^{1}

^{2}

^{3}

^{1}

^{*}

^{4}

^{5}

^{6}

^{2}

^{3}

^{7}

^{1}

^{5}

^{8}

^{5}

^{1}

The authors have declared that no competing interests exist.

Conceived and designed the experiments: ME SMB IM. Performed the experiments: ME SMB. Analyzed the data: ME SMB. Contributed reagents/materials/analysis tools: ME SMB IM MFM OFP ALH MP PJM VV PV ML. Wrote the paper: ME SMB IM MFM OFP ALH MP PJM VV PV ML.

Chromosome evolution has been demonstrated to have profound effects on diversification rates and speciation in angiosperms. While polyploidy has predated some major radiations in plants, it has also been related to decreased diversification rates. There has been comparatively little attention to the evolutionary role of gains and losses of single chromosomes, which may or not entail changes in the DNA content (then called aneuploidy or dysploidy, respectively). In this study we investigate the role of chromosome number transitions and of possible associated genome size changes in angiosperm evolution. We model the tempo and mode of chromosome number evolution and its possible correlation with patterns of cladogenesis in 15 angiosperm clades. Inferred polyploid transitions are distributed more frequently towards recent times than single chromosome gains and losses. This is likely because the latter events do not entail changes in DNA content and are probably due to fission or fusion events (dysploidy), as revealed by an analysis of the relationship between genome size and chromosome number. Our results support the general pattern that recently originated polyploids fail to persist, and suggest that dysploidy may have comparatively longer-term persistence than polyploidy. Changes in chromosome number associated with dysploidy were typically observed across the phylogenies based on a chi-square analysis, consistent with these changes being neutral with respect to diversification.

While variation in chromosome number is widespread among plants, its role in species diversification has long been debated

In this study, we consider gains and losses of single chromosomes as processes that entail (i) change in DNA content (aneuploidy: duplication or losses of chromosomes), or (ii) little or negligible change in DNA content (dysploidy: fission and/or fusion;

A central issue in the study of chromosome number evolution concerns the relative importance of polyploid transitions versus gains and losses of single chromosomes

Here, we perform a phylogenetic comparative analysis of chromosome evolution and lineage diversification in 15 flowering plant clades to estimate the relative importance of polyploidy and gains and losses of single chromosomes in the evolution of angiosperms. The particular aims of this study are to evaluate (i) previous hypotheses about the role of polyploidy in angiosperm diversification; (ii) the persistence of gains and losses of single chromosomes (including a priori aneuploidy and dysploidy) along angiosperm evolution; and (iii) the relative distribution and timing of polyploidy and gains and losses of single chromosomes across angiosperm phylogenies.

We used molecular phylogenies from 15 angiosperm groups (

Focal group (order, family) | Species richness | S_{phy} |
S_{counts} |
2 |
Centromeretype |

12 | 12 | 12 | 24 – 96 | Monocentric | |

Orchidinae (Asparagales, Orchidaceae) | ca. 1800 | 103 | 73 | 10–82 | Monocentric |

21 | 19 | 18 | 9 – 45 | Monocentric | |

49 | 47 | 47 | 17 – 51 | Monocentric | |

Resedaceae (Brassicales) | ca. 85 | 66 | 35 | 6 – 40 | Monocentric |

14 | 14 | 14 | 9 – 70 | Monocentric | |

ca. 74 | 66 | 55 | 8 – 80 | Monocentric | |

Antirrhineae (Lamiales, Plantaginaceae) | ca. 326 | 44 | 36 | 6 – 18 | Monocentric |

ca. 530 | 61 | 56 | 6 – 12 | Monocentric | |

Cistaceae (Malvales) | ca. 180 | 47 | 45 | 5 – 24 | Monocentric |

Cariceae (Poales, Cyperaceae) | ca. 2000 | 135 | 100 | 6 – 57 | Holocentric |

ca. 90 | 57 | 57 | 26 – 43 | Holocentric | |

ca. 70 | 35 | 21 | 30 – 46 | Holocentric | |

ca. 70 | 38 | 25 | 30 – 42 | Holocentric | |

ca. 70 | 56 | 50 | 8 – 110 | Monocentric |

Given the dated molecular phylogenies and the assignments of chromosome numbers to the tips, we aimed to infer the location of chromosome number transitions using the ChromEvol methodology

