Cryo-EM Structure of the Human FLCN-FNIP2-Rag-Ragulator Complex.

LEAD CONTACT AND MATERIALS AVAILABILITY

The reagents generated in this study are available with no restriction. Further information and requests for resources and reagents should be directed to the Lead Contact, David M. Sabatini ( sabatini@wi.mit.edu ).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

HEK-293T cells were obtained from American Type Culture Collection (ATCC) and were maintained in an incubator setting at 37°C and 5% CO 2 . They were cultured in Dulbecco’s Modified Eagle Medium (DMEM), supplemented with 10% IFS, 2 mM glutamine, 100 IU/ml penicillin, and 100 μg/ml streptomycin. HEK-293 FreeStyle cells were obtained from Thermal Fisher Scientific and were maintained in a Multitron shaker set at 37°C, 125 r pm, 8% CO 2 , and 80% humidity. They were cultured in FreeStyle 293 Expression Medium, supplemented with 100 IU/ml penicillin and 100 μg/ml streptomycin. BL21(DE3) E. Coli strain was grown at 37°C in a Multitron shaker in LB media. To induce protein expression, BL21(DE3) E. Coli strain was transformed by the corresponding plasmids (see below), propagated at 37°C, and induc ed with IPTG at 18°C. METHOD DETAILS Protein Preparation The Rag GTPase heterodimer was purified as previously described ( Shen et al., 2017 ). In brief, co-expression of C-terminally His-tagged RagA(T21N) and tagless RagC was induced by 0.5 mM IPTG in BL21(DE3) cells. The heterodimer was sequentially purified through Q Sepharose, Ni-NTA, MonoQ, and Superdex 200 columns. Ragulator was purified as previously described ( Shen and Sabatini, 2018 ). In brief, the five subunits of Ragulator were co-transformed and expressed in BL21(DE3) strain. The pentameric complex was sequentially purified through GST, Ni-NTA, MonoQ, and Superdex 200 columns. Human FLCN-FNIP2 was expressed and purified as previously described ( Shen et al., 2017 ; Tsun et al., 2013 ). In brief, HA-tagged FLCN were co-expressed with Flag-tagged FNIP2 in FreeStyle 293 cells. The complex was isolated using Flag-M2 beads and further purified by a Superdex 200 gel-filtration column. The FLCN-FNIP2-Rag-Ragulator complex was assembled as follows. 1 mg of purified FLCN-FNIP2 was incubated with 1 mg of RagA(T21N)-RagC and 2 mg of Ragulator in a total volume of 1 ml for ten hours at 4°C. The assembly solution contains 50 mM NaHEPES (pH 7.4), 100 mM NaCl, 2 mM MgCl 2 , 2 mM DTT, 0.1% CHAPS, 100 μM GppNHp, and 100 μM GDP. The nanomeric complex was separate from excess Rag GTPases and Ragulator on a Superdex 200 column. The FLCN-FNIP2-Rag-Ragulator complex was concentrated to 8 μg/μl in a 100kDa molecular weight cut-off concentrator and ultracentrifuged at 100,000× g for 30 minutes immediately prior to freezing grids.

Cryo grids preparation

Cryo grids were prepared immediately after protein purification. 400-mesh Quantifoil 1.2/1.3 Cu grid (Quantifoil, Großlöbichau Germany) was made hydrophilic by glow discharging for 60 seconds with a current of 15 mA in a Pelico EasiGlow system. Cryo grids were prepared using an FEI Mark IV Vitrobot (FEI, part of Thermo Fisher Scientific, Hillsboro, OR). The chamber of the Vitrobot was kept at 4°C and 100% relative humidity. The blotting time was 3 seconds with an equipment-specific blotting force set at 3. 3 μl of sample was applied to the glow-discharged grid and then rapidly plunge-frozen into a liquid ethane bath.

