HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via ISI Web of Science (4) Citing Articles via Google Scholar Google Scholar Articles by Harbaugh, D. T. Articles by Baldwin, B. G. Search for Related Content PubMed Articles by Harbaugh, D. T. Articles by Baldwin, B. G. Agricola Articles by Harbaugh, D. T. Articles by Baldwin, B. G.

Phylogeny and biogeography of the sandalwoods (Santalum, Santalaceae): repeated dispersals throughout the Pacific¹

Department of Integrative Biology, University and Jepson Herbaria, University of California, Berkeley, 1001 Valley Life Sciences Building, Berkeley, California 94720-2465 USA

Received for publication November 3, 2006. Accepted for publication May 8, 2007.

ABSTRACT

Results of the first genus-wide phylogenetic analysis for Santalum (Santalaceae), using a combination of 18S–26S nuclear ribosomal (ITS, ETS) and chloroplast (3' trnK intron) DNA sequences, provide new perspectives on relationships and biogeographic patterns among the widespread and economically important sandalwoods. Congruent trees based on maximum parsimony, maximum likelihood, and Bayesian methods support an origin of Santalum in Australia and at least five putatively bird-mediated, long-distance dispersal events out of Australia, with two colonizations of Melanesia, two of the Hawaiian Islands, and one of the Juan Fernandez Islands. The phylogenetic data also provide the best available evidence for plant dispersal out of the Hawaiian Islands to the Bonin Islands and eastern Polynesia. Inability to reject rate constancy of Santalum ITS evolution and use of fossil-based calibrations yielded estimates for timing of speciation and colonization events in the Pacific, with dates of 1.0–1.5 million yr ago (Ma) and 0.4–0.6 Ma for onset of diversification of the two Hawaiian lineages. The results indicate that the previously recognized sections Polynesica, Santalum, and Solenantha, the widespread Australian species S. lanceolatum, and the Hawaiian species S. freycinetianum are not monophyletic and need taxonomic revision, which is currently being pursued.

Key Words: ETS • ITS • long-distance dispersal • molecular phylogeny • Pacific biogeography • Santalaceae • Santalum • 3' • trnK intron

Covering half of the Earth's surface, the Pacific Ocean is dotted with over 20 000 islands (Carson, 1996 ; Keast, 1996 ), including the world's most isolated major island group, the volcanic Hawaiian Islands, over 2000 miles from the nearest continent. Derivation of the Pacific Island flora has intrigued botanists for the past century (Guppy, 1906 ; Campbell, 1919 ; Brown, 1921 , 1935 ; Fosberg, 1948 ; Thorne, 1963 ; Carlquist, 1967 , 1970 , 1974 , 1996 ; Sakai et al., 1995 ; Carson, 1996 ; Mueller-Dombois and Fosberg, 1998 ; Wagner et al., 1999), and much progress in understanding Pacific Basin biogeographic patterns has been made over the last 15–20 years with molecular phylogenetic reconstruction of dispersal patterns and timing of colonization events. These phylogenetic hypotheses, for example, have revealed that the Hawaiian angiosperm flora has been derived from across the globe, including from other Pacific Islands (Howarth et al., 1997 ; Wright et al., 2000 , 2001 ; Gemmill et al., 2002 ; Nepokroeff et al., 2003 ), the Americas (Baldwin et al., 1991 ; DeJoode and Wendel, 1992 ; Baldwin and Robichaux, 1995 ; Howarth et al., 1997 ; Pax et al., 1997 ; Vargas et al., 1998 ; Costello and Motley, 2001 ; Lindqvist and Albert, 2002 ; Wanntorp and Wanntorp, 2003 ), the subarctic (Ballard and Sytsma, 2000 ; Wagner et al., 2005 ), Australia (Howarth et al., 2003 ), Southeast Asia (Hao et al., 2004 ), and Africa (Seelanan et al., 1997 ; Kim et al., 1998 ). Several studies have revealed single colonizations of the Hawaiian Islands of unresolved origin (Molvray et al., 1999 ; Powell and Kron, 2002 ; Steinmann and Porter, 2002 ; Hughes et al., 2003 ; Morden and Gregoritza, 2005 ); for example, Hawaiian Tetramolopium (Okada et al., 1997 ; Lowrey et al., 2005 ), Bidens (Ganders et al., 2000 ), and Cyrtandra (Cronk et al., 2005 ) have sister clades in the South Pacific, leaving open the possibility of dispersal out of the Hawaiian Islands to the South Pacific. Few instances of multiple dispersal events within one angiosperm genus to the Hawaiian Islands have been discovered, including two for Rubus (Howarth et al., 1997 ) and three for Scaevola (Howarth et al., 2003 ).

Until now, the phylogeny and biogeography of one of the most widespread naturally occurring plant groups in the Pacific, the sandalwoods, have not been examined using modern analytical methods. Sandalwood is commonly known for its fragrant heartwood oil used to scent incense and perfume (Cherrier, 1991 ; Whistler, 1992 ; Barrett and Fox, 1995 ; Kepler, 1998 ; Payne, 2003 ). The sandalwood genus, Santalum L. (Santalaceae), is a hemiparasitic genus that includes 15 extant species, approximately 14 varieties, and one recently extinct species, distributed throughout India, Australia, and the Pacific Islands (Table 1). It is distributed as far southeast as the Juan Fernandez Islands, only 600 miles off the coast of Chile, and as far northwest as the Bonin Islands, 600 miles south of Honshu, Japan (van Balgooy, 1960 , 1971 ; van Balgooy et al., 1996 ). Santalum was chosen for research because of the potential for fine-scale resolution of biogeographic patterns in a particularly widespread, indigenous Pacific angiosperm group.

Demand for the valuable oil of sandalwoods has resulted in drastic over-harvesting; Santalum is one of the most heavily exploited groups of plants across its range (Brennan and Merlin, 1991 ). The species endemic to the Juan Fernandez Islands (S. fernandezianum) became extinct in the last century because of human exploitation (Darwin, 1839 ; Stuessy et al., 1992 ). The last standing tree of S. yasi in Samoa is reported to have been cut within the past few years (A. Whistler, University of Hawaii, personal observation). One variety in the Hawaiian Islands (S. freycinetianum var. lanaiense) is currently listed as endangered by the U.S. Fish and Wildlife Service, and two other taxa are considered threatened by extinction, S. insulare var. hendersonense (Henderson Island) and S. boninense (Bonin Islands) (Maina et al., 1988 ; Waldren et al., 1995 ). Understanding the genetic diversity and relationships of these taxa is crucial to appropriately manage and conserve the existing populations of sandalwoods.

