# Source paper archive - Source title: Evolution of herbicide resistance through co-option of an ethylene-associated nitrilase - Author(s): Ishizawa, H., Nakanomori, L., Hengphasatporn, K., Nishimura, K., Ishizaka, M., Akimoto, C. T., Su, D., Sugahara, M., Endo, M., Suzuki, R., Sato, M. P., Yasuda, K., Shigeta, Y., Gondo, T., Uchino, A., Iwakami, S. - Source landing URL: https://www.biorxiv.org/content/10.64898/2026.11.06.764846v1 - PDF URL: https://www.biorxiv.org/content/10.64898/2026.11.06.764846v1.full.pdf - PDF SHA-256: 8bf817fa54ae141731fdc1ccf72e2d3279182eb532c036683edeaf0496ef5044 - Converter: pymupdf4llm 0.0.27 - Limits: Automated PDF text extraction; reading order, equations, tables and formatting may be incomplete or incorrect. Images are omitted, and no OCR was performed. Markup and links are neutralized for safe downloading. Consult the original PDF for the authoritative version. --- [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) **Evolution of herbicide resistance through co-option of an ethylene-** **associated nitrilase** Hinata Ishizawa [1†], Luna Nakanomori [2†], Kowit Hengphasatporn [3], Kazusa Nishimura [4], Masumi Ishizaka [5], Chiharu Akimoto-Tomiyama [6], Danrui Su [7], Manami Sugahara [8], Masaki Endo [6], Rintaro Suzuki [5], Mitsuhiko P. Sato [9], Kentaro Yasuda [10], Yasuteru Shigeta [3], Takahiro Gondo [11], Akira Uchino [12], Satoshi Iwakami [7,13] 1 Graduate School of Agriculture, Kyoto University, Kitashirakawa-Oiwake-cho, Kyoto 606 8502 JAPAN 2 Graduate school of Agriculture, Tokyo University of Agriculture and Technology, Fuchu, Tokyo 183-8509 JAPAN 3 Center for Computational Sciences, University of Tsukuba, Tsukuba 305-8573 JAPAN 4 Graduate School of Environmental, Life, Natural Science and Technology, Okayama University, Okayama 700-8530 JAPAN 5 Research Center for Advanced Analysis, National Agriculture and Food Research Organization, Tsukuba, Ibaraki 305-8518 JAPAN 6 Institute of Agrobiological Sciences, National Agriculture and Food Research Organization, Tsukuba, Ibaraki 305-8634 JAPAN 7 United Graduate School of Agricultural Science, Tokyo University of Agriculture and Technology, Fuchu, Tokyo 183-8509 JAPAN 8 Faculty of Agriculture, University of Miyazaki, Miyazaki 889-2192 JAPAN 9 Kazusa DNA Research Institute, Chiba 292-0818 JAPAN 10 Agri-Innovation Education and Research Center, Akita Prefectural University, Akita 010 0451 JAPAN 11 Frontier Science Research Center, University of Miyazaki, Miyazaki 889-2192 JAPAN 12 Central Region Agricultural Research Center, National Agriculture and Food Research Organization, Tsu, Mie 514-2392 JAPAN 13 Institute of Agriculture, Tokyo University of Agriculture and Technology, Fuchu, Tokyo 183 8509 JAPAN - These authors contributed equally to this work. ∗ Correspondence to: S. Iwakami (iwakamis@go.tuat.ac.jp) [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 1 **Abstract** 2 Adaptive evolution frequently proceeds through the co-option of pre-existing biological 3 functions, yet the molecular mechanisms by which such co-option generates new adaptive 4 traits over contemporary evolutionary timescales remain poorly understood. Herbicide 5 resistant weeds provide a powerful model for addressing this question because enhanced 6 herbicide metabolism has repeatedly evolved within only a few decades. Here, we show that 7 resistance to cyhalofop-butyl in _Echinochloa crus-galli_ evolved through the co-option of 8 _EcNIT4A2_, which encodes a nitrilase that normally functions in ethylene-associated cyanide 9 detoxification. Genetic mapping, transcriptomics and transgenic analyses identified 10 _EcNIT4A2_ as the major determinant of resistance. A D317G substitution enhanced its 11 herbicide-detoxifying activity. Molecular dynamics simulations indicated that this substitution 12 facilitates substrate access to the catalytic pocket. A promoter structural variant markedly 13 increased _EcNIT4A2_ expression. Our findings reveal how coding and regulatory changes 14 can jointly drive the co-option and optimization of a conserved metabolic function during rapid 15 adaptation under intense anthropogenic selection. 16 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 17 **Introduction** 18 Adaptive evolution frequently proceeds through the modification and redeployment of pre 19 existing biological systems rather than the evolution of entirely new ones. [1,2] In plants, 20 evolutionary innovation is repeatedly linked to gene duplication and the subsequent co-option 21 of existing genes into new functional contexts. [3] Such co-option may be particularly crucial for 22 traits that evolve over contemporary timescales, where adaptation must arise within very few 23 generations. However, the molecular processes driving evolutionary co-option over such 24 short timescales remain largely unknown. 25 Herbicide-resistant weeds pose an escalating threat to sustainable agriculture and 26 global food security by compromising the efficacy of essential weed control tools. [4–6] Beyond 27 its agricultural importance, the repeated and independent evolution of herbicide resistance 28 under exceptionally strong anthropogenic selection provides an outstanding model for 29 investigating contemporary adaptive evolution. [7,8] One common outcome of this selection is 30 the evolution of metabolic herbicide resistance through enhanced detoxification capacity. [4,9] 31 Unlike mammals, in which several enzyme families have specialized roles in 32 xenobiotic metabolism, plants do not appear to possess enzyme families dedicated primarily 33 to detoxification of foreign chemicals. [10] Thus, herbicide-metabolizing enzymes are generally 34 believed to have evolved from enzymes whose primary functions lie in endogenous or 35 specialized metabolism. [11–15] Nevertheless, research into metabolic resistance has largely 36 focused on a limited set of enzyme families already known for their capacity to metabolize 37 xenobiotics, particularly cytochrome P450 monooxygenases and glutathione S 38 transferases. [16,17] Whether enzymes with established roles in endogenous metabolism can 39 also be recruited for herbicide detoxification therefore remains largely unexplored. 40 Here we investigate the molecular basis of resistance to the acetyl-CoA carboxylase 41 (ACCase) inhibitor cyhalofop-butyl (CB) in the globally problematic grass weed _Echinochloa_ 42 _crus-galli_ . [18] We identify EcNIT4A2, a NIT4 nitrilase involved in ethylene-associated cyanide 43 metabolism, as the major determinant of CB resistance. By integrating quantitative genetics, 44 comparative transcriptomics, functional analyses and molecular dynamics simulations, we 45 show that resistance arose through the co-option of this endogenous metabolic enzyme. A 46 promoter structural variant (SV) markedly increased _EcNIT4A2_ expression, whereas a single 47 amino acid substitution enhanced its herbicide-detoxifying activity by facilitating substrate 48 access to the catalytic pocket. Together, our findings reveal how regulatory and coding 49 changes can jointly recruit and optimize a conserved metabolic function during rapid 50 adaptation under intense anthropogenic selection. 51 52 **Results** [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 53 **The R1 line rapidly detoxifies cyhalofop-acid to cyhalofop-diacid** 54 Previous studies showed that two CB-resistant lines collected in 2010 from nearby rice fields 55 in Japan exhibited distinct responses to different herbicides despite their close geographic 56 proximity (Fig. 1a,b). [18,19] Because neither line carried known target-site resistance 57 mutations, [18] we investigated whether enhanced herbicide metabolism underlies CB 58 resistance. 59 CB is a proherbicide with little direct herbicidal activity and is hydrolysed in plants 60 to cyhalofop-acid, its herbicidally active form. [20] Further conversion of cyhalofop-acid to 61 cyhalofop-diacid inactivates the herbicide. LC–MS/MS analysis showed that cyhalofop-acid 62 accumulated rapidly in the susceptible S1 line following CB treatment, whereas its 63 abundance remained low in both resistant lines (Fig. 1c). Cyhalofop-diacid accumulated to 64 high levels only in R1, indicating rapid detoxification of cyhalofop-acid in this line. Because 65 CB abundance varied among samples, we directly assessed sensitivity to cyhalofop-acid. 66 Both resistant lines were less sensitive than S1 (Fig. 1d), demonstrating that their resistance 67 extends to the active acid itself. These results identify rapid conversion of cyhalofop-acid to 68 cyhalofop-diacid as the basis of resistance in R1; subsequent analyses therefore focused on 69 this line. 70 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 71 72 83 40.7) explaining 69.9% of the phenotypic variance (Fig. 2d). Composite interval mapping 84 independently identified the same QTL. Fine mapping of 15 recombinant individuals using 85 nine indel markers narrowed the locus to a 1.42-Mb interval between H045 and H008 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 86 (32,291,204–33,706,248 bp), containing 24 gene models. Marker H046 at 32.80 Mb showed 87 complete co-segregation with the resistance phenotype. 