The frequency and importance of polyploidy in tropical rainforest tree radiations

1039 Number of Supporting Information Files 3

1576

618

2979 23 24 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 2 Summary 25 • The presence of two or more copies of the genome in an organism, termed ‘polyploidy’, is a 26 crucial force in plant evolution, generating genetic, phenotypic and ecological diversity. The 27 Amazonian tree flora is the most species-rich on Earth, and largely arose as a result of rapid 28 evolutionary radiations. While polyploidy is an important catalyst of rapid radiations, it 29 remains poorly studied in tropical tree radiations. 30 • We examined ploidy variation across Inga (Fabaceae), a characteristic Amazonian tree 31 radiation, using DNA sequence data from 1305 loci for 189/282 Inga species. We then tested 32 whether polyploid species experience more positive selection than diploids, particularly in 33 loci underlying chemical defence against herbivory, which is a key ecological pressure 34 affecting rainforest tree diversification. 35 • We show that tetraploidy occurs in 14% (N=27) of the Inga species we sequenced, with 36 several widespread species showing geographical ploidy variation, alongside minimal 37 phylogenetic signal in ploidy which suggests recurrent polyploidisation. Interestingly, we 38 found more loci under selection in polyploids than diploids, most notably in chemical defence 39 loci. 40 • Our results show that polyploidy has arisen independently in several Inga species, and that 41 polyploidisation can lead to elevated selection in chemical defence, helping to shape 42 ecological interactions and influence diversification in Inga. 43 44 45

Radiation; Diversification; Phylogenomics; Amazon; Rainforest; Fabaceae; Selection; 46 Whole genome duplication 47 48 49

50 Polyploidy, whole genome duplication that results in the presence of more than two sets of 51 chromosomes in an organism, is a central force in plant evolution. All angiosperm species have 52 experienced at least one round of historical polyploidy, while approximately 35% of angiosperm 53 species are more recent polyploids based on their chromosome counts (Wood et al., 2009). Polyploidy 54 can catalyse rapid speciation by promoting instantaneous reproductive isolation between a new 55 polyploid lineage and its diploid progenitor(s) (Coyne & Orr 2004; Van de Peer et al., 2017; though 56 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 3 see Brown et al., 2024), by increasing the likelihood of chromosome mismatching and resultant 57 sterility of hybrid offspring (Wood et al., 2009). Polyploidy also has the potential to generate 58 evolutionary novelty through gene duplication and subsequent biased retention of adaptive genetic 59 variation (Flagel & Wendel 2009; Birchler & Yang 2022). Specifically, genes may take on new fates 60 (neofunctionalisation) or partition previous fates (subfunctionalisation) following duplication that 61

from polyploidy (Flagel & Wendel 2009). Polyploidisation may involve hybridisation 62 (allopolyploidy), or occur within a species (autopolyploidy), and an increasing body of ecological 63 work shows that polyploids may have increased adaptive potential, allowing them to occupy new 64 niche space driven by physiological or anatomical trait differences relative to their parental 65 progenitors (Van de Peer et al., 2017). 66 While polyploidy is well-established as a driver of plant species diversification in temperate floras, 67 there is less evidence for polyploidy in the tropics (Rice et al., 2019), particularly in tropical 68 rainforests. It is unclear whether the scarcity of observed polyploidy in tropical plants is due to a lack 69 of data or biological differences in the propensity for polyploid formation between the tropical and 70 temperate zones. The species richness and geographical remoteness of many tropical rainforest 71 environments has meant that, in the past, generating representative datasets to identify the number of 72 polyploid origins with certainty was extremely difficult. A major challenge for inferring polyploid 73 origins for tropical tree clades was the lack of well-resolved phylogenies for these species-rich groups, 74 because many of these groups arose through rapid speciation that made phylogenetic inference with 75 single loci difficult (Koenen et al., 2015). However, with the advent of genomic tools such as target-76 capture (Gnirke et al., 2009; Andermann et al., 2020) that allow inference of well-resolved 77 phylogenetic trees for these speciose groups (such as Inga (Nicholls et al., 2015; Schley et al., 2025), 78 alongside novel methods that use target-capture sequencing data to estimate ploidy bioinformatically 79 (e.g. nQuack (Gaynor et al., 2024)), there is now scope for addressing key questions about ploidy 80 evolution in speciose tropical plant groups at scale. 81 The paucity of studies aiming to understand ploidy evolution in species-rich tropical rainforest trees is 82 a prominent knowledge gap because tropical rainforests, and in particular the Amazon, have more tree 83 species than anywhere else on Earth (Ulloa Ulloa et al., 2017; Cardoso et al., 2017). Exactly how this 84 species-richness arose remains enigmatic, but there is mounting evidence that much of Amazonia’s 85 tree diversity arose through rapid evolutionary radiations that gave rise to species-rich tree genera 86 (e.g. Inga (Richardson et al., 2001), Trichillia (Koenen et al., 2015), Guatteria (Erkens et al., 2007)). 87 Polyploidy, in concert with related phenomena such as hybridisation, is well-known as a catalyst of 88 rapid evolutionary radiations (Barrier et al., 1999; Landis et al., 2018; Meudt et al., 2021), akin to 89 those that generated much of Amazonia’s tree diversity. However, there are no case studies of which 90 we are aware that examine whether polyploidy has influenced rapid radiation in the Amazonian flora. 91 The genus Inga (Fabaceae) exemplifies the rapid evolutionary radiations that gave rise to much of 92 Amazonia’s tree diversity, with >300 species that have diversified in the last 5-7Ma (Pennington, 93 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 4 1997; Richardson et al., 2001; Ringelberg et al., 2023). Most Inga species are thought to be 94 contemporary diploids, evidenced by a handful of diploid chromosome counts (2n = 26), with higher 95 chromosome counts in a few Inga species suggesting that tetraploidy does occur (2n = 4x = 52) 96 (Hanson, 1995; Figueiredo et al., 2014). Moreover, the legume family within which Inga is nested is 97 of ancient polyploid origin (Koenen et al., 2021), and introgressive hybridisation (that can lead to 98 allopolyploidy) is widespread in Inga (Schley et al., 2025). 99 One aspect by which polyploidy may fuel adaptation and diversification in Inga is by generating 100 diversity in chemical defences against herbivory. Rainforest trees are subject to high insect herbivory 101 pressure (Kursar et al., 2009; Forrister et al., 2019), and so herbivory is likely to be a powerful 102 selective factor influencing speciation in rainforests (Coley et al., 2018). Phylogenetic work shows 103 divergence in chemical defences between species in Inga (Kursar et al., 2009), with close relatives 104 often differing greatly in their chemical defences (Forrister et al., 2023). Negative frequency-105 dependent processes in rainforest communities are driven by herbivory (Janzen, 1970; Connell, 1971; 106 Forrister et al., 2019), implying a selective advantage for novel, rare defences that allow escape from 107 local herbivores. Genome duplication has been suggested to influence chemical defence evolution and 108 diversification in temperate herbs (e.g. in the Brassicaceae-Pieris butterfly ‘chemical arms race’ 109 (Edger et al., 2015)), and these duplication events are thought to be a common source of evolutionary 110 novelty in defence compounds (Ober, 2010; Moore et al., 2014; Endara et al., 2023). This raises the 111 prospect that novelty in chemical defence fuelled by polyploidy may have played a role in Inga’s 112 diversification. 113 Here, we use target-capture data at a phylogenetic scale for Inga to investigate polyploidy in tropical 114 trees, where it has been poorly studied despite its demonstrated importance in plant evolution. This 115 will help us to understand the evolution of Amazonian tree diversity more broadly, because Inga 116 exemplifies the species-rich tree genera that underwent similar recent radiations and make up the bulk 117 of Amazonian tree species. We estimate contemporary ploidy levels across Inga, following which we 118 assess whether putative polyploid lineages are the result of allo- or auto-tetraploidy events. Finally, 119 we assess whether polyploidisation is associated with innovation in biosynthetic genes that underlie 120 chemical defence in Inga, evidenced by more selection in those loci. 121 Accordingly, here we ask three key questions: 122 123 1) What is the distribution of polyploidy across the radiation of Inga? 124 2) What proportion of polyploids are allo-or-autopolyploids? 125 3) Is polyploidy associated with elevated selection on chemical defences? 126 127 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 5

