Predominantly inverse modulation of gene expression in genomically unbalanced disomic haploid maize.

The phenotypic consequences of the addition or subtraction of part of a chromosome is more severe than changing the dosage of the whole genome. By crossing diploid trisomies to a haploid inducer, we identified 17 distal segmental haploid disomies that cover ∼80% of the maize genome. Disomic haploids provide a level of genomic imbalance that is not ordinarily achievable in multicellular eukaryotes, allowing the impact to be stronger and more easily studied. Transcriptome size estimates revealed that a few disomies inversely modulate most of the transcriptome. Based on RNA sequencing, the expression levels of genes located on the varied chromosome arms (cis) in disomies ranged from being proportional to chromosomal dosage (dosage effect) to showing dosage compensation with no expression change with dosage. For genes not located on the varied chromosome arm (trans), an obvious trans-acting effect can be observed, with the majority showing a decreased modulation (inverse effect). The extent of dosage compensation of varied cis genes correlates with the extent of trans inverse effects across the 17 genomic regions studied. The results also have implications for the role of stoichiometry in gene expression, the control of quantitative traits, and the evolution of dosage-sensitive genes.

Genomic imbalance determines positive and negative modulation of gene expression in diploid maize.

Genomic imbalance caused by changing the dosage of individual chromosomes (aneuploidy) has a more detrimental effect than varying the dosage of complete sets of chromosomes (ploidy). We examined the impact of both increased and decreased dosage of 15 distal and 1 interstitial chromosomal regions via RNA-seq of maize (Zea mays) mature leaf tissue to reveal new aspects of genomic imbalance. The results indicate that significant changes in gene expression in aneuploids occur both on the varied chromosome (cis) and the remainder of the genome (trans), with a wider spread of modulation compared with the whole-ploidy series of haploid to tetraploid. In general, cis genes in aneuploids range from a gene-dosage effect to dosage compensation, whereas for trans genes the most common effect is an inverse correlation in that expression is modulated toward the opposite direction of the varied chromosomal dosage, although positive modulations also occur. Furthermore, this analysis revealed the existence of increased and decreased effects in which the expression of many genes under genome imbalance are modulated toward the same direction regardless of increased or decreased chromosomal dosage, which is predicted from kinetic considerations of multicomponent molecular interactions. The findings provide novel insights into understanding mechanistic aspects of gene regulation.

Sequence of the supernumerary B chromosome of maize provides insight into its drive mechanism and evolution.

B chromosomes are enigmatic elements in thousands of plant and animal genomes that persist in populations despite being nonessential. They circumvent the laws of Mendelian inheritance but the molecular mechanisms underlying this behavior remain unknown. Here we present the sequence, annotation, and analysis of the maize B chromosome providing insight into its drive mechanism. The sequence assembly reveals detailed locations of the elements involved with the cis and trans functions of its drive mechanism, consisting of nondisjunction at the second pollen mitosis and preferential fertilization of the egg by the B-containing sperm. We identified 758 protein-coding genes in 125.9 Mb of B chromosome sequence, of which at least 88 are expressed. Our results demonstrate that transposable elements in the B chromosome are shared with the standard A chromosome set but multiple lines of evidence fail to detect a syntenic genic region in the A chromosomes, suggesting a distant origin. The current gene content is a result of continuous transfer from the A chromosomal complement over an extended evolutionary time with subsequent degradation but with selection for maintenance of this nonvital chromosome.

Effect of aneuploidy of a non-essential chromosome on gene expression in maize.

The non-essential supernumerary maize (Zea mays) B chromosome (B) has recently been shown to contain active genes and to be capable of impacting gene expression of the A chromosomes. However, the effect of the B chromosome on gene expression is still unclear. In addition, it is unknown whether the accumulation of the B chromosome has a cumulative effect on gene expression. To examine these questions, the global expression of genes, microRNAs (miRNAs), and transposable elements (TEs) of leaf tissue of maize W22 plants with 0-7 copies of the B chromosome was studied. All experimental genotypes with B chromosomes displayed a trend of upregulated gene expression for a subset of A-located genes compared to the control. Over 3000 A-located genes are significantly differentially expressed in all experimental genotypes with the B chromosome relative to the control. Modulations of these genes are largely determined by the presence rather than the copy number of the B chromosome. By contrast, the expression of most B-located genes is positively correlated with B copy number, showing a proportional gene dosage effect. The B chromosome also causes increased expression of A-located miRNAs. Differentially expressed miRNAs potentially regulate their targets in a cascade of effects. Furthermore, the varied copy number of the B chromosome leads to the differential expression of A-located and B-located TEs. The findings provide novel insights into the function and properties of the B chromosome.

