Cysteine-based crosslinking approach for characterization of oligomeric pore-forming proteins in the mitochondrial membranes

Abstract

Mitochondria are important not only to healthy but also dying cells. In particular, apoptotic cell death initiates when the mitochondrial outer membrane is permeabilized by Bax, a protein of the Bcl-2 family. Bax shares a structural fold with some α-helical bacterial pore-forming toxins before these proteins actively engage membranes. Despite decades of intensive research, the structures of the pores formed by these proteins are mostly unknown, mainly because the pores are assembled by different numbers of the proteins whose conformation and interaction are highly dynamic. Site-specific crosslinking of the pore-forming proteins in cellular membranes where the pores are assembled is a powerful approach to assess the biological pore structure, dynamics and function. In this chapter, we describe a cysteine-based site-specific crosslinking protocol for the Bax protein in the mitochondrial membrane. We discuss the expected results and the resulting structural-functional models for the pore-forming Bax oligomer, in comparison with other crosslinking approaches that have been used to study other mitochondrial protein complexes. At the end, we highlight the advantages of the crosslinking approaches as well as the limitations and alternative approaches.

1. Introduction

Mitochondria are the most important organelle in all eukaryotic cells. They are factories where many metabolic reactions take place, where many proteins are either produced within from the mitochondrial genome or imported from the cytosol after they are produced based on the cellular genome in the nucleus, where many lipids begin or end their synthesis, and where all the ATP is produced as the most used energy token for the cell. To produce ATP, an electrochemical gradient is generated across the inner membrane of mitochondria, resulting in a proton flow that powers the ATP synthase to covert ADP to ATP. Mitochondria also have an outer membrane. The double membranes segregate the mitochondria to two compartments, the matrix surrounded by the inner membrane and the intermembrane space between the inner and outer membranes, in which different reactions occur generating different products. The two membranes also carry out different reactions and have their own goods to make.

It is hence not a surprise that the permeability of both mitochondrial membranes is tightly controlled to allow certain, but not other, molecules to cross. The examples are voltage-dependent anion channel (VDAC) in the outer membrane and adenine nucleotide translocator (ANT) in the inner membrane that have regulated conductance for ions, metabolites and ATP-ADP exchange. The ATP synthase also has a channel to let protons flow through, producing a force that spins a transmembrane rotor to power the enzyme. In contrast to these integral membrane proteins that have a small channel to let ions or small molecules to cross the membranes in a controllable way, i.e., the voltage-gated ion channel of VDAC, there are large pores in the mitochondrial membranes. The translocases in the outer and inner membranes (TOM and TIM) contain pores that are large enough to allow proteins, at least in their extended unfolded state, to move from the cytoplasm to the mitochondrial intermembrane space or matrix. Comparing to these large pores that are part of the healthy cells, even larger pores are formed in the stressed cells. For example, the mysterious mitochondrial permeability transition pore, and the deadly pores formed by Bax and Bak proteins of the Bcl-2 family in the outer membrane that can be tens or even hundreds of nanometers wide in the tubular mitochondria whose dimension is in the order of micrometer (McArthur et al., 2018; Salvador-Gallego et al., 2016).

These mega pores are formed by oligomers of the same protein, i.e., Bax. Because the pores are oligomeric in nature, and heterogeneous in size, conventional structural biology approaches such as X-ray crystallography and nuclear magnetic resonance are not very useful to solve their three-dimensional structures in toto. The revolutionary cryo-electron microscopy is useful to review structures of large complexes that are homogeneous in size and shape. As an example, Gasdermin pore structure that may be formed in the mitochondrial membranes to release mitochondrial proteins and DNA to trigger inflammatory cytokine secretion and cell death was determined recently (Ruan, Xia, Liu, Lieberman, & Wu, 2018). In comparison to the Gasdermin pore that is formed by 27 or 28 Gasdermin molecules and has a diameter of 18 nm, Bax pore sizes are much larger and much more variable, perhaps formed by tens to hundreds of Bax proteins that interacted via variable regions for variable durations (Cosentino & Garcia-Saez, 2017). Thus, even cryo-electron microscopy may not be good enough to solve their structures.

Fortunately, crosslinking techniques come to the rescue, since they can detect the proximity between two sites in the same protein and hence the conformation and dynamics of the protein, or the proximity of two or more proteins in a complex and hence the protein-protein interaction and dynamics. More importantly, they can monitor the conformation and interaction as well as the dynamics of the proteins and complexes in situ, e.g., in the mitochondrial outer membrane where the conformationally dynamic Bax proteins form the structurally dynamic oligomeric pore. And, they are relatively cheaper, easy and fast to do as we and other have demonstrated. For example, we used site-specific photo and chemical crosslinking to investigate the homo-oligomeric Bax interactions in the mitochondrial membranes as well as the interactions of Bax with other proteins of the Bcl-2 family (Chi et al., 2020; Ding et al., 2014; Zhang et al., 2016). Kluck and colleagues analyzed the Bak oligomerization in the mitochondria isolated from cells using disulfide crosslinking (Dewson et al., 2008; Uren et al., 2017). Walensky laboratory coupled chemical crosslinking with mass spectrometry to study the Bax oligomers induced by detergent (Hauseman et al., 2020).

Here, we describe a cysteine-based site-specific crosslinking protocol that we have used to study the pore complex formed by active Bax proteins in the mitochondria isolated from animal tissue. We discuss the crosslinking data that we have used to model the structures of different dimeric interfaces in the oligomeric pore. We outline cautions that one must take when interpreting the crosslinking data in order to draw correct conclusions about the protein interaction detected by the crosslinking. Furthermore, we compare different crosslinking methods that have been used to study other mitochondrial protein complexes. At the end, we list the advantages and limitations of the crosslinking approach, and the alternative approaches that can also be used to study functional protein interactions in the mitochondrial membranes.

