| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systemics and Phytogeography |
2Department of Ecology and Evolutionary Biology, 569 Dabney Hall, The University of Tennessee, Knoxville, Tennessee 37996 USA; 3Department of Biological and Environmental Sciences, 615 McCallie Avenue, The University of Tennessee, Chattanooga, Tennessee 37403 USA
Received for publication March 10, 2005. Accepted for publication August 31, 2005.
ABSTRACT
The North American plums are a closely related group that are not easily circumscribed, have overlapping morphologies, and are known to hybridize. We previously showed that the North American plums are a closely related, monophyletic group of taxa with little to no cpDNA sequence divergence between taxa. In that study, we came to the unanticipated conclusion that relationships inferred among the taxa contrast sharply with previously defined groups based on morphological characters. Here the aim was to determine if the primary cpDNA haplotypes identified in our earlier study are confined to the taxa in which they were initially observed. The cpDNA rpL16 intron was sequenced for 207 accessions of the 17 North American plum taxa plus Prunus texana. The results show that many taxa contain more than one of the three primary cpDNA haplotypes. Aside from the results found in sect. Prunocerasus, this study has broader implications for phylogenetics in general. The common practice of choosing a single exemplar to represent a taxon can be profoundly misleading in closely related groups. In hindsight, the possibility existed in our earlier study that we could have chosen a different combination of exemplars, which could have resulted in a different inferred phylogeny.
Key Words: cpDNA • hybrid • phylogeny • phylogeography • plum • Prunocerasus • Prunus • Rosaceae
Morphological taxonomy has been notoriously difficult within the North American plums, Prunus L. subg. Prunus sect. Prunocerasus Koehne (Rosaceae), because species boundaries are blurred by interspecific similarities and intraspecific variation and likely by interspecific introgressive hybridization (Waugh, 1899
, 1901
; Hedrick et al., 1911
; Wight, 1915
; Rehder, 1940
; Shaw and Small, 2004
). This has historically led to species oversplit within sect. Prunocerasus where species were named based on characters as taxonomically questionable as pubescence and geography. However, relatively little taxonomic work has been done on the North American plums in the last ca. 60 years, and throughout North America taxonomists have settled on names in a fashion that brings to mind lineage sorting. Because there are few discrete morphological characters available for constructing a cladistic hypothesis for sect. Prunocerasus, we are relegated to DNA sequence data to provide insight into the relationships among the taxa within the section.
In a previous study (Shaw and Small, 2004
) that employed seven noncoding cpDNA regions (totaling 4375 bp), we showed that the North American plums are a monophyletic assemblage of taxa but that there is little genetic divergence between species. In that study, we selected single representatives of each of the 17 commonly accepted North American plum taxa plus P. texana and showed that the majority segregate into a tritomy sister to P. subcordata, the only species from western North America (Fig. 1). Although these three primary clades were each supported by high bootstrap values, there was little to no phylogenetic resolution within clades. An unexpected result of that study was that the composition of the three clades contrasts sharply with previously defined groups based on morphological characters (Waugh, 1899
, 1901
; Wight, 1915
; Rehder, 1940
; see fig. 3 in Shaw and Small, 2004
)—each of the three clades contain taxa that are morphologically more similar to taxa of other clades. Comparison of the species within the Chickasaw clade (Fig. 1; see also fig. 3 of Shaw and Small, 2004
) highlights the lack of genetic divergence and the discontinuity between the cpDNA phylogeny and classical morphological groups. Of six species within the Chickasaw clade, the 4375 bp of noncoding cpDNA sequences were identical with the exception of two autapomorphies. While some of the taxa within this clade are nearly identical morphologically, others are easily distinguished from the rest. The lack of genetic divergence within these three primary clades coupled with the obvious discontinuity with respect to longstanding morphological groups leads to the question of whether or not chloroplast lineages are monophyletic within the taxa of sect. Prunocerasus. This question can be addressed though the additional sampling of multiple accessions of each species within the section in a phylogeographic-level sampling scheme.
