Journal of Immunological Methods 303 (2005) 122 – 134 www.elsevier.com/locate/jim
Research paper
A human ImmunoChip cDNA microarray provides a comprehensive tool to study immune responses T. Nikula a,d,*, A. West a,1, M. Katajamaa a,d,1, T. Lo¨nnberg a,1, R. Sara a, T. Aittokallio b, O.S. Nevalainen c, R. Lahesmaa a a Turku Centre for Biotechnology, FIN-20521, Turku, Finland Department of Mathematics, University of Turku, FIN-20014, Turku, Finland c Department of Information Technology, University of Turku, FIN-20520, Turku, Finland d National Graduate School in Informational and Structural Biology, FIN-20521, Turku, Finland b
Received 23 September 2004; received in revised form 2 March 2005; accepted 10 June 2005 Available online 29 June 2005
Abstract DNA microarray technology has developed rapidly in recent years and has become an essential tool, providing novel approaches to biomedical research. In this paper, we describe a self-designed ImmunoChip cDNA array for immunological research. With a comprehensive selection of genes of interest, we can focus on key signalling pathways and molecular mechanisms at relatively low cost compared to commercial platforms which are usually targeted at global screening of gene expression. To validate the efficiency of the ImmunoChip, we studied T helper cell polarization to functionally distinct subsets (Th1 and Th2). We also developed a tool for quality control of cDNA microarrays that assesses the technical quality of an ImmunoChip. The information produced with the quality control tool is shown to be valuable for extracting correct information from cDNA microarrays. Gene expression measurements with ImmunoChip are in agreement with the results obtained using oligonucleotide microarrays and with published quantitative RT-PCR data. The ImmunoChip provides reliable measurements and gives new insights into various aspects of human immune responses. D 2005 Elsevier B.V. All rights reserved. Keywords: Immunology; Microarray; T helper cell differentiation; Quality control
1. Introduction
* Corresponding author. Turku Centre for Biotechnology, FIN20521, Turku, Finland. Tel.: +358 2 333 8013; fax: +358 2 333 8000. E-mail address:
[email protected] (T. Nikula). 1 These authors contributed equally to this article. 0022-1759/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.jim.2005.06.004
In the past few years microarray technologies have become essential tools in biomedical research and several different applications using either cDNA or oligonucleotide based microarrays for the simultaneous monitoring of thousands of mRNA transcription levels have been published (Duggan et al., 1999;
T. Nikula et al. / Journal of Immunological Methods 303 (2005) 122–134
Lockhart et al., 1996; Schena et al., 1995). A growing number of sequences in the databases, together with further development of the techniques, continue to provide improved array formats and new commercial applications. However, despite the fast progress in the field of bioinformatics, the amount of both biological and technical replicates required for reliable biological results increases the number of arrays needed for clinical trials or disease profiling and screening. Utilization of the commercial platforms and large arrays is relatively expensive, especially when multiple samples are analyzed. Instead of using arrays for global screening, self-designed targeted arrays may be a method-of-choice for focused studies, often following the genome-wide gene expression profiling. Certain specialized arrays for different purposes– mostly for cancer studies but also for immunology– have already been described (Alizadeh et al., 1999; Brown and Botstein, 1999; Fitzgerald and Guimbellot, 2001; Khan et al., 1999; Lorenz et al., 2003). The self-designed ImmunoChip cDNA microarray described in this paper has been developed for studies on signal transduction and molecular networks in the pathogenesis of immune-mediated diseases. The aim of the present work was to evaluate the first version of the ImmunoChip containing approximately 2000 different genes by applying it to identify differentially expressed genes in T helper cell subtypes (Th1 and Th2). Polarization of naı¨ve CD4+ T cells to Th1 and Th2 cells, whose functions have been well characterized, is triggered by activation through TCR/ CD28 and directed by IL-12 and IL-4, respectively (Glimcher and Murphy, 2000; Mosmann et al., 1986; Romagnani, 1996; Seder and Paul, 1994). Because of the importance of T helper cells in immune responses, studying them at different stages of activation and differentiation is indispensable. The use of commercially available oligonucleotide microarrays has already been shown to be an effective way to explore gene expression in T cells (Chen et al., 2003; Chtanova et al., 2001; Hamalainen et al., 2001; Lund et al., 2003a; Rogge et al., 2000). Here, the validation of the self-designed cDNA microarray was based on two sets of experiments. First, gene expression profiles of CD4+ lymphocytes polarized towards Th1 and Th2 cells for 7 and 14 days were studied. Second, the effects of 48 h cytokine stimulation on human CD4+ T cells were explored. We also assessed the quality and
123
reproducibility of replicate spots on the slide and slideto-slide variation between biological replicates using a tool implemented by our group to control the quality of self-made DNA microarrays.
