Protein interactions are involved in essentially all aspects of life at the molecular level. Mass spectrometry based methods have been used for many years for the unbiased identification of proteins interacting with bait molecules. This workflow is known as affinity purification combined with mass spectrometry (AP-MS) and complements genetic methods such as yeast two hybrid screens (Rual JF et al, Nature, 2005).
History of interaction proteomics
One key challenge of the AP-MS approach is the fact that affinity purification samples do not only contain specific interaction partners, but also unspecific background proteins, which bind to the affinity matrix or the reaction vessel. This meant that researchers had either to apply stringent washing conditions, which leads to the loss of weak interactors, or accept high false positive rates (proteins falsely described as interactors). The tandem affinity purification (TAP) method circumvented some of these limitations by applying two elution steps, yielding rather clean purifications, but at the cost of losing weak interaction partners and requiring high amounts of input material (Rigaut G et al, Nature Biotechnology, 1999 ).
The TAP-tagging approach is still widely employed today and was used to generate the first large scale interactome datasets of model organisms such as the budding yeast (Gavin AC et al, Nature, 2002, Ho Y et al, Nature, 2002, Gavin AC et al, Nature, 2006, Krogan NJ et al, Nature 2006).
However, despite all advances in purification methods, researchers widely use them in combination with non-quantitative mass spectrometry and have to control the false positive rate by blacklisting ‘sticky’ proteins that were identified in control purifications. With the advent of fast, high-resolution mass spectrometers, it became even more apparent that even ‘clean’ purifications do in fact contain a large number of proteins, so that blacklisting proteins will effectively lead to false negatives.
SILAC-based quantitative interaction proteomics
Several years ago, we demonstrated that quantitative proteomics sidesteps essentially all of the drawbacks of non-quantitative methods (Blagoev B et al, Nature Biotechnology, 2003).
The concept is simple: two affinity purifications are performed, one with the bait molecule of choice, and one negative control. Different isotope labels, e.g. light and heavy SILAC, are chosen for the two purifications, and labels are swapped for replicate experiments. The eluates are mixed and analyzed together on the mass spectrometer. In this setup, unspecific binders appear with 1:1 ratios, whereas specific binders show high ratios in the forward experiments and low ratios in the reverse experiments (where labeling conditions have been swapped)
This principle can be applied in many different formats and we have used it to identify sequence-specific DNA and RNA binders (Mittler G et al,Genome research, 2009, Butter F et al, Proc Natl Acad Sci USA, 2009), interactors of modified peptides, such as phosphorylated tyrosines (Schulze WX and Mann M, J Biol Chem, 2004, Hanke S and Mann M, Mol Cell Proteomics, 2009) and modified histone tails (Vermeulen M et al, Cell, 2007, Vermeulen M and Eberl HC, Cell 2010), and of full length proteins (Selbach M and Mann M, Nature methods 2006, Hubner et al, J Cell Biology 2010).
|Nucleic acid-protein interactions
Full length protein-protein interactions
Label-free quantitative interaction proteomics
We have recently developed novel label-free quantification algorithms, which are now implemented in MaxQuant. These algorithms performs pairwise comparisons of peptide intensities across all runs to increase the quantitative information. This algorithm now allows us to identify interaction partners from unlabeled AP-MS samples with comparable reliability to SILAC-based approaches. In order to improve data quality, we found it important to improve the reproducibility of the experimental workflow by carrying out affinity purifications in 96-well format or fully automated on a robotic platform, and to analyze at least 3 replicate experiments.
Label-free interaction proteomics can be easily scaled up to screen the interactomes of a large number of baits in parallel, and is applicable to systems that are not readily amenable to metabolic labeling, such as tissues. We have successfully used this approach for interaction mapping of full-length proteins and for peptide pull-downs from mouse tissues.