The avocado genome informs deep angiosperm phylogeny, highlights introgressive hybridization, and reveals pathogen-influenced gene space adaptation Martha Rendón-Anayaa,b,1, Enrique Ibarra-Laclettea,c,1, Alfonso Méndez-Bravoa,d, Tianying Lane, Chunfang Zhengf, Lorenzo Carretero-Pauletg, Claudia Anahí Perez-Torresa,c, Alejandra Chacón-Lópeza, Gustavo Hernandez-Guzmána,h,i, Tien-Hao Change, Kimberly M. Farre, W. Brad Barbazukj, Srikar Chamalak, Marek Mutwill, Devendra Shivharel, David Alvarez-Poncem, Neena Mittern, Alice Haywardn, Stephen Fletchern, Julio Rozaso,p, Alejandro Sánchez Graciao,p, David Kuhnq, Alejandro F. Barrientos-Priegor, Jarkko Salojärvil, Pablo Librados,t, David Sankofff, Alfredo Herrera-Estrellaa, Victor A. Alberte,l,2, and Luis Herrera-Estrellaa,u,2 a Unidad de Genomica Avanzada/Langebio, Centro de Investigación y de Estudios Avanzados, Irapuato 36821, México; bDepartment of Plant Biology, Uppsala BioCenter, Swedish University of Agricultural Sciences, SE-750 07 Uppsala, Sweden; cRed de Estudios Moleculares Avanzados, Instituto de Ecología A.C., 91070 Xalapa, México; dEscuela Nacional de Estudios Superiores, Laboratorio Nacional de Análisis y Síntesis Ecológica, Universidad Nacional Autónoma de México, 58190 Morelia, México; eDepartment of Biological Sciences, University at Buffalo, Buffalo, NY 14260; fDepartment of Mathematics and Statistics, University of Ottawa, Ottawa, ON, Canada K1N 6N5; gCenter for Plant Systems Biology, Vlaams Instituut voor Biotechnologie (VIB), University of Ghent, 9052 Ghent, Belgium; hDepartamento de Alimentos, Universidad de Guanajuato, 36500 Irapuato, México; iDivisión de Ciencias de la Vida, Universidad de Guanajuato, 36500 Irapuato, México; jDepartment of Biology, University of Florida, Gainesville, FL 32611; kDepartment of Pathology, Immunology and Laboratory Medicine, University of Florida, Gainesville, FL 32610; lSchool of Biological Sciences, Nanyang Technological University, Singapore 637551; m Department of Biology, University of Nevada, Reno, NV 89557; nCentre for Horticultural Science, Queensland Alliance for Agriculture and Food Innovation, The University of Queensland, St. Lucia, QLD 4072, Australia; oDepartament de Genètica, Microbiologia i Estadística, Universitat de Barcelona, 08007 Barcelona, Spain; pInstitut de Recerca de la Biodiversitat, Universitat de Barcelona, 08007 Barcelona, Spain; qSubtropical Horticulture Research PLANT BIOLOGY Station, Agricultural Research Service, US Department of Agriculture, Miami, FL 33158; rPosgrado en Horticultura, Departamento de Fitotecnia, Universidad Autónoma Chapingo, 56230 Texcoco, México; sCentre for GeoGenetics, Natural History Museum of Denmark, 1017 Copenhagen, Denmark; tLaboratoire d’Anthropobiologie Moléculaire et d’Imagerie de Synthèse, CNRS Unité Mixte de Recherche 5288, Université de Toulouse, Université Paul Sabatier, 31330 Toulouse, France; and uDepartment of Plant and Soil Science, Texas Tech University, Lubbock, TX 79409 Contributed by Luis Herrera-Estrella, July 9, 2019 (sent for review December 31, 2018; reviewed by Todd P. Michael and Yves Van de Peer) The avocado, Persea americana, is a fruit crop of immense impor- flowering plant lineages, comprising about 285,000 species tance to Mexican agriculture with an increasing demand world- combined (2). Avocados are a vital crop for Mexico, from wide. Avocado lies in the anciently diverged magnoliid clade of which almost 50% of all avocado exports originate, valued at angiosperms, which has a controversial phylogenetic position rel- ative to eudicots and monocots. We sequenced the nuclear ge- Significance nomes of the Mexican avocado race, P. americana var. drymifolia, and the most commercially popular hybrid cultivar, Hass, and an- The avocado is a nutritious, economically important fruit species chored the latter to chromosomes using a genetic map. Resequenc- that occupies an unresolved position near the earliest evolutionary ing of Guatemalan and West Indian varieties revealed that ∼39% branchings of flowering plants. Our nuclear genome sequences of of the Hass genome represents Guatemalan source regions intro- Mexican and Hass variety avocados inform ancient evolutionary gressed into a Mexican race background. Some introgressed blocks relationships and genome doublings and the admixed nature of are extremely large, consistent with the recent origin of the culti- Hass and provide a look at how pathogen interactions have sha- var. The avocado lineage experienced 2 lineage-specific polyploidy ped the avocado’s more recent genomic evolutionary history. events during its evolutionary history. Although gene-tree/ species-tree phylogenomic results are inconclusive, syntenic ortho- Author contributions: N.M., A.H.-E., V.A.A., and L.H.-E. designed research; M.R.-A., E.I.-L., log distances to other species place avocado as sister to the enor- A.M.-B., C.A.P.-T., A.C.-L., G.H.-G., N.M., A.H., D.K., A.F.B.-P., D. Sankoff, and V.A.A. per- mous monocot and eudicot lineages combined. Duplicate genes formed research; N.M., D.K., A.F.B.-P., and L.H.-E. contributed new reagents/analytic tools; descending from polyploidy augmented the transcription factor M.R.-A., E.I.-L., A.M.-B., T.L., C.Z., L.C.-P., C.A.P.-T., A.C.-L., G.H.-G., T.-H.C., K.M.F., W.B.B., S.C., M.M., D. Shivhare, D.A.-P., A.H., S.F., J.R., A.S.G., J.S., P.L., D. Sankoff, A.H.-E., V.A.A., diversity of avocado, while tandem duplicates enhanced the sec- and L.H.-E. analyzed data; and M.R.-A., E.I.-L., A.H.-E., V.A.A., and L.H.-E. wrote the paper. ondary metabolism of the species. Phenylpropanoid biosynthesis, Reviewers: T.P.M., J. Craig Venter Institute; and Y.V.d.P., Ghent University. known to be elicited by Colletotrichum (anthracnose) pathogen The authors declare no conflict of interest. infection in avocado, is one enriched function among tandems. This open access article is distributed under Creative Commons Attribution-NonCommercial- Furthermore, transcriptome data show that tandem duplicates NoDerivatives License 4.0 (CC BY-NC-ND). are significantly up- and down-regulated in response to anthrac- Data deposition: Bioproject: PRJNA508502. Biosamples: SAMN10523735, SAMN10523720, nose infection, whereas polyploid duplicates are not, supporting SAMN10523736, SAMN10523738, SAMN10523739, SAMN10523746, SAMN10523747, the general view that collections of tandem duplicates contribute SAMN10523748, SAMN10523749, SAMN10523750, SAMN10523752, SAMN10523753, evolutionarily recent “tuning knobs” in the genome adaptive land- and SAMN10523756. SRA submission: SUB4878870. Whole Genome Shotgun projects scapes of given species. have been deposited at DDBJ/ENA/GenBank (accession nos. SDXN00000000 and SDSS00000000). The versions described in this paper are version SDXN01000000 and SDSS01000000 (P. americana var. drymifolia and P. americana cultivar Hass, respectively). | avocado genome angiosperm phylogeny | genome duplications | The genome assemblies and annotations are available at https://genomevolution.org/ | Phytophthora genome evolution CoGe/SearchResults.pl?s=29305&p=genome and https://genomevolution.org/coge/ SearchResults.pl?s=29302&p=genome (P. americana var. drymifolia and P. americana cultivar Hass, respectively). T he avocado, Persea americana, is a commercially important tree fruit species in the Lauraceae family, otherwise known for the spices cinnamon, bay leaves, and sassafras (gumbo filé) 1 2 M.R.