Proteome Size Research Articles

Bioinformatic Comparison of Bacterial Secretomes

The rapid increasing number of completed bacterial genomes provides a good opportunity to compare their proteomes. This study was undertaken to specifically compare and contrast their secretomes—the fraction of the proteome with predicted N-terminal signal sequences, both type I and type II. A total of 176 theoretical bacterial proteomes were examined using the ExProt program. Compared with the Gram-positives, the Gram-negative bacteria were found, on average, to contain a larger number of potential Sec-dependent sequences. In the Gram-negative bacteria but not in the others, there was a positive correlation between proteome size and secretome size, while there was no correlation between secretome size and pathogenicity. Within the Gram-negative bacteria, intracellular pathogens were found to have the smallest secretomes. However, the secretomes of certain bacteria did not fit into the observed pattern. Specifically, the secretome of Borrelia burgdoferi has an unusually large number of putative lipoproteins, and the signal peptides of mycoplasmas show closer sequence similarity to those of the Gram-negative bacteria. Our analysis also suggests that even for a theoretical minimal genome of 300 open reading frames, a fraction of this gene pool (up to a maximum of 20%) may code for proteins with Sec-dependent signal sequences.

Accurate and Scalable Techniques for the Complex/Pathway Membership Problem in Protein Networks

A protein network shows physical interactions as well as functional associations. An important usage of such networks is to discover unknown members of partially known complexes and pathways. A number of methods exist for such analyses, and they can be divided into two main categories based on their treatment of highly connected proteins. In this paper, we show that methods that are not affected by the degree (number of linkages) of a protein give more accurate predictions for certain complexes and pathways. We propose a network flow-based technique to compute the association probability of a pair of proteins. We extend the proposed technique using hierarchical clustering in order to scale well with the size of proteome. We also show that top-k queries are not suitable for a large number of cases, and threshold queries are more meaningful in these cases. Network flow technique with clustering is able to optimize meaningful threshold queries and answer them with high efficiency compared to a similar method that uses Monte Carlo simulation.

FASTKD2 Nonsense Mutation in an Infantile Mitochondrial Encephalomyopathy Associated with Cytochrome C Oxidase Deficiency

In two siblings we found a mitochondrial encephalomyopathy, characterized by developmental delay, hemiplegia, convulsions, asymmetrical brain atrophy, and low cytochrome c oxidase (COX) activity in skeletal muscle. The disease locus was identified on chromosome 2 by homozygosity mapping; candidate genes were prioritized for their known or predicted mitochondrial localization and then sequenced in probands and controls. A homozygous nonsense mutation in the KIAA0971 gene segregated with the disease in the proband family. The corresponding protein is known as fas activated serine-threonine kinase domain 2, FASTKD2. Confocal immunofluorescence colocalized a tagged recombinant FASTKD2 protein with mitochondrial markers, and membrane-potential-dependent in vitro mitochondrial import was demonstrated in isolated mitochondria. In staurosporine-induced-apoptosis experiments, decreased nuclear fragmentation was detected in treated mutant versus control fibroblasts. In conclusion, we found a loss-of-function mutation in a gene segregating with a peculiar mitochondrial encephalomyopathy associated with COX deficiency in skeletal muscle. The corresponding protein is localized in the mitochondrial inner compartment. Preliminary data indicate that FASTKD2 plays a role in mitochondrial apoptosis.

Impacts of protein–protein interaction domains on organism and network complexity

It has been a puzzle that genome or proteome sizes are not correlated with the complexity of the organisms. Although alternative splicing and noncoding and regulatory elements explain some of the differences, the complexity of the protein interaction network and regulatory network may provide additional explanations. Here, we collected 642 domains that mediate protein-protein interactions (PPIs) and examined the evolution of the PPI domains and its impact on organismal complexity and PPI network complexity. In agreement with previous more general studies of protein domains, a significant expansion of PPI domains per proteome was found in metazoa. We also found both the number and coverage of PPI domains per protein increased. However, a better correlation with complexity was seen with increasing PPI domain coverage per protein, so that proteins in complex organisms are more compact and specialized in PPI. Such a structural adaptation of the proteins is correlated with the number of interactions that the proteins can make in PPI networks, and seems to be a more favorable way to increase network connectivity than other structural adaptations.

