Aim:  Predictions of species’ responses to accelerating global climate change require an understanding of historical range shifts. However, large-scale phylogeographic studies of Eastern North American understory plant taxa are relatively scarce. Here we employ ecological niche modelling and genetic analyses for inference of optimal pawpaw habitat in the past and future. 

Location: Twenty-six states in the eastern United States

Taxon: Asimina triloba (L.) Dunal (Annonaceae)

Methods:  The present-day niche of Asimina triloba was modelled in Maxent using seven bioclimatic variables, elevation, and location data from field samples and herbarium specimens. To model historically optimal habitats, the present-day model was projected onto rasters of seven bioclimatic variables and elevation representing the last glacial maximum (~22,000 years before present [YBP]) and the mid-Holocene (~6,000 YBP). Predicted habitat suitability for 2070 was also modelled. Additionally, 62 populations were genotyped with nine nuclear microsatellite loci and statistically analyzed. Levels and partitioning of genetic variation within and among populations were estimated within a geographic context.

Results:  Models indicate that optimal habitat 22,000 YBP was severely restricted to now-submerged Gulf of Mexico and southeastern U.S. coastlines. By 2070 models suggest that optimal habitat will expand substantially northward relative to the present. Species-level genetic diversity (H_E = 0.765) was high and genetic structure among populations was moderate (G_ST = 0.202). Structure indicates that there are two population clusters straddling the Appalachian Mountains.

Main conclusions:  Models suggest that 22,000 YBP A. triloba was restricted to two major refugia in narrow bands of now-submerged habitat and one small inland refugium in southeastern Alabama and southwestern Georgia. Molecular data are consistent and suggest that the two eastern refugia expanded to give rise to the eastern cluster which is characterized by higher genetic diversity. The Texas/Louisiana refugium likely gave rise to populations in the western cluster, characterized by lower genetic diversity. 

Georeferenced Occurrence Data

GPS coordinates were obtained from 82 naturally occurring populations throughout the United States as well as from 18 herbarium specimens collected after 1992 (herbaria: DEK, ILLS, IND, KSP, MO, TROY and USF). Prior to 1992, the U.S. government scrambled GPS signals rendering them too inaccurate to be useful. 

Genetic Data

Leaf samples were collected from naturally occuring plants and snap-frozen in liquid nitrogen for transport to the University of Georgia. Frozen leaf tissue (~ 0.05 g) was ground to a powder and genomic DNA was extracted using a modified CTAB protocol (Doyle & Doyle, 1983). DNA quality and quantity were evaluated using a ND-1000 Nanodrop spectrophotometer. Nine nuclear microsatellite loci were amplified with markers developed by Lu et al. (2011). A 3-primer PCR protocol was used whereby a CAG-tag sequence (Hauswaldt & Glenn, 2003) was added to the 5’ end of one primer and a third fluorescently labeled (FAM, HEX, or NED) primer identical to the CAG-tag was included in the reaction. PCR amplification was carried out in 12.5 mL reaction volumes consisting of 0.26 mM molecular grade ddH2O, 1 mM One Taq standard reaction buffer (New England Biolabs, Ipswich, MA), 0.7 mM MgCl solution (Sigma-Aldrich), 1 mM bovine serum albumin (Thermo-Fisher), 0.2 mM dNTP mix (New England Biolabs, Ipswich, MA), 0.04 mM CAG-tagged primer, 0.4 mM unlabeled primer, 0.36 mM universal fluorescently labelled CAG primer, 0.2 mM One Taq Hot Start DNA polymerase (New England Biolabs, Ipswich, MA), and 50 ng template DNA.

Genetic Data

The dataset consists of codominant neutral nuclear markers.  Missing data is designated as "0".