| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
0 Department of Systematic Botany, Evolutionary Biology Centre, Uppsala University, Norbyvägen 18D, SE-752 36 Uppsala, Sweden
Received for publication October 19, 1999. Accepted for publication February 10, 2000.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: cpDNA • Ixoroideae • morphology • nuclear ITS sequences • phylogeny • rbcL • Rubiaceae
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
), but have until recently received remarkably little attention from phylogenetic systematists compared to other large families like Asteraceae and Orchidaceae. In recent years, however, the number of higher level phylogenetic studies in Rubiaceae has increased considerably (see, e.g., Bremer and Jansen, 1991
; Bremer, Andreasen, and Olsson, 1995
; Natali, Manen, and Ehrendorfer, 1995
; Andersson and Rova, 1999
, for molecular studies; Andersson, 1995, 1996
; Delprete, 1996
; Persson, 1996
, for morphological studies; and Andreasen and Bremer, 1996
; Bremer, 1997
, for combined molecular and morphological analyses). Renewed interest in systematic relationships of Rubiaceae can be attributed in part to a recent survey and classification of the family by Robbrecht (1988)
. The situation in Rubiaceae exemplifies the importance of phylogenetic hypotheses as a base to stimulate new classifications based on phylogenetic relationships and points to gaps in our systematic knowledge.
In Robbrecht's classification, Rubiaceae are divided into four subfamilies: Rubioideae, Cinchonoideae, Antirheoideae, and Ixoroideae. The major difference compared with the two other classifications of the family presented during the century (Verdcourt, 1958
; Bremekamp, 1966
) is the much wider delimitation of the subfamily Antirheoideae. It has now been shown, by different investigators (Bremer, Andreasen, and Olsson, 1995
; Rova and Andersson, 1995
; Andersson, 1996
; Delprete, 1996
; Young et al., 1996
; Andersson and Rova, 1999
), that this subfamily is polyphyletic and, as a consequence, the remaining subfamilies need to be recircumscribed. The subfamily Rubioideae seems to be well supported as a group (Bremer, 1996
; Manen and Natali, 1996
), but the exact delimitation of Cinchonoideae and Ixoroideae remains to be further explored. In addition, tribal delimitations, especially in the large "Gardenieae-Ixoreae" complex (= Ixoroideae sensu Robbrecht), are still not settled.
The Ixoroideae sensu Robbrecht comprise about one-fifth of all Rubiaceae genera and consist of five tribes: Gardenieae, Octotropideae, Pavetteae, Aulacocalyceae, and Coffeeae. Two of the more well-known Rubiaceae genera are members of Ixoroideae: the economically important Coffea and the often cultivated Gardenia. The distribution of the subfamily is concentrated in the paleotropics, but the Gardenieae-Gardeniinae and the type genus Ixora extend to the tropics of the New World. Typical characteristics used in combination to delimit the Ixoroideae include secondary pollen presentation, entire interpetiolar stipules, contorted aestivation, and different kinds of fleshy fruits. Fruit characters have been very important for classification of the entire family and at the end of the nineteenth century (e.g., Schumann, 1891)
the number of seeds per carpel served as the base for classification. As ovule number varies from one to numerous in Ixoroideae this approach placed taxa of this group in different tribes and subfamilies. Taxa with contorted aestivation and numerous seeds were placed in Gardenieae, while those with one seed belonged to "Ixoreae" (correct name should be Coffeeae; Darwin, 1976
). Bremekamp (1934)
recognized this inconsistency and moved a group of genera with drupes from Gardenieae to "Ixoreae." He placed these two tribes together with Vanguerieae, Chiococceae, Cremasporeae, Coptosapelteae, and Achranthereae, in the subfamily Ixoroideae (Bremekamp, 1966
). Ixoroideae were restricted by Robbrecht (1988)
to the genera Bremekamp included in Gardenieae, "Ixoreae," and Cremasporeae. Additionally, Robbrecht moved Vanguerieae and Chiococceae to subfamily Antirheoideae, while the other two smaller tribes were included in Cinchonoideae.
As for the tribal classification of Ixoroideae several new tribal delimitations have been proposed (Fig. 1). The main part of Bremekamp's "Ixoreae" (1934) is now housed in Pavetteae (Robbrecht, 1984
). Genera with axillary inflorescences were transferred to a separate subtribe of the Gardenieae (Diplosporinae), leaving Coffeeae restricted to only Coffea and Psilanthus (Robbrecht and Puff, 1986
). Some genera with superior embryo radicles and axillary inflorescences variously treated in Gardenieae (Bremekamp, 1934
) or Cremasporeae (Bremekamp, 1966
), were merged in tribe Octotropideae (Robbrecht, 1980
; Robbrecht, Bridson, and Deb, 1993
). In addition, the small tribe Aulacocalyceae was erected for genera with superior embryo radicles but terminal inflorescences (Robbrecht and Puff, 1986
).
|
) treating Gardenieae, the largest tribe in Ixoroideae, and a molecular analysis (Persson, 1998
) using the rps16 intron and trnL-trnF spacer of the chloroplast DNA (cp DNA). Focusing on the whole subfamily, Andreasen and Bremer (1996)
used rbcL sequence and morphological analyses of four of the five Ixoroideae tribes. Recently, ITS (internal transcribed spacer) data was produced to assess its usefulness in the Ixoroideae (Andreasen, Baldwin, and Bremer, 1999
). Despite differences in taxon sampling and results, these cladistic analyses agree in the exclusion of the Diplosporinae taxa from Gardenieae and the exclusion of the genus Posoqueria from the same tribe.
Combined analyses are becoming increasingly more common in phylogenetic studies (see, e.g., Qiu et al., 1999
; Soltis, Soltis, and Chase, 1999
). Morphological data, as well as molecular data both from the chloroplast and the nuclear DNA, are analyzed simultaneously. Using not only more data, but data from different, independent sources, strengthens the phylogenetic hypothesis and can in some instances reveal inconsistencies between data sets (reviewed in de Queiroz, Donoghue, and Kim, 1995
; Huelsenbeck, Bull, and Clifford, 1996
; Nixon and Carpenter, 1997
). For example, since cpDNA is transferred more readily across lineages than nuclear DNA (see, e.g., Rieseberg and Wendel, 1993
), combined cpDNA-nuclear DNA studies have the potential of discovering effects of hybridization events on evolutionary patterns.
