Introduction

Cell division control protein 42 homolog (Cdc42) is a member of the Rho GTPase family and was identified in 1990 [1]. It plays a pivotal role in the regulation of cell polarity, vesicle trafficking, cytoskeletal and microtubule dynamics, as well as cell proliferation and adhesion [2,3,4,5]. Thus, Cdc42 is closely related to various pathogenic processes such as cancer, cardiovascular diseases, and neuronal degenerative diseases [6,7,8].

Like other Rho GTPase members, Cdc42 acts as an essential molecular switch in intracellular signaling networks. To perform this function, Cdc42 cycles between inactive guanine diphosphate (GDP)-bound state and active guanine triphosphate (GTP)-bound state, which is controlled by guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs). GEFs simulate the nucleotide exchange from GDP to GTP, resulting in the formation of Cdc42-GTP and the following activation of its downstream effectors [9]. GAPs, which assist with the hydrolysis of GTP, inactivate Cdc42 by promoting the formation of Cdc42-GDP [10].

Additionally, Cdc42 also cycles between membrane-associated and cytosolic states. Membrane attachment is a prerequisite for the activation of Cdc42, which depends on the post-translational modification of Cdc42 by prenylation. Prenylation is the process that a geranylgeranyl isoprenoid is attached to the Cys188 of -CAAX (CAAX is a conserved sequence in the C-terminus of Rho GTPases, where C means cysteine, A can be any aliphatic amino acid, and X can be any amino acid) in Cdc42, followed by -AAX peptide cleaved from Cdc42 and a carboxymethyl group added to the prenylated Cys188 [11]. Here, we shortened the geranylgeranylated and carboxymethylated Cys188 as non-standard Cys188 for convenience.

GDP dissociation inhibitors (GDIs, another type regulators of Rho GTPases) can extract the Cdc42 from membrane to cytosol by binding the geranylgeranyl group of the non-standard Cys188 to control its subcellular distribution and its cycle between membrane-associated (active) and cytosolic (inactive) states [12]. In cytosol, GDIs bind the inactive GDP-bound form of Cdc42 to maintain Cdc42 in an inactive state. GDIs also act as chaperones for Cdc42 to protect it from degradation [13]. Thus, GDIs are multifunctional regulators for Rho GTPases.

Unlike a large number of GEFs (82 members) and GAPs (69 members), only 3 members of GDIs are expressed in mammalian cells for Rho GTPases: RhoGDI1 (also known as RhoGDI1α), RhoGDI2 (also known as RhoGDIβ or LY-GDI or D4-GDI), and RhoGDI3 (also known as RhoGDI1γ) [14]. Cdc42 has been found to interact with both RhoGDI1 and RhoGDI2, but it is preferentially and effectively regulated by RhoGDI1 [15].

Evidence is accumulating that Cdc42 plays a critical role in cancer cell proliferation, survival, migration and invasion [16]. And the elevated expression and activity of Cdc42 are detected in multiple cancers, such as ovarian, breast, prostate and gastric cancers [16]. In fact, the expression level of Cdc42 in breast cancer lysates is discovered much higher than that in normal tissues in the same patients [17]. In addition, the hyperactivated Cdc42 is found to promote prostate cancer metastasis via Ack1 [18], and hepatocellular carcinoma (HCC) invasion via EMT [19]. It is interesting to find that the silence of Cdc42 can inhibit neuroblastoma cell proliferation and transformation [20]. The inhibition of Cdc42 by miR-133 can suppress gastric cancer cell proliferation and migration through Cdc42/PAKs signal pathway [21]. Moreover, small molecule inhibitors of Cdc42 can also block cancer cell invasion and metastasis. For example, we previously found that the ZCL278 can suppress the migration of prostate cancer cell [22] and the ZCL367 can reduce the proliferation of lung cancer cell [23]. Besides, a highly Cdc42-selective inhibitor, CID2950007 can inhibit ovarian cancer cell migration [24]. These findings suggest that Cdc42 plays a significant role in cancer and the inhibition of Cdc42 is beneficial to block the development of Cdc42-related cancers. Accordingly, Cdc42 is a potentially important therapeutic target for cancers.

