PCAT1 contains two identical PEP domains, raising the question of whether two copies of the substrate can bind and be translocated simultaneously. To address this question, we used native MS to determine the maximum binding capacity of PCAT1.

First, we tested if the PCAT1 homodimer remains intact during native MS analysis in two different detergents: n-undecyl-β-D-maltopyranoside (UDM) and octyl tetraethylene glycol ether (C₈E₄). Previous structural and biochemical characterization of the PCAT1 transporter was performed in UDM (Lin et al., 2015). C₈E₄ has been shown to require the lowest activation energy for detergent removal in the gas phase with minimal destabilization of membrane protein complexes during MS analysis (Reading et al., 2015). The mass spectra obtained in both detergents at optimal MS parameters (Figure 1—figure supplement 1) yielded a single peak series with a measured mass consistent with the mass of dimeric PCAT1 (Figure 1A and Table 1).

To obtain a stable PCAT1-CtA complex, we used a proteolytic-deficient mutant (C21A) that can bind but does not cleave the substrate (Lin et al., 2015). Two-fold molar excess of the substrate was mixed with PCAT1 prior to MS analysis. In both detergents, we observed two main assemblies of 172.3 and 182.5 kDa, corresponding to dimeric PCAT1 bound to one or two CtA, respectively (Figure 1B, Table 1). Based on the relative peak intensities with correction for gas-phase dissociation, the PCAT1 dimer with two CtA bound (1:2 complex) is the dominant assembly in the presence of excess CtA (75 ± 2% in UDM and 84 ± 2% in C₈E₄) (Figure 1—figure supplement 2 and Table 1). The lower abundance of the 1:2 complex in UDM is likely due to the higher collision activation energies needed to remove UDM that lead to increased protein complex dissociation (Figure 1—figure supplements 1 and 2).

Next, we determined the cryo-EM structure of the PCAT1(C21A)-CtA complex in the presence of excess CtA and in the absence of ATP. The overall resolution is approximately 3.4 Å (Figure 2, Table 2, Figure 2—figure supplements 1–4). Side-chain densities are prominent in the TMDs, allowing unambiguous residue assignment (Figure 2—figure supplement 3). The density of the PEPs and NBDs show clear definition of secondary structure, permitting docking of the crystal structures of isolated PEP and NBD. Extra densities, unaccounted by PCAT1, are found on the surface of both PEP domains, likely corresponding to the N-terminal region of CtA (Figure 2—figure supplement 3). The final model, containing residues 8–722 of PCAT1 and two leader peptides of CtA, was refined against the EM density to excellent geometry and statistics (Table 2).

Consistent with the native MS analysis, two copies of CtA were bound to a single PCAT1 transporter (Figure 2). In this ATP-free, substrate-bound conformation, the TM cavity is continuous with the cytosol and closed off to the extracellular space. The two PEP domains, each binding a CtA molecule, interact with the transporter core at the TMD/NBD interface. The structure is two-fold symmetric, with the exception of the C-terminal regions of the substrate, a point that we will discuss in detail below.

The sequence of the CtA leader peptide is homologous to that of other PCAT substrates in Gram-positive bacteria (Figure 3A). The cleavage site contains a consensus sequence of L(−12)XXXE(−8)L(−7)XXXXG(−2)G(−1) (Figure 3A). Residues 8–25 of the leader peptide form a L-shaped structure that wraps around the PEP domain, with the N-terminus in the cytosol and the C-terminal end at the lateral opening of the translocation pathway (Figure 3B and C). Residues 15–21 (i.e, positions −10 to −4 of the consensus sequence) form a short helix, docking onto a hydrophobic groove on PEP (Figures 3B and 4A). The double-glycine motif (G23 and G24), inserted to the active site of PEP, interacts with catalytic residues C21A and H99 (Figure 3C). The total buried surface of the leader peptide by the PEP domain is approximately 740 Ų.

The substrate-binding cleft at the cleavage site is very narrow, which explains the specificity for the double-glycine motif in proteolysis. The Cβ atom of C21 (mutated to alanine) is about 5.5 Å away from G24, consistent with C21 serving as a nucleophile to attack the substrate backbone (Figure 3C, Figure 3—figure supplement 1) (Schechter and Berger, 1967). H99 orients the N atom of its imidazole ring towards the C21A, consistent with its role to polarize the attacking C21 (Husain and Lowe, 1968; Wu and Tai, 2004). The side chain of D115 forms a hydrogen bond with H99, maintaining H99 in an electronegative state and a catalytically favorable position (Vernet et al., 1995). In addition, the highly conserved Q15 residue is located in the vicinity, suggesting that it may function as an oxyanion hole that stabilizes the tetrahedral intermediate (Ménard et al., 1991). The configuration of the active site is typical of a cysteine protease, indicating that PCAT1 processes the CtA substrate via the common catalytic thiol mechanism (Drenth et al., 1968; Kamphuis et al., 1984).

