BADGE, BeadsArray for the Detection of Gene Expression, a High-Throughput Diagnostic Bioassay

Evaluation of Specificity of Hybridization between Targets and Capture Probes Coupled to Microspheres

Oligonucleotides with unique sequences of 25 bases were selected from each of 20 Arabidopsis genes and used as capture probes in the first experiment (Table 1). Because NTSR is a human gene that is absent from the Arabidopsis genome, it was chosen as a negative control. Among the 20 genes, most are disease-related genes, whereas UBQ5 and ACT2 are constitutively expressed genes that serve as internal controls (Table 1).

The capture probes derived from the 20 Arabidopsis genes and the human gene NTSR were coupled to different color-coded microspheres in separate reaction tubes and then mixed together for multiplexed assays. Targets (nucleotides to be analyzed) were oligonucleotides complementary to capture probes and tagged with biotin on the 5′ end. The targets were mixed in four tubes in various amounts (5, 10, 20, or 40 fmoles) as indicated in Table 2. In the hexaplexed assay (Fig. 1A), only the first six targets in Table 2 (PR1, PDF, PAD3, UBQ5, ACT2, and NTSR) and their corresponding capture probes coupled to microspheres were added to the reaction. Phycoerythrin (PE)-conjugated streptavidin was added to the reaction to detect bound targets that were biotinylated. The signal of each target hybridized to its specific capture probe coupled to microspheres was determined by the fluorescence intensity of phycoerythrin. A reaction tube with all the components except the targets was used as a negative control. At least 200 microspheres of each set were analyzed by the Luminex¹⁰⁰ system. The Mean Fluorescence Intensity (MFI) of the counted microspheres, which corresponds to the number of targets bound, was calculated and displayed by the Luminex software. The icosaaplexed assay (Fig. 1B) was conducted the same way as the hexaplexed assay except that all 20 targets and 20 sets of microspheres were mixed in the reaction, as indicated in Table 2. These two experiments were designed to address two questions: Are the hybridization signals linearly related to the added targets in the multiplexed assay? Does hybridization between capture probes and targets remain specific as multiplicity increases? As shown in Figure 1A, the hybridization signals from the six targets varied with the added target amounts in a linear fashion, and the signal from the negative control NTSR was not detectable above background. In the experiment shown in Figure 1B, 20 targets were added to the assay. Again, linear signals were obtained, and the signal detected from the negative control NTSR was at the background level. The results from the two experiments showed that the specificity of hybridization did not change while increasing multiplicity from 6 to 20 in the assays. The hybridization between each capture probe and target was specific and linearly proportional. However, when the targets of some genes (e.g., PR5 and AIG2) were present at high concentrations, the signals reached a plateau, as shown in Figure 1B. It is possible that this problem could be avoided by coupling capture probes to microspheres with better efficiencies.

Spiking Experiments in a Whole Population of Labeled Arabidopsis cRNA

Next, we tested the specificity of hybridization in a whole population of biotin-labeled Arabidopsis cRNA. In this experiment, we also determined the range of target amount over which hybridization signals were linearly correlated so that the levels of target genes could be accurately quantified in a population of cRNA. To this end, M13 target oligonucleotides were added to a population of biotin-labeled Arabidopsis cRNA (Fig. 2). The M13 capture probe is a 25-mer oligonucleotide derived from the M13 phage that is absent in the Arabidopsis genome, and the M13 target is another oligonucleotide complementary to the M13 capture probe and tagged with biotin at the 5′ end. The M13 target oligonucleotide was spiked into 10, 20, or 30 μg of cRNA at 0.1, 1.0, 10.0, and 100.0 fmoles, respectively. The tubes from these spiked reactions plus the control tubes (without cRNA) were hybridized at 48°C for 1 h. The signal of the M13 target was detected by measuring the bound streptavidin-phycoerythrin on the microspheres coupled with the M13 capture probe. Assuming that the average size of each gene in a population of Arabidopsis cRNA is 1000 bases and the average molecular weight of each base is 300, then 0.1 fmole would correspond to 30 pg of a gene. Therefore, spiking 0.1, 1.0, 10.0, and 100.0 fmoles of the known target into 10 μg of cRNA is equivalent to the molecular ratios of 1:300,000, 1:30,000, 1:3,000, and 1:300, respectively. As shown in Figure 2, the hybridization signals of M13 target in the absence of cRNA (control) and in the presence of 10, 20, or 30 μg of cRNA were almost identical. This result showed that specific hybridization between the M13 capture probe and its target was not affected by the presence of a large population of biotin-labeled Arabidopsis cRNA, indicating the feasibility of using the assay for analyzing specific hybridization reactions in biological samples. Furthermore, as indicated in Figure 2, the hybridization signal was proportional to the target amount between the frequencies (ratio of a specific target gene to a population of genes) of 1:30,000 and 1:300 (with addition of 1–100 fmoles of target). The minimum amount of detectable target in the linear range is approximately 1.0 fmole for this BeadsArray system under this condition, which corresponds to the level of moderately abundant transcripts in a total population of 10 μg of RNA. We define a moderately abundant transcript as one with 10–30 copies per cell. This experiment showed that the BADGE assay seems capable of measuring transcripts with moderate to abundant expression levels in cells.

