Knock-out mutants from an En-1 mutagenized Arabidopsis thaliana population generate phenylpropanoid biosynthesis phenotypes

Basic molecular biology techniques were applied according to Sambrook et al. (20). Radiolabeled probes were derived from gel-purified DNA fragments as described for AtCHS and UBQ (17). The Col FLS probe was derived from a full-length FLS cDNA (U84259) isolated from a λZAPII whole plant cDNA library by using a probe derived from EST T76883. The FLS transcriptional start site was determined by 5′ rapid amplification of cDNA ends as described (21). The F3H transcriptional start site was determined by using premixed reagents (5′/3′ rapid amplification of cDNA ends kit, Boehringer Mannheim) according to the manufacturer’s recommendations. Sequences of plasmid DNA and PCR fragments were determined on Applied Biosystems 377 sequencers by using dye-terminator chemistry. Oligonucleotides were purchased from MWG (Ebersbach, Germany). Thin layer chromatography (TLC) standards were obtained from Sigma and Roth (Karlsruhe, Germany).

† The En-1 mutagenized A. thaliana poplulation of 3,000 lines (22) was extended by 5,000 lines to a total of 8,000 lines carrying ≈48,000 independent En-1 insertions. The new lines were created in a similar way as the original 3,000 lines. First, 28 lines derived from parent G11 (23) and 3 lines derived from parent G69, all carrying 8 to 16 En-1 insertions, were selected. The 5,000 new lines consisted of 150 progeny of the 31 selected plants propagated for 6 additional single-seed descent generations and 400 of the original single-seed descent lines (23) in generation 12. Finally, leaf material for DNA preparations and seeds were sampled from all lines. Plant genomic DNA was prepared as described (22).

To identify mutations in the initial population, 2,000 S₆ families of 20 individuals each were analyzed on soil for segregation of recessive mutant phenotypes, and 20 to 40 S₆ seedlings from 3,000 lines were screened on MS agar (50 mg/liter ampicillin, no sucrose; 16 h fluorescent light per day). The mutant traits scored on agar were germination, seedling growth, pigmentation, root length, and seed size. Traits scored on soil were flowering time, plant size, plant and leaf shape, trichome morphology, epicuticular wax, flower morphology, and fertility.

PCR-based screens were performed according to a three dimensional grid as described (22). The following primers specific for the respective Col allele were used: oUH8 (5′-GAATCCGTCGCTAAACCGGAAGATTCG-3′) and oUH10 (5′-GGAGGATTATCATCTCCGGTTAGTTCCG-3′) for FLS, oSA4 (5′-AGGAACTTTGACTGAGCTAGCCGGAGAGTC-3′) and oSA7 (5′-GAGGCGAGCAAGCTCCAAATCTCTTCCC-3′) for F3H.

For dark adaptation, 8-week-old plants were transferred to complete darkness for a total of 64 h. Light induction experiments were carried out by irradiating dark-adapted plants for 24 h with continuous white light (≈150 microeinstein/m²⋅s) in a phytochamber. Immediately after harvesting, leaves were frozen in liquid nitrogen and were stored until use at −80°C. Total RNA was isolated as described (21). tt6 was obtained from Nottingham Anabidopsis Stock Centre (stock number NW87).

DNA fragments flanking En-1 insertions were mapped by using the Lister and Dean (24) recombinant inbred lines (RILs) or on the Landsberg erecta (Ler) × Cape Verde Islands RIL map developed recently (25). Phenotypic markers were used in allelism tests to assign the new mutant phenotypes to previously mapped mutants. Molecular and phenotypic markers were combined in one map by using classical map landmarks, although this procedure causes some inaccuracy because of the restricted number of classical visible markers included in the RIL maps.

