An asymmetric nautilus-like HflK/C assembly controls FtsH proteolysis of membrane proteins

FtsH•HflK/C purification without protein overexpression.

We engineered an E. coli strain in which the chromosomal ftsH gene was modified to encode a C-terminal FLAG epitope and then used anti-FLAG antibody resin to purify tagged FtsH and associated HflK and HflC (Supplementary Figure 2A; see Methods). Addition of the FLAG tag did not slow E. coli growth at 37°C (Supplementary Figure 2B), despite FtsH protease activity being essential for cell viability (Herman et al., 1993). Thus, we inferred that the tag did not markedly affect activity in vivo. In preparation for single-particle cryo-EM, we used n-dodecyl-β-D-maltoside (DDM) or glyco-diosgenin (GDN) detergents to solubilize the FtsH•HflK/C complex prior to purification. Following DDM solubilization, the purified FtsH•HflK/C complex was active in ATP hydrolysis and ATP-dependent protein degradation (Supplementary Figure 2C-D).

Structural features of an asymmetric FtsH•HflK/C complex.

We refined a ~4.4 Å resolution cryo-EM structure of the DDM-solubilized complex without imposed symmetry, revealing an asymmetric HflK/C complex resembling a nautilus shell with two internal FtsH hexamers (Figure 1A-C; Supplementary Figures 3-4). A similar cryo-EM structure, albeit at lower resolution, was obtained when the complex was solubilized with the detergent GDN (Supplementary Figures 5), indicating that the nautilus-like structure was not unique to a specific detergent. Moreover, this nautilus-like structure was not the result of preferential orientations or preferential positioning at the air-water interface, as the 3D-FSC (Tan et al., 2017) measure of sphericity was 0.89 with diverse projection angles observed (Supplementary Figure 4C-D). Furthermore, tomographic reconstructions revealed random particle orientations on the EM grid (Supplementary Figure 6).

HflK and HflC share sequence and structural homology and were difficult to distinguish in the 4.4-Å density map of the DDM-solubilized complex. However, after signal subtraction and local refinement in C1 (see Methods), we refined the periplasmic portion of the nautilus structure farthest from the membrane, which resembles a hat, to 3.5-Å resolution (Supplementary Figure 7-8). This structure allowed visualization of side chains and unambiguous assignment of ~110 C-terminal residues of 24 alternating HflK and HflC subunits (Figure 1D). The corresponding hat region of the C4-symmetric HflK/C structure (pdb 7WI3) (Qiao et al., 2022) aligned to the asymmetric nautilus hat with an RMSD of 2.3 Å for the 2,545 common Cα positions. SPFH domains, present in each HflK and HflC subunit, formed the membrane-proximal sides of the nautilus chamber, but only 21 of 24 these domains could be modeled (Figures 1-2). The entryway into the nautilus chamber was located near the ‘missing’ SPFH domains, which presumably adopted multiple conformations or were proteolytically truncated. The final model, which had good geometry (Table 1), included residues 79-355 for most HflK subunits, residues 1-160 and 191-329 for most HflC subunits, and residues 31-97 of the periplasmic domains of two FtsH hexamers.

The nautilus chamber in our structure contained two hexamers composed of FtsH periplasmic domains, with one hexamer better resolved than the other. In their closed C4-symmetric structure, Qiao et al. identified contacts between Arg¹⁹¹ of HflK and residues 62-64 of the periplasmic domain of FtsH. Although the resolution of our asymmetric nautilus structure was not sufficient to confidently assign these exact contacts, the backbones in both structures were positioned similarly (Figures 1B, 2). Because the tightly closed HflK/C cage in the C4-symmetric structure (Qiao et al., 2022) contained four FtsH hexamers, whereas our open nautilus chamber contained fewer FtsH hexamers, we hypothesized that chamber curvature might be proportional to the number of internal FtsH hexamers. To test this hypothesis, we performed focused classification, using a mask encompassing the periplasmic hexamer of FtsH with lower occupancy (Supplementary Figure 3). After 3D-classification and refinement of the classified particles, assemblies with either one or two FtsH hexamers were evident (Supplementary Figure 9-10). Comparison of these maps (Figure 2A-B) revealed no significant structural rearrangements in the nautilus shell, indicating that chamber curvature does not simply increase with the number of enclosed FtsH hexamers. Moreover, when we overproduced HflK and HflC prior to affinity purification via a FLAG-tag on HflC, the resulting cryo-EM structure (Figure 2C, Supplementary Figures 11-12) had a nautilus-like HflK/C chamber lacking FtsH density, but with curvature similar to our structures with one or two FtsH hexamers. In each instance, our maps were clearly distinct from the C4-symmertrized structure from Qiao et al. (Figure 2D).

