Fleshy Fruits in Liliflorous Monocots

Fleshy fruits occur in several monocot orders and families, and it is generally assumed that they have been derived from capsular fruits many times during the evolution of monocot lineages. Huber hypothesized in 1969 that most capsules in Asparagales are derived secondarily from berries and that this transformation was correlated with the evolution of phytomelan-coated seeds, a pivotal character in his circumscription of Asparagales as part of reclassifying Liliaceae s.l. Dahlgren and co-workers suggested several parallel derivations and "reversals" in this character, e.g., the transformation sequence trifollicular fruits __... capsules __... berries __... capsules __... berries. Mapping of fleshy fruits on a phylogeny based on molecular characters indicates that Asparagales do not have fleshy fruits as a basal character. Dahlgren's "cyclic character evolution" hypothesis is not supported by the distribution of dry and fleshy fruits, and there is no obvious correlation between baccate fruits and phytomelaniferous seeds in Asparagales. Phytomelaniferous seeds are not an evident synapomorphy of Asparagales as presently circumscribed. The anatomy and development of different capsular and baccate fruits in selected genera are studied in an ongoing project to reveal homologies and establish an adequate fruit typology. Some observations of texture and dehiscence structures in dry and fleshy capsules and in typical berries from hypogynous and epigynous flowers are reported in this paper.


INTRODUCTION
Fruit characters have been given considerable attention in both classic and modern monocot taxonomy.Capsules have generally been considered the "basic" fruit type, from which berries, drupes, and caryopses or nuts have evolved (Dahlgren and Clifford 1982;Dahlgren et a!. 1985).Derived fruits may characterize large groups at familial or ordinal level such as Arales (fleshy indehiscent fruits) or Poales (dry indehiscent fruits).
In the liliiflorous monocots (the families that in APG II [Angiosperm Phylogeny Group 2003] constitute Asparagales, Liliales, Dioscoreales, and Pandanales) the "typical" gynoecium is a three-carpellate superior ovary with axile placentation, a single style and several ovules, but degree of fusion and occurrence of nectaries varies, and epigyny occurs in several clades (Rudall 2002).Texture of the pericarp is another highly variable character, ranging from leathery capsules and soft-walled, partially or irregularly dehiscing fruits to typical berries.Presence of fleshy fruits has been an important distinguishing character for some traditional groupings at family rank (e.g., Smilacaceae, Convallariaceae, Luzuriagaceae, Philesiaceae), and several genera have occasionally been raised to family level with special reference to this trait (e.g., Dianella, Behnia, • Petermannia).The occurrence and evolution of berries in Asparagales was given special attention by Huber and Dahlgren, who pioneered the modern research in monocot phylogeny (see below).
Information on ontogeny, anatomy, and mode of dehiscence or seed release of monocot fruits is, however, very scanty.Even the occurrence and variation of fruit types within some genera are not completely known.In an ongoing study we seek to provide basic information about the nature of fleshy monocot fruits in the broadest sense in order to test hypotheses about homology and evolution of this trait and possible correlation with other traits, e.g., seed characters.Structure and opening mechanisms of capsular fruits in polymorphic families and genera may be included to generate a broader perspective of character evolution.

Cyclic Evolution of Fruit Types in Asparagales?
Huber (1969) was the first to recognize the asparagoid families as a division at ordinal level of the very broad Liliaceae sensu Engler (1892) and Krause (1930).Huber's rather informal orders were subdivided into "Familiengruppen," arranged in ascending order according to an assessment of evolutionary status (see also Huber 1977).The first "Familiengruppe" out of eight in Asparagales listed Rhipogonaceae, Smilacaceae, Philesiaceae (with Lapageria), Luzuriagaceae (including Behnia), Petermanniaceae, Ruscaceae, Convallariaceae, Nolinaceae, Asparagaceae, and Herreriaceae.This assemblage, which includes most of the berryfruited families of Asparagales sensu Huber, was clearly positioned as "basal" in the order in his bubble diagram (Huber 1969: Fig. 35).
