A longitudinal study of gene expression in healthy individuals
- Chris Karlovich†1Email author,
- Guillemette Duchateau-Nguyen†2,
- Andrea Johnson†3,
- Patricia McLoughlin†2,
- Mercidita Navarro4,
- Carole Fleurbaey2,
- Lori Steiner1,
- Michel Tessier2,
- Tracy Nguyen1,
- Monika Wilhelm-Seiler2 and
- John P Caulfield5
© Karlovich et al; licensee BioMed Central Ltd. 2009
Received: 20 June 2008
Accepted: 07 June 2009
Published: 07 June 2009
The use of gene expression in venous blood either as a pharmacodynamic marker in clinical trials of drugs or as a diagnostic test requires knowledge of the variability in expression over time in healthy volunteers. Here we defined a normal range of gene expression over 6 months in the blood of four cohorts of healthy men and women who were stratified by age (22–55 years and > 55 years) and gender.
Eleven immunomodulatory genes likely to play important roles in inflammatory conditions such as rheumatoid arthritis and infection in addition to four genes typically used as reference genes were examined by quantitative reverse transcription-polymerase chain reaction (qRT-PCR), as well as the full genome as represented by Affymetrix HG U133 Plus 2.0 microarrays.
Gene expression levels as assessed by qRT-PCR and microarray were relatively stable over time with ~2% of genes as measured by microarray showing intra-subject differences over time periods longer than one month. Fifteen genes varied by gender. The eleven genes examined by qRT-PCR remained within a limited dynamic range for all individuals. Specifically, for the seven most stably expressed genes (CXCL1, HMOX1, IL1RN, IL1B, IL6R, PTGS2, and TNF), 95% of all samples profiled fell within 1.5–2.5 Ct, the equivalent of a 4- to 6-fold dynamic range. Two subjects who experienced severe adverse events of cancer and anemia, had microarray gene expression profiles that were distinct from normal while subjects who experienced an infection had only slightly elevated levels of inflammatory markers.
This study defines the range and variability of gene expression in healthy men and women over a six-month period. These parameters can be used to estimate the number of subjects needed to observe significant differences from normal gene expression in clinical studies. A set of genes that varied by gender was also identified as were a set of genes with elevated expression in a subject with iron deficiency anemia and another subject being treated for lung cancer.
Gene expression profiling studies in venous blood are used to explore transcriptional differences between diseased and healthy individuals, identify biomarkers that may identify the appropriate therapy for individual patients, monitor a therapy and pharmacodynamic responses to new drugs, or define prognosis. For example, tumor molecular profiles have been identified that select preferred therapy and predict recurrence [1–3]. For autoimmune diseases, where the disease tissue is not readily accessible, peripheral blood has become a surrogate tissue by default . Venous blood contains cells that will migrate into and/or have migrated from the inflamed tissue, potentially allowing the identification of patients with disease by expression profiling of whole blood. To date, studies of rheumatoid arthritis, systemic lupus erythematosus (SLE), Crohn's disease, and multiple sclerosis (MS) patients have identified expression profiles that may be useful in diagnosis [5, 6].
Knowledge of the variability of gene expression over time in healthy men and women will help in the design of studies where drugs are expected to modulate gene expression in the diseased population. It will be particularly important to clearly establish the extent of normal variation for cases where the difference between normal and diseased gene expression profiles is subtle.
Clinical studies need to control for multiple sources of variability since gene expression can be influenced by psychological stress [7, 8], exercise [9, 10] and eating . In vitro, inflammation markers are rapidly induced in peripheral blood mononuclear cells (PBMC) by infectious agents [12–14]. In vivo gene expression profiles of individuals exposed to a bacterial toxin appeared to be different from those of healthy individuals not exposed to the toxin . Gene expression is also sensitive to the method by which the RNA from blood is isolated [16, 17].
Several previous studies have examined gene expression in the peripheral blood of normal individuals [15, 18–21]. These investigators found inter-individual variation as well as genes that uniquely identified the subjects being studied. The results were obtained using either peripheral blood mononuclear cells (PBMC) or whole blood. Gene expression profiles differ greatly among the different blood cell types [22, 23]. However, the isolation of the different cell types requires capable laboratories that are not always available at all centers involved in large clinical trials. Whole blood can be sampled reliably with the PAXgene Blood RNA System [24, 25]. Thus, the studies here were carried out using this collection method.
The gene expression studies mentioned previously were largely limited to a single platform for assays (microarray or qRT-PCR), small numbers of subjects with analyses often restricted to a short sampling time frame, or even a single time point [15, 18–21]. This study represents the largest reported study of longitudinal genomic sampling utilizing both microarray and qRT-PCR in an age and gender stratified healthy patient population, to the best of our knowledge. Moreover, the study was performed under controlled conditions with all samples drawn in the morning from fasted individuals at fixed times relative to the start of the study.
