Zea mays Drought‐Overly Sensitive1/TUBA4 Is Wilty3, and Transcriptome Co‐Expression Analysis of Shoot Meristem Mutant Tissues Reveals Wilty2/TUB6:Wi3 Interactions Associated With Stem Vascular Bundle Development

ABSTRACT

Plant vasculature is essential for the transport of water, nutrients, and signaling molecules across organs, while also providing critical mechanical support for growth and development. Disruptions in vascular bundle formation can therefore lead to severe physiological and developmental defects. In maize, ethyl methanesulfonate (EMS)‐induced dominant nonallelic Wilty mutants exhibit a pronounced wilting phenotype even under well‐watered conditions, indicating underlying defects in vascular function. In this study, we characterized the Wi3 mutant, identified as ZmDrought‐Overly‐Sensitive1/DOS1, and compared it with the previously described Wi2 mutant to uncover shared mechanisms underlying their phenotypes. We provide evidence, by bulk segregant resequencing linkage disequilibrium of SNPs adjacent to the causal Wilty SNPs in respective ß‐ and α‐tubulin genes, for the personal communication from Gerry Neuffer that Wi2/ß‐tub6 provenance is from ACR‐related stock, whereas Wi3/α‐tub4 allele is from Mo17, not B73 as claimed by the authors who cloned Dos1. Histochemical staining and Fourier‐transform infrared (FTIR) spectroscopy of vascular bundles in Wi3 indicated apparent alterations in cellulose and lignin content consistent with those observed in Wi2. Transcriptome analysis of shoot meristems further indicated that similar sets of genes and pathways are differentially expressed in both mutants, suggesting convergence on common biological pathways. Using bulk‐segregant whole‐genome resequencing, we identified alpha‐tubulin4 (TUA4) as the causal gene in Wi3 (ZmDOS1), harboring a C‐to‐T substitution within the N‐terminal GTPase‐binding domain. This mutation results in a glutamic acid196‐to‐lysine substitution. Given that α‐ and β‐tubulin subunits heterodimerize, and in many plants and animal mutant alleles are dominant‐negative gains‐of‐function, we infer Wi2, Wi3, and likely Wi4, based on very similar FTIR biophysical difference spectra, may act as effectors of vascular bundle cell wall deposition, potentially involving vesicle trafficking as recently shown for asymmetric cell divisions in maize stomatal development. Together, these findings highlight the functional interdependence of tubulin subunits and provide a plausible mechanistic framework for the striking biophysical, transcriptomic, and phenotypic similarities observed between Wi2, Wi3/ZmDOS1, and Wi4 mutants.

1. Introduction

Patterned secondary cell wall (SCW) structures of xylem vessels are vital for rapid high‐volume transport of water and nutrients from roots to leaves in angiosperms, whereas specialized and less efficient xylem tracheids are the basis for the tallest organisms on the planet ( Sequoia sempervirens ; coast redwood) that pull water against the gravity vector under negative pressure from transpiration to a height of 115 m without drought‐induced embolism. Formation of vasculature was a crucial step in basal plant evolution allowing colonization of land. However, vascular patterning and ontogeny is lesser studied in monocots than dicots, where procambial meristems differentiate vascular bundles (VBs) in a highly regulated process involving cell specialization of xylem and phloem, elongation, localized secondary cell wall deposition, programmed cell death, and autolysis that continues in secondary cambial meristems (Furuta et al. 2014; Saß and Schneider 2024). Several anatomical differences exist between monocot and dicot VB development, with the formation of secondary vasculature or cork from undifferentiated cambium in woody dicots being an obvious difference. The monocots lack a cambium ring of parallel rows of cells formed by secondary meristem, which separates the xylem from the phloem vasculature (Shi et al. 2019). It remains to be elucidated whether the dicots gained, or the grasses lost the ability to form cambium stem cells over evolutionary time (Roodt et al. 2019). However, because the monocots lack this cambial formation, vascular patterning needs to be precise to allow efficient water conductance. Defects in development of vasculature in monocots can lead to extreme phenotypes. One such class of mutants in maize, which may be affected in some aspects of VB development are the Wilty mutants. The classic phenotype is wilting of leaves in well‐watered conditions, unlike the wilty phenotypes of drought stress hormone abscisic acid biosynthesis mutants in Arabidopsis (Rock and Zeevaart 1991) and tomato (Taylor et al. 1988; Burbidge et al. 1999).

The single gene spontaneous recessive wilty1 (wi1) mutant (MT) of maize was first isolated by Oliver Nelson in 1950s and reported to have stunted growth and wilting phenotype at the tips of the youngest leaves three weeks after emergence. The wi1 mutant, known only by phenotype but mapped to chr6, has normal stomata number and size but abnormal immature or absent metaxylem vessels in stems, suggesting a defect in the VB morphogenesis and/or differentiation (Postlethwait and Nelson 1957). Another recessive mutation, wilty5 (wi5) isolated from M₂ population of a Mutator transposon line also displays stunted growth and wilting phenotype along with progressive leaf chlorosis after the three‐leaf stage. The wi5 mutation was mapped using 3800 progenies to a genetic interval of 50 kbp on chr5 containing a 20 bp deletion in an open reading frame (ORF) encoding for endo‐1,4‐β‐xylanase from glycosyl dehydration family10. The cloned mutation was shown to be responsible for the phenotypic changes by transgene complementation with the wild type (WT) promoter driving Wi5 cDNA ORF in progeny of crosses with wi5 (Hu et al. 2020). Other maize recessive wilty mutants have been cloned and provide insights into water transport efficiency, namely, necrotic upper tips1/nut1 encoding a NAC transcription factor homologous to Arabidopsis VASCULAR‐RELATED NAC‐DOMAIN 6/ANAC101 induced by flowering and localized to the developing protoxylem of root, stem, and leaf sheath, but not metaxylem (Dong et al. 2020), and drought sensitive1/Zm00001d001765, an ortholog of Arabidopsis Exocyst70A1 involved in tethering of post‐Golgi vesicles and deposition in the plasma membrane of CASPARIAN STRIP MEMBRANE DOMAIN PROTEINs, CELLULOSE SYNTHASE7, and polar transport of auxin (Drdová et al. 2013; Kalmbach et al. 2017). ZmIQD27 was cloned and shown expressed in the root meristematic zone where secondary cell wall deposition is initiated, and loss‐of‐function iqd27 drought‐sensitive mutants exhibit a microtubular arrangement disorder (Li et al. 2023). Drought sensitive1 gene is regulated by NUT1 during maize root stele and stem proto‐ and meta‐xylem to mediate development of secondary wall annular ring patterning, lignin deposition, and cellular differentiation (Zhu et al. 2026).

There exist three dominant Wilty mutants of maize; all ethyl methane sulfonate (EMS)‐induced: Wilty 2 (Wi2; chr3; 11 cM from T3‐9c), Wilty3 (Wi3; unmapped), and Wi‐2445 (Wi4; mapped 25 cM distal to T5‐9c translocation) (Anderson 1938; Neuffer 1989, 1990). Wi2/3/4 heterozygous plants have normal amounts of abscisic acid and induction of drought‐responsive gene markers (Rock and Ng 1999). The Wi2 MT was initially reported by the authors of wi5 cloning to encode for beta tubulin6b (TUB6b) gene (Zhao et al. 2015; Dhingra 2021; Huang et al. 2024), and subsequently validated by us (Dhingra 2021), with an additional weak Wi2 allele ZmA457 L215F substitution documented that segregates 3:1 WT:drought sensitive (Huang et al. 2024). The dominant Wilty mutants show varying degrees of wilting phenotype with Wi2 being the least wilted (Wi2 ^A457 homozygote is fertile) (Huang et al. 2024) and stunted, whereas Wi3 has an intermediate phenotype and Wi4 has the most severe phenotype.

Because fibrous protein “intermediate filaments” such as keratins, vimentins, desmins, and laminins are not found in plants, and cell walls of plants preclude cell migration as a mechanism in morphogenesis of tissues and organs, the actin and tubulin cytoskeletal microfilaments are of fundamental importance for plant growth, morphogenesis, development, and response to environmental perturbations (Li and Staiger 2018; Livanos and Müller 2019). Tubulin heterodimers of alpha (GTP‐binding) and beta (GTP/GDP‐binding) subunits contribute to the polar formation of hollow columns of microtubule protofilaments involved in oriented asymmetric cell divisions, meiosis, polarity of cell growth, and intracellular trafficking. In organisms with lower genome complexity, mutations in tubulin genes tend to be lethal (Neff et al. 1983). Mutations in alpha‐ and beta‐tubulins both impair vesicle trafficking in animal neurons, leading to similar developmental and, in some cases, severe neurological phenotypes (Buscaglia et al. 2020; Hoff et al. 2022), and mutations reported have been heterozygous missense mutations, supporting altered protein function rather than haploinsufficiency as a primary genetic etiology. However, in plants, due to genetic redundancy, mutations in tubulin genes have been reported to be subtle and nonlethal (Breviario et al. 2013). Plants with mutations in cytoskeletal components or when treated with cytoskeletal protein polymerization inhibitors display phenotypic similarities to plants with cell wall synthesis defects like dwarfism, alterations in cell wall composition, and swollen epidermal cells (Sugimoto et al. 2003; Buschmann and Lloyd 2008; Endler and Persson 2011). The cytoskeleton is vital for trafficking of cell wall components synthesized in the endoplasmic reticulum and the Golgi apparatus (Crowell et al. 2009; Bashline et al. 2014). Cortical microtubules define the site of insertion of the cellulose synthase complex (CSC) and guide the direction of the CSC during the process of wall cellulose synthesis (Paredez et al. 2006). Another polymer in secondary plant cell walls, lignin, provides hydrophobicity and rigidity to tracheary elements, which allows them to withstand the negative pressures of vertical capillary water transport (Turner and Somerville 1997). Lignin molecules are generated by the radial coupling of monolignols (hydroxycinnamyl alcohols): guaiacyl (G), sinapyl (S), and hyrdoxylcoumaroyl (H) units (Dixon et al. 2001). Lignin composition varies across plant species and tissues; gymnosperms are rich in G units and low in H units, whereas dicot lignin is mainly composed of G and S units (Weng and Chapple 2010).