Dahlgren accepted Huber's arrangement of Asparagales throughout his classifications (Dahlgren 1975(Dahlgren , 1980;;G. Dahlgren 1989), although some of Huber's families were eventually transferred to other orders.The "Familiengrup-pen" were not formalized, but are to some extend reflected in the sequence of families.Dahlgren and Clifford ( 1982: 219) noted about berries in monocotyledons: "Their distribution is not random but concentrated in certain orders ... they abound amongst hygrophilous terrestrial plants growing in shady or semi-shady forests, such plants being especially concentrated in the Arales, Cyclanthales, and Taccales and in some genera of the Dioscoreales and in several families in the Asparagales" and, citing personal communication with Huber: "it is possible to interpret the capsules in most Asparagales (but not in the Liliales) as secondarily derived from berries.During the baccaceous stage of the fruit, the seeds lost their ordinary type of testa, which when the fruits became capsular again was substituted by a phytomelan crust."Dahlgren et a!. ( 1985) stated no explicit hypothesis about the basal fruit type in Asparagales, but may have inferred from Huber that berries could be the ancestral fruit type of a monophyletic Asparagales."A postulated evolutionary sequence of fruit types in the ancestors of Asparagales, in the Liliiflorae, represents an interesting cyclic process.[ ... ] The capsules of many Asparagales might be secondary to berries, which are probably, in their turn, derived from capsules with phytomelan-coated seeds.If these surmises are true, the Haemanthus berry could have been derived via the sequence trifollicular fruits --7 capsules-?berries --7 capsules --7 berries!" (Dahlgren et al. 1985: 87).They assumed that changes from capsule to berry and from berry to capsule had taken place several times and suggested that a berry-fruited stage could have been an important historical constraint leading to the evolution of phytomelan in seed coats, the most distinctive character (but not synapomorphy!) of Asparagales as originally circumscribed.

MATERIALS AND METHODS
The phylogeny used in the present investigation is based upon sequences from seven genes, two multicopy nuclear genes (ISS and 26S rDNA [ribosomal DNA]), four plastid genes (atpB, matK, ndhF, and rbcL), and a single mitochondrial gene (atp l ), and is identical to the phylogeny presented by Chase et al. (2006).The analysis includes 141 taxa, 16 of which are dicot outgroups (Quiet al. 1999).The remaining 125 taxa are all monocots and placeholders for nearly all families recognized in the APG II (2003), except a few small families in Alismatales, which were not included.However, due to missing data, three families (Hydatellaceae and the achlorophyllous Corsiaceae and Triuridaceae) were also excluded from the final analysis.Detailed information about primers, accession data, GenBank accession numbers, etc. will be published elsewhere.The analyses were done in PAUP* vers.4.0bl0 (Swofford 2001).To avoid using extensive swapping on suboptimal tree islands, 500 random input replicates were used saving only 20 trees per replicate, using only SPR (subtree-pruning-regrafting) swapping.The trees found in this manner were subjected to TBR (treebisection-reconnection) swapping to find all trees.However, this did not yield further trees (length = 62,873; consistency index = 0.54; retention index = 0.48).The tree shown here (Fig. 1) is a simplified strict consensus tree of these trees.Branch support was calculated using bootstrapping, and the values and a detailed discussion of the topology may be found in Chase et al. (2006).
The occurrence of fleshy fruits, including succulent but dehiscing capsules, as well as typical berries (in the sense of Roth 1977 andJudd 1985) was mapped on this phylogeny (Fig. l) together with phytomelaniferous seeds, i.e., seeds with a phytomelan crust formed in the outer integument (Huber 1969;Danilova eta!. 1995;Graven eta!. 1998), and position of the ovary.A formal optimization procedure is not feasible at the family level; as most terminals are polymorphic for the characters mapped, this would leave most of the internal nodes with ambiguous states.Source data for the mapping (Appendix l) are mainly obtained from Huber (1969), treatments in Kubitzki (l998a), and direct observations of live or conserved material.
