StellARray Gene Expression System Revealing Profiles with Unbiased Significance Application Examples

In this paper, we present three application examples demonstrating the utility of the StellARray Gene Expression System to reveal gene expression level changes in diverse biological contexts such as toxicology, cancer, and stem cell differentiation.

By combining Clonetics and Poietics Primary Human Cells with the StellARray Gene Expression System, all from Lonza, the researcher is provided with a synergistic system to reveal gross and subtle changes in gene expression when analyzing in vitro models of human tissues. This is accomplished easily in 96- and 384-well formatted StellARray qPCR Arrays using a standard qPCR instrument and a generic SYBR Green-based Reagent Master Mix. The Global Pattern Recognition (GPR) Data Analysis Tool is optimally suited to generate a ranked list of significantly changed genes within a qPCR dataset. GPR overcomes the inconsistencies associated with conventional single gene normalization procedures by eliminating a priori normalizer selection. Overall, the results show how the StellARray Gene Expression System eliminates false positives and provides TRUE results that are backed by a rigorous statistical analysis.

Introduction

Molecular profiling is widely regarded to be integral for advancing life science research and clinical diagnosis in the 21st century. It is increasingly recognized that subsets of genes (hundreds rather than thousands) provide the most reliable gene profile definitions of any specific biological process.

We have developed a novel solution - The StellARray Gene Expression System - to meet the need of biomedical scientists for reliable profiling of biologically focused gene sets. It is based on quantitative polymerase chain reaction (qPCR), the gold standard for accurate measurements of gene expression, having an inherently wider dynamic range when compared to microarrays.

The StellARray Gene Expression System

The StellARray Gene Expression System is comprised of three parts:

1. StellARray qPCR Array
2. Global Pattern Recognition (GPR) Data Analysis Tool
3. GeneSieve Query

Each StellARray qPCR Array is compiled to provide the highest information content with a scientifically relevant context. Each qPCR array is either a 96- or 384-well formatted, ‘wet lab’ validated group of gene primer sets, focused on specific biological processes. A single gene quality control example illustrating the accuracy and reproducibility of amplification using the StellARray qPCR Array is shown in (Figures 1a & b). StellARray qPCR Arrays are used with standard qPCR instruments and SYBR Green-based Reagents. They provide a streamlined pathway to successful molecular profiling experiments.

\

 

Figures 1a & b: Real-time PCR data showing amplification curves of quadruplicate wells from a StellARray qPCR Array under standardised reaction conditions. 1a. Example of highly reproducible amplication curve data. 1b. A single peak indicating a single PCR product.

The Global Pattern Recognition (GPR) Analysis Tool has solved one of the most fundamental problems facing experimentation using qPCR - “How do I analyze the data and determine REAL changes in gene expression?” The answer to this question is found in the GPR algorithm, which makes gene expression analysis simple, fast and reliable.

The StellARray Gene Expression System also includes the GeneSieve Query, bringing a new level of search capability to researchers by merging databases containing genecentric biological information in a new and innovative way.

GeneSieve Query can filter or “sieve” thousands of genes for a custom population of real-time PCR arrays that have a unifying biological theme, such as a particular disease or pathway of interest.

Why is there a need for the Global Pattern Recognition (GPR) Algorithm?

The traditional approach to measure gene expression changes from qPCR data has been to normalize the results of a gene of interest with respect to a housekeeping gene (a.k.a. a reference or normalizer gene). The general assumption is that the level of expression of the normalizer gene does not change in the context of the experiment and can be used to normalize the variability in RNA quantity between individual samples. By normalizing to a housekeeping gene, in theory, a magnitude of change can be calculated between groups of samples for a gene of interest. However, this mode of analysis is greatly complicated by the fact that housekeeping genes commonly used as normalizers (e.g., GAPDH, ß-actin, and HPRT) can themselves change in apparent expression when comparing tissues or cells in different states (Bustin 2000; Schmittgen et al. 2000; Goidin et al. 2001; Hamalainen et al. 2001).

