Quantification of Protein Interaction Network Dynamics using Multiplexed Co-Immunoprecipitation


Your institution must subscribe to JoVE's Biochemistry section to access this content.

Fill out the form below to receive a free trial or learn more about access:



Quantitative Multiplex Immunoprecipitation (QMI) uses flow cytometry for sensitive detection of differences in the abundance of targeted protein-protein interactions between two samples. QMI can be performed using a small amount of biomaterial, does not require genetically engineered tags, and can be adapted for any previously defined protein interaction network.

Cite this Article

Copy Citation | Download Citations | Reprints and Permissions

Brown, E. A., Neier, S. C., Neuhauser, C., Schrum, A. G., Smith, S. E. Quantification of Protein Interaction Network Dynamics using Multiplexed Co-Immunoprecipitation. J. Vis. Exp. (150), e60029, doi:10.3791/60029 (2019).


Dynamic protein-protein interactions control cellular behavior, from motility to DNA replication to signal transduction. However, monitoring dynamic interactions among multiple proteins in a protein interaction network is technically difficult. Here, we present a protocol for Quantitative Multiplex Immunoprecipitation (QMI), which allows quantitative assessment of fold changes in protein interactions based on relative fluorescence measurements of Proteins in Shared Complexes detected by Exposed Surface epitopes (PiSCES). In QMI, protein complexes from cell lysates are immunoprecipitated onto microspheres, and then probed with a labeled antibody for a different protein in order to quantify the abundance of PiSCES. Immunoprecipitation antibodies are conjugated to different MagBead spectral regions, which allows a flow cytometer to differentiate multiple parallel immunoprecipitations and simultaneously quantify the amount of probe antibody associated with each. QMI does not require genetic tagging and can be performed using minimal biomaterial compared to other immunoprecipitation methods. QMI can be adapted for any defined group of interacting proteins, and has thus far been used to characterize signaling networks in T cells and neuronal glutamate synapses. Results have led to new hypothesis generation with potential diagnostic and therapeutic applications. This protocol includes instructions to perform QMI, from the initial antibody panel selection through to running assays and analyzing data. The initial assembly of a QMI assay involves screening antibodies to generate a panel, and empirically determining an appropriate lysis buffer. The subsequent reagent preparation includes covalently coupling immunoprecipitation antibodies to MagBeads, and biotinylating probe antibodies so they can be labeled by a streptavidin-conjugated fluorophore. To run the assay, lysate is mixed with MagBeads overnight, and then beads are divided and incubated with different probe antibodies, and then a fluorophore label, and read by flow cytometry. Two statistical tests are performed to identify PiSCES that differ significantly between experimental conditions, and results are visualized using heatmaps or node-edge diagrams.


Dynamic protein-protein interactions constitute the molecular signaling cascades and motile structures that are the functional basis of most cellular physiology1. These processes are often depicted as linear signaling pathways that switch between steady states based on single inputs, but experimental and modeling data clearly show that they function as integrated networks2,3,4. In the case of G proteins, different receptors often have the ability to activate the same G protein, and a single receptor can also activate more than one type of G protein5,6. In order for the relatively small number of G protein classes to specifically modulate a vast array of cellular functions such as synaptic transmission, hormone regulation, and cell migration, cells must both integrate and differentiate these signals4,5. Evidence has shown that this signal specificity, for G proteins as well as others, is primarily derived on the basis of finely tuned protein-protein interactions and their temporal dynamics1,3,4,5,6,7. Because signaling networks are comprised of dynamic protein complexes with multiple inputs, outputs, and feedback loops, a single perturbation has the opportunity to alter the overall homeostatic balance of a cell's physiology4,7. It is now widely agreed that signaling should be examined from a network perspective in order to better understand how the integration of multiple inputs controls discrete cellular functions in health and disease7,8,9,10,11,12,13. In light of this, Quantitative Multiplex Immunoprecipitation (QMI) was developed to gather medium-throughput, quantitative data about fold changes in dynamic protein interaction networks.

QMI is an antibody-based assay in which cell lysate is incubated with a panel of immunoprecipitation antibodies that are covalently coupled to magnetic beads containing distinct ratios of fluorescent dyes. Having specific antibodies coupled to distinct magnetic bead classes allows for simultaneous co-immunoprecipitation of multiple target proteins from the same lysate. Following immunoprecipitation (IP), magnetic beads are incubated with a second, fluorophore-conjugated probe antibody (or biotinylated antibody in conjunction with fluorophore-conjugated streptavidin). Co-associations between the proteins recognized by each IP antibody-probe antibody pair, or PiSCES (proteins in shared complexes detected by exposed surface epitopes), are then detected by flow cytometry and can be quantitatively compared between different sample conditions14. Illustrations in Figure 1 show the steps involved in running a QMI assay, including a diagram of magnetic beads with immunoprecipitated protein complexes labeled by fluorescently conjugated probe antibodies (Figure 1C).

The sensitivity of QMI depends on the protein concentration of the lysate relative to the number of magnetic beads used for immunoprecipitation, and achieving a resolution to detect 10% fold changes requires only a small amount of starting material compared to other co-IP methods14,15. For example, the amount of starting material used in QMI is similar to that required for a sandwich Enzyme-Linked ImmunoSorbent Assay (ELISA), but multiple interactions are detected in a single QMI assay. QMI assays using 20 IPs and 20 probe targets have been performed using 1-5 x 105 primary T cells isolated from a 4 mm skin biopsy, P2 synaptosomal preparations from a 3 mm coronal section of mouse prefrontal cortex, or 3 x 106 cultured mouse primary cortical neurons14,16,17. This sensitivity makes QMI useful for analysis of cells or tissue with limited availability, such as clinical samples.