The classification follows APG II ( 2003), but without the optional broad circumscriptions of Alliaceae, Asparagaceae, and Xanthorrhoeaceae in agreement with Pires et a!. (2006) in order to facilitate comparison.Hesperocallidaceae are here included in Agavaceae.We have retained the APG II name "Asparagales," although it is antedated by Orchidales and Iridales.The liliiflorous monocots ("lilioid monocots" of Rudall 2002Rudall , 2003) ) are a grade comprising the orders Petrosaviales, Pandanales, Dioscoreales, Liliales, and Asparagales.The commelinid clade, which has been found to be sister to Asparagales (Chase et al. 1995a(Chase et al. , b, 2000;;APG II 2003), is usually not covered by this term, which largely corresponds to superorder Lilianae in Dahlgren and Clifford ( 1982) or Liliiflorae in Dahlgren et al. (1985) without most of Pandanales.
Observations on selected dry and fleshy liliiflorous fruits were made on material collected in nature or from living specimens in cultivation (Table 1).Standard histological techniques were applied according to Ruzin (1999).Fruits for sectioning in Paraplast® were fixed in FAA and 10 f.lm thick transverse sections were stained in Johansen's Safranin!Fast Green (S/FG).Fruits for sectioning in glycolmethacrylate were fixed in a mixture of 2% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M phosphate buffer, and 3 f.lmthick transverse sections were stained in Toluidine Blue 0 (TB) or with the Periodic Acid Schiff reaction followed by Aniline Blue Black ("Amido Black") (PAS/ABB).Observations were concentrated on the upper part of the fruits.Microscopy images were digitized directly from the Leica DM RXA 2 microscope using a Hamamatsu C-2400 (ORCA II) camera.

Distribution of Fleshy Fruits and Phytomelaniferous Seeds
Fleshy fruits are found in 24 of the 39 liliiflorous (including Pandanales and Petrosaviales) terminals of the phylogeny (Fig. 1 ).Only five of these (all in Liliales) are not polymorphic.Fleshy fruits are found in all terminals of Liliales except Campynemataceae, and in all of these there are some species with typical berries.In Asparagales, fleshy fruits occur in half of the terminals, and only in seven out of these 12 are true berries recorded.Berries are formed in epigynous flowers in six terminals, scattered in three orders.
gales, e.g., in Zingiberales, need verification.Six terminals are polymorphic for this character, and polymorphism exists also at the generic level, e.g., in Leucojum (Amaryllidaceae).Five of the 24 terminals of Asparagales do not have this trait.
Comparison of character distribution data from different sources is complicated by the fluctuating taxonomy of liliiflorous monocots.Luzuriagaceae (Liliales) for example, are always characterized by baccate fruits, but Huber (1969) and Dahlgren et a!. ( 1985) referred this family to Asparagales and included genera here referred to as Agavaceae (Behnia), Laxmanniaceae (Eustrephus), and Hemerocallidaceae (Geitonoplesium).Appendix 1 of the present paper contains an annotated list of families with records of fleshy fruits appearing as terminals in Fig. l, with notes on ovary position and phytomelaniferous seeds.

Fruit Structure
The pericarp of all fruits studied here is composed of an outer (abaxial) epidermis, an inner (adaxial) epidermis, and between them the mesocarp, which consists of ground tissue, mostly parenchyma, and vascular tissue.In most of the species studied so far the outer epidermis has cells with rather thin primary walls having a thin cuticle.In hypogynous flowers the peri carp derives from the carpels only, while in epigynous flowers it may be construed as also including extracarpellary tissues.The fruits are three-carpellate, forming three separate loculi and a central column where the septa meet.Generally, five distinct vascular strands composed of one or more vascular bundles can be identified in each carpel.One single median (dorsal) strand in the middle of the carpel, two marginal (ventral) strands in the septum near or in the central column, and two lateral strands in the pericarp wall where two neighbor carpels fuse, forming the septum.Both the lateral and the marginal strands from neighboring carpels often fuse and form a single strand.Moreover, several more or less prominent vascular bundles can be observed.In epigynous flowers the vascularization is more complex and in Galanthus nivalis three vascular bundles are present at the median and lateral positions (Fig. 2).The two outer bundles of the median and the lateral strands continue into the outer tepa! and stamen or the inner tepa! and stamen, respectively.The inner vascular bundle of the median strand continues into the style while the corresponding bundle of the lateral strand diminishes and ends.