18S rRNA is another normalizer that intuitively and experimentally seems more stable, but even 18S can vary in comparison to other genes when analyzed by sensitive qPCR techniques (Bustin 2000, Akilesh et al., 2003). Additionally, the use of 18S rRNA preempts the use of oligo-dT as a firststrand cDNA synthesis primer as 18S rRNA is not poly-adenylated and will yield varying amounts when using purified mRNA via oligo-dT based mRNA purification systems. Any small variation in the normalizer amplification would therefore compromise the analysis of the complete qPCR data set.

(Figure 2) illustrates the vagaries of reliance on single gene normalization. In this example, the p-value distribution for 384 genes is compared by normalizing to either 18S or GAPDH. It can be clearly seen that there is statistical discordance with respect to many of the genes determined to be statistically significant changes by either method alone. This illustrates the disparity that can result from single gene normalization.

Figure 2:  Comparison of the p-value distributions of genes normalized to either 18S or GAPDH illustrates the disparity of results produced by single gene normalization. This example shows values from the Human Lymphoma and Leukemia 384 StellARray qPCR Array (Cat. No. 00188333) for cDNA derived from Human Lymphoma compared to Normal Adjacent Tissue RNA (Applied Biosystems, Inc., Part No. AM7268). Genes on the X-axis are sorted according to p-values in ascending order (left to right) relative to the 18S normalization. Data points for each unique gene, normalized to either 18S or GAPDH, are located in the same vertical plane. Overall, quadrants containing data points for both 18S and GAPDH (blue and red) represent genes in statistical agreement between both normalization routines. Specifically, quadrant A (lower left) shows those genes that are significantly different between the two groups when normalized relative to either 18S or GAPDH. Quadrant B (upper left) shows those genes determined to be significantly different when normalized to 18S, but not by GAPDH. Quadrant C (lower right) shows those genes determined to be significantly different when normalized to GAPDH, but not by 18S. Quadrant D (upper right) shows those genes that are not significantly different when normalized to either GAPDH or 18S.

Global Pattern Recognition (GPR) Algorithm, modified for the StellARray Gene Expression System

To combat the above mentioned problems, access to a modified, proprietary GPR algorithm is provided as part of the StellARray Gene Expression System, which is optimally suited to generate a ranked list of significantly changed genes within a qPCR dataset. This unique algorithm and accompanying software overcomes the problem of identifying invariant normalizers by globally positioning changes in gene expression with respect to all genes within an experiment.

GPR provides a true statistical analysis of results based on internal consistency of the data, which makes GPR optimally suited to detect small, but reproducible, changes in gene expression. GPR also eliminates data skewing that results from the variability associated with single gene normalizers.

Compared to conventional analysis methods, the data skew correction provided by the GPR global normalization process produces a linear p-value distribution with a lower false positive rate resulting in more accurate data (Figure 3).

Figure 3: P-value distribution of single gene normalized values for three commonly used ‘Normalizers’ compared to GPR. Each data point represents a single p-value with respect to a single gene. Genes for each method have been independently sorted by p-value. The data was derived from the Human Lymphoma and Leukemia 384 StellARray qPCR Array (Cat. No. 00188333) by determining the expression level changes in cDNA derived from Human Lymphoma compared to Normal Adjacent Tissue RNA (Applied Biosystems, Inc., Part No. AM7268). For illustrative purposes, data above a p-value of 0.5 has been excluded from the graph. The GPR global normalization process produces a linear p-value distribution with a lower false positive rate resulting in more accurate data and solid gene leads.

Only after the genes are statistically evaluated for replicate consistency is the magnitude of the change calculated. An example of the GPR output is illustrated in (Figure 4).

Figure 4: An example of GPR data formatting for data analysis purposes. The magnitude of fold change is calculated after statistical ranking by p-value has been performed.

Application of the StellARray Gene Expression System

Materials and Methods

Since true biological replicates are not necessarily available, it is acceptable to generate samples from split cultures of cells that have been treated as individual samples. This includes RNA purifications and cDNA syntheses and are referred to as biotechnical replicates (‘biotech reps’, for short).