RhoGDI1, as an endogenous inhibitor protein of Cdc42, is closely related to the occurrence and development of various human cancers. RhoGDI1 depletion results in constitutive activation of Cdc42 at molecular level [25]. RhoGDI1’s expression is significantly reduced in breast cancer specimens [25] and malignant gliomas [26]. And reduced RhoGDI1 expression is also associated with worsening clinical prognosis of HCC patients [27] as well as lower survival of bladder cancer patients [28]. These findings indicate that the expression of RhoGDI1 is negatively related to the activity of Cdc42. Thereby, it is critical to understand how RhoGDI1 prevents nucleotide exchange and maintains Cdc42 in a GDP-bound inactive state.

Researchers find the N-terminal helix-loop-helix motif of RhoGDI1 is necessary to inhibit nucleotide dissociation from Cdc42 through structural and biochemical analyses [29]. Further, the crystal structure of Cdc42 in complex with RhoGDI1 shows that the extensive direct contact between the helix-loop-helix motif of the RhoGDI1 and the switch regions of Cdc42 might lead to the blocking of the dissociation of GDP from Cdc42 through steric hindrance [30]. However, the inhibition process of Cdc42 by RhoGDI1 is dynamic, which makes it difficult to be captured through experiments. Thus, the detailed inhibition mechanism of Cdc42 by RhoGDI1 is still unclear.

In this study, to investigate the inhibition mechanism of Cdc42 by RhoGDI1, we performed explicit molecular dynamics (MD) simulations on GDP-bound Cdc42 with or without RhoGDI1. During the simulations, we discovered that switch regions showed large conformational changes without RhoGDI1, which is beneficial to open the nucleotide binding pocket and to promote GDP dissociation. Whereas, RhoGDI1 plays a crucial role in maintaining the intramolecular interactions of Cdc42, such as hydrogen bonds between Asp57 and Thr17, Phe37, Gln39, and hydrophobic interactions between Phe37 and Leu20, Ile21, Thr24, Leu55, to keep closed conformation of Cdc42. In addition, RhoGDI1 forms extensive interactions with Cdc42. The interactions between helix-loop-helix motif and switch regions are directly and primarily responsible for the closed conformation of Cdc42 to prevent GDP dissociation. The interactions between β-sandwich and Cdc42 (especially residue Arg66 and geranylgeranyl group) are essential for their stable binding. The electrostatic interactions between the N-terminus of RhoGDI1 and the C-terminal hypervariable region of Cdc42 contribute to their binding and specificity. This inhibition mechanism of Cdc42 by RhoGDI1 at atomic level provide essential information for discovering molecules that replace RhoGDI1 with a higher binding affinity to Cdc42 or enhance the binding of RhoGDI1 to Cdc42 to inhibit the activity of Cdc42.

Materials and methods

System preparation

The initial structures of Cdc42-GDP and RhoGDI1-Cdc42-GDP complexes were built based on the complex crystal structure of RhoGDI1-Cdc42-GDP including RhoGDI1 from Bos Taurus and Cdc42 from Homo Sapiens (PDB ID:1DOA). There are six different amino acids between RhoGDI1 from Bos Taurus and Homo Sapiens, including Arg74, Thr81, Met162, Asn176, Arg186 and Glu201 in Bos Taurus, corresponding to Gly74, Ser81, Val162, Ser176, Lys186, and Asp201 in Homo Sapiens. In addition, a non-standard Cys188 exists in the C-terminus of Cdc42 of the two complexes [31]. We first mutated the six different residues from Bos Taurus to Homo Sapiens and optimized them in Maestro [32]. Then, we prepared the two structures in Protein Preparation wizard to fill in the missing side chains and atoms, assign bond orders, release the possible steric clashes, and predict the protonation state of histidine residues using PROPKA [33] at pH 7.0. The final protonation states of histidine residues were identified based on the prediction results and their surrounding residues.