The dehiscing fruits in the taxa studied here are loculicidal and open distally.The carpel walls split through the median vascular bundles and in the central column where the septa meet, resulting in three valves with the placenta on a longitudinal rim in the middle (Fig. 3, 4).Some of the capsular fruits studied deviate from a typical capsule in having a fleshy mesocarp when dehiscing (Fig. 4), while berries are indehiscent with a fleshy mesocarp at maturity.Dry capsules.-Hyacinthoidesnon-scripta, Ornithogalum narbonense, Leucojum aestivum, and Narcissus sp.represent taxa with dry capsules.In Hyacinthoides Fabr., the median bundle lies in a narrow zone of the pericarp and the xylem forms a wide V-shaped wedge with protoxylem at the base and two groups of metaxylem separated by parenchyma inside the "V."The phloem outside the metaxylem is also located in two groups separated by parenchyma (Fig. 5).At maturity the cells of the inner epidermis have thick, lignified Cultivated ornamental** outer cell walls except for a group of thin-walled, non-lignified cells, small in transection facing the median vascular bundle (Fig. 6).In Ornithogalum narbonense the xylem in the median vascular bundle forms a "V" with parenchyma in the wedge (Fig. 7) as in Hyacinthoides.Large, thin-walled sclerenchyma cells are present both outside the xylem towards the inner epidermis and outside the phloem towards the outer epidermis.The cells of the inner epidermis have lignified secondary cell walls with U-shaped thickenings (seen transversally, Fig. 7) like the cell walls of the inner epidermis in Hyacinthoides, but in 0. narbonense also the small cells facing the median bundle have U-shaped thickenings.In Narcissus as in 0. narbonense the inner epidermis has lignified and U-shaped secondary cell walls (Fig. 8, 9).In addition, below the inner epidermal cells inside the median vascular bundle is a lignified hypodermal layer in contact with the lignified bundle sheath cells outside the xylem part of the vascular bundle (Fig. 8).
With progressing dehydration at maturity the outer epidermis and the mesocarp shrink while most of the inner epidermal cells maintain their shape due to the thick, lignified cell walls, a fact important for the dehiscence process.In Hyacinthoides the small thin-walled cells in the inner epidermis rupture as do the parenchyma cells and the protoxylem, and the median vascular bundle breaks up along a median line with protoxylem and thin-walled parenchyma cells, splitting the phloem and the metaxylem in half (Fig. 6, 10).The final split of the carpels takes place in the outer epidermis where the cells form a slight indentation towards the vascular bundle.In 0. narbonense the median vascular bundle breaks up as in Hyacinthoides and later the thinwalled sclerenchyma cells separate like the small inner epidermal cells with the U-shaped lignified wall (Fig. 11).
The central column in the mature capsule is separated into three parts where the septa meet.In Leucojum the inner epidermal cells have prominent U-shaped secondary cell walls, but they are replaced by smaller unlignified cells near the central column.The inner epidermis bursts here and the column splits more or less irregularly through the continuation of the stylar canal in this species (Fig. 12) as in 0. narbonense and Narcissus.Groups of thin-walled sclerenchyma cells occur in the central column of the species studied.inner epidermi s in 0. nutans have relati vely thi n and o nly slig htly li gni fied , U-s haped secondary wall thi cke nings (Fig. 13), and as in 0 .narbon.ense,cell s fac ing the medi an vasc ul ar bundle are li gni fied (F ig.14).The thi ck, matu re mesocarp consists of large, vacuolated cells except fo r some thin-walled sclerenchy ma cell s abutting the medi an vasc ul ar bundl e in a narrow region of the pericarp.At dehi scence, the median vasc ular bundle , w hich has a very fl at and w ide V-shaped xyl em area (Fig. 14) , splits media ll y along parenc hy ma cell s as in Hyacin.thoidesand 0 .narbonense.The centra l column has sma ll groups of thin-walled sclerenchyma cell s an d splits as in 0 .narbonense.
In the median strand of Galanthus L. frui ts th e inner vascul ar bund le represents the carpe l bund le.In thi s bund le th e xy le m fo rms a " V " as the metaxylem is di vided by a row of parenchy ma cells (Fig .15), it is thus very simil ar to the medi an vascul ar bundle seen in dry capsul es, while no special structure was o bser ved in the o uter vascular bundles.The inner epidermis of the Ga.lan.th.us pericarp consists of cell s w ith primary wall s onl y, the o uter epide rmi s has thin primary cell walls with a di stinct c uticle, and towards maturi ty the shrinking mesocarp cell s and large intercellul ar s paces for m the spo ngy ti ssue of thi s fruit (Fig .16).At dehiscence the septa split in the continuation of the stylar canal.