|
; Soltis et al., 1992
), relatively few researchers have addressed the frequency of chloroplast sharing among species in closely related groups. To date, most phylogenetic investigations have used the strategy of sampling one or few representatives per species (as opposed to "population" studies that involve sampling multiple accessions within a single species or closely related species pair). In sampling a single representative per species, one makes the following assumptions: (1) the study species are monophyletic, (2) the study species represent genetically and reproductively isolated lineages that are not reticulating through introgressive hybridization (which is important because the lack of recombination of the chloroplast genome means that it introgresses as a linked block and can therefore be a source of error in molecular systematics [Doyle, 1992
; Rieseberg and Wendel, 1993
; Riesberg et al., 1996
]), (3) the boundaries between the study species are well characterized, and (4) the exemplar specimens are accurately identified. Therefore, in closely related groups of species, where one or more of these assumptions may be violated, sampling multiple individuals from each of the representative species is the only means of accurately assessing phylogenetic relationships. For example, Whittemore and Schaal (1991)
cautioned that a single accession of each species could have been misleading in their study of hybridizing oak species. Funk and Omland (2003
, p. 414) wrote that ideally one should "include all species believed a priori to be closely related (i.e., congeners), maximize the geographic diversity of samples and the number of samples collected from areas of sympatry between study species [this same suggestion was made by Matos and Schaal (2000)
], and sample broadly from known sources of biological variation (subspecies, ecotypes, morphological variants, etc.)." Furthermore, Funk and Omland (2003)
call for a new tradition of "congeneric phylogeography" where phylogenetics and phylogeography are united yielding a more population-level sampling scheme that will improve the resolution of evolutionary relationships among closely related entities.
While the concept of phylogeography (Avise, 1987
) has had a large impact on research in animal systems over the last 18 years, it has been relatively slowly applied to botanical systems because of the relatively slow mutation rate of organellar DNA in plants (Schaal et al., 1998
). This may be remedied to some extent after more informative regions of the cpDNA molecule have been better characterized (e.g., Shaw et al., 2005
; J. Shaw, E. Lickey, E. Schilling, and R. Small, unpublished data). Botanical phylogeography has successfully been applied to plant systems in Eurasia/North Africa (e.g., Comes and Abott, 2001
; Besnard et al., 2002
; Palme and Vendramin, 2002
; Burban and Petit, 2003
; Dane et al., 2003
; Grivet and Petit, 2003; Palme et al., 2003
) and Japan and East Asia (Chiang et al., 2001
; Fujii et al., 2002
; Honjo et al., 2004; Kanno et al., 2004). Comparatively, the botanical phylogeography of North America has been much less studied, and of those studies, most have been restricted either in the number of species or in geography (e.g., Parks et al., 1994
; Sewell et al., 1996
; Manos et al., 1999
; Tremblay and Schoen, 1999
; Abbott et al., 2000
; Sanjur et al., 2002
; Dobes et al., 2004
; Jørgensen and Mauricio, 2004
).
Most studies to investigate interspecific vs. intraspecific cpDNA variation are "population" in nature and confined to relatively few species (e.g., Mason-Gamer et al., 1995
; Comes and Abbott, 1998
, 1999
; Maskas and Cruzan, 2000
; Matos and Schaal, 2000
), but a few studies have compared interspecific vs. intraspecific cpDNA variation among a number of hybridizing species (e.g., Whittemore and Schaal, 1991
; Bain and Jansen, 1997; Dumolin-Lapegue et al., 1997
; Manos et al., 1999
; Comes and Abbott, 2001
; Gardner et al., 2004
; Kanno et al., 2004). A common thread throughout these studies is the occurrence of chloroplast sharing among closely related species. Furthermore, if hybridization is frequent, chloroplasts may be distributed geographically instead of taxonomically (Quercus, Whittemore and Schaal, 1991
; Quercus, Dumolin-Lapegue et al., 1997
; Eucalyptus, Jackson et al., 1999
; the Pinus montezumae complex, Matos and Schaal, 2000
).
