| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics and Phytogeography |
University of California—Davis, Department of Pomology, One Shields Avenue, Davis, California 95616 USA
Received for publication January 15, 2002. Accepted for publication May 16, 2002.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: alcohol dehydrogenase • likelihood • molecular clock • parsimony • Prunus • s6pdh
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
) of shrubs and trees, is an important component of Northern Hemisphere forest (Fedorov et al., 1941
; Elias, 1980
; Yü et al., 1986
) and desert (Mason, 1913
; Browicz and Zohary, 1996
) communities. A significant number of species are also found in the tropics worldwide (Kalkmann, 1965
; Brako and Zarucchi, 1993
). Many species are cultivated worldwide for their fruits, such as cherries (P. cerasus and P. avium), apricot (P. armeniaca), almond (P. dulcis), peach (P. persica), and plums (several species of subgenus Prunus) (Moore and Ballington, 1990
). Prunus serotina, the North American black cherry, is valued for its timber. Several species are ornamentals (Krüssmann, 1986
), in particular flowering cherries of subgenus Cerasus (Kuitert, 1999
). Wild North American plums (subgenus Prunus, section Prunocerasus) were an important source of food for Native Americans (Wight, 1915
) and are used locally to make tarts and jams. Finally, species of Prunus play an important ecological role by providing food for wildlife (Elias, 1980
; M. Beck, University of Nevada, personal communication). In addition to their economic, cultural, and ecological importance, species of Prunus display considerable morphological variation in terms of types of inflorescences and fruits and an interesting geographic distribution. The evolution of some characters most likely arose as adaptations to special habitats, such as the presence of dry fruits in desert species.
Prunus is the largest genus in the subfamily Amygdaloideae, which also includes Exochorda, Maddenia, Oemleria, Prinsepia, and Pygeum in the most recent taxonomic treatment of Rosaceae by Takhtajan (1997)
. This classification is supported by evidence from wood anatomy (Zhang, 1992
) and chromosome number (Goldblatt, 1976
). Molecular data have shown that Maddenia may be nested within Prunus (Lee and Wen, 2001
) and Oemleria, Prinsepia, and Exochorda are monophyletic (Morgan, Soltis, and Robertson, 1994
; Potter et al., 2002
). However, these two clades are not strongly supported as sister to each other. A phylogeny of Rosaceae using rbcL weakly supported Amygdaloideae sensu Takhtajan (Morgan, Soltis, and Robertson, 1994
), while other molecular data show no relationship between Prunus and the clade including Oemleria, Prinsepia, and Exochorda (Potter et al., 2002
). Pygeum has not been included in the mentioned phylogenetic studies; thus, its relationships to Prunus remain to be tested by molecular cladistics.
Historically, Prunus has been classified into five subgenera: Amygdalus, Cerasus, Laurocerasus, Padus, and Prunus. Other authors have chosen to treat those subgenera as separate genera (reviewed by McVaugh, 1951
). In this paper we follow Rehder's (1940)
classification of the genus into five subgenera with the addition of subgenus Emplectocladus (Torr.) Sargent (http://ajbsupp.botany.org/v89/).
Prunus has been the subject of two recent phylogenetic studies (Bortiri et al., 2001
; Lee and Wen, 2001
). In a maximum parsimony analysis of combined internal transcribed spacer (ITS) and trnL-trnF data, we found that most species fall into two groups. Subgenera Cerasus, Padus, and Laurocerasus together are supported as monophyletic and a second group consisting of subgenera Amygdalus, Prunus, Emplectocladus, and the section Microcerasus (subgenus Cerasus) was found in the strict consensus of most parsimonious trees but lacked significant bootstrap support (Bortiri et al., 2001
). Lee and Wen's (2001)
results, based on ITS data, were similar in that they also found the same two major groups. When the two data sets, ITS and trnL-trnF, were analyzed separately, however, they supported different scenarios for the relationships of subgenus Cerasus (Bortiri et al., 2001
). According to the chloroplast trnL-trnF spacer, Cerasus is placed in a group with subgenera Amygdalus, Prunus, Emplectocladus, and the sections Penarmeniaca and Microcerasus. The nuclear ITS region, on the other hand, supported a clade composed of subgenera Cerasus, Padus, and Laurocerasus (Bortiri et al., 2001
).
One of the shortcomings of those previous studies was the lack of support for some deep nodes, in particular those in the group that included the subgenera Emplectocladus, Amygdalus, and Prunus. This outcome was attributed to high homoplasy levels in ITS, perhaps caused by rapid speciation and early hybridization (Bortiri et al., 2001
).
