed by BRET and by western blotting for the phosphorylated regulatory threonine of ERMs were very similar [IC50 (BRET): E=30.2 nM, R=8.9 nM, M=8.0 nM; EC50 (WB)=63.8 nM)] (Fig. 2D,E). Such correlation with the phosphorylation event indicates that the sensors readily detect an activation process, thereby demonstrating the sensitivity of the ERM BRET biosensors. In addition, we found that individual ERM biosensors presented similar IC50 for activation by calyculin A, as calculated by BRET (Fig. 2D), consistent with the notion that the three ERMs have similar sensitivity to phosphatases.
To further validate that the decrease of BRET signals reveals the existence of a pool of closed-inactive ERMs at the plasma membrane, we next introduced a phospho-mimetic mutation (T to D) on the regulatory threonine of individual E,R,M-rLucII. This mutation promotes ERM opening and open-active conformation in cells (Fievet et al., 2004). Confirming our hypothesis, we found that individual ERM opening via the T to D mutation promoted a net decrease in BRET signals when compared to their wild-type counterparts (Fig. 2G–I). To visualize the pool of closed-inactive ERMs at the plasma membrane, we next monitored BRET signals using BRET-based imaging (Kobayashi et al., 2019). As reported for endogenous ezrin (Fehon et al., 2010), ezrin-rLucII localized both in the cytoplasm and at the plasma membrane (Fig. 2J, total luminescence). As expected from the spectrometric BRET signals detected, BRET signals can readily be detected at the plasma membrane further demonstrating that rLucII fused to the C-terminus of ezrin and rGFP-CAAX are in close proximity at the plasma membrane (Fig. 2J, BRET2). Also consistent with the spectrometric data presented above, upon calyculin A activation, ezrin biosensor BRET signals decreased specifically at the plasma membrane while ezrin-rLucII signals remained constant (Fig. 2J–L). This indicates that ezrin did not leave the plasma membrane but rather transitioned into an open conformation upon phosphatase inhibition. Taken together, these findings demonstrate that ERMs stably associate with the plasma membrane in their closed conformation.
We next assessed whether these ERM biosensors could also monitor ERM inhibition. According to our finding that a pool of inactive ERMs associates with the plasma membrane, an increase of BRET signals should follow ERM inhibition since the closing of ERMs at the plasma membrane will bring the luciferase donor closer to the rGFP acceptor. Although staurosporine promotes ERM dephosphorylation and inactivation (Fig. 1B–D), we did not detect any significant changes in the BRET signals after staurosporine treatment with the radixin-rLucII and moesin-rLucII biosensors, and only a slight increase with the ezrin-rLucII biosensor (Fig. 2A–C). BRET-based imaging confirmed these observations, and we did not measure any significant changes in ezrin-rLucII biosensor BRET signals at the plasma membrane upon staurosporine treatment (Fig. 2J,K) nor any change in its distribution (Fig. 2L). In addition, the introduction of a non-phosphorylatable mutation of the regulatory threonine (T to A) in the individual E,R,M-rLucII constructs inactivating ERMs (Fievet et al., 2004) did not significantly affect the BRET signals of each bimolecular sensor in comparison with their wild-type counterpart (Fig. 2G–I). This lack of signal variation following inhibition could result from an equilibrium between an increase of signals from the plasma membrane associated inactive conformation and a decrease of signals due to the detachment of ERMs from the plasma membrane to the cytoplasm.
Engineering ebBRET-based biosensors to monitor ERM closing at the plasma membrane
The equilibrium between inactive ERMs at the plasma membrane and in the cytoplasm makes it difficult to measure the closing of ERMs at the plasma membrane using the first ebBRET-based biosensors. To monitor ERM closing, we designed a second set of ebBRET-based ERM biosensors that are restricted to the plasma membrane. To this aim, we stably associated the individual E,R,M-rLucII fusions to the plasma membrane using a myristoylation (Myr) motif followed by a polybasic (PB) amino acid sequence (hereafter referred to collectively as individual MyrPB-E,R,M-rLucII) (Fig. 3A). We first used sequential fractionation of cytosol and membranes to verify that the MyrPB motif targets ERMs at the plasma membrane (Fig. 3B,C). We found that, as previously reported for endogenous ezrin (Gautreau et al., 2000), ezrin-rLucII localizes both in the cytosol and in the membrane fractions (Fig. 3B,C). In a marked contrast, MyrPB-ezrin-rLucII was almost exclusively found in the membrane fraction. The association of MyrPB-ezrin-rLucII with the plasma membrane was further confirmed using BRET-based imaging. MyrPB-ezrin-rLucII was significantly enriched at the plasma membrane when compared with the ezrin-rLucII biosensor (Fig. 3D,E).
