Introduction
Principles
Materials
Methods
DNA amplification
Polyacrylamide gel electrophoresis and silver staining
Sample storage
Image processing and data analysis
Comments
Optimization
Reaction components
Thermal cycling parameters
Primer design
Data analysis
Troubleshooting
Reproducibility
Controlling contamination
Miscellaneous
Introduction
The generation of characteristic signatures from virtually any nucleic acid, even of anonymous nature, has been made possible by the invention of nucleic acid scanning (Livak et al. 1992, Bassam et al. 1995). These signatures are composed of arbitrary collections of amplification products that result from the targeting of a multiplicity of anonymous sites (amplicons) in template DNA or RNA molecules. The targeted sites are amplified with one or more oligodeoxynucleotides to produce arbitrary but entirely characteristic "fingerprint" patterns (Williams et al. 1990, Welsh and McClelland 1990, Caetano-Anollés et al. 1991). Nucleic acid scanning can be used in a wide range of applications including DNA and RNA fingerprinting, genome mapping, marker-assisted breeding and molecular evolution, systematics, and population biology (reviewed in Caetano-Anollés 1993, 1994, 1996, Rafalski and Tingey 1993). Three techniques were originally described: randomly amplified polymorphic DNA (RAPD) (Williams et al. 1990), arbitrarily primed PCR (AP-PCR) (Welsh and McClelland 1990), and DNA amplification fingerprinting (DAF) (Caetano-Anollés et al. 1991). Collectively termed multiple arbitrary amplicon profiling (MAAP), these techniques reveal polymorphic DNA or RNA without prior knowledge of sequence or cloned and characterized probes. Nucleic acid scanning produces amplification profiles of varying complexity, sometimes very simple (RAPD) but sometimes quite complex and comparable to multilocus DNA fingerprinting (DAF). Several variations of the original strategies provide higher multiplex ratios and detection of polymorphic DNA. For example, AFLP analysis produces highly complex profiles by arbitrary amplification of restriction fragments ligated to adaptor cassettes with hemi-specific primers harboring adaptor-complementary 5' termini (Vos et al. 1995). Other strategies increase profile complexity and polymorphic DNA levels by prior digestion of template DNA with one or more restriction endonucleases (Caetano-Anollés et al. 1993), when using mini-hairpin primers (Caetano-Anollés and Gresshoff 1994), or when producing arbitrary signatures from amplification products (ASAP) (Caetano-Anollés and Gresshoff 1996). Finally, primers can be designed to be complementary to consensus elements representing interspersed repetitive sequences (IRS) (such as the human Alu sequences), satellite sequences (including mini- and microsatellites), transposable elements, telomere associated sequences (TAS), and coding sequences such as histones, rRNA and tRNA genes.
DNA amplification fingerprinting (DAF) is a general nucleic acid amplification technique that uses at least one primer of at least 5 nucleotides (nt) in length to produce characteristic and highly informative DNA patterns (Caetano-Anollés et al. 1991). These patterns are adequately resolved by polyacrylamide gel electrophoresis and silver staining, but can be produced using agarose gel electrophoresis, denaturing gradient gel electrophoresis (DGGE), temperature sweep gel electrophoresis (TSGE) or by automated analysis using DNA sequencers or capillary electrophoresis (CE). DAF can be distinguished from other genome scanning techniques by the high primer-to-template ratios, simplicity, excellent reproducibility and high multiplex ratios.
The DAF technique, in its different forms, can been used to fingerprint a wide variety of genomes including those of high (plants and animals) and low (fungi, bacteria, mitochondrial and plastid DNA) complexity, as well as subgenomic fragments like PCR amplified products, cloned DNA and cDNA populations. In this section we will describe the DAF technique in detail, highlighting the role of primer design and appropriate optimization of experimental variables. Return
Principles
The amplification of nucleic acids with arbitrary primers is mainly driven by the interaction between primer, template annealing sites and enzyme, and determined by complex kinetic and thermodynamic processes. The outcome is the generation of a population of polynucleotide products usually representing those genomic regions (amplicons) that have been predominantly amplified. A model to explain the amplification of DNA with arbitrary primers was proposed by Caetano-Anollés et al. (1992) [see also Caetano-Anollés 1993, 1994]. The model is based on the competitive effects of primer-template and template-template interactions established predominantly during the first few cycles of the amplification process (Caetano-Anollés et al. 1992; Caetano-Anollés and Gresshoff 1994). During the first few temperature cycles a group of cognate amplicons is selected for amplification during a "template screening" phase driven by primer-template-enzyme interactions that can accomodate primer-template mismatching events. First-round amplification products are initially single-stranded but have palindromic termini that can establish template-template interactions, forming hairpin-loops and duplexes. The primer will have to recognize and displace these structures to allow enzyme anchoring and primer extension. In subsequent rounds of amplification the different species of the reaction tend to establish and equilibrium, while the rare primer-template duplexes are enzymatically transformed into accumulating amplification products.
