Published online before print November 21, 2003
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
* Department of Pediatrics and Cancer Center, University of Massachusetts School of Medicine, Worcester; and the
Department of Genetics, Boyer Center for Molecular Medicine, Yale University School of Medicine, New Haven, Connecticut
1 Correspondence: Department of Pediatrics and Cancer Center, University of Massachusetts Medical School, LRB 404, 364 Plantation Street, Worcester, MA 01655. E-mail: peter.newburger{at}umassmed.edu
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Key Words: mRNA • transcription • transcription factor • chromatin
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
A widespread but incorrect view of the neutrophil portrays it as a short-lived, terminally differentiated cell with a highly condensed nucleus and hence, unable to induce gene expression. However, in spite of its dense chromatin and paucity of total mRNA and ribosomes, the neutrophil has long been known to be capable of transcription-dependent synthesis of heat-shock protein [5 ], to express mRNA-encoding phagocytic receptors [6 ], and to modulate RNA synthesis in response to lectin stimulation or glucocorticoid treatment [7 ]. Neutrophils can up-regulate genes involved in phagocytic function, such as the genes encoding gp91-phox and p22-phox, the two components of the phagocyte cytochrome b [8 , 9 ]. In addition, a variety of stimuli, defined and complex, up-regulates c-fos expression [10 , 11 ].
Neutrophils not only respond to cytokines but also participate in the "cytokine network" by secretion of interleukin (IL) and colony-stimulating factor (CSF) molecules, with regulation at the level of gene expression demonstrated in some studies. Stimulation with bacterial lipopolysaccharide (LPS) induces neutrophils to synthesize IL-1
and -1ß [12
, 13
], as well as IL-1 receptor (IL-1R) antagonist [14
]. Granulocyte macrophage (GM)-CSF induces neutrophil expression of IL-6 [15
], and 5' lipoxygenase [16
]. Tumor necrosis factor- (TNF-) expression by neutrophils occurs in response to LPS, GM-CSF, and G-CSF [17
], and the secreted cytokine creates a feedback loop for further neutrophil activation. In other positive-feedback loops, neutrophils augment their own recruitment by up-regulating expression of IL-8 and G-CSF [13
, 18
]. Conversely, incubation of neutrophils with G-CSF induces expression of interferon- (IFN-) [19
], which provides negative feedback to diminish myelopoiesis and neutrophil activation.
Several recent high-throughput studies have expanded the scope of neutrophil gene expression by showing a wide range of mRNAs present in unstimulated neutrophils and dramatic changes in transcript levels induced by stimulation with bacteria and other agonists [20
21
22
23
]. Our previous studies using cDNA-display analysis identified 
600 known genes and expressed sequence tags differentially expressed by human neutrophils exposed to bacteria and encoding a variety of cytokines, receptors, apoptosis-regulating products, and membrane-trafficking regulators [20
].
Fessler et al. [21 ] performed genomic and proteomic analyses of neutrophil activation by LPS, noting the up-regulation of 100 distinct genes and focusing on the attenuation of the response by pretreatment with a p38 kinase inhibitor. Kobayashi et al. [22 , 24 ] used oligonucleotide arrays to compare changes in human neutrophil transcript levels over a 6-h time-course, accompanying phagocytosis of opsonized particles mediated by Fc receptors for immunoglobulin G (IgG) versus complement receptors. Many differentially expressed genes related to at least three distinct apoptotic pathways, in contrast with a more limited number of genes involved in host defense. The onset of apoptosis was marked by changes in expression of nearly 200 genes involved in membrane trafficking and cell metabolism, including enzymes necessary for energy metabolism and antioxidant defenses [24 ]. Yang et al. [23 ] used microarray analysis to examine neutrophil activation by antineutrophil cytoplasmic antibodies from kidney disease patients, focusing on the induction of human differentiation-dependent gene 2.
In the present study, we have examined in detail the time-course of changes in gene expression during neutrophil activation and have compared the expression profiles induced by stimulation with live, opsonized Escherichia coli K-12 enterobacteria, soluble E. coli LPS, and the bacterially derived chemoattractant formyl-methionyl-leucyl-phenylalanine (fMLP). In particular, our data analysis has focused on the early-response genes, which should represent the most direct effects of exogenous stimuli and include the transcriptional regulators of subsequent changes in expression of the effector genes described in previous studies.
