The superroot2 suppressor mutant designated 494 was identified in an EMS-mutagenized homozygous sur2-1gl1 population [17]. Its adventitious rooting-related phenotypic characterization and coarse genetic mapping have been recently described [17]. Briefly, genetic mapping was accomplished using 160 phenotyped mutant plants collected from a F₂ population derived from a cross between 494 (Wassilewskija 4, Ws-4 background) and atr4-1, an allele of sur2 in Columbia-0 (Col-0) background [26]. The mapping strategy and the molecular markers used have been previously described [5].

Approximately 1 g samples were collected from pools of 25 homozygous sur2-1gl1, homozygous 494 (F₃ generation after the 4^th backcross) and homozygous 2035 (F₃ generation after the 2^nd backcross) mutant plants, respectively, and bulked. DNA was extracted from each sample as described by Hanania et al. [27] and eluted in 200 µl of water. The concentration and quality of the DNA were determined using a NanoDrop 2000 spectrophotometer (Thermo Scientific) then the samples were freeze-dried for shipping. Extracts from each line containing more than 5 µg of high-quality DNA, meeting the quality requirements for sequencing, were sequenced at Beijing Genomics Institute Shenzhen (BGI) using the Illumina HiSeq 2000 high-throughput sequencing system. The sequencing conditions, data processing procedures and alignment results are summarized in Tables S1 and S2 in File S1. DNA libraries covering 4.29 giga base pairs (Gb), 4.54 Gb and 4.29 Gb (clean data) were generated for sur2-1gl1, 494 and 2035, respectively. Very high quality data were obtained from all samples, with Q20 values (error rate <1%) of 96.88 for sur2-1gl1, 96.2 for 494 and 96.8 for 2035, respectively. The deep-sequencing datasets have been submitted to the European Nucleotide Archive (ENA), and are available through the following link: https://www.ebi.ac.uk/ena (accession ID: ERA296179).

The Col-0 (TAIR10) sequence was used as reference genome, while both sur2-1gl1 and 494 are in Wassilewskija-4 (Ws-4, N5390) background. Consequently, to construct a Ws-4 reference genome for the mutant 494, sequencing reads of the parental genotype sur2-1gl1 were first aligned against the reference genome sequence (TAIR10) using SOAP2 (http://soap.genomics.org.cn/soapaligner.html), and SNPs of the sequenced genome were detected using SOAPsnp (http://soap.genomics.org.cn/soapsnp.html). Using SNPs from the sur2-1gl1 sequence the corresponding sites in the TAIR10 were replaced, and the newly constructed sur2-1gl1 genome was subsequently used as reference for 494. To identify the 494-specific mutation, credible SNPs (which are likely mutations between 494 and sur2-1gl1) were filtered from credible loci differing between 494 and sur2-1gl1-TAIR10. These are relatively reliable loci filtered using the following criteria: consensus quality ⩾20 (error rate <1%), total depth ⩾5 and ⩽50 (to avoid copy number variation, CNV), and estimate copy number of the site <2. Only homozygous 494 SNPs were considered further. We used the same approach to identify the causal mutation in the mutant 2035 (data not shown).

The putatively causal mutation identified in the mapped region of 494 by analyzing the NGS dataset was subsequently re-confirmed by Sanger sequencing. To confirm the splicing defect and intron retention in the 494 mRNA, total RNA was extracted from seedlings of both 494 and sur2-1gl1 mutants using the RNAquous isolation kit (Ambion), treated with rDNase I, using the DNA-free Kit (Ambion), and subsequently cDNA was synthesized using the iScript cDNA Synthesis Kit (BIO-RAD), according to the manufacturers instructions. Furthermore, gene-specific primers (At4g02570.F_TGGCTATCCCGCTTCTTCTA and At4g02570.R_TTGCAAACACAACCAGCAAT) spanning the splicing site altered by the 494 mutation were used to amplify the cDNAs, using standard PCR procedures.

To genotype the 494 point mutation, new derived cleaved-amplified polymorphic sequence (dCAPS) primers (494+EcoNI.F_CTTGCCCTGATTACCTGTTGAA and 494+EcoNI.R_ GCCACTCTCTCCCTCTCCTT) were designed using dCAPS Finder 2.0 software (http://helix.wustl.edu/dcaps/dcaps.html) [28]. One mismatch (underlined) was introduced in the F primer to incorporate a restriction site in the PCR product of one allele. After amplification, the PCR products were digested with EcoNI (Fermentas Fast Digest) following the manufacturer’s recommendations and electrophoretically separated on a 4% agarose gel. The wild type yielded two fragments of 133 and 16 bp, respectively, while the 494 allele gave one band of 149 bp.

