The Effects of Nuclear DNA Content (C-value) on the Quality and Utility of AFLP Fingerprints

Abstract

• Background and Aims Nuclear DNA content (C-value) varies approximately 1000-fold across the angiosperms, and this variation has been reported to have an effect on the quality of AFLP fingerprints. Various methods have been proposed for circumventing the problems associated with small and large genomes. Here we investigate the range of nuclear DNA contents across which the standard AFLP protocol can be used.

• Methods AFLP fingerprinting was conducted on an automated platform using the standard protocol (with 3 + 3 selective bases) in which DNA fragments are visualized as bands. Species with nuclear DNA contents ranging from 1C = 0·2 to 32·35 pg were included, and the total number of bands and the number of polymorphic bands were counted. For the species with the smallest C-value (Bixa orellana) and for one of the species with a large C-value (Damasonium alisma), alternative protocols using 2 + 3 and 3 + 4 selective bases, respectively, were also used.

• Key Results Acceptable AFLP traces were obtained using the standard protocol with 1C-values of 0·30–8·43 pg. Below this range, the quality was improved by using 2 + 3 selective bases. Above this range, the traces were generally characterized by a few strongly amplifying bands and noisy baselines. Damasonium alisma, however, gave more even traces, probably due to it being a tetraploid.

• Conclusions We propose that for known polyploids, genome size is a more useful indicator than the 1C-value in deciding which AFLP protocol to use. Thus, knowledge of ploidy (allowing estimation of genome size) and C-value are both important. For small genomes, the number of interpretable bands can be increased by decreasing the number of selective bases. For larger genomes, increasing the number of bases does not necessarily decrease the number of bands as predicted. The presence of a small number of strongly amplifying bands is likely to be linked to the presence of repetitive DNA sequences in high copy number in taxa with large genomes.

INTRODUCTION

Nuclear DNA content in angiosperms varies approximately 1000-fold (1C = approx. 0·1 to 127·4 pg [1 pg = 980 Mb], where 1C is the amount of DNA in the unreplicated gametic nucleus of an organism), affecting a wide range of characteristics, including rate of cell division, sensitivity to radiation, ecological behaviour in plant communities and optimal environment for crops (summarized in Bennett and Smith, 1976; Bennett and Leitch, 1995). Nuclear DNA content can evolve by various mechanisms, some of which have relatively small effects (including length evolution in non-coding DNA; Petrov, 2001, 2002) and others which have major effects (including polyploidy and active transposable element activity; e.g. SanMiguel et al., 1996; Bennetzen, 2002; Bennetzen et al., 2005). Polyploidy has a clear effect on nuclear DNA content in the context of closely related taxa, e.g. Triticum aestivum is a hexaploid and its C-value (1C = 17·33 pg) is approximately three times larger than those of related diploid taxa in Triticeae (e.g. Hordeum vulgare and Aegilops spp.). There is, however, evidence of genome downsizing in polyploids relative to their diploid progenitors in some cases (Leitch and Bennett, 2004). Genome size in diploids is the same as the 1C-value, but in polyploids, with two or more ancestral genomes, the 1C-value will exceed the genome size approximately twofold in tetraploids, threefold in hexaploids, etc. Especially in the case of allopolyploids, dividing the 1C-value by two or three as appropriate will only give a mean genome size (see Bennett et al., 1998).

Despite its relatively high ploidy, the C-value of hexaploid T. aestivum is only approximately half that of the diploid Cypripedium calceolus (1C = 32·35 pg) and only 14 % of that of tetraploid Fritillaria assyriaca, so it is clear that polyploidy does not account for all the observed variation. SanMiguel et al. (1996) demonstrated that at least 50 % of the nuclear genome of Zea mays consists of retrotransposons, and, if this is a general pattern, retrotransposon activity appears to be a good candidate for explaining the existence of such large genomes. Retrotransposons have been detected in some taxa with large genomes, including Lilium henryi (Smyth et al., 1989; Leeton and Smyth, 1993; C-value unknown, but Lilium spp. that have been measured have 1C = 32·75–43·20 pg) and Iris brevicaulis and I. fulva (1C = approx. 10 pg), with one group of retrotransposons alone accounting for 6–10 % of the genome (Kentner et al., 2003).

