Allosteric inactivation of an engineered optogenetic GTPase

Significance

Control of cell signaling activity in proteins by light is one of the primary goals of optogenetics. The hybrid light-receptor/cell-signaling protein Cdc42Lov was engineered recently as an optogenetic tool, employing a novel allosteric strategy that results in photoinhibition. In contrast to previous activation designs, the mechanism of inactivation of GTPase signaling activity via insertion of the LOV2 domain in cell division control protein 42 (Cdc42) is only apparent at a detailed structural and dynamic level. NMR characterization of dark and mutationally “lit” forms reveals the allosteric interdomain perturbations and bidirectional signaling, the knowledge of which will enhance future applications of this design strategy.

Abstract

Optogenetics is a technique for establishing direct spatiotemporal control over molecular function within living cells using light. Light application induces conformational changes within targeted proteins that produce changes in function. One of the applications of optogenetic tools is an allosteric control of proteins via light-sensing domain (LOV2), which allows direct and robust control of protein function. Computational studies supported by cellular imaging demonstrated that application of light allosterically inhibited signaling proteins Vav2, ITSN, and Rac1, but the structural and dynamic basis of such control has yet to be elucidated by experiment. Here, using NMR spectroscopy, we discover principles of action of allosteric control of cell division control protein 42 (CDC42), a small GTPase involved in cell signaling. Both LOV2 and Cdc42 employ flexibility in their function to switch between “dark”/“lit” or active/inactive states, respectively. By conjoining Cdc42 and phototropin1 LOV2 domains into the bi-switchable fusion Cdc42Lov, application of light—or alternatively, mutation in LOV2 to mimic light absorption—allosterically inhibits Cdc42 downstream signaling. The flow and patterning of allosteric transduction in this flexible system are well suited to observation by NMR. Close monitoring of the structural and dynamic properties of dark versus “lit” states of Cdc42Lov revealed lit-induced allosteric perturbations that extend to Cdc42’s downstream effector binding site. Chemical shift perturbations for lit mimic, I539E, have distinct regions of sensitivity, and both the domains are coupled together, leading to bidirectional interdomain signaling. Insights gained from this optoallosteric design will increase our ability to control response sensitivity in future designs.

Over the last several years, there has been great interest in controlling the activity of proteins with light (1–8). The field of “optogenetics” has expanded our understanding, in situ, of the role of specific proteins in signal transduction and, especially, how neuronal processes are controlled (9–11). In the cellular environment, proteins are naturally controlled in many ways, including by allosteric regulation (5, 9–13). Recently, an optogenetic approach was employed to regulate the activity of signal transduction proteins allosterically, by engineering in a modulatory light-sensing domain (6, 14, 15). A computational strategy was employed for introducing the optogenetic sensor—a light-oxygen-voltage sensing domain (LOV)—into loops of proteins that regulate cell motility, which included the two GTPases cell division control protein 42 (Cdc42) and Rac1 (6, 13, 16). In the designs, light-induced changes of the LOV domain are transferred into perturbations to the GTPase, resulting in allosteric regulation of activity (6).

Rho family GTPases constitute a family of small (~21 kDa) signaling G proteins that bind and hydrolyze guanosine triphosphate (GTP) (17). Members of this family, such as Cdc42 and Rac1, act as molecular switches by shuttling between inactive and active states upon binding to guanosine diphosphate (GDP) and GTP (18). In the abovementioned study, Cdc42 was engineered by the introduction of the Avena sativa phototropin1 LOV2 domain such that it is regulatable by light in vivo (6), enabling spatiotemporal control. The general model of LOV2 action is that, upon application of blue light and absorption by flavin mononucleotide (FMN) cofactor, the LOV2 domain Jα-helix becomes disordered (19, 20). Because the LOV2 domain is at the domain junction and allosterically linked to the nucleotide binding site, the Jα deformation transfers into the Cdc42 domain to down-regulate GTPase effector binding activity, resulting in photoinhibition (6).

