The Sec1–Munc18 protein VPS33B forms a uniquely bidirectional complex with VPS16B

Two or more copies of VPS33B and VPS16B are present in the complex

To determine whether multiple copies of VPS33B and/or VPS16B are present in the complex, we utilized differentially tagged proteins selectively expressed in HEK293 cells, as shown in Figure 2, A–D. Hemagglutinin (HA)-VPS33B, Myc-VPS16B, and VPS16B-FLAG coexpressed in HEK293 cells (Fig. 2A) revealed that VPS16B-FLAG could co-IP Myc-VPS16B (Fig. 2B). Similarly, when HA-VPS33B, Myc-VPS16B, and VPS33B-FLAG were cotransfected (Fig. 2C), VPS33B-FLAG could co-IP HA-VPS33B (Fig. 2D). These results indicate that in mammalian cells, both native and tagged VPS33B and VPS16B form a large stable complex containing more than one copy of each protein.

Open in a separate window

Figure 2

The VPS33B–VPS16B complex contains multiple copies of each component.A–D, SDS-PAGE immunoblots show differentially tagged VPS33B and VPS16B expressed in transfected HEK293 cells can co-IP a second copy of each protein. A, probing for FLAG, HA, Myc, and GAPDH (bottom) of lysates from HEK293 cells cotransfected with HA-VPS33B + Myc-VPS16B, and with/without VPS16B-FLAG as indicated. B, in lysates from triple-transfected cells, anti-FLAG antibody co-IP’s HA-tagged VPS33B and Myc-tagged VPS16B, confirming at least two copies of VPS16B in the complex. C, probing of lysates from cells cotransfected with HA-VPS33B + Myc-VPS16B, and with/without VPS33B-FLAG. D, in lysates from triple-transfected cells, anti-FLAG antibody co-IP’s HA-tagged VPS33B and Myc-tagged VPS16B, confirming at least two copies of VPS33B in the complex. co-IP, coimmunoprecipitation; HA, hemagglutinin; HEK293, human embryonic kidney 293 cell line; VPS, vacuolar protein sorting protein.

Human VPS33B and VPS16B expressed in yeast also form a high MW complex

To delineate the structure of the VPS33B–VPS16B complex, we required the sizeable amounts of protein typically generated via cellular expression systems. When expressed in Escherichia coli, we found that VPS33B and VPS16B were mostly present in insoluble inclusions, and whereas expression in Sf9 and Sf21 insect cells using baculovirus yielded stable complex, the yield was insufficient for structural analysis. We then found that human VPS33B and VPS16B can be expressed at higher levels in S. cerevisiae using the pBEVY-T vector, which contains a constitutive bidirectional glyceraldehyde-3-phosphate dehydrogenase-alcohol dehydrogenase I fused promoter suitable for coexpression of two proteins at similar levels (44). Untagged VPS33B and VPS16B-3xFLAG purified from yeast using FLAG affinity purification followed by anion exchange chromatography revealed the expected MWs on BN-PAGE immunoblots (Fig. 3A) and showed no significant impurities on Coomassie blue-stained gels (Fig. 3B). The expected predominant ∼480 kDa complex was observed on a stained BN-PAGE gel (Fig. 3C), from which the band was extracted and after trypsin digestion analyzed by mass spectrometry, which detected primarily VPS33B and VPS16B (Table S1). The only additional protein detected was the yeast actin-binding protein YFR016C. Only two YFR016C peptides with total spectral count of two were detected compared with spectral counts of more than 200 for VPS33B and VPS16B peptides; therefore, this is probably not a specific interactor with human VPS33B–VPS16B. VPS33B–VPS16B-3xFLAG affinity purified from yeast was also examined using circular dichroism, which relies on differential absorption of left- and right-handed circularly polarized light of varying wavelengths by protein secondary structures and can be used to assess the folded state of protein (45). The VPS33B–VPS16B-3xFLAG complex was observed to be well folded and primarily alpha helical (Fig. 3D).