-A. and E.I.-L. contributed equally to this work. To whom correspondence may be addressed. Email: vaalbert@buffalo.edu or lherrerae@ cinvestav.mx. (1). Lauraceae is contained within the early diverging magnoliid This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. lineage of angiosperms, which at about 11,000 total species is 1073/pnas.1822129116/-/DCSupplemental. minuscule in comparison to the dominant eudicot and monocot Published online August 6, 2019. www.pnas.org/cgi/doi/10.1073/pnas.1822129116 PNAS | August 20, 2019 | vol. 116 | no. 34 | 17081–17089 about $2.5 billion US dollars.* Although the avocado has an We also anchored the Hass genome to an avocado genetic ancient cultivation history in Mexico and Central to South map (12). Two large mapping populations of 1,339 trees were America (5), its extreme worldwide popularity as an oily, nutty- genotyped with 5,050 SNP markers from transcribed genes, and flavored fruit with highly beneficial nutritional properties dates the resulting map was used to order the Hass scaffolds into 12 mainly from the early 20th century (6). Cultivated avocados oc- linkage groups, matching the avocado haploid chromosome cur in 3 landraces with possibly independent cultivation origins number (see, e.g., chromosome 4, Fig. 2A). The total length of that reflect their current distribution: the Mexican, Guatemalan, the anchored genome accounts for 46.2% of the Hass assembly and West Indian varieties (6). The principal industrial avocado and represents 915 scaffolds, 361 of which could be oriented (SI cultivar is known as Hass, after the grower who first patented it Appendix, section 1.5). in 1935. Hass represents a hybrid between the Guatemalan and Mexican races, but its precise breeding history is unknown (6, 7). SNPs, Population Structure, and the Parentage of Hass. To study Here, we generate and analyze the complete genome sequences avocado from a population genomic perspective, we resequenced of a Hass individual and a representative of the highland Mexi- accessions of different races and cultivars and mapped the reads can landrace, Persea americana var. drymifolia. We also study against the Hass reference genome assembly (SI Appendix, Table genome resequencing data for other Mexican individuals, as well S2). The estimated depth of coverage ranged from 3.3 to 39×, as Guatemalan and West Indian accessions. We use these data to with breadth of coverage between 70 and 92% (SI Appendix, study the admixed origin of the Hass cultivar and demonstrate its section 3.2). Given the uneven sequencing coverage, we used ANGSD to call SNPs across the entire (unanchored) genome racewise parentage more precisely. We evaluate the phylogenetic assembly, followed by a stringent pruning based on per-site origin of avocado among angiosperms and provide information depth, minor allele frequencies, and linkage disequilibrium, that on avocado’s unique polyploid ancestry. The adaptive landscape resulted in 179,029 high-quality SNP variants. Phylogenetic, of the avocado genome in terms of its duplicate gene functional principal component, and identity-by-state (IBS) analyses de- diversity was also explored. We further evaluate gene expression rived from this dataset (Fig. 1A and SI Appendix, section 4) patterns during the defense response of Hass avocado to an- cluster the samples belonging to the Hass cultivar and Guate- thracnose disease and how this is partitioned by gene duplication malan variety into 2 groups as expected according to their ge- mechanisms. netic background. Principal component analysis of genome-wide Results and Discussion SNPs showed relative uniformity in Costa Rican/West Indian/ Guatemalan group but strong heterogeneity within the Mexican Plant Material, Genome Assembly, and Annotation. Due to growing subpopulation, wherein the unusual accession Tiny Charly is a market demand, 90% of cultivated avocado corresponds to the divergent sample (SI Appendix, Fig. S15). SNPhylo (13) results cultivar Hass, which in Mexico is commonly grafted on Mexican reflected the poor fit of the SNP data to a bifurcating tree by race (P. americana var. drymifolia) rootstock (6). This practice embedding the hybrid Hass within an otherwise Mexican clade, 1 makes it possible to maintain high productivity as the indigenous known parent of this hybrid cultivar (SI Appendix, Fig. S13). race is well-adapted to Mexican highland soils. The Hass cultivar Furthermore, in that lineage’s sister group, Guatemalan acces- and Mexican race were chosen to generate reference genomes sions were derived within an otherwise Costa Rican/West Indian (SI Appendix, section 1). Additionally, to explore the genetic lineage, suggesting an admixed origin involving Guatemalan and diversity available in avocado, we resequenced representative other sources. Phylogenetic patterns generated from chromosome- individuals from the 3 avocado botanical varieties (vars. dry- wise SNP subsets (based only on contigs anchored to chromo- mifolia, guatemalensis, and americana), including the disease- somes) recapitulated these relationships for 7 of avocado’s 12 resistant rootstock Velvick (8), an additional Hass specimen chromosomes, whereas 1 chromosome supported Hass to be and the early flowering/fruiting Hass somatic mutant Carmen sister to the Costa Rican/West Indian lineage, perhaps reflective Hass cultivar [otherwise known as Mendez No. 1 (6, 9)], as well as of chromosomal differences in the admixture proportions of this wild avocados of the West Indian variety (P. americana var. costar- hybrid cultivar (SI Appendix, Fig. S14 and Dataset S6). Fur- icensis). Persea shiedeana (the edible coyo), a species relatively thermore, IBS clustering placed Hass intermediate between the closely related to P. americana (10), was also included (SI Guatemalan and Mexican subpopulations, agreeing with the Appendix, Tables S1 and S2). hybrid nature of this variety (SI Appendix, Fig. S16). De novo and evidence-directed annotation revealed a similar To account for further evidence of admixture in the Hass number of protein coding genes in each genome: 22,441 from the reference genome, we used NGSAdmix (14) modeling different Mexican race and 24,616 from Hass (Table 1, SI Appendix, sec- possible numbers of source populations (K = 1 to 6) (SI Ap- tion 2, and Datasets S2 and S3). We next used the Benchmarking pendix, section 5 and Fig. S17). The Akaike information criterion Universal Single-Copy Orthologs (BUSCO) software to estimate (AIC) indicated K = 1 as the preferred model, reflecting poor the presence of 1,440 conserved embryophyte single-copy genes population structuring within avocado as a whole. However, (11) in the annotations, leading to estimated completeness per- since