The above structures were further prepared with the tLeap module of the AMBER20 package [34]. The AMBER force field ff14SB [35] was used for proteins. The Generalized AMBER force field (GAFF) [36] was used for GDP and the non-standard Cys188. The parameters of GDP were taken from the AMBER parameter database (http://amber.manchester.ac.uk/), while those of non-standard Cys188 were generated in Antechamber. The non-standard Cys188 was first optimized and then its electrostatic potential was calculated using the Gaussian 09 package [37]. After the calculation of the restrained electrostatic potential charges of the non-standard Cys188, its parameters were generated. Both of the systems were solvated by TIP3P water molecules, with the size of the box to ensure a distance of at least 10 Å between the protein atoms and the box boundaries [38]. Eventually, 3 and 11 Na+ ions were respectively added to the Cdc42-GDP and RhoGDI1-Cdc42-GDP systems to neutralize them.

Molecular dynamics simulations

All simulations were performed using the AMBER20 package. Firstly, both of the solvated systems were subjected to two-round energy minimizations. In the first round, 5000 steps energy minimization was carried out with the protein atoms being restrained by 200 kcal/mol/Å2 force, with the first 1000 steps using the steepest descent algorithm and the last 4000 steps using the conjugate gradient algorithm. In the second round, another 5000 steps energy minimization was carried out with no restraint. Then, both of the systems were heated from 0 to 300 K using the Langevin [39] thermostat within 100 ps (each comprising 20 ps heating from 0 to 100 K; 20 ps heating from 100 to 200 K; 20 ps heating from 200 to 300 K; 40 ps heating at 300 K) and equilibrated to 1 atm within 100 ps with weak restraints (2.0 kcal/mol/Å2) on protein atoms. Subsequently, another 100 ps equilibration was performed at a constant temperature of 300 K and a constant pressure of 1 atm without any constraint. Finally, each system was submitted to triple-parallel 250 ns simulations under the constant pressure periodic boundary at a constant temperature of 300 K using the PMEMD module.

During the simulations, an integration step of 2 fs was used for the MD simulations. The Particle Mesh Ewald (PME) method [40] was employed to calculate the long-range electrostatic interactions within the systems. An 8 Å cutoff was employed for the non-bonded van der Waals interactions. SHAKE [41] algorithm was used to perform constraints on all covalent bonds involving hydrogen atoms to remove the bond stretching freedom. Coordinates were saved every 1000 steps.

Molecular dynamics data analysis

CPPTRAJ of AMBER20 was used to analyze the MD simulation results. The root-mean-square deviation (RMSD) and the root-mean-square fluctuation (RMSF) of Cdc42-GDP and RhoGDI1-Cdc42-GDP during the simulations were calculated to investigate the fluctuation of the two systems. The first snapshot was set as reference in the calculations.

The radius of gyration (Rg) of Switch I and Switch II in both of the two systems were calculated to elucidate their overall conformational change during the simulations. The solvent accessible surface area (SASA) of Switch I and Switch II of Cdc42 was calculated to study the accessibility of the switch regions. Hydrogen bond interactions with donor–acceptor distance less than 3.5 Å and donor-H-acceptor angle more than 120° were calculated.

The cluster analysis was performed with the average linkage hierarchical agglomerative (bottom-up) approach in CPPTRAJ module of AMBER20 based on the RMSD of main chain atoms of Cdc42 [42]. Every ten frames in both of the two systems were clustered. The representitive conformations for each cluster were extracted for further analysis.

The Principal Component Analysis (PCA) [43] characterizes the directions of the main motion mode of Cdc42 in the two systems. The average coordinate of Cdc42 during the simulations in each of the two systems was calculated as the reference structures in the PCA calculation. Then, the trajectories of the two systems were projected into the corresponding first three principal components based on the RMSD of Cdc42 in the two systems.

Moreover, dynamic cross-correlation matrix (DCCM) [44] analysis was calculated to reveal the correlated motions between Cdc42 and RhoGDI1 based on all the non-hydrogen atoms in RhoGDI1-Cdc42-GDP system.

Results and discussion

Structural features of the Cdc42/RhoGDI1 complex

As a member of the Rho GTPase family, the N-terminal domain of Cdc42 contains two conserved functional elements, Switch I (residues 27–42) and Switch II (residues 57–72), which are the vital binding regions for GDP or GTP and will undergo obvious conformational changes during nucleotide exchange [45] (Fig. 1a, b). In addition, the C-terminal geranylgeranyl group of Cdc42 which inserts into the C-terminal hydrophobic pocket of RhoGDI1 also contributes to strengthen the binding between Cdc42 and RhoGDI1 [30].