The present study extends the information base for healthy individuals by measuring the stability and dynamic range of gene expression in healthy subjects over an extended period of time, six months. qRT-PCR was used to define the range of gene expression in 28 individuals. Eleven immunomodulatory genes likely to play important roles in inflammatory conditions and four reference genes were examined. In addition, 22 subjects were assessed by high-density oligonucleotide microarray, two of whom experienced a severe adverse event during the six-month study period.
The study was conducted in accordance with good clinical practice, the Declaration of Helsinki, and appropriate regulatory guidelines. An independent ethics committee approved the protocol. Subjects gave written informed consent to participate in the study.
Blood samples were obtained from 80 healthy volunteers enrolled at a single site in Strasbourg, France. All subjects were enrolled in January and early February, 2003 and had their last visit in August. Subjects were enrolled into one of four cohorts:
Cohort 1: males 20–55 years of age
Cohort 2: males > 55 years of age
Cohort 3: females 20–55 years of age
Cohort 4: females > 55 years of age
Subjects were enrolled in the study if they were at least 20 years of age and were healthy, as determined by medical history, physical examination, and standard laboratory test values. Exclusion criteria included abnormal laboratory test results, BMI > 32, hypertension, and a positive urine pregnancy test, or positive urine HCV, HBV or HIV test. Any subject who had a concomitant disease or condition that could interfere with clinical evaluations, was an active smoker, or was taking medication on a regular basis was also excluded. Former smoking status was not captured.
A subset of 20 subjects was randomly selected from this study population for microarray analysis. The subset consisted of five subjects from each cohort. Two additional subjects who suffered from serious adverse events during the study were added subsequently. Samples from those 22 subjects were taken at five time points (Baseline, Day 14, Day 28, Day 90 and Day 180).
Quantitative RT-PCR was originally performed on all 80 subjects at each of the five time points for all analytes described in the study. An analysis of the data revealed that mean gene expression in all analytes varied significantly by time point. Due to flaws in the initial experimental design, we could not discern the degree to which the differences were attributable to a true biological effect (e.g. a seasonal effect of gene expression) or experimental bias (e.g. a "batch" effect).
The qPCR study was therefore repeated with 28 subjects chosen at random from the larger set of 80 where sufficient RNA remained. The 28-subject study was designed to distinguish between biological and experimental variability, and to minimize experimental variability. The 28 subjects were assayed by qRT-PCR at four time points (Baseline, Day 28, Day 90 and Day 180). Samples from eight subjects were analyzed on both qRT-PCR and microarray platforms.
RNA processing and quantitation
Whole blood (2.5 mL) was collected into PAXgene tubes (Becton-Dickinson Diagnostics; Hombrechtikon, Switzerland) and frozen immediately at -20°C. The tubes were shipped frozen and stored at -80°C for a period ranging from one week to six months before RNA was extracted. RNA extractions were performed in batches in the order in which the PAXgene tubes were received. Total RNA was extracted using the PAXgene 96 Blood RNA kit (Qiagen; Hilden, Germany) according to the manufacturer's instructions. RNA quality was assessed on an Agilent Bioanalyzer 2100 (Agilent Technologies; Palo Alto, CA). Any RNA sample with an RNA integrity number (RIN) < 6.5 was rejected and a replacement sample was obtained from a replicate PAXgene tube. The mean yield of RNA from all extractions with RIN ≥ 6.5 was 7.0 μg per tube (1 SD = 2.6 μg).
Quantitative RT-PCR assays used the fluorescent dye SYBR Green to monitor amplicon formation in a one step format in which the RT step was performed in the same tube as the PCR. All 384-well plates were profiled on an ABI Prism 7900 HT Sequence Detection System (Applied Biosystems Inc.; Foster City, CA). All reactions were performed with a single reagent lot. The cycling parameters were as follows: Step 1: 50°C 2 min for Uracyl-N-Glycosylase (UNG) digest followed by 95°C 1 min to destroy UNG activity; Step 2: 60°C 30 minutes for reverse transcription; Step 3: 50 Cycles of two-step PCR. Each cycle included 95°C 30 sec (denaturation) followed by 60°C 30 sec (re-annealing and extension) Step 4: Slow ramp from 60°C to 95°C to collect data for dissociation curves. UNG was used to eliminate the possibility of carryover contamination.
One well of RNA diluent was run as a negative control for each RNA specimen. RNA diluent consisted of 10 mM TRIS pH 8.0, 0.1 mM EDTA, 0.020 mg/mL poly r(A) RNA (GE Healthcare), and 0.09% sodium azide (w/v). "High input" and "low input" positive controls assessing expression of B2M were run on each plate. The target for the positive controls was human blood peripheral leukocyte total RNA (Clontech) (100 ng for high input and 1 ng for low input). After a run, the plate was inspected to see that Ct values for high input and low input controls fell within a specified range.
Samples were quantified using ribogreen, a nucleic acid stain. After the concentration was determined, the samples were diluted in RNA diluent. 2 ng of RNA were used in a 10 μl single-tube RT-PCR reaction for all assays except IL-6, where 20 ng were used. All reactions were set up in triplicate using a Biomek FX Laboratory Automation Workstation (Beckman Coulter; Fullerton, CA). The primers used for each assay and gene accession numbers are shown in Additional File 1.
