The DH and PH Domains of Trio Coordinately Engage Rho GTPases for their Efficient Activation

Structure of Trio in complex with nucleotide-depleted Rac1

A complex between the N-terminal DH/PH cassette of Trio (residues 1226 – 1536) and residues 1-189 (C189S) of Rac1 yielded poorly diffracting crystals. Since several structures of Rho GTPases indicate that the C-terminal polybasic tail is not typically well-structured, nor directly involved in binding GEFs,^{13; 19; 28; 42} a truncated version of Rac1 lacking this region and ending at residue 177 (Rac177) was crystallized with the N-terminal DH/PH cassette of Trio and used for structure determination of the Trio•Rac1 complex. Previous work from this lab has shown that full-length, unprenylated Rac1 and Rac177 exhibited identical capacity to bind guanine nucleotides and to be activated by the DH/PH cassette of Tiam1 in vitro.¹² Phases were determined for a native data set collected at the SER-CAT beamline (ID-22, Advanced Photon Source) using Trio DH/PH⁴³ and Rac1¹² as a model for molecular replacement. The final structure has an R_work = 22.3% and an R_free = 24.9% and incorporates data from 19.4Å– 2.0 Å (see Table 1 for data collection and refinement statistics).

The overall domain architecture of the Trio•Rac1 structure is similar to that seen in other complexes of Dbl-family GEFs and their GTPases; the DH domain of Trio consists of a bundle of six α-helices while the PH domain consists of a core seven strand β-sandwich with three inter-strand loops and a capping C-terminal helix^{10; 13; 28; 42; 44-46} (Figure 2(a) and 3). The majority of the interface between Rac1 and Trio, which buries ∼2500 Ų of surface area, is mediated by the DH domain. However, the complex also shows significant interactions between Rac1 and the PH domain of Trio that occur predominantly through residues in the β3/β4 loop and His1410 (Figure 2(a) and (b)), similar to the interactions seen between Dbs and its cognate GTPases.^{13; 28}

There is excellent electron density for Rac1, the DH domain, and the parts of the PH domain that contact Rac1 (Figure 2(c) and Table 1). In contrast, electron density is poor for the remainder of the PH domain, especially the β1/β2 and β6/β7 loops, which have not been modeled. The unusually high average B-factor for the PH domain (Table 1) suggests that it is inherently mobile. To illustrate the predicted motion of the PH domain, we have displayed the thermal ellipses for each atom of the Trio•Rac1 structure (Figure 2(d)). These ellipses represent the anisotropic B-values obtained after TLS refinement^{47; 48} and indicate that a portion of the PH domain is highly mobile. This observation is consistent with several reports that indicate many DH-associated PH domains have significant conformational mobility and are often difficult to model when crystallized.^{28; 42; 44} One notable exception to this generalization is the structure of Trio crystallized without bound GTPase.⁴³ However, in this case, the PH domain is locked into place by several crystal contacts.

Interestingly, the anisotropic B-values and the corresponding thermal ellipses associated with residues of the PH domain that mediate the interface with Rac1, including those in the β3/β4 loop, are comparable to those for atoms in the DH domain and Rac1 (Figure 2(d) and data not shown). While the majority of the PH domain has few stabilizing interactions, the interaction of the β3/β4 loop with Rac1 appears to restrict its motion. Furthermore, a simulated annealing omit map generated from the final model after omitting residues in the interface between the PH domain and Rac1 and the final 2F_o-F_c map both show clear density for residues mediating this interface (Figure 2(c)). Thus, the interactions between the DH-associated PH domain of Trio and Rac1 illustrated in Figure 2(b) are not modeling artifacts

Comparison of Trio•Rac1 to Dbs•Cdc42

Superimposition of the structures of Trio•Rac1 with the corresponding DH/PH portion of Trio in isolation indicates that an ∼10° rotation of the PH domain with respect to the DH domain (toward Rac1) occurs upon complex formation. This shift mimics a similar rearrangement within the DH/PH portion of Dbs upon engagement of Cdc42 or RhoA^{13; 28} (Figure 4(a) and (b)). Consistent with sequence conservation between Dbs and Trio, the molecular interactions between the PH domain of Trio and Rac1 are analogous to those that mediate the interface between the PH domain of Dbs and Cdc42 (Figure 4(c) and (d)). Specifically, the interactions between Asp65 (Rac1), His1410 (Trio, DH domain) and Tyr1472 (Trio, PH domain) mimic the interactions between Asp65 (Cdc42), His814 (Dbs, DH domain), and Tyr889 (Dbs, PH domain). In addition, interactions between Asp65 (Rac1), Gln1430 (Trio, PH domain), and a bridging water molecule are recapitulated in the Dbs•Cdc42 complex. Other similar interactions also occur within both structures, i.e., Ser1470 within the PH domain of Trio interacts with His103 of Rac1 while Lys885 of the PH domain of Dbs interacts with His103 of Cdc42.

