Tree Genetics & Genomes

, Volume 9, Issue 1, pp 107–116

Comparative chloroplast and nuclear DNA analysis of Castanea species in the southern region of the USA


  • Xiaowei Li
    • Department of HorticultureAuburn University
    • Department of HorticultureAuburn University
Original Paper

DOI: 10.1007/s11295-012-0538-z

Cite this article as:
Li, X. & Dane, F. Tree Genetics & Genomes (2013) 9: 107. doi:10.1007/s11295-012-0538-z


Boundaries between American Castanea species (Castanea dentata, the American chestnut and C. pumila var. pumila, the Allegheny chinkapin, and var. ozarkensis, the Ozark chinkapin) have been difficult to establish because of intraspecific variation, interspecific similarities and the incidence of chestnut blight, which has prevented trees from maturing. In this study, informative chloroplast (cp) DNA and nuclear sequences from Castanea taxa were analyzed to gain a better understanding of their phylogeography in North America. Our emphasis has been on the most southern Castanea population in the Appalachian region, known for its morphological diversity. This Ruffner Mountain (Alabama) population shows a high number of unique haplotypes, which can be divided into two main groups. One group shares homology with the widespread and evolutionarily recent C. dentata haplotype. The other group shares homology with American chestnuts and Allegheny chinkapin taxa from southern states. This group has been the result of recent and more ancient cp capture and hybridization, indicative of hybrid zone clustering and glacial refugial origin. The range of C. pumila must have been more extensive along the Coastal Plains region, since only a few mutations separate the Ozark chinkapin from the main Allegheny chinkapin haplotype. The geographic origin of the American Castanea species complex appears to be in the Gulf Coast region.


Castanea dentataCastanea pumilaChestnutAllegheny chinkapinOzark chinkapincpDNANuclear SNPsHybridization


The genus Castanea (Fagaceae), which includes seven species, is widely distributed in the Northern Hemisphere. Two species and two varieties can be found on the North American continent. American chestnut (Castanea dentata (Marsh.) Borkh.), with its characteristic three nuts per bur, once was the dominant canopy tree species in the Appalachian forest ecosystem. It possessed a remarkable array of desirable characteristics, grew very rapidly, often to great size, had outstanding form and wood quality, and provided food and revenue for rural communities. It ranged from Maine and southern Ontario to Mississippi, and from the Atlantic coast to the Appalachian Mountains and the Ohio Valley (Burnham 1988; Delcourt 2002; American chinkapin, with one nut per bur, consists of two varieties, the Allegheny chinkapin (Castanea pumila Mill. var. pumila) and Ozark chinkapin (C. pumila var. ozarkensis) (Johnson 1988). Ozark chinkapin is distributed in the extreme southwest of Missouri, northwest of Arkansas, and the extreme eastern portion of Oklahoma ( The Allegheny chinkapin performs best on well-drained soils in full sun or partial shade. Its range of adaptation is from northern Florida; west to Texas and Oklahoma; north to Kentucky, Virginia, and Maryland; and along the Atlantic coastal plain to Cape Cod, Massachusetts. The chinkapin is not economically important for nut and timber production because of its small nut and tree sizes and only provides a food source and community for wildlife (Payne et al. 1994).

Unfortunately, the American chinkapin as well as American chestnut are considered endangered species due to a devastating disease, chestnut blight, caused by the bark-inhabiting canker fungus (Cryphonectria parasitica). This fungus was introduced from Asia in the late 19th century (Milgroom et al. 1996). The disease has reduced the American chestnut from an important timber- and nut-producing tree to an understory shrub (Anagnostakis 1987; Kubisiak et al. 1997). The American chinkapin, which is closely related to the American chestnut, has almost been extirpated from several states in the southern US (Paillet 1993).

