| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics |
2Royal Botanic Gardens, Mrs Macquaries Road, Sydney, NSW 2000, Australia; 3School of Biological Science, University of New South Wales, Sydney, NSW 2052, Australia; and 4School of Tropical Biology, James Cook University, PO Box 6811, Cairns, Queensland 4870, Australia.
Received for publication January 5, 2001. Accepted for publication May 8, 2001.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: matK; • molecular phylogeny • morphology • Myrtaceae
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
who concluded, from a cladistic analysis based on morphological and anatomical characters, that these subfamilies must be abandoned. In their initial analysis of Myrtales, they identified a clade containing Myrtaceae plus Heteropyxis and Psiloxylon, the last two of which are often placed in monogeneric families. Using Heteropyxis and Psiloxylon as outgroups, they then analyzed the Myrtaceae, exemplified by 14 terminal taxa. One particularly interesting conclusion emerged from this analysis: Acmena, Syzygium, and related genera constituted a group of fleshy-fruited genera that was separate from the other genera usually assigned to the Myrtoideae. This distinction had already been flagged by strong evidence from wood and bark anatomy (Ingle and Dadswell, 1953
; Chattaway, 1959
).
However, one limitation of this analysis was the marked disparity in the size of their terminal taxa: four were monotypic genera and the remainder were groups of from two to more than ten, often large, genera. The authors were aware that their emphasis on problematic taxa led to their 14 terminal taxa being "of unequal status" (Johnson and Briggs, 1984
, p. 733). The most notable example of this was the lumping of a heterogeneous collection of taxa into a single terminal taxon, the Metrosideros group, which they themselves admitted would be better treated as smaller units.
The present study began as an attempt to score a wider range of genera from this assemblage. In the process, we reviewed the characters used by Johnson and Briggs (1984)
, removing autapomorphies and the weighted scoring. However, the cladogram produced by analyzing the revised data showed little resolution due to extensive homoplasy (Wilson et al., 1994
).
Conti, Litt, and Sytsma (1996)
used rbcL sequence data to study relationships in the order Myrtales. Because of the breadth of the study, only five genera of Myrtaceae were included plus Heteropyxis and Psiloxylon. One of the surprising results was that a strongly supported clade of representatives of the family Vochysiaceae appeared to be the sister-group of Myrtaceae sensu stricto (s.s.). However, in a more detailed analysis of the same data, Conti et al. (1997)
showed that this relationship was not strong and their maximum likelihood analysis showed Vochysiaceae to be sister to Myrtaceae sensu lato (s.l.), including Heteropyxis and Psiloxylon. The more recent rbcL analysis of the eudicots by Savolainen et al. (2000)
also supported the latter topology.
Gadek, Wilson, and Quinn (1996)
presented a preliminary molecular analysis of Myrtaceae and its near relatives based on partial sequences of the chloroplast gene matK. This study, which concentrated on the position of Heteropyxis and Psiloxylon, included 13 genera of Myrtaceae s.s. as well as various outgroups. Their analysis led to a number of conclusions with taxonomic implications: (1) there is a lack of support for separate family status for Heteropyxis and Psiloxylon; (2) there is strong support for the monophyly of the myrtoid group, excluding Syzygium; (3) there is strong support for Johnson and Briggs' (1984)
Chamelaucium alliance, including elements formerly referred to the separate subfamily Chamelaucioideae (e.g., by Schmid, 1980
); and (4) the Metrosideros alliance is probably polyphyletic. The study by Gadek, Wilson, and Quinn (1996)
also led to conclusions relating to the evolution of the ovary and the fruit within the family, i.e., that the superior, or almost superior, ovary is plesiomorphic and that fleshy fruits have arisen more than once.
This paper critically examines the morphological characters that have been used in defining relationships of genera and tribes (alliances) in Myrtaceae and integrates these with an analysis of a complementary taxon matrix comprising molecular sequences of matK. This is not intended to be a comprehensive molecular analysis, rather an attempt to examine critically the morphological characters commonly used in identifying affinities in the family.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
or because they had a substantial literature of nonmolecular characters, particularly wood anatomy. Once the morphological data set was fixed, a matching molecular data set was compiled for the same exemplar species.
|
. All data were checked against the original sources cited in that paper. Autapomorphies (e.g., their characters 17, 18, 19, 20, 34, 43, 46, 47, and 53) have been excluded, and the scoring of other characters has been modified. Where more than one character scored the same information, which amounted to character weighting, these have been combined (e.g., their characters 2, 3, and 4 have been combined in our character 2; 59 and 60 are combined in 43). In all, 45 characters are retained (Table 2). Four new characters are added: 4, 11, 25, and 40. The standard scoring employed here is "unordered multistate" without any of the weighting used in the Johnson and Briggs approach. Qualea was used as a representative of the outgroup, Vochysiaceae, this being the exemplar taxon used by Chase et al. (1993)
, Conti, Litt, and Sytsma (1996)
, and Conti et al. (1997)
. Morphological data for this family are derived from Flores (1993a, b)
, Kawasaki (1998)
, A. Litt (Yale University, personal communication), and P. G. Wilson (personal observations).
|
, we have taken data from Ingle and Dadswell (1953)
and augmented it with data for Psiloxylon (Schmid, 1980
), Heteropyxis (Stern and Brizicky, 1958
; Schmid, 1980
), Kunzea (Patel, 1994
), Calyptranthes (Dias-Leme, Gasson, and Nic Lughada, 1995
), Anetholea (Wilson, O'Brien, and Quinn, 2000
), and the Vochysiaceae (Quirk, 1980
). The terminology of Ingle and Dadswell (1953)
is adopted for wood anatomy. Wood characters for some American taxa were inferred from the general descriptions provided by Record and Hess (1943)
. Myrtaceae generally have fibers with distinctly bordered pits, i.e., fiber-tracheids (character 3); this is usually interpreted as the plesiomorphic condition. These were recorded for most myrtaceous genera in this study, plus Heteropyxis. A smaller number of myrtaceous genera have fibers with simple to minutely bordered pits, i.e., libriform fibers. This condition also occurs in Psiloxylon and in many genera of the Vochysiaceae, all of which have septate fibers (character 4), these being crystalliferous in Psiloxylon.
