Institute of Food Research, Norwich Research Park, Colney, Norwich, United Kingdom
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
interindividual variation; microarray; transcriptomic
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Such information is essential to provide a basis for a comprehensive understanding of normal tissue functions and the impact of genetic and environmental factors on these. These data are also needed for the development of more robust designs for clinical and nutritional intervention studies involving complex gene expression analysis.
In the study described here, microarrays were used to analyze gene transcription profiles in peripheral blood mononuclear cells (PBMC) isolated from 18 (10 female and 8 male) apparently healthy human volunteers. Each volunteer provided a series of five blood samples taken first thing in the morning, after an overnight fast, at weekly intervals, with each sample being taken sequentially on a different day of the normal working week.
PBMC from venous blood samples were selected for analysis because blood is an accessible tissue that is used in many clinical and nutritional studies. Peripheral blood leukocytes are often used as surrogates for predicting potential effects in inaccessible tissues. Their validity in this respect is likely to be highly dependent on the context of the research, and thorough evaluation is still required to further validate their use. Definition of the normal degree of variation, as well as the responsiveness of gene expression in these cells to environmental factors, will provide a key component in this process.
Whole blood is an example of a complex tissue that contains a number of different cell types. Each of these cell types expresses a unique pattern of gene transcription relating to its specific function. Whitney et al. (16) have demonstrated a clear correlation between the expression of specific groups of genes in whole blood RNA isolates and the representation of cell types (erythrocytes and leukocyte subsets) in these samples. In the current study, this source of variation was reduced by using PBMC instead of whole blood. PBMC were isolated immediately after collection of blood using a protocol designed to minimize sample processing time and maximize the consistency of processing between samples. The content of different leukocyte subsets in the PBMC preparations was determined by flow cytometric analysis. Total RNA from each PBMC sample was analyzed in duplicate on oligonucleotide-based arrays produced in-house (each containing features representing 13,927 genes). Of the genes represented on these arrays, reliable signal (in at least 95% of the samples) was obtained for 8,489 genes. The transcript profiles of these genes were examined for apparent effects of sex, age, body mass index (BMI), and the presence of varying proportions of different leukocyte subsets within the PBMC preparations. The variation in transcription levels for most genes across sample sets from individuals was comparatively small, with average within-subject coefficients of variation <20% for 80% of the genes examined. However, a high proportion of genes (39 or 58% depending on the method of analysis) exhibited significant interindividual differences in transcript levels. As might be expected, it proved possible to distinguish perfectly between samples from each donor by applying unsupervised hierarchical clustering, using subsets of these genes (431 down to 137) defined by statistical comparisons of increasing stringency between volunteers. The genes present in these sets were observed to cover a wide range of biological functions (based on gene ontology annotation) including notable clusters of genes involved in immune function, interferon-regulated genes, histones, and X- or Y-linked genes.
These data extend the currently very limited information base that specifically describes the degree of normal variation in the expression of genes in healthy human tissues. The results suggest that genetic and environmental factors elicit substantial interindividual differences in the expression of many genes, but also that the overall expression profile in PBMC is comparatively stable within an individual over time.
![]() |
MATERIALS AND METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Volunteers and blood samples.
Approval for the study was obtained from both the Institute of Food Research Human Research Governance Committee and the Norwich Local Research Ethics Committee. Male and female volunteers (aged between 20 and 45 yr) were recruited and included on the study if they satisfied the following inclusion criteria: were nonsmokers (and had not smoked for at least 6 mo before the study), were not taking regular prescribed medication (including hormone replacement therapy and oral contraception), were not pregnant and had not been pregnant during the 12 mo before the study, had not received inoculations within 4 mo of starting the study or planned to during the study, had not donated or intended to donate blood within 16 wk of the first and last study samples, were not diagnosed with any long-term medical condition (e.g., diabetes, hemophilia, cardiovascular disease, glaucoma, anemia, gastrointestinal disease), and were not experiencing symptoms of allergy (individuals with allergy controlled by avoidance were accepted provided they were not currently experiencing symptoms). Written informed consent was obtained from each volunteer before commencing the study.