Over one quarter of named Santalum species are endemic to the Hawaiian Islands, and differing hypotheses have been proposed to explain the number and sources of colonizations to this area. Several authors have suggested that the diversity of Santalum in the Hawaiian Islands is a result of two independent colonizations, one Polynesian and one Austral in origin (Skottsberg, 1930b ; Fosberg, 1948 ; Wagner et al., 1999 ). Two of the Hawaiian species (S. ellipticum, S. paniculatum) have been placed in the Hawaiian endemic section Hawaiiensia and are believed to have resulted from a colonization event from Polynesia, where Santalum is represented by one recognized species, S. insulare (Skottsberg, 1930a , b ; Wagner et al., 1999 ). The other two Hawaiian species (S. freycinetianum, S. haleakalae) were recognized as a second Hawaiian endemic section Solenantha by Tuyama (1939) , although Skottsberg (1930b) and Wagner et al. (1999) placed them in section Santalum, implying a West Pacific or Austral origin. In contrast, Merlin and Van Ravenswaay (1990) asserted that all of the Hawaiian species are the result of adaptive radiation from a single colonizing ancestor. Such rapid ecological diversification often occurs in island archipelagos (see, e.g., Carlquist, 1974 ; Schluter, 2002 ).

Various regional, floristic, or taxonomic treatments discuss relationships and distributional patterns within Santalum (Brown, 1810 ; Candolle, 1856 ; Bentham and von Muller, 1873 ; Hillebrand, 1888 ; Bailey, 1902 ; Rock, 1916 ; Skottsberg, 1927 , 1930a , b , 1934 , 1938 , 1944 ; Sprague and Summerhayes, 1927 ; Brown, 1935 ; Tuyama, 1939 ; Degener, 1937 ; Stemmermann, 1980 ; Sykes, 1980 ; Hewson and George, 1984 ; St. John, 1984 ; Fosberg and Sachét, 1985 ; Hallé, 1988 ). However, all previous studies have lacked a molecular phylogenetic context for evaluating evolutionary and biogeographic questions in the genus. Here, phylogenetic hypotheses based on molecular data are used for the first time to investigate the monophyly and relationships of taxa and to examine biogeographic patterns, including the timing, number, and source(s) of colonization events to—and from—the Hawaiian Islands.

MATERIALS AND METHODS

Taxonomic sampling
Taxonomic sampling was designed to cover the range of morphological, ecological, and biogeographical diversity in the genus without over-reliance on previous taxonomies. A total of 87 specimens were used in the analyses (85 ingroup, two outgroup; Appendix 1). All previously recognized extant and extinct species and most previously recognized varieties were sampled. To test species concepts and detect possible intraspecific differentiation in widespread species, we analyzed multiple exemplars representing different populations within species. Two outgroup species, Colpoon compressum Berg. and Osyris alba L., were chosen to root the phylogeny, based on the molecular phylogeny of the Santalales (Nickrent and Malécot, 2001 ).

Molecular character sampling
The molecular characters chosen for this analysis were selected based on their utility in previous studies to resolve species relationships. The internal transcribed spacer (ITS) and external transcribed spacer (ETS) regions of 18S–26S nuclear ribosomal DNA have similar utility and have proven to be useful sources of phylogenetic characters, especially in recently diverged lineages (Baldwin et al., 1995 ; Baldwin and Markos, 1998 ; Baldwin and Sanderson, 1998 ). The 3' trnK intron has been shown to be among the most rapidly evolving regions in the chloroplast genome (Johnson and Soltis, 1995 ; Shaw et al., 2005 ).

Extraction and amplification of DNA
Leaf samples from fresh, silica-dried (Chase and Hillis, 1991 ), and herbarium specimens were ground to fine powders, from which total genomic DNAs were extracted using the DNeasy Plant Mini Kit (Qiagen, Valencia, California, USA). For the polymerase chain reaction (PCR) amplifications, total genomic DNA was diluted in distilled water from 1 : 1 to 1 : 50, depending on the age and drying method of the leaf material, with older herbarium material requiring less dilution.

The 3' trnK intron was amplified using the published universal primers matK8 and trnK2R (Steele and Vilgalys, 1994 ; Johnson and Soltis, 1995 ) and for ITS using the published primers ITS4 and ITS-I (White et al., 1990 ; Urbatsch et al., 2000 ). A group-specific ETS primer was designed after an initial amplification using the Expand High Fidelity PCR System Kit (Roche Molecular Biochemicals, Indianapolis, Indiana, USA) with published primers 18S-IGS and 26S-IGS (Baldwin and Markos, 1998 ), using the following cycling parameters: 95°C for 4 min followed by 35 cycles of 95°C for 30 s, 50.5°C for 45 s, and 72°C for 4 min. The cycling ended with 50.5°C for 1 min 20 s and 72°C for 8 min. The group-specific primer DH-ETS1 [5' CGC CAR TGT GCR YGA YGC GT 3'] was used with the universal primer 18S-ETS to amplify the ETS region (Baldwin and Markos, 1998 ). Fragments of the 3' trnK intron, ITS, and ETS regions were amplified using Bioneer AccuPower PCR Pre-Mix tubes (Bionexus, Oakland, California, USA) in 20-µL reactions on a thermal cycler with the following parameters: 96°C for 1 min followed by 40 cycles of 96°C for 10 s, 48°C for 30 s, and 72°C for 20 s (plus 4 s per cycle). The cycling ended with 72°C for 7 min.

Sequencing and alignment
PCR products were cleaned using the Exo-Sap Pre-Sequencing Kit (USB Corp., Cleveland, Ohio, USA), followed by standard cycle-sequencing protocols using the ABI Prism BigDye Terminator kit, version 2.0 or 3.1 (Applied Biosystems, Foster City, California, USA) and precipitation of sequencing products with 75% isopropanol. The same primers used in amplification were used in cycle-sequencing, except the internal ITS-5A primer (Downie and Katz-Downie, 1996 ) was used instead of ITS-I. Direct sequencing of both the 5' and 3' products was done using an ABI 377 (Applied Biosystems, Foster City, California, USA) automated sequencer. Sequences were visualized and edited using the ABI Prism software Sequence Navigator.