Typica l berries.-In the hypogyno us Ma ianth.emumracemosum and epigyno us H aemanthus a.lbifios the inner epidermi s has thin pr imary cell walls, the outer epidermi s in both species has thicker cell wall s and has a di stinct, th in cuticle in Haemanthus and a moderately thick cuticle with cuticular ridges in Maianthemum F. H. Wigg.The mesocarp is composed of large thin-walled and vacuolated cells and small, rather few vascular strands (Fig. 17).In Maianthemum only the median vascular bundles are seen in cross section (Fig. 18); in Haemanthus more strands are visible.The central column consists of homogenous parenchymatous tissue, in Maianthemum with small1ateral vascular bundles, in Haemanthus with relatively larger bundles.No lignified cells were observed in the mesocarp and inner epidermis.The structure of the pericarp of M. racemosum resembles the description by Kraus (1949) for M. bifolium L.
The range of variation in the pericarp of the berries is insufficiently known at this stage of the investigation, but observations on Paris quadrifolia L. (Melanthiaceae, Liliales) revealed a very thin pericarp composed of regular cell layers with thin cell walls, large vacuoles and large and irregular intercellular spaces, especially in the septa.Only very small median and marginal vascular bundles were visible in fruits of this species.The outer epidermis has thick outer periclinal walls and a cuticle with distinct cuticular ridges.

Fruit and Seed Character Phylogeny
The traditional notion of the dry dehiscent fruit as symplesiomorphic in the liliiflorous orders is not manifested by the character map presented here (Fig. 1).Fleshy fruits are frequent in Alismatales, and the fruits of Acorus L. are reportedly berry-like (Thompson 2000).Dahlgren et al. (1985) discussed the possibility of trilocular capsules being the oldest fruit type in monocotyledons as a whole.However, the chosen outgroups in phylogenetic analyses of all monocots exhibit a range of fruit types and there is as yet no stable hypothesis of what constitute the sister group to all monocots and which fruit types may occur in it.Givnish et al. (2006) inferred that fleshy fruits (using a less inclusive definition, not counting Acorus) evolved independently 21 times and were lost 11 times in an optimization of this character on a tree of 282 exemplars representing 74 monocot families.
Fleshy fruits seem to have evolved at least twice in Dioscoreales in the family Dioscoreaceae (Caddick et al. 2002a, b).In Liliales, berry-like fruits occur in most families and may be a synapomorphy above the family level, e.g., uniting Smilacaceae, Philesiaceae, and Rhipogonaceae.The baccate fruits of Liliaceae seem to arise phylogenetically late within the family (Rudall et al. 2000b) complicating an optimization with an early transition to berries in the Liliaceae-Smilacaceae clade.Berry-like fruits may be plesiomorphic in Melanthiaceae.In the genus Trillium capsules seem to be derived from berries in connection with a shift in dispersal strategy towards myrmecochory (Kawano and Kato 1995).The latter is one of the few possible cases of "secondary capsules" as envisioned by Dahlgren et al. (1985).
It is clear from the phylogeny shown in Fig. 1 that Asparagales as now understood are not derived from a group of plants with baccate fruits.The basic asparagalean families of Huber ( 1969) and Dahlgren et al. (1985) are now referred to Liliales (Rhipogonaceae, Smilacaceae, Philesiaceae-Lu-zuriagaceae, and Petermanniaceae) or to the higher Asparagales (Behniaceae, here included in Agavaceae), Ruscaceae s.l.(including Convallariaceae, Dracaenaceae, and Nolinaceae), Asparagaceae, and Herreriaceae (here included in Agavaceae).The largest conceivable clade with synapomorphic fleshy fruits may be the Laxmanniaceae-Asparagaceae branch, but it would require reversals to dry fruits in all three families.In a morphological-molecular phylogenetic analysis of 26 genera of Ruscaceae s.l., Asparagaceae, and Agavaceae (Rudall et al. 2000a) it was hypothesized that berry-like fruits have evolved independently in Ruscaceae and Asparagaceae, the dry fruits of the No/ina-group in Ruscaceae being a symplesiomorphy.The topology of the tree with terminals from Ruscaceae in Givnish et al. (2006) differs from Rudall et al. (2000a) and suggests that the dry fruits of Nolina Michx.may be apomorphic in the family.Densely sampled phylogenies with several representatives from all genera are needed to form plausible hypotheses of character polarity and levels of synapomorphy in the families of this order.
