# Source paper archive - Source title: Large-scale forward genetics reveals genetically separable components of plant cell proliferative capacity - Author(s): Verdu-Navarro, F., Espin-Veracruz, A., Weiss, J., Kacprzyk, J., Navarro, P. J., Parras, D., Borraz-Moron, D., Salas, E., Egea-Cortines, M. - Source landing URL: https://www.biorxiv.org/content/10.64898/2026.10.30.762808v1 - PDF URL: https://www.biorxiv.org/content/10.64898/2026.10.30.762808v1.full.pdf - PDF SHA-256: f86b5ced9cc269afda6a5d53aa0c15c86258ba1d52ba996101fb472014b10ca7 - 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/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) # 1 Large-scale forward genetics reveals genetically separable components of 2 plant cell proliferative capacity 3 4 Fuensanta Verdú-Navarro [1,2], Ángela Espín-Veracruz [1], Julia Weiss [1], Joanna Kacprzyk [3], 5 Pedro J. Navarro-Lorente [4], Dolores Parras-Burgos [5], David Borraz-Moron [4], Eduardo 6 Salas-Fernández [4], Marcos Egea-Cortines [1*] 7 1- Universidad Politécnica de Cartagena, member of European University of 8 Technology EUT+, Instituto de Biotecnología Vegetal, 30203 Cartagena, Spain 9 2- Bioprocess Pilot Plant, BIONET, 30320 Fuente Álamo, Spain 10 3-School of Biology and Environmental Science, University College Dublin, D04 11 VIW4, Ireland 12 4- Universidad Politécnica de Cartagena, member of European University of 13 Technology EUT+, División de Sistemas e Ingeniería Electrónica, 30203 Cartagena, 14 Spain 15 5- Universidad Politécnica de Cartagena, member of European University of 16 Technology EUT+, Grupo de Crear y Diseñar en Ingeniería, 30203 Cartagena, Spain 17 18 *Corresponding author: Marcos.egea@upct.es 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 1 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 34 35 36 37 **Graphical Abstract** 38 Overwiew of the study. Image created with Biorender. 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 2 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 54 55 **Summary** 56 Plant cell cultures underpin regeneration, transformation and emerging 57 applications in biotechnology and plant cell agriculture, yet the genetic 58 determinants of proliferative performance remain poorly characterized. Here, we 59 combined large-scale forward genetics with automated image-based 60 phenotyping to quantify 27,648 calli representing 2,880 Arabidopsis thaliana T 61 DNA insertion lines. Of these, 2,456 lines produced calli smaller than the WT 62 population mean, whereas only 190 produced larger calli, revealing a strongly 63 asymmetric phenotypic distribution. Enhanced callus growth was not 64 consistently associated with increased vegetative growth, indicating that these 65 traits can be genetically uncoupled. Validation of selected high-growth mutants 66 in suspension culture identified lines with enhanced biomass accumulation, 67 while serial propagation revealed independent variation in sustained proliferative 68 capacity. Thus, callus growth, whole-plant growth, short-term suspension 69 biomass accumulation and sustained proliferation represent distinct, not 70 necessarily coupled components of proliferative performance. These findings 71 establish a scalable framework for identifying genetic determinants relevant to 72 the development of improved plant cell production lines. 73 74 75 76 77 78 79 3 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 80 81 **Highlights** 82 - Established a high-throughput platform for quantitative phenotyping of plant 83 callus growth. 84 - Screened 2,880 Arabidopsis T-DNA lines, revealing extensive genetic variation 85 in callus proliferative capacity. 86 - Identified mutants with enhanced biomass accumulation or sustained 87 proliferative capacity in suspension culture. 88 - Established a framework for the genetic improvement of plant cell production 89 platforms. 90 91 **Keywords:** Callus proliferation; Forward genetics; High-throughput phenotyping; 92 Computer vision; T-DNA insertion mutants; Enhanced biomass accumulation; Plant cell 93 suspension culture; Sustained proliferative capacity; Serial subculture; Plant cell 94 bioprocessing 95 96 97 98 99 100 101 102 103 104 105 106 107 4 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 108 109 110 **Introduction** 111 112 Plant cell and callus cultures underpin plant transformation, regeneration and an 113 expanding range of industrial applications, including the production of recombinant 114 proteins, high-value metabolites and emerging cellular agriculture products [1–3] . Despite 115 their growing industrial relevance, efforts to improve plant cell cultures have focused 116 predominantly on medium optimization, elicitation strategies and bioprocess 117 engineering. Unlike microbial and mammalian production hosts, plant cell cultures have 118 not been subjected to systematic large-scale genetic improvement for enhanced 119 proliferative capacity. 120 121 Genetic studies of plant cell cultures have focused predominantly on regeneration 122 competence rather than industrial performance. Landmark discoveries identifying 123 regulators such as PLT, WOX and WIND1 established the molecular basis of callus 124 induction, totipotency and de novo shoot regeneration, enabling regeneration of 125 previously recalcitrant species [4,5] . Likewise, manipulation of individual cell-cycle 126 regulators demonstrated that callus proliferation can be genetically modified. For 127 example, overexpression of _CYCD3_ enhances callus induction and growth in 128 Arabidopsis [6,7] ., while expression of a rice _R2 CDK_ -activating kinase is sufficient to 129 induce callus formation in tobacco explants in the absence of cytokinin [8] . These studies 130 established that proliferation is genetically tuneable, but they examined individual 131 developmental regulators rather than the genetic architecture of proliferative capacity 132 itself. Consequently, proliferative capacity has not been systematically explored as a 133 quantitative industrial trait for improving transformation efficiency, biomass 134 accumulation and the performance of plant cell production platforms. 135 136 A major barrier to treating proliferative capacity as a quantitative trait has been the lack 137 of scalable phenotyping protocols. Unlike traits such as flowering time or plant 138 architecture, callus growth has traditionally required manual image analysis or 139 destructive measurements, limiting the number of genotypes that could be evaluated. 140 Recent advances in automated imaging and AI-based image segmentation now enable 141 objective, high-throughput quantification of callus size in large image datasets [9–11] . 5 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 142 Combined with the availability of genome-wide Arabidopsis T-DNA insertion 143 collections [12], these technologies now make forward genetic dissection of proliferative 144 capacity feasible at population scale. 