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

This Article Abstract Full Text (PDF) Supplementary Data 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 (9) Citing Articles via Google Scholar Google Scholar Articles by Motley, T. J. Articles by Delprete, P. G. Search for Related Content PubMed Articles by Motley, T. J. Articles by Delprete, P. G. Agricola Articles by Motley, T. J. Articles by Delprete, P. G.

Molecular systematics of the Catesbaeeae-Chiococceae complex (Rubiaceae): flower and fruit evolution and biogeographic implications¹

²Lewis B. and Dorothy Cullman Program for Molecular Systematics Studies, The New York Botanical Garden, Bronx, New York 10458-5126 USA; ³Institute of Systematic Botany, The New York Botanical Garden, Bronx, New York 10458-5126 USA

Received for publication December 12, 2003. Accepted for publication September 30, 2004.

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS LITERATURE CITED

 
The classification of the Catesbaeeae and Chiococceae tribes, along with that of the entire Rubiaceae, has long been debated. The Catesbaeeae-Chiococceae complex (CCC) includes approximately 28 genera and 190 species primarily concentrated in the Greater Antilles (nearly 70% of the species), Central and South America, and in the western Pacific (three genera). Previous molecular studies, with broad sampling of the Rubiaceae, have shown the CCC to be a monophyletic group. The present study is a more detailed examination of the generic relationships within the CCC using two data sets, the nuclear ribosomal ITS regions and the trnL-F chloroplast intron and spacer. Maximum parsimony analyses lend further support to the previous hypotheses that the CCC is monophyletic and sister to Strumpfia maritima. However, within the complex several genera do not form monophyletic groups. Previous studies of the Rubiaceae suggest that the ancestral fruit type in the CCC is a multiseeded capsule. Indehiscent, fleshy fruits appear to have evolved three to four times within this lineage. Changes in floral morphologies within the complex tend to correspond to cladogenesis among and within genera. Finally, molecular analyses suggest one or possibly two long-distance dispersals from the Americas to the western Pacific.

Key Words: biogeography • Caribbean • Catesbaeeae • Chiococceae • flower evolution • fruit evolution • islands • ITS • Neotropics • Pacific • Rubiaceae • systematics • trnL-F

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS LITERATURE CITED

 
The Rubiaceae is one of the largest and most diverse families of flowering plants, with approximately 650 genera and 13 000 species (Delprete, 2004 ), mostly of pantropical distribution. Historically fruit and seed characters have been used to infer the classification of the family and in several cases to define subfamilies (de Candolle, 1830 ; Hooker, 1873 ; Schumann, 1891 ). The subfamilial and tribal classification went through several minor modifications until Bremekamp (1966) , using many additional characters, divided the family into eight subfamilies and 43 tribes. The last comprehensive system of classification was proposed by Robbrecht (1988 , 1993 ), in which he recognized four subfamilies and 44 tribes. Recent phylogenetic evidence based on results of molecular studies with broad sampling throughout the family have revealed that the Rubiaceae should be best treated as three subfamilies— Cinchonoideae, Ixoroideae, and Rubioideae (e.g., rbcL, Bremer et al., 1995 ; rps16, Andersson and Rova, 1999 , rbcL and ndhF, Bremer et al., 1999 ; trnL-F, Rova et al., 2002 ). In addition, it has been demonstrated that fleshiness of the mesocarp, placentation, and ovule number are variable within monophyletic groups at the tribal level and above (Bremer, 1992 ; Delprete, 1996 ; Bremer and Manen, 2000 ).

Historically, the tribes Catesbaeeae, Condamineeae, and Chiococceae have been variously circumscribed. Hooker (1873) divided the Rubiaceae into three series: Series A, species with many ovules per locule; Series B, species with two ovules per locule; and Series C, species with a single ovule per locule. Following these criteria, he placed the tribes Condamineeae and Catesbaeeae in Series A, the tribe Chiococceae in Series C, and further divided the Condamineeae on the basis of fruit types into the subtribes Condamineeae, Pinckneyinae, and Portlandiinae.

Verdcourt (1958) treated the tribe Condamineeae sensu Hooker (1873) as a subtribe of the Rondeletieae, without recognizing any further subdivisions, and placed the Rondeletieae with the Chiococceae and the Catesbaeeae, among other tribes, in the subfamily Cinchonoideae. Robbrecht (1988) moved the Chiococceae to the subfamily Antirrheoideae, the Condamineeae as defined by Hooker (1873) in the subfamily Cinchonoideae, and treated the Catesbaeeae as tribus incertae sedis (for a more complete discussion of the taxonomic history see Delprete, 1996 ).

Based on phylogenetic analyses using molecular and morphological data Bremer (1992) transferred the subtribe Portlandiinae, characterized by multiseeded capsules, into the tribe Chiococceae, characterized by two-seeded, drupaceous fruits. Subsequently, Delprete (1996) , using a morphology-based phylogeny, included the Portlandiinae in the Catesbaeeae, placed the Chiococceae as the sister tribe and created an informal Exostema group (including Badusa, Exostema, and Morierina). In the same work, he merged the two remaining subtribes of Hooker of the Condamineeae into the Rondeletieae. Phylogenetic studies in the Rubiaceae using rbcL (Bremer et al., 1995 ), rps16 (Andersson and Rova, 1999 ), and trnL-F (Rova et al., 2002 ) have corroborated some of the ideas presented by Delprete's (1996) tribal circumscription, but contradicted his circumscription of the tribes Catesbaeeae and Chiococceae, and instead revealed that the two groups are intermixed in a single monophyletic lineage.

However, in each of the molecular studies of Catesbaeeae and Chiococceae, the relationships among genera were not fully resolved and the generic sampling was incomplete. As a result, the monophyly of most genera included in these tribes has never been tested, and delimitations of several genera are still debated. Aiello (1979) , in her treatment of the Portlandia complex, segregated several genera from Portlandia, reducing it to a Jamaican endemic genus. The separation of Cubanola, Nernstia (= Cigarilla A. Aiello), Osa, and Thogsennia from Portlandia was based principally on fruit and seed characters, and has been supported by recent morphological studies (Ochoterena, 2000 ), but has not yet been tested with molecular characters. Molecular studies have been conducted on two genera in the tribes, Exostema s.l. (McDowell and Bremer, 1998 ; McDowell et al., 2003 ) and Erithalis (Negrón-Ortiz and Watson, 2002 ), but the relationships of these genera within the context of the entire tribe are unresolved.