Open in a separate window

Figure 3

Human VPS33B–VPS16B complex expressed in yeast is stable and well folded with a molecular weight of approximately 315 kDa.A, BN-PAGE immunoblots probed for VPS33B and VPS16B show the presence of both in human VPS33B–VPS16B-3xFLAG complexes affinity purified from yeast; major band comigrates with 480 kDa marker (arrowhead) as in HEK293 and DAMI cell lysates (Fig. 1A). B, Coomassie blue–stained SDS-PAGE gel indicates VPS33B and VPS16B are the only proteins present in purified complex, also shown in native form. C, on stained BN-PAGE gel. D, circular dichroism analysis of purified complex indicates mostly helical secondary structure. E, Representative SEC-MALS chromatogram for purified complex. F, SDS-PAGE immunoblots detected VPS33B and VPS16B in the major peak (fraction 2); results yielded a mean calculated molecular weight of 315 kDa for the complex (see Fig. S1 for light scattering data). BN-PAGE, blue native polyacrylamide gel electrophoresis; DAMI, megakaryoblastic DAMI cell line; HEK293, human embryonic kidney 293 cell line; VPS, vacuolar protein sorting protein.

VPS33B and VPS16B form a 315 kDa MW complex

The size of VPS33B–VPS16B has not previously been determined, beyond an estimate of the hydrodynamic radius of a recombinant complex determined by SEC (41). We used SEC–MALS to determine the MW of purified VPS33B–VPS16B (Figs. 3E and S1), detected in a peak fraction confirmed by immunoblotting to contain VPS33B and VPS16B (Fig. 3F). Results from analysis of three independent purifications gave MW estimates for VPS33B–VPS16B of 310 kDa ± 9%, 327 kDa ± 3.3%, and 310 kDa ± 0.4%. These values give an average MW of ∼315 kDa, which is larger than the minimal size predicted from co-IP experiments that indicated the presence of at least two copies each of VPS33B and VPS16B (Fig. 2) for a total minimum MW of ∼261 kDa (VPS33B: 70.6 × 2 = 141.2 kDa and VPS16B-3xFLAG: 60.8 × 2 = 121.6 kDa).

The VPS33B–VPS16B complex contains two copies of VPS33B and three copies of VPS16B

To determine the ratio of components within VPS33B–VPS16B, we expressed both proteins in yeast with short N-terminal HA tags and confirmed recovery of affinity-purified complex by SDS-PAGE (Fig. 4A). Immunoblots of serially diluted samples were probed with primary anti-HA antibody and secondary Alexa-488-labeled antibody, and nonsaturated sets of bands were quantified by scanning (Fig. 4, B and C). We observed that the VPS33B band (band 1, upper) represented 39.1% of the signal and the VPS16B band (band 2, lower) represented 60.9%, indicating a 2:3 ratio of VPS33B to VPS16B in the complex. The calculated MW of 324 kDa containing two copies of VPS33B (141.2 kDa) and three copies of VPS16B-3xFLAG (182.4 kDa) is comparable to the SEC–MALS determined average MW of ∼315 kDa.

Open in a separate window

Figure 4

The VPS33B–VPS16B complex contains two VPS33B and three VPS16B proteins.A, SDS-PAGE gel of HA-VPS33B–HA-VPS16B-3xFLAG complex purified from yeast lysate via FLAG affinity purification and anion exchange chromatography. B, serially diluted samples of complex (total protein shown for each lane) were blotted and probed with monoclonal anti-HA primary antibody and Alexa Fluor 488–conjugated secondary antibody; box indicates 10 ng bands used for quantification (see the Experimental procedures section). C, analysis of fluorescence signal showed 39.1% of total corresponded to HA-VPS33B and 60.9% to HA-VPS16B-3xFLAG, indicating a 2:3 ratio of these proteins in the complex. HA, hemagglutinin tag; VPS, vacuolar protein sorting protein.

