Rohwer.2000.Toward a phylogenetic classification of the Lauraceae-evidence from matk sequences

SAVE OFFLINE

Systematic Botany (2000), 25(1): pp. 60–71 q Copyright 2000 by the American Society of Plant Taxonomists

Toward a Phylogenetic Classification of the Lauraceae: Evidence from matK Sequences JENS G. ROHWER Institut fu¨r Spezielle Botanik, Johannes Gutenberg-Universita¨t, 55099 Mainz, Germany Communicating Editor: Alan Whittemore ABSTRACT. The larger part of the matK gene and the (392) adjacent spacer have been sequenced and used for phylogenetic analyses in 48 species of Lauraceae from all parts of their geographical range, and in three outgroup taxa. Except for the aberrant genera Hypodaphnis and Cassytha, the genetic divergence within the family is surprisingly low. In spite of this, several clades receive sufficient support to change our current concepts of relationships within the family. Particularly well supported is a Beilschmiedia–Cryptocarya clade that had been recognized by wood anatomy, but not in most of the recent morphological systems. The separation between taxa with involucrate and non-involucrate inflorescences, which had been one of the basic subdivisions in all systems of the family so far, is not supported by the present data. Instead, there seems to be an early division into a Gondwanan group and a Laurasian–South American group.

outgroups for the present study because they were the species most easily available of the two families. Taxon Sampling. The taxon sample within the Lauraceae was designed to include as much morphological variation as possible, and to allow at least a preliminary judgement about the major subdivisions of the systems mentioned above. Some additional taxa (e.g., Aspidostemon Rohwer & H.G.Richt., Eusideroxylon Teijsm. & Binn. and Neocinnamomum H.Liu) would have been desirable, but so far no fresh or silica-dried material was available of these taxa, and all attempts to extract amplifiable DNA from herbarium material failed in these cases. Some large, widespread and possibly not monophyletic genera are represented by several species from different morphological and/or geographical groups. The plant material came from various different sources (Table 1). Some was collected in botanic gardens, some in the field, and a number of samples were taken from herbarium specimens. In the first case voucher specimens were made (except when the entire plant was too small to remove a twig), and later compared with reliably identified material in HBG and K. Material collected by Jorge Gomez-Laurito in Costa Rica unfortunately came without voucher specimens. However, the identifications of this very experienced collector are usually correct, and a comparison of the leaf fragments in the herbarium provided no reason to suspect that they might be erroneous. In herbarium material from AAU and HBG, the correct generic identification was checked in each case. The specific identity often could not be ascertained beyond

The Lauraceae are among the floristically most important tropical woody families (Gentry 1988), but relationships within the family are still poorly understood. Essentially the same set of characters has been used in nearly all classification systems proposed so far (Nees 1836; Meissner 1864; Bentham 1880; Mez 1889; Pax 1891; Kostermans 1957; Hutchinson 1964; Rohwer 1993a), yet the results of different studies have been very different, depending on the authors’ opinion on the reliability of these characters. Van der Werff and Richter (1996) summarized the different systems, reaching the conclusion that the most characteristic feature common to all of them was that they are not widely accepted. The greatest single contribution towards a better understanding of the Lauraceae in recent decades came from the addition of an entirely new set of characters, i.e., from wood anatomy (Richter 1981). Therefore, it seemed most promising to further expand the range of characters by including molecular evidence. The chloroplast gene matK was chosen for this purpose because its potential for resolving intrafamilial relationships had been proven in earlier studies (see, e.g., Johnson and Soltis 1995; Hilu and Liang 1997). MATERIALS

AND

METHODS

Choice of Outgroup. Hernandiaceae and Monimiaceae s.str. are the closest relatives of the Lauraceae according to an analysis of six plastid genome regions by Renner (1999). Hernandia nymphaeaefolia (Hernandiaceae), Peumus boldus and Tambourissa religiosa (both Monimiaceae) were chosen as 60

2000]

61

ROHWER: LAURACEAE

TABLE 1. Material examined. Taxon

Actinodaphne borneensis Meisn. Aiouea dubia (Humb., Bonpl. & Kunth) Mez Alseodaphne perakensis (Gamble) Kosterm. Aniba affinis (Meisn.) Mez Apollonias barbujana (Cav.) Bornm. Beilschmiedia berteroana (Gay) Kosterm. Beilschmiedia mexicana (Mez) Kosterm. Beilschmiedia roxburghiana Nees Beilschmiedia tawa (A. Cunn.) Kirk Caryodaphnopsis tonkinensis (Lecomte) Airy Shaw Cassytha ciliolata Nees Chlorocardium rodiei (R.H. Schomb.) Rohwer, H.G. Richt. & van der Werff Cinnamomum camphora (L.) J. Presl Cinnamomum chavarrianum (Hammel) Kosterm. Cinnamomum cinnamomifolium (Humb., Bonpl. & Kunth) Kosterm. Cinnamomum verum J. Presl Cryptocarya alba (Molina) Looser Cryptocarya triplinervis R.Br. Dehaasia cuneata (Blume) Blume Dicypellium caryophyllaceum (Mart.) Nees Endiandra pubens Meisn. Endlicheria formosa A.C. Sm. Endlicheria verticillata Mez Hernandia nymphaeaefolia (C. Presl) Kubitzki Hypodaphnis zenkeri (Engl.) Stapf Laurus azorica (Seub.) Franco Licaria triandra (Sw.) Kosterm. Lindera benzoin (L.) Blume Litsea glutinosa (Lour.) C.B. Rob. Litsea japonica (Thunb.) Juss. Mezilaurus opaca Kubitzki & van der Werff Nectandra coriacea (Sw.) Griseb. Nectandra riparia Rohwer Nectandra salicifolia (Humb., Bonpl. & Kunth) Nees Neolitsea sericea (Blume) Koidz. Ocotea foetens (Ait.) Baill. Ocotea paulii C.K. Allen Persea americana Mill. Persea grijsii (Hance) Kosterm. Persea indica (L.) Spreng. Persea lingue (Ruiz. & Pav.) Nees Peumus boldus Molina Phoebe formosana (Hayata) Hayata Pleurothyrium costanense van der Werff Povedadaphne quadriporata W.C. Burger Rhodostemonodaphne grandis (Mez) Rohwer Sassafras tzumu (Hemsl.) Hemsl. Tambourissa religiosa (Tul.) A.DC. Umbellularia californica (Hook. & Arn.) Nutt. Urbanodendron verrucosum (Nees) Mez Williamodendron glaucophyllum (van der Werff) Kubitzki & H.G. Richt.