There are significantly different levels of resistance to chestnut blight among Castanea species. The Asian species (Chinese chestnut, C. mollissima BL.; Seguin chestnut, C. seguinii Dode; Chinese chinkapin, C. henryi Rehd. & Wils.; and Japanese chestnut, C. crenata Sieb. & Zucc.) exhibit high levels of resistance, while the European (C. sativa Mill.) and North American species are susceptible (Burnham 1988; Huang et al. 1996; Kubisiak et al. 1997). Although the deadly fungus cannot attack the root systems of these species, it destroys the shoots, the new life form underneath the forest floor. Almost always, the fungus kills the shoots before the trees flower. So reforestation cannot occur before the disease is controlled. From the early 1950s, many methods have been used to eradicate or to control this devastating disease. However, none of these approaches have been effective in orchards or forests (Griffin 2008). The American Chestnut Foundation (TACF), a nonprofit organization, aims to recover the American chestnut tree via an intensive breeding program using the Chinese chestnut (C. mollissima) species as a source of blight resistance and recurrent backcrosses to regionally adapted American chestnut trees ( Research has indicated that natural hybridization between Castanea species does occur (Rutter et al. 1991; Dane 2009) and accurate and unambiguous identification of Castanea species and their interspecific hybrids is needed for incorporation of American chestnuts into the breeding program.

Chloroplast (cp) genomes of plants are highly conserved in gene order and gene content and can provide phylogenetically useful information at various taxonomic levels (Avise 2004; Ames et al. 2007). Sequences from noncoding regions of the cp genome are often used in systematic analysis because such regions tend to evolve relatively rapidly and provide higher percentages of variable and informative characters as compared to cpDNA coding sequences (Taberlet et al. 1991). Several regions (such as trnT-L, ndhF, and ndhC with higher mutation rates) have been used to analyze phylogenetic and phylogeographic relationships between Castanea species (Lang et al. 2007; Dane and Lang 2008; Binkley 2008; Dane 2009). Sequencing results supported the genus Castanea as a monophyletic group with C. crenata (the Japanese chestnut) as basal. The North American species are supported as a clade with C. pumila var. ozarkensis as basal, sister to the group comprising C. pumila var. pumila and C. dentata (Lang et al. 2007). The genealogical network approach separates adjacent cp haplotypes with single mutation steps, guiding ancestral haplotypes at internal branching points, whereas younger haplotypes are placed toward more external or tip positions (Posada and Crandell 2001; Jakob and Blattner 2006). The cp network of American Castanea is partitioned into a small Ozark group and a large group of Allegheny chinkapin and American chestnut with shared haplotypes (Dane 2009). Boundaries between the American Castanea species are difficult to establish due to intraspecific variation, interspecific similarities, and interspecific hybridization. All species in genus are diploid (2n = 2x = 24) and hybridize freely (Jaynes 1975; Rutter et al. 1991). In these taxonomically uncertain groups, the integration of molecular and morphological data should help resolve systematic questions about their evolutionary relationships and origin.

Although the plant nuclear genome has a large size and large number and diversity of genes, conserved regions are often used for phylogenetic studies (Soltis et al. 2009). In plant studies, nuclear markers, usually combined with cpDNA, are used to check genetic diversity, differentiation, and phylogenetic hypotheses of species, genus, and even family (Muir et al. 2004). Also nuclear DNA has been investigated for plant species delineation and inference of evolutionary relationships (Eidesen et al. 2007; Soltis and Soltis 2009). In this study, we compared and analyzed sequences from different Castanea populations based on several cpDNA regions (trnT-L, trnL, ndhF, ndhC, ycf9, and rpl16) and two polymorphic nuclear regions, with the intent to assess the extent of haplotype sharing and hybridization between Castanea species in order to gain a better understanding of their phylogeography in North America. Our emphasis is on the most southern Castanea population in the Appalachian region, known for its morphological diversity. Historical factors such as the last glacial period and subsequent recolonization of species into lands newly released by ice have played a role in determining the present patterns of genetic variation in Castanea populations (Davis 1983; Dane and Lang 2008). Many tree species that once inhabited areas north of the glacial boundary were restricted to smaller ranges south of ice or to refugia along the Gulf Coast (Delcourt 2002) with clustering of contact and hybrid zones and suture zones in central and eastern Alabama (Swenson and Howard 2005). Interglacial pollen records of Castanea in fossil lakes in Florida date back to 35,000 to 30,000 BP (Watts and Hansen 1994). Analysis of preserved pollen samples suggest that Castanea migrated very slowly northward and arrived in Connecticut only 2,000 years ago (Davis 1983). This study investigates if southern populations harbor ancestral haplotypes and show higher incidences of haplotype sharing as a result of interspecific hybridization.

