Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

In asthma and other allergic diseases, the dramatic increase in the number of eosinophils at the site of antigen challenge has implicated this cell type in the pathophysiology of these diseases (1). The specific accumulation of eosinophils is controlled by a number of different cytokines, including interleukin (IL)-3, IL-5, and granulocyte macrophage colony-stimulating factor (GM-CSF), and chemokines such as eotaxin (3, 4). These polypeptides exert considerable influence over a range of eosinophil functions through a variety of signaling pathways. Some signaling pathways are specific to eosinophils, whereas others are common to a number of cell types. To drive a particular phenotypic change, the cell not only must be exposed to the correct repertoire of cytokines, but also must possess the requisite signal transduction pathways. IL-5 is responsible for promoting several phenotype changes in eosinophils during their maturation (2). Initially, this cytokine plays a critical role in the development of the eosinophil and the control of its release from the bone marrow into the circulation. In addition to its hematopoietic effects, IL-5 is a key survival factor for eosinophils (5). Indeed, in the absence of IL-5 the average life of an eosinophil is in the range of 18 to 30 h, whereas in the presence of IL-5 this is extended to 7 to 8 d. Therefore, the frequency of apoptosis within the eosinophil population is a crucial factor in determining the steady-state number of eosinophils at the site of inflammation.

Physiologically, the role of the eosinophil is in host defense and particularly in responding to parasitic infection (11). Thus, eosinophils are armed with a range of cytotoxic proteins as well as enzymes able to generate reactive oxygen metabolites with the intent of killing parasites. Humans have evolved to tolerate a certain amount of bystander tissue damage in preference to a sustained parasitic infection. By contrast, in asthma the inappropriate recruitment of eosinophils results in persistent tissue damage with no countering benefit (2, 4). In an attempt to regulate the tissue damage that is believed to contribute to the underlying disease pathology, we pursued the discovery of genes whose modulation would induce the apoptosis, or inhibit the survival, of eosinophils. Because IL-5 is central to these processes in eosinophils, we focused our attention on the genes regulated by this cytokine. Considerable literature already surrounds IL-5 and the pathways it transduces (reviewed in 12). In an attempt to fill the gaps in our understanding, we used high-density microarray transcription profiling of eosinophil genes to characterize their response to IL-5. Moreover, the signal transduction literature surrounding this cytokine would be beneficial for the validation of our experimental model and microarray platform. Consequently, robust identification of genes known to belong to these pathways would lend credibility to the data acquired on genes of hitherto unknown function.

From the plethora of IL-5 effects, we expected a variety of genes to be regulated in primary eosinophils. To facilitate the identification of genes that govern the apoptosis and survivability of eosinophils, we looked for an alternative cellular model whose survival was also dependent on IL-5. The cell line TF1.8 has previously been shown to be dependent on IL-5 for survival. We hypothesized that genes whose regulation was similarly coordinated in both cellular systems would have a higher probability of having an antiapoptotic function compared with the remaining pool of regulated genes. In addition, we examined the tissue distribution of the regulated genes to identify those that were restricted or ubiquitous in their expression compared with the eosinophil. Overlaying the tissue distribution information on top of the IL-5 regulation data should provide an insight into the genes that are exclusive to the eosinophil survival process compared with general apoptotic/ antiapoptotic processes. Thus, genes whose function is exclusive to the eosinophil may be more likely to result in pharmaceutical targets whose modulation induces fewer side effects in other cell types.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Cells

The TF1.8 cell line was originally cloned by Prof. Colin Sanderson, Institute for Child Health Research, Perth, Western Australia. This is a derivative of the TF1 human erythroleukemic cell line, which displays increased responsiveness to IL-5 and retains responsiveness to the hemapoietic growth factors normally active on TF1 cells, including GM-CSF, IL-3, and erythropoietin. Primary eosinophils and TF1.8 cells were cultured in RPMI media without phenol red and glutamine (GIBCO BRL, Carlsbad, CA), 10% fetal calf serum (FCS), 1 mM sodium pyruvate, 2 mM L-glutamine, and 10 pM IL-5, and maintained in culture at a density of 10⁶ cells/ml and 3 × 10⁴ cells/ml, respectively.

