A comparison of plant chromosome number variation among Corsica, Sardinia and Sicily, the three largest Mediterranean islands

Abstract

Chromosome counts of the vascular floras of Corsica, Sardinia, and Sicily were compared quantitatively, based on data stored in the online database “Chrobase.it”, and analysed to determine any trend associated to the latitudinal range. Mean chromosome number (CN) ± SD, frequency of B-chromosomes occurrence (fB), frequency of odd CN, and ICNH (Index of Chromosome Number Heterogeneity) were calculated for each island’s flora. The relative frequency of putative even ploidy levels was also calculated for the most frequent base numbers (x = 7, 8, 9) as the percentage of cytotype frequency value relative to the total sum of cytotypes. Our results show that ICNH and mean CN are lower in Sicily and higher in Sardinia and Corsica; furthermore, the relative frequencies of cytotypes of putative even polyploid series decrease with increasing ploidy levels, which can be modelled by an exponential decay of the general form f(x) = e^(ax+b), with different parameters in the different islands. Such results are in agreement with previous researches showing an increase of mean CNs across Italy and Europe and a similar trend, modelled by an exponential decay, as regards the ploidy levels of endemic taxa of the Italian flora.

Keywords:

• chromosome number database
• cytogeography
• France
• Italy
• polyploidy

Introduction

The quantitative analysis of chromosome number (CN) variation in plants has been a long-studied research subject. Pioneer scholars inferred that polyploidy increases with distance from the equator both in boreal (Löve and Löve 1957; Hanelt 1966) and austral hemispheres (Hair 1966; Hanelt 1966), and that the lowest percentages of polyploids occur in subtropical and warm-temperate regions (Stebbins 1971).

The interest in this research field has been revived by the availability of large datasets stored in electronic format, such as Chrobase.it (Bedini et al. 2010 onwards), Chromosome Number Database (Góralski et al. 2009 onwards) and many others (see Peruzzi and Bedini 2014 for a detailed list of both plant and animal groups), besides the very recently established Chromosome Counts Database (CCDB), online since the end of November 2014 (Rice et al. 2015).

Plant CN databases have been used to probe systematic comparisons of geographical or taxonomical groups of plants (Gacek et al. 2011; Peruzzi et al. 2011; Bedini et al. 2012a, 2012b, 2012c; Góralski et al. 2013, 2014). For instance, a significant increase in mean CN was detected along a bioclimatic/latitudinal gradient in Italy (Bedini et al. 2012a) and in Southern to Central Europe (Peruzzi et al. 2012); the same authors showed that specific orders and families are characterized by peculiar CN variation patterns (Bedini et al. 2012b). In combination with animal CN databases, plant CN databases allowed the first comparison of CN variations in selected plant and animal groups (Peruzzi et al. 2014).

In this work, we exploited the wealth of data stored in Chrobase.it (Bedini et al. 2010 onwards; Garbari et al. 2012) to answer the following questions relevant to the biogeography of Central Mediterranean islands. (a) Are patterns of CN variations in vascular floras of Central Mediterranean islands related to their latitudinal range? (b) How are such patterns – if any – linked to the Central Mediterranean and Central European context?

Material and methods

Data source

CNs were taken from the online database Chrobase.it (Bedini et al. 2010 onwards). Chrobase.it stores the available karyological information about Italian vascular flora, in terms of CN (2n and/or n) and B-chromosome occurrence, along with main geographic-administrative data and literature references (Bedini et al. 2012a). By querying the database on geographic provenance of count data, we were able to extract chromosome counts recorded for three Mediterranean islands: Sicily, Sardinia, and Corsica. The main features of the available chromosome counts available for each island and their geographical localization are reported in Table 1.

Table 1. Geographical localization and features of the available chromosome counts for the three largest Mediterranean islands.

	Sicily	Sardinia	Corsica
Coordinates (dd) of centroid	37.5908° N14.1506° E	40.0857° N9.0310° E	42.1515° N9.1056° E
Counts (n)	1357	798	393
Cytotypes (n)	991	530	261
Taxa (n, % of flora)	879 (28%)	438 (17%)	228 (8%)
Range (min–max CN)	6–174	6–182	8–216
fOCN	0.0343	0.0754	0.0517
fB	0.0353	0.0716	0.0080
ICNH	18.55	25.72	24.17
Mean ± SD	27.27 ± 17.67^a	32.07 ± 22.58^b	33.80 ± 28.85^b

Means differ significantly (ANOVA: df = 2, F = 14,120, p < 0.001); Tukey post-hoc pairwise comparisons are given as superscript letter (same letter = no significant differences).

