SEARCH

SEARCH BY CITATION

Keywords:

  • microarrays;
  • interferon;
  • TNF-alpha;
  • FLIP;
  • apoptosis

Abstract

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

Summary. Microarrays provide a powerful tool for the study of haemopoietic stem and progenitor cells (HSC). Because of the low frequency of HSC, it is rarely feasible to obtain enough mRNA for microarray hybridizations, and amplification will be necessary. Antisense RNA (aRNA) amplification is reported to give high-fidelity amplification, but most studies have used only qualitative validation. Before applying aRNA amplification to the study of HSC, we wished to determine its fidelity and reproducibility, and whether statistically significant results can be obtained. We found that aRNA amplification introduced biases into relative RNA abundance. However, these biases were extremely consistent, and valid comparisons could be made, if amplified RNA was compared with amplified RNA. By applying this method to the effect of interferon-γ and tumour necrosis factor-α on normal primary CD34+ HSC, biologically significant differences could be detected, including potential mechanisms for resistance of CD34+ cells to CD95-mediated apoptosis and evidence of the differentiating effects of the cytokines. Differences of twofold or less were detected, and most of these differences attained statistical significance after triplicate experiments. These data demonstrate that aRNA amplification can be used with microarray hybridization to study the transcriptional profiles of small numbers of primary CD34+ HSC.

There is great interest in the use of microarrays to study gene expression in human stem and progenitor cells (HSC), both during normal differentiation and in disease states. However, as CD34+ HSC constitute only 1–4% of normal bone marrow mononuclear cells, it is not possible to routinely obtain sufficient cell numbers to extract enough RNA for microarray analysis. The problem is further accentuated when studying rarer, more primitive, subsets of the HSC population. Thus amplification of the starting RNA will be a necessary requirement.

Antisense RNA (aRNA) amplification is a linear rather than an exponential process, and it has been reported to give high-fidelity mRNA amplification (Wang et al, 2000). This technique is now in widespread use in microarray studies. However, most studies have used the technique with no (Miyazato et al, 2001; Fink et al, 2002; Mori et al, 2002) or only qualitative validation (Luo et al, 1999; Baugh et al, 2001; Ernst et al, 2002; Scheidl et al, 2002; Sotiriou et al, 2002), and there have been few quantitative systematic studies of its fidelity and reproducibility (Pabon et al, 2001; Scheidl et al, 2002). Indeed, careful examination of the published data does not fully support the claim of high-fidelity amplification. As microarrays become an established technique, stricter scientific criteria are being applied to experiments, and the need for quantitative statistical evaluation of data is becoming accepted as mandatory. We, therefore, wished to verify the quantitative applicability of aRNA amplification and microarray hybridization to studies of gene expression in HSC.

We show here that aRNA amplification introduced significant biases into the expression profile but, nevertheless, quantitative and statistically significant data could be obtained from experiments with CD34+ HSC.

Cell lines.  Cell lines were maintained in Roswell Park Memorial Institute (RPMI)-1640 medium supplemented with 10% fetal calf serum (FCS), 2 mmol/l glutamine, 100 U/ml penicillin and 100 U/ml streptomycin at 37°C in 5% CO2 and 100% humidity. Cells were counted using a haemocytometer and passaged regularly.

CD34+ cell isolation and treatment with interferon-γ (IFN-γ) and tumour necrosis factor-α (TNF-α). Bone marrow was aspirated from the iliac crest of volunteer donors, following informed consent as approved by the Local Ethical Committee, and collected in 1 ml of 10 IU/ml of preservative-free heparin (Leo Laboratories, Princes Risborough, UK). The aspirate was diluted 1:1 in Iscove's modified Dulbecco's medium [IMDM; Gibco Invitrogen, Paisley, UK (Gibco)] of osmolality 340 mOsm/kg, supplemented with 10% FCS, 10% horse serum, 100 IU/ml penicillin–streptomycin [Sigma Aldrich, Poole, UK (Sigma)] and 10−6 mol/l hydrocortisone succinate (Sigma) (henceforth referred to as IMDM-340 medium). Mononuclear cells were isolated by centrifuging on Ficoll–hypaque at 400 g for 20 min and were then washed twice in plain IMDM. CD34+ cells from this mononuclear cell population were then positively selected with anti-CD34 and magnetically labelled anti-IgG microbeads (MACS; Miltenyi Biotech, Bisley, UK), using the manufacturer's protocol.

