Modeling stochastic gene expression: implications for haploinsufficiency

Severe pancreas hypoplasia and multicystic renal dysplasia in two human fetuses carrying novel HNF1beta/MODY5 mutations

Heterozygous mutations in the HNF1beta/vHNF1/TCF2 gene cause maturity-onset diabetes of the young (MODY5), associated with severe renal disease and abnormal genital tract. Here, we characterize two fetuses, a 27-week male and a 31.5-week female, carrying novel mutations in exons 2 and 7 of HNF1beta, respectively. Although these mutations were predicted to have different functional consequences, both fetuses displayed highly similar phenotypes. They presented one of the most severe phenotypes described in HNF1beta carriers: bilateral enlarged polycystic kidneys, severe pancreas hypoplasia and abnormal genital tract. Consistent with this, we detected high levels of HNF1beta transcripts in 8-week human embryos in the mesonephros and metanephric kidney and in the epithelium of pancreas. Renal histology and immunohistochemistry analyses of mutant fetuses revealed cysts derived from all nephron segments with multilayered epithelia and dysplastic regions, accompanied by a marked increase in the expression of beta-catenin and E-cadherin. A significant proportion of cysts still expressed the cystic renal disease proteins, polycystin-1, polycystin-2, fibrocystin and uromodulin, implying that cyst formation may result from a deregulation of cell-cell adhesion and/or the Wnt/beta-catenin signaling pathway. Both fetuses exhibited a severe pancreatic hypoplasia with underdeveloped and disorganized acini, together with an absence of ventral pancreatic-derived tissue. beta-catenin and E-cadherin were strongly downregulated in the exocrine and endocrine compartments, and the islets lacked the transporter essential for glucose-sensing GLUT2, indicating a beta-cell maturation defect. This study provides evidence of differential gene-dosage requirements for HNF1beta in normal human kidney and pancreas differentiation and increases our understanding of the etiology of MODY5 disorder.

Defining a link with autosomal-dominant polycystic kidney disease in mice with congenitally low expression of Pkd1

Mouse models for autosomal-dominant polycystic kidney disease (ADPKD), derived from homozygous targeted disruption of Pkd1 gene, generally die in utero or perinatally because of systemic defects. We introduced a loxP site and a loxP-flanked mc1-neo cassette into introns 30 and 34, respectively, of the Pkd1 locus to generate a conditional, targeted mutation. Significantly, before excision of the floxed exons and mc1-neo from the targeted locus by Cre recombinase, mice homozygous for the targeted allele appeared normal at birth but developed polycystic kidney disease with a slower progression than that of Pkd-null mice. Further, the homozygotes continued to produce low levels of full-length Pkd1-encoded protein, suggesting that slight Pkd1 expression is sufficient for renal cyst formation in ADPKD. In this viable model, up-regulation of heparin-binding epidermal growth factor-like growth factor accompanied increased epidermal growth factor receptor signaling, which may be involved in abnormal proliferation of the cyst-lining epithelia. Increased apoptosis in cyst epithelia was only observed in the later period that correlated with the cyst regression. Abnormalities in Na(+)/K(+)-ATPase, aquaporin-2, and vasopressin V2 receptor expression were also identified. This mouse model may be suitable for further studies of progression and therapeutic interventions of ADPKD.

Spatial modeling of dimerization reaction dynamics in the plasma membrane: Monte Carlo vs. continuum differential equations

Bimolecular reactions in the plasma membrane, such as receptor dimerization, are a key signaling step for many signaling systems. For receptors to dimerize, they must first diffuse until a collision happens, upon which a dimerization reaction may occur. Therefore, study of the dynamics of cell signaling on the membrane may require the use of a spatial modeling framework. Despite the availability of spatial simulation methods, e.g., stochastic spatial Monte Carlo (MC) simulation and partial differential equation (PDE) based approaches, many biological models invoke well-mixed assumptions without completely evaluating the importance of spatial organization. Whether one is to utilize a spatial or non-spatial simulation framework is therefore an important decision. In order to evaluate the importance of spatial effects a priori, i.e., without performing simulations, we have assessed the applicability of a dimensionless number, known as second Damköhler number (Da), defined here as the ratio of time scales of collision and reaction, for 2-dimensional bimolecular reactions. Our study shows that dimerization reactions in the plasma membrane with Da approximately >0.1 (tested in the receptor density range of 10(2)-10(5)/microm(2)) require spatial modeling. We also evaluated the effective reaction rate constants of MC and simple deterministic PDEs. Our simulations show that the effective reaction rate constant decreases with time due to time dependent changes in the spatial distribution of receptors. As a result, the effective reaction rate constant of simple PDEs can differ from that of MC by up to two orders of magnitude. Furthermore, we show that the fluctuations in the number of copies of signaling proteins (noise) may also depend on the diffusion properties of the system. Finally, we used the spatial MC model to explore the effect of plasma membrane heterogeneities, such as receptor localization and reduced diffusivity, on the dimerization rate. Interestingly, our simulations show that localization of epidermal growth factor receptor (EGFR) can cause the diffusion limited dimerization rate to be up to two orders of magnitude higher at higher average receptor densities reported for cancer cells, as compared to a normal cell.

Origins of extrinsic variability in eukaryotic gene expression

Variable gene expression within a clonal population of cells has been implicated in a number of important processes including mutation and evolution, determination of cell fates and the development of genetic disease. Recent studies have demonstrated that a significant component of expression variability arises from extrinsic factors thought to influence multiple genes simultaneously, yet the biological origins of this extrinsic variability have received little attention. Here we combine computational modelling with fluorescence data generated from multiple promoter-gene inserts in Saccharomyces cerevisiae to identify two major sources of extrinsic variability. One unavoidable source arising from the coupling of gene expression with population dynamics leads to a ubiquitous lower limit for expression variability. A second source, which is modelled as originating from a common upstream transcription factor, exemplifies how regulatory networks can convert noise in upstream regulator expression into extrinsic noise at the output of a target gene. Our results highlight the importance of the interplay of gene regulatory networks with population heterogeneity for understanding the origins of cellular diversity.

Is ectopic expression caused by deregulatory mutations or due to gene-regulation leaks with evolutionary potential?

