Medical Device Link.

Originally Published IVD Technology July 2003

Molecular Diagnostics

Figure 1. Conceptual schematics of the Flow-Thru chip. A silicon wafer with organized microchannels open at two ends allows for the flow of fluids through the channels. Oligonucleotide probes occupy 100 channels and are bound to the inner surface of the channels. Click to enlarge.

Researchers have been demanding tools that will enable them to access globally and systematically the rapidly growing amount of molecular information. This information could help researchers and IVD manufacturers identify the precise genes and encoded proteins that mark a disease state, discover new therapeutic targets, and investigate pathways of interest. 

The molecular field is shifting gears from performing global whole-genome analyses to performing more-detailed, gene-specific analyses. These detailed analyses are intended to better understand patterns of gene expression for a small, focused subset of genes that serve as a diagnostic set upon which conclusions are made. 

Accordingly, the second generation of focused arrays is now required to study the expression of genes and proteins involved in regulatory pathways and specific diseases. By focusing on well-characterized pathways, scientists can obtain answers to questions in areas such as predictive toxicology and cancer tumor type classification.^1–3

Figure 2. Expression raw data of 20 genes in 226 samples (137 normal samples versus 89 tumor samples) were exported and analyzed by the Partek Pro 2000. To configure dissimilarity for multidimensional scaling, metric method and Euclidean distance function were used. Normal samples are represented by red squares, and tumor samples are represented by green squares. Click to enlarge.

Microarrays serve as an important tool to query biological phenomena and obtain differential gene-expression information on relevant sets of genes. More recently, arrays with a small number of probes and custom gene content, as well as theme arrays that focus on certain diseases, pathways, or gene families, are recognized to be best suited as diagnostic tools. A comparison of the salient features of the 4-D array and the standard 2-D array reveals the advantages of the 4-D array (see Table I).

Microarray technology draws from a combination of disciplines including surface chemistry, microfabrication, engineering, bioinformatics, and molecular biology. In a conventional two-dimensional array, oligonucleotides or cDNA representing specific genes are immobilized on a solid support made of glass, silicon, plastic, or nylon. The attachment of DNA is either through a covalent or hydrophobic attachment. 

Recent developments in microarrays include improving the chemistry for attachment, increasing the flatness of the glass, and testing the polish surface to improve accuracy. Plastic supports and gel-modified surfaces on glass slides are also on the market, and lend some depth to the binding of probes.

There are a number of applications for microarrays including the following: transcription profiling for target discovery; lead optimization; exon mapping; single nucleotide polymorphism analysis and genotyping; identifying molecular markers for tumor classification; predicting gene function; and diagnostic and prognostic analysis of diseases.

Market Niche

Figure 3. Representative images of two different chips after the hybridizations with a normal colon (a) and a colon tumor (b) sample. Chip probes for all genes were selected based on the expression profile differences between normal samples and colon tumor samples. These differences for medium- and high-expressing genes are visible by the human eye, but are not visible for low-expressing genes. Click to enlarge.

The molecular marketplace is changing rapidly. According to some studies, the total market for molecular diagnostics is expected to grow at a compound annual rate of 16% and reach $5 billion worldwide by 2007.⁴ As sophisticated nucleic acid–testing systems, microarrays have the potential to become an important tool in the molecular diagnostics market and contribute to this growth. 

The main advantage of microarrays is that they can be used to create large sets of data, identify markers or candidate genes, and predict results compatible with clinical outcomes. These advantages indicate that medium-density arrays have a strong position in clinical diagnostics. What was once a theoretical consideration that gene expression patterns are biologically and clinically relevant is now a reality.

Several reports have indicated that measuring gene expression patterns using DNA microarrays based on molecular signatures provides a high probability of prognosis for patients.^5–7 These signatures can be used to identify patients that need further treatment or those for whom treatment can be spared. Currently, microarrays are being used to distinguish normal tissue from tumors as well as to subtype tumors. 