QMI can be adapted for any previously defined protein interaction network (provided that antibodies are available), and to date has been developed to analyze the T cell antigen receptor (TCR) signalosome and a subset of proteins at glutamatergic synapses in neurons17,18. In studies of T cell receptor signaling, QMI was first used to identify stimulation-induced changes in PiSCES, and then to distinguish autoimmune patients from a control group, detect endogenous autoimmune signaling, and finally to generate a hypothesis involving an unbalanced disease-associated subnetwork of interactions14. More recently, the same QMI panel was used to determine that thymocyte selection is determined by quantitative rather than qualitative differences in TCR-associated protein signaling19. In neurons, QMI was used to describe input-specific rearrangement of a protein interaction network for distinct types of input signals in a manner which supports newly emerging models of synaptic plasticity17. Additionally, this synaptic QMI panel was used to identify differences in seven mouse models of autism, cluster the models into subgroups based on their PiSCES biosignatures, and accurately hypothesize a shared molecular deficit that was previously unrecognized in one of the models16. A similar approach could be used to screen for other subgroups that might respond to different drug treatments, or assign drugs to specific responsive subgroups. QMI has potential applications in diagnostics, patient sub-typing, and drug development, in addition to basic science.

To assemble a QMI antibody panel, initial antibody screening and selection protocols are described in Section 1, below. Once antibody panels are identified, protocols for conjugation of the selected antibodies to magnetic beads for IP, and for biotinylation of the selected probe antibodies, are described in Section 2. The protocol for running the QMI assay on cell or tissue lysates is described in Section 3. Finally, since a single experiment can generate ~5 x 105 individual datapoints, instructions and computer codes to assist in data processing, analysis, and visualization are provided in Section 4. An overview of the workflow described in sections 2-4 is shown in Figure 1.

Subscription Required. Please recommend JoVE to your librarian.


1. Assay design

  1. Candidate antibody preparation
    1. For each protein of interest, choose 3 to 5 antibodies to screen. When possible, use monoclonal antibodies that recognize different epitopes. Also include one non-specific control antibody.
    2. To remove Tris, perform buffer exchange by adding the antibody to a 30 kDa spin filter, spinning down to its minimum volume, adding 500 µL of phosphate-buffered saline (PBS), and repeating 3 times. To remove carrier proteins, perform antibody purification according to the manufacturer's protocol (see Table of Materials for specific purification recommendation).
      NOTE: This is done because carrier proteins and buffers with free amine groups (such as Tris) will react with COOH groups and quench the subsequent bead coupling and biotinylation reactions. Ensure that all antibodies are purified (no carrier proteins) and in a buffer free of primary amines (i.e., no Tris).
    3. Couple each antibody to carboxylate modified latex (CML) beads as described by Davis and Schrum20. To conserve the antibody, scale down bead coupling reactions by up to 1/5 (i.e., 3.6 x 106 beads with 10 µL of 0.2-1 mg/mL antibody).
    4. Estimate bead numbers using a hemocytometer (typically ~108/mL) and store at 4 °C. Beads have been stored for over a year and used successfully in QMI assays, but shelf life or expiration dates have not been formally established. NaN3 in the B/S buffer prevents bacterial growth.
    5. Biotinylate a portion of each antibody (see section 2.2 below). Store at 4 °C.
    6. Confirm effective CML bead coupling and accurate counting by staining 1 x 105 beads with a PE-conjugated antibody reactive to the species in which the antibody was raised and reading on a flow cytometer.
    7. Confirm antibody biotinylation by dot blot using streptavidin-HRP.
    8. Once the lab has generated reagents that are known to be effective, use those reagents as positive controls in confirmation reactions in steps 1.1.5 and 1.1.6.
  2. Antibody screening by IP-FCM (immunoprecipitation detected by flow cytometry)
    1. Decide on an appropriate screening lysate. For this and all other pre-QMI screening steps (anything included in this section 1: Assay Design), do not use biosamples with limited availability. Instead, choose a comparable control material such as wildtype mouse tissue, cell lines, or normal human donor tissue that is not a limiting resource.
      NOTE: Choosing a lysis buffer for this assay is not trivial, and is discussed in section 1.5: Detergent Selection, as well as in the penultimate paragraph of the Discussion section. Standard lysis buffers have a base of 150 mM NaCl, 50 mM Tris (pH 7.4), 10 mM sodium fluoride, 2 mM sodium orthovanadate, and protease and phosphatase inhibitor cocktails. Detergents compatible with QMI include 1% NP-40, 1% Digitonin, 0.1-1% Triton X-100, and 0.5-1% deoxycholate14,16,17,18,19,21.
    2. Calculate the total volume of lysate to be used for each IP, using 10 µL for each IP-probe combination to be screened. If screening X probe antibodies and using one IgG control, each IP will use (X+1) * (10 µL) * (1.1 for pipetting error); X+1 is to account for the required IgG probe control. For example, in a 3x3 antibody screen, each IP should use 44 ul. Remember to include an IgG IP control (see example screening setup in Figure 2).
    3. Calculate the CML bead number to be used. If you are screening X probe antibodies, use [(X+1) * 5 x 104 beads]. Ideally, 5,000 beads per well will result in >2,000 beads per well being read by the flow cytometer. For example, in a 3 x 3 antibody screen, each IP should use 20,000 beads, which is approximately 0.66 µL of prepared CML bead stock from step 1.1.3 (beads should first be quantified using a hemocytometer to ensure accuracy).
    4. Incubate the volume of lysate from step 1.2.2 with the volume of each CML bead to be screened from step 1.2.3, overnight at 4 °C with rotation to prevent beads from settling. Typically, perform incubations in the first column of a 96-well PCR plate, and cap with PCR tube strip caps.
    5. Spin down CML beads at 3,200 x g for 1 min and remove lysate by a single, rapid flicking of the plate over the sink. A tiny white pellet should be visible at the bottom of each well both before and after flicking.
    6. Resuspend CML beads in a volume of FlyP buffer to equal 20 µL for each bead-probe pair; for X probe antibodies, use (X+1) * (20 µL) * (1.1 for pipetting error), similar to step 1.2.2. For a 3 x 3 screen, resuspend in 88 µL of FlyP buffer. FlyP buffer is 100 mM NaCl, 50 mM Tris pH 7.4, 1% BSA, 0.01% NaN3.
    7. Distribute each IP across (X+1) wells of a 96-well PCR plate, where X is the number of probe antibodies being screened, using 20 µL/well. Figure 2A shows an example screening setup.
    8. Wash 2 additional times using 200 µL of FlyP buffer per well. Spin the plate as in step 1.2.5 and flick the plate to remove wash buffer after each wash. Pellets will be extremely small, but should be visible after each wash.
    9. For each biotinylated antibody to be screened, calculate the total volume as (Y+1) * 1.1 * 50 µL, where Y is the number of IP antibodies being screened. Dilute antibody to a working concentration in this volume of FlyP buffer, typically starting with 1:100 dilution of a 0.5 mg/mL stock.
    10. Distribute each diluted antibody down each column of the 96 well plate, and ensure CML beads are resuspended.
    11. Incubate at 4 °C for 1 h, either with rotation or pipetting at 15 min intervals to ensure CML beads remain in suspension.
    12. Wash 3x in 200 µL of FlyP buffer, for each wash centrifuge and remove lysate as in step 1.2.5.
    13. Resuspend all CML beads in 50 µL of 1:200 Streptavidin-PE in FlyP buffer.
    14. Incubate at 4 °C in the dark for 30 min.
    15. Wash 3x in 200 µL of FlyP buffer, for each wash centrifuge and remove lysate as in step 1.2.5.
    16. Resuspend in 200 µL of FlyP buffer, and then run on a flow cytometer.
  3. Choosing Antibodies to Include in Assay
    1. Gate on size using FSC-H vs SSC-H, and eliminate doublets using FSC-H vs FSC-A.
    2. Generate histograms of PE fluorescence intensity and overlay both IgG controls (IgG bead-test probe, test bead-IgG probe) onto tested pairs (Figure 2).
    3. Look for a bead-probe pair that gives clear signal over noise (Figure 2B). Additionally, it is not ideal to use the same antibody for both bead and probe. Differential epitope recognition maximizes chances of observing interactions because some epitopes may be occluded in certain protein complexes. If there are no acceptable options, repeat screen with additional antibodies.
  4. Confirmation of Antibody Specificity
    1. To ensure antibody specificity for the intended targets, use a lysate sample in which the target has been knocked out; for example, a knockout mouse or an RNAi cell line. Alternatively, use lysate from a target-negative cell line in which the target protein has been artificially expressed.
    2. Perform IP-FCM as described in step 1.2, modifying to fit the experiment.
  5. Detergent Selection
    1. As detergents are critical in co-IP experiments, empirically test different variations to ensure that the assay has maximum likelihood of detecting changes. To start, choose a relatively small panel of interactions (4-8) that are known to change in a given condition and/or are of particular interest to your study.
    2. Using the non-fluorescent, antibody-conjugated CML beads made for initial screens, perform IP-FCM as described in 1.2 using varied lysis buffer detergent conditions. Detergent screens can be performed with detergent as the only variable, or with different cell conditions for each detergent. Always use IgG controls for both beads and probes, since detergents occasionally produce unexpected background in some IP-Probe combinations.
    3. Based on the MFIs from the screen, choose a detergent that optimizes the signal for the PiSCES of interest. It is likely that some compromises will need to be made22.

