| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Molecular Diagnostics and Genetics |
1 Nichols Institute, Quest Diagnostics, San Juan Capistrano, CA.2 Thermo Electron, Inc. Point of Care and Rapid Diagnostics, 331 S. 104th St., Louisville, CO 80027.
aAuthor for correspondence. E-mail rob.jenison{at}thermo.com.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: The CF PortraitTM chip contains an 8 x 8 array of capture probes and controls to detect all requisite alleles. Single-tube multiplex PCR with 15 biotin-labeled primer pairs was used to amplify sequences containing all single-nucleotide polymorphisms to be interrogated. Detection of a thin-film signal created by hybridization of multiplex PCR-amplified DNA to complementary capture probes was performed with an automated image analysis instrument, NucleoSightTM. Allele classification, data formatting, and uploading to a laboratory information system were fully automated.
Results: The described platform correctly classified all mutations and polymorphisms and can screen 
1300 patient samples in a 10-h shift. Final validation was performed by two separate 1000-sample comparisons with Roche CF Gold line probe strips and the Applera CF OLA, Ver 3.0. The CF Portrait Biochip made no errors during this validation, whereas the Applera assay made seven miscalls of the IVS-8 5T/7T/9T polymorphism
Conclusions: The CF Portrait platform is an automated, high-throughput, DNA chip-based assay capable of accurately classifying all CF mutations in the recommended screening panel, including the IVS-8 5T/7T/9T polymorphism.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
30 years, CF remains a chronic, debilitating disease with emotional and financial costs to individuals, families, and society (4)(5).
In 1989, the gene responsible for CF was cloned, and the most common CF mutation (
F508) was identified (6)(7)(8). Although F508 accounts for 60% of CF mutations, more than 1000 separate CF-causing mutations in the CF transmembrane regulator protein (CFTR) have been described (9).
The American College of Medical Genetics (ACMG) published a recommended panel of 25 mutations and 6 polymorphisms for population-based CF screening in May 2001 (2). In October of that year the American College of Obstetrics and Gynecology (ACOG) recommended that CF carrier detection be offered to all pregnant Caucasian couples (1). Since that time, testing volumes have increased dramatically (10) and continue to increase.
We describe the development and technical validation of a DNA chip based on an interference-based detection system of nucleic acids on optically coated silicon chips (11).
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
dna samples
Genomic DNA samples were purified from peripheral blood specimens as described previously (10). For the technical validation, samples of known genotype were purchased (Coriell), or genotype was determined by the Applera CF OLA ASR, Ver. 3.0 (Applera Diagnostics) or the CF Gold line probe assay (Roche Molecular Systems) before testing. For the platform comparison, patient samples of previously unknown genotype were tested at Nichols Institute. In the first study, 1092 patient samples that had been tested with the CF Gold line probe assay between September 2002 and January 2003 were randomly chosen for testing on the CF Portrait chip in February 2003. For the second comparison, 1076 samples that had been tested with the Applera CF OLA ASR, Ver. 3.0, platform in January 2003 were tested on the CF Portrait chip in February 2003.
Although information such as patient ethnicity and pertinent family history was requested for each patient, in practice, this information was rarely provided. Because the indication was not routinely provided, we could not determine whether the test was for carrier detection vs mutation detection in a patient with CF or for infertility evaluations.
assay principle
The hybridization-based single-nucleotide polymorphism (SNP) platform described here was configured on a multilayered, optically coated silicon surface (Fig. 1
, A and B) that was highly reflective with a gold appearance under white light. Surface color was a consequence of the low reflectance of blue wavelengths and high reflectance of red wavelengths of light from the surface (Fig. 1C
). Capture probes were covalently attached to the surface. Biotin-labeled PCR amplicon was denatured and hybridized to the complementary, allele-specific capture probes. The surface-immobilized duplex was measured by the binding of an anti-biotin antibody conjugated to horseradish peroxidase (HRP). In the presence of a precipitating substrate for HRP, a molecular thin film was deposited on the surface (Fig. 1B
). The additional mass on the surface altered the pathlength of light reflected from the optical surface layers and attenuated specific wavelengths of light through destructive interference (Fig. 1C
). The result was an apparent surface color change from gold to blue. The surfaces were designed so that a minimal thickness increase (20Å) caused a color change to which the human eye is maximally sensitive. We have previously demonstrated an analytical limit of detection of 10 fmol/L or 60 000 copies of single-stranded DNA in a model system (11).