With the MyrPB-E,R,M-rLucII biosensors, we expect to measure an increase in BRET signals upon ERM inactivation and a decrease upon activation. In control cells, we found that the BRET signals of these three biosensors were saturable in titration experiments showing the specificity of the BRET signals (Fig. 3F–H). As predicted, ERM inactivation by staurosporine significantly increased BRET signals of individual MyrPB-E,R,M-rLucII sensors (Fig. 3F–H). To verify that this effect was not due to an unexpected modification of the composition of the plasma membrane caused by staurosporine treatment, we constructed a control biosensor composed of rLucII directly fused to the MyrPB plasma membrane targeting motif (MyrPB-rLucII). Treatment with staurosporine had no effect on the BRET signals between this plasma membrane-targeted rLucII and rGFP-CAAX (Fig. 3I). As observed with the first ebBRET-based biosensors (Fig. 2A–C), ERM activation with calyculin A decreased BRET signals of the three MyrPB-E,R,M-rLucII biosensors (Fig. 3F–H) while it did not affect the BRET signals of the control MyrPB-rLucII biosensor (Fig. 3I). We then assessed the respective EC50 and IC50 of staurosporine and calyculin A on individual MyrPB-E,R,M-rLucII and endogenous ERMs measured by BRET and western blotting, respectively. We found that the endogenous ERMs and MyrPB-E,R,M-rLucII biosensors presented a similar EC50 and IC50 for calyculin A and staurosporine within the nanomolar range (Fig. 3J–L).
Finally, since ERMs can form homo- and hetero-oligomers at the plasma membrane (Berryman et al., 1995; Gautreau et al., 2000), we verified that oligomerization with endogenous ERMs does not interfere with BRET signals of this second set of ebBRET-based biosensors. To test this, we depleted ezrin using two different shRNAs. As a result of silent DNA mutations introduced in the MyrPB-ezrin-rLucII, one of these shRNA (EZR#9) reduced endogenous ezrin protein levels without affecting the protein levels or the Luciferase activity of exogenous MyrPB-ezrin-rLucII (Fig. S1A,B). The other shRNA (EZR#8) reduced the expression levels of both proteins (Fig. S1A,B). We found that specific depletion of endogenous ezrin did not affect the BRET signals of any of the MyrPB-E,R,M-rLucII biosensors at steady state or upon treatments with kinase or phosphatase inhibitors (Fig. S1C–E).
Taken together, these results suggest that this second set of ebBRET-based biosensors monitor both activation and inhibition following treatment with phosphatase and kinase inhibitors, respectively.
The MyrPB-ezrin-rLucII biosensor probes ezrin conformational changes at the plasma membrane
Ezrin being the most characterized ERM during cancer metastasis (Clucas and Valderrama, 2014), we decided to focus on this paralog to further our understanding of its regulation. We tested whether the MyrPB-ezrin-rLucII BRET signals can be modulated by introducing the T567D phospho-mimetic or T567A non-phosphorylatable mutations to change its conformation. Confirming that the MyrPB-ezrin-rLucII biosensor probes ezrin conformational states, we observed that BRET signals decreased for the MyrPB-ezrinT567D-rLucII mutant and increased for the MyrPB-ezrinT567A-rLucII biosensor mutant (Fig. 4A). As expected, we also found that the BRET signals from MyrPB-ezrin-rLucII phospho-mutants were not decreased or increased upon calyculin A or staurosporine treatments, respectively (Fig. S2). Furthermore, the live BRET signals of MyrPB-ezrin-rLucII at the plasma membrane significantly decreased upon calyculin A activation and increased upon staurosporine inactivation, without any significant change of the signal luminescence intensity of MyrPB-ezrin-rLucII (Fig. 4B–D).