Sequences of the various mutants were verified by direct sequencing of PCR fragments. Initially, Col F3H genomic DNA fragments were synthesized by using primers designed according to published Ler sequences (26). For the Col F3H sequence, see GenBank accession no. AF064064. Genomic sequences flanking En-1 insertions were amplified with an En-1-specific primer (En205 or En8130 for 5′ and 3′ ends, respectively; ref. 22) and the respective gene specific primer.

A. thaliana genomic DNA was isolated as described (27). PCR conditions, primers, and polymorphic sites were as follows: 50 μl total volume, ≈20 ng genomic DNA, 20 pM of each primer, 1 unit Taq Pol, and buffer (Boehringer Mannheim); cycle program: 5 min, 94°C; 40× (1 min, 94°C; 1 min, 55°C; and 1 min, 72°C); and 10 min, 72°C; FLS primers oUH5 (5′-GAATCCCTAATAACGTCTCCG-3′) and oUH1982 (5′-TTACATATCCGCCATTGTTTCCGGC-3′) and F3H primers oSA4 and oSA21 (5′-CACTTTCACCCATCCTTCAGGCTTATTTG-3′); restriction polymorphisms: BssHII site was present only in Col FLS, and BclI site was present only in Ler F3H.

The FLS ORF, modified at the start and stop codons by PCR, was inserted into NcoI/BamHI-digested pQE60 (Qiagen, Hilden, Germany), resulting in the plasmid pQE60-FLS. Escherichia coli strain SG13009∷pUBS520 (28) was used for protein production as described (29). As a negative control, cells containing pQE60 were included. Bacteria from a 50-ml culture were resuspended in 3 ml Hepes buffer (0.2 M Hepes/KOH, pH 7.0/1 mM EDTA/200 μg/ml lysozyme) and were incubated on ice for 30 min with slight mixing every 5 min. The suspension was centrifuged (13,000 × g, 10 min, 4°C), and the supernatant (crude extract) was used for FLS activity assays. A 200-μl reaction mixture contained 3,000 dpm ¹⁴C-labeled dihydrokaempferol, 10 μl crude extract, 5 mM ascorbate, 50 μM FeSO₄, 0.25 mM 2-oxoglutarate, and 50 mM Hepes/KOH (pH 7.0) and was incubated at 37°C for 30 min (30). Flavonoids were extracted twice with ethyl acetate and were chromatographed on cellulose plates (Merck) by using chloroform/acetic acid/water (10:9:1) as solvent. Unlabeled dihydrokaempferol, kaempferol, and naringenin were co-chromatographed as standards and were visualized under UV light. The radiolabeled substances were detected by radioimaging (Fuji Fujix BAS1000).

Leaves were incubated in 2 M HCl for 16 h at RT and subsequently for 20 min at 100°C. Flavonoid aglycons were extracted with isoamyl alcohol, were developed on cellulose TLC plates (Merck) with Forestal [water/acetic acid/HCl (32%) 10:30:1] or chloroform/acetic acid/water (10:9:1), and were visualized under 366-nm UV light. Standards were kaempferol and quercetin.

A population of 8,000 A. thaliana lines containing, on average, six independent En-1 insertions (48,000 in total) was generated by single-seed descent. The number of insertions per line varied between 1 and 20. In a sub-population of 3,000 lines, 153 mutant phenotypes (5%) were detected in an initial screen (Table 1). Of the mutations, ≈70% were genetically unstable, indicating the presence of En-1 at the mutant locus. The data show that En-1 is mobile in A. thaliana and causes a broad spectrum of mutant phenotypes.

The localization of En-1 insertions was determined by using three different strategies. First, mutants with known phenotypes were crossed with lines containing the corresponding mapped mutation to test for allelism. Second, A. thaliana sequences next to the En-1 insertions were isolated by PCR and were mapped by using RILs (loci named “B plus number” in Fig. 1). Third, map positions were derived from genes containing En-1 insertions (see below) whose locations were known. Fig. 1 shows that En-1 insertions occurred in all chromosomes. The detection of several independent GL2 and PIN mutants indicates a high number of insertions at the lower end of chromosome I. Despite these clusters, the data indicate a wide distribution of insertion sites over the entire A. thaliana genome.

