Introduction

Genetic variations, which are major factors contributing to the diversity and evolution of all organisms, have long been used to enhance cultivated plant lineages. Artificial mutagenesis induces genetic variations and is frequently used in plant breeding programs. Ionizing radiation is one of the most common mutagens in these programs because the associated methods are relatively simple and are applicable for all plant materials and species1,2,3. Because radiation tracks are allocated randomly in the target nuclei, the radiation-induced DNA damages likely occur randomly in the genome4,5. However, repair of the damaged DNA could not be equally proceeded at respective regions in the genome. For example, it is well known that structurally altered DNA lesions are excised and repaired more effectively in transcribed regions than un-transcribed regions6. It was also reported that DNA double-strand breaks in transcriptionally active regions was accurately repaired by error-free homologous recombination pathway in human cells7,8. Configuration of chromatin at DNA lesions is also important for determining their repair efficiency9. Very recently, using large datasets of genomics and epigenomics in Arabidopsis, spontaneous mutations were shown to occur non-randomly in the genome10. Despite of the recent progress, a long-standing theory on the randomness of radiation-induced mutations in the genome has not been fully verified in higher plants. If the radiation-induced mutations occur in non-random, more efficient mutation breeding program might be explored.

Mutation analysis in irradiated M1 plants is important for addressing directly mutagenic effects of radiations, and the obtained knowledge would be helpful to plant breeding especially for vegetatively propagated plants. Radiation generates unique mutational patterns in the cells of irradiated tissues. Thus, the irradiated M1 plants comprise a mixture of cells with varying mutational patterns. Because of the complexity of M1 plants, radiation-induced mutations have generally been analyzed in the self-pollinated plants derived from the irradiated M1 plants (i.e., M2 or subsequent generations)11,12. However, some types of radiation-induced mutations, including large deletions and chromosomal rearrangements, are hardly transmitted from one generation to the next13. Thus, to accurately evaluate the randomness of the mutations induced by ionizing radiation, the irradiated M1 plants that harbor all types of mutations should be used for developing a simple experimental system that enables a comparative analysis of mutations at multiple genomic loci.

Several heterogenous marker systems have been used to detect mutations in mutagenized M1 plants. The β-glucuronidase-encoding gene (GUS) is a typical example. Constructs for the functional silencing of GUS are introduced into the target genome prior to mutagenesis. When a mutation restores GUS expression, cells harboring the restored GUS in mutagenized plant tissues are detected on the basis of blue pigmentation14. Several GUS constructs designed to detect specific types of DNA changes, such as base substitution and homologous recombination, have been used15,16. A single genome may simultaneously contain several GUS constructs, but all of the constructs will produce the same blue pigmentation when at least one of them restores GUS expression. Accordingly, it is difficult to discriminate between mutations at multiple loci in a single plant via phenotypic examinations. Another heterogenous marker gene, rpsL (encoding the ribosomal S12 protein in Escherichia coli), has also been used to detect mutations in the M1 genome. Because mutations in rpsL result in streptomycin-resistant E. coli, plasmids rescued from transgenic plants with rpsL in their genome are inserted into E. coli cells, which are screened on the basis of streptomycin resistance and analyzed to reveal mutations in the rpsL region17. In contrast to GUS constructs, a single rpsL construct can detect multiple types of mutations, including single base substitutions and insertions/deletions (indels), if they occurred within its genic region (375 bp). However, when multiple rpsL constructs are introduced into one genome, it is difficult to determine the original genomic location of the mutated rpsL gene. Therefore, although these systems can provide researchers with valuable information regarding the mutations in plant genomes caused by ionizing radiation18,19, they are not suitable for detecting a wide range of radiation-induced mutations at multiple loci in a single plant.