Having shown that the MyrPB-ezrin-rLucII biosensor faithfully monitors ezrin conformation changes upon chemical activation or inhibition, we then tested whether the BRET signals of this biosensor can be modulated by depleting upstream regulators of ERMs. We previously reported that the regulatory threonine of dMoesin, the sole ERM protein in Drosophila, is phosphorylated by the Ser/Thr kinase Slik and is dephosphorylated by the Ser/Thr phosphatase PP1-87B (Carreno et al., 2008; De Jamblinne et al., 2020; Roubinet et al., 2011). In mammals, SLK, the human ortholog of Slik and the close paralogue of LOK, phosphorylates ERM proteins (Machicoane et al., 2014) whereas PPP1CA, an ortholog of PP1-87B, dephosphorylates them (Canals et al., 2012). We thus knocked down SLK and PPP1CA by shRNA, and we confirmed that these orthologs also regulate ERM phosphorylation in HEK293T cells by western blotting (Fig. 4E). Confirming the robustness of the individual MyrPB-E,R,M-rLucII biosensors, we observed that the BRET signals associated with ezrin, radixin and moesin increased upon SLK depletion and decreased upon PPP1CA depletion (Fig. 4F; Fig. S3A,B). Taken together, these results show that membrane-restricted ERM BRET biosensors precisely monitor both activation and inhibition of individual ERM by sensing their open and closed conformation at the plasma membrane.
Finally, we assessed whether the MyrPB-ezrin-rLucII ebBRET-based biosensor is suitable for high-throughput screening. To this aim, we measured the Z-factor (Zhang et al., 1999; denoted Z′) of activation and inhibition in HEK293T cells seeded in 384-well plate using automated procedures. Vehicle (DMSO), calyculin A or staurosporine were randomly distributed on cells stably expressing the MyrPB-ezrin-rLucII and rGFP-CAAX bimolecular ebBRET-based biosensor. At 1 h after treatment, BRET signals were read and we found that calyculin A and staurosporine induced a highly reproducible decrease or increase of MyrPB-ezrin-rLucII BRET signal, respectively, when compared with vehicle alone (Fig. 4G). The Z factor of ezrin activation, measured by a BRET signal decrease (Z′=0.87, calyculin A) and ezrin inhibition, measured by a BRET signal increase (Z′=0.59, staurosporine) are both compatible with high-throughput screening procedures, showing that this biosensor will be valuable to identify compounds that modulate ezrin activation after screening of compound libraries.
Phosphorylation of T567 controls ezrin opening
Phosphorylation of the C-terminal conserved threonine residue is a key mechanism for ERM activation. However, other amino acid residues have been reported to directly contribute to ERM opening and activation. We next aimed to test whether the BRET biosensors were able to assess the contribution of these amino acids for the regulation of ezrin. Cyclin-dependent kinase 5 (Cdk5) has been shown to activate ezrin by phosphorylating the T235 of its FERM domain (Yang and Hinds, 2003). In the closed ezrin conformation, T235 directly opposes T567 at the interface between the FERM and C-ERMAD domains (Pearson et al., 2000). Phosphorylation on T235 was hence proposed to promote ezrin opening by bringing negative charges to the FERM domain that abrogate its intramolecular binding to the C-ERMAD. In accordance with this hypothesis, we found that phospho-mimetic mutation of T235 (ezrinT235D) opens ezrin as indicated by the decrease of BRET signals associated with MyrPB-ezrinT235D-rLucII (Fig. 5A). However, this decrease was much less pronounced than the one observed with MyrPB-ezrinT567D-rLucII (Fig. 4A). In addition, we found that calyculin A or staurosporine treatments were able to modulate the opening of MyrPB-ezrinT235D-rLucII, as seen by the decrease or increase in its associated BRET signals (Fig. 5B). This is in a marked contrast with what we observed with MyrPB-ezrinT567D-rLucII, where calyculin A or staurosporine treatments did not promote ezrin opening or closing, respectively (Fig. S2). This suggests that regulation of ezrin by calyculin A or staurosporine is achieved mainly by modulating the phosphorylation state of T567. Thus, although phosphorylation of T235 can promote ezrin opening, regulation of T567 phosphorylation appears to be the main event that regulates ezrin.
Binding to PtdIns(4,5)P2 is a prerequisite for ezrin opening and activation
Binding to PtdIns(4,5)P2 has been shown to recruit ERM at the plasma membrane (Barret et al., 2000). Mutation of two lysine doublets into asparagine (denoted KK253,254NN and KK262,263NN) on ezrin abrogates its binding to PtdIns(4,5)P2 and its subsequent phosphorylation on T567 (Barret et al., 2000; Fievet et al., 2004). By mutating the PtdIns(4,5)P2-binding site within the first iteration of the ezrin biosensor (ezrinK253,254,262,263N-rLucII), we confirmed that PtdIns(4,5)P2 promotes recruitment of ezrin to the plasma membrane. Indeed, the BRET signals associated with ezrinK253,254,262,263N-rLucII were reduced when compared to ezrin-rLucII (Fig. 5C). We also found that ezrinK253,254,262,263N-rLucII was only slightly sensitive to calyculin A-induced activation when compared to ezrin-rLucII (Fig. 5D). This suggests that ezrin activation mainly occurs at the plasma membrane.
