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

This Article Abstract Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (14) Citing Articles via Google Scholar Google Scholar Articles by Tsou, C.-H. Search for Related Content PubMed Articles by Tsou, C.-H. Agricola Articles by Tsou, C.-H.

Early floral development of Camellioideae (Theaceae)¹

^a Institute of Botany, Academia Sinica, Nankang, Taipei, Taiwan 115, Republic of China

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
The early floral development of Camellioideae was studied. Two major evolutionary lineages were recognized for this subfamily. The earlier evolved lineage (Camellia, Polyspora, and Pyrenaria) has normally 11–14 perianth members, which are initiated in a continuous spiral and are differentiated into sepals and petals at late floral development, and numerous stamens initiated individually and centrifugally on the whole androecial region. The later derived lineage (Franklinia, Hartia, Schima, and Stewartia) has five sepals and five petals arranged in two whorls, and numerous individual stamens originating centrifugally from the five petal-opposed zones. Hartia-Stewartia and Franklinia-Schima further diverged as two branches — the former is characterized by having androecial fascicles and axile-basal placentation. The androecial fascicle is considered to be derived within this subfamily. The latter exhibits a higher degree of carpellary congenital fusion and axile-central placentaion, and as a whole, is concluded to be the most advanced group in the Camellioideae. A taxonomic treatment of the Camellioideae at the tribal level is also proposed.

Key Words: Camellieae • Camellioideae • floral ontogeny • Gordonieae • phylogeny • Stewartiinae • Theaceae

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
This is a part of a series of studies on the Theaceae, which hopes to improve our understanding of the features and the phylogeny of this family (Tsou, 1995, 1997). The flowers of Camellioideae are rather uniform in gross morphology. In general, they are solitary, medium to large, showy, actinomorphic, and with 10–14 sepals and petals, numerous exserted stamens, and a superior ovary. Such a character combination is crucial in defining the Camellioideae. Nonetheless, the high degree of uniformity of the camellioid flowers makes the floral characters much less useful in the taxonomy at tribal and generic levels. Earlier authors, such as Melchior (1925, 1964) and Airy-Shaw (1936), used a single floral character in their tribal assignments, i.e., whether the sepals and petals gradate abruptly or not, whereas later authors, such as Sealy (1958) and Keng (1962), considered this criterion untenable and used only fruit and seed characters in their keys to the tribes, subtribes, or genera of the Camellioideae.

In the last decade, floral development has come to be indispensable for an integrated study of any higher rank taxon (Leins, Tucker, and Endress, 1988; Endress, 1994; Tucker, 1996). In the Camellioideae, Erbar (1986) studied Stewartia pseudocamellia, and Sugiyama (1991) examined Camellia japonica. In S. pseudocamellia (Erbar, 1986), the five sepals appear in a whorl as do the five petals. The androecium and gynoecium both originate from a deeply sunken floral apex; the androecium first develops as a five-fascicle cluster of primordia, then individual stamen primordia appear on its surface. The gynoecium is five-carpellate, and the placentation is axial-basal. In C. japonica (Sugiyama, 1991), the 13–18 perianth primordia are initiated in a single spiral, with sepals and petals indistinguishable at early developmental stages. The floral apex is slightly concave. No androecial fascicles are initiated, and the individual stamens appear directly on the surface of the androecial region in a spiral-centrifugal order. The gynoecium consists of three carpels, and the placentation is axile-central. In short, for the early floral development of all four floral categories, S. pseudocamellia and C. japonica differ greatly in their aestivation, structure of androecium, number of perianth members, and their placentation.

No one has seriously addressed the evolutionary nature of the uniformity of mature camellioid flowers. Did the flowers of this subfamily evolve so slowly that high homogeneity has been maintained? Or, is this similarity merely superficial and thus represents a convergence? Erbar's (1986) and Sugiyama's (1991) ontogenetic works, when considered together, have already indicated that the great similarity between the flowers of Camellia and Stewartia is indeed superficial and that very different patterns of early floral development have evolved in the Camellioideae.

Because the floral ontogeny of only two genera in the Camellioideae have been investigated and because early floral development seems to contain valuable information about the evolution of the Camellioideae, a subject poorly understood and a topic rarely discussed, I undertook the present study to bridge the wide gaps in our knowledge of floral development in the Camellioideae and to discover more useful characters for the phylogenetic reconstruction of this subfamily. Among the ten camellioid genera currently recognized by me (Apterosperma, Camellia s.l., Franklinia, Gordonia s.s., Hartia, Polyspora, Pyrenaria s.l., Schima, Sinopyrenaria, and Stewartia), the floral ontogeny of seven genera was examined using scanning electron microscopy. Apterosperma, Gordonia s.s., and Sinopyrenaria are not documented due to the lack of very young material; nevertheless, their small/medium buds and flowers were available and examined under a dissecting microscope for a more comprehensive analysis.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Floral buds of different sizes and open flowers of 12 species from ten camellioid genera were collected from various countries (Table 1). They were fixed in FAA (5 parts formalin:5 parts acetic acid:90 parts 70% ethyl alcohol) immediately after detachment. After being fixed for at least 48 h, plant materials were transferred to 50% ethyl alcohol. Very young floral buds of nine species from seven genera (Table 1) were carefully dissected under a dissecting microscope and ultrasonically cleaned in 50% ethyl alcohol, dehydrated through an ethyl alcohol:acetone series, and then dried with a Hitachi HCP-2 critical point dryer. Pieces of material were mounted on aluminum stubs with Scotch double adhesive tape, coated with gold in a Hitachi IB-2 ion coater, and then observed with a Zeiss-950 scanning electron microscope. The remaining three samples with small and medium but no very young floral buds were dissected and observed under a dissecting microscope.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Two major genus groups can be recognized on the basis of perianth phyllotaxy and the timing of differentiation of sepals and petals. In Group I (Camellia, Polyspora, and Pyrenaria), the 11–16 perianth members are initiated quincuncially and differentiate into sepals or petals only at late floral development. Distinctions among these three genera are rather limited, and subgrouping appears unnecessary. In Group II (Hartia, Stewartia, Franklinia, and Schima), the calyx and corolla are distinct at inception, and the five sepals and five petals are each arranged in one whorl though the members of each whorl are of spiral initiation. Within Group II, two subgroups are recognizable on the basis of the very different patterns of androecial and gynoecial development. In Group IIa (Hartia and Stewartia), individual stamens are initiated from five androecial fascicles and the placentation is axile-basal, whereas in Group IIb (Franklinia and Schima), no androecial fascicles are formed and the placentation is axile-central.

