Tracing the origin of non-brittle rachis alleles in wheat

10 The evolution of non-brittle rachises, controlled by the TtBTR1 genes, was a key step during 11 wheat domestication. Here, using k-mer-based approaches applied to a large diversity panel, 12 we refine previous estimates of the geographical and temporal origins of the three known 13 Ttbtr1 loss-of-function alleles and show that they emerged in distinct wild emmer 14 subpopulations. For this, we generated a chromosome -scale dika wheat assembly carrying 15 the recently discovered Ttbtr1-Ab allele. Our analyses reveal that the Ttbtr1-A alleles reside 16 on an introgression from the southern judaicum wild emmer population into northern wild 17 emmer wheat, providing an explanation for the long -standing debate about the Ttbtr1-A 18 origin. We further demonstrate that the Ttbtr1-Aa and Ttbtr1-B alleles are already present in 19 wild emmer wheat, with evidence indicating that they arose prior to the advent of agriculture. 20 Together, these findings support a model in which key domestication genes in domesticated 21 crops were selected and combined from standing genetic variation in wild relatives. 22 23 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 3

24 The transition to agriculture in the Near Eas t marked one of humankind’s most profound 25 sociocultural transformations, laying the foundation for modern civilizations. This shift was 26 closely tied to the domestication of plants and animals. In cereals, a key target of domestication 27 was rachis brittleness 1-4. While the wild progenitors of cereals shed grains at maturity through 28 brittle internodes, cultivation favored spikes that remained intact for harvesting. Key genes 29 contributing to rachis brittleness in Triticeae were first discovered in barley, where loss-of-function 30 mutations in either the BTR1 or BTR2 gene confer the non-brittle phenotype3,5. 31 32 In wheat, non-brittleness is similarly caused by loss-of-function mutations in BTR1. Three Ttbtr1 33 alleles have been identified in domesticated wheats of the emmer lineage, which includes both 34 durum wheat (Triticum turgidum ssp. durum) and bread wheat (T. aestivum), the two economically 35 most important wheat species today . Initially, only two Ttbtr1 alleles were known, Ttbtr1-Aa on 36 chromosome 3A and Ttbtr1-B on chromosome 3B . The combination of Ttbtr1-Aa and Ttbtr1-B 37

in non-brittleness, leading to the view that the emergence of this trait in the emmer wheat 38 lineage was monophyletic. The Ttbtr1-Aa allele carries a two-base-pair deletion, whereas Ttbtr1-39 B harbors a ~4 kb insertion2,4. Both mutations disrupt the TtBTR1 coding sequence. More recently, 40 a second Ttbtr1-A allele, Ttbtr1-Ab, was discovered. It carries a 5,029 bp retrotransposon insertion 41 and has been found in tetraploid T. turgidum ssp. carthlicum (dika wheat) as well as in some 42 domesticated tetraploid wheat accessions from Ethiopia, revealing two independent origins of the 43 Ttbtr1-A allele6. The exact evolutionary origins of Ttbtr1 and domesticated wheat remain 44 contested7-9, a debate further complicated by the mosaic haplotype composition of cereal genomes 45 shaped by extensive gene flow among populations10-13. 46 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 4 47

and Discussion 48 To investigate the origin and distribution of the recently discovered Ttbtr1-Ab allele6, as well as 49 the evolution of non -brittleness, we first generated a chromosome -scale assembly of the Ttbtr1-50 Ab-carrying T. turgidum ssp. carthlicum accession CWI 22960 by combining PacBio HiFi reads 51 with Hi-C. The assembly spanned 10.88 Gb with a contig N50 of 42.6 Mb (Supplementary Table 52 1). In this accession, Ttbtr1-Ab carried the 5,029 bp retrotransposon insertion at position 459 bp, 53 with two identical 248 bp long terminal repeats (LTRs). Next, we performed an allelic diversity 54 analysis at Ttbtr1-A and Ttbtr1-B using a k-mer database derived