On their first visit, the volunteers' heights and weights were measured, their blood pressures and pulses monitored, and urine and fasting venous blood samples taken for the purpose of general health screening. Volunteers were excluded from further participation in the study according to the following criteria: BMI was <17 or >31 kg/m2; blood pressure was <100/50 or >140/90 mmHg; pulse exceeded 90 beats/min; analysis of urine revealed pH, protein, glucose, ketones, urobilinogen, bilirubin, leukocytes, nitrites, or blood/hemoglobin levels outside normal ranges; or fasting blood glucose, hemoglobin, full blood cell count, or urea and electrolyte concentrations fell outside the normal range and were considered, on review by a consultant physician, to be sufficient cause for exclusion.
Volunteers who met the inclusion criteria and were not excluded for one or more of the reasons above commenced a series of five visits at 1-wk intervals to the Institute's Human Nutrition Unit to provide overnight-fasted venous blood samples. The visits were timed such that each of the samples from one volunteer was collected on sequential mornings of the week (Monday to Friday). Ten female and eight male volunteers completed the study. All five samples were collected from 16 of the volunteers, but one sample was missed out by each of the 2 remaining volunteers (both male). On each visit, the volunteers were asked whether they had been unwell during the previous week, and any illness that they reported or over-the-counter medication they had taken was recorded. All samples were anonymized.
Sample processing.
Blood samples (60 ml) were drawn from an antecubital vein in each volunteer's forearm. Each subsequent step involved in the processing for the isolation of PBMC RNA was performed without delay after collection of the blood. The majority of each sample (56 ml) was transferred to a set of seven BD Vacutainer CPT tubes (8-ml draw volume each), containing sodium citrate as anti-coagulant, for isolation of PBMC according to the manufacturer's instructions and as detailed below. The remaining portion (4 ml) of each blood sample was transferred into an EDTA anti-coagulant tube for isolation and storage (at 80°C) of plasma after centrifugation at 2,000 g for 20 min at 20°C.
After centrifugation for 20 min at 1,500 g, 20°C, the unopened BD Vacutainer CPT tubes were inverted gently five times to mix the plasma plus mononuclear cell layer present above the polyester gel. The entire content of each tube above the gel layer was decanted into fresh tubes, with all the preparations originating from a single blood sample being pooled at this time. Two portions of the cell suspension (5 and 0.7 ml) were removed for PBMC protein extraction and for flow cytometric analysis (described below). The remaining cell suspension was diluted with 1 vol of physiological saline and then centrifuged at 300 g, 20°C, for 15 min. The liquid was decanted, and total RNA was extracted from the cell pellet using RNeasy midikits according to the manufacturer's instructions.
Flow cytometric analysis of PBMC preparations.
The 0.7-ml portion of cell suspension from each sample was collected and washed twice with 1 ml of saline (0.9% wt/vol) containing 0.15% (wt/vol) EDTA (SE buffer). Cells were recovered after each wash by centrifugation at 300 g. Cell pellets were resuspended in 0.5 ml of SE buffer, and aliquots were removed for immunofluorescent antibody staining.
Fluorescently labeled antibodies against CD3, CD14, and CD19 were used to detect T cells, B cells, and monocytes, respectively, and a combination of antibodies against CD16 and CD56 was used detect natural killer cells. Antibody titrations were performed to determine optimal dilutions for each antibody used. Equal volumes (50 µl of each) of freshly prepared antibody cocktail and cell suspension (2 x 106 cells/ml) were mixed together and incubated for 15 min at room temperature in the dark. Stained cells were washed twice with phosphate-buffered saline containing 1% (wt/vol) bovine serum albumin, 1% (wt/vol) sodium azide, and 0.15% (wt/vol) EDTA. Resuspended cells were fixed with 1% (wt/vol; final concentration) paraformaldehyde and subsequently stored for up to 7 days at 4°C until flow cytometric analysis and data acquisition. Fluorescence data were acquired on a six-parameter Beckman Coulter Altra flow cytometer equipped with a 488-nm argon ion and 635-nm HeNe lasers and analyzed using Expo32 software (Beckman Coulter, High Wycombe, UK).
Microarray production and analysis.
Microarrays were produced using a commercial set of 14,001 (7072mers) oligonucleotides (Operon human oligonucleotide set version 1.0; Operon Biotechnologies Technologies, Cologne, Germany), which consisted of 13,971 gene-specific oligonucleotides, 29 negative controls, and 1 positive control (an equimolar mixture of the 13,971 gene-specific oligonucleotides). A set of additional controls was added, including six yeast gene-specific oligonucleotides and Cy3- and Cy5-labeled PCR products for an Escherichia coli gene (FixC) that were used as "landing lights" to confirm the correct printing order and orientation of oligonucleotides on the arrays. The oligonucleotides were printed onto epoxy-coated slides using a Stanford design arraying robot (14). Before hybridization, the printed arrays were postprocessed according to the slide manufacturer's instructions.