2. Materials and methods 2.1. Array construction and sequencing The human ImmunoChips were printed on glass slides at the Finnish DNA Microarray Centre, Turku Centre for Biotechnology (http://microarrays.btk.fi/). Sequences to be spotted were picked from Research Genetics’ Human Sequence Verified 40 K Library or cloned in-house. The sequences represent roughly 2000 genes implicated in lymphocyte activation or differentiation including cytokines, chemokines and their receptors, transcription factors and genes involved in signalling, apoptosis and cell cycle regulation. The total number of different spots on the array is 2400; some of the spots represent different cDNA clones from the same gene and some are empty spots that do not contain DNA. The spots are located on the array in three columns and four rows (Fig. 1). There are 12 sub-arrays each containing 600 (24 25) different spots. The sub-arrays are identical in the same row but different from each other in the same column. Thus, the three technical replicates of each sequence are placed on different parts of the slide. All the cDNA clones giving an interesting expression signal on the array were sequenced with Applied Biosystems ABI PRISM 3100 Genetic Analyzer using universal primers. A total of 490 cDNA clones, representing a quarter of the genes on the array, were selected from the same plates used for spotting. Sequencing results were blasted against databases to verify the results. 2.2. In vitro polarization and activation of Th1 and Th2 cells In the first set of experiments, three separate trials were performed using cells extracted from neonatal cord blood of three individuals. Mononuclear cells were isolated using Ficoll Paque Isolation (Amersham Pharmacia Biotech, Uppsala, Sweden) and CD4+ lymphocytes were recovered with magnetic beads (Dynal
124
T. Nikula et al. / Journal of Immunological Methods 303 (2005) 122–134
Fig. 1. Technical design of an ImmunoChip microarray (adapted from Aittokallio et al., 2003). Each 25 24 sub-array contains 600 probes, which are spotted three times using the same printing tip. The number of spots within an array row is 1800. Each of the three array columns contains 2400 different probes, representing all the genes monitored with the microarray.
Biotech, Oslo, Norway). The CD4+ cells were stimulated in the presence of 100 ng/ml PHA (Murex Diagnostics, Chatillon, France) and irradiated (60 Gy) with CD32-B7 transfected fibroblasts (Yang et al., 1995). The final densities of the cell cultures were 1 106 CD4+ cells and 0.5 106 feeder cells in 1 ml. The cells were grown in Yssel’s medium (Irvine Scientific, Santa Ana, CA) supplemented with 1% AB-serum (Red Cross, Helsinki, Finland). Differentiation was primed with 2.5 ng/ml of IL-12 (R&D Systems, Minneapolis, MN) for Th1 cells and with 10 Ag/ml of anti-IL-12 (R&D Systems) and 10 ng/ml IL-4 (R&D Systems) for Th2 cells. To enhance the proliferation of the lymphocytes, 40 U/ml of IL-2
(R&D Systems) was added to the cultures after 48 h. Cells were split at days 3 and 10. At day 7, part of the cells was collected and the rest was re-stimulated and cultured for another 7 days as described above. At 7 and 14 days, half of the collected cells were activated by plate-bound aCD3 (500 ng/well for coating) and 0.5 Ag/ml soluble aCD28 (Immunotech, Marseille, France) for 6 h (Lund et al., 2003a); the other half was treated similarly without antibodies. Polarization efficiency was controlled by intracellular FACS staining of IFNg and IL-4. For the second set of experiments, separate cell cultures were prepared as previously described (Lund et al., 2003a). Briefly, mononuclear cells were isolated
T. Nikula et al. / Journal of Immunological Methods 303 (2005) 122–134
125
from human neonatal cord blood and the CD4+ cells were further separated as described above. One aliquot of the cells was reserved as an untreated sample of naı¨ve T cells. The rest of the cells were cultured for 48 h. Cells were stimulated with aCD28 and platebound aCD3. Commitment to Th1 and Th2 lineages was induced using IL-12 and IL-4 with anti-IL-12, respectively, in the absence or presence of TGF-h (3 ng/ml, R&D Systems). To obtain sufficient amount of material for the hybridizations, the isolation and culture of the cells were repeated once. Harvested cells from both cultures were pooled before further processing.
slip. Hybridizations were performed overnight at 65 8C in hybridization chambers (Corning), with humidity maintained by 3 SSC. After hybridization, the slides were dipped in 0.5 SSC/0.1% SDS to remove the coverslip. Slides were then washed by agitation for 15 min in 0.5 SSC/ 0.1% SDS, 15 min in 0.5 SSC/0.01% SDS, 2 min in 0.06 SSC and 1 min in 0.06 SSC again. All washing steps were performed at room temperature. Finally, the slides were dried by centrifugation (1 min, 700 rpm).
2.3. RNA extraction and labelling
In the first set of experiments, CD4+ T cells were isolated from neonatal cord blood from three individuals. The cells were differentiated in vitro to Th1 and Th2 cells. The samples of Th1 and Th2 cells before and after activation were hybridized once against the reference sample at two time points (7 and 14 days). For each treatment, we thus have three biological replicates giving a total number of 24 hybridizations. As there are three technical replicates on the array, 72 expression measurements are made for a single sequence in this experimental setting. In the second set of experiments, we pooled the cord blood CD4+ T cells obtained from three individuals to obtain sufficient RNA for the hybridizations. The cells were either untreated or treated with cytokines for 48 h as indicated. Pooled TGF-h-treated samples were hybridized three times against the reference sample, which together with three technical replicates gives a total of nine expression measurements for a single sequence. Those 48 h samples that were not treated with TGF-h were hybridized twice against the reference sample resulting in six individual data points.