The present study addresses two issues in Prunus subg. Prunus sect. Prunocerasus. Hybrid speciation has been hypothesized to occur within the genus Prunus (Watkins, 1976
; Brettin et al., 2000
; Mohanty et al., 2000
) in general, and specifically in sect. Prunocerasus (Waugh, 1899
, 1901
; Steyermark, 1963
). Because the species of sect. Prunocerasus are capable of natural hybridization (Hedrick et al., 1911
; Wight, 1915
; Flory, 1938
; Rehder, 1940
), we sampled extensively within each of the putative taxa to determine if the three primary cpDNA haplotypes observed by Shaw and Small (2004)
(Fig. 1) are confined to the taxa in which they were initially observed. The second issue is phylogeographic in nature. Because we are sampling heavily within each of the putative taxa of sect. Prunocerasus, we can analyze the distribution of cpDNA haplotypes to assess geographic structure in their distribution.
MATERIALS AND METHODS
Taxon sampling
Using both wild-collected and herbarium specimens (TENN, APSU, BRIT, MICH, BH), leaf material was selected from multiple accessions of each taxon in an attempt to sample each taxon throughout its respective range. A total of 207 accessions from the 17 North American plum taxa plus P. texana (18 taxa) were included (Table 1). All of the commonly accepted members of sect. Prunocerasus were sampled, according to Small (1933)
, Rehder (1940)
, Fernald (1950)
, Bailey and Bailey (1941)
, Blackburn (1952)
, Gleason (1952)
, Steyermark (1963)
, Radford et al. (1968)
, Correll and Johnston (1970)
, Duncan and Duncan (1988)
, Godfrey (1988)
, Wunderlin (1988)
, Gleason and Cronquist (1991)
, Smith (1994)
, and Wofford and Chester (2002)
. A single sample of the federally endangered P. geniculata was obtained from Carl Weakley at the Archbold Biological Station in Lake Placid, Florida, USA.
|
DNA sequences
The rpL16 intron region was chosen because Shaw and Small (2004)
showed it to contain at least one synapomorphic character for each of the three primary clades revealed in that study. Also, the characters identifying the three primary clades are all within approximately 650 bp from the 3' end of the intron and could therefore be obtained by one sequencing reaction using the rpL1516R primer (see Laboratory procedures).
Laboratory procedures
DNA was extracted from leaves using the DNeasy Plant Mini kit (Qiagen, Valencia, California, USA). The polymerase chain reaction (PCR) was performed using Eppendorf (Westbury, New York, USA) Mastercycler gradient or Mastercycler personal thermal cyclers in 50 µL volumes with the following reaction components: 1 µL template DNA (
10– 100 ng), 1x ExTaq buffer (PanVera/TaKaRa, Madison, Wisconsin, USA), 200 µmol/L each dNTP, 3.0 mmol/L MgCl2, 0.1 µmol/L each primer, and 1.25 units ExTaq (PanVera/TaKaRa). Reactions included bovine serum albumin at a final concentration of 0.2 µg/µL, which is known to improve amplification from difficult templates.
PCR and sequencing primers rpL16F71 (GCT ATG CTT AGT GTG TGA CTC GTT G) and rpL16R1516 (CCC TTC ATT CTT CCT CTA TGT TG) for the rpL16 intron are from Small et al. (1998) and Shaw et al. (2005)
. The PCR protocol described next was preceded by template DNA denaturation at 80°C for 5 min and followed by a final extension step of 5 min at 65°C. The PCR cycling conditions were 30 cycles of denaturation at 95°C for 1 min, primer annealing at 50°C for 1 min, followed by a ramp of 0.3°C/s to 65°C, and primer extension at 65°C for 4 min.
PCR products were checked on 1% agarose gels before being cleaned with ExoSAP-IT (USB, Cleveland, Ohio, USA). In most cases, DNA sequencing was performed using only the reverse primer rpL16R1516; however, in a few cases, the forward primer was needed. All DNA sequencing was performed with the ABI Prism BigDye Terminator Cycle Sequencing Ready Reaction kit, v. 2.0 or 3.1 (Perkin-Elmer/Applied Biosystems, Foster City, California, USA) and electrophoresed and detected on an ABI Prism 3100 automated sequencer (University of Tennessee Molecular Biology Resource Facility, Knoxville, Tennessee, USA). Sequencher 3.1.1 (Gene Codes, Ann Arbor, Michigan, USA) was used to edit the DNA strands. All sequences have been deposited in GenBank, accession numbers AY773482-AY773670 and AY500638-AY500655, and are listed in Table 1.