Sequences of the ITS and ETS regions and 3' trnK intron were aligned in PAUP* 4.0b10 (Swofford, 2002 ) by eye. All Santalum sequences were easily aligned, but ambiguities arose in the alignment of sections of ITS sequences between the outgroup and Santalum. For these small sections of outgroup sequence, bases were replaced with question marks to represent unknown data because homology could not be assessed confidently.

Phylogenetic analyses
Maximum parsimony (MP) analyses were performed on separate (ITS, ETS, or 3' trnK intron) and combined (ITS, ETS, and 3' trnK intron) data sets, with characters unordered and unweighted. To account for information in the insertions and deletions (indels) inferred from aligned sequences, we recoded gaps greater than two base pairs as binary characters, regardless of length, using the simple gap-coding approach of Simmons and Ochoterena (2000) . This method treats identical gaps in sequences as potential evidence for common ancestry. A total of 14 gap characters were included (ITS, 10; ETS, 4; 3' trnK intron, 0). Congruence among the three nuclear and chloroplast data sets was tested using the incongruence length difference (ILD) test (Farris et al., 1994 , 1995 ), as implemented in PAUP* 4.0b10 (as the "partition homogeneity test"; Swofford, 2002 ), with 1000 ILD replicates, 10 000 random addition sequences per replicate, tree-bisection-reconnection (TBR) branch swapping, and MulTrees off.

For ITS, ETS, and combined (ITS, ETS, and 3' trnK intron) data sets, heuristic searches were performed in PAUP* with 100 random-addition sequences, TBR branch swapping, and MulTrees on. For the smaller 3' trnK intron data set, a branch-and-bound search was performed. Before all MP analyses were ran, identical sequences in parsimony informative characters were removed. Santalum fernandezianum was not included in the nuclear or combined data sets because the ITS and ETS regions could not be amplified.

Parsimony bootstrap support values were calculated using PAUP*. For the nuclear and combined data sets, 1000 bootstrap replicates were run with a full heuristic search, 10 000 random-addition sequences, TBR branch swapping, and MulTrees off (holding one tree at each step). For bootstrap analysis of the 3' trnK intron data set, the same parameters were used except MulTrees was left on, without exceeding tree-storage capacity.

A maximum likelihood (ML) analysis in PAUP* was run on the combined data set (minus the gap characters) using the best-fit model TrN+G, as chosen by Modeltest v3.6 (Posada and Crandall, 1998 ). A heuristic search was performed with 10 random-addition sequence replicates and TBR branch-swapping. Taxa with identical sequences were removed prior to analyses. ML bootstrap values were computed in PAUP* by running 100 replicates with a full heuristic search using 10 random-addition sequences, nearest-neighbor-interchange (NNI) branch swapping, and MulTrees on.

A Bayesian analysis was conducted on the same data set used in the ML analysis using MrBayes version 3.1.2 (Ronquist and Huelsenbeck, 2003 ). The same nucleotide substitution model as in the ML analysis was used. The analysis was performed for three million generations with four chains, with samples taken every 100 generations. Upon completion of the analysis, likelihood scores were plotted against generation number to determine the burn-in period, which was approximately 50 000 generations. All of the trees were loaded into PAUP*, and after removing the first 500 trees, a majority rule consensus tree was generated to compute posterior probabilities.

The conservative Shimodaira–Hasegawa (S-H) test was used to examine the significance of alternative hypotheses of relationships in Santalum, including the monophyly of taxa and geographical areas (Shimodaira and Hasegawa, 1999 ; Goldman et al., 2000 ). An S-H test was conducted for each phylogenetic hypothesis generated from ML analysis under a particular topological constraint. S-H tests were performed in PAUP* using RELL (reestimated log likelihood) bootstrap with 1000 replicates and a one-tailed test, and the likelihood settings the same as in the ML analysis.

Molecular clock test and divergence time estimation
Each of the gene regions analyzed was tested for clock-like evolution, i.e., evolution at a roughly constant rate over time (Zuckerkandl and Pauling, 1965 ), across lineages using Felsenstein's (1988) likelihood ratio test and the ML tree topology from analysis of combined data. For each gene, likelihoods of the ML tree with and without enforcement of a molecular clock were compared using PAUP*. If rate constancy was not rejected, then divergence times and nucleotide substitution rates were estimated using the computer program r8s v. 1.71 (Sanderson, 2003 ). Because of the lack of a fossil record in Santalum, absolute timing of divergence events within Santalum was conservatively (over)estimated using the external calibration of 32–48 million years ago (Ma) for the split of Santalum-Osyris, as estimated by the fossil-calibrated molecular phylogeny of Wikström et al. (2001) . The divergence times in Wikström et al. (2001) were chosen for this study because they represent a conservative range based on both parsimony and maximum likelihood estimates.

Biogeographic analyses
Geographical areas were mapped onto the combined ML tree (= Bayesian tree) using the parsimony criterion to elucidate major, natural dispersal patterns in the genus. Geographic distribution was converted into six unordered characters, (1, Outgroup: Europe, Africa; 2, Austral: Australia, Papua New Guinea; 3, Melanesia: Fiji, Vanuatu, New Caledonia; 4, Hawaiian Islands; 5 eastern Polynesia: Austral Is., Cook Is. Marquesas Is., Pitcairn Is., Society Is.; 6, Bonin Islands) and traced onto the tree using MacClade v.4.06 (Maddison and Maddison, 2000 ). These areas were delimited based on their discreteness and geological histories. Specimens collected in areas where they are known to have been introduced or cultivated were coded from their area of natural origin.

RESULTS

Phylogenetic analyses
The total aligned length for the 3' trnK intron was 404 characters, of which only 21 (5.2%) were parsimony informative. Parsimony analysis of the 3' trnK intron data resulted in 11 MP trees (length [L] = 47, consistency index [CI] = 0.8958, retention index [RI] = 0.8571; not shown). Because of the low number of parsimony-informative characters, the chloroplast data set resulted in trees that are poorly resolved and weakly supported.

