DEFECTIVELY ORGANIZED TRIBUTARIES 5 is not required for leaf venation patterning in Arabidopsis thaliana

SUMMARY The search for genetic regulators of leaf venation patterning started over 30 years ago, primarily focused on mutant screens in the eudicotyledon Arabidopsis thaliana. Developmental perturbations in either cotyledons or true leaves led to the identification of transcription factors required to elaborate the characteristic reticulated vein network. An ortholog of one of these, the C2H2 zinc finger protein DEFECTIVELY ORGANIZED TRIBUTARIES 5 (AtDOT5), was recently identified through transcriptomics as a candidate regulator of parallel venation in maize (Zea mays) leaves. To elucidate how AtDOT5 regulates vein patterning, we generated three independent loss‐of‐function mutations by gene editing in Arabidopsis. Surprisingly, none of them exhibited any obvious phenotypic perturbations. To reconcile our findings with earlier reports, we re‐evaluated the original Atdot5‐1 and Atdot5‐2 alleles. By genome sequencing, we show that reported mutations at the Atdot5‐1 locus are actually polymorphisms between Landsberg erecta and Columbia ecotypes, and that other mutations present in the background most likely cause the pleiotropic mutant phenotype observed. We further show that a T‐DNA insertion in the Atdot5‐2 locus has no impact on leaf venation patterns when segregated from other T‐DNA insertions present in the original line. We thus conclude that AtDOT5 plays no role in leaf venation patterning in Arabidopsis.


INTRODUCTION
Vascular plants rely on a complex network of veins to transport water, nutrients and sugar throughout the plant and to serve as structural support. How the vascular network is established and maintained has been the focus of research for decades, spanning disciplines from physiology and development to hydraulics and mathematics (reviewed in Nelson & Dengler, 1997, Sack & Scoffoni, 2013, De Rybel et al., 2016, Perico et al., 2022. Early physiological studies revealed that auxin induces vascular formation after wounding, leading to the hypothesis that canalization of auxin flow through developing veins facilitates selforganization of the vascular network (Sachs, 1969;Sachs, 1981). Attempts to validate this hypothesis at the molecular level are still ongoing, with new discoveries alternately supporting or refuting the possibility (reviewed in Rolland-Lagan & Prusinkiewicz, 2005, Bennett et al., 2014, Ravichandran et al., 2020, Perico et al., 2022. Central to these endeavors are screens for mutants with perturbed vascular patterning that enable causative mutations to be identified and thus genetic regulators of the patterning process to be revealed. Mutants with perturbed vascular patterning in roots and/or shoots have been reported in both eudicotyledons and monocotyledons, with most studies to date focused on the eudicot species Arabidopsis thaliana (e.g., Candela et al., 1999;Carland et al., 1999;Hardtke & Berleth, 1998;Petricka et al., 2008;Scarpella & Meijer, 2004;Smillie et al., 2012).
The extent to which patterning mechanisms are conserved in eudicot leaves that elaborate reticulate vein networks and monocot leaves that develop parallel veins is unknown; however, transcriptomic studies identified an ortholog of the Arabidopsis DEFECTIVELY ORGANIZED TRIBUTARIES 5 (AtDOT5) gene as a candidate regulator of leaf venation patterning in maize (Zea mays) . This observation suggests that aspects of the patterning process may be shared in eudicots and monocots. The Arabidopsis defectively organized tributaries (Atdot) mutants were isolated from an extensive forward genetics screen for altered vein patterning in young leaves of approximately 30 000 individuals (Petricka et al., 2008). The plants were derived from three different mutagenized populations, two generated in the Columbia-0 (Col-0) background using either diepoxybutane (Clay & Nelson, 2005) or activation tagging mutagenesis (Weigel et al., 2000) and one obtained by Dissociation (Ds) transposon mutagenesis in the Landsberg erecta (Ler) background (Bancroft et al., 1992). The pleiotropic Atdot5-1 mutant was isolated from the Ds mutagenized Ler background and was classically mapped to the At1g13290 (AtDOT5) locus (Petricka et al., 2008), which encodes a WIP family C2H2 zinc finger protein with predicted transcription factor activity (Appelhagen et al., 2010). Atdot5-1 mutants exhibited narrower leaves with misaligned veins, delayed leaf initiation, reduced apical dominance, short roots and enhanced auxin sensitivity. The Atdot5-1 allele differed from wild type at four amino acid positions, all outside of the C2H2 and WIP domains. A second allele, Atdot5-2, contained a T-DNA insertion upstream of the C2H2 domain that conditioned an embryo lethal phenotype (Alonso et al., 2003;Petricka et al., 2008). A second T-DNA insertion in the line containing the Atdot5-2 allele was subsequently reported in the 5 0 untranslated region (UTR) of At2g26740 (https://abrc.osu.edu/stocks/631530), which encodes one of the Arabidopsis SOLUBLE EPOXIDE HYDROLASES (AtSEH) proteins (Kiyosue et al., 1994;Pineau et al., 2017). Constitutive expression of the AtDOT5 genomic sequence from wild-type Ler was reported to only partially complement the leaf initiation defects of the Atdot5-1 mutant (Petricka et al., 2008). To date, the mechanism by which AtDOT5 regulates vein patterning in Arabidopsis has not been elucidated.
Given the potential for conserved function of AtDOT5 orthologs in both eudicot and monocot leaf vein patterning, we sought to determine the mechanism of gene function. To this end we used CRISPR/Cas9 to generate three independent loss-of-function alleles in Arabidopsis. Here we report the characterization of those mutants and a reanalysis of the original mutant alleles.