Total RNA from the cultured cells was isolated once with Trizol reagent (Gibco BRLR) and further purified with Qiagen’s RNeasy columns. Peripheral blood mononuclear cells from 13 normal donors of the Finnish Red Cross were isolated by Ficoll gradient centrifugation (Amersham Biosciences) to create a reference sample for all hybridizations. Total RNA from the reference sample was isolated twice with Trizol reagent (Gibco BRLR). In addition to measurement of the A 260/A 280 absorbance ratio, agarose gel electrophoresis was performed to determine the quality of RNA. For each hybridization, equal amounts of total RNA from the reference pool and the target sample were taken and labelled directly during cDNA synthesis. The reference sample was labelled with FluoroLinkk Cy3-dUTP (Amersham Pharmacia Biotech) and the samples of interest with Cy5-dUTP (Amersham Pharmacia Biotech). The labelled reference and the desired sample were mixed with hybridization reagents in one tube before hybridization. The labelling protocol is available through a website of the Finnish DNA Microarray Centre (http://microarrays. btk.fi/public/Resources.shtml). 2.4. Hybridization The slides were fixed with UV light (90 mJ) before hybridization and prehybridized for 30 min at 50 8C with a buffer containing 5 SSC, 0.1% SDS and 1% BSA. The hybridization mixture (http://microarrays. btk.fi/public/Resources.shtml) containing the labelled cDNA was placed on the array under a glass cover-
2.5. Experimental design
2.6. cDNA microarray data analysis Separate images for Cy3 and Cy5 dyes were acquired using a ScanArrayk 5000 laser-scanning microscope (Packard BioSciences), and the images were then combined for data reduction in QuantArrayR microarray analysis software from the same company. Gene transcript levels were determined from the fluorescence intensities of the scanned data image files. Systematic variation in the measured intensity levels
126
T. Nikula et al. / Journal of Immunological Methods 303 (2005) 122–134
was removed using the intensity-dependent normalization with robust scatter plot smoother function, as described previously (Yang et al., 2002). In the first set of experiments, the expression levels of each sequence on the array were compared between Th1 and Th2 cells. A gene was considered as differentially expressed between the groups compared if at least a 1.5-fold difference in the average expression level was observed and the difference was also statistically significant at risk level p b 0.05. The statistical significance was assessed on the basis of the two-sided paired t-test statistics computed on data groups consisting of three replicates from the three independent hybridizations. The second set of experiments consisted of gene expression data from cells polarized for 48 h to Th1 or Th2 direction and treated with TGF-h. Analysis of the data using the same criteria as used for the first set of experiments would have resulted in a large amount of results. Therefore, different detection limits were set for this experiment. The expression level of a gene was considered to be significantly altered if the observed change in transcript level was two-fold or greater, and if the sequence in question had a p-value less than 0.05. These rather stringent criteria were selected to limit the resulting number of differentially expressed genes. 2.7. Tool for quality control We developed a tool for hybridization quality control, called RRPlot (http://bioinformatics.btk.fi/ public/rrplot). Although this tool has been designed for quality control of microarrays manufactured at the Finnish DNA Microarray Centre, it can also be applied to any array that fulfils two prerequisites. First, the array design should contain at least two identical sets of replicate spots, since many of the quality measures are based on observing differences between two sets of technical replicates. Second, the array should include some spots that do not contain DNA. These bempty spotsQ are used both to evaluate the quality of the spotting and to measure the background intensity levels. The main idea of the RRPlot is not to give a binary decision concerning the quality of a microarray, but to provide the user with a set of visuali-
zations and measures, which help in determining the quality of the hybridization. Data processing in RRPlot consists of two main stages: (i) data filtering and normalization; (ii) drawing of plots indicating the quality of the hybridization. Since spots with low intensity are usually excluded from further analysis, they can also be excluded from the quality control in RRPlot. Data filtering is used to remove the spots with intensities near background values of the hybridized array. A cut-off value for filtering is calculated based on the spots containing no DNA. The remaining data is normalized with the LOWESS method (Yang et al., 2002). In the next processing stage, several plots are drawn from the data. These can be categorized into three groups, which show the distribution of spot and background intensities across the whole array, the differences between technical replicates (Fig. 2), and intensities of the predefined empty spots (http://bioinformatics. btk.fi/public/rrplot/RRPlot_manual.pdf). 2.8. Affymetrix arrays Total RNA of the 7 day samples (activated and untreated Th1 and Th2 cells) from two different individuals was hybridized to HG-U133A oligonucleotide arrays from Affymetrix. Samples were prepared with the Affymetrix small sample preparation according to the instructions and recommendations provided by the manufacturer (Affymetrix, Santa Clara, CA, USA; http://www.affymetrix.com/). The data analysis was performed at three consecutive levels. At the detection level, each probe was assigned a status of present, absent or marginal. At the comparison level of the analysis, the signal log ratio between Th1 and Th2 cell samples was determined. Because of the differences between the two platforms, instead of comparing expression levels or signal log ratios, results from the Affymetrix analysis were compared to the results obtained with the ImmunoChip cDNA microarray in terms of the genes found to be differentially expressed between Th1 and Th2 samples. 2.9. Computing inter-platform agreement The inter-platform agreement for comparisons between ImmunoChip and Q-RT-PCR or ImmunoChip
T. Nikula et al. / Journal of Immunological Methods 303 (2005) 122–134
127
Fig. 2. The RRPlot quality control program produces an overview of the quality of a single hybridization. Logarithmic ratios of three technical replicates are presented as a pair-wise comparison between the replicates. Each comparison is shown in a matrix, which has the same configuration as the replicates on the array. The colour of a matrix element expresses the difference between the values of the replicates. The colour scale on the right shows the relationship between the colour and absolute difference in ratios. The selected example images indicate good consensus between replicates 1 and 2, whereas replicate 3 is most different from these two and the differing observations are non-randomly distributed on the array.