Some of the areas that are targeted by microarrays in the diagnostics field are pharmacogenomics, disease risk projections, and disease predisposition studies. Other clinical applications for DNA arrays include:

• Discriminating subtypes of cancers such as melanoma, lymphoma, and breast cancer, with emphasis on gene expression patterns, tumor progression, and treatment resistance. 
• Rapid identification of infectious diseases; and subtyping bacteria, viruses, and parasites, including antibiotic resistance.

Figure 4. The signal-to-noise value of each gene was normalized to GAPDH. Fold change was calculated between tumor and normal samples for each gene. Click to enlarge.

However, conventional microarrays face several challenges that may hinder them from better serving the diagnostics market and being adopted in clinical diagnostics. End-users are concerned about cost, ease of use, reliability, and reproducibility. Technological challenges include reliability of the assay, sensitivity of detection, standardization and normalization, and data analysis and interpretation. There are also market issues, such as intellectual property issues and competitive advantages over current and existing technologies.

To resolve these challenges, any microarray platform will have to: 

• Package the array in a clinically friendly kit with all the reagents required to make it simple for a technician to run a test and produce reproducible and statistically significant results.
• Package the image-capture software and data analysis tools to make them easy to use and understand.
• Provide annotation for the genes and gene sets such that the test results can be easily understood and explained by physicians who may not be well versed in interpreting genomics data.
• Increase the uniformity and reliability of the assay to a point where variability starting from sample prep to data analysis is minimized, with the goal of increasing the reproducibility of the assay.
• Improve the numerical analysis tools for simple clustering methods that are essential for accurate diagnosis.

4-D Array Systems

Table I. Salient features of standard 2-D flat arrays as compared with the 4-D array. Click to enlarge.

For a microarray platform to be a reliable and easy-to-use tool for clinical research and diagnostics, it is important that the platform provides an increase in sample throughput; a decrease in variability, labor, sample size, and assay time; as well as being completely automated. 4-D array systems are medium- to low-density platforms that offer differential advantages from the more common flat-glass-slide arrays. 

The porous microchannels in these systems provide an organized surface, resulting in uniformity in spot size and spot morphology. These channels also provide greater surface-to-volume ratio compared with a flat slide, enhancing the responsiveness of the assay in addition to reducing assay time and reagent volume.

The Flow-Thru chip by MetriGenix Inc. (Gaithersburg, MD) used in 4-D arrays consists of a porous silicon chip that is 1 ¥ 1 cm and 0.5 mm thick. Silicon offers advantages since it is pristine and well ordered compared with glass. Within the chips, there are approximately half a million microchannels that are coated on the inner surface. The noninterconnected microchannels (10 µm in diameter and 500 µm in length) connect the upper and lower surfaces of the chip in a manner that allows fluid to flow through them (see Figure 1). 

Table II. A breakdown of the samples from a gene expression database that contained normal and clinically distinct cancerous colon samples. Click to enlarge.

Silicon wafers are modified to form an intermediate film layer for biomolecule immobilization. The intermediate layer is produced by immersing the wafers in a solution of 2% glycidoxylpropyltrimethoxysilane in methanol. The silanized wafers are cleaned by sonication and dried at 85°C. 

The wafers are divided into chip sections using a photolithographic step during a pore formation process and then broken into individual chips prior to printing. Gene-specific oligonucleotide probes with 5' amino group modifications are covalently bound to the inner surfaces of the microchannels, increasing the binding capacity 100-fold compared with a similar-sized spot on a flat surface.

Printed arrays have gene-specific spots that are approximately 100 µm in diameter, occupying about 100 channels. Oligonucleotide probes are designed to be specific for the genes of interest using proprietary probe-design software. Hybridization occurs via automated microfluidics, which provides uniform hybridization with a low sample volume. Targets labeled with biotin are detected via chemiluminescence using a Streptavidin-HRP conjugate. 

During hybridization, the labeled targets flow through the 10-µm microchannels, minimizing the diffusion distance required for a target molecule to find and hybridize to the bound capture probes. 
Other components in this 4-D microarray system include the following:

• An automated fluidic processing station for the 4-D array, composed of a fluid pumping and solenoid valving system for buffer and reagent delivery. 
• A detection instrument with a cooled 12-bit CCD camera that has a 1.3 megapixel resolution. 
• An image capture and image analysis program. 

