Multiple quantitative trait loci contribute tolerance to bacterial canker incited by Pseudomonas syringae pv. actinidiae in kiwifruit (Actinidia chinensis)

Pseudomonas syringae pv. actinidiae (Psa) Biovar 3, a virulent, canker-inducing pathogen is an economic threat to the kiwifruit (Actinidia spp.) industry worldwide. The commercially grown diploid (2x) A. chinensis var. chinensis is more susceptible to Psa than tetraploid and hexaploid kiwifruit. However information on the genetic loci modulating Psa resistance in kiwifruit is not available. Here we report mapping of quantitative trait loci (QTLs) regulating tolerance to Psa in a diploid kiwifruit population, derived from a cross between an elite Psa-susceptible ‘Hort16A’ and a tolerant male breeding parent P1. Using high-density genetic maps and intensive phenotyping, we identified a single QTL for Psa tolerance on Linkage Group (LG) 27 of ‘Hort16A’ revealing 16-19% phenotypic variance and candidate alleles for susceptibility and tolerance at this loci. In addition, six minor QTLs were identified in P1 on distinct LGs, exerting 4-9% variance. Complete tolerance in the F1 population is attained by additive effects from ‘Hort16A’ and P1 QTLs providing evidence that divergent genetic pathways fend-off virulent Psa strain. Two different bioassays further identified new QTLs for tissue-specific responses to Psa. Transcriptome analysis of Psa-tolerant and susceptible genotypes in field revealed hallmarks of basal defense and provided candidate RNA-biomarkers for screening Psa tolerance.


8
Abstract: 1 9 Pseudomonas syringae pv. actinidiae (Psa) Biovar 3, a virulent, canker-inducing pathogen is an 2 0 economic threat to the kiwifruit (Actinidia spp.) industry worldwide. The commercially grown 2 1 diploid (2x) A. chinensis var. chinensis is more susceptible to Psa than tetraploid and hexaploid 2 2 kiwifruit. However information on the genetic loci modulating Psa resistance in kiwifruit is not 2 3 available. Here we report mapping of quantitative trait loci (QTLs) regulating tolerance to Psa in 2 4 a diploid kiwifruit population, derived from a cross between an elite Psa-susceptible 'Hort16A' 2 5 and a tolerant male breeding parent P1. Using high-density genetic maps and intensive 2 6 phenotyping, we identified a single QTL for Psa tolerance on Linkage Group (LG) 27 of 2 7 'Hort16A' revealing 16-19% phenotypic variance and candidate alleles for susceptibility and 2 8 tolerance at this loci. In addition, six minor QTLs were identified in P1 on distinct LGs, exerting 2 9

Introduction 3 8
Pseudomonas syringae is a hemi-biotrophic bacterial complex 1 that can infect a range of plant 3 9 species. It comprises pathovars which cause similar symptoms on their host plants and several 4 0 pathovars can lead to severe crop loss. P. syringae pv. actinidiae (Psa) infects several species of 4 1 Actinidia (kiwifruit) 2,3 and virulent Psa strains induce a range of symptoms on the main stem of 4 2 the vine, foliage, floral buds and fruits 4 . Psa pathovar's strains can be grouped into five biovars 4 3 based on their genetic and biological characteristics 4,5 . Strains of biovar 3, previously called 4 4 Psa-V (referred to here as Psa), are currently the most aggressive and were responsible for infections and symptoms in Actinidia species emerged from Japan, China, Korea and Italy from 5 2 1984 to 1994 2,3,13,14,16,17,19 . The symptoms include cankers on trunk and leaders, cane death and 5 3 stem collapse, discharge of red and milky exudates (ooze) from cankers, canes and abaxial leaf 5 4 surfaces, tip browning, angular leaf necrosis (sometimes with chlorotic halos), shoot and leaf 5 5 wilt, bud browning and flower blight. Strains of Psa infect Actinidia species with varying degrees 5 6 of virulence, indicating a classical host-pathogen evolutionary relationship 5,9,[20][21][22][23][24] . 