| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Biology, University of Missouri-St. Louis, St. Louis, Missouri 63121; and The Missouri Botanical Garden, St. Louis, Missouri 63166
Received for publication November 2, 1998. Accepted for publication February 23, 1999.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: Atherospermataceae • Calycanthaceae • Gomortegaceae • Hernandiaceae • Lauraceae • Monimiaceae • Siparunaceae • Trimeniaceae
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
2400 species. As well as being economically important as a source of hardwoods, the group is of great phylogenetic interest because Laurales are among the oldest known flowering plants. Paleobotanical interest has focused especially on Chloranthaceae, Calycanthaceae, and Lauraceae, families with fossil records that go back to the Early and mid-Cretaceous (Drinnan et al., 1990
; Pedersen et al., 1991
; Herendeen, Crepet, and Nixon, 1993
; Crane, Friis, and Pedersen, 1994
; Crepet and Nixon, 1994
; Friis et al., 1994
; Friis, Crane, and Pedersen, 1997
; Eklund and Kvacek, 1998
). With the advent of molecular systematic data, however, it has become clear that Chloranthaceae do not belong in Laurales (contra Takthajan, 1973, 1997, and Thorne, 1974, in press
). More recently, attention has turned to another putative lauralean taxon, the monotypic Amborellaceae. Amborella trichopoda (Baillon, 1869
) was placed as sister to all other angiosperms or in a grade with Illicium, Austrobaileya, Schisandra, the Piperales, and Nymphaeaceae in the 18S rDNA trees of Soltis et al. (1997a)
, and in a clade with Austrobaileya, Illicium, and Nymphaeaceae in a combined analysis of rbcL and 18S sequences (Soltis et al., 1997b
; compare also Fig. 4B in Chase et al., 1993
). An analysis of combined atpB, rbcL, and 18S sequences again places Amborella as sister to all other angiosperms (M. Chase, D. Soltis, and P. Soltis, personal communication, 1998).
Prior to molecular studies, Amborellaceae and Chloranthaceae had been interpreted as close to Trimeniaceae (Endress and Sampson, 1983
; Philipson, 1993a
). Trimeniaceae are a family with four or five species in one or two genera, for which the first discovered species, Trimenia weinmanniifolia (Seemann, 1865
) and Piptocalyx moorei (Oliver in Bentham, 1870
; = Trimenia moorei (Oliver) Philipson), were regarded as probably belonging in Monimiaceae by Bentham and Hooker (1880)
because their flowers shared some traits with those of Xymalos. Xymalos is one of two Monimiaceae with numerous stamens but a single carpel. Monimiaceae at the time also encompassed Amborella. Similar to Chloranthaceae, Amborella and Trimenia have flower and inflorescence modifications typical of wind-pollination, such as inconspicuous flowers, single ovules, stigmas with large receptive surfaces, reduced or lacking tepals or petals, and separation of sexual function, making them similar to each other and difficult to homologize with animal-pollinated potential relatives. As will be shown here, Trimeniaceae do not belong in Laurales. Instead, Trimenia is closest to Illicium, Schisandra, and Austrobaileya.
Partly because the circumscription of Laurales (summarized in Table 1) is unclear, partly because their second-largest family, Monimiaceae, as traditionally conceived (e.g., Money, Bailey, and Swamy, 1950
; Philipson, 1993b
; Takhtajan, 1997
) is blatantly polyphyletic (Renner, 1998
), relationships of and within Laurales have remained obscure. Hallier (1905)
, who in his Natürliche (phylogenetische) System der Blütenpflanzen first interpreted the Laurales as a natural group, thought them characterized by perigynous flowers, that is, flowers with a well-developed fleshy receptacle enveloping the carpels. As circumscribed by Hallier, Laurales comprised three families, the Calycanthaceae, the Monimiaceae s.l. (sensu lato), into which Hallier had sunk another highly aberrant group, the monospecific Gomortegaceae (Reiche, 1896
), and the Lauraceae, which in Hallier's conception included the Hernandiaceae. Note that the most divergent member of the Calycanthaceae, Calycanthus (Idiospermum) australiensis, was first described in 1912, after Hallier's paper. Calycanthaceae and Monimiaceae were seen as related to Lauraceae, and the entire order was seen as derived from ("abstammend von") Magnoliales (Hallier, 1905
). The Magnoliales were characterized by hypogynous flowers, i.e., with the carpels exposed on the receptacle rather than embedded in it. Together, Laurales and Magnoliales formed the Polycarpicae (both had ancestrally numerous carpels per flower). Chloranthaceae do not have carpels embedded in the receptacle nor do they have numerous carpels, and Hallier (1905)
accordingly did not include them in the Laurales; instead he placed them with Piperaceae and Saururaceae in Piperales. He either had not seen material of Amborella, Piptocalyx, and Trimenia, all of which lack cup-shaped receptacles and thus might have been deemed misfits in Laurales, or else he was not concerned about intrafamilial variation in receptacle development. As it was, Trimeniaceae and Amborellaceae remained buried in Monimiaceae until 1917 and 1948 when they were raised to family rank by Gibbs and Pichon, respectively (which became widely accepted after the landmark Monimiaceae study by Money, Bailey and Swamy, 1950
). No morphologist excluded them from Laurales, however.