Rays (character 6) of Myrtaceae are never homogeneous; they have been scored as either weakly or strongly heterogeneous, following the distinctions used by Ingle and Dadswell (1953)
. In their discussion, they define these two conditions as (1) strongly heterogeneous, uniseriate rays or uniseriate margins of multicellular rays consisting of few to many rows of upright cells; or (2) weakly heterogeneous, multiseriate rays with a single row of square to upright marginal cells. In genera with the second type, there is no tendency for the rays to be of two distinct widths, a phenomenon found frequently in the first type.
Johnson and Briggs' (1984)
character 12 (presence or absence of crystalliferous fibers) appears to be an error in terminology. As far as we are aware, and as we have noted above, Psiloxylon is the only genus included here that has crystalliferous fibers; these were cited by Baas and Zweypfenning (1979)
as a possible basis for postulating an affinity with Lythraceae. The character is excluded here as an autapomorphy. Johnson and Briggs presumably had the character "presence or absence of crystal strands in the axial parenchyma" in mind; these are relatively common and have been scored here as character 10.
Oil ducts (character 11) are a feature of a small number of genera but are quite variable in form. We have scored two basic types, as seen in transverse section (TS) of the petiole: small ducts both inside and outside the vascular strand (Welch, 1923
), as in Lophostemon (state 1); and ducts (usually relatively large) occurring only within the vascular strand, as in Corymbia (state 2). We have scored Arillastrum as positive for the latter state but ducts do not occur in the petioles of all individuals (P. G. Wilson, personal observation); however, "oil tubes" are reported to occur in other tissues (Dawson, 1970
). In Arillastrum, these may be a characteristic of juvenile plants as they are in some other members of the eucalypt group (Carr and Carr, 1969
). Johnson and Briggs (1984)
scored oil ducts as three different characters, distinguishing between stem ducts and petiole ducts; this needs further investigation, but the study by Carr and Carr demonstrated that they were not independent of each other, at least in eucalypts. Various members of Vochysiaceae, including Qualea, are reported to have "traumatic axial gum ducts" in their stems (Quirk, 1980
). Although Ingle and Dadswell (1953
, p. 397) noted that these are similar to those that occur in some members of the eucalypt group, we do not consider these gum ducts to be homologous with the oil ducts in Myrtaceae, and they are not scored as such.
For indumentum (character 12), we have scored three states: glabrous, "standard," and multicellular hairs. "Standard" hairs are defined by Briggs and Johnson (1979)
as "acute, rather thick-walled, and unicellular (with no basal cell)," but they list a number of variants, including hairs that are curled, bent, basally saccate, and two-armed. In Johnson and Briggs (1984)
, five different hair types are scored but four of these are autapomorphies and have no value in terms of phylogenetic reconstruction; for the purposes of this study, these four are considered variants of the standard hair type. It is possible that the thin walled, blunt-ended hair type has phylogenetic significance, but probably only within the eucalypt clade (Ladiges and Humphries, 1983
). Multicellular hairs occur in Angophora (Johnson, 1972
; Ladiges, 1984
) and have been recorded for Allosyncarpia (Blake, 1977
), a taxon not included in the present analysis. They are also found very rarely in Syzygium but they are not typical of the genus and have not been scored here. Arillastrum has short, branched hairs on the hypanthium (one of the autapomorphic characters in the Johnson and Briggs analysis), and Dawson (1970)
reported short, multicellular hairs on the epicotyl. The latter, however, could not be independently confirmed, so multicellular hairs are not scored for this taxon in the present analysis. On present knowledge, Eucalyptopsis is glabrous.
"Bristle-glands" (character 13) are distinctive structures characteristic of some members of the eucalypt group (Johnson, 1972
; Ladiges, 1984
). We have followed Johnson and Briggs (1984)
in scoring these as present (state 2) in Angophora (and Corymbia) and incipient in other eucalypts (state 1). Ladiges (1984)
has shown that emergent glands are present on juvenile foliage of the "Monocalyptus" group of eucalypts and absent in the "Symphiomyrtus" group. No member of the latter group is scored in the present analysis but, in this case, the absence of bristle glands appears to be the result of secondary loss rather than absence, as in the rest of the family.
Stipules (character 14) have been scored as absent, small, or hair-like. Small stipules (state 1) are paired but mostly inconspicuous and visible only on new growth near the apex of the shoot; only rarely, for example in Calytrix, are they readily discernible. Hair-like, or rudimentary, stipules (state 2) are often more numerous and have been documented and illustrated in Myrtaceae by Weberling (1956, 1966, 2000)
; these are not particularly easy to observe where herbarium material is limited and, in these cases, we have scored the character as unknown (?).
Vegetative phyllotaxy (character 15) was scored as two characters by Johnson and Briggs (1984)
but we have treated it as one character with three states. Whereas the exemplar species of Xanthostemon (X. chrysanthus) has spiral phyllotaxis throughout (state 2), there are a few species that show an additional state, i.e., spiral seedling phyllotaxis that changes to opposite (Wilson, 1990
); this state may have arisen more than once in this genus. The only other genus of Myrtaceae known to show this state is Barongia (Wilson and Hyland, 1988
).
