The modular approach to method validation initially proposed by [9] and also discussed elsewhere [CO-EXTRA project, http://www.coextra.eu/research_themes/topics202.html, Codex Committee on methods of analysis and sampling, ftp://ftp.fao.org/codex/Alinorm10/al33_23e.pdf; 25], entails that each step of the analytical procedure is validated as a stand-alone module and needs to meet acceptance criteria before the output of one module is transferred as input to the next; its practical implementation would offer several advantages at enforcement level:
- laboratories could select the preferred and ideally validated DNA extraction procedure providing acceptable yield and DNA quality fit for the PCR-based module;
- DNA extraction protocols could be applied to different matrices provided the acceptance criteria set for DNA quality are met;
- the modular approach would make it possible to reduce the number of target-taxon reference systems in food analysis, particularly when multiple events from the same ingredient (i.e. same plant species) in a sample could be quantified in relation to only one validated taxon-specific system, with the practical consequence of favouring less laborious, but time-effective and cost-saving practices.
Incidentally, general endorsement of the modular approach would also support wider acceptance of flexible application of accreditations across regions and organisations, thus cutting several laboratory constraints and burdens [25].
Recently, the European Network of GMO Laboratories (ENGL, http://gmo-crl.jrc.ec.europa.eu/guidancedocs.htm) updated a guidance document on how methods for GMO analysis should be evaluated and validated in the context of the European legislation. It included a minimum set of acceptance criteria to appraise the quality of DNA extracts consisting mainly of running inhibition tests and applying stringent acceptance criteria. This is a move in the direction of underpinning the modularity principle and providing tools for critically combining the DNA extraction module with the analytical module (qPCR) in a single method. Generally speaking, criteria to appraise DNA fitness to access downstream qPCR applications should take in consideration three elements: i) DNA in reaction should be enough to result in efficient target amplification. This is achieved following the reaction set-up described in the validated method. If DNA yield is a limiting factor from the DNA extraction module, it is essential to load in reaction an amount of GM-target sequence above its limit of quantification (LOQ); ii) the DNA should be of sufficient length to be amplified in qPCR. This has basically to do with the amplicon length which are normally quite short in qPCR methodologies, therefore in case of highly processed matrices it is important to combine the DNA extraction method which does not add to the fragmentation state of DNA molecules; iii) finally, DNA extracts should be adequately free from PCR inhibitors. This is recognised as a relevant point by the scientific community and it is mentioned in ISO norms (20, 21).
A proposal to implement the modular principle and apply a quality 'filter' on DNA extracts, focusing on the reaction efficiencies of dilution series of DNA extracts, was also put forward independently [12].
Consequently, the primary aim of this study was to investigate whether different DNA extraction methods could be interchangeable and if DNA acceptance criteria could be applied to make the two modules independent.
The extraction of nucleic acids from biological material requires cell lysis, inactivation of cellular nucleases and separation of the desired nucleic acid from cellular debris. Three extraction methods were selected in this study due to their wide application in GMO field and plant molecular biology and to the difference on lyses buffer and on DNA isolation technique. The CTAB-based protocol (http://mbg.jrc.ec.europa.eu/capacitybuilding/docs-manual-EN.htm, for extensive review) is particularly suitable for the elimination of polysaccharides and polyphenolic compounds otherwise affecting the DNA purity. Plant cells can be lysed with the ionic detergent CTAB, which forms an insoluble complex with nucleic acids in a low-salt environment. Under these conditions, polysaccharides, phenolic compounds and other contaminants remain in the supernatant and can be washed away. The DNA complex is solubilised by raising the salt concentration and precipitated with ethanol or isopropanol; Dellaporta-derived method uses first a thermal lyses in the presence of Tris HCl, EDTA, CTAB and β- mercaptoethanol, followed by deprotenisation and removal of contaminants by phenol-chloroform precipitation; the Wizard™ ('Wizard') magnetic DNA purification for food kit makes use of a thermal lyses and magnetic bead technology for the DNA isolation in a mobile solid phase. Solid phase techniques simplify nucleic acid purification since they can replace several steps of centrifugation, organic extraction and phase separation with a single, rapid magnetic separation step.
The analysis performed on the full set of data - assessing three matrices and three extraction methods - revealed significant interaction effects that tended to decline (in probabilistic terms) when data reduction was performed by applying acceptance criteria to inhibition tests (see Table 3). However, such interaction effects (that is, violation of the assumption of module independence) were mainly due to application of the Wizard DNA extraction method to the biscuit matrix (see Figure 5) and disappeared when these data were excluded from the analysis, combined with application of quality criteria from inhibition tests. Wizard-based DNA extraction showed a significant overestimation of the % GM DNA concentration, particularly in the biscuit matrix (see Figure 5). Similar findings were reported by [14] on flours of certified reference material and by other authors [29] when the DNA extract was passed through the magnetic beads purification step. Alike results were obtained for a complex feedstuff, where the Wizard magnetic DNA extraction method introduced a bias in the ratio between the two analytes, with the results showing a significant increase (up to 20 times), much more than observed by [15] in flours of certified reference material with regard to the % GM obtained when the DNA extract was passed through the magnetic beads purification step (Berben, G., CRA-W, Gembloux, Belgium, personal communication).