Based on recent systematic studies, the Catesbaeeae-Chiococceae Complex (CCC) encompasses 28 genera and about 190 species (Table 1; Borhidi and Acuña, 1971 ; Jérémie and Hallé, 1976 ; Borhidi and Muñiz, 1975 ; Borhidi et al., 1977 ; Borhidi, 1980 , 2002 , 2003 ; Ridsdale, 1982 ; Aiello and Borhidi, 1986 ; Lorence, 1986 ; Villareal, 1987 ; Delprete and Nee, 1997 ; Huysmans et al., 1999 ).

View larger version (26K): [in this window] [in a new window]   Fig. 1. Distribution map of Catesbaeeae-Chiococceae complex (CCC) species. Circled areas indicate geographic regions with approximate numbers of species per region, B = The Bahamas, C = Cuba, CA = Central America, F = Florida, H = Hispaniola, J = Jamaica, M = Mexico, NC = New Caledonia, P = Puerto Rico, SA = South America, and WP = Western Pacific (except New Caledonia). Dashed line indicates the Andesite Line

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS LITERATURE CITED

 
Taxon sampling
Four outgroup taxa were selected from the subfamily Cinchonoideae (Strumpfia, Guettarda, Chione, and Cinchona species). These taxa were selected based on previous results of Rova et al. (2002) . Among the ingroup taxa, 61 species are included (plus four duplicate accessions of four polymorphic taxa) from 24 of the 28 genera included in the CCC (See Data Supplement accompanying online version of this article). The four genera not included are Nernstia from Central America and Mexico and the Cuban endemics, Thogsennia, Ceuthocarpus, and Shaferocharis; all but the last, with three species, are monotypic. Additionally, the genera Molopanthera Turcz., Mastixiodendron Melch., Merrill & Perry, Placocarpa Hook.f., and Werhamia S. Moore which have been included in the CCC, but excluded from this group based on evidence from recent studies (Delprete, 1996 ; Delprete and Nee, 1997 ; Huysmans et al., 1999 ; Rova et al., 2002 ), were not sampled. Only species from which both ITS and trnL-F data were available were included in the analyses presented here. Most large genera are represented by several species; however, some genera that are species rich in Cuba (Phialanthus, Schmidtottia and Scolosanthus) are under-represented.

DNA extraction
Leaf samples were collected either in silica gel or from herbarium sheets. Genomic DNA was extracted from approximately 1 cm² of dried leaf tissue using a modified CTAB methodology. Leaf material was ground in a lysing matrix "A" tube (Qbiogene, Carlsbad, California, USA) and pulverized for 15 s in a Fastprep machine FP-120 (Qbiogene) bead mill at speed 5. Subsequently, 500 µL of Carlson Lysis Buffer (2 g CTAB, 8.18 g NaCl, 0.745 g EDTA, 10 mL 1 mol/L Tris/HCl pH 7.0, nanopure water to 100 mL, verified to pH 9.5, autoclaved, with 1 g PEG 4000 added when cool) and 75 µL of ß-mercaptoethanol were added to each tube and incubated at 74°C with occasional shaking for 60–90 min. Following incubation, 575 µL of SEVAG (24 : 1 chloroform : isoamyl alcohol) were added to all tubes which were placed on a tipping board for 30 min at room temperature. The tubes were then centrifuged at 14 000 rpm for 1 min, and 350 µL of supernatant were removed and added to new tubes containing 1050 µL of NaI solution, 20 µL Glassmilk, and 4 µL TBE modifier (Qbiogene). The tubes were placed on a tipping board for 30 min at room temperature. Afterwards, the tubes were centrifuged at 14 000 rpm for 1 min, and all of the supernatant was discarded. Next, each Glassmilk pellet was washed three times with 800 µL and once with 150 µL of ice cold New Wash solution (Qbiogene). After the final wash, all of the New Wash was aspirated from the Glassmilk pellet and 50 µL of 10 mmol/L Tris-Cl (pH 8.5) elution buffer were added to resuspend the DNA. The tubes were incubated at 55°C for 10 min and then centrifuged for 1 min at 14 000 rpm. The supernatant containing the DNA was removed and transferred to new tubes and stored at –20°C.

DNA amplification
DNA was amplified using the polymerase chain reaction (PCR; Mullis and Faloona, 1987 ). PCR reactions were performed in a 25 µL mixture consisting of 2.5 µL 10x buffer with MgCl₂ (Perkin Elmer, Foster City, California, USA), 9.3 µL autoclaved water, 2.5 µL BSA (bovine serum albumin), 2.5 µL dNTP, 1 µL each of two 20 µmol/L primers, 5 µL betaine, 0.2 µL Taq polymerase (Qiagen Valencia, California, USA), and 1 µL of genomic DNA. All PCR and cycle sequencing reactions were run on a Gene Amp PCR system 9600 (Applied Biosystems, Foster City, California, USA). Amplification of trnL-F region utilized external primers "c" (5'-CGAAATCGGTAGACGCTACG-3') and "f" (5'-ATTTGAACTGGTGACACGAG-3') and the internal primers "e" (5'-GGTTCAACTCCCTCTATCCC-3') and "d" (5'-GGGGATAGAGGGACTTGAAC-3') (Taberlet et al., 1991 ). The PCR conditions for amplification of the trnL-F region were: 1 cycle 94°C for 3 min; 32 cycles of 94°C for 45 s, 52°C for 30 s, 72°C for 1 min 30 s; and 1 cycle 74°C for 7 min, hold 4°C. The ITS region was amplified using forward (5'-CCTTATCATTAAGAGGAAGGAG-3') and reverse (5'-TATGCTTAAAYTCAGCGGGT-3') primers and when necessary two additional internal primers were employed, (5'-GCTACGTTCTTCATCGATGC-3') and (5'-GCATCGATGAAGAACGTAGC-3') (modified from White et al., 1990 ; Baldwin, 1992 ). The PCR conditions for amplification of the ITS region were: 1 cycle 97°C for 50 s; 30 cycles of 97°C for 50 s, 53°C for 50 s, 72°C for 1 min 50 s; and 1 cycle 72°C for 7 min, hold 4°C.

DNA sequencing
To detect successfully amplified products and the possible contamination of negative controls, PCR products were examined on 1% agarose gels stained with ethidium bromide and visualized under ultraviolet light. Amplified products were purified with spin columns from the QIAquick PCR purification kit (Qiagen) following protocols provided by the manufacturer. Purified products were cycle sequenced with dye terminator ABI Prism Ready reaction mix (Applied Biosystems) using dRhodamine or Big Dye v1.0 (1/8 reaction) and 5% dimethyl sulfoxide. Cycle sequencing conditions were: 1 cycle 95°C for 1 min; 32 cycles of 96°C for 10 s, 50°C for 5 s, 60°C for 3 min; and hold 4°C. Products were purified via gel filtration over Sephadex G-50 (Amersham Pharmacia Biotech, Piscataway, New Jersey, USA) and dehydrated in a Speed Vac (Savant Speed Vac Systems, Albertville, Minnesota, USA). The DNA was resuspended in 2.2 µL of formamide (83.5%) and EDTA blue-dextran loading dye (16.5%), heated at 95°C for 2 min and immediately placed on ice. Sequencing products were separated on 5% denaturing polyacrylamide gels on an ABI Prism 377XL DNA sequencer (Applied Biosystems).