The yield, purity, and integrity of each RNA sample was determined by spectrophotometric analysis at 260 and 280 nm and by use of an Agilent 2100 bioanalyzer with RNA 6000 Nano chips (Agilent Technologies, South Queensferry, UK) and RNA 6000 ladder (Ambion Europe, Huntingdon, UK). The absorbance ratio at 260 and 280 nm (A260/A280) for all samples was between 1.8 and 2.2. The 28S/18S rRNA ratios were all between 1.5 and 2.1. The average RNA yield was 97 ± 34 µg (range 41175 µg).
Cy5-labeled cDNA was produced by reverse transcription of test RNA samples (20 µg each), using an established indirect labeling protocol with minor modifications (6). Cy3-labeled reference cDNA was produced from the Universal Human Reference RNA, prepared according to the manufacturer's instructions, in the same manner as the PBMC samples.
Each test sample, together with a reference sample, was hybridized to two arrays on separate occasions. Initially, 10 batches of hybridizations (each consisting of 1618 arrays) were set up. To avoid any potential for introducing experimental bias, the samples were randomly allocated to the different batches. The randomization was performed with two restraints; each batch contained no more than one sample originating from any given volunteer and either three or four samples numbered 15 (where sample 1 was the 1st sample obtained from a volunteer and sample 5 the last). An 11th batch of arrays was set up to accommodate repeat analyses of samples for which uneven hybridization or background signal meant that a large number of features had to be rejected during image analysis of one of the first two technical replicates. In all, 184 arrays were included in the final analysis, representing duplicate arrays for 70, triplicate arrays for 13, and single arrays for 5 samples.
Hybridizations were performed overnight at 70°C according to an established protocol (7). After hybridization, the slides were washed in 1x saline sodium citrate solution (SSC), 0.2% (wt/vol) SDS, at 65°C with constant agitation to remove the coverslips. The slides were then transferred to fresh 1x SSC, 0.2% (wt/vol) SDS, and incubated at 65°C with constant mixing for 15 min. The slides were washed in 1x SSC twice for 5 min each, followed by three 5-min washes in 0.2x SSC. Finally, they were dried by centrifugation at 250 g for 10 min at room temperature. The arrays were scanned using an Agilent G2565BA microarray scanner system (Agilent Technologies). The array data and corresponding sample information (volunteer no., sex, BMI, age range, and percentages of different leukocyte subsets) were made available in MIAME-compliant format by submission to the ArrayExpress database (http://www.ebi.ac.uk/arrayexpress) with array design accession number A-MEXP-170 and experiment accession number E-TABM-7.
Array image processing, data manipulation, and normalization.
The array images were analyzed with Agilent G2567AA Feature Extraction Software version 7.1 (Agilent Technologies), using the "wholespot" method with minimum feature signal background subtraction. Bad features, identified by visual inspection, were removed by manual flagging. The final data were exported into Excel, and additional unreliable data points were removed if they met any of the following criteria: saturation of any pixels within the feature in either channel, feature size <25 pixels, feature not found in either channel, or background-corrected signal intensity <2.6 x SD of background in both channels. Where the background-corrected signal was >2.6 x SD of background signal in only one channel, a surrogate value (equal to the SD of the background) was used for the other channel.
The data were analyzed using two approaches in different software packages. GeneSpring 7.2 (Agilent Technologies) was employed to analyze the effects of discrete variables (i.e., sex and volunteer) and for subsequent hierarchical clustering. R (http://www.R-project.org) was used to enable analysis of the effects of continuous variables (e.g., age, BMI, and percentages of different leukocyte subsets within samples). For the GeneSpring analysis, the edited raw data were imported and then normalized in a multistage process. First the Cy3 and Cy5 data for samples in batch 7 were reversed to account for inadvertent dye swap performed during sample preparation in this batch. Second, the data for each array underwent regional (by print pin) LoWess normalization. Finally, to account for any batch-specific effects, the normalized Cy5/Cy3 ratio for each gene on each array was normalized against the mean ratio for that gene calculated from all arrays within the corresponding batch.
Within R the raw data set was normalized initially using the same two first steps as for GeneSpring. The LoWess-normalized data were finally subjected to Quantile normalization using the Limma package (http://bioinf.wehi.edu.au/limma) to ensure consistent empirical distribution of intensities across arrays and across channels (17).
