A framework for the automated analysis of subcellular patterns in human protein atlas images

The systematic study of subcellular location patterns is required to fully characterize the human proteome, as subcellular location provides critical context necessary for understanding a protein's function. The analysis of tens of thousands of expressed proteins for the many cell types and cellular conditions under which they may be found creates a need for automated subcellular pattern analysis. We therefore describe the application of automated methods, previously developed and validated by our laboratory on fluorescence micrographs of cultured cell lines, to analyze subcellular patterns in tissue images from the Human Protein Atlas. The Atlas currently contains images of over 3000 protein patterns in various human tissues obtained using immunohistochemistry. We chose a 16 protein subset from the Atlas that reflects the major classes of subcellular location. We then separated DNA and protein staining in the images, extracted various features from each image, and trained a support vector machine classifier to recognize the protein patterns. Our results show that our system can distinguish the patterns with 83% accuracy in 45 different tissues, and when only the most confident classifications are considered, this rises to 97%. These results are encouraging given that the tissues contain many different cell types organized in different manners, and that the Atlas images are of moderate resolution. The approach described is an important starting point for automatically assigning subcellular locations on a proteome-wide basis for collections of tissue images such as the Atlas.     

A Global Multiregional Proteomic Map of the Human Cerebral Cortex

The Brodmann area(BA)-based map is one of the most widely used cortical maps for studies of human brain functions and in clinical practice; however, the molecular architecture of BAs remains unknown. The present study provided a global multiregional proteomic map of the human cerebral cortex by analyzing 29 BAs. These 29 BAs were grouped into 6 clusters based on similarities in proteomic patterns: the motor and sensory cluster, vision cluster, auditory and Broca’s area cluster, Wernicke’s area c...     

Analysis of the human serum proteome

Changes in serum proteins that signal histopathological states, such as cancer, are useful diagnostic and prognostic biomarkers. Unfortunately, the large dynamic concentration range of proteins in serum makes it a challenging proteome to effectively characterize. Typically, methods to deplete highly abundant proteins to decrease this dynamic protein concentration range are employed, yet such depletion results in removal of important low abundant proteins. A multi-dimensional peptide separation strategy utilizing conventional separation techniques combined with tandem mass spectrometry (MS/MS) was employed for a proteome analysis of human serum. Serum proteins were digested with trypsin and resolved into 20 fractions by ampholyte-free liquid phase isoelectric focusing. These 20 peptide fractions were further fractionated by strong cation-exchange chromatography, each of which was analyzed by microcapillary reversed-phase liquid chromatography coupled online with MS/MS analysis. This investigation resulted in the identification of 1444 unique proteins in serum. Proteins from all functional classes, cellular localization, and abundance levels were identified. This study illustrates that a majority of lower abundance proteins identified in serum are present as secreted or shed species by cells as a result of signalling, necrosis, apoptosis, and hemolysis. These findings show that the protein content of serum is quite reflective of the overall profile of the human organism and a conventional multidimensional fractionation strategy combined with MS/MS is entirely capable of characterizing a significant fraction of the serum proteome. We have constructed a publicly available human serum proteomic database (http://bpp.nci.nih.gov) to provide a reference resource to facilitate future investigations of the vast archive of pathophysiological content in serum. These authors contributed equally to this work.     

湖南烙铁头蛇毒蛋白组分的双向电泳分析

烙铁头蛇是世界上剧毒的蛇种之一,其所携带的毒素能够导致严重的机体损伤。应用蛋白质双向电泳技术,对湖南烙铁头蛇蛇毒蛋白的蛋白质组分进行分析。通过等电聚焦和SDS-PAGE凝胶电泳分析获得完整的烙铁头蛇毒全蛋白质的图谱,经胶体考马斯亮蓝染色后,应用PDQuest软件对蛋白表达谱进行分析。通过等电聚焦和SDS-PAGE凝胶电泳有83个蛋白质组分被检测出来。其中大约90.00%的蛋白质的相对分子质量(Mr)分布在15～45 kDa之间,大约72.29%的蛋白质等电点(pI)在4.0～7.0之间。通过对烙铁头蛇毒的蛋白组学研究,获得其蛇毒蛋白质组分的表征特点,为后续进一步研究各组分的身份和潜在功能奠定基础,既可以提出新的治疗方案又可以为新的药理应用提供宝贵资源。     