This photoinhibitory LOV2 insertion strategy is fundamentally different when compared to other LOV2-based optogenetic modulations of activity. In most cases, LOV2 has been employed in an activating scheme in which light absorption results in complete domain dissociation or uncaging of the protein of interest (21–24), or in dimerization of a split protein (21, 25, 26). Domain-level uncaging has more recently been used in a circularly permutated LOV2 system to either activate or inhibit activities in new ways (23). The photoinhibitory LOV2 insertion strategy (6, 27), however, stands in contrast to uncaging/dissociation in that modulation of activity occurs while the target domain interfacial structures are presumably retained, and that allosteric regulation operates through a more subtle propagation of perturbed interactions or dynamics. Importantly, this allosteric photomodulation strategy is being employed in a growing number of applications, such as with enzymes (28, 29) and, notably, with optogenetic nanobodies (30) and monobodies (31), and nanocomputing devices (8), reflecting its general utility. Yet, some of those studies arrive at the best designs through either mild screening or trial and error with respect to insertion sites, and specific LOV2 insertions can unpredictably lead to either inhibition or activation. These powerful tools are clearly advancing optogenetic technology, but how activity is allosterically modulated is generally not well understood.

It is therefore important to gather more molecular insight into this allosteric optogenetic mechanism. Presumably, target domain (e.g., Cdc42) inhibition results from an allosteric transfer of some kind of strain and/or conformational transition, although it is possible that Jα-helix unwinding might lead to target domain unfolding. Obtaining more mechanistic detail about how photoinhibition is achieved will be important to improve the design. A similar question is whether close tethering of the Jα-helix to the target protein in these designs degrades LOV2’s innate allosteric light responsiveness. Ultimately, the major question is how light absorption in LOV2 allosterically propagates into Cdc42, and what structural regions are affected.

To address these questions and gain a detailed look into an “optoallosteric” protein, we have undertaken a study of the hybrid photoinhibitable protein Cdc42Lov, which to date has only been characterized in vivo (6). Cdc42 function was assayed in vitro with purified components to obtain a quantitative assessment of optogenetic function, and a detailed NMR analysis has been carried out on Cdc42Lov to assess the allosteric mechanism. NMR is a suitable method to probe Cdc42Lov not only because both LOV2 and Cdc42 domains are intrinsically flexible as they are both “switch” proteins, but also because there is likely some interdomain flexibility that may be modulated in the lit state. Both dark and “lit” forms were studied via NMR chemical shift changes, with the “lit” form stabilized through the use of a previously characterized LOV2 domain mutation, I539E (32). Extensive NMR chemical shift assignment and transverse relaxation-optimized spectroscopy-heteronuclear single quantum coherence (TROSY-HSQC) spectral changes revealed that overall the Cdc42 domain remains structurally intact in the lit form, and that key functional regions of both domains are perturbed structurally and dynamically, providing insight into the allosteric mechanism. Interestingly, allosteric communication was found to be bidirectional, as binding of Pak1 effector to Cdc42 resulted in significant changes in critical photosensitive regions of LOV2.

Results and Discussion

The computational design and in vivo function of Cdc42Lov, a light-responsive, allosteric fusion of human Cdc42 and the second LOV domain from A. sativa phototropin1 (AsLOV2), was described previously (6). In this design, the LOV2 domain is inserted into the β3-β4 loop of Cdc42, such that it is bordered by the N-terminal segment of Cdc42 (1 to 47) in the preceding residues and the C-terminal segment of Cdc42 (48 to 178) in the following residues (Fig. 1 and SI Appendix, Fig. S1). This mode of insertion positions the LOV2 domain at a different surface from the nucleotide and Cdc42 effector binding surfaces and results in photoinhibition of Cdc42 activity. In vivo, the inhibitory effect was a twofold-to-threefold activity change for Cdc42Lov, but no characterization has been carried out for single steps in vitro, such as how light affects Cdc42 effector binding. Thus, basic mechanistic information on how the Cdc42Lov domain junction partitions and directs destabilizing energy, as well as how this impacts functional interactions, will deepen our understanding of this novel allosteric design. We sought to track the communication pathway linking LOV2 and Cdc42 domains using mutations that mimic light absorption and monitoring with NMR spectroscopy.