RESULTS & DISCUSSION
Gene edited loss-of-function Atdot5 mutants show no vein patterning or morphological defects To generate null alleles of AtDOT5, CRISPR/Cas9 was used in conjunction with a single guide RNA (sgRNA) that was designed to target the first exon of the At1g13290 locus. The guide was predicted to bind 100 bp downstream from the start codon and to induce mutations that would disrupt both the WIP and the C2H2 zinc finger domains (Figure 1a). T1 plants were screened for potential mutations by genomic PCR and three different loss-of-function alleles were identified in which deletions or insertions led to a premature stop codon (Figure 1a). Two alleles, containing a 5-bp deletion (C11.1_4) and a 1-bp insertion (C11.4_A3), arose in the same background (line C11). In order to segregate away any potential interactors from the shared C11 background that might influence the mutant phenotype, the two lines were independently backcrossed as the male parent to wild-type Col-0 and then selfed to F2. The third allele, containing a 1-bp deletion, was isolated as a heterozygous, transgene free T2 line and was fixed as homozygous (F12.12_5) in the T3 generation. For all three mutant alleles, segregating wild-type siblings were fixed in the same generation for use as controls in phenotyping experiments.
To characterize the mutant phenotype, homozygous mutants from all three independent lines were grown to maturity alongside wild-type controls. Unexpectedly, mutant plants were developmentally indistinguishable from segregating wild types at all stages of development. Leaf initiation was unperturbed (Figure 1b,c), there was no loss of apical dominance or delayed flowering (Figure 1d, e), floral organs developed normally and the plants were fertile (Figure 1f,g). Furthermore, cross-sections revealed no differences in patterning of the stem vasculature (Figure 1h,i) and paradermal views revealed normal leaf venation networks (Figure 1j-m). As such, contrary to previous reports, these results indicated that AtDOT5 function is not necessary for leaf venation patterning or for the regulation of overall plant morphology.
The Atdot5-2 allele conditions no obvious developmental defects To try and reconcile our finding that AtDOT5 is dispensable for normal plant development with earlier reports, we re-evaluated the genotype and phenotype of the original Atdot5-2 mutant line. The Atdot5-2 allele was acquired as a homozygous line in the Col-0 background (SALK_ 148869c). In this line a T-DNA is inserted in the first exon of AtDOT5 (At1g13290), upstream of the WIP and C2H2 zinc finger domains ( Figure 2a). The genotype was confirmed as homozygous by genomic PCR using primers flanking the T-DNA insertion point together with a primer in the T-DNA left border ( Figure S1). Importantly, the SALK_148869c line differs from the original SALK_148869 line in that the second insertion at the At2g26740 locus has been segregated away ( Figure S1) and the mutant phenotype is not embryo lethal. The phenotype of the SALK_148869c mutant should thus reflect loss of function of AtDOT5. In experiments analogous to those performed with the gene edited Atdot5 loss-of-function alleles (Figure 1), the Atdot5-2 mutant phenotype was compared to that of Col-0 wild-type plants. Atdot5-2 mutants exhibited normal leaf, rosette and flower development, no loss of apical dominance and full fertility (Figure 2b-g). Crucially, no changes were observed in either stem (Figure 2h,i) or leaf vasculature patterning (Figure 2j-m). These data are consistent with our hypothesis that loss of AtDOT5 function does not lead to perturbed vein patterning and suggest that the reported embryo lethal phenotype in the original SALK_148869 line (Petricka et al., 2008) was caused by loss of function of the soluble epoxide hydrolase encoded by the At2g26740 locus. It is notable that this Developmental defects exhibited by the Atdot5-1 mutant cannot be explained by mutations in the AtDOT5 gene Given that the new gene edited alleles and the Atdot5-2 allele, all of which are in the Col-0 background, do not lead to perturbed leaf venation patterns or to any general morphological defects, we considered whether the Atdot5-1 mutant phenotype was specific to the Ler background and thus to any genetic interactions present therein. To this end, we re-evaluated the phenotype and genotype of the original Atdot5-1 mutant line. Throughout development, Atdot5-1 mutants exhibited a pleiotropic phenotype characterized by variable seedling  (Figure 3j,l) with fewer tertiary veins evident, most of the quaternary veins absent and higher order veins completely absent. These observations are consistent with the report that suggested AtDOT5 influences multiple aspects of plant development, including leaf venation patterning (Petricka et al., 2008).
In an attempt to explain the conflicting phenotypic differences between Atdot5-1 versus the gene edited and Atdot5-2 loss-of-function mutants, we compared the AtDOT5 locus in Ler and Col-0 Arabidopsis accessions. To this end, the coding sequence (CDS) of AtDOT5 (At1g13750) from the recently published Ler genome (Zapata et al., 2016) was aligned to the Col-0 AtDOT5 (At1g13290) reference sequence (TAIR). The two sequences differ by 10 single nucleotide polymorphisms (SNPs), six of which result in amino acid changes at positions 36 (T to A), 64 (S to T), 244 (S to G), 267 (V to E), 269 (E to K) and 294 (Y to C). In the original report by Petricka et al. (2008), nine point mutations were reported to alter the sequence of the AtDOT5 gene in the Atdot5-1 mutant background, with four of them resulting in amino acid changes at positions 46 (T to A), 64 (S to T), 244 (S to G) and 294 (Y to C). The Ler genome sequence reveals that three of the four amino acid changes reported as mutagenesisinduced in the Atdot5-1 background correspond to natural variation present between Ler and Col-0. We could not validate the fourth change (46 -T to A) in the genome of Ler or in the genome of the Atdot5-1 mutant (see below). As such, mutations in AtDOT5 cannot explain the phenotypic changes observed in the Atdot5-1 mutant. To identify potential causative mutations underlying the Atdot5-1 phenotype, the genome of mutant plants was sequenced at 30-fold coverage. When the sequence of the AtDOT5 locus in the mutant background was aligned with the published Ler genome, no significant changes were identified in AtDOT5 or in any of the 10 adjacent genes either upstream or downstream ( Figure S2). As such, we cannot explain how partial complementation of the Atdot5-1 mutant phenotype was achieved when sequences from this genomic region were used in transgenic complementation experiments (Petricka et al., 2008). Given that the Atdot5-1 line was identified in a collection of transposon-tagged mutants, we next looked for evidence of transgenes and/or transposons in the Atdot5-1 mutant genome. A single transgene insertion was identified, corresponding to the transposon-containing construct used in the mutagenesis process (Bancroft et al., 1992). Transgene reassembly from sequence traces showed that the Ds element was inserted in the CDS of the kanamycin resistance gene (NPTII) in the Atdot5-1 mutant line, instead of in the streptomycin phosphotransferase gene where it was positioned in the original transformation construct (Figure 4a,b). This observation suggests that the Ds was transactivated by an autonomous Activator element at some point since the original lines were generated. Transposition of Ds is further evidenced by duplication of an 8-bp sequence 'GCAGCTGT' at the insertion point in NPTII.
Although not well supported, a single pair of reads, one read corresponding to the Ds element and its paired read mapping to positions 11 454 793-11 454 942 on chromosome 4, suggests that a Ds element might also be inserted upstream of the final exon of AtLer-4G47010, the Ler ortholog of AT4G20370 in Col-0 that encodes TWIN SISTER OF FT (TSF). However, because loss-of-function mutations in TSF do not condition phenotypes similar to Atdot5-1 (Yamaguchi et al., 2005) and there was just a single read, we discounted the possibility that the insertion was significant. No other Ds insertion events were detected in the genome.
The sequence flanking the Ds-containing transgene in the Atdot5-1 genome indicates that the transgene is inserted between the DEFORMED ROOTS AND LEAVES (DRL1) locus (At1G13870) and a gene of unknown function (At1G13880) (Figure 4a). DRL1 was first identified by Ds tagging with the reported loss-of-function drl1-1 mutant caused by a transposed Ds (tDs) in the DRL1 CDS (Bancroft et al., 1993). Genomic PCR confirmed that the Dscontaining transgene is present at the same genomic location in both Atdot5-1 and drl1-1 mutants, suggesting that it may have been the source of the tDs that inserted into the DRL1 coding region ( Figure S3). Although some aspects of the Atdot5-1 mutant phenotype resemble those found in drl1 mutants, some are noticeably different. For example, drl1-1 mutants do not form inflorescences (Bancroft  , 1993) and drl1-2 mutants show no venation patterning defects even though leaves are narrower than wild type (Nelissen et al., 2003). As such, we hypothesized that the transgene insertion upstream of the DRL1 locus in Atdot5-1 has no functional significance. This hypothesis is supported by the fact that any phenotypic consequences of the insertion would have been segregating in the progenitor lines of the drl1-1 mutant, and none were reported (Bancroft et al., 1993). As a final verification, we assessed whether DRL1 gene expression is perturbed in Atdot5-1 mutants. DRL1 encodes a putative elongator associated protein that is expressed in all organs during wild-type development (Nelissen et al., 2003). Notably, quantitative reverse transcriptase-PCR (qRT-PCR) using RNA extracted from pooled 7-day-old seedlings demonstrated that transcript levels are not significantly different between wildtype Ler and Atdot5-1 mutant lines (Figure 4c), and the genome sequence reveals no mutations in the DRL1 CDS. We thus conclude that the pleiotropic Atdot5-1 mutant phenotype is not caused by a transposon or transgene insertion in the genome, nor by any loss or gain of DRL1 function.
To identify potential causative mutations in the Atdot5-1 genome, sequence variations that have the potential to disrupt gene function were identified. A total of eight exonic frameshift deletions and 10 exonic frameshift insertions were identified (Table 1). Previous reports of phenotypes associated with loss of function at two of the identified loci -AT3G49360 (Xiong et al., 2009) and At5G54600 (Liu et al., 2013) suggest that the frameshifts observed at these loci are not responsible for the Atdot5-1 phenotype. The other sequence variants cannot be eliminated as potential causative mutations without further investigation.