The En-1-mutagenized population was screened for knock-out (null) alleles at 109 loci, and En-1 insertions were recovered for 55 genes (Fig. 1). In 50% of the cases, two or more different alleles were found. Insertions were identified by using a three-dimensional, PCR-based screening strategy and were confirmed in individual progeny of the selected plant. For final proof, the sequence of the diagnostic PCR product was determined. Of all insertions studied so far, 37% were located outside the genes, 40% in exons, 15% in introns, and 8% in exon/intron border regions.

The 37% insertions located close to, but not inside, the gene did not result in gene disruption. To isolate a knock-out allele, the ability of En-1 to transpose to linked sites was exploited. The success of this approach was exemplified for four genes: AtKC1 (potassium channel 1, U73325), AtKAT2 (voltage-gated potassium channel, U25694), AtKCO1 (outward rectifying potassium channel, X97323), and a PIN homologue (AtPIN2). From each of the respective homozygous plants, ≈2,000 progeny were analyzed by the three-dimensional, PCR-based screening strategy. In each case, an insertion located in exonic sequences was recovered while the original insertion, located 1 to 3 kilobases distant from the new insertion site, had disappeared. We conclude that En-1 is a powerful tool for gene disruption in A. thaliana.

Using the reverse genetic screen described above, we isolated a putative knock-out mutant (line P90) for the gene corresponding to EST T76883. This EST displays 63% amino acid sequence similarity to the functionally identified P. hybrida FLS and corresponds to a putative AtFLS gene (19). The original P90 plant was homozygous for the insertion; all 71 offspring tested contained the insertional allele. En-1 was inserted in the second intron of the gene (Fig. 2A), and the corresponding transcript was barely detectable even under light conditions inducing strong transcript accumulation in the wild type (Fig. 2B). The homozygous mutant was impaired strongly in its ability to produce the flavonol kaempferol (Fig. 2C). These results indicate that a loss-of-function mutant for AtFLS was indeed obtained.

Because the En-1-mutagenised population consists of multicopy lines, the causal relationship between an observed phenotype of the recessive mutant and a particular insertion has to be established. Phenotypic reversion to wild type on excision of En-1 from the locus provides this proof. Five phenotypic revertants were identified among the offspring (S₁) of the original homozygous P90 plant. All were heterozygous at the FLS locus; three of the revertant alleles contained footprints, and two contained the wild-type sequence (Fig. 2A). Fig. 2C shows the phenotypic reversion of plant 14. DNA gel-blot analysis of offspring from plant 14 revealed a segregation ratio of 9:13:5 (FLS/FLS: fls-1∷En/FLS: fls-1∷En/fls-1∷En), similar to the expected 1:2:1 ratio. All analyzed S₁ plants, including the revertants and offspring of plant 14, were shown to be descendants of the originally isolated mutant plant; that is, they displayed the P90-specific En-1 insertion pattern (data not shown). These data demonstrate a causal relationship between the presence of En-1 at the FLS locus and the flavonol-deficient phenotype of line P90.

Finally, the activity of the enzyme encoded by the FLS locus was verified biochemically. By using a full-length cDNA, the enzyme was expressed in E. coli. The reaction catalyzed by FLS is illustrated in Fig. 3A. The recombinant protein displayed FLS activity (Fig. 3B), assigning FLS function to the gene defined by EST T76883.

Results from a cross between tt6 and P90 demonstrated that the previous assumption that the TT6 locus encodes FLS was no longer tenable: full complementation was observed; that is, the F₁ plants displayed wild-type phenotype in terms of both seed color and flavonol content. In addition, analysis of the F₂ from a cross between tt6 and Col confirmed independent segregation of the tt phenotype and a FLS-specific cleared amplified polymorphic system marker.