we know Hass is admixed a priori, we chose the smallest centages of 85% and 86.3% for Hass and Mexican avocado, (most parsimonious) K for which Hass admixture appears (K = 3). respectively (Table 1). The Mexican race was sequenced using This criterion predicts the following 3 populations: 1) P. schie- the short-read, high-coverage Illumina sequencing platform, deana, 2) the West Indian plus the Guatemalan varieties, and the while the Hass genome was sequenced using the long-read 3) Mexican accessions (Fig. 1B). Combining the IBS and Pacific Biosciences sequencing technology. Given the similar NGSAdmix observations, we specifically calculated the contri- BUSCO scores, we used the larger Hass genome assembly for bution of Guatemalan and Mexican backgrounds into the Hass downstream single-nucleotide polymorphism (SNP) calling, as subpopulation. EIGMIX (15) revealed that the greatest admix- PacBio technology lowers the probability of contig misassembly ture proportion, 61%, stemmed from the Mexican race (SI Ap- and permits incorporation of substantially more repetitive DNA pendix, Fig. S18 and Table S8). sequence and genes lying within it into the assembled genome, Although based on ∼46% anchoring of scaffolds to chromo- which might have otherwise been missed. somes, we investigated chromosomewise signatures of admixture in the Hass genome (SI Appendix, section 5.2). We calculated the ^f , ^f , and dXY estimators of introgression and divergence (SI d dM *Based on information taken from refs. 3 and 4. Appendix, Fig. S19 and Dataset S7) according to Martin et al. 17082 | www.pnas.org/cgi/doi/10.1073/pnas.1822129116 Rendón-Anaya et al. Table 1. General statistics of the avocado assemblies and their annotations P. americana assembly Metrics Var. drymifolia Hass cultivar No. of contigs 99,957 8,135 Total length of contigs, bp 668,137,248 912,697,600 No. of scaffolds 42,722 — Total length of scaffolds, bp 823,419,498 — Longest contig/scaffold, bp 254,240/4,610,966 2,811,280/– Mean contig/scaffold length, bp 6,684/19,274 112,194/– N50 contig/scaffold length, bp 11,724/323,854 296,371/– Assembly in scaffolded contigs, % 87.6 0 Assembly in unscaffolded contigs, % 12.4 100 Protein coding genes (% BUSCO completeness) 22,441 (86.3%) 24,616 (85%) (16) in nonoverlapping 100-kb windows, controlling the di- Guatemalan origin across the 12 chromosomes (Fig. 2, SI Ap- rectionality of gene flow from the Guatemalan race into Hass pendix, section 5.2 and Figs. S21 and S23–S34). Chromosome 4 versus the Mexican race into Hass, setting P. schiedeana as the illustrates these analyses well, demonstrating that a huge Gua- outgroup and leaving Tiny Charly out of the Mexican sub- temalan block, which could encompass an entire chromosome population to avoid the bias this divergent accession could in- arm, is present in the Hass genome (Fig. 2A). The length of this troduce into calculations (Fig. 2B). Genomic regions that behave Guatemalan-derived block, uninterrupted by recombination, as ^f dM outliers can be distinguished as introgressed from ances- PLANT BIOLOGY reflects the extremely recent hybrid origin of the cultivar. tral variation if the absolute genetic distance dXY is also reduced We also calculated the level of nucleotide diversity [π (17, 18)] between donor (P3) and receptor (P2). In the presence of gene in each population (Mexican, Guatemalan, and Hass), the FST flow, genomic windows coalesce more recently than the lineage index (19) to determine regions of high differentiation between split, so the magnitude of reduction in P2 − P3 dXY is greater than varieties, and Tajima’s D (20) in order to evaluate any deviations in the case where recombination and hybridization are absent. from neutral evolution (SI Appendix, section 5.3 and Dataset S7). We evaluated several ^f dM cutoffs (Q50, 75, and 90; SI Appendix, We observed that Hass has the lowest nucleotide diversity (π = Fig. S20) and observed a remarkable reduction of genetic di- 0.06) and very high Tajima’s D in all chromosomes (genomic vergence in the scenario where gene flow occurs from the average of 1.5), as expected for individuals derived from a recent Guatemalan race into Hass. Considering those blocks with founder event and clonally propagated; these values contrast ^f with the low, positive Tajima’s D values in the Mexican and dM½Guat−Hass > 0.174 (Q50), dXY ½Guat−Hass < 0.113 (mean di- Guatemalan populations (genomic averages of 0.19 and 0.11, vergence between subpopulations) and ^f < 0.114 dM½Drym−Hass respectively; SI Appendix, Fig. S22). In the case of chromosome 4, (Q50), we were able to define 840 high-confidence regions of the FST index between the Mexican race and Hass corroborates our A B Mendez reference var. costaricensis merged 0.4 var. drymifolia var. guatemalensis Hass Velvick West Indian 0.2 reference P. schiedeana PC1 (16.7%) 069 02 0.0 CH G 11 CH G 10 Velvick CH G 07 0.2 Tiny Charly CH CR 25 P. schiedeana CH G 30 0.4 0.2 0.0 0.2 PC2 (13.6%) Fig. 1. Population genomic structure of avocado. (A) Principal component analysis (PCA) of genome-wide SNPs reveals population groupings among races and varieties. The Guatemalan and West Indian/Costa Rican accessions are closely related, while the Mexican (P. americana var. drymifolia) specimens are more diverse, with the unusual individual Tiny Charly drawn toward the outgroup species P. schiedeana by PC2. Hass and its sport Mendez are tightly clustered and intermediate between Mexican and Guatemalan and West Indian/Costa Rican on PC2. (B) NGSAdmix analysis reveals similar population structure at K = 3. The P. schiedeana outgroup is distinct, and the Hass reference genome is revealed to be admixed between Guatemalan–West Indian and Mexican source populations, the Mexican source clearly contributing greater than 50%. Rendón-Anaya et al. PNAS | August 20, 2019 | vol. 116 | no. 34 | 17083 A B C 0.0e+00 5.0e+06 chr position 1.0e+07 1.5e+07 2.0e+07 0.0 0.2 0.4 0.1 0.2 0.3 0.4 0.5 0.2 0.0 0.2 0.4 0.6 1 0 1 2 Fst dxy FdM Tajima's D guatemalensis guatemalensis_Hass drymifolia drymifolia_Hass Hass Fig. 2. SNP diversity analysis reveals the hybrid genomic background of Hass avocado. (A) Twenty-two megabases (Mb) of anchored DNA on chromosome 4 exemplify the hybrid nature of Hass, in which genomic introgression from the Guatemalan avocado race (var. guatemalensis) occurred into a Mexican (var. drymifolia) genetic background. (B) While in the top chromosome arm the blue trend line shows a low differentiation index (FST) between Hass and the Mexican subpopulation as well as a high introgression signal ð^f d Þ from var. drymifolia into Hass, these signals should not be misinterpreted as introgression events, since the absolute genetic divergence ðdXY Þ between both sets of accessions does not vary along the chromosome. The lower arm of the chromosome, however, has inverted trends, where our estimators describe an elevated introgression signal from var. guatemalensis into Hass, as confirmed by the decay in dXY (red trend line), and higher FST between Hass and Mexican accessions. (C) No