To genotype sur2-1, which carries a 61 bp insertion in the cytochrome P450 CYP83B1 gene [29], primers were designed (Sur2-1F_AGCTTGGTTTCGGACAGTACAC and Sur2-1R_ACTTAGATCAACGGTGCCTGAT) that amplify a 237 bp fragment in sur2-1, and a 176 bp fragment in wild type, respectively.

To genotype the axr6-3 and cul1-7 alleles used for complementation tests, we used the genotyping primers and conditions described by Gilkerson et al. [24].

Mutants and corresponding wild-type plants cultivated in vitro were characterized and their auxin contents were quantified as previously described [17]. For phenotypic evaluation of soil-grown plants, seeds were first germinated in vitro then the resulting seedlings were transferred into pots, which were placed in growth chambers providing short day (8 h darkness/16 h light) conditions at 22°C/18°C (light/dark temperatures). Plants under all growth conditions were visually inspected at time-points corresponding to selected developmental stages.

Genetic and physiological studies have shown that adventitious root (AR) formation is a heritable quantitative genetic trait controlled by multiple endogenous and environmental factors (reviewed in: [30]–[33]). Little is known about the molecular mechanisms controlling this developmental process, but we have recently started to unveil the complex regulatory mechanisms controlling AR formation. Using Arabidopsis thaliana as a model, we have shown that auxin and light signaling play essential roles in regulating AR formation on Arabidopsis hypocotyls [34]–[36]. We have also obtained preliminary indications that different regulatory pathways control lateral root and AR initiation in the hypocotyl, although both types of roots originate from pericycle cells [37], [38].

To substantiate these findings we screened seedlings obtained from EMS-mutagenized homozygous superroot2-1glabra1 (sur2-1gl1) seeds, aiming to identify Arabidopsis mutants that produce a nearly normal main root system but have specific impairment in AR formation on etiolated hypocotyls [17]. The mutant designated 494 was identified during that screen. The mutation it carries suppresses the AR phenotype of the superroot2-1 (sur2-1) mutant [39], and the suppressor mutant develops significantly fewer AR on the hypocotyl than sur2-1gl1, despite retaining similarly high endogenous IAA levels [17].

Homozygous 494 mutant plants selected from a homozygous M₃ population were backcrossed four times with the sur2-1gl1 parental genotype to remove EMS-induced SNPs not associated with the phenotype. In parallel, homozygous 494 M₃ mutant plants were outcrossed with the atr4-1 mutant, carrying an allele of the sur2 mutant in Col-0 background [26], to generate a mapping population. Lastly, to identify potential alleles, allelic tests were conducted by crossing the mutant with other mutants isolated in the screen.

The 494 mutation was mapped on genomic DNA extracted from phenotyped mutant seedlings that produced fewer AR than sur2-1gl1, identified in a segregating F₂ mapping population grown in vitro as previously described [5]. Coarse mapping was completed using newly identified INsertions/DELetions (INDEL) markers, as described in Materials and Methods. A flowchart of the NGS-assisted mapping approach, which can be applied for identifying mutations in both reference and non-reference Arabidopsis accession backgrounds, is shown in Figure 1.

Even after applying the filtering regime described in Materials and Methods we identified 33 mutations scattered across the five chromosomes by comparing the 494 and newly constructed sur2-1gl1 reference sequences, 25 of which were canonical C:G-to-T:A EMS-induced changes (Table S3 in File S1). Twenty-one of these mutations are situated in non-coding sequences (one in a 3′UTR, one in a 5′UTR, two in transposons, four in introns, and 13 in intergenic regions). Of the 12 mutations affecting the CDS, only six were non-synonymous. Interestingly, the only mutation located in the genomic interval 1.062.516 bp - 2.821.733 bp defined by the mapping markers UPSC_4-1062 and UPSC_4-2821 in the top of chromosome 4, is a synonymous G-to-A substitution at position 1.130.414 bp (Figure 2A). That mutation, located at the junction between the 5^th exon and the 6^th intron of the CUL1 gene (AT4G02570), affects the splicing efficiency of the 6^th intron, as confirmed by comparing the PCR amplification products of the sur2-1gl1 and 494 cDNAs (Figure 2B). Gene-specific primers spanning the splicing site amplified a 463 bp fragment in sur2-1gl1, while in addition to the correct spliced variant a larger 551 bp amplicon was detected in 494 (Figure 2B). The size difference of 88 bp between the 494 amplicons corresponds to the size of the 6^th intron. These observations show that although correct splicing occurs in 494 and some wild-type protein is produced, (probably less efficiently than in sur2-1gl1), the 494 point mutation leads to intron retention in the 494 mRNA, translation of which is predicted to yield a truncated protein product.