Statistical analysis.
The cross-gene error model within GeneSpring was used to further filter the data to remove any features for which the signal intensity was too low to be considered reliable (12). The average value of base/proportional error for the entire set of arrays was calculated using this model (based on deviation from 1). Only genes for which the normalized control (Cy3) signal intensity was greater than this value in at least 95% of samples were used in subsequent analyses.
Within GeneSpring normalized data were analyzed using Welch's t-test to investigate effects of gender on expression of individual genes. Welch's one-way ANOVA was applied to investigate volunteer-specific effects. In each case, multiple-test corrections (Benjamini and Hochberg false discovery rate or Holm) were applied. To provide a basis for examining intraindividual variation, coefficients of variation (CVs) were calculated for each gene for each volunteer's set of five samples using the averaged normalized ratios of the technical replicates. The overall average of the CVs for each gene was calculated for the sample sets from all 18 volunteers. One- and two-dimensional hierarchical clustering of selected gene lists was performed using standard correlations.
The same genes that passed the GeneSpring cross-gene error model filter, described above, were used for statistical analysis within R. The lmFit function from the Limma package was used to fit a linear model for each gene in the series of arrays, with the log ratio of gene expression as the dependent variable and independent variables sex, age, and BMI, as well as percentages of different leukocyte subsets in the samples, as determined by flow cytometry. When leukocyte subsets were included in the model, only the array data sets for which corresponding complete flow cytometric data had been obtained were included in the analysis (157 array data sets). Hybridization batch number was also incorporated into all the models to account for possible batch-specific effects. Moderated t-statistics and log odds of differential expression were created by empirical Bayes shrinkage of the standard errors toward a common value, using the function eBayes (also from Limma).
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Data normalization.
A first round of data analysis within GeneSpring, in which the data were not normalized by batch, revealed significant batch-to-batch differences (data not shown). Batchwise normalization of the data was achieved by dividing the ratio for each gene by the mean of the ratios for that gene in that batch. No significant effects of batch were evident following this additional normalization step (i.e., there were no significant differences using Welch's ANOVA with Benjamini and Hochberg multiple-test correction with a false discovery rate of 5%).
Of the genes represented on the arrays, reliable signal (determined according to the criteria set out in MATERIALS AND METHODS) was obtained for 8,489 genes in at least 95% of samples. The expression profiles of these genes were examined for effects of sex, age, and BMI; for the presence of varying proportions of different leukocyte subsets; and for patterns specific to individual volunteers.
Effects of different cell types.
T cells were the most abundant leukocyte subset within PBMC preparations, accounting for 59.4 ± 7.5% of all cells (mean ± SD). It was assumed, therefore, that these cells would provide the predominant features of the PBMC sample transcriptional profiles, with additional contributions from other cell subsets varying in a manner depending on their relative abundance in each sample. On this basis, the effects of the presence of varying amounts of other cell types present (i.e., B cells, granulocytes, natural killer cells, and monocytes) on the sample transcription profiles were analyzed. B cells and granulocytes were the least abundant cells in the PBMC preparations at 5.3 ± 2.0 and 0.1 ± 0.2% of the total cell count, respectively. Natural killer cells and monocytes were present as 12.3 ± 6.3 and 21.2 ± 6.3%, respectively, of the total cells within the PBMC samples. There were no detectable effects on the expression of any genes due to variations in the percentages of B cells or granulocytes. The expression of 28 genes correlated significantly with the abundance of natural killer cells (Table 1). Among these were four killer cell immunoglobulin-like receptor genes (KIR2DL2, KIR2DL3, KIR3DL2, and KIR2DS3) and two killer cell lectin-like receptors (KLRF1 and KLRC2), all of which correlated positively with the percentage of natural killer cells. The expression of 67 genes correlated significantly with the percentage of monocytes present (the 50 genes with the lowest P values are shown in Table 2). These included genes involved in diverse cellular functions such as immune, inflammatory, and defense responses; transcriptional regulation; and cell proliferation and development.
|
|
|
Effects of age.
