Broad patterns of gene expression revealed by clustering analysis of tumor and normal colon tissues probed by oligonucleotide arrays

Oligonucleotide arrays can provide a broad picture of the state of the cell, by monitoring the expression level of thousands of genes at the same time. It is of interest to develop techniques for extracting useful information from the resulting data sets. Here we report the application of a two-way clustering method for analyzing a data set consisting of the expression patterns of different cell types. Gene expression in 40 tumor and 22 normal colon tissue samples was analyzed with an Affymetrix oligonucleotide array complementary to more than 6,500 human genes. An efficient two-way clustering algorithm was applied to both the genes and the tissues, revealing broad coherent patterns that suggest a high degree of organization underlying gene expression in these tissues. Coregulated families of genes clustered together, as demonstrated for the ribosomal proteins. Clustering also separated cancerous from noncancerous tissue and cell lines from in vivo tissues on the basis of subtle distributed patterns of genes even when expression of individual genes varied only slightly between the tissues. Two-way clustering thus may be of use both in classifying genes into functional groups and in classifying tissues based on gene expression.

Robustness in simple biochemical networks

Cells use complex networks of interacting molecular components to transfer and process information. These "computational devices of living cells" are responsible for many important cellular processes, including cell-cycle regulation and signal transduction. Here we address the issue of the sensitivity of the networks to variations in their biochemical parameters. We propose a mechanism for robust adaptation in simple signal transduction networks. We show that this mechanism applies in particular to bacterial chemotaxis. This is demonstrated within a quantitative model which explains, in a unified way, many aspects of chemotaxis, including proper responses to chemical gradients. The adaptation property is a consequence of the network's connectivity and does not require the 'fine-tuning' of parameters. We argue that the key properties of biochemical networks should be robust in order to ensure their proper functioning.

Robustness in bacterial chemotaxis

Networks of interacting proteins orchestrate the responses of living cells to a variety of external stimuli, but how sensitive is the functioning of these protein networks to variations in their biochemical parameters? One possibility is that to achieve appropriate function, the reaction rate constants and enzyme concentrations need to be adjusted in a precise manner, and any deviation from these 'fine-tuned' values ruins the network's performance. An alternative possibility is that key properties of biochemical networks are robust; that is, they are insensitive to the precise values of the biochemical parameters. Here we address this issue in experiments using chemotaxis of Escherichia coli, one of the best-characterized sensory systems. We focus on how response and adaptation to attractant signals vary with systematic changes in the intracellular concentration of the components of the chemotaxis network. We find that some properties, such as steady-state behaviour and adaptation time, show strong variations in response to varying protein concentrations. In contrast, the precision of adaptation is robust and does not vary with the protein concentrations. This is consistent with a recently proposed molecular mechanism for exact adaptation, where robustness is a direct consequence of the network's architecture.

Noise in protein expression scales with natural protein abundance

Noise in gene expression is generated at multiple levels, such as transcription and translation, chromatin remodeling and pathway-specific regulation. Studies of individual promoters have suggested different dominating noise sources, raising the question of whether a general trend exists across a large number of genes and conditions. We examined the variation in the expression levels of 43 Saccharomyces cerevisiae proteins, in cells grown under 11 experimental conditions. For all classes of genes and under all conditions, the expression variance was approximately proportional to the mean; the same scaling was observed at steady state and during the transient responses to the perturbations. Theoretical analysis suggests that this scaling behavior reflects variability in mRNA copy number, resulting from random 'birth and death' of mRNA molecules or from promoter fluctuations. Deviation of coexpressed genes from this general trend, including high noise in stress-related genes and low noise in proteasomal genes, may indicate fluctuations in pathway-specific regulators or a differential activation pattern of the underlying gene promoters.