and Affymetrix data was calculated with the SAS system (V.8). With each platform the resulting genes were divided into three categories (expressed more by Th1 cells, equal expression, and expressed more
by Th2 cells) and unweighted Cohen’s j-statistics with exact one-sided p-values were calculated between the genes common to both platforms (Gwet, 2002).
128
T. Nikula et al. / Journal of Immunological Methods 303 (2005) 122–134
3. Results 3.1. ImmunoChip quality From the spotted sequences on the ImmunoChip, we sequenced 25%, i.e. 490 clones, that were found to be differentially expressed. The majority of these genes (407) were confirmed by sequencing, whereas the rest (83) were found to represent a gene other than the annotated one. Thus, 83% of the Research Genetics’ cDNA clones were correctly annotated based on sequence comparisons. That is fairly consistent with previous reports on the fidelity of Research Genetics’ cDNA clones and other IMAGE (Integrated Molecular Analysis of Genomes and their Expression) consortium cDNA clones (Halgren et al., 2001; Taylor et al., 2001).
The average correlation between technical replicates as calculated from raw expression data on the array was found to be fairly good (Pearson q = 0.895), but filtering and normalization of the data still increased it markedly (Pearson q = 0.930). Surprisingly, normalization without filtering decreased the average correlation between three technical replicates (Pearson q = 0.578) clearly demonstrating the importance of the filtering step. Data generated with RRPlot are illustrated in Fig. 3. Cut-off levels for the intensities are similar for both channels in all arrays and do not seem to affect correlations between technical replicates or the amount of technical replicates that do not meet the criteria required to be included in the analysis. The reliability of the expression data obtained by the ImmunoChip was also assessed by comparison with previously published quantitative RT-PCR (Q-
Fig. 3. Evaluation of the hybridization quality of the first set of experiments using statistical data obtained from the RRPlot quality control tool. The x-axis stands for 24 hybridizations in this experimental set. Pearson’s correlations between technical replicates on the arrays (1–24) are on the y-axis; the raw data obtained from the arrays and the lowes normalized data before and after filtering in the RRPlot are represented with solid lines. Proportions of technical replicates that have been filtered out after quality control are represented with a dashed line. Cut-off intensities (10e 3) for both Cy3 and Cy5 channels are represented with dotted lines.
T. Nikula et al. / Journal of Immunological Methods 303 (2005) 122–134
RT-PCR) results (Table 1) (Hamalainen et al., 2000, 2001; Lund et al., 2003a,b). As indicated in Table 1, the results of the present study are in agreement with the previous results at both 7 (j = 0.56, p = 0.0011) and 14 days (j = 0.47, p = 0.0082). However, the results from j-statistics should be interpreted with caution since the number of common genes used in the calculation was rather small. 3.2. Differences in gene expression of T helper cell subtypes after 7 and 14 days of polarization, and 6 h of activation To validate the self-designed ImmunoChip cDNA microarray we characterized the gene expression between T helper subsets 1 and 2 by using two experimental setups. In the first set of experiments, CD4+ cells were cultured for 7 or 14 days in polarizing conditions and half the cells were further activated via the CD3 and CD28 receptors. Altogether, the expression of 134 genes was detected as differentially expressed between Th1 and Th2 cells. Based on functional classification, the majority of these genes Table 1 Results of the present study were in agreement with the previously published Q-RT-PCR results at both 7 (j = 0.56, p = 0.0011) and 14 days (j = 0.47, p = 0.0082) Gene
RT-PCR (7 daysa)
ImmunoChip (7 days)
RT-PCR (14 daysb)
ImmunoChip (14 days)
DUSP6 STAT4 FGFR CCR2 IFNGRB CCR4 GADD45B IL-18R CXCR3 TNFA RANTES IL12RB2 SLAM CCR1 IFNc
Th2 Th2 – – – – Th1 Th1 Th1 Th1 Th1 Th1 Th1 Th1 Th1
– Th2 – – – – – – Th1 Th1 – Th1 Th1 Th1 Th1
– Th2 – Th1 Th2 Th2 – – – Th1 Th1 Th1 Th1 Th1 –
– Th2 Th2 Th1 – Th2 – Th1 Th1 – Th1 Th1 Th1 – Th1
Th1 or Th2 indicates stronger expression in either cell type, otherwise no difference between cell types was detected (–). A 1.5-fold difference with p b 0.05 was considered to be differentially expressed in the ImmunoChip data. a Lund et al. (2003a,b). b Hamalainen et al. (2000, 2001).