145 146 Here, we combined large-scale forward genetics in _Arabidopsis thaliana_ with 147 standardized callus induction and deep-learning-based object detection and image 148 segmentation to systematically identify genetic determinants of callus growth. By 149 screening 2,880 homozygous T-DNA insertion lines, we asked whether growth-related 150 phenotypes identified in a high-throughput image-based screen could reveal candidate 151 loci for improving plant cell cultures. We further evaluated representative high-growth 152 mutants to determine whether increased callus size translated into enhanced biomass 153 accumulation and sustained proliferation in suspension culture. Together, our results 154 establish a scalable framework for identifying genetic determinants of plant cell 155 proliferative performance and candidate targets for the development of improved plant 156 cell production lines. 157 158 **Results** 159 **Testing cell cycle related genes as a proof of concept** 160 Previous studies have demonstrated that mutations in _Arabidopsis thaliana_ _KIP_ 161 _RELATED PROTEIN_ _KRP_ ( _INHIBITOR OF CYCLIN KINASE- ICK_ ) genes, which 162 encode cyclin-dependent kinase inhibitors, result in enhanced cell proliferation and 163 increased organ size. Double and triple _KRP_ mutants display enlarged leaves [13] . 164 Furthermore quintuple ICK/KRP knockouts exhibit increased cell proliferation, larger 165 organs and seeds, and enhanced callus growth [14,15] . 166 Based on this evidence, we hypothesized that T-DNA insertion lines affecting cell 167 cycle-related genes would display detectable alterations in callus size under our 168 induction conditions. Because the aim was to develop a scalable forward screen, calli 169 were induced directly from seeds [16,17], avoiding the need to generate leaf explants for 170 each mutant line and biological replicate. The callus induction medium contained 171 Murashige Skoog salts (MS), B5 vitamins 2,4 Dichloro phenoxyacetic acid and 6 172 Benzyl Aminopurine. To test this as a proof of concept, we selected 13 homozygous T 173 DNA insertion lines targeting cell cycle genes including _AtKRP1_, _AtKRP5_, _AtKRP7_, 174 and _AtSIAMESE_ (Table S1). Mutant and wild-type seeds were randomly distributed 6 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 175 across plates (Figure S1), and germination rate and callus growth were evaluated by 176 visual inspection. 177 Several mutants displayed altered germination and readily detectable differences in 178 callus size relative to Col-0 (Figure S1). Mutations in _SIM_ **,** _AtKRP1_ **,** and one _AtKRP7_ 179 allele produced smaller calli, whereas a second independent _AtKRP7_ allele exhibited 180 increased callus size. Previous studies reported enhanced callus development only in 181 higher-order _KRP_ mutants [15], whereas our results indicate that individual alleles can 182 generate detectable phenotypes under the screening conditions employed here. These 183 observations validated the experimental design and indicated that a genome-scale 184 forward genetic screen would be feasible despite the functional redundancy expected 185 among cell-cycle regulators. 186 187 **A scalable imaging and AI pipeline enables high-throughput callus phenotyping** 188 Reliable quantitative comparison of callus size across thousands of genotypes requires 189 highly standardized image acquisition. Variations in illumination, camera position or 190 plate orientation introduce systematic errors that can obscure relatively small 191 phenotypic differences. To minimize these sources of variation, we developed a custom 192 built dark chamber housing the imaging system (Figure 1 a,b). Plates were illuminated 193 using uniform white backlight illumination and imaged with an OAK-1 RGB camera 194 positioned perpendicular to the plate under fixed geometric conditions, enabling 195 consistent pixel-based area quantification. 196 Captured images were transmitted to an NVIDIA Jetson TX2 processing unit, where 197 preprocessing, image acquisition control, and inference were executed through a custom 198 graphical interface (see Materials and Methods). 199 To enable quantitative comparison across thousands of independent genotypes, key 200 experimental variables were standardized, including callus induction conditions, seed 201 distribution, plate orientation during image acquisition and automated genotype 202 assignment. 7 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 203 Prior to large-scale screening, a deep-learning model was trained to distinguish 204 proliferating calli from non-germinated seeds. A ground-truth dataset comprising 1,230 205 manually annotated calli, together with associated non-germinated-seed annotations 206 from 26 Petri dish images, was used for model training and evaluation. The resulting 207 model accurately detected calli and non-germinated seeds across all plates (Figure 1c– 208 e). Following object detection, individual calli were segmented using HSV color-space 209 thresholding and contour extraction, enabling pixel-based area quantification. 210 Detected calli were automatically assigned to their corresponding genotypes based on 211 their position within the predefined plate sectors. Callus area measurements were 212 exported for downstream statistical analysis, enabling standardized high-throughput 213 quantification of callus size across the mutant population. 214 215 **High-throughput screening of a T-DNA insertion population for callus** 216 **proliferation** 217 The screening platform was applied to 2,880 homozygous Arabidopsis T-DNA insertion 218 lines distributed across 192 Petri plates (12 × 12 cm) (Table S2), each containing 15 219 mutant lines and one plate-matched Col-0 control. 220 Of the 2,880 lines screened, 723 showed reduced or null germination (Table S3). After 221 excluding 234 lines with complete absence of germination, a total of 27,648 calli were 222 analyzed 28 days after plating, comprising 1,728 WT and 25,920 mutant calli (Figure 2 223 a). Across the full dataset, the distribution of callus areas in the mutagenized population 224 was shifted toward smaller values relative to WT ( _t_ -test, _P_ < 2.2 × 10 � ¹ � ; Figure 2a). 225 The screen identified mutants displaying reduced germination as well as both reduced 226 and increased callus size (Figure 2b–d). 227 Lines with fewer than four germinated seedlings were retained for descriptive analysis 228 but excluded from quantitative size comparisons. Thirteen plates in which the WT 229 control displayed abnormal growth were identified as batch outliers and excluded from 230 subsequent analyses. 231 Using the mean WT callus area (6,430 pixels) as a population reference, 2,456 mutant 232 lines produced smaller calli than WT, whereas only 190 produced larger calli (Tables S4 233 and S5), indicating a strong asymmetry in the distribution of growth phenotypes. To 8 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 234 identify candidate lines while accounting for plate-to-plate variation, each mutant was 235 subsequently compared with the Col-0 control grown on the same plate using two-sided 236 _t_ -tests. At a nominal significance threshold of _P_ < 0.05, 988 lines showed reduced callus 237 growth and 19 showed increased callus growth relative to their plate-matched WT 238 controls (Table S6). 239 After Benjamini–Hochberg correction for multiple testing, one high-growth line 240 remained significant at FDR < 0.05. Because the primary screen was designed to 241 identify candidates for subsequent experimental validation, high-growth lines were 242 prioritized on the basis of their plate-matched phenotype and confirmation of the 243 corresponding T-DNA insertion. Fifteen high-growth lines were ultimately advanced 244 for further characterization (Table S7; Figure 3). 