Mechanisms of noise-resistance in genetic oscillators

A wide range of organisms use circadian clocks to keep internal sense of daily time and regulate their behavior accordingly. Most of these clocks use intracellular genetic networks based on positive and negative regulatory elements. The integration of these "circuits" at the cellular level imposes strong constraints on their functioning and design. Here, we study a recently proposed model [Barkai, N. & Leibler, S. (2000) Nature (London), 403, 267-268] that incorporates just the essential elements found experimentally. We show that this type of oscillator is driven mainly by two elements: the concentration of a repressor protein and the dynamics of an activator protein forming an inactive complex with the repressor. Thus, the clock does not need to rely on mRNA dynamics to oscillate, which makes it especially resistant to fluctuations. Oscillations can be present even when the time average of the number of mRNA molecules goes below one. Under some conditions, this oscillator is not only resistant to but, paradoxically, also enhanced by the intrinsic biochemical noise.

Revealing modular organization in the yeast transcriptional network

Standard clustering methods can classify genes successfully when applied to relatively small data sets, but have limited use in the analysis of large-scale expression data, mainly owing to their assignment of a gene to a single cluster. Here we propose an alternative method for the global analysis of genome-wide expression data. Our approach assigns genes to context-dependent and potentially overlapping 'transcription modules', thus overcoming the main limitations of traditional clustering methods. We use our method to elucidate regulatory properties of cellular pathways and to characterize cis-regulatory elements. By applying our algorithm systematically to all of the available expression data on Saccharomyces cerevisiae, we identify a comprehensive set of overlapping transcriptional modules. Our results provide functional predictions for numerous genes, identify relations between modules and present a global view on the transcriptional network.

Robustness of the BMP morphogen gradient in Drosophila embryonic patterning

Developmental patterning relies on morphogen gradients, which generally involve feedback loops to buffer against perturbations caused by fluctuations in gene dosage and expression. Although many gene components involved in such feedback loops have been identified, how they work together to generate a robust pattern remains unclear. Here we study the network of extracellular proteins that patterns the dorsal region of the Drosophila embryo by establishing a graded activation of the bone morphogenic protein (BMP) pathway. We find that the BMP activation gradient itself is robust to changes in gene dosage. Computational search for networks that support robustness shows that transport of the BMP class ligands (Scw and Dpp) into the dorsal midline by the BMP inhibitor Sog is the key event in this patterning process. The mechanism underlying robustness relies on the ability to store an excess of signalling molecules in a restricted spatial domain where Sog is largely absent. It requires extensive diffusion of the BMP-Sog complexes, coupled with restricted diffusion of the free ligands. We show experimentally that Dpp is widely diffusible in the presence of Sog but tightly localized in its absence, thus validating a central prediction of our theoretical study.

Two strategies for gene regulation by promoter nucleosomes

Chromatin structure is central for the regulation of gene expression, but its genome-wide organization is only beginning to be understood. Here, we examine the connection between patterns of nucleosome occupancy and the capacity to modulate gene expression upon changing conditions, i.e., transcriptional plasticity. By analyzing genome-wide data of nucleosome positioning in yeast, we find that the presence of nucleosomes close to the transcription start site is associated with high transcriptional plasticity, while nucleosomes at more distant upstream positions are negatively correlated with transcriptional plasticity. Based on this, we identify two typical promoter structures associated with low or high plasticity, respectively. The first class is characterized by a relatively large nucleosome-free region close to the start site coupled with well-positioned nucleosomes further upstream, whereas the second class displays a more evenly distributed and dynamic nucleosome positioning, with high occupancy close to the start site. The two classes are further distinguished by multiple promoter features, including histone turnover, binding site locations, H2A.Z occupancy, expression noise, and expression diversity. Analysis of nucleosome positioning in human promoters reproduces the main observations. Our results suggest two distinct strategies for gene regulation by chromatin, which are selectively employed by different genes.