5 7 Screening of thousands of Actinidia genotypes from 24 taxa in the breeding program at The New 5 8 Zealand Institute for Plant & Food Research Limited (PFR) for tolerance to natural and artificial 5 9 Psa infections 25,26 revealed that diploid (2x) A. chinensis var. chinensis are more susceptible to 6 0 Psa infection than tetraploid (4x) A. chinensis var. chinensis, which in turn are more susceptible 6 1 than diploid and hexaploid (6x) A. chinensis var. deliciosa [25][26][27] . Many species outside the A. 6 2 chinensis complex are more tolerant to Psa than A. chinensis and the germplasm holds diverse 6 3 genetic potential for Psa tolerance 28 . Information on the genetic markers and molecular 6 4 mechanisms associated with Psa tolerance and resistance in the commercial cultivars producing 6 5 taxas including A. chinensis, A. deliciosa and A. arguta is however limited. In this study we 6 6 provide the first detailed view of the genetic loci modulating Psa tolerance and tissue-specific 6 7 response in diploid A. chinensis, utilizing an intensively phenotyped population of seedlings 6 8 developed from a cross between Psa-susceptible 'Hort16A' and a tolerant breeding parent (P1), 6 9 as our experimental material for quantitative trait locus (QTL) analysis. 7 0

1
Intensive phenotyping targets diverse developmental stages and environmental conditions 7 2 Initially, a pilot population of 53 genotypes from 'Hort16A' × P1 were replicated 3 times and 7 3 phenotyped following natural field infection with Psa. The response to Psa infection in the 7 4 expanded population was measured on 236 genotypes of 'Hort16A'×P1 population, which were 7 5 clonally replicated ~30 times. Phenotyping of the population was performed under field 7 6 conditions following natural infection as well as using two bioassays (scheme for phenotyping is 7 7 laid out in Supplementary Fig.1a). Multiple phenotypes were recorded in field ( Fig. 1a-e) to 7 8 develop a combined score referred to as Psa_score_Field (Fig. 2a). The mean clonal repeatability 7 9 for this score was 0.65, while the repeatability of clonal means at 0.8. For the stab assay 26 , 8 0 various tissue-specific phenotypic responses were recorded, including Stem_necrosis, 8 1 Leaf_spots, Ooze, Stem_collapse, Tip_death and Wilt ( Fig. 1f-k, Fig. 2b), with repeatability of 8 2 clonal means for these scores as 0.60, 0.766, 0.64, 0.79, 0.78 and 0.71 respectively. A 8 3 Psa_score_Stab was also calculated (Fig. 2b) from all phenotypes assessed in the Stab assay (see 8 4 Experimental Procedures). In the flood bioassay, adapted from 29 , overall health was scored at 8 5 weekly intervals post-inoculation (Flood Assay/FA_Week1 to FA_Week5) (Fig. 1l). The 8 6 frequency distribution of phenotypes and Psa_scores revealed that most exhibited non-Normal 8 7 genomes. The maps for 'Hort16A' and P1 encompassed a total genetic distance of 3,499 cM and 1 0 1 3,875 cM, respectively, with an average density of 1 marker/ 2cM for both parents. All predicted 1 0 2 29 LGs were constructed for 'Hort16A'; however some were fragmented in P1 (LGs 3, 16, 19, 1 0 3 23, 25, 27). 