Functional analysis of the interface between the PH domain of Trio and Rac1

To test the functional significance of the interactions between Rac1 and the PH domain of Trio, mutations designed to disrupt these interactions were introduced into the DH/PH fragment used for crystallization (Figure 5) and purified mutant proteins were assayed for exchange activity on soluble wild-type Rac1 (residues 1-189, c189s) (Figure 5(a)). Relative to the equivalent wild-type DH/PH fragment of Trio, mutation of either His1410 to Ala, Gln1430 to Ala, or Tyr1472 to Phe significantly decreases exchange activity. Similar to Dbs,²⁸ mutation of Tyr1472 to Phe produces the largest reduction in exchange activity. Substitution of Tyr 1472 to Phe was previously assessed under similar conditions⁴³ and the two sets of measurements are consistent. Unsurprisingly, complementary mutations within Rac1 designed to disrupt interaction with the PH domain of Trio reduced exchange by the wild-type fragment of Trio (Figure 5(c)), further confirming the functional importance of engaging Rac1 through the PH domain for catalyzed exchange. The D65A and R66A mutations introduced in Rac1 lie within the switch region responsible for binding nucleotide. However, rates of spontaneous exchange for these mutant forms of Rac1 were no more than 2-fold higher than wild-type Rac1 (data not shown) and these small differences cannot account for the relatively large decrement in the capacity of Trio to activate them. The spontaneous rate of exchange for Rac1 H103A was essentially identical to wild-type Rac1.

In this study, Trio was crystallized in complex with nucleotide-depleted Rac1, but Trio alsoexchanges on RhoG both in vitro^{29; 43} and in vivo,^{29; 31; 40; 41} and does so approximately three times more efficiently in vitro.⁴³ RhoG is 80% homologous to Rac1, 65% identical, and the residues in nucleotide-depleted Rac1 that are buried in the interface with Trio are 100% identical in RhoG (data not shown). Consequently, mutants of Trio that have reduced exchange activity on Rac1 were also tested for activity on soluble, wild-type RhoG (Figure 5(b)). Predictably, mutations within the PH domain of Trio also reduced its exchange activity on RhoG, with Trio Y1472F exchanging the least efficiently. These data strongly suggest that the interface between Trio and RhoG is highly similar, if not identical, to its interface with Rac1. Circular dichroism spectroscopy (Figure 5(d)) and gel filtration chromatography (data not shown) confirmed that all mutant forms of Trio were properly folded and monodisperse.

Wild-type Trio robustly activates RhoG through its N-terminal DH/PH cassette to promote neurite outgrowth in PC-12 cells.³¹ Furthermore, mutant forms of Trio lacking the N-terminal PH domain exhibit a reduced capacity to both activate RhoG in vitro and induce neurite formation in PC-12 cells.^{29; 31} While these studies indicate the importance of the N-terminal PH domain of Trio for guanine nucleotide exchange of its cognate GTPases and its attendant morphological consequences, they do not address the underlying mechanistic causes for these results. Therefore, to test the effects of disrupting specific interactions between the N-terminal PH domain of Trio and its cognate GTPases, the single substitutions described above were introduced into full-length Trio (3038 amino acids) and the capacity of these substituted forms of Trio to induce neurite outgrowth of PC-12 cells was quantified (Figure 6). Consistent with the inability of these mutant forms of Trio to activate Rac1 and RhoG efficiently in vitro, these singly-substituted forms of full-length Trio are also significantly compromised in their capacity to induce neurite outgrowth in PC-12 cells. Since these substitutions do not perturb the overall fold of the isolated DH/PH fragment of Trio, it is highly unlikely that they do so within the context of full-length Trio.