BICOSS: Bayesian iterative conditional stochastic search for GWAS.

BACKGROUND: Single marker analysis (SMA) with linear mixed models for genome wide association studies has uncovered the contribution of genetic variants to many observed phenotypes. However, SMA has weak false discovery control. In addition, when a few variants have large effect sizes, SMA has low statistical power to detect small and medium effect sizes, leading to low recall of true causal single nucleotide polymorphisms (SNPs). RESULTS: We present the Bayesian Iterative Conditional Stochastic Search (BICOSS) method that controls false discovery rate and increases recall of variants with small and medium effect sizes. BICOSS iterates between a screening step and a Bayesian model selection step. A simulation study shows that, when compared to SMA, BICOSS dramatically reduces false discovery rate and allows for smaller effect sizes to be discovered. Finally, two real world applications show the utility and flexibility of BICOSS. CONCLUSIONS: When compared to widely used SMA, BICOSS provides higher recall of true SNPs while dramatically reducing false discovery rate.

Global impacts of chromosomal imbalance on gene expression in Arabidopsis and other taxa.

Changes in dosage of part of the genome (aneuploidy) have long been known to produce much more severe phenotypic consequences than changes in the number of whole genomes (ploidy). To examine the basis of these differences, global gene expression in mature leaf tissue for all five trisomies and in diploids, triploids, and tetraploids of Arabidopsis thaliana was studied. The trisomies displayed a greater spread of expression modulation than the ploidy series. In general, expression of genes on the varied chromosome ranged from compensation to dosage effect, whereas genes from the remainder of the genome ranged from no effect to reduced expression approaching the inverse level of chromosomal imbalance (2/3). Genome-wide DNA methylation was examined in each genotype and found to shift most prominently with trisomy 4 but otherwise exhibited little change, indicating that genetic imbalance is generally mechanistically unrelated to DNA methylation. Independent analysis of gene functional classes demonstrated that ribosomal, proteasomal, and gene body methylated genes were less modulated compared with all classes of genes, whereas transcription factors, signal transduction components, and organelle-targeted protein genes were more tightly inversely affected. Comparing transcription factors and their targets in the trisomies and in expression networks revealed considerable discordance, illustrating that altered regulatory stoichiometry is a major contributor to genetic imbalance. Reanalysis of published data on gene expression in disomic yeast and trisomic mouse cells detected similar stoichiometric effects across broad phylogenetic taxa, and indicated that these effects reflect normal gene regulatory processes.

Conditions of embryo culture from days 5 to 7 of development alter the DNA methylome of the bovine fetus at day 86 of gestation.

PURPOSE: We tested whether in vitro production (IVP) causes changes in DNA methylation in fetal liver and skeletal muscle and if exposure of cultured embryos to colony-stimulating factor 2 (CSF2) alters DNA methylation. METHODS: Female fetuses were produced by artificial insemination or transfer of an IVP embryo. Embryos were treated from days 5 to 7 after fertilization with CSF2 or vehicle. DNA methylation in fetal liver and skeletal muscle was determined by post-bisulfite adaptor tagging-based sequencing. The degree of DNA methylation for CpG sites in 50-bp windows of the promoter region 500 bp upstream of the transcriptional start site was compared between treatments. RESULTS: For liver, there were 12 genes (6% of those analyzed) in which DNA methylation was affected by treatment, with one 50-bp window per gene affected by treatment. For muscle, the degree of DNA methylation was affected by treatment for 32 windows (19% of the total windows analyzed) representing 28 distinct genes (23% of analyzed genes). For 19 of the 28 genes in muscle, the greatest deviation in DNA methylation was for the CSF2 group. CONCLUSION: Results are consistent with alterations in the methylome being one of the mechanisms by which IVP can result in altered fetal development and postnatal function in the resultant offspring. In addition, results indicate that maternally derived cell-signaling molecules can regulate the pattern of DNA methylation.

Modeling allele-specific expression at the gene and SNP levels simultaneously by a Bayesian logistic mixed regression model.