|
surface preparation
Details of the preparation of the optically coated silicon wafers have been described previously (11). Briefly, Si3N4 was applied to crystalline silicon wafers in a vapor deposition chamber (Nordiko). The polymer, T-structure aminoalkyl polydimethyl siloxane (TSPS; United Chemical Technologies) was applied by use of a spin coater (Machine Technologies, Inc.) and cured at 150 °C for 24 h to create a thin film biosensor surface. An amino-functionalized surface [TSPS/poly(Lys-Phe)] was prepared by passively adsorbing 50 µg/L poly(Lys-Phe) (Sigma) in 1x phosphate-buffered saline containing 2 mol/L NaCl (pH 6) overnight at room temperature.
probe immobilization
Oligonucleotides were synthesized with a 5' amino modifier C6 phosphoramidite (Glen Research), resuspended in water, and stored at –70 °C until use. The positive control was synthesized with a 3' biotin phosphoramidite. Capture probes were activated as described previously (11)(12)(13). The activated DNA was brought to 750 µL with water, aliquoted into microtiter plates, and lyophilized (Virtis). Frozen DNA in plates was stored at –20 °C for up to 2 months. The dried, activated probes were resuspended with 50 µL of spotting solution [0.1 mol/L sodium phosphate (pH 7.8) containing 100 mL/L glycerol], and 10 nL was applied to the amino-functionalized wafer by use of a commercial printer (BioDot). The wafers were stored for 2 h at room temperature in a humid Petri dish to prevent drying. After immobilization of the allele-specific oligonucleotides (ASOs), the quality of the printed arrays was assessed with a machine vision system interfaced with an XYZ stage (New England Affiliated Technologies). The spotted solution was illuminated, and the image for each array was magnified to interrogate spot number, spot diameter, linearity of rows and columns, and squareness. Mean spot diameter was 500 µm, and center-to-center spacing was 750 µm between neighboring spots. Arrays not passing specific quality-control criteria were marked with a scribe pin to prevent subsequent processing. Wafers were then "stripped" of loosely adsorbed probe with 1 mg/L sodium dodecyl sulfate at 45 °C for at least 2 h, washed with water, dried, and stored at room temperature.
plate manufacture
Each wafer contained 126 arrays, and as configured, production of 3000 arrays/day per printer with >95% overall array yield was possible. The wafers were cut into 49-mm2 chips by use of a scribe/break process (Dynatex) that was an adaptation of a semiconductor process developed to dry-process liquid-sensitive devices. The spotted wafer was mounted on a hoop backed with adhesive tape, and a computer-controlled head moved a diamond tip scribe across the wafer with a controlled force and speed. The diamond scribe head moved in the XY plane to produce scribe lines in the desired pattern over the surface of the wafer. After completion of the scribe process, the scribe/break system then realigned the wafer and began the break process by applying a controlled force to the backside of the adhesive tape, using a breaker bar to break the wafer into chips.
The process of placing the chips into the microtiter plates was accomplished with a modified Magnum 9000X pick-and-place system (ESC Manufacturing). The depth of the microtiter well necessitated an increased vertical travel distance of the pick head. The length of the pick head was modified, and a new cam was designed and manufactured to increase the vertical travel range of the head assembly. The pick-and-place system used pattern recognition to assure that all chips had the correct geometry. If a chip had been marked as defective during previous processing, the system would recognize this and not place the defective chip in the microtiter plate. The system applied a controlled amount of epoxy (3M) to the bottom of the microtiter plate well, placed the chip on top of the epoxy in the well, and verified that the chip had been correctly placed. Finally, the plate was inspected for correct orientation and flatness of the chips by a NucleoSight instrument adapted for use as an optical inspection station.
pcr
PCR reactions were performed in hard shell-microplates (MJ Research). Fifteen amplicons of the CFTR gene, covering all of the mutations in the ACMG CF panel (see Table 1 in the Data Supplement that accompanies the online version of this article athttp://www.clinchem.org/content/vol50/issue5/ ) were amplified in 13.5-µL total volume reactions. The reaction contained 1.0 U of Qiagen HotStart Taq DNA polymerase, 1x Qiagen PCR buffer, 1x Qiagen Q solution, 1.4 mM MgCl2, 0.37 mM each deoxynucleotide triphosphate, 0.74 µg/mL bovine serum albumin, and 15 appropriate primer pairs with final concentrations ranging from 0.04 to 0.320 µM (see Table 2 in the online Data Supplement). HPLC-purified 5'-biotin-labeled primers were purchased from Midland Scientific and were >90% pure.