In a PCR-based reverse genetic screen for En-1 insertions in the F3H gene, three plants were detected that produced offspring with tt (yellow) seeds. Additional f3h lines were detected by screening for tt seeds (forward screens) and subsequent molecular analysis of the F3H locus. The sequences of two En-1 insertion alleles and derived footprint alleles as well as an additional footprint allele are depicted in Fig. 4. TLC analyses of the f3h knock-out lines detected no flavonoids in these plants, which is consistent with the complete loss of testa pigments. Homozygous lines carrying these f3h alleles did not complement the tt6 mutation in test crosses, and, in the above-mentioned F₂ from a tt6 × Col cross, the pale brown tt6 phenotype co-segregated with the f3h locus (Table 2). Finally, the f3h allele from tt6 was sequenced, and a single point mutation in the ORF was detected, which caused premature termination of translation (Fig. 4). Taken together, these data prove that the TT6 locus encodes F3H and thus identify the last missing mutant for the early steps of flavonoid biosynthesis.

We isolated an A. thaliana line carrying an insertion of En-1 at the FLS locus (allele fls-1∷En). The mutant displays a phenotype that proves the in vivo function of FLS. En-1 integrated in the second intron of the FLS gene and caused a characteristic three-bp target site duplication (10, 31). Although the FLS ORF and the conserved 5′ and 3′ splice site sequences of the intron were not affected, almost no FLS mRNA was detectable. Thus, integration of the 8.3-kilobase En-1 into an intron was apparently sufficient to cause essentially a null allele. Absence of mRNA could be caused by premature termination of transcription or by defects in mRNA maturation. The very low amount of mRNA detected can be attributed to sectors derived from somatic reversion events. The detection of a wild type-sized restriction fragment by DNA gel-blot analysis and the residual amounts of kaempferol found in plants homozygous for the fls-1∷En allele support the assumption of such somatic reversion events. The almost unaltered quercetin level in the fls mutant may indicate the presence of another, related enzyme with higher affinity for dihydroquercetin than for dihydrokaempferol as substrate. Previous low-stringency DNA gel-blot experiments also provided evidence for a FLS-related gene in the A. thaliana genome (19).

Seeds of fls plants show a brown color indistinguishable from that of wild-type seeds, suggesting that FLS is not required for pigment production in the testa. Condensed tannins and anthocyanins probably serve as testa pigments (Fig. 3A). The wild type-like seed color of fls plants also explains why this locus was not among the tt mutants (14, 32).

Mapping data suggested that TT6 does not encode FLS. tt6 was placed to the bottom of chromosome III (14) whereas FLS maps at the top of chromosome V (marker U1 at 23 centimorgans of the RIL map, and ref. 19). TLC analyses of tt6 leaves revealed a reduction of both FLS-derived flavonols and dihydroflavanone reductase-derived anthocyanidins (data not shown). Similar reductions have been reported for other organs of tt6 plants (15, 33). These data suggested that, in tt6, an early enzyme of the flavonoid pathway (Fig. 3A) was affected. Mutants in CHS (tt4) and chalcone-flavanone isomerase (CFI, tt5) already have been established. F3H, encoding the third enzyme of the pathway, was, like tt6, mapped to the bottom of chromosome III (marker F3H at 68 centimorgans of the RIL map, and ref. 26). Therefore, F3H was a strong candidate for the gene mutated in tt6. Genetic complementation analysis and sequencing of the tt6 f3h allele yielded final proof that the TT6 locus encodes F3H.