We next searched for structural heterogeneity in our DDM-solubilized FtsH•HflK/C dataset using cryoDRGN (Kinman et al., 2023; Sun et al., 2023; Zhong et al., 2021). This analysis revealed the expected nautilus-like structures with either one or two FtsH hexamers and additional differences in the width of the entryway into the nautilus chamber (Supplementary Movie 1), consistent with flexibility in the SPFH domains of the terminal HflK/C subunits. Notably, however, closed HflK/C structures or complexes with four FtsH hexamers were not observed.

Unusual curvature of the membrane domain of FtsH•HflK/C.

The density map of DDM-solubilized FtsH•HflK/C showed the approximate positions of lipid head groups in the bilayer (Supplementary Figure 4), but individual lipid molecules could not be resolved or modeled. Notably, however, the bilayer curved inward within the complex, whereas the bacterial inner membrane as a whole curves outward. This unusual membrane geometry was not a function of detergent solubilization of the complex, as similar curvature was observed in a lower-resolution (~12Å) 3D reconstruction of the affinity purified FtsH•HflK/C complex following extraction with the nanodisc-forming polymer Carboxy-DIBMA, and native lipids (Figure 3A, Supplementary Figures 13-14). Notably, a 2D class average side view of these particles showed pronounced inward membrane curvature within the complex and reversal of this curvature as the membrane extended beyond the embedded FtsH•HflK/C complex (Figure 3B). Density for the cytoplasmic portions of both FtsH hexamers in the Carboxy-DIBMA map was far stronger than in the higher-resolution map of the DDM-solubilized complex, which may reflect the presence of native lipids or other differences in the protocols used to purify the complexes.

FtsH•HflK/C enhances lipid scrambling.

Unusual membrane curvature often correlates with enhanced rates of polar lipid head group flipping from one side of a bilayer to the other (Devaux et al., 2008; Jarsch et al., 2016). This activity can be assayed by reconstituting proteoliposomes with fluorescent NBD-labeled lipids (Ghanbarpour et al, 2021; Ploier & Menon, 2016), which orient randomly in proteoliposomes, with approximately half of the fluorescent lipids initially in the inner leaflet and half in the outer leaflet. Treatment with dithionite, which does not penetrate liposomes, quenches the fluorescence of outward-facing NBD groups, thereby reducing the fluorescence to ~50% of the initial level (Figure 4A, left). If proteins that catalyze lipid scrambling are present in the membrane, dithionite treatment can further reduce fluorescence, as NBD-labeled lipids initially facing the lumen can be translocated to the outside, where they are exposed to dithionite quenching (Figure 4A, right). Using this assay, we found that FtsH•HflK/C and FtsH catalyze lipid scrambling (Figure 4B-C). Although this activity did not require ATP hydrolysis by FtsH, we postulate that it may help this ATP-fueled protease to extract membrane proteins from the lipid bilayer during degradation (see Discussion).

HflK/C impacts intracellular protein levels.

To investigate how HflK/C affects E. coli protein levels, we used mass spectrometry to quantify the steady-state abundance of ~2,250 proteins in wild-type cells or mutant cells lacking HflK/C (ΔhflK/C) (Figure 5). Of these proteins, ~88% showed no significant differences, ~8% were more abundant in ΔhflK/C cells, and ~4% were less abundant in these mutant cells. We considered that FtsH•HflK/C could degrade proteins in the ‘more abundant’ category faster than FtsH alone, and that HflK/C could restrict degradation of proteins in the ‘less abundant’ category. However, other possibilities include indirect effects of hflK/C genetic deletion, such as an impact on FtsH-driven degradation of proteins that regulate the transcription, translation, or turnover of these differentially abundant proteins. Notably, proteins in the ‘more abundant’ category included the membrane proteins DadA, PspC, SecY, and YlaC, which are known to be degraded by FtsH (Arends et al., 2016; Singh and Darwin, 2011; Westphal et al., 2012). To identify proteins that directly interact with FtsH, we affinity-purified FtsH from wild-type or ΔhflK/C cells, and analyzed the eluate by mass spectrometry. Remarkably, some of the differentially abundant proteins identified above, including DadA and AstC, also exhibited HflK/C-dependent association with FtsH (Extended Data Table 1).