The vast range of nuclear DNA content, the different causes of evolution in DNA content and the relatively high percentage of the genome made up of highly repetitive sequences, at least in some taxa, have a marked effect on many traits of organisms, and DNA C-value has been referred to as a ‘key character in biology and biodiversity’ (Bennett et al., 2000). Recently, it has been demonstrated that increase in nuclear DNA content is correlated with increased risk of extinction in plants (Vinogradov, 2003), possibly due to ‘selfish’ DNA being maladaptive.

In addition to its effect on biological traits, nuclear DNA content also has a phylogenetic pattern (e.g. Leitch et al., 1998; Soltis et al., 2003; Leitch et al., 2005) and is one of the major factors taken into account in the choice of model organisms. Thus, taxa initially chosen for sequencing projects all had relatively small genomes.

It has been suggested that genome size is likely to have an effect on multilocus genetic fingerprinting techniques that assay the nuclear genome. More specifically, earlier studies have indicated that a large nuclear DNA content has indeed caused problems in some cases (Costa et al., 2000; Han et al., 2000; Fay and Cowan, 2001; Koebner et al., 2001; Fay et al., 2002b; see below for further information). Here we investigate the effect of variation in nuclear DNA content on one widely used genetic fingerprinting technique, AFLP.

Vos et al. (1995) developed this technique, which they called AFLP (amplified fragment length polymorphism) because of its similarity to RFLP (restriction fragment length polymorphism). Both techniques involve the cutting of genomic DNA with restriction endonucleases followed by the visualization of the resulting DNA fragments as bands. Polymorphisms detected in both methods can arise from loss or gain of restriction sites or insertions/deletions between restriction sites. Major differences between the two techniques, however, are that AFLP relies on detection of PCR products by electrophoresis instead of Southern hybridization as used in RFLP, and that AFLP is generally a dominant marker system whereas RFLP is a co-dominant marker system (i.e. in AFLP it is not normally possible to distinguish between homozygotes and heterozygotes at an individual locus). Vos et al. stated that ‘the AFLP technique provides a novel and very powerful DNA fingerprinting technique for DNAs of any origin or complexity’. Since the publication of the AFLP technique, it has quickly become one of the more widely used methods of DNA fingerprinting for crops and wild plant species (e.g. Mueller and Wolfenbarger, 1999; Ridout and Donini, 1999; Fay and Krauss, 2003).

For AFLP, genomic DNA is restricted with two different restriction endonucleases, and a subset of the resulting fragments is amplified using forward and reverse primers each with 1–4 (usually three) additional bases. The fragments are then visualized using radioactivity, silver staining or fluorescent dyes (the last for use with an automated sequencer). The resulting AFLP markers tend to show a strongly asymmetric size distribution, with a much higher proportion of smaller than larger fragments, and this pattern does not appear to be affected by GC content or by genome size (Vekemans et al., 2002). Vekemans et al., however, studied species with relatively small genomes (Phaseolus lunatus, 1C = 0·7 pg, and Lolium perenne, 1C = 2·08 pg) and their modelling was based on genomes (up to 0·01 pg) much smaller than even the smallest angiosperm genomes. The effect of large genome size on fragment size distribution was thus not addressed.

AFLP markers have been shown to be widely distributed across the nuclear genome in Oryza sativa (Maheshwaran et al., 1997; Zhu et al., 1998), Zea mays (Vuylsteke et al., 1999) and Pinus taeda (Remington et al., 1999). In addition to the other characteristics of this technique, including reliability and high levels of polymorphism, AFLP thus provides a good general fingerprinting system. There is evidence, however, that AFLP markers show some clustering on genetic maps. For example, in Arabidopsis thaliana, while markers were found spread across different parts of the genetic map, there was evidence of some clustering around the centromeric regions of some linkage groups (Alonso-Blanco et al., 1998). At least some AFLP markers consist of or include highly repetitive sequences and these may represent parts of retrotransposons in some cases (Reamon-Büttner et al., 1999).