In this example, the toxicology and stem cell applications, we used quadruplicate biotech replicates per group. In some cases, for example, it is acceptable to generate data sets derived from technical replicates. In this context, a technical replicate refers to data sets that are derived from single or pooled cDNA’s that are then split into triplicates (at least) and analyzed as individual replicates.

For the cancer application the data sets were derived from technical triplicates.

Sample Preparation

RNA purification

All RNA purifications were carried out using the RNeasy Mini Kit (RNeasy Mini Kit, Cat. No. 74104, Qiagen, Inc.) and following the manufacturer’s recommendations for cultured cells. One microliter (1µl) of SUPERase-In ([20 U/µl] Cat. No AM2694, Applied Biosystems, Inc.) was added to the bottom of the RNA elution receiver tube (just prior to elution) to improve RNA stability. The RNA quantity was determined using the RiboGreen Assay Kit (Quant-iT RiboGreen RNA Assay Kit, Cat. No. R11490, Invitrogen, Inc.). The RNA quality was determined using the BioAnalyzer 2100 microfluidics-based platform (2100 Bioanalyzer System, Cat. No. G2940CA, Agilent, Inc.).

cDNA Synthesis

Approximately 200ng total RNA was used as template per cDNA reaction using the qScript cDNA Synthesis Kit (qScript cDNA Synthesis Kit, Cat. No. 95047, Quanta Biosciences, Inc.) following the manufacturer’s recommendations.

Cell Culture

All cells were cultured using the specific growth and differentiation guidelines for each cell type.

Toxicology - Clonetics Ready Heps Fresh Human Hepatocytes (Cat. No. CC-2703, Lonza) were cultured on collagen and allowed to recover from shipping for 24 hours in HCM Hepatocyte Culture Medium SingleQuots Kit (Cat. No. CC-4182, Lonza) at 37°C with 5% (v/v) CO2. A 1M stock solution (in ethanol) of Acetaminophen (Cat. No. A7085, Sigma-Aldrich) was used to produce medium containing a final concentration of 10mM Acetaminophen in fresh HCM Medium for the treatment plate. HCM Medium was removed from the cells and 2ml of medium, treated or untreated, was added to the cells. Cells were cultured at 37°C with 5% (v/v) CO2 incubation for 4 hours. Cells were trypsinized, scraped and processed for RNA and cDNA as described above.

Cancer - FirstChoice Human Tumor/ Normal Adjacent Tissue RNA (Cat. No. AM7268 Human Lymphoma/Normal Adjacent Tissue RNA, Applied Biosystems, Inc.) was used as the source for RNA and cDNA was synthesized as described above.

Stem Cells - hMSC Human Mesenchymal Stem Cells (Cat. No. PT-2501, Lonza) were cultured in MSC Growth Medium in T75 flasks for 1 week for expansion (MSCGM Mesenchymal Stem Cell Growth Medium BulletKit, Cat. No. PT-3001, Lonza) at 37°C with 5% (v/v) CO2 incubation. For stem cell osteogenic differentiation, hMSC’s were plated in 24-well plates at a density of 3100 cells/cm² in Osteogenic Basal Medium using the hMSC Mesenchymal Stem Cell Osteogenic Differentiation BulletKit (Cat. No, PT-3002, Lonza). Differentiation was initiated 2 days after initial plating using the supplied SingleQuots Factors. Medium was changed every 2-3 days. Select wells were stained for positive mineralization with OsteoImage Mineralization Assay (Cat. No. PA-1503, Lonza) at day 21. Cells were harvested on day 22 of differentiation and processed for RNA and cDNA as described above.

