Clinical Genotyping Technical Guide

Table of Contents

Letter from the Editor
Index of Experts
Q1: What guidelines for sample collection help to ensure high-accuracy genotyping?
Q2: What levels of evidence are needed to consider a DNA variant clinically relevant?
Q3: How do you determine which technology to use to find particular variants?
Q4: What validation techniques do you employ to ensure assay reproducibility?
Q5: What criteria determine the threshold to call a variant for clinical purposes?
Q6: How do you implement quality controls to check your data for errors?
List of Resources

Download the PDF version here

Letter from the Editor

While SNPs, CNVs, and chromosomal rearrangements are just a few of the genetic variations that are commonly being studied today, newer types of analyses are attempting to correlate these changes to disease. One type of change that has seen a lot of attention has been the single nucleotide polymorphism, but both general and clinical research have come a long way since the creation of the HapMap and the ensuing wave of genome-wide association studies.

As genotyping has moved beyond simply identifying and cataloging genetic variants to making these alterations relevant to disease, clinical genotyping has become a focus for an increasing number of labs. Clinical genotyping presents numerous challenges, many of them related to operating and maintaining labs that meet CLIA guidelines. While some labs are already there, others are still trying to upgrade in order to confront the recurring issues of ensuring careful DNA extraction, avoiding contamination, and maintaining good sample handling and storage techniques. This is especially relevant as genotyping moves to the clinic, where complete, accurate results can have a significant impact on a patient's life.

Whether genotyping samples against known genetic variants, discovering new variants, or genotyping for pathogen resistance, there's a whole world out there when it comes to clinical genotyping. In this guide, we bring you advice from leaders in the field whose expertise runs the gamut, but which is primarily focused on clinical genotyping to both predict and confirm disease susceptibility. Experts offer their thoughts on the most effective ways to handle specimens, the best technologies to use for genotyping assays, and the most reliable methods to ensure producibility, specificity, and sensitivity in each type of assay. So sit back, relax (that is, if you’re already CLIA certified), and enjoy the tips.

— Jeanene Swanson

Index of Experts

Doug Bintzler
DNA Core Facility
University of Cincinnati

Betsy Bove
Clinical Molecular Genetics Laboratory
Fox Chase Cancer Center

Christopher Plowe
Center for Vaccine Development
University of Maryland School of Medicine

Katia Sol-Church
Biomolecular Core Lab
AI duPont Hospital for Children

Robert Welch
Core Genotyping Facility
National Cancer Institute

Q1: What guidelines for sample collection help to ensure high-accuracy genotyping?

Many of the samples we test in our facility come from human DNA. Therefore, any method of collection must ensure safety as well as prevent possible contamination of the sample during collection and isolation of the DNA. We have set up GLP protocols for sample collection and isolation that require the use of laboratory safety equipment. We maintain similar protocols during DNA isolation and testing. Proper labeling is also required during collection so that each sample can easily be identified and stored. Using proper scientific techniques and GLP protocols has worked to ensure the accuracy of our genotyping.

— Doug Bintzler

We sample peripheral blood for germline testing, or paraffin-embedded tissue for somatic tumor testing. Quality control is dictated and regulated by clinical laboratory standards to ensure sample integrity for sensitive and specific molecular testing. Our laboratory's standard operating procedures manual details instruction for type of collection container, for proper specimen labeling and handling, and for preparing fresh cut slides from paraffin-embedded block when requested. There are written criteria for rejection of unacceptable or possibly commingled specimens. A system is documented to positively identify all patient specimens, specimen types, and aliquots through all phases of the analysis, including specimen receipt, nucleic acid extraction, nucleic acid quantification, hybridization, detection, documentation, and storage. Samples are stored properly to minimize degradation of nucleic acids.

— Betsy Bove

We do much of our malaria parasite genotyping on filter paper blood samples, which are air-dried and placed in individual resealable plastic bags, each with a desiccant. We typically write dates and unique identifiers right on the sample at the time of collection to minimize the likelihood of mislabeling.