In order to test the hypotheses of evolution and relationships in Prunus set forth by our previous work with ITS and trnL-trnF, we started studies on the nuclear gene s6pdh, which encodes NADP+-dependent sorbitol-6-phosphate dehydrogenase (S6PDH [Enzyme Commission 1.1.1.200]) (Yamaki and Ishikawa, 1986
). In the Rosaceae, with the exception of Rosoideae sensu stricto (Wallaart, 1980
; Morgan, Soltis, and Robertson, 1994
), the most important products of photosynthesis found in the phloem are sucrose and the sugar alcohol sorbitol (Webb and Burley, 1962
). Sorbitol-6-phosphate is produced as a result of the reduction of glucose-6-phosphate by S6PDH, which uses NADP+ as cofactor. S6PDH has been purified from leaves of loquat (Eriobotrya japonica) (Hirai, 1981
) and apple seedlings (Malus domestica) (Kanayama and Yamaki, 1993
). Tao, Uratsu, and Dandekar (1995)
transformed tobacco (Nicotiana tabacum) plants, which do not produce sorbitol, with the Malus domestica s6pdh cDNA sequence and showed that leaves from transformed plants accumulate sorbitol. In addition, the concentration of sorbitol correlated positively with the expression of S6PDH. The genomic sequence of Malus domestica s6pdh consists of six exons totaling 3.4 kilobases (kb), with a TC simple repeat in intron 3 (Bains et al., 1998
). Studies on polymorphisms of the TC microsatellite in intron 3 of M. domestica indicate the presence of two copies of s6pdh in that species and one in species of Prunus (S. Arulsekar, University of California, Davis, personal communication). Malus belongs to subfamily Maloideae, which has chromosome numbers x = 15, 17; whereas the rest of the family have chromosome numbers x = 7, 8, 9. It has been postulated that subfamily Maloideae is an ancient polyploid (e.g., Sax, 1933
; Gladkova, 1972
). For that reason we expected s6pdh to be a single-copy gene in diploid taxa of Rosaceae other than Maloideae.
In this paper we present a phylogenetic study of Prunus using parsimony and maximum likelihood methods as determined by sequence data from the nuclear gene s6pdh. We compare the results obtained using the two methods of phylogenetic reconstruction. We also compare rates of evolution of s6pdh with those of ITS and the trnL-trnF spacer and test whether rates of evolution in s6pdh are uniform across all lineages of Prunus. Finally, we combine data from all three markers to reconstruct the phylogeny of the genus using parsimony and maximum likelihood methods. The outcomes produced by both methods are discussed in relation to work carried out in previous studies. This paper aims to understand the species phylogeny of Prunus. Data from s6pdh will complement evidence from the ITS and trnL-trnF regions used previously (Bortiri et al., 2001
). The main objectives are (a) to examine the phylogenetic relationships of subgenus Cerasus, (b) to test the monophyly of the group composed of subgenera Prunus, Amygdalus, Emplectocladus, and section Microcerasus, and (c) to generate data that will support a robust phylogeny of Prunus.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). This sample includes all subgenera and most sections of Prunus (Mason, 1913
; Rehder, 1940
). DNA extractions were performed according to the cetyltrimethylammonium bromide (CTAB) method (Doyle and Doyle, 1987
) except the extractions of Oemleria cerasiformis (Hook. & Arn.) Landon and Exochorda racemosa (Lindl.) Rehder, which followed a different protocol (Dellaporta, Wood, and Hicks, 1983
). All collected vouchers are deposited at the University of California-Davis Herbarium (DAV). Information on the origin of the plant material, as well as GenBank accession numbers, are stored at the American Journal of Botany supplementary data web site (http://ajbsupp.botany.org/v89/).
Amplification and sequencing
We obtained s6pdh sequences from two different sources: direct polymerase chain reaction (PCR) sequencing and sequencing of cloned PCR products. In addition, we repeated some of the PCR and cloning reactions in order to detect amplification of two or more paralogous genes and to make sure that artifacts during PCR or cloning were not masking or altering the outcome substantially.
The primers used for amplification are shown in Table 1 and their relative positions on s6pdh are in Fig. 1. All primers, with the exception of s6pdh-p and 77L25, were based on the cDNA sequence of Malus domestica (Kanayama et al., 1992
; GenBank accession number D11080). We designed s6pdh-p to be used in combination with s6pdh-k on species of Prunus. A second primer, designated 77L25, was created to use in combination with s6pdh-a for amplification of intron 1 of Prunus species (Fig. 1). The s6pdh-p and 77L25 primers were based on a cloned sequence of s6pdh from P. dulcis that had been obtained with s6pdh-k and s6pdh-j. All primers were designed with the help of Oligos® 5.0 (National Biosciences, Plymouth, Minnesota, USA) and purchased from Genosys Biotechnologies (The Woodlands, Texas, USA). Sequences from Pyrus caucasica Fed., Oemleria, Amelanchier alnifolia (Nutt.), Gillenia stipulata (Torr.) Sarg., Kageneckia oblonga Ruiz & Pav., Vauquelinia californica (Torr.) Sarg., and Sorbus were produced by PCR amplification with primers s6pdh-k and s6pdh-j. Sorbaria sorbifolia (L.) A. Braun and Spiraea cantoniensis Lour. were amplified with the primer combinations s6pdh-k/s6pdh-h and s6pdh-c/s6pdh-j, which yield overlapping PCR products (Fig. 1). We obtained partial sequences from Exochorda racemosa, Holodiscus microphyllus Rydb., and Chamaebatiaria millefolium (Torr.) Maxim., by amplification with primers s6pdh-e and s6pdh-j. For the combined analyses of s6pdh, ITS, and the trnL-trnF spacer we used previously published sequences (Bortiri et al., 2001
). We also obtained new ITS and trnL-trnF sequences from Prunus caroliniana, P. microcarpa, and P. lusitanica using protocols described previously (Bortiri et al., 2001
).
|
|
For direct sequencing we purified the PCR products by separation in 0.8% agarose gels, excised the band of appropriate size, and purified the DNA with a QIAquick gel extraction kit (QIAGEN, Valencia, California, USA) according to the manufacturer's specifications. Direct sequencing of PCR fragments was performed using primers s6pdh-k, s6pdh-p, and s6pdh-c. We used primers s6pdh-a and 77L25 for sequencing of intron 1.
