Abstract
Molecular switches are central to signal transduction in protein interaction networks. One switch protein can independently regulate distinct cellular processes, but the molecular mechanisms enabling this functional multi-specificity remain unclear. Here we integrate system-scale cellular and biophysical measurements to study how a paradigm switch, the small GTPase Ran/Gsp1, achieves its functional multi-specificity. We make 56 targeted point mutations to individual interactions of Ran/Gsp1 and show through quantitative, systematic genetic and physical interaction mapping that Ran/Gsp1 interface perturbations have widespread cellular consequences that cluster by biological processes but, unexpectedly, not by the targeted interactions. Instead, the cellular consequences of the interface mutations group by their biophysical effects on kinetic parameters of the GTPase switch cycle, and cycle kinetics are allosterically tuned by distal interface mutations. We propose that the functional multi-specificity of Ran/Gsp1 is encoded by a differential sensitivity of biological processes to different kinetic parameters of the Gsp1 switch cycle, and that Gsp1 partners binding to the sites of distal mutations act as allosteric regulators of the switch. Similar mechanisms may underlie biological regulation by other GTPases and biological switches. Finally, our integrative platform to determine the quantitative consequences of cellular perturbations may help explain the effects of disease mutations targeting central switches.
Proteins perform their cellular functions within networks of interactions with many partners (1, 2). This complexity raises the fundamental question of functional specificity: How can different functions be controlled individually with the required precision and accuracy, when distinct cellular processes are interconnected and often even share common regulators? Moreover, in highly interconnected networks even a small perturbation targeting individual interactions, introduced by posttranslational modifications, point mutations, or drug binding, could be magnified through the network and have widespread cellular consequences. Protein mutations in disease are enriched in protein-protein interfaces (3, 4), but it is unclear whether the consequences of these mutations can be explained primarily by their effects on individual interactions. Similarly, drug compounds are typically designed against specific targets but could affect cellular functions more broadly. Determining the extent and mechanism by which molecular perturbations affect interconnected biological processes requires an approach that quantifies effects on both the cellular network and on the molecular functions of the targeted protein (Fig. 1a).
To develop such an approach, we targeted a central molecular switch, a GTPase. GTPases belong to a class of common biological motifs, where a two-state switch is controlled by regulators with opposing functions (6, 7) (Fig. 1a). For GTPases, the two states of the switch are defined by the conformation of the GTPase in either the GTP- or GDP-bound forms, and the interconversion between the two states is catalyzed by guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs) (Fig. 1b). Switch motifs are often multi-specific, defined here as regulating several different processes (8). This multi-specificity raises the question of how a single switch motif differentially controls diverse processes at the cellular level.
In this study, we sought to uncover the mechanistic basis of functional multi-specificity in the small GTPase Gsp1 (the S. cerevisiae homolog of human Ran, which shares 83% amino acid identity with Gsp1), which is a single molecular switch with one main GEF and one main GAP (9). Gsp1 regulates nucleocytoplasmic transport of proteins (10, 11) and RNA (12, 13), cell cycle progression (14), RNA processing (15) and nuclear envelope assembly (16). Gsp1/Ran forms direct physical interactions with a large number of partners, and high-resolution crystal structures of Gsp1/Ran in complex with 16 different binding partners are known (Fig. S1, Supplementary File 1 Table 1). We reasoned that by placing defined point mutations in Gsp1 interfaces with these partners, we could differentially perturb subsets of biological processes regulated by Gsp1. We then determined the functional consequences of these Gsp1 mutations on diverse biological processes in S. cerevisiae using quantitative genetic interaction mapping, measured changes to the physical interaction network using affinity purification mass spectrometry (AP-MS), and finally quantified molecular changes on the Gsp1 switch motif using biophysical studies in vitro (Fig. 1 a, b).
Targeted perturbations to GTPase interaction interfaces
To target each of the 16 known interaction interfaces of Gsp1, we designed 56 S. cerevisiae strains with genomically integrated point mutations in the GSP1 gene (Fig. 1c, Fig. S1, Supplementary File 1 Tables 2, 3). To avoid simultaneously affecting all Gsp1 functions and to create viable mutant strains (as Gsp1 is essential), we excluded mutations in the Gsp1 nucleotide binding site and the switch I and II regions (17). We confirmed by Western blot that the mutant Gsp1 protein levels were close to the endogenous wild-type levels (Fig. S2).
Genetic interactions of Gsp1 mutants
To determine the functional consequences of the Gsp1 interface mutations, we performed a genetic interaction (GI) screen in S. cerevisiae using the epistatic mini-array profile (E-MAP) approach (18, 19). We measured growth of each GSP1 point mutant in the context of an array of single gene knockouts, resulting in a quantitative functional profile of up to 1444 GI values for each GSP1 point mutant. Similarity of genetic interaction profiles often indicates shared functions. The 56 GSP1 point mutants fell into two clusters, 23 ‘strong’ mutants with rich GI profiles containing 9-373 significant interactions (Fig. 1d), and 33 ‘weak’ mutants with 0-8 significant interactions (Fig. S3a, Methods and Supplementary File 1 Fig. 1). The strong mutants covered eleven Gsp1 sequence positions and all 16 structurally characterized Gsp1 protein interaction interfaces (Fig. 1e). Remarkably, twelve of the GSP1 interface point mutants had a greater number of significant GIs than an average deletion of a non-essential S. cerevisiae gene, and six GSP1 point mutants had more GIs than an average temperature sensitive mutant of an essential gene from a previously published large-scale S. cerevisiae GI map (20) (Fig. 1f, Supplementary File 2). The GIs of the designed Gsp1 interface mutations spanned diverse biological processes known to be linked to Gsp1, including mRNA transport, the dynein/dynactin pathway, tRNA modification by the elongator complex, and spindle assembly regulation. Furthermore, unbiased hierarchical clustering of S. cerevisiae genes solely by their GI profiles with the 56 GSP1 point mutants also grouped many other genes by their biological complex or pathway membership such as members of the Hog1 signaling pathway, SWR1 and Rpd3L complexes, and mitochondrial proteins (Fig. 1d, Fig. S3b). Taken together, the GI analysis reveals expansive functional consequences of Gsp1 interface point mutations - similar in magnitude to the effects typically observed for deleting entire genes - that illuminate many of the biological functions of GSP1.
In contrast to their clustering of biological processes, the GI profiles of the Gsp1 point mutants did not group based on their location in the Gsp1 partner interfaces (Fig. 1e). For example, strains with GSP1 mutations at residues Thr34 (T34E, T34Q) and Asp79 (D79S, D79A) have similar GI profiles (Fig. 1d) but these mutations are in different interaction interfaces (Fig. 1e) on opposite sides of the Gsp1 structure (Fig. 1c). This observation was unexpected and contrary to our initial expectation that Gsp1 achieves its functional specificity by interacting with different partners and, accordingly, targeting different protein interfaces should affect distinct functions of Gsp1. To analyze this finding further and quantify the functional similarities between individual GSP1 mutants across most biological processes in S. cerevisiae, we compared the GSP1 mutant GI profiles to profiles from 3358 S. cerevisiae genes (20) using Pearson correlations. In this analysis, significant positive correlations signify functional relationships (21) (Supplementary File 3, Supplementary File 1 Table 4, Fig. S3c). Strikingly, GI profiles of GSP1 mutants and of Gsp1 physical interaction partners were on average no more similar to each other in instances where the Gsp1 mutation was located in the partner interface than when the mutation was not (Fig. 1g, Fig. S3d). These results suggest that the rich functional profiles of GSP1 mutants cannot simply be explained by considering only the interface or partner interaction targeted by the point mutation.