FtsH•HflK/C purification.

A C-terminal FLAG-tagged variant of FtsH was introduced into E. coli BL21(DE3) at the endogenous FtsH locus using the no-SCAR (Scarless Cas9 Assisted Recombineering) system (Reisch & Prather, 2017). Briefly, the spacer sequence GTCGCCTAACTGCTCTGACA targeting the C-terminus of FtsH was closed into the temperature sensitive pKDsg-XXX plasmid, and introduced into BL21(DE3) cells containing the pCas9cr4 plasmid. λ-Red recombineering (Datsenko & Wanner, 2000) was induced, followed by transformation with a 500 bp double-stranded gene block (IDT) containing a mutation in the PAM sequence that appended the sequence coding for “SGSGDYKDDDK” to the C-terminus of FtsH. The chromosomal mutation was selected for by inducing Cas9 and subsequently verified by amplifying the region using PCR and DNA sequencing. BL21(DE3) cells expressing this C-terminal FLAG-tagged FtsH were grown overnight in 50 mL of Luria broth (LB), transferred to a 1:1 LB:2xYT medium mixture, and grown for an additional 20 h to an A₆₀₀ of ~5.5. Cells were harvested by centrifugation at 4,500 rpm for 30 min and resuspended in buffer A [20mM HEPES (pH 7.5), 400mM NaCl, 100mM KCl, 10% glycerol, 1mM DTT] before flash freezing in liquid N₂ for storage at −80°C.

For cell lysis, resuspended cells were sonicated (QSonica) on ice using 30% power with cycles of 10 s on and 30 s off for a total duration of 4 min. Following sonication, cells were centrifuged at 12,100 x g for 20 min in a JL25.50 rotor. The supernatant was isolated and then centrifuged at 105,000 x g for 1 h using a 45 Ti rotor. The pelleted membrane was resuspended in 5 mL of buffer A and disrupted using a Dounce homogenizer. Detergent n-dodecyl-β-D-maltoside (DDM) was added to a final concentration of 1%, followed by platform rotation at 4°C for 1 h. The mixture was centrifuged for 20 min at 35,000 rpm using a TLA 100.3 rotor, incubated with anti-Flag M2 resin overnight, washed with a buffer containing 0.03% DDM, and finally eluted with 0.25 mg/mL Flag peptide (APEXBio) according to the manufacturer’s directions. The same protocol was used for glyco-diosgenin (GDN) sample preparation, except 2% GDN was used for membrane solubilization and 0.03% GDN was employed for the wash and elution steps.

The overexpressed HflK/C sample was prepared using the petDUET plasmid as previously reported (Ma et al., 2022). Briefly, HflC followed by a C-terminal Flag-tag was cloned into the first multicloning site, and HflK with an N-terminal octa-histidine tag was cloned into the second site. BL21 (DE3) cells harboring the plasmid were grown at 37°C from an overnight cell culture to an OD of ~1. Expression was induced by the addition of 0.2mM IPTG and continued overnight at 18°C. Purification was performed similarly to the procedure described for the endogenously expressed FtsH-HflK/C solubilized in DDM.

Carboxy-DIBMA extraction.

Carboxy-DIBMA was synthesized analogously to Sulfo-DIBMA as described (Glueck et al., 2022). Membranes, isolated as described above, were resuspended in buffer B [50mM Tris-HCl (pH 8.0), 100mM NaCl, 5mM MgSO₄, 100μM ZnCl₂, and 3mM β-mercaptoethanol], to a final concentration of approximately 35 mg/mL. To this suspension, 22 mg/mL of Carboxy-DIBMA solution and 1 mM ATP were added, bringing the total volume to 5 mL. The mixture was gently rotated overnight at 4°C to facilitate solubilization. Subsequently, samples were centrifuged at 30,000 x g, yielding a clear supernatant. This supernatant was incubated with anti-FLAG M2 magnetic beads for 2 h at 4°C with gentle rotation. After incubation, the sample was centrifuged at 300 x g, the supernatant was carefully removed with a magnet, and beads were washed with 15 mL of buffer B. Finally, the sample was eluted using 0.5 mg/mL FLAG peptide. The eluted sample was concentrated, analyzed, and quantified after SDS-PAGE, using bovine serum albumin as a standard.