Image collection and processing

Two data sets were collected on two different 300 kV FEI Titan Krios cryo electron microscopes (FEI) at HHMI Janelia Research Campus. The first data set was collected on Janelia Krios2 with spherical aberration Cs of 2.7 mm and equipped with a Gatan K2 Summit camera. The final exposure was collected in dose fractionation mode on the K2 camera at a calibrated magnification of 38,168, corresponding to 1.31 Å per physical pixel in the image (0.655 Å per super-resolution pixel). The dose rate on the specimen was set to be 5.83 electron per Å 2 per second and total exposure time was 10 s, resulting in a total dose of 58.3 electrons per Å 2 . With dose fractionation set at 0.25 s per frame, each movie series contained 40 frames and each frame received a dose of 1.46 electrons per Å 2 . Fully automated data collection was carried out using SerialEM with a nominal defocus range set from −1.5 to −3 m. A total of 4535 dose fractionation movies were collected in this session. One round of data processing was done on this dataset. Beam-induced motion were measured, corrected, and dose-weighted at 1.46 electron/Å2 per frame with data binned by 2 using cisTEM ( Grant et al., 2018 ). CTF determination for each movie series was calculated by amplitude averaging of every 3 frames using cisTEM. Automated particle picking using ab inito mode was carried out in cisTEM on all the micrographs and 927042 particles were extracted. Two rounds of reference-free 2D classification with CTF correction was performed in cisTEM to throw away bad particles. 321015 particles were kept for further processing. Ab initio 3D initial model was generated using CryoSparc ( Punjani et al., 2017 ). Subsequent 3D refinement was performed in cisTEM to generate a 3D reconstruction density map. Fourier Shell Correlation at a criteria of 0.143 reported resolution of 6.34 Å for this map. To improve the resolution of the reconstruction, a larger data set was collected at higher magnification on Janelia Krios1. This Krios microscope is equipped with a spherical aberration corrector (Cs=0.01 mm), an energy filter (Gatan GIF Quantum) and a post-GIF Gatan K2 Summit direct electron detector. The final exposure was collected in dose fractionation mode on the K2 camera at a calibrated magnification of 48,077, corresponding to 1.04 Å per physical pixel (0.52 Å per super-resolution pixel). The dose rate on the specimen was set to be 9.25 electron per Å 2 per second and total exposure time was 6.4 s, resulting in a total dose of 59.2 electrons per Å 2 . With dose fractionation set at 0.16 s per frame, each movie series contained 40 frames and each frame received a dose of 1.48 electrons per Å 2 . Fully automated data collection was carried out using SerialEM ( Mastronarde, 2005 ) with a nominal defocus range set from −1.5 to −3 μm. A total of 13,549 dose fractionation movies were collected in this session. The recorded movies were corrected for drift using Relion’s MotionCor2 implementation ( Zheng et al., 2017 ; Zivanov et al., 2019 ), and contrast transfer function (CTF) parameters were determined using GCTF ( Zhang, 2016 ). From our previous reconstitution attempts we learned that the Rag-Ragulator-FLCN-FNIP2 structure is rather elongated, and therefore considered the possibility that our previous picking efforts were missing the small orthogonal views of the protein complex. To address this, we created two sets of manually-picked particles (1,000 each) that contained either the elongated or the small orthogonal views of the protein. These two hand-selected particle sets were then used to train a deep-learning particle picker, crYOLO ( Wagner et al., 2019 ), which produced two large particle sets of two different box sizes. The larger box size of 360 pixels, corresponding to 374 Å, yielded 1,340,028 particles, and a smaller box size of 172 pixels, corresponding to 179 Å, produced 724,249 particles. After extraction and downscaling, the resulting two particle sets were processed separately in Relion ( Zivanov et al., 2018 ). Three rounds of reference-free 2D classifications were used to remove incorrectly selected particles and those of incomplete complexes. The remaining particles (sets of 864,167 and 519,928) were combined, and a strict distance cut-off of 250 Å was used to discard duplicates. The resulting clean set of 1,384,095 particles was re-extracted at full size, and used in 3D classifications. Our earlier 6.34 Å map served as a starting model, and in the first round of 3D classifications produced two (out of four) classes of high quality. These classes were combined together, and subsequently used in iterative cycles of per-particle CTF refinement and per-particle motion correction in Relion ( Zivanov et al., 2018 ). After convergence, and another round of 3D classifications, we identified two (out of four) classes that showed the highest level of structural detail. These 3D classes were combined and the resulting 615,470 particles were used in further 3D refinement. The final map obtained from these particles was sharpened with a B-factor of −22 Å 2 and estimated at 3.31 Å resolution, according to ‘gold standard’ Fourier shell correlation (FSC) of 0.143 ( Fig. S1 ). Local resolution was estimated using Relion to extend from 3.0 to 5.0 Å resolution ( Fig. S1 ).