For the reactions that were later to be cloned we loaded a 15 µL aliquot of the finished PCR reaction in 0.8% agarose gels for electrophoresis to identify successful PCR amplifications. We cloned the PCR products from the remaining volume of successful reactions using a TOPO TA Cloning® kit (Invitrogen Corporation, Carlsbad, California, USA) according to manufacturer's specifications except that the reaction volumes were halved. We picked 4–10 colonies from each cloning reaction and grew them overnight in 10 mL of Luria-Bertani medium with ampicillin. We extracted the plasmid DNA using a QIAprep® Spin Miniprep kit (QIAGEN) and confirmed the presence of inserts by restricting approximately 2 µg of the mini-prepped plasmid DNA with 1.2 units of EcoRI. We selected clones that had inserts and sequenced them in both directions with primers T7 and T3. All sequencing was performed at the Plant Genetics Facility (UC Davis), which uses an ABI Prism sequencer (Applied Biosystems).
In order to search for proteins related to S6PDH we conducted a BLAST search in GenBank using the deduced amino acid sequence of Malus domestica (GenBank accession number GBAN-P28475).
Southern blots
We performed a Southern blot analysis to estimate the number of genes with sequence similarity to s6pdh that are present in the genome of several Rosaceae. DNA was extracted from 4 g of fresh leaves using the method of Doyle and Doyle (1987)
. We restricted 5 µg aliquots of DNA with EcoRI, EcoRV, DraI, and HindIII. The restricted DNA fragments were separated by electrophoresis in 0.8% agarose gels and transferred to membranes according to standard procedures (Sambrook, Fritsch, and Maniatis, 1989
). The probe was prepared by PCR amplification of a Malus domestica cDNA clone (Kanayama et al., 1992
) using primers s6pdh-a and s6pdh-j, which yield a fragment of 0.93 kb, and labeled by primer extension using 32P dNTPs (Sambrook, Fritsch, and Maniatis, 1989
). The membranes containing the restricted fragments were hybridized overnight at 60°C and washed at 60°C in 2x SSPE/0.1% SDS buffer three times. The membranes were then exposed onto an X-ray film for 1 h and developed.
Sequence alignments
The sequences were edited in Sequencher version 4.1.2 (Gene Codes Corporation, Ann Arbor, Michigan, USA) and two data sets were prepared. The first data set included only the coding sequences of s6pdh. To prepare this data set we used the View Translation command in Sequencher to find the exon/intron boundaries and edited out the intron sequences using the deduced amino acid sequence of Malus domestica as a guide. These DNA coding sequences were imported into Se-Al version 2.0a3 (Rambaut, 2001
) and aligned as amino acid sequences using the universal genetic code. The aligned nucleotide sequences were then exported in a NEXUS format. The second data set was prepared by manually aligning the introns from all Prunus taxa using the first data set as a scaffold. Intron sequences from outgroup taxa were excluded because they were too divergent and were treated as missing data. Insertions and deletions (indels) introduced by the alignment procedure were treated as missing values.
We used ClustalX (Thompson et al., 1997
) to align the ITS, trnL-trnF, and the S6PDH (amino acids) data sets. The latter consisted of nine GenBank protein accessions found from the BLAST search, plus the deduced amino acid sequence of S6PDH from Malus domestica, P. ilicifolia, and P. dulcis. Six indels, two from the ITS data set and four from the trnL-trnF alignment, were scored as separate characters. We took a conservative approach to scoring indels as additional characters, considering only those that were more than two sites long with unequivocally aligned boundaries, a measure taken by previous authors (Wojciechowski, Sanderson, and Hu, 1999
).
Phylogenetic analyses
We used PAUP* 4.0b8 (Swofford, 2001
) to find most parsimonious trees (MPT) and trees of maximum likelihood (MLT). Unless otherwise indicated, parsimony searches used a heuristic strategy with 100 replicates of random taxon addition, tree bisection-reconnection (TBR) and no constraints on the maximum numbers of trees to be saved each step. We conducted two different heuristic searches for MLT. In the first place, we used the "as is" addition strategy instead of random additions. For the second search we used all trees saved from a parsimony analysis as starting trees. Measures of homoplasy are given as consistency indices excluding uninformative characters (CI*) and retention index (RI). We estimated the statistical support for each node by performing 1000 bootstrap repetitions (Felsenstein, 1985
) with simple taxon addition. For the bootstrap analysis of s6pdh we saved up to 1000 trees each time. We obtained decay values (Bremer, 1988
) for each node by using AutoDecay version 4.0.2 (Eriksson, 1998
) and TreeView version 1.6.0 (Page, 1996
). For the parsimony analysis of the S6PDHs and related dehydrogenases data set we used a Branch and Bound search strategy and statistical support for each node was calculated by performing 1000 Branch and Bound bootstrap repetitions. We calculated P values for incongruence length difference (ILD) tests (Farris et al., 1994a
, b
) as a measure of incongruence among the three data partitions. The tests were implemented in PAUP* (Swofford, 2001
) by running 500 repetitions of the partition homogeneity test, with TBR, and a maximum of 1000 trees saved at each step.