Isolation of Eosinophils

Eosinophils were purified from 200 ml of whole blood, anticoagulated with 1:9 vol/vol sodium citrate solution, as described elsewhere (13). Whole blood was fractionated by differential centrifugation (400 × g for 20 min at room temperature) through histopaque using Acuspin tubes (Sigma, St. Louis, MO). Mixed granulocytes were recovered with the red-cell layer, which was resuspended in 6% dextran and left to stand for 45 min at room temperature. The supernatants were pooled and granulocytes pelleted by centrifugation (300 × g for 5 min at 4°C). The cell pellet was resuspended in phosphate-buffered saline and contaminating red cells were lysed by hypotonic shock (20 ml of water for 30 s followed by 20 ml of 2× Hanks' balanced salt solution). The mixed granulocytes were incubated with anti-CD16 immunomagnetic beads (Miltenyi Biotech, Bergish Gladbach, Germany) for 40 min at 4°C to bind neutrophils, and were passed through a magnetic column; pure eosinophils were obtained by negative selection. Eosinophils were resuspended in RPMI 1640 (GIBCO BRL) containing 10% FCS; a sample was stained with Kimura and counted, and the percentage purity was enumerated. Samples were routinely > 95% pure.

Cellular Assays

For the IL-5-withdrawal experiments in TF1.8 cells, the cells were spun down, washed repeatedly, and resuspended at a density of 1 × 10⁵ cells/ml in growth media without the IL-5 supplement. Primary eosinophils were purified and cultured in media as described earlier in the presence or absence of IL-5. Cells were harvested and subjected to apoptosis assay or RNA extraction before microarray analysis at the appropriate time points, as indicated in the figure captions.

Caspase 3 assays were performed using the Clontech ApoAlert caspase 3 assay kits according to the manufacturer's instructions. Ac-Asp-Glu-Val-Asp(DEVD)-7-Amino-4-trifluoromethyl coumarin(AFC) or Ac-Asp-Glu-Val-Asp(DEVD)-7-Amino-4-methyl coumarin(AMC) at 10 mM (in dimethyl sulfoxide) was obtained from Promega (Madison, WI) or Alexis (San Diego, CA). Samples were incubated for 30 min at 37°C in FLUOstar. The DEVD-AMC and DEVD-AFC fluorescence values were measured at excitation and emission wavelengths of 390 and 495 nm, respectively.

After 1 h, a further measurement was recorded to determine the end-point fluorescence. Phosphatidylserine externalization was monitored by fluorescence-activated cell sorter (Beckman Coulter, Fullerton, CA) analysis using the Promega ApoAlert Annexin V-FITC Apoptosis Kit according to the manufacturer's procedures.

Extraction of Total Cellular RNA

Total cellular RNA (cRNA) was extracted from eosinophils and TF1.8 cells using a modified RNAzolB method described elsewhere (14). Purified eosinophils or TF1.8 cells (with or without IL-5) were pelleted by centrifugation (3,000 × g for 5 min at 4°C) at the appropriate time points and lysed in 0.2 ml RNAzolB/10⁶ cells. RNA was isolated by double phenol chloroform extraction followed by precipitation of the aqueous phase with isopropanol using standard procedures. RNA was quantified by spectrophotometry (Absorbance [A]260 nm), and quality was assessed by electrophoresis through formamide/formaldehyde agarose gel.