Real-time PCR

Quantitative PCR data was collected using the three application- specific 384-well StellARray qPCR Arrays: Toxicology Application - Human General Toxicology 384 StellARray qPCR Array (Lonza Cat. No. 00188331) Cancer Application - Human Lymphoma & Leukemia 384 StellARray qPCR Array (Lonza Cat. No. 00188333) Stem Cell Differentiation Application - Human Stem Cell 384 StellARray qPCR Array (Lonza Cat. No. 00188335)

Each StellARray qPCR Array well was loaded with 10 microliters of sample-specific, SYBR Green Master Mix containing a chemically modified hot-start Taq Polymerase. The array was heat sealed, and run on a 7900HT Sequence Detection System (Applied Biosystems, Inc.) using default cycling parameters for 40 cycles (1 cycle of 50°C for 2 minutes, 1 cycle of 95°C for 10 minutes, and 40 cycles of 95°C for 15 seconds and 60°C for 1 minute). Fluorescence data was acquired during the 60°C anneal/extension plateau. Post-run data collection involved the setting of a common threshold (Ct) across all arrays within an experiment, exportation and collation of the Ct values, and analysis via GPR.

GPR algorithm

Data input for GPR consists of a list of Ct values, derived directly from real-time PCR instruments, for each sample (control or experimental). After designating the control and experimental sets of Ct values, GPR filters expression data to separate genes into two groups - genes considered for analysis (G) and genes that can be used as potential normalizers (N). Genes that are not expressed in either sample are not considered further. After filtering the data into two sets, GPR performs a proprietary global normalization and statistical analysis by comparing each G to each N. The magnitude of change in expression (“fold change”) for each gene is subsequently determined using the ten best N genes, as defined within each experiment.

Results

Toxicology

In concert with Clonetics Ready Heps Fresh Primary Human Hepatocytes, we evaluated the GPR-specific expression profile of the effect of exposure to Acetaminophen (APAP). This classic gene expression profiling example involves the demonstration of Acetaminophen-derived hepatotoxic effects.

GPR analysis yielded 61 significantly changing genes (‘hitters’) from 383 potential targets (see Table 1, Figure 5) using the Human General Toxicology 384 StellARray qPCR Array (Lonza Cat. No. 00188331). Ingenuity Pathway Analysis (Ingenuity Systems) (Figure 5a) identified the following top 5 canonical pathways associated with the GPR hitters:

1. NRF2-mediated Oxidative Stress Response
2. IL10 Signaling
3. Hepatic Fibrosis / Hepatic Stellate Cell (HSC) Activation
4. Aryl Hydrocarbon Receptor Signaling
5. PXR/RXR Activation

Table 1: Toxicology - GPR results derived from Acetaminophen treated Clonetics Ready Heps Primary Human Hepatocytes. Results shown represent genes whose expression levels are statistically different between the treated and untreated groups and fold change values are displayed with respect to the treated group.

Figure 5: Graphical representation of Table 1. Toxicology GPR results.

Figure 5a: Ingenuity Pathway Analysis generated pathway associated with expression levels of MTTP and TNF, the two highest statistically ranked genes, and other associated genes.

Cancer

A lymphoma is a type of solid neoplasm that originates from lymphocytes. The FirstChoice Human Tumor RNA sample is derived from a large B-cell type lymphoma, which arose in the small intestine of a 61 year old female. We analyzed a tumor and matched normal sample from a single individual. As expected from the tumor type, there is a significant enrichment for genes associated with a B-cell lineage in the GPR results.

GPR analysis yielded 75 significantly changing genes (‘hitters’) from 383 potential targets (see Table 2, Figure 6) using the Human Lymphoma & Leukemia 384 StellARray qPCR Array (Lonza Cat. No. 00188333). Ingenuity Pathway Analysis (Ingenuity Systems) (Figure 6a) identified the following top 6 canonical pathways associated with the GPR hitters:

1. B Cell Receptor Signaling
2. IL2 Signaling p53 Signaling Cell
3. Cycle: G2/M
4. DNA Damage Checkpoint Regulation
5. Aryl Hydrocarbon Receptor Signaling
6. Glucocorticoid Receptor Signaling

Table 2: Cancer - GPR results derived from FirstChoice Human Tumor and Normal Adjacent Tissue RNA. The tumor RNA sample is derived from a large B cell type lymphoma, which arose in the small intestine of a 61-year-old female.