Physical interactions of Gsp1 mutants
To investigate further why the GI profiles of Gsp1 mutations did not group based on targeted specific physical interactions of Gsp1, we sought to determine changes to the physical interaction network in response to the mutations (for example, mutations could fail to target the intended interactions). We tagged wild-type Gsp1 and 28 mutants covering all interface residues shown in Fig. 1e with an amino- or carboxy-terminal 3xFLAG tag and quantified the abundance of each of the ‘prey’ partner proteins in complex with Gsp1 by AP-MS (Fig. 2a, Fig. S4, Supplementary File 4). We refer to the prey partner protein abundance in the pulled-down Gsp1 complexes simply as “abundance” below. As expected, Gsp1 mutations in the cores of interaction interfaces decreased the abundance of the prey partner targeted by each specific mutation (Fig. 2b, large half-circles representing location of the mutation in the interface core are predominantly red, indicating a decreased interaction with the targeted prey partner). However, when considering the changes in abundance of all physical interaction partners observed in the AP-MS experiments (core regulators Rna1 and Srm1, and effectors Yrb1, Kap95, Pse1 and Srp1), we found that several of these interaction partners showed significant changes in interactions with almost all Gsp1 mutants (Fig. 2c). For example, the point mutations at residue Thr34 are in the core of the interface with Yrb1 (Fig. 2d) and accordingly resulted in lower abundance of Yrb1 compared to wild type (Fig. 2b) but also significantly affected the interactions with other partners (Fig. 2c). Over the entire dataset, the most notable change in abundance occurred with the two core GTPase cycle regulators, GAP (Rna1) and GEF (Srm1) (Fig. 2c, Fig. S4d, Supplementary File 1 Table 5), despite most mutations not being in the Rna1 or Srm1 interfaces (Fig. 2d-f). In summary, the AP-MS experiments confirm that the point mutations affected the targeted interactions but also reveal unexpected, extensive changes to the physical interaction network of Gsp1 that cannot simply be explained by the interface location of the mutations.
Molecular mechanism of Gsp1 point mutants
The AP-MS experiments showed that most Gsp1 interface mutations significantly altered physical interactions with the two principal GTPase regulators, GAP and GEF. To address the question whether the mutations act directly or indirectly (i.e. by altering the competition between physical interaction partners in the cell), we recombinantly expressed and purified wild-type Gsp1 and 22 mutants that expressed and purified well. We then measured the effect of the mutations on GAP-mediated GTP hydrolysis and GEF-mediated nucleotide exchange in vitro (Fig. 3a, b, Fig. S5, Supplementary File 1 Figures 2, 3, and Tables 6, 7). Of the 22 Gsp1 point mutants, 14 (all of which had strong GI profiles except K132H) showed 3-to >200-fold effect on either or both of the GAP- or GEF-mediated reactions (Fig. S5e). In particular, mutations that are not in the interface with the GAP both increased (3-fold, R108G mutant) and decreased (10-fold, R78K, T34E/Q/A, and D79S mutants) the catalytic efficiency of GAP-mediated GTP hydrolysis, compared to wild-type Gsp1 (Fig. 3a). As expected, mutations in the interface with the GEF (K101, and R108) decreased the catalytic efficiency of GEF-mediated nucleotide exchange >40-fold. However, other mutations not in the GEF interface (R78K, Y157A) also decreased the efficiency notably (3-to 10-fold, Fig. 3b). These results show that Gsp1 interface mutations are capable of modulating the GTPase cycle by directly affecting GTP hydrolysis and nucleotide exchange catalyzed by the two core switch regulators, GAP and GEF. Moreover, since seven out of the 14 mutations with larger than 3-fold effects are located outside of the interfaces with either the GAP or the GEF as well as outside the known switch regions, our data suggest considerable, previously unappreciated, allostery in the GTPase switch.
To probe the mechanism of these allosteric effects, we examined the impact of Gsp1 point mutations on the conformational distribution in the active site of GTP-bound Gsp1 using 1D 31P nuclear magnetic resonance (NMR) spectroscopy. Similar to the human homolog Ran (22), 1D 31P NMR spectra of S. cerevisiae wild-type Gsp1:GTP showed two distinct peaks for the γ-phosphate of bound GTP arising from differences in the local chemical environment of the γ-phosphate in each of two distinct conformations (termed γ1 and γ2) (Fig. S6a). While 87% of wild-type Gsp1:GTP was in the γ2 state conformation, the populations of the γ2 state in Gsp1 mutants ranged from close to 0% for T34E and T34Q, to close to 100% for H141R, Y157A, and K132H (Fig. 3c). Remarkably, these interface sites are at least 18 Å away from the γ phosphate of Gsp1-bound GTP (Fig. 3d, e).
Prior work on human Ran (22) and the related GTPase Ras (23) had implicated the γ2 state conformation in effector binding. Here we observe a striking linear relationship when plotting the γ2 state population in the different Gsp1 mutants against their effects on relative catalytic efficiency of GAP-mediated GTP hydrolysis (Fig. 3f) and intrinsic GTP hydrolysis (Supplementary File 1 Table 8, Fig. S6b, c). These experiments directly quantify the allosteric effects of the mutations and indicate that the γ2 state represents the hydrolysis-competent conformation of Gsp1:GTP. Exceptions to the linear relationship are the K132H mutation, which is in the core of the GAP interface and is hence expected to directly affect the interaction with the GAP, and the D79S and R78K mutations, which are on the edge of the GTPase switch II region (from residues 69 to 77) and could lead to additional perturbations of the nucleotide binding site geometry. Taken together, the 31P NMR and kinetic data support a molecular mechanism whereby Gsp1 interface mutations allosterically shift the conformational distribution at the active site which in turn alters the GTPase switch cycle. Interestingly, the allosteric mutations we show here to tune the distribution of the catalytically competent state of Gsp1 (T34, H141, Q147, and Y157, Fig. 3d, e) do not overlap with either allosteric inhibitor pocket successfully targeted by small molecule inhibitors in Ras (24–26) (Fig. S6d).
Our in vitro analysis showed that most Gsp1 interface mutations affect GEF-catalyzed nucleotide exchange and GAP-catalyzed GTP hydrolysis differentially (Fig. 3a,b, and Fig. S5e). To determine to what extent these effects explain changes in the physical interaction network of Gsp1 (Fig. 2), we compared in vitro kinetic and our AP-MS data (Fig. 4a). We found that all Gsp1 mutants that affected the efficiency (kcat/Km) of GEF-catalyzed nucleotide exchange more than the efficiency of GAP-catalyzed GTP hydrolysis (points above the diagonal in Fig. 4a) showed a larger decrease in the abundance of pulled-down GEF compared to pulled-down GAP (teal points), while the opposite was the case for the mutants below the diagonal (orange points in Fig. 4a). Similar but weaker relationships were observed for other prey proteins (Fig. S7). We conclude that Gsp1 interface mutations allosterically perturb the GTPase cycle, and that the direction of the cycle perturbation is a good predictor of altered physical interactions with the two main cycle regulators, even in the context of many other potentially competing partner proteins.