Polymorphisms in nucleic acid scanning result from changes in DNA sequence, initially within primer-defined sites in the genome. However, they can also arise from the deletion, insertion or inversion of a priming site or segments between priming sites, and from comformational changes in DNA that would alter the efficiency of amplification or priming. These DNA polymorphims can become useful markers both in general fingerprinting or mapping applications, directly or when converted into sequence-tagged sites (STS).
A typical DAF reaction involves the use of a single primer usually composed of 8 to 12 nt, a length that approaches the minimum configuration for DNA amplification (Caetano-Anollés et al. 1992, Vincent et al. 1994), amplification conditions that discriminate bonafide amplicons from those of artifactual origin, and a non-stringent reaction environment that ensures the reproducible targeting of multiple sites (Bassam et al. 1992). To guarantee optimal performance and reproducibility, a good understanding of primer design and a careful optimization exercise of amplification parameters is required. Please note that assumptions made for the PCR may not hold true when amplifying nucleic acids with arbitrary primers. Return
Materials
1. Reaction buffer: Stoffel (STF) buffer, 100 mM Tris-HCl, 100 mM KCl (pH 8.3). The following buffers stocks can also be used: TTNK10 (200 mM Tris.HCl, 1% Triton X-100, 40 mM ammonium sulfate, and 100 mM KCl), TTK10 (200 mM Tris.HCl, 1% Triton X-100, and 100 mM KCl), TTK30 (200 mM Tris.HCl, 1% Triton X-100, 40 mM ammonium sulfate, and 300 mM KCl), and TB (100 mM Tricine). All these buffers are adjusted to pH 8.6.
2. Deoxynucleoside triphosphate stock solution: 2 mM of each dNTP. If dNTP solutions are made from dry reagents, the solution should be adjusted to pH 7.5 with unbuffered 0.1 M Tris or 0.1 M NaOH, checking pH by spotting 1 µL aliquots on a strip of pH paper.
3. Magnesium solution: 25 mM or 100 mM magnesium chloride solution.
4. Oligodeoxinucleotide primer: 30 µM or 300 µM stock solution.
5. Thermostable DNA polymerase enzyme. Truncated versions of Thermus aquaticus DNA polymerase, such as AmpliTaq Stoffel fragment (Perkin-Elmer, Norwalk, CT), are preferred.
6. Template: 1-10 ng/µl stock solutions.
7. High quality deionized double distilled water (more than 10 megaohm cm), preferably HPLC grade.
Store all reagents at -20°C. However, keep the magnesium stock in use at 4°C to ensure reagent uniformity. Return
Methods
A DAF protocol usually involves two major steps: DNA amplification, and the separation and visualization of amplification products. DNA amplification usually takes 3-8 h of experimentation time depending on the thermocycler apparatus used. This time can decrease with availability of faster thermocycling units. DNA separation in the typical laboratory is usually done by polyacrylamide gel electrophoresis (PAGE) and individually resolved fragments are generally identified by silver staining. These two procedures may take about 1.5-2.5 h. Alternatively, real-time analysis of amplification products is possible but limited to proper instrumentation, such as access to DNA sequencers or capillary electrophoresis (CE) units.
DNA amplification
1. Assemble a typical amplification mixture in 20 µl total volume by adding components in the following order: 10.2 µl water, 2 µl deoxynucleoside triphosphates (stock with 2 mM of each dNTP), 1.2 µl magnesium chloride (25 or 100 mM stock), 2 µl STF reaction buffer, 0.6 µl DNA polymerase enzyme (10 units/µl AmpliTaq Stoffel fragment), 2 µl oligonucleotide primer (30-300 µM), and 2 µl template (usually diluted to 1-10 ng/µl). Shorter primers require higher oligonucleotide concentrations and low complexity genomes higher magnesium levels. TTNK10, TTK10, TTK30 and TB buffers can also be used successfully (Caetano-Anollés et al. 1994). When possible, prepare a master mix with reagents that are common to avoid pipetting errors and aliquot the mix into 0.2 ml or 0.5 ml microcentrifuge tubes. The total volume of the reaction can be decreased to 10 µl by generally doubling concentrations of reagents in stocks, or alternatively by using more accurate pipettors. This precautions are necessary to ensure reproducibility.
2. If necessary, cover the amplification mix with 1-2 drops of heavy mineral oil. With certain thermal cyclers the mineral oil layer may not be required.
3. Amplify the DNA in the temperature cycler for the desired number of cycles (usually 35). Several protocols can be used depending on primer, template and thermal cyler selected. Generally, use two-step cycles of 20 s at 96C and 20 s at 30C or 50C in an block-based thermocycler (Ericomp, San Diego, CA), and three-step cycles of 30 s at 96C, 30 s at 30C or 50C and 30 s at 72C in an oven thermocycler (Bios, New Haven, CT).
4. Retrieve amplification mixtures by adding 100-200 µl of chloroform to each tube and pipetting out the aqueous droplet, or directly by using a long pipette tip or by rolling the amplification mixture over Parafilm.
