| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
2 Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3DS, UK; 3 Royal Botanic Garden, Edinburgh, Scotland, EH3 5LR, UK; 4 Institute of Cell and Molecular Biology, University of Edinburgh, Rutherford Building, King's Buildings, Mayfield Road, Edinburgh, Scotland, EH9 3JR, UK; and 5 Smith College, Northampton, Massachusetts 01063 USA
Received for publication January 8, 1999. Accepted for publication December 17, 1999.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: molecular phylogeny • rbcL • Rhamnaceae • trnL-F
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
50 genera and 900 species. They are characterized by flowers with petal-opposed stamens (obhaplostemony) and a tendency towards xeromorphism. Obhaplostemony is a relatively rare feature in angiosperms, and this has resulted in Rhamnaceae being associated with other families such as Vitaceae and Cornaceae exhibiting this arrangement. The xeromorphic adaptations exhibited by some members of the family include reduced or absent leaves, crowding of leaves, shortening of branch axes, presence of thorns or spines, and a low, shrubby habit. There are few plants of economic value in Rhamnaceae, the most notable being the jujube (Ziziphus jujuba), a fruit tree, and the ornamental shrubs Ceanothus and Colletia.
The history of the taxonomic relationships of Rhamnaceae is presented in Table 1. Two patterns have generally been followed: either Rhamnaceae have been placed with groups such as Vitaceae on the basis of shared floral features (Takhtajan, 1980
; Cronquist, 1988
) or with Elaeagnaceae on the basis of shared vegetative characteristics (Thorne, 1992
; Takhtajan, 1997
).
|
. The most recent revision of the entire family by Suessenguth (1953)
recognized Hooker's (1862) five tribes and included a total of 58 genera and 984 species. A summary of Suessenguth's system is presented in Table 2. Prior to this study Rhamnaceae comprised 49 genera (a few genera have been described subsequent to Suessenguth's treatment and others are now treated as congeneric; see Richardson et al., 2000
). The suprageneric or tribal classification of Rhamnaceae had been based largely on fruit characters, which in Suessenguth's system resulted in the circumscription of two large and otherwise heterogeneous tribes, Rhamneae Hook. f. and Zizipheae Brongn. (= Paliureae Reiss. ex Endl.).
|
An analysis of sequences of the plastid gene rbcL for 499 species of angiosperms (Chase et al., 1993
) showed that Rhamnaceae are part of a weakly supported group also containing Rosaceae, Urticales, and Fagales. Further studies using rbcL (Soltis et al., 1995
) indicated a close relationship between Elaeagnaceae and Rhamnaceae. Studies using 18S nuclear ribosomal DNA, atpB and rbcL sequence data (Soltis et al., 1997
; Savolainen et al., 2000
) have supported the link between Rhamnaceae and Elaeagnaceae. Sequence data for a plastid noncoding region have placed Barbeyaceae and Dirachmaceae in association with Rhamnaceae (Thulin et al., 1998
). The occurrence of nitrogen-fixing symbioses in some Rhamnaceae, Elaeagnaceae, Ulmaceae, and Rosaceae offers further support for a close relationship among these families (Soltis et al., 1995
; Swensen, 1996
).
Taxa from the above families were included in this analysis in an attempt to refine ideas about relationships among them as well as among genera within Rhamnaceae. Sequences were obtained from two regions of the plastid genome for 66 taxa in Rhamnaceae and related families. Sequence data from rbcL have been widely applied at the intrafamilial level, for example in Dipsacales (Donoghue, 1992
), Geraniaceae (Price and Palmer, 1993
), Cornaceae (Xiang et al., 1993
), Saxifragaceae sensu stricto (Soltis et al., 1993
), Rosaceae (Morgan, Soltis, and Robertson, 1994
), Droseraceae (Williams, Albert, and Chase, 1994
), Zygophyllaceae (Sheahan and Chase, 1996
), Themidaceae (Fay and Chase, 1996
), Lecythidaceae (Morton et al., 1997
), and Plumbaginaceae (Lledó et al., 1998
). We have extended the sampling for the trnL (UAA) 5' intron and the intergenic spacer between the trnL (UAA) 3' exon and trnF (GAA; Taberlet et al., 1991
) from that of Thulin et al. (1998)
. This region, subsequently referred to as trnL-F, has been used in suprageneric phylogenetic analysis of Asparagales (Fay et al., 2000
) and Haemodoraceae (Hopper et al., 1999
). The results of the combined analysis of these data were used in part to redefine the suprageneric classification of Rhamnaceae (Richardson et al., 2000
). A discussion of the comparative utility in phylogenetic analysis and aspects of the molecular evolution of rbcL and trnL-F data are also presented.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Soltis et al., 1995
; Thulin et al., 1998
) had shown this family to be the sister of the rest.
|
1.0 g fresh, 0.2–0.25 g silica gel dried leaves or 0.1–0.2 g of material from herbarium sheets using a 2X CTAB (hexadecyltrimethylammonium bromide) method modified from Doyle and Doyle (1987)
; DNA was precipitated using isopropanol instead of ethanol because it was found to be more reliable for Rhamnaceae, and DNA extracted from herbarium material was left to precipitate for at least 3 wk at -20°C (Fay et al., 1998
) because this has been shown to give better yields. The reasons for this are unclear, but it could be due to the presence of altered secondary compounds that form as a result of the degradation associated with drying (making the DNA more difficult to precipitate) or simply because the DNA from herbarium specimens is degraded and therefore takes longer to precipitate. DNA was extracted from herbarium specimens for 21 of the 66 taxa. All DNA samples were purified on cesium chloride/ethidium bromide gradients (1.55 g/mL).