We used Modeltest version 3.04 (Posada and Crandall, 1998
, 2001
) to select evolution models for the maximum likelihood analyses. Modeltest calculates the operator 
= 2(lnL1 – lnL0) (Goldman, 1993
) where L0, the null hypothesis, is a simpler model than L1. The test evaluates several models of evolution and selects the one that fails to be rejected by an alternative and more complex model. Simpler models, which have fewer parameters, are preferred to more complex ones when there is no significant difference in how well they fit the data because the use of additional parameters introduces a potential source of error (Nei and Kumar, 2000
). The models of molecular evolution used in this study are given in the Appendix.
To test the hypothesis that rates of evolution of s6pdh are constant in all the lineages of Prunus (molecular clock evolution) we used the same operator ( = 2[lnL1 – lnL0]), where L0 is the likelihood when equal rates are forced on all branches and L1 is the score without such constraint. The outcome was evaluated in a 
2 distribution with N – 2 degrees of freedom, where N = number of taxa, and significance of 0.01 (Felsenstein, 1981
; Yang, Goldman, and Friday, 1995
; Huelsenbeck and Rannala, 1997
). The null hypothesis, i.e., evolution in a clockwise fashion, will fail to be rejected if is greater than the value 20.01(N – 2).
As an estimation of the rate of evolution we calculated the pairwise divergence among Prunus species for s6pdh, ITS, and the trnL-trnF spacer in PAUP*. We found that different models explain different sets of data significantly better. However, for the purposes of pairwise divergence, we wanted to use a uniform criterion. Thus, we selected a TrN model (Tamura and Nei, 1993
), with a gamma shape parameter = 0.5, to calculate the pairwise distances for all three markers.
We carried out the SH (Shimodaira and Hasegawa, 1999
) likelihood test of topologies for the trees obtained by ML and MP analyses of the combined data sets. The test was performed in PAUP* (Swofford, 2001
) by running 1000 RELL nonparametric bootstrap repetitions using the data that produced the trees being tested. P values are then given for each topology and a particular tree topology will fail to be rejected in favor of the ML tree if its P value is >0.05 (Shimodaira and Hasegawa, 1999
). Likelihood-based tests of alternative phylogenetic hypothesis have been reviewed recently (Goldman, Anderson, and Rodrigo, 2000
).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
|
The PCR amplification with s6pdh-k and s6pdh-p always yielded a single band, as visualized in agarose gels, of approximately 1.3 kb. The s6pdh sequences obtained from Prunus and the other Rosaceae genera studied here have the same organization as in Malus domestica. The sequences varied in length from 1218 base pairs (bp) in Oemleria to 1309 bp in P. maritima and P. salicina, and the gene is organized into 6 exons and 5 introns (Fig. 1). We also confirmed that the TC repeat in intron 3 is present in all species of Prunus.
Southern blotting of s6pdh
The blots corresponding to P. persica, P. dulcis, Spiraea betulifolia Pall., Aruncus dioicus (Walter) Fernald, and Neillia sinensis Oliv. (Fig. 3A) indicate that in most cases one region strongly hybridizes to the probe, represented by a band with strong signal in the cases of Spiraea (lanes 1–4), Prunus persica (5–8), P. dulcis (9–10), Aruncus (11–14), and Neillia (15–18). The lanes corresponding to P. persica have one band with very strong signal (Fig. 3A, lanes 5 and 6) or two bands of approximately equal intensity (lanes 7 and 8). One EcoRV site was confirmed in P. dulcis by sequencing (Fig. 3B) and its presence in P. persica is very likely given that both species have the same EcoRV restriction pattern (Fig. 3A). Prunus persica and P. dulcis have one site each for EcoRI, DraI, and HindIII (Fig. 3B, C), and this is reflected in their restriction patterns with these enzymes. The fragment represented by the band with the strongest signal in lanes 5 and 6 (P. persica) and lane 10 (P. dulcis) (Fig. 3A) is present as two bands of lower molecular mass in lanes 7 and 8 (P. persica) (Fig. 3A) due to a restriction in the middle of s6pdh by DraI and HindIII (Fig. 3C). The presence of a high molecular mass band in lane 5 (P. persica) was unexpected and may be due to a failure to restrict by EcoRI.
|
|
We found divergent sequences in P. caroliniana and P. emarginata. The direct PCR sequence and one of the two sequenced clones of P. caroliniana (clone "A") are most closely related to P. ilicifolia but a second clone (P. caroliniana-2) is nested within a clade composed of P. virginiana, P. laurocerasus, P. lusitanica, and a P. emarginata clone (Fig. 4A). The two divergent P. caroliniana clones differ at 16 out of 251 amino acids. The P. emarginata "direct PCR" sequence was sister to P. takesimensis-P. fruticosa. We repeated sequencing of s6pdh from a different accession of P. emarginata ("994" in Fig. 4A) and found four clones to form a monophyletic group unrelated to any of the other two P. emarginata sequences.
