2.1 Study area and sample collection
Our fieldwork covered 14 historical distribution sites of C. yucatanicus on the northern coast of the Yucatan Peninsula, Mexico (Figure 1). In each site, we conducted intensive surveys along transects combining with stops that lasted six minutes each 100m, with the help of a playback; for this, we downloaded xeno-canto vocalizations of C. yucatanicus (https://xeno-canto.org/), which were recorded between 2010 and 2016, including recordings of couples and solitary individuals. We played the vocalizations for one minute with a one-minute break, repeating this process three times. We sampled during three days per site, from 07:00-11:00 and from 16:00-19:00 hours, from March 2015 to February 2016. We captured 186 individuals using 12 meters long mist nets, but only 130 were considered in this study which correspond to a mean of 9 individuals per site.
We captured individuals marked with a unique combination of colored Darvick leg bands. Then, we extracted blood samples (2-3μl) from the brachial vein [39]. We ensured that individuals did not present bleeding when they were released. We collected blood in tubes with K3EDTA at 15%, and samples were stored at -20 ° C until processing in the laboratory. Access to protected areas and collections was made with the SGPA / DGVS / 06821/14 and SGPA / DGVS / 007765/15 permits.
2.2 Laboratory process
We extracted DNA using a modification of the cell lysis method and phenol-chloroform isoamyl alcohol [40]. Successful DNA extractions were detected by electrophoresis in 1.5% agarose gels with SYBR Gold Staining, at a constant voltage of 100 V for 30 min. Samples were genotyped using seven microsatellites described by Barr et al. [41] for C. brunneicapillus and standardized for this study for C. yucatanicus (Table 1). For amplification, we used a mixture containing 1 µL of DNA, 3 µL of master mix Taq DNA Polymerase (InvitrogenTM), 2.7 µL of ultrapure water and 0.3 µL of primer, for a final volume of 6 µL. We preheated PCRs reactions at 94 °C for 3 min and then performed 39 cycles with the following steps: denaturation at 94 °C for 1 min, alignment at a specific temperature for each primer (Table 1) for 1 min and extension to 72 °C for 1 min. We maintained extension at 72 °C for 10 min and it was allowed cooling up 10 °C. We visualized PCR products by electrophoresis in 2% agarose gels with SYBR Gold Staining, at a constant voltage of 100 V for 30 min. Sizing was made by capillary electrophoresis in an Applied Biosystems automatic sequencer, using LIZ-600 as the internal size standard. We analyzed electropherograms using Peak Scanner 1.0 (Applied Biosystems).
2.3 Genetic diversity analysis
We quantified genetic diversity through allele richness (Na), Shannon diversity index (I) and expected (He) heterozygosity using Genalex [42]. We conducted an analysis to determine if populations have experimented a recent bottleneck in Bottleneck program [43] using Wilcoxon test. To do this, we evaluated
Table 1. Microsatellite loci information used to analyze the genetic diversity of Campylorhynchus yucatanicus in the north tip of the Yucatan Peninsula, Mexico. Total number of alleles (AT) and alignment temperature (TA). 1 Barr et al. 2015
ID
|
Locus1
|
Repetition units
|
Primer sequence (5´-3´)
|
AT
|
Length
(pb)
|
TA (°C)
|
Locus1
|
CACW3-01
|
(ATT)5G(TTA)4(TTG)6TTATTG(TTGTTA)3(TCA)9
|
F: ACTGTTCACCCTTGGACCTG
R: TGTCTGGAAACCACTGAAGAAC
|
6
|
168-188
|
|
Locus2
|
CACW3-03
|
(CTA)5CTG(CTA)8(ATA)10
|
F: TCCTGAAATGTAATTCAGACACC
|
5
|
259-279
|
57.6
|
|
|
|
R: CAGAGTGCTACTTAAATTGATTCTTTC
|
|
|
|
Locus3
|
CACW3-05
|
(TGT)5
|
F: GATGCATATTGTCAGAGTTCCAC
|
5
|
131-149
|
57.6
|
|
|
|
R: CTGGACTGAGCTAACAAATGATG
|
|
|
|
Locus4
|
CACW3-11
|
(ATA)5(AAC)6AAT(AAC)4(AAT)3AG(TAA)4
|
F: TTCTCCTCCCTCTACCTCCTTT
|
8
|
180-204
|
54
|
|
|
|
R: GTGACAACAGAAAATTCCCTTTA
|
|
|
|
Locus5
|
CACW4-01
|
(GTAT)6GAATCTG(TCTA)11
|
F: TTTTGCCTAATAAACTGGCTGAC
|
3
|
122-133
|
54
|
|
|
|
R: CACAGAACCACAACCTACATGG
|
|
|
|
Locus7
|
CACW4-04
|
(TCTA)14
|
F: TCTCACGTCTTACCATCCTGTG
|
5
|
241-257
|
57.6
|
|
|
|
R: TTGATACTTGAAACTCTCCTTCTGTC
|
|
|
|
Locus9
|
CACW4-09
|
(GATG)22
|
F: GCTAACTGAAAGGGATTGTTGG
|
5
|
92-116
|
59
|
|
|
|
R: TTTCTGGCATGTTTCCTGTC
|
|
|
|
our data with 70% ratio explained by a Step Mutation Model (SMM) and 30% explained by the Infinite Alleles Model (AMI) in a two-phase model (TPM) as recommended for microsatellites analysis [43].