The total aligned length for the ITS region (ITS1, 5.8S gene, and ITS2) was 738 characters, of which 179 (24.3%) were parsimony informative, not including 10 recoded gap characters. Aligned length for the sequenced ETS region was 453 characters, of which 153 (33.8%) were parsimony informative, not including four recoded gap characters. Separate MP analyses of these two data sets resulted in strict consensus trees (not shown) that are congruent with one another and with the strict consensus of combined trees in Fig. 1. The ETS strict-consensus topology is more poorly resolved than the strict consensus of ITS trees, except S. boninense and S. freycinetianum var. pyrularium are resolved within the S. insulare clade in the ETS tree and unresolved in the ITS tree. Additionally, a sister-group relationship between S. yasi and S. album was resolved in the ETS analysis but not in the ITS analysis.

View larger version (56K): [in this window] [in a new window]   Fig 1. Santalum phylogeny estimated from combined ITS, ETS, and 3' trnK data. Maximum likelihood (ML) and Bayesian topologies are identical and shown here. The maximum parsimony strict-consensus topology differs from this tree only in the clade indicated by an asterisk (*). Numbers above branches represent support values from (left to right) ML bootstrap, Bayesian, and MP bootstrap analyses. Section Santalum is highlighted in gray, and previously recognized Santalum sections and the formerly recognized genus Eucarya are indicated on the right of the phylogeny.

The combined data matrix contained 1595 aligned sequence characters and 14 recoded gap characters for 84 Santalum specimens and two outgroup taxa. The MP analysis yielded 42 122 trees (L = 798, CI = 0.8202, RI = 0.9166). The ML analysis yielded one tree (–ln L = 6331.8378; Fig. 1) that is well resolved, with most major clades having high support. The ML tree has an identical topology to the Bayesian tree; the MP tree differs only in resolution within S. acuminatum (Fig. 1). ML and MP bootstrap values are similar, with Bayesian posterior probabilities generally higher (Fig. 1).

The combined MP analysis resulted in a strict-consensus tree that is congruent with, but more fully resolved than, consensus trees from the separate ITS and ETS MP analyses. The relationships of S. macgregorii, S. freycinetianum var. pyrularium, S. boninense, and S. obtusifolium were resolved in the combined analyses only. The only clade that was unresolved in the combined MP analysis but resolved in a separate data tree is within S. acuminatum. The topology within this clade in the ITS tree is not the same as for the ML and Bayesian trees from the combined data set (Fig. 1).

Despite the lack of resolution and support in the 3' trnK intron tree (not shown), analyses of the cpDNA data provided the only evidence for a sister-group relationship between the extinct S. fernandezianum and S. acuminatum (because no nuclear material was successfully sequenced). Topological incongruities between the cpDNA and nuclear rDNA trees involved cpDNA clades supported by only one character change. For example, S. macgregorii and S. boninense swap positions in the 3' trnK intron and combined trees. Resolution within S. acuminatum in the cpDNA tree is identical to the combined ML and Bayesian trees but not to the combined MP tree. Based on results of the ILD test, these incongruities may result simply from insufficient characters in the cpDNA data set to counteract homoplasy, although lineage sorting and hybridization cannot be ruled out.

Taxonomic groupings
All but two of the named species are resolved as monophyletic (Fig. 1). Santalum freycinetianum appears to be paraphyletic, with a weakly supported node uniting S. boninense and S. insulare with S. freycinetianum var. pyrularium, to the exclusion of other S. freycinetianum varieties. Based on nuclear rDNA and combined analyses, the widespread Australian taxon S. lanceolatum represents two well-supported and widely separated clades; one is sister to S. austrocaledonicum, and the other is a member of a clade including S. macgregorii, S. ellipticum, and S. paniculatum.

Regarding the named sections within Santalum (see Fig. 1), the endemic Hawaiian section Hawaiiensia is monophyletic, while the other Hawaiian endemic section Solenantha is paraphyletic, with S. insulare (sect. Polynesica) and S. boninense (sect. Santalum) nested within it. If S. fernandezianum is sister to S. acuminatum, as seen in the 3' trnK intron tree (not shown), then section Polynesica is polyphyletic. Section Santalum is paraphyletic, with the other three sections nested within it.

Divergence times
Rate constancy of molecular evolution across lineages was rejected for all gene regions except ITS1+ITS2 (–2 ln LR = 57.8, df = 48, 0.25 > P > 0.10). Given the conservative calibration of the base of Santalum using the published dates for the split of Santalum-Osyris, the range of substitution rates estimated for Santalum ITS1+ITS2 is at least 2.49 x 10^–9 to 3.74 x 10^–9 substitutions/site/year. The chronogram showing the range of estimated divergence times is in Fig 2.

View larger version (24K): [in this window] [in a new window]   Fig. 2. The ITS1+2 chronogram of Santalum, showing the range of divergence time estimates in millions of years, using the calibration for the split of Santalum-Osyris from Wikström et al. (2001) .

View larger version (42K): [in this window] [in a new window]   Fig. 3. Biogeographical patterns of Santalum. (A) The long-distance dispersal patterns of Santalum are demonstrated by optimizing the native geographical areas using parsimony on the maximum likelihood (ML)/Bayesian topology estimated from combined ITS, ETS, and 3' trnK intron data. Arrows indicate the hypothesized movement of Santalum by people. (B) Map of the current distribution and hypothesized natural dispersal patterns of Santalum, inferred from optimization of geographical areas as shown in Fig. 3A. Solid arrows represent unequivocal reconstruction, and dotted lines represent equivocal reconstructions of natural dispersal patterns. The dispersal to Juan Fernandez Islands is resolved based on the 3' trnK intron phylogeny (not shown).

Although S. fernandezianum was not included in the combined analyses (Fig. 1), in the 3' trnK intron tree (not shown) this endemic of the Juan Fernandez Islands is nested among Australian taxa, indicating an origin from that continent. These results show at least five long-distance dispersal events from Australia to the Pacific and at least one dispersal from the Hawaiian Islands in Santalum, as shown in Fig. 3B.