The Expression Levels of Pathogenesis-Related Genes in Arabidopsis as a Function of Time Following BTH Treatment Were Profiled by the BADGE Assay

In response to pathogen attack or chemical exposure, a series of pathogenesis-related (PR) genes is activated in many plant species. The expression pattern of these PR genes is important for understanding the defense mechanism in plants (Glazebrook et al. 1997; Glazebrook 1999). Transgenic plants that overexpress various PR genes show enhanced disease resistance against certain pathogens (Broglie et al. 1989; Alexander et al. 1993; Liu et al. 1995). Therefore, a method for monitoring expression levels of a group of PR genes simultaneously is helpful for elucidation of signal transduction networks in plants and useful for screening induced and repressed PR marker genes in mutant plants. Benzothiadiazole (BTH) is a well-studied chemical activator of PR genes in many plants including Arabidopsis, tobacco, and other important agricultural plants (Gorlach et al. 1996; Lawton et al. 1996). We studied the induction of a group of PR genes in Arabidopsis as a function of time following BTH treatment.

In our experiments, total RNA was isolated from Arabidopsis leaves that were harvested at various times (1–9 d) after the application of BTH or water (control). The total RNA was transcribed into cDNA by reverse transcription using oligo(dT). Double-stranded cDNA was synthesized and biotin-labeled cRNA was subsequently obtained through an in vitro transcription reaction using T7 RNA polymerase and biotinylated UTP. The labeled cRNA was hybridized to premixed sets of 20 microspheres at 48°C for 1 h. Then, the expression levels of the 20 genes were measured simultaneously in the multiplexed assay through the detection of streptavidin-conjugated phycoerythrin (Fig. 3). The results obtained from BTH-treated samples are shown in the left panels of Figure 3 (Fig. 3A,C,E,G) and the corresponding results obtained from water-treated samples (controls) are shown in the right panels (Fig. 3B,D,F,H). The 20 genes measured in these experiments were organized into four groups according to their properties and expression patterns.

The genes in the first group (Fig. 3A,B) including PR1, PR5, and PRX1 were induced by BTH. The expression levels of these genes reached their peaks on day 3 following BTH treatment and remained elevated until day 5 (Fig. 3A), after which the signals declined. In contrast, the expression levels of these genes in all the control samples (water-treated Arabidopsis) remained low and unchanged (Fig. 3B). Compared with the controls, the message levels of PRX1, PR1, and PR5 were increased ∼3-fold to 10-fold in these samples by 3 d after treatment with BTH.

The genes in the second group (Fig. 3C,D) did not change their expression levels significantly in either BTH- (Fig. 3C) or water-treated (Fig. 3D) samples. Throughout the whole experiment, very stable expression levels were observed for this group of genes, and no elevated signals were detected in the BTH-treated samples. This indicated that the regulation of the pathogenesis-related genes in the second group is different from that of the three genes in the first group.

In contrast with the genes in the first group, the expression levels of the genes in the third group (Fig. 3E,F) were repressed by BTH. As shown in the figure, the expression levels of PAL, ELI3, and AOS decreased ∼2-fold to 4-fold in the BTH-treated samples compared with the controls on day 3. After day 5, the expression levels of the three genes began to recover.

The genes in the fourth group (Fig. 3G,H) were constitutively expressed and served as internal controls. As expected, their expression levels remained unchanged in both BTH- (Fig. 3G) and water-treated (Fig. 3H) samples. The signal from M13 stayed at the background level. Measured by another established technology, TaqMan, the expression levels of TRX3 and APX3 in the induced and noninduced Arabidopsis samples were generally unchanged.