evidence for selective sweeps or domestication signatures were identified; Mexican and Guatemalan subpopulations displayed neutral D values while Hass maintained extreme D values at the theoretical upper limit of the estimator (∼2). Such positive values reflect a “bottlenecked” origin with clonal expansion after the very recent foundation of the cultivar only a few decades ago. Each dot in the plots corresponds to statistics for SNP data in nonoverlapping 100-kb windows (confidence interval of 0.90 for graphical smoothed conditional means). Apparent centromeric regions are located at around 10 Mb, where FST, ^f d and dXY intersect and Tajima’s D for Hass decreases. previous conclusions on admixture, showing that approximately these events with respect to the gamma hexaploidy (23, 24) and half of the chromosome corresponds to the Mexican background species splits using Ks density plots of orthologous and paralogous whereas the other half has its origin in the Guatemalan pop- gene pairs between avocado, Amborella, and Vitis (SI Appendix, ulation (Fig. 2B). section 6.2). Avocado polyploidy events are lineage-specific as both postdate the divergence of the avocado lineage from Whole-Genome Duplication History. Next, we investigated P. common ancestry with either Amborella or Vitis (Fig. 3C and SI americana genome structural history and relatedness of the Appendix, Fig. S36). Considerable fractionation (alternative dele- species to other major groups of angiosperms (SI Appendix, tion of duplicated genes between polyploid subgenomes) since section 6). We used the CoGe SynMap tool (21) to examine these 2 polyploidy events is observable in blockwise relationships of avocado self:self and avocado:Amborella synteny. Amborella, the about 4:1 (Fig. 3B and SI Appendix, Fig. S35). The blockwise re- single living representative of the sister lineage to all other an- lationship of 4:1 for avocado:Amborella suggests, though does not giosperms, is known to show a 1:3 syntenic block relation- prove, that the 2 events were WGDs and not triplications. Further ship compared with Vitis (grapevine) and its ancient hexaploid quantitative analysis using well-conserved orthologous syntenic structure shared with all core eudicots (22–25). As such, the “superblocks” (26) between avocado and 15 other angiosperms Amborella genome displays no additional whole-genome dupli- strongly supported the conclusion that the most recent polyploidy cations (WGDs) since the angiosperm last common ancestor. event in avocado was a WGD and not a triplication, as in the an- Consequently, paralogous syntenic blocks discovered within cient gamma hexaploidy event (SI Appendix, section 6.3). self:self and avocado:Amborella plots could reflect WGDs unique to the avocado lineage that occurred since last common Phylogenomic Analysis of Avocado’s Placement among Angiosperms. ancestry with Amborella. Using these approaches, we discovered To further corroborate the placement of these 2 polyploidy that the avocado genomes show evidence for 2 ancient poly- events as specific to the avocado lineage, we generated phylo- ploidy events (Fig. 3 A–C). We investigated the relative timing of genetic trees of representatives of the major angiosperm lineages 17084 | www.pnas.org/cgi/doi/10.1073/pnas.1822129116 Rendón-Anaya et al. A B PLANT BIOLOGY C Fig. 3. Phylogenomic and whole-genome duplication history of avocado. (A) An ultrametric time tree based on universally present single-copy protein sequences depicts 1 of 3 common resolutions of Persea (Magnoliidae) relationships to other flowering plants. This topology, showing avocado sister to monocots plus eudicots, mirrors phylogenetic relationships derived from syntenic distances. Here, the split time between the last common ancestor of av- ocado and the monocot/eudicot crown group is less than 4 million y. Pie charts at 50% positions on branches show proportions of gene gains (orange) versus losses (blue) as determined by BadiRate’s birth–death–innovation model. Yellow–green (greater–lesser) heat map to the right of the tree depicts relative numbers of genes in the modern genomes. Syntenic analysis revealed 2 independent WGD events (red stars) during avocado’s evolutionary history. (B) Hass avocado (bottom 4 genomic blocks) shows 4:1 intercalated syntenic relationships with Amborella (upper block). (C) Syntenic homologs in avocado show a bimodal Ks distribution suggestive of 2 polyploidy events (numbered 1 and 2; cyan: Hass:Hass paralogs; green: Hass:drymifolia homologs) following the split between magnoliids and Amborella (red syntenic homologs). These events postdate the species split between Vitis and avocado (purple syntenic homologs) and so are independent of the gamma triplication that underlies Vitis. Rendón-Anaya et al. PNAS | August 20, 2019 | vol. 116 | no. 34 | 17085 using 2 data forms: coding sequence alignments and modal dis- comparatively small over time (39, 40). We observe a correlation tances within large collections of syntenic orthologs between between turnover rates and branch lengths at the multigene family species pairs (SI Appendix, section 7). level (SI Appendix, Fig. S46), suggesting pervasive copy number Single-copy gene families [presumed unambiguous orthologs, variation (CNV) in the ancestral species, possibly exacerbated by those that returned to single copy following duplicate deletions WGD and subsequent fractionation processes. Short phylogenetic after the various polyploidy events in flowering plant history branches, representing rapid speciation events, increase the in- (27)] were retrieved from orthogroup classification of 19 angio- cidence of ILS in phylogeny reconstruction since extinctions of al- sperm proteomes, including those of avocado, Amborella, and ternative duplicate copies within ancestral populations (e.g., unfixed representatives of monocots and eudicots (SI Appendix, section CNVs) further break up branches that are nearly time-coincident 3.1 and Dataset S4). Phylogenetic trees based on 176 stringently already (41). According to BadiRate estimates, the temporal im- filtered single-copy gene alignments (SI Appendix, section 7.1 pact of ILS on turnover rates extended well beyond 10 Mya, a time and Datasets S8 and S9) gave different results for amino acid frame exceeding the branch length of the lineage that existed versus inferred codon data. Based on protein sequences, avocado immediately prior to avocado divergence from other species, was resolved as sister to monocots plus eudicots (i.e., branching which varied in age from only 7.4 to as little as 3.8 My (Fig. 3A before their divergence from each other; cf. refs. 28 and 29), and SI Appendix, Table S11 and Figs. S46 and S47). This implies whereas from coding sequences avocado was placed as sister to that the 3 different placements of avocado among angiosperms monocots only (cf. ref. 30) (SI Appendix, Figs. S40 and S41, re- may be impossible to discriminate among for purely biological spectively). In a different analysis we included Gnetum (a gym- reasons (cf. ref. 42). Yet, 1 of the 3 