Here, we document characterization of the causal mutation of Wi3‐N1614 dominant allele of maize (Neuffer 1990; Dhingra 2021), generated in the Mo17 background (G. Neuffer, pers. comm. 2018). Our biophysical characterization by Fourier transform infrared spectroscopy of extracted Wi2, Wi3, and Wi4 stem VBs and transcriptome analyses of Wi2 and Wi3 heterozygote shoot meristems indicate that Wi3 and Wi4 dominant mutations are most likely functionally related to Wi2/beta tubulin6 (Zhao et al. 2015; Huang et al. 2024), consistent with a hetero‐dimerization model for tubulin function with many examples in plants and animals for dominant‐negative MT phenotypes. Bulk‐segregant whole genome resequencing of pools of segregating Wi3 mutants and WT siblings identified alpha tubulin4 (TUA4) as the causal gene for the Wi3 MT allele, with independent support from absolute linkage disequilibrium of > 160 Mo17‐unique SNPs flanking the Wi3 causal mutation; the cultivar provenance of the sole Wi3 allele. The causal SNP is a C → T transition, which changes residue 196 glutamic acid to lysine in a highly conserved GTP‐binding domain of TUA. Notably, the same C➔T E196K amino acid substitution mutation was reported as causal for ZmDrought Overly Sensitive1 without documentation for claimed B73 provenance (Huang et al. 2024). The identical E196K lesion, together with the absolute linkage disequilibrium of > 160 Mo17‐specific SNPs flanking the Wi3‐N1614 mutation, indicates that Dos1 and Wi3 represent the same mutation arising in the Mo17 background; a B73 origin for Dos1 was reported but not documented. Stocks corresponding to these alleles are publicly available for redistribution without limitation through the Maize Genetics Cooperation Stock Center. We hereby honor the memory and life work of Dr. Myron Gerald Neuffer (March 4, 1922–October 8, 2019) without whose pioneering creation and selfless cataloguing of thousands of unique maize mutants this work would not have been possible.

2. Materials and Methods

2.1. Plant Materials

The seeds for the maize stock lines 605F:wi1,Y1,pl1 (genetic background line “2006‐980‐10/980‐5”), 301D:Wi2‐N1540 (genetic background line “1997‐4677‐3/4666‐4”; propagation in W23/L317 genotypes, but originally from colored aleurone “mongrel ACR” stock of Gerry Neuffer [pers. comm.]), U840G:Wi3‐N1614 (“2006P‐279‐8/240‐4, W23/L317”; provenance from Mo17 [Gerry Neuffer; pers. comm.]), and 512E:Wi4‐N2445A (“2011‐B73‐39/676‐8”; provenance from A632 [Gerry Neuffer, pers. comm.]) were obtained from the Maize Genetics Cooperation Stock Center (University of Illinois, Urbana; http://maizecoop.cropsci.uiuc.edu/) and progeny were observed to segregate 1:1 for Wilty dominant phenotypes (Figure S1). The mutants were originally generated by pollen from providential stock treated with EMS (G. Neuffer, pers. comm.). Individual Wi2 and Wi3 heterozygous plants grown in the field manifesting Wilty symptoms at the seven to eight leaf stage, before (Wi2) and during (Wi3) internode elongation associated with the transition to tasseling, were crossed into the reference genome B73 stock background (obtained from the US National Plant Germplasm System, accession PI550473) and back‐crossed (BC) to B73 for three generations in the years 2015‐17 to generate mapping populations. For bulk‐segregant resequencing Analysis (BSA‐seq) and RNA‐seq, segregating sibling plant tissue samples were frozen individually from dissected shoot meristems (Figure S2) from 168 heterozygous Wi2/+ individuals and 178 homozygous WT siblings, and from 92 heterozygous Wi3/+ individuals and 90 WT siblings. The cataloged frozen tissue from individuals preserved the ability to genotype rare recombination breakpoints near the causal loci in subsequent assays. Figure S3 shows the workflow and conceptual rationale for BSA‐seq.

A genetic test of Wi2 × Wi3 heterozygous gene interaction was performed by crossing Wi2/+ and Wi3/+ heterozygotes and growing 29 progeny plants in the field from September 15 to November 20 when first freeze temperature ended the experiment. Leaves of individual seedlings were extracted for genotyping, and end of growing season physiological parameters measured for plant height, diameter of stalk at base, and dry biomass. A chi‐square test for goodness‐of‐fit was performed under a hypothesis of 1:2:1 phenotypes for WT:single allele (Wi2 or Wi3 alone):double allele Wi2‐plus‐Wi3 mode according to Mendel's Second Law.

2.2. DNA Extraction for Bulk‐Segregant Whole Genome Resequencing ANALYSIS

For creating the bulk‐segregant pools, equal amounts of fresh individual meristem tissue were pooled in aliquots of 25 individuals scored for phenotype. Genomic DNA was extracted from pools of 25 individuals ground to powder with mortar and pestle under liquid N₂ using the Qiagen DNAeasy Plant DNA extraction kit and quantified using Nanodrop spectrophotometer (ND‐1000, Thermo Fisher, Waltham, MA) and qualified as high molecular weight by agarose gel electrophoresis. The extracted gDNA was sheared as per the recommended 550 bp insert size using Covaris M220 ultra sonicator (Woburn, MA). Two hundred nanograms of DNA was suspended in 52.5 μL of resuspension buffer (Illumina TruSeq Nano DNA kit; San Diego, CA), transferred to Covaris tubes, and subjected to fragmentation with 50 W peak incident power, 20% duty‐cycle factor, 200 cycles per burst for 45 s at 20°C. Genomic libraries were prepared with the Illumina TruSeq Nano DNA library prep kit according to manufacturer protocol. IDT for Illumina UD DNA dual indexes were used for bar‐coding the libraries and completed libraries were assayed for insert size (average 550 bp) and quantified on Agilent 2100 Bioanalyzer using High‐Sensitivity DNA chip (Catalogue #5067‐4626). The equimolar pool of the indexed genomic libraries was subjected to the adapter blocking reagent (Illumina cat # 20024144) and sequenced on an S4 lane of Illumina NovaSeq 6000, 2 × 150 bp, as fee‐for‐service by the University of California‐San Francisco IGM Genomics Center.

2.3. BSA‐seq Bioinformatics Pipeline

The raw reads obtained from the BSA‐seq in fastq format were aligned to the Z. mays B73 whole genome reference assembly_v4, https://www.ncbi.nlm.nih.gov/datasets/genome/GCA_000005005.6/ (Jiao et al. 2017) using Bowtie2 v2.3.4 (Langmead and Salzberg 2012) (Table S1). For resequencing data analysis, the raw reads were processed further using Picard Tools (http://broadinstitute.github.io/picard/), BWA‐aligner v.0.7.17 and Genome Analysis ToolKit (GATK; Broad Institute) pipeline (Van der Auwera et al. 2013) to identify de novo single nucleotide polymorphisms (SNPs) and insertion/deletions (Indels). SNPs and Indel datasets from HapMap2 (Chia et al. 2012) and HapMap3 (Bukowski et al. 2018) were used to filter out the known SNPs/Indels in the genome data per chromosome. The GATK HaplotypeCaller tool was used to create the Genotype GVCF files subjected to a merged joint‐calling “parent–child–trio” analysis using B73 as one parent, Wi3/+ sibs as the other parent, and WT sibs library as child. The R package “QTLseqr” (Mansfeld and Grumet 2018) was used to visualize and map candidate causal mutations to a chromosome.

We speculated that SNP/Indels within the intergenic regions would most likely not be causal for Wilty phenotypes, with the caveat mutations in cis‐regulatory regions can function to effect linked gene expression. Regions spanning the mapping intervals for the causal Wi2 and Wi3 mutations were filtered to keep only the homozygous (WT) SNPs/Indels for the datasets and heterozygous reference for the MT Wilty datasets using the SnpSift function of the SnpEff_v5 tool (Cingolani et al. 2012). The de novo variants were further annotated based on codon translation and the effect of a SNP/Indel on the annotated gene structure. All SNPs and Indels within the ~10 Mb intervals identified for mapping the mutations were categorized into low, moderate, and high functional effect candidate SNPs/Indels and cross‐checked for effect of the SNP/Indel in structure of the respective genes visualized in the Phytozome web browser (https://phytozome‐next.jgi.doe.gov/info/Zmays_RefGen_V4).

Because the provenance of Wi3 from EMS of pollen was known (G. Neuffer, pers. comm. 2018) to be cultivar Missouri17 (Mo17), all high confidence B73 (four SRA accessions: ERR3288215–ERR3288218), and Mo17 SNPs (three SRA accessions: SRR1575517, SRR5976302, and SRR8907080) within 106 kb distal and 115 kb proximal interval relative to centromere at chr5:101.6–104.8 Mbp flanking the claimed Wi3 causal SNP chr5:9836290 C➔ T were downloaded from SNPVersity (Version4) website https://wgs.maizegdb.org/ (Andorf et al. 2025). These provenance‐validated unique SNPs in the GVCF Trio output were manually “lifted” from NAM5.0 output coordinates at https://www.ebi.ac.uk/eva/?Home to B73 ver4 coordinates after curating each unique European Variation Archive identifier from the Genetic Variation track at https://jbrowse.maizegdb.org/. The annotated dataset was scrutinized for recombination breakpoint evidences.

To check for the percentage of chloroplast and the mitochondrial genome contamination in our genomic libraries, the raw data were mapped to both the Z. mays chloroplast (PRJNA124; NC_001666.2) (Maier et al. 1995) and the mitochondrial genomes (PRJNA16736; NC_007982.1) (Clifton et al. 2004). For the Wi3 MT genomic libraries, 1.0% of the raw reads mapped to the chloroplast and 0.10% of reads mapped to the mitochondrial genome. Similarly, for the WT sib libraries, 1.3% of the reads mapped to the chloroplast genome and 0.12% of reads mapped to mitochondrial genome (Figure S2).

2.4. Bulk‐Segregant Total RNA‐seq Library Preparation

For bulk‐segregant mRNA‐seq (BSR‐seq), shoot meristem tissues were harvested from plants at a common developmental stage, 40 days after planting, for all genotypes. For each genotype, meristem tissue from 25 individuals scored for the respective Wi2/+ and Wi3/+ MT or wild‐type sib phenotype was pooled independently for each replicate library; MT and wild‐type sib pools were collected in parallel from the same planting to ensure stage‐matched comparisons. This yielded a total of 12 libraries (two MT genotypes and their respective wild‐type sibs and three biological replicates per genotype). Because differential expression was determined within each genotype, comparing each MT to its contemporaneously harvested wild‐type sib pool, stage‐related effects are internally controlled within each comparison. Total RNA was extracted according to manufacturer protocol using the Sigma‐Aldrich Spectrum Plant Total RNA extraction kit. The RNA was quantified on Nanodrop ND‐1000 spectrophotometer and RNA quality and quantity assayed on the Agilent 2100 Bioanalyzer instrument using Agilent RNA 6000 Nano Total RNA analysis kit (Catalogue # 5067‐1511). One microgram of total RNA was taken forward per sample to prepare mRNA‐seq libraries by Illumina TruSeq Stranded total RNA seq kit (RiboZero Plant) with time‐dependent hydrolysis step to yield an average RNA fragment size of 150 nt according to the manufacturer protocol. IDT for Illumina UD RNA dual indexes (Catalogue # 20020591) were used for indexing and finished libraries run on 6% native PAGE gel and bands between ~260 and 310 bp were excised and purified by subjecting to gel breaker (Thermo Fisher Scientific 1st Engineering Gel Breaker Tubes) as described (Mittal et al. 2023). The individual libraries were validated and quantified with an Agilent 2100 Bioanalyzer using High‐sensitivity DNA chip. The equimolar pool of the transcriptomic libraries was subjected to Illumina adapter blocking reagent (Cat # 20024144) and sequenced 1× 100 bp on an Illumina NovaSeq6000 S4 platform as fee‐for‐service from UCLA Sequencing and Genotyping Core Facility in two independent runs, providing a technical replicate that was analyzed by principal component analysis (see below, and Table S1).

