Duplication of symbiotic Lysin Motif-receptors predates the evolution of nitrogen-fixing nodule symbiosis

Duplication in LysM-receptors predate nodulation 1 corresponding author: Rene Geurts rene.geurts@wur.nl 2 3 Duplication of symbiotic Lysin Motif-receptors predates the evolution of nitrogen-fixing nodule 4 symbiosis 5 Luuk Rutten, Kana Miyata, Yuda Purwana Roswanjaya, Rik Huisman, Fengjiao Bu, Marijke 6 Hartog, Sidney Linders, Robin van Velzen, Arjan van Zeijl, Ton Bisseling, Wouter Kohlen 7 and Rene Geurts* 8 # These authors contributed equally to this work. 9 10 1. Laboratory of Molecular Biology, Department of Plant Sciences, Wageningen University, 11 Droevendaalsesteeg 1, 6708PB Wageningen, The Netherlands. 2. Centre of Technology for 12 Agricultural Production, Agency for the Assessment and Application of Technology (BPPT), 13 Jakarta, Indonesia. 3. Biosystematics Group, Department of Plant Sciences, Wageningen 14 University, 6708 PB, Wageningen, The Netherlands. 15 16 ONE SENTENCE SUMMARY: 17 Four lysin motif receptor kinases controlling rhizobium nodule formation in the non-legume 18 Parasponia evolved after two ancient duplications. 19 20 AUTHOR CONTRIBUTIONS: 21 RG, LR and KM designed the research; LR, KM, YPR, RH, FB, MH, and SL performed research; 22 LR, KM, YPR, RvV, WK and RG analysed data; and LR, KM, TB and RG wrote the manuscript. 23 24 FUNDING INFORMATION: 25 This work was supported by an NWO-VICI grant (865.13.001) to RG, the ENSA project funded by 26 the Bill & Melinda Gates Foundation to the University of Cambridge to RG, NWO-VENI grant 27 Plant Physiology Preview. Published on July 21, 2020, as DOI:10.1104/pp.19.01420


INTRODUCTION
-7-genus we analysed the LYK3 genomic region of two additional Parasponia and three non-168 nodulating species of the closely related genus Trema. This revealed that the duplication of LYK3 169 exon1 is present in all species investigated and occurred twice, where the most distal exon 1 copy 170 was lost in parasponia ( Figure 2A, Figure S2A). The encoded pre-mRNAs both splice into a shared 171 second exon (Figure 2). Each exon1 copy contains a putative transcription and translation start site, 172 which allows for differential expression of the variants ( Figure 2B-C). Genes of the LYK-I clade 173 have a highly conserved intron-exon structure (Zhang et al., 2009). In most cases, the first exon 174 encodes the extracellular domain comprising the signal peptide and three LysM motifs. So, the 175 parasponia PanLYK3 gene encodes two protein variants, named PanLYK3.1 and PanLYK3.2, that 176 differ in their extracellular domain ( Figure S2B). 177 The LYR-IA orthogroup represents the legume LCO receptors MtNFP, LjNFR5 and pea (Pisum 178 sativum) PsSYM10 (Madsen et al., 2003;Arrighi et al., 2006;Buendia et al., 2016;Miyata et al., 179 2016). Previously, we have shown that Parasponia species harbour two genes in this orthogroup, 180 PanNFP1 and PanNFP2 in parasponia, of which the latter is more closely related to 181 MtNFP/LjNFR5 (van Velzen et al., 2018b). PanNFP1 and PanNFP2 originated from an ancient 182 duplication. Phylogenetic reconstruction, including additional nodulating and non-nodulating 183 species, supported the occurrence of NFP-I and NFP-II subclades in the LYR-IA orthogroup and 184 showed that this duplication associates with the origin of the nitrogen-fixing clade (Figure 3; Data 185 set S2). Several Actinorhizal species possess gene copies in both NFP subclades; including Datisca 186 glomerata, Dryas drummondii, and Ceanothus thyrsiflorius. All these species nodulate with 187 diazotrophic Frankia species of taxonomic cluster-II, which possess LCO biosynthesis genes. An 188 NFP-II-type orthologous gene is notably absent in Actinorhizal species that are exclusively 189 nodulated by Frankia species of cluster-I or cluster-III that lack LCO biosynthesis genes; e.g. Alnus 190 glutinosa and Casuarina glauca ( Figure 3) (Pawlowski and Demchenko, 2012;Griesmann et al., 191 2018;Salgado et al., 2018;Van Nguyen et al., 2019). In line with what was reported for the non-192 nodulating Rosales species (van Velzen et al., 2018b), NFP-II-type pseudogenes can be found in the 193 genomes of the non-nodulating Fagales species Castanea mollissima and Quercus fagus. This 194 shows a strict association of the presence of a functional NFP-II-type gene and  nodulation, suggesting that the NFP-II subclade represents LCO receptors that function exclusively 196 in nodulation. 