AFLP has a number of advantages over pre-existing techniques. Of particular relevance to work on rare plants, only a small quantity of DNA (typically 0·5 µg per restriction) is used because the technique is based on PCR, meaning that tiny quantities of plant material are required in most cases. In addition to this, the fingerprints (visualized as traces on automated platforms) are highly reproducible and consist of many markers, allowing greater discernment between closely related plants than other techniques such as random amplified polymorphic DNAs (RAPD) and microsatellites. As a result of these advantages, AFLP has now been successfully used in projects on a range of different species, including germplasm assessment (e.g. in Miscanthus; Hodkinson et al., 2002), conservation-based studies of Astragalus cremnophylax (Travis et al., 1996), Populus euphratica (Fay et al., 1999), Medusagyne oppositifolia and Rothmannia annae (Fay et al., 2000), Tecophilaea cyanocrocus and Tulipa sprengeri (Maunder et al., 2001), Grevillea scapigera (Krauss et al., 2002) and Cosmos atrosanguineus (Wilkinson et al., 2003), studies of species delimitation in Calopogon spp. (Goldman et al., 2004), Dactylorhiza spp. (Hedrén et al., 2001) and Phylica spp. (Richardson et al., 2003), paternity analysis in Persoonia mollis (Krauss, 2000), and studies of hybridization in Sorbus spp. (Fay et al., 2002a) and Schoenoplectus spp. (Fay et al., 2003).

In the original paper describing the method, Vos et al. (1995) showed that it was possible to obtain AFLP fingerprints from organisms with a wide range in DNA content (phage λ, Autographa californica Nuclear Polyhedrosis Virus (AcNPV), Acinetobacter, Saccharomyces and Homo sapiens, plus the angiosperms—in order of increasing C-value—Arabidopsis thaliana, Cucumis sativus, Solanum lycopersicon, Brassica napus, Lactuca sativa, Zea mays and Hordeum vulgare) with 1C-values up to 5440 Mb (5·5 pg) by using primers with differing numbers of selective nucleotides (1–3) at the 3′ end based on the EcoRI and MseI restriction sites, giving 2–6 selective nucleotides in total. The angiosperms chosen by Vos et al. have 1C-values ranging from 157 Mb (0·16 pg) in A. thaliana (Bennett et al., 2003) to 5440 Mb (5·55 pg) in H. vulgare (Bennett and Smith, 1976), and these all yielded acceptable AFLP fingerprints with either 2 + 3 or 3 + 3 selective nucleotides. In the AFLP™ Plant Mapping Protocol (Applied Biosystems, 1996), the use of 2 + 3 selective bases for ‘small plant genomes’ (50–500 Mb; approx. 0·05–0·51 pg) and 3 + 3 selective bases for ‘regular plant genomes’ (500–6000 Mb, 0·51–6·12 pg) is recommended. However, angiosperms show a far wider range in C-value (up to 1C = 127·4 pg in Fritillaria assyriaca; Bennett and Smith, 1976) than that sampled by Vos et al., and various authors have reported difficulties in obtaining high quality AFLP using the standard technique with angiosperms possessing larger C-values (1C > 14700 Mb; 15 pg) e.g. Alstroemeria spp. (Han et al., 2000), Pinus pinaster (Costa et al., 2000), Cypripedium calceolus (Fay and Cowan, 2001; Fay et al., 2002a), Triticum aestivum (Koebner et al., 2001) and Cephalanthera longifolia (Fay et al., 2002a).

The wide range in C-values observed in angiosperms will have a marked effect on the number of fragments produced in restriction digests, and using the standard AFLP technique, we have observed that species with high C-values (1C > 15 pg) tend to produce traces with large numbers of weakly amplifying and often co-migrating bands (Fay and Cowan, 2001; Fay et al., 2002a). As a result, only relatively few bands can be reliably scored, and these tend to be present in most or all individuals. The variation present is normally found in the weak bands and, due to this problem, the level of variation identified is almost certainly an underestimate.

There are, however, modifications to the standard AFLP technique that can generate reliable fingerprints for some problematic species. Perhaps the simplest modification is to change the number of bases in one or both of the selective primers (Vos et al., 1995). Decreasing the number of bases can generate higher quality traces in species with smaller genomes (e.g. Vos et al., 1995; Inocencio et al., in press). Increasing the number of bases reduces the number of fragments amplified and can generate cleaner genetic fingerprints in species with larger genomes. By using a 4-base extension on the MseI primer in conjunction with a more stringent PCR protocol rather than the standard AFLP technique, Krauss (1999) was able to generate cleaner fingerprints and assign paternity to offspring in Persoonia mollis (Proteaceae; C-value unknown). Similarly, Costa et al. (2000) used a 4-base MseI primer extension with Pinus pinaster (1C = 24·35 pg), while Han et al. (1999, 2000) used a 4-base MseI primer extension together with a 4-base EcoRI primer extension in Alstroemeria spp. (1C = 22·1–39·5 pg) following a modified pre-selective PCR amplification. In vandaceous orchid hybrids (C-values unknown), clean and reproducible AFLP fingerprints were obtained using a 4-base extension on the EcoRI primer (Chen et al., 1999). In our trials, however, we have failed to produce high quality AFLP traces for several taxa with large genomes, notably Cypripedium calceolus (1C = 32·35 pg) and Cephalanthera longifolia (1C = 16·8 pg).