from whole-genome sequencing 55 data of 2,130 domesticated and 463 wild wheat accessions (Supplementary Table 2, 3), including 56 both tetraploids and hexaploids. In total, 98.6% of the domesticated accessions carried the 2 bp 57 deletion characteristic for Ttbtr1-Aa. In contrast, only 30 (1.4%) domesticated wheat accessions 58 harbored the recently identified Ttbtr1-Ab allele. The 30 accessions represent tetraploid wheats, 59 13 of which are classified as T. turgidum ssp. carthlicum (dika wheat), 13 domesticated Ethiopian 60 tetraploids, two T. turgidum ssp. durum, one T. turgidum ssp. polonicum, and one T. turgidum ssp. 61 turanicum (Supplementary Table 2). 62 63 No additional loss-of-function allele was found for Ttbtr1-B. Among the 463 wild emmer wheat 64 accessions, nine carried the Ttbtr1-Aa allele (Ttbtr1-Aa / TtBtr1-B), and seven wild emmer wheat 65 accessions had the Ttbtr1-B allele (TtBtr1-A / Ttbtr1-B) (Supplementary Table 3). Five wild emmer 66 wheat accessions caried both Ttbtr1-Aa and Ttbtr1-B, which can be due to accession 67 misclassification or gene flow from domesticated wheat . No wild emmer wheat lines were found 68 carrying the newly identified Ttbtr1-Ab allele. 69 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 5 70 Wild emmer wheat can be grouped into three main clades, (i) a northeastern population comprising 71 accessions collected from present -day southern Anatolia, Iran, and Iraq (EST_WE W), (ii) a 72 Southern Levant population (LEV_WEW), and (iii) race judaicum (JUD_WEW) found around the 73 Sea of Galilee4,6,14. A k-mer-based phylogeny across the 463 wild emmer wheat accessions and 30 74 tetraploid domesticated wheat accessions recovered these three major wild emmer wheat clades. 75 The LEV_WEW population further split into two subgroups (LEV_WEW-1 and LEV_WEW-2) 76 along the north -south gradient, while the EST_WEW population resolved into five subgroups 77 (EST_WEW-1 to EST_WEW -5), highly associated with the geographical provenance of the 78 accessions from west to east (Fig. 1a, b; Supplementary Figs. 1-3; Supplementary Table 3). 79 80 To further refine the phylogeny of the TtBtr1 loci, we selected non-recombining 300-kb genomic 81 segments surrounding TtBTR1-A and TtBTR1-B and constructed two phylogenetic trees for the 82 complete diversity panel based on identity-by-state scores, using the assembly of Chinese Spring15 83 (hexaploid wheat, Ttbtr1-Aa / Ttbtr1-B) as a reference (Fig. 1c, Supplementary Fig. 4, 5). For 84 Ttbtr1-A, the domesticated wheat accessions carrying Ttbtr1-Aa and Ttbtr1-Ab clustered into two 85 distinct branches of the phylogenetic tree (Fig. 1c). Most of the wild emmer wheat accessions with 86 Ttbtr1-Aa belong to the northern subpopulation EST_WEW-5 from Iran and Iraq, whereas the 87 closest wild emmer wheat relatives with the wild-type TtBtr1-A allele belong to the northern 88 subpopulation EST_WEW-3 from southern Anatolia, for both Ttbtr1-Aa and Ttbtr1-Ab (Fig. 1c). 89 This broader northern distribution region has previously been recognized as the most likely origin 90 of Ttbtr1-Aa, although its precise geographic origin and subpopulation could not be determined 7-91 9. Notably, we found that Ttbtr1-Aa and Ttbtr1-Ab reside within a genomic introgression from the 92 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 6 judaicum subpopulation from Southern Levant. This introgression has become widespread across 93 northern wild emmer wheats (Fig. 1d), accounting for the difficulty in pinpointing the exact origin 94 of Ttbtr1-A, but is mostly absent in the Southern Levant wild emmer wheat . The widespread 95 distribution of the judaicum introgression in EST_WEW is consistent with the previously reported 96 low genetic diversity at the Ttbtr1-A locus6. The longest version of the judaicum introgression was 97 found in CWI 22960 and two other T. turgidum ssp. carthlicum accessions, spanning around 140 98 Mb. 99 100 Consistent with previous reports, the