Whole genome expression profiles were generated for all samples at scheduled time points (Days 1, 14, 28, 90, and 180) in the subset of 22 subjects with the exception of the Day 180 time point for subject 174 who died prior to the end of the study. A total number of 109 microarrays were hybridized. The Day 1, 90, and 180 samples were obtained in January/February, April/May, and July/August, respectively, and the mRNA was extracted in the order received (see "RNA processing and quantitation"). Additionally, the samples were processed for microarrays in two batches. The first batch was with samples up to Day 28, and the second with samples taken at Day 90 and Day 180. One microgram of total RNA from each blood sample was used to generate biotinylated cRNA, using the Agilent Low RNA Input Linear Amplification procedure (Version 2), with minor modifications (Agilent Technologies GmbH; Waldbronn, Germany); modification: biotinylated UTP and CTP (Enzo Life Sciences; Farmingdale, NY) were used for the in vitro transcription reaction, each at a final concentration of 1 mM. cRNA was purified using Promega SV 96 DNA isolation kit (automated method), as per the manufacturer's instructions (Promega Corporation, Madison, WI) or Qiagen's RNeasy purification kit (manual method) (Qiagen GmbH; Hilden, Germany); performance equivalency of the manual versus automated methods were assessed in terms of yield, quality and reproducibility with control universal human reference RNA and control whole blood RNA (data not shown). The cRNA samples were hybridized overnight to Affymetrix U133 Plus 2.0 full genome oligonucleotide arrays and then stained with Streptavidin-Phycoerythrin according to the manufacturer's instructions (Affymetrix Inc, CA, USA). Arrays were scanned using a GeneChip Scanner 3000 (Affymetrix) and signal intensities were calculated automatically by GeneChip Operating Software (GCOS, version 1.0; Affymetrix). Gene signal intensities were computed using the MAS 5.0 algorithm (component of GCOS 1.0 software). Signal intensities were normalized using a quantile-quantile method . All normalized data were log2-transformed prior to analysis to down-weight the influence of high expression values. The data have been deposited in NCBI's Gene Expression Omnibus (Edgar et al., 2002) and are accessible through GEO Series accession number GSE16028 http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE16028.
Units of cycle threshold (Ct), were used for the expression data for all analyses. Outliers were defined as data points falling ≥ 3 standard deviations (SD) outside the mean Ct value, averaged over all time points and individuals, within a gene. A graphical assessment was performed to look for possible seasonal effects.
In addition to the normalization to total RNA input that was performed, a strategy of normalization to four reference genes as a function of average gene expression weighted by their variances was considered. Several aspects of our study led us to conclude that reference gene normalization was introducing several sources of systematic variability into the data. This variability was estimated to be of appreciable size (analyses not shown) and was not present in the total RNA normalized data because it had been controlled for through study design. This variability was due to differences among reactions and differences across plates. The differences among reactions were due to the necessity of having reference genes measured in external reactions, because SYBR green assays cannot be multiplexed. The differences among plates were due to our 28-subject substudy design where a single gene was assayed per plate, which allowed us to make all analyses on within-plate comparisons only. It is known that the precision that can be achieved with reference gene normalization is limited by the variability present in the reference genes . Since we could not increase the precision of the immunomodulatory gene expression data with the reference genes, and we had alternatively controlled for appreciable amounts of systematic variability through study design, the strategy of normalization to reference genes was abandoned.
A linear mixed model with fixed effects for gender, the interaction of gender and time, and age, and a random effect for subject was applied to the qRT-PCR data. From this model, effects were tested and within-individual variation in gene expression over time was estimated. The analysis was implemented using SAS Proc Mixed.
An exploratory analysis was first performed to assess outliers and to determine if any clusters of subjects could be observed. For this purpose, a correspondence analysis was used. This analysis reduces the complexity of the data and facilitates their interpretation by finding combinations of genes that best explain the variability in the entire data set . The analysis was performed using the statistical package XlStat 6.0 (Addinsoft; New York, NY).
Identification of differentially expressed genes
Of the potential 54,675 probe sets on the U133 Plus 2.0 microarray, 34,573 probe sets were present in at least one sample (among the 109 microarrays hybridized) and were used for the analysis. A statistical analysis was performed to determine if expression profiles could indicate significant gender, age, and time effects on gene expression. A linear mixed-effect model  was built independently for each of the 34,573 selected probe sets using log2-transformed expression values and the analysis was implemented using the function lme in SPLUS.
where y is the normalized signal intensity of probe set p in individual i at time t, β gender , β age , and β time are respectively the gender, age and time effects and α is the individual random effect. The variances were denoted (between-subjects variance) for and (within-subject variance) for ε itp .