CONCLUSION
Gene editing technologies allow for unambiguous assignment of mutant phenotypes to loss-of-function alleles. This advance has enabled more robust hypotheses of gene function to be proposed than was previously possible. Genetic screens that utilized highly mutagenic chemicals or non-specific insertion tags such as T-DNA or transposons inevitably led to 'noisy' genomes that could mask the actual mutation of interest, and RNAi suppression lines similarly led to imprecise interpretations of gene function. The discovery by gene editing that the long-standing AUXIN BINDING PROTEIN played no role in auxin homeostasis is probably one of the most high-profile cases of mistaken identity in plant biology (Gao et al., 2015) but others have been reported (Bergelson et al., 2016;Yoshida et al., 2018) and there will undoubtedly be more over the coming years. In this context, we have shown here that the AtDOT5 gene does not regulate venation patterning in the Arabidopsis leaf. This finding raises the question of whether the gene name should be changed. If so, a process needs to be established to enable changes to be made in a systematic way that can also be applied to any future findings of mistaken gene identity. Columbia-0 (Col-0) seeds were originally obtained from Lehle Seeds but have been propagated in the lab for many years. Seeds were sown directly on soil (Levington Seed modular compost), stratified at 4°C for 2 days to break dormancy and then transferred to a controlled environment chamber (CER) with a set temperature of 21°C and a 16-h photoperiod. Col-0 plants used to generate CRISPR lines were grown in the greenhouse under the same temperature and photoperiod conditions as the CER.