2.5. BSR‐seq Bioinformatics Analysis

The raw reads obtained in the fastq format were quality checked by using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Quality control metrics for gDNAseq and RNAseq libraries are provided in Table S1. The reads were then subjected to filtration step to remove Z. mays ribosomal RNA (rRNA), transfer RNA (tRNA), small nuclear (snRNA; spliceosome uridine‐rich U‐RNAs), and small nucleolar RNA (snoRNAs; Box C/D and Box H/ACA that guide chemical modifications of other RNAs) (Kalvari et al. 2018) (https://rfam.org/) using the ‐unmap function of the Bowtie aligner (bowtie‐1.1.2 version) (Langmead et al. 2009). The raw reads were either mapped to the 39,756 Z. mays B73 cDNAs in the NAM5.0 reference file “Zmays_833_Zm‐B73‐REFERENCE‐NAM‐5.0.55.transcript_primaryTranscriptOnly.fa” (GCA_000005005.6, available for download at https://phytozome‐next.jgi.doe.gov/) for “gene level” analysis, or the 72,539 transcript variants file “Zmays_833_Zm‐B73‐REFERENCE‐NAM‐5.0.55.transcript.fa” for “probe‐level” analysis of candidate gene splicing variants using the kallisto pseudoaligner tool in sense‐strand only mode (rf‐stranded) (Bray et al. 2016). As input for MapMan gene ontology analysis, and for gene (or transcript variants for causal candidate genes) level analysis reported in Table S2, the outputs from kallisto mappings were analyzed using DESeq2 (Love et al. 2014). The 25,767 primary transcripts with automated statistical significance “padj” value real number output were used as inputs for MapMan, to reduce “shot noise” of low‐expressed genes after “shrinkage” to account for expression variances across all genotype libraries. The R package “ggplot” (Wickham 2016) was used to generate PCA (principal component analysis) of the transcriptome datasets.

To view the various metabolic processes and pathways differentially regulated in the MT RNA‐seq datasets, output from DESeq2 for 25,767 genes with false discovery rate padj < 0.05 in both Wi2 and Wi3 MT datasets (6890 genes) were analyzed with MapMan and PageMan v3.6.0RC1 analysis using gene ontology mappings from Mercator ver.4.8 and “X4.5 Vesicle trafficking R5.0” modified pathway mapping as reference (downloaded from https://mapman.gabipd.org/mapmanstore) (Schwacke et al. 2019). Table S2 list all 39,591 primary transcript “gene level” differential expression results corresponding to the MapMan gene ontology bin notations. For overrepresentation of biological process bins, two different tests were conducted: Fisher exact tests for overrepresented and underrepresented bins for both up‐ and down‐regulated genes; and a bin‐wise corrected rank sum test for overrepresentation of gene ontology bins; both tests were false‐positive corrected according to Benjamini and Hochberg (1995) (Benjamini and Hochberg 1995).

2.6. Indel Marker PCR‐Based Chromosome Walking and Wi2 and Wi3 Interval Fine Mapping

To fine map the Wi2 and Wi3 genomic locus intervals, Indel marker primer‐based PCR chromosome walking was performed (Gallavotti and Whipple 2015). The output from GATK analysis identified numerous Indels in the QTLseqr‐delimited mapping interval. Indels of 15 bp or longer in length were chosen and primers were designed to flank these Indels by 100–150 bp to generate ~300 bp amplicon sizes for ease of scoring amplicon polymorphisms in pools of MT sibling or WT genomic DNA template pools amplified and run on 3.5% agarose gels. Genomic DNA extracted from four or more distinct pools of 25 independent homozygous WT individuals was used as template for genotyping of Indel PCR markers. The candidate causal mutations were further verified using Sanger sequencing (Azenta Life Sciences, Burlington, MA) of PCR amplicons for Wi3 (Figure S7). The details of the primers used are listed in Table S3. For each PCR reaction, 25 μL reaction volume was created using 12.5 μL of GoTaq Hot Start master mix (Promega, Madison WI), one μL of DNA extract at a concentration of 10–20 ng/μL, one μL of 1 mM primer mix, and nuclease‐free water. The initial denaturation was carried out at 95°C for 5 min, followed by 30 cycles of denaturation at 95°C (30 s), annealing temperature at 61°C (30 s), extension at 72°C (30 s), and final round of extension (polishing) at 72°C for 7 min. These reaction volumes and settings were used for all primer sets.

2.7. Fourier Transform Infra‐Red/FTIR Attenuated Total Reflectance Difference Spectra Analysis of Dissected Vascular Bundles

VBs were isolated from the internodes of field‐grown plants by the process of “retting,” a primitive method for production of bast fibers from jute, hemp, and flax stalks (Hooper 1931). In the process of retting, the parenchyma tissues are digested by the action of the bacteria and other endemic micro‐organisms leaving behind strands of VBs (Shane 2000). Outer layers of the sheath were removed to expose the middle stem containing internodes, which were transversely and longitudinally cut into ~15‐cm length pieces and steeped in buckets of water covered with foil stored at room temperature for 1 month. The VBs were manually removed from digested pith tissues within the outer lignified stem “shell.” The isolated VBs were washed with distilled water and stored in 50% ethanol (Figure S4).

The isolated VBs from pith internodes were ground to fine powder using liquid nitrogen in SPEX freeze mill 6870 (SPEX sample prep USA). For dehydration of the ground powder, approximately 100–300 mg of ground tissue was placed in a microfuge tube at room temperature and 1 mL of freshly prepared 70% ethanol was added and kept at 70°C for 1 h. The supernatant was removed and the step was repeated to remove soluble components and starch (Dampanaboina et al. 2021). The supernatant was removed and 1 mL of 100% methanol was added and vortexed briefly. The solution was then centrifuged at 13,000 rpm and the supernatant was discarded and 1 mL of (1:1) chloroform:methanol was added to the pellet, vortexed and centrifuged. The supernatant was removed and 1 mL of 100% acetone was added to the pellet, vortexed and centrifuged. The supernatant was removed and the pellet was dried at 37°C overnight. The pellet was analyzed by FTIR in universal attenuated total reflectance (UATR) mode on a Perkin Elmer LAS Instrument (Shelton, CT) available at the TTU Fiber and Biopolymer Research Institute. The technique consisted of placing the dried powdered sample as packed pellet on a 2‐mm diameter ZnSe‐Diamond composite crystal (previously scanned as reference) with deuterated tri‐glycine sulphate (DTGS) as the mid‐infrared detector standard. Pressure was applied on the sample to ensure contact between the incident infrared beam and the sample. Fifteen FTIR spectra were collected from each sample at a spectrum resolution of 4 cm⁻¹, with 32 co‐added scans, over the range of 650–4000 cm⁻¹ and peak integration performed. The baseline was corrected and normalized and the data saved in ASCII (.dat) format. A difference spectrum of wavenumber between MT and control WT sample was obtained by subtracting the MT values from the WT and plotted graphically.

PCA was performed on the matrix of dataset values of 15 replicate scans from each sample of MT and WT siblings. The R software packages ggbiplot (Vu and Friendly 2023), cluster (Maechler 2026), vegan (Oksanen 2025), and ggfortify (Tang and Horikoshi 2024) were used to assess variability across FTIR spectra relative to the mean of the population.

2.8. Histochemical Characterization of Cross Sections of 6th Internode Stalks

The sixth internode of 106‐day‐old field‐grown plants of Wi2/+ and Wi3/+ and their respective WT sibs were harvested and ~1‐cm‐long internode pieces fixed in FAA (formaldehyde, acetic acid, and ethanol) for subsequent paraffin‐embedding and microtome‐sectioning to prepare permanent microscope slides. FAAT fixing solution was made with 50 mL ethyl alcohol, 5 mL glacial acetic acid, 10 mL of 35%–40% formaldehyde, and 35 mL of distilled water for a volume of 100 mL. The samples were fixed, embedded in paraffin, sectioned on a microtome at 6‐μm thickness, and stained with 1% aqueous safranin‐O (Sigma‐Aldrich #S2255) for 10–20 min, counterstained with 0.5% fast green FCF (Sigma‐Aldrich #F7258) in 95% ethanol for 10–40 min (Ma et al. 1993).

Sixth internodes were freshly harvested from 60‐day‐old segregating MT and WT sib plants grown in the field and 90‐μm‐thick cross sections were made using a Vibratome 3000 (Technical Products International, Model 074018; Campden Instruments, Loughborough, England). Fresh/live cross‐sections were bathed in freshly prepared histochemical stains for rapid visualization of cellular secondary wall components (Mitra and Loqué 2014), as described below. Stem cross‐sections were imaged using an Olympus BX41 light microscope at the Texas Tech College of Arts and Sciences Microscopy Core facility. Pith cell area was quantified with ImageJ (Fiji) (Schindelin et al. 2012) by converting the image to 8 bit and then applying threshold to generate binary images followed by automated particle analysis to measure individual cell areas (200‐infinity cell size, 0.5–1 circularity). Water masking was used to create the estimated lines; a binary mask creation technique used in image segmentation to isolate specific, often connected, objects by treating them like a topographic surface. Vascular anatomical traits, including VB area, metaxylem vessel area, and protoxylem cell areas were measured by manually outlining the respective structures using the freehand selection tool in ImageJ. Multiple cells or vascular structures were selected from each of two technical replicate sequential cross section slides to obtain representative measurements for those protoxylem cells showing a complete secondary cell wall annular ring (as opposed to those bundles often showing protoxylem lacunae; a lysigenous cavity of intercellular space formed from dissolution and disintegration of the cell that leaves a hole surrounded by remnants of wall). The R package ggplot2 (Wickham 2016) was used to create violin box plots for comparison. The assumptions of normality and homogeneity of variance were evaluated using the Shapiro–Wilk test and Levene's test, respectively. Several traits exhibited significant heteroscedasticity and departures from normal distribution; therefore, nonparametric methods were employed. Group differences were assessed using Kruskal–Wallis rank sum test. When significant differences were detected, post hoc pairwise comparisons were conducted using Dunn's test with Bonferroni correction to control Type‐I errors due to multiple testing. Statistical significance was determined at α = 0.05.