In this study of the subfamily Ixoroideae we use data from morphology, restriction site mapping of the cpDNA, and sequences of the chloroplast gene rbcL and the ITS region in the nuclear, ribosomal DNA. The present study examines tribal and generic relationships in Ixoroideae and proposes improvements in the present classification.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; see Table 1), sampling >40% of the genera. In addition, representatives from the tribes Vanguerieae, Alberteae, Hippotideae, and the genus Emmenopterys (incertae sedis) were included based on results of previous analyses (Bremer and Jansen, 1991
; Bremer, Andreasen, and Olsson, 1995
; Andreasen, Baldwin, and Bremer, 1999
; Bremer et al., 1999
). Outgroup taxa used were Cinchona, Luculia, Nauclea, Sarcocephalus, and Hallea, which all have been found to be outside Ixoroideae s.l. (sensu lato) in earlier analyses (see, e.g., Bremer, Andreasen, and Olsson, 1995
). In Robbrecht's classification they belong to the tribes Cinchoneae, Coptosapelteae, and Naucleeae and were included in his subfamily Cinchonoideae.
|
; Bridson and Verdcourt, 1988
; see Table 1). Different accessions of Nauclea were used in the rbcL and restriction enzyme studies and of Vangueria for rbcL and ITS. Restriction site information from Vangueriopsis was substituted for Vangueria.
Robbrecht's classification (Robbrecht, 1988, 1993
) will be followed when discussing the results of our analyses, unless otherwise stated.
Missing taxa
For some taxa we do not have information from all of the four data sets. The rbcL sequence data set consists of 82 taxa, and, of these, 75 are represented in the morphological data set, 26 in the restriction enzyme data set, and 21 in the ITS sequence data set. As a consequence of the missing taxa, 
30% of the data entries in the combined data matrix are missing. Combining the data sets despite the lack of overlap in taxa represented was preferred since it allowed us to consider all available information simultaneously, and for all of the taxa. Analyses on a subset of the taxa may result in spurious relationships due to incomplete sampling and long branch attraction (Huelsenbeck and Hillis, 1993
).
Morphology
The base for the morphological matrix was a compilation of Ixoroideae characters from Robbrecht and Puff (1986)
and investigations of herbarium and fresh material. Additional information was compiled from literature (see Andreasen and Bremer, 1996
). For taxa not included in Andreasen and Bremer (1996)
, the following sources were also consulted: Bentham and Hooker, 1862–1883
; Hemsley, 1916
; Pitard, 1922–1924
; Cuatrecasas, 1953
; Cavaco, 1965
; Heine and Hallé, 1970
; Ridsdale, van den Brink, and Koek-Noorman, 1972
; Verdcourt, 1981, 1983
; Tirvengadum, 1983
; Puff, Robbrecht, and Randrianasolo, 1984
; Robbrecht, 1984
; Bridson and Robbrecht, 1985b
; Tirvengadum and Robbrecht, 1985
; Ali and Robbrecht, 1991
; Owens et al., 1993
; Puff and Rohrhofer, 1993
; Puff et al., 1996
; De Block, 1997, 1998
; De Block and Robbrecht, 1997)
. The 49 characters (Appendix 1) used in the parsimony analysis are included in the morphological matrix (Appendix 2). Genera were used as terminal taxa unless the assumption of monophyly could be questioned. In those cases the sequenced species was used as the terminal taxon in the morphological matrix (for further discussion see Andreasen and Bremer, 1996
).
Molecular methods
Total DNA was extracted from fresh or silica gel dried leaf material using the standard CTAB procedure (Saghai-Maroof et al., 1984
; Doyle and Doyle, 1987
). In the case of Heinsenia, herbarium material was used. The DNA was further purified with cesium chloride/ethidium bromide gradient centrifugation or QiaQuick's polymerase chain reaction (PCR) purification kit (QIAGEN, Valencia, California, USA).
rbcL sequence data
Of the 82 rbcL sequences in this study 24 are new (see Table 1 for EMBL number, voucher information for new sequences, and references for earlier published sequences). The rbcL region was amplified using 5' and 3' PCR (polymerase chain reaction) primers designed by Olmstead et al. (1992)
, or, for some problematic DNAs and for automated sequencing, internal primers designed by G. Zurawski (DNAX Research Institute, Palo Alto, California, USA). For manual sequencing the double-stranded DNA was asymmetrically amplified to obtain single-stranded DNA (Kaltenboeck et al., 1992
). Single-stranded DNA was cleaned with QiaQuick's PCR purification kit and sequenced with the standard dideoxy chain termination reaction (Sanger, Nicklen, and Coulson, 1977
), using the following internal primers (listed with permission from G. Zurawski), numbered by the corresponding first position of rbcL of Zea mays (R = reverse): rbcLz-1 5'-ATGTCACCACAAACAGAAACTAAAGCAAGT, rbcLz-234 5'-CGTTACAAAGGACGATGCTACCACATCGA, rbcLz-427 5'-GCTTATTCAAAAACTTTCCAAGGCCCGCC, rbcLz-674 5'-TTTATAAATCACAAGCCGAAACTGGTGAAATC, rbcLz-895 5'-GCAGTTATTGATAGACAGAAAAATCATGGT, rbcLz-1020 5'-ACTTTAGGTTTTGTTGATTTATTGCGCGATGATT, rbcLz-1204 5'-TTTGGTGGAGGAACTTTAGGACACCCTTGGGG, rbcLz-346R 5'-AAATACGTTACCCACAATGGAAGTAAATAT, rbcLz-674R 5'-GATTTCGCCTGTTTCGGCTTGTGCTTTATAAA, rbcLz-1020R 5'-ATCATCGCGCAATAAATCAACAAAACCTAAAGT, rbcLz-1204R 5'-CCCTAAGGGTGTCCTAAAGTTTCTCCACC, rbcLz-1375R 5'-AATTTGATCTCCTTCCATATTTCGCA. Both strands were sequenced to obtain at least partial sequence overlap. Five DNAs were sequenced with automated sequencing: Emmenopterys, Dictyandra, Tarenna supra-axillaris, Hyperacanthus, and Coddia. For these, double-stranded DNA was cycle sequenced using Perkin-Elmer's FS kit and GeneAmp PCR Systems 9600 (Perkin-Elmer Applied Biosystems, Foster City, California, USA). Ethanol precipitation was used to clean the sequenced products before loading the samples on an ABI Prism 377 Automated Sequencer (Perkin-Elmer Applied Biosystems). Nucleotide positions 27 to 1428 in the resulting rbcL sequences were used in the subsequent phylogenetic analyses. The variable 3' end of rbcL after position 1428 was aligned manually, and blocks of nucleotides were recoded into present/absent characters (Fig. 2). Using blocks, rather than the actual nucleotides, were preferred since the alignment was difficult between more divergent taxa. For some parts of the matrix, certain blocks of nucleotides were assessed as homologous and used as characters in the analyses. If taxa had just a few nucleotides in the 3' region, the alignment was deemed too uncertain for inclusion of these nucleotides in any of the block characters (e.g., "AAA" in Gardenia augusta-Genipa). By not including the actual nucleotides in the analyses we may lose information, but by including them we would run the risk of incorrectly grouping taxa based on uncertain homologies. For Ixora biflora, Duroia, and Scyphiphora we were unable to obtain sequences of the 3' end of rbcL.