Maximum likelihood analysis
In order to reduce computational time in all analyses of ML we included only one sequence from each species of Prunus. For P. caroliniana and P. emarginata we included one from each group of divergent sequences. We further reduced the number of taxa in this step by excluding Chamabaetiaria, Holodiscus, and Exochorda because their sequences were incomplete (see Materials and methods). After running Modeltest we selected a general time reversible model (Tamura and Nei, 1993
) (see Appendix for parameters and models used for all maximum likelihood analyses). The first search, using the "as is" addition sequence, recovered one MLT (Fig. 4B) (score –Ln = 7720.96), which shows no topological conflict with the strict consensus of MPT. A maximum parsimony analysis of this reduced data set found 98 MPT. When we swapped all 98 MPT using a maximum likelihood criterion we recovered the same tree as in the first ML search (Fig. 4B). There are no conflicts between the MPT and MLT, except that the MLT is more resolved. Prunus persica and P. dulcis, subgenus Amygdalus, are unresolved in a polytomy with the rest of the genus. The rest of the genus is divided into three lineages joined in a polytomy: P. fasciculata (subgenus Emplectocladus), subgenus Prunus (including Microcerasus), and a clade composed of subgenera Cerasus, Padus, and Laurocerasus.
Rates of evolution in Prunus
We calculated the means of pairwise divergences for the coding regions of s6pdh, all sequenced introns of s6pdh, ITS, and trnL-trnF among species of Prunus as estimations of the rates of evolution (Table 2). We did not attempt to calibrate a molecular clock using those markers because the rates depart significantly from a molecular clock (see below). The highest divergences are found in intron 4 of s6pdh, which shows approximately twice as much divergence as the other s6pdh introns and both ITS1 and ITS2. The lowest average divergence occurred in the chloroplast trnL-trnF spacer. Coding regions of s6pdh have a slightly higher rate than trnL-trnF, whereas the introns (with the exception of intron 4) have a rate comparable to ITS.
|
).
|
). In addition, it has been reported that combination of partitions found to be incongruent by the ILD test can still increase phylogenetic accuracy (Joy and Conn, 2001
), especially when the conflict is between nodes that show poor support in the separate analyses (Krzywinski, Wilkerson, and Besansky, 2001
). In some cases combining incongruent data sets is necessary to recover the true relationships (Yoder, Irwin, and Payseur, 2001
). Finally, we advocate analyses of combined partitions in addition to separate analyses on philosophical grounds, since combined analyses maximize the phylogenetic information available and are most consistent with the principle of maximum parsimony (Nixon and Carpenter, 1996
). For the analyses of combined ITS, trnL-trnF, and s6pdh data sets we included four outgroups (Oemleria, Exochorda, Spiraea, and Sorbaria) and 22 species of Prunus. We excluded P. caroliniana-2 and the two P. emarginata sequences (774-B and 994) because we believe they are divergent paralogs (see DISCUSSION). The matrix had 2760 characters, of which 384 characters were parsimony-informative, and 226 were informative among the ingroup (s6pdh: 148, ITS: 60, trnL-trnF: 18). The heuristic search recovered nine MPT of length = 1592, CI* = 0.58, and RI = 0.61 (Fig. 5A). In the strict consensus Prunus is divided into two clades. One composed of subgenera Cerasus, Laurocerasus, and Padus. In the second clade the subgenera Emplectocladus, Prunus (including Microcerasus and Penarmeniaca), and Amygdalus are joined in a polytomy.
|
) model of evolution on the combined data set (Appendix). The MLT has no conflict with the consensus of the nine MPT and is unresolved for P. fasciculata and subgenera Amygdalus and Prunus. The likelihoods of the nine MPT, evaluated under maximum likelihood settings used to obtain the MLT, ranged from 12 058.78 (tree in Fig. 5A) to 12 061.18. The SH test indicated that there is no significant difference in the likelihood between the trees obtained by parsimony and likelihood analyses, with P values ranging between 0.93 and 0.88. |
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
) and Paeonia (Sang, Donoghue, and Zhang, 1997
), PgiC in Clarkia (Gottlieb and Ford, 1996
), GBSSI, or waxy, in the Rosaceae (Evans et al., 2000
), and phytochromes A and C in angiosperms (Mathews and Donoghue, 1999
).
The s6pdh gene belongs to a family of genes that encode NADP+-dependent aldo-keto reductases. Related enzymes are found not only among plants but also in animals and fungi. Mannitol-6-phosphate dehydrogenases (M6PDH) from Apium, Arabidopsis, and Orobanche are the proteins most closely related to S6PDH of Rosaceae. M6PDH catalyzes the reduction of mannose into mannitol, a reaction equivalent to the reduction of glucose into sorbitol by S6PDH. Mannitol is the most common sugar alcohol in angiosperms, while sorbitol is produced by Rosaceae (Zimmerman and Ziegler, 1975
; Loescher, 1987
), with the exception of Rosoideae sensu stricto (Fig. 1 in Potter et al., 2002
).
We do not know whether m6pdh, in addition to s6pdh, is present in the Rosaceae, but based on the information from the Southern blots it is evident that other loci with sequence similarity to s6pdh are found in the genome of several sorbitol-producing Rosaceae. This suggests that s6pdh originated via duplication of m6pdh-like gene in the common ancestor of the sorbitol-producing Rosaceae. This would explain the presence of more than one region with sequence similarity to s6pdh in several Rosaceae (Fig. 3A), whereas Fragaria (a genus of Rosoideae sensu stricto) and Rhamnus (Rhamnaceae) had only one weak band (data not shown). It is important to point out that the presence of an m6pdh-like gene does not mean the plant produces significant amounts of mannitol, i.e., neither Rosoideae sensu stricto (s.s.) nor Rhamnaceae have been found to produce sugar alcohols (Zimmerman and Ziegler, 1975
; Wallart, 1980).