Voucher

Provenance

EMBL No.

Atkins 450 (AAU) Madsen 75433 (AAU) van Balgooy 2186 (HBG) Ziburski 547 (HBG) Rohwer 151 (MJG) Zo¨llner 21411 (HBG) Reyna 1404 (HBG) Gerlach s.n. (MJG) Chase 5519 (K) Frodin 2073 (AAU)

Brunei Ecuador Malaysia Bot. Gard. Bonn Bot Gard. Bonn 323-90 Chile El Salvador Bot. Gard. Munich Bot. Gard. Kew Borneo

AJ247142 AJ247143 AJ247144 AJ247145 AJ247146 AJ247147 AJ247148 AJ247149 AJ247150 AJ247151

Rohwer 175 (MJG) Polak 281 (HBG)

Bot. Gard. Mainz Guyana

AJ247152 AJ247153

Rohwer 164 (MJG) Gomez-Laurito (–)

Bot. Gard. Mainz Costa Rica

AJ247154 AJ247155

Thomsen 8942 (AAU)

Ecuador

AJ247156

Rohwer 165 (MJG) Chase 5521 (K) Chase 5522 (K) de Wilde 12742 (HBG) Pires 16756 (HBG) Constable 4876 (HBG) Balslev 97178 (AAU) Va´squez 9123 (AAU) Rohwer 166 (MJG) Leeuwenberg 5557 (HBG) Rohwer 168 (MJG) Zona s.n. (FTG) Rohwer 170 (MJG) Larsen 33442 (AAU) Rohwer 156 (MJG) Revilla 1226 (HBG) Zona s.n. (FTG) Kvist 1533 (AAU) Gomez-Laurito (–)

Bot. Gard. Mainz Bot. Gard. Kew Bot. Gard. Kew Sumatra Brazil Australia Ecuador Peru Bot. Gard. Mainz Cameroun Bot. Gard. Mainz Miami, FTG # X-1-172 Bot. Gard. Mainz Thailand Bot. Gard. Bonn 573-90 Peru Miami, FTG # X-4-50 Peru Costa Rica

AJ247157 AJ247158 AJ247159 AJ247160 AJ247161 AJ247162 AJ247163 AJ247164 AJ247165 AJ247166 AJ247167 AJ247168 AJ247169 AJ247170 AJ247171 AJ247172 AJ247173 AJ247174 AJ247175

Rohwer 171 (MJG) Rohwer 157 (MJG) Gomez-Laurito (–) Rohwer 181 (MJG) Rohwer 163 (MJG) Rohwer 172 (MJG) Hecker (–) Rohwer 182 (MJG) Rohwer 158 (MJG) Bosque CB-J-03 (HBG) Gomez-Laurito (–) Zak 4671 (HBG) Rohwer 160 (MJG) Kramer s.n. (MJG) Rohwer 173 (MJG) Heringer 18577 (HBG) Burger 12348 (HBG)

Bot. Gard. Mainz Bot. Gard. Bonn 575-90 Costa Rica Bot. Gard. Mainz Bot. Gard. Hamburg Bot. Gard. Mainz Bot. Gard. Marburg Bot. Gard. Hamburg Bot. Gard. Bonn 577-90 Venezuela Costa Rica Ecuador Bot. Gard. Bonn 3991 Bot. Gard. Heidelberg Bot. Gard. Mainz Brazil Costa Rica

AJ247176 AJ247177 AJ247178 AJ247179 AJ247180 AJ247181 AJ247182 AJ247183 AJ247184 AJ247185 AJ247186 AJ247187 AJ247188 AJ247189 AJ247190 AJ247191 AJ247192

62

TABLE 2. Primers used for amplification. Positions within matK are relative to tobacco. The last primer is almost identical with trnKII (Steele and Vilgalys 1994), shifted to 59 by one base. Its position is given as it appears in the alignment used for the analysis. Position

448 805 941 1084 1166 1318 1422 1847

[Volume 25

SYSTEMATIC BOTANY

Type

Sequence

forward forward reverse reverse forward reverse forward reverse reverse

59-GTGTCAGATATACTAATACC-39 59-ACCCTATGGTTGTTCAAAGAC-39 59-GTCTTTGAACAACCATAGGGT-39 59-CCGGTTGAGACCACAAGT-39 59-CTATTAAGAAATTCGAGACC-39 59-ACGGCTTACTAATGGGATGCC-39 59-TGTGCTAGAACTTTGTCTCG-39 59-TTGGGAAGATCAAAGAAAGA-39 59-ACTAGTCGGATGGAGTAGA-39

doubt, due to lack of reliable treatments for most genera and most parts of the world. DNA Extraction. Total genomic DNA from the material, whether fresh, silica-gel dried or from the herbarium, was extracted using the DNeasyTM (QIAGEN) or NucleoSpin-PlantTM (Macherey-Nagel) extraction kits. Amplification. About 1100 base pairs (bp) of the 39-end of the matK gene (starting from a position equivalent to nucleotide 448 of the tobacco matK gene) plus about 400 bp of the adjacent spacer were amplified by standard polymerase chain reaction (PCR) using a Grant Autogene II thermocycler. Usually only short fragments , 500 bp could be amplified from herbarium material, and in longer amplified fragments the maximum readable length starting from any primer usually varied between about 400 and 800 bp. Therefore, it was necessary to design a number of internal primers (Table 2). The PCR products were purified by agarose gel electrophoresis followed by gel extraction using the QIAquickTM (QIAGEN) or NucleoSpin-ExtractTM (Macherey-Nagel) kits. Sequencing. Cycle sequencing reactions in both directions were performed using aliquots of the purified PCR products (0.5 to 8 ml, depending on the DNA concentration), the Thermosequenase KitTM (Amersham) and one of the primers (Table 2). Samples were analyzed by GENterprise, Mainz, with automated sequencing systems (P-E Applied Biosystems ABI 373, 377). Sequence Alignment. The different sequence fragments of any one taxon were aligned automatically using SequencherTM 3.0 (GeneCodes), and then checked manually. The fully edited sequences,