Berries or berry-like fruits are formed in epigynous flowers in six families scattered in the three orders Dioscoreales, Liliales, and Asparagales.Ovary position seems not to be correlated with fruit texture, which mainly develops in the latest of the three developmental phases (initiation-differential growth-tissue differentiation) recognized by Endress (1994).
It is assumed that fleshy fruits, in particular true berries, may evolve as an adaptation to hygrophilous habitats such as forest understory.Correlation with other characters, such as rhizomatous growth habit and net-veined leaves, was discussed by Patterson and Givnish (2002) and Givnish et al. (2006).
The frequent occurrence of phytomelaniferous seeds in Asparagales was conjectured to be associated with character transformation from berry-like fruits to "secondary capsules" by Huber (1969) and Dahlgren and co-workers (Dahlgren and Clifford 1982;Dahlgren and Rasmussen 1983;Dahlgren et al. 1985), the crust being a protection of the seeds not needed in a berry.The change in the view of monocot evolution generated by analyses of molecular data has eliminated the basal "baccaceous stage" of Huber (pers. comm. cited in Dahlgren and Clifford 1982: 220), which was the basis for this hypothesis.Phytomelaniferous seeds were not considered a synapomorphy of all Asparagales by Huber, but with the exclusion of some of Huber's families from the order it became a possible basal trait in Dahlgren's classifications.It is often implied to be a synapomorphy of Asparagales in systematic textbooks and surveys (Kubitzki et al. 1998;Judd et al. 1999;Bremer et al. 2003).However, a comprehensive reconstruction of the evolution of this character has not been attempted and this paper may be the first to address the question.Dahlgren and Rasmussen (1983) refrained from a diagrammatic consideration of character evolution of the "Asparagales complex" because of uncertainty about the monophy1y of the families and the excessive homoplasy that would have to be accepted in any hypothesis.
Phytomelaniferous seeds are not a likely synapomorphy of Asparagales in the present phylogeny if Orchidaceae are included in the order.Even if Orchidaceae are excluded, this character will be highly homoplastic.Meerow et al. ( 1999) regarded phytomelan-crusted seeds as an ancestral state of Asparagales and inferred that this trait has been lost five times in Amaryllidaceae.It seems very difficult to estimate the extent of synapomorphy of this trait considering the present uncertainty about the basal branching of Asparagales and the still incomplete knowledge of phylogeny within many families.Phytomelan-crusted seeds are a frequent and distinctive character of Asparagales, but it should not be emphasized as an indisputable synapomorphy of the entire order.
Confirming the occurrence of phytomelaniferous seeds requires anatomical studies of different ontogenetic stages, and the information concerning distribution of this character comes mostly from the comprehensive survey by Huber ( 1969).Observations of crustose dark seeds in families where phytomelaniferous seeds have not been recorded should be investigated carefully.The only detailed investigation of the formation and chemistry of phytomelan in Asparagales using modern analytical tools are the studies of Gasteria verrucosa (Mill.)H. Duval (Asphodelaceae) by Wittich andGraven (1995, 1998) and Graven et al. (1998).A more thorough understanding of this trait and if possible recognition of more states than just presence or absence is highly desirable.

Evolution of Fruit Structure
Whether dry or fleshy, or evolved from hypogynous or epigynous flowers, the dehiscent fruits examined here are remarkably similar in details of the median vascular bundle in the carpel and the inner epidermis.They possess the Vshaped xylem with a middle strand of parenchyma cells, and the cells of the inner epidermis have lignified, U-shaped secondary cell walls except in Galanthus.These traits are supposed to play a role in the regular opening of the dry capsules.The inner epidermal cells facing the median vascular bundle vary slightly in wall thickness and the amount of lignification of the outer wall; a lignified hypodermis was observed in Narcissus.The parenchymatous strand through the V-shaped xylem wedge resembles the loculicidal dehiscence line in Lilium L. depicted by Roth (1977) and may be homologous.