Flavonoid pigments are conserved in many plant species and are frequently used to detect spontaneous and artificially-induced mutations20,21. The flavonoid biosynthetic pathway in the model plant Arabidopsis has been extensively analyzed at the molecular level. All of the flavonoid compounds in Arabidopsis are synthesized in a single biosynthetic pathway, and the genes encoding the enzymes catalyzing the pathway reactions have been identified22. The brown coloration of mature Arabidopsis seeds is due to the oxidation of the non-colored proanthocyanidins23. The disruption of any of the biosynthetic steps in this pathway decreases the accumulation of brown pigments in the seed coat, resulting in the so-called transparent testa (tt) phenotype. Notably, mature seeds of some tt mutants are differentially colored, making them visually distinguishable24. This indicates that each tt mutant can be discriminated according to the color of the mature seeds. Another important feature is that the flavonoid biosynthetic pathway in Arabidopsis consists of only one set of genes in the genome. This enables the detection of mutations in the flavonoid biosynthesis-related genes in M1 plants on the basis of the loss of heterozygosity (LOH). The detection of LOH in the M1 generation would also decrease the number of plants that must be analyzed compared with the detection of homozygous mutants in the M2 or later generations. In this study, we tried to develop a system for detecting mutations at multiple loci in the M1 generation by examining flavonoid pigmentation in Arabidopsis seeds. The discrimination of two mutants according to seed color was supported by the detection of mutations in the genes responsible for the phenotype. Using this system, we revealed that the mutation frequencies differed significantly at two loci (TT4 and TT8).

Results

Isolation and characterization of pab2 and pab3 mutants

Using a mutagenized M2 population derived from banyuls (ban) plants, we previously isolated a mutant with a pale-ban (pab) phenotype, which was designated as a pab1 mutant25. The subsequent screening of this population detected two new pab mutants (pab2-1 and pab3-1), both of which had immature seeds with relatively low anthocyanin contents (Fig. 1a). Because the pab2 and pab3 phenotypes were only detected in the ban background, we referred to these mutants as pab2 ban and pab3 ban, respectively. The results of the rough-mapping analysis for the pab2-1 ban and pab3-1 ban mutants indicated that the genes responsible for their strong anthocyanin-less phenotypes were located near the TT4 and TT8 loci, respectively. The TT4 gene encodes the first enzyme of the flavonoid biosynthetic pathway26, whereas the TT8 gene encodes a seed coat-specific transcription factor that is necessary for the production of the “late” enzymes of the pathway27. The examination of the sequences in the TT4 and TT8 regions revealed mutations in these two genes. More specifically, a large fragment, including most of the TT4 region, was deleted in pab2-1 ban (Fig. 1c). In pab3-1 ban, a single nucleotide deletion (T146∆) was detected in TT8, which resulted in a truncated tt8 protein that contained only 58 amino acids, making it considerably shorter than the wild-type TT8 protein, which comprises 518 amino acids (Fig. 1c).

Figure 1
figure 1

Characterization of pab2-1 ban and pab3-1 ban. Seed color at the immature (a) and mature (b) stages in ban, pab2-1 ban, and pab3-1 ban. (c) Schematic representation of mutations in the TT4 and TT8 regions for pab2-1 ban (top) and pab3-1 ban (bottom), respectively. Pale blue and black boxes represent exons and untranslated regions, respectively. Mutations in pab2-1 ban and pab3-1 ban are indicated in red. Green bars indicate the region used for the qPCR analysis. Gray and white boxes below the gene (labeled with a–i) indicate the amplified and non-amplified fragments, respectively, for the PCR performed using site-specific primers and template DNA from the pab2-1 ban and pab3-1 ban mutants.

Full size image

At the immature seed stage, the phenotypes of the pab2-1 ban and pab3-1 ban mutants were similar, but they obviously differed from the phenotype of the original ban mutant (Fig. 1a). However, there were clear differences in the pab2-1 ban and pab3-1 ban mature seeds, which were pale yellow and pale brown, respectively (Fig. 1b). The phenotypic features of the immature and mature seeds prompted us to develop a new mutation detection system for the M1 plants as described in the following section.

