Gradual genome size evolution and polyploidy in Allium from the Qinghai–Tibetan Plateau

Abstract

Background and Aims

Genome size is an important plant trait, with substantial interspecies variation. The mechanisms and selective pressures underlying genome size evolution are important topics in evolutionary biology. There is considerable diversity in Allium from the Qinghai–Tibetan Plateau, where genome size variation and related evolutionary mechanisms are poorly understood.

Methods

We reconstructed the Allium phylogeny using DNA sequences from 71 species. We also estimated genome sizes of 62 species, and determined chromosome numbers in 65 species. We examined the phylogenetic signal associated with genome size variation, and tested how well the data fit different evolutionary models. Correlations between genome size variations and seed mass, altitude and 19 bioclimatic factors were determined.

Key Results

Allium genome sizes differed substantially between species and within diploids, triploids, tetraploids, hexaploids and octaploids. Size per monoploid genome (1Cx) tended to decrease with increasing ploidy levels. Allium polyploids tended to grow at a higher altitude than diploids. The phylogenetic tree was divided into three evolutionary branches. The genomes in Clade I were mostly close to the ancestral genome (18.781 pg) while those in Clades II and III tended to expand and contract, respectively. A weak phylogenetic signal was detected for Allium genome size. Furthermore, significant positive correlations were detected between genome size and seed mass, as well as between genome size and altitude. However, genome size was not correlated with 19 bioclimatic variables.

Conclusions

Allium genome size shows gradual evolution, followed by subsequent adaptive radiation. The three well-supported Allium clades are consistent with previous studies. The evolutionary patterns in different Allium clades revealed genome contraction, expansion and relative stasis. The Allium species in Clade II may follow adaptive radiation. The genome contraction in Clade III may be due to DNA loss after polyploidization. Allium genome size might be influenced by selective pressure due to the conditions on the Qinghai–Tibetan Plateau (low temperature, high UV irradiation and abundant phosphate in the soil).

INTRODUCTION

Genome size is the total DNA amount in the unreplicated haploid nucleus (Greilhuber, 2005). It is an important consideration when examining plant evolution, because it represents the necessary baseline information for whole-genome sequencing, and it relationships to morphological characteristics, physiological indices and polyploidization have been reported (Beaulieu et al., 2008). Genome size differs substantially among species, varying by over 2400-fold among angiosperms (Pellicer et al., 2010). There are probably several reasons for the differences in genome size, including recombination rate, tandem repeats (Tiley and Burleigh, 2015), transposable elements (TEs) and polyploidization; however, the relative contribution of each of these factors probably varies between species (Bennetzen et al., 2005). Remarkably, genome size variations may also be linked to several other factors, including evolutionary processes (Rees et al., 1966; Laurie and Bennett, 1985; Leitch and Leitch, 2013) such as selection (Bennetzen et al., 2005), as well as abiotic factors, including elevation (Suda et al., 2009), rainfall (Qiu et al., 2019) and soil nutrient content (Guignard et al., 2016). Clearly, ecological factors contribute to genome size variations (Carta and Peruzzi, 2016).

Genome size is also related to the phenotype, karyotype and seed characteristics (Kang et al., 2014; Jordan et al., 2015). For example, there are strong correlations between genome size and cell size, guard cell length and seed mass (Knight et al., 2005). Additionally, a significant correlation between genome size and karyotype traits (e.g. total karyotype length and mean chromosome length) was observed in Echinops (Asteraceae) (Garnatje et al., 2004). Notably, a previous study demonstrated that the effects of the factors influencing genome size are greater at lower taxonomic levels than at higher taxonomic levels, implying that phylogenetic scale is an important consideration for studies on genome size evolution (Dušková et al., 2010). The genus is an ideal taxonomic level for investigating plant genome size evolution (Hawkins et al., 2008). Unfortunately, there have been relatively few studies addressing genome size variations in a genus and their relationship with the phenotype, karyotype and ecological factors (Kang et al., 2014; Jordan et al., 2015).

The Qinghai–Tibetan Plateau, which is characterized by high elevation, low pressure, low temperature and frequent precipitation (rain and snow), has long been regarded as a ‘natural laboratory’ for investigating plant evolution and environmental adaptation. However, there is limited information available regarding the genome size of plants growing there. The richness of species and high differentiation among Allium species make them ideal for studying genome size evolution on the Qinghai–Tibetan Plateau. Allium is a monophyletic taxon comprising about 971 species in the northern hemisphere, with ~111 species growing on or surrounding the Qinghai–Tibetan Plateau (Hauenschild et al., 2017). There is substantial interest in the relationship between genome size and biological characteristics. For example, previous investigations revealed significant positive correlations between genome size and karyotype characters (e.g. chromosome length and volume), as well as between genome size and seed mass in Allium (Jones and Rees, 1968; Labani and Elkington, 1987; Ohri et al., 1998). The application of model-based phylogenetic comparative methods provides a reliable means to study the mode and tempo of genome size evolution (Kang et al., 2014); however, there are only a few reports regarding Allium. Significant interspecific variations in genome size were detected on the basis of ITS phylogeny (Ricroch et al., 2005). Moreover, the mode and tempo of genome size evolution in Allium subgenus Melanocrommyum were reported to be punctuated evolution and species-specific adaptation according to the ITS phylogeny (Gurushidze et al., 2012). In the present study, we investigate Allium genome size variations and evaluate the relationships between genome size and seed mass, and as well as ecological factors in a phylogenetic framework. We also clarify the mode and tempo of Allium genome size evolution.