88 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 89 90 _**EcNIT4A2**_ **is highly expressed in the R1 line** 91 To identify genes associated with CB resistance in R1, we first compared the leaf 92 transcriptomes of untreated R1, R2 and susceptible S1 plants. A total of 1,372 genes were 93 specifically upregulated in R1 relative to both R2 and S1 (Fig. 3a). We then performed an 94 independent RNA-seq analysis comparing R1 with three CB-susceptible populations 95 collected from geographically distinct regions of Japan. This second analysis identified 304 96 genes specifically upregulated in R1, 67 of which were shared with the first dataset (Fig. 3b). 97 Of the 24 gene models located within the 1.42-Mb resistance interval (Fig. 2d), only 98 _EcNIT4A2_ was consistently upregulated in R1 in both transcriptome analyses. Consistent 99 with its candidacy, RT-qPCR analysis showed that high _EcNIT4A2_ expression co-segregated 100 with resistance in S1 × R1 F2:3 families (Fig. 3c). 101 _EcNIT4A2_ encodes a NIT4 nitrilase, an enzyme with a conserved function in 102 cyanide detoxification associated with ethylene biosynthesis. [22,23] Ethylene production 103 generates cyanide as a by-product, which is incorporated into β-cyanoalanine and 104 subsequently hydrolysed by NIT4 to produce asparagine or aspartate (Fig. 3d). Thus, the 105 genetic and transcriptomic analyses identified an enzyme whose endogenous function 106 involves nitrile hydrolysis, the same reaction underlying detoxification of cyhalofop-acid to 107 cyhalofop-diacid. 108 Whereas NIT4 occurs as a single-copy gene in most dicots, grasses possess two 109 distinct paralogs, NIT4A and NIT4B (Fig. 3e). [24] These paralogs are typically arranged in 110 tandem in grass genomes (Fig. 3f). In the hexaploid _E. crus-galli_ genome, tandem _NIT4A–_ 111 _NIT4B_ pairs occur on homoeologous chromosomes 7A, 7B and 7C, resulting in six NIT4 112 copies. Of these, _EcNIT4A2_, the B-subgenome _NIT4A_ copy on chromosome 7B, was 113 uniquely overexpressed in R1 (Fig. 3g). 114 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 115 126 independent transgenic lines were obtained. Both exhibited resistance to CB in callus assays 127 (Fig. S3), and their regenerated progeny were resistant in both foliar application and sterile [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 128 culture assays (Fig. 4b, Fig. S3). RNA-seq analysis showed that _NIT4A_ transcript abundance 129 in resistant transgenic progeny was generally several hundred to over 1,000 TPM, 130 substantially exceeding endogenous _NIT4A2_ expression in wild-type _E. phyllopogon_ (~40 131 TPM). Together, these results demonstrate that overexpression of _EcNIT4A2_ is sufficient to 132 confer CB resistance in _Echinochloa_ . 133 We next examined whether overexpression of wild-type rice NIT4A could also 134 enhance CB tolerance. Overexpression of _OsNIT4A_ had little effect in callus assays (Fig. S4) 135 but conferred a significant increase in CB tolerance at the whole-plant level (Fig. 4c,d). To 136 assess the contribution of endogenous NIT4 activity to basal CB tolerance, we disrupted 137 _OsNIT4A_ and _OsNIT4B_ using CRISPR/Cas9. Disruption of either _OsNIT4A_ or _OsNIT4B_ 138 increased CB sensitivity by approximately two orders of magnitude, as both mutants showed 139 growth inhibition at 0.1 μM CB comparable to or greater than that of WT plants at 20 μM CB 140 (Fig. 4e,f). Grass NIT4A and NIT4B proteins can form heteromeric complexes. [29–32] If these 141 complexes contribute to cyhalofop-acid detoxification, disruption of either gene would be 142 expected to reduce heteromeric NIT4 activity, potentially explaining why both knockouts 143 showed similar increases in CB sensitivity. Together, these results indicate that endogenous 144 NIT4 activity contributes to basal CB tolerance in rice. Notably, the modest effect of _OsNIT4A_ 145 overexpression contrasts with the strong resistance conferred by the R1 _EcNIT4A2_ allele. [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 146 147 154 D317G. This position is well-conserved as aspartic acid (D) among grass NIT4A proteins, 155 including those from rice and _Echinochloa_ (Fig. 5b). Functional comparison showed that 156 expression of the D317 _EcNIT4A2_ allele from S1 conferred little resistance to CB as observed 157 for rice _NIT4A_ (Fig. S4), whereas the G317 allele from R1 conferred strong resistance (Fig. [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 158 5c, Fig. S5). Expression of the two other _E. crus-galli NIT4A_ homeologues, _EcNIT4A1_ and 159 _EcNIT4A3_, both of which encode D317, likewise failed to confer substantial resistance (Fig. 160 5c). Introduction of D317G into rice NIT4A also resulted in a marked increase in CB 161 resistance (Fig. 5d). Together, these results identify D317G as the coding mutation 162 responsible for enhanced NIT4A-mediated CB resistance in _E. crus-galli_ . 163 To explain the enhanced activity conferred by D317G, we examined the structural 164 organization of grass NIT4 complexes. The only experimentally determined NIT4 structure is 165 the Arabidopsis cryo-EM complex, in which the substrate pocket is formed by three adjacent 166 subunits (pocket, lid and lock). [33] In this structure, D317 from the lock subunit forms an 167 intersubunit contact with R95 in the adjacent lid subunit. Arabidopsis has a single NIT4, 168 whereas grasses possess two paralogous subfamilies, NIT4A and NIT4B, proposed to 169 function as heteromers, [29–32] implying that several NIT4A–NIT4B assemblies may exist. 170 Because residue 317 contributes to the pocket only from the lock subunit [33] and its potential 171 interacting residue at position 95 differs between the alternative lid subunits (P95 in 172 EcNIT4A2 and R95 in EcNIT4B2), we modelled all eight possible assemblies containing 173 either the S1 or R1 EcNIT4A2 at the lock position and either EcNIT4A2 or EcNIT4B2 at the 174 lid and pocket positions (Fig. 5e). 175 To compare substrate access among these assemblies, we performed enhanced 176 sampling molecular dynamics (MD) simulations designed to sample the entry of cyhalofop 177 acid into the catalytic pocket. Across 200 LB-PaCS-MD cycles, the ligand approached the 178 catalytic pocket in all eight assemblies (Fig. S6a), but the extent of approach varied 179 substantially with subunit composition. A [R] –B–B showed the most favorable substrate access 180 and the widest lid–lock separation among the eight modeled assemblies in the final LB 181 PaCS-MD structures (10.18 Å; Fig. 5e). Free-energy landscapes supported this trend, with 182 A [R] –B–B reaching the smallest annotated ligand–pocket distance (6.09 Å), compared with 183 11.55 Å in A [S] –B–B (Fig. S6b). The remaining assemblies fell between these extremes. 184 Together, these results suggest that D317G enhances substrate access in a subunit 185 dependent manner, with the strongest effect observed in A [R] –B–B. 186 Extended simulations of A [S] –B–B and A [R] –B–B further supported this difference (Fig. 187 5f and Supplementary Videos S1 and S2). In A [S] –B–B, the ligand remained near the pocket 188 entrance, with a mean ligand–C197 distance of 13.51 ± 1.18 Å, whereas in A [R] –B–B, it 189 penetrated further into the pocket (7.18 ± 0.24 Å). The R95–residue 317 distance was also 190 greater in A [R] –B–B (9.40 ± 1.42 Å) than in A [S] –B–B (6.16 ± 0.24 Å), consistent with a wider 191 lid–lock opening. Additional contact analyses revealed more persistent ligand–lock 192 interactions in A [S] –B–B and transient R95–D317 hydrogen bonding during the early 193 simulation stages (Fig. S6c,d). In a representative bound conformation of A [R] –B–B (Fig. 5g), [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 194 the aryl moiety of cyhalofop-acid packs against R95 of the NIT4B lid, while its carboxylate 195 group interacts with E76, K163 and E170, positioning the nitrile carbon 5.99 Å from C197. 