2.8.1. Maüle Stain

The cross‐sections were oxidized in 1% KMnO₄ solution for 15 min and then washed three times with distilled water. The samples were then washed with 1 M HCl for 60 s and then given a final wash with distilled water before aqueous mounting on alkalized glycerol (nine parts glycerol with one part phosphate buffered saline, adjusting the pH to 8.5–9.0 with 0.1 M NaOH) for bright field observation under an Olympus BX41 epifluorescence microscope at the TTU College of Arts and Sciences Microscopy Core facility.

2.8.2. Phloroglucinol:HCl Stain (Wiesner's Reaction)

Eighteen percent (1% w/v) phloroglucinol stain was freshly prepared and the cross‐sections were dipped in the stain for 3 min. The sections were then mounted on glass slides in a drop of acidified (75% glycerol, 15% water, and 10% concentrated H₂SO₄) glycerol and observed under bright field illumination.

2.8.3. Calcofluor White and Aniline Blue Stains

The cross‐sections were stained in 1% calcofluor white stain for 5 min and then washed with distilled water or stained with Aniline blue stain for five to 7 min and washed with distilled water. The stained cross‐sections were then mounted on slides with water and observed under UV light source on an Olympus BX41 epifluorescence microscope. The fluorescence obtained from the UV‐specific stains and autofluorescence of unstained tissues was quantified using ImageJ software (Table S8) (Schindelin et al. 2012).

3. Results

3.1. Wilty2 and Wilty3 are Semidominant Alleles That Interact Genetically, and Condition Similar Transcriptome Associated with Cytoskeleton and Cell‐wall Pathways

Neuffer (Neuffer 1990) reported that Wi3 was not an allele of semidominant Wi2 (on chr3 (Neuffer 1989)) based on progeny from a Wi2/+, Wi3/+ double allele backcross to normal/WT assorted independently, giving 3:1 ratio of Wilty:wildtype progeny. Table 1 shows results of progeny grown in the field in different years between Wi2/+ and Wi3/+ heterozygote crosses, consistent with semidominant allele interactions in progeny to condition a more severe phenotype. Selfed Wi3/+ progeny showed ~¼ progeny with severe wilty phenotype (data not shown), similar to Wi2/+ self‐fertilizations (Neuffer 1989), consistent with semidominant genetic additive effects of Wi2 plus Wi3 alleles conditioning very severe wilty physiology. We note that these gene‐interaction tests are based on modest population sizes and on morphological categorization (wilting severity, plant height, stalk diameter, and dry biomass); although the observed segregation is statistically consistent with the expected 1:2:1 ratio (Table 1), molecular genotyping of the progeny at the causal TUA4 (Wi3) and β‐tubulin (Wi2) lesions identified in this study would provide independent, genotype‐based confirmation of the additive interaction and is a priority for future analysis. Our field results (including early manifestation of severe wiltiness after germination) differ from those of Huang et al. (Huang et al. 2024) who claimed that at 25 days after germination in the greenhouse of F₁ progeny of a double‐heterozygote Wi2/+, Dos1/+ reciprocal F₁ crosses (each progeny of mapping cross to Mo17 and/or B73) gave 3:1 normal: wilty phenotypes, suggesting that Wi2 and Dos1 single MT alleles do not condition semidominant Wilty phenotypes at an early stage of growth. We have a different interpretation of the discrepancy between Huang et al. (Huang et al. 2024) and our results in Table 1 (BC₃ descendants of our initial mapping crosses): at the F₁ stage our mapping population Wi2/+ and Wi3/+ stock crosses to B73 manifested a strong heterosis/hybrid vigor phenotype, a known phenomenon in first generation hybrids between Stiff Stalk/dent inbreds like B73 and Non‐Stiff Stalk Lancaster Sure–derived elite line Mo17 (Bornowski et al. 2021). We also speculate that the Wi2 ^A457 L215F allele phenotype, described by Huang et al. (Huang et al. 2024) as recessive, may be due to hybrid vigor because their mapping population was from crosses to B73. There are no known cases to our knowledge of beta tubulin L215 mutations causing other than a dominant‐negative phenotype. L215 is in the helix 6–7 loop that participates in binding the anticancer drug paclitaxel. The L215 residue faces the microtubule lumen and forms interdimer contacts plus lateral protofilament interactions important for microtubule assembly, consistent with prior studies that show mutations conferring paclitaxel resistance occur with equal frequency in both α‐ and β‐tubulin (Schibler and Cabral 1986; Nogales et al. 1999; Wang et al. 2006).

TABLE 1.

Gene interaction test by phenotyping physiological parameters of field‐grown progeny of Wi2/+ × Wi3/+ crosses.

Experiment year (number progeny tested)	Number of mature plants severely wilted and height < 100 cm	Number of plants moderately wilted and reduced height: 116–144 cm	Number of normal/wild type plants height > 172 cm	Chi square probability 1:2:1 severe: moderate: normal
2021 (9)	1	4	4	0.35
Hypothesis:	Both Wi2 and Wi3 mutant alleles	Either Wi2 or Wi3 single allele	No mutant alleles
	Number plants severe stalk diameter at base < 3.6 cm and dry biomass < 3.8 g	Number plants moderate stalk diameter at base < 6.3 cm and dry biomass < 14.9 g	Number of plants normal stalk diameter at base > 6.5 cm, and biomass > 15.3 g	Chi square probability 1:2:1 severe: moderate: normal
2024 (29)	11	11	7	0.25

BSR‐seq of pools of meristem tissues of Wi2 and Wi3 heterozygous and WT siblings were analyzed. We obtained a total of 83.9‐m clean paired‐end reads for the six Wi3 and WT sibling libraries (Table S1). DESeq2 analysis of the BSR‐seq showed 10,155 genes significantly differentially regulated with padj < 0.05 for Wi2 and 12,848 genes differentially regulated for Wi3 with padj < 0.05 (Table S2). Principal component analysis revealed spatially coordinated clusters corresponding to the Wi2 and Wi3 MT and WT libraries, and lack of a significant batch effect for technical replicate library runs (Figure S5). We interpret the x‐axis PCA dimension represents variances approximating the older age of the Wi3 samples at harvest (at start of bolting/tasseling; meristems were beginning to green, Figure S2) versus Wi2 samples harvested from the field 2 weeks earlier; this is supported by evidence from quantification of chloroplast and mitochondrial DNAs in the genomic DNA libraries showing Wi2 samples had a relatively greater percentage of ctDNA and mtDNA than Wi3 (Table S1).

Interestingly, comparison of the two MT shoot meristem transcriptome data sets revealed that they shared over half (6890) of their significantly differentially expressed genes, where 87.2% of these were both up and both down, concordantly (Table S2), suggesting that Wi2/Tubulin‐beta6 and Wi3 mutants condition phenotypic effects by the same biological process; namely, tubulin heterodimer binding partners. Therefore, we employed MapMan Gene Ontology analysis, which further revealed significant overrepresentation of 20 major metabolism and ontology process bins associated principally with cytoskeletal organization, vesicle trafficking, and cell cycle regulation in both mutants (Figure 1). The Pearson correlation coefficient (R ² = 0.77) for the significantly overrepresented bin Z scores shared across mutants represents a very good fit to a model where the Wi2 and Wi3 gene products functionally and mechanistically interact. In particular, genes related to DNA replication/repair, nucleotide metabolism, chromatin remodeling/assembly/histones, chromosome segregation, mitosis/meiosis, cytokinesis, and phragmoplast microtubule/kinesin motor organization were up‐regulated (Table 2), as were a large number of dual tyrosine kinase‐like/ser‐thr family genes, and genes involved mRNA degradation, small interfering RNA, PPR‐type organelle RNA editing, and mRNA splicing pathways (the latter by virtue of down‐regulated genes being significantly underrepresented). On the other hand, nuclear‐encoded photosynthetic genes were significantly down‐regulated in both MT meristems, inversely correlated with cortical microtubule rearrangements into highly ordered, transverse arrays (perpendicular to the axis of growth) in normal response to light (Mathur and Hülskamp 2002). We speculate a plausible mechanism could be related to Arabidopsis snowy cotyledon3/QWRF1 plant‐specific function in microtubule and peroxisome‐associated chloroplast biogenesis and anisotropic cell expansion (Albrecht et al. 2010; Ma et al. 2021); the maize homologue Zm00001eb400630 is significantly up‐regulated in Wi3 and up in Wi2 mutants (Table S2). Another possible mechanism is via chloroplast retrograde signaling to histone remodeling during shoot meristem greening (Quevedo et al. 2025), consistent with observed up‐regulation of PPR organelle RNA processing genes (Table 2) as a response to energy deficits (Kendrick et al. 2022). In addition, down‐regulated were heat shock/chaperone genes, genes involved in lipid droplet‐associated metabolism (intrinsically related as a hub to vesicle trafficking), abscisic acid signaling/PP2C cladeA, BBX and HD‐ZipI/II transcription factor families, and major intrinsic solute transporter family genes (Table 2). In Arabidopsis, cytoplasmic mRNA decay factors (DCP5, LSM1–7, and XRN4) regulate the stability of mRNAs involved in abscisic acid signaling, where ABA induces microtubule disassembly via MICROTUBULE‐RELATED E3 LIGASE57 necessary for stomatal closure during drought stress (Dou et al. 2021; Wang et al. 2023), the homologues of which (Zm00001eb232400, Zm00001eb226390) were significantly deranged in Wi2. AtBBX18/19 regulate photomorphogenesis and hypocotyl elongation, a process dependent on microtubule orientation, by antagonizing abscisic acid signaling and binding E3 ubiquitin ligase COP1 (Wang et al. 2015; Bai et al. 2019), and their maize homologues Zm00001eb103370 and Zm00001eb319460 were concordantly down‐regulated in Wi2 and Wi3 shoot meristems (Table S2).

FIGURE 1.

Co‐differential Gene Ontology expression correlation analysis shows Wi2/+ and Wi3/+ shoot meristem tissues are affected in the same ways as hypothesized tubulin heterodimer binding partners. Data points represent statistically significant MapMan Wilcoxon rank sum overrepresented Gene Ontology bin Z scores for both mutants. The orange data points are cases where an independent Fisher exact test found bins “23” (vesicle trafficking [781 genes]) and “20.1” (ubiquitin proteasome [966 genes]) were significantly underrepresented for both up‐ and down‐regulated genes, indicating suppression effect on pathways; “11.3/4” (chromatin organization: histone modification/methylation/nucleosome remodeling [250/116/119 genes, respectively]) and “15.1” (pre‐mRNA splicing [273 genes]) and “16.1” (mRNA degradation [101 genes]) were significantly underrepresented for down‐regulation; whereas “50.3.1079” (membrane protein of unknown function; phytozome chromatin assembly factor 1 subunit A (CHAF1A) family [6 genes]) and “24.3.1” (solute transport.channel‐mediated Membrane Intrinsic Protein/MIP [27 genes]) were both overrepresented for significantly down‐regulated elements. See Table S5 for list of 80 overrepresented bins and numbers of genes per bin. Z scores for metabolism and process bins are calculated for all differentially expressed genes padj < 0.05, Benjamini‐Hochberg false positive adjusted for both, or either Wi2/+, or Wi3/+ mutants.