It has long been thought that gene expression is tightly regulated in multicellular eukaryotes, so that expression profiles match functional profiles. This conception emerged from the assumption that gene activity is synonymous with gene function. This paradigm was first challenged by comparative protein electrophoresis studies showing extensive differences in expression patterns among related species. The paradigm is now being challenged by evolutionary transcriptomics using microarray technologies. Most gene expression profiles display features that lack any obvious functional significance. The so-called "ectopic" expression refers to the expression of genes at times and locations where the target gene is not known to have a function. However, ectopic expression might be associated with genuine function even if this function is not essential or has yet to be ascertained. Alternatively, ectopic expression might come about as a superfluous by-product of regulatory systems, which would call for a revision of prevailing ideas about the specificity of gene regulation. We herein review available evidence for ectopic expression and the hypotheses proposed for its origin and evolution. We propose that ectopic expression must be regarded as part of an integrated phenotypic whole. It seems likely that ectopic expression represents a leak in the evolution of regulatory systems, but one that is endowed with considerable evolutionary possibilities.

Modeling T cell antigen discrimination based on feedback control of digital ERK responses

T-lymphocyte activation displays a remarkable combination of speed, sensitivity, and discrimination in response to peptide-major histocompatibility complex (pMHC) ligand engagement of clonally distributed antigen receptors (T cell receptors or TCRs). Even a few foreign pMHCs on the surface of an antigen-presenting cell trigger effective signaling within seconds, whereas 1 x 10(5)-1 x 10(6) self-pMHC ligands that may differ from the foreign stimulus by only a single amino acid fail to elicit this response. No existing model accounts for this nearly absolute distinction between closely related TCR ligands while also preserving the other canonical features of T-cell responses. Here we document the unexpected highly amplified and digital nature of extracellular signal-regulated kinase (ERK) activation in T cells. Based on this observation and evidence that competing positive- and negative-feedback loops contribute to TCR ligand discrimination, we constructed a new mathematical model of proximal TCR-dependent signaling. The model made clear that competition between a digital positive feedback based on ERK activity and an analog negative feedback involving SH2 domain-containing tyrosine phosphatase (SHP-1) was critical for defining a sharp ligand-discrimination threshold while preserving a rapid and sensitive response. Several nontrivial predictions of this model, including the notion that this threshold is highly sensitive to small changes in SHP-1 expression levels during cellular differentiation, were confirmed by experiment. These results combining computation and experiment reveal that ligand discrimination by T cells is controlled by the dynamics of competing feedback loops that regulate a high-gain digital amplifier, which is itself modulated during differentiation by alterations in the intracellular concentrations of key enzymes. The organization of the signaling network that we model here may be a prototypic solution to the problem of achieving ligand selectivity, low noise, and high sensitivity in biological responses.

Graded and binary responses in stochastic gene expression

Recently, several theoretical and experimental studies have been undertaken to probe the effect of stochasticity on gene expression (GE). In experiments, the GE response to an inducing signal in a cell, measured by the amount of mRNAs/proteins synthesized, is found to be either graded or binary. The latter type of response gives rise to a bimodal distribution in protein levels in an ensemble of cells. One possible origin of binary response is cellular bistability achieved through positive feedback or autoregulation. In this paper, we study a simple, stochastic model of GE and show that the origin of binary response lies exclusively in stochasticity. The transitions between the active and inactive states of the gene are random in nature. Graded and binary responses occur in the model depending on the relative stability of the activated and deactivated gene states with respect to that of mRNAs/proteins. The theoretical results on binary response provide a good description of the 'all-or-none' phenomenon observed in an eukaryotic system.

Ultrasensitization: switch-like regulation of cellular signaling by transcriptional induction

Cellular signaling networks are subject to transcriptional and proteolytic regulation under both physiological and pathological conditions. For example, the expression of proteins subject to covalent modification by phosphorylation is known to be altered upon cellular differentiation or during carcinogenesis. However, it is unclear how moderate alterations in protein expression can bring about large changes in signal transmission as, for example, observed in the case of haploinsufficiency, where halving the expression of signaling proteins abrogates cellular function. By modeling a fundamental motif of signal transduction, the phosphorylation–dephosphorylation cycle, we show that minor alterations in the concentration of the protein subject to phosphorylation (or the phosphatase) can affect signal transmission in a highly ultrasensitive fashion. This “ultrasensitization” is strongly favored by substrate sequestration on the catalyzing enzymes, and can be observed with experimentally measured enzymatic rate constants. Furthermore, we show that coordinated transcription of multiple proteins (i.e., synexpression) within a protein kinase cascade results in even more pronounced all-or-none behavior with respect to signal transmission. Finally, we demonstrate that ultrasensitization can account for specificity and modularity in the regulation of cellular signal transduction. Ultrasensitization can result in all-or-none cell-fate decisions and in highly specific cellular regulation. Additionally, switch-like phenomena such as ultrasensitization are known to contribute to bistability, oscillations, noise reduction, and cellular heterogeneity.

Hormones and other external stimuli induce cellular transitions such as cell division or differentiation by regulating gene expression. Hormone-induced cellular transitions are known to occur in a switch-like fashion: while weak background stimuli are rejected, cellular transitions proceed fully as soon as a threshold hormone concentration is exceeded. Earlier studies have described several mechanisms whereby such a switch-like behavior can be realized in intracellular communication via signal transduction networks, which convert hormonal signals into alterations in gene expression.

The authors demonstrate how switch-like behavior can be further enhanced downstream of hormone-induced gene expression. They show that even minor (hormone-induced) alterations in gene expression can dramatically affect the activity of intracellular signal transduction networks, and thereby modify cellular behavior. This phenomenon has been termed “ultrasensitization.”

Ultrasensitization can explain the pronounced dosage sensitivity observed for many disease-associated signal transduction proteins: for example, the mutation of one of two alleles (gene copies), resulting in a 2-fold reduction of gene expression, can already initiate disease progression. Although such sensitivity towards mutations is potentially harmful, the fact that cells nevertheless exhibit ultrasensitization suggests that somehow cells benefit from ultrasensitization. The authors illustrate how ultrasensitization improves the specificity and efficiency of cell-to-cell communication and contributes to cellular memory.

Tumor development: haploinsufficiency and local network assembly

According to the current models, tumor development is a continuous process of mutation accumulation, leading to several intermediate phenotypes and final phases of autonomy, unlimited growth and metastasis. One of the most important events in that process is the initial destabilization of cellular pathways that subsequently allow mutations to accumulate. The mechanisms involved in that stage are not clear. In principle, the estimated very low mutation frequency in human or mouse cells would suggest that accumulating the required number of mutations for tumor development should be a statistically unlikely event. However, this theory is contradicted by the high incidence of cancers. Here we discuss the role of protein haploinsufficiency as a contributor to the initial phases of tumor development, and suggest possible mechanisms that might be involved in that process.