The Proteomic Constraint and Its Role in Molecular Evolution

Recently, the concept of a "Proteomic Constraint" was introduced to explain the frequency of genetic code deviations in mitochondrial genomes. The Proteomic Constraint was proposed to be proportional to the size of the mitochondrially encoded proteome, hence small proteomes are expected to experience smaller total numbers of errors resulting from genetic code deviations, leading to less likelihood of causing lethality. The concept is now extended to encompass several other aspects of the genetic information system. When the Proteomic Constraint is small, it is proposed that there is little selective pressure to evolve or maintain error correction mechanisms, as a result of the smaller total number of errors that accumulate. Conversely, a large Proteomic Constraint is proposed to result in a correspondingly large selective pressure to evolve or maintain error correction mechanisms. Differences in the size of the Proteomic Constraint can help to explain differences in replicational, transcriptional, and translational fidelities between genomes. A key piece of evidence is the existence of negative power law relationships between proteome size and error rates; these are demonstrated to be diagnostic of the action of the Proteomic Constraint. The Proteomic Constraint is argued to be a major factor determining mutation rates in a diverse range of DNA genomes, implying that mutation rates are clock like. A small Proteomic Constraint partly explains why RNA viruses possess high mutation rates. A reduced Proteomic Constraint in intracellular pathogenic bacteria predicts a drift upwards in mutation rates. Differences in the Proteomic Constraint also appear to be linked to differences in recombination rates between eukaryotes. In addition, a reduced Proteomic Constraint may explain features of resident genomes, such as loss of DNA repair pathways, increased substitution rates, and AT biases, in addition to the occurrence of genetic code deviations. Thus, it is argued that the Proteomic Constraint is a universal factor that influences a wide range of properties of the genetic information system.

What Does Systems Biology Mean for Drug Development?

The complexity and flexibility of cellular architectures is increasingly recognized by impressive progress on the side of molecular analytics, i.e. proteomics, genomics and metabolomics. One of the messages from systems biology is that the number of molecular species in cellular networks is orders of magnitude bigger than anticipated by genomic analysis, in particular by fast posttranslational modifications of proteins. The requirements to manage external signals, integrate spatiotemporal signal transduction inside an organism and at the same time optimizing networks of biochemical and chemical reactions result in chemically extremely fine tuned molecular entities. Chemical side reactions of enzymatic activity, like e.g. random oxidative damage of proteins by free radicals during aging constantly introduce epigenetic alterations of protein targets. These events gradually and on an individual stochastic scale, keep modifying activities of these targets, and their affinities and selectivities towards biological and pharmacological ligands. One further message is that many of the key reactions in living systems are essentially based on interactions of low affinities and even low selectivities. This principle is responsible for the enormous flexibility and redundancy of cellular circuitries. So, in complex disorders like cancer or neurodegenerative diseases, which are rooted in relatively subtle and multimodal dysfunction of important physiologic pathways, drug discovery programs based on the concept of high affinity/high specificity compounds ("one-target, one-disease"), which still dominate the pharmaceutical industry increasingly turn out to be unsuccessful. Despite improvements in rational drug design and high throughput screening methods, the number of novel, single-target drugs fell much behind expectations during the past decade and the treatment of "complex diseases" remains a most pressing medical need. Currently a change of paradigm can be observed with regard to a new focus on agents that modulate multiple targets simultaneously. Targeting cellular function as a system rather than on the level of the single protein molecule significantly increases the size of the drugable proteome and is expected to introduce novel classes of multi-target drugs with fewer adverse effects and toxicity. Multiple target approaches have recently been used to design medications against atherosclerosis, cancer, depression, psychosis and neurodegenerative diseases. A focussed approach towards "systemic" drugs will certainly require the development of novel computational and mathematical concepts for appropriate modelling of complex data and extraction of "screenable" information from biological systems essentially ruled by deterministic chaotic processes on a background of individual stochasticity.