Levant region represents the most likely origin of Ttbtr1-B. 101 Five of the seven wild emmer wheat accessions carrying Ttbtr1-B fall within the EST_WEW-1 102 subpopulation. While most of the accessions belonging to EST_WEW -1 have an unknown 103 collection site, the accessions with known geographical location were all from Lebanon. The 104 genetically closest wild emmer wheat accessions with the wild-type TtBtr1-B allele belong to the 105 Southern Levant subpopulations. The five genetically closest wild emmer wheat accessions with 106 wild-type TtBtr1-B were collected from Lebanon (3 accessions), Syria, and Israel (Fig. 1c). Taken 107 together, these phylogenetic patterns point to the northern margin of the Southern Levant region 108 as the most probable origin of Ttbtr1-B. 109 110 The occurrence of Ttbtr1-Aa and Ttbtr1-B in wild emmer wheat may reflect either post-111 domestication gene flow from domesticated wheat or standing genetic variation that predates 112 agriculture, analogous to the teosinte glume architecture (tga1) allele in maize16 and the btr1 allele 113 in barley10. To estimate the timing of Ttbtr1-B emergence, we dated the associated retrotransposon 114 insertion by evaluating SNPs between its two long terminal repeats17. The previously reported ~4 115 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 7 kb insertion in Ttbtr1-B was an underestimation, resulting from the limitations of short-read-based 116 genome assemblies4. Ttbtr1-B sequences from 13 long-read-based domesticated wheat assemblies 117 (Supplementary Table 4) revealed a 11.97 kb retrotransposon insertion with two long terminal 118 repeats of approximately 3,894 bp (with minor size differences caused by homopolymers) . In 119 addition, we mapped raw reads from the seven wild emmer wheat accessions carrying Ttbtr1-B 120 (TtBtr1-Aa / Ttbtr1-B), four wild emmer wheat accessions carrying both Ttbtr1-Aa and Ttbtr1-B, 121 and six domesticated tetraploid wheat accessions to the Chinese Spring reference genome15. SNPs 122 were then called across the LTRs. Phylogenetic analyses revealed that some Ttbtr1-B-carrying wild 123 emmer wheat accessions clustered with domesticated wheat, consistent with post-domestication 124 gene flow, while other wild emmer wheat accessions formed a distinct clade defined by private 125 SNPs unique to this group (Fig. 1e). Among the latter are TA11213 and GT004, two wild emmer 126 wheat accessions from Lebanon belonging to subpopulation EST_WEW -1. LTR dating based on 127 the 13 chromosome-scale assemblies indicated a transposon insertion age of ~100,000 ± 30,000 128 years (Supplementary Table 5) . While molecular dating relies on assumptions about mutation 129 rates17, even our most recent estimate (49,000 ± 22,000 years) places the origin of Ttbtr1-B around 130 27,000 years ago, well before the beginning of agriculture. A pre-domestication origin of Ttbtr-1B 131 is further supported by the SNP-based phylogenetic analysis, which showed that some LTR -132 specific SNPs occur exclusively in wild emmer wheat and are absent from domesticated wheat. In 133 addition, this timing is consistent with the finding of domestic-type wild emmer wheat spikes at 134 the Ohalo II archaeological site before the advent of agriculture18. The long terminal repeats of the 135 retrotransposon inserted in Ttbtr1-Ab were too short for dating. We thus used a SNP-based dating 136 approach10, estimating the emergence of the Ttbtr1-Aa and Ttbtr1-Ab mutations approximately 137 38,000 and 33,600 years ago. 138 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 8 139 Together, our findings support a model in which key mutations for rachis brittleness arose in 140 different wild emmer wheat subpopulations and were maintained in natural or early human -141 managed settings. Wild emmer wheat accessions carrying a single non-brittle rachis allele showed 142 brittle/semi-brittle rachises (Fig. 2). Plants with semi-brittle rachises have been found in stands of 143 wild emmer wheat, and it has been reported that environmental factors such as humidity and 144 temperature likely influence spike maturation and shattering19. The presence of a single non-brittle 145 rachis allele in wild emmer wheat therefore likely has no strong fitness effect and spikes will 146 eventually shatter. The combination of Ttbtr1-A and Ttbtr1-B alleles through hybridization, 147 resulting in fully domesticated wheat with non-brittle rachis, likely occurred multiple times, giving 148 rise to distinct wheat subspecies. 