Drosophila sturtevanti

Drosophila sturtevanti View in CoL samples

A t o t a l o f 1 2 6 i n d i v i d u a l s f r o m n i n e populations of D. sturtevanti were collected from two phytophysiognomy types of the Atlantic Forest of three Brazilian geographic regions ( Table 1, Fig. 1 View Fig ). The collections were performed with the permission of the Instituto Chico Mendes de Conservação da Biodiversidade – ICMBio, a regulatory agency responsible for environmental studies in Brazil (permission number 46752). The representative specimens were stored in 100% ethanol in the collection of the Museu de Zoologia da Universidade de São Paulo, Brazil. All flies were captured using closed traps containing fermented banana bait ( Penariol et al. 2008). The identification of males was carried out by analyzing the aedeagus using specific identification keys ( Freire-Maia and Pavan 1949; Mourão and Bicudo 1967; Vilela and Bächli 1990; Souza et al. 2014). Samples were preserved in 96% ethanol and kept at 4°C until DNA extraction.

DNA extraction and microsatellite analyses

The extraction of genomic DNA was performed by individual maceration of each sampled male using the Promega kit. Eleven microsatellites (Dsturt_B, Dsturt_D, Dsturt_E, Dsturt_G, Dsturt_I, Dsturt_J, Dsturt_K, Dsturt_L, Dsturt_M, Dsturt_N and Dsturt_O) described for D. sturtevanti (Roman, B.E., Trava, B. M.; Madi-Ravazzi, L., submitted) were amplified through polymerase chain reaction (PCR) in a total volume of 25 µl, containing 0.5 µl of Taq DNA polymerase, 2.5 µl of 10x buffer, 0.2 mM dNTP, 0.2 pmol of each primer, 1.5 mM MgCl 2 and 3 ng of DNA. Touchdown PCR was performed for the Dsturt_B, Dsturt_D, Dsturt_I, Dsturt_K and Dsturt_M loci as follows: denaturation cycle at 94°C for 2 minutes; 2 repetitions of 10 cycles at 94°C for 1 minute, 65°C for 1 minute (-1°C per cycle), and 72°C for 2 minutes; ending with 18 cycles at 94°C for 1 minute, 55°C for 1 minute and 72°C for 5 minutes. For all other loci, a specific primer annealing temperature (Ta) was applied: 53°C - Dsturt_ N; 55°C - Dsturt_G; 56°C - Dsturt_E and Dsturt_L; 57°C - Dsturt_O; 65°C - Dsturt_J. The PCR conditions for these loci were as follows: 94°C for 2 minutes; 30 cycles of 94°C for 1 minute, specific Ta for 1 minute and 72°C for 2 minutes; and ending with 72°C for 5 minutes. PCR amplification products were visualized in 6% polyacrylamide gel and stained with 15% silver nitrate ( Sanguinetti et al. 1994).

Statistical analysis

Population genetic structure and diversities were assessed using 10–15 individuals from each of the nine populations. The mean number of alleles (Na), effective number of alleles (Ne), expected heterozygosity (H E), observed heterozygosity (H O), number of private alleles (Np), frequency of private alleles (Ap) and fixation index (F) were calculated for each population in GenAlEx software v.6.51b2 ( Peakall and Smouse 2006 2012). The allelic polymorphic information content (PIC) was obtained using CERVUS software v.3.0.7 ( Kalinowski et al. 2007). The mean allelic richness (Ar)—which is an unbiased measure of the number of alleles estimated independently of the sample size, allowing for comparison between different sample sizes ( El Mousadik and Petit 1996)—was calculated in FSTAT v.2.93 ( Goudet 2001). Departures from Hardy-Weinberg equilibrium (HWE) at each locus within populations were estimated in the GenAlEx software. All levels of significance were determined after a sequential Bonferroni correction for multiple tests ( Holm 1979). The FreeNA software ( Chapuis and Estoup 2007) was used to estimate null allele frequencies (An) for each locus and population following the Expectation Maximization (EM) algorithm of Dempster et al. (1977). A population bottleneck test, using adjusted frequencies for the presence of null alleles, was performed in BOTTLENECK software ( Cornuet and Luikart 1996; Piry et al. 1999) to test the occurrence of recent demographic events. The program BOTTLENECK computed the distribution of the expected heterozygosity from the observed number of alleles when given the sample size under the assumption of mutation-drift equilibrium. The simulation of the coalescent process of n genes was performed under the two-phase model (TPM), using the Wilcoxon signed-rank test, setting the parameters as 90% single-step mutations, 10% multiplestep mutations, a variance of the geometric distribution of 12, and 1,000 iterations. These settings correspond to sensible parameter values for most microsatellites, considering that fewer than 20 loci were used ( Cornuet and Luikart 1996; Piry et al. 1999).