Affymetrix Microarray Complementary DNA Labeling, Hybridization, and Scanning

These methodologies are described in detail by Affymetrix (Palo Alto, CA). Essentially, poly(A+) messenger RNA (mRNA) was extracted and purified from total RNA (5 to 100 µg) using a Micro poly(A+) kit (Ambion, Austin, TX). The complementary DNA (cDNA) was synthesized (from 0.5 to 5 µg mRNA) using the Superscript Choice kit (GIBCO BRL), incorporating a T7-(dT)24 primer (Genset, Paris, France). Purified cDNA (up to 2 µg) was in vitro transcribed using the MEGAscript T7 Kit (Ambion), incorporating Biotin-11-cytidine triphosphate and Biotin-16-uridine triphosphate (final concentration 1.875 mM; Sigma/Enzo, New York, NY). After purification, in vitro cRNA was fragmented in buffer containing magnesium at 94°C. Labeled cDNA was hybridized to the microarray while rotating at 60 rpm, for approximately 16 h at 45°C. After hybridization, the microarray was washed using the Affymetrix Fluidics Station in buffer containing biotinylated antistreptavidin antibody (Vector Laboratories, Burlingame, CA; 10 min, 25°C) and stained with streptavidin phycoerythrin (SAPE) (final concentration 10 µg/ml; Molecular Probes, Eugene, OR) for 10 min at 25°C. Subsequently, the microarray was washed, restained with SAPE (10 min, 25°C), and washed again before measuring the fluorescence bound to the microarray at 570 nm in an Affymetrix scanner.

Microarray Data Processing

We identified a set of 1,476 genes on the Affymetrix Human 6800 microarray that were consistently expressed across a range of tissues at a variety of expression levels. These 1,476 genes were then used to normalize the data from different microarray experiments. Briefly, the expression levels of the preselected genes were scaled to a "standard experiment" and the geometric mean of the scaling factors was calculated. This value served as the normalization factor for all genes represented on the microarray.

The fold change between the level of expression of any two genes on two different microarrays was calculated according to the protocol supplied by Affymetrix. Subsequently the relative expression level () was determined using the following algorithm, in which T is the treatment expression value; C is the control value, and  is the fold change.

Images of scanned Affymetrix GeneChips were processed using the software and parameter settings suggested by the manufacturer (Affymetrix). The processed fluorescence values were placed into a table containing m rows, which represented the m genes, and n columns that corresponded to the n replica experiments. The n replica experiments were considered to be independent. Using this assumption, each of the n replicates was compared with the others in a pairwise manner. This process resulted in a large number of replicate pairs. The expression values for each pair were plotted against those of other pairs in a scatter plot in which the distribution of points was around the x = y diagonal. The data from these scatter plots were combined for all genes and rotated by /4. This manipulation transformed the x = y axis from the latter plot to the ln(I) = 0, which represents the signal-to-noise distribution derived from comparison of all replicates. To determine a signal-to-noise profile, a sliding histogram algorithm along the y axis was applied (15). A threshold was set at 80% of the maximum noise value. Genes with intensities below this threshold value were considered within the experimental noise.

TaqMan Reverse Transcriptase Polymerase Chain Reaction

cDNA was reverse transcribed from 10 µl of total RNA using Multiscribe reverse transcriptase (RT) (Applied Biosystems, Foster City, CA) in a 50-µl reaction volume with random hexamer as the primer. A negative control reaction lacking RT was also performed for each RNA sample. TaqMan primers and probe were designed using Primer Express software (version 1.0). TaqMan polymerase chain reaction (PCR) reactions were performed in MicroAmp Optical 96-well Reaction Plates and Optical Caps. The quantity of 5 µl of each cDNA sample and standard curve sample was amplified by TaqMan PCR in a reaction volume of 25 µl. Each 25-µl reaction consisted of 1 × TaqMan Universal PCR Master Mix; 5.5 mM MgCl₂; 0.8 mM each of deoxyadenosine triphosphate, deoxycytidine triphosphate, and deoxyguanidine triphosphate; 1.6 mM deoxyurdine triphosphate; 0.625 U AmpliTaq Gold DNA polymerase; 0.25 U AmpErase uracil N-glycolase; 200 nM forward and reverse primer; 100 nM TaqMan probe; and 5 µl template. Amplification was as follows: 50°C for 2 min and 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. Fluorescence emission was monitored at every cycle using the ABI PRISM 7700 Sequence Detection System, and this information was automatically converted to amplification plots using the ABI PRISM 7700 Sequence Detection System software. The data were quantitated by extrapolation from the standard curve, normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and the means ± standard deviation were plotted.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Eosinophils can readily be isolated from human peripheral blood and inhibited from entering apoptosis by the addition of IL-5 to the culture medium. Frequently, more than one method is necessary to confirm the apoptotic status of a cell population. Figure 1 shows the change in two classic apoptotic markers, phosphatidylserine externalization and caspase 3 activity, in eosinophils after titration of IL-5 from 0 to 10 pM in the culture medium. In both experiments, approximately 1 to 10 pM IL-5 appeared sufficient to inhibit apoptosis by > 95% (Figure 1). Higher concentrations of IL-5 did not completely inhibit caspase 3 activity, suggesting that the cells continue to apoptose (Figure 1B). This result may explain the observation that terminally differentiated eosinophils are viable in culture for only up to 7 to 10 d. After 24 h in the absence of IL-5, nearly 50% of the cells were apoptotic by annexin V binding to phosphatidylserine (Figure 1A). In contrast, nearly 90% of the cells were apoptotic by 48 h (Figure 1A). Interestingly, in our hands, apoptosis was maximal at 24 h after isolation using caspase 3 as the marker, consistent with this being an earlier stage marker of apoptosis (data not shown).

