Quick viewing(Text Mode)

Correlations Among Fruit Traits and Evolution of Different Fruits Within

Correlations Among Fruit Traits and Evolution of Different Fruits Within

BolankdJournul oJ/h Linnean Sociplv (20OO), I.?.?: 303-326. With G figures doi:l0.lOOS/lmjl. 1999.0340, available online at http://www.idealibrary.rom on I hi"

Correlations among hittraits and evolution of different hits within

GUDRUN CLAUSING'*, KARSTEN MEYER' AND SUSANNE S. RENNER' 'Institut$r Spezielh Botanik, Universitat Main<, 0-55099 Main< Germany, 'Department of Biology, Universip of Missouri-St. Louis, 8001 Natural Bridge Rd., St. Louis, M063121- 4499, us;4

RecciwdJu~1999; accepted for publicdwn Februaty 2000

The anatomy and morphology of nearly mature in 85 mainly palaeotropical species of Melastomataceae were examined using microtome- and hand-sectioning, and differential staining. Much structural heterogeneity was observed in both capsules and . Mul- tivariate analyses of 31 of the 52 characters recorded for each species, revealed that indehiscence is associated with fusion of and tissues, placenta persistence, lack of a persistent endocarp, and a dearth of srlereids in these tissues, while dehiscenre is correlated with the opposite states and a persistent exocarp. Other characters such as lignification or fleshiness of tissues do not show a consistent association with . Break down of broad fruit types, such as '' and '', into their individual morphological and anatomical traits shows how unusual fruit types, such as woody berries, fleshy capsules, and capsules containing fleshy placentas (display fruits), which are common in palaeotropical Melastomeae and Dissochaeteae, contribute to a loosening of expected correlations. Thus, discriminant analysis clearly differentiated display fruits from the other fruit types because of their combination of fleshy placentas with a persistent endocarp and absence of ovary/hypanthium fusion. The evolution of fruit types within Melastomataceae, and especially Dissochaeteae, and their reliability as phylogenetic indicators is discussed in the light of molecular phylogenies for these groups that show that berries and capsules evolved several times independently, explaining the observed heterogeneity of outwardly similar fruits. Fruit diversity within Melnrtoma, a monophyletic of 22 species, provides an example of the plasticity afforded by the particular construction of Melastomataceae fruits, which has contributed to ecological diversification in melastome dispersal. 0 2000 The Linnean Society of London

ADDITIONAL KEY WORDS:-berries ~ capsules ~ dispersal mechanisms - display fruits - Dissochaeteae - multivariate analysis - phylogeny.

CONTENTS

Introduction...... 304 Material and methods ...... 305 Results ...... 308 General anatomy and morphology of melastome fruits ...... 308

* Corresponding author. E-mail: [email protected] 303 OO2&4O74/OO/07O303 + 24 $35.00/0 0 2000 The IdinnranSociety of London 304 G. CLAUSING ETA.

Characters and their states ...... 310 Multivariate analyses ...... 312 Discussion ...... 314 Structural heterogeneity of berries and capsules ...... 314 Fruit evolution in Melastomataceae, especially in the Dissochaeteae/Sonerileae complex ...... 316 Fruit variation at the intrageneric level: the case of ..... 319 Fruit types as phylogenetic indicators ...... 320 Fruit character plasticity and dispersal ecology ...... 320 Acknowledgements ...... 321 References ...... 322

INTRODUCTION

Fruit types such as capsule, , , and berry have sometimes been regarded as evolutionarily conservative (Stebbins, 1974; Spjut, 1994); indeed, many older classification systems accorded fruit type a high taxonomic weight. However, molecular phylogenetic studies are showing that fruits often are unreliable indicators of natural alliances and that their use in higher-level classification has created artificial groups (: Plunkett, Soltis & Soltis (1996); Lamiales d.:Wagstaff & Olmstead (1997); : Johnson & Briggs (1 984); Ranunculaceae: Hoot (1995); Rhamnaceae: Richardson & Medan (pers. corn., Sept. 1997); : Rohrer, Robertson & Phipps (1991), Morgan, Soltis & Robertson (1994); Saxi- fragaceae: Soltis, Soltis & Bothel (1990)).In , for example, tracing of fruit evolution on a molecular phylogenetic suggests that fleshy fruits have evolved 12 times independently within that family (Bremer & Eriksson, 1992; Bremer, Andreasen & Olsson, 1995). Moreover, molecular data are showing that fruit evolution may be rapid when morphogenetic genes of major effect are involved that can cause dramatic morphological differences (cf. Kadereit , 1994). The striking homoplasy of fruit types may have two causes. First, it may result from strong directional selection on fruit characters; the latter are highly adaptive and directly correlated with the dispersal agent. For example, fleshy fruits normally are associated with dissemination by animals. Second, fruit characters may appear more homoplastic than they really are because of incorrect homology assessments. This is a problem particularly in large tropical families, such as Rubiaceae and Melastomataceae, where detailed morphological and anatomical studies based on appropriately preserved material are scarce (but see Igersheim, 1993; Rohwer, 1996). Here we investigate the morphological and anatomical plasticity of fruits in Melastomataceae. Melastomataceae are part of , which include families with both dry and fleshy fruits. Examples are Myrtaceae (capsules, nuts, , and berries), Lythraceae s.1. (capsules and berries), Onagraceae (capsules, nuts, berries), and Melastomataceae (capsules, berries, and several intermediate types). Johnson and Briggs ( 1984) concluded from their morphology-based phylogenetic study of Myrtaceae that the traditional three subfamilies delimited by succulent and indehiscent versus dry and dehiscent fruits had to be abandoned since fleshiness of the ovary, and fleshiness of both ovary and hypanthium, each arose independently at least twice. In Lythraceae, berries of Sonneratiu and rZlnicu evolved independently (Graham, Crisci & Hoch, 1993), and in Onagraceae, the nut-like fmits of Heterguuru originated from ancestors that had septicidal Clarh-type capsules (Sytsma & Gottlieb, 1986). FRUIT TRAITS IN MEIASTOMATACEAE 305

These examples from close relatives of the Melastomataceae (Conti, Litt & Sytsma, 1996) suggest that phylogenetic investigations, followed by critical morphological homology assessments, are needed if fruit characters are to be used in phylogenetic analyses of this family. Tribal classifications of Melastomataceae (De Candolle, 1828; Naudin, 1851; Triana, 1871; Cogniaux, 1891; Renner, 1993) have relied heavily on fruit types, both as diagnostic characters as well as supposed phylogenetic markers (Renner, 1993). Based on outgroup comparison, Renner considered dry capsules as plesiomorphic for the family and fleshy fruits from inferior ovaries with embedded in pulp (her ‘true berries’) as apomorphic. Because she was aware of the plasticity of fruit characters, especially the different origins of fleshy fruit walls and modes of dehiscence, she used only one fruit character in her analysis, namely the presence and absence of true berries, suggesting that further critical examination of fruit attributes was needed (Renner, 1993: 529). The present study examines the fruit anatomy and morphology of 85 species of Melastomataceae, representing the mainly palaeotropical Astronieae, Dissochaeteae, Kibessieae, Melastomeae, and Sonerileae, and a few neotropical Blakeeae, Me- lastomeae, Miconieae, and Bertolonieae (for tribal classification, see Material and methods). Our major goal was to investigate whether the terms capsule and berry as traditionally applied in these groups describe structurally homogeneous phenotypes. To this end we performed multivariate analyses of 31 anatomical and morphological characters to identifjr states that might correlate with fruit dehiscence (or a lack thereof). We also investigated the pattern of plasticity of fruit characters at different hierarchical levels and within monophyletic groups, as identified by molecular sequence data, and tried to relate it to ecological determinants.