generally confirmed by forward and reverse sequencing reactions, have been deposited in EMBL under the accession numbers indicated in Table 1. The sequences of different taxa were aligned manually. The alignment is available from EMBL (ds39429) and from the author. Alignment in the coding region was unproblematical, as only two insertion events had to be postulated, both in Peumus boldus, involving one and two AAA codons, respectively. Alignment in the spacer, although involving several indels, was generally straightforward, too. Bases 1656–1661 and 1793–1799, however, were excluded from the phylogenetic analysis because of uncertain alignment. Bases 1656–1661 represent a duplication present in Dehaasia cuneata and Pleurothyrium costanense but differing in length in the two species. Including it would create either one or four (probably spurious) synapomorphies between the two species, and it greatly increases the calculating time without changing the result. Bases 1793–1799 are in the middle of a sequence fragment of three to seven consecutive thymines and six to eight guanines. Indels of one to five bases in many taxa need to be postulated here, and their optimum position in the alignment seemed to be different in different groups of taxa. Otherwise, gaps were treated as missing data. In the final matrix, they added up to 2115 of the 64907 matrix cells, or 3.26%. All indels, however, were reintroduced into the matrix as presence-absence characters. When two or more taxa turned out to be identical in the sequence fragment analysed, only one of them was included in the phylogenetic analysis. Phylogenetic Analysis. Parsimony analyses were conducted as heuristic searches using PAUP (version 3.1.1., Swofford 1993, and test versions of 4.0b), with 100 random stepwise additions and the options COLLAPSE, MULPARS and TBR on. As all replicates yielded identical results, among themselves and with addition sequence CLOSEST, this latter option was chosen for the bootstrap analysis. Although most of the 100 bootstrap replicates were relatively fast, one of them (sometimes as late as replicate #78) invariably ran into overflow, with more than 24,000 equally parsimonious trees. To avoid this, MAXTREES was set to 1000 in the final attempt. This limit was hit only twice during the search. RESULTS Sequence Variability. The sequence divergence among the 51 taxa examined was surprisingly low.

2000]

63

ROHWER: LAURACEAE

In three pairs of taxa, two of them from different genera, it was even zero, i.e., the sequences turned out to be identical in all 1360 nucleotide positions and 21 indels included in the analysis. This was the case in Aniba affinis and Urbanodendron verrucosum, in Persea grijsii and Phoebe formosana, and in the two neotropical Cinnamomum species examined, C. chavarrianum and C. cinnamomifolium, which are morphologically quite distinct. Dicypellium caryophyllaceum differed from Aniba affinis and Urbanodendron verrucosum only in two of the positions that had been excluded because of uncertain alignment. Therefore, only one taxon from each of these groups was included in the final data matrix, which thus consisted of 47 taxa and 1381 characters. 923 of these characters were constant, 289 uninformative (275 bp plus 14 indels), and 169 informative (12.2%, or 162 bp 5 11.9%). Within the Lauraceae, 1031 characters were constant (1025 bp plus six indels), 216 uninformative (208 bp plus eight indels), and 134 were informative (9.7%, or 127 bp 5 9.3%). The maximum divergence found within the Lauraceae was 151 bp (11.1%, or 159 characters including indels 5 11.5%), between Cassytha ciliolata and Hypodaphnis zenkeri, and the mean divergence was 39.9 bp (2.9%, or 42.7 characters including indels 5 3.1%). The precise number of consecutive adenines in Cassytha ciliolata from position 1599 onwards could not be ascertained beyond doubt, because the polymerase would not amplify correctly across such a long poly-A stretch. Accepted for the analysis were 22, which appeared most likely from five sequencing attempts and fits into the alignment without having to postulate further indels. Parsimony Analysis. The parsimony analysis yielded 100 equally parsimonious trees (Fig. 1; Length 5 679, consistency index CI 5 0.798, retention index RI 5 0.845). Excluding the characters coding for indels slightly changed the tree scores but not the trees themselves. It is obvious that most of the topological instability among the 100 trees results from the low sequence divergence within a very large terminal group, which is unresolved in the strict consensus tree, whereas in the 50% majority consensus tree (Fig. 2) it is subdivided into a large (sub)terminal polytomy and a smaller Persea group. Sister group to this weakly resolved majority of the family is a clade including Chlorocardium, Mezilaurus and Williamodendron, followed by Caryodaphnopsis and a clade consisting of Beilschmiedia, Endiandra and Cryptocarya. Cassytha and Hypodaphnis are found at the base of the family.

DISCUSSION Phylogenetic Implications. Despite some gaps in the taxon sample and poor resolution among many taxa, the present analysis does yield significant information about the phylogenetic position of several clades or individual taxa within the Lauraceae. These will be discussed in the following. HYPODAPHNIS. This is the only lauraceous genus with a truly inferior ovary, and the only one endemic in tropical Africa. In most of the recent systems (Kostermans 1957; Richter 1981; Rohwer 1993a) it had been placed somewhere near, though not immediately close to Cryptocarya, in which the ovary is completely enclosed in the floral tube but not fused with it. Embryological evidence (Heo et al. 1998), however, strongly pleads against a close relationship with Cryptocarya. This analysis shows that Hypodaphnis is even more aberrant than expected. It has so many characters in common with the outgroup taxa that it seems to be sister to the rest of the Lauraceae. This position, however, is not well supported yet, especially with respect to Cassytha. The branch separating the two failed to reach 50% in the bootstrap analysis, whereas the branch separating Cassytha and Hypodaphnis from the rest of the Lauraceae received somewhat better, though still not very strong support. In the distance matrix, Hypodaphnis is closest to Beilschmiedia berteroana (84 differences), followed by B. mexicana, B. tawa and Endiandra pubens (all 86 differences). These are still very large distances compared to those among the rest of the family, and Hypodaphnis is certainly not the closest relative of Beilschmiedia, but it still may turn out to belong to the base of the Beilschmiedia– Cryptocarya clade. CASSYTHA. This hemiparasitic twiner is by far the most aberrant genus of the Lauraceae, irrespective of whether morphological, embryological (Heo et al. 1998) or the present molecular data are considered. It was placed in a subfamily of its own by Kostermans (1957), whereas Rohwer (1993a) placed it in his Cryptocarya group. The former placement is supported by embryological data (Heo et al. 1998), although there is only one probable synapomorphy for the Lauraceae excluding Cassytha, viz. the ab initio nuclear type of endosperm formation. All other embryological characters, in which Cassytha differs from the rest of the family, are autapomorphies of the parasite. The reason for placing Cassytha in the Cryptocarya group was mainly the fruit construction shared with the genera of that group, viz. with the receptacular tube

