TILLING is a good alternative to more direct DNA modifying techniques since seed mutagenesis is easy to apply and relatively independent of genome size and organization [46
]. To determine the feasibility of developing an oat TILLING population initial experiments were performed to find an optimal balance between mutation frequency and lethality. Since the oat genome size is very large, it is crucial that the mutation frequency in the final TILLING-population is high. Our aim was to produce a population where mutations in all genes can be found with a high redundancy without having to produce a very large population size, which requires more advanced logistics in handling and maintenance. Screening large populations also becomes laborious, expensive and technically more demanding. On the other hand, a too high mutation frequency will increase the frequency of deleterious mutations, which if too numerous will kill the plant. After testing different EMS concentrations we settled for 0.9% (v/v) or 86.98 mM, which gave a germination rate of 37%. This EMS concentration is in the range of what has been used for wheat (0.75%, 1.0% or 1.2% v/v) [29
] and maize (1% v/v) [30
], whereas for barley three different concentrations were used (10, 20, 30 mM corresponding to 0.1%, 0.2%, 0.3% v/v) [25
]. The concentration used for rice was much higher (1.5% v/v) [32
four genomes, denoted A, B, C and D, have been identified. In nature, 5 different combinations of these genomes, namely AA, CC, AABB, AACC and AACCDD can be found [47
]. Belinda, like other commonly cultivated A. sativa
species, is a natural allo-hexaploid that contains three genomes AA, CC and DD. Such a redundant genomic organization could be a potential complication when identifying specific phenotypes from the TILLING-population. The presence of a specific allele on all three genomes would mean that even if one is mutated other alleles could compensate for the lost function. However, as has been shown in hexaploid wheat, the three genomes often contribute differently to the expression of certain alleles, which also can also vary even in different tissues of the plant [48
]. When inspecting the oat M2 segregating TILLING-population visible phenotypes were indeed observed in about 5% of the plants (Figure ). This is surprisingly high considering that in wheat visible phenotypes were seen in only 0.5% of the cases [29
]. To extend the resolution of the phenotypic observations we therefore decided to test a specific biochemical pathway.
We stained individual M3 seeds from 1824 randomly chosen individual lines with phloroglucinol-HCI (Wiesner test) [41
]. This test is generally considered to be indicative of aldehyde end groups present in, for example cinnamyl aldehydes and is thus indicative of lignin content and/or structure [49
]. The Wiesner test revealed differences in the staining pattern in approximately 1% of the screened seeds. Again, this is a surprisingly high frequency, which corroborated the phenotypic observations. To verify this, total lignin in seeds from the mutant lines were quantified with the acetyl bromide method. In this assay, bromine replacement of α-carbon-OH groups produce a brown-red colored complex, which is dissolved in acetic acid. The intensity of the color is proportional to the amount of lignin [50
]. The assay confirmed all 17 of the earlier identified mutants (Figure ). Since it is not likely that independent mutations have simultaneously occurred in the same lignin gene in all of the three oat genomes, this indicates that at least some mutations are dominant in a way that is normally seen in a diploid plant. More studies are needed to determine if oat is different from wheat in this respect.
To test the mutation frequency more directly on the DNA level, two different approaches were used - RAPD and MALDI-TOF. One was more generally aimed at the whole genome, while the other was directed at specific genes.
By optimising a RAPD-PCR based method for oat genomic DNA the average mutation frequency for the whole genome was calculated. An advantage with this method is that it is relatively fast and simple. When analysing DNA samples from 252 different lines, 4 abnormal electrophoresis patterns were found, corresponding to a mutation frequency of approximately 1 per 20 kb (data not shown) assuming that only a single bp is sufficient to prevent primer binding. However, this might not always be the case and therefore the calculated mutation frequency will rather underestimate than overestimate the true frequency.