Another modification to AFLP involves the use of methylation-sensitive restriction enzymes such as PstI in preference to EcoRI for ‘large and complex’ conifer genomes [Picea abies (1C = 18·6 pg); Paglia and Morgante, 1998] and Sse83871 for Triticum aestivum (1C = 17·3 pg; Donini et al., 1997). However, Sse83871 is sensitive to DNA methylation, and this could lead to polymorphic bands being detected that are due to different levels of methylation between tissues and samples rather than differences in DNA sequences. Growth of plant material under controlled conditions and use of the same tissue type for all individuals are likely to minimize this problem (Donini et al., 1997), but such approaches will be difficult or impossible to apply to materials collected under uncontrolled growth conditions in nature and where the amount and type of tissue available are limited due to scarcity of the species involved.

To circumscribe the range of nuclear DNA content over which the standard AFLP protocol on an automated platform is of use with wild species, we present data here for angiosperms with a range of nuclear DNA contents (1C = 0·20–32·35 pg). We also provide preliminary data relating to the use of 2 + 3 and 3 + 4 selective bases for two species.

MATERIALS AND METHODS

DNA samples

Representative DNA samples were taken from the DNA Bank at the Royal Botanic Gardens, Kew. The species and the number of samples included are listed in Table 1.

Table 1.

Performance of species with a range of nuclear DNA contents (1C = 0·2–32·35 pg) with the standard AFLP protocol

Species	Ploidy	Ch. no.1	1C-value (pg)2	Genome size3 (pg)	No. of bands scored	No. of bands per sample	No. of polymorphic bands (%)	Primer combination4	No. of samples
Bixa orellana	2 x	14	0·205	0·20	40	33–38	10 (25·0)	E-AAG + M-CAA	4
Zostera marina	2 x	12	0·326	0·32	44	27–33	22 (50·0)	E-AAC + M-CAG	5
Sorbus aria	2 x	34	∼0·657	0·658	50	36–40	25 (50·0)	E-ACT + M-CAC	6
Bromus interruptus	4 x	28	8·439	4·22	92	67–84	35 (38·0)	E-ACT + M-CAG	15
Cephalanthera longifolia	2 x	32 (34)	16·8010	16·80	51	50–51	1 (2·0)	E-ACG + M-CTG	6
Damasonium alisma	6 x	42	23·6211	7·87	26	23–26	3 (11·5)	E-ACT + M-CAG	3
Tulipa sprengeri	2 x	24	∼24·0012	24·00	10	9–10	1 (10·0)	E-ACA + M-CAG	5
Cypripedium calceolus	2 x	20	32·3513	32·35	14	13–14	1 (7·1)	E-AAC + M-CAA	7

¹Chromosome number.

²The DNA amount in the unreplicated gametic nucleus of an organism is referred to as its C-value, irrespective of the ploidy level of the taxon.

³The average DNA amount per genome i.e. 1C-value divided by two for tetraploids and by three for hexaploids. For diploids, this is equivalent to the 1C-value.

⁴The first primer is based on EcoRI (E) and the second on MseI (M). Each has three selective bases as indicated after the hyphen.

⁵Hanson et al., 2001.

⁶Hanson et al., 2003.

⁷This value is approximate as it is that given for the two diploid Sorbus species for which the C-value is known (Bennett and Leitch, 2003).

⁸The whole subfamily Maloideae is thought to have a tetraploid origin (Dickson et al., 1992). If this is the case the genome size is 0·33 pg.

⁹Bennett and Smith, 1976.

¹⁰Bennett and Leitch, 2003.

¹¹L. Hanson, Royal Botanic Gardens, Kew, unpubl. res.

¹²This value is approximate as it is the mean value for diploid Tulipa spp. for which 1C-values are known (Bennett and Leitch, 2003).

¹³Bennett and Leitch, 2003.