Identification of a candidate tumor suppressor gene RHOBTB1 located at a novel allelic loss region 10q21 in head and neck cancer

PURPOSE

Aims of the study are to narrow-down the hotspot region on 10q21 defined by previous genome-wide loss of heterozygosity (LOH) analysis in head and neck squamous cell carcinomas (HNSCC) and to define candidate tumor suppressor genes (TSG) concerned with 10q21.

MATERIALS AND METHODS

LOH analysis was carried out with ten polymorphic microsatellite markers. Expression analysis was performed by semi-quantitative RT-PCR, and mutation analysis by PCR and direct sequencing.

RESULTS

LOH analysis on 10q21 in 52 HNSCC indicated distinctive and frequent allelic loss at D10S589 (42%). Among flanking genes, we found the RHOBTB1 gene as a candidate TSG, since an intragenic marker demonstrated the highest LOH (44%). Expression analysis revealed down-regulation of RHOBTB1 mRNA in 37% of tumors. Interestingly, all the five tumors that showed decreased expression of RHOBTB1 were accompanied with LOH, supporting the haploinsufficiency and class 2 TSG characteristics of RHOBTB1. No pathogenic mutation of RHOBTB1 was found. Furthermore, another gene within the region, EGR2, was also taken under scope. LOH frequencies around the EGR2 gene were relatively low (23 and 33%). Albeit semi-quantitative expression analysis of EGR2 demonstrated downregulation in 45% of tumor samples, no relation was found between the expression levels and LOH status.

CONCLUSION

Frequent allelic loss and decreased expression of RHOBTB1 suggested that this gene has a role in tumorigenesis of a subset of HNSCC.

Genomic analysis of single cytokeratin-positive cells from bone marrow reveals early mutational events in breast cancer

Chromosomal instability in human breast cancer is known to take place before mammary neoplasias display morphological signs of invasion. We describe here the unexpected finding of a tumor cell population with normal karyotypes isolated from bone marrow of breast cancer patients. By analyzing the same single cells for chromosomal aberrations, subchromosomal allelic losses, and gene amplifications, we confirmed their malignant origin and delineated the sequence of genomic events during breast cancer progression. On this trajectory of genomic progression, we identified a subpopulation of patients with very early HER2 amplification. Because early changes have the highest probability of being shared by genetically unstable tumor cells, the genetic characterization of disseminated tumor cells provides a novel rationale for selecting patients for targeted therapies.

Stochasticity in gene expression: from theories to phenotypes

Genetically identical cells exposed to the same environmental conditions can show significant variation in molecular content and marked differences in phenotypic characteristics. This variability is linked to stochasticity in gene expression, which is generally viewed as having detrimental effects on cellular function with potential implications for disease. However, stochasticity in gene expression can also be advantageous. It can provide the flexibility needed by cells to adapt to fluctuating environments or respond to sudden stresses, and a mechanism by which population heterogeneity can be established during cellular differentiation and development.

Developmental instability of theDrosophilawing as an index of genomic perturbation and altered cell proliferation

We experimentally induced different levels of instability affecting the development of specific wing regions of Drosophila melanogaster using the UAS-GAL4 system. A common index of developmental instability is fluctuating asymmetry (FA), that is, random differences between body sides of single individuals. We studied the FA in transgenic strains carrying random genomic insertions (UAS strains), as well as insertions in the regulatory region of genes involved in the organization of wing development (GAL4 strains). In addition, the expression of genes that increase (dp110 and 3622) or decrease (dPTEN) cell proliferation was ectopically induced. Our results are related to different levels of perturbation. Through the first kind of perturbation, genome integrity was compromised by the insertion of foreign DNA. In all cases, we observed a general increase in FA, although it was rarely found significant. The second kind of perturbation involved a modification of genes controlling wing development through the insertion of a GAL4 sequence in their promoter region. The third kind involved the ectopic expression of genes controlling cell proliferation. Our results show that (i) the level of FA is connected with the level of morphological perturbation induced, (ii) FA increase was higher in the wing regions that were the target of the genetic perturbation, and (iii) developmental instability was also observed in regions that were not directly addressed by the perturbation. The results were discussed on the basis of the running models about Drosophila wing development.

Noise characteristics of feed forward loops

A prominent feature of gene transcription regulatory networks is the presence in large numbers of motifs, i.e., patterns of interconnection, in the networks. One such motif is the feed forward loop (FFL) consisting of three genes X, Y and Z. The protein product x of X controls the synthesis of protein product y of Y. Proteins x and y jointly regulate the synthesis of z proteins from the gene Z. The FFLs, depending on the nature of the regulating interactions, can be of eight different types which can again be classified into two categories: coherent and incoherent. In this paper, we study the noise characteristics of FFLs using the Langevin formalism and the Monte Carlo simulation technique based on the Gillespie algorithm. We calculate the variances around the mean protein levels in the steady states of the FFLs and find that, in the case of coherent FFLs, the most abundant FFL, namely, the type-1 coherent FFL, is the least noisy. This is shown to be true for all parameter values when the FFLs operate above their thresholds of activation/repression. In the case of incoherent FFLs, no such general conclusion can be shown. The results suggest possible relationships between noise, functionality and abundance.

Energy constraints on the evolution of gene expression

I here estimate the energy cost of changes in gene expression for several thousand genes in the yeast Saccharomyces cerevisiae. A doubling of gene expression, as it occurs in a gene duplication event, is significantly selected against for all genes for which expression data is available. It carries a median selective disadvantage of s > 10(-5), several times greater than the selection coefficient s = 1.47 x 10(-7) below which genetic drift dominates a mutant's fate. When considered separately, increases in messenger RNA expression or protein expression by more than a factor 2 also have significant energy costs for most genes. This means that the evolution of transcription and translation rates is not an evolutionarily neutral process. They are under active selection opposing them. My estimates are based on genome-scale information of gene expression in the yeast S. cerevisiae as well as information on the energy cost of biosynthesizing amino acids and nucleotides.

A stochastic version of corticosteriod pharmacogenomic model

The purpose of this study was to develop a stochastic version of corticosteriod fifth generation pharmacogenomic model. The Gillespie algorithm was used to generate the independent time courses of the receptor messenger RNA (mRNA). Initial parameters for the stochastic simulation were adapted from the study by Jin et al. The result obtained from the proposed stochastic model showed an overall agreement with the deterministic fifth generation model. This study suggested that because the stochastic model takes into account the "noise" nature of gene regulation, it would have potential application in pharmacogenomic modeling.