5. Dilute the samples 5-10 fold prior to electrophoresis to avoid gel overloading.
Polyacrylamide gel electrophoresis and silver staining
DNA amplification products are usually electrophoresed in 0.45 mm thick polyacrylamide slab gels (8x10 cm) backed on polyester film, and the DNA can be silver stained using the procedure of Bassam et al. (1991). A description of the separation and visualization of amplification products can be found in section 2. Polyester-backed gels can be preserved for many years by drying and can be stored in photographic albums. These gels avoid the costs of photography and allow easy densitometrical scanning of DNA profiles. They also act as repositories of DNA and as experimental records. The silver stained amplification products can be isolated from the dried gels (Weaver et al. 1994; see Separation).
Sample storage
Only a small aliquot (3-6%) of the amplification reaction is generally used for electrophoretic separation, leaving the rest of the sample for future reference or analysis. Samples can be stored at 4C and kept for years without visible degradation of amplification products. Evaporation can be prevented by maintaining the oil overlays.
Image processing and data analysis
Silver stained profiles can be scanned and the images stored in electronic files. For example, fingerprints can be scanned with an Apple Color One scanner (Apple Computer, Cupertino, CA) and the Ofoto program (version 2.02; Light Spource Computer Images Inc., Salinas, CA), and the images subsequently analyzed using the Think Pascal program "Image" for the Macintosh computer (Version 1.45; Wayne Rasband, NIH; Internet, wayne@helix.nih.gov) using a gel analysis macro. Scanned images can be analyzed with a variety of computer programs. For example, Dendron (Solltech Inc., Oakdale, IA) permits manipulation of lanes and gels for proper alignment with molecular weights, constructs composite images from different gels, finds and assigns polymorphisms, computes similarity coefficients, and generates dendrograms that permit for example the analysis of microbial isolates. Return
Comments
Optimization
The annealing of a single arbitrary primer to short and complementary inverted repeats that are in near proximity but scattered throughout the template and the successful strand-extension of the annealed oligonucleotides are the basis of the amplification reaction. Targeting of template sites occurs under a non-stringent reaction environment that is both adequate for annealing of short primers and specific enough to provide discrimination of legitimate and illegitimate amplicons. How can the amplification process be described in such a context? In general terms, DNA amplification can be characterized by three parameters: specificity, efficiency (i.e. yield) and fidelity. These parameters are strongly influenced by the different components of the reaction (such as primer, magnesium, and deoxynucleoside triphosphate concentrations) and are modulated by thermocycling conditions (such as annealing temperature). The judicious design of arbitrary primer and careful optimization of amplification components will ultimately result in reproducible and efficient amplification.
Based on these considerations, each investigator should set up its own amplification protocol by choosing appropriate reaction components, concentrations and thermocycling conditions. To do so one must determine the widest range of values for a particular reaction component or thermocycling condition within which amplification parameters exhibit little or no variation. This range of values defines a "reproducibility window" (Bassam and Bentley 1994) and provides a mesure of central tendency with which to define an experimental concentration, temperature or cycle number and avoid borderline conditions.
The optimization of the amplification reaction is a laborious process as a large group of interacting factors have profound effects on fingerprint, product number and efficiency of amplification. It relies on the sequential investigation of each variable and the design of large experiments. It should be noted that in reality optimum conditions are seldomly identified. Optimization has preceded several RAPD studies and rendered widely different optima (Akopyanz et al. 1992, Carlson et al. 1991, Devos and Gale 1992, Fekete et al. 1992, Munthali et al. 1992, Nadeau et al. 1992). This is in contrast to the few reports of optimization using DAF, where similar conditions were found optimal for the amplification of different organims such as centipedegrass (Weaver et al. 1995), bermudagrass (Caetano-Anollés et al. 1995), Petunia (Cerny et al. 1996), Chrysanthemum (Scott et al. 1996) and soybean (G. Caetano-Anollés, unpublished). This consistency only holds true if genomes of similar complexity are examined, and appears related to the tolerance of the DAF reaction to wide changes in amplification conditions (Caetano-Anollés 1994). Table 1 shows recommended conditions for amplification of genomes of low and high complexity. These conditions should serve as a start for a DAF optimization exercise.
| Reagents |
Plants & animals |
Fungi & bacteria |
||
|
Optimal |
Suggested |
Optimal |
Suggested |
|
|
Primer (µM) |
2-10 |
3-6 |
3-9 |
3-6 |
|
Template (ng/µl) |
0.01-2 |
0.1 |
0.1-1 |
0.1 |
|
Mg (mM) |
1-8 |
1.5 |
4-8 |
6 |
|
dNTPs (µM) |
50-300 |
200 |
50-300 |
200 |
|
Enzyme (U/µl)b |
0.2-0.8 |
0.3 |
0.2-2 |
0.3 |
a Buffer (10 mM Tris.HCl and 10 mM KCl; pH 8.3)
b AmpliTaq Stoffel DNA polymerase.