245 246 **Candidate loci span diverse biological functions** 247 Candidate loci identified in the screen encoded proteins associated with chromatin 248 organization, transcriptional regulation, hormone signalling, metabolism, transport and 249 cell wall-related processes. However, Gene Set Enrichment Analysis did not identify 250 any category that remained significant after false discovery rate correction (Table S8,9; 251 Figure S2). These results suggest that callus growth is influenced by diverse biological 252 processes rather than a single dominant genetic pathway. 253 254 **Genetic determinants of reduced callus proliferation** 255 To further resolve genetic contributors to callus proliferation, we examined the mutants 256 displaying the strongest negative growth phenotypes. Among the 50 lines with the 257 smallest calli (Table S4), 45 also exhibited reduced germination rates (1–3 out of 9 258 seeds germinated). The most severe phenotypes corresponded to mutations affecting 259 genes encoding a MA3 domain-containing protein (AT3G48390) involved in 260 translational regulation, the DNAJ-containing protein THERMOSENSITIVE MALE 261 STERILE 1 (AT3G08970), the protein kinase AtPERK9 (AT1G68690), the RNA 262 polymerase II elongation factor MDA7.18 (AT5G56120), and a heavy metal 263 transport/detoxification protein (AT1G56210). These genes encode proteins involved in 264 translation, protein folding, intracellular transport and transcriptional regulation 265 266 In contrast, a subset of mutants showing strongly reduced callus proliferation did not 267 display impaired germination, suggesting more specific effects on post-embryonic 9 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 268 proliferative capacity. These included mutations in the copper transporter COPT2 269 (AT3G46900), an unannotated SNARE-domain protein (AT4G22850), an oligomeric 270 Golgi complex subunit-like protein (AT4G24840), a rhomboid-related intramembrane 271 serine protease (AT5G25640), and a Mo25 family protein (AT2G03410). These loci 272 represent candidates affecting proliferative growth after germination. 273 274 **Genetic determinants of enhanced callus proliferation** 275 276 To further characterize the strongest candidates, we examined the subset of 15 mutant 277 lines displaying significantly increased growth relative to their plate-matched WT 278 controls (Figure 3). T-DNA insertions were distributed across distinct loci, in 1000 279 upstream, 5’ UTRs, exons and introns (Table S7). 280 281 The genes identified were associated with four broad and partially overlapping 282 biological processes: mitochondrial and organelle function (NDB1, AT1G79120, 283 AT5G46460), growth and developmental regulation (EAR1, RHF2a, BBX21, WRKY4, 284 HSP90 co-chaperone p23-2), stress and nutrient signalling (ANAC053, EAR1, 285 WRKY4, PHL3/UNE16, AT4G28260), and cellular and metabolic functions (TLP2, 286 LCAT-PLA, FMOGS-OX7). Several of these genes remain poorly characterized despite 287 having predicted molecular functions, including AT1G17030, AT1G79120, 288 AT4G28260 and AT5G46460. 289 290 We next examined the molecular consequences of the T-DNA insertions by measuring 291 transcript abundance directly in callus tissue. The effects of the insertions varied 292 substantially among the selected lines. Four lines _,_ AT5G22000 _(_ AtRHF2a _),_ 293 AT5G22090 (EAR1), AT4G28260 (acyl-UDP-N-acetylglucosamine O-acyltransferase) 294 and AT3G10500 (ANAC053), corresponded to knockout alleles. Three alleles 295 corresponding to WRKY4 (AT1G13960), a calcium-dependent phospholipase 296 (AT4G19860) and a co-chaperone HSP90p23-2 (AT3G03773) were significantly down 297 regulated. 298 299 In contrast, six genes were significantly upregulated: _PHL3/UNE16_ (AT4G13640), 300 _NDB1(AT4G28220_, _FMOGS-OX7_ (AT1G12160), Pentatricopeptide repeat (PPR) 301 superfamily protein (AT5G46460) AT1G79120 and AT1G17030, with expression 10 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 302 ranging from 1.38- to 4.27-fold that of the wild type. Two lines, corresponding to TLP2 303 (AT2G18280) and BBX21 (AT1G75540), transcript abundance did not differ 304 significantly from the wild type despite their increased callus-growth phenotype. 305 Therefore, the phenotype of these lines cannot be directly associated with altered 306 expression of the corresponding genes. The possibility that the phenotype results from 307 additional mutations present in these lines, or from effects of the T-DNA insertion not 308 reflected in transcript abundance, cannot be excluded. Consequently, these genes were 309 not considered further in the functional interpretation of the growth phenotype. 310 311 Among the high-growth candidates, _NDB1_ (AT4G28220), encoding an external 312 mitochondrial NAD(P)H dehydrogenase, has previously been implicated in vegetative 313 biomass accumulation, with loss-of-function mutants showing reduced biomass [18] . The 314 SALK_087720C line identified in our screen carries a T-DNA insertion in the 5 ′ region 315 and showed a 4.27-fold increase in _NDB1_ expression in callus (p < 0.001), together with 316 increased callus growth. Thus, the increased-expression phenotype observed here is 317 consistent with the previously reported reduction in biomass following loss of _NDB1_ 318 function and identifies _NDB1_ as a candidate for callus growth enhancement through 319 enhanced expression. 320 321 AT4G13640 encodes PHR1-LIKE 3, a GARP-family transcription factor associated 322 with phosphate-responsive regulation. In the SALK_113627C line, the T-DNA insertion 323 is located approximately 1 kb upstream of the coding sequence. Expression analysis 324 showed a 1.35-fold increase in AT4G13640 transcript abundance relative to wild type 325 (p = 0.003), indicating that the enhanced callus-growth phenotype in this line is 326 associated with increased expression of PHL3. 327 Several of the remaining genes have previously described functions that provide 328 biological context for their increased callus-growth phenotypes. EAR1 was identified as 329 a knockout in our screen, consistent with previous observations that loss of EAR1 330 function can promote growth under drought conditions [19] . RHF2a was also identified as 331 a knockout and has a redundant function with RHF1a during Arabidopsis 332 gametogenesis [20] . The p23-2 line showed strongly reduced expression (0.27-fold 333 relative to WT); loss of Atp23-2 has previously been associated with reduced root 334 growth and impaired root meristem maintenance [21] . ANAC053, which was identified 11 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 335 here as a knockout, encodes a membrane-associated NAC transcription factor 336 implicated in ROS production during drought-induced leaf senescence [22] . 337 338 Taken together, these candidates illustrate the diverse genetic basis of callus 339 proliferation and provide multiple entry points for the genetic improvement of plant cell 340 cultures. 341 342 **Enhanced Callus Growth Does Not Predict Vegetative Growth** 343 To determine whether mutations promoting callus proliferation also enhanced whole 344 plant growth, selected high-growth mutants were grown in soil under standard 345 conditions and vegetative growth was evaluated 28 days after sowing. 