MATERIALS AND METHODS

Plant materials

Supplementary Data Table S1 lists 83 taxa studied, which represent 65 species from 12/15 subgenera. A total of 62 species are used for genome size analysis, 63 species used for seed mass analysis and 46 species used for ploidy analysis (Table S2). Genome size, 2C-value, seed mass, chromosome number and ploidy level data are also presented (Table S2). Combined with data downloaded from the GenBank database, the final dataset includes 71 Allium species (Table S3). Leucocoryne pauciflora, Tristagma uniflorum, Tulbaghia fragrans, Ipheion uniflorum, Nothoscordum gracile and Dichelostemma multiflorum are used as outgroups.

All Allium samples were collected from the Qinghai–Tibetan Plateau and then grown in a glasshouse at the Kunming Institute of Botany, Chinese Academy of Sciences. Voucher specimens were deposited in the Herbarium of the Kunming Institute of Botany, Chinese Academy of Sciences (KUN).

Chromosome number

Root tips were pretreated with 8-hydroxyquinoline (0.002 mol L^–1) at 20–21 °C for 4–5 h, fixed in Carnoy solution (ethanol/acetic acid = 3:1) for 50 min at 4 °C, and then dissociated in a solution (1 m HCl/45 % acetic acid = 1:1) at 60 °C for 30 s. We used acetic orcein (1 %) to stain for 2–3 h and observed root tips on a glass slide (Wang et al., 2013). Three to five individuals were studied, and at least five metaphase plates from each individual were studied for counting chromosome numbers.

Genome size estimation

Genome size was measured using CyFlow Space-3000 (Partec, Germany). Triticum aestivum ‘Chinese Spring’ (2C = 35.60 pg; International Wheat Genome Sequencing Consortium, 2014) and Secale cereale (2C = 16.19 pg; Doležel et al., 1998; Li et al., 2021) were chosen as internal standards. The 1Cx-value was defined as the genome size. Leaf material was cut with a razor blade in a culture dish using WPB nuclear solution buffer (Doležel et al., 2007). After filtering cellular debris using disposable filters (30 µm), the nuclear suspension was stained for 10 min using 150 µL propidium iodide (50 μg mL^–1). At least 5000–10 000 nuclei were detected for all samples. The genome size value was the average of one to three accessions, and three individuals per accession were measured. To evaluate the results, the average of coefficient of variation values (CV) <5 % was regarded as reliable. On the basis of the peak of internal standard and Allium species, Allium genome sizes were calculated (Supplementary Data Fig. S1):

$${\displaystyle \begin{array}{ll}2\mathrm{C}-\mathrm{v}\mathrm{a}\mathrm{l}\mathrm{u}\mathrm{e}\left(\mathrm{p}\mathrm{g}\right)=& (\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}\mathrm{s}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{e}\mathrm{p}\mathrm{e}\mathrm{a}\mathrm{k}\times 2\mathrm{C}-\mathrm{v}\mathrm{a}\mathrm{l}\mathrm{u}\mathrm{e}\left(\mathrm{p}\mathrm{g}\right)\\ & \mathrm{o}\mathrm{f}\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{d}\mathrm{s}\mathrm{p}\mathrm{e}\mathrm{c}\mathrm{i}\mathrm{e}\mathrm{s})/{\displaystyle \mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{d}\mathrm{p}\mathrm{e}\mathrm{a}\mathrm{k}}\end{array}}$$

$${\displaystyle \begin{array}{l}{\displaystyle \mathrm{G}\mathrm{e}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{s}\mathrm{i}\mathrm{z}\mathrm{e}=2\mathrm{C}-\mathrm{v}\mathrm{a}\mathrm{l}\mathrm{u}\mathrm{e}\left(\mathrm{p}\mathrm{g}\right)/\mathrm{p}\mathrm{l}\mathrm{o}\mathrm{i}\mathrm{d}\mathrm{y}\mathrm{l}\mathrm{e}\mathrm{v}\mathrm{e}\mathrm{l}}\end{array}}$$

Seed mass

Seed mass was calculated using the hundred-grain method. To estimate the seed mass of the investigated taxa, one to three accessions were measured, and the average value of eight repeats of 100 seeds each was recorded as the seed mass.

DNA extraction, PCR amplification and sequencing

A nuclear locus, ITS, and chloroplast loci (rps16, matK, atpB-rbcL, rpl32-trnL and trnQ-rps16) were used to reconstruct a phylogenetic tree (Friesen, 2006; Hauenschild et al., 2017). In this study, 235 of the 372 sequences were previously published and obtained from NCBI GenBank. The DNA sequences for the nuclear locus (ITS: 61 species) and five plastid loci (rps16: 44 species, rpl32-trnL: 39 species, matK: 46 species, atpB-rbcL: 18 species, and trnQ-rps16: 12 species) were downloaded (Supplementary Data Table S3). The remaining 84 sequences were newly generated by DNA extraction, polymerase chain reaction (PCR) amplification and sequencing. These sequences were available at the National Genomics Data Center (https://ngdc.cncb.ac.cn/) under BioProject number CRA005207. Genomic DNA extraction used the CTAB method (Doyle and Doyle, 1987). The volume of PCR amplification was 20 µL including 2.5 µL MgCl₂ (1.5 mmol L^–1), 2.5 µL dNTP mixture (0.2 mmol L^–1), 1 µL primers (0.2 µmol L^–1), 0.3 µL Tap polymerase, 2 µL genomic DNA as template, 2 µL PCR buffer and 9.7 µL ddH₂O.