The VPS33B–VPS16B complex has a two-lobed overall structure

We pursued structural determination by negative-staining EM and cryo-EM. Attempts to perform high-resolution cryo-EM with purified complexes were unsuccessful since the complexes became heterogeneous and produced very low-resolution 2D and 3D classes. It is likely that this was due to the cryo-EM sample preparation causing unfavorable interactions between the VPS33B–VPS16B-3xFLAG complex and the air–water interface and/or flexibility within the complex. Flexibility was also an obstacle to initial structural determination of the intact HOPS complex by cryo-EM, requiring crosslinking and negative-staining EM (46). We therefore focused our efforts on negative-staining EM, utilizing glutaraldehyde crosslinking to stabilize the complex. Negative-staining EM of the VPS33B–VPS16B complex revealed a two-lobed configuration, approximately 15 nm along the major axis. 2D classification of complexes revealed a slight bend between the two lobes. The complex appears to contain two ball-like structures connected by a thinner stalk (Fig. 5, A and B). Validation of this structure came from SAXS of affinity purified VPS33B–VPS16B-3xFLAG (Fig. S2, A and B). The results revealed VPS33B–VPS16B has a radius of gyration of 7.27 ± 0.14 nm and hydrodynamic radius of 9.38 ± 0.18 nm, dimensions consistent with our negative-staining results.

Open in a separate window

Figure 5

Negative-staining EM of VPS33B–VPS16B complex.A, representative micrograph of negative-stained glutaraldehyde-crosslinked purified VPS33B–VPS16B-3xFLAG complex; circles indicate representative particles used for 2D class averaging. Micrographs collected at 25,000× magnification with image pixel size of 2.61 Å; bar represents 200 nm. B, 2D class averages generated from 407 particles reveal an elongated two-lobed complex structure; bar represents 10 nm. VPS, vacuolar protein sorting protein.

Localization of VPS33B molecules within the complex

To localize VPS33B, we tagged the C terminus with a biotinylation signal to allow attachment of avidin, which can be resolved by negative-staining EM of purified VPS33B–VPS16B. VPS33B containing the C-terminal biotinylation signal formed a complex with VPS16B that could be purified from yeast, as shown in SDS-PAGE and BN-PAGE gels (Fig. 6, A and B). Subsequent attachment of avidin tetramers and negative-staining EM revealed avidin extensions on opposing ends of the bilobed complex, indicating the positioning of VPS33B molecules within the complex in a trans orientation (Fig. 6, C and D).

Open in a separate window

Figure 6

Localization and orientation of VPS33B within complex via avidin labeling of biotinylated VPS33B.A, SDS-PAGE gel showing avidin-labeled VPS33B–VPS16B; avidin binding to biotinylated VPS33B was disrupted by preheating the sample to 95 °C for 5 min. B, BN-PAGE gel shows that addition of the biotinylation signal does not disrupt formation of VPS33B–VPS16B complex (arrowhead). C, 2D class averages derived from negative-staining EM of crosslinked avidin-labeled VPS33B–VPS16B show avidin bound to opposing sides of the broad lobes of the complex. Classes were generated from 234 particles. The scale bar represents 10 nm. D, 2D class showing representation of avidin and VPS33B overlaid to indicate position of VPS33B subunits (71 kDa native protein + 8 kDa biotinylation signal) and avidin (tetramer with total molecular weight of 67 kDa). The scale bar represents 10 nm. BN-PAGE, blue native polyacrylamide gel electrophoresis; VPS, vacuolar protein sorting protein.

N- and C-terminal regions of VPS16B are not required for the VPS33B–VPS16B complex

To determine regions of VPS33B and VPS16B required for complex formation, we made a series of truncations and assessed complex assembly. We observed that truncations in VPS33B (VPS33B [202–617], VPS33B [322–617], VPS33B [1–360], and VPS33B [1–401]) were not tolerated, resulting in lack of expression in yeast. However, deletions in VPS16B (VPS16B [143–493], VPS16B [143–316]), when expressed with full-length VPS33B, were affinity purified (Fig. 7, A and B). Complexes containing both VPS16B truncations could be isolated on BN-PAGE (Fig. 7, C and D) and were able to form bilobed complexes similar to intact VPS16B (Fig. 7, E and F). We confirmed that the composition of two copies of VPS33B and three copies of truncated VPS16B was maintained in these complexes (Fig. S3). These data indicate that the core portion of VPS16B, VPS16B (143–316), is required for VPS33B–VPS16B interaction in the complex. The N and C termini of VPS16B may have additional roles in complex function.