|
; Graybeal, 1998
). Taxon addition minimizes the effects of spurious groupings among clades rich in autapomorphies, i.e., relatively long branches, which are liable to produce long-branch-attraction artifacts, while large amounts of sequence data increase phylogenetic signal. Previous large-scale molecular studies of basal angiosperms have included a small number of core lauralean taxa. Thus, Qiu et al. (1993)
and the angiosperm-wide rbcL analysis of Chase et al. (1993)
both included Calycanthus, Chimonanthus, and Idiospermum of Calycanthaceae, Hernandia and Gyrocarpus of Hernandiaceae, Cinnamomum and Persea of Lauraceae, and Hedycarya of Monimiaceae s.str. (sensu stricto). The 18S analysis of Soltis et al. (1997a)
included Calycanthus and Sassafras. The combined rbcL + atpB analysis of Savolainen et al. (in press)
includes the same taxa as Chase et al. (1993)
plus an additional member of Monimiaceae (Kibara) and minus one genus each of Calycanthaceae and Hernandiaceae. Other studies that combined morphological characters and rDNA (Doyle, Donoghue, and Zimmer, 1994
) or rbcL (Nandi, Chase, and Endress, 1998
) have included, respectively, Calycanthus, Persea, and Hedycarya, and Chimonanthus. Lacking from all these analyses are Atherospermataceae, Gomortegaceae, Siparunaceae, and Trimeniaceae, as well as first-branching members of Lauraceae and Monimiaceae s.str. The present study addresses the circumscription and internal relationships of Laurales using a two-pronged approach. Broad rbcL sampling of all relevant clades is used to establish whether the perigynous-flowered basal angiosperm families form a monophyletic group (thus evaluating Hallier's concept of Laurales). Then, dense taxon sampling and data from five additional genome regions (the rpl16 intron and the trnT-trnL, trnL-trnF, atpB-rbcL, and psbA-trnH intergenic spacers) were used to resolve relationships within Laurales. Fifteen morphological characters that have traditionally been of key importance in the classification of Laurales were studied on the molecular trees a posteriori as well as being included directly in the analysis in a total evidence approach.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
50 genera in three tribes is represented by four genera (following the classification of van der Werff and Richter, 1996
). The Cryptocaryeae are represented by Cryptocarya, the Perseeae by Persea and Cinnamomum, and the Laureae by Litsea and Sassafras. The monophyly of Lauraceae is unquestioned (J. Rohwer, personal communication 1998, University of Mainz; A. Chanderbali, personal communication, 1998, University of Missouri-St. Louis), and ongoing molecular systematic work on the family by J. Rohwer and A. Chanderbali guided the choice of taxa to be included. In addition to 40 rbcL sequences from putative Laurales (see Table 2), 42 sequences were included to represent all lineages that have been suggested by earlier authors (Hallier, 1905
; Endress, 1972
; Thorne, 1974, in press
; Takhtajan, 1973, 1997
; Cronquist, 1981, 1988
; Kubitzki, 1993a
) or in published or ongoing molecular work (cf. Introduction) as potential lauralean outgroups: Magnoliales s.l. (sensu lato) (Annonaceae, Degeneriaceae, Eupomatiaceae, Himantandraceae, Magnoliaceae, Myristicaceae), Winterales (Canellaceae, Winteraceae), and monocots (Araceae). Because molecular systematic analyses have shown that Amborella groups with the "expanded Nymphaeales" (cf. Fig. 1 for members of this informal grouping), rbcL sequences were included of Austrobaileyaceae, Cabombaceae, Nymphaeaceae, Illiciaceae, and Schisandraceae. Note that Austrobaileyaceae have sometimes also been included in Laurales, e.g., by Melchior (1964)
, Smith (1972)
, Takhtajan (e.g., 1973
), Thorne (1974)
, and Walker (1976)
. Because of the unclear position of Chloranthaceae among basal angiosperms, representatives of Aristolochiaceae, Lactoridaceae, Piperaceae, and Saururaceae were also included to cover all lineages with which Chloranthaceae might group. The tree was rooted with seven rbcL sequences from Ephedra, Gnetum, and Welwitschia.
|
|
; the latter work accords the former Hernandia subgenus Hazomalania generic rank). Monimiaceae are represented by four genera from their two main lineages (Renner, 1998
, and unpublished data), and Siparunaceae are represented by one species each of their two genera (Renner, 1998
, and unpublished data). Lauraceae are represented by species from the same three tribes as in the broadscale rbcL study. Magnoliaceae and Myristicaceae are used as outgroups based on Hallier's morphological hypothesis outlined above, the results of Nandi, Chase, and Endress (1998
, fig. 4), and unpublished analyses of combined rbcL, atpB, and 18S sequences (M. Chase, personal communication, 1998, Royal Botanic Gardens Kew; M. Chase, D. Soltis, P. Soltis, unpublished data).
|
). The plant material was either ground in Eppendorf tubes with sterilized sea-sand or in porcelain mortars under liquid nitrogen. The rbcL gene was then amplified using a 26-nucleotide forward primer (1F, 5'-ATG TCA CCA CAA ACA GAA ACT AAA GC-3') and a 24-nucleotide reverse primer (1460R, CTT TTA GTA AAA GAT TGG GCC GAG). Primer sequences are those of Fay, Swensen, and Chase (1997)
. To amplify the entire gene, two internal primers were used in addition to the two mentioned above, namely 724R (3'-CAT GTA CCT GCA GTA GC-5') and 636F (5'-GCG TTG GAG AGA TCG TTT CT-3'); where the 636F primer did not work, a complement of the 724R primer was used instead. To detect sequencing errors, the rbcL sequences were translated into amino acids using the program SeqPup (which was also used to align sequences; below) and then compared with Fig. 4 in Kellogg and Juliano (1997)
, which lists variable and invariable sites in 499 seed plant rbcL sequences. Alternative amino acids found in Laurales all occurred at sites already known to be variable within angiosperms, and no stop codons were found in any of the sequences. A total of 1434 nucleotides, from positions 30–1395 of the rbcL exon, plus 39 base pairs (bp) following the exon were sequenced, but only the first 1331 nucleotides were used in the analyses because several of the sequences downloaded from GenBank ended at around position 1331.
Five plastid chloroplast intron and spacer regions were used, in addition to rbcL gene sequences, to resolve intralauralean relationships. The first of these was the large intron that interrupts the rpl16 gene in all angiosperms thus far investigated (cf. Kelchner and Clark, 1997
; Baum, Small, and Wendel, 1998
). Primers 1067F (5'-CTT CCT CTA TGT TGT TTA CG-3') and 18R (3'-GCT ATG CTT AGT GTG TGA CTC-5') designed by C. B. Asmussen (personal communication, 1997, University of Copenhagen) were used to amplify an 800-bp long region including the entire intron. The completed alignment (with gaps) comprised 1064 positions.
To amplify the trnT-trnL and trnL-trnF intergenic spacer regions, I used the universal primers a, b, e, and f of Taberlet et al. (1991)
. The trnT-trnL sequences ranged in length from 569 nucleotides (Illigera) to 895 nucleotides (Knema), with the Monimiaceae, Hernandiaceae, and Lauraceae sequences being similar to each other but differing strikingly from the remaining sequences near the 5' end of the locus. The first 212 nucleotides were therefore excluded from the analysis so that the completed alignment (472 positions with gaps) comprised only the second half of the spacer. The trnL-trnF data set included 365 nucleotides and the completed alignment comprised 451 positions.