2. Materials and equipment

2.1. Materials

2.1.1. Reagents for molecular biology

• Vectors for construction of DNA plasmids for in vitro protein synthesis, e.g., pSPUTK (Stratagene) or pGEM series (Promega). These vectors contain promoters for T7 and SP6 RNA polymerases that flank a multiple cloning region where a protein-coding DNA sequence can be inserted. In vitro transcription of the resulting plasmid generates the mRNA. And in turn, in vitro translation of the mRNA generates the protein

• PCR reagents to produce the protein-coding DNA sequence with restriction endonuclease sites on both ends

• Restriction endonucleases, and DNA ligase to clone the protein-coding DNA sequence into the vector

• E. coli strains, growth media, antibiotics, transformation reagents, and plasmid DNA preparation kit (QIAprep Spin Miniprep kit, Qiagen) to produce the protein-coding plasmid

2.1.2. Reagents for protein synthesis

Note: Nuclease-free ultrapure water (NF H₂O) is obtained from Milli-Q Direct 8 water purification system (Millipore), autoclaved in bottles that are baked at 210 °C for overnight, and stored at room temperature (~22 °C), unless indicated otherwise. All solutions are prepared in NF H₂O with chemicals of the highest purity from Sigma or Thermo Fisher Scientific, filtered with 0.2 μm filter, and stored at 4 °C, unless indicated otherwise.

• Reagents for in vitro transcription to produce mRNAs, and in vitro translation to produce proteins are previously described (Lin, Johnson, & Zhang, 2019).

• Transcription/Translation-coupled (TNT) wheat germ extract system (Promega), see 3.1.3

• Cycloheximide, 200 μg/mL, prepare 1 mL, aliquot 10 μL for multiple uses, store at −20 °C

2.1.3. Reagents for activation and localization of proteins to mitochondria

• BH3 peptide: containing Bax residues 53–86 as described before (Tan et al., 2006), solubilized 10 mM in DMSO, dilute to 370 μM with NF H₂O, aliquot 30 μL for single use, store at −80 °C

• Mitochondria: isolated from bak knockout mouse liver as described (Pogmore, Pemberton, Chi, & Andrews, 2016), aliquot 10 μL for single use, stored at −80 °C

• AT-KCl buffer, prepare 50 mL according to Table 1, store at 4 °C

• N-ethylmaleimide (NEM), 1 and 0.4 M in 100% ethanol, prepare 1 mL in a 1.5-mL microfuge tube, wrap the tube with aluminum foil, store at 4 °C

  Note: Stop use the NEM solution when the color changes to green.

• Succinate, 1 M, prepare 10 mL, store at 4 °C

• AT-KCl-Succinate buffer, prepare 50 mL by mixing 0.25 mL 1 M succinate with 49.75 mL AT-KCL buffer so that [succinate] = 5 mM, store at 4 °C

• ATP, 0.1 M, prepare 1 mL, aliquots 10 μL for single use, store at −80 °C

• Phosphocreatine (CP), 4 mM, prepare 1 mL, aliquots 10 μL for single use, store at −80 °C

• Creatine phosphokinase (CPK), 4 mg/mL, prepare 1 mL, aliquots 10 μL for single use, store at −80 °C

• DTT, 50 mM, prepare 1 mL, aliquots 20 μL for single use, store at −20 °C

• Low Salt buffer, prepare 50 mL according to Table 2, filter with 0.2 μm filter, store at 4 °C

Table 1.

AT-KCl buffer.

Component	Volume (mL)	Final concentration
NF H2O	29.8
1 M Trehalose	15	300 mM
1 M HEPES, adjust to pH 7.5 with 5 M KOH	0.5	10 mM
1 M KCl	4	80 mM
0.5 M EGTA, adjust to pH 7.7 with 5 M NaOH	0.1	1 mM
0.5 M EDTA, adjust to pH 8.0 with 5 M NaOH	0.1	1 mM
10% (w/v) BSA	0.5	0.1% (w/v)
Final Volume	50

Table 2.

Low salt buffer.

Component	Volume (mL)	Final concentration
NF H2O	47.32
1 M HEPES, adjust to pH 7.5 with 5 M KOH	1.25	25 mM
4 M CH3COOK, adjust to pH 7.5 with 5 M KOH	1.38	110 mM
1 M Mg(CH3COO)2	0.05	1 mM
Final Volume	50

2.1.4. Reagents for protein crosslinking

• NaAsO₂, 0.2 M, prepare 1 mL, store at 4 °C

• Copper (II)(1,10-Phenanthroline)₃ (CuPhe) solution, prepare according to Table 3, aliquot 0.5 mL in microfuge tubes for multiple uses, wrap the tubes with aluminum foil, store at 4 °C

• Quench solution-I, freshly prepare according to Table 4, enough to stop 20 crosslinking reactions in 3.3.1, keep at room temperature before use

• Quench solution-II, prepare in a 15-mL tube according to Table 5, wrap the tube with aluminum foil, store at 4 °C

• Sample buffer, prepare according to Table 6, store at 22 °C

• CCl₃COOH, 25% (w/v), prepare 100 mL, store at 4 °C

• Acetone-HCl solution, mix 90 mL acetone and 1 mL 0.1 M HCl, store at 4 °C

• Reagents for SDS Polyacrylamide Gel Electrophoresis (SDS-PAGE) are previously described (Lin et al., 2019)

• [Methyl-¹⁴C] methylated protein molecular weight markers (PerkinElmer)

• PageRuler prestained protein ladder (Thermo Fisher Scientific)

Table 3.

CuPhe solution.

Component	Quantity	Final concentration
NF H2O	8 mL
100% ethanol	2 mL	20% (v/v)
CuSO4	0.075 g	30 mM
1,10-Phenanthroline	0.18 g	100 mM
Final Volume	10 mL

Table 4.

Quench solution-I.

Component	Volume (μL)	Final concentration (mM)
0.5 M EDTA, adjust to pH 8.0 with 5 M NaOH	216	390
0.4 M NEM	55.2	80
0.2 M NaAsO2	5.5	4
Final Volume	276.7

Table 5.

Quench solution-II.

Component	Volume (mL)	Final concentration (mM)
NF H2O	4.9
0.5 M EDTA, adjust to pH 8.0 with 5 M NaOH	4	200
0.4 M NEM	1	40
0.2 M NaAsO2	0.1	2
Final Volume	10

Table 6.

Sample buffer.