— Christopher Plowe

Most of our clinical samples are collected off site and shipped to our lab for processing and genotyping. Though in many of the off site clinical settings, where it can be difficult to implement standard/strict guidelines for sample collection, we request that all specimens be identified, collection tubes clearly labeled, and that the shipment include consent forms from the patients and/or legal guardians. Any ambiguity in the sample ID is usually resolved by a re-sampling of the patient. To avoid possibility of contamination when collecting pediatric buccal cell specimen, we provide individually wrapped collection kits that include tubes with cell lysis buffer, cytobrush, and gloves in addition to a complete set of instructions. We discourage the transfer of specimen to a second container to avoid cross-contamination between samples. Ideally, we request that live cells and blood be shipped to us at room temperature within 48 hours of collection to reduce tissue degradation and loss of viability. For in-house patient procedures, the lab staff is routinely on standby to collect the specimens directly from the clinic or from the OR. Biopsies that are used to establish a patient's fibroblast culture are collected in DMEM or RPMI media for transport to the cell culture facility. All other tissues are snap-frozen in liquid nitrogen, and stored at -70°C until further processing, or formalin-fixed. We have validated procedures in place in the lab for extracting DNA from whole blood, buccal cells, formalin fixed, paraffin-embedded samples, frozen tissues, fresh biopsies, and from primary cells lines established from a patient's tissues. DNA quality is assessed on a Nanodrop by the spectrophotometer ratio 260/280 (typically we considered ~1.8 a good ratio). To maintain a single direction workflow, we have set up distinct pre-PCR and post-PCR staging areas and have dedicated equipment to avoid contamination and ensure high quality genotyping.

All genotyping data workflow is performed in compliance with Good Laboratory Practices following CLIA guidelines to ensure data quality and sample tracking. We keep current a database of all individuals enrolled in the clinical study (probands and available parents) to help keep track of the different molecular tests and DNA variants identified.

— Katia Sol-Church

For our lab we cannot influence the collection or the extraction but recommend that methods are used to produce high-quality/high molecular weight genomic DNA. However, since we can’t influence the collection, one of the most important parts of our genotyping process is sample handling and qualification. Samples are accurately measured for volume and concentration using methods that minimize dilution and are reproducible. Once accurate estimates of DNA mass and concentration are determined, samples are genetically "fingerprinted" using a panel of highly polymorphic microsatellite loci, including the amelogenin locus. This data is used to QC the samples as

1. it indicates the ability to be amplified and genotyped, and
2. the data produced can be used to determine identity (if the sample is a duplicate of another sample, intentional or unintentional), confirm gender (as compared to the reported gender), indicate contamination, and as a reference if we are ever to receive DNA from this sample in the future.

— Bob Welch

Q2: What levels of evidence are needed to consider a DNA variant clinically relevant?

Our laboratory is service-oriented. Therefore, individual users and their research usually determine the relevance of a clinical variant. The identification of a single variant may be relevant for many of the genetically based diseases that are studied. We may test hundreds of samples for our users in search of a single variant.

— Doug Bintzler

To consider a DNA variant clinically relevant, we rely on the consensus of the most recently published scientific literature. Relevance is traditionally determined by all or a combination of the following: functional analysis, co-migration of the variant with affected members of multiple validated family pedigrees, percentage of affected individuals with the variant in the testing population, conservation of sequence at the position of the variant among species, and penetrance among those individuals carrying the variant.

— Betsy Bove

For a new variant, we will go back to the original sample and repeat everything from DNA extraction through PCR and genotyping or sequencing. For new variants detected by allele-specific PCR or restriction digestion, we will confirm by sequencing. When an apparent new haplotype is detected based on multiple polymorphisms, we have to make sure it is a real haplotype and not an artifact of polyclonal malaria infection, so we will clone the target sequence and resequence it to confirm a unique and novel haplotype.

— Christopher Plowe

Working in a children's hospital, we are well aware of the impact that reporting of variants of unknown significance has on the patient's family and clinical team. So we make sure that there can be no doubt that the variant identified in our lab is indeed clinically relevant. The level of evidence required greatly depends on the disorder under study. DNA variants associated with disorders caused by de novo "gain of function" germline mutations are easily identifiable. The first step in the variant validation process is to verify that it is not present in the child's unaffected parents. When parents are not available for testing we try to determine if the DNA variant (single point mutation, deletion, rearrangement) will result in the synthesis of a mutated protein. Most of our patients carry mutations in important domains of the protein and some of the variants are already known to result in gain of function. New variants should be evaluated biologically to determine their effect on the protein's cellular function.