Alternative topologies
Results of tests for alternative topologies (S-H test) indicated rejection of monophyly of the Hawaiian sandalwoods (S. ellipticum, S. freycinetianum, S. haleakalae, and S. paniculatum; P = 0.000*), thereby providing additional evidence for multiple, independent colonizations of the Hawaiian Islands. The hypothesis of polyphyly of S. lanceolatum was reinforced by the significant rejection of monophyly for that species (P = 0.002*). Monophyly could not be rejected for S. freycinetianum (P = 0.159), S. freycinetianum with S. boninense (P = 0.169), or S. freycinetianum with S. insulare (P = 0.141); relationships among these three taxa are evidently tenuous. However, the best trees from the MP, ML, and Bayesian analyses indicate paraphyly of S. freycinetianum and yield an unequivocal resolution of dispersal out of the Hawaiian Islands, based on parsimony optimization of areas on those trees.

DISCUSSION

Classification
Results of this study indicate that the sectional classification of Santalum needs substantial revision. Previous taxonomic groupings and hypotheses of relationships in Santalum have been based entirely on a few morphological characters, such as floral tube color, ratio of corolla length to width, and placement of the ovary. For example, section Santalum is described as usually having reddish corollas that are longer than wide and partly superior ovaries (Skottsberg, 1930b ; Stemmermann, 1980 ; Wagner et al., 1999 ). Tuyama (1939) separated the Hawaiian members of section Santalum into the endemic section Solenantha based on their longer perianth tubes, smaller ovaries, and absence of hairs proximal to the filaments. Section Hawaiiensia is described as having white, green, brown, or orange corollas that are as wide as long, and inferior ovaries (Skottsberg, 1930b ; Stemmermann, 1980 ; Wagner et al., 1999 ). Section Polynesica is described as appearing similar to Hawaiiensia but with ovaries that are partly superior (Skottsberg, 1930b ).

Molecular phylogenetic data for Hawaiian and Polynesian taxa reveal problems with taxonomic emphasis on conspicuous corolla and ovary characters in Santalum. On the basis of morphological similarities in such characters, Skottsberg (1930a , b ) hypothesized that sections Hawaiiensia and Polynesica were closely related, and Fosberg and Sachét (1985) treated section Polynesica as a synonym of section Hawaiiensia. Results of molecular phylogenetic analyses, however, demonstrate that sections Hawaiiensia and Polynesica are more closely related to other taxa of Santalum than to one another (Fig. 1) and should not be united taxonomically; floral tube color and proportions and placement of the ovary have considerable intraspecific variation and homoplasy in this group and are therefore unreliable indicators of evolutionary relationships.

Competing taxonomic treatments and hypotheses about relationships of the Juan Fernandez endemic, Santalum fernandezianum, can now be re-evaluated based on phylogenetic data. Lack of availability of much floral material of this extinct species has contributed to uncertainty about its taxonomic placement. Sprague and Summerhayes (1927) transferred S. fernandezianum to Mida, a genus that they allied with the formerly recognized Australian genus Eucarya (= Santalum pro parte), because of the short, cupular perianth tubes and sessile stigmas in both the Juan Fernandez endemic and the type species of Mida (M. salicifolia). Skottsberg (1930b) treated the Juan Fernandez species within Santalum sect. Polynesica. The chloroplast phylogeny (not shown) suggests that S. fernandezianum is sister to S. acuminatum, formerly in Eucarya; evidently, Sprague and Summerhayes (1927) were not so far off after all.

Results of this study show a high amount of lineage diversity in the sole living species of Santalum sect. Polynesica, S. insulare, in accord with morphological and taxonomic diversity in these South Pacific sandalwoods. Three strongly supported clades of S. insulare discovered from combined molecular data represent three different areas: (1) Marquesas Islands, (2) Society Islands, and (3) Henderson, Austral, and Cook islands (Fig. 1). The basal trichotomy may be explained by an early, rapid set of dispersal events to these three regions, with subsequent radiation within each area. In a recent phylogeographic study, Butaud et al. (2005) identified the same three geographical affinities and determined that S. insulare was centered in the Cook and Austral islands, with subsequent dispersal to the Marquesas and the Society Islands.

Phylogenetic findings for S. lanceolatum indicate previously undetected diversity in this widespread Australian taxon. Based on floral morphology, Skottsberg (1930a) suggested that S. lanceolatum is most closely related to another Australian species, S. macgregorii. Phylogenetic results from combined data indicate two clades of S. lanceolatum—one, including all specimens from Northern Australia, is resolved as most closely related to S. macgregorii, as proposed by Skottsberg, and sect. Hawaiiensia (Fig. 1). Though no hypotheses have been set forth to separate S. lanceolatum into two species, considerable variation in leaf shape, color, and size have been reported across its geographic range (Applegate et al., 1990 ); for example, collections from Northern Queensland have been described as having smaller flowers and discolored leaves (Hewson and George, 1984 ). Results from analysis of 3' waxy sequences of Santalum and morphological considerations offer additional perspectives on relationships and appropriate taxonomic treatment of members of S. lanceolatum; a complete taxonomic revision of the genus will be published as a monograph (D. T. Harbaugh, unpublished manuscript).

Biogeography and timing of divergence events
The conservative estimates of substitution rates for Santalum ITS1+ITS2 (2.49 x 10^–9 to 3.74 x 10^–9 substitutions/site/year) fall well within the range of published rates for ITS1+ITS2 in other woody angiosperms (Richardson, 2001 ; Kay and Whittall, 2006 ). The ITS chronogram (Fig. 2) shows most of the major lineages appearing no later than 6.3–9.5 Ma and radiating since 1.0–1.5 Ma. These results indicate either biased extinction of earlier branches, or rapid, recent radiation in Santalum, which is consistent with dispersal and isolation on islands.

Biogeographic reconstructions indicate an Australian origin of Santalum. Multiple instances of long-distance dispersal out of Australia and between Pacific Islands to account for most major disjunctions in the Pacific distribution in the genus. The only two islands on which Santalum naturally grows that are not of volcanic origin are New Caledonia and New Guinea, both previously connected to Australia. New Caledonia separated from Australia 65–74 Ma (Keast, 1996 ), long before the origin of the genus, based on the fossil-calibrated ITS tree (Fig. 2). Therefore, long-distance dispersal must account for the occurrence of S. austrocaledonicum on New Caledonia. New Guinea and Australia were periodically joined until about 7000 years ago because of sea level fluctuations associated with repeated glaciations over the past 2 million years (Keast, 1996 ). Based on the molecular chronogram (Fig. 2), the stem lineage of S. macgregorii, which is endemic to Papua New Guinea, diverged 2.5–3.7 Ma, before the vicariance event that separated New Guinea from the Australian landmass.