Materials and methods

Plant material and morphological analyses

C. dentata samples from four populations (KY, NC, GA, and AL) in the southern region of the US and one population in the northeast (CT) were used with representative samples from populations across the species range (Table 1). Major emphasis was on the most southern population at Ruffner Mountain Nature Center near Birmingham, Alabama (Fig. 1) because of its morphological diversity. Similarly, representative C. pumila var. ozarkensis and var. pumila were included for comparative analysis and C. sativa, C. mollissima, and C. crenata as outgroup taxa (Table 1). Wherever possible fresh leaf samples were collected and examined by F. Hebard at TACF, Meadowview, VA using diagnostic quantitative and qualitative characteristics such as presence, density, shape, and length of hairs on leaves, and color of stem, as described by Jaynes (1975) and Johnson (1988).
Table 1

Collection site information of Castanea samples and cpDNA haplotypes


Sample ID

County, state, or country


C. dentata

AL-TAL (T2, T12)

Talladeega, AL

2 HD11

C. dentata


Lacon, AL

2 HD2

C. dentata

NC-C2, NC-C4,

Coweeta County, NC

2 HD1

NC-C9, NC-C11, NC-C14, NC-C22, NC-C27, NC-C37

6 HD5

NC-C58, NC-C60

2 HD7

NC-C45, NC-C51

2 HP20

C. dentata

KY-LW11, KY-LW23

Laurel and Whitney County, KY


KY-LW3, KY-LW14,KY-LW18, KY-LW21

4 HD4

KY-LW26, KY-LW32

2 HD5

C. dentata


Lula Lake, GA

1 HD7

GA-LL16, GA-LL17, GA-LL-20, GA-LL36, GA-LL38, GA-LL42, GA-LL43

7 HD6

C. dentata

AL-5, AL-6, AL-IZ1, AL-7CN, AL-AL2, AL-M18, AL-M33, AL-M38, AL-M60, AL-M61

Ruffner Mountain Nature Center, Birmingham, AL

10 HD12

1 HD7


1 HD13


AL-M001, AL-M30, AL-M31, AL-XL1, AL-M35, AL-


11 HR2

1 HR1

M36, AL-M37, AL-M40, AL-M65, AL-M67, AL-M68


C. dentata


Amherst, Maine


C. dentata


Ontario, Canada


C. dentata

BA22, BA33


2 HD1

C. dentata

CT1, CT2, CT3, CT4, CT5, CT6, (C D1 or C dent)


6 HD1

C. pumila var. ozarkensis

CCRgammaT3 (P7), CCRetaT3 (P9)

Russellville, AR


C. pumila var. ozarkensis

AR-E1-3, AR-E9-3, AR-M1-8, AR-B4-3

Sylamore Ranger District, AR

4 HO2


1 HO3

C. pumila var. ozarkensis


Oregon Ct, Missouri

1 HO1

C. pumila var. ozarkensis



2 HO2,

1 HO1

C. pumila var. pumila


Varnamtown, NC

3 HP21

C. pumila var. pumila

VA-A3, VA-A4

Iron Mountain, VA

2 HP11

C. pumila var. pumila

VA-B5, VA-12, VA-18

Iron Mountain, VA

3 HP1

C. pumila var. pumila



1 HP23

C. pumila var. pumila


Saucier, MS


C. pumila var. pumila


Eglin Air Force Base, FL

1 HP3

FL-E3, FL-P4, FL-G6, FL-I3

4 HP4

C. pumila var. pumila

NC1, NC4

Pisgah Forest, NC

2 HP1

C. pumila var. pumila


Rabun Ct, GA

1 HP7


C sat1

HHR3T2, New Haven, CT


C. sativa

C sat7



C. crenata

C cre1

NLR34T6, New Haven, CT


C. crenata

C cre5

WL∆, New Haven, CT


C. mollissima

C mol2

SLR1T15, New Haven, CT


C. mollissima

C mol4

WLR38T4, New Haven, CT
Fig. 1

Taxa distribution at Ruffner Mountain, Alabama. Taxa 4CN, 7CN, AL2, M18, M33, M38, M60, M68, 5, 6, IZ1, and PT-20 can be characterized morphologically as type I (haplotypes HD7, HD12, and HD13), and taxa XL1, M30, M31, M34, M35, M36, M37, M40, M001, M61, M65, and M67 as type II (HR1, and HR2)