2. Multiplex reagent preparation

  1. Magnetic bead coupling
    1. Using the magnetic bead region map, select bead regions to use in a pattern that minimizes risk of cross-detection. Magnetic bead typically smear up and to the right, so avoid bead regions that are diagonally adjacent. Beads from every other column of the bead diagram shown on the Luminex website are recommended (https://www.luminexcorp.com/magplex-microspheres/).
    2. Prepare the carrier-free antibody at 0.1 mg/mL in PBS (as in 1.1.2) in 250 µL. Keep on ice for later use.
    3. Vortex magnetic beads extensively, and then aliquot 250 µL into an amber microcentrifuge tube (to protect beads from photobleaching).
    4. Magnetically separate magnetic beads for 60 s and remove the supernatant.
    5. Add 250 µL of MES buffer (50 mM MES pH 6.0, 1 mM EDTA), vortex, magnetically separate for 60 s, and remove the supernatant. Repeat and resuspend magnetic beads in 200 µL of MES buffer.
    6. Add 40 µL of MES to a 2 mg of single-use tube of Sulfo-NHS to make a 50 mg/mL stock.
    7. Add 25 µL of freshly made Sulfo-NHS to the magnetic beads . Vortex.
    8. Add 25 µL of 50 mg/mL freshly dissolved EDAC [1-ethyl-3-(-3-dimethylaminopropyl) carbodiimide hydrochloride, also called EDC] in MES buffer. Vortex.
    9. Cover and shake on a vortexer with a tube-holding attachment for 20 min at room temp, 1000 rpm.
    10. Magnetically separate for 60 s and remove the supernatant.
    11. Resuspend in 500 µL of PBS, vortex, magnetically separate for 60 s and remove the supernatant. Repeat.
    12. Resuspend in 250 µL of antibody solution from step 2.1.2. Vortex.
    13. Incubate 2 h at room temp with shaking on a vortexer at 1000 rpm.
    14. Add 500 µL of PBS to the magnetic beads, vortex, magnetically separate for 60 s, and remove the supernatant.
    15. Add 750 µL of Blocking/Storage (B/S) buffer (1% BSA in PBS pH 7.4, 0.01% NaN3). Cover and incubate 30 min at room temp, 1000 rpm.
    16. Magnetically separate for 60 s and remove the supernatant. Resuspend in 100 µL of B/S buffer.
    17. Store at 4 °C. Beads have been stored for over a year and used successfully in QMI assays, but shelf life or expiration dates have not been formally established. NaN3 in the B/S buffer should prevent bacterial growth.
    18. Validate magnetic bead coupling by staining ~0.25 µL of coupled magnetic beads with a fluorescent anti-host species secondary and reading on a flow cytometer, as in step 1.1.5.
  2. Biotinylation
    1. Ensure that antibodies are in PBS with no carrier protein.
    2. Calculate the total µg of antibody to be biotinylated (100-200 µg recommended for use in multiplex, 25-50 µg recommended for screening).
    3. Prepare fresh 10 mM sulfo-NHS-biotin (can be done by adding 224 µL of ddH2O to a 1 mg no-weigh tube).
    4. Add 1 µL of 10 mM sulfo-NHS-biotin per 25 µg of antibody, vortex or pipette up and down to mix.
    5. Incubate at room temp for 1 h.
    6. Incubate at 4 °C for 1 h.
    7. Use a 30 kDa spin filter to remove unbound biotin and stop the reaction. Add 500 µL of PBS and spin the column until the minimum volume is reached. Add 500 µL of additional PBS and repeat for 3 total buffer exchanges.
    8. Estimate concentration by measuring the absorbance of 1-2 µL on a spectrophotometer, and then bring the antibody concentration to 0.5 mg/mL.
    9. Store at 4 ˚C.