1 0 4 QTL mapping from field phenotype scores confirmed oligogenic nature of Psa field 1 0 5 tolerance 1 0 6 A QTL for control of field tolerance to Psa, Psa_score_Field, was identified in 'Hort16A' on the 1 0 7 upper arm of LG27 ( Fig. 3a) using multiple models for QTL discovery. At a LOD score of 7.02 1 0 8 (Fig. 3a), the location of the LG27 QTL on the Red5 genome (version 1.69.0) 33 is between ~3.4 1 0 9 to 4.6 Mbp. The LG27 QTL was also identified for Psa_score-Field, in 'Hort16A' from the pilot 1 1 0 trial ( bp allele u which is associated with susceptibility. 1 1 7 Using interval mapping and KW analysis, six QTLs were identified in P1, for Psa_score_Field 1 1 8 indicating Psa tolerance is multigenic in P1. A single QTL with LOD score above 3 was located 1 1 9 on the upper arm of LG22 (Fig. 3b), while three additional QTLs on LGs 3.1, 15 and 24 1 2 0 ( Supplementary Fig. 4), as well as two KW QTLs on LG14 (S14_5310060, K value > 9, P < 1 2 1 0.0001) and LG28 (S28_1476180, K value > 7, P < 0.0001). From these, the effect of favorable 1 2 2 grandparent alleles from P1 on field tolerance was verified from at least 3 QTLs i.e., by analysis 1 2 3 of an SSR marker designed in the region underlying the LG22 QTL (SSRLG22_8032664) (Fig  1  2  4 3g and Supplementary Table1), a SNP marker E6P3 designed in a putative cell wall protein 1 2 5 encoding gene Acc15766 within the LG14 QTL ( Supplementary Fig. 5 Since the QTL identified in 'Hort16A' has the highest effect, we predicted that this locus could 1 3 3 be linked to susceptibility observed in diploid kiwifruit breeding parents. For this purpose, we 1 3 4 performed validation of the LG27 QTL in another field grown A. chinensis population, derived 1 3 5 from two Psa-susceptible parents'Hort22D' and P2, using SSRLG27_439F5R5 located ~3 kb 1 3 6 distant from SSRLG27_439F4R4. Data revealed that the 430/432 bp allele linked in cis with the 1 3 7 previously described allele v was associated with tolerance in this population and was present in 1 3 8 both parents, while the 440 bp allele z (size 440 bp) was linked to susceptibility and present in 1 3 9 'Hort22D' (Supplementary Fig. 6b The pilot field population was phenotyped monthly for symptoms from natural Psa infection  4  1  5  between 2013 and 2015 and the data used to develop the phenotypic scoring for the expanded  4  1  6 population, for which phenotyping was monthly from February 2017 to September 2018. Traits 4 1 7 scored included cane death, ooze, shoot death and tip death (Fig. 1). Presence/absence of leaf 4 1 8 spots was not recorded, as scores in the pilot study exhibited high between-plant variability. A 4 1 9 cumulative Psa score (Psa_score_Field) was calculated, with removals due to oozing and >50% 4 2 0 cane death scoring twice as highly as tip death and shoot death. Least squares mean (LS Mean) 4 2 1 was calculated for each individual genotype, based on this score. 4 2 2 The bioassays were performed in controlled environments, with the stab bioassay 26  genotypes were phenotyped using the stab bioassay, with 35 batches phenotyped across three 4 2 7 years. Details are in Supplemental Methods S1. The flood bioassay 29 was performed by flooding 4 2 8 six biological replicates of each genotype with Psa, that had been grown on tissue-culture media 4 2 9 in an aseptic growth medium in a tub for 4 to 6 weeks. Details are provided in Supplemental 4 3 0 Methods S1. 