Gene amplification and purification
For most taxa, amplification of the rbcL exon was carried out in two overlapping halves using forward primers beginning at positions 1 and 636 and reverse primers beginning at position 724 and at a downstream ribosomal control site (Fay et al., 1998
). DNA from some herbarium specimens had to be amplified in shorter pieces using forward primers beginning at positions 636 and 895 and reverse primers beginning at position 1024 and the downstream site. Amplification of the trnL-F region (Taberlet et al., 1991
) was carried out using the forward primer c and the reverse primer f, and some taxa had to be amplified in shorter pieces using the primer pairs c and d and e and f. The d and e primers are exact complements so these sequences have a 20 base-pair (bp) gap where the primer site is located. Amplification products were purified using Magic mini-columns (Promega, Southampton, Hampshire, UK) or QIAquick columns (Qiagen, Crawley, West Sussex, UK), following protocols provided by the manufacturers.
DNA sequencing
Standard dideoxy methods using S35 (for 12 rbcL sequences) or modified dideoxy cycle sequencing with dye terminators run on an ABI 373A or 377 automated sequencer (according to the manufacturer's protocols; Applied Biosystems, Inc., Warrington, Cheshire, UK) were used to sequence the amplification products directly. Automated sequence output files were edited and assembled using Sequence Navigator and Autoassembler (Applied Biosystems Inc.). All sequences have been submitted to GenBank (for accession numbers, see Table 3).
Sequence alignment
Alignment of rbcL sequences was easily performed manually because of the absence of insertions or deletions. An initial alignment was performed for five trnL-F sequences using Clustal version 1.61 (Higgins, Bleasby and Fuchs, 1992
). Subsequent sequences were aligned manually (aligned matrices available at http://www.botany.org/bsa/ajbsupp/v86/s01-01.html and from the first and last authors j.richardson@rbge.org.uk/m.chase@rbgkew.org.uk).
After alignment of the trnL-F matrix, a matrix of insertion/deletion characters was prepared (characters were coded as present or absent). These characters were given weight equal to that of all single characters in the matrix. A large deletion can mask other smaller deletions, and taxa that have these larger deletions are coded as unknown for deletions that occur entirely within them. For example, there is a deletion between positions 891 and 941 for some taxa, whereas in other taxa there are smaller deletions between these positions, which we have coded as missing for taxa with the larger deletion.
We used 1399 rbcL characters and 1258 trnL-F characters. The ends were clipped from the sequences to remove primer sites (i.e., 20 bp from beginning of rbcL, 24 bp from the beginning and 28 bp from the end of trnL-F). Two regions of 59 and 16 bp of the trnL-F matrix were too ambiguous to be confidently aligned and so were excluded from all analyses.
Phylogenetic analysis
Data were analyzed using the parsimony algorithm of the software package PAUP version 3.1.1 for Macintosh (Swofford, 1993
). Searches were conducted on the separate rbcL and trnL-F data sets (which included the matrix of 16 trnL-F indel characters) and on both data sets combined. Tree searches were conducted under the equal and unordered weights criterion (Fitch parsimony; Fitch, 1971
) with 1000 random sequence additions and TBR (tree bisection-reconnection) swapping, but permitting only five trees to be held at each step. The limit on the number of trees held at each step was implemented to cut down the time spent searching on suboptimal trees. All shortest trees collected in the 1000 replicates were then used as starting trees for another round of heuristic search, and all these trees were swapped on to completion or swapped on until 6000 trees were produced, at which point we limited the number of trees and swapped on the 6000 trees collected. This search strategy should find islands of equally parsimonious trees (Maddison, 1991
). To investigate the existence of islands, we then used one of the shortest trees as a starting tree and if all 6000 trees found produced the same strict consensus tree, then we concluded that only one island of trees was present (all shortest trees appeared to be from one island, so this topic will not be discussed further).
Successive approximations weighting (SW; Farris, 1969
) was then carried out on these trees with a limit of ten trees per replicate and ten replicates per round. The trees collected in the ten replicates were then swapped to completion (or until we reached 6000 trees, as above) before the next round of weighting was implemented. Further rounds of SW continued until tree lengths were the same in two consecutive rounds. Characters were reweighted according to their rescaled consistency indices (RC), with a base weight of 1000. This procedure was designed to downweight or eliminate characters that were highly homoplasious.
One thousand replicates of the bootstrap (Felsenstein, 1985
) were then carried out with the successive weights applied. We applied the following scheme of support: bootstrap values of 50–74% represent weak support, 75–84% moderate support, and 85–100% strong support. Unless otherwise noted, we use the word "support" to mean internal support, in this case estimated by the bootstrap.
MacClade version 3.05 (Maddison and Maddison, 1992
) was used to calculate the number of steps and consistency index (CI) and retention index (RI) for each codon position in the rbcL analysis (Table 4), and CI and RI values of indel characters from the trnL-F matrix (Table 5). MacClade was also used to plot the number of unambiguous steps per character optimized on the most parsimonious SW tree from the combined analysis and the number of characters per number of steps on both the trnL-F and rbcL trees. The Fitch CI and RI were each calculated for transitions and transversions on the SW tree of the combined analysis using a step matrix in which transitions were weighted to zero, thus calculating the CI and RIs of transversions on the combined tree, and from these we calculated the number and CI/RI of transitions (Table 6). Tree lengths for all matrices are given in Table 7.
|
|
|
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
|
Within Rhamnaceae, strongly bootstrap-supported major groups are identified: a ziziphoid group that has Elaeagnaceae as a sister group; an ampeloziziphoid group that contains the genera Ampeloziziphus, Doerpfeldia, and Bathiorhamnus (the inclusion of Ventilago in this group does not occur in all Fitch trees and recieves <50% SW bootstrap support); and a rhamnoid group that has the ampeloziziphoid group as its sister.