64

SYSTEMATIC BOTANY

[Volume 25

FIG. 1. One of 100 equally parsimonious trees (L 5 679, CI 5 0.798, RI 5 0.845). Numbers are branch lengths; length 1 omitted for the sake of legibility.

2000]

ROHWER: LAURACEAE

65

FIG. 2. 50% majority consensus of the 100 equally parsimonious matK trees. Majority percentages above the lines, bootstrap values below the lines. Columns on the right show (1.) groups discussed in the text, (2.) inflorescence structures (bot. 5 botryoid, from winter buds; inv. 5 involucrate), and (3.) distribution of the genus and the species (boldface). AF 5 Africa, AS 5 Asia, AU 5 Australia, CI 5 Canary Islands, MA 5 Madagascar, ME 5 Mediterranean, NA 5 North America, NT 5 neotropical plus South American temperate, NZ 5 New Zealand, PT 5 pantropical. The bars on the far right show the inferred distribution of the ancestors of the major clades.

66

SYSTEMATIC BOTANY

completely enclosing the ovary (at least) at fruit maturity. This, however, could be a parallelism. The present analysis does not clarify this question. Neither of the two alternative positions is well supported in this analysis, but both remain possible. Like many other parasitic plants (Nickrent et al. 1998), Cassytha shows an increased rate of macromolecular evolution. Its average distance from the remainder of the family is 147.8 characters, whereas the average distance among the Lauraceae excluding Cassytha is only 36.9 characters. Its apparent position between Hypodaphnis and the Beilschmiedia–Cryptocarya clade thus may be influenced by long-branch attraction. Within the Lauraceae, Cassytha shows the greatest distance from Hypodaphnis (159 differences), next to which it appears in the cladogram, but like Hypodaphnis it is closest to Beilschmiedia berteroana (134 differences). This further underscores the ambiguity at the base of the present matK tree. The basal clades hopefully will become more reliably supported once sequences of the two genera with semi-inferior ovaries, Eusideroxylon Teijsm. & Binnend. and Potoxylon Kosterm., become available. BEILSCHMIEDIA, CRYPTOCARYA, AND ENDIANDRA. The clade uniting these three genera is 100% supported in the bootstrap analysis, and a close relationship among the three genera is also evident in wood and bark anatomy (Richter 1981), inflorescence structure (van der Werff and Richter 1996) and several embryological characters (Heo et al. 1998). Therefore, van der Werff and Richter (1996) placed the three genera in the Cryptocaryeae, together with Potameia and Triadodaphne, which have not been examined here. In other morphologically based systems, however, Beilschmiedia and Endiandra on the one hand, and Cryptocarya on the other, had usually been separated to a greater (Kostermans 1957) or lesser degree (Rohwer 1993a), because they represent opposite extremes in the development of the floral tube. This difference, however, can be only of minor importance, judged from the overwhelming support for a close relationship from several unrelated character sets. Beilschmiedia is one of the genera that were suspected to be perhaps not monophyletic, because of considerable morphological variation and an enormous range of distribution, from Mexico and Japan in the north to Chile and New Zealand in the south. The four species examined here (from Central and South America, tropical Asia, and New Zealand) clearly do belong to the same clade, but it should be noted that so far an African representative is lacking. The po-

[Volume 25

sition of Endiandra within Beilschmiedia in the consensus trees is not very strongly supported in the bootstrap analysis, but it is conceivable that Endiandra may have arisen from Beilschmiedia through reduction of the six outer stamens. The sister group to the Beilschmiedia–Cryptocarya clade, including all remaining genera, is well supported by the molecular data, but is too diverse to be defined in morphological or anatomical terms. CARYODAPHNOPSIS. Kostermans (1957) placed this genus in the synonymy of Persea, but reinstated it later (Kostermans 1974). The present author (Rohwer 1993a) likewise chose a position close to Persea, although with some hesitation. Richter (1981), on the other hand, placed Caryodaphnopsis distally in the Beilschmiedia–Cryptocarya branch of his phylogenetic diagram, a position supported by the embryological data of Heo et al. (1998). The molecular data now suggest a position between the two groups, but clearly remote from both. Although well supported by the bootstrap values, this may not be the final arrangement. DNA amplification was possible from only a single specimen, and the signal was not very strong. Therefore, there is still a slight possibility of errors in the sequence, although (with a total of 15 sequencing reactions) all attempts have been made to ensure their correctness. Caryodaphnopsis shows a remarkable disjunction between Southeast Asia and South America (van der Werff and Richter 1985). Although both morphology and wood anatomy leave no doubt about the circumscription of the genus, it would be interesting to get sequences of South American species as well. From a morphological point of view, the Asian Neocinnamomum (not examined here) appears to be close to Caryodaphnopsis, whereas anatomically it seems closer to Mezilaurus (see below). The sister group to Caryodaphnopsis is very strongly supported by the present macromolecular data, but as it still includes the majority of the genera, it also encompasses such a wide range of variation that it cannot be characterized morphologically with any degree of confidence. According to the results of Heo et al. (1998), it seems to be characterized by an amoeboid tapetum in the anthers, as opposed to a glandular tapetum in the other genera, but the taxon stample studied so far is still too small to accept this as a general characteristic. CHLOROCARDIUM, MEZILAURUS, AND WILLIAMODENDRON. Although the inclusion of Chlorocardium in this clade is reasonably supported by the present data, it cannot yet be accepted as proven.

