Sec7 regulatory domains scaffold autoinhibited and active conformations

The late stages of Golgi maturation involve a series of sequential trafficking events in which cargo-laden vesicles are produced and targeted to multiple distinct subcellular destinations. Each of these vesicle biogenesis events requires activation of an Arf GTPase by the Sec7/BIG guanine nucleotide exchange factor (GEF). Sec7 localization and activity is regulated by autoinhibition, positive feedback, and interaction with other GTPases. Although these mechanisms have been characterized biochemically, we lack a clear picture of how GEF localization and activity is modulated by these signals. Here we report the cryoEM structure of full-length Sec7 in its autoinhibited form, revealing the architecture of its multiple regulatory domains. We use functional experiments to determine the basis for autoinhibition and use structural predictions to produce a model for an active conformation of the GEF that is supported empirically. This study therefore elucidates the conformational transition that Sec7 undergoes to become active on the organelle membrane surface.

SI Appendix, Figure S1 -Legend on following page SI Appendix, Figure S1 -T.terrestris Sec7 purification and CryoEM data processing workflow.a, SDS-PAGE of purified T. terrestris Sec7 expressed in P. pastoris.b, Tryptophan fluorescence GEF activity assay demonstrating that the purified T. terrestris Sec7 construct is capable of myristoylated-Arf1 activation in the presence of TGN liposomes (see Methods).Note that the observed rate and sigmoidal shape of the curve are consistent with autoinhibitory behavior previously determined for S. cerevisiae Sec7 (1).c, Representative cryoEM micrograph.d, CryoEM data processing workflow.S2 -Heterogeneous Reconstruction and UMAP analysis by cryoDRGN.Dimer particles were focus-refined on a monomer (without symmetry expansion or subtraction) and used to train a cryoDRGN variability model (2).a, Cryo EM map of the Sec7 dimer reconstruction (left), and three example cryoDRGN generated maps highlighting the flexibility of the Sec7 dimer.b, All maps generated by cryoDRGN.Maps 6-10 significantly lack density for a second monomer, perhaps due to flexibility.Red arrows indicate maps with little/no density for the GEF domain bound to HDS2.We note that this only occurs in the monomer that was not focused on during refinement, and therefore could be artifactual.S1

SI Appendix, Video S2
Hypothetical model for the conformational changes Sec7 undergoes when switching from the autoinhibited to active states.A model of the inactive conformation containing all residues of T. terrestris Sec7 was generated by SWISS-MODEL (10) templated with the cryoEM experimental model, and then a morphing transition was created using the AlphaFold predicted model of T. terrestris Sec7 in ChimeraX (11).The first few frames of this morph are looped in the beginning to highlight the presumed spontaneous, but infrequent, dissociation of the GEF domain from the HDS2 domain.Once Sec7 adopts the active conformation, it can be stabilized by binding to the activated forms of regulatory GTPases such as Arl1 (depicted in blue), allowing activation of the Arf1 GTPase substrate (depicted in red).

Purification of T. terrestris Sec7 for cryoEM
After transformation, 5 colonies of CFY4970 were patched together on a fresh Zeocin plate, and cultured with autoinduction media as described (15).Cells were collected by centrifugation after 48hrs, and yeast cell paste was flash frozen in liquid nitrogen in small aliquots.Cell paste was then lysed in a Spex cryogenic mill (6875D) for 15 cycles, 15 cps, 2 min rest between cycles, and powder was stored at -80 until further processing.

Purification of S. cerevisiae proteins for biochemical analysis
S. cerevisiae Sec7 constructs were grown, induced, and harvested as described above.
Batch affinity purification was performed with resin that had been fragmented by sonication (see below).The clarified lysate was incubated with the resin for 2 hours at 4 °C with rotation.After binding the resin was washed five times with 10 ml wash buffer (50 mM Hepes, pH 7.4, 450 mM NaCl, 5% glycerol, 40 mM Imidazole, 1 mM DTT, and 0.1% CHAPS), and transferred to a fresh tube after the third wash.Sec7 was eluted from the NiNTA resin by 3C protease cleavage overnight at 4 °C.This elution was further purified by size exclusion chromatography using a Sepharose 6 increase 10/300 (SEC buffer: 25 mM Hepes, pH 7.4, 250 mM NaCl, 5% glycerol, 1 mM DTT).SEC fractions were analyzed by SDS-PAGE, pooled, concentrated, flash frozen in liquid nitrogen, and stored at -80 for later use.

Preparation of fragmented affinity resin
In order to increase the surface area of the Ni-NTA resin, the resin was resuspended in water to make a 20% slurry, then sonicated at 80% power for three minutes with a macro-tip, (cycles: 20s on / 10s off).Following sonication we equilibrated the fragmented resin 5x with buffer at a lower speed (1000 rpm) to remove resin fines.