Other strongly supported groups within these larger groups include: (1) in the ziziphoid group: (a) Pomaderreae Reiss. ex Endl., a group of Australian taxa (Pomaderris, Siegfriedia, Spyridium, and Trymalium), (b) Ceanothus, (c) Phyliceae Reiss. ex Endl., a group of taxa with a southern African center of distribution (Phylica, Nesiota, and Noltea), (d) Colubrina, (e) Paliureae, including Ziziphus, Paliurus, and Hovenia, and (f) Gouanieae (excluding Pleuranthodes), and (2) in the rhamnoid group: Rhamneae, a clade composed of Karwinskia, Condalia, Krugiodendron, Reynosia, Rhamnella, Rhamnidium, Berchemia, Sageretia, Rhamnus, Frangula, and Scutia with the monotypic tribe Maesopsideae being sister to Rhamneae.
Analysis of trnL-F data
The aligned trnL-F data matrix had 595 variable characters and 353 potentially informative characters out of a total of 1258 characters used (i.e., 28%). The heuristic search produced >6000 equally parsimonious Fitch trees with 1339 steps, CI = 0.67 (0.57 excluding autapomorphies), and RI = 0.75. Application of SW produced >6000 trees with a length of 652 105 steps, CI = 0.87 (0.76 excluding autapomorphies), and RI = 0.91. The Fitch length of the SW trees was 1339, i.e., the weighted trees were a subset of the Fitch trees from the same island. Figure 2 shows one of the weighted trees with Fitch branch lengths (ACCTRAN optimization) and SW bootstrap percentages; branches that collapse in the strict consensus tree of the Fitch analysis are marked with a solid arrow, and those not present in the strict consensus of the weighted tree are marked with an open arrow. The performance of the indel characters is shown in Table 5. The average CI was 0.84, and the average RI was 0.90.
|
Within Rhamnaceae, the strongly supported major groups identified in the rbcL analysis (i.e., the ziziphoid, rhamnoid, and ampeloziziphoid groups) receive further support. The inclusion of Ventilago in the rhamnoid group and not the ampeloziziphoid group (as in the rbcL tree) is strongly supported. Generally speaking, the generic relationships and the larger clades identified are highly congruent with the rbcL results. We did not use any congruence metrics because these are too coarse to be useful (Wiens, 1998
) and instead preferred node by node inspection to determine if evidence of strongly bootstrap-supported incongruence was present (i.e., less resolution and lower bootstrap percentages in the combined trees).
Analysis of combined rbcL and trnL-F data
The combined matrix produced 324 Fitch trees with a length of 2559 steps, a CI = 0.59 (0.48 excluding autapomorphies) and RI = 0.70. With SW there was only one tree with two trichotomies, tree length of 1 068 277 steps, CI = 0.85 (0.71 excluding autapomorphies), and RI = 0.88. Figure 3 shows this single tree with its Fitch branch lengths (ACCTRAN optimization) and SW bootstrap values; branches that collapse in the strict consensus tree of the Fitch analysis are marked with a solid arrow. The Fitch length of this tree was 2559 steps (i.e., it was one of the Fitch trees from the same island).
|
)/sampling error (Huelsenbeck, Bull, and Cunningham, 1996
).
Molecular evolution
Figure 4 shows a plot of the number of steps per character optimized on the single most parsimonious SW tree from the combined analysis. The trnL-F region has a more even distribution of substitutions along its length than rbcL. Figure 5 shows the number of steps per character on both the trnL-F and rbcL trees, indicating that some rbcL characters change up to 16 times, whereas the trnL-F positions change up to nine times only. We consider the existence of such highly homoplasious positions to justify the use of SW, which downweights only those characters that change frequently based on the best RC from any tree.
|
|
The transition/transversion ratio for rbcL calculated on the single combined SW tree was 1.17. The trnL-F transition/transversion ratios based on the combined SW tree were calculated separately for the intron, exon, intergenic spacer, and they were (258/282) 0.91, (6/1) 6.0, and (340/342) 0.99, respectively. Thus, for rbcL there is a bias for transitions, whereas noncoding regions have no bias. Transitions have higher CIs and RIs (Table 6) than tranversions for both rbcL and trnL-F.
Table 7 shows the tree lengths when analyzed alone for rbcL and trnL-F as well as the number of steps for rbcL and trnL-F data sets optimized on the single combined SW tree. Both of the separate analyses underestimate the number of substitutions relative to the combined tree. The trnL-F region had a 1339/1347 difference, which is a 0.6% underestimate of change in the trnL-F tree compared to the combined tree. The rbcL gene had a 1174/1194 difference, which is a 1.7% underestimate of change in the rbcL tree compared to the combined tree. Thus the rbcL tree was less accurate than the trnL-F tree. We consider the combined tree more accurate because overall bootstrap percentages are higher than either rbcL or trnL-F analyzed separately.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
), and third positions in codons are expected to be more variable than first or second positions, as is the case with this rbcL data set (Table 4). Direct combination of matrices should enhance the common although perhaps partially to wholly overshadowed patterns, and this could result in a slightly different but more strongly supported pattern than either matrix analyzed separately (as is the case here). Because of the degenerate nature of the genetic code, first and second positions in a codon are under higher levels of direct selection, and therefore fewer of these can change than third positions. In noncoding regions such as trnL-F there is probably less functional constraint than there is in rbcL (constraints in noncoding regions could involve ribosome control sites and other structural features). Given similar constraints among noncoding characters, rates of change among noncoding characters should be more similar, and this is what we found, i.e., trnL-F has a more even pattern of change than rbcL (Fig. 4). Also, on average each rbcL position changes more often than in trnL-F, and rbcL has more hypervariable positions than trnL-F (Fig. 5). This uneven pattern of variation in rbcL makes it harder to detect all changes (i.e., all instances of homoplasy) in such positions and is therefore more likely to produce misrepresentations of relationships (i.e., underestimates in the actual amount of change). As compared to the combined tree, which is better supported by the bootstrap, the rbcL tree misses more homoplasious changes and has a less accurate overall topology (rbcL is a 1.7% underestimate of the combined tree, whereas trnL-F is only a 0.6% underestimate).