We used 2 µL of extracted DNA in the PCR reactions. Typically, 86–87 patient samples were amplified in the 96-well plate with 4 wells used for positive controls (a minimum of 3 calibrators and 1 rotating control, all from ACMG-recommended CF screening mutations), including a 5T control, a wild-type control, and 3 no-DNA quality controls with 1 interspersed within the plate and 2 in the last wells of each plate. The thermocyling conditions included an initial hot-start step at 95 °C for 15 min, followed by 34 cycles of denaturation at 94 °C for 15 s, annealing at 56 °C for 20 s, and polymerization at 72 °C for 30 s. Cycling culminated with a final extension at 72 °C for 5 min and then cooled to 4 °C until further use.
image analysis
The NucleoSight system used to read and identify genotypes was a modification of commercial hardware and software. A commercially available imaging system, including a black-and-white camera mounted above an automated XY stage, was modified and reinforced to provide a stiffer support structure, and an exterior protective case was constructed for environmental isolation (see Fig. 1
in the online Data Supplement). A black-and-white camera was chosen because of the relative ease of intensity data analysis compared with color imaging.
The software directed reading of the chips, allele classification, formatting, printing, uploading, and storing of data. We began with the commercial product IPLab 3.5 (Scanalytics, Inc.) and programmed a series of scripts to configure the instrument, control the stage, acquire images, and store images and data. An extension to IPLab was developed to locate and analyze spot values automatically.
The image analysis software development will be described in detail elsewhere. Briefly, potential spots were located by use of an algorithm based on the Hough (14) transform. The high-quality gradient operators of Ando (15) were applied to the image, and the results were processed to generate gradient magnitude and direction information. A threshold operation was applied to the magnitude to select those strong enough to be examined further. A voting algorithm was applied to the threshold magnitude in conjunction with the gradient direction information to locate the center of potential spots. Implementation of the voting algorithm enforced a size range for acceptable spots as well; circular-looking artifacts that were too large or too small to be legitimate reacted spots were ignored. Pixels that received enough "votes" above threshold were considered potential spot centers.
Once potential spots were located, an array of rows and columns was deduced by heuristic reasoning to create a template for genotype classifications (16). The signal from each spot was composed of gradient information combined with spot size, shape, and edge quality. Use of all of these factors in analysis allowed accurate identification and measurement of true spots and rejection of surface contamination such as scratches. The software analyzed the spot signal values for the wild-type and mutant ASO hybridization for each mutation and used a ratio to classify the genotype for that particular pair. For wild-type/mutant spot pairs, the ratio was calculated as: WT/(WT + MU), where WT is the signal from the wild-type spot and MU is the signal from the mutant spot.
General rules and truth tables were constructed to classify all possible permutations and combinations of genotypes; the data were then formatted for review. At the discretion of the user, the software can be configured to transfer formatted data or to interface with a LIS to initially upload patient information and then to download the genotype classification. Ratio thresholds used in the calculations are user adjustable. The system also has facilities to store all images and data on CD-ROM. The analysis of each image required 200 ms. Infrequently, the software was unable to make a genotype classification because of spot defects or the lack of complete data. The defects can include misshapen spots and spots that are too large or too small as printed by the spotter to be interpreted by the image analysis software within the acceptable size range. Assay artifacts, such as poor washing, can also leave streaks of color on the chip surface that interfere with image analysis. Lack of signal except on the control spots indicated PCR failure, lack of delivery of PCR amplicon, or another critical assay reagent. In these instances image analysis failure was reported. The user was prompted to review the image and could then make a "manual" call, if sufficiently good data were available, or could choose to retest the sample. A printout of the graphical image allowed laboratory technologists and directors to review the NucleoSight allele classifications, overrule automated classifications, make manual genotyping decisions, and then finalize genotype assignments and initiate uploads into the LIS.
assay procedure
Two protocols were optimized: a semiautomated platform for intermediate throughput (250 samples per technologist per 8-h shift) and an automated protocol (1300 samples per technologist per 10-h shift). For the semiautomated platform, a plate washer (Thermo Electron) and a PTC 200 thermocycler (MJ Research) with a flat block and heated lid were used. An 8-channel multipipettor (Thermo Electron) aided in pipetting steps of the assay.