EMS mutagenesis typically induces hundreds of randomly distributed mutations per genome [3], [15] that can hamper the direct identification of causal mutations when using NGS data alone. Performing sufficient allelic tests (using existing alleles or T-DNA insertion lines), or complementing all of the candidates by transforming mutants with corresponding wild-type sequences would clearly be extremely tedious and generally unfeasible. The number of candidate mutations can be reduced by repeated backcrossing to the parental genotype before sequencing or by directly sequencing genomes carrying two or more independent alleles isolated in the same screen, followed by a search of unique mutations located in the same gene in multiple genomes. The first (repeated backcrossing) option is time-consuming, even if the focal organism has short generation times like Arabidopsis, and not always very efficient for removing EMS-induced SNPs that are not associated with the phenotype of interest. However, backcrossing is not essential for identifying EMS-induced mutations through NGS-assisted mapping, nor is the number of backcrosses important, since the strategy yielded very similar results when applied to mutants backcrossed four or two times (Table S3 and Table S4 in File S1; [40]). In the example discussed here, a high number of potential candidate mutations were identified after filtering, even if mutant plants were backcrossed four times before sequencing. Second, a typical non-saturating EMS mutagenesis screen provides the likelihood of detecting only single alleles [41], therefore direct sequencing of the mutant will have to be supported by mapping [12], [42]. Furthermore, in our particular case we found that mapping-by-sequencing is less suitable for identifying mutations with a weak phenotypic penetrance because it is difficult or even sometime impossible to trace mutant seedlings correctly in pooled F₂ recombinants, or contaminants that occasionally occur [5]. Thus, sequencing such pools could have generated misleading or truncated information as described by [40] and sequencing a pool of homozygous F₃ mutants turned to be a more reliable alternative.

Confirmation by Sanger sequencing of the point mutation identified in the CUL1 gene (AT4G02570), the only mutation situated in the mapped region, prompted us to consider it as a strong candidate for the 494 suppressor phenotype and we named this new allele cul1-494.

To demonstrate unambiguously that the cul1-494 mutation confers the suppressor phenotype we conducted complementation tests using three viable Col-0 alleles axr6-3 [23], cul1-6 [18], and cul1-7 [24]. They all failed to complement cul1-494, confirming that cul1-494 is a new viable allele of CUL1. Moreover, both axr6-3sur2-1 and cul1-7sur2-1 double mutants produce fewer AR than sur2-1 (Figure 3), confirming that the axr6-3 and cul1-7 mutations suppress the AR phenotype of sur2-1 in a similar way to cul1-494.

Phenotypically, cul1-494 mutant differ in several respects from those carrying other known viable CUL1 mutant alleles, notably its developmental defects are less pronounced. The first allele, axr6-3 was isolated in a screen for tir1-1 enhancers designed to identify genes required for SCF^TIR1-mediated auxin responses [23]. axr6-3 plants exhibit impaired auxin responses, reduced apical dominance, delayed senescence, reduced male fertility, and aberrant flower development. In addition, this allele is temperature-sensitive, mutant plants being sterile at 22°C, but fertile at 18°C. The axr6-3 point mutation, located at the N terminus of the protein, interferes with Aux/IAA protein degradation and prevents the assembly of SCF^TIR1 complexes by disrupting ASK1 binding. cul1-6, a recessive viable allele that affects interaction with the SCF regulatory protein CAND1 (CULLIN ASSOCIATED AND NEDDYLATION DISSOCIATED), was isolated in a screen for mutants resistant to sirtinol [18]. cul1-6 plants have defects in seedling and adult morphology, including delayed leaf emergence, retarded root growth, reduced apical dominance, curled leaves and altered floral morphology. In addition to reduced auxin sensitivity, cul1-6 seedlings display reduced sensitivity to other hormones including jasmonic acid, the cytokinin 6-benzyladenine and ethylene, and are hyposensitive to red and blue light [18]. The third, viable, missense, recessive Col-0 CUL1 allele, cul1-7, was identified from a screen designed to isolate mutants with defective degradation of an Aux/IAA-luciferase (IAA1-LUC) fusion protein [24]. The cul1-7 mutation affects subunit interaction at the CUL1 C-terminus. The mutant displays pleiotropic developmental defects similar to axr6-3 and cul1-6 (dwarfed with reduced apical dominance and curly leaves). Another viable CUL1 allele with a point mutation located at the C-terminus, icu13, recently isolated in the Enkheim-2 (En-2) background, shows developmental defects that have been associated with alterations of auxin signaling [25]. Traits of the mutant plants include mild leaf hyponasty, increased numbers of vegetative leaves, early bolting (with short flower stems), and reduced apical dominance relative to wild-type.