128 Target-Capture DNA Sequencing and Phylogenomic Analysis 129 We analysed target capture DNA sequencing data from 189 of 282 accepted Inga species (67%), most 130 of which were taken from previous studies (Nicholls et al., 2015; Schley et al., 2025), comprising one 131 accession for each of the 189 species. We then sequenced an additional seven accessions from five of 132 these Inga species (Inga capitata, Inga cylindrica, Inga heterophylla, Inga laurina and Inga striata), 133 collected from different geographical regions, because they displayed chromosome counts that 134 conflicted with our preliminary ploidy estimation (Hanson, 1995; Figueiredo et al., 2014). These extra 135 accessions were sequenced using the same protocol as Schley et al. (2025), detailed below. Sampling 136 was based on a taxonomically-verified list of accepted Inga species (WCVP, 2020) and included an 137 unbiased selection of species that span the Inga phylogenetic tree, the remainder of which were either 138 too rare or too degraded to source collections for sequencing. We also sampled one individual from 139 Inga’s sister genus Zygia from Nicholls et al. (2015) as an outgroup. Accession information, including 140 species, sampling location, collector and data source is detailed in Table S1, Supporting Information. 141 In-depth details of DNA sequencing, assembly, alignment and phylogenetic inference are described in 142 Schley et al. (2025). Briefly, DNA library preparation, enrichment and sequencing were carried out 143 either by Arbor BioSciences (Ann Arbor, MI, USA) or the University of Exeter sequencing service 144 (Exeter, UK) with the NEBnext Ultra II FS protocol (New England Biolabs, Ipswich, MA, USA). 145 Targeted bait capture was performed using the ‘Mimobaits’ bait set (Nicholls et al., 2015; Koenen et 146 al., 2020) with the MyBaits protocol v.2 and 3 (Arbor Biosciences, Ann Arbor, MI, USA). The 147 Mimobaits set targets 1320 loci specific to the Mimosoideae subfamily within which Inga is nested, 148 including 113 genes underlying anti-herbivore defence chemistry in Inga (hereafter ‘defence 149 chemistry’ loci). The Mimobaits set additionally targets 1044 ‘single-copy phylogenetically 150 informative’ loci, 109 ‘differentially expressed’ loci and 54 un-annotated ‘miscellaneous’ loci 151 showing phylogenetic signal (further details in Nicholls et al. (2015) and Schley et al. (2025)). 152 Enriched libraries were sequenced using the NovaSeq 6000 platform with a paired-end 150bp run. 153 DNA sequencing reads were quality-checked with FASTQC 0.11.3 (Andrews, 2010), adapters were 154 removed and bases were filtered using TRIMMOMATIC 0.3.6 (Bolger et al., 2014) (< 2 mismatches, 155 palindrome clip threshold = 30, simple clip threshold = 10, quality score threshold < 28), with reads < 156 36bp removed. Filtered reads were then assembled into target loci with HYBPIPER v.1.2 (Johnson et 157 al., 2016) with a minimum coverage cut-off of 5×. Targeted loci were then aligned by locus using 158 1000 iterations in MAFFT 7.453 (Katoh & Standley 2013) with the ‘—adjustdirectionaccurately’ 159 flag, and were cleaned using the ‘-automated1’ flag in trimAl 1.3 (Capella-Gutiérrez et al., 2009), 160 resulting in 1305 refined alignments for Inga. Alignment summaries detailing proportions of variable 161 sites and missing data are presented in Schley et al. (2025). Gene trees were inferred for each locus 162 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 6 alignment using IQ-TREE (Nguyen et al., 2015) by selecting the best-fit substitution model (-MFP) 163 while reducing the impact of severe model violations (-bnni) with 1000 ultrafast bootstrap replicates. 164 Following this, a ‘species tree’ for Inga was inferred with the best-scoring IQtrees using ASTRALMP 165 5.15.5 under the default parameters (Zhang et al., 2018). 166 167 Estimating Ploidy at Phylogenetic Scale 168 All analyses in this study were conducted on the UK Crop Diversity Bioinformatics HPC Resource 169 (Percival/i1 Alwyn et al., 2025). We estimated the ploidies of all sequenced Inga accessions using 170 NQUACK (Gaynor et al., 2024). NQUACK improves on the model selection tools implemented in 171 previous ploidy estimation tools (e.g. NQUIRE (Weiß et al., 2018)) to predict ploidy more accurately, 172 as well as providing tools to filter sequence data further before ploidy estimation. We used NQUACK’s 173 prepare_data function to convert BAMs from HybPiper into NQUACK text files, following which we 174 used the process_data function to convert to a count of allelic depths. For this we excluded sites with 175 a minimum depth of 10, an estimated sequencing error rate of 0.01, with allele frequency truncation 176 between 0.15 and 0.85 (to remove sites most likely attributed to noise) and used no maximum depth 177 filter, following the suggestions of the NQUACK tutorial 178 https://mlgaynor.com/nQuack/articles/BasicExample.html. Then, we tested many different models 179 under different allelic ratio distributions using quackNormal, quackBeta and quackBetaBinom 180 commands, following which we selected the best model using the quackit function, based on that with 181 the lowest Bayesian Information Criterion (BIC). To corroborate ploidy estimations, we also ran 1000 182 bootstrapping replicates under the best model for each species, testing between the ‘diploid’, ‘triploid’ 183 and ‘tetraploid’ mixtures in each bootstrap replicate. 184 185 Assessing allopolyploidy vs autopolyploidy 186 We used NQUACK to estimate whether the inferred tetraploid Inga species in our target capture dataset 187 were allotetraploid, resulting from hybridisation, or autotetraploid, resulting from whole genome 188 duplication within a species. To do this, we estimated alpha values (i.e., the proportion of sites with a 189 certain allelic ratio) for allelic ratios of 0.25 (corresponding to a tetraploid genotype of aaab), 0.5 190 (aabb) and 0.75 (abbb). Estimated alpha values can indicate the mode of inheritance, which is directly 191 related to mode of polyploidy (i.e. disomic inheritance in allotetraploids vs tetrasomic inheritance in 192 autotetraploids). Specifically, with tetratomic inheritance, we expect that the model will assign 193 relatively equal proportions (i.e. alpha values) to each class of heterozygotes (aaab, aabb, abbb) 194 resulting in a tri-modal distribution that indicates an autopolyploid. In contrast, we expect a higher 195 proportion at 0.5 (aabb genotypes) in an allopolyploid due to disomic inheritance of two divergent 196 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 7 subgenomes. To estimate alpha, we ran the expectation maximization algorithm with either the 197 normal-uniform distribution (with the emstepNUA function) or the beta-uniform distribution (with the 198 emstepBU function), based on which distribution was inferred to be the best-fit to the data in the 199 initial NQUACK polyploidy estimation. We used the default starting parameters for each parameter 200 (avec = 0.3, 0.3, 0.3, 0.1; mvec = 0.25, 0.50, 0.75; svec = 0.01, 0.01, 0.01), and plotted the allelic 201 ratios for each sample using the R function hist(). We assigned a species as allopolyploid if its alpha 202 value at an allelic ratio of 0.5 was larger than its alpha values at both 0.25 and 0.75. 203 204 Modelling Phylogenetic Signal of Ploidy 205 We assessed whether the modes of ploidy we inferred showed phylogenetic signal across our Inga 206 phylogenetic tree, which could indicate shared polyploidy events between related taxa. To do this, we 207 used the phylo.D() (Fritz & Purvis 2010) function in the R (R Development Core Team, 2013) 208 package caper (Orme et al., 2013), estimating Fritz’ D statistic as a measure of phylogenetic signal 209 with 10000 permutations to assess the significance of observed D values. D values closer to 1 indicate 210 the random distribution of a trait across the phylogeny, with values closer to 0 indicating a trait 211 evolves by Brownian motion across the tree. Similarly, values of D 1 suggest overdispersion of a trait. Following this, we plotted our inferred ploidies from 214 NQUACK onto our Inga phylogenetic tree in Phytools (Revell, 2012). 215 216 Testing for Selection in Diploids vs. Polyploids 217 Polyploidy can greatly influence the outcomes of selection through increasing available genetic 218 variation. Therefore, we tested whether each of our target-capture loci experienced positive selection 219 (i.e., more non-synonymous nucleotide changes than synonymous changes) on at least one branch and 220 at least one site using BUSTED (Murrell et al., 2015), comparing between tetraploid and diploid 221 species inferred with NQUACK. As input for BUSTED, we used previously prepared codon-aware 222 alignments for 1207 of the Mimobaits loci from Schley et al. (2025), prepared using OMM_MACSE 223 (Ranwez et al., 2011; Ranwez et al., 2021). These alignments contain the same samples from Schley 224 et al. (2025) as were used above to infer ploidy with NQUACK, excluding the newly sampled 225 accessions from this study. We used the ploidies inferred with NQUACK to run seven separate 226 BUSTED analyses: one where polyploid taxa (N=27) were set as ‘foreground’ taxa for selection 227 testing, and six further analyses containing between 27-28 diploid taxa each (randomly selected 228 without replacement) set as ‘foreground’ taxa. We used similar numbers of taxa across diploid and 229 tetraploid BUSTED runs to ensure valid comparisons between ploidy levels - BUSTED searches for 230 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 8 evidence of selection at any site in any foreground taxa in an alignment, meaning that alignments with 231 more foreground taxa have a higher chance of inferring selection. The taxa included in each run are 232 shown in Table S2 (Supporting Information). We accounted for false positives in our selection tests 233 by adjusting the P-values output by BUSTED with a 5% FDR (false discovery rate) in R (Benjamini 234 & Hochberg 1995). 235 Next, we used χ 2 tests in R to examine two hypotheses – firstly, whether the number of loci under 236 selection (i.e., loci with a BUSTED FDR P-value < 0.05) was associated with locus annotation in 237 each BUSTED run (i.e., for each of the six diploid runs and the run with only polyploids). The second 238 hypothesis was whether the number of loci under selection was associated with ploidy within each 239 locus annotation type (‘defence chemistry’, ‘differentially expressed’, ‘single-copy phylogenetically 240 informative’ and ‘miscellaneous’). We performed six separate tests between diploid taxa and 241 polyploid taxa – one for each of the six diploid BUSTED runs - against the same polyploid BUSTED 242 run each time. We also visualised the counts and percentages of loci under selection from each 243 BUSTED ploidy based on their locus annotation using boxplots in R. 244 245