Open in a separate window

Figure 7

Truncated VPS16B forms a stable complex with VPS33B.A, Coomassie blue–stained SDS-PAGE gel comparing purified VPS33B–VPS16B complex containing wildtype or N-terminal truncated VPS16B. B, SDS-PAGE gel of complex containing VPS33B and amino acids 143 to 316 of VPS16B. C, BN-PAGE gel of VPS33B in complex with amino acids 143 to 493 of VPS16B (arrowhead); truncation does not appear to disrupt complex formation. D, BN-PAGE gel of VPS33B in complex with amino acids 143 to 316 of VPS16B (arrowhead); truncation does not appear to disrupt complex formation. E, 2D class averages from negative-staining EM of crosslinked VPS33B–VPS16B(143–493) complex showing a two-lobed structure similar to full-length complex (Fig. 5B), indicating the stable association of subunits when N terminus of VPS16B is absent. Classes were generated from 308 particles. F, 2D class averages from negative-staining EM of crosslinked VPS33B–VPS16B(143–316) complex showing a two-lobed structure similar to full-length complex, indicating the stable association of subunits when N- and C-terminal regions of VPS16B are absent. Classes were generated from 424 particles. The scale bars represent 10 nm. VPS, vacuolar protein sorting protein.

ARC-causing L30P variant in VPS33B abrogates complex formation

ARC-causative single amino acid substitutions of VPS33B are cataloged in the Leiden Open Variation Database (https://grenada.lumc.nl/LSDB_list/lsdbs/VPS33B) (47). When the ARC-causing VPS33B variant L30P was coexpressed with VPS16B-3xFLAG and protein was FLAG affinity purified, we observed the absence of the typical VPS33B band together with the appearance of a prominent ∼70 kDa band on SDS-PAGE (Fig. 8A), which mass spectrometry identified as yeast HSP70-family proteins (Table S2). Immunoblotting confirmed the absence of both VPS33B and VPS16B-FLAG (Fig. 8, B and C) in coexpressing yeast cell lysates, consistent with loss of both proteins in humans with this variant (48), suggesting that the L30P region of VPS33B is important for stable complex formation. In contrast, similar examination of the ARC-associated VPS33B variants S243F and H344D showed they could be recovered when coexpressed with VPS16B-3xFLAG in yeast (Fig. 8A), although immunoblotting indicated slightly lower expression of VPS33B and VPS16B (Fig. 8, B and C). Both variants formed complexes with VPS16B detected by BN-PAGE (Fig. 8D), which showed the expected bilobed structure when assessed by negative-staining EM (Fig. 8, E and F). The interaction of VPS16B (VIPAS39) with S243F VPS33B has been suggested in a previous study (49); however, interaction with the H344D variant has not been reported.

Open in a separate window

Figure 8

Effects of ARC-syndrome causative VPS33B variants L30P, S243F, and H344D on complex formation.A, SDS-PAGE gel of complexes containing VPS33B L30P, S243F, and H344D purified from yeast shows VPS33B is mostly absent in the L30P sample and VPS16B is also reduced. B, immunoblotting for VPS33B in yeast lysates expressing VPS33B wildtype, L30P, S243F, and H344D along with wildtype VPS16B-3xFLAG show VPS33B is undetectable in the L30P lysate. C, immunoblotting for VPS16B-3xFLAG in yeast lysates from cultures expressing VPS33B wildtype, L30P, S243F, and H344D shows FLAG signal is weak in the L30P lysate. For (B) and (C), nitrocellulose membranes were blotted for α-tubulin as loading control without stripping. D, silver-stained BN-PAGE gel of complexes containing VPS33B S243F or H344D showing that these variants do not disrupt complex formation. Complex indicated by arrowhead. E, 2D class averages from negative-staining EM of VPS33B–VPS16B-3xFLAG complexes containing VPS33B S243F mutation. Classes generated from 257 particles. The scale bar represents 10 nm. F, 2D class averages from negative-staining EM of VPS33B–VPS16B-3xFLAG complexes containing VPS33B H344D mutation. Classes generated from 259 particles. The scale bar represents 10 nm. ARC, arthrogryposis, renal dysfunction, and cholestasis; BN-PAGE, blue native polyacrylamide gel electrophoresis; VPS, vacuolar protein sorting protein.