Component	Volume (mL)	Final concentration
NF H2O	6.9
1 M Tris	12	240 mM
0.5 M EDTA, adjust to pH 8.0 with 5 M NaOH	1.5	15 mM
100% (v/v) glycerol	15	30% (v/v)
25% (w/v) SDS	14.6	7.3% (w/v)
Add 2 mg bromophenol blue		0.004% (w/v)
Final Volume	50

2.2. Equipment

• Pipettes: Rainin Classic, 1000 μL, 200 μL, 20 μL, and 10 μL

• Pipet-Aid pipette controller: Drummond Scientific

• Centrifuge: Eppendorf, model 5415R, refrigerated with rotor for 1.5-mL microfuge tube

• Water bath incubator: Fisher Scientific, model Isotemp 215

• Dry bath incubator: Fisher Scientific

• Vortex: Fisher Scientific, model Genie 2

• Vacuum concentrator: Thermo Fisher Scientific, model SpeedVAC

• Vertical gel electrophoresis apparatus: GibcoBRL, model V16

• Power supply: Bio-Rad, model PowerPac 3000

• Gyrotory shaker: New Brunswick, model G2

• Gel dryer: Bio-Rad, model 583

• Vacuum pump: Precision Scientific, model DD90

• Phosphor-imager: Fujifilm FLA-9000 multipurpose image scanner

• Phosphor-imaging plate, Fujifilm BAS storage phosphor screen, type MS, 20 × 25 cm

• UV/Vis spectrophotometer: Beckman, model DU520

• Benchtop pH meter: Orion, model Star A111

• Balances: Ohaus, models Standard and Explorer Precision

• Programmable thermal controller: MJ Research, model PTC-100

• Refrigerators: 4 °C, −20 °C, and −80 °C

3. Protocol

3.1. In vitro protein synthesis

Timing: Week 1 and Week 2, Day 1

3.1.1. Construction of DNA plasmids for in vitro synthesis of Bax proteins with no, single or double Cys residues

1. The human Bax DNA was inserted between the NcoI and EcoRV sites in pSPUTK vector (previously available from Stratagene; now available from the corresponding author upon request; can be replaced by other vectors that are currently available, e.g., pGEM series from Promega). As we described previously (Ding et al., 2014; Zhang et al., 2016), the resulting plasmid with the start codon of the Bax DNA overlaps with the NcoI site (CCATGG) and the stop codon 5′ to the EcoRV site was used to generate Bax RNA in vitro through SP6 RNA polymerase catalyzed transcription. And in turn, the Bax mRNA was used to produce Bax protein in vitro through wheat germ extract or rabbit reticulocyte lysate-based translation system. Both translation systems are commercially available (Promega). Alternatively, we used the Bax plasmid to produce the Bax protein in vitro using transcription/translation coupled (TNT) system based on wheat germ extract or rabbit reticulocyte lysate (Promega) (Chi et al., 2020; Zhang et al., 2016). Since the TNT system is the simplest one to produce proteins, we describe it in 3.1.3

2. To construct the plasmids coding for Bax mutants with no, single or double Cys residues, the Bax plasmid described in (a) was mutated, first by changing the two endogenous Cys codons to Ser codons to generate the Cys-null Bax plasmid (Bax C0), and second by changing codon(s) at specified position(s) to Cys codon(s) to generate the single or double-Cys Bax plasmid(s), e.g., the single-Cys Bax I31C, E69C, or E146C or double-Cys Bax L59C/M79C plasmid with the codon(s) for Ile³¹, Glu⁶⁹, Glu¹⁴⁶ or Leu⁵⁹ and Met⁷⁹ changed to the codon(s) for Cys, respectively. QuickChange kit (Agilent) was used to generate these Bax mutant plasmids according to the manufacturer’s protocol. All the resulting plasmids were verified by DNA sequencing.

3.1.2. Expression and purification of DNA plasmids for the single or double-Cys, or Cys-null Bax

1. The plasmids from 3.1.1 were amplified in E. coli DH5α or XL1-Blue cells and purified using QIAprep Spin Miniprep kit (Qiagen). The resulting plasmids in NF H₂O were quantified by measuring A₂₆₀ of 50-fold diluted samples and calculating the concentrations of the undiluted stocks using the following equation:

   $$[PlasmidStock]=A_{260}\times 2.5(\mu g∕\mu l)$$

2. The plasmid stocks were then diluted so that the [plasmid] = 0.15 μg/μl; store at −20 °C for multiple uses.

Critical: The A₂₆₀/A₂₈₀ ratio obtained from the diluted plasmids must be 1.8–2.0 so that the purity of the plasmids is sufficient for the TNT reaction in 3.1.3.

Pause Point: Pause for weekend.

Timing: Week 2, Day 1, Hour 1–2.

3.1.3. In vitro synthesis of single or double-Cys, or Cys-null Bax proteins using the transcription/translation-coupled (TNT) wheat germ extract system (Promega) by following the manufacturer’s protocol. An example is given below

1. Assemble a master TNT reaction in a 1.5-ml microfuge tube on ice for 10 5-μL individual reactions according to Table 7.

2. Aliquot 4.3 μL of the master TNT reaction for each individual reaction to a 1.5-mL microfuge tube, and add 0.7 μL of 0.15 μg/μL plasmid for the single or double-Cys, or Cys-null Bax so that [plasmid] = 0.02 μg/μL.

3. Conduct the TNT reaction in a 30 °C water bath for 90 min to produce the single or double-Cys, or Cys-null Bax protein.

4. Stop the TNT reaction by adding 0.5 μL 200 μg/mL cycloheximide so that [cycloheximide] = 20 μg/mL, and incubating at 22 °C for 10 min.

Table 7.

Master TNT reaction.

Componenta	Volume (μL)	Final concentration
NF H2O	10.8
TNT Wheat Germ Extract	25	1/2 dilution
TNT Reaction Buffer	2	1/25 dilution
TNT RNA Polymerase (SP6)	1	1/50 dilution
Amino Acid Mixture, Minus Methionine, 1 mM	1	20 μM
RNasin Ribonuclease Inhibitor, 40 U/μL	1	0.8 U/μL
[35S]Methionine, 40.2 μCi/μLb	2.5	2 μCi/μL
Gently mix by pipetting, and centrifuge briefly
Final Volume	43.3	For 10 5-μL TNT reaction

^aNote: Add the components to the microfuge tube according to the order in Table 7.