In the case of the intramarginal vein (character 16), Johnson and Briggs (1984)
consider brochidodromy to be the plesiomorphic state, with strengthening of the vein arches leading to apomorphic states in which an intramarginal vein and/or various degrees of acrodromy are apparent. Examples of the wide variation that occurs in the family can be seen in Klucking (1988)
and Christophel and Hyland (1993)
. Although Johnson and Briggs (1984)
considered that this tendency had occurred along several separate lineages, we have been unable to distinguish separate states of acrodromy; all the derived forms are scored as a single apomorphic state.
Inflorescence characters follow Briggs and Johnson (1979)
but scoring differs from Johnson and Briggs (1984)
; inflorescence phyllotaxy (character 17), for example, combines two of their characters. For "inflorescence branching" (character 18) we have redefined their scoring in terms of the actual inflorescence types (as scored in Briggs and Johnson, 1979
), with two levels of reduction between panicle (state 0) and monad (state 3). It should be noted that indeterminate (blastotelic) raceme-like aggregations of monads are scored according to the unit inflorescence, i.e., the monad. In the monad and triad, in particular, the distinction between the pedicel and the anthopodium (character 20) has not been widely recognized, even after Briggs and Johnson (1979)
pointed out its significance.
The androecium (character 21) is rather complex and difficult to score. The outgroup, Vochysiaceae, is basically obhaplostemonous (A. Litt, personal communication) but the androecium is consistently reduced to a single fertile stamen (see Fig. 6 in Kawasaki, 1998
) and up to four staminodes; we have scored this as an autapomorphy (state 3). The diplostemonous condition (state 0) is found in Psiloxylon (Schmid, 1980
), and a partial loss of the antesepalous whorl (state 1) seems to have occurred in Heteropyxis (Stern and Brizicky, 1958
; Fernandes, 1971
). In Myrtaceae s.s., Johnson and Briggs (1984
, pp. 739–741) conclude that the basic condition is obdiplostemony with frequent suppression of the antesepalous whorl. They drew their conclusions from published observations of small groups of antesepalous stamens in some specimens of Arillastrum (Dawson, 1970
) and from the observation by Bunniger (1972)
of antesepalous staminal primordia developing after the antepetalous primordia in one species of Luma (as Myrceugenella). In fact, in their Table 5, Johnson and Briggs scored the myrtoids as a whole as showing this state, even though Bunniger did not observe antesepalous primordia in the other myrtoid taxa he studied (Psidium and Eugenia) and Payer (1857)
did not record them in Myrtus. We interpret the situation in Luma as either a vestige of the plesiomorphic condition or a reversion towards obdiplostemony.
|
has provided confirmation that, in most genera in the family, staminal primordia arise at, or adjacent to, the bases of the developing petals. In the genera they have studied, stamen initiation has occurred either within a prestaminal bulge or directly on the floral apex. They have argued (1996) that the view of Myrtaceae being basically obhaplostemonous with subsequent proliferation is not supported by their observations, since the prestaminal bulge cannot be interpreted as a single stamen primordium. The basal position of Psiloxylon in the Myrtaceae s.l. has been supported by molecular data (Gadek, Wilson, and Quinn, 1996
), and this gives some support to Schmid's (1980)
suggestion that the plesiomorphic condition in the family was likely to be diplostemony. Thus, the usual situation in the Myrtaceae s.s., although apparently a de novo development, is no doubt derived from this plesiomorphic state, perhaps via some form of obdiplostemony, and we continue use of the classical terminology as a shorthand.
Many groups, however, have yet to be investigated and, in the absence of firm data, taxa have mainly been scored as obhaplostemonous (state 2). We have not scored obdiplostemony, as it has not definitely been recorded for any of the taxa included here. For reasons stated above, Arillastrum has been scored as showing partial loss of the antesepalous whorl (state 1); its near relative, Eucalyptopsis, has been scored similarly, following Johnson and Briggs (1984)
. Thaleropia and Tristania occasionally have stamens that appear to be initiated in an antesepalous position (Orlovich, Drinnan, and Ladiges, 1996
), and these taxa also have been scored as showing partial loss of the antesepalous whorl.
Stamen proliferation (character 22) proceeds in various ways from the staminal primordia. Stamens appear in a single row (state 1) but in some taxa further rows arise centripetally so that the mature androecium has two to many rows (state 2). Alternatively, development of the group of stamens without any lateral spreading can result in a staminal fascicle (state 3). The final form of the androecium depends on whether the stamens expand radially to occupy the available space on the rim of the hypanthium or remain clustered in front of the petals, with or without fusion. As noted by Orlovich, Drinnan, and Ladiges (1996, 1999)
, staminal clustering is usually the result of stamen initiation on a distinct antepetalous prestaminal bulge, whereas a more-or-less continuous ring of stamens often results from the staminal initials arising directly on the hypanthial wall in front of the developing petal primordium. The aestivation of the stamens in the mature bud (character 23) may be a significant character; they are erect in Heteropyxis and Psiloxylon (Schmid, 1980
) and Vochysiaceae, but inflexed in Lythraceae, Melastomataceae, and most Myrtaceae.
Character 25, "anther dehiscence," has been added. Anthers are uniformly tetralocular in the family, but Tobe and Raven (1987, 1990)
have shown that Heteropyxis and Psiloxylon have anthers in which each locule opens by a separate slit (state 2). This is in contrast to most other genera in which the pair of locules in each anther cell open by a common slit (state 1). The latter state is clearly plesiomorphic, and the only other genus in Myrtales that has the former type of anther dehiscence is Rhynchocalyx (Rhynchocalycaceae).