Development of a mutation detection system for mutagenized plants using double heterozygotes for TT4 and TT8

The cross-pollination of the pab2-1 ban and pab3-1 ban mutants generated plants with a double heterozygous genotype for TT4 and TT8 in the ban background (genotype: TT4/tt4 TT8/tt8 ban/ban). The double heterozygous plants had a phenotype that was similar to that of the ban single mutant (e.g., red immature seeds and dark brown mature seeds). A deleterious mutation in the wild-type allele of the heterozygous TT4 or TT8 locus will result in a non-functional anthocyanin biosynthetic pathway. A mutated cell is expected to divide and form a mutant sector in the double heterozygous plant. When the sector harboring the deleterious mutation in the TT4 or TT8 locus contributes to seed coat formation (i.e., the terminal tissue of the mutated somatic cells), the seeds will have an anthocyanin-less phenotype. The anthocyanin-less seeds were clearly distinguished from the non-mutated anthocyanin-containing seeds at the immature seed stage (Fig. 2). Additionally, after the remaining mutant siliques matured and dried, the anthocyanin-less seeds were pale yellow (tt4) or pale brown (tt8), making it possible to easily differentiate between these seeds (Fig. 2). Thus, mutations at two distinct loci (TT4 or TT8) were detectable on the basis of the seed color at the immature and mature stages.

Figure 2
figure 2

Strategy for detecting mutations at multiple loci in the M1 generation. The cross-pollination of pab2-1 ban and pab3-1 ban generated double heterozygous plants for TT4 and TT8. The double heterozygotes were irradiated with ion beams. When the wild-type allele at the heterozygous TT4 or TT8 locus was mutated by irradiation, anthocyanin-less immature seeds were produced by the tissues with the mutated tt4 or tt8 allele in the M1 plants. These seeds were easily discriminated from the red immature seeds produced by the tissue with non-mutated TT4 and TT8 alleles. The M1 plants producing anthocyanin-less immature seeds were grown until the seeds matured to determine whether their mature seeds were pale yellow (tt4-type) or pale brown (tt8-type).

Full size image

The double heterozygous seedlings were mutagenized by irradiating them with carbon ions. The survival curve of the wild-type seedlings indicated that the shoulder dose (i.e., the dose at which the survival rate decreased sharply from nearly 100%) was about 30 Gy at 1 day after germination (DAG) (Supplementary Fig. S1). We selected 15 Gy (i.e., half the shoulder dose) for the mutagenesis of the double heterozygous seedlings with carbon ion beams. The irradiated double heterozygous plants were grown until flowering and seed formation. Immature seeds were removed from the young siliques of the irradiated plants and then their pigmentation levels were investigated using a stereomicroscope. None of the approximately 400 non-irradiated double heterozygous plants produced anthocyanin-less immature seeds. Of the 2588 double heterozygous seedlings irradiated with carbon ion beams, 19 produced anthocyanin-less immature seeds. The 19 candidate mutants showed strong phenotypes with nearly complete lack of anthocyanins, and candidates showing subtle reduced anthocyanins were not isolated. The candidate mutants showed normal morphology in general, except for the anthocyanin-less feature. We then compared the candidate mutants with the parental pab2-1 ban (tt4 ban) and pab3-1 ban (tt8 ban) plants in terms of the mature seed colors. Among the 19 candidate mutants, two had pale yellow tt4-type mature seeds, whereas the other 17 had pale brown tt8-type mature seeds (Table 1, Supplementary Fig. S2). If TT4 and TT8 were considered to be equally mutated (i.e., a probability for each TT4 and TT8 mutation is 0.5), this unequal distribution of tt4 and tt8 was significant (p < 0.001; two-sided binomial test). On the other hand, including the 5′- and 3′-untranslated regions, TT8 comprises 4643 bp, which is about 2.6-times longer than TT4, which consists of 1789 bp. Taking the size difference into consideration (i.e., a probability for TT4 mutation is assumed as 0.28 [1789 (1789 + 4643)−1]), distribution of two tt4 and 17 tt8 mutants was not significant (p > 0.06; one-sided binomial test).

Table 1 Screening of anthocyanin-less seeds produced by the double heterozygous plants irradiated with carbon ion beams.
Full size table