The purified CD34+ cells were resuspended in IMDM-340 medium. TNF-α and IFN-γ (R & D Systems, Abingdon, UK) were added to final concentrations of 10 ng/ml. Treated cells are referred to as IFN/TNF treated. The remaining CD34+ cells were incubated with no added cytokines. Following addition of the cytokines, the cells were incubated for 16 h at 37°C and 5% CO2. Following incubation, about 500 000 IFN/TNF-treated and untreated cells were washed, and RNA extracted as described below.

The purity of the cells and efficacy of the IFN/TNF treatment was assessed by flow cytometry. Briefly, cells were washed once in phosphate-buffered saline (PBS) with 5% FCS. Human gamma globulin (20 µl) was added and cells were incubated on ice for 10 min to block non-specific binding. The following antibodies or matched isotype controls were added: CD34-tricolor (581; Caltag, Towcester, UK), CD38-phycoerythrin (CD38-PE; HIT-2; BD Pharmingen, Oxford, UK), CD95-fluorescein isothiocyanate (CD95-FITC; LOB-3/17 Serotec, Kidlington, UK). Cells were incubated for 30 min on ice, washed (in PBS with 5% FCS) and resuspended in 2% formaldehyde, and analysed on a FACScan flow cytometer (Becton-Dickinson, Oxford, UK). Acquisition and analysis was performed using cellquest software.

RNA extraction.  RNA was extracted using an Rneasy minikit (Qiagen, Crawley, UK), following the manufacturer's protocol, and recovered in diethyl pyrocarbonate (DEPC)-treated water. The yield and purity of the RNA was determined by spectrophotometry, and the integrity of the RNA was checked by 0·8% agarose gel electrophoresis and ethidium bromide staining. The RNA was stored at −70°C. mRNA was purified using a Micropoly(A) Pure kit (Ambion, Huntington, UK), dissolved in 20 μl DEPC-treated water and stored at −70°C.

Construction of the array.  The first series of experiments were performed with array SGHMS01, consisting of elements designed to a subset of human genes important in infection and immunity. Further experiments were performed with a second generation array, SGHMS02, which carried additional elements in particular for additional genes relevant to haemopoiesis. A gene list for both arrays is available at http://www.sghms.ac.uk/depts/medmicro/bugs/CoopGrp.htm. The arrays were based on exon-specific polymerase chain reaction (PCR) products designed to all the predicted splice variants for the genes according to the ensembl database release at the time of construction. The design aimed to achieve minimal homology between the elements on the array and all the splice variants of genes represented on the array. This was achieved by designing primer sets to all exons giving an amplicon of greater than 120 base pairs using primer3 software. These amplicons were compared with the original exon sequences using the blastn (nucleotide basic local alignment search tool) program, and primers giving the largest amplicons with single BLAST hits lying closest to the 3′ end of the gene sequence for each splice variant were selected. Where this was not possible owing to the absence of a unique exon in a one-splice variant, multiple primer sets were selected for which a combination of amplicons would provide a unique signature when arrayed as separate elements on the array. Amplicons were amplified from genomic DNA obtained from the CEM human T-lymphoblastic cell line, using 5 U Hot Star Taq (Qiagen), 1 µmol/l each primer, 1 × PCR buffer (Qiagen), and amplified at 95°C for 15 min followed by 40 cycles of 95°C, 55°C, 72°C for 30 s each followed by 5 min extension at 72°C. All liquid handling and PCR was carried out using a RoboAmp2400 (MWG). PCR products were assessed for their quality by agarose gel electrophoresis; products not passing set quality criteria were repeated. PCR products were postprocessed by isopropanol precipitation in the presence of glycogen and resupended in 50% (v/v) dimethylsulphoxide for printing. DNA was printed on poly-l-lysine-coated (Sigma) glass slides using a MicrogridII (Biorobotics, Haslingfield, UK). Arrays were postprint processed essentially as described (Eisen & Brown, 1999). Briefly, the slides were hydrated by water vapour for 5 s, dried on a heating block for a further 5 s and the DNA was then ultra-violet crosslinked to the poly l-lysine (2000 mj) using a Stratolinker (Stratagene, Amsterdam, the Netherlands). Slides were prewashed in 1 × saline sodium citrate (SSC) for 60 s followed by 0·06 × SSC for 60 s; they were then washed in 1·8% (w/v) succinic anhydride in 1-methyl-2-pyrrolidinone (Sigma) activated by the addition of sodium borate to 4 mmol/l for 15 min. Slides were washed in a boiling water bath for 2 min and finally washed in 95% (v/v) ethanol for 1 min. Slides were dried by centrifugation and stored dessicated in the dark until required.