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Isolation and stimulation of leukocytes
Neutrophils were isolated from venous blood (freshly drawn at 8–9 AM) of healthy volunteers, using dextran sedimentation and centrifugation through Ficoll-Paque Plus (Pharmacia, Uppsala, Sweden), as described previously [20
]. Morphologically, the neutrophil preparations were more than 99% pure except for the presence of 1–3% eosinophils and <0.75% monocytes. Freshly purified neutrophils, 2.5 x 108 cells for each time point, were suspended at 2 x 106 ml-1 in 80–100 ml RPMI medium (Life Technologies), supplemented with 10% heat-inactivated donor serum (from a previous blood draw, frozen after heat-inactivation at 56°C for 30 min). Neutrophils were equilibrated to room temperature, then stimulated by addition of opsonized E. coli (20:1 ratio of E. coli to neutrophils), fMLP (10-7 M), or LPS (10 ng/mL), and incubated at 37°C with gentle agitation in a temperature-controlled water bath. For each time-course, a time-zero control was harvested before stimulation and subsequent samples at 10, 20, 30, and 60 min of incubation, plus a 120-min time-point for E. coli and LPS. Rapid chilling and then centrifugation at 1100 g at 4°C terminated the incubation. Because of inter-donor differences in gene expression [25
], we adopted a repeated-measures design of the experiments to treat the data as within-subjects variables to isolate each set of time-points for each neutrophil isolation from each donor. Each stimulation time-course was repeated at least three times with different donors.
Human monocytes were prepared and stimulated as described previously [26
]. In brief, peripheral blood mononuclear cells were separated from leukopheresis samples by Ficoll-Paque density gradient centrifugation and allowed to adhere onto serum-coated, plastic, six-well, tissue-culture dishes (2x106 cells/well) for 2 h at 37°C in a 5% CO2 incubator. Nonadherent cells were removed by gentle washes, and half the wells with the purified monocytes adhering to the plastic were harvested for total RNA using RNAeasy (Qiagen, Valencia, CA). Monocytes were immunostained with anti-CD-14 antibody to assure 
95% purity.
The manufacturer certified all reagents—serum, buffers, media, and containers—as nonpyrogenic. Peripheral blood and leukopheresis samples were obtained from normal adult volunteers. The University of Massachusetts Medical Center Committees on Protection of Human Subjects in Research (Worcester) and Yale University Human Investigation Committee (New Haven, CT) approved all procedures and consent forms at their respective sites.
RNA sampling and oligonucleotide array hybridization
Total RNA was extracted from neutrophils by a guanidine HCl method, as described previously [27
, 28
]. RNA samples were digested with DNase I at 37°C for 30 min and cleaned by passage through RNeasy mini-spin columns (Qiagen). Running aliquots on native agarose gels monintored the quality of RNA. Each RNA sample (10 µg) was used as starting material for cRNA preparation, as described previously [29
]. In vitro transcription by T7 RNA polymerase (Ambion, Austin, TX) was performed in a nucleotide triphosphate mixture containing the biotinylated reagents Bio-11-CTP and Bio-16-UTP (Enzo Life Sciences, Famingdale, NY). Labeled cRNA samples and fragmentations were checked on native agarose gels, and 15 µg fragmentation products were hybridized to HG_U95A version 2 GeneChip arrays (Affymetrix, Santa Clara, CA), according to the manufacturer’s instructions.