Results from the molecular phylogenetic analyses uphold and extend the hypothesis of Skottsberg (1930b) , Fosberg (1948) , and Wagner et al. (1999) that two long-distance dispersal events account for the diversity of Santalum in the Hawaiian Islands, with at least one being Austral in origin. The tree-based biogeographic reconstructions (Fig. 3) indicate that each of the two Hawaiian colonization events originated in Australia or Papua New Guinea. According to W. L. Wagner and J. Price (Smithsonian Institution, personal observation), only about 5% of the native Hawaiian angiosperm founder lineages descended from Australian ancestors; two independent dispersal events from Australia represent approximately one seventh of all natural Hawaiian colonizations of Austral origin. Divergence-time estimates indicate that the first dispersal event from Australia to the Hawaiian Islands occurred ca. 1.0–1.5 Ma, resulting in the S. freycinetianum/S. haleakalae grade, and the second from Australia or Papua New Guinea (which was connected to Australia at the time of dispersal) occurred only ca. 411 to 616 thousand years ago, resulting in the S. ellipticum/S. paniculatum clade. These divergence-time estimates assume that ages of the most recent common ancestors of each Hawaiian clade correspond with timing of dispersal out of Australia (both colonizations could be somewhat older if extinction has pruned the Hawaiian stem lineages). Both of these dispersal events to the Hawaiian Islands are well within the age range of all the modern, high Hawaiian Islands or island groups except the island of Hawai‘i, which formed within the last 400 000 years (Carson and Clague, 1995 ).

The dispersal patterns of Santalum are extraordinary and unique for Hawaiian plants. Not only is Santalum one of very few examples of multiple colonizations of the Hawaiian Islands by angiosperms, it is the only angiosperm group for which a molecular phylogeny has yielded an unequivocal parsimony reconstruction of dispersal out of the Hawaiian Islands, possibly twice, to the Bonin Islands and eastern Polynesia. In Bonin, only 1% of the flora is derived from the Pacific, with a majority of taxa from Asian sources (van Balgooy, 1960 ). Eastern Polynesia and the Hawaiian Islands share many lineages with closely related species in each area, although most of these lineages are believed to have originated in Asia, Australia, and the Pacific, with dispersal northward from Polynesia to the Hawaiian Islands (Brown, 1935 ; Mueller-Dombois and Fosberg, 1998 ). Santalum is a good example of why regional subdivision of the Pacific flora is hazardous; movements of plants throughout the Pacific can be highly individualistic (Carlquist, 1996 ; van Balgooy et al., 1996 ).

Several specimens of S. album from India and Australia were identical across all three genes, indicating a recent dispersal, consistent with the movement of this species by people. Santalum album from Indonesia was not included in this study because material was unavailable; Indonesia has been reported to be the possible source area of S. album in India and Australia (Roxburgh, 1820 ; Sprague and Summerhayes, 1927 ; Tuyama, 1939 ; St. John, 1947 ; McKinnell, 1990 ; Rai, 1990 ). If so, a short-distance dispersal between Australia and Indonesia could have occurred, and then long-distance dispersal by people to India may have followed.

Modes of dispersal
Various mechanisms allow for long-distance, chance dispersal of plants across oceanic islands, such as wind, rafting, floating in ocean currents, or movement by birds (Campbell, 1919 ; Thorne, 1963 ; Cruden, 1966 ; Carlquist, 1966 , 1967 , 1996 ; Cain et al., 2000 ). The red, purple, or black fleshy fruits of sandalwoods are attractive to birds; in addition, the thick, hard endocarp can probably protect the seed from mechanical damage or digestion, so transport in the gut of frugivorous birds is the most likely means for long-distance dispersal of Santalum (Guppy, 1906 ; Carlquist, 1970 , 1974 ; Sakai et al., 1995 ). Fruits of several species of Santalum have been observed being eaten by birds (S. acuminatum, S. lanceolatum, S. murrayanum, S. spicatum [Ford and Paton, 1986 ]; S. album [Ridley, 1930 ; Johnson, 1981 ]; S. austrocaledonicum [Bottin et al., 2005 ]). Pacific golden plovers (Pluvialis pulva) and Pacific pigeons (Ducula pacifica) are important seed dispersers and must be considered as potential vectors in the movement of sandalwoods throughout the Pacific (Carlquist, 1966 ; McConkey et al., 2004 ).

Regardless of the precise means of dispersal in sandalwoods, historical biogeographic patterns resolved in this study underscore the extraordinary ability of plants to disperse without the aid of people throughout the Pacific islands, as suggested earlier by Carlquist (1974). The biogeographic hypotheses proposed here and new evidence for polyphyletic taxa and cryptic biodiversity in Santalum are subjects of continuing molecular and cytogenetic investigation by D. T. Harbaugh. Enhanced phylogenetic understanding and taxonomic revision of the sandalwoods are critical to properly manage and conserve these economically and culturally important plants for future generations.

APPENDIX 1.

Specimens sampled for phylogenetic analyses. Numbers after taxon name follow numbering in Figs. 1, 2, and 3A. A dash indicates missing data. Locality abbreviations are given for the Hawaiian Islands, USA (HI) and for the following states in Australia: New South Wales (NSW), Northern Territory (NT), Queensland (QLD), South Australia (SA), Victoria (VIC), and Western Australia (WA). Herbarium abbreviations follow standard herbarium acronyms from the Index Herbariorum.

Taxon—Sample no.: GenBank accessions ITS, ETS, 3' trnK intron; locality; Voucher specimen (herbarium).

Colpoon compressumBerg.— EF569288, EF569374, EF569233; South Africa; Harbaugh 63 (UC).

Osyris albaL.— EF569287, EF569373, EF569232; Greece; Harbaugh 61 (UC).

