Page 9 - Voda_Dapporto_alii_2016

www.nature.com/scientificreports/

nestedness and turnover as the mean value of the ratio between the faunistic dissimilarities obtained by nest-
edness and Sørensen indexes (see Supplementary Information and Supplementary Data) . This ratio has been
41
computed for all species, as well as for the rare and widespread species groups, separately. We tested for the effect
of species dispersal tendency and their frequency at source on their occupancy on islands. Measurements of the
dispersal tendency in butterflies are complex and, in previous studies, have been mostly based on the agreement
among subjective evaluations made by experts, and less commonly by using objective species traits. Here we com-
bined the indexes provided by four studies 42–45 by standardizing their range between 0 (low dispersal) and 1 (high
dispersal) and computed, for each species, an average dispersal tendency based on the available measurements. To
estimate the frequency at source we counted the number of cells of 0.1 × 0.1 degrees between 34° and 40° latitude
and 7° and 18° longitude, where each of the 32 investigated species has been reported. As standardized sources we
46
used “CKmap2000”, an online checklist of the Italian butterfly fauna , and the database of the Butterfly Diversity
and Evolution Lab at The Institute of Evolutionary Biology (CSIC-UPF), Spain. The frequency of each species on
islands was regressed against dispersal tendency and frequency at source, and the relative importance of these
factors was tested by hierarchical partitioning of variance.

Genetic analyses and identification of study units. Using standard sequencing procedures (see
Supplementary Information) we obtained cytochrome c oxidase subunit I (COI) sequences for 1044 specimens
(Supplementary Table S2) from the study islands and from five surrounding areas: southern Italy (Calabria), east-
ern Sicily (>14° longitude), western Sicily (<14° longitude), Tunisia and Algeria. We considered only the islands
for which we had available more than 80% of the reported fauna. This led to the exclusion of Linosa and left ten
islands and a total of 15 areas to be analysed.
The butterfly species currently recognized by taxonomists show different levels of intraspecific genetic diver-
gence and they can have an unbalanced contribution to the overall biogeographic pattern 31,47 . To reduce this
bias, we identified as units the groups of individuals having COI p-distances less than 3%, a measure that, in
Lepidoptera, was reported to separate more than 90% of recognized sister species . By applying this threshold we
48
identified 29 units (termed species) closely matching the taxonomy proposed by Fauna Europaea (www.faunaeur.
org). The list of species is provided in Supplementary Table S1.
Overall population differentiation pattern. To examine patterns of genetic differentiation for each spe-
49
cies in the study area we constructed haplotype networks with the TCS Network method in PopART (http://
popart.otago.ac.nz). We calculated the genetic uncorrected p-distances among all sequenced specimens for each
species, as well as two measures for population differentiation: Dst and Gst :
50
Dst = Ht − Hs
where Ht represents the average intraspecific p-distances for all specimens of a given species, and Hs is the aver-
age of the intra-population p-distances. Thus, Dst represents average genetic differentiation among populations
in p-distance units. The second measure (Gst) is a standardized index defined as:
Gst = Dst/Ht
representing which fraction of the total genetic differentiation is encompassed by differentiation among popula-
tions . This index ranges from negative values to 1 (complete differentiation). Negative values (intra-area differ-
50
entiation higher than inter-area differentiation) can have different subtle meanings, but they are mostly generated
51
as a bias due to relatively small sample sizes; usually they are set to zero and we used this solution for two cases
in our study. We also calculated pairwise Gst among pairs of populations using the following formula:
Gst ij, = Dst /Ht
ij,
representing which fraction of the overall genetic diversity (Ht) is expressed by inter-population diversification
(Dst i,j ) between a given pair of areas (i and j).
By using the Gst pairwise matrices for each species, we calculated the mean of the available values of the
corresponding cells to produce the final mean Gst matrix, representing the degree of genetic differentiation
among areas based on all species. A PCoA was applied to this matrix to obtain the overall genetic pattern among
areas. Subsequently, we aligned this configuration with the geographic location of the areas by using the “pro-
crustes” analysis from the “vegan” R package (Supplementary Data). To visualize the pattern of similarity among
islands over geographic space, we projected the final configuration of average Gst values among areas in the RGB
space 24,52 using the R package “recluster”. The colour resemblance of the resulting dots is directly proportional to
the genetic differentiation among the communities. These dots were then plotted on a map, where we outlined
the − 100 m depth contour as a reliable reconstruction of land during the LGM .
20
Islands located between two genetically contrasting sources can show intermediate communities if: i) islands
host species characterized by genetic variation and individuals belong to different sources, or ii) islands host only
a reduced fraction of species characterized by genetic differentiation, and therefore look equally similar to both
sources. To visualize the pattern of genetic similarity and to compare the degree of genetic differentiation among
areas we computed for each area the mean Dst as the mean of all the Dst values for the species available for each
area. We then plotted the relationship between the PCoA configuration and the mean Dst of each area in 3D. If
the first hypothesis is met, we expect that intermediate islands will have similar mean Dst to other islands, while
if the second is true we expect that intermediate islands will have lower mean Dst values.

Analysis of genetic differentiation among populations. We tested for correlation between Dst and
both dispersal tendency and island occupancy (number of studied islands occupied by each species) using the

Scientific RepoRts | 6:28828 | DOI: 10.1038/srep28828 9