3. Quantitative multiplex immunoprecipitation

  1. Plate layout
    NOTE: This assay works best when performed using 96-well plates and 2-4 sample conditions.
    1. Always run appropriate controls (i.e., stimulated v. unstimulated cells) on the same plate in order to detect changes between conditions. Distribute each sample horizontally across the plate, and use each column for a different probe antibody. A set of technical replicates for each probe should be run immediately after the first set. See Figure 3 for an example.
    2. Carefully document the plate layout to facilitate accurate plate loading and analysis.
  2. Sample preparation & immunoprecipitation (Day 1)
    1. Lyse tissue or cells in appropriate detergent with protease and phosphatase inhibitors and incubate on ice for 15 min. Take care to keep the lysate cold at all times.
      NOTE: The exact quantity of stating biomaterial and lysate protein concentration must be empirically determined, and some examples of previously used samples are listed in the third paragraph of the introduction. In general, in the range of 200 µL of 2 mg/mL protein per sample has been successful in the past for 20 IP and 20 probe targets, but ideal inputs for each antibody panel and cell or tissue type must be determined empirically.
    2. Spin down at 4 °C for 15 min at 16,000 x g to remove membranes and debris; keep supernatant as lysate.
    3. Perform a BCA Assay or similar to determine protein concentrations, and then normalize protein concentration between samples. If using cells, begin with an equal number of cells per condition and normalization is optional.
    4. Prepare a master magnetic bead mix that contains ~250 magnetic beads of each class per well in the assay. Adjust the bead numbers after data analysis so that in future assays an average of 110 beads of each class will be read per well.
      NOTE: Example calculation: (New bead volume) = [(Run Average) / 110] * (previous bead volume). Bead volumes should be adjusted in this way about every 8 runs or as needed. Typically, 3-4 µL of each magnetic bead(prepared as above) are used for a 2-plate experiment.
    5. Wash the magnetic bead mix 2x in FlyP buffer with magnetic separation, and then resuspend in FlyP buffer. For resuspension, use 10 µL per sample per plate. FlyP buffer is 100 mM NaCl, 50 mM Tris pH 7.4, 1% BSA, 0.01% NaN3.
    6. After thoroughly vortexing the magnetic bead mix, aliquot 10 µL into ice cold microcentrifuge tubes (one tube per sample). Add equal volumes of lysate (with normalized concentrations) to each tube for immunoprecipitation.
    7. Aliquot the lysate-magnetic bead mixture into one tube for each plate being run; e.g. for a 2-plate experiment, split the lysate into two tubes. Place tubes on a rotator at 4 °C overnight for immunoprecipitation, covered to keep out light.
  3. Running the assay (Day 2)
    1. Start with the lysate-magnetic bead tubes for Plate #1. Use a magnetic bead rack to remove lysate from the magnetic beads, and reserve lysate for future analysis. Wash beads 2x in 500 µL of ice cold FlyP buffer. Keep tubes tubes always on ice or at 4 °C.
    2. Calculate resuspension volume as (number of probes) * (2 technical replicates) * (25 µL per well) * (1.1 for pipetting error). Resuspend IPs in calculated volume of ice-cold FlyP buffer.
    3. After thoroughly resuspending magnetic beads by gentle pipetting, distribute 25 µL per well across a flat-bottomed 96 well plate, on ice.
    4. In a different 96-well plate, dilute biotinylated probe antibodies to 2x working concentration (working concentration is typically 1:100 or 1:200, empirically determined) in FlyP buffer so that their order matches the columns on the plate layout (see Figure 3). The final volume of probe antibodies at the working concentration will be 50 µL per well, so the volume of each 2x antibody prepared should be (25 µL) * (number of biological samples) * (2 technical replicates) * (1.1 for pipetting error).
    5. Use a multichannel pipette to distribute 25 µL of each probe antibody dilution into the magnetic bead-containing assay plate.
    6. Shake on a horizontal plate shaker to mix and resuspend the magnetic beads, and then incubate at 4 °C for 1 h, shaking at 450 rpm in the dark.
    7. Wash 3x with FlyP buffer on a magnetic plate washer at 4 °C.
    8. Resuspend the magnetic beads in 50 µL of 1:200 Streptavidin-PE.
    9. Shake to mix and resuspend beads, and then incubate at 4 °C for 30 min, shaking at 450 rpm in the dark.
    10. Wash 3x with FlyP buffer on a magnetic plate washer at 4 °C.
    11. Resuspend in 125 µL of FlyP buffer.
    12. Shake for 1 min at 900 rpm to thoroughly resuspend beads.
    13. Run on refrigerated flow cytometer (see diagram in Figure S1). Use the "high RP1 target" setting in the flow cytometer software, and a stop condition of 1,000 beads per region (greatly overshooting the number that should be in any individual well to prevent the machine from stopping prematurely) and sample volume of 80 µL.
    14. Pause the run half way through and resuspend the beads to prevent settling.
    15. Export data files in the .xml format.
    16. Repeat the process for the remaining plates, starting at step 3.3.1.