Fig. 1
figure 1

Structure of Cdc42-GDP in complex with RhoGDI1. a Schematic representation of human Cdc42. b Crystal structure of GDP (red stick)-bound Cdc42 (grey cartoon) with RhoGDI1 (yellow cartoon). Switch I, Switch II and the C-terminal hypervariable domain of Cdc42 are colored in orange, cyan, and pink, respectively. The secondary structures of Cdc42 and GDP are shown in the same color in the following figures

Full size image

As a negative regulator of Cdc42, the structural features of RhoGDI1 allow it to establish extensive interactions with Cdc42. As shown in Fig. 1b, the N-terminal helix-loop-helix motif of RhoGDI1 interacts with the switch regions of Cdc42 to assist the binding between Cdc42 and RhoGDI1 and inhibit GDP’s dissociation. The C-terminal domain of RhoGDI1, presenting as an antiparallel β-pleated sheet (β-sandwich), forms a hydrophobic pocket that can bind with the C-terminus of Cdc42.

Here, we performed explicit MD simulations on the Cdc42 with or without RhoGDI1 to explore the impacts of RhoGDI1 on the structural and dynamical behaviors of Cdc42, and further to investigate the inhibition mechanism of Cdc42 by RhoGDI1.

RhoGDI1 stabilizes the conformation of Cdc42

The Principal Component Analysis (PCA) and cluster analysis were carried out to characterize the major motions of Cdc42 in the two systems. The trajectories of Cdc42 in each of the two systems were projected into the corresponding three principal components. And these trajectories were clustered into four clusters using the average linkage hierarchical agglomerative approach (Fig. S1). In the PCA projection map, different clusters were shown in different colors. As shown in Fig. 2a, b, the distribution of Cdc42’s conformation in Cdc42-GDP system is broader than that in RhoGDI1-Cdc42-GDP system, especially in the direction of PC1, which implies that Cdc42 is more flexible and fluctuant in Cdc42-GDP system than in RhoGDI1-Cdc42-GDP system.

Fig. 2
figure 2

The projection of the three principal components in Cdc42-GDP (a) and RhoGDI1-Cdc42-GDP (b) systems. The conformations of Cdc42 in the triple-parallel simulations were clustered and showed in different colors: cluster 1, orange; cluster 2, green; cluster 3, red; cluster 4, blue. c The RMSF of Cdc42 in Cdc42-GDP (red) and RhoGDI1-Cdc42-GDP (blue) systems. d The RMSF of RhoGDI1 in RhoGDI1-Cdc42-GDP (blue) system. Averages were obtained from the triple-parallel simulations, and the error bars correspond to the standard deviation

Full size image

To further quantify the dynamic conformational changes throughout the simulations, the RMSD of the trajectories relative to the first frame of the two systems was calculated. As shown in Fig. S2, the RMSD of Cdc42-GDP (~ 4 Å) is much larger than that of RhoGDI1-Cdc42-GDP (~ 2 Å), indicating that the conformation of Cdc42 is much more fluctuant in Cdc42-GDP than in RhoGDI1-Cdc42-GDP, which is consistent with the PCA results.

To investigate the fluctuation of each residue, the RMSF of Cdc42 and RhoGDI1 in both of the two systems during the simulations was calculated (Fig. 2c, d). As shown in Fig. 2c, the RMSF values of the switch regions and the C-terminal hypervariable region of Cdc42 are much different in the two systems. The much lower RMSF values of these three regions in the RhoGDI1-Cdc42-GDP system indicate the much greater stability of these regions in Cdc42 when it bound with RhoGDI1, which is consistent with the B factor values of the crystal structures [30]. Additionally, in the RhoGDI1-Cdc42-GDP system, the N-terminus 25 residues of RhoGDI1 showed high flexibility (Fig. 2d), which is also consistent with the B factor values of the crystal structures [30].

In a word, without RhoGDI1, Cdc42 has more flexible conformations, especially in switch regions and the C-terminal hypervariable region. RhoGDI1 can stabilize the conformation of Cdc42.