|
) and ITS 4 (White et al., 1990
) and the PCR ingredients and parameters followed Baldwin (1992)
. The amplified DNA was sequenced using automated sequencing as for the six rbcL sequences. Both strands were sequenced using sequencing primers ITS 5 (White et al., 1990
), or ITS 1-TM (McDowell and Bremer, 1998
), ITS4, ITS 3, and ITS 2 (White et al., 1990
). For details about the sequencing procedure and for criteria used in alignment of the ITS sequences see Andreasen, Baldwin, and Bremer (1999)
.
Restriction enzyme data
Of the taxa in Table 1, 26 were included in the restriction enzyme study (indicated with * in Table 1). The purified DNA was digested with nine restriction endonucleases (EcoRV, ClaI, AvaI, BclI, DraI, BamHI, NcoI, SacI, and HindIII) following the specifications of the manufacturer. The DNA fragments were separated by agarose electrophoreses (1% gels) and transferred to nylon filters by bidirectional blotting (Palmer, 1982, 1986
). The filters were hybridized with 16 cloned chloroplast probes from Lactuca (Table 2, see also Table 2 in Jansen and Palmer, 1987
). Probes were labelled with 
-32P dATP via nick translation. Hybridizations were carried out at 65°C and the DNA fragments were visualized by autoradiography. Fragments <0.4 kb were not visualized. The lengths of the DNA fragments were estimated by comparing them to lambda DNA fragments with known length. For each restriction enzyme, a restriction map containing all taxa was constructed and the sites of the different taxa were aligned relatively to each other. The overlap hybridization method described in Palmer (1986)
was used to order the fragments. Restriction site occurrences and absences were used as characters (Table 3). For Luculia, Vangueria, Tarenna, Ixora, Coffea, Mitriostigma, and Gardenia character data from the restriction enzyme study of Bremer and Jansen (1991)
were added to the matrix (characters 38–61; data matrix with the 61 informative characters in Appendix 3). No informative sites were identified for SacI and the rbcL region was not included in the analyses.
|
|
), using heuristic methods and with all character changes weighted equally. The following options were used: TBR branch swapping, Mulpars on, Steepest descent off, random addition sequences (at least 100 replicates). The partition homogeneity test (Farris et al., 1995
) (1000 replicates, Mulpars off, five TBR replicates) was used to assess incongruencies between the matrices. Support for the groupings in the trees was evaluated using jackknife analysis (Farris et al., 1996
) with 10 000 replicates, Mulpars off, and five TBR replicates for the combined and the restriction site analyses. In addition, Bremer support values, b (Bremer, 1988, 1994
; Källersjö et al., 1992
), were obtained using PAUP* 4.0b2 and the program Autodecay 4.0.1 (Eriksson, 1998
) with ten replicates of random addition for each constraint tree. The program MacClade (Maddison and Maddison, 1992
) was used for ACCTRAN reconstruction of character evolution and MacVector 4.1 (Kodak Scientific Imaging Systems, New Haven, Connecticut, USA) for translation of rbcL DNA sequences into amino acids. |
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
The rbcL analysis (excluding the variable 3' end of the gene) resulted in 134 978 equally parsimonious trees, 692 steps long (CI = 0.37, and retention index, RI = 0.67). The strict consensus tree of these trees is shown in Fig. 3. The restriction enzyme data analysis gave 3228 trees (CI = 0.59, RI = 0.72) and the consensus is rather unresolved (Fig. 4). The morphological matrix resulted in 8827 trees but because of the low number of characters (49) in comparison to the number of taxa (75) there is little overall support, and the strict consensus tree is very unresolved (not shown). The analysis of the 21 ITS sequences resulted in nine most parsimonious tree (98 informative characters plus 13 gap characters coded as present or absent; trees are 386 steps, CI = 0.48, RI = 0.44; see Andreasen, Baldwin, and Bremer, 1999, for more information). The partition homogeneity test between submatrices ITS, morphology, and chloroplast gave a P value of 5%. Combining the data sets (including the recoded 3' end characters of rbcL) gave 110 equally parsimonious trees, 1830 steps long (CI = 0.44, RI = 0.60). The strict consensus tree is shown in Fig. 5. The resolution of this combined tree is much higher than when the data sets are analyzed separately.
|
|
|
) is strongly supported (Fig. 5; jackknife 100%, b = 10) and so is the Ixoroideae s.s. taxa (jackknife 100%, b = 10). Other major groupings, however, are not strongly supported, except for the clade including Coffeeae to Gardenieae (jackknife 99%, b = 8). Mussaenda, Pentagonia, and Emmenopterys of the Cinchonoideae are basal in Ixoroideae s.l. and genera of the subfamily Antirheoideae (Alberta, Canthium, and Vangueria) are nested within the Ixoroideae s.s. There are clades corresponding to the tribes Vanguerieae (jackknife 100%), Pavetteae (excluding Ixora and allies), Coffeeae (including four taxa from Gardenieae-Diplosporinae), and Gardenieae-Gardeniinae. Octotropideae form one monophyletic group with the genus Cremaspora (Gardenieae-Diplosporinae) as their sister.