346 High-growth mutants displayed diverse vegetative phenotypes. Some lines exhibited 347 larger rosettes than WT (Figure 4a–g), whereas others were smaller (Figure 4h–k) or 348 showed no obvious differences from WT (Figure 4l–o). 349 These observations demonstrate that enhanced callus proliferation is not necessarily 350 accompanied by increased vegetative growth, supporting the conclusion that in vitro 351 proliferative capacity can be genetically uncoupled from whole-plant biomass 352 accumulation. 353 354 355 **High-Growth Callus Mutants Reveal Distinct Suspension Culture Phenotypes** 356 To determine whether increased callus size on solid medium predicted improved 357 suspension-culture performance, calli from five high-growth mutants were expanded on 358 Petri dishes and equal amounts of biomass were used to inoculate liquid cultures. 359 Growth trajectories were analysed using linear mixed-effects models fitted to log 360 transformed sedimented biomass measurements, allowing line-specific slopes to be 361 interpreted as specific biomass accumulation rates (Figure 5). 362 Two independent lines exhibited enhanced biomass accumulation relative to Col-0. 363 After correction for initial inoculum volume, SALK_036374C and SALK_086651C 364 accumulated approximately 1.27- and 2.47-fold the biomass of WT, respectively. These 365 lines represent two different molecular outcomes of T-DNA insertion identified above: 366 SALK_036374C is a knockout of AT4G28260, whereas SALK_086651C shows 12 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 367 increased expression of FMOGS-OX7 (AT1G12160). Thus, improved suspension 368 culture growth was associated with both loss of gene function and increased gene 369 expression. 370 During liquid-culture validation, maintenance of the cultures through successive 371 passages revealed substantial differences in their capacity to sustain proliferation. Col-0 372 cultures showed a marked decline in proliferative capacity by the fifth passage and 373 could no longer be maintained after the sixth passage. Several mutant lines lost growth 374 capacity earlier, including SALK_108025C after three passages and SALK_086651C 375 and SALK_113627C after four passages. In contrast, SALK_023787C and 376 SALK_036374C remained vigorously proliferative through seven successive passages, 377 when the experiment was terminated. These observations reveal an additional source of 378 genetic variation among the selected lines and suggest that short-term biomass 379 accumulation and the capacity to sustain proliferation across successive passages 380 represent distinct, not necessarily coupled, components of suspension-culture 381 performance. 382 Together, these results demonstrate that image-based phenotypic screening on solid 383 medium can identify genetic variation affecting multiple traits relevant to suspension 384 culture, including biomass accumulation and sustained proliferative capacity, providing 385 a practical strategy for the genetic development of superior plant cell production lines. 386 387 388 **Discussion** 389 Callus proliferation is central to regeneration, transformation, and plant cell–based 390 manufacturing, yet optimization strategies have traditionally focused on media 391 composition rather than genetic improvement. Here, we establish a scalable framework 392 that treats callus growth as a quantitative genetic trait by integrating forward genetics 393 with automated imaging and AI-assisted image analysis. This approach enabled the 394 analysis of 27,648 calli and the identification of candidate loci modulating proliferative 395 capacity, including rare mutations capable of enhancing growth. 396 13 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 397 Callus formation and proliferation are frequently regarded as predominantly hormone 398 driven processes [23] . Although extensive work has focused on optimizing exogenous 399 growth regulator combinations, genome-scale identification of genetic determinants 400 controlling proliferative capacity has remained largely unexplored. By isolating callus 401 directly from seeds under standardized conditions, we systematically quantified growth 402 variation across 2,880 mutant lines and demonstrated substantial genetic control of 403 proliferative capacity. The resulting phenotypic distribution was markedly asymmetric: 404 while numerous mutations reduced callus size, only a small subset enhanced 405 proliferation. Such a skewed distribution suggests that proliferative growth operates 406 near a constrained optimum, where most perturbations disrupt functions required for 407 growth, whereas rare mutations relieve regulatory constraints. Growth-enhancing alleles 408 may therefore reflect the release of regulatory constraints rather than disruption of core 409 biosynthetic processes, underscoring their potential value for biotechnological 410 optimization. Together, the predominance of moderate-effect alleles and the limited 411 number of loci remaining significant after multiple-testing correction suggest that callus 412 growth is governed by a highly distributed genetic architecture rather than by a small 413 number of major-effect regulators. The functional diversity of candidate loci, together 414 with the absence of significantly enriched categories after multiple-testing correction, 415 further supports a model in which callus growth is influenced by multiple biological 416 processes rather than a single dominant pathway. 417 418 One initial working hypothesis was that enhanced callus growth would predominantly 419 arise from mutations affecting core cell-cycle regulators. However, such genes were 420 notably absent from the high-growth category. Instead, several loci associated with 421 enhanced callus growth lacked functional annotation. This finding suggests that the 422 genetic basis of enhanced proliferative capacity remains incompletely characterized and 423 that forward genetic screening can uncover regulators that would not readily be 424 predicted from current functional annotations. 425 426 A key outcome of this study is that enhanced callus growth does not universally 427 translate into increased plant growth. Several mutants displaying larger calli exhibited 14 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 428 reduced or unchanged vegetative growth under soil conditions, indicating that growth in 429 dedifferentiated, hormone-rich environments can be genetically uncoupled from organ 430 growth. These observations suggest that callus growth and whole-plant growth are at 431 least partially independent traits, governed by overlapping but non-identical regulatory 432 mechanisms. 433 434 The primary objective of this screen was to identify genetic determinants capable of 435 improving growth in suspension systems relevant to bioreactor-based production. To 436 assess the translational value of the primary screen, representative high-growth mutants 437 were selected for liquid-culture validation. This validation was not intended as an 438 exhaustive characterization of all candidate loci, but rather as a test of whether enhanced 439 growth on solid medium could predict improved performance in liquid culture. Both 440 SALK_086651C and SALK_036374C accumulated more biomass than Col-0 during 441 the culture period. After correction for the initial inoculum volume, SALK_036374C 442 and SALK_086651C exhibited approximately 1.27-fold and 2.47-fold greater net 443 biomass gains than the wild type, respectively. Although these differences were 444 measured over single seven-day culture cycles, they demonstrate that genetic variation 445 identified through solid-medium screening can translate into altered biomass 446 accumulation in suspension culture. 