BACKGROUND: High-throughput sequencing experiments, which can determine allele origins, have been used to assess genome-wide allele-specific expression. Despite the amount of data generated from high-throughput experiments, statistical methods are often too simplistic to understand the complexity of gene expression. Specifically, existing methods do not test allele-specific expression (ASE) of a gene as a whole and variation in ASE within a gene across exons separately and simultaneously. RESULTS: We propose a generalized linear mixed model to close these gaps, incorporating variations due to genes, single nucleotide polymorphisms (SNPs), and biological replicates. To improve reliability of statistical inferences, we assign priors on each effect in the model so that information is shared across genes in the entire genome. We utilize Bayesian model selection to test the hypothesis of ASE for each gene and variations across SNPs within a gene. We apply our method to four tissue types in a bovine study to de novo detect ASE genes in the bovine genome, and uncover intriguing predictions of regulatory ASEs across gene exons and across tissue types. We compared our method to competing approaches through simulation studies that mimicked the real datasets. The R package, BLMRM, that implements our proposed algorithm, is publicly available for download at https://github.com/JingXieMIZZOU/BLMRM . CONCLUSIONS: We will show that the proposed method exhibits improved control of the false discovery rate and improved power over existing methods when SNP variation and biological variation are present. Besides, our method also maintains low computational requirements that allows for whole genome analysis.

Lean maternal hyperglycemia alters offspring lipid metabolism and susceptibility to diet-induced obesity in mice†.

We previously developed a model of gestational diabetes mellitus (GDM) in which dams exhibit glucose intolerance, insulin resistance, and reduced insulin response to glucose challenge only during pregnancy, without accompanying obesity. Here, we aimed to determine how lean gestational glucose intolerance affects offspring risk of metabolic dysfunction. One cohort of offspring was sacrificed at 19 weeks, and one at 31 weeks, with half of the second cohort placed on a high-fat, high-sucrose diet (HFHS) at 23 weeks. Exposure to maternal glucose intolerance increased weights of HFHS-fed offspring. Chow-fed offspring of GDM dams exhibited higher body fat percentages at 4, 12, and 20 weeks of age. At 28 weeks, offspring of GDM dams fed the HFHS but not the chow diet (CD) also had higher body fat percentages than offspring of controls (CON). Exposure to GDM increased the respiratory quotient (Vol CO2/Vol O2) in offspring. Maternal GDM increased adipose mRNA levels of peroxisome proliferator-activated receptor gamma (Pparg) and adiponectin (Adipoq) in 31-week-old CD-fed male offspring, and increased mRNA levels of insulin receptor (Insr) and lipoprotein lipase (Lpl) in 31-week-old male offspring on both diets. In liver at 31 weeks, mRNA levels of peroxisome proliferator-activated receptor alpha (Ppara) were elevated in CD-fed male offspring of GDM dams, and male offspring of GDM dams exhibited higher mRNA levels of Insr on both diets. Neither fasting insulin nor glucose tolerance was affected by exposure to GDM. Our findings show that GDM comprising glucose intolerance only during pregnancy programs increased adiposity in offspring, and suggests increased insulin sensitivity of subcutaneous adipose tissue.

A Bayesian hidden Markov model for detecting differentially methylated regions.

Alterations in DNA methylation have been linked to the development and progression of many diseases. The bisulfite sequencing technique presents methylation profiles at base resolution. Count data on methylated and unmethylated reads provide information on the methylation level at each CpG site. As more bisulfite sequencing data become available, these data are increasingly needed to infer methylation aberrations in diseases. Automated and powerful algorithms also need to be developed to accurately identify differentially methylated regions between treatment groups. This study adopts a Bayesian approach using the hidden Markov model to account for inherent dependence in read count data. Given the expense of sequencing experiments, few replicates are available for each treatment group. A Bayesian approach that borrows information across an entire chromosome improves the reliability of statistical inferences. The proposed hidden Markov model considers location dependence among genomic loci by incorporating correlation structures as a function of genomic distance. An iterative algorithm based on expectation-maximization is designed for parameter estimation. Methylation states are inferred by identifying the optimal sequence of latent states from observations. Real datasets and simulation studies that mimic the real datasets are used to illustrate the reliability and success of the proposed method.

Detecting differentially expressed genes for syndromes by considering change in mean and dispersion simultaneously.