Santalum acuminatum(R.Br.) A.DC.—1: EF569290, EF569376, EF569236; SA; Harbaugh 94 (UC). 2: EF569289, EF569375, EF569235; NSW; Harbaugh 125 (UC). 3: EF569291, EF569377, EF569237; VIC; Harbaugh 116 (UC). 4: EF569292, EF569378, EF569238; WA; Harbaugh 142 (UC). 5: EF569294, —, —; WA; Stauffer and Royce 5343 (UC). 6: EF569295, —, —; WA; Harbaugh 129 (UC). 7: EF569296, —, —; WA; Harbaugh 132 (UC). 8: EF569293, EF569379, EF569239; WA; Harbaugh 151 (UC).

Santalum albumL.—1: EF569317, EF569394, EF569252; India; Harbaugh 33 (UC). 2: EF569320, —, —; India; Garcia 60 (UC). 3: EF569319, EF569396, EF569254; India; Pant s.n (UC). 4: EF569321, —, —; Melville Island, NT; Dunlop 1976 (CANB). 5: EF569318, EF569395, EF569253; NT; Harbaugh 65 (UC).

Santalum austrocaledonicumViell. var.austrocaledonicum—1: EF569325, EF569399, —; Santo, Vanuatu; Page SnSo12 (JCT). 2: EF569326, EF569400, EF569257; Vanuatu; Garcia 39 (UC). 3: EF569327, EF569401, EF569258; Erromango, Vanuatu; Page SnEr16 (JCT). 4: EF569328, EF569402, EF569259; Grande Terre, New Caledonia; Bottin GT46 (ALF). 5: EF569331, EF569405, EF569261; Lifou, New Caledonia; Bottin L90 (ALF).

Santalum austrocaledonicumViell. var.minutumN.Hallé—1: EF569329, EF569403, —; Grande Terre, New Caledonia; Suprin 1948 (NOU). 2: EF569330, EF569404, EF569260; Grande Terre, New Caledonia; Bottin GT57 (ALF).

Santalum boninense(Nakai) Tuyama—1: EF569350, EF569419, EF569273; Hahajima, Bonin Islands, Japan; Tuyama s.n. (UC). 2: EF569351, EF569420, EF569274; Hahajima, Bonin Islands, Japan; Kato 060001 (UC). 3: EF569352, EF569421, EF569275; Chichijima, Bonin Islands, Japan; Kato 060002 (UC).

Santalum ellipticumGaudich. var.ellipticum—1: EF569361, EF569429, EF569282; O‘ahu, HI; Harbaugh 07 (UC). 2: EF569365, —, —; O‘ahu, HI; Harbaugh 15 (UC). 3: EF569367, EF569431, EF569284; O‘ahu, HI; Welton et al. 162 (BISH). 4: EF569366, EF569430, EF569283; O‘ahu, HI; Takeuchi 3223 (BISH). 5: EF569364, —, —; Kaua‘i, HI; Flynn et al. 2760 (BISH). 6: EF569363, —, —; Lana‘i, HI; Imada and Puttock 2003–193 (BISH). 7: EF569362, —, —; Maui, HI; Nagata 4448 (BISH).

Santalum ellipticumGaudich. var.littorale(Rock) Skotts.— EF569368, EF569432, EF569285; O‘ahu, HI; Harbaugh 21 (UC).

Santalum fernandezianumF.Phil— —, —, EF569234; Masatierra, Juan Fernandez Islands; Johow s.n. (UPP).

Santalum freycinetianumGaudich. var.freycinetianum—1: EF569359, EF569427, EF569280; O‘ahu, HI; Harbaugh 24 (UC). 2: EF569360, EF569428, EF569281; O‘ahu, HI; Lee s.n. (BISH). 3: EF569357, EF569425, EF569278; Moloka‘i, HI; Russell s.n. (BISH).

Santalum freycinetianumGaudich. var. lanaienseRock—1: EF569355, —, —; Maui, HI; Oppenheimer et al. H20011 (BISH). 2: EF569356, EF569424, EF569277; Maui, HI; Smith 2900 (UC). Santalum freycinetianumGaudich. var.pyrularium(A.Gray) Stemmerm.—1: EF569354, EF569423, EF569276; Kaua‘i, HI; Harbaugh 05 (UC). 2: EF569353, EF569422, —; Kaua‘i, HI; Flynn et al. 2744 (BISH).

Santalum haleakalaeHillebr.—EF569358, EF569426, EF569279; Maui, HI; D. Harbaugh 42 (UC).

Santalum insulareBertero ex A.DC. var.alticolaFosberg & Sachét— EF569338, EF569411, —; Tahiti, Society Islands, French Polynesia; Gagne and Montgomery s.n. (BISH).

Santalum insulareBertero ex A.DC. var.hendersonense(F.Br.) Fosberg & Sachét—1: EF569341, EF569414, —; Henderson, Pitcairn Islands; Florence et al. 10908 (BISH). 2: EF569346, EF569418, EF569271; Henderson, Pitcairn Islands; Tait 70 (UC).

Santalum insulareBertero ex A.DC. var.insulare—EF569344, EF569416, EF569269; Tahiti, Society Islands, French Polynesia. Butaud T116 (ALF).

Santalum insulareBertero ex A.DC. var.marchionenseSkottsb.—1: EF569342, EF569415, —; Ua Pou, Marquesas Islands, French Polynesia; Wood and Meyer 10377 (BISH). 2: EF569343, —, —; Tahuata, Marquesas Islands, French Polynesia; Perlman 18383 (BISH). 3: EF569337, EF569410, EF569266; Hiva Oa, Marquesas Islands, French Polynesia; Butaud HO24 (ALF). 4: EF569339, EF569412, EF569267; Nuku Hiva, Marquesas Islands, French Polynesia; Butaud NH8 (ALF).

Santalum insulareBertero ex A.DC. var.mitiaroSykes—EF569340, EF569413, EF569268; Mitiaro, Cook Islands; Whistler 5913 (BISH).

Santalum insulareBertero ex A.DC. var.raiateense(J.W.Moore) Fosberg & Sachét—1: EF569347, —, —; Mo‘orea, Society Islands, French Polynesia; Bouvet M6 (ALF). 2: EF569345, EF569417, EF569270; Raiatea, Society Islands, French Polynesia; Butaud R4 (ALF).