Accentuate the Negative:Proteome Comparisons Using the Negative Proteome Database

The availability of complete genome sequence information for diverse organisms including model genetic organisms has ushered in a new era of protein sequence comparisons making it possible to search for commonalities among entire proteomes using the Basic Local Alignment Search Tool (BLAST). Although the identification and analysis of proteins shared by humans and model organisms has proven an invaluable tool to understanding gene function, the sets of proteins unique to a given model organism's proteome have remained largely unexplored. We have constructed a searchable database that allows biologists to identify proteins unique to a given proteome. The Negative Proteome Database (NPD) is populated with pair-wise protein sequence comparisons between each of the following proteomes: Homo sapiens, Mus musculus, Drosophila melanogaster, Caenorhabditis elegans, Saccharomyces cerevisiae, Dictyostelium discoideum, Chlamydomonus reinhardti, Escherichia coli K12, Arabidopsis thaliana and Methanoscarcina acetivorans. Our analysis of negative proteome datasets using the NPD has thus far revealed 107 proteins in humans that may be involved in motile cilia function, 1628 potential pesticide target proteins in flies, 659 proteins shared by flies and humans that are not represented in the less neurologically complex worm proteome, and 180 nuclear encoded human disease associated proteins that are absent from the fly proteome. The NPD is the only online resource where users can quickly perform complex negative and positive comparisons of model organism proteomes. We anticipate that the NPD and the adaptable algorithm which can readily be used to duplicate this analysis on custom sets of proteomes will be an invaluable tool in the investigation of organism specific protein sets.     

Combining experimental and predicted datasets for determination of the subcellular location of proteins in Arabidopsis

Substantial experimental datasets defining the subcellular location of Arabidopsis (Arabidopsis thaliana) proteins have been reported in the literature in the form of organelle proteomes built from mass spectrometry data (approximately 2,500 proteins). Subcellular location for specific proteins has also been published based on imaging of chimeric fluorescent fusion proteins in intact cells (approximately 900 proteins). Further, the more diverse history of biochemical determination of subcellular location is stored in the entries of the Swiss-Prot database for the products of many Arabidopsis genes (approximately 1,800 proteins). Combined with the range of bioinformatic targeting prediction tools and comparative genomic analysis, these experimental datasets provide a powerful basis for defining the final location of proteins within the wide variety of subcellular structures present inside Arabidopsis cells. We have analyzed these published experimental and prediction data to answer a range of substantial questions facing researchers about the veracity of these approaches to determining protein location and their interrelatedness. We have merged these data to form the subcellular location database for Arabidopsis proteins (SUBA), providing an integrated understanding of protein location, encompassing the plastid, mitochondrion, peroxisome, nucleus, plasma membrane, endoplasmic reticulum, vacuole, Golgi, cytoskeleton structures, and cytosol (www.suba.bcs.uwa.edu.au). This includes data on more than 4,400 nonredundant Arabidopsis protein sequences. We also provide researchers with an online resource that may be used to query protein sets or protein families and determine whether predicted or experimental location data exist; to analyze the nature of contamination between published proteome sets; and/or for building theoretical subcellular proteomes in Arabidopsis using the latest experimental data.     

Proteome cataloging and relative quantification of Francisella tularensis tularensis strain Schu4 in 2D PAGE using preparative isoelectric focusing

The protein complement of whole cell extract of the bacterium Francisella tularensis tularensis was analyzed using two-dimensional electrophoresis with preparative isoelectric focusing in the first dimension. The format allows the quantification of relative protein abundance by linear densitometry and extends the potential dynamic range of protein detection by as much as an order of magnitude. The relative abundance and rank order of 136 unique proteins identified in F. tularensis tularensis were established. It is estimated that 16% of the moderately to highly expressed proteins and 8% of all predicted non-pseudogenes were identified by comparing this proteome information with the relative abundance of mRNA as measured by microarray. This rank-ordered proteome list provides an important resource for understanding the pathogenesis of F. tularensis and is a tool for the selection and design of synthetic vaccines. This method represents a useful additional technique to improve whole proteome analyses of simple organisms.     