197 198 Parasponia PanNFP1, PanNFP2, PanLYK1 and PanLYK3 can perceive rhizobium LCOs 199 -9-variants) and PanNFP2 as the most likely candidates to encode rhizobium LCO receptors in 201 parasponia. We noted that, in contrast to PanLYK3, PanLYK1 is exclusively expressed in roots and 202 nodule tissue ( Figure 2B), suggesting this gene may also function in a symbiotic context. Therefore, 203 we decided to include this gene in further studies. Finally, we included also PanNFP1, since an 204 earlier study based on RNA interference (RNAi) in transformed parasponia roots showed that this 205 gene functions in nodulation (Op den Camp et al., 2011). To test whether these four parasponia 206 genes can function as rhizobium LCO receptors, we conducted two complementary experiments. 207 First, we introduced parasponia receptor pairs into a lotus Ljnfr1;Ljnfr5 double mutant aiming to 208 determine whether these parasponia receptors can trans-complement for LCO-induced Ca 2+ 209 oscillation. Second, we generated CRISPR-Cas9 knockout mutants in parasponia to study their role 210 in nodulation. 211 We selected lotus for trans-complementation studies as its microbial host Mesorhizobium loti strain 212 R7A can also nodulate parasponia ( Figure  For the trans-complementation constructs, we included the nuclear localized calcium sensor R-217 GECO1.2, allowing visualization of nuclear Ca 2+ oscillations (Zhao et al., 2011). In wild-type lotus 218 roots, Ca 2+ oscillation was most strong in young root hair cells, whereas this response is not 219 recorded in the Ljnfr1-1;Ljnfr5-2 double mutant (Figure S3I,J; movie S1) (Miwa et al., 2006). 220 Analysing the transgenic roots expressing parasponia receptor combinations revealed that nine out 221 of eleven tested combinations elicit Ca 2+ oscillation, although less regular in shape and frequency 222 when compared to the positive control ( Figure 4B; movie S2). Interestingly, the receptor 223 combinations PanLYK1;LjNFR5 and LjNFR1;PanNFP2 did not elicit any Ca 2+ oscillation response, 224 whereas both parasponia receptors are -at least partially-functional as an M. loti LCO receptor 225 when combined with a parasponia counterpart ( Figure 4B). Upon inoculation with M. loti R7A, 226 only nodule-like structures were observed on roots trans-complemented with different parasponia 227 receptor combinations (4 weeks post-inoculation), but not with heterologous receptor pairs (Table  228 S2). We sectioned the largest nodule-like structures, which were present on PanLYK3.2;PanNFP2 229 and PanLYK1;PanNFP1 transformed plants. This showed the absence of intracellular rhizobium 230 infections ( Figure S3K-P). Taken We recently established an efficient Agrobacterium tumefaciens-mediated transformation protocol 237 for parasponia, which allows the generation of CRISPR-Cas9 mutant plantlets in a ~3 month 238 timeframe (van Zeijl et al., 2018;Wardhani et al., 2019). This enabled us to test by mutagenesis 239 whether PanLYK1, PanLYK3, PanNFP1 and PanNFP2 are essential for rhizobium-induced nodule 240 formation. We aimed to generate small deletions of 100-300 bp in the area covering the LysM 241 domains by using two or three single guide RNAs (sgRNAs) that have no potential high identity 242 off-targets. In the case of PanLYK3 the transmembrane domain was targeted in order to mutate both 243 alternative start variants. Additionally, we targeted specifically PanLYK3.1 and PanLYK3.2 by 244 designing specific guides on the first exon. Selected single guides only had off-targets with at least 245 three mismatches or two indels, based on alignments to the parasponia reference genome. Shoots 246 regenerated after A. tumefaciens-mediated co-cultivation were genotyped using PCR and 247 https://plantphysiol.org Downloaded on November 5, 2020. -Published by Copyright (c) 2020 American Society of Plant Biologists. All rights reserved.