196 Together, these findings suggest that D317G facilitates substrate entry by altering the lid– 198 199 200 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 202 To investigate the basis of _EcNIT4A2_ overexpression, we compared the genomic region 203 surrounding _EcNIT4A2_ between the S1 and R1 genomes assembled using HiFi sequencing. 204 In the S1 genome, _EcNIT4B2_ was located upstream of _EcNIT4A2_, with an approximately 205 4.2-kb intergenic region separating the _EcNIT4B_ 2 stop codon and the _EcNIT4A_ 2 start codon 206 (Fig. 6a). The sequence immediately upstream of the _EcNIT4A2_ coding region was identical 207 between S1 and R1 for the first 106 bp. However, the adjacent 3,349-bp region in S1 was 208 replaced by a 313-bp sequence in R1 (Fig. 6a). The remaining intergenic sequence was 209 nearly identical between the two alleles. 210 To further examine the relationship between promoter genotype and _EcNIT4A2_ 211 expression, we genotyped the promoter region in the progeny previously analyzed by qRT 212 PCR (Fig. 3g). All individuals carrying the R-type promoter exhibited elevated _EcNIT4A2_ 213 expression, whereas individuals carrying the S-type promoter showed expression levels 214 comparable to the susceptible parent (Fig. 6b), indicating complete co-segregation of the 215 promoter SV and _EcNIT4A2_ overexpression. 216 To test whether this promoter SV contributes to _EcNIT4A2_ overexpression, the 217 _EcNIT4B2_ – _EcNIT4A2_ intergenic regions from the S1 and R1 lines were fused to a NanoLuc 218 reporter and analyzed using a dual-luciferase assay in transiently transformed rice calli (Fig. 219 6c). Whereas the S1 promoter drove only weak reporter expression, the R1 promoter 220 exhibited approximately 10-fold higher normalized NanoLuc activity (NanoLuc/F-LUC). 221 Together, these findings demonstrate that a promoter SV is responsible for _EcNIT4A2_ 222 overexpression, thereby providing the regulatory basis for its recruitment into herbicide 223 detoxification. [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 225 226 **Discussion** 227 The co-option of pre-existing biological functions is an important mechanism of evolutionary 228 innovation, yet how such transitions occur over contemporary evolutionary timescales 229 remains poorly understood. Here, we demonstrate that resistance to the herbicide cyhalofop 230 butyl evolved through the co-option of EcNIT4A2, a nitrilase involved in ethylene-associated 231 cyanide metabolism. Resistance arose through the combined effects of a promoter structural 232 variant that markedly increased _EcNIT4A2_ expression and a coding mutation that enhanced 233 its herbicide-metabolizing activity. In contrast, the independently resistant R2 line showed 234 neither elevated _EcNIT4A2_ expression nor cyhalofop-diacid accumulation, suggesting that 235 alternative resistance mechanisms may operate in this line. Together, our findings provide a 236 molecular example of how regulatory and coding changes can jointly recruit and optimize a 237 conserved metabolic enzyme during rapid adaptation under anthropogenic selection. 238 The evolutionary history of plant nitrilases suggests that this enzyme family is 239 particularly amenable to functional diversification. [23] Although ancestral NIT4 enzymes 240 function in cyanide detoxification, gene duplication has repeatedly been followed by changes 241 in substrate specificity. In Brassicaceae, for example, duplicated NIT1–3 enzymes acquired 242 the ability to metabolize glucosinolate-derived nitriles, [24,33,34] whereas in sorghum, duplication 243 and functional divergence of _NIT4B_ gave rise to paralogs specialized for dhurrin turnover. [29] [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 244 The evolution of cyhalofop metabolism by EcNIT4A2 extends this recurring pattern of nitrilase 245 diversification into the context of contemporary adaptation. Unlike the endogenous 246 metabolites involved in these earlier examples, CB is a synthetic herbicide introduced only 247 in the 1990s, [35] illustrating how endogenous metabolic enzymes can be recruited to detoxify 248 entirely novel anthropogenic compounds over short evolutionary timescales. 249 Resistance in R1 involved changes in both _EcNIT4A2_ expression and the function 250 of its encoded protein. Elevated expression of detoxification genes is common in metabolic 251 herbicide resistance, [e.g. 36–38] whereas direct evidence that adaptive amino acid substitutions 252 can enhance herbicide-metabolizing activity remains scare. In the R1 line, a promoter SV 253 markedly increased _EcNIT4A2_ expression, while the D317G substitution substantially 254 enhanced its capacity to metabolize cyhalofop-acid. Importantly, disruption of either 255 _OsNIT4A_ or _OsNIT4B_ greatly increased CB sensitivity, indicating that endogenous NIT4 256 complexes already contribute to basal cyhalofop-acid detoxification in plants. The 257 Arabidopsis NIT4 structure revealed an R95–D317 lid–lock interaction, and D317 was 258 proposed as a target for altering substrate size. [33] Our structural analyses suggest that the 259 naturally occurring D317G substitution disrupts this interaction, widening the substrate-entry 260 channel without altering the catalytic tetrad. This effect was observed only in heteromeric 261 assemblies, consistent with the proposed heteromeric organization of grass NIT4. [29–32] 262 Together, these findings suggest that herbicide resistance arose through enhancement of a 263 pre-existing metabolic capacity rather than through the emergence of an entirely new 264 herbicide-metabolizing activity. 265 The evolutionary origins of the regulatory and coding changes underlying 266 EcNIT4A2-mediated resistance remain unresolved. One explanation for the rapid evolution 267 frequently observed in metabolic herbicide resistance is that selection acts primarily on 268 standing genetic variation rather than relying solely on newly arising beneficial mutations. [39] 269 Notably, a promoter configuration resembling the R1 allele is present in the published 270 genome of an _E. crus-galli_ accession originating from China, [40] raising the possibility that at 271 least the regulatory component of resistance arose from pre-existing variation. In contrast, 272 D317 is highly conserved among plant NIT4 proteins, suggesting stronger evolutionary 273 constraint on this coding position. Determining whether the regulatory and coding variants 274 arose independently, sequentially or from pre-existing variation will require broader 275 population-level analyses of _EcNIT4A2_ diversity. 276 Because NIT4 performs an important physiological role in cyanide detoxification, its 277 recruitment for herbicide metabolism raises the question of how such evolutionary 278 modification could occur without compromising its endogenous function. One possible 279 explanation is functional redundancy. _E. crus-galli_ is hexaploid and carries multiple NIT4 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 280 homeologues, and _EcNIT4A2_ is expressed at only very low levels in the susceptible lines 281 (Fig. 3g). Thus, this particular copy may make only a limited contribution to basal NIT4 282 function, potentially allowing it to diverge while other copies maintain the ancestral function. 283 Gene duplication is widely recognized as a mechanism that facilitates such evolutionary 284 innovation by relaxing functional constraints on individual copies. [41,42] Consistent with this 285 interpretation, resistant plants carrying the highly expressed D317G allele showed no 286 obvious developmental defects under our experimental conditions, although the effect of 287 D317G on native β-cyanoalanine metabolism remains unresolved. Functional redundancy 288 among _NIT4_ copies may therefore have provided the evolutionary flexibility that enabled 289 _EcNIT4A2_ to be co-opted for herbicide detoxification. 290 Repeated recruitment of conserved genes and pathways is a common feature of 291 evolutionary innovation over deep evolutionary timescales. [3] Our findings show that the same 292 principle can operate during contemporary adaptation, with regulatory and coding changes 293 jointly enhancing a pre-existing metabolic capacity within only a few decades. More broadly, 294 this work provides a molecular example of how conserved metabolic functions can be co 295 opted and optimized under intense anthropogenic selection. 