The PCR amplification conditions were as follows: ITS: 94 °C for 5 min, followed by 30 cycles of 45 s at 94 °C, 55 °C for 45 s and 72 °C for 1 min, and final extension for 7 min at 72 °C; rps16: 95 °C for 4 min, followed by 35 cycles of 30 s at 95 °C, 51.3 °C for 1 min and 72 °C for 1.5 min, with a final extension for 10 min at 72 °C; matK: 94 °C for 3 min; followed by 32 cycles of 30 s at 94 °C, 55 °C for 30 s and 72 °C for 2 min, with a final extension for 10 min at 72 °C; atpB-rbcL: 94 °C for 3 min; followed by 30 cycles of 45 s at 94 °C, 50 °C for 1 min and 72 °C for 1 min, with final extension for 10 min at 72 °C; rpl32-trnL: 80 °C for 5 min; followed by 30 cycles of 1 min at 95 °C, 50 °C for 1 min and 65 °C for 4 min, with a final extension for 5 min at 65 °C; trnQ-rps16: 94 °C for 4 min; followed by 32 cycles of 45 s at 94 °C, 53 °C for 90 s and 72 °C for 70 s, with a final extension for 10 min at 72 °C.

The polyethylene glycol (PEG) precipitation procedure was used to purify the PCR products. We used ClustalX 1.83 to align sequences (Thompson et al., 1997), and Bioedit to manually edit the sequences (Hall, 1999).

Phylogenetic analysis

We conducted a comprehensive species-level analysis of phylogenetic relationships using the DNA sequences of a nuclear locus (ITS) and chloroplast loci (rps16, rpl32-trnL, matK, atpB-rbcL and trnQ-rps16). Some unavailable sequences were treated as missing data. Bayesian inference (BI), Modeltest 3.7 and MrBayes 3.1.2 were respectively used to reconstruct phylogenetic trees, determine the optimal model of molecular evolution and implement Bayesian inference under a Markov chain Monte Carlo (MCMC) algorithm (Huelsenbeck and Ronquist, 2001; Posada and Crandall, 1998; Posada and Buckley, 2004). PAUP* 4.0 b10 was used to calculate the consensus tree (Swofford, 2003). The incongruence length difference (ILD) test was used to evaluate the congruence of the nuclear and plastid datasets (Farris et al., 1994).

Ancestral genome size

To clarify the evolutionary history of Allium genome size, we reconstructed the ancestral genome size using the square-change parsimony reconstruction method in Mesquite 2.75 (Madison and Madison, 2011). Based on the above phylogenetic tree using MrBayes 3.1.2, the file g. nex. run 1. Was used for ancestral genome calculations in Mesquite 2.75.

Models of trait evolution

To detect the mode and tempo of Allium genome size, we used the generalized least-squares method in Bayes Traits (Pagel and Meade, 2007). In the analysis, the average genome size of species was used to evaluate evolutionary models and for comparative analysis. The median values of 19 bioclimatic variables were determined for each region in which species were collected. The parameters Pagel’s lambda (λ), kappa (κ) and delta (δ) were respectively used to confirm the phylogenetic association, mode and tempo of the genome size evolution. The parameter λ represents the phylogenetic signal, which assesses the contribution of phylogeny to the interspecific covariance of a given trait. Specifically, λ = 0 suggests trait evolution independent of phylogeny, while λ = 1 indicates phylogenetic relationships can effectively predict trait variation patterns (pure Brownian motion process). Additionally, 0 < λ < 1 represents varying degrees in the phylogenetic signal. The parameter κ reflects the contributions of trait evolution to the punctuated and gradual mode in a phylogeny. Specifically, κ = 0 suggests trait evolution does not depend on branch length, which is consistent with a punctuated mode of evolution, whereas κ = 1 suggests that trait evolution depends directly on branch length. Moreover, κ < 1 reflects a greater proportion of evolution in shorter branches, while κ > 1 reflects a greater proportion of evolution in longer branches (gradual mode). The parameter δ is used to test differential evolutionary rates over time and to rescale the phylogeny according to the stability of the evolutionary rate: δ = 1 (gradual evolution). Additionally, δ < 1 suggests a temporally early trait evolution, which is a signature of adaptation radiation, whereas δ > 1 indicates longer paths contributed disproportionately to trait evolution, which represents accelerated evolution over time (species-specific adaptation). The most appropriate models were tested according to likelihood ratio tests (LRTs) (Huelsenbeck and Rannala, 1997).

Phylogenetic regression analysis

We used the phylogenetic generalized least-squares (PGLS) method to evaluate the relationship between genome size and traits. Phylogenetic regression was carried out using a phylogenetic tree, in which internal branches were all multiplied by λ and tip branches were left at their original length (Revell, 2010). Regarding the phylogenetic signal, if λ = 0, this is equal to ordinary (non-phylogenetic) least-squares regression (OLS), and OLS supposes a star phylogeny that residual variation is distinct among species; if λ = 1, PGLS and phylogenetic independent contrasts (PIC) regression are functionally equivalent (Felsenstein, 1985); if 0 < λ < 1, the OLS and the PIC methods are unsuitable, due respectively to their under- and overestimation of the impact of phylogeny (Hernández et al., 2013). In the present study, regression modelling with three types was tested, with λ equal to 0, 1 and estimated values using the R package CAPER. Furthermore, seed mass, altitude and 19 bioclimatic variables served as dependent variables, whereas genome size served as the independent variable (Orme et al., 2012). Median values for the 19 bioclimatic variables were obtained from WorldClim (http://worldclim.org). Correlation analysis involving both OLS and PGLS models were performed using the CAPER package.