447 Even modest differences in growth rate can become consequential during repeated 448 cultivation cycles as growth in liquid culture is exponential. Indeed, a 1.7-fold increase 449 in daily growth rate has previously been associated with a reduction in bioreactor 450 cultivation time from 21 to 10 days [24] . Viewed in this context, the 1.27-fold and 2.47 451 fold increases in PCV gain observed for SALK_036374C and SALK_086651C, 452 respectively, highlight the potential value of genetically improving proliferative 453 capacity in plant cell cultures. However, validation under bioreactor conditions will be 454 required to determine the resulting effects on process productivity. 455 The diversity of candidate loci identified in the screen provides multiple routes for the 456 genetic improvement of plant cell proliferation. While high-growth mutants identify loci 457 whose disruption or altered expression can enhance proliferation, low-growth alleles 458 may reveal positive regulators of proliferative capacity and therefore potential targets 15 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 459 for overexpression or gene-dosage strategies. Beyond targeted engineering, these 460 candidate loci may facilitate germplasm screening for naturally occurring allelic or 461 expression variants associated with enhanced in vitro proliferation. The scale of the 462 dataset therefore provides a resource for systematic prioritization of candidate loci and 463 future combinatorial approaches to improving plant cell production systems. 464 465 Beyond short-term biomass accumulation, serial propagation revealed substantial 466 genetic variation in the ability of cultures to sustain proliferation over successive 467 passages. This trait is particularly relevant to long-term and continuous bioprocessing, 468 where maintenance of a viable and proliferatively competent cell population is required 469 over extended cultivation periods. Long-established plant cell lines such as tobacco BY 470 2, Arabidopsis T87 and our _Nicotiana benthamiana_ LAB-3.1416JA line illustrate that 471 plant cells can acquire or maintain proliferative capacity over very prolonged culture 472 periods [24–27] . However, exceptionally long-term proliferative capacity can only be 473 established retrospectively, whereas sustained proliferation across serial passages 474 provides a prospectively measurable phenotype for identifying candidate lines with 475 extended proliferative potential. Importantly, our results indicate that this trait is not 476 necessarily coupled to short-term biomass accumulation: SALK_086651C showed 477 strongly enhanced biomass accumulation but lost proliferative capacity after four 478 passages, whereas SALK_023787C and SALK_036374C retained vigorous 479 proliferation through seven passages. Sustained proliferative capacity therefore emerges 480 as a distinct and potentially important target for the genetic development of plant cell 481 lines intended for extended and continuous production strategies, including perfusion 482 based cultivation. 483 **Concluding Remarks** 484 Our study establishes callus proliferative capacity as a genetically tractable trait and 485 provides a scalable framework for identifying genetic determinants relevant to plant cell 486 biotechnology. Importantly, callus growth, whole-plant growth, short-term biomass 487 accumulation, and sustained proliferation are not necessarily coupled, indicating that 488 plant cell performance comprises distinct components that can potentially be optimized 489 independently. Integrating genetic screening with automated phenotyping and 16 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 490 suspension-culture validation therefore provides a route toward the development of 491 plant cell lines tailored for in vitro production. Future work should determine whether 492 these traits can be transferred across species, combined with high product yields, and 493 maintained under bioreactor conditions. 494 # 495 Experimental procedures 496 **Plant material and in vitro culture conditions** 497 498 Arabidopsis thaliana T-DNA insertion lines from the SALK collection (NASC N27941; 499 ABRC CS27941) were obtained from the Nottingham Arabidopsis Stock Centre 500 (NASC) (Alonso et al., 2003). Columbia-0 (Col-0) was used as the wild-type control. 501 For the proof-of-concept experiment, T-DNA lines affecting cell cycle-related genes 502 were selected from the collection (Table S1). A total of 2,880 mutant lines were 503 screened in the large-scale experiment (Table S2). 504 505 Seeds were surface-sterilized in 20% (v/v) commercial bleach containing 37 g L � ¹ 506 sodium hypochlorite and supplemented with 250 µL L � ¹ Tween-20 for 20 min, rinsed 507 four times with sterile distilled water, and stratified at 4 °C for 24 h. 508 509 For the proof-of-concept experiment, seeds were germinated directly on callus induction 510 medium containing MS salts with B5 vitamins, 2,4-D and 6-BAP (Sello et al., 2017). 511 Each plate was divided into 16 sectors containing 13 mutant lines and three Col-0 512 controls, with nine seeds sown per sector. Three independent plates were prepared 513 (Figure S1). 514 515 For the large-scale screen, seeds were sown directly on callus induction medium in 12 × 516 12 cm square Petri dishes containing 70 mL of medium comprising MS salts with B5 517 vitamins, 30 g L � ¹ sucrose, 10 mg L � ¹ picloram, 0.2 mg L � ¹ kinetin, and 3.5 g L � ¹ 518 Gelrite. Plates were divided into 16 equal sectors, with nine seeds of a single genotype 519 sown in each sector. Each plate contained 15 mutant lines and one Col-0 control, with 17 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 520 Col-0 consistently positioned in the upper-left sector. Plates were cultured at 23 °C 521 under continuous light for 28 days before imaging. 522 523 **Image preprocessing and plate segmentation** 524 Due to the high density of individuals per plate (144 seeds), two recurrent challenges 525 were observed: (i) overlapping calli and (ii) growth extending beyond predefined grid 526 sectors. To standardize image geometry and enable subsequent detection and spatial 527 assignment of individual calli, a preprocessing pipeline based on classical computer 528 vision methods was implemented prior to deep learning–based analysis. 529 Preprocessing steps included grayscale conversion, adaptive inverse thresholding, 530 contour detection, and morphological filtering. The largest detected contour 531 corresponding to the Petri dish was identified, and the image was cropped and 532 geometrically normalized to ensure consistent scaling across plates. Each plate was then 533 subdivided into 16 predefined spatial sectors corresponding to genotype positions. 534 This procedure enabled robust extraction of the plate region and established a 535 standardized spatial framework for subsequent assignment of detected calli to genotype 536 positions. 