In addition to recruiting ezrin at the plasma membrane, PtdIns(4,5)P2 has also been proposed to prime ezrin opening, a necessary step to allow the LOK kinase to wedge between the FERM and C-ERMAD to fully open and phosphorylate this ERM (Pelaseyed et al., 2017). Confirming the hypothesis that PtdIns(4,5)P2 binding is a key mechanism for ezrin opening, we found that targeting ezrin to the plasma membrane while abrogating binding to PtdIns(4,5)P2 promotes ezrin closing. Mutation of the PtdIns(4,5)P2-binding site within MyrPB-ezrin-rLucII (MyrPB-ezrinK253,254,262,263N-rLucII) significantly increased BRET signals when compared to MyrPB-ezrin-rLucII (Fig. 5C). Interestingly, while calyculin A was still able to promote MyrPB-ezrinK253,254,262,263N-rLucII opening as observed by a decrease in BRET signals, staurosporine did not promote further closing of MyrPB-ezrinK253,254,262,263N at the difference of MyrPB-ezrin (Fig. 5E).
Closed-inactive cortical ezrin serve as a reserve of proteins that can be quickly and locally activated
The existence of a pool of closed ERMs that stably associates with the plasma membrane (see Fig. 2) raises the possibility that these ERMs serve as a reserve of proteins that can be rapidly and locally activated. We explored this hypothesis by comparing the timing of ezrin conformational opening by BRET with the timing of ezrin recruitment at the plasma membrane using live-cell imaging in HEK293T cells expressing ezrin–GFP. Upon phosphatase inhibition, cortical closed-inactive ezrin opened very rapidly as the BRET signals started to decrease at 1 min (the earliest time point tested) and reached a plateau at ∼4 min for both ezrin-rLucII and MyrPB-ezrin-rLucII biosensors (Fig. 6A). In the meantime, ezrin–GFP was significantly recruited at the plasma membrane only after 8 min of activation (Fig. 6B,C). In parallel, we observed that, upon activation, phosphorylation of the regulatory threonine of ERMs also started very rapidly (∼2 min, the earliest time point tested) but reached a plateau only after 10 min; 6 min after the cortical pool of closed-inactive ERMs was fully activated (10 min versus 4 min) (Fig. 6A,D). Although these differences of timing could be due to the use of different molecular tools, we hypothesized that the cortical pool of plasma membrane associated-inactive ezrin is activated before the de novo recruitment of cytoplasmic inactive ezrin. We tested this hypothesis by comparing the profile of phosphorylation of ezrin-rLucII with those of the plasma membrane associated MyrPB-ezrin-rLucII after activation (Fig. 6E,F). We found that whereas the maximal phosphorylation of MyrPB-ezrin-rLucII was achieved rapidly after 2 min of activation by calyculin A, ezrin-rLucII phosphorylation increased more gradually to reach a plateau at 10 min. In accordance with these findings, we performed membrane/cytosol fractionation experiments, which revealed that, following calyculin A treatment, ERM phosphorylation precedes their presence in the membrane fraction. We found that the presence of ERM phosphorylation in the membrane fraction strongly increases after 4 min of treatment (the earliest time point tested) whereas ezrin enrichment was detected only after 8 min (Fig. 6G,H; Fig. S4). This confirms that ERMs are constitutively associated with the plasma membrane under basal condition, which is consistent with previous reports and with their critical role in cortical regulation (Berryman et al., 1995; Fievet et al., 2004; Gautreau et al., 2000). Taken together, these data show that the cortical pool of closed-inactive ezrin is activated first. Then, upon sustained activation, cytoplasmic ezrin is recruited at the plasma membrane to be concomitantly phosphorylated and activated (see model, Fig. 7).
Here, we developed and characterized a set of ebBRET-based biosensors that measure the activation of individual ERMs. These biosensors locally monitor the conformational changes of ERMs at the plasma membrane that dictate their activity. We demonstrated that these biosensors faithfully measure ERM regulation in live cells following activation and inhibition. In addition, these biosensors revealed that a pool of closed-inactive ERMs at the plasma membrane can be rapidly and locally open and activated upon stimulation. This finding adds a supplementary layer of refinement in the model of regulation of ERM proteins.