RESULTS

Chromosome counts and ploidy levels

Of the 65 Allium species includes in this study (Supplementary Data Table S2), 42 are diploids, four are triploids, 15 are tetraploids, three are hexaploids and one is octaploid. Polyploids in Allium prefer to grow at higher altitude than diploids (Fig. 1). No correlation between ploidy level and latitude was observed (Fig. S2).

Fig. 1.

Boxplots illustrating the variability of diploidy and polyploidy with altitude. The outlined central box depicts the middle 50 % of the data extending from the upper to lower quartile; the horizontal bar is at the median. The ends of the vertical lines indicate the minimum and maximum data values, unless outliers are present. The P value indicates significant difference between diploid and polyploid groups at different altitude (P = 0.0064 < 0.01, t-test).

Genome size variation in Allium

Figure 2 presents the genome size variability among 62 Allium species, including 39 diploid species, four triploid species, 15 tetraploid species, three hexaploid species and one octaploid species. The Allium genome size ranges from 2.61 pg (A. rude) to 55.78 pg (A. cyathophorum var. farreri), a 21.37-fold variablility (Supplementary Data Table S2). The somatic chromosomes are shown in Fig. S3. Of the diploid species, the estimated genome size ranges from 7.35 pg (A. chrysocephalum; 2n = 16) to 55.78 pg (A. cyathophorum var. farreri; 2n = 16) (7.59-fold), with a mean of 18.08 pg. Among the triploids, the estimated genome size ranges from 5.69 pg (A. korolkowii; 2n = 24) to 24.34 pg (A. ovalifolium; 2n = 24) (4.28-fold), with a mean of 13.21 pg. For the tetraploids, the estimated genome size ranges from 4.20 pg (A. ramosum; 2n = 28) to 16.91 pg (A. bidentatum; 2n = 32) (4.03-fold), with a mean of 9.62 pg. Among the hexaploids, the estimated genome size ranges from 2.97 pg (A. fasciculatum; 2n = 42) to 7.55 pg (A. polyrhizum; 2n = 54) (2.54-fold), with a mean of 5.74 pg. Allium rude is the only octaploid species included in this study, and it has the smallest genome (2.61 pg). The average genome size is greater for the diploids than for the triploids, tetraploids, hexaploids and octaploids. Moreover, genome size tends to decrease with increasing ploidy levels (Fig. 3).

Fig. 2.

Genome size estimates for 62 Allium species. Histogram bars represent the average of population means. Error bars represent the standard error of the population means. Species exhibiting cytotypic variation (diploidy, triploidy, tetraploidy, hexaploidy and octaploidy) are indicated.

Fig. 3.

Boxplots illustrating the variability in genome size at different ploidy levels. The outlined central box depicts the middle 50 % of the data extending from the upper to lower quartile; the horizontal bar is at the median. The ends of the vertical lines indicate the minimum and maximum data values, unless outliers are present.

Patterns of genome size evolution in Allium

The ILD test revealed significant incongruency between the ITS and chloroplast DNA (cpDNA) (rpl32-trnL, atpB-rbcL, matK, rps16 and trnQ-rps16) datasets (P = 0.01). Therefore, a combined analysis of ITS and cpDNA sequence data was not performed. Figure 4 presents the Allium phylogenetic tree with genome sizes and ploidy levels determined on the basis of ITS. The reconstructed Allium ancestral genome is estimated to be 18.781 pg (Fig. 5). The variability in the cpDNA data is insufficient for resolving the relationships within Allium (Supplementary Data Fig. S4). The following analysis is based largely on the phylogenetic reconstruction using the ITS dataset.

Fig. 4.

Molecular phyloneny of Allium used in the PGLS analysis. Genome size and ploidy levels mapped on the ITS phylogenetic tree of 71 Allium species. Leucocoryne pauciflora, Tristagma uniflorum, Tulbaghia fragrans, Ipheion uniflorum, Nothoscordum gracile and Dichelostemma multiflorum were used as outgroups.

Fig. 5.

Reconstruction of the Allium ancestral genome; different genome sizes are represented by different colours. #Genome sizes are not available.

Pagel’s λ is greater than 0 (λ = 0.0003), implying that genome size is weakly correlated with phylogeny in Allium (Table 1). The κ value (κ < 1, Table 1) and the δ value (δ < 1, Table 1) respectively indicate that the mode and tempo of the Allium genome size evolution are gradual evolution and adaptive radiation.

Table 1.

Maximum-likelihood estimates for the parameters lambda, kappa, and delta, calculated according to the generalized least square method in BayesTraits. All estimates were calculated under the assumptions of the random walk model.