Because of the high similarity in floral development between or among the members of the same (sub-)group, I take the (sub-)groups as the units of description so as to avoid unnecessary repetition.

Group I: Camellia, Polyspora, and Pyrenaria
Perianth initiation
The 11–16 perianth primordia are initiated in a spiral sequence (Figs. 1, 14, 22, 23, 34), with a parastichy number of most likely two or three. The young perianth members are homogeneous in appearance and their bases are basically triangular (Figs. 14, 23, 34). In Polyspora axillaris and Pyrenaria shinkoensis, the young perianth members remain basally triangular for a period of time, and the circumference of floral buds is thus multiangular (Figs. 23, 25, 34, 37), whereas the bases of young perianth members of Camellia soon become crescent shaped and the floral buds appear rounded in outline (Figs. 1, 14). During late floral development, the outer perianth members remain small, stout, and brownish, but the inner five to seven members become large, flattened, and light-colored. However, this morphological transition is gradual, and there is no clear-cut demarcation between sepals and petals in fully developed flowers.

View larger version (176K): [in this window] [in a new window]   Figs. 1–12. Floral development of Camellia hengchunensis and C. tenuifolia . Figs. 1–8 . C. hengchunensis . 1. Floral axis after the last perianth member is just initiated, showing the concave floral apex, round circumference of the axis, and spiral arrangement of perianth members. Scale bar = 200 µm. 2. Floral apex showing the initiation of earlier stamens (S) on the ring-like androecial primordium and individual carpels (C). Scale bar = 50 µm. 3. Initiation of stamens of the second whorl. Carpel primordia appear as rounded mounds. Scale bar = 100 µm. 4. Nearly all the stamen primordia have emerged. Carpel primordia are expanded more in lateral directions. Scale bar = 100 µm. 5. All stamens are initiated. Young carpels have already fused and assumed an ascidiate form. Note the unusual opposite arrangement between stamens of inner and outer whorls (arrows). Scale bar = 100 µm. 6. Three carpels are highly fused. A trifid fissure is left open at the apex. Scale bar = 100 µm. 7. An occasional four-carpellate gynoecium. Scale bar = 100 µm. 8. Young stigma (Sg) composed of three branchlets arranged in bilateral symmetry. The one on the left is shorter and slightly outcurved. Scale bar = 200 µm. Figs. 9–12 . C. tenuifolia . 9. Floral bud of a similar stage to that of C. hengchunensis in Fig. 5 ; they are highly similar in floral structure. Scale bar = 100 µm. 10. Floral bud of a similar stage to that of C. hengchunensis in Fig. 6 ; they are still not distinguishable. Scale bar = 200 µm. 11. Young gynoecium. Note the trichomes on the ovary surface and the unfused carpellary apexes. Scale bar = 200 µm. 12. Young stigma with three identical stigmatic branchlets arranged in radial symmetry. Scale bar = 200 µm.

View larger version (228K): [in this window] [in a new window]   Figs. 13–21. Floral development of Camellia sinensis . 13. Floral apex showing initiation of earliest stamens (*) from the ring-like androecial primordium and three carpel primordia (arrows). Scale bar = 50 µm. 14. Top view of a floral bud. Note the round circumference and spirally arranged perianth members (1–12). Scale bar = 200 µm. 15. Higher magnification of Fig. 14 , showing the initiation of stamens (S) of the second whorl and three carpels (C). Scale bar = 100 µm. 16. Initiation of stamens of the third whorl, three carpels are laterally expanded but not yet fused. Scale bar = 100 µm. 17. Initiation of stamens about to finish. Carpels are fused laterally at the lower part. Scale bar = 200 µm. 18. Higher magnification of Fig. 17 , showing the ascidiate carpels. Scale bar = 100 µm. 19. Stage slightly after that in Figs. 17 and 18 , where carpels are much elongated individually. Scale bar = 200 µm. 20. Floral bud with front stamens removed showing the horizontal bottom line of gynoecium and androecium. Scale bar = 200 µm. 21. Gynoecium from a medium bud showing the hairy ovary and the three stigmatic branchlets bent 90°. Scale bar = 400 µm.

View larger version (181K): [in this window] [in a new window]   Figs. 22–33. Floral development of Polyspora axillaris . 22. Floral axis after the perianth initiation is just completed, showing the concave floral apex and the clockwise spiral of perianth phyllotaxy. Scale bar = 200 µm. 23. The same floral axis as in Fig. 22 , showing the angular circumference of floral axis and the triangular bases of perianth members. Scale bar = 200 µm. 24. Floral apex showing initiation of earlier stamens (S) on the ring-like androecial primordium. Carpel primordia are not yet visible. Scale bar = 50 µm. 25. Floral apex with young carpels and stamens. Note the triangular bases of most perianth members. Scale bar = 200 µm. 26. High magnification of Fig. 25 showing the initiation of five carpel primordia (C) and stamens (S) of the second whorl. Scale bar = 100 µm. 27. Initiation of stamens of the second and the third whorls. The five carpel primordia (C) appear as round mounds. Scale bar = 100 µm. 28. Young carpels not yet fused but already appearing ascidiate. Scale bar = 40 µm. 29. Young carpels just slightly fused at the base. The carpellary chamber is distinct (arrow). Scale bar = 40 µm. 30. High magnification of three fused carpels showing their carpellary chambers (arrows). Scale bar = 40 µm. 31. Young gynoecium showing the closing of carpellary chambers (arrows). Scale bar = 100 µm. 32. Older floral bud. Its outermost stamens have just originated whereas inner stamens have already differentiated into anthers and filaments, and the gynoecium is well developed. Scale bar = 200 µm. 33. Longitudinal view of a dissected floral bud showing the superior position of the ovary (O) and young stamens (S). Scale bar = 200 µm.