537 538 **Deep learning–based detection of calli and seeds** 539 540 Object detection was performed using the YOLO11 architecture [28] trained to identify 541 two classes: (i) calli and (ii) non-germinated seeds. A ground-truth dataset comprising 542 1,230 manually annotated calli along with associated non-germinated seed annotations 543 from 26 Petri dish images was generated using the Labelme open-source tool (Wada et 544 al., 2021). To avoid data leakage, the dataset was partitioned at the image level into 545 training (18 images, n=851 calli), validation (5 images, n=254 calli), and testing (3 546 images, n=125 calli) subsets. 547 Multiple YOLO11 configurations were evaluated. All models demonstrated high 548 generalization performance, with precision and recall values ≥ 0.96 and mAP 50 values up 549 to 0.994 on the test set. The YOLO11n configuration was selected for deployment due 550 to its strong performance and computational efficiency. 551 18 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 552 Following bounding box detection, individual callus regions of interest were segmented 553 using HSV color-space thresholding and contour extraction to generate precise pixel 554 level masks. The area was computed directly from the number of pixels contained 555 within each mask. Because acquisition geometry and imaging parameters were fixed 556 across experiments, pixel area was used as a consistent proxy for callus size. 557 558 **Growth quantification and data aggregation** 559 Detected objects were assigned to genotypes based on their spatial position within each 560 of the 16 predefined sectors. Non-germinated seeds were recorded for germination rate 561 estimation but excluded from callus area calculations. 562 For each genotype on each plate, pixel areas of all detected calli (up to nine per 563 genotype) were exported automatically to structured .xlsx files. Mean callus area for 564 each genotype was calculated for downstream statistical analysis. 565 566 **Gene set enrichment analysis** 567 We used the Webgestalt interface to identify enriched pathways using default settings 29 568 . 569 570 **Vegetative growth assessment** 571 For vegetative growth assessment, seedlings were grown vertically on plates at 23 °C 572 under continuous light for 14 days and then transplanted to soil 573 (perlite:vermiculite:substrate, 1:1:0.5). Plants were acclimated under a covered tray for 574 5–7 days and then grown in a 12 h light/12 h dark photoperiod (23 °C day/18 °C night). 575 Plants were watered twice per week with ATM solution [30] . Images were acquired at 28 576 days after stratification, corresponding to 14 days post-transplantation. 577 578 **Liquid culture analysis** 579 Calli from selected high-growth lines were propagated on solid medium, mechanically 580 disrupted, and used to inoculate liquid cultures. For each line, 4 g fresh callus biomass 581 was used to inoculate 16 mL fresh medium in 100 mL Erlenmeyer flasks. Cultures were 582 maintained at 23 °C under light conditions with shaking at 120 rpm. Biomass 583 accumulation was monitored using a custom 3D-printed sedimentation device (HD 19 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 584 PLA) designed in SolidWorks and produced from an .STL file (Supplementary Material 585 S1). 586 587 Packed cell volume (PCV) was measured daily by allowing cultures to sediment for 10 588 min. Cultures were subsequently maintained by serial subculture, transferring growing 589 cultures into fresh medium under the same conditions. Successive subcultures were 590 recorded as passages (P1, P2, etc.). When a culture lost proliferative capacity, the 591 corresponding line was re-established in liquid culture from callus propagated on solid 592 medium when additional biological replicates were required. 593 594 Growth trajectories were analysed using linear mixed-effects models fitted to log 595 transformed sedimentation measurements: log(PCV) ~ Day × Line, with replicate as a 596 random effect using R 4.3.2. Estimated slope differences between each mutant and Col 597 0 were obtained using emtrends (emmeans), and contrasts were computed using a 598 treatment-versus-control comparison. Overall growth was calculated by subtracting the 599 initial volume from the final volume for each sample. 600 601 **Gene expression analysis of T-DNA mutant lines** 602 Leaf samples were collected from soil-grown plants of each line, callus samples were 603 processed directly from in vitro cultures. Total RNA was extracted using the 604 NucleoSpin ® RNA kit according to the manufacturer’s protocol. For the conversion to 605 cDNA, we used the Maxima™ H Minus cDNA Synthesis Master Mix commercial kit 606 containing DNases from ThermoFisher ©. Transcript presence or absence for candidate 607 genes was assessed by endpoint RT-PCR in leaves. For selected lines showing 608 expression, transcript levels were quantified by RT-qPCR using SYBR Green chemistry 609 and normalized to SAND (AT2G28390) as the reference gene using callus tissue. 610 Relative expression was calculated using the 2^- ΔΔ Ct method [31] . We used at least three 611 biological replicas and two technical per extraction. Primer sequences are provided in 612 Supplementary Table S10. 613 614 **Acknowledgements** 615 This work is funded by Proyecto PID2021-127933OB-C21; financed by 616 MICIU/AEI/10.13039/501100011033 and FEDER to M.E.-C. and J.W.; and Catedra 20 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 617 BIONET in Biotechnology, Grant DIN2020-011559, funded by 618 MICIU/AEI/10.13039/501100011033 and by “European Union 619 NextGenerationEU/PRTR”, to F.V.-N. We would like to thank Dr. Qikun Liu for 620 comments on the manuscript. 621 622 **Author contributions** 623 Conceptualization, F.V-N, J.W, J.K, P.J.N-L, D.P.B and M.E-C.; Methodology, F.V-N, 624 J.W, J.K, P.J.N-L, D.P.B and M.E-C.; Investigation, F.V-N, A.E-V, J.W, J.K, P.J.N-L, 625 D.P.B, D.B-M, E.S-F, and M.E-C.; Writing – Original Draft, F.V-N, J.W and M.E-C; 626 Writing – Review and Editing, F.V-N, A.E-V, J.W, J.K, P.J.N-L, D.P.B, D.B-M, E.S-F, 627 and M.E-C Funding Acquisition, F.V-N, J.W, and M.E-C.; Resources, J.W, J.K, P.J.N 628 L, D.P.B, and M.E-C; Supervision, J.W, J.K, P.J.N-L, and M.E-C. 629 630 **Declarations of interest** 631 The authors declare no competing interests. Fuensanta Verdú is employed by BIONET, 632 but the company was not involved in the study design, data acquisition, data analysis, 633 manuscript preparation, or publication decisions. 634 635 **Use of AI** 636 During the preparation of this work, the authors used ChatGPT (OpenAI) for English 637 language editing and to assist in the development and refinement of R scripts used for 638 data analysis and figure preparation. The authors reviewed and verified all AI-assisted 639 outputs and take full responsibility for the content of the manuscript. 640 All figures were prepared by the authors, using Inkspace and Biorender 641 642 **Figure legends** 643 **Figure 1. Automated image acquisition and quantitative phenotyping of callus** 644 **growth.** (a) Schematic representation of the image acquisition device. (b) Photograph of 645 the imaging apparatus. (c) Image of a screening plate. (d) Detection of individual calli 646 using the trained YOLO11 model. Red labels indicate automatically assigned sector and 647 callus identifiers. (e) Segmentation and area quantification of individual calli. Blue 648 labels indicate the calculated area of each callus in pixels. 