296 297 **Methods** 298 **Plant materials** 299 _Echinochloa crus-galli_ var. _formosensis_ lines S1 (formerly Ecf99), R1 (formerly Ecf108), and 300 R2 (formerly Ecf27) were used. The lines were originated in Okayama city, Japan. The details 301 of herbicide sensitivity are described in elsewhere. [18,19] The crossing of S1 and R1 was 302 performed following hot water emasculation method in rice. [43] The success of hybridization 303 was confirmed by sequencing the _ALS3_ gene in the resultant plants, following PCR 304 amplification with primers listed in Table S1. The F1 plant was self-pollinated to generate F2 305 progeny. Individual F2 plants were subsequently self-pollinated to produce F2:3 families. 306 _E. crus-galli_ var. _crus-galli_ lines EcTU018, EcTU019, and EcTU006, originating from 307 Hokkaido, Niigata, and Aichi, respectively, were designated S2, S3, and S4, respectively. 308 They were all sensitive to CB (Fig. S7). _E. phyllopogon_ AEC02-17 line [28] and rice ( _Oryza_ 309 _sativa_ L. ‘Nipponbare’) were used for transformation. 310 311 **CB sensitivity assay** 312 Foliar application of a commercial formulation of CB was performed essentially as described 313 previously. [19] Briefly, plants were grown under controlled environmental conditions (28 ± 5 °C, 314 13-h photoperiod). At the 4- or 5-leaf stage, plants were treated with Clincher 30% EW 315 (Corteva Japan) supplemented with 0.1% (v/v) Surfactant 30 (Maruwa Biochemical) at 300 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 316 g active ingredient (a.i.)/ha. Herbicide sensitivity was evaluated 10 or 21 days after treatment, 317 depending on the experiment. 318 For sensitivity assays in culture tubes, seeds were surface-sterilized by vacuum 319 infiltration in sodium hypochlorite (~2.5% active chlorine) containing 0.01% (v/v) Tween 20 320 for 15 min and then washed three times with sterile water. The sterilized seeds were 321 germinated under 25°C and transferred to MS medium supplemented with 0.05% (v/v) Plant 322 Preservative Mixture (Plant Cell Technology). _E. crus-galli_ seedlings were grown at 25 °C for 323 7 days, whereas rice seedlings were grown at 30°C for 5 days, both under a 12-h photoperiod. 324 For rice callus sensitivity assays, 9–12 independent transgenic calli were cultured 325 on N6D medium containing CB, as described previously. [37] For _E. phyllopogon_ assays, two 326 independently transformed callus lines were examined, with 12 callus pieces per line 327 assayed on a single plate at each CB concentration. 328 329 **Vector construction** 330 Unless otherwise indicated, coding sequences were PCR-amplified from cDNA using the 331 primers listed in Table S1. The PCR products were subcloned into the pCR-Blunt vector using 332 the Zero Blunt PCR Cloning Kit (Thermo Fisher Scientific). The coding sequences of 333 _CYP81A68_ (OK483200.1) and _EcGSTU23_ (EC_v6.g055562) were synthesized by GenScript 334 (Tokyo, Japan). The coding sequence of _CYP89B16_ (Eph1.1ch9A.g032540.1), previously 335 identified in the R511 line of _E. phyllopogon_, [44] was amplified from R511 cDNA. The D317G 336 substitution was introduced into the _OsNIT4A_ coding sequence by site-directed mutagenesis 337 using the primers listed in Table S1. 338 For construction of overexpression vectors, pCAMBIA1390-sGFP [45] was digested 339 with BsrGI and SalI, and the respective coding sequences were inserted into the linearized 340 vector using either the In-Fusion HD Cloning Kit (Takara) or the SLiCE method. [46] The 341 pAHC25 [47] vector was digested with SmaI and SacI, and the _EcNIT4A2_ coding sequence was 342 inserted into the linearized vector using SLiCE. Binary CRISPR/Cas9 vectors carrying gene 343 specific single-guide RNAs were constructed as described previously. [48] The protospacer– 344 PAM sequences used for _OsNIT4A_ (5′-GCGGAGGTGGAGATGAACGGCGG-3′) and 345 _OsNIT4B_ (5′-ACGGAGGTGGAGATGAACGCCGG-3′) were designed using CRISPR-P 346 v2.0. [49] 347 To construct the dual-luciferase reporter vectors, the firefly luciferase (FLUC) coding 348 sequence together with the NOS terminator (NOS-T) was PCR-amplified from pGWB535 349 (Addgene plasmid no. 74873) using primer set pL1. The 2×CaMV 35S promoter was 350 amplified from pCAMBIA1390-sGFP using primer set pL2. These fragments were assembled 351 into the linearized pUC19 vector supplied with the In-Fusion HD Cloning Kit to generate a [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 352 2×35S::FLUC::NOS-T construct. The resulting 2×35S::FLUC::NOS-T cassette was 353 subsequently amplified using primer set pL3. Separately, the 35S terminator (35S-T) was 354 amplified from pB2GW7 [50] using primer set pL4, and the NanoLuc luciferase (NLUC) coding 355 sequence was amplified from pNL1.1.PGK (Promega) using primer set pL5. These fragments 356 were assembled into the linearized pUC19 vector supplied with the In-Fusion HD Cloning Kit 357 to generate an intermediate construct. The intermediate plasmid was digested with AscI, and 358 the 2×35S::FLUC::NOS-T cassette was inserted by In-Fusion cloning to generate pLUC2. 359 To construct pLUC4, pB2GW7 was digested with MauBI and HindIII. The NOS 360 T::FLUC::2×35S::NLUC::35S-T cassette was PCR-amplified from pLUC2 using primer set 361 pL6 and inserted into the linearized pB2GW7 backbone by In-Fusion cloning. The S1 and R1 362 _EcNIT4A2_ promoter fragments were PCR-amplified from genomic DNA and subcloned into 363 the pCR-Blunt vector. The promoter fragments were re-amplified using primer set pL7 and 364 inserted into AscI-linearized pLUC4 by In-Fusion cloning. All final constructs were verified by 365 Sanger sequencing. All primers used for vector construction are listed in Table S1. 366 367 **Plant transformation** 368 Rice transformation was performed as previously described. [37] Genome-edited T0 plants 369 were genotyped by Sanger sequencing of PCR amplicons spanning the target site (Table 370 S1), and biallelic mutants were selected for further analyses. 371 _E. phyllopogon_ transformation was performed as previously described. [28] Briefly, 372 embryogenic calli were induced on MS medium [51] supplemented with 2.0 mg L [-1] 2,4-D, 0.1 373 mg L [-1] 6-benzylaminopurine and 5 µM copper sulfate. The embryogenic calli were 374 transformed using a self-built particle bombardment apparatus. [52] Bombardment was carried 375 out using gold particles (1.0 µm diameter; Biorad). The transformants were selected with 5 376 mg L [-1] bialaphos. Plants were regenerated by transferring the calli on MS medium 377 supplemented with 0.02 mg L [-1] naphthaleneacetic acid (NAA) and 2 mg L [-1] kinetin and 5 mg 378 L [-1] bialaphos. Regenerated plants were tested for the presence of the EcNIT4A2-NOS 379 terminator by PCR (Table S1). Subsequently, PCR-positive plants were transferred to soil 380 and grown to maturity. 381 382 **Dual luciferase assay** 383 Calli co-cultured with Agrobacterium for 1 day at 28°C were washed with sterile water, and 384 50 mg of callus tissue was snap-frozen in liquid nitrogen and stored at −80°C until use. The 385 frozen tissue was ground and resuspended in Passive Lysis Buffer (Promega). Firefly and 386 NanoLuc luciferase activities were measured using the Nano-Glo Dual-Luciferase Reporter 387 Assay System (Promega) according to the manufacturer's instructions. Luminescence was [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 388 detected using a GloMax Explorer (Promega). 389 390 **Real-time PCR** 391 Total RNA was extracted from the shoot of a single plant with three biological replicates, and 392 500 ng of RNA was used to synthesize cDNA with the PrimeScript IV 1st strand cDNA 393 Synthesis Mix (Takara). Quantitative PCR was performed using THUNDERBIRD Next SYBR 394 qPCR Mix (TOYOBO) on a QuantStudio 3 (ThermoFischer). The primers used in this study 395 are listed in Table S1. Primer concentrations for _EcEIF4B_ and _EcNIT4A2_ were 100 μM and 396 300 μM, respectively. Samples with _EcNIT4A2_ signals below the detection limit after 40 397 cycles were assigned a Ct value of 40 for downstream analyses. Data was analyzed using 398 the ΔΔCt method. [53] 399 400 **Nucleic acid extraction** 401 Total RNA was extracted using a Plant Total RNA Mini Kit (Favorgen). The extracted RNA 402 was treated with TURBO DNA- _free_ Kit (Thermo Fisher Scientific). For PCR and dpMIG-seq 403 analysis, DNA was extracted from leaves of each plant with the cetyltrimethylammonium 404 bromide method. [54] For HiFi sequencing, DNA was extracted using NucleoBond HMW DNA 405 (Takara). 