Cryo-EM sample preparation.

DDM- or GDN-solubilized samples were prepared by applying 2.5 μL of ~0.75 mg/mL of FtsH•HflK/C complex onto homemade graphene-supported 200-mesh Quantifoil 2/1 copper grids (Grassetti et al., 2023), pre-treated with UV/ozone for 7 min using a Bioforce UV/ozone cleaner. Sample-loaded grids were blotted for 4 s with a blot force of +4 at 6°C and 95% relative humidity using a FEI Vitrobot Mk IV instrument. The same concentration of the overexpressed HflK/C sample was used for grid preparation. However, in this case, 2-nm carbon-supported 200-mesh Quantifoil 2/1 copper grids, which had been glow-discharged for 20 s in an easiGlow glow discharger (Pelco) at 15 mA, were utilized. For the Carboxy-DIBMA-extracted sample, a concentration of ~0.25 mg/mL of the FtsH•HflK/C complex was applied to 2-nm carbon-supported 200-mesh Quantifoil 2/1 copper grids, also glow-discharged for 20 s in an easiGlow glow discharger (Pelco) at 15 mA, and the sample was vitrified as above.

Cryo-EM data collection.

For the DDM-solubilized FtsH•HflK/C complex, 25,179 movies were collected with EPU using aberration-free image shift (AFIS) and hole-clustering method on a Titan Krios G3i with an acceleration voltage of 300 kV and magnification of 130,000x, detected in super-resolution mode on a Gatan K3 detector for an effective pixel size of 0.654 Å (binned by 2). Movies were collected as 40 frames with a defocus range from −0.05 to −1.75 μm and a total exposure per specimen of 46.1 e⁻/Ų. For the DDM-solubilized and overexpressed HflK/C complex, movies were collected as 40 frames with a defocus range from −0.5 to −1.75 μm and a total exposure per specimen of 47.2 e⁻/Ų using AFIS hole clustering. For the Carboxy-DIBMA extracted sample, movies (11,989) were collected as 40 frames with a defocus range from −0.30 to −1.75 μm and a total exposure per specimen of 44.6 e⁻/Ų using AFIS hole clustering. An additional data collection session on this sample produced 32,907 movies, collected with a defocus range from −0.75 to −2.0 μm and a total exposure per specimen of 44.4 e⁻/Ų using AFIS hole clustering.

Cryo-EM pre-processing and particle picking.

For the DDM-solubilized FtsH•HflK/C complex, data processing was performed using cryoSPARC (v4.0) and default parameters unless noted (see Supplementary Figure 3). Raw movies (25,179) were pre-processed using ‘Patch motion correction’, and ‘Patch CTF estimation’. Particles (73,171) were picked using the ‘Blob-picker’ tool applied to 1,000 micrographs selected at random. Particles were extracted (box size 640px, Fourier cropped to 360px) and classified using the ‘2D classification’ utility. The entire set of micrographs were then picked with the ‘Template picker’ tool, using a single class from 2D classification. Particles were extracted (box size 640px, Fourier-cropped to 360px) and subjected to three rounds of 2D classification, resulting in the selection of 334,256 particles as a preliminary stack.

For the DDM-solubilized overexpressed HflK/C complex, data processing was performed using cryoSPARC (v4.0) and default parameters unless noted (see Supplementary Figure 11). Raw movies (16,055) were pre-processed and particles picked as described above, except that the extraction-box-size was 900px (Fourier cropped to 300px). After four rounds of 2D classification, 295,185 particles were selected as a preliminary stack.

For the FtsH•HflK/C complex extracted with Carboxy-DIBMA, data processing was carried out using cryoSPARC (v4.0) with default settings unless noted (see Supplementary Figure 13). Raw movies were collected over two sessions. Data from these sessions were pre-processed individually using ‘Patch motion correction’ and ‘Patch CTF estimation’. Particles were separately extracted (box size of 1,000px, Fourier cropped to 360px). After several rounds of 2D classification of particles from the isolated datasets, the particles were pooled and subjected to two additional rounds of 2D classification, resulting in the final particle stack of 47,449 particles.