RhoGDI1 maintains the switch regions in a closed conformation

Cluster analysis reveals significant conformational differences of Cdc42, particularly switch regions, in the two systems (Fig. S1). The superposition of the representative structures in the Cdc42-GDP system during the simulations shows that Switch I undergoes a large conformational change and gradually gets farther from GDP in the Cdc42-GDP system: cluster 1 (38.5%) and cluster 2 (29.5%) are in partially open states, cluster 3 (21.7%) is in a closed state, cluster 4 (10.3%) is in an entirely open state. Conversely, all of the four clusters of the RhoGDI1-Cdc42-GDP system maintain closed conformation (Fig. S1), which further supports that the presence of RhoGDI1 impedes the motion of the switch regions.

We used two parameter pairs to quantitate the different conformational changes of Cdc42, especially the switch regions (Fig. 3a). One is defined by the distance (d) of Thr35’s Cα-GDP’s Pβ, which reflects the conformational changes of the Switch I. Thr35 located in Switch I of Cdc42 plays essential roles for its effectors’ binding [46]. Chymotrypsin digestion experiments found that Switch I of Cdc42T35A showed slow proteolytic cleavage, indicating the mutation of Thr35 led to reduced conformational flexibility of Switch I [43]. Moreover, in vitro binding assays found that Cdc42T35A had a less binding affinity to GEFs than that of the wild type, which could affect Cdc42’s activation [47]. The other is defined by the dihedral (φ) of the Cα atoms of Thr25, Asp76, Phe82, and Arg66, which monitors the conformational rotations of the helix of Switch II. Residues Thr25, Asp76, and Phe82 are stable during the simulations, and thus the change of φ caused by Arg66 can reflect the flexibility of Switch II.

Fig. 3
figure 3

a Two parameters to describe the flexibility of Switch I and Switch II. The distance pair (d, Thr35’s Cα-GDP’s Pβ) was shown in red arrow. The Cα atoms of Thr25, Asp76, Phe82, and Arg66 defined for dihedral (φ) were shown in red point. b, c The probability distributions of the d and φ in the two systems during the triple-parallel simulations

Full size image

The probability distribution of d in the RhoGDI1-Cdc42-GDP system can be roughly divided into three adjacent intervals, 6.8–7.2 Å, 8.0–8.2 Å and 8.8–9.2 Å, suggesting a relatively stable and closed conformation of the Switch I. However, the d values exhibit a broader distribution and separate into four distant intervals, 8.0–8.5 Å, 10.2–10.5 Å, 18.3–18.7 Å and 22.5–23 Å, in the Cdc42-GDP system, indicating a much more flexible and open conformation of the Switch I (Fig. 3b). Meanwhile, the φ values reached the peak at ~ 53° in RhoGDI1-Cdc42-GDP, whereas at ~ 58° in Cdc42-GDP (Fig. 3c), suggesting Switch II showed more flexibility without RhoGDI1. Accordingly, both switch regions are more flexible in Cdc42-GDP than in RhoGDI1-Cdc42-GDP. The nucleotide binding pocket surrounded by switch regions varied from closed to open in Cdc42-GDP system, which are beneficial for the dissociation of GDP. In the presence of RhoGDI1, the nucleotide binding pocket maintained closed to inhibit the dissociation of GDP.

In addition, the radius of gyration (Rg) for the Switch I and Switch II in the two systems were calculated to elucidate the overall conformational change for switch regions. During the simulations, Rg maintains 12.5–13 Å in the RhoGDI1-Cdc42-GDP, whereas Rg increases from an average value of ∼12.5 Å to ∼14 Å in the Cdc42-GDP, which indicates that the switch regions undergo a gradually opening process without RhoGDI1 (Fig. 4a). Moreover, it is also identified that the interaction of RhoGDI1 with Cdc42 could result in a more compact and closed conformation of Cdc42 in switch regions.

Fig. 4
figure 4

a Rg for Switch I and Switch II. b, c SASA for Switch I (b) and Switch II (c)

Full size image

To investigate the solvent accessibility of Switch I and Switch II, their solvent accessible surface areas (SASAs) were calculated. As shown in Fig. 4b, c, the SASAs of Switch I and Switch II are around 1500 Å2 and 1400 Å2 in the Cdc42-GDP system, significantly higher than those in RhoGDI1-Cdc42-GDP system, which suggests that the interactions between RhoGDI1 and switch regions reduce their exposure. In short, RhoGDI1 confines switch regions within a closed conformation to limit the flexibility of Switch I, which exerts an important influence on maintaining the inactive conformation of Cdc42.