2000]

ROHWER: LAURACEAE

The three genera share hard and heavy wood, but the wood structure of Chlorocardium is clearly different from that of the other two genera (Richter 1985; Kubitzki and Richter 1987; Rohwer et al. 1991). In particular, it should be noted that Mezilaurus and Williamodendron have distinctly bordered pits in their wood fibers, like the genera discussed so far (except the herbaceous Cassytha). So that appears to be a plesiomorphic character for the Lauraceae. Chlorocardium, on the other hand, has fibers with simple pits only, like the remaining genera. This could be a synapomorphy if Chlorocardium was wrongly placed here. Morphologically, Chlorocardium has so little in common with Mezilaurus and Williamodendron that a close relationship would be rather surprising. Chlorocardium has tetramerous to irregular flowers with numerous (12–20) stamens, whereas the two other genera have trimerous flowers with only three stamens. The close relationship between Mezilaurus and Williamodendron, on the other hand, is absolutely certain. They had even been placed in the same genus by van der Werff (1987). Both morphological and anatomical data suggest that Sextonia van der Werff (not examined here) may belong to the Mezilaurus clade as well, whereas embryological characters suggest that it belongs to the Beilschmiedia–Cryptocarya group (Heo et al. 1998, as Ocotea rubra Mez). The clade including all remaining genera is 100% supported in the bootstrap analysis and relatively poorly resolved. It includes the Laureae and most of the Ocotea group in the sense of Rohwer (1993a), excluding Mezilaurus and Williamodendron. Van der Werff and Richter (1996) did not suggest any definite placement for these two genera (and several others), but otherwise the group discussed here coincides with their Perseeae plus Laureae. It includes both the genera with umbellate, involucrate (partial) inflorescences and those with basically thyrsoid inflorescences showing regular lateral and terminal dichasia. There have been indications that these genera might be related, but still it is surprising how close they are according to the matK data. Richter (1981) hinted at a relationship by drawing a short common ‘‘stem’’ uniting these two of the three major branches in his phylogenetic diagram, but it did not seem as if this relationship was particularly close. The genera with umbellate and those with thyrsoid inflorescences plus Mezilaurus and Hypodaphnis also had been put together in one group by Heo et al. (1998), but since that group comprised all genera except Cassytha and those of the much smaller and more distinct Beilschmiedia

67

group, this did not necessarily mean a close relationship either. From a morphological point of view, it appears most plausible to assume that inflorescences with a regular thyrsoid structure evolved from the less regularly paniculate inflorescences that characterize all genera below this node. The taxa with involucrate inflorescences would then have evolved later within this group. There is indeed some morphological evidence suggesting that supposedly umbellate inflorescences were derived from a thyrsoid pattern (V. Ruge, unpubl. data). The molecular data do not allow any definite conclusion yet on whether this interpretation is correct or not. THE PERSEA GROUP. This clade seems to be a real phylogenetic alliance, even though it is neither present in the strict consensus nor supported by the bootstrap analysis. However, its instability is due to just three bases common to Apollonias barbujana (Persea group) and Sassafras tzumu (terminal group) and different from (most of) the other Lauraceae. Deleting either one of the two species or excluding a single particularly variable base in the spacer region (position 1623) greatly reduces the number of equally parsimonious trees (to 18 or 25), makes the Persea clade appear in the strict consensus, and increases the bootstrap support for this clade to 51– 63%. When one of the two species is deleted, however, the Persea clade appears as a member of the subterminal polytomy rather than as its sister group. Morphologically, the Persea clade is characterized by the presence of regularly thyrsoid inflorescences and the absence of a cupule. The genera seem to be so closely related that their delimitation against each other had been controversial. Alseodaphne and Phoebe had been included in Persea by Bentham (1880), who also recognized that Apollonias differed from (section) Phoebe in nothing but the number of pollen sacs. Kostermans remarked the same about Alseodaphne (1973a) and Dehaasia (1973b), which indeed come out as sister groups here. Earlier, Kostermans (1957) had followed Bentham in treating Alseodaphne as a synonym of Persea. It should also be noted that the Persea clade does not include the neotropical species of Cinnamomum, which had been included in Phoebe until Kostermans (1961) recognized their true affinities. THE TERMINAL GROUP. The most surprising result of this study is certainly the low degree of sequence differentiation within a group comprising both the genera with umbellate, involucrate inflorescences (Actinodaphne, Laurus, Lindera, Litsea, Neolitsea, Umbellularia) and the genera with basically thyrsoid inflorescences, except those of the Persea

68

SYSTEMATIC BOTANY

group discussed above. The maximum sequence divergence within this group is 13 bases (0.9%), and the average divergence is just 7.3 bases (0.5%). Morphologically, the two types of inflorescence represent the most clearcut distinction among the woody Lauraceae. Therefore, the involucrate group has been recognized in all systems so far on a relatively high hierarchic level, mostly as tribe Laureae. Such a high rank becomes questionable, even though a sequence analysis of a more rapidly evolving marker may still show that the presently unresolved ‘‘Laureae’’ are a monophyletic group. They might even be the sister group to all Lauraceae with thyrsoid inflorescences, i.e., the Persea group plus the remainder of the terminal group. In morphological terms, it is difficult to find any synapomorphy that unites the entire terminal group. Perhaps the development of a true cupule, a bowl- or cup-shaped structure subtending the fruit, may be regarded as such. It is present in most genera within the group, but unfortunately not in all species. Especially several taxa with involucrate inflorescences do not have a cupule, and in some Ocotea species it is greatly reduced. Outside the terminal group, there are species with scarcely changed or swollen fruiting pedicels, with small, discoid structures subtending the fruit or with floral tubes enclosing the entire ovary even in fruit, but the typical cupshaped structures do not occur. The only exception to this rule is the genus Chlorocardium, in which one species has a larger discoid cupule, whereas the second species shows an almost closed subglobular structure. So it remains questionable whether the cupule can be be regarded as characteristic of the terminal group. Within the terminal group, only one clade deserves to be discussed at this stage, namely the clade uniting Pleurothyrium costanense and Nectandra riparia, a representative of the core Nectandras in the sense of Rohwer (1993b). Morphologically, the two groups share papillose tepals, more or less fleshy stamens and a rather deep receptacle, but they differ in the arrangement of their pollen sacs and the size of their staminal glands. The species of the Nectandra coriacea group are not part of this clade, which indicates that Nectandra, as presently circumscribed, may be paraphyletic or even diphyletic. Biogeographic Considerations. All families of the order Laurales as presently circumscribed (Renner 1999), except the monotypic Gomortegaceae, are present in both the Old and the New World. Most of them are either clearly centered in the southern hemisphere (Atherospermataceae, Gomor-