View larger version (233K): [in this window] [in a new window]   Figs. 34–42. Floral development of Pyrenaria shinkoensis . 34. Floral axis after the perianth initiation is just completed, showing the concave floral apex, triangular bases of most perianth members, and the counterclockwise spiral of perianth phyllotaxy. Scale bar = 200 µm. 35. Floral apex showing the initiation of earlier stamens (S) on the ring-like androecial primordium and a carpel primordium (*).Scale bar = 40 µm. 36. Initiation of stamens (S) of the second whorl and early development of carpels (C). Scale bar = 100 µm. 37. Floral axis showing the clockwise spiral of perianth phyllotaxy, triangular bases of perianth members, and only two carpel primordia. Scale bar = 200 µm. 38. Floral apex showing the initiation of stamens (S) of the second and third whorls and three separated carpel primordia (C). Scale bar = 100 µm. 39. Three young carpels (C) of different sizes, showing the continuation of phyllotaxy from perianth members to carpels. Scale bar = 200 µm. 40. Three young carpels well fused and assuming an ascidiate form. Scale bar = 100 µm. 41. Nearly all stamens initiated. The androecium and gynoecium are slightly deformed due to the pressure of the innermost perianth members. Scale bar = 200 µm. 42. Floral bud with partial stamens removed, showing the horizontal surface of the receptacle. Scale bar = 200 µm.

There are a total of 35–40 stamen primordia arranged in two whorls in Camellia hengchunensis and C. tenuifolia (Figs. 6, 10), ~100 in three whorls in C. sinensis (Fig. 17), ~300 in four (–five) whorls in Polyspora axillaris (Fig. 32), and 150–190 stamens in three (–four) whorls in Pyrenaria shinkoensis (Fig. 41). The just-emerged stamen primordium ranges from 40 to 45 µm in diameter in these three genera (Figs. 3, 15, 25, 36). The androecial region is finally fully occupied with stamen primordia. At the beginning of stamen initiation, the androecial ring primordium of the five species studied is similar in size, but it differs considerably by the time stamen initiation ceases (Table 2). Therefore, the total number of stamen primordia a floral bud would finally contain is correlated with the outward expansion of the ring primordium by the time stamen initiation is complete.

In samples of Camellia and Pyrenaria, the three carpel primordia in a gynoecium usually originate sequentially but unrelated to the phyllotaxis of stamens and perianth members; nevertheless, in a few buds of Pyrenaria shinkoensis, the initiation sequence of the carpels resumed the two-fifths spiral sequence finished by the perianth members (Fig. 39). In samples of Polyspora, the initiation sequence of the five carpels in a gynoecium cannot be ascertained due to the very weak size difference among the carpel primordia (Figs. 25–27).

Though the early gynoecial development of the five samples examined is similar, as summarized above, differences do exist among the three genera. First, Polyspora axillaris and Pyrenaria shinkoensis have a moderately long style ending with a capitate stigma, whereas the three Camellia species examined have a short to very short style ending with a long and branching stigma (Figs. 8, 12, 21). In Camellia, each stigmatic branch has its own pollen-transmitting duct, and all of the pollen-transmitting ducts are confluent in the stylar region. The three stigmatic branches of C. hengchunensis differ in length and are of bilateral symmetry (Fig. 8). Second, the five carpels of a P. axillaris flower typically do not initiate until the stamens of the second whorl start to emerge (Figs. 24–26), whereas in the four samples from the other two genera carpel primordia initiate synchronously with the stamen primordia of the first whorl (Figs. 2, 13, 35).

Group IIa: Hartia and Stewartia
Erbar (1986) has published an intensive account of the floral development of Stewartia pseudocamellia. Though my SEM photographs of the same species do not show additional information to that of Erbar (1986) and though the early floral development of Hartia villosa is highly similar to that of S. pseudocamellia, I will describe early floral development of Hartia and Stewartia in detail to illustrate my different interpretation.

Perianth initiation
The five sepal primordia (Fig. 43) and the five petal primordia (Fig. 44) are situated in a whorl though initiated in a two-fifths spiral sequence, whereas the spiral of calyx is not continued by the corolla. Thus the petals are not in episepalous positions. This whorled arrangement may result from a comparatively rather rapid emergence of the five sepal/petal primordia relative to the elongation of the floral apex. Within the calyx whorl, the bases of the later four (the second to the fifth) young sepals are much compressed laterally (Fig. 45). Later, the five young sepals fuse together at their very base. The floral apex is fairly narrow at this time; the small floral apex together with the surrounding wall of the fused sepal bases cast a view of a tube base (Figs. 45, 51). In this connection, it may be pertinent to note that in Hartia and Stewartia the characteristically winged and boat-shaped leaf petioles enclose vegetative and floral buds tightly in the radial direction during their early development. Such pressure may contribute to the much narrowed "tube base" in this case. During the entire perianth ontogeny, the circumference of the young buds is rounded (Figs. 43, 45).