Data analysis
Affymetrix MicroArray Suite version 5.0 managed raw data output, with average intensity (target signal) set as 150. The absolute expression (detection) analysis used the statistical algorithm of Microarray Suite 5.0. To reduce noise in E. coli time-course samples as a result of bacterial RNA contamination, raw signals were fitted to the Li and Hung [30
] and Li and Wong [31
] noise model, implemented in DNA chip-analyzer software (dChip; http://www.dchip.org/). The probe-sensitivity index file generated from all of the non-E. coli time-course samples was applied to all data, including the E. coli samples, to generate model-based expression index (MBEI) values. The Log2-transformed MBEIs were first filtered by a flooring filter; that is, at least one group (stimulus and time) needed to have signal values up to the average signal level of marginal calls [Log2(MBEI)=7 or MicroArray Suite 5.0 signal=128 (for details, see Table S1 in the online supplementary material http://www.jleukbio.org/cgi/full/jlb.0903412/DC1)]. The floored MBEI data were used for an F-test of each time-course by significance analysis of microarrays (SAM) software [32
], a multiple testing approach based on permutation resampling and designed for microarray analysis involving small numbers of replicates. The MBEI data without log transformation were used for a group-comparison analysis, implemented in dChip, which calculates weighted group means to reduce bias and uses a lower 90% confidence bound of relative change for filtering. The "Affy" package of the R project BioConductor (http://www.bioconductor.org/) [29
, 33
], which is based on different assumptions and uses a different normalization approach, was used to generate the median polish estimates of expression levels. The data normalized by Affy were subjected to filtering with same floor filter as described above and to the SAM multiclass test using F statistics with the same setting.
Genes were classified as showing a significant change in expression level if they met the following criteria for an arbitrary threshold of relative change in comparisons of group means and a P-value threshold in statistical hypothesis testing: For at least for one time-point and stimulus, the weighted mean of the gene’s expression level had a relative change 1.8-fold (lower 90% confidence bound-of-fold change) and an absolute difference 64 compared with its corresponding time-zero group; and for at least one stimulation time-course, the gene must be rejected by SAM using F statistics with a false discovery rate set at 0.01.
Genes with significantly different levels of expression over the time-course, as identified by the approach described above, were then subjected to hierarchical clustering to group the genes according to similarity defined by correlation coefficient. The k-means cluster was used to test the stability of the hierarchical clusters. The functional annotations of significant genes were organized with GoSurfer software (http://biosun1.harvard.edu/complab/gosurfer). Small, relevant branches and unresolved genes (with limited or ambiguous Gene Ontology annotation) were merged manually, based on information in National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/) databases.
Real-time reverse transcriptase-polymerase chain reaction (RT-PCR)
RT and PCR analysis using real-time PCR verified microarray findings for multiple genes [34
]. Target regions for PCR primers were selected from the human oligonucleotide set for use in microarray and PCR measures of RNA designed by the Harvard-Lipper Center for Computational Genetics [35
] (http://arep.med.harvard.edu/probes.htm) and based on the human UniGene dataset Build #158 (http://www.ncbi.nlm.nih.gov/); or from Affymetrix probe selection regions within the target sequence of the Affymetrix HG_U133 set (http://www.affymetrix.com/analysis/download_center.affx), based on human UniGene Build #133. Primers were designed to yield amplicons of 100–150 bp.
Aliquots of the same RNA samples used for microarray analyses served as templates for the RT reactions. First-strand cDNA was synthesized from 50 ng DNaseI-treated, total RNA using SuperScript II RT (Life Technologies, Rockville, MD). Quantitative PCR was performed on a Smart Cycler (Cepheid, Sunnyvale, CA), using SYBR Green I fluorescent dye (QuantiTech SYBR Green PCR, Qiagen). A comparative threshold cycle (Ct) method was used to process the real-time PCR data [36
]. A panel of six constitutively expressed genes, most from the Affymetrix HG-U133A set normalization controls (http://www.affymetrix.com/support/technical/byproduct.affx?product=hgu133), was chosen as reference genes for normalization. Results were calculated as the normalized difference in Ct for a given time-point relative to the time-zero baseline (
Ct), calculated by subtracting the Ct of the reference gene from the Ct of the target gene for each time point (Ct) and then subtracting the Ct,cb of time zero from the Ct,q of time-point. Amplicon size and specificity were verified by agarose gel electrophoresis.
Principal component analysis of transcription factors
To analyze the differences and similarities between neutrophils and monocytes in resting or activated states, we examined the expression profiles of 474 transcription factor genes that had at least one sample with an Affymetrix "present call". We performed a principal component analysis among 31 datasets representing resting neutrophils, neutrophils stimulated with E. coli for 30 or 120 min, resting monocytes, and monocytes stimulated with E. coli for 120 min. As a preliminary step in searching for differentially expressed transcription factor genes among the five different cell types, we excluded 180 transcription factor genes for which we did not have complete data or genes that have Affymetrix "absent calls" in all samples. To verify that the 31 samples are separable into five distinctive groups, despite the expected sample similarities among related cell types, we visualized their distribution in a reduced, two-dimensional space. To reduce the dimensionality from 474 (the number of transcription factors used in profiling each sample), we used a variant of principal components analysis. This procedure allows projection of the data onto a two-dimensional space spanned by the two leading principal components, which are the linear combinations of the normalized expression values of the 474 transcription factors that capture most of the variance of the complete data.