[Volume 25

tegaceae, Monimiaceae s.str.), or they are predominantly tropical in both hemispheres (Hernandiaceae, Lauraceae, Siparunaceae), with the majority of their species and/or their presumably primitive characters south of the equator. Only the Calycanthaceae, which occupy a marginal position within the order, are centered in the northern hemisphere. Therefore, the Laurales are often interpreted as a Gondwana group. The basal clades within the Lauraceae are compatible with such an interpretation. Cassytha, Cryptocarya, and Beilschmiedia are the only three fully pantropical genera in the Lauraceae, and all three are well represented in the southern hemisphere, in South America, Africa, and Australia. The tropical West African Hypodaphnis may be seen as a Gondwanan relict, like the similarly distributed Glossocalyx Benth. in the Siparunaceae (see Renner 1998), whereas the Australasian Endiandra can be interpreted as a derivative from Beilschmiedia-like ancestors. The situation changes in the distal part of the cladogram. Caryodaphnopsis (see above) is known from South America and Southeast Asia, but neither from Africa nor from Australia. This distribution is somewhat similar to that of the Macleaya/ Bocconia clade in the Papaveraceae described by Blattner and Kadereit (1995), which they interpreted as derived from the more common East Asian– eastern North American disjunction. Based on geological data, they discussed the possible age of the disjunction between Eocene (maximum) and Pliocene (minimum), without reaching a definite conclusion. For Caryodaphnopsis, no age estimate can be given so far, as only one (Asian) species has been examined here. The clade including Chlorocardium, Mezilaurus, and Williamodendron is entirely neotropical. Its position in the cladogram would still allow an origin from the Gondwanan stock of the family, in contrast to its sister group (see below). The present restriction of this clade to South America may be regarded as evidence that the common ancestor with its sister group had been South American as well, and this in turn would suggest that Laurasia was colonized via South America. Although palaeogeographic reconstructions (e.g., Krutzsch 1989; Briggs 1996) differ in details, this appears to be one of the two most likely routes (the other is via Africa) for any angiosperm group that migrated between Gondwana and Laurasia before the Mid-Cretaceous (see below). It would, however, push back its origin at least into the Jurassic. On the other hand, Chlorocardium, Mezilaurus and Williamodendron are both

2000]

ROHWER: LAURACEAE

morphologically and anatomically so different from the taxa in their sister group that it is tempting to regard them as representatives of an otherwise extinct lineage, which is only distantly related to the remaining taxa and therefore of little relevance for their origin. The taxa in the sister group to the previous clade can be sorted into several geographical groups. The largest group in terms of species numbers is formed by Asian–American disjunct genera (Cinnamomum, Lindera, Litsea, Persea, Sassafras). Morphologically it is a relatively diverse assemblage, including taxa with and without involucra, and with and without cupules. This suggests that migration between the Old World and the New World occurred more than once within this group. The idea that the disjunctions might have been caused by the breakup of Gondwana can be ruled out based on the weak differentiation in the chloroplast genome. These five American–Asian disjunct genera thus clearly represent a Laurasian element within the family. Lauraceae with a modern-looking flower structure (but inflorescences unmatched in recent material) have been present for at least about 100 million years in an area probably corresponding to the Normapolles provice of Laurasia (Drinnan et al. 1990; Crane et al. 1994; Herendeen et al. 1994; Eklund and Kvacˇek 1998; Eklund, in press, and references therein), whereas comparably well preserved Cretaceous fossils from the Southern Hemisphere are not known so far. The five Asian–American disjunct genera have overlapping but different distributions. Sassafras is entirely temperate, showing the common E Asian– eastern N American disjunction reviewed by Boufford and Spongberg (1983). Lindera has a similar distribution in the New World, but in the Old World it has a wider range, reaching Sri Lanka and N Australia. Persea ranges across a slightly smaller latitudinal range than Lindera in the Old World, whereas in the New World it reaches less to the north but as far as Chile to the south. In addition, it has a species in the Canary Islands. Cinnamomum has a slightly smaller range than Persea in the New World, from Mexico and Cuba to Uruguay, but a much wider range in the Old World, from Pakistan to Fiji, and from Japan to SE Australia. Litsea is the largest and (within this group) most widely distributed genus in the Old World, reaching even New Zealand. In the New World, in contrast, it is represented by only a few species in the southeastern USA, Mexico, and Central America. At least in the Americas all these genera prefer temperate to