^bNote: [³⁵S]Methionine decays and the radioactivity at the time of TNT reaction is calculated using the following equation:

$$A_t=A_0{e}^{\left(-0.693T∕T_{1∕2}\right)}$$

For this example: A_T = the radioactivity at the time of TNT reaction = 40.2 μCi/μL, A₀ = the radioactivity at the time specified by the manufacturer = 43.5 μCi/μL, T = the time between the manufacturer specified and the TNT reaction = 10 days, and T_1/2 = the time for the radioactivity of [³⁵S] isotope to decay to half of the manufacturer specified activity = 87.4 days.

Critical: Handling RNA: mRNAs and tRNAs for in vitro translation are very labile. Nucleases are secreted by humans, and transferred in our finger oil to any surface we touch. We therefore must avoid any contact between a mRNA or tRNA solution and any surface that may have been touched by us or any solution that may have been contaminated by nucleases. When transcriptions or translations do not work, it is invariably because of a nuclease contamination that degrades the mRNAs and tRNAs. Thus, stringent nuclease-free solutions and techniques are absolutely the most critical requirement for achieving successful transcription and translation.

3.2. In vitro targeting Bax proteins to mitochondria and fractionation

3.2.1. Isolation of mitochondria from bak knockout mouse liver as described before (Pogmore et al., 2016). The resulting mitochondria lack not only Bak protein due to the gene knockout but Bax protein due to its cytoplasmic localization in the liver cells. Therefore, there is no endogenous Bax and Bak in the mitochondria that can form complexes with the single or double-Cys Bax, and reduce the number of Bax complexes that can be crosslinked via the introduced Cys residues.

Timing: Week 2, Day 1, Hour 3–4.

3.2.2. Preparation of mitochondria before use for Bax targeting.

1. 10 min before the TNT reactions complete, take one aliquot of frozen mitochondria in 1.5-mL microfuge tube (~10 μL, [protein] = 50 mg/mL, in AT-KCl buffer without KCl) from −80 °C; thaw the mitochondria by holding the tube between fingers; place on ice.

   Note: 10 μL of the frozen mitochondria are enough for targeting of six different Bax proteins; increase accordingly for experiments with more proteins.

2. Dilute the mitochondria by adding 490 μL AT-KCl buffer so that [protein] = 1 mg/mL; gently mix by pipetting; briefly centrifuge.

3. Treat the mitochondria with NEM to block sulfhydryls in the mitochondrial proteins so they will not interfere with the crosslinking reaction via sulfhydryls of Bax proteins in 3.3: Add 9 μL 1 M NEM to the mitochondria so that [NEM] = 18 mM; incubate on ice for 10 min; centrifuge at 5000g and 4 °C for 2 min; wash the mitochondria pellet with 500 μL AT-KCl buffer by gently pipetting; centrifuge again to pellet the mitochondria.

4. Resuspend the mitochondria pellet with 250 μL AT-KCl-Succinate buffer by gently pipetting until there are no visible clumps of mitochondria; briefly centrifuge; the resulting mitochondria’s [protein] = 2 mg/mL.

5. Add 5 μL 0.1 M ATP, 0.6 μL 4 mg/mL CPK, and 0.6 μL 4 mM CP to the mitochondria so that [ATP] = 2 mM, [CPK] = 10 μg/mL, and [CP] = 10 μM; gently mix by pipetting; briefly centrifuge; keep the mitochondria on ice and use within 2 h.

3.2.3. Activation and localization of Bax proteins to mitochondria

1. To each TNT reaction from 3.1.3, add 35 μL of the mitochondria from 3.2.2, 0.9 μL 50 mM DTT, and 1.7 μL 370 μM BH3 peptide so that the mitochondrial [protein] = 1.6 mg/mE, [DTT] = 1 mM, and [BH3 peptide] = 15 μM; gently mix by tapping the tube; briefly centrifuge; incubate at 37 °C for 60 min to allow the BH3 peptide to activate the soluble Bax proteins resulting in integral Bax proteins that interact to form oligomeric pores in the mitochondrial outer membrane.

2. Separate of the mitochondrial Bax proteins from the soluble proteins by centrifuging the samples from (a) at 13,000g and 4 °C for 5 min; use pipette to transfer each supernatant to a 1.5-mL microfuge tube for analysis of the soluble Bax proteins, if needed; see (Zhang et al., 2016) for an example.

3. Resuspend each mitochondria pellet with the associated Bax proteins in 85 μL low salt buffer by gently pipetting.

4. To each mitochondria sample, add 0.4 μL 0.2 M NaAsO₂ so that [NaAsO₂] = 1 mM; gently mix by pipetting; incubate on ice for 5 min. This treatment decreases residual DTT that otherwise will inhibit the crosslinking in Section 3.3.

3.3. Crosslinking of mitochondrial Bax proteins via cysteine residues

Timing: Week 2, Day 1, Hour 5.

3.3.1. Disulfide crosslinking of the mitochondrial Bax proteins.

1. Split each sample from 3.2.3 to two 40-μL aliquots in 1.5-mL microfuge tubes.

2. To one aliquot from each sample, add 0.4 μL CuPhe solution so that [CuSO₄] = 0.3 mM, and [1,10-phenanthroline] = 1 mM; incubate on ice for 30 min to let CuPhe, a redox catalyst, induce disulfide crosslinking of the Bax proteins.

3. To stop each crosslinking reaction, add 10 μL of freshly prepared Quench solution-I so that [EDTA] = 77 mM, [NEM] = 15.5 mM, and [NaAsO₂] = 0.8 mM; incubate at 37 °C for 10 min. The resulting samples contain Bax proteins that are crosslinked for “30 min.”

4. To the other aliquot from each sample, add 10 μL Quench solution-I to block the disulfide crosslinking before adding 0.4 μL of CuPhe solution; incubate on ice for 30 min and then at 37 °C for 10 min. The resulting samples contain Bax proteins that are crosslinked for “0 min,” serving as the negative controls.

