Contemporary climate influence on variability patterns of Anadenanthera colubrina var.cebil, a key species in seasonally dry tropical forests

María Victoria García · María Eugenia Barrandeguy · Kathleen Prinz

Abstract The distribution of many plant species has been shaped by climate changes, and their current phenotypic and genetic variability reflect microclimatically suitable habitats.This study relates contemporary climate to variability patterns of phenotypic traits and molecular markers in the Argentinean distribution of Anadenanthera colubrina var.cebil, as well as to identify the most relevant phenotypic trait or molecular marker associated with those patterns.Individuals from four populations in both biogeographic provinces, Paranaense and Yungas, were investigated.Multivariate analyses and multiple linear regressions were carried out to determine relationships among phenotypic traits and nuclear microsatellites, respectively, to climatic variables, and to identify the phenotypic traits as well as nuclear microsatellite loci most sensitive to climate.Two and three clusters of individuals were detected based on genetic and phenotypic data, respectively.Only clusters based on genetic data reflected the biogeographic origin of individuals.Reproductive traits were the most relevant indicators of climatic effects.One microsatellite locus Ac41.1 appeared to be non-neutral presenting a strong correlation with climate variable temperature seasonality.Our findings show complex patterns of genetic and phenotypic variability in the Argentinean distribution of A.colubrina var.cebil related to the present or contemporary climate, and provides an example for an integrative approach to better understand climate impact on contemporary genetic and phenotypic variability in light of global climate change.

Keywords Contemporary climate · Curupay · Genetic variability · Phenotypic variability · Seasonally dry tropical forests

Introduction

Biogeographical distribution patterns of plant species are primary limited by climate.These patterns are also shaped by combinations of isolation by distance, genetic drift, selection and environmental conditions, along with other factors.Information from genetic, demographic and ecological approaches must therefore be integrated in order to understand the driving forces of contemporary distribution patterns of plant species so that sustainable population strategies can be developed (Lefèvre et al.2014).

Populations of a species are affected differently over its range by spatio-temporal climate changes so that populations vary in their level of adaption (Davis et al.2005).The response by plant populations is affected by the magnitude, rate and duration of climate changes.Such response also depends on the phenotypic and ecological plasticity of individual genotypes, the distribution and nature of genetic variation for relevant traits, the extent of gene flow among populations through dispersal of both pollen and seeds, and the demographic processes of populations (Davis et al.2005).Current relationships between phenotypes also reflect either historical habitat tracking or ongoing adaptation in local habitats.These processes operate at different spatial and temporal scales on different levels of biological organization.On the one hand, habitat-tracking relationships between phenotypic traits and environments result from regional processes such as dispersal and ecological sorting of species; on the other, adaptation relationships reflect natural selection among alternative alleles at short distances (Oberle and Schaal 2011).Both perspectives, i.e., habitat tracking and adaptation, highlight how alternative responses to climate change may generate different relationships between genetic variation, trait variation, and geographic distribution of genetic variability of populations (Oberle and Schaal 2011).

Periods of variable climate, including glacial-interglacial cycles with strong changes in temperature, precipitation and CO2concentration, in the past ca.2.5 × 106years have operated on modern plant taxa (Davis and Shaw 2001).Hence, the current distribution of genetic variations may reflect responses to historic climate changes (range), and current microclimate heterogeneity (landscape distribution pattern) (e.g., Oberle and Schaal 2011).

Pleistocene glacial and interglacial periods of time have repeatedly and significantly influenced the geographical distribution of plants and animals in the temperate latitudes.Changes in effective precipitation have induced forests or woodlands to become savanna and vice versa in South America (van der Hammen 1974).The Pleistocene Arc Theory hypothesized that seasonally dry tropical forests (SDTFs) had expanded and merged during the drier glacial periods of the Quaternary, and contracted and fragmented during the moister interglacial periods (Pennington et al.2000).Thus, their forest species have experienced repetitive cycles of fragmentation and expansion during these climatic changes (Mogni et al.2015).