A walkaway automated protocol using a Freedom Genesis 200 Tecan robotic workstation (TRW) equipped with a Robotic Manipulator arm was also optimized. The TRW was fitted with a TEMO 96-channel robotic pipetting station, a 96-channel plate washer, and two MJ Research thermocyclers equipped with remote Alpha docks with Power Bonnet and flat blocks. Incubation hotels and open hotels were also placed on the TRW, and TeStack provided the necessary supply of disposable pipette tip boxes. Tecan-provided Gemini pipetting software and FACTS scheduling software were used.
For both the semiautomated and the automated protocols, the assay involved diluting PCR products (1.8 µL) in 8.2 µL of distilled, deionized H2O and denaturing the double-stranded DNA by use of 10 µL of a 0.1 mol/L NaOH–20 mmol/L EDTA (pH 8) solution (both reagents from Amresco). Hybridization solution (180 µL; 5x standard saline citrate–1 mg/L sodium dodecyl sulfate–5 mg/L BlockAid; Thermo Electron) was added to CF Portrait plate wells and incubated at 51 °C for prehybridization for 20 min. After 10 min of alkaline denaturation, amplicons were added to the prehybridization buffer in the plates. Hybridization was performed at 51 °C for 10 min. Plates were then washed with wash buffer A (1 mg/L sodium dodecyl sulfate in 0.1x standard saline citrate) followed by wash buffer B (0.1x standard saline citrate). We added 200 µL of 1 µg/mL conjugate solution (peroxidase-conjugated IgG fraction monoclonal mouse anti-biotin, made in hybridization buffer; Jackson Research) to each well, and samples were incubated at room temperature for 10 min followed by washing with wash buffer A followed by buffer B. We then added 200 µL of 3,3',5,5'-tetramethylbenzidine reagent (BioFX), incubated the plates for 5 min at room temperature, and then washed them with distilled, deionized H2O. Chips were finally washed with methanol manually and dried upside down in a 65 °C oven. This semiautomated assay required 70 min to process a 96-well plate.
In the automated protocol, an operator placed between one and eight CF Portrait plates (96–768 samples) and corresponding numbers of 96-well plates containing PCR-amplified patient DNA and controls on the automated platform workspace. Liquid reagents and disposable consumables were also placed on the automated platform. The FACTS scheduling software coordinated the entire procedure based on the number of plates to be analyzed. On the TRW, a maximum throughput of eight plates in 5.33 h can be achieved. Considering that 87 patient samples can be analyzed per plate, 696 patient samples could be analyzed during that time frame. Once the robotic protocol was initiated, no intervention was required, making it a walkaway system until the final manual methanol wash. Dried plates were placed on the NucleoSight imaging station and analyzed as described.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
F508, and 507 were designed empirically.
We chose stringent assay conditions in an attempt to limit nonspecific hybridization. As a result, AT-rich sequences tended to be longer, causing potentially reduced discrimination. Two approaches were successful in improving discrimination for some of the high-AT-content (>70%) ASO pairs. As a primary approach, a second artificial base change was introduced into the ASOs for the mutant alleles 621 + 1C>T and I507. The introduction of an artificial base change remote from the actual mutation influences overall stability of the duplex and increases the discrimination of the ASO. To find the optimum location for the artificial or secondary mutation, nearest-neighbor calculations were used. Use of this strategy increased the difference in calculated Tms (Tm) between the wild-type and mutant alleles from 5.1 °C to 9.4 °C and 5.9 °C to 9.1 °C for 621 + 1G>T and I507, respectively.
The second method of increasing ASO discrimination for AT-rich ASOs involved incorporation of modified bases into the capture probe to affect the Tm. After failing to achieve acceptable discrimination for the 2184 delA mutation using the above strategy, we introduced modified bases into the ASO for the mutant allele. The addition of 5-methylcytosine and 2-aminopurine adjacent to the deletion improved discrimination dramatically. Interestingly, the addition of the modified bases, although altering the measured Tm of the probe, did not alter the measured Tm of the ASO probe set (data not shown). However, the performance of the modified probes was superior to their unmodified counterparts with an increase in the signal-to-noise ratio from 3 to 22 under optimized conditions. The nature of this effect is currently under investigation.