Stochasticity or the fatal ‘imperfection’ of cloning

The concept of clone is analysed with the aim of exploring the limits to which a phenotype can be said to be determined geneticaly. First of all, mutations that result from the replication, topological manipulation or lesion of DNA introduce a source of heritable variation in an otherwise identical genetic background. But more important, stochastic effects in many biological processes may superimpose a phenotypic variation which is not encoded in the genome. The source of stochasticity ranges from the random selection of alleles or whole chromosomes to be expressed in small cell populations, to fluctuations in processes such as gene expression, due to limiting amounts of the players involved. The picture emerging is that the term clone is a statistical over-simplification representing a series of individuals having essentially the same genome but capable of exhibiting wide phenotypic variation. Finally, to what extent fluctuations in biological processes, usually thought of as noise, are in fact signal is also discussed.

The selective values of alleles in a molecular network model are context dependent

Classical quantitative genetics has applied linear modeling to the problem of mapping genotypic to phenotypic variation. Much of this theory was developed prior to the availability of molecular biology. The current understanding of the mechanisms of gene expression indicates the importance of nonlinear effects resulting from gene interactions. We provide a bridge between genetics and gene network theories by relating key concepts from quantitative genetics to the parameters, variables, and performance functions of genetic networks. We illustrate this methodology by simulating the genetic switch controlling galactose metabolism in yeast and its response to selection for a population of individuals. Results indicate that genes have heterogeneous contributions to phenotypes and that additive and nonadditive effects are context dependent. Early cycles of selection suggest strong additive effects attributed to some genes. Later cycles suggest the presence of strong context-dependent nonadditive effects that are conditional on the outcomes of earlier selection cycles. A single favorable allele cannot be consistently identified for most loci. These results highlight the complications that can arise with the presence of nonlinear effects associated with genes acting in networks when selection is conducted on a population of individuals segregating for the genes contributing to the network.

Self-organization vs Watchmaker: stochastic gene expression and cell differentiation

Cell differentiation and organism development are traditionally described in deterministic terms of program and design, echoing a conventional clockwork perception of the cell on another scale. However, the current experimental reality of stochastic gene expression and cell plasticity is poorly consistent with the ideas of design, purpose and determinism, suggesting that the habit of classico-mechanistic interpretation of life phenomena may handicap our ability to adequately comprehend and model biological systems. An alternative conceptualization of cell differentiation and development is proposed where the developing organism is viewed as a dynamic self-organizing system of adaptive interacting agents. This alternative interpretation appears to be more consistent with the probabilistic nature of gene expression and the phenomena of cell plasticity, and is coterminous with the novel emerging image of the cell as a self-organizing molecular system. I suggest that stochasticity, as a principle of differentiation and adaptation, and self-organization, as a concept of emergence, have the potential to provide an interpretational framework that unites phenomena across different scales of biological organization, from molecules to societies.

Self-consistent proteomic field theory of stochastic gene switches

We present a self-consistent field approximation approach to the problem of the genetic switch composed of two mutually repressing/activating genes. The protein and DNA state dynamics are treated stochastically and on an equal footing. In this approach the mean influence of the proteomic cloud created by one gene on the action of another is self-consistently computed. Within this approximation a broad range of stochastic genetic switches may be solved exactly in terms of finding the probability distribution and its moments. A much larger class of problems, such as genetic networks and cascades, also remain exactly solvable with this approximation. We discuss, in depth, certain specific types of basic switches used by biological systems and compare their behavior to the expectation for a deterministic switch.

Lowering of Pkd1 expression is sufficient to cause polycystic kidney disease

Autosomal dominant polycystic kidney disease (ADPKD) is a major cause of renal failure and is characterized by the formation of many fluid-filled cysts in the kidneys. It is a systemic disorder that is caused by mutations in PKD1 or PKD2. Homozygous inactivation of these genes at the cellular level, by a 'two-hit' mechanism, has been implicated in cyst formation but does not seem to be the sole mechanism for cystogenesis. We have generated a novel mouse model with a hypomorphic Pkd1 allele, Pkd1(nl), harbouring an intronic neomycin-selectable marker. This selection cassette causes aberrant splicing of intron 1, yielding only 13-20% normally spliced Pkd1 transcripts in the majority of homozygous Pkd1(nl) mice. Homozygous Pkd1(nl) mice are viable, showing bilaterally enlarged polycystic kidneys. This is in contrast to homozygous knock-out mice, which are embryonic lethal, and heterozygous knock-out mice that show only a very mild cystic phenotype. In addition, homozygous Pkd1(nl) mice showed dilatations of pancreatic and liver bile ducts, and the mice had cardiovascular abnormalities, pathogenic features similar to the human ADPKD phenotype. Removal of the neomycin selection-cassette restored the phenotype of wild-type mice. These results show that a reduced dosage of Pkd1 is sufficient to initiate cystogenesis and vascular defects and indicate that low Pkd1 gene expression levels can overcome the embryonic lethality seen in Pkd1 knock-out mice. We propose that in patients reduced PKD1 expression of the normal allele below a critical level, due to genetic, environmental or stochastic factors, may lead to cyst formation in the kidneys and other clinical features of ADPKD.

FLI1 monoallelic expression combined with its hemizygous loss underlies Paris-Trousseau/Jacobsen thrombopenia

Paris-Trousseau syndrome (PTS; also known as Jacobsen syndrome) is characterized by several congenital anomalies including a dysmegakaryopoiesis with two morphologically distinct populations of megakaryocytes (MKs). PTS patients harbor deletions on the long arm of chromosome 11, including the FLI1 gene, which encodes a transcription factor essential for megakaryopoiesis. We show here that lentivirus-mediated overexpression of FLI1 in patient CD34(+) cells restores the megakaryopoiesis in vitro, indicating that FLI1 hemizygous deletion contributes to the PTS hematopoietic defects. FISH analysis on pre-mRNA and single-cell RT-PCR revealed that FLI1 expression is mainly monoallelic in CD41(+)CD42(-) progenitors, while it is predominantly biallelic in the other stages of megakaryopoiesis. In PTS cells, the hemizygous deletion of FLI1 generates a subpopulation of CD41(+)CD42(-) cells completely lacking FLI1 transcription. We propose that the absence of FLI1 expression in these CD41(+)CD42(-) cells might prevent their differentiation, which could explain the segregation of the PTS MKs into two subpopulations: one normal and one composed of small immature MKs undergoing a massive lysis, presumably originating from either FLI1(+) or FLI1(-) CD41(+)CD42(-) cells, respectively. Thus, we point to the role of transient monoallelic expression of a gene essential for differentiation in the genesis of human haploinsufficiency-associated disease and suggest that such a mechanism may be involved in the pathogenesis of other congenital or acquired genetic diseases.