Currently, the SDTFs are a biome with a wide and fragmented distribution, found from Mexico to Argentina and throughout the Caribbean.They occur on fertile soils where rainfall is less than ～ 1800 mm per year, with a period of 3 to 6 months receiving less than 100 mm per month, during which the vegetation is mostly deciduous (DRYFLOR 2016).Forests exhibit highly fragmentary distribution and form a ‘dry diagonal’ of woody vegetation between the Caatinga in the northeast and the Andean piedmont dry forests in the southwest of South America (Prado and Gibbs 1993).These forests occur as large, well-defined nuclei (e.g., Caatinga in the northeast) and as smaller enclaves within other vegetation types (e.g., Cerrado and Chaco) (Caetano et al.2008).The transition between savannas and forests in the Cerrado is a consequence of nitrogen deposition (Bueno et al.2018), while floristic compositions across SDTFs appear to be related to temperature regime (Neves et al.2015).Prado and Gibbs (1993) and Pennington et al.(2000) compared the current distribution of dry forest species across the South American tropics and showed that over 100 phylogenetically unrelated species have similar geographic patterns, forming four disjunct dry forest nuclei: Caatinga, Misiones, Chiquitano and Piedmont (Fig.1).The SDTFs have been understood to be a metacommunity (biome) for woody plant clades (Pennington et al.2009), while Prado and Gibbs (1993) established that Fabaceae and Bignoniaceae are the most dominant families in these forests.

The southernmost distribution of SDTFs is located in Argentina where these forests are distributed in the Paranaense and Yungas biogeographic provinces (Cabrera 1994) in the north-east and north-west of the country, respectively (Fig.1).These biogeographic provinces are located in the Misiones and Piedmont SDTFs’s nuclei, respectively, and contain the highest biodiversity in the country (Brown et al.2001; Di Bittetti et al.2003).Misiones and Piedmont nuclei occur as enclaves within the domain of Chaco, areas of woodlands and xeromorphic forests that occur on the less well-drained soils of Paraguay, Argentina and Bolivia (Spichiger et al.1995), with the Argentinean SDTFs nuclei mostly deciduous and different from the surrounding xeromorphic forests in the Chaco domain.

Fig.1 Distribution of Seasonally Dry Tropical Forests (SDTFs) in South America and Argentinean sampled populations

The most typical species in the SDTFs isAnadenanthera colubrina(Vell.) Brenan (Leguminosae, Caesalpinioideae), and owing to its distribution, is considered as a key species of SDTFs as it is found to be either dominant or frequent in all dry forest nuclei of South America (Prado 2000).Presently, is a dominant species while has been involved in the cyclic expansion-retreat migrations of biomes during the climate changes of the Pleistocene (Prado and Gibbs 1993).

Anadenanthera colubrinavar.cebil(LPWG 2017), locally known as curupay, is a long-lived, semi-deciduous canopy species that can attain heights of 35 m.It has compound bipinnate leaves with specialized ant glands (extrafloral nectaries, specialized nectar-producing glands).It also has hermaphroditic flowers with male and female parts in inflorescences and long legume fruit (von Altschul 1964; Justiniano and Fredericksen 1998; Cialdella 2000; Klitgård and Lewis 2010).It has been suggested that the mating system ofA.colubrinavar.cebilis predominantly outcrossing (Cialdella 2000).Bees are the main pollinators, and seeds are dispersed by self-dispersal or wind dispersed following pod dehiscence (Justiniano and Fredericksen 1998; De Noir et al.2002).