5. Add 65 μL 25% (w/v) CCl₃COOH to each sample from (c) and (d) and incubate on ice to precipitate the proteins.

Critical: 3.3.1 must be done as soon as possible to avoid oxidation of the samples by oxygen in the solutions prior to the crosslinking reactions, which increases the background crosslinked protein bands in the “0 min” controls on the phosphor-images of SDS-PAGE gel.

Pause Point: Pause for overnight.

3.4. SDS-PAGE and phosphor-imaging of crosslinked Bax proteins

Timing: Week 2, Day 2, Hour 1–6.

3.4.1. Sample preparation for SDS-PAGE.

1. Precipitate the proteins in the samples from 3.3 by centrifuging at 13,000g and 4 °C for 15 min; remove the supernatants by pipetting; wash each protein pellet with 350 μL Acetone-HCl solution by vortex; centrifuge again and remove the supernatants; dry the protein pellets in vacuum concentrator at 22 °C for 20 min.

2. Mix 250 μL Quench solution-II with 250 μL Sample buffer at 22 °C.

3. Resuspend each pellet in 25 μL of the Quench solution-II/Sample buffer mixture by vortex; briefly centrifuge.

4. Aliquot 10 μL from each “30 min” sample; add 1 μL 2-mercaptoethanol so that [2-mercaptoethanol] = 10% (v/v); vortex and briefly centrifuge, resulting in the reduced “30 min” samples.

5. Mix 1 μL of [Methyl-¹⁴C] methylated protein molecular weight markers with 9 μL of Quench-Sample Buffer.

6. Incubate all the “30 min” crosslinked and reduced, “0 min” crosslinked samples, and the protein markers at 65 °C for 30 min; briefly centrifuge.

3.4.2. SDS-PAGE

1. Set up SDS-PAGE apparatus according to laboratory or manufacture protocol.

   Note: We make gels for SDS-PAGE according to Lin et al. (2019).

2. Load 10 μL of the samples from 3.4.1 per well in gel.

Critical: Separate the reduced sample set from the nonreduced crosslinked sample set and the protein markers by a well loaded with some leftover mitochondria from 3.2.3 solubilized in Quench solution-II/Sample buffer mixture without 2-mercaptoethanol as described in 3.4.1. Otherwise, the crossover reducing agent from the reduced samples will affect the migration or integrity of the crosslinked samples during SDS-PAGE.

Alternative: Nonradioactive PageRuler prestained protein ladder (5 μL mixed with 5 μL of Quench solution-II/Sample buffer mixture) can be used instead of the radioactive protein molecular weight markers. Since the bands of these proteins are visible on the gel, their positions can be painted by radioactive ink made with the leftover samples from 3.4.8 that contains radioactive Bax proteins after the gel is dry on a filter paper. The radioactive paint allows the positions of the non-radioactive protein ladder to be detected by phosphor-imaging in 3.4.10.

• (c)Perform SDS-PAGE according to laboratory or manufacture protocol to separate the crosslinked Bax proteins from the monomers.

  Note: We perform SDS-PAGE according to Lin et al. (2019).

Timing: Week 2, Day 2, Hour 7–8; and Day 3–5.

3.4.3. Phosphor-imaging

1. Perform phosphor-imaging according to Lin et al. (2019) to detect the radioactive Bax proteins and their crosslinked products. A phosphor-image acquired from Fujifilm FLA-9000 Multipurpose Image Scanner controlled by Fujifilm Image Reader FLA-9000 Program is shown in Fig. 1A.

2. Analyze the phosphor-image using Fujifilm Multi Gauge Program. A structural model that fits the crosslinking data from the Bax I31C mutant is shown in Fig. 1B. Other structural models that fit the crosslinking data from Bax E69C and L59C/M79C mutants are described previously (Zhang et al., 2016).

Fig. 1.

Crosslinking of the mitochondrial Bax proteins and a dimer model for Bax α-helix 1. (A) The in vitro synthesized [³⁵S]Met-labeled single-Cys (I31C, E69C or E146C) or double-Cys (L59C/M79C), or Cys-null (C0) Bax protein was activated and localized to the mitochondria that were then isolated. CuPhe solution was added to the resulting mitochondria to induce disulfide crosslinking of the Bax proteins for 30 min, and Quench solution was then added to stop the crosslinking reactions, resulting in the “30 min” crosslinked samples. For the “0 min” crosslinked controls, Quench solution was added before the addition of CuPhe. All the samples and controls were analyzed by non-reducing SDS-PAGE and phosphor-imaging. In the resulting phosphor-images, the closed circles indicate Bax monomers, the arrows indicate disulfide-linked Bax dimers, and the molecular weights (MW) indicate the positions of prestained protein ladder. (B) A model for a dimer formed two parallel Bax α-helix 1 structures was generated computationally to fit the crosslinking data obtained from Bax I31C shown in panel A and other Bax mutants with Cys residues positioned in the α-helix 1 (data not shown) using the method we described before for modeling of another dimer formed by Bax α-helices 2, 3, and 4 to fit the crosslinking data from Bax E69C and other mutants; see fig. 3 in Zhang et al. (2016). PyMOL Molecular Graphics System (Schrödinger, LLC) was used to depict the model. The two helices are indicated by α1 and α1′. The termini of α1 are labeled by ${\mathsf{NH}}_3^{+}$ and COO⁻. The amino acid residues are shown as thin lines, except of the residues, I31 in α1 and I31′ in α1′, which formed a disulfide bond after they were replaced by Cys are shown as thick sticks with their α carbon atoms linked by a dashed line with the distance indicated. For Discussion below, the L27 and L27′ are depicted as the I31 and I31′.

4. Results and discussion

The most intuitive interpretations of the crosslinking data such as that shown in Fig. 1A are that, if a crosslinked band is detected for two proteins, the proteins form a dimer via the regions where the two Cys residues are located. The stronger the crosslinked band, the stronger the protein-protein interaction, and the more stable and longer lasting protein complex. If no crosslinking band is detected, the proteins do not interact via the Cys-containing regions. However, one should be cautious about these interpretations based on the following considerations.