Fig. 1.

Optogenetic design of allosteric inactivation by light. (A) Gene construct of Cdc42Lov showing insertion of LOV2 domain into Cdc42, ~330 residues. Natural numbering of each domain is retained. On top is LOV2 domain and bottom is Cdc42 numbering. (B) Schematic of allosteric mechanism showing changes from active state to inactive state upon light illumination. Effector is shown in magenta. (C) ITC data for PAK1 effector peptide binding to Cdc42Lov^{WT, CA} (Left, K_d = 0.3 µM) and light activation mimicking I539E Cdc42Lov^{LM, CA} (Right, K_d = 88 µM).

Several mutations were employed to control states of both domains. For simplicity and relation to familiar residue numbering of Cdc42 and AsLOV2, we retain the respective single domain numbering. The I539E (LOV2 domain) mutation mimics the effects of blue light absorption and was, therefore, used as a proxy for the “lit” state of Cdc42Lov (LM) (21). While “dark” state mutants also exist (C450A) (DM), these were typically not needed since measurements on wildtype (WT) protein were made in the absence of light. The Q61L mutation in Cdc42 is known to be constitutively active (33) and was used to favor the GTPase active conformation. GTPase activity is modulated by the exchange of GDP/GTP that toggles inactive vs. active conformations (34, 35). In Cdc42Lov, light absorption may alter GDP/GTP exchange or directly alter the effector binding site, or some combination of both. Because we found that Cdc42Lov allosteric function exists in vitro by using GDP in conjunction with the Q61L mutation (below), NMR studies were conducted with GDP nucleotide, along with the Q61L mutation, to favor the active conformation (36) (denoted by CA = constitutively active) as noted (Materials and Methods).

“Lit” (LM) Conformation Allosterically Inhibits Cdc42 Effector Binding In Vitro.

The effect of the I539E lit mutation on Cdc42Lov (Cdc42Lov^LM) was tested by measuring binding affinity to Cdc42 downstream effector P21-activated kinase (Pak1). Pak1 is a serine/threonine kinase that only binds to the active form of Cdc42 (37), and the interaction can be shown with binding of a ~50-residue Pak1 segment to the constitutively active (CA) Q61L mutant of Cdc42 (38, 39). Pak1 peptide binding was evaluated for Cdc42Lov^WT,CA and Cdc42Lov^LM,CA using isothermal titration calorimetry (ITC). As expected, the unlit (i.e., WT LOV2 sequence) protein showed high-affinity binding with K_d = 0.3 µM, indicating that the Cdc42 domain retains full effector binding function (Fig. 1C and SI Appendix, Table S1). By contrast, mutationally “lit” Cdc42Lov^LM,CA showed greatly reduced binding affinity, with K_d ~90 µM (Fig. 1C and SI Appendix, Table S1). Thus, in vitro, the light-mimicking LOV2 domain perturbation is allosterically transduced to the Cdc42 domain to drastically reduce (>100-fold) effector binding function.

NMR Spectral Comparison of Cdc42Lov^WT and Cdc42Lov^LM.