The atpB-rbcL spacer (Golenberg et al., 1993
) was amplified using the forward primer ‘oligo 2’ of Manen, Natali, and Ehrendorfer (1994)
; a primer complementary to the rbcL forward primer 1F (sequence given above) was used for the reverse reaction. A total of 740 nucleotides were used in the analysis, and the completed atpB-rbcL alignment (with gaps) comprised 898 positions.
The 390-bp long psbA-trnH intergenic spacer was amplified using the psbA-F forward and trnH-R reverse primers designed by Sang, Crawford, and Stuessy (1997)
. Their forward primer amplifies the last 41 base pairs of the psbA gene, a region that varies little among the 25 members of Laurales. However, alignment difficulties occurred toward the 3' end of this rapidly evolving spacer, and therefore only the first 170 nucleotides of each sequence were used in the analysis. The completed psbA-trnH alignment with gaps was 186 positions long.
PCR amplification followed standard protocols. PCR products were cleaned either by running the entire product on a low-melting point agarose gel and then recovering the amplified DNA with the help of QIAquick gel extraction kits (QIAGEN, 1997
) or by using the QIAquick PCR purification columns directly without a prior gel purification step. Double-stranded PCR products were used as sequencing templates, and sequencing was done on an ABI 377 automated sequencer (University of Missouri—DNA Core Sequencing Facility). Usually (but not for all psbA-trnH spacer sequences; above), both strands of DNA were sequenced and used to generate a consensus sequence using Sequencher version 3.1 (GeneCodes Corporation, 1998). All alignment was done manually in Sequencher and/or in SeqPup version 0.6 (D. Gilbert, Indiana University, Bloomington, 1996). None of the insertions or deletions in the five intron and spacer sequence data sets were potentially informative at the between-family level, and gaps were therefore not used as characters. A total of 170 sequences, including a new Hedyosmum sequence, were generated for this study and its precursor (Renner, 1998
) and have been deposited in GenBank (see Tables 2 and 3 for accession numbers).
Morphology
The morphological, palynological, and karyological characters scored for Laurales (Table 4) were the same as in a previously published matrix (Renner, Schwarzbach, and Lohmann, 1997
), which includes descriptions of characters and their states, and lists literature and herbarium sources used. Of the taxa used in that study, two of the outgroups (Austrobaileyaceae and Winteraceae) and six of the nine genera of Monimiaceae s.str. were dropped because molecular data indicated they were irrrelevant to an analysis of intralauralean family relationships; on the other hand, Palmeria was added because it is among the first-branching Monimiaceae (Renner, 1998
). For the new matrix, the terminal taxa Atherospermataceae, Calycanthaceae, Hernandiaceae, and Lauraceae were replaced by their genera sampled in the molecular study, and Knema (Myristicaceae) was added as another outgroup. For character 1, vegetative phyllotaxy, I followed Nandi, Chase, and Endress (1998)
in combining the states "alternate" and "spiral-alternate" into one state, while for character 5, anther dehiscence, I added the third state, "opening by two laterally hinged valves," to account for variation within Monimiaceae and Hernandiaceae. Character states for Knema are taken from Kühn and Kubitzki (1993)
.
|
Bootstrap support (Felsenstein, 1985
) for each clade in the intralauralean analyses was estimated based on 100 or 1000 replications, using the closest taxon addition and TBR branch swapping options. For the broadscale rbcL analysis, relative clade support was assessed via two parsimony jackknife analyses, each with 10 000 replicates (Farris et al., 1996 [1997]
). Bremer support (Bremer, 1988, 1994
), also called the decay index, was calculated as outlined in Bremer (1994)
.
Morphological characters were investigated, using MacClade 3.0, by optimizing most parsimonious state assignments onto the best molecular trees of Laurales found (Maddison and Maddison, 1992
). In a second step, they were appended to the combined molecular matrix and included in a heuristic analysis to investigate their effect on clade support.
Data concatenation and ambiguity coding
Parsimony analyses of the six data sets followed by bootstrapping showed that there were no statistically well-supported ("hard") incongruencies among the topologies based on these data, and they were therefore combined. (See the Results section: Effects of amount of data and missing data, for further exploration of individual data sets.) For the combined 6-genome regions-28-taxon analysis, sequences from the same species (usually from a single total DNA extract) were spliced together with the following exceptions (Tables 2, 3): a Doryphora aromatica rbcL sequence was combined with rpl16, trnT-trnL, trnL-trnF, atpB-rbcL, and psbA-trnH sequences from Doryphora sassafras. Doryphora comprises just these two species. In the Lauraceae, spacer and intron sequences from Beilschmiedia obovata and Litsea glaucescens were supplemented by rbcL sequences from Cryptocarya obovata and Litsea japonica, respectively. Cryptocarya and Beilschmiedia are both members of the Cryptocaryeae. For Siparuna, four sequences from S. aspera were combined with one from S. lepidota and one from S. guianensis. For Laurelia sempervirens and Glossocalyx longicuspis, the sequence to be spliced came from their respective closest relatives, which were also in the data set (compare Table 3). Specifically, the genus Laurelia (Atherospermataceae) comprises two species that grouped with each other in the rpl16 data set (63% bootstrap support), while the other four data sets contained too few informative nucleotide changes to yield statistically well-supported groupings within atherosperms with the exception of an Atherosperma-Nemuaron clade that appeared in two of the data sets. Basing myself mainly on the rpl16 data, I decided to use the L. novae-zelandiae trnT-trnL sequence to complement the five other sequences from L. sempervirens rather than dropping L. sempervirens from the analysis (see below for an alternative approach). Glossocalyx is a monotypic genus that based on morphological, rbcL, and trnL-trnF data is sister to Siparuna (Renner, Schwarzbach, and Lohmann, 1997
; Renner, 1998
); a Siparuna trnT-trnL sequence was used to complement the Glossocalyx set of sequences.