4. Data analysis

NOTE: The ANC code was designed to compare two conditions from N = 4 experiments, each with 2 technical replicates for each condition. For example, cells are stimulated four independent times, QMI is run on four different days on control (unstimulated) and stimulated cells, with technical replicates as above, and data analysis proceeds as described below.

  1. Adaptive non-parametric with adjustable alpha cutoff (ANC)
    1. Open MATLAB and set the active directory to a folder containing the ANC program components and the .xml files exported from the flow cytometer.
    2. Fill in the "ANC input" file to reflect the details of the experimental design. The example file included in Supplementary File has been pre-filled to run the example data, also provided.
    3. Run the program, which will write a .csv file into the active directory. The file reports PiSCES that are significantly different, at a false positive (alpha) level of 0.05, between Control and Experimental conditions, in all 4 experimental replicates, or at least 3/4 replicates.
    4. Note 'ANC hits,' which are defined as PiSCES with significant differences in at least 3 experimental replicates, represented as 3/4∩4/4 in the file, for use in step 4.3.1.
  2. Weighted correlation network analysis23 (CNA)
    1. Paste-transpose the column titles of the data file output by Matlab ending in "_MFI.CSV" into the first row of a new excel sheet. Add the columns "experiment" for experiment number, and "treatment", for experimental treatment, or any other variables to be analyzed. Save this file as "traits.csv".
    2. Open R studio and set the working directory to a folder containing the "_MFI.CSV" and "TRAITS.CSV" files.
    3. Run the R commands as indicated in the commented command file and the detailed in the instructions included with the files. The WCNA modules significantly correlated with each experimental trait are output as a graphic file, and the correlation of each interaction IPi_ProbeJ with each module is output as a .csv file.
    4. Note 'CNA hits,' which are defined as interactions with module membership (MM) > 0.7 and p < 0.05 for membership in a module that was identified as significantly correlated with the experimental variable of interest, for use in step 4.3.1.
  3. Positive 'hits' & visualization
    1. For each interaction in the "3/4∩4/4 hits" list in the ANC output file (from step 4.1.4), identify if that interaction is also a "CNA hit" by checking the CNA output file (see step 4.2.6). Create a new column that indicates if each ANC hit is also a CNA hit.
    2. Calculate the average log2 fold change value for each ANC∩CNA hit by averaging the values given in the ANC output spreadsheet "_Hits.csv" from step 4.1.3. Convert values to log2 fold change before averaging. For interactions that were significant in only 3/4 replicates, delete the outlier value.
    3. Make a spreadsheet with each ANC∩CNA hit listed as an IP in one column, a probe in the second column, and the fold change value in the third column. Use this spreadsheet to create a node-edge diagram in Cytoscape by importing the file as a network.

Subscription Required. Please recommend JoVE to your librarian.

Representative Results

Antibody Screening
Figure 2B shows the results of a screen for the protein Connexin36. Most IP_probe combinations produce no signal over IgG controls. IP with the monoclonal antibody 1E5 and probe with either 1E5 or the polyclonal antibody 6200 produces a rightward shift in the bead distribution compared to IgG controls. Here, IP 1E5 and probe 6200poly were selected to avoid using the same antibody as IP and probe, both to reduce the probability of a non-specific protein being recognized by two independent antibodies, and to increase the chance of detecting co-associations using different epitopes. It is best to choose an IP_probe combination with at least 1-2 log higher MFI compared to IgG controls, but occasionally pairs producing weaker MFIs that are consistently distinguishable from controls may be used if no alternatives are identified. Figure 2C shows a specificity validation experiment for the 1E5-6200poly combination. Lysate from 293 cells transfected with a Connexin36 plasmid produced a ~1.5-log rightward shift in the bead distribution, while untransfected cells overlapped with the IgG controls. When confirming the specificity of a pair, negative control lysate from a knockout animal or cell line without the target protein should have an MFI similar to the IgG controls.

Bead Coupling
A typical magnetic bead coupling quality control reaction will give an MFI 3-4 logs above background when stained with a secondary antibody conjugated to a fluorophore with a brightness index between 3 and 5 (such as PE or FITC). Figure 4 shows a typical quality control reaction comparing the conjugation of a new magnetic bead compared to the older batch being replaced.