cDNA synthesis and T7 amplification.  Total RNA (1–3 μg) was mixed with 1 μg oligo dT-T7 primer [AAA CGA CGG CCA GTG AAT TGT AAT ACG ACT CAC TAT AGG CGC TTT TTT TTT TTT TTT; (Invitrogen, Paisley, UK) purified by polyacrylamide gel electrophoresis], heated at 70°C for 4 min and chilled on ice. Reverse transcription was performed in a total volume of 20 μl containing 4 μl first-strand buffer (Gibco), 2 μl 0·1 mol/l dithiothreitol (DTT), 2 μl 10 mmol/l dNTPs, 1 μl RnaseOUT (Gibco) and 400 U Superscript II reverse transcriptase (Gibco) at 42°C for 90 min. The mixture was then heated at 70°C for 15 min and cooled immediately on ice. For second-strand synthesis the reaction was made up to 160 μl, containing 32 μl second-strand buffer (Gibco), 3 μl dNTPs (10 mmol/l each), 0·7 μl RNase H (2 U/ml; Gibco) and 4 μl DNA polymerase I (10 U/ml; Gibco). The reaction was incubated for 2 h at 16°C and terminated by adding 12 μl 250 mmol/l EDTA. cDNA was purified by phenol:chloroform extraction and ethanol precipitation. The pellet was resuspended in 8 μl water and the double-stranded cDNA was assessed by ethidium bromide spot assay.

In-vitro transcription was performed using a T7 MEGAscript kit (Ambion) in a final reaction volume of 20 μl. The reaction mixture was incubated at 37°C for 4 h. DNase 1 (1 μl, 2 U/μl) was added to the reaction and re-incubated at 37°C for 15 min. The amplified RNA was precipitated using lithium chloride, washed in 70% ethanol and dissolved in 20 μl DEPC-treated water. The yield of the amplified RNA was assessed by optical density. The product size was determined by formaldehyde gel electrophoresis of denatured RNA in 1% agarose containing 1 × N-morpholinol propanesulphonic acid (MOPS) buffer [0·02 mol/l MOPS, 8 mmol/l sodium acetate and 1 mmol/l EDTA (pH 8·0)] and 18% of 12·3 mol/l formaldehyde and staining with SYBR Gold.

RNA labelling and microarray hybridization.  mRNA or aRNA (2 μg), or 50 μg total RNA, were mixed with 1 μl random primers (3 μg/ml; Gibco) in a final volume of 11 μl, heated at 70°C for 10 min and chilled on ice. The RNA was labelled with Cy3 or Cy5 by reverse transcription in a 25-μl reaction containing 5 μl first-strand buffer, 2·5 μl 0·1 mol/l DTT, 2·3 μl dNTPs (5 mmol/l dATP, 5 mmol/l dGTP, 5 mmol/l dTTP and 2 mmol/l dCTP), 1·7 μl Cy3 or Cy5-labelled dCTP (Amersham) and 2·5 μl Superscript II (200 U/ml; Gibco). The tube was incubated for 10 min at 25°C protected from light, followed by 90 min at 42°C. The products of each tube were mixed and the RNA was hydrolysed by adding 2·5 μl of 2·5 mol/l NaOH and incubation for 15 min at 37°C in the dark. The reaction was neutralized using 25 μl of 2 mol/l HEPES free acid. The labelled cDNA was purified using a MinElute column (Qiagen) and eluted in 13·1 μl water.

The microarray slide was incubated in preheated prehybridization solution [8·75 ml 20 × SSC, 0·5 ml 10% sodium dodecyl sulphate (SDS), 5 ml 0·1% bovine serum albumin and 35·75 ml water] for 20 min at 65°C. The slide was then rinsed thoroughly in 400 ml water for 1 min followed by a rinse in 400 ml isopropanol for 1 min and dried by centrifugation.