Focal group (order, family) | Best supported model |
x (P>0.05) | Gains | Losses | PP | Demi-PP | P value from Chi-square for Polyploidy |
P value from Chi-square for gains and losses of single chromosomes |
Type of gains and losses of single chromosomes |

CRD | 24 (0.92) | 0 | 0 | 2 | 5 | 0.2255 (obs recent > exp recent, obs ancient < exp ancient) | - | - | |

Orchidinae (Asparagales, Orchidaceae) | CRD | 21 (0.88), 22 (0.11) | 1 | 32 | 3 | 1 | 0.0084 (obs recent > exp recent, obs ancient < exp ancient) | 0.3477 (obs recent > exp recent, obs ancient < exp ancient) | Dysploidy |

CRDE | 9 (0.99) | 0 | 0 | 6 | 0 | 0.0028 (obs recent > exp recent, obs ancient < exp ancient) | - | - | |

CRD | 17 (0.96) | 0 | 0 | 11 | 11 | - | - | ||

Resedaceae (Brassicales) | CRDE | 3 (0.256), 4 (0.25), 2(0.20), 5 (0.15), 1 (0.08), 6 (0.05) | 21 | 0 | 10 | 1 | 0.3140 (obs recent > exp recent, obs ancient < exp ancient) | 0.7685 (obs recent < exp recent, obs ancient < exp ancient) | Dysploidy |

LR | None P>0.05 | 920 | 1270 | 46 | 0 | 0.7384 (obs recent > exp recent, obs ancient < exp ancient) | Undetermined | ||

CRDE | 10 (0.45), 9 (0.36), 5 (0.10) | 2 | 3 | 9 | 1 | Undetermined | |||

Antirrhineae (Lamiales, Plantaginaceae) | LRDE | 9 (0.96) | 0 | 14 | 6 | 1 | 0.5354 (obs recent < exp recent, obs ancient > exp ancient) | 0.5866 (obs recent < exp recent, obs ancient > exp ancient) | Dysploidy |

CRD | 6 (0.99) | 1 | 0 | 4 | 2 | 0.3355 (obs recent < exp recent, obs ancient > exp ancient) | 0.7407 (obs recent > exp recent, obs ancient < exp ancient) | Dysploidy | |

Cistaceae (Malvales) | CR | 4 (0.87), 2 (0.08) | 7 | 0 | 9 | 0 | 0.1743 (obs recent > exp recent, obs ancient < exp ancient) | 0.3006 (obs recent > exp recent, obs ancient < exp ancient) | Dysploidy |

Cariceae (Poales, Cyperaceae) | LR | None P>0.05 | 3480 | 3699 | 3 | 0 | 0.0269 (obs recent > exp recent, obs ancient < exp ancient) | Dysploidy | |

LRND | None P>0.05 | 1176 | 1086 | 0 | 0 | - | 0.9765 (obs recent > exp recent, obs ancient < exp ancient) | Dysploidy | |

CRND | 38 (0.32), 37 (0.23), 39 (0.23), 36 (0.09), 40 (0.08) | 74 | 101 | 0 | 0 | - | 0.2936 (obs recent >exp recent, obs ancient < exp ancient) | Dysploidy | |

CRND | 38 (0.27), 39(0.24), 37 (0.19), 40 (0.13), 36 (0.08) | 12 | 34 | 0 | 0 | - | 0.3439 (obs recent > exp recent, obs ancient < exp ancient) | Dysploidy | |

CRD | 25 (0.08), 24 (0.08), 26 (0.08), 23 (0.08), 27 (0.07), 22 (0.07), 28 (0.06), 21 (0.06), 29(0.05), 20 (0.05) | 0 | 216 | 29 | 0 | 0.4278 (obs recent > exp recent, obs ancient < exp ancient) | 0.1471 (obs recent < exp recent, obs ancient > exp ancient) | Dysploidy |