246 Model comparison in NQUACK using BIC recovered a tetraploid model as the best fit for 27 of the 189 247 Inga accessions we tested (14.3%), where each accession belonged to a different Inga species (Table 248 S1, Supporting Information). Of these putative tetraploids, bootstrap resampling in NQUACK 249 recovered a tetraploid model most frequently (>600/1000 BS) in 15/27 accessions. The remainder 250 recovered either more bootstraps for the diploid model (seven accessions) or had similar numbers of 251 bootstraps for both tetraploid and diploid models (five accessions). In one accession, 252 Inga_killipiana_WFR_2627, the best model fit was that parameterising a diploid mixture, while 253 675/1000 bootstrap replicates recovered the tetraploid model. The best-fit NQUACK frequency 254 distribution was either beta-uniform or normal-uniform for all accessions. Model fits, ploidy 255 estimates, bootstrapping and alpha value estimates, in addition to the amount of missing data 256 recovered for each of the 189 Inga accessions in Schley et al. (2025), are available in Table S1, 257 Supporting Information. 258 In total, we recovered 11 putative allotetraploid accessions (Fig. 1) out of all 27 inferred tetraploids. 259 Interestingly, all accessions that alpha-value comparison recovered as putative allotetraploids (which 260 all had a tetraploid model as the best fit to their read data in the NQUACK BIC comparison) recovered 261 higher numbers of bootstraps fitting the diploid model, or fitting both the diploid and tetraploid 262 models at a similar proportion. This is reflected by their allele frequency spectra (Fig. S1, Supporting 263 Information). However, it is also worth noting that several of these putative allopolyploid accessions 264 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 9 also had noisy allele frequency spectra, particularly Inga_gereauana_VAS_14326 and 265 Inga_goldmanii_BCI_9620, which both had poor locus recovery in Schley et al. (2025). 266 We also recovered geographical ploidy variation in four out of the five widespread Inga species for 267 which we sampled more than one accession. We inferred one tetraploid and one diploid accession 268 from each of Inga heterophylla, Inga laurina and Inga capitata, as well as two diploid and one 269 tetraploid accession for Inga striata (Table S1, Supporting Information), with ploidy varying 270 depending on where accessions were collected. Interestingly, for Inga striata, the two diploid 271 accessions were both collected from southeastern Brazil. 272 273 Figure 1: Ploidy, estimated for each individual with nQuack, mapped on the phylogenetic tree of Inga. The 274 ASTRAL phylogenetic tree was inferred by previous work (Schley et al., 2025) based on 1305 loci from the 275 Mimobaits target bait capture set, and contains a single accession per species. Circles on species names indicate 276 ploidy of the accession, with green indicating diploidy and red indicating tetraploidy. Putative allopolyploids, 277 inferred using the distribution of allelic ratios and estimated using nQuack’s alpha parameter, are marked on the 278 tree with red arrows. Clades are annotated first by intrageneric subclade as in Schley et al. (2025), and then with 279 the broader clades within Inga s.s. in which they are nested (Redhair clade, Hairy clade, Fast clade). In 280 shortened subclade annotations, ‘Leiocal.’ = Leiocalycina subclade, ‘Poepp.’ = Poeppigiana subclade, ‘M. 281 grade’ = Microcalyx grade, ‘Umbel.’ = Umbellifera subclade. Exemplar line drawings, modified from 282 Pennington (1997), are shown for several species, with species inferred by nQuack to be tetraploid shown in red 283 text with a red dashed box around them. 284 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 285 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 11 The accessions (hereafter, representing species) that we inferred to be tetraploids were widely-spread 286 across the phylogenetic tree of Inga, with few closely related species being polyploids (Figure 1). Our 287 modelling of ploidy shifts across the Inga phylogenetic tree recovered weak evidence of phylogenetic 288 signal (D=0.834), which was neither significantly different from a random distribution of ploidy 289 across the phylogeny (P = 0.312) or a distribution resulting from Brownian motion (P = 0.089) (Table 290 S3, Supporting Information). 291 Within the tetraploid Inga species (N=27), BUSTED inferred positive selection in 63.463% of all loci, 292 whereas within the six subsets of diploid Inga species (N= 27-28 per set) BUSTED inferred positive 293 selection in between 43.910 - 62.054% of loci. Across these BUSTED analyses, more loci were under 294 selection in the tetraploid species than the diploid species for every locus annotation type (Figure 2; 295 Table S4, Supporting Information). This difference was particularly prominent in chemical defence 296 loci (Figure 2; Fig. S2, Supporting Information). Our χ 2 tests only showed a significant association 297 between selection result and locus annotation in the BUSTED run for diploid set 2 only (χ 2= 8.349, 298 df=3, P = 0.0039; Table S5, Supporting Information). Interestingly, we also found significant 299 associations between ploidy and the number of loci under selection, with the most prominent 300 relationship evident in the ‘Defence chemistry’ loci (maximum χ 2 = 17.431, df = 1, P = 2.979x10-5) 301 and ‘Single Copy Phylogenetically Informative’ loci (maximum χ 2 = 62.539, df = 1, P = 2.612x10-15) 302 (Table S6, Supporting Information). The ‘Differentially expressed’ loci also showed a significant 303 relationship between selection score and ploidy, but only in diploid sets 1-3 (Table S6, Supporting 304 Information). 305 306 Figure 2: Box plots indicating the percentages (top panel) and counts (bottom panel) of Inga target capture loci 307 inferred to be under selection by BUSTED ('under selection' if FDR-corrected P value <0.05) for the ‘tetraploid’ 308 and six randomised ‘diploid’ BUSTED runs, all of which comprised ca. 27 taxa. The x-axis indicates the target 309 capture locus annotation and bar colours indicate whether each counted locus was from the ‘polyploid’ or 310 ‘diploid’ BUSTED run. For the diploid box, the dark bar represents the median, the top and bottom edges of 311 each box represent the first and third quartiles, while the white circles represent outliers. The tetraploid point 312 and dotted line represents the number of loci under selection from the single polyploid BUSTED run. The total 313 number loci in each locus annotation class are indicated in italics on the x-axis beneath the Annotation name. 314 315 316 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 317 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 13

318 Our analyses recovered well-supported cases of polyploidy in Inga, a speciose tree radiation 319 characteristic of tropical American rainforests. Most of these events occurred independently, both as a 320

