Use of normalization methods for analysis of microarrays containing a high degree of gene effects.

BACKGROUND: High-throughput microarrays are widely used to study gene expression across tissues and developmental stages. Analysis of gene expression data is challenging in these experiments due to the presence of significant percentages of differentially expressed genes (DEG) observed between tissues and developmental stages. Data normalization methods that are widely used today are not designed for data with a large proportion of tissue or gene effects. RESULTS: In our current study, we describe a novel two-dimensional nonparametric normalization method for analyzing microarray data which functions well in the absence or presence of large numbers of gene effects. Rather than relying on an assumption of low variability among most genes, the method implements a unique peak selection strategy to distinguish DEG from genes that are invariant in expression, prior to nonlinear curve fitting. We compared the method under simulated and experimental conditions with five alternative nonlinear normalization approaches: quantile, lowess, robust lowess, invariant set, and cross-correlation (Xcorr). Simulations included various percentages of simulated DEG and the experimental data used is from publicly available datasets known to be difficult to analyze due to the presence of approximately 34% DEG. CONCLUSION: We have demonstrated that the new method provides considerable improvement in the accuracy of data normalization when large proportions of gene effects are present. The performance improvement is mostly attributed to its variable selection component, which is designed to separate expression invariant genes from DEG. Adding this key component of the new method to alternative normalization approaches rescues the most of the sensitivity of these methods to gene effects. The results indicate that our method may be used without prior knowledge of or assumptions about housekeeping genes to normalize microarrays that are quite different.

Efficacy of the farnesyltransferase inhibitor R115777 in a rat mammary tumor model: role of Ha-ras mutations and use of microarray analysis in identifying potential targets.

Rats treated with the alkylating agent methylnitrosourea (MNU) develop multiple, hormonally dependent mammary tumors. Roughly 50% of the tumors have Ha-ras mutation, whereas 50% do not. The MNU-induced rat mammary tumor model was employed to examine the therapeutic efficacy of the farnesyltransferase inhibitor (FTI), R115777, and to examine the use of genomics in identifying susceptible tumors as well as identifying genes whose expression are modulated by FTI treatment. In animals bearing palpable mammary tumors (< 7 mm diameter), we performed a surgical biopsy, and 3 days following the biopsy, rats were treated with R115777 (50 mg/kg body wt/day) by gavage. Tumors with Ha-ras mutations underwent profound regression, with nearly 90% showing complete regressions within 4 weeks. In contrast, the non-Ha-ras mutation-bearing tumors yielded a more variable response, although roughly half of the non-Ha-ras mutation tumors underwent significant regression. These results show that although all tumors appear to respond to the FTI inhibitor the tumors with Ha-ras mutations were exquisitely sensitive. We employed a microarray approach to define potential targets and the mechanism of action of R115777 in Ha-ras mutant or wildtype tumors following treatment with FTI. In addition, we determined whether gene expression prior to FTI treatment can be used to differentiate highly sensitive tumors (Ha-ras mutant) and tumors with variable sensitivity (Ha-ras wildtype). Untreated or FTI-treated (4 days at 50 mg/kg body wt) tumors (Ha-ras mutant or wildtype) were examined using oligonucleotide arrays. A significant number of genes were differentially expressed in control rat mammary tumors with or without an activated Ha-ras mutation, suggesting that a microarray analysis might differentiate highly sensitive and variably sensitive tumors. Most of the genes whose expressions were modulated by FTI in tumors were independent of Ha-ras status and were presumably modulated by effects on farnesylation of proteins other than Ha-ras. However, treatment of Ha-ras-mutated mammary tumors with R155777 results in preferential modulation of genes involved in ras-MAP kinase signal transduction pathway and in decreased expression of many genes involved with cell proliferation. In contrast, several classes of genes are altered in rat mammary tumors without a mutated Ha-ras, suggesting that non-ras targets are involved. Ras pathway related genes, p53, WT1 and PCNA, were preferentially modulated in Ha-ras-mutated tumors, whereas modulation of genes in the G-protein pathway, various cytochrome p450s and RB1 are involved in Ha-ras wildtype tumors. Elucidation of gene expression changes in FTI-treated or control rat mammary adenocarcinomas will help in identifying potential pharmacodynamic markers of FTI treatment as well as potential molecular targets of R115777 and other FTIs.

Altered gene expression profile in mouse bladder cancers induced by hydroxybutyl(butyl)nitrosamine.