149 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 9 150 Figure 1. Origin of non-brittle rachis in wheat. a, Map showing the collection sites and the wild emmer 151 wheat accessions analyzed in this study. Only accessions with known collection sites are shown. IRQ, Iraq; 152 IRN, Islamic Republic of Iran; JOR, Hashemite Kingdom of Jordan; SAU, Kingdom of Saudi Arabia; SYR, 153 Syrian Arab Republic; TUR, Türkiye. Jittering is applied for map readability. Colors for the different 154 subpopulations and symbols for the TtBtr1 allele combinations will be used throughout the figure. b, k-mer-155 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 10 based network representation of the wild emmer wheat population structure. Accessions with a normalized 156 distance closer than 0.61 are connected. c, k-mer-based p hylogenetic t rees across the 300-kb genomic 157 segments surrounding the TtBTR1-A (left) and the TtBTR1-B loci (right). For TtBTR1-A (left) only samples 158 showing identity-by-state across the whole locus are shown to increase the resolution. For the TtBTR1-B 159 locus (right), only a subset of accessions from distinct populations is reported for the readability of the tree. 160 Bootstrap support values are represented by node thickness . d, Genome identity heatmap across 161 chromosome 3A (position 61.4-70 Mb). The wild emmer wheat Zavitan, belonging to the judaicum race, 162 was used as a reference. Dark regions indicate identity -by-state to Zavitan, while red color indicates 163 genomic segments that are distant from Zavitan. The Ttbtr1-Aa and Ttbtr1-Ab alleles are located on a 164 judaicum introgression that is widespread across the northern wild emmer wheat group (EST_WEW) but 165 absent in the Southern Levant wild emmer wheat group (LEV_WEW). e, Phylogenetic tree obtained 166 comparing the SNPs presents on the long terminal repeats (LTRs) of the transposable element inserted in 167 the Ttbtr1-B allele. Samples can be identified by the numbers reported in the panel and in Supplementary 168 Tables 2 and 3. 169 170 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 11 171 Figure 2. Phenotypic representation for wild emmer wheat spikes carrying Ttbtr1-Aa, Ttbtr1-B or both at 172 the same time. All spikes show a semi-brittle phenotype comparable to the one of domesticated emmer. 173 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 12

174 175 T. turgidum ssp. carthlicum (CWI 22960) assembly 176 High molecular weight DNA was extracted from dark-treated 2-week-old seedling s using the 177 Qiagen Genomic Tip kit. The output of four PacBio Revio SMRT cells (in total 364.62 Gb, 33.5-178 fold coverage, 23.99 million reads, 15,211 bp average read length) was assembled with hifiasm (v. 179 0.19.8)20 with default settings. The 799 million Illumina short reads generated by the sequencing 180 of a Hi -C library were first mapped to the CWI 22960 contig -level assembly with BWA-MEM 181 using the Arima Genomic mapping pipeline ’s default settings 182 (https://github.com/ArimaGenomics/mapping_pipeline). The resulting file was then processed 183 with YaHS (v. 1.1)21. A few rounds of manual curations were performed with the help of Juicebox 184 (v. 2.15) 22 and CHROMEISTER 23. PacBio Revio sequencing was performed in the KAUST 185 Bioscience Core Lab. Hi-C library preparation and sequencing was done by CNRGV as a service. 