In a previous study, we carried out research confined to the southernmost distribution and included four populations located on the sites where itis well-represented in Argentina.The presence of homogeneous clusters for both phenotypic and molecular genetic variability were identified and where individuals were assigned to their biogeographic provinces of origin (García et al.2014).In Brazil, the distribution of the Caatinga nuclei occur as enclaves within the Caatinga domain and has shown both spatial structural and significant relationships with environmental variables, i.e., geo-climatic variables that determine the availability of ground water over time (Santos et al.2012).This information is not presently available for the rest of SDTFs distribution in South America.From our previous results and in recognition of large microclimatic diversity related to topographic diversity (Ligier et al.1985), along Paranaense and Yungas biogeographic provinces, a new analysis was carried out from the dataset analyzed in García et al.(2014) under the hypothesis that detected clusters could be the result of spatial climate change across the southernmost range of the species.

To highlight to the accepted climatic variables underlying changeable patterns inA.colubrinavar.cebilbetween the Argentinean SDTFs nuclei, the aims of this study are: (1) to relate contemporary climate (temperature and precipitation) to phenotypic traits and molecular markers so as to understand the variability patterns; and, (2) to identify the most relevant phenotypic trait or microsatellite loci associated with the distribution of environmental variability.These aims intend to answer the following questions: Are variability patterns inA.colubrinavar.cebilin the Argentinean SDTFs consequences of contemporary climate? If it is true, what are the climatic variables operating on the variability patterns? Also, which do phenotypic trait or microsatellite loci show climatic influence and define the detected variability patterns ofA.colubrinavar.cebilin the Argentinean SDTFs?

Materials and methods

System of study

Four populations ofA.colubrinavar.cebilwere analyzed in two different biogeographic provinces, Paranaense and Yungas.Soils in the Paranaense are well-drained, extremely acid, with low nutrient availability.The climate is subtropical with mild winters and warm summers with frequent rain (Ligier et al.1985).In the Yungas, the climate is also subtropical with a dry season characterized by mild, rainy winters and warm summers.There is large microclimatic diversity related to topography (Ligier et al.1985).

The populations studied were: Candelaria (Cand) (27°26′58.2″S-55°44′20.22″W 104 m, a.s.l.), and Santa Ana (SA) (27°25′55.92″S-55°34′16.68″W 153 m, a.s.l.) in the Parananense Province, and Tucumán (T) (26°47′26.10″S-65°18′58.14″W 610 m, a.s.l.) and Jujuy (J) (23°45′15.012″S-64°51′12.996″W 800 m, a.s.l.) in the Yungas Province (Fig.1).Twenty individuals were sampled in Candelaria, 16 in Santa Ana, 14 in Tucumán and 19 in Jujuy.The sampling methodology is described in García et al.(2014).

Phenotypic traits

Variability of phenotypic traits, i.e., traits whose value is defined by both genotype and environment, was evaluated by means of eight vegetative and five reproductive traits.The vegetative traits were: number of pairs of leaflets (NPL), distance mean between leaflets (DMBL), length of medium leaflet (LML), width of medium leaflet (WML), length/width medium leaflet Ratio (L/WML), length of leaf (LL), width of leaf (WL) and relation of leaf length/width (L/WL); reproductive traits were: length of fruit (LF), width of fruit (WF), number of seeds per fruit (NSF), length of seed (LS) and width of seed (WS).The length and width of seed were measured in all seeds of fruit analysed.According to availability, traits were measured in three to five leaves and five to eight fruits per individual.Leaves were dried between paper towels kept in herbarium condition while fruits and seeds were dried at room temperature and kept in paper bags.Leaves and fruit traits were measured with a ruler while seeds were measured with a digital caliber.

Molecular markers and genotyping

Eight nuclear microsatellite loci developed forA.colubrinavar.cebil(Barrandeguy et al.2012) were analyzed:Ac34.3,Ac48.1,Ac11.2,Ac28.3,Ac157.1,Ac41.1,Ac172.1, andAc162.1.The genotyping methodology is described in Barrandeguy et al.(2014).

Data analyses

A new analysis of the dataset in García et al.(2014) was performed in order to address the research questions of the current study.Data analysis included methodologies for illustrating the differentiation between biogeographic provinces based on phenotypic and molecular data.