Sequence alignment
Sequences were edited and aligned in Sequencher version 3.1.2 (Gene Codes, Ann Arbor, Michigan, USA) followed by manual refinement. Indels were treated as missing data. Additionally, in the trnL-F data set (and the trnL-F data in the combined analysis) indels of equal length that occurred in more than one sequence were considered as homologous and scored as separate binary characters added to the data matrix (Simmons and Ochoterena, 2000 ). Indels in the aligned ITS data for ingroup taxa which occurred in more than one sequence were 1 bp (base pair)(or 2 bp in a single case) in length and were not coded as characters, because these motifs are prone to sequencing, and reading errors (Goldenberg et al., 1993 ; Oxelman et al., 1997 ; Andersson and Rova, 1999 ).

Phylogenetic analysis
The alignment was analyzed using PAUP* 4.0b10 (Swofford, 2000 ). Minimal length trees were generated using a heuristic search, with 1000 random addition sequence replicates, with Tree-bisection-reconnection (TBR) branch swapping, and multiple parsimonious trees option (MULPARS) in effect. Uninformative characters were included in analyses except, as noted, for the calculation of alternative tree statistics. Tree statistics included the consistency index (CI; Kluge and Farris, 1969 ) and retention index (RI; Farris, 1989 ). Relative internal branch support was estimated with bootstrap analysis (Felsenstein, 1985 ) with 1000 replicates with TBR branch swapping and simple taxon addition. Data sets (ITS and trnL-F) were analyzed independently and combined using a total evidence approach (Kluge, 1989 ; Chippindale and Wiens, 1994 ; Nixon and Carpenter, 1996 ). Bootstrap percentages are described as high (85–100%), moderate (75–84%), and low (50–74%). Furthermore, Branch Support analysis using AutoDecay version 4.0 (Eriksson, 1998 ) based on comparing suboptimal trees with minimum-length ones (Bremer, 1994 ) were used to provide an additional measure of branch support.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS LITERATURE CITED

 
The trnL-F data set had an aligned length of 1061 nucleotides, of which 105 were parsimony informative. Ten parsimony informative indels were scored as additional characters; three were present in multiple genera, four were shared by species within a single genus, and three were present in multiple accessions of a single species. Most indels ranged between 5 and 10 bps in length, except for a single 200-bp deletion that was present in both Exostema acuminatum accessions. Additionally, a poly-T microsatellite varying from 1 to16 bp in length (beginning at bp 945 character position) was excluded from the analysis. Analysis of the trnL-F data resulted in 59 049 most parsimonious trees (MPT) of 298 steps in length, a CI = 0.876, and a RI = 0.934 (Fig. 2). Analysis of the trnL-F data without indels produced >100 000 MPTs 10 steps shorter with a very similar topology. Differences in the bootstrap consensus tree without the indel data from the tree consensus tree with indel data were that Coutarea was unresolved, and that the Exostema caribaeum-Solenandra mexicana clade and the polytomy supporting the Erithalis and Chiococca Clades lacked support. Furthermore, the Exostema acuminatum clade was weakly supported (50% bootstrap) in a polytomy with the Portlandia-Catesbaea clade and a Bikkia-Chiococca clade was resolved, a clade not retrieved in the trnL-F indel analysis(results not shown). The ITS data set had an aligned length of 692 nucleotides, of which 216 were parsimony informative. Fifty-nine indels were present in the data set, 25 (ranging from 1to 14 bps in length) were restricted to outgroup taxa. Of the remaining 34 (ranging from 1 to 4 bps in length) 19 were parsimony uninformative. The informative indels were, except for one 2 bp indel, only 1 bp in length. Analysis of the ITS data resulted in 64 515 MPT of 957 steps in length, a CI = 0.518, and RI = 0.740 (Fig. 2).

View larger version (77K): [in this window] [in a new window]   Fig. 2. The strict consensus trees from the two independent analyses. The tree on the left is the strict consensus of 59 049 most parsimonious trees (length = 298 steps, CI = 0.876, RI = 0.934) obtained from the trnL-F data set. The tree on the right is the strict consensus of in 64 515 most parsimonious trees (length = 957 steps, CI = 0.518, RI = 0.740) obtained from the ITS data set. Numbers above the branches are bootstrap values, numbers in parentheses are decay values, and arrows indicate the areas of incongruency in results from the chloroplast and nuclear data sets

View larger version (65K): [in this window] [in a new window]   Fig. 3. The strict consensus tree of 94 742 most parsimonious trees (length = 1278 steps, CI = 0.591, RI = 0.7795) from the combined (trnL-F and ITS) data set. Numbers above branches are bootstrap values and numbers in parentheses are decay values. The members of the monophyletic Catesbaeeae-Chiococceae complex (CCC) is indicated by the rightmost bracket and the two Pacific clades are indicated by the smaller brackets. The symbols prior to the terminal names indicate fruit types (+ = capsular; 0 = drupaceous; x = baccate) and flower types (C = Chiococca type; E = Exostema type, P = Portlandia type). Letters in parentheses following terminal names indicate distribution ranges of the species: B = The Bahamas, C = Cuba, CA = Central America, CAR = widespread Caribbean, F = Florida, H = Hispaniola, J = Jamaica, M = Mexico, NC = New Caledonia, NG = New Guinea, P = Palau, PR = Puerto Rico, SA = South America, SI = Solomon Islands, V = Vanuatu, and WP = Western Pacific (except New Caledonia). Letters below the branches indicate subclades, the uppercase letters designate three of the five subclades in the CCC and the lowercase letters designate the four subclades within subclade C

The species of Exostema sect. Exostema sensu McDowell sampled in this study (E. acuminatum, E. caribaeum, E. nitens, and E. spinosa) are, with the exception of the E. nitens-E. spinosa clade, mostly unresolved. The remaining Exostema species form a weakly supported clade (Fig. 3, clade A). Within this clade are two subclades: one clade contains Exostema species placed in section Pitonia and the other clade contains species of section Brachyantha (McDowell, 1996 ; McDowell and Bremer, 1998 ) or more recently the genus Solenandra sensu Borhidi (2002) with the exception of S. selleana. The two clades have bootstrap support of 100 and 71%, respectively. Exostema selleanum was placed by McDowell in section Brachyantha (McDowell, 1996 ; McDowell and Bremer, 1998 ) and included by Borhidi (2002) within Solenandra, but in our combined analysis Solenandra selleana (= E. selleanum) is nested within Exostema section Pitonia. This species was included in the first two phylogenetic studies of Exostema (McDowell, 1996 ; McDowell and Bremer, 1998 ), but was not included in the more recent study that included broader sampling and provided better resolution (McDowell et al., 2003 ). This unexpected result, the placement of S. selleana within the Exostema section Pitonia clade, has been checked with laboratory notes and the voucher specimen is consistent with the type material of S. selleana.