PCR analyses of the anthocyanin-less seed coat-enriched DNA

To confirm the presence of the de novo mutation responsible for LOH at the TT4 and TT8 loci in the anthocyanin-less seeds, we first tried to isolate seed coat-specific DNA by separating all of the endosperm and embryo cells from the seed coat, because the seed color is derived from the seed coat tissues, but not the endosperm and embryo24. However, it was difficult to obtain sufficient amounts of seed coat-specific DNA to conduct a subsequent analysis. This was likely because of the loss of seed coat cells during the dissection of a limited number of mutant seeds. Therefore, we decided to use seed coat-enriched samples, instead of seed coat-specific samples. To verify the quality of the obtained seed coat-enriched DNA, immature hybrid seeds produced by adding pollen with the ban-4/ban-4 genotype (ban-4 means the fourth allele of ban28) to the stigma with the BAN/BAN genotype were used as the model materials. In the model hybrids, the seed coat cells had the maternal BAN/BAN genotype, whereas the embryo cells had a mixture of maternal and paternal genotypes (BAN/ban-4). The PCR amplification of the BAN and ban-4 fragments followed by the cleaved amplified polymorphic sequence (CAPS) analysis29 involving the restriction enzyme ScrFI, which digests the BAN allele, but not the ban-4 allele, generated maternal BAN bands, but not the paternal ban-4 band, for the seed coat DNA (Supplementary Fig. S3). This banding pattern was consistent with the predicted genotype and suggests that the obtained seed coat-enriched DNA was suitable for investigating the mutations responsible for LOH in anthocyanin-less immature seeds.

Four mutants with anthocyanin-less phenotypes were randomly selected at the immature seed stage for the dissection and examination of their seed coats. However, we were unable to determine whether these seeds were tt4- or tt8-type seeds. After the remaining seeds for the four selected mutants were allowed to mature and dry, three (123-1, 49-4, and 39-2) and one (43-7) mutants were revealed to produce tt8- and tt4-type seeds, respectively. Therefore, we expected that the mutations responsible for LOH would be found at the corresponding loci in the four mutants. We first examined the possibility of small mutations at the TT4 and TT8 loci by performing a site-specific PCR amplification and then analyzing the amplicon sequences. All four mutants produced amplification patterns for the TT4 and TT8 regions (e.g., Supplementary Fig. S4) that were similar to those of the control DNA (i.e., from non-irradiated double heterozygous plants). The sequencing of the amplified fragments did not detect any mutations responsible for LOH in TT4 and TT8. This was probably due to the contamination of the seed coat-enriched tissues with other non-mutated cells (e.g., endosperm cells), and this explanation was supported by the results on quantitative real-time PCR and dosage analyses (see below). We next addressed another possibility that deletions larger than the amplified fragment occurred at the TT4 or TT8 locus. Such a deletion would be difficult to detect during a site-specific PCR analysis. Accordingly, a quantitative real-time PCR (qPCR) analysis of the TT4 and TT8 regions was performed, with the ratio of the quantity of the TT4 fragments to that of the TT8 fragments normalized against the corresponding ratio for the non-irradiated double heterozygous plants. The relative TT4:TT8 fragment ratio would be less than 1 if a mutant lacked TT4, but contained TT8. Conversely, the relative TT4:TT8 fragment ratio would be greater than 1 if a mutant lacked TT8, but contained TT4. The qPCR analysis indicated that three mutants (123-1, 49-4, and 39-2, all of which exhibited the tt8-type phenotype) had a relative TT4:TT8 fragment ratio of approximately 1.5 or higher, whereas the other mutant (43-7, which exhibited the tt4-type phenotype) had a relative ratio of approximately 0.25 (Fig. 3). These results suggest that the three pale brown tt8-type mutants and the one pale yellow tt4-type mutant harbor deletion-type mutations in TT8 and TT4, respectively. These findings were in accordance with our phenotype-based discrimination.

Figure 3
figure 3

Quantification of the TT4 and TT8 fragments in DNA from anthocyanin-less seed coat-enriched tissues on the basis of a qPCR analysis. The TT4:TT8 fragment ratios normalized against the corresponding ratio in the non-irradiated double heterozygous plants (red dotted line) are provided for the four mutants. Data are presented as the mean value and standard error from three biological replicates.

Full size image

The target sequence used for the qPCR analysis of TT4 was present in the wild-type TT4 allele, but not in the pab2-1 allele (Fig. 1c). Nevertheless, some TT4 fragments were detected in the tt4-type mutant seed coat (43-7), reflecting the presence of wild-type TT4 fragments in the seed coat-enriched DNA. This was consistent with the detection of non-mutated wild-type TT4 DNA fragments during the site-specific PCR analysis of the mutant DNA (Supplementary Fig. S4) that might include small amount of other non-mutated cells like endosperm cells.