Model Building and Refinement

Atomic models were prepared with Coot ( Emsley et al., 2010 ). We first fit in the available structures into our cryo-EM density map, including RagC-NBD in its GppNHp bound form (PDB: 3LLU), Ragulator-Rag(CRD) (PDB: 6EHR), FLCN-DENN (PDB: 3V42). Other parts of the model were built de novo by tracing the backbone, following domain topology, and registering the bulky residues based on the secondary structure predictions by I-TASSER ( Roy et al., 2010 ; Yang et al., 2015 ; Zhang, 2008 ) and Jpred ( Drozdetskiy et al., 2015 ). Real-space refinements of FLCN-FNIP2-Rag-Ragulator were performed using PHENIX ( Adams et al., 2010 ; Liebschner et al., 2019 ) with secondary structure restraints. MolProbity ( Chen et al., 2010 ) was used to evaluate the geometries of the structural model. Corrected Fourier shell correlation curves were calculated between refined atomic model and the cryo-EM density map.

Preparation of Cell Lysates and Immunoprecipitates

Cell lysates and immunoprecipitates were prepared as previously established ( Shen et al., 2017 ). In brief, two million HEK-293T cells were first plated onto a 10 cm dish. Twenty-four hours later, the cells were transfected with the plasmids indicated in the figure panels. Thirty-six hours later, cells were rinsed once with PBS and lysed in Triton Lysis Buffer (40 mM NaHEPES, pH 7.4; 5 mM MgCl 2 ; 10 mM Na 4 P 2 O 7 ; 10 mM sodium β-glycerol phosphate; 1% Triton X-100; and one tablet of EDTA-free protease inhibitor (Roche) per 25 ml of buffer). The lysates were cleared and immunoprecipitated with FLAG-M2 affinity gel. Following immunoprecipitation, the gel was washed once with Triton Lysis Buffer and three times with Triton Lysis Buffer supplemented with 500 mM sodium chloride. Immunoprecipitated proteins were denatured by SDS buffer, resolved by SDS-PAGE gels, and analyzed by immunoblotting.

Kinetic Measurements

To determine the stimulatory effect of FLCN-FNIP2 on the hydrolysis rate of the Rag GTPases under single turnover conditions, increasing amounts of FLCN-FNIP2 was incubated with 50 nM Rag GTPase heterodimer that is preloaded with ~ 0.1 nM of γ- 32 P-GTP. Small aliquots of the reaction were withdrawn at different time points and quenched by 0.75 M KH 2 PO 4 (pH 3.3). The time points were then expanded by thin layer chromatography (TLC) plates, and imaged and quantified with phosphorimaging screens. Linear regression was used to fit the fraction of phosphate against time, to generate the observed rate constants ( k obsd ). The k obsd s were then fit to a single binding equation.

QUANTIFICATION AND STATISTICAL ANALYSIS

All the kinetic assays were measured at least three times, and the results were reported with Mean ± SEM.

DATA AND CODE AVAILABILITY

Atomic coordinates and structure factors have been deposited in the Protein Data Bank (PDB) under the accession number 6ULG. Electron density maps have been deposited in EM Data Bank under the accession number EMD-20814.

LEAD CONTACT AND MATERIALS AVAILABILITY

The reagents generated in this study are available with no restriction. Further information and requests for resources and reagents should be directed to the Lead Contact, David M. Sabatini ( sabatini@wi.mit.edu ).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

HEK-293T cells were obtained from American Type Culture Collection (ATCC) and were maintained in an incubator setting at 37°C and 5% CO 2 . They were cultured in Dulbecco’s Modified Eagle Medium (DMEM), supplemented with 10% IFS, 2 mM glutamine, 100 IU/ml penicillin, and 100 μg/ml streptomycin. HEK-293 FreeStyle cells were obtained from Thermal Fisher Scientific and were maintained in a Multitron shaker set at 37°C, 125 r pm, 8% CO 2 , and 80% humidity. They were cultured in FreeStyle 293 Expression Medium, supplemented with 100 IU/ml penicillin and 100 μg/ml streptomycin. BL21(DE3) E. Coli strain was grown at 37°C in a Multitron shaker in LB media. To induce protein expression, BL21(DE3) E. Coli strain was transformed by the corresponding plasmids (see below), propagated at 37°C, and induc ed with IPTG at 18°C.