The total number of cytotypes retained for each dataset (SI: Sicily; SA: Sardinia; CO: Corsica) was obtained by excluding counts in multiple copy (i.e. the same chromosome number for the same species). Any n counts (a minority in the three datasets) were transformed to 2n. Families circumscription followed APG III (2009), Chase and Reveal (2011), Christenhusz, Reveal, et al. (2011) and Christenhusz, Zhang, et al. (2011).

Data analysis

Similarly to Bedini et al. (2012a, 2012b), the following data were calculated for each dataset: mean CN ± SD, frequency of B-chromosomes occurrence (fB), and frequency of odd CN, not considering B-chromosomes (fOCN). In addition, the ICNH (Index of Chromosome Number Heterogeneity) was calculated for each dataset following Peruzzi et al. (2014). According to these authors, the ICNH is calculated according to the following formula:

(1)

where a is CN, b is SD of CN, c is %(fB+fOCN). Its value can vary from 0 (if no CN variation occurs in a group, i.e. that group would be represented by a vertical line only varying in length) to +∞, albeit very high values can be reached only theoretically. This value corresponds to the square root of the area of the ideal triangle built in a three-variables radar plot, where the vertices of the triangle are defined by mean CN, SD of CN, and %(fB+fOCN). This triangle gives a graphical representation of CN variation in a group.

ANOVA was used to test statistical differences in CN among considered groups, after checking homoscedasticity (Levene test), and Tukey HSD test was used for multiple comparisons. R software, version 3.0.3, with stats package (R Core Team 2014) was used to perform these tests.

For the three most frequent basic chromosome numbers, x = 7, x = 8 and x = 9, we first computed the matching putative artioploid series from 2n = 2x to 2n = 10x; for each series, we extracted from each dataset (SI, SA, CO) the frequency of cytotypes (FC) matching each series’ item. Odd ploidy levels were not considered since they were very rare in our dataset, with the only exception of 2n = 27 (see Results).

As the actual basic chromosome number is unknown for most of the taxa of the islands’ floras, we used FC as proxy of the true frequency of artioploid cytotypes (the implications of this assumption are considered in the Discussion) and referred to it as “putative” frequency of artioploid cytotypes. We were thus able to calculate the putative relative frequency for each series (x = 7, 8, 9) as the percentage of each FC value relative to the total sum of cytotypes of the same series.

Finally, we plotted putative relative frequency against ploidy level and tested their relationship by means of nonlinear models based on least-squares estimates in R software (nls function; Rossiter 2009): power, logarithmic, and exponential curves were fitted to the plotted data and their goodness of fit was assessed by analysis of residuals and the Kolmogorov–Smirnov test as suggested by Straume and Johnson (2010); pairwise fit comparison was performed by F-test. Both tests were run with R software, version 3.0.3, stats package (R Core Team 2014).

Results

Table 1 summarizes the number of counts, cytotypes and taxa for each dataset, along with the range and mean of CN number, fB, fOCN and ICNH (see also Figure 1).

Figure 1. Comparison of three-variables radar plot, among Corsica, ICNH = 24.17; Sardinia, ICNH = 25.72; Sicily, ICNH = 18.75.

The species with the highest number of cytotypes are Genista monspessulana (L.) L.A.S.Johnson (2n = 46, 48 + 0-4B) in Sicily; G. sulcitana Vals. (2n = 18 + 0-2B; 27 + 0-2B) in Sardinia; and Crocus minimus DC. (2n = 24, 25, 26, 27, 28, 29, 30) in Corsica. The highest CN is 2n = 174 in Sicily (Pteris vittata L.); 2n = 182 in Sardinia (Colchicum gonarei Camarda); and 2n = 216 in Corsica (C. corsicum Baker). At the lowest end of the CN range there are Hypochaeris cretensis (L.) Bory & Chaub., 2n = 6, in Sicily and Sardinia; and Hypochaeris robertia (Sch.Bip.) Fiori, 2n = 8, in Corsica.