As discussed above, combining matrices should strengthen only shared signal, which is likely to be the phylogenetic one (see also Chase and Cox, 1998
). In general, similar weakly supported patterns of separate data sets would be expected to be more strongly supported when combined. Finally, combining data sets detects evidence for additional substitutions not found in the individual matrices, thus permitting more accurate overall character reconstruction and estimates of relationships. As a result, combined trees might be expected to be longer than any of the individual matrix trees because combined matrices should recover more of the unobserved substitutions, which is the case in this study (Table 7). The greater underestimate in change for rbcL compared to trnL-F may have resulted in the spurious nesting of Elaeagnaceae, Barbeyaceae, and Dirachmaceae within Rhamnaceae in the rbcL analysis.
A further rbcL analysis was run in which the monophyly of Rhamnaceae was constrained. This analysis produced a Fitch tree of 1175 steps, i.e., only one step longer than the nonconstrained analysis. Such underestimates of change on single gene matrices highlight the limitations of too little data in which patterns are too weak for accurate reconstruction (i.e. sampling error sensu Huelsenbeck et al., 1997
), not the unreliability of parsimony as an optimality criterion. The following sections of the discussion will focus mainly on the combined tree, which we view as the most accurate, for the reasons explained above.
Thirteen of the 16 indel characters from the trnL-F data set were nonhomoplasious synapomorphies. Therefore, in this analysis, indel characters appear to be good phylogenetic markers. Six indels appear to be unique in sequence, and the other ten are copies or near copies of adjacent regions.
In this data set, coding regions have a transition bias, whereas introns or nontranscribed spacers have no apparent bias. The lack of bias in noncoding trnL-F (Table 6) is consistent with the findings of Morton and Clegg (1995)
who demonstrated that substitutions in noncoding regions of the plastid genome were affected by the two immediately flanking bases. When both the 5' and 3' flanking nucleotides are G or C, Morton and Clegg found that only 25% of the observed substitutions were transversions, whereas if the flanking nucleotides were both A or T 57% of the substitutions were transversions. Because noncoding regions of the plastid genome are more A/T rich, the relative proportion of transversions increases, resulting in a more even transition/transversion ratio.
Differential rates of change (codon positions or ts/tv) have been used to justify relative weighting schemes in phylogenetic analyses (e.g., Fu, 1998
; Smith, 1998
; Zink and Blackwell, 1998
). The enhanced performance of transitions (Table 6) and third codon positions in rbcL as indicated by CI and RI values (Table 4) shows that differential weighting of whole classes of characters is not justified.
Relationships of Rhamnaceae
The Dirachma/Barbeya alliance is strongly supported by the bootstrap analysis. This clade is the sister group of Rhamnaceae in the combined tree with moderate support. Thulin et al. (1998)
suggested that the families Barbeyaceae and Dirachmaceae should be retained because they differ so significantly in morphology, and the results here also indicate that this would be the best circumscription for these families given the large number of morphological and molecular differences between them and the other families. Greater sampling from within the urticalean families and Rosaceae may result in a better placement of Barbeyaceae and Dirachmaceae, but their combination of traits otherwise restricted to either Rhamnaceae or the urticalean families (Thulin et al., 1998
) appears to indicate either a position as obtained here or as sister to the urticalean families is appropriate.
Relationships within Rhamnaceae
Classification based solely on DNA sequence data should be treated with caution unless corroborated by evidence from other sources, but the rbcL/trnL-F data have indicated patterns that were not apparent from previous non-phylogenetic studies of morphology and anatomy. The single SW tree from the combined analysis shows that Rhamnaceae are a well-supported monophyletic group and also provides support for some of Suessenguth's tribes. However, these data show a division of Rhamnaceae into three clades supported by bootstrap values of 99 or 100, but for which there are no obvious morphological apomorphies (no formal morphological phylogenetic analysis is presented here, but see Richardson et al., 2000
). Such groups were described as "cryptic clades" (Wojciechowski et al., 1993
) in a study that identified a strongly supported clade of dysploid North American Astragalus, which was found to be supported by three different lines of genotypic evidence (chromosomal, nuclear rDNA, and plastid DNA) but for which there were no morphological apomorphies. The authors suggested that the group should be given an informal name, and we have likewise chosen to adopt informal names for the three cryptic clades identified here. The only morphological evidence for these groups comes from studies of gynoecium ontogenesis (Medan, 1988
). A molecular phylogenetic study of Saxifragaceae (Soltis, Soltis, and Thompson, 1992
) indicated that inferior or superior ovaries, which had been used in the classification of Lithophragma, were homoplasious. However, a study of gynoecium ontogenesis revealed that patterns in the initial development of the ovary were consistent with the molecular phylogeny. Monophyletic groups within Lithophragma could be defined on the basis of whether the floral apex in the initial stages of development is more or less flat or with a circular depression. Subsequent ontogenetic development leads to superior or inferior ovaries from both meristem conditions, and thus ovary conditions at maturity are not always homologous. Without ontogenetic investigation this would seem to represent a case of parallel evolution. A similar phenomenon could be occurring in Rhamnaceae; Medan (1988)
studied the shape of the floral apex and the degree of intercalary growth at carpellary bases in 17 genera of Rhamnaceae. In some taxa the floral apex is more or less flat at the time of primordia differentiation, e.g., Condalia and Rhamnus, and these taxa usually go on to form superior ovaries. In other taxa the floral apex shows a circular depression at the time of primordia differentiation, e.g., Colletia, Noltea, Phylica, and Pomaderris, and these taxa usually go on to form inferior or semi-inferior ovaries. We could therefore have a situation in Rhamnaceae similar to that of Lithophragma in which the latter stages of ontogenetic development of floral apices may obscure the initial patterns. The limited sampling in the Medan (1988)
study could be expanded and provide morphological character support for the cryptic clades defined by the molecular data.