MATERIAL AND METHODS

Fruits of 85 species of Melastomataceae, representing 3 1 genera and eight tribes were investigated (see Appendix in which we follow the classification of Renner (1 993); we recognize Bertolonieae and Dissochaeteae semu Cogniaux (189 1) based on molecular phylogenetic results (Clausing, 1999; Clausing & Renner, in press), although the circumscription of these tribes at present is unclear). Fruits came from alcohol samples (70 spp.) collected in South-east , Madagascar, and , from specimens (7 spp.), or from cultivated at the Botanical Garden Mainz (8 spp.). We analysed nearly mature fruits. Cross-sections of fruits of ten species were made using a microtome and then stained with toluodine blue, which stains cell walls; the remaining material was hand-sectioned. Herbarium material was boiled prior to sectioning. Manual sections were stained with phloroglucine+ HC1 or with iodine in potassium iodide solution, staining lignin and , respectively. They were preserved in Hoyer’s solution (Kearns & Inouye, 1993), which bleaches , bringing out lignified cells and mineral crystals. Fifty-two characters were coded in binary fashion (Table 1) and scored for 85 species. The original uncoded observations are available upon request from the first author. Multistate characters were split into multiple characters. Thus, the 3-state character ‘sclereid abundance’ was divided into ‘few sclereids @resent/other)’, ‘intermediate abundancy @resent/other)’, and ‘many sclereids (present/other)’. Definitions used in this and other problematic coding situations are given below 306 G. CLAUSINC ETAL.

TABLE1. Characters and their states; see text for definitions of problematic states. Of the 52 characters listed here atid scored for all species, the 31 shown in Tables 2-6 were included in the contingency analyses

Tissue Character State

Plarenta Persistenre ahscnt (0) present (I) Endorarp Rrsistrnce alxcnt (0) present (I) Lignification ahsent (0). present (I) Cell pattern mosaic-like (I) other (0) Ditto undulate (I) other (0) Dittn elongate (I) other (0) Ditto elongate-undulate (I) other (0) Mesorarp '~hickness thin (0) thick (I) Sclerrid presenre absent (0) present (1) Few sclereids present (I) other (0) Intermediate numher of sclereids prrsent (I) other (0) Many sclereids present (I) other (0) Sclercid distribution scattered (I) other (0) Ditto rlustered (I) other (0) Ditto in a ring (I) other (0) Calcium oxalate presencc ahsent (0) present (I) Stiirrh granule presence allsent (0) present (I) Hypanthium Thickness thin (0) thick (I) Srlereid presence ahsent (0) present (I) Few sclereids present (I) other (0) Intermediate number of srlereids present (I) other (0) Many srlereids presrnt (I) other (0) Srlereid distribution scattered (I) other (0) Ditto clustered (I) other (0) Ditto in a ring (I) other (0) Calcium oxalatr presence absent (0) present (I) Starch granule presence ahsent (0) present (1) Size of inner rells relative to rells larger (I) other (0) Ditto same size (I) other (0) Ditto smaller (1) other (0) Outer epidermis cell surfare smooth (I) other (0) Fusrd tissues Thickness thin (0) thirk (I) Sclerrid presence absent (0) present (1) Few srlereids present (I) other (0) Intermediate numher of sclrreids present (I) other (0) Many sclereids present (I) other (0) Sclereid distribution scattered (I) other (0) Ditto clustered (I) other (0) Diito in a ring (1) other (0) Calcium oxalate presence absent (0) present (I) Starch granule prescnce ahsent (0) present (I) Ovary and hypanthium free present (I) other (0) Extent of ovary/hypanthium fusion complete (I) other (0) Ditto partial (I) other (0) Exocarp Distinrtness in mature fruits absent (0) present (I) Lignifiration absent (0) present (I) Size of exocarp cells relative to mesocarp cells larger (I) other (0) Ditto same size (I) other (0) Dittn smaller (I) other (0) Ovary Length relative to hypanthium longrr (1) other (0) Ditto same length (I) other (0) Ditto smaller (<2/3) (I) other (0) FRUIT 'I'MITS IN MELAST'0MATAC:FAE 307

TABLE2. Standardized canonical coefficients of a discriminant analysis that included 81 spccies and nine fruit characters. Values indicate the contribution of each variable to the ca- nonical axis separating dehiscent and indehiscent fruits

Character and state Can I

Placenta persistence (0, I) -0.21 Endocarp persistence (0, I) 0.75 Outer epidermis cells smooth (I, 0) 0.47 Fused tissues (0, I) 0.16 Ovary/ hypanthium fusion complete (I, 0) -0.78 Ovary/hypanthium fusion partial (I, 0) -0.44 Exocarp distinctness at maturity (0, I) 0.00 Ovary longer than hypanthium (I, 0) 0.34 Ovary as long as hypanthium (I, 0) 0.42

(Results: Characters and their states). Of the 52 characters scored, five were excluded from multivariate analysis because breaking-up of multistate characters for binary coding resulted in the statistical redundancy of seven characters. Another 14 characters were excluded because not all characters applied to all species or had been observed in all species. For example, characters referring to fused tissues cannot be scored in species with free ovaries. The multivariate matrix for all characters and species comprised 42 12 cells, and empty cells due to such characters preclude analysis. Of the excluded characters, five dealt with endocarp lignification and cell pattern, and nine with types of sclereid distribution in the mesocarp, hypanthium, or fused tissues. To determine which characters were associated with dehiscence or indehiscence we first performed a discriminant analysis that included nine characters scored for all species (Table 2) and pooled display fruits (i.e. fruits that dehisce to expose fleshy placentas) with dehiscent fruits. Blakea paucgora, Centradenia grandgora, Medinilh homoeandra, and Sonerila malgaritacea were excluded from this analysis because they lacked values for some characters. In a second discriminant analysis that included 60 species for which the relevant data were available, placenta fleshiness (absent/ present), was added as a character, and fruits were categorized as indehiscent-fleshy ( =soft berries), dehiscent-dry (= capsules), indehiscent-dry (dry berries), or dehiscent- fleshy (=display fruit). Two further discriminant analyses were run on subsets of species, one consisting of the 28 species with a fused ovary and hypanthium (two additional such species had to be excluded because of missing values), the other of the 52 species lacking a fused ovary and hypanthium (four such species had missing values). For the first subset, an additional five characters were included in the analysis that apply to fused tissues, yielding a total of 14 characters (Table 3), while for the second subset a total of 25 characters could be used (Table 4). To characterize the major axes of variation in fruit characters we used principal component analysis on the full set of 81 species. Analyses were carried out with the statistics program SAS 6.12 using procedures CANDISC and PRINCOMP (SAS Institute Inc., 1987). Fruit types were also traced on a molecular phylogenetic tree for Melastomataceae and Memecylaceae (Clausing & Renner, in press) using MacClade 3.04 (Maddison & Maddison, 1992), with ACCTR4N character state optimization. 308 C. CLAUSING ET AL.

‘hu3. Standardized canonical coefficients of a discriminant analysis that included 81 species and nine fruit characters, and for which fruits were separated into the four categories indehiscent-fleshy, dehiscent-dry, in- dehiscent-dry, and dehiscent-fleshy ( = display fruits). High values indicate high contributions of the respective character to the canonical axis separating dchiscent-fleshy fruits from indehiscent or dehiscent-dry fruits. Thus, ovary/ hypanthium fusion and fleshy placentas explain much of the difference between fruit types, while placenta persistence per Je does not explain differences between fruit types

Charartrr and state Can I Can2

Squared canonical rorrelation K’ 0.54 -0.50 Placenta fleshy (0, I) 0.83 4.65 Placenta persistence (0, I) 0.00 0.00 F,nclocarp persistence (0, 1) 0.40 0.63 Fused tissues (0, 1) 0.25 -0.13 Ovary/hypanthium fusion complete (I, 0) -0.83 -0.23 Ovary/hypanthium fusion partial (I, 0) -0.17 -0.46 Exocarp distinrtness at maturity (0, I) 0.27 0.36 Ovary longer than hypanthium (I, 0) 0.03 0.40 Ovary as long as hypanthium (I, 0) 0.17 0.29

TABLE4. Standardized canonical coefficients of a discriminant analysis that included 28 species with fused tissues and 14 fruit characters. Values indicate the contributions of the characters to the canonical axis separating dehiscent and indehiscent fruits. Thus, endocarp persistence and abundant calcium oxalate in the fused tissues vs few sclereids in fused tissues most clearly separate dehiscent from indehiscent fruits

‘Character and state Can I

Placenta prrsistence (0, I) -0.59 Endocarp pemistencr (0, I) 0.87 Hypmthium outer rpidrrmis cells smooth (I, 0) 0.56 Fused tissues thick (0, I) -0.55 Ovary/hypanthium fusion complete (I, 0) -0.44 Ovary/hypanthium fusion partial (I, 0) - 0.02 Exocarp distinctness at maturity (0, 1) 0.00 Ovary longrr than hypanthium (I, 0) -0.06 Ovary as long as hypdnihium (I, 0) 0.1 1 Sclereid prrsencc in fused tissues (0, I) 0.30 Fused tissues with few sclereids (I, 0) -0.74 Fused tissues with an intermediate numlier of sclereids (I, 0) 0.07 Calcium oxalate presence in fused tissues (0, I) 0.66 Starch ,pnule presence in fused tissues (0, I) 0.06