-11-subsequent sequence analysis to detect potential mutations at the CRISPR target sites. Only T 0 248 shoots with a >75 bp deletion between the two target sites or edits generating a frameshift were 249 considered for propagation and subsequent further evaluation. At least two independent mutant 250 alleles were generated per gene, with the exception of Panlyk3.1 for which only a single suitable 251 allele could be identified (Data set S3). Putative off-target sites that occur in coding sequence 252 regions were amplified by PCR and subsequently sequenced by sanger sequencing. Subsequently, 253 PanNFP1 was sequenced in PanNFP2 lines, and PanNFP2 in PanNFP1 lines. (Data set S3). No 254 off-target mutations at these locations were identified. The selected tissue culture lines were in vitro 255 propagated and rooted, so they could be used for experimentation. 256 We compared the nodulation phenotype of Panlyk1, Panlyk3, Pannfp1 and Pannfp2 knockout 257 mutants in independent experiments, using empty vector (EV) transformed lines as control ( Figure  258 5; Figure S4). All three independent Pannfp2 mutant lines showed to be unable to form nodules or 259 nodule-like structures (5 weeks post inoculation, wpi) with strain Mesorhizobium plurifarium 260 BOR2, demonstrating the requirement for this gene in the nodulation trait ( Figure 5A). 261 Additionally, we noted a reduced nodulation efficiency of all three independent Pannfp1 mutant 262 lines. This is in line with earlier findings using RNAi to target PanNFP1 in A. rhizogenes-263 transformed parasponia roots (Op den Camp et al., 2011), demonstrating that Pannfp1 controls 264 nodulation efficiency, but is not essential for rhizobium intracellular infection. Previously, we 265 reported that PanNFP1 RNAi-nodules have a strong infection phenotype when inoculated with the 266 Sinorhizobium fredii strain NGR234 (Op den Camp et al., 2011). We did not observe such an 267 infection phenotype in nodules induced by M. plurifarium BOR2 on Pannfp1 knockout mutant 268 plants ( Figure S4). In order to determine whether the Pannfp1 infection phenotype is strain 269 dependent, we nodulated plants, also with S. fredii NGR234. This strain showed to be less optimal 270 under the chosen conditions (agroperlite supplemented with EKM medium and S. fredii 271 NGR234.pHC60 at OD 0.05). In an effort to optimize nodulation efficiency with this strain, we 272 used river sand and scored nodulation 8 weeks post-inoculation. Under these conditions, no 273 difference between Pannfp1 and EV-control was observed. Nodules formed on Pannfp1 mutant 274 plants were infected normally ( Figure S4). 275 Similarly to Pannfp1 mutant plants inoculated with M. plurifarium BOR2, we found a reduced 276 nodulation efficiency in parasponia Panlyk3 knockout mutants, but not in Panlyk3.1 and Panlyk3.2 277 variant specific mutant alleles, nor in Panlyk1 mutants ( Figure 5; Figure S4). To determine whether 278 nodules formed on Panlyk1 and Panlyk3 mutants have an infection phenotype, we analysed thin 279 sections. In contrast to legumes, parasponia doesn't guide rhizobia in infection threads towards the 280 nodule primordia. Instead, rhizobia enter via apoplastic cracks in epidermis and cortex, and only 281 https://plantphysiol.org Downloaded on November 5, 2020. -Published by Copyright (c) 2020 American Society of Plant Biologists. All rights reserved.