different tree topologies was nosperm) and Selaginella (a nonseed plant) in orthogroup preferred based on AIC contrasts under the FR model: the to- classification to generate a rooted species tree from all gene pology wherein magnoliids are sister to monocots plus eudicots trees (4,694) that contained one or more (i.e., paralogous) gene (Fig. 3A and SI Appendix, Fig. S47). copies from all species (SI Appendix, section 7.2). Here, avocado was resolved as sister to eudicots only (SI Appendix, Fig. S42), a Functional Enrichments in Duplicate Gene Space. Duplicate gene result similarly found in transcriptome-based analyses of large collections within plant genomes mainly derive from 2 processes, numbers of species (26, 31). In an altogether different approach local and ongoing tandem duplication events, many of which may (32, 33), we generated a neighbor-joining tree based on modal be recent, and global and often ancient polyploidy events dissimilarity scores from thousands of syntenically validated wherein entire gene complements are duplicated (43). Sub- ortholog pairs generated by the SynMap function on the CoGe functionalization and/or neofunctionalization of duplicate gene platform (21) (SI Appendix, section 7.3). Here, avocado was copies (44) results in retained descendants of duplication events again placed as sister to monocots plus eudicots, as in Fig. 3A (SI that have differentially escaped the otherwise usual fate of Appendix, Fig. S44). duplicates—pseudogenization—through functional divergence. Apparently, the early branching orders of the angiosperms are Tandem duplication is problematic for genes that are part of extremely difficult to determine using protein coding sequences. dosage-sensitive transcriptional regulatory networks, or for genes This problem is due in part to sequence parallelism/reversal over that code for parts of multiprotein complexes (45). Such func- deep time, limitations in taxon sampling (including unknown tions are more likely to remain among the surviving duplicate extinctions), biases in sequence-based ortholog versus paralog complements stemming from precisely dosage-balancing poly- determination, but clearly also to the relatively coincident ploidy events (46). On the other hand, dosage-responsive func- branching times of the species involved (see figure 6 of ref. 34 tions such as secondary metabolism (including biochemical and also below). Rapid species divergences can lead to real gene- pathway addition) are among those most likely to survive as sub- tree/species-tree discordances through enhanced occurrence of or neofunctionalized tandem duplicates (45). These patterns incomplete lineage sorting (ILS), wherein polymorphic allele have been repeatedly observed among plant genomes, wherein states in ancestral populations do not have enough time to fix secondary metabolic function is most prevalent among tandems, according to the species tree (35–37). and transcriptional function is enriched among polyploid dupli- In an experimental approach to the problem, we further in- cates (e.g., refs. 46 and 47). The avocado genome provides no vestigated the possible role of ILS using gene family turnover exception to this rule; we identified precisely these over- analysis as incorporated in BadiRate (38) (SI Appendix, section representation patterns among Gene Ontology (GO) and Kyoto 7.4). Trees with the 3 alternative placements of avocado were Encyclopedia of Genes and Genomes (KEGG) categories for converted into time-calibrated ultrametric trees, and the likeli- these different classes of gene duplicates, separated using the hoods of duplicate gains versus losses were evaluated under 4 CoGe platform (SI Appendix, section 8). Among 2,433 total different branch models (SI Appendix, section 7.4). The AIC polyploid duplicates, “regulation of transcription, DNA-tem- clearly favored free-rates (FR) models, supporting heterogeneous plated” was significantly overrepresented by 352 genes (Dataset rates of multigene family evolution across lineages (Dataset S10). S12). Enriched functions among tandem duplicates was highly Interestingly, such uneven rates of gene turnover cannot be entirely illustrative of the secondary metabolic landscape particular to explained by lineage-specific WGD/whole-genome triplication avocado (Dataset S12). We show that “phenylpropanoid bio- (WGT) events, given that FR models fit multigene family data synthesis” and closely related KEGG pathways (Dataset S13) better than WGD/WGT models alone. Additionally, allowing are significantly enriched among tandem duplicates (P = turnover rates to vary in each short branch (<10 My) also improved 2.08e-08; Fisher’s exact test, Bonferroni-corrected). This likelihood and AIC values, although the fit was still worse than functional enrichment in a long-lived tree may have evolved in under the FR model (Dataset S10). That FR models fit gene count response to pathogen infection, including Colletotrichum (an- data significantly better could be explained by their flexibility to thracnose) and Phytophthora cinnamomi (avocado root rot), accommodate variation that is not explicitly accounted for by cur- both of which are reported to activate the phenylpropanoid rent turnover models, such as gene copy variation within species. biosynthetic pathways in avocado (48–50). Several GO func- Intraspecific variation, segregating in an ancestral population, can tional enrichments among avocado tandems (for example, be inherited differently by 2 splitting lineages, which will thus start “1,3-beta-D-glucan synthase activity” and “regulation of cell diverging with a significant fraction of differentiation. This pre- shape”; P = 1.64e-05 and 0.00258, respectively; Fisher’s exact dicts that divergence will be inflated for short branches, and that test, Bonferroni-corrected) relate to callose synthase activity this bias will become negligible as divergence times increase, be- (51, 52), a recently discovered avocado defense mechanism against cause its relative contribution to the total divergence tends to be P. cinnamomi (53). Other significantly enriched GOs include 17086 | www.pnas.org/cgi/doi/10.1073/pnas.1822129116 Rendón-Anaya et al. “phenylpropanoid metabolic process,” “lignin biosynthetic process,” (P = 0.038984; Fisher’s exact test, Bonferroni-corrected). Among and “UDP-glycosyltransferase activity” (P = 0.00142, 7.36e-07 genes with this annotation were 4 homologs of XYLOGLUCAN and 5.16e-07, respectively; Fisher’s exact test, Bonferroni- ENDOTRANSGLUCOSYLASE/HYDROLASE 22 [XTH22; also corrected), categories directly or closely related to phenyl- known as TOUCH 4 (TCH4) (66)]. XTH22 and similar genes propanoid biosynthesis (54, 55). The lignin functional enrich- encoding cell-wall–modifying proteins have been shown to up- or ment, for example, includes diverse tandemly duplicated genes down-regulate after Citrus Huanglongbing infection (67), whitefly involved in many pathway-interrelated processes, including ho- infestation (65), and herbivore (68) or mechanical stimulation (69), mologs of both biosynthetic and