21 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 649 650 **Figure 2. Large-scale screening reveals extensive genetic variation in callus** 651 **growth.** (a) Distribution of callus area (pixels) in 1,728 wild-type (Col-0) and 25,920 652 mutant calli. The distributions differed significantly (two-sample t-test, _P_ < 2.2 × 653 10 � ¹ � ). Representative screening plates illustrating (b) lines with low germination rate, 654 (c) a mutant line producing smaller calli than Col-0, and (d) a mutant line producing 655 larger calli than Col-0. In each plate, the Col-0 control occupies the upper-left sector. 656 657 658 **Figure 3.** **Enhanced callus growth in 15 T-DNA insertion lines.** Boxplots show callus 659 area (pixels) for 15 mutant lines with significantly larger calli than their respective 660 plate-matched Col-0 controls (nominal P < 0.05). Each panel displays the SALK 661 identifier of the mutant and its corresponding Col-0 control. The affected genes are (a) 662 AT1G17030, (b) AT5G22000, (c) AT3G03773, (d) AT4G28220, (e) AT1G13960, (f) 663 AT1G12160, (g) AT2G18280, (h) AT4G28260, (i) AT1G79120, (j) AT1G75540, (k) 664 AT3G10500, (l) AT5G22090, (m) AT4G19860, (n) AT5G46460, and (o) AT4G13640. 665 Panels are ordered to correspond to the same mutant lines shown in Figure 4. 666 667 **Figure 4.** **Enhanced Callus Growth Does Not Predict Vegetative Growth.** 668 Representative vegetative phenotypes of the 15 high-callus-growth mutant lines after 669 two weeks of growth on plates followed by two weeks in soil. Col-0 is shown on the left 670 of each panel and the corresponding mutant on the right. Images show one 671 representative plant from ten plants analysed per genotype. (a) SALK_091444C 672 (AT1G17030), (b) SALK_023787C (AT5G22000), (c) SALK_126538C (AT3G03773), 673 (d) SALK_087720C (AT4G28220), (e) SALK_121735C (AT1G13960), (f) 674 SALK_086651C (AT1G12160), (g) SALK_058100C (AT2G18280), (h) 675 SALK_036374C (AT4G28260), (i) SALK_017387C (AT1G79120), (j) 676 SALK_113836C (AT1G75540), (k) SALK_009578C (AT3G10500), (l) 677 SALK_108025C (AT5G22090), (m) SALK_061713C (AT4G19860), (n) 678 SALK_033891C (AT5G46460), (o) SALK_113627C (AT4G13640). 679 680 22 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 681 **Figure 5. High-growth callus mutants reveal distinct suspension-culture** 682 **phenotypes.** (a) Growth trajectories of Col-0 and five high-callus-growth mutant lines 683 in liquid culture: SALK_023787C (AT5G22000), SALK_036374C (AT4G28260), 684 SALK_086651C (AT1G12160), SALK_108025C (AT5G22090), and SALK_113627C 685 (AT4G13640). Growth trajectories were analysed using a linear mixed-effects model 686 fitted to log-transformed packed cell volume (PCV) measurements. (b) Estimated 687 growth rates for Col-0 and the five mutant lines, with confidence intervals derived from 688 the model. Representative suspension cultures of Col-0 (left) and SALK_086651C 689 (right) at (c) day 1 and (d) day 7 after inoculation. Images show representative flasks 690 from eight Col-0 and four SALK_086651C biological replicates. 691 692 693 **Supplemental information** 694 695 **Figure S1. Proof-of-concept analysis of selected cell-cycle mutants affecting callus** 696 **development.** (a) Plate layout used for the proof-of-concept experiment. Each plate 697 contained nine seeds per sector, comprising 13 mutant lines and three Col-0 controls, 698 for a total of 144 seeds per plate. Thirteen alleles representing eleven genes were 699 analysed (Table S1). Col-0 sectors are indicated in blue, mutants producing smaller calli 700 in orange, and the mutant producing larger calli in green. (b) T-DNA insertion positions 701 in AT1G49620 ( _AtKRP7_ ). 702 703 **Figure S2. Gene set enrichment analysis of lines with reduced and enhanced callus** 704 **growth.** 705 **Supplementary material S1** . STL file to print the Erlenmeyer holder on a standard 3D 706 printer-Available upon request 707 **Table S1.** Arabidopsis T-DNA lines affecting genes related to cell cycle used for proof 708 of concept. 709 **Table S2** . Arabidopsis T-DNA lines tested from NASC collection. 710 **Table S3.** Arabidopsis T-DNA lines showing low or null germination rates. 711 **Table S4.** Arabidopsis T-DNA lines showing decreased growth compared to the WT 712 average. 23 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 713 **Table S5.** Arabidopsis T-DNA lines showing increased growth compared to the WT 714 average. 715 **Table S6** . Arabidopsis T-DNA lines showing significant smaller calli area than their 716 respective plate control (WT). 717 **Table S7** . Predicted T-DNA insertion sites, primers, and expression analysis of high 718 callus-growth mutant lines. 719 **Table S8.** Results of Gene Enrichment Analysis for the genes affected by T-DNA 720 insertions causing decreased growth compared to the WT average. 721 **Table S9.** Results of Gene Enrichment Analysis for the genes affected by T-DNA 722 insertions causing increased growth compared to the WT average. 723 724 **Data availability statement** 725 The data that supports the findings of this study are available in supporting information 726 of this article. 727 728 Declaration of generative AI and AI-assisted technologies in the manuscript 729 preparation process 730 During the preparation of this work, the author(s) used CHATGPT 5.1 for English 731 polishing and improve R-scripts for figures. The author(s) reviewed and edited the 732 output as needed and take full responsibility for the content of the published 733 article. 734 735 736 **Bibliography** 737 1. Fischer, R., and Buyel, J.F. (2020). Molecular farming – The slope of enlightenment. 738 Biotechnol. Adv. _40_, 107519. https://doi.org/10.1016/j.biotechadv.2020.107519. 739 2. Rademacher, T., Sack, M., Blessing, D., Fischer, R., Holland, T., and Buyel, J. (2019). 740 Plant cell packs: a scalable platform for recombinant protein production and metabolic 741 engineering. Plant Biotechnol. J. _17_, 1560–1566. https://doi.org/10.1111/pbi.13081. 742 3. Schillberg, S., Raven, N., Spiegel, H., Rasche, S., and Buntru, M. (2019). Critical Analysis 743 of the Commercial Potential of Plants for the Production of Recombinant Proteins. Front. 744 Plant Sci. _10_ . https://doi.org/10.3389/fpls.2019.00720. 745 4. Ikeuchi, M., Favero, D.S., Sakamoto, Y., Iwase, A., Coleman, D., Rymen, B., and 746 Sugimoto, K. (2019). Molecular Mechanisms of Plant Regeneration. Annu. Rev. Plant Biol. 747 _70_, 377–406. https://doi.org/10.1146/annurev-arplant-050718-100434. 24 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 748 5. Maher, M.F., Nasti, R.A., Vollbrecht, M., Starker, C.G., Clark, M.D., and Voytas, D.F. 749 (2020). Plant gene editing through de novo induction of meristems. Nat. Biotechnol. _38_, 750 84–89. https://doi.org/10.1038/s41587-019-0337-2. 751 6. Cockcroft, C.E., den Boer, B.G., Healy, J.M., and Murray, J.A. (2000). Cyclin D control of 752 growth rate in plants. Nature _405_, 575–579. 753 7. Riou-Khamlichi, C., Huntley, R., Jacqmard, A., and Murray, J.A. (1999). Cytokinin 754 activation of Arabidopsis cell division through a D-type cyclin. Science _283_, 1541–1544. 