CryoEM data processing
Movies were motion-corrected and dose-weighted using MotionCor2 (17), and micrograph defocus values were estimated using GCTF (18).Micrographs were manually inspected and culled to 3,401 usable micrographs, which were imported into CryoSPARC for particle picking (19).Defocus values were estimated with patch-CTF estimation in CryoSPARC and 'blob picker' was used to pick an initial set of particles.2D-classification was used to generate templates for template picking.An initial set of 938,642 particles was used for Ab-initio model generation and 3D-classification was performed iteratively using heterogeneous 3D refinement to generate a final set of 296,177 particles.
These particles were re-extracted in RELION 3.1 (20) and assigned GCTF-estimated per-particle defocus values.Iterative rounds of CTF refinement and Bayesian polishing improved the resolution from 7.14 Å to 5.3 Å. 3D classification was attempted, but no improvement was attainable for the dimer reconstruction.Focused refinement on a monomer(21) was performed using a mask including a single full monomer and a small portion of the other (corresponding approximately to the HDS4 domain) for particle subtraction after symmetry expansion.A second monomer mask containing only a single monomer was used for refinement.After several iterations of CTF refinement, fixed angle 3D classification isolated 196,888 symmetry expanded particles which produced a 3.7 Å (0.143 FSC) reconstruction after iterative CTF refinement.The published crystal structure of the DCB-HUS domain and AlphaFold prediction (12,22) were used for guidance with de novo building in the few regions with poor side chain density.A monomeric atomic model was refined into the monomer map using Real Space Refine (23) in Phenix (24).A model for the dimer was then generated and refined into a composite dimer map produced from the monomer map and the consensus dimer map using Phenix Combine Maps, and validated in Phenix (24)(25)(26).See Tables 1 and 2 and SI Appendix, Figures S1 and S3.
For cryoDRGN analysis, we used TOPAZ to increase the likelihood of rare particles (27) (this did not improve resolution of the monomer).Starting with 1,240,146 topaz picked particles heterogeneous 3D classification in CryoSPARC generated a final stack of 280,528 particles that were then used to generate a dimer map with C2 symmetry imposed (6.3 A), followed by a focused monomer refinement without subtraction before cryoDRGN training and analysis (SI Appendix, Figure S2) (2).

Fluorescence microscopy
Cells were grown overnight at 30 °C in liquid selection media (-Leu) to an OD of 0.6.
Cells were allowed to settle on a coverslip dish (MatTek) for 10 min, and washed with fresh media.Imaging for SI Appendix, Figure S4 was done using a CSU-X spinning-disk confocal system (Intelligent Imaging Innovations) with a DMI6000 B microscope (Leica), 10031.46NA oil immersion objective, and a QuantME EMCCD camera (Photometrics).Imaging was done using a DeltaVision Elite system equipped with an Olympus IX-71 inverted microscope, a DV Elite complementary metal-oxide semiconductor camera, a ×100/1.4NA oil objective, and a DV Light SSI 7 Color illumination system with Live Cell Speed Option with DV Elite filter sets.Exposure and laser power were adjusted according to intensity, and were kept the same for all specimens being compared in an experiment.The brightness/intensity was equivalently adjusted across all images in an experiment using ImageJ.

Liposome preparation
Synthetic TGN liposomes were prepared as described previously (16).In brief -lipid stocks in chloroform were combined in a pear-shaped flask to produce a lipid mixture mimicking that of the yeast TGN(28) (SI Appendix, Table S3).Chloroform was evaporated slowly in a rotary evaporator heated to ~37 °C, then rehydrated in HK buffer (20mM HEPES pH 7.5, 150 mM KOAc) at 37 °C overnight.After gentle resuspension, the mixture was extruded through 100 nm filters 21 times and stored at 4 °C for no more than 1 month.

Liposome flotation membrane binding assay
Liposome flotation was performed as described (29).Briefly -100 nm liposomes were loaded with Arf1 by EDTA exchange of GMPPNP for a final concentration of 250 µM lipid and the indicated Arf1 concentration.Sec7 constructs were added to a final concentration of 550 nM, and incubated at room temperature for 1 hour.Then 2.5 M sucrose in HK (20mM HEPES pH 7.5, 150 mM KOAc) was added to 1 M final concentration, and 80 µl was transferred to a polycarbonate tube.Then a 100 ul layer of 0.75 M Sucrose was added, followed by 20 µl of HK.
This was centrifuged at 20 oC in a TLA100 rotor for 30 minutes, and the top layer collected for SDS-PAGE analysis (12% acrylamide gel).