In two cases, I decided against matching and fusion of terminals. Gomortegaceae comprise a single species, Gomortega nitida, for which no trnT-trnL sequence could be generated. Likewise, Atherosperma comprises a single species from which I was unable to amplify trnL-trnF. Both sequences were coded as missing ("nnnn") and then spliced in with the remaining five Gomortega and Atherosperma sequences, which added 472 and 451 ambiguous sites, respectively (out of 4402 total nucleotide sites used for each taxon). These 923 ambiguous sites represent 0.74% of all nucleotide positions (923 of 28 x 4402 = 123 256 positions).
I explored the following alternative approaches to the problem of missing sequences. First, the two sequences that contained missing characters, Gomortega and Atherosperma, were excluded from the analysis. Second, the two data sets containing "holes" that could not be filled by splicing (trnT-trnL for Gomortega; trnL-trnF for Atherosperma) were sequentially and simultaneously excluded.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Structure of the Laurales
The matrix for the 25 core lauralean taxa (plus three outgroups) included 4402 characters of which 898 were informative. An unconstrained search yielded four equally parsimonious trees of length 2569 (including all characters), with a CI of 0.743 and a RI of 0.741, located in a single island found in all 1000 random taxon addition replicates. The four trees, which differed from each other only in the placement of the three first-branching genera of Atherospermataceae (Fig. 2), were rooted at the branch between Magnoliales and the remaining taxa based on published and unpublished analyses that show Magnoliales and Laurales as sister groups (Nandi, Chase, and Endress, 1998
, fig. 4; Magnoliales here are seen as including Annonales and Myristicaceae; M. Chase, personal communication, 1998, Royal Botanic Gardens Kew; M. Chase, D. Soltis, and P. Soltis, unpublished data).
|
Effects of amount of data and missing data
Of the six molecular data sets, four by themselves yield statistically supported resolution at the family level (>50% bootstrap support): rpl16 and atpB-rbcL support the Siparunaceae (atherosperm-Gomortega) clade (at 82 and 71%, respectively) as well as the atherosperm-Gomortega clade (at 74 and 60%, respectively), while rbcL supports only the atherosperm-Gomortega sister-group relationship (at 54%). Rpl16 and atpB-rbcL also both support the node making Calycanthaceae sister to all remaining families (at 85 and 58%, respectively), and atpB-rbcL and trnT-trnL both support the Hernandiaceae-Monimiaceae-Lauraceae clade (at 59 and 77%, respectively). No "hard" topological conflicts among trees based on individual or variously combined data sets were found.
To explore the effects of absolute numbers of informative characters included and of particular sets of data, such as the trnT-trnL or trnL-trnF matrices that each included 450 ambiguous sites due to lacking sequences (Materials and Methods), I sequentially added or excluded data sets. These searches and bootstrap analyses were run using the closest taxon addition option rather than random taxon addition. The exclusion of the trnT-trnL data (472 characters of which 176 were parsimony informative) yielded six equally parsimonious trees (L = 2112, CI = 0.746, RI = 0.738), the strict consensus of which was identical to the tree resulting from the full analysis, except that the Hernandiaceae-(Lauraceae-Monimiaceae) clade collapsed. When bootstrapped, support for the remaining clades remained virtually the same as in the full analysis. The exclusion of the trnL-trnF data (451 characters of which 122 were parsimony informative) yielded four equally parsimonious trees (L = 2208, CI = 0.750, RI = 0.746). Their strict consensus was identical to the topology found in the full analysis except for intra-atherosperm relationships, which were less resolved; clade support was almost unaffected. The exclusion of the atpB-rbcL data (898 characters of which 172 were parsimony informative) resulted in three equally parsimonious trees (L = 2015, CI = 0.745, RI = 0.744) with the same general topology as before, but support values for the Hernandiaceae-(Lauraceae-Monimiaceae) and Lauraceae-Monimiaceae clades dropped to 55%. The exclusion of the rbcL data (1331 characters of which 135 were informative) resulted in four shortest trees (L = 2114, CI = 0.766, RI = 0.758) with the same topology as found in the full analysis except that the Lauraceae-Monimiaceae node now had only 49% bootstrap support; support for the remaining clades remained essentially unchanged. The exclusion of the rpl16 data (1064 characters of which 250 were parsimony informative) resulted in lower support for the node making Calycanthaceae sister to all other Laurales (49%) and for the Monimiaceae-Lauraceae node (48%). Finally, exclusion of the psbA-trnH data (186 characters of which 43 were parsimony informative) had almost no effect on topology or bootstrap values.
To explore the effects of the 472 and 451 ambiguous sites contained in the concatenated sequences of Gomortega and Atherosperma, an analysis was run with both these taxa excluded. This yielded two shortest trees of the same topologies as found in the full analysis (L = 2499, CI = 0.749, RI = 0.742). Bootstrap percentages remained virtually unchanged. Next, I re-included Gomortega and Atherosperma and instead excluded Illigera and Hernandia because these and other Hernandiaceae are among the longest branches in the analysis (Fig. 2). This resulted in two shortest trees of the same general topology as found before except that the Monimiaceae no longer grouped with the Lauraceae but instead with the remaining two Hernandiaceae (43% bootstrap support). The exclusion of any two (of four) Monimiaceae or two (of three) Calycanthaceae, by contrast, did not significantly affect topology or clade support.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
, who found that when the total number of characters is held constant, accuracy is much higher if the characters are distributed across a larger number of taxa. Similar conclusions, namely that denser species sampling greatly improves the ability of an analysis to reconstruct phylogeny, have been reached by others (Lecointre et al., 1993, 1994
; Hillis, 1996
; Purvis and Quickie, 1997
). Although I did not address the relationship between adding taxa vs. adding characters systematically, the experimental exclusion or inclusion of taxa vs. characters showed the sometimes striking impact of lowering taxon density. Thus, the exclusion of two Hernandiaceae, a group particularly rich in nucleotide changes and thus easily forming long branches, had the same impact as an "across the board" exclusion of up to a quarter of all informative characters (cf. Results). These results underline the importance of sampling taxa specifically so as to enhance signal by breaking up long branches, which in the Laurales involves sampling several Hernandiaceae.