Figure 2 shows a plot of the frequency of even CN for Sicily (a), Sardinia (c), Corsica (e); in general, CN frequency is low for odd CNs, with few exceptions; as regards even values, their frequency is low when CN < 2n = 14 and CN > 2n = 48; it reaches intermediate values for 2n = 18 < CN ≤ 2n = 48; it peaks for 2n = 12 < CN < 2n = 20; in particular, in Sicily the modal CN is 2n = 18, followed by 2n = 14 and 2n = 16, while in SA the modal CN is still 2n = 18, but the second highest is 2n = 36, the third 2n = 16; in CO 2n = 18 is followed by 2n = 16 and 2n = 28.

Figure 2. Scatter plots of frequency of sporophytic CNs and of relative frequency of putative ploidy levels of the most frequent base numbers (x = 7, 8, 9) for Sicily (a, b), Sardinia (c, d), and Corsica (e, f). Dashed lines are the best fit curve. Expression and goodness-of-fit parameters are given for each plot. S = squared sum of residuals; p = probability of the null hypothesis of the Kolmogorov–Smirnov test.

The even putative polyploid series of the three most common base numbers (x = 7, 8, 9: Figure 2a, c, e; black squares, diamonds, and triangles respectively) follow a decreasing trend. Pairwise comparison of goodness of fit revealed that the experimental data are best fit by an exponential decay function of the general form f(x) = e^(ax+b) (data not shown) with different a, b parameters for the three islands, determining the different values for diploid and putative polyploid cytotypes. The three polyploid series are plotted as relative frequency (%) of the putative ploidy level for Sicily (Figure 2b), Sardinia (Figure 2d), and Corsica (Figure 2f), along with the best fit curve.

Discussion

The Sicilian dataset coverage is about 28% of vascular plants, the Sardinian dataset about 17% (based on data summarized by Peruzzi 2010), and the Corsican dataset about 8% (Jeanmonod and Gamisans 2007); therefore, our results might not be fully representative of the islands’ floras.

The use of a putative ploidy level, relying on a simple algorithm to automatically associate somatic CNs to basic numbers, might be questioned because it does not necessarily reflect the true ploidy level of a given cytotype: for instance, the somatic number 2n = 56 might be interpreted as 2n = 7x = 56 or 2n = 8x = 56, i.e. as the octoploid of x = 7, or as the eptaploid of x = 8, or as the result of one or more polyploidization events followed by disploidy as hinted for angiosperm groups by Rice et al. (2015). While the use of the true x for each taxon would yield more accurate results, this datum is not readily available for all taxa. Plus, our results, in agreement with previous studies (Bedini et al. 2012a, 2012b, 2012c) support the following conclusions:

(1)	odd sporophytic CNs are less frequent than even ones; and
(2)	high sporophytic CNs are less frequent than low ones.

For this reason, any ambiguous case has been solved by attributing conflicting CNs to the lowest even ploidy level within the considered series, although odd ploidy levels or higher ploidy levels were theoretically possible.

Therefore, we believe that any bias due to the overlooking of odd or high ploidy level cytotypes bears little significance in the light of the high number of data analysed in our datasets.

The increase in mean CN with latitude is also in agreement with previous hypotheses (Löve and Löve 1957; Hanelt 1966) and with the results of a quantitative analysis of vascular floras of Italy, Poland, and Slovakia (Peruzzi et al. 2012). However, in this work we were able to demonstrate the increase at a much finer scale within the Mediterranean Sea, paralleled by an increase in ICNH (Sicily versus Corsica/Sardinia). Concerning the latter parameter, according to the classification made by Peruzzi et al. (2014), the heterogeneity value obtained for Sicily can be classified as “Low” (< 20), while those obtained for Corsica and Sardinia can be classified as “High” (> 20).

In addition, we were able to fit a curve to the data, showing that putative polyploid cytotypes are linked to the respective diploid cytotypes by a fixed ratio following an exponential decay. The general shape of the “ploidy level” curve is in agreement with previous studies carried out on the Italian flora (Bedini et al. 2012a) and on its endemic component (Bedini et al. 2012c). Furthermore, such a curve might fit well the ploidy index (pix) distribution reported by Rice et al. (2015, Figure 2) concerning plants harbouring intraspecific chromosome number variation: the tetraploid (pix = 2), hexaploid (pix = 3), octoploid (pix = 4), decaploid (pix = 5), and dodecaploid (pix = 6) visually appear to be linked by an exponential decay relationship, but the authors did not elaborate on this issue.

Given the increasing evidence of this peculiar type of relationship among ploidy levels, further studies need to clarify its biological meaning, hitherto unexplained.

Disclosure statement

No potential conflict of interest was reported by the authors.