69

subtropical or tropical-montane conditions; they are weakly if at all represented in the tropical lowlands. This preference for cooler climates certainly facilitated their migration along northern routes in the Tertiary. The second largest (or largest in number of genera) geographical group within the weakly differentiated terminal clade of the Lauraceae is formed by taxa restricted to tropical America (Aiouea, Aniba, Endlicheria, Licaria, Nectandra, Pleurothyrium, Povedadaphne, Rhodostemonodaphne, Urbanodendron). Morphologically, this group is relatively homogeneous, with thyrsoid inflorescences, invariably regular trimerous flowers, and distinct cupules. The largest genus in the Neotropics, Ocotea, shares the same suite of characters, but besides about 300 neotropical species it also has one species in the Canary Islands, seven in Africa and about 50 in Madagascar. The present analysis does not allow any judgement yet whether Ocotea is monophyletic or not. If it is, then it should be noted that the presumably most advanced species, especially the obviously dioecious ones (with the other sex rudimentary or absent), are found almost exclusively in South America, whereas the Old World species are at least apparently hermaphrodite. This observation, together with the relatively weak separation of Ocotea from its closest relatives in Central America as opposed to a much sharper discontinuity in South America, led me earlier to speculate about a geologically relatively recent immigration of Ocotea to South America (Rohwer 1986). This view would be compatible with the interpretation given for the American–Asian disjunct genera above, but then it remains an open question whether Ocotea migrated via an Atlantic rather than Pacific route. A third geographical group is formed by the genera restricted to (Austral–)Asia. They can easily be sorted into derivatives of either the Litsea alliance (Actinodaphne, Neolitsea) or the Persea alliance (Alseodaphne, Dehaasia, Phoebe). Thus they are undoubtedly derived from the Laurasian stock. The two Canary Island relicts not discussed so far show the same pattern: Laurus, of which a second species occurs in the Mediterranean region, differs from Lindera (i.e., a member of the Litsea group) mainly in its flower symmetry. Apollonias, on the other hand, differs from Phoebe (Persea group) only in the number of pollen sacs per anther. In fact, the second species of Apollonias (A. arnottii Nees), which occurs in India, is often regarded as a species of Phoebe that has lost the upper pair of pollen sacs. Such a reduction is known from several genera, and

70

[Volume 25

SYSTEMATIC BOTANY

is relatively common in South American Persea (Rohwer 1994). The Californian Umbellularia is certainly a relict as well, and the presence of a cupule excludes it from the Persea group. Still its position is somewhat ambiguous: Like the genera of the Litsea group it has involucrate inflorescences, but its flowers and fruits are more similar to Cinnamomum or Ocotea. Outlook. Preliminary as it may be, this study helps to define tasks for further research. To improve the resolution in the basal part of the present cladogram, it will be necessary to get material of as many as possible of the other genera that Richter (1981) placed in the Beilschmiedia–Cryptocarya branch of his phylogenetic diagram and of their presumed relatives (Aspidostemon Rohwer & H.G.Richt., Brassiodendron C.K.Allen, Dahlgrenodendron J.J.M.van der Merwe & A.E.van Wyk, Eusideroxylon Teijsm. & Binn., Hexapora Hook.f., Potameia Thouars, Potoxylon Kosterm., Syndiclis Hook.f., Triadodaphne Kosterm.). Unfortunately, many of these are mono- or oligotypic and quite rare. Hexapora still seems to be known from the type only. The addition of representatives of small aberrant genera that do not show distinct similarities to the Beilschmiedia–Cryptocarya group (Anaueria Kosterm., Cinnadenia Kosterm., Neocinnamomum H.Liu and Sextonia van der Werff) will most likely change the topology in the middle part of the cladogram. The greatest challenge at present seems to be the resolution of the distal polytomy, and research is under way to find an appropriate marker for this purpose. Resolving the phylogenetic pathways within that group will allow much more reliable interpretations of the character evolution and the biogeographic history of the Lauraceae. ACKNOWLEDGEMENTS. I thank the director of the Institut fu¨r Spezielle Botanik in Mainz, Prof. J. W. Kadereit, for giving me the opportunity to learn the molecular techniques and to work in his lab. Thanks are also due to the entire staff for their help with any kind of problem that arose during my work. I am particularly grateful to Mike Thiv, who guided my initial steps into the field of DNA sequence analysis. The directors of the herbaria AAU and HBG, Henrik Balslev and Klaus Kubitzki, generously allowed me to remove numerous fragments from herbarium specimens. Jorge Gomez Laurito is gratefully acknowledged for collecting material for my studies in Costa Rica. Mark Chase kindly sent DNA extracts of several species from Kew Gardens. Gu¨nther Gerlach (Munich), Peter Leins and Klaus Kramer (Heidelberg), Wolfram Lobin (Bonn), Carsten Schirarend (Hamburg), and Scott Zona (Miami) kindly sent silica-dried material from their respective Botanical Gardens and/or allowed me to collect material

there, and Ulrich Hecker brought back a specimen from Marburg. Joachim Kadereit, Mike Thiv and the two reviewers, Susanne Renner and Henk van der Werff, are gratefully acknowledged for their helpful comments on an earlier draft of this paper. This study was supported by grant RO 700/3–1 from the Deutsche Forschungsgemeinschaft.

LITERATURE CITED BENTHAM, G. 1880. Laurineae. Pp. 146–168 in Genera plantarum vol. 3, 1, eds. G. Bentham and J. D. Hooker. London: Lovell Reeve & Co, Williams & Norgate. BLATTNER, F. R. and J. W. KADEREIT. 1995. Three intercontinental disjunctions in Papaveraceae subfamily Chelidonioideae: evidence from chloroplast DNA. Plant Systematics and Evolution, Supplement 9: 147–157. BOUFFORD, D. E. and S. A. SPONGBERG. 1983. Eastern Asian–eastern North American phytogeographical relationships—a history from the time of Linnaeus to the twentieth century. Annals of the Missouri Botanical Garden 70: 423–439. BRIGGS, J. C. 1996. Global biogeography. Amsterdam: Elsevier. CRANE, P. R., E. M. FRIIS and K. R. PEDERSEN. 1994. Paleobotanical evidence on the early radiation of magnoliid angiosperms. Plant Systematics and Evolution, Supplement 8: 51–72. DRINNAN, A. N., P. R. CRANE, E. M. FRIIS and K. R. PEDERSEN. 1990. Lauraceous flowers from the Potomacgroup (Mid-Cretaceous) of eastern North America. Botanical Gazette 151: 370–384. EKLUND, H. (in press). Lauraceous flowers from the Late Cretaceous of North Carolina, U.S.A. Botanical Journal of the Linnean Society. and J. KVACˇEK 1998. Lauraceous inflorescences and flowers from the Cenomanian of Bohemia (Czech Republic, Central Europe). International Journal of Plant Sciences 159: 668–686. GENTRY, A. H. 1988. Changes in plant community diversity and floristic composition on environmental and geographical gradients. Annals of the Missouri Botanical Garden 75: 1–34. HEO, K., H. VAN DER WERFF and H. TOBE. 1998. Embryology and relationships of Lauraceae (Laurales). Botanical Journal of the Linnean Society 126: 295–322. HERENDEEN, P. S., W. L. CREPET, and K. C. NIXON. 1994. Fossil flowers and pollen of Lauraceae from the Upper Cretaceous of New Jersey. Plant Systematics and Evolution 189: 29–40. HILU, K. W. and H. LIANG. 1997. The matK gene: sequence variation and application in plant systematics. American Journal of Botany 84: 830–839. HUTCHINSON, J. 1964. The genera of flowering plants (Dicotyledonae), vol. 1. Oxford: Clarendon Press. JOHNSON, L. A. and D. E. SOLTIS. 1995. Phylogenetic inference in Saxifragaceae sensu stricto and Gilia (Pole-