186 187 Whole-genome sequencing 188 Genomic DNA from the 25 wheat accessions was extracted from one or two young leaves of a 189 single plant following the CTAB protocol described by Abrouk et al .24. PCR-free library 190 preparation and sequencing at 12-fold coverage were performed as a service by Psomagen Inc. 191 (Rockville, Maryland, United States). 192 193 Whole-genome sequencing data collection. 194 Publicly available wheat sequencing data were retrieved from NCBI from the following project 195 numbers: PRJNA1007489, PRJNA759292, PRJNA628827, PRJNA476679, PRJNA310175, 196 PRJNA688544, PRJNA1070409, PRJNA1188632, PRJNA439156, PRJNA476679, 197 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 13 PRJNA596843, PRJNA597250, PRJNA663409, PRJNA669381, PRJNA670578, 198 PRJNA694980, PRJNA714281, PRJNA722149, PRJNA729723, PRJNA744310, PRJNA745496, 199 PRJNA759292, PRJNA771357, PRJNA790490, PRJNA820989, PRJNA877303, PRJNA900700, 200 PRJNA918327, PRJNA956839, PRJNA986484, PRJNA986532; from EBI-ENA from the 201 following project numbers: PRJEB61424, PRJEB22687, PRJEB44721, PRJEB45541, 202 PRJEB48529, PRJEB49351; and from the Genome Sequence Archive in the BIG Data Center from 203 the following project numbers: PRJCA004228, PRJCA004273, PRJCA005979, PRJCA009783, 204 PRJCA019508, PRJCA019636, PRJCA021345. 205 206 k-mer counting 207 Raw Illumina reads were cleaned with Trimmomatic (v. 0.39) 25 with the following settings: 208 LEADING:3 TRAILING:3 SLIDINGWINDOW:4:25 MINLEN:75. 31 -mers were counted with 209 KMC3 (v. 3.1.2)26. 210 211 k-mer-based phylogeny 212 The pairwise intersections between k-mer sets representing different accessions were computed 213 with the FastIBS KDBIntersect function (https://github.com/githubcbrc/FastIBS). To account for 214 different coverages, which influences the total number of k-mers in a set, we applied a ‘reduction 215 factor’ to each comparison value depending on the number of total k-mers present in each of the 216 two datasets. To compute the ‘reduction factor’ , a Random Forest Regressor was trained on the 217 surface obtained with the intersections of k-mer set s, simulating different coverages for two 218 datasets (CRR061704 and SRR11670754). For CRR061704, 19 datasets corresponding to 4.5-fold 219 to 16-fold coverage were obtained, while for SRR11670754, 28 datasets corresponding to 4.5-fold 220 to 26 -fold coverage were obtained (Supplementary Fig. 6) . This process was executed with a 221 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 14 Python (v. 3.8.8) script with the package scikit -learn (v. 1.2.2 ; 222 https://github.com/emilecg/btr_analysis). 223 PCA and Hierarchical clustering were executed with a Python (v. 3.8.8) script with the package 224 scikit-learn (v. 1.2.2). 225 The results obtained from the ‘reduction factor’ subtraction were then normalized to have values 226 in a 0 to 1 range. To achieve this, we applied the following formula, where X is the value of the 227 normalized comparison, Min is the minimum value obtained from all the normalized comparisons, 228 and Max is the maximum value obtained from all the normalized comparisons: 229 230 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 = 1 − 𝑋 − 𝑀𝑖𝑛 𝑀𝑎𝑥 − 𝑀𝑖𝑛 231 232 The network from the distance matrix was built using the Python package NetworkX (v3.4.2)27. 233 234 Fastibsmapper – allelic diversity analysis 235 In order to perform the allele diversity analysis, we first used TtBtr1-A and TtBtr1-B gene 236 sequences from Zavitan4 as query to search the genome of CWI 86942 using BLAST (v. 2.16.0)28 237 and extracted with Samtools (v. 1.16.1)29. 238 The gene sequences were used as reference for a FastIBS fastibsmapper run against all the k-mer 239 sets generated in this study. The resulting files were stacked in two tables, plotted with a Python 240 script (available at https://github.com/emilecg/ btr_analysis) and were visually inspecte d to 241 determine allelic diversity across the accessions. 