While characterizing the first set of ERM biosensors, we confirmed the existence of a pool of inactive ERM proteins that stably associates with the plasma membrane and that have a closed conformation. This fraction of closed ezrin proteins could correspond to the head-to-tail inactive oligomers that were previously identified at the plasma membrane of epithelial cells (Gautreau et al., 2000). In addition, our data confirm a new model of ERM activation (Pelaseyed et al., 2017). In this model, PtdIns(4,5)P2 primes ezrin opening at the plasma membrane, allowing the Ser/Thr kinase LOK to wedge in between the FERM domain and the C-ERMAD to fully open the protein. Here, by introducing mutations that abrogate ezrin binding to PtdIns(4,5)P2, we confirmed that this phosphoinositide is necessary to properly control the opening and activation of ezrin. While the available molecular tools did not allow us to test it, the new model of ERM activation implies that ERM proteins are already associated with the plasma membrane in their closed conformation before activation. In accordance with this model, we demonstrated that a pool of inactive-closed ERMs is stably associated with the cortex.
Generating one specific cell shape that promotes one specific biological function relies on the integration of actin filaments at the plasma membrane by proteins such as ERMs. Ezrin, radixin and moesin regulate several functions that require a rapid remodeling of cell shape such as cell division or cell migration. In agreement with the rapid timeline associated with these functions, we discovered that the cortical pool of closed ezrin can be activated within 1 min (see Fig. 6A), even before the other pool of cytosolic ezrin is recruited and activated at the plasma membrane (see Fig. 6B,C; model in Fig. 7). These results are consistent with previous observations showing that ezrin cycles between active and inactive states within 2 min (Viswanatha et al., 2012). Interestingly, we observed that, upon calyculin A treatment, BRET signals associated with ezrin-rLucII are stable over time (see Fig. 6A), even after 9 min, when cytosolic ezrin is recruited (see Fig. 6B,C). Since BRET signals occur when the rLucII and rGFP-CAAX are in close proximity at the plasma membrane, this suggests that the recruitment of cytosolic closed-inactive ERMs is immediately followed by their opening and activation. This discovery contributes to a better comprehension of how ERMs are open and activated at the plasma membrane. However, we still need more information to understand how the trafficking of cytosolic ERMs toward the plasma membrane can be triggered after stimulation.
Until now, the main approach used to study ERM activation was to follow the phosphorylation of the C-ERMAD regulatory threonine residue using a specific phospho-antibody through western blotting or immunofluorescence. However, this approach based on this p-ERM antibody presents several limitations. First, this antibody does not discriminate between the phosphorylated form of the three ERMs. Therefore, it is not possible to visualize the level of activation of each individual ERM in cells by immunofluorescence. In addition, since ezrin and radixin have a similar apparent molecular mass, they cannot be separated by regular SDS-PAGE. Here, we developed biosensors that monitor the activity of each ERM independently. Second, ERM phosphorylation at their regulatory C-ERMAD threonine has been shown to stabilize their active open conformation, and therefore is commonly associated with their activation state. However, ERMs were also found to be activated without being phosphorylated at this conserved residue. For instance, phosphorylation of T235 of the FERM domain of ezrin has been proposed to be sufficient to open and activate this ERM (Yang and Hinds, 2003). We confirmed this role by showing that a phospho-mimetic mutation of T235 on ezrin promotes its opening. We brought experimental evidence that the ebBRET-based conformational biosensors directly monitor the closed-inactive and open-active forms of ERMs at the plasma membrane and thus bypass the requirement that the C-ERMAD threonine is phosphorylated. Third, the use of antibodies is not compatible with real-time analysis in live cells. Thus, assessing sequential activation of ERMs by western blotting or immunofluorescence requires complex experimental approaches with different samples for every time point. This greatly increases the variability between conditions and experiments. In the present study, we showed that the ERM bimolecular biosensors allow monitoring of ERM activation within the same cell population over time. We took advantage of this to compare the activation kinetics of the different pools of ezrin. In addition, using BRET-based imaging, we showed that the level of ERM activation can be visualized in real-time in living cells.