Within the abovementioned clades there is strong support for the grouping of Myonima, Versteegia, and the Ixora species (jackknife 100%, b = 9) Leptactina and Dictyandra (Pavetteae; b = 10, jackknife 100%), and Coffea, Paracoffea, and Psilanthus (= Coffeeae s.s.; jackknife 99%, b = 5). In Gardenieae the African sister genera Oxyanthus and Mitriostigma have strong support (jackknife 100%, b = 10) and the neotropical genera Glossostipula, Duroia, Alibertia, and Borojoa form a well-supported group (jackknife 93%, b = 5).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
investigated how nonrandom missing entries (i.e., taxa lacking certain types of data) affect the accuracy of estimating a true phylogeny or a tree obtained using a complete data set. They found that three factors affect the accuracy (the probability of estimating the true phylogeny) of the analyses: the type of data missing, the total number of characters in the analysis, and the proportion of characters lacking data. The latter parameter had surprisingly little effect on accuracy in Wiens and Reeder's analyses. Accuracy was very similar when taxa were 50, 75, or 100% complete. In comparison, the taxon completeness in our analysis varied within that range: 57–100%. Wiens and Reeder's results suggest that the larger the number of characters in a combined analysis, the smaller the cost in accuracy associated with combining data sets with nonoverlapping taxon sampling. No decrease in accuracy occurred when including 50% incomplete taxa in data matrices with 174 characters. Even if further studies are needed to assess this effect fully, our data set includes a comparatively large number of informative characters (416), so the decrease in accuracy can be assumed to be small, especially given comparatively high taxon overlap between data sets.
Support and incongruence
By using all available information, though incomplete for some taxa, and analyzing it simultaneously we maximize the information content of the analysis (i.e., "simultaneous analysis," Nixon and Carpenter, 1997
; also called "total evidence," Kluge, 1989
). Our results suggest that combining data results in more robustly supported and resolved topologies (compare Figs. 3 and 4 with Fig. 5). For example, Coffeeae s.s. are supported by a jackknife fraction of 75% in the rbcL analysis and, in the combined analysis, the support is increased to 99%. The partition homogeneity test resulted in a P value of 5%, which would indicate that the data matrices are incongruent if = 0.05 is the chosen cut-off level. Inspecting the separate trees, however, reveals no well-supported topological incongruencies, which is not very surprising since the consensus trees from the separate analyses are fairly unresolved. In addition, there is no decrease in support values for any of the clades between the separate and combined trees, indicating that there is no "real" incongruence caused by biological phenomena, e.g., hybridization, but the incongruent topologies are most likely a result of sampling error. Earlier studies (Sullivan, 1996
; Cunningham, 1997
) have suggested that the 0.05 level is indeed too conservative for real data sets and that P < 0.001 is more reasonable. The assumption for the homogeneity test is that the characters are drawn at random from an indefinite universe of characters, an assumption we know is violated, e.g., when using particular DNA regions with possible evolutionary constraints.
Generally, there is least support for intertribal relationships and also for larger clades within Gardenieae-Gardeniinae. The cause for the low support is the low number of characters at these nodes, which is especially pertinent when looking at rbcL sequence data only (the collapsed consensus tree in Fig. 3). This pattern of few characters at intertribal branches might indicate comparatively fast speciations (branchings per unit of time) at the time when these groups originated or that mutations produced were not fixed but have changed again later on (homoplastic characters).
Phylogenetic implications
Our results suggest that the Cinchonoideae genera (Pentagonia, Mussaenda, and Emmenopterys; Robbrecht, 1988, 1993) are part of the basal branches of the ingroup ("basal Ixoroids," see Fig. 5), thus supporting results of previous analyses (Bremer, Andreasen, and Olsson, 1995
; Bremer and Thulin, 1998
; Bremer et al., 1999
). The relationships of these "basal Ixoroids" will be addressed in future analyses by including more taxa. However, a new and noteworthy relationship with this group is shown by the genus Posoqueria, which has been included as an aberrant member in Gardenieae (Robbrecht, 1988
). Exclusion of Posoqueria from Gardenieae and Ixoroideae s.s. is strongly supported here (jackknife 100%, b = 10 steps). The present analysis suggests that Posoqueria is a member of or closely related to the recently resurrected tribe Mussaendeae (Bremer and Thulin, 1998
), but the exact position must await more inclusive taxon sampling.
Our analysis supports the earlier findings that the subfamily Antirheoideae is polyphyletic (Bremer, 1996
; Rova and Andersson, 1995
; Andersson, 1996
; Delprete, 1996
; Young et al., 1996
; Andersson and Rova, 1999
) since the Antirheoideae tribes Vanguerieae and Alberteae are nested in the subfamily Ixoroideae. Initially, Bremekamp (1934)
placed Alberteae (including Cremaspora and some Octotropideae and Aulacocalyceae genera) in Ixoroideae. Later (1952) he excluded Alberta and Nematostylis from the subfamily because he thought they lacked the secondary pollen mechanism (see Puff, Robbrecht, and Randrianasolo, 1984
). Verdcourt (1958)
, on the other hand, limited Alberteae to Alberta and Nematostylis and placed it close to "Ixoreae". In our analysis it is sister to Vanguerieae and clearly part of Ixoroideae. Morphological characters supporting this sister relationship are, e.g., the superior embryo radicles and endocarps with apical splits.