of whole-genome duplication within a species (leading to autopolyploids) and hybridisation 321 (leading to allopolyploids). Interestingly, we also recovered strong evidence for elevated selection in 322 polyploid Inga species relative to diploids, most notably within loci underlying chemical defence 323 against herbivory. Our study is, to our knowledge, the only one to have explored ploidy evolution at 324 detailed phylogenetic scale in a large rainforest tree radiation, having done so with an extensive target 325 capture dataset comprising >1300 loci sequenced for 189 Inga species. Our results show that 326 polyploidy has occurred recurrently across Inga over a broad geographic scale, with polyploids 327 experiencing significantly more selection than diploids, suggesting polyploidy may be more important 328 in tropical plant taxa than previously assumed. 329 330 Ploidy estimation and geographical ploidy variation 331 Our ploidy inference using NQUACK suggested that 27/189 sampled Inga species were tetraploids 332 (Table S1, Supporting Information). While we may have missed rare cytotypes in some species by 333 using the single-accession-per-species dataset of Schley et al. (2025), our NQUACK ploidy estimates 334 were congruent with 22 of 23 available chromosome counts for the focal Inga species in this study 335 (Hanson, 1995; Figueiredo et al., 2014) (Table S1, Supporting Information). Furthermore, our 336 NQUACK ploidy estimates for five widespread Inga species, for which we sequenced multiple 337 accessions, were consistent with observed geographical variation in chromosome count for four of 338 these species (Hanson, 1995; Figueiredo et al., 2014). This was with the exception of Inga cylindrica, 339 all accessions of which we inferred to be diploid (Table S1, Supporting Information). 340 One species that displayed geographical variation in ploidy was Inga laurina. The Ecuadorian 341 accession of I. laurina in this study (TI_1387) was inferred to be diploid, whereas the Brazilian 342 accession (JMF_1409) was inferred to be tetraploid (Table S1, Supporting Information). This is 343 congruent with previous work, which recovered chromosome counts from Brazilian accessions of I. 344 laurina of both 2n=26 and 2n=52 (Figueiredo et al., 2014). These counts correspond to diploid and 345 tetraploid individuals, respectively, because the base chromosome number in Inga is x=13 (Shibata, 346 1962; Hanson, 1995; Pennington, 1997). Our ploidy estimates for I. laurina are also congruent with 347 the recent autotetraploid reference genome that was sequenced for a Brazilian accession of I. laurina 348 (Schley et al., 2024). In a similar vein, the Brazilian accession of Inga capitata used in this study 349 (M54A) was inferred to be tetraploid, whereas the Ecuadorian accession (TI_687) was inferred to be 350 diploid. This agrees with Figueiredo et al. (2014), who recovered a chromosome count of 2n=52 for 351 Brazilian accessions of I. capitata. Moreover, our Brazilian accessions of Inga striata (FCPG_752, 352 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 14 LCS_627) and I. heterophylla (M_1399) were inferred to be diploid in our analysis, mirroring 353 previous chromosome counts of 2n=26 for Brazilian accessions of these species (Figueiredo et al., 354 2014), while we inferred our Peruvian accessions of both species to be tetraploid (Table S1, 355 Supporting Information). Only the three accessions of Inga cylindrica in our study (MB_ZNC 356 (Brazil), FG_35 (French Guiana) and TI_553 (Ecuador)) contrasted with previous chromosome 357 counts - we inferred all three accessions to be diploid, whereas Figueiredo et al. (2014) recovered a 358 chromosome count of 2n=52 for Brazilian accessions of this species. We consider it likely that I. 359 cylindrica shows intraspecific ploidy variation that was not detected in this study, likely because both 360 studies are based on different accessions. This requires confirmation with future work. 361 Geographical variation in ploidy is common in plant species (Suda et al., 2007; Kolář et al., 2017), 362 particularly in those with widespread distributions. Kolář et al. (2017) reviewed 69 studies of ploidy 363 variation within plant species, and found spatial segregation of ploidy within widespread species in 364 81% of the studies they reviewed. This is interesting, given that polyploids should be less likely to 365 become established due to ‘minority cytotype exclusion’, in which assortative mating within 366 cytotypes is selected for (hence preventing the generation of novel polyploids) due to reduced fitness 367 of offspring produced through inter-cytotype reproduction (Levin, 1975; Felber, 1991). This is likely 368 to be particularly true in outcrossing species such as Inga (Koptur, 1984). 369 However, polyploid lineages do still form and persist despite such frequency-dependent 370 disadvantages, often overcoming minority cytotype exclusion in settings where they experience 371 increased competitive ability or higher fecundity (Fowler & Levin 1984). Differences in competitive 372 ability in varying environments may explain why we found ploidy variation in species such as Inga 373 laurina and I. capitata, which are among the most broadly-distributed Inga species (Pennington, 374 1997). These species span a gradient of rainfall, which decreases from western to eastern Amazonia 375 (Espinoza Villar et al., 2009). Given that drought significantly impacts the mortality of rainforest trees 376 (Rowland et al., 2015), and elevated drought tolerance is evident in some polyploid trees (Diallo et 377 al., 2016; Ræbild et al., 2024), selection against drought-prone diploids in drier environments may 378 maintain ploidy variation within these Inga species. Alternatively, without reproductive isolation, 379 persistence of polyploids and their diploid progenitors is still probable (Gaynor et al., 2025) and has 380 been observed in many mixed-cytotype species (Bartolić et al., 2024). Polyploids can persist and 381 outcompete their diploid progenitors in such cases because inter-cytotype mating leads to gamete 382 wastage in diploids, resulting in polyploid individuals gaining a competitive advantage, particularly in 383 varying environments (Gaynor et al., 2025). 384 However, ploidy variation within widespread Inga species may also reflect taxonomic uncertainty or 385 incipient divergence of new lineages. For example, Inga laurina’s distribution spans the tropical 386 Americas, and as a result this species displays a high degree of morphological variation. This has led 387 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 15 to the suggestion that Inga laurina could actually comprise several cryptic species that remain 388 undescribed (Dexter et al., 2010). Similarly, we inferred intraspecific ploidy variation in Inga 389 capitata, which is found across the tropical Americas and is highly variable across its range, both 390 morphologically and chemically (Forrister et al., 2023), again raising the prospect of cryptic diversity 391 within this species. 392 393 Polyploidy arose repeatedly in Inga through hybridisation and WGD 394 Fifteen of the 27 Inga species that we inferred to be tetraploid recovered strong bootstrap support for a 395 tetraploid model (>600/1000 BS, Table S1, Supporting Information), and alpha value estimation in 396 NQUACK suggested that these species were autotetraploids. The remainder either had more bootstrap 397 replicates fitting a diploid model (n=7) or had similar numbers of bootstraps for both diploid and 398 tetraploid models (n=5), nearly all of which had alpha value estimates, suggesting that they were 399 allotetraploids. In contrast, one species (Inga killipiana) recovered a diploid model as best-fit, but 400 recovered more bootstraps for the tetraploid model (675/1000 BS). 401 The apparent disparity between ‘best-fitting’ model mixtures and the results of bootstrap replicates in 402 these seven species can be explained by differences in inheritance patterns between allotetraploids and 403 autotetraploids. Allotetraploids are likely to recover similar allelic ratios to diploids for biallelic sites, 404 with elevated frequencies around an allelic ratio of 0.5 in heterozygotes. We recovered exactly this in 405 both our alpha-value estimates and allele frequency histograms for these species (Table S1; Figure S1, 406 Supporting Information), suggesting that they were allotetraploids. Elevated frequencies around 0.5 407 occur in allotetraploids because they experience disomic inheritance, resulting from independent 408 segregation of their two divergent subgenomes, which leads to a high proportion of aabb genotypes 409 (Ranallo-Benavidez et al., 2020). Conversely, autotetraploids undergo tetrasomic inheritance, where 410 all four sets of homologous chromosomes can pair, resulting in a higher ratio of aaab or abbb 411 genotypes (corresponding to allelic ratios of 0.25 or 0.75, respectively) (Lv et al., 2024). We 412 recovered similar patterns for our putative autotetraploid species based on our alpha-value estimates 413 and allele frequency histograms (Table S1; Figure S1, Supporting Information). 414 Introgression has occurred frequently across Inga’s evolutionary history, involving many species 415 (Schley et al., 2025). This reticulate history may also explain several putative allotetraploid species 416 that we inferred with NQUACK, as allopolyploidy results from hybridisation that brings together two 417 separate subgenomes into one descendent lineage (Van de Peer et al., 2017). Indeed, some of the 418 accessions that we inferred to be tetraploid in this study were implicated in introgression in Schley et 419 al. (2025) (e.g. the putative allotetraploid Inga_chrysantha_KD_7021). This species displayed a high 420 degree of introgressed genetic variation in Schley et al. (2025), as indicated by the gamma statistics 421 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 16 estimated using PhyloNetworks (Solís-Lemus et al., 2017), suggesting that these allotetraploidy 422 events are relatively recent. 423 Our NQUACK analyses also recovered one accession (Inga killipiana WFR_2627) for which the best-424 supported model mixture was diploid, but recovered 675/1000 bootstrap replicates that suggested a 425 tetraploid model fits best (Table S1, Supporting Information). While it is most likely that this occurred 426 through model estimation error, it is also possible that this species was previously an autotetraploid 427 (as indicated by alpha-value estimation in NQUACK) that then underwent rediploidisation, resulting in 428 its genome retaining ‘vestigial evidence of past polyploidy’ (Wendel, 2015). 429 More broadly, our phylogenetic signal estimates suggest that polyploidy has evolved multiple times 430 independently in Inga, or at least has not been retained in entire clades (Figure 1). We recovered 431 comparatively low phylogenetic signal for polyploidy across our Inga tree – our analyses indicated 432 that polyploidy was not phylogenetically clustered (D = 0.834), and evolved across the tree in a 433 fashion that was not significantly different from random evolution (P = 0.312) or Brownian motion (P 434 = 0.089) (Table S3, Supporting Information). This suggests that the inferred polyploidy events are all 435 relatively recent, likely occurring independently several times across the Inga phylogenetic tree, 436 implying that polyploidy may be an unstable state in Inga. Recent polyploids are often considered an 437 evolutionary ‘dead end’, exhibiting higher extinction rates than diploids over time (Arrigo & Barker 438 2012; Van de Peer et al., 2017). Inga is a very young radiation (Richardson et al., 2001), which may 439 help to explain why the polyploid Inga species that did arise have persisted despite the fitness 440 disadvantages of polyploidy in outcrossing trees – these polyploids have simply not yet had enough 441 time to go extinct. However, we also recovered elevated selection in polyploid species relative to 442 diploids, suggesting selective mechanisms that may favour polyploidy, as detailed below. 443 444 Polyploids experience more widespread selection than diploids 445 We found evidence for elevated positive selection in tetraploid Inga species relative to diploid 446 species, with 63.463% of the 1207 loci we analysed under selection in tetraploid species, compared to 447 between 43.91 - 62.054% of those loci in diploid species (Figure 2; Table S4, Supporting 448 information). Tetraploids had more loci under selection than diploids across all locus annotation 449 types, most significantly in defence chemistry loci and single-copy phylogenetically informative loci 450 (Table S6, Supporting Information). Furthermore, Defence Chemistry loci in tetraploids showed the 451 highest proportion of loci under selection of any annotation class (Figure 2; Figure S2, Supporting 452 information). 453 The higher frequency of loci under positive selection in polyploid Inga species likely results from the 454 expansion of genetic variation on which selection can act that results from polyploidy. With more 455 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 17 copies of the genome there is more potential for genetic variation to accumulate, both in the case of 456 allotetraploidy and autotetraploidy (Soltis & Soltis 2000; Soltis et al., 2015). More copies of the 457 genome also mean that there is higher functional redundancy among gene copies, which can initially 458 relax selection on duplicated genes such that they can accumulate variation and, eventually, take on 459 entirely new functions (‘neofunctionalisation’) (Lynch & Conery 2000; Flagel & Wendel 2009). For 460 this reason, polyploidy is often associated with elevated phenotypic and ecological diversity in plants, 461 and can lead to adaptation to novel niches (Levin, 1983; Ramsey, 2011). This can explain the higher 462 degree of non-synonymous nucleotide changes that we found within polyploid Inga species, 463 particularly for functional genes such as those underlying chemical defence. Tropical rainforest trees 464 such as Inga are subject to relentless insect herbivory, and as a result possess a wide array of chemical 465 defences against herbivores (Kursar et al., 2009; Coley et al., 2018; Forrister et al., 2023). Thus, 466 herbivory represents a significant niche axis in Inga, and polyploidy may help to generate diversity in 467 chemical defence that allows herbivore escape. Indeed, polyploidy and gene duplication are well-468 documented as catalysts of chemical defence evolution in many plant groups (e.g., in the Brassicaceae 469 (Edger et al., 2015)) and, as a result, gene duplication is hypothesised to be a common source of 470 evolutionary novelty in plant defence compounds (Ober, 2010; Moore et al., 2014). 471 However, it is also worth noting that accumulation of deleterious mutations may also explain the 472 elevated selection we observed in polyploids. While polysomic masking in polyploid species can 473 reduce the effect of recessive mutations on the phenotype, resulting in less positive selection 474 (Haldane, 1932; Baduel et al., 2018), weakening of purifying selection can also lead to elevated 475 proportions of non-synonymous mutations (Paape et al., 2018). This occurs because polyploid 476 genomes accumulate deleterious sites more rapidly, resulting in higher genetic diversity at non-477 synonymous sites when compared to their diploid relatives (Conover & Wendel 2022). A productive 478 direction for future work would be to explore whether the observed non-synonymous mutations found 479 in the chemical defence loci of polyploid Ingas are indeed deleterious, or whether they lead to 480 chemical novelty. The latter outcome is to be expected given that reassortment and modification of 481 existing secondary chemistry through gene duplication is how chemical defence is hypothesised to 482 evolve in Inga (termed the ‘Lego chemistry’ model (Forrister et al., 2023)). 483 484 Polyploidy and the evolution of tropical tree diversity 485 While there is a distinct lack of studies that examine the fine-scale patterns of ploidy evolution in 486 tropical rainforests, polyploidy in plants is in general more frequent at higher latitudes, resulting in a 487 ‘latitudinal polyploidy gradient’ (Rice et al., 2019; Hagen et al., 2024). This latitudinal gradient likely 488