TABLE 2.

Significant transcriptional effects of Wi2 and Wi3 mutations on biological processes revealed by MapMan/PageMan analysis (Mercator ver 4.8 nomenclature) of overrepresented/enriched hierarchical proteome terms aligned with Kyoto Encyclopedia of Genes and Genomes Orthology/KEGG. See Table S2 for gene lists, and Table S5 for lists of Wilcoxon rank sum and select Fisher exact test statistics.

Bin, subprocess	Biological process	Genes quantified	Wi2/+, Wi3/+ Overrepresented padj‐value	Wi2 Test statistic “Z score” a	Wi3 Test statistic “Z score” a
12.1	Cell division. DNA replication	152	8E−13, 0	7.17	+ ∞
11.1.1	Chromatin organization.chromatin structure. DNA wrapping	68	8E−7, 0	5.03	+ ∞
13	DNA damage response	136	3E−6, 2E−6	4.72	4.88
12.4.3	Cell division.cytokinesis.phragmoplast microtubule organization	15	0.06, 3E−5	1.9	4.25
1.1	Photosynthesis.photophosphorylation	333	2E−12, 2E−35	−4.42	−4.47
22.1	Cytoskeleton organization.microtubular network	172	5E−10, 0	6.22	+ ∞
22.1.3	Kinesin microtubule‐based motor protein activities	68	9E−8, 0	5.36	+ ∞
22.1.2.12	Cytoskeleton organization.microtubular network. Kinesin microtubule‐based motor protein activities. Kinesin‐14	24	0.09, 8E−4	1.68	3.36
19.8.9	Protein control.chaperone activity.smallHsp holdase activity	23	3E−6, 2E−4	−4.71	−3.85
19.8.5.1.1	Protein control.chaperone activity.cytosolic Hsp70‐1/2/3/4/5	7	0.07, 0.05	−1.81	−2.00
6	Nucleotide metabolism	158	0.008, 0.02	2.65	2.40
5.7	Lipid metabolism.cytoplasmic lipid droplet‐associated activity	34	0.008, 0.005	−2.65	−2.86
18.4.1	Protein modification.phosphorylation. TKL protein kinase activity	501	9E−6, 0.04	4.46	2.13
18.5.1.2.1	Protein modification.ser/thr phosphatase activities.clade A Mn/Mg protein phosphatase	15	3E−5, 2E−4	−4.18	−3.74
10.1	Phytohormone action.abscisic acid	78	0.01, 0.01	−2.58	−2.54
14.5.4.5	RNA biosynthesis. DNA‐binding transcriptional regulation.other all‐alpha‐helix DNA‐binding. BBX family	32	0.002, 0.003	−3.21	−3.06
16.5.2	RNA homeostasis.mRNA silencing.transacting siRNAs	16	0.18, 0.05	1.35	2.00
14.5.3.1.1	RNA biosynthesis. DNA‐binding transcriptional regulation.helix‐turn‐helix. HOMEO domain. HD‐ZIP I/II	36	5E−4, 0.06	−3.50	−1.92
50.2.18.87	Putative activities. RNA processing.putative PPR‐type organelle RNA editing factor	166	2E−10, 0	6.42	+ ∞
29.1.2	Plant organogenesis.leaf formation.abaxial polarity	19	0.14, 0.005	1.49	2.80
24.1.3	Solute transport.primary active. ABC‐type transporter family	88	0.02, 0.04	2.41	2.09
24.3.1	Solute.transport.channel‐mediated. MIP‐type intrinsic activity	27	0.002, 0.22	−3.10	−2.75 b
23¶	Vesicle trafficking; up‐regulation down‐regulation	781	0.93, 0.45	−3.18 −7.85	−6.88 −3.80
20.1¶	Protein homeostasis.ubiquitin‐proteosome; up regulation down regulation	966	0.81, 0.06	−7.09 −6.22	−2.34 −2.97
11.3.1¶	Chromatin organization.histone modification.methylation; down	116	0.38, 0.57	−2.10	−2.01
11.4¶	Chromatin organization.nucleosome remodeling; down	119	0.89, 0.90	−3.36	−3.07
50.3.1079¶	Unknown.membrane protein/chromatin assembly factor1A‐like; up	6	0.33, 0.004	2.46	2.99
16.1¶	RNA homeostasis.mRNA degradation; down regulation	101	0.50, 0.68	−2.72	−2.44
15.1¶	RNA processing.spliceosome‐mediated pre‐mRNA; down	273	0.80, 0.74	−4.15	−2.44

^a Subbin Wilcoxon rank sum over representation analysis of bin values; setting = 1 (L2FC > 1.0) and Benjamini–Hochberg multiple testing corrected padj. The algorithm tests the median fold‐change (overall positive or negative for bin) within the ontological group against the median fold‐change of all genes not in that group. Z score ∼standard deviations from median.

^b Fisher exact test result for overrepresented down regulation; padj shown for Wilcoxon rank sum result.

Fisher exact test results showing those large bins significantly underrepresented, confirmed by nonsignificant padj values, indicating these pathways are suppressed in Wi2 and Wi3 mutants. See Figure 1 and Table S4.

Remarkably, Fisher exact tests conducted on very large gene bins “vesicle trafficking” (781 genes; Figure S6) and “ubiquitin/proteosome protein homeostasis” (966 genes) showed dual significant underrepresentations for both up‐ and down‐regulated genes in Wi2 and Wi3 mutants (Figure 1 and Table 2). One possibility is that these correlated transcriptional signatures reflect altered regulation of two tightly linked homeostatic pathways: histone H2B mono‐ubiquitination affecting tubulin depolymerization (Zhou et al. 2017), mono‐ubiquitination otherwise directing transmembrane proteins into vesicles (endocytosis), and poly‐ubiquitination marking proteins for stress‐induced sorting and/or degradation at the proteosome or endosomal/vacuolar/multivesicular bodies before or after transport (Zientara‐Rytter and Sirko 2016; Yu and Xie 2017). As an isolated case in point: the exocytic vesicle trafficking component Drought‐Sensitive1/Zm00001eb065570/EXO70A that functions in xylem vessel development (Zhu et al. 2026) was significantly up‐regulated in Wi2/+ shoot meristems, but down‐regulated slightly in Wi3/+ MT (Table S4). Three other examples of systems exploration that identified trafficking and histone remodeling, or ubiquitin‐related genes significantly deranged in Wi2 and Wi3 were up‐regulation of Zm00001eb082920/HISTONE‐LYSINE N‐METHYLTRANSFERASE ASHH1; its homologue SETD2/HUNTINGTIN‐INTERACTING‐P1 in animals binds clathrin, is phosphorylated by receptor tyrosine kinases, and is involved in endocytosis (Koonin et al. 2004; Ames et al. 2013; Bettencourt et al. 2021) (https://www.ncbi.nlm.nih.gov/Structure/cdd/cddsrv.cgi?uid=KOG4442). For ubiquitin‐related genes, we identified down‐regulation of Zm00001eb059010 annotated as xylose isomerase but with an N‐terminal tetratricopeptide‐like domain homologous to retrograde trafficking COPI coatomer subunit epsilon, and significant up‐regulation in Wi2 and Wi3 mutants of UTG8‐binding cargo autophagic receptor RPN10/Zm00001eb026350, a conserved component of the 19S regulatory complex of the 26S proteasome as deranged expression outliers (Marshall et al. 2019) (Table S4). We recognize that hypothetical mechanisms suggested by the transcript‐level data are drawn from statistically buffered datasets inferring mRNA abundance does not directly report pathway activity. Thus, establishing whether vesicle trafficking or proteasomal function is altered would require direct protein‐level or imaging evidence (e.g., vesicle‐marker or CESA trafficking dynamics). Notwithstanding, a recent publication showed that DOS11/Wi3/TUA4 is a direct effector and marker of vesicle trafficking to control asymmetric cell divisions impacting cell plates and abnormal cell wall deposition, among other insights into the role of tubulin partners in stomatal development (Duan et al. 2026).

3.2. Wi2 and Wi3 Mutants Have More Abundant per Area But Smaller VBs in Internode Pith and Altered Cell Wall Compositions Revealed by Histology and Fourier Transform Infrared/FTIR Spectroscopy

Histological examination of cross‐sections of sixth internodes in Wi2 and Wi3 compared to WT sibs showed that heterozygous mutants had smaller VBs and a dearth of defined protoxylem cell walls at the expense of abundant lacunae (Figure 2). To further examine the histochemical properties of the VBs, fresh sixth internodes from 60‐day‐old field‐grown plants were subjected to histochemical staining using a range of specific dyes. Cross‐sections (90 μm thick) were prepared using a vibratome from segregating Wi2 and Wi3 MT lines and their WT sibs. Bright‐field imaging of unstained sections confirmed that VBs in the Wi2 and Wi3 MT lines were smaller than those in the WT siblings when observed under the same magnification (Figure 3A; below).

FIGURE 2.

Light (and UV autofluorescence; lowest panel) micrographs of Wi2/+ and Wi3/+ mutant and normal sibling vascular bundle (VB) histology from cross sections of internode 6, 106 d.a.s. stained with fast green (phloem, cellulosic cell walls, and cytoplasm) and safranin O (lignified xylem walls). Wild type panel is labeled for cell types: ph, metaphloem (arrowhead); mxv, metaxylem vessels (arrows); x, tracheary element; la, lacuna; obl, obliterated protophloem. Note smaller VBs in Wi2 and Wi3 mutants. Panels are 100× mag scale bar ~200 μm.

FIGURE 3.

Histochemical staining of Wi2/+ and Wi3/+ and Wild type siblings of fresh cross sections from internode six of 60‐day‐old field‐grown plants. Fresh‐cut stems were sectioned 90 μm thick with a vibratome, stained and visualized under bright field (A–C) and UV epifluorescence (D–F). (A) VBs under bright field without any stain. (B) Maüle's stain, specific for lignin S subunit (purple) and G subunit (brown). (C) Phloroglucinol stain, which stains H lignin red. (D) Aniline blue stain. (E) Calcofluor white stain. (F) autofluorescence visualized under UV light excitation. Scale bars ~400 μm (10 × magnification) and ~200 μm (20× magnification).