129
were either enzymes or cell surface molecules. The rest were distributed quite evenly between the following categories: cytokines, chemokines and other ligands; genes related to transcriptional regulation; structural molecules and intracellular trafficking genes; unclassified genes (Fig. 4). As expected based on previous DNA microarray results describing Th cell polarization (Hamalainen et al., 2001; Lund et al., 2003b; Rogge et al., 2000), the amount of upregulated transcripts was higher in Th1 samples than in Th2 samples (Table 2A). As an exception, CD3/ CD28 activated Th2 cells at day 14 showed more upregulated transcripts than Th1 cells. In Fig. 5, all differentially expressed genes are presented along with the expression profile over the treatments. Genes with corresponding expression between 7 day Th1 and Th2 samples on both Affymetrix oligonucleotide arrays and ImmunoChip are marked with an asterisk next to the gene abbreviation (j = 0.30, p = 0.0003 for polarized and j = 0.43, p b 0.0001 for polarized and CD3/CD28 activated samples). There are only two genes, IL12RB2 on the Th1 side and STAT4 on the Th2 side, which were equally expressed in all sample types. However, in addition to these two, several other genes previously implicated in T helper cell differentiation were detected. In the polarized samples at day 7, 36 transcripts were more expressed in Th1 cells and 10 transcripts were more expressed in Th2 cells (Table 2A). The differentially expressed genes included IFN-c, SCYA4, IL12RB2, SLAM, CD38, GZMB, TNFA, GBP2 and TNFSF10 in Th1 cells and GATA3, ID2 and STAT4 in Th2 cells as has previously been described (Hamalainen et al., 2001; Lund et al., 2003a; Rogge et al., 2000). After 6 h of CD3/CD28 activation, the amount of genes differentially expressed between Th1 and Th2 cells polarized for 7 days increased to 65. Among the 45 genes up-regulated in Th1 cells, 13 have also been detected in previous array studies: IL12RB2, STAT1, GZMB, GBP2, RANTES, CCR1, TNFSF10, API2, IL18R1, TANK, SCYB10 and PSCDBP (Hamalainen et al., 2001; Lund et al., 2003a; Rogge et al., 2000). Seven out of twenty genes up-regulated in Th2 cells in these conditions have been reported as differentially expressed between Th1 and Th2 cells: IFNGR2, GATA3, ID2, DUSP6, CCR4, STAT4 and FGFR1 (Hamalainen et al., 2001; Lund et al., 2003a; Rogge et al., 2000).
130
T. Nikula et al. / Journal of Immunological Methods 303 (2005) 122–134
Fig. 4. Functional categories of the 134 differentially expressed genes in the first set of experiments. The percentages of these genes classified into six functional categories are illustrated as a pie diagram.
The increase in polarization time from 7 to 14 days results in an increased amount of differentially expressed genes. In total, 44 and 24 up-regulated genes were detected in Th1 and Th2 cells, respectively (Table 2A). At the 14 day time point, Th1 cell expression included IL12RB2, RANTES, CXCR3, SCYA4, GZMB, CCR2, STAT1, GBP2, MIHC/API2, TANK, CASP1, IL18R1, IFN-c and PSCDBP, which have all been implicated in the differentiation process Table 2 The distribution of genes expressed in Th1 and Th2 cells after 7 and 14 days polarization and further activation (A) and up- and downregulated genes after 48 h polarization in the presence and absence of TGF-h (B) (A)
7 days 7 days + 6 h 14 days 14 days + 6 h anti-CD3/CD28 anti-CD3/CD28
Th1 Th2 Total
36 10 46
(B)
Th1 Th1 + versus TGF-h Th0
Up-regulated 8 Down-regulated 12 Total 20
45 20 65
2 8 10
44 24 68
10 31 41
Th2 versus Th0
Th2 + TGF-h
7 13 20
3 11 14
(Hamalainen et al., 2001; Lund et al., 2003a; Rogge et al., 2000). In Th2 cells, four previously detected T cell differentiation genes, GATA3, STAT4, FGFR1 and CCR4, were found. Importantly, CD3/CD28 activation of 14 day polarized cells resulted in a decrease in the amount of differentially expressed genes and more transcripts were detected in Th2 than in Th1 cells. However, the expression of genes previously reported to be differentially expressed by Th1 and Th2 cells such as SCYA4, IFN-c, SCYA3, IL12RB2 and SLAM up-regulated in Th1 cells and FGFR1, IFNGR1/2, STAT4 and ID2 up-regulated in Th2 cells were detected. In addition to those genes previously reported to be differentially expressed by Th1 or Th2 cells, several other genes implicated in lymphocyte proliferation, differentiation and activation were differentially expressed between Th1 and Th2 cells at the time points studied (Fig. 5). 3.3. Gene expression differences after 48 h of culture In the second set of experiments, CD4+ cells were cultured for 48 h under polarizing conditions in the absence or presence of TGF-h. Thus, TGF-htreated samples were compared with polarized cells
T. Nikula et al. / Journal of Immunological Methods 303 (2005) 122–134
131
Fig. 5. Polarization and activation-induced differences in Th1 and Th2 gene expression. Measured gene expression differences are arranged according to the time points and treatments (A–D). The colour intensity scale (E) indicates the signal log ratios between Th1 and Th2 cells. For those genes that are represented on the array by several clones, the mean log ratio is used. In 7 day samples (A and B) an asterisk next to the gene abbreviation indicates that a corresponding result between Th1 and Th2 cells was obtained with Affymetrix oligonucleotide arrays (j = 0.30, p = 0.0003 for polarized and j = 0.43, p b 0.0001 for polarized and CD3/CD28 activated samples). The functional groups of the genes are indicated as numbers (1, cell surface molecules; 2, cytokines, chemokines and other ligands; 3, enzymes and pathway molecules; 4, structural molecules and intracellular trafficking; 5, transcriptional regulation; 6, unclassified).
