Myelination of the central nervous system (CNS) is critical to vertebrate nervous systems for efficient neural signaling. CNS myelination occurs as oligodendrocytes terminally differentiate, a process regulated in part by the myelin regulatory factor, MYRF. Using bioinformatics and extensive biochemical and functional assays, we find that MYRF is generated as an integral membrane protein that must be processed to release its transcription factor domain from the membrane. In contrast to most membrane-bound transcription factors, MYRF proteolysis seems constitutive and independent of cell- and tissue-type, as we demonstrate by reconstitution in E. coli and yeast. The apparent absence of physiological cues raises the question as to how and why MYRF is processed. By using computational methods capable of recognizing extremely divergent sequence homology, we identified a MYRF protein domain distantly related to bacteriophage tailspike proteins. Although occurring in otherwise unrelated proteins, the phage domains are known to chaperone the tailspike proteins' trimerization and auto-cleavage, raising the hypothesis that the MYRF domain might contribute to a novel activation method for a membrane-bound transcription factor. We find that the MYRF domain indeed serves as an intramolecular chaperone that facilitates MYRF trimerization and proteolysis. Functional assays confirm that the chaperone domain-mediated auto-proteolysis is essential both for MYRF's transcriptional activity and its ability to promote oligodendrocyte maturation. This work thus reveals a previously unknown key step in CNS myelination. These data also reconcile conflicting observations of this protein family, different members of which have been identified as transmembrane or nuclear proteins. Finally, our data illustrate a remarkable evolutionary repurposing between bacteriophages and eukaryotes, with a chaperone domain capable of catalyzing trimerization-dependent auto-proteolysis in two entirely distinct protein and cellular contexts, in one case participating in bacteriophage tailspike maturation and in the other activating a key transcription factor for CNS myelination.
Membrane-bound transcription factors are synthesized as integral membrane proteins, but are proteolytically cleaved in response to relevant cues, untethering their transcription factor domains from the membrane to control gene expression in the nucleus. Here, we find that the myelin regulatory factor MYRF, a major transcriptional regulator of oligodendrocyte differentiation and central nervous system myelination, is also a membrane-bound transcription factor. In marked contrast to most well-known membrane-bound transcription factors, cleavage of MYRF appears to be unconditional. Surprisingly, this processing is performed by a protein domain shared with bacteriophages in otherwise unrelated proteins, where the domain is critical to the folding and proteolytic maturation of virus tailspikes. In addition to revealing a previously unknown key step in central nervous system myelination, this work also illustrates a remarkable example of evolutionary repurposing between bacteriophages and eukaryotes, with the same protein domain capable of catalyzing trimerization-dependent auto-proteolysis in two completely distinct protein and cellular contexts.
Gram-negative bacteria produce outer membrane vesicles (OMVs) that package and deliver proteins, small molecules, and DNA to prokaryotic and eukaryotic cells. The molecular details of OMV biogenesis have not been fully elucidated, but peptidoglycan-associated outer membrane proteins that tether the outer membrane to the underlying peptidoglycan have been shown to be critical for OMV formation in multiple Enterobacteriaceae. In this study, we demonstrate that the peptidoglycan-associated outer membrane proteins OprF and OprI, but not OprL, impact production of OMVs by the opportunistic pathogen Pseudomonas aeruginosa. Interestingly, OprF does not appear to be important for tethering the outer membrane to peptidoglycan but instead impacts OMV formation through modulation of the levels of the Pseudomonas quinolone signal (PQS), a quorum signal previously shown by our laboratory to be critical for OMV formation. Thus, the mechanism by which OprF impacts OMV formation is distinct from that for other peptidoglycan-associated outer membrane proteins, including OprI.
Mass spectrometry (MS)-based shotgun proteomics allows protein identifications even in complex biological samples. Protein abundances can then be estimated from the counts of MS/MS spectra attributable to each protein, provided that one corrects for differential MS-detectability of the contributing peptides. We describe the use of a method, APEX, which calculates Absolute Protein EXpression levels based on learned correction factors, MS/MS spectral counts, and each protein's probability of correct identification.