The level of genetic differentiation among populations was verified using multiple approaches: indexes of genetic differentiation, a neighbor-joining tree, the use of a nonspatial Bayesian algorithm and analyses of molecular variance (AMOVA). The FreeNA software was used to estimate two genetic differentiation indexes based on the ENA method ( Chapuis and Estoup 2007): DC - Cavalli-Sforza and Edwards (1967) genetic distances, and F ST ( Weir 1996). For these indexes 95% confidence intervals (C.I.) were obtained using bootstrap resampling over loci. Wright (1978) qualitative classification of the genetic differentiation among populations (spatial analyses) was applied accordingly to the F ST values obtained: ‘low’ (0–0.05), ‘moderate’ (0.05–0.15), ‘high’ (0.15–0.25) and ‘very high’ (> 0.25). The correlations between the genetic differentiation above (DC and F ST) and geographic distance by the Mantel tests (with 9999 random permutations) were performed in GenAlEx software. A neighbor-joining tree ( Saitou and Nei 1987) was obtained through maximum likelihood analysis ( Felsenstein 1981) on adjusted allele frequencies (considering the presence of null alleles) of microsatellite data using PHYLIP software (version 3.7a; Felsenstein 2009). To evaluate support for the branches, a bootstrap analysis ( Felsenstein 1985) was performed 1000 times. An unrooted tree was constructed using Cavalli-Sforza and Edwards (1967) distances, using a branch and bound algorithm, the majority rule option and with random addition of populations. The cluster-based Bayesian method was performed with the software STRUCTURE v.2.3.4 ( Pritchard et al. 2000), considering the presence of null alleles and the Dsturt_I locus as X or Y linked in the genotypic matrix. For this analysis, the admixture hypothesis was used assuming the existence of correlated allele frequencies, in which each sample is partially composed of the genome of each ancestral population. Together with the allele frequency model, they allow the log likelihood L (K) for the data to be obtained. This model is considered to be the most appropriate when the a priori origin and the degree of isolation of the studied populations are unknown ( Pritchard et al. 2000). The prior probability, i.e., the probability that an individual belongs to any reference K population, is defined as l/K. The K value was fixed from 1 to 11 using 10,000 burn-in, 500,000 Markov chain Monte Carlo (MCMC) replicates after burn-in and 25 iterations. After obtaining the results, a bar graph was generated with the CLUMPAK tool ( Kopelman et al. 2015) using the best number of clusters (K) obtained by the STRUCTURE HARVESTER tool ( Earl and Von Holdt 2012), according to the Evanno test ( Evanno et al. 2005; Earl and Von Holdt 2012).

The AMOVA was run in the Arlequin software (version 3.5.2.2, Excoffier and Lischer 2010), using the adjusted genetic frequencies and 1,000 permutations, to test four assumptions about the distribution of genetic variability of Brazilian D. sturtevanti populations: 1) with no grouping; 2) grouping according to geographic regions (Northeast, Southeast and South regions); 3) grouping according to Atlantic Forest phytophysiognomies (dense ombrophilous forest and semideciduous seasonal forest); and 4) groupings according to the result of the cluster-based Bayesian method performed with STRUCTURE software. The second grouping was proposed based on a previous morphological study ( Segala 2019) using D. sturtevanti, ( Segala 2019) using D. sturtevanti and also to test correlation between genetic variability distribution and isolation by distance. The third grouping was tested based on phylogeography and microsatellite studies with D. ornatifrons ( Gustani et al. 2015; Zorzato 2015), another Neotropical forest dweeling species, which suggested that there is correlation between genetic variability and Atlantic Forest phytophysionomies. In the absense of a genetic structure, slightly negative variation can be obtained only by chance, because the true value estimated is zero ( Huang et al. 2021).