Encoding of Gsp1 multi-specificity
We next asked whether the allosteric effect of the mutations on the balance of the GTPase cycle, rather than the interface in which a mutation is made, could better explain the functional effects observed in the cellular GI profiles. We clustered the GI profiles of the Gsp1 mutants based on correlation with the GI profiles of 3358 S. cerevisiae genes (20). We then compared clustering of these GI profile correlations (using the 278 genes with significant correlations to Gsp1 mutants, Fig. 4b, Fig. S8a) with the biophysical effects of the mutations on the efficiencies of GAP-catalyzed GTP hydrolysis and GEF-catalyzed nucleotide exchange. Remarkably, the Gsp1 mutant GI profile clustering mirrored an approximate ordering by the in vitro mutant effects on the GTPase cycle: relative GEF efficiency systematically decreased with increasing row number and relative GAP efficiency systematically increased (Fig. 4b). A clear outlier of this ordering is the K101R mutant, which primarily affects GEF-mediated nucleotide exchange in vitro but, by GI profiles, groups with mutations affecting the efficiency of GTP hydrolysis. A lysine at the corresponding position (Lys99) is annotated to be acetylated in mammalian systems (27), and this acetylation was shown to reduce the efficiency of nucleotide release from Ran:GDP:GEF complexes (28). We hypothesize that while our K101R mutation affected the interaction with the GEF, it also likely broke a critical mechanism by which the cell reduces GEF activity, phenocopying the mutants with reduced GTP hydrolysis activity. This observation suggests the possibility that the allosteric sites discovered here might be used to control cellular functions via posttranslational modifications.
Finally, we asked whether our biophysical measurements of how the different Gsp1 mutants perturb the GTPase cycle regulation could provide insight into Gsp1’s functional multi-specificity, i.e. its ability to distinctly regulate multiple biological processes. Clustering the S. cerevisiae genes and Gsp1 mutants based on correlations of their respective GI profiles revealed that genes fall into one of three categories (Fig. 4b, Fig. S8a): (i) genes that correlate strongly with all or most of the Gsp1 point mutants (most strikingly genes in cluster 5, but also clusters 3, 4, and 6), irrespective of the direction of GTPase cycle perturbation; (ii) genes in cluster 1, but also cluster 2, that correlate primarily with mutants more perturbed in the GTP hydrolysis side of the cycle (orange bars) and (iii) genes in cluster 7 that correlate primarily with mutants more perturbed in the nucleotide exchange side of the cycle (teal bars). Most importantly, this clustering distinguishes between biological processes, since genes with shared biological function (Supplementary File 5) all predominantly fall into one of the three categories defined above (Fig. 4c, Fig. S8b-e and 9). Overall, our analysis suggests that distinct processes regulated by Gsp1, such as nuclear transport, spindle assembly, and tRNA modification (Fig. 4c) as well as 5 mRNA capping, transcription regulation, and cell polarity (Fig. S9b-d) are differentially sensitive to perturbations of each side of the cycle.
Discussion
Only five years after the discovery of the small GTPase Ran, Rush et al. (29) proposed that Ran must act by two different mechanisms: one in which the cycling of the GTPase is most important (‘Rab paradigm’), and the other in which the amount of “active” Ran:GTP is most important (‘Ras paradigm’). Our findings lead to a model where Ran/Gsp1 acts by three different paradigms that are defined by the sensitivity of different biological processes to perturbations of different characteristics of the Gsp1 GTPase cycle, i.e. the ability to (i) cycle, (ii) turn off by hydrolyzing to Gsp1:GDP, and (iii) activate by producing Gsp1:GTP. Other effects such as direct perturbations of interactions and binding partner competition undoubtedly also play a role in modulating the detailed effects of our Gsp1 mutations. Nevertheless, our model explains to a remarkable degree how a single molecular switch motif can differentially control subsets of biological processes by using one of the three functional modes.
The discovery of several allosteric sites (positions 34, 141, 147, and 157) in the model molecular switch Gsp1 both explains the widespread functional consequences we observe for single amino acid point mutations at interaction surfaces of Gsp1 and has important implications for revising our understanding of GTPase switch regulation. We show that mutations in distal interaction interfaces allosterically modulate the switch cycle. This finding demonstrates thermodynamic coupling between interaction interfaces and the classical switch region in the active site and thereby suggests that partners binding to distal sites also regulate the switch by affecting conformational equilibria at the active site. This hypothesis is supported by evidence that the Yrb1 homolog RanBP1 modulates GAP activity (22, 30, 31). Our data provide a mechanistic explanation, where mutations at allosteric sites, including Thr34 in the Yrb1 binding interface, tune the population of Gsp1 in a hydrolytically-primed conformation. Since the overall switch mechanism is conserved across the small GTPase fold, we propose that thermodynamic coupling between distal interfaces and functional conformational changes may be a more general mechanism to regulate other GTPase switches, and may aid in the development of allosteric inhibitors.
Our observation of widespread functional effects induced by relatively small mutational perturbations is reminiscent of the ultrasensitivity achievable in biological motifs with opposing regulators (6). While switch-like ultrasensitivity is typically described for systems controlled by covalent modifications (such as phosphorylation), our results, as well as the observations that cellular levels of small GTPase regulators require tight control (32, 33), corroborate a model of ultra-sensitivity for GTPase conformational switches (34). While we investigated the changes to the GTPase cycle caused by mutations, similar effects on regulation could be exerted by partner binding or posttranslational modification.
Finally, deriving a model that explains the cellular multi-specificity of GTPases by differential sensitivity of biological processes to distinct parameters of the switch cycle was enabled by a quantitative analysis that integrated functional genomics, proteomics, and biophysics. Given the prevalence of biological two-state switch motifs controlled by opposing regulators (kinase/phosphatase, acetylase/deacetylase) (35), we envision this approach to be fruitful for other studies of cellular regulation and to be extended to mammalian systems using CRISPR-based approaches to yield mechanistic insights into the drastic consequences of disease mutations targeting central molecular switches.
Methods
Point mutations in genomic Gsp1 sequence
We identified all residues in Gsp1 that comprised the interfaces with Gsp1 binding partners for which co-complex crystal structures with Gsp1 were available (Supplementary File 1 Table 1, Fig. S1a). Residues comprising interface core, support or rim were defined based on relative change in solvent accessible surface area (ΔrASA), as previously defined (5), compared to the empirical maximum solvent accessible surface area for each of the 20 amino acids (36). Accessible surface area was calculated using the bio3d R package (37). All the custom code for interface analysis from co-complex crystal structures is provided in the associated code repository in the structure analysis folder. We avoided Gsp1 residues that are within 5 Å of the nucleotide (GDP or GTP) in any of the structures or that are within the canonical small GTPase switch regions (P-loop, switch loop I, and switch loop II). We then mutated residues that are located in interface cores (defined as residues that bury more than 25% of their surface upon complex formation, as previously defined (5), Supplementary File 1 Table 2, Fig. S1b) into amino acid residues with a range of properties (differing in size, charge and polarity) and attempted to make stable and viable S. cerevisiae strains carrying a genomic Gsp1 point mutation coupled to nourseothricin (clonNAT / nourseothricin, Werner BioAgents GmbH, CAS 96736-11-7) resistance. The list of attempted mutants is provided in Supplementary File 1 Table 3. The genomic construct was designed to minimally disrupt the non-coding sequences known at the time, including the 5 UTR and 3 UTR, as well as the putative regulatory elements in the downstream gene Sec72 (Supplementary File 1 Fig. 4). The GSP1 genomic region was cloned into a pCR2.1-TOPO vector (Invitrogen) and point mutations in the GSP1 coding sequence were introduced using the QuikChange™ Site-Directed Mutagenesis (Stratagene, La Jolla) protocol. S. cerevisiae strains containing mutant GSP1 genes were regularly confirmed by sequencing the Gsp1 genomic region.