View larger version (16K): [in this window] [in a new window]   Figure 1.   (A) Primary eosinophil apoptosis after treatment with different doses of IL-5 after 24 h and 48 h in culture. Apoptosis was measured by phosphatidylserine externalization using an annexin V antibody. (B) Primary eosinophil apoptosis after treatment with different doses of IL-5 after 24 h in culture. Apoptosis was measured by caspase 3 activity. Samples were analyzed in duplicate; mean values ± range are shown.

In an attempt to identify genes that are associated with driving cells into or rescuing cells from apoptosis, microarray transcription profiles were obtained for human eosinophils in the presence or absence of IL-5. To focus on primary changes, we harvested the eosinophil mRNA for microarray profiling immediately before and 1 h after the addition of IL-5. To ensure that the expression changes were directly related to IL-5, a treatment control sample (no IL-5) was set up and harvested at 1 h (see MATERIALS AND METHODS for details). Although additional time points would have broadened the scope of the study, this was prohibited by the difficulty of obtaining subjects to harvest the required number of eosinophils for microarray.

Approximately 2,409 eosinophil genes were detected present on the Affymetrix chip before and after exposure to IL-5 (data not shown). From the absolute expression data, the fold change (treatment versus control) was calculated for each gene on the array. For those genes whose expression was undetectable in one population and present in the other (e.g., induction of expression from absent to present), the "absent" value was set to the noise level of the chip. Thus, a highly conservative method was used which calculated the minimum value for the fold change for genes regulated in this manner. Biologically independent experiments in at least triplicate were performed and the mean fold change values shown. Figure 2 presents those genes whose expression changed by more than 2-fold after addition of IL-5 to eosinophils. The majority of these genes (73) were upregulated, whereas only seven genes were downregulated. These genes fell into one of several categories with respect to predicted biologic function; adhesion, e.g., intercellular adhesion molecule (ICAM)-1, CD24; recruitment and activation, e.g., IL-8, extracellular regulated kinase (ERK)-3, CCR-1, CD69; and apoptosis/survival, e.g., Pim-1, early growth response (EGR)-1.

View larger version (21K): [in this window] [in a new window]   Figure 2.   The bar graph represents those genes regulated >2-fold in primary eosinophils after IL-5 treatment for 1 h, relative to an untreated control sample prepared at 1 h. The black bars represent the mean fold changes from three independent experiments. The error bars represent the standard error for each gene. All genes were annotated using available public databases. Where the gene name is absent, no annotation was available from public databases.

The three most strongly upregulated genes were EGR-1, CD69, and CCR-1. Interestingly, two other members of the EGR zinc finger transcription factor family, EGR-2 and EGR-3, were also upregulated by IL-5. Proliferative responses and the proteins that regulate these processes are often associated with antiapoptotic effects. Previously, both CD69 and CCR-1 were found upregulated in eosinophils by IL-5 (see 2 and references therein). The two most strongly downregulated genes were CCAAT/enhancer binding protein (C/EBP) and regulator of G-protein signaling (RGS)-2. RGS-2 is a member of a diverse family of proteins that share a common motif that interacts directly with the G- subunit of G-protein coupled receptors (16). Interestingly, the interplay of C/EBP family members has been reported to regulate the terminal differentiation of a number of hemapoietic lineages, where C/EBP appears to play a particularly important role in eosinophil and neutrophil lineage commitment and maturation (17).