4 3 1 Bacterial inoculations for assessment of growth curve in resistant vs. susceptible plants 4 3 2 Assessment of the growth curve for Psa in 'Hort16A' and P1 was performed using multiple 4 3 3 biological replicates in the greenhouse, as described for the stab test bioassay. Young potted 4 3 4 kiwifruit plants were inoculated with Psa, on 8 to 10 biological replicates of each genotype in 4 3 5 February, 2018. Further details are provided in Supplemental Methods S1. 4 3 6 Genotyping, genetic maps and QTL mapping 4 3 7 DNA was extracted from freeze-dried leaves using the Cetyl trimethylammonium bromide 4 3 8 (CTAB) method 84 . GBS libraries were prepared for 53 individuals from the pilot population and 4 3 9 236 individuals from the expanded population, as well as the two parents, using the method 4 4 0 described in detail by 85 , modified from the standard GBS protocol 30 . The individual and pooled 4 4 1 libraries were checked for quality with a Fragment Analyser (Advanced Analytical) and pooled 4 4 2 libraries with satisfactory QC were dried down and dispatched to the Australian Genome 4 4 3 Research Facility (AGRF) for single-end sequencing on an Illumina® HiSeq™ platform. The 4 4 4 sequencing reads were de-multiplexed based on GBS library preparation barcodes using the ea-4 4 5 utils.1.1.2-537 package and those reads starting with the approved barcode immediately followed 4 4 6 by the remnant of the BamHI cut site sequence were retained for further analysis. Variant calling 4 4 7 and genotyping was performed using TASSEL v3.0 and 5.0 and ~60,000 and 80,000 SNP calls 4 4 8 were generated for the individuals in the two populations, respectively. SNP calling was 4 4 9 performed using an early version of the represented 70% coverage of the expanded population. In our data sets (Supplementary Data 1), 4 5 5 Red5 markers begin with S, whereas markers generated from 'Hongyang' begin with HY, 4 5 6 followed by the number of the linkage group and the position of the marker on the respective 4 5 7 physical genome (for example S1_10661198 or HY10_1385907). The SNP data were 4 5 8 subsequently filtered to obtain 9,875 and 9,327 SNP markers polymorphic between 'Hort16A' 4 5 9 and P1, respectively (3,364 for P1 in pilot study). JoinMap v 5.0 86 was used to develop genetic 4 6 0 linkage maps for both the parents, at a LOD score between 15 and 22. QTL mapping was 4 6 1 performed using the rQTL package 87 and MapQTL5 software 88 . Multiple QTL models, 4 6 2 including Maximum likelihood (EM), Haley-Knott regression, Non-parametric, multiple 4 6 3 imputation and Kruskal-Wallis analysis (KW) were employed for single QTL scans. 4 6 4

RNA-seq and RT-qRT-PCR 4 6 5
Total RNA was extracted from healthy young leaves, at the sixth to ninth position from the 4 6 6 apical leaf, from genotypes in the field that were segregated into three groups based on 4 6 7 susceptibility to Psa. The first two groups were of two-year-old plants that were defined 4 6 8 respectively as: 1) tolerant/medium tolerant (Psa-TMT), consisting of three relatively Psa-4 6 9 tolerant genotypes, including P1, and 2) susceptible (Psa-Sus), comprising three fully Psa-4 7 0 susceptible genotypes including 'Hort16A' based on observations of 8 to 11 biological 4 7 1 replicates. The third group (Psa-FT) exhibited field tolerance to Psa over four years. RNA 4 7 2 extraction was from samples snap frozen with liquid nitrogen, using the Spectrum Total Plant 4 7 3 RNA kit (Sigma-Aldridge, Auckland, New Zealand) and QC was performed with the Fragment 4 7 4 Analyzer to select RNA with RNA Integrity Number (RIN) of 7.1-8.2. Samples from three 4 7 5 biological replicates were pooled. Library preparation