The labelled purified cDNA was mixed with hybridization solution (4 μl filtered 20 × SSC and 2·9 μl filtered 2% SDS) and heated for 2 min at 95°C followed by a brief centrifugation. The mixture was hybridized to the slide under a glass cover slip, using a hybridization cassette (Genpak, New Milton, UK), at 65°C overnight.

The hybridized slide was washed for 2 min in wash A (20 ml 20 × SSC, 2 ml 10% SDS and 378 ml water) at 65°C and for 2 min in wash B (1·2 ml 20 × SSC and 398·8 ml water) at room temperature with continuous agitation. The slide was washed once more in fresh wash B and dried by centrifugation.

Slide scanning and data analysis.  The dye bound to each array element was visualized using a dual confocal laser scanner (Affymetrix 418). The image was quantified using imagene 4·2 software (Biodiscovery, Marina del Ray, CA, USA), and the data transferred to genespring 4·2 (Silicon Genetics, Redwood City, CA, USA) for normalization and interpretation. Fluorescence was normalized against all genes, usually at the 70th percentile, and the data displayed as a scatter plot. A threshold was set in each channel at the mean +3 standard deviations of the negative spots. This corresponded well with the start of ‘fishtailing’ of the distribution. Replicate data were combined within genespring 4·2. Statistical significance was evaluated by the Students t-test, assuming unequal variances, and combined with expression data via a Venn diagram.

Results

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

In initial experiments, the CD34+ AML cell line KG1 was used as a model system. PolyA+ RNA was reverse transcribed using a combined T7 promoter/oligo dT primer and converted to double-stranded cDNA by RNAse H/DNA polymerase I treatment. This cDNA was transcribed for 2, 4 or 8 h by T7 RNA polymerase using the Ambion Megascribe in-vitro transcription kit. The resulting aRNA was evaluated by formaldehyde gel electrophoresis (Fig 1). At 2 and 4 h, transcripts ranged from 500 to 4000 nt with a mode around 2000 nt. Yield increased from 2 to 4 h. However, by 8 h, both the yield and size of the aRNA was decreased, indicating RNA degradation. Using smaller amounts of cDNA still resulted in the optimum yield at 4 h of incubation. Therefore, all subsequent in-vitro transcription reactions were standardized at 4 h. Starting from 300 ng of mRNA, the total yield of aRNA was typically 30 µg. This represented a 100-fold amplification by weight or, allowing for the size of the RNA, about a 300-fold molar amplification.

image

Figure 1. Denaturing gel electrophoresis of aRNA. aRNA was synthesized by in-vitro transcription for 2, 4 or 8 h and analysed by formaldehyde gel electrophoresis. The 28S and 18S ribosomal bands of total RNA are shown as a size marker.

Download figure to PowerPoint

In order to evaluate the fidelity of amplification, the aRNA was compared with the starting mRNA by microarray hybridization. To determine the intrinsic variability in labelling, hybridization and scanning, aRNA labelled with Cy5 was competitively hybridized to the SGHMS01 array against the same aRNA labelled with Cy3, and results were analysed using imagene 4·1 and genespring 4·1 or 4·2. Figure 2A shows a scatter plot of normalized Cy5 versus Cy3 intensity for each element on the array. The results of the self:self comparison showed extremely tight concordance (r = 0·95) with one outlier due to consistent differential labelling. aRNA labelled with Cy5 was then compared with mRNA labelled with Cy3 (Fig 2B). It was clear that aRNA amplification had introduced changes of threefold and greater in the relative abundance of a proportion of transcripts, with a resulting correlation of only 0·5. Thus comparison of amplified with unamplified RNA would only allow confident detection of differences in gene expression of the order of 5–10-fold or greater.

image

Figure 2. Scatter plots of microarray hybridizations of amplified aRNA to array SGHMS01. In each panel, the central line represents the line of equal normalized expression in the Cy5 and Cy3 channels. The upper and lower lines represent a twofold increase and twofold decrease, respectively, in the Cy5 versus the Cy3 channel. (A) Self versus self: KG1 aRNA (Cy5 labelled) against the same RNA (Cy3 labelled); (B) aRNA versus mRNA: KG1 aRNA (Cy5) against the starting mRNA (Cy3); (C) independent amplifications: KG1 aRNA (Cy5) against an independent amplification from the same mRNA (Cy3); (D) independent amplifications: K562 aRNA (Cy5) against an independent amplification from the same mRNA (Cy3); (E) aRNA versus mRNA: K562 aRNA (Cy5) against the starting mRNA (Cy3).