Construct design and plant transformation
A short sgRNA was designed against the first exon of the At1g13290 gene reference sequence using the CRISPOR online tool (Concordet & Haeussler, 2018). Constructs were generated using parts of a modular cloning system based on Golden Gate technology (Weber et al., 2011). The guide sequence was integrated by PCR into an RNA scaffold derived from EC15768 (Hughes & Langdale, 2022), modified by the addition of 34 bp (5 0 -CTAGACCCAGCTTTCTTGTACAAAGTTGGCATTA-3 0 ) at the 3 0 end as found on pICH86966 (Nekrasov et al., 2013). The scaffold was assembled into a Golden Gate level 1 module (position 3, reverse) downstream of the AtU6-26 promoter. The promoter was amplified by PCR from Col-0 genomic DNA using the following primers: AtU6-26pF: 5 0 -cactctgtggtctcaGGAGAAGCTTCGTTGAACAACGGA-3 0 and AtU6-26pR 5 0 -cactctgtggtctcaCAATCACTACTTCGACTC TAGCTG-3 0 , containing 4-bp sequences and BpiI restriction sites compatible with the PU Level 0 vector, EC41295. The Cas9p gene (Ma et al., 2015) obtained by synthesis as an SC module (CDS) was cloned under the control of the AtYAO (At4g05410) promoter (Yan et al., 2015) in a Golden Gate level 1 module corresponding to position 2, forward. The AtYAO promoter was obtained by PCR from genomic DNA with primers pAtYAO-F (5 0 -cactctgtggtctcaGGAGACCCAAATCAACAGCTGCAA-3 0 ) and pAtYAO-R (5 0 -cactctgtggtctcaCATTTCTTCTCTCTCTCACTCCCT CT-3 0 ) and cloned as described above into a PU Level 0 vector, EC41295. To terminate transcription the t-NOS terminator module, EC41421, was used. To allow for the selection of transgenic plants, a module containing the bar gene fused to the Agrobacterium tumefaciens nopaline synthase (NOS) promoter and terminator was cloned adjacent to the T-DNA left border of the pICSL4723 Golden Gate level 2 backbone, in position 1, reverse. Arabidopsis Col-0 plants were transformed by floral dipping (Clough & Bent, 1998).