at the Australian Genome Research 4 7 6 Facility used the TruSeq Stranded kit and subsequent paired-end Illumina® sequencing 4 7 7 employed the NovaSeq6000 platform. An average of ~19 million, 150 bp paired-end reads were 4 7 8 retrieved for each sample (~6 Gb) and read sequences of low-quality, ribosomal RNA as well as 4 7 9 adaptors were filtered out using Trimmomatic 89 and SortMeRna 90 . RNA-seq reads were aligned 4 8 0 to the Red5 reference gene models using STAR and differential expression analysis was 4 8 1 performed using DESeq2 91 . The details concerning the target file, read statistics and DEGs are 4 8 2 provided in Supplementary Data2. 4 8 3 For RT-qRT-PCR on field samples, RNA was extracted (as described above) from 3-7 clonal 4 8 4 replicates of each genotype from Psa-TMT and Psa-Sus group from heathy leaf tissues. RNA 4 8 5 was extracted from leaf tissues of 5 different genotypes in Psa-FT group, as these genotypes are 4 8 6 not clonally replicated. For RT-qRT-PCR on infected tissues, samples were harvested from the 4 8 7 leaf tissues used for assessment of bacterial growth curve as described above. A 10mm leaf disc 4 8 8 was harvested from the region infected with Psa, at 0, 6, 24 and 48 hrs post-infection, and frozen 4 8 9 in liquid nitrogen. One leaf disc was harvested from a single clonal replicate of 'Hort16A' and P1 4 9 0 per time point and three biological replicates were harvested at each time point. Data from 0 and 4 9 1 24 hrs time is only presented in the study. Total RNA (~ 2 μ g) was treated with DNase I (Roche 4 9 2 Applied Sciences) and used for cDNA synthesis using SuperScript IV Reverse 4 9 3 Transcriptase (Life Technologies-Invitrogen). The cDNA was diluted 20-fold and used for 4 9 4 qRT-PCR employing a LightCycler ® 480 SYBR Green 1 Master PCR labelling kit (Roche 4 9 5 Applied Sciences) and RotorGene 3000 Real time PCR machine (Corbett Research, Sydney, 4 9 6 Australia). Relative transcript abundance was determined relative to the mean of the expression 4 9 7 of Actin and Ubiquitin genes in the same sample. Comparative quantification was performed as 4 9 8 described 92 . Primers used for genes are provided in Supplementary Table 6. 4 9 9 SSR and SNP marker design and screening 5 0 0 Repeats were identified manually in the genome sequence underlying the QTLs. PCR primers for 5 0 1 SSR markers were designed using Primer3 and employed to screen DNA extracted from the 5 0 2 populations 93 . Analysis and scoring of the alleles in the amplicons was performed on a Hitachi 5 0 3 ABI3500 Applied Biosystems genetic analyzer. Primers were also designed around SNPs in the 5 0 4 genes identified in the genomic sequence of Red5 underlying the QTLs. The SNP markers were 5 0 5 screened using real-time High Resolution Melting analysis 94 . All primer sequences are provided 5 0 6 in Supplementary Supplementary We would like to acknowledge AgMARDT New Zealand for a post-doctoral fellowship to JT 5 1 0 and funding the work. Financial support was also provided by; 1) PFR Strategic Science 5 1 1 Investment Funds Breeding Technology Development, 2) Kiwifruit Breeding Programme and 3) 5 1 2 KRIP Phenotying Bioassays. We acknowledge the assistance of Belinda Diepenheim, Andrew 5 1 3 Mullan, Renata Blissett, Bruce Dobson, Matt Spier, Judith Rees, Janet Phipps, Deidre Cornish, 5 1 4 Janet Yu, Jenny Oldham and Jaqui Wallace. PCRs, data analysis, genetic map construction, QTL mapping, designed the SNP and qPCR 5 1 9 markers. JT, SEG and DC wrote the manuscript.