The Validation of the BADGE Assay

To evaluate the validity and feasibility of an assay, it is necessary to examine reproducibility and linearity of the assay, and to compare data with that obtained by other established technologies. To address reproducibility, all of the experiments described in the previous sections were repeated three to five times. Based on the five separate experiments described in Figure 3, the maximum standard deviation of expression signals in these assays was 35% of the mean. The standard deviation for most of the genes ranged from 5% to 20% of the mean. For the 20 genes studied in Figure 3, only three genes showed standard deviations of 30%–35%. Therefore, we concluded that for the genes analyzed in these experiments, the hybridization signals differed by no more than 35%. Different labeling efficiency, random fragmentation, and other factors could contribute to the variation. We considered this level of variation acceptable because the reproducibility of many microarray systems is such that only differences greater than twofold are significant.

To determine if the signals are linearly related to the amounts of target cRNA, we varied cRNA amounts (1.25 μg, 2.5 μg, 5.0 μg, and 10 μg) in separate assays using the Arabidopsis samples from 3 d of treatment with BTH or water. Again, we simultaneously measured the expression levels of 20 genes as discussed in the experiment presented in Figure 3. To make the graph clear, here we only present the results of the genes from the first group (induced by BTH) and the internal control UBQ4 in these assays. As shown in Figure 4A, linear signals were detected from 2.5 μg to 10 μg of biotin-labeled cRNA that was prepared from samples treated by BTH. The signals from 1.25-μg samples were low for the genes PRX1 and UBQ4 (∼2× higher than background), indicative of the sensitivity limit of the assay. If 2 μg of cRNA is the minimal amount to get reasonable signals, it means that for genes expressed with moderate abundance (e.g., UBQ4), their expression levels should be easily detectable from samples prepared from 1–2 μg of total RNA using this BADGE assay. Again, if one assumes that the average size of all mRNAs is 1000 bases in the Arabidopsis genome and the average molecular weight of the four bases is 300, a gene with moderate abundance in 2 μg of a total population of RNA is at the level ∼0.7 fmole. For a second time, this experiment showed that the minimal amount of detectable target gene for this BADGE assay was ∼1 fmole, which is consistent with the result obtained from the spiking experiment shown in Figure 2. The results shown in Figure 4B are the expression levels of PR1, PR5, PRX1, and UBQ4 from control Arabidopsis samples that were treated with water. The signals for PR1, PR5, and PRX1 remained at low levels as anticipated (Fig. 4B). In contrast, the expression level of the internal control gene UBQ4 was always linearly correlated to the added target amount (Fig. 4A,B).

GeneChip technology from Affymetrix is widely accepted as a reliable method for monitoring expression levels of several thousand of genes simultaneously. A GeneChip specifically made for Arabidopsis was used to measure the expression of genes in Arabidopsis leaves after treatment with BTH for 3 d. Aliquots of the same samples were also analyzed by the BADGE assay using the Luminex¹⁰⁰ system. The results from the BADGE assay were compared with that represented on the Arabidopsis GeneChip (Fig. 5). GST1 and PRX1 were not included in this figure because the two genes were not found in the pre-made Arabidopsis GeneChip.

We normalized the hybridization signals of all the genes detected by GeneChip and BADGE against the expression level of UBQ4. The relative hybridization signals were plotted versus their corresponding genes (Fig. 5A). As shown in Figure 5A, the trends of the curves obtained by the two technologies were very similar. Although the actual values of the relative hybridization signals to UBQ4 were not exactly the same for the two technologies, signals detected from genes with high expression levels by GeneChip were always high as detected by BADGE, and signals detected from genes with low expression levels appeared consistently low. The data obtained from GeneChip technology strongly supported the data from the BADGE assay.

Our experiments showed that different capture probes chosen from the same gene gave similar results. To check if a single capture probe is sufficient, we also designed multiple capture probes from each gene and then we mixed either three or six capture probes together in equal molar ratios and coupled them together to the same set of microspheres. We hybridized cRNA from an Arabidopsis sample to the three-probe (3P) and six-probe (6P) microspheres, respectively. The data were normalized against the signal of UBQ4 as described above and also plotted together in Figure 5A. The relative expression signals for all the genes were very similar. According to the information provided by Luminex, the diameter of a single microsphere is 5.6 μm and can be coupled to 100,000–500,000 molecules of capture probes on the surface. Multiple capture probes were supposed to occupy the surface of a microsphere randomly and evenly when they were mixed in equal molar ratio in the coupling reactions. The use of multiple capture probes coupled to microspheres could avoid the problem of possible inefficient binding that was associated with a single capture probe.