Dataset	Parameter	Estimated value	Interpretation
Genome size	Lambda (contribution of phylogeny)	0.0003 (0 < λ < 1)	Weak phylogenetic signal
	Kappa (mode of evolution)	–1e+200 (κ < 1)	Gradual evolution
	Delta (tempo of evolution)	–264.451 (δ < 1)	Adaptive radiation

Correlation between genome size and seed mass

Maximum-likelihood tests of the continuous models suggested that genome size and seed mass best fit the Brownian motion model (Table 2). The PIC mode reveals a significant positive relationship between genome size and seed mass (R² = 0.211, P < 0.001; Table 3), which is consistent with our linear quantile regression (Fig. 6).

Table 2.

Model selection statistics for the evolution of Allium genome size, seed mass and altitude.

Model	Parameters	Log likelihood	k	AICc
Genome size
BM	–	–262.616	2	531.589
Lambda	λ = 1.089e–198	–264.451	3	537.508
Delta	δ = 8.596e–32	–264.451	3	537.508
Kappa	κ = 2.141e–83	–1e+200	3	2e+200
OU	–	–Inf	3	Inf
Seed mass
BM	–	–307.642	2	621.642
Lambda	λ = 1.424e–91	–307.642	3	623.889
Delta	δ = 1.268e–125	–307.642	3	623.889
Kappa	κ = 7.123e–194	–e+200	3	2e+200
OU	–	–Inf	3	Inf
Altitude
BM	–	–602.006	2	1210.371
Lambda	λ = 1.095e–49	–600.848	3	1210.301
Delta	δ = 5.649e–61	–602.006	3	1212.619
Kappa	κ = 2.835e–169	–1e+200	3	2e+200
OU	–	–	3	—

Table 3.

Summary of Allium trait correlation as estimated by three types of linear regression models: ordinary least-squares regression (OLS: i.e. non-phylogenetic regression), phylogenetic independent contrasts (PIC) and phylogenetic generalized least-squares (PGLS). Lh: log likelihood; b: slope; Significance levels are denoted by asterisks: *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Model	Seed mass–genome size				Altitude–genome size
	Lh	b	R  2	P	Lh	b	R  2	P
OLS (λ = 0)	–305.465	0.521	0.0749	0.0209*	–598.795	36.553	0.0933	0.00959**
PIC (λ = 1)	–334.373	1.308	0.2110	5.639e–05***	–615.514	38.382	0.077	0.0189*
PGLS (λ = MLE)	–305.290	0.465	0.0641	0.0332*	–598.215	36.946	0.010	0.0068**

Fig. 6.

Scatter plot and quantile regression revealing the relationship between genome size and seed mass, and between altitude and genome size. The dashed grey lines correspond to the quantiles (0.15, 0.30, 0.45, 0.60, 0.75, 0.90), the solid blue line represents the median fit to the data, and the red line represents the least-squares estimate for the conditional mean function.

Correlation between genome size and altitude

Maximum-likelihood tests of the continuous models indicated that genome size and altitude best fit the lambda model (Table 2). The OLS mode detects a significant positive relationship between genome size and altitude (R² = 0.093, P < 0.01; Table 3), which is consistent with our linear quantile regression (Fig. 6).

Correlation between genome size and bioclimatic variables

Under the PGLS model, the PGLS and OLS analysis revealed no correlation between genome size and 19 bioclimatic variables (Table 4).

Table 4.

Summary of the correlations between bioclimatic variables and genome size as estimated by ordinary least squares regression (OLS) and phylogenetic generalized least squares (PGLS) analyses. Significance levels are denoted by asterisks: *, P < 0.05; **, P < 0.01; ***, P < 0.001.

	R  2 (OLS)	P	R  2 (PGLS)	P
BIO1 = Annual mean temperature	0.009278	0.472	0.04109	0.127
BIO2 = Mean diurnal range (mean of monthly (max temp − min temp))	0.008793	0.4838	0.004734	0.6078
BIO3 = Isothermality (BIO2/BIO7) (*100)	0.01453	0.3674	0.01437	0.3701
BIO4 = Temperature seasonality (standard deviation *100)	0.02399	0.2457	0.02169	0.2699
BIO5 = Max temperature of warmest month	0.01934	0.2979	0.01955	0.2951
BIO6 = Min temperature of coldest month	6.981e–06	0.9843	0.0002464	0.9069
BIO7 = Temperature annual range (BIO5 − BIO6)	0.03088	0.187	0.02712	0.2167
BIO8 = Mean temperature of wettest quarter	0.01755	0.3215	0.01878	0.305
BIO9 = Mean temperature of driest quarter	0.001019	0.812	0.002945	0.6858
BIO10 = Mean temperature of warmest quarter	0.01829	0.3115	0.02048	0.2838
BIO11 = Mean temperature of coldest quarter	5.156e–08	0.9987	0.000427	0.8776
BIO12 = Annual precipitation	0.008087	0.502	0.002466	0.7112
BIO13 = Precipitation of wettest month	0.001545	0.7696	0.0003854	0.8837
BIO14 = Precipitation of driest month	0.0008474	0.8283	0.001007	0.8131
BIO15 = Precipitation seasonality (coefficient of variation)	0.0002819	0.9005	0.0007633	0.8369
BIO16 = Precipitation of wettest quarter	0.001496	0.7731	0.0001141	0.9366
BIO17 = Precipitation of driest quarter	0.002072	0.7344	0.001209	0.7955
BIO18 = Precipitation of warmest quarter	0.001566	0.768	0.0001568	0.9257
BIO19 = Precipitation of coldest quarter	0.0006486	0.8495	0.0005248	0.8645