LG performed field phenotyping. SH performed 5 2 0 stab assay. HB developed SSR markers and performed validation of the markers. CB, AC, KT 5 2 1 and MM performed flood assay. EM, AW and KF performed replication of the genotypes in the 5 2 2 tissue culture and greenhouse. CW performed validation of the breeding parents from the 5 2 3 germplasm. CD performed GBS analysis. RC developed circos view of QTLs and RNA-seq data. 5 2 4 MK performed DNA extraction for the validation population. JT, DH and LG performed 5 2 5 statistical analysis of the data. JV performed bacterial growth assessments. JM performed 5 2 6 validation of the LG27 QTL in the breeding germplasm. KH managed orchard plantations of the 5 2 7 genotypes. 5 2 8 Competing interests 5 2 9 I declare that the authors have no competing interests as defined by Nature Research, or other 5 3 0 interests that might be perceived to influence the results and/or discussion reported in this paper. 5 3 1 Cane_death (d) Ooze and (e) Shoot_death. Phenotypes observed in the stab bioassay include f) 5 3 5

Figure Legends
Stem_necrosis, g) Leaf_spots, h) Ooze, i) Stem_collapse, j) Tip_death, k) Wilt and l) is a 5 3 6 representative flood assay (FA_Week3) phenotype for disease response. displays the progression of susceptibility from left to right, while the y-axis represents frequency 5 4 0 in the population. b) LSM of phenotypes from the stab assay, including Stem_necrosis, 5 4 1 Stem_collapse, Tip_death, Psa_score_Stab, Ooze, Leaf_spot and Wilt. c) Means of the health 5 4 2 score from Flood bioassays (FA_Week 1 to FA_Week5). The WSTATISTIC is from the 5 4 3 Shapiro-Wilks test for the null hypothesis that the distribution is normal. Phenotypic scores with 5 4 4 P<0.001 are rejected for the hypothesis that these distributions are normal. d) Principal 5 4 5 components analysis on the correlation matrix of the field assessment, flood assay and stab assay 5 4 6 measures. Genotypes are shown as points and measurements are shown as vectors (lines pointing 5 4 7 from the origin) defined by their correlation with the three principal components. Psa_score_Field. From stab assay phenotypes major QTLs on: c) LG13 in 'Hort16A' for 5 5 2 Stem_necrosis, and in P1 on d) the upper arm of LG27 for Ooze, e) LG10 for Wilt and f) LG1 5 5 3 for Psa_score_Stab. SNPs at peaks are indicated. g) shows dot plot analysis of the allelotypes of 5 5 4 markers underlying quantitative trait loci derived from the Psa_score_Field for the population 5 5 5 'Hort 16A' x P1'. Psa in the field. RNA-seq was performed on young healthy leaf tissues of field-grown plants 5 5 8 belonging to three groups based on relative tolerance / susceptibility. The first group included 5 5 9 three relatively Psa-tolerant plants (Tolerant to Medium Tolerant /Psa-TMT). The second group 5 6 0 included three fully Psa-susceptible genotypes, including 'Hort16A' (Psa-Sus). All had been 5 6 1 exposed to Psa for 1 year in the field. The third group represents the three most tolerant 5 6 2 0 genotypes, tolerant over 3 years in the field (Psa-FT). a) and b) show heat-maps for the genome 5 6 3 wide differential expression (DE) analysis in Psa-TMT vs Psa-Sus and Psa-FT vs Psa-Sus 5 6 4 respectively. c and d) are plots of Principal component analysis for the DE in Psa-TMT vs Psa-5 6 5 Sus and Psa-FT vs Psa-Sus respectively. e) shows volcano plots for the DE, with significantly 5 6 6 (Padj <0.01/log 10 padj, logfold2) upregulated and downregulated genes highlighted in red and 5 6 7 blue respectively in the two comparisons, Psa-TMT vs Psa-Sus and Psa-FT