Variation of the input concentrations of capture ASOs spotted on the chip was essential to allow unambiguous genotype discrimination on a single chip. The optimum input concentration of each capture ASO was determined empirically with known genotype targets (Fig. 2
). To discover and optimize capture probe sequences, we used multiplex PCR-amplified genomic DNA of known genotypes. We designed the multiplex PCR to contain all 25 mutations and 6 polymorphisms recommended by the ACMG (see Table 1 in the online Data Supplement). This required 15 primer sets that amplified the appropriate genomic sequences (see Materials and Methods). For homozygous targets, specific hybridization ratios were varied to provide maximum signal in the absence of detectable noise; probe concentrations were varied to give a 1:1 signal for heterozygous samples. Final probe input concentrations varied over a 20-fold range. Optimized probe sequences and their respective input spotting concentrations are available in Table 3 of the online Data Supplement.
|
An interesting observation was made during ASO development. In the initial stages of development we did not have genomic control DNA for one of the mutations, 2184delA. We were using a synthetic, double-stranded oligonucleotide as a positive control for this sequence, and the assay seemed to be performing well. However, when we eventually obtained a genomic control for this mutation, the assay did not detect this mutation under standard assay conditions, causing us to consider other probe sequences. This was almost certainly attributable to complexity differences between the synthetic oligonucleotide control and PCR-amplified genomic DNA and indicated the importance of using genomic DNA for assay development.
technical validation
Once the composition of the chip and the assay conditions were finalized, several experiments were conducted to validate the performance of the chip with known (Coriell) or previously genotyped samples (Roche or Applera CF OLA assays). Because the experiments were performed over a period of several months and the available controls changed, the same samples were not used for all of the experiments. However, all mutations of the ACMG/ACOG panel were tested in each experiment. This provided additional validation of the assay because different samples were used in various experiments. In addition to wild-type controls and heterozygotes for each ACMG mutation and polymorphism, DNA from 12 compound heterozygotes (F508/1898 + 1G>A, 711 + 1G>T/F508, G85E/621 + 1G>T, 3659delC/F508, 3120 + 1G>A/621 + 1G>T, R347P/G551D, A455E/F508, R560T/dF508, R553X/F508, 621 + 1G>T/F508, 621 + 1G>T/711 + 1G>T, R117H/F508, and I506V/F508) and DNA from 4 homozygous patients (F508 and 2789 + 5G>A, 3849 + 10kbC>T, and G542X) was used in validation experiments.
All samples were correctly classified and matched with their appropriate interpretive comments. Fig. 3
shows a photograph of the 96-well plate with embedded chips and chip imaging results for all of the ACMG-recommended mutations and polymorphisms. Expected visual spot patterns, as is shown in Fig. 3C
, were observed with the exception of a compound heterozygote (I506V/F508) and a homozygous mutation F508/F508 in exon 10. In both cases the expected signal was missing for one or more wild-type probe spots. Compound mutations within the same codon or on nearby codons can affect the performance of other probes detecting nearby sequence regions. For example, wild-type probes for I506V and I507V share common sequences with the F508 probe, and I507 and F508C have identical wild-type sequence (see Table 2 in the online Data Supplement); this is not surprising because they detect mutations on three contiguous codons. As a consequence, when the F508/F508 homozygous mutant sample was tested the F508C, I507, I506V, and I507V wild-type probes all lost activity, and in the case of the I506V/F508 compound heterozygote, the I507V wild-type probe lost activity. The effect, however, was reproducible; therefore, the software could be taught to recognize this pattern.
|
To get a quantitative assessment of probe performance, we evaluated the corresponding numerical data output for the entire set of mutant alleles (Fig. 4
). To determine variability in probe performance we used only the signal intensity of each reacted probe spot to calculate the ratio. The gradient-based method described in the Materials and Methods creates a binary analysis of the data that was not appropriate for measuring variability in the performance of the probe set. Homozygous wild-type and heterozygous samples tested in this series of plates were clearly differentiated, showing excellent probe specificity. Hybridization of PCR-amplified homozygous samples produced unambiguous and highly specific signals; the signal-to-noise ratios, as defined by the ratio of wild-type probe signal divided by mutant probe signal for each allele, ranged from 10 to 600 with a mean value of 108. This performance correlated well with calculated discrimination values; the mean calculated Tm for the probe set was 8 °C, which corresponded to a calculated affinity difference of 100-fold.