Although IL-5 is a potent survival factor for eosinophils, it has also been shown to regulate genes associated with other physiologic processes, such as adhesion, migration, and activation. For example, a number of other genes, including IL-8 and ICAM-1, were found regulated in eosinophils by us and by others (reviewed in 2). However, these proteins have not been shown to play a direct antiapoptotic role in either eosinophils or other cell types (2, 5). In addition, microarray experiments of this type are likely to identify bystander genes. We hypothesized that genes whose regulation was similarly coordinated in a distinct cellular model, whose survival was also dependent on IL-5, would have a higher probability of having an antiapoptotic function compared with the remaining pool of regulated genes from either cellular model.

TF1.8 cells are derived from erythroleukemia and demonstrate a dependence on IL-5 for their survival (Figure 3). Physiologically, this system responded to IL-5 in a way similar to eosinophils, as illustrated by the concentration of IL-5 (5 to 10 pM) required to maximally inhibit apoptosis in these cells. Further, similar to eosinophils, complete inhibition of apoptosis was not observed even at the highest concentrations of IL-5. Lowering the IL-5 concentration on TF1.8 cells to 0.01 pM or below effectively prevented this cytokine from inhibiting apoptosis (Figure 3).

View larger version (14K): [in this window] [in a new window]   Figure 3.   The level of apoptosis, as measured by caspase 3 activity, in TF1.8 cells at different doses of IL-5. Samples were analyzed in duplicate; mean values ± range are shown.

Microarray transcription profiles were obtained for TF1.8 cells at 0, 1, 4, and 8 h after the withdrawal of IL-5 from the culture medium. A treatment control group was also set up. It should be noted that in these experiments, the cells maintained in the presence of IL-5 were the control group. Interestingly, no significant changes in gene expression were observed 1 h after IL-5 withdrawal, and so this time point was not considered further. By contrast, a number of changes were observed at 4 and 8 h after IL-5 withdrawal from TF1.8 cells compared with the respective control populations. The total number of TF1.8 genes detected on the Affymetrix chip was in the range of 2,316 to 2,352 (data not shown).

By comparison with IL-5-stimulated eosinophils, the balance between up- and downregulated genes after IL-5 withdrawal from TF1.8 cells was approximately equal (Figure 4). Of the 30 genes observed to change their expression at 4 h after IL-5 withdrawal, 16 were upregulated and 14 downregulated (Figure 4A); whereas, at 8 h, 57 genes changed their expression level, of which 28 were up- and 29 were downregulated (Figure 4B). Many of the genes whose expression was observed to change at 4 h continued to change at 8 h. However, a number of genes demonstrated no further change at 8 h compared with 4 h or were observed to change at 8 h and not 4 h after IL-5 withdrawal (Figure 4). Temporal changes in gene expression may offer insight into the pathways controlling phenotype change. Interestingly, the TF1.8 cells appeared to take longer to sense and respond to IL-5, illustrated by the absence of significant expression changes at 1 h after IL-5 withdrawal, compared with the response of eosinophils after addition of IL-5. The reason for this difference is unclear, although we suggest that it may be due to the process of withdrawal compared with addition of IL-5 or a consequence of the distinct cell types.

View larger version (25K): [in this window] [in a new window]   View larger version (21K): [in this window] [in a new window]   Figure 4.   (A) The bar graph represents those genes regulated > 2-fold in TF1.8 cells 4 h after IL-5 withdrawal, relative to the 4-h control cells exposed to IL-5. (B) The bar graph represents the genes regulated > 2-fold in TF1.8 cells 8 h after IL-5 withdrawal, relative to the 8-h control cells exposed to IL-5. Black bars represent the mean fold changes from three independent experiments. The error bars represent the standard error for each gene. All genes were annotated using available public databases. Where the gene name is absent, no annotation was available from public databases.