DNA extraction, PCR, and nucleotide sequencing

DNA was extracted from nuts of C. pumila using the DNeasy™ plant mini kit (Qiagen, Valencia, CA). DNA extractions were made from fresh leaf material using CTAB (hexadecyl trimethylammonium bromide) method (Kubisiak and Roberds 2003). In addition to published cpDNA sequences (Taberlet et al 1991; Lang et al. 2006), the variable regions rpl16 intron, ycf9-trnGM in the large single copy (LSC) and ndhF region in the small single copy (SSC) of the cp genome were chosen for analysis (Heinze 2002; Lang et al. 2007). Taxa from almost all populations with the exception of the Connecticut and Ruffner Mountain (AL) populations were reanalyzed with sequence information from the variable ndhC-trnV2F region (Shaw et al. 2007). cpDNA haplotypes were defined on basis of the combined sequence information from six regions which had previously been shown to be polymorphic within and between Castanea species (Lang et al. 2007; Shaw et al 2007). C. dentata haplotypes are classified as HD, C. pumila var. pumila as HP, var. ozarkensis as HO, and Ruffner Mountain as HR. Wound or blight fungus infection responsive expressed sequence tags (ESTs) from American and European chestnut available in the GenBank database ( were used for the design of primers to identify sequence polymorphisms unique to each Castanea species. Primer pair 17, designed using American chestnut EST (BG835820), and primer pair 126 (CR628009; Casasoli et al. 2006) were used in this study to amplify small intronic nuclear regions (Table 2). Double-stranded DNA amplifications were performed in a 55-μL volume containing 1× PCR buffer (New England Biolabs, Ipswich, MA), 200 μmol/L of each dNTP, 0.2 μmol/L of each primer, 2 U of Taq polymerase (New England Biolabs), and 2.5 μL template DNA (50 ng/L). PCR products were purified using Qiaquick PCR purification kit (Qiagen, Valencia, CA) to remove excess primers and dNTPs. Sequencing of PCR products was conducted by Auburn Genomics and Sequencing Lab with the ABI3100 sequencer (Applied Biosystems Inc., Foster City, CA).
Table 2

Primers used for nuclear DNA sequencing analysis (from 5′ to 3′)







17 F


55/72 °C






126 F


53/72 °C

Casasoli et al. (2006)





Sequence alignment and data analyses

Multiple alignments of the sequences were carried out at ClustalW at the default setting, using the AlignX program implemented in the Vector NTI software, and adjusted manually. Gaps were introduced in the alignment in order to optimize positional homology. Single base indels were cross-checked to the original chromatograms to verify that they were not sequencing artifacts missed during base calling. Indels that were potentially parsimonious were scored and added to the end of the datasets as present (1) or absent (0) type characters. Gaps with overlaps were considered nested and treated as single multistate character according to Simmons and Ochotreana (2000). Areas of ambiguous alignment were excluded. A maximum parsimony analysis was conducted using PAUP version 4.0 software (Swofford 2000). Sequences were aligned with cpDNA sequence information from other informative C. dentata and C. pumila populations (Dane and Lang 2008; Lang et al. 2006, 2007; Dane 2009) (Table 1). Nuclear DNA sequences (17 and 126 regions) were aligned with sequences from other known Castanea taxa. Haplotype networks were developed manually, since TCS (Clement et al. 2000) did not provide enough resolution for these species with shared haplotypes and large indels.


North American Castanea taxa are taxonomically challenging especially in the southern Appalachian regions where species ranges overlap. Classification is based primarily on leaf morphology and where possible, variation in fruit and cupule numbers. It is difficult to discern American Castanea species on basis of leaf morphological characteristics, since the criteria used are influenced by environmental and climatic factors. The most southern Castanea population at the Ruffner Mountain Nature Center, Alabama's unique Red Mountain ridge near Birmingham, can be divided into a group (I, 12 samples) of trees with long simple hairs mainly on midribs of the glabrous leaves, occasionally with a few interveinal stellate hairs off the midrib and purplish stem, typical of C. dentata. The leaves of the second group (II, 12 samples) have simple hairs on midrib, minor veins, interveinal regions, and stalked glandular hairs, indicative of intermediacy in morphological characteristics, since leaves of C. pumila var. pumila and var. ozarkensis are smaller, variable, and puberulent to tomentose. Group I samples occur mostly in the southeast region of the nature center (Fig. 1) and chinkapin-type samples in the northeast region. The type II shrubs might be the result of hybridization between the American chestnut and Allegheny chinkapin or remnants of once more extensive C. dentata populations.