Data Analysis
In each experiment, ANC compares the fluorescence distributions of each magnetic bead class in each well (i.e. all possible IP_Probe combinations) between a user-defined control and experimental condition. It assigns a p-value to each combination that reflects the probability that the beads have been sampled from identical populations based on Kolomogrov-Shmirnov (K-S) statistics. The program then calculates the K-S p-value required to produce a false-positive rate of 0.05 by correcting for multiple comparisons and accounting for technical variability (differences between the technical replicates). IP_probe combinations (PiSCES) whose K-S test p-value falls below the calculated cut-off in all four experiments, or at least 3/4 experiments (3/4∩4/4) are identified. Since the p-value cut-offs differ depending on these different levels of stringency, occasionally PiSCES will be identified in 4/4 but not 3/4∩4/4, so separate lists are calculated. For detailed ANC equations, see (Smith et al. 2016).14 Details about WCNA analysis and results are discussed thoroughly by Langfelder et al.23

Data Presentation
ANC and CNA23 analyses are performed to identify PiSCES that both (1) show significant fold changes between experimental conditions in at least 3/4 of runs and (2) belong to a CNA module that is correlated with the experimental variable. These high-confidence PiSCES that are identified by two independent statistical approaches are referred to as ANC∩CNA PiSCES. These interactions can be visualized as a node-edge diagram using the open-source software Cytoscape (Figure 5a) or as a heatmap by using the R code and analysis instructions included in the supplementary material (Figure 5b).

Figure 1
Figure 1. Overview of Quantitative Multiplex Immunoprecipitation. (a) Previously screened antibodies are covalently coupled to different classes of magnetic beads in separate reactions. (b) Overnight, protein complexes are immunoprecipitated using a mixture of the antibody-coupled magnetic beads. (c) Co-immunoprecipitated proteins are labeled by a probe antibody and a fluorophore. (d) Magnetic beads and labeled protein complexes are run through a refrigerated flow cytometer to quantify relative amounts of proteins occurring in shared complexes. See Figure S1 for schematic details of custom refrigeration. (e) The flow cytometer Manager Software separates MagBeads by class and (f) displays fluorescence histograms from each bead region. (g) Data are exported as .xml files and analyzed by two independent statistical approaches. Only PiSCES identified by both analyses are reported using heatmap and node-edge diagram visualizations. Please click here to view a larger version of this figure.

Figure 2
Figure 2. Connexin 36 antibody screening using IP-FCM. (a) IP-FCM was performed on mouse brain lysate using a 4x4 panel of Connexin 36 (Cx36) CML beads and probes. Lysate was immunoprecipitated with each CML bead in a separate row of the plate. After washes, each bead was distributed across its row so that one probe antibody can be added per column. (b) Most antibody combinations show no signal (orange) over IgG background (gray, blue). The 1E5 IP with the 6200Poly probe shows acceptable positive signal. The 1E5 bead/probe and 6200Poly bead/probe pairs each show acceptable signal, but it is not ideal to use the same antibody for both bead and probe. Differential epitope recognition maximizes chances of observing interactions because some epitopes may be occluded in certain protein complexes. The 6200Poly bead with the 1E5 probe gives the strongest signal and was chosen to use in the multiplex assay pending specificity confirmation. (c) IP-FCM using the pair of Cx36 antibodies selected from screening was performed on the lysate of 293T cells transfected with Cx36 and non-transfected controls. There is clear signal from the Cx36-transfected cells, but the non-transfected cells are indistinguishable from IgG bead/probe controls. Please click here to view a larger version of this figure.

Figure 3
Figure 3. Example plate layout. A 4-condition multiplex is set up in a 96-well plate. Samples 1-4 are loaded in consecutive rows (each biological sample represented here by a different color), and technical replicates are loaded in the same order in the following 4 rows. One probe is used per column. Please click here to view a larger version of this figure.

Figure 4
Figure 4. A typical quality control reaction comparing the conjugation of a new MagBead compared to the older batch being replaced. The bead gives an MFI 2-4 logs above background, and the new batch has an MFI similar to that of the old batch. Please click here to view a larger version of this figure.

Figure 5
Figure 5. QMI identifies synaptic PiSCES that change in magnitude following 5 minutes of NMDA stimulation in cultured cortical neurons. A QMI experiment compared NMDA stimulated vs. unstimulated (ACSF control) neurons. PiSCES that were identified by both ANC and CNA analyses are presented. (a) In a node-edge diagram produced using the open source software Cytoscape, the nodes indicate the antibody targets (proteins) that were included as IPs and probes in the QMI panel. The edges represent ANC∩CNA PiSCES, with the color and thickness of the edge indicating the direction and magnitude of the fold-change between NMDA treatment and control. PiSCES that did not change between the NMDA and control conditions are not included in the figure. (b) A heatmap produced in R using the Heatmap.2 function represents the same ANC∩CNA PiSCES. ComBAT-normalized, log2 MFI values are normalized by row to account for data that spans ~3 logs, and the relative MFI for each experimental replicate is show to demonstrate the relative magnitude and consistency of each reported PiSCES. The data and code required to reproduce these figures are included in the Supplementary File. Please click here to view a larger version of this figure.

Supplementary Figure 1
Figure S1: Diagrams of custom refrigeration of the flow cytometry system. The flow cytometer’s array reader must be kept at room temperature, but the lower portion (the microplate platform) must be refrigerated to maintain PiSCES during analysis. See the Table of Materials for model information about flow cytometer and sandwich prep refrigerator used. (a) The upper attachments and food storage bins were removed from a sandwich prep refrigerator. The microplate platform was placed on the metal supports meant to hold the plastic food storage bins. The plastic housing of the microplate platform was removed to make it fit. A custom plexiglass platform was built with measurements shown in (b) to cover the upper opening of the refrigerator. The plexiglass was insulated with ½" foam insulation cut to match the size of the plexiglass, and sealed the gap with insulating tape. A hole was then drilled through the plexiglass to allow the sample needle from the flow cytometer assay reader to access the microplate platform when extended. A black coupling device that was originally screwed into the top of the microplate platform was removed, and screwed back into the microplate platform from above the plexiglass, which aided in alignment. A door in the Plexiglass cover allows user access to the microplate carrier when it is extended out of the unit. Note that the flow cytometer software will alert the user that the plate carrier is too cold, but the user can override the warning and run cooled QMI experiments. (c) Photograph of the assembled system. (d) Detail of the front right corner, as drawn in (a), showing assembly of plexiglass cover and insulation. (e) Detail of the sample needle aligned above the holes. (f) Detail of the shaved-down section of insulation that allows airflow under the unit. (g) Detail of the open door showing the flow cytometer microplate platform below. Please click here to view a larger version of this figure.