Sequence alignment

The sequences of Dbs (NP 079255), Trio (NP 009049), Duo (NP 003938), Dbl (NP 005360), Duet (NP 008995), PRex1 (NP 065871), p63RhoGEF (NP 891992), Lfc (NP 004714), neuroblastoma (NP 005263), and obscurin (NP 443075) shown in Figure 1(b) were aligned using Clustal X.⁵⁵ The NCBI accession numbers are given in parentheses.

Protein preparation for guanine nucleotide exchange assays

Human Trio DH/PH (residues 1226 – 1535) (kindly provided by Dr. Yi Zheng, Cincinnati Children's Hospital Medical Center), was encoded as a fusion with an N-terminal His₆-tag in pET15a (Novagen), expressed in the BL21 (DE3) E. coli strain, and purified similarly to published protocols.^{28; 56} Briefly, transformed cells were grown in LB media containing 0.1 mg/ml ampicillin at 37°C to an OD₆₀₀ of 0.7 (mid-log phase) and induced with 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) at 27°C for 5 hours. Harvested cells were resuspended in 50 mM NaH₂PO₄, pH 8, 300 mM NaCl (buffer A), and 5mM imidazole and lysed with an Emulsi-Flex C5 (Avestin). Lysate was clarified by centrifugation at > 125,000 × g for 30 minutes prior to loading the supernatant onto a Ni²⁺-Sepharose affinity column (GE Healthcare) equilibrated in buffer A containing 5 mM imidazole. The column was washed with buffer A containing 55 mM imidazole and Trio eluted with buffer A containing 400 mM imidazole. The protein was further purified using a 26/60 Sephacryl-200 size exclusion column (GE Healthcare) equilibrated with 50 mM Tris, pH 8, 2 mM DTT, 2mM EDTA, 150 mM NaCl, and 5% glycerol. Fractions containing Trio were pooled, concentrated, and stored at −80°C.

Human Rac1 (residues 1-189, C189S) was expressed from pET21a (Novagen) in the BL21 (DE3) E. coli strain. Transformed cells were grown and induced similarly to cells expressing Trio. Harvested cells were resuspended in 10 mM MES, pH 6, 2 mM DTT, 10% glycerol, 1 mM MgCl₂ (buffer B) and 10 mM NaCl prior to lysis and clarification as described above. Supernatant was loaded onto an SP-Sepharose Fast Flow 26/10 column (GE Healthcare). Rac1 was eluted from the column using buffer B with an increasing gradient of NaCl and further purified using a 26/60 Sephacryl-200 size exclusion column (GE Healthcare) equilibrated with 50 mM Tris, pH 8, 2 mM DTT, 2mM MgCl₂, 150 mM NaCl, and 5% glycerol. Fractions containing Rac1 were pooled, concentrated, and stored at −80°C.

Human RhoG (residues 1-188, C188S) was expressed from the pGEX4TEV2 vector⁵⁷ with an N-terminal GST-tag in the BL21 (DE3) E. coli strain. Transformed cells were grown in LB media containing 0.1 mg/ml ampicillin at 37°C to an OD₆₀₀ of 0.7 (mid-log phase) and induced with 1 mM IPTG at 20°C for 16 hours to 18 hours. Cells were resuspended in buffer C (150 mM NaCl, 20 mM Tris, pH 8, 2 mM DTT, 1 mM MgCl₂, 10 μM GDP, and 5% glycerol) prior to lysis and clarification as described above. Supernatant was loaded onto a GST-Sepharose affinity column (GE Healthcare) equilibrated in buffer C; the column was washed with buffer C, and RhoG eluted with buffer C containing 10 mM glutathione (reduced). The GST tag was removed by cutting with TEV while dialyzing against buffer C overnight. The protein was further dialyzed against buffer D (10 mM MES, pH 6, 2 mM DTT, 2 mM MgCl₂, 10 μM GDP, and 5% glycerol) and applied to a Source-S 16/10 column (GE Healthcare) equilibrated in buffer D. RhoG was eluted with buffer D containing an increasing concentration of NaCl. The fractions containing RhoG were then loaded onto a GST-Sepharose affinity column equilibrated in buffer C to remove residual amounts of GST-tagged protein. The flow-through from the GST column was buffer exchanged into buffer C, concentrated, and stored at −80°C.

Mutations in Trio DH/PH and Rac1 were made using the Quikchange site directed mutagenesis kit (Stratagene) following manufacturer's instructions. Sequences were verified using automated sequencing. Mutants of Trio and Rac1 were purified as described above for wild-type Trio and Rac1, respectively.