The Portlandia-Catesbaea clade (Fig. 3, clade B), in general contains species with large, campanulate flowers (except for the 12 small-flowered species of Catesbaea). Exostema acuminatum, which is weakly supported (59%) as sister to the other members of the clade, is also an exception to this trend. Cubanola domingensis is sister to Phyllacanthus grisebachianus, which is sister to a trichotomy of clades each containing species in the genera Osa, Catesbaea, and Portlandia-Isidorea. The Osa clade, like the genus, is monotypic. The Catesbaea clade has 99% bootstrap support and is also a unresolved trichotomy. Two clades are represented by C. parviflora and C. spinosa, and the third clade contains three species with C. fuertesii being sister to C. glabra and C. holacantha. Portlandia and Isidorea are strongly supported (90% bootstrap support) as sister taxa and each genus is strongly supported as monophyletic, confirming the results presented by Delprete and Motley (2003) .

The Bikkia-Chiococca clade (Fig. 3, clade C) is a well-supported clade (100% bootstrap support) composed of four subclades in an unresolved polytomy: the coastal Bikkia subclade (subclade a), the Cuban subclade (subclade b), the New Caledonia Bikkia subclade (subclade c), and the Chiococca subclade (subclade d). Bikkia, like Exostema, is polyphyletic. The coastal Bikkia clade has high bootstrap support (88%) and includes Badusa palauensis, which is sister to Bikkia palauensis, Bi. pancheri, and the widespread species Bi. tetandra. Typically, these Bikkia species are from coastal habitats and have similar morphologies (solitary flowers with white funnel-shaped corollas). The New Caledonia Bikkia clade contains Siemensia pendula (Cuban endemic) as a weakly supported (53% bootstrap) sister to an unresolved clade of Bikkia species endemic to New Caledonia and often occurring on ultramafic soils and Morierina, a monotypic genus, endemic to New Caledonia. The clade containing the endemic New Caledonian Bikkia species and Morierina clade is strongly supported (100%). The Cuban subclade (81% support) consists of the genera Schmidtottia, Ceratopyxis, and Phialanthus. Schmidtottia is sister to the latter two genera.

The Chiococca subclade (subclade d) contains most of the indehiscent-fruited CCC genera apart from Catesbaea and Phialanthus. It comprises two groups—a paraphyletic Chiococca lineage, which includes Asemnantha, and an unresolved Erithalis-Salzmannia-Scolosanthus lineage. The monophyly of the subclade containing Chiococca is weakly supported (50% bootstrap) due to the tenuous grouping of C. pubescens with the other species of Chiococca; however, the inclusion of Asemnantha inside of the Chiococca clade is strongly supported (100%). The relationship among the genera of the Erithalis-Salzmannia-Scolosanthus subclade is unresolved (due to incongruence between the two data sets); however, both Erithalis (96%) and Scolosanthus (100%) are strongly supported as monophyletic genera. In the ITS analyses Salzmannia is weakly supported (52% bootstrap) as sister to Erithalis, but in the trnL-F data analysis Salzmannia is placed as sister (75% bootstrap support) to Scolosanthus.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS LITERATURE CITED

 
Systematic relationships
Tribal phylogeny of the Catesbaeeae-Chiococceae complex
The phylogenetic relationships among the genera of the Catesbaeeae and Chiococceae tribes and the identification of the two tribes as a single monophyletic assemblage (Rova, 1999 ; Rova et al., 2002 ) was further substantiated by the more rigorous sampling of the ingroup taxa and combined analysis of sequence data from both the nrDNA ITS and cpDNA trnL-F regions used in this study. No previous taxonomic treatments of the tribes have monophyletic circumscriptions. However, Bremer et al. (1995) came the closest to defining the genera in the lineage when they suggested a modification of Chiococceae sensu Bremer (1992) to possibly include the members of the Catesbaeeae.

Historically, Hooker (1873) first circumscribed the Chiococceae as a tribe of 11 genera. Allenanthus Standley, Chione D C., and Placocarpa, which he included in the tribe, have all been shown in numerous later studies (Bremer, 1992 ; Bremer et al., 1995 ; Delprete, 1996 ; Huysmans et al., 1999 ; Rova et al., 2002 ) to not be closely related to the other genera in the lineage. They are morphologically easily distinguished by a combination of floral and pollen characters. The remaining eight genera included by Hooker (Asemnantha, Ceratopyxis, Chiococca, Erithalis, Phialanthus, Salzmannia, and Scolosanthus, excluding the unsampled Shaferocharis) are paraphyletic. In our study these genera form two (subclades b and d) of the four subclades producing a polytomy in the Bikkia-Chiococca clade (clade C). Bremer (1992) circumscribed the tribe Chiococceae as a much broader unit. She included the genera of the subtribe Portlandiinae (Bikkia, Ceuthocarpus, Coutaportla, Coutarea, Cubanola, Isidorea, Nernstia, Osa, Portlandia, Schmidtottia, Siemensia, and Thogsennia), all formerly placed in the Condamineeae by Robbrecht (1988) . Additionally, her circumscription contained Hintonia, the Exostema group (including Badusa and Morierina), and the eight genera included by Hooker, (listed above), all of which have basal stamen attachment. However, Phialanthus was excluded from this circumscription because the anthers are ovate rather than linear and the stamens, while having basal attachment, are free rather than fused into a basal ring (Bremer, 1992 ). Furthermore, members of the Catesbaeeae and Chiococceae (Catesbaea, Phyllacanthus, and Shaferocharis) were not included in Bremer's original (1992) circumscription, but in subsequent studies (Bremer and Struwe, 1992 ; Bremer et al., 1995 ) the latter strongly suggesting the inclusion of the Catesbaeeae or as sister to the Chiococceae. Later, Delprete (1996) , based on a phylogenetic analysis using morphological data, emended the Catesbaeeae sensu Hooker to include the Portlandiinae. He also recognized the Chiococceae sensu Hooker and an informal Exostema group (Badusa, Exostema, and Morierina), which, based on morphological analyses, were not included in either of the two monophyletic tribes. The trnL-F and ITS data independently and combined support the hypothesis that the Chiococceae sensu Bremer, the Catesbaeeae sensu Hooker, and Phialanthus combined are a monophyletic lineage. Only a combination of two morphological characters define the CCC, anther attachment at base of corolla tube and spinulose pollen, but neither one is unique to this group. It is perhaps the lack of a single synapomorphic character that hindered the previous tribal treatments from defining a monophyletic, generic circumscription of the lineage.