Dosage analysis of the anthocyanin-less seed coat-enriched DNA

To generate additional evidence of the large deletions at the heterozygous TT loci, genome-wide DNA fragment dosage variations were evaluated using the seed coat-enriched DNA. Short reads obtained by next-generation sequencing were mapped to the Arabidopsis reference genome, and the reads in non-overlapping 100-kb chromosomal bins were counted. The relative read depth (RRD) was calculated by dividing the average read depth for the respective bins by the median depth for five chromosomes in each line. When RRD was plotted to the corresponding chromosomal region, the large deletions were depicted as consecutive plots with RRD values that were lower than those in non-deleted regions that had a theoretical RRD of 1. For the TT8 locus on chromosome 4, one mutant (43-7) had an RRD value of approximately 1, whereas the other three mutants (123-1, 49-4, and 39-2) had an RRD value of approximately 0.65 (Fig. 4, top). In contrast, for the TT4 locus on chromosome 5, one mutant (43-7) had an RRD value of approximately 0.65, whereas the other three mutants (123-1, 49-4, and 39-2) had an RRD value of 1 (Fig. 4, bottom). Because each plot in Fig. 4 represents 100 kb, the regions with low RRD values appeared to be extremely large (i.e., Mb-order) in the four anthocyanin-less seed coats (1.5, 1.9, 2.6, and 1.5 Mb in 123-1, 49-4, 39-2, and 43-7, respectively). Hence, the three tt8-type mutants (123-1, 49-4, and 39-2) and the one tt4-type mutant (43-7) have de novo mutations responsible for the LOH at either the TT8 or TT4 locus. Moreover, all of these mutations were large deletions. This was consistent with the qPCR data as well as the phenotype-based discrimination (Fig. 3).

Figure 4
figure 4

Dosage variations on chromosomes 4 and 5 for the four anthocyanin-less seed coat-enriched DNA samples (123-1 (blue circles), 49-4 (orange circles), 39-2 (gray triangles), and 43-7 (gold crosses)). The approximate positions of the heterozygous TT8 and TT4 loci are indicated by pale orange bars on chromosomes 4 and 5, respectively.

Full size image

Discussion

To date, comparing mutation frequencies at multiple loci in the M1 genome has been challenging. We recently used the LOH of three anthocyanin biosynthesis-related genes to detect genome-wide mutations in gamma-irradiated M1 Arabidopsis leaves30. In this earlier study, an anthocyanin-less sector formed in the leaves of the irradiated M1 plants when one of the wild-type alleles at three heterozygous loci was mutated. However, unlike the current study, the mutated locus causing the anthocyanin-less phenotype was not identified on the basis of the observed phenotype alone because visible characteristics, including color intensity, were not distinguishable among the LOH of the three genes. In the present study, we focused on the differences in the colors of mature Arabidopsis seeds, which revealed that mutations in TT4 and TT8 can be clearly distinguished by simply examining the seed phenotype of irradiated double heterozygous M1 plants. The introduction of the ban mutation in the genetic background facilitated the screening for LOH at the immature seed stage (Fig. 2). This enabled the screening of seed color mutants from multiple inflorescence stems in a single M1 plant. The mutation frequency, which was calculated as the number of plants that produced anthocyanin-less seeds/total number of examined plants, was 0.7% (two tt genes as targets in this study). This was substantially higher than the 0.06% mutation frequency in the previous screening of anthocyanin-less stems of Arabidopsis M2 plants (seven tt genes as targets)11. Thus, the method used in the current study can effectively detect mutations. Additionally, screening at the immature seed stage allowed for the extraction of genomic DNA from fresh seed coats. The analyses of the seed coat-enriched DNA indicated that the four selected mutants have deletion-type mutations in either TT4 or TT8 (Figs. 3 and 4). The results were consistent with the phenotype-based predicted mutations, confirming the reliability of the mutation detection system developed in this study.