Similarities and differences in genome-wide expression data of six organisms

Comparing genomic properties of different organisms is of fundamental importance in the study of biological and evolutionary principles. Although differences among organisms are often attributed to differential gene expression, genome-wide comparative analysis thus far has been based primarily on genomic sequence information. We present a comparative study of large datasets of expression profiles from six evolutionarily distant organisms: S. cerevisiae, C. elegans, E. coli, A. thaliana, D. melanogaster, and H. sapiens. We use genomic sequence information to connect these data and compare global and modular properties of the transcription programs. Linking genes whose expression profiles are similar, we find that for all organisms the connectivity distribution follows a power-law, highly connected genes tend to be essential and conserved, and the expression program is highly modular. We reveal the modular structure by decomposing each set of expression data into coexpressed modules. Functionally related sets of genes are frequently coexpressed in multiple organisms. Yet their relative importance to the transcription program and their regulatory relationships vary among organisms. Our results demonstrate the potential of combining sequence and expression data for improving functional gene annotation and expanding our understanding of how gene expression and diversity evolved.

Defining transcription modules using large-scale gene expression data

MOTIVATION

Large-scale gene expression data comprising a variety of cellular conditions hold the promise of a global view on the transcription program. While conventional clustering algorithms have been successfully applied to smaller datasets, the utility of many algorithms for the analysis of large-scale data is limited by their inability to capture combinatorial and condition-specific co-regulation. In addition, there is an increasing need to integrate the rapidly accumulating body of other high-throughput biological data with the expression analysis. In a previous work, we introduced the signature algorithm, which overcomes the problems of conventional clustering and allows for intuitive integration of additional biological data. However, this approach is constrained by the comprehensiveness of relevant external data and its lacking ability to capture hierarchical modularity.

METHODS

We present a novel method for the analysis of large-scale expression data, which assigns genes into context-dependent and potentially overlapping regulatory units. We introduce the notion of a transcription module as a self-consistent regulatory unit consisting of a set of co-regulated genes as well as the experimental conditions that induce their co-regulation. Self-consistency is defined by a rigorous mathematical criterion. We propose an efficient algorithm to identify such modules, which is based on the iterative application of the signature algorithm. A threshold parameter that determines the resolution of the modular decomposition is introduced.

RESULTS

The method is applied systematically to over 1000 expression profiles of the yeast Saccharomyces cerevisiae, and the results are presented using two complementary visualization schemes we developed. The average biological coherence, as measured by the conservation of putative cis-regulatory motifs between four related yeast species, is higher for transcription modules than for clusters identified by other methods applied to the same dataset. Our method is related to singular value decomposition (SVD) and to the pairwise average linkage clustering algorithm. It extends SVD by filtering out noise in the expression data and offering variable resolution to reveal hierarchical organization. It furthermore has the advantage over both methods of capturing overlapping modules in the presence of combinatorial regulation.

SUPPLEMENTARY INFORMATION

http://www.weizmann.ac.il/~barkai/modules

Rewiring of the Yeast Transcriptional Network Through the Evolution of Motif Usage

Recent experiments revealed large-scale differences in the transcription programs of related species, yet little is known about the genetic basis underlying the evolution of gene expression and its contribution to phenotypic diversity. Here we describe a large-scale modulation of the yeast transcription program that is connected to the emergence of the capacity for rapid anaerobic growth. Genes coding for mitochondrial and cytoplasmic ribosomal proteins display a strongly correlated expression pattern in Candida albicans, but this correlation is lost in the fermentative yeast Saccharomyces cerevisiae. We provide evidence that this change in gene expression is connected to the loss of a specific cis-regulatory element from dozens of genes following the apparent whole-genome duplication event. Our results shed new light on the genetic mechanisms underlying the large-scale evolution of transcriptional networks.

A genetic signature of interspecies variations in gene expression

Phenotypic diversity is generated through changes in gene structure or gene regulation. The availability of full genomic sequences allows for the analysis of gene sequence evolution. In contrast, little is known about the principles driving the evolution of gene expression. Here we describe the differential transcriptional response of four closely related yeast species to a variety of environmental stresses. Genes containing a TATA box in their promoters show an increased interspecies variability in expression, independent of their functional association. Examining additional data sets, we find that this enhanced expression divergence of TATA-containing genes is consistent across all eukaryotes studied to date, including nematodes, fruit flies, plants and mammals. TATA-dependent regulation may enhance the sensitivity of gene expression to genetic perturbations, thus facilitating expression divergence at particular genetic loci.