AFLP fingerprinting

AFLP was performed according to the AFLP™ Plant Mapping Protocol (Applied Biosystems, 1996). For each specimen, 0·5 µg of DNA was digested using the restriction enzymes EcoRI and MseI and adaptors were ligated onto the resulting fragments. A preselective round of PCR amplification was performed using primers that were complimentary to the EcoRI and MseI adaptors and had one extra 3′ base in the case of the standard and large genome protocols, or no additional bases on the EcoRI-based primer in the small genome protocol (the MseI-based primer still has one additional base as in the other protocols). In the standard protocol, a second selective round of PCR was performed using two extra 3′ bases (three in total) on each primer. For the small genome example (Bixa orellana), one extra 3′ selective base was used on the EcoR1-based primer (two in total) and two extra 3′ bases (three in total) on the MseI-based primer. In the case of the large genome example (Damasonium alisma), reactions were carried out in the same manner except that two extra 3′ bases were added to the EcoR1 primer and three extra 3′ bases (four in total) to the MseI primer. In all cases, the EcoRI-based primer in the selective amplification had a 5′ fluorescent label.

Initially a primer trial was conducted for each species using 9–12 different normal genome selective primer combinations (3 + 3 selective bases). From this, the primer combinations that gave the clearest traces with the most interpretable bands were chosen. In the cases where large or small genome selective primer combinations were used, a further primer trial was conducted and appropriate primers chosen again on the basis of clarity and number of bands. The primer combinations used in the main study are listed in Tables 1–3.

Table 2.

The effect of reducing the number of selective bases from six to five on AFLP in Bixa orellana (a species with a small genome)

Species	1C-value (pg)	No. of bands scored	No. of bands per sample	No. of polymorphic bands (%)	Primer combination	No. of samples
B. orellana	0·20	40	33–38	10 (25·0)	E-AAG + M-CAA	4
B. orellana	0·20	91	72–81	25 (27·5)	E-TG + M-CAA	4

Table 3.

The effect of increasing the number of selective bases from six to seven on AFLP in Damasonium alisma

Species	1C-value (pg)	No. of bands scored	No. of bands per sample	No. of polymorphic bands (%)	Primer combination	No. of samples
D. alisma	23·62	26	23–26	3 (11·5)	E-ACT + M-CAG	3
D. alisma	23·62	67	65	4 (6·0)	E-ACT + M-CAG + A	2
D. alisma	23·62	54	48–53	7 (13·0)	E-ACT + M-CAG + C	2
D. alisma	23·62	41	36–42	6 (14·6)	E-ACT + M-CAG + G	2
D. alisma	23·62	49	44–47	7 (14·3)	E-ACT + M-CAG + T	2

For the large genome primer combinations, the average number of bands scored was 53. The overall percentage polymorphism for the large genome primer combinations was 12·0 %.

The reactions were separated on a 5·0 % polyacrylamide gel using an ABI 377 automated sequencer. GeneScan 2·1 and Genotyper 2·0 (Applied Biosystems) were used to analyse bands.

Data analysis

Bands were scored as either present or absent for all individuals and exported from Genotyper into a Microsoft Excel spreadsheet. We then calculated the total number of scored bands per species, the number and percentage of polymorphic bands, and the ranges of band number per sample.

RESULTS

Standard protocol

We obtained AFLP traces for all species tested. However, the quality of the traces and the number of bands that could be scored was variable (Fig. 1; Table 1). Although quality of AFLP traces is difficult to assess quantitatively, those obtained for Bixa orellana (the species with the smallest C-value tested) were clearly suboptimal, with relatively few peaks, most of which amplified comparatively weakly. In contrast, those obtained for Zostera marina (with a slightly larger C-value) appeared better with clearer bands, although the total number of bands scored was not much greater. Sorbus aria and Bromus interruptus both performed well, even though the C-values differ more than tenfold. Although the number of bands scored for S. aria was greater than that for Z. marina, the number of strongly amplifying bands was lower (Fig. 1), possibly due to the suggested tetraploid origin of Maloideae (see Discussion). Despite their relative C-values, Damasonium alisma produced clearer traces than Cephalanthera longifolia and Tulipa sprengeri, although this could be due to D. alisma being a tetraploid (see Discussion). Cypripedium calceolus gave poor traces with relatively few, strongly amplifying bands, and low levels of polymorphism. In D. alisma, C. longifolia, T. sprengeri and C. calceolus, the percentage of polymorphic bands was low.