Proposed structure of the VPS33B–VPS16B complex

To refine our model of the 2:3 VPS33B–VPS16B complex, we combined our EM data for the overall bilobed structure (Fig. 5) and the location of VPS33B subunits (Fig. 6D) with predicted structures derived from the AlphaFold Protein Structure Database (https://alphafold.ebi.ac.uk/), generated via an artificial intelligence system that has produced highly accurate 3D protein structure predictions for 200 million proteins from amino acid sequences (50, 51, 52). The structure for VPS33B (https://alphafold.ebi.ac.uk/entry/{"type":"entrez-protein","attrs":{"text":"Q9H267","term_id":"313104046","term_text":"Q9H267"}}Q9H267) contains several alpha helices folded into a ball (Fig. 9A, left upper panel—red dots indicate C-terminal serine residues) where the per-residue confidence score (pLDDT) is very high (pLDDT >90) as indicated by the large proportion of blue in the ribbon structure. Other regions (indicated in yellow) have low to very low confidence, and we modeled these next to each other in the proposed trans arrangement of VPS33B proteins. The predicted structure for VPS16B (VIPAS39; https://alphafold.ebi.ac.uk/entry/{"type":"entrez-protein","attrs":{"text":"Q9H9C1","term_id":"41016926","term_text":"Q9H9C1"}}Q9H9C1) is elongate (Fig. 9A, lower left) with a core region of alpha helices at very high confidence (outlined in green), and a less structured N-terminal region (amino acids 1–152) with low to very low confidence (shown in yellow). Guided by this information and a 3D volume derived from our EM data (Fig. 9A, right panel top), we propose that in the VPS33B–VPS16B complex, two well-folded VPS33B molecules comprise either end in an opposing trans orientation, whereas three VPS16B molecules are located between them (Fig. 9A, right panel bottom).

Open in a separate window

Figure 9

Proposed model of the VPS33B–VPS16B complex.A, top left, two copies of the AlphaFold predicted VPS33B structure (https://alphafold.ebi.ac.uk/entry/{"type":"entrez-protein","attrs":{"text":"Q9H267","term_id":"313104046","term_text":"Q9H267"}}Q9H267) arranged in trans, consistent with our structural data (Fig. 6; C-terminal serine 617 highlighted in red). Bottom left, two copies of the AlphaFold VPS16B structure (https://alphafold.ebi.ac.uk/entry/{"type":"entrez-protein","attrs":{"text":"Q9H9C1","term_id":"41016926","term_text":"Q9H9C1"}}Q9H9C1) with high-confidence region highlighted in green outline. Right, VPS33B–VPS16B complex 3D volume (gray) generated from negative-staining EM data (Fig. 5), above proposed model of the complex assembled from two VPS33B molecules bridged by three VPS16B. B, AlphaFold predicted structure of VPS16B with highlighted amino acids 143 to 316 (turquoise) we have identified as necessary for complex formation. C, AlphaFold predicted structure of VPS33B showing ARC-causing amino acid changes L30P, S243F, and H344D highlighted in red; the structures are rotated −90° along the Y-axis relative to each other. For AlphaFold predicted structures in (A) and (C), blue ribbon indicates per-residue confidence scores >70 (confident and very confident), and yellow ribbon indicates per-residue confidence scores <70 (low and very low confidence). ARC, arthrogryposis, renal dysfunction, and cholestasis; VPS, vacuolar protein sorting protein.

BN-PAGE

Precast Invitrogen NativePAGE 4 to 16% Bis–Tris Protein Gels (Thermo Fisher; catalog no.: BN1002BOX) in an XCell SureLock Mini gel tank (Thermo Fisher; catalog no.: EI0001) were utilized. Sample buffer, running, and transfer buffers were prepared according to the manufacturer’s specifications. Gels were run at room temperature at constant 150 V for 3 h.

DAMI cell cultures

DAMI cells were grown to confluence in Iscove’s media with 5% horse serum, in a 37 °C incubator with 5% CO₂ injection. For BN-PAGE, cells were pelleted at 400g for 3 min and lysed in lysis buffer (50 mM Tris, 150 mM NaCl, 1 mM EGTA, 0.5% Triton X-100, pH 7.4) for 1 h on ice. Lysate was cleared at 214,000g for 1 h at 4 °C, and the supernatant was dialyzed in 1 l 1× native sample buffer overnight at 4 °C.