from both a higher frequency of polyploid formation at higher latitudes (due to harsher 489 environmental conditions that can induce polyploidy (De Storme & Geelen 2013; Lohaus & Van de 490 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 18 Peer 2016)) and the greater ability of polyploids to colonise new environments as a result of self-491 compatibility, elevated phenotypic plasticity and increased adaptive potential (Van de Peer et al., 492 2017). Our results are congruent with this - we found a relatively low proportion of polyploids in Inga 493 (ca. 14% of the 189 species that we sampled), falling below estimates of polyploid prevalence across 494 angiosperms (35% (Wood et al., 2009)) and even below the estimated prevalence of polyploids in 495 tropical American rainforests made by a previous study (ca. 30% (Rice et al., 2019)). Indeed, in their 496 analysis of polyploid prevalence in plants, Rice et al. found that the tropical rainforests of South 497 America, Central Africa and Borneo held the lowest proportions of polyploid plant species of 498 anywhere they surveyed, suggesting significant barriers to polyploid formation and persistence in 499 tropical rainforests. One aspect that may explain this paucity of polyploids is the low nutrient 500 availability in most tropical rainforests (Place, 2001), particularly of phosphorus. Phosphorus is a key 501 nutrient required to build nucleic acids and so, in the absence of available Phosphorus, species that 502 need to build larger genomes (such as polyploids) are likely to be at a selective disadvantage (Rice et 503 al., 2019; Morton et al., 2024). 504 The highly competitive environment of tropical rainforests may also preclude the establishment of 505 new polyploid lineages (Rice et al., 2019). This is in part due to the high abundance and species 506 richness of existing, pre-adapted diploid competitors in environments that have been relatively stable 507 for tens of millions of years when compared to higher latitudes (Cheng et al., 2013). This is further 508 highlighted by the fact that other tropical regions (such as Hawaii) host many polyploid species that 509 resulted from similarly rapid radiations as that observed in Inga (Barrier et al., 1999), but with one 510 key difference - these insular species diversified in the presence of fewer competitors, following 511 colonisation of new oceanic islands (Wilson, 1961; Yoder et al., 2010). 512 The relatively low degree of polyploidy we recovered suggests that, on one hand, polyploidy has not 513 greatly influenced the diversification of Inga, likely arising from independent events and that were not 514 shared across clades. This may be true more broadly, based on the few other rainforest tree genera that 515 have been studied at narrower scale. For example, cytological studies suggest that two species of 516 Shorea (Dipterocarpaceae) out of a total of 47 Shorea species in tropical Asia are polyploids (Kaur et 517 al., 1986). Similarly, microsatellite data from Afzelia in tropical Africa suggests that four of the six 518 Afzelia species are polyploids (Donkpegan et al., 2017). Accordingly, based on existing evidence, the 519 diminished prevalence of polyploidy that we found in Inga appears to be representative in other 520 tropical rainforest trees. 521 On the other hand, the elevated selection that we found in polyploid Inga species, particularly in loci 522 relating to chemical defence against herbivory, suggests that polyploidy has still had an effect on the 523 evolutionary history of Inga. Chemical divergence is well-established as a driver of divergent 524 evolution between closely-related Inga species (Forrister et al., 2023), and polyploidy (alongside gene 525 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 19 duplication) is well established as a catalyst for the evolution of defence chemical diversity in plants 526 (Ober, 2010; Moore et al., 2014). Together, this suggests that the elevated proportion of non-527 synonymous mutations we found in chemical defence loci within polyploid Inga species likely results 528 from the increased genetic diversity that usually occurs following polyploidisation. 529 Finally, it is of great importance to note the dearth of studies that explore polyploidy in the tropics, 530 particularly of finer-scale polyploid dynamics. For example, of the 69 studies examined by Kolář et 531 al. (2017) in their review of intraspecific ploidy variation at fine geographical scale, they found only 532 one study of a tropical savanna tree species (Senegalia senegal, Fabaceae), and found none for 533 tropical rainforest trees. This highlights the necessity for future detailed studies of polyploidy in 534 tropical rainforest tree clades, because it is these species that make up the bulk of the world’s tree 535 diversity (Ulloa Ulloa et al., 2017; Cardoso et al., 2017). 536 537