S. cerevisiae genetics and genetic interaction mapping
S. cerevisiae transformation
To generate MAT:α strains with Gsp1 point mutations the entire cassette was amplified by PCR using S. cerevisiae transformation forward and reverse primers, and S. cerevisiae was transformed into the starting SGA MAT:α his3D1; leu2D0; ura3D0; LYS2þ; can1::STE2pr-SpHIS5 (SpHIS5 is the S. pombe HIS5 gene); lyp1D::STE3pr-LEU2 strain from (38) as described below. Primers for amplifying the GSP1 genomic region: S. cerevisiae Transformation FWD: GTATGATCAACTTTTCCT-CACCTTTTAAGTTTGTTTCG S. cerevisiae Transformation REV: GATTGGAGAAACCAACCCAAATTTTACAC-CACAA
DNA competet S. cerevisiae cells were made using a LiAc protocol. The final transformation mixture contained 10 mM LiAc (Lithium acetate dihydrate, 98%, extra pure, ACROS Organics™, CAS 6108-17-4), 50 µg ssDNA (UltraPure™ Salmon Sperm DNA Solution, Invitrogen, 15632011), 30% sterile-filtered PEG 8000 (Poly(ethylene glycol), BioUltra, 8,000, Sigma-Aldrich, 89510-250G-F). A S. cerevisiae pellet of approximately 25 µl was mixed with 15 µl of linear DNA PCR product and 240 µl of the transformation mixture, and heat shocked at 42°C for 40 minutes. Transformed cells were grown on YPD (20 g Bacto™ Peptone (CAT # 211820, BD Diagnostic Systems), 10 g Bacto™ Yeast Extract (CAT # 212720 BD), and 20 g Dextrose (CAT # D16-3, Fisher Chemicals) per 1-liter medium) + clonNAT plates and incubated at 30°C for 3 to 6 days. Many colonies that appeared after 24-48 hours carried the clonNAT cassette but not the GSP1 point mutation, or the 3xFLAG tag. Cells were therefore sparsely plated and plates were incubated for a longer period of time after which colonies of different sizes were picked and the mutant strains were confirmed by sequencing.
Epistatic mini-array profiling (E-MAP) of Gsp1 point mutants
Genetic interactions of all viable GSP1 point mutant (PM-GSP1-clonNAT) strains were identified by epistatic miniarray profile (E-MAP) screens (19, 38) using a previously constructed array library of 1,536 KAN-marked (kanamycin) mutant strains assembled from the S. cerevisiae deletion collection (39) and the DAmP (decreased abundance by mRNA perturbation) strain collection (40), covering genes involved in a wide variety of cellular processes (18, 41). The E-MAP screen was conducted as previously described in Collins et al. (38), using the HT Colony Grid Analyzer Java program (21) and E-MAP toolbox for MATLAB to extract colony sizes of double mutant strains and a statistical scoring scheme to compute genetic interaction scores. Genetic interaction scores represent the average of 3-5 independent replicate screens. Reproducibility was assessed as previously described (21) by comparing individual scores to the average score for each mutant:gene pair, with the two values showing strong correlation across the dataset (Pearson correlation coefficient = 0.83, Supplementary File 1 Figure 5).
Hierarchical clustering of E-MAP genetic interaction data
All E-MAP library DAmP strains as well as library strains showing poor reproducibility were discarded, leaving 1444 out of the original 1536 library genes. Averaged S-scores of genetic interactions between wild type and point mutant Gsp1 and the 1444 S. cerevisiae genes are provided in Supplementary File 2. Hierarchical clustering on the genetic interaction profiles was performed using the average linkage method and the pairwise Pearson correlation coefficient as a distance metric. To identify clusters of functionally related library genes, the hierarchical clustering tree was cut to produce 1200 clusters, resulting in 43 clusters with 3 or more members. Biological function descriptions for genes in these clusters were extracted from the Saccharomyces Genome Database (SGD) (42). Clusters of genes representing common functions (complexes, pathways or biological functions) were selected by manual inspection and represented in the main text Fig. 1d and Fig. S3b. All of the custom code for E-MAP analysis is provided in the E-MAP analysis GitHub repository. Clustered heatmaps were produced using the ComplexHeatmap package (43).
Scaling of published genetic interaction data to the E-MAP format
To enable comparison of GSP1 point mutant genetic interaction profiles to profiles of other S. cerevisiae genes, published Synthetic Gene Array (SGA) genetic interaction data (20) from CellMap.org (44) were scaled to the E-MAP format using a published non-linear scaling method (45). Briefly, 75,314 genetic interaction pairs present in both the SGA and a previously described E-MAP dataset used to study chromatin biology (46) were ordered by genetic interaction score and partitioned into 500 equally sized bins separately for each dataset. Bin size (150 pairs per bin) was chosen to provide enough bins for fitting the scaling spline (described below) while still maintaining a large number of pairs per bin such that the mean could be used as a high confidence estimate of the score values in each bin. Scaling factors were computed that scaled the mean of each SGA bin to match the mean of the corresponding E-MAP bin. A non-linear univariate spline was fit through the scaling factors, providing a scaling function that was subsequently applied to each SGA score. The distribution of scores of shared interactions between the scaled SGA and the E-MAP chromatin library was similar to that between replicates in the E-MAP chromatin library, matching what was seen in the previously published scaling of SGA data to E-MAP format (45) (Supplementary File 1 Fig. 6). The SGA genetic interaction scores are taken from CellMap.org (44). The scaling code is provided in SGA scaling GitHub repository.
Significance of genetic interactions
The S-score metric used in scoring genetic interactions measured by the E-MAP method has been previously characterized in terms of confidence that any given averaged S-score represents a significant interaction (21). We fit a spline to data points from Fig. 4c from Collins et al (21), allowing us to provide an approximate confidence estimate for each of our measured GSP1 and scaled S. cerevisiae SGA genetic interaction scores. The SGA dataset (20) is accompanied by p-values as well as its own recommendations for a threshold at which individual interactions are considered significant. We plotted the SGA score scaled to E-MAP format vs. the associated p-value (negative log-transformed, Supplementary File 1 Fig. 1a) and found the distribution to have a similar shape to the confidence function for S-scores (Supplementary File 1 Fig. 1b). For example, a 95% confidence threshold is associated with E-MAP S-scores less than −4 or greater than 5, while the median p-value of scaled SGA scores is less than 0.05 for scores less than −5 or greater than 3. We ultimately elected to use a significance cutoff of absolute S-score greater than 3. This threshold corresponds to an estimated confidence value of 0.83 for S-scores less than −3 and 0.65 for S-scores greater than 3. We compared these values to the intermediate significance threshold recommended for the SGA data from Ref. (20), which was p-value < 0.05 and absolute SGA score > 0.08. After scaling to E-MAP format, this threshold corresponds to scaled S-scores less than −2.97 or greater than 2.25, below our chosen threshold of −3 and 3.