BACKGROUND: Using next-generation sequencing technology to measure gene expression, an empirically intriguing question concerns the identification of differentially expressed genes across treatment groups. Existing methods aim to identify genes whose mean expressions differ among treatment groups by assuming equal dispersion across all groups. For syndromes, however, various combinations of gene expression alterations can result in the same disease, leading to greater heteroscedasticity in the biological replicates in the disease group compared to the normal group. Traditional methods that only consider changes in the mean will fail to fully analyze gene expression in such a scenario. In addition, sequencing technology is relatively expensive; most labs can only afford a few replicates per treatment group, which poses further challenges to reliably estimating the mean and dispersion under each treatment condition. RESULTS: We designed an empirical Bayes method and a pooled permutation test to simultaneously consider the change in mean and dispersion across treatment groups. We further computed confidence intervals based on Bayes estimates to identify differentially expressed genes that are unique to each disease sample as well as those that are common across all disease samples. We illustrated our method by applying it to gene expression data from a large offspring syndrome experiment, which motivated this study. We compared our method to competing approaches through simulation studies that mimicked the real datasets to demonstrate the effectiveness of our proposed method. CONCLUSIONS: We will show that, compared to popular methods that only aim to find the difference in the mean, our method can capture greater variation in the disease group to effectively identify differentially expressed genes for syndromes.

Glutamine supplementation enhances development of in vitro-produced porcine embryos and increases leucine consumption from the medium.

Improper composition of culture medium contributes to reduced viability of in vitro-produced embryos. Glutamine (Gln) is a crucial amino acid for preimplantation embryos as it supports proliferation and is involved in many different biosynthetic pathways. Previous transcriptional profiling revealed several upregulated genes related to Gln transport and metabolism in in vitro-produced porcine blastocysts compared to in vivo-produced counterparts, indicating a potential deficiency in the culture medium. Therefore, the objective of this study was to determine the effects of Gln supplementation on in vitro-produced porcine embryo development, gene expression, and metabolism. Cleaved embryos were selected and cultured in MU2 medium supplemented with 1 mM Gln (control), 3.75 mM Gln (+Gln), 3.75 mM GlutaMAX (+Max), or 3.75 mM alanine (+Ala) until day 6. Embryos cultured with +Gln or +Max had increased development to the blastocyst stage and total number of nuclei compared to the control (P < 0.05). Moreover, expression of misregulated transcripts involved in glutamine and glutamate transport and metabolism was corrected when embryos were cultured with +Gln or +Max. Metabolomics analysis revealed increased production of glutamine and glutamate into the medium by embryos cultured with +Max and increased consumption of leucine by embryos cultured with +Gln or +Max. As an indicator of cellular health, mitochondrial membrane potential was increased when embryos were cultured with +Max which was coincident with decreased apoptosis in these blastocysts. Lastly, two embryo transfers by using embryos cultured with +Max resulted in viable piglets, confirming that this treatment is consistent with in vivo developmental competence.

Characterization of global loss of imprinting in fetal overgrowth syndrome induced by assisted reproduction.

Embryos generated with the use of assisted reproductive technologies (ART) can develop overgrowth syndromes. In ruminants, the condition is referred to as large offspring syndrome (LOS) and exhibits variable phenotypic abnormalities including overgrowth, enlarged tongue, and abdominal wall defects. These characteristics recapitulate those observed in the human loss-of-imprinting (LOI) overgrowth syndrome Beckwith-Wiedemann (BWS). We have recently shown LOI at the KCNQ1 locus in LOS, the most common epimutation in BWS. Although the first case of ART-induced LOS was reported in 1995, studies have not yet determined the extent of LOI in this condition. Here, we determined allele-specific expression of imprinted genes previously identified in human and/or mouse in day ∼105 Bos taurus indicus × Bos taurus taurus F1 hybrid control and LOS fetuses using RNAseq. Our analysis allowed us to determine the monoallelic expression of 20 genes in tissues of control fetuses. LOS fetuses displayed variable LOI compared with controls. Biallelic expression of imprinted genes in LOS was associated with tissue-specific hypomethylation of the normally methylated parental allele. In addition, a positive correlation was observed between body weight and the number of biallelically expressed imprinted genes in LOS fetuses. Furthermore, not only was there loss of allele-specific expression of imprinted genes in LOS, but also differential transcript amounts of these genes between control and overgrown fetuses. In summary, we characterized previously unidentified imprinted genes in bovines and identified misregulation of imprinting at multiple loci in LOS. We concluded that LOS is a multilocus LOI syndrome, as is BWS.

What if we ignore the random effects when analyzing RNA-seq data in a multifactor experiment.