Guanine nucleotide exchange assays

Nucleotide exchange was measured using a fluorescence based assay, similar to published protocols,^{28; 58} in which N-methylanthraniloyl (mant)-GTP was loaded onto the GTPase. Spectroscopic analysis was carried out using a Perkin-Elmer LS 55 spectrometer at 20°C. The exchange assay mixture containing 20 mM Tris pH 7.5, 50 mM NaCl, 10 mM MgCl₂, 1 mM DTT, 100 uM mant-GTP, and 2 uM GTPase was allowed to equilibrate with constant stirring. Trio was then added at 50 or 400 nM for exchange assays with RhoG or Rac1, respectively, and nucleotide exchange was measured by monitoring the decrease in intrinsic tryptophan fluorescence (λ_ex=295 nm, λ_em=335 nm) of the GTPase due to the binding of mant-GTP. The data were fit to one phase exponential decay curves using the program GraphPad Prism™ in order to determine the rate of nucleotide exchange.

Protein preparation for formation of Trio/Rac1 complex

Human Trio DH/PH (residues 1226 – 1536) was cloned in a pPROEX-HTa vector (Invitrogen) using NcoI and XhoI cleavage sites, expressed as a fusion protein with an N-terminal His₆-tag in BL21 (DE3)s E. coli strain, and eluted from a Ni²⁺-Sepharose affinity column (GE Healthcare) as described above. The His₆-tag was removed by cleavage with TEV while dialyzing (16 hrs) against a buffer containing 50 mM Tris, pH 8, 2 mM DTT, 2mM EDTA, 150 mM NaCl, and 5% glycerol. The removal of the His₆-tag was confirmed using SDS-PAGE. The protein was then stored at 4°C until it was used to form a complex with Rac1.

Rac177 was expressed from the pET15b vector (Novagen) in BL21 (DE3)s E .coli strain and protein was purified according to published protocols.¹² Minor variations include that the 45% ammonium sulfate pellet was resuspended in 20 mM Tris, pH 8, 2 mM DTT, 1 mM MgCl₂, and 10 μM GDP and dialyzed extensively against buffer E (20 mM Tris pH 8, 10 mM NaCl, 2 mM DTT, 1 mM MgCl₂, and 10 μM GDP) to remove the ammonium sulfate and increase the pH. The resuspended pellet was loaded onto a Q-Sepharose Fast Flow 26/10 column (GE Healthcare) instead of a size-exclusion column. Rac177 was eluted from the column using buffer E with increasing amounts NaCl. The fractions containing Rac177 were pooled together and stored at 4°C before being used to form a complex with Trio.

Crystallization of the Trio/Rac177 complex

The Trio/Rac177 complex was formed in the presence of an excess of nucleotide-free Rac177 in 20 mM Tris, pH 8, 2 mM DTT, 4mM EDTA, 200 mM NaCl, and 5% glycerol. The complex was purified on a 26/60 Sephacryl-200 size exclusion column equilibrated with 20 mM Tris, pH 8, 2 mM DTT, 4 mM EDTA, 200 mM NaCl, and 5% glycerol. Fractions containing purified complex were pooled together, dialyzed into a buffer containing 50 mM NaCl, 10 mM Tris, 2 mM EDTA, and 2 mM DTT, concentrated to ∼21 mg/ml, and stored at −80°C.

Trio/Rac177 crystals were obtained by vapor diffusion at 18 °C. Drops were formed by combining equal volumes of protein complex and reservoir solution (100 mM sodium cacodylate pH 5.5 – 6.5, 14 – 18% (w/v) PEG 8000 (FLUKA), and 300-500 mM calcium acetate). Crystals typically appeared after 2 – 3 days and grew to final dimensions of 0.1 × 0.1 × 0.05 mm after several days. Crystals were cryoprotected by increasing the glycerol concentration of the drop to 21% (v/v) in 3% increments. Cryoprotected crystals were then suspended in a rayon loop (Hampton Research) and snap frozen in liquid nitrogen. The crystals belong to the space group P2₁2₁2 with unit cell parameters a = 97.492 Å, b = 108.558 Å, and c = 53.416 Å.

Data collection and structure determination

A native data set was collected using a single frozen crystal at the SER-CAT beamline (ID-22, Advanced Photon Source). Data were integrated and scaled using DENZO and SCALEPACK⁵⁹. Phases were calculated by molecular replacement using the Trio DH/PH⁴³ and Rac1¹² structures. The calculations were performed using both AMORE⁶⁰ and PHASER⁶¹ from the CCP4 suite of programs,⁶² which yielded identical solutions.