242 243 FastIBS 244 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 15 FastIBS (https://github.com/githubcbrc/FastIBS) was used to identify wild emmer wheat 245 accessions carrying a non-recombinant, identical-by-state haplotype block surrounding TtBtr1-Aa 246 and TtBtr1-B using Chinese Spring as a reference, and TtBtr1-Ab using the CWI 22960 assembly 247 as a reference. Based on the FastIBS results, plotted as heatmaps, we were able to identify the non-248 recombining wild emmer wheat accessions for the 300-kb genomic segment surrounding the three 249 loci. The three 300 -kb loci from Chinese Spring and CWI 22960 where extracted from their 250 respective assemblies using Samtools (v. 1.16.1)29 to be used as references for the local phylogeny 251 analyses. 252 253 Local phylogeny for TtBtr1-A and TtBtr1-B 254 Three 300-kb and one 30-kb segments surrounding TtBtr1-A and TtBtr1-b were used as reference 255 for a FastIBS fastibsmapper run using all the k-mer sets generated in this study. The generation of 256 the phylogenetic tree from the stacked output of fastibsmapper required four distinct steps. In the 257 first step, all the reported drops in k-mer coverage with a maximum depth of 25 and a width of 5 258 bases were stored as a separated file. We defined a drop in k-mer coverage following the described 259 criteria as a ‘valley’. This file was inspected to discard valleys that occurred in a single accession, 260 all the other valleys were retained and saved in an output file. Two valleys were considered the 261 same if their start and end positions occurred within 5 bases. This file was then processed in the 262 third step: the whole list of valleys was organized in a presence absence matrix. The matrix was 263 then converted manually into a PHYLIP format and given to IQ-TREE (v. 3.0.1)30 as an input. A 264 first tree was produced to assess the general topology. After that, two more informative trees were 265 produced with a subset including all wild emmer wheat accessions and a subset of the domesticated 266 wheat accessions considering the very high level of similarity among this second group. The model 267 with the best fit was selected by IQ-TREE (GTR2+FO+R4) to construct the tree. Bootstrap support 268 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 16 values were calculated from 1,000 replicates. The tree plot was made with the online tool iTOL (v. 269 7.2.2)31. The trees were further pruned for tree readability. The three first steps were implemented 270 with three different Python scripts available at https://github.com/emilecg/btr_analysis. 271 272 Time estimation of the Ttbtr1-B retrotransposon insertion 273 Clustal Omega (v. 1.2.4)32 was used to align the LTR sequences extracted from 13 PacBio -based 274 genomes (Supplementary Table 4). Raw reads from eleven wild emmer wheat accessions and six 275 domesticated tetraploid accessions were mapped to the Chinese Spring assembly 15 using BWA-276 MEM (v. 0.7.17)33 and SNPs were called over the two LTRs using bcftools mpileup (v. 1.16) 29 277 with default settings. A manual inspection of the mapping results was performed to select true 278 SNPs. The time estimation of the retrotransposon’s insertion in to Ttbtr1-B was performed 279 according to Wicker et al. 202217. 280 281 Time estimation of the Ttbtr1-A loss-of-function mutation events 282 This estimate is based on the work of Guo et al. 202510 in barley. Whole-genome sequencing reads 283 from 61 accessions (35 wild emmer wheat accession and 26 domesticated wheat accessions) 284 carrying a non-recombinant identical-by-state haplotype in the Ttbtr1-A locus were mapped to the 285 Zavitan genome using Minimap2 (v. 2.24)34 using default short-read settings. Bcftools (v. 1.16)29 286 was used to call SNPs retaining the ones with the following quality settings: -q 20 -Q 20. A filtering 287 step was then added to retain only biallelic SNPs. Homozygous SNPs showing a depth lower than 288 2 and higher than 50 were set to