Obviously, the regulation of a gene after addition of IL-5 would be expected to be in the opposite orientation after IL-5 withdrawal. Therefore, we queried the data for those genes that were upregulated by IL-5 in eosinophils and downregulated after IL-5 withdrawal from TF1.8 cells. The genes fulfilling these criteria were Pim-1, DSP-5 (hVH3, B23), CD24, and SLP-76 (Table 1). Both Pim-1 and DSP-5 have previously been reported as antiapoptotic factors, and a strong role can also be evoked for CD24 and SLP-76 in antiapoptotic signaling pathways (see DISCUSSION). A similar query to identify genes downregulated in eosinophils and upregulated in TF1.8 cells by IL-5 did not identify any genes.

To validate the data from the Affymetrix chip experiments, the regulation of Pim-1, DSP-5, CD24, and SLP-76 was confirmed using quantitative real-time PCR (TaqMan). Figure 5A shows the expression of the four genes in primary eosinophils, obtained from a new donor, in the presence or absence of IL-5. All four genes were upregulated greater than 2-fold with IL-5 treatment, consistent with previous observations using the gene chips. From these and other data, we suggest that differences in the absolute fold change values are due to the donor variability. Similar studies were also performed on RNA from TF1.8 cells (Figure 5B). All the genes were shown to be downregulated by TaqMan after IL-5 withdrawal consistent with the Affymetrix chip experiments.

View larger version (22K): [in this window] [in a new window]   Figure 5.   TaqMan quantitative PCR analysis was performed in the presence (+) (grey bars) or absence () (black bars) of the cytokine IL-5. (A) The relative expression of Pim-1, DSP-5, CD24, and SLP-76 in primary eosinophils with or without IL-5 for 1 h. (B) The relative expression of Pim-1, DSP-5, CD24, and SLP-76 in TF1.8 cells 4 and 8 h after IL-5 withdrawal. The expression of all the genes was normalized with respect to GAPDH levels. The error bars represent the variation between duplicate wells from the same cDNA sample.

The distribution of the coregulated genes across a range of tissues, including bone marrow, brain, kidney, liver, and lung, is presented in Figure 6A. Of the four genes, Pim-1 and SLP-76 were most restricted to eosinophils. By contrast, DSP-5 did not show a high level of expression in any of the tissues examined. Although DSP-5 was called absent in kidney and liver, these values were close to the sensitivity threshold of the Affymetrix chips (see MATERIALS AND METHODS) such that the level of restriction could not be accurately measured. Previously, DSP-5 was shown to have the highest level of expression in brain, pancreas, and placenta, and also to be expressed in heart, kidney, and liver. The sensitivity of Affymetrix chips was observed to vary between genes and is in part dependent on the quality of the probe sets on the chip. The increases in expression of Pim-1 and DSP-5 after treatment of eosinophils with IL-5 were 6.5 and 4.3, respectively. Therefore, the level of expression in IL-5-treated eosinophils was in the range of 19- to 184-fold higher for Pim-1 and 3- to 4.3-fold higher for DSP-5 where detected, compared with bone marrow, brain, kidney, liver, and lung.

View larger version (25K): [in this window] [in a new window]   Figure 6.   (A) The relative expression and tissue distribution of the genes Pim-1, DSP-5, CD24, and SLP-76. (B) The tissue distribution and relative expression levels of genes known to exhibit a restricted pattern of expression in eosinophils and/or hemapoietic cells. The genes are: A, 15-lipoxygenase; B, Fyn binding protein; C, hematopoietic cell kinase; D, P-selectin ligand; E, CCR-3, eotaxin receptor; F, CLC protein; and G, CXCR4 chemokine receptor.