Circumscription of the Laurales
The broadscale rbcL analysis shows that a monophyletic order Laurales includes the following seven lineages (using available family names): Atherospermataceae, Calycanthaceae (incl. Idiospermaceae), Gomortegaceae, Hernandiaceae, Lauraceae, Monimiaceae, and Siparunaceae. The best contender for a morphological synapomorphy for these seven families is their perigynous flowers, in which the carpel(s) is/are often deeply embedded in a fleshy receptacle that may or may not become woody in fruit (Hallier, 1905
; Endress and Igersheim, 1997
). Other characters often listed as typically lauralean, such as uniovulate carpels, unilacunar nodes, opposite leaves, and inaperturate pollen, while being useful for distinguishing Laurales from their presumed closest relatives (Magnoliales?), vary within Laurales and/or are also found in a number of other basal angiosperms.
In the single most-parsimonious tree resulting from the combined rbcL + morphology analysis of Nandi, Chase, and Endress (1998)
, Laurales appear as sister to a clade comprising Myristicaceae, Magnoliales, and Annonales (= Magnoliales s.l.), albeit with less than 50% bootstrap support. A problem in this study that conceivably contributed to this low support is that the Laurales were circumscribed as including Calycanthaceae, Monimiaceae s.l., Lauraceae, Hernandiaceae s.str., and Gyrocarpaceae (mislabelled as "Gyrostemonaceae" on p. 199; P. Endress, personal communication, September 1998). Gomortegaceae were not included in the study. For the molecular matrix the highly autapomorphic Calycanthaceae (compare the length of their subtending branch in Fig. 2) were chosen as a place holder, while for the morphological matrix presumed apomorphic tendencies were scored across the entire order. Thus, Laurales were scored as having valvate anther opening (character 230, absent/present), inaperturate or monosulcate pollen (character 129 with four states), a reticulate sexine (character 135 with four states), marginal placentation (character 241 with four states), and oil or proteins stored in the endosperm (character 161 with four states). These characters are variable within Laurales (as are many other characters scored as monomorphic) and do not occur together in any of the families.
Families excluded from Laurales
Of the excluded families (see Results) none was morphologically strongly linked with the core Laurales. Generally, the inclusion of Amborellaceae, Chloranthaceae, Trimeniaceae, and sometimes Austrobaileyaceae in Laurales was based on such widespread lower angiosperm traits as unilacunar nodes, opposite leaves, or a spiral floral phyllotaxy, and in the case of the first three also the single apical ovule. Typically, Chloranthaceae were thought to be closest to Trimeniaceae and Amborellaceae (Endress and Sampson, 1983
; Sampson and Endress, 1984
; Endress, 1987
), which in turn remained in Laurales mostly because of tradition (they had originally been described as Monimiaceae; see Introduction). In a strikingly inconsistent decision, Cronquist (1981)
excluded Chloranthaceae from Laurales but left Amborella and Trimenia in the order, stating "[Amborella] is clearly a member of the Laurales, in which its primitively vesselless wood, alternate leaves, essentially hypogynous flowers, several carpels, abundant endosperm, and stamens dehiscent by slits mark it as an archaic type. The virtual absence of ethereal oil cells is anomalous in the group..." Takhtajan (1997)
adopts the same line, adding "Trimeniaceae also approach Chloranthaceae, which I prefer to put in a separate order." In keeping with a phenetic approach to classification, he excluded Calycanthaceae from Laurales, placing them in a monotypic order next to Laurales.
Although earlier work had stressed the overall similarity between Chloranthaceae, Trimeniaceae, and Amborellaceae (see especially Endress and Sampson, 1983
), these families do not group together based on rbcL. Rather, Amborella groups with Nymphaeaceae/Cabombaceae, and Trimenia with Schisandra/Illicum/Austrobaileya, while Chloranthaceae occupy an isolated position (Fig. 1). Sequences of 18S and atpB are currently being produced for Trimenia to add this taxon to combined rbcL + atpB + 18S analyses.
Relationships within Laurales
Within Laurales, the oldest split is between Calycanthaceae and the remaining six families, which in turn form two clades, the Siparunaceae (Atherospermataceae-Gomortegaceae) and the Hernandiaceae (Monimiaceae-Lauraceae). Calycanthaceae, Lauraceae, and Hernandiaceae are clearly monophyletic. Whether or not the single species of Idiospermum is recognized as a family (Idiospermaceae) depends on taste (Stevens, 1997
; Backlund and Bremer, 1998
) as does the recognition of Gyrocarpaceae. As shown by Shutts (1960)
, Kubitzki (1969)
, and others, the two subgroups of Hernandiaceae, Gyrocarpus and Sparattanthelium on the one hand and Hernandia and Illigera on the other, are divided by a major morphological gap, which historically led to their repeated assignment to different higher taxa.
Several morphological characters are consistent with the molecular results (as is also apparent from the morphological analysis of Donoghue and Doyle [1989
] once Amborellaceae, Trimeniaceae, Chloranthaceae, and Austrobaileyaceae are removed from the Laurales). The deep split between Calycanthaceae and the remaining families is paralleled by at least three characters. (1) Calycanthaceae have two lateral ovules per carpel (of which only the lower one forms a mature embryo sac; Schaeppi, 1953
; Nicely, 1965
), while all other families have a single ovule per carpel. (2) Calycanthaceae have disulculate columellate pollen, while the remaining families with one exception have inaperturate pollen with rather thin, often spinulose exines. The disulculate state in Calycanthaceae may have been acquired relatively recently in their evolution, since a fossil flower attributed to that family (Friis et al., 1994
) has monosulcate pollen. Another difference between extant Calycanthaceae and the fossil is that modern calycanth pollen has a nearly closed tectum, while the fossil has a reticulate tectum. If the ancestral lauralean pollen is assumed to have been monosulcate with a thick exine, the Calycanthaceae would have retained the thick exine but modified the single sulcus into two sulculi, while their sister lineage would have lost both the thick exine and the sulcus. Inaperturate pollen with thin, granular-spinulose exines as found in the sister clade to calycanths is rare in basal angiosperms and therefore represents a clear synapomorphy. (3) A third character, paralleling the ovule and pollen characters, is the presence or absence of floral nectary glands. Calycanthaceae are beetle-pollinated and devoid of nectaries (except for isolated nectarogeneous fields on the inner tepals of Chimonanthus; Vogel, 1998
), whereas large nectary glands on the filament bases characterize Lauraceae, Hernandiaceae, first-branching Monimiaceae, Atherospermataceae, and Gomortegaceae. The only lineages without filament glands are Palmeria and other higher Monimiaceae and Siparunaceae (Renner, Schwarzbach, and Lohmann, 1997
), both apparently cases of secondary loss. Glands were likely lost in conjunction with increasing closure of flowers, which as far as known are pollinated by small non-nectar feeding beetles in Monimiaceae (Lorence, 1985
) and by ovipositing (again, non-nectar foraging) gall midges in Siparunaceae (Feil and Renner, 1991
; Feil, 1992; see Renner, Schwarzbach, and Lohmann [1997]
for details on pollen tube growth in Siparuna). Lorence (1985
, p. 66) observed flies as pollinators in some species of Tambourissa that offer mucilaginous stigma exudates as a reward (in addition to pollen) and hypothesized a secondary switch from beetle to fly pollination.