cpDNA haplotype networks and phylogenetic analyses

C. dentata populations in the southern range of the species show a higher number of cpDNA haplotypes, with three to four haplotypes in populations from North Carolina, Kentucky, and Georgia, and seven haplotypes in Alabama (Table 1). Analysis of unique cpDNA haplotypes from American chestnut populations with indels or multinucleotide hairpin-type mutations coded as extra binary characters was conducted using comparative sequences from representative C. pumila haplotypes and C. sativa as outgroup. This resulted in a most parsimonious tree with 103 steps, a consistency index (CI) of 0.6139 and a retention index (RI) of 0.8152. The 50 % majority rule consensus tree is presented in Fig. 2. Four clades are evident from the cpDNA phylogeny.
Fig. 2

Fifty percent majority rule consensus of 240 trees inferred from comparative analysis of Castanea dentata, C. pumila var. pumila, and var. ozarkensis cpDNA sequences (CI = 0.6139 , RI = 0.8152) using C. sativa as outgroup

One clade is composed of several C. dentata haplotypes and one C. pumila var. pumila haplotype from northeastern Georgia, characterized by two unique deletions (12 and 72 bp) at trnT-L. Indel variability at this region has been used to separate populations into a unique and widely distributed C. dentata haplotype, HD1, and northern populations are often fixed for this haplotype (Kubisiak and Roberds 2003, 2006). Haplotype HD1 is also characterized by a 35-bp insertion at ndhF and 49 bp deletion at ndhC. However, indel variability at these regions does exist in haplotypes of this clade. The insertion at ndhF (35 bp), for example, is absent from one Ruffner Mountain haplotype (HD13). The Ruffner Mountain type I taxa (haplotypes HD12 and HD13) have a multinucleotide hairpin change at a section of the ndhC region, which has been deleted in HD1. These multinucleotide changes also occur in other American chestnuts (HD2) and in the Allegheny chinkapin population from northeastern GA (HP7). To better understand processes resulting in patterns found in these cpDNA haplotypes, a network approach was used manually to develop a stepwise mutation model phylogeny with indels, SNPs, and the multinucleotide change coded as single mutational events (Posada and Crandall 2001). Only in a few cases was the presence of indels or SNPs inconsistent because of lack of contemporary intermediate haplotypes. In those cases, the most likely haplotype network was developed (Fig. 3).
Fig. 3

Haplotype network of Castanea dentata with closely related C. pumila var. pumila, indicating unique indels and mutational events. Circle size reflects frequency of haplotypes, hash marks are single mutational events (transitions or transversions), diamond multinucleotide SNP, and boxes indicate indels with length in base pair

The second clade (Figs. 2 and 4) includes a mix of taxa collected from Allegheny chinkapin and American chestnut populations in KY, VA, NC, and GA and, and Ruffner Mountain. The Ruffner Mountain type II haplotypes (HR1 and HR2) show several unique deletions at ndhF (Fig. 4) with one exception (HD7). Most of the taxa in this clade exhibit minor mutational differences, with the exception of haplotype HP11 which has the 35-bp insertion at ndhF, characteristic of C. dentata haplotype HD1, and occurs at low frequency in an Allegheny chinkapin population in Virginia.
Fig. 4

Haplotype network represents the minimum number of mutational events between major Ozark chinkapin (C. pumila var. ozarkensis), Allegheny chinkapin (C. pumila var. pumila), and American chestnut (C. dentata) haplotypes. Circle size reflects haplotype frequency; hash marks transversions, transitions, or cpSSRs; boxes indels (with length in base pair)

The third clade contains Florida Allegheny chinkapin taxa, which occur in populations at the Eglin Air Force Base and harbor unique haplotypes, distinguishable from others by several indels and SNPs at the ndhC and ycf9 regions. The fourth clade contains both Ozark haplotypes from Arkansas and Missouri (HO1, HO2, and HO3) and the most common Allegheny chinkapin haplotype (HP1) widely distributed across Georgia, North Carolina, and Virginia and a haplotype (HP20) from North Carolina.