Pollen form (character 26) is scored following the categories used by Patel, Skvarla, and Raven (1984)
, augmented by data from Pike (1956)
and Gadek and Martin (1981)
. Pollen of Vochysiaceae is longicolpate (which we have scored as plesiomorphic) but grains are oblate spheroidal rather than oblate (Erdtman, 1952
, p. 452). For pollen, Johnson and Briggs (1984)
scored two separate characters, i.e., "loss of syncolporate condition" (which was scored as an autapomorphy for Kjellbergiodendron) and "pollen with very large polar islands" (which was scored as an autapomorphy for Psiloxylon). We have scored a single character with three states.
Ovary vascularization (character 30) could not be determined with any certainty in taxa with a single locule (e.g., Darwinia) or with basal placentation (e.g., Lysicarpus). Scoring of "mixed" vascularization follows Johnson and Briggs (1984
, character 45).
With character 38, "form of the cotyledons," the folded state is very diverse, ranging from cotyledons individually folded as in Arillastrum, to the obvolute/convolute type seen in Tristaniopsis and Lophostemon. The complex ruminate cotyledons of Acmena and the variable, thick cotyledons of Lindsayomyrtus (Hyland and van Steenis, 1973
) are scored as uncertain (?) but the thick cotyledons of Syzygium australe are partly interlocking and are scored as folded (state 1).
The mode of embryo sac formation (character 39) is considered by Tobe and Raven (1987)
to be quite significant. The vast majority of Myrtalean taxa (including all Myrtaceae) that have been examined have the Polygonum type of development but Heteropyxis and Psiloxylon have the Allium type. As Tobe and Raven (1987, 1990)
point out, the Allium type of development is rare, being recorded elsewhere in Myrtales only in Alzatea (Alzateaceae).
The number of seeds relative to the number of ovules per locule (character 43) combines characters 59 and 60 of Johnson and Briggs (1984)
. Seed number is related to reproductive strategy, which is itself partly correlated with fruit type. Most genera in the present analysis are scored as plesiomorphic even though fertile seeds are frequently only a small percentage of seed-like objects (seed + "chaff") in the fruit; this is because there is no fixed pattern of seed set and all ovules appear to have the potential to become seeds. The second character state (seeds few per fruit) has been scored for two groups of genera: those with ovulodes that never become seeds and those with a more-or-less fixed pattern of seed set. The third character state (single seed per fruit) is mostly found in indehiscent fruits, the exception, in this data set, being Whiteodendron.
The presence of oil glands (character 44) that produce essential oils is one of the fundamental features of the family Myrtaceae s.s. Heteropyxis also has oil-bearing glands in the leaves, and these produce an oil that contains >90% monoterpenes (Weyerstahl et al., 1992
), a rather simple complement of oils. Psiloxylon has leaf glands (Scott, 1980
) but these do not produce any essential oils (J. J. Brophy, University of New South Wales, personal communication); glands have long been known to occur in this genus but were not considered different from similar glands that occur in Bixaceae and Flacourtiaceae (de Cordemoy, 1895
).
Base chromosome number (character 45) is a potentially useful character in some parts of the family. There is considerable dysploid reduction within the Chamelaucium suballiance, with considerable variation in haploid number within genera. This trend is scored under state 2; in this database there is only one example, n = 6 in Darwinia fascicularis, although within the genus it varies from n = 5 to 11 (Smith-White, 1954
; Briggs, 1962
; Rye, 1979
). There are two records for Vochysiaceae, both in Vochysia (n = 11, Goldblatt, 1979
; n = 12, Berry, 1987
), and these data are extrapolated to provide the ambiguity scored for Qualea in this analysis.
These characters have been scored (Table 3) for the same set of exemplar taxa for which sequence data are available, so as to score some of the variation within the groups that were the terminal taxa in the Johnson and Briggs (1984)
analysis. Hence, the number of taxa has been considerably expanded. The nomenclature used has also been updated.
|
Molecular database
DNA was extracted either by the hot CTAB (hexadecyltrimethyl ammonium bromide) method of Doyle and Doyle (1990)
or using the plant DNeasy Minikit (Qiagen, Clifton Hill, Victoria, Australia) following a liquid nitrogen grind. Double-stranded templates were amplified in a Thermocycler (Perkin Elmer, Norwalk, Connecticut, USA) and purified in Wizard Preps PCR (polymerase chain reaction) columns (Promega, Annandale, New South Wales, Australia). Sequences were obtained by manual sequencing using radionucleotide labelling (Gadek, Wilson, and Quinn, 1996
) or automated sequencing on an ABI Prism 377 DNA Sequencer (Perkin Elmer, Norwalk, Connecticut, USA) using fluorescent dye-labelled terminators. Electropherograms were processed using ABI software, Factura and Autoassembler. Sequences were aligned using DNA and Protein Sequence Alignment software (DAPSA; Dr. E. Harley, University of Cape Town, Cape Town, South Africa). Primers used were as per O'Brien, Quinn, and Wilson (2000)
.
At present, neither morphological nor DNA data provide robust evidence of any taxon within Myrtales being the unambiguous sister to Myrtaceae s.l. Memecylon and Qualea were included as representatives of possible sister families, Memecylaceae and Vochysiaceae, as indicated by Conti, Litt, and Sytsma (1996)
and Conti et al. (1997)
. Rhynchocalyx was included as a more distant Myrtalean representative, Arabidopsis and Sinapis were included to represent the closely related order Brassicales [Capparales] (Chase et al., 1993
; Angiosperm Phylogeny Group, 1998
), and an unambiguously distant taxon, Saxifraga, was used to root the analysis.