METHOD DETAILS Protein Preparation The Rag GTPase heterodimer was purified as previously described ( Shen et al., 2017 ). In brief, co-expression of C-terminally His-tagged RagA(T21N) and tagless RagC was induced by 0.5 mM IPTG in BL21(DE3) cells. The heterodimer was sequentially purified through Q Sepharose, Ni-NTA, MonoQ, and Superdex 200 columns. Ragulator was purified as previously described ( Shen and Sabatini, 2018 ). In brief, the five subunits of Ragulator were co-transformed and expressed in BL21(DE3) strain. The pentameric complex was sequentially purified through GST, Ni-NTA, MonoQ, and Superdex 200 columns. Human FLCN-FNIP2 was expressed and purified as previously described ( Shen et al., 2017 ; Tsun et al., 2013 ). In brief, HA-tagged FLCN were co-expressed with Flag-tagged FNIP2 in FreeStyle 293 cells. The complex was isolated using Flag-M2 beads and further purified by a Superdex 200 gel-filtration column. The FLCN-FNIP2-Rag-Ragulator complex was assembled as follows. 1 mg of purified FLCN-FNIP2 was incubated with 1 mg of RagA(T21N)-RagC and 2 mg of Ragulator in a total volume of 1 ml for ten hours at 4°C. The assembly solution contains 50 mM NaHEPES (pH 7.4), 100 mM NaCl, 2 mM MgCl 2 , 2 mM DTT, 0.1% CHAPS, 100 μM GppNHp, and 100 μM GDP. The nanomeric complex was separate from excess Rag GTPases and Ragulator on a Superdex 200 column. The FLCN-FNIP2-Rag-Ragulator complex was concentrated to 8 μg/μl in a 100kDa molecular weight cut-off concentrator and ultracentrifuged at 100,000× g for 30 minutes immediately prior to freezing grids.

Cryo grids preparation

Cryo grids were prepared immediately after protein purification. 400-mesh Quantifoil 1.2/1.3 Cu grid (Quantifoil, Großlöbichau Germany) was made hydrophilic by glow discharging for 60 seconds with a current of 15 mA in a Pelico EasiGlow system. Cryo grids were prepared using an FEI Mark IV Vitrobot (FEI, part of Thermo Fisher Scientific, Hillsboro, OR). The chamber of the Vitrobot was kept at 4°C and 100% relative humidity. The blotting time was 3 seconds with an equipment-specific blotting force set at 3. 3 μl of sample was applied to the glow-discharged grid and then rapidly plunge-frozen into a liquid ethane bath.

