Resolving Affinity Purified Protein Complexes by Blue Native PAGE and Protein Correlation Profiling

Most proteins act in association with others; hence, it is crucial to characterize these functional units in order to fully understand biological processes. Affinity purification coupled to mass spectrometry (AP-MS) has become the method of choice for identifying protein-protein interactions. However, conventional AP-MS studies provide information on protein interactions, but the organizational information is lost. To address this issue, we developed a strategy to unravel the distinct functional assemblies a protein might be involved in, by resolving affinity-purified protein complexes prior to their characterization by mass spectrometry. Protein complexes isolated through affinity purification of a bait protein using an epitope tag and competitive elution are separated through blue native electrophoresis. Comparison of protein migration profiles through correlation profiling using quantitative mass spectrometry allows assignment of interacting proteins to distinct molecular entities. This method is able to resolve protein complexes of close molecular weights that might not be resolved by traditional chromatographic techniques such as gel filtration. With little more work than conventional AP-geLC-MS/MS, we demonstrate this strategy may in many cases be adequate for obtaining protein complex topological information concomitantly to identifying protein interactions.


Introduction
In cells, most proteins perform their functions through transitory protein-protein interactions or by forming stable protein assemblies. Characterizing protein interactions is crucial for fully understanding cellular processes. Affinity purification in combination with mass spectrometry (AP-MS) is one of the most commonly applied strategies to identify native protein interactions. Significant improvements in instrument capabilities achieved in the last decade have made this approach extremely powerful. It is important to note that the interactions identified by AP-MS experiments include a mixture of direct and indirect associations between bait and preys. In addition, often proteins take part in several different complexes within the same cellular context, which might have different biological roles, and therefore the interactors that are identified by AP-MS might represent a mix of distinct protein assemblies or functional entities. It is not possible to derive such topological information a priori from the mono-dimensional lists of proteins generated by simple AP-MS experiments. However, the technique can be exploited further to define the architecture of protein complexes by combining it with one or more methods to resolve these assemblies.
In order to resolve the topology of protein interactions identified through AP-MS, several strategies have been applied. One approach is to perform iterative AP-MS experiments using the preys identified in a previous round of experiments as baits 1 . Although very informative, this is quite a labor intensive task both experimentally and analytically. Protein cross-linking in combination with mass spectrometry is increasingly being used to derive topological information on protein complexes 2,3,4,5 . However, computational analysis of cross-linked peptides still remains a challenging task and hence is the bottleneck in the workflow. The advent of MS-cleavable cross-linking reagents should facilitate mapping of the amino acid residues that are in close proximity in interacting proteins 6,7 . Another alternative is to combine affinity purification with prior orthogonal separation techniques 8,9 . Chromatographic fractionation by gel filtration or ion exchange, or sucrose gradient fractionation have also recently been used in combination with quantitative mass spectrometry to describe multiprotein complexes at a system-wide level, by-passing the complex isolation step 10,11,12,13 . Blue native polyacrylamide gel electrophoresis (BN-PAGE) has been widely applied to investigate native protein interactions, typically those involving mitochondrial membrane protein complexes 14 . This separation technique was also recently used in combination with label-free protein quantification and correlation profiling, not only on mitochondrial complexes 15,16,17 , but also for unravelling other protein complexes from whole cells 18,19 . We hypothesized that the combination of affinity purification with subsequent native fractionation approaches and quantitative MS should provide a useful strategy for resolving multiple protein assemblies containing a particular protein.  1. Thaw the cell pellet in a water bath at 37 °C until it starts to melt. Take the tube out of the bath, swirl it gently until the pellet is completely thawed, and immediately place the tube containing the cell suspension on ice. 2. Add 5 mL of ice-cold lysis buffer (containing DTT and protease inhibitors) to the cell suspension and swirl to mix. Incubate in ice for 10 min. 3. Transfer the lysate to a cold Dounce homogenizer. Lyse with 20-30 strokes using the tight pestle (until there is no noticeable viscosity).

Isolation of Native Protein Complexes by FLAG Affinity Purification
Transfer the homogenate to cold microcentrifuge tubes. 4. Centrifuge at 18,000 x g for 15 min at 4 °C. 5. Transfer the cleared lysate to a clean cold tube. Leave 50 μL of lysate behind. Take a 35 μL aliquot of lysate to measure protein concentration and monitor the purification. 6. Remove the PBS-0.1% Tween-20 from the beads. Resuspend the beads with 1 mL of the lysate and mix with the rest of the lysate.
Incubate the mixture in a rotating wheel at 4 °C for 1-2 h. 7. Collect the beads on the side of the tube with the magnet. Collect a 30 μL aliquot of supernatant and discard the rest of the supernatant. Resuspend the beads in 0.5 mL of IPP150 buffer (10 mM Tris-HCl pH 8, 150 mM NaCl, 1mM EDTA, 0.1% NP-40) by pipetting 4-6 times. Transfer to a 1.5 mL cold tube. 8. Repeat the washes with 0.5 mL of IPP150 buffer two more times. 9. Wash beads three times with 0.5 mL of FLAG native elution buffer (20 mM Bis-Tris pH 7, 20 mM NaCl, 0.02% Nonidet P-40, 1 mM EDTA, 200 mM ε-aminocaproic acid). With the last wash, transfer the beads to a new cold tube. Remove buffer thoroughly. 10. Resuspend the beads in 100 μL of 200 μg/mL 3x FLAG peptide in native elution buffer. Incubate at 4 °C for 10 min with gentle rotation. 11. Collect the beads with the magnet and transfer the supernatant to a cold new tube. 12. Repeat steps 1.3.10-1.3.11 twice. Pool all the eluates. 13. Concentrate the eluate down to 25 μL in a centrifugal filter unit (10 kDa nominal molecular weight limit cut-off, PES) by centrifugation at 10,000 x g at 4 °C. Transfer the concentrated eluate to a new cold tube and add 50% glycerol for a final concentration of 5%. The concentrated eluate can be kept at 4 °C overnight if required.