View larger version (176K): [in this window] [in a new window]   Figs. 43–54. Floral development of Hartia villosa and Stewartia pseudocamellia . Figs. 43–50 . Hartia villosa . 43. Five sepals (K) initiated spirally but arranged in one whorl. The first petal (P) appears in an inner whorl. Scale bar = 100 µm. 44. Five petal primordia (P) initiated spirally but arranged in one whorl. The androecial fascicle primordium (*) in front of petal 1 is just initiated. Scale bar = 50 µm. 45. Young floral bud showing its round circumference, five sepals (K) and five petals (P) in two whorls, developing androecial fascicles (*), and the invaginated floral apex. Scale bar = 200 µm. 46. Higher magnification of Fig. 45 . Note the initiation of stamen primordia (S) on androecial fascicles (AF) and the size difference among the androecial fascicles. Scale bar = 40 µm. 47. Floral apex showing the difference of size and number of stamens of the five androecial fascicles (AF). Carpel primordia (C) are already evident. Scale bar = 100 µm. 48. Higher magnification of Fig. 47 ; young carpels (C) are isolated, much elongated, and slightly concave at the ventral side. Scale bar = 40 µm. 49. Upper part of a young gynoecium showing five carpels well-fused laterally and the five-lobed stigma. Scale bar = 200 µm. 50. Longitudinal view of a dissected gynoecium. Note that carpels are poorly fused at ventral side and ovules (Ov) are in axile-basal positions. Scale bar = 200 µm. Figs. 51–54 . Stewartia pseudocamellia . 51. Floral apex after the innermost petal (P5) is just originated, showing five developing androecial fascicle primordia (*) and the earliest carpel primordium (C). Scale bar = 100 µm. 52. Androecial fascicles (AF) and carpel primordia (C) in opposite positions. Note stamens (S) are initiated on the the androecial fascicle in centrifugal direction. Scale bar = 100 µm. 53. Floral apex with five epipetalous androecial fascicles (AF) and young carpels (C). Carpels are fused laterally but not ventrally. Scale bar = 200 µm. 54. Longitudinal view of a dissected gynoecium. Carpels are poorly fused ventrally and ovules (Ov) assume axile-basal positons (Ov). Scale bar = 200 µm.

Gynoecium initiation
During the initiation of the lowermost/innermost one or two rows of stamens, carpel primordia appear one by one right opposite to the five androecial fascicles (Fig. 48); consequently, the five petals, five stamen groups, and five carpels are arranged in five radiate lines (Fig. 53). The carpel primordia are tangentially elongated, rather than hemispherical, bulges at inception (Fig. 48). In S. pseudocamellia, the inception of the five carpels is neither synchronous nor of a typical spiral. The two carpel primordia spatially associated with the first two androecial fascicles indeed develop first, but the other three carpels are usually of similar size, so their initiation sequence cannot be determined. In H. villosa, the carpels are so sunken that the carpel initiation sequence cannot be detected. In both taxa, the five carpels soon reach similar size, and grow faster in the longitudinal rather than other directions before starting to roll in at their individual bases (Fig. 53). Subsequently, each carpellary chamber becomes evident. Then the five carpels begin to fuse laterally at their bases to form a ring-like syncarpous gynoecium. The syncarpous gynoecium elongates fast and soon differentiates into an ovary, a short style, and a big five-lobed stigma (Fig. 49). The pollen-transmitting canal is five-armed (Fig. 49). The five carpels are very poorly fused along their ventral sides even at floral maturity. The five carpels are plicate. Within each ovarian locule, ovules are originated from the slope connecting the ventral sides and the locule base. Thus the placentation may be designated as axial-basal (Figs. 50, 54).

Group IIb: Franklinia and Schima
Perianth initiation of Schima superba var. kankoensis
The pattern of perianth initiation of S. superba var. kankoensis is basically the same as that of Hartia + Stewartia in that the five sepals and the five petals are arranged in two whorls (Figs. 55, 57) though initiated in a two-fifths spiral sequence, respectively, such that the later three sepals are compressed laterally and the five sepals soon fuse together at their very bases (Fig. 57). But S. superba var. kankoensis is apparently different in having its young floral apex broad and slightly convex after the petal initiation is completed (Figs. 55, 56). By that time, the isolaterally pentagonal floral apex has differentiated into two distinct parts: a central gynoecial bulge and five peripheral slopes, i.e., the androecial part (Fig. 56).

View larger version (179K): [in this window] [in a new window]   Figs. 55–66. Floral development of Schima superba var. kankoensis . 55. Young floral bud showing its round circumference, whorled arrangement of five sepals (K) and five petals (P), and the convex floral apex. Scale bar = 200 µm. 56. Floral apex more or less pentagonal, showing the central convex gynoecial region (G) and the peripheral androecial region (A). Scale bar = 100 µm. 57. Young floral bud showing whorled arrangement of sepals (K) and petals (P) and early initiation of stamens and gynoecium. Sepals become fused at the base (arrows). Scale bar = 200 µm. 58. Higher magnificaion of Fig. 57 . Stamen primordia (S) are initiated from five petal-opposed zones, which are separated by five grooves (arrows). Within each zone stamen primordia are initiated from two sides inwards. On the gynoecial primordium (G) one carpel primordium (*) appears in front of petal 1 (P). Scale bar = 100 µm. 59. Floral apex showing early initiation of androecium and gynoecium. The five androecial zones are separated by short grooves (arrows), with one to four stamen primordia (S) in each zone. Scale bar = 100 µm. 60. Floral apex with one whorl of stamen primordia and five epipetalous carpel primordia. Most grooves (arrows) between the androecial zones are still evident. Scale bar = 100 µm. 61. Floral apex with two whorls of stamen primordia and five young carpels. Androecial zonation can still be recognized by a few remaining grooves (arrows). Young carpels are highly fused and in an ascidiate form already. Scale bar = 100 µm. 62. Foral bud of similar stage as that in Fig. 61 , showing a groove (arrow) between two androecial zones. Scale bar = 50 µm. 63. Gynoecium with five highly fused carpels. The gynoecium is below its surrounding stamens. Scale bar = 100 µm. 64. Floral apex with stamen initiation completed. Note the elevation of the receptacle beneath the androecium. Scale bar = 200 µm. 65. Young floral bud showing the round circumference, the fusion of sepals at the base, and the concave floral apex. Scale bar = 500 µm. 66. Floral apex showing a syncarpous gynoecium and ~ 70 stamens. Note the stamens (*) of the outermost whorl are mostly horizontally oriented. Scale bar = 200 µm.