Analyses
Heuristic searches were conducted in PAUP* version 4.0b8 (Swofford, 2001
) using TBR (tree bisection reconnection) branch-swapping to recover the most-parsimonious trees. In all analyses, 100 replicates of random taxon addition were conducted with the heuristic search option, in order to establish whether there was more than one island of trees. To assess the impact of homoplasy in the molecular data, the trees generated were then used as starting trees in heuristic searches employing successive approximations weighting (SAW) according to the mean rescaled consistency index (RC) with the base weight set at 1 (Farris, 1969
). This was repeated until tree length reached a constant value. In the combined analysis, all data were given the same weight. Relative support for the clades identified by parsimony analysis was estimated using the bootstrap option in PAUP. On the separate analyses of the nonmolecular and molecular data, the numbers of trees found in some replicates exhausted available computer memory. For this reason the number of trees saved in each replicate was limited to 10, and the number of replicates was set at 5000. No such limit was necessary in the analysis of the combined data; in this case the number of replicates was 500. Decay values were determined using AutoDecay (version 4.0.2'; Eriksson, 1999
) and heuristic searches in PAUP* with 30 replicates of random taxon addition. Hypotheses of character evolution were examined by using MacClade (version 3.08a; Maddison and Maddison, 1999
) to plot the morphological data onto trees generated from the combined data. Constraint trees were constructed in MacClade and imported into PAUP in order to conduct constraint analyses to test alternative hypotheses of evolution of nonmolecular characters.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
) between them is not resolved. Monophyly of Myrtaceae s.s. is supported by 77% bootstrap (bs), and +3 decay. Several subclades receive support: the Acmena group (79% bs, +4 decay), the Eucalypt group (67% bs, +3 decay), and the Tristania group (89% bs, +3 decay). The Myrtoideae s.s. (the Myrtoid group) is also identified, but both the group and its internal topology collapse at +1 step. Relationships between the remaining taxa are unresolved.
|
). The first 103–107 bp were missing from the sequences of Corymbia, Ristantia, and Tristania, and sequences for Eucalyptopsis, Rhynchocalyx, and Xanthomyrtus had 238, 434, and 188 bases, respectively, undetermined near the middle of the gene. All sequences were complete at the 3' end except Psiloxylon, which lacked 19 bases. Alignment required 21 indels, each involving up to 12 complete codons. The aligned database consisted of 1601 bp and included 496 potentially informative characters. Ten indels were potentially informative (Table 4); these were scored as additional characters and added to the database. Uncorrected pairwise sequence divergences ranged up to 7.2% (Rhodamnia to Babingtonia) within the ingroup.
|
|
The successive approximations weighting analysis produced three equally parsimonious trees of 859.7 steps (CI = 0.80, RI = 0.84, RC = 0.75), but the strict consensus of these showed only one change (cf. Fig. 2) affecting an unsupported part of the tree: clade H became the sister group to clade D. All the well-supported clades were still present, and there was an increase in bootstrap support for some of the weakly supported clades.
An analysis of the combined data (all characters unweighted) yielded two equally parsimonious trees of 1217 steps (Fig. 3; CI = 0.51, RI = 0.62, RC = 0.39). Again, all the labelled clades identified in Fig. 2 were retrieved. The differences between these trees involve only clades that are unsupported or weakly supported in Fig. 2. The Melaleuca group (clade L) is now sister to the Eucalypt group (clade K). Several of the individual taxa that were placed in less well-supported positions have also moved. Osbornia and Xanthomyrtus are now the respective sister taxa of the Acmena group (clade A) and Myrtoid group (clade I); in the latter case there is some support for the association (75% bs, +3 decay). Xanthostemon is now the first lineage to diverge within Myrtaceae s.s., and the relationships of Syncarpia are incompletely resolved.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
The Lophostemon group (clade M, Fig. 2) represents an early diverging lineage within Myrtaceae s.s. Johnson and Briggs (1984)
scored Kjellbergiodendron and Whiteodendron separately, and their analysis placed these genera together as sister to a clade comprising Lindsayomyrtus and the Acmena group. The molecular analysis confirms the sister relationship of the first two genera but neither Lindsayomyrtus nor the Acmena group appear to be closely related to them; instead Kjellbergiodendron and Whiteodendron are robustly grouped with Lophostemon.
Although there is some support (75% bs) for the link between Xanthostemon and clade M in the molecular data, the remainder of the Metrosideros group of Johnson and Briggs (1984
, their Table 3) are scattered through the cladogram. Metrosideros clusters robustly with Tepualia (clade D, 100% bs), Thaleropia (formerly Metrosideros sect. Adnatae J.W. Dawson) is sister to Tristania (clade B, 100%), as suggested by Wilson (1993)
, and Tristaniopsis, Ristantia, and Lysicarpus are grouped (99%) in clade H. None of these clades shows any affinity with another in this group.
There is a very strongly defined Eucalypt group, and within it there is 94% bs support for Corymbia as sister to Angophora rather than to Eucalyptus s.s.; this is supported by the distribution of indel 5 (see above and Fig. 2). This pattern of relationship is in agreement with other molecular studies (Udovicic, McFadden, and Ladiges, 1995
; Ladiges and Udovicic, 2000
), is supported by morphological evidence (Ladiges, 1984
), and lends credence to the segregation of Corymbia from Eucalyptus s.l. (Hill and Johnson, 1995
).
The Leptospermum group of Johnson and Briggs (1984)
, which included both the Leptospermum and Calothamnus suballiances, is not retrieved. On the sample represented here, the Chamelaucium group (clade F) is strongly indicated as sister to the Leptospermum suballiance (Briggs and Johnson, 1979
; clade E), while the Calothamnus suballiance (= Beaufortia suballiance of Johnson and Briggs, 1983
) are placed at a distance in clade L.