The significance of the effects was tested with t-tests. Because of the numerous t-tests performed, we needed to use a correction procedure, called the False Discovery Rate (FDR) , which controls the expected proportion of genes erroneously identified as differentially expressed genes. For this study, an FDR of 5% was chosen. For each probe set, a Shapiro test was performed to check whether the linear mixed-effect models correctly fit the data. Only 17,329 probe sets for which the Shapiro P-value > 0.1% were kept for further analysis.
Using the variance components estimated with the linear mixed-effect model built for each of the 17, 329 probe sets, we determined for each probe set p the intraclass correlation coefficient among expression measures within a subject over time by calculating the term , with the variance computed between subjects, and the within-subject variance. Values of the correlation range between 1 (i.e. no variation within subject) and 0 (i.e. high variation within subject).
We used the graphical representation proposed by Bland and Altman  to determine the level of agreement between gene expression values measured at different time points. This type of graph plots the difference between two measures as a function of their means. A good agreement between different measures is observed when the cloud of data points is located around 0 and from the absence of a trend (i.e., slope of a robust linear fit of the scatter plot close to 0). In the graphs shown the differences on the vertical axis and the means on the horizontal axis were not computed from the normalized expression values but from the effects determined by the linear mixed-effect model using time as a factor.
Subjects who experienced an adverse event had an additional sample collected following recovery from their infections. Eighteen of the eighty subjects in the study had a total of 24 infections. Infections in 7 patients were ongoing at the time of a scheduled sample collection, and infections in 11 patients ended within fourteen days before a collection. The other seven infections in these patients occurred outside these time frames and were not included in this analysis. There was a wide range of times between the scheduled blood draw and the recovery blood draw (9–51 days), with most recovery period blood draws occurring between 9–22 days. The time between the end of the infection and the recovery period blood draw ranged from 14–97 days, with most occurring between 14 to 23 days after the end of the event.
Results and Discussion
An overview of the study population and hematology measures is presented first, followed by gene expression data from both individuals who remained healthy throughout the study, and those who experienced an adverse event.
Study population characteristics
Counts of patient samples by study, gender and age group.
Main Study (n = 80)
qRT-PCR Study (n = 28)
Microarray Study (n = 22)
≤ 55 Years
> 55 Years
≤ 55 Years
> 55 Years
A subset of 28 individuals was chosen from the larger study population and analyzed for the expression of selected genes by qRT-PCR at four time points (Baseline, Day 28, Day 90 and Day 180) (Table 1). Gene expression of twenty-two subjects was assessed by high-density microarray. Samples from eight subjects were analyzed on both qRT-PCR and microarray platforms. All other samples in the two subsets were assayed on only one platform.
Hematology values of all subjects over all time points.
Mean ± SD (109/L)
Study Range (109/L)
Lab Normal Range (109/L)
Within-Subject Estimate of Variability (109/L)
6.25 ± 1.43
3.38 ± 1.07
2.16 ± 0.63
0.48 ± 0.13
0.03 ± 0.02
0.19 ± 0.12
242 ± 54
Range of expression in selected immunomodulatory and reference genes by qRT-PCR
McLoughlin et al. evaluated expression of 48 inflammation and immune-related transcripts by qRT-PCR in healthy individuals at a single time point, including 10 of the genes in the present study. They found a similarly narrow range of expression between subjects, including standard deviation values that were within 0.2 Ct of that observed here for most targets examined. Two of the targets with the highest variability in our study, MMP9 and SERPINE1, showed similarly higher variability in that study.
While PPP1CA and B2M had a relatively narrow range of expression, GAPDH and 18S rRNA were more variable among individuals than several of the immunomodulatory genes and therefore were not suitable for normalizing the data set. GAPDH is elevated in T cells upon activation [32, 34] and can fluctuate in response to the changing energy demands of cells . The results presented here support the notion that normalizing qRT-PCR data by input RNA amount is a reasonable approach and may introduce less variability than the use of a single reference gene.
Variation in gene expression over time in samples hybridized on microarrays
Within-individual variability in gene expression over time
Estimates of different components of within-individual variation (qRT-PCR study).
Variance Within Time Point (Ct 2)
Covariance Between Time Points (Ct 2)
Total Within-Individual Variance (Ct 2)
Correlation Between Time Points
Age and gender effects
No genes analyzed by either qRT-PCR or microarray were observed to be differentially regulated between the two age groups. An age effect has been observed previously with RNA profiling studies [19, 21, 39], and might have been expected here, given the large number of genes measured. One explanation is that the broad age range within a given group in this study may have prevented the detection of such an effect. Another is that blood might be less susceptible to senescence effects when compared to other tissues such as brain or muscle [40, 41].
Genes differentially expressed between genders with log-transformed signal intensities ≥ 7 (Microarray study).