Von Guttenberg (1971) observed lignification of the inner epidermis in several different lilioid genera and regarded this, in addition to the shape and orientation of the epidermal cells, to be part of the opening mechanism.Also Tilton and Horner (1983) mentioned thickened outer tangential cell walls of the inner epidermis in Ornithogalum caudatum Jacq.The rather complex vascularization of carpels and ovules in several liliiflorous fruits, especially in species of Melanthiaceae and Colchicaceae, was described in a series of papers by Sterling (1980 and papers cited therein), Utech and Kawano (1981), and Utech (1986 and papers cited therein), but information about structure of the median vascular strand, the inner epidermis, or other structural traits of relevance for opening mechanisms was not provided.
The fruits of species with fleshy capsules lie on the ground when ripe, the peduncle decays and the fruits are often completely detached at dehiscence.In Ornithogalum nutans the inner epidermis of the ripe fleshy capsule has distinct cell wall thickenings, although less pronounced than in species with dry capsules, indicating an opening mechanism similar to the one in the dry capsules.However, Galanthus nivalis lacks cell wall thickenings in the inner epidermis and may dehisce in a different way.Galanthus capsules have been observed to open at high humidity (Buxbaum 1934;Frederiksen pers. obs.).
The fleshy, dehiscent fruits of Ornithogalum nutans and Galanthus are most likely derived from dry capsules similar to the ones found in other species of Ornithogalum, or in Leucojum and Narcissus, respectively.The seeds in Galanthus and some species of Leucojum have prominent elaiosomes.This and the procumbent nature of the fruits are no doubt adaptations to myrmecochory (Buxbaum 1934).In Trillium (Melanthiaceae) capsules are hypothesized to have evolved from typical berries in connection with a change from endozoochory to myrmecochory (Kawano and Kato 1995;Farmer and Schilling 2002).
The berries examined in this study differ mainly from the capsules in the extent of the mesocarp, and they do not have lignified secondary cell wall thickenings of the inner epidermis, or xylem with a preformed parenchymatous line of dehiscence in the median vascular strand of the carpel.There is relatively little structural variation between the two berries examined, and no significant difference was observed between berries formed in the hypogynous and epigynous flower.In preliminary observations of other liliiflorous berries we have found different ways of attaining the succulence of the fruits and various mechanisms for partial opening.However, it is still uncertain to what extent polarity of character evolution between dry or succulent and dehiscent or indehiscent fruits can be inferred from structural studies.
The characters observed in this study must be evaluated in a context of historical ecology.The history of fruits is a reflection of the history of changes in life strategy and invasion of ecological niches and may be expected to show certain correlations to or form patterns with other functional characters, a phenomenon referred to as concerted convergence (Givnish 1997;Givnish and Sytsma 1997;Patterson and Givnish 2002;Givnish et al. 2006) or iterative evolutionary themes (Rudall 2003).The importance of phylogenetic approaches to studies of plant reproductive systems has gained increasing recognition within the last decade (Endress 1994;Silverton et al. 1997;Weller and Sakai 1999), and it is expected that new insight in causes and directions of evolution can be obtained from integrated studies of phylogenetics, structural development, and ecology.
str. (fl II Ruscaceae (') ...__ ___ Laxmanniaceae (fl(~)..------Agavaceae (fl (~) Fig. I.-Mapping of fleshy fruits (berry symbol) with indication of ovary position (epigynous or hypogynous perianth remnants) and phytomelaniferous seeds (black seed symbol) on a phylogeny of liliiflorous monocots (simplified strict consensus of the trees found byChase et al. 2006).Brackets indicate that a character is found in some but not all species of a terminal.Absence of fruit symbol indicates that fleshy fruits do not occur in this family.Absence of a seed symbol indicates that phytomelaniferous seeds do not occur or have not been detected in this family.See Appendix I for notes on occurrences.

Table 1 .
Material used in anatomical studies.* = Material cultivated in the botanical garden, University of Copenhagen.** = Material from private gardens.Vouchers are kept at the Botanical Laboratory, University of Copenhagen.