2000]

ROHWER: LAURACEAE

moniaceae) using matK sequences. Annals of the Missouri Botanical Garden 82: 149–175. KOSTERMANS, A. J. G. H. 1957. Lauraceae. Pengumuman Balai Besar Penjelidikan Kehutanan Indonesia 57: 1– 64. 1961. The new world species of Cinnamomum Trew. (Lauraceae). Reinwardtia 6: 17–24. 1973a. A synopsis of Alseodaphne Nees (Lauraceae). Candollea 28: 93–136. 1973b. A synopsis of the genus Dehaasia Bl. (Lauraceae). Botanische Jahrbu¨cher fu¨r Systematik, Pflanzengeschichte und Pflanzengeographie 93: 424–480. 1974. A monograph of Caryodaphnopsis A. Shaw. Reinwardtia 9: 123–137. KRUTZSCH, W. 1989. Paleogeography and historical phytogeography (paleochorology) in the Neophyticum. Plant Systematics and Evolution 162: 5–61. KUBITZKI, K. and H. G. RICHTER. 1987. Williamodendron Kubitzki & Richter, a new genus of neotropical Lauraceae. Botanische Jahrbu¨cher fu¨r Systematik, Pflanzengeschichte und Pflanzengeographie 109: 49–58. MEISSNER, C. F. 1864. Lauraceae. Pp. 1–260 in Prodromus systematis naturalis regni vegetabilis vol. 15, 2, ed. A. de Candolle. Paris: Fortin, Masson et soc. MEZ, C. 1889. Lauraceae Americanae. Jahrbuch des Ko¨niglichen Botanischen Gartens und des Botanischen Museums zu Berlin 5: 1–556. NEES VON ESENBECK, C. G. D. 1836. Systema Laurinarum. Berlin: Sumtibus Veitii et sociorum. NICKRENT, D. L., R. J. DUFF, A. E. COLWELL, A. D. WOLFE, N. D. YOUNG, K. E. STEINER, and C. W. DE PAMPHILIS. 1998. Molecular phylogenetic and evolutionary studies of parasitic plants. Pp. 211–241 in Molecular systematics in plants II: DNA sequencing, eds. D. E. Soltis, P. S. Soltis and J. J. Doyle. Boston, Dordrecht, London: Kluwer. PAX, F. 1891. Lauraceae. Pp. 106–126 in Die natu¨rlichen Pflanzenfamilien vol. III, 2, eds. A. Engler and K. Prantl. Leipzig: Engelmann. RENNER, S. S. 1998. Phylogenetic affinities of Monimiaceae based on cpDNA and spacer sequences. Perspectives

71

in Plant Ecolology, Evolution and Systematics 1: 61– 77. 1999. Circumscription and phylogeny of Laurales: Evidence from molecular and morphological data. American Journal of Botany 86: 1301–1315. RICHTER, H. G. 1981. Anatomie des sekunda¨ren Xylems und der Rinde der Lauraceae. Sonderba¨nde des Naturwissenschaftlichen Vereins in Hamburg 5: 1–148. 1985. Wood and bark anatomy of Lauraceae. II. Licaria Aublet. IAWA Bulletin, N.S. 6: 187–199. ROHWER, J. G. 1986. Prodromus einer Monographie der Gattung Ocotea Aubl. (Lauraceae). Mitteilungen aus dem Institut fu¨r Allgemeine Botanik Hamburg 20: 3– 278. 1993a. Lauraceae. Pp. 366–391 in The families and genera of vascular plants, vol. 2, eds. K. Kubitzki, J. G. Rohwer and V. Bittrich. Berlin, Heidelberg, New York, London, Paris, Tokyo, Hong Kong, Barcelona, Budapest: Springer. 1993b. Lauraceae: Nectandra. Flora Neotropica Monograph No. 60. New York: The New York Botanical Garden. 1994. A note on the evolution of stamens in the Laurales, with emphasis on the Lauraceae. Botanica Acta 107: 103–110. , H. G. RICHTER, and H. VAN DER WERFF. 1991. Two new genera of neotropical Lauraceae and critical remarks on the generic delimitation. Annals of the Missouri Botanical Garden 78: 388–400. STEELE, K. P. and R. VILGALYS. 1994. Phylogenetic analysis of Polemoniaceae using nucleotide sequences of the plastid gene matK. Systematic Botany 19: 126–142. SWOFFORD, D. L. 1993. PAUP: phylogenetic analysis using parsimony, version 3.1.1. Champaign, IL: Illinois Natural History Survey. WERFF, H. VAN DER 1987. A revision of Mezilaurus (Lauraceae). Annals of the Missouri Botanical Garden 74: 153–182. and H. G. RICHTER. 1985. Caryodaphnopsis AiryShaw (Lauraceae), a genus new to the Neotropics. Systematic Botany 10: 166–173. and 1996: Toward an improved classification of Lauraceae. Annals of the Missouri Botanical Garden 83: 409–418.