To put into context the relative levels and specificity of expression of Pim-1, DSP-5, CD24, and SLP-76 in eosinophils, we queried the data for genes already known to be restricted in their expression to eosinophils (Figure 6B). In agreement with previous reports, the Charcot-Leydon crystal (CLC) protein mRNA was one of the most abundant mRNAs expressed in eosinophils. Even though the mRNA encoding FYN binding protein was not expressed to the same level as the CLC protein, it still exhibited a highly restricted pattern of expression. The remaining genes represented in Figure 6B showed intermediate levels of expression in eosinophils. Pim-1 was expressed toward the lower end of this range and exhibited a restricted expression profile in untreated eosinophils close but not equal to these genes. By contrast, DSP-5 and CD24 were not restricted in their expression to eosinophils.

Figure 7 compares the relative expression of Pim-1, DSP-5, CD24, and SLP-76 across different hemopoietic cell types. This analysis indicates that the expression of CD24, SLP-76, and Pim-1 is particularly high in, but not exclusive to, eosinophils. Of particular note is that the relative expression of Pim-1, DSP-5, and CD24 appeared to be significantly higher in eosinophils compared with neutrophils. There is also a suggestion that Pim-1 and DSP-5 may also be regulated by other proinflammatory signals in both the B- and T-cell environments.

View larger version (30K): [in this window] [in a new window]   Figure 7.   The relative expression of the genes Pim-1, DSP-5, CD24, and SLP-76 in different hemopoietic cells.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

This report describes how two model systems have been integrated to identify genes that play a role in the survival (and apoptosis) of eosinophils in response to IL-5, a key cytokine influencing hematopoietic cell development. We believe that isolated primary human eosinophils represent the most relevant cellular model to study the effects of human IL-5. This view is supported by a wealth of literature characterizing the effects of IL-5 on leukocytes and eosinophils in particular (4, 5, 10, 12). However, eosinophils have a number of drawbacks, not least of which is the ability to isolate significant numbers of fresh, highly pure cells. Consequently, gaps exist in our knowledge of IL-5 signaling pathways and precisely how apoptosis is prevented. We applied microarray technology to help identify these gaps and refined the analysis by integrating a second cellular model, also dependent on IL-5, into the study. The effect of IL-5 on TF1.8 cells and the induction of apoptosis after its withdrawal was characterized and shown to have similarities to primary eosinophils.

Considering the number of phenotypic changes regulated by IL-5, it was surprising to find relatively few regulated genes. Of the 80 genes observed to change in eosinophils, nine are known transcription factors and at least 13 more play critical roles in various transcription and mRNA processing functions. Thus, we suggest that many of the phenotypic changes promoted by IL-5 are through secondary responses controlled by, for example, the transcription proteins identified herein.

A number of the genes identified in this study were previously shown as regulated in eosinophils by IL-5 (reviewed in 2, 5, and 12). Many of these genes have also been associated with known eosinophil functional responses such as recruitment and activation (CCR-1, CD66, CD44, IL-8). The ability to confirm previous studies was considered to be an essential component of the validation process for this microarray platform. Additional regulated genes in both eosinophils and TF1.8 cells were also observed after IL-5 stimulation or withdrawal, respectively. Although not the major focus here, many of these genes have no known role in the physiology of the eosinophil and offer interesting opportunities for further study. The hypothesis tested in this study was that genes coregulated in both eosinophils and TF1.8 cells would have a higher probability of having an antiapoptotic function compared with the remaining pool of regulated genes. This hypothesis was supported by the data that identified four genes (SLP-76, Pim-1, DSP-5, and CD24) which can be closely linked to antiapoptotic mechanisms.

The protein SLP-76 is an adapter protein that has recently been shown to lie downstream of Lyn and Syk coupling the T-cell receptor (CD3) to its downstream signaling cascade (18). SLP-76 has been shown to interact after phosphorylation by Lyn or Syk, with Grb-2 via its proline-rich domain and with the nucleotide exchange factor Vav (19, 20). Lyn and Syk are members of the Src family of protein tyrosine kinases and together with Jak-2 are also coupled to the common -subunit of the IL-5 receptor (21). This common -subunit is shared with the IL-3 and GM-CSF receptors. Consequently, Lyn and Syk play critical roles in conducting the IL-5 response and have been shown using antisense oligonucleotides to be partly responsible for the survival effects of IL-5 in eosinophils (24, 25). Therefore, it is not unreasonable to suggest that SLP-76 lies downstream of Lyn and Syk in eosinophils contributing to the antiapoptotic response.