Some DNA variants are identified that cannot be tested functionally while they may be a marker for disease predisposition or severity or indeed be a simple polymorphism with no clinical relevance to the disorder under study. We routinely screen a database for SNPs that are associated or co-segregate with certain clinical phenotypes, and look at the frequency of the variant in case/control epidemiological studies. There are limitations in addressing the general problem of causality of sequence variants.

— Katia Sol-Church

We are looking for markers at this stage. Assuming those markers, however, are marking some clinically relevant variant, then I would suggest that convincing evidence would include replication of that variant in more than one dataset with adequate power and demonstration of the biological significance of why the variant is clinically significant.

— Bob Welch

Q3: How do you determine which technology to use to find particular variants?

Because most of the variants determined in this laboratory are unknown, our laboratory is somewhat limited in technology for detecting DNA variants. Our method of choice is automated sequencing. However, samples related to environmental studies generally involve AFLP and RFLP.

— Doug Bintzler

The type, any known ramification, and sequence-specific characteristics of the variant will determine the technology used for detection. High-volume testing of DNA that would traditionally be directly sequenced we have validated using enzyme mutation detection (EMD). This assay is more cost effective and less labor intensive. Low-volume tests are directly sequenced either for general scanning or site-specific alteration detection. Alterations that exhibit DNA conformation change and are routinely tested can be detected using the gel-based heteroduplex mobility assay (HMA). Known fragment length variation of specific DNA alterations are detected by electropherogram readout using capillary electrophoresis, technology traditionally used for direct sequencing. Other gene alterations may be detected indirectly. One example of this is the detection of microsatellite instability (MSI) in non-coding regions of the genome, consistent with inadequate DNA repair.

— Betsy Bove

We have to consider several variables, including the circumstances under which the technology will be used. For many of our assays we want our colleagues in resource-limited developing country labs to be able to perform them, so we may choose a more robust and inexpensive but less sensitive method such as allele-specific restriction digestion. It also depends on the target sequence — if there are too many single nucleotide polymorphisms in close proximity, for example, pyrosequencing assays will be hard to develop and we may rely on standard sequencing. Another consideration is whether we are interested only in looking for known variants, in which case a simple allele-specific assay is adequate, or if we want to be able to detect unknown sequence variations, in which case we must use standard DNA sequencing.

— Christopher Plowe

The technology choice depends not only on the type of variants to be genotyped and quality of the DNA, but also on the size of the test population. In a previous association study on familial osteoporosis, we were able to take advantage of the high throughput capabilities of the Illumina platform to genotype a large number of SNPs to identify nearby genetically linked susceptibility genes; however, the current study has a much smaller number of samples available for genotyping. My research lab is set up in a small pediatric hospital and many of the current disorders under study are quite rare. While pediatric specimens are referred to my lab from many sites around the US and the world, the size of our cohort precludes the use of high-throughput technology in our genotyping efforts. Thus for the present study, direct DNA sequencing utilizing the ABI 3130xl Genetic Analyzer platform is the most practical approach to genotyping. Mutations most frequently occur in a certain region of the genes, thus PCR is performed to amplify a specific region, which is then subjected to sequencing. Alternative techniques to direct sequencing are often dictated by the quality of the clinical sample that is available for testing. When dealing with DNA samples of poor quality/degraded (as is often the case for DNA isolated from archived formalin-fixed, paraffin-embedded tissues) direct sequencing sometimes fails to provide clear genotyping data. We have a better chance in getting reliable data from degraded samples with pyrosequencing technology using the Biotage PSQ96 platform. Pyrosequencing allows for the analysis of short amplicons, which are more readily obtained from sub-optimal DNA than larger amplicons. For the same reason, another technique we have used in our lab to successfully genotype samples of varying quality is the Taqman-based allelic discrimination assay that we run for mid-scale genotyping projects. The format used is a 384-well format of the real-time PCR 7900 platform from Applied Biosystems. We buy off-the-shelf, ready-made assays from Applied Biosystems as these usually perform well and require minimal optimization if any.