A variety of genetic alterations and gene expression changes are involved in the pathogenesis of bladder tumor. To explore these changes, oligonucleotide array analysis was performed on RNA obtained from carcinogen-induced mouse bladder tumors and normal mouse bladder epithelia using Affymetrix (Santa Clara, CA) MGU74Av2 GeneChips. Analysis yielded 1164 known genes that were changed in the tumors. Certain of the upregulated genes included EGFR-Ras signaling genes, transcription factors, cell cycle-related genes, and intracellular signaling cascade genes. However, downregulated genes include mitogen-activated protein kinases, cell cycle checkpoint genes, Rab subfamily genes, Rho subfamily genes, and SH2 and SH3 domains-related genes. These genes are involved in a broad range of different pathways including control of cell proliferation, differentiation, cell cycle, signal transduction, and apoptosis. Using the pathway visualization tool GenMAPP, we found that several genes, including TbR-I, STAT1, Smad1, Smad2, Jun, NFkappaB, and so on, in the TGF-beta signaling pathway and p115 RhoGEF, RhoGDI3, MEKK4A/MEKK4B, PI3KA, and JNK in the G13 signaling pathway were differentially expressed in the tumors. In summary, we have determined the expression profiles of genes differentially expressed during mouse bladder tumorigenesis. Our results suggest that activation of the EGFR-Ras pathway, uncontrolled cell cycle, aberrant transcription factors, and G13 and TGF-beta pathways are involved, and the cross-talk between these pathways seems to play important roles in mouse bladder tumorigenesis.

Budesonide exerts its chemopreventive efficacy during mouse lung tumorigenesis by modulating gene expressions.

Budesonide, a glucocorticoid, was proven to be a highly effective agent in preventing the development of lung tumors in A/J mice. In a lung tumor bioassay, budesonide produced 70% inhibition of tumor multiplicity and 94% reduction of total tumor load compared to benzopyrene (B[a]P) treated mice. Gene expression array analysis was performed on mouse lung tumors from this bioassay using Affymetrix U74Av2 GeneChips to determine gene expression changes associated with budesonide treatment. We found 363 genes that were changed between lung tumors induced by treatment with B[a]P and similar tumors treated with budesonide. Among them, 243 genes were overexpressed and 120 genes were underexpressed after budesonide treatment. In addition, 108 genes differentially expressed during mouse lung tumorigenesis (50 genes overexpressed and 58 genes underexpressed) were modulated back to normal levels after budesonide treatment when compared with the controls group. These genes are involved in a broad range of different pathways including control of cell cycle, signal transduction, and apoptosis and may play a role in the observed preventive effect. Our results suggest that budesonide exerts its effects of chemoprevention through growth arrest via Mad2/3 and through apoptosis via Bim/Blk and, by inference, caspase-8/9. Using the pathway visualization tool GenMapp, G protein pathway and MAPK cascade were also regulated by budesonide. Thus, we have determined, for the first time, the expression profiles of genes modulated by budesonide during murine lung tumorigenesis. Our results indicate that the chemopreventive effects of budesonide in the mouse lung tumorigenesis assay involved increase and decrease expression of a wide variety of genes in multiple signaling pathways.

A chemically induced model for squamous cell carcinoma of the lung in mice: histopathology and strain susceptibility.

Lung cancer, primarily associated with tobacco use, is the leading cause of cancer morbidity and mortality in the United States. Squamous cell carcinoma (SCC) is one of the four major histological types of lung cancer. Although there are several established models for lung adenoma and adenocarcinomas, there is no well-established mouse model for lung SCC. We treated eight different inbred strains of mice with N-nitroso-tris-chloroethylurea by skin painting and found that this regimen induced lung SCCs in five strains of mouse (SWR/J, NIH Swiss, A/J, BALB/cJ, and FVB/J) but not in the others (AKR/J, 129/svJ, and C57BL/6J). Mouse lung SCCs have similar histopathological features and keratin staining to human SCC. Moreover, a wide spectrum of abnormal lung squamous phenotypes including hyperplasia, metaplasia, carcinoma in situ, and invasive carcinoma, were observed. There are strain-specific differences in susceptibility to Lscc induction by N-nitroso-tris-chloroethylurea with NIH Swiss, A/J, and SWR/J mice developing scores of SCCs whereas the resistant strains AKR/J, 129/svJ, and C57BL/6J failed to develop any SCCs. FVB/J and BALB/cJ mice had an intermediate response. We conducted whole-genome linkage disequilibrium analysis in seven strains of mice, divided into three phenotype categories of susceptibility, using Fisher's exact test applied to 6,128 markers in publically available databases. Three markers were found significantly associated with susceptibility to SCC with the P < 0.05. They were D1Mit169, D3Mit178, and D18Mit91. Interestingly, none of these sites overlap with the major susceptibility loci associated with lung adenoma/adenocarcinoma development in mice. The mouse SCC described here is highly significant for preclinical studies of lung cancer chemopreventive agents because most human trials have been conducted against precancerous lesions for SCC. Furthermore, this model can be used in determining genetic modifiers that contribute to susceptibility or resistance to lung SCC development.