Image collection and processing

Two data sets were collected on two different 300 kV FEI Titan Krios cryo electron microscopes (FEI) at HHMI Janelia Research Campus. The first data set was collected on Janelia Krios2 with spherical aberration Cs of 2.7 mm and equipped with a Gatan K2 Summit camera. The final exposure was collected in dose fractionation mode on the K2 camera at a calibrated magnification of 38,168, corresponding to 1.31 Å per physical pixel in the image (0.655 Å per super-resolution pixel). The dose rate on the specimen was set to be 5.83 electron per Å 2 per second and total exposure time was 10 s, resulting in a total dose of 58.3 electrons per Å 2 . With dose fractionation set at 0.25 s per frame, each movie series contained 40 frames and each frame received a dose of 1.46 electrons per Å 2 . Fully automated data collection was carried out using SerialEM with a nominal defocus range set from −1.5 to −3 m. A total of 4535 dose fractionation movies were collected in this session. One round of data processing was done on this dataset. Beam-induced motion were measured, corrected, and dose-weighted at 1.46 electron/Å2 per frame with data binned by 2 using cisTEM ( Grant et al., 2018 ). CTF determination for each movie series was calculated by amplitude averaging of every 3 frames using cisTEM. Automated particle picking using ab inito mode was carried out in cisTEM on all the micrographs and 927042 particles were extracted. Two rounds of reference-free 2D classification with CTF correction was performed in cisTEM to throw away bad particles. 321015 particles were kept for further processing. Ab initio 3D initial model was generated using CryoSparc ( Punjani et al., 2017 ). Subsequent 3D refinement was performed in cisTEM to generate a 3D reconstruction density map. Fourier Shell Correlation at a criteria of 0.143 reported resolution of 6.34 Å for this map. To improve the resolution of the reconstruction, a larger data set was collected at higher magnification on Janelia Krios1. This Krios microscope is equipped with a spherical aberration corrector (Cs=0.01 mm), an energy filter (Gatan GIF Quantum) and a post-GIF Gatan K2 Summit direct electron detector. The final exposure was collected in dose fractionation mode on the K2 camera at a calibrated magnification of 48,077, corresponding to 1.04 Å per physical pixel (0.52 Å per super-resolution pixel). The dose rate on the specimen was set to be 9.25 electron per Å 2 per second and total exposure time was 6.4 s, resulting in a total dose of 59.2 electrons per Å 2 . With dose fractionation set at 0.16 s per frame, each movie series contained 40 frames and each frame received a dose of 1.48 electrons per Å 2 . Fully automated data collection was carried out using SerialEM ( Mastronarde, 2005 ) with a nominal defocus range set from −1.5 to −3 μm. A total of 13,549 dose fractionation movies were collected in this session. The recorded movies were corrected for drift using Relion’s MotionCor2 implementation ( Zheng et al., 2017 ; Zivanov et al., 2019 ), and contrast transfer function (CTF) parameters were determined using GCTF ( Zhang, 2016 ). From our previous reconstitution attempts we learned that the Rag-Ragulator-FLCN-FNIP2 structure is rather elongated, and therefore considered the possibility that our previous picking efforts were missing the small orthogonal views of the protein complex. To address this, we created two sets of manually-picked particles (1,000 each) that contained either the elongated or the small orthogonal views of the protein. These two hand-selected particle sets were then used to train a deep-learning particle picker, crYOLO ( Wagner et al., 2019 ), which produced two large particle sets of two different box sizes. The larger box size of 360 pixels, corresponding to 374 Å, yielded 1,340,028 particles, and a smaller box size of 172 pixels, corresponding to 179 Å, produced 724,249 particles. After extraction and downscaling, the resulting two particle sets were processed separately in Relion ( Zivanov et al., 2018 ). Three rounds of reference-free 2D classifications were used to remove incorrectly selected particles and those of incomplete complexes. The remaining particles (sets of 864,167 and 519,928) were combined, and a strict distance cut-off of 250 Å was used to discard duplicates. The resulting clean set of 1,384,095 particles was re-extracted at full size, and used in 3D classifications. Our earlier 6.34 Å map served as a starting model, and in the first round of 3D classifications produced two (out of four) classes of high quality. These classes were combined together, and subsequently used in iterative cycles of per-particle CTF refinement and per-particle motion correction in Relion ( Zivanov et al., 2018 ). After convergence, and another round of 3D classifications, we identified two (out of four) classes that showed the highest level of structural detail. These 3D classes were combined and the resulting 615,470 particles were used in further 3D refinement. The final map obtained from these particles was sharpened with a B-factor of −22 Å 2 and estimated at 3.31 Å resolution, according to ‘gold standard’ Fourier shell correlation (FSC) of 0.143 ( Fig. S1 ). Local resolution was estimated using Relion to extend from 3.0 to 5.0 Å resolution ( Fig. S1 ).

Model Building and Refinement

Atomic models were prepared with Coot ( Emsley et al., 2010 ). We first fit in the available structures into our cryo-EM density map, including RagC-NBD in its GppNHp bound form (PDB: 3LLU), Ragulator-Rag(CRD) (PDB: 6EHR), FLCN-DENN (PDB: 3V42). Other parts of the model were built de novo by tracing the backbone, following domain topology, and registering the bulky residues based on the secondary structure predictions by I-TASSER ( Roy et al., 2010 ; Yang et al., 2015 ; Zhang, 2008 ) and Jpred ( Drozdetskiy et al., 2015 ). Real-space refinements of FLCN-FNIP2-Rag-Ragulator were performed using PHENIX ( Adams et al., 2010 ; Liebschner et al., 2019 ) with secondary structure restraints. MolProbity ( Chen et al., 2010 ) was used to evaluate the geometries of the structural model. Corrected Fourier shell correlation curves were calculated between refined atomic model and the cryo-EM density map.