Santalum insulareBertero ex A.DC. var.raivavenseF.Br.—1: EF569348, —, EF569272; Raivavae, Austral Islands; McCormack s.n. (CHR). 2: EF569349, —, —; Raivavae, Austral Islands; Butaud RV1 (ALF).

Santalum lanceolatumR.Br.—1: EF569297, EF569380, EF569240; QLD; Page SnH05 (JCT). 2: EF569302, EF569383, EF569243; SA; Lay 213 (UC). 3: EF569307, —, —; SA; Harbaugh 101 (UC). 4: EF569308, —, —; SA; Stauffer and Wilson 5432 (UC). 5: EF569301, —, —; NSW; Lally and Landsberg 298 (CANB). 6: EF569298, EF569381, EF569241; QLD; Clemens s.n. (UC). 7: EF569299, EF569382, EF569242; NSW; Harbaugh 124 (UC). 8: EF569300, —, —; QLD; Fraser 12 (CANB). 9: EF569303, EF569384, EF569244; NT; Harbaugh 77 (UC). 10: EF569304, —, —; NT; Slee and Craven 2524 (CANB). 11: EF569306, EF569386, EF569246; QLD; Page SnCW24 (JCT). 12: EF569305, EF569385, EF569245; QLD; Page SnCW21 (JCT).

Santalum macgregoriiF.Muell.—1: EF569322, —, —; Apanaipai Village, Papua New Guinea; Unknown AA631 (UC). 2: EF569323, EF569397, EF569255; Ishea Village, Papua New Guinea; Unknown AA643 (UC). 3: EF569324, EF569398, EF569256; Leke Oalai Village, Papua New Guinea; Unknown AA603 (UC).

Santalum murrayanum(T.Mitch.) C.A.Gardner—1: EF569309, EF569387, EF569247; SA; Eichler 16206 (UC). 2: EF569310, EF569388, —; VIC; Harbaugh 121 (UC).

Santalum obtusifoliumR.Br.—1: EF569313, EF569390, —; NSW; Stauffer et al. 5692 (UC). 2: EF569311, —, —; NSW; Telford 10878 (CBG). 3: EF569312, EF569389, EF569248; QLD; Stauffer and Everist 5505 (UC).

Santalum paniculatumHook. & Arn. var.paniculatum—1: EF569369, EF569433, EF569286; Hawai‘i, HI; Harbaugh 47 (UC). 2: EF569370, —, —; Hawai‘i, HI; Harbaugh 50 (UC). 3: EF569371, —, —; Hawai‘i, HI; Perlman et al. 10289 (BISH).

Santalum paniculatumHook. & Arn. var.pilgeri(Rock) Stemmerm.— EF569372, —, —; Hawai‘i, HI; Takeuchi and Shinabukuro 5810 (BISH).

Santalum spicatum(R.Br.) A.DC.—1: EF569315, EF569392, EF569250; WA; Stauffer and Royce 5314 (UC). 2: EF569316, EF569393, EF569251; WA; Harbaugh 144 (UC). 3: EF569314, EF569391, EF569249; SA; Stauffer and Wilson 5426 (UC).

Santalum yasiSeem.—1: EF569332, EF569406, EF569262; Fiji; Garcia 59 (UC). 2: EF569333, EF569407, EF569263; Eua, Tonga; A. Whistler 7360 (BISH). 3: EF569336, —, —; Fiji; Garcia 62 (UC). 4: EF569335, EF569409, EF569265; Fiji; Garcia 63 (UC). 5: EF569334, EF569408, EF569264; Tonga; Larsen s.n. (UC).

FOOTNOTES

¹ The authors thank V. Garcia, R. Kirkpatrick, B. Mishler, A. Moore, A. Murdock, W. L. Wagner, and an anonymous reviewer for invaluable comments on the manuscript; L. Bottin, J.-M. Bouvet, J.-F. Butaud, V. Garcia, B. Gunn, M. Ito, A. Larsen, T. Page, and A. R. Smith for providing plant material; P. Ala, K. Harbaugh, B. Lepschi, F. Murdoch, L. Ocampo, T. Page, and D. Reynaud for assistance with field collection; the UC Molecular Phylogenetics Laboratory staff and B. Wessa for support of lab work; C. Hoisington for drawing Fig. 3B; and J.-F. Butaud, M. Byrne, D. Nickrent, and L. Thomson for helpful discussions. This study, part of the Ph.D. dissertation research of D.T.H. supervised by B. Baldwin at the University of California, Berkeley, was partly funded by a National Science Foundation Graduate Research Fellowship, a Department of Integrative Biology Summer Research Grant (UC Berkeley), Botany in Action (Phipps Conservatory and Botanical Gardens), an American Society for Plant Taxonomists Graduate Research Grant, a Hunt Institute for Botanical Documentation Lawrence Memorial Award, and a Santa Barbara Scholarship Foundation Graduate Fellowship to D.T.H.

² Author for correspondence (e-mail: danicah{at}berkeley.edu )

LITERATURE CITED

Applegate G. B. Chamberlain J. C. Daruhi G. Feigelson J. L. Hamilton L. McKinnell F. H. Neil P. E. Rai S. N. Rodehn B. Statham P. C. Stemmermann L.. 1990. Sandalwood in the Pacific, a state-of-knowledge synthesis and summary from the April 1990 symposium. In L. Hamilton, C. E. Conrad [eds.] Proceedings of the Symposium on Sandalwood in the Pacific, 1990,Honolulu, Hawaii, USA,1–11. Pacific Southwest Research Station, U. S. Department of Agriculture Forest Service, Berkeley, California, USA..

Bailey F. M.. 1902. Santalum. In The Queensland flora, 1384-1387. H. J. Diddams and Co., Brisbane, Australia..

Baldwin B. G. Kyhos D. W. Dvorak J. Carr G. D.. 1991. Chloroplast DNA evidence for a North American origin of the Hawaiian silversword alliance (Asteraceae). Proceedings of the National Academy of Sciences, USA 88: 1840-1843..[Abstract/Free Full Text]

Baldwin B. G. Markos S.. 1998. Phylogenetic utility of the external transcribed spacer (ETS) of 18S-26S rDNA: congruence of ETS and ITS trees of Calycadenia (Compositae). Molecular Phylogenetics and Evolution 10: 449-463..[CrossRef][ISI]