Haploinsufficiency for tumour suppressor genes: when you don't need to go all the way

Classical tumour suppressor genes are thought to require mutation or loss of both alleles to facilitate tumour progression. However, it has become clear over the last few years that for some genes, haploinsufficiency, which is loss of only one allele, may contribute to carcinogenesis. These effects can either be directly attributable to the reduction in gene dosage or may act in concert with other oncogenic or haploinsufficient events. Here we describe the genes that undergo this phenomenon and discuss possible mechanisms that allow haploinsufficiency to display a phenotype and facilitate the pathogenesis of cancer.

Intrinsic and external noise in an auto-regulatory genetic network

A single gene auto-regulatory network is analysed. The main goal is to investigate the effects of the negative and positive feedbacks on the intrinsic and external noises. The central finding of this paper is that: for the intrinsic noise, both the negative and positive feedback regulations increase the fluctuation strength of mRNA levels (where the fluctuation strength is measured by the Fano factor for both the fluctuations of mRNAs and proteins), and the negative feedback decreases, but the positive feedback increases, the fluctuation strength of proteins; for the external noise, the negative feedback not only increase the fluctuation strength of mRNA levels but also the fluctuation strength of proteins, and though the effect of the positive feedback on the fluctuation strength of mRNA levels depends on the size of positive feedback parameter k, the positive feedback must decrease the fluctuation strength of proteins.

Myristoylation of the fus1 protein is required for tumor suppression in human lung cancer cells

FUS1 is a novel tumor suppressor gene identified in the human chromosome 3p21.3 region that is deleted in many cancers. Using surface-enhanced laser desorption/ionization mass spectrometric analysis on an anti-Fus1-antibody-capture ProteinChip array, we identified wild-type Fus1 as an N-myristoylated protein. N-myristoylation is a protein modification process in which a 14-carbon myristoyl group is cotranslationally and covalently added to the NH2-terminal glycine residue of the nascent polypeptide. Loss of expression or a defect of myristoylation of the Fus1 protein was observed in human primary lung cancer and cancer cell lines. A myristoylation-deficient mutant of the Fus1 protein abrogated its ability to inhibit tumor cell-induced clonogenicity in vitro, to induce apoptosis in lung tumor cells, and to suppress the growth of tumor xenografts and lung metastases in vivo and rendered it susceptible to rapid proteasome-dependent degradation. Our results show that myristoylation is required for Fus1-mediated tumor-suppressing activity and suggest a novel mechanism for the inactivation of tumor suppressors in lung cancer and a role for deficient posttranslational modification in tumor suppressor-gene-mediated carcinogenesis.

The Selective Values of Alleles in a Molecular Network Model Are Context Dependent

Classical quantitative genetics has applied linear modeling to the problem of mapping genotypic to phenotypic variation. Much of this theory was developed prior to the availability of molecular biology. The current understanding of the mechanisms of gene expression indicates the importance of nonlinear effects resulting from gene interactions. We provide a bridge between genetics and gene network theories by relating key concepts from quantitative genetics to the parameters, variables, and performance functions of genetic networks. We illustrate this methodology by simulating the genetic switch controlling galactose metabolism in yeast and its response to selection for a population of individuals. Results indicate that genes have heterogeneous contributions to phenotypes and that additive and nonadditive effects are context dependent. Early cycles of selection suggest strong additive effects attributed to some genes. Later cycles suggest the presence of strong context-dependent nonadditive effects that are conditional on the outcomes of earlier selection cycles. A single favorable allele cannot be consistently identified for most loci. These results highlight the complications that can arise with the presence of nonlinear effects associated with genes acting in networks when selection is conducted on a population of individuals segregating for the genes contributing to the network.

Transcriptional networks controlling pancreatic development and beta cell function

Transcription factors provide the genetic instructions that drive pancreatic development and enable mature beta cells to function properly. To understand fully how this is accomplished, it is necessary to unravel the regulatory networks formed by transcription factors acting on their genomic targets. This article discusses recent advances in our understanding of how transcriptional networks control early pancreas organogenesis, embryonic endocrine cell formation and the differentiated function of adult beta cells. We discuss how mutations in several transcription factor genes involved in such networks cause Maturity onset diabetes of the young (MODY). Finally, we propose that pancreatic gene programs might be manipulated to generate beta cells or to enhance the function of existing beta cells, thereby providing a possible treatment of different forms of diabetes.

Fluctuations in transcription factor binding can explain the graded and binary responses observed in inducible gene expression

Inducible genes are expressed in the presence of an external stimulus. Individual cells may exhibit either a binary or graded response to such signals. It has been hypothesized that the chemical kinetics of transcription factor/DNA interactions can account for both these scenarios (EMBO J. 9(9) (1990) 2835; BioEssays 14(5) (1992) 341). To explore this question, we have conducted work based on the experimental results of Fiering et al. (Genes Dev. 4 (10) (1990) 1823). In these experiments, three upstream NF-AT binding sites control transcription of the lacZ gene, which codes for the enzyme beta-Galactosidase. The experimental data show a binary response for this system. We consider the effects of fluctuations in NF-AT binding on the response of the system. Our modeling results are in good qualitative agreement with the experimental data, and illustrate how the binary and graded responses can stem from the same underlying mechanism.

Silencing the Drosophila ribosomal protein L14 gene using targeted RNA interference causes distinct somatic anomalies

The Drosophila Minutes are haploinsufficient mutations that are defective in ribosomal protein (rp) production, resulting in short, thin bristles, delayed development and recessive lethality. In a Minute fly, the amount of rp gene messenger RNA (mRNA) is reduced to >or=50% of the normal amount of gene product, and becomes rate limiting for ribosome biogenesis, cell proliferation and growth. Haploinsufficiency increases the vulnerability to complete loss of gene function (homozygous null state) if hit by a second mutation. Because of the homozygous lethality, it has only been possible to study the effects of Minute mutations in heterozygous animals. To be able to study the consequences of a loss-of-function of an rp gene (0%>mRNA<50%) in developing and differentiated cells we used heritable RNA interference (RNAi) in combination with the yeast GAL4/UAS binary system to spatiotemporally knock down the ribosomal protein L14 (RpL14) gene. We show, at the RNA and phenotypic levels, that RNAi efficiently reduces RpL14 gene expression throughout development, causing lethality and distinct and dramatic somatic anomalies in both developing and differentiated cells.