Homo sapiens non-functional folate binding protein mrna, complete cds
Eukaryotic translation initiation factor 1a, x-linked
Transmembrane protein with egf-like and two follistatin-like domains 2
Eukaryotic translation initiation factor 2, subunit 3 gamma, 52 kda
Ribosomal protein s4, x-linked
Similar to dna segment, chr 11, brigham + womens genetics 0434 expressed
Eukaryotic translation elongation factor 1 alpha 1
Eukaryotic translation elongation factor 1 alpha 1
Speckle-type poz protein
Erbb2 interacting protein
Kinase interacting with leukemia-associated gene (stathmin)
High mobility group nucleosomal binding domain 4
Chromosome 10 open reading frame 45
HIV tat specific factor 1
Guanine nucleotide binding protein (g protein), gamma 2
High mobility group nucleosomal binding domain 4
High mobility group nucleosomal binding domain 4
Effect of two severe adverse events on gene expression
Subject 121 had anemia secondary to rectal bleeding with a progressive drop in red cell count from baseline to Day 90. The subject's bleeding was stopped surgically on Day 100; on Day 180 red cell parameters had returned to the normal range. The correspondence analysis data points on Days 1, 14, and 28 are outside of the main cluster with the Day 90 profile reaching a maximum distance from the others. The Day 180 point has returned to normal. RBC-associated genes with elevated expression levels included ferrochetalase, carbonic anhydrase, ALAS2, erythrocyte membrane protein band 4.2, glycophorin A and B, and 2,3-bisphosphoglycerate mutase. Some of the genes over-expressed in subject 121, were correlated previously to reticulocyte expressed genes: SLC4A1, EPP42, BCL2L1 and BNIp3L . ALAS2 and carbonic anhydrase have also been identified as genes that are down-regulated in anemia during acute renal allograft rejection .
Subject 174 developed lung cancer and began chemotherapy on approximately Day 70. She was a strong outlier in the correspondence analysis on Day 90. Her monocyte count on that day was severely depressed to 6.5% of the baseline value (from 0.46 × 109/L to 0.03 × 109/L). All other white cell types were within normal limits on Days 1, 14, 28, and 90. For subject 174, 31 genes were found to be up-regulated. The gene ACRBP was elevated four- fold above the other subjects on Day 90. ACRBP is a member of the cancer/testis family of antigens, is immunogenic, and has been detected in different tumor types . The growth arrest-specific (GAS)2-like 1 gene, an actin-associated protein expressed at high levels in growth-arrested cells , was found to be 1.5-fold higher in subject 174, compared to other individuals at Day 90. The changes in gene expression observed in this subject correlated with the onset of chemotherapy and may be largely attributable to treatment. However, for both these cases data is required in larger numbers of patients to draw firm conclusions.
Effect of infection on gene expression
In this study with 80 healthy male and female subjects ≥ 20 years of age, a normal range of gene expression was established for 11 inflammation-related genes and four housekeeping genes. Each gene was stably expressed, independent of age and gender. There was no apparent correlation between values more than three standard deviations from the mean with adverse events or hematology values. Nineteen subjects who had infections at the time of or within two weeks before their scheduled blood draws gave an additional sample two weeks after the resolution of their infections. The samples collected during the active infection showed slight elevations from the expression levels seen in uninfected subjects. However, the samples taken two weeks later showed increased expression of TGFβ1.
Gene expression levels, as measured by microarray analysis, appeared to be constant over one month; however, over three months, a small percentage of genes appeared to vary. We observed that intra-individual gene correlations differ greatly depending on signal intensity. A small proportion of genes were found to be differentially regulated according to gender. Differential gene regulation by age (in subjects 25–55 years of age versus subjects > 55 years of age) was not observed. Elevated expression levels of red blood cell-associated genes were observed in one subject who experienced progressive anemia secondary to blood loss. Included among those genes were ferrochetalase, carbonic anhydrase, ALAS2, erythrocyte membrane protein band 4.2, glycophorin A and B, and 2,3-bisphosphoglycerate mutase.
Use of RNA expression studies in clinical trials
Estimates of sample sizes required to achieve a certain power to detect changes in gene expression of a particular magnitude can be calculated from the estimates of variance components given in the results section, results from other studies, and the estimated effect size from in vitro work.
Samples should be analyzed at the end of the experiment rather than on an ongoing basis. A bias was introduced when samples from the present study were processed and analyzed with different reagent lots over time in sequential batches This type of bias, also recently described by Yang et al. , could lead to the identification of many artifacts among differentially expressed gene sets.
Small elevations in inflammatory gene expression produced during upper respiratory infections generally return to normal levels of expression within two weeks after the infection has resolved.
Time and/or seasonal effects may be a factor in trials lasting longer than one month.
Gender and age effects are not likely to be problematic in populations from 20 to 65 years of age. The assessment of gene expression profiles in children or very old individuals might lead to different conclusions.
We thank Emanuelle Erhart-Gaudel and the staff of the Roche Clinical Pharmacology Unit in Strasbourg, France, for conducting the clinical trial, David Trollinger of Source Precision Medicine for discussions regarding the genes chosen for qRT-PCR, Xiaoning Wu and Sean Chien for providing technical support for the qRT-PCR experiments, and Laurent Essioux and Brian Rhees for comments regarding the statistical analyses. The study was funded by the Pharmaceutical and Diagnostic Divisions of Hoffman-LaRoche. All authors read and approved the final manuscript.