-12-form infection threads to penetrate nodule cells. Once inside, infection threads develop into fixation 282 threads, which are wider -having two phyla of bacteria aligned compared to one in infection 283 threads-and possess a thinner cell wall (Lancelle and Torrey, 1984;Lancelle and Torrey, 1985). 284 Panlyk1 mutant nodules showed no defects in infection thread structure or the transition from 285 https://plantphysiol.org Downloaded on November 5, 2020. -Published by Copyright (c) 2020 American Society of Plant Biologists. All rights reserved.  Figure S4). To confirm that the infection phenotype 289 is a result of a full Panlyk3 knockout mutation, we sectioned 28 nodules of the independent 290 knockout line Panlyk3-c3. This revealed similar results; 11 nodules normally infected, 11 contained 291 only infection threads, and 6 nodules with an intermediate phenotype. Next, we determined whether 292 this infection phenotype is controlled specifically by either PanLYK3.1 or PanLYK3.2, which 293 showed not to be the case ( Figure S4). As ~50% of the nodules formed on the parasponia Panlyk3 294 mutant plants displayed a wild-type phenotype, it suggests redundancy in gene functioning. 295 Interestingly, S. fredii NGR234 could not nodulate Panlyk3 mutants, which suggest this strain is 296 fully dependent on PanLYK3 controlled signal transduction ( Figure S4). 297 As parasponia did not experience any gene duplication events in the LYK-Ib clade, PanLYK1 in the 298 LYK-Ia clade is the closest homolog of PanLYK3. In order to investigate whether the PanLYK1 299 gene is functionally redundant with PanLYK3 in cases of M. plurifarium BOR2 inoculation, we 300 generated a Panlyk1;Panlyk3 double mutant. To do so, a binary construct with the two sgRNAs 301 targeting PanLYK1 was used for re-transformation of the Panlyk3 mutant (line a2). We obtained 302 three independent Panlyk1;Panlyk3 mutants (data set S3). M. plurifarium BOR2 inoculation 303 experiments revealed that all Panlyk1;Panlyk3 double mutant lines were unable to form any nodule 304 or nodule-like structure ( Figure 6). To confirm that the nodulation minus phenotype in the 305 Panlyk1;Panlyk3 lines is not due to any off-target mutation, we conducted complementation studies 306 using A. rhizogenes-mediated root transformation. As the putative promoter of PanLYK3 is rather 307 complex due to the occurrence of alternative transcriptional start sites (Figure 2), we used the 308 tumefaciens mediated transformation showed that a 2.75 kbps PanNFP2 upstream region can be 327 used to functionally complement the parasponia Pannfp2 mutant when using a PanNFP2 CRISPR-328 resistant allele (PanNFP2cr). Two independent lines formed 7±7 and 4±1 nodules 5 weeks post 329 inoculation with M. plurifarium BOR2 (Figure 7). However, when we used PanNFP1 driven by the 330   Figure S6B). To test whether ROS bursts can also be triggered by 364 rhizobium LCOs, we used the extracts of M. loti R7A and Rhizobium tropici CIAT899. These two 365 strains can nodulate parasponia but produce structurally different LCOs (López-Lara et al., 1995;366 Folch-Mallol et al., 1996). However, neither triggered a ROS burst in parasponia roots, (  PanPT4, PanVPY, PanD27, PanRAD1 and PanRAM1 ( Figure S7). Also, this suggested that 394 PanNFP1 is expressed higher than PanNFP2 under these conditions ( Figure S7). However, no 395 significant differential regulation of any of the studied LysM-type receptor encoding genes was 396 detected between phosphate starved control roots and mycorrhized root samples ( Figure  alexa488 stained roots showed that besides the level of colonization, also the morphology of the few 408 arbuscules that were formed was affected in Panlyk1;Panlyk3 