The projections of the samples onto the leading principal components are computed by applying the singular value decomposition technique on the transformed data matrix. The latter is obtained by a preprocessing, normalization step, as described recently [37 ]. We selected a normalization procedure based on the concept that two genes and likewise, two samples, whose expression profiles differ only by a multiplicative constant of proportionality, are really behaving in the same way. That is, we consider two genes (each represented by a row of the matrix) or two samples (represented by the columns of the matrix) to be equivalent if their expression profiles differ by a multiplicative constant. This adjustment involves independent rescaling of the rows and columns as described previously [37 ].
To identify the differentially expressed transcription factors that have at least one cell type with a mean normalized-expression value significantly different from means of samples of other cell types, we applied a one-way ANOVA [38 ] and its nonparametric Kruskal-Wallis test [39 ]. Transcription factors that return P values <0.0001 using the classical ANOVA or <0.01 using its nonparametric version cast doubt on the null hypothesis that the means of all cell types are equal. Small P values suggest that at least one sample mean is significantly different than the other sample means. By applying the classical one-way ANOVA, we identified a subset of 96 transcription factors with P values <0.0001. In the ANOVA test, it is assumed that all sample populations have equal variance and are normally distributed (in addition to the assumption that all observations are mutually independent); therefore, we also applied the less-restrictive Kruskal-Wallis test on this subset of 96 genes. We found that 90% of genes have P values <0.001, indicating that both tests select similar lists of differentially expressed genes. Once we identified a differentially expressed transcription factor, we determined which pairs of the normalized expression means of the different cell types are significantly different using a multiple comparison procedure [40 ] based on Tukey’s honestly significant difference criterion as a conservative test. This procedure is designed to provide an upper bound on the probability that any comparison will be incorrectly found significant. The output contains the estimated difference in means of each pair of cell types and an adjustable confidence interval for the difference, which we fixed at 99.99%. If the confidence interval does contain zero, the difference would not be significant at the 0.0001 level. Thus, if the confidence interval of the difference of means of two cell types does not contain zero, we considered the transcription factor to have significant differential expression between this pair of cell types.
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Although there is no consensus approach for the identification of a transcriptome, an absolute detection analysis by the microarray manufacturer’s procedure provides a good estimate. For Affymetrix HG_U95Av2 chips, which display 10,000 transcripts represented by 12,561 probe sets, the total number of neutrophil transcripts eliciting a present signal constituted 4400 probe sets (35% of the total), representing 3500 unique entries (for details, see Table S1 in the online supplementary material). The numbers of transcripts detected as present in resting and stimulated neutrophils varied only slightly within a narrow range and correspond closely to a previous measure of neutrophil gene expression using similar microarrays [22
] and to an alternative calculation of the present data by dChip software. Microarray measurements of transcriptome size in other cell systems have reported similar proportions of positive signals [41
42
43
44
].
Using criteria detailed above, 976 unique entries in the UniGene (Build #95) database were identified as differentially expressed in response to stimulation by three microbe-derived stimuli: particulate-opsonized E. coli, soluble LPS, and soluble chemoattractant fMLP (see Fig. 1 ). Of these identified entries, 126 represent genes that currently have no functional annotation in the public databases and so were not included in further analysis. Among the 850 annotated response genes, only 16 genes had group mean signals below the average absent level (30.68) at time zero (resting), and 84 genes had a group mean signal under the average marginal level (96.73). The paucity of low or undetectable signals indicates that most changes occurred within a detectable range, rather than representing absolute "on/off" switching, and suggests that most of the cellular response is derived from the active transcriptome, through wide-ranging, quantitative fine-tuning of gene expression.