Protein Interaction Network. Double Exponential Model

proper theoretical description of the distribution of the node degree for yeast protein-protein interaction network was investigated to deal with the observed discrepancy between usually proposed models and the existing data. power law or the generalized power law with exponential cut-off were shown to be inaccurate within a wide range of degree values. Proposed linear-combination-of-exponentialdecays- method exactly characterizing the distribution by the spectrum of decay constants revealed two separate parameter domains. A consequent hypothesis that the node degree distribution could follow the universal double exponential law was successfully verified by selected model comparison using the AIC criterion. BIND and DIP data for H. pylori, E. coli, S. cerevisiae, D. melanogaster, C. elegans and A. thaliana were used for this purpose. A linear change in the magnitude of the distribution components with proteome size was observed, manifesting the evolutional stability of the process of developing the protein interaction network. Proposed kinetic model of protein evolution, considering the two hypothetical protein classes, first, with a relatively rapid emerging rate and a short characteristic residence time, and the second one, with the opposite properties, analytically described the nature of bi-exponential pattern. model presents a situation in which evolutionary conserved proteins increase their interactions due to specific kinetic conditions. Thus, we oppose the opinion that the majority of such interactions are biologically significant, and, therefore the older parts of interactome are more complex. We believe that our interactome results support the hypothesis of Stuart Kaufman, presented in his book The Origin of Order, that random mutations and natural selection constitute the origin of order and complexity.

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.

Towards defining the nuclear proteome

Direct evidence is reported for 2,568 mammalian proteins within the nuclear proteome, consisting of at least 14% of the entire proteome.

DBD––taxonomically broad transcription factor predictions: new content and functionality

DNA-binding domain (DBD) is a database of predicted sequence-specific DNA-binding transcription factors (TFs) for all publicly available proteomes. The proteomes have increased from 150 in the initial version of DBD to over 700 in the current version. All predicted TFs must contain a significant match to a hidden Markov model representing a sequence-specific DNA-binding domain family. Access to TF predictions is provided through http://transcriptionfactor.org, where new search options are now provided such as searching by gene names in model organisms, searching for all proteins in a particular DBD family and specific organism. We illustrate the application of this type of search facility by contrasting trends of DBD family occurrence throughout the tree of life, highlighting the clear partition between eukaryotic and prokaryotic DBD expansions. The website content has been expanded to include dedicated pages for each TF containing domain assignment details, gene names, links to external databases and links to TFs with similar domain arrangements. We compare the increase in number of predicted TFs with proteome size in eukaryotes and prokaryotes. Eukaryotes follow a slower rate of increase in TFs than prokaryotes, which could be due to the presence of splice variants or an increase in combinatorial control.

Towards completion of the Earth's proteome

New protein sequences are deposited in databases at an accelerating pace; however, many of these are homologous to known proteins and could be considered redundant. If all historical releases of the protein database are analysed using the original sequence-clustering procedure described here, the fraction of newly sequenced proteins that are redundant is increasing. We interpret this as an indication that the sequencing of the Earth's proteome--the complete set of proteins on Earth--is approaching completion. We estimate the approximate size of the Earth's proteome to be 5 million sequences, most of which will be identified during the next 5 years. As the Earth's proteome nears completion, cluster analysis of the protein database will become essential to identify under-explored taxa to which future sequencing efforts should be directed and to focus research on protein families without experimental characterization.

Non‐electrophoretic differential detergent fractionation proteomics using frozen whole organs

Non-electrophoretic methods based on two-dimensional liquid chromatography followed directly by tandem mass spectrometry (2D-LC/MS(2)) have become the preferred method for high-throughput expression proteomics and are widely applied to fresh tissues. Pre-fractionation techniques are also used in combination with 2D-LC/MS(2) to both increase the proteome size and to assign cellular locations. Data from such experiments have become central to systems biology analyses. Here we apply a differential detergent (pre)fractionation (DDF) followed by 2D-LC/MS(2) to frozen archival tissues. Our results show that by using frozen archival tissues, we do not lose proteome coverage or the ability to assign proteins to cellular compartments. In addition, we were able to assign 'biological process' Gene Ontology (GO) annotations, which will facilitate systems biological modeling of our proteomics data.

Occurrence of Copper Proteins through the Three Domains of Life: A Bioinformatic Approach

In high-throughput genome-level protein investigation efforts, such as Structural Genomics, the systematic experimental characterization of metal-binding properties (i.e., the investigation of the metalloproteome) is not always pursued and remains far from trivial. In the present work, we have applied a bioinformatic approach to investigate the occurrence of (putative) copper-binding proteins in 57 different organisms spanning the entire tree of life. We found that the size of the copper proteome is generally less than 1% of the total proteome of an organism, in both eukaryotes and prokaryotes. The occurrence of copper-binding proteins is relatively scarce when compared to that of zinc-binding proteins and of non-heme iron proteins. This may be due to both poorer bioavailability (in particular with respect to iron in the ancient world) and the complexity of copper chemistry and the risks associated with it, which may have adversely affected natural selection of copper-binding proteins. The present analysis shows that there is a strong relationship between the metal coordination sphere and protein function. A network involving proteins having roles in both copper transport and respiration was identified, parts or all of which are detected in the majority of the organisms examined.