Molecular profiling of mouse lung tumors: association with tumor progression, lung development, and human lung adenocarcinomas.

We have performed oligonucleotide array analysis on various murine lung tissues [normal lungs, lung adenomas, and lung adenocarcinomas (ACs)] using Affymetrix U74Av2 GeneChips to examine the complex genetic changes occurring during lung carcinogenesis. Analysis yielded 20 novel genes differentially expressed in both lung adenomas and ACs versus normal lungs, including the tumor suppressor APC2 and the oncogene Ros 1. In addition, 50 genes were found to be differentially expressed in lung adenomas versus lung ACs, including the differentiation factor Hox C6, the oncogene Ets 2, and the Ras nuclear transport factor, nuclear transport factor 2. To understand the potential relationship between genes expressed in murine lung tumors and its relationship to altered gene expression observed during embryogenesis and postnatal development, tissues from embryonic lungs and from lungs of mice up to 4 weeks following birth were examined using Affymetrix U74Av2 GeneChips. From this analysis, approximately 1300 genes were determined to exhibit differential expression in fetal lung versus postnatal lung. When we compared lung adenomas, lung ACs, and normal lung parenchyma, 24 developmentally regulated genes were found aberrantly expressed in lung tumors; these included the cell cycle control factor CDC5, the cellular differentiation factor TEA domain 4, and the proapoptotic factor BNIP 2. Finally, we compared the murine lung tumor gene expression data to the expression of genes in human lung cancer, in order to assess the relevance of murine lung cancer models in the study of human AC formation. When the 17 human lung ACs and six human lung large cell carcinomas were examined, it was found that 13 of the 17 human lung ACs clustered tightly together in a pattern that was different from the remaining four human lung ACs and six large cell carcinomas, which exhibited a different pattern. Interestingly, the mouse lung adenomas appeared similar to 13 clustered ACs, while mouse lung ACs appeared more similar in pattern to the group consisting of four ACs and six large-cell carcinomas (LCCs). Nevertheless, when compared with the combined human ACs, 39 genes with similar expression changes in murine lung tumors and human ACs/LCCs were identified, such as the oncogene-related BCL7B, the cell cycle regulator CDK4, and the proapoptotic Endophilin B1. Overall, we have determined, for the first time, the expression profiles during murine lung tumor progression and have established, at the molecular level, an association between murine lung tumorigenesis and lung development. We have also attempted to compare the expression profiles found in mouse lung cancers and those in human lung ACs.

A high performance test of differential gene expression for oligonucleotide arrays.

Logit-t employs a logit-transformation for normalization followed by statistical testing at the probe-level. Using four publicly-available datasets, together providing 2,710 known positive incidences of differential expression and 2,913,813 known negative incidences, performance of statistical tests were: Logit-t provided 75% positive-predictive value, compared with 5% for Affymetrix Microarray Suite 5, 6% for dChip perfect match (PM)-only, and 9% for Robust Multi-array Analysis at the p < 0.01 threshold. Logit-t provided 70% sensitivity, Microarray Suite 5 provided 46%, dChip provided 53% and Robust Multi-array Analysis provided 63%.

Fine mapping and identification of candidate pulmonary adenoma susceptibility 1 genes using advanced intercross lines.