Fig. 1.

Representative AFLP traces for species with nuclear DNA contents ranging from 0·2 to 32·35 pg. The traces are arranged in order of increasing nuclear DNA content with the smallest (Bixa orellana) at the top and the largest (Cypripedium calceolus) at the bottom.

Small genome protocol

For Bixa orellana (Table 2), reduction of the number of selective bases from six to five increased the quality of the traces (Fig. 2) and the number of bands scored, but the percentage of polymorphic bands was similar (25·0–27·5 %) using both protocols.

Fig. 2.

Representative AFLP traces for Bixa orellana, showing the improvement achieved by using the small genome protocol (below) instead of the standard protocol (above).

Large genome protocol

For D. alisma (Table 3, Fig. 3) increasing the number of selective bases from six to seven increased the number of bands scored by approximately twofold from 26 to an average of 53 (range 41–67) per primer combination. The overall percentage of polymorphic bands (12·0 %), however, remained similar.

Fig. 3.

Representative AFLP traces for Damasonium alisma, showing those obtained using the standard protocol (top) and the large genome protocol (below).

DISCUSSION

Nuclear DNA content and AFLP quality

As previously recorded (see Introduction), nuclear DNA content is demonstrated to have a strong effect on the quality of AFLP. With the small genomes tested here, the quality of the traces is not a major problem with the standard protocol, but relatively few bands are detected. In the case of Bixa orellana, the number of bands was increased approximately twofold (Table 2) by decreasing the number of selective bases from six to five (the ‘small genome’ protocol of Applied Biosystems). With both protocols we detected a similar level of polymorphism (25·0–27·5 %). Thus it appears that the estimate of genetic diversity obtained with this species is similar regardless of the protocol used. However, the larger number of bands detected means that the small genome protocol results in more resolving power between individuals.

As C-value increases, we saw an improvement in the quality of the traces (Fig. 1) and a general increase in the number of bands detected (Table 1) up to a nuclear DNA content of 1C = 8·43 pg (Bromus interruptus). The quality of the traces for Sorbus aria was a possible exception to the general trend, although band number scored fits into the increasing series. Maloideae (the subfamily of Rosaceae to which Sorbus belongs) are believed to have a polyploid origin and the mean nuclear DNA content of ‘diploids’ in Maloideae is more than twice that found in other subfamilies (0·73 pg in Maloideae vs. 0·28–0·33 pg in the other subfamilies; Dickson et al., 1992), and thus the ‘true’ genome size of Sorbus aria may be approximately 0·33 pg, rather than 0·65. If this is the case, then this may explain the unexpectedly poor quality of the traces.

With genomes larger than that of Bromus interruptus, the quality of the traces and the number of bands detected decreased. It is difficult to assess the effect of nuclear DNA content on levels of polymorphism detected based on the data presented here alone, although in both cases where we present data for the same species using different protocols (Bixa orellana and Damasonium alisma), the percentage polymorphism is similar, regardless of the protocol. For D. alisma, the effect of increasing the number of selective bases from six to seven initially appears counterintuitive, as the number of bands recorded increased approximately twofold rather than decreasing by fourfold as would have been predicted on theoretical grounds. However, this increase in band number is due to an increase in the number that were interpretable rather than an increase in overall band number. In some cases, other types of data suggest that the figures for percentage polymorphism (a measure of genetic variability) are an underestimate in the case of species with large genomes. In Cephalanthera longifolia, populations have been shown to be as genetically variable as other outbreeding orchid species using allozymes (Scacchi et al., 1991), as is also the case in the related species C. rubra (Scacchi et al., 1991; Brzosko and Wróblewska, 2003). Using three plastid microsatellite loci, we were able to detect four different haplotypes using the same range of material of C. longifolia for which we only detected two genotypes here (Micheneau, 2002; C. Micheneau et al., Royal Botanic Gardens, Kew, UK, unpubl. res.). In Cypripedium calceolus, populations have been shown to have relatively high genetic diversity using allozymes (Brzosko et al., 2002), and using only one variable plastid microsatellite we detected four different haplotypes for the same range of material for which we only detected two genotypes here (Fay and Cowan, 2001). Using a larger number of length-variable plastid DNA sequences, we can now recognize several additional haplotypes (M. F. Fay et al., unpubl. res.).