plants. In wild type plants, many 409 cortical cells were filled with arbuscules that were finely branched and occupied most of the cell. In 410 contrast, the few hyphae that enter cortical cells in the Panlyk1;Panlyk3 mutant were unable to form 411 mature arbuscules, either because the fungus fails to switch to fine branching, or because a limited 412 number of fine branches is made ( Figure 9). As both Panlyk1  We used parasponia as a comparative system to legumes to obtain insight into the evolutionary 424 trajectory of LysM-type rhizobium LCO receptors. By conducting phylogenetic analysis, trans-425 complementation studies in a lotus LCO receptor double mutant, and CRISPR-Cas9 mutagenesis in 426 parasponia, we identified four LysM-type receptors that function in LCO-driven nodulation in a 427 non-legume. Two of these, PanLYK3 and PanNFP2, are putative orthologs to known legume 428 rhizobium LCO receptors LjNFR1/MtLYK3 and LjNFR5/MtNFP, respectively. As the Parasponia 429 and legume lineages diverged early in the nitrogen-fixing clade (>100 MYA), the use of 430 orthologous genes for rhizobium LCO perception supports the hypothesis of a shared evolutionary 431 origin of LCO-driven nodulation. In contrast to legumes, symbiotic LysM-type receptors in 432 Parasponia did not experience recent duplication events. Instead, the Parasponia symbiotic LysM-433 type LCO receptors evolved following two ancient duplications. We hypothesize that the PanNFP1 434 -PanNFP2 duplication associates with the origin of the nitrogen-fixation clade, whereas in case of 435 PanLYK1 and PanLYK3, the duplication occurred prior to the birth of the nitrogen-fixing clade. 436 This makes it most probable that the capability of these receptors to perceive LCOs predates the 437 evolution of the nitrogen-fixing nodulation trait. -23-legume and non-legume NFP-II type proteins that is distinct from NFP-I ( Figure S9). This also 457 supports the hypothesis that NFP1-NFP2 duplicated at the root of the nitrogen-fixing clade. is involved in this process, the so-called entry receptor (Ardourel et al., 1994). Such entry receptors 469 have not yet been fully characterized, but MtLYK3 may carry out such functions, as they control 470 rhizobium infection (Limpens et al., 2003;Smit et al., 2007)  and PanLYK3, which may explain the reported mycorrhization phenotype on PanNFP1 RNAi roots 534 ( Figure S10). Studies presented here using CRISPR-Cas9 knockout mutant lines revealed 535 substantial biological variation in mycorrhization efficiency of parasponia roots, which may have 536 hindered the observation of minor quantitative AM symbiosis phenotypes. To rule out that 537 PanNFP1 and PanNFP2 may function redundantly to control AM symbiosis, we analysed a 538 Pannfp1;Pannfp2 double mutant. Also, these lines showed to be effectively mycorrhized. 539 Therefore, we conclude that our current mutant phenotype analysis does not find support for 540 essential functioning of parasponia PanNFP1 and PanNFP2 in AM symbiosis by. 541 The study presented here provided insight into the evolutionary trajectory of symbiotic LCO LysM-542 type receptors. By using parasponia as a comparative system to legumes, we revealed two ancestral 543 duplications of LysM-type LCO receptors that predate, and coincide with, the evolution of nitrogen-544 fixing nodules. The strict association of genes in the NFP-II clade with LCO-driven nodulation 545 strongly suggests that this gene was co-opted to function specifically in this symbiosis, making 546 NFP2 a target in approaches to engineer LCO-driven nodulation in non-leguminous plants. 547

LysM-type receptor phylogeny reconstructions 550