Data analysis
Alignment of DNA sequences was initially performed with ClustalX (Thompson et al., 2001
), with subsequent manual adjustment by eye in MacClade v. 4.0 (Sinauer, Sunderland, Massachusetts, USA). Variable positions in the data matrix were double checked against the original chromatogram files to make sure that all base calls were true at all variable positions. In all cases, alignment of potentially informative positions was unambiguous. Indels were coded as binary characters except in the case of a single poly-A/ T run, which was omitted from the data set because it may be a PCR artifact and not reflective of the phylogenetic history of the group.
Analysis of phylogenetic relationships was conducted using the optimality criterion of maximum parsimony. Searches for most-parsimonious trees were executed in PAUP* version 4.0b10 (Swofford, 2002
) by a heuristic search with tree-bisection-reconnection (TBR) branch swapping and 1000 random sequence addition replicates with the "collapse zero-length branches" option in effect. Bootstrap support (Felsenstein, 1985
) was estimated with 1000 replications of heuristic search and simple taxon addition with the constraint of 10 000 000 rearrangements per replicate. Both the consistency and retention indices (CI and RI, respectively) were used to assess the amount of homoplasy present in the data set.
Relationships among the chloroplast haplotypes were also inferred using the software TCS v. 1.13 (Clement et al., 2000
), which implements a statistical parsimony approach to estimating gene genealogies. TCS was performed as an alternative means of analysis because it can infer ancestral or intermediate haplotypes (as opposed to assuming that these haplotypes are extinct). For the TCS analysis, we used the same sequence alignment that was used in the PAUP* analysis described.
Geographic distribution of haplotypes
To assess the distribution of haplotypes within each of the putative taxa of sect. Prunocerasus, distribution maps were created using range information from the U.S. Department of Agriculture national PLANTS database (USDA, 2002
) and from Little (1971
, 1976
, 1977
), Gleason and Cronquist (1991)
, and Correll and Johnston (1970)
. Haplotypes found within each of the taxa were then mapped onto the distribution maps. Additional maps were created irrespective of taxon to show overall haplotype distribution.
RESULTS
rpL16 intron analysis and inference of chloro-haplotypes
The aligned data set consisted of 797 bp from the 3' end of the 1-kb rpL16 intron. Within the aligned matrix were 23 parsimony informative characters (including multibase-pair indels coded as binary characters) and 10 variable but parsimony uninformative characters. In the maximum parsimony analysis, a heuristic search found three equally parsimonious trees of 34 steps with high consistency and retention indices, 0.97 and 0.99, respectively. Bootstrap values are shown above the branches in Fig. 2a–c. The topology of the genealogy generated with TCS (Fig. 3) is consistent with the phylogeny generated with PAUP* (Fig. 2).
|
|
Phylogenetic results and haplotypic distribution
Because 207 accessions could not be placed on the same figure, the phylogeny of sect. Prunocerasus is shown in the strict consensus tree in Fig. 2a–c. Each of the three figures details one of the three primary clades. Within sect. Prunocerasus, the backbone of the tree shows an initial split in the section between the northwestern species, P. subcordata, and the remaining species. A single accession of P. umbellata var. umbellata was placed in a grade between P. subcordata and P. texana (see also the gene genealogy in Fig. 3). This haplotype is unique and therefore interesting; despite other sampled accessions being from the same general geographic location, this haplotype was observed only once. Four accessions of P. texana, currently classified in subg. Amygdalus (Wight, 1913
), were positioned with strong support as sister to the remaining taxa (inside of P. subcordata and the peculiar P. umbellata var. umbellata accession). The 199 remaining accessions segregated into three clades (A = American, B = Beach, and C = Chickasaw) that are each strongly supported. These three primary clades directly correspond to those found in Shaw and Small (2004)
, and the exemplars used in that study are denoted in Fig. 2a–c with an asterisk. Within each of the three primary clades, relationships are weakly resolved because of the lack of sequence divergence. However, the three figures show that many of the taxa of sect. Prunocerasus are para- or polyphyletic. Representatives of most taxa were found in two or three of the primary clades (compare Fig. 2a–c, or see Table 1). In fact, of those taxa of the three primary clades, P. angustifolia and P. alleghaniensis var. davisii are the only ones where multiple accessions all resolve in the same clade.