Finally, ERMs were recently identified as promising therapeutic targets against metastasis (Clucas and Valderrama, 2014; Ghaffari et al., 2019; Ren and Khanna, 2014). The development of new ERM therapeutic inhibitors requires high-throughput screening procedures that are not compatible with the existing experimental approaches to study ERMs. BRET-based biosensors have already been widely used in the past years in genetic and chemical screens. Here, we developed ERM bimolecular biosensors that present high and reproducible BRET signals in 384-well plates, and hence are compatible with high-throughput screening. We are currently using these new BRET-based biosensors to identify new compounds targeting ERM activity and therefore their potential ability to inhibit the metastatic progression associated with ERMs.
MATERIALS AND METHODS
Reagents and inhibitors
Coelenterazine 400a (Deep Blue C) and methoxy e-CTZ (Prolume Purple) were purchased from NanoLight Technology (#340 and #369, respectively). Calyculin A was purchased from Sigma (#C5552). Staurosporine was purchased from ApexBio (#A8192).
The rGFP-CAAX construct was previously described (Namkung et al., 2016). Ezrin-rLucII, radixin-rLucII and moesin-rLucII constructs were obtained by subcloning PCR amplified ezrin, radixin or moesin into pcDNA 3.1 hygro(+) GFP10-rLucII vector (Picard et al., 2018) digested with NheI and AgeI to replace GFP10. MyrPB-E,R,M-rLucII constructs were obtained by adding a myristoylation sequence followed by a polybasic motif (MyrPB, MGCTLSAEDKAAVERSKMAVQSPKKGLLQRLFKRQHQTIPRVAVQNAAIRSGGSGGSGGSGGSNAAIRS) to the E,R,M-rLucII constructs. The negative control MyrPB-rLucII was obtained by subcloning the MyrPB sequence into the pCDNA 3.1 hygro(+) GFP10-rLucII vector. All following point mutations were obtained by inverse PCR from the wild-type constructs. The constitutively active and inactive (MyrPB-) ezrin-rLucII mutants (MyrPB-)ezrinT567D-rLucII and (MyrPB-)ezrinT567A-rLucII were constructed using the following primers: (forward) 5′-AAGATCTGCGGCAGATCCGGCAGGGCAACACCAAGC-3′ and (reverse) 5′-TGTACTTGTCCCGGCCTTGCCTCATGTTCTCG-3′ for the T567D mutant; (forward) 5′-AAGCTTTGCGGCAGATCCGGCAGGGCAACACCAAGC-3′ and (reverse) 5′-TGTACTTGTCCCGGCCTTGCCTCATGTTCTCG-3′ for the T567A mutant. Primers used for the MyrPB-ezrinT235D-rLucII mutant were: (forward) 3′-CCCAAAGATTGGCTTTCCTTGG-5′ and (reverse) 3′-TCTAACTTATCATCTTTCTCATAAATATTCAGTCC-5′. (MyrPB)-ezrin-rlucII mutants for the PdtIns(4,5)P2-binding motif [(MyrPB)-ezrinK253,254,262,263N-rLucII] were obtained using the following primers: (forward) 5′-CCCATCGACAACAACGCACCTGACTTTGTGTTTTATGC-3′ and (reverse) 5′-TTTAATGACAAAGTTGTTGTCATTGAAAGAGATGTTCC-3′. For radixinT564D-rLucII and radixinT564A-rLucII, primers used were: (forward) 5′-AAGATCTGCGACAGATTCGACAAGGCAATACAAAGC-3′ and (reverse) 5′-TGTACTTATCACGGCCTGCTTTAACATTCTCAGC-3′ for the T564D mutant; (forward) 5′-AAGCTTTGCGACAGATTCGACAAGGCAATACAAAGC-3′ and (reverse) 5′-TGTACTTATCACGGCCTGCTTTAACATTCTCAGC-3′ for the T564A mutant. Finally, primers used for moesinT558D-rLucII and MoesinT558A-rLucII mutants were: (forward) 5′-AAGATCTGCGCCAGATCCGGCAGGGCAACACCAAGC-3′ and (reverse) 5′-TGTATTTGTCTCGGCCCAGTCGCATGTTCTCAGC-3′ for T558D mutant; (forward) 5′-AAGCTTTGCGCCAGATCCGGCAGGGCAACACCAAGC-3′ and (reverse) 5′-TGTATTTGTCTCGGCCCAGTCGCATGTTCTCAGC-3′ for T558A mutant. Ezrin–GFP was obtained by subcloning PCR amplified ezrin into pEGFP-N1 vector (Clontech). All PCRs were performed using Phusion High-Fidelity DNA Polymerase (New England Biolabs). MISSION shRNA constructs were obtained from Sigma in pLKO.1-puro vectors: SLK (TRCN0000000897) and PPP1CA (TRCN0000002453).