The Gardenieae
The exclusion of representatives of Gardenieae-Diplosporinae from Gardenieae is strongly supported (40 extra steps are required to make the taxa of the Diplosporinae a monophyletic group and part of the Gardenieae clade). This was shown for Tricalysia and Cremaspora in earlier studies (Andreasen and Bremer, 1996
; Persson, 1996, 1998
) and now for Diplospora, the type genus of Diplosporinae and also for Scyphiphora (Fig. 5). These Diplosporinae genera are found in three different clades in the trees: in the Coffeeae, close to Octotropideae, and the Ixora group, respectively (see below).
Three taxa placed in other tribes are part of the Gardenieae clade in our analysis. One is the genus Duperrea, which was included in Pavetteae as a genus with unknown affinity (Bridson and Robbrecht, 1985a
). De Block and Robbrecht (1997)
found that members of this Indo-Chinese genus lack seed coats and also adaxial excavation in the seeds, characteristics that are typical for members of Pavetteae. Instead, the seeds are embedded in placental pulp, a feature typical of Gardenieae-Gardeniinae. The relationship of Duperrea with Gardenieae-Gardeniinae is corroborated in our analysis (Fig. 5), although the support is low.
Aoranthe is an African genus which was segregated from the Asian Porterandia and placed in the tribe Isertieae (Somers, 1988
). This segregation is supported in the present study, although both Aoranthe and Porterandia are part of the Gardenieae clade. Andersson (1996)
and Bremer and Thulin (1998)
found it to be part of Ixoroideae s.s., but did not address the question of its position within the subfamily further. Persson (1996, 1998)
also found it to be part of Gardenieae, but in the Gardenia group and not close to Oxyanthus and Mitriostigma as in our study. Aoranthe shares with Oxyanthus and Mitriostigma the characters of pseudo-axillary flowers, presence of domatia, and the sculptured thickenings of the exotesta.
Heinsenia was placed, together with its close ally Aulacocalyx and four other genera, in tribe Aulacocalyceae by Robbrecht and Puff (1986)
. Morphologically, Aulacocalyceae shares characteristics with the Gardenieae-Gardeniinae in the terminal inflorescences and fruits with seeds embedded in placental pulp, but differs in having a superior embryo radicle and lacking a seed coat. Despite these differences, our results support Heinsenia as part of the Gardenieae-Gardeniinae. Although support is low, it is interesting to note the sister relationship of Heinsenia with Rothmannia in the analysis. The branching pattern of Heinsenia and Rothmannia is very similar, and the corolla of both genera is often white with colored spots inside (Keay, 1958
; Bridson and Verdcourt, 1988
). As recognition of the tribe Aulacocalyceae would render the Gardenieae-Gardeniinae paraphyletic, the best solution is to sink Aulacocalyceae into Gardenieae.
Within the Gardenieae-Gardeniinae Robbrecht and Puff (1986)
suggested the following three informal groups based on morphological similarities: (1) the taxa with pollen grains in tetrads, (2) a group of neotropical genera with unisexual flowers, and (3) Aidia and allies with small, spherical fruits.
The first group (1) with pollen in tetrads (character 44, Appendix 1) was supported in a morphological analysis by Persson (1996)
. Later, based on chloroplast data, he found that this character must have arisen at least three times (Persson, 1998
). In the present analysis, taxa with pollen in tetrads are found in essentially the same three groups as found by Persson (1998)
(see Fig. 5): in Gardenia, which groups with the monad genera Genipa, Kailarsenia, and Coddia; in the Oxyanthus group; and in a group consisting of Euclinia, Calochone, Randia, Casasia, and Rosenbergiodendron (the last genus with pollen in monads). In addition, Massularia has pollen in larger clumps, i.e., massulae. The relationships within Gardenieae are rather unresolved and weakly supported, but reconstructing this character requires three (four with Massularia) independent origins in all the equally parsimonious trees.
The genera Duroia, Alibertia, Borojoa, and Genipa have unisexual flowers (group 2, character 8, Appendix 1) and were proposed to form a natural, neotropical group in Gardenieae. The three first genera form a clade (jackknife 93%, b = 5) including Glossostipula. Glossostipula also has unisexual flowers and is a newly described neotropical genus for some taxa previously referred to Genipa and Randia (Lorence, 1986
). The sequenced Genipa species, Genipa americana, is not part of that clade, a result congruent with the analyses presented by Persson (1996, 1998)
, which also indicated that Genipa, as currently circumscribed, may be polyphyletic. Genipa americana comprises a clade together with Gardenia and Kailarsenia.
Aidia and Oxyceros are representatives for the third group with small, spherical fruits (the "Aidia group" sensu Persson, 1996
). They form a clade with the Australian Randia moorei and the African genus Hyperacanthus but the support is low (jackknife 65%, b = 2). Thorns occur in Randia moorei, Oxyceros, and in Hyperacanthus, and Aidia, Oxyceros, and Hyperacanthus all have porate pollen and endotesta with sculptured thickenings.
Randia is today commonly accepted as a strictly neotropical genus (e.g., Fagerlind, 1942
; Tirvengadum, 1978
) and two of the included species (R. aculeata and R. truncata) belong to Randia s.s. Many of the paleotropical species need further studies, e.g., the New Guinean species included in the analysis, Randia demicostata, is sister to Sukunia, which is proposed to be close to Trukia (Smith and Darwin, 1988
). Randia fitzalanii, an Australian species that is part of this clade has been placed in Trukia (Fosberg, 1987
). Recently, Puttock (1999)
and Puttock and Quinn (1999)
performed morphological analyses of Asian-Australian Gardenieae and considered it appropriate to include Trukia, Sukunia, and possibly Porterandia in Atractocarpus. Our results support these relationships, although our analysis includes a more limited sampling compared to Puttock and Quinn's analysis.