The geographical position of sect. Prunocerasus accessions and their observed haplotypes were mapped and are shown in Fig. 4a–k. The individual taxon maps show that most taxa in the section contain more than one cpDNA haplotype. Only two taxa, P. subcordata and P. texana, were found to possess unique haplotypes that were not found in any other taxon (Fig. 4a), excluding the one peculiar Umbellata (pU) haplotype found in one accession of P. umbellata.
|
Phylogenetic relationships within sect. Prunocerasus
In a previous study (Shaw and Small, 2004
), we reported that the majority of sect. Prunocerasus species resolve in three primary clades inside of P. subcordata (Fig. 1). Because the relationships revealed in that study were incongruous with classical morphological assemblages (Waugh, 1899
, 1901
; Wight, 1915
; Rehder, 1940
), we were motivated to survey several accessions of each of the taxa to test taxon monophyly. Here we report the organellar para- and polyphyly of several of the taxa of sect. Prunocerasus, the North American plums.
Prunus subcordata is unique among the species of sect. Prunocerasus because it is the only one with a western North American distribution. All of the other taxa in the section have ranges east of the Rocky Mountains. The sampled accessions of P. subcordata have identical rpL16 sequences, the S haplotype (Fig. 4a). This haplotype was supported by three characters and was not observed in any other taxon. Four accessions of Prunus texana, a species endemic to Texas with pubescent fruits, that is currently classified as a peach in subg. Amygdalus (Wight, 1913
), had identical rpL16 sequences, haplotype T (Fig. 4a). Two synapomorphic characters support the monophyly of P. texana, and this clade is sister to the remaining species, inside of P. subcordata and a single P. umbellata var. umbellata accession (haplotype pU). More work needs to be done to determine whether P. texana is a pubescent-fruited plum or contains a plum chloroplast in peach "clothing."
Prunus geniculata is a federally endangered species restricted to four counties in the sand–pine scrub habitat of the Lake Wales Ridge region of central Florida—an ancient dune system habitat known to harbor many endemic species in addition to P. geniculata (Christman and Judd, 1990
; Dobson et al., 1997
). This species was only sampled once because of its protected status and the rarity of herbarium material. The single accession contained the B haplotype (Fig. 4a), but there is an autapomorphy (possible synapomorphy?) in P. geniculata just beyond where the data set had to be trimmed. This is worth mentioning because only two different B haplotypes (B and B1) were identified, compared to seven different C and 10 different A haplotypes. This possible additional B haplotype is also significant because it may prove useful in our future work concerning gene flow and hybridization of this federally protected species. Another taxon that was only sampled once is P. maritima var. gravesii (Fig. 4a). This is a particularly interesting taxon because since its discovery in 1894 it has only been known as a single clonal individual near Groton, Connecticut, USA (Anderson, 1980
). Prunus maritima var. gravesii is currently presumed extirpated from the wild, and the single accession of this taxon contained the B haplotype, a haplotype shared by several (but not all) accessions of P. maritima var. maritima. Aside from P. geniculata and P. maritima var. gravesii, only P. angustifolia and P. alleghaniensis var. davisii were invariant and both contained only the C haplotype (Fig. 4b, d).