The hybridization signals obtained from experiments using one-, three-, and six-probe microspheres were all different. Theoretically, for the BADGE assay, the probability of hybridizing to either single- or multiple-probe microspheres for target genes (cRNA) should be equal because the densities of capture probes per microsphere were the same (100,000–500,000 molecules per microsphere regardless of whether they were mono or multiple probes) if we assumed the association and dissociation constants of each capture probe to its target sequence were the same. However, many factors could result in different hybridization signals in practice. For instance, the random fragmentation of target genes in cRNA samples, the different hybridization efficiencies of various capture probes, and the different coupling efficiencies of multiple capture probes to microspheres. In spite of these factors, the expression profiles of the genes studied in these experiments (using 1P, 3P, and 6P microspheres) agreed well between experiments and were consistent with the results obtained from the GeneChip experiment. Our experiments indicated that single and multiple capture probes worked equally well. Although none of our designed capture probes failed in the BADGE assay, to avoid the potential problem of possible inefficient binding arising from a single capture probe, we recommend using multiple capture probes, say, three, for the experiments.

We next compared the fold changes in hybridization signals of BTH-treated samples to the control samples (Fig. 5B). The fold change in hybridization signal for each gene was calculated as the signal from the BTH-treated sample over the signal from the water-treated sample. Consistently detected by BeadsArray (1P, 3P, and 6P microspheres) and GeneChip, the expression levels of PR5 and PR1 were significantly induced by BTH at 3 d after treatment (Fig. 5B), but the expression levels of other genes were not changed dramatically in the two samples.

Capture Probe and Its Coupling to Microspheres

A unique sequence of 25 bases within a region of 600 bases from the 3′ end of a target gene was chosen as a capture probe. Multiple capture probes could be selected from the same gene at different positions. The T_m of the chosen capture probe usually ranged from 50°C to 65°C and minimal secondary structure was preferred. Capture probes were chosen using commercial software (Vector NTI). All capture-probe oligonucleotides were synthesized with 5′-amino uni-linker (Oligos Etc.) and then covalently linked to carboxylated fluorochrome microspheres (Luminex). Specifically, 5 × 10⁶ carboxylated microspheres were pelleted in a microcentrifuge for 1 min at maximum speed, and the supernatant was carefully removed using a pipette without disturbing the microspheres. The microspheres were resuspended in 50 μL of a buffer containing 0.1 M MES (Sigma) at pH 4.5. The amino-substituted capture probe was dissolved in ddH₂O at a concentration of 1 mM, and 1 μL of the solution (containing 1 nmole of capture-probe oligonucleotides) was added to the microspheres for the coupling reaction. The coupling reaction was initiated by adding 2.5 μL of freshly made 10 mg/mL 1-ethyl-3-(3-dimethyl-aminopropyl) carbodiimide hydrochloride (EDC; Aldrich). The mixture of microspheres, capture probes, and EDC was vortexed briefly and incubated at room temperature for 30 min in the dark. After the 30-min incubation, a second 2.5 μL of freshly prepared EDC solution (10 mg/mL) was added to the reaction and incubated for an additional 30 min. This step was repeated for a total of three times (three EDC additions). During the incubations, the reaction was occasionally mixed by finger flicking the tube to keep the microspheres in suspension. After the coupling reaction, 1 mL of 0.02% Tween 20 (BioRad) was added to the microspheres. The solution was mixed well and centrifuged in a microcentrifuge for 1 min at maximum speed. The supernatant containing free capture-probe oligonucleotides and excess EDC was carefully removed. The microspheres were washed again in 1 mL of 0.1% SDS (Ambion) to ensure the removal of free capture probe and EDC. In the end, the capture-probe-conjugated microspheres were resuspended in 100 μL of a buffer containing 0.1 M MES at pH 4.5. The coupled microspheres were stored at 4°C in a dark box and stayed stable for at least 6 mo. The microspheres were diluted in TE buffer and counted using a cell counter slide under 100× magnification. For a single hybridization assay, about 7500 coupled microspheres of each set were used. The coupling efficiencies and hybridization specificity were evaluated by hybridizing the coupled microspheres to their corresponding biotinylated complementary oligonucleotides.

Plant Materials and Chemical Application

Wild-type Arabidopsis thaliana ecotype Columbia (Col) plants were grown in soil at 22°C in growth chambers programmed for a 12-h-light (8000–10,000 lx) and 12-h-dark cycle for 3 wk. Benzothiadiazole (BTH) was freshly added in sterile water to reach a final concentration of 300 μM of active ingredient and then applied as a fine mist to 3-week-old Arabidopsis, ∼0.5–1.0 mL per plant. Another set of control Arabidopsis grown in the same conditions was sprayed with a fine mist of sterile water. Leaves were harvested from plants 1, 3, 5, 7, and 9 d after BTH or water treatment and immediately frozen in liquid nitrogen. Leaf samples were stored at −80°C. For each experiment, three pots of Arabidopsis were treated and collected. The water- or BTH-treated plants were placed in different growth chambers under identical conditions to avoid cross induction.