Subcellular proteomics for determining iron-limited remodeling of plastids in the model diatom Thalassiosira pseudonana (Bacillariophyta)

Diatoms are important primary producers in the world's oceans, yet their growth is constrained in large regions by low bioavailable iron (Fe). Low-Fe stress-induced limitation of primary production is due to requirements for Fe in components of essential metabolic pathways including photosynthesis and other chloroplast plastid functions. Studies have shown that under low-Fe stress, diatoms alter plastid-specific processes, including components of electron transport. These physiological changes suggest changes of protein content and in protein abundances within the diatom plastid. While in silico predictions provide putative information on plastid-localized proteins, knowledge of diatom plastid proteins remains limited in comparison to well-studied model photosynthetic organisms. To address this, we employed shotgun proteomics to investigate the proteome of subcellular plastid-enriched fractions from Thalassiosira pseudonana to gain a better understanding of how the plastid proteome is remodeled in response to Fe limitation. Using mass spectrometry-based peptide identification and quantification, we analyzed T. pseudonana grown under Fe-replete and -limiting conditions. Through these analyses, we inferred the relative quantities of each protein, revealing that Fe limitation regulates major metabolic pathways in the plastid, including the Calvin cycle. Additionally, we observed changes in the expression of light-harvesting proteins. In silico localization predictions of proteins identified in this plastid-enriched proteome allowed for an in-depth comparison of theoretical versus observed plastid-localization, providing evidence for the potential of additional protein import pathways into the diatom plastid.     

Generation and validation of affinity reagents on a proteome-wide level

There is a need for protein-specific affinity reagents to explore the gene products encoded by the genome. Recently, systematic efforts to generate validated affinity reagents on a whole human proteome level have been initiated. There are several issues for such efforts, including choice of antigen, type of affinity reagent, and the subsequent validation of the generated protein-specific binders. The advantages and disadvantages with the different approaches are discussed and the problems related to quality assessment of antibodies to be used in multi-platform applications are addressed. This review also describes the efforts to create a virtual resource of validated antibodies using a community-based portal and summarizes the status and visions for the publicly available human protein atlas (http://www.proteinatlas.org) showing the human protein profiles in a large number of normal and cancer tissues as well as a large set of human cell lines.     

Toward a confocal subcellular atlas of the human proteome

Information on protein localization on the subcellular level is important to map and characterize the proteome and to better understand cellular functions of proteins. Here we report on a pilot study of 466 proteins in three human cell lines aimed to allow large scale confocal microscopy analysis using protein-specific antibodies. Approximately 3000 high resolution images were generated, and more than 80% of the analyzed proteins could be classified in one or multiple subcellular compartment(s). The localizations of the proteins showed, in many cases, good agreement with the Gene Ontology localization prediction model. This is the first large scale antibody-based study to localize proteins into subcellular compartments using antibodies and confocal microscopy. The results suggest that this approach might be a valuable tool in conjunction with predictive models for protein localization.     

Subcellular fractionation enhances proteome coverage of pancreatic duct cells

ObjectivesSubcellular fractionation of whole cell lysates offers a means of simplifying protein mixtures, potentially permitting greater depth of proteomic analysis. Here we compare proteins identified from pancreatic duct cells (PaDC) following organelle enrichment to those identified from PaDC whole cell lysates to determine if the additional procedures of subcellular fractionation increase proteome coverage.MethodsWe used differential centrifugation to enrich for nuclear, mitochondrial, membrane, and cytosolic proteins. We then compared – via mass spectrometry-based analysis – the number of proteins identified from these four fractions with four biological replicates of PaDC whole cell lysates.ResultsWe identified similar numbers of proteins among all samples investigated. In total, 1658 non-redundant proteins were identified in the replicate samples, while 2196 were identified in the subcellular fractionation samples, corresponding to a 30% increase. Additionally, we noted that each organelle fraction was in fact enriched with proteins specific to the targeted organelle.ConclusionsSubcellular fractionation of PaDC resulted in greater proteome coverage compared to PaDC whole cell lysate analysis. Although more labor intensive and time consuming, subcellular fractionation provides greater proteome coverage, and enriches for compartmentalized sub-populations of proteins. Application of this subcellular fractionation strategy allows for a greater depth of proteomic analysis and thus a better understanding of the cellular mechanisms of pancreatic disease.     