The remaining 172 accessions of 12 taxa are scattered among the three primary clades (Fig. 2a–c). Within each of these clades, relationships are weakly resolved because of the lack of sequence divergence. Even still, this study clearly shows that chloroplast types are shared among most of the North American plum taxa of sect. Prunocerasus. Examination of the phylogeny (Fig. 2–2c) and the haplotype maps (Fig. 4– 4k) for many of the taxa reveals that most taxa contain more than one of the three primary haplotypes. All three primary haplotypes were found in P. gracilis (Fig. 4e). The three primary haplotypes plus one basal haplotype (pU) were found in P. umbellata (including P. u. var. injucunda) (Fig. 4k). Haplotypes A and C were found in P. alleghaniensis var. alleghaniensis (Fig. 4b), P. americana var. americana and P. americana var. lanata (Fig. 4c), P. hortulana (Fig. 4f), P. mexicana (Fig. 4g), P. munsoniana (Fig. 4h), P. nigra (Fig. 4i), and P. rivularis (Fig. 4j). Haplotypes B and C were found in P. maritima var. maritima (Fig. 4a).
The findings of this study prompt speculation regarding the distinctiveness of several of the taxa in the section. Many workers have commented on the inability to reliably morphologically distinguish many of the species. For example, Duncan and Duncan (1988
, p. 303) wrote that P. umbellata "varies considerably and has been divided into varieties by others, based mainly on hairiness." In another example in which taxa have been distinguished based mainly on pubescence, Robertson (1974)
noted, "P. americana and P. mexicana [incl. P. americana var. lanata] need to be studied in detail to ascertain their distinctiveness and distribution." Because of the level of para- and polyphyly observed in many of the taxa in sect. Prunocerasus and the lack of distinctive morphological characters in the section, we feel that taxonomic "lumping" may be in order. We are currently analyzing a parallel nuclear DNA data set to further address this question.
Within several of the 12 taxa that have more than one haplotype, there is a haplotype that is found with a higher frequency (Table 1). For example, 35 accessions of P. americana var. americana contain the A haplotype (resolving in the American clade), while six accessions have the C haplotype and resolve in the Chickasaw clade. If we generate a phylogeny based on the most commonly found haplotype in each taxon, the phylogeny more closely matches classical morphological assemblages than the one illustrated in Fig. 3 of Shaw and Small (2004)
(compare Fig. 5 to Fig. 1, which is modified from Shaw and Small [2004]
). Of the 12 taxa with more than one haplotype, two (P. nigra and P. hortulana) had haplotypic frequencies close to 50% and are therefore shown in Fig. 5 with dashed lines because they could have been positioned in either the American or Chickasaw clades. This is worth mentioning in the context of their morphology. Prunus nigra shares morphological characters with some members of the American clade, P. americana var. americana and P. mexicana—all three are trees, have relatively large flowers, and similar leaf shape. These three species were allied by earlier workers (Waugh, 1899
, 1901
; Wight, 1915
). But P. nigra also has glandular leaf teeth and calyx lobes, characteristics shared by P. angustifolia and P. munsoniana of the Chickasaw clade. Because P. nigra shares morphological characters with the dominant members of both the American and the Chickasaw clades, it is not surprising that P. nigra contains the American and Chickasaw chloroplast types in equal frequency.
|
, 1901
) thought that this species was the center taxon of a hybrid swarm, and Steyermark (1963)
noted that some individuals appear to be hybrids between it and P. mexicana.
In our earlier study, P. maritima var. maritima and P. maritima var. gravesii were shown to be sister to P. geniculata in the Beach clade (Fig. 1). These results were unexpected because P. maritima and P. geniculata are morphologically very different. Prunus geniculata is a diminutive species unlike any other in the section and, in our opinion, is morphologically closer to P. angustifolia than to any other taxon in the section. Furthermore, P. maritima var. maritima is morphologically closer to P. umbellata var. umbellata, P. alleghaniensis var. alleghaniensis, and P. gracilis and was allied to these species by earlier workers (Waugh, 1899
, 1901
; Wight, 1915
). In this study, we sampled 14 accessions of P. maritima var. maritima, nine of which contained the C haplotype and five the B haplotype (Fig. 4a). Interestingly, while most P. maritima var. maritima accessions contain the C haplotype and morphologically this species more closely resembles the species of the Chickasaw clade, the variety P. maritima var. gravesii (known as a single clonal individual since 1894) contains the B haplotype and is geographically embedded in the center of the P. maritima var. maritima B haplotype region.