Silver staining of polyacrylamide gels

SDS-PAGE or BN-PAGE gels that had been fixed (40% methanol, 10% acetic acid, 50% Milli-Q water v/v/v) for at least 15 min were equilibrated for 1 h in 50 ml Milli-Q water. The gels were rinsed once with Milli-Q water, reduced with sodium thiosulfate (20 mg in 100 ml Milli-Q water), and then rinsed twice with Milli-Q water for 30 s per rinse. Gels were stained in a clean glass container with 100 ml staining solution (100 mg silver nitrate in 100 ml Milli-Q water) for 25 to 30 min. Gels were rinsed twice with a small amount of developing solution (3 g sodium carbonate and 1 grain or approximately 1 mg of sodium thiosulfate in 100 ml Milli-Q water, supplemented with 140 μl 37% formaldehyde immediately before use) and developed in the remaining solution until satisfactorily stained, usually 30 s to 1 min. Development was stopped by replacing the developing solution with 1% v/v glacial acetic acid in Milli-Q water and allowing the gel to equilibrate for 20 min. All solutions are prepared fresh as needed.

Protein purification from yeast

BJ2168 yeast containing the selected pBEVY-T plasmid growing on Trp D.O. agar media was used to seed 5 ml Trp D.O. liquid culture for overnight growth at 30 °C. About 24 h later, the liquid culture was used to seed a 60 ml Trp D.O. culture. After 24 h, this culture was used to seed five 1.4 l YPD cultures in 2800 ml baffle-less Fernbach flasks. The final growth was completed in a 30 °C incubator for 18 h with 210 rpm orbital agitation. Cells were pelleted using a JLA 9.1000 rotor for 15 min at 3000g and washed once with Milli-Q water. Each growth yielded 90 to 100 g washed cell pellet. Yeast was resuspended in 50 ml lysis buffer (20 mM Tris [pH 7.6], 200 mM NaCl, 10% sucrose, and 5 mM EDTA) with protease inhibitor (0.033 g 6-aminocaproic acid, 0.052 g 4-aminobenzamidine-2HCl, and 100 μl 2% PMSF in isopropanol), and lysed using a French Press at 1000 psi. Lysate was cleared by ultracentrifugation at 50,000g for 1 h at 4 °C using a Beckman SW28 rotor, and supernatant was filtered using a 0.45 μm syringe filter. Equal portions of cleared lysate were split between two 500 μl drip columns of m2 mouse anti FLAG-conjugated agarose beads (Sigma; catalog no.: A2220). After lysate has flowed through, beads were washed first with 10 ml high salt wash buffer (20 mM Tris, 400 mM NaCl, pH 7.6) and then with 10 ml low salt wash buffer (20 mM Tris, 200 mM NaCl, pH 7.6) without disturbing the packed bed. Aliquots of 3xFLAG peptide were diluted to a final concentration of 133 μg/ml using the low salt buffer. Each FLAG affinity column was incubated three times with 500 μl diluted FLAG peptide, for 15 min per incubation, followed by one incubation with 500 μl low salt buffer.

Ion exchange chromatography

For ion exchange chromatography, 1 ml GE HiTrap Q-HP anion exchange columns (VWR; catalog no.: CA95055-866L) and a GE AKTA Pure FPLC system were used. Purification was carried out in a 4 °C cold cabinet. The column was first equilibrated with at least five column volumes of low salt buffer (20 mM Tris, 200 mM NaCl, pH 7.6), and elutions 2, 3, and 4 from FLAG affinity column were then injected using a 5 ml injection loop prefilled with low salt buffer. Protein was eluted using a 200 to 450 mM NaCl gradient with 20 mM Tris (pH 7.6) over 15 1 ml fractions, at a flow rate of 1 ml/min. Samples were concentrated and exchanged into 20 mM Hepes buffer (pH 8.0) with 500 mM NaCl using Amicon mini concentrators with 100 kDa MW cutoff (Sigma; catalog no.: UFC510024).

Avidin labeling of VPS33B-biotin

The biotinylation signal was added to C terminus of VPS33B by overlapping PCR as described previously. The biotinylation signal consists of amino acids 49 to 123 of Propionibacterium shermanii transcarboxylase (methylmalonyl-CoA carboxyltransferase 1.3S subunit; UniProt: {"type":"entrez-protein","attrs":{"text":"P02904","term_id":"114847","term_text":"P02904"}}P02904), corresponding to its biotinyl-binding domain. The amino acid sequence of this biotinylation signal is AGKAGEGEIPAPLAGTVSKILVKEGDTVKAGQTVLVLEAMKMETEINAPTDGKVEKVLVKERDAVQGGQGLIKIG

When expressed in yeast, this sequence is recognized by the endogenous biotinylation machinery and lysine 41 (bolded and underlined in the aforementioned sequence) becomes biotinylated.