DISCUSSION

Polyploidy evolution of Allium on the Qinghai–Tibetan Plateau

Our results regarding the taxonomic distribution of ploidy levels in Allium confirm that of the 65 Allium species examined in this study, 42 are diploids and the rest are polyploids. Cytotypes show an interesting geographical structure, with Allium polyploids preferentially growing at higher altitude than diploids (Fig. 1); there is no correlation between ploidy level and latitude (Supplementary Data Fig. S2). This supports the hypotheses that polyploids are better adapted to extreme climates than diploids. However, some Allium diploids (e.g. A. tubiflorum, A. chrysanthum, A. prattii and A. atrosanguineum) are distributed at high altitude (Jing et al., 1999; Xue et al., 2000; Tang et al., 2005; Zhang et al., 2009; Table S1), and some Allium polyploids (e.g. A. polyrhizum, A. macrostemon, A. senescens and A. strictum) grow at low altitude (Zhu and Xu, 1999; Tang et al., 2005; Table S1). This may be because the dramatic climatic oscillations of the Quaternary caused some species on the Qinghai–Tibetan Plateau to retreat to low-altitude refugia, whereas some species survived in the on the platform (Qiu et al., 2011). Expansion of cooler and drier habitats was caused by global climate fluctuations, with several species, including Allium species, colonizing these habitats to diversify the local flora from the late Miocene to Pliocene (Xie et al., 2019; Han et al., 2020). Polyploidization events are important for the ecological radiation of Allium (Han et al., 2020). For instance, Allium przewalskianum (tetraploidy) resulted from at least eight independent autopolyploidization events that enabled it to colonize and survive in the arid conditions of the Qinghai–Tibetan Plateau (Wu et al., 2010). The small distribution ranges of some Allium species may be attributed to temporal constraints (i.e. their recent origin) (Paule et al., 2017).

Genome size variation and correlation with phylogeny

In the current study, we first comprehensively analysed the Allium genome size variation and evolution on the Qinghai–Tibetan Plateau. We present the results of our initial analysis of phylogenetic trends and the correlation between seed mass/altitude and genome size variation based on the well-resolved ITS phylogeny. Genome sizes vary substantially among the Allium diploids, triploids, tetraploids and hexaploids. A previous study showed that a large proportion of the Allium genome (44–65 %) consisted of repetitive sequences (Ranjekar et al., 1978), especially moderately repetitive sequences (Narayan, 1988). These repeating DNA segments were interspersed with lowly repetitive sequences during the evolution of various species (Narayan, 1988). Allium had one of the largest genomes among taxa in the alpine subnival belt in the Hengduan Mountains, and recent cytological analysis indicated that diploids were common, implying large genome size may be related to chromosome doubling fusion or duplicate gene copies at the chromosome level, not simply chromosome doubling (Sun, 2020). Remarkably, Allium genome size tends to decrease with increasing ploidy levels. This might reflect genome contractions after polyploidization, which is a general trend observed in angiosperms (Leitch and Bennett, 2004). Deletion of repetitive DNA sequences and TE dynamics were probably the most important factors influencing the genomic contraction (Bennetzen et al., 2005). There is cytological evidence of the loss of chromosomal segments following polyploidization. A previous investigation showed that telomeric heterochromatin of Secale cereal lost up to 100 % (Gustafson and Bennett, 1982). Therefore, repetitive sequences and chromosome doubling fusion may be the main reasons for Allium genome size variation.

Obvious differences in genome size evolution are observed among the three well-supported Allium clades (Fig. 4), which is consistent with previous studies (Hauenschild et al., 2017; Xie et al., 2020). The genomes of Clade I Allium species show minimal variations, with estimated genome sizes that were similar to the estimated size of the reconstructed ancestral genome (18.781 pg). Two noteworthy exceptions are the diploid A. wallichii (26.17 pg) and the hexaploid A. fasciculatum (2.97 pg). Accordingly, the A. wallichii genome is larger than the genomes of most diploid species. The considerable abundance of repetitive sequences in the Allium genome (44-65 %) (Ranjekar et al., 1978) suggested that the relatively large diploid A. wallichii genome may have resulted from an increase in the number of repetitive sequences. The small genome of hexaploid A. fasciculatum may be related to DNA loss during polyploidization. The genomes of most of the Clade II Allium species have experienced genomic expansion relative to the ancestral reconstructed estimates (Fig. 4). Interestingly, these species are diploids, with the exception of the triploid A. ovalifolium, and they grow at high altitude (2245–4564 m) except for A. maowenense (1500 m). A significant positive relationship between genome size and altitude was tested by the OLS model and the linear quantile regression model (Table 3; Fig. 6). The correlation between genome size and ecological factors is discussed further below. Additionally, adaptive radiation may have occurred in the species in this clade. This possibility is also addressed below. The genomes of the Allium species in Clade III have experienced genome contraction compared to the ancestral reconstructed estimate (Fig. 4). Surprisingly, A. cyathophorum var. farreri, which is a diploid, has the largest genome among these species, possibly because of an increase in the abundance of repetitive sequences. The precise mechanism underlying this change in genome size will need to be elucidated in future investigations. Clade III includes diploids, triploids, tetraploids, hexaploids and octaploids. DNA loss during polyploidization may be the main cause of the genome contraction in Clade III. The differences among the three Allium clades are interesting and confirm that the evolutionary trends in genome size variation are inconsistent with the previous result for genome size expansion (Hawkins et al., 2009).

