Here, we improved one-colour IP-FCM to develop a powerful technology platform for the high-throughput generation of multidimensional and quantitative data that serves as a new input for quantitative analysis. As a pertinent example, we showed its application to mathematical modelling of early signalling events in T-cell activation.
The main advantages of multi-colour IP-FCM are: (i) precise protein data with a high dynamic range, (ii) normalized data on protein phosphorylations and interactions, (iii) multiple parameters quantified simultaneously using two bead sizes and multiple colours, (iv) absolute values, in contrast to relative ones, can be obtained, (v) time effective, (vi) adaptable to a 96-well format for the generation of large data sets, (vii) except for a flow cytometer no special equipment is needed, and (viii) the assay can be run at native or denaturing conditions, depending on the availability of antibodies.
For IP-FCM, 5000 beads were measured and the geometric MFI was taken, in contrast to WB where only one quantification per point was done. This might contribute to the smaller error of IP-FCM, in addition of uneven WB transfer 
. Furthermore, the high-throughput nature of IP-FCM easily allows measurement of more replicas than triplicates. This is very limited in IP-WB, and efforts are undertaken to develop SDS-PAGE and WB systems in which large numbers of samples can be applied 
We used PE-labelled calibration beads to obtain absolute values. Since defined FITC-labelled beads are also available, one could retrieve absolute values from two-colour IP-FCM.
High specificity of the stainings is reached by the “sandwich assay” nature of IP-FCM. Very few antibodies are truly monospecific, the majority also bind to at least one other cellular antigen. The sandwich assays achieve superb selectivity without the size fractionation afforded by WB, because the specificities of two different antibodies are exploited. Hence IP-FCM is fundamentally more specific compared to assays where only one antibody is employed, such as intracellular FCM staining. IP-FCM might also be used for quantification of other stimulus-induced events, as e.g. ubiquitinylation, methylation or proteolysis; and for identification of stimulus-specific changes in subcellular localization by cell fractionation prior to IP.
However, IP-FCM neither yields information on the protein size as does IP-WB, nor on individual cells as does intracellular staining for FCM 
. It is also not suited to identify novel phosphorylation sites or interactions. When using native conditions for the IP and general anti-phospho-tyrosine antibodies for the staining step (as in ), one should consider that the antibody might probe all accessible phospho-tyrosines of the purified protein complex. For example, the phospho-tyrosine signal in is a mixture of phospho-CD3 and associated phospho-proteins, such as phospho-ZAP70. If this is not desired, we recommend a denaturation step before the IP. Another potential drawback of IP-FCM is the fact that epitopes might be spatially blocked by bound proteins, conformational changes or covalent modifications. Again, denaturation could be of advantage, as we did when measuring phospho-Erk levels.
IP-FCM is best suited to generate large quantitative, multidimensional data sets on protein phosphorylations and interactions that are already known and for which good antibodies exist.
Sensitivity of IP-FCM might be enhanced by increasing the concentration of the lysate (lysis of cells in smaller volume), by reducing the number of beads used per sample, by increasing the concentration of the staining antibodies or by using a primary and secondary staining reagent, such as a biotinylated first antibody and fluorophore-coupled streptavidin.
Using multi-colour IP-FCM, we reconstructed with high quantitative accuracy the dynamics of phosphorylations at the TCR-CD3 and ZAP70, which have previously been partially characterized by IP-WB and one-colour IP-FCM 
. These data have allowed us to develop a mechanistic model of the underlying TCR-CD3-ZAP70 interaction and reversible phosphorylations and (unlike the IP-WB data with their large error bars) have forced the model to precisely reproduce the kinetics of ZAP70 recruitment and phosphorylation. As a result, the model has correctly predicted the temporal relation of two key ZAP70 phosphorylations, comparatively early phosphorylation of Y319 by Lck and delayed trans-autophosphorylation of Y493. Thus, an earlier suggestion that phosphorylation at Y319 required for Y493 phosphorylation 
, is enforced by our study.
Unexpectedly, our data also showed that pervanadate stimulation of the cells led to a transient decrease of the pY319-ZAP70/ZAP70 ratio at the TCR-CD3. Due to the large errors of IP-WB, this conclusion could not be drawn. Importantly, the mathematical model demonstrated that the initial massive recruitment of non-phosphorylated ZAP70 was responsible for this seemingly counter-intuitive kinetic behavior. Indeed, we could experimentally verify that a small amount of pY319-ZAP70 is pre-bound to the TCR-CD3 in resting cells and that the bulk of ZAP70 in the cytosol is in the non-phosphorylated state. Thus, recruitment of the cytosolic ZAP70 pool to the TCR-CD3 upon stimulation leads to a transient decrease of the pY319-ZAP70/ZAP70 ratio at the TCR-CD3. At later time points the ratio increases, due to phosphorylation of ZAP70 bound to the TCR-CD3.
The quantitative agreements between data and mathematical simulations corroborates the underlying mechanistic model, underscoring the need for de novo phosphorylation of ZAP70 recruited to the TCR-CD3, followed by trans-autophosphorylation of ZAP70 molecules. In conclusion, the high accuracy and sensitivity of IP-FCM is suited to elucidate the temporal coding of cell signalling events to unrivalled accuracy.