This circumstance is also substantiated by the fuzzy logic-based assessment which revealed quite a high score (0.75) for the fuzzy-logic indicator (see Table 4), associated with lower performance from the biscuit/Wizard DNA extraction combination (i.e. high quantification bias).
Such analysis gives a clue to module independence. Analysis of variance found that applying data-filtering criteria before running the qPCR showed some effectiveness. It made it possible to assess the interaction between the factors involved in the process leading to GM DNA quantification. Although application of acceptance criteria to inhibition tests is certainly useful, its effectiveness for making the matrices independent of the DNA extraction method needs further investigation. Moreover, this can be achieved at the cost of excluding many data, e.g. those obtained when the Wizard extraction method was applied to feed (see Table 1). Cases like this are a clue that the wrong method is being applied to a given matrix.
In this study, low 260:230 ratios were observed (especially with Wizard but also with Dellaporta, when the latter was applied to feed and biscuit, see Table 1), possibly due to the presence of contaminants absorbing light at 230 nm (i.e. ethylenediaminetetraacetic acid, phenol and guanidine thiocyanate), but this was not related to the DNA yield. This study also considered whether the well-known laboratory practice to estimate the purity of DNA extracts based on absorbance ratios at defined wavelengths (260:280 and 260:230 in the ultraviolet spectrum) was in agreement with assessment of DNA quality by means of inhibition runs. Basically, no significant convergence was seen in this respect (see Table 1). Absorbance ratios of 260:280 and 260:230 have long been used in molecular biology to obtain an estimate of the purity of nucleic acid preparations intended for numerous applications, such as Southern and Northern blotting, reverse transcription, cloning and generation of libraries [30]. However, these common ways of assessing DNA purity are insufficient when it comes to decide whether a DNA extract is fit-for-purpose of application to qPCR, where parameters such as method linearity over a dynamic range and amplification efficiency come into play [[12, 19] and http://gmo-crl.jrc.ec.europa.eu/doc/Min_Perf_Requirements_Analytical_methods.pdf]. The proposed acceptance criteria for assessment of DNA quality are hence intended to 'filter' for those DNA extracts suitable for downstream qPCR application.
Determination of the absorbance of a DNA solution in the UV spectrum - and conversion of the result into a concentration measurement - is a common technique. Recently fluorescence response of intercalating dyes, particularly the PicoGreen dsDNA quantitation reagent, are gaining more and more acceptance due to their reliability, sensitivity, extended dynamic range and lesser interaction with single-strand nucleic acids [31]. Consequently, this method for determining DNA concentration is becoming a well-established procedure in GMO testing laboratories too.
A recent study [32] found that UV and fluorimetric determination (average between the PicoGreen and the Hoescht dye methods of determining DNA concentration) agreed quite well when DNA concentration was measured from DNA samples extracted by the CTAB procedure from a relatively simple and unprocessed matrix such as NK603 certified reference material, at different percentages of GM content. However, once the CRM materials were processed by sonication or heat treatment this agreement was lost and, as the level of DNA degradation was forced, the amount of DNA measured increased when the UV method was used but, importantly, declined using the fluorescent dye method. The findings presented in this study mirror such behaviour and show that the combination of CTAB with 1% RRS certified reference material produced fairly close agreement between the estimates of DNA yields either via UV (260 nm) or PicoGreen (see Table 1). This study found different outcomes when other DNA extraction methods were used and DNA extracts were quantified by UV and fluorescence. Processed materials (biscuit and feed) showed the most striking deviations between spectrophotometric and fluorimetric results. This is probably due to the ability of short or single-strand nucleic acid fragments to interfere with UV rays more than with PicoGreen dye. A qualitative assessment of the DNA extracts by visual inspection of the results of the gel electrophoresis run (see Figure 2) does indeed agree more with the yield data from fluorescence methods than with those obtained from spectrophotometric determination. Gel analysis suggests that per each DNA isolation procedure, DNA extracted from the feed, and even more clearly from the biscuit, matrix suffered extensive damage compared to an unprocessed matrix (CRM 1%): the signs of broad degradation are quite striking, with DNA fragments dispersed everywhere between high and very low molecular weight. A tendency to select for low molecular weight DNA strands might be identified with the Wizard beads and Dellaporta methods. By contrast, DNA extracted with the CTAB and Dellaporta methods from certified reference material show a distinctly high molecular weight band.
Hence, the 'liberal' estimation of DNA content by UV could not adequately reflect the relative abundance of intact and amplifiable targets and could affect GMO quantification. In other words, if DNA is severely damaged the less abundant reaction target, typically the GM analyte, might fall close to or below the LOQ with practical consequences for the ability to quantify the target accurately. In order to quantify GMOs in food correctly, the degree of feed processing and the amount of degraded DNA or its average molecular weight are not paramount. Instead, the crucial factors are that the average length of the extracted DNA molecules must be longer than the amplicon sizes, the amount of the less abundant target sequence must be above the practical LOQ [33], there must be no imbalance between the GM/reference targets (if induced by industrial processing and/or extraction protocols) and there must be well-defined but affordable acceptance criteria for the quality of DNA extracts, as this study shows. Comparable reaction efficiencies between the food/feed sample and the reference curve [12] should also be considered further grounds for scientific consensus.