In our hands, pyrosequencing has also shown to be more sensitive than sequencing or TaqMan at detecting rare alleles. However, in one instance of a DNA variant associated with mosaicism, we observed increased sensitivity in the detection of rare alleles using the "low-tech" agarose gel-based restriction fragment length polymorphism (RFLP).

Microsatellite (aka short tandem repeat/STR) genotyping is also a method widely used in my lab. First of all, STR-PCR is a great tool to determine the quality of the DNA. We amplify in multiplex chromosome-specific amplicons ranging from 90 to 500 bp in length. While high-quality DNA will amplify all loci equally well, degraded or sub-optimal DNA will perform poorly (if at all) with the larger amplicons.

We routinely use STR-PCR genotyping to identify loss of heterozygosity (LOH) in cancer samples and define the flow of alleles between the proband and parental samples. Automated STR genotyping is routinely performed on the ABI 310 or ABI 3130xl genetic analyzer using chromosomespecific microsatellite markers in multiplex PCR setup. We also use the highly polymorphic STR marker sets to rule out sample mix-up for clinical labs and verify sample purity.

Gene duplication is also a type of DNA variant that is used in clinical research to genotype pediatric samples. In our molecular diagnostic lab, one such clinical research test has been developed to identify probands with duplicated genes using the semi-quantitative capabilities of capillary electrophoresis on the ABI 310, and is currently offered as a diagnostic test to clinicians.

— Katia Sol-Church

The technology or genotype platform depends on three major factors. 1) What are the needs of the study design (how many variants are needed) at the time of investigation? 2) How much and what quality of biospecimen/DNA is available for genotyping? 3) What is the ultimate cost/timeframe for completion of the project? Weighing these three factors in concert determines our choice for the most appropriate genotyping technology.

— Bob Welch

Q4: What validation techniques do you employ to ensure assay reproducibility?

Our validation methods are similar to the steps we use to determine quality control. Because we always employ a commercial standard and an in-house standard, we can run comparisons from each test to determine if the results are reproducible. Generally, variations in the standards may indicate changes in whatever instrument is used in the test. In the development or improvement of any technique, the in-house standards provide an effective means for validation.

— Doug Bintzler

A newly developed clinical assay will be validated before being put into clinical practice. To ensure reproducibility of the assay, samples previously tested using a formerly validated assay will be tested using the new assay. Alternatively, blinded samples from another laboratory performing the same assay or assay type will be tested to confirm sensitivity and specificity of previous testing results. Blinded sample exchange with the reference laboratory is clear validation of the sensitivity and specificity.

— Betsy Bove

We use negative and positive controls at all steps as a matter of routine. Our lab serves as a reference lab for collaborating labs, so when they have a novel or unexpected result, they will send us blinded samples and we repeat the genotyping with no knowledge of their results. Discrepant results are repeated and confirmed by sequencing, preferably from the original clinical sample rather than extracted DNA.

— Christopher Plowe

We spend much time on assay development. Optimized conditions are routinely worked out on a Robocycler (Stratagene) using a gradient block feature, which allows testing of a wide range of annealing temperatures under different magnesium concentrations or buffer conditions. We use a series of control DNA isolated from unaffected individuals of various ethnic backgrounds to validate the assays, and use SOPs to minimize technical variability. Each clinical specimen is tested alongside a water control to ensure the integrity of PCR reagents and technique. Periodically running a positive control validates robustness of the assay and control for reagent performance. Typically we test DNA isolated from two different tissue sources from the same individual to ensure reproducibility of the data, in combination with independent amplification of the region of interest using two sets of non-overlapping primers. For cells derived from patient biopsies, two independent cell lines are established as biological replicates. Likewise, for direct DNA extraction from a patient biopsy, two independent extractions are performed. Some of the allelic discrimination assays are set up in duplicate or triplicate to ensure reproducibility. Discordance between duplicates results in re-testing.