- Perou CM, Sorlie T, Eisen MB, Rijn van de M, Jeffrey SS, Rees CA, Pollack JR, Ross DT, Johnsen H, Akslen LA, et al: Molecular portraits of human breast tumours. Nature. 2000, 406: 747-752. 10.1038/35021093.View ArticlePubMedGoogle Scholar
- van't Veer LJ, Dai H, Vijver van de MJ, He YD, Hart AA, Mao M, Peterse HL, Kooy van der K, Marton MJ, Witteveen AT, et al: Gene expression profiling predicts clinical outcome of breast cancer. Nature. 2002, 415: 530-536. 10.1038/415530a.View ArticleGoogle Scholar
- Paik S, Shak S, Tang G, Kim C, Baker J, Cronin M, Baehner FL, Walker MG, Watson D, Park T, et al: A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. N Engl J Med. 2004, 351: 2817-2826. 10.1056/NEJMoa041588.View ArticlePubMedGoogle Scholar
- Burczynski ME, Dorner AJ: Transcriptional profiling of peripheral blood cells in clinical pharmacogenomic studies. Pharmacogenomics. 2006, 7: 187-202. 10.2217/146224188.8.131.52.View ArticlePubMedGoogle Scholar
- Aune TM, Maas K, Parker J, Moore JH, Olsen NJ: Profiles of gene expression in human autoimmune disease. Cell Biochem Biophys. 2004, 40: 81-96. 10.1385/CBB:40:2:081.View ArticlePubMedGoogle Scholar
- Baechler EC, Batliwalla FM, Reed AM, Peterson EJ, Gaffney PM, Moser KL, Gregersen PK, Behrens TW: Gene expression profiling in human autoimmunity. Immunol Rev. 2006, 210: 120-137. 10.1111/j.0105-2896.2006.00367.x.View ArticlePubMedGoogle Scholar
- Bierhaus A, Wolf J, Andrassy M, Rohleder N, Humpert PM, Petrov D, Ferstl R, von Eynatten M, Wendt T, Rudofsky G, et al: A mechanism converting psychosocial stress into mononuclear cell activation. Proc Natl Acad Sci USA. 2003, 100: 1920-1925. 10.1073/pnas.0438019100.View ArticlePubMedPubMed CentralGoogle Scholar
- Nagabhushan M, Mathews HL, Witek-Janusek L: Aberrant nuclear expression of AP-1 and NFkappaB in lymphocytes of women stressed by the experience of breast biopsy. Brain Behav Immun. 2001, 15: 78-84. 10.1006/brbi.2000.0589.View ArticlePubMedGoogle Scholar
- Petersen AM, Pedersen BK: The anti-inflammatory effect of exercise. J Appl Physiol. 2005, 98: 1154-1162. 10.1152/japplphysiol.00164.2004.View ArticlePubMedGoogle Scholar
- Sureda A, Ferrer MD, Tauler P, Maestre I, Aguilo A, Cordova A, Tur JA, Roche E, Pons A: Intense physical activity enhances neutrophil antioxidant enzyme gene expression. Immunocytochemistry evidence for catalase secretion. Free Radic Res. 2007, 41: 874-883. 10.1080/10715760701416459.View ArticlePubMedGoogle Scholar
- van Erk MJ, Blom WA, van Ommen B, Hendriks HF: High-protein and high-carbohydrate breakfasts differentially change the transcriptome of human blood cells. Am J Clin Nutr. 2006, 84: 1233-1241.PubMedGoogle Scholar
- Boldrick JC, Alizadeh AA, Diehn M, Dudoit S, Liu CL, Belcher CE, Botstein D, Staudt LM, Brown PO, Relman DA: Stereotyped and specific gene expression programs in human innate immune responses to bacteria. Proc Natl Acad Sci USA. 2002, 99: 972-977. 10.1073/pnas.231625398.View ArticlePubMedPubMed CentralGoogle Scholar
- Spurrell JC, Wiehler S, Zaheer RS, Sanders SP, Proud D: Human airway epithelial cells produce IP-10 (CXCL10) in vitro and in vivo upon rhinovirus infection. Am J Physiol Lung Cell Mol Physiol. 2005, 289: L85-95. 10.1152/ajplung.00397.2004.View ArticlePubMedGoogle Scholar
- Stockl J, Vetr H, Majdic O, Zlabinger G, Kuechler E, Knapp W: Human major group rhinoviruses downmodulate the accessory function of monocytes by inducing IL-10. J Clin Invest. 1999, 104: 957-965. 10.1172/JCI7255.View ArticlePubMedPubMed CentralGoogle Scholar
- McLoughlin K, Turteltaub K, Bankaitis-Davis D, Gerren R, Siconolfi L, Storm K, Cheronis J, Trollinger D, Macejak D, Tryon V, Bevilacqua M: Limited dynamic range of immune response gene expression observed in healthy blood donors using RT-PCR. Mol Med. 2006, 12: 185-195. 10.2119/2006-00018.McLoughlin.View ArticlePubMedPubMed CentralGoogle Scholar