Supplementary File
Supplementary File. Data and code required to reproduce these figures. Please click here to download this file.

Subscription Required. Please recommend JoVE to your librarian.


The QMI assay requires substantial investment in antibody panel development, equipment and reagents, but once the assay is established, one can collect high-dimensional data observing protein interaction networks as they respond to experimentally-controlled stimuli. Technically, QMI requires careful pipetting and tracking of sample and antibody well locations. Carefully labeling the assay plates is useful, as is making a detailed template of well locations on paper, which is then saved for data analysis. The importance of keeping the beads and lysate cold at all times, including in the flow cytometer microplate carrier (see Figure S1 for custom refrigeration instructions) cannot be overstated. Protein interactions will rapidly dissociate at room temperature, and early attempts at using an unmodified, room temperature flow cytometer ended with the identification of many temperature-labile interactions, but not those that changed with the intended stimulation.

QMI is an antibody-based assay, so the initial selection of antibodies is critical. Monoclonal or recombinant antibodies should be used whenever possible to reduce variability in results. Polyclonals show lot-to-lot variation, but peptide-based polyclonals to a short epitope seem to be relatively stable over time. Drift can be minimized by buying large batches of antibodies; this also allows for custom-orders of carrier-free antibodies which precludes the need to purify antibodies using Melon Gel and spin columns, and the associated antibody loss.

It is also important to note that, because detecting a signal is reliant upon available epitopes, the lack of a signal does not necessarily indicate the lack of an interaction, a limitation that is common with other protein interaction methodologies.14 Further, when a signal is detected, it is impossible to unambiguously state weather the protein interaction is direct (A interacts with B) or indirect (A interacts with X and Y, which then interact with B), which is why the observed interactions are referred to as PiSCES rather than protein-protein interactions (PPI), which may imply direct binding. A limitation of all antibody-based methods that should be kept in mind is that the addition of antibodies may disrupt or stabilize protein complexes. Another limitation of using flow cytometry rather than western blots is that size information to confirm antibody specificity is not available. To overcome this limitation, IgG controls are used in screening each antibody pair, and specificity is confirmed with knock-out or knock-in cell lines or animals before proceeding with QMI experiments (section 1.4).

IgG controls are not used in the QMI assay because each IgG produces a different level of background signal, making it impossible to know the correct background value to subtract. For example, if IP (X)_probe IgG gives an MFI of 100 and IP IgG_probe Y gives an MFI of 200, which background value should be subtracted from IP X_probe Y? Similarly, sometimes undetected interactions (e.g., IP X probe Z) will have a lower MFI than the nonspecific IgG interactions. To account for this limitation of not knowing the absolute MFI signal, PiSCES are not reported solely for being detected above an arbitrary background level. Instead, only PiSCES that change in response to a given stimulation are reported. While high MFI can be caused by nonspecific noise, this noise would not be expected to change with stimulation. In addition, a portion (10-20%) of condition-dependent interactions observed are generally confirmed by a second method, typically IP-western. This confirmation is analogous to confirming high-throughput RNA sequencing results with RT-PCR and is meant to increase confidence QMI results.

Expression effects influencing QMI results cannot be ruled out without additional tests, because QMI does not distinguish between increased absolute levels of a protein and increased homo-multimerization of a protein. To minimize uncertainty regarding expression, experiments can be performed using acute treatment conditions with short timescales that minimize potential changes in protein expression levels. Other methods are needed to rule out expression effects in chronic treatment conditions or primary patient samples.

It is vital to select an appropriate lysis buffer for QMI. Too weak of a detergent can leave membranes intact and hold together proteins that are not in complex, while too strong a detergent can destroy protein complexes. Additional factors such as the presence of calcium or its chelators can dramatically affect PiSCES and should be carefully considered before screening antibodies to include in a QMI panel. For IP-western experiments, lysis conditions are usually optimized for each PiSCES on a case-by-case basis, but the best conditions for detecting a single PiSCES may not translate to other PiSCES in the same protein network22. Detergent selection presents a chicken-and-egg dilemma, in that a lysis buffer is needed to screen antibody candidates, but a panel of antibodies is needed to screen for an appropriate lysis buffer. While not a perfect solution, one can select a small panel of beads and probes that are of particular interest and/or have known associations or dissociations in response to a stimulus, and testing their behavior under different lysis conditions on the CML beads initially used for screening antibodies (step 1.1.3). An ideal detergent should allow for both reliable detection of PiSCES and recapitulation of known physiological protein behavior (association/dissociation) with a given stimulus. If there is any concern that a detergent does not fully solubilize membranes, a negative control antibody can be added that would only give signal if two proteins were linked by membrane24. When appropriate lysis buffers are selected, changes in even weak interactions - such as those between a kinase and substrate - can be reliably detected (e.g. TCR-LCK)14.

Previous work using QMI in neurons and T cells has both carefully confirmed previous findings in order to increase confidence in the validity of QMI results, and generated new hypotheses that led to discoveries about signal transduction and disease pathways. In the future, QMI can be adapted to other protein interaction networks and expanded up to 500 proteins with the current microsphere classes available. Using QMI to study how networks of multi-protein complexes change in response to stimuli as they control cellular processes has the potential to yield important insights into both health and disease.

Subscription Required. Please recommend JoVE to your librarian.


The authors have no conflicts of interest to disclose.


The authors wish to acknowledge Tessa Davis for important contributions to QMI assay development, and current and former members of the Smith and Schrum labs for technical guidance and intellectual input. This work was funded by NIMH grants R01 MH113545 and R00 MH 102244.