Fleshy-fruited taxa do not form a monophyletic group. The matK analysis confirms that the Acmena group has arisen independently of the Myrtoid group and that Anetholea (Wilson, O'Brien, and Quinn, 2000
) is a member of it. An interesting result is that Xanthomyrtus does not cluster with the Myrtoid group but is loosely associated with the Tristania group; this is in agreement with the assertion, based on seed and embryo morphology, that the genus does not belong in the Myrtoideae s.s. (Landrum and Stevenson, 1986
). It is clear from the analyses presented here that, based on its position in the phylogeny, the group of genera identified here as the Myrtoid group do not form a taxon of subfamilial rank; they would be more appropriately placed in a tribe, i.e., Myrteae.
Although there is good character support for most of the major clades in the family, the relationships between them are mostly unsupported (<50% bs). This may be the result of rapid evolutionary diversification within the family in response to the strongly selective environmental changes that occurred at the time of increasing aridity in the Oligocene/Miocene, resulting in the widespread appearance of sclerophyllous vegetation. Alternatively, it could be the result of early genetic isolation of the progenitors of the major clades in the family followed by "gradualistic" change. The fact that the successive approximations weighting analysis did not produce any increase in resolution or character support for clades is interpreted as evidence that the lack of a well-supported resolution of basal relationships in the family is due to insufficient characters rather than excessive homoplasy. This tends to favor the first alternative described above. Hence, additional sequence data from a region of similar variability is needed to address this situation, rather than data from a more conservative region.
Virtually all clades in the family have representatives in Australia, which suggests that these progenitors of the family as we know it were located in east Gondwana in the late Cretaceous, although the sister taxa to the Myrtaceae s.s., Psiloxylon and Heteropyxis, appear to be of west Gondwanan origin. Recent reassessment of the fossil evidence suggests the family may have been present by the middle Eocene (Manchester, Dilcher, and Wing, 1998
).
The estimate of phylogeny derived from the analysis of the morphological data (Fig. 1) is poorly resolved and mostly very weakly supported, despite the reevaluation of the characters. Two separate clades of fleshy-fruited taxa are recognized (the Myrtoid and Acmena groups), but Kjellbergiodendron and Xanthomyrtus are ungrouped. Johnson and Briggs (1984)
recognized Acmena and Syzygium as members of a lineage separate from the remaining fleshy-fruited taxa, and this group, together with Anetholea (Wilson, O'Brien, and Quinn, 2000
), is one of the most strongly supported clades in the morphological analysis (79% bs, +4 decay).
The combination of the nonmolecular and molecular data did not affect the recognition of any of the labelled clades (cf. Figs. 2 and 3). It did, however, result in some degree of support (75% bs, +3 decay) for a link between Xanthomyrtus and the Myrtoid group. The different placements of Xanthomyrtus in the molecular and combined analyses indicates the need for caution in drawing firm conclusions on the relationships of this genus. Further data are needed to test the alternative hypotheses. Otherwise, changes between Figs. 2 and 3 are unsupported in either case. Osbornia, which is weakly associated with the Melaleuca group in Fig. 2, is the unsupported sister to the Acmena group in Fig. 3. There is now a sister relationship (<50% bs) between the Eucalypt and Melaleuca groups (clades K and L). The two fleshy-fruited groups (clades A and I) are now included, along with the Backhousia group, Osbornia and Xanthomyrtus, in an unsupported monophyletic assemblage (clade P).
There is marked homoplasy in many of the nonmolecular characters. Two 2-state characters (6 and 10) require 6 steps each on Fig. 1. Almost every morphological character displays some degree of homoplasy on the molecular and combined estimates of phylogeny. In some cases there is extreme homoplasy: e.g., characters 20 (presence/absence of an anthopodium) and 38 (flat/folded cotyledons) each require 9 steps on Fig. 3, while the four-state character 18 (inflorescence branching) requires a total of 20 steps on Fig. 3. Nevertheless, there are some characters that do correlate with strongly supported features of the molecular-based topology.
The separation of Psiloxylon and Heteropyxis from Myrtaceae s.s. is confirmed by (1) the stamens in the mature bud being straight rather than inflexed (character 23/state 0, Fig. 4; this state also occurs in the greatly reduced androecium of Qualea), (2) the anther cells dehiscing by separate slits (25/2, Fig. 4), and (3) the bisporic embryo sac (39/1, Fig. 5). Vasicentric tracheids (5/1, Fig. 5) also distinguish Myrtaceae s.s., but with secondary losses in the Acmena group and Eucalyptopsis. The base chromosome number for Myrtaceae s.l. is probably x = 12, (45/1), but the distinctiveness of this state is reduced by the equivocal records (11 and 12) from the Vochysiaceae. Psiloxylon and Heteropyxis, on the other hand, are linked to Myrtaceae by the presence of secretory cavities in the leaves and of glands in the anther connective (character 24, states 1 or 2). The most parsimonious explanation for the lack of essential oils in the leaves of Psiloxylon (44/0) is that the pathway for their production has been lost in that lineage.
|
|
). The eucalypt clade is distinguished by the presence of ovulodes (42/1) and by the low numbers of seeds relative to ovules (43/1), but the latter character is somewhat difficult to measure quantitatively.
Vessel-ray pitting provides some support for both the Acmena and Eucalypt groups, although in each case the defining state appears homoplastic (Fig. 5). Large isodiametric pits (2/3) characterize the Eucalypt group but have been recorded in Melaleuca leucadendra (Ingle and Dadswell, 1953
). Other species of Melaleuca, including the one used in the present study, M. viridiflora, have small, nontracheidal pitting (2/1). Large, somewhat scalariform pits (2/2) are a synapomorphy for the Acmena group but are also found in Kjellbergiodendron within the Lophostemon group. Constraint analysis on the combined data set shows that a single origin for this state requires an additional 18 steps.