Optimization can be achieved in different ways. The influence of different reaction components on the DAF reaction was determined for bacterial (Bassam et al. 1992) and turfgrass DNA (Weaver et al. 1995) by using an iterative process of analysis. This optimization strategy is based on a simple matrix analysis in which several values for those experimental variables determined a priori to be most important are tested in combination with each of other variables. This approach simplifies the otherwise overwhelming full matrix analysis. Similarly, Wolff et al. (1993) used a fractional factorial design to study the significance of several reaction components in RAPD analysis of Chrysanthemum. Finally, the Taguchi method (Taguchi 1986) has also been applied to the study of interactions between specific reaction components in PCR and RAPD (Cobb and Clarkson 1994).
Because of simplicity and proved success in industrial process design, the Taguchi method offers a simple optimization alternative (Cobb and Clarkson 1994). A number of reaction components at three different concentrations spanning expected reproducibility windows are laid in an orthogonal array that defines only a limited number of amplification reactions. The measurement of an amplification parameter (such as product yield) is then used to estimate a quadratic loss function (termed signal-to-noise ratio, S). For each component the optimal conditions are those that give the largest S values. Polynomial regression generates S curves whose maximum represent reaction optima. Furthermore, a number of progressive trials can be used to better define those initially relevant factors. The results of Taguchi optimization rely on the assumption that optima lie within the range tested and do not originate from tightly defined or vey relaxed peaks. Results also depict the way an amplification parameter is quantitated. An optimization of bacterial DNA fingerprinting such as that described by Bassam et al. (1992) will therefore require Taguchi optimization for depictors of amplification efficiency (product yield and range of products amplified) and specificity (number of products, range of products amplified, departure from a consensus fingerprint, and reproducibility), and a series of optimization experiments that will progressively define the different optima.
Reaction components
Primer-template ratio: The most important variable in the amplification reaction is the ratio between concentrations of primer and template. Operationally, both primer concentration and length define the different nucleic acid scanning techniques (RAPD, AP-PCR and DAF). DAF uses over 10 times more primer than RAPD (at least 3 µM), and uses high primer:template mass ratios ranging 5-50,000.
Template: DAF can reproducibly amplify very low DNA template levels and tolerates template concentrations that span over a thousand-fold range (Table 2). DNA concentrations of 0.1 ng/µl produce consistent DAF fingerprints from most plant and animal genomes. However, too little template causes amplicon stoichiometric misrepresentation and lack of reproducibility. This is particularly important in the analysis of low complexity genomes (viruses, bacteria and most fungi) where a template concentration of at least 1 ng/µl should be used (Bassam et al. 1992, Caetano-Anollés and Gresshoff 1994b).
A reasonable effort to use relatively good quality DNA should also be invested. Avoid DNA isolation methods that produce severely degraded DNA or do not eliminate contaminants that can inhibit the activity of the DNA polymerase. It should be noted that for certain genomes, purity and integrity of DNA has minimal effect on the amplification reaction. Therefore, invest some time in designing a suitable DNA extraction protocol that provides high throughput and consistent amplification of your template material. Recently, some rapid DNA extraction protocols have been described (Williams and Ronald 1994, Guidet 1994).
Enzyme: The activity of thermostable DNA polymerases is highly variable (Bej and Mahbubani 1994). This variability is particularly evident in nucleic acid scanning applications where even subtle differences in specificities manifest as changes in efficiency, multiplex ratio and fingerprint composition (Bassam et al. 1992, Schierwater and Ender 1993, Aldrich and Cullis 1993, G. Caetano-Anollés and B.J. Bassam, unpublished). Different eubacterial DNA polymerases exhibit widely different optima and result in variant DNA patterns. In general, truncated DNA polymerases such as Thermus aquaticus Stoffel fragment, produce better defined fingerprints with stronger products of low molecular weight ( less than 500 bp), are more thermostable, and have a broader tolerance for wide magnesium concentrations. Overall, these thermostable equivalents of Klenow fragment DNA polymerase are more tolerant of experimental variables (Bassam et al., 1992) and are here recommended. Choosing a polymerase is usually accompanied by the selection of appropriate buffer components. Recommended buffers are generally tailored for the PCR and should only serve as the starting point for an optimization exercise to establish the influence of ionic components and concentrations.
Ionic components: Ionic components are crucial determinants of amplification. Magnesium is one important example. Consistent fingerprints can be obtained with relatively low levels of magnesium (1.5-4 mM) for plant and animal DNA and with high levels (4-8 mM) for bacteria and fungi (Table 2). However, magnesium requirements are dependent on the counterion and other buffer components. Its activity is also modulated by the concentrations of primer, template and deoxyribonucleoside triphosphates (Weaver 1994, Weaver et al. 1995). An excess of any of these components can inhibit the amplification reaction due to the sequestration of free magnesium cation. In turn, an excess of magnesium levels decrease amplification stringency and increase primer-template mismatching. Potasium, ammonium and detergents like Triton X-100 alter amplification efficiency and specificity with Tris or Tricine buffers (Caetano-Anollés et al. 1994). In contrast, pH had little effect. A study of the effect of these components defined reaction buffers that use an uniform magnesium sulfate concentration (4 mM) to amplify templates of low and high complexity (Caetano-Anollés et al. 1994). These buffers were formulated for use with the Stoffel enzyme but had deleterious effects on the activity of other enzymes such as VentR DNA polymerase (New England Biolabs, Beverly, MA). In general, our results argue against the possible formulation of a universal buffer for use with enzymes from different sources.