The suppressor mutant 494 was identified based on its AR-related phenotype, and characterized in connection to this developmental process [17]. A comparison of four-day old in vitro etiolated seedlings of 494 and cul1-494 mutants with axr6-3, cul1-7, cul1-6 and the parental controls Ws-4, sur2-1gl1 and Col-0, respectively, is shown in Figure 4A. Both 494 [17] and cul1-494 mutants displayed a reduced apical hook, like axr6-3 and cul1-6. However, seven days after transfer to light both 494 and cul1-494 grew better in vitro than axr6-3, cul1-7 and cul1-6 (Figure 4B). When grown in soil cul1-494 flowers earlier than its control Ws-4 (Figure 4C), and like axr6-3 [23] has reduced male fertility. In contrast, 494 is fertile, and except for a reduced number of rosette leaves and slightly reduced height, does not display any other obvious developmental defects, despite retaining similar endogenous auxin contents to sur2-1gl1 [17]. These finding show that the cul1-494 mutation not only suppresses the AR formation but also other developmental defects associated with high auxin content, such as epinastic cotyledons and leaves, long petioles and small leaf blades (Figure 4B and C). The cul1-494 single mutant produces similar numbers of AR and lateral roots to the corresponding wild type, Ws-4, while axr6-3, cul1-7, cul1-6 do not develop AR and produce fewer lateral roots than the wild type (Figure 4B). These observations suggest that cul1-494 is a weaker allele than the others. Nevertheless, the suppression of AR production in 494 is likely due to weak auxin responses (since AtCUL1 is a component of SCF-type ubiquitin ligase complexes containing the F-box TIR1, which are essential for auxin responses [43]) arising from functional perturbation of SCF, as observed in the other mutants [18], [23]–[25]. This hypothesis is consistent with our previous findings that expression of three auxin-inducible GH3 genes (GH3.3, GH3.5, and GH3.6) is strongly down-regulated in 494 suppressor mutant, despite its elevated endogenous auxin content [17]. These three GH3 genes are essential for AR development in Arabidopsis hypocotyls [35], [36] and their expression level correlates with the number of ARs [17].

The cul1-494 mutant shares, to a certain extent, the pleiotropic perturbations associated with the other viable CUL1 mutant alleles and, more specifically, the fertility defect. AtCUL1 is a component not only of SCF-type ubiquitin ligase complexes containing the F-box TIR1 essential for auxin responses [43], but also complexes containing COI1, an F-box protein that mediates jasmonate signaling [44]. Furthermore, analyses of effects of point mutations in AtCUL1 have shown it plays an essential role in responses to jasmonates [18], [45], which are important regulators of plant development and responses to environmental stresses [46]. Jasmonic acid (JA) has demonstrated importance for flower development and fertility [47], [48], and we recently showed that it inhibits AR formation downstream of auxin signaling in Arabidopsis hypocotyls [36]. Therefore, the low number of ARs produced by the suppressor mutant 494 is most likely due to a defect in SCF^TIR1 functions, but not in SCF^COI1 functions.

Our recent publication indicates that, in addition to auxin responses, ethylene biosynthesis is potentially impaired in the 494 mutant [17]. Interestingly, cul1-6 also reportedly has reduced sensitivity to ethylene [18], and disturbed responses to several other hormones known to require SCF function have been observed in different CUL1 mutants [18], [24]. This is consistent with observations that all the hormones known to be involved in the control of AR formation interact through complex crosstalk [33], which remains to be fully elucidated.