Download figure to PowerPoint

The changes in RNA abundances introduced by aRNA amplification could represent random variation or biases in the amplification of specific genes. To test these possibilities, two independent amplifications from the same starting mRNA were compared with each other by microarray hybridization (Fig 2C). The results showed strong concordance (r = 0·99), with no obviously increased variability compared with a self:self comparison. It was clear by visual inspection that differences in gene expression of only twofold could be confidently detected against this background. To confirm these results, experiments were repeated with mRNA from the K562 cell line (Fig 2D and E), which showed essentially identical results to those obtained with KG1 cells.

For studying small numbers of cells, there are advantages in not having to purify poly A+ RNA. We, therefore, investigated whether reliable aRNA amplification could be obtained starting from total RNA. KG1 cell total RNA was reverse transcribed using the oligo dT-T7 promoter primer, and the double-stranded DNA was used for in-vitro transcription as above. The size distribution of the aRNA was indistinguishable from that obtained from polyA+ RNA. Three micrograms of total RNA yielded 30 µg aRNA. This was higher than the yield obtained with purified polyA+ RNA: assuming that polyA+ RNA constitutes around 3% of the total RNA, amplification was 300-fold by weight or about 1000-fold molar amplification.

The fidelity of amplification was again investigated by microarray hybridization. As with mRNA, there was a poor correlation of aRNA with the starting RNA (r = 0·64; Fig 3A), as against the intrinsic variability when Cy5-labelled aRNA was compared with the same RNA labelled with Cy3 (r = 0·97; Fig 3B). When two independently amplified aRNAs from the same starting total RNA were compared, variability was equivalent to that with one aRNA compared with itself (r = 0·99; Fig 3C).

image

Figure 3. Scatter plots of microarray hybridizations of amplified aRNA to array SGHMS02. In each panel, the central line represents the line of equal normalized expression in the Cy5 and Cy3 channels. The upper and lower lines represent a twofold increase and twofold decrease, respectively, in the Cy5 versus the Cy3 channel. (A) aRNA versus total RNA: KG1 aRNA (Cy5) against the starting total RNA (Cy3); (B) self versus self: KG1 aRNA (Cy5 labelled) against the same RNA (Cy3 labelled); (C) independent amplifications: KG1 aRNA (Cy5) against an independent amplification from the same total RNA (Cy3); (D) independent amplifications from primary CD34 cell RNA: CD34 cell aRNA (Cy5) against an independent amplification from the same total RNA (Cy3).

Download figure to PowerPoint

Having shown the consistency of aRNA amplification in a model system, we then proceeded to perform the same experiment on primary CD34+ cell RNA. CD34+ cells were purified from bone marrow mononuclear cells of a normal donor by immunomagnetic bead separation. Total RNA was extracted from 2·5 × 105 CD34-positive cells, reverse transcribed and aRNA was generated by in-vitro transcription. aRNA (6–13 μg) was obtained in duplicate amplifications. Again, comparison of two independent amplifications showed an excellent correlation (r = 0·99; Fig 3D).

The results above indicate in principle that valid results could be obtained from primary CD34+ cells using these methods. We, therefore, proceeded to confirm this in practice. Many early microarray papers published qualitative data without statistical validation by replicate experiments (in part because of cost considerations), but this is no longer considered ideal practice. We, therefore, examined whether independent experiments using amplified RNA could generate statistically significant results. As a model system in which to test this, we studied the effects of a combination of INF-γ and TNF-α.

In three separate experiments, CD34+ cells from different donors were incubated for 16 h with or without a combination of INF-γ and TNF-α, and the gene expression profiles of treated and untreated cells were compared using the SGHMS02 array. The data from the three experiments were then combined and analysed using genespring software.