The structure of the PEP domain is highly conserved among different PCATs. The overall root-mean-square deviations (RMSDs) among the structures of ComA (Ishii et al., 2010) (PDB: 3K8U), LahT147 (Bobeica et al., 2019) (PDB: 6MPZ), and the PEP domain of PCAT1 are approximately 1 Å (Figure 3—figure supplement 2). In addition, substrates of the different PEPs share a consensus leader peptide sequence (Figure 3A), yet each PEP specifically recognizes one or a small subset of substrates.

What are the common substrate recognition motifs and what are the unique features that confer specificity? The leader peptides of CtA and the LahT147 substrate both contain a two-turn helix consisting of residues from positions −10 to −4, docking into a shallow groove on the surface of the PEP (Figure 4A and B). A hydrophobic knot, formed by the two leucine residues at positions −7 and −12 of the leader peptide and two hydrophobic residues on PEP, is observed in both structures (Figure 4B). To test the importance of this interaction, we mutated the conserved leucine residues in CtA and estimated its affinity for PCAT1(C21A) using a pull-down assay (Figure 4D). Compared to the wt CtA, the L(−7)A and L(−12)A mutants have much lower affinity for PCAT1. The pulled down PCAT1 could only be detected by Western blot even at the highest concentration of CtA tested (Figure 4D). We found hydrophobic residues A55 and I59 on PCAT1 that are within the Van der Waals radius of L(−7)A and L(−12)A, and are conserved in LahT147 as well (Figure 4B). Furthermore, reciprocal mutations introduced to residues A55 and I59 of PCAT1 lowered its affinity for CtA (Figure 4E) by more than eight fold, consistent with the role that these residues play on the CtA recognition. This hydrophobic interaction is likely a common feature shared by different PEP domains. Alanine substitution of either one of the two leucine residues in the ComA substrate decreased its affinity by approximately twenty fold (Kotake et al., 2008). Similarly, LahT147 does not process precursor peptides with substitutions at L(−7) or L(−12) (Bobeica et al., 2019).

Mutational data at the −4 position indicate that this residue is also important for the PEP-substrate interaction (Figure 4D and Kotake et al., 2008); however, in contrast to the highly conserved L(−7) and L(−12) residues, the amino acid identity at position −4 varies among different substrates (Figure 3A and C) and appears to correlate with the interacting residue on PEP. For example, M (−4) of CtA packs against G135 of PCAT1 (Figure 4C). In LahT147, a much smaller residue V(−4) of the substrate interacts with a larger residue I137 on PEP (Figure 4C). Therefore, it seems that the leader peptide position −4 confers specificity among different PCATs.

In contrast to the leader peptide, the cargo region (residues 25–90) is flexible and less defined. Inside the translocation pathway, we observe density of elongated shape packed along the interior surface of the cavity shown in the composite cryo-EM map of PCAT1 and CtA (Figure 5A, colored blue). To test if this density corresponds to the C-terminal region of CtA, single cysteines were introduced at six positions downstream of the leader peptide (Figure 5B). We also placed single cysteine, in three position, in the otherwise cysteine-free PCAT1: K275C and A433C on the interior surface of the TM cavity and K417C in an extracellular loop outside the translocation pathway (Figure 5C). The proximity of the cysteine pairs was analyzed by mixing CtA with PCAT1 in the presence of the oxidizing agent CuPhen or the reducing agent DTT, and then using Western blot to detect the appearance of PCAT1-CtA crosslinked product (Figure 5B and C). Mass spectrometry analysis was performed on three samples excised from the SDS-PAGE gel. In each case, the expected cross-linked peptide containing the engineered disulfide bond was observed. Cysteines at multiple positions along CtA can be crosslinked to K275C and A433C on PCAT1, suggesting that the cargo inside the cavity is unstructured and flexible, consistent with the amorphous nature of the EM density (Figure 5A). The cysteine at position 38 of CtA can be weakly crosslinked to K275C but not to K433C, which is consistent with the structure. K38 is near the leader peptide at the cytoplasmic side of the membrane and is not able to reach the extracellular ceiling of the cavity prior to proteolytic cleavage. No crosslinking product was observed for the cysteine placed outside the TM cavity (K417C), indicating that the reaction is specific (Figure 5B and C).

