Gossypium hirsutum cv. TM-1 and mutant Ghpericdh were used here. TM-1 is a standard genetic line of upland cotton, obtained from USDA-ARS, College Station, TX, United States. The Ghpericdh is a deficiency mutant of the GH_D13G1452 gene, generated from CCRI49 as a receptor overexpressing a glyphosate resistance gene g10evo (Tan, 2016), which was developed by our lab for five generations, and its wild-type (non-1007), a non-transgenic line with normal GH_D13G1452 separated from the selfing transgenic plant T0, were used in the experiment. CCRI49 is a conventional cotton cultivar, which was provided by the Cotton Research Institute, Chinese Academy of Agricultural Sciences. All cotton plants were grown in the Agricultural Station field from May to September (Zhejiang University, Hangzhou) or in a greenhouse at 28^°C/25^°C under a 14-h photoperiod with a light intensity of 35000 lx.

DNA and total RNA were extracted from fresh young leaves. cDNA synthesis, RT-PCR, and qPCR were performed according to Cao et al. (2021). Fusion primer and nested integrated PCR (FPNI-PCR), used for flanking sequence amplification, were designed according to methods described previously (Li, 2016; Xu, 2017). Primers in this study are listed in Supplementary Table 1.

The genomic DNA of 30 μg was digested completely with Hinde, separated by 0.8% gel electrophoresis, transferred into a nylon membrane (Amersham, United Kingdom), and hybridized with digoxin-labeled DNA fragments of g10evo at 65^°C overnight. The signaling was detected by the image analyzer FLA-5100 (FUJIFILM, Japan). Detailed procedures were as described in the DIG High Prime DNA Labeling and Detection Starter Kit II (Roche, Switzerland).

Full-length open-reading frames (ORFs) of GH_D13G1452 fused with a superfolder green fluorescent protein (sGFP) on its N-terminus or C-terminus and driven by CaMV35S (pCAMBIA1300 vector), transiently co-expressed with the known peroxisomal markers 984 fused with mCherry in epidermal cells of tobacco via Agrobacterium (Nelson et al., 2007). The GFP and mCherry fluorescence in epidermal cells of tobacco was detected and photographed by a laser confocal microscope (Olympus, Japan) after injection from 48 to 72 h.

The complete cDNA of gene g10evo (Tan, 2016), amplified from Deinococcus Radiodurans, was ligated into the overexpression vectors pCAMBIA-1300 (driven by the CaMV35s promoter) and transformed into Gossypium hirsutum cv. CCRI49 according to methods described previously (Yan, 2011). The transformants were selected on a selective medium containing 2.0 M glyphosate.

The TM-1 seedlings were transferred into a plastic bucket of 1 L with a 1/2 MS culture medium. When growing to a two-leaf stage, plants were stressed by 10^–4 M IAA, 10^–5 M GA, 10^–7 M ABA, 10^–4 M JA, 10^–3 M SA, 15% PEG6000, and 0.1-M NaCl for 3, 6, and 24 h with the non-treatment group as control. The roots were used for extracting RNA. The tendency (T) of the gene relative expression level over time under different treatments was calculated as T = 2^–(ΔCt) of the treatment group - 2^–(ΔCt) of the control group.

The plant height was measured from the base of hypocotyl to the first fully expanded true leaf. Here, the cotyledon is defined as the first node, the first leaf of the main stem is the second node, and so on. The distance between two adjacent leaves is defined as internode length. Internode length = Plant height/node numbers (each line repeated 6 individual plants). The germination rate was computed as the proportion of the germinated seeds in 15 days to the total seeds used (three replicates with 50 seeds in each one). Seed germination was carried out under a greenhouse at 28^°C/25^°C under a 14-h photoperiod with a light intensity of 35,000 lx.

The experiment was started in the morning after 10 h of the dark cycle. Cotyledon detached from cotyledon-stage seedlings was floated on the incubation medium (10-mM MES, 50-mM KCl, 100-μM CaCl₂, pH = 6.1) in Petri dishes under light for 2 h to make stomata open fully (Xi et al., 2019). The lower epidermis was peeled off, transferred to a drop of incubation medium on a glass slide, and immediately observed under a microscope (Nikon Eclipse Ni, Japan) for stomata and photographed. The long axis and short axis of the stomatal aperture were measured to calculate the opening degree (OD). OD = short axis/long axis. Three individuals were randomly selected from each genotype, and three visual fields were randomly selected from each cotyledon.

Pairwise sequence comparisons were conducted using ClustalW (Thompson et al., 1994). Sequence similarities were analyzed with BioEdit (Tippmann, 2004). The phylogenetic tree was constructed by the neighbor-joining method with a bootstrap replication of 500 by using MEAG 5 (Tamura et al., 2011).

Whole-genome resequencing of the Ghpericdh mutant was carried out with Oxford Nanopore Technologies (ONT) by Novogene (Tianjin, China) (Supplementary Table 2). Based on the resequencing data, the reads that could be mapped to both the g10evo gene and the cotton genome were extracted with NextGenMap-LR software (Sedlazeck et al., 2018). Then, the BAM files were obtained by comparing the extracted reads with the cotton genome. The VCR files were obtained by detecting the structural variation of the extracted BAM files using sniffles software (Sedlazeck et al., 2018). Based on the results of variation detection, the insertion sequence was extracted and compared with the g10evo gene for similarity analysis.