The averaged data are shown in Fig 4. Eight genes showed > 2-fold upregulation and three genes showed> 2-fold downregulation. All of these were significant at P < 0·01. Our previous experiments would suggest that even genes with a less than twofold change might be analysable. Thus, of 14 genes showing > 1·7-fold upregulation, 13 were significant and, of nine genes showing > 1·7-fold downregulation, eight were significant (P < 0·05 or better). Details are given in Table I. In one experiment, dye labelling was reversed (dye flip), and the same up and downregulated genes were observed.

image

Figure 4. Scatter plots of microarray hybridization to array SGHMS02 of amplified aRNA from IFN/TNF-treated CD34 cells against untreated cells. Two upregulated genes (MIG, VCAM) are not visible on this plot, because their uninduced levels were below the threshold. One upregulated gene (TNFRSF6) is represented twice, for two splice variants. The central line represents the line of equal normalized expression in the Cy5 and Cy3 channels. The upper and lower lines represent a twofold increase and twofold decrease, respectively, in the Cy5 versus the Cy3 channel.

Download figure to PowerPoint

Table I.  Genes showing statistically significant upregulation or downregulation in the presence of IFNγ/TNFα.
Gene nameDescription 
≥ 2-fold upregulation
  MIGIFNγ-inducible monokineP < 0·01
  VCAMVascular cell adhesion moleculeP < 0·01
  STAT1Signal transducer and activator of transcription 1P < 0·01
  IRF1Interferon responsive factor 1P < 0·01
  CASPER1Caspase 8 inhibitor; FLIPP < 0·01
  CD38Haemopoietic progenitor cell antigenP < 0·01
  TNFRSF6TNF receptor superfamily member 6; Fas; CD95P < 0·01
  MX1Interferon-regulated GTP binding proteinP < 0·01
≥ 1·7-fold upregulation
  HLACHuman lymphocyte antigenP < 0·05
  SOD2Superoxide dismutaseP < 0·01
  PIG7Lipopolysaccharide-inducible TNFα controlling gene; LITAFP < 0·05
  CASP1Caspase 1P < 0·01
  TNFRSF5TNF receptor superfamily member 5P < 0·05
  JAK2Janus kinase 2P < 0·01
≥ 1·7-fold downregulation P < 0·05
  VEGFVascular endothelial growth factorP < 0·05
  MAPK12Mitogen-activated protein kinase p38 gamma; ERK6; SAPK3P < 0·05
  CD34Haemopoietic stem and progenitor cell antigenP < 0·05
  HSPA6Heat shock 70-kDa protein 6P < 0·01
  IL1R2Interleukin 1 receptor, type 2P < 0·05
≥ 2-fold downregulation
  CD14Monocyte differentiation antigen; LPS receptorP < 0·01
  FOSP55-c-FOS proto-oncogeneP < 0·01
  BADPro-apoptotic BCL-2 binding proteinP < 0·01

Discussion

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

Commitment and differentiation of haematopoietic stem cells involves changes in gene transcription, regulated by specific and non-specific transcription factors. DNA microarrays provide a powerful tool for investigating patterns of gene expression, but they require significant amounts of starting RNA.

The antisense RNA amplification method of Kacharmina et al (1999) has long been a powerful tool to study gene expression when only small numbers of purified cells are available. A recent paper showing the application of aRNA amplification to microarray studies reported the high fidelity of this technique (Wang et al, 2000). On the basis of this, many groups have applied the aRNA method to microarray studies without further validation (Miyazato et al, 2001; Fink et al, 2002; Mori et al, 2002), while other groups have undertaken qualitative validation (Luo et al, 1999; Baugh et al, 2001; Ernst et al, 2002; Scheidl et al, 2002; Sotiriou et al, 2002) to show that the lists of overexpressed and underexpressed genes obtained using amplified material are equivalent to those using primary RNA. Only a few groups have directly addressed the quantitative validation of aRNA amplification (Pabon et al, 2001; Scheidl et al, 2002).

As microarray methods change from a novel to an established technique, stricter scientific criteria are being applied to experiments, and the need for quantitative statistical evaluation of data is becoming accepted as mandatory. Thus quantitative validation of amplification is essential.