Ab initio reconstruction, global refinement, and model building.

For the FtsH•HflK/C complex solubilized in DDM, ‘Ab-initio’ reconstruction using two classes was performed, with a resulting single class model and particle stack (184,019 particles) selected for subsequent ‘Homogenous refinement’ followed by ‘Non-uniform refinement’. A mask was then created in ChimeraX (Pettersen et al., 2021) that only included the periplasmic portion of HflK/C and FtsH, excluding the cytoplasmic domain of FtsH, which exhibited substantial flexibility. Using this mask, ‘Local refinement’ was performed, leading to a ~4.4 Å GS-FSC resolution map we termed “map a”. To separate single vs. double FtsH hexamers within the HflK/C assembly, a mask was generated that encompassed the FtsH periplasmic domain with a lower occupancy. Using this mask, ‘3D classification’ (20 classes) was performed. Based on visual inspection, particles assigned to maps bearing one FtsH hexamer were pooled and subjected to ‘Local refinement’ using the original mask, producing “map I”. A parallel operation was performed on particles assigned to maps bearing two hexamers, resulting in “map II”.

To generate a higher-resolution map for the ‘hat’ portion of the complex a signal subtraction and local refinement strategy was employed (see Supplementary Figure 7). Briefly, two masks were generated using map a - one encompassing on the hat portion, and another containing the remainder of map a, which was used for signal subtraction after first locally aligning the particles using the non-hat mask using the ‘Local refinement’ utility. Next, the signal-subtracted particles and the mask around the hat portion were used to generate an initial volume using the ‘Homogeneous reconstruct only’ utility followed by a ‘Local refinement’ using the same mask. After application of the ‘Global CTF refinement’ followed by ‘Local CTF refinement’, the final map reached ~3.5 Å GS-FSC resolution.

For the overexpressed HflK/C complex solubilized in DDM, ‘Ab-initio reconstruction’ using three classes was performed, and one class was selected (106,858 particles). ‘Homogeneous refinement’ followed by ‘Non-uniform refinement’ were then employed, producing a map that was used as an initial model (after low-pass filtering) in a final round of ‘Non-uniform refinement’ followed by ‘Local refinement’ with the original particle stack (230,567 particles). After conducting both global and local CTF refinements, the final map achieved a GS-FSC resolution of ~9.8 Å.

For the FtsH•HflK/C complex extracted with Carboxy-DIBMA, ‘Ab-initio reconstruction’ was carried out on the 47,449 particles using two classes. Of these, 24,011 particles were selected for ‘Non-uniform refinement’, followed by global and local CTF refinement, and an additional round of ‘Local refinement.

Model building was performed using a combination of ChimeraX-1.3 (Pettersen et al., 2021), Coot (0.9.4) (Casanal et al., 2020), and Phenix (1.14) (Liebschner et al., 2019). Local resolution was estimated by cryoSPARC implementation of monoRes (Vilas et al., 2018).

Tilt-series acquisition.

The GDN-solubilized FtsH•HflK/C sample was used for tomographic data collection. A total of 18 tilt series were acquired on a Titan Krios G3i microscope operating at 300kV. Movies were collected at a nominal magnification of 64,000x on a BioQuantum K3 detector operating in super-resolution mode, and were pre-binned at collection time to a physical pixel size of 1.36 Å, with 6 frames per movie. Tilt series were collected following a dose-symmetric scheme from −58° to 58° in 2° increments, beginning from 0°, and each tilt series was collected at a nominal defocus stepped between −3 μm and −5 μm. The total electron dose per tilt series was 117.2 e⁻/Ų. The energy-filter slit width was 20 eV during acquisition, and the zero-loss peak was refined between each tilt movie acquisition.

Tomogram reconstruction and sub-tomogram averaging.