A Theoretical Model for the Regulation of Sex-lethal, a Gene That Controls Sex Determination and Dosage Compensation in Drosophila melanogaster

Cell fate commitment relies upon making a choice between different developmental pathways and subsequently remembering that choice. Experimental studies have thoroughly investigated this central theme in biology for sex determination. In the somatic cells of Drosophila melanogaster, Sex-lethal (Sxl) is the master regulatory gene that specifies sexual identity. We have developed a theoretical model for the initial sex-specific regulation of Sxl expression. The model is based on the well-documented molecular details of the system and uses a stochastic formulation of transcription. Numerical simulations allow quantitative assessment of the role of different regulatory mechanisms in achieving a robust switch. We establish on a formal basis that the autoregulatory loop involved in the alternative splicing of Sxl primary transcripts generates an all-or-none bistable behavior and constitutes an efficient stabilization and memorization device. The model indicates that production of a small amount of early Sxl proteins leaves the autoregulatory loop in its off state. Numerical simulations of mutant genotypes enable us to reproduce and explain the phenotypic effects of perturbations induced in the dosage of genes whose products participate in the early Sxl promoter activation.

The Engineering of Gene Regulatory Networks

The rapid accumulation of genetic information and advancement of experimental techniques have opened a new frontier in biomedical engineering. With the availability of well-characterized components from natural gene networks, the stage has been set for the engineering of artificial gene regulatory networks with sophisticated computational and functional capabilities. In these efforts, the ability to construct, analyze, and interpret qualitative and quantitative models is becoming increasingly important. In this review, we consider the current state of gene network engineering from a combined experimental and modeling perspective. We discuss how networks with increased complexity are being constructed from simple modular components and how quantitative deterministic and stochastic modeling of these modules may provide the foundation for accurate in silico representations of gene regulatory network function in vivo.

Stochastic gene expression as a many-body problem

Gene expression has a stochastic component because of the single-molecule nature of the gene and the small number of copies of individual DNA-binding proteins in the cell. We show how the statistics of such systems can be mapped onto quantum many-body problems. The dynamics of a single gene switch resembles the spin-boson model of a two-site polaron or an electron transfer reaction. Networks of switches can be approximately described as quantum spin systems by using an appropriate variational principle. In this way, the concept of frustration for magnetic systems can be taken over into gene networks. The landscape of stable attractors depends on the degree and style of frustration, much as for neural networks. We show the number of attractors, which may represent cell types, is much smaller for appropriately designed weakly frustrated stochastic networks than for randomly connected networks.

Haploinsufficiency at the alpha-synuclein gene underlies phenotypic severity in familial Parkinson's disease

To date, two point mutations, G209A and G88C, have been reported in the coding region of the alpha-synuclein gene in autosomal dominant familial Parkinson's disease. When translated, these lead to the missense mutations Ala53Thr and Ala30Pro, respectively. Reduced mRNA expression of the G209A allele was reported recently in a Greek-American family. Here, we show that alpha-synuclein mRNA is normally expressed in blood cells and report the results of an analysis of alpha-synuclein mRNA and protein expression in lymphoblastoid cell lines established from kindreds with the G209A and G88C mutations. mRNA expression was characterized using a TaqMan real-time quantitative reverse transcriptase-polymerase chain reaction (RT-PCR) assay. We assessed five affected and three unaffected members of a German family with the G88C mutation and two affected members in different, unrelated Greek families with the G209A mutation. The ratio of wild-type to mutant alpha-synuclein allele expression ranged from 2.2 to 9.2 in the affected individuals with a severe clinical phenotype. The ratios of the expression levels of the wild-type to mutant alleles were only slightly decreased in mild cases and were less than 1.0 in two asymptomatic heterozygotes. Sequence analysis of the RT-PCR products showed only the presence of G in position 88 and G in position 209 in severely affected heterozygotes of the German and Greek families, respectively. High performance liquid chromatography/mass spectrometry demonstrated that, relative to wild-type alpha-synuclein, there is a reduction of Ala30Pro alpha-synuclein in lymphoblastoid cell lines originating from severely affected, but not mildly affected G88C/+ heterozygotes. Taken together, these data indicate that there is haploinsufficiency at the alpha-synuclein gene and that the ratio of expression of the wild-type to mutant alleles correlates with the severity of the clinical phenotype. Furthermore, these findings suggest that haploinsufficiency of alpha-synuclein mutations may contribute to disease progression in these forms of familial Parkinson's disease.

Selective instability: maternal effort and the evolution of gene activation and deactivation rates

We previously used simulations of gene expression to demonstrate that rapid activation and deactivation rates stabilized outcomes in stochastic systems. We hypothesized that transient single allele inactivation of an autosomal gene during gametogenesis or very early embryogenesis could have a selective advantage if it permits the functional sampling of each allele and precludes committing maternal effort to an embryo with a deleterious mutation. To test this hypothesis, we simulated the evolution of gene expression activation and deactivation rates and imposed two different selective pressures on the populations: (a). late selection against individuals that cannot maintain a threshold level of gene product that occurs after the investment of maternal effort (i.e., after birth); or (b). early selection: in addition to late selection, maintenance of the gene product above a threshold level was necessary for early development prior to commitment of maternal effort. We found that the opportunity to save reproductive effort from early selection caused the evolution of higher deactivation rates and lower activation rates than in the late selection condition. Thus, we predict that in the special case where early selection can save maternal investment in non-viable offspring, gene expression activation rates and deactivation rates might be selected to permit sampling of the product from each allele.

Near-critical phenomena in intracellular metabolite pools

The supply and consumption of metabolites in living cells are catalyzed by enzymes. Here we consider two of the simplest schemes where one substrate is eliminated through Michaelis-Menten kinetics, and where two types of substrates are joined together by an enzyme. It is demonstrated how steady-state substrate concentrations can change ultrasensitively in response to changes in their supply rates and how this is coupled to slow relaxation back to steady state after a perturbation. In the one-substrate system, such near-critical behavior occurs when the supply rate approaches the maximal elimination rate, and in the two-substrate system it occurs when the rates of substrate supply are almost balanced. As systems that operate near criticality tend to display large random fluctuations, we also carried out a stochastic analysis using analytical approximations of master equations and compared the results with molecular-level Monte Carlo simulations. It was found that the significance of random fluctuations was directly coupled to the steady-state sensitivity and that the two substrates can fluctuate greatly because they are anticorrelated in such a way that the product formation rate displays only small variation. Basic relations are highlighted and biological implications are discussed.