|
To validate the performance of the CF Portrait DNA Chip, we first performed a comparison with the CF Gold line probe assay. For this comparison, 1092 samples, including 3 fixed positive controls (F508/F508, 5T/WT, and 3659delC/F508) and 1 rotating control of known genotype on each 96-well plate, were analyzed after their amplification and compared with the previously reported genotypes generated with the CF Gold strips. Successful genotyping was obtained for 997 of the 1092 samples (91%) with 88% automated allele classifications and 11% manual classifications. The failure rate of 9% was higher than that observed with the Roche strips (5%) and was most likely a product of DNA quality and not assay sensitivity. As stated in the Materials and Methods, the DNA samples were not fresh, and the chips were observed to have gross failure consistent with PCR failure. There were no discrepancies for genotype classification assignments for either the CF mutations or for the polymorphisms in the Roche comparison.
In this series, there were 17 F508 heterozygous patient samples, 1 F508 homozygous sample, 2 R117H heterozygous samples, and 1 heterozygous patient sample each for I148T, G542X, R553X, R347P, and 2789 + 5G>A, for a total of 26 mutant alleles. Additional mutant alleles detected in the control samples included three fixed control samples (F508 homozygous, 5T/WT, 3659delC/F508) on every plate and two heterozygous samples (R560T and 1078delT) and one heterozygous sample each for R334W, A455E, R347P, R117H, I507, I507V, G551D, and 1717-1G>A as rotating controls.
In a second comparison with the Applera CF OLA, Ver. 3.0, we analyzed 1076 samples, including 3 fixed positive controls (F508/F508, 5T/WT, and N1303K/WT) and 1 rotating control of known genotype. In all, 53 samples failed initial analysis in the Applera series and required repeat analysis compared with 54 samples in the CF Portrait DNA Chip series, yielding an acceptable repeat rate of 5% for both platforms. Automated classification calls were made for 972 of 1022 results (94%). In this comparison, there were 19 F508 heterozygous patient samples, 3 I148T heterozygous samples, 3 R117H heterozygous and 1 R117H homozygous samples, 2 W1282X heterozygous samples, and 1 heterozygous patient sample each for G551D, R553X, R1162X, and 3849 + 10kBC>T, for a total of 36 mutant alleles. Additional mutant alleles detected for this study included fixed controls F508 homozygous, 5T/WT, and a N1303K heterozygous sample on all plates, and one heterozygous sample each for R560T, G542X, R553X, W1282X, 2184delA, G85E, I148T, 621 + 1G>T, R334W, R117H, 1078delT, and 1717-1G>A as rotating controls.
There were no discrepant results for CF mutations; however, there were seven discrepancies for the IVS-8 polymorphism. In all cases, DNA sequencing analysis was performed to determine the correct genotype, and in all cases the CF Portrait DNA chip had assigned the correct genotype with the errors being assigned to the Applera OLA assay. Five patients were genotyped as 7T/7T by the Applera assay, whereas the actual genotype, as determined by the CF Portrait DNA chip and sequencing, was 7T/9T. A single patient was genotyped as 7T/9T by the Applera assay, whereas the actual genotype was 7T/7T, and another patient was genotyped as 5T/5T by the Applera assay, with the actual genotype being 5T/7T.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In this report we have described a novel high-throughput platform for highly multiplexed SNP genotyping using the CF Portrait chip. The described detection platform relies on appropriate configuration of optical surface coatings on a silicon wafer such that binding of a target sequence of interest to its complementary surface immobilized capture probe transduces a detectable color change on the surface. Layers were selected based on chemical and optical criteria such that thickness increases on the order of 20Å were detected, creating a sensitive detection surface (11). This silicon-based chip is similar to those previously used to detect protein, nucleic acid, and carbohydrate targets in clinical samples associated with viral or bacterial pathogens (12)(13)(20)(21)(22)(23). The creation of a high-throughput platform was accomplished by use of off-the-shelf semiconductor processes to embed chips in each well of a 96-well plate. A slight modification of an off-the-shelf image acquisition system combined with custom image software enabled rapid analysis of chip images and classification of results. The resulting platform allowed a single technologist to process 1300 PCR-amplified samples, or roughly 40 000 genotypes, in a 10-h shift.