Phloroglucinol (Wiesner) staining, which detects lignin‐associated p‐hydroxycinnamyl aldehyde groups and produces a characteristic cherry‐pink coloration in lignified tissues (Adler 1977), showed slightly weaker brightfield staining in Wi2 and Wi3 MT VBs compared to WT sibs (Figure 3C). Similarly, Mäule staining, which distinguishes syringyl (S) from guaiacyl (G) lignin units (Mitra and Loqué 2014), also exhibited reduced coloration in the MT VBs (Figure 3B). To assess other cell wall components, aniline blue staining, which detects callose (β‐1,3‐glucans), was visualized under UV epifluorescence and showed stronger fluorescence in the Wi2 and Wi3 MT VBs relative to WT when normalized to UV autofluorescence, suggesting increased callose deposition (Figure 3D and Figure S10). Calcofluor white staining, which binds to cellulose and other β‐glucans in plant cell walls, also produced slightly stronger fluorescence in the MTs, indicating higher cellulose content in the MT VBs (Figure 3E and Figure S10). Unstained sections were additionally examined for UV autofluorescence, which primarily arises from natural fluorophores such as lignin in plant cell walls and could be an independent blue wavelength autofluorescence normalization parameter. Under identical imaging conditions, Wi2 and Wi3 MT VBs exhibited lower autofluorescence compared to WT VBs, consistent with apparent reduced lignin accumulation in the mutants (Figure 3F and Figure S10). Together, these results suggest that Wi2 and Wi3 MT VBs exhibit altered cell wall composition, particularly with respect to lignin and callose deposition. These histochemical comparisons are semiquantitative and provide indirect support for altered wall composition rather than absolute quantification of individual wall polymers.

Given that Wi2 was previously characterized as having decreased inner diameters of protoxylem vessels (Huang et al. 2024), we quantified with ImageJ the proto‐ and meta‐xylem cross‐sectional areas in our histology specimen bundles showing clear secondary walls (not lacunae, which were common: lysigenous cavities from dissolution and disintegration of the cell that leaves a hole with only remnants of wall; see Figure 2 “Mutant” column) from 106‐day‐old field‐grown plants and confirmed the protoxylem had smaller sizes for Wi2, as well as for Wi3 heterozygotes (Figure 4) which is in line with the claim of Huang et al. (Huang et al. 2024) for significantly smaller Dos1/+ protoxylem diameters versus WT assayed 30 days after germination. Similarly, we examined the cortical pith cell size, VB size and metaxylem vessel sizes. We found that there was little difference in cortical pith cell sizes, but significantly smaller VBs, metaxylem, and protoxylem cell sizes for both Wi2 and Wi3 heterozygous mutants compared to WTs. Surprisingly, even with relatively larger VBs for Wi3 genotypes (possibly due to being older), MT Wi3 heterozygotes had significantly smaller metaxylem compared to Wi2 heterozygote mutants (Figure 4C).

FIGURE 4.

Violin plots with overlaid boxplots for comparison of anatomical cell sizes compared among four genotypes. (A) Pith cortical cells, n = 576; (B) Vascular bundle size; n = 26; (C) Metaxylem vessel cell size, n = 20; and (D) Protoxylem cell size, n = 20 in cross‐sections of sixth internode histology specimens from segregating sibs of Wilty2/+ and Wilty3/+ heterozygotes in the field. Group differences were assessed using Kruskal–Wallis rank sum test with Bonferroni correction. Statistical significance (different lower case letters) was determined at α = 0.05.

Because the cell wall components of the VBs of the two Wilty mutants showed significant differences for some or the other histochemical stain used above, we performed FTIR–universal attenuated total reflectance/UATR analysis of the pith‐extracted internode VBs obtained from the process of retting on Wilty2/3/4 heterozygous field‐grown mutants with wi1 (lacks metaxylem vessels) serving as the reference positive control. The FTIR–Universal Attenuated Total Reflectance/UATR spectroscopy is a proven method to identify specific functional groups in polysaccharides extracted from plant cell walls like cellulose, hemicellulose, pectin, and various phenolic acid groups including lignin analyzed across 1800‐ to 800‐cm⁻¹ range (Liu et al. 2021). The difference spectra of wi1 versus WT showed two peaks: a decreased CH=CH bending of lignin (wave # 920‐983) and increased C‐O stretch in cellulose (wave # 1062; Figure 5) (Chen et al. 1998; Alonso‐Simón et al. 2011). The FTIR difference spectra revealed consistent patterns of cell wall physico‐chemistry across the Wi2, Wi3, and Wi4 MT VB powders relative to their WT sibs, distinct from the positive control wi1 (Figure 5). Reduced peaks at ~971 cm⁻¹ (Wi2) and ~965 cm⁻¹ (Wi3), associated with arabinose/pectin components (Kline et al. 2010) suggest lower levels of these polysaccharides in the MT VBs. In contrast, elevated peaks corresponding to cellulose C–O stretching were observed at 1020/1063 and 1109 cm⁻¹ (Wi2), 1018/1059 cm⁻¹ (Wi3), and 997/1021/1056 cm⁻¹ (Wi4), indicating increased cellulose content in the MT cell walls (Alonso‐Simón et al. 2011). Across all three dominant mutants, negative peaks at ~1231–1235 cm⁻¹ correspond to reduced guaiacyl (G) lignin C–O stretching, whereas positive peaks at ~1319–1320 cm⁻¹ suggest relatively higher syringyl (S) lignin content (Kline et al. 2010). Additionally, negative peaks in the range of ~1702–1740 cm⁻¹, including peaks at 1732 cm⁻¹ (Wi3) and 1738 cm⁻¹ (Wi4), indicate reduced phenolic esters and esterified uronic acids associated with cell wall polysaccharides (Alonso‐Simón et al. 2011). A distinct peak at 1656 cm⁻¹ observed in Wi4 may correspond to the amide I band of proteins (Alonso‐Simón et al. 2011). Principal component analysis (PCA) of the FTIR‐ATR spectra separated MT and WT samples into distinct clusters, indicating clear differences in cell wall composition between genotypes (Figure 5, insets). The overall FTIR spectral profiles of Wi2, Wi3, and Wi4 were highly similar, supporting the histochemical staining results (Figures 2 and 3) and suggesting that VB cell walls are similarly altered across these dominant Wilty mutants. We emphasize that these FTIR and histochemical data are correlative and provide indirect evidence of altered wall composition; they do not establish a causal link between the tubulin mutations, vesicle trafficking, and wall deposition. Moreover, while the close spectral similarity of Wi4 to Wi2 and Wi3 is consistent with a shared cell‐wall defect, it does not by itself establish the molecular identity or function of the Wi4 gene.

FIGURE 5.

(above) Fourier transform infra‐red attenuated total reflectance difference spectra of powdered pith VBs isolated from retting the internodes of Wi2 and Wi3 mutant and wild‐type siblings. Elevated wave number ~1060 absorbance suggests increases in cellulose in mutants, decreased G lignin and esterified uronic acid (wave # 1231 and 1738, respectively, increased S lignin (wave # 1320), increased pectin (wave # ~1018), and decreased CH=CH bending of lignin (wave # 920‐983). Insets: Principal component analysis of datasets show good reproducibility (green = mutant; pink = wild type), n = 11).

3.3. Alpha Tubulin4 (TUA4) Identified as the Causal Wi3 Gene C➔ T E196K Mutation Mapped to Short Arm of Chromosome 5

For identification of the causal mutation in Wi3, two genomic libraries were prepared with pooled genomic DNAs (Figure S3 and Table S1). Based on the delta SNP index obtained from the QTLseqr, the Wi3 mutation was mapped to short arm of chromosome 5 where the raw data showed a significant peak above the 99% confidence range from coordinates 5–~22 Mbp (Figure 6). This interval contained 214,519 SNPs/Indels in the raw data and 27,277 SNPs/Indels after filtering out Hapamp2 (Chia et al. 2012) and HapMap3 (Bukowski et al. 2018) known SNPs/Indels. Because EMS mutagenesis primarily induces GC:AT transitions in plants, we filtered out all the indels from the mapping interval to obtaine 10,425 de novo SNPs, which were inherited as heterozygous SNPs for the Wi3 MT parent and homozygous reference for the WT sib/child in a parent–child–trio analysis with B73 short reads. Next, the SnpEff annotation of all de novo SNPs annotated as intergenic was filtered out, which left 5607 SNPs in the mapping interval. Out of the 5607 de novo SNPs, there were 18 high, 42 low, and 214 moderate de novo SNPs for a total of 274 SNPs, which impacted the gene structure of various genes within the mapping interval. We had BSA‐seq data from the other maize mutants available in the lab namely, Wi2 MT, Wi2 WT sib (Dhingra 2021), Wi4 MT, Wi4 WT sib (data not shown), alb‐605F (albino) MT and WT sib (Dhingra 2021), and wi1 MT (data not shown). Thus, we reasoned that the candidate causal SNP should be unique and not found in any of the above data sets to be inherited in a heterozygous manner. Thus, we filtered out all the SNPs inherited in a heterozygous manner in the above‐mentioned data sets for chromosome 5. After this filtration step, we obtained 11 moderate de novo SNPs in different genes within the mapping interval (Table 3).

FIGURE 6.

The Wi3 mutations (lower panel; upper right) are mapped to short arm of Chr. 5, respectively, after filtering out the known SNPs/Indels of hapmap2 and hapmap3 from the raw data. Single peaks in a ~2 Mbp plateau for Wi3 are obtained with more than 99 percentile confidence intervals. [upper panel deleted in revision].

TABLE 3.

Quality metric parameters for causal Wi3 candidates on chr5 called by GATK on BSA short read libraries, identified by SnpEff software. Bold shows read depth evidence for cloning claim, based on causal candidate alternate or ref. SNP with no evidence of recombination in WT and mutant mapping library reads.