406 407 **Genome assembly** 408 HMW DNA was sequenced on a PacBio Revio platform at BGI using their standard library 409 preparation protocol. HiFi reads were assembled with hifiasm (v0.25.0). [55] Nuclear contigs 410 were extracted using Tiara (v1.0.3) [56] and aligned to the reference genome of _E. crus-galli_ [40] 411 using RagTag (v2.1.0). [57] Gene annotation was transferred from the reference genome using 412 Liftoff (v1.6.3). [58] 413 414 **RNA-sequencing analyses** 415 Germinated plants (R1, R2, and S1) were grown hydroponically to the 2.5-leaf stage in a 416 growth chamber at 25°C under a 12-h photoperiod (∼300 μmol m⁻² s⁻¹) as described 417 previously. [59] Shoots from three plants were snap-frozen in liquid nitrogen and stored at 418 −80 °C until RNA extraction. RNA samples were sent to AZENTA for library preparation using 419 their standard protocol. The libraries were sequenced on the HiSeq 4000 platform (150-bp 420 paired-end reads). 421 For the other RNA-seq experiments, RNA was extracted from the shoots of 2-leaf 422 stage plants grown in a growth chamber under a 14-h photoperiod at 25 ± 5°C with a high 423 pressure sodium lamp. Libraries for S1, S2, S3, and S4 were prepared using the NEBNext [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 424 Ultra II RNA Library Prep Kit for Illumina (NEB, Tokyo, Japan). Sequencing was performed 425 by AZENTA on a NovaSeq 6000 platform (150-bp paired-end reads). RNA samples from 426 transgenic _E. phyllopogon_ were sent to BGI for library preparation using their standard 427 protocol. The libraries were sequenced on the DNBSEQ platform (150-bp paired-end reads). 428 Except for the clean data from BGI, raw data were processed using fastp (v0.23.2) [60] 429 with the following parameters: -q 30 -l 120 -w 16 --detect_adapter_for_pe. Filtered reads 430 were mapped to the reference genome using STAR (v2.7.9), [61] and read counts were 431 quantified using RSEM (v1.3.2). [62] Differentially expressed genes were identified using the R 432 package edgeR (v3.38.4). [63] Genes specifically upregulated in each line were identified by 433 comparing the target line with the other lines using pairwise contrasts in a generalized linear 434 model framework. Genes were defined as specifically upregulated when they showed 435 significant upregulation in both comparisons (FDR < 0.05 and logFC > 1). 436 437 **dpMIG-seq and genetic analysis** 438 Genomic DNA was extracted from F2 individuals as well as the parental lines (S1 and R1). 439 dpMIG-seq libraries were prepared as reported previously. [21] Paired-end sequencing (150 bp) 440 were performed on a DNBSEQ-G400 platform. Raw reads were processed using fastp 441 (v0.23.4) [60] with the parameters -q 30 -l 120 -w 16 --detect_adapter_for_pe. Filtered reads 442 were aligned to the _E. crus-galli_ genome using strobealign (v0.15.0), [64] and alignments were 443 coordinate-sorted with SAMtools (v1.11). [65] BAM files were filtered using bamutils (-mismatch 444 3 -properpair) (v0.5.9). [66] Variants were called using SAMtools mpileup and BCFtools 445 BCFtools (v1.19). [65] 446 A linkage map was constructed using LEP-MAP3 (Rastas, 2017). [67] Markers were 447 assigned to chromosomes based on their physical positions in the reference genome. [40] 448 Markers showing segregation distortion or linkage relationships inconsistent with the 449 reference genome were excluded, and marker order and genetic distances were estimated 450 using the Haldane mapping function. 451 QTL analysis was performed using the R/qtl package (Broman et al., 2003). [68] The 452 F 2 individuals were genotyped by dpMIG-seq, and resistance phenotypes of the 453 corresponding F 2:3 families were used for QTL mapping. Interval mapping and composite 454 interval mapping were performed, and genome-wide significance thresholds were 455 determined by 1,000 permutation tests. Confidence intervals were defined as 1.5-LOD 456 support intervals. Additive and dominance effects, together with the proportion of phenotypic 457 variance explained, were estimated for the detected QTL. 458 To refine the resistance locus, recombinant individuals identified within the QTL 459 region by dpMIG-seq were selected for fine mapping. Nine indel markers were developed [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 460 from whole-genome resequencing data of the QTL region (Table S1) and used to genotype 461 the recombinant individuals. The QTL interval was refined by comparing graphical genotypes 462 with resistance phenotypes and identifying recombination breakpoints delimiting the 463 resistance locus. 464 465 **LC-MS/MS** 466 4-leaf stage plants were treated with 5 mM CB by applying 1 μL at each of three spots (3 μL 467 per plant). Plants were sampled at 0, 3, 8, and 24 h after application, and two plants were 468 pooled as one sample. Each sample was extracted with 300 μL of 80% (v/v) methanol and 469 transferred to a 1.5-mL microcentrifuge tube. After centrifugation at 15,000 rpm. for 10 min, 470 the supernatant was subjected to LC–MS/MS analysis. Metabolites were analyzed using a 471 Nexera X2 UHPLC system (Shimadzu) coupled to a TripleTOF [TM] 5600+ mass spectrometer 472 (SCIEX) equipped with an electrospray ionization (ESI) source (DuoSpray [TM] ion source). 473 Chromatographic separation was performed on a SunFire C18 column (2.1 × 150 mm, 3.5 474 μm; Waters) fitted with an OptiGuard C18 guard column (1.0 × 15 mm, Optimize 475 Technologies) maintained at 40°C. The mobile phases consisted of (A) water containing 10 476 mM formic acid and 10 mM ammonium formate and (B) methanol. The gradient was 10% B 477 at 0 min, increased linearly to 95% B at 15 min, maintained until 25 min. The injection volume 478 was 1 μL. 479 CB was analyzed in positive ion mode, whereas cyhalofop-acid and cyhalofop − 480 diacid were analyzed in negative ion mode. The ion spray voltages were +5500 V and 481 4500 V for positive and negative modes, respectively, and the source temperature was 550°C. – 482 CB, cyhalofop-acid, and cyhalofop-diacid were quantified by targeted LC MS/MS using the 483 transitions _m/z_ 375.1 ([M+NH4]+) → 256.07793, _m/z_ 300.1 ([M-H] [−] ) → 228.04663, and _m/z_ 484 319.1 ([M-H] [ −] ) → 247.04121, respectively. 485 486 **Phylogeny analyses of NIT4 genes** 487 Amino acid sequences of NIT4 of _A. thaliana_, _Nicotiana tabacum_, _Lupinus angustifolius_, 488 _Saccharum officinarum_, _Hordeum vulgare_, _O. sativa_, _Populus trichocarpa_, _Triticum aestivum_ 489 and _Zea mays_ were aligned with MAFFT (v7.520). [69] A maximum-likelihood phylogenetic tree 490 was inferred from the aligned sequences using IQ-TREE (v.3.1.2) [70] and visualized using iTOL 491 (v7.6). [71] The local synteny around NIT4 was visualized with clinker (v0.0.21). [72] Sequence 492 logo was generated with grass NIT4A sequences (Table S2) using ggseqlogo (v0.2). [73] 493 494 **Homology Modeling and MD simulation of NIT4 fibril** 495 Eight distinct NIT4 fibril systems were constructed from trimeric assemblies where each [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 496 monomer contributed specific pocket, lid, and lock regions. The pocket and lid regions were 497 assigned to either EcNIT4A2_S (designated as S) or EcNIT4B2 (designated as B), whereas 498 the lock region was assigned to either EcNIT4A2_S (S) or EcNIT4A2_R1 (designated as R). 499 This combinatorial scheme yielded eight systems designated as A [S] –A [S] –A [S], A [S] –B–B, A [S] –A [S] – 500 B, A [S] –B– A [S], A [R] –A [S] –A [S], A [R] –B–B, A [R] –A [S] –B, and A [R] –B– A [S], where the three-letter code 501 indicates the identity of the pocket, lid, and lock regions, respectively. Structural models were 502 generated using SWISS-MODEL [74] with cryo-EM structure 6I5T [33] as the template. Each 503 modeled trimeric fibril complex was prepared for MD simulation using the AMBER24 504 package. [75] Protein topologies were generated with the AMBER ff19SB force field, while 505 parameters for cyhalofop-acid ligand were derived from the General AMBER Force Field. 506 Ligand atomic partial charges were assigned via the restrained electrostatic potential method 507 following density functional theory calculations at the B3LYP/6-31G(d) level using 508 Gaussian16. [76] The ANTECHAMBER module was employed to generate ligand topology and 509 parameter files compatible with the AMBER force field. Complexes were solvated in 510 truncated octahedral TIP3P boxes with a 10 Å buffer and neutralized with Na [+] or Cl [−] . 