In the mutation detection system developed here, candidate mutants were screened based on the visible phenotypes in immature seeds (i.e., differences in pigmentation of the immature seeds) (Fig. 2). It should be noted that all candidate mutants isolated in this study showed nearly complete loss of anthocyanins in immature seeds, and candidates showing subtle reduced pigmentation were not isolated. It is theoretically possible to isolate mutants with weak phenotypes, because the immature seed does not contain any other pigments, and in fact, we previously isolated a mutant with pale pink to orange pigmentation from a mutagenized M2 population of ban plants25. Subsequent molecular analysis of the four randomly selected mutants unraveled that they harbored extremely large deletions (> 1.5 Mb) including heterozygous TT loci (Fig. 4). It is obvious that a large deletion has more chance to change a function of TT4 or TT8 than other types of mutations such as single base substitution, small indel, and chromosomal rearrangement whose breakpoints were located within an intragenic region13. Thus, it would be reasonable that large deletions are the majority of mutations responsible for LOH, although mutations other than large deletions might occur at the TT loci in the remaining candidate mutants. On the other hand, a combination of site-specific PCR, qPCR and dosage analysis used in this study would be difficult to identify the breakpoints of inversions and translocations within the intragenic region. To detect any types of mutations at the heterozygous TT loci including junctions of structural variations, whole-genome resequencing using seed coat-enriched DNA would be effective, as done in the previous report30.

The qPCR analysis revealed that all four analyzed mutants have large deletions in TT4 or TT8 (Fig. 3). The qPCR data were verified by the dosage analysis, which demonstrated that large deletions were responsible for LOH in all four examined mutants, and that all of the deletions at TT loci were extremely large (≥ 1.5 Mb) (Fig. 4). Consistent with this finding, we recently detected the LOH of three anthocyanin biosynthesis-related genes in gamma-irradiated Arabidopsis M1 leaves, and observed that the lack of anthocyanin in four of five leaves was due to extremely large deletions (0.5–5.0 Mb), whereas the anthocyanin-less phenotype of one leaf was caused by a small mutation at one of the three heterozygous loci30. In contrast to the mutations in the M1 genomes causing the anthocyanin-less phenotypes, large mutations (i.e., deletions) and small mutations (i.e., single base substitutions and small indels) are reportedly responsible for anthocyanin-less phenotypes at similar frequencies in Arabidopsis M3 mutants induced by carbon ion beams31. Focusing on the large deletions, 30 large deletions (≥ 149 bp) were detected by array comparative genomic hybridization in 13 Arabidopsis M3 mutants induced by argon and iron ion beams, but an extremely large deletion (> 1 Mb) was not found32. Similarly, out of 22 large (≥ 100 bp) deletions found in 16 Arabidopsis M3 mutants induced by argon and carbon ion beams, single extremely large deletion (> 1 Mb) was found33. Because extremely large deletions in the M1 genome will hardly be transmitted to the next generation13, the number of extremely large deletions affecting the phenotype will decrease in the M2 or later generations. Thus, substantial number of large deletions, which are induced more frequently by ionizing radiation than by chemical mutagens such as ethyl methanesulfonate2, have not been effectively detected during standard mutation analyses of genomes from the M2 or later generations, but they can be explored more thoroughly by analyzing the M1 genome. Considering it is useful for detecting a wide range of mutations30, an analysis of the M1 genome is also important for accurately determining the occurrence of mutations in plant genomes induced by ionizing radiation.

On the basis of the mutation detection system developed in this study, the tt4 phenotype (two plants) was 8.5-times less frequently observed than the tt8 phenotype (17 plants) following the carbon ion irradiation (Table 1). The unequal distribution may be associated with the size difference between TT4 and TT8, because a probability of two tt4 mutants out of 19 candidate mutants was not significant in the binomial test when mutation frequency per DNA length was assumed to be equal. However, all analyzed four mutants harbored extremely large deletions and most of the breakpoints of the deletions were located far from the TT genes (Fig. 4). This observation implies that the unequal distribution of the tt mutants may be caused by other reasons such as chromosomal location of the marker genes. Further analyses using different marker gene sets are necessary to give a solid explanation for the mutation frequency differences. In this study, because pab2-1 ban (tt4 ban) and pab3-1 ban (tt8 ban) were used to detect mutations, two TT genes that differed substantially in terms of size (TT4 and TT8) were compared regarding their mutation frequencies. In addition to tt4 and tt8, there are some other tt mutants with unique seed colors24. For example, tt9 seeds are tan-like or grayish34, whereas tt19 seeds at the ripening stage are pale brown, but the intensity of the coloration increases after a long-term desiccation35. The TT9 (5794 bp) and TT19 (1246 bp) genes are similar in size to TT8 and TT4, respectively. Artificial double mutants (i.e., ban and other tt mutants) that produce seeds with varying colors, making them distinguishable from one another, can be used to construct multiple heterozygous plants for detecting de novo mutations at their TT loci and for comparing their mutation frequencies in M1 plants. These analyses will help to clarify whether mutation frequencies at specific genomic loci can vary following an ionizing radiation treatment of plants.