Tomographic data processing generally followed Powell et al., 2024, with the following modifications. Tilt movies were aligned and initial CTF parameters estimated in Warp (v1.0.9) (Tegunov and Cramer, 2019). Tilt-series stacks were exported and aligned in IMOD (v4.11.12) using patch tracking (Mastronarde and Held, 2017). Tilt series were re-imported into Warp with alignment parameters and CTF parameters were fit using each full-tilt series. Tomograms were reconstructed and deconvolved in Warp using a pixel size of 10 Å/px. All particles were manually picked using Cube (n=1,578 particles). Particles were exported from Warp with pixel size of 8 Å/px and box size of 64 px³. RELION (v3.1.3) (Zivanov et al., 2018) was used for initial model generation and subsequent particle refinement with C1 symmetry, reaching an unmasked resolution of 24.4 Å. Refined particles were imported to M (v1.0.9) (Tegunov et al., 2021) and subjected to iterative refinement of image warp, volume warp, particle pose, and stage angle parameters, ultimately reaching a masked resolution of 17.9 Å. Particles were re-extracted from M with pixel size of 8 Å/px and a box size of 64 px³, re-refined in RELION, re-imported to M, and re-refined in M, reaching a final resolution of 17.3 Å. Copies of this final reconstruction were mapped back in the source tomogram and visualized with ChimeraX using the subtomo2chimerax functionality implemented in tomoDRGN (Powell and Davis, 2024).

Mass spectrometry.

For steady-state protein comparisons between wild-type and ΔhflK/C E. coli strains, cells were grown in LB and harvested via centrifugation at 4°C at an A₆₀₀ of ~1. Cell pellets were frozen at −80°C for future use. Thawed cell pellets were washed with buffer MS1 [20mM Tris-HCl (pH 7.5), 100mM NH₄Cl, 10mM MgCl₂, 0.5mM EDTA, and 6mM β-mercaptoethanol], followed by the addition of SDS (10% final concentration) and 0.25 U/μL benzonase (EMD Millipore). After incubation for 10 min at room temperature, lysates were cleared via centrifugation at 21,000 x g for 8 min. DTT (final concentration of 20 mM) was introduced to the lysate, followed by heating at 65°C to reduce disulfide bonds. After cooling to 30°C, iodoacetamide (final concentration of 40 mM) was added for 30 mins to alkylate free cysteines. Subsequent acidification, column preparation and sample binding to S-Trap^™ micro (≤ 100 μg) spin columns and column washing followed the manufacturer’s protocol (ProtiFi). A trypsin solution (Promega Inc.) was combined with 200 μL of digestion buffer and applied to the spin columns containing bound proteins followed by a 1 h incubation at 42°C. Peptides were eluted using a sequential addition of 50 mM TAEB and 0.2% (v/v) formic acid. Further elution with acetonitrile-formic acid mix (35μL of 50% acetonitrile, 0.2% formic acid) helped maximize recovery of hydrophobic peptides. Finally, peptides were dried in a speed-vac and resuspended in buffer MS2 [4% acetonitrile, 0.1% formic acid] for subsequent mass spectrometry analysis.

Peptides (~1.5μg) were injected into an Ultimate 3000 UHPLC system (Thermo Scientific) comprising a PepMap100 C18 (75μm ID x 2cm, particle size 3μm, pore size 100A; Thermo Fisher Scientific) precolumn upstream of a PepMap C18 (75μm ID x 2cm, particle size 3μm, pore size 100A; Thermo Fisher Scientific) analytical column. Sample loading proceeded at 300 nL/min in 98% buffer LCMS_A [0.1% formic acid in water], 2% LCMS_B [0.1% formic acid in acetonitrile]. After washing the precolumn with 20 column volumes of LCMS_A, peptides were resolved using a 120-min linear gradient from 4% to 30% LCMS_B. Peptides were ionized by nanospray electrospray ionization and analyzed on a Q Exactive HF-X mass spectrometer (Thermo Scientific). For data-dependent acquisitions, MS1 full scans (350-1400 m/z range) were collected at a resolution of 60,000 (AGC 3x10⁶, 50ms maximum injection time). HCD fragmentation of the 12 most abundant precursor ions was performed at 25% NCE. The fragment analysis (MS2) was performed with a resolution of 15,000 (AGC 1x10⁵, 100ms maximum injection time, 2.2 m/z isolation window). For data-independent acquisitions, MS1 full scans (350-1400 m/z range) were collected at a resolution of 120,000 (AGC 3x10⁶, 50ms maximum injection time). Fragment analysis (MS2) was subdivided into 25 DIA isolation windows of varying widths (see Supplementary Table 1) using a resolution of 30,000 (AGC 1x10⁶, 70ms maximum injection time). HCD fragmentation was performed at 25% NCE. The resulting raw DIA files were analyzed using Spectronaut (v15), with a search database containing the E. coli Uniprot database concatenated to a database containing common contaminants. In silico digests were performed using Trypsin/P and LysC. The false discovery rate was set at 1% at both the peptide and protein levels. Normalization of all replicates and generation of the volcano plot were performed using Spectronaut.