Group 1: rhamnoid clade
This clade is divided into three strongly supported subgroups. The first of these comprises Rhamneae with genera such as Rhamnus and Berchemia that have drupaceous fruits, superior ovaries and a nectariferous disc either partly or totally adnate to the calyx tube. The interrelationships of the genera within this group are not particularly well supported. The second subgroup, the monotypic tribe Maesopsideae, consists of the monotypic genus Maesopsis, which is a sister to Rhamneae. It is so morphologically divergent that its inclusion in Rhamneae seems unwarranted. Ventilagineae are the third distinct subgroup with strong support as sister to the Maesopsideae–Rhamneae alliance. All members of this third tribe are climbers with apically winged fruits and semi-inferior ovaries. We had no success with herbarium DNA for Smythea, which is the only other genus previously placed in this tribe. However, this genus is similar in morphology to Ventilago and should therefore be tentatively included within Ventilagineae.
Group 2: ampeloziziphoid clade
This group consists of three highly divergent genera that have palmately veined leaves and drupaceous fruits: Ampeloziziphus, a monotypic Brazilian climber with semi-inferior ovaries and a thick nectariferous disc; Doerpfeldia, a monotypic tree from Cuba with small leaves and a superior ovary thinly covered by the nectariferous disc; and Bathiorhamnus, a genus of two Madagascan tree species with a superior ovary and thick nectariferous disc. There are, however, no exclusive morphological similarities linking these genera. The high levels of molecular divergence between these genera and their highly disjunct distribution indicates that they are only distantly related, and it is likely that they are remnants of groups that were formerly more diverse and widespread. These three genera have each been placed in a separate tribe because of their highly divergent nature (Richardson et al., 2000
).
Group 3: ziziphoid clade
The third major clade comprises genera that usually have semi-inferior to inferior ovaries and capsular fruits. There are, however, exceptions to this, e.g., Ziziphus has drupaceous fruits. In addition some genera of the tribe Colletieae have superior ovaries or drupaceous fruits. This ziziphoid group may be further split into a number of subgroups. Suessenguth's more derived tribes Colletieae and Gouanieae are strongly supported monophyletic groups. There are some differences between rbcL and trnL-F regarding the position of Pleuranthodes, but moving this taxon from its unresolved position in the rbcL tree to within Gouanieae using MacClade does not significantly increase tree length. With trnL-F, Pleuranthodes is strongly supported in an embedded position in Gouanieae. We view this shift as merely an indication that rbcL has no clear signal for Pleuranthodes, which is indicated by strong support at several nodes as a member of Gouanieae in the combined tree. Gouanieae are climbers with tendrils and longitudinally winged fruits; Colletieae are a group of strongly armed trees or shrubs. Australian Pomaderreae are characterized by the presence of stellate hairs.
Ziziphus, Paliurus, and Hovenia make up another strongly supported tribe in the combined analysis, Paliureae. Hovenia appears to have a close relationship with Ziziphus and Paliurus in that they both have cymose inflorescences, a base chromosome number of x = 12, and a similar pollen exine structure, and on the basis of this evidence Hovenia is placed in Paliureae. The strongly supported, predominantly South African Phyliceae, consisting of Phylica, Nesiota, and Noltea, also appear distinct and are generally characterized by having an ericoid shrubby habit, inferior ovaries, and leaves with revolute margins and tomentose undersurfaces.
Colubrina, which includes trees or shrubs with the nectariferous disc filling the receptacle and surrounding the ovary, and the similar Lasiodiscus were always thought to be closely related (Johnston, 1971
; Figueiredo, 1995
), but only the rbcL matrix produced trees in which Colubrina and Lasiodiscus form a clade. Further sampling of Lasiodiscus and studies of other sequences might be necessary to lend more molecular support for a Colubrina/Lasiodiscus grouping. The two morphologically similar genera may eventually be treated as a distinct tribe, but at this time there is insufficient evidence to recognize this group.
The affinities of a number of other genera are unclear. The arborescent genus Alphitonia from Malaysia, Australia, and the western Pacific has fruit exocarps that are thick, spongy, and friable at maturity, and Emmenosperma is similar in having red arillate seeds persisting on the receptacle after dehiscence. Again further evidence is needed to place these two genera together in a separate tribe. According to the trnL-F and combined analyses, Schistocarpaea appears to be closely related to Colletieae, in spite of having few supporting morphological characters.
The North American genus Ceanothus is characterized by having receptacles and nectariferous discs persisting on the pedicel, and its relationship with the other clades is unresolved. Ceanothus, and a number of genera in Colletieae (Colletia, Discaria, Kentrothamnus, Retanilla, and Trevoa) engage in root nodular fixation of nitrogen in a symbiotic association with the actinomycete bacterium Frankia (Baker and Schwintzer, 1990
). Soltis et al. (1995)
showed that root-nodulating angiosperms fall within one clade and may share a genetic predisposition to nodulation, even though most of the members of this clade do not nodulate. Root-nodulating members of Rhamnaceae also fall within one clade, the ziziphoids, in which most of the members do not nodulate, but relationships between groups within this clade are not clearly resolved. It could be that with additional data Ceanothus and Colletieae could end up as sister taxa, in which case root nodulation has arisen only once within the family (with a subsequent loss in Adolphia).