 View larger version (63K): [in a new window] |
Figure 1. Hierarchical clustering of genes expressed in human neutrophils and responsive to stimulation by opsonized E. coli, LPS, and fMLP. The "heat map" presents up- and down-regulation of mRNA levels detected on human genome HG95 microarrays, normalized and filtered as described in Materials and Methods. Rows correspond to different genes, and the columns represent the various time-points after exposure to each stimulus [time 0 (unstimulated) and times 30, 60, 90, and 120 min for E. coli and LPS; times 30, 60, and 90 min for fMLP, from left to right, respectively). (a) The final clustering tree, with omission of the initial branch between the upper, down-regulated cluster (blue dendogram) and the lower, up-regulated clusters. (Purple and red dendograms indicate early/transient and late up-regulation, respectively.) (b) The expanded cluster shows transient responses in greater detail, and gene symbols are indicated to the right of each row. The scales below the heat maps indicate the relative changes in gene expression represented by the range of colors.
|
 View larger version (16K): [in a new window] |
Figure 2. Unique and overlapping responses to neutrophil stimuli. The Venn diagram presents the number of genes that showed significant changes in expression in response to the indicated stimuli. (a) Repressed genes; (b) up-regulated genes.
|
The largest group, surprisingly, is a collection of 245 genes involved in the multiple layers regulating gene expression, including 148 transcription-related genes (Table 1 ) encoding transcription factors, transcription activators, chromatin remodeling and modifying proteins, DNA helicases and other DNA-binding proteins, as well as 95 regulators of protein synthesis or stability (Table 2 ), including translation factors, ubiquitination machinery or small protein-conjugating enzymes, and chaperones. The two processes of transcription and protein synthesis/stability are not completely distinct. For example, polyamine metabolism alters DNA–protein interactions [48 ] but also contributes to an autoregulatory feedback loop controlled by translational frameshifting [49 ]. Conversely, chaperone proteins not only facilitate proper folding of polypeptides and protect fully translated proteins from aggregation but also stimulate the activity of histone-deacetylase complexes [50 ], which are essential for remodeling chromatin structure and transcriptional accessibility [51 ].
|
View this table: [in a new window] |
Table 1. Differentially Expressed Genes Involved in Transcriptional Regulation
|
|
View this table: [in a new window] |
Table 2. Differentially Expressed Genes Involved in Protein Synthesis and Stability
|
(PTPRE)], which were up-regulated by all stimuli. The other major groups include 80 cell migration and cytoskeletal-regulation genes [such as intercellular adhesion molecule-1 (ICAM1), urokinase plasminogen activator receptor (PLAUR), and 
-actin (ACTG)1], 57 host defense genes [such as Toll/IL-1R-containing molecule domain-containing adaptor-inducing IFN-ß (TRIF), CD83, and oxidative stress response 1 (OSR1), induced in all responses], and 55 cytokine and cytokine-signaling molecules [such as CXC chemokine ligand (CXCL)1, IL1B, TNF, and nuclear factor (NF)KB immediate-early (IE)]. One of the host-defense genes, CD83, encodes an Ig supergene family member, best known as a marker for mature dendritic cells [52
], but has also been identified as a differentially regulated gene in neutrophil and macrophage-transcription profiles [22
, 47
, 53
]. This gene underwent early and sustained up-regulation from marginal expression in resting cells to a more than 20-fold higher level in cells responding to all three stimuli. CD83 surface antigen has been reported previously in neutrophils cultured in vitro with IFN- or GM-CSF [54
] or harvested during acute bacterial infection in vivo [55
].
Neutrophils exposed to the exogenous microbe-derived stimuli also showed distinct, temporal patterns in the resultant changes in gene expression. By hierarchical clustering, four-time clusters were identified, including early down, transient up, early up, and late up clusters (Fig. 3
). The early down-regulated or repressed cluster, consisting mainly of E. coli response genes, showed a rapid declining pattern that occurred within the first hour of stimulation. In addition, within the 2-h time course, a group of 60 genes, listed in Table 3
, exhibited a particularly finely tuned pattern of transient up-regulation and then a return to the original level in response to one or more of the stimuli. Some genes in this cluster [such as prostaglandin-endoperoxide synthase-2 (PTGS2) oncostatin M, (OSM), and TOB1] showed no change or a sustained response in other studies [21
, 22
]. However, the transient pattern in our system was confirmed by real time RT-PCR analysis (see below).