In summary, the molecular data support Kubitzki's (1993c)
view that "the Calycanthaceae are certainly the most aberrant element, yet not as primitive as claimed by Loconte and Stevenson (1991)
, but rather autapomorphic."
The Monimiaceae-Lauraceae-Hernandiaceae clade
These three families formed a moderately well-supported clade (Fig. 2) that morphologically appears supported mainly by the apical position of their ovules (the ovules are inserted at or near the locule apex). Their sister clade settled on basal ovules (ovules inserted at or near the base of the locule) except for Gomortega (below), while Calycanthaceae have two lateral ovules. Some earlier workers have stressed the close relationship of Monimiaceae, Lauraceae, and Hernandiaceae (Shutts, 1960
; Takhtajan, 1973
; Rohwer, 1993
), but assessments of relationships were handicapped by the prevailing broad concept of Monimiaceae, which greatly confused the picture of character variability in this family. As long as Siparunaceae and Atherospermataceae are included in Monimiaceae, the last combine basal and apical ovules, tetrasporangiate and disporangiate stamens with anthers opening by slits or apically hinged valves, wood with narrow rays and wood with very broad rays, and inaperturate thin-exined pollen as well as meridionosulcate columellate pollen.
The sister-group relationship between Lauraceae and Monimiaceae (Fig. 2) found here implies three parallelisms in Lauraceae and Hernandiaceae or reversals in Monimiaceae. First, the fixed unicarpellate condition found in both families must have evolved independently. Elsewhere within Laurales, solitary carpels evolved in Calycanthaceae (Idiospermum) and within higher Monimiaceae (Xymalos, Hennecartia), but all remaining groups, including Monimiaceae, have numerous carpels per flower. Second, the fixation of stamen and tepal numbers in flowers of Lauraceae (which are 3-merous) and Hernandiaceae [which are 3–4(–6) or 4–8-merous] likely evolved independently, and future morphological analyses should no longer treat the "fixed number of floral parts" as a single character as done here. Third, the absence of endosperm from the mature seeds of Lauraceae and Hernandiaceae must have evolved independently. Endosperm is also lost in Calycanthaceae.
The molecular phylogeny implies that disporangiate two-valvate anthers in Lauraceae and Hernandiaceae evolved independently, confirming the traditional view that tetrasporangiate stamens are basal in Lauraceae (Rohwer, 1993
; Crane, Friis, and Pedersen, 1994
). Tetrasporangiate stamens are plesiomorphic in the order, and they open by slits in two of the seven families of Laurales, Calycanthaceae and Monimiaceae. Notably, Monimiaceae anthers may dehisce longicidally or rarely transversally and may have short lateral incisions at the top and bottom of the slits, which results in a saloon door-like opening of the thecae (Baillon, 1869
, fig. 339; Endress and Hufford, 1989
, figs. 83–84). This can be interpreted as either a tendency towards valvate anther opening or a remnant of a former valvate dehiscence. The disporangiate anthers of Hernandiaceae dehisce in various ways; Gyrocarpus and Sparattanthelium have apically hinged valves, while Illigera and Hernandia (except the single species of H. subgenus Hazomalania) have laterally hinged valves. In Hazomalania, according to Kubitzki (1993b)
"the most primitive element of the family," valves are hinged apically (work on a molecular phylogeny of Hernandiaceae is in progress; Renner, unpublished data). Friis et al. (1994)
have described a calycanthaceous fossil flower, Virginianthus, that had tetrasporangiate anthers opening by laterally hinged valves. It is thus possible that valvate dehiscence (with the valves laterally hinged) is ancestral in Laurales and was lost in Calycanthaceae and Monimiaceae, however, this interpretation hinges on the correct placement of Virginianthus.
The Atherospermataceae-Gomortegaceae-Siparunaceae clade
This clade receives 91% bootstrap support (Fig. 2) and agrees with several nonmolecular characters. Thus, all three families have disporangiate two-valvate stamens. Atherospermataceae and Siparunaceae have a chromosome number of n = 22 (Philipson, 1993b
), while Gomortega nitida has n = 21 (Goldblatt, 1976
; under the synonymous name G. keule). The ancestral condition in the clade is basal ovules (in contrast to the condition in its sister clade). The gynoecium of Gomortega is unique in Laurales in being syncarpous (two to three merous) and inferior (Leinfellner, 1968
; Endress and Igersheim, 1997
), and it may be surmised that the evolution of apical ovules in Gomortega occurred with the fusion and embedding of the ovary. Gomortegaceae have traditionally been placed close to, or even in, Monimiaceae. For example, Philipson (1987)
stated "The coherence of the Monimiaceae and the gap which separates them from other families of Laurales justify its continued recognition as a single family. Gomortegaceae, with a single species, represents the sole possible reservation. If its syncarpous inferior gynoecium is not considered sufficient to separate it from the Monimiaceae, then its union with that family would be a solution preferable to the fragmentation of the Monimiaceae." By contrast Rohwer (1993)
thought "Gomortega looks definitely lauraceous, and at least a cross-section of the ovary is needed to reveal features incompatible with Lauraceae." That the true relations of Gomortega lie with Atherospermataceae was first recognized by Schodde (1969, 1970
) based on a phenetic analysis of some 40 morphological characters. A cladistic analysis of morphological characters (Renner, Schwarzbach, and Lohmann, 1997
) also found Atherospermataceae and Gomortegaceae as sister groups, but support for this was weak. Among the few morphological or anatomical features suggesting a sister-group relationship between Atherospermataceae and Gomortega is the peculiar structure of their sieve tube plastids (Behnke, 1981, 1988
). Most Monimiaceae s.str. and all Lauraceae, Hernandiaceae, and Siparunaceae have protein-containing plastids of the same kind as found in most Magnoliaceae and many Myristicaceae (Psc-type plastids in the notation of Behnke, 1988
). By contrast, Calycanthaceae, Atherospermataceae, and Gomortega have plastids of the rare Pcsf-type (Behnke, 1988
).