The crosslinking between two proteins would occur when a Cys in one protein is close to a Cys in other protein in a complex formed by the two proteins. For example, a disulfide crosslinking occurs between two single-Cys Bax E69C proteins, each with a Cys replaced Glu⁶⁹ in the sequence, as shown in Fig. 1A, lane 8. As expected, the apparent molecular weight of the crosslinked proteins is about twice of each protein’s molecular weight. The crosslinked dimer shifts to the monomer when the sample is reduced as described in 3.4.1 d) before SDS-PAGE as expected from a disulfide-linked dimer; see this data in fig. 3A and the related Appendix fig. S2B of Zhang et al. (2016). The crosslinking does not occur in the “0 min” crosslinked control (lane 7) so the oxidation of the sulfhydryls of the Cys residues is required for the disulfide bond to form. No crosslinking is detected with the Cys-null Bax C0 protein (lanes 3–4) so the crosslinking is dependent on the Cys residue. The disulfide crosslinking also occurs between two double-Cys Bax L59C/M79C proteins, each with two Cys substituted the Leu⁵⁹ and Met⁷⁹, and requires the oxidation and dependent on the Cys residues as well (lanes 9–10). The other single-Cys mutant, Bax I31C or E146C, also generates a disulfide crosslinked dimer (lanes 1–2 or 5–6), albeit at much lower levels than the E69C dimer.

The negative “0 min” crosslinked control is necessary for one to claim the positive crosslinking detected in the respective “30 min” crosslinked sample. Occasionally, one may see a crosslinked protein band in the “0 min” control. An interpretation of this false positive data would be that the control sample is oxidized such that the proteins are crosslinked before the Quench solution is added in 3.3.1 d) to block the crosslinking. To avoid this, on should use the fresh-made sample under reducing condition, which is Critical as noted in 3.3.1. Alternatively, the Quench solution may deteriorate, thereby unable to block the crosslinking. To counter that, we prepare the Quench solution immediately prior to the use in 3.3.1 mainly because one of the components, NEM, is unstable in aqueous solution and sensitive to light, although the NEM stock in ethanol is stable at 4 °C without light.

We have never seen a crosslinked band from the Cys-null Bax. However, if a Cys-null protein generates a “crosslinked” band in the “30 min” lane of a non-reducing SDS-PAGE gel, one should first check whether the band disappears in a reducing SDS-PAGE gel. Second, one should check whether the band also appears in the “0 min” control of the Cys-null protein. Because if the answer is no from the first check, but yes from the second, the band is not generated by the disulfide crosslinking reaction, but rather by other means. For example, the TNT reaction may produce not only the Cys-null protein but a mysterious protein that has the same molecular weight as the crosslinked Cys-null proteins. The mysterious protein and other potential background proteins that the TNT may produce can be easily revealed by running a control TNT without the Cys-null mRNA.

Based on typical geometries of disulfide bonds determined from a survey of known protein structures (Ozhogina & Bominaar, 2009), one may conclude that the crosslinked Cys residues are less than 8 Å apart in terms of the distance between their α-carbons. For example, the distance between the α-carbons of two Cys residues that replace the I31 and I31′ would be 6.8 Å according to the structural model in Fig. 1B. As we documented in fig. 1 of Zhang et al. (2016), the crosslink data from the L59C/M79C and other Cys-containing Bax mutants could be fit by the structure of a symmetrical dimer formed by two Bax regions containing the α-helices 2 to 5, which was determined by crystallography (Czabotar et al., 2013). This led to the conclusion that two active Bax proteins in the mitochondrial outer membrane form a symmetrical dimer structure via their α-helices 2 to 5 regions, similar to the crystal structure (PDB ID: 4BDU). In contrast, the crosslinking via the two E69C residues is unexpected since their α-carbons are 12.2 Å apart in the crystal structure. This crosslinking data suggests that the conformation of the dimer formed by the α-helices 2 to 5 of full-length Bax protein at the mitochondrial membrane is more dynamic than the crystal structure such that the distance between the two E69C residues can be less than 8 Å at least for some time. Alternatively, the dimer may rearrange after the α-helix 5 is released from the dimer and inserts into the membrane, resulting in a dimer without the α-helix 5 in which the two E69C residues are closer than those in the α-helices 2 to 5 dimer. See fig. 3 of Zhang et al. (2016) for more evidence and logic for the dimer transition and a structural model for the dimer without the α-helix 5.

By comparing the arrow-indicated bands in lanes 8 and 10 of Fig. 1A, one may see that the two Bax E69C proteins that are linked by the single disulfide between two Cys⁶⁹ residues migrate slower in the SDS-PAGE gel than the two Bax L59C/M79C proteins that are linked by the double disulfides, one between Cys⁵⁹ and Cys^79,, and other between Cys⁷⁹ and Cys^59’. Note that similar to the depiction of the residues in the two protomers of the α-helix 1 dimer in Fig. 1B, the two Cys residues without the prime symbol are of the first protomer in the α-helices 2 to 5 dimer, whereas the two Cys residues with the prime symbol are of the second protomer. These data are expected with two SDS-charged protein complexes that are crosslinked differently and hence have different shapes that run through the gel matrix with different speeds in an electrical field. These data further support the symmetrical dimer model in which the two α-helices 2 to 3 regions are antiparallel (Zhang et al., 2016). Thus, careful analysis of the crosslinking data from a carefully designed experiment can reveal more structural information about a protein complex.

Moreover, the model for another Bax dimer via α-helix 9 is shown in fig. 7 of Zhang et al. (2016). This model is not based a known structure but on the crosslinking data from Bax proteins with Cys residues in the α-helix 9. A computational transmembrane α-helical dimer model built on the conserved GxxxG/A dimerization motif in α-helix 9 fits the crosslinking data. Thus, crosslinking and computational modeling can generate a high-resolution dimer model without a known structure determined by X-ray crystallography, NMR or cryo-EM. With this said, atomic resolution structures of protein complexes determined by structural biologists are highly valuable to design of crosslinking experiments. For example, based on the structures one can position the Cys residues precisely to where the crosslinking can occur to bridge the interacting proteins. Even if a structure that is not of a protein complex but of the protein monomer that can form the complex, it is still useful since one can position the Cys residues on the surface that one suspects the binding would occur. Without any structures, one must fish out the interacting regions by position Cys or other reactive residues such as Lys throughout the interacting proteins, performing tedious crosslinking experiment, and hoping to detect the dimer from some Cys or Lys residues. Then one can structurally model the protein complex, and further test the structural model by more crosslinking with better positioned Cys or Lys residues guided by the model. On the other hand, the crosslinking approach can validate the crystal, NMR and cryo-EM structures to see if they represent the structures form by the full-length proteins and in the cognate membranes. Since most of the crystal, NMR and cryo-EM structures are determined without membranes and of truncated proteins, this validation is essential before one commits more resources to assess functional implications of the structures.