The density inside the TM cavity appears to be connected to only one of the two leader peptides (Figure 5A and D), raising the question whether one or two cargos are inside the TM cavity. We asked if the TM cavity is large enough to accommodate two cargos. Using the 3V server (Voss and Gerstein, 2010), the volume of the TM cavity is estimated to be 15,000 ų. Considering the average density of a small protein is around 1.35 g/cm (Voss and Gerstein, 2010; Fischer et al., 2009), the size of a CtA cargo is approximately 9000 ų, a conservative estimate for an unstructured protein. Thus, it is likely that only one cargo is inserted into the TM cavity. We termed the cargo inside the TM cavity the ‘translocating CtA’ as it is positioned for cleavage and translocation. The density for the cargo region is weak and amorphous, suggesting that there is no specific interaction between the cargo and transporter, and the substrate specificity is conferred through the leader peptide only. For the other substrate, the non-translocating CtA, no density is observed beyond S26, two residues beyond the double glycine motif (Figure 5D and Figure 2—figure supplement 3E). Furthermore, G23, G24, and S25 form a kink, orienting the C-terminus towards the cytoplasm (Figure 5D and Figure 2—figure supplement 3E). There is also biochemical evidence that the C-terminus of CtA is accessible: the PCAT1-CtA complex can be pulled down using an antibody against a tag placed at the C-terminus of CtA (Lin et al., 2015). Based on these observations, we suggest that the non-translocating cargo is located in the cytosol, flexibly linked to the leader peptide.

Comparing the structures of PCAT1 in the presence and absence of CtA shows that in the substrate-bound conformation, the intracellular opening of the TM cavity (formed from TM3 and TM4) is approximately 3 Å wider and the two NBDs are further apart (Figure 6). These structural differences are not influenced by crystal packing, as regions involved in lattice contacts do not undergo conformational changes upon CtA binding (Figure 6—figure supplement 1).

The conformational change of PCAT1 is opposite to that of the multidrug transporter MRP1, in which substrate binding brings the two NBDs closer (Johnson and Chen, 2017). MRP1, like many other ABC transporters, has higher ATPase activity in the presence of its substrate (Johnson and Chen, 2017). In contrast, addition of CtA reduces the ATPase activity of PCAT1 by about 10% for the wild-type protein and 50% for the cleavage-incompetent C21A mutant (Lin et al., 2015). As ATP hydrolysis requires NBD dimerization, the structural observation correlates well with the distinct biochemical property of this protein-conducting ABC transporter. The insertion of a large cargo into the TM cavity would likely decrease the rate of NBD dimerization thereby slowing down ATP hydrolysis.

Binding of the leader peptide also induces local conformational changes within the PEP domain. The largest displacement, approximately 5 Å, occurs in a loop (residues 91–98) preceding the catalytic residue H99 (Figure 3—figure supplement 2C). As a result, in the substrate-bound structure, a path between the catalytic site and the TM cavity opens up permitting insertion of the C-terminal cargo into the translocation pathway at the lateral opening between TM3 and TM4. (Figure 6A).

The wt and mutant PCAT1 genes were subcloned into the pMCSG20 vector with an N-terminal glutathione-S-transferase (GST) tag and a Tobacco Etch Virus (TEV) protease cleavage site. Protein was expressed and purified as previously described (Lin et al., 2015). Briefly, E. coli strain BL21(DE3) codon plus (RIL) cells expressing PCAT1 were lysed and solubilized in buffer containing 1% n-dodecyl-β-D-maltoside (DDM; Anatrace), 50 mM Tris pH 7.0, 500 mM NaCl, 10% glycerol, and 5 mM DTT. PCAT1 was enriched on Glutathione Sepharose 4B resin, washed with buffer containing 50 mM Tris pH 7.0, 500 mM NaCl, 10% glycerol, 5 mM DTT, and 2 mM n-undecyl-β-D-maltopyranoside (UDM; Anatrace). The GST tag was removed by cleavage with TEV protease, and PCAT1 was further purified using a Superdex 200 increase column (GE Healthcare) in a buffer containing 50 mM Tris pH 7.0, 150 mM NaCl, 2 mM UDM.

The gene encoding CtA was subcloned into the pMCSG7 vector with an N-terminal TEV-cleavable 6x His tag and a C-terminal 3x Flag or HA-tag. RIL cells expressing CtA were resuspended in lysis buffer (50 mM Tris pH 7.0, 150 mM NaCl, and 10% glycerol), lysed by three passes through a high-pressure homogenizer (Emulsiflex C-3; Avestin), and centrifuged at 80,000 g for 40 min to isolate inclusion bodies. The pellet was washed extensively with lysis buffer plus 1% Triton X-100, then with lysis buffer alone, before solubilizing in 8 M urea. The denatured protein was purified on cobalt affinity resin (Clontech Laboratories) and refolded via dialysis in lysis buffer plus 5 mM DTT. The His-tag was removed by cleavage with TEV protease and the protein were further purified by gel-filtration chromatography (Superdex 75 HiLoad 16/60, GE Healthcare).