Total RNA Extraction and cRNA Labeling

Frozen tissue samples were ground to a fine powder in liquid nitrogen using a mortar and pestle. The tissue powder in liquid nitrogen was transferred to a 50-mL sterile tube. We did not allow the samples to thaw. The samples were uniformly homogenized in either TRIzol reagent (GIBCO BRL) or RNAwiz reagent (Ambion) for about 30 sec using a rotor-stator homogenizer Tissue Tearor (Biospec Products). Total RNA was extracted from the tissue homogenates following the manufacturers' protocols and quantitated by absorbance at 260 nm. Total RNA (10 μg) of each sample was converted to double-stranded cDNA using an oligo(dT) primer plus a 5′-end T7 RNA polymerase promoter sequence via the Superscript Choice System for cDNA Synthesis (GIBCO BRL). Double-stranded cDNA was purified by phenol–chloroform extraction, phase lock gel (5 Prime–3 Prime, Inc.), and ethanol precipitation. In vitro transcription was applied on the cDNA samples using T7 RNA polymerase (RNA Transcript Labeling Kit; Enzo Diagnostics). The cRNA was labeled by incorporating biotinylated UTP (Boehringer Mannheim) using a ratio of 1:3 of labeled to unlabeled UTP in the reaction. Next, the labeled cRNA was purified by ethanol precipitation and quantitated by taking absorbance at 260 nm. Before hybridization to capture-probe-coupled microspheres, the cRNA was randomly fragmented by incubating at 94°C for 35 min.

Hybridization of Targets to Capture-Probe-Coupled Microspheres

The 1× hybridization buffer contained 3 M Tetramethylammonium-chloride (TMAC; Sigma), 0.1% SDS, 50 mM Tris-chloride (pH 8.0), and 4 mM EDTA (pH 8.0). The stock hybridization solution was prepared as 1.5× and stored at 50°C to prevent precipitation. In the first step, target samples containing 10 μg of labeled cRNA or a desired amount of nucleic acid fragments in 20 μL were denatured by heating at 95°C for 10 min. Capture-probe-conjugated microspheres (∼7500 beads per color) were mixed in 40 μL of 1.5× hybridization buffer and subsequently added to the denatured target samples. The hybridization mixture was quickly vortexed and incubated at 48°C for 1 h in an Eppendorf microtube incubator (Eppendorf Scientific) with a shaking speed of 300 rpm. The hybridization conditions for DNA fragments were slightly different. For example, for the two experiments shown in Figure 1, A and B, the hybridization temperature was set at 55°C and the time was 15 min. After incubation, the hybridization mixture was spun for 1 min at 14,000 g in a microcentrifuge. Supernatant was carefully removed with a pipette without disturbing the microspheres. The microspheres were washed by adding 50 μL of 1× hybridization solution, mixed by finger flicking, and incubated at 48°C for 5 min without shaking. Microspheres were collected by centrifugation for 1 min at maximum speed and the supernatant was removed. The microspheres were washed a total of three times, then 50 μL of 1× TMAC and 0.5 μL of 1 mg/mL streptavidin-conjugated R-phycoerythrin (Molecular Probes) were added to the microspheres. The solution was briefly vortexed and incubated in the dark at room temperature for 10 min. The microspheres (35 μL) were analyzed on the Luminex¹⁰⁰ system, and at least 200 events of each set of microspheres were counted.

Flow Cytometry Analysis

The Luminex¹⁰⁰ system integrates lasers, optics, fluidics, electronics, and signal processing to identify each set of color-coded microspheres from every other set, and to measure the total fluorescence at the microsphere's surface for each set to quantify the amount of bound target. Each set of microspheres was distinguished by assigned fluorescence color code (different percentage of red and orange) inside the microspheres. The fluorescence associated on the surface of each microsphere corresponding to amounts of targets was detected and measured by a laser detector. The MFI of each subset of microspheres was calculated and displayed on a monitor by a digital signal processor and the Luminex proprietary software. The Luminex¹⁰⁰ system detects fluorescent dyes with an excitation wavelength of ∼532 nm and an emission wavelength ∼580 nm. For each experiment, 200 events of each subset of microspheres were analyzed on the Luminex¹⁰⁰ system to obtain an MFI that was representative of the whole population of each set of microspheres.