GI profile correlation measurements
Of the 1444 library genes in the GSP1 point mutant GI profile map, 1129 were present in the SGA dataset from Ref. (20). Pairwise Pearson correlation coefficients were computed between all GSP1 point mutants and SGA gene profiles, and all profiles trimmed to include only genetic interaction measurements with the 1129 shared library genes. Due to the relative sparsity of genetic interaction profiles, pairwise comparisons are dominated by high numbers of non-significant interactions. Accordingly, we did not consider correlations with GSP1 point mutants or SGA gene profiles that did not have significant genetic interactions (absolute scaled S-score greater than 3, see above) with at least 10 of the 1129 library genes. This requirement removed all weak Gsp1 point mutants and one strong mutant (R108A) from the correlation analysis (as they had at most nine genetic interactions with absolute score greater than 3), leaving 22 strong mutants and 3383 S. cerevisiae SGA genes to be included in the correlation analysis. All Pearson correlations and their p-values between Gsp1 mutants and S. cerevisiae genes, including all correlations that did not pass our significance filtering procedures, are provided in Supplementary File 3. The subset of Pearson correlations between Gsp1 point mutants and Gsp1 partners with available co-complex X-ray crystal structures, used to make the point plots in Figs. 1g and 7c, d, are also available in Supplementary File 1 Table 4. Statistical significance of correlations was computed using both two-sided and one-sided (positive) t-tests adjusted for multiple hypothesis testing using both the Bonferroni method and the FDR method, which controls the false discovery rate (47). All p-values reported in the text and figures are one-sided (positive) and corrected by the FDR method, unless otherwise stated. Custom code for genetic interaction profile correlation calculations and filtering is provided in the accompanying E-MAP correlation GitHub repository.
Significance testing was used to filter out S. cerevisiae gene SGA profiles that did not show a significant correlation (one-sided positive, Bonferroni-adjusted) with at least two GSP1 point mutants profile. In total, 278 S. cerevisiae genes from the SGA had a significant profile correlation (one-sided positive, Bonferroni-adjusted) with at least two GSP1 point mutants and were therefore included in the correlation analysis shown in Fig. 4b and Fig. S8a. After this filtering step, the one-sided p-values were used to populate a matrix of 22 mutants vs. 278 genes, and hierarchical clustering was performed using Ward’s method. Pearson correlation between correlation vectors was used as a distance metric for the mutant (row) clustering, while Euclidean distance was selected for the gene (column) clustering, due to the column vectors being relatively short (22 mutants per column vs. 278 genes per row) and thus sensitive to outliers when clustered using Pearson correlations as the distance metric.
For the gene set analysis we decreased the stringency of inclusion of S. cerevisiae SGA genes to include all genes with a significant profile correlation (one-sided positive, Bonferroni-adjusted) with one or more Gsp1 mutants, which added another 201 genes, resulting in 479 genes. We made the gene sets larger to increase our confidence in connecting the patterns of correlations between S. cerevisiae genes and Gsp1 mutants, and GTPase cycle parameters represented in Fig. 4c and Fig. S9. Manually curated gene sets of S. cerevisiae genes with significant correlations with Gsp1 mutants are provided in Supplementary File 5.
Protein expression levels by Western Blot
S. cerevisiae strains were grown at 30°C in YPD medium (20 g Bacto™ Peptone (CAT # 211820, BD Diagnostic Systems), 10 g Bacto™ Yeast Extract (CAT # 212720 BD), and 20 g Dextrose (CAT # D16-3, Fisher Chemicals) per 1 L medium) for 1.5 - 2 hours until OD600 reached 0.3. Cell culture aliquots of 1 ml were centrifuged for 3 minutes at 21,000 x g and resuspended in 30 µl of phosphate buffered saline (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4, pH = 7.4) and 10 µl of SDS-PAGE Sample Buffer (CAT # 161-0747, BioRad), to a final SDS concentration of 1%, and 2 mM beta-mercaptoethanol. Lysates were run (3 µl for most, and 6 µl for slow growing mutants with lower OD600) on Stain-Free gels (4-20%, CAT # 4568096, Bio-Rad, Tris/Glycine SDS Buffer (CAT # 161-0732, BioRad)). After electrophoresis, the gel was scanned for total protein quantification and the proteins were subsequently transferred to an Immobilon-FL PVDF membrane (CAT # IPF00010, EMD Millipore). The membrane was probed with Rabbit anti-RAN (CAT # PA 1-5783, ThermoFisher Scientific) primary, and Goat anti-Rabbit-IgG(H+L)-HRP (CAT # 31460, Thermo Fisher) secondary antibodies. The membrane was developed using Super Signal West Femto substrate (CAT # 34096, Thermo Fisher), and scanned and analyzed with Image Lab software on a ChemiDoc MP (BioRad). Each blot had at least one wild-type (WT-GSP1-clonNAT) and at least one MAT:α strain control. The total protein levels (TPMUT) for each Gsp1 point mutant lane were then normalized to the wild-type (WT-GSP1-clonNAT) lane of the corresponding blot (TPWT), providing an adjustment value to account for differences in loading between lanes
To compute the relative expression of a Gsp1 point mutant, the density (DMUT) of the Western blot bands corresponding to the Gsp1 point mutant was divided by the total protein adjustment and finally normalized against the same value for the wild type Gsp1, i.e.
Note that for blots with a single WT lane, aWT = 1. For blots with more than one WT lane included, aWT was computed for each WT lane by normalizing to the average TP across all WT lanes, and the average adjusted WT density (DWT/aWT) across all WT lanes was used for computing the relative expression of point mutants. Example Western blots are provided in Supplementary File 1 Fig. 7, and the final protein expression level data for all the mutants are shown in Fig. S2. Identifying Gsp1 complexes in S. cerevisiae with affinity purification mass spectrometry (AP-MS)
S. cerevisiae cell lysate preparation
S. cerevisiae strains for AP-MS were grown in YAPD medium (120 mg adenine hemisulfate salt (CAT # A9126, SIGMA), 10 g Bacto yeast extract (CAT # BD 212720), 20 g Bacto peptone (CAT # BD 211820), 20 g dextrose (D-glucose D16-3 Fisher Chemicals) per 1 L of medium). Each strain was grown at 30°C for 12 to 24 h to OD600 of 1-1.5. The cells were harvested by centrifugation at 3000 RCF for 3 minutes and the pellet was washed in 50 ml of ice-cold ddH2O, followed by a wash in 50 ml of 2x lysis buffer (200 mM HEPES pH 7.5, 200 mM KCl, 2 mM MgCl2, 30 µM GTP (Guanosine 5-triphosphate sodium salt hydrate, CAT # G8877, Sigma-Aldrich), 1 mM Dithiothreitol (Promega V3151), 0.1% IGEPAL CA-630 (CAT # I8896, Sigma-Aldrich), and 10% glycerol). Each pellet of approximately 500 µl was then resuspended in 500 µl of 2X lysis buffer supplemented with protease inhibitors without EDTA (cOmplete, Mini, EDTA-free Protease Inhibitor Cocktail, CAT # 11836170001, Roche) and dripped through a syringe into liquid nitrogen. The frozen S. cerevisiae cell pellets were lysed in liquid nitrogen with a SPEX™ SamplePrep 6870 Freezer/Mill™.