The purified PCAT1 samples were buffer exchanged into native MS solution (200 mM ammonium acetate pH 7.5 with either 0.058% UDM or 0.5% C₈E₄) using Zeba microspin desalting columns (Thermo Scientific) with a 40 kDa molecular weight cut-off (MWCO). The typical concentrations used for native MS analysis were 4 µM PCAT1 monomer for the PCAT1-only sample and 4 µM PCAT1(C21A) monomer + 8 µM CtA (a two-fold excess of CtA). An aliquot (2–3 µL) of the sample was loaded into a gold-coated quartz capillary that was fabricated in-house. The sample was then electrosprayed into an Exactive Plus EMR instrument (Thermo Fisher Scientific) using a static nanospray source. The MS parameters used include: spray voltage, 1.0–1.4 kV; capillary temperature, 125°C; S-lens RF level, 200; resolving power, 8750 or 17,500 at m/z of 200; AGC target, 1 × 10⁶; number of microscans, 5; maximum injection time, 200 ms; injection flatapole, 8 V; interflatapole, 4 V; bent flatapole, 4 V; ultrahigh vacuum pressure, 7–10 × 10⁻¹⁰ mbar; total number of scans, 100. The in-source dissociation (ISD) and higher-energy collisional dissociation (HCD) parameters were varied and optimized accordingly (see Figure 1—figure supplements 1 and 2). Mass calibration was performed using cesium iodide. The acquired MS spectra were visualized using Thermo Xcalibur Qual Browser (version 3.0.63) and deconvolution was performed either manually or using UniDec version 3.2 (Marty et al., 2015; Reid et al., 2019). The deconvolved spectra from UniDec were plotted using the m/z software (Proteometrics LLC). Experimental masses were reported as the average mass ± standard deviation (S.D.) across all calculated mass values obtained within the observed charge state series.

Purified PCAT1(C21A) (5 mg/ml, 62 µM monomer) was mixed with 62 µM refolded refolded CtA and incubated on ice for 30 mins. About 3 µl of sample was applied onto glow-discharged holey carbon grids (Quantifoil gold R1.2–1.3), incubated for 20 s at 100% humidity, and blotted with filter paper for 3 s before being plunge-frozen into liquid ethane using a Vitrobot Mark IV (FEI). A dataset of 3478 movies was collected on the Titan Krios Transmission Electron Microscope (FEI) outfitted with a K2 Summit direct electron detector (Gatan) with a super-resolution pixel size of 0.545 Å using Leginon (Suloway et al., 2009). The electron dose rate was eight electrons/pixel/s with a total exposure time of 12 s resulting in a total electron dose of 80 electrons/Ų over 60 frames.

The procedure for image processing is summarized in Figure 2—figure supplement 1 . Movie frames were corrected using a gain reference and binned by a factor of two to a pixel size of 1.09 Å. Movie frames were aligned using Unblur (Grant and Grigorieff, 2015) and the effective contrast transfer function (CTF) was determined from frame-summed micrographs using CTFFIND4 (Rohou and Grigorieff, 2015). Templates for auto-picking were generated from 2D classes generated from 5000 manually picked particles in RELION (Scheres, 2016). After manual inspection to remove false positives, 572,800 automated picked particles were extracted with a box size of 300 pixels and subjected to drift correction using alignparts_lmbfgs (Rubinstein and Brubaker, 2015). The resulting particles were 2D-classified into 150 classes after which 383,002 particles were selected. An ab initio 3D model with C2 symmetry, generated from CryoSPARC (Punjani et al., 2017), was low-passed filtered to 60 Å and used as an initial model for 3D Classification in RELION (Scheres, 2016). The most populated class was further refined in RELION to 4 Å resolution. A smoothed mask, excluding the detergent micelle, was created and used for 3D classification without alignment and subsequent local refinement in RELION. The final 3D reconstruction with C2 symmetry yielded a 3.9 Å map.