vs Psa-Sus. The grey 5 6 8 dots indicate non-significantly expressed genes, whereas the green dots highlight the genes that 5 6 9 are differentially expressed in common between Psa-TMT vs Psa-Sus and Psa-FT vs Psa-Sus. f) 5 7 0 shows the DE genes in common or unique, respectively, between Psa-TMT vs Psa-Sus and Psa-5 7 1 FT vs Psa-Sus. as well as RNA-seq data associated with Psa-tolerant and susceptible genotypes, anchored on the 5 7 4 chromosomes of the Red5 genome version 1.69.0. Tracks A and B represent differentially 5 7 5 expressed genes (DEGs) with logFC +2 and above in fully tolerant (Psa-FT) vs susceptible (Psa-5 7 6 Sus) and tolerant to medium tolerant (PsaTMT) vs Psa-Sus genotypes, respectively. On track A, 5 7 7 blue circles are upregulated and red circles are downregulated genes in Psa-FT compared to Psa-5 7 8 Sus genotypes. On track B, blue circles are downregulated and red circles are upregulated genes 5 7 9 in Psa-TMT compared to Psa-Sus genotypes. Genes with logFC, between 1 and -1, are 5 8 0 represented by green circles. Increase in circle diameter indicates increasing logFC value. reverse transcription polymerase chain reaction (RT-qRT-PCR). Data represents mean relative 5 9 0 gene expression of the candidate genes, in three to seven clonal replicates for each genotype, to 5 9 1 the mean of Actin and Ubiquitin genes and plotted using geom_boxplot. Asterisks represents 5 9 2 statistically significant differences in the relative expression of the candidate genes in genotypes 5 9 3 of Psa-TMT and Psa-FT compared to the mean relative expression of genes in Psa-Sus genotypes 5 9 4 using Student's t-test. 5 9 5 5 9 6 References 5 9 7    h  y  t  o  p  a  t  h  o  l  o  g  y   5  3  ,  5  1  3  -5  3  9  ,  d  o  i  :  1  0  .  1  1  4  6  /  a  n  n  u  r  e  v  -p  h  y  t  o  -1  0  2  3  1  3  -0  4  5  9  1  4  7  2  5 ( 2 0 1 5  M  a  n  d  i  p  r  o  p  a  m  i  d  t  a  r  g  e  t  s  t  h  e  c  e  l  l  u  l  o  s  e  s  y  n  t  h  a  s  e  -l  i  k  e  P  i  C  e  s  A  3  t  o  i  n  h  i  b  i  t  c  e  l  l  8  0  8  w  a  l  l  b  i  o  s  y  n  t  h  e  s  i  s  i  n  t  h  e  o  o  m  y  c  e  t  e  p  l  a  n  t  p  a  t  h  o  g  e  n  ,  P  h  y  t  o  p  h  t  h  o  r  a  i  n  f  e  s  t  a  n  s  .   M  o  l  e  c  u  l  a  r   8  0  9   P  l  a  n  t  P  a  t  h  o  l  o  g  y   1  1  ,  2  2  7  -2  4  3  ,  d  o  i  :  d  o  i  :  1  0  .  1  1  1  1  /  j  .  1  3  6  4  -3  7  0  3  .  2  0  0  9  .  0  0  6  0  4  .  x  (  2  0  1  0  )  .  8  1  0  7  5  M  i  l  l  e  r  ,  J  .  C  .  ,  C  h  e  z  e  m  ,  W  .  R  .  &  C  l  a  y  ,  N  .  K  .  T  e  r  n  a  r  y  W  D  4  0  R  e  p  e  a  t  -C  o  n  t  a  i  n  i  n  g  P  r  o  t  e  i  n  8  1  1  C  o  m  p  l  e  x  e  s  :  E  v  o  l  u  t  i  o  n  ,  C  o  m  p  o  s  i  t  i  o  n  a  n  d  R  o  l  e  s  i  n  P  l  a  n  t  I  m  m  u  n  i  t T  r  i  m  m  o  m  a  t  i  c  :  a  f  l  e  x  i  b  l  e  t  r  i  m  m  e  r  f  o  r  I  l  l  u  m  i  n  a  8  5  1  s  e  q  u  e  n  c  e  d  a  t  a  .   B  i  o  i  n  f  o  r  m  a  t  i  c  s   3  0  ,  2  1  1  4  -2  1  2  0  (  2  0  1  4  )  .  8  5  2  9  0  K  o  p  y  l  o  v  a  ,  E  .  ,  N  o  é  ,  L  .  &  T  o  u  z  e  t  ,  H  .  S  o  r  t  M  e  R  N  A  :  f  a  s  t  a  n  d  a  c  c  u  r  a  t  e  f  i  l  t  e  r  i  n  g  o  f  r  i  b  o  s  o  m  a  l  8  5  3  R  N  A  s  i  n  m  e  t  a  t  r  a  n  s  c  r  i  p  t  o  m  i  c  d  a  t  a  .   B  i  o  i  n  f  o  r  m  a  t  i  c  s   2  8  ,  3  2  1  1  -3  2  1  7  (  2  0  1