To gain structural insight into the optogenetic allosteric mechanism of Cdc42Lov, NMR signals of “lit” Cdc42Lov^LM were compared to “unlit” Cdc42Lov^WT. We confirmed that the 2D ¹H–¹⁵N HSQC spectrum of the wild-type protein was nearly identical compared to Cdc42Lov^DM (SI Appendix, Fig. S2) due to the absence of visible light in the magnet. While both lit and unlit forms show spectra with numerous peaks and significant chemical shift dispersion indicative of folded proteins, the unlit (WT) form has more peaks with uniform intensity (Fig. 2). Spectral overlay suggests the presence of some slow μs-ms dynamics in the lit state. The greater degree of variability in peak intensities, along with some peaks collapsed into the center region, in the lit form spectrum suggests a subpopulation of destabilized species or regions of partially unfolded structure in the lit form (Fig. 2). Interestingly, circular dichroism (CD) spectra of both forms showed only a slight difference (SI Appendix, Fig. S3), indicating that the lit form appears to retain nearly all of its helical structure, and size exclusion chromatography with multi-angle static light scattering (SEC-MALS) also indicated similar sizes for lit and unlit states (SI Appendix, Fig. S3). Given that the gross structural features of lit and unlit Cdc42Lov appear similar, the NMR signals were analyzed in greater detail to detect differences in the two forms.

Fig. 2.

TROSY-HSQC of unlit and lit states of Cdc42Lov design. (A) Cdc42Lov^WT “unlit” HSQC. (B) Cdc42Lov^LM “lit.” (C) Backbone assignment of Cdc42Lov^WT. Assignments are shown on a model generated using the SWISS modeler. In green are assigned residues, gray unassigned, and in red are the unassigned switch regions of the Cdc42 domain.

Light-Triggered Interdomain Force Transduction Monitored by Chemical Shift Perturbations (CSPs).

To gain structural insight into how light allosterically transmits from LOV2 to Cdc42, CSPs were recorded. As the design couples initial LOV2 domain perturbation to the Cdc42 domain, significant CSPs in the LOV2 domain, especially from Jα-helix unwinding, are expected. Backbone chemical shifts for Cdc42Lov^WT and Cdc42Lov^LM were assigned using standard triple-resonance methods on perdeuterated proteins (Methods and Materials). Assignments for unlit and lit forms were made for 68% and 62% of nonproline residues, respectively, with peak broadening from dynamic behavior being the primary reason for unassigned residues (Fig. 2C). In principle, CSPs can give insights into adjustments made in both domains and trace critical residues that lie along the communication pathway between LOV2 and Cdc42 functional sites (Fig. 3A). Lit form-induced chemical shift effects are shown in Fig. 3. One of the key LOV2 domain residues is C450, which is known to form a covalent adduct with FMN upon illumination (39). The unlit state has a prominent amide peak for C450 which becomes completely broadened in Cdc42Lov^LM (Fig. 4A). Similarly, Q513, which serves as a “glutamine lever” that induces unwinding of the Jα-helix, is also completely broadened in Cdc42Lov^LM (Fig. 4B) (40). Adduct formation and glutamine lever movement are coupled events, and their broadening suggests that the coupling is retained in Cdc42Lov^LM. Overall, this behavior in the LOV2 domain is fully consistent with I539E serving as an effective proxy for light absorption.

Fig. 3.

Impact of I539E mutation (light mimic) on Cdc42Lov. (A) Top, Chemical shift perturbation map shown in white-to-red scale showing increasingly perturbed residues. Broadened residues are in cyan, and unassigned residues are in black. Bottom, histogram showing CSPs for Cdc42 and LOV2 domain. PAK1-interacting region is shown with orange ovals. Broadened residues are shown as blue bars on the plot. (B) Intact-residue map based on the intensity of peaks in “lit” state (as described in main text). Intact residues are highlighted in magenta, cyan are residues with nonzero CSPs or broadened residues, and unassigned residues are in black. FMN and GDP are shown in green sticks and spheres, respectively, and Pak1 (PDB 1e0a) is shown in orange. The side chain of residues involved in signal transduction for LOV2 is shown in yellow sticks.

Fig. 4.

CSP peak comparison of unlit and lit states. Cdc42Lov^WT “unlit” peaks are shown in black and Cdc42Lov^LM I539E “lit” mimic peaks are in red.