To compare the genome wide mutation frequency to the frequency in specific genes, two different genes, each encoding characteristics important for oat breeders, were chosen as models. The first gene, AsPAL1
, encodes a protein that catalyses the first step of the phenyl-propanoid pathway [51
]. The second gene, AsCslF6
, encodes an enzyme involved in the biosynthesis of mixed link 1-3, 1-4 β-D-glucans (β-glucans). β-glucans are water-soluble fibres common in the cell wall of the Poaceae
family and are synthesised by one or several members of a very large cellulose synthase superfamily, from which the CslC
genes seem to be most relevant [52
]. These two genes were screened for mutations by MALDI-TOF.
By adapting a MALDI-TOF based SNP screening method to oat TILLING, and by direct DNA sequencing, mutations for the two target genes were identified. In total 533 kb were covered and 16 mutations were found (Table ), giving an average mutation frequency of 1 per 33.3 kb. This discrepancy from the RAPD-PCR estimation could reflect the fact that, in contrast to the genome wide estimate where local differences are evened out, factors such as chromosomal location, degree of closed or open chromatin, gene redundancy level etc., can influence the result when focusing on more specific regions. Additional investigations are therefore needed to explain more precisely why the mutation frequency in the chosen genes appears to be lower than the general frequency.
Overall, the mutation frequency in the oat TILLING-population thus appears to be in the same range as previously found in hexaploid wheat (1 mutation/24 kb) [29
]. This is much higher than what was described in barley (1 mutation/1000 kb) [25
], maize (1 mutation/500 kb) [30
], rice (1 mutation/530 kb) [32
] and other dicots [28
]. The specific reasons for these differences are not yet known, but presumably larger polyploid genomes like oat and wheat can tolerate more mutations before lethality occurs.
Successful TILLING not only depends on the population size and the mutation frequency, the method used to identify specific mutations in the population is equally important. This is especially a concern with a large, hexaploid genome like oat. Up to now, all, published high-throughput TILLING procedures have utilized a CEL I based mismatch-cleavage enzyme on heteroduplex formations followed by a detection of end-labeled cleavage products on electrophoretic gels [30
]. On the other hand MALDI-TOF based methods to identify SNP-mutations also have an inherent capacity to combine straightforward sample preparation with high sensitivity and high-throughput and are well documented, especially in clinical applications [36
]. However, to our knowledge MALDI-TOF methods have not been systematically exploited in TILLING. One drawback with current MALDI-methods is their reliance on specialised equipment and software for identifying the mutations. Another potential problem for large scale screening programs is the high costs of the chemicals needed. On the other hand, MS based methods are sensitive, robust and reliable, which minimise redundant experiments and reduce the cost. They also allow the identification of homozygous mutations. In addition, using several alternative cleavage patterns from the same sample, mutations can be pinpointed down to the single bp level, potentially eliminating the need for confirmative DNA sequencing.
Thus, the MALDI-TOF method used here indeed detected mutations with an, up until now, 100% precision. The final percentage is likely to decrease somewhat as sample size increases. In any case, MALDI-TOF will be good complement or possibly an alternative to CEL I/Li-COR-screening provided that the per-sample cost could be brought down to an acceptable level and throughput increased. Without automation, MALDI-screening is still a low throughput method only allowing one person to screen approximately 288 samples per week, assuming 8-hour workdays. We are currently in the process of optimising and automating the protocol for large-scale screening.
A comprehensive analysis of the spectrum of mutations generated by EMS in Arabidopsis
was published in 2003 [24
]. It was shown that about 50% of all mutations were missense mutations that altered an amino acid, while approximately 5% of the mutations were nonsense resulting in premature termination of the target protein. Furthermore, in Arabidopsis
, maize and wheat the majority of all mutations were the expected GC to AT transitions [29
]. In rice on the other hand, this distribution was somewhat different and only 70% of recorded mutations were GC to AT [32
]. In oat, we have so far identified 16 mutations, out of which 15 were GC to AT transitions (Table ). Thus, oat seems to be similar to wheat in this respect. Most mutations were missense and silent. So far no nonsense mutation have been found (Table ). We did not find any examples of small deletions or other rearrangements either, which indicates that the EMS mutagenesis worked as expected.