GH_D13G1452 has a high similarity of 84.4% with the peroxisomal ICDH from soybeans in the amino acid sequence (Table 1) and belongs to the peroxisomal ICDH group based on phylogenetic analysis (Figure 1A). The peroxisomal ICDHs possess the type-I peroxisomal targeting signal (PTS1), a tripeptide sequence typically found at the C terminus of peroxisomal proteins (Gould et al., 1989), such as SKL existed in perICDHs of soybeans and SRL in Arabidopsis perICDH, while Proline-Lysine-Leucine (PKL) was observed in the C-terminal of GH_D13G1452 (Figure 1B). Subcellular localization further confirmed that the peroxisomal targeting signal was located at the C-terminus of GH_D13G1452, while the peroxisomal signals disappeared when their C-terminus fused GFP (Figure 1C). Based on the above results, we named GH_D13G1452 as GhperICDH.

Transcripts Per Million (TPM) data of upland cotton transcriptome (Hu et al., 2019) downloaded from CottonFGD were used to analyze the gene expression pattern in different tissues and organs. The GhperICDH was expressed in different tissues and organs with a peak level in stem (30.39 TPM) and had variable levels at different stages of the ovule and developed fibers with peak level (31.7 TPM) in ovules of 15 days post-anthesis (DPA) (Figure 1D). Cis-elements analysis showed that a large number of environmental response elements are found in the 3-kb region upstream of GhperICDH, besides the core elements of the promoter (Figure 1E), which indicates that GhperICDH responds to a variety of stress conditions. Treating with IAA, GA, ABA, JA, SA, PEG, and salt for 3, 6, and 24 h, the expression trend of GhperICDH showed that this gene had obvious responses to ABA, JA, SA, PEG, and salt (Figure 1F).

A GhperICDH-deficiency mutant with glyphosate resistance was identified, and we named it Ghpericdh. Southern blotting displayed one copy of g10evo in the mutant (Figure 2A). Ghpericdh plants could grow normally under the recommended concentration of glyphosate isopropylamine in a field (1.37 kg⋅ai⋅hm^–2), while became damaged and displayed leaf malformation under high concentrations (4.10 and 6.83 kg⋅ai⋅hm^–2). Wild-type (WT) plants died under any concentration of glyphosate isopropylamine (Figure 2B). FPNI-PCR revealed the accurate position of g10evo, which was located in the middle of the 12th intron of GH_D13G1452 (Hu et al., 2019; Figure 2C). Genome variation detection of the mutant found only one site on 45,469,231 nt of chromosome D13 of G. hirsutum cv. TM-1 genome containing an insertion sequence that possessed a high similarity of 92.4% with the g10evo gene, which was consistent with the result of FPNI-PCR. GH_D13G1452 did not express in mutant Ghpericdh but did express in WT (Figure 2D). As controls, the homologous gene GH_A13G1507 and the housekeeping gene GhUBQ7 did normally express in both mutant and WT (Figure 2D and Supplementary Figure 1). Based on the genome information of mutant Ghpericdh, two specific molecular markers 80Ln/LBSP2 and Rb2b/80R only amplified in Ghpericdh were designed (Figure 2E and Supplementary Table 1).

The mutant Ghpericdh displays a dwarf phenotype (Figure 3A), but it can grow, flower, and bear fruits. Whether the plants growing naturally in the field or cultured in the greenhouse, the plant height and the internode length of mutants were significantly lower than those of WT (Figure 3B). The seeds from the mutant plants were shorter than WT seeds (Figure 3C). The greenhouse condition caused a decline in seed size of the mutant, which kept a similar trend in the WT as well (Figure 3C). Interestingly, the germination rate of seeds from greenhouse-grown mutant plants was dramatically lower than that from field-grown plants, i.e., 8.7 vs. 69.5%. In comparison, the seed germination rate from WT only decreased from 97.5 to 84.8% (Figure 3C). An observation from the Ghpericdh kernel found that seeds from the greenhouse appeared black (Figure 3D, marked by the red arrow) that became necrotic gradually during germination (Figure 3E), which led to growth stagnation. Similar to the Arabidopsis pericdh mutant, most stomata of Ghpericdh kept semi-closed (OD = 0.25-0.5) under light, while those of WT stayed open (OD ≧0.5) (Supplementary Figure 2).

F₁ heterozygotes were obtained from the reciprocal crosses between Ghpericdh and WT. The plant height of F₁ plants at the two-leaf stage was 11.6 cm between Ghpericdh (9.9 cm) and WT (16.5 cm) (Figure 3F). Among the 83 F₂ plants from the self-pollinated F₁ plant, the plant numbers of mutant/heterozygous/WT phenotype of plant height were 19/39/25, respectively, that the segregation conformed to 1:2:1 (χ² = 1.17, p > 0.1). The genotype of F₂ plants was investigated with GH_D13G1452-specific primer 80Ln/80R and Ghpericdh-specific molecular markers 80Ln/LBSP2 and Rb2b/80R, which showed that Ghpericdh homozygotes accounted for 19.3%, heterozygotes accounted for 49.4%, and GH_D13G1452 homozygotes accounted for 31.3%. The correlation coefficient between plant height and the GH_D13G1452 gene was 0.8. Therefore, GH_D13G1452 functioned as an incomplete dominance on plant height. The relatively low proportion of the mutant type in F₂ might be related to the reduction of the seed germination rate caused by the GH_D13G1452 deletion.