Self-enhanced ligand degradation underlies robustness of morphogen gradients

Morphogen gradients provide long-range positional information by extending across a developing field. To ensure reproducible patterning, their profile is invariable despite genetic or environmental fluctuations. Common models assume a morphogen profile that decays exponentially. Here, we show that exponential profiles cannot, at the same time, buffer fluctuations in morphogen production rate and define long-range gradients. To comply with both requirements, morphogens should decay rapidly close to their source but at a significantly slower rate over most of the field. Numerical search revealed two network designs that support robustness to fluctuations in morphogen production rate. In both cases, morphogens enhance their own degradation, leading to a higher degradation rate close to their source. This is achieved through reciprocal interactions between the morphogen and its receptor. The two robust networks are consistent with properties of the Wg and Hh morphogens in the Drosophila wing disc and provide novel insights into their function.

Principles of transcriptional control in the metabolic network of Saccharomyces cerevisiae

Cellular networks are subject to extensive regulation, which modifies the availability and efficiency of connections between components in response to external conditions. Thus far, studies of large-scale networks have focused on their connectivity, but have not considered how the modulation of this connectivity might also determine network properties. To address this issue, we analyzed how the coordinated expression of enzymes shapes the metabolic network of Saccharomyces cerevisiae. By integrating large-scale expression data with the structural description of the metabolic network, we systematically characterized the transcriptional regulation of metabolic pathways. The analysis revealed recurrent patterns, which may represent design principles of metabolic gene regulation. First, we find that transcription regulation biases metabolic flow toward linearity by coexpressing only distinct branches at metabolic branchpoints. Second, individual isozymes were often separately coregulated with distinct processes, providing a means of reducing crosstalk between pathways using a common reaction. Finally, transcriptional regulation defined a hierarchical organization of metabolic pathways into groups of varying expression coherence. These results emphasize the utility of incorporating regulatory information when analyzing properties of large-scale cellular networks.

Iterative signature algorithm for the analysis of large-scale gene expression data

We present an approach for the analysis of genome-wide expression data. Our method is designed to overcome the limitations of traditional techniques, when applied to large-scale data. Rather than alloting each gene to a single cluster, we assign both genes and conditions to context-dependent and potentially overlapping transcription modules. We provide a rigorous definition of a transcription module as the object to be retrieved from the expression data. An efficient algorithm, which searches for the modules encoded in the data by iteratively refining sets of genes and conditions until they match this definition, is established. Each iteration involves a linear map, induced by the normalized expression matrix, followed by the application of a threshold function. We argue that our method is in fact a generalization of singular value decomposition, which corresponds to the special case where no threshold is applied. We show analytically that for noisy expression data our approach leads to better classification due to the implementation of the threshold. This result is confirmed by numerical analyses based on in silico expression data. We discuss briefly results obtained by applying our algorithm to expression data from the yeast Saccharomyces cerevisiae.

Pre-steady-state decoding of the Bicoid morphogen gradient

Morphogen gradients are established by the localized production and subsequent diffusion of signaling molecules. It is generally assumed that cell fates are induced only after morphogen profiles have reached their steady state. Yet, patterning processes during early development occur rapidly, and tissue patterning may precede the convergence of the gradient to its steady state. Here we consider the implications of pre-steady-state decoding of the Bicoid morphogen gradient for patterning of the anterior–posterior axis of the Drosophila embryo. Quantitative analysis of the shift in the expression domains of several Bicoid targets (gap genes) upon alteration of bcd dosage, as well as a temporal analysis of a reporter for Bicoid activity, suggest that a transient decoding mechanism is employed in this setting. We show that decoding the pre-steady-state morphogen profile can reduce patterning errors caused by fluctuations in the rate of morphogen production. This can explain the surprisingly small shifts in gap and pair-rule gene expression domains observed in response to alterations in bcd dosage.