regulatory genes encoding the latter provoking a Botrytis-protective response. In a different HYDROXYCINNAMOYL-COA SHIKIMATE/QUINATE pathogen response, up-regulation of XTH22 occurs in concert with HYDROXYCINNAMOYL TRANSFERASE (HCT), CINNAMYL pectin digestion in Pseudomonas-sensitive Arabidopsis lines that ALCOHOL DEHYDROGENASE 5 (CAD5), LACCASE 17 overexpress IDA-LIKE 6 [IDL6 (70)]. (LAC17), CAFFEATE O-METHYLTRANSFERASE 1 (COMT1), PEROXIDASE 52 (PRX52), NAC DOMAIN CONTAINING Conclusions PROTEIN 12 (NAC012), and NAC SECONDARY WALL Our genomes of Mexican and Hass avocados provide the req- THICKENING PROMOTING FACTOR 1 (NST1). As could uisite resources for genome-wide association studies to identify be expected from the above, the GOs “defense response” and important traits among natural avocado genetic diversity present “defense response to fungus” are significantly enriched among in Mesoamerica, to develop genome-assisted breeding and ge- tandem duplicates (P = 0.000165 and 0.0167, respectively; netic modification efforts crucial for the improvement of this Fisher’s exact test, Bonferroni-corrected), as has been discovered long-life-cycle crop, to fight threatening avocado diseases, and to for other plant genomes, and involving many different gene optimize growth and desirable phenotypic traits. We anchored families and responses. Tandem O-methyltransferases homolo- almost half of the sequenced Hass genome to a genetic map, gous to COMT1 may also contribute to synthesis of the phe- providing linkage information for genetic variation on 12 chro- nylpropanoid derivative and insecticide estragole (56), which is mosomes. We resequenced 10 genomes representing small largely responsible for the anise-like leaf scent and fruit taste of populations of Guatemalan, West Indian, Mexican, and Hass- PLANT BIOLOGY many avocado cultivars, particularly of the Mexican race (57). related cultivars—and the genome of the closely related spe- Another relevant enriched GO category among tandems is cies P. schiedeana—in order to call SNPs and study genetic di- “ethylene-activated signaling pathway” (P = 0.000463; Fisher’s versity among these chromosomes. Analyses of admixture and exact test, Bonferroni-corrected), which annotates many differ- introgression clearly highlighted the hybrid origin of Hass avo- ent transcription factor duplicates. Ethylene signaling factors cado, pointed to its Mexican and Guatemalan progenitor races, such as ERF1 (represented by 2 homologs) are heavily involved and showed Hass to contain Guatemalan introgression in ap- in pathogen-induced responses, including to infection by Colle- proximately one-third of its genome. Introgressed blocks of totrichum and other necrotrophic fungi (58–61). Also identified chromosome arm size matched expectation based on Hass’s re- are 3 homologs of EIN3, a transcription factor that initiates cent (20th century) origin. We uncovered 2 ancient polyploidy downstream ethylene responses, including fruit ripening (62). events that occurred in the lineage leading to avocado and Avocado fruit matures on the tree in a process that involves conclude that these were independent from genome duplications ethylene synthesis and signaling, while it does not ripen until or triplications known to have occurred in other angiosperm harvested—a desirable trait that allows growers to delay har- clades. We contributed to solving the problem of magnoliid vesting for several months (63). phylogenetic relationships to other major angiosperm clades by Given the ancient derivation of avocado’s retained polyploid showing that thousands of syntenic orthologs among 14 species duplicates, most tandem duplicates in the genome are expected support an arrangement wherein the magnoliid clade branched to be of more recent origin, having been generated by ongoing off before the split between monocots and eudicots. However, gene birth–death–innovation processes that operate in all this resolution is tentative, with coding sequence phylogenomics eukaryotic genomes. As such, sub- or neofunctionalized tandem inconclusive and gene family birth/death analysis suggesting ap- duplicates that survive the usual fate of duplicated genes— preciable duplicate gene turnover—and therefore enhanced pseudogenization—should be enriched in functions that fine- possibility for ILS—during what appears to have been a nearly tune a given species’ recent selective environment. In the case coincident radiation of the major angiosperm clades. We also of avocado, response to fungal pathogens is precisely reflected in studied the adaptive landscape of the avocado genome through its tandemly duplicated gene complement. functional enrichment analyses of its mechanistically distinct duplicate gene collections, that is, tandem versus polyploid du- Differential Expression of Tandem versus Polyploid Duplicates. Fol- plicates. Tandem duplicates were enriched with many potentially lowing our prediction that many tandem duplicates fixed in the important metabolic responses that may include relatively recent avocado genome may have evolved under relatively recent adaptation against fungal pathogens. In contrast, ancient poly- pathogen pressure, we examined differential expression of Hass ploid duplicates, which originated in 2 distinct waves, were genes after treatment with the anthracnose causal agent (64) (SI enriched with transcriptional regulatory functions reflective of Appendix, section 9). Hass transcriptome reads for untreated core physiological and developmental processes. We discovered control versus pathogen-treated were mapped to Hass gene that tandem duplicates were more dynamically transcribed fol- models using Kallisto (65), normalized to transcript-per-million lowing anthracnose infection, and that some of the up-regulated values and thresholded by identifying genes with treatment/ genes could be related to defense responses. In sum, our work control log2 fold change outside of the [2, −2] interval. Tandems paves the way for genomics-assisted avocado improvement (1). were significantly enriched among up-regulated (P = 3.536e-09; Fisher’s exact test) and down-regulated genes (P = 7.274e-07), Data Availability. Bioproject: PRJNA508502. Biosamples: whereas polyploid duplicates did not show enrichment (SI Ap- SAMN10523735, SAMN10523720, SAMN10523736, SAMN10523738, pendix, section 9). We interpret these results to indicate that SAMN10523739, SAMN10523746, SAMN10523747, SAMN10523748; tandem duplicates are the most dynamic component of the av- SAMN10523749, SAMN10523750, SAMN10523752, SAMN10523753, ocado duplicate gene space under pathogen treatment. and SAMN10523756. SRA submission: SUB4878870. Whole We also examined functional enrichments for up- versus down- Genome Shotgun projects have been deposited at DDBJ/ENA/ regulated tandem duplicates (Dataset S16). The only significantly GenBank under the accession nos. SDXN00000000 and enriched category was xyloglucan:xyloglucosyl transferase activity SDSS00000000. The versions described in this paper are versions Rendón-Anaya et al. PNAS | August 20, 2019 | vol. 116 | no. 34 | 17087 SDXN01000000 and SDSS01000000 (P. americana var. drymifolia National Center for Genome Resources for PacBio single-molecule real-time and P. americana cultivar Hass, respectively). The genome assem- sequencing. A P. americana var. drymifolia reference individual was se- blies and annotations are available at https://genomevolution.org/ quenced using different fragment-size libraries (∼0.5, 1, 3, 5, or 8 kb). For CoGe/SearchResults.pl?s=29305&p=genome and https:// detailed information about assembly, annotation, and other bioinformatic genomevolution.org/coge/SearchResults.pl?s=29302&p=genome analysis see SI Appendix. (P. americana var. drymifolia and P. americana cultivar Hass, respectively). ACKNOWLEDGMENTS. This project was funded in large part by Grant 00126261from the Secretaria de Agrícultura, Ganadería, Recursos Pesqueros y Alimentos/Consejo Nacional de Ciencia y Tecnología sectorial program to Materials and Methods L.H.