Control, exploitation and tolerance of intracellular noise

Noise has many roles in biological function, including generation of errors in DNA replication leading to mutation and evolution, noise-driven divergence of cell fates, noise-induced amplification of signals, and maintenance of the quantitative individuality of cells. Yet there is order to the behaviour and development of cells. They operate within strict parameters and in many cases this behaviour seems robust, implying that noise is largely filtered by the system. How can we explain the use, rejection and sensitivity to noise that is found in biological systems? An exploration of the sources and consequences of noise calls for the use of stochastic models.

Increased noise as an effect of haploinsufficiency of the tumor-suppressor gene neurofibromatosis type 1 in vitro

In human diseases related to tumor-suppressor genes, it is suggested that only the complete loss of the protein results in specific symptoms such as tumor formation, whereas simple reduction of protein quantity to 50%, called haploinsufficiency, essentially does not affect cellular behavior. Using a model of gene expression, it was presumed that haploinsufficiency is related to an increased noise in gene expression also in vivo [Cook, D. L., Gerber, A. N. & Tapscott, S. J. (1998) Proc. Natl. Acad. Sci. USA 95, 15641-15646]. Here, we demonstrate that haploinsufficiency of the tumor-suppressor gene neurofibromatosis type 1 (NF1) results in an increased variation of dendrite formation in cultured NF1 melanocytes. These morphological differences between NF1 and control melanocytes can be described by a mathematical model in which the cell is considered to be a self-organized automaton. The model describes the adjustment of the cells to a set point and includes a noise term that allows for stochastic processes. It describes the experimental data of control and NF1 melanocytes. In the cells haploinsufficient for NF1 we found an altered signal-to-noise ratio detectable as increased variation in dendrite formation in two of three investigated morphological parameters. We also suggest that in vivo NF1 haploinsufficiency results in an increased noise in cellular regulation and that this effect of haploinsufficiency may be found also in other tumor suppressors.

APC-dependent suppression of colon carcinogenesis by PPARgamma

Activation of PPARgamma by synthetic ligands, such as thiazolidinediones, stimulates adipogenesis and improves insulin sensitivity. Although thiazolidinediones represent a major therapy for type 2 diabetes, conflicting studies showing that these agents can increase or decrease colonic tumors in mice have raised concerns about the role of PPARgamma in colon cancer. To analyze critically the role of this receptor, we have used mice heterozygous for Ppargamma with both chemical and genetic models of this malignancy. Heterozygous loss of PPARgamma causes an increase in beta-catenin levels and a greater incidence of colon cancer when animals are treated with azoxymethane. However, mice with preexisting damage to Apc, a regulator of beta-catenin, develop tumors in a manner insensitive to the status of PPARgamma. These data show that PPARgamma can suppress beta-catenin levels and colon carcinogenesis but only before damage to the APC/beta-catenin pathway. This finding suggests a potentially important use for PPARgamma ligands as chemopreventative agents in colon cancer.

A genetic switch in pancreatic beta-cells: implications for differentiation and haploinsufficiency

Heterozygous mutations in the genes encoding transcriptional regulators hepatocyte nuclear factor (HNF)-1alpha and HNF-4alpha cause a form of diabetes known as maturity-onset diabetes of the young (MODY). Haploinsufficiency of HNF-1alpha or HNF-4alpha results in MODY because of defective function of pancreatic islet cells. In contrast, homozygous null mutations in mouse models lead to widespread and profound gene expression defects in multiple cell types. Thus, it is not surprising that HNF-1alpha function is now known to have distinct properties in pancreatic beta-cells. It controls a complex tissue-selective genetic network that is activated when pancreatic cells differentiate, and allows these cells to maintain critical specialized functions. The network contains an indispensable core component formed by a positive cross-regulatory feedback circuit between HNF-1alpha and HNF-4alpha. This type of circuit configuration can exhibit a switch-like behavior with two stable states. In the default active state, it can serve to perpetuate network activity in differentiated beta-cells. However, the loss of one HNF-1alpha or HNF-4alpha allele can increase the probability that the feedback circuit is permanently switched off, resulting in decreased expression of all four alleles selectively in beta-cells. Such a model can serve to rationalize key aspects of the pathogenic mechanism in MODY.

Mutually regulated expression of Pax6 and Six3 and its implications for the Pax6 haploinsufficient lens phenotype

Pax6 is a key regulator of eye development in vertebrates and invertebrates, and heterozygous loss-of-function mutations of the mouse Pax6 gene result in the Small eye phenotype, in which a small lens is a constant feature. To provide an understanding of the mechanisms underlying this haploinsufficient phenotype, we evaluated in Pax6 heterozygous mice the effects of reduced Pax6 gene dosage on the activity of other transcription factors regulating eye formation. We found that Six3 expression was specifically reduced in lenses of Pax6 heterozygous mouse embryos. Interactions between orthologous genes from the Pax and Six families have been identified in Drosophila and vertebrate species, and we examined the control of Pax6 and Six3 gene expression in the developing mouse lens. Using in vitro and transgenic approaches, we found that either transcription factor binds regulatory sequences from the counterpart gene and that both genes mutually activate their expression. These studies define a functional relationship in the lens in which Six3 expression is dosage-dependent on Pax6 and where, conversely, Six3 activates Pax6. Accordingly, we show a rescue of the Pax6 haploinsufficient lens phenotype after lens-specific expression of Six3 in transgenic mice. This phenotypic rescue was accompanied by cell proliferation and activation of the platelet-derived growth factor alpha-R/cyclin D1 signaling pathway. Our findings thus provide a mechanism implicating gene regulatory interactions between Pax6 and Six3 in the tissue-specific defects found in Pax6 heterozygous mice.