In line with the coupling design, the I539E mutation results in extensive CSPs and peak broadening for various secondary structures surrounding the FMN binding pocket. This includes residues in β-strands G, H, and I, such as R451, I466, D471, V478, Q479, F494, L496, M499, Y508, F509, and L514 (Fig. 4 and SI Appendix, Fig. S4), with these residues showing either peak shifting or disappearance/broadening, indicative of a structural rearrangement. Although light activation leads to the unwinding of the Jα-helix in native LOV2 (6, 28), strict CSPs do not indicate complete unwinding of the Jα-helix in Cdc42Lov^LM. While this may result from stabilization conferred from confinement by Cdc42, CSPs alone can be misinterpreted since they are based entirely on the peak position of assigned residues, and only require that peaks be present and assignable. Closer inspection showed that some assigned peaks in the LOV2 domain are greatly reduced in their intensity (or broadened). Thus, while CSPs can track perturbations, they can be less sensitive to changes in dynamics or local stability.

To address this caveat, we included another metric for perturbation based on intensities. Here, the residues are binned as “intact” if the peak position and intensity are comparable to the unlit form (lit peak is in close vicinity and intensity approximately at least 40% of unlit), or else it is identified as “not intact” (SI Appendix, Table S2). This intact-residue map (Fig. 3B) provides another level of understanding for lit-form perturbations and shows which protein regions retain stably folded structure. The core region with FMN binding pocket shows an equal mixture of intact and nonintact residues (Fig. 3B and SI Appendix, Fig. S5). Almost all residues of the Jα-helix including I539 appear as nonintact except L546 (Fig. 4F and SI Appendix, Fig. S5) and G547, part of the hinge region, and L531. Together, CSPs and the intactness map show that although the Jα-helix in the LOV2 domain appears to have some residues that are minimally perturbed, the entire Jα-helix is undergoing transitions or conformational change.

Interestingly, HSQC peak perturbations are also observed in the Cdc42 domain of Cdc42Lov^LM. These perturbations are somewhat weaker compared to those in the LOV2 domain but nevertheless indicate that Cdc42Lov^LM is transducing the light-induced effect into Cdc42. A major region of interest is the Pak1-binding interface since application of light alters Pak1 binding (Fig. 1C) (38). Residues in Cdc42 that interact with Pak1 are 21 to 25 (α2), 36 to 47 (β3), 67 to 72 (switch 2 or α5), and 166 to 178 (α13) (orange “ovals” in Fig. 3A). Of these residues, approximately half were assigned and were used to monitor the impact of I539E/“light” perturbation. The majority of these were observed in the C-terminal segment, within α13, by either exhibiting CSPs or broadening (Fig. 3A and SI Appendix, Fig. S6). Specifically, residues K166, D170, I173, and L174 are impacted by the I539E mutation and thus are likely involved in allosterically reducing Pak1-binding affinity (Figs. 3A and 4C). While most of the N-terminal segments of the Cdc42 domain could not be assigned, there was notable broadening in a few residues within α2, including I23 (Fig. 4D) which makes extensive contacts with F81 of Pak1. It is likely that switches 1 and 2 are also perturbed by I539E, but these regions could not be assigned in any state, which is typical for Cdc42 and GTPases in general (41). Overall, upon considering the residues comprising the Pak1-binding interface, of the residues that are observable for tracking CSPs, nearly every one of these residues (orange ovals in Fig. 3) in close proximity to Pak1 exhibits nonzero CSPs. This confirms that allosteric photoinhibition operates by directed perturbations to the binding interface.