For the preparation of samples for mass spectrometric analysis of the pull-down of FtsH from wild-type and ΔhflK/C strains, the same purification protocol for the FtsH•HflK/C complex describe above was followed, using GDN detergent to retain weak interactions between FtsH and its substrates. After running an SDS-PAGE gel of the FLAG resin eluate from WT and ΔhflK/C cells, the entire gel lane for each condition was excised and sent to the proteomic core facility at the Koch Institute for Cancer Research (MIT) for data analysis. There, samples were analyzed via DDA on a Q Exactive HF-X mass spectrometer coupled to a Ultimate 3000 UHPLC. Co-purifying proteins in each condition were quantified using spectral counts, normalized to that of FtsH in each sample.

Biochemical assays.

Expression and purification of FtsH-bearing samples followed protocols listed above, using the following buffer substitutions. Buffer A was substituted with buffer BA1 [50mM Tris-HCl (pH 8.0), 1mM MgCl₂, 100mM KCl, 10% glycerol, and 3mM β-mercaptoethanol] during cell lysis, and with buffer BA2 [50mM Tris-HCl (pH 8.0), 100mM NaCl, 5mM MgCl₂, 100μM ZnCl₂, 10% glycerol, 3mM β-mercaptoethanol] when resuspending the pelleted membranes. Additionally, prior to elution, the resin was washed using buffer BA3 [50mM Tris-HCl (pH 8.0), 400mM NaCl, 5mM MgCl₂, 100μM ZnCl₂, 0.03% DDM and 10% glycerol]. The protein was eluted with 0.25 mg/mL FLAG peptide (APEXBio) in buffer BA5 [50mM Tris-HCl (pH 8.0), 100mM NaCl, 5mM MgCl₂, 100μM ZnCl₂, 10% glycerol, and 0.03% DDM] according to the manufacturer’s directions.

The λ-CII substrate was produced using HisTag-SUMO-λ-CII in the pET22b plasmid. The BL21(DE3) cells harboring the plasmid were grown to an OD₆₀₀ of ~0.8 at 37°C. Protein expression was induced by addition of 0.5mM of IPTG, and the culture was grown overnight at 18°C. Cells were harvested by centrifuging at 4,000rpm for 25 min at 4°C. The cell pellet was resuspended in buffer BA6 [20mM HEPES (pH 8.0), 400mM NaCl, 100mM KCl, and 10% glycerol], and stored at −80°C.

For purification, the cells were thawed and incubated with 1 mM DTT and protease inhibitor cocktail prior to lysis using 30% power with cycles of 10 s on and 30 s off for a total duration of 3.5 min. The sample was centrifuged at 4,500 rpm for 30 min in a JA25.50 rotor to remove cell debris. The supernatant was transferred to a new tube and incubated with pre-washed Ni-NTA resin for 30 minutes at 4°C in buffer BA6 supplemented with 5mM imidazole. The resin was washed with 200 mL of buffer BA6 supplemented with 15mM imidazole to remove impurities. The resin was then incubated with the Sumo Protease overnight at 4°C in buffer BA6 containing 5mM imidazole. The next day, the cleaved protein was collected as HisTag-SUMO remained bound to the resin. The cleaved protein was concentrated using a 10-kDa cutoff centrifugal unit and loaded onto a S200 (16/600) gel filtration column. Fractions were collected, analyzed using SDS-PAGE, concentrated, aliquoted, and flash-frozen in liquid nitrogen for future use.

Degradation assays were performed at 37 °C using 5μM λ-CII, ~0.3μM of FtsH•HflK/C complex in buffer BA7 [25mM HEPES (pH 7.5), 150mM NaCl, 5mM MgCl₂, 3mM β-mercaptoethanol,100μM ZnCl₂, 5mM ATP, 0.009% DDM]. The concentrations of FtsH/HflK/C were calculated using a BSA standard after running SDS PAGE. All components except β-CII were initially mixed at 37°C for 5 min. λ-CII, which had also been preincubated at 37°C, was then added to initiate degradation, which was monitored by SDS-PAGE gel analysis.