(Th1, Th2), which in turn were compared with unpolarized activated cells (Th0). To specify the effects of each treatment, the mean expression levels of replicate hybridizations were compared between samples that had received different treatments. In CD3/CD28-activated cells versus non-activated cells, 229 up-regulated and 108 down-regulated
genes met the criteria for differential expression (data not shown). The effects of polarization were considerably more limited (Table 2B). Fig. 6 depicts the gene expression profiles over treatments. The expression of 12 genes was up-regulated and the expression of 8 genes was down-regulated when cells polarized to Th1 and Th2 directions were
132
T. Nikula et al. / Journal of Immunological Methods 303 (2005) 122–134
Fig. 6. The target genes of IL-12, IL-4 and TGF-h. Cells were cultured in the absence or presence of TGF-h. Differentially expressed genes after IL-12 (A), IL-4 (B) and TGF-h (C, D) treatment are colour-coded with green for down- and red for up-regulation. The colour intensity scale (E) depicts the relationship between intensity and log ratio of the difference in expression. For genes with multiple probe sequences the mean log ratio is used. The functional groups of the genes are represented as numbers (1, cell surface molecules; 2, cytokines, chemokines and other ligands; 3, enzymes and pathway molecules; 4, structural molecules and intracellular trafficking; 5, transcriptional regulation; 6, unclassified).
compared. Among the genes differentially expressed SERPINB1, IFN-c, IL12RB2 and GZMB were also up-regulated in corresponding samples analyzed by Affymetrix oligonucleotide arrays (Lund et al., 2003a). Similarly, regulation of STAT1, MX1, LTB, GATA3, GZMB and DUSP6 after polarization to Th2 cells, GZMB and RANTES in TGF-h-treated Th1 cells and RGS1 in TGF-h-treated Th2 cells is in concordance with our previous results obtained by Affymetrix oligonucleotide arrays (Lund et al., 2003a).
4. Discussion We have described and validated a cDNA microarray for studies of human immune responses. The applicability of the ImmunoChip was validated with two different sets of experiments on T helper cell polarization. The quality control tool proved to be valuable for extracting correct information from cDNA microarrays and both validation studies demonstrated clearly the value of the human ImmunoChip as a tool to analyze gene expression. Despite the differences in platform and sample preparation, our results obtained by the human ImmunoChip are supported by previous results (Chen et al., 2003; Chtanova et al., 2001; Hamalainen et al., 2001;
Lund et al., 2003a,b; Rogge et al., 2000). As the comparison between different array platforms is rather complex and also the sample preparation differs in most of the reference studies, we have compared differentially expressed genes manually. In two recent studies (Lee et al., 2003; Yuen et al., 2002) dealing with the congruence between the cDNA and oligonucleotide microarrays, these two platforms were regarded as complementary rather than as competitors. Moreover, the two platforms are suggested to be suitable for mutual validation. We compared the results obtained by the two platforms by analyzing the gene expression of samples from day 7. Samples from two of the three individuals were hybridized on both platforms. As illustrated in Fig. 5, the majority of the genes that are detected by the ImmunoChip are confirmed with oligonucleotide arrays. The other time points used in this study, i.e. 48 h and 14 days, have also been studied previously by our group using oligonucleotide arrays (Hamalainen et al., 2001; Lund et al., 2003a). Although these studies were performed independently, the results are in agreement with the present study. Further confirmation of the results obtained by the ImmunoChip comes from comparisons with the QRT-PCR results (Table 1) indicating the expression of several genes implicated in T helper cell differentiation after 7 or 14 days (Hamalainen et al., 2000, 2001; Lund et al., 2003a,b).