The number of alleles per locus was registered by counting.The evaluation of linkage disequilibrium and the estimation of the presence of null alleles and genotyping errors have been described in García et al.(2014).

To illustrate the differentiation between biogeographic provinces, unrooted distance trees for phenotypic and molecular data were constructed using the software Darwin 5.0.84 (Perrier and Jacquemoud-Collet 2006).Matrices of pairwise phenotypic and genetic distances were computed using the Euclidian distance index.Unrooted distance trees were constructed for phenotypic traits and molecular markers using the weighted neighbor-joining method proposed by Saitou and Nei (1987), and the unweighted neighbor-joining method, respectively, proposed by Gascuel (1997).The robustness of each node was assessed by bootstrapping for all phenotypic traits, loci and alleles with 1000 replications.

A principal component analysis (PCA) was performed to demonstrate the variance of phenotypic traits along a set of principal axes in the A-space.This analysis is based on the distance matrix and was performed on standardized and centered population mean values using the Multivariate Statistical Package (MVSP) (Kovach 1995).Relative significances of characters were analyzed for the first three components.The most relevant locus associated with the distribution of molecular variability was observed using a principal coordinate analysis (PCoA) based on Nei’s genetic distances (Nei 1978), calculated with GenAlEx 6.4 (Peakall and Smouse 2006).Relative significances of loci were also analyzed for the first three coordinates.

Statistical analyses

A specific statistical approach for analysing possible relationships between phenotypic and molecular data were performed, looking for an explicative contemporary climate variable responsible for the variability patterns of natural Argentinean populations ofA.colubrinavar.cebil.

Means, median values and coefficients of variation were calculated for the quantitative traits under study at the biogeographic provinces level.Statistical differences for all quantitative traits were determined by means of a paired-ttest (Steel and Torrie 1980) using GraphPad (http://www.graph pad.com/ quick calcs/ ttest1.cfm).

Multiple linear regressions (MLR) can be regaded as an extension of straight-line regression analysis (which involves only one independent variable), to the situation in which more than one independent variable must be considered (Kleinbaum et al.2008).In this way, it allows an analysis of relationships between multiple predictor and single predicted variables, and has become one of the key components in the molecular ecologist’s analytical toolbox (Frasier 2016).These analyses were performed to examine relationships among all phenotypic traits and 19 bioclimatic variables derived from the WorldClim database (Hijmans et al.2005) using the Rcmdr (Fox 2005, 2007).DIVA GIS (Hijmans et al.2001) was used to extract each variable’s value according to each individual’s geographic position.

A basic assumption in multiple linear regression modelling is the independence of explanatory variables (Kleinbaum et al.2008), i.e., there is no linear relationship among the explanatory variables.A case of the explanatory variables being highly correlated is referred to as multicolinearity.In this way, twelve out of nineteen variables were excluded from this analysis.Therefore, seven temperature-related variables were included in the maximum model for MLR.All variables are detailed in Table 1.We specified the model with the lowest Akaike information criterion (AIC) score following a bidirectional elimination strategy for selection of the variables, i.e., a combination between forward selection model and backward elimination model (Kleinbaum et al.2008).

Canonical correspondence analyses (CCA) were performed for each locus in order to understand relationships among its alleles and climatic variables using the vegan package (Oksanen et al.2011).Ordination axes represent linear combinations of climatic variables (TerBraak 1986).Six climatic variables for both temperature (x1-x6) and precipitation (x8-x13) were included (Table 1).Alleles were converted into a single variable based on presence or absence using the method of Smouse and Williams (1982).For codominant markers in a diploid genome, the score of a single allelic variable is ‘1’ for homozygous presence, ‘0.5’ for heterozygous presence, and ‘0’ for homozygous absence (Westfall and Conkle 1992).The number of allelic variables for each locus represents the number of alleles minus one (Smouse and Williams 1982).