-E.; Grant 05-2018 from the Governor University Research Initiative pro- P. americana var. drymifolia was obtained from the germplasm bank of the gram from the state of Texas; Howard Hughes Medical Institute Grant Instituto Nacional de Investigaciones Forestales y Agropecuarias in Uruapan 55005946 to L.H.-E.; Grants 0922742 and 1442190 to V.A.A., N.M., and and the Hass and Carmen Hass cultivars were collected from a commercial A.H. from the National Science Foundation; Horticulture Innovation Aus- orchard in Tingambato, both in Michoacán, Mexico. The remaining rese- tralia Ltd; and the Australian Bureau of Agricultural and Resource Econom- quenced accessions were obtained from the Fundación Salvador Sánchez ics and Sciences. We thank the Fundación Salvador Sánchez Colín–Centro de Colín germplasm bank located at La Cruz Experimental Center at Coatepec Investigaciones Científicas y Tecnológicas de Aguacate en el Estado de Méx- Harinas in the state of Mexico. The Velvick rootstock was provided by the ico, S.C. for providing avocado specimens. We also thank Araceli Fernandez University of Queensland, Australia. DNA was extracted from young leaves and Emanuel Villafan, administrators of the high-performance computing of single individuals for all cultivars sequenced. For the reference genome of systems at Laboratorio Nacional de Genomica para la Biodiversidad and the Hass cultivar, high-quality megabase-sized DNA was submitted to the Instituto Nacional de Ecología, respectively. 1. A. S. Chanderbali et al., Persea americana (avocado): Bringing ancient flowers to fruit 28. M. J. Moore, C. D. Bell, P. S. Soltis, D. E. Soltis, Using plastid genome-scale data to in the genomics era. BioEssays 30, 386–396 (2008). resolve enigmatic relationships among basal angiosperms. Proc. Natl. Acad. Sci. U.S.A. 2. M. J. Christenhusz, J. W. Byng, The number of known plants species in the world and 104, 19363–19368 (2007). its annual increase. Phytotaxa 261, 201–217 (2016). 29. D. E. Soltis et al., Angiosperm phylogeny: 17 genes, 640 taxa. Am. J. Bot. 98, 704–730 3. Statista Research Department, Avocado industry–Statistics & facts. https://www.statista. (2011). com/topics/3108/avocadoindustry/. Accessed 29 July 2019. 30. D. E. Soltis et al., Angiosperm phylogeny inferred from 18S rDNA, rbcL, and atpB 4. Fresh Plaza, Mexico: Avocado exports generate 2.5 billion dollars. https://www. sequences. Bot. J. Linn. Soc. 133, 381–461 (2000). freshplaza.com/article/178230/Mexico-Avocado-exportsgenerate-2.5-billion-dollars/. 31. N. J. Wickett et al., Phylotranscriptomic analysis of the origin and early diversification Accessed 29 July 2019. of land plants. Proc. Natl. Acad. Sci. U.S.A. 111, E4859–E4868 (2014). 5. M. E. Galindo-Tovar, A. M. Arzate-Fernández, N. Ogata-Aguilar, I. Landero-Torres, 32. D. Sankoff, C. Zheng, E. Lyons, H. Tang, “The trees in the peaks” in International The avocado (Persea americana, Lauraceae) crop in Mesoamerica: 10,000 years of Conference on Algorithms for Computational Biology, I. Holmes, C. Martin-Vide, history. Harv. Pap. Bot. 12, 325–334 (2007). M. A. Vega-Rodríguez, Eds. (Springer, 2016), pp. 3–14. 6. B. A. Schaffer, B. N. Wolstenholme, A. W. Whiley, The Avocado: Botany, Production 33. D. Sankoff et al., Models for similarity distributions of syntenic homologs and appli- and Uses (CABI, 2013). cations to phylogenomics. IEEE/ACM Trans. Comput. Biol. Bioinformatics 10.1109/ 7. H. Chen, P. L. Morrell, V. E. Ashworth, M. de la Cruz, M. T. Clegg, Tracing the geo- TCBB.2018.2849377 (2018). graphic origins of major avocado cultivars. J. Hered. 100, 56–65 (2009). 34. L. Zeng et al., Resolution of deep angiosperm phylogeny using conserved nuclear 8. S. Willingham et al., Rootstock influences postharvest anthracnose development genes and estimates of early divergence times. Nat. Commun. 5, 4956 (2014). in‘Hass’ avocado. Aust. J. Agric. Res. 52, 1017–1022 (2001). 35. A. Hobolth, O. F. Christensen, T. Mailund, M. H. Schierup, Genomic relationships and 9. C. Illsley-Granich, R. Brokaw, S. Ochoa-Ascencio, “Hass Carmen, a precocious flower- speciation times of human, chimpanzee, and gorilla inferred from a coalescent hid- ing avocado tree” in Proceedings VII World Avocado Congress (2011), pp. 5–9. den Markov model. PLoS Genet. 3, e7 (2007). 10. G. Furnier, M. Cummings, M. Clegg, Evolution of the avocados as revealed by DNA 36. D. A. Pollard, V. N. Iyer, A. M. Moses, M. B. Eisen, Widespread discordance of gene restriction fragment variation. J. Hered. 81, 183–188 (1990). trees with species tree in Drosophila: Evidence for incomplete lineage sorting. PLoS 11. F. A. Simão, R. M. Waterhouse, P. Ioannidis, E. V. Kriventseva, E. M. Zdobnov, BUSCO: Genet. 2, e173 (2006). Assessing genome assembly and annotation completeness with single-copy orthologs. 37. J. B. Whitfield, P. J. Lockhart, Deciphering ancient rapid radiations. Trends Ecol. Evol. Bioinformatics 31, 3210–3212 (2015). 22, 258–265 (2007). 12. D. Kuhn et al., Application of genomic tools to avocado (Persea americana) breeding: 38. P. Librado, F. G. Vieira, J. Rozas, BadiRate: Estimating family turnover rates by SNP discovery for genotyping and germplasm characterization. Sci. Hortic. (Amsterdam) likelihood-based methods. Bioinformatics 28, 279–281 (2012). 246, 1–11 (2019). 39. D. Charlesworth, Don’t Forget the Ancestral Polymorphisms (Nature Publishing 13. T. H. Lee, H. Guo, X. Wang, C. Kim, A. H. Paterson, SNPhylo: A pipeline to construct a Group, 2010). phylogenetic tree from huge SNP data. BMC Genomics 15, 162 (2014). 40. G. I. Peterson, J. Masel, Quantitative prediction of molecular clock and ka/ks at short 14. L. Skotte, T. S. Korneliussen, A. Albrechtsen, Estimating individual admixture pro- timescales. Mol. Biol. Evol. 26, 2595–2603 (2009). portions from next generation sequencing data. Genetics 195, 693–702 (2013). 41. M. D. Rasmussen, M. Kellis, Unified modeling of gene duplication, loss, and co- 15. X. Zheng, B. S. Weir, Eigenanalysis of SNP data with an identity by descent in- alescence using a locus tree. Genome Res. 22, 755–765 (2012). terpretation. Theor. Popul. Biol. 107, 65–76 (2016). 