The DNA sequence used to encode this biotinylation signal is GCCGGTAAGGCCGGAGAGGGCGAGATTCCCGCTCCGCTGGCCGGCACCGTCTCCAAGATCCTCGTGAAGGAGGGTGACACGGTCAAGGCTGGTCAGACCGTGCTCGTTCTCGAGGCCATGAAGATGGAGACCGAGATCAACGCTCCCACCGACGGCAAGGTCGAGAAGGTCCTTGTCAAGGAGCGTGACGCCGTGCAGGGCGGTCAGGGTCTCATCAAGATCGGCTGA.

Egg white avidin stock (BioShop; catalog no.: AVD407) was reconstituted in Milli-Q water at a concentration of 20 mg/ml and stored frozen at −20 °C in 50 μl aliquots. For use, 25 μl of avidin stock was diluted with 1 ml 20 mM Hepes buffer (pH 8.0) with 200 mM NaCl to 0.5 mg/ml final concentration.

Biotinylated VPS33B in complex with VPS16B-3xFLAG was expressed in yeast and loaded onto FLAG-affinity column as described previously. After washing with low salt wash buffer, 1 ml diluted avidin was introduced to the column and allowed to flow through, labeling biotinylated VPS33B immobilized on FLAG-affinity column. Excess avidin was removed by washing with 20 mM Hepes/200 mM NaCl. Elution was performed as aforementioned, except with FLAG peptide diluted in 20 mM Hepes/200 mM NaCl instead of low salt buffer. For quality control, samples of biotinylated VPS33B with avidin were mixed with SDS-PAGE running buffer and either incubated at 95 °C for 5 min or stored at room temperature for the same time, and both samples were then resolved on SDS-PAGE gel. The absence of free VPS33B-Bio band in sample that was not heated indicates robust avidin labeling.

Mass spectrometry

Bands were excised from Coomassie-stained gels and collected in a 1.5 ml Eppendorf tube. After washing with 50 μl of 50 mM ammonium bicarbonate, remaining Coomassie was removed with 50 μl of 50% v/v acetonitrile/25 mM ammonium bicarbonate. The sample was reduced with 30 μl of 10 mM DTT in 50 mM ammonium bicarbonate for 30 min at 56 °C and then alkylated with 30 μl of 100 mM iodoacetamide for 15 min at room temperature. Excess water was then removed by washing the excised band with 50 μl of 50% acetonitrile/25 mM ammonium bicarbonate as aforementioned. In-gel digestion of protein sample was performed by covering the excised band with 50 μl of 13 ng/μl MS-grade trypsin (Thermo Fisher Scientific; catalog no.: 90057) in 50 mM ammonium bicarbonate and incubating overnight at 37 °C. Peptides were extracted from the gel by washing twice alternately with 5% formic acid and 100% acetonitrile and collecting the supernatant (after 10 min incubation). Peptides were dried in a SpeedVac and reconstituted in 5 μl of 0.1% v/v formic acid in Milli-Q water. C18 Ziptip cleanup and MALDI-MS were performed by the SPARC BioCentre, The Hospital for Sick Children, Toronto, Canada.

SAXS

VPS33B/VPS16B-3xFLAG in solution at 0.92 mg/ml with 20 mM Tris (pH 7.6) and 320 mM NaCl was used for SAXS. Fresh buffer showed no significant scattering. SAXS data were collected in 24 × 5 min accumulations. SAXS was performed in 1D line collimation mode on an Anton Paar SAXSpace System with SAXSdrive, version 2.02.250.13277. SAXStreat, version 1.06.003.8398 was used to export data in a suitable format for analysis with SAXSquant, version 4.1.1.8319.

Circular dichroism

Circular dichroism was performed using a Jasco J-1500 circular dichroism spectrophotometer. Pure protein sample at 0.1 mg/ml in 500 mM NaCl and 20 mM phosphate buffer (pH 8.0) was loaded in a Hellma High Precision quartz cell with path length of 1 mm. The buffer had no significant absorbance in the 190 to 280 nm range. Data were collected at 20 nm/min from 280 to 190 nm.