 View larger version (32K): [in a new window] |
Figure 3. Temporal clustering of responses. The bar graphs show the numbers of differentially expressed genes, divided by temporal pattern (indicated above each graph) and functional category of the genes (indicated in the left margin). The numbers of responding genes are indicated in the scale at the top of each graph. Colors represent the different stimuli, as shown below.
|
|
View this table: [in a new window] |
Table 3. Transiently Up-Regulated Genes
|
It is also noteworthy that many of the down-regulated genes function as repressors. For example, the transcriptional corepressor MAX dimerization protein 1 (MAD) [56 ] was down-regulated by all three stimuli. Neutrophils also down-regulated the genes encoding chromobox homologue 1 (CBX1), a major component of heterochromatin [57 ], and SMART/histone deacetylase (HDAC)1-associated repressor protein (SHARP), which acts as a transcriptional corepressor through the recruitment of histone deacetylase complexes [58 ]. Both were rapidly down-regulated in response to E. coli; CBX1 was also significantly down-regulated by LPS. Thus, the diminished expression of these genes could paradoxically contribute to the net activation response of the cell.
Compared with the down-regulated responses, the sets of up-regulated genes appeared to be somewhat more specific to each stimulus. Although the number of genes up-regulated by all three stimuli was larger than the set that all three repressed, the temporal patterns of down-regulation were not identical. In addition, each stimulus induced expression of a unique group of genes: 60 by E. coli, 54 by fMLP, and 103 by LPS. This divergence may reflect differences in receptors and signal-transduction pathways activated by each stimulus, although many of the pathways eventually converge on the same cytoplasmic transcription factors, such as NF-
B, Smad, and signal transducer and activator of transcription (STAT) family members.
Validation of microarray data by real-time quantitative PCR
Thirty well-annotated genes with significant changes in expression level were picked for validation by real-time quantitative RT-PCR (a complete list is available as Table S3 in the online supplementary material). These included up-regulated or transiently up-regulated genes (including OSM, IL8, DNAJB1, PTGS2, and CD69) and down-regulated genes (including HLX1, CNOT3, FBXO9, FOXO1A, and BRD8). Each target gene was examined in 15 reactions using samples from three stimulation time-course experiments. The specificity of each reaction was confirmed by melting-curve analysis and agarose-gel visualization of the PCR products.
Figure 4
presents a plot of real-time quantitative PCR data, expressed as Ct (where the change in transcript abundance is a logarithmic function of 2-CT), versus the linear change in oligonucleotide array signal. The data were filtered to show only points with a greater than 1.25-fold change from time zero. Almost all of the interrogations are located in the first and third quadrants, indicating good concordance between the results from real-time PCR and from oligonucleotide arrays for up-regulated and down-regulated changes (91% and 88% concordance, respectively, with correlation R2=0.49). As expected, we observed greater changes in expression by real-time quantitative PCR (along the logarithmic x-axis) than by oligonucleotide array (on the linear y-axis). RT-PCR has a larger dynamic range and sensitivity than microarray analysis hybridization and tends to generate data showing a greater magnitude of change. Thus, the technique is generally used to validate observed trends but is not expected to duplicate the fold changes obtained in chip experiments [59
, 60
].
 View larger version (20K): [in a new window] |
Figure 4. Comparison of changes in gene expression measured by real-time quantitative PCR versus oligonucleotide array. Each point on the scatter plot represents the difference in expression level of a given gene relative to the time-zero expression level, as measured in the same RNA sample by the two techniques. The PCR data are expressed as normalized differences in Ct, representing the difference between time zero (base line) and a specific time point, normalized to a reference gene that showed the closest relative amplification efficiency within a panel of constantly expressed normalization control genes; thus, the change in transcript abundance is a logarithmic function of 2-CT. Microarray hybridization signals are expressed as fold changes, representing the ratios of weighted group means generated by the "combine comparisons" function of dChip software.