We detected more than one region with sequence similarity to s6pdh in the genome of several Rosaceae by Southern blotting, and we found divergent sequences in P. caroliniana and P. emarginata. However, sequencing of several clones of P. maritima, as well as repeating PCR and cloning on some accessions (P. fasciculata, P. armeniaca, P. andersonii), failed to yield paralogous genes in those taxa. All of the sequences examined had the characteristic TC repeat in intron 3. Previous experiments involving the use of this microsatellite in P. dulcis and P. persica (S. Arulsekar, personal communication) are consistent with s6pdh being a single-copy gene in those species.
The presence of divergent paralogs in P. caroliniana (subgenus Laurocerasus) and P. emarginata (subgenus Cerasus) could be due to gene duplication by polyploidization. Alternatively, the paralogs could have originated by a duplication of the genomic region containing the gene, as in the case of m6pdh of A. thaliana (Lin et al., 1999
). Several species of Prunus are reportedly polyploids (Darlington and Wylie, 1955
), in particular all examined species of subgenera Laurocerasus, Padus, and several crop species. Prunus caroliniana has been reported as tetraploid (Robertson, 1974
). We did not find information on the chromosome numbers of P. emarginata and therefore we cannot rule out the possibility it is a polyploid. The presence of two bands as a result of PCR amplification of intron 1 of P. ilicifolia, a tetraploid species (Parfitt et al., 1990
), may represent two divergent paralogs resulting from a polyploidization event. We cannot confirm this because we sequenced only one clone of the s6pdh PCR product for P. ilicifolia.
In summary, the evidence here presented, from southern blotting and sequencing, indicates that s6pdh is a low-copy gene. More than one region is likely to be amplified from species of subgenera Padus and Laurocerasus, which are polyploids. On the other hand, amplification from diploid species, e.g., subgenera Emplectocladus, Prunus, and Amygdalus, should be straightforward. One exception would be polyploid plum species (P. domestica, P. spinosa), where multiple copies should be found.
The occurrence of an s6pdh duplication event in Prunus can have its advantages. One of the problems with reconstructing a robust phylogeny of Prunus is the genetic distance to other Rosaceae genera, which complicates homology assessment through alignment. The presence of a gene duplication in Prunus could avoid this problem altogether by providing the root of the tree at the duplication event (Gottlieb and Ford, 1996
; Gaut et al., 1999
; Mathews and Donoghue, 1999
; Vieira, Vieira, and Charlesworth, 1999
; Citerne, Möller, and Cronk, 2000
). It is important to note that this will be the case if the duplication is a synapomorphy of the whole genus.
Molecular evolution and utility of s6pdh
Sequences from s6pdh have more than twice the number of informative sites as ITS among species of Prunus (153 vs. 60) but are slightly less homoplastic (CI is 0.6 in s6pdh and 0.57 in ITS). While the coding sequences of s6pdh evolve at a rate similar to those in the trnL-trnF spacer, the introns have a significantly faster rate, comparable to ITS or higher, as in the case of intron 4. Despite rates of evolution similar to or higher than ITS and comparable levels of homoplasy, s6pdh failed to provide significant support for the branching order of early lineages in the genus. ITS and trnL-trnF also show branches of zero or near-zero length in this region of the tree. A similar phenomenon, i.e., short branches corresponding to the early-splitting lineages, has been observed in other groups of angiosperms (Rohwer, 2000
; Fishbein et al., 2001
; Potter et al., 2002
) and attributed to rapid radiation. Nevertheless, s6pdh supported the same general topology as ITS and the combined ITS and trnL-trnF data sets (Lee and Wen, 2001
; Bortiri et al., 2001
), which will be discussed below.
Rates of evolution of s6pdh are not constant across all branches, as demonstrated by the failure of the data to be explained in terms of a molecular clock. This unequal rate is most likely due to the long branches in the clade including subgenera Padus and Laurocerasus (Fig. 5A–B). We have not investigated the source of these uneven evolution rates but differences in generation time should not be a reason because all species of Prunus are perennials (trees and shrubs). We considered the possibility of pseudogenes being the cause of this acceleration because many species of subgenera Padus and Laurocerasus are polyploids. However, we think this is unlikely for several reasons. In the first place, all sequences appear to encode functional proteins, i.e., there are no stop codons anywhere in the coding sequences, and their deduced amino acid sequences are highly conserved with respect to the rest of the genus. Secondly, the two divergent copies of P. caroliniana show a very high rate of evolution compared to the rest of Prunus (Figs. 4B, 5B). We would expect only one, the silenced copy, to have a significantly higher rate as a result of lower selective pressure. Finally, the same pattern, i.e., acceleration of the rate of molecular evolution in Padus and Laurocerasus, is seen in both ITS and trnL-trnF (data not shown, but see Fig. 4 in Bortiri et al., 2001
), although to a lesser extent in the cpDNA.
Phylogenetic relationships
In agreement with other molecular data (Bortiri et al., 2001
; Potter et al., 2002
), s6pdh does not support a close relationship between Prunus and Oemleria-Exochorda. On the other hand, s6pdh found the clades Oemleria-Exochorda and Sorbaria-Chamaebatiaria, which have been strongly supported by other molecular data (Morgan et al., 1994
; Potter et al., 2002
). Similarly, Gillenia is sister to Maloideae according to s6pdh, a relationship also obtained by analysis of cpDNA (Potter et al., 2002
).