FLAG immunoprecipitation
FLAG immunoprecipitations were performed as previously described (48, 49). Details are as follows. For FLAG immunoprecipitations, frozen samples were initially kept at room temperature for 5 minutes and then placed on ice or at 4°C in all subsequent steps, unless indicated otherwise. Following the addition of 1.5 − 3.0 ml Suspension Buffer (0.1 M HEPES pH 7.5, 0.1 M KCl, 1 mM MgCl2, 15 µM GTP, and 0.5 mM Dithiothreitol) supplemented with cOmplete mini EDTA-free protease and PhosSTOP phosphatase inhibitor cocktails (Roche), samples were incubated on a rotator for at least 10 minutes and then adjusted to 6 ml total volume with additional Suspension Buffer supplemented with inhibitors before centrifugation at 18,000 rpm for 10 minutes. Anti-FLAG M2 Affinity Gel beads (50 µl slurry; Sigma-Aldrich) were washed twice with 1 ml Suspension Buffer. After reserving 50 µl, the remaining supernatant and anti-FLAG M2 Affinity Gel beads were combined and incubated for 2 hours on a tube rotator. Beads were then collected by centrifugation at 300 rpm for 5 minutes and washed three times. For each wash step, beads were alternately suspended in 1 ml Suspension Buffer and collected by centrifugation at 2,000 rpm for 5 minutes. After removing residual wash buffer, proteins were eluted in 42 µl 0.1 mg/ml 3xFLAG peptide, 0.05% RapiGest SF Surfactant (Waters Corporation) in Suspension Buffer by gently agitating beads on a vortex mixer at room temperature for 30 minutes. Immunoprecipitated proteins (4 µl) were resolved on 4-20% Criterion Tris-HCl Precast gels (BioRad) and visualized by silver stain (Pierce Silver Stain Kit; Thermo Scientific) (Supplementary File 1 Fig. 8) before submitting 10 µl of each sample for mass spectrometry. At least three independent biological replicates were performed for each FLAG-tagged protein and the untagged negative control.
Liquid chromatography with tandem mass spectrometry (LC-MS/MS) analysis
To prepare samples for LC-MS/MS analysis, immunoprecipitated protein (10 µl) was denatured and reduced in 2 M urea, 10 mM NH4HCO3, and 2 mM Dithiothreitol for 30 minutes at 60°C with constant shaking, alkylated in the dark with 2 mM iodoacetamide for 45 minutes at room temperature and digested overnight at 37°C with 80 ng trypsin (Promega). Following digestion, peptides were acidified with formic acid and desalted using C18 Zip-Tips (Millipore) according to the manufacturer’s specifications. Samples were re-suspended in 4% formic acid, 2% acetonitrile solution, and separated by a 75-minute reversed-phase gradient over a nanoflow C18 column (Dr. Maisch). Peptides were directly injected into a Q-Exactive Plus mass spectrometer (Thermo), with all MS1 and MS2 spectra collected in the orbitrap. Raw MS data were searched against the S. cerevisiae proteome (SGD sequences downloaded January 13, 2015) using the default settings in MaxQuant (version 1.5.7.4), with a match-between-runs enabled (50, 51). Peptides and proteins were filtered to 1% false discovery rate in MaxQuant, and identified proteins were then subjected to protein-protein interaction scoring using SAINTexpress (52). Label free quantification and statistical analysis were performed using MSstats (53). Fold change in abundance of preys for 3xFLAG-tagged Gsp1 point mutants was always calculated compared to the wild type Gsp1 with the corresponding tag. All AP-MS data are available from the PRIDE repository under the PXD016338 identifier. Custom code for analysis of AP-MS data is available from the accompanying AP-MS GitHub repository. Fold change values between prey abundance between the mutant and wild type Gsp1 and the corresponding FDR adjusted p-values are provided in Supplementary File 4. The intersection of all prey proteins identified at least once with both the amino- or carboxy-terminal 3xFLAG tag, and their interquartile ranges (IQR) of log2(fold change) values across all the Gsp1 mutants, are provided in Supplementary File 1 Table 5. Quality of data and reproducibility between replicates was assessed based on correlations of peptide counts between replicates (Supplementary File 1 Figs. 9, 10).
Biochemical and biophysical assays
Protein purifications
All proteins were expressed from a pET-28 a (+) vector with a N-terminal 6xHis tag in E. coli strain BL21 (DE3) in the presence of 50 mg/L Kanamycin. GEF (Srm1 from S. cerevisiae, (Uniprot P21827)) was purified as Δ1-27Srm1 and GAP (Rna1 from S. pombe, Uniprot P41391) as a full-length protein. ScΔ1-27Srm1 and SpRna1 were expressed in 2xYT medium (10 g NaCl, 10 g yeast extract (BD BactoTM Yeast Extract #212720), 16 g tryp-tone (Fisher, BP1421) per 1 L of medium) overnight at 25°C upon addition of 300 µmol/L Isopropyl-β-D-thiogalactoside (IPTG). Gsp1 variants were expressed by autoinduction for 60 hours at 20°C (54). The autoinduction medium consisted of ZY medium (10 g/L tryptone, 5 g/L yeast extract) supplemented with the following stock mixtures: 20xNPS (1M Na2HPO4, 1 M KH2PO4, and 0.5 M (NH4)2SO4), 50x 5052 (25% glycerol, 2.5% glucose, and 10% α-lactose monohydrate), 1000x trace metal mixture (50 mM FeCl3, 20 mM CaCl2, 10 mM each of MnCl2 and ZnSO4, and 2 mM each of CoCl2, CuCl2, NiCl2, Na2MoO4, Na2SeO3, and H3BO3 in 60 mM HCl). Cells were lysed in 50 mM Tris pH 7.5, 500 mM NaCl, 10 mM imidazole, and 2 mM β-mercaptoethanol using a microfluidizer from Microfluidics. For Gsp1 purifications, the lysis buffer was also supplemented with 10 mM MgCl2. The His-tagged proteins were purified on Ni-NTA resin (Thermo Scientific #88222) and washed into a buffer containing 50 mM Tris (pH 7.5) and 100 mM NaCl, with 5 mM MgCl2 for Gsp1 proteins. The N-terminal His-tag was digested at room temperature overnight using up to 12 NIH Units per mL of bovine thrombin (Sigma-Aldrich T4648-10KU). Proteins were then purified using size exclusion chromatography (HiLoad 26/600 Superdex 200 pg column from GE Healthcare), and purity was confirmed to be at least 90% by SDS polyacrylamide gel electrophoresis. Samples were concentrated on 10 kDa spin filter columns (Amicon Catalog # UFC901024) into storage buffer (50 mM Tris pH 7.5, 150 mM NaCl, 1 mM Dithiothreitol). Storage buffer for Gsp1 proteins was supplemented with 5 mM MgCl2.. Protein concentrations were confirmed by measuring at 10-50x dilution using a Nanodrop (ThermoScientific). The extinction coefficient at 280 nm used for nucleotide (GDP or GTP) bound Gsp1 was 37675 M-1 cm-1, as described in (55). The ratio of absorbance at 260 nm and 280 nm for purified Gsp1 bound to GDP was 0.76. Extinction coefficients for other proteins were estimated based on their primary protein sequence using the ProtParam tool. Concentrated proteins were flash-frozen and stored at −80 °C.