POS	REF	ALT	Wilty_3_MT	Wilty_3_WT	Type of mutation	Score	Gene id	Gene description	SNP position substitution	Amino acid substitution
2478958	C	T	0/1:98,21:119:99:329,0,2888	0/0:44,0:44:99:0,112,1710	missense_variant	MODERATE	Zm00001d012969	NEDD8‐activating enzyme E1 regulatory subunit AXR1	c.325G>A	p.Glu109Lys
4706889	G	A	0/1:1,2:3:36:40,0,36	./.:0,0:0:.:0,0,0	missense_variant	MODERATE	Zm00001d013100	Myosin‐binding protein 7	c.487G>A	p.Val163Met
6082910	C	T	0/1:6,5:11:99:129,0,159	0/0:24,0:24:36:0,36,765	missense_variant	MODERATE	Zm00001d013182	BTB/POZ domain‐containing protein	c.280G>A	p.Val94Met
6149796	G	A	0/1:69,42:111:99:0|1:6149794_C_T:1553,0,3064	0/0:83,0:83:99:.:.:0,101,1800	missense_variant&splice_region_variant	MODERATE	Zm00001d013188	Seven transmembrane domain protein	c.632G>A	p.Gly211Asp
6724042	C	T	0/1:8,5:13:99:122,0,216	0/0:17,0:17:3:0,3467	missense_variant	MODERATE	Zm00001d013223	Retrovirus‐related Pol polyprotein LINE‐1	c.1462G>A	p.Ala488Thr
7406529	C	T	0/1:63,47:110:99:0|1:7406520_C_T:1851,0,2540	0/0:38,0:38:99:.:.:0,102,1197	missense_variant	MODERATE	Zm00001d013265	MYB‐related‐transcription factor 106 description: Transcriptional adapter ADA2b	c.977G>A	p.Gly326Asp
7529463	C	T	0/1:71,33:104:99:791,0,2060	0/0:33,0:33:99:0,99,1048	missense_variant	MODERATE	Zm00001d013271	RNA binding (RRM/RBD/RNP motifs) family protein	c.352G>A	p.Glu118Lys
7711359	C	T	0/1:69,61:130:99:1538,0,1912	0/0:36,0:36:99:0,99,1485	missense_variant	MODERATE	Zm00001d013278	PHD‐transcription factor 25 description: Acyl‐CoA N‐acyltransferase with RING/FYVE/PHD‐type zinc finger protein	c.2609G>A	p.Arg870Lys
9836290	C¶	T*	0/1:38,30:68:99:781,0,1031	0/0:34,0:34:99:0,99,1243	missense_variant	MODERATE	Zm00001d013367	Alpha tubulin4 description: Tubulin alpha‐4 chain	c.586G>A	p.Glu196Lys
20825393	C	T	0/1:24,5:29:56:56,0,618	0/0:54,0:54:25:0,25,1502	missense_variant	MODERATE	Zm00001d013802	Protein coding	c.97G>A	p.Ala33Thr
22724930	G	A	0/1:34,7:41:76:76,0,976	0/0:68,0:68:99:0,104,1800	missense_variant	MODERATE	Zm00001d013854	Protein coding	c.100C>T	p.Pro34Ser

^a Causal candidate SNP showing no evidence of recombination in deeply interrogated mapping pool libraries; see VCF output in Table S6.

^b Allele validated by Sanger sequencing of PCR amplicon. See Figure S7.

In this interval, we identified a candidate SNP in alpha tubulin4 (TUA4) (Zm00001d013367/Zm00001eb215710) harboring C‐to‐T missense mutation at ver4 coordinate 9,836,290 bp (Table S6). The causal SNP changed glutamic acid residue 196, a negatively charged amino acid to lysine, a positively charged amino acid in the TUA4 protein. The causal SNP was annotated to the Tubulin FtsZ/GTPase binding domain (residues 3–214; InterPro PF00091) and an alpha helix highly conserved in plants and animals (Figure S9) (Nogales et al. 1998; Kelley et al. 2015).

To rule out other possible candidate genes within the mapping interval, we performed PCR for the informative Indels mined from the GATK calls. Primers (Table S3) were designed around the Indel markers mapping between the causal SNP located in the single top candidate gene ZmTUA4 and other flanking candidate genes. Three primer pairs flanking Indels near the candidate gene ZmTUA4 revealed increasing homozygosity for the WT amplicons closer to the candidate gene (Figure 7). Wi3‐P2 primer pair maps 138,290 bp north from the causal SNP and showed no evidence of recombination break points in all four WT sib pools of ~25 individuals, whereas, in the Wi3 MT pool, two bands are seen as evidence of heterozygous segregation of the two alleles.

FIGURE 7.

Position of primers designed flanking Indels of > 15 bp nearby the candidate causal gene. For Primer pair W3‐P2, which is located 138,290 bp away from the causal SNP, there was no evidence of recombination breakpoints in any mutant pools, which show only a single amplicon wild type reference band from genomic DNAs. Yellow stars represent recombination breakpoints in wildtype pools.

Notably, the causal mutation identified in Wi3‐N1614 corresponds to a C → T transition in TUA4, resulting in the E196K amino acid substitution independently described as ZmDrought‐overly‐sensitive1 (Huang et al. 2024). Table S6 lists an analysis of > 168 cases of Wi3‐adjacent Mo17 absolute SNP linkage disequilibria spanning eight north breakpoints (from 20 to 116 kbp away from causal) and at least nine south recombination breakpoints (from 8 to 116 kbp from causal) documented in our Wi3 Variant Calling Format—GATK parent–child–trio BSA output, capturing high‐confidence absolute genotype REF calls for B73 homozygous (non‐Mo17) reference, wild‐type sib homozygous non‐Mo17 reference, and Wi3/+ ALT heterozygous Mo17‐unique provenance SNPs. VCF2Dif analysis (Xu et al. 2025) also showed that the Wi3 MT segregating pool shares 119 Alt SNPs with Mo17 with the 50 kb region upstream and downstream of the causal mutation (Figure S8B), providing additional evidence for its provenance. Within the query interval for Wi3, Mo17 shows 94.8% (276/291) similarity in ALT SNPs. These cases (Table S6) prove that Dos1 is Wi3, because in the absence of evidence that Dos1 was isolated from B73 genotype as claimed (Huang et al. 2024), the cultivar provenance of the sole Wi3 allele, identical to Dos1 causal C➔ T SNP, is Mo17 as reported (G. Neuffer, pers. comm. 2018). Although it is theoretically possible Dos1 is an independent isolate of Wi3 ^E196K, the Wi2‐N1540 and Wi3‐N1614 mutants are available from the Maize Genetics Stock Center and seeds carrying these mutants have been distributed to several researchers.

For Wi2, the evidence for provenance‐traceable relationships with 1500 sequenced accessions (Andorf et al. 2025) was stronger than for the Wi3‐ specific SNPs, with perfect concordance for ~110–150 SNPs on either side flanking the causal Wi2 mutation. The highest similarities were with the maize lines ND167 and B91, which shared ~90% of ALT SNPs (607/670 overall in ~190 kbp interval for ND167); these lines and nearest‐identified neighbor line Mo13 were sequenced by the University of Nebraska‐Lincoln. This supports that the Neuffer “mongrel ACR” stock referenced in Maize Genome Database annotations for numerous characterized Neuffer batch‐correlated loci (e.g., N1330, N1339A, N1528, N1530, N1544, and N1687) and ND167/B91 stocks are closely related genetic germplasms (Figure S8A). For lines ND167 and B91, we found the first differences for documented high confidence SNPVersity ALT alleles, compared to B73, was at 12,219 bp distal north (113 SNPs spanned, for both ND167 and B91), and 29,061 bp distal south (156 SNPs spanned for ND167 and 108 spanned for B91) from the causal Wi2 SNP mutation (Table S7). On the other hand, the genotype for Mo13 SNPs did not maintain such long stretches of matching ALT alleles compared to Wi2/+ SNP library calls. Thus, the provenance for these two publicly available lines is likely derived from, or share a common progenitor, with “mongrel ACR” stock used to generate the Wilty2‐N1540 mutation, based on perfectly concordant nearby flanking SNPs versus the next nearest Nebraska‐sequenced line Mo13 (Table S7 and Figure S8A).

There are 15 paralogues of Wi2/tubulin beta6 (which has three isoform splicing variants) and seven paralogues of Wi3/tubulin alpha4 (two isoform splicing variants) annotated in B73 maize. It was of interest to examine the BSRseq transcript variant levels for the mutated genes to find that Wi2 transcript splice variant T002, which has a different 5′ intron1 donor site seven nucleotides upstream from the primary transcript “GT” donor in the annotated 5′ UTR, was down‐regulated significantly in Wi2 meristems, but up‐regulated significantly in meristems of Wi3 heterozygote MT (Table S2, rows 10–16), as was Wi3 primary transcript, raising questions about structure:function of Wi2/3 variants and feedback regulation across tubulin genes. There is evidence for a “tubulin economy” in plants and animals, which adjusts transcriptional up‐expression of related tubulin genes in a cell type‐specific manner when specific isotypes are disrupted (Fertuzinhos et al. 2022).

4. Discussion

It is known that presence/absence variants are more prevalent than anticipated in inbred temperate maize stocks (Darracq et al. 2018) and a substantial portion of variation lies outside the principal maize B73 reference genome (Hirsch et al. 2014), a potential “weak link” for BSA‐seq approaches to clone novel mutants of unknown genetic provenance. Our claim that Dos1 is Wi3 relies on the fact that the provenance of Wi3 from mutagenesis of Mo17 stock (Table S6 and Figure S8B) is known, as is the provenance of Wi2 from a “mongrel ACR” stock (G. Neuffer, pers. comm.), which we infer from Figure S8 should be most closely related to highly inbred, Germplasm Resources Information Network/GRIN‐ available traditional nonstiff stalk group varieties B91 (PI 527700), ND167 (NSL 32726), and Mo13 (PI 558529) (Hansey et al. 2011; Mural et al. 2022) (Figure S8A).

Alpha and beta tubulin polymer proteins make up the dynamic array of microtubules, which reorganize in response to environmental and developmental cues. Several studies have reported mutants of tubulin in Arabidopsis to have twisted shoot and root growth with specific preference for left/right handedness (Thitamadee et al. 2002; Ishida and Hashimoto 2007; Ishida et al. 2007; Nakamura and Hashimoto 2009). In rice the dominant Tid1–1 (Twisted dwarf 1–1), MT of TUA exhibited a twisting and dwarf phenotype (Sunohara et al. 2009). Huang et al. (Huang et al. 2024) demonstrated that ectopic expression in Arabidopsis of MT maize allele α‐tubulin4 (ZmDOS1) alters microtubule array orientation, underscoring its key role in microtubule organization and secondary cell wall deposition.