T. Nikula et al. / Journal of Immunological Methods 303 (2005) 122–134
The technical quality of the array was observed to be good and the quality control tool we have developed proved to be very useful in assessing the quality of the hybridizations on ImmunoChip. Expression measurements originating from three replicate spots on the array are extremely valuable for both quality measurements and visualization of the hybridization results, and they can also improve the normalization (Aittokallio et al., 2003; Fan et al., 2004). The RRPlot quality control tool provides a user-friendly visualization of the quality of a single hybridization, where potential problems caused by array handling are quickly localized (Fig. 2). Together with statistical tests, such visualization reveals unwanted factors in the technical replicates and improper data can then be removed from further analysis. As a result we have several choices for handling the data measurements for each sequence: use the median of three good technical replicates; use the average of two remaining good technical replicates; use the only technical replicate accepted; or omit the sequence. In our validation experiment, the hybridizations were found to be of high quality without severe problems in any of the slides. Besides the discovery of genes differentially expressed in response to a particular treatment, gene expression profiling provides a convenient way to elucidate entire pathways. Currently, pathway data are already available through the internet (Dahlquist et al., 2002; Kanehisa and Goto, 2000). Analyzing the gene expression at a pathway level often reveals more than what can be observed by measuring the gene expression differences alone (Mootha et al., 2003). The ImmunoChip provides an excellent focused platform to study the effects of knock down (Hammond et al., 2001) or over-expression on pathways involved in signalling in the cells of the immune system. Furthermore, gene expression data can also be used in combination with other data sources, which allows a thorough analysis (Whitney et al., 2003). To date most of the studies exploiting DNA microarrays to study immunological questions have consisted of relatively few biological and/or technical replicates and few time points. To complement gene expression analysis on a global scale, analysis of the transcripts with immunological relevance provides an excellent tool for detailed follow-up studies. A benefit of this scheme is that it accomplishes at a sufficient level of costs without losing indispensible results. We
133
have used this strategy in a number of studies and found it both feasible and affordable. The current, extended version of the human ImmunoChip contains roughly 5000 genes in three replicates. Genes included in the extended version of the array are primarily selected based on our results obtained by large-scale studies with Affymetrix oligonucleotide arrays (Chen et al., 2003; Lund et al., 2003a, 2004). The selection of genes on the ImmunoChip can be further broadened as new results become available. Together with effective quality control and robust normalization and data analysis methods, the array technology will continue to provide us with new insights into the molecular mechanisms of immune response.
Acknowledgements We thank Arja Reinikainen, Pa¨ivi Junni, Pa¨ivi Haaranen, Eveliina Virtanen and the Finnish DNA Microarray Centre for skilful technical assistance. This work was supported by the Academy of Finland (grant 203 632), National Technology Agency and Turku University Hospital Fund.
References Aittokallio, T., Kurki, M., Nevalainen, O., Nikula, T., West, A., Lahesmaa, R., 2003. Computational strategies for analyzing data in gene expression microarray experiments. J. Bioinform. Comput. Biol. 1, 541. Alizadeh, A., Eisen, M., Davis, R.E., Ma, C., Sabet, H., Tran, T., Powell, J.I., Yang, L., Marti, G.E., Moore, D.T., Hudson Jr., J.R., Chan, W.C., Greiner, T., Weisenburger, D., Armitage, J.O., Lossos, I., Levy, R., Botstein, D., Brown, P.O., Staudt, L.M., 1999. The lymphochip: a specialized cDNA microarray for the genomic-scale analysis of gene expression in normal and malignant lymphocytes. Cold Spring Harbor Symp. Quant. Biol. 64, 71. Brown, P.O., Botstein, D., 1999. Exploring the new world of the genome with DNA microarrays. Nat. Genet. 21, 33. Chen, Z., Lund, R., Aittokallio, T., Kosonen, M., Nevalainen, O., Lahesmaa, R., 2003. Identification of novel IL-4/Stat6-regulated genes in T lymphocytes. J. Immunol. 171, 3627. Chtanova, T., Kemp, R.A., Sutherland, A.P., Ronchese, F., Mackay, C.R., 2001. Gene microarrays reveal extensive differential gene expression in both CD4(+) and CD8(+) type 1 and type 2 T cells. J. Immunol. 167, 3057. Dahlquist, K.D., Salomonis, N., Vranizan, K., Lawlor, S.C., Conklin, B.R., 2002. GenMAPP, a new tool for viewing and analyzing microarray data on biological pathways. Nat. Genet. 31, 19.
134
T. Nikula et al. / Journal of Immunological Methods 303 (2005) 122–134