In our hands every attempt to purify the S. cerevisiae homologue of GAP (Rna1, Uniprot P11745) from E. coli yielded a protein that eluted in the void volume on the Sephadex 200 size exclusion column, indicating that the protein is forming soluble higher-order oligomers. We were, however, successful in purifying the S. pombe homologue of GAP (Rna1, Uniprot P41391) as a monomer of high purity as described above. S. pombe and S. cerevisiae Rna1 proteins have an overall 43% sequence identity and 72% sequence similarity. The X-ray crystal structure of Ran GTPase and its GAP used in our analyses is a co-complex structure of the S. pombe homolog of Rna1 (PDB: 15kd), human RAN, and human RANBP1 (Supplementary File 1 Table 1). We used the purified S. pombe homolog of Rna1 in all of our GTP hydrolysis kinetic experiments.
Circular dichroism (CD) spectroscopy of protein thermostability
Samples for CD analysis were prepared at approximately 2 µM Gsp1 in 2 mM HEPES pH 7.5, 5 mM NaCl, 200 µM MgCl2, and 50 µM Dithiothreitol. CD spectra were recorded at 25°C using 2 mm cuvettes (Starna, 21-Q-2) in a JASCO J-710 CD-spectrometer (Serial #9079119). The bandwidth was 2 nm, rate of scanning 20 nm/min, data pitch 0.2 nm, and response time 8 s. Each CD spectrum represents the accumulation of 5 scans. Buffer spectra were subtracted from the sample spectra using the Spectra Manager software Version 1.53.01 from JASCO Corporation. Temperature melts were performed from 25°C - 95°C, monitoring at 210 nm, using a data pitch of 0.5°C and a temperature slope of 1°C per minute. As all thermal melts of wild type and mutant Gsp1 proteins were irreversible, only apparent Tm was estimated (Supplementary File 1 Fig. 11) and is reported in Supplementary File 1 Table 9.
GTP loading of Gsp1
Gsp1 variants for GTPase assays as well as for 31P NMR spectroscopy were first loaded with GTP by incubation in the presence of 20-fold excess GTP (Guanosine 5-Triphosphate, Disodium Salt, CAT # 371701, Calbiochem) in 50 mM Tris HCl pH 7.5, 100 mM NaCl, 5 mM MgCl2. Exchange of GDP for GTP was initiated by the addition of 10 mM EDTA. Reactions were incubated for 3 hours at 4°C and stopped by addition of 1 M MgCl2 to a final concentration of 20 mM MgCl2 to quench the EDTA. GTP-loaded protein was buffer exchanged into either NMR buffer or the GTPase assay buffer using NAP-5 Sephadex G-25 DNA Grade columns (GE Healthcare # 17085301).
Reverse phase high performance liquid chromatography (HPLC)
Analysis of bound nucleotide was performed using reverse-phase chromatography as previously described (55) using a C18 column (HAISIL TS Targa C18, particle size 5 µm, pore size 120 Å, dimensions 150 × 4.6 mm, Higgins Analytical # TS-1546-C185). The column was preceded by a precolumn filter (The Nest Group, Inc, Part # UA318, requires 0.5 µm frits, Part # UA102) and a C18 guard column (HAICart SS Cartridge Column, HAISIL Targa C18, 3.2×20 mm, 5µm, 120 Å Higgins Analytical # TF-0232-C185, requires a Guard Holder Kit, Higgins Analytical # HK-GUARD-FF). To prepare the nucleotide for analysis, a Gsp1 sample was first diluted to a concentration of 25-30 µM and a volume of 40 µl. The protein was denatured by addition of 2.5 µl of 10% perchloric acid (HClO4). The pH was raised by addition of 1.75 µl 4 M sodium acetate (CH3COONa) pH 4.0. The nucleotide was separated from the precipitated protein before application to the column by spinning at 20,000 x g for 20 minutes. 30 µl of supernatant was withdrawn and mixed 1:1 with reverse-phase buffer (10 mM tetra-n-butylammonium bromide, 100 mM KH2PO4 / K2HPO4, pH 6.5, 0.2 mM NaN3). 20 µl of sample was injected onto the equilibrated column, and was run isocratically in 92.5% reverse-phase buffer, 7.5% acetonitrile at a flow rate of 1 ml/min for 35 min (20 column volumes). Nucleotide retention was measured by monitoring absorbance at both 254 nm and 280 nm. Example HPLC reverse phase chromatogram of GTP-loaded wild type Gsp1 is shown in Supplementary File 1 Fig. 12.
NMR Spectroscopy
Gsp1 samples for 31P NMR spectroscopy were first loaded with GTP as described above, and buffer exchanged into NMR Buffer (D2O with 50 mM Tris-HCl pH 7.4, 5 mM MgCl2, 2 mM Dithiothreitol). Final sample concentrations were between 250 µM and 2 mM, and 400 ul of sample was loaded into 5 mm Shigemi advanced microtubes matched to D2O (BMS-005TB; Shigemi Co. Ltd, Tokyo, Japan.). 31P NMR experiments were performed on a Bruker Avance III 600 MHz NMR spectrometer with a 5 mm BBFO Z-gradient Probe. Spectra were acquired and processed with the Bruker TopSpin software (version 4.0.3). Indirect chemical shift referencing for 31P to DSS (2 mM Sucrose, 0.5 mM DSS, 2 mM NaN3 in 90% H2O + 10% D2O; water-suppression standard) was done using the IUPAC-IUB recommended ratios (56). Spectra were recorded at 25°C using the pulse and acquire program zg (TopSpin 3.6.0), with an acquisition time of 280 milliseconds, a recycle delay of 3.84 seconds, and a 65° hard pulse. 4,096 complex points were acquired over the course of 4,096 scans and a total acquisition time of 4.75 hours. Spectra were zero-filled once and multiplied with an exponential window function (EM) with a line-broadening of 6 Hz (LB = 6) prior to Fourier transformation. Peaks were integrated using the auto-integrate function in TopSpin 4.0.7, and peak areas were referenced to the bound GTP-β peak of each spectrum. Values for the fraction of each variant in state 2 were computed by taking the area of the GTP-γ2 peak and dividing by the sum of the two GTP-γ peak areas.