— Katia Sol-Church

Here we are really looking for genotype accuracy, or, is the assay producing the "correct" genotype? Genotype assay reproducibility is really sample set dependent so every time we genotype different samples we look for reproducibility (see question 6.) However, for new assays we always want to examine if they are performing accurately first. So all newly developed assays are genotyped on a control set of individuals (usually CEPH or HapMap controls); these individuals have been genotyped for the specific variants previously either by our facility (by sequencing the region or by previous genotyping assays on a different platform) or by others as in the HapMap project. For example, if we develop an assay for a new variant and the international HapMap project has genotyped this variant on the HapMap samples, as our first order of business when we determine the assay is optimized we will also genotype the HapMap samples and ensure that the genotypes that we are getting match those that were obtained by the HapMap. There are sometimes significant differences, and in these cases we may still choose to genotype the SNP but will note that the data may be suspect and rely on QC measures to ensure genotype reproducibility.

— Bob Welch

Q5: What criteria determine the threshold to call a variant for clinical purposes?

The threshold values of a particular variant depend on the technology employed. In the detection of an unknown variant related to an identified genetic related disease, 25 percent may be considered an appropriate threshold as long as the variant is confirmed on both strands. The environmental studies we have participated in may have a threshold value as low as 1 percent, in which case any variant is often considered relevant.

— Doug Bintzler

This is a tough one, since different techniques have different thresholds of detection, and most of the samples we examine are from polyclonal malaria infections, which also contain human as well as parasite DNA. There is no reliable way that I know of to determine with confidence whether a faintly positive result represents a minority variant in a mixed infection or a false positive result. The best way to approach this is to reproduce mixed infections by preparing serial dilutions with known ratios of variants, and then determining thresholds for true positive calls within a reasonable margin of error.

— Christopher Plowe

Sequencing data generated on the ABI Prism 3130xl instrument is analyzed using KB base caller, a software providing quality value (QV) for each base to help flag low-quality regions. We usually consider calls acceptable when QV>20 (probability of error is less than 1 percent). We set the threshold for the detection of mixed base at 25 percent and choose the IUB code option for mixed base assignment. While this setting is appropriate for detection of mixed base for most patients carrying heterozygous germline mutation, it does not allow automatic detection of rare alleles in cases of mixed base resulting from mosaicism levels of less than 20 percent. Thus, care is always taken to verify the data manually via visual inspection of the electropherogram. Other genotyping technologies such as pyrosequencing are more apt at reproducibly detecting rare alleles down to a threshold of 5 percent. For this reason pyrosequencing has been the method of choice in my lab for detection of rare alleles associated with mosaic patients.

When using Allelic Discrimination TaqMan assay for SNP genotyping, we commonly use the SDS software auto caller feature (confidence interval at 99 percent) and automatic threshold. As for other techniques careful visual review of the data (AD plot) may warrant changing the threshold manually for a more accurate allele call.

— Katia Sol-Church

All genotype calling is done in an automated fashion using platform-dependent software. For the most part, pilot data (raw data with known genotypes and wide variety of sample qualities) is run through the analytical software package until metrics are fine-tuned so that genotypes are called as completely as possible and accurately. SOPs are developed to train analysts to use a series of these metrics to filter out bad data and report data that is deemed good. At the end of the genotype calling, QC reviews look at sample and assay completion rates, sample and assay concordance rates (using laboratory blinded duplicates), assay fit to HWE on controls, and assay performance when the assay was validated.

— Bob Welch

Q6: How do you implement quality controls to check your data for errors?

We employ several standards to each test to measure both the validity of the test as well as the reproducibility. The standards provided with each test kit and a negative control generally provide a good test of quality. However, we have also developed several in-house standards. Each in-house standard was prepared using collection methods and protocols similar to the samples that are tested and is affected by the same problems that may also affect the test samples. By comparing the results of the in-house standards we can monitor the quality of the test and measure any deviations. Each researcher may elect to include a standard with their samples as a second method for measuring the quality.

— Doug Bintzler

Quality control for genotype data is multi-layered. Standard controls are included in each assay, clear outcomes stipulated in validated protocols. Where practical and appropriate, positive/negative and internal controls, water and sensitivity controls, and run-to-run controls are incorporated into each assay experiment. This panel assures reproducibility, specificity, sensitivity, and absence of contaminant. Sequencing data is produced for both the forward and reverse directions, and evaluated against valid genomic databases using standardized computer software. Sequencing data printouts are evaluated visually as well to search for alterations occurring in not less than 10 percent of extracted DNA material. Clinical data is reviewed by assay technologist and laboratory manager.