Gynoecium initiation of S. superba var. kankoensis
As stated before, the gynoecium and androecium begin to develop at about the same time (Figs. 56, 58). The common primordium of the gynoecium rises up slowly until nearly half of the stamens of the innermost whorl have originated, when the bulges of individual carpels emerge quincuncially, each opposing a stamen-initiation zone and a petal (Figs. 58, 59, 60). The individual carpel primordium was observable by its apical part only, indicating a high degree of congenital fusion between the five carpel primordia at their bases (Figs. 58, 59, 60). Before the stamens of the innermost whorl are all initiated the first carpel primordium has already assumed the ascidiate form (Fig. 60). The five young carpels then grow to a comparable size. Their ventral synascidiate region is prominent and just slightly lower than the tips of individual primordia (Figs. 61–63). After this stage, the syncarpous gynoecium gradually becomes urn shaped (Figs. 63–66). During subsequent development, a slender style and a capitate stigma are differentiated. The placentation is axile-central. The base of the ovarian cavities is slightly embedded within the receptacle, but the ovary position has not yet reached the condition of typical semi-inferior.

Franklinia alatamaha
Limited by the material available, the earliest floral development of F. alatamaha was not followed. Nevertheless, the floral developmental characteristics of Franklinia, as revealed from the available material (Figs. 67–69), are quite similar to those of S. superba var. kankoensis (Figs. 63, 64, 66). The youngest stage observed is represented by the co-existence of three-to-four stamen whorls and a gynoecium of five similar-sized and highly fused ascidiate young carpels (Fig. 67). In Fig. 67, a short groove on the androecial slope is visible, implying that this androecium is likely to be composed initially of five stamen-initiation zones as in Schima. In short, in terms of floral development, the overall middle stages of F. alatamaha shown by Figs. 67–69 are very similar to the comparable stages of S. superba var. kankoensis, particularly when one takes into consideration the very early carpellary fusion (compare Fig. 67 with Fig. 63) and the developmental manner of the receptacle (compare Figs. 68, 69 with Figs. 64, 66). Among other crucial similarities between Franklinia and S. superba (data not shown for Franklinia) are the whorled arrangement of sepals and petals, the petal-opposed carpels, and the axile-central placentation.

View larger version (79K): [in this window] [in a new window]   Figs. 67–69. Floral development of Franklinia alatamaha . 67. Young floral bud with highly fused carpels while stamen initiation has not yet been completed. Note a remaining groove (arrow) between two groups of stamens and the receptacle beneath the androecium much elevated. Scale bar = 200 µm. 68. Floral apex with around four whorls of stamens and a syncarpous gynoecium. Scale bar = 100 µm. 69. Older floral bud with stamen initiation completed. The gynoecium is now closed at the apex. Stamens of outer whorls are mostly in horizontal orientation. Scale bars = 200 µm.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Major floral developmental characters
Compared with the gross morphology of mature flowers, the early floral development of Camellioideae is far more diverse: genera may express differently in all or some of the four categories of floral organs, and many ontogenetic characters exhibt interesting intergeneric variations. The most important is that some two-state characters covary in presumably related genera within the subfamily. These major characters are as follows: (1) the degree of congenital carpellary fusion (CCF) – low vs. high; (2) the arrangement of perianth-members – spiral vs. whorled; (3) the timing of differentiation between sepals and petals – occurring at late floral development vs. in the very beginning of their inception; (4) the androecial zonation for stamen initiation – zonated vs. unzonated; (5) androecial fascicles – presence vs. absence; (6) placentation – axile-central vs. axile-basal; (7) ovary position – superior vs. intermediate between superior and semi-inferior.

The polarity of character 1 can be determined by consensus. The trend of syncarpy from apocarpy has long been supposed to be one major evolutionary trend in the angiosperms (Eames, 1961; Stebbins, 1974; Endress, 1994); though cases of secondary apocarpy have been reported (Endress, Jenny, and Fallen, 1983; Fallen, 1986; Jenny, 1988; Ramp, 1988) from the following few unrelated orders: Malvales (Sterculiaceae), Sapindales (Rutaceae, Simaroubaceae), and Gentianales (Apocynaceae, Asclepiadaceae). In highly advanced plant groups, the syncarpous ovary is normally inferior and developed from a highly congenitally fused common primordium (Sattler, 1973). It is believed that during the modification process from apocarpy to syncarpy, a gradual strengthening of CCF must take place. Accordingly, within Camellioideae, an on-average rather primitive group among those syncarpous ones, the CCF may be reasonably assumed to have evolved from a low to high level.

The whorled state of character 2 is generally considered to be a prerequisite for the evolution from aposepaly to synsepaly and from apopetaly to sympetaly (Endress, 1987). It is also more prevalent than the spiral state in those dicots of middle and higher evolutionary levels (Endress, 1987, 1994). Nevertheless, because reversal from whorled back to spiral pattern is possible (Endress, 1987), the polarity determination of character 2 is thus in need of a more intensive analysis.

As for character 3, I find no reliable criteria to judge the relative primitiveness between early and late differentiation of sepals and petals in the camellioid flowers.

Characters 4 and 5 are interrelated. Generally speaking, for a multistaminate androecium in the dicots, individual stamens may arise as primary primordia on the receptacle, as secondary primordia from a ring-like primary primordium, or as secondary ones within the fixed-numbered, separated primary primordia (Ronse Decraene and Smets, 1992; Endress, 1994, fig. 2.37). In the Camellioideae, the latter two patterns occur in Group I and Group II, respectively. In many unrelated nonmagnoliid dicot orders the co-occurrence of these two patterns of multistaminate androecia in a family is fairly common (Ronse Decraene and Smets, 1992, Table 1), and the transition between these two patterns is poorly understood. In addition, the debate on the origin of androecial fascicles in angiosperms remains unsettled (Leins and Erbar, 1991; Ronse Decraene and Smets, 1992); therefore, for Camellioideae, the polarity of characters 4 and 5 cannot be determined at this moment.