Orthogroups containing LysM-type receptor kinases of parasponia, generated in a previous study 551 (van Velzen et al., 2018b), were combined and re-aligned into a single alignment using 552 -26-Fabales, Fagales, Cucurbitales and Rosales species were downloaded and local BLAST analysis 558 was conducted using Geneious R8.1.9 (Biomatters Ltd, UK) to search for additional unannotated 559 LYK-I and LYR-IA protein sequences. Pseudogenes were annotated manually based on the closest 560 functional ortholog so that a protein sequence could be deduced. Correct protein sequences were 561 aligned using MAFFTV7.017 and subsequently manually curated. The deduced amino acid 562 sequence was subsequently added to the alignment if the alignment length was at least 70% of the 563 was run at least three times. Trees were rooted to outgroup angiosperm species Amborella 571 trichopoda. UF Bootstrap Branch supports >98 were omitted for visual clarity. Gene names, 572 accession numbers and alignment file of identified homologs can be found in Data set S1 for LYK-I 573 and Data set S2 for LYR-IA, and Table S1 for parasponia. 574

Vector constructs 581
All vectors generated for this study were created using golden gate cloning (Engler et al., 2009). 582 Backbones and binary vectors were derived from the golden gate molecular toolbox (Engler et al., 583 2014). Parasponia LysM-type receptor cDNA clones were sequence synthesized as level 0 modules, 584 including silent mutations in golden gate BsaI or BpiI restriction sites. Golden gate compatible 585 clones of LjNFR1 and LjNFR5 promoters, CDS and terminators were obtained from Arhus 586 University, Denmark. The calcium signalling reporter pLjUBQ1:R-GECO1.2 was published 587 previously (Kelner et al., 2018). The generation and assembly of parasponia CRISPR constructs 588 were done as published previously (van Zeijl et al., 2018). For hairy root transformation, a modified 589 level 2 standard vector carrying spectinomycin instead of kanamycin resistance was created. All 590 https://plantphysiol.org Downloaded on November 5, 2020. -Published by Copyright (c) 2020 American Society of Plant Biologists. All rights reserved.
-27-sgRNAs were expressed using the AtU6 promoter. All Golden Gate binary vectors were verified by 591 restriction digestion and DNA sequencing before transformation. A list of primers and constructs 592 can be found in Table S3 and S4.  593 Genotyping and off-target analysis 594 All sgRNA targets were designed using the Geneious R10 CRISPR design tool, which picks targets 595 on the principles described in Doench et al. (2014). To be selected Guide RNAs must have no 596 potential target sites in the genome with (i) Less than three mismatches or (ii) less than two indels. 597 Known off-target locations in CDS regions were PCR amplified and sequenced. No off-target 598 mutations at these sites were detected. Genotypes and known off-target locations of CRISPR 599 mutants used in this study can be found in Data set S3. Primers used for the creation of sgRNAs and 600 subsequent sequencing of mutants and off-targets are listed in Table S4. 601

Rhizobium LCO isolation 613
To isolate rhizobium LCOs the plasmid pMP604 containing an auto-active NodD protein was 614 introduced in M. loti R7A and R. tropici CIAT899 (Spaink et al., 1989;López-Lara et al., 1995). 615 LCOs were extracted from a 750 ml liquid culture, OD600=0.5, grown at 28°C in minimal medium 616 and 1h shaking. The butanol phase was transferred and subsequently evaporated (water bath 40˚C). 620 Pellet was dissolved in 75ml methanol, tested for Nod-factor activity and stored at -20˚C for later 621 https://plantphysiol.org Downloaded on November 5, 2020. -Published by Copyright (c) 2020 American Society of Plant Biologists. All rights reserved.