— Betsy Bove

This depends on the reasons for genotyping. Much of what we do is surveillance for drug-resistant malaria, where we are measuring the prevalence of resistance markers. For this kind of genotyping it is less important to have a high degree of confidence in each individual result, so for such routine work we rely mainly on positive and negative controls. If a clinical decision were going to be based on an individual result, we would have to institute more stringent quality control measures such as those we use when confirming an unexpected or novel result.

— Christopher Plowe

In most direct sequencing results, signal strength greater than 40 indicates a good quality reaction and weaker signal may indicate data quality problems. While calls are made automatically and quite accurately on any given sequencing platform, there are instances when errors occur during base calling due to chemistry or instrument run artifacts. Most of us have observed base-calling errors such as missed mixed base call and missing bases resulting from compression effects. These errors can be easily missed if one only looks at the text output of the genotyping results. So to ensure highest quality control of the data we routinely have at least two experienced scientists visually inspect each electropherogram. The criteria for visual QC include verification of the overall signal strength, peak shape, and height. Each questionable call is inspected for concordance on the other strand. In case of base ambiguity/discordance between the forward and reverse reaction, the sample is subjected to resequencing. We routinely spend countless hours reviewing and scanning sequence data from our clinical samples to ensure scientific integrity and clinical relevance.

For pyrosequencing QC we make use of the built-in QC control to validate base calls. Inspection of the sample pyrograms verifies that the sequence context surrounding the genetic variation is correct. We must see perfect concordance in the genotyping call between biological replicates and the technical duplicates.

In the case of TaqMan Allelic Discrimination (AD), although it can be run as an end point assay, we routinely examine the amplification plot and review baseline and threshold for each sample as part of our QC. For some of the "precious" and depleted DNA samples of sub-optimal quality, AD calls need to be made manually. In our hands, we found that the multi-component data plot (which displays the contribution of each dye over the duration of the real-time run) can help us validate some of the trickier genotyping calls.

— Katia Sol-Church

Two major ways:

1. For each assay whether panel (multiplex or single) samples with known genotype are always run within a batch (i.e. one sample of a 96 well plate). From the validation of the assay we have control samples where we know the genotype. After genotype data is called, comparison of the genotype calls are made to the previously generated data to ensure that the calls are accurate.
2. For each dataset a number of samples, usually 10%, are duplicated in a random fashion, and blinded to the laboratory. After data analysis the genotypes for these pairs are checked to be concordant, ensuring reproducibility. In certain assay types certain control samples that form a trio (mother - father - child) are also run to determine if any Mendelian transmission errors are observed. Finally controls (non-affected samples) are usually checked for deviation to Hardy Weinberg Equilibrium (HWE).

— Bob Welch

List of Resources

For further reading, check out the following publications and websites.


Publications

Hunter DJ, et al. (2007) A genome-wide association study identifies alleles in FGFR2 associated with risk of sporadic postmenopausal breast cancer. Nature Genetics 39(7):870-4.

Easton DF, et al. (2007) Genome-wide association study identifies novel breast cancer susceptibility loci. Nature 447(7148):1087-93.

Yeager M, et al. (2007) Genome-wide association study of prostate cancer identifies a second locus at 8q24. Nature Genetics 39(5):645-9.

Couch FJ, et al. (2007) AURKA F31I polymorphism and breast cancer risk in BRCA1 and BRCA2 mutation carriers: a consortium of investigators of modifiers of BRCA1/2 study. Cancer Epidemiol Biomarkers Prev. 16(7):1416-21.

Tchou J, et al. (2007) Large genomic rearrangement in BRCA1 and BRCA2 and clinical characteristics of men with breast cancer in the United States. Clin Breast Cancer 7(8):627-33.

Websites

American Association of Blood Banks
http://www.aabb.org

Association of Biomolecular Resource Facilities
http://www.abrf.org/index.cfm/group.show/ FragmentAnalysis.40.htm

Marshfield Clinic
http://www.marshfieldclinic.org/research/ pages/index.aspx

NIST Biotechnology Division
http://www.cstl.nist.gov/div831/