Erbar (1988) claimed that "Several similarities in flower development between Stewartia and the primitive cactus Pereskia indicate a relationship between Dilleniidae and Caryophyllidae." She then considered the complex androecia (i.e., androecial fascicles) to be archaic within the Dilleniidae (Erbar, 1986, 1988). This viewpoint is, in my opinion, questionable. Androecial fascicles occur in numerous phylogenetically unrelated families of Dilleniidae, Rosidae, and Centrospermae. A comparison between the androecial fascicle development of Hartia + Stewartia (my work) and that of the Metrosideros group of Myrtaceae (Orlovich et al., 1996) reveals a much greater similarity than that between Stewartia and Pereskia. In Hartia + Stewartia and the Metrosideros group the androecial fascicles are initiated from the petal-opposed regions of a deeply concave floral apex, whereas in Pereskia the fascicles are alternate with petals and emerge from a convex floral apex (Leins and Schwitalla, 1986). Furthermore, the myrtaceous androecial fascicles are within-family secondary structures (Johnson and Briggs, 1984) and might not be homologous among different taxa (Drinnan and Ladiges, 1991). Therefore, because the very strong resemblance between Hartia + Stewartia and the Metrosideros group in the developmental features of androecial fascicles is undoubtedly superficial, Erbar's (1988) evolutionary conclusion based on connecting Stewartia with Pereskia is simply not convincing. Since the androecial structure may be highly variable within a family, whether the androecial fascicle is a primitive or derived structure within the Camellioideae has to be ascertained through an integrated analysis of the subfamily.

As to the placentation (character 6), Hartia and Stewartia have axile, basally positioned ovules in each locule, whereas the other eight camellioid genera have a typical axile-central placentation. These two types of placentation can be referred to the type E (axile-central) and type K (axile-basal), respectively, of Stebbins' scheme (1974: fig. 12–3) on the evolution of placentation. Stebbins considered the axile-central type to be ancestral to the axile-basal type; however, whether his hypothesis is applicable in the Camellioideae needs verification. Keng (1962) described the placentas of Hartia and Stewartia as parietal and proposed that the camellioid placentation has probably evolved from parietal to axile (e.g., from Stewartia through Hartia to Gordonia s.l., Tutcheria, and Camellia). I do not concur with Keng's interpretation on the placentas of Hartia and Stewartia as parietal. There is no parietal placentation in the Camellioideae.

The polarity of character 7 can be determined on the basis that the evolution from superior to inferior ovary is an overwhelming trend in angiosperms (Cronquist, 1988; Endress, 1994), although a case of reversal has been reported from the Araliaceae (Eyde and Tseng, 1969). In the Camellioideae, the position of the very young ovaries of Schima and Franklinia (Group IIb) is intermediate between superior and semi-inferior, yet these become superficially similar to an inferior ovary at their maturity in that petals and filaments are fused basally, with the fused part appressing and enclosing the ovary tightly. Such an ontogenetic modification apparently foreshadows the inferior ovary position. Therefore, the polarity of the character 7 can be determined without hesitation.

Cladistic analysis of ten characters
Among the above seven floral-developmental characters, only two (1 and 7) can be definitely determined for their polarity. A more intensive polarity analysis is consequently required for a sufficient understanding of the floral evolution in the Camellioideae. For this purpose, three additional characters from other aspects were lumped together with the foregoing seven (in a continuing sequence, see Table 3) for a cladistic analysis employing PAUP 3.1.1 (Swofford, 1993). Characters 8 and 9 are embryological (Tsou, 1996, 1997), and character 10 is cytological. In each of the seven genera of the three (sub-)groups, the baseline data accumulation of all of these ten characters is satisfactory. Particularly significant is that within each (sub-)group all, except character 10 (in Subgroup IIa), of these ten characters are expressed consistently. [Character 10 in Subgroup IIa, N = 18 in Hartia, whereas N = 15 in Stewartia, apparently represents a transition between Group I (N = 15) and Subgroup IIb (N = 18).] To my knowledge, no other salient characters of the Camellioideae have been studied for all these seven genera and have the character state uniformly expressed in each (sub-)group.

Neither Ternstroemioideae nor any other extant taxon was chosen as the outgroup for the current analysis because Ternstroemioideae as well as those thealean and ebenalean families is rather dissimilar to the Camellioideae in the overall expression of these ten characters. Instead, an assumed ancestor was taken as the outgroup for convenience. It was simply defined as a taxon expressing the primitive states of the four characters (characters 1, 7, 8, and 9) whose polarity has been assessed (Tables 3, 4).

View larger version (9K): [in this window] [in a new window]   Fig. 70. The single most-parsimonious tree based on data matrix of Table 4 . This is the same shortest tree (12 steps, CI = 1.00, RI = 1.00) generated by PAUP using either exhausitive, branch-and-bound, or heuristic search options. The first taxon (Assumed ancestor) was used as the outgroup. Each black bar indicates one change of state. Both characters 8 and 9 have three states and two changes.

In addition, Keng's (1962) suggestion that the axile-central placentation is derived within Camellioideae is also rejected. I consider axile-central placentation to be ancestral in this subfamily. In the lineage of Hartia-Stewartia the central axis of the ovary has been reduced to the ovarian base, and the axile-basal placentation was consequently derived.

Tribal taxonomy and phylogeny of Camellioideae
Though produced chiefly for the understanding of the floral evolution in Camellioideae, the cladogram (Fig. 70) simultaneously, in my opinion, presents the major taxonomic differentiation within the subfamily. Thus, the (sub-)groups I, IIa, and IIb — originally distinguished by their distinct perianth phyllotaxy, timing of the differentiation between sepals and petals, and patterns of androecial and gynoecial development, and further supported by the chromosome numbers and embryological data — can be directly converted to formal taxonomic categories as follows:

Tribe Camellieae (= Group I)—Camellia s.l., Polyspora, Pyrenaria s.l., Sinopyrenaria (Laplacea and Parapyrenaria with generic status uncertain, may be merely an element of Polyspora and Pyrenaria, respectively.)