The molecular trees from the combined analysis have been used in conjunction with a phylogenetic analysis of morphological data to recircumscribe the tribal classification of Rhamnaceae (Richardson et al., 2000
). Tribes were delimited on the basis of strong bootstrap support in the molecular results with additional support from morphology (summarized in Table 8).
|
stated that "Rhamnaceae are so well represented both in tropical and temperate regions that it is difficult to trace the history of the family." The Rose Creek flower was ascribed to Rosaceae by Basinger and Dilcher (1984)
, but it is clearly a member of Rhamnaceae with obhaplostemonous flowers and "rhamnaceous" pollen. This plant has been dated to 94–96 million years old, and thus it establishes this as the minimum age for the family. Given such an age, it is likely that Rhamnaceae had already become cosmopolitan before the continents became widely separated, and the following discussion is based upon this assumption.
Two general patterns in the distribution of the three major groups within Rhamnaceae can be observed. The ampeloziziphoid group illustrates a pattern of disjunction also found in other groups between northern South America and Madagascar (e.g., Fay et al., 1998
). In this case there are long branch lengths and a lack of morphological similarities, indicating that this group is ancient and probably had a much wider distribution subsequently reduced by extinction, particularly in Africa. The other major groups likely had a similarly wide distribution not reduced by extinction to the same extent as the ampeloziziphoid group. Overlaid on this pattern, is another, presumably post-Gondwanan, in which groups are more or less restricted to individual plates or later separating plates (e.g., Colletieae in South America, Australia, and New Zealand, Phyliceae in Africa, and Pomaderreae in Australia).
The ziziphoid group is cosmopolitan with a predominantly southern hemisphere distribution and could be of Gondwanan origin with the exception of Ceanothus, which has a North American distribution. This indicates that either this whole southern group was widespread throughout Gondwanaland and parts of Laurasia (in what is now North America) and subsequently contracted, largely to the southern hemisphere, or that Ceanothus arrived at its present location by long-distance dispersal. California has many examples of relictual taxa from lineages that are otherwise restricted to the Old World or the southern hemisphere; these include species of Paeonia (Paeoniaceae), Odontostomum (Tecophilaeaceae; Brummitt et al., 1998
) and Fremontodendron (Malvaceae; Bayer et al., 1999
). Ceanothus is sister to other clades within the ziziphoid group and morphologically highly autapomorphous (Richardson et al., 2000
); therefore we do not consider it to be a recent derivative of one of these clades and thus the most likely explanation for its present distribution is that it is relictual and reasonably old (65 million years).
Gouanieae have a similar distribution to the ampeloziziphoid group, with some genera of the group also being found in Africa. Colubrina is predominantly found in northern South America, although species are also found in Asia, Hawaii, Madagascar, and South Africa. Lasiodiscus is found in Africa and Madagascar, and this distribution may represent the remnants of previously more widespread groups that are now only found on Madagascar or in rain and coastal forests in the tropical parts of sub-Saharan Africa and east Africa. Alphitonia, Pomaderreae, and Schistocarpaea are Australasian taxa, representing isolated clades. Colletieae are a mostly South American group, but two species of Discaria are found in Australia and New Zealand. This is a southern hemisphere disjunction, which is also found in other groups such as Orthrosanthus, Libertia, Berberidopsis and Eucryphia, and these are probably relicts of formerly more widespread groups that were present throughout southern South America, east Antarctica, Tasmania, New Zealand, and eastern Australia (Africa and Madagascar having split earlier from Gondwana).
Within Rhamneae, relationships are not clearly resolved by trnL-F and rbcL sequence data. A more in-depth molecular study using a more variable region and additional taxon sampling is needed to clarify relationships before any biogeographical conclusions can be drawn. However, they do form a strongly supported monophyletic unit that has a wide distribution throughout the tropics and into northern temperate regions. Ventilagineae, found in the Old World tropics but with a center of diversity in India, could have had a Gondwanan origin and subsequently spread into Asia when India collided with Asia. More species in each genus throughout the family need to be analyzed to make a fine-scale biogeographic assessment.
General conclusions
Rhamnaceae are an old monophyletic family with Barbeyaceae and Dirachmaceae forming their sister. Further data from other fields such as anatomy or chemistry are necessary to provide more evidence for the three "cryptic clades" resolved by the molecular data. Although there is strong molecular support for these major divisions in Rhamnaceae, we have been unable to identify morphological characters that could adequately describe these groups (see also, Richardson et al., 2000
).
What is clear from these results is that the tribes Rhamneae and Zizipheae as circumscribed by Suessenguth are unnatural, and a reclassification of some tribes in Rhamnaceae is necessary. All of the tribes proposed by Richardson et al. (2000)
are strongly supported by the bootstrap in the combined analysis. The molecular data indicate that many morphological character states have evolved in parallel (e.g., leaf venation patterns, fruit type, and pollen exine architecture), but it is not a simple matter of morphology vs. molecules. Classifications based on one particular morphological character (such as Suessenguth's reliance on fruit characters) often do not compare well with those based on many morphological characters. A classification based largely on molecular data with the support of some morphological characters seems a better solution.
|
FOOTNOTES |
|---|
6 Author for correspondence (email: j.richardson{at}rbge.org.uk
) Current address: Royal Botanic Garden, Inverleith Row, Edinburgh, Scotland, EH3 5LR.