Cell culture, transfection and infection
HEK293T human kidney cells were maintained in Dulbecco's modified Eagle's medium (DMEM; 4.5 g/l D-glucose, L-glutamine, 110 mg/l sodium pyruvate; Thermo Fisher Scientific #11995073) supplemented with 10% fetal bovine serum (FBS; Life Invitrogen #12483020) and 1% penicillin-streptomycin antibiotics (Thermo Fisher Scientific #15070063) at 37°C with 5% CO2. For transfection, cells (350,000 cells/ml) were resuspended in DMEM supplemented with 10% FBS and transfected with 1 µg of total DNA (50 ng rLucII construct, 300 ng rGFP-CAAX and 650 ng salmon sperm DNA) using linear polyethyleneimine (PEI, Alfa Aesar #43896) as transfecting agent with a PEI:DNA ratio of 3:1. Cells were then plated in white 96-well culture plates (VWR #82050-736) at a concentration of 35,000 cells per well and were incubated for 48 h prior to BRET measurement. For immunoblotting experiments, cells (350,000 cells/ml) were transfected with 1 µg of rLucII construct using PEI. Cells were then plated on regular six-well culture plates (VWR #82050-842) at a concentration of 350,000 cells per well and incubated for 48 h prior lysis.
For lentiviral infection, cells were seeded at 5000 cells per well in 96-well culture plates in DMEM supplemented with 10% FBS and 5 µg/ml polybrene (Sigma #H9268). Lentiviruses were added to the medium and cells were incubated for 24 h. The next day, cells were transfected with 100 ng total DNA using PEI (as described above) and incubated for an additional 24 h. Finally, infected cells were selected with 2 µg/ml puromycin (EMD Millipore #540222) for 48 h.
Transfected cells with BRET biosensors were washed with Hank's balanced salt solution (HBSS, Thermo Fisher Scientific #14065056). Calyculin A and staurosporine were added at 100 nM for 10 min and 30 min, respectively, unless mentioned otherwise. Coelenterazine 400a diluted in HBSS was added 5 min prior to the reading at a final concentration of 2.5 µM. BRET was monitored with a Tecan Infinite 200 PRO multifunctional microplate reader (Tecan) equipped with BLUE1 (370–480 nm; donor) and GREEN1 (520–570 nm; acceptor) filters. BRET signals was calculated as a ratio by dividing the acceptor emission value by the donor emission value. For automated procedures, inhibitors were added using an Echo 555 acoustic dispenser (Labcyte) while BRET2 was monitored with a Synergy NEO HTS microplate reader (Biotek).
At 48 h before imaging, HEK293T cells were transfected as described before with rLucII-tagged BRET donors and rGFP-CAAX acceptor and plated in 35 mm glass bottom culture dishes (MatTek #P35GC-0-14-C). The day of the experiment, cells were washed with HBSS and inhibitors were added. rLucII substrate methoxy e-CTZ was added at a final concentration of 10 µM prior to the acquisition. Images and analysis were obtained as previously described (Kobayashi et al., 2019). An inverted microscope (Nikon Eclipse Ti-U) equipped with a 60× objective lens (Nikon CFI Apochromat TIRF) and EM-CCD camera (Nuvu Cameras HNu 512) was used to acquire images. The EM-CCD camera was set in photon counting mode, and 100 successive frames of 100 ms exposure each (total 10 s) were acquired alternatively without filter (total luminescence) or with a 480 nm longpass filter (acceptor) using a filter changer (Sutter Lambda 10-2) inserted before the camera. The acquisitions were repeated five times and all the photon counting images taken for 50 s were integrated. BRET values were finally obtained by dividing the acceptor counts by the total luminescence counts pixelwise.