Mass Spectrometry Analysis
Note: All subsequent steps should be performed in a laminar flow hood if possible to ensure cleanliness of the samples. All solutions should be prepared with HPLC-grade water. Discuss this protocol with the Mass Spectrometry laboratory that will carry out the analysis. 3. Mass spectrometry NOTE: Analyze peptides on a high resolution mass spectrometer using methods compatible with shotgun proteomics. Data dependent acquisition is the most commonly used LC-tandem MS set-up for this type of analysis and should be optimized for efficient peptide identification, minimizing redundant sequencing and candidate selection over background noise. Mass spectra in the first stage mass analysis scan (MS1) should be recorded in profile mode. It is recommended to preferentially select doubly and triply charged ions for second stage mass analysis (MS2). A mass range of m/z 400-1,600 is suitable to identify most peptides between 8-30 amino acids range. Here, a protocol for analysis using an Orbitrap mass spectrometer is provided.

Data Analysis
1. Protein identification and quantification using MaxQuant 1. Use the freely available MaxQuant software to identify and quantify proteins in each gel fraction from the mass spectrometry raw data. NOTE: MaxQuant works on data generated by data-dependent acquisition shotgun proteomics approaches. The MS data from the blue native gel fractions should be analyzed in batch as separate experiments and not as fractions of an experiment combining all gel slices. Check the iBAQ value box to perform stoichiometry calculations. Detailed protocols for database search and protein quantification using MaxQuant are described in 21 (Figure 1). A matrix will appear containing all protein identifications in rows, with the gel fractions as columns. 3. Remove entries corresponding to reverse hits, proteins only identified by site and potential contaminants by selecting Filter rows based on categorical column in the Filter rows dropdown menu (Figure 2). 4. Normalize the fraction intensities of each protein across the profile against the total protein intensity by selecting Divide in the Normalize dropdown menu, then Sum (Figure 3). A new matrix appears. 5. Display the migration profile plots of all proteins by selecting Profile Plot in the Visualization dropdown menu (Figure 4). 6. Select Filter rows based on valid values in the Filter rows dropdown menu. Enter 1 in the number of valid values required. 7. Perform hierarchical clustering of identified proteins across all blue native gel fractions by selecting Hierarchical clustering in the Clustering/PCA dropdown menu. Uncheck the Columns tree box (Figure 5). The clusters are visualized in a heat-map in the Clustering tab next to the matrix (Figure 6). 8. Locate the cluster containing the bait protein in the dendrogram. Other components of the cluster represent proteins co-migrating with the bait and are therefore potential interacting proteins/subunits of the same complex.