The murine homologue of Pim-1 has been shown to act as a survival factor to inhibit apoptosis in myeloid cells deprived of cytokines (26). The mechanism by which Pim-1 functions is obscure, although it was reported to act in part, but not exclusively, through the Bcl-2 family of proteins. These data fit well with a number of reports within the literature that show a link, albeit variable, between the Bcl-2 family of proteins and eosinophil apoptosis (27). Recently, Pim-1 has also been shown to protect hematopoietic FDC cells from genotoxin-induced apoptosis (30). Interestingly, Pim-1 is regulated in murine Nb2 lymphoma cells by signal transducer and activator of transcription (STAT) proteins and is coupled to the common -subunit of the IL-3 receptor by Jak-2 (31). Others have suggested that Jak-2 (and Lyn) signals converge on the Raf-1 pathway, inasmuch as Raf-1 has been shown to be critical for eosinophil survival (5). However, on the basis of our transcription profiling studies, we suggest that Pim-1 lies downstream of the Jak-STAT proteins and propose a signaling cascade for the IL-5 receptor similar to that reported for the IL-3 receptor (31).

The mitogen-activated protein kinases (MAPKs) are a family of serine/threonine kinases involved in the regulation of a wide range of cellular responses, including cell proliferation, differentiation, and survival (32). The activation of Jun amino-terminal kinase and p38 MAPKs is generally associated with the promotion of apoptosis, whereas the MEK-ERK kinases promote survival (33). DSP-5 is a member of a growing family of dual-specificity phosphatases acting on the MAPK superfamily (34). Although DSP-5 has been shown to inactivate ERK-1, it is unclear whether ERK-1 is the natural substrate of this enzyme in vivo or whether DSP-5 has primarily a pro- or antiapoptotic bias (35, 36). From the tissue distribution profile of the mRNA, it would appear that the function of DSP-5 is not exclusive to the eosinophil; however, further investigation of this interesting family of proteins is in progress.

The role evoked for CD24 in IL-5-induced survival is likely to be indirect. CD24 is heavily glycosylated and anchored to the plasma membrane by glycosyl phosphatidylinositol enabling it to be coupled to the intracellular signaling pathways via Lyn kinase (37, 38). The fact that CD24 is coupled to Lyn kinase, which is the same kinase that signals the IL-5 survival response, suggests a way in which this receptor may enhance the survival of eosinophils. Interestingly, CD24 is also a ligand for P-selectin (39); and the process of adhesion, albeit not specifically via P-selectin interactions, has previously been shown to stimulate survival of eosinophils (for review, see 3).

Profiling of eosinophil mRNA has identified a number of genes and promoted their role in the survival mechanisms of this cell. Further studies to elaborate on the functional biology of these genes are underway. This report has also identified a number of genes regulated by IL-5, currently unconnected to eosinophil survival processes. Some of these genes have previously been shown to play a role in other aspects of eosinophil biology, such as adhesion or activation, whereas the function of others remains unknown.

Microarrays are proving to be a powerful tool to highlight genes of potential interest in many different biologic paradigms. The technical reproducibility and sensitivity of the systems are improving and all the cellular responses that we have investigated have shown consistent changes in their gene expression profiles. Currently, the hurdle faced is the routine interpretation of this information to identify genes that are causal and specific to the phenotypic change of interest and not bystanders, artifacts, or part of normal cellular fluctuations. One method we routinely employ to filter out nonspecific gene changes is to combine different cellular models focused on the same experimental paradigm and to look for the common (or distinct) changes. As demonstrated herein, this approach can be used to focus attention on a subset of genes for further investigation which is more likely to contain genes relevant to the phenotypic change under study. Combination of these data with tissue distribution analysis, readily generated by microarrays, can add another filter suggesting possible exclusivity or non-exclusivity of function for genes of interest. The ability to accurately focus on a small subset of genes is critical, because follow-up studies to validate (or not) the function of these genes remain relatively time-consuming and labor-intensive.