In vitro Arf activation (GEF) assay
GEF activity was determined by measuring the native Tryptophan fluorescence of Arf1, as described previously (16).Briefly, to a final volume of 150 µl in HKM (20mM HEPES pH 7.5, 150 mM KOAc, 2 mM MgCl 2, 1 mM DTT), 200 µM 100 nm TGN liposomes, Sec7 construct (with concentration as detailed below), 1 µM myristoylated Arf1, and 200 µM GTP were added to a quartz cuvette.For Figures 2 and 4, GEF was added to a final concentration of 100 nM.For Figure 6b, 20 nM GEF was added.For Figure 6a, 60 nM GEF was added.For Figure 5g and SI Appendix, Figure S6c, 200 nM GEF was added.Tryptophan fluorescence (297.5 nm excitation, 340 nm emission) was measured using a fluorometer, and between additions the reaction was well mixed and the fluorescence was allowed to stabilize.The order listed is the order components were added, except for reactions with preloaded Ypt31.For these reactions, 500 nM prenylated Ypt31 was loaded onto liposomes by EDTA exchange of GMPPNP at 30 °C for 30 minutes.The presence of GMPPNP in the liposome mixture dictated that Arf1 was added last.Curves were fitted in GraphPad PRISM 10 using a nonlinear regression (one phase association) accounting for drift if apparent in the data to extrapolate the rate constant, which was then divided by the GEF concentration to calculate the exchange rate.

Statistical analysis
All assays were performed in triplicate, and statistical analysis was performed using GraphPad PRISM 10.GEF assay comparisons were analyzed by unpaired parametric t-test.
Liposome floatation comparisons were analyzed by paired ratio t-test.
c, Plots of UMAP analysis colored by reconstruction weight of the maps in b (top) or by density of particles (bottom).d, Isolated UMAP distribution plots of each map in b.SI Appendix, Figure S3 -Focused refinement of a Sec7 monomer.a, Map generated by focused refinement of an entire single monomer using symmetry expanded particles gave the highest resolution, 3.7 Å overall.Local resolution determined using RELION (3) was used to color the map as indicated.b, Model built and refined using the map in a, colored by domains as indicated.SI Appendix, Figure S4 -Legend on following page SI Appendix, Figure S4 -Detailed view of HDS2 interface.a, Multiple sequence alignment of the HDS2 interface in Sec7 and Gea2 homologs.Boxed regions indicate the surface exposed loops that contact the GEF domain.b, Key residues of the HDS2 interface are shown.c, Comparison with the autoinhibited Grp1 crystal structure.Grp1 autoinhibition involves a basic patch in the PH domain that interacts with its GEF domain in a distinct manner from that of the Sec7 HDS2-GEF domain interaction.d, Western blot for α-GFP-Sec7, showing the L1376D construct is expressed similarly to WT after shuffling.e, Microscopy of GFP-Sec7 constructs after shuffling in sec7Δ::KanMX strain background.f, Example SDS-PAGE gel of purified protein constructs used in assays shown in Fig. 2d,e and Fig. 4a-d.SI Appendix, Figure S5 -Predicted structural model of Sec7 in an active conformation.a, AlphaFold prediction with Arf1 bound to the GEF domain as in the Gea2-Arf1 cryoEM structure (4) shows the catalytic surface of the GEF domain is accessible in this conformation.b, Positions of the GEF linkers (757-816 and 1067-1082 in S. cerevisiae) surround the HUS box in the AlphaFold prediction, and physically connect the GEF domain to the activating surface of the DCB/HUS domains.Note residues 1040-1082 correspond to the GEF-HDS1 linker identified in the Gea2 cryoEM structure (4).SI Appendix, Figure S6 -SEC/MALS of monomeric Sec7 construct and membrane binding.a, Purified monomeric Sec7 eluted ~2.5 ml later than WT Sec7 over a Superose 6 10/300 column (left), and was verified by MALS to be monomeric (right).b, Quantification of reaction rate constants for triplicate measurements as reported in Fig. 5g.n.s.: not significant, * p < 0.05 c, Representative traces of ΔN17-Arf1 exchange catalyzed by the indicated GEF constructs.SI Appendix, Figure S7 -Dimerization and membrane binding.a, Representative SDS-PAGE of membrane float assay.b, Quantification of triplicate measurements in a. n.s.: not significant, * p < 0.05, * p < 0.01 SI Appendix, Figure S8 -Raw data for Fig. 6d.Representative trace of GEF assay shown in Fig 6d.SI Appendix, Figure S9 -The autoinhibited conformation of Sec7 is not compatible with stable membrane binding.Comparison of Gea2 cryoEM (a), autoinhibited Sec7 cryoEM (b), and AlphaFold-predicted Sec7 (c) structures on the membrane surface.

-Plasmids
* All constructs encode full-length proteins unless otherwise noted.SI Appendix,