Name Company Catalog Number Comments
96-well flat bottomed plates Bio Rad 171025001
96-well PCR plates VWR 82006-704
Bioplex 200 System with HTF Bio Rad 171000205 modiefied to keep partially refrigerated, see Figure S1 for details
Bio-Plex Pro Wash Station Bio Rad 30034376
BSA Sigma
CML beads Invitrogen C37481
EDTA Sigma E6758
EZ-Link Sulfo-NHS-Biotin Thermo Scientific A39256
MagPlex Microspheres Luminex MC12xxx-01 xxx is the 3 digit bead region
Melon Gel IgG Spin Purification Kit Thermo Scientific 45206 used for antibody purification
MES Sigma M3671
Microplate film, non-sterile USA Scientific 2920-0000
Phosphotase inhibitor cocktail #2 Sigma P5726
Protease inhibitor cocktail Sigma P8340
Sandwich Prep Refrigerator Norlake SMP 36 15 for custom refrigeration of Bioplex 200
Sodium fluoride Sigma 201154
Sodium orthovanadate Sigma 450243
Streptavidin-PE BioLegend 405204
Sulfo NHS Thermo Scientific A39269
Tris Fisher Scientific BP152



  1. Pawson, T., Nash, P. Protein-protein interactions define specificity in signal transduction. Genes and Development. 14, (9), 1027-1047 (2000).
  2. Rodríguez-Jorge, O., et al. Cooperation between T cell receptor and Toll-like receptor 5 signaling for CD4. Science Signaling. 12, (577), (2019).
  3. Bugaj, L. J., et al. Cancer mutations and targeted drugs can disrupt dynamic signal encoding by the Ras-Erk pathway. Science. 361, (6405), (2018).
  4. Pawson, T. Dynamic control of signaling by modular adaptor proteins. Current Opinion in Cell Biology. 19, (2), 112-116 (2007).
  5. Hamm, H. E. The many faces of G protein signaling. Journal of Biological Chemistry. 273, (2), 669-672 (1998).
  6. Lohse, M. J., Hofmann, K. P. Spatial and Temporal Aspects of Signaling by G-Protein-Coupled Receptors. Molecular Pharmacology. 88, (3), 572-578 (2015).
  7. Eltschinger, S., Loewith, R. TOR Complexes and the Maintenance of Cellular Homeostasis. Trends in Cell Biology. 26, (2), 148-159 (2016).
  8. Papin, J. A., Hunter, T., Palsson, B. O., Subramaniam, S. Reconstruction of cellular signalling networks and analysis of their properties. Nature Reviews Molecular Cell Biology. 6, (2), 99-111 (2005).
  9. Komarova, N. L., Zou, X., Nie, Q., Bardwell, L. A theoretical framework for specificity in cell signaling. Molecular Systems Biology. 1, 23 (2005).
  10. Schrum, A. G., Gil, D. Robustness and Specificity in Signal Transduction via Physiologic Protein Interaction Networks. Clinical and Experimental Pharmacology. 2, (3), (2012).
  11. De Las Rivas, J., Fontanillo, C. Protein-protein interactions essentials: key concepts to building and analyzing interactome networks. PLoS Computational Biology. 6, (6), 1000807 (2010).
  12. Ron, D., Walter, P. Signal integration in the endoplasmic reticulum unfolded protein response. Nature Reviews Molecular Cell Biology. 8, (7), 519-529 (2007).
  13. Bogdan, A. R., Miyazawa, M., Hashimoto, K., Tsuji, Y. Regulators of Iron Homeostasis: New Players in Metabolism, Cell Death, and Disease. Trends in Biochemical Sciences. 41, (3), 274-286 (2016).
  14. Smith, S. E., et al. Multiplex matrix network analysis of protein complexes in the human TCR signalosome. Science Signaling. 9, (439), 7 (2016).
  15. Schrum, A. G., et al. High-sensitivity detection and quantitative analysis of native protein-protein interactions and multiprotein complexes by flow cytometry. Science's STKE. 2007, (389), (2007).
  16. Brown, E. A., et al. Clustering the autisms using glutamate synapse protein interaction networks from cortical and hippocampal tissue of seven mouse models. Molecular Autism. 9, (1), 48 (2018).
  17. Lautz, J. D., Brown, E. A., Williams VanSchoiack, A. A., Smith, S. E. P. Synaptic activity induces input-specific rearrangements in a targeted synaptic protein interaction network. Journal of Neurochemistry. 146, (5), 540-559 (2018).
  18. Smith, S. E., et al. Signalling protein complexes isolated from primary human skin-resident T cells can be analysed by Multiplex IP-FCM. Experimental Dermatology. 23, (4), 272-273 (2014).
  19. Neier, S. C., et al. The early proximal αβ TCR signalosome specifies thymic selection outcome through a quantitative protein interaction network. Science Immunology. 4, (32), (2019).
  20. Davis, T. R., Schrum, A. G. IP-FCM: immunoprecipitation detected by flow cytometry. Journal of Visualized Expiriments. (46), (2010).
  21. Smith, S. E., et al. IP-FCM measures physiologic protein-protein interactions modulated by signal transduction and small-molecule drug inhibition. PLoS One. 7, (9), 45722 (2012).
  22. Lautz, J., et al. Activity-dependent changes in synaptic protein complex composition are consistent in different detergents despite differential solubility. BioRXiv. (2019).
  23. Langfelder, P., Horvath, S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics. 9, 559 (2008).
  24. Schrum, A. G., Gil, D., Turka, L. A., Palmer, E. Physical and functional bivalency observed among TCR/CD3 complexes isolated from primary T cells. Journal of Immunology. 187, (2), 870-878 (2011).



    Post a Question / Comment / Request

    You must be signed in to post a comment. Please or create an account.

    Usage Statistics