Phylogeographic inference
A strength of the phylogeographic approach stems from an increased sampling strategy, as compared to most phylogenetic studies. This allows evolutionary questions to be addressed in more detail and to be placed in a geographical context. The phylogeographic approach will doubtless further our knowledge of closely related species, like the North American plums, for which factors like hybridization, introgression, or lineage sorting can mask "true" evolutionary relationships in studies lacking thorough sampling. This study highlights the need for a new tradition of "congeneric phylogeography" (Funk and Omland, 2003
) in which phylogenetics and phylogeography are joined, yielding a more population-level sampling scheme that will improve the resolution of evolutionary relationships among closely related biological entities.
Prunus is a morphologically complicated genus of the northern hemisphere with about 200 species of small trees or shrubs (Rehder, 1940
; Robertson, 1974
). Although three broad phylogenetic studies have focused on Prunus (Bortiri et al., 2001
; Lee and Wen, 2001
; Shaw and Small, 2004
), none have adequately sampled outside Prunus to establish a root and pinpoint the center of origin for the genus. Furthermore, in studies spanning Rosaceae, Prunus has been allied with a clade that includes Exochorda Lindl., Oemleria Reichb., and Prinsepia Royle (other genera of Takhtajan's [1997]
Amygdaloideae) (Morgan et al., 1994
), within an array of "spiraeoid" lineages, or as sister to Maloideae s.l. (Potter, 2003
). Without knowledge of the sister taxon to Prunus, nor even a clear sister species to sect. Prunocerasus, the migration of sect. Prunocerasus species to and throughout North America cannot be unequivocally established—leaving us with the question of whether sect. Prunocerasus arrived in North America across the Bering or North Atlantic land bridge.
With the exceptions P. subcordata and P. texana, multiple accessions of the remaining taxa are scattered among three different clades whose relationships to one another are unresolved despite sequencing seven noncoding cpDNA regions (4375 aligned nucleotide positions) (Shaw and Small, 2004
). This tritomy may be "real" in the sense that all three primary haplotypes appear to be very close in age. Based on the philosophy that refugial plant populations have higher levels of genetic diversity (Demesure et al., 1996
; Dumolin-Lapegue et al., 1997
; King and Ferris, 1998
), and because the three primary haplotypes were only found intermingled in Texas (along with P. texana and the pU haplotype), this area appears to have been a refugium or point of origin for many sect. Prunocerasus members.
Over the last 2 million years there have been 16–20 cycles of glaciation when plants species underwent southern migrations and northern reintroductions (Hays et al., 1969
; Davis, 1983
). Cyclical glaciation could account for multiple bouts of range expansion and divergence followed by range contraction and hybridization. This may explain both the morphological complexity and the randomness of the taxonomic distribution of the three primary haplotypes.
Even though individual taxon haplotype maps reveal no obvious pattern, a clear pattern emerges if all of the haplotypes are mapped together regardless of the taxa in which they were observed (Fig. 6a, b). The A (American) haplotype was found in 11 taxa (P. alleghaniensis var. alleghaniensis, P. americana var. americana, P. americana var. lanata, P. gracilis, P. hortulana, P. mexicana, P. munsoniana, P. nigra, P. rivularis, P. umbellata var. umbellata, and P. umbellata var. injucunda) (Fig. 4b, c, e–k). This haplotype is mostly confined to the interior of the USA. The B (Beach) haplotype was only found in eastern central Texas, the panhandle of Florida, and the easternmost portions of Cape Cod, Massachusetts, and Long Island, New York in the USA (Fig. 6a). This haplotype was found in five different taxa: P. geniculata, P. gracilis, P. maritima var. maritima, P. maritima var. gravesii, and P. umbellata var. umbellata (Fig. 4a, e, k). The distribution of the B haplotype is engaging because it was only found in restricted, isolated areas of the Atlantic and Gulf Coastal Plain floristic province (Fig. 6a).