The APEX-based calculations consist of three parts: (1) Using training data, peptide sequences and their sequence properties, a model is built that can be used to estimate MS-detectability (Oi) for any given protein. (2) Absolute abundances of proteins measured in an MS/MS experiment are calculated with information from spectral counts, identification probabilities and the learned Oi -values. (3) Simple statistics allow for significance analysis of differential expression in two distinct biological samples, i.e., measuring relative protein abundances. APEX-based protein abundances span more than four orders of magnitude and are applicable to mixtures of hundreds to thousands of proteins from any type of organism.
Quantitative proteomics; Protein expression; Label-free mass spectrometry; Spectral counting
Recent advances in next-generation DNA sequencing and proteomics provide an unprecedented ability to survey mRNA and protein abundances. Such proteome-wide surveys are illuminating the extent to which different aspects of gene expression help to regulate cellular protein abundances. Current data demonstrate a substantial role for regulatory processes occurring after mRNA is made — that is, post-transcriptional, translational and protein degradation regulation — in controlling steady-state protein abundances. Intriguing observations are also emerging in relation to cells following perturbation, single-cell studies and the apparent evolutionary conservation of protein and mRNA abundances. Here, we summarize current understanding of the major factors regulating protein expression.
We have described a protocol for performing high-throughput immunofluorescence microscopy on microarrays of yeast cells. This approach employs immunostaining of spheroplasted yeast cells printed as high-density cell microarrays, followed by imaging using automated microscopy. A yeast spheroplast microarray can contain more than 5,000 printed spots, each containing cells from a given yeast strain, and is thus suitable for genome-wide screens focusing on single cell phenotypes, such as systematic localization or co-localization studies or genetic assays for genes affecting probed targets. We demonstrate the use of yeast spheroplast microarrays to probe microtubule and spindle defects across a collection of yeast strains harboring tetracycline-down-regulatable alleles of essential genes.
Yeast; immunofluorescence; high-throughput microscopy; cell microarrays; microtubule
It has been hypothesized that components of enzymatic pathways might organize into intracellular assemblies to improve their catalytic efficiency or lead to coordinate regulation. Accordingly, de novo purine biosynthesis enzymes may form a purinosome in the absence of purines, and a punctate intracellular body has been identified as the purinosome. We investigated the mechanism by which human de novo purine biosynthetic enzymes might be organized into purinosomes, especially under differing cellular conditions. Irregardless of the activity of bodies formed by endogenous enzymes, we demonstrate that intracellular bodies formed by transiently transfected, fluorescently tagged human purine biosynthesis proteins are best explained as protein aggregation.
Motivation: A number of computational methods have been proposed that predict protein–protein interactions (PPIs) based on protein sequence features. Since the number of potential non-interacting protein pairs (negative PPIs) is very high both in absolute terms and in comparison to that of interacting protein pairs (positive PPIs), computational prediction methods rely upon subsets of negative PPIs for training and validation. Hence, the need arises for subset sampling for negative PPIs.
Results: We clarify that there are two fundamentally different types of subset sampling for negative PPIs. One is subset sampling for cross-validated testing, where one desires unbiased subsets so that predictive performance estimated with them can be safely assumed to generalize to the population level. The other is subset sampling for training, where one desires the subsets that best train predictive algorithms, even if these subsets are biased. We show that confusion between these two fundamentally different types of subset sampling led one study recently published in Bioinformatics to the erroneous conclusion that predictive algorithms based on protein sequence features are hardly better than random in predicting PPIs. Rather, both protein sequence features and the ‘hubbiness’ of interacting proteins contribute to effective prediction of PPIs. We provide guidance for appropriate use of random versus balanced sampling.
Availability: The datasets used for this study are available at http://www.marcottelab.org/PPINegativeDataSampling.