Golgi matrix protein gene, Golga3/Mea2, rearranged and re-expressed in pachytene spermatocytes restores spermatogenesis in the mouse

In a transgenic mouse, Golga3/Mea2 gene (human homolog: GOLGA3/golgin-160) was disrupted by a translocation at the site of the transgene integration. Exons 8-24 of the disrupted gene remained intact and formed a fusion gene (DeltaMea2) with the antisense strand of E. coli-derived transgene by means of a cryptic splice signal in there. The protein product of DeltaMea2, virtually a form truncated to 2/3 of the normal size, localized to Golgi apparatus of pachytene spermatocytes and round spermatids. DeltaMea2 expression was specific to the testis, but varied among separate seminiferous tubules. It also showed variation among homozygous individuals from 0.5 to 4.3% of the wild type (wt) level. At the lowest levels, neither spermatids nor spermatozoa were present in the homozygous testes, but when the expression of DeltaMea2 increased to 4.3% of the wt level, high sperm production was restored and a sporadic (1/22) fertile homozygous male was obtained. The earliest apoptotic degeneration of pachytene spermatocytes evidenced at 17 dpp in homozygous testes in some discrete seminiferous tubules was preceded by DeltaMea2 expression in a variegated fashion at 16 dpp. These results consistently indicated that in homozygous testes, the pachytene spermatocytes which failed to express DeltaMea2 may undergo apoptotic degeneration. Golga3/Mea2, and DeltaMea2 in homozygotes, in a certain excessive amount may be important for survival of pachytene spermatocytes in the mouse.

Exploring the etiology of haploinsufficiency

The focus of this essay is the phenomenon of haploinsufficiency (HI), a manifestation of genetic dominance that arises when only one allele of a normally diploid locus is present. Specifically, I examine the nature of HI for transcription factor genes. Although the concept of HI applies to many such genes, there is a potentially large variety of mechanisms that underlie it. Even when the phenomenon is linked in all cases to reduced absolute gene expression levels, there are several well-documented cases where the explanation is not reduced expression per se but altered stoichiometry. I will discuss the notion of haploimbalance in general and evaluate the property of transcriptional synergy within the context of HI. This kind of non-linear behaviour can probably explain a large proportion of the cases of HI, as well as the variability in most HI phenotypes and the fact that several factors in the same pathway may be dosage sensitive. For the sake of generality, a theoretical analysis of simple non-linear HI systems is also attempted. This article is certainly another preliminary exploration of the complex matters of HI, which remain an intellectual challenge from many points of view.

Stochasticity in transcriptional regulation: origins, consequences, and mathematical representations

Transcriptional regulation is an inherently noisy process. The origins of this stochastic behavior can be traced to the random transitions among the discrete chemical states of operators that control the transcription rate and to finite number fluctuations in the biochemical reactions for the synthesis and degradation of transcripts. We develop stochastic models to which these random reactions are intrinsic and a series of simpler models derived explicitly from the first as approximations in different parameter regimes. This innate stochasticity can have both a quantitative and qualitative impact on the behavior of gene-regulatory networks. We introduce a natural generalization of deterministic bifurcations for classification of stochastic systems and show that simple noisy genetic switches have rich bifurcation structures; among them, bifurcations driven solely by changing the rate of operator fluctuations even as the underlying deterministic system remains unchanged. We find stochastic bistability where the deterministic equations predict monostability and vice-versa. We derive and solve equations for the mean waiting times for spontaneous transitions between quasistable states in these switches.

Random monoallelic expression of three genes clustered within 60 kb of mouse t complex genomic DNA

Mammals achieve gene dosage control by (1) random X-chromosome inactivation in females, (2) parental origin-specific imprinting of selected autosomal genes, and (3) random autosomal inactivation. Genes belonging to the third category of epigenetic phenomenon are just now emerging, with only six identified so far. Here we report three additional genes, Nubp2, Igfals, and Jsap1, that show 50%-methylated CpG sites by Southern blot analyses and primarily monoallelic expression in single-cell allele-specific RT-PCR analysis of bone marrow stromal cells and hepatocytes. Furthermore, we show that, in contrast to X inactivation, alleles can switch between active and inactive states during the formation of daughter cells. These three genes are the first in their category to exist as a tight cluster, in the proximal region of mouse chromosome 17, providing a thus far unique example of a region of autosomal random monoallelic expression.

Autosomal dominant polycystic kidney disease: modification of disease progression

Autosomal dominant polycystic kidney disease is a common inherited disorder, which is characterised by the formation of fluid-filled cysts in both kidneys that leads to progressive renal failure. Mutations in two genes, PKD1 and PKD2, are associated with the disorder. We describe the various factors that cause variation in disease progression between patients. These include whether the patient has a germline mutation in the PKD1 or in the PKD2 gene, and the nature of the mutation. Detection of mutations in PKD1 is complicated, but the total number identified is rising and will enable genotype-to-phenotype studies. Another factor affecting disease progression is the occurrence of somatic mutations in PKD genes. Furthermore, modifying genes might directly affect the function of polycystins by affecting the rate of somatic mutations or the rate of protein interactions, or they might affect cystogenesis itself or clinical factors associated with disease progression. Finally, environmental factors that speed up or slow down progress towards chronic renal failure have been identified in rodents.

Intrinsic noise in gene regulatory networks

Cells are intrinsically noisy biochemical reactors: low reactant numbers can lead to significant statistical fluctuations in molecule numbers and reaction rates. Here we use an analytic model to investigate the emergent noise properties of genetic systems. We find for a single gene that noise is essentially determined at the translational level, and that the mean and variance of protein concentration can be independently controlled. The noise strength immediately following single gene induction is almost twice the final steady-state value. We find that fluctuations in the concentrations of a regulatory protein can propagate through a genetic cascade; translational noise control could explain the inefficient translation rates observed for genes encoding such regulatory proteins. For an autoregulatory protein, we demonstrate that negative feedback efficiently decreases system noise. The model can be used to predict the noise characteristics of networks of arbitrary connectivity. The general procedure is further illustrated for an autocatalytic protein and a bistable genetic switch. The analysis of intrinsic noise reveals biological roles of gene network structures and can lead to a deeper understanding of their evolutionary origin.

The art of the probable: system control in the adaptive immune system

The immune system provides very effective host defense against infectious agents. Although many details are known about the cells and molecules involved, a broader "systems engineering" view of this complex system is just beginning to emerge. Here the argument is put forward that stochastic events, potent amplification mechanisms, feedback controls, and heterogeneity arising from spatially dispersed cell interactions give rise to many of the gross properties of the immune system. A better appreciation of these underlying features will not only add to our basic understanding of how immunity develops or goes awry, but also illuminate new directions for manipulating the system in prophylactic and therapeutic settings.