The Coffeeae
The tribe Coffeeae was restricted to Coffea and Psilanthus (Paracoffea in our analysis is usually included in Psilanthus) by Robbrecht and Puff (1986)
: Tricalysia and related genera, which were considered to be closely related to Coffea (Bremekamp, 1934
), were moved to the subtribe Diplosporinae of Gardenieae by Robbrecht and Puff (1986)
. According to our results, Coffeeae should include not only Tricalysia and Diplospora (Gardenieae-Diplosporinae) but also Bertiera (Gardenieae-Gardeniinae). The position of Bertiera in this clade was indicated in previous analyses (Andreasen and Bremer, 1996
), and our analyses presented here include the type species, B. guianensis, strengthening the inclusion of Bertiera in the Coffeeae. The Coffeeae clade is moderately supported (jackknife 76%, b = 4).
The position of Diplospora in the Coffeeae clade is new, but since Tricalysia and Diplospora have been considered congeneric (Schumann, 1891)
, the position of Diplospora close to Tricalysia is not surprising. The inclusion of these Diplosporinae taxa in Coffeeae is fairly strongly supported (jackknife 86%, b = 3). Diplospora was recently split into Diplospora s.s. and Discospermum (Ali and Robbrecht, 1991
), and the sequenced species, Diplospora polysperma, was considered to be a Discospermum by these authors. Ali and Robbrecht (1991)
refrained from making a combination under that genus, however, because the origin of the material was unknown. We have coded it as Discospermum in the morphological matrix. Discospermum has much larger fruits than Diplospora, with many seeds frequently embedded in placental tissue, a feature reminiscent of the type of fruit found in many genera of Gardenieae-Gardeniinae. Thus, Ali and Robbrecht (1991)
stated that Discospermum "links the Diplosporinae with the Gardeniinae and supports the rank (subtribal) given to these." Our results do not no support such a relationship (Fig. 5), since at least some genera of Diplosporinae belong in Coffeeae. Sericanthe, which was recently separated from Tricalysia (Robbrecht, 1978
), probably belongs in this group too since the two genera was considered closely related (Robbrecht, 1978
).
The Octotropideae
All the included taxa of the Octotropideae form a monophyletic group in the combined tree (Fig. 5). The internal groupings are not well supported, but the presumed "primitive" genera (Robbrecht and Puff, 1986
) with many seeds, Fernelia and Ramosmania, are supported as sister taxa in our results (jackknife 81%, b = 3). The sister-group relationship between Pouchetia and Hypobathrum is supported (jackknife 69%), and they were both placed in the "central group" (Robbrecht and Puff, 1986
). This group included genera with small fruits, relatively few ovules, and similar thickenings in the exotesta.
A member of the Gardenieae-Diplosporinae, Cremaspora, is sister to tribe Octotropideae, although the support is low (60%, b = 3). The taxonomic history of this genus is complex, partly because the orientation of its embryo radicle was misinterpreted (e.g., Schumann, 1891)
. Bremekamp (1934)
placed Cremaspora in the tribe Cremasporeae together with some genera that Robbrecht (1988, 1993)
included in Octotropideae and Aulacocalyceae. Verdcourt (1958)
included it in "Ixoreae"-Cremasporinae, but excluded the Aulacocalyceae genus Heinsenia (discussed above). As noted by Ali and Robbrecht (1991)
and others, there are strong similarities between the taxa placed in Gardenieae-Diplosporinae and Octotropideae. While such similarities support inclusion of Cremaspora in Octotropideae, the position of Cremaspora is not well supported in our analyses, and we suggest a placement in a tribe of its own, and resurrect Bremekamp's (1934)
Cremasporeae.
The Pavetteae
The Pavetteae in its current circumscription (Robbrecht, 1993
) are paraphyletic according to our results (Fig. 5). The included taxa of this tribe are found in two clades (excluding Duperrea, which, as discussed above, is part of the Gardenieae-Gardeniinae clade), which conform to the two informal groups proposed by Bridson and Robbrecht (1985a)
: Ixora, Myonima, and Versteegia in one clade (see below) and the rest of the taxa, Pavetteae s.s., higher up in the tree as sister to Gardenieae. The support for the clade consisting of Coffeeae to Gardenieae, including Pavetteae s.s. is strong (jackknife 99%, b = 8). If monophyly of Pavetteae is included as a constraint in the analysis, 29 extra steps are required, which is fairly high compared to the Bremer support values of up to 14, which were obtained in the analyses. Within Pavetteae s.s., the close relationship between Leptactina and Dictyandra (Robbrecht, 1984
) is strongly suggested (jackknife 100%, b = 10). Tarenna s.l. (Bridson, 1979
) is a heterogeneous genus as both Pavetta and Rutidea are nested within it. One of the characters used for delimiting Pavetteae (Robbrecht, 1984
), seeds with adaxial excavation and placenta within the seed cavities (character 39) must, according to the analysis, be interpreted as a parallelism occurring in this group and also in Ixora and allies.
The grouping of Ixora with the genera Myonima and Versteegia, is strongly supported (bootstrap 100%, b = 9), and this clade occupies a basally diverged position in the Ixoroideae s.s. These genera exhibit many common characteristics and are considered to be closely related (Ridsdale, van den Brink, and Koek-Noorman, 1972
; Verdcourt, 1983
; Jansen, De Block, and Smets, 1997
). Versteegia is a cauliflorous genus from New Guinea with salmon-pink flowers, a flower color that is uncommon in Rubiaceae but found also in Ixora. Myonima differs from Ixora in having a much shorter corolla tube and frequently three- to seven-locular ovary. Supporting characters in the analysis are: four corolla lobes (character 13), elongated anthers (character 20), and the parallelism shared with Pavetta and allies, seeds with adaxial excavation (see above). Some of the other genera included in the group around Ixora (Bridson and Robbrecht, 1985a
; Table 5 in Robbrecht and Puff, 1986
) probably also belong here.