The different En-1-derived f3h alleles (Fig. 4) were independent isolates because the respective lines displayed unrelated En-1 insertion patterns in DNA gel-blot analyses (data not shown). All of these f3h alleles caused a tt phenotype very similar to that of tt4. Such a phenotype would be expected for a null mutation in a single-copy gene encoding an early step in the nonbranched part of flavonoid biosynthesis. The severe tt phenotype of the f3h-4f and f3h-5f alleles suggests that the frameshift mutations in the second F3H exon lead to completely nonfunctional F3H proteins whereas the point mutation in tt6 might be leaky, allowing the production of small amounts of functional enzyme. This interpretation is supported by the results from the tt6 × Col cross, which do not indicate the presence of an additional locus providing F3H activity. Because the premature stop codon in the f3h–1(tt6) allele causes a shorter protein than two of the footprints, the leakiness might be caused by some sort of suppression of the UAG stop codon. We also observed a slightly reduced level of f3h mRNA in light-induced tt6 in comparison with wild type (data not shown), a feature that can be explained by mRNA destabilization by premature stop codons (34).

The two classes of allelic mutations isolated from the En-1 population complete the set of A. thaliana mutants in the first part of phenylpropanoid metabolism. Mutants defective in the other early enzymes, including dihydroflavanone reductase (tt3), have been known for some time (35, 36). In the f3h lines, the last enzymatic step before the branch points toward either anthocyanidins or flavonols is blocked. As might have been expected from the biochemistry of flavonoid biosynthesis, this mutant was among the tt collection, although the leakiness of the tt6 allele caused initial problems with the correct assignment of the mutant. The fls-1∷En allele now provides access to a mutant specific for the flavonol branch of the pathway. This mutant will be important for determining the exact contribution of flavonols to the shield of UV protective compounds (37, 38).

Our objective was to develop a simple system that allows the study of the function of principally every gene of A. thaliana. The success rate of 50% for recovery of En-1 insertions in the genes analyzed was expected on the basis of the calculated number of insertions in the population, indicating that the sensitivity of the applied PCR screening technique was sufficient. The whole population originated from five plants selected to contain En-1 but not the original T-DNA. It is therefore likely that the observed clusters reflect the insertion sites already present in the first generation, which were conserved by transposition to linked sites (39). We estimate that, during setup of the population, on average 10% of the En-1 copies transposed per generation, in close agreement with the previously calculated excision frequency of 8% (10). However, the transposition activity varied in an unpredictable manner from generation to generation. As a consequence, several single-seed descent generations were required to generate new insertions, although there was no need for large-scale crosses, selections, or numerous transformations as required for two-element transposon systems or T-DNA insertion mutagenesis.

No obvious structural prerequisites for specific En-1 insertions were detected. En-1 inserted in introns, exons, and promoter regions and downstream of coding regions in both orientations relative to the target gene. The only exceptions from the otherwise random insertions obtained were the two independent f3h alleles containing En-1 insertions at the same site, though with opposite orientation. More data are required to determine whether independent transpositions to the very same site and reversion by leaving the identical footprint occurred just by chance or as a result of sequence-dependent insertion/excision.

The moderate activity of En-1 seems to ensure that unstable mutants can be analyzed rather easily. On the other hand, En-1 mobility is sufficiently high to facilitate the establishment of causal relationships between the phenotype and the affected gene through the isolation of wild-type revertants (germinal reversions). However, a complication associated with autonomous transposable elements is somatic reversion. This feature may be extremely useful for studying, for example, cell autonomy, although it interferes with the molecular and biochemical analysis of mutants because chimeric plants have to be studied. The isolation of stable footprint alleles as described here for F3H overcomes this problem. To isolate stable alleles of a given tagged locus, a PCR-based screen for the absence of En-1 should be performed with heterozygous plants generated by crosses with Col wild type. The apparent tendency of En-1 to transpose to linked sites has been used to generate insertions in exons by using lines in which En-1 was located close to the gene and can also be used for local saturation. These features of the En-1 mutagenized population are a clear advantage when compared with T-DNA-derived insertion mutant collections. The En-1-mutagenized A. thaliana population is therefore likely to serve as a useful tool in future efforts to assign function to the increasing number of putative genes detected during sequencing of the A. thaliana genome.