|
LITERATURE CITED |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Baker, D. W., and C. R. Schwintzer. 1990 Introduction. In C. R. Schwintzer and J. D. Tjepkema [eds.], The biology of Frankia and actinorhizal plants. Academic Press, San Diego, California, USA
Basinger, J., and D. Dilcher. 1984 Ancient bisexual flowers. Science 224: 511–513
Bayer, C., M. F. Fay, A. Y. de Bruijn, V. Savolainen, C. M. Morton, K. Kubitski, and M. W. Chase. 1999 Support for an expanded family concept of Malvaceae within a recircumscribed order Malvales: a combined analysis of plastid atpB and rbcL DNA sequences. Botanical Journal of the Linnean Society 129: 267–303[CrossRef]
Brummitt, R. K., H. Banks, M. A. T. Johnson, K. A. Doherty, K. Jones, M. W. Chase, and P. R. Rudall. 1998 Taxonomy of Cyanastroideae (Tecophilaeaceae): a multidisciplinary approach. Kew Bulletin 53: 769–803[CrossRef]
Chase, M. W., et al. 1993 Phylogenetics of seed plants: an analysis of nucleotide sequences from the plastid gene rbcL. Annals of the Missouri Botanical Garden 80: 528–580[CrossRef][ISI]
———, and A. V. Cox. 1998 Gene sequences, collaboration and analysis of large data sets. Australian Systematic Botany 11: 215–229[CrossRef][ISI]
Cronquist, A. 1988 The evolution and classification of flowering plants, 2nd. ed. New York Botanical Garden, New York, New York, USA
Donoghue, M. J., R. G. Olmstead, J. F. Smith, and J. D. Palmer. 1992 Phylogenetic relationships of Dipsacales based on rbcL sequences. Annals of the Missouri Botanical Garden 79: 333–345
Doyle, J. J., and J. L. Doyle. 1987 A rapid DNA isolation procedure from small quantities of fresh leaf tissue. Phytochemical Bulletin 19: 11–15
Farris, J. S. 1969 A successive approximations approach to character weighting. Systematic Zoology 18: 374–385[CrossRef]
Fay, M. F., C. Bayer, W. S. Alverson, A. Y. de Bruijn, and M. W. Chase. 1998 Plastid rbcL sequence data indicate a close affinity between Diegodendron and Bixa. Taxon 47: 43–50
———, and M. W. Chase. 1996 Resurrection of Themidaceae for the Brodiea alliance, and recircumscription of Alliaceae, Amaryllidaceae and Agapanthoideae. Taxon 45: 441–451[CrossRef][ISI]
———, P. J. Rudall, S. Sullivan, K. L. Stobart, A. Y. de Bruijn, G. Reeves, F. Qamaraz-Zaman, W.-P. Hong, J. Joseph, W. J. Hahn, J. G. Conran, and M. W. Chase. 2000 Phylogenetic studies of Asparagales based on four plastid DNA loci. In K. L. Wilson and D. A. Morrison [eds.], Monocots—systematics and evolution, Proceedings of the Second International Conference on the Comparative Biology of the Monocots, Sydney, September, 1998, vol. 1. CSIRO, Melbourne, Australia
Felsenstein, J. 1985 Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39: 783–791[CrossRef][ISI]
Figueiredo, E. 1995 A revision of Lasiodiscus (Rhamnaceae). Kew Bulletin 50: 495–526[CrossRef]
Fitch, W. M. 1971 Toward defining the course of evolution: minimum change for a specified tree topology. Systematic Zoology 20: 406–416[CrossRef][ISI]
Fu, J. 1998 Toward the phylogeny of the family Lacertidae: implications from mitochondrial DNA 12S and 16S gene sequences (Reptilia: Squamata). Molecular Phylogenetics and Evolution 9: 118–130[CrossRef][ISI][Medline]
Higgins, D. G., A. J. Bleasby, and R. Fuchs. 1992 CLUSTAL: a new multiple sequence alignment program. Computer Applications in Biosciences 8: 189–191
Hooker, J. D. 1862 Rhamnaceae. In G. Bentham and J. D. Hooker [eds.], Genera Plantarum 1: 371–386. Reeve and Co., London, UK
Hopper, S. D., M. F. Fay, M. Rosetto, and M. W. Chase. 1999 A molecular phylogenetic analysis of the bloodroot and kangaroo paw family, Haemodoraceae: taxonomic, biogeographic and conservation implications. Botanical Journal of the Linnean Society 131: 285–299[CrossRef]
Huelsenbeck, J. P., J. J. Bull, and C. W. Cunningham. 1996 Combining data in phylogenetic analyses. Trends in Ecology and Evolution 11: 152–158[CrossRef]
Hutchinson, J. 1959 The families of flowering plants, vol. 1, Dicotyledons. Clarendon Press, Oxford, UK
Johnston, M. C. 1971 Revision of Colubrina (Rhamnaceae). Brittonia 23: 2–53[CrossRef][ISI]
Lledó, M. D., M. B. Crespo, K. M. Cameron, M. F. Fay, and M. W. Chase. 1998 Systematics of Plumbaginaceae based upon cladistic analysis of rbcL sequence data. Systematic Botany 23: 21–29
Maddison, D. R. 1991 The discovery and importance of multiple islands of most-parsimonious trees. Systematic Zoology 40: 315–328[CrossRef]
Maddison, W. P., and D. R. Maddison. 1992 MacClade, analysis of phylogeny and character evolution, version 3.04. Sinauer, Sunderland, MA
Medan, D. 1988 Gynoecium ontogenesis in the Rhamnaceae: a comparative study. In P. Leins, S. C. Tucker, and P. K. Endress [eds.], Aspects of floral development, 133–141. Cramer, Berlin, Stuttgart, Germany
Morgan, D. R., D. E. Soltis, and K. R. Robertson. 1994 Systematic and evolutionary implications of rbcL sequence variation in Rosaceae. American Journal of Botany 81: 890–903[CrossRef][ISI]