RESULTS

General anatomy and morphology of melmtome bits

Of the 85 species in our sample, 25 had dry dehiscent fruits (capsules), eight dry and lignified but indehiscent fruits (dry berries), 36 fleshy indehiscent fruits (soft FRUIT TRAIlS IN MEIASI'OMATACEAE 309

Figure 1. Examples of melastome fruits. A-C, capsules. A, Dionycha bqjent B, Oxypora exba; C, Sonerila obliqua; D & E, Berries: D, Medinilla subemsa; E, bracteata; F, Pternandra hirlella.

berries), and 16 fleshy fruits that dehisce (Figs 1, 4). The latter kind of fruit was termed display fruit by Wiehler (1983) because, when mature, these capsule-like fruits split to display bluish-black, cream, or reddish juicy placentas in which small 310 G. CIAUSING ET AL seeds are embedded. The general shape of melastome fruits varies from globose to campanulate or urceolate (Fig. l), with some capsules being obpyramidal (e.g. Sarcopyramis nepalensk) or cubic (e.g. Phyllagathis dispar). In Dionycha (Fig. 1A) and Dichaetanthera, half of the ovary is exposed due to a short cup-shaped or campanulate hypanthium. Melastome fruits develop from epigynous or perigynous , and the fruit wall thus consists of two parts of different morphological origin. The pericarp derives from the ovary and consists of endo-, meso-, and exocarp. [IThe terms endo- and exocarp are used in the strict sense (Cave, 1869)) referring only to the inner and outer cell-layer of the ovary, respectively]. The outer fruit wall derives from the hypanthium and includes inner and outer hypanthium epidermis and hypanthium ground tissue. The most important differences in the anatomy of mature fruits are the extent of fusion of ovary and hypanthium, presence or absence of an endocarp, and the location and degree of lignification (the importance of these characters was also evident in all statistical analyses; see Results: Multivariate analyses). is predominantly axillary, but some genera have basal-axillary placentation or placentas located on the ovary wall between the septs. Mesocarp and hypanthium tissues comprise several layers of parenchymatous cells that are usually larger than the epidermal cells. An inner hypanthium epidermis and an exocarp occur only in parts of the fruit where ovary and hypanthium are not fused. In species that have pockets that house the in stage, these are located between pericarp and hypanthium tissue (Fig. 4E). They appear as compressed lacunae covered by epidermis. The epidermis on the inside of the pockets eventually grades into the exocarp and that on the outside continues into the inner epidermis of the hypanthium. Enations or hairs may be found on both epidermises and on the exocarp. Bristles, uni- or multiseriate hairs, glandular hairs or scales are common on Melastomataceae fruits, although some fruits are entirely glabrous, e.g. those of Medinilla and Sonen’la.

Characters and their states

See Table 1. Placentation. Placentation is predominantly axillary. In Kibessieae (Rernandra; Fig. 4F), the placentas are located on the ovary wall between the septs, and in some Astronieae and Dissochaeteae (Creochiton) they are positioned basal-axillary. Placentas can disintegrate or persist in mature fruits. In the statistical analyses this character showed a consistently high association with fruit type (below). Where the placenta disintegrates, the seeds lie loosely in the locule or are embedded in fruit pulp. The latter is the case in most species of Medinilla (Fig. 4C). Where the placenta persists it can be club-shaped, V- to T-shaped, branched, or stalked (Fig. 2). In most species of Melastoma, a fleshy placenta fills the entire (Figs 4A, 6C). Rarely does the placenta consist of only the central , as for instance in Blastus borneensis. Endocarp. Three endocarp characters were coded, viz. whether the endocarp dis- integrates or persists in mature fruits, whether it is lignified, and which of four patterns its cells form in surface view. Endocarp cell patterns were categorized into four states-mosaic-like, undulate (the commonest situation), elongate-undulate, or 31 1

Figure 2. Placenta shapes. A, fleshy; B, stalked; C, club-shaped; D, V- or T-shaped; E, branched.

Figure 3. Endocarp cell patterns in surface view. A, mosaic pattern; B, undulate pattern; C, elongate- undulate pattern; D, elongate pattern. elongate (Fig. 3)-but the character was excluded from the multivariate analyses (Material and methods). Mesocarp. The mesocarp consists of parenchymatous cells that may be dead in mature fruits. Mesocarp thickness is highly variable, with the number of cell layers varying between 3 and 25; 3-12 layers were coded as thin, and more than 12 as thick. Another mesocarp character showing continuous variation is the number and 312 G. CIAUSING E7‘4L distribution of sclereids, which are elongate and/or branched cells with lignified walls. We recorded the presence or absence of sclereids and three classes of sclereid abundance, namely few sclereids (c0.4 per one degree and cell layer), intermediate abundance (0.5-0.8 per one degree and cell layer), and many sclereids ( > 1 per one degree and cell layer). Sclereids may be scattered throughout the tissue, clustered in small groups, or form closed rings, but these patterns were not included in the statistical analyses. Further mesocarp characters concerned the presence of calcium oxalate crystals (found in most species) and starch granules (found in Medinilla and Plethiandra). Hypanthium. The hypanthium ground tissue is very similar to that of the mesocarp. Like the mesocarp, it is characterized by a varying density and distribution of sclereids and calcium oxalate crystals, with sclereid and crystal abundance normally being higher than in the mesocarp. Definition of classes of abundance was as for the mesocarp. In two instances, Arthmstemma ciliaturn and Blastus borneensis, the subepidermal cell layers are lignified. Another hypanthium character included in some statistical analyses was the size of the inner hypanthium epidermis cells relative to those of the exocarp. We distinguished two states, viz. whether the cells of the inner hypanthium epidermis are smaller (the commonest condition) or larger than those of the exocarp. Different from exocarp cells, cells of the inner hypanthium epidermis remain unlignified. The outer hypanthium epidermis cells can be smooth, convex or papillose (coding, see Table 1). Fused tissues. Where ovary and hypanthium are fused, their tissues become in- distinguishable, resulting in the category ‘fused tissues’ for which the same characters were recorded as for the mesocarp and hypanthium. The fusion df ovary and hypanthium can be complete or partial, with much plasticity even within genera, especially in Medinilla and Melastoma. Difficulties in determining the extent of fusion arise from the pockets, which extend downwards into the region where ovary and hypanthium are adnate and which appear as compressed lacunae in cross-sections (Fig. 4E). We scored ovaries as free if no septs connected them to the hypanthium. Exocaq. Four exocarp characters were included in the statistical analyses, namely whether the exocarp was distinct in mature fruits, whether it was lignified, and what the size of its cells was relative to those of the mesocarp. Ovaly. An ovary character included in the statistical analyses addressed its length relative to that of the hypanthium, the states being ovary longer than hypanthium, of the same length, or shorter.

Multivariate anabses

The discriminant analysis of nine characters for 81 species (see Material and methods; Appendix) was able to separate dehiscent and indehiscent fruits (Wilk’s lambda = 0.48, F= 9.7, df= 8,72, RO.0001). Although character distributions over- lapped to some degree, dehiscence accounted for 52% of the variation in the nine characters. Dehiscence was positively correlated with endocarp persistence and negatively with ovary/hypanthium fusion (whether complete or partial) and a persistent placenta (Table 2). FRUIT TRAITS IN MELASTOMATACFAE 313 A B