Contact: firstname.lastname@example.org; email@example.com
Supplementary Information: Supplementary data are available at Bioinformatics online.
Proteins play major roles in most biological processes; as a consequence, protein expression levels are highly regulated. While extensive post-transcriptional, translational and protein degradation control clearly influence protein concentration and functionality, it is often thought that protein abundances are primarily determined by the abundances of the corresponding mRNAs. Hence surprisingly, a recent study showed that abundances of orthologous nematode and fly proteins correlate better than their corresponding mRNA abundances. We tested if this phenomenon is general by collecting and testing matching large-scale protein and mRNA expression datasets from seven different species: two bacteria, yeast, nematode, fly, human, and plant. We find that steady-state abundances of proteins show significantly higher correlation across these diverse phylogenetic taxa than the abundances of their corresponding mRNAs (p=0.0008, paired Wilcoxon). These data support the presence of strong selective pressure to maintain protein abundances during evolution, even when mRNA abundances diverge.
Increasing knowledge about the organization of proteins into complexes, systems, and pathways has led to a flowering of theoretical approaches for exploiting this knowledge in order to better learn the functions of proteins and their roles underlying phenotypic traits and diseases. Much of this body of theory has been developed and tested in model organisms, relying on their relative simplicity and genetic and biochemical tractability to accelerate the research. In this review, we discuss several of the major approaches for computationally integrating proteomics and genomics observations into integrated protein networks, then applying guilt-by-association in these networks in order to identify genes underlying traits. Recent trends in this field include a rising appreciation of the modular network organization of proteins underlying traits or mutational phenotypes, and how to exploit such protein modularity using computational approaches related to the internet search algorithm PageRank. Many protein network-based predictions have recently been experimentally confirmed in yeast, worms, plants, and mice, and several successful approaches in model organisms have been directly translated to analyze human disease, with notable recent applications to glioma and breast cancer prognosis.
Data integration; Function prediction; Humans; Model organisms; Phenotype prediction; Protein interaction networks
Shotgun proteomics using mass spectrometry is a powerful method for protein identification but suffers limited sensitivity in complex samples. Integrating peptide identifications from multiple database search engines is a promising strategy to increase the number of peptide identifications and reduce the volume of unassigned tandem mass spectra. Existing methods pool statistical significance scores such as p-values or posterior probabilities of peptide-spectrum matches (PSMs) from multiple search engines after high scoring peptides have been assigned to spectra, but these methods lack reliable control of identification error rates as data are integrated from different search engines. We developed a statistically coherent method for integrative analysis, termed MSblender. MSblender converts raw search scores from search engines into a probability score for all possible PSMs and properly accounts for the correlation between search scores. The method reliably estimates false discovery rates and identifies more PSMs than any single search engine at the same false discovery rate. Increased identifications increment spectral counts for all detected proteins and allow quantification of proteins that would not have been quantified by individual search engines. We also demonstrate that enhanced quantification contributes to improve sensitivity in differential expression analyses.
integrative analysis; database search; peptide identification
Haploinsufficiency, wherein a single functional copy of a gene is insufficient to maintain normal function, is a major cause of dominant disease. Human disease studies have identified several hundred haploinsufficient (HI) genes. We have compiled a map of 1,079 haplosufficient (HS) genes by systematic identification of genes unambiguously and repeatedly compromised by copy number variation among 8,458 apparently healthy individuals and contrasted the genomic, evolutionary, functional, and network properties between these HS genes and known HI genes. We found that HI genes are typically longer and have more conserved coding sequences and promoters than HS genes. HI genes exhibit higher levels of expression during early development and greater tissue specificity. Moreover, within a probabilistic human functional interaction network HI genes have more interaction partners and greater network proximity to other known HI genes. We built a predictive model on the basis of these differences and annotated 12,443 genes with their predicted probability of being haploinsufficient. We validated these predictions of haploinsufficiency by demonstrating that genes with a high predicted probability of exhibiting haploinsufficiency are enriched among genes implicated in human dominant diseases and among genes causing abnormal phenotypes in heterozygous knockout mice. We have transformed these gene-based haploinsufficiency predictions into haploinsufficiency scores for genic deletions, which we demonstrate to better discriminate between pathogenic and benign deletions than consideration of the deletion size or numbers of genes deleted. These robust predictions of haploinsufficiency support clinical interpretation of novel loss-of-function variants and prioritization of variants and genes for follow-up studies.