511 Systems were minimized for 1000 restrained steepest-descent and 1000 unrestrained 512 conjugate-gradient steps, heated from 0 to 300 K over 100 ps under NVT, and equilibrated 513 for 1 ns under NPT at 300 K and 1.013 bar. Langevin dynamics (2.0 ps [−1] ) and the Monte 514 Carlo barostat were used. All MD simulations employed a 2 fs integration time step with the 515 SHAKE algorithm applied to constrain bonds involving hydrogen atoms. Long-range 516 electrostatic interactions were treated with the particle mesh Ewald method using a real 517 space cutoff of 10 Å, and the same cutoff was applied for short-range van der Waals 518 interactions. 519 520 **Enhanced Sampling via LB-PaCS-MD Simulation** 521 To efficiently sample the binding pathway of cyhalofop-acid into the fibril pocket, the Ligand 522 Binding Parallel Cascade Selection Molecular Dynamics (LB-PaCS-MD) approach was 523 employed through the PaCS-Q workflow. [77] For each of the eight systems, enhanced 524 sampling was conducted over 220 cycles, with 5 independent replicas (candidates) per cycle 525 and a production length of 100 ps per replica. At the end of each cycle, candidate trajectories 526 were ranked based on the ligand–pocket center-of-mass distance ( _d_ Lig–pocket ), and the five 527 snapshots exhibiting the shortest distances were selected as seed structures for the next 528 cycle. The selection algorithm employed a likelihood-maximization approach with additional [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 529 weighting factors to prevent artificial biasing toward unrealistic conformations. Selection 530 parameters were optimized through preliminary validation runs to ensure adequate sampling 531 of intermediate states. This iterative selection scheme progressively biases the ensemble 532 toward the bound state while preserving the stochastic nature of unbiased MD within each 533 short replica. Convergence was assessed by monitoring the evolution of _d_ Lig–pocket and 534 ensuring that fluctuations in this distance reached a plateau over the final 50 cycles. 535 536 **Free-Energy Landscape and Binding-State Analysis** 537 Two-dimensional free energy landscapes (2D-FEL) were constructed using _d_ Lig–pocket and the 538 number of protein-ligand contacts as reaction coordinates. Contact analysis was performed 539 using a distance cutoff of 4.0 Å between heavy atoms. Free energy surfaces were calculated 540 using the pyEMMA [78] package with Markov State Model (MSM) analysis, discretizing data 541 into 50×50 grids and computing free energy as F(i,j) = - _k_ B _T_ ln P(i,j), where P(i,j) represents 542 the probability of observing the system in bin (i,j). Global minimum states (GMS) were 543 identified as the lowest free energy regions in each 2D-FEL. Statistical significance of binding 544 affinity differences was assessed through bootstrapping analysis of trajectory data (n=1000 545 bootstrap samples). Representative structures from GMS were extracted and analyzed for 546 detailed protein-ligand interactions using VMD and in-house Python scripts. 547 548 **Statistics** 549 For box plots, the centre line indicates the median, boxes represent the interquartile range 550 (IQR), whiskers extend to 1.5 × IQR, and points indicate individual biological replicates. Violin 551 plots show the kernel density distribution. The centre line indicates the median, boxes 552 indicate the interquartile range, whiskers extend to 1.5 × IQR, and points represent individual 553 biological replicates. Two-sided _t_ -test was permed using the tidyplots R package. [79] Relative 554 shoot height was analyzed by one-way ANOVA followed by Dunnett’s multiple-comparison 555 test comparing each transgenic line with WT, using the multcomp R package. [80] 556 557 **Acknowledgements** 558 We thank Dr. Hiroyuki Kasahara for helpful discussions on the enzymatic properties of NIT4. 559 Computations were partially performed on the NIG supercomputer at ROIS National Institute 560 of Genetics. This work was supported in part by the JSPS KAKENHI Grant (JP19K06066, to 561 A.U., JP23K23612, to S.I.), JST PRESTO (JPMJPR24N2, to S.I.), Japan Association for 562 Advancement of Phyto-Regulators (JAPR) (JA-2022-1 to A.U.), and the FLOuRISH 563 Fellowship Program of TUAT to D.S. 564 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 565 **Author contribution** 566 S.I. and A.U. conceived the project. H.I., L.N., S.I. and A.U. characterized plants. M.I. and 567 C.A.T conducted herbicide metabolism study. L.N., D.S. and S.I. designed and performed 568 dual luciferase assay. H.I. and S.I. conducted RNA-seq analyses. L.N., S.I. and M.P.S. 569 conducted genome analyses. H.I., L.N., M.S., T.G. and S.I. transformed plants. M.E. 570 generated rice knockout mutants. K.Y. provided _E. phyllopogon_ seeds. K.N. and H.I. 571 performed genetic mapping. K.H., Y.S. and R.S. performed structure analysis. S.I., H.I., L.N. 572 and D.S. performed other data analyses. S.I., K.H. and K.N. wrote the paper with feedback 573 from all authors. 574 575 **Ethics declarations** 576 Competing interests 577 The authors declare no competing interests. 578 579 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 580 **References** 581 1. True, J. R. & Carroll, S. B. Gene co-option in physiological and morphological evolution. 582 _Annu. Rev. Cell Dev. Biol._ **18**, 53–80 (2002). 583 2. Jacob, F. Evolution and tinkering. _Science_ **196**, 1161–1166 (1977). 584 3. Di Stilio, V. S. & Garcia, A. G. K. Gene co-option across scales shapes plant diversity. 585 _Curr. Biol._ **36**, R620–R628 (2026). 586 4. Powles, S. B. & Yu, Q. Evolution in action: plants resistant to herbicides. _Annu. Rev._ 587 _Plant Biol._ **61**, 317–347 (2010). 588 5. Heap, I. Global perspective of herbicide-resistant weeds. _Pest Manag. Sci._ **70**, 1306– 589 1315 (2014). 590 6. Palumbi, S. R. Humans as the world’s greatest evolutionary force. _Science_ **293**, 1786– 591 1790 (2001). 592 7. Baucom, R. S. Evolutionary and ecological insights from herbicide-resistant weeds: what 593 have we learned about plant adaptation, and what is left to uncover? _New Phytol._ **223**, 594 68–82 (2019). 595 8. Kreiner, J. M. _et al._ Rapid weed adaptation and range expansion in response to 596 agriculture over the past two centuries. _Science_ **378**, 1079–1085 (2022). 597 9. Yu, Q. & Powles, S. Metabolism-based herbicide resistance and cross-resistance in crop 598 weeds: a threat to herbicide sustainability and global crop production. _Plant Physiol._ **166**, 599 1106–1118 (2014). 600 10. Edwards, R., Dixon, D. P., Cummins, I., Brazier-Hicks, M. & Skipsey, M. New 601 perspectives on the metabolism and detoxification of synthetic compounds in plants. in 602 _Plant Ecophysiology_ (eds. Schröder, P. & Collins, C. D.) 125–148 (Springer Netherlands, 603 Dordrecht, 2011). 604 11. Höfer, R. _et al._ Dual function of the cytochrome P450 CYP76 family from _Arabidopsis_ 605 _thaliana_ in the metabolism of monoterpenols and phenylurea herbicides. _Plant Physiol._ 606 **166**, 1149–1161 (2014). 607 12. Abdollahi, F. _et al._ Innate promiscuity of the CYP706 family of P450 enzymes provides 608 a suitable context for the evolution of dinitroaniline resistance in weed. _New Phytol._ **229**, 609 3253–3268 (2021). 610 13. Dimaano, N. G. & Iwakami, S. Cytochrome P450-mediated herbicide metabolism in [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 611 plants: current understanding and prospects. _Pest Manag. Sci._ **77**, 22–32 (2021). 612 14. Dixon, D. P., Skipsey, M. & Edwards, R. Roles for glutathione transferases in plant 613 secondary metabolism. _Phytochemistry_ **71**, 338–350 (2010). 614 15. Iwakami, S. & Yamaguchi, T. Identifying resistance‐conferring cytochrome P450s in 615 weeds: experimental strategies and practical considerations. _Pest Manag. Sci._ 616 https://doi.org/10.1002/ps.71279 (2026) doi:10.1002/ps.71279. 617 16. Gaines, T. A. _et al._ Mechanisms of evolved herbicide resistance. _J. Biol. Chem._ **295**, 618 10307–10330 (2020). 619 17. Yuan, J. S., Tranel, P. J. & Stewart, C. N., Jr. Non-target-site herbicide resistance: a 620 family business. _Trends Plant Sci._ **12**, 6–13 (2007). 621 18. Iwakami, S. _et al._ Multiple-herbicide resistance in _Echinochloa crus-galli_ var. 622 _formosensis_, an allohexaploid weed species, in dry-seeded rice. _Pestic. Biochem._ 623 _Physiol._ **119**, 1–8 (2015). 