Thermal cycling parameters
A number of thermal cycling parameters should be considered, including annealing and template denaturation temperatures, cycle number, temperature effects on enzyme and nucleic acids, and times of annealing, denaturation and strand extension. Temperature affects the interaction between enzyme and nucleic acid species and the kinetics of the amplification reaction. For example, annealing temperature causes changes in yield, number and distribution of amplification products (Caetano-Anollés et al. 1992). When using octamer primers, annealing temperatures as high as 65C can be used.
Annealing temperature in RAPD analysis has a major impact on reproducibility and quality of banding patterns. Furthermore, an optimization of the cycling program for RAPD has shown that variables such as denaturation, annealing and extension times were very important (Yu and Pauls 1992). RAPD profiles even change with the cycling program and kind of thermal cycler used (Penner et al. 1993), as subtle changes in the temperature profile within the actual reaction tube alter amplification. In contrast, identical DAF profiles can be produced using different thermocyclers and many of thermal cycling parameters have minimal effect in the DAF reaction once fixed within a range. Examples include denaturing temperature, cycle number, and times of annealing, denaturation and strand extension. However, it is important to minimize denaturation time so as to extend the life of the enzyme and to provide enough time of annealing (especially during first cycles) as to maximize stochastic processes.
DAF amplification products are usually up to 2,000 bp in size. When using truncated DNA polymerases their average length decreases to about less than 500 bp. We found that a two step cycling between annealing (30-60C) and denaturation (90-96C) temperatures provides optimal yield and reprodicibiliy. The customary extension step was not needed if heating and cooling rates (10-20 C/min) provided adequate time for primer extension. If not, a short (30 s) extension step at 68-72C can be performed. Other cycle regimes that keep enzyme exposure to high temperatures to a minimum can also be used with little if no variation in the DAF profile. For example, Bassam and Bentley (1994) use a "touch-down" alternative resembling that of Yu and Paul (1993): 94C for 2 min, followed by 35 cycles of 90C for 30 s and 50C for 1 min. Finally, cycle number should be kept to a minimum (usually 35 cycles) to avoid "plateau" amplification effects (see Caetano-Anollés 1993).
Primer design
Specificity in the PCR is related to the ability to amplify only the targeted site; the predicted product being either present or absent. In DAF, a multiplicity of arbitrary sites is targeted and therefore specificity is expressed as the ability to produce a "consensus" fingerprint characteristic of such a multiplex reaction. In the search for this "most parsimonious" fingerprint, the nucleic acid scanning reaction must be defined as much as possible. This can be generally accomplished by minimizing the number of interactions established within and between the different primer and template species during amplification. A simple way to do this is to decrease primer length (Caetano-Anollés et al. 1992) or to block interactions between the termini of amplification products by the incorporation of a mini-hairpin at the 5' terminus of the primer (Caetano-Anollés and Gresshoff 1994). While primer length can be reduced down to 5 nt, these oligonucleotides prime inefficiently and have to be used at high concentrations, sometimes leading to inhibitory effects from concentrated primer stocks. Furthermore, their use is compromised by the existence of palindromic termini in first-round amplification products capable of forming hairpin loops and of interfering with the amplification of certain products (Caetano-Anollés et al. 1992). In turn, increasing primer length or decreasing genome complexity favours mismatching during primer annealing allowing generation of amplification products in those cases where none are to be expected on theoretical grounds (Caetano-Anollés 1994). The existence of "multiple mismatch annealing" events (cf. Venugopal et al. 1993) should therefore be decreased to a minimum. Our studies indicate that while amplification requires of a primer of at least 5 nt and annealing sites with perfect homology to the first 5 or 6 nt from the 3' terminus (Caetano-Anollés et al. 1992, Caetano-Anollés 1994, Caetano-Anollés and Gresshoff 1994), a length of 8 nt provides a good compromise between efficiency and specificity.
Fingerprint complexity and detection of DNA polymorphisms is inherently dependent on primer sequence, primer length and the number of primers in the reaction. Several factors should be considered when designing oligonucleotide sequences. For PCR, ideal primers should have 40-60% GC content and a 3'-terminal clamp of 2-3 nt, should be free of palindromes, repetitive motifs, excessive degeneracy and long stretches of purines or pyrimidines, and primer pairs should have similar size (18-25 nt), melting temperatures (Tm) and nucleotide ratios (Roux 1995). Furthermore, annealing temperatures used in the PCR reaction should straddle 2-10C below the calculated Tm values for primer-template sequences. None of these requirements are necessary or fulfilled when using arbitrary primers. In particular, several studies also demonstrate insensitivity of DAF to primer GC content (Caetano-Anollés 1994, Prabhu and Gresshoff 1994). Overall, results suggest that primer annealing is governed by the kinetic component of the reaction rather than by thermodynamic parameters.