Figure 4. Cross sections of melastome fruits. A, ; B, Oxyspora eea;C, Medinilla submsa; D, Sonerih malgeniacea; E, Dirsochaeta bracteatu; F, pternandra hirtella. 314 C. CLAUSING ETAL. If fleshiness of the placenta was included in the data set and fruits were placed in the four categories indehiscent-fleshy (soft berries), dehiscent-dry (capsules), indehiscent-dry (dry berries), and dehiscent-fleshy (display fruit), discriminant analysis found three axes of which the first two were most significant (WWs lambda=0.22, F= 4.8, df= 24,168, RO.000 1; n = 60 species). The first axis primarily separated dehiscent-fleshy ( = display) fruits from indehiscent-fleshy and indehiscent-dry fruits (RO.0001,p=54%). It was associated positively with a fleshy placenta and a persistent endocarp. It was negatively associated with ovary/hypanthium fusion (Table 3). The second canonical axis separated dehiscent-dry fruits (=capsules) from the three other types (RO.001, p=50%) and was positively associated with a persistent endocarp and negatively with fusion of hypanthium and ovary. Discriminant analysis of 28 species with a fused ovary and hypanthium, using 14 characters, again separated dehiscent and indehiscent fruits, with 56% of the variance in the data set accounted for by dehiscence, but only marginal significance because of the small sample size (Wilk's lambda = 0.44, F= 1.4, df= 13,14, P= 0.27). As in the analyses based on larger sample sizes, dehiscence was positively correlated with a persistent endocarp and negatively with a persistent placenta and complete ovary/ hypanthium fusion (Table 4). Weaker correlations existed between dehiscence and the presence of calcium oxalate crystals and sclereids. The 25 characters included for the 52 species lacking fused tissues easily discriminated dehiscent and indehiscent fruits, with 83% of the variance attributable to mode of dehiscence (Wilk's lambda = 0.17, F= 5.6, df= 24,27, RO.000 1). The highest correlations existed between dehiscence and the presence of sclereids in the mesocarp, followed by a persistent endocarp (Table 5). Principal component analysis based on nine characters revealed a single dominant axis that accounted for 40% of the total variation present in the data. Species at one end of the axis have fruits with a complete fusion of ovary and hypanthium tissues, while fruits at the other end have distinct exocarps and only partial ovary/ hypanthium fusion (Table 6).

DISCUSSION

The general picture that emerges from the five multivariate analyses is that a small number of characters, addressing the persistence and fusion of tissues, are consistently correlated with mode of fruit dehiscence. Indehiscence is associated with ovary/hypanthium fusion, placenta persistence, lack of a persistent endocarp, and a dearth of sclereids in these tissues, while dehiscence is correlated with the opposite states and a persistent exocarp. Specific to the Old World are display fruits, which in the principal component analysis were clearly differentiated from other fruit types, mainly because of their combination of features otherwise associated with dehiscent fruits, such as persistent endocarp and absence of ovary/hypanthium fusion, with a feature otherwise found in indehiscent fruits, namely persistent, fleshy placentas.

Structural heterOgaep of berries and capsules

Berries. In New World Blakeeae and Miconieae, berries (i.e. indehiscent fruits) usually have fleshy placentas and fused tissues, while in the Old World this correlation is FRUIT TRAITS IN MELASTOMATACEAE 315

TABLE5. Standardized canonical coefficients ofa discriminant analysis that included 52 species without fused tissues and 25 fruit characters. Values indicate the contributionsofeach character to the canonical axis separating dehiscent and indehiscent fruits. Characters that most clearly separate dehiscent from indehiscent fruits are sclereids in the mesocarp and endocarp per- sistence. Sclereid presence vs presence in low or intermediate numbers make inverse con- tributions to the separation of fruit types because of O/ 1 coding (cf. Table 1)

Character and state Can 1

Placenta persistence (0, I) -0.30 Endocarp persistence (0, I) 1.13 Hypanthium outer epidermis cells smooth (1, 0) 0.77 Hypanthium thick (0, I) 0.28 Ovary/hypanthium fusion complete (1, 0) -0.19 Ovary/hypanthium fusion partial (1, 0) -0.11 Exocarp distinctness at maturity (0, I) 0.00 Ovary longer than hypanthium (1, 0) 0.44 Ovary as long as hypanthium (1, 0) 0.25 Sclereid presence in hypanthium (0, 1) - 0.80

Hypanthium with few sclereids (1, 0) 0.15 Hypanthium with an intermediate number of sclereids (I, 0) - 0.6 I Calcium oxalate presence in hypanthium (0, I) -0.18 Starch granule presence in hypanthium (0, 1) -0.32 Inner epidermis cells larger than hypanthium cells (I, 0) - 0.20 Inner epidermis cells of same size as hypanthium cells (I, 0) 0.14 Mesocarp thick (0, 1) -0.04 Sclereid presence in mesocarp (0, 1) -1.15 Mesocarp with few sclereids (I, 0) 1.36 Mesocarp with an intermediate number of sclereids (1, 0) 1.10 Calcium oxalate presence in mesocarp (0, I) -0.07 Starch granule presence in mesocarp (0, I) 0.43 Exocarp lignification (0, 1) 0.22 Exocarp cells larger than mesocarp cells (I, 0) 0.45 Exocarp cells of same size as mesocarp cells (I, 0) -0.41

TABLE6. Relative contributions of nine fruit characters in a principal component analysis of fruits from 81 species. As in Tables 2-5, values indicate the contributions of each character to the canonical axis separating dehiscent and indehiscent fruits. Thus, presence or absence of fused tissues, exocarp distinctness in mature fruits, and ovary/hypanthium fusion contribute most strongly to the canonical axis separating dehiscent and indehiscent fruits

Character and state Prin 1

Placenta persistence (0, I) 0.1 1 Endocarp persistence (0, 1) 0.22 Hypanthium outer epidermis cells smooth 0.08 Fused tissues (0, 1) - 0.50 Ovarylhypanthium fusion complete (1, 0) - 0.49 Ovary/hypanthium fusion partial (I, 0) 0.44 Exocarp distinctness at maturity (0, I) 0.50 Ovary longer than hypanthium (1, 0) -0.01 Ovary as long as hypanthium (1, 0) 0.09

less strong. This is mainly due to the Old World genus Melastoma, heavily represented in our sample, which has fleshy placentas, but nearly free or only partially fused ovaries and hypanthia, and the Dissochaeta alliance, also strongly represented in the 316 G. CLAUSINC ETAL. sample, which contains many species with dry or woody berries (Figs lE, 4E). Examples of soft berries with fleshy placentas are Lorga spmceana and Clidemia donnell- smithii (Miconieae), and Medinilla amplectern, and Pachycentria glauca (Dissochaeteae). Soft berries are characterized by the absence of lignified tissues and a dearth of sclereids in the fruit walls, and often by an early-degrading endocarp. Hard berries are characterized by a persistent endocarp as seen in Dissochaeta rgormata and D. aJniS, in which mesocarp and hypanthium ground tissues are not fused and in which the thick ground tissue contains a dense ring of sclereids. Some miconiean berries, such as those of Bellucia aequiloba, also have a dense ring of sclereids in the ground tissue, but their ovary and hypanthium are completely fused and there is no endocarp. In all species of the neotropical genus Blaha investigated, pericarp and hypanthium are fused and heavily sclerified, and in B. mtundijilia the endocarp is even lignified. As shown by these examples, berries in Blakeeae, Dissochaeteae, and Miconieae are so variable in structure that they cannot be considered a single character, the presence or absence of which can be coded in binary fashion as done by Renner (1993). Capsules. Roughly half the species sampled had dehiscent fruits. As with berries, capsules in Melastomataceae are very heterogenous, but less of that variation is represented in our sample, which lacks representatives of Microlicieae and Merianieae (all of which have capsules) and includes only one Bertolonieae, a group with particularly diverse capsules (see illustrations in Baumgratz, 1983-85). In our sample, dehiscence is associated with a persistent endocarp and unfused or only partially fused hypanthium and ovary tissues. Exceptions are some species of Sonerila (Sonerileae; Figs lC, 4D) and ptentandra (Kibessieae; Figs lF, 4F). Pternandra fruits represent intermediates between berries and capsules, with variation in fruit dehiscence among species. In some, the fruit walls consist of fused hypanthium and ovary tissues and are strikingly hard due to an abundance of sclereids scattered or clustered in the fruit wall. Such fruits were referred to as capsulae baccatae by De Candolle (1828). In Sonerila, capsules may have a more or less completely fused hypanthium and ovary, their tissues being indistinguishable (Fig. 4D). If anything, only the endocarp is lignified, and the capsules then resemble the woody berries found in the Dis- sochaeteae. In other lineages, capsule lignification is also highly variable. Melastomeae, for example, comprise capsules with lignified endocarp, exocarp, and subepidermal layers of the hypanthium ground tissue (e.g. Arthrostemma ciliaturn), capsules with lignified endocarp and exocarp (e.g. Ebouchina uroilleana), capsules in which only the endocarp is lignified (e.g. Dichaetanthera and Dionycha), and capsules lacking lignification (some species of Melastoma; Fig. 4A).