Kinetic measurements of GTP hydrolysis
Kinetic parameters of the GTP hydrolysis reaction were determined using a protocol similar to one previously described (57). Gsp1 samples for GTP hydrolysis kinetic assays were first loaded with GTP as described above. GTP hydrolysis was monitored by measuring fluorescence of the E. coli phosphate-binding protein labeled with 7-Diethylamino-3-[N-(2-maleimidoethyl)carbamoyl]coumarin (MDCC) (phosphate sensor, CAT # PV4406, Thermo Fisher) upon binding of the free phosphate GTP hydrolysis product (excitation at 425 nm, emission at 457 nm). All experiments were performed in GTPase assay buffer (40 mM HEPES pH 7.5, 100 mM NaCl, 4 mM MgCl2, 1 mM Dithiothreitol) at 30°C in 100 µl reaction volume using a Synergy H1 plate reader from BioTek, using Corning 3881 96-well half-area clear-bottom non-binding surface plates. The phosphate sensor at 20 µM and 50 µM concentrations was calibrated with a range of concentrations of K2HPO4 using only the data in the linear range. For most mutants a concentration of 1 nM GAP (SpRna1, Rna1 from S. pombe) was used. In order to run the time courses to completion, for mutants with low kcat/Km enzyme concentrations of 2-7 nM were used. Initially we collected time course data for all Gsp1 variants at approximately 8 µM concentration of loaded Gsp1:GTP with 1 nM GAP and 20 µM phosphate sensor. Exact concentrations of loaded Gsp1:GTP were then estimated based on the plateau fluorescence and the sensor calibration parameters. Each time course was then fitted to an integrated Michaelis Menten equation using a Lambert W function, as previously described (58). We analysed the data with the custom-made software DELA written by David Lambright from University of Massachusetts Medical School and generously made available to us as freeware. If the estimated Km was higher than 1 µM, we repeated the time course kinetic experiments with higher concentration of Gsp1:GTP of approximately tenfold above the Km. The Michaelis Menten kcat and Km parameters and their standard deviations were calculated from at least 2-9 technical replicates from 1-4 independently GTP-loaded Gsp1 samples (Supplementary File 1 Table 6). For each individual GAP-mediated GTP hydrolysis experiment, a control experiment with the same concentration of GTP-loaded Gsp1 and the same concentration of sensor, but without added GAP was ran in parallel. The first 100 seconds of these data were used to determine the baseline fluorescence, and the rest of the data were linearly fit to estimate intrinsic GTP hydrolysis rate (Supplementary File 1 Table 8). Example data and integrated Michaelis-Menten fits are show in Supplementary File 1 Fig. 2.
Kinetic measurements of Srm1 mediated nucleotide exchange
Kinetic parameters of GEF mediated nucleotide exchange were determined using a fluorescence resonance energy transfer (FRET) based protocol (59). Each Gsp1 variant was purified as a Gsp1:GDP complex, as confirmed by reverse phase chromatography. Nucleotide exchange from GDP to mant-GTP (2’-(or-3’)-O-(N-Methylanthraniloyl) Guanosine 5-Triphosphate, CAT # NU-206L, Jena Biosciences) was monitored by measuring a decrease in intrinsic Gsp1 tryptophan fluorescence (295 nm excitation, 335 nm detection) due to FRET upon binding of the mant group. Each time course was measured in GEF assay buffer (40 mM HEPES pH 7.5, 100 mM NaCl, 4 mM MgCl2, 1 mM Dithiothreitol) with excess of mant-GTP. For most variants of Gsp1 we measured time courses at Gsp1:GDP concentrations ranging from 0.25 to 12 µM with an excess mant-GTP concentration of 200 µM. For Gsp1 variants with high Km values that had to be measured at concentrations of up to 200 µM we used an excess of 1000 µM mant-GTP. All kinetic measurements were done at 30°C in 100 µl reaction volume using 5 nM GEF (Δ1-27Srm1), except for higher concentrations of the mutants with high Km values that were measured at 20 nM GEF. Data were collected in a Synergy H1 plate reader from BioTek, using Corning 3686 96-well half-area non-binding surface plates. For low concentrations of Gsp1:GDP the time course data were fit to a combination of two exponential decays: where knuc. exch is the rate constant of the GDP to mant-GTP exchange, kbg is the rate constant of background decay due to photo-bleaching, and span1 and span2 are the fluorescence amplitudes for the two processes. For high concentrations of substrate, or for mutants with very low rates, the initial velocity was determined by a linear fit to the initial 10-20% of the data. The kinetic parameters of the nucleotide exchange were determined by fitting a Michaelis-Menten equation to an average of 38 data points (ranging from 17 to 91) per Gsp1 point mutant for a range of substrate concentrations from [Gsp1:GDP] = 0.25 µM to [Gsp1:GDP] » Km. Example data and Michaelis-Menten fits are shown in Supplementary File 1 Fig. 3. Michaelis-Menten kcat and Km parameters for GEF-mediated nucleotide exchange are provided in Supplementary File 1 Table 7. The errors of the kcat and the Km parameters were determined from the standard error of the exponential fit of the Michaelis-Menten equation to the data. The error of the catalytic efficiency (kcat/Km) was calculated by adding the standard errors of the individual parameters and normalizing it for the values of the parameters
All the custom written code for fitting and analysis of kinetics data is provided in the accompanying kinetics GitHub repository.
Author Contributions
T.P., C.J.P.M., N.J.K. and T.K. identified and developed the core questions. T.P. and C.J.P.M. performed the bulk of the experiments and data analysis. J.X. and T.P. performed the E-MAP screens. G.M.J. performed the pull-down experiments. D.L.S. and R.K. performed the MS experiments and together with T.P. analyzed the data. N.O. contributed to design of Gsp1 mutants. H.B. contributed to E-MAP analysis. M.J.S.K. suggested the NMR studies. C.J.P.M. and M.J.S.K. performed the NMR experiments and analyzed the data. Y.Z. performed the Western blot experiments. T.P., C.J.P.M., and Y.Z. prepared the proteins. T.P. performed the kinetics experiments. T.P., C.J.P.M. and T.K. wrote the manuscript with contributions from the other authors. N.J.K. and T.K. over-saw the project.
Data Availability
The mass spectrometry proteomics data have been deposited to the PRIDE proteomics data repository with the dataset identifier PXD016338 and are available as Supplementary Tables. Raw biophysics data (cycle kinetics, CD, and NMR), and E-MAP S-scores, scaled SGA (44) scores, and their correlations are available from a GitHub repository All other data that support the findings of this study are available within the paper and its Supplementary Files.
Code Availability
Custom written R and Python scripts are available without restrictions from a a Gsp1 manuscript GitHub repository
Competing Interests
The authors declare no competing interests.
Acknowledgements
We thank Roxana Ordonez for contributions to the design of mutations; Cristina Melero, Deborah Jeon, Shivani Mathur, Raina Danbi Kim, and Kale Kundert for technical help; Maru Jaime Garza for contributions to the conformational analysis by NMR; Colm Ryan for advice on E-MAP analysis; David Lambright for guidance and software to analyze kinetics of GAP-mediated hydrolysis; and Dave Agard, Geeta Narlikar, James Fraser and Janet M. Thornton for stimulating discussions; Ricardo Henriques and his lab for the preprint template. This work was supported by a grant from the National Institutes of Health (R01-GM117189) to T.K., and a Sir Henry Wellcome Postdoctoral Fellowship (101614/Z/13/Z) to T.P. C.J.P.M. is a UCSF Discovery Fellow. T.K. is a Chan Zuckerberg Biohub investigator.