The sister taxon to Ixora and allies (the Ixora group) in the combined analysis is the mangrove genus Scyphiphora. Puff and Rohrhofer (1993)
investigated the morphology in this problematic taxon and suggested a position in Ixoroideae s.s., most probably in the Gardenieae-Diplosporinae group based on the axillary inflorescences and pauciovulate carpels. Recent analyses by Bremer et al. (1999)
confirm the position of Scyphiphora in the Ixoroideae, although few taxa from the subfamily were included. The position of Scyphiphora as sister to the Ixora group in the trees resulting from the combined analysis conflicts with the position close to Vanguerieae inferred in the rbcL trees. The support for the relationship with the Ixora group is not very strong (jackknife 70%, b = 3), and the support in the separate rbcL analysis even lower. A character supporting the position of Scyphiphora as sister to the Ixora group are the number of corolla lobes (four; character 13, Appendix 1), an uncommon feature in the subfamily. A position close to any of the Gardenieae-Diplosporinae taxa is strongly rejected as the support is high (jackknife 99%, b = 8) for the upper clade including the Gardenieae-Diplosporinae taxa and Coffeeae, Octotropideae, Pavetteae s.s., and Gardenieae s.s. We tentatively include Scyphiphora in Ixoreae.
The subfamily Ixoroideae and the tribes included in it are in need of new circumscriptions based on the results of our analyses. The exact delimitation of the subfamily (i.e., Ixoroideae s.l., see, e.g., Bremer and Thulin, 1998
; Bremer et al., 1999
) will not be dealt with here, as more taxa must be investigated before stable conclusions can be drawn. However, the results of the present study suggest the following new tribal delimitations in the Ixoroideae s.s. based on the included taxa.
Coffeeae DC., Ann. Mus. Hist. Nat. (Paris) 9: 217 (1807). Type: Coffea L.
Shrubs, treelets, rarely trees, or geofrutices. Stipules interpetiolar, entire. Raphides absent. Inflorescences axillary and paired at nodes, or rarely terminal. Corolla lobes contorted to the left. Secondary pollen presentation absent (Coffea, Psilanthus, probably also in Diplospora and Discospermum) or present (Tricalysia and Bertiera). Stigmatic lobes free over most of their length (except in Bertiera). Ovary two-locular, placenta axile with one to many ovules. Fruit berry-like or with more or less dry wall and placental outgrowths. Embryo radicle inferior or lateral. Endosperm entire (ruminate in some species of Tricalysia). Exotesta cells without thickenings, with thickenings along outer tangential walls, or along radial and inner tangential walls (Bertiera). Pollen grains in monads, three-to-four colporate.
Genera included: Coffea, Psilanthus (including Paracoffea), Diplospora, Discospermum, Tricalysia, Bertiera. Genus that probably belongs here: Sericanthe.
Cremasporeae Bremek. ex S. P. Darwin, Taxon 25: 601 (1976). Type: Cremaspora Benth.
Shrubs, treelets, or sometimes lianas. Stipules interpetiolar, entire, deciduous. Raphides absent. Inflorescences axillary and paired at nodes. Corolla lobes contorted to the left. Secondary pollen presentation present. Stigmatic lobes fused over most of their length. Ovary two-locular, axile placenta with one ovule in each locule. Fruit with leathery wall and (1)–2 seeds. Embryo radicle inferior. Endosperm entire. Exotesta with thickenings in radial walls. Pollen grains in monads, three-colporate.
Genus included: Cremaspora.
Gardenieae A. Rich. ex DC., Prodr. 4: 342, 367 (1830). Type: Gardenia Ellis.
Shrubs, trees, or lianas. Stipules interpetiolar, entire. Raphides absent. Inflorescences terminal, rarely pseudo-axillary, very rarely truly axillary and paired at nodes. Aestivation contorted to the left, rarely to the right. Secondary pollen presentation present. Stigma club shaped (dividing lobes rare). Ovary with 2(–9) carpels, axile or parietal placentas with one to many ovules. Fruit with more or less dry wall, or berry-like, with seeds imbedded in the fleshy placenta. Embryo radicle orientation variable. Endosperm entire. Seeds mostly exotestal (seed-coat reduced in Duperrea, and Heinsenia), usually with thickenings along radial and inner tangential wall. Pollen grains in monads or sometimes in tetrads, three-porate or three-colporate.
Genera included:
From Gardenieae-Gardeniinae: Aidia, Alibertia, Borojoa, Burchellia, Calochone, Casasia, Coddia, Didymosalpinx, Duroia, Euclinia, Gardenia, Genipa, Glossostipula, Hyperacanthus, Kailarsenia, Massularia, Mitriostigma, Oxyanthus, Oxyceros, Porterandia, Randia, Rosenbergiodendron, Rothmannia, Schumanniophyton, Sukunia.
From Aulacocalyceae: Heinsenia.
From Pavetteae: Duperrea.
Genera that probably should be included: most of Robbrecht's Gardenieae-Gardeniinae (1988, 1993) except those mentioned below, and Aulacocalyx.
Genera excluded: Gardenieae-Gardeniinae: Posoqueria, exact position still uncertain, in vicinity of Mussaendeae. Bertiera should be moved to Coffeeae.
Gardenieae-Diplosporinae: Tricalysia, Diplospora, and Discospermum to Coffeeae. Cremaspora to Cremasporeae, Scyphiphora to be tentatively accommodated in Ixoreae.
Ixoreae A. Gray, Proc. Amer. Acad. Arts 4: 39 (1858). Type: Ixora L.
Shrubs, or small trees. Stipules interpetiolar, entire. Petioles articulate. Raphides absent. Inflorescences terminal. Aestivation contorted to the left. Secondary pollen presentation usually present. Stigmatic lobes free over most of their length. Ovary 2(–7) locular, axile placentas with one ovule per carpel. Fruits fleshy. Seeds with adaxial excavation. Embryo radicle inferior. Endosperm entire. Seeds exotestal, with thickenings (as anastomosing ribs) along outer tangential wall, or rarely absent. Pollen grains in monads, three-colporate.
Genera included: Ixora, Myonima, Versteegia.
Genera that probably belong here: Captaincookia, Doricera. Scyphiphora is tentatively included here although the description is not altered to include its characteristics.
Pavetteae A. Rich. ex Dumort., Anal. Fam. Pl. (Rubiaceae): 33 (1829). Type Pavetta L.