Using the ChromEvol software (evaluatePPDist option), we calculated the observed chromosome number transitions per unit of time that occurred relatively recently (four temporal strategies: tip branches, from present to 10% of total time, from present to 25% of total time and from present to 50% of total time) and those that occurred deeper in time (the rest of the tree). We calculated the expected number of each type of chromosome number transition (polyploidizations plus demipolyploidizations and gains plus losses of single chromosomes) along the tree assuming that they occur homogenously over time as the null hypothesis (total number of events inferred along the tree divided by the total time). Then chi-square was used to test whether the number of observed transitions along external and internal branches for each temporal level is significantly different than the number of transitions under the expectation of constant transition rate along the tree (null hypothesis). P values smaller than 0.002 were considered significant to reject the null hypothesis. We selected this conservative P-value because our tests are two-tailed and because we did multiple tests and therefore Bonferroni correction is required. Nevertheless, P-values smaller than 0.025 were considered as marginal support to reject the null hypothesis.

Gains and losses of single chromosomes encompass different phenomena with various expected outcomes: 1) Aneuploidy: duplication or loss of a chromosome including its DNA content, and 2) Dysploidy: chromosome fusion/fission that do not result in changes in DNA content. In addition, losing chromosomes after polyploidization has a different outcome than when there is not previous polyploidization (in the first case, although genes may be lost, there are extra copies of the genes). In order to differentiate between different patterns of gains and losses of single chromosomes for lineages for which these transitions were inferred, we gathered genome size values from the Plant DNA C-values database (data.kew.org/cvalues/) and references therein. We then analyzed the possible correlation between genome size and chromosome numbers (excluding species affected by polyploidy) both visually and using linear models as implemented in the function

The studied groups comprise a high cytogenetic variability, with chromosome numbers ranging from ^{−1})) and polyploid transitions (0.0013–0.7888 #transitions my^{−1}) (

Our analyses (at four different temporal levels) comparing the expected and the observed number of transitions per time reveal similar results: First, the number of polyploid transitions towards the tips of the trees is significantly higher than expected under the null hypothesis (constant polyploidization rate through time) in four of the 12 datasets with polyploidy, in two data sets there is marginal support to reject the null hypothesis whereas for the remaining six data sets the results were congruent with it (

In this study, we have compared the expected vs. observed number of chromosome transitions in a temporal context following several alternative strategies (tips vs. rest of the chronogram tree, present to 10% of total time vs. rest of the chronogram tree, present to 25% of total time vs. rest of the chronogram tree, and present to 50% of total time vs. rest of the chronogram tree) obtaining identical results. Inferred polyploidizations were distributed closer to the tips of the trees than expected under constant chromosome transition rate (null hypothesis) for half of datasets under all tests (

Gains and losses of single chromosomes may originate via many disparate mechanisms, which entail small changes in the chromosome number and changes (aneuploidy) or not (dysploidy) in DNA content (

Our results suggest that most of the inferred gains and losses of single chromosomes are dysploid events, and therefore these transitions mostly correspond to changes in chromosome number without changes in DNA content. Gene balance theory predicts that loss or duplication of a subset of chromosomes (aneuploidy) should be more strongly selected against than whole-genome duplication

Interestingly, holocentric chromosomes (without localized centromere) evolve almost strictly by fission and fusion which do not convey changes in DNA content and those rearrangements are generally neutral or nearly so because of the effects of the diffuse centromere

To sum up, our results support the hypothesis that dysploidy is less disadvantageous than polyploidy in terms of generating long term persisting lineages

A. Glossary of cytogenetic terms used throughout the article. B. Results and discussion concerning the individual datasets used for our study.

(DOCX)

Table S1. Parameters of the best supported model inferred for each group with ChromEvol. Table S2. Species, diploid chromosome number (

(DOCX)

Zip file with graphs of phylogenetic trees with haploid chromosome numbers in the tips and phylogentic trees in parenthetical format from chromEvol analyses with inferred mutation events.

(ZIP)

We thank the Academic Editor Dr. Gabriel AB Marais, Dr. Aretuza Sousa and two anonymous reviewers for their critical comments on the manuscript.

^{nd}Ed. New York: Columbia University Press. 432 p.