Results

Distribution of phenotypic and genetic variability across biogeographic regions

Patterns of phenotypic variation were different in each province (Fig.2).Five out of eight vegetative traits showed higher averages in Paranaense than in the Yungas, whereas the means of reproductive traits were higher values in the latter, except for WS (seed width) (Fig.2).Coefficients of variation had moderate values from 7 to 35% in the Paranaense and 9.1 to 23.5% in the Yungas, with the highest in NSF (number of seeds/fruit).

Fig.2 Variation of quantitative traits a Means and CV (coefficient of variation) by biogeographic provinces; b Box and whisker plots for each vegetative trait by biogeographic provinces; c Box and whisker plots for each reproductive trait by biogeographic provinces (NPL number of pairs of leaflets, MDBL mean distance between leaflets, LML length of medium leaflet, WML width of medium leaflet, L/WML length/width of medium leaflet, LL length of leaf, WL width of leaf, L/WL length/width of leaf, LF length of fruit, WF width of fruit, NSF number of seeds per fruit, LS length of seed, WS width of seed; measurements in cm, ns, not significant; *P ＜ 0.05)

The weighted neighbor-joining tree for phenotypic traits defined three main groups: individuals from Yungas clustered in separate groups, with individuals from Paranaense mostly clustered together in one group, although a high proportion were grouped with individuals from Tucumán (Yungas).Several branches presented bootstrap values higher than 60 (Fig.3a).

Individuals from different biogeographic provinces were separated along the first main component axis in the ordination analysis for phenotypic traits (Fig.4a).The percentage of variation summarized by the first two component axes explained 66.3%, with fruit length, fruit width, number of seeds per fruit, seed length and width of seed as the most explanatory traits.The reproductive traits correlated to the first PCA axis explain 36% of the total variation, and these traits were most relevant in the discrimination of the biogeographic provinces (Table 2).

Table 1 Bioclimatic variables from the WorldClim database for Paranaense and Yungas biogeographic province (http://www.diva-gis.org/ Data)

Microsatellite markers (simple sequence repeats markers, SSR) are used extensively for obtaining information about population’s differentiation and their structure.In fact, they are useful in population and landscape genetics.SSRs reveal high polymorphism because they are multi allelic andco-dominant markers (Pournosrat et al.2018).In our study, genotypic frequencies did not show evidence of LD or null alleles.Analysis of genetic variability was reported in García et al.(2014).

Genetic relationships between biogeographic provinces were illustrated by the unweighted neighbor-joining tree for molecular markers (Fig.3b).Individuals from different populations were grouped according to their biogeographic province of origin.Several branches also presented bootstrap values higher than 50.

Fig.3 a Phenotypic relationships among individuals of Anadenanthera colubrina var.cebil from Yungas and Paranaense biogeographic provinces illustrated by a weighted neighbor-joining tree; b genetic relationships among individuals of A.colubrina var.cebil from Yungas and Paranaense biogeographic provinces illustrated by a weighted neighbor-joining tree; numbers indicate bootstrap values after 1000 replications (only bootstraps higher than 50% are shown); (J) Jujuy, (T) Tucumán), (Cand) Candelaria and (SA) Santa Ana)

The ordination analysis of molecular markers showed two main groups according to origin (Fig.4b).The percentage of variation was mostly represented by the first three coordinates which explained 63%.The highest correlation to the first coordinate was found for locusAc41.1, accounting for 48% of the variation (Table 3).Similar groupings of individuals in the bi-dimensional plot was shown in the ordination analysis performed without locusAc41.1 (Fig.S1).This locus was excluded from the present analysis due to its highFSTvalue and presumed non neutral nature (García et al.2014).