Comparative Genomics of Protists: New Insights into the Evolution of Eukaryotic Signal Transduction and Gene Regulation

Data from protist genomes suggest that eukaryotes show enormous variability in their gene complements, especially of genes coding regulatory proteins. Overall counts of eukaryotic signaling proteins show weak nonlinear scaling with proteome size, but individual superfamilies of signaling domains might show vast expansions in certain protists. Alteration of domain architectural complexity of signaling proteins and repeated lineage-specific reshaping of architectures might have played a major role in the emergence of new signaling interactions in different eukaryotes. Lateral transfer of various signaling domains from bacteria or from hosts, in parasites such as apicomplexans, appears to also have played a major role in the origin of new functional networks. Lineage-specific expansion of regulatory proteins, particularly of transcription factors, has played a critical role in the adaptive radiation of different protist lineages. Comparative genomics allows objective reconstruction of the ancestral conditions and subsequent diversification of several regulatory systems involved in phosphorylation, cyclic nucleotide signaling, Ubiquitin conjugation, chromatin remodeling, and posttranscriptional gene silencing.

M.I.3 Nuclear genes involved in mitochondrial syndromes

Maternally inherited mitochondrial DNA encodes only 13 subunits of the respiratory complexes. All other proteins involved in a functional respiratory chain are nucleus-encoded and imported into mitochondria. Because of the extreme complexity of OXPHOS, the number of candidate disease genes is enormous, and tends to coincide with the size of the mitochondrial proteome itself (>1000 gene products). Important advancements have recently been made in three main areas: (a) defects associated with mtDNA instability; (b) defects of mtDNA translation; (c) defects of coenzyme Q biosynthesis. The first group comprises Mendelian traits characterized by either the accumulation of multiple mtDNA deletions or tissue-specific mtDNA depletion. Mutations in polymerase gamma, the core enzyme of mtDNA replication, and in several other factors related to mtDNA maintenance, are responsible for a continuum clinical spectrum ranging from autosomal dominant or recessive PEO, to spino-cerebellar ataxia with epilepsy, to Alpers’ syndrome. The second group comprises severe encephalopathies caused by mutations in mitochondrial translation factors, but also a peculiar multisystemic disease characterized by mitochondrial myopathy with lactic acidosis and sideroblastic anemia. MLASA is caused by mutations in pseudouridylate synthase 1, an isomerase that converts uridine into pseudouridine of both cytosolic and mitochondrial tRNAs. The third group consists of primary CoQ10 deficiency, a potentially treatable condition with a clinical spectrum that encompasses at least five major phenotypes. Mutations in the CoQ10 biosynthetic genes, COQ2, PDSS1, and PDSS2, have been identified in children with severe multisystemic disease, while the other forms still await genetic characterization.