Contrary to the report of Wang et al (2000), we found that aRNA amplification is of only limited fidelity and the relative abundance of individual genes may change by up to fivefold during the process. Our protocol differs from that previously reported in using RNA-primed second-strand cDNA synthesis rather than a template switch primer. However, other studies using template switching have indicated a lack of fidelity (Baugh et al, 2001; Pabon et al, 2001; Scheidl et al, 2002), and the data of Wang et al (2000) do not fully support the claim of high fidelity. Furthermore, a template switch primer would be expected to introduce biases of its own: the method was developed to selectively clone only those messages which achieve full-length reverse transcription in first-strand cDNA synthesis.

We have not investigated the origin of the individual biases, but there are many obvious possibilities, and different biases may be important for different sequences. Biases could arise during the original cDNA synthesis. As all sequences are transcribed from the same (T7) promoter, differences in the rate of initiation are unlikely, but termination of transcription is likely to be sequence specific, affecting the length of individual aRNA transcripts and thus their relative signal strength after labelling. Also, as the efficiency of Cy dye labelling can be sequence specific, biases may be introduced for some transcripts when the antisense sequence labels with different efficiency to the sense sequence.

Our results showed that the biases in amplification are extremely consistent. Thus, in principle, amplified RNA can be used to obtain quantitatively, not just qualitatively, consistent results, as long as amplified RNA is compared with amplified RNA. We then showed that this technique could be directly applied to studying the transcriptional response of primary CD34+ HSC and that statistically significant results can be obtained.

Inflammatory cytokines are known to have an inhibitory effect on colony formation in culture, and are hypothesized to have a role in the pathophysiology of aplastic anaemia in vivo (Young, 2000). We have previously studied the effect of IFN-γ/TNF-α on purified CD34+ cells in vitro and shown a decrease in primitive CD34+CD38 cells, probably due to differentiation, and a large increase in expression of CD95 (unpublished observations). However, the CD34+ cells were resistant to CD95-mediated apoptosis, as previously reported (Barcena et al, 1999; Josefsen et al, 1999).

Microarray studies could, in principle, help elucidate both the differentiation process and the resistance to apoptosis. Although the current studies have only used a limited gene set and at a single timepoint, some interesting results have already begun to emerge. Many of the genes upregulated in these cells are already known to be IFN inducible in other cell types, for instance MIG, IRF1 and MX1. CD95/Fas is also seen to be upregulated. Even with surface Fas expression, CD34+ cells appear to be resistant to the induction of apoptosis by agonistic antibodies or soluble Fas-ligand. While these may be relatively weak inducers, we have recently extended this work to show that the cells are also resistant to membrane-bound Fas-ligand (unpublished observations). It has recently been suggested that this resistance to Fas-mediated induction of apoptosis is due to expression of the caspase 8 inhibitor FLIP/CASPER {FLICE [FADD (Fas-associated death domain protein)-like interleukin 1 beta-converting enzyme]-inhibitory protein/Caspase-8-related protein} (Kim et al, 2002). Thus it is interesting that upregulation of CASPER mRNA was seen in cells treated with IFN/TNF. Conversely, the pro-apoptotic BCL-2 family protein, BAD (bcl-xL/bcl-2 associated death promoter), is downregulated, and this may represent another mechanism of resistance to apoptosis. Another interesting observation was the downregulation of CD34 and upregulation of CD38, which correlated with flow cytometry data under the same conditions. This is more likely to be an effect on cell differentiation than a direct transcriptional effect of IFN/TNF on the CD34 and CD38 genes, and more detailed studies are in progress to confirm this.

Thus, despite the rarity of the CD34+ HSC population and the biases inherent in aRNA amplification, our results show that the transcriptional behaviour of normal CD34+ HSC can be studied using microarrays and yield statistically significant results.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

This work was supported by a grant from the UK Leukaemia Research Fund. M.A. is in receipt of a fellowship from the Arab Republic of Egypt. K.L. is supported by the Medical Research Council.