missing. The heterozygous SNPs were set to missing if the depths 289 of both alleles were not greater or equal than three. We then used Beagle (v. 5.4) 35 to phase the 290 SNP matrix to be used as an input for GEV A (v. 1) 36. We added two artificial SNPs: one in the 291 position of the 2 bp deletion causing the Ttbtr1-Aa allele and another in the position of the 292 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 17 retrotransposon insertion causing the Ttbtr1-Ab allele. We then used GEV A to infer the age of the 293 two haplotypes surrounding the two causal variants using default settings and the mutation rate 294 determined in Brachypodium distachyon (6.13 × 10-9). Finally, we doubled the ages estimated by 295 GEV A considering the highly homozygous nature of the wheat genome as it was done for barley 296 by Guo et al.10. We reported the average of the results of 20 independent runs of the haplotype 297 ages determined by the molecular clock model. 298 299 Data availability 300 The Illumina whole -genome sequencing data generated in this study, the Hi -C reads generated 301 from CWI 22960, and the CWI 22960 assembly are available at ENA under BioProject number 302 PRJEB101210. The PacBio HiFi reads are available at ENA under BioProject number 303 PRJEB101209. 304 305 Code availability 306 All the custom python scripts used are available at https://github.com/emilecg/btr_analysis. 307 308

309 We thank Natalia Arango López for the DNA extractions. We thank Yael Lev -Mirom, Assaf 310 Distelfeld, and Curtis Pozniak for critical feedback. This publication is based upon work supported 311 by KAUST award ORFS-CRG12-2024-6474. 312 313 Author Contributions 314 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 18 S.G.K. and E.C.-G. conceived the experiments and wrote the manuscript. E.C.-G assembled the T. 315 turgidum ssp. carthlicum genome, built the phylogeny and analyzed the TtBtr1 alleles. T.W. 316 performed the dating of the Ttbtr1-B loss-of-function event. 317 318 Competing interests 319 The authors declare no competing interests. 320 321 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 19

322 1. Pourkheirandish, M. et al. On the origin of the non -brittle rachis trait of domesticated 323 einkorn wheat. Frontiers in Plant Science 8, 2031 (2018). 324 2. Nave, M. et al. Wheat domestication in light of haplotype analyses of the Brittle rachis 1 325 genes (BTR1-A and BTR1-B). Plant Science 285, 193-199 (2019). 326 3. Pourkheirandish, M. et al. Evolution of the grain dispersal system in barley. Cell 162, 527-327 539 (2015). 328 4. Avni, R. et al. Wild emmer genome architecture and diversity elucidate wheat evolution 329 and domestication. Science 357, 93-97 (2017). 330 5. Civáñ, P. & Brown, T.A. A novel mutation conferring the nonbrittle phenotype of cultivated 331 barley. New Phytologist 214, 468-472 (2017). 332 6. Lev-Mirom, Y . et al. Desert cave grains uncover ancient tetraploid wheat dispersion routes. 333 Nature Plants in press(2026). 334 7. Lev-Mirom, Y . & Distelfeld, A. Where was wheat domesticated? Nature Plants 9, 1201-335 1202 (2023). 336 8. Zhao, X.B. et al. Population genomics unravels the Holocene history of bread wheat and 337 its relatives. Nature Plants 9, 403-419 (2023). 338 9. Zhao, X., Guo, Y . & Lu, F. Reply to: Where was wheat domesticated? Nature Plants 9, 339 1203-1206 (2023). 340 10. Guo, Y . et al. A haplotype-based evolutionary history of barley domestication. Nature 341 (2025). 342 11. Cavalet-Giorsa, E. et al. Origin and evolution of the bread wheat D genome. Nature 633, 343 848-855 (2024). 344 12. He, F. et al. Exome sequencing highlights the role of wild-relative introgression in shaping 345 the adaptive landscape of the wheat genome. Nature Genetics 51, 896-904 (2019). 346 13. Wang, Z.H. et al. Dispersed emergence and protracted domestication of polyploid wheat 347 uncovered by mosaic ancestral haploblock inference. Nature Communications 13, 3891 348 (2022). 