Constraints imposed by non‐functional protein–protein interactions on gene expression and proteome size

Crowded intracellular environments present a challenge for proteins to form functional specific complexes while reducing non‐functional interactions with promiscuous non‐functional partners. Here we show how the need to minimize the waste of resources to non‐functional interactions limits the proteome diversity and the average concentration of co‐expressed and co‐localized proteins. Using the results of high‐throughput Yeast 2‐Hybrid experiments, we estimate the characteristic strength of non‐functional protein–protein interactions. By combining these data with the strengths of specific interactions, we assess the fraction of time proteins spend tied up in non‐functional interactions as a function of their overall concentration. This allows us to sketch the phase diagram for baker's yeast cells using the experimentally measured concentrations and subcellular localization of their proteins. The positions of yeast compartments on the phase diagram are consistent with our hypothesis that the yeast proteome has evolved to operate closely to the upper limit of its size, whereas keeping individual protein concentrations sufficiently low to reduce non‐functional interactions. These findings have implication for conceptual understanding of intracellular compartmentalization, multicellularity and differentiation.     

Benchmarks for flexible and rigid transcription factor-DNA docking

Background

Although structural domains in proteins (SDs) are important, half of the regions in the human proteome are currently left with no SD assignments. These unassigned regions consist not only of novel SDs, but also of intrinsically disordered (ID) regions since proteins, especially those in eukaryotes, generally contain a significant fraction of ID regions. As ID regions can be inferred from amino acid sequences, a method that combines SD and ID region assignments can determine the fractions of SDs and ID regions in any proteome.

Results

In contrast to other available ID prediction programs that merely identify likely ID regions, the DICHOT system we previously developed classifies the entire protein sequence into SDs and ID regions. Application of DICHOT to the human proteome revealed that residue-wise ID regions constitute 35%, SDs with similarity to PDB structures comprise 52%, while SDs with no similarity to PDB structures account for the remaining 13%. The last group consists of novel structural domains, termed cryptic domains, which serve as good targets of structural genomics. The DICHOT method applied to the proteomes of other model organisms indicated that eukaryotes generally have high ID contents, while prokaryotes do not. In human proteins, ID contents differ among subcellular localizations: nuclear proteins had the highest residue-wise ID fraction (47%), while mitochondrial proteins exhibited the lowest (13%). Phosphorylation and O-linked glycosylation sites were found to be located preferentially in ID regions. As O-linked glycans are attached to residues in the extracellular regions of proteins, the modification is likely to protect the ID regions from proteolytic cleavage in the extracellular environment. Alternative splicing events tend to occur more frequently in ID regions. We interpret this as evidence that natural selection is operating at the protein level in alternative splicing.

Conclusions

We classified entire regions of proteins into the two categories, SDs and ID regions and thereby obtained various kinds of complete genome-wide statistics. The results of the present study are important basic information for understanding protein structural architectures and have been made publicly available at http://spock.genes.nig.ac.jp/~genome/DICHOT.     

Characterization of human proteins with different subcellular localizations by topological and biological properties

Knowing the protein localization can provide valuable information resource for elucidating protein function. In recent years, with the advances of human genomics and proteomics, it is possible to characterize human proteins that are located in different subcellular localizations. In this study, we used the topological properties and biological properties to characterize human proteins with six subcellular localizations. Almost all of these properties were found to be significantly different among six protein categories. Network topology analysis indicated that several significant topological properties, including the degree and k-core, were higher for the mitochondrial proteins. Biological property analysis showed that the nuclear proteins appeared to be correlated with important biological function. We hope these findings may provide some important help for comprehensive understanding the biological function of proteins, and prediction of protein subcellular localizations in human.