References

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References
  • Barcena, A., Muench, M.O., Song, K.S., Ohkubo, T. & Harrison, M.R. (1999) Role of CD95/Fas and its ligand in the regulation of the growth of human CD34(++) CD38(-) fetal liver cells. Experimental Hematology, 27, 14281439.
  • Baugh, L.R., Hill, A.A., Brown, E.L. & Hunter, C.P. (2001) Quantitative analysis of mRNA amplification by in vitro transcription. Nucleic Acids Research, 29, e29.
  • Eisen, M.B. & Brown, P.O. (1999) DNA arrays for analysis of gene expression. Methods in Enzymology, 303, 179205.
  • Ernst, T., Hergenhahn, M., Kenzelmann, M., Cohen, C.D., Bonrouhi, M., Weninger, A., Klaren, R., Grone, E.F., Wiesel, M., Gudemann, C., Kuster, J., Schott, W., Staehler, G., Kretzler, M., Hollstein, M. & Grone, H.J. (2002) Decrease and gain of gene expression are equally discriminatory markers for prostate carcinoma: a gene expression analysis on total and microdissected prostate tissue. American Journal of Pathology, 160, 21692180.
  • Fink, L., Kohlhoff, S., Stein, M.M., Hanze, J., Weissmann, N., Rose, F., Akkayagil, E., Manz, D., Grimminger, F., Seeger, W. & Bohle, R.M. (2002) cDNA array hybridization after laser-assisted microdissection from nonneoplastic tissue. American Journal of Pathology, 160, 8190.
  • Josefsen, D., Myklebust, J.H., Lynch, D.H., Stokke, T., Blomhoff, H.K. & Smeland, E.B. (1999) Fas ligand promotes cell survival of immature human bone marrow CD34+CD38– hematopoietic progenitor cells by suppressing apoptosis. Experimental Hematology, 27, 14511459.
  • Kacharmina, J.E., Crino, P.B. & Eberwine, J. (1999) Preparation of cDNA from single cells and subcellular regions. Methods in Enzymology, 303, 318.
  • Kim, H., Whartenby, K.A., Georgantas, III, R.W., Wingard., J. & Civin, C.I. (2002) Human CD34+ hematopoietic stem/progenitor cells express high levels of FLIP and are resistant to Fas-mediated apoptosis. Stem Cells, 20, 174182.
  • Luo, L., Salunga, R.C., Guo, H., Bittner, A., Joy, K.C., Galindo, J.E., Xiao, H., Rogers, K.E., Wan, J.S., Jackson, M.R. & Erlander, M.G. (1999) Gene expression profiles of laser-captured adjacent neuronal subtypes. Nature Medicine, 5, 117122.
  • Miyazato, A., Ueno, S., Ohmine, K., Ueda, M., Yoshida, K., Yamashita, Y., Kaneko, T., Mori, M., Kirito, K., Toshima, M., Nakamura, Y., Saito, K., Kano, Y., Furusawa, S., Ozawa, K. & Mano, H. (2001) Identification of myelodysplastic syndrome-specific genes by DNA microarray analysis with purified hematopoietic stem cell fraction. Blood, 98, 422427.
  • Mori, M., Mimori, K., Yoshikawa, Y., Shibuta, K., Utsunomiya, T., Sadanaga, N., Tanaka, F., Matsuyama, A., Inoue, H. & Sugimachi, K. (2002) Analysis of the gene-expression profile regarding the progression of human gastric carcinoma. Surgery, 131, s39s47.
  • Pabon, C., Modrusan, Z., Ruvolo, M.V., Coleman, I.M., Daniel, S., Yue, H. & Arnold, Jr, L.J. (2001) Optimized T7 amplification system for microarray analysis. Biotechniques, 31, 874879.
  • Scheidl, S.J., Nilsson, S., Kalen, M., Hellstrom, M., Takemoto, M., Hakansson, J. & Lindahl, P. (2002) mRNA expression profiling of laser microbeam microdissected cells from slender embryonic structures. American Journal of Pathology, 160, 801813.
  • Sotiriou, C., Khanna, C., Jazaeri, A.A., Petersen, D. & Liu, E.T. (2002) Core biopsies can be used to distinguish differences in expression profiling by cDNA microarrays. Journal of Molecular Diagnostics, 4, 3036.
  • Wang, E., Miller, L.D., Ohnmacht, G.A., Liu, E.T. & Marincola, F.M. (2000) High-fidelity mRNA amplification for gene profiling. Nature Biotechnology, 18, 457459.
  • Young, N.S. (2000) Hematopoietic cell destruction by immune mechanisms in acquired aplastic anemia. Seminars in Hematology, 37, 314.