Humans, like most complex organisms, have two copies of most genes in their genome, one from the mother and one from the father. This redundancy provides a back-up copy for most genes, should one copy be lost through mutation. For a minority of genes, one functional copy is not enough to sustain normal human function, and mutations causing the loss of function of one of the copies of such genes are a major cause of childhood developmental diseases. Over the past 20 years medical geneticists have identified over 300 such genes, but it is not known how many of the 22,000 genes in our genome may also be sensitive to gene loss. By comparing these ∼300 genes known to be sensitive to gene loss with over 1,000 genes where loss of a single copy does not result in disease, we have identified some key evolutionary and functional similarities between genes sensitive to loss of a single copy. We have used these similarities to predict for most genes in the genome, whether loss of a single copy is likely to result in disease. These predictions will help in the interpretation of mutations seen in patients.
The Gram-negative bacterium Pseudomonas aeruginosa is a common cause of chronic airway infections in individuals with the heritable disease cystic fibrosis (CF). After prolonged colonization of the CF lung, P. aeruginosa becomes highly resistant to host clearance and antibiotic treatment; therefore, understanding how this bacterium evolves during chronic infection is important for identifying beneficial adaptations that could be targeted therapeutically. To identify potential adaptive traits of P. aeruginosa during chronic infection, we carried out global transcriptomic profiling of chronological clonal isolates obtained from 3 individuals with CF. Isolates were collected sequentially over periods ranging from 3 months to 8 years, representing up to 39,000 in vivo generations. We identified 24 genes that were commonly regulated by all 3 P. aeruginosa lineages, including several genes encoding traits previously shown to be important for in vivo growth. Our results reveal that parallel evolution occurs in the CF lung and that at least a proportion of the traits identified are beneficial for P. aeruginosa chronic colonization of the CF lung.
Deadly diseases like AIDS, malaria, and tuberculosis are the result of long-term chronic infections. Pathogens that cause chronic infections adapt to the host environment, avoiding the immune response and resisting antimicrobial agents. Studies of pathogen adaptation are therefore important for understanding how the efficacy of current therapeutics may change upon prolonged infection. One notorious chronic pathogen is Pseudomonas aeruginosa, a bacterium that causes long-term infections in individuals with the heritable disease cystic fibrosis (CF). We used gene expression profiles to identify 24 genes that commonly changed expression over time in 3 P. aeruginosa lineages, indicating that these changes occur in parallel in the lungs of individuals with CF. Several of these genes have previously been shown to encode traits critical for in vivo-relevant processes, suggesting that they are likely beneficial adaptations important for chronic colonization of the CF lung.
Edge detection is a signal processing algorithm common in artificial intelligence and image recognition programs. We have constructed a genetically encoded edge detection algorithm that programs an isogenic community of E.coli to sense an image of light, communicate to identify the light-dark edges, and visually present the result of the computation. The algorithm is implemented using multiple genetic circuits. An engineered light sensor enables cells to distinguish between light and dark regions. In the dark, cells produce a diffusible chemical signal that diffuses into light regions. Genetic logic gates are used so that only cells that sense light and the diffusible signal produce a positive output. A mathematical model constructed from first principles and parameterized with experimental measurements of the component circuits predicts the performance of the complete program. Quantitatively accurate models will facilitate the engineering of more complex biological behaviors and inform bottom-up studies of natural genetic regulatory networks.