It was previously thought that cell fates were determined by morphogen gradients only after steady state was established. Here the authors show fate may precede gradient steady state.

Subdivision of naive fields of cells into separate cell populations, distinguished by the unique combinations of genes they express, constitutes a major aspect of organism development. Classically, this involves the generation of gradients of signaling molecules (morphogens), which induce distinct cell fates in a concentration-dependent manner. It has been generally assumed that morphogen gradients are interpreted only after they reach a spatially fixed, steady-state profile. Our study re-examines this assumption for the classical case of the Bicoid morphogen, a transcription factor that is distributed as a gradient in the early Drosophila embryo. We propose and present evidence for dynamic, pre-steady-state decoding of the Bicoid profile. We further show that such dynamic decoding can directly account for the surprisingly small shifts in the expression domains of target genes, observed in response to altered Bicoid dosage, without invoking additional mechanisms or contributing factors. Pre-steady-state decoding can thus provide a simple explanation for the relative robustness of this classical morphogen system, which has been a long-standing problem.

Noise propagation and signaling sensitivity in biological networks: a role for positive feedback

Interactions between genes and proteins are crucial for efficient processing of internal or external signals, but this connectivity also amplifies stochastic fluctuations by propagating noise between components. Linear (unbranched) cascades were shown to exhibit an interplay between the sensitivity to changes in input signals and the ability to buffer noise. We searched for biological circuits that can maintain signaling sensitivity while minimizing noise propagation, focusing on cases where the noise is characterized by rapid fluctuations. Negative feedback can buffer this type of noise, but this buffering comes at the expense of an even greater reduction in signaling sensitivity. By systematically analyzing three-component circuits, we identify positive feedback as a central motif allowing for the buffering of propagated noise while maintaining sensitivity to long-term changes in input signals. We show analytically that noise reduction in the presence of positive feedback results from improved averaging of rapid fluctuations over time, and discuss in detail a particular implementation in the control of nutrient homeostasis in yeast. As the design of biological networks optimizes for multiple constraints, positive feedback can be used to improve sensitivity without a compromise in the ability to buffer propagated noise.

Biological circuits need to be sensitive to changes in environmental signals but at the same time buffer rapid fluctuations (noise) that might be imposed on this input. In this paper, we analyze the interplay between sensitivity to signals and the ability to buffer noise. Previous studies reported that negative feedback attenuates noise. We show, however, that this ability comes at the expense of an even more dramatic reduction in sensitivity. In fact, when comparing systems of the same sensitivity, a system with negative feedback is more amenable to noise than a system without such feedback. We searched for small biological circuits that can buffer noise while maintaining high sensitivity, and found that positive feedback exhibits this property. This ability of positive feedback to buffer noise reflects its slowed-down dynamics. We discuss general requirements for the function of positive feedback as a noise-filtering device and describe a particular implementation that appears to function in yeast nutrient homeostasis. Our study emphasizes the need to consider multiple constraints when analyzing the design logic of biological networks.

Principles of cellular resource allocation revealed by condition-dependent proteome profiling

Growing cells coordinate protein translation with metabolic rates. Central to this coordination is ribosome production. Ribosomes drive cell growth, but translation of ribosomal proteins competes with production of non-ribosomal proteins. Theory shows that cell growth is maximized when all expressed ribosomes are constantly translating. To examine whether budding yeast function at this limit of full ribosomal usage, we profiled the proteomes of cells growing in different environments. We find that cells produce excess ribosomal proteins, amounting to a constant ≈8% of the proteome. Accordingly, ≈25% of ribosomal proteins expressed in rapidly growing cells does not contribute to translation. Further, this fraction increases as growth rate decreases and these excess ribosomal proteins are employed when translation demands unexpectedly increase. We suggest that steadily growing cells prepare for conditions that demand increased translation by producing excess ribosomes, at the expense of lower steady-state growth rate.