As discussed in context of Bax, crosslinking data can lead to structural models for dimers formed by different regions of an oligomerizing protein. To build a model for an oligomer from two or more dimer models, a first principle is that the oligomer is formed by a protein via multiple dimer interfaces that can be the same or different. For example, using A-B to represent a protein with two binding surfaces, A and B, an oligomer of A-B can be either represented by A-B:A-B:A-B… with the same B:A interface, or by A-B:B-A:A-B… with two different B:B and A:A interfaces. We have used this first principle to build an oligomer model for Bax pore in the mitochondrial membrane as shown in fig. 11 of Zhang et al. (2016). We have validated the oligomer model by crosslinking of Bax proteins via Cys residues located in both α-helices 2 to 5 and α-helix 9, and observing the oligomers crosslinked via both dimer interfaces. Importantly, we have demonstrated the functional relevance of the oligomer model by structure-guided mutagenesis that is expected to disrupt each dimeric interaction, followed by mitochondrial membrane permeabilization assay that reveals the mutations’ impact on Bax pore formation.

The distances between the dimers in an oligomer can be estimated from the dimension of the protein monomer and the relative location of the binding surfaces (e.g., the distance between A and B in the protein A-B). If the monomer structure is not known or the structure changes greatly when the monomers form the dimer, one must determine the distance between the binding surfaces experimentally. One may use longer molecular rulers to measure this intramolecular distance, e.g., crosslinking two residues by a bifunctional crosslinker with long spacer arm up to 54 Å (i.e., PEG12-SPDP from Thermo Scientific Pierce), DEER between two paramagnetic centers (up to 60 Å for membrane proteins and 100 Å for soluble proteins (Jeschke, 2012)), and FRET from a donor to an acceptor (up to 170 Å using Thermo Fisher Scientific Alexa Fluor 594/647 donor-acceptor pair with the Forster distance R₀ = 85 Å (Johnson, 2005)).

As mentioned above, the crosslinking band generated by Bax I31C or E146C is weaker than that generated by Bax E69C. In general, reasons for “weak” vs “strong” crosslinking include:

1. A smaller number of a mutant protein is localized to the mitochondria than another mutant. Since the number of protein complexes that are formed by the proteins are dependent on the number of the proteins, the fewer the proteins in the mitochondria, the fewer the protein complexes, and hence, the fewer of the proteins that are crosslinked. This is an interpretation of the data from Bax I31C since fewer proteins are in the mitochondria subjected to the crosslinking, judged by the lower intensity of the protein monomer band in the gel compared to the E69C monomer band. Normalization of the dimer band intensity by the total protein (dimer plus monomer) band intensity can only partially reduce the apparent difference between the crosslinking efficiencies of the two mutants but not between the monomer-dimer equilibria that are determined not only by the binding constant but the concentration of each mutant in the mitochondria.

2. The interaction via an interface is weaker than that via another interface. This could be the interpretation of the data from Bax E146C because similar numbers of the E146C and E69C mutants are localized to the mitochondria, judged by the similar intensities of their monomer bands. Since the number of the E146C dimer is less than that of the E69C dimer as judged by the intensities of their dimer bands, the dimeric interaction via the E146C-containing regions would be weaker than that via the E69C-containing regions. A similar conclusion was drawn for the interactions of Bak, a structural and functional homolog of Bax, in part based on the crosslinking data (Uren et al., 2017). Thus, the interaction between the core regions of Bak is stronger than that between the other regions.

In general, a stable, and hence static, structure of a complex would increase the crosslinking between two proteins in the complex, resulting in more crosslinked dimers that can be detected by SDS-PAGE and phosphor-imaging, western blot or dye staining. On the other hand, a less stable, and hence dynamic or transient, structure of a complex would decrease the crosslinking, resulting in fewer crosslinked dimers. This is how a crosslinking approach could distinguish these different protein complexes. However, whether one sees more or fewer crosslinked dimers on a gel would also depend on other factors, e.g., (1) above, and (3–4) below. Thus, to one extreme, the crosslinking would not distinguish the complexes. But to the other extreme, if all the other factors are the same, such that in a well-controlled experiment, a protein complex is more or less stable due to a point mutation that does not change any other properties of the interacting protein, the crosslinking would distinguish the wild-type complex from the mutated one.

• (3)Similar to other point mutations, a Cys mutation in a region may weaken the protein interaction, whereas another Cys mutation in another region may not. To ascertain the Cys mutations do not adversely affect the protein function that is dependent on the protein interaction, one should assay the function of the mutants and compare to that of the wild-type protein. For example, we determined the function of Cys-containing Bax mutants in the mitochondrial outer membrane permeabilization before we used the mutants in the crosslinking experiments to determine the interaction of Bax proteins as shown in figure EV1 of Zhang et al. (2016).
• (4)Two Cys residues are worse positioned than the other two Cys residues to detect the same interaction. Since the efficiency of a disulfide formation is dictated by the geometry of two Cys residues, a weaker crosslinking is expected when the two Cys residues are in a less favorable geometry for the disulfide formation. Using the structural model for Bax α-helix 1 dimer as an example, the I31C and I31C′ residues are 6.7 Å apart, perhaps too far away to form a disulfide bond efficiently. One may place the Cys residues to other locations that are closer, e.g., L27C and L27C′ that are 4.2 Å apart, and test if this increased the disulfide formation. The redox state of the environment for the Cys residues also dictates the efficiency of the disulfide formation. Thus, if two sulfhydryls are located in a reduced environment that cannot be oxidized, a disulfide will not form.