ATP hydrolysis was measured using a coupled enzymatic reaction (Norby, 1988) in which NADH oxidation to NAD+ reduces absorbance at 340 nm (Δε = 6.22 mM⁻¹cm⁻¹) using a SpectraMax M5 plate reader and a 384-well assay plate (Corning, 3575). A 20X stock ATPase reaction mix contained 20μL of pyruvate kinase and lactic dehydrogenase from rabbit muscle (P0294, Sigma Aldrich), 10μL of 200mM NADH, grade II (CAS# 606688), 15μL of 1M phosphoenolpyruvate (Sigma Aldrich, 10108294001) in 25mM HEPES-KOH (pH 7.6), and 25μL of 200mM ATP. For assays, FtsH₆ (0.5 μM) was present in 10μL of buffer BA7. After 5 min of equilibration at 37°C, the ATPase assay was initiated by addition of an equal volume of 1X ATPase reaction mix.

Lipid preparation.

All lipids, including POPC (#850457C), POPE (#850757C), and NBD-PE (#810153C), were purchased from Avanti lipids, and liposomes were prepared as described (Ghanbarpour et al., 2021, Ploier and Menton, 2016, Marek and Gunther-Pomorski, 2016)). Briefly, a mixture of POPC (90% of total lipid by weight), POPE (9.5%), and NBD-lipid (0.5%, either NBD-PE or NBD-PC) solubilized in chloroform was mixed and dried under a nitrogen stream followed by vacuum to form a lipid film. The dry lipid film was subsequently resuspended in buffer LP1 [50mM HEPES (pH 7.6), 200mM NaCl] to create a 10.5mM lipid stock, which was incubated at 37°C for 10 min, followed by resuspension and ten freeze-thaw cycles. Finally, the lipid mixture underwent 30 cycles of extrusion through a 400-nm polycarbonate filter to complete the preparation.

Proteoliposome preparation.

Liposomes, with a lipid concentration of 5.25mM in a total volume of 250μL, were subjected to destabilization by adding Triton X-100. The concentration of Triton X-100 used was determined based on a prior swelling assay (Ghanbarpour et al. 2021). This destabilization process was carried out at room temperature for 2-3 h. Proteins solubilized in detergent were then introduced to the destabilized liposomes, and the mixture was rotated gently for 1 h to ensure mixing. Removal of detergent was achieved in three stages using pre-washed biobeads (Bio-Rad #152-3920). Initially, 20mg of biobeads were added to the mixture, which was then incubated at room temperature for 1 h. Subsequently, 20mg of additional biobeads were added, and the mixture was allowed to rotate at room temperature for 2 h. In the final step, the mixture was transferred into a new tube containing 40mg of fresh biobeads and rotated at 4°C overnight to complete detergent removal, prior to removing biobeads by pipetting.

Scramblase assay.

Lipid scramblase activity was monitored at 30°C using 96-well plates containing 100 μL of either liposomes or proteoliposomes. The assay tracked the decrease in NBD fluorescence (excitation 460 nm; emission 538 nm) upon the introduction of dithionite to a final concentration of 5mM using a BioTek Synergy H1 Hybrid Multi-Mode Reader. Subsequently, an additional dose of dithionite (5mM) along with Triton X-100 (0.5%) was added to dissolve liposomes and ensure NBD reduction.

To test for proteolipid disruption, a 1-myristoyl-2-C6-NBD-PE bovine serum albumin (BSA) back-extraction assay (Ghanbarpour et al., 2021; Marek and Gunther-Pomorski, 2016; Ploier and Menon, 2016) was carried out adding 3mg/mL BSA to liposomes or proteoliposomes instead of dithionite. Like dithionite, BSA interacts exclusively with NBD-PE from the outer leaflet of liposomes, but BSA quenches fluorescence less effectively than dithionite. The protocol used above was used for the NBD-glucose assay, except NBD-lipids were excluded from liposomes or proteoliposomes and NBD-glucose at a concentration of 12.6 μM was introduced during the liposome-destabilization phase. Glucose leakage was assessed after 2 days of dialysis against buffer free of NBD-glucose.