Identifying differentially expressed (DE) genes between different conditions is one of the main goals of RNA-seq data analysis. Although a large amount of RNA-seq data were produced for two-group comparison with small sample sizes at early stage, more and more RNA-seq data are being produced in the setting of complex experimental designs such as split-plot designs and repeated measure designs. Data arising from such experiments are traditionally analyzed by mixed-effects models. Therefore an appropriate statistical approach for analyzing RNA-seq data from such designs should be generalized linear mixed models (GLMM) or similar approaches that allow for random effects. However, common practices for analyzing such data in literature either treat random effects as fixed or completely ignore the experimental design and focus on two-group comparison using partial data. In this paper, we examine the effect of ignoring the random effects when analyzing RNA-seq data. We accomplish this goal by comparing the standard GLMM model to the methods that ignore the random effects through simulation studies and real data analysis. Our studies show that, ignoring random effects in a multi-factor experiment can lead to the increase of the false positives among the top selected genes or lower power when the nominal FDR level is controlled.

Identification and functional characterization of soybean root hair microRNAs expressed in response to Bradyrhizobium japonicum infection.

Three soybean [Glycine max (L) Merr.] small RNA libraries were generated and sequenced using the Illumina platform to examine the role of miRNAs during soybean nodulation. The small RNA libraries were derived from root hairs inoculated with Bradyrhizobium japonicum (In_RH) or mock-inoculated with water (Un_RH), as well as from the comparable inoculated stripped root samples (i.e. inoculated roots with the root hairs removed). Sequencing of these libraries identified a total of 114 miRNAs, including 22 novel miRNAs. A comparison of miRNA abundance among the 114 miRNAs identified 66 miRNAs that were differentially expressed between root hairs and stripped roots, and 48 miRNAs that were differentially regulated in infected root hairs in response to B. japonicum when compared to uninfected root hairs (P ≤ 0.05). A parallel analysis of RNA ends (PARE) library was constructed and sequenced to reveal a total of 405 soybean miRNA targets, with most predicted to encode transcription factors or proteins involved in protein modification, protein degradation and hormone pathways. The roles of gma-miR4416 and gma-miR2606b during nodulation were further analysed. Ectopic expression of these two miRNAs in soybean roots resulted in significant changes in nodule numbers. miRNA target information suggested that gma-miR2606b regulates a Mannosyl-oligosaccharide 1, 2-alpha-mannosidase gene, while gma-miR4416 regulates the expression of a rhizobium-induced peroxidase 1 (RIP1)-like peroxidase gene, GmRIP1, during nodulation.

Modeling the next generation sequencing read count data for DNA copy number variant study.

As one of the most recent advanced technologies developed for biomedical research, the next generation sequencing (NGS) technology has opened more opportunities for scientific discovery of genetic information. The NGS technology is particularly useful in elucidating a genome for the analysis of DNA copy number variants (CNVs). The study of CNVs is important as many genetic studies have led to the conclusion that cancer development, genetic disorders, and other diseases are usually relevant to CNVs on the genome. One way to analyze the NGS data for detecting boundaries of CNV regions on a chromosome or a genome is to phrase the problem as a statistical change point detection problem presented in the read count data. We therefore provide a statistical change point model to help detect CNVs using the NGS read count data. We use a Bayesian approach to incorporate possible parameter changes in the underlying distribution of the NGS read count data. Posterior probabilities for the change point inferences are derived. Extensive simulation studies have shown advantages of our proposed methods. The proposed methods are also applied to a publicly available lung cancer cell line NGS dataset, and CNV regions on this cell line are successfully identified.

Identification of microRNAs and their mRNA targets during soybean nodule development: functional analysis of the role of miR393j-3p in soybean nodulation.

Plant microRNAs (miRNAs) play important regulatory roles in a number of developmental processes. The present work investigated the roles of miRNAs during nodule development in the crop legume soybean (Glycine max). Fifteen soybean small RNA libraries were sequenced from different stages of nodule development, including young nodules, mature nodules and senescent nodules. In order to identify the regulatory targets of the miRNAs, five parallel analysis of RNA ends (PARE) libraries were also sequenced from the same stages of nodule development. Sequencing identified 284 miRNAs, including 178 novel soybean miRNAs. Analysis of miRNA abundance identified 139 miRNAs whose expression was significantly regulated during nodule development, including 12 miRNAs whose expression changed > 10-fold. Analysis of the PARE libraries identified 533 miRNA targets, including three nodulation-related genes and eight nodule-specific genes. miR393j-3p was selected for detailed analysis as its expression was significantly regulated during nodule formation, and it targeted a nodulin gene, Early Nodulin 93 (ENOD93). Strong, ectopic expression of miR393j-3p, as well as RNAi silencing of ENOD93 expression, significantly reduced nodule formation. The data indicate that miR393j-3p regulation of ENOD93 mRNA abundance is a key control point for soybean nodule formation.