While palindromic sequences should always be avoided, it should be noted that amplification failure of some primers or production of few amplification products can be related to the many known compositional inhomogeneities characteristic of DNA sequence. Using optimized conditions, DAF renders scorable patterns in more than 90% of primers studied.
Arbitrary primers can be used in pairwise combinations in what has been termed multiplex DAF. This strategy (Caetano-Anollés et al. 1991) has reportedly increased detection of polymorphic DNA (Callahan et al. 1993, Micheli et al. 1993). It also permits the combinatorial use of a limited set or primers. For example, a set of 10 oligonucleotides can be used in 90 different paiwise combinations (excluding those where they are used alone). However, a small fraction of amplification products (about 20%) is actually generated by each contributing primer. These fingerprint overlaps only partially compromise the potential of this multiplex DAF approach.
Data analysis
Fingerprint evaluation: A fingerprint pattern is only informative if it can be compared to other patterns. To do so the pattern has to be evaluated for matching components. For example, when DAF samples are separated by electrophoresis, individual bands within each lane have to be "sized" relative to molecular weigh standards and then compared (matched) in the search for co-migrating bands. This process of "band scoring" is dependent on parameters such as electrophoresis (type and conditions used), band detection, and quality of DNA and amplification reagents. For example, during electrophoresis mobility of DNA fragments can shift due to irregularities of the electric field or protein or polysaccharide impurities in the samples. These problems can be usually corrected by running molecular weight standards every few lanes or by including them together with the samples, by using monomorphic products as internal references, and by running samples at least in duplicate. Scanned images of gels can also be manipulated to correct artifactual electrophoretic irregularities, such as the common "smiling" effect on PAGE in which centrally located bands exhibit higher mobilities, and warping due to variation in electric field.
During electrophoresis some bands are not well resolved as others. It is good practice to score bands within a region of the gel where fingerprint patterns are well discernable. For example, using a standard PAGE protocol for DAF analysis, only those products that are less than 500-700 bp in length are generally scored. Bands should be scored consistently throught the different samples to be compared. In some cases, co-migrating bands exhibit differences in intensity which could represent the dosage of the amplification products (perhaps indicative of the homozygous or heterozygous state) or existence of more than one co-migrating band (probably representing more than one locus). Exclude from the analysis those products that exhibit wide but unexplainable variations in intensity or lend themselves to ambiguous interpretation.
Band scoring can be arduous if done by the eye and hand. Bands are sized and matched directly on gels, autoradiographic or photographic films, or photocopies on transparency overlays. The presence, absence and intensity of a band in a particular location is noted relative to the weight standards, usually with the help of rulers or alignment devices and a lightbox. Scoring by the eye and hand is demanding and may suffer from error and bias from the investigator. However, image analysis can now be automated with appropriate hardware and software tools. The image can be recorded via a high-resolution video camera, scanning device, or phosphorimage analyzer, edited to reduce background and allow band alignment, and then analyzed for band matching. The analysis usually takes into account the mobility and intensity of the bands permitting to set thresholds of tolerance for each of these parameters. Automated image analysis has increased the level of accuracy with which fingerprints are analyzed and has simplified enormously the task (Gill et al. 1991).
Interpretation: The dataset generated following band scoring and analysis of fingerprints needs to be interpreted according to each particular application. Applications include the assesment of genetic diversity, the identification of genotype, and segregation and linkage analysis of molecular markers. DAF patterns can be treated as multilocus fingerprint data, and are therefore subject to the advantages and limitations of the system (Caetano-Anollés 1994). For example, allelic pairs cannot be assigned to each other and allele frequencies cannot be calculated for each loci. Co-migrating bands can represent heterologous loci, and individual bands may not represent individual characters. Remember that DAF bands can be used as markers in genetic studies. They are largely dominant, and are therefore more accurate and better suited for genetic mapping in backcross, recombinant inbred or haploid populations than in F2 populations. Be aware that some markers can exhibit atypical segregation. In soybean, 75% of DAF markers show true Mendelian inheritance while the rest segregate in a uniparental way, being either of maternal (probably cytoplasmic) or paternal origin (Prabhu and Gresshoff 1994).
Quantification of pairwise similarities or differences between fingerprints can be done using different algorithms. Most commonly, a "similarity index" is calculated as depictor of band sharing. For example, Jaccard's similarity coefficent (J) takes into consideration only those matches between bands that are present (Jaccard 1908):
J=sp/sp+(1-s)
where p is the proportion of shared bands present in both samples and s is the proportion of the total number of bands that are shared either present or absent in both samples. In turn, the Nei and Li's coefficient (N) measures the proportion of bands shared as the result of being inherited from a common ancestor and represents the proportion of bands present and shared in both samples divided by the average of the proportion of bands present in each sample (Nei and Li 1979).
