1 Introduction
The majority of cellular functions are carried out by macromolecular assemblies of proteins known as protein complexes. Understanding protein–protein interactions (PPIs) and the complexes
they induce systematically in the context of the cellular environment would define the biochemical state of cells and vastly increase
our knowledge on the genotype to phenotype relationship. A
potentially new view of biology could emerge from this knowledge,
where disease phenotypes are not viewed as the consequence of
aberrant gene expression but as the effects of molecular perturbations on the interconnected PPI network. How proteins interact to
generate organized modules and extended PPI networks and how
modules dynamically adapt to (gene expression) perturbation are
key questions to explain the pleiotropic effects of genes.
While systematic analysis of proteins in biological samples has
been greatly facilitated by mass spectrometry (MS) and is by now
routine, the similarly systematic characterization of PPIs and the
complexes they form has remained challenging. The prevalent
methods, yeast two hybrid analysis and affinity co-purification of
interacting proteins are labor intensive, sequential, and not applicable to primary tissue. Consequently, these methods are poorly
suited to detect changes in PPI’s and complexes in cells or tissues
in different biochemical states.
Structural, compositional, or stoichiometric differences in protein complexes in different functional states have traditionally been
determined by the methods of structural biology, Cryo-EM single
particle analysis, NMR, and X-ray crystallography. In recent years,
several structural proteomic techniques such as native proteomics
[1], thermal profiling [2], cross-linking mass spectrometry [3],
hydrogen–deuterium exchange (HDX) [4], and limited proteolysis
(LiP-MS) [5] were developed to support and complement structural data of protein complexes [6]. Moreover, other proteomic
techniques were developed to analyze protein–protein interactions
with increased throughput and sensitivity, such as affinity purification mass spectrometry (AP-MS) [7], proximity labelling (BioID)
[8], and biochemical fractionation mass spectrometry (BF-MS) [9–
12]. For AP-MS, a bait protein is used to purify interaction partners
along with the bait protein, and the resulting PPI network generated by reciprocal purifications could be used to infer network
modules [13–15], for example, using graph mining algorithms
[16] or statistical approaches [17]. In BioID methods, a promiscuous biotin ligase BirA is fused to a target protein that, when
expressed in cells, biotinylates proteins in its close proximity to
the engineered protein and thereby identifies the molecular environment of the protein of interest. In general, structural proteomic
methods are characterized by lack of scalability and low throughput. Indeed, they require protein engineering and protein
270
Andrea Fossati et al.
The majority of cellular functions are carried out by macromolecular assemblies of proteins known as protein complexes. Understanding protein–protein interactions (PPIs) and the complexes
they induce systematically in the context of the cellular environment would define the biochemical state of cells and vastly increase
our knowledge on the genotype to phenotype relationship. A
potentially new view of biology could emerge from this knowledge,
where disease phenotypes are not viewed as the consequence of
aberrant gene expression but as the effects of molecular perturbations on the interconnected PPI network. How proteins interact to
generate organized modules and extended PPI networks and how
modules dynamically adapt to (gene expression) perturbation are
key questions to explain the pleiotropic effects of genes.
While systematic analysis of proteins in biological samples has
been greatly facilitated by mass spectrometry (MS) and is by now
routine, the similarly systematic characterization of PPIs and the
complexes they form has remained challenging. The prevalent
methods, yeast two hybrid analysis and affinity co-purification of
interacting proteins are labor intensive, sequential, and not applicable to primary tissue. Consequently, these methods are poorly
suited to detect changes in PPI’s and complexes in cells or tissues
in different biochemical states.
Structural, compositional, or stoichiometric differences in protein complexes in different functional states have traditionally been
determined by the methods of structural biology, Cryo-EM single
particle analysis, NMR, and X-ray crystallography. In recent years,
several structural proteomic techniques such as native proteomics
[1], thermal profiling [2], cross-linking mass spectrometry [3],
hydrogen–deuterium exchange (HDX) [4], and limited proteolysis
(LiP-MS) [5] were developed to support and complement structural data of protein complexes [6]. Moreover, other proteomic
techniques were developed to analyze protein–protein interactions
with increased throughput and sensitivity, such as affinity purification mass spectrometry (AP-MS) [7], proximity labelling (BioID)
[8], and biochemical fractionation mass spectrometry (BF-MS) [9–
12]. For AP-MS, a bait protein is used to purify interaction partners
along with the bait protein, and the resulting PPI network generated by reciprocal purifications could be used to infer network
modules [13–15], for example, using graph mining algorithms
[16] or statistical approaches [17]. In BioID methods, a promiscuous biotin ligase BirA is fused to a target protein that, when
expressed in cells, biotinylates proteins in its close proximity to
the engineered protein and thereby identifies the molecular environment of the protein of interest. In general, structural proteomic
methods are characterized by lack of scalability and low throughput. Indeed, they require protein engineering and protein
270
Andrea Fossati et al.