Morton, B. R., and M. T. Clegg. 1995 Neighbouring base composition is strongly correlated with base substitution bias in a region of the chloroplast genome. Journal of Molecular Evolution 41: 597–603[ISI][Medline]
Morton, C. M., S. A. Mori, G. T. Prance, K. G. Carol, and M. W. Chase. 1997 Phylogenetic relationships of Lecythidaceae: a cladistic analysis using rbcL sequence and morphological data. American Journal of Botany 84: 530–540[Abstract]
Price, R., and J. D. Palmer. 1993 Phylogenetic relationships of the Geraniaceae and Geraniales from rbcL sequence comparisons. Annals of the Missouri Botanical Garden 80: 661–671[CrossRef][ISI]
Raven, P. H., and D. I. Axelrod. 1974 Angiosperm biogeography and past continental movements. Annals of the Missouri Botanical Garden 61: 539–673[CrossRef][ISI]
Richardson, J. E., M. F. Fay, Q. C. B. Cronk, and M. W. Chase. 2000 A revision of the tribal classification of Rhamnaceae. Kew Bulletin 55: 311–340[CrossRef]
Savolainen, V., M. W. Chase, S. B. Hoot, C. M. Morton, D. E. Soltis, C. Bayer, M. F. Fay, A. Y. de Bruijn, S. Sullivan, and Y.-L. Qiu. 2000 Phylogenetics of flowering plants based upon a combined analysis of plastid atpB and rbcL gene sequences. Systematic Biology 49: 306–362[CrossRef][ISI][Medline]
Seelanen, T., A. Schnabel, and J. Wendel. 1997 Congruence and consensus in the cotton tribe (Malvaceae). Systematic Botany 22: 259–290[CrossRef][ISI]
Sheahan, M. C., and M. W. Chase. 1996 A phylogenetic analysis of Zygophyllaceae R.Br. based on morphological, anatomical and rbcL DNA sequence data. Botanical Journal of the Linnean Society 122: 279–300[CrossRef]
Smith, M. F. 1998 Phylogenetic relationships and geographic structure in pocket gophers in the genus Thomomys. Molecular Phylogenetics and Evolution 9: 1–14
Soltis, D. E., D. R. Morgan, A. Grable, P. S. Soltis, and R. Kuzoff. 1993 Molecular systematics of Saxifragaceae sensu stricto. American Journal of Botany 80: 1056–1081[CrossRef][ISI]
———, P. S. Soltis, D. R. Morgan, S. M. Swensen, B. C. Mullins, J. M. Dowd, and P. Martin. 1995 Chloroplast gene sequence data suggest a single origin of the predisposition for symbiotic nitrogen fixation in angiosperms. Proceedings of the National Academy of Sciences, USA 92: 2647–2651
———, ———, D. L. Nickrent, L. A. Johnson, W. J. Hahn, S. B. Hoot, J. A. Sweere, R. K. Kuzoff, K. A. Kron, M. W. Chase, S. M. Swensen, E. A. Zimmer, S. M. Chaw, L. J. Gillespie, W. J. Kress, and K. J. Sytsma. 1997 Angiosperm phylogeny inferred from 18S ribosomal DNA sequences. Annals of the Missouri Botanical Garden 84: 1–49
———, ———, and J. N. Thompson. 1992 Chloroplast DNA variation in Lithophragma (Saxifragaceae). Systematic Botany 17: 607–619[CrossRef][ISI]
Suessenguth, K. 1953 Rhamnaceae, Vitaceae, Leeaceae. In A. Engler and K. Prantl [eds.], Die Naturlichen Pflanzenfamilien 2, Aufl., 20d. Duncker and Humbolt, Berlin, Germany
Swensen, S. M. 1996 The evolution of actinorhizal symbioses: evidence for multiple origins of the symbiotic association. American Journal of Botany 83: 1503–1512[CrossRef][ISI]
Swofford, D. L. 1993 PAUP: phylogenetic analysis using parsimony, version 3.1.1. Computer program distributed by the the Illinois Natural History Survey, Champaign, Illinois, USA
Taberlet, P., L. Gielly, G. Pautou, and J. Bouvet. 1991 Universal primers for amplification of three non-coding regions of chloroplast DNA. Plant Molecular Biology 17: 1105–1109[CrossRef][ISI][Medline]
Takhtajan, A. 1980 Outline of the classification of flowering plants (Magnoliophyta). Botanical Review 46: 226–359
———. 1997 Diversity and classification of flowering plants. Columbia University Press, New York, New York, USA
Thorne, R. F. 1992 Classification and geography of flowering plants. Botanical Review 58: 225–348
Thulin, M, B. Bremer, J. E. Richardson, J. Niklasson, M. F. Fay, and M. W. Chase. 1998 Family relationships of the enigmatic rosid genera Barbeya and Dirachma from the Horn of Africa region. Plant Systematics and Evolution 213: 103–119[CrossRef][ISI]
Wiens, J. J. 1998 Testing phylogenetic methods with tree congruence: phylogenetic analysis of polymorphic morphological characters in phrynosomatid lizards. Systematic Biology 47: 427–444[CrossRef][ISI]
Williams, E. S., V. A. Albert, and M. W. Chase. 1994 Relationships of Droseraceae: a cladistic analysis of rbcL sequence and morphogical data. American Journal of Botany 81: 1027–1037[CrossRef][ISI]
Wojciechowski, M. F., M. J. Sanderson, B. G. Baldwin, and M. J. Donoghue. 1993 Monophyly of aneuploid Astragalus (Fabaceae): evidence from nuclear ribosomal DNA internal transcribed spacer sequences. American Journal of Botany 80: 711–721[CrossRef][ISI]
Xiang, Q.-Y., D. E. Soltis, D. R. Morgan, and P. S. Soltis. 1993 Phylogenetic relationships of Cornus L. sensu lato and putative relatives inferred from rbcL sequence data. Annals of the Missouri Botanical Garden 80: 723–734