Tribe Gordonieae (= Group II)—

Subtribe Stewartiinae (= Subgroup IIa): Hartia, Stewartia

Subtribe Gordoniinae (= Subgroup IIb): Apterosperma, Franklinia, Gordonia s.s., Schima

The classification proposed here matches well with Airy-Shaw's (1936) and Melchior's (1964) tribal/subtribal treatments, but differs from the others to a greater extent (Melchior, 1925; Sealy, 1958; Keng, 1962; Deng and Baas, 1991). This is mainly because the floral character employed by Airy-Shaw and Melchior, whether sepals and petals are sharply dissimilar, is very significant because this character is associated with two fundamental transitions in the early floral development of Camellioideae, i.e., the arrangement of perianth members and the timing of the differentiation between sepals and petals. On the other hand, most of the other treatments usually overlooked floral characters and put too much weight on fruit characters, which lead to mistakes like lumping Polyspora (including Laplacea) with Gordonia s.s. and Schima because they had capsules and winged seeds in common. In fact, such a grouping is not reliable because the wing of Polyspora seeds is developed from the mesophyll of the raphe whereas that of Schima seeds is derived from the epidermis of outer integument on the antiraphe side (Tsou, 1997). On the other hand, Gordonia s.s. and Schima having an intimate relationship with Franklinia, a genus without winged seeds, is strongly supported by cytology (Table 3), embryology (Table 3 and unpublished data), floral development (the present study), and molecular data (Prince and Parks, 1997). Apterosperma, another genus without winged seeds, seems phylogenetically close to these three genera based on the limited information on floral features (the present study) and pseudopollen (Tsou, 1996). In the Camellioideae the evolution of fruit and seed seems less conservative and much more diversified than the evolution of early floral development. Implications of fruit and seed characters in phylogenetic considerations among the camellioid genera need to be confirmed with developmental data.

The phylogenetic relationships among the camellioid tribes/subtribes have been proposed only by Keng (1962). According to his Fig. 30, he considered Camellieae (including Camellia s.l. and Pyrenaria s.l.) to be the latest evolved and Gordonieae (including Franklinia, Gordonia s.s., Polyspora, and Schima) the earliest branch. My results in Fig. 70 suggest that, on average, Camellieae (Camellia s.l., Polyspora, and Pyrenaria) appears to be the most primitive branch and subtribe Gordoniinae (Apterosperma, Franklinia, Gordonia s.s., and Schima) of the Gordonieae the most advanced. Among the latter four genera, Schima is proposed as the most advanced genus within the Camellioideae. Such a phylogenetic hypothesis is congruent with my earlier conclusion on Camellia, Franklinia, and Schima based on eight embryological characters (Tsou, 1997).

	FOOTNOTES

 
¹ The author thanks Scott Mori (Franklinia alatamaha and Stewartia pseudocamellia ), Linda Prince (Gordonia lasianthus ), and Hong Wong (Sinopyrenaria yunnanensis ) for providing materials; Hua-Gu Yeh, Yu-Hsing Fon, and Hong Wang from mainland China, and Yu-Pin Cheng, Rey-Fen Lin, Mong-Whai Su, and Kuo-Hsing Wang from Taiwan for help with field work; Rey-Fen Lin and Choryn-Man Tseng for technical support; D. Tyler Rainsbury for correcting English; and Dr. Chung-Fu Shen for his critical comments on this manuscript. This study was supported by the Institute of Botany, Academia Sinica, as well as grants NSC 82–0211-B-001–054 and NSC 86–2311-B-001–011 from the National Science Council, the Republic of China.

² FAX: +886–2-782–7954 (e-mail:botsou{at}ccvax.sinica.edu.tw ).

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Ackerman, W. L.1971Genetic and cytological studies with Camellia and related genera. Technical Bulletin Number 1427, Agricultural Research Service, USDA, U.S. Government Printing office, Washington, DC.

Airy-shaw, H. K.1936Notes on the genus Schima and on the classification of the Theaceae-Camellioideae. Kew Bulletin 1936: 496–499.

Chang, H.-T.1976Apterosperma––Genus Novum Theacearum. Journal of Sun Yatsen University 2: 90–92 (in Chinese).

Cronquist, A.1988The evolution and classification of flowering plants. New York Botanical Garden, Bronx, NY.

Deng, L., and P. Baas.1991The wood anatomy of the Theaceae. IAWA Bulletin (New Series) 12: 333–353.

Drinnan, A. N., and P. Y. L. Ladiges.1991Floral development in the "Symphyomyrtus group" of eucalyptus (Eucalyptus: Myrtaceae). Australian Systematic Botany 4: 553–562.

Eames, A. J.1961Morphology of the angiosperms. McGraw-Hill, New York, NY.

Endress, P. K.1987Floral phyllotaxis and floral evolution. Botanishe Jahrbucher fur Systematik 108: 417–438.

———.1994Diversity and evolutionary biology of tropical flowers. Cambridge University Press, Cambridge.

———, M. Jenny, and M. E. Fallen.1983Convergent elaboration of apocarpous gynoecia in higher advanced dicotyledons (Sapindales, Malvales, Gentianales). Nordic Journal of Botany 3: 293–300.

Erbar, C.1986Untersuchungen zur Entwicklung der spiraligen Blute von Stewartia pseudocamellia (Theaceae). Botanishe Jahrbucher fur Systematik 106: 391–407.

———.1988Early developmental patterns in flowers and their value for systematics. In P. Leins, S. C. Tucker, and P. K. Endress [eds.], Aspects of floral development, 7–24. J. Cramer, Berlin.

Eyde, R. H., and C. C. Tseng.1969Flower of Tetraplasandra gymnocarpa hypogyny with epigynous ancestry. Science (Washington, D.C.) 166: 506–508.

Fallen, M. E.1986Floral structure in the Apocynaceae: morphological, functional, and evolutionary aspects. Botanishe Jahrbucher fur Systematik 106: 245–286.

Gill, B. S., S. S. Bir, M. S. Sidhu, and V. K. Singhal.1984Chromosome number reports LXXXIV. Taxon 33: 536–539.

Hu, H.-H.1956Sinopyrenaria and Yunnanea, two new genera of Theaceae from Yunnan, China. Acta Phytotaxonomica 5: 279–284.

Huang, S.-F., Z.-F. Zhao, Z. Y. Chen, S.-Z. Chen, and X.-X. Huang.1985Preliminary report on chromosome numbers of plants. Asian Forestry of Science and Technology 1: 1–15.