To get better insight into how light-induced signaling propagates into Cdc42 and effector binding regions, we looked at the connecting domain hinge region and other distal parts of Cdc42. At the hinge, between loops β3 and β4, a few residues, L546, G547 (both in LOV2), and G48 (Cdc42), are slightly perturbed but still appear intact, suggestive of the strong connection between the two domains. Furthermore, the CSPs in the Jα-helix propagate to the adjacent β4 strand of the Cdc42 domain (Fig. 3A). T52, L53, and G54 are a few of the β4 strand residues that experience perturbation due to the unwinding of Jα-helix (Fig. 4 H and I). Thus, Jα-helix unwinding is translating structural changes to the connected β4 strand consistent with the design hypothesis that these strands mediate the propagation of signals allosterically. However, light-induced changes extend beyond the immediately connected strands to the Jα-helix. For example, residues in the β6 strand (F78 to F82) show peak shifting or broadening (Fig. 3A). F78 and V80 make a triad with distal residue I101, that also experiences a light-induced CSP (Figs. 3A and 4J). Except for L79, all other residues in the β6 strand are broadened, confirming that the I539E mimic of light absorption penetrates central regions of Cdc42.

The other primary region for observing structural perturbations is the nucleotide-binding site because of the functional coupling between nucleotide status (GDP vs. GTP) and effector binding in GTPases. In Cdc42, the nucleotide-binding site is formed by the switch 1 (28–40) and switch 2 (57 to 74) regions (27, 28), and the three loops formed by residues 12 to 17 (connecting β1-α2), 115 to 120, and 158 to 161 (connecting β12-α13) (SI Appendix, Fig. S6). Perturbations to the switch regions may be expected via translation through the connecting strands β3 and β4. However, because of the intrinsic dynamic nature (loss of NMR signals) of the two switches, we could only observe perturbations to the three loops. Specifically, L160 shows a sizeable shift perturbation (0.05 ppm) (Fig. 4N and SI Appendix, Fig. S6), and T115 is broadened, indicating a mixture of structural perturbation and conformational instability in the nucleotide-binding site upon I539E mutation (Fig. 3A).

Unlike other main regions of Cdc42 domain, the core region Q116-K150 shows little perturbation and remains intact, suggestive of controlled transmission with directed allosteric communication (i.e., some regions are entirely unperturbed) from the LOV2 domain (Figs. 3 and 4). This unperturbed intact core region is followed by a C-terminal helix (α13) that showed a high magnitude of light-dependent CSP. Interestingly, α13 is not only in close proximity to the nucleotide-binding site via L160 but also involved in interacting with downstream effector Pak1 (as mentioned before).

Collectively, these HSQC peak perturbations show that Cdc42Lov^LM is undergoing allosteric conformational change, with both domains showing perturbation in the lit-mimicked state. The conformational adjustments in the Cdc42 domain are mainly around regions that are in close proximity to the nucleotide-binding region and C-terminal helix. Also, there is a structurally unperturbed intact region suggestive of controlled transmission of signals. Overall, the data support that this Cdc42Lov^LM mimic, an observed “lit” state, is working according to the strategy of inserting the LOV2 domain and achieves the goal by altering the chemical shifts of β3 (they can no longer be assigned) and β4.

Pak1-Bound CSPs Confirming Bidirectional Allosteric Communication.

To further evaluate the coupling of domains, we observed how Pak1 binding (downstream effector) affects Cdc42Lov. Since Pak1 only binds to the active state of Cdc42, we used Cdc42Lov^WT,CA (Q61L, constitutively active) to study the complex. Upon binding Pak1, substantial CSPs were observed in α2 (K18-T27), α7 (F90-E95), α10 (D122-K135), and α11(A142-D148) in the Cdc42 domain (Fig. 5 and SI Appendix, Fig. S7 and Table S3), all of which have been shown to undergo conformational changes upon binding Pak1 (33, 42, 43) (Fig. 5 and SI Appendix, Fig. S7). In addition to the above helices, extensive broadening and perturbations were observed in the α13 (L165-E178) helix (SI Appendix, Fig. S8). Interestingly, α13 is also perturbed in I539E lit mimic, suggesting a possible coupled behavior between both domains.

Fig. 5.