Ancestral or ancient haplotypes can be identified using comparative sequence analysis. At the rpl16 and trnL regions, known for their deep evolutionary splits in the genus (Lang et al. 2007), ancestral regions are shared between C. sativa, C. dentata haplotype HD1, and the Florida Allegheny chinkapin population (haplotype HP3). Figure 4 illustrates the close relationship between the Ozark and the most common Allegheny chinkapin haplotype (HP1) found in Virginia and North Carolina, indicative of a once more widespread range of C. pumila across the Coastal Plains region. Allegheny chinkapin haplotype HP3 from northern Florida is closely related to haplotypes from American chestnut populations from GA, KY, AL, and NC. A low number of mutational events separate HP3 from C. dentata haplotypes HD5–7 (>3 mutations, Fig. 4), C. dentata HD1 (five indels as mutational events), but a higher number from Ozark haplotypes (HO1–3). Predictions assume that older haplotypes can be characterized by a high number of descending lineages (Posada and Crandall 2001). This would indicate that the Allegheny chinkapin haplotype HP3, located at an intermediate position of the network, serves as a bridge between the Ozark haplotypes (without trnT-L deletions) and evolutionary recent C. dentata networks (with trnT-L deletions, Figs. 3 and 4) and can be considered as ancient, while other haplotypes are more derived.

Nuclear DNA marker analysis

Primers designed from wound-specific Castanea ESTs (Schafleitner and Wilhelm 2002) were used to screen genomic DNA from different Castanea species to identify sequence polymorphisms specific for each species. Only intronic regions amplified with two primer pairs (17 and 126) were used here for combined sequence analysis (680 bp), since other amplified regions showed intraspecific variability which could not be used for species identification (Dane, unpublished). Asian and European chestnut species were used as outgroup for parsimony analysis (Table 1). The presence of conserved SNPs can divide the taxa into a C. pumila clade (Fig. 5), which contains Ozark and Allegheny chinkapin and the Ruffner Mountain type II taxa (with the exception of AL-4CN, cpDNA haplotype HD7). The second clade contains taxa from American chestnut populations across the species range and many different cpDNA haplotypes. The species specificity of these nuclear SNPs (two at region 126; one at region 17) was tested using C. dentata and C. pumila taxa from across the range of each species. cpDNA haplotypes classified as HD (and C. sativa haplotypes) show the C. dentata nuclear type, while the Ruffner Mountain type II taxa (HR haplotypes) and HP and HO haplotypes show the C. pumila nuclear type. American chestnut samples from southern populations without the evolutionary recent (12 and 72 bp) deletions at trnT-L are placed in the C. dentata group with HD1 (Fig. 5). Asian species show unique SNPs at region 126. GA-JJ7 shows heterozygosity at those sites, indicative of a recent interspecific hybrid with an Asian Castanea species as the paternal parent. Similar results were observed for a known F1 hybrid (KY110) between C. mollissima × C. dentata. Thus recent interspecific hybrids can be identified using cp and nuclear SNPs.
Fig. 5

Fifty percent majority rule consensus inferred from comparative analysis of 680 bp of Castanea nuclear sequences using C. sativa, C. mollissima, and C. crenata as outgroup


Throughout history, geographical and climatic events have had an effect on the range of North American plant species (Hewitt 2000). More specifically, glacial events of the Pleistocene have impacted the distribution of American Castanea species and resulted in repeated glacial cycles with postglacial expansion from southern refuge areas into northeastern North America (Davis 1983). This large geologically and topologically complex area is characterized by several terrestrial discontinuities such as the Appalachian mountain range and Apalachicola, and Tombigbee and Mississippi river basins (Soltis et al. 2006). Refugia in southern Appalachians around central and eastern Alabama/Georgia and clustering of hybrid zones and suture zones have been predicted for several plant and animal species (Davis 1983; Griffin and Barrett 2004; Swenson and Howard 2005; Amato et al. 2008; Morris et al. 2008). Predictions of refugial origin might be determined by locating populations of increased intra- and interspecific genetic variation (Hewitt 2000; Petit et al. 2003; Griffin and Barrett 2004). Our research focused on the Ruffner Mountain Castanea population, the most southerly population in the Appalachian Mountain region. This population of shrubs shows morphological and molecular diversity, pointing to clustering of contact, hybrid zones, and possibly glacial refugial origin. It contains several unique cpDNA haplotypes and can be divided into two groups based on leaf morphology, cp, nuclear indels, and SNPs.