Phylogenetic position of Strumpfia
Strumpfia has long been a genus with uncertain affinities in the Rubiaceae (Hooker, 1873 ; Schumann, 1891 ; Bremekamp, 1966 ; Bridson and Robbrecht, 1985 ; Robbrecht, 1988 ; Igersheim, 1993 ). Igersheim (1993) , in a detailed morphological and anatomical study of the genus, noted that the characters of Strumpfia do not fit well into any of Robbrecht's (1988) subfamilies or tribes and identified the characters in Strumpfia that overlap with the characters defining the groups in Robbrecht's hierarchical classification. Although a monotypic tribe was suggested, Igersheim (1993) felt the formal recognition of a tribe was premature and would be "too easy a solution" to the problem. However, molecular studies (Rova, 1999 ; Rova et al., 2002 ) have shown that Strumpfia belongs to the subfamily Cinchonoideae and based on trnL-F and rps16 data is sister to the CCC (Rova, 1999 ; Rova et al., 2002 ). In there work the question of whether it should be included as a member of the CCC was still unsettled. In all analyses, both independent and combined, our data support the work of Rova (1999) and Rova et al. (2002) , in that Strumpfia is sister to the CCC. The combination of morphological characteristics of the androecium, plurilocular pyrenes, and verrucose pollen, present in Strumpfia, further differentiate it from the CCC and other genera of the Cinchonoideae (Igersheim, 1993 ). Based on these apomorphies we opt not to treat it as a member of the CCC but instead as a monotypic tribe.

The Exostema complex
In a recent study McDowell et al. (2003) showed that Exostema is a paraphyletic genus with respect to Chiococca, Coutarea, and Erithalis, contrary to the monophyly accepted in earlier investigations (McDowell, 1996 ; McDowell and Bremer, 1998 ). Exostema is here shown to be a polyphyletic group; however, a large paraphyletic lineage cannot be completely ruled out. In general, three of the four Exostema clades retrieved by McDowell et al. (2003) correspond to the clades retrieved in the combined analysis (except for the collapse of the E. caribaeum branch from the E. spinosum-E. nitens clade, and the fact that the South American taxa, which formed an independent clade in the McDowell study, but were not included here). Borhidi (2002) transferred the Caribbean taxa of section Brachyantha to the genus Solenandra prior to the study of McDowell et al. (2003) , who were apparently unaware of the taxonomic change.

In our study, section Exostema sensu McDowell (1996) , represented by E. caribaeum, E. nitens, and E. spinosum, was monophyletic in the ITS analysis (and weakly supported as sister to E. mexicanum in the trnL-F data set), but formed two groups in the combined analyses due to the collapse of the branch resolving E. caribaeum as sister to the other two species. Section Exostema was circumscribed by McDowell (1996) as having axillary inflorescences with one or a few flowers.

The other two Exostema sections sensu McDowell, section Brachyantha (Solenandra ixoroides, S. mexicana, S. parviflora, and S. selleana) and section Pitonia (E. ellipticum, E. lineatum, and E. longiflorum), both have terminal inflorescences with many flowers. They differ from each other by the presence of smaller flowers, with diurnal floral fragrance and basipetally arranged seeds in the former section, and the presence of larger flowers with nocturnal floral fragrance and acropetally arranged seeds in the latter section. In our combined analyses, these two sections formed sister clades in a monophyletic lineage, as was the case in the recent phylogeny of McDowell et al. (2003) . However, in our study, Solenandra selleana was nested within the Pitonia clade of Exostema with 100% bootstrap support. Unfortunately, this taxon was not included in the most recent study of McDowell et al. (2003) , and, upon re-examining the voucher specimen and sequence data, our placement of this species within the section Pitonia is confirmed and indicates the genus Solenandra is not monophyletic. Furthermore, in the McDowell et al. (2003) study, the South American Exostema species and Coutarea form an unresolved sister clade to this lineage. Our data set does not include samples of the South American Exostema species; however, in the trnL-F analyses, Coutarea is sister to a clade containing members of sect. Pitonia and S. selleana, a relationship not supported in either the ITS or combined analyses.

The last Exostema clade contains E. acuminatum, which is weakly supported as sister to the Portlandia-Catesbaea clade in the combined analyses. This species, along with E. salicifolium, formerly placed in section Exostema, was also resolved by McDowell et al. (2003) as a separate clade, but there are few or no morphological characters separating these taxa from the rest of the section (McDowell et al., 2003 ). Our examinations of specimens of E. acuminatum and E. caribaeum (both in Exostema section Exostema, sensu McDowell), which sometimes occur sympatrically, could only discern two morphological differences among the species. Exostema acuminatum flowers are reddish-pink at later stages of anthesis (not yellow), and the corolla is more shallowly lobed (less than half corolla length). An additional difference was that the two E. acuminatum accessions both had a 200-bp deletion in the trnL-F sequences that was absent from all other Exostema species. While it is tempting to elevate all three sections of Exostema and the E. acuminatum clade to generic level, we prefer to refrain from doing so until further morphological and molecular evidence is available and all species are represented in the study.

The Pacific genera
Bikkia is a genus of approximately 20 species (Darwin, 1985 ). Eleven species occur in New Caledonia, 10 of which are endemic to the main island (Jérémie and Hallé, 1976 ). The remaining species are distributed from New Guinea, Philippines, the Moluccas, Micronesia, Fiji, Tonga, and Niue to the Wallis Islands (Airy Shaw, 1973 ). Our analyses show that Bikkia, as presently treated, is a polyphyletic genus. One clade consists of species endemic to New Caledonia with colorful, campanulate corollas and the monotypic genus Morierina, which is also endemic to New Caledonia (Vieillard, 1865 ; Brongnart and Gris, 1871 ). The other Bikkia species form the coastal Bikkia clade, which is sister to Badusa palauensis, the only representative sampled of the Pacific genus Badusa (a widespread genus of three species; Ridsdale, 1982 ; Soejarto et al., 1996 ). The Bikkia species in this clade are typically coastal species with white, funnel-shaped corollas. The species of this group included in these analyses include: B. tetandra, a widespread species throughout the Mariana Islands, Fiji, and Western Polynesia, which is the type species of the genus; B. pancheri from the Isle of Pines, New Caledonia (which also occurs in the New Hebrides, Solomon Islands, and New Britain); and Bikkia palauensis.