N=2sp/2sp+(1-s)
The N coefficient has been recommended for analysis of multilocus fingerprints (Lamboy 1994a). Please note that these two coefficients do not consider which of unshared bands belong to which sample. These coefficients can be corrected to accomodate the existence of artifactual false-negative or false-positive bands if they were present in the amplification reactions (Lamboy 1994a,b). These and other coefficients can be used in phenetic and phylogenetic analysis of fingerprints by using a number of statistical analysis software packages that use ordination techniques [such as "principal component analysis" (PCA) or "principal coordinate analysis" (PCO)], distance matrix or cluster analysis methods [such as "unweighted pair group method using arithmetic average" (UPGMA) or "neighbour-joining" (NJ) algorithms] or parsimony strategies [such as "phylogenetic analysis using parsimony" (PAUP)]. Examples of cluster analysis applications include PHYLIP (J. Felsenstein, Dept. of Genetics, Univ. of Washington, Seattle, WA) and NTSYS (Exeter Software, Setauket, NY), and of parsimony analysis applications PHYLIP, PAUP (D.L. Swofford, Illinois Natural History Survey, Champaign, IL), McClade (W.P Madison and D.R. Maddison, Sinauer Assoc., Sunderland, MA), and Hennig86 (J.S. Harris, Port Jefferson Sta., New York, NY).
For the analysis of mapping populations in linkage analysis and genetic mapping, calculation of recombination values and map distances can be performed with a group of other computer programs, such as LINKAGE-1 (Suiter et al. 1983), MAPMAKER (Lander et al. 1987), GMENDEL (Liu and Knapp 1990), or JOINMAP (Stam 1993).
Cost and other practical considerations: The cost per fingerprint can be estimated to be US$1.27 for DAF and US$1.08 for RAPD analysis. On a per sample basis, this cost is comparable to that of RFLP and other PCR-based fingerprints, but do not include labor, laboratory equipment and space. About two thirds of these expenses correspond to the polymerase enzyme. Therefore, those users capable of preparing their own enzyme can decrease costs considerably. Several enzyme purification protocols are available (eg. Pluthero 1993, Harrell and Hart 1994). The cost of the primer is negligible (about 1%) but can be considerable when constructing a primer library for genetic mapping or fingerprinting purposes. Primer synthesis can now provide oligonucleotides at about US$0.01/base per nmol.
An important decision is that of a thermal cycler. A wide range of thermal cyclers is available with varied heat sources, and mechanisms of heat transfer and dissipation. Heat source can be electrical, electronic (Peltier) or irradiation-based (visible or IR light). Heat transfer can be via metallic blocks, air or liquids. Heat dissipation can be achieved by convection in air-driven (oven) units, thermal conductivity in block-based units (Peltier or water cooling), or can be based on robotic transfer of tubes to blocks preset at lower temperatures (Stratagene, La Jolla, CA). Some thermal cyclers eliminate condensation by the uniform heating of the tubes and can avoid the use of oil overlays. Other thermal cyclers have sample temperature probes that accurately control cycling parameters. Finally, some units maximize heat transfer by using capillary tubes or thin-walled tubes; these units usually result in very fast cycling reactions. Depending on needs and performance, any of these units can be used. For example, the oven based thermal cyclers can accomodate a big number of tubes (500-1000), but are relatively slow. Light-driven units allow the use of very small samples, are fast, but can accomodate only few capillaries.
Thermocyclers capable of using 96-well plates or tube assemblies are very convenient and compatible with robotic or multichannel pipettors that may be used to assemble the RAPD reaction. Thermocyclers with heated lids do not require adding oil to the PCR reactions and are therefore most convenient. Some state-of-the-art thermocyclers are compatible with 384-well plates, increasing sample throughput significantly. Back
Trouble-shooting
Trouble-shooting can be a painful experience, especially when the source of the problem is not readily apparent. Two scenarios are the most commonly observed in the laboratory: (1) amplification fail for no apparent reason, and (2) spurious bands appear in the amplification or control reactions. These problems usually arise because of poor optimization, contaminated reagents, or inadequate handling of reagents and assemblage of the reaction mix.
Reproducibility
DAF profiles are highly reliable and appear free of artifactual bands. Reproducible fingerprints can be produced from replicate DNA preparations by different operators in independent experiments over time and using oven or thermal block-based cyclers.
Reproducibility requires that rare amplification events be either completely avoided or freely accomodated. Rare annealing events occur even under stringent amplification conditions and usually involve primer-template mismatching early in the thermocycling reaction. They can result in spurious products as the rare events are later amplified with high efficiency. Remember that once mismatched annealing occurs the resulting amplification products have annealing sites perfectly complementary to the primer. Artifactual non-genetic variation in RAPD analysis has been shown to be considerable in the absence of appropriate optimization of primer and template concentrations and annealing temperature (Ellsworth et al. 1993; Muralidharan and Wakeland 1993). Furthermore, RAPD analysis was found subject to day-to-day and lab-to-lab variability, and even to depend on the thermocycler used (Devos and Gale 1992, Meunier and Grimont 1993, Penner et al. 1993). RAPD inconsistencies probably arise from the use low primer-template ratios that can result in borderline experimental conditions ill-defined during optimization by the inadequate resolution of amplification products. In contrast, DAF profiles are produced with minimal experimental variability mainly because the high primer-template ratios used in the amplification reaction guarantees the free accomodation of the rare amplification events. These products of rare events are proportionally represented in the final fingerprint only after withstanding competition in a liberal "jungle" world where the most-fit becomes the winner.