Evolutionary recent deletions at the trnT-L cp region have been used to distinguish C. dentata from other Castanea populations. Deletions at trnT-L are absent from all Asian and European chestnut species and American and Chinese chinkapin (C. henryi) species (Lang et al. 2006). However, American chestnut trees from southern populations do not always carry the above-mentioned deletions, while some Allegheny chinkapin trees do; thus, the characteristic deletions can only be used to detect the evolutionarily recent C. dentata haplotypes. The major 49-bp deletion at the ndhC region similarly is evolutionarily recent, since it has only been detected in C. dentata haplotype HD1, not in other Castanea species. The 35-bp insertion at ndhF has been observed in several C. dentata haplotypes (HD2, HD1, and HD12), the C. pumila var. pumila haplotype from northeastern Georgia (HP7) and from Virginia (HP11), but not in other Castanea species. Thus American chestnut haplotypes without these characteristic indels are more ancient, as the Ruffner Mountain haplotype HD13, since it lacks the (35 bp) ndhF insertion observed in haplotype HD12 at Ruffner Mountain (Fig. 2). Since the C. pumila var. pumila haplotype (HP7) shares the characteristic C. dentata indels, hybridization must have occurred over time between the species. Also, the unique type II (HR) haplotypes of the Ruffner Mountain population (with C. pumila nuclear SNPs) group with American chestnuts from southern states and Allegheny chinkapins from across the species range, clearly a hybrid group of taxa. Sharing of cpDNA haplotypes has been detected among many species, especially those with overlapping geographical distribution, and has been attributed to gene flow and introgression or incomplete sorting of ancient lineages (Avise 2004; Guicking et al. 2011). It is known that hybridization has played an important role during plant speciation and evolution (Soltis and Soltis 2009). Boundaries between the American Castanea species have been difficult to establish especially in southern states because of intraspecific variation, interspecific similarities, and clearly hybridization in contact zones. This was also found by Binkley (2008), who focused on a Castanea population with variable leaf morphology in Georgia. The American chestnut was driven to clonal propagation by the chestnut blight epidemic. Lack of sexual reproduction over the past 70 years has produced a significant change in the population genetics of the species, favoring heterozygous individuals with higher growth rates (Stilwell et al. 2003).

The range of C. pumila must have been more extensive along the Coastal Plains region, since only a few cpDNA mutations separate the Ozark chinkapin haplotypes distributed in Arkansas and Missouri from the common and widely distributed Allegheny chinkapin haplotype found in Virginia and North Carolina. Several other Allegheny chinkapin haplotypes from Virginia are separated by more mutational events from the Ozark chinkapin haplotypes (Fig. 4). The Florida C. pumila haplotype (HP3) serves as intermediate in the network between Ozark and Allegheny chinkapin and several different C. dentata haplotypes, which points to the Gulf Coast region as the center of origin for Castanea on the American continent. The nuclear markers separate the American Castanea taxa into a chinkapin group (which includes the type II Ruffner Mountain taxa) and an American chestnut group (Figs. 25) of evolutionary ancient and more recent cpDNA haplotypes. Alabama is known as a divergence hotspot for closely related plant and animal species (Swenson and Howard 2005; Soltis et al. 2006; Rissler and Smith 2010). This area has a diversity of geological substrata, relatively unaffected by Pleistocene glaciations, and harbors multiple relic and endemic species. Species moving north and south during glacial advances and retreats could have had populations entering multiple geological formations driving divergence. Since reproductive barriers are not absolute between Castanea taxa (Johnson 1988; Rutter et al. 1991), hybridizations occurred frequently in the overlapping regions. This explains the complex phylogeographic patterns observed in these American Castanea species. Further studies utilizing single copy nuclear gene phylogenies are needed to obtain more information on the evolutionary relationships of this hybrid species complex.


The authors thank The American Chestnut Foundation for partial financial support and members from its state chapters, especially Martin Schulman and Fred Hebard, for help with collection and identification of field sampling.

Copyright information

© Springer-Verlag 2012