2.4 Genetic structure analysis
We determined whether genotypic frequencies at each locus were under Hardy-Weinberg equilibrium (HWE) and evaluated linkage disequilibrium (LD) among pairs of loci in Genepop 4.6 [44]. To characterize the genetic structure of the species, we used several methods. Bayesian clustering analysis was applied to identify genetic groups, without prior assignment of individuals to a given population, in Structure 2.3 [45]. To determine the most probable value of K, which could be interpreted as the optimal number of genetic groups or true clusters, we ran K from 1 to 14 with 10 simulations for each K using 10,000 iterations before beginning analysis and 50,000 iterations in the Markov Chain Monte Carlo (MCMC). We used the method proposed by Evanno et al. [46] to define the value of K considering the distribution of ΔK in Structure Harvester [47].
Bayesian clustering analysis were also performed in the Geneland 4.0 library [48, 49, 50, 51] for R 3.4.1 [52], in which we assumed a spatial model with alleles not correlated. For this analysis, we performed 1x106 MCMC iterations. We considered several populations, or groups (K), from 1 to 14, and 1 iteration was saved every 100. Uncertainty of 100 m for the geographic coordinates was assumed, and we evaluated the MCMC convergence using 10 repetitions in each analysis.
Analysis of molecular variance (AMOVA) allowed evaluation of the percentage of variation between genetic groups identified by Structure and among sites in the same group and among individuals. This analysis was carried out in Genalex [53]
2.5 Landscape composition and genetic diversity: node level
We defined 14 plots that resulted from a buffer area of 2 km around capture points at each site. To characterize the landscape's composition and structure, we used three images of the Sentinel 2 satellite (20 m spatial resolution, April 2016). We manually digitized patches of know vegetation previously ground truth during bird mist netting, which were classified into four classes with different types of vegetation and land use: (1) adequate habitat preserved, (2) disturbed habitat, (3) unsuitable habitat, and (4) secondary vegetation habitat with some human intervention. The first category consisted of conserved fragments with coastal thorn scrub forest and dune vegetation complex [32]. Disturbed habitats included human settlements, roads, and areas with bare soil because of deforestation. The third category included habitats where C. yucatanicus has not been registered according to the literature [17, 26, 30, 31] and our field observations. Finally, in the fourth category, we considered types of secondary vegetation that maintain elements of the original vegetation of coastal thorn scrub forest such as agaves and cacti [32] and that are subjected to different human activities uses such as livestock.
Structure and composition of the landscape, as well as its degree of fragmentation, were recorded for each plot [7] in Patch Analyst [53]. We selected the following variables: Shannon’s patch equitability index (SEI), proportion of suitable habitat (CA1), patch diversity index (SDI), proportion of disturbed habitat (CA2), risk index for proximity to human settlements (PA), edge density of the appropriate habitat patches (ED1), average form index of suitable habitat patches (MSI14), distance to human settlements (SA), number of patches of suitable habitat (NumP14), and average size of patches of suitable habitat (MedPS14). We calculated the risk index by proximity of human settlements (PA) with the following formula PA=PobTotal/S*100, where PobTotal is total population of nearest human settlement and S is the distance to the settlement.
We proposed a priori 12 models or hypotheses to describe the relationship between the genetic diversity of C. yucatanicus populations and the configuration of the landscape. These models were evaluated through Akaike Information Criterion (AIC) [54]. The hypotheses proposed were based on the previous knowledge obtained in the literature and in our field observations. Models were constructed using linear regressions in R [52]. Expected heterozygosity (He) of C. yucatanicus in the sampling sites constituted our response variable, while the predictive or explanatory variables were those obtained in Patch Analyst.
We selected the model with lowest AIC [54] because it corresponds to greater support of data. The difference between the AIC value of the best model and the AIC of the remaining models (Δ AIC) made it possible to evaluate their relative hierarchical organization. Those models with ΔAIC < 2 were considered as better and equally competitive. Models with 2 < ΔAIC < 4 were considered as partially informative. We calculated Akaike's weight (ωi) [54, 55] to evaluate the relative likelihood of models to be plausible. Regression coefficients (R2) were calculated to verify the fit of data and for models with more than one parameter and adjusted R2 was taken into account. We calculated F statistic and P values to determine the statistical significance of regressions.
2.6 Genetic distances and landscape resistance
We analyzed the relationship between genetic distance and landscape resistance between sites through linear regression. We also explored the relationship between genetic distance and the spatial Euclidean distance between sites. Calculation of the FST statistic was performed in Genalex 6.5 [42] and modified according to the formula FST mod=FST/(1-FST) formula [56]. We selected FST mod for the following connectivity analyses because the other calculated variables that describe the genetic distance between sites, such as the coefficient of genetic differentiation between populations (GST), showed significant correlations.
The resistance offered by landscape to the movement of individuals was evaluated using circuit theory [57] in Circuitscape [58]. The term resistance was used as an antonym of landscape connectivity, which is defined as the degree to which landscape facilitates the movement of individuals [59] that can move randomly between two sites or nodes. In this context, nodes can be habitat fragments, populations, or points in the landscape, among whom connectivity could be evaluated [58]. We used the centroid of capture points at each collection site to establish focal nodes in this study. The map of the Yucatan Peninsula that contained the C. yucatanicus potential distribution allowed acquiring more accurate resistance models [60]. The resistance layer used to estimate cost or resistance between nodes was used in previous studies for the same species [61].