|
 View larger version (11K): [in a new window] |
Figure 5. Principal component analysis of transcription-factor expression by neutrophils and monocytes. The graph shows the projection of 31 samples onto a subspace spanned by the two leading principal components. Using a normalized matrix, projections of the samples onto the second and third principal components for normalized profiles of each cell type are shown by the symbols indicated in the figure insert. Clustering of samples is evident, although the identities of the samples were not used in performing the projection.
|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The total number of genes identified in this study is comparable with that reported in microarray analyses by Kobayashi et al. [22 , 53 ], of which 40% of the genes were found to be differentially expressed under the various experimental conditions of the previously published and present studies. In particular, down-regulation of critical inflammatory response genes, as first described by Kobayishi et al. [53 ], was also observed in our experiments using E. coli stimulation. For example, the bacterial agonist caused down-regulation of platelet/endothelial cell adhesion molecule 1 and IL-17R, but the changes in these genes were not significant in fMLP and LPS experiments.
The diversity of the other sets of genes identified in the different experiments probably derives from differences in experimental design, such as preparation of neutrophils, time points for examination of expression profiles, and diversity of stimuli. The current studies focused on the early transcriptional changes and comparisons of "complete" particulate bacterial stimuli and with two of its soluble components. In comparison with the previous findings, we identified fewer metabolism and vesicle-transport genes but more chromatin-binding genes and translation regulators. Most importantly, 75% of the transiently regulated genes found in our study (mostly at early time points) were not identified in other work on the neutrophil response. As expected, we did not detect events observed by DeLeo’s laboratory [22
, 24
, 53
] at later time points, such as down-regulation of IL-8Rß (IL8RB) and IL-10R (IL10RA) at 3–6 h after stimulation. Overall, the differences between the sets of genes identified in the previous and current studies probably reflect temporal changes in the transcription of early- and late-response genes as well as the diversity of the responses to complex particulate stimuli compared with the individual priming or activating components. Thus, the response to opsonized bacteria appears to represent a complex set of interacting up- and down-regulating changes rather than a simple summation of responses to individual agonist molecules.
In our functional classification of identified genes, the category of regulators and mediators of gene expression included genes encoding molecules ranging from general RNA polymerase II transcription factors to protein chaperones. Some of the transcription factors, such as members of the fos, jun, and NF-B families, were predictable findings, as they have been studied individually in phagocytic cells. However, the present studies confirm and considerably expand our previous analysis [20
] that indicated the operation of a much broader array of transcriptional regulators in neutrophils.
Other differentially expressed, regulatory genes included the protein kinase CK2-phosphorylated HDAC2 (class I HDAC) and MYST histone acetyltransferase 1 (MYST1), which allow chromatins to fluctuate between the condensed and decondensed states [63 , 64 ]. HDAC1 and -2 are components of the large, multisubunit complexes Sin3 and NuRD, which are recruited by transcription factors such as Mad [65 ]. Ubiquitination mediators, other small protein-conjugating enzymes, and chaperones have not been reported in previous transcription profiles of cells in the innate-immune system. They not only direct proteasome-mediated degradation of proteins but also affect chromatin structure and thus serve as members of the multilayered regulation mechanism that coordinates gene expression with other integral biological processes [66 , 67 ].
In addition to the temporal clustering of genes, we also observed rather different profiles associated with each stimulus. The responses of neutrophils to soluble and particulate stimuli are initiated by multiple surface receptors and mediated by a web of signal-transduction pathways [68 , 69 ]. The three agonists used in the current study use quite different signaling mechanisms. Opsonized, particulate stimuli such as bacteria are recognized by antibody and complement receptors, and the live organisms also engage the cell surface through fimbrial proteins and fimbriae-mediated delivery of LPS [70 71 72 ]. LPS molecules vary in structure and form of delivery among bacteria, which can lead to recruitment of different pattern-recognition molecules and signaling pathways [73 , 74 ]. Thus, the differences in responses may reflect the additional receptors engaged by the opsonized bacteria and the differences in structure and presentation of LPS in soluble versus bacterial surface forms. Formylated peptides such as fMLP engage a single receptor in the seven-transmembrane helix family but activate multiple intracellular signaling proteins including p38 MAPK, phosphatidylinositol-3 kinase/Akt, and extracellular-regulated kinase [75 ]. Our previous study of neutrophil responses to virulent and nonvirulent gram-negative bacterial species showed that Yersinia pestis was able to suppress many host-defense genes compared with E. coli [20 ]; the present studies indicate that even the nonvirulent organism is able to play a more dynamic role in shaping the inflammatory response than do simple soluble agonists.