While capturing the image, the substrate is continuously pumped through the chip to ensure an enzyme limited reaction, resulting in higher signal intensities per unit time compared with static chemiluminescence-detection methods used in typical membrane arrays.³ The detection station uses an automatic exposure routine to take advantage of the maximum dynamic range of 3 logs of the detection system. 

These instruments have to be calibrated regularly to keep the fluidics uniform. One of the biggest challenges has been to reduce the variability that may be caused due to the instruments. In addition, changes in temperature tend to introduce bubbles in the tubing and chambers, which can lead to inconsistent data.

4-D Colon Cancer Array

While colorectal cancer is one of the most common types of cancer in both men and women, it is highly treatable and often curable when detected early, before progressing to an invasive cancer. The ability to identify patients who may be at a higher risk to develop colon cancer also provides better management of the disease. Molecular classification of tumors using gene expression microarrays offers the potential to alter the practice of surgical pathology and oncology.^8–9 Further validation of the identified colon cancer signature genes requires platforms with good sensitivity and precision, ease of use, and higher throughput capability to accommodate large sample volumes and data comparison. 

An analysis was conducted using 291 samples from a gene expression database that contained normal colon and clinically distinct colon cancer classifications. According to a breakdown of the sample classifications in the database, normal, adenoma or benign tumor, and adenocarcinoma samples were included in the statistical analysis (see Table II). A correlation was performed on the expression profile of 312 housekeeping-gene fragments in the samples, and showed 29 samples (20 normal and 9 adenocarcinoma) had correlations less then 0.1. These samples were removed from the analysis. 

In order to find genes with significant differential gene-expression levels between normal and tumor samples, genes were filtered using the Z-test algorithm with a cutoff of |Z-score| > 8.0. A total of 202 genes were selected based on this process. 

Raw expression data for these 202 genes in a total of 226 samples were then exported to a software package and analyzed using a linear discriminate analysis. This analysis allowed the gene set to be reduced to a smaller set of 20 genes. The 20 genes can differentiate normal samples from tumor colon samples (see Figure 2). These 20 genes can also be used as potential biomarkers to separate normal and tumor samples with a certain degree of prediction power. 

A similar approach was taken to identify genes with significant differential expression levels between benign and malignant samples, and early- and late-stage tumor samples. Filtering of the genes was performed with a cutoff of |Z-score| > 4.8. A total of 100 genes were selected which differentiated benign and malignant samples. At the same time, a set of 11 genes was selected which differentiated early- and late-stage tumor samples. After using a linear discriminate analysis, the number of genes highly associated with malignant disease was reduced from 100 to 23 genes. 

The average prediction error was calculated from a cross-validation analysis of the prediction model with the smaller refined gene sets. The error rates were 4.2%, 3.5%, and 6.9% for normal versus tumor samples, benign versus malignant samples, and early- versus late-stage tumor samples, respectively. 

A 4-D colon cancer array by MetriGenix also has these signature genes available to evaluate gene expression differences between normal colon and colon tumor samples. This array contains sense-strand oligonucleotide probes designed to capture the antisense sequence of the targets. The probes are 60 bases long with a specific delta G, GC-content, and melting temperature. In order to minimize the possibility of probe cross-hybridization, repetitive sequences or sequences with low complexity were masked prior to probe search. Each probe sequence is searched against sequence databases to ensure minimal similarity to any gene in the genomic database. 

Commercially available total RNA from normal colon and colon tumor tissue was purchased, and cRNA was isolated from total RNA.¹⁰ The labeled target and hybridization of the arrays were carried out as described in the instruction manual.¹¹

A 4-D colon cancer chip was used to determine whether these signature genes are differentially expressed in normal colon and colon tumor samples based on an analysis of the 4-D data from six normal colon samples and six colon tumor samples. Representative images of two different colon cancer arrays after hybridization with normal colon RNA and colon tumor RNA were generated (see Figure 3). Quantitative RT-polymerase chain reaction (PCR) was also performed on 20% of the total genes. 