- Baechler EC, Batliwalla FM, Karypis G, Gaffney PM, Moser K, Ortmann WA, Espe KJ, Balasubramanian S, Hughes KM, Chan JP, et al: Expression levels for many genes in human peripheral blood cells are highly sensitive to ex vivo incubation. Genes Immun. 2004, 5: 347-353. 10.1038/sj.gene.6364098.View ArticlePubMedGoogle Scholar
- Hartel C, Bein G, Muller-Steinhardt M, Kluter H: Ex vivo induction of cytokine mRNA expression in human blood samples. J Immunol Methods. 2001, 249: 63-71. 10.1016/S0022-1759(00)00334-3.View ArticlePubMedGoogle Scholar
- Campbell C, Vernon SD, Karem KL, Nisenbaum R, Unger ER: Assessment of normal variability in peripheral blood gene expression. Dis Markers. 2002, 18: 201-206.View ArticlePubMedGoogle Scholar
- Eady JJ, Wortley GM, Wormstone YM, Hughes JC, Astley SB, Foxall RJ, Doleman JF, Elliott RM: Variation in gene expression profiles of peripheral blood mononuclear cells from healthy volunteers. Physiol Genomics. 2005, 22: 402-411. 10.1152/physiolgenomics.00080.2005.View ArticlePubMedGoogle Scholar
- Radich JP, Mao M, Stepaniants S, Biery M, Castle J, Ward T, Schimmack G, Kobayashi S, Carleton M, Lampe J, Linsley PS: Individual-specific variation of gene expression in peripheral blood leukocytes. Genomics. 2004, 83: 980-988. 10.1016/j.ygeno.2003.12.013.View ArticlePubMedGoogle Scholar
- Whitney AR, Diehn M, Popper SJ, Alizadeh AA, Boldrick JC, Relman DA, Brown PO: Individuality and variation in gene expression patterns in human blood. Proc Natl Acad Sci USA. 2003, 100: 1896-1901. 10.1073/pnas.252784499.View ArticlePubMedPubMed CentralGoogle Scholar
- Cobb JP, Mindrinos MN, Miller-Graziano C, Calvano SE, Baker HV, Xiao W, Laudanski K, Brownstein BH, Elson CM, Hayden DL, et al: Application of genome-wide expression analysis to human health and disease. Proc Natl Acad Sci USA. 2005, 102: 4801-4806. 10.1073/pnas.0409768102.View ArticlePubMedPubMed CentralGoogle Scholar
- Palmer C, Diehn M, Alizadeh AA, Brown PO: Cell-type specific gene expression profiles of leukocytes in human peripheral blood. BMC Genomics. 2006, 7: 115-10.1186/1471-2164-7-115.View ArticlePubMedPubMed CentralGoogle Scholar
- Fan H, Hegde PS: The transcriptome in blood: challenges and solutions for robust expression profiling. Curr Mol Med. 2005, 5: 3-10. 10.2174/1566524053152861.View ArticlePubMedGoogle Scholar
- Kim SJ, Dix DJ, Thompson KE, Murrell RN, Schmid JE, Gallagher JE, Rockett JC: Effects of storage, RNA extraction, genechip type, and donor sex on gene expression profiling of human whole blood. Clin Chem. 2007, 53: 1038-1045. 10.1373/clinchem.2006.078436.View ArticlePubMedGoogle Scholar
- Bolstad BM, Irizarry RA, Astrand M, Speed TP: A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics. 2003, 19: 185-193. 10.1093/bioinformatics/19.2.185.View ArticlePubMedGoogle Scholar
- Huggett J, Dheda K, Bustin S, Zumla A: Real-time RT-PCR normalisation; strategies and considerations. Genes Immun. 2005, 6: 279-284. 10.1038/sj.gene.6364190.View ArticlePubMedGoogle Scholar
- Fellenberg K, Hauser NC, Brors B, Neutzner A, Hoheisel JD, Vingron M: Correspondence analysis applied to microarray data. Proc Natl Acad Sci USA. 2001, 98: 10781-10786. 10.1073/pnas.181597298.View ArticlePubMedPubMed CentralGoogle Scholar
- Pinheiro JC, Bates DM: Mixed-effects models in S and S-PLUS. 2000, New York: SpringerView ArticleGoogle Scholar
- Benjamini Y, Hochberg Y: Controlling the False Discovery Rate – a Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society Series B-Methodological. 1995, 57: 289-300.Google Scholar
- Bland JM, Altman DG: Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986, 1: 307-310.View ArticlePubMedGoogle Scholar
- Bas A, Forsberg G, Hammarstrom S, Hammarstrom ML: Utility of the housekeeping genes 18S rRNA, beta-actin and glyceraldehyde-3-phosphate-dehydrogenase for normalization in real-time quantitative reverse transcriptase-polymerase chain reaction analysis of gene expression in human T lymphocytes. Scand J Immunol. 2004, 59: 566-573. 10.1111/j.0300-9475.2004.01440.x.View ArticlePubMedGoogle Scholar