Pak1 binding (A) Overlay of ¹⁵N TROSY-HSQC of Cdc42Lov^WT,CA and Cdc42Lov^WT,CA Pak1 bound in black and red, respectively. (B) Chemical shift perturbation map for Pak1 binding to Cdc42Lov^WT,CA. Color coding as in Fig. 3B.

Pak1 binding introduced changes not only in the Cdc42 domain but also in the LOV2 domain (Fig. 5 and SI Appendix, Fig. S7). Overall perturbations in the LOV2 domain are in the core region and around the Jα-helix. Closer inspection showed that R451, a residue next to C450, shows chemical shift changes upon Pak1 binding, indicative of allosteric communication to the key residue involved in adduct formation in the lit state. Similarly, residues close to the glutamine (Q513) lever, I510 and G511, were found perturbed upon Pak1 binding. These were the sites responsive to I539E (light mimic), yet it is striking that these same sites are clearly perturbed upon distal binding of Pak1. Similarly, in the Cdc42 domain, I101, L160, and a majority of the α13 helix were all perturbed by both I539E mutation and Pak1 binding. Taken together, these in-common sites of perturbation indicate that the nature of domain coupling leads to bidirectional interdomain signaling (SI Appendix, Fig. S9). While this confirms the allosteric interactions as shown in numerous cases (44–46), in general, long-range communication need not be bidirectional (47).

In this study, we have tracked the allosteric response of an optogenetically encoded structural distortion in an engineered two-domain hybrid protein, Cdc42Lov. Though the allosteric photoinhibition of Cdc42 function was consistent with the design, knowledge of the operative mechanism has been lacking. Due to the flexible nature of the interdomain linkage, we have opted for a solution-based approach to reveal clues into the allosteric mechanism. NMR characterization of the light-mimicking response of I539E has the advantage of site-specific tracking using chemical shifts and peak intensities that report on dynamics. Monitoring of chemical shifts allowed regions in Cdc42Lov to be identified as experiencing structural perturbation and dynamic switching behavior—as well as general regions of the protein that remain structurally intact vs. locally destabilized—in response to light absorption. As protein engineering advances to adopt similar, improved, or alternative strategies to implement allosteric functional properties, the high-resolution solution approach afforded by NMR should be a valuable tool for assessing such mechanisms, especially when systems possess flexibility that may make other structural approaches more challenging.

Materials and Methods

Protein Expression and Purification.

The photosensitive construct of Cdc42 using LOV2 insertion (Cdc42Lov, Fig. 1 and SI Appendix, Fig. S1) was designed and synthesized as described previously (6) (Bio Basic Inc.) with an often-used Q61L mutation that confers constitutively active Cdc42 (Cdc42^CA). The C450A and I539E, “dark” and “lit” mutants, as well as Pak1-binding protein (Pak1) domain constructs with a C terminus 6xHis-tag [PET23-PBP(65-109)-N-Cys-His6] were expressed in Escherichia coli and purified as described in SI Appendix, Materials and Methods.

ITC.

ITC samples for all variants and Pak1 peptide were prepared in NMR buffer with Tris (2-carboxyethyl) phosphine (TCEP) as a reducing agent at pH 7.5. Experiments were conducted on a MicroCal AutoITC 200 (Malvern Panalytical). Protein concentration in the cell was set to 20 μM and 150 μM, for unlit and lit states, respectively. Pak1 concentration was 10 times higher depending upon cell protein concentration. Raw thermograms were integrated using NITPIC (48) and fitted in SEDPHAT (49) software. The single-site binding model was used for data fitting.

NMR Spectroscopy.

Standard transverse relaxation-optimized spectroscopy (TROSY) triple resonance and ¹H–¹⁵N heteronuclear single quantum coherence (TROSY-HSQC) experiments were used for backbone assignment and CSP calculations (50), respectively, as described in SI Appendix, Materials and Methods.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Cdc42Lov chemical shift assignment data have been submitted to the BMRB (Biological Magnetic Resonance Bank), ID 51681. All study data are included in the article and/or SI Appendix.