There were no conflicts between the topologies obtained by parsimony and maximum likelihood analyses of s6pdh, and the phylogenies of Prunus based on this gene were congruent with those based on ITS and trnL-trnF in most respects. The consensus is that subgenus Cerasus is nested within subgenera Padus and Laurocerasus, all the species of section Microcerasus included in these studies are nested within subgenus Prunus, and subgenera Amygdalus and Emplectocladus are sister to subgenus Prunus. In our previous work on Prunus we found a discrepancy regarding the phylogenetic relationships of subgenus Cerasus (Bortiri et al., 2001
). Phylogeny reconstruction with the cpDNA marker trnL-trnF strongly supported a clade of subgenera Padus and Laurocerasus and placed Cerasus in a group with the rest of Prunus. On the other hand, subgenus Cerasus was nested in a clade with subgenera Padus and Laurocerasus based on data from ITS alone (Lee and Wen, 2001
; Bortiri et al., 2001
). Both maximum parsimony and likelihood analyses of s6pdh agree with the latter scenario and suggest that hybridization may have taken place in the evolution of species of Prunus. However, support from s6pdh was relatively weak because the node (Cerasus, Padus, Laurocerasus), which received 66% in the bootstrap analysis, collapsed in the strict consensus (Fig. 4A).
In agreement with ITS, s6pdh reconstructed a monophyletic subgenus Prunus, which includes not only the sections traditionally assigned to it (Prunus—Old World plums, Prunocerasus—New World plums, and Armeniaca—apricots), but also sections Microcerasus (formerly in subgenus Cerasus) and Penarmeniaca. The latter was assigned to subgenus Prunus by Mason (1913)
but not included in the treatment of Prunus by Rehder (1940)
. Section Microcerasus, represented here by P. besseyi, P. microcarpa, and P. tomentosa, a group of shrubs from the Old and New World, has been previously classified in subgenus Cerasus (Rehder, 1940
; Yü et al., 1986
) or as a separate subgenus, Lithocerasus (Ingram, 1948
). Isozyme data (Mowrey and Werner, 1990
) and hybridization studies (Kataoka, Sugiura, and Tomana, 1988
), in addition to our previous work on ITS and trnL-trnF (Bortiri et al., 2001
), have pointed out its closer relationships to subgenus Prunus rather than to Cerasus. We are presently working on expanding the sampling within this group but the data so far indicates that Microcerasus is an artificial group because it is polyphyletic in all analyses, with the exception of a P. tomentosa-P. bifrons clade found by the ITS and trnL-trnF combined data sets (Bortiri et al., 2001
).
Section Penarmeniaca includes P. andersonii and P. fremontii, two dry-fruited species from southwestern USA and Mexico. Sequences from s6pdh place them as sister to P. maritima and P. mexicana, two plum species from eastern North America (Fig. 4A–B), but ITS (Bortiri et al., 2001
) and the combined data set suggest a sister relationships with P. besseyi (Fig. 5A–B), another North American species. Prunus besseyi was found to be sister to the rest of subgenus Prunus by s6pdh (Fig. 4A–B). Thus, there are some discrepancies between the two nuclear markers (ITS and s6pdh) in the topologies they support for these species. This conflict between data sets could be due to lineage sorting, horizontal transfer, or gene duplication and extinction (Page, 1993
; Maddison, 1995
, 1997
; Wendel and Doyle, 1998
). It is also possible that the conflict is caused by convergent evolution ("long-branch attraction"). However, we think this is less likely because maximum likelihood, which is less prone to obtain the wrong phylogeny under these conditions of unequal rate of evolution (see below), reconstructed the same phylogeny as parsimony.
Prunus fasciculata, another dry-fruited species from southwestern USA and Mexico, has been considered related to the almonds (subgenus Amygdalus) by Rehder (1940)
. However, results from s6pdh, in accordance with our previous study based on ITS and trnL-trnF (Bortiri et al., 2001
), suggest that P. fasciculata diverged early in the evolution of Prunus and is not a member of Amygdalus. Its phylogenetic position as sister to a Prunus-Amygdalus-Penarmeniaca-Microcerasus alliance (Bortiri et al., 2001
) is only weakly supported by the combined data set (Fig. 5A–B). In preliminary analyses of s6pdh, P. fasciculata was found to be sister to Cerasus, Laurocerasus, and Padus. We tested this hypothesis by creating a constrained tree with that topology and found that its likelihood was not significantly lower than the MLT and the nine MPT (P = 0.818). Therefore, the precise phylogenetic position of P. fasciculata is still unresolved.
The monophyly of subgenus Amygdalus (P. dulcis and P. persica, almond and peach, respectively) is supported by ITS + trnL-trnF + s6pdh (Fig. 5A–B), and its position as sister to Prunus is in accordance with previous results (Bortiri et al., 2001
; Lee and Wen, 2001
). Lee and Wen (2001)
found Amygdalus to be paraphyletic, contrary to our findings (Bortiri et al., 2001
). The discrepancy is explained by the inclusion in their study of P. tenella Batsch., a species of section Chamaeamygdalus Spach. We have recently tested this hypothesis by sequencing ITS and trnL-trnF of P. petunnikovii Litv., a species closely related to P. tenella, and obtained the same results (data not shown) as Lee and Wen (2001)
; i.e., Amygdalus, as treated by Browicz and Zohary (1996)
is not monophyletic. Additional sampling would be necessary to test relationships among all sections of subgenus Amygdalus.