755 8. Yamaguchi, M., Kato, H., Yoshida, S., Yamamura, S., Uchimiya, H., and Umeda, M. 756 (2003). Control of in vitro organogenesis by cyclin-dependent kinase activities in plants. 757 Proc. Natl. Acad. Sci. _100_, 8019–8023. https://doi.org/10.1073/pnas.1332637100. 758 9. Furbank, R.T., and Tester, M. (2011). Phenomics - technologies to relieve the phenotyping 759 bottleneck. Trends Plant Sci. _16_, 635–644. https://doi.org/10.1016/J.Tplants.2011.09.005. 760 10. Navarro, P.J., Salas, E., Rosique, F., Ruszczak, B., and Egea-Cortines, M. (2025). 761 Conditional generative adversarial network for segmentation and image reconstruction of 762 grape berries in variable illumination conditions. Expert Syst. Appl. _287_, 128076. 763 https://doi.org/10.1016/j.eswa.2025.128076. 764 11. Perez-Sanz, F., Navarro, P.J., and Egea-Cortines, M. (2017). Plant phenomics: an overview 765 of image acquisition technologies and image data analysis algorithms. GigaScience _6_, 766 gix092. https://doi.org/10.1093/gigascience/gix092. 767 12. Alonso, J.M., Stepanova, A.N., Leisse, T.J., Kim, C.J., Chen, H., Shinn, P., Stevenson, 768 D.K., Zimmerman, J., Barajas, P., Cheuk, R., et al. (2003). Genome-wide insertional 769 mutagenesis of Arabidopsis thaliana. Science _301_, 653–657. 770 https://doi.org/10.1126/science.1086391. 771 13. Sizani, B.L., Kalve, S., Markakis, M.N., Domagalska, M.A., Stelmaszewska, J., 772 AbdElgawad, H., Zhao, X., De Veylder, L., De Vos, D., Broeckhove, J., et al. (2019). 773 Multiple mechanisms explain how reduced expression increases leaf size of Arabidopsis 774 thaliana. New Phytol. _221_, 1345–1358. https://doi.org/10.1111/nph.15458. 775 14. Cheng, Y., Cao, L., Wang, S., Li, Y., Shi, X., Liu, H., Li, L., Zhang, Z., Fowke, L.C., 776 Wang, H., et al. (2013). Downregulation of multiple CDK inhibitor / genes upregulates the 777 E2F pathway and increases cell proliferation, and organ and seed sizes in Arabidopsis. Plant 778 J. _75_, 642–655. https://doi.org/10.1111/tpj.12228. 779 15. Cheng, Y., Liu, H., Cao, L., Wang, S., Li, Y., Zhang, Y., Jiang, W., Zhou, Y., and Wang, 780 H. (2015). Down-regulation of multiple CDK inhibitor ICK/KRP genes promotes cell 781 proliferation, callus induction and plant regeneration in Arabidopsis. Front. Plant Sci. _6_, 782 825. 783 16. Huang, B.C., and Yeoman, M.M. (1984). Callus proliferation and morphogenesis in tissue 784 cultures of _Arabidopsis thaliana_ L. Plant Sci. Lett. _33_, 353–363. 785 https://doi.org/10.1016/0304-4211(84)90027-0. 786 17. von Recklinghausen, I.R., Iwanowska, A., Kieft, H., Mordhorst, A.P., Schel, J.H., and van 787 Lammeren, A.A. (2000). Structure and development of somatic embryos formed 788 inArabidopsis thaliana pt mutant callus cultures derived from seedlings. Protoplasma _211_, 789 217–224. 25 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 790 18. Wallström, S.V., Florez-Sarasa, I., Araújo, W.L., Aidemark, M., Fernández-Fernández, M., 791 Fernie, A.R., Ribas-Carbó, M., and Rasmusson, A.G. (2014). Suppression of the External 792 Mitochondrial NADPH Dehydrogenase, NDB1, in _Arabidopsis thaliana_ Affects Central 793 Metabolism and Vegetative Growth. Mol. Plant _7_, 356–368. 794 https://doi.org/10.1093/mp/sst115. 795 19. Wang, K., He, J., Zhao, Y., Wu, T., Zhou, X., Ding, Y., Kong, L., Wang, X., Wang, Y., Li, 796 J., et al. (2018). EAR1 Negatively Regulates ABA Signaling by Enhancing 2C Protein 797 Phosphatase Activity. Plant Cell _30_, 815–834. https://doi.org/10.1105/tpc.17.00875. 798 20. Liu, J., Zhang, Y., Qin, G., Tsuge, T., Sakaguchi, N., Luo, G., Sun, K., Shi, D., Aki, S., 799 Zheng, N., et al. (2008). Targeted Degradation of the Cyclin-Dependent Kinase Inhibitor 800 ICK4/KRP6 by RING-Type E3 Ligases Is Essential for Mitotic Cell Cycle Progression 801 during Arabidopsis Gametogenesis. Plant Cell _20_, 1538–1554. 802 https://doi.org/10.1105/tpc.108.059741. 803 21. D’Alessandro, S., Golin, S., Hardtke, C.S., Lo Schiavo, F., and Zottini, M. (2015). The co804 chaperone p23 controls root development through the modulation of auxin distribution in 805 the Arabidopsis root meristem. J. Exp. Bot. _66_, 5113–5122. 806 https://doi.org/10.1093/jxb/erv330. 807 22. Lee, S., Seo, P.J., Lee, H.-J., and Park, C.-M. (2012). A NAC transcription factor NTL4 808 promotes reactive oxygen species production during drought-induced leaf senescence in 809 Arabidopsis. Plant J. Cell Mol. Biol. _70_, 831–844. https://doi.org/10.1111/j.1365810 313X.2012.04932.x. 811 23. Efferth, T. (2019). Biotechnology Applications of Plant Callus Cultures. Engineering _5_, 50– 812 59. https://doi.org/10.1016/j.eng.2018.11.006. 813 24. Verdú-Navarro, F., Moreno-Cid, J.A., Weiss, J., and Egea-Cortines, M. (2025). Cascade 814 Oxygen Control Enhances Growth of Nicotiana benthamiana Cell Cultures in Stirred-Tank 815 Bioreactors. Plants _14_, 2879. 816 25. Koizumi, N., Yamada, H., Kiba, T., Nakamichi, N., Yamashino, T., and Mizuno, T. (2004). 817 Characterization of Arabidopsis T87 cultured cells with special reference to cytokinin 818 responses. Plant Cell Physiol. _45_, S224–S224. 819 26. Kwiatkowska, A., Zebrowski, J., Oklejewicz, B., Czarnik, J., Halibart-Puzio, J., and Wnuk, 820 M. (2014). The age-dependent epigenetic and physiological changes in an _Arabidopsis_ T87 821 cell suspension culture during long-term cultivation. Biochem. Biophys. Res. Commun. 822 _447_, 285–291. https://doi.org/10.1016/j.bbrc.2014.03.141. 823 27. Nagata, T., Nemoto, Y., and Hasezawa, S. (1992). Tobacco BY-2 Cell Line as the “HeLa” 824 Cell in the Cell Biology of Higher Plants. In International Review of Cytology, K. W. Jeon 825 and M. Friedlander, eds. (Academic Press), pp. 1–30. https://doi.org/10.1016/S0074826 7696(08)62452-3. 827 28. Khanam, R., and Hussain, M. (2024). YOLOv11: An Overview of the Key Architectural 828 Enhancements. Preprint at arXiv, https://doi.org/10.48550/arXiv.2410.17725 829 https://doi.org/10.48550/arXiv.2410.17725. 830 29. Elizarraras, J.M., Liao, Y., Shi, Z., Zhu, Q., Pico, A.R., and Zhang, B. (2024). WebGestalt 831 2024: faster gene set analysis and new support for metabolomics and multi-omics. Nucleic 832 Acids Res. _52_, W415–W421. https://doi.org/10.1093/nar/gkae456. 26 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.762808; this version posted November 3, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.762808) (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-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 833 30. Haughn, G.W., and Somerville, C. (1986). Sulfonylurea-resistant mutants of Arabidopsis 834 thaliana. Mol. Gen. Genet. MGG _204_, 430–434. https://doi.org/10.1007/BF00331020. 835 31. Livak, K.J., and Schmittgen, T.D. (2001). Analysis of Relative Gene Expression Data 836 Using Real-Time Quantitative PCR and the 2− ΔΔ CT Method. Methods _25_, 402–408. 837 https://doi.org/10.1006/meth.2001.1262. 838 839 27 (c) (d) (e) (a) (b) (c) (d) (a) (b) (c) (d) (e) (f) (g) (i) (j) (k) (l) (m) (n) (o) (a) (d) (g) (j) (b) (c) (e) (f) (h) (i) (k) (l) (m) (n) (0) |Col1|(b) Growth -rate differences<br>SALK 086651C<br>_<br>SALK 036374C<br>_<br>Mutant<br>SALK 113627C<br>_<br>SALK 023787C<br>_<br>SALK 108025C<br>_<br>-0 .02 0 .00 0 .02 0 .04<br>Δgrowth rate vsCol0 (perday)<br>(d)| |---|---| |1.2<br>1.3<br>1.4<br>1.5<br>1.6<br>1.7<br>2<br>4<br>6<br>Day after inoculation<br>Sedimented measure (cm)<br>Line<br>Col0<br>SALK_023787C<br>SALK_036374C<br>SALK_086651C<br>SALK_108025C<br>SALK_113627C||