Knowing these caveats, one should not conclude whether the dimerization of one protein is of a higher affinity than the dimerization of other protein solely based on the crosslinking band intensity from the one is higher than the other. Since the two Cys-containing proteins are different, and they are both different from the wild-type protein, one should not conclude that the protein-protein interaction via one dimer interface is stronger than the interaction via other interface because the crosslinked band via the two Cys residues positioned in the one interface is stronger than the crosslinked band via the two Cys residues in the other interface.

With all the cautions that one must take to correctly interpret the crosslinking data, a crosslinking study will be useful in the following ways. As mentioned above, a crosslinking study can validate a high-resolution structure of a protein complex. In particular, if the structure is determined form part of the protein, the crosslinking study with the whole protein can determine whether the partial structure is used by the whole protein in the complex formation. More relevant to membrane proteins, such as the mitochondrial Bax protein, the structures are usually determined in solution or in crystals that lack membranes. These structures must be validated by experiments that use the proteins in the membranes, such as the crosslinking experiment with the Bax protein in the mitochondrial membrane. Moreover, there is no atomic resolution structure for any Bax pore in any membrane, so the crosslinking approach has been useful to generate structural models for the pores, which allow further elucidate the structure and function of the pore. Expanding the scope, various crosslinking approaches have been used to study other mitochondrial protein complexes and their dynamic structures and functions such as TIM (Banerjee, Gunsel, & Mokranjac, 2017), and TOM (Shiota et al., 2015; Shiota, Mabuchi, Tanaka-Yamano, Yamano, & Endo, 2011) with the results matched well with the later cryo-EM structures (Araiso et al., 2019; Araiso, Imai, & Endo, 2020; Bausewein et al., 2017; Tucker & Park, 2019). Even more powerful, when crosslinking approaches are combined with mass spectrometry, the mitochondrial protein interactome has been mapped at high structural resolutions, e.g., electron transport chain complexes and ATP synthase (Schweppe et al., 2017). Therefore, we are on shining crossroads of science toward full molecular and even atomic knowledge of a super important organelle that entered our cells as endosymbiotic bacteria 1.45 billion years ago (Martin & Mentel, 2010). No matter which road we forward and which vehicle we drive, fruitful reward will be ample. So, TAKE OFF and ENJOY!

5. Advantages

• Protein production by TNT is faster and cheaper than that by expression in cells

• The radioactive proteins produced from TNT can be detected by phosphor-imaging even at very low levels comparing to immunoblotting that is a less sensitive detection method and required specific antibodies

• Disulfide crosslinking uses “zero length” crosslinker. Longer crosslinkers that can link sulfhydryls are available (e.g., from Thermo Fisher Scientific) to crosslink Cys residues that are more distal in the interface. As an example, we use bismaleimidohexane (BMH), a homo-bifunctional crosslinker with two sulfhydryl-reactive moieties linked by a spacer arm of 13.8 Å, to crosslink Bax and Bim, a Bax activator, via Cys residues placed in the suspected binding surfaces (Chi et al., 2020).

• Hetero-bifunctional crosslinkers with a sulfhydryl-reactive group at one end and a photo-reactive group at the other end (available from Thermo Fisher Scientific) can be used to crosslink two interacting proteins of which one contains a Cys and the other does not. In addition, since photo-crosslinkers are less selective reacting with almost any atoms nearby as discussed above, they are useful to initially screen the binding regions before Cys-based crosslinking that can provide more structural information about the binding (Zhang et al., 2010).

• Distance constraints provided by various crosslinking methods are useful to model interface structures in protein complexes using inhouse or online structure computation programs (Ding et al., 2014; Zhang et al., 2016). The resulting interface structural models can feed molecular dynamics simulations that can refine the models or explore the interaction landscapes (Liao et al., 2016). For more information regarding computation and simulation of interacting protein structures, see (Ozdemir, Nussinov, Gursoy, & Keskin, 2019; Slater, Miller, & Kontoyianni, 2020).

• Cys serves as a specific labeling site in protein, useful for probing the protein conformation and environment with other techniques that requires site-specific labeling, e.g., fluorescence spectroscopy with fluorescent dyes (Johnson, 2005), EPR or NMR spectroscopy with paramagnetic probes, and accessibility to water soluble or lipophilic labeling agents. These techniques are complementary to the crosslinking and hence can verify and expand the research

6. Limitations and alternatives

• The phosphor-imaging requires expensive equipment

• Radioactive materials require license to use and cautions to handle

• The mitochondria from a specific gene-knockout mouse will be clean so one does not worry about the endogenous protein competing with the Cys mutant protein for complex formation, which would reduce the crosslinking via the Cys. One may use the wild-type mouse if the Cys mutant can still form enough complexes for the crosslinking. Or, one may use liposomes without any proteins, but they are not native membranes. If one wants to detect the crosslinking of the Cys mutant with the endogenous wild type protein that does not have any Cys or have Cys but it is not in the right position for the crosslinking, one may use a hetero-bifunctional crosslinker with a sulfhydryl reactive group at one end and a photo reactive group such as aryl azide (e.g., phenyl, hydroxyphenyl, or nitrophenyl azide) or diazirine at the other end, so that the two proteins can be linked by the crosslinker via the Cys in the mutant and an atom in the wild type protein that can be reached by the crosslinker. Note that when aryl azide is activated by ultraviolet photon, it forms a nitrene group that can initiate addition reactions with double bonds, insertion into C-H and N-H sites, or subsequent ring expansion to react with a nucleophile (e.g., primary amines). Photo-activation of diazirine creates reactive carbene intermediates that can form covalent bonds through addition reactions with any amino acid side chain or peptide backbone. These reagents are available from Thermo Fisher Scientific

• The crosslinking results are qualitative as discussed in Section 4. In comparison, FRET-based in vitro binding assay is quantitative (Lovell et al., 2008; Pogmore et al., 2016). FRET-based microscopy can even visualize and quantify protein interactions in live cells (Osterlund, Liu, & Andrews, 2015). However, it is difficult to distinguish the interactions via different binding interfaces in an oligomeric protein complex using FRET. Using the Bax oligomer as an example, when a FRET is detected between two fluorescent Bax proteins, is the binding mediated by the two α-helices 2 to 5 regions or the two α-helix 1 regions?