Agreed with the MD simulations, NMR spectroscopy showed a high level of disorder in Switch I and Switch II [48]. In fact, the structural flexibility of switch regions is critical for interactions of Cdc42 with GEFs, GAPs, or downstream effectors.

RhoGDI1 changes the intramolecular interactions of Cdc42

To explore the reason why the switch regions maintain a closed conformation in the RhoGDI1-Cdc42-GDP system at the molecular level, the intramolecular interactions of Cdc42 in the two systems were analyzed. As shown in Fig. 5a, c, in the RhoGDI1-Cdc42-GDP system, the Asp57 directly forms hydrogen bonds with Thr17, Phe37, and Gln39, while the Phe37 forms hydrophobic interactions with Leu20, Ile21, Thr24 and Leu55. However, these interactions disappeared in the Cdc42-GDP system (Fig. 5b, d), indicating that the switch regions are more flexible and the GDP is much easier to dissociate in the Cdc42-GDP system. Accordingly, these hydrogen bonds and hydrophobic interactions induced by RhoGDI1 are essential to stabilize the closed conformation of Cdc42 in the RhoGDI1-Cdc42-GDP system.

Fig. 5
figure 5

Intramolecular interactions between the switch regions. The hydrogen bond interactions of switch regions in RhoGDI1-Cdc42-GDP (a) and Cdc42-GDP (b) systems. The hydrophobic interactions of switch regions in RhoGDI1-Cdc42-GDP (c) and Cdc42-GDP (d) systems. Structures are taken from representative conformations of the largest cluster in the two systems. Hydrogen bonds are indicated by red dashed lines

Full size image

RhoGDI1 forms extensive interactions with Cdc42

To investigate the correlation and motion between RhoGDI1 and Cdc42, the dynamic cross-correlation matrix (DCCM) for each residue in the RhoGDI1-Cdc42-GDP system was calculated. The higher correlation indicates that there is more intensive interaction within the complex. As shown in Fig. 6, the areas i and ii marked by black dashed box represent the positive correlation between the helix-loop-helix motif of RhoGDI1 and Switch I, Switch II of Cdc42, which indicates that the helix-loop-helix motif of RhoGDI1 interacts dominantly with the switch regions of Cdc42.

Fig. 6
figure 6

DCCM map representing the covariance of residues between RhoGDI1 and Cdc42. The x-axis indicates the residue number of Cdc42 and the y-axis indicates the residue number of RhoGDI1. The cross-correlation values are depicted on a scale from blue to red

Full size image

By carefully analyzing the complex structure of RhoGDI1 and Cdc42, their interactions are mainly concentrated in three parts. Firstly, the N-terminal helix-loop-helix motif of RhoGDI1 forms hydrophobic and hydrogen bond interactions, with the switch regions of Cdc42 to keep them into a closed conformation to prevent the dissociation of GDP (Fig. 7a, b). Secondly, RhoGDI1 residues located in the β-sandwich form stable hydrogen bond or salt bridge interactions with Cdc42 to assist in their binding (Fig. 7c). And the hydrophobic interior of the β-sandwich can bind the geranylgeranyl group in the C-terminus of Cdc42 by forming extensive hydrophobic interactions (Fig. 7d). Finally, the negatively charged and flexible N-terminus 25 residues forms electrostatic interactions with the C-terminal hypervariable regions of Cdc42 (Fig. 7d).

Fig. 7
figure 7

The interactions between RhoGDI1 (yellow cartoon) and Cdc42 (grey cartoon). Switch I, Switch II and the C-terminal hypervariable domain of Cdc42 are colored in orange, cyan, and pink, respectively. Hydrogen bonds and salt bridges are indicated by red dashed lines