When tested with total RNA from normal colon and colon tumor tissues, the 4-D array results were consistent with expression data from earlier reports and quantitative RT-PCR (see Figure 4). The fold change directionality is consistent for all genes across the three platforms. The exception was gene 3, which is a very low abundant gene in both normal colon and colon tumor samples. When gene abundance is low it is difficult to calculate the fold change, resulting in discrepancy in the directionality in gene 3.

This 4-D colon cancer array is also designed to understand the molecular basis of colon neoplasia, such that this knowledge can be used to produce improved diagnostic and therapeutic approaches for this disease. The data generated should provide valuable insight into specific roles of the genes, either individually or as a group, in the development and progression from normal colonic epithelium through benign adenomas to invasive colon cancer.

Conclusion

	Mridula Iyer, PhD, is product manager; Reena Philip, PhD, is a project leader and scientist in the R&D department; Hoyt E. Matthai is vice president of operations; Eric Eastman, PhD, is senior vice president and chief scientific officer; and Andrew J. O’Beirne, DrPH, is president and chief executive officer at MetriGenix Inc. (Gaithersburg, MD). They can be reached via miyer@metrigenix.com,  rphilip@metrigenix.com,  hmatthai@metrigenix.com,  eeastman@metrigenix.com, and aobeirne@metrigenix.com,  respectively.

The 4-D array platform has been developed to be fast, robust, and user-friendly. Although this platform is currently targeted for use in basic and medical research, future developments are intended to eventually move this technology to actual use in clinical diagnostics. A second-generation high-throughput microarray platform is also being developed that is capable of processing 96 arrays simultaneously while maintaining the advantages of the 4-D system. 

References

1. RS Thomas et al., “Identification of Toxicologically Predictive Gene Sets Using cDNA Microarrays,” Molecular Pharmacology 60, no. 6 (2001): 1189–1194.

2. J Khan et al., “Classification and Diagnostic Prediction of Cancers Using Gene Expression Profiling and Artificial Neural Networks,” Nature Medicine 7, no. 6 (2001) 673–679.

3. B Cheek et al., “Chemiluminescence Detection for Hybridization Assays on the Flow-Thru Chip, a Three-Dimensional Microchannel Biochip,” Analytical Chemistry 73 (2001): 5777–5783.

4. Molecular Diagnostics: Transforming the Pharmaceutical Market. (Westborough, MA; Drug and Market Development Publishing, 2002).

5. C Sotiriou et al., “Gene Expression Profiles Derived from Fine Needle Aspiration Correlated with Response to Systemic Chemotherapy in Breast Cancer,” Breast Cancer Research 4, no. 3 (2002): 1–8

6. M Takahashi et al., “Gene Expression Profiling of Clear Cell Renal Cell Carcinoma: Gene Identification and Prognostic Classification,” Proceedings of the National Academy of Sciences 98, no. 17 (2001): 9754–9759.
7. JB Welsh et al., “Analysis of Gene Expression Profiles in Normal and Neoplastic Ovarian Tissue Samples Identifies Candidate Molecular Markers of Epithelial Ovarian Cancer,” Proceedings of the National Academy of Sciences 98, no. 3 (2001): 1176–1181.

8. TT Zou et al., “Application of cDNA Microarrays to Generate Molecular Taxonomy Capable of Distinguishing Between Colon Cancer and Normal Colon,” Oncogene 2, no. 31 (2002): 4855–4862.
9. TJ Giordano et al., “Organ Specific Molecular Classification of Primary Lung, Colon, and Ovarian Adenocarcinoma Using Gene Expression Profiles,” American Journal of Pathology 159, no. 4 (2001): 1231–1238.

10. DJ Lockhart et al., “Expression Monitoring by Hybridization to High-Density Oligonucleotide Arrays,” Nature Biotechnology 14, no. 13 (1996): 1675–1680.

11. MGX 4D Array Instruction Manual (Gaithersburg, MD: MetriGenix, 2002). 

Copyright ©2003 IVD Technology