538 This work was supported by a Natural Environment Research Council standard grant (grant number 539 NE/V012258/1). Sequencing was funded partly by the Biotechnology and Biological Sciences 540 Research Council, grant number BB/P022898/1. J.N.’s work was funded by National Science 541 Foundation Standard and Dimensions of Biodiversity grants, numbers DEB-0640630 and DEB-542 1135733. M.L.G. was funded by an NSF Postdoctoral Research Fellowship in Biology (DBI-543 2410238). This project utilised equipment funded by the Wellcome Trust (Multi-User Equipment 544 Grant award number 218247/Z/19/Z). The authors acknowledge the Research/Scientific Computing 545 teams at The James Hutton Institute and NIAB for providing computational resources and technical 546 support for the ‘UK's Crop Diversity Bioinformatics HPC’ (BBSRC grant BB/S019669/1), use of 547 which has contributed to the results reported within this paper. Many thanks to Karen Moore and 548 Audrey Farbos for their superlative help with generating the data via the Exeter Sequencing Service. 549 550 Competing interests 551 The authors declare no competing interests 552 553 Author contributions 554 This study was conceived by R.J.S, R.T.P. and A.D.T. Analyses were performed by R.J.S., with 555 guidance from M.L.G., F.P., C.K., A.D.T. Sequence data were produced by R.P., J.A.N, and C.K. 556 using material collected from herbaria and silica collections by R.P., G.P.L. and K.G.D. The 557 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 20 manuscript was written by R.J.S. with contributions from R.P., J.A.N., M.L.G., G.P.L., F.P., K.G.D., 558 C.K., R.T.P., and A.D.T. 559 560 Data Availability 561 The data that support the findings of this study are openly available from online repositories. The 562 accession numbers for all data collated from previous studies and those newly submitted for this study 563 are found in Supplementary Table S1. All nucleotide sequence data produced by this study are 564 available on the NCBI Sequence Read Archive under the study accession PRJEB84192. 565 566