Fig.4 a Principal component analysis (PCA) of quantitative traits in individuals of Anadenanthera colubrina var.cebil from Yungas and Paranaense biogeographic provinces; b principal coordinate analysis (PCoA) of genetic data resulted from SSR analysis of individuals of A.colubrina var.cebil fromYungas and Paranaense biogeographic provinces

How climatic variability shapes traits and genetic variability

Based on the general linear model that includes a combination of seven climatic variables (x1-x7) to explain phenotypic variability, Akaike information criterion (AIC) scores suggested different models for different quantitative traits.Five out of eight vegetative and all reproductive traits showed statistical significance for the multiple regression models (Table 4).The general model explains phenotypic variability of number of leaf pairs, while a model defined by a combination of temperature variablesx2,x3,x4,x5, andx6explains phenotypic variability of length of medium leaflet, length/width of medium leaflet and width of leaf.The vegetative trait leaf length/width is explained by a model defined by combination of temperature variablesx1,x2,x3andx6.The general model also explains phenotypic variability of seed traits (LS and WS), while a model defined by the combination of temperature variablesx2,x5andx6explains phenotypic variability of width of fruit.A model defined solely by the temperature variablex6explains the phenotypic variability of number of seeds per fruit.

The climatic variable minimum temperature of coldest month (MTCM-x6) was involved in all models used to explain phenotypic variability considering both vegetative and reproductive traits.

Locus-specific CCAs showed that approximately 80% of allelic patterns were explained by temperature variables, as indicated by the first two axes for three out of six loci.For locusAc11.2, all temperature variables explained 28% of the total variation in the data, while the first two canonical axes contained 76% of the variation, and as a result, 21% of the total variation was represented in the graph (Fig.5a).For locusAc41.1, the six temperature variables explained 28% of the total variation in the data, while the first two canonical axes contained 80% of the variation, and as a result, 22% of the total variation was represented in the graphic (Fig.5b).Finally, for locusAc162.1, the six temperature variables explained 26% of the total variation in the data while the first two canonical axes contained 85% of the variation, and as a result 22% of the total variation was represented in the graphic (Fig.5c).Although temperature CCA analysis for these loci showed similar percentages of total variation, distribution of individuals in the bi-dimensional plots corresponds to their biogeographic province of origin only for locusAc41.1.In locusAc41.1, the variablesx4,x5andx6are strongly related to the first canonical axis indicated by the length of arrows.The variablesx5andx6showed an increased gradient to individuals representing the Paranaense biogeographic province while variablex4showed an increased gradient to individuals originated from the Yungas biogeographic province (Fig.5b).

Locus-specific CCAs showed that about 79% of allelic patterns were explained by precipitation variables as indicated by the first two axes for three out of six loci.For locusAc11.2, all precipitation variables explained 29% of the total variation in the data while the first two canonical axes contained 72% of the variation, and thus 21% of the total variation was represented in the graphic (Fig.5d).For locusAc41.1, the six precipitation variables explained 28% of the total variation while the first two canonical axes contained 81% of the variation, and 22% of the total variation was represented in the graphic (Fig.5e).Finally, for locusAc162.1, the six precipitation variables explained 26% of the total variation, while the first two canonical axes contained 86% of the variation, and as a result 22% of the total variation was represented in the graphic (Fig.5f).In a similar way for temperature variables analysis, the distribution of individuals in the bi-dimensional plots corresponds to their biogeographic province of origin only for locusAc41.1, although precipitation CCA analysis for these loci showed similar percentages of total variation.In locusAc41.1, all variables are related to the first canonical axis indicated by the length of arrows.The variablesx8,x9, x11,x12andx13showed an increased gradient to individuals representing the Paranaense, while variablex10showed an increased gradient to individuals originated in the Yungas (Fig.5e).

Fig.5 Canonical correspondence analysis for temperature and precipitation variables

Climatic CCA for locusAc41.1 showed an increased gradient to individuals from Yungas for variablesx4andx11, indicating temperature and precipitation seasonality, respectively.

Discussion

In the current study, the influence of the present climate on genetic and phenotypic variability in Argentinean populations ofA.colubrinavar.cebilwas analyzed.