Plants are essential sources of food, fiber and renewable energy. Effective methods for manipulating plant traits have important agricultural and economic consequences. We introduce a rational approach for associating genes with plant traits by combined use of a genome-scale functional network and targeted reverse genetic screening. We present a probabilistic network (AraNet) of functional associations among 19,647 (73%) genes of the reference flowering plant Arabidopsis thaliana. AraNet associations have measured precision greater than literature-based protein interactions (21%) for 55% of genes, and are highly predictive for diverse biological pathways. Using AraNet, we found a 10-fold enrichment in identifying early seedling development genes. By interrogating network neighborhoods, we identify At1g80710 (now Drought sensitive 1; Drs1) and At3g05090 (now Lateral root stimulator 1; Lrs1) as novel regulators of drought sensitivity and lateral root development, respectively. AraNet (http://www.functionalnet.org/aranet/) provides a global resource for plant gene function identification and genetic dissection of plant traits.
The planar cell polarity (PCP) signaling pathway is essential for embryonic development because it governs diverse cellular behaviors, and the “core PCP” proteins, such as Dishevelled and Frizzled, have been extensively characterized1–4. By contrast, the “PCP effector” proteins, such as Intu and Fuz, remain largely unstudied5, 6. These proteins are essential for PCP signaling, but they have never been investigated in a mammal and their cell biological activities remain entirely unknown. We report here that Fuz mutant mice display neural tube defects, skeletal dysmorphologies, and Hedgehog signaling defects stemming from disrupted ciliogenesis. Using bioinformatics and imaging of an in vivo mucociliary epithelium, we establish a central role for Fuz in membrane trafficking, showing that Fuz is essential for trafficking of cargo to basal bodies and to the apical tips of cilia. Fuz is also essential for exocytosis in secretory cells. Finally, we identify a novel, Rab-related small GTPase as a Fuz interaction partner that is also essential for ciliogenesis and secretion. These results are significant because they provide novel insights into the mechanisms by which developmental regulatory systems like PCP signaling interface with fundamental cellular systems such as the vesicle trafficking machinery.
Many genes are toxic when overexpressed, but general mechanisms for this toxicity have proven elusive. Vavouri et al. (2009) find that intrinsic protein disorder and promiscuous molecular interactions are strong determinants of dosage sensitivity, explaining in part the toxicity of dosage-sensitive oncogenes in mice and humans.
The step by step assembly process from preribosome in the nucleus to translation-competent 60S ribosome subunit in the cytoplasm is revealed (also see Kemmler et al. in this issue).
The ribosome stalk is essential for recruitment of translation factors. In yeast, P0 and Rpl12 correspond to bacterial L10 and L11 and form the stalk base of mature ribosomes, whereas Mrt4 is a nuclear paralogue of P0. In this study, we show that the dual-specificity phosphatase Yvh1 is required for the release of Mrt4 from the pre-60S subunits. Deletion of YVH1 leads to the persistence of Mrt4 on pre-60S subunits in the cytoplasm. A mutation in Mrt4 at the protein–RNA interface bypasses the requirement for Yvh1. Pre-60S subunits associated with Yvh1 contain Rpl12 but lack both Mrt4 and P0. These results suggest a linear series of events in which Yvh1 binds to the pre-60S subunit to displace Mrt4. Subsequently, P0 loads onto the subunit to assemble the mature stalk, and Yvh1 is released. The initial assembly of the ribosome with Mrt4 may provide functional compartmentalization of ribosome assembly in addition to the spatial separation afforded by the nuclear envelope.
Cellular processes often depend on stable physical associations between proteins. Despite recent progress, knowledge of the composition of human protein complexes remains limited. To close this gap, we applied an integrative global proteomic profiling approach, based on chromatographic separation of cultured human cell extracts into more than one thousand biochemical fractions which were subsequently analyzed by quantitative tandem mass spectrometry, to systematically identify a network of 13,993 high-confidence physical interactions among 3,006 stably-associated soluble human proteins. Most of the 622 putative protein complexes we report are linked to core biological processes, and encompass both candidate disease genes and unnanotated proteins to inform on mechanism. Strikingly, whereas larger multi-protein assemblies tend to be more extensively annotated and evolutionarily conserved, human protein complexes with 5 or fewer subunits are far more likely to be functionally un-annotated or restricted to vertebrates, suggesting more recent functional innovations.