The large array of transcription factors in the neutrophil’s early-response repertoire far exceeds the small number investigated in most previous studies of myeloid function and development, which generally focused on the analysis of the promoters of single genes or on gene knockouts in mice. The total number was comparable with that reported in microarray analyses by Kobayashi et al. [22
, 53
]. Equally striking was the prominence of chromatin-regulatory genes in this repertoire. Gene regulation at the level of chromatin structure has recently been demonstrated for short-term NF-B-induced changes of cytokine and major histocompatibility complex gene expression in T and B cells [76
77
78
]. The present study indicates that alterations of histone acetylation and chromatin structure may also play a role in the innate-immune system by mediating the rapid changes in gene expression that we have observed in neutrophils. The role and interplay of these transcription factors and chromatin-remodeling elements in normal myeloid development and in the response of normal neutrophils to activation are ripe for further intensive study.
Received September 30, 2003; accepted October 22, 2003.
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
and ß genes by human blood polymorphonuclear leukocytes J. Clin. Invest. 87,1312-1321 Proc. Natl. Acad. Sci. USA 87,6758-6761 Blood 75,17-19
C(T)) method Methods 25,402-408[CrossRef][Medline] signalling pathway EMBO J. 21,5417-5426[CrossRef][Medline]B-dependent assembly of an enhanceosome-like complex on the promoter region of apoptosis inhibitor Bfl-1/A1 Mol. Cell. Biol. 23,2749-2761This article has been cited by other articles:
 |
|
 |
|
  X. Zhang, Z. Lian, C. Padden, M. B. Gerstein, J. Rozowsky, M. Snyder, T. R. Gingeras, P. Kapranov, S. M. Weissman, and P. E. Newburger A myelopoiesis-associated regulatory intergenic noncoding RNA transcript within the human HOXA cluster Blood, March 12, 2009; 113(11): 2526 - 2534. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Radom-Aizik, F. Zaldivar Jr., S.-Y. Leu, P. Galassetti, and D. M. Cooper Effects of 30 min of aerobic exercise on gene expression in human neutrophils J Appl Physiol, January 1, 2008; 104(1): 236 - 243. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Pestka and H.-R. Zhou Toll-Like Receptor Priming Sensitizes Macrophages to Proinflammatory Cytokine Gene Induction by Deoxynivalenol and Other Toxicants Toxicol. Sci., August 1, 2006; 92(2): 445 - 455. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. Takizawa, T. C. Smith, T. Nebl, J. L. Crowley, S. J. Palmieri, L. M. Lifshitz, A. G. Ehrhardt, L. M. Hoffman, M. C. Beckerle, and E. J. Luna Supervillin modulation of focal adhesions involving TRIP6/ZRP-1 J. Cell Biol., July 31, 2006; 174(3): 447 - 458. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. C. Parker, M. K. B. Whyte, S. K. Dower, and I. Sabroe The expression and roles of Toll-like receptors in the biology of the human neutrophil J. Leukoc. Biol., June 1, 2005; 77(6): 886 - 892. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. H. Connolly, V. J. Caiozzo, F. Zaldivar, D. Nemet, J. Larson, S.-p. Hung, J. D. Heck, G. W. Hatfield, and D. M. Cooper Effects of exercise on gene expression in human peripheral blood mononuclear cells J Appl Physiol, October 1, 2004; 97(4): 1461 - 1469. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. J. Melendez and F. B. M. Ibrahim Antisense Knockdown of Sphingosine Kinase 1 in Human Macrophages Inhibits C5a Receptor-Dependent Signal Transduction, Ca2+ Signals, Enzyme Release, Cytokine Production, and Chemotaxis J. Immunol., August 1, 2004; 173(3): 1596 - 1603. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Kluger, D. P. Tuck, J. T. Chang, Y. Nakayama, R. Poddar, N. Kohya, Z. Lian, A. B. Nasr, H. R. Halaban, D. S. Krause, et al. Lineage specificity of gene expression patterns PNAS, April 27, 2004; 101(17): 6508 - 6513. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