Distribution of phenotypic and genetic variability across biogeographic regions

Reproductive traits showed the highest discriminatory power and were differently affected in the two biogeographic provinces (i.e., the average of NSF, number of seeds/fruit) (Fig.2a, c), possibly indicating a different adaptive response among individuals in these provinces.These traits were also the most relevant in the discrimination of biogeographic provinces of origin (Fig.4a; Table 2).The variation of this reproductive trait was higher in the Paranaense (35%) than in the Yungas (23.5%).Higher regeneration rates and a higher adaptive potential could therefore be expected in disturbed areas of the Paranaense.Since peripheral (younger)populations show reduced adaptive potential as a result of their isolation from other populations (Volis et al.2014), the results also indicate that populations from the Yungas may be younger than populations from the Paranaense, in a similar way that peripheral populations which have expanded their geographic ranges from core (older) populations.High differentiation between biogeographic regions may be explained by different climatic conditions, whereas on small scales, environmental variation may allow for local adaptation, and thus differentiation of populations (Joshi et al.2001).Differences at small scales could be a consequence of phenotypic plasticity or adaptive responses to selective pressures.An increased understanding of the roles of plasticity requires a ‘whole organism’ approach, taking into consideration that organisms are integrated complex phenotypes (Forsman 2015).However, certain phenotypic traits could be considered as adaptive traits, and hence, their patterns of variation may reflect selective pressure.Reproductive traits are relevant indicators of the capacity of individuals and populations to persist from generation to generation.Also, these phenotypic traits are rather stable, and thus are often included in taxonomic analyses.

Table 2 Relative significance of quantitative traits to the three first principal components

Table 3 Relative significance of molecular markers to the three first principal coordinates

Table 4 Multiple linear regressions between climatic variables and phenotypic traits

No obvious geographical distribution of phenotypic variability was observed by the weighted neighbour-joining tree despite a robust grouping in three well-defined groups (Fig.3a).In contrast, the distribution of genetic variability by geographical origin can be considered a consequence of genetic drift resulting from patch distribution of analyzed populations (Fig.3b).In the same sense, previous studies have detected high genetic differentiation in nuclear and chloroplast genomes between Argentinean populations ofA.colubrinavar.cebilfrom both biogeographic provinces, emphasizing the role of ancient fragmentation of natural populations in the north (Barrandeguy et al.2014).The study of historical development of SDTFs by demographic analysis of cpSSRs data inA.colubrinavar.cebilalso identified footprints of Post-Last Glacial Maximum expansion events indicated by the presence of rare haplotypes in Piedmont nuclei (Barrandeguy et al.2016).In addition, patch distribution of the populations in the Paranaense affected the genetic structure of populations, whereas gene flow by pollen maintains high genetic variation and preserves the effects of genetic drift in populations of the Yungas biogeographic province (Barrandeguy et al.2014).

How climatic variability shapes traits and genetic variability

In our study, multiple linear regressions were used to understand climate effects on morphological and genetic variation patterns.Hence, phenotypic differences between both biogeographic provinces may be explained by local adaptation to minimum temperature of coldest month (MTCM-x6).Despite the fact that both biogeographic provinces are characterized by mild winters, differences in the MTCM are visible.The mean temperature of the coldest month reaches about 10 °C in the Paranaense and 6 °C in the Yungas (Table 1).Lower temperatures in the Yungas may therefore lead to an increasing NSF (number of seeds/fruit), whereas higher temperatures in the coldest months have a positive effect on the vegetative traits in Paranaense.NSF is a functional trait relative to a variety of attributes involved in individual fitness (w), being a functional morpho-physiophenological trait which impacts fitness indirectly via its effects in performance traits, and on that, the environment plays a role of a sieve, defining which plants are able to be persist in the community (Violle et al.2007).A 2 °C rise in temperature has already been shown to impact seed responses, including seed production, seed mass, seedling emergence and establishment, and soil seed bank dynamics (Cochrane et al.2011).Soil seed banks are one of the main contributors to species persistence and coexistence in variable habitats, allowing populations to re-establish, e.g., after extended drought (Ooi et al.2009).The direct effect of temperature on seeds and seed bank dynamics therefore suggests that any temperature increase related to climate change will be a critical driver determining species persistence and coexistence in such habitats (Ooi et al.2009).