624 19. Iwakami, S. _et al._ Syntenic analysis of ACCase loci and target-site-resistance mutations 625 in cyhalofop-butyl resistant _Echinochloa crus-galli_ var. _crus-galli_ in Japan. _Pest Manag._ 626 _Sci._ **80**, 627–636 (2024). 627 20. Ruiz-Santaella, J. P., Heredia, A. & Prado, R. D. Basis of selectivity of cyhalofop-butyl in 628 _Oryza sativa_ L. _Planta_ **223**, 191–199 (2006). 629 21. Nishimura, K. _et al._ Degenerate oligonucleotide primer MIG-seq: an effective PCR 630 based method for high-throughput genotyping. _Plant J._ **118**, 2296–2317 (2024). 631 22. Howden, A. J. M., Jill Harrison, C. & Preston, G. M. A conserved mechanism for nitrile 632 metabolism in bacteria and plants. **57**, 243–253 (2009). 633 23. Piotrowski, M. Primary or secondary? Versatile nitrilases in plant metabolism. 634 _Phytochemistry_ **69**, 2655–2667 (2008). 635 24. Janowitz, T., Trompetter, I. & Piotrowski, M. Evolution of nitrilases in glucosinolate 636 containing plants. _Phytochemistry_ **70**, 1680–1686 (2009). 637 25. Pan, L. _et al._ CYP81A68 confers metabolic resistance to ALS and ACCase-inhibiting 638 herbicides and its epigenetic regulation in _Echinochloa crus-galli_ . _J. Hazard. Mater._ **428**, 639 128225 (2022). 640 26. Feng, T. _et al._ A natural glutathione _S_ -transferase gene _GSTU23_ confers metabolic 641 resistance to metamifop in _Echinochloa crus-galli_ . _Int. J. Biol. Macromol._ 134078 (2024). [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 642 27. Qi, J. _et al._ _EcCYP89B16_ confers metabolic herbicide resistance and is associated with 643 Calvin cycle adjustments in _Echinochloa crus-galli_ . _Plant Physiol._ **201**, kiag161 (2026). 644 28. Gondo, T., Sugahara, M., Ng, H. M., Yasuda, K. & Iwakami, S. Development of a stable 645 genetic transformation system in _Echinochloa phyllopogon_ . _Research Square_ 646 https://doi.org/10.21203/rs.3.rs-9989784/v1 (2026) doi:10.21203/rs.3.rs-9989784/v1. 647 29. Jenrich, R. _et al._ Evolution of heteromeric nitrilase complexes in Poaceae with new 648 functions in nitrile metabolism. _Proc. Natl. Acad. Sci. U. S. A._ **104**, 18848–18853 (2007). 649 30. Song, M. _et al._ OsNAR2.1 Interaction with OsNIT1 and OsNIT2 Functions in Root-growth 650 Responses to Nitrate and Ammonium. _Plant Physiol._ **183**, 289–303 (2020). 651 31. Kriechbaumer, V. _et al._ Maize nitrilases have a dual role in auxin homeostasis and beta 652 cyanoalanine hydrolysis. _J. Exp. Bot._ **58**, 4225–4233 (2007). 653 32. Wójcikowska, B. The role of nitrilases in auxin biosynthesis, plant development and 654 stress responses. _Plant Cell Physiol._ **67**, 115–129 (2026). 655 33. Mulelu, A. E., Kirykowicz, A. M. & Woodward, J. D. Cryo-EM and directed evolution 656 reveal how Arabidopsis nitrilase specificity is influenced by its quaternary structure. 657 _Commun Biol_ **2**, 260 (2019). 658 34. Woodward, J. D., Trompetter, I., Sewell, B. T. & Piotrowski, M. Substrate specificity of 659 plant nitrilase complexes is affected by their helical twist. _Commun Biol_ **1**, 186 (2018). 660 35. Jhala, A. J. _et al._ Acetyl-CoA carboxylase (ACCase)-inhibiting herbicides: Past, present, 661 and future. _Weed Technol._ **40**, 1–45 (2026). 662 36. Iwakami, S. _et al._ Cytochrome P450 CYP81A12 and CYP81A21 are associated with 663 resistance to two acetolactate synthase inhibitors in _Echinochloa phyllopogon_ . _Plant_ 664 _Physiol._ **165**, 618–629 (2014). 665 37. Suda, H. _et al._ Transcriptionally linked simultaneous overexpression of P450 genes for 666 broad-spectrum herbicide resistance. _Plant Physiol._ **192**, 3017–3029 (2023). 667 38. Han, H. _et al._ Cytochrome P450 CYP81A10v7 in _Lolium rigidum_ confers metabolic 668 resistance to herbicides across at least five modes of action. _Plant J._ **105**, 79–92 (2021). 669 39. Hawkins, N. J., Bass, C., Dixon, A. & Neve, P. The evolutionary origins of pesticide 670 resistance. _Biol. Rev. Camb. Philos. Soc._ **94**, 135–155 (2019). 671 40. Wu, D. _et al._ Genomic insights into the evolution of _Echinochloa_ species as weed and 672 orphan crop. _Nat. Commun._ **13**, 689 (2022). [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 673 41. Ohno, S. _Evolution by Gene Duplication_ . (Springer, New York, NY, 1971). 674 42. Soltis, P. S., Marchant, D. B., Van de Peer, Y. & Soltis, D. E. Polyploidy and genome 675 evolution in plants. _Curr. Opin. Genet. Dev._ **35**, 119–125 (2015). 676 43. Kondo, Y. Artificial hybridization of rice by the hot-water emasculation method. (In 677 Japanese). _Agric. and Hort_ **14**, 41–52 (1939). 678 44. Iwakami, S. _et al._ Cytochrome P450 genes induced by bispyribac-sodium treatment in a 679 multiple-herbicide-resistant biotype of _Echinochloa phyllopogon_ . _Pest Manag. Sci._ **70**, 680 549–558 (2014). 681 45. Toki, S. _et al._ Early infection of scutellum tissue with _Agrobacterium_ allows high-speed 682 transformation of rice. _Plant J._ **47**, 969–976 (2006). 683 46. Motohashi, K. A simple and efficient seamless DNA cloning method using SLiCE from 684 _Escherichia coli_ laboratory strains and its application to SLiP site-directed mutagenesis. 685 _BMC Biotechnol._ **15**, 47 (2015). 686 47. Christensen, A. H. & Quail, P. H. Ubiquitin promoter-based vectors for high-level 687 expression of selectable and/or screenable marker genes in monocotyledonous plants. 688 _Transgenic Res._ **5**, 213–218 (1996). 689 48. Mikami, M., Toki, S. & Endo, M. Comparison of CRISPR/Cas9 expression constructs for 690 efficient targeted mutagenesis in rice. _Plant Mol. Biol._ **88**, 561–572 (2015). 691 49. Liu, H. _et al._ CRISPR-P 2.0: An improved CRISPR-Cas9 tool for genome editing in plants. 692 _Mol. Plant_ **10**, 530–532 (2017). 693 50. Karimi, M., Inzé, D. & Depicker, A. GATEWAY [TM] vectors for _Agrobacterium_ -mediated 694 plant transformation. _Trends Plant Sci._ **7**, 193–195 (2002). 695 51. Murashige, T. & Skoog, F. A revised medium for rapid growth and bio assays with 696 tobacco tissue cultures. _Physiol. Plant._ **15**, 473–497 (1962). 697 52. Akashi, R., Yuge, C., Gondo, T., Kawamura, O. & Hoffmann, F. Bialaphos-resistant cells 698 of dallisgrass ( _Paspalum dilatatum_ Poir.) through particle bombardment with a simple 699 self-built inflow gun. _Jpn. J. Grassl. Sci._ **47**, 588–593 (2002). 700 53. Schmittgen, T. D. & Livak, K. J. Analyzing real-time PCR data by the comparative C(T) 701 method. _Nat. Protoc._ **3**, 1101–1108 (2008). 702 54. Murray, M. G. & Thompson, W. F. Rapid isolation of high molecular weight plant DNA. 703 _Nucleic Acids Res._ **8**, 4321–4326 (1980). [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 704 55. Cheng, H., Concepcion, G. T., Feng, X., Zhang, H. & Li, H. Haplotype-resolved de novo 705 assembly using phased assembly graphs with hifiasm. _Nat. Methods_ **18**, 170–175 706 (2021). 707 56. Karlicki, M., Antonowicz, S. & Karnkowska, A. Tiara: deep learning-based classification 708 system for eukaryotic sequences. _Bioinformatics_ **38**, 344–350 (2022). 709 57. Alonge, M. _et al._ Automated assembly scaffolding using RagTag elevates a new tomato 710 system for high-throughput genome editing. _Genome Biol._ **23**, 258 (2022). 711 58. Shumate, A. & Salzberg, S. L. Liftoff: accurate mapping of gene annotations. 712 _Bioinformatics_ **37**, 1639–1643 (2021). 713 59. Iwakami, S. _et al._ CYP81A P450s are involved in concomitant cross-resistance to 714 acetolactate synthase and acetyl-CoA carboxylase herbicides in _Echinochloa_ 715 _phyllopogon_ . _New Phytol._ **221**, 2112–2122 (2019). 716 60. Chen, S., Zhou, Y., Chen, Y. & Gu, J. fastp: an ultra-fast all-in-one FASTQ preprocessor. 717 _Bioinformatics_ **34**, i884–i890 (2018). 718 61. Dobin, A. _et al._ STAR: ultrafast universal RNA-seq aligner. _Bioinformatics_ **29**, 15–21 719 (2013). 720 62. Li, B. & Dewey, C. N. RSEM: accurate transcript quantification from RNA-Seq data with 721 or without a reference genome. _BMC Bioinformatics_ **12**, 323 (2011). 722 63. Robinson, M. D., McCarthy, D. J. & Smyth, G. K. edgeR: a Bioconductor package for 723 differential expression analysis of digital gene expression data. _Bioinformatics_ **26**, 139– 724 140 (2010). 725 64. Sahlin, K. Strobealign: flexible seed size enables ultra-fast and accurate read alignment. 726 _Genome Biol._ **23**, 260 (2022). 727 65. Danecek, P. _et al._ Twelve years of SAMtools and BCFtools. _Gigascience_ **10**, giab008 728 (2021). 729 66. Breese, M. R. & Liu, Y. NGSUtils: a software suite for analyzing and manipulating next 730 generation sequencing datasets. _Bioinformatics_ **29**, 494–496 (2013). 