After treatment, HEK293T cells were washed with ice-cold phosphate-buffered saline (PBS) and lysed in TLB buffer (40 mM HEPES, 1 mM EDTA, 120 mM NaCl, 10 mM NaPPi, 10% glycerol, 1% Triton X-100, 0.1% SDS) supplemented with both phosphatase and protease inhibitors [phosphatase inhibitor cocktail (PIC, Sigma #P2850), 1 mM sodium orthovanadate (Na3VO4, Sigma #S6508), 5 mM β-glycerophosphate (Sigma #G6251), 1 mM PMSF (Sigma #P7626) and anti-protease cocktail (Sigma #4693132001)]. After Bradford quantification, cell lysates were diluted with sample buffer (200 mM Tris-HCl pH 6.8, 8% SDS, 0.4% Bromophenol Blue, 40% glycerol and 412 mM β-mercaptoethanol). Samples were resolved by 8% SDS-PAGE and transferred to nitrocellulose membranes (pore 0.2 μm, VWR #27376-991). Membranes were blocked in TBS-Tween (25 mM Tris-HCl pH 8, 125 mM NaCl and 0.1% Tween 20) supplemented with 2% BSA for 1 h before overnight incubation with primary antibodies at 4°C. Primary antibodies used were: rabbit anti-ERM (1:1000, Cell Signaling #3142), rabbit anti-ezrin (1:1000, Cell Signaling #3145), rabbit anti-radixin (1:1000, Cell Signaling #2636), rabbit anti-moesin (1:1000, Cell Signaling #3150), rabbit anti-phospho-ERM (1:5000; Roubinet et al., 2011), mouse anti-actin (1:5000, Sigma #MAB1501), rabbit anti-SLK (1:500, Cerdalane #A300-499A) and rabbit anti-PPP1CA (1:1000, Cedarlane #A300-904A-M). Goat anti-rabbit-IgG HRP antibody (1:10,000, Santa Cruz Biotechnology #sc-2004) and goat anti-mouse-IgG HRP (1:10,000, Santa Cruz Biotechnology #sc-516102) were used as secondary antibodies. Finally, protein detection was performed using Amersham ECL western blotting detection reagent (GE Healthcare #CA95038-564L). Immunoblot quantifications were performed using ImageJ software (NIH).
Cytosol and membrane fractionation
Fractionation experiments were performed as previously described (Gautreau et al., 2000). Briefly, HEK293T cells were scraped off and mechanically disrupted using a 27 G syringe in cold buffer (10 mM HEPES, 1 mM EDTA, 150 mM NaCl pH 7.4) supplemented with both phosphatase and protease inhibitors (phosphatase inhibitor cocktail, 1 mM Na3VO4, 5 mM β-glycerophosphate, 1 mM PMSF and anti-protease cocktail). After a first centrifugation at 600 g for 10 min at 4°C to remove debris and nuclei, the supernatant was then subjected to a 25-min centrifuge at 100,000 g using a S120AT3 rotor in a Sorvall Discovery M150SE centrifuge (Hitachi). Membranes were recovered in the pellet, while the cytosolic fraction was taken as the supernatant.
At 48 h before the experiment, HEK293T cells were transfected with ezrin–GFP in 35 mm glass bottom culture dishes (MatTek #P35GC-0-14-C) as described above. The day of the experiment, images were acquired every minute using Yokogawa CSU-X1 5000 spinning disk confocal microscope. Quantification of the signals at the plasma membrane was performed using Image J software (NIH).
All data from BRET and western blot analysis were analyzed using GraphPad PRISM software (GraphPad Software, La Jolla, CA, USA). All data are represented by the mean±s.d. of multiple independent experiments. Microscopy images were prepared using Image J software (NIH) and Photoshop (Adobe).
Conceptualization: K.L., B.D., C.L., M.B., S.C.; Methodology: C.L., M.B.; Validation: K.L.; Formal analysis: K.L., B.D., Y.Y.H., C.L., M.B., S.C.; Investigation: K.L., B.D., Y.Y.H., A.P., M.H., H.K., C.L., M.B.; Data curation: S.C.; Writing - original draft: K.L., S.C.; Writing - review & editing: M.B., S.C.; Visualization: K.L.; Supervision: M.B., S.C.; Project administration: S.C.; Funding acquisition: M.B., S.C.
This work has been supported by Canadian Cancer Society Research Institute (CCSRI) Innovation Grant (705892 to S.C.) and a Foundation Grant from the Canadian Institute for Health Research (148431 to M.B.). K.L. held a doctoral scholarship from IRIC and from Montreal University's Molecular Biology Program as well as a Études Supérieures et Postdoctorales (ESP) fellowship from Montreal University. Y.Y.H. held a master scholarship from IRIC. M.B. holds the Canada Research Chair in Signals Transduction and Molecular Pharmacology.
Peer review history
The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/134/7/jcs255307/
The authors declare no competing or financial interests.