Duggan, D.J., Bittner, M., Chen, Y., Meltzer, P., Trent, J.M., 1999. Expression profiling using cDNA microarrays. Nat. Genet. 21, 10. Fan, J., Tam, P., Woude, G.V., Ren, Y., 2004. Normalization and analysis of cDNA microarrays using within-array replications applied to neuroblastoma cell response to a cytokine. Proc. Natl. Acad. Sci. U. S. A. 101, 1135. Fitzgerald, D.A., Guimbellot, J.S., 2001. Tailored arrays. Scientist 15, 26. Glimcher, L.H., Murphy, K.M., 2000. Lineage commitment in the immune system: the T helper lymphocyte grows up. Genes Dev. 14, 1693. Gwet, K., 2002. Computing inter-rater reliability with the SAS system. Series: Statistical Methods for Inter-Rater Reliability Assessment, vol. 3. Halgren, R.G., Fielden, M.R., Fong, C.J., Zacharewski, T.R., 2001. Assessment of clone identity and sequence fidelity for 1189 IMAGE cDNA clones. Nucleic Acids Res. 29, 582. Hamalainen, H., Meissner, S., Lahesmaa, R., 2000. Signaling lymphocytic activation molecule (SLAM) is differentially expressed in human Th1 and Th2 cells. J. Immunol. Methods 242, 9. Hamalainen, H., Zhou, H., Chou, W., Hashizume, H., Heller, R., Lahesmaa, R., 2001. Distinct gene expression profiles of human type 1 and type 2 T helper cells. Genome Biol. 2 (RESEARCH0022). Hammond, S.M., Caudy, A.A., Hannon, G.J., 2001. Post-transcriptional gene silencing by double-stranded RNA. Nat. Rev., Genet. 2, 110. Kanehisa, M., Goto, S., 2000. KEGG: Kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 28, 27. Khan, J., Saal, L.H., Bittner, M.L., Chen, Y., Trent, J.M., Meltzer, P.S., 1999. Expression profiling in cancer using cDNA microarrays. Electrophoresis 20, 223. Lee, J.K., Bussey, K.J., Gwadry, F.G., Reinhold, W., Riddick, G., Pelletier, S.L., Nishizuka, S., Szakacs, G., Annereau, J.P., Shankavaram, U., Lababidi, S., Smith, L.H., Gottesman, M.M., Weinstein, J.N., 2003. Comparing cDNA and oligonucleotide array data: concordance of gene expression across platforms for the NCI-60 cancer cells. Genome Biol. 4, R82. Lockhart, D.J., Dong, H., Byrne, M.C., Follettie, M.T., Gallo, M.V., Chee, M.S., Mittmann, M., Wang, C., Kobayashi, M., Horton, H., Brown, E.L., 1996. Expression monitoring by hybridization to high-density oligonucleotide arrays (see comments). Nat. Biotechnol. 14, 1675. Lorenz, M.G., Cortes, L.M., Lorenz, J.J., Liu, E.T., 2003. Strategy for the design of custom cDNA microarrays. BioTechniques 34, 1264. Lund, R., Aittokallio, T., Nevalainen, O., Lahesmaa, R., 2003a. Identification of novel genes regulated by IL-12, IL-4, or TGF-beta during the early polarization of CD4+ lymphocytes. J. Immunol. 171, 5328.
Lund, R.J., Ylikoski, E.K., Aittokallio, T., Nevalainen, O., Lahesmaa, R., 2003b. Kinetics and STAT4- or STAT6-mediated regulation of genes involved in lymphocyte polarization to Th1 and Th2 cells. Eur. J. Immunol. 33, 1105. Lund, R.J., Chen, Z., Scheinin, J., Lahesmaa, R., 2004. Early target genes of IL-12 and STAT4 signaling in Th cells. J. Immunol. 172, 6775. Mootha, V.K., Lindgren, C.M., Eriksson, K.F., Subramanian, A., Sihag, S., Lehar, J., Puigserver, P., Carlsson, E., Ridderstrale, M., Laurila, E., Houstis, N., Daly, M.J., Patterson, N., Mesirov, J.P., Golub, T.R., Tamayo, P., Spiegelman, B., Lander, E.S., Hirschhorn, J.N., Altshuler, D., Groop, L.C., 2003. PGC1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat. Genet. 34, 267. Mosmann, T.R., Cherwinski, H., Bond, M.W., Giedlin, M.A., Coffman, R.L., 1986. Two types of murine helper T cell clone: I. Definition according to profiles of lymphokine activities and secreted proteins. J. Immunol. 136, 2348. Rogge, L., Bianchi, E., Biffi, M., Bono, E., Chang, S.Y., Alexander, H., Santini, C., Ferrari, G., Sinigaglia, L., Seiler, M., Neeb, M., Mous, J., Sinigaglia, F., Certa, U., 2000. Transcript imaging of the development of human T helper cells using oligonucleotide arrays. Nat. Genet. 25, 96. Romagnani, S., 1996. Th1 and Th2 in human diseases. Clin. Immunol. Immunopathol. 80, 225. Schena, M., Shalon, D., Davis, R.W., Brown, P.O., 1995. Quantitative monitoring of gene expression patterns with a complementary DNA microarray (see comments). Science 270, 467. Seder, R.A., Paul, W.E., 1994. Acquisition of lymphokine-producing phenotype by CD4+ T cells. Annu. Rev. Immunol. 12, 635. Taylor, E., Cogdell, D., Coombes, K., Hu, L., Ramdas, L., Tabor, A., Hamilton, S., Zhang, W., 2001. Sequence verification as quality-control step for production of cDNA microarrays. BioTechniques 31, 62. Whitney, A.R., Diehn, M., Popper, S.J., Alizadeh, A.A., Boldrick, J.C., Relman, D.A., Brown, P.O., 2003. Individuality and variation in gene expression patterns in human blood. Proc. Natl. Acad. Sci. U. S. A. 100, 1896. Yang, L.P., Byun, D.G., Demeure, C.E., Vezzio, N., Delespesse, G., 1995. Default development of cloned human naive CD4 T cells into interleukin-4- and interleukin-5-producing effector cells. Eur. J. Immunol. 25, 3517. Yang, Y.H., Dudoit, S., Luu, P., Lin, D.M., Peng, V., Ngai, J., Speed, T.P., 2002. Normalization for cDNA microarray data: a robust composite method addressing single and multiple slide systematic variation. Nucleic Acids Res. 30, e15. Yuen, T., Wurmbach, E., Pfeffer, R.L., Ebersole, B.J., Sealfon, S.C., 2002. Accuracy and calibration of commercial oligonucleotide and custom cDNA microarrays. Nucleic Acids Res. 30, e48.