Effects of phylogenetic methods, amount of data, and number of taxa
The use of maximum likelihood methods for phylogenetic inference has been advocated as a way to avoid inconsistency due to "long branch attraction" effect (Felsenstein, 1978
; Hendy and Penny, 1989
). Several authors have studied the performance of parsimony, likelihood, and other methods under conditions of unequal rates of evolution (Huelsenbeck and Hillis, 1993
; Siddall, 1998
; Swofford et al., 2001
) and the effect of taxon and data addition to improve parsimony accuracy (Kim, 1996
; Graybeal, 1998
; Poe and Swofford, 1999
).
In order to address the concern of reconstructing a misleading phylogeny using only one method, we decided to use parsimony and maximum likelihood and compare the results. Here we show that, despite the presence of unequal rates of evolution in lineages of Prunus, both methods reconstruct the same phylogeny and the results are very similar to those obtained previously using a larger number of taxa (Bortiri et al., 2001
). Prunus is a very large genus with over 200 species, of which we included only 22; however, we took care to select representatives of all groups of species that were previously shown to be monophyletic (Lee and Wen, 2001
; Bortiri et al., 2001
), thereby reducing the sample size while minimizing the effect on the overall topology. Nevertheless, a large group of species from the tropics is missing here. These species, many of which were formerly classified in genus Pygeum (Kalkman, 1965
), were found to be sister to genus Prunus in a cladistic study of wood anatomy characters (Zhang, 1992
). Thus, this group of species is very important to a better understanding of the natural history of Prunus because of their geographic distribution (Prunus is distributed largely in temperate areas of the Northern Hemisphere) and their putative phylogenetic position. We believe that future studies of the phylogeny of Prunus should include these tropical representatives of subgenus Laurocerasus, including the species formerly classified in genus Pygeum.
To conclude, s6pdh appears to be a single copy gene in all diploid species of Prunus, with the possible exception in P. emarginata. The rates of evolution of exon sequences of s6pdh are comparable to trnL-trnF, while the introns are closer to ITS, with the exception of intron 4, where the rates are almost twice as high. An acceleration in the rates of all three markers is observed in the Cerasus-Laurocerasus-Padus clade. Maximum likelihood and parsimony analyses of s6pdh confirmed some findings from previous work with ITS and trnL-trnF. An early divergence in the evolution of Prunus separated Laurocerasus and Padus from the rest of genus. Subgenus Prunus includes section Microcerasus. The two nuclear markers, ITS and s6pdh, support slightly different phylogenies within subgenus Prunus, which are most likely due to lineage sorting. Future work in this genus should concentrate on species from the tropics, classified in subgenus Laurocerasus, and, to a less degree, on additional sampling of section Microcerasus.
|
|
FOOTNOTES |
|---|
2 Author for reprint requests (pebortiri{at}ucdavis.edu
)
|
LITERATURE CITED |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Bortiri E. S.-H. Oh J. Jiang S. Baggett A. Granger C. Weeks M. Buckingham D. Potter D. E. Parfitt 2001 Phylogeny and systematics of Prunus (Rosaceae) as determined by sequence analysis of ITS and the chloroplast trnL-trnF spacer DNA. Systematic Botany 26: 797-807
Brako L. J. L. Zarucchi 1993 Catalogue of the flowering plants and gymnosperms of Peru. Missouri Botanical Garden, St. Louis, Missouri, USA
Bremer K. 1988 The limits of amino acid sequence data in angiosperm phylogenetic reconstruction. Evolution 42: 795-803[CrossRef][ISI]
Browicz K. D. Zohary 1996 The genus Amygdalus L. (Rosaceae): species relationships, distribution, and evolution under domestication. Genetic Resources and Crop Evolution 43: 229-247
Citerne H. L. M. Möller Q. C. B. Cronk 2000 Diversity of cycloidea-like genes in Gesneriaceae in relation to floral symmetry. Annals of Botany 86: 167-176
Cunningham C. W. 1997 Is congruence between data partitions a reliable predictor of phylogenetic accuracy? Empirically testing an iterative procedure for choosing among phylogenetic methods. Systematic Biology 46: 733-740
Darlington C. D. A. P. Wylie 1955 Chromosome atlas of flowering plants. George Allen and Unwin, London, UK
Dellaporta S. L. J. Wood J. B. Hicks 1983 A plant DNA minipreparation: version II. Plant Molecular Biology Reporter 1: 19-21
Doyle J. J. J. L. Doyle 1987 A rapid DNA isolation procedure for small quantities of fresh leaf tissue. Phytochemistry Bulletin 19: 11-15
Elias T. S. 1980 Trees of North America. Van Nostrand Reinhold, New York, New York, USA
Eriksson T. 1998 AutoDecay, version 4. Department of Botany, Stockholm University, Stockholm. Available at: http://www.bergianska.se/personal/TorstenE/
Evans R. C. L. A. Alice C. S. Campbell E. A. Kellogg T. A. Dickinson 2000 The granule-bound synthase (GBSSI) gene in the Rosaceae: multiple loci and phylogenetic utility. Molecular Phylogenetics and Evolution 17: 388-400
Farris J. S. M. Källersjö A. G. Kluge C. Bult 1994a Permutations. Cladistics 10: 65-76[CrossRef][ISI]