The DNA topology implies a return from inaperturate pollen with granulate exines, as found in most Laurales including Siparunaceae and Gomortega, to tectate-columellate exines and aperturate pollen along the stem lineage of Atherospermataceae. Transitions between granular and columellate tectal and/or supratectal sculptures have also occurred in the presumed closest relative of Laurales, the Magnoliales, in Annonaceae, Magnoliaceae, and Myristicaceae (Walker, 1976
; Doyle and Le Thomas, 1994, 1996
). Interestingly, Hesse and Kubitzki (1983)
described Gomortega pollen as having short columellae, which might indicate a tendency towards thickening of the exine already along the stem lineage of atherosperms and Gomortega. The apertures in atherosperms are of the meridionosulcate or disulcate kind, which is very rare (Sampson, 1996
, 1997) and possibly a hint of their de novo evolution.
Siparunaceae have many autapomorphic characters, such as disporangiate anthers that open by a single flap (a unique state in the Magnoliidae; Endress and Hufford, 1989
, p. 77), unitegmic ovules (Renner, Schwarzbach, and Lohmann, 1997
), and flowers closed by a roof.
Taxonomic conclusions
Laurales comprise seven major lineages for which family names are available. Trimeniaceae (like Amborellaceae, Austrobaileyaceae, and Chloranthaceae) are not part of Laurales. The recognition of Gyrocarpaceae and Idiospermaceae as families separate from Hernandiaceae and Calycanthaceae, respectively, is purely a matter of ranking, whereas the Monimiaceae s.l. are polyphyletic and need to be dismantled, as already argued by Schodde (1970)
. The information provided by this study should allow a clearer interpretation of the morphological and chemical characters found in Laurales (along the lines of Nandi, Chase, and Endress, 1998
) and help to represent the group appropriately in the broader scale studies needed to link it to putative sister groups, such as the Magnoliales.
|
FOOTNOTES |
|---|
2
E-mail: biosrenn{at}admiral.umsl.edu
|
LITERATURE CITED |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Angiosperm Phylogeny Group. 1998 An ordinal classification for the families of flowering plants. Annals of the Missorui Botanical Garden 85: 531–553.
Baillon, H. E. 1869 Histoire des Plantes, vol. 1. L. Hachette et Cie, Paris.
Backlund, A., and K. Bremer. 1998 To be or not to be—principles of classification and monotypic plant families. Taxon 47: 391–400.[CrossRef][ISI]
Baum, D. A., R. L. Small, and J. F. Wendel. 1998 Biogeography and floral evolution of baobabs (Adansonia, Bombacaceae) as inferred from multiple dat sets. Systematic Biology 47: 181–207.[CrossRef][ISI][Medline]
Behnke, H.-D. 1981 Sieve-element characters. Nordic Journal of Botany 1: 381–400.
———. 1988 Sieve-element plastids, phloem protein, and evolution of flowering plants: III. Magnoliidae. Taxon 37: 699–732.[CrossRef][ISI]
Bentham, G. 1870 Flora Australiensis, vol. V. Lovell Reeve & Co., London.
———, and J. D. Hooker f. 1880 Genera Plantarum, vol. 3. Lovell Reeve & Co., London.
Bremer, K. 1988 The limits of amino acid sequence data in angiosperm phylogenetic reconstruction. Evolution 42: 795–803.[CrossRef][ISI]
———. 1994 Branch support and tree stability. Cladistics 10: 295–304.[CrossRef][ISI]
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 Journal of Systematic Botany 11: 215–229.[CrossRef]
Crane, P. R., E. M. Friis, and K. R. Pedersen. 1994 Palaeobotanical evidence on the early radiation of magnoliid angiosperms. Plant Systematics and Evolution, Supplement 8: 51–72.
Crepet, W. L., and K. C. Nixon. 1994 Flowers of Turonian Magnoliidae and their implications. Plant Systematics and Evolution, Supplement 8: 73–91.
Cronquist, A. 1981 An integrated system of classification of flowering plants. Columbia University Press, New York, NY.
———. 1988 The evolution and classification of flowering plants. New York Botanical Garden, Bronx, NY.
Donoghue, M. J., and J. A. Doyle. 1989 Phylogenetic analysis of angiosperms and the relationships of Hamamelidae. In P. R. Crane and S. Blackmore [eds.], Evolution, systematics, and fossil history of the Hamamelidae, vol. 1, 17–45. Clarendon Press, Oxford.
Doyle, J. A., M. J. Donoghue, and E. A. Zimmer. 1994 Integration of morphological and ribosomal RNA data on the origin of angiosperms. Annals of the Missouri Botanical Garden 81: 419–450.[CrossRef][ISI]
———, and A. Le Thomas. 1994 Cladistic analysis and pollen evolution in Annonaceae. Acta Botanica Gallica 141: 149–170.[ISI]
———, and ———. 1996. Phylogenetic analysis and character evolution in Annonaceae. Bulletin du Muséum national d'Histoire naturelle, section B, Adansonia 18: 279–334.
Drinnan, A. N., P. R. Crane, E. M. Friis, and K. R. Pedersen. 1990 Lauraceous flowers from the Potomac Group (mid-Cretaceous) of eastern North America. Botanical Gazette 151: 370–384.[CrossRef][ISI]
Duvall, M. R., M. T. Clegg, M. W. Chase, W. D. Clark, W. J. Kress, H. G. Hills, L. E. Eguiarte, J. F. Smith, B. S. Gaut, E. A. Zimmer, and G. H. Learn, Jr. 1993 Phylogenetic hypotheses for the monocotyledons constructed from rbcL sequence data. Annals of the Missouri Botanical Garden 80: 607–619.[CrossRef][ISI]
Eklund, H., and J. Kvacek. 1998 Lauraceous inflorescences and flowers from the Cenomanian of Bohemia (Czech Republic, Central Europe). International Journal of Plant Sciences 159: 668–686.[CrossRef]
Endress, P. K. 1972 Zur vergleichenden Entwicklungsmorphologie, Embryologie und Systematik bei Laurales. Botanische Jahrbücher für Systematik 92: 331–428.