Here we describe human spotted cell chips, a technology for determining cellular state across arrays of cells subjected to chemical or genetic perturbation. Cells are grown and treated under standard tissue culture conditions before being fixed and printed onto replicate glass slides, effectively decoupling the experimental conditions from the assay technique. Each slide is then probed using immunofluorescence or other optical reporter and assayed by automated microscopy. We show potential applications of the cell chip by assaying HeLa and A549 samples for changes in target protein abundance (of the dsRNA-activated protein kinase PKR), subcellular localization (nuclear translocation of NFκB) and activation state (phosphorylation of STAT1 and of the p38 and JNK stress kinases) in response to treatment by several chemical effectors (anisomycin, TNFα, and interferon), and we demonstrate scalability by printing a chip with ∼4,700 discrete samples of HeLa cells. Coupling this technology to high-throughput methods for culturing and treating cell lines could enable researchers to examine the impact of exogenous effectors on the same population of experimentally treated cells across multiple reporter targets potentially representing a variety of molecular systems, thus producing a highly multiplexed dataset with minimized experimental variance and at reduced reagent cost compared to alternative techniques. The ability to prepare and store chips also allows researchers to follow up on observations gleaned from initial screens with maximal repeatability.
Motivation: High-throughput protein identification experiments based on tandem mass spectrometry (MS/MS) often suffer from low sensitivity and low-confidence protein identifications. In a typical shotgun proteomics experiment, it is assumed that all proteins are equally likely to be present. However, there is often other evidence to suggest that a protein is present and confidence in individual protein identification can be updated accordingly.
Results: We develop a method that analyzes MS/MS experiments in the larger context of the biological processes active in a cell. Our method, MSNet, improves protein identification in shotgun proteomics experiments by considering information on functional associations from a gene functional network. MSNet substantially increases the number of proteins identified in the sample at a given error rate. We identify 8–29% more proteins than the original MS experiment when applied to yeast grown in different experimental conditions analyzed on different MS/MS instruments, and 37% more proteins in a human sample. We validate up to 94% of our identifications in yeast by presence in ground-truth reference sets.
Availability and Implementation: Software and datasets are available at http://aug.csres.utexas.edu/msnet
Contact: firstname.lastname@example.org, email@example.com
Supplementary information: Supplementary data are available at Bioinformatics online.
Motivation: Tandem mass spectrometry (MS/MS) offers fast and reliable characterization of complex protein mixtures, but suffers from low sensitivity in protein identification. In a typical shotgun proteomics experiment, it is assumed that all proteins are equally likely to be present. However, there is often other information available, e.g. the probability of a protein's presence is likely to correlate with its mRNA concentration.
Results: We develop a Bayesian score that estimates the posterior probability of a protein's presence in the sample given its identification in an MS/MS experiment and its mRNA concentration measured under similar experimental conditions. Our method, MSpresso, substantially increases the number of proteins identified in an MS/MS experiment at the same error rate, e.g. in yeast, MSpresso increases the number of proteins identified by ∼40%. We apply MSpresso to data from different MS/MS instruments, experimental conditions and organisms (Escherichia coli, human), and predict 19–63% more proteins across the different datasets. MSpresso demonstrates that incorporating prior knowledge of protein presence into shotgun proteomics experiments can substantially improve protein identification scores.
Availability and Implementation: Software is available upon request from the authors. Mass spectrometry datasets and supplementary information are available from http://www.marcottelab.org/MSpresso/.
Contact: firstname.lastname@example.org; email@example.com
Supplementary Information: Supplementary data website: http://www.marcottelab.org/MSpresso/.