Fruit characters show an interesting distribution on the combined estimate of phylogeny (Fig. 6). The plesiomorphic state in the family is a dehiscent fruit, but indehiscence arises in four lineages: Darwinia, Kjellbergiodendron, Psiloxylon, and in clade P. The combined data provide strong evidence for multiple origins of indehiscence, a single origin requiring an additional 85 steps on the tree (51 steps excluding Psiloxylon). Fleshiness of the fruit is also shown to be homoplastic: three origins are indicated. A single origin requires an additional 66 steps on the tree. In all taxa with fleshy fruit, except Psiloxylon, there is the development of a fleshy hypanthium (34/1). In Psiloxylon, however, the fruit develops from a superior ovary, and this condition of a fleshy ovary wall devoid of a hypanthium is unique to this taxon. Even excluding Psiloxylon, a single origin of fleshy-fruitedness within Myrtaceae s.s. requires an additional 33 steps. We reiterate that the leathery fruits of Lindsayomyrtus (developed mostly from the ovary wall) and Osbornia (from the hypanthium) have not been scored as fleshy, nor are the indehiscent fruits of Darwinia, from the Chamelaucium alliance, in any sense fleshy. Hence there is very strong evidence for separate origins of fleshy fruit within the Myrtoid and Acmena groups, as proposed by Johnson and Briggs (1984)
. It should be noted, however, that these two groups are placed within clade P and that they therefore appear to have been derived from a common ancestor with indehiscent fruit (Fig. 6). Note, too, that Kunzea, which overwhelmingly has dry dehiscent fruits, also contains some species with dry indehiscent fruits (e.g., K. cambagei) and at least one with fleshy indehiscent fruits (K. pomifera).
The "standard" Myrtaceous unicellular hair is shown to be plesiomorphic for the family. The indumentum has been lost in clades A, F, K, and in Lindsayomyrtus and Whiteodendron. The absence of "standard" hairs provides one of the distinctions between the Acmena and Myrtoid groups (Briggs and Johnson, 1979
). The eucalypts are also well known for their lack of these simple hairs, although various groups within this clade do have distinctive hair types (bristle glands and thin-walled, blunt-ended hairs) as discussed above. In the topology presented here, however, indumentum has been lost on a number of occasions, so that it is a relatively weak indicator of affinity.
There are additional characters that appear to be potentially informative but require further investigation. The Backhousia clade (C) seems to be characterized by a particular embryo type, having a curved hypocotyl with the cotyledons bent to one side, so that they are incumbent with respect to it; both genera are confined to Australia. The Metrosideros clade (D) is apparently characterized by having only five vascular traces in the hypanthium, a significant feature pointed out by Wilson (1993)
. From the limited data available, it seems likely that the plesiomorphic state is ten major veins but this needs to be confirmed. It is clear that Psiloxylon has ten veins (Fig. 3 in Schmid, 1980
), and it seems likely that Heteropyxis does, too (Fig. 1 in Stern and Brizicky, 1958
).
The Leptospermum clade (G) is made up of two robust subclades (E and F). The Leptospermum subclade (E) is discussed more fully in O'Brien, Quinn, and Wilson (2000)
, and the topology is similar to that found by Johnson and Briggs (1984)
, except that the Melaleuca group is excluded. The Chamelaucium subclade (F) may be distinguished by the distinctive embryo, which has two small cotyledons on a slender neck folded down against a relatively massive hypocotyl. By contrast, the embryos of members of the Leptospermum subclade are relatively unspecialized: they are straight, with the cotyledons usually longer than the hypocotyl.
There are no clear synapomorphies for the Tristaniopsis clade (H). Most taxa have staminal bundles of some sort, sometimes indistinct, and sterile anthers occur in Lysicarpus and Ristantia. The Myrtoid group (clade I) is defined morphologically by the combination of the fleshy fruit, the presence of indumentum of "standard" hairs, and the multiple whorls of stamens, although there are notable exceptions to this last feature. Traditionally, this group has been divided into three subtribes based on embryo morphology: the Myrtinae (curved/coiled embryos with small cotyledons), Myrciinae (embryos with foliaceous cotyledons), and Eugeniinae (fleshy embryo with plano-convex cotyledons). This seems to have some support here, with three taxa of Myrtinae (Archirhodomyrtus, Pilidiostigma, and Rhodamnia) forming a single clade (Figs. 2 and 3) but this distinction is not sustained in a wider molecular analysis (P. G. Wilson and M. M. O'Brien, unpublished data). This wider analysis indicates that most Australian Myrtinae form a clade distinct from other Myrtinae and marked by a 36-bp indel, but relatively few taxa have been sampled. Further work is needed, particularly to establish if the New Zealand and New Caledonian Myrtinae are sister to the Australian or southern South American taxa. Preliminary data show that two genera from New Zealand do not have the large indel but New Caledonian genera are yet to be sampled. It is perhaps significant that chromones, uncommon components of essential oils, have been found in some species of both New Zealand and southern South American taxa (Weyerstahl, Marschall, and Landrum, 1992
).
For the Eucalyptus clade (K) the only additional character that may be informative, apart from the indumentum and wood anatomical characters mentioned above, is the lack of rudimentary stipules (Weberling, 1956
). The Corymbia + Angophora clade (J) seems to have one clear synapomorphy, i.e., the presence of bristle glands bearing four cap cells with micropapillae (Ladiges and Humphries, 1983
; Ladiges, 1984
). Other possible characters, such as stem oil ducts and seed morphology, require more detailed analysis.