Effects of the use of assisted reproduction and high-caloric diet consumption on body weight and cardiovascular health of juvenile mouse offspring.

Maternal obesity and the use of assisted reproductive technologies (ART) are two suboptimal developmental environments that can lead to offspring obesity and cardiovascular disease. We hypothesized that these environments independently and synergistically adversely affect the offspring's weight and cardiovascular performance at ~7 weeks of age. Mice were fed either 24% fat and 17.5% high-fructose (HF) corn syrup or maintenance chow (5% fat; low-fat, no-fructose (LF)). Dams were subdivided into no ART and ART groups. ART embryos were cultured in Whitten's medium and transferred into pseudopregnant recipients consuming the same diet as the donor. Offspring were fed the same diet as the mother. Body weights (BW) were measured weekly and mean arterial pressure (MAP) was collected through carotid artery catheterization at killing (55±0.5 days old). Expression of genes involved in cardiovascular remodeling was measured in thoracic aorta using qRT-PCR, and levels of reactive oxygen species (ROS) were measured intracellularly and extracellularly in mesenteric resistance arteries. ART resulted in increased BW at weaning. This effect decreased over time and diet was the predominant determinant of BW by killing. Males had greater MAP than females (P=0.002) and HF consumption was associated with greater MAP regardless of sex (P<0.05). Gene expression was affected by sex (P<0.05) and diet (P<0.1). Lastly, the use of ART resulted in offspring with increased intracellular ROS (P=0.05). In summary, exposure to an obesogenic diet pre- and/or post-natally affects weight, MAP, and gene expression while ART increases oxidative stress in mesenteric resistance arteries of juvenile offspring, no synergistic effects were observed.

Impacts of Salmonella enteritidis infection on liver transcriptome in broilers.

Salmonella enterica serovar enteritidis is an enteric bacterium that can contaminate chicken eggs and meat, resulting in production losses and consumer illness. To provide insight into the systemic metabolic effects of S. enteritidis infection, liver samples were harvested 10-days postinfection from broiler hens. Hepatic global gene expression levels were assessed using a chicken 44K Agilent microarray. Forty-four genes were differentially expressed at a significance level of q value < 0.05. One hundred eighty-three genes were differentially expressed at a suggestive significance level of q value < 0.1. A predominance of downregulation existed among significantly differentially expressed genes. Cell cycle and metabolism networks were created from the differentially expressed genes. Mitochondria-mediated apoptosis, electron transport, peptidase activity, vein constriction, cell differentiation, IL-2 signaling, Jak-Stat signaling, B-cell receptor signaling, GDP/GTP exchange, and protein recycling were among the functions of the differentially expressed genes that were down-regulated in response to S. enteritidis. The effects of S. enteritidis infection on the liver transcriptome profiles of broilers reflect a predominance of downregulation of genes with cell cycle and metabolic functions. The most pronounced response was the downregulation of genes that function in metabolic pathways, inflammation, and mitochondria-mediated apoptosis. These results provide insight into important systemic metabolic mechanisms that are active in the chicken liver in response to S. enteritidis infection at 10-days postinfection.

Borrowing information across genes and experiments for improved error variance estimation in microarray data analysis.

Statistical inference for microarray experiments usually involves the estimation of error variance for each gene. Because the sample size available for each gene is often low, the usual unbiased estimator of the error variance can be unreliable. Shrinkage methods, including empirical Bayes approaches that borrow information across genes to produce more stable estimates, have been developed in recent years. Because the same microarray platform is often used for at least several experiments to study similar biological systems, there is an opportunity to improve variance estimation further by borrowing information not only across genes but also across experiments. We propose a lognormal model for error variances that involves random gene effects and random experiment effects. Based on the model, we develop an empirical Bayes estimator of the error variance for each combination of gene and experiment and call this estimator BAGE because information is Borrowed Across Genes and Experiments. A permutation strategy is used to make inference about the differential expression status of each gene. Simulation studies with data generated from different probability models and real microarray data show that our method outperforms existing approaches.