Shrubs, small trees, or lianas. Stipules interpetiolar, entire (fimbriate in Rutidea). Raphides absent. Inflorescences terminal or sometimes pseudo-axillary. Aestivation contorted to the left. Secondary pollen presentation usually present. Stigmatic lobes fused over most of their length. Ovary with two locular, axile placentas with one to many ovules. Fruits fleshy. Seeds with adaxial excavation. Embryo-radicle inferior or lateral. Endosperm entire or slightly to deeply ruminate. Seeds exotestal, with thickenings (as a continuous plate) along outer tangential wall, or absent. Pollen grains in monads, 3–4(–5)-colporate.
Genera included: Pavetta, Rutidea, Tarenna, Leptactina, Dictyandra.
Genera that probably belong here: Cladoceras, Tennantia, Coleactina.
Genera excluded: Ixora, Myonima, and Versteegia to Ixoreae. Duperrea to Gardenieae.
Genera that probably should be excluded: Captaincookia, Doricera.
In conclusion, the combined analysis resulted in higher degree of resolution than when the data sets were analyzed alone, supporting the view that, when possible, all available data should be analyzed simultaneously to maximize explanatory power (see, e.g., Nixon and Carpenter, 1997
). By adding other, independent data sets (even if taxon sampling is not complete) greater resolution is achieved, exemplifying that combined analyses can remedy the incapacity of one data set to resolve a phylogeny.
The results of the combined analyses suggest several new and important relationships in the subfamily Ixoroideae, and, as a result, new circumscriptions at tribal level are needed. The tribe Ixoreae should be resurrected for Ixora and allies, which are excluded from Pavetteae. Scyphiphora is probably related to Ixoreae and we tentatively include it there. Heinsenia, representing Aulacocalyceae, is part of Gardenieae, as is Duperrea (earlier placed in Pavetteae) and Aoranthe (Isertieae). There is strong support for the inclusion of the Gardenieae-Diplosporinae taxa Diplospora, Discospermum, Tricalysia, and also Bertiera (Gardenieae-Gardeniinae) in Coffeeae. Cremaspora is sister to Octotropideae but is best housed in a tribe of its own, Cremasporeae. The inclusion of the two Antirheoideae tribes Vanguerieae and Alberteae in Ixoroideae is strongly supported, as is the exclusion of Posoqueria from Gardenieae and Ixoroideae s.s.
|
|
|
|
|
|
FOOTNOTES |
|---|
3 Author for reprint requests, current address: The Jepson Herbarium and Department of Integrative Biology, 1001 Valley Life Sciences Bldg. 2465, University of California, Berkeley, California 94720-2465 (Katarina.Andreasen{at}systbot.uu.se
).
|
LITERATURE CITED |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Andersson, L. 1995 Tribes and genera of the Cinchoneae complex (Rubiaceae). Annals of the Missouri Botanical Garden 82: 409–427.[CrossRef][ISI]
———. 1996 Circumscription of the tribe Isertieae (Rubiaceae). Opera Botanica Belgica 7: 139–164.
———, and J. H. E. Rova. 1999 The rps16 intron and the phylogeny of the Rubioideae. Plant Systematics and Evolution 214: 161–186.[CrossRef][ISI]
Andreasen, K., and B. Bremer. 1996 Phylogeny of the subfamily Ixoroideae (Rubiaceae). Opera Botanica Belgica 7: 119–138.
———, B. G. Baldwin, and B. Bremer. 1999 Phylogenetic utility of the nuclear rDNA ITS region in the subfamily Ixoroideae (Rubiaceae): comparisons with cpDNA rbcL sequence data. Plant Systematics and Evolution 217: 119–135.[CrossRef][ISI]
Baldwin, B. G. 1992 Phylogenetic utility of the internal transcribed spacers of nuclear ribosomal DNA in plants: an example from the Compositae. Molecular Phylogenetics and Evolution 1: 3–16.[CrossRef][Medline]
Bentham, G., and J. D. Hooker. 1862–1883 Genera Plantarum. Lovell, Reeve & Co., Williams & Norgate, London, UK.
Bremekamp, C. E. B. 1934 A monograph of the genus Pavetta L. Repertorium Specierum Novarum Regni Vegetabilis 37: 1–208.
———. 1952 The African species of Oldenlandia L. sensu Hiern et K. Schumann. Verhandelingen der Koninklijke Nederlandsche Akademie van Wetenschappen, Afdeeling Natuurkunde, Tweede sectie 18: 11–24.
———. 1966 Remarks on the position, the delimitation and the subdivision of the Rubiaceae. Acta Botanica Neerlandica 15: 1–33.
Bremer, B. 1996 Phylogenetic studies within Rubiaceae and relationships to other families based on molecular data. Opera Botanica Belgica 7: 33–50.
———. 1997 Combined and separate analyses of morphological and molecular data in the plant family Rubiaceae. Cladistics 12: 21–40.[ISI]
———, K. Andreasen, and D. Olsson. 1995 Subfamilial and tribal relationships in the Rubiaceae based on rbcL sequence data. Annals of Missouri Botanical Garden 82: 383–397.[CrossRef][ISI]
———, and R. K. Jansen. 1991 Comparative restriction site mapping of chloroplast DNA implies new phylogenetic relationships within Rubiaceae. American Journal of Botany 78: 198–213.[CrossRef][ISI]
———, ———, B. Oxelman, M. Backlund, H. Lantz, and K.-J. Kim. 1999 More characters or more taxa for a robust phylogeny—a case study from the coffee family (Rubiaceae). Systematic Biology 48: 413–435.[CrossRef][ISI][Medline]
———, and M. Thulin. 1998 Collapse of Isertieae, re-establishment of Mussaendeae, and a new genus of Sabiceeae (Rubiaceae): phylogenetic relationships based on rbcL data. Plant Systematics and Evolution 211: 71–92.[CrossRef][ISI]
Bremer, K. 1988 The limits of amino acid sequence data in angiosperm phylogenetic reconstruction. Evolution 42: 795–803.
= taxa with pollen in tetrads. Support for nodes is indicated (Bremer support >1 above branches, jackknife >50% below)