Representative Results
The workflow of the Affinity purification Blue native protein Correlation profiling by Mass Spectrometry (ABC-MS) strategy is depicted in Figure  7. Native protein complexes around a protein of interest are isolated by affinity purification using antibodies against an epitope tag (in this case FLAG) and competitive elution. The complexes are resolved by BN-PAGE, and the whole gel lane is excised into 48 sections, and prepared for shotgun LC-MS/MS. Quantitative MS information is used to generate a migration profile for each identified protein across the blue native separation. Proteins that interact to form a distinct complex display similar migration profiles with superimposable peaks. When a protein of interest takes part in more than one assembly, multiple peaks are observed in its migration profile, given the sub-complexes are within the resolving power of the blue native gel. A systematic comparison of the migration profiles can be achieved by protein correlation profiling using hierarchical clustering. The dendrogram and the peak intensities for all fractions are visualized in a heat-map, facilitating the identification of interacting proteins that belong to distinct protein complexes (Figure 8).
We used this strategy to analyze the interacting partners of Mta2, a core subunit of the NuRD chromatin remodeling complex 20 . As shown in Figure 9, the migration profile of Mta2 displayed two peaks of distinct intensities between 700 kDa and 1.2 MDa, with the lower mass peak displaying higher abundance. Other NuRD core subunits, including Mta1/3, Hdac1/2 and Mbd3, showed identical separation pattern, albeit the peaks for some of the subunits, namely Chd4, Gatad2a/b and Rbbp4/7, had the inverse abundance distribution (Figure 9A, B). In contrast, the profile of Cdk2ap1, a regulatory factor that recruits the NuRD complex to Wnt gene promoters 23 , only displayed the higher mass peak ( Figure  9C), and so did Sall4, a transcriptional repressor that has also been shown to bind NuRD 20,24,25 . Thus, fractionation of affinity purified Mta2associated proteins by blue native PAGE was able to resolve two different forms of the NuRD complex.
ABC-MS also allows the identification of novel interactors whilst assigning them to particular protein entities. Through examination of the proteins clusters and fraction intensities represented in a heat-map we detected a strong correlation between NuRD subunits and Wdr5, a regulatory subunit of the MLL methyltransferase complex 26 , with Wdr5 displaying two migration peaks coincident with the two NuRD peaks (Figure 8), suggesting a novel interaction between Wdr5 and NuRD. We confirmed this interaction by co-immunoprecipitation and co-migration in size exclusion chromatography 20 .
The choice of distance metric calculation in the hierarchical clustering will influence the shape of the clusters and hence the correlations. We recommend experimenting with the different metrics to achieve the best fit with existing interaction knowledge of the protein or complex of interest. Generally we achieved best results with the Manhattan (L1) distance metric 20 . Pearson correlation or Euclidean distance metrics have also been reported for identifying complexes based on alternative fractionation techniques 10,11,12 .
The quantitative MS data obtained from the blue native fractions can also be used to determine the stoichiometry of protein complexes. The iBAQ (intensity based absolute quantification) value provides a measure of relative abundance of the identified proteins 27,28 . The iBAQ values for the protein complex members in all the fractions within a migration profile peak are normalized to that of the bait protein to obtain relative amounts. The normalized iBAQs across a profile peak for a given protein complex member should follow a horizontal trend, and the trend values reflect the stoichiometry of the interacting proteins relative to the bait protein. For a more detailed representation of stoichiometry calculations, see 20 .

Discussion
Here, we describe the use of affinity purification followed by blue native gel electrophoresis in combination with quantitative mass spectrometry to resolve protein complexes. This approach offers a method to unravel one-dimensional protein interaction lists into functional protein assemblies.
We demonstrate the method based on the use of epitope-tagged proteins. However, if a cell line expressing a tagged protein is not available, an alternative might be to use antibodies against the protein of interest, provided there is a peptide available to achieve native competitive elution. The amount of starting material and quantity of beads might need to be modified depending on the expression level of the target protein. We typically perform this protocol with 2-5 x 10 8 cells starting material, and this is sufficient even for proteins with low expression level. The lysis buffer composition should be chosen empirically to achieve near to complete solubilization of the bait. This might be challenging for some types of proteins, in particular chromatin binding or membrane proteins. Alternative options include increasing the amount of salt, provided that the protein complex under study is stable in high salt, or using sonication and/or nuclease treatment for chromatin binding proteins 20,29 . In the case of membrane proteins, switching detergent to DDM or digitonin may be advisable 30 . In the case of DNA binding proteins, it is useful to include  20 . Complete removal of nucleic acids ensures that the interactions detected occur between proteins and are not mediated by DNA.
A critical element to consider when using this approach is the stability of the complex under investigation. The procedure is long and may involve preserving the protein complex overnight. We have achieved good success with two chromatin remodeling complexes (D. Bode and M. Pardo, data not shown), but this should be evaluated. The affinity purification step may be shortened if required.
Alternative techniques that allow native fractionation, such as size exclusion chromatography, have been widely used for over 50 years to characterize protein complexes. We and others have shown that the resolution of blue native PAGE is superior to that achieved using size exclusion chromatography 20,31,32,33 . Another advantage of blue native PAGE is that it does not require chromatography systems, which are expensive, but rather uses protein electrophoretic equipment that is widespread in laboratories. In terms of hands-on time, this method does not involve more work than the traditional geLC-MS/MS approach or offline chromatographic fractionation. However, as most fractionation techniques do, it has a limit to their resolution. Complexes that are very homogeneous or close in mass and shape may be beyond the resolution offered by blue native PAGE, and hence the protocol as reported here might not be universally successful in resolving distinct complexes with shared subunits. Encouragingly, we have been successful in separating two very similar tetrameric complexes sharing three subunits (M. Pardo, manuscript in preparation).
Since mass spectrometry has become increasingly sensitive, even minute amounts of non-specific interactors and contaminants can be detected in AP-MS samples. The approach presented here might aid in the discrimination of real interactors from background contamination in the affinity purification by focusing the attention on proteins with migration peaks that coincide with bait migration peaks at higher molecular weight than that of the monomeric proteins.
Several groups have used fractionation techniques followed by protein correlation profiling to delineate protein complexes at cellular scale without the need for previous isolation 10,11,12,34 . However, this can result in failure to detect sub-stoichiometric interactions. The incorporation of an enrichment step through affinity purification can help to overcome this. The approach described here should be generally useful for exploring the topology of protein complexes and unraveling the multiple complexes a given protein takes part in within the same cellular context. The strategy is simple and amenable to laboratories that may not have expensive chromatographic fractionation equipment to resolve protein complexes.

Disclosures
The authors have nothing to disclose.