Phylogenetic signal and evolutionary pattern

Phylogenetic signal varies among phylogenetic/taxonomic levels (Kamilar and Cooper, 2013). At higher taxonomic levels, genome size is highly dependent on phylogeny (Beaulieu et al., 2007; Knight et al., 2010; Whitney et al., 2010). Within individual taxonomic groups, a strong phylogenetic signal associated with genome size changes has been observed in a few studies, including the following genera: Orobanche (λ = 1; Weiss-Schneeweiss et al., 2006), Chenopodium (λ = 0.987; Mandák et al., 2016), Filago (λ = 0.934; Andrés-Sánchez et al., 2013), Hieracium (λ = 0.908; Chrtek et al., 2009) and Primulina (λ = 0.900; Kang et al., 2014). In the current study, the weak phylogenetic signal associated with the Allium genome sizes implies that genome size variation is minimally related to phylogenetic divergence (Table 1). A strong phylogenetic signal implies that genome size is not strongly influenced by selective pressure. Taxon-specific responses possibly obscure phylogenetic signal if genome size has been directly impacted by selective pressures (Lysak et al., 2009). This may also reflect a gradual evolution of the trait, during which selective pressures or adaptive radiation are important (Andrés-Sánchez et al., 2013).

The κ data in this study (κ < 1; Table 1) indicate that genome size evolves gradually and that this evolution is greater in the short branch. For other taxa, a gradual genome size evolution has been reported for Hieracium (Chrtek et al., 2009), Filago (Andrés-Sánchez et al., 2013) and Primulina (Kang et al., 2014), whereas punctuated evolution was reported for Orobanche (Weiss-Schneeweiss et al., 2006), Liliaceae (Leitch et al., 2007) and Allium subgen. Melanocrommyum (Gurushidze et al., 2012). Surprisingly, the results of this study are inconsistent with those of the previous examination of Allium subgen. Melanocrommyum. This discrepancy may be explained by the differences in the geographical distributions and habitats between the studies. More specifically, the Allium species in the current study are from the Qinghai–Tibetan Plateau, whereas the previously investigated Allium subgen. Melanocrommyum species were from Germany. Additionally, previous research indicated that genome size changes immediately occurred after speciation, during which punctuated evolution was more likely than a gradual evolution in genome size (Gregory, 2004). High stability in intraspecific genome size (Greilhuber, 1998; Gregory, 2004) and the rapid genome reorganization after speciation (Rieseberg et al., 1995) suggest punctuated evolution is more realistic than gradual evolution. However, this will need to be verified by analysing genome size evolution and phylogenetic relationships in more taxa.

Our δ data indicate adaptive radiation during the evolution of Allium genome size, involving a temporally early trait evolution or ‘early burst’ (δ < 1; Table 1). A rapid rate of trait change at the early evolutionary phase and a slower rate at later phases may reflect adaptive radiation ((Pagel et al., 2004)). Coincidentally, a previous study demonstrated that the Allium QTP-clade diversified most from the start of the radiation (2.34 Mya), with global cooling and climate oscillations in the Quaternary identified as the major reasons for the increased speciation rates (Hauenschild et al., 2017). Hence, Allium species on the Qinghai–Tibetan Plateau are highly diverse. Furthermore, our result is in accordance with the weak phylogenetic signal associated with Allium genome size. Thus, Allium genome size is directly affected by selective pressure, which is a feature of early adaptive radiation.

Correlated evolution of traits

Correlations among the evolution of genome size, seed mass and ecological factors were evaluated according to OLS (non-phylogenetic), PIC (phylogenetically independent contrasts) and PGLS (phylogenetic generalized least-squares) models. Our LRT results indicated the genome size and seed mass data fit the PIC model better than the OLS and PIC models. However, the genome size and altitude data fit the OLS model better than the PIC and PGLS models (Table 2).

Genome size and seed mass

Genome size variation affects numerous cellular parameters, including chromosome size, nuclear volume and cellular volume (Bennett, 1987). These are considered to be nucleotype effects (Bennett, 1972), which lead to variations in phenotypic traits (e.g. seed mass) (Beaulieu et al., 2008). Because seed mass variations are agronomically and ecologically important, clarifying the underlying molecular mechanism is critical (Beaulieu et al., 2007). Several previous studies revealed the positive correlation between genome size and seed mass among different accessions in one taxon and across species within the same genera or families (Caceres et al., 1998; Leishman, 1999; Knight and Ackerly, 2002). However, no significant correlation was detected between these two traits in 92 species in the subgenus Phillodineae of the genus Acacia (Gallagher et al., 2011).