Preparation of Cell Lysates and Immunoprecipitates

Cell lysates and immunoprecipitates were prepared as previously established ( Shen et al., 2017 ). In brief, two million HEK-293T cells were first plated onto a 10 cm dish. Twenty-four hours later, the cells were transfected with the plasmids indicated in the figure panels. Thirty-six hours later, cells were rinsed once with PBS and lysed in Triton Lysis Buffer (40 mM NaHEPES, pH 7.4; 5 mM MgCl 2 ; 10 mM Na 4 P 2 O 7 ; 10 mM sodium β-glycerol phosphate; 1% Triton X-100; and one tablet of EDTA-free protease inhibitor (Roche) per 25 ml of buffer). The lysates were cleared and immunoprecipitated with FLAG-M2 affinity gel. Following immunoprecipitation, the gel was washed once with Triton Lysis Buffer and three times with Triton Lysis Buffer supplemented with 500 mM sodium chloride. Immunoprecipitated proteins were denatured by SDS buffer, resolved by SDS-PAGE gels, and analyzed by immunoblotting.

Kinetic Measurements

To determine the stimulatory effect of FLCN-FNIP2 on the hydrolysis rate of the Rag GTPases under single turnover conditions, increasing amounts of FLCN-FNIP2 was incubated with 50 nM Rag GTPase heterodimer that is preloaded with ~ 0.1 nM of γ- 32 P-GTP. Small aliquots of the reaction were withdrawn at different time points and quenched by 0.75 M KH 2 PO 4 (pH 3.3). The time points were then expanded by thin layer chromatography (TLC) plates, and imaged and quantified with phosphorimaging screens. Linear regression was used to fit the fraction of phosphate against time, to generate the observed rate constants ( k obsd ). The k obsd s were then fit to a single binding equation.

Supplementary Material 1 2 Supplementary Figure 1. Structural determination of the FLCN-FNIP2-Rag-Ragulator complex ( Related to Figure 1 ) A. Workflow for data processing of the FLCN-FNIP2-Rag-Ragulator dataset. B. Sample images for the 2D clustering of the FLCN-FNIP2-Rag-Ragulator complex. Views of the complex from different perspectives can be observed and clustered from the dataset. C & D. Half-set gold-standard Fourier shell correlation (FCS) (A) and map-model FSC (B) for the FLCN-FNIP2-Rag-Ragulator. E. Local resolution of the FLCN-FNIP2-Rag-Ragulator cryo-EM density map. Supplementary Figure 2. Structural model for FLCN-FNIP2 ( Related to Figure 2 ) A-C. Sample regions from the cryo-EM density maps and the fitted structural model for α-helical (A), β-strand (B), and loop (C) regions of FLCN. Secondary structures and bulky side chains can be registered and resolved at the current resolution. D. Architecture of FLCN, domain arrangement, and registration of secondary structure. E. Architecture of FNIP2, domain arrangement, and registration of secondary structure. Supplementary Figure 3. Nucleotide binding of the Rag GTPases ( Related to Figure 3 ) A & B. Bound nucleotide and the corresponding cryo-EM density for RagA (A) and RagC (B). Switch I of RagA vanishes as indicated by the dashed line. Switch I of RagC is shown in blue. Supplementary Figure 4. Stimulated hydrolysis assay to probe the effect of FLCN-FNIP2 ( Related to Figure 4 ) A. Sample time course of a GTP hydrolysis reaction when the reaction was let to reach completion. Only ~50% of the GTP was hydrolyzed, suggesting the bound GTP from only one subunit was stimulated to hydrolyze by FLCN-FNIP2. This experiment was repeated twice and a representative data set is shown here. B. Specificity of GTP hydrolysis reaction using mutant RagA(Q66L) or RagC(Q120L). The Rag GTPase heterodimers carrying the RagC(Q120L) mutation abolish the stimulatory effect of FLCN-FNIP2, while those carrying the RagA(Q66L) mutation maintain it. This experiment was repeated twice and a representative data set is shown here. C. Single turnover GTP hydrolysis assay to determine the influence of Ragulator on the stimulatory effect of FLCN-FNIP2. D. Ragulator has mild impact on the stimulatory effect of FLCN-FNIP2. When the Rag GTPases bind Ragulator, FLCN-FNIP2 stimulates GTP hydrolysis to a similar extent as the Rag GTPases alone. This experiment was repeated twice and a representative data set is shown here.