Polarizing cells extensively restructure cellular components in a spatially and temporally coupled manner along the major axis of cellular extension. Budding yeast are a useful model of polarized growth, helping to define many molecular components of this conserved process. Besides budding, yeast cells also differentiate upon treatment with pheromone from the opposite mating type, forming a mating projection (the ‘shmoo’) by directional restructuring of the cytoskeleton, localized vesicular transport and overall reorganization of the cytosol. To characterize the proteomic localization changes accompanying polarized growth, we developed and implemented a novel cell microarray-based imaging assay for measuring the spatial redistribution of a large fraction of the yeast proteome, and applied this assay to identify proteins localized along the mating projection following pheromone treatment. We further trained a machine learning algorithm to refine the cell imaging screen, identifying additional shmoo-localized proteins. In all, we identified 74 proteins that specifically localize to the mating projection, including previously uncharacterized proteins (Ycr043c, Ydr348c, Yer071c, Ymr295c, and Yor304c-a) and known polarization complexes such as the exocyst. Functional analysis of these proteins, coupled with quantitative analysis of individual organelle movements during shmoo formation, suggests a model in which the basic machinery for cell polarization is generally conserved between processes forming the bud and the shmoo, with a distinct subset of proteins used only for shmoo formation. The net effect is a defined ordering of major organelles along the polarization axis, with specific proteins implicated at the proximal growth tip.
Upon sensing mating pheromone, budding yeast cells form a mating projection (the ‘shmoo’) that serves as a model for polarized cell growth, involving cytoskeletal/cytosolic restructuring and directed vesicular transport. We developed a cell microarray-based imaging assay for measuring localization of the yeast proteome during polarized growth. We find major organelles ordered along the polarization axis, localize 74 proteins to the growth tip, and observe adaptive reuse of general polarization machinery.
Proteomics; polarized growth; subcellular localization; pheromone response; yeast
Proteins interact in complex protein–protein interaction (PPI) networks whose topological properties—such as scale-free topology, hierarchical modularity, and dissortativity—have suggested models of network evolution. Currently preferred models invoke preferential attachment or gene duplication and divergence to produce networks whose topology matches that observed for real PPIs, thus supporting these as likely models for network evolution. Here, we show that the interaction density and homodimeric frequency are highly protein age–dependent in real PPI networks in a manner which does not agree with these canonical models. In light of these results, we propose an alternative stochastic model, which adds each protein sequentially to a growing network in a manner analogous to protein crystal growth (CG) in solution. The key ideas are (1) interaction probability increases with availability of unoccupied interaction surface, thus following an anti-preferential attachment rule, (2) as a network grows, highly connected sub-networks emerge into protein modules or complexes, and (3) once a new protein is committed to a module, further connections tend to be localized within that module. The CG model produces PPI networks consistent in both topology and age distributions with real PPI networks and is well supported by the spatial arrangement of protein complexes of known 3-D structure, suggesting a plausible physical mechanism for network evolution.
Proteins function together forming stable protein complexes or transient interactions in various cellular processes, such as gene regulation and signaling. Here, we address the basic question of how these networks of interacting proteins evolve. This is an important problem, as the structures of such networks underlie important features of biological systems, such as functional modularity, error-tolerance, and stability. It is not yet known how these network architectures originate or what driving forces underlie the observed network structure. Several models have been proposed over the past decade—in particular, a “rich get richer” model (preferential attachment) and a model based upon gene duplication and divergence—often based only on network topologies. Here, we show that real yeast protein interaction networks show a unique age distribution among interacting proteins, which rules out these canonical models. In light of these results, we developed a simple, alternative model based on well-established physical principles, analogous to the process of growing protein crystals in solution. The model better explains many features of real PPI networks, including the network topologies, their characteristic age distributions, and the spatial distribution of subunits of differing ages within protein complexes, suggesting a plausible physical mechanism of network evolution.