In the present study, Allium genome size is positively correlated with seed mass according to the PIC regression and linear quantile regression (Table 3; Fig. 6), which is consistent with the findings of a previous study (Ohri et al., 1998). A general trend among species was that large genomes were present in large cells that divide relatively slowly (Knight and Beaulieu, 2008). Previous reports implied that genome size and guard cell size were strongly correlated across angiosperms (Beaulieu et al., 2008). Cells in the seeds of diploid and tetraploid species may differ dramatically in terms of size (Linkies et al., 2010). Seed mass is a key component of plant adaptation (Wang et al., 2020), and it is more highly correlated with plant size than with other factors (e.g. dispersal mode and environmental conditions) (Moles et al., 2005a, b). The correlation between plant size and seed mass may suggest there were common genetic components that determined seed size, plant size and specific plant organ size (Linkies et al., 2010). During seed development, transcription factors may play an important role in seed mass (Linkies et al., 2010). For example, in a previous study on seed phenotypes, the seed mass of the shb1-D mutant in Arabidopsis thaliana (L.) Heynh. increased largely because of coordinated endosperm enlargement as well as embryo development; additionally, compared with a wild-type control, the mutants had larger cells, accumulated more proteins and fatty acids, and showed delayed embryo development (Zhou et al., 2009). TEs were also probably important for genome size evolution (Bennetzen et al., 2005). The effects of genome size changes on the seed developmental process may be similar to the effects of loss-of-function mutations to transcription factors, and this might lead to changes in cell proliferation and developmental timing (Linkies et al., 2010).

Genome size and ecological factors

Genome size is usually regarded as an evolutionary character, suggesting that variation is not random, but generally occurs in response to external environmental fluctuations (Pellicer et al., 2018). A related issue is whether genome size varies predictably as a function of ecological factors, indicating ecological constraints on genome size. Associations between genome size and ecological factors have been explored in some plant groups at various taxonomic levels (Knight et al., 2005; Díez et al., 2013; Carta and Peruzzi, 2016; Qiu et al., 2019). In the present study, a significant positive relationship between Allium genome size and altitude was revealed by the OLS model and linear quantile regression (Table 3; Fig. 6). This is consistent with the results of a previous study that revealed the patterns in genome size variation along ecogeographical gradients (Guo et al., 2018). Our analysis of the relationship between Allium genome size and 19 bioclimatic variables did not detect any significant correlation (Table 4). Thus, ecological factors are probably relatively unimportant for determining Allium genome size, which is in accordance with the results of an earlier analysis of A. oleraceum (Duchoslav et al., 2013). Yet, we cannot exclude the possibility that selective pressure is an important factor.

The Qinghai–Tibetan Plateau is characterized by low air pressure, low temperature, high UV irradiation, and abundant precipitation in the form of rain and snow. The positive correlation between genome size and altitude may be related to the ability of Allium species to grow at low temperatures and their frost resistance. This ability to grow in cold conditions is facilitated by growth (i.e. cell division) in the previous season and by cell expansion during the early part of the season with low temperatures (Grime and Mowforth, 1982). Species that are relatively resistant to cold stress tend to have large genomes (Macgillivray and Grime, 1995; Knight and Ackerly, 2002). Large genomes lead to an increased ecological burden because of the greater need for macro-nutrients, especially nitrogen and phosphorus, to produce nucleic acids (Vitousek et al., 2010; Šmarda et al., 2013). The large genomes of alpine plants may be acceptable because soils at high altitudes are generally rich in phosphate (Körner, 1989). Nevertheless, phosphate is often a limited nutrient for DNA biosynthesis and plant growth (Raven et al., 1986). A previous study detected a correlation between large genomes and phosphate-rich soil (Hanson et al., 2001), suggesting there may be a correlation between alpine habitats and large genomes. Notably, there is also evidence of a relationship between UV-damage and genome size (Sparrow and Miksche, 1961). In the present study, we showed that plants with large genomes growing at high altitude may be affected by cell expansion early in the preceding favourable season, and then were subsequently exposed to low temperatures, phosphate-rich soil and increased UV-damage. Future studies should more precisely characterize the above-mentioned mechanisms to further clarify the evolutionary forces driving genome size variation on the Qinghai–Tibetan Plateau.

CONCLUSIONS

Genome size of Allium from the Qinghai–Tibetan Plateau varies considerably among species and within diploids, triploids, tetraploids, hexaploids and octaploids. This variability is probably influenced by repetitive sequences and chromosome doubling fusion. Genome size tends to decrease with increasing ploidy levels, probably because of DNA loss after polyploidization. The cytotypes in this study show an interesting geographical structure, with Allium polyploids preferentially distributed at higher altitude than diploids. The evolutionary pattern of Allium genome size is gradual evolution, followed by adaptive radiation. The Allium phylogenetic clades reflect genomic contraction, expansion and relative stasis. The weak phylogenetic signal for Allium genome sizes suggests that genome size variation is little associated with phylogenetic divergence. Furthermore, genome size is positively correlated with altitude. However, there is no correlation between genome size and 19 bioclimatic variables. Therefore, Allium genome size may be influenced by selection pressures on the Qinghai–Tibetan Plateau (i.e. low temperature, high UV irradiation and high soil phosphate).

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Fig. S1. Histograms of the relative flow-cytometric genome size measurements in taxa of Allium. Fig. S2. Boxplots illustrating the variability of of diploidy and polyploidy with latitude. Fig. S3. Somatic metaphase chromosome numbers in Allium species investigated in this study. Fig. S4. Phylogenetic tree of Allium based on the combined chloroplast sequences. Table S1. Voucher information of Allium in this study. Table S2. Chromosome numbers, ploidy levels, seed mass, genome size and chromosome length of Allium. Table S3. Accession numbers for DNA sequence of Allium species and the outgroup used for phylogenetic reconstruction.

FUNDING

This work was supported by the National Natural Science Foundation of China (31800189), the Natural Science Foundation of Yunnan Province (202001AU070071), and the Second Tibetan Plateau Scientific Expedition and Research (STEP) programme (2019QZKK0502).