Full size image

Hydrophobic interactions dominante the binding between the helix-loop-helix of RhoGDI1 and the switch regions of Cdc42. As shown in Fig. 7a, b, residue Leu48 of RhoGDI1 forms hydrophobic interactions with Val36 of Switch I and Ala59 of Switch II, residues Ile35, Ile38, Leu56, and Ile122 of RhoGDI1 form hydrophobic interactions with residues Leu67 and Leu70 of Switch II. To investigate the stalibity of these hydrophobic interactions, the average distances of the center of mass (COM) of these hydrophobic residue sidechains during the last 100 ns in the RhoGDI1-Cdc42-GDP system were calculated. As shown in Fig. 8a–c, almost all the average distances are around 6 Å, indicating these hydrophobic interactions are very stable during the simulations. In addition, Asp45 and Ser47 of RhoGDI1 form hydrogen bond interactions with Thr35 of Swich I, and Gln61 as well as Tyr64 of Switch II. These findings are consistent with the experimental results that the RhoGDI1 cannot slow down the dissociation rate of GDP when the helix-loop-helix motif was removed manually in the RhoGDI1-Cdc42-GDP complex [29].

Fig. 8
figure 8

a and b The COM of residues’ sidechains is defined for the distance calculations. c The mean distance value and standard deviation of residue pairs d1-d9 during the last 100 ns from the triple-parallel simulations

Full size image

The β-sandwich of RhoGDI1 forms hydrogen bonds and salt bridges with Cdc42 (Fig. 7c). It is worth noting that the Arg66 of Switch II froms salt bridge with Asp185 and hydrogen bonds with Pro30 and Ile122 of RhoGDI1. Meanwhile, His103 of Cdc42 forms a hydrogen bond with Asp184. All the fractions of these hydrogen bonds are more than 70%, indicating that these hydrogen bonds are persistently stable in the simulation process (Fig. 9). In vitro binding assays found that overexpression of the Cdc42R66A mutation is incapable of binding with RhoGDI1 [49]. Thereby, Arg66 performs a critical role in connecting the switch regions of Cdc42 with RhoGDI1.

Fig. 9
figure 9

a Six hydrogen bonds or salt bridges are defined for calculations. b Fractions of each hydrogen bond or salt bridge during the last 100 ns of the triple-parallel simulations

Full size image

The hydrophobic interior of the β-sandwich of RhoGDI1 can bind the geranylgeranyl group in the C-terminus of Cdc42 by forming hydrophobic interactions. These hydrophobic interactions play key roles in helping RhoGDI1 to extract Cdc42 from membrane to cytosol, which effectively shield it from the aqueous milieu and keep it in a stable inactive state in the cytosol. However, these interactions have little effect on the inhibition of GDP dissociation [29, 50]. In fact, the GDIΔ59 which contained almost exclusively the β-sandwich still can bind with GTPase in the high nanomolar range [29]. Thus, the interactions between RhoGDI1’s β-sandwich and Cdc42 mainly contribute to their stable binding.

In addition, the N-terminal residues Glu17, Glu19, and Glu20 of RhoGDI1 can form electrostatic interactions with the C-terminal hypervariable region of Cdc42 including residues Lys184, Arg186 and Arg187 (Fig. 7d). And we calculated the electrostatic energies between these negative residues of RhoGDI1 and the Cdc42. As shown in Fig. 10, residues Glu17 and Glu19 contribute considerable electrostatic energies, up to 48 kcal/mol and 38 kcal/mol. In fact, the electrostatic interaction between the C-terminal hypervariable region of Rho GTPases and the negatively charged N-terminus of RhoGDIs mediated the specificity of Rho GTPase-GDI interactions [51]. Furthermore, the negatively charged and flexible N terminus of RhoGDI1 was a pivotal regulator in the cytoplasm/membrane cycles of the RhoGDI-Rac complex [52]. Therefore, the electrostatic interaction between the N-termins 25 residues of RhoGDI1 and the C-terminal hypervariable region of Cdc42 are also important for their complex formation.

Fig. 10
figure 10

The electrostatic energy between Glu17, Glu19 and Glu20 of RhoGDI1 and the C-terminal hypervariable regions of Cdc42 (residues 180–189)

Full size image

Proposed inhibition mechanism of Cdc42 by RhoGDI1

Previously, we have revealed the dissociation mechanism of GDP from Cdc42 via DOCK9. In the Cdc42-DOCK9-GDP system, the flexible switch regions underwent a conformational change to interact with DOCK9, which changes the orientations of Lys16, Thr17, Cys18 and Phe28 of Cdc42 to weaken the interactions between Cdc42 and GDP and further promote the release of GDP [53]. Conversely, RhoGDI1 maintains Cdc42 into a closed nucleotide binding pocket and prevents Cdc42’s activation. Focusing on the effect of RhoGDI1 on Cdc42, we carried out MD simulations and dynamics analyses on the Cdc42-GDP and RhoGDI1-Cdc42-GDP systems, and further supposed the inhibition mechanism of Cdc42 by RhoGDI1 as shown in Fig. 11.