Mapping the Human Proteome Using Antibodies

A great need exists for the generation of protein-specific affinity reagents to explore the human proteome. With this background, the HUPO Antibody Initiative (HAI) was started in 2005 with the objective to promote the generation and use of well validated antibodies toward human proteins. The initiative is currently chaired by Mathias Uhlen with Michael Snyder (Yale University, New Haven, CT) and Peter Hudson (Commonwealth Scientific and Industrial Research Organisation (CSIRO), Parkville, Victoria, Australia) as co-chairs. Several workshops have since been held to coordinate efforts and funding in the area. At the end of 2005, the National Cancer Institute (NCI) of the National Institutes of Health organized a workshop in which the conclusion was that validated and well characterized affinity capture reagents (e.g. antibodies, aptamers, and affibodies) will play a key role in proteomics research platforms for the prevention, early detection, treatment, and monitoring of cancer (1). Recently (May 2007), the Protein Capture Tools proposal was selected by the National Institutes of Health Roadmap initiative for staged implementation (www.nihroadmap.nih.gov). Similarly a European Union-funded program, ProteomeBinder, started in 2006 with the aim to coordinate efforts in Europe involving more than 20 research groups from both the European Union and the United States (2). The mission of HAI is to endorse the production of well characterized sets of renewable (preferably monoclonal) antibodies or other affinity reagents for various protein assays and to ultimately generate paired antibodies/affinity reagents for all human proteins produced as a sustainable resource. The paired antibodies should preferably be produced toward two separate and non-overlapping epitopes of the same protein target to facilitate the validation of specificity by allowing comparisons between the paired antibodies with regard to staining pattern across various analysis platforms. Paired antibodies also allow for coupled assays with high signal-tonoise ratios such as capture/detection (sandwich) assays and other assays with requirement for dual binding to increase specificity and sensitivity. The presence of a multitude of protein isoforms, such as splice variants, combinatorial variants, post-translational modifications, specific proteolysis, and genetic variability (alleles), makes the human proteome space huge, and the initial ambition of a genome-wide antibody-based proteomics effort has therefore been suggested to be directed to the nonredundant proteome defined as a representative protein from every gene locus (3). The current estimate of the size of the non-redundant proteome is 23,000 proteins (Ensembl, www.ensembl.org), and at present more than 16,000 of those have been manually annotated by the Swiss-Prot/UniProt effort (www.uniprot.org). The HUPO Antibody Initiative consists of two separate activities: 1) the development of a virtual antibody resource of validated antibodies and 2) a protein atlas for the expression and localization of human proteins in normal and disease tissues. The Antibody Resource database is aimed to produce a comprehensive catalogue of validated antibodies through a community-based effort. The objective is to create a format for submission of data on affinity reagents along with application-specific validation. The primary data from the quality assurance should preferably be submitted and made available through the database. This initiative depends on input from a large number of academic groups and commercial companies. A pilot version of the resource database will be presented at the fourth annual HAI workshop to be held in Seoul, Korea on Sunday October 7, 2007. The workshop will focus on the progress and time plans for the antibody portal. The Human Protein Atlas program (www.proteinatlas.org) is aimed to provide a comprehensive and annotated database of high resolution images showing tissue profiles in normal and cancer tissues (4). Version 2.0 of this publicly available database was released at the 6th annual HUPO conference in Long Beach in 2006 doubling the number of proteins to more than 1300 human proteins with 1.2 million immunohistochemistry images. A new version (3.0) will be released October 8, 2007 at the 7th annual HUPO conference in Seoul, Korea, and the plan is to again double the content of the database and to release at least 2 million images representing more than 2500 human proteins. The short term objective of this effort is to generate data for 8000 proteins by the end of 2009, and the mission is to generate a draft version of the complete proteome by 2015, 10 years after the public release of the first version of the Protein Atlas portal.

The relationships between the isoelectric point and: length of proteins, taxonomy and ecology of organisms

BackgroundThe distribution of isoelectric point (pI) of proteins in a proteome is universal for all organisms. It is bimodal dividing the proteome into two sets of acidic and basic proteins. Different species however have different abundance of acidic and basic proteins that may be correlated with taxonomy, subcellular localization, ecological niche of organisms and proteome size.ResultsWe have analysed 1784 proteomes encoded by chromosomes of Archaea, Bacteria, Eukaryota, and also mitochondria, plastids, prokaryotic plasmids, phages and viruses. We have found significant correlation in more than 95% of proteomes between the protein length and pI in proteomes – positive for acidic proteins and negative for the basic ones. Plastids, viruses and plasmids encode more basic proteomes while chromosomes of Archaea, Bacteria, Eukaryota, mitochondria and phages more acidic ones. Mitochondrial proteomes of Viridiplantae, Protista and Fungi are more basic than Metazoa. It results from the presence of basic proteins in the former proteomes and their absence from the latter ones and is related with reduction of metazoan genomes. Significant correlation was found between the pI bias of proteomes encoded by prokaryotic chromosomes and proteomes encoded by plasmids but there is no correlation between eukaryotic nuclear-coded proteomes and proteomes encoded by organelles. Detailed analyses of prokaryotic proteomes showed significant relationships between pI distribution and habitat, relation to the host cell and salinity of the environment, but no significant correlation with oxygen and temperature requirements. The salinity is positively correlated with acidicity of proteomes. Host-associated organisms and especially intracellular species have more basic proteomes than free-living ones. The higher rate of mutations accumulation in the intracellular parasites and endosymbionts is responsible for the basicity of their tiny proteomes that explains the observed positive correlation between the decrease of genome size and the increase of basicity of proteomes. The results indicate that even conserved proteins subjected to strong selectional constraints follow the global trend in the pI distribution.ConclusionThe distribution of pI of proteins in proteomes shows clear relationships with length of proteins, subcellular localization, taxonomy and ecology of organisms. The distribution is also strongly affected by mutational pressure especially in intracellular organisms.