In the present study, we used newly developed F(11) generation mouse advanced intercross lines (AIL) to fine map Pas1-3 quantitative trait loci (QTL). The (A/J x C57BL/6) F(11) AIL mouse population was created by crossing lung tumor-resistant C57BL/6 mice with lung tumor-susceptible A/J mice. By selectively genotyping 30% of the population, we have confirmed the Pas1 QTL and narrowed it to an interval of approximately 1.0 cM or 1.3 Mb in the vicinity of the Kras2 gene. The Pas2 QTL was detected by both ANOVA and regression analysis but not by MapMaker EXP/QTL software. In addition, an interaction between the Pas1 and Pas2 QTLs was revealed. However, the Pas3 QTL was not confirmed in this study. It was either lost during the development of the AIL or too weak to be detected using AIL. The Pas1 locus is now sufficiently fine-mapped that candidate gene(s) for the Pas1 locus can be characterized by positional cloning. In this study, all 27 of the known or predicted genes located in the Pas1 candidate region were characterized as possible candidate Pas genes. Six genes were selected for additional analyses because of their relevant function in tumorigenesis or allelic changes between A/J and C57BL/6 mice. The Lrmp gene bears amino acid polymorphisms among various mouse strains that are highly correlated with the Pas1 allele status. The Pas1c1 gene (RIKEN Ak016641), encoding an intermediate filament tail domain-containing protein, produces alternatively spliced transcripts across inbred strains of mice, and its splicing pattern cosegregates with the Pas1 allele. The genetic and expression data support these two genes as strong candidates for the Pas1 locus. Of the other four genes (Eca39, RIKEN Ak015530, mHoj-1, and Krag), no functional polymorphisms or differential gene expression were found in Eca39, mHoj-1, and Krag between lung tumor-susceptible and -resistant strains. The Ak015530 carries an amino acid polymorphism, but this polymorphism does not cosegregate with mouse lung tumor susceptibility. Thus, these 4 genes are less likely candidates for the Pas1 locus.

Group A Streptococcus gene expression in humans and cynomolgus macaques with acute pharyngitis.

The molecular mechanisms used by group A Streptococcus (GAS) to survive on the host mucosal surface and cause acute pharyngitis are poorly understood. To provide new information about GAS host-pathogen interactions, we used real-time reverse transcription-PCR (RT-PCR) to analyze transcripts of 17 GAS genes in throat swab specimens taken from 18 pediatric patients with pharyngitis. The expression of known and putative virulence genes and regulatory genes (including genes in seven two-component regulatory systems) was studied. Several known and previously uncharacterized GAS virulence gene regulators were highly expressed compared to the constitutively expressed control gene proS. To examine in vivo gene transcription in a controlled setting, three cynomolgus macaques were infected with strain MGAS5005, an organism that is genetically representative of most serotype M1 strains recovered from pharyngitis and invasive disease episodes in North America and Western Europe. These three animals developed clinical signs and symptoms of GAS pharyngitis and seroconverted to several GAS extracellular proteins. Real-time RT-PCR analysis of throat swab material collected at intervals throughout a 12-day infection protocol indicated that expression profiles of a subset of GAS genes accurately reflected the profiles observed in the human pediatric patients. The results of our study demonstrate that analysis of in vivo GAS gene expression is feasible in throat swab specimens obtained from infected human and nonhuman primates. In addition, we conclude that the cynomolgus macaque is a useful nonhuman primate model for the study of molecular events contributing to acute pharyngitis caused by GAS.

Theoretical and experimental comparisons of gene expression indexes for oligonucleotide arrays.

MOTIVATION: Oligonucleotide expression arrays exhibit systematic and reproducible variation produced by the multiple distinct probes used to represent a gene. Recently, a gene expression index has been proposed that explicitly models probe effects, and provides improved fits of hybridization intensity for arrays containing perfect match (PM) and mismatch (MM) probe pairs. RESULTS: Here we use a combination of analytical arguments and empirical data to show directly that the estimates provided by model-based expression indexes are superior to those provided by commercial software. The improvement is greatest for genes in which probe effects vary substantially, and modeling the PM and MM intensities separately is superior to using the PM-MM differences. To empirically compare expression indexes, we designed a mixing experiment involving three groups of human fibroblast cells (serum starved, serum stimulated, and a 50:50 mixture of starved/stimulated), with six replicate HuGeneFL arrays in each group. Careful spiking of control genes provides evidence that 88-98% of the genes on the array are detectably transcribed, and that the model-based estimates can accurately detect the presence versus absence of a gene. The use of extensive replication from single RNA sources enables exploration of the technical variability of the array.

Linkage disequilibrium mapping of novel lung tumor susceptibility quantitative trait loci in mice.

Linkage disequilibrium (LD) has been used to map chromosomal regions regulating quantitative traits, also called quantitative trait loci (QTLs). With the increasing number of available mouse polymorphic genetic markers, LD can be estimated for the purpose of fine-mapping a given QTL or in the identification of novel QTLs. A whole-genome LD analysis was conducted for mapping mouse lung tumor susceptibility QTLs in 25 strains of mice with known susceptibility to lung cancer using 5638 genetic markers. A total of 63 markers were found to be significantly associated with lung tumor susceptibility, many of which were novel QTLs. This study demonstrates the feasibility of using LD to map QTLs on a whole genome level. Further characterization of the newly identified lung tumor susceptibility QTLs may lead to the identification of genes whose human homologue may predispose some individuals to lung cancer.