567 Andermann T, Torres Jiménez MF, Matos-Maraví P, Batista R, Blanco-Pastor JL, Gustafsson 568 ALS, Kistler L, Liberal IM, Oxelman B, Bacon CD. 2020. A guide to carrying out a phylogenomic 569 target sequence capture project. Frontiers in Genetics 10: 1407. 570 Andrews S. 2010. FastQC: a quality control tool for high throughput sequence data. Available online 571 at http://www.bioinformatics.babraham.ac.uk/projects/fastqc. v. 0.11.9. Cambridge, UK. Retrieved 572 01/01/2020. 573 Arrigo N, Barker MS. 2012. Rarely successful polyploids and their legacy in plant genomes. 574 Current Opinion in Plant Biology 15: 140-146. 575 Baduel P, Bray S, Vallejo-Marin M, Kolář F, Yant L. 2018. The “Polyploid Hop”: shifting 576 challenges and opportunities over the evolutionary lifespan of genome duplications. Frontiers in 577 Ecology and Evolution 6: 117. 578 Barrier M, Baldwin BG, Robichaux RH, Purugganan MD. 1999. Interspecific hybrid ancestry of a 579 plant adaptive radiation: allopolyploidy of the Hawaiian silversword alliance (Asteraceae) inferred 580 from floral homeotic gene duplications. Molecular Biology and Evolution 16: 1105-1113. 581 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 21 Bartolić P, Morgan EJ, Padilla-García N, Kolář F. 2024. Ploidy as a leaky reproductive barrier: 582 mechanisms, rates and evolutionary significance of interploidy gene flow. Annals of Botany 134: 537-583 550. 584 Benjamini Y, Hochberg Y. 1995. Controlling the false discovery rate: a practical and powerful 585 approach to multiple testing. J. R. Stat. Soc. B.:289-300. 586 Birchler JA, Yang H. 2022. The multiple fates of gene duplications: deletion, hypofunctionalization, 587 subfunctionalization, neofunctionalization, dosage balance constraints, and neutral variation. The 588 Plant Cell 34: 2466-2474. 589 Bolger AM, Lohse M, Usadel B. 2014. Trimmomatic: a flexible trimmer for Illumina sequence data. 590 Bioinformatics 30: 2114-2120. 591 Brown MR, Abbott RJ, Twyford AD. 2024. The emerging importance of cross/i1 ploidy 592 hybridisation and introgression. Molecular Ecology 33: e17315. 593 Capella-Gutiérrez S, Silla-Martínez JM, Gabaldón T. 2009. trimAl: a tool for automated 594 alignment trimming in large-scale phylogenetic analyses. Bioinformatics 25 15:1972-1973. 595 Cardoso D, Sarkinen T, Alexander S, Amorim AM, Bittrich V, Celis M, Daly DC, Fiaschi P, 596 Funk VA, Giacomin LL, Goldenberg R, Heiden G, Iganci J, Kelloff CL, Knapp S, Cavalcante 597 de Lima H, Machado AFP, Dos Santos RM, Mello-Silva R, Michelangeli FA, Mitchell J, 598 Moonlight P, de Moraes, P. L. R., Mori SA, Nunes TS, Pennington TD, Pirani JR, Prance GT, 599 de Queiroz LP, Rapini A, Riina R, Rincon CAV, Roque N, Shimizu G, Sobral M, Stehmann JR, 600 Stevens WD, Taylor CM, Trovo M, van den Berg C, van der Werff H, Viana PL, Zartman CE, 601 Forzza RC. 2017. Amazon plant diversity revealed by a taxonomically verified species list. Proc. 602 Natl. Acad. Sci. U. S. A. 114 40:10695-10700. 603 Cheng H, Sinha A, Cruz FW, Wang X, Edwards RL, d’Horta FM, Ribas CC, Vuille M, Stott 604 LD, Auler AS. 2013. Climate change patterns in Amazonia and biodiversity. Nat. Comm. 4:1411. 605 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 22 Coley PD, Endara M, Kursar TA. 2018. Consequences of interspecific variation in defenses and 606 herbivore host choice for the ecology and evolution of Inga, a speciose rainforest tree. Oecologia 187: 607 361-376. 608 Connell JH. 1971. On the role of natural enemies in preventing competitive exclusion in some marine 609 animals and in rain forest trees. In: Den Boer PJ, andGradwell GR, eds. Dynamics of populations. 610 Wageningen, The Netherlands:Centre for Agricultural Publishing and Documentation, . 611 Conover JL, Wendel JF. 2022. Deleterious mutations accumulate faster in allopolyploid than diploid 612 cotton (Gossypium) and unequally between subgenomes. Molecular Biology and Evolution 39: 613 msac024. 614 Coyne JA, and Orr HA. 2004. Speciation. Sunderland, MA: Sinauer. 615 De Storme N, Geelen D. 2013. Sexual polyploidization in plants–cytological mechanisms and 616 molecular regulation. New Phytologist 198: 670-684. 617 Dexter KG, Pennington TD, Cunningham CW. 2010. Using DNA to assess errors in tropical tree 618 identifications: How often are ecologists wrong and when does it matter? Ecol. Monogr. 80 2:267-286. 619 Diallo AM, Nielsen LR, Kjær ED, Petersen KK, Ræbild A. 2016. Polyploidy can confer 620 superiority to West African Acacia senegal (L.) Willd. trees. Frontiers in Plant Science 7: 821. 621 Donkpegan AS, Doucet J, Migliore J, Duminil J, Dainou K, Pineiro R, Wieringa JJ, 622 Champluvier D, Hardy OJ. 2017. Evolution in African tropical trees displaying ploidy-habitat 623 association: the genus Afzelia (Leguminosae). Molecular Phylogenetics and Evolution 107: 270-281. 624 Edger PP, Heidel-Fischer HM, Bekaert M, Rota J, Glöckner G, Platts AE, Heckel DG, Der JP, 625 Wafula EK, Tang M. 2015. The butterfly plant arms-race escalated by gene and genome duplications. 626 Proceedings of the National Academy of Sciences 112: 8362-8366. 627 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 23 Endara M, Forrister DL, Coley PD. 2023. The evolutionary ecology of plant chemical defenses: 628 from molecules to communities. Annual Review of Ecology, Evolution, and Systematics 54: 107-127. 629 Erkens RH, Chatrou LW, Maas JW, van der Niet T, Savolainen V. 2007. A rapid diversification 630 of rainforest trees (Guatteria; Annonaceae) following dispersal from Central into South America. Mol. 631 Phylogenet. Evol. 44 1:399-411. 632 Espinoza Villar JC, Ronchail J, Guyot JL, Cochonneau G, Naziano F, Lavado W, De Oliveira E, 633 Pombosa R, Vauchel P. 2009. Spatio/i1 temporal rainfall variability in the Amazon basin countries 634 (Brazil, Peru, Bolivia, Colombia, and Ecuador). International Journal of Climatology: A Journal of 635 the Royal Meteorological Society 29: 1574-1594. 636 Felber F. 1991. Establishment of a tetraploid cytotype in a diploid population: effect of relative 637 fitness of the cytotypes. Journal of Evolutionary Biology 4: 195-207. 638 Figueiredo MF, Bruno R, Barros e Silva, A E, Nascimento S, Oliveira IG, Felix LP. 2014. 639 Intraspecific and interspecific polyploidy of Brazilian species of the genus Inga (Leguminosae: 640 Mimosoideae). Genetics and Molecular Research 13: 3395-3403. 641 Flagel LE, Wendel JF. 2009. Gene duplication and evolutionary novelty in plants. New Phytologist 642 183: 557-564. 643 Forrister DL, Endara M, Soule AJ, Younkin GC, Mills AG, Lokvam J, Dexter KG, Pennington 644 RT, Kidner CA, Nicholls JA. 2023. Diversity and divergence: evolution of secondary metabolism in 645 the tropical tree genus Inga. New Phytologist 237: 631-642. 646 Forrister DL, Endara M, Younkin GC, Coley PD, Kursar TA. 2019. Herbivores as drivers of 647 negative density dependence in tropical forest saplings. Science 363: 1213-1216. 648 Fowler NL, Levin DA. 1984. Ecological constraints on the establishment of a novel polyploid in 649 competition with its diploid progenitor. The American Naturalist 124: 703-711. 650 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 24 Fritz SA, Purvis A. 2010. Selectivity in mammalian extinction risk and threat types: a new measure 651 of phylogenetic signal strength in binary traits. Conservation Biology 24: 1042-1051. 652 Gaynor ML, Kortessis N, Soltis DE, Soltis PS, Ponciano JM. 2025. Dynamics of mixed-ploidy 653 populations under demographic and environmental stochasticities. The American Naturalist 205: 413-654 434. 655 Gaynor ML, Landis JB, O'Connor TK, Laport RG, Doyle JJ, Soltis DE, Ponciano JM, Soltis PS. 656 2024. nQuack: An R package for predicting ploidal level from sequence data using site/i1 based 657 heterozygosity. Applications in Plant Sciences 12: e11606. 658 Gnirke A, Melnikov A, Maguire J, Rogov P, LeProust EM, Brockman W, Fennell T, 659 Giannoukos G, Fisher S, Russ C. 2009. Solution hybrid selection with ultra-long oligonucleotides 660 for massively parallel targeted sequencing. Nature Biotechnology 27: 182-189. 661 Hagen ER, Vasconcelos T, Boyko JD, Beaulieu JM. 2024. Investigating historical drivers of 662 latitudinal gradients in polyploid plant biogeography: A multiclade perspective. American Journal of 663 Botany 111: e16356. 664 Haldane JBS. 1932. The causes of evolution. Longmans, Green, New York. 665 Hanson L. 1995. Some new chromosome counts in the genus Inga (Leguminosae: Mimosoideae). 666 Kew Bulletin: 801-804. 667 Janzen DH. 1970. Herbivores and the number of tree species in tropical forests. The American 668 Naturalist 104: 501-528. 669 Johnson MG, Gardner EM, Liu Y, Medina R, Goffinet B, Shaw AJ, Zerega NJ, Wickett NJ. 670 2016. HybPiper: Extracting coding sequence and introns for phylogenetics from high/i1 throughput 671 sequencing reads using target enrichment. App. Plant. Sci. 4 7:1600016. 672 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 25 Katoh K, Standley DM. 2013. MAFFT: Multiple Sequence Alignment Software Version 7: 673 improvements in performance and usability. Molecular Biology and Evolution 30: 772-780. 674 10.1093/molbev/mst010. 675 Kaur A, Jong K, Sands VE, Soepadmo E. 1986. Cytoembryology of some Malaysian dipterocarps, 676 with some evidence of apomixis. Botanical Journal of the Linnean Society 92: 75-88. 677 Koenen EJ, Clarkson JJ, Pennington TD, Chatrou LW. 2015. Recently evolved diversity and 678 convergent radiations of rainforest mahoganies (Meliaceae) shed new light on the origins of rainforest 679 hyperdiversity. New Phytol. 207 2:327-339. 680 Koenen EJ, Kidner C, de Souza ÉR, Simon MF, Iganci JR, Nicholls JA, Brown GK, De Queiroz 681 LP, Luckow M, Lewis GP. 2020. Hybrid capture of 964 nuclear genes resolves evolutionary 682 relationships in the mimosoid legumes and reveals the polytomous origins of a large pantropical 683 radiation. American Journal of Botany 107: 1710-1735. 684 Koenen EJ, Ojeda DI, Bakker FT, Wieringa JJ, Kidner C, Hardy OJ, Pennington RT, 685 Herendeen PS, Bruneau A, Hughes CE. 2021. The origin of the legumes is a complex 686 paleopolyploid phylogenomic tangle closely associated with the Cretaceous–Paleogene (K–Pg) mass 687 extinction event. Systematic Biology 70: 508-526. 688 Kolář F, Č ertner M, Suda J, Schönswetter P, Husband BC. 2017. Mixed-ploidy species: progress 689 and opportunities in polyploid research. Trends in Plant Science 22: 1041-1055. 690 Koptur S. 1984. Outcrossing and pollinator limitation of fruit set: breeding systems of neotropical 691 Inga trees (Fabaceae: Mimosoideae). Evolution 38:1130-1143. 692 Kursar TA, Dexter KG, Lokvam J, Pennington RT, Richardson JE, Weber MG, Murakami ET, 693 Drake C, McGregor R, Coley PD. 2009. The evolution of antiherbivore defenses and their 694 contribution to species coexistence in the tropical tree genus Inga. Proc. Natl. Acad. Sci. U. S. A. 106 695 43:18073-18078. 696 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 26 Landis JB, Soltis DE, Li Z, Marx HE, Barker MS, Tank DC, Soltis PS. 2018. Impact of 697 whole/i1 genome duplication events on diversification rates in angiosperms. American Journal of 698 Botany 105: 348-363. 699 Levin DA. 1983. Polyploidy and novelty in flowering plants. The American Naturalist 122: 1-25. 700 Levin DA. 1975. Minority cytotype exclusion in local plant populations. Taxon 24: 35-43. 701 Lohaus R, Van de Peer Y. 2016. Of dups and dinos: evolution at the K/Pg boundary. Current 702 Opinion in Plant Biology 30: 62-69. 703 Lv Z, Addo Nyarko C, Ramtekey V, Behn H, Mason AS. 2024. Defining autopolyploidy: Cytology, 704 genetics, and taxonomy. American Journal of Botany 111: e16292. 705 Lynch M, Conery JS. 2000. The evolutionary fate and consequences of duplicate genes. Science 290: 706 1151-1155. 707 Meudt HM, Albach DC, Tanentzap AJ, Igea J, Newmarch SC, Brandt AJ, Lee WG, Tate JA. 708 2021. Polyploidy on islands: its emergence and importance for diversification. Frontiers in Plant 709 Science 12: 637214. 710 Moore BD, Andrew RL, Külheim C, Foley WJ. 2014. Explaining intraspecific diversity in plant 711 secondary metabolites in an ecological context. New Phytologist 201: 733-750. 712 Morton JA, Arnillas CA, Biedermann L, Borer ET, Brudvig LA, Buckley YM, Cadotte MW, 713 Davies K, Donohue I, Ebeling A. 2024. Genome size influences plant growth and biodiversity 714 responses to nutrient fertilization in diverse grassland communities. PLoS Biology 22: e3002927. 715 Murrell B, Weaver S, Smith MD, Wertheim JO, Murrell S, Aylward A, Eren K, Pollner T, 716 Martin DP, Smith DM. 2015. Gene-wide identification of episodic selection. Molecular Biology and 717 Evolution 32: 1365-1371. 718 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 27 Nguyen L, Schmidt HA, Von Haeseler A, Minh BQ. 2015. IQ-TREE: a fast and effective stochastic 719 algorithm for estimating maximum-likelihood phylogenies. Molecular Biology and Evolution 32: 268-720 274. 721 Nicholls JA, Pennington RT, Koenen EJ, Hughes CE, Hearn J, Bunnefeld L, Dexter KG, Stone 722 GN, Kidner CA. 2015. Using targeted enrichment of nuclear genes to increase phylogenetic 723 resolution in the neotropical rain forest genus Inga (Leguminosae: Mimosoideae). Frontiers in Plant 724 Science 6: 710. 725 Ober D. 2010. Gene duplications and the time thereafter–examples from plant secondary metabolism. 726 Plant Biology 12: 570-577. 727 Orme D, Freckleton R, Thomas G, Petzoldt T, Fritz S, Isaac N, Pearse W. 2013. The caper 728 package: comparative analysis of phylogenetics and evolution in R. v. 5. https://cran.r-729 project.org/web/packages/caper/vignettes/caper.pdf. Retrieved 01/01/2020. 730 Paape T, Briskine RV, Halstead-Nussloch G, Lischer HE, Shimizu-Inatsugi R, Hatakeyama M, 731 Tanaka K, Nishiyama T, Sabirov R, Sese J. 2018. Patterns of polymorphism and selection in the 732 subgenomes of the allopolyploid Arabidopsis kamchatica. Nature Communications 9: 3909. 733 Pennington TD. 1997. The genus inga: Botany. Royal Botanic Gardens. 734 Percival/i4 Alwyn L, Barnes I, Clark MD, Cockram J, Coffey MP, Jones S, Kersey PJ, Kidner 735 CA, Kosiol C, Li B. 2025. UKCropDiversity/i1 HPC: A collaborative high/i1 performance computing 736 resource approach for sustainable agriculture and biodiversity conservation. Plants, People, Planet 7: 737 969-977. 738 Place SE. 2001. Tropical rainforests: Latin american nature and society in transition. Lanham, MD, 739 USA: Rowman & Littlefield. 740 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 28 R Development Core Team. 2013. R: A language and environment for statistical computing. 741 Available at: http://www.R-project.org/. v. 3.6. R Foundation for Statistical Computing, Vienna, 742 Austria. 743 Ræbild A, Anamthawat-Jónsson K, Egertsdotter U, Immanen J, Jensen AM, Koutouleas A, 744 Martens HJ, Nieminen K, Olofsson JK, Röper A. 2024. Polyploidy–A tool in adapting trees to 745 future climate changes? A review of polyploidy in trees. Forest Ecology and Management 560: 746 121767. 747 Ramsey J. 2011. Polyploidy and ecological adaptation in wild yarrow. Proceedings of the National 748 Academy of Sciences 108: 7096-7101. 749 Ranallo-Benavidez TR, Jaron KS, Schatz MC. 2020. GenomeScope 2.0 and Smudgeplot for 750 reference-free profiling of polyploid genomes. Nature Communications 11: 1432. 751 Ranwez V, Chantret N, and Delsuc F. 2021. Aligning protein-coding nucleotide sequences with 752 MACSE. In: Katoh K, ed. Multiple sequence alignment: Methods and protocols. New York, NY, 753 USA:Springer, 51. 754 Ranwez V, Harispe S, Delsuc F, Douzery EJ. 2011. MACSE: Multiple Alignment of Coding 755 SEquences accounting for frameshifts and stop codons. PloS One 6: e22594. 756 Revell LJ. 2012. phytools: an R package for phylogenetic comparative biology (and other things). 757