The described platform has several attractive features for high-throughput applications: The CF Portrait chip is highly accurate. In a rigorous comparison with two other commercial platforms in use, CF Portrait demonstrated 100% concordance for mutation detection with both the Roche strip and with the Applera CF OLA system. Seven discordant results that the Applera CF OLA gave for the IVS-8 5/7/9T polymorphism were sequenced and correctly identified by the CF Portrait chip. This high degree of accuracy can be attributed to low nonspecific binding to the chip surface, which allows for a high signal-to-noise ratio with optimized surface probes. The materials used to create the optical layers are chemically inert, and the surface has limited surface roughness with <1 nm in thickness variance as measured by atomic force microscopy (23). Application of target nucleic acids and proteins to the surface in the presence of clinical samples, including whole blood, serum, sputum, and feces, have shown no adverse effects on assay performance (12)(13)(20)(21)(22).
The CF Portrait chip was manufactured with existing semiconductor processes that are currently used to make inexpensive integrated circuit chips on the scale of >50 million/year. As configured here, these processes can produce 1 million chips/year, limited by the spotting printer, and can be scaled by adding additional instruments. These factors, combined with the low material cost of silicon wafers and oligonucleotides, allowed us to make arrays on a scale necessary to supply the high-throughput testing needs of the clinical diagnostics laboratory in a cost-effective process.
A fundamental issue with thin-film generation on optical surfaces is that the color change transitions from darker than the background chip color to lighter as more thickness is deposited; the increased thickness is a function of the amount of analyte present (11). Classic intensity analysis is insufficient to detect these "white spots". To solve this problem, we modified a method used to identify craters from satellite images of extraterrestrial moons (24). Craters have many similarities to spots on our surface, such as being more or less rounded, variable in size, and both lighter or darker than the surrounding background. Analysis of the intensity gradient near the edge of a spot allowed us to distinguish between authentic reactions and pseudo-spots formed by nonspecific binding of material surrounding an unreacted zone. An added benefit of using this approach was that we also were able to detect other types of anomalous spots. "Donuts" or "blowouts", the appearance of a ring of signal on the edge of a spot only, are frequently observed problems in spot detection. We have observed that highly reacted probe spots can have a thick film that is sufficiently unstable that it flakes away in the center, possibly because of mechanical instability. We have been able to detect donuts by this analysis, thereby correctly measuring strong reactions that were missed by simple intensity analysis.
As a result of the Human Genome Project, molecular testing is playing an increasing role in medical practice. An impediment to more widespread use of molecular testing is the complexity and high cost of performing molecular assays, especially when more than one mutation must be assayed for each patient. As data from the Human Genome Project are mined further, testing for new panels of multiple SNPs and mutations will become indicated. We have created a highly accurate, relatively inexpensive, automated assay system that is easily adaptable for other multiple mutation assays that do not require sophisticated equipment or expertise.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
2 Current address: Sirna Therapeutics, 2945 Wilderness Place, Boulder, CO.
3 Nonstandard abbreviations: CF, cystic fibrosis; CFTR, cystic fibrosis transmembrane conductance regulator; ACMG, American College of Medical Genetics; ACOG, American College of Obstetrics and Gynecology; LIS, laboratory information system; SNP, single-nucleotide polymorphism; HRP, horseradish peroxidase; TSPS, T-structure polydimethyl siloxane; ASO, allele-specific oligonucleotide; TRW, Tecan robotic workstation; and Tm, melting temperature.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  X. Lin, J. A. Flint, M. Azaro, T. Coradetti, W. M. Kopacka, D. L. Streck, Z. Wang, J. Dermody, and W. Mandecki Microtransponder-Based Multiplex Assay for Genotyping Cystic Fibrosis Clin. Chem., July 1, 2007; 53(7): 1372 - 1376. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. M. Strom, R. Janeszco, F. Quan, S.-b. Wang, A. Buller, M. McGinniss, and W. Sun Technical Validation of a Tm Biosciences Luminex-Based Multiplex Assay for Detecting the American College of Medical Genetics Recommended Cystic Fibrosis Mutation Panel J. Mol. Diagn., July 1, 2006; 8(3): 371 - 375. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
, heterozygous samples.