349 14. Adhikari, L. et al. Dissecting the population structure, diversity and genetic architecture of 350 disease resistance in wild emmer wheat (Triticum turgidum subsp. dicoccoides). Research 351 Square, https://doi.org/10.21203/rs.3.rs-4909521/v1 (2024). 352 15. Liu, S.C. et al. A telomere-to-telomere genome assembly coupled with multi -omic data 353 provides insights into the evolution of hexaploid bread wheat. Nature Genetics 57, 1008-354 1020 (2025). 355 16. Fairbanks, R.A. & Ross-Ibarra, J. An ancient origin of the naked grains of maize. Proc Natl 356 Acad Sci U S A 122, e2503748122 (2025). 357 17. Wicker, T. et al. Transposable element populations shed light on the evolutionary history 358 of wheat and the complex co -evolution of autonomous and non -autonomous 359 retrotransposons. Advanced Genetics 3, 2100022 (2022). 360 18. Snir, A. et al. The origin of cultivation and proto -weeds, long before neolithic farming. 361 PLOS ONE 10, e0131422 (2015). 362 19. Peleg, Z., Abbo, S. & Gopher, A. When half is more than the whole: Wheat domestication 363 syndrome reconsidered. Evolutionary Applications 15, 2002-2009 (2022). 364 20. Cheng, H.Y ., Concepcion, G.T., Feng, X.W., Zhang, H.W. & Li, H. Haplotype-resolved de 365 novo assembly using phased assembly graphs with hifiasm. Nature Methods 18, 170-175 366 (2021). 367 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint 20 21. Zhou, C., McCarthy, S.A. & Durbin, R. YaHS: yet another Hi -C scaffolding tool. 368 Bioinformatics 39, btac808 (2022). 369 22. Durand, N.C. et al. Juicebox provides a visualization system for Hi -C contact maps with 370 unlimited zoom. Cell Systems 3, 99-101 (2016). 371 23. Pérez-Wohlfeil, E., Diaz-del-Pino, S. & Trelles, O. Ultra-fast genome comparison for large-372 scale genomic experiments. Scientific Reports 9, 10274 (2019). 373 24. Abrouk, M. et al. Fonio millet genome unlocks African orphan crop diversity for 374 agriculture in a changing climate. Nature Communications 11, 4488 (2020). 375 25. Bolger, A.M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina 376 sequence data. Bioinformatics 30, 2114-2120 (2014). 377 26. Kokot, M., Długosz, M. & Deorowicz, S. KMC 3: counting and manipulating k -mer 378 statistics. Bioinformatics 33, 2759-2761 (2017). 379 27. Hagberg, A.A., Schult, D.A. & Swart, P.J. Exploring network structure, dynamics, and 380 function using NetworkX. in Proceedings of the 7th Python in Science Conference 381 (SciPy2008) (eds Varoquaux, G., Vaught, T. & Millman, J.) 11 -15 (Pasadena, CA USA, 382 2008). 383 28. Camacho, C. et al. BLAST+: architecture and applications. BMC Bioinformatics 10, 421 384 (2009). 385 29. Danecek, P. et al. Twelve years of SAMtools and BCFtools. Gigascience 10, giab008 386 (2021). 387 30. Wong, T.K.F. et al. IQ-TREE 3: phylogenomic inference software using complex 388 evolutionary models. EcoEvoRxiv, https://doi.org/10.32942/X2P62N (2025). 389 31. Letunic, I. & Bork, P. Interactive Tree of Life (iTOL) v6: recent updates to the phylogenetic 390 tree display and annotation tool. Nucleic Acids Research 52, 78-82 (2024). 391 32. Sievers, F. et al. Fast, scalable generation of high -quality protein multiple sequence 392 alignments using Clustal Omega. Molecular Systems Biology 7, 539 (2011). 393 33. Li, H. Aligning sequence reads, clone sequences and assembly contigs with BWA -MEM. 394 arXiv preprint arXiv:1303.3997 (2013). 395 34. Li, H. Minimap2: pairwise alignment for nucleotide sequences. Bioinformatics 34, 3094-396 3100 (2018). 397 35. Browning, B.L., Tian, X., Zhou, Y . & Browning, S.R. Fast two-stage phasing of large-scale 398 sequence data. The American Journal of Human Genetics 108, 1880-1890 (2021). 399 36. Albers, P.K. & McVean, G. Dating genomic variants and shared ancestry in population -400 scale sequencing data. PLOS Biology 18, e3000586 (2020). 401 402 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 28, 2026. ; https://doi.org/10.64898/2026.02.26.708176doi: bioRxiv preprint