|
wrote that there were approximately 60 species that fit this distribution pattern and that most are of conservative habits and habitats (not weedy and therefore not likely to have spread there because of human-mediated introductions). While explanation for this disjunct distribution pattern is still unclear, Reznicek (1994)
proposed that many species migrated into the Great Lakes Region through dispersal jumps between areas of suitable habitat created along major postglacial drainage channels (see Reznicek, 1994
, for a thorough discussion of this hypothesis as well as discussion of other Coastal Plain taxa that are disjunct to the Great Lakes). It is very interesting to point out that the distribution of the C haplotype, although found in 14 taxa in sect. Prunocerasus, is nearly identical to the distribution of Rhexia virginica L. (see Fig. 4 in Reznicek, 1994
). This distribution was also recently observed in Trillium grandiflorum (Michaux) Salisbury cpDNA (Griffin and Barrett, 2004
). Furthermore, according to Jackson and Singer (1997)
, species whose ranges are disjunct between the Coastal Plain and the Great Lakes region are typically rare or absent in the North American interior, a pattern that is also congruent with the distribution of the C haplotype in sect. Prunocerasus. Plum genetic structure is not only related to the occurrence of refugial zones and particular biogeographic conditions of North America, but also to human influence in terms of dispersal or disturbance. In the early stages of this study, we were concerned that the inconsistency of haplotypes observed (while we were accumulating this data set) may have been due to human mediated dispersal and hybridization. However, our impression is that sampling was thorough enough to place confidence in the larger patterns observed in the data. Because plums have undoubtedly been spread by humans, we are less inclined to comment on the less frequent haplotypes and their distributions.
Broader ramifications for phylogenetics
Earlier workers cautioned that, in phylogenetic studies of closely related species, using only a single accession of each species could be misleading (Whittemore and Schaal, 1991
; Matos and Schaal, 2000
; Funk and Omland, 2003
), and they recommended more thorough sampling to circumvent potential problems. This study exemplifies that reality. In closely related groups of species, sampling multiple individuals from each of the representative taxa is the only means of accurately assessing phylogenetic relationships.
In Prunus sect. Prunocerasus, 14 of 18 taxa contained more than one chloro-haplotype. In hindsight, the possibility existed that in our earlier study we could have chosen different combinations of exemplars, each of which could have resulted in a different phylogeny. In fact, because nine taxa contain both the A and C haplotypes; one taxon contains the A and B haplotypes; one taxon contains the A, B, and C haplotypes; and one taxon contains the A, B, C, and pU haplotypes, the possibility existed that we could have generated any one of 12 288 different "species" trees depending on the accessions chosen.
The high degree of chloroplast sharing that was observed in this study may be more common than currently recognized. Chat et al. (2004)
found that five of eight species of Kiwi (Actinidia Lindley, Actinidiaceae) were polyphyletic in their cpDNA. In another example, McKinnon et al. (1999)
showed that five of seven closely related Eucalyptus LHérit. (Myrtaceae) species were polyphyletic in their cpDNA, and Jackson et al. (1999)
showed that cpDNA variation did not conform to subspecies boundaries but was instead geographically distributed in Eucalyptus globulus Labill. These results highlight the need for increased sampling in phylogenetic studies of closely related taxa.
Conclusions
Three main conclusions can be drawn from this study. The first concerns the para- and polyphyly of chloro-haplotypes observed in 12 of 17 taxa in Prunus subg. Prunus sect. Prunocerasus. Although lineage sorting from a polymorphic ancestor cannot be ruled out as the cause for the observed pattern, we feel that a hypothesis of chloroplast sharing through past and current hybridization is a more likely explanation because most of the species in the section are known to hybridize. Secondly, four accessions of the Texas peachbush, P. texana, all contain the same haplotype (T), which is strongly supported as being embedded within sect. Prunocerasus species. Our future nDNA work may help to further illuminate the taxonomic position of this species. Third, and perhaps the most important general conclusion, this study highlights the need for more thorough sampling in phylogenetic investigations of closely related taxa. Most phylogenetic studies utilizing cpDNA data commonly include one or a few individual(s) per taxon; therefore, in closely related groups, the true phylogeny might not be reflected but rather the obscuring effects of hybridization and introgression or lineage sorting from ancestral polymorphisms.
List of taxa used in this investigation, source and voucher numbers, and GenBank accession numbers.