The New Caledonia endemic species of Bikkia are typically found in the inland forests and have been treated in the past as part of separate genera (Cormigonus Rafin. nom. nud., Thiollierea Montr., and Grisia Brongn.; Jérémie and Hallé, 1976 ), so the formal segregation of the two Bikkia clades is not unprecedented. However, the placement of Morierina within the endemic New Caledonia Bikkia clade is surprising. Morierina differs from Bikkia in having narrow, tubular flowers with exserted anthers and nonreduplicate corollas, typical in Exostema type flowers, rather than the Portlandia type of Bikkia (Vieillard, 1865 ; Delprete, 1996 ). Additionally, Morierina montana is a large tree found in densely forested areas (T. Motley, personal observations), whereas, the Bikkia species in New Caledonia are shrubs on ultrabasic soils. This relationship could represent a case of morphological diversification, in which case Morierina should be a member of the New Caledonian Bikkia species. This hypothesis is suggested by the congruent placement (although poorly supported) in the individual analyses (Fig. 2) where Morierina is nested inside of the Bikkia clade. An alternate hypothesis is that Morierina is simply sister to the New Caledonian Bikkia species, a case that gains support from reduced resolution in this clade in the combined analyses (Fig. 3). In either case, Morierina represents a morphological change, perhaps driven by a shift to a different ecological niche and/or pollination syndrome. The sister relationship of Badusa to the other coastal Bikkia clade also suggests an adaptive pollinator shift as in the case of Morierina. Thus a shift in floral morphology is perhaps not an uncommon evolutionary event. Badusa flowers, although shorter in the length of the corolla tube, are similar in shape to Morierina (Exostema type) and the Bikkia species have the Portlandia type flowers (Ridsdale, 1982 ; Darwin, 1985 ; Delprete, 1996 ). Badusa is a genus of three species (Ridsdale, 1982 ; Soejarto et al., 1996 ) endemic to the South Pacific, and represented here by only one species, B. palauensis.

Chiococca, Asemnantha, and Salzmannia
Chiococca is a Neotropical genus of about 20 species, with the greatest species diversity in Mexico and Central America, but with a few species extending into the Caribbean and South America (Standley, 1934 ). The species are scandent shrubs or vines, with axillary inflorescences and campanulate to slightly urceolate corollas. They have bilocular ovaries with axial placentation, and a single, apically attached ovule per locule, and their fruits are drupaceous, with woody pyrenes (Standley, 1934 ). The most widespread and polymorphic species of the genus is Chiococca alba, a scandent shrub that can grow into a vine up to 15 m tall and ranges from southern North America (Florida Keys) to Argentina. Asemnantha is a monotypic genus of small shrubs endemic to southern Mexico and northern Central America, and differs from Chiococca by having the stamens fused at the base to form a ring, rather than being free (Standley, 1934 ). In both the separate and combined analyses Asemnantha is strongly supported as a member of the Chiococca clade (Fig. 3, clade C, subclade d).

Salzmannia is a monotypic genus of scandent shrubs and woody vines from the coastal forests of Brazil, traditionally associated with Chiococca because of their morphological similarities, although its distinctness has been questioned (Schumann, 1889 , 1891 ). Morphologically Salzmannia differs from Chiococca and Asemnantha only by the presence of glabrous filaments, corolla lobes not recurved, and persistent, leaf-like bracts subtending the inflorescence (Hooker, 1873 ; Schumann, 1891 ). In this study, Salzmannia was found to be more closely related to genera principally from the Greater and Lesser Antilles (i.e., Scolosanthus and Erithalis, the latter also occurring in Florida) than to Asemnantha with which it shares many morphological synapomorphies (Bremer, 1992 ). Salzmannia was weakly supported (52%) as sister to Erithalis in the ITS tree and more strongly supported sister to Scolosanthus (75%) in the trnL-F tree. These three genera form a trichotomy (74%) in the combined analysis, which supports the recognition of Salzmannia as a monotypic genus.

Erithalis and Scolosanthus
Erithalis is a genus of 8–10 species distinguished by ovaries with 3–5 locules, rather than 2, as in the rest of the CCC. The results of this study correspond to those of Negrón-Ortiz and Watson (2002) , which supported the monophyly of this genus. Scolosanthus a genus of 20 species (Liogier, 1962 , 1995 ) represented by four species in this study, is well supported (100% bootstrap) as a monophyletic genus.

Catesbaea and Phyllacanthus
Catesbaea is a genus of about 16 species distributed throughout the Greater Antilles, with one species in the Bahamas and southern Florida, and one species in the Lesser Antilles (Liogier, 1995 ). Our study sampled five species, which formed a strongly supported clade (99%), confirming the monophyly of the genus. This genus has the most variable corollas in the CCC, both in size and in shape. For example, of the species sampled, C. fuertesii, C. glabra, C. holacantha, and C. parviflora have small, campanulate corollas (less than 2 cm long), whereas C. spinosa (the type species of the genus) has large, funnel-shaped corollas 4–17 cm long (Liogier, 1995 ; Delprete, 1996 ). Although the basal lineage within Catesbaea is not resolved, comparison with sister groups suggests that the campanulate corollas are derived from funnel-shaped corollas within Catesbaea.

Phyllacanthus grisebachianus is a species endemic to Cuba (Liogier, 1962 ), is probably now extinct due to habitat loss caused by sugar cane cultivation (Oviedo et al., 1988 ). Phyllacanthus shares many morphological similarities with Catesbaea, but it can be distinguished from Catesbaea by having large, flattened, triangular thorns and carpels with uniseriate ovules (vs. terete, needle-like thorns and multiseriate ovules; Liogier, 1962 ; Delprete, 1996 ). Rova (1999) succeeded in extracting DNA from this species by grinding one thorn from a specimen more than 50 yr old, which is the most recent of the two known collections ever made of this taxon. In our trnL-F analysis, Phyllacanthus nested within the Catesbaea clade (a result also recovered by Rova et al., 2002 ); however, this relationship collapsed in the ITS and combined analyses. Because the Phyllacanthus sequences are incomplete due to degraded DNA, its relationship to Catesbaea is still uncertain. Morphologically, Phyllacanthus flowers are nearly identical in size, shape, and color to those of C. flaviflora Urb. Furthermore, based on the trnL-F data (analyses both here and in Rova et al., 2002 ) and the morphological similarities with Catesbaea we suspect that it is best to include Phyllacanthus in Catesbaea, and return to its original binomial, Catesbaea phyllacantha Hook. f. (Hooker, 1871 ).

Portlandia, and the related genera Isidorea, Osa, and Cubanola
Aiello (1979) circumscribed Portlandia as a genus of five species endemic to Jamaica with large, funnel-shaped corollas, imbricate aestivation, basal insertion of stamen, basal fusion of filaments, anthers basifixed and linear, and seeds horizontally arranged and wingless. In a recent revision (Delprete and Motley, 2003 ), Portlandia was shown to be a monophyletic genus of seven Jamaican species. Isidorea, which is also shown here to be monophyletic, is sister to Portlandia. These genera share the morphological features listed above, and the only difference between the two is the presence of pungent apices on the leaves and stipules of Isidorea (Aiello, 1979 ). While all the recognized species of Portlandia were sampled in this study, only four species of Isidorea from the Dominican Republic were sampled (the Cuban species were unsampled); however, the two genera are strongly supported as sister monophyletic lineages in the combined analyses.

Cubanola is a genus of two species, one endemic to Hispaniola and the other to Cuba (Aiello, 1979 ). In the combined and trnL-F analyses (but unresolved in the ITS analyses) Cubanola is sister to a clade containing Catesbaea, Phyllacanthus, Portlandia, Isidorea, and Osa (a monotypic genus endemic to the Peninsula de Osa, Costa Rica). The four large-flowered genera, with corollas of some species exceeding 25 cm in length, along with Catesbaea, Phyllacanthus, and perhaps Exostema acuminatum, represent a monophyletic group of closely related genera. The internal relationships within this clade for the most part support the morphological work of Aiello (1979) , who proposed their generic segregation.

Relationships of additional Cuban genera
Among the other genera of the CCC not previously discussed, four are principally Cuban or are endemic to Cuba. Schmidtottia (16 species endemic to serpentine soils of eastern Cuba), Ceratopyxis (a single species occurring in the limestone haystack mountains of eastern Cuba), and Phialanthus (17 of 20 species occurring in Cuba) (Liogier, 1962 ) form a well-supported clade (81%) in the combined analyses. The species in this clade encompass a diverse assemblage floral and fruit morphologies. Schmidtottia has capsular fruits and medium-sized, narrowly campanulate to narrowly infundibuliform corollas (Portlandia type). Ceratopyxis also has capsular fruits and small, narrowly funnel-shaped flowers. In contrast, the species of Phialanthus have drupaceous fruits and minute, narrowly campanulate corollas (Chiococca type) (Liogier, 1962 ). In our study, Schmidtottia was sister to Ceratopyxis and two accessions of Phialanthus. Although sampling within the large genera of this clade was limited, our analyses indicate an independent derivation of fleshy fruits from capsules in this Cuban lineage.

Siemensia is a monotypic genus endemic to the limestone haystack mountains of the Province of Pinar del Rio, western Cuba (Liogier, 1962 ). In the ITS phylogeny, Siemensia was placed as sister to the New Caledonian Bikkia clade, whereas in the trnL-F analysis it was placed in a polytomy with the coastal Bikkia clade and the Chiococca clade. In the combined analysis Siemensia was sister to the New Caledonian Bikkia clade. Although it was poorly supported (53%), this evidence suggests a possible close relationship of this Cuban endemic species to the Pacific taxa.

Relationships of genera from Mexico and Central and South America
Hintonia is a genus of three species (Ochoterena, 2000 ) endemic to Mexico and northern Central America. Coutarea is a genus of two species—C. hexandra, ranging from Mexico to Argentina, and C. andrei Standl., endemic to the Loja Province of Southern Ecuador (Delprete, 1999 ). These two genera are isolated in our strict consensus trees. Coutaportla is a genus of three or four species from Mexico and Central America (Lorence, 1986 , 1999 ; Villarreal, 1987 ). Recently, Borhidi (2003) transferred C. guatemalensis (Standl.) Lorence to his new monotypic genus Lorencea (not sampled). Because only one species of Coutaportla was sampled in this study, we cannot discuss this segregation of Lorencea based on these data.

Fruit and flower evolution
The second objective of this study was to examine flower and fruit evolution in the CCC using a phylogenetic framework. Flower characters uniting the CCC are stamen attachment at the base of the corolla (Bremer, 1992 ; Delprete, 1996 ), pollen with perforate and microechinate tectum, and a smooth, often perforated layer under the inner nexine layer (Huysmans et al., 1999 ). Floral morphology varies within the CCC mainly in size, shape, and merosity. Delprete (1996) described three flower types (Table 1; Fig. 4): (1) the Portlandia type with long ([5–]10–27 cm), campanulate (Fig. 4A) to tubular-funnelform (Fig. 4B) and shallowly lobed corollas, linear anthers, and a long style with four-parted, adnate, elongate-clavate stigma; (2) the Chiococca type (Fig. 4C) with short (0.3–2 cm long), campanulate, and deeply lobed corollas, oblong anthers, and a short style and two-lobed stigmas; and (3) the Exostema type (Fig. 4D) with short or long (2–20 cm), narrowly tubular, reflexed-lobed corollas, linear, exserted anthers, and a long style with capitate or linear stigmas. Fruits vary greatly among the members of the CCC (Table 1). The mesocarp can be dry, leathery, or fleshy; placentation is axial or apical; dehiscence is loculicidal, septicidal or absent; and seeds are winged or unwinged, and flattened or globose. The genera of the CCC can exhibit almost any combination of these fruit characters.

View larger version (89K): [in this window] [in a new window]   Fig. 4. Flower types in the Catesbaeeae-Chiococceae complex (CCC). The Portlandia type (A–B) with either large (A) long, campanulate (Bikkia macrophylla) or (B) tubular-funnelform corollas (Portlandia platantha). (C) The Chiococca type (Erithalis harrisii) with short, campanulate, deeply lobed corollas. (D) The Exostema type with short or long, narrowly tubular, lobed corollas (Exostema caribaeum). Scale bars = 1 cm

Presuming the ancestral condition in the CCC included capsular fruits, as is the case in most of the subfamily and closely related taxa (Rova et al., 2002 ), drupaceous fruits seem to have evolved two separate times within the CCC (Fig. 3), once in Phialanthus and a second time in the Chiococca clade (including Chiococca, Asemnantha, Salzmannia, Erithalis, and Scolosanthus). The baccate fruit type has evolved at least one time and possibly twice in Catesbaea and Phyllacanthus. The genus Chiococca has one of the broadest geographical distributions in the CCC, and its spread probably due to endochorous dispersal by birds eating the small, drupaceous fruit. Scolosanthus, Erithalis, and Phialanthus, also have similar drupaceous fruits and Catesbaea, with baccate fruits, all tend to have widespread distributions that are facilitated by animal dispersal (Table 1). However, Coutarea hexandra, Exostema caribaeum, the coastal Bikkia species, and Badusa are also widespread and have capsular fruits and wind-dispersed seeds. Thus no clear trend is evident that would suggest one fruit type is more advantageous over another for dispersability.

The differing floral morphologies probably reflect adaptive shifts for pollinator specialization. Unfortunately, actual pollination data for the CCC are lacking for this group, although several species have been the subject of field observations. The Exostema type flowers (narrowly, tubular, lobed corollas; linear, exserted anthers; and a long style) are typical of moth or butterfly pollination; the Chiococca type flowers (short, campanulate, and deeply lobed corollas; oblong anthers; and a short style) are typical of entomophilous (particularly bee) pollination; and the Portlandia type flowers (long, campanulate to tubular-funnelform corollas; linear anthers; and a long style) are characteristic of bird- and bat-pollinated flowers (Faegri and Pijl, 1979 ; Delprete and Motley, 2003 ).

The Exostema type flowers seem to have evolved at least three times in the CCC. Because of the poor resolution among the clades of Exostema s.l. it is not possible to more precisely determine the number of events. However, among the Exostema clades the corolla color variation and differences in production of floral fragrances (diurnal vs. nocturnal) seem to reflect cladogenesis in these groups. This is especially evident between the sister clades corresponding to sections Brachyantha (= Solenandra) and Pitonia in McDowell's (1996) classification. Apart from the species of Exostema s.l., the Exostema type floral morphology seems to have evolved independently twice in the Pacific, once in Badusa and again in Morierina. Both genera are most closely related to species with Portlandia type flowers (Fig. 3).

The Chiococca type flower has apparently evolved three or four times (Catesbaea spp., Phyllacanthus, Phialanthus, and the Chiococca clade). These three events correspond to the same groups that have evolved fleshy, indehiscent fruits. This flower type is characteristic of mellitophily (Faegri and Pijl, 1979 ). Catesbaea is the genus with the most variable corollas in the complex, both in size and shape. Twelve species have Chiococca type flowers (<2 cm long, campanulate corollas), and four species have Portlandia type flowers (4–17 cm long, funnel-shaped corollas; Delprete, 1996 ). Because Catesbaea is within a clade (Fig. 3, clade B) of genera that principally have Portlandia type flowers, the change in flower type appears to be an adaptive shift from bird or bat pollination to entomophily. It will be interesting to determine in future analyses whether the two floral types in Catesbaea are resolved in sister clades in a manner similar to the species with floral differences seen in the Exostema clade (sections Brachyantha and Pitonia) described above. Because only a single large-flowered species was included in the present analysis, the effects of pollinator shifts in the evolution of the genus cannot be determined.

The Portlandia type flowers are widely distributed throughout the CCC phylogenetic tree and appear to have originated independently five or more times in the group. The corolla shape ranges from infundibuliform-salverform in Portlandia, Isidorea, Hintonia, Coutaportla, Schmidtottia, and the coastal Bikkia species, to campanulate in Osa, Coutarea, Siemensia, Cubanola, and four species of Catesbaea, and the New Caledonian Bikkia species. The differences in floral structure and color seen between the two Bikkia clades seems to reflect different pollination strategies. The coastal Bikkia species have white, upright, salverform flowers characteristic of flowers that attract bats and moths (Faegri and Pijl, 1979 ). The New Caledonian Bikkia species have brightly colored campanulate flowers (red, yellow, purple, and pink) like those of species adapted to ornithophily (Faegri and Pijl, 1979 ). The white-colored, campanulate flowers of Osa, Cubanola, Siemensia, and large-flowered Catesbaea species that have a long corolla tube, (9–12 cm) are most likely pollinated by long-tongued moths or possibly bats. The flowers of Coutarea hexandra vary from white to pink, purple, red, or yellow with a broad, obconical, and slightly asymmetrical floral tube that suggests an adaptation to a wide range of bird pollinators. Among the species in both Portlandia and Isidorea, there are also differences in floral changes in color, shape, and size, although the differences are not as pronounced as the variation in Catesbaea. Within Portlandia there are four red-flowered species and three white-flowered species (Delprete and Motley, 2003 ). Corolla length varies from 2.5–5.4 cm in P. proctorii to 10– 22 cm in P. grandiflora (Aiello, 1979 ). The species with the smaller red corollas, P. proctorii, were observed being visited by the red-billed streamertail (Trochilus polytomus polytomus) (Delprete and Motley, 2003 ). Anecdotal evidence from foresters and field observations indicate the salverform, white corolla flowers of P. grandiflora were always collected in the field with damage due to aggressive floral visitors. The flowers open in the evening and produce large amount sweet-smelling floral fragrance (T. Motley and P. Delprete, personal observations), which strongly suggest bat pollination. No pollination data are available for the other species. However, the red flower morphology seems to be the ancestral state in Portlandia. Portlandia coccinea is sister to P. microsepala, both red-flowered species, which are sister to an unresolved clade of both white- and red-flowered species (Fig. 3). Shifts in floral morphology and adaptation to various pollinators seem to have occurred often among and within the genera of the CCC.

In general, taxa with Chiococca type flowers have drupaceous fruits, with a single, pendulous, apical ovule per locule (Ceratopyxis is an exception to this trend in having woody, septicidal capsules), and those with Exostema or Portlandia type flowers have few to many axial ovules, and either woody, septicidal capsules or leathery, baccate fruits in Catesbaea. However, because many of the characters appear homoplasious, our preliminary compartmental classifications of fruit and fruit types are being re-evaluated thorough careful studies (H. Ochoterena, P. Delprete, and T. Motley, unpublished data), and, although our discussions on flower morphology and pollinators are based mostly on generalizations, we hope it will spur more detailed field studies in the future.

Biogeography
The third objective of this study was to better understand the origins of the biogeographic disjunction between the Caribbean and Pacific genera. The Greater Antilles is the most species-rich area and is likely the center of origin for the CCC (Fig. 1). However, alternate hypotheses (i.e., South American or Pacific origins) cannot be completely ruled out. Working under the assumption that the CCC is of American origin we attempt to determine the number of long-distance dispersal events that were involved in the evolution and establishment of the three Pacific genera (Bikkia, Badusa, and Morierina). We propose three biogeographic hypotheses, which we hope to test using the phylogeny of the CCC: (1) origin in the Greater Antilles, and prior to the formation of the Central American landbridge (ca. 60–80 my (million years ago); Briggs, 1994 ; Iturralde-Vinent and MacPhee, 1999 ), dispersal into the Pacific basin; (2) origin in the Greater Antilles, followed by a series of dispersals to South or Central America and subsequently to the Pacific; and (3) origin in Central or South America with dispersals to Caribbean and the Pacific. Unfortunately, due to the lack of resolution among the major clades of the lineages and the uncertain placement of several continental taxa, we are not able to determine the area of origin