42. A. Suh, L. Smeds, H. Ellegren, The dynamics of incomplete lineage sorting across the 16. S. H. Martin, J. W. Davey, C. D. Jiggins, Evaluating the use of ABBA-BABA statistics to ancient adaptive radiation of neoavian birds. PLoS Biol. 13, e1002224 (2015). locate introgressed loci. Mol. Biol. Evol. 32, 244–257 (2015). 43. M. Lynch, The Origins of Genome Architecture (Sinauer Associates, Sunderland, MA, 17. M. Nei, W.-H. Li, Mathematical model for studying genetic variation in terms of re- 2007), vol. 98. striction endonucleases. Proc. Natl. Acad. Sci. U.S.A. 76, 5269–5273 (1979). 44. A. Force et al., Preservation of duplicate genes by complementary, degenerative 18. G. A. Watterson, On the number of segregating sites in genetical models without recombination. Theor. Popul. Biol. 7, 256–276 (1975). mutations. Genetics 151, 1531–1545 (1999). 19. K. E. Holsinger, B. S. Weir, Genetics in geographically structured populations: De- 45. M. Freeling, Bias in plant gene content following different sorts of duplication: fining, estimating and interpreting F(ST). Nat. Rev. Genet. 10, 639–650 (2009). Tandem, whole-genome, segmental, or by transposition. Annu. Rev. Plant Biol. 60, 20. F. Tajima, Statistical method for testing the neutral mutation hypothesis by DNA 433–453 (2009). polymorphism. Genetics 123, 585–595 (1989). 46. F. Cheng et al., Gene retention, fractionation and subgenome differences in polyploid 21. E. Lyons, B. Pedersen, J. Kane, M. Freeling, The value of nonmodel genomes and an plants. Nat. Plants 4, 258–268 (2018). example using SynMap within CoGe to dissect the hexaploidy that predates the ro- 47. J. Salojärvi et al., Genome sequencing and population genomic analyses provide in- sids. Trop. Plant Biol. 1, 181–190 (2008). sights into the adaptive landscape of silver birch. Nat. Genet. 49, 904–912 (2017). 22. V. A. Albert et al.; Amborella Genome Project, The Amborella genome and the 48. D. Beno-Moualem, D. Prusky, Early events during quiescent infection development by evolution of flowering plants. Science 342, 1241089 (2013). Colletotrichum gloeosporioides in unripe avocado fruits. Phytopathology 90, 553–559 23. H. Tang et al., Unraveling ancient hexaploidy through multiply-aligned angiosperm (2000). gene maps. Genome Res. 18, 1944–1954 (2008). 49. C. H. Acosta-Muñiz et al., Identification of avocado (Persea americana) root proteins 24. Y. Jiao et al., A genome triplication associated with early diversification of the core induced by infection with the oomycete Phytophthora cinnamomi using a proteomic eudicots. Genome Biol. 13, R3 (2012). approach. Physiol. Plant. 144, 59–72 (2012). 25. O. Jaillon et al., The grapevine genome sequence suggests ancestral hexaploidization 50. J. Engelbrecht, N. Van den Berg, Expression of defence-related genes against Phy- in major angiosperm phyla. Nature 449, 463–467 (2007). tophthora cinnamomi in five avocado rootstocks. S. Afr. J. Sci. 109, 1–8 (2013). 26. C. Zheng, D. Santos Muñoz, V. A. Albert, D. Sankoff, Syntenic block overlap multi- 51. E. Luna et al., Callose deposition: A multifaceted plant defense response. Mol. Plant plicities with a panel of reference genomes provide a signature of ancient poly- Microbe Interact. 24, 183–193 (2011). ploidization events. BMC Genomics 16 (suppl. 10), S8 (2015). 52. L. Eshraghi et al., Phosphite primed defence responses and enhanced expression of 27. C. Zheng, E. Chen, V. A. Albert, E. Lyons, D. Sankoff, Ancient eudicot hexaploidy defence genes in Arabidopsis thaliana infected with Phytophthora cinnamomi. Plant meets ancestral eurosid gene order. BMC Genomics 14 (suppl. 7), S3 (2013). Pathol. 60, 1086–1095 (2011). 17088 | www.pnas.org/cgi/doi/10.1073/pnas.1822129116 Rendón-Anaya et al. 53. N. v. d. Berg, J. B. Christie, T. A. S. Aveling, J. Engelbrecht, Callose and β-1,3-glucanase 61. W. F. Broekaert, S. L. Delauré, M. F. De Bolle, B. P. Cammue, The role of ethylene in inhibit Phytophthora cinnamomi in a resistant avocado rootstock. Plant Pathol. 67, host-pathogen interactions. Annu. Rev. Phytopathol. 44, 393–416 (2006). 1150–1160 (2018). 62. P. Lü et al., Genome encode analyses reveal the basis of convergent evolution of 54. J. Le Roy, B. Huss, A. Creach, S. Hawkins, G. Neutelings, Glycosylation is a major fleshy fruit ripening. Nat. Plants 4, 784–791 (2018). regulator of phenylpropanoid availability and biological activity in plants. Front. 63. E. M. Yahia, A. B. Woolf, “Avocado (Persea americana Mill.)” in Postharvest Biology Plant Sci. 7, 735 (2016). and Technology of Tropical and Subtropical Fruits: Açai to Citrus, E. Yahia, Ed. 55. N. D. Bonawitz, C. Chapple, The genetics of lignin biosynthesis: Connecting genotype (Woodhead Publishing, 2011), pp. 125–186e. to phenotype. Annu. Rev. Genet. 44, 337–363 (2010). 64. L.-Á. Xoca-Orozco et al., Transcriptomic analysis of avocado hass (Persea americana Mill) 56. C. L. Chang, I. K. Cho, Q. X. Li, Insecticidal activity of basil oil, trans-anethole, estra- in the interaction system fruit-chitosan-Colletotrichum. Front. Plant Sci. 8, 956 (2017). gole, and linalool to adult fruit flies of Ceratitis capitata, Bactrocera dorsalis, and 65. N. L. Bray, H. Pimentel, P. Melsted, L. Pachter, Near-optimal probabilistic RNA-seq Bactrocera cucurbitae. J. Econ. Entomol. 102, 203–209 (2009). quantification. Nat. Biotechnol. 34, 525–527 (2016). 57. J. R. King, R. J. Knight, Occurrence and assay of estragole in the leaves of various 66. W. Xu et al., Arabidopsis TCH4, regulated by hormones and the environment, encodes avocado cultivars. J. Agric. Food Chem. 35, 842–844 (1987). a xyloglucan endotransglycosylase. Plant Cell 7, 1555–1567 (1995). 58. X. Zhu et al., The wheat ethylene response factor transcription factor pathogen- 67. Y. Hu et al., Comparative transcriptome analysis unveils the tolerance mechanisms of induced ERF1 mediates host responses to both the necrotrophic pathogen Rhizoc- Citrus hystrix in response to ‘Candidatus Liberibacter asiaticus’ infection. PLoS One 12, tonia cerealis and freezing stresses. Plant Physiol. 164, 1499–1514 (2014). e0189229 (2017). 59. L. Xing et al., Overexpression of ERF1-V from Haynaldia villosa can enhance the re- 68. N. H. Davila Olivas et al., Effect of prior drought and pathogen stress on Arabidopsis sistance of wheat to powdery mildew and increase the tolerance to salt and drought transcriptome changes to caterpillar herbivory. New Phytol. 210, 1344–1356 (2016). stresses. Front. Plant Sci. 8, 1948 (2017). 69. R. H. Tomas-Grau et al., Soft mechanical stimulation induces a defense response 60. I. Diniz et al., A first insight into the involvement of phytohormones pathways in against Botrytis cinerea in strawberry. Plant Cell Rep. 37, 239–250 (2018). coffee resistance and susceptibility to Colletotrichum kahawae. PLoS One 12, 70. X. Wang et al., IDL6-HAE/HSL2 impacts pectin degradation and resistance to Pseudo- e0178159 (2017). monas syringae pv tomato DC3000 in Arabidopsis leaves. Plant J. 89, 250–263 (2017). PLANT BIOLOGY Rendón-Anaya et al. PNAS | August 20, 2019 | vol. 116 | no. 34 | 17089