Model building and structure refinement

Electron density maps were calculated using CNS.⁶³ The interactive graphics program O was used for the majority of the model building,⁶⁴ but the molecular graphics program COOT⁶⁵ was also used to add water molecules. Most of the subsequent refinement was performed using CNS;⁶³ however, due the conformational mobility of the PH domain, CNS refinement was unable to produce clear density for this region of the molecule. Thus, TLS refinement from the CCP4 program Refmac5^{47; 48} was used to improve the quality of the electron density maps. Using Rac1, the DH domain, the PH domain, and the waters as TLS “groups” and employing tight geometrical restraints (matrix diagonal weighing term = 0.06), a final model with R_work = 22.3% and R_free = 24.9% was produced. In addition, the torsion angles for every non-glycine residue are within the most favored or allowed regions of the Ramachandran diagram, as determined by PROCHECK.⁶⁶ Please refer to Table 1 for refinement statistics.

The simulated annealing omit map used to validate the interactions seen at the interface between Rac1 and the PH domain of Trio was generated from the final coordinates after omitting residues involved in this interface (residues 64 – 68, and 102 – 104 from Rac1 and residues 1405 – 1411, 1429 – 1431, and 1469 – 1473 from Trio).

All images of protein structures, with the exception of Figure 2(d), were generated using Pymol.⁶⁷ The images in Figure 2(d) displaying the theoretical anisotropic motion of the atoms⁶⁸ were generated using CCP4MG.^{69; 70} Pymol⁶⁷ was also used to calculate buried surface areas.

Protein Data Bank accession codes

The atomic coordinates and structure factors (code 2NZ8) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ.

Cell culture and transfection

To introduce point mutations (H1410A, Q1430A, and Y1472F) into EGFP-Trio FL wild-type (human) (generously donated by Dr. Anne Debant, Centre de Recherche en Biochimie Macromoléculaire, Montpellier, France), a section of Trio containing the N-terminal DH/PH cassette (Trio DH/PH-long) was subcloned into the pMCSG7 vector.⁷¹ Mutations were introduced using the Quikchange site directed mutagenesis kit (Stratagene) following manufacturer's instructions and verified by automated sequencing. The DH/PH cassette of Trio was then digested from Trio DH/PH-long and subcloned into EGFP-Trio FL.

PC-12 cells were cultured in DMEM containing 10% horse serum (HS), 5% fetal bovine serum (FBS), and 1x penicillin/streptomycin (100 units/ml penicillin; 0.1 mg/ml streptomycin) at 37°C in the presence of 5% CO₂. For transfection, cells were seeded at a density of 500,000 cells per well on glass coverslips coated with rat-tail collagen type I (BD biosciences) in 6-well dishes. Cells were cultured for 18 hrs and then transfected with 1 μg of EGFP-C2 vector (Clonetech), EGFP-RhoG wild-type (human) (generously donated by Dr. Keith Burridge, UNC-Chapel Hill), or EGFP-Trio FL (wild-type and mutant forms) per well using Lipofectamine plus (Invitrogen) according to the manufacturer's instructions. Cells were fixed 48 to 72 hours post transfection for 10 minutes with 3.7% (v/v) paraformaldehyde in PBS, permeabilized with 0.3% Triton X-100, stained for actin with Alexa Fluor 546 phalloidan (Invitrogen), washed, and mounted using Fluorsave Reagent (Calbiochem) according the manufacturer's instructions.

Image collection and data processing for neurite outgrowth assays

Fixed cells were observed using an Olympus Fluoview 300 laser scanning confocal microscope with a 60X PL APO oil immersion objective when scoring for neurite positive cells. Neurite outgrowth activity was determined by positively scoring transfected cells displaying one or more neurites greater than one cell body in length. Cells found in large clumps (> 10 cells) were excluded from analysis. Displayed results represent the average of 3 independent experiments in which at least 100 cells per condition were counted for each experiment. In addition, the studies were blinded to conceal which slides contained cells that had been transfected with wild-type and mutant versions of Trio.

Images were obtained using an Olympus Fluoview 1000 laser scanning confocal microscope with 60× PL APO oil immersion objective in 20 – 30 micron XYZ stacks. Final images were processed in ImageJ⁷² and represent a standard deviation projection of the compiled stacks.