Multivariate analysis, i.e., a sensitive analysis to detect low genetic differences across loci (Grivet et al.2008), revealed thatAc41.1 was the most explanatory locus of the total genetic variation, and its allele frequencies reflected a biogeographical pattern in both CCA analyses.Allele patterns of locusAc41.1 indicate an environmental dependence upon temperature and precipitation seasonality, (x4andx11, respectively).Similarly, distribution patterns of the Caatinga nuclei tree flora is primarily related with the availability of ground water over time (Santos et al.2012).The highest value ofFST(0.215) was also found for this locus.Variation inFSTamong loci may indicate the effects of selection if applied over a wide geographic range of individuals (Beaumont and Nichols 1996).In fact, this supports the hypothesized non-neutral nature of locusAc41.1 (García et al.2014).Several studies have increased the amount of evidence about the non-neutrality for some SSRs because they could be linked to genomic regions under selection (Nielsen et al.2006; Lazrek et al.2009; Shi et al.2011).Loci of ecological relevance may be considered as either potentially adaptive or as linked to the genes or genomic regions under selection (Manel et al.2012).Thus,Ac41.1may be regarded as a signature of selective sweep, suggesting a probable linkage with advantageous alleles for adaptive traits (Hamilton 2009).

Our study provides an example for an integrative approach to better understand climate impact on contemporary genetic and phenotypic variations.The genetic locusAc41.1 and the phenotypic trait NSF (number of seeds/fruit) indicate climatic effects, and can thus be considered practical tools for the development of management and conservation strategies.In view of rapid climate changes, an assessment of their impact on multiple levels of plant biological processes is urgently needed.The information is not only relevant for single species, but also for species interaction, soil biochemistry and numerous other environmental factors that are affected by climate (Riordan et al.2016).

Conclusions

In this study, we show that differences in contemporary genetic and phenotypic variability in Argentinean populations ofA.colubrinavar.cebilare linked to climate effects.Phenotypic variability, considering both vegetative and reproductive traits, was explained by minimum temperature of the coldest month, while the distribution of alleles of locusAc41.1 showed a geographical pattern related to temperature and precipitation seasonality.It may be possibly linked to a gene region under selection.Relationships to climatic variables may, therefore, indicate adaptation of individuals to particular conditions in their local populations.Large genetic resources are essential to ensure adaptation to changing environmental conditions (Lefèvre et al.2014).Conservation and sustainable management of tree species requires detailed understanding of their genetic diversity and distribution of genetic resources (Boshier and Amaral 2004; McGaughran et al.2014), i.e., understanding the evolutionary and ecological processes and adaptations to changing environmental conditions which can only be achieved if neutral markers and phenotypic traits are combined (Tripiana et al.2007).Climate impact on the genetic structure of populations emphasizes the value of considering the environment as a cause of intraspecific patterns of variability.This is largely due to the fact that similar genetic patterns may result from diverse processes such as adaptation, dispersal limitation and genetic drift.To better understand the interactions of populations with their climate, forthcoming research evaluating populations across the whole Argentinean distribution ofA.colubrinavar.cebilwill integrate genetic and phenotypic data with techniques in spatial modelling and in future studies, ecological niche modelling based on climate predictions should be considered as another methodological approach to generate predictions about the geographical distribution of genetic variation in response to climate change.Our study provides evidence regarding the important role in shaping the phenotypic and genetic variation in which contemporary climate plays in the phenotypic and genetic variation inA.colubrinavar.cebil.In this way, although its distribution may be the result of historic climates, its phenotypic and genetic make-up is most certainly affected by the contemporary climate.

AcknowledgmentsM.V.García wishes to thank the Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET) for the fellowship within the framework of “Programa de Financiamiento Parcial para Estadías en el Exterior para Investigadores Asistentes”.