The movie frames were also motion-corrected using MotionCor2 software (Zheng et al., 2017), and CTF estimation was calculated using gCTF both implemented in Relion 3 (Zhang, 2016). Subsequently, the final set of selected particle were re-extracted and subjected to three iterations of Bayesian particle polishing (Zivanov et al., 2019), CTF refinement (Zivanov et al., 2018), and local refinement with C2 symmetry in Relion 3. One round of refinement in C1 was performed to release the symmetry from C2 to C1. Subsequent 3D classification without image alignment was performed. The 3D classes were manually inspected for differences in local asymmetry. The particles belonging to a 3D class that appeared to have an opposite orientation were rotated 180° along the symmetry axis. One round of local refinement with C1 symmetry yielded 3.6 Å resolution in Relion. The final round of refinement was performed using CryoSPARC produces a map of 3.37 Å resolution (Punjani et al., 2017). The resolution was estimated using Fourier Shell Correlation (FSC) with 0.143 cutoff implemented in the 3DFSC web application (Tan et al., 2017) Figure 2—figure supplement 4. Local resolution estimation from two CryoSPARC half maps was performed in CryoSPARC Figure 2—figure supplement 2.

The crystal structure of PCAT1 (PDB:4RY2) was placed into a sharpened cryo-EM map (sharpening factor, −80 Ų) using rigid body fitting in Chimera (Pettersen et al., 2004) followed by manual adjustments in Coot (Emsley and Cowtan, 2004). The final model consists of residues 9–722 of PCAT1, residues 8–29 of the translocating CtA, and residues 8–25 of the non-translocating CtA.

The model was initially refined against one working half map in real-space by PHENIX (Adams et al., 2010), followed by rounds of refinement in reciprocal space using REFMAC (Brown et al., 2015), with secondary structure and reference restraints derived from ProSMART (Nicholls et al., 2014). The quality of the final model was evaluated by MolProbity (Chen et al., 2010). To assess the degree of overfitting, we calculated the FSC curves between the model and working half map, the non-working half map, and the full map using SPIDER (Frank et al., 1996).

To perform disulfide cross-linking accessibility assay, we constructed a cysteine-free PCAT1 by replacing all nine cysteines in PCAT1 with serine. Single cysteine substitutions at positions 275, 417, or 433 were introduced to the cysteine-free PCAT1 construct by site-directed mutagenesis. Mutant CtA (0.6 µM) containing an introduced cysteine residue and a HA tag was mixed with equimolar single-cysteine PCAT1 in buffer containing 50 mM HEPES pH 7.0, 150 mM NaCl, and 2 mM UDM. The reaction mixture was incubated at room temperature for 10 min in the presence of 15 µM Cu-Phenanthroline or 50 mM DTT before being analyzed by SDS-PAGE. For Western blotting, the cross-linked product was visualized with a primary anti-HA mouse monoclonal antibody (1:10000) and a secondary goat anti-mouse Alexa Fluor 680 antibody (1:1000). In addition, we have verified the identity of the cross-linked products by mass spectrometry for the following pairs: PCAT1 K275C-CtA K38C, PCAT1 K275C-CtA K44C, and PCAT1 K275C-CtA K65C. MS analysis was performed by the Rockefeller University Proteomics Resource Center.

To assess residues on CtA for PEP binding, PCAT1 (C21A)-bound GST sepharose resin was used to pull down wt or mutant CtA. CtA titration was performed by mixing 20 µl of Glutathione resin with 0.2, 0.4, 0.8, 1.6, or 3.2 nmol of CtA in 100 µl reaction buffer. The reaction mixture was incubated for 30 min on ice, after which the resin was washed twice with 400 µl wash buffer (50 mM Tris pH 7.0, 500 mM NaCl, 10% glycerol, 2 mM UDM, 5 mM DTT). The samples were analyzed by SDS-PAGE as well as Western with M2 anti-flag mouse antibody as primary antibody (1:5000) and goat anti-mouse Alexa Fluor 680 antibody as a secondary antibody (1:1000).

To assess the reciprocal binding residues on PCAT1, site-directed mutagenesis was performed to introduce mutation to PCAT1 (C21A), after which the pulldown experiment was performed using the C-terminal 3xFlag tagged CtA as a bait. For each reaction, 0.6 µM of PCAT1 was incubated on ice for 10 min with 1.2, 2.4, 4.8, 9.6, or 19.2 µM of CtA in 100 µl reaction buffer (50 mM HEPES pH 7.0, 150 mM NaCl, 2 mM UDM, 5 mM DTT). To capture the PCAT1 (C21A)-CtA complex, 15 µl of Anti-Flag M2 agarose affinity gel (Sigma-Aldrich) was added to the reaction and incubated for 30 min. The agarose affinity gel was washed twice with 400 µl reaction buffer. Samples analyzed by SDS-PAGE.

Structure figures were prepared using the program PyMOL (Schrodinger, 2015), UCSF Chimera (Pettersen et al., 2004), and UCSF ChimeraX (Goddard et al., 2018).