———. 1987 The Chloranthaceae: reproductive structures and phylogenetic position. Botanische Jahrbücher für Systematik 109: 153–226.
———, and L. D. Hufford. 1989 The diversity of stamen structures and dehiscence patterns among Magnoliidae. Botanical Journal of the Linnean Society 100: 45–85.
———, and T. Igersheim. 1997 Gynoecium diversity and systematics of the Laurales. Botanical Journal of the Linnean Society 125: 93–168.[CrossRef]
———, and F. B. Sampson. 1983 Floral structure and relationships of the Trimeniaceae (Laurales). Journal of the Arnold Arboretum 64: 447–473.[ISI]
Farris, J. S., V. A. Albert, M. Källersjö, D. Lipscomb, and A. G. Kluge. 1996 [1997] Parsimony jackknifing outperforms neighbor-joining. Cladistics 12: 99–124.[CrossRef][ISI]
Fay, M. F., S. M. Swensen, and M. W. Chase. 1997 Taxonomic affinities of Medusagyne oppositifolia (Medusagynaceae). Kew Bulletin 52: 111–120.[CrossRef]
Feil, J. P. 1992 Reproductive ecology of dioecious Siparuna (Monimiaceae) in Ecuador—a case of gall midge pollination. Botanical Journal of the Linnean Society 110: 171–203.
———, and S. S. Renner. 1991 The pollination of Siparuna (Monimiaceae) by gall-midges (Cecidomyiidae): another likely ancient association. American Journal of Botany 78: 186.
Felsenstein, J. 1985 Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39: 783–791.[CrossRef][ISI]
Friis, E. M., P. C. Crane, and K. R. Pedersen. 1997 Fossil history of magnoliid angiosperms. In K. Iwatsuki and P. H. Raven [eds.], Evolution and diversification of land plants, 121–156. Springer-Verlag, Tokyo.
———, H. Eklund, K. R. Pedersen, and P. R. Crane. 1994 Virginianthus calycanthoides gen. et sp. nov.: a calycanthaceous flower from the Potomac Group (Early Cretaceous) of eastern North America. International Journal of Plant Sciences 155: 772–785.[CrossRef][ISI]
Gibbs, L.S. 1917 A contribution to the phytogeography and flora of the Arfak mountains. Taylor and Francis, London.
Goldblatt, P. 1976 Chromosome number in Gomortega keule. Annals of the Missouri Botanical Garden 63: 207–208.
Golenberg, E. M., M. T. Clegg, M. L. Durbin, J. Doebley, and D. P. Ma. 1993 Evolution of a noncoding region of the chloroplast genome. Molecular Phylogenetics and Evolution 2: 52–64.[CrossRef][Medline]
———, D. E. Giannasi, M. T. Clegg, C. J. Smiley, M. Durbin, D. Henderson and G. Zurawski. 1990 Chloroplast DNA sequences from a Miocene Magnolia species. Nature 344: 656–658.[CrossRef][Medline]
Graybeal, A. 1998 Is it better to add taxa or characters to a difficult phylogenetic problem? Systematic Biology 74: 9–17.
Hallier, H. 1905 Ein zweiter Entwurf des natürlichen (phylogenetischen) Systems der Blütenpflanzen. Berichte der Deutschen Botanischen Gesellschaft 23: 85–91.
Hasebe, M., R. Kofuji, M. Ito, M. Kato, K. Iwatsuki, and K. Ueda. 1992 Phylogeny of gymnosperms inferred from rbcL gene sequences. Botanical Magazine, Tokyo 105: 673–679.[CrossRef][ISI]
Herendeen, P. S., W. L. Crepet, and K. C. Nixon. 1993 Chloranthus-like stamens from the Upper Cretaceous of New Jersey. American Journal of Botany 80: 865–871.
Hesse, M., and K. Kubitzki. 1983 The sporoderm ultrastructure in Persea, Nectandra, Hernandia, Gomortega and some other lauralean genera. Plant Systematics and Evolution 141: 299–311.[CrossRef][ISI]
Hillis, D. M. 1996 Inferring complex phylogenies. Nature 383: 130–131.[CrossRef][Medline]
Kelchner, S. A., and L. G. Clark. 1997 Molecular evolution and phylogenetic utility of the chloroplast rpl16 intron in Chusquea and the Bambusoideae (Poaceae). Molecular Phylogenetics and Evolution 8: 385–397.[CrossRef][ISI][Medline]
Kellogg, E. A., and N. D. Juliano. 1997 The structure and function of RuBisCo and their implications for systematic studies. American Journal of Botany 84: 413–428.[Abstract]
Kubitzki, K. 1969 Monographie der Hernandiaceen. Botanische Jahrbücher für Systematik 89: 78–148.
———. 1993a Introduction. In K. Kubitzki, J. G. Rohwer, and V. Bittrich [eds.], The families and genera of vascular plants, vol. 2, 1–12. Springer Verlag, Berlin.
———. 1993b Hernandiaceae. In K. Kubitzki, J. G. Rohwer, and V. Bittrich [eds.], The families and genera of vascular plants, vol. 2, 334–338. Springer Verlag, Berlin.
———. 1993c Calycanthaceae. In K. Kubitzki, J. G. Rohwer, and V. Bittrich [eds.], The families and genera of vascular plants, vol. 2, 197–200. Springer Verlag, Berlin.
Kühn, U., and K. Kubitzki. 1993 Myristicaceae. In K. Kubitzki, J. G. Rohwer, and V. Bittrich [eds.], The families and genera of vascular plants, vol. 2, 457–467. Springer Verlag, Berlin.
Lecointre, G., H. Philippe, H. L. Van Le, and H. Le Guyader. 1993 Species sampling has a major impact on phylogenetic inference. Molecular Phylogenetics and Evolution 2: 205–224.[CrossRef][Medline]
———, ———, ———, and ———. 1994 How many nucleotides are required to resolve a phylogenet
