Abstract
Most proteins function by forming complexes within a dynamic
interconnected network that underlies various biological mechanisms. To
systematically investigate such interactomes, high-throughput techniques
including CF-MS have been developed to capture, identify, and quantify
protein-protein interactions (PPIs) in large-scale. Compared to other
techniques, CF-MS allows the global identification and quantification of
native protein complexes in one setting, without genetic manipulation
and overexpression. Furthermore, quantitative CF-MS can potentially
elucidate the distribution of a protein in multiple co-elution features,
informing the stoichiometries and dynamics of a target protein complex.
In this issue, Youssef et al. (Proteomics 2023, XX, XXXX-XXXX) combined
multiplex CF-MS and an in-house algorithm to study the dynamics of the
PPI network for Escherichia coli grown under ten different
conditions. While the results demonstrated that while most proteins
remained stable, the authors were able to detect disrupted interactions
that were growth condition-specific. Further bioinformatics analyses
also revealed biophysical properties and structural patterns that govern
such a response.
(156 Words )
Youssef et al. conducted a study
using multiplex co-fractionation mass spectrometry (mCF-MS) and a custom
software pipeline to investigate protein-protein interactions (PPI) inEscherichia coli under ten different growth conditions [1].
The results revealed dynamic rewiring of protein networks, and
subsequent bioinformatics analyses shed light on the biophysical
properties and structural patterns that govern the response to
environmental stimuli.
Bacteria such as E. coli possess mechanisms to adapt to changing
physiological demands through genetic and biochemical processes [2].
These mechanisms involve regulating protein abundance and the dynamic
formation of multi-subunit assemblies. Currently, comprehensive studies
of these complex ”interactomes” mainly rely on high-throughput
strategies that combine biochemical purification/fractionation with
MS-based techniques [3]. A related concept for capturing
protein-protein interactions (PPIs) arises from the spatiotemporal
co-behaviour of proteins, including co-expression, co-localization,
co-aggregation, or co-fractionation. Co-fractionation occurs when
proteins from the same assembly comigrate from an analytical column
under native conditions, suggesting potential co-localisation [4].
In this study, E. coli K-12 was subjected to ten different growth
conditions, including various culture media and stressors. The authors
analysed the E. coli interactome under each condition using
mCF-MS, which involved fractionation of native protein complexes by
high-performance liquid chromatography (IEX-HPLC). Subsequently, a
bottom-up proteomics workflow with isobaric TMT labelling facilitated
the identification and quantification of condition-specific protein
complex elution patterns.
A critical aspect of this work is a two-module data analysis pipeline
developed by Youssef et al. The first module performs pre-processing
steps, including data normalisation, smoothing, and filtering, to
construct protein profiles and correlate them with reference PPI
databases, including STRING and BioGRID [5,7]. This pre-processing
workflow addresses the disparity between peptide-based TMT measurements
and the quantification of proteins inferred from tryptic peptides. This
workflow also dynamically captures PPIs dynamically, distinguishing it
from existing software tools [6,8,9]. Meanwhile, the second module
computes a conditional similarity score for each predicted PPI pair from
the mCF-MS data compared to a reference interactome. To estimate the
degree of interactome remodelling, the authors introduced a similarity
score derived from two distinct data metrics: co-elution (qualitative
measure of HPLC retention time shifts) and co-abundance (quantitative
measure of intensity fold changes).
Using this pipeline, the authors generated a reference interactome
comprising 6,152 high-confidence pairwise interactions from all ten
mCF/MS datasets. Most proteins (68.3%) formed a single complex, while
the rest participated in multiple complexes. The overall extent of
interactome remodelling remained relatively stable in all conditions,
with less than 5% showing high remodelling scores indicative of
disrupted interactions. Notably, under galactose as the primary carbon
source, a protein complex mainly composed of galactose metabolism
enzymes was extensively formed. The high temperature (42 °C) resulted in
the decomposition of a remodelled top-ranking complex into two
subcomplexes—one consisting of heat shock response proteins from the
’hsl’ family and the other consisting of hydrogenase proteins from the
’hya’ gene family. The authors also identified pathways that were
influential in driving remodelling within each growth condition. The
protein modification machinery consistently showed a significant impact
across all conditions, highlighting the role of dynamic
post-translational modifications.
The authors conducted an analysis to explore the associations between
different protein traits and integrated remodelling scores within each
growth condition. First, they found a negative correlation between
remodelling scores and protein intensities, indicating that proteins
with lower abundance were more prone to remodelling. Furthermore, at
near-zero growth and stationary phase, the stability of ancient protein
interactions (based on evolution age) was found to be favoured.
Furthermore, membrane proteins exhibited relatively higher stability
compared to their cytosolic and periplasmic counterparts, and proteins
undergoing phosphorylation tended to be more stable. On the other hand,
proteins with intrinsically disordered structures had lower median
remodelling scores, indicating their tendency to form constitutive
assemblies. Additionally, highly connected hub proteins involved in
multiple interactions demonstrated greater stability, while the number
of complexes in which a protein participated had a lesser impact. To
facilitate the exploration of results, the authors also developed an
interactive web application that visualises the dynamic profiles ofE. coli mCF-MS.
At the end of the manuscript, the authors discussed the limitations of
their workflows. One drawback pertained to the freeze-thawing and
sonication of the E. coli cells prior to co-fractionation. They
also acknowledged their sheer focus on large-scale interactome
remodelling patterns, while the ten different E. coli culture
conditions would have provided ample material for additional biological
experiments. Notwithstanding, the experimental and computational
pipelines presented here have contributed to proof-of-concept and a
ready-to-use workflow for future investigations into interactome
dynamics in different biological systems.