In this study, we provide evidence of cloning of Wi3 dominant MT E196K of maize, which manifests a wilting phenotype even in well‐watered conditions. Cloning evidence was facilitated by employing the “EVIL twin” method (Addo‐Quaye et al. 2017) to filter out SNPs, which were heterozygous in wi1 MT, albino MT, wi1/alb WT sib, Wi2 MT, Wi2 WT sib, Wi4 MT, and Wi4 WT sib to reduce background SNP noise in highly polymorphic maize. Wilting phenotypes caused by defects in ABA hormone biosynthesis or response have been reported in sunflower (Pugliesi et al. 1994), lotus (Biswas et al. 2009), tomato (Tal 1966; Taylor et al. 1988; Burbidge et al. 1999), tobacco (Rousselin et al. 1992), capsicum (Tal et al. 1974), potato (Jong et al. 2001), and pea (McAdam et al. 2015) in addition to model plants Arabidopsis and maize. However, it has become clear from recent reports that defects in VB development can also lead to a wilting phenotype in plants. Interestingly, the Wi3 transcriptome data showed high similarities to Wi2 transcriptome dataset (Figure 1 and Table 2) for genes differentially repressed in the same biological processes, in particular large gene sets comprising vesicle trafficking and the ubiquitin‐proteasome system (UPS), both which play critical, regulatory roles in controlling the stability and turnover of proteins associated with the microtubule cytoskeleton, that is, hypocotyl elongation (Lian et al. 2017). UPS primarily functions by degrading microtubule‐associated proteins and controlling microtubule dynamics. While tubulin is not ubiquitinated, the proteins that bind to it are, allowing the cell to rapidly change microtubule organization. All paralogous and orthologous α and ß tubulin gene pairs in poplar show evidence of strong purifying selection, consistent with expression data showing tubulin family members are preferentially expressed in root, leaf, or stem, whereas TUB paralogs exhibit redundancy in expression patterns (Mao et al. 2025). In cotton, the five characterized TUAs have been reported to be highly expressed during seed fiber development and during the onset of fiber secondary cell wall synthesis (Whittaker and Triplett 1999).

Microtubules play an important role in marking the site of CSC at the plasma membrane (Crowell et al. 2009; Gutierrez et al. 2009). The microtubule organization 1 (mor1) MT of Arabidopsis has decreased density of cortical microtubules and increased crystalline cellulose abundance (Fujita et al. 2012). Histochemical staining for lignin using Wiesner/phloroglucinol: HCl and Maüle stains (Figure 3) and FTIR analysis of VBs (Figure 5) support that Wi3 and Wi2 mutants had lower lignin content in VBs consistent with the role of cortical microtubules in biosynthesis of hemicellulose and lignin monomers produced in the Golgi complex (Farquharson 2018). Takenaka et al. (Takenaka et al. 2018) demonstrated that treating a CESA7 (CELLULOSE SYNTHASE SUBUNIT 7) MT with microtubule depolymerizing drug oryzalin led to disruption of helically patterned xylan and lignin deposition during the process of differentiation of protoxylem vessel cells.

Mutations in both α‐ and β‐tubulin are expected to produce similar downstream effects on vesicle trafficking pathways because these proteins function as obligate heterodimers that form the fundamental building blocks of microtubules. Proper microtubule assembly depends on tightly coordinated folding, dimerization, and polymerization of α/β‐tubulin; therefore, disruption in either subunit can compromise microtubule integrity, dynamics, and organization. Such defects may directly impair the ability of motor proteins, including kinesins, to bind and move along microtubule tracks, potentially altering vesicle transport. In most documented cases, these mutations exhibit dominant‐negative effects, where incorporation of defective tubulin subunits disrupts the function of the entire microtubule network. Consistent with this, studies in diverse systems have shown that mutations in α‐tubulin (e.g., TUBA1A) and β‐tubulin (e.g., TUBB2B and TUBB3) result in comparable defects in vesicle trafficking, particularly in animal neurons, where they lead to developmental abnormalities and cellular dysfunction (Niwa et al. 2013; Romaniello et al. 2015; Buscaglia et al. 2020; Hoff et al. 2022; Qu et al. 2025). Our transcriptome data are consistent with this expectation and the recent results of Duan et al. (2026) but do not by themselves establish altered vesicle transport in Wi2 or Wi3; direct protein‐level or imaging evidence (e.g., vesicle‐markers or CESA trafficking dynamics) would be required to confirm trafficking changes. It is noteworthy that maize leaf‐rolling and short internode MT swl3 encodes a predicted esterase/lipase localized to the endoplasmic reticulum (Li et al. 2026) that positively regulates drought tolerance, making it an excellent candidate marker for tubulin‐mediated vesicle trafficking in xylem internode development.

In conclusion, we identified the canonical Mo17 cultivar as source of Wi3/Dos1 semidominant MT allele of alpha‐tubulin4 (TUA4). Functional validation of TUA4 as the causal gene for this lesion has already been established: Huang et al. (Huang et al. 2024) recapitulated the dominant Dos1 phenotype by transgenically expressing the MT genomic allele (ZmDOS1pro:ZmDos1) under its native promoter in a WT background, the definitive test for a dominant gain‐of‐function allele. Because Wi3 carries the identical TUA4 C‐to‐T (E196K) substitution mapped here by BSA‐seq, this published validation applies directly, and no additional complementation or CRISPR recreation is required to establish causality. Because two of the three dominant Wilty mutants of maize are tubulin heterodimer mutants, we speculate based on our FITR results (Figure 5) that Wi4 may encode a tubulin gene, consistent with Wi4 allele conditioning a semidominant phenotype.

Author Contributions

Conceptualization, methodology, funding acquisition: Chris Rock. Material preparation, data collection: Anuradha Dhingra, Indra Adhikari, Chris Rock. Formal analysis and investigation: Anuradha Dhingra, Chris Rock, Indra Adhikari. Writing – original draft preparation: Anuradha Dhingra, Chris Rock. Writing – review and editing: Chris Rock, Indra Adhikari, Anuradha Dhingra.

Funding

Partial financial support was from the USDA‐TTU Cooperative Agreement: Ogallala Aquifer Program Project #3090‐13000‐015‐11S “Water Management to Sustain the Economic Activity from the Ogallala Aquifer on the Southern High Plains.”

Conflicts of Interest

The authors declare no conflicts of interest. C.D.R. did consultancy work for Richard Bagdasarian Inc., resulting in an unrestricted gift to the TTU Foundation for use by C.D.R., which defrayed some costs of reagents and fee‐for‐service sequencing.

Supporting information

Table S1: gDNA BSA‐seq and RNA libraries quality control parameters.

Table S2: Primary transcript gene‐level sense stranded RNA‐seq differential expression analyses of Wi2/+ and Wi3/+ genotypes versus WT sibs (sorted by shallower library Wi3 DE decreasing padj).

Table S3: Sequence of the primers used for Indel walking and genotyping Wi2 and Wi3 mutant alleles.

Table S4: Fisher exact test extreme cases in ontology bin “vesicle trafficking” significantly underrepresented (less than expected by chance) for both up‐ and down‐regulated genes, supporting that vesicle trafficking is co‐suppressed by Wi2 and Wi3 mutations.

Table S5: Wilcoxon rank sum and fisher exact test overrepresented bin statistics from gene‐level DE results of MapMan analysis of Wi2/+ and Wi3/+ genotypes.

Table S6: Proof that the Wilty3 allele E196L originated from the Mo17 cultivar background, based on parent–child–trio Variant Calling Format SNPs within ~116 kbp proximal and distal to the Wi3 causal mutation, spanning 19 recombination breakpoints and validated for every case of 168 known Mo17‐specific SNPs.

Table S7: Evidence that the Wilty2 background EMS mutagenized stock “Neuffer mongrel ACR” is most closely related to Germplasm Resources Information Network/GRIN‐ available traditional nonstiff stalk group varieties ND167 (NSL 32726) and B91 (PI 527700).

Table S8: Quantification of the fluorescence obtained from the UV‐specific stains and autofluorescence of unstained stem cross section tissues using ImageJ software.

Figure S1: The Wi2/+ and Wi3/+ mutant phenotypes versus wild type siblings (+/+). The WT sibs have grown to full height and do not show any symptoms of wilting under well‐watered conditions. The Wilty2/+ and Wi3/+ individuals show reduction in height and severe leaf wilting and leaf tip scorching phenotypes. The elongation of the mutant stalk is stunted with a corresponding reduction in biomass at maturity. A yardstick (0.914 m) is placed between the WT and mutant plant specimens for scale.

Figure S2: Meristem tissues used for source of genomic DNA and total RNA extractions for BSA‐seq and BSR‐seq libraries of Wilty2 (upper panel) and Wi3 (lower panel; note tassel organogenesis) mutant and WT sibs. The shoot of the plants was bisected longitudinally to locate and dissect the actively dividing basal shoot meristem region (highlighted inset box).

Figure S3: The workflow for bulk‐segregant whole genome resequencing (BSA‐seq). Lines segregating for the Wi3 phenotypes were back crossed into the reference genome B73 stock background for three generations (BC₃). For BSA‐ and BSR‐seq, segregating plants obtained from BC3 were grown in field to generate the mapping population.

Figure S4: Vascular bundles (VBs) isolated from internodes by the process of retting.

Figure S5: Principal component analysis (PCA) of the Wi2/+ and Wi3/+ and WT sib sense strand RNA‐seq library runs analyzed at gene and batch level.

Figure S6: MapMan pathway analysis “X4.5 Vesicle trafficking R5.0” (modified from https://mapman.gabipd.org/mapmanstore) in “Clusterview” showing suppression of differential expression in both Wi2 (left datapoints in horizontal graphs) and Wi3 (right datapoints) shoot meristems. Clusterview shows each individual gene in a sub‐bin in gray, and the mean behavior of all genes within that cluster as bold red lines flanked by 95% confidence interval red lines. The vast majority of bin 23 genes are less differentially expressed in both Wi2 and Wi3 mutant meristems than the median fold‐change of all genes not in that ontological bin. Z scores on the y‐axes of graphs ∼standard deviations from median.

Figure S7: pld370189‐sup‐0002‐Supplementary_Material.docx. Wilty3 allele Sanger sequencing. The G‐A missense mutation (Glu196‐Lys) is highlighted with the red circle. Other known SNPs are annotated manually from MaizeGDB (https://maizegdb.org/).

Figure S8: (above) Phylogenetic clustering of the Wi3 and Wi2 candidates‐centered local haplotype across selected maize accessions using the nearest 25 shared SNP overlap sites around the converted candidate interval on NAM5 Chr5 (Wi3) and Chr3 (Wi2). Branch length represents mean genotype mismatch across these local SNPs. Labels include each accession and run ID together with the number of matching sites among the nearest 10 and 25 local SNP's (n10/n25) and within the 50 kb candidate‐centered window (not for Wi2 [panel A] because of higher mismatches).

Figure S9: pld370189‐sup‐0002‐Supplementary_Material.docx. Zea mays Alpha Tubulin4 (TUA4) predicted secondary structure (green alpha helix; blue beta sheet) of Wi3 MT protein sequence obtained using Protein Homology/AnalogY Recognition Engine/Phyre2 software⁹⁵. The causal SNP substitution glutamic acid to lysine (K) is indicated by star.

Figure S10: Quantification of fluorescence obtained from UV‐specific stains and autofluorescence of unstained stem cross‐section tissues of Wi2 and Wi3 MT and WT sib plants using ImageJ software.

Data Availability Statement

The datasets generated and analyzed during the current study are available in NCBI Sequence Read Archive (SRA); https://www.ncbi.nlm.nih.gov/, BioProjects PRJNA1047017 (Wi2) and PRJNA1047022 (Wi3).