Comparative gene expression analysis by differential clustering approach: application to the Candida albicans transcription program

Differences in gene expression underlie many of the phenotypic variations between related organisms, yet approaches to characterize such differences on a genome-wide scale are not well developed. Here, we introduce the “differential clustering algorithm” for revealing conserved and diverged co-expression patterns. Our approach is applied at different levels of organization, ranging from pair-wise correlations within specific groups of functionally linked genes, to higher-order correlations between such groups. Using the differential clustering algorithm, we systematically compared the transcription program of the fungal pathogen Candida albicans with that of the model organism Saccharomyces cerevisiae. Many of the identified differences are related to the differential requirement for mitochondrial function in the two yeasts. Distinct regulation patterns of cell cycle genes and of amino acid metabolic genes were also revealed and, in some cases, could be linked to the differential appearance of cis-regulatory elements in the gene promoter regions. Our study provides a comprehensive framework for comparative gene expression analysis and a rich source of hypotheses for uncharacterized open reading frames and putative cis-regulatory elements in C.albicans.

Candida albicans is a fungal inhabitant of the intestinal tract of most healthy humans. It becomes a serious and often lethal pathogen in people with a weak immune system. C. albicans is a distant relative of the well-studied baker's yeast, Saccharomyces cerevisiae. It is now possible to determine the degree to which these two fungi have similar or different patterns of transcription.

Here, methods were developed that comprehensively compare the expression patterns of S. cerevisiae and C. albicans. A novel algorithm was used to determine if the expression of groups of genes in one organism are fully, partially, or not at all similar in the other organism. This algorithm was first applied to pre-defined groups of genes predicted to have similar functions and was then used to compare the global organization of the transcription programs between the two organisms.

The analysis revealed that the expression patterns reflect the different metabolic preferences of the two yeasts. The authors also found that amino acid metabolism regulation is more differentiated in C.albicans. Furthermore, the different expression patterns can be traced down to the use of different regulatory sequences. This study provides a comprehensive framework for comparative gene expression analysis, as well as a Web site with interactive analysis tools, which allow the development of hypotheses concerning uncharacterized genes and the sequences that regulate them.

A biological solution to a fundamental distributed computing problem

Computational and biological systems are often distributed so that processors (cells) jointly solve a task, without any of them receiving all inputs or observing all outputs. Maximal independent set (MIS) selection is a fundamental distributed computing procedure that seeks to elect a set of local leaders in a network. A variant of this problem is solved during the development of the fly's nervous system, when sensory organ precursor (SOP) cells are chosen. By studying SOP selection, we derived a fast algorithm for MIS selection that combines two attractive features. First, processors do not need to know their degree; second, it has an optimal message complexity while only using one-bit messages. Our findings suggest that simple and efficient algorithms can be developed on the basis of biologically derived insights.

Expression noise and acetylation profiles distinguish HDAC functions

Gene expression shows a significant variation (noise) between genetically identical cells. Noise depends on the gene expression process regulated by the chromatin environment. We screened for chromatin factors that modulate noise in S. cerevisiae and analyzed the results using a theoretical model that infers regulatory mechanisms from the noise versus mean relationship. Distinct activities of the Rpd3(L) and Set3 histone deacetylase complexes were predicted. Both HDACs repressed expression. Yet, Rpd3(L)C decreased the frequency of transcriptional bursts, while Set3C decreased the burst size, as did H2B monoubiquitination (ubH2B). We mapped the acetylation of H3 lysine 9 (H3K9ac) upon deletion of multiple subunits of Set3C and Rpd3(L)C and of ubH2B effectors. ubH2B and Set3C appear to function in the same pathway to reduce the probability that an elongating PolII produces a functional transcript (PolII processivity), while Rpd3(L)C likely represses gene expression at a step preceding elongation.