731 67. Rastas, P. Lep-MAP3: robust linkage mapping even for low-coverage whole genome 732 sequencing data. _Bioinformatics_ **33**, 3726–3732 (2017). 733 68. Broman, K. W., Wu, H., Sen, S. & Churchill, G. A. R/qtl: QTL mapping in experimental 734 crosses. _Bioinformatics_ **19**, 889–890 (2003). [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 735 69. Katoh, K. & Standley, D. M. MAFFT multiple sequence alignment software version 7: 736 improvements in performance and usability. _Mol. Biol. Evol._ **30**, 772–780 (2013). 737 70. Nguyen, L. T., Schmidt, H. A., Von, H. A. & Minh, B. Q. IQ-TREE: a fast and effective 738 stochastic algorithm for estimating maximum-likelihood phylogenies. _Molecular biology_ 739 _and evolution_ **32**, 268–274 (2015). 740 71. Letunic, I. & Bork, P. Interactive Tree of Life (iTOL) v6: recent updates to the phylogenetic 741 tree display and annotation tool. _Nucleic Acids Res._ **52**, W78–W82 (2024). 742 72. Gilchrist, C. L. M. & Chooi, Y.-H. Clinker & clustermap.js: Automatic generation of gene 743 cluster comparison figures. _Bioinformatics_ **37**, 2473–2475 (2021). 744 73. Wagih, O. ggseqlogo: a versatile R package for drawing sequence logos. _Bioinformatics_ 745 **33**, 3645–3647 (2017). 746 74. Waterhouse, A. _et al._ SWISS-MODEL: homology modelling of protein structures and 747 complexes. _Nucleic Acids Res._ **46**, W296–W303 (2018). 748 75. Case, D. A. _et al._ Amber 2024. _University of California, San Francisco_ . 749 76. Frisch, M. J. _et al._ Gaussian 16 Rev. C.01. _Wallingford, CT_ (2016). 750 77. Duan, L., Hengphasatporn, K. & Shigeta, Y. PaCS-Q: Python toolkits for path sampling 751 in MD and QM/MM MD simulation. _J. Chem. Inf. Model._ **65**, 6441–6445 (2025). 752 78. Senne, M., Trendelkamp-Schroer, B., Mey, A. S. J. S., Schütte, C. & Noé, F. EMMA: A 753 software package for Markov model building and analysis. _J. Chem. Theory Comput._ **8**, 754 2223–2238 (2012). 755 79. Engler, J. B. Tidyplots empowers life scientists with easy code-based data visualization. 756 _Imeta_ **4**, e70018 (2025). 757 80. Hothorn, T., Bretz, F. & Westfall, P. Simultaneous inference in general parametric models. 758 _Biom. J._ **50**, 346–363 (2008). 759 760 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 761 **Figure legends** 762 **Figure 1. Physiological characterization of cyhalofop-butyl resistance in** _**Echinochloa**_ 763 _**crus-galli**_ 764 **a,** Previously reported cross-resistance profiles of the susceptible line (S1) and two resistant 765 lines (R1 and R2). **b,** Representative phenotypes of S1, R1 and R2 following foliar application 766 of cyhalofop-butyl. **c,** Time-course LC–MS/MS analysis of cyhalofop-butyl metabolism to 767 cyhalofop-acid and cyhalofop-diacid in S1, R1 and R2. Data represent means ± s.e.m. (n = 768 3). **d,** Cyhalofop-acid sensitvitiy. Representative images were obtained after culture in liquid 769 medium containing 6 μM cyhalofop-acid. Growth was quantified relative to untreated controls. 770 Data represent means ± s.e.m. (n = 3). 771 772 **Figure 2. Genetic mapping identifies a major cyhalofop-butyl (CB) resistance locus** 773 **a,** Sequence chromatograms confirming successful hybridization of the S1 × R1 F1 plant at 774 a diagnostic SNP. **b,** Segregation of CB resistance in the F2 population derived from S1 × 775 R1. **c,** Representative phenotypes of F2:3 families derived from resistant and susceptible F2 776 plants. **d,** Genome-wide LOD profile from interval mapping (IM). The dashed line indicates 777 the significance threshold. **e,** Fine mapping of the resistance locus using recombinant 778 individuals narrowed the candidate interval to a 1.42-Mb region between markers H045 and 779 H008. RI, resistance indices. 780 781 **Fig. 3 Independent transcriptome analyses identify R1-specific upregulation of** 782 _**EcNIT4A2**_ **, a nitrilase in the cyanide-detoxification pathway.** 783 **a,** Collection sites of the resistant lines R1 and R2 and susceptible lines S1–S4 in Japan. **b,** 784 Venn diagrams of differentially expressed genes identified in two independent RNA-seq 785 analyses. The R1-specific gene sets from the first (R1, R2 and S1) and second (R1 and S2– 786 S4) analyses shared 67 genes. **c,** RT–qPCR analysis of _EcNIT4A2_ expression in S1, R1 and 787 seven sensitive and seven resistant F 3 families. Expression is shown relative to S1. Points 788 represent biological replicates (n=3); bars indicate mean ± s.e.m. **d,** Ethylene biosynthesis 789 and the associated cyanide-detoxification pathway involving β-cyanoalanine synthase (β 790 CAS) and NIT4. SAM, S-adenosyl-L-methionine; ACC, 1-aminocyclopropane-1-carboxylic 791 acid. **e,** Maximum-likelihood phylogeny of plant NIT4 proteins. **f,** Microsynteny of the _NIT4A_ [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 792 and _NIT4B_ regions in grasses. Connecting ribbons indicate sequence identity. **g,** Transcript 793 abundance of the three _EcNIT4A_ and three _EcNIT4B_ homoeologues in S1, R2 and R1 in the 794 first RNA-seq dataset (n = 3 biological replicates). *, R1-specific increase in _EcNIT4A2_ 795 identified by edgeR analysis. 796 797 **Figure 4. Functional validation of** _**EcNIT4A2**_ **as the cyhalofop-butyl (CB) resistance** 798 **gene.** 799 **a,** Rice calli expressing _EcNIT4A2_ or _EcNIT4B2_ cultured on medium containing CB. **b,** 800 _Echinochloa phyllopogon_ plants transformed with _EcNIT4A2_ . Relative _NIT4A2_ expression, 801 including both transgenic _EcNIT4A2_ and endogenous _EpNIT4A2_ transcripts, was quantified 802 by RNA-seq. **c,** RT–qPCR analysis of _OsNIT4A_ expression in _OsNIT4A_ -overexpressing rice 803 plants ( _n_ = 4). **d,** CB sensitivity was assessed by measuring relative shoot length following 804 treatment with 300 µM CB (n = 15). * and ** indicate _P_ < 0.05 and _P_ < 0.01, respectively, 805 compared with WT (Dunnett’s multiple-comparison test). Representative phenotypes of rice 806 plants overexpressing _OsNIT4A_ following treatment with 300 µM CB. **e,** Design of guide 807 RNAs targeting _OsNIT4A_ and _OsNIT4B_ in rice. **f,** CB sensitivity of the knockout mutants. 808 Data are presented as mean ± s.e.m (n=3). Yellow bar, 2 cm. 809 810 **Figure 5 Structural basis of enhanced cyhalofop metabolism by the D317G** 811 **substitution in EcNIT4A2.** 812 **a,** Distribution of five amino acid substitutions among EcNIT4A2 from R1 and S1 and the 813 homoeologous EcNIT4A1 and EcNIT4A3 proteins. Filled squares indicate residues differing 814 from S1 EcNIT4A2. **b,** Sequence logo of residues 311–322 from grass NIT4A homologues. 815 **c,** Rice calli expressing R1 or S1 _EcNIT4A2_, _EcNIT4A1_ or _EcNIT4A3_ on medium containing 816 the indicated concentrations of cyhalofop butyl. **d,** Rice calli expressing wild-type _OsNIT4A_ 817 or _OsNIT4A_ carrying D317G on medium containing the indicated concentrations of cyhalofop 818 butyl. **e,** Eight modelled EcNIT4A–EcNIT4B assemblies (lock–lid–pocket) for the A [S] (upper 819 panel) and A [R] (lower panel) variants. Lid–lock distances were measured between residue 820 Cα atoms in the final structures obtained from LB-PaCS-MD simulations. EcNIT4A2 occupies 821 the lock position throughout. **f,** Binding poses of A [S] –B–B (left) and A [R] –B–B (right), showing 822 lock (chain A, tan), lid (chain B, blue) and pocket (chain C, grey). Insets show cyhalofop-acid 823 relative to the catalytic C195. Lower panels, ligand–C195 (blue) and R93–residue 317 824 (purple) distances over simulation progression (mean ± s.d.). **g,** Minimum-energy 825 conformation of A [R] –B–B. Inset shows cyhalofop-acid packed against R93, with the 826 carboxylate coordinated by K161, E168 and E74 and the nitrile carbon 5.99 Å from C195. [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.06.764846; this version posted November 11, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.06.764846) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 827 Residue positions are reported using AtNIT4-equivalent numbering (PDB 6I5T). 828 829 **Figure 6. A promoter structural variant enhances EcNIT4A2 expression** 830 **a,** Comparison of the _EcNIT4A2_ promoter regions in S1 and R1. **b,** PCR validation of the 831 promoter structural variant in representative susceptible and resistant F 3 plants. **c,** Promoter 832 activity of the S1 and R1 _EcNIT4A2_ promoter alleles measured using a dual-luciferase 833 reporter assay in rice calli. Firefly luciferase activity was normalized to NanoLuc luciferase. 834 **, _P_ < 0.01 (two-sided Student’s _t_ -test).