Epitope-Driven TB Vaccine Development: A Streamlined Approach Using Immuno-Informatics, ELISpot Assays, and HLA Transgenic Mice

New vaccine candidates that might better control the worldwide prevalence of Mycobacterium tuberculosis (Mtb) have yet to be described. Strong CD4+ T cell-mediated immune response (CMI) is correlated with protection from the development of TB disease; however, the selection of suitable vaccine antigens has been thwarted by the size and complexity of the (Mtb) proteome, and by the relative difficulty of delivering these antigens in the right immunological context. One possible solution is to develop immunotherapeutic vaccines for TB that are based on T cell epitopes representing multiple antigens. This text illustrates the stepwise development of epitope-driven vaccines from in silico epitope mapping to testing the vaccine in a live Mtb challenge model. First, we used the whole genome Mtb microarray to identify bacterial proteins expressed under the conditions thought to model Mtb survival and replication in human macrophages. Eighteen of these proteins were also found by Behr et al. to be absent from at least one strain of BCG; the sequences of these eighteen proteins were then screened for T-cell epitopes using the immuno-informatics algorithm, EpiMatrix. Of the seventeen representative epitopes evaluated in ELISpot assays, all seventeen were confirmed to elicit interferon (IFN)-gamma secretion by PBMC from Mtb-exposed subjects. A parallel live Mtb challenge study in mice showed prototype epitope-based TB vaccines to be robustly immunogenic but not as effective as BCG. These experiments illustrate the use of immuno-informatics tools for vaccine development and describe a pathway for the development of a more effective, epitope-driven, immunotherapeutic vaccine for TB.

How to design multi-target drugs

Despite improved rational drug design and a remarkable progress in genomic, proteomic and high-throughput screening methods, the number of novel, single-target drugs has fallen far behind expectations during the past decade. Multi-target drugs multiply the number of pharmacologically relevant target molecules by introducing a set of indirect, network-dependent effects. Parallel with this, the low-affinity binding of multi-target drugs eases the constraints of druggability and significantly increases the size of the druggable proteome. These effects tremendously expand the number of potential drug targets and introduce novel classes of multi-target drugs with smaller side effects and toxicity. Here, the authors review the recent progress in this field, compare possible network attack strategies and propose several methods to find target-sets for multi-target drugs.

Global Gene Expression Patterns ofNostoc punctiformein Steady-State Dinitrogen-Grown Heterocyst-Containing Cultures and at Single Time Points during the Differentiation of Akinetes and Hormogonia

The vegetative cells of the filamentous cyanobacterium Nostoc punctiforme can differentiate into three mutually exclusive cell types: nitrogen-fixing heterocysts, spore-like akinetes, and motile hormogomium filaments. A DNA microarray consisting of 6,893 N. punctiforme genes was used to identify the global transcription patterns at single time points in the three developmental states, compared to those in ammonium-grown time zero cultures. Analysis of ammonium-grown cultures yielded a transcriptome of 2,935 genes, which is nearly twice the size of a soluble proteome. The NH(4)(+)-grown transcriptome was enriched in genes encoding core metabolic functions. A steady-state N(2)-grown (heterocyst-containing) culture showed differential transcription of 495 genes, 373 of which were up-regulated. The majority of the up-regulated genes were predicted from studies of heterocyst differentiation and N(2) fixation; other genes are candidates for more detailed genetic analysis. Three days into the developmental process, akinetes showed a similar number of differentially expressed genes (497 genes), which were equally up- and down-regulated. The down-regulated genes were enriched in core metabolic functions, consistent with entry into a nongrowth state. There were relatively few adaptive genes up-regulated in 3-day akinetes, and there was little overlap with putative heterocyst developmental genes. There were 1,827 differentially transcribed genes in 24-h hormogonia, which was nearly fivefold greater than the number in akinete-forming or N(2)-fixing cultures. The majority of the up-regulated adaptive genes were genes encoding proteins for signal transduction and transcriptional regulation, which is characteristic of a motile filament that is poised to sense and respond to the environment. The greatest fraction of the 883 down-regulated genes was involved in core metabolism, also consistent with entry into a nongrowth state. The differentiation of heterocysts (steady state, N(2) grown), akinetes, and hormogonia appears to involve the up-regulation of genes distinct for each state.