- Silver N, Best S, Jiang J, Thein SL: Selection of housekeeping genes for gene expression studies in human reticulocytes using real-time PCR. BMC Mol Biol. 2006, 7: 33-10.1186/1471-2199-7-33.View ArticlePubMedPubMed CentralGoogle Scholar
- Radonic A, Thulke S, Mackay IM, Landt O, Siegert W, Nitsche A: Guideline to reference gene selection for quantitative real-time PCR. Biochem Biophys Res Commun. 2004, 313: 856-862. 10.1016/j.bbrc.2003.11.177.View ArticlePubMedGoogle Scholar
- Barber RD, Harmer DW, Coleman RA, Clark BJ: GAPDH as a housekeeping gene: analysis of GAPDH mRNA expression in a panel of 72 human tissues. Physiol Genomics. 2005, 21: 389-395. 10.1152/physiolgenomics.00025.2005.View ArticlePubMedGoogle Scholar
- Li X, Kim J, Zhou J, Gu W, Quigg R: Use of signal thresholds to determine significant changes in microarray data analyses. Genet Mol Biol. 2005, 28: 191-200.View ArticleGoogle Scholar
- Dobbin KK, Beer DG, Meyerson M, Yeatman TJ, Gerald WL, Jacobson JW, Conley B, Buetow KH, Heiskanen M, Simon RM, et al: Interlaboratory comparability study of cancer gene expression analysis using oligonucleotide microarrays. Clin Cancer Res. 2005, 11: 565-572.PubMedGoogle Scholar
- Zhang J, Finney RP, Clifford RJ, Derr LK, Buetow KH: Detecting false expression signals in high-density oligonucleotide arrays by an in silico approach. Genomics. 2005, 85: 297-308. 10.1016/j.ygeno.2004.11.004.View ArticlePubMedGoogle Scholar
- Tang Y, Lu A, Ran R, Aronow BJ, Schorry EK, Hopkin RJ, Gilbert DL, Glauser TA, Hershey AD, Richtand NW, et al: Human blood genomics: distinct profiles for gender, age and neurofibromatosis type 1. Brain Res Mol Brain Res. 2004, 132: 155-167. 10.1016/j.molbrainres.2003.10.014.View ArticlePubMedGoogle Scholar
- Welle S, Brooks AI, Delehanty JM, Needler N, Thornton CA: Gene expression profile of aging in human muscle. Physiol Genomics. 2003, 14: 149-159.View ArticlePubMedGoogle Scholar
- Galvin JE, Ginsberg SD: Expression profiling in the aging brain: a perspective. Ageing Res Rev. 2005, 4: 529-547. 10.1016/j.arr.2005.06.009.View ArticlePubMedGoogle Scholar
- Fisher EM, Beer-Romero P, Brown LG, Ridley A, McNeil JA, Lawrence JB, Willard HF, Bieber FR, Page DC: Homologous ribosomal protein genes on the human X and Y chromosomes: escape from X inactivation and possible implications for Turner syndrome. Cell. 1990, 63: 1205-1218. 10.1016/0092-8674(90)90416-C.View ArticlePubMedGoogle Scholar
- Punyadeera C, Dassen H, Klomp J, Dunselman G, Kamps R, Dijcks F, Ederveen A, de Goeij A, Groothuis P: Oestrogen-modulated gene expression in the human endometrium. Cell Mol Life Sci. 2005, 62: 239-250. 10.1007/s00018-004-4435-y.View ArticlePubMedGoogle Scholar
- Chua MS, Barry C, Chen X, Salvatierra O, Sarwal MM: Molecular profiling of anemia in acute renal allograft rejection using DNA microarrays. Am J Transplant. 2003, 3: 17-22. 10.1034/j.1600-6143.2003.30104.x.View ArticlePubMedGoogle Scholar
- Ono T, Kurashige T, Harada N, Noguchi Y, Saika T, Niikawa N, Aoe M, Nakamura S, Higashi T, Hiraki A, et al: Identification of proacrosin binding protein sp32 precursor as a human cancer/testis antigen. Proc Natl Acad Sci USA. 2001, 98: 3282-3287. 10.1073/pnas.041625098.View ArticlePubMedPubMed CentralGoogle Scholar
- Schneider C, King RM, Philipson L: Genes specifically expressed at growth arrest of mammalian cells. Cell. 1988, 54: 787-793. 10.1016/S0092-8674(88)91065-3.View ArticlePubMedGoogle Scholar
- Yang H, Harrington CA, Vartanian K, Coldren CD, Hall R, Churchill GA: Randomization in laboratory procedure is key to obtaining reproducible microarray results. PLoS ONE. 2008, 3: e3724-10.1371/journal.pone.0003724.View ArticlePubMedPubMed CentralGoogle Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1755-8794/2/33/prepub
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.