The decrease in the number of AFLP bands scored in these larger genomes is due to the fact that many of the loci are not amplified strongly enough to be scored, rather than to a real decrease in the number of loci. This is evident from the noisy baseline in the traces for the species with larger genomes (Fig. 1). This noise represents weakly amplifying loci that do not reach the threshold for detection. Compounding this, in many cases it appears that the number of weak bands is such that many of them are of similar enough size to make reliable interpretation impossible. The weak amplification of many loci contrasts markedly with the strong amplification of the loci producing bands that can be scored. At least some of these strong bands represent repetitive, high copy loci, and it has been suggested that some may represent parts of transposable elements (Reamon-Büttner et al., 1999). In Cypripedium calceolus, we have cloned some of the strongly amplifying bands, and some are indeed parts of retrotransposons. In one case, we have cloned 136 slightly different copies of one band, including 32 from one individual, supporting the idea that the strongly amplifying bands are present in high copy number (M. F. Fay et al., unpubl. res.).

Ploidy, nuclear DNA content and AFLP

On the basis of the data presented here, it appears that in polyploids the genome size (rather than the C-value) may be the factor effecting quality of AFLP as total nuclear DNA content increases. Thus the quality of the AFLP traces for Damasonium alisma (a tetraploid) is markedly superior not only to that for traces of Tulipa sprengeri (a diploid with a similar 1C-value) but also to that for those of Cephalanthera longifolia (a diploid with a 1C-value approximately 71 % of that of Damasonium alisma). On this basis, the quality of the traces for Bromus interruptus (a tetraploid) may be better than would be expected for a diploid with a similar 1C-value. Conversely, the quality of the traces for Sorbus aria was lower than expected, and this may be due to the proposed polyploid origin of Maloideae. The relatively high quality of the traces for Damasonium alisma may indicate that the standard protocol can be used with species where the genome size is ≤12·0 pg.

Other studies have used AFLP to investigate taxa of variable ploidy. In Dactylorhiza, the diploid species D. fuchsii and D. incarnata, the apparently autotetraploid D. maculata (derived from a yet to be identified diploid) and a range of allotetraploids, derived from hybridization events between D. fuchsii or D. maculata on the one hand and D. incarnata on the other, were studied (Hedrén et al., 2001). Similar studies include those on diploid, triploid and tetraploid Miscanthus (Hodkinson et al., 2002), diploid and hexaploid Calopogon (Goldman et al., 2004) and tetraploid to dodecaploid (or possibly duodevigintiploid) Dupontia (Brysting et al., 2004), and plants with higher ploidy levels produced AFLP traces of similar quality to those with lower ploidy levels. In all these cases, there was no evidence of decrease in band number with higher C-value, and the number of interpretable bands showed a general increase with ploidy level within each genus. It thus appears that there is a general pattern for polyploids to give higher quality AFLP traces than their C-value might indicate, whereas increases in DNA content per genome (>15·0 pg) cause a decrease in AFLP quality.

Choosing the appropriate AFLP protocol

It appears that polyploids give better AFLP traces than would be expected based solely on their 1C-value, and we therefore recommend the use of the genome size (rather than nuclear DNA content) as a more useful indicator of the utility of different AFLP protocols for polyploids. On the basis of the data presented here, we suggest that the standard protocol with 3 + 3 selective bases is useful for species where the genome size (not the 1C-value in polyploids) is in the range 0·3–12·0 pg (somewhat broader than the range of 0·51–6·12 pg, recommended by Applied Biosystems, 1996). Below this range, reduction in the number of selective bases can increase the number of bands scored, but does not necessarily reveal a higher percentage polymorphism. Above this range, increasing the number of selective bases may be of utility in cases where the genome size is less than approx. 15 pg. For genomes >15 pg, the presence of repetitive loci in high copy number may mean that the reliable interpretation of loci present in low copy number (those that are likely to be polymorphic) is not achievable. In these cases, the use of other types of genetic markers that focus on particular regions of the genome are probably more appropriate. Nuclear microsatellites are one such marker type, although they may also not be ideal for species with large genomes. In addition to the long development time compared to AFLP, increasing genome size has been shown to have a negative effect on amplification of microsatellites, and it is consequently more difficult to work with microsatellites in species with large genomes (Garner, 2002). Thus it appears that nuclear genome size may have a more general impact on genetic fingerprinting.