Production of complex fingerprints (such as those in DAF) also ensures that the absence of a particular amplification product will not affect the probability of amplification of other products. This effect can be quite severe in RAPD analysis because of the low number of amplification products produced and the characteristics of the amplification reaction. Competition between amplification products, especially during the early cycles, could turn a stochastic event into alternative RAPD outcomes. Remember the rule of thumb: the higher the multiplex ratio the higher the reproducibility of a fingerprint.
Finally, intraexperiment variability has been commonly observed in RAPD analysis as the result of incorrectly prepared DNA containing ethanol-precipitable contaminants (Micheli et al. 1994). This potential problem can also affect DAF and AP-PCR techniques, but can be diagnosed by amplifying serially diluted DNA: reliable fingerprints should be obtained within a large reproducibility window in these experiments. Furthermore, it is good practice to amplify several 10-fold dilutions of DNA and confirm reliable fingerprints.
Nucleic acid scanning reactions must produce identical results in different laboratories over time. Penner et al. (1993) have shown "transportability" problems in RAPD analysis mainly due to variations in the amplification cycle. Transportability has not yet been tested appropriately in DAF.
Controlling contamination
Maintain a nucleic acid-free clean environment in which the only DNA that enters the reaction is the template added by the investigator (Dieffenbach and Dveksler 1993, Dragon 1993). Standard DAF can be considered a contamination-insensitive amplification technique as long as adequate levels of template DNA are used (more than 0.1 ng/µl) and no isolation and cloning of amplification products is done in the area. While there is no need to have a separate sample preparation workplace for DNA extraction and dilution, pre-amplification and post-amplification areas should be separated either physically or by working in contained environments (Dieffenbach and Dveksler 1993). Sterilize bench areas routinely with ultraviolet (UV) light. UV light reduces contamination by several orders of magnitude but is less effective with DNA fragments shorter than 300 bp (Sarkar and Sommer 1990, 1991). Therefore expose bench and materials to prolongued UV doses.
Sample preparation and re-amplification reagents should be rendered free from nucleic acid contamination. Handle reagents and manage the laboratory bench much as described by Dragon (1993). Gloves should be worn during sample and reagent preparation, setting up of amplification reactions, and retrieving amplified samples for analysis. Use sterile double-distilled water in all operations. All reagents should be prepared in large volumes, aliquoted, and if possible maintained frozen at -20C. This will guarantee reagent consistency and a decrease of contamination. If possible, use aerosol barrier pipette tips or positive-displacement pipettes for sample preparation reagents and try to maintain a set of dedicated pipettes for pre- and post-amplification activities. Always use sterilized pipette tips and disposable sterile bottles or tubes.
While standard DAF does not require the use of barrier or positive-displacement pipettes, minimize the production of aerosols by briefly centrifuging the tubes prior to opening. Do not pop-open tubes. Remember that airborn contamination in aerosols is the main cause of false positives in the bench. For maximum efficiency and minimum contamination prepare "ready-to-use" master mixes containing all reagents except for one or two missing components. Exert caution when preparing and using primer stock solutions, and include routine negative and positive amplification controls. As a final rule, establish a unidirectional traffic flow from the pre-amplification area to a contamination-contained post-amplification workplace.
Miscellaneous
A number of complicating factors can alter the outcome of nucleic acid amplification techniques such as the PCR (Roux 1995). Many of these factors affect nucleic acid scanning much in the same way. For example, pipetting can influence reproducibility by inadequate delivery of reagents. It is strongly recommended to assemble master mixes containing all but one crucial or variable reagent. When running reactions in multiple thermal cyclers, make sure that all units are calibrated in their performance. Also be aware of temperature inhomogeneities during thermocycling. Well-to-well variations can have devastating effects on reproducibility.
Absence of amplification is also a common problem. It may indicate the presence of inhibitors in the DNA sample, such as ionic detergents, stains (eg. bromophenol blue), phenol and heparin. It may also indicate the recalcitrant behavior of template DNA. Try increasing denaturation temperature and the length of the denaturation step, at least during the first cycle.
Finally, a degraded primer stcok can result in amplification failure. Most commercially-obtained primers are of satisfactory quality. However, some primer stocks can become degraded and can exhibit concentrations lower than nominal. Primers stored as a frozen solution occasionally deteriorate especially when frequently thawed and re-frozen, and should be checked periodically. Many failures are also due to inadequate mixing of the primer stock after thawing. Always thaw the stock solution completely and mix thoroughly before use. Return