Fig. 11
figure 11

Schematic of the proposed inhibition mechanism of RhoGDI1 to Cdc42

Full size image

Without RhoGDI1, the switch regions of Cdc42 are flexible, especially Switch I. The intramolecular interactions between the switch regions, such as the hydrogen bonds between Asp57 and Thr17, Phe37, Gln39, and the hydrophobic interactions between Phe37 and Leu20, Ile21, Thr24, Leu55, are disappeared. In addition, the conformations with semi-open and open states which favor the dissociation of GDP and subsequent the activation of Cdc42 are dominant in the Cdc42-GDP system during the simulations.

In the presence of RhoGDI1, it maintains the intramolecular interactions around nucleotide binding pocket to keep it into closed conformation to further prevent GDP dissociation. RhoGDI1 forms extensive interactions with Cdc42. The hydrophobic interactions between helix-loop-helix motif of RhoGDI1 (residues Ile35, Ile38, Leu48, Leu56, and Ile122) and the switch regions (residues Val36, Ala59, Leu67, and Leu70) of Cdc42 play direct and primary roles in maintaining the closed conformation of Cdc42. The hydrophobic interactions between the β-sandwich of RhoGDI1 and the geranylgeranyl group in the C-terminus of Cdc42 play important roles in extracting Cdc42 from membrane to cytosol. And the polar interactions between β-sandwich (residues Pro30, Ile122, Asp184, and Asp185) and Cdc42 (Arg66 and His103) stabilize the binding of RhoGDI1-Cdc42. In addition, the electrostatic interactions between the N-terminus 25 residues of RhoGDI1 and the C-terminal hypervariable region of Cdc42 also assist in their binding and specificity.

In short, RhoGDI1 can inhibit the activation of Cdc42 by extracting it from membrane to cytosol and maintaining it in closed conformation through extensive interactions.

Conclusion

The hyperactivation of Cdc42 is involved in the activation of multiple signaling pathways and leads to the development of various pathological disorders in humans, especially cancer’s progress and metastasis. RhoGDI1, which can prevent GDP’s dissociation and inhibit Cdc42’s activation, is a crucial negative regulator of Cdc42. Therefore, an in-depth understanding of the inhibition mechanism of Cdc42 by RhoGDI1 is critical for further drug development targeting Cdc42-related cancers. We combined MD simulations and a series of dynamics analyses on Cdc42-GDP and RhoGDI1-Cdc42-GDP systems to investigate this inhibition mechanism. The results suggest that switch regions, especially Switch I, exhibits significant conformational flexibility without RhoGDI1, which facilitates the opening of the nucleotide binding pocket and the dissociation of GDP. However, RhoGDI1 stabilizes the intramolecular interactions of Cdc42 around the nucleotide binding pocket and keeps the switch regions closed through extensive interactions. Among them, the interactions between the helix-loop-helix motif of RhoGDI1 and switch regions of Cdc42 play direct and primary roles in maintaining the closed conformation of Cdc42 to inhibit GDP dissociation. The interactions between the β-sandwich of RhoGDI1 and Arg66 as well as geranylgeranyl group of Cdc42 are crucial for stabilizing their binding. The electrostatic interactions between the N-terminus 25 residues of RhoGDI1 and the C-terminal hypervariable region of Cdc42 also assist in their binding affinity and specificity. These results are in good agreement with existing experimental data and provide a comprehensive understanding on how RhoGDI1 inhibits Cdc42’s activation. These findings provide insights for the discovery of stabilizers of Cdc42-RhoGDI1 complex or molecules replacing RhoGDI1 with a higher binding affinity to Cdc42 to inhibit Cdc42’s activation, which are beneficial for the discovery of novel drugs targeting on Cdc42-related diseases.