in Ecology and Evolution 3 2:217-223. 758 Rice A, Šmarda P, Novosolov M, Drori M, Glick L, Sabath N, Meiri S, Belmaker J, Mayrose I. 759 2019. The global biogeography of polyploid plants. Nature Ecology & Evolution 3: 265-273. 760 Richardson JE, Pennington RT, Pennington TD, Hollingsworth PM. 2001. Rapid diversification 761 of a species-rich genus of neotropical rain forest trees. Science 293 5538:2242-2245. 762 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 29 Ringelberg JJ, Koenen EJ, Sauter B, Aebli A, Rando JG, Iganci JR, de Queiroz LP, Murphy DJ, 763 Gaudeul M, Bruneau A. 2023. Precipitation is the main axis of tropical plant phylogenetic turnover 764 across space and time. Science Advances 9: eade4954. 765 Rowland L, da Costa, Antonio Carlos Lola, Galbraith DR, Oliveira RS, Binks OJ, Oliveira AA, 766 Pullen AM, Doughty CE, Metcalfe DB, Vasconcelos SS. 2015. Death from drought in tropical 767 forests is triggered by hydraulics not carbon starvation. Nature 528: 119-122. 768 Schley RJ&, Pennington RT, Twyford AD, Dexter KG, Kidner C, Michael TP, of Life, 769 Wellcome Sanger Institute Tree. 2024. The genome sequence of Inga laurina (Sw.) Willd. 770 Wellcome Open Research 9: 567. 771 Schley RJ&, Piñeiro Portela R, Nicholls JA, Pezzini FF, Farbos A, Lewis GP, Ringelberg JJ, 772 Kidner C, Twyford AD, Dexter KG, et al. 2025. Rampant Reticulation in a Rapid Radiation of 773 Tropical Trees - Insights from Inga (Fabaceae). Systematic 774 Biologyhttps://doi.org/10.1093/sysbio/syaf027. 775 Shibata K. 1962. Cytological studies on some wild and cultivated plants of Colômbia. II. Species of 776 Mimosaceae and Galegeae. Japanese Journal of Genetics 37: 410. 777 Solís-Lemus C, Bastide P, Ané C. 2017. PhyloNetworks: a package for phylogenetic networks. Mol. 778 Biol. Evol. 34 12:3292-3298. 779 Soltis PS, Marchant DB, Van de Peer Y, Soltis DE. 2015. Polyploidy and genome evolution in 780 plants. Current Opinion in Genetics & Development 35: 119-125. 781 Soltis PS, Soltis DE. 2000. The role of genetic and genomic attributes in the success of polyploids. 782 Proceedings of the National Academy of Sciences 97: 7051-7057. 783 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 30 Suda J, Weiss/i4 Schneeweiss H, Tribsch A, Schneeweiss GM, Trávní č ek P, Schönswetter P. 2007. 784 Complex distribution patterns of di/i1 , tetra/i1 , and hexaploid cytotypes in the European high mountain 785 plant Senecio carniolicus (Asteraceae). American Journal of Botany 94: 1391-1401. 786 Ulloa Ulloa C, Acevedo-Rodriguez P, Beck S, Belgrano MJ, Bernal R, Berry PE, Brako L, Celis 787 M, Davidse G, Forzza RC, Gradstein SR, Hokche O, Leon B, Leon-Yanez S, Magill RE, Neill 788 DA, Nee M, Raven PH, Stimmel H, Strong MT, Villasenor JL, Zarucchi JL, Zuloaga FO, 789 Jorgensen PM. 2017. An integrated assessment of the vascular plant species of the Americas. Science 790 358 6370:1614-1617. 791 Van de Peer Y, Mizrachi E, Marchal K. 2017. The evolutionary significance of polyploidy. Nature 792 Reviews Genetics 18: 411-424. 793 WCVP. 2020. World Checklist of Vascular Plants, version 2.0. Facilitated by the Royal Botanic 794 Gardens, Kew. Published on the Internet; http://wcvp.science.kew.org/ Retrieved 31st January 2020. 795 Weiß CL, Pais M, Cano LM, Kamoun S, Burbano HA. 2018. nQuire: a statistical framework for 796 ploidy estimation using next generation sequencing. BMC Bioinformatics 19: 1-8. 797 Wendel JF. 2015. The wondrous cycles of polyploidy in plants. American Journal of Botany 102. 798 Wilson EO. 1961. The nature of the taxon cycle in the Melanesian ant fauna. The American 799 Naturalist 95: 169-193. 800 Wood TE, Takebayashi N, Barker MS, Mayrose I, Greenspoon PB, Rieseberg LH. 2009. The 801 frequency of polyploid speciation in vascular plants. Proceedings of the National Academy of 802 Sciences 106: 13875-13879. 803 Yoder JB, Clancey E, Des Roches S, Eastman JM, Gentry L, Godsoe W, Hagey TJ, Jochimsen 804 D, Oswald BP, Robertson J. 2010. Ecological opportunity and the origin of adaptive radiations. 805 Journal of Evolutionary Biology 23: 1581-1596. 806 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint 31 Zhang C, Rabiee M, Sayyari E, Mirarab S. 2018. ASTRAL-III: polynomial time species tree 807 reconstruction from partially resolved gene trees. BMC Bioinformatics 19: 153. 808 809 Supporting Information 810 Additional supporting information may be found in Supporting Information 1 in the online version of 811 this article. 812 • Figure S1: nQuack allele frequency histograms for putative tetraploids 813 • Figure S2: Boxplot showing number of defence chemistry loci under selection for each ploidy 814 • Table S1: Ploidy inference and sample information (Available on Dryad) 815 • Table S2: Taxa included in BUSTED analyses: polyploids and six sets of diploids 816 • Table S3: Ploidy phylogenetic signal modelling results 817 • Table S4: Counts of loci under selection for each BUSTED ploidy run 818 • Table S5: BUSTED selection χ 2 results: selection x annotation type for each ploidy level 819 • Table S6: BUSTED selection χ 2 results: selection x ploidy for each annotation type 820 .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint .CC-BY-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 2, 2025. ; https://doi.org/10.1101/2025.07.28.667178doi: bioRxiv preprint