-28-use. The concentration of active LCOs was estimated by using LjNIN induction in lotus wild type 622 Gifu roots, 3h post-application. The lowest active dilution was estimated to be ~10 -10 M. 623

Calcium oscillation quantification 638
Calcium spiking experiments were performed on a Leica TCS SP8 HyD confocal microscope 639 equipped with a water lens HC plan-Apochromat CS2 40x/1.0. Transformed root segments 640 expressing R-GECO1.2 were selected and incubated with 500x diluted LCO extract (estimated to 641 represent ~10 -9 M) in nitrate-free ½ Hoagland's medium (Hoagland et al., 1950) on a glass slide 642 with coverslip. Images were taken at 5s intervals for a minimum of 20 minutes per sample using an 643 excitation wavelength of 552 nm and emission spectrum 585-620 nm. It is possible to monitor a 644 large number of nuclei per root sample. However, only epidermal and especially root hairs showed 645 to be responsive. Therefore, total nuclei numbers vary largely between samples. Video recordings 646 of Imaged root samples were exported to ImageJ1.50i (Collins, 2007) Fisher Scientific, Waltham, MA, USA), roots were incubated in 10% (w/v) KOH at 60°C for 3 h. 692 Then roots were washed three times in phosphate-buffered saline (PBS) (150 mM NaCl, 10 mM 693 Na 2 HPO 4 and 1.8 mM KH 2 PO 4 , pH 7.4), and incubated in 0.2 μg ml -1 WGA-Alexafluor 488 in PBS 694 at room temperature for 16 h. For RNA-isolation, parasponia WT plants were grown according to 695 conditions above. RNA was isolated according to protocols published in (Op den Camp et al., 2011;696 van Velzen et al., 2018b). Mock inoculated plants were harvested as control. Three independent 697 biological replicates were taken per sample. Expression was determined using RNA-seq. Reads 698 were mapped using kallisto (Bray et al., 2016). Expression values and differential expression were 699 determined using sleuth (Pimentel et al., 2017). Differentially expressed genes (Benjamini-700 Hochberg multiple testing corrected q-value <= 0.05) 701 qPCR analysis of panNFPi cDNA samples. 702 PanNFPi cDNA samples were generated previously (Op den Camp et al., 2011). qPCR was 703 performed in 10 μl reactions using 2x iQ SYBR Green Super-mix (Bio-Rad, United States). PCR 704 reaction was executed on a CFX Connect optical cycler, according to the manufacturer's protocol 705 (Bio-Rad, United States). Three technical replicates per cDNA sample were used. Data analysis and 706 statistical analysis of biological replicates was performed using CFX Manager 3.0 software (Bio-707 Rad, United States). Gene expression was normalized against Reference genes PanACTIN and 708 PanEF1alpha. Primers can be found in Table S4. 709

ROS assay 710
Parasponia plantlets were grown on rooting medium (van Zeijl et al., 2018) for 4 weeks at 28°C 711 before the treatment. Roots, submerged in water, were cut into approximately 1cm pieces. Each 712 well of a black 96 well flat bottom polystyrene plate (Nunc) was filled with 10 root pieces. 10 713 replicates per line were analysed. After filling the wells, the plate was kept 5 hours in 28 °C. After 714 incubation, the water was replaced with 100 µl of assay solution containing 0.5 µM L-012 715 (FUJIFILM Wako Chemicals), 10 µg/ml Horseradish peroxidase (Sigma), and respective elicitors (;716 CO7 (ELICITYL)  Parasponia plantlets were grown on rooting medium (van Zeijl et al., 2018) for 4 weeks at 28°C 722 before the treatment. About 200mg of roots were cut while submerged in water and collected in a 723 PCR-tube. Root segments were incubated for 5 hours at 28 °C before treatment. Root pieces were 724 treated with water containing 100 μM CO7 (ELICITYL) for 10 min. After incubation, roots were 725 immediately frozen in liquid nitrogen. Samples were homogenized using metal beads. Total root 726 protein was extracted in a buffer containing 50 mM Tris·HCl (pH 7.5), 150 mM KCl, 1 mM EDTA 727 (pH 7.5), 0.1% w/v Triton X-100, 1 mM DTT, complete protease inhibitors (Roche), and phosstop 728  Table S1.