Jenny, M.1988Different gynoecium types in Sterculiaceae: ontogeny and functional aspects. In P. Leins, S. C. Tucker, and P. K. Endress [eds.], Aspects of floral development, 225–236. J. Cramer, Berlin.

Johnson, L. A. S., and B. G. Briggs.1984Myrtales and Myrtaceae –– a phylogenetic analysis. Annals of the Missouri Botanical Garden 71: 700–756. [CrossRef][ISI]

Keng, H.1962Comparative morphological studies in Theaceae. University of California Publications in Botany 33: 269–384.

Kobuski, C. E.1951Studies in the Theaceae, XXI The species of Theaceae indigenous to the United States. Journal of the Arnold Arboretum 32: 123–138.

Kondo, K.1977Chromosome numbers in the Genus Camellia. Biotropica 9(2): 86–94.

Leins, P., and C. Erbar.1991Fascicled androecia in Dilleniidae and some remarks on the Garcinia androecium. Botanica Acta 104: 336–344. [ISI]

———, and S. Schwitalla.1986Studien an Cactaceen - Bluten I. Einige Bemerkungen zur Blutenentwicklung von Pereskia. Beitrage zur Biologie der Pflanzen 60: 313–323.

———, S. C. Tucker, and P. K. Endress.1988Aspects of floral development. J. Cramer, Berlin.

Li, J.1996A systematic study on the genera Stewartia and Hartia (Theaceae). Acta Phytotaxonomica Sinica 34: 48–67. (in Chinese, with English summary)

Mehra, P. N.1976Cytology of Himalayan hardwood. Sree Saraswaaty Press, Calcutta.

Melchior, H.1925Theaceae. In A. Engler and E. Prantl [eds.], Die naturlichen Pflanzenfamilien, 2d ed., vol. 21, 109–154. Leipzig.

———.1964Theaceae. In A. Engler, Syllabus der Pflanzenfamilien, Band II, 166–168. Gebruer Borntraeger, Berlin.

Oginuma, K., Z. Gu, L. Xia, and K. Kondo.1994Karyomorphology of some Theaceae from China and Singapore. La Kromosomo 2–73: 2498–2503.

Ono, M.1977Cytotaxonomical studies on the flowering plants endemic to the Bonin Islands. Memoir of the National Science Museum, Tokyo 10: 63–80.

Orlovich, D. A., A. N. Drinna, and P.Y. Ladiges.1996Floral development in the Metrosideros group (Myrtaceae) with special emphasis on the androecium. Telopea 6: 689–719.

Prince, L. M. and C. Parks.1997Phylogenetic relationships of the Theoideae (Theaceae) based on two chloroplast DNA sequence data sets. American Journal of Botany 84 (Supplement) : 225 (Abstract).

Ramp, E.1988Struktur, Funktion und systematische Bedeutung des Gynoeciums bei Rutaceae und Simaroubaceae. Ph.D. dissertation, University of Zurich, Zurich.

Ronse Decraene, L. P., and E. F. Smets.1992Complex polyandry in the Magnoliatae: definition, distribution and systematic value. Nordic Journal of Botany 12: 621–649. [ISI]

Santamour, F. S.1963Chromosome number in Theaceae. Morris Arboretum Bulletin 14: 51–53.

Sattler, R.1973Organogenesis of flowers: a photographic text-atlas. University of Toronto Press, Toronto.

Sealy, J. R.1958A revision of the genus Camellia. Royal Horticultural Society, London.

Stebbins, G. L.1974Flowering plants: evolution above the species level. Belknap Press of Harvard University Press, Cambridge, MA.

Swofford, D. L.1993PAUP: phylogenetic analysis using parsimony, version 3.1.1. Illinois Natural History Survey, Champaign, IL.

Sugiyama, M.1991Scanning electron microscopy observation on early ontogeny of the flower of Camellia japonica L. Journal of Japanese Botany 66: 295–299.

Tsou, C.-H.1995Embryology of Theaceae — anther and ovule development of Adinandra, Cleyera, and Eurya. Journal of Plant Research 108: 77–86. [CrossRef][ISI]

———.1996The pseudopollen of the Camellioideae of Theaceae: its development and systematic significance. American Journal of Botany 83 (Supplement): 198–199 (Abstract).

———.1997Embryology of the Theaceae — anther and ovule development of Camellia,Franklinia, and Schima. American Journal of Botany 84: 369–381. [Abstract]

Tucker, S. C.1996Floral evolution and development: testing hypotheses. American Journal of Botany 83 (Supplement):

Wood, C. E.1959The genera of Theaceae of the southeastern United States. Journal of the Arnold Arboretum 40: 413–419.

Zhang W.-J., and T.-L. Ming.1995Karyotypical study of Sect. Archecamellia of Genus Camellia. Acta Botanica Yunnanica 17: 48–54. (in Chinese, with English summary)

This article has been cited by other articles:

				L. P. Ronse De Craene Are Petals Sterile Stamens or Bracts? The Origin and Evolution of Petals in the Core Eudicots Ann. Bot., September 1, 2007; 100(3): 621 - 630. [Abstract] [Full Text] [PDF]

				L. P. RONSE DE CRAENE Floral Development of Berberidopsis corallina: a Crucial Link in the Evolution of Flowers in the Core Eudicots Ann. Bot., November 1, 2004; 94(5): 741 - 751. [Abstract] [Full Text] [PDF]

				A. A. Anderberg, C. Rydin, and M. Kallersjo Phylogenetic relationships in the order Ericales s.l.: analyses of molecular data from five genes from the plastid and mitochondrial genomes Am. J. Botany, April 1, 2002; 89(4): 677 - 687. [Abstract] [Full Text] [PDF]

				L. M. Prince and C. R. Parks Phylogenetic relationships of Theaceae inferred from chloroplast DNA sequence data Am. J. Botany, December 1, 2001; 88(12): 2309 - 2320. [Abstract] [Full Text] [PDF]