When a linear regression model incorporating sex, age, and BMI as independent variables was employed, 31 genes were identified that exhibited apparent age-dependent expression. When the full model including leukocyte cell subsets was used, the number decreased to 24 (Table 4). Predominant among these age-dependent genes were immunoglobulins (12 of 24), transcript levels of all of which decreased significantly with age. These included different rearrangements of the variable regions for both heavy and light chains plus the immunoglobulin J chain. There was no significant correlation between B-cell number in the PBMC preparations and age.
|
|
|
|
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
To our knowledge, only two studies have been published that relate directly to the one described here (10, 16). Radich et al. (10) used a similar study design but focused on gene transcript profiles in mixed leukocyte populations from whole blood samples. Whitney et al. (16) performed a study that included analysis of expression profiles, both in mixed leukocytes from whole blood and PBMC from healthy volunteers.
Overall, there were some notable common features between the results obtained from the current study and those described by Whitney et al. (16) and Radich et al. (10). For example, a number of major histocompatibility class II genes were identified as varying between donors [HLA-DQ ( and ß) chains in all 3 studies and HLA-RDß in the current study and that of Whitney et al.]. Also notable was the common theme of interindividual variation in specific interferon-regulated genes. Whitney et al. identified a cluster of six interferon-regulated genes (OAS3, MTAP44, INADL, MX1, IFIT44L/GS3686, and IFIT1) that varied among individuals. Radich et al. identified a cluster of five such genes that overlapped partially with those described by Whitney et al. (CIG5, IFIT1, IFIT4, MX1, and USP18).
In the study described here, when two-dimensional hierarchical clustering was performed with 431 genes identified as varying between individuals using Holm multiple-test correction with a cutoff value of 103, a cluster of 9 genes was identified, of which 7 were genes known to be interferon regulated (FER1L3, Ly6E, G1P2/IFI15/ISG15, OAS3, IFI44, IFI44L/GS3686, and STAT1) and 1 belonged to a family for which there is evidence of interferon regulation (LAP3) (8, 13). The remaining gene represents a novel candidate for interferon regulation (SMAP52). In addition, MX1 was identified as varying between individuals in the current study when the less-stringent Benjamini and Hochberg multiple-test correction with a false discovery rate of 5% was applied. Finally, in both the current study and that described by Whitney et al., immunoglobulin genes (including different forms of light and heavy chains and the J chain) were identified as being highly represented in the lists of genes that decreased in expression in an age-dependent manner. This appears to be in accordance with the finding that serum IgG and IgM levels reduce with increasing age (3).
However, the full extent of overlap between the genes of interest identified in the three studies as varying significantly with the parameters investigated was rather limited. For example, of the 72 genes identified by Radich et al. (10) as varying between individuals in mixed leukocyte populations from whole blood, only 22 were present in the list of 3,302 genes identified in the current study as varying significantly between volunteers. Of the 20 genes that were found to be most significantly differentially expressed between volunteers in the current study, 3 genes were also identified as varying in expression with sex in both the current study and the study of Whitney et al. (16) (XIST, CYorf15B, and UTY), 1 gene was common to the list of genes identified as varying between individuals in the study of Radich et al. (GNG11), and only 1 gene was found to vary between individuals in all three studies (HLA-RDß chain). The majority of the 3,302 genes identified as varying significantly in transcript levels among volunteers represent findings unique to the study presented here.
Differences between the genes identified and the interpretation of results in the various studies discussed here are likely to have resulted from technical differences in the array platforms used, the genes represented on the arrays, and the analytical procedures used. Variation in the leukocyte subsets analyzed (e.g., leukocytes from whole blood vs. isolated PBMC) will also have had a bearing on the results, which makes direct comparison more difficult. In the present study, a comparatively small number of genes were identified for which the transcript levels correlated with differences in leukocyte subsets in the PBMC samples analyzed. This suggests that use of PBMC rather than whole blood is an effective approach for reducing intersample variation.
From their analyses, Whitney et al. (16) concluded that there were intrinsic individual differences in the expression of a number of genes in PBMC, but that these were not the dominant source of variation in gene expression among the samples. The conclusions of the current study tend to suggest the reverse. Within-individual variation in the expression of most genes was low. However, a large proportion of the genes analyzed (up to 39%) exhibited statistically significant differences between individuals.
This difference in interpretation is most likely due to factors of study design and statistical analysis. For the study described here, the number of samples analyzed and, in particular, the greater and consistent number of samples obtained from each volunteer would have increased the statistical power and enhanced detection of interindividual differences in gene expression compared with the equivalent component of the study of Whitney et al. (16). Also, the samples for the current study were collected according to a more closely controlled regimen: at weekly intervals and always first thing in the morning after an overnight fast. The samples for the study by Whitney et al. were collected at variable intervals, the volunteers were not specifically fasted, and the samples were taken at different times of day. This reflects differences in the aims of the two studies rather than any study design fault. However, such factors may well have served to increase the within-individual variation observed. Indeed Whitney et al. were able to identify a number of genes whose expression varied according to the time the samples were collected (16). It was not possible to evaluate acute effects of food intake in either study.