Figure 6: Graphical representation of Table 2. Cancer GPR results.

Figure 6a: Ingenuity Pathway Analysis generated pathway associated with expression levels of CD83, the highest statistically ranked gene, and other associated genes.

Stem Cell Differentiation

Cell differentiation is a transition of a cell from one cell type to another and involves a switch from one pattern of gene expression to another. Mesenchymal stem cells, derived from the bone marrow, can give rise to stromal cells, fat cells, and types of bone cells. In this application, differentiation conditions were selected to drive the hMSC cells to an osteogenic phenotype.

GPR analysis yielded 87 significantly changing genes (‘hitters’) from 383 potential targets (see Table 3, Figure 7) using the Human Stem Cell 384 StellARray qPCR Array (Lonza Cat. No. 00188335). Ingenuity Pathway Analysis (Ingenuity Systems) (Figure 7a) identified the following top 8 canonical pathways associated with the GPR hitters:

1. Hepatic Fibrosis / Hepatic Stellate Cell Activation
2. Wnt/β-catenin Signaling
3. IL6 Signaling
4. Aryl Hydrocarbon Receptor Signaling
5. Acute Phase Response Signaling
6. Ephrin Receptor Signaling
7. Cell Cycle: G1/S Checkpoint Regulation
8. JAK/STAT Signaling

Table 3: Stem Cell Differentiation - GPR results derived from Poietics hMSC Human Mesenchymal Stem Cells that have been cultured under conditions to stimulate osteogenic differentiation. Results shown represent genes whose expression levels are statistically different between the differentiated and undifferentiated groups. Fold Change values are displayed with respect to the differentiated group.

Figure 7: Graphical representation of Table 3. Cancer GPR results.

Figure 7a: Ingenuity Pathway Analysis generated pathway associated with expression level of ANPEP, the highest statistically ranked gene, and other associated genes.

Summary

The StellARray Gene Expression System revealed gene expression level changes in three diverse biological contexts. Clonetics and Poietics Primary Human Cells from Lonza were challenged with various experimental conditions and the StellARray Gene Expression System revealed true changes in gene expression levels. The StellARray Gene Expression System shows obvious merit for the purpose of accurately measuring gene expression levels within in vitro models of human tissues as well as tumor samples. By utilizing objective designs of gene content within a StellARray qPCR Array and the novel Global Pattern Recognition Analysis Tool, a statistically significant gene expression profile was easily obtained.

Using existing qPCR platforms and standard chemistries, reliable information can be derived from relatively small sets of target genes without the error associated with a priori chosen normalizer genes and fold-change biases.

References

Review - Absolute quantification of mRNA using real-time reverse transcription polymerase chain reaction assays. Bustin SA. (2000) Journal of Molecular Endocrinology 25, 169-193.

Effect of experimental treatment on housekeeping gene expression: validation by real-time, quantitative RT-PCR. Schmittgen TD, Zakrajsek BA. (2000) J Biochem Biophys Methods. Nov 20; 46(1-2), 69-81.

Ribosomal 18S RNA prevails over glyceraldehyde-3-phosphate dehydrogenase and beta-actin genes as internal standard for quantitative comparison of mRNA levels in invasive and noninvasive human melanoma cell subpopulations. Goidin D, Mamessier A, Staquet MJ, Schmitt D, Berthier-Vergnes O. (2001) Anal Biochem, Aug 1;295(1):17-21.

Identification and validation of endogenous reference genes for expression profiling of T helper cell differentiation by quantitative real-time RT-PCR. Hamalainen HK, Tubman JC, Vikman S, Kyrola T, Ylikoski E, Warrington JA, Lahesmaa R. (2001) Anal Biochem, Dec 1;299(1):63-70.

Customized Molecular Phenotyping by Quantitative Gene Expression and Pattern Recognition Analysis. Shreeram Akilesh, Daniel J. Shaffer, and Derry Roopenian. (2003) Genome Research 13:1719-1727.