Methods

Plant materials

The Arabidopsis ban-4/anthocyanin spotted testa mutant (Columbia (Col-0) accession), which was isolated by Tanaka et al. in our laboratory28, was used in this study. The ban-4 mutant, which has a 49-bp deletion in ANTHOCYANIDIN REDUCTASE, accumulates anthocyanins (i.e., compounds closely related to proanthocyanidins) in the immature seed coat. Therefore, ban has red immature seeds36,37. A mutagenized M2 population of the ban mutant25 was used for screening pab phenotypes. Another ban mutant, ban (Nö), which is produced by the insertion of the Ds transposon in the Nössen accession (pst13633)38,39 and was provided by RIKEN Genomic Sciences Center, was used for the mapping analysis of the newly isolated pab mutants. All of the Arabidopsis materials were grown in growth cabinets under standard conditions at 23 °C with a 16 h-light/8 h-dark cycle. Experimental research on plants including the collection of plant material was performed in accordance with relevant institutional, national, and international guidelines and legislation.

Screening and characterization of pab mutants

The coloration of immature seeds was investigated by opening the young siliques of M2 individuals under a stereomicroscope (Leica Microsystems, Wetzlar, Germany) at around 6–7 days after flowering. Two mutants (pab2-1 ban and pab3-1 ban) with pab phenotypes that were identified in the mutagenized M2 population of the ban plants were self-pollinated to obtain their M3 seeds. The rough-mapping analysis was conducted using the F2 populations of pab2-1 ban or pab3-1 ban and ban (Nö) to evaluate the linkage with the CAPS and simple sequence length polymorphism (SSLP) markers and determine the approximate position of the mutated pab genes29,40. The candidate genes responsible for the pab2-1 and pab3-1 phenotypes (i.e., TT4 and TT8, respectively) were characterized according to a PCR amplification using site-specific primers (Supplementary Table S1). The PCR (40 cycles) was completed using the AmpliTaq Gold 360 polymerase (Thermo Fisher Scientific, Waltham, MA). The amplified fragments for the TT8 region were analyzed by Sanger sequencing to identify small mutations responsible for the pab3-1 phenotype.

Irradiation of double heterozygous plants with carbon ions

Double heterozygous seeds (F1 seeds) were obtained by cross-pollinating pab2-1 ban and pab3-1 ban plants. The seeds were sown on Murashige and Skoog medium containing 2.5% sucrose in plates under sterile conditions. After a 4-day vernalization at 4 °C in darkness, the plates were transferred to a growth cabinet for an incubation at 23 °C under continuous light. Before the mutagenesis experiment, the sensitivity of wild-type Arabidopsis seedlings to different doses of accelerated carbon ions (17.3 MeV/u 12C5+) was assessed at Takasaki Ion Accelerators for Advanced Radiation Application (TIARA) by calculating the survival rates of plants irradiated at 1 DAG. Plants were considered to have survived the treatment if they had greenish true leaves at 4 weeks after irradiation. A survival curve was drawn according to the single-hit multitarget theory as previously described41. In this study, 15 Gy was used for the mutagenesis of the double heterozygous seedlings at 1 DAG. At around 5–7 days after the irradiation, the seedlings were transplanted into a soil mixture comprising vermiculite (Hakugen, Tokyo, Japan) and TM-2 (Takii Seed, Kyoto, Japan) (1:1). They were then incubated in a growth cabinet under standard conditions at 23 °C with a 16 h-light/8 h-dark cycle until flowering and seed formation.