We used two parallel approaches and software systems to analyze the data. This enabled analyses of both discrete and continuous variables. Because microarray data interrogation is still a developing field, there were inevitably slight differences in the data normalization processes used for the different approaches, and the statistical approaches were quite distinct. However, for sex (the only parameter investigated using both approaches) it was reassuring to note the substantial overlap in genes identified by the two methods as exhibiting sex-specific differences in transcript levels. The genes that were identified using both approaches were generally those for which the calculated P values were lowest. This suggests that both approaches were valid.
Sample processing before RNA isolation introduces the potential for generation of artifacts resulting from alterations in the expression of certain genes during processing. Indeed, it has been suggested that the typical gene transcription profile of RNA prepared from isolated PBMC may carry a "fingerprint" that is symptomatic of a stress response (16). Additionally, changes in gene expression have been reported as a result of varying delays between collection of blood samples and subsequent RNA isolation (10). In the current study, all possible practical measures were taken to ensure that the PBMC were always isolated as quickly as possible after obtaining blood samples. Just as important, the isolation protocol was set up with a view to maximizing the consistency of the process and minimizing operator differences. Thus, while it is possible that changes in gene expression did arise during sample processing, the objective to keep this to a minimum and as consistent as possible between samples was achieved.
It is not possible to be certain which of the factors discussed above are primarily responsible for the differences in gene transcription profiles reported between the various studies. However, these differences do not mean that any of the results are necessarily invalid. Indeed, these types of apparent discrepancy are common in all biological research. The particular issue with microarray studies is that the volume of data generated is so much larger than was previously achievable that discrepancies may promote confusion rather than provide clarification. Clearly, the findings that are most consistent among the studies offer an obvious point for initial follow-up, but the other data should not be ignored. These studies serve to highlight the fact that improved standardization in the design, execution, and analysis of array studies, to accompany the established standards of array data reporting, would facilitate cross-study comparisons.
The genes that were found to vary most within individuals represent an intriguing group. The possibility that some, perhaps all, of the enhanced between-sample variation observed for these genes may have arisen from technical errors due to array design factors cannot be ruled out. However, the presence of a large number of immunoglobulin genes within this group that cosegregated when analyzed by hierarchical clustering lends credence to the idea that this process did not select random genes that varied solely because of technical factors. Rather, it seems more likely that many of these genes do indeed vary significantly in expression at the level of transcription from day to day. If so, these correspond to genes that would appear to be exquisitely sensitive to environmental factors.
As discussed above, clusters of immunoglobulin genes were also identified as varying in mRNA transcript levels significantly with donor age and between volunteers. This suggests that both genetic and environmental factors are important in the regulation of their expression, consistent with what we already known about immunoglobulin function, regulation of gene expression, and interindividual variations in response to immune challenges.
In conclusion, this study extends the limited information base currently available that describes normal patterns of gene expression and variation at the level of mRNA in human tissues. The data generated have been made freely available and should represent a useful resource for the design of future studies. A host of new candidate genes for regulation by genetic, life-stage, and environmental factors were identified. These could be used as a basis for hypothesis development. Substantial differences in gene expression profiles were identified between individuals. Conversely, PBMC gene transcription profiles were found to be remarkably consistent within serial samples obtained from individuals. This suggests that the scope for future clinical and nutritional intervention studies using gene expression profiling is considerable, provided care is taken in the study design to account for interindividual differences and in sample preparation to avoid, or at least minimize, the introduction of processing artifacts. The baseline data generated should find use in the execution of power calculations for the design of such studies.
![]() |
GRANTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
ACKNOWLEDGMENTS |
---|
The Institute of Food Research is a member organization of The European Nutrigenomics Organisation (NuGO).
![]() |
FOOTNOTES |
---|
Address for reprint requests and other correspondence: R. M. Elliott, Institute of Food Research, Norwich Research Park, Colney, Norwich, United Kingdom (e-mail: ruan.elliott{at}bbsrc.ac.uk).
10.1152/physiolgenomics.00080.2005
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|