SEC–MALS

SEC–MALS was performed by the Biophysical Core Facility at the Peter Gilgan Centre for Research and Learning. A 1260 Infinity II bio-inert HPLC with AdvanceBio SEC300A 2.7 μm 4.6 × 300 mm column was used. SEC–MALS was performed at 4 °C with 20 mM Hepes (pH 8.0) with 500 mM NaCl at 0.2 ml/min flow rate. Data were collected using a MiniDAWNTREOS MALS, Quasi-Elastic Light Scattering, and OptiLabT-rEXRefractive Index detection instrument (Wyatt Technology), along with UV₂₈₀ chromatogram. Samples were centrifuged at >20,000g for 10 to 15 min at 4 °C to remove any aggregates or insoluble particulates. About 20 to 25 μl of sample at 0.25 to 0.5 mg/ml were injected. Fractions were collected every 30 s (100 μl per fraction). Chromatogram data were analyzed using ASTRA software (Wyatt) with a UV₂₈₀ 0.1% of 0.903 ml/(mg cm), dn/dc of 0.185, and refractive index 0 to the supplied buffer to determine the MWs of the protein chromatographic peaks.

Quantitative immunoblotting

Complexes containing N-terminal HA-tagged VPS33B and N-terminal HA-tagged VPS16B-3xFLAG were purified from yeast using FLAG affinity followed by anion exchange as described. Pure protein concentrations were measured by A₂₈₀ nm using a NanoDrop 1000 spectrophotometer. Serial dilutions were loaded on a 10% SDS-PAGE gel, with 0.04 μg as the maximum total protein in a lane followed by a dilution series with 0.01, 0.0025, 0.000625, and 0.0003125 μg. Proteins were transferred to nitrocellulose membrane by wet transfer, for 1:30 h at 100 V. The membrane was blocked with 4% w/v bovine serum albumin in Tris-buffered saline with Tween-20. Monoclonal mouse anti HA was diluted 1:1000 in blocking buffer. Secondary antibody was Alexa Fluor 488–conjugated donkey antimouse diluted 1:5000 in blocking buffer. After 1 h incubation with secondary antibody in the dark, the membrane was washed with PBS before imaging using a BioRad ChemiDoc XRS+. Analysis was performed using Image Lab software, version 6.1 with manual lane assignment and automatic band detection at the low-sensitivity band detection setting.

Complex crosslinking

Affinity-purified VPS33B–VPS16B-3xFLAG (0.3–0.5 mg/ml) in 20 mM Hepes (pH 8.0) and 500 mM NaCl was crosslinked with 0.01% v/v glutaraldehyde at 42 °C for 5 min. Glutaraldehyde was quenched by adding 1/20 volume of 1 M Tris (pH 7.8). Samples were either used immediately for EM grid preparation or stored at 4 °C for up to 24 h before use. For avidin-labeled biotinylated VPS33B, crosslinking was performed in 20 mM Hepes (pH 8.0) and 200 mM NaCl.

Negative-staining EM

Carbon-coated copper–rhodium grids were prepared according to Booth et al. (70). Protein complexes were diluted to between 0.01 and 0.05 mg/ml in 20 mM Hepes (pH 8.0) with 500 mM NaCl, except for avidin-labeled biotinylated VPS33B, which was diluted in 20 mM Hepes (pH 8.0) with 200 mM NaCl. Grids were glow discharged for 15 s, then 4 μl of sample was pipetted onto the grid and allowed to bind for 2 min at room temperature. Excess sample was removed by edge blotting onto Whatman filter paper (Sigma; catalog no.: Z240079). Grids were washed three times in 50 μl drops of Milli-Q water and stained with 2% uranyl formate in Milli-Q water. Excess uranyl formate was removed by edge blotting, and grids were allowed to air dry. Micrographs were collected on an FEI Tecnai F20 TEM at 25,000× magnification using a Gatan Orius SC1000 CCD, with a pixel size of 2.61 Å. Relion 2.1 was used for manual particle picking and 2D classification. AlphaFold was used for ab initio 3D reconstruction.