Screening of anthocyanin-less seeds in the irradiated double heterozygous plants

At the immature seed stage, anthocyanin-less seeds were screened in the irradiated and non-irradiated populations of double heterozygous plants as described for the screening of the pab mutant. When anthocyanin-less seeds were not detected in three or more siliques on the first inflorescence stem, the basal region of the first stem was cut and the second and subsequent inflorescence stems were induced to form new siliques from different cell lineages. A total of eight or more siliques from at least three inflorescence stems per plant were analyzed in terms of the color of the immature seeds.

When anthocyanin-less immature seeds were detected in the detached young silique, the corresponding inflorescence stem was allowed to continue to grow to obtain mature seeds in the same lineage as the anthocyanin-less immature seeds, making it possible to compare the mature seed color with that of the parental pab2-1 ban and pab3-1 ban plants.

Binomial tests42 were performed to assess whether a probability for the observed distribution of two mutant types (tt4-type and tt8-type) was significantly different to an expected probability that mutation frequencies are assumed to be equal, or biased depending on the gene size, between TT4 and TT8.

PCR analysis using genomic DNA from the seed coat-enriched tissues of anthocyanin-less seeds

To examine the mutations responsible for the anthocyanin-less seed phenotype in the M1 generation, we collected the seed coats from the respective mutant seeds because the embryo and endosperm contain the paternal and maternal genotypes in the M2 generation, whereas the seed coat is composed of only the maternal genotype43. When anthocyanin-less immature seeds were detected in a detached young silique during the screening, the remaining fresh immature seeds in the same silique (about 5–15 seeds) were used in bulk to collect the seed coat cells because they were considered to have the same maternal genotype. The immature seeds were gently pushed on a glass slide with forceps to separate the seed coat from the embryo. The immature seed coats separated from the embryo were pooled in a distilled water droplet, washed gently several times to remove endosperm cells, and transferred to a mortar. All of the dissection steps were completed with samples maintained on ice. The seed coat-enriched tissues were ground in liquid nitrogen using a pestle. Genomic DNA was then extracted from the ground material using the DNeasy Plant Mini kit (Qiagen, Hilden, Germany).

The genomic DNA extracted from the seed coat-enriched tissues served as the template for the site-specific PCR amplification of the TT4 and TT8 regions using the same primers as those used for the pab2-1 and pab3-1 analyses. The amplicons were purified using the QIAquick PCR Purification kit (Qiagen) and then analyzed by Sanger sequencing.

The TT4 and TT8 DNA fragments in the seed coat-enriched DNA were quantified by a qPCR analysis, which was performed using the KAPA SYBR Fast qPCR kit (Kapa Biosystems, Wilmington, MA) and the CFX96 Real-Time System (Bio-Rad, Hercules, CA). A melting curve analysis was performed to confirm the specificity of the amplification by the primers listed in Table S1. The TT4 and TT8 DNA fragments were quantified according to the comparative Ct method44 and then expressed as the TT4:TT8 ratio, which was normalized against the corresponding ratio in the non-irradiated control plants.

Dosage analysis

Using 1–5 ng seed coat-enriched DNA, the next-generation sequencing libraries were constructed using the KAPA HyperPlus Library Preparation Kit (Kapa Biosystems) and the IDT for Illumina unique dual indexes (Illumina, San Diego, CA). The pooled libraries were sequenced on the Illumina NovaSeq 6000 and NextSeq 500 platforms, with PE150 chemistry. Low-quality reads and adapter sequences were removed from the raw data using Illumiprocessor (version 2.0.9; https://illumiprocessor.readthedocs.io/en/latest/). The remaining clean reads were mapped to the Arabidopsis reference genome (TAIR10.27; https://www.arabidopsis.org/) using the Burrows–Wheeler Aligner (version 0.7.5; http://bio-bwa.sourceforge.net/), SAMtools (version 1.3.1; http://samtools.sourceforge.net/), and Picard-tools (version 1.119; https://broadinstitute.github.Io/picard/). The obtained BAM files were used for the dosage analysis as previously described45, with minor modifications. The bin size was set as 100 kb, and the average read depth in the respective bins was determined using the goleft depth algorithm (https://github.com/brentp/goleft). The read depth in each bin was normalized against the median read depth for all five chromosomes in each line. The resulting RRD values in each bin were plotted to the corresponding chromosomal region.