The members of the Melaleuca clade (L) seem typically to have spicate inflorescences and thickly woody fruits with inconspicuous or deciduous sepals. However, variants, such as capitate inflorescences, do occur. A similar inflorescence type is found in Kunzea baxteri but that species has conspicuous, persistent sepals and does not have thickly woody fruits. Ladiges et al. (1999)
, in a much broader study of the Melaleuca group, suggest that the presence of staminal bundles is the plesiomorphic condition in this clade. Staminal bundles are of widespread occurrence in the family s.s. and are found in five of the clades in Fig. 2. Embryo type does not seem significant because embryos with flat cotyledons and embryos with obvolute cotyledons both occur in genera in the group.
Taxa in clade M share some potentially informative character states. All three genera have oil ducts: Kjellbergiodendron and Whiteodendron have relatively wide oil ducts in the pith of their petioles but in Lophostemon they are smaller in diameter and also occur in the cortex. A virtually identical arrangement is found in Syncarpia, which is reflected in the nonmolecular tree, but there is no support for a close relationship of Syncarpia and Lophostemon in the molecular analysis. Kjellbergiodendron and Whiteodendron also appear to have a similar embryo type with a larger, outer cotyledon that encloses a smaller, inner one.
Available information from bark anatomy is limited, but also corroborates some groupings of genera. Bamber (1962)
presents data that support the Melaleuca group (clade L) and the Leptospermum subgroup (clade E) and strongly affirm certain genera, like Choricarpia, Syncarpia, and Metrosideros. A broader survey might provide evidence for other groups as well.
Conclusions
The matK sequence data allowed the identification of 11 robust groupings within the Myrtaceae s.l. However, apart from the placement of the Psiloxylon group as the first diverging lineage, relationships between the groups are either unresolved or weakly supported by the data. Clearly further sequence data are required before a reliable estimate can be made of the deeper relationships within the family. An increase in taxon sampling is also needed to define the limits of the infrafamilial taxonomic groups recognized here. There is strong support in the molecular data for a number of the alliances and suballiances defined by Briggs and Johnson (1979)
, i.e., the Acmena, Backhousia, and Eucalyptus alliances, the Myrtoideae s.s., and the Calothamnus and Chamelaucium suballiances. On the other hand, the Metrosideros and Leptospermum alliances are shown to be polyphyletic; neither was retrieved in analyses of either database. Several smaller informal taxonomic groupings are defined from among the members of the Metrosideros alliance: the Tristania, Tristaniopsis, Metrosideros, and Lophostemon groups. Despite the very limited sampling within groups in this analysis, there is robust support in the molecular data for the generic distinction between Eucalyptus and Corymbia.
Analysis of a revised and updated database of nonmolecular characters yielded only limited resolution of relationships. Despite this, there was some congruence between the morphological and molecular estimates of relationships. Two separate fleshy-fruited clades, the Acmena and Myrtoid groups, were identified in both analyses, as were the Eucalypt and Tristania groups. An analysis of the combined databases retrieved all clades identified as strongly supported on the molecular data. Estimates of the phylogeny based on the molecular data and also the combined data reveal extensive homoplasy within the family in a range of nonmolecular characters, including wood, fruit, and floral characters, which is in line with the limited resolution obtained from the nonmolecular analysis.
|
FOOTNOTES |
|---|
5 Author for reprint requests (peter.wilson{at}rbgsyd.nsw.gov.au
).
6 Present address: Department of Botany, University of Tasmania, PO Box 252C, Hobart, Tasmania 7001, Australia.
|
LITERATURE CITED |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Baas P. R. C. V. J. Zweypfenning 1979 Wood anatomy of the Lythraceae. Acta Botanica Neerlandica 28: 117-155[ISI]
Bamber R. K. 1962 The anatomy of the barks of Leptospermoideae. Australian Journal of Botany 10: 25-54
Berry P. 1987 Vochysiaceae. Chromosome number reports. Taxon 36: 493
Blake S. T. 1977 Allosyncarpia ternata, a new genus and species of Myrtaceae subfamily Leptospermoideae from northern Australia. Austrobaileya 1: 43-46
Briggs B. G. 1962 The New South Wales species of Darwinia. Contributions from the New South Wales National Herbarium 3: 129-150
Briggs B. G. L. A. S. Johnson 1979 Evolution in the Myrtaceae: evidence from inflorescence structure. Proceedings of the Linnean Society of New South Wales 102: 158-256
Bunniger L. 1972 Untersuchungen über die morphologische Natur des Hypanthiums bei Myrtales- und Thymeleales-Familien. II. Myrtaceae. III. Vergleich mit den Thymelaeaceae. Beiträge zür Biologie der Pflanzen 48: 79-156
Carr S. G. M. D. J. Carr 1969 Oil glands and ducts in Eucalyptus L'Hérit. I. The phloem and the pith. Australian Journal of Botany 17: 471-513[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]
Chattaway M. M. 1959 The anatomy of bark. VII. Species of Eugenia (s.l.). Tropical Woods 111: 1-14
Christophel D. C. B. P. M. Hyland 1993 Leaf atlas of Australian tropical rain forest trees. CSIRO, East Melbourne, Victoria, Australia
Conti E. S. A. Graham A. Litt P. G. Wilson B. G. Briggs L. A. S. Johnson K. J. Sytsma 1997 [1998] Interfamilial relationships in Myrtales: molecular phylogeny and patterns of morphological evolution. Systematic Botany 22: 629-647[CrossRef][ISI]
Conti E. A. Litt K. J. Sytsma 1996 Circumscription of Myrtales and their relationships to other rosids: evidence from rbcL sequence data. American Journal of Botany 83: 221-233[CrossRef][ISI]
Dawson J. W. 1970 Pacific capsular Myrtaceae 1. Reproductive morphology of Arillastrum gummiferum Panch. ex Baillon (New Caledonia). Blumea 18: 441-445