Genotyping
To genotype CRISPR lines, 96 T1 basta resistant seedlings were analyzed using a CAPS marker designed to cut the wild-type sequence at the predicted editing site (3 bp into the guide sequence from the PAM) ( Figure S4). The undigested fragment was sequenced to identify the nature of the mutations. The T2 progeny were screened to select mutant lines free of the transgene. When it was not possible to segregate the mutation from the transgene, the lines were backcrossed onto Col-0.
Three biological replicates for Ler, four biological replicates for the Atdot5-1 mutant and three technical replicates for each sample were run. The Cq values and primer efficiencies were calculated for each sample using the R package qpcR (Ritz & Spiess, 2008) and relative expression levels were calculated using the EasyqpcR package based on previously published algorithms (Hellemans et al., 2007). The box plot was generated using RStudio.

Histology
To analyze leaf venation patterns, leaf 5 from approximately 3week-old (CRISPR mutants, Salk lines, wild-type Ler and Col-0) or approximately 4-week-old (Atdot5-1) plants was fixed in 3:1 ethanol:acetic acid and chlorophyll was cleared by successive washes with 70% ethanol followed by overnight incubation in histoclear. Leaf vasculature was imaged using a Leica S9i stereomicroscope against a dark background.
Young stem segments were cut 1 cm from the base of the inflorescence soon after bolting. Where possible the plants were selected to have stems of equivalent heights. In the case of the Atdot5-1 mutant and the Ler wild-type control, lateral shoots were used instead of the main inflorescence because the mutant lacked apical dominance. One-centimeter segments were fixed overnight using 3:1 ethanol:acetic acid and then infiltrated with paraffin wax in a Tissue-Tek VIP machine (Sakura, www.sakura.eu) following a protocol described previously (Hughes et al., 2019). Tenmicrometer sections were stained using a 1% Safranin solution in 50% ethanol, counterstained with a 0.04% fast-green solution in 95% ethanol and mounted in DPX mounting medium. Brightfield images were obtained using Leica LASX software and a DFCT000T camera fitted on a Leica DMRB microscope.

AUTHOR CONTRIBUTIONS
DV and JAL conceived and designed the experiments. DV carried out the experiments and analyzed the data. DV and JAL wrote the manuscript.

ACKNOWLEDGMENTS
We are grateful to Julie Bull, Roxaana Clayton and Lizzie Jamieson for technical support; John Baker for photography; Steve Kelly for help with bioinformatic analyses; and Tom Hughes, Sophie Johnson, Julia Lambret-Frotte, Chiara Perico, Sovanna Tan and Maricris Zaidem for discussion throughout the experimental work and during manuscript preparation. This work was funded by the Bill and Melinda Gates Foundation C 4 Rice grant awarded to the University of Oxford ( -20192019-2024.

CONFLICT OF INTEREST
The authors have no competing interests to declare.

DATA AVAILABILITY STATEMENT
All data generated or analyzed during this study are included in this published article and its supplementary information files, except for the sequence reads of the Atdot5-1 mutant genome which are available at ArrayExpress https://www.ebi.ac.uk/arrayexpress under accession number E-MTAB-12010.

SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article. Figure S1. Validation of T-DNA insertion in AtDOT5 in the SALK_148869c line. Figure S2. Snapshots of sequencing reads for genes flanking AtDOT5 in the Atdot5-1 genome. Figure S3. Confirmation of shared transgene position in Atdot5-1 and Atdrl1-1 mutants. Figure S4. PCR assay to genotype gene edited alleles of AtDOT5.