Fruit evolution in Melacrtomataceae, especial4 in the Dissochae&adSonm'leae complex

The heterogeneity of melastome capsules and berries suggests that both may have evolved several times within the family. In addition to the present morphological- anatomical survey, molecular phylogenetic work (Clausing, 1999; Clausing & Renner, in press) shows that Renner's (1993) grouping of Blakeeae, Dissochaeteae, and Miconieae on the basis of shared soft berries was erroneous. Evidence from combined FRUIT TRAITS IN MELASTOMATACME 317

hits Rhynchanthera Microlicia Microlicieae unordered Lavoisiera Soft berry Rhexia Arthrostemma Dry berry Monochaetum Dissotis Woody berry Tibouchina Pteroplepis Melastomead Fleshy capsule Melastoma Rhexieae Osbeckia 0DV capsule Dichaetanthera Centradenia Drupe Nepsera m Aciotos Equivocal Diplectria Medinilla Calvoa Amphiblemma Dissochaetead Driessenia Sonerileae complex Blastus Phyllagathis Blakea - Blakeeae Molena Monolena Bertolonieae? Bertolonia ntrazygia Maieta lbcoca Mic o ni e a e Leandra Clidemia Gmffenrieda Merianin Merianieae Adelobotrys Bellucia - ? Astronia Astronia Astronieae Pternandm Pternandra Kibessieae Mouriri Mouriri Memecylaceae Memecylon Olinia Penaea Crypteronia outgroups Myrtus Eugenia Ludwigiu

Figure 5. Fruit evolution in Melastomataceae and Memecylaceae as traced under ACCTRAN optimization on the highest likelihood tree obtained under the general time-reversible model, using concatenated rbcL, rp116, and ndhF sequences (Clausing & Renner, in press). Soft berries are also found in some members of Melastornu (Melastomeae) and Puchycentria (Dissochaeteae). rbcL, ndhF, and rpll6 sequence data (Fig. 5) indicates that soft berries evolved at least three times and dry berries at least twice within Melastomataceae. In addition to the three cases shown in Figure 5, soft berries also evolved in Melastoma (Melastomeae) and Puch_centriu (Dissochaeteae). An ndhF phylogeny of the Dis- sochaeteae/Sonerileae complex further suggests that soft berries may have evolved four times within that alliance alone (Clausing, 1999). There also appear to have been two switches to dry berries and two switches to berries with a dense ring of sclereids in the hypanthium. 318 G. CLAUSINC ETAL.

L. 0

w FRUIT TRAITS IN MELASTOMATACEAE 319

The present study included fruits from 15 genera and many species of Dis- aochaeteae/Sonerileae (Appendix), and their anatomy and morpholo,gy varied greatly. For example, woodiness h some dissochaetcan hits is achieved by scler- ification of the hypanthium ground tissue and/or the mesocarp (e.g. Dissochaeta bracteata; Fig. 4E), while in others, such as Dissochaeta afik or D. r$ormata, hardness is due to a closed ring of sclereids in the hypanthium ground tissue. The anatomy of these woody berries is very different from that of soft berries found in other Dissochaeteae (compare Dissochaeta bracteata, Fig. 4E, and Medinilla suberosa, Fig. 4C). Another example of the plasticity of fruits in this alliance is provided by Kendrickia and its sister group Catanthera. The former has a thick-walled capsule that opens with four regular longitudinal cracks (Clawing, pers. obs.), while Catanthera has soft berries. Both genera are -climbers and form a well-supported monophyletic group- Sonerileae sequenced so far (Amphiblemma, Blastus, Calvoa, Driesseniu, and Ptyllagathk) appear nested within Dissochaeteae (Fig. 5). Capsules in these genera often have a lignified endocarp and sometimes also a lignified exocarp, but at least in Driessenia glandul&ra the hypanthium is fleshy and there is no lignification, making the fruits similar to those of some Dissochaeteae, such as Medinilla serpens.

Fruit variation at the intrageneric level: the case @Melastoma

Another example of fruit variation among closely related species as indicated by ndhF sequences (Meyer, 1999; Renner & Meyer, submitted) is provided by Mehtoma, a member of the pantropical Melastomeae (including Rhexieae), which appear to be monophyletic (Fig. 5).Melastoma species have a persistent endocarp that is either lignified or parenchymatous (Fig. 4A). The exocarp and the inner epidermis of the hypanthium are always distinct so that mesocarp and hypanthium ground tissue can be distinguished. The fusion of ovary and hypanthium ranges from nearly free to up to 3/4 fused. Mesocarp and hypanthium vary in number of cell layers and distribution of sclereids and calcium oxalate crystals. Often, the subendocarpal, subexocarpal, or subepidermal layers are stuffed with calcium oxalate crystals. Figure 6 shows three fruit types that have evolved within this genus of 22 species (Meyer, in press). The commonest fruit type is represented by Melastoma beccarianum (Fig. 6C), with irregularly transversal splitting of woody fruit walls and seeds embedded in a soft or solid pulp formed by the placentas. In Melastomataceae, such display fruits are restricted to Melastoma, and discriminant analysis was able to clearly distinguish these fruits from the three other fruit types in our sample (above). Similar display fruits occur in Gesneriaceae (Smith & Carroll, 1997) and Marcgraavia (Meyer, pers. obs.). The seeds are dispersed by birds that feed on the fleshy, sweet placentas (e.g. Gross, 1993 for Melastoma malabathricum (as M. am)). A second type of fruit is found in M. orientale (Fig. 6B), which also has a fleshy placenta but does not split open. It contains no sclereids in mesocarp and hypanthium ground tissue. Ovary and hypanthium are fused for 3/4 of their length. Like soft berries in Blakeeae, Dissochaeteae, OF Miconieae, this fruit shows a striking increase in size during ripening, but in anatomy it is very different from those fruits, indicating that berries in Melastoma evolved independently. Finally, a third fruit type is found in M. pellegriniunum (Fig. 6A), the capsules of which open by apical pores. The capsule of this species is characterized by the 320 C. CLAWING LTAL. complete drying out of the placenta at maturity so that the seeds lie loosely in the ligdied endocarp from where they are released through the pores. The hypanthium is parenchymatous with ten lignified nerves that persist when the fruit has shed its seeds, a striking parallelism to the capsules of Oxyspora exba (Figs lB, 4B) and Astronia smihciilia. Capsules of A. smilaciifolia also have a persistent endocarp, but it is not lignified. Their star-like dehiscence results from the irregular disintegration of the fruit wall (which consists of the fused ovary and hypanthium), with sharply triangular parts of the wall persisting as a dome-shaped frame. Capsules of 0. exba have a heavily lignified and persistent endocarp and a hypanthium with prominent vascular bundles. The seeds are released loculicidally through five longitudinal slits in the endocarp.

Fruit gpes as phylogaetic indicators

Berries may evolved three times (Fig. 5) within the family, and similar to the examples from other families listed in the introduction, the use of berries as a supposed synapomorphy of Blakeeae, Dissochaeteae, and Miconieae (Renner, 1993) resulted in the unnatural grouping of a large number of palaeotropical and neotropical genera that do not form a monophyletic clade. Erroneous homology assumptions may have partly resulted from a lack of detailed studies of fruit morphology and anatomy. In Melastomataceae, as in other Myrtales, fruit characters are highly variable even at the intrageneric level (as evidenced by Melastoma), and they may rarely be reliable indicators of monophyletic groups. This does not mean that carefully coded fruit-anatomical characters should not be included in a phylogenetic analysis of the family. Indeed, in some other families, fruit characters have been found to be non-homoplastic even at higher hierarchical levels. Thus, Rohwer (1 996) showed that fruit-anatomical traits are valuable for recognizing generic relationships in . A phylogenetic study of Cornus based on cpDNA restriction site and morphological data also found the two major clades to be congruent with the presence of iridoid glucosides and fruit colour (Xiang et al., 1996): one clade contains the blue-fruited, the other the red-fruited dogwoods, as postulated purely on morphological grounds by Eyde (1985). Also in Eythrina, a phylogeny based on cpDNA restriction site and morphological data indicates that exocarp ornamentation and texture are phylogenetically informative (Bruneau, 1996), and in a study of phylogenetic relationships and dispersal system evolution in , too, fruit characteristics were found to be non-homoplasious synapomorphies of certain groups of genera (Snijman & Linder, 1996). All this supports Stevens' (1991) general conclusion that qualitative morphological characters require critical examination and exact coding. In the end, however, the reliability of fruit characters in any one group can only be tested by considering molecular and morphological data together.

Fruit character plasticig and dirpersal ecology

Fruit character plasticity in Melastomataceae must relate to shifts in mechanisms, which in turn relate to habitat. Thus, changes from fleshy berries to dry capsules may correlate with colonization of more open habitats where wind dispersal may be less costly than bird dispersal or where frugivorous understorey FRUIT TRAITS IN MELASTOMATACME 32 I

birds may be less abundant (Stiles & Rosselli, 1993). Alternatively, capsular species may undergo selection for increased fleshiness of their exocarp and placentas upon entering more closed forest habitats. Stiles and Rosselli found that in the neotropics the pattern of species richness of small mashers, such as manakins and tanagers, conforms closely to the distribution patterns of berry-fruited melastomes (i.e. Mi- conieae and Blakeeae). These birds, like Miconieae and Blakeeae, are most diverse in wet forests at middle elevations. Similar relationships may exist in the Old World although there are no detailed studies of fruit handling and dispersal of palaeotropical melastomes by birds. Species of Medinilla and other soft-berried Dissochaeteae occurring in montane forests (700-2000 m) at Mt. Kinabalu National Park (Borneo) depend on small to medium-sized birds living in flocks and on small mammals for dispersal of their seeds (Clausing, 1999). Another selective factor may be . In melastomes, an epiphytic habit is strongly associated with baccate fruits adapted to bird dispersal. For example, in the neotropics, 85% of epiphytic melastomes produce berries (Renner, 1986). Some 300 of the 350 Dissochaeteae known are facultative or obligate epiphytes (Clausing, 1999), and thus, a combination of similar growth forms (namely climbing forms) and habitats (tropical forests) may have promoted independent evolution of bird- adapted berries in Dissochaeteae and Miconieae. Once juicy berries had evolved, further adaptive shifts became possible. While most Miconieae offer small red or blue berries, a few have much larger yellow ones adapted for dispersal by bats or monkeys (Renner, 1989), and this may then have selected for hardening of these berries as seen in Bellucia. Little is known about the biophysical details of capsule function in melastomes (but see Weber, 1987; Stone & Weber, 1988; Cellinese, 1997, on rain splash dispersal in Sonerileae). However, wind tunnel experiments in other groups have shown that capsule morphology can influence the shape of seed shadows (e.g. Blattner & Kadereit, 199 1). This suggests that parallelisms in capsule morphology and anatomy, such as the sclerified vascular bundles that function in capsule opening in Astronia, Ovspora eea,and Melastoma pellegrinianum, may reflect similar selective conditions in terms of seed dispersal in these species’ habitats. Capsular species also may undergo selection for increased fleshiness of their pericarp and placentas upon entering more closed forest habitats. Together, our results indicate that berries and capsules evolved several times within Melastomataceae, with anatomical evidence supporting molecular phylo- genetic findings. Ultimately, the exceptional evolutionary flexibility of melastomes fruits may stem from the structure of myrtalean flowers in which hypanthium and ovary tissues are both involved in the construction of the fruit. This allows different degrees of tissue fusion, thickening, hardening, or disintegration, which in turn permits fruit morphology to change dramatically even between closely related species.

ACKNOWLEDGEMENTS

We thank R.E. Ricklefs and H. Bruelheide for statistical advice, J.W. Kadereit and an anonymous reviewer for constructive criticism, J. Rohwer for comments on

, an early version of the manuscript, N. Cellinese and J. Regalado for identifjmg species of Sonerileae, A. Berg for drawing Figures 1,2, and 6, and Barbara Dittmann 322 G. CIAUSING ETdL. for preparing microtome sections. Financial support for this project came from the Deutsche Forschungsgemeinschaft (grant RE/603/2- 1 to S.S. Renner). Field work and collecting of fruit samples was conducted with kind permission of the Perancang Ekonomi Unit, Kuala Lumpur, and Sabah Parks, Kota Kinabalu, Sabah. Logistic support came from Kinabalu Park, Sabah, the Sabah Department in Sandakan, the Institute of Biodiversity and Environmental Conservation of the Universiti , Sarawak, the Department of Botany of the National University of Hanoi, and the Department of Biology of Prince of Songkhla University, Hat Yai, Thailand.

REFERENCES

Baumgratz JFA. 1983-1985 [1988]. Morfologia dos frutos e sementes de Melastomaticeas brasileiras. Archivos do Jardim Botanic0 do Rio de Janeim 27: 1 13-1 55. Blattner F, Kadereit JW. 1991. Patterns of seed dispersal in two species of Papaveer L. under near- natural conditions. 185: 55-64. Boissieu MH de. 1912. Une MClastomacCe asiatique d’un genre africain. Bulletin de la SociM Botanique de France 59 33C332. Bremer By Andreasen K, Olsson D. 1995. Subfamilial and tribal relationships in the Rubiaceae based on rbc L sequence data. Annals ofthe Missouri Botanical Garden 82: 383-397. Bremer By Eriksson 0. 1992. Evolution of fruit characters and dispersal modes in the tropical family Rubiaceae. Biological Journal ofthe Linnean So&o 47: 7S95. Bruneau A. 1996. Phylogenetic and biogeographical patterns in Erythrina (Leguminosae: Phaseolcae) as inferred from morphological and DNA characters. Systematic Botany 21: 587405. Candolle de AP. 1828. Melastomaceae. Rudmmus 3: 99-202. Cave C. 1869. Structure et dkveloppement du fruit. Annales des Sciences Naturelles; Botanique, St%s 5 10 123- 190. Cellinese N. 1997. Notes on the systematics and biogeography of the Sonerila generic alliance (Melastomataceae)with special focus on fruit characters. Tivpical Biodiverno 4: 83-93. Clausing G. 1999. Die Systematikder Dissochaeteae und ihre Stellung innerhalb der Melastomataceae. Ph.D. dissertation,Johannes Gutenberg-Universitat, Mainz. Clausing G, Renner SS. (in press). Molecular phylogeneticsof Melastomataceae and Memecylaceae: implications for character evolution. American Journal ofBotany. Cogniaux CA. 1891. Melastomaceae. In: De Candolle A, De Candolle C, eds. Monographiue Phan- emgamarum vol7. Paris: G. Masson, 1-1256 Conti E, Litt A, Sytsma I(J. 1996. Circumscription of Myrtales and their relationships to other : evidence from rbcL sequence data. American Journal ofBotany 83: 221-233. Eyde RH. 1985. The case for monkey-mediated evolution in big-bracted dogwoods. ArnoMia 45: 2-9. Graham SA, Crisci JV,Hoch PC. 1993. Cladistic analysis of Lythraceae smu lato based on morphological characters. Botanical Journal ofthe Linnean Socieo 113: 1-33. Gross CL. 1993. The breeding system and of Melrrttoma afine (Melastomataceae);a pioneer in tropical . Biotropica 25: 468-474. Hoot SB. 1995. Phylogeny of Ranunculaceae based on preliminary atpB, rbcL and 18s nuclear ribosomal DNA sequence data. In: Jensen U, Kadereit JW,eds. The .ystematics and evolution ofthe Ranunculjfforae. Wen: Springer, 24 1-25 1, Igersheim A. 1993. The characters states of the Caribbean monotypic endemic Strumpja (Rubiaceae). NoNordicJournal ofBotay 13: 545-559. Johnson US, Briggs BG. 1984. Myrtales and Myrtaceae - a phylogenetic analysis. Annalr ofthe Missouri Botanical Garden 71: 700-756. Kadereit JW. 1994. Molecules and morphology, phylogenetics and genetics. Botania Acta 107: 369-373. Kearns CA, Inouye DW. 1993. lichniquesforpollinatwn biologkts. Niwot, Colorado: Colorado University Press. Maddison WP, Maddison DR. 1992. MacClade: anabsis ofptjylogeny and character evolution (version 3). Sunderland, M4 Sinauer & Associates. FRUIT TRAITS IN MELASTOMATACEAE 323

Morgan DRYSoltis DE, Robertson KR. 1994. Systematic and evolutionary implications of rbcL sequence variation in Rosaceae. American Journal ofBotany 81: 89Ck903. Meyer K. 1999. Phylogenic und Systematik der Gattung Melastoma (Melastomataceae)unter besonderer BerUcksichtigung des M. malabathhum-Komplexes. Ph.D. dissertation, Johannes Gutenberg- Universitiit, Mainz. Meyer K. (in press). Revision of the Southeast Asian Genus Melastoma (Melastomataceae).Blumen. Naudin CV. 1851. Melastomacearum monographicae descriptiones. Annalps des Sciences Naturelles; Botanique, Series 3 17: 305-382. Plunkett GM, Soltis DE, Soltis PS. 1996. Evolutionary patterns in Apiaceae: inferences based on mat K sequence data. &stematic Botany 21: 477-495. Renner SS. 1986. The neotropical epiphytic Melastomataceae: phytogeographic patterns, fruit types, and floral biology. Sellyana 9: 104-1 1 1. Renner SS. 1989. A survey of reproductive biology in neotropical Melastomataceaeand Memecylaceae. Annals ofthe Mksouri Botanical Garden 76: 496-518. Renner SS. 1993. Phylogcny and classification of the Melastomataceae. Nordic Journal of Botany 13 5 19-540. Renner SS, Clausing G, Cellinese N, Meyer K. (in press). Melastomataceae. In: Larsen K, Nielsen I, eds. Flora of77iailand. Rohrer JR, Robertson KR, Phipps JB. 1991. Variation in structure among fruits of Maloideae (Rosaceae). Amerimn Journal of Botany 78: 16 17-1 635. Rohwer J. 1996. Die Frucht- und Samenstrukturen der Oleaceae - eine vergleichend-anatomische Untersuchung. Bibliotheca Botanica 148 1-1 77. SAS Institute. 1987. SAS 6.12. $AS/STATguidejr personal computers. Cary, NC. SmithJF, Carroll CL. 1997. A cladistic analysis of the tribe Episcieae (Gesneriaceae)based on ndhF sequences: origin of morphological characters. $stmatic Botany 22: 7 13-724. Snijman DAYLinder HP. 1996. Phylogenetic relationships, seed characters, and dispersal system evolution in Amaryllideae (Amaryllidaceae).Annals of the Missouri Botanical Garden 83: 362-386. Soltis DE, Soltis PS, Bothel KD. 1990. Chloroplast DNA evidence for the origins of monotypic Bensoniella and Conimitella (Saxifragaceae).Systematic Botany 15: 349-362. Spjut RW. 1994. A systematic treatment of fruit types. Memoirs of the New rOrk Botuniral Garden 70 1-182. Stebbins GL. 1974. Flowerirlg plank evolution above the species level. Cambridge: Harvard University Press. Stevens PF. 1991. Character states, morphological variation, and phylogenetic analysis: a review. &&matic Botuny 16: 553-583. Stiles FG, Rosselli L. 1993. Consumption of fruits of the Melastomataceae by birds: how diffuse is coevolution? EgetatiO 107-108: 57-73. Stone BC, Weber A. 1988. A new species of Phyllugathk (Melastomataccae) from the Endau-Rompin proposed National Park, Malaysia. Aoceedings of the Academy of Natural Sriences of Philadebhia 139: 307-313. Sytsma yT, Gottlieb LD. 1986. Chloroplast DNA evidence for the origin of the genus Hetemgaura from a species of Clarkia (Onagraceae). Aoceedirlgs ofthe National Academy ofthe United States ofAmerica 83: 5554-5557. TrianaJJ. 1871. Les Mtlastomactes. Eansactions ofthe Linnean Socieo oflandon, Botany 28 1-188. Wagstaff SJ, Ohtead RG. 1997. Phylogeny of Labiatae and Verbenaceae inferred from rbcL sequences. Systematic Botany 22: 165-179. Weber A. 1987. Two new species of Phyllqathk related to I! tuberculata (Melastomataceae) from Peninsular Malaysia. &stmatics and Evolution 157: 187-199. Wiehler H. 1983. A synopsis of the neotropical Gesneriaceae. Selbyana 6: 1-249. Xiang Q-Y, Brunsfeld SJ, Soltis DE, Soltis PS. 1996. Phylogenetic relationships in Cornus based on chloroplast DNA restriction sites: implications for biogeography and character evolution. 9~tematicBotany 21: 515-534. Nw 4-

APPENDIX

Fruit material investigated species and collecting locations. Classification follows Renner (1993), but recognizes Bertolonieae and Dissochaeteae smCogniaux (1891) based on molecular phylogenetic results (Clawing, 1999; Clausing & Renner, in press) although the final circumscription and disposition of these tribes is still unclear.

Species Tribe Material Collector Collecting locality

Anerintlcirtur fllal7nthus King Sonerileae alcohol sample Clausing 70 Cameron Highlands, Pahang, Malaysia Anninclcirh~~manop/IyuUr Bakh.f. Sonerileae alcohol sample Clausing 248 Gurulau Ridge, Sabah, Malaysia Artk&mma ciliaiiun Pav. ex G.Don. Melastomeae living plant Greenhouse BG Mainz Asbvnio sm&@ia Triana Astronieae alcohol sample Clausing 189 Bukit Matang, Sarawak, Malaysia Beccarhthus sp. Astronieae herbarium specimen Coods ct al. NGF 40347 New Guinea BeIlucia acgutloba Pilg. Miconieae alcohol sample Cid d al. 5196 Cruzeiro do Sul, Acre, Brazil p Belhcia pcntmnna Naudin Miconieae alcohol sample Renner 990 Rio Branco, Acre, Brazil Bertobnia mocuIalD DC. Bertolonieae living plant Greenhouse BG Maim F BMca m2omola Donn.Sm. Blakeeae herbarium specimen Wilbur & Stone 9861 Heredia, Costa Rica 5 BMca pauc&ra Gleason Blakeeae herbarium specimen Mori & Kallunki 5900 Chiriqui, Panama BIakca rnatndijilia D.Don Blakeeae herbarium specimen Luteyn & Cotton 10894 Carchi, 3 BhhbomMlsir Cogn. Sonerileae alcohol sample Clausing 25 Mt. Kinabalu N. P., Sabah, Malaysia Cahnthcrn taunmm (Men.) Regalado Dissochaeteae alcohol sample Clawing 109 Mt. Kinabalu N. P., Sabah, Malaysia Catanthna tefrandra Stapf Dissochaeteae alcohol sample Clausing 2 14 Mt. Kinabalu N. P., Sabah, Malaysia Cm&admia grand$Xa (Schltdl.) Endl. Melastomeae living plant Greenhouse BG Mainz Clidcmia donncll-sn~iCogn. Miconieae living plant Greenhouse BG Maim Cmhhmanticola (Ridl.) Veldkamp Dissochaeteae herbarium specimen Clemens 34455 Sabah, Malaysia Dichaetanha arbma Baker Melastomeae alcohol sample Clausing 281 Antananarivo, Madagascar Dichactanthna asPm;na Cogn. Melastomeae alcohol sample Clausing 280 Antananarivo, Madagascar hychbopn Naudin Melastomeae alcohol sample Clausing 300 Antsiranana, Madagascar Dtphtrk becccariana (Cogn.) Kunae (**) Dissochaeteae alcohol sample Clausing 146 Bukit Matang, Sarawak, Malaysia Dtphtrk latiilia (Tria~)Kunae (**) Dissochaeteae alcohol sample Clawing 264 Bako N. P., Sarawak, Malaysia Dirrochnrto afiir Forth.) Clausing (***)= Dissochaeteae alcohol sample Clawing 65 Cameron whlands, Pahang, Malaysia Mmknemomsa (Jack) Bakh.f. DiFsocharla muuclata Hook. f. ex Triana Dissochaeteae alcohol sample Clawing 234 Poring Hot Springs, Sabah, Malaysia Dissochacla beccahna Cogn. Dissochaeteae alcohol sample Clawing 245 Ranau, Sabah, Malaysia APPENDIX-continued

Species Tribe Material Collector Collecting locality

Dircochoetn bructazfa (Jack) Blume Dissochaeteae alcohol sample Clausing 149 Bukit Matang, Sarawak, Malaysia Dirsachaeta celebica Blume Dissochaeteae alcohol sample clausing 46 Poring Hot Springs, Sabah, Malaysia Dissochaetu divaricatu (wid.) G.Don (***) = Dissochaeteae alcohol sample Clausing 32 Poring Hot Springs, Sabah, Malaysia Dtplecbia diuaricata (Wid.) Kuntze Dirrothocta gln6ra Merr. subsp. kinabaiunrir Dissochaeteae alcohol sample Clawing 33 Mt. Kinabalu N. P., Sabah, Malaysia Veldkamp (***)=oipkbia gln6ra (Merr.) M.P.Nayar subsp. kinabalumrir (Veldkamp) 3 J.F.Maxwell 2 Dirsac~gradis (Jack) Blume Dissochaeteae alcohol sample clausing 74 Kepong, Pahang, Malaysia el Dksocihactn intcmudia Blume Dissochaeteae alcohol sample Clausing 64 Cameron Highlands, Pahang, Malaysia 4 Dirs~haetunformatn Clawing (***)= Manolmcs Dissochaeteae alcohol samples ‘Clausing 119; ‘Sabah, Malaysia; Songkhla, Thailand’ stellulata (Jack) Bakh.f. Meyer 9623’ E Dirsochaetu s@uf& (Blume) Backer ex Clausing Dissochaeteae alcohol sample Clawing 240 Ranau, Sabah, Malaysia 2 (***)=Lhphbia ~@~lat;(Blume) Kuntze Greenhouse BG Mainz DirsotirfiUtiCosa (Brenan) Brenan & Keay Melastomeae living plant z Dtiessmia glonduligcro Stapf Sonerileae alcohol sample Clawing 221 Mt. Kinabalu N. P., Sabah, Malaysia Gr& UiFcosa H.Perrier Sonerileae alcohol sample Clawing 304 Anuiranana, Madagascar Loya spnuem Benth. ex Triana Miconieae alcohol sample Renner 237 Manaus, Amazonas, Brazil Me&illa amphh Regalado Dissochaeteae alcohol sample clausing 9 Mt. Kinabalu N. P, Sabah, Malaysia Medinilla basalturn Jum. & H.Pemer Dissochaeteae living plant Greenhouse BG Mainz Medinilla clarkei King Dissochaeteae alcohol sample Clawing 60 Cameron Highlands, Pahang, Malaysia il4 Medinilla coni$di~ Baker ex. H.Pemer Dissochaeteae alcohol sample Clawing 317 Antsiranana, Madagascar > Medhilla crasszzlk~(Reinw. ex Blume) Blume Dissochaeteae alcohol sample Clawing 6 Mt. Kinabalu N. P., Sabah, Malaysia Medinilla~qellfiaJum. & H.Perrier Dissochaeteae alcohol sample Clawing 3 16 Anuiranana, Madagascar hhtindla homoeandra (Stapf) M.P.Nayar Dissochaeteae alcohol sample Clausing 4 Mt. Kinabalu N. P., Sabah, Malaysia Medinill0 kinabbis Regalado Dissochaeteae alcohol sample Clausing 227 Mt. Kinabalu N. P., Sabah, Malaysia Medmilla kmJora Rid. Dissochaeteae alcohol sample Clawing 191a Sarawak, Malaysia Medkilla sipens Sapf Dissochaeteae alcohol sample Clausing 268 Bau, Sarawak, Malaysia Medinilla sp. Dissochaeteae alcohol sample Clausing 309 Antsiranana, Madagascar Medinilla sprciOsa (Reinw. ex Blume) Blume Dissochaeteae alcohol sample Clawing 26 Mt. Kinabalu N. P., Sabah, Malaysia Medinilla s&-phanos@ia Stapf Dissochaeteae alcohol sample Clausing 17 Mt. Kinabalu N. P., Sabah, Malaysia Medinill0 subemsa Regalado Dissochaeteae alcohol sample clausing 220 Mt. Kinabalu N. P., Sabah, Malaysia Melatoma beccarianwn Cogn. Melastomeae alcohol sample Clausing 249 Gurulau Rige, Sabah, Malaysia Mclatoma imbrinrtum Wall. ex C.B.Clarke Melastomeae alcohol sample Clawing 112 Mt. Kinabalu N. P., Sabah, Malaysia Mehtoma malabathrinrm L. Melastomeae alcohol sample Meyer 95 13 Songkhla, Thailand continued APPENDIX-confinued

Species Tribe Material Collector Collecting locality

Mehtoma ohkGuillaumin Melastomeae alcohol sample Meyer 96 17 Chantaburi, Thailand A.iclarloma pik@bwn (H.Boissieu) K.Meyer Melastomeae alcohol sample Mryer 96 19 Trat, Thailand (***)=Dicsoris pelkgnnbwn H.Boissieu Melastoma ptrnkmce Rid. Melastomeae alcohol sample Meyer 9624 Perak, Malaysia Melastoma sanguinnnn Sims Melastomeae alcohol sample Meyer 95 12 Songkhla, Thailand Mtlastoma sefthnmriUm Lour. Melastomeae alcohol sample Tu & Meyer s.n. Vinh Phu, Vietnam Mehhmu sp. Melastomeae alcohol sample Clausing 200 Mt. Penrissen, Sarawak, Malaysia Melashmu aclutinoswn Ridl. Melastomeae alcohol sample Meyer 9633- Perak, Malaysia Monolma primuhJora Hook.f. Sonerileae living plant Greenhouse BG Mainz Osbeckia aurafo H.Penier Melastomeae alcohol sample Clausing 324 Antsiranana, Madagascar Osbeclab ncpalmrrc Hook. Melastomeae herbarium specimen Meyer 9543 Pahang, Malaysia Osbeckia slclloro Buch.-Ham. ex Ker Gawl. Melastomeae alcohol sample Meyer 9602 hi,Thailand Ogspora bullah (Griff.)J.F.MaxweU Sonerileae alcohol sample Clausing 6 1 Cameron Highlands, Pahang, Malaysia Oppora e+ (Jack) J.F.Maxwell Sonerileae alcohol sample Clawing 63 Cameron Highlands, Pahang, Malaysia Pachyccnhia conrhitto (Blume) Blume Dissochaeteae alcohol sample Clausing 263 Bako N. P., Sarawak, Malaysia Po&enmh ghca Triana (***) Dissochaeteae alcohol sample Clausing 156 Bako N. P., Sarawak, Malaysia fiyltigathis elliptica Stapf Sonerileae alcohol sample Clausing 2 I2 Mt. Kinabalu N. P., Sabah, Malaysia fiyllogathis pnanlha Korth. Sonerileae alcohol sample Clausing 190 Bukit Matang, Sarawak, Malaysia Pleihinndra hookm’ Stapf Dissochaeteae alcohol sample Clausing 168 Penrissen, Sarawak, Malaysia PInnandra comtlcscnrr Jack Kibessieae alcohol sample Clausiig 175 Mt. Santubong, Sarawak, Malaysia PInnandra cogndM.P.Nayar Kibessieae alcohol sample Clawing 47 Mt. Kinabalu N. P., Sabah, Malaysia pkmandra CGhinataJack Kibessieae alcohol sample Clawing 75 Kepong, Pahang, Malaysia Ptemandra hi&& (Cogn.) M.P.Nayar Kibessieae alcohol sample Clausing 180, 162 Bukir Matang and Mt. Santubong, Sarawak, Malaysia Sarcopyramis nepalnrrir Wall. Sonerileae alcohol sample Tu & Meyer 0004 Lao Cai, Vietnam Sonnila bercariatla Cogn. Sonerileae alcohol sample Clausing 206 Mt. Penrissen, Sarawak, Malaysia Sonetih magaritocra (Lindl.) Triana Sonerileae living plants Greenhouse BG Mainz Som’h neruulosa Ridl. Sone rile a e alcohol sample Clausing 5 I Mt. Kinabalu N. P., Sabah, Malaysia Som’h obliqua Korth. onerileae alcohol sample Clawing 170 Bukit Timah, Singapore sonerila sp. Sonerileae alcohol sample Clawing 224 Mt. Kinabalu N. P., Sabah, Malaysia Sowih tmu$&~ Blume Sonerileae alcohol sample Clawing 2 17 Mt. Kinabalu N. P., Sabah, Malaysia Tzbouchina untilL?m (DC.) Cogn. Melastomeae herbarium specimen Meyer 9537 Sabah, Malaysia